A Sponsored Supplement to Science SPReading The importance ... · SPReading the word: The importance of binding kinetics Originally published 5 June 2015 in SCIENCE sciencemag.org

A Sponsored Supplement to Science

Sponsored by Produced by the

Science/AAAS Custom Publishing Office

SPReading the word: The importance of binding kinetics

http://sciencemag.org


http://gehealthcare.com

gelifesciences.com/biacoreGE, GE monogram, and Biacore are trademarks of General Electric Company.© 2017 General Electric Company.GE Healthcare Bio-Sciences AB, Björkgatan 30, 751 84 Uppsala, SwedenFor local office contact information, visit gelifesciences.com/contact.

29273499 AA 06/2017

Quality Biacore™ consumablesfor excellent assaysNo matter what you want to get out of your interaction analysis, GE Healthcare has developed a range of tools designed specifically to make Biacore assays as easy and reliable as possible. The complete toolbox is backed up by stringent production methods and quality control.

• A sensor surface for every need

• Kits to save you time and effort

• Buffers and solutions for convenience

http://gelifesciences.com/biacore


Introductions

2 Using SPR to transform precision medicine Jackie Oberst, Ph.D. Sean Sanders, Ph.D. Science/AAAS

3 Pushing the boundaries of discovery with SPR Christina Burtsoff-Asp Product Marketing Manager, Life Sciences GE Healthcare

Research articles

4 Targeting DnaN for tuberculosis therapy using novel griselimycins Angela Kling, Peer Lukat, Deepak V. Almeida et al.

11 Therapeutic bispecific antibodies cross the blood- brain barrier in nonhuman primates Y. Joy Yu, Jasvinder K. Atwal, Yin Zhang et al.

20 Heparin is an activating ligand of the orphan receptor tyrosine kinase ALK Phillip B. Murray, Irit Lax, Andrey Reshetnyak et al.

White paper

27 Deeper insights into biological realities GE Healthcare

Technical note

32 Achieving data-driven decisions with real-time interaction analyses GE Healthcare

gelifesciences.com/biacoreGE, GE monogram, and Biacore are trademarks of General Electric Company.© 2017 General Electric Company.GE Healthcare Bio-Sciences AB, Björkgatan 30, 751 84 Uppsala, SwedenFor local office contact information, visit gelifesciences.com/contact.

29273499 AA 06/2017

Quality Biacore™ consumablesfor excellent assaysNo matter what you want to get out of your interaction analysis, GE Healthcare has developed a range of tools designed specifically to make Biacore assays as easy and reliable as possible. The complete toolbox is backed up by stringent production methods and quality control.

• A sensor surface for every need

• Kits to save you time and effort

• Buffers and solutions for convenience

1


Table of contents

SCIENCE sciencemag.org

ROGER GONCALVES, ASSOCIATE SALES DIRECTORCustom Publishing Europe, Middle East, and [email protected]+41-43-243-1358

© 2017 by The American Association for the Advancement of Science. All rights reserved. 28 July 2017

Editors: Jackie Oberst, Ph.D.; Sean Sanders, Ph.D.Proofreader/Copyeditor: Bob FrenchDesigner: Amy Hardcastle

About the cover: Surface plasmon resonance (SPR) has become a go-to-tool for antibody characterization as well as for other applications. Today, researchers are increasingly focused on the development of novel antibody treatments, including antibody–drug conjugates (shown) for cancer.

Cover image provided courtesy of GE Healthcare.

This booklet was produced by the Science/AAAS Custom Publishing Office and sponsored by GE Healthcare.


mailto:rgoncalves%40science-int.co.uk%0D?subject=

2

Using SPR to transform precision medicine

F ans of the sci-fi series Star Trek are familiar with the show’s fictional “tricorders,” handheld devices used for sensor scanning and data analysis and recording. Yet these futuristic instruments are now, in a sense, entering the

real world due to the advent of biosensors, devices that transform the detection of biological elements—antibodies, nucleic acids, cell receptors, enzymes, among others—into signals (e.g., optical, electrochemical) that can be more easily measured and quantified.

The technology behind this modern-day measuring equipment is surface plasmon resonance (SPR), a process through which electrons on a metal surface are excited by a polarized light source, creating charge-density waves called plasmons. These plasmons then bend the light at a specific angle (known as the resonance angle) that is picked up through a detector. As molecules bind and dissociate from the metal surface, the resonance angle changes, giving an interaction profile (including binding kinetics, specificity, concentration, and affinity) that is recorded in real time. Chemicals such as antigens can be bound to a surface and exposed to blood or tissue culture media to see which antibodies bind best to them. Similarly, DNA or enzymes can be tethered to see which proteins affix to them. The possibilities for this technology include proteomics, immunogenicity, and drug discovery, but really are limitless.

However, the best use for SPR is precision medicine. This tool could help to more quickly identify molecules or compounds that could aid in treatment of diseases. Conversely, it could also help pinpoint which patients would respond best to a particular drug or vaccine.

Included in this booklet are articles from Science, Science Translational Medicine, and Science Signaling that detail the prospects for SPR technology. The first article determines that a candidate tuberculosis drug, a peptide from Streptomyces called griselimycin, functions by binding to a portion of the culprit bacteria’s chromosomes called DnaN. The second describes how drugs that need to cross the blood-brain barrier (BBB) can do so more effectively by using bispecific antibodies that attach to the BBB’s transferrin receptors; in this case, an enzyme combined with these antibodies was delivered to the brain to reduce brain amyloid-ß in mice and possibly reverse neurological diseases such as Alzheimer’s. The third article identifies heparin as a potential ligand for the orphan receptor anaplastic lymphoma kinase (ALK), which has been implicated in such cancers as lung adenocarcinoma and neuroblastoma. This mechanism could provide an approach for developing better ALK-targeted cancer therapies.

To reference Star Trek once again, its opening monologue spelled out a starship's mission—but biosensors such as SPR can indeed take researchers on experimental “voyages” that will enable them “to boldly go” where no researcher has gone before. Although perhaps not yet at the “final frontier” of precision medicine, researchers and technology developers are certainly at the cutting edge.

Jackie Oberst, Ph.D.Sean Sanders, Ph.D.Custom Publishing OfficeScience/AAAS


sciencemag.org SCIENCE

Jackie Oberst

Sean Sanders


3

T oday, the healthcare industry is increasingly turning to precision medicine to improve patient outcomes and address the rising global incidence of lifestyle and chronic diseases while keeping costs as low as possible.

GE Healthcare’s Life Sciences business is on a mission to accelerate precision medicine by helping researchers, pharmaceutical compa-nies, and clinicians to discover, make, and use new medicines and therapies. We provide expertise and tools for a wide range of ap-plications, including basic research into cells, proteins, and drug dis-covery. We also supply technologies that support large-scale manu-facturing of vaccines and biopharmaceuticals—the fastest-growing class of medicines—to battle some of the toughest diseases facing the world, from diabetes to cancer to autoimmune diseases.

Understanding the nature of interactions between molecules is fundamentally important in the life sciences. In the 1980s, pioneering research in surface plasmon resonance (SPR) led GE to launch the world’s first SPR-based analytical instrument, called Biacore. Since then, Biacore SPR technology has been used in thousands of peer-reviewed scientific publications, allowing real-time, label-free detection and characterization of biomolecular interactions.

GE’s Biacore SPR systems provide detailed information about biomolecular interactions, without the need for labeling. From microliters of sample, SPR provides information on affinity, kinetics, specificity, comparability, and active concentration for molecular interactions. Binding partners can range from low molecular weight fragments to cells. With automated runs and tailored evaluation tools, Biacore SPR systems help obtain trustworthy results quickly, enabling researchers to rapidly understand the nature and strength of the interaction they are studying.

Healthcare needs more precise diagnoses and treatments to improve patient outcomes, fight disease, and help reduce the $350 billion wasted each year on poorly targeted medicines. The speed and reliability of information gained from SPR research is shaping precision medicine and allowing for the development of better patient stratification, improved treatments, and better patient outcomes.

Christina Burtsoff-AspProduct Marketing Manager, Life Sciences GE Healthcare

Pushing the boundaries of discovery with SPR

Introductions



4


sciencemag.org SCIENCEOriginally published 5 June 2015 in SCIENCE

RESEARCH ARTICLE◥

ANTIBIOTICS

Targeting DnaN for tuberculosistherapy using novel griselimycinsAngela Kling,1,2* Peer Lukat,1,2,3* Deepak V. Almeida,4,5 Armin Bauer,6 Evelyne Fontaine,7

Sylvie Sordello,7 Nestor Zaburannyi,1,2 Jennifer Herrmann,1,2 Silke C. Wenzel,1,2

Claudia König,6 Nicole C. Ammerman,4,5 María Belén Barrio,7 Kai Borchers,6

Florence Bordon-Pallier,8 Mark Brönstrup,3,6 Gilles Courtemanche,7 Martin Gerlitz,6

Michel Geslin,7 Peter Hammann,9 Dirk W. Heinz,2,3 Holger Hoffmann,6 Sylvie Klieber,10

Markus Kohlmann,6 Michael Kurz,6 Christine Lair,7 Hans Matter,6 Eric Nuermberger,4

Sandeep Tyagi,4 Laurent Fraisse,7 Jacques H. Grosset,4,5 Sophie Lagrange,7 Rolf Müller1,2†

The discovery of Streptomyces-produced streptomycin founded the age of tuberculosistherapy. Despite the subsequent development of a curative regimen for this disease,tuberculosis remains a worldwide problem, and the emergence of multidrug-resistantMycobacterium tuberculosis has prioritized the need for new drugs. Here we show that newoptimized derivatives from Streptomyces-derived griselimycin are highly active againstM. tuberculosis, both in vitro and in vivo, by inhibiting the DNA polymerase sliding clampDnaN. We discovered that resistance to griselimycins, occurring at very low frequency, isassociated with amplification of a chromosomal segment containing dnaN, as well as theori site. Our results demonstrate that griselimycins have high translational potential fortuberculosis treatment, validate DnaN as an antimicrobial target, and capture the processof antibiotic pressure-induced gene amplification.

The discovery of streptomycin, a natural anti-biotic produced by Streptomyces griseus,marked the beginning of two formativedisciplines within the field of infectiousdiseases—namely, the study of bacterial-

derived (rather than fungal- or plant-derived) me-dicinal compounds and the drug treatment oftuberculosis (TB) (1). This achievement initiateddecades of research in the discovery and use ofanti-TB drugs, ultimately leading to the develop-ment of the 6-month, multidrug regimen currentlyused for the cure of drug-susceptible TB (2). Un-fortunately, failures in the implementation of thiscurative regimen, which are partly due to the chal-lenges of its complex and lengthy nature, haveled to the development and transmission of drug-

resistant strains ofMycobacterium tuberculosis.Today, TB remains an enormous global healthburden, causing an estimated 1.3 million deathsand 8.7 million new cases in 2012, and a growingpercentage of TB (more than 30% of new cases insome countries) is multidrug-resistant (3). Thus,new drugs addressing novel M. tuberculosis tar-gets are needed to provide different therapy op-tions for patients with drug-resistant TB and alsoto both shorten and simplify treatment of drug-sensitive TB. Ideally, these new drugs shouldbe combined in regimens tackling both drug-sensitive and drug-resistant TB, representing aparadigm shift toward more universally usefulTB treatment regimens.Bacterial-derived natural products remain a

rich source for antibacterial lead compounds. Infact, ~80% of the currently used antibiotics areeither directly derived from bacterial metabolicpathways or represent structural derivatives ofmetaboliteswith improved pharmaceutical prop-erties (4). However, due to the reduced interest indevelopment of antibacterial drugs in the lastdecades of the 20th century, quite a number ofpromising natural product leads were not ad-vanced to clinical development. Recently, naturalproduct and antibiotic research has been revi-talized, not only because of the urgent need toidentify novel antibiotics but also owing to ad-vanced technologies becoming available. Thus,researchers are now enabled to overcome hurdlesin natural product research, such as target iden-tification by deciphering the self-resistance mech-anisms in producer strains throughwhole-genomesequencing and compound optimization by ge-

netic engineering. Successful recent applicationsof these capabilities include the derivation ofsemisynthetic spectinamides found to be highlyactive against both drug-resistant and -susceptibleM. tuberculosis strains (5) and the identifica-tion of InhA as the mycobacterial target of theDactylosporangium fulvum–produced pyrido-mycin (6).In a search for neglected antibiotics with high

anti-TB potential, Sanofi reinvestigated griseli-mycin (GM) (Fig. 1), a cyclic peptide that was iso-lated from two strains of Streptomyces identifiedin the 1960s (7). GM was found to have anti-bacterial activity specifically against organismswithin the Corynebacterineae suborder, notablyincludingMycobacterium species, which promptedthe company Rhône-Poulenc to pursue develop-ment of GM as an anti-TB drug. The first hu-man studies were promising but revealed poorpharmacokinetics of GM, in particular short plas-ma elimination half-life after oral administration(8, 9). Following elucidation of the compound’sstructure (10, 11), a derivatization program wasinitiated to find GM analogs with improved phar-macokinetic properties (12, 13); however, this pro-gram was terminated in the 1970s after rifampin(RIF) became available for TB treatment. Becauseof earlier reports of the effectiveness ofGMagainstdrug-resistantM. tuberculosis (14, 15), we recentlyreinitiated studies on this natural product leadwith the ultimate goal of introducing a highly ac-tive, stable, and safe derivative of this compoundclass into the TB drug development pipeline.

Development of GM analogs

Our primary optimization goals for GM were toincrease its potency, metabolic stability, and ex-posure. Metabolic stability profiling of natural,less abundant analogs of GM identified Pro8 as amain site ofmetabolic degradation, supported bythe finding that the methyl derivative [methyl-griselimycin (MGM)] (Fig. 1) wasmarkedly morestable than GM itself after incubation with hu-man livermicrosomes (Table 1). Because only verysmall amounts of MGM are produced naturally,a total synthesis route was elaborated to provideaccess to MGM and related analogs (see supple-mentary text). Structure-activity relationships ofnew synthetic GM analogs resulting from thisapproach confirmed that incorporation of sub-stituents at Pro8 led to metabolically highly stablecompounds and also indicated that increasinglipophilicity considerably increased in vivo ex-posure in plasma and lungs of mice, as well asin vitro activity againstM. tuberculosis (see sup-plementary text). From these efforts, cyclohexyl-griselimycin (CGM) (Fig. 1) was identified. Theminimum inhibitory concentration (MIC) valuesof CGM were 0.06 and 0.2 mg/ml for the drug-susceptibleM. tuberculosis strainH37Rv in brothculture and within macrophage-like (RAW264.7)cells, respectively (Table 1). CGM exhibited time-dependent bactericidal activity in vitro (Fig. 2A).Although the unbound fraction of CGM in plasmawas low (0.3 and 0.4% in human and mouseplasma, respectively), the MIC shift of CGM inthe presence of human or mouse sera was only

RESEARCH

1106 5 JUNE 2015 • VOL 348 ISSUE 6239 sciencemag.org SCIENCE

1Department of Microbial Natural Products, HelmholtzInstitute for Pharmaceutical Research Saarland (HIPS),Helmholtz Centre for Infection Research and PharmaceuticalBiotechnology, Saarland University, 66123 Saarbrücken,Germany. 2German Centre for Infection Research (DZIF),Partner Site Hannover-Braunschweig, Hannover, Germany.3Helmholtz Centre for Infection Research (HZI), 38124Braunschweig, Germany. 4Center for Tuberculosis Research,Johns Hopkins University School of Medicine, Baltimore, MD21231, USA. 5KwaZulu-Natal Research Institute forTuberculosis and HIV (K-RITH), Durban 4001, South Africa.6Sanofi-Aventis R&D, LGCR/Chemistry, Industriepark Höchst,65926 Frankfurt am Main, Germany. 7Sanofi-Aventis R&D,Infectious Diseases Therapeutic Strategic Unit, 31036Toulouse, France. 8Sanofi-Aventis R&D, Strategy, SciencePolicy & External Innovation (S&I), 75008 Paris, France.9Sanofi-Aventis R&D, Infectious Diseases TherapeuticStrategic Unit, 65926 Frankfurt, Germany. 10Sanofi-AventisR&D, Disposition Safety and Animal Research, 34184Montpellier, France.*These authors contributed equally to this work.†Corresponding author. E-mail: [email protected]

on

May

11,

201

7ht

tp://

scie

nce.

scie

ncem

ag.o

rg/

Dow

nloa

ded

from


mailto:rolf.mueller%40helmholtz-hzi.de?subject=

5

Research articles

SCIENCE sciencemag.orgOriginally published 5 June 2015 in SCIENCE

around five- or sevenfold, respectively (Table 1).The MIC values of CGM were similar for a rangeofM. tuberculosis strains from different lineages(representing geographical and evolutionary di-versity), as well as for strains monoresistant tofirst- or second-line anti-TB drugs (table S1),demonstrating a lack of cross-resistance. GM,MGM, and CGM were not active against M.tuberculosis under hypoxic conditions in whichthe bacteriawerenot activelymultiplying (Table 1).CGM exhibited optimized adsorption, distri-

bution, metabolism, and excretion properties;that is, high oral bioavailability (89%), moder-ate total plasma clearance (1.1 liter/hour perkilogram), and a large volume of distribution(5.5 liter/kg). Exposure and half-life inmice (sup-portive of once-daily dosing) were higher thanfor GM and MGM. In addition, over the 30- to100-mg/kg dose range, CGM exposure in plasmaand lung increased roughly with dose propor-tionality and continued to increase over the 100-to 600-mg/kg dose range (table S2). Moreover,

CGM exhibited limited potential for drug-druginteractions, as neither CYP induction nor inhi-bition was observed, and the contribution ofCYP to degradation was balanced, indicatingthat induction of CYPs by other drugs should notaffect CGMexposures (Table 1). CGM (at concen-trations up to 5000 mg/ml) did not increase thenumber of revertant colonies in Ames II test-ing with TA98 and mixed strains of Salmonellatyphimurium, indicating a lack of mutagenicity.Additionally, no chromosomal aberrations wereobserved in CGM-exposed mammalian cells, astested with L5178Y cells (at CGM concentrationsup to 1000 mg/ml, with or without metabolic ac-tivation) and with CHO-K1 cells (fig. S1), indicat-ing a lack of genotoxicity.

CGM activity in mouse models of TB

To assess the in vivo activity of CGM, we con-ducted dose-ranging experiments using mousemodels of acute and chronic TB (see supplemen-tarymaterials andmethods). In the acute model,which is used to test the antimicrobial activity ofcompounds and regimens against bacteria active-ly multiplying in vivo, mice were aerosol-infectedwith M. tuberculosis, and oral administration ofCGM (at daily doses ranging from 10 to 600mg/kg)was initiated on the day after infection. After4 weeks, all of the untreated control mice haddied, whereas all of the mice treated with anydose of CGM survived. As expected, mice treatedwith isoniazid (INH) at 10 mg/kg experienced adecrease of ~3 log10 colony-forming units (CFUs)in the lungs,whereasmice receivingRIFat 10mg/kg,which is expected to have poor initial activity inthis model (16), did not die but did experiencebacterial growth in the lungs (Fig. 2B). In micereceiving 10 and 25 mg/kg of CGM, neither thedevelopment of gross lung lesions (fig. S2) norbacterial growth could be prevented (Fig. 2B). To-tal prevention of bacterial growth and gross lunglesions occurred in mice treated with 50 mg/kg,defining the minimal effective dose of CGM.The CFU count decreased by ~2 log10 in thelungs comparedwith the CFU count at initiationof treatment with a daily 100-mg/kg dose, de-fining the minimal bactericidal dose. In micetreated with 200-, 400-, and 600-mg/kg doses,a dose-dependent decline in lung CFU countswas observed (P < 0.0001), and the lungs ofmice receiving the 600-mg/kg daily dose wereculture-negative after 4 weeks of treatment. Inthe chronic model, which is used to test forantimicrobial activity against a stable bacterialpopulation in vivo, mice were aerosol-infected,achieving a low implantation of 2.21 log10 CFUper lung, and treatment was initiated 4 weekslater when a stable, host-contained infectionwas established at nearly 7 log10 CFU per lung.After 4 weeks of treatment, all doses of CGMexhibited some degree of activity, with doses of50 mg/kg and higher resulting in statisticallysignificant differences in lung CFU counts fromthe untreated control (P ≤ 0.01) (Fig. 2C andfig. S2). Treatment with CGM at 100 mg/kg re-sulted in a decrease in lung CFU counts similarto that observed with RIF at 10 mg/kg. In both

SCIENCE sciencemag.org 5 JUNE 2015 • VOL 348 ISSUE 6239 1107

Table 1. Optimization parameters for GM and derivatives. Cmax, maximum concentration; AUC, areaunder the concentration curve; Vss, volume of distribution; t½, half-life; nd, not determined. Pharmaco-kinetic parameters (Cmax, AUC, and t½) were determined after a single oral administration of 30mg/kg ofthe test compound inmice. For oral bioavailability, a single oral dose was compared to a single intravenousdose of 3 mg/kg. F = [AUCoral]/[AUCiv], the ratio of exposure of an equivalent dose after nonintravenous(in this case, oral) and intravenous administration as a measure of bioavailability.

Optimization parameterCompound

GM MGM CGM

MIC (mg/ml) for M. tuberculosis in liquid culture 1 0.6 0.06MIC (mg/ml) for M. tuberculosis in liquid culture

containing 25% human serum1.2 0.9 0.3

MIC (mg/ml) for M. tuberculosis in liquid culturecontaining 25% mouse serum

1.2 0.8 0.4

MIC (mg/ml) for M. tuberculosis in liquid cultureunder anaerobic conditions

>33 >33 >35

MIC (mg/ml) for M. tuberculosis in macrophages 6.2 2.1 0.2Unbound fraction (%) in human plasma nd 7.9 0.3Unbound fraction (%) in mouse plasma nd 11.9 0.4Metabolic stability in human liver microsomes

(% remaining compound after 20 min)62 100 86

Metabolic degradation/clearance fromhuman liver microsomes (ml/min per mg)

187.5 3 6.75

CYP1A2, CYP2B6, and CYP3A induction (Yes/No) nd nd NoCYP3A4 inhibition (mg/ml) >33 >33 >35Contribution of CYP3A4 to degradation in

human liver microsomes (%)38 nd 19

Cmax (ng/ml) in plasma 3900 2820 2620Cmax (ng/g) in lung tissue 2580 8550 8850AUC (ng·hour/ml) in plasma 5100 6600 23,000AUC (ng·hour/ml) in lung tissue 5900 12,000 70,000Clearance (liters/hour per kg) 2.4 2.1 1.1Vss (liters/kg) 1.2 1.6 5.5t½ (hours) in plasma 2.0 2.4 4.3t½ (hours) in lung tissue 3.9 4.3 4.3Oral bioavailability (F) (%) 48 47 89

RESEARCH | RESEARCH ARTICLE

Fig. 1. Chemical structure of GM andderivatives. Substitutions at Pro8

increased metabolic stability.

on

May

11,

201

7ht

tp://

scie

nce.

scie

ncem

ag.o

rg/

Dow

nloa

ded

from


6



of these experiments, no overt adverse effectsof CGM administration were observed in treatedmice.Due to the promising bactericidal activity of

CGM when administered as monotherapy, weevaluated the activity of this compound when ad-ministered together with first-line anti-TB drugsin the mouse model of TB chemotherapy. Stan-dard TB treatment comprises a 2-month intensivephase of daily RIF, INH, pyrazinamide (PZA), andethambutol, followed by a 4-month continuationphase of daily RIF and INH. We designed an ex-periment to assess the bactericidal activity of CGM(at theminimal bactericidal dose of 100mg/kg perday) used alone and in combination during thefirst 3 months of treatment. Mice were aerosol-infectedwithM. tuberculosis (implantation of 3.26log10 CFU per lung), and treatment was initiated2weeks after infection, when the lung CFU counthad reached 7.43 log10. In these experimentalconditions, CGM alone was as active as INH

(the most bactericidal first-line drug), reducingCFU counts in the lungs of mice by 3.78 log10over 3 months, indicating that on a molar basis,CGM (1196 g/mol) had equivalent potency toINH (137 g/mol). However, at any time point, thecombination of INH and CGM was as active aseither INHorCGMalone. The combination of CGMwith PZA was as active as the standard combi-nation of INH, RIF, and PZA, reducing lung CFUcounts in the mice by 5.5 log10 over 3 months(table S3). Additionally, the combination of CGMwith RIF was much more active than the stan-dard combination (additional 2 log10 CFUdecline),nearly resulting inmouse lung culture conversionby week 12 of treatment. Even more impressivewas the activity of the three-drug combinationof RIF, PZA, andCGM that resulted inmouse lungculture conversion after only 8weeks of treatment,representing killing of >7 log10 CFU in the lungs.In contrast and as expected, in the mice receivingthe standard drug combination, lung CFU counts

declined by ~4.4 and ~5.7 log10 by weeks 8 and12, respectively (Fig. 2D and table S3). To assessthe activity of CGM (at 100 mg/kg per day) dur-ing the continuation phase, mice received thestandard regimen for the first 8 weeks of treat-ment, and then CGM was administered alone orin combination duringweeks 8 to 20. The bacterialpopulation remaining during the continuationphase had survived the initial 8 weeks of treat-ment andmay have been enriched for organismsthat are phenotypically tolerant to the antibiotics(i.e., the so-called “persisters”). The replacementof either of the drugs in the standard continua-tion regimen (RIF and INH) with CGM did notaffect the time to mouse lung culture conversion(20 weeks), and no statistically significant differ-ences were observed in lung CFU counts betweenthese groups at any time point (table S4). Thebactericidal activity of the combination of INHwith CGM was not significantly different fromthat of CGM alone at any time point. However,


Fig. 2. In vitro and in vivo CGM activity against M. tuberculosis. (A)Activity of CGM against M. tuberculosis in 7H9 broth culture. The numberfollowing the compound abbreviation indicates the concentration in mi-crograms per milliliter. (B) CGM dose-ranging in the mouse model of acuteTB: CFU counts after 4 weeks of treatment. Data represent the mean lunglog10 CFU counts (five mice per group). Day 0 indicates lung CFU counts attreatment initiation (the day after infection). The number following the com-pound abbreviation indicates the dose in milligrams per kilogram per day. (C)CGM dose-ranging in the mouse model of chronicTB: CFU counts after 4 weeksof treatment. Data represent the mean lung log10 CFU counts (5 mice

per group). Day 0 indicates lung CFU counts at treatment initiation (28 daysafter infection). The number following the compound abbreviation indicatesthe dose in milligrams per kilogram per day. (D) CGM administered in com-bination with anti-TB drugs during the intensive phase of treatment in amouse model of TB. Data represent the mean lung log10 CFU counts (5 to10 mice per group). Drug doses: RIF, 10 mg/kg per day; INH, 10 mg/kg perday; PZA, 150 mg/kg per day; CGM, 100 mg/kg per day. w, weeks. Standardtreatment is 8 weeks of daily RIF + INH + PZA, followed by daily RIF + INH.Error bars in all panels represent SD. Dotted lines in (B) and (C) indicate theCFU counts in the lungs when treatment was initiated.


on

May

11,

201

7ht

tp://

scie

nce.

scie

ncem

ag.o

rg/

Dow

nloa

ded

from

of these experiments, no overt adverse effectsof CGM administration were observed in treatedmice.Due to the promising bactericidal activity of

CGM when administered as monotherapy, weevaluated the activity of this compound when ad-ministered together with first-line anti-TB drugsin the mouse model of TB chemotherapy. Stan-dard TB treatment comprises a 2-month intensivephase of daily RIF, INH, pyrazinamide (PZA), andethambutol, followed by a 4-month continuationphase of daily RIF and INH. We designed an ex-periment to assess the bactericidal activity of CGM(at theminimal bactericidal dose of 100mg/kg perday) used alone and in combination during thefirst 3 months of treatment. Mice were aerosol-infectedwithM. tuberculosis (implantation of 3.26log10 CFU per lung), and treatment was initiated2weeks after infection, when the lung CFU counthad reached 7.43 log10. In these experimentalconditions, CGM alone was as active as INH

(the most bactericidal first-line drug), reducingCFU counts in the lungs of mice by 3.78 log10over 3 months, indicating that on a molar basis,CGM (1196 g/mol) had equivalent potency toINH (137 g/mol). However, at any time point, thecombination of INH and CGM was as active aseither INHorCGMalone. The combination of CGMwith PZA was as active as the standard combi-nation of INH, RIF, and PZA, reducing lung CFUcounts in the mice by 5.5 log10 over 3 months(table S3). Additionally, the combination of CGMwith RIF was much more active than the stan-dard combination (additional 2 log10 CFUdecline),nearly resulting inmouse lung culture conversionby week 12 of treatment. Even more impressivewas the activity of the three-drug combinationof RIF, PZA, andCGM that resulted inmouse lungculture conversion after only 8weeks of treatment,representing killing of >7 log10 CFU in the lungs.In contrast and as expected, in the mice receivingthe standard drug combination, lung CFU counts

declined by ~4.4 and ~5.7 log10 by weeks 8 and12, respectively (Fig. 2D and table S3). To assessthe activity of CGM (at 100 mg/kg per day) dur-ing the continuation phase, mice received thestandard regimen for the first 8 weeks of treat-ment, and then CGM was administered alone orin combination duringweeks 8 to 20. The bacterialpopulation remaining during the continuationphase had survived the initial 8 weeks of treat-ment andmay have been enriched for organismsthat are phenotypically tolerant to the antibiotics(i.e., the so-called “persisters”). The replacementof either of the drugs in the standard continua-tion regimen (RIF and INH) with CGM did notaffect the time to mouse lung culture conversion(20 weeks), and no statistically significant differ-ences were observed in lung CFU counts betweenthese groups at any time point (table S4). Thebactericidal activity of the combination of INHwith CGM was not significantly different fromthat of CGM alone at any time point. However,


Fig. 2. In vitro and in vivo CGM activity against M. tuberculosis. (A)Activity of CGM against M. tuberculosis in 7H9 broth culture. The numberfollowing the compound abbreviation indicates the concentration in mi-crograms per milliliter. (B) CGM dose-ranging in the mouse model of acuteTB: CFU counts after 4 weeks of treatment. Data represent the mean lunglog10 CFU counts (five mice per group). Day 0 indicates lung CFU counts attreatment initiation (the day after infection). The number following the com-pound abbreviation indicates the dose in milligrams per kilogram per day. (C)CGM dose-ranging in the mouse model of chronicTB: CFU counts after 4 weeksof treatment. Data represent the mean lung log10 CFU counts (5 mice

per group). Day 0 indicates lung CFU counts at treatment initiation (28 daysafter infection). The number following the compound abbreviation indicatesthe dose in milligrams per kilogram per day. (D) CGM administered in com-bination with anti-TB drugs during the intensive phase of treatment in amouse model of TB. Data represent the mean lung log10 CFU counts (5 to10 mice per group). Drug doses: RIF, 10 mg/kg per day; INH, 10 mg/kg perday; PZA, 150 mg/kg per day; CGM, 100 mg/kg per day. w, weeks. Standardtreatment is 8 weeks of daily RIF + INH + PZA, followed by daily RIF + INH.Error bars in all panels represent SD. Dotted lines in (B) and (C) indicate theCFU counts in the lungs when treatment was initiated.


on

May

11,

201

7ht

tp://

scie

nce.

scie

ncem

ag.o

rg/

Dow

nloa

ded

from


7

Research articles


this combination was significantly more bacte-ricidal than INH alone after 16 (P ≤ 0.05) and 20(P ≤ 0.0001) weeks of treatment, with the effectbeing additive. Thus, taken together, our data in-dicate that CGM is highly active against activelyreplicating M. tuberculosis, both in vitro and invivo, and also exhibits bactericidal activity againstnonreplicating bacteria in vivo.

Self-resistance to GM in Streptomyces

The demonstrated activity of CGM against bothdrug-susceptible and -resistant M. tuberculosis(14, 15) and the notable enhancement of the bac-tericidal activity of the standard TB drug regimenwith the addition of CGM in mice suggested thatthe mechanism of action of CGM may be differ-ent from other anti-TB drugs. GM and MGM arenaturally produced by some species of Strepto-myces, and we identified the biosynthetic genecluster responsible for GM and MGM productionin S. sp. DSM-40835 (17), a strain naturally resist-ant to GM, and examined it for the presence of apossible resistance-conferring component. Wefound a homolog of the dnaN gene, which en-codes the sliding clamp (also called the b clamp)of DNA polymerase, within the nonribosomalpeptide megasynthetase cassette for GM synthe-sis (fig. S3A and table S5). This dnaN homolog,annotated as griR, encoded a protein with 51%identity to DnaN from the same strain (fig. S3,B and C). To examine the effect of GriR on sus-ceptibility to GM, we introduced the gene intoStreptomyces coelicolor A3(2), a strain suscepti-ble to this natural product (and naturally lack-

ing the entire GM biosynthesis cassette) (18).The presence of the griR-expressing plasmid al-lowed the strain to survive in the presence of GM(fig. S3D and table S6), suggesting that over-expression of griRmediated GM resistance. Com-parison of the DnaN sequence from differentmicrobial species revealed a clear cluster breakbetween GM-sensitive and -resistant bacteria basedon homology of the sliding clamp (fig. S3C andtable S6). Taken together, these data suggest thatGM and its derivatives target DnaN and may thusinterfere with DNA replication.

GM resistance in mycobacteria

The intriguing relationship between dnaN se-quence, copynumber of the gene, and resistance toGM observed in Streptomyces compelled us to ex-amine the mechanism(s) of resistance to this com-pound in mycobacteria. Using the fast-growing,nonpathogenic, andGM-sensitive (MIC: 4.5 mg/ml)M. smegmatis, we selected GM-resistant bac-teria in vitro. The GM-resistant M. smegmatis(which occurred at an extremely low frequency;at a GM concentration of 10 mg/ml the frequencywas 5 × 10−10) exhibited an altered, elongatedcellular morphology (Fig. 3A), as has been pre-viously observed in GM-exposed M. tuberculosis(9). The GM-resistant bacteria were cross-resistantto MGM and CGM, but not to RIF or any othertested antibiotic (table S7). Genome sequence com-parisons between thewild-type (WT) parent strainand mutants generated via stepwise exposure toincreasing concentrations of GM (up to 40 mg/ml)revealed amplification of a chromosomal segment

containing dnaN, with a single point mutation inthe Pribnow box of the promoter region (115 G >A) (19) (tables S8 and S9), suggesting that myco-bacterial resistance to GM, like that of Strepto-myces, is mediated by amplification of DnaN. Anumber of inconsistent single-nucleotide poly-morphisms (SNPs)were also observed.Allmutantsanalyzed contained head-to-tail repeats of a chro-mosomal segment, but these segments or ampli-cons varied in size (ranging from 12 to 28 kb inlength) and copy number (ranging from 3 to 49copies), and occurrence of the amplicons wasobserved at different steps in the selection pro-cess (Fig. 3B, fig S4, and table S8). In addition todnaN and its promoter, the amplicons containeddnaA, recF, gyrB, and up to more than 20 ad-ditional genes (table S9); they also all includedthe ori site. Amplification of a chromosomal seg-ment was confirmed by Southern blot (Fig. 3, Cand D). Based on the high copy number of theamplicons that include the ori site, formation ofan extrachromosomal element could not be ruledout. However, plasmids were not detected by ei-ther standard plasmid isolation methods or aga-rose gel analysis. Yet in both M. smegmatis andM. tuberculosis, plasmids containing the sameori as the chromosome are usuallymaintained ata relatively low copy number (20, 21), and natu-rally occurring plasmids inmycobacteria are knownto be difficult to isolate (22).Although the exact sequence of the amplicons

varied amongGM-resistantmutants, the 115 G >Atransition mutation in the dnaN promoter oc-curred in allmutants at the2.5-mg/mlGMexposure


Fig. 3. Emergence of GM resistance in M. smegmatis by targetamplification. (A) Phase-contrast images of WTand GM-resistantM. smegmatis cells. Scale bars, 10 mm. (B) Overview of the geneticchanges associated with increased exposure to GM forM. smegmatismutants 1.GM2.5 to 1.GM40 (subscripts indicate GM concentration inmicrograms per milliliter when the mutant was obtained). Red starsrepresent the 115 G > A mutation in the Pribnow box of the dnaNpromoter region. (C) Southern blot and (D) fragmentation patternof EcoRI-digested genomic DNA of WTM. smegmatis and GM-resistant mutants. EcoRI cuts downstream of the dnaN gene inside the amplicon and upstream ofthe dnaN gene outside the amplicon, resulting in a 15.9-kb fragment for samples without amplicons and 15.9-kb and 10.2-kb fragments for samples withamplicons. EcoRI restriction sites are indicated in red; the binding site of the probe (in the dnaN gene) is indicated in blue.


on

May

11,

201

7ht

tp://

scie

nce.

scie

ncem

ag.o

rg/

Dow

nloa

ded

from


8



step, leading to anMIC upshift from the baseline4 mg/ml to 8 to 16 mg/ml (table S8). This pointmutation was not consistently maintained in allsteps of increasingGM concentration, being eitherreplaced or complemented in mutants exposedto higher GM concentrations by amplification ofa dnaN-containing chromosomal segment, whichled to resistance up to or greater than 64 mg/ml.Resistance to GM was accompanied by consider-able fitness costs, whereby M. smegmatis growthwas negatively correlatedwith increasing levels ofresistance (fig. S5), in some cases growing sopoorly that we could not determine MIC values(table S8). Our data indicate that when GM isremoved from the growth media, these geneticandphenotypicmodifications begin to be reversed,including a decrease in amplicon copy number, adownshift in MIC, and less growth retardation(fig. S5 and table S8).The unprecedented mechanism of gaining re-

sistance to GM in our in vitro experiments withM. smegmatis prompted us to determine the rel-evance of these findings during in vivo infectionwith M. tuberculosis. To select resistant bacteriain vivo, nude mice were aerosol-infected withthe fully drug-susceptibleM. tuberculosis strainH37Rv and then treated by monotherapy witheither INH (5 mg/kg) or CGM (100 mg/kg). With-out the pressure from a functional immune sys-tem,monotherapy (when administered once daily,5 days per week) with even the most bactericidaldrug (INH) cannot prevent bacterial multiplica-tion in these mice when drug levels in the bloodfall below the therapeutic concentration, which,based on the half-life of INH (~1.5 hours) (23)andCGM(~4hours) (Table 1) in themouse, shouldoccur daily and over the weekends. As anticipated,after 4 weeks of treatment, the lung bacterialloads began to increase in mice receiving eitherINH or CGM (fig. S6A). CFUs isolated from theCGM-treated mice were more than 10 times lesssensitive to CGM (MIC: 1 to 2 mg/ml) (see pro-portion of CFUs growing onCGM-containing agarplates in table S10). Using the agar proportionmethod, the CGM-resistantM. tuberculosis werefound to be fully susceptible to INH and RIF, aswell as moxifloxacin and streptomycin. Genomicanalysis of CGM-resistant colonies revealed anamplification of a 10.3-kb chromosomal segmentcontaining the dnaN gene, together with fourother genes and the ori site (fig. S6B and tableS11), which was present in the CGM-resistantM.tuberculosis but absent in the WT parent strain.No SNPs were observed in the CGM-resistantbacteria compared with the WT parent strain.Thus, the mechanism of resistance observedin vitro with M. smegmatis was also observedin vivo with M. tuberculosis.

Characterization and crystalstructure analysis of theGM-DnaN interaction

To confirmour genomic-based findings thatDnaNis the target of GM and its derivatives, we usedsurface plasmon resonance (SPR) to characterizethe binding of GM, MGM, and CGM with DnaNfrom M. smegmatis, M. tuberculosis, and Esche-

richia coli, as well as with the human slidingclamp [proliferating cell nuclear antigen (PCNA)].Binding was also characterized with GriR, theDnaNhomolog from theGMbiosynthesis clusterin Streptomyces. SPR analysis demonstrated bind-ing of GM to the mycobacterial sliding clampswith high affinity (equilibrium dissociation con-stant KD: 8.3 × 10−11 M and 1.0 × 10−10 M for M.smegmatis andM. tuberculosis DnaN proteins,respectively) and a fast recognition rate andslowdissociation from the protein (M. smegmatis,association rate constant ka: 2.2 × 107 M−1 s−1,dissociation rate constant kd: 1.9 × 10−3 s−1; M.tuberculosis, ka: 8.6 × 106M−1 s−1, kd: 8.4 × 10−4 s−1)(Table 2 and fig. S7). No binding was detectedbetween GM and the human sliding clamp, butbinding was observed at a significantly lower lev-el with the E. coli clamp (KD: 6.5 × 10−7 M),despite the lack of in vitro activity of GM againstthis organism (table S6). This lack of activity couldbe associated with the fast dissociation of GM

from the E. coli DnaN compared with the myco-bacterial sliding clamps (fig. S7), or it may becaused by efflux and/or impaired penetration inE. coli. Binding of GM was also observed to theGMresistance-conferringGriR, but at a significant-ly lower level compared with the mycobacterialDnaN proteins. It is possible that this weakerbinding may be sufficient for self-resistance inStreptomyces. Binding of GM to GriR showed akinetic profile similar to that of the E. coli DnaN,with fast association and dissociation rates butwith somewhat lower affinity. MGM and CGM ex-hibited kinetic profiles similar to GM when inter-acting with the M. tuberculosis sliding clamp,whereas the interaction ofM. smegmatisDnaNwith CGM was characterized by a slower dis-sociation rate (kd: 4.1 × 10−4 s−1) and a slightlyslower recognition rate (ka: 4.5 × 106 M−1 s−1).Thus, it may be the increasing lipophilicity (thusinfluencing the ability to traverse the cell wall)of the GM derivatives, rather than changes in


Table 2. SPR-based kinetic parameters of sliding clamp interactions with GM, MGM, and CGM. ForM. smegmatis andM. tuberculosis DnaN, the KD values were determined from the ratio between the kineticrate constants (ka/kd), and the dissociative half-life t1/2 was calculated by ln2/kd. For interactions with faston and off rates (E. coli and S. sp. DSM-40835 DnaN proteins), KD values were determined by steady-stateaffinity analyses from the dependence of steady-state binding levels on analyte concentrations. Datarepresent mean and SD from three independent experiments. SPR sensorgrams are presented in fig. S7.na, not applicable.

Interactants KD (M)[SD]

ka (M−1 s−1)[SD]

kd (s−1)[SD]

t1/2 (s)[SD]Protein Compound

DnaNM. smegmatis

GM 8.3 × 10−11

[3.9 × 10−11]2.2 × 107

[5.4 × 105]1.9 × 10−3

[8.7 × 10−4]450[160]

MGM 9.9 × 10−10

[5.7 × 10−11]1.6 × 107

[5.5 × 106]1.3 × 10−3

[1.9 × 10−4]545[179]

CGM 1.2 × 10−11

[4.2 × 10−11]4.5 × 106

[3.4 × 106]4.1 × 10−4

[1.7 × 10−4]1998[785]

DnaNM. tuberculosis

GM 1.0 × 10−10

[8.1 × 10−10]8.6 × 106

[2.5 × 106]8.4 × 10−4

[1.9 × 10−4]863[183]

MGM 1.1 × 10−10

[1.1 × 10−11]6.5 × 106

[2.1 × 106]7.1 × 10−4

[2.1 × 10−4]1080[361]

CGM 2.0 × 10−10

[7.9 × 10−11]2.6 × 106

[7.8 × 105]4.7 × 10−4

[1.4 × 10−4]1659[643]

DnaNE. coli

GM 6.5 × 10−7

[9.5 × 10−8]na na na

MGM 8.4 × 10−7

[5.2 × 10−8]na na na

CGM 6.6 × 10−7

[3.0 × 10−8]na na na

GriRS. sp.DSM-40835

GM 5.8 × 10−6

[1.4 × 10−6]na na na

MGM 6.6 × 10−6

[6.8 × 10−7]na na na

CGM 6.5 × 10−6

[3.2 × 10−7]na na na

PCNA GM No binding detectedMGM No binding detectedCGM No binding detected


on

May

11,

201

7ht

tp://

scie

nce.

scie

ncem

ag.o

rg/

Dow

nloa

ded

from


9

Research articles


binding kinetics, that explains their more potentantimycobacterial activity compared with GM.To gain a better understanding of the molec-

ular interactions responsible for GM-DnaN com-plex formation, cocrystal structures of GM andCGMbound tomycobacterial DnaNproteinsweredetermined. The structures ofM. smegmatisDnaNwith GM and CGM were refined to resolutions of2.1 and 2.3 Å, respectively, whereas the structuresof theM. tuberculosis DnaN with bound GM andCGM were refined to resolutions of 2.2 and 1.9 Å,respectively (table S12). Both inhibitors were welldefined in the electron density maps (fig. S8). GMand CGM bound to a hydrophobic cleft betweendomains II and III of the sliding clamp (Fig. 4A),which has been identified previously as the pep-tide interaction site responsible for the bindingof DNA polymerases and other DNA-modifyingenzymes (24, 25). Binding of GM is mainly ac-complished via hydrophobic interactions.Only twoamino acid residues are involved in direct hydro-gen bonding: Arg181 (or Arg183) forM. smegmatis(or M. tuberculosis) forms two hydrogen bondsto the ligand, one via its backbone carbonyloxygen and one via its guanidinemoiety, whereasArg394 (or Arg399) for M. smegmatis (or M.tuberculosis) is involved in hydrogen bondingthrough its backbone amide nitrogen. Addition-ally, a bridging water molecule is coordinated

between the ligand and Pro392 (or Pro397) for M.smegmatis (orM. tuberculosis) (Fig. 4B and fig.S9). Subsite one of the peptide-binding pocketis occupiedby the cyclic part of the ligand,whereassubsite two harbors the linear part of the GMmolecule (Fig. 4C). This linear section and theadjacent half of themacrocycle superimpose verywell for GM and CGM, whereas differences arevisible on the opposite side of the macrocycle(figs. S10 and S11A). The ligands seem to be lesstightly bound in this area, thus exhibiting ahigher degree of flexibility (fig. S11B). The addi-tional cyclohexyl moiety of CGM protrudes intothe solvent and is involved in the formation ofcrystal contacts (fig. S12); the crystal packingmightinduce the observed differences in CGM bindingas compared with GM and thus may not be phys-iologically relevant.Binding of GM to the peptide interaction site

of DnaN should lead to inhibition of the DnaNinteraction with the polymerase III a subunit(DnaE1). To test this, we used SPR to assess theeffect ofGMon the binding ofDnaNwith aDnaE1peptide that contains the internal DnaN bindingmotif and is known to be essential for processivereplication in E. coli (26). Binding of this peptidedecreased by half when the sliding clamp wassaturated with GM (fig. S13). Inhibition of DnaE1binding to DnaN should result in inefficient rep-

lication with decreased processivity and may leadto DNA strand breaks. Induction of the SOS re-sponsewas observed inGM-exposedM. smegmatis(fig. S14).Several functional and structural studies have

indicated that the sliding clampmay be a feasibleantibacterial target that can be addressed by smallmolecules and peptidic inhibitors (24, 27–29).Although inhibition of DNA replication doesnot necessarily lead to immediate bacterial celldeath, blocking the replication process may in-duce other lethal damage within the cell, such asinduction of the SOS response (fig. S14) and/ortoxin-antitoxin–like stress, as in the case of thequinolones, which block DNA replication by tar-geting topoisomerases but kill bacteria rapidly(30). In fact, DnaN was recently shown to be tar-geted by a toxin-antitoxin system in Caulobactercrescentus (31). The cell-elongation phenotype ofGM-exposed mycobacteria indicates that block-ing DNA replication induces a wider cellular re-sponse; this cell-elongation effect has also beenreported in M. fortuitum upon exposure to thefluoroquinolone ofloxacin (32). Furthermore, inboth Gram-positive and Gram-negative bacteria,the sliding clamp has been demonstrated to in-teract with DNA repair proteins, including manyinduced during the SOS response (33). Thus, theinteraction of GMs with DnaN could also affectDNA repair, contributing to the bactericidal ac-tivity of the compounds. Whereas previously re-ported DnaN binders inhibited bacterial growthwith only moderate potency, GM and its deriv-atives represent the first DnaN inhibitors withpicomolar affinity to the target and high efficacy(MIC values ≤1.0 mg/ml) against mycobacteria.Known small-molecule DnaN binders targetedthe deep hydrophobic pocket of subsite one,whereas subsite two remained empty. Peptidicinhibitors and the natural interacting peptidestarget both subsites; the linear segment and partof the macrocycle of GM superimposed very wellwith their backbone, thereby mimicking theconformation of a linear peptide bound to DnaN(fig. S15). The specific inhibition of mycobacteriaby GM, which should leave commensal micro-biota intact, constitutes an additional advantagewith respect to the selection of resistance.

Conclusion

Total synthesis of new GMs led to the optimizedCGM, which demonstrated a clear potential forTB treatment by targeting the DNA polymerasesliding clamp. Currently, CGM and a few otherGM derivatives are being further profiled to se-lect the best drug candidate to move forward indevelopment, with CGM being one of the leadcompounds pursued in late preclinical studiesinvolving broad toxicity testing. In this work, wehave demonstrated the validity of DnaN as a drugtarget and have identified direct small-molecule–DnaN interactions that probably interrupt es-sential polymerase and DNA repair activities inmycobacteria, leading to killing both in vitro andin vivo. Through the selection of GM-or CGM-resistant bacteria, we discovered that resistanceto these compounds is possible, albeit at a very


Fig. 4. Crystal structure of the M. smegmatis slidingclamp (DnaN) in complex with GM. (A) Crystal structureof the M. smegmatis DnaN dimer in complex with GM at aresolution of 2.1 Å. GM binds to a hydrophobic pocket be-tween domains II and III, which is known as the protein-protein interaction site responsible for therecruitment of DNA polymerases by the sliding clamp. The cartoon representation of the left half of thehomodimeric ring is colored from theN terminus (blue) to the C terminus (red).The second half of the ringis shown as surface representation colored according to the electrostatic surface potential, rangingfrom –5 kT/e (red) to +5 kT/e (blue) (k, Boltzmann’s constant; T, temperature; e, charge of an electron).(B) Interactions between the ligand and the protein. The residues involved in hydrophobic contacts withGM are represented as gray sticks. The residues also involved in hydrogen bonding with the ligand arerepresented as green sticks. Hydrogen bonds are indicated by dashed green lines. F, Phe; R, Arg; M, Met;P, Pro;T,Thr; L, Leu;V,Val. (C) Surface representation of the binding pocket.The surface is colored accordingto the electrostatic surface potential. Both subsites of the binding pocket are indicated by circles.


on

May

11,

201

7ht

tp://

scie

nce.

scie

ncem

ag.o

rg/

Dow

nloa

ded

from


10



low frequency, and is mediated through multi-plication of dnaN, whereby the resistance to GMsincreases with increasing dnaN amplification.The nature of the amplicons, consisting of clonalrepeats that differ slightly between bacterial iso-lates but all include the ori site, and the observeddecrease in target amplification in the absence ofGM suggest that the amplification is connected tosevere fitness costs, as observedby the slowgrowthof the mutants. As exposure to GM triggeredchromosomal duplications in a concentration-dependent manner in vitro, GM could possiblyserve as a tool to study the mechanism of chro-mosomal duplication events in mycobacteria thathave contributed to the natural variation betweendifferent lineages. Because there is no preexistingresistance to GMs due to the different mode ofaction compared with that of existing TB drugs,and because resistance occurs at an extremely lowfrequency and is associated with a severe fitnesscost, this new series has the potential to contrib-ute to drug regimens with utility for patients withboth drug-sensitive and drug-resistant TB. In ad-dition, our work has confirmed the sliding clampas an antimicrobial target and revealed an unusualmechanism of conferring drug resistance that isapplicable to the wider antibiotic discovery field.

REFERENCES AND NOTES

1. A. Smith, Ed., Br. Med. J. 1, 278–279 (1946).2. British Thoracic Society, Br. J. Dis. Chest 78, 330–336 (1984).3. World Health Organization, “Antimicrobial resistance: Global

report on surveillance” (World Health Organization Press,Geneva, Switzerland, 2014).

4. D. J. Newman, G. M. Cragg, J. Nat. Prod. 75, 311–335(2012).

5. R. E. Lee et al., Nat. Med. 20, 152–158 (2014).6. R. C. Hartkoorn et al., EMBO Mol. Med. 4, 1032–1042 (2012).7. B. Terlain, J. P. Thomas, Bull. Soc. Chim. Fr. 6, 2349–2356

(1971).8. F. Bénazet et al., in Antibiotics – Advances in Research,

Production and Clinical Use: Proceedings of the Congress onAntibiotics held in Prague, 15-19 June, 1964, M. Herold, Z. Gabriel,Eds. (Butterworths, London, 1966), pp. 262–264.

9. H. Noufflard-Guy-Loé, S. Berteaux, Rev. Tuberc. Pneumol.(Paris) 29, 301–326 (1965).

10. B. Terlain, J. P. Thomas, C. R. Hebd. Seances Acad. Sci. Ser. C269, 1546–1549 (1969).

11. B. Terlain, J. P. Thomas, Bull. Soc. Chim. Fr. 6, 2357–2362(1971).

12. G. Jolles, New cyclopeptides, G.B. Patent 1,252,553 (1971).13. J. Bouchaudon, Nouveau cyclopeptide, sa préparation et les

médicaments qui le contiennent, Fr. Patent 2,469,395 (1981).14. Anonymous, New antibiotic product, its preparation and

compositions containing it, G.B. Patent 966,124 (1964).15. M. Toyohara, Ann. Inst. Pasteur Microbiol. 138, 737–744 (1987).16. E. L. Nuermberger, “The role of the mouse model in the

evaluation of new antituberculosis drugs,” in AntituberculosisChemotherapy, P. R. Donald, P. D. Van Helden, Eds. (Karger,Basel, Switzerland, 2011), pp. 145–152, vol. 40, chap. 15.

17. M. Broenstrup et al., Gene cluster for biosynthesis ofgriselimycin and methylgriselimycin. PCT Int. Appl. WO2013/053857 (2013).

18. S. D. Bentley et al., Nature 417, 141–147 (2002).19. L. Salazar, E. Guerrero, Y. Casart, L. Turcios, F. Bartoli,

Microbiology 149, 773–784 (2003).20. M. H. Qin, M. V. Madiraju, M. Rajagopalan, Gene 233, 121–130

(1999).21. L. Salazar et al., Mol. Microbiol. 20, 283–293 (1996).22. F. Movahedzadeh, W. Bitter, Methods Mol. Biol. 465, 217–228

(2009).23. J. Grosset, B. Ji, in Mycobacteria Volume II Chemotherapy,

P. R. J. Gangadharam, P. A. Jenkins, Eds. (Chapman and Hall,New York, 1998), pp. 51–97.

24. K. A. Bunting, S. M. Roe, L. H. Pearl, EMBO J. 22, 5883–5892(2003).

25. D. Y. Burnouf et al., J. Mol. Biol. 335, 1187–1197 (2004).26. P. R. Dohrmann, C. S. McHenry, J. Mol. Biol. 350, 228–239

(2005).27. G. Wijffels et al., J. Med. Chem. 54, 4831–4838 (2011).28. Z. Yin et al., J. Med. Chem. 57, 2799–2806 (2014).29. R. E. Georgescu et al., Proc. Natl. Acad. Sci. U.S.A. 105,

11116–11121 (2008).30. K. Drlica, M. Malik, R. J. Kerns, X. Zhao, Antimicrob. Agents

Chemother. 52, 385–392 (2008).31. C. D. Aakre, T. N. Phung, D. Huang, M. T. Laub, Mol. Cell 52,

617–628 (2013).32. H. Saito, T. Watanabe, T. Hirata, Zentralbl. Bakteriol. Mikrobiol.

Hyg. A 268, 448–455 (1988).33. G. Wijffels, B. Dalrymple, K. Kongsuwan, N. E. Dixon, IUBMB

Life 57, 413–419 (2005).

ACKNOWLEDGMENTS

We thank the staff at beamlines P11 at PETRA III (DESY, Hamburg,Germany), X06DA at the Swiss Light Source (PSI, Villigen,Switzerland), and BL14.1 at BESSY II (Berlin, Germany) forassistance during data collection. R.M.’s research laboratory wassupported by the German Ministry for Education and Research(BMBF) with grant fZ031598. Sanofi funded its contribution. D.V.A.,N.C.A., and J.H.G. are supported by the KwaZulu-Natal ResearchInstitute for Tuberculosis and HIV and the Howard Hughes MedicalInstitute. D.V.A., S.T., E.N., and J.H.G. were supported by twogrants from Sanofi-Aventis (Amd. 14, grant 111937, and Amd. 7,grant 103671). We thank C. Wylegalla for technical assistance. Weacknowledge the GMAK group at HZI for assistance with genomesequencing of bacterial mutants; S. Franzblau from the Universityof Illinois at Chicago for performing the in vitro MIC assays bothunder hypoxic conditions and with different strains of M.tuberculosis (testing agreement) and for the gift of luciferase-expressing M. tuberculosis; and P. Brodin from the PasteurInstitute, Lille, France, for the gift of GFP-expressing M. tuberculosisH37Rv. We thank A. Upton from the TB Alliance for analysis of themanuscript. The sequence of the Streptomyces sp. DSM-40835GM biosynthetic gene cluster has been deposited in GenBank(accession number KP211414). Atomic coordinates and structurefactors for the reported crystal structures have been deposited inthe Protein Data Bank (identification numbers listed in table S12).

A.B., E.F., S.S., M.B.B., F.B.-P., M.B., G.C., M.G., P.H., S.K., C.L.,H.M., C.K., K.B., M.Ger., H.H., M.Ko., M.Ku., L.F., and S.L. areemployed by Sanofi-Aventis R&D. Sanofi has filed patentapplications on GM derivatives. Author contributions: A.K.performed resistance studies and DnaN expression andphysicochemical analysis of the interaction of the protein withGM derivatives. P.L. performed expression of M. smegmatis DnaNand all structural biology experiments. P.L. and D.W.H. wereinvolved in structural analyses and data interpretation. P.H. andF.B.P. identified the GM series from the Rhône-Poulenc archives.G.C., M.Ges., H.H., A.B., and E.F. designed and synthesized allGM derivatives. S.S., M.B.B., E.F., and A.B. managed the drugoptimization program. S.L., M.B., and L.F. supervised the overallGM program at Sanofi. C.L. and M.B.B. designed and performedthe in vitro experiments with M. tuberculosis H37Rv. S.K. andM.Ko. designed and performed the pharmacokinetic evaluations.M.Ko. and K.B. elucidated the metabolic degradation of GM and itsderivatives. M.Ger. and M.Ku. performed analytics and structureconfirmation of GM and its derivatives. M.Ku. and H.M. performedconformational analysis and designed models for GM derivativeswith improved pharmacokinetic properties. D.V.A., N.C.A., E.N.,S.T., and J.H.G. designed and performed all animal infectionexperiments, including associated assays, and analyzed allresulting data. N.Z. analyzed all genome data. J.H. performedactivity assays and identified and analyzed resistant M. smegmatisstrains. S.C.W. and C.K. identified and characterized the GMbiosynthetic gene cluster. A.K., P.L., A.B., N.C.A., S.L., M.B., andR.M. conceived the studies and wrote the paper. A.K. and P.L.contributed equally to the study. All authors discussed the resultsand commented on the manuscript.

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/348/6239/1106/suppl/DC1Materials and MethodsSupplementary TextFigs. S1 to S15Tables S1 to S12References (34–76)

12 December 2014; accepted 16 April 201510.1126/science.aaa4690

REPORTS◥

MULTIFERROICS

Magnetoelectric domain control inmultiferroic TbMnO3Masakazu Matsubara,1,2*† Sebastian Manz,1*† Masahito Mochizuki,3,4 Teresa Kubacka,5

Ayato Iyama,6 Nadir Aliouane,7 Tsuyoshi Kimura,6 Steven L. Johnson,5

Dennis Meier,1 Manfred Fiebig1

The manipulation of domains by external fields in ferroic materials is of major interest forapplications. In multiferroics with strongly coupled magnetic and electric order, however,the magnetoelectric coupling on the level of the domains is largely unexplored. Weinvestigated the field-induced domain dynamics of TbMnO3 in the multiferroic groundstate and across a first-order spin-flop transition. In spite of the discontinuous nature ofthis transition, the reorientation of the order parameters is deterministic and preserves themultiferroic domain pattern. Landau-Lifshitz-Gilbert simulations reveal that this behavioris intrinsic. Such magnetoelectric correlations in spin-driven ferroelectrics may lead todomain wall–based nanoelectronics devices.

The entanglement of magnetic and electriclong-range orders in magnetically inducedferroelectrics is a key to the manipulationof dielectric properties by magnetic fieldsand vice versa. Long believed to be a low-

temperature phenomenon, these strong magneto-electric correlations are now gaining importancein the light of recent studies that report spin-driven ferroelectricity in conditions approachingroom temperature (1–3).


RESEARCH

on

May

11,

201

7ht

tp://

scie

nce.

scie

ncem

ag.o

rg/

Dow

nloa

ded

from





























Life 57, 413–419 (2005).

ACKNOWLEDGMENTS






REPORTS◥

MULTIFERROICS








RESEARCHon June 19, 2017

http://science.sciencemag.org/

Dow

nloaded from





























Life 57, 413–419 (2005).

ACKNOWLEDGMENTS






REPORTS◥

MULTIFERROICS








RESEARCH

on June 19, 2017http://science.sciencem

ag.org/D

ownloaded from





























Life 57, 413–419 (2005).

ACKNOWLEDGMENTS






REPORTS◥

MULTIFERROICS








RESEARCH


ag.org/D

ownloaded from





























Life 57, 413–419 (2005).

ACKNOWLEDGMENTS






REPORTS◥

MULTIFERROICS








RESEARCH


ag.org/D

ownloaded from





























Life 57, 413–419 (2005).

ACKNOWLEDGMENTS






REPORTS◥

MULTIFERROICS








RESEARCH


ag.org/D

ownloaded from





























Life 57, 413–419 (2005).

ACKNOWLEDGMENTS






REPORTS◥

MULTIFERROICS








RESEARCH


ag.org/D

ownloaded from


http://www.sciencemag.org/content/348/6239/1106/suppl/DC1

11

Research articles

SCIENCE sciencemag.org Originally published 5 Nov 2014 in SCIENCE TRANSLATIONAL MEDICINE

INTRODUCTION

The blood-brain barrier (BBB) remains a formidable obstacle for developing therapeutics to treat neurological disease, particularly for large molecules such as antibodies (1–3).

Previous efforts exploiting the transferrin (Tf)/transferrin receptor (TfR) pathway to enhance large-molecule uptake in the brain via receptor-mediated transcytosis have focused exclusively on rodent TfR binding antibodies (4, 5). Thus, it remains to be shown if the TfR pathway can be used to cross the BBB in primates. TfR pathway validation for brain delivery in primates is critical if this approach is to be used in a clinical setting for central nervous system disease targets with access restricted by the BBB. Here, we describe the generation and characterization of primate-specific TfR bispecific antibodies designed to boost antibody uptake into the primate brain.

RESULTS

Generation of primate bispecific anti-TfR/BACE1 antibody variants

Previously, anti-TfRA, an antibody that binds selectively to mouse TfR, was engineered to investigate the TfR pathway for brain uptake and safety liabilities in mice (6, 7). We showed that by reducing affinity for TfR in either the bivalent or the monovalent (bispecific) format, brain uptake and biodistribution were improved. Here, we have now generated a new antibody to primate TfR, which cross-reacts with both human and monkey TfR expressed on human embryonic kidney (HEK) 293 cells (Fig. 1A). This antibody was then humanized to generate anti-TfR1. Anti-TfR1 does not block ligand binding to TfR, as shown by assessing the binding of anti-TfR1 to TfR in the presence or absence of Tf (Fig. 1B) and human hemochromatosis (HFE) protein (Fig. 1C). Consistent with its nonblocking properties, the anti-TfR1 binding site was mapped to the apical domain, a site distant from these ligand binding sites (Fig. 1D). To assess TfR affinity relationships for brain uptake, we generated a second lower-affinity variant called anti-TfR2 by introducing a single alanine substitution into CDR-L3 (complementarity determining region L3) of the anti-TfR1 light chain (Fig. 1E). Monovalent TfR binding of anti-TfR2 showed a 20- to 25-fold decrease in affinity to both human and monkey TfR (Fig. 1E). Furthermore, there was an about fourfold decrease in monovalent

binding affinity observed from human to monkey TfR for both anti-TfR variants, which is particularly relevant because previous studies have shown that modest changes in affinity can affect brain uptake (6, 8).

Using “knob into holes” technology, a meth-od that promotes heterodimerization of two different half-antibodies into a bispecific im-munoglobulin G (IgG) antibody (9), primate cross-reactive anti-TfR variants were combined with an antibody directed against a target in the brain, β-secretase (BACE1), thus generating anti-TfR/BACE1 bispecific antibodies (Fig. 1E). BACE1 is an aspartyl protease responsible for the first of two cleavage events in amyloid pre-cursor protein (APP) that give rise to amyloid-β (Aβ) peptides, which accumulate as plaques in the brains of patients with Alzheimer’s disease (10). Although BACE1 is a therapeutic target for Alzheimer’s disease, inhibition of BACE1 also provides an ideal pharmacodynamic (PD) signal for assessing the activity of antibodies that cross the BBB. To assess brain activity of the TfR bispecific antibody after systemic deliv-ery in primates, we used a previously described anti-BACE1 antibody that binds to mouse, mon-key, and human BACE1 (11). As expected, both anti-TfR1/BACE1 and anti-TfR2/BACE1 were equally potent at reducing Aβ production in mouse neurons (Fig. 1F).

Characterization of a human TfR knock-in mouse model

Brain uptake of the anti-TfR/BACE1 bispecific antibodies was first tested using a human TfR knock-in mouse, generated by introducing the human TFRC complementary DNA (cDNA) af-ter exon 1 of the mouse Tfrc gene (fig. S1). To confirm TfR species specificity, we dosed wild-type and human TfR (TFRC) knock-in mice with control IgG, mouse-specific anti-TfRA, or primate-specific anti-TfR1. We observed anti-TfR antibody localization in the cortical vasculature of mouse brain 1 hour after dos-ing wild-type mice with mouse-specific anti-TfRA only, whereas primate-specific anti-TfR1 only stained human TfR knock-in mice (Fig. 2A). Brain antibody concentrations mea-sured 1 day after dosing with anti-TfR/BACE1 (20 mg/kg) demonstrated brain uptake in a species-specific manner (Fig. 2B). Enhanced TfR-dependent clearance was observed in both the hemizygous and homozygous hu-man TfR knock-in mice when dosed with the primate-specific anti-TfR1/BACE1 (fig. S1B). This is likely a result of elevated human TfR protein expression across multiple tissues in both the hemizygous and homozygous mice (fig. S1C). We also noted that red blood cell counts and hemoglobin concentrations were altered between the wild-type and human TfR homozyogous mice but remained normal in hemizygous mice (fig. S1D). The hemizygous genotype was thus selected for subsequent dosing studies to best approximate wild-type TfR conditions.

Using therapeutic antibodies that need to cross the blood-brain barrier (BBB) to treat neurological disease is a difficult challenge. We have shown that bispecific antibodies with optimized binding to the transferrin receptor (TfR) that target β-secretase (BACE1) can cross the BBB and reduce brain amyloid-β (Aβ) in mice. Can TfR enhance antibody uptake in the primate brain? We describe two humanized TfR/BACE1 bispecific antibody variants. Using a human TfR knock-in mouse, we observed that anti-TfR/BACE1 antibodies could cross the BBB and reduce brain Aβ in a TfR affinity–dependent fashion. Intravenous dos-ing of monkeys with anti-TfR/BACE1 antibodies also reduced Aβ both in cerebral spinal fluid and in brain tissue, and the degree of reduction correlated with the brain concen-tration of anti-TfR/BACE1 antibody. These results demonstrate that the TfR bispecific antibody platform can robustly and safely deliver therapeutic antibody across the BBB in the primate brain.

Therapeutic bispecific antibodies cross the blood-brain barrier in nonhuman primatesY. Joy Yu,1* Jasvinder K. Atwal,1* Yin Zhang,2 Raymond K. Tong,3 Kristin R. Wildsmith,4 Christine Tan,2 Nga Bien-Ly,1 Maria Hersom,1 Janice A. Maloney,1 William J. Meilandt,1 Daniela Bumbaca,4 Kapil Gadkar,4 Kwame Hoyte,5 Wilman Luk,5 Yanmei Lu,5 James A. Ernst,3 Kimberly Scearce-Levie,1 Jessica A. Couch,4 Mark S. Dennis,2 Ryan J. Watts1†

1Department of Neuroscience, Genentech Inc., 1 DNA Way, South San Francisco, CA 94080, USA. 2Antibody Engineering, Genentech Inc., South San Francisco, CA 94080, USA. 3Protein Chemistry, Genentech Inc., South San Francisco, CA 94080, USA. ⁴Development Sciences, Genentech Inc., South San Francisco, CA 94080, USA. 5Biochemical and Cellular Pharmacology, Genentech Inc., South San Francisco, CA 94080, USA.*These authors contributed equally to this work.†Corresponding author. E-mail: [email protected]

RESEARCH ARTICLE◥

ANTIBIOTICS

















RESEARCH



on

May

11,

201

7ht

tp://

scie

nce.

scie

ncem

ag.o

rg/

Dow

nloa

ded

from

ANTIBODY THERAPEUTICS


mailto:rwatts%40gene.com?subject=

12



whereas primate-specific anti-TfR1 only stained human TfR knock-inmice (Fig. 2A). Brain antibody concentrationsmeasured 1day after dos-ing with anti-TfR/BACE1 (20 mg/kg) demonstrated brain uptake in aspecies-specific manner (Fig. 2B). Enhanced TfR-dependent clearance

was observed in both the hemizygous and homozygous human TfRknock-in mice when dosed with the primate-specific anti-TfR1/BACE1(fig. S1B). This is likely a result of elevated humanTfR protein expressionacross multiple tissues in both the hemizygous and homozygous mice

A

100 101 102 103 104

FL2-H: FL2-height

0

20

40

60

80

100

100 101 102 103 104

FL2-H: FL2-height

0

20

40

60

80

100

UntransfectedHEK293 + cynoTfRHEK293 + huTfRSecondary PE only

No primary + Anti-TfR (parental)

Antibody concentration (nM)

Anti-TfR1

Anti-TfR1 + holoTfAnti-TfRC12

Anti-TfRC12 + holoTf

Bo

un

d a

nti

bo

dy

(OD

650

nm

)

B

10–1 10 0 10 1 10 2 10 3 10 40.0

0.5

1.0

1.5

% o

f max

% o

f max

Antibody concentration (nM)

Control IgGAnti-TfR1

Anti-TfRC12

C

10–2 10–1 100 101 102 1030.0

0.2

0.4

0.6

0.8

1.0

1.2

Bo

un

d h

um

an T

fR E

CD

(O

D 6

50 n

m)

Humanized anti-TfR variant

Human TfR KD

CynoTfR KD

Anti-TfR1/BACE1

Anti-TfR2/BACE1

10 nM

270 nM

37 nM

810 nM

Anti-BACE1Anti-TfR1/BACE1Anti-TfR2/BACE1

Antibody concentration (log10

nM)

D

E F

–2 –1 0 1 2 3 40

50

100

150

200

HuTfRECD

HuTfR apical domain

Bin

din

g (O

D 6

50 n

m)

Anti-TfR1

Anti-TfRC12

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

A x-

40 (p

g/m

l)

Fig. 1. Generation and characterization of primate binding anti-TfR/BACE1 bispecific antibodies. (A) Fluorescence-activated cell sorting (FACS)analysis ofHEK293 cells in the red (FL2-H) fluorescent channel transfectedwithhuman (hu) or monkey (cyno) TfR after treatment with anti-TfR hybridomaantibody supernatant (10 mg/ml). A FACS shift was also detectedwith untrans-fected HEK293 cells (orange), due to anti-TfR binding to endogenous humanTfR on HEK293 cells. PE, R-phycoerythrin. (B) Enzyme-linked immunosorbentassay (ELISA) binding assay showed that binding between anti-TfR1 and TfRwas not affected by human holo (iron-saturated) Tf (1 mg/ml), whereasbinding of TfR to anti-TfRC12, an antibody that targets the Tf-binding region,was blocked by human holo Tf. OD, optical density. (C) Interaction betweenhemochromatosis protein and human TfR was not affected by anti-TfR1 anti-body. (D) Anti-TfR1 antibody binds to the human TfR apical domain. Biotinyl-

ated human TfR extracellular domain (ECD) or TfR–apical domain phage wasadded to MaxiSorp plates coated with anti-TfR1 or anti-TfRC12 antibodies, andboundphagewasdetectedwithhorseradishperoxidase (HRP)–streptavidin orHRP–anti-M13. Data points represent replicates for anti-TfR1 binding to HuTfRECD (circle) or HuTfR apical domain (triangle), and anti-TfRC12 binding toHuTfRECD (square) or HuTfR apical domain (inverted triangle). (E) Top: Schematicshowing anti-TfR1/BACE1 bispecific antibody. Bottom: Affinity of anti-TfR1/BACE1 antibody and anti-TfR2/BACE1 antibody for human and monkey TfRextracellular domain as measured by Biacore T100 using single-cycle kinetics.KD, dissociation constant. (F) Primary mouse cortical neurons show a dose-dependent inhibition of BACE1 after treatment with anti-TfR1/BACE1 or anti-TfR2/BACE1 antibodies as measured by cellular Abx-40 production via ELISA.Values plotted are means ± SEM.

R E S EARCH ART I C L E

www.ScienceTranslationalMedicine.org 5 November 2014 Vol 6 Issue 261 261ra154 2Affinity-dependent brain uptake of anti-TfR/BACE1 variants in human TfR knock-in miceWe next assessed the brain uptake and activity of the anti-TfR/BACE1 bispecific antibody affinity variants in the hemizygous human TfR mice after a single 50 mg/kg intravenous dose.

We observed that the lower-affinity anti-TfR2/BACE1 antibody showed higher peripheral plasma and brain concentrations than did the anti-TfR1/BACE1 antibody (Fig. 2, C and D). Nevertheless, both bispecific antibodies reduced Aβ in the plasma and brain to the same extent at each assessed time point (Fig. 2, E and

F), suggesting that antibody concentrations exceeded the in vivo concentrations needed to drive Aβ reduction. This resulted in maximal BACE1 inhibition: maximum in vivo reduction of Aβ by anti-BACE1 antibody inhibition was ~55 to 65% based on allosteric and cellular mechanisms (11, 12).


13

Research articles


AControl IgG Anti-TfRA Anti-TfR1

KI/

KI

WT

Control IgG

Anti-TfR1/BACE1

Anti-TfR2/BACE1

Bra

in h

uTf

R le

vels

(no

rmal

ized

to

act

in)

Control IgG Anti-TfR1/BACE1 Anti-TfR2/BACE1

huTfR

Actin

Bra

in a

nti

bo

dy

con

cen

trat

ion

(n

M)

Days after dose

Days after dose Days after dose

0

10

20

30

40

50 WT

WT/KI

KI/KI

Bra

in a

nti

bo

dy

con

cen

trat

ion

(n

M)

Control IgG Anti-TfRA/BACE1 Anti-TfR1/BACE1

B

C D

E F

G

0.00

0.05

0.10

0.15

****

n.s.

****

n.s.

****

**

n.s.

****

38

98

Pla

sma

anti

bo

dy

con

cen

trat

ion

(µ

M)

Days after dose

0 2 4 6 80.01

0.1

1

10Control IgGAnti-TfR1/BACE1Anti-TfR2/BACE1

0 2 4 6 80

10

20

30

40

0 2 4 6 80

50

100

150

200

0 2 4 6 80

1000

2000

3000

4000

Pla

sma

A x-

40 (

pg

/ml)

Bra

in A

x-40

(p

g/g

)

Fig. 2. Anti-TfR/BACE1 antibody variants show affinity-dependentbrain uptake in a human TfR knock-in mouse model. (A) Immuno-histochemical localization of injected antibody in wild-type (WT) and humanTfR knock-in mouse brain sections 1 hour after an intravenous injection(5 mg/kg) of control IgG, anti-TfRA (murine-specific), or anti-TfR1 (primate-specific). Vascular localization of anti-TfRA was observed only in WT miceandabsent in thehuTfR knock-inmouse.Conversely, anti-TfR1 stainingwasonlyobserved in the brain vasculature of human TfR knock-in mouse and absent intheWTmouse. Scale bar, 50 mm. (B) Brain concentrations of antibody 24 hoursafter an intravenous injection (20 mg/kg) of control IgG, anti-TfRA/BACE1 anti-body, or anti-TfR1/BACE1 antibody. Means ± SEM, n = 6 per group; **P ≤ 0.01and ****P ≤ 0.0001, compared to control IgG. Statistical tests: one-way analysisof variance (ANOVA) with Dunnett’s multiple comparisons test compared tocontrol IgG–dosedanimals. n.s., not significant. (C toF) Antibody concentrations

in plasma (C) and in brain (D) and Abx-40 concentrations in plasma (E) and brain(F) of human TfR knock-in mice after an intravenous injection (50 mg/kg) ofcontrol IgG, anti-TfR1/BACE1 antibody, or anti-TfR2/BACE1 antibody. Means ±SEM, n = 6 per group. (G) Western blot of human TfR in mouse cortical braintissue1dayafter intravenousdose (50mg/kg)of controlor TfR/BACE1bispecificantibodies (n = 6 per group). Data points represent individual animals dosedwith control IgG (circle), anti-TfR1/BACE1 (square), or anti-TfR2/BACE1 (triangle).Quantification of Western blot bands represents means ± SEM of human TfRnormalized to actin. ****P ≤ 0.0001. Statistical test: one-way ANOVA withDunnett’smultiple comparisons test compared to control IgG–dosed animals.

AQ17


www.ScienceTranslationalMedicine.org 5 November 2014 Vol 6 Issue 261 261ra154 3


KI/

KI

WT

Control IgG

Anti-TfR1/BACE1

Anti-TfR2/BACE1

Bra

in h

uTf

R le

vels

(no

rmal

ized

to

act

in)


huTfR

Actin

Bra

in a

nti

bo

dy

con

cen

trat

ion

(n

M)

Days after dose


0

10

20

30

40

50 WT

WT/KI

KI/KI

Bra

in a

nti

bo

dy

con

cen

trat

ion

(n

M)


B

C D

E F

G

0.00

0.05

0.10

0.15

****

n.s.

****

n.s.

****

**

n.s.

****

38

98

Pla

sma

anti

bo

dy

con

cen

trat

ion

(µ

M)

Days after dose

0 2 4 6 80.01

0.1

1


0 2 4 6 80

10

20

30

40

0 2 4 6 80

50

100

150

200

0 2 4 6 80

1000

2000

3000

4000

Pla

sma

A x-

40 (

pg

/ml)

Bra

in A

x-40

(p

g/g

)



AQ17




KI/

KI

WT

Control IgG

Anti-TfR1/BACE1

Anti-TfR2/BACE1

Bra

in h

uTf

R le

vels

(no

rmal

ized

to

act

in)


huTfR

Actin

Bra

in a

nti

bo

dy

con

cen

trat

ion

(n

M)

Days after dose


0

10

20

30

40

50 WT

WT/KI

KI/KI

Bra

in a

nti

bo

dy

con

cen

trat

ion

(n

M)


B

C D

E F

G

0.00

0.05

0.10

0.15

****

n.s.

****

n.s.

****

**

n.s.

****

38

98

Pla

sma

anti

bo

dy

con

cen

trat

ion

(µ

M)

Days after dose

0 2 4 6 80.01

0.1

1


0 2 4 6 80

10

20

30

40

0 2 4 6 80

50

100

150

200

0 2 4 6 80

1000

2000

3000

4000

Pla

sma

A x-

40 (

pg

/ml)

Bra

in A

x-40

(p

g/g

)



AQ17




KI/

KI

WT

Control IgG

Anti-TfR1/BACE1

Anti-TfR2/BACE1

Bra

in h

uTf

R le

vels

(no

rmal

ized

to

act

in)


huTfR

Actin

Bra

in a

nti

bo

dy

con

cen

trat

ion

(n

M)

Days after dose


0

10

20

30

40

50 WT

WT/KI

KI/KI

Bra

in a

nti

bo

dy

con

cen

trat

ion

(n

M)


B

C D

E F

G

0.00

0.05

0.10

0.15

****

n.s.

****

n.s.

****

**

n.s.

****

38

98

Pla

sma

anti

bo

dy

con

cen

trat

ion

(µ

M)

Days after dose

0 2 4 6 80.01

0.1

1


0 2 4 6 80

10

20

30

40

0 2 4 6 80

50

100

150

200

0 2 4 6 80

1000

2000

3000

4000

Pla

sma

A x-

40 (

pg

/ml)

Bra

in A

x-40

(p

g/g

)



AQ17




14



Investigation of the impact of anti-TfR bi-specific antibody dosing on brain TfR protein levels revealed an affinity-driven degradation of TfR, with only high-affinity anti-TfR1/BACE1 antibody significantly reducing TfR protein levels 1 day after dosing (Fig. 2G). The affin-ity relationship with TfR degradation was re-capitulated in vitro in both human (HEK293) and monkey (COS7) cells, where high-affinity monovalent anti-TfR1 binding triggered TfR degradation (fig. S2, A and B). This is a liabil-ity we have previously observed with high- affinity monovalent mouse TfR antibodies in vivo and limits brain uptake of TfR bispecific antibodies after multiple dosing (13). These data demonstrate that the lower-affinity anti-TfR2/BACE1 has more desirable therapeutic properties when dosed in the human TfR knock-in mouse model.

Brain uptake and activity of anti-TfR/BACE1 antibody variants in nonhuman primates

Key steps in validating TfR as a viable path-way for therapeutic antibody delivery across the BBB include the demonstration of robust brain uptake and the assessment of initial safety risks after administration of TfR bispe-cific antibodies to primates. Cynomolgus mon-keys were dosed intravenously with anti-TfR1/BACE1 antibody, anti-TfR2/BACE1 antibody, or control antibody at 30 mg/kg to assess brain antibody concentrations and BACE1 inhibi-tion. As noted, both anti-TfR1/BACE1 and anti-TfR2/BACE1 bind to human TfR about fourfold tighter than to monkey TfR (Fig. 1E), likely because of species sequence differences in the TfR apical domain. Peripheral serum concen-trations of anti-TfR/BACE1 bispecific antibod-ies correlated with TfR binding affinity, with higher-affinity anti-TfR1/BACE1 antibody ex-hibiting faster clearance than anti-TfR2/BACE1 antibody (Fig. 3A). Both bispecific antibodies reduced peripheral plasma Aβ concentrations by greater than 50% (Fig. 3B). Strikingly, the anti-TfR1/BACE1 antibody showed a robust and sustained reduction in Aβ concentrations in cerebrospinal fluid (CSF) (Fig. 3C), a result that was corroborated by an equally robust reduction in the sAPPβ (soluble APPβ)/sAPPα ratio in CSF (Fig. 3D and fig. S3, A and B), fur-ther verifying inhibition of BACE1.

We have previously shown in mouse that anti-TfR/BACE1 antibody substantially im-proves brain uptake and subsequent Aβ re-duction relative to anti-BACE1 antibody (7). We thus conducted a similar comparison in nonhuman primates by measuring brain con-centrations of the anti-TfR/BACE1 antibody variants, anti-BACE1 antibody, and control antibody from monkey brains harvested 1 day after a single 30 mg/kg intravenous dose. Di-rectly correlating with Aβ and sAPPβ/sAPPα reductions in CSF (Fig. 3, C and D), dosing with anti-TfR1/BACE1 antibody led to the high-est concentration of bispecific antibody in all

monkey brain regions assayed (Fig. 3E). Fur-thermore, reductions in brain Aβ concentra-tions also correlated with reductions in CSF Aβ concentrations (Fig. 3F); however, the extent of Aβ reduction was more modest in brain relative to CSF, consistent with what has been reported (14). Anti-TfR2/BACE1 an-tibody showed moderate brain uptake cor-relating with moderate reductions in CSF Aβ and sAPPβ/sAPPα ratio (compare Fig. 3E with Fig. 3, C and D). The reduced affinity of anti-TfR2 antibody for monkey TfR, relative to human TfR, explained the poorer brain uptake and relatively weaker CSF/brain PD effect seen in monkey compared to the hu-man TfR knock-in mice. Compared to control antibody, anti-BACE1 antibody showed low brain uptake and no reduction in brain Aβ (Fig. 3, E and F). The PK (pharmacokinetic)/PD relationship shown in monkeys with anti-TfR/BACE1 illustrates the substantial im-provement in brain uptake with the TfR bi-specific antibody, providing evidence that TfR is a robust receptor-mediated pathway in the BBB of primates.

No reticulocyte loss with anti-TfR/BACE1 in monkeys

In previous work, we reported that TfR- positive circulating reticulocytes in mice pre-sented a major safety liability for anti-TfR antibodies, posing a significant concern for chronic dosing paradigms using an anti-TfR antibody platform (7). Attenuating antibody effector function abolished acute clinical signs, although residual reticulocyte loss was still observed. Although primates have sub-stantially fewer TfR-positive circulating retic-ulocytes in comparison to rodents, a critical assessment of the impact of our newly gener-ated anti-TfR bispecific antibodies on the he-matopoietic system in primates is warranted, with a particular focus on reticulocytes.

Time course analysis of reticulocytes, red blood cells, immature reticulocytes (which ex-press the highest levels of TfR), hemoglobin, serum iron (Fe2+), and total iron-binding ca-pacity (TIBC) for monkeys dosed intravenous-ly with anti-TfR1/BACE1 or anti-TfR2/BACE1 (30 mg/kg) revealed no sign of reticulocyte loss (Fig. 4, A and C), impact on the hemato-poietic system (Fig. 4, B and D), or significant changes in serum iron concentrations (Fig. 4E) or TIBC (an indirect measure of circulat-ing Tf concentrations) (Fig. 4F). The lack of impact on hematological and iron-related blood parameters further confirmed the non-blocking nature of these primate-specific anti-TfR bispecific antibodies in vivo (Fig. 1,B and C). Notably, both anti-TfR/BACE1 bispecific antibodies tested in these studies lacked Fcg receptor binding; thus, the possibility remains that anti-TfR antibodies with normal effector function could deplete TfR-positive cells that reside in bone marrow, as was previously shown in mice (7).

As we had seen robust TfR degradation with the high-affinity anti-TfR1/BACE1 an-tibody in the human TfR knock-in mice, we also assayed TfR protein levels in monkey brain after anti-TfR/BACE1 antibody dosing. Western blot analysis of brain TfR protein 1 day after dosing in monkeys did not show changes in TfR protein levels with either anti-TfR/BACE1 bispecific antibody (fig. S2C).

DISCUSSION

Utilization of molecular “lifts” to cross the BBB has been experimentally investigated for several decades (1–3); however, success in translating these findings to humans re-mains elusive. One of the most studied BBB receptors, TfR, has shown promise in rodents but has not been assayed in primates. Fur-thermore, safety findings in mice that raise concerns about using TfR to enable large therapeutics such as antibodies to cross the BBB have recently been described (7). To determine if TfR is a viable target for drug delivery in humans, we generated several primate binding anti-TfR/BACE1 bispecific antibody variants. We tested these anti-TfR/BACE1 antibodies both in a newly generated human TfR knock-in mouse model and in monkeys. The data we obtained, including no acute safety observations or reticulocyte loss in monkeys, suggest that TfR is indeed a vi-able target for drug delivery across the BBB in nonhuman primates and is worthy of further drug development.

Comparison of these data from the primate-specific anti-TfR/BACE1 antibodies in human TfR knock-in mice and monkeys revealed an optimal anti-TfR affinity that is neither too high nor too low. These results add critical in-sight into previous studies using mouse-spe-cific anti-TfR antibodies (6, 7, 13), thus identi-fying a cross-species principle that is likely to be intrinsic to the biology of the TfR pathway at the BBB. We propose a model (Fig. 5) in which brain uptake of anti-TfR–based anti-body therapeutics is based on identifying the optimal affinity of the antibody for TfR. The data used to generate this model are based on single-arm (monovalent) binding to TfR, be-cause previous work with bivalent antibodies only reflects the high-affinity context based on avidity from both antibody arms binding to TfR (6). We have shown that high-affinity monovalent binding to TfR drives rapid an-tibody clearance and TfR degradation, thus limiting sustained brain exposure to anti-body. In contrast, very low affinity binding to TfR results in substantially better periph-eral blood antibody exposure as a result of reduced target-mediated clearance; however, maximal brain uptake is also reduced because of the extremely low affinity of antibody for TfR. Ultimately, the identification of an opti-mal affinity for TfR leads to sustained brain concentrations of antibody above a therapeu-tic threshold that is dependent on the potency


15

Research articles


A

C

E

Days after dose

D

B

F

0 5 10 150.1

1

10

100

1000

0 5 10 150

50

100

150


–5 0 5 10 150

50

100

150

–5 0 5 10 150.0

0.5

1.0

1.5

0 1 2 3

StriatumCortex

ThalamusHippocampus

MidbrainCerebellum

% Injected dose/kg brain

StriatumCortex

ThalamusHippocampus

MidbrainCerebellum

StriatumCortex

ThalamusHippocampus

MidbrainCerebellum

StriatumCortex

ThalamusHippocampus

MidbrainCerebellum

*

****

**

**

***

0 20 40 60–40 –20 80

CSF

(% o

f bas

elin

e)B

rain

(rel

ativ

eto

co

ntr

ol I

gG

)

Days after dose

Control IgGAnti-TfR1/BACE1Anti-TfR2/BACE1

Control IgGAnti-TfR1/BACE1Anti-TfR2/BACE1Anti-BACE1


Pl

asm

a A

x-40

(% o

f bas

elin

e)

CSF

A x-

40 (%

of b

asel

ine)

% A x-40

reduction

Fig. 3. Brainuptakeandactivity of anti-TfR/BACE1antibody variants innonhuman primates. (A to D) Serum antibody concentrations (A), plasmaAbx-40 concentrations (B), CSF Abx-40 concentrations (C), and sAPPb/sAPParatio in CSF (D) were measured in cynomolgus monkeys after 30 mg/kg in-travenous dosing. Means ± SEM, n = 5 per group. (E) Brain antibody uptakeas calculated by percent of injected dose per kilogram of tissue in differentbrain structures 24 hours after a 30 mg/kg dose. Because antibody uptakeacross different brain regions was comparable, we calculated the averageconcentration across these regions for each animal and then determineda groupmean to represent overall brain exposure for the different antibodies.

The mean brain antibody concentrations (±SEM, in nM) were as follows: con-trol IgG, 0.61 ± 0.16; anti-TfR1/BACE1, 9.39 ± 0.50; anti-TfR2/BACE1, 2.55 ± 0.44;and anti-BACE1, 0.20 ± 0.02. (F) CSF and brain (cortex) Abx-40 concentrationsmeasured 24 hours after a 30 mg/kg dose. Mean percent reduction in Abx-40was calculated relative to baseline (CSF) or control IgG (brain). Data pointsrepresent individual animals dosed with control IgG (circle), anti-TfR1/BACE1(square), anti-TfR2/BACE1 (triangle), or anti-BACE1 (inverted triangle). Statis-tical tests: one-way ANOVA (E and F) with Bonferroni multiple comparisonstest compared to control IgG–dosed animals. *P < 0.05, ***P < 0.001, and****P < 0.0001.



A

C

E

Days after dose

D

B

F

0 5 10 150.1

1

10

100

1000

0 5 10 150

50

100

150


–5 0 5 10 150

50

100

150

–5 0 5 10 150.0

0.5

1.0

1.5

0 1 2 3

StriatumCortex

ThalamusHippocampus

MidbrainCerebellum

% Injected dose/kg brain

StriatumCortex

ThalamusHippocampus

MidbrainCerebellum

StriatumCortex

ThalamusHippocampus

MidbrainCerebellum

StriatumCortex

ThalamusHippocampus

MidbrainCerebellum

*

****

**

**

***

0 20 40 60–40 –20 80

CSF

(% o

f bas

elin

e)B

rain

(rel

ativ

eto

co

ntr

ol I

gG

)

Days after dose




Pl

asm

a A

x-40

(% o

f bas

elin

e)

CSF

A x-

40 (%

of b

asel

ine)

% A x-40

reduction

Fig. 3. Brainuptakeandactivity of anti-TfR/BACE1antibody variants innonhuman primates. (A to D) Serum antibody concentrations (A), plasmaAbx-40 concentrations (B), CSF Abx-40 concentrations (C), and sAPPb/sAPParatio in CSF (D) were measured in cynomolgus monkeys after 30 mg/kg in-travenous dosing. Means ± SEM, n = 5 per group. (E) Brain antibody uptakeas calculated by percent of injected dose per kilogram of tissue in differentbrain structures 24 hours after a 30 mg/kg dose. Because antibody uptakeacross different brain regions was comparable, we calculated the averageconcentration across these regions for each animal and then determineda groupmean to represent overall brain exposure for the different antibodies.

The mean brain antibody concentrations (±SEM, in nM) were as follows: con-trol IgG, 0.61 ± 0.16; anti-TfR1/BACE1, 9.39 ± 0.50; anti-TfR2/BACE1, 2.55 ± 0.44;and anti-BACE1, 0.20 ± 0.02. (F) CSF and brain (cortex) Abx-40 concentrationsmeasured 24 hours after a 30 mg/kg dose. Mean percent reduction in Abx-40was calculated relative to baseline (CSF) or control IgG (brain). Data pointsrepresent individual animals dosed with control IgG (circle), anti-TfR1/BACE1(square), anti-TfR2/BACE1 (triangle), or anti-BACE1 (inverted triangle). Statis-tical tests: one-way ANOVA (E and F) with Bonferroni multiple comparisonstest compared to control IgG–dosed animals. *P < 0.05, ***P < 0.001, and****P < 0.0001.




16



these findings to humans remains elusive. One of themost studied BBBreceptors, TfR, has shown promise in rodents but has not been assayedin primates. Furthermore, safety findings in mice that raise concernsabout using TfR to enable large therapeutics such as antibodies tocross the BBB have recently been described (7). To determine if TfRis a viable target for drug delivery in humans, we generated severalprimate binding anti-TfR/BACE1 bispecific antibody variants. Wetested these anti-TfR/BACE1 antibodies both in a newly generated hu-man TfR knock-inmousemodel and inmonkeys. The data we obtained,including no acute safety observations or reticulocyte loss in monkeys,

suggest that TfR is indeed a viable target for drug delivery across theBBB in nonhuman primates and is worthy of further drug development.

Comparison of these data from the primate-specific anti-TfR/BACE1antibodies in human TfR knock-in mice and monkeys revealed an op-timal anti-TfR affinity that is neither too high nor too low. These resultsadd critical insight into previous studies using mouse-specific anti-TfR antibodies (6, 7, 13), thus identifying a cross-species principle thatis likely to be intrinsic to the biology of the TfR pathway at the BBB.We propose a model (Fig. 5) in which brain uptake of anti-TfR–basedantibody therapeutics is based on identifying the optimal affinity of the

–10 0 10 20 300

50

100

150

Days after dose

A

–10 0 10 20 300

2

4

6

8

Days after dose

B

C

Hem

og

lob

in (g

/dl)

D

–10 0 10 20 300

10

20

30

–10 0 10 20 300

5

10

15


100

200

300

–10 0 10 20 300

Days after dose

–10 0 10 20 30

200

400

600

Days after dose

0

E F

Reti

culo

cyte

s (1

09 /lit

er)


Imm

atu

re re

ticu

locy

te

frac

tio

n (%

)

Fig. 4. Safety analysis in monkeys after anti-TfR/BACE1 antibody dos-ing. Blood was collected for hematology from all animals (n = 5 pergroup) at 11 and 7 days before dosing and at 1 hour and 1, 7, 14, and30 days after dosing. (A to F) Reticulocyte counts (A), red blood cell counts

(B), immature reticulocyte fraction (C), hemoglobin concentration (D), iron(Fe2+) concentration (E), and TIBC (F) were measured after dosing with30 mg/kg of either one of the anti-TfR/BACE1 bispecific antibodies. Plottedvalues are means ± SEM.


www.ScienceTranslationalMedicine.org 5 November 2014 Vol 6 Issue 261 261ra154 6of the therapeutic arm (for example, BACE1). The exact optimal affinity may vary from species to species and from central nervous system target to target.

A key difference in TfR biology between rodents and primates was functionally test-ed in this study, namely, the determination of reticulocyte loss after dosing with anti-TfR antibody. We have shown previously that circulating reticulocytes in mice have substantially higher TfR protein expression than those of primates (7). The finding that neither anti-TfR/BACE1 antibody variant caused reticulocyte loss in monkeys con-

firms this difference functionally (Fig. 4) and raises an important consideration: that TfR biology is not completely conserved be-tween species. We propose that one reason moderate-affinity anti-TfR1/BACE1 antibody performs better in monkeys relative to the human TfR knock-in mouse is because of the about fourfold shift in affinity between antibody binding to human and monkey TfR. However, an equally important con-sideration is that the differences in biology between species may also lead to signifi-cantly different outcomes (for example, re-ticulocyte loss). These differences may in-

clude TfR expression, BBB physiology, and/or antibody kinetics. As such, it is important to note that whereas rodent models may be very useful in identifying and providing the initial validation of BBB targets, the devel-opment of clinical candidates will require extensive investigation in nonhuman pri-mates, including repeat-dosing studies to establish long-term safety and activity. In conclusion, these data illustrate the prom-ise of targeting TfR as a potentially safe and robust target to boost antibody uptake in primate brain, laying the foundation for future testing in humans.

SECTION | RESEARCH ARTICLES


17

Research articles


antibody for TfR. The data used to generate this model are based onsingle-arm (monovalent) binding to TfR, because previous work withbivalent antibodies only reflects the high-affinity context based on avid-ity from both antibody arms binding to TfR (6). We have shown thathigh-affinity monovalent binding to TfR drives rapid antibody clear-ance and TfR degradation, thus limiting sustained brain exposure toantibody. In contrast, very low affinity binding to TfR results in subs-tantially better peripheral blood antibody exposure as a result of reducedtarget-mediated clearance; however, maximal brain uptake is also re-duced because of the extremely low affinity of antibody for TfR. Ulti-mately, the identification of an optimal affinity for TfR leads tosustained brain concentrations of antibody above a therapeutic thresh-old that is dependent on the potency of the therapeutic arm (for exam-ple, BACE1). The exact optimal affinitymay vary from species to speciesand from central nervous system target to target.

A key difference in TfR biology between rodents and primates wasfunctionally tested in this study, namely, the determination of reticulo-cyte loss after dosing with anti-TfR antibody.We have shown previous-ly that circulating reticulocytes in mice have substantially higher TfRprotein expression than those of primates (7). The finding that neitheranti-TfR/BACE1 antibody variant caused reticulocyte loss in monkeysconfirms this difference functionally (Fig. 4) and raises an importantconsideration: that TfR biology is not completely conserved between

species. We propose that one reasonmoderate-affinity anti-TfR1/BACE1 anti-body performs better in monkeys relativeto the human TfR knock-in mouse is be-cause of the about fourfold shift in affinitybetween antibody binding to human andmonkey TfR. However, an equally impor-tant consideration is that the differencesin biology between species may also leadto significantly different outcomes (for ex-ample, reticulocyte loss). These differ-ences may include TfR expression, BBBphysiology, and/or antibody kinetics. Assuch, it is important to note that whereasrodentmodels may be very useful in iden-tifying and providing the initial validationof BBB targets, the development of clinicalcandidates will require extensive investi-gation in nonhuman primates, includingrepeat-dosing studies to establish long-term safety and activity. In conclusion,these data illustrate the promise of target-ing TfR as a potentially safe and robusttarget to boost antibody uptake in primatebrain, laying the foundation for futuretesting in humans.

MATERIALS AND METHODS

Study designThe purpose of this study was to generateprimate binding TfR bispecific antibodiesand determine if they can cross theBBB innonhuman primates as a means to deliver

antibody therapeutics to the central nervous system. Human/monkeycross-reactive anti-TfR bispecific antibodies were also assessed in humanTfR knock-in mice. For in vivo studies, sample sizes of six C57Bl/6 miceor five cynomolgus monkeys per experimental group were used to ac-count for biological variability among animals and for reliable statisticalanalysis (detailed methods provided below). For in vitro studies usingHEK293 andCHO (Chinese hamster ovary) cell lines, three independentexperiments were performed, with technical triplicates per condition.In vivo data collection occurred for a predetermined period of timebased on previous TfR antibody PK studies. Data were collected blindlyand/or via assay automation. All data were included, and no animalswere excluded from the study.

Human/monkey TfR antibody generationand characterization

Antibody generation. BALB/c mice were coimmunized with hu-man and cyno TfR ECD. Mice demonstrating high serum titers by TfRELISA and binding to HEK293 cells transiently expressing human ormonkey TfR by FACS were fused using X63.Ag8.653 mouse myelomacells. After fusion, hybridoma supernatants were again screened byELISA and FACS. Clones that bound both human and cyno TfR in thepresence of 1 mM human holo-Tf (R&D, 2914-HT) were selected forfurther characterization by surface plasmon resonance (SPR) analysis.

Bra

in a

nti

bo

dy

exp

osu

re

TfR affinity

Low affinity

-Reduced max uptake

-No TfR degradation

-Slow clearance

Optimal affinity

-Sustained uptake & exposure

-Minimal TfR degradation

-Moderate clearance

Target concentration(therapeutic potency)

High affinity

-Reduced exposure

-TfR degradation

-Rapid clearance

Fig. 5. Relationship between TfR affinity and sustained antibody uptake by brain tissue. Data acrossmultiple species and acrossmultiple TfR antibody affinities suggested amodel relating TfR affinity (x axis) tobrain antibody exposure (y axis). Monovalent anti-TfR bispecific antibodies with high affinity (red) show re-duced sustained exposure in both brain and periphery as a result of rapid clearance and TfR degradation.Low-affinity binding to TfR (blue) results in reduced maximal brain uptake; however, TfR protein levels arenot altered, and both systemic and brain antibody concentrations are sustained as a result of slow clearance.An optimal affinity for TfR (green) shows sustained presence of antibody in brain above therapeutic targetconcentrations (threshold depends on the potency of the therapeutic arm, dashed line). Minimal TfR deg-radation and moderate anti-TfR antibody clearance are observed with an optimal TfR affinity. The exactoptimized TfR affinity depends on the species/model being investigated and possibly other properties ofthe selected TfR antibody (for example, TfR expression and anti-TfR antibody epitopes).




Study design

The purpose of this study was to generate primate binding TfR bispecific antibodies and determine if they can cross the BBB in nonhuman primates as a means to deliver antibody therapeutics to the central nervous system. Human/monkey cross-reactive anti-TfR bispecific antibodies were also assessed in human TfR knock-in mice. For in vivo studies, sample sizes of six C57Bl/6 mice or five cynomolgus monkeys per experimental group were used to account for biological variability among animals and for reliable statistical analysis (detailed methods provided below). For in vitro studies using HEK293 and CHO (Chinese hamster ovary) cell lines, three independent experiments were performed, with technical triplicates per condition. In vivo data collection occurred for a predetermined period of time based on previous TfR antibody PK studies. Data were col lected blindly and/or via assay automation. All data were included, and no animals were excluded from the study.

Human/monkey TfR antibody generation and characterization

Antibody generation. BALB/c mice were coimmunized with human and cyno TfR ECD. Mice demonstrating high serum titers by TfR ELISA and binding to HEK293 cells tran-siently expressing human or monkey TfR by FACS were fused using X63.Ag8.653 mouse myeloma cells. After fusion, hybridoma su-pernatants were again screened by ELISA and FACS. Clones that bound both human and cyno TfR in the presence of 1 μM human holo-Tf (R&D, 2914-HT) were selected for further characterization by surface plasmon resonance (SPR) analysis. A clone that had similar affinity to both human and cyno TfR was humanized to generate anti-TfR1. Anti-TfRC12 was derived from a synthetic antibody phage library that was panned against hu-man TfR ECD. Binding of anti-TfRC12 to TfR was blocked by human holo-Tf. Human iso-type control IgG was anti-human glycopro-tein D antibody.

Affinity measurements. SPR analysis was performed using a Biacore T100 in HBS-P buffer. Biotinylated human or cyno

TfR ECD was captured on a CM5 chip ran-domly coupled with streptavidin. Affinities of serially diluted TfR1/BACE1 or TfR2/BACE1 were determined using single-cycle kinetics.

Tf competition. MaxiSorp plates were coated with human TfR ECD (2 μg/ml) in phosphate-buffered saline (PBS). Human holo-Tf (1 mg/ml) or PBS was added to the plate for 40 min. Antibody was then added to the wells containing holo-Tf or PBS. Bound antibody was detected with HRP–Goat anti-Human Fcg antibody (Jackson ImmunoRe-search, 109-036-098).

HFE competition. A mixture of biotinyl-ated human TfR ECD (2 μg/ml) and serially diluted antibody was added to MaxiSorp plates coated with HFE protein (1 μg/ml) (Novus Biologicals, H00003077-Q01) in PBS for 1 hour. Bound biotinylated human TfR ECD was detected with HRP-streptavidin (GE Healthcare, RPN 4401V).

TfR epitope evaluation. Immobilized anti-TfR1 and anti-TfRC12 were evaluated for binding to biotinylated human TfR ECD or monovalent M13 phage displaying the human TfR apical domain. Antibodies were coated at 1 μg/ml in PBS on MaxiSorp


18



plates. Bound biotinylated human TfR ECD or TfR apical domain phage was detected with HRP-streptavidin (GE Healthcare, RPN 4401V) or HRP–anti-M13 (GE Healthcare, 27-9421-01), respectively.

Bispecific antibody generation. Anti-TfR1, anti-TfR2, and anti-BACE1 were used to generate bispecific antibodies using the knob-in-hole bispecific antibody con-struction technology (6, 9). In addition to the knob and hole mutations in the Fc for anti-TfR (hole) and anti-BACE1 (knob), both half-antibodies contain mutations in the Fc region that abrogated effector function (N297G). The knob and hole half-antibodies were expressed and purified separately from CHO, and the bispecific antibody was assem-bled by reductive annealing. After assembly, bispecific antibodies were purified by hydro-phobic interaction chromatography. Homo-geneity and purity of the final antibodies were confirmed by liquid chromatography–mass spectroscopy.

Cellular Aβx-40 assays

Aβx-40

production was measured in dissoci-ated cortical neuronal cultures prepared from E16.5 C57BL/6J mice. Neurons were plated at 2.4 × 104 cells per well in a 96-well plate and grown for 5 days in vitro. Fresh medium containing anti-BACE1 or anti-TfR/BACE1 was incubated with the neurons for 24 hours. Medium was harvested and assayed for Aβ

x-40 using a sensitive mouse Aβ

x-40 ELISA (see

below for details). Aβx-40

values were normal-ized for cell viability, as determined using the CellTiter-Glo Luminescent Cell Viability As-say (Promega). Experiments were performed at least three times, and each point in each experiment was repeated in duplicate. Data were plotted using a four-parameter nonlin-ear regression curve using GraphPad Prism.

Generation of human TfR knock-in mouse

The construct for targeting human TFRC cDNA into the C57BL/6 Tfrc locus in embry-onic stem (ES) cells was made using a com-bination of recombineering (15, 16) and stan-dard molecular cloning techniques. Briefly, a cassette (human TFRC cDNA, SV40 pA, and frt-PGK-em7-Neo-BGHpA-frt) flanked by short homologies to the mouse Tfrc gene was used to modify a Tfrc C57BL/6J bacterial arti-ficial chromosome (BAC) (RP23 BAC library) by recombineering. The human TFRC cDNA cassette was inserted at the endogenous ATG, and the remainder of Tfrc exon 2 plus the be-ginning of intron 2 were deleted. The targeted region in the BAC was then retrieved into pBlight-TK (15) along with flanking genomic Tfrc sequences as homology arms for ES cell targeting. Specifically, the 2950–base pair (bp) 5′ homology arm corresponds to (assembly NCBI37/mm9) chr.16:32,610,333-32,613,282,

and the 2599-bp 3′ homology arm corre-sponds to chr.16:32,613,320-32,615,918. The fi-nal vector was confirmed by DNA sequencing. The Tfrc/TFRC KI vector was linearized with Not I, and C57BL/6N C2 ES cells were tar-geted using standard methods (G418-positive and gancyclovir-negative selection). Positive clones were identified using polymerase chain reaction (PCR) and TaqMan analysis and con-firmed by sequencing of the modified locus. Correctly targeted ES cells were transfected with an Flpe plasmid to remove Neo, and ES cells were then injected into blastocysts using standard techniques. Germline transmission was obtained after crossing resulting chime-ras with C57BL/6N females.

Measuring antibody concentrations and mouse Aβx-40 in brain and plasma

The animals’ care was in accordance with Ge-nentech Institutional Animal Care and Use Committee (IACUC) guidelines. For wild-type mice, female C57BL/6 mice ages 6 to 8 weeks were used. For huTfR KI/WT and KI/KI mice, about 8-week-old males and females were used. Mice were intravenously injected with antibody and taken down at the indicated time after injection. Before perfusion with PBS, whole blood was collected in plasma Microtainer tubes (BD Diagnostics) and spun down at 14,000 rpm for 2 min. Plasma super-natant was isolated for antibody and mouse Aβ

x-40 measurements. Brains were extracted,

and one hemi-brain was homogenized in 1% NP-40 (Calbiochem) in PBS containing cOm-plete Mini, EDTA-free protease inhibitor cock-tail tablets (Roche Diagnostics). Homogenized brain samples were rotated at 4°C for 1 hour before spinning at 14,000 rpm for 20 min. Supernatant was isolated for brain antibody measurement. For brain Aβ

x-40 measurements,

the other hemi-brain was homogenized in 5 M guanidine hydrochloride buffer, and samples were rotated for 3 hours at room temperature before diluting (1:10) in 0.25% casein and 5 mM EDTA (pH 8.0) in PBS containing freshly added aprotinin (20 mg/ml) and leupeptin (10 mg/ml). Diluted homogenates were spun at 14,000 rpm for 20 min, and supernatants were isolated for mouse Aβ

x-40 measurements.

PK assays. Antibody concentrations in mouse serum and brain samples were mea-sured using an ELISA. Nunc 384-well Maxi-Sorp immunoplates were coated with F(ab’)2 fragment of donkey anti-human IgG, Fc frag-ment–specific polyclonal antibody (Jackson ImmunoResearch). After blocking the plates, each antibody was used as a standard to quantify the respective antibody concentra-tions. Standards and samples were incubated on plates for 2 hours at room temperature with mild agitation. Bound antibody was de-tected with HRP-conjugated F(ab’)2 goat anti-human IgG, Fc-specific polyclonal antibody (Jackson ImmunoResearch). Concentrations were determined from the standard curve

using a four-parameter nonlinear regression program. The assay had lower limit of quan-titation (LLOQ) values of 3.12 ng/ml in serum and 1.56 ng/ml in brain.

PD assays. Aβx-40

concentrations in mouse neuronal culture supernatants, plasma, and brain samples were measured using an ELI-SA similar to methods for PK analysis above. Briefly, rabbit polyclonal antibody specific for the C terminus of Aβ

40 (Millipore) was coated

onto plates, and biotinylated anti-mouse Aβ monoclonal antibody M3.2 (Covance) was used for detection. The assay had LLOQ val-ues of 1.96 pg/ml in plasma and 39.1 pg/g in brain.

Western blot analysis

Mouse and monkey brain tissue was isolated and homogenized in 1% NP-40 with protease inhibitors as described above. About 20 μg was loaded onto 4 to 12% Bis-Tris Novex gels (Life Technologies). Gels were transferred onto ni-trocellulose membranes using the iBlot sys-tem (Life Technologies), and Western blotting was performed using Odyssey blocking buffer reagents and secondary antibodies (LI-COR). Human and mouse cross-reactive mouse anti-TfR (Life Technologies, clone H68.4; 1:1000) was used to detect total TfR, whereas human TfR was detected by goat anti-human TfR (R&D, AF2474; 1:2000). Rabbit anti–β-actin (Abcam, 8227; 1:2000) served as a loading control. Western membranes were imaged and quantified using manufacturer-supplied software and system (Odyssey/LI-COR).

Immunohistochemistry

Wild-type and human TfR knock-in mice were intravenously injected with control IgG, anti-TfRA, or anti-TfR1 (5 mg/kg), followed by PBS perfusion 1 hour after dosing. Brains were drop-fixed in 4% paraformaldehyde overnight at 4°C, followed by 30% sucrose overnight at 4°C. Brain tissue were sectioned at 35-μm thickness on a sliding microtome, blocked for 1 to 3 hours in 5% bovine serum albumin (BSA) 0.3% Triton, and incubated with 1:200 Alexa Fluor 488 anti-human secondary anti-body (Life Technologies) in 1% BSA 0.3% Tri-ton for 2 hours at room temperature. Mount-ed slides were subsequently analyzed by Leica fluorescence microscopy.

Cynomolgus monkey PK/PD studies

Animals (male cynomolgus macaque aged 3 to 5 years; five animals per group) were surgically prepared with indwelling cannulae inserted into the cisterna magna and connected to a subcutaneous access port to permit CSF sampling. Control IgG (anti-gD, targeting glycoprotein D of herpes simplex virus), anti-TfR1/BACE1, or anti-TfR2/BACE1 was delivered at 30 mg/kg via an intravenous bolus injection into the


19

Research articles


saphenous vein. CSF and blood samples were collected at various time points from 11 days before dosing up to 60 days post-dose. Samples were collected at the same time of the day.

A follow-up study to measure brain antibody uptake and Aβ levels was performed using surgically cannulated animals. This study included both naïve animals and animals dosed in the above PK/PD (after minimum 100-day drug washout). Animals were not re-administered the same treatment they received in the PK/PD study. Control IgG, anti-TfR1/BACE1, anti-TfR2/BACE1, or anti-BACE1 was delivered at 30 mg/kg via an intravenous bolus injection into the saphenous vein. Twenty-four hours post-dose, brains were perfused and harvested. Brain regions were subdissected and immediately frozen. Brain samples were prepared for PK and Aβ

x-40 PD ELISAs using

protocols identical to those for mouse tissue described above.

All animal protocols were approved by the IACUC of Genentech and were conducted in accordance with the National Institutes of Health’s Guide for the Care and Use of Labora-tory Animals.

PK assays. Total antibody concentrations in monkey serum, CSF, and brain samples were measured using monkey-adsorbed sheep anti-human IgG polyclonal antibody (Bind-ing Site) as coat and a monkey-adsorbed goat anti-human IgG antibody conjugated to HRP (Bethyl) as detection. The assay had an LLOQ value of 39.1 ng/ml in serum or CSF and 3.1 ng/ml in brain.

PD assays. The concentrations of total cyno Aβ

x-40 in plasma were determined

using an Aβ electroluminescence (ECL) assay, and the concentrations of total cyno Aβ

x-40 in CSF and brain were determined

using a sandwich ELISA. In both cases, the capture antibody, specific for the C terminus of Aβ

x-40 (Millipore), was precoated

on the plates, and the anti-Aβ monoclonal antibody 6E10 (Covance) was used for detection. The assays had LLOQ values of 60 pg/ml in plasma, 100 pg/ml in CSF, and 50 pg/ml in brain. CSF concentrations of sAPPα and sAPPβ were determined with a sAPPα/sAPPβ multiplex ECL assay. The anti-Aβ monoclonal antibody 6E10 was used to capture sAPPα, whereas an antibody directed against amino acids 591 to 596 of APP was used to capture sAPPβ. Both analytes were detected with an antibody directed against the N terminus of APP. CSF was thawed on ice and then diluted 1:10 into 1% BSA in tris-buffered saline–Tween 20. The assay had LLOQ values of 0.05 and 0.03 ng/ml for sAPPα and sAPPβ, respectively.

Hematology and blood chemistry assays in cynomolgus monkeys and human TfR knock-in miceFor evaluation of red-cell indices in cynomolgus monkeys, about 0.5 ml of blood was collected into K2 EDTA–containing tubes from all animals at 11 and 7 days before dosing and at 1 hour and 1, 7, 14, and 30 days post-dose. The collected blood was assayed for standard hematological parameters within 2 hours after collection using the Siemens ADVIA 120. For evaluation of clinical chemistry parameters including Fe2+ and TIBC, about 2 ml of blood was collected after an overnight fast from all animals at 11 and 7 days before dosing and at 7, 14, and 30 days post-dose. The collected blood was separated into serum by centrifugation at 4000 rpm for 10 min at 4°C within 1 hour after collection. The serum samples were then assayed for standard clinical chemistry parameters using the Roche COBAS INTEGRA 400 plus.

For evaluation of red-cell indices in huTfR KI mice, about 0.1 ml of blood was collected into K2 EDTA–containing tubes from all animals and assayed for standard hematological parameters within 2 hours of collection using the Sysmex XT-2000iV (Sysmex).

Statistical analysisAll values are expressed as means ± SEM, un-less otherwise indicated, and P values were assessed by ordinary one-way ANOVA, with Dunnett’s multiple comparisons test. Corre-lation analysis between brain TfR and anti-body concentrations was performed using Graphpad Prism version 6.

REFERENCES AND NOTES 1. A. R. Jones, E. V. Shusta, Pharm. Res. 24, 1759–1771 (2007). 2. J. Lichota et al., J. Neurochem. 113, 1–13 (2010). 3. Y. J. Yu, R. J. Watts, Neurotherapeutics 10, 459–472 (2013). 4. R. D. Bell, M. D. Ehlers, Neuron 81, 1–3 (2014). 5. R. J. Watts, M. S. Dennis, Curr. Opin. Chem. Biol. 17, 393–399 (2013). 6. Y. J. Yu et al., Sci. Transl. Med. 3, 84ra44 (2011). 7. J. A. Couch et al., Sci. Transl. Med. 5, 183ra57 (2013). 8. J. Niewoehner et al., Neuron 81, 49–60 (2014). 9. J. B. Ridgway, L. G. Presta, P. Carter, Protein Eng. 9, 617–621 (1996).10. R. Yan, R. Vassar, Lancet Neurol. 13, 319–329 (2014).11. J. K. Atwal et al., Sci. Transl. Med. 3, 84ra43 (2011).

12. L. Zhou et al., J. Biol. Chem. 286, 8677–8687 (2011).13. N. Bien-Ly et al., J. Exp. Med. 211, 233–244 (2014).14. Y. Lu et al., J. Pharmacol. Exp. Ther. 342, 366–375 (2012). 15. S. Warming, R. A. Rachel, N. A. Jenkins, N. G. Copeland,, Mol. Cell. Biol. 26, 6913–6922 (2006).16. P. Liu, N. A. Jenkins, N. G. Copeland, Genome Res. 13, 476–484 (2003).

ACKNOWLEDGMENTSWe thank Y. Chen and X. Chen for generation of anti-TfRC12; M. Mathieu for constructing the TfR apical domain phage; S. Warming, M. Roose-Girma, and B. Newman for design and generation of human TfR knock-in mice; H. Solanoy for in vivo assistance; K. Stark and A. Oldendrop for study implementation; and S. Schauer and K. Peng for sample processing. We also thank our colleagues at Transgenic Technology core laboratories for technical assistance. Author contributions: R.J.W., Y.J.Y., and J.K.A. designed the project. R.J.W., Y.J.Y., J.K.A., K.R.W., J.A.M., W.J. M., K.G., D.B., K.S.-L., J.A.E., J.A.C., and M.S.D. designed, performed, oversaw, and analyzed various in vivo experiments. M.S.D., Y.Z., and C.T. generated and subsequently engineered the antibodies and performed in vitro binding experiments. R.K.T. and J.A.E. purified and refolded the bispecific antibodies and antigens. K.H., W.L., and Y.L. ran and analyzed PK/PD assays. N.B.-L., M.H., and J.A.M. designed, conducted, and analyzed various in vitro experiments. R.J.W., Y.J.Y., and J.K.A. wrote the manuscript with extensive comments from M.S.D. Competing interests: All authors are paid employees of Genentech Inc. Genentech has filed patents on the subject matter of this paper.

Submitted 18 June 2014 Accepted 9 October 2014 Published 5 November 2014 10.1126/scitranslmed.3009835

Citation: Y. J. Yu, J. K. Atwal, Y. Zhang, R. K. Tong, K. R. Wildsmith, C. Tan, N. Bien-Ly, M. Hersom, J. A. Maloney, W. J. Meilandt, D. Bumbaca, K. Gadkar, K. Hoyte, W. Luk, Y. Lu, J. A. Ernst, K. Scearce-Levie, J. A. Couch, M. S. Dennis, R. J. Watts, Therapeutic bispecific antibodies cross the blood-brain barrier in nonhuman primates. Sci. Transl. Med. 6, 261ra154 (2014).

SUPPLEMENTARY MATERIALSwww.sciencetranslationalmedicine.org/cgi/content/full/6/261/261ra154/DC1Materials and MethodsFig. S1. Generation and characterization of a human TfR knock-in mouse model.Fig. S2. In vitro and in vivo characterization of TfR degradation related to species and affinity.Fig. S3. PD analysis in monkey after anti-TfR/BACE1 antibody dosing.


http://www.sciencetranslationalmedicine.org/cgi/content/full/6/261/261ra154/DC1

20



INTRODUCTION

Receptor tyrosine kinases (RTKs) and their respective ligands play important regulatory roles throughout develop-ment and stimulate many critical cellular processes. Aberrant signaling by these

molecules is implicated in various diseases, especially cancer. Very few RTKs in mammals have no known ligand and are considered “or-phan” RTKs. One such receptor is anaplastic lymphoma kinase (ALK), which is present almost exclusively in the nervous system, pri-marily during embryonic development. More-over, the gene encoding ALK is mutated or overexpressed in several cancers (1–3). Genetic fusions that result in the joining the kinase do-main of ALK with various other proteins (for example, EML4-ALK and NMP-ALK) are the main drivers of a subset of non–small cell lung carcinoma and anaplastic large cell lymphoma cases (2), whereas acquired somatic or inher-ited germline mutations in ALK, or ALK over-expression, are implicated in neuroblastoma and anaplastic thyroid cancer (1, 3).

Pleiotrophin (PTN) and Midkine (MK),

which are two related heparin-binding ex-tracellular molecules, were reported to be physiological ligands for ALK (4, 5). However, subsequent studies were unable to confirm these results (6–10). Ligand interactions with RTKs induce autophosphorylation, receptor endocytosis, and activation of downstream signaling such as signaling pathways involv-ing the kinases AKT and extracellular signal–regulated kinase 1/2 (ERK 1/2) (11). To date, no ligand has been identified for ALK that meets these criteria (6–10).

The extracellular domain (ECD) of mam-malian ALK consists of two MAM (meprin/A5/protein tyrosine phosphatase Mu) do-mains, which flank an LDL-A (low-density li-poprotein class A) domain and are N-terminal to a glycine-rich region and a putative epider-mal growth factor (EGF)–like domain (11, 12). The ECD is unique among RTKs, sharing high sequence similarity only with leukocyte tyro-sine kinase (LTK) in the glycine-rich region and EGF domain. Like ALK, LTK is an orphan RTK (2, 11). There is a conserved, highly ba-sic 249–amino acid N-terminal region (NTR) in mature vertebrate ALKs. The NTR of ALK has no reported function and is not conserved among invertebrate ALKs or any other pro-tein. Moreover, ligands that bind invertebrate ALKs (Jeb in Drosophila melanogaster and HEN1 in Caenorhabditis elegans) are not con-served in mammals (2).

Here, we found that heparin is a specific, high-affinity ligand that directly binds the

NTR of ALK. Short heparin chains bound monovalently to ALK ECD and antagonized ALK activation; longer heparin chains in-duced ALK dimerization, activation, and downstream signaling in cultured neuroblas-toma cells. We developed monoclonal anti-bodies (mAbs) that bound to the NTR of ALK, inhibited heparin binding, and prevented heparin-induced ALK activation. Thus, hepa-rin and perhaps other sulfated proteoglycans can function as ALK ligands or co-ligands, suggesting a mechanism similar to binding and activation of fibroblast growth factor re-ceptors (FGFRs) by heparin and FGF (11).

RESULTS

Heparin activates ALK

The heparin-binding molecules PTN and MK are putative ligands for ALK. However, there are discrepancies among various reports re-garding this role for PTN and MK (4–10). To investigate the ability of PTN and MK to activate ALK, we tested whether PTN or MK could activate endogenous ALK in cultured neuroblastoma cells (NB1) in the presence or absence of heparin. Similar to other RTKs, ALK autophosphorylation at several tyrosine residues can be used to monitor activation using antibodies against phosphorylated ty-rosine or phosphorylation-specific antibodies targeting ALK (11). Exposing cells to mixtures of heparin and PTN or heparin and MK stimu-lated robust autophosphorylation of ALK (Fig. 1A). However, heparin alone was sufficient to induce ALK autophosphorylation to a similar degree as heparin mixed with PTN or heparin mixed with MK (Fig. 1A), indicating that PTN and MK were entirely dispensable for ALK activation. Moreover, heparin-induced ALK autophosphorylation and phosphorylation of downstream signaling proteins AKT and ERK 1/2 were concentration-dependent (Fig. 1B and fig. S1). Similar to ligand-mediated stimu-lation of other RTKs (11), heparin stimulation of NB1 cells induced ALK internalization (Fig. 1C and fig. S1).

We developed mAbs directed against the ECD of ALK and tested their ability to acti-vate or inhibit ALK. The mAb αALK1 stimu-lated phosphorylation of ALK and ERK 1/2 (Fig. 1A) and internalization of ALK (Fig. 1C and fig. S1) in NB1 cells. In contrast, a differ-ent mAb, αALK2, inhibited the ability of hep-arin to induce phosphorylation of ALK and ERK 1/2 (Fig. 1A).

Heparin is a specific, high-affinity ligand for ALK

The above results suggested that heparin is a ligand or co-ligand for ALK, similar to the role of heparin as a co-ligand for all four members of the FGFR family of RTKs (11, 13, 14). Moreover, RPTPσ (receptor protein tyrosine phosphatase σ), which also

Heparin is an activating ligand of the orphan receptor tyrosine kinase ALKPhillip B. Murray,1 Irit Lax,1 Andrey Reshetnyak,1 Gwenda F. Ligon,2 Jay S. Lillquist,2 Edward J. Natoli Jr.,2 Xiarong Shi,1 Ewa Folta-Stogniew,3 Murat Gunel,4,5 Diego Alvarado,2 Joseph Schlessinger1*

Anaplastic lymphoma kinase (ALK) is one of the few remaining “orphan” receptor tyrosine kinases (RTKs) in which the ligands are unknown. Ligand-mediated activation of RTKs is important throughout development. ALK is particularly relevant to the development of the nervous system. Increased activation of RTKs by mutation, genetic amplification, or signals from the stroma contributes to disease progression and acquired drug resistance in cancer. Aberrant activation of ALK occurs in subsets of lung adenocarcinoma, neuro-blastoma, and other cancers. We found that heparin is a ligand that binds specifically to the ALK extracellular domain. Whereas heparins with short chain lengths bound to ALK in a monovalent manner and did not activate the receptor, longer heparin chains induced ALK dimerization and activation in cultured neuroblastoma cells. Heparin lacking N- and O-linked sulfate groups or other glycosaminoglycans with sulfation patterns different than heparin failed to activate ALK. Moreover, antibodies that bound to the extracellular domain of ALK interfered with heparin binding and prevented heparin-mediated activa-tion of ALK. Thus, heparin and perhaps related glycosaminoglycans function as ligands for ALK, revealing a potential mechanism for the regulation of ALK activity in vivo and suggesting an approach for developing ALK-targeted therapies for cancer.

1Department of Pharmacology, Yale University School of Medicine, New Haven, CT 06520, USA. 2Kolltan Pharmaceuticals, New Haven, CT 06520, USA. 3Keck Foundation Biotechnology Resource Laboratory, Yale University, New Haven, CT 06520, USA. 4Department of Neurosurgery, Yale University School of Medicine, New Haven, CT 06520, USA. 5Department of Genetics, Yale University School of Medicine, New Haven, CT 06520, USA.*Corresponding author. E-mail: [email protected]

Originally published 20 January 2015 in SCIENCE SIGNALING

RESEARCH ARTICLE◥

ANTIBIOTICS

















RESEARCH



on

May

11,

201

7ht

tp://

scie

nce.

scie

ncem

ag.o

rg/

Dow

nloa

ded

from

CELL BIOLOGY


mailto:joseph.schlessinger%40yale.edu?subject=

21

Research articles

SCIENCE sciencemag.org Originally published 20 January 2015 in SCIENCE SIGNALING

plays a role in the development of the nervous system, binds directly to glycosaminoglycans (15). Therefore, we investigated the ability of heparin to bind directly and specifically to the ALK ECD.

Analysis of the NTR of ALK revealed a putative heparin-binding motif (Fig. 2A), similar to the motif found in FGFRs that is critical for the ability of heparin to act as a co-ligand for FGFR activation (13, 14). Deletion of the NTR eliminated the ability of heparin to promote autophosphorylation of ALK in transiently transfected human embryonic kidney (HEK) 293 cells (fig. S2), indicating that the NTR is required for ALK activation by heparin.

We characterized the biophysical proper-ties of heparin binding to mammalian ALK in vitro. We used the ECD from Canis famil-iaris (dog ALK), which shares 91.2% identity to the human ECD (16). Notably, full-length ALK ECD (FL-ECD), but not a truncation mutant of the ALK ECD lacking the NTR (ΔN-ECD), could be purified using hepa-rin-Sepharose chromatography (Fig. 2B).

Mutating the basic residues in the putative heparin-binding motif of FL-ECD to acidic residues also inhibited binding to the hep-arin-Sepharose beads (Fig. 2B). Thus, these data suggest that heparin directly binds to the NTR domain of ALK.

To measure the affinity and specific-ity of binding between heparin and ALK, we used surface plasmon resonance (SPR). We immobilized biotinylated heparin to a NeutrAvidin surface and tested binding with various concentrations of FL-ECD and ΔN-ECD. The FL-ECD bound with high af-finity (K

D = 151 nM) (Fig. 3A), similar to the

range of KDs of heparin binding to FGFs and

FGFRs (14). In contrast, ΔN-ECD bound too weakly for a K

D to be estimated (Fig. 3B).

Moreover, the disaccharide heparin mi-metic sucrose octasulfate (SOS) inhibited binding of FL-ECD to immobilized heparin at a median inhibitory concentration (IC

50)

of 6.5 μM and an inhibition constant (Ki) of

2.25 μM (Fig. 3, C and D), indicating a high degree of specificity for the interaction be-tween heparin and ALK.

Heparin chain length correlates with affinity for ALK and ALK oligomerizationWe hypothesized that the stoichiometry of the heparin-ALK interaction would depend on heparin chain length, with increasing chain length resulting in greater degrees of ALK oligomerization. The experiments used to determine binding affinity by SPR used heparin with heterogenous chain lengths. To test if there was a correlation between hepa-rin chain length and ALK-binding affinity and oligomerization state, we used purified heparins with various degrees of polymeriza-tion (dp), where 1 dp is equal to 1 heparin di-saccharide unit. However, it is important to note that these purified heparins were pools of different lengths defined by an average chain length. We used isothermal titration calorimetry (ITC) to quantify the interaction between FL-ECD and heparins with various average chain lengths. For heparin with an average chain length of dp8–9, dp15, or dp25, the molar ratio and affinity of FL-ECD for heparin indicated approximately monovalent (K

D = 505 nM), bivalent (K

D = 200 nM), and

tetravalent or pentavalent (KD = 80 nM) bind-

ing, respectively (Fig. 4, A to C), suggesting that the enhanced affinity of ALK oligomer-ized on heparin of longer chain lengths is likely due to avidity associated with this interaction.

We used the molecular weight of the ALK-heparin complexes in solution to relate the molar ratio of heparin-ALK binding to ALK stoichiometry by subjecting FL-ECD and FL-ECD–heparin complexes to size exclusion chromatography coupled to a multiangle laser light scattering (SEC-MALLS) detector. In the absence of heparin or in the presence of heparins with an average chain length ≤dp10, FL-ECD eluted from the column with an apparent molecular weight equivalent to monomers (Fig. 4D and fig. S3). In contrast, in the presence of heparin dp15, a large proportion of FL-ECD eluted as an apparent dimer (Fig. 4D). In the presence of heparin dp25, FL-ECD eluted across a broad range of apparent molecular weights greater than that of monomeric or dimeric FL-ECD (Fig. 4D), suggesting a broad range of higher-order ALK oligomerization likely due, at least in part, to the heterogeneity of chain lengths in the average dp25 heparin pool. Unlike FL-ECD, ΔN-ECD did not elute at a different apparent molecular weight in the presence of heparin dp25 (Fig. 4E), consistent with the observations that ΔN-ECD did not bind to heparin-Sepharose or to the heparin-coated SPR surface (Fig. 2B).

Heparins of specific chain length and sulfation patterns activate ALK

To relate the stoichiometry of ALK-heparin binding to physiological activation of the receptor, we stimulated NB1 cells with different concentrations of heparin with different

similar degree as heparin mixed with PTN or heparin mixed with MK(Fig. 1A), indicating that PTN and MK were entirely dispensable forALK activation. Moreover, heparin-induced ALK autophosphorylationand phosphorylation of downstream signaling proteins AKT and ERK1/2 were concentration-dependent (Fig. 1B and fig. S1). Similar to ligand-mediated stimulation of other RTKs (11), heparin stimulation of NB1 cellsinduced ALK internalization (Fig. 1C and fig. S1).

We developed mAbs directed against the ECD of ALK and tested theirability to activate or inhibit ALK. The mAb aALK1 stimulated phospho-rylation of ALK and ERK 1/2 (Fig. 1A) and internalization of ALK (Fig.1C and fig. S1) in NB1 cells. In contrast, a different mAb, aALK2,inhibited the ability of heparin to induce phosphorylation of ALK andERK 1/2 (Fig. 1A).

Heparin is a specific, high-affinity ligand for ALKThe above results suggested that heparin is a ligand or co-ligand for ALK,similar to the role of heparin as a co-ligand for all four members of theFGFR family of RTKs (11, 13, 14). Moreover, RPTPs (receptor proteintyrosine phosphatase s), which also plays a role in the development of thenervous system, binds directly to glycosaminoglycans (15). Therefore, we

investigated the ability of heparin to bind directly and specifically to theALK ECD.

Analysis of the NTR of ALK revealed a putative heparin-binding motif(Fig. 2A), similar to the motif found in FGFRs that is critical for the abilityof heparin to act as a co-ligand for FGFR activation (13, 14). Deletion of theNTR eliminated the ability of heparin to promote autophosphorylation ofALK in transiently transfected human embryonic kidney (HEK) 293 cells(fig. S2), indicating that the NTR is required for ALK activation by heparin.

We characterized the biophysical properties of heparin binding tomammalian ALK in vitro. We used the ECD from Canis familiaris (dogALK), which shares 91.2% identity to the human ECD (16). Notably,full-length ALK ECD (FL-ECD), but not a truncation mutant of theALK ECD lacking the NTR (DN-ECD), could be purified using heparin-Sepharose chromatography (Fig. 2B). Mutating the basic residues in the pu-tative heparin-binding motif of FL-ECD to acidic residues also inhibitedbinding to the heparin-Sepharose beads (Fig. 2B). Thus, these data suggestthat heparin directly binds to the NTR domain of ALK.

To measure the affinity and specificity of binding between heparinand ALK, we used surface plasmon resonance (SPR). We immobilizedbiotinylated heparin to a NeutrAvidin surface and tested binding withvarious concentrations of FL-ECD and DN-ECD. The FL-ECD boundwith high affinity (KD = 151 nM) (Fig. 3A), similar to the range ofKDs of heparin binding to FGFs and FGFRs (14). In contrast, DN-ECD bound too weakly for a KD to be estimated (Fig. 3B). Moreover,the disaccharide heparin mimetic sucrose octasulfate (SOS) inhibitedbinding of FL-ECD to immobilized heparin at a median inhibitory con-centration (IC50) of 6.5 mM and an inhibition constant (Ki) of 2.25 mM(Fig. 3, C and D), indicating a high degree of specificity for the interactionbetween heparin and ALK.

Heparin chain length correlates with affinity for ALKand ALK oligomerizationWe hypothesized that the stoichiometry of the heparin-ALK interactionwould depend on heparin chain length, with increasing chain lengthresulting in greater degrees of ALK oligomerization. The experimentsused to determine binding affinity by SPR used heparin with heterog-enous chain lengths. To test if there was a correlation between heparinchain length and ALK-binding affinity and oligomerization state, weused purified heparins with various degrees of polymerization (dp),where 1 dp is equal to 1 heparin disaccharide unit. However, it is im-portant to note that these purified heparins were pools of differentlengths defined by an average chain length. We used isothermal titra-tion calorimetry (ITC) to quantify the interaction between FL-ECD andheparins with various average chain lengths. For heparin with an averagechain length of dp8–9, dp15, or dp25, the molar ratio and affinity ofFL-ECD for heparin indicated approximately monovalent (KD = 505 nM),bivalent (KD = 200 nM), and tetravalent or pentavalent (KD = 80 nM)binding, respectively (Fig. 4, A to C), suggesting that the enhanced affinityof ALK oligomerized on heparin of longer chain lengths is likely due toavidity associated with this interaction.

We used the molecular weight of the ALK-heparin complexes in solu-tion to relate the molar ratio of heparin-ALK binding to ALK stoichiometryby subjecting FL-ECD and FL-ECD–heparin complexes to size exclusionchromatography coupled to a multiangle laser light scattering (SEC-MALLS)detector. In the absence of heparin or in the presence of heparins with anaverage chain length ≤dp10, FL-ECD eluted from the column with an ap-parent molecular weight equivalent to monomers (Fig. 4D and fig. S3). Incontrast, in the presence of heparin dp15, a large proportion of FL-ECDeluted as an apparent dimer (Fig. 4D). In the presence of heparin dp25, FL-ECD eluted across a broad range of apparent molecular weights greater than

Fig. 1. Heparin activates ALK inNB1 cells. (A and B) Westernblot for Tyr1604 phosphorylated

ALK (pALK), Ser473 phosphorylated AKT (pAKT), and Thr202 or Thr204 phos-phorylated ERK1 and ERK2 (pERK 1/2) using lysates from NB1 cells. In(A), cells were exposed to PTN (100 nM), MK (100 nM), heparin (HEP,10 mg/ml), aALK2 (10 nM), or aALK1 (10 nM) alone or in the indicatedcombinations for 10 min. In (B), cells were exposed to heparin at dif-ferent concentrations (100 to 0.01 mg/ml) for 10 min. MW, molecularweight. (C) Western blot for endogenous ALK using isolated cell surface–biotinylated proteins or lysates of NB1 cells exposed to heparin (10 mg/ml) oraALK1 (10 nM) for 6 hours. For (A) to (C), “-” indicates unstimulated nega-tive control cells, and the blots are representative of at least two inde-pendent experiments.

R E S E A R C H A R T I C L E

www.SCIENCESIGNALING.org 20 January 2015 Vol 8 Issue 360 ra6 2

on May 11, 2017

http://stke.sciencemag.org/

Dow

nloaded from


22



that of monomeric or dimeric FL-ECD (Fig.4D), suggesting a broad range of higher-order ALK oligomerization likely due, atleast in part, to the heterogeneity of chainlengths in the average dp25 heparin pool.Unlike FL-ECD, DN-ECD did not elute ata different apparent molecular weight in thepresence of heparin dp25 (Fig. 4E),consistent with the observations that DN-ECD did not bind to heparin-Sepharose orto the heparin-coated SPR surface (Fig. 2B).

Heparins of specific chain lengthand sulfation patterns activate ALKTo relate the stoichiometry of ALK-heparinbinding to physiological activation of the re-ceptor, we stimulated NB1 cells with differ-ent concentrations of heparin with differentchain lengths and measured ALK autophos-phorylation by enzyme-linked immunosor-bent assay (ELISA). Heparin with averagechain lengths greater than dp15, but not lessthan dp15, induced autophosphorylation ofALK (Fig. 5A), suggesting that ALK di-merization and higher-order oligomerizationpromote receptor activation.

We tested whether SOS, which is equiv-alent to heparin with a chain length of dp1and thus should not induce dimerization,inhibited heparin-induced activation of ALKin NB1 cells. We found that SOS reducedthe ability of mixed chain length heparin to promote autophosphorylationof ALK and phosphorylation of AKT and ERK 1/2 in a concentration-dependent manner (Fig. 5B).

Heparin is an experimental proxy for physiological ligands used to studythe effects of glycosaminoglycans on signaling by RTKs (2, 13, 14, 17).Glycosaminoglycans other than heparin are likely to be the physiologicallyrelevant ALK ligand or co-ligand in vivo. Moreover, the sulfation pattern of

glycosaminoglycans can influence binding and activation state of receptors,as observed with RPTPs (14). Therefore, we tested whether exposing NB1cells to differentially sulfated heparin or other glycosaminoglycans with dif-ferent sulfation patterns activated ALK. We found that heparin, oversulfatedheparin, and dextran sulfate induced autophosphorylation of ALK, whereaschondroitin sulfate, heparan sulfate, and various other specifically desulfatedheparins did not (Fig. 5, C and D).

Fig. 2. Heparin directly binds the NTR in the ECD of ALK. (A) Amino acidalignment of the NTR of ALK (amino acids 44 to 69) with the heparin-binding motif of FGFR. (B) Coomassie-stained protein gel of elutionfractions from heparin-Sepharose chromatography. Top gel: FL-ECD

and DN-ECD. Bottom gel: FL-ECD with mutations in the putative heparin-binding motif (mutation set 1: FL-ECDR48E,R51E,K52E or mutation set 2:FL-ECDR65E,R69E). Gels are representative of two independentexperiments.

Fig. 3. Heparin binds to ALK with high affinity and specificity. (A to C) SPR sensograms where a titration ofFL-ECD or DN-ECD or a single concentration of FL-ECD in the presence of different concentrations of SOS(10 mM to 26 nM) are injected over a heparin-coated surface. RU, response units. Graphs are traces fromthe same surface and representative of three independent experiments. The average binding affinity (KD)of FL-ECD for the heparin-coated surface across three experiments was calculated using a steady-statemodel. (D) Graph of inhibition of FL-ECD binding to the heparin surface by SOS calculated from three in-dependent experiments as shown in (C). Data are means ± SD. The Ki was calculated to be 2.25 ± 0.70 mM.



on May 11, 2017


Dow

nloaded from

that of monomeric or dimeric FL-ECD (Fig.4D), suggesting a broad range of higher-order ALK oligomerization likely due, atleast in part, to the heterogeneity of chainlengths in the average dp25 heparin pool.Unlike FL-ECD, DN-ECD did not elute ata different apparent molecular weight in thepresence of heparin dp25 (Fig. 4E),consistent with the observations that DN-ECD did not bind to heparin-Sepharose orto the heparin-coated SPR surface (Fig. 2B).










on May 11, 2017


Dow

nloaded from that of monomeric or dimeric FL-ECD (Fig.

4D), suggesting a broad range of higher-order ALK oligomerization likely due, atleast in part, to the heterogeneity of chainlengths in the average dp25 heparin pool.Unlike FL-ECD, DN-ECD did not elute ata different apparent molecular weight in thepresence of heparin dp25 (Fig. 4E),consistent with the observations that DN-ECD did not bind to heparin-Sepharose orto the heparin-coated SPR surface (Fig. 2B).










on May 11, 2017


Dow

nloaded from

chain lengths and measured ALK autophosphorylation by enzyme-linked immunosorbent assay (ELISA). Heparin with average chain lengths greater than dp15, but not less than dp15, induced autophosphorylation of ALK (Fig. 5A),

suggesting that ALK dimerization and higher-order oligomerization promote receptor activation.

We tested whether SOS, which is equivalent to heparin with a chain length of dp1 and thus should not induce dimerization, inhibited

heparin-induced activation of ALK in NB1 cells. We found that SOS reduced the ability of mixed chain length heparin to promote autophosphorylation of ALK and phosphorylation of AKT and ERK 1/2 in a concentration-dependent manner (Fig. 5B).


23

Research articles


An antibody targeting the ECD of ALK inhibits heparinbinding and heparin-induced activation of ALKmAbs that directly inhibit ligand binding or receptor dimerization are usefulfor treatments of various clinical indications. ALK plays a role in neuroblas-toma pathogenesis (1, 2). Therefore, we tested several mAbs targeting eitherthe N-terminal or C-terminal half of the human ALK ECD for the ability toinhibit heparin binding using an SPR competition assay. The mAb aALK3decreased the SPR refractive index (Fig. 6A), indicating that aALK3disrupted the interaction of FL-ECD with the heparin-coated surface(increased Koff for FL-ECD). In contrast, seven other mAbs increasedthe SPR refractive index (Fig. 6A), indicating that these mAbs did notcompete with FL-ECD for heparin binding and presumably bound at anindependent site on FL-ECD, increasing the total mass on the heparin-coated surface. In the absence of FL-ECD, aALK3 and six of the otherseven mAbs did not bind to the heparin-coated surface (fig. S4), indicatingvery little nonspecific direct binding to the surface.

Immunoblot analysis indicated that aALK3 bound to the NTR of ALK(fig. S5), suggesting that aALK3 directly competes with heparin at theheparin-binding site and should prevent activation of ALK by heparin.Thus, we tested whether aALK3 inhibited the ability of heparin to induce

autophosphorylation of ALK. The whole immunoglobulin G (IgG) aALK3mAb stimulated ALK autophosphorylation in NB1 cells (fig. S6), sug-gesting that the antibody caused receptor dimerization, which is an estab-lished mechanism for antibodies that bind RTKs and act as agonists (11).Thus, we generated a monovalent Fab fragment derived from aALK3(aALK3-Fab). aALK3-Fab inhibited heparin-induced autophosphorylationof ALK with an estimated IC50 <100 nM (Fig. 6B). Likewise, aALK3-Fabinhibited heparin-induced phosphorylation of AKT and ERK 1/2 (Fig. 6B).

DISCUSSION

We found that heparin bound to ALK and induced ALK tyrosine auto-phosphorylation, suggesting a role for sulfated glycosaminoglycans andassociated proteoglycans as physiological ligands for ALK. One possiblemechanism of ALK activation is that sulfated carbohydrate moieties of aproteoglycan bind to the positively charged region in the NTR of ALKand the protein core of the same proteoglycan binds to another regionin ALK. Another potential mechanism is that the sulfated carbohydratemoieties of the proteoglycan could act together with a second, yet to be iden-tified, protein ligand to stimulate ALK dimerization and activation similar

Fig. 4. ALK-binding affinity and oligomerization depends on heparin chainlength. (A to C) Results of ITC for heparin with the indicated average chainlength titrated into a solution of FL-ECD. Top: Heat evolved from injection.Bottom: Integration of the heat released in top panels as a function of hep-arin to FL-ECD in the cell. Titration curves are representative of twoindependent experiments. Molar ratio is the average ratio of heparin toFL-ECD at which heat evolved is half maximal. KD is the average bindingaffinity of FL-ECD for heparin. (D and E) Results of SEC-MALLS for FL-ECD

(D) or DN-ECD (E) mixed with heparin with the indicated average chainlength. Solid lines represent the normalized refractive index in arbitraryunits (AU) of FL-ECD or DN-ECD alone or as higher-order complexes.Dotted lines represent MWs of each species measured using MALLS.The measured MWs correspond to oligomeric FL-ECD species formationupon mixing with heparin. No complex formation is observed when heparinis mixed with DN-ECD. Graphs are representative traces from two inde-pendent experiments.



on May 11, 2017


Dow

nloaded from

Heparin is an experimental proxy for physiological ligands used to study the effects of glycosaminoglycans on signaling by RTKs (2, 13, 14, 17). Glycosaminoglycans other than heparin are likely to be the physiologically relevant ALK ligand or co-ligand in vivo. Moreover, the sulfation pattern of glycosaminoglycans can influence binding and activation state of receptors, as observed with RPTPσ (14). Therefore, we tested whether exposing NB1 cells to differentially sulfated heparin or other glycosaminoglycans with different sulfation patterns activated ALK. We found that heparin, oversulfated heparin, and dextran sulfate induced autophosphorylation of ALK, whereas chondroitin sulfate, heparan sulfate, and various other specifically desulfated heparins did not (Fig. 5, C and D).

An antibody targeting the ECD of ALK inhibits heparin binding and heparin-induced activation of ALK

mAbs that directly inhibit ligand binding or receptor dimerization are useful for treat-ments of various clinical indications. ALK plays a role in neuroblastoma pathogenesis (1, 2). Therefore, we tested several mAbs tar-geting either the N-terminal or C-terminal half of the human ALK ECD for the ability to inhibit heparin binding using an SPR compe-tition assay. The mAb αALK3 decreased the SPR refractive index (Fig. 6A), indicating that αALK3 disrupted the interaction of FL-ECD with the heparin-coated surface (increased K

off for FL-ECD). In contrast, seven other

mAbs increased the SPR refractive index (Fig. 6A), indicating that these mAbs did not com-pete with FL-ECD for heparin binding and

presumably bound at an independent site on FL-ECD, increasing the total mass on the heparin-coated surface. In the absence of FL-ECD, αALK3 and six of the other seven mAbs did not bind to the heparin-coated surface (fig. S4), indicating very little nonspe-cific direct binding to the surface.

Immunoblot analysis indicated that αALK3 bound to the NTR of ALK (fig. S5), suggesting that αALK3 directly competes with heparin at the heparin-binding site and should prevent activation of ALK by heparin. Thus, we tested whether αALK3 inhibited the ability of hepa-rin to induce autophosphorylation of ALK. The whole immunoglobulin G (IgG) αALK3 mAb stimulated ALK autophosphorylation in NB1 cells (fig. S6), suggesting that the an-tibody caused receptor dimerization, which is an established mechanism for antibodies that bind RTKs and act as agonists (11). Thus, we

An antibody targeting the ECD of ALK inhibits heparinbinding and heparin-induced activation of ALKmAbs that directly inhibit ligand binding or receptor dimerization are usefulfor treatments of various clinical indications. ALK plays a role in neuroblas-toma pathogenesis (1, 2). Therefore, we tested several mAbs targeting eitherthe N-terminal or C-terminal half of the human ALK ECD for the ability toinhibit heparin binding using an SPR competition assay. The mAb aALK3decreased the SPR refractive index (Fig. 6A), indicating that aALK3disrupted the interaction of FL-ECD with the heparin-coated surface(increased Koff for FL-ECD). In contrast, seven other mAbs increasedthe SPR refractive index (Fig. 6A), indicating that these mAbs did notcompete with FL-ECD for heparin binding and presumably bound at anindependent site on FL-ECD, increasing the total mass on the heparin-coated surface. In the absence of FL-ECD, aALK3 and six of the otherseven mAbs did not bind to the heparin-coated surface (fig. S4), indicatingvery little nonspecific direct binding to the surface.

Immunoblot analysis indicated that aALK3 bound to the NTR of ALK(fig. S5), suggesting that aALK3 directly competes with heparin at theheparin-binding site and should prevent activation of ALK by heparin.Thus, we tested whether aALK3 inhibited the ability of heparin to induce

autophosphorylation of ALK. The whole immunoglobulin G (IgG) aALK3mAb stimulated ALK autophosphorylation in NB1 cells (fig. S6), sug-gesting that the antibody caused receptor dimerization, which is an estab-lished mechanism for antibodies that bind RTKs and act as agonists (11).Thus, we generated a monovalent Fab fragment derived from aALK3(aALK3-Fab). aALK3-Fab inhibited heparin-induced autophosphorylationof ALK with an estimated IC50 <100 nM (Fig. 6B). Likewise, aALK3-Fabinhibited heparin-induced phosphorylation of AKT and ERK 1/2 (Fig. 6B).

DISCUSSION

We found that heparin bound to ALK and induced ALK tyrosine auto-phosphorylation, suggesting a role for sulfated glycosaminoglycans andassociated proteoglycans as physiological ligands for ALK. One possiblemechanism of ALK activation is that sulfated carbohydrate moieties of aproteoglycan bind to the positively charged region in the NTR of ALKand the protein core of the same proteoglycan binds to another regionin ALK. Another potential mechanism is that the sulfated carbohydratemoieties of the proteoglycan could act together with a second, yet to be iden-tified, protein ligand to stimulate ALK dimerization and activation similar

Fig. 4. ALK-binding affinity and oligomerization depends on heparin chainlength. (A to C) Results of ITC for heparin with the indicated average chainlength titrated into a solution of FL-ECD. Top: Heat evolved from injection.Bottom: Integration of the heat released in top panels as a function of hep-arin to FL-ECD in the cell. Titration curves are representative of twoindependent experiments. Molar ratio is the average ratio of heparin toFL-ECD at which heat evolved is half maximal. KD is the average bindingaffinity of FL-ECD for heparin. (D and E) Results of SEC-MALLS for FL-ECD

(D) or DN-ECD (E) mixed with heparin with the indicated average chainlength. Solid lines represent the normalized refractive index in arbitraryunits (AU) of FL-ECD or DN-ECD alone or as higher-order complexes.Dotted lines represent MWs of each species measured using MALLS.The measured MWs correspond to oligomeric FL-ECD species formationupon mixing with heparin. No complex formation is observed when heparinis mixed with DN-ECD. Graphs are representative traces from two inde-pendent experiments.



on May 11, 2017


Dow

nloaded from


24



to the paradigm established for heparin and FGF stimulation of FGFR (11).LTK and ALK share homology in the glycine-rich region and EGF-likerepeat. The glycine-rich region and EGF-like repeat comprise the entire

ECD of LTK, suggesting that these regions likely serve as the binding sitefor an unidentified ligand. Thus, although ALK and LTK may shareidentical or related ligands, ALK is likely activated by two or more co-ligands:a sulfated glycosaminoglycan-linked proteoglycan that binds the NTR and aligand that binds to the glycine-rich and EGF-like repeat of the ECD.

The identification of heparin as a ligand or co-ligand provides insightand tools for investigating ALK biology. Little is known about the func-tion of mammalian ALK. In Drosophila, constitutive activation of ALKby Jeb is essential for maintaining normal brain metabolism during star-vation (18). In addition to its role in neurobiology, ALK signaling has beenimplicated in driving several types of cancer (1–3). Similar to other RTKs,ligand-mediated increases in ALK signaling may contribute to tumori-genesis. Thus, the discovery of heparin as a bona fide ALK ligand providesmechanistic understanding of ALK activation and potential methods toinhibit oncogenic ALK by targeting its ECD.


Cloning, expression, and purification of FL-ECD,DN-ECD, and heparin-binding motif mutationsThe nucleotide sequence coding for amino acids 1 to 1037 of dog ALK(encoding FL-ECD) (16) was synthesized by Blue Heron. An 8X-His tagwas added to the 3′ end, followed by a stop codon. An Xho I site wasadded upstream of the start codon, and an Xba I site was added directlyafter the stop codon. The construct was subcloned into pcDNA3.3. DN-ECDwas produced by polymerase chain reaction–mediated deletion (19) of aminoacids 21 to 263 using FL-ECD as the template. The mutations in the heparin-binding motif were produced using site-directed mutagenesis using theQuikChange II Site-Directed Mutagenesis protocol developed by Stratagene(catalog no. 200523). All plasmids were verified by sequencing of the com-plete open reading frame.

To produce recombinant His-tagged FL-ECD and DN-ECD, plasmidswere transiently transfected into HEK293-S cells grown in Dulbecco’smodified Eagle’s medium–F12 with 10% fetal bovine serum (FBS) and1% penicillin and streptomycin using the Lipofectamine 2000 as per themanufacturer’s instructions (Invitrogen). The culture medium was switched

Fig. 5. Heparins of specific chain length and sulfation patterns activateALK. (A) Graphs represent the autophosphorylation of ALK asmeasured by ELISA in lysates of NB1 cells exposed to heparins withthe indicated average chain lengths at various concentrations for10min. (B) Western blot for the indicated proteins using lysates ofNB1 cells exposed to mixed chain length heparin (10 mg/ml) andSOS at various concentrations for 15 min. pMAPK, phosphorylatedmitogen-activated protein kinase. (C) Graphs representing theautophosphorylation of ALK measured by ELISA in lysates of NB1cells exposed to the indicated heparins and heparin derivatives atvarious concentrations for 15min. Various desulfatedheparin deriva-tives were used that lack sulfates on either the iduronic acid or gluco-samine subunits of heparin as follows: O-linked sulfates on carbon-2of iduronic acid; O-linked sulfates on carbon-6 of glucosamine;O-linked sulfates on carbon-2 and on carbon-6 glucosamine; N-linkedsulfates of glucosamine; andO- andN-linked sulfates on iduronic acidand glucosamine. (D) Western blot for the indicated proteins usinglysates of NB1 cells exposed to the indicated glycosaminoglycans(10mg/ml). For (A) and (C), data aremeans±SDof three independentexperiments. For (B) and (D), blots are representative of two inde-pendent experiments.

Fig. 6. The mAb aALK3 inhibits heparin binding and activation of ALK. (A)SPR sensogram of various IgG mAbs (100 nM) injected over FL-ECDbound to a heparin-coated surface. Data are representative of threeindependent experiments. (B) Western blot for the indicated antibodiesusing lysates of NB1 cells exposed to mixed chain length heparin (10 mg/ml)and the indicated concentrations of aALK3-Fab for 15 min. The blots arerepresentative of two independent experiments.



on May 11, 2017


Dow

nloaded from











on May 11, 2017


Dow

nloaded from











on May 11, 2017


Dow

nloaded from


25

Research articles


generated a monovalent Fab fragment derived from αALK3 (αALK3-Fab). αALK3-Fab inhib-ited heparin-induced autophosphorylation of ALK with an estimated IC

50 <100 nM (Fig.

6B). Likewise, αALK3-Fab inhibited heparin-induced phosphorylation of AKT and ERK 1/2 (Fig. 6B).

DISCUSSIONWe found that heparin bound to ALK and in-duced ALK tyrosine autophosphorylation, sug-gesting a role for sulfated glycosaminoglycans and associated proteoglycans as physiological ligands for ALK. One possible mechanism of ALK activation is that sulfated carbohydrate moieties of a proteoglycan bind to the posi-tively charged region in the NTR of ALK and the protein core of the same proteoglycan binds to another region in ALK. Another po-tential mechanism is that the sulfated carbo-hydrate moieties of the proteoglycan could act together with a second, yet to be identified, protein ligand to stimulate ALK dimeriza-tion and activation similar to the paradigm established for heparin and FGF stimulation of FGFR (11). LTK and ALK share homology in the glycine-rich region and EGF-like repeat. The glycine-rich region and EGF-like repeat comprise the entire ECD of LTK, suggesting that these regions likely serve as the binding site for an unidentified ligand. Thus, although ALK and LTK may share identical or related ligands, ALK is likely activated by two or more co-ligands: a sulfated glycosaminoglycan-linked proteoglycan that binds the NTR and a ligand that binds to the glycine-rich and EGF-like repeat of the ECD.

The identification of heparin as a ligand or co-ligand provides insight and tools for investigating ALK biology. Little is known about the function of mammalian ALK. In Drosophila, constitutive activation of ALK by Jeb is essential for maintaining normal brain metabolism during starvation (18). In addition to its role in neurobiology, ALK signaling has been implicated in driving several types of cancer (1–3). Similar to other RTKs, ligand-mediated increases in ALK signaling may contribute to tumorigenesis. Thus, the discovery of heparin as a bona fide ALK ligand provides mechanistic understanding of ALK activation and potential methods to inhibit oncogenic ALK by targeting its ECD.


Cloning, expression, and purification of FL-ECD, ΔN-ECD, and heparin-binding motif mutations

The nucleotide sequence coding for amino acids 1 to 1037 of dog ALK (encoding FL-ECD) (16) was synthesized by Blue Heron. An 8X-His tag was added to the 3′ end, fol-lowed by a stop codon. An Xho I site was added upstream of the start codon, and an

Xba I site was added directly after the stop codon. The construct was subcloned into pcDNA3.3. ΔN-ECD was produced by polymerase chain reaction–mediated de-letion (19) of amino acids 21 to 263 using FL-ECD as the template. The mutations in the heparin-binding motif were produced using site-directed mutagenesis using the QuikChange II Site-Directed Mutagenesis protocol developed by Stratagene (catalog no. 200523). All plasmids were verified by sequencing of the complete open reading frame.

To produce recombinant His-tagged FL-ECD and ΔN-ECD, plasmids were transiently transfected into HEK293-S cells grown in Dulbecco’s modified Eagle’s medium–F12 with 10% fetal bovine serum (FBS) and 1% penicillin and streptomycin using the Lipo-fectamine 2000 as per the manufacturer’s instructions (Invitrogen). The culture me-dium was switched to Opti-MEM just before transfection. Transfected cells were incu-bated in Opti-MEM for 4 days. The medium was collected and clarified by centrifugation and vacuum filtration (0.45-μm polyvinyli-dene difluoride filter).

To isolate His-tagged proteins, nickel-Sepharose excel beads (GE Healthcare) were added to the clarified medium and incubated overnight at 4°C with agitation. Beads were washed with 30 column vol-umes (CVs) of phosphate-buffered saline (PBS) and 30 CVs of 20 mM imidazole in buffer A [25 mM Hepes, 150 mM NaCl, and 10% glycerol (pH 7.4)]. Elution was per-formed with 250 mM imidazole in buffer A. Fractions were collected and further purified by SEC on a HiLoad Superdex 200 column preequilibrated with buffer A. Fractions containing His-tagged proteins (estimated to be >97% pure) were collected, combined, and concentrated to 10 mg/ml using a 30,000-MWCO (molecular weight cutoff) concentrator device (Sartorius Ste-dim). For heparin-Sepharose chromatogra-phy experiments, eluted protein from the nickel column was dialyzed against buffer A without imidazole and then incubated with heparin-Sepharose beads (GE Health-care) overnight. Heparin-Sepharose beads were washed with 50 CVs of buffer A and eluted in 25 mM Hepes, 1 M NaCl, and 10% glycerol at pH 7.4.

Generation of mAbs and αALK3-Fab

mAbs were purified from hybridoma cul-tures, derived from mice immunized with the ALK ectodomain (either M1-G460 or T637-S1038, which was synthesized, cloned, and expressed with 8X-His tags as described above for the dog FL-ECD and ΔN-ECD). Conditioned hybridoma medium was passed through a protein A column and washed with PBS, and mAbs were eluted with 0.1 M glycine (pH 2.7) and im-

mediately neutralized with 1 M tris (pH 7.5). The buffer containing the purified mAbs was changed to PBS by tangential flow filtration.

αALK3-Fab was generated from the pa-rental αALK3 IgG antibody and purified us-ing the Fab Preparation Kit (Thermo-Pierce, catalog no. 44985). The Fab fragment was further purified by SEC, and the buffer was changed to PBS using a HiLoad Superdex 200 26/600 column.

NTR epitope mapping

The NTR of human ALK used for the ex-periment shown in fig. S5 was cloned into pET28b (Novagen) and expressed in BL21-(de3) Escherichia coli cells. Inclusion bodies were isolated and solubilized in Laemmli sample buffer for use in SDS–polyacrylamide gel electrophoresis (SDS-PAGE). Solubilized inclusion bodies containing the NTR were used for Western blot.

ALK phosphorylation and signaling assays in NB1 cells

NB1 cells obtained from the American Type Culture Collection were cultured in RPMI supplemented with 10% FBS and 1% penicil-lin and streptomycin. NB1 cells were incu-bated with the indicated concentrations of heparin for 10 min at 37°C. For experiments with mAbs or αALK3-Fab, antibodies were incubated with NB1 cells before exposure to heparin. SOS, PTN, MK, and differentially sulfated glycans were added to NB1 cells at the same time as heparin. Different chain length heparins, various glycosaminogly-cans, and desulfated heparins were pur-chased from Neoparin (catalog nos. GT8021, GT8011, GT8086, GT8085, GT80840, GT8083, GT8082, GT6030, GT6014, GT6013, GT6012, GT6011, and GT6020). Unfractionated hepa-rin was purchased from Sigma (catalog no. H3393-25KU).

For Western blots, cell lysates were re-solved by SDS-PAGE and transferred to ni-trocellulose membranes, and membranes were incubated with antibodies against phosphorylated and total ALK, ERK 1/2, and AKT and β-tubulin (Cell Signaling Technol-ogy, catalog nos. 3633, 3333, 4695, 4370, 9272, 4060, and 2128).

For ELISAs, cell lysates were incubated in 96-well plates coated with an ALK antibody (Cell Signaling Technology, catalog no. 3791). Plates were rinsed with PBS and then incu-bated with a horseradish peroxidase (HRP)–conjugated phosphotyrosine antibody (R&D Systems, catalog no. HAM1676). HRP ac-tivity was detected with SuperSignal West Pico Chemiluminescent Substrate (Thermo-Pierce) using a plate reader (BioTek). Lumi-nescence was normalized to the maximal observed value and plotted as a function of the log10-transformed concentration of the glycosaminoglycan.


26



ALK internalization assayNB1 cells were exposed to heparin or αALK1 for 6 hours. Cells were washed with PBS three times and incubated with cell- impermeable Sulfo-NHS-LC-Biotin (0.5 mg/ml) (Thermo-Pierce, catalog no. 21435) in PBS for 1 hour. Then, cells were washed with PBS containing 0.1 M glycine and lysed with 50 mM Hepes, 150 nM NaCl, 1 mM EDTA, 1 mM EGTA, 10% glycerol, 1% Triton X-100, 25 mM NaF, 1 mM Na

3VO

4, 1 mM phenylmethylsul-

fonyl fluoride, and Roche Complete Protease Inhibitor Cocktail. Biotinylated proteins were enriched with 50 μl of NeutrAvidin agarose beads (Thermo-Pierce, catalog no. 29200) in 500 μl of lysate at room temperature for 1 hour, eluted overnight with sample buffer containing β-mercaptoethanol, and resolved by SDS-PAGE. Protein was transferred to ni-trocellulose, and membranes were immuno- blotted with an antibody for ALK.

Surface plasmon resonanceSPR experiments were performed using a Biacore T100 instrument (GE Healthcare) at 25°C. All reagents were dialyzed into a buffer composed of 25 mM Hepes, 150 mM NaCl, and 10% glycerol at pH 7.4. Biotinylated heparin (≥97% pure, Sigma, catalog no. B9806-10MG) was further purified using PD-10 prepacked columns (GE Healthcare, catalog no. 17-0851-01) to remove free biotin. Biotinylated heparin was immobilized on an assembled NeutrAvidin (Thermo-Pierce) surface (amine coupled on a CM4 Series-S Biacore chip). Three surfaces were produced with different concentrations of biotinylated heparin by varying contact time from 48 to 240 s. Three-fold dilutions of FL-ECD and ΔN-ECD were in-jected sequentially and in random order over a reference surface without heparin and three heparin surfaces. The surface was regenerated between cycles with 2.5 M NaCl and 5 mM ace-tic acid at pH 4.5. For the SOS competition as-say, serial dilutions of SOS were preincubated with 0.350 μM FL-ECD and then injected over the surface. IC50 and K

i values were derived

by plotting the log-transformed concentration of SOS as a function of fractional bound FL-ECD at saturation and applying the Cheng-Prusoff equation.

Isothermal titration calorimetry

ITC assays were performed using a MicroCal VP-ITC instrument (Malvern) with a 1.3-ml cell volume and 250-μl ligand syringe with 25 mM Hepes, 150 mM NaCl, and 10% glycerol at pH 7.4 at 25°C. Each macromolecule and ligand was extensively dialyzed against this buffer. Heparins of defined average length were purchased from Neoparin. For dp25 heparin binding to ALK, 1.43 ml of 8.3 μM ALK was placed in the cell. Two hundred fifty microliters of 44 μM dp25 heparin was titrated in 8-μl increments. For dp15 heparin binding to ALK, 1.43 ml of 6 μM

REFERENCES AND NOTES 1. T. J. Pugh et al., Nat. Genet. 45, 279–284 (2013). 2. B. Hallberg, R. H. Palmer, Nat. Rev. Cancer 13, 685–700 (2013). 3. A. K. Murugan, M. Xing, Cancer Res. 71, 4403–4411 (2011). 4. G. E. Stoica et al., J. Biol. Chem. 277, 35990– 35998 (2002). 5. G. E. Stoica et al., J. Biol. Chem. 276, 16772–16779 (2001). 6. W. G. Dirks et al., Int. J. Cancer 100, 49–56 (2002).

7. T. Mathivet, P. Mazot, M. Vigny, Cell. Signal. 19, 2434–2443 (2007). 8. I. Miyake et al., Oncogene 21, 5823–5834 (2002). 9 . C. Moog-Lutz et al., J. Biol. Chem. 280, 26039– 26048 (2005). 10. A. Motegi, J. Fujimoto, M. Kotani, H. Sakuraba, T. Yamamoto, J. Cell Sci. 117, 3319–3329 (2004). 11. M. A. Lemmon, J. Schlessinger, Cell 141, 1117–1134 (2010). 12. L. A. Kelley, M. J. Sternberg, Nat. Protoc. 4, 363–371 (2009). 13. J. Schlessinger et al., Mol. Cell 6, 743–750 (2000). 14. O. A. Ibrahimi, F. Zhang, S. C. Lang Hrstka, M. Mohammadi, R. J. Linhardt, Biochemistry 43, 4724–4730 (2004). 15. C. H. Coles et al., Science 332, 484–488 (2011). 16. W. J. Kent et al., Genome Res. 12, 996–1006 (2002). 17. D. L. Rabenstein, Nat. Prod. Rep. 19, 312–331 (2002). 18. L. Y. Cheng et al., Cell 146, 435–447 (2011). 19. M. D. Hansson, K. Rzeznicka, M. Rosenbäck, M. Hansson, N. Sirijovski, Anal. Biochem. 375, 373–375 (2008). 20. E. Folta-Stogniew, K. R. Williams, J. Biomol. Tech. 10, 51–63 (1999).

ACKNOWLEDGMENTS We thank J. H. Bae, J. Mohanty, S. Lee, and N. Kucera for discussion and assistance. Funding: This work was supported by a grant from Kolltan Pharmaceuticals (I.L.). Author contributions: P.B.M. designed and performed biophysical- and cell-based experiments and analyses, prepared the figures, and wrote the manuscript. I.L., G.F.L., J.S.L., E.J.N., X.S., and D.A. designed and performed cell-based assays. A.R., E.F.-S., and M.G. helped design or perform biophysical assays. D.A. and J.S. designed the overall study, supervised the experiments, and wrote the manuscript. Competing interests: The authors declare the following competing financial interests: J.S. is the founder and consultant of Kolltan. G.F.L., J.S.L., E.J.N., and D.A. are employees of Kolltan. Data and materials availability: Data and materials are available on request from J.S.

Submitted 15 September 2014Accepted 23 December 2014Final Publication 20 January 201510.1126/scisignal.2005916Citation: P. B. Murray, I. Lax, A. Reshetnyak, G. F. Ligon, J. S. Lillquist, E. J. Natoli Jr., X. Shi, E. Folta-Stogniew, M. Gunel, D. Alvarado, J. Schlessinger, Heparin is an activating ligand of the orphan receptor tyrosine kinase ALK. Sci. Signal. 8, ra6 (2015).

SUPPLEMENTARY MATERIALSwww.sciencesignaling.org/cgi/content/full/8/360/ra6/DC1Fig. S1. Quantification of heparin-induced ERK 1/2 activation and ALK internalization.Fig. S2. Heparin induces activation of full-length ALK, but not ΔN-ALK.Fig. S3. Heparins with short chain lengths do not induce dimerization of ALK.Fig. S4. Background binding of mAbs to heparin-only SPR surface.Fig. S5. Epitope mapping of αALK3 by Western blot.Fig. S6. Activation of ALK by mAbs.

ALK was placed in the cell. Two hundred fifty microliters of 60 μM dp25 heparin was titrated in 8-μl increments. For dp8–9 heparin binding to ALK, 1.43 ml of 10 μM ALK was placed in the cell. Two hundred fifty microliters of 150 μM dp25 heparin was titrated in 10-μl increments. Data were collected and then processed, corrected for heat of dilution, and analyzed using “Origin 5.0 with MicroCal ITC feature” software. Data were fit to a one-site model by nonlinear least squares regression from which the calculated affinities and stoichiometries were derived.

Size exclusion chromatography–multiangle laser light scattering

The light scattering data were collected us-ing a Superose 6 column in tandem with a Superdex 75, 10/300, HR SEC column (for FL-ECD experiments) or a Superose 6 column in tandem with a second Superose 6 column (for ΔN-ECD experiments) (GE Healthcare), connected to high-performance liquid chro-matography (HPLC) System (Agilent 1200, Agilent Technologies) equipped with an auto-sampler. The elution from SEC was monitored by a photodiode array ultraviolet (UV) and visible light detector (Agilent Technologies), differential refractometer (Optilab rEX, Wy-att Corp.), and a static and dynamic MALLS detector [HELEOS II with QELS (quasielastic light scattering) capability, Wyatt Corp.]. The SEC detection system was equilibrated in 150 mM NaCl, 25 mM Hepes (pH 7.4), and 10% glycerol at a flow rate of 0.4 ml/min. Chem-Station software (Agilent Technologies) was used to control the HPLC and data collec-tion from the multiwavelength photodiode array UV and visible light detector. ASTRA software (Wyatt Corp.) was used to collect data from the refractive index detector and the light scattering detectors, and the UV trace at 280 nm, sent from the photodiode array detector, was recorded. Average mo-lecular masses were determined across the entire elution profile in the intervals of 1 s from static light scattering measurement us-ing ASTRA software as previously described (20). Heparin-ALK complexes were prepared by mixing heparin (purchased from Neopa-rin) in excess with FL-ECD.


http://www.sciencesignaling.org/cgi/content/full/8/360/ra6/DC1

Activated Gα13

K204A

C-terminal

PHD

H

RH

LARG

GE Healthcare

White paper, Biacore™ systems for label-free interaction analysis

Deeper insights into biological realitiesDetection of interactionsInteractions between biomolecules are the key drivers of biological processes. For example, protein-protein interactions ensure signal transduction across membranes via G protein-coupled receptor (GPCR), maintain the structure of complexes by chaperones, or enable the enzymatic modification of proteins via post-translational phosphorylation. Hundreds of thousands of interactions have been identified and collected in databases, for example IntAct (http://www.ebi.ac.uk/intact/). Such a database allows the interactions to be assembled into pathways and studied further. Yet the majority of protein interactions probably remain undiscovered.

Homing in on the nature of interactionsGPCRs are transmembrane receptors that sense molecules outside the cell and activate signal transduction pathways, leading to cellular responses. The work of Suzuki and co-workers (1), based at Tokyo University, has increased the understanding of the interactions and guanine-nucleotide-dependent conformational changes involved in transient GPCR-signaling. Their work represents a typical approach to the study of interactions, in this case between LARG, a guanine nucleotide exchange factor for Rho, and Gα

13, a

GPCR subunit. The study starts with detailed interaction analysis, and concludes by looking at the thermodynamics of conformational changes.

Which domains interact?The first step for Suzuki and co-workers was to determine if interactions occur, and their dependency on effector molecules. They used a cell-based assay with cells transfected with myc-tagged deletion constructs of LARG with different combinations of domains, and Gα

13. They

showed that Gα13

interacts directly with LARG through its RH domain, DH/PH domains, and C-terminal region (Fig 1). The next step was to characterize these interactions with SPR-based assays using Biacore 3000.

SPR-based analysis confirms the interactions and provides more detailStudies were based on two variants of Gα

13: Gα

i /13, a chimera

of Gα13

, and Gαi /13

KA, which has a mutation, K204A, in the domain that recognizes the RH domain of LARG. These variants could be activated with AMF (AlCl

3, MgCl

2, and NaF).

The variants of Gα13

were immobilized on the surface of Sensor Chip CM5 using amine coupling chemistry, and activated by including AMF in the running buffer (Fig 2). Solutions of LARG constructs at different concentrations were flowed over the sensor surface in the presence or absence of AMF, followed by a regeneration buffer.

The results for Gαi /13

clearly showed that AMF activation is necessary for all interactions and that these involve RH, DH, and PH domains of LARG. Comparing LARG construct DPC (DH, PH, and C-terminal) and DH/PH (DH and PH domains, only) indicated that the C-terminal is also involved. Suzuki and co-workers also used Gα

i /13KA to confirm that

the mutation at Lys-204 (K204A) significantly reduced the affinity of Gα

13 for the RH domain of LARG, without affecting

interactions with the DH and PH domains.

Fig 1. Leukemia-associated RhoGEF (LARG) is a guanine nucleotide exchange factor for Rho and has a number of domains that can interact with Gα

13:

RH: Regulator of G protein Signaling (RGS) domain, which is required for activity.

DH: Dbl homology (DH) or RhoGEF domain. This consists of an approximately 150 amino acid region that induces LARG to displace GDP.

PH: Pleckstrin homology domain. This domain can bind phosphatidylinositol of membranes and thereby recruit proteins to be directed to cellular compartments or interact in signal transduction pathways. The PH domain can increase catalytic efficiency.

The position of the K204A mutation in the RH-binding region of Gαi /13

KA is also shown.

27

White paper



0 60 120 180

Resp

onse

(RU

)

AMF(+)

80

60

40

20

0

Gαi/13

Domains in construct

Activated Inactivated Activated Inactivated

Gαi/13KA

0 60 120 180

Resp

onse

(RU

) 80

60

40

20

0

-20

0 60 120 180

Resp

onse

(RU

) 160

120

80

40

0

0 60 120 180

Resp

onse

(RU

) 80

60

40

20

0

0 60 120 180

Time (s)

Time (s)

Time (s)

Time (s)

Time (s)

Resp

onse

(RU

) 604020

0-20-40-60-80

0 60 120 180

AMF(-)

80

60

40

20

0

0 60 120 180

80

60

40

20

0

-20

0 60 120 180

160

120

80

40

0

0 60 120 180

80

60

40

20

0

0 60 120 180

Time (s)

Low signal

Low signal Low signal

Low signal

Low signal Low signal Low signal

Low signal

Low signal

Low signal

Time (s)

Time (s)

Time (s)

Time (s)

604020

0-20-40-60-80

0 60 120 180

Resp

onse

(RU

)

AMF(+)

80

60

40

20

0

0 60 120 180

Resp

onse

(RU

) 80

60

40

20

0

-20

0 60 120 180Re

spon

se (R

U) 160

120

80

40

0

0 60 120 180

Resp

onse

(RU

) 80

60

40

20

0

0 60 120 180

Time (s)

Time (s)

Time (s)

Time (s)

Time (s)

Resp

onse

(RU

) 604020

0-20-40-60-80

0 60 120 180

AMF(-)

80

60

40

20

0

0 60 120 180

80

60

40

20

0

-20

0 60 120 180

160

120

80

40

0

0 60 120 180

80

60

40

20

0

0 60 120 180

Time (s)

Time (s)

Time (s)

Time (s)

Time (s)

604020

0-20-40-60-80

RH DH PH C

RH

RH

DH PH

DH PH C

DH PH

LARG-RHvan’t Hoff

LARG-RDPvan’t Hoff

In K

D

1/T × 103

In K

D

1/T × 103

3.46 3.50 3.54 3.58

-18.00

-18.05

-18.10

-18.15

-18.20

-18.25

3.46 3.50 3.54 3.58

-18.00

-18.05

-18.10

-18.15

-18.20

-18.25

FreeGα

13

+LARG-RDP

Association∆G

ass= 36

Equilibrium∆G = -47

Dissociation∆G

diss = -83

TransitionGα

13

LARG-RDP

BoundGα

13

-LARG-RDP

∆G

(kJ

mol

-1)

Reaction state

Determining the extent of the interactionThe raw data represented in these sensorgrams was used to determine kinetic data: association rate constant (k

a),

dissociation rate constant (kd) and equilibrium constant (K

D).

The relative affinity (in parentheses) of the LARG constructs for Gα

i /13 increased in the order RH (×1) < RH-DH-PH (×2)

< RH-DH-PH-C terminal (×20). The interaction involving the C-terminal therefore decreased the dissociation rate of the Gα

i /13-LARG complex considerably. The association of

the RH domain with the Gα13

surface, including Lys-204, probably induces the conformational change of the DH/PH domains of LARG necessary for its biological action.

What are the dynamics?Suzuki and co-workers decided to investigate the dynamics of the conformational changes induced by Gα

13-LARG

binding (especially the contribution of the RH-domain and the DH/PH domains) by determining the thermodynamics of these interactions. This involved using Biacore T100 to analyze the interactions between LARG and Gα

i /13

immobilized on Sensor Chip CM5 (Fig 3).

Fig 2. Kinetics of binding of LARG to Gαi /13

or Gαi /13

KA immobilized on a sensor surface and analyzed using Biacore 3000. Gα

i /13 and Gα

i /13KA proteins were immobilized on Sensor Chip CM5 on separate spots. The

concentrations of proteins were in the nM to μM range, depending on the protein. (Adapted from Figure 1C in Suzuki, N. et al., J. Biol. Chem. 284, 5000–5009 [2009]).

The interaction between RH-DH-PH and Gα13

had a higher free energy change than the RH/Gα

13 interaction, which is

consistent with the higher affinity of the LARG-RH-DH-PH construct compared with the LARG-RH construct (Fig 3). Of particular interest was the fact that the Gα

13-RH interaction

is enthalpy-driven and entropically unfavorable. This contrasted with the binding of the RH-DH-PH construct to Gα

13, which is less favorable than LARG-RH with respect

to enthalpy, but is entropy-driven with a positive ΔS0. In general, a large negative heat capacity change (ΔC

p0)

in a protein-protein interaction indicates the removal of water-accessible hydrophobic surface area coupled to conformational changes. Indeed, the researchers were able to conclude that the interaction between Gα13 and the RH domain of LARG triggers conformational changes that bury an exposed hydrophobic surface to create a large complementary surface, thereby promoting complex formation.

Fig 3. Thermodynamic analysis of Gα13

-LARG interaction. (A) Thermodynamic analysis of the Gα

13-LARG complex formation and dissociation

through its RH and DH domain. van’t Hoff plots of the experimental data are shown. (B) Schematic reaction profile of the thermodynamic energies at the different states of Gα13

-LARG interaction. The thermodynamic parameters at an equilibrium state and at a transition state were estimated from van’t Hoff plots and Eyring plots. (Adapted from Figure 4 in Suzuki, N. et al., J. Biol. Chem. 284, 5000–5009 [2009]).

(A)

(B)

28



-100 0 100 200

Resp

onse

(RU

) 40

30

20

10

0

Time (s)-100 400 900 1400 1900

Resp

onse

(RU

) 80

60

40

20

0

Time (s)

-100 0 100 200 300

Resp

onse

(RU

)

30

20

10

0

Time (s)-100 0 100 200

Resp

onse

(RU

) 50

40

30

20

10

0

Time (s)

-50 0 50 100 150

Resp

onse

(RU

)

30

20

10

0

Time (s)

KineticsThe cell is a dynamic system. So, identifying interactions is just the first step in interaction analysis and most often leads to an investigation of kinetics, to answer the following questions:

• How fast do molecules bind (association)?

• How fast do complexes fall apart (dissociation)?

Kinetics therefore determine whether a complex forms or dissociates within a given time span. Association and dissociation measurements can also be used to determine how much complex is formed at equilibrium (the steady state where association balances dissociation). This is the affinity of the interaction.

A stable basis for kinetics studiesDetermining the residence time or half-life of a drug on its target receptor is a critical point in drug discovery. SPR-based assays are routinely used in biophysical screening of soluble drug targets to determine equilibrium binding constants, kinetic rate constants and thermodynamic parameters. It becomes more challenging to analyze membrane protein targets, which can be destabilized by detergent-extraction and lose their native ligand-binding capability. This problem is particularly acute in the study of GPCRs, a very important group of drug targets. One method for engineering stability into receptors, called Stabilized Receptor, or StaR™, has enabled the study of receptors that were previously difficult to analyze, including GPCRs. These receptors can be purified in large quantities, retain correct folding, and are stabilized to such an extent that they can readily be used in binding assays. Robertson and co-workers at Heptares Therapeutics in the UK, have used SPR-based assays to analyze the kinetics of interactions between drug candidates and StaR molecules (2).

Designing in protein stabilityThe researchers first determined the success of mutagenesis by determining the thermostability of over three hundred mutations, and combinations of mutations with a method involving stabilization with a radioactive agonist. Stability with respect to detergents was also tested. The most promising candidates were then selected for recombination and kinetic analysis.

Determining kinetics with more accuracy and throughputThe kinetics of interactions involving GPCRs are usually determined by measuring dissociation rates of a radioactive ligand in the presence or absence of the test compound. Having confirmed that radio-ligand binding studies and

SPR-based assays gave similar results for affinity, Robertson and co-workers found that SPR-based assays enabled analysis of kinetics on solid-phase and thereby provided a simpler, higher throughput, and more accurate method compared to conventional methods.

The researchers used SPR to examine the kinetics of one stabilized construct, A

2A-Star2, in more detail. The kinetic

characterization of five A2A

ligands, with relative molecular masses in the range 285 to 345, using SPR detection assays is shown in Figure 4. The binding was concentration dependent, and showed high reproducibility between triplicates. Binding parameters could be determined after fitting the sensorgrams to a 1:1 model. The range of association and dissociation rate constants displayed by these antagonists gave a 10 000-fold range in affinities. In addition, the validity of the approach was confirmed by the fact that affinity constants closely correlated with those obtained in equilibrium binding studies on the wild-type receptor in membranes.

Fig 4. Kinetic characterization of five A2A

ligands using SPR. A2A

StaR2 was immobilized on Sensor Chip CM5 using amine coupling. On- and off-rates of antagonist binding were then determined for each compound, (Adapted from Figure 6 in Robertson, N. et al., Neuropharmacology 60, 36–44 (2011)).

The ability to stabilize receptors opens up new possibilities for the analysis of relatively unstable GPCRs, using crystallography to probe structure, and SPR to analyze kinetics. This approach should lead to the identification of ligand candidates with the desired interaction properties.

29

White paper


Anti-GST antibody

GST U

bL

Dextran matrix

GST U

b

GST U

bL-inse

rt

USP4-DID2

Resp

onse

(RU

)

Time (s) -200 0 200 400 600 800

125

100

75

50

25

0

(B)

(A)

Resp

onse

(RU

)Time (s)

-200 0 200 400 600 800

150

100

50

0

(C)

Resp

onse

(RU

)

Time (s) -200 0 200 400 600 800

40

20

0

(D)

(E) (F)

GST-Ub GST-insert

Equi

libriu

m r

espo

nse

Concentration (µM) 0.01 0.1 1 10

150

100

50

0

GST-insert

Equi

libriu

m r

espo

nse


40

30

20

10

0

GST-UbL

KD: 1.36 ±0.06 µMK

D: 1.32 ±0.07 µMK

D: 1.39 ±0.08 µM

Ki: 1.4 ±0.08 µM

USP4CDK

D: 0.64 µM

USP4-D1D2 K

D: 0.15 µM

Equi

libriu

m r

espo

nse


100

50

0

GST-Ub

GST-UbL

Free

USP

4-D

1D2

(µM

)

Ub (µM) 0.1 0.3 1 3.2 10

0.8

0.6

0.4

0.2

0.0

µcal

/skc

al/m

ol o

f inj

ecta

nt

µcal

/skc

al/m

ol o

f inj

ecta

nt

Time (min)

Molar ratio

0 10 20 30 40 50

0.0 0.5 1.0

0.100.050.00

-0.05-0.10-0.15-0.20-0.25-0.30-0.350.00

-2.00-4.00-6.00-8.00

-10.00-12.00-14.00-16.00

Time (min)

Molar ratio

0 10 20 30 40 50

0.0 0.5 1.0 1.5

0.300.200.100.00

-0.10-0.20-0.30-0.40-0.50-0.600.00

-2.00-4.00-6.00-8.00

-10.00-12.00

Structure and functionThe structure of a biomolecule is usually critical to its function. The structure may be static, but is more often dynamic and influences, or is influenced by, interactions with other biomolecules. Very often, as in the case of enzymes, changes in structure are instrumental in the function.

Protein structure may be determined by sequencing and then applying biophysical methods such as X-ray crystallography and NMR to determine the three-dimensional, structure in interaction with cofactors where necessary. Understanding the roles played by conformational changes and structure in interactions requires application of other methods, such as biochemical assays and biophysical techniques, such as SPR-based assays.

Find out how proteins work by looking at their domainsThe three-dimensional structure of a protein can be determined from its sequence and by applying biophysical techniques, such as X-ray crystallography and NMR spectroscopy. The structural domains within the protein can then be dissected out and isolated, which then makes it possible to apply SPR-based assays to investigate interactions between domains. This deepens the understanding of the dynamics that link structure and function. Luna-Vargas and co-workers, based at The Netherlands Cancer Institute, have applied this approach in the study of an enzyme involved in the processing of ubiquitin (3).

Ubiquitin (Ub) is a small, ubiquitous regulatory protein with a number of functions, including being used by conjugating enzymes to label proteins that are to be processed. This discovery led to the award of the Nobel Prize for chemistry to Aaron Ciechanover, Ayram Hershko, and Irwin Rose in 2004. Deubiquinating enzymes (DUBs), such as USP4, balance this labeling activity by removing Ub from the target molecule. Luna-Vargas and co-workers used SPR detection to investigate the structural elements of ubiquitin-specific protease 4 (USP4) that control its deubiquinating activity.

Determining the structural elements to be isolatedThe researchers first determined the structure of the catalytic domain of USP4, which consists of two regions, D1 and D2, using X-ray crystallography. They also prepared a number of constructs comprising various combinations of the protein domains - for example a construct with an insert between D1 and D2 that contained an Ubiquitin-like domain (Ubl), as in the native protein. In vitro de-ubiquitinating assays indicated that including the Ubl-insert in USP4-D1D2 (USP4-D1-Ubl-insert-D2) reduced catalytic activity, which suggested that the Ubl-insert inhibits the DUB activity of USP4 through an autoregulatory mechanism.

Exploring the interactionsThe researchers then used SPR-based assays to determine if the Ubl-insert interacts directly with the catalytic domain, USP4-D1D2. GST (glutathione S-transferase)-fused fragments were immobilized on α-GST antibodies that had been lysine-coupled to Sensor Chip CM5 (Fig 5A). USP4D1-D2 was flowed over the surface and interactions were detected using Biacore T100. Luna-Vargas and co-workers showed that the affinities of the Ubl-insert, and Ubl alone, for USP4-D1D2 are indeed similar to that of Ub itself (Fig 5B-D).

Fig 5. Ubiquitin competes with the insert or Ubl-domain for binding to USP4–D1D2. (A) GST-fused fragments were immobilized on α-GST antibodies that had been immobilized on a sensor chip and USP4D1-D2 was then flowed over the sensor surface. (B-D) Interaction of Ub and the insert fragments with USP4–D1D2 was analyzed by SPR-based assays. Top: (B) GST-tagged Ub, (C) GST-insert and (D) GST-Ubl domain. Bottom: Langmuir binding curves. (E) Competition experiment with immobilized GST insert on USP4–D1D2 with varying concentrations of Ub. A one-site competition-binding model was fitted (Ki 1.4 μM). (F) The interaction of Ub with USP4–D1D2 (left) and with full-length USP4CD (right) was studied by ITC. Thermodynamic values were: USP4–D1D2 (ΔH = -14.3 kcal/mol and ΔS = -16.9 cal/mol/deg), for USP4CD (ΔH = -11.4 kcal/mol and ΔS = -10.0 cal/mol/deg) (Adapted from Figure 3 in Luna-Vargas, M .P. et al., EMBO Rep. 12, 365–72 [2011]).

30



D1D2

Ubl

UblTarg

et

Ubl

D1D2

Ub

D1D2

D1D2

Ubl

UblTarg

et

Ubl

D1D2

Ub

D1D2

US

P3 9 C D

US

P 3 9 C D

Ubl competes with UbSPR-based detection was used in a competition assay, involving flowing USP4-D1D2 over the immobilized Ubl-insert in the presence of increasing amounts of Ub. The data could be fitted to a one-site competition binding model, showing that the USP4 insert containing Ubl competes with Ub for binding to the catalytic domain, USP4-D1D2. Additional experiments, using Isothermal Titration Calorimeter, showed that the Kd values of the constructs D1-Ubl-insert-D2, and USP4-D1D2 for Ub were comparable.

Data from SPR-based assays indicated that Ub binds to D1-Ubl-insert-D2 with slower off- and on-rates than when binding to USP4-D1D2. The researchers concluded that the insert prevents both rapid binding and rapid release of

the Ub substrate, thus allowing competitive binding. The similarity of the KD values for purified Ubl and Ubl-insert (Fig 5B-D) suggested that the Ubl domain is the functional part of the insert. SPR-based analysis was also used to demonstrate that the Ubl domain could bind to the catalytic domain of other DUBs. Luna-Vargas and co-workers summarized their data in a model in which DUB activity is partially inhibited by the Ubl domain, which binds to the Ub-binding region of USP4, thus preventing Ub substrate binding (Fig 6).

Explore moreWhat do you need to get the confidence to take the next step in your research or to submit your next paper for publication? Just imagine what a better understanding of molecular function and activity could do for your research.

Biacore systems can provide key data in real-time to discriminate crucial differences in affinity, even for interactions where challenging targets are involved.

Biacore systems are designed to help you to generate decisive, information-rich data, that will help answer key questions concerning the nature of binding.

References1. Suzuki, N. et al. Activation of Leukemia-associated RhoGEF by G13 with Significant

Conformational Rearrangements in the Interface. J. Biol. Chem. 284, 5000–5009 (2009).

2. Robertson, N. et al. The properties of thermostabilised G protein-coupled receptors (StaRs) and their use in drug discovery. Neuropharmacology 60, 36–44 (2011).

3. Luna-Vargas, M .P. et al. Ubiquitin-specific protease 4 is inhibited by its ubiquitin-like domain. EMBO Rep. 12, 365–72 (2011).

(A) (B)

Fig 6. Model for Ubl domain inhibition on USP4. (A) Schematic model of the auto-inhibitory role of the Ubl domain in USP4. (B) Other USP enzymes, such as USP39, may relieve the inhibition by binding to the Ubl domain (Adapted from Figure 5 in Luna-Vargas, M .P. et al., EMBO Rep. 12, 365-72 [2011]).

gelifesciences.com/biacoreGE, the GE Monogram, and Biacore are trademarks of General Electric Company.StaR is a trademark of Heptares Therapeutics. All other third-party trademarks are the property of their respective owners.© 2017 General Electric Company All goods and services are sold subject to the terms and conditions of sale of the company within GE Healthcare which supplies them. A copy of these terms and conditions is available on request. Contact your local GE Healthcare representative for the most current information.GE Healthcare Bio-Sciences AB, Björkgatan 30, 751 84 Uppsala, SwedenFor local office contact information, visit gelifesciences.com/contact.

29270161 AA 06/2017

31

White paper


GE Healthcare

Achieving data-driven decisions with real-time interaction analysesIntroductionBiological processes are “real-time” events, driven and regulated by dynamic interactions between key molecules.

End-point techniques such as ELISA offer a snapshot view of interactions providing only basic information, such as overall binding strength (affinity). The affinity depends on the ratio of on- and off-rates so that equal affinity interactions can have very different kinetic properties, resulting in different biological responses.

Biacore SPR systems can provide key data in real-time to discriminate these crucial differences, even for interactions where challenging targets are involved.

Biacore systems are designed to help you to generate decisive, information-rich data in real time. This information will help answer the following key questions concerning the nature of the binding.

How strong?Affinity is a steady-state measurement made at equilibrium of a binding event and reflects the strength of an attraction between molecules.

How fast?Binding kinetics determine how fast/slow a complex forms or dissociates within a given time span and allow calculation of association and dissociation rate constants.

How much?Qualitative and quantitative determination of active analyte binding to a target protein.

How specific?Is the molecule specific for its target? Does the antibody recognize multiple derivatives? The flexibility in the Biacore assay design allows rapid assessment of cross-reactivity and specificity.

Data-driven understandingThe high information data provided by label-free interaction analysis enables scientists to fully understand binding events between almost any types of biologically relevant interactants.

• Understand the relationship between molecular interaction and function

• Screen for hits and optimize leads based on selectivity, affinity, and kinetics

• Examine interactions of ions, small molecules, and multidomain proteins or viruses with targets

• Screen and characterize antibodies and proteins based on yes/no binding, affinity, and kinetics from the fastest on-rates to the slowest off-rates

• Quantitate protein by measuring the concentration of active protein with retained biological function

SPR principleDuring SPR analysis, one of the interacting molecules is immobilized on a sensor surface, while the potential interacting partner flows over the sensor surface in solution. Interactions between the two are detected in real-time through changes in mass concentration close to the sensor surface. Binding data is presented in a sensorgram, where SPR responses in resonance units (RU) are plotted versus time (Fig 1).

Advantages of kineticsInteractions characterized by similar affinities can have very different kinetic properties, resulting in different biological responses. By resolving affinity into on- and off-rates, comprehensive information is obtained on how the dynamics of molecular interactions relate to protein function. On-rates reflect recognition between interacting partners while off-rates indicate stability of the complex.

Technical note, Biacore™ systems for label-free interaction analysis

32




-40 -20 0 20 40 60 80 100 120 140 160Time (s)

0

0.5

1.0

1.5

2.0

Resp

onse

(RU

)

0 1 2 3 4 5 6 7 8 9Time (h)

0

20

40

60

80

Blo

cked

targ

et (%

)

Baseline(buffer)

Association(sample solution)

Dissociation(buffer)

Resp

onse

(RU

)

Time

Report point

Sensorgram

Baseline

Bindingresponse

×

gelifesciences.com/biacoreGE, the GE Monogram, and Biacore are trademarks of General Electric Company.© 2017 General Electric Company All goods and services are sold subject to the terms and conditions of sale of the company within GE Healthcare which supplies them. A copy of these terms and conditions is available on request. Contact your local GE Healthcare representative for the most current information.GE Healthcare Bio-Sciences AB, Björkgatan 30, 751 84 Uppsala, SwedenFor local office contact information, visit gelifesciences.com/contact.

29270160AA 06/2017

Fig 2. Blue sensorgram illustrates rapid kinetics. Frequent administration of low dose is required to block target. Red sensorgram illustrates slow kinetics. Infrequent administration of high dose blocks target long after injection.

Importance of outstanding SPR sensitivity for reliable binding resultsFull characterization of antigen-antibody interactions is of great importance when assessing the suitability of antibodies as therapeutic or diagnostic tools. Ranking of strong binders can be complicated by avidity effects, and the dissociation rate will appear slower than in reality. To clearly differentiate strong-binding antibodies in terms of dissociation rates, it is necessary to use low levels of immobilized binding partner to obtain clean, avidity-free interaction studies. Obtaining accurate data from the low immobilization levels requires a highly sensitive SPR sensor for the analysis. Sensitivity and low baseline noise are equally important to reliably detect and profile interactions involving very small compounds or targets with low activity levels. High sensitivity with low noise enables resolution of sensorgrams at sub-resonance unit levels, exemplified in Figure 3.

Fig 1. The sensorgram provides real-time information about binding profiles with binding responses measured in resonance units (RU). Association: interactions in solution bind to molecules on sensor chip surface. Dissociation: Binders allowed to dissociate from molecule on sensor chip surface. Any remaining bound sample molecules may be removed in a regeneration step that prepares the surface for the next sample injection.

Fig 3. The high sensitivity of Biacore 8K enables confident analysis of fast on-rates. Sensorgram showing binding of melagatran to thrombin: k

a 4.0 × 107 M-1 s-1; k

d 0.014 s-1.

ConclusionUnderstanding the nature of interactions between molecules is fundamentally important for increased understanding of biological processes. The sensitivity and performance of Biacore systems provide high-quality molecular interactions data in a range of fields, including research, drug discovery, development, and quality control.

This information provides an extra dimension which can be crucial in supporting hit-to-lead development (Fig 2).

Kinetic properties play an important role in pharmacokinetics for drug development. A compound that shows rapid binding will have a quick effect, but if dissociation is also fast the effect will be short-lived. On the other hand, a compound with slower binding kinetics might need more time to reach full effect but will not need to be given as often.

-40 -20 0 20 40 60 80 100 120 140 160Time (s)

0

0.5

1.0

1.5

2.0Re

spon

se (R

U)

0 1 2 3 4 5 6 7 8 9Time (h)

0

20

40

60

80

Blo

cked

targ

et (%

)

Baseline(buffer)

Association(sample solution)

Dissociation(buffer)

Resp

onse

(RU

)

Time

Report point

Sensorgram

Baseline

Bindingresponse

×

gelifesciences.com/biacoreGE, the GE Monogram, and Biacore are trademarks of General Electric Company.© 2017 General Electric Company All goods and services are sold subject to the terms and conditions of sale of the company within GE Healthcare which supplies them. A copy of these terms and conditions is available on request. Contact your local GE Healthcare representative for the most current information.GE Healthcare Bio-Sciences AB, Björkgatan 30, 751 84 Uppsala, SwedenFor local office contact information, visit gelifesciences.com/contact.

29270160AA 06/2017

Fig 2. Blue sensorgram illustrates rapid kinetics. Frequent administration of low dose is required to block target. Red sensorgram illustrates slow kinetics. Infrequent administration of high dose blocks target long after injection.

Importance of outstanding SPR sensitivity for reliable binding resultsFull characterization of antigen-antibody interactions is of great importance when assessing the suitability of antibodies as therapeutic or diagnostic tools. Ranking of strong binders can be complicated by avidity effects, and the dissociation rate will appear slower than in reality. To clearly differentiate strong-binding antibodies in terms of dissociation rates, it is necessary to use low levels of immobilized binding partner to obtain clean, avidity-free interaction studies. Obtaining accurate data from the low immobilization levels requires a highly sensitive SPR sensor for the analysis. Sensitivity and low baseline noise are equally important to reliably detect and profile interactions involving very small compounds or targets with low activity levels. High sensitivity with low noise enables resolution of sensorgrams at sub-resonance unit levels, exemplified in Figure 3.

Fig 1. The sensorgram provides real-time information about binding profiles with binding responses measured in resonance units (RU). Association: interactions in solution bind to molecules on sensor chip surface. Dissociation: Binders allowed to dissociate from molecule on sensor chip surface. Any remaining bound sample molecules may be removed in a regeneration step that prepares the surface for the next sample injection.

Fig 3. The high sensitivity of Biacore 8K enables confident analysis of fast on-rates. Sensorgram showing binding of melagatran to thrombin: k

a 4.0 × 107 M-1 s-1; k

d 0.014 s-1.

ConclusionUnderstanding the nature of interactions between molecules is fundamentally important for increased understanding of biological processes. The sensitivity and performance of Biacore systems provide high-quality molecular interactions data in a range of fields, including research, drug discovery, development, and quality control.

This information provides an extra dimension which can be crucial in supporting hit-to-lead development (Fig 2).

Kinetic properties play an important role in pharmacokinetics for drug development. A compound that shows rapid binding will have a quick effect, but if dissociation is also fast the effect will be short-lived. On the other hand, a compound with slower binding kinetics might need more time to reach full effect but will not need to be given as often.

33

Technical note


Receive your FREE printed copy of theBiacore™ Sensor Surface HandbookThe Biacore Sensor Surface Handbook provides a general guide to the design and use of sensor surfaces for Biacore analyses, and covers:

• Properties of the sensor surface• Approaches to immobilizing ligand• Practical aspects of ligand immobilization• Regeneration of the sensor surface• Sensor chip storage and re-use

Order your FREE printed copy today.

To benefit from this offer, go to gelifesciences.com/biacoreclub, login using your password or sign up by using your Biacore instrument serial number and the latest software product key.

The following special offers are available to members of our Biacore club until 30 September 2017.

To benefit from this offer go to our Biacore club, gelifesciences.com/biacoreclub

GE Healthcare

GE, the GE Monogram, and Biacore are trademarks of General Electric Company. © 2017 General Electric Company. GE Healthcare Bio-Sciences AB, Björkgatan 30, 751 84 Uppsala, Sweden.

06/2017

gelifesciences.com

http://aaas.org

http://forceforscience.org/join

Receive your FREE printed copy of theBiacore™ Sensor Surface HandbookThe Biacore Sensor Surface Handbook provides a general guide to the design and use of sensor surfaces for Biacore analyses, and covers:

• Properties of the sensor surface• Approaches to immobilizing ligand• Practical aspects of ligand immobilization• Regeneration of the sensor surface• Sensor chip storage and re-use

Order your FREE printed copy today.

To benefit from this offer, go to gelifesciences.com/biacoreclub, login using your password or sign up by using your Biacore instrument serial number and the latest software product key.

The following special offers are available to members of our Biacore club until 30 September 2017.

To benefit from this offer go to our Biacore club, gelifesciences.com/biacoreclub

GE Healthcare

GE, the GE Monogram, and Biacore are trademarks of General Electric Company. © 2017 General Electric Company. GE Healthcare Bio-Sciences AB, Björkgatan 30, 751 84 Uppsala, Sweden.

06/2017

gelifesciences.com

http://gelifesciences.com

http://gelifesciences.com/biacoreclub


GE, GE monogram, and Biacore are trademarks of General Electric Company. © 2017 General Electric Company. First published June. 2016. GE Healthcare Bio-Sciences Corp., 100 Results Way, Marlborough, MA 01752, USA For local office contact information, visit gelifesciences.com/contact.

29215287 AA 06/2016

It’s a hit. For sure.Introducing Biacore™ 8K, the newest member of the Biacore family of products. Biacore 8K efficiently delivers binding data with the quality you expect while meeting tomorrow’s challenges in small molecule and biotherapeutic screening and characterization.

Discover more, more efficiently.

• Single solution for interaction analysis in both screening and characterization

• Screening of 2300 small molecule fragments in a day

• High-quality kinetic characterization of 64 interactions in 4 h

• 60 h unattended runtime with queueing abilities and rapid multi-run evaluations

• Confident interaction analysis of small molecules and complex targets such as GPCRs

• Confident differentiation of high-affinity binders

Discover deeper insights more efficiently.

gelifesciences.com/biacore8k

Request a demo today.

032-0062-UK_GE_Biacore_8K_Ad_062117_VF_PRINT.indd 1 6/21/17 6:39 AM

http://gelifesciences.com/biacore8k

Documents

A Sponsored Supplement to Science SPReading The importance ... · SPReading the word: The importance of binding kinetics Originally published 5 June 2015 in SCIENCE sciencemag.org