Paclitaxel induces calcium oscillations via an inositol1,4,5-trisphosphate receptor and neuronal calciumsensor 1-dependent mechanismWolfgang Boehmerle*†, Ute Splittgerber‡, Michael B. Lazarus‡, Kathleen M. McKenzie‡, David G. Johnston*,David J. Austin‡, and Barbara E. Ehrlich*§

Departments of *Pharmacology and ‡Chemistry, Yale University, New Haven, CT 06520; and †Neuroscience Research Centre,Charite Universitaetsmedizin Berlin, 10117, Berlin, Germany

Edited by Solomon H. Snyder, Johns Hopkins University School of Medicine, Baltimore, MD, and approved October 4, 2006 (received for reviewAugust 23, 2006)

Taxol (Paclitaxel) is an important natural product for the treatmentof solid tumors. Despite a well documented tubulin-stabilizingeffect, many side effects of taxol therapy cannot be explained bycytoskeletal mechanisms. In the present study submicromolar con-centrations of taxol, mimicking concentrations found in patients,induced cytosolic calcium (Ca2�) oscillations in a human neuronalcell line. These oscillations were independent of extracellular andmitochondrial Ca2� but dependent on intact signaling via thephosphoinositide signaling pathway. We identified a taxol bindingprotein, neuronal Ca2� sensor 1 (NCS-1), a Ca2� binding proteinthat interacts with the inositol 1,4,5-trisphosphate receptor from ahuman brain cDNA phage display library. Taxol increased bindingof NCS-1 to the inositol 1,4,5-trisphosphate receptor. Short hairpinRNA-mediated knockdown of NCS-1 in the same cell line abrogatedthe response to taxol but not to other agonists stimulating thephosphoinositide signaling pathway. These findings are importantfor studies involving taxol as a research tool in cell biology and mayhelp to devise new strategies for the management of side effectsinduced by taxol therapy.

calcium imaging � calcium release � display cloning � drug-induced sideeffects � hypersensitivity reactions

Taxol (Paclitaxel) is among the most commonly used che-motherapeutic drugs in clinical practice for treatment of

ovarian and breast cancer, non-small-cell lung carcinoma, andAIDS-related Kaposi’s sarcoma, as well as skin, head, and neckcancers (reviewed in ref. 1). First reports of the antitumoractivity of an isolate from the bark of the pacific yew Taxusbrevifolia date back to 1964, but taxol’s main mechanism ofaction, the promotion of tubulin polymerization and stabili-zation of existing microtubules, was not discovered until 1979(2). Despite its well developed mode of action in microtubulestabilization (2), there is evidence that taxol exhibits non-microtubule-associated biological functions (3). Only a fewtaxol binding proteins other than tubulin have been found.Several heat shock proteins identified in macrophage celllysates (4) and the antiapoptotic protein Bcl-2 have beenshown to interact with taxol. The interaction with BCl-2 maybe important for the proapoptotic effects of taxol (5). Taxolfrequently induces additional side effects such as acute hyper-sensitivity reactions, cardiac conduction disturbances, andneurosensory symptoms (6). The etiology of these potentiallydose- and therapy-limiting side effects is still poorly under-stood and difficult to explain with the known taxol interactionpartners.

One interesting suggestion for understanding these side ef-fects comes from the observation that taxol exerts effects oncytosolic Ca2� signaling. When concentrations of 8,500 ng�ml(10 �M) taxol were applied, opening of the mitochondrialpermeability transition pore was observed (7, 8). One caveat ofthese studies is that high concentrations of taxol were used,

whereas in most clinical applications even the maximum plasmaconcentration does not exceed 3,600 ng�ml (4.3 �mol�liter) (9),and steady-state plasma concentrations are even lower withreported values between 85 and 850 ng�ml (10).

In this study we aimed to determine whether much lowerconcentrations of taxol could alter cytosolic Ca2� signaling and,if so, to characterize the involved pathways. We found that taxolin submicromolar concentrations induced oscillatory changes incytosolic Ca2� in an inositol 1,4,5-trisphosphate receptor(InsP3R)-dependent manner. Because there is no direct inter-action between tubulin and the InsP3R, we used a C-7 biotin-ylated taxol probe and a display cloning procedure (11) toinvestigate the possibility of non-tubulin taxol binding proteins.We have cloned a binding partner from a T7 bacteriophagehuman brain cDNA library. The isolated protein has beenidentified as neuronal Ca2� sensor 1 (NCS-1), which is a memberof a family of Ca2� binding proteins (12) and has recently beenshown to modulate InsP3R-dependent Ca2� signaling (13). In-triguingly, taxol increased binding of NCS-1 to the InsP3R andshort hairpin RNA (shRNA)-mediated knockdown of NCS-1abrogated taxol-induced Ca2� oscillations. These findings sug-gest the need for caution when using taxol in cell biologicalstudies where it is often added for microtubule stabilization andvisualization. Furthermore, these findings introduce a pathwayfor the understanding of side effects specific to taxol therapy andmay contribute to the future development of more effectivederivatives.

ResultsEffects of Low Taxol Concentrations on Intracellular Ca2� in HumanNeuroblastoma Cells. We monitored intracellular Ca2� changes inthe human neuroblastoma cell line SH-SY5Y with the fluores-cent dye Fluo-4�AM. The investigations reported here usedtaxol concentrations mimicking steady-state concentrations ob-served in patients (10). Addition of taxol at a concentration of800 ng�ml (937 nM) evoked an increase in intracellular Ca2�,typically within the first 40 seconds after bath application. Thisinitial increase was followed by subsequent Ca2� increases, thuscreating an oscillatory pattern (Fig. 1A).

Author contributions: W.B., D.J.A., and B.E.E. designed research; W.B., U.S., M.B.L., K.M.M.,and D.G.J., performed research; W.B., U.S., M.B.L., and K.M.M. analyzed data; and W.B.,B.E.E., and D.J.A. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS direct submission.

Abbreviations: ER, endoplasmic reticulum; InsP3R, inositol 1,4,5-trisphosphate receptor;NCS-1, neuronal Ca2� sensor 1; RyR, ryanodine receptor; 2-APB, 2-aminoethoxydiphenyl-borate; shRNA, short hairpin RNA; PNP, peripheral neuropathy.

§To whom correspondence should be addressed at: Department of Pharmacology, YaleUniversity School of Medicine, 333 Cedar Street, New Haven, CT 06520-8066. E-mail:barbara.ehrlich@yale.edu.

© 2006 by The National Academy of Sciences of the USA

18356–18361 � PNAS � November 28, 2006 � vol. 103 � no. 48 www.pnas.org�cgi�doi�10.1073�pnas.0607240103

Dow

nloa

ded

by g

uest

on

May

22,

202

0

Because taxol was dissolved in a 1:1 mixture of cremophore ELand absolute ethanol, a formulation that has been suggested tohave biological effects of its own (reviewed in ref. 14), we alsotested the effects of the vehicle at the same concentration (0.4�l�ml) as that used in the experiments testing the effect of taxolon Ca2� signals. In the taxol group �50% of all cells oscillatedupon treatment, which was significantly different from thevehicle group. Likewise, more cells responded with transients totaxol treatment compared with vehicle-treated cells, with the

majority of cells not responding to vehicle alone (Table 1).Spontaneous activity of untreated cells [173�5] was not signifi-cantly different from vehicle treatment in any of the threecategories, confirming that vehicle at the concentrations used inthe present study does not have an effect on intracellular Ca2�

signaling.Next, we tested the effect of a range of taxol concentrations.

At the lowest concentrations of 0.8 ng�ml and 8 ng�ml, oscilla-tions were not observed (0.6 � 0.6% [167�4] oscillating at 0.8ng�ml, and 0.5% � 0.5% [147�4] at 8 ng�ml). After addition of80 ng�ml, 30 � 10% [318�6] of the entire population of cellsoscillated, a trend that continued at 200 ng�ml (39 � 13%[199�5]) (P � 0.05). Application of 1,600 ng�ml taxol inducedoscillations in 53 � 10% [172�6] of all cells, which was notsignificantly different from treatment with 800 ng�ml, indicatingthat the response saturates when �50% of all cells oscillate. Byusing a sigmoidal fit to the data, the calculated EC50 for evokingan oscillatory Ca2� response was 83 ng�ml taxol (Fig. 1B).

To measure the regularity of the taxol-induced Ca2� oscilla-tions we used power spectral analysis as described previously(15). We found that the Ca2� oscillations could be describedadequately with one major peak (example in Fig. 1C), indicatinga regular oscillatory response. The average oscillation frequencywas 12.8 � 0.9 mHz (n � 191 cells).

Neither Extracellular nor Mitochondrial Ca2� Is Required for Taxol-Induced Oscillations. To determine the contribution of extracel-lular Ca2� to the taxol-induced oscillations, cells were observedin a Ca2�-free solution (0 Ca2� plus 10 mM EGTA added to theextracellular solution). This treatment did not abolish the initialresponse, but there was a slight, yet not significant, reduction inthe percentage of cells producing an oscillatory response to 800ng�ml taxol (Table 1). However, power spectral analysis revealeda significantly (P � 0.01) reduced mean oscillation frequency of7.2 � 0.7 mHz compared with measurements when extracellularCa2� was present. We also observed in several experiments adecreasing amplitude of the response (Fig. 1D), which, togetherwith the reduced oscillatory frequency, suggests that, althoughextracellular Ca2� is not necessary for the initial response, it isrequired to maintain the response, presumably to prevent de-pletion of intracellular Ca2� stores.

Because previous studies have shown a direct effect of hightaxol concentrations on mitochondrial permeability (7, 8), it wasimportant to determine whether this observation would beobtained with the much lower concentrations used in this study.Cells were preincubated for �5 min in 5 �M cyclosporin Abefore imaging, a treatment previously shown to abrogate mi-tochondrial permeability increases in response to 10 �M taxol(7). Addition of 800 ng�ml taxol to cyclosporin A pretreated cells

Fig. 1. Taxol induces Ca2� oscillations independent of extracellular andmitochondrial Ca2�. (A) Representative normalized Ca2� changes induced by800 ng�ml taxol (arrow). (B) Taxol-induced oscillations are concentration-dependent with a calculated EC50 of 83 ng�ml (arrow). (C) Power spectralanalysis of the cell shown in A reveals that the dominant peak of taxol-inducedCa2� oscillations occurs at �12 mHz. (D–F) Ca2�-free solution (D) (10 mMEGTA), preincubation of cells with cyclosporin A (E), and treatment with themitochondrial uncoupler FCCP (F) did not block oscillations induced by 800ng�ml taxol (arrow). A, 10 �M ATP; T, 3 �M thapsigargin.

Table 1. Percentage of cells responding to various treatments as described in Results

Treatment Nonresponding % Transient % Oscillation % [n�N]

Vehicle 77.2 � 8.5 19.6 � 7.7 3.2 � 1.7 [169�5]800 ng�ml taxol 11.1 � 3.6** 36.4 � 2.5 (NS) 52.6 � 3.5** [357�7]Ca2�-free � 800 ng�ml taxol 17 � 12.9** 49 � 5.8 (NS) 34 � 9.3* [232�5]5 �M cyclosporin A � 800 ng�ml taxol 2.2 � 1.8** 42.8 � 14.4 (NS) 55.1 � 15.3** [222�5]1 �M FCCP � 800 ng�ml taxol 12.5 � 6.8** 40.4 � 7.8 (NS) 47.1 � 12.6** [219�5]10 �M thapsigargin � 800 ng�ml taxol 99 � 1 (NS) 1 � 1 (NS) 0 (NS) [210�5]70 �M dantrolene � 800 ng�ml taxol 13 � 4.6** 63.5 � 10.4** 23.6 � 7.1 (NS) [120�4]5 �M xestospongin C � 800 ng�ml taxol 92.8 � 3 (NS) 6.8 � 2.8 (NS) 0.3 � 0.3 (NS) [216�5]20 �M 2-APB � 800 ng�ml taxol 98.2 � 0.9 (NS) 1.8 � 0.9 (NS) 0 (NS) [269�6]5 �M U73122 � 800 ng�ml taxol 87.3 � 4.4 (NS) 11.5 � 5.1 (NS) 1.1 � 1.1 (NS) [97�4]Scrambled shRNA � 800 ng�ml taxol 3.7 � 2.8** 49 � 14.5 (NS) 47.3 � 14.4** [108�5]NCS-1 shRNA � 800 ng�ml taxol 71.7 � 9.3 (NS) 23 � 6.4 (NS) 5.3 � 3.4 (NS) [77�5]

Statistics: ANOVA (Dunnett) compared with vehicle. NS, not significant; *, P � 0.05; **, P � 0.01.

Boehmerle et al. PNAS � November 28, 2006 � vol. 103 � no. 48 � 18357

PHYS

IOLO

GY

Dow

nloa

ded

by g

uest

on

May

22,

202

0

induced oscillations in 55% of all cells, which was significantlydifferent from vehicle stimulation (Fig. 1E) but similar to cellsnot treated with cyclosporin A (Table 1). Furthermore, theability to alter the Ca2� oscillations with complete depolariza-tion of the mitochondrial membrane potential �m was tested.The addition of the mitochondrial uncoupler FCCP (1 �M) ledto a fast and complete release of mitochondrial Ca2� (first peakin Fig. 1F), and further addition of 10 �M FCCP did not produceany additional effects on cytosolic Ca2�. Subsequent treatmentwith 800 ng�ml taxol still induced cytosolic Ca2� oscillations(Fig. 1F), which were significantly different from the response tovehicle (Table 1).

These findings were supported by direct observation of themitochondrial membrane potential using the fluorescent poten-tiometric dye rhodamine 123. We observed a highly specificlabeling of mitochondria, which did not change when 800 ng�mltaxol was added, indicating a stable �m. In addition, there wereno changes observed in the shape of the mitochondria, whereas1 �M FCCP rapidly and completely depolarized the mitochon-dria and abolished mitochondrial labeling (Fig. 4, which ispublished as supporting information on the PNAS web site). Aninvolvement of mitochondrial permeability transition in theresponse to low concentrations of taxol thus seems unlikely.

Taxol-Induced Ca2� Oscillations Depend on Ca2� Stored in the Endo-plasmic Reticulum (ER) and the InsP3R. To further dissect themechanism of the observed taxol-induced oscillations, the Ca2�

stored in the ER was depleted with thapsigargin, an inhibitor ofthe sarcoplasmic-ER Ca2� ATPase. After the initial Ca2�

release from the ER, cytosolic Ca2� concentrations returned tobaseline levels, and addition of 800 ng�ml taxol did not elicit anyfurther Ca2� response (Fig. 2A). No cell produced an oscillatoryresponse after stimulation with taxol (Table 1). Ca2� releasefrom the ER is thus required for the induction of Ca2� oscilla-tions by low concentrations of taxol.

Ca2� can be released from the ER through the ryanodinereceptor (RyR) or InsP3R families. RyR were blocked with thecell-permeant inhibitor dantrolene. After preincubation for atleast 5 min with 70 �M dantrolene, 24% of all cells produced

an oscillatory response after stimulation with 800 ng�ml taxol(Table 1). This was different from cells stimulated with taxolin the absence of dantrolene and cells treated with vehiclealone, albeit the latter effect failed statistical significancebecause of increased variability in the response of dantrolene-treated cells. Analysis of the power spectrum density revealeda mean oscillatory frequency of 10 � 1.1 mHz, which was notsignificantly different from the mean oscillatory frequencyobserved in taxol-stimulated cells in the absence of dantrolene.We also observed a reduction of the amplitude of the Ca2�

peaks during the oscillation in dantrolene-treated cells (Fig.2B). Because RyR are important for Ca2�-induced Ca2�

release (CICR) (16), it is likely that RyR participate intaxol-induced Ca2� oscillations via CICR-mediated amplifi-cation of the signal. Because of the reduced amplitude of theCa2� signal, fewer cells met the criteria (Materials and Meth-ods) we had defined for an oscillation. This observation iscomparable to previous studies that showed that blockage ofthe RyR with dantrolene altered the shape of the response butwas not required for the initiation of the response (17).

To study the role of the InsP3R in taxol-induced Ca2� oscil-lations, cells were treated with the InsP3R inhibitors 2-amino-ethoxydiphenylborate (2-APB) or xestospongin C. Because 100�M 2-APB has also been shown to block the mitochondrialpermeability transition pore (18), we used a concentration fivetimes lower. Ca2� oscillations induced by 800 ng�ml taxol wereabolished in cells pretreated for at least 5 min with either 20 �M2-APB or 5 �M xestospongin C. In some cells a small transientresponse was observed, but most cells did not respond at all tostimulation with taxol (Fig. 2C and Table 1). This observationstrongly supports the conclusion that the InsP3R is required forthe oscillatory response of cytosolic Ca2� to submicromolar taxolconcentrations.

To determine whether taxol is able to activate the InsP3R inthe absence of InsP3 or whether it modulates the response toInsP3, we blocked the formation of InsP3 with the cell-permeantphospholipase C inhibitor U-73122. Incubation for 7 min in 5 �MU-73122 abrogated the response (Fig. 2D and Table 1). Theseresults demonstrate that the oscillatory response to low taxolconcentrations depends on phospholipase C activity, as well asactivation of the InsP3R.

Screening and Identification of NCS-1 with a Biotinylated Taxol Probe.Because taxol mainly interacts with tubulin and the observedoscillatory Ca2� response depends on the InsP3R it was impor-tant to know whether tubulin interacts with this receptor andwhether addition of 800 ng�ml taxol would cause any change inthe interaction. However, no indication was found for an asso-ciation of these proteins when using coimmunoprecipitationfrom cerebellar lysate. Only the protein directly associated withthe immunoprecipitating antibody could be detected regardlessof the presence of taxol (Fig. 3A).

Because known taxol interaction partners could not explainour observations and several taxol-specific side effects involvethe peripheral nervous system (6) we screened a human brain T7cDNA phage display library for taxol binding using C-7-biotin-ylated-taxol (7-bio-taxol) and a modified display cloning proce-dure (11). Each selection round in display cloning involved threedistinct steps: (i) binding of the phage to the solid support, (ii)washing away unbound phage, and (iii) eluting bound phageparticles, which were used for the next round of selection. Afterthe first round of affinity selection, 16 clones were selected andanalyzed by PCR amplification of the insert region and DNAsequencing. None of the 16 clones contained an expressible gene.After the second round of affinity selection, 3 of 16 clones werefound to encode the full ORF of NCS-1 (Fig. 5A, which ispublished as supporting information on the PNAS web site) (19).The coding sequence was in-frame with the coding sequence of

Fig. 2. Taxol-induced Ca2� oscillations depend on ER Ca2�, InsP3R, and InsP3

but not RyR. (A) Taxol-induced Ca2� responses are abolished after depletionof ER–Ca2� with thapsigargin. (B) Preincubation with 70 �M dantrolenedecreases the response amplitude to 800 ng�ml taxol (arrow). (C and D)Preincubation with the InsP3R inhibitors xestospongin C (5 �M; black trace) or2-APB (20 �M; gray trace) (C) and treatment with the phospholipase C inhib-itor U-73122 (5 �M) (D) abolish the response to 800 ng�ml taxol (arrow). A, 10�M ATP; T, 3 �M thapsigargin.

18358 � www.pnas.org�cgi�doi�10.1073�pnas.0607240103 Boehmerle et al.

Dow

nloa

ded

by g

uest

on

May

22,

202

0

cp10, the T7 coat protein, as expected for a properly displayedprotein. After the third round of selection 8 of 16 clonescontained NCS-1, with four clones not having inserts at all. Thepredominance of NCS-1 in the rescued clones strongly suggestedthat NCS-1 was being affinity-amplified by 7-bio-taxol. In addi-tion, the absence of additional ORFs, as seen in other selections(20) and the appearance of phage clones not containing any geneinserts, suggested that NCS-1 was the primary binding proteinfor taxol in this cDNA library. When using biotinylated phorbolester and identical selection procedures, none of the isolatedclones were NCS-1, even after five rounds of selection.

To validate the affinity of 7-bio-taxol for the NCS-1 protein,an on-phage concentration-dependent binding study was per-formed. The phage solutions were titered and plotted as afunction of probe concentration for each well. Nonlinearregression analysis of the data gave an EC50 value of 728 � 44ng�ml (557 � 34 nM) (Fig. 3B). This result is intriguingbecause the concentration is within the range observed inpatients (9) and, as outlined above, sufficient to induce Ca2�

oscillations. Crystal structure analysis of human NCS-1 hasshown that, upon treatment with Ca2�, three of its four EFmotifs bind the metal, causing a large conformational shift inthe C-terminal portion of the protein and exposing a largehydrophobic cleft, which is postulated to serve as a proteindocking site that forms in response to cellular Ca2� f lux (19).It is possible that this region of the protein also serves as thebinding site for the taxol probe molecule. Given the large sizeof this hydrophobic pocket, we evaluated four other biotin-ylated natural products, as well as a molecule that matches thelinker portion of 7-bio-taxol for their capability to bind NCS-1phage. None of the additional probe molecules, bio-FK506,bio-phorbol ester, biof lavokavin A, biobengamide, and thebiotin-LC-LC linker, exhibited any appreciable binding toNCS-1 phage (Fig. 5B). Together these results indicate thatbio-taxol’s affinity for the NCS-1-expressing phage clone isspecific and not due to nonspecific hydrophobic interaction.

NCS-1 Knockdown Abrogates Taxol-Induced Ca2� Oscillations. Be-cause taxol binds to NCS-1 but not to the InsP3R and NCS-1 iscapable of positively modulating InsP3-mediated Ca2� release(13), we hypothesized that the observed oscillatory response tosubmicromolar concentrations of taxol might be due to analteration of the NCS-1–InsP3R interaction. To study whetherthe binding of NCS-1 to the InsP3R is altered in the presence of800 ng�ml taxol, NCS-1 was coimmunoprecipitated with InsP3Rfrom cerebellar lysate. The amount of NCS-1 bound to theInsP3R was increased in the presence of taxol (Fig. 3C).

To further establish the importance of the NCS-1–InsP3Rinteraction for the observed effects, NCS-1 was knocked downby transient transfection with a vector coexpressing anti NCS-1shRNA and GFP. The coexpression with GFP allowed identi-fication of those cells expressing the NCS-1 shRNA, which wouldcontain less NCS-1 protein. Ca2� transients were monitored incells by using the red fluorescent dye Rhod-2�AM. Expressionof NCS-1 shRNA resulted in a reduction in the immunosignal forNCS-1 by �80% (n � 3 independent experiments; P � 0.001)(Fig. 3D).

Interestingly, the oscillatory Ca2� response to 800 ng�ml taxolwas abrogated in NCS-1 knockdown cells compared with cellstransfected with a vector expressing a scrambled shRNA se-quence that does not target any known gene (Table 1). Therewere no responses to taxol stimulation in 72% of NCS-1 knock-down cells compared with 4% (P � 0.001) of cells transfectedwith scrambled shRNA (Fig. 3E). The responses of the NCS-1knockdown cells were not significantly different from vehicletreatment (Table 1). These results indicate that NCS-1 is causallyinvolved in the Ca2� response to low taxol concentrations.

Because NCS-1 positively modulates the InsP3R-mediatedCa2� release (13), we tested whether NCS-1 knockdown cellswere still able to oscillate in response to stimulation with lowconcentrations of ATP, a robust activator of InsP3-dependentCa2� oscillations. Induction of Ca2� oscillations was not affectedin NCS-1 knockdown cells with 42 � 7% [70�4] of all cellsoscillating upon stimulation with 0.75 �M ATP compared with38 � 13% [64�4] (not significant) of cells transfected withscrambled shRNA (Fig. 3F). This result supports the suggestionthat the observed abrogation of taxol-induced Ca2� oscillationsin NCS-1 knockdown cells is taxol-specific and not a nonspecificside effect of the NCS-1 knockdown.

DiscussionThe aim of this study was to investigate the effects of submicro-molar concentrations of taxol, as found in patients undergoingchemotherapy for solid tumors (9, 14), on the Ca2� homeostasisof a neuronal cell line. We found a rapid induction of cytosolicCa2� oscillations, which, at least initially, did not depend on Ca2�

f luxes from the extracellular space or on the release of mito-

Fig. 3. Taxol binds to NCS-1, an interaction that enhances binding of NCS-1to InsP3R. (A) Coimmunoprecipitation of �-tubulin and InsP3R from mousecerebellar lysate. Lanes, from left to right, show mouse cerebellar lysate, beadstreated with preimmune serum but no specific antibody, immunoprecipitatewith anti-InsP3R1, immunoprecipitate with anti-InsP3R1 and 800 ng�ml taxol,immunoprecipitate with anti-�-tubulin, and immunoprecipitate with anti-�-tubulin and 800 ng�ml taxol. The immunoblot was probed with anti-InsP3R inUpper and anti-�-tubulin in Lower. (B) Binding analysis of 7-bio-taxol 4 withNCS-1 phage. Phage rescue titer is reported in pfu as a function of probeconcentration. (C) Taxol increases the binding of NCS-1 to the InsP3R. Lanes,from left to right, show mouse cerebellar lysate, beads treated with preim-mune serum, immunoprecipitate with anti-InsP3R, and immunoprecipitatewith anti-InsP3R and 800 ng�ml taxol. The immunoblot was probed withanti-InsP3R1 in Upper and anti-NCS-1 in Lower. (D) Cells transiently transfectedwith a vector expressing NCS-1 shRNA as well as GFP showed a significantreduction (�80%) of the immunosignal compared with cells expressing scram-bled shRNA. (E) Oscillations induced by 800 ng�ml taxol (arrow) were abro-gated in NCS-1 knockdown cells (red trace) but unaffected in cells expressingscrambled shRNA (black trace). (F) Oscillations induced by 0.75 �M ATP wereunaffected by NCS-1 knockdown (red trace) compared with scrambled shRNA-expressing cells (black trace). T, 3 �M thapsigargin.

Boehmerle et al. PNAS � November 28, 2006 � vol. 103 � no. 48 � 18359

PHYS

IOLO

GY

Dow

nloa

ded

by g

uest

on

May

22,

202

0

chondrial Ca2�. A further dissection of the molecular basis ofthis signal revealed a dependence on Ca2� release from the ERand, more specifically, the InsP3R. Release of Ca2� through theRyR does not seem to be necessary for the initiation of the Ca2�

transient under these conditions, but activation of the RyR isused for amplification of the Ca2� signal. Because known taxolbinding proteins could not explain this effect, we used a humanbrain cDNA phage display library and affinity chromatographywith a biotinylated taxol probe to screen for new interactionpartners of taxol. By using this procedure, a phage clone thatencodes the NCS-1 protein was isolated. Subsequent bindinganalyses confirmed the affinity of taxol for this protein andestablished both probe- and Ca2�-dependent binding character-istics. Intriguingly, the concentration range for the affinity ofbio-taxol for NCS-1 was well within the concentration rangeinducing Ca2� oscillations.

Recently we demonstrated that NCS-1 enhances InsP3R ac-tivity both at the single-channel level and in intact cells (13). Thisled us to hypothesize that the observed oscillatory response tosubmicromolar concentrations of taxol is due to a NCS-1-mediated increase in InsP3R activity, which should be abrogatedin cells where NCS-1 is knocked down. Indeed, we found anincreased binding of NCS-1 to the InsP3R in the presence oftaxol, and when NCS-1 was knocked down the Ca2� response wasnot significantly different from vehicle treatment. The lattereffect appears to be taxol-specific, because NCS-1 knockdowncells oscillated to the same extent as cells with normal NCS-1levels after stimulation with low concentrations of ATP. The keyfindings of this study have implications for various taxol therapy-related side effects, discussed below and outlined in a model(Fig. 6, which is published as supporting information on thePNAS web site).

One side effect, which almost led to the discontinuation ofclinical trials with taxol, is the high incidence of severe hyper-sensitivity reactions (21). This problem could be addressed witha stringent pretreatment of all patients with corticosteroids aswell as H1 and H2 antagonists (21). However, even withpretreatment, between 1% and 3% of all patients develop majorhypersensitivity reactions (6). Initially it was thought that thevehicle cremophore EL was responsible for these hypersensitiv-ity reactions (reviewed in ref. 14), but subsequent elegant in vivostudies linked this phenomenon to sensory nerve peptides suchas Substance P and calcitonin gene-related peptide (CGRP)(22). By using these newer findings it was suggested that releaseof the peptide mediators would cause mast cell activation withhistamine release. Notably, NCS-1 is highly expressed by dorsalroot ganglion cells and has been found to be colocalized withCGRP in peripheral nerve terminals innervating blood vessels(23). In light of our findings, it seems likely that binding of taxolto NCS-1 in peripheral nerves could facilitate the release ofneuropeptides in a Ca2�-dependent fashion, which can thentrigger a hypersensitivity reaction.

Another side effect with a high incidence is cardiac arrhythmia(24). This phenomenon does not seem to be linked to cytotox-icity and could be reproduced in an in vitro preparation (25).Although the InsP3R is not playing a dominant role in excit-ation–contraction coupling in cardiomyocytes, several studiessuggest a role for the InsP3R in cardiac arrythmogenesis (26).Because NCS-1 is also expressed in the heart (27), one directprediction from our observations, that taxol positively modulatesthe InsP3R in a NCS-1-dependent manner, would be a positiveinotropic effect (26); this has been observed in taxol-treatedpapillary muscles (25).

NCS-1 is also highly expressed in neuronal tissues. Becausetaxol does not seem to cross an intact blood–brain barrier (28),neuronal side effects should occur in the peripheral nervoussystem. In fact, peripheral neuropathy (PNP) is another frequentmajor dose-dependent and therapy-limiting side effect of taxol

chemotherapy that is still poorly understood (reviewed in ref.29). In the context of the present study, it is interesting to notethat inhibition of the Ca2�-activated calpain proteases has aprotective effect against taxol-induced sensory neuropathy invivo and taxol causes activation of both calpains and caspases inPC12 cells (30). Furthermore, the Ca2�-permeable nonselectivecation channel transient receptor potential vanilloid 4 (TRPV4),a receptor located in the plasma membrane, has been shown tobe essential in taxol-induced PNP in rats (31). This last obser-vation is intriguing because TRPV4 currents were shown to bepotentiated by increases in intracellular Ca2� (32). It thus seemslikely that cytosolic Ca2� oscillations in dorsal root ganglianeurons are induced by taxol binding to NCS-1 and subsequentpositive modulation of the InsP3R. These effects, in turn, arepotentiated in a TRP-dependent manner and lead to the acti-vation of Ca2� activated proteases with the result of cell mal-function and death, resulting in PNP. This model offers anexplanation for the apparent beneficial effects of Ca2� channelblockers in the treatment of PNP (29).

Taxol is also frequently used as a pharmacological tool in cellbiology, where it is used to arrest cells in the G2 phase, studymicrotubule function, and visualize the microtubule cytoskele-ton with fluorophore-conjugated derivatives. In the context ofthe present finding that taxol also affects intracellular Ca2�

signaling, care should be given when interpreting the resultsobtained with taxol as a microtubule-modifying drug.

In summary, our observations (that taxol has a new bindingpartner, NCS-1, and that binding to NCS-1 leads to initiation ofcytosolic Ca2� oscillations) suggest that taxol’s effects when usedas a research tool will be more complex than originally expectedand that our model (Fig. 6) may help to devise new strategies forthe management of side effects induced by taxol therapy.

Materials and MethodsPlasmids. The GeneClip U1 Hairpin Cloning Systems (Promega,Madison, WI) was used as vector for expression of NCS-1 orscrambled shRNA. The shRNA template for NCS-1 was GGCT-TCCAGAAGATCTACA, and as scrambled sequence for con-trols we used GGCTTCGTGAAGGTCTATA.

Cell Cultures and Transfection. The human neuroblastoma cell lineSH-SY5Y (American Type Culture Collection, Manassas, VA)was cultured and transfected as described previously (33).

Ca2� Imaging. Cells were incubated (30 min at 37°C in 5% CO2)in Hepes buffer containing either 5 �M Fluo-4�AM or 6 �MRhod-2�AM (Molecular Probes, Invitrogen, Carlsbad, CA)together with 0.1% Pluronic F-127 (Molecular Probes, Invitro-gen). The Hepes medium contained 130 mM NaCl, 4.7 mM KCl,1 mM MgSO4, 1.2 mM KH2PO4, 1.3 mM CaCl2, 20 mM Hepes,and 5 mM glucose (pH 7.4). Coverslips were mounted in achamber (Warner Instruments, Hamden, CT) and transferred toa Zeiss LSM 510 META scanning laser confocal microscopeequipped with a C-Apochromat �40�1.2 water immersion ob-jective (Zeiss, Thornwood, NY). Images were acquired at 0.33Hz. All drugs were bath-applied. To identify cells expressingshRNA with GFP, transfected cells were examined by usingfluorescence excitation at 488 nm in a multitrack configurationto minimize crosstalk. Whole-cell f luorescence was measured bydefining each cell as one region of interest. A given cell wasconsidered to oscillate when at least three Ca2� transients(deflections from �20% from baseline) were recorded over thetime monitored, usually 10 min. Cells that did not respond tostimulation and to control stimulation with either 10 �M ATP(A) or 3 �M thapsigargin (T), as indicated in Results, wereexcluded from evaluation. Inhibitors were added after the dyeincubation; times and concentrations are indicated in Results.Ca2�-induced fluorescence intensity ratio F�F0 was plotted as a

18360 � www.pnas.org�cgi�doi�10.1073�pnas.0607240103 Boehmerle et al.

Dow

nloa

ded

by g

uest

on

May

22,

202

0

function of time with F0 as an average of the first five points ofthe baseline. To perform power spectrum analysis, we used analgorithm written in MATLAB as described previously (15).

Statistical Analysis. Data are expressed as mean � SEM or asrepresentative traces. [n�N] describes the number of cells studied(n) in N independent cultures. Statistical analysis of the differ-ences between multiple groups was performed by using aone-way ANOVA (Dunnett multiple-comparisons test) (Instat;GraphPad, San Diego, CA) for two groups using t test(SigmaPlot, Systat, Richmond, CA); P � 0.05 was consideredstatistically significant.

Immunoprecipitation and Western Blot Analysis. Lysate preparation,immunoprecipitation, and immunoblotting were performed asdescribed previously (13). Antibodies used were as follows:anti-NCS-1 (Santa Cruz Biotechnology, Santa Cruz, CA), anti-�-actin (Abcam, Cambridge, MA), anti-�-tubulin (Covance,Berkeley, CA), and anti-InsP3R1 (17). For immunoprecipitation,cerebellar lysate was incubated with antibody in the presence of50 �M free Ca2� and 800 ng�ml taxol as indicated in Results.NCS-1 knockdown blots were quantified by scanning densitom-etry by using UN-SCAN-IT (Silk Scientific, Orem, UT) normal-izing NCS-1 expression to the �-actin loading control.

Display Cloning. Human brain polyA� mRNA (Clontech, Moun-tain View, CA) was used to create a cDNA phage display librarywith the OrientExpress cDNA library synthesis kit (Novagen, LaJolla, CA). Phage lysate generation for T7 cDNA libraries andindividual T7 phage clones were prepared by infection into logphase BLT5403 Escherichia coli. The oligonucleotide sequencingprimers T7Forward (5-TCTTCGCCCAGAAGCTGCAG) andT7Reverse (5-CCTCCTTTCAGCAAAAAACCCC) wereused for both PCR amplification and DNA sequence analysis ofthe T7 phage-displayed inserts. C7-bio-taxol agarose columnswere prepared by incubating a slurry of avidin agarose resin fromPierce (Rockford, IL) in PBS with C7-bio-taxol. The column wasextensively washed with wash buffer (50 mM Tris, 150 mM NaCl,

0.5 mM CaCl2, and 1 mM MgSO4), after which an aliquot of thecleared human brain T7 cDNA phage display lysate was addedto the resin. Nonspecifically bound phage was removed bywashing with wash buffer. Bound phage were eluted with 1%SDS in PBS, and the eluent was diluted with LB media (1:1).Phage rescue titer was found to be 26 � 106 pfu�ml after the firstround of selection, 54 � 106 pfu�ml after the second round, and3,000 � 106 pfu�ml after the third round. Random selection ofphage plaques and sequencing yielded multiple copies of anidentical clone, encoding NCS-1 protein. The clone TT3.5 wasselected for use in all subsequent studies.

For probe-dependent NCS-1 phage binding studies, serialdilutions of each biotinylated probe were incubated with aNeutrAvidin-coated microtiter plate (Pierce), and subsequentlya biotin block (1 mM) was performed to reduce backgroundbinding. An aliquot of the NCS-1 phage lysate was added to eachwell and incubated overnight at 4°C. Nonbinding phage wereremoved by washing, and bound phage were eluted by treatmentwith 1% SDS in TBS. The eluate was titered by serial dilutionand dropping triplicates onto bacterial plates. The resultingplaques were counted and plotted as a function of incubatedbiotinylated probe. Information about the additional methodsused for Figs. 4–6 can be found in Supporting ExperimentalProcedures, which is published as supporting information on thePNAS web site.

We thank Manuel Estrada, Per Uhlen, Anurag Varshney, Sven-EricJordt, Brenda DeGray, Felix Heidrich, and Victor K. Chung for invalu-able advice regarding the design of the experiments and thoughtfuldiscussions and comments on the manuscript. We also thank M. VenkataRami Reddy (University of California, La Jolla, CA) for the synthesis ofC7-bio-taxol and Andreas Jeromin (University of Texas, Austin, TX) andJolanta Vidugiriene (Promega, Madison, WI) for shRNA reagents. Thiswork was supported by National Institutes of Health Grants GM63496(to B.E.E.) and GM59673 (to D.J.A.), the Patterson Research Trust(B.E.E.), National Center for Complementary and Alternative MedicineSmall Business Innovation Research Phase I Grant R43 AT0 0324-01 (toAncile Pharmaceuticals), and a German National Merit Foundationscholarship (to W.B.). D.J.A. was a Fellow of the Alfred P. SloanFoundation.

1. Mekhail TM, Markman M (2002) Expert Opin Pharmacother 3:755–766.2. Schiff PB, Fant J, Horwitz SB (1979) Nature 277:665–667.3. Blagosklonny MV, Fojo T (1999) Int J Cancer 83:151–156.4. Byrd CA, Bornmann W, Erdjument-Bromage H, Tempst P, Pavletich N, Rosen

N, Nathan CF, Ding A (1999) Proc Natl Acad Sci USA 96:5645–5650.5. Rodi DJ, Janes RW, Sanganee HJ, Holton RA, Wallace BA, Makowski L

(1999) J Mol Biol 285:197–203.6. Rowinsky EK, Eisenhauer EA, Chaudhry V, Arbuck SG, Donehower RC

(1993) Semin Oncol 20:1–15.7. Kidd JF, Pilkington MF, Schell MJ, Fogarty KE, Skepper JN, Taylor CW,

Thorn P (2002) J Biol Chem 277:6504–6510.8. Mironov SL, Ivannikov MV, Johansson M (2005) J Biol Chem 280:715–721.9. Huizing MT, Keung AC, Rosing H, van der Kuij V, ten Bokkel Huininkqq WW,

Mandjes IM, Dubbelman AC, Pinedo HM, Beijnen JH (1993) J Clin Oncol11:2127–2135.

10. Rowinsky EK, Jiroutek M, Bonomi P, Johnson D, Baker SD (1999) Clin CancerRes 5:767–774.

11. McKenzie KM, Videlock EJ, Splittgerber U, Austin DJ (2004) Angew Chem IntEd Engl 43:4052–4055.

12. Burgoyne RD, Weiss JL (2001) Biochem J 353:1–12.13. Schlecker C, Boehmerle W, Jeromin A, DeGray B, Varshney A, Sharma Y,

Szigeti-Buck K, Ehrlich BE (2006) J Clin Invest 116:1668–1674.14. Gelderblom H, Verweij J, Nooter K, Sparreboom A (2001) Eur J Cancer

37:1590–1598.15. Uhlen P (2004) Sci STKE, pl15.16. Friel DD, Tsien RW (1992) J Physiol (London) 450:217–246.17. Johenning FW, Zochowski M, Conway SJ, Holmes AB, Koulen P, Ehrlich BE

(2002) J Neurosci 22:5344–5353.18. Chinopoulos C, Starkov AA, Fiskum G (2003) J Biol Chem 278:27382–27389.

19. Bourne Y, Dannenberg J, Pollmann V, Marchot P, Pongs O (2001) J Biol Chem276:11949–11955.

20. Videlock EJ, Chung VK, Mohan MA, Strok TM, Austin DJ (2004) J Am ChemSoc 126:3730–3731.

21. Weiss RB, Donehower RC, Wiernik PH, Ohnuma T, Gralla RJ, Trump DL,Baker JR, Jr, Van Echo DA, Von Hoff DD, Leyland-Jones B (1990) J ClinOncol 8:1263–1268.

22. Itoh Y, Sendo T, Hirakawa T, Goromaru T, Takasaki S, Yahata H, Nakano H,Oishi R (2004) Am J Respir Crit Care Med 169:113–119.

23. Averill S, Robson LG, Jeromin A, Priestley JV (2004) Neuroscience 123:419–427.

24. Rowinsky EK, McGuire WP, Guarnieri T, Fisherman JS, Christian MC,Donehower RC (1991) J Clin Oncol 9:1704–1712.

25. Alloatti G, Penna C, Gallo MP, Levi RC, Bombardelli E, Appendino G (1998)J Pharmacol Exp Ther 284:561–567.

26. Zima AV, Blatter LA (2004) J Physiol (London) 555:607–615.27. Nakamura TY, Sturm E, Pountney DJ, Orenzoff B, Artman M, Coetzee WA

(2003) Pediatr Res 53:554–557.28. Eiseman JL, Eddington ND, Leslie J, MacAuley C, Sentz DL, Zuhowski M,

Kujawa JM, Young D, Egorin MJ (1994) Cancer Chemother Pharmacol34:465–471.

29. Mielke S, Sparreboom A, Mross K (2006) Eur J Cancer 42:24–30.30. Wang MS, Davis AA, Culver DG, Wang Q, Powers JC, Glass JD (2004) Brain

127:671–679.31. Alessandri-Haber N, Dina OA, Yeh JJ, Parada CA, Reichling DB, Levine JD

(2004) J Neurosci 24:4444–4452.32. Strotmann R, Schultz G, Plant TD (2003) J Biol Chem 278:26541–26549.33. Estrada M, Uhlen P, Ehrlich BE (2006) J Cell Sci 119:733–743.

Boehmerle et al. PNAS � November 28, 2006 � vol. 103 � no. 48 � 18361

PHYS

IOLO

GY

Dow

nloa

ded

by g

uest

on

May

22,

202

0