of March 24, 2018.This information is current as

Nephritic Factorsfor Therapy of Renal Disease Associated withAlternative Pathway Convertase: Potential A Humanized Antibody That Regulates the

Ronald P. Taylor, Oscar Llorca and Claire L. HarrisLindorfer, Santiago Rodriguez de Cordoba, B. Paul Morgan, Danielle Paixão-Cavalcante, Eva Torreira, Margaret A.

http://www.jimmunol.org/content/192/10/4844doi: 10.4049/jimmunol.1303131April 2014;

2014; 192:4844-4851; Prepublished online 11J Immunol 

MaterialSupplementary

1.DCSupplementalhttp://www.jimmunol.org/content/suppl/2014/04/11/jimmunol.130313

Referenceshttp://www.jimmunol.org/content/192/10/4844.full#ref-list-1

, 12 of which you can access for free at: cites 36 articlesThis article

        average*  

4 weeks from acceptance to publicationFast Publication! •    

Every submission reviewed by practicing scientistsNo Triage! •    

from submission to initial decisionRapid Reviews! 30 days* •    

Submit online. ?The JIWhy

Subscriptionhttp://jimmunol.org/subscription

is online at: The Journal of ImmunologyInformation about subscribing to

Permissionshttp://www.aai.org/About/Publications/JI/copyright.htmlSubmit copyright permission requests at:

Email Alertshttp://jimmunol.org/alertsReceive free email-alerts when new articles cite this article. Sign up at:

Print ISSN: 0022-1767 Online ISSN: 1550-6606. Immunologists, Inc. All rights reserved.Copyright © 2014 by The American Association of1451 Rockville Pike, Suite 650, Rockville, MD 20852The American Association of Immunologists, Inc.,

is published twice each month byThe Journal of Immunology

by guest on March 24, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

by guest on March 24, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

The Journal of Immunology

A Humanized Antibody That Regulates the AlternativePathway Convertase: Potential for Therapy of Renal DiseaseAssociated with Nephritic Factors

Danielle Paixao-Cavalcante,*,1 Eva Torreira,† Margaret A. Lindorfer,‡

Santiago Rodriguez de Cordoba,† B. Paul Morgan,* Ronald P. Taylor,‡ Oscar Llorca,† and

Claire L. Harris*

Dysregulation of the complement alternative pathway can cause disease in various organs that may be life-threatening. Severe

alternative pathway dysregulation can be triggered by autoantibodies to the C3 convertase, termed nephritic factors, which cause

pathological stabilization of the convertase enzyme and confer resistance to innate control mechanisms; unregulated complement

consumption followed by deposition of C3 fragments in tissues ensues. The mAb, 3E7, and its humanized derivative, H17, have been

shown previously to specifically bind activated C3 and prevent binding of both the activating protein, factor B, and the inhibitor,

factor H, which are opposite effects that complicate its potential for therapy. Using ligand binding assays, functional assays, and

electron microscopy, we show that these Abs bind C3b via a site that overlaps the binding site on C3 for the Ba domain within factor

B, thereby blocking an interaction essential for convertase formation. Both Abs also bind the preformed convertase, C3bBb, and

provide powerful inhibition of complement activation by preventing cleavage of C3. Critically, the Abs also bound and inhibited C3

cleavage by the nephritic factor–stabilized convertase. We suggest that by preventing enzyme formation and/or cleavage of C3 to

its active downstream fragments, H17 may be an effective therapy for conditions caused by severe dysregulation of the C3

convertase and, in particular, those that involve nephritic factors, such as dense deposit disease. The Journal of Immunology,

2014, 192: 4844–4851.

Complement is part of innate immunity, with key roles indefense against pathogens through opsonization and ly-sis, clearance of apoptotic cells, handling of immune

complexes, and modulation of adaptive immune responses (1).Complement can be triggered via three activation pathways: theclassical, alternative (AP), and lectin pathways, all leading to thegeneration of a C3 cleaving enzyme, or convertase, the central andmost important step of the activation cascade. Cleavage of C3generates C3b, which covalently links to target cells, bindingfactor B (fB) in an Mg2+-dependent manner to form C3bB. Thisproenzyme is activated by factor D (fD), generating the active C3convertase, C3bBb. Binding of properdin (P) stabilizes this oth-erwise labile complex. Each C3 convertase cleaves many C3 to

C3b, thus providing exponential amplification of the pathway.Complement activation progresses by formation of the C5 cleav-ing enzyme, resulting in generation of C5a and C5b. C5a isa proinflammatory peptide with anaphylactic and chemotacticproperties, whereas C5b binds the next complement component,C6, marking the start of the terminal pathway that culminates information of the cytolytic membrane attack complex (MAC) (2).The AP “ticks over” constantly in plasma. Spontaneous hy-

drolysis of C3 generates a C3b-like molecule, C3(H2O), that bindsfB, which is then processed by fD to form a fluid-phase enzyme,C3(H2O)Bb, that cleaves C3 to C3b, thus “priming” the AP forimmediate activation (3). C3b generated in the fluid phase israpidly inactivated, thus preventing uncontrolled consumption of

*Institute of Infection & Immunity, School of Medicine, Cardiff University, CardiffCF14 4XN, United Kingdom; †Centro de Investigaciones Biologicas, Consejo Supe-rior de Investigaciones Cientıficas, 28040 Madrid, Spain; and ‡Department of Bio-chemistry and Molecular Genetics, University of Virginia School of Medicine,Charlottesville, VA 22908

1Current address: Cristalia Produtos Quımicos Farmaceuticos Limitada, Itapira, SaoPaulo, Brazil.

Received for publication November 22, 2013. Accepted for publication March 17,2014.

This work was supported by Kidneeds, USA (to C.L.H. and B.P.M.), the MedicalResearch Council (Grant G0701298 to C.L.H. and B.P.M.), the Ramon Areces Foun-dation (to O.L.), the Autonomous Region of Madrid (Grant S2010/BMD-2316 to S.R.d.C.and O.L.), the Spanish Ministry of Science and Innovation (Grant SAF2011-22988 to O.L.and Grant SAF2011-26583 to S.R.d.C.), a Kidney Research UK Post-Doctoral Fellowship(Grant PDF5/2010 to D.P.-C.), the Red Tematica de Investigacion Cooperativa en Cancerfrom the Instituto de Salud Carlos III (Grant RD06/0020/1001 to O.L.), the Fundacion RenalInigo Alvarez de Toledo (to S.R.d.C.), the Ciber de Enfermedades Raras (to S.R.d.C.), andthe 7th Framework Programme European Union project EURenOmics (to S.R.d.C.).

C.L.H., R.P.T., S.R.d.C., O.L., and B.P.M. conceived and planned the project; D.P.-C.,E.T., O.L., and C.L.H. designed and performed the research experiments and analyzedthe data; D.P.-C., C.L.H., O.L., S.R.d.C., B.P.M., and R.P.T. drafted and edited themanuscript; R.P.T. and M.A.L. contributed vital reagents.

The structures presented in this article have been submitted to the EMDataBankdatabase (http://www.emdatabank.org/index.html) under accession number EMD-2553.

This work was presented in abstract form at the 23rd International ComplementWorkshop, August 1–5, 2010, New York, NY.

Address correspondence and reprint requests to Prof. Claire L. Harris, Institute ofInfection & Immunity, School of Medicine, Cardiff University, Henry WellcomeBuilding, Heath Park, Cardiff CF14 4XN, United Kingdom. E-mail address:harriscl@cardiff.ac.uk

The online version of this article contains supplemental material.

Abbreviations used in this article: aHUS, atypical hemolytic uremic syndrome; AP,alternative pathway; C3NeF, C3 nephritic factor; DAF, decay accelerating factor;DDD, dense deposit disease; E, erythrocyte; EA, Ab-coated E; EM, electron micros-copy; fB, factor B; fD, factor D; fH, factor H; HBSP, HEPES-buffered saline sup-plemented with 0.005% surfactant P20; HBSPMg, HBSP supplemented with 1 mMMgCl2; MAC, membrane attack complex; MG, macroglobulin; P, properdin; PDB,Protein Data Bank; PNH, paroxysmal nocturnal hemoglobinuria; RU, resonanceunits; sDAF, soluble recombinant DAF comprising domains 1–4; SPR, surfaceplasmon resonance.

Copyright� 2014 by TheAmerican Association of Immunologists, Inc. 0022-1767/14/$16.00

www.jimmunol.org/cgi/doi/10.4049/jimmunol.1303131

by guest on March 24, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

complement in plasma; however, a proportion binds indiscrimi-nately to any cell in its vicinity and, if not strictly regulated, candrive complement activation and cause damage to host cells.Damage to self is restricted by numerous complement regulatoryproteins present in the fluid phase (including factor H [fH]) and oncell membranes including CD55, CD35, and CD46. These regu-lators act by accelerating natural decay of C3bBb or by acting ascofactors for the proteolytic inactivation of C3b by the plasmaprotease factor I (4, 5).In health, complement is in homeostatic balance; activation in

plasma occurs at a low level, and regulation prevents significantdeposition of the central component, C3b, and limits further activationexcept on pathogens. The capacity of complement to initiate quicklyand amplify efficiently means that any disturbance in homeostasis canbe devastating to health (6). “Dysregulation” of the central compo-nents of the amplification loop, C3, fB, fD, or the control protein, fH,can cause acute or chronic inflammation and contribute to the pa-thologies associated with diverse diseases, including rheumatoidarthritis, systemic lupus erythematosus, glomerulonephritis, multiplesclerosis, sepsis, asthma, and ischemia/reperfusion injuries. In each,complement activation drives a “vicious cycle” of inflammation andtissue damage (7).It is now established that the prototypic complement dysregulation-

associated diseases, dense deposit disease (DDD), atypical hemolyticuremic syndrome (aHUS), and age-related macular degeneration, areeach associated with mutations and/or polymorphisms in the com-ponents and regulators of the AP C3 convertase (8, 9). Severedysregulation is also triggered by autoantibodies against com-plement components, complexes, or regulators. Abs that interferewith function of fH are found in some aHUS and DDD patients.Abs that bind the AP C3 convertase, C3bBb, known as C3 nephriticfactors (C3NeF), are present in .80% of patients with DDD (10, 11).Once bound to the C3 convertase, C3NeF stabilizes the C3bBb com-plex, increasing its half-life and preventing regulation by complementregulatory proteins such as fH (12). This stabilized C3 convertaseconsumes intact C3, thereby generating large amounts of fluid-phaseactivated C3 fragments (C3b, iC3b, C3dg), which locate in the kidney.Understanding of the mechanisms behind complement-mediated

diseases has advanced enormously in recent years, revealing newavenues for development of specific therapies that target differentsteps within the complement cascade. These include solublerecombinant complement regulatory proteins, such as TP10 (13),and mAbs that block complement proteins, such as eculizumab(anti-C5) (14) and TA106 (anti-fB Fab; Alexion Pharmaceuticals)(15). Small-molecule inhibitors that prevent C3 activation are intrials for disorders such as age-related macular degeneration(compstatin) (16, 17). The most successful anticomplement ther-apeutic to date is eculizumab, approved by the Food and DrugAdministration for treatment of paroxysmal nocturnal hemoglo-binuria (PNH) and aHUS, diseases in which dysregulated APtriggers terminal pathway activation, which plays a critical role inpathology. However, not all complement-mediated diseases occuras a consequence of C5a and MAC activities; in DDD, there ismarked activation of C3, and fluid-phase C3 activation fragments(primarily iC3b) trapped in the glomerular basement membrane inthe kidney are implicated in driving pathology (18).ThemAb, 3E7, and its humanized chimeric counterpart, H17, bind

C3b, C3b(H2O), and iC3b, but not native C3 or C4/C4b (19, 20).The mAbs compete with fB and fH for binding to C3b, preventingformation of C3 convertase and generation of iC3b, respectively;however, their precise mechanism of action is unclear (19). In anin vitro model of PNH, they inhibited deposition of C3b andabolished the hemolysis of PNH erythrocytes (E) (21). Inhibitionwas due to mAb binding to fluid phase C3(H2O), preventing AP

tickover and deposition of C3b on the surface of the E. However, itis not clear whether the mAbs bind and influence the AP convertase,which is important particularly in the context of the pathogenicC3NeF-stabilized AP convertase. We describe in this article themechanism by which 3E7/H17 binding influences AP convertaseactivity, whether unstabilized or stabilized by C3NeF; these data aresupported by structural analysis of the mAb in complex with C3b.The data suggest that mAb H17 may be an effective therapy indisorders mediated by dysregulation of the AP C3 convertase andparticularly in C3NeF-associated diseases such as DDD.

Materials and MethodsmAbs and complement components

The mAb 3E7, and its humanized derivative H17, have been describedpreviously (19). Fab fragments of H17 were produced by digestion withpapain. C3 was prepared by classical chromatography using an establishedprotocol (22); fB was affinity purified from EDTA plasma on anti-Bb(mAb JC1, in-house) HiTrap column (GE Healthcare). fD was purchasedfrom Complement Technology (Tyler, TX). C3b was generated by APactivation by coincubating pure C3, fB, and fD as described previously(23), followed by anion exchange chromatography (Mono Q; GE Health-care) and size exclusion chromatography (Superdex 200; GE Healthcare).Concentrations of pure proteins were assessed using absorbance at A280and the following extinction coefficients: fB: 1.43 cm21(mg/ml)21; C3:0.98 cm21(mg/ml)21; Ig 1.4 cm21(mg/ml)21. Soluble recombinant decayaccelerating factor (DAF) comprising domains 1–4 (sDAF) was gifted byProf. Susan Lea (University of Oxford). A strongly C3NeF+ Ig preparationwas donated by Dr. Margarita Lopez-Trascasa (Hospital Universitario de LaPaz, Madrid, Spain).

Surface plasmon resonance studies

Surface plasmon resonance (SPR) analyses were conducted at 25˚C ona BIAcore T100 (GE Healthcare), except for kinetic data, which wereperformed on a BIAcore 3000; kinetic/affinity data were collected at 30 ml/min and functional/blocking data at 20 ml/min. Proteins were amine-coupled to a CM5 (carboxymethylated dextran) chip (NHS/EDC cou-pling kit; GE Healthcare). Interactions were analyzed in HEPES-bufferedsaline (10 mM HEPES pH 7.4, 150 mM NaCl) supplemented with 0.005%surfactant P20 (HBSP) and either 1 mM MgCl2 (HBSPMg) or 1 mMNi2SO4 (HBSNi). Regeneration was achieved using 10 mM sodium acetatepH 4, 1M NaCl. Data were evaluated using Biaevaluation 4.1 (BIAcore3000) or Biaevaluation 1.1 software (BIAcore T100). Affinity of mAb forC3b was measured by immobilizing the mAb (335–400 resonance units[RU]) and flowing C3b in HBSP at concentrations indicated. Data wereanalyzed using the 1:1 Langmuir binding model. C3b was denselyimmobilized (1200 RU) and mAb flowed to test avidity effects.

mAb (20 mg/ml) was flowed over a C3b surface (1200 RU) in HBSPMg,followed by fB (70 mg/ml), to test whether mAb 3E7 or H17 blocked fBbinding to C3b. mAbs (varying concentrations as indicated) were flowedover the C3b-coated surface in HBSPMg before flowing fB (140 mg/ml)and fD (1 mg/ml), to test whether mAb blocked convertase formation. fB(140 mg/ml) and fD (1 mg/ml) were flowed across the C3b surface inHBSPMg, followed by immediate injection of mAb (10 mg/ml), to in-vestigate whether mAb bound preformed convertase. Convertase on thesurface was decayed using sDAF (two injections), and residual mAbbinding was measured and compared with that bound to C3b in the ab-sence of convertase formation. A high concentration of fB (350 mg/ml)was flowed in HBSNi before flowing mAb (20 mg/ml) to test whether fBprevented mAb binding to C3b; binding of mAb to surface with andwithout fB was compared.

To test whether mAbs affected the capacity of chip-bound convertase tocleave C3 to C3b, we formed C3 convertase by flowing a mix containing fB(as indicated) and fD (1 mg/ml) in HBSPMg in the presence or absence ofC3NeF-containing IgG (IND4; 360 mg/ml). When C3NeF Ig was flowed,all normal (nonstabilized) convertase was removed by two injections ofsDAF (12). Either mAbs (100 mg/ml) or buffer was flowed across thesurface, followed by injection of native C3. Cleavage of C3 to nascent C3bresulted in covalent deposition on the chip surface apparent as an increasedRU in the sensorgram.

Hemolysis assay

Sheep E were sensitized by incubating 1 vol of 4% E (v/v) with 1 vol of1:4000 Amboceptor (Siemens Healthcare, Marburg, Germany) for 30 min at37˚C. Ab-coated E (EA) were washed twice in complement fixation diluent

The Journal of Immunology 4845

by guest on March 24, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

(Oxoid, Basingstoke, U.K.) and coated with C3b by incubating with serumdepleted of fB and fH, and containing the C5 inhibitor OmCI, as previ-ously described (24). C3b-coated EA (C3b-EA) at 2% in AP buffer (5 mMsodium barbitone pH 7.4, 150 mM NaCl, 7 mM MgCl2, 10 mM EGTA pH7.4) were incubated with fB (1 mg/ml) and fD (0.5 mg/ml) for 15 min at37˚C to form C3 convertase. Convertase-coated EAwere incubated (20 min,25˚C) with increasing concentrations of mAb 3E7 or fH in AP buffercontaining 20 mM EDTA. Lysis was developed by adding 50 ml fB/fHdepleted serum diluted 1:25 in PBS/20 mM EDTA. Percent lysis wascalculated as described previously.

Electron microscopy of C3b–mAb complexes

C3b (6 mg) was incubated with an approximate 3-fold molar excess of H17Fab fragments (6 mg) in 25 mM Tris HCl pH 7.4, 150 mM NaCl, and 10 mMDTT for 1 h at room temperature. C3b–Fab complex was purified by gelfiltration chromatography using a Superdex 200 PC 3.2/30 column (GEHealthcare). Fractions were analyzed by SDS-PAGE.

Immediately after gel filtration, 5–10 ml of the peak fraction containingthe C3b-H17 Fab complex was diluted to 10 mg/ml in the same buffer andadsorbed to glow-discharged, carbon-coated copper grids and stained with

0.2% uranyl formate. Grids were imaged on a JEOL JEM-1230 electronmicroscope at 100 kV. Micrographs were collected under low-dose con-ditions (∼10e2/A2 per exposure) using a 4k 3 4k TemCam-F416 camera(TVIPS) at 2.28 A/pixel. A total of 7227 images of C3b-H17 was selectedand binned to 4.56 A/pixel after contrast transfer function correction.Particles were subjected to reference-free alignment and classification, andwere refined using angular refinement methods in EMAN (25). An ab initiomodel for refinement was obtained using the random conical tilt methodapplied to those averages where the Fab was evident (Supplemental Fig. 1).The high correlation between those averages obtained after angular re-finement and the reference-free averages supported the correctness of thefinal structure. The resolution of the structure (30 A) was estimated usingthe Fourier Shell Correlation method and a 0.5 correlation coefficient. Apseudoatomic model was obtained by fitting the atomic structure of C3b(Protein Data Bank [PDB] ID 2I07) and a Fab (PDB ID 1H0D; chains Aand B) obtained from the PDB within the electron microscopy (EM) densityof C3b-H17 using Chimera (26). The top fitting solution showed 0.89 and0.93 cross-correlation coefficients between the x-ray and the EM structuresfor C3b and the Fab, respectively.

EM database

The structure of the C3b–H17 complex has been deposited in theEMDataBank database (http://www.emdatabank.org/index.html) under acces-sion number EMD-2553.

ResultsmAbs 3E7 and H17 bind with high affinity to C3b, free, and inthe AP convertase

Previous studies have shown that mAb 3E7 binds to C3b andC3(H2O). Affinity of the interaction was investigated in this studyusing SPR; 3E7 (Fig. 1A) was immobilized on the surface ofa CM5 sensor chip, and C3b was flowed across. Affinity (KD) was35 nM as determined using the Langmuir 1:1 binding model (x2 ,1.0, indicating a good fit to the model with little heterogeneity).H17 is a humanized, chimeric form of mAb 3E7 in which the

FIGURE 1. Kinetics of interaction of 3E7/H17 with C3b. (A) 3E7 or (B)

H17 was immobilized on the surface of a CM5 chip, and C3b flowed across

at indicated concentrations (1:2 serial dilution). Binding was monitored and

kinetics evaluated using the 1:1 Langmuir binding model; KD (KD) of 3E7

was 35 nM and of H17 was 28 nM (mean of two different experiments using

multiple concentrations and global fitting). (C) C3b (1200 RU) was immo-

bilized on the chip and 3E7 was flowed across to investigate how mAbs

would bind to a C3b-coated surface; avidity effects markedly alter the

binding kinetics, indicating the binding interaction that may take place

in vivo on a tissue attacked by complement and coated in C3b.

FIGURE 2. 3E7/H17 bind to the preformed convertase, but not the

proenzyme. (A) C3 convertase was formed by flowing a mix containing fB

and fD over immobilized C3b (1200 RU); 3E7/H17 was immediately

flowed across the convertase, followed by sDAF (two injections) to remove

Bb from C3b (black line). 3E7/H17 was also flowed in the absence of

preformed convertase (gray line), and the two mAb injects were aligned.

The amount of mAb bound was the same irrespective of whether con-

vertase had been preformed on the surface. (B) Binding to proenzyme was

assessed by flowing high concentration of fB in Ni2+-containing buffer (to

stabilize) across C3b to form proenzyme followed by mAb. Binding of

mAb was compared with that achieved in the absence of fB. Solid black

line represents convertase followed by mAb; gray line represents mAb on

bare C3b surface; dashed line represents convertase only to illustrate

proenzyme decay curve.

4846 THERAPEUTIC CONTROL OF COMPLEMENT ALTERNATIVE PATHWAY

by guest on March 24, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

mouse Fc region and CH1 domain is replaced with human IgG1constant regions (19). Affinity of H17 for C3b was similar to thatof the parent Ab (28 nM; Fig. 1B). The interaction was reversedand 3E7 was flowed across a high-density C3b surface (1200 RU).Affinity cannot be accurately measured in this latter orientationbecause of avidity effects and cross-linking of C3b via both Fabarms of 3E7; however, the data illustrate that 3E7 binds tightly toC3b-opsonised surfaces (Fig. 1C).To test whether 3E7/H17 bound C3b in the AP convertase,

C3bBb, binding of mAb was compared before and after con-vertase formation on the surface of the BIAcore chip. To enableaccurate quantitation of mAb binding to C3b in the convertase, weremoved Bb bound to C3b after flowing mAb using sDAF. BothmAbs bound equally well to C3b and C3bBb on the surface, dem-onstrating that mAb binding to C3b is not hampered by the presenceof Bb in the complex (Fig. 2A). High concentration of fB was flowedover the C3b surface in nickel-containing buffer (to stabilize theC3bB proenzyme complex) to investigate whether the Ba domain inintact fB interfered with mAb binding. The amount of 3E7 or H17bound to the C3b surface in the absence or presence of fB wascompared; data in Fig. 2B illustrate that intact fB blocked binding ofeither mAb.

mAbs 3E7 and H17 prevent AP C3 convertase formation byblocking binding of fB

The earlier data demonstrate that the mAbs bound C3b in thepreformed C3 convertase, but not the proenzyme, implying that themAbs and the Ba domain share a binding site on C3b. It remainedimportant to assess whether the mAbs inhibited convertase for-mation. The surface was saturated with mAbs before flowing fB toanalyze the effect of 3E7/H17 on fB binding to immobilized C3b.Binding of mAbs to C3b completely prevented subsequent at-tachment of fB and formation of the proenzyme, C3bB (Fig. 3A).Both mAbs also prevented convertase formation in a dose-dependentmanner (Fig. 3B).

mAb 3E7 inhibits lysis of target cells bearing preformed C3convertase

The capacity of 3E7 to inhibit lysis of cells targeted by the AP wasinvestigated by coating sheep E with AP convertase, C3bBb. Cellswere incubated with increasing concentrations of 3E7 before de-

veloping lysis with plasma/EDTA (depleted of fH). Data in Fig. 4demonstrate that 3E7 blocks AP-mediated lysis of E. Because oflimited amounts of material, H17 was not tested in this assay, butDiLillo (19) reported that mAbs 3E7 and H17 were equally ef-fective in blocking AP-mediated hemolysis of rabbit E.

mAbs 3E7 and H17 block C3-cleaving activity of both normaland C3NeF-stabilized AP convertase

Cleavage of C3 by the convertase was assessed in real time bySPR to test whether mAb binding affects C3 convertase activity.Convertase was formed by flowing fB and fD over immobilizedC3b, then mAb was flowed across to bind to free C3b and C3bBb.C3 convertase activity was determined by flowing the substrate,C3, over the surface to generate nascent C3b that binds the chipsurface through its exposed thioester. In the absence of mAb,convertase-generated C3b deposition was evident on the chipsurface (Fig. 5A); however, preincubation with mAb preventedC3b deposition, demonstrating that mAb blocked interaction ofthe convertase with C3. To test the influence of C3NeF, we

FIGURE 3. 3E7/H17 block proenzyme and

convertase formation. (A) C3b (1200 RU) was

immobilized on the chip surface and fB flowed

across to form proenzyme C3bB; fB was injected

as indicated (gray line). Binding of fB was mark-

edly decreased if 3E7 or H17 was first bound to the

surface (black line); the two fB injections were

aligned. (B) Various amounts of 3E7/H17 were

bound to the C3b surface (as indicated but not

shown) before flowing fB and fD; the amount of

Ab bound to the surface was controlled through

multiple injections of mAb at 10 mg/ml until the

required amount (in RU) was bound. The surface

was regenerated to baseline between cycles, and

identical concentrations of fB and fD were flowed

in each case. Convertase (C3bBb) formation was

inhibited by the presence of mAb because of

blockade of fB binding.

FIGURE 4. 3E7 inhibits convertase formed on cell surfaces. C3 con-

vertase was generated on the surface of C3b-coated sheep E by adding fB

and fD. The cells were incubated with increasing concentrations of 3E7 in

the presence of EDTA (to prevent further convertase formation). Lysis was

developed by adding NHS depleted of fB and fH in EDTA to develop

MAC. Percentage of lysis was calculated as follows: percentage lysis =

100 3 (A410 test sample 2 A410 0% control)/(A410 100% lysis 2 A4100%

control). Means and SD (n = 2) are displayed.

The Journal of Immunology 4847

by guest on March 24, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

carried out a similar experiment where C3 convertase wasformed in the presence of C3NeF-containing IgG; in this case,all normal (nonstabilized) convertase was removed using sDAF

before flowing C3. Incubation of the highly active, C3NeF-stabilized convertase with 3E7 or H17 blocked C3 cleavage(Fig. 5B).

FIGURE 5. 3E7/H17 inhibit both normal and

C3NeF-stabilized convertase by preventing C3

cleavage. (A) C3 convertase was formed on the

surface of a CM5 chip by flowing fB and fD across

immobilized C3b. Either 3E7 (gray sensorgram) or

buffer (black sensorgram) was flowed across the

surface followed by convertase substrate, C3.

Binding of mAb blocked C3b deposition by the

convertase, C3bBb. (B) C3NeF-containing C3

convertase was formed on the surface of the chip

by flowing fB and fD across immobilized C3b in

the presence of Ig-containing C3NeF. Two injec-

tions of sDAF decayed any normal convertase

(nonstabilized) and ensured that only NeF-stabi-

lized convertase remained. Either 3E7/H17 mAb

(gray sensorgram) or buffer (black sensorgram)

was flowed across the surface followed by con-

vertase substrate, C3. Only a short pulse of C3 was

passed over the surface because of the high activity

of the C3NeF-stabilized enzyme. Binding of 3E7/

H17 blocked C3b deposition.

FIGURE 6. Three-dimensional structure of C3b-H17 Fab. (A) The C3b–H17 complex was purified by gel filtration chromatography. Fractions were

analyzed by SDS-PAGE (bottom panel), revealing two peaks, the first corresponding to the complex and the second to the free excess H17. Peak fraction 15

was selected for its observation by EM. (B) EM of C3b-H17. The peak fraction was observed in the microscope revealing several views of the complex,

including side views with the typical structural features of C3b being evident, and tilted views where density for H17 could be visualized. A two-di-

mensional average of single-molecule images corresponding to each type of view is shown. Scale bar, 10 nm. (C) Structure of C3b-H17. Two views of the

pseudoatomic model of C3b-H17 with the EM density represented as a white transparent density where the atomic structures of C3b (red color; PDB ID

2I07) and Fab (blue color) have been fitted. The location of H17 on C3b clashes with the binding of the Ba fragment from fB, as determined by the location

of the three SCRs in the crystal structure of C3Bb (32). Scale bar, 3 nm.

4848 THERAPEUTIC CONTROL OF COMPLEMENT ALTERNATIVE PATHWAY

by guest on March 24, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

Structural basis for complement inhibition by mAbs 3E7 and H17

The C3b–H17 Fab complex was isolated by size-exclusion chro-matography (Fig. 6A) and analyzed by EM to determine the mo-lecular basis for the observed effects of 3E7 and H17 (SupplementalFig. 1). Images of individual molecules of the complex were col-lected, and reference-free two-dimensional averages of this data setrevealed that complexes were interacting with the support film inseveral distinct orientations (Fig. 6B). Strikingly, those views wherethe triangular shape of C3b was more evident (“side view”) revealedno apparent density that could be assigned to H17. This suggestedthat H17 projected perpendicular to C3b, and thus was not obviousin this orientation (Fig. 6B). In contrast, “tilted views” of C3b-H17revealed a density absent in EM images of C3b alone (27), corre-sponding to H17 Fab.A 30-A resolution structure of the C3b–H17 Fab complex was

obtained by angular refinement of the images collected (Fig. 6C).C3b-H17 revealed two distinct regions of density that could bereadily assigned as C3b and H17 after computational fitting of theatomic structures of C3b (PDB ID 2I07) (28) and a Fab (PDB ID1H0D) within the density of the EM structure (Fig. 6C). Thecrystal structures of C3b and Fab perfectly matched within the EMstructure (cross-correlation coefficients of 0.89 and 0.93 for C3band Fab, respectively), thus annotating every domain in thecomplex. The exact epitope mapped by the Fab could not be de-termined at this resolution, but the pseudoatomic model of C3b-H17 Fab revealed that H17 bound a site placed within the C3bmacroglobulin 6 (MG6)-MG7 region, which is exposed as a con-sequence of the C3 to C3b conformational change; this site over-lapped with the Ba binding site (Fig. 7). As predicted from thesingle-molecule images (Fig. 6B, Supplemental Fig. 1), H17 proj-ects outward and perpendicular to C3b (Fig. 6C).

DiscussionThe AP C3 convertase, C3bBb, is the key enzyme of the com-plement system, delivering tickover activation and amplificationthat are essential to roles in immune defense and waste disposal.AP activation and dysregulation also contribute to many diseaseprocesses (7); this can be caused by polymorphisms and mutationsin AP proteins or by autoantibodies against AP components (6).C3NeF, autoantibodies that bind the AP C3 convertase, cause APdysregulation by inhibiting both spontaneous and accelerateddecay (12). To date, no specific therapies to counter the effects ofC3Nef have been developed.The mAb 3E7 and the humanized derivative, H17, were previously

shown to bind C3b, iC3b, and C3(H2O), but not native C3 (19). Inthis study, we confirmed that binding to C3b was high affinity andstable. Binding of the mAb to C3b prevented the binding of fB toform the proenzyme, C3bB, and thus stopped further AP activation.The mAb also efficiently bound preformed C3bBb, implying thatthe mAb binding site overlapped the Ba-binding region on C3b.The importance of Ba binding in formation of the proenzyme hasbeen emphasized in several earlier studies (27, 29, 30). Importantly,we also found that 3E7/H17 binding to the C3 convertase abrogatedits C3 cleaving capacity and prevented further complement activa-tion. The mAbs therefore have multiple effects as inhibitors of APconvertase formation and blockers of AP convertase C3-cleavingcapacity. The compound effect is powerful inhibition of AP con-vertase formation and function, raising the possibility that the mAbscould play roles in therapy of AP-driven diseases. Although bindingof 3E7/H17 prevents fH binding (data not shown), this will have nopathogenic effect, as we show that the mAb-bound convertase isnonfunctional.C3NeF stabilize the AP convertase and are likely pathogenic

because the convertase continues to cleave C3, causing unregulated

activation of the AP. To test whether 3E7/H17 could be used astherapy in diseases associated with C3NeF, the C3NeF-stabilizedconvertase was generated using IgG isolated from the serum ofa patient with a proven high-titer pathological C3NeF. C3 wasadded as substrate to test whether bound 3E7/H17 blocked C3-cleaving activity; the mAb completely blocked C3 cleavage bythe C3NeF-stabilized enzyme. The mAb will therefore inhibit theactivity of C3NeF in two ways: first, by preventing the formation ofC3bBb by binding nascent C3b; and second, by binding to thepreformed NeF-stabilized convertase and inhibiting C3-cleavingactivity. Notably, although all C3NeF stabilize the convertaseand most prevent accelerated decay, they are heterogeneous be-tween, and possibly within, patients (12). So far, the inhibitoryeffects of 3E7/H17 have only been demonstrated for one C3NeF.It is entirely possible that some C3NeF-stabilized C3 convertaseswill be resistant to mAb inhibition; however, even in this even-tuality, the mAb will still inhibit AP activation by preventing fBbinding to C3b, and thus convertase formation.To shed further light on the mechanism by which 3E7/H17

affected convertase formation and activity, we analyzed com-plexes of C3b and a Fab fragment of H17 using high-resolution EM.The components of the complex were obvious in the images ob-tained with the Fab binding in the vicinity of the MG6 and MG7domains of C3b. The transition from C3 to C3b is accompanied bymajor structural rearrangements that include a substantial shift of

FIGURE 7. Model for H17 activity. (A) A cartoon representing a side

view of C3b shows that H17 can access C3b. (B) H17 cannot bind to C3bB

because of steric clashes with the Ba domain. (C) H17 partially clashes with

Bb in the structure of the C3bBb convertase (PDB ID 2WIN) (41), which

would impede H17 binding to the AP C3 convertase. We hypothesize that

the flexibility of the C345C-Bb module in the C3bBb convertase could

allow H17 binding by displacing the Bb region (27). C3bBb is shown with

the MG ring facing the viewer for clarity. The locations of C345C, the MG

ring, and the TED domain in each cartoon model are labeled.

The Journal of Immunology 4849

by guest on March 24, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

the MG7 domain, likely explaining the finding that the mAb bindsC3b, iC3b, and C3(H2O), but not native C3 (28). The structure ofthe proenzyme C3bB (PDB 2XWJ) shows that the three SCRs thatcomprise Ba interact with the a’NT and CUB domains of C3b andocclude binding sites for decay accelerators, fH and DAF (31, 32).These Ba-binding domains are immediately adjacent to MG6 andMG7; indeed, a’NT and MG6 form one continuous exposed re-gion that includes binding sites for fB, fH, and CR1 (28, 32).Binding of the Fab thus likely impedes subsequent binding of fBby occluding the Ba binding site and, when bound to the con-vertase, directly interferes with cleavage of substrate. Binding ofthe mAb could cause distortion of the MG ring in C3b, thuspreventing C3 binding; alternatively, it might displace the Bbserine protease domain away from the substrate (C3), thus pre-venting cleavage (Fig. 7). It is also possible that, through bindingto MG7, the Ab prevents the substrate, C3, from binding to theconvertase. The crystal structure of cobra venom factor with C5implicates this domain in intermolecular interactions, and a dis-ease-associated 2-aa deletion within the MG7 domain has beenshown to prevent substrate binding (33, 34). The binding site(s)of C3NeF on the C3 convertase has yet to be mapped; our datademonstrate that, at least for the C3NeF tested, these are distinctfrom the mAb and substrate binding domains described earlier.Ab-based therapeutics represent one of the fastest growing

sectors of the pharmaceutical industry and, with the rise of the anti-C5 mAb eculizumab, are now impacting on complement-drivendiseases (35, 36). Different diseases may require that differentsteps of the complement cascade are blocked by targeting a spe-cific component (37, 38). Eculizumab abolishes the generation ofC5a and MAC formation, thereby reducing tissue damage causedby activation of the terminal pathway. Additional strategies fortreatment include upstream inhibition of C3 activation, which canbe mediated by 3E7/H17, as well as the newly developed CR2-fHchimeric construct, TT30 (21, 39). Several other mAb targeting APcomponents (fD, fB) are in development (37), although none hasyet been tested in C3NeF-driven pathologies. Inhibiting the APconvertase carries the risk for increasing susceptibility to bacterialinfection; however, judicious use of prophylactic antibiotics, oradministration during acute flares of disease, should minimize thisrisk. The humanized mAb, H17, shown in this study to be effectivein controlling the AP C3 convertase, is an exciting new candidatefor therapy in many diseases driven by dysregulation of the C3convertase, including aHUS and PNH. The amount of therapeuticAb needed to block a neoepitope on C3b will be far less than thatrequired to block native C3, although the mAb can also bind toC3(H2O) (19), which could act as a decoy and raise the effectivedose required for blocking pathologic complement activation. Theexact dosing of Ab will depend on the nature of the disease, fluidphase versus surface, and the extent of complement dysregulationat the level of C3. The demonstration that the mAb bind a highlyactive, C3NeF-stabilized convertase and inhibit its capacity tocleave C3 opens the possibility that the mAb could provide the firstspecific therapy for DDD, a disease that currently lacks effectivetreatment options; 80% of DDD cases are associated with presenceof C3NeF. The mAbs are specific for the human C3 convertase, andthus cannot be tested in rodent models of DDD. They do, however,inhibit AP activation in primate serum (21), raising the possibilitythat a primate model of DDD could be developed to test thetherapeutic effects of the mAb (40).

AcknowledgmentsWe thank Dr. Leslie Casey and EluSys Therapeutics for generously provid-

ing mAb H17.

DisclosuresR.P.T. owns stock in EluSys, a start-up company that is not publicly traded.

S.R.d.C. has provided consultation to Alexion Pharmaceuticals. B.P.M. has

provided consultation to Baxter. C.L.H. has an employment contract with

GlaxoSmithKline; the entire study was completed and the manuscript as-

sembled before this employment. All other authors declare no competing

financial interests.

References1. Walport, M. J. 1993. The Roche Rheumatology Prize Lecture. Complement

deficiency and disease. Br. J. Rheumatol. 32: 269–273.2. Morgan, B. P. 2000. The complement system: an overview. Methods Mol. Biol.

150: 1–13.3. Pangburn, M. K., R. D. Schreiber, and H. J. M€uller-Eberhard. 1981. Formation of

the initial C3 convertase of the alternative complement pathway. Acquisition ofC3b-like activities by spontaneous hydrolysis of the putative thioester in nativeC3. J. Exp. Med. 154: 856–867.

4. Morgan, B. P., and C. L. Harris. 1999. Complement Regulatory Proteins. Aca-demic Press, London.

5. Morgan, B. P., and S. Meri. 1994. Membrane proteins that protect againstcomplement lysis. Springer Semin. Immunopathol. 15: 369–396.

6. de Cordoba, S. R., A. Tortajada, C. L. Harris, and B. P. Morgan. 2012. Com-plement dysregulation and disease: from genes and proteins to diagnostics anddrugs. Immunobiology 217: 1034–1046.

7. Holers, V. M. 2008. The spectrum of complement alternative pathway-mediateddiseases. Immunol. Rev. 223: 300–316.

8. Pickering, M. C., and H. T. Cook. 2008. Translational mini-review series oncomplement factor H: renal diseases associated with complement factor H: novelinsights from humans and animals. Clin. Exp. Immunol. 151: 210–230.

9. Harris, C. L., M. Heurich, S. Rodriguez de Cordoba, and B. P. Morgan. 2012.The complotype: dictating risk for inflammation and infection. Trends Immunol.33: 513–521.

10. Spitzer, R. E., E. H. Vallota, J. Forristal, E. Sudora, A. Stitzel, N. C. Davis, andC. D. West. 1969. Serum C’3 lytic system in patients with glomerulonephritis.Science 164: 436–437.

11. Smith, R. J. H., C. L. Harris, and M. C. Pickering. 2011. Dense deposit disease.Mol. Immunol. 48: 1604–1610.

12. Paixao-Cavalcante, D., M. Lopez-Trascasa, L. Skattum, P. C. Giclas,T. H. Goodship, S. R. de Cordoba, L. Truedsson, B. P. Morgan, and C. L. Harris.2012. Sensitive and specific assays for C3 nephritic factors clarify mechanismsunderlying complement dysregulation. Kidney Int. 82: 1084–1092.

13. Weisman, H. F., T. Bartow, M. K. Leppo, M. P. Boyle, H. C. Marsh, Jr.,G. R. Carson, K. H. Roux, M. L. Weisfeldt, and D. T. Fearon. 1990.Recombinant soluble CR1 suppressed complement activation, inflammation, andnecrosis associated with reperfusion of ischemic myocardium. Trans. Assoc. Am.Physicians 103: 64–72.

14. Kaplan, M. 2002. Eculizumab (Alexion). Curr. Opin. Investig. Drugs 3: 1017–1023.

15. Thurman, J. M., D. M. Kraus, G. Girardi, D. Hourcade, H. J. Kang, P. A. Royer,L. M. Mitchell, P. C. Giclas, J. Salmon, G. Gilkeson, and V. M. Holers. 2005. Anovel inhibitor of the alternative complement pathway prevents antiphospholipidantibody-induced pregnancy loss in mice. Mol. Immunol. 42: 87–97.

16. Sahu, A., B. K. Kay, and J. D. Lambris. 1996. Inhibition of human complementby a C3-binding peptide isolated from a phage-displayed random peptide library.J. Immunol. 157: 884–891.

17. Ricklin, D., and J. D. Lambris. 2008. Compstatin: a complement inhibitor on itsway to clinical application. Adv. Exp. Med. Biol. 632: 273–292.

18. Paixao-Cavalcante, D., S. Hanson, M. Botto, H. T. Cook, and M. C. Pickering.2009. Factor H facilitates the clearance of GBM bound iC3b by controlling C3activation in fluid phase. Mol. Immunol. 46: 1942–1950.

19. DiLillo, D. J., A. W. Pawluczkowycz, W. Peng, A. D. Kennedy, P. V. Beum,M. A. Lindorfer, and R. P. Taylor. 2006. Selective and efficient inhibition of thealternative pathway of complement by a mAb that recognizes C3b/iC3b. Mol.Immunol. 43: 1010–1019.

20. Pawluczkowycz, A. W., M. A. Lindorfer, J. N. Waitumbi, and R. P. Taylor. 2007.Hematin promotes complement alternative pathway-mediated deposition of C3activation fragments on human erythrocytes: potential implications for thepathogenesis of anemia in malaria. J. Immunol. 179: 5543–5552.

21. Lindorfer, M. A., A. W. Pawluczkowycz, E. M. Peek, K. Hickman, R. P. Taylor,and C. J. Parker. 2010. A novel approach to preventing the hemolysis of par-oxysmal nocturnal hemoglobinuria: both complement-mediated cytolysis and C3deposition are blocked by a monoclonal antibody specific for the alternativepathway of complement. Blood 115: 2283–2291.

22. Harrison, R. A. 1996. Purification, assay and characterization of complementproteins from plasma. In Weir’s Handbook of Experimental Immunology. Vol. II.Cell Surface and Messenger Molecules of the Immune System. L. A. Herzenberg,D. M. Weir, L. A. Herzenberg, and C. Blackwell, eds. Blackwell Science,Cambridge, MA, p. 75.71–75.50.

23. Heurich, M., R. Martınez-Barricarte, N. J. Francis, D. L. Roberts, S. Rodrıguezde Cordoba, B. P. Morgan, and C. L. Harris. 2011. Common polymorphisms inC3, factor B, and factor H collaborate to determine systemic complement ac-tivity and disease risk. Proc. Natl. Acad. Sci. USA 108: 8761–8766.

24. Tortajada, A., T. Montes, R. Martınez-Barricarte, B. P. Morgan, C. L. Harris, andS. R. de Cordoba. 2009. The disease-protective complement factor H allotypic

4850 THERAPEUTIC CONTROL OF COMPLEMENT ALTERNATIVE PATHWAY

by guest on March 24, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

variant Ile62 shows increased binding affinity for C3b and enhanced cofactoractivity. Hum. Mol. Genet. 18: 3452–3461.

25. Ludtke, S. J. 2010. 3-D structures of macromolecules using single-particleanalysis in EMAN. Methods Mol. Biol. 673: 157–173.

26. Goddard, T. D., C. C. Huang, and T. E. Ferrin. 2007. Visualizing density mapswith UCSF Chimera. J. Struct. Biol. 157: 281–287.

27. Torreira, E., A. Tortajada, T. Montes, S. Rodrıguez de Cordoba, and O. Llorca.2009. 3D structure of the C3bB complex provides insights into the activation andregulation of the complement alternative pathway convertase. Proc. Natl. Acad.Sci. USA 106: 882–887.

28. Janssen, B. J., A. Christodoulidou, A. McCarthy, J. D. Lambris, and P. Gros.2006. Structure of C3b reveals conformational changes that underlie comple-ment activity. Nature 444: 213–216.

29. Hourcade, D. E., L. M. Wagner, and T. J. Oglesby. 1995. Analysis of the shortconsensus repeats of human complement factor B by site-directed mutagenesis.J. Biol. Chem. 270: 19716–19722.

30. Pryzdial, E. L., and D. E. Isenman. 1987. Alternative complement pathway ac-tivation fragment Ba binds to C3b. Evidence that formation of the factor B-C3bcomplex involves two discrete points of contact. J. Biol. Chem. 262: 1519–1525.

31. Torreira, E., A. Tortajada, T. Montes, S. Rodriguez de Cordoba, and O. Llorca.2009. 3D structure of the C3bB complex provides insights into the activation andregulation of the complement alternative pathway convertase. Proc. Natl. Acad.Sci. U S A. 106: 882–887.

32. Forneris, F., D. Ricklin, J. Wu, A. Tzekou, R. S. Wallace, J. D. Lambris, andP. Gros. 2010. Structures of C3b in complex with factors B and D give insightinto complement convertase formation. Science 330: 1816–1820.

33. Martınez-Barricarte, R., M. Heurich, F. Valdes-Canedo, E. Vazquez-Martul,E. Torreira, T. Montes, A. Tortajada, S. Pinto, M. Lopez-Trascasa, B. P. Morgan,et al. 2010. Human C3 mutation reveals a mechanism of dense deposit disease

pathogenesis and provides insights into complement activation and regulation. J.Clin. Invest. 120: 3702–3712.

34. Laursen, N. S., K. R. Andersen, I. Braren, E. Spillner, L. Sottrup-Jensen, andG. R. Andersen. 2011. Substrate recognition by complement convertases revealedin the C5-cobra venom factor complex. EMBO J. 30: 606–616.

35. Zuber, J., F. Fakhouri, L. T. Roumenina, C. Loirat, and V. Fremeaux-Bacchi;French Study Group for aHUS/C3G. 2012. Use of eculizumab for atypicalhaemolytic uraemic syndrome and C3 glomerulopathies. Nat. Rev. Nephrol. 8:643–657.

36. Hill, A., P. Hillmen, S. J. Richards, D. Elebute, J. C. Marsh, J. Chan,C. F. Mojcik, and R. P. Rother. 2005. Sustained response and long-term safetyof eculizumab in paroxysmal nocturnal hemoglobinuria. Blood. 106: 2559–2565.

37. Ricklin, D., and J. D. Lambris. 2007. Complement-targeted therapeutics. Nat.Biotechnol. 25: 1265–1275.

38. Ricklin, D., and J. D. Lambris. 2013. Complement in immune and inflammatorydisorders: therapeutic interventions. J. Immunol. 190: 3839–3847.

39. Risitano, A. M., R. Notaro, C. Pascariello, M. Sica, L. del Vecchio,C. J. Horvath, M. Fridkis-Hareli, C. Selleri, M. A. Lindorfer, R. P. Taylor, et al.2012. The complement receptor 2/factor H fusion protein TT30 protects par-oxysmal nocturnal hemoglobinuria erythrocytes from complement-mediatedhemolysis and C3 fragment. Blood 119: 6307–6316.

40. Gozalo, A. S., L. I. Cheng, M. E. St Claire, J. M. Ward, and W. R. Elkins. 2008.Pathology of captive moustached tamarins (Saguinus mystax). Comp. Med. 58:188–195.

41. Rooijakkers, S. H., J. Wu, M. Ruyken, R. van Domselaar, K. L. Planken,A. Tzekou, D. Ricklin, J. D. Lambris, B. J. Janssen, J. A. van Strijp, and P. Gros.2009. Structural and functional implications of the alternative complementpathway C3 convertase stabilized by a staphylococcal inhibitor. Nat. Immunol.10: 721–727.

The Journal of Immunology 4851

by guest on March 24, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from