TLR Adaptor Protein MYD88 Mediates Sensitivity to HDAC Inhibitors via a Cytokine-Dependent Mechanism

Abnormal epigenetic control is a common feature of tumorigenesis, and aberrant lysine acetylation is believed to take on an important role in driving the malignant phenotype (1). Lysine acetylation is regulated by two groups of enzymes, with histone acetyl-transferases (HAT) mediating the acetylation event, and histone deacetylases (HDAC) providing the deacetylation event (2). This type of relationship, often referred to as “writer, reader, and eraser”, plays an instrumental role in epigenetic control, where lysine acetylation provides a key regulatory event (1). Furthermore, HAT and HDAC, and bromodomain reader proteins, are families with multiple members, reflecting the diverse biological roles that these enzymes and reader proteins undertake (2, 3). Lysine acetylation occurs on proteins with diverse biological functions, for example, many proteins involved with metabolism and cell signaling (4, 5), and thus acetylation cannot be viewed as a modification that is restricted to the nucleus and epigenetic control.

There is increasing recognition that lysine acetylation is deregulated in cancer, either through mutation in HAT genes or conversely deregulation of HDAC activity (6).

HDACs have been increasingly recognized as a relevant therapeutic target in neoplastic disease (7, 8). Small-molecule inhibition of HDAC activity has antiproliferative effects on cancer cells, including cell-cycle arrest and apoptosis (9), and consequently there has been considerable interest in progressing HDAC inhibitors into clinical studies (2, 10, 11). Some HDAC inhibitors have been approved for clinical use, for example SAHA/vorinostat for treating cutaneous T-cell lymphoma (CTCL), and panobinostat for multiple myeloma (10, 12, 13). It remains unclear how HDAC inhibitors should be optimally deployed for clinical benefit, reflecting the pleiotropic and diverse cellular mechanisms that are controlled by acetylation, and lack of knowledge about how these mechanisms are influenced by HDAC inhibitors and in malignant disease.

To address this question, we have taken an unbiased loss-of-function screening approach to identify pathways that influence the effect of HDAC inhibitors on cancer cells. Here, we report that the Toll-like receptor (TLR) adaptor protein MYD88 regulates cell sensitivity to HDAC inhibitor treatment. MYD88 (myeloid differentiation primary response gene 88) augments TLR-dependent signaling, thereby affecting the level of extracellular cytokines, which act in part through an autocrine mechanism to sensitize cells to HDAC inhibition. Notably, MYD88 is a direct target for lysine acetylation, and its deacetylation is mediated by HDAC6. Significantly, hypoacetylation of MYD88 causes enhanced IL6 levels, which sensitizes cells to HDAC inhibition. MYD88 is expressed at high levels and its gene is a target for somatic mutation in hematologic malignancy (14, 15), and cells expressing the most frequently observed mutation, L265P, exhibit enhanced sensitivity to HDAC inhibitors. Furthermore, MYD88 is a component of a wider acetylation signature in the ABC subgroup of diffuse large B-cell lymphoma (DLBCL). Thus, the role of MYD88 in TLR-dependent signaling defines a novel mechanism that sensitizes tumor cells to HDAC inhibitors. Our results bear on the apparent sensitivity of malignancies of hematologic origin to HDAC inhibitor–based therapy.

U2OS cells were from ATCC. Cells were regularly tested with a Lonza mycoplasma kit. Regular immunofluorescence microscopy with nuclear staining confirmed a negative mycoplasma result (last negative test November 2015). Cells were passaged twice a week and used until 16 passages. The inducible U2OS-stable cell lines were thawed before each experiment. RIVA and HBL-1 cells were from Prof. Alison Banham (May 2013; Haemato-oncology, Nuffield Division of Clinical Laboratory Sciences, John Radcliffe Hospital, Oxford, United Kingdom) at passage 8. MYD88 L265P mutation status was confirmed in both cell lines (December 2013) by DNA sequencing. Mycoplasma testing was carried out as above (last negative test May 2016). U2OS cells were maintained in DMEM supplemented with 10% FCS and 1% penicillin/streptomycin (Invitrogen). RIVA and HBL-1 cells were maintained in RPMI supplemented with the same. The inducible FLAG-MYD88 cells were created in parental U2OS cells using the Clontech Tet-On gene expression system (Clontech). Ectopic FLAG-protein expression was induced by doxycycline according to manufacturer's instructions. HDAC6 shRNA knockdown U2OS cell lines were created in U2OS cells using HDAC6 or control shRNA lentiviral particles (Santa Cruz Biotechnology), and stable clones were selected for puromycin resistance as per the manufacturer's protocol.

Cells were transfected using GeneJuice (Merck) and harvested 24 to 72 hours after transfection. U2OS cells were transfected with siRNAs against HDAC6 (Dharmacon, M-00349900), MYD88 (Smart-Pool Dharmacon), and nontargeting control NT2 (Dharmacon) using oligofectamine (Invitrogen) to a final concentration of 50 nmol/L before harvesting. Lymphoma cells were transfected with MYD88 or control GFP siRNA mixed in OptiMEM and RNAiMax according to the manufacturer's instructions. The mixture was added to 2 × 10⁶ cells in 1-mL medium (2 mL total) for 6 hours, seeded into a 6-cm dish, and incubated for a further 4 days. One day before harvesting, SAHA was added to the transfected cells.

The PCMV-FLAG-WT MYD88 mutant derivatives L265P and K132R were created using oligonucleotides designed in accordance with Stratagene's QuikChange Site-Directed Mutagenesis Kit. MYD88 was identified by a loss-of-function screen as described in ref. 16.

The following antibodies were used: FLAG-M2 (Sigma), PARP (Cell Signaling Technology), acetylated tubulin (Sigma), actin (Sigma), HDAC6 (Millipore), phospho-Chk1 S317 (Millipore), Chk1 (Cell Signaling Technology), cleaved caspase-3 (Cell Signaling Technology), MYD88 (Cell Signaling Technology for immunoblotting and Abcam for IHC), TLR4 (Cell Signaling Technology), p53 (Santa Cruz Biotechnology), IL6 (Abcam), IL8 (Abcam), acetylated lysine (Cell Signaling Technology), and NFκB p65 (Santa Cruz Biotechnology). HDAC inhibitors were obtained from Cayman Chemical, Sigma, and Chemietek, and the MYD88 inhibitor T1617932 from AOBIOUS Inc.

Immunoblotting, immunoprecipitation, and nuclear extract preparation were carried out as described previously (8). Representative blots from at least three experiments are shown.

U2OS-Tet-FLAG-MYD88–inducible cells were seeded at 1 × 10³ cells per well in 6-well plates. FLAG-MYD88 was induced for 4 hours prior to 24 hours of drug treatment as indicated. At 9 days after drug treatment, cells were washed twice in PBS, fixed in methanol for 10 minutes, and stained with crystal violet solution for 30 minutes before extensive washing with water. The colonies were then counted. Flow cytometry was performed as described previously (16).

Cells were transfected with β-galactosidase (β-gal) and Bcl-xl reporter plasmids and expressing vector as described previously (17). Triplicate samples were transferred to a 96-well plate and processed with a WinGlo 96 Microplate luminometer (Promega), and relative activity calculated according to transfection efficiency indicated by β-gal activity.

The levels of six cytokines, IL1β, IL12p, IL10, IL6, TNFα, and IL8, were analyzed using a cytometric bead array as described by the manufacturer (BD Biosciences).

FLAG-MYD88–stable cell lines were prepared as described previously (18, 19). Samples were digested with trypsin (0.2 μg/μL) overnight at 37°C and processed for mass spectrometry as described.

IHC on paraffin-embedded formalin-fixed tissue was performed exactly as described previously (20).

A dataset of 14,816 gene signatures was created by merging signatures downloaded from http://lymphochip.nih.gov/signaturedb/ (SignatureDB), http://www.broadinstitute.org/gsea/msigdb/index.jsp MSigDB V4 (MSigDB C1–C7), http://compbio.dfci.harvard.edu/genesigdb/ Gene Signature Database V4 (GeneSigDB), UniProt Keywords (UniProtKB; refs. 21–24). Gene ontology (GO) and annotation files were downloaded from http://www.geneontology.org (03/17/2015). The structure from the .obo and annotations from the .goa files were used to build a GO gene set. Enrichment of gene sets against signatures was assessed using a hypergeometric test, where the draw is the gene set genes, the successes are the signature genes, and the population is the number of genes present on the platform. The top 500 most MYD88-correlated genes within ABC and GC DLBCL were determined across 10 DLBCL datasets encompassing 2,030 cases (Care and colleagues; manuscript submitted). These 500 genes were used for GO and gene signature enrichment analysis.

On the basis of the top 500 in ABC DLBCL, acetylation events were assigned to the 500 genes/proteins. Annotated acetylation and acetylation sites at lysine residues of proteins (experiment based, N-terminal acetylation excluded) were obtained from the following databases: Compendium of Protein Lysine Modifications (CPML), PhosphoSite Plus, Phosida, and UniProt. On the basis of the heatmaps and the signatures, the genes in the MYD88-related ABC subgroup DLBCL nonimmune clusters were assigned to the respective clusters chr3p21/chr3p22, MCM complex/DNA unwinding, metabolic process (including pentose and organophosphate pathways), macromolecule organization, and adhesion. Acetylation events of the genes in the different clusters were assigned and a protein list of the occurring proteins in the clusters (nonoverlapping) was compiled. The percentage of acetylation was calculated for the top 500 genes ABC DLBCL, the total number of genes from ABC DLBCL nonimmune clusters, and the individual clusters from ABC DLBCL nonimmune clusters.

MYD88 was identified in a genome-wide loss-of-function screen for genes that regulate cell sensitivity to HDAC inhibitors using SAHA, as described previously (16). The effect of MYD88 in regulating cell sensitivity to HDAC inhibitors was confirmed both by short-term silencing and ectopic MYD88 expression; treating cells with MYD88 siRNA reduced sensitivity to apoptosis upon SAHA treatment and conversely expressing ectopic MYD88 enhanced sensitivity to SAHA (Fig. 1A and Supplementary Fig. S1A). Furthermore, a stable inducible cell line expressing MYD88 exhibited enhanced apoptosis to SAHA (Supplementary Fig. S1B). A similar effect was observed in cell colony formation assays, where the inhibitory effect of SAHA was enhanced upon stable expression of ectopic MYD88 relative to uninduced cells (Fig. 1B). Moreover, MYD88 sensitized cells to a variety of other HDAC inhibitors, such as panobinostat, but not mechanistically unrelated agents (Supplementary Table S1). These results establish that MYD88 regulates cell sensitivity to HDAC inhibitor treatment.

One of the major effector arms of MYD88-dependent TLR signaling regulates cytokine gene expression, via NFκB-dependent transcription (25). To explore the mechanistic role of MYD88, we tested the impact of MYD88 on cytokine expression and thereafter the influence of cytokines on HDAC inhibitor–treated cells. The increased level of ectopic MYD88 caused the activation of NFκB and NFκB-responsive transcription (Fig. 2A and Supplementary Fig. S1C). We then measured the extracellular level of different cytokines regulated by the NFκB pathway, including IL1β, IL6, IL8, IL10, IL12, and TNFα, and observed that IL6 was elevated upon MYD88 expression and reduced by SAHA treatment (Fig. 2B). The level of IL8 was also enhanced by MYD88 but in contrast to IL6 remained high upon SAHA treatment. The effect of MYD88 expression was insignificant on the other cytokines measured under these assay conditions (Fig. 2B).

We addressed whether the increase in the level of extracellular IL6 or IL8 dependent upon MYD88 influenced cell sensitivity to HDAC inhibitors using neutralizing antibodies to inactivate IL6, which we added to the medium of cells expressing inducible ectopic MYD88 and then treated the cells with SAHA (Fig. 2C and D). We observed that the enhanced level of apoptosis (increased PARP and sub-G₁ cells) that occurred upon MYD88 expression with SAHA treatment was reduced by IL6 neutralization (Fig. 2C and D). In contrast, neutralization of IL8, under the same experimental conditions of MYD88 expression, had a modest effect on SAHA-dependent apoptosis (Supplementary Fig. S1D), suggesting that IL6 rather than IL8 is the more relevant effector in these conditions. The increased level of extracellular IL6 dependent upon MYD88 expression therefore augments cell sensitivity to HDAC inhibitors.

To explore the mechanism through which IL6 sensitizes cells to HDAC inhibitors, we considered an influence on cell-cycle progression, an idea that is consistent with previous reports describing the growth promoting effects of IL6 (26, 27). We compared MYD88- induced to uninduced cells, and noticed impaired cell-cycle progression, reflecting an increase in the size of the G₂ phase and reduced S-phase population, upon treatment with anti-IL6 antibody (25% − doxycycline to 37% + doxycycline); the anti-IL6–treated MYD88-induced cells exhibited a profile typical of cells stalled in G₂ phase (Supplementary Fig. S1E, i and ii; compare doxycycline +/− with anti-IL6 treatment). Because the cell-cycle profile was reminiscent of an active G₂ checkpoint (3, 28), we investigated whether anti-IL6–treated cells had an active G₂ checkpoint. Chk1 acts at the G₂–M phase transition in response to cellular stress (28), and S317 phosphorylation is a marker for active Chk1 kinase (3). Enhanced phosphorylation occurred at S317 upon IL6 neutralization (Supplementary Fig. S1E, iii), consistent with a role of Chk1 in the G₂ checkpoint and the increased G₂ phase cells, which appeared upon IL6 neutralization. An increase in phosphorylation at S317 also occurred in uninduced cells, although the effect in MYD88-induced cells was far more apparent (Supplementary Fig. S1E, iii). These results suggest that an effect of IL6 is to allow cell-cycle progression.

Having established that MYD88 regulates cell sensitivity to HDAC inhibition, we turned our attention to the control of MYD88 activity. One possibility that we considered was a direct influence of HDAC on MYD88; for example, if MYD88 was itself acetylated. Significantly, we identified lysine acetylation on both ectopic and endogenous MYD88 protein (Fig. 3A and Supplementary Fig. S1F), and then by tandem mass spectrometry identified a single acetylated lysine (K) residue (K132; Fig. 3A). Furthermore, the low but detectable acetylation level on MYD88 contrasted with its mutant derivative, carrying a lysine to arginine substitution (K132R), which showed reduced acetylation (Fig. 3B), suggesting that K132R is an acetylation loss-of-function mutant. Most importantly, K132R exhibited a longer half-life relative to wild-type MYD88 (Fig. 3C), suggesting that acetylation influenced the stability of MYD88. We then compared the properties of wild-type MYD88 to K132R, beginning with HDAC inhibitor sensitivity and observed that K132R expression sensitized cells to a higher level of apoptosis upon HDAC inhibitor treatment compared with wild-type MYD88 (PARP cleavage and sub-G₁ cells; Fig. 3D and E). Furthermore, the heightened effect of K132R coincided with an increased extracellular IL6 level compared with wild-type MYD88 (Fig. 3F). MYD88 is thus a direct target for acetylation, and acetylation impacts on its biochemical and biological properties.

We considered HDAC6 as a possible deacetylase, given its cytoplasmic colocalization with MYD88 and role in regulating cytoplasmic substrates (29), and assessed the effect of silencing HDAC6 on MYD88. We observed in HDAC6-depleted cells a reduced level of MYD88 and concomitant reduction in IL6 (Fig. 4A and Supplementary Fig. S2A–S2C). Furthermore, in cells treated with tubastatin A, a small-molecule inhibitor of HDAC6 (30), MYD88 levels were also reduced (Supplementary Fig. S2B and S2C); the acetylation of tubulin, a recognized substrate of HDAC6 (31), increased under tubastatin A treatment (Supplementary Fig. S2C). Overall, the results highlight HDAC6 as a key regulator of MYD88 and more generally suggest that acetylation controls the activity of MYD88.

MYD88 is expressed at high levels in the hematopoietic lineage and somatic mutation in MYD88, resulting in the L265P substitution observed in the ABC subtype of DLBCL and frequently in Waldenstrom macroglobulinemia (14, 15). L265P is believed to result in an oncogenic gain-of-function mutation, which augments signaling through NFκB (14). We reasoned therefore that L265P may impact on MYD88 as a sensitizer to HDAC inhibitor treatment. When the effect of L265P was compared with wild-type MYD88, an enhanced level of apoptosis was apparent upon HDAC inhibitor treatment and increased IL6 production was evident (Fig. 4B and C). Furthermore, L265P displayed a lower level of acetylation compared with its wild-type counterpart (Fig. 4D).

To establish the importance of the results for lymphoma cells, we used two pathologically relevant ABC-like DLBCL cell lines, RIVA and HBL-1, representing either wild-type MYD88 (RIVA) or a heterozygous somatic L265P mutation, respectively. We found that HBL-1 cells, expressing heterozygous MYD88 L265P, were more sensitive than RIVA cells to apoptosis (Fig. 5A and B). Moreover, the apoptosis seen upon HDAC inhibitor treatment required MYD88, as depleting MYD88 in RIVA cells reduced cell sensitivity to SAHA treatment (Fig. 5C) and TLR signaling, as treating cells with an anti-TLR4 antibody, which blocks TLR signaling (32), reduced the level of apoptosis (Fig. 5D). We then examined the level of cytokine production by each cell line and observed that extracellular IL6 was higher in HBL-1 compared with RIVA cells, in contrast to IL8 (Fig. 5E). Furthermore, neutralizing extracellular IL6 with anti-IL6 antibody caused a reduction in HDAC inhibitor–induced apoptosis (caspase-3 cleavage and sub-G₁ cells) in HBL-1 compared with RIVA cells (Fig. 5F and G), and conversely treatment with lipopolysaccharide (LPS) to augment extracellular cytokine levels increased sensitivity to SAHA treatment (Supplementary Fig. S2D and S2E). Moreover, taking a chemical biology approach to inactivate MYD88 using the small-molecule MYD88 dimerization inhibitor T6167923 (33) on cells treated with LPS caused a significant reduction in the level of SAHA-induced apoptosis (Supplementary Fig. S2D). Altogether, these results suggest that MYD88, by virtue of its ability to act through TLR signaling and augment extracellular IL6 levels, regulates cell sensitivity to HDAC inhibition.

We also examined by IHC MYD88 expression in tissue microarray biopsies from patients with DLBCL compared with biopsies taken from patients with prostate and colorectal cancer, and noncancerous tonsillar tissue biopsies; Waldenstrom macroglobulinemia sections stained for MYD88 served as a positive control (Supplementary Fig. S3). High levels of MYD88 were apparent in the DLBCL sections, but not prostate and colorectal cancer (Supplementary Fig. S3A–S3E). The conclusions drawn from the cell-based studies relating to the role of MYD88 in lymphoma cell lines could therefore be relevant to the clinical disease.

As MYD88 is believed to be important in the pathology of the ABC subgroup of DLBCL (14) and because MYD88 confers cell sensitivity to HDAC inhibitors, we reasoned that there might be a group of genes whose expression coincided with MYD88, that were functionally related with the role of MYD88 in HDAC inhibition. We thus took a bioinformatics approach to determine at the genome-wide level whether a gene signature correlated with MYD88 expression by interrogating gene expression datasets derived from DLBCL clinical biopsies. Accordingly, we ranked the 500 most correlated genes with MYD88 expression and clustered them according to GO descriptors (Fig. 6A). There was some overlap between the GO clusters within the ABC and GC subgroups, mostly related to immunologic GO terms and therefore likely to reflect a shared signature of host immune response (Fig. 6A and Supplementary Fig. S4). Notably, a number of ABC subgroup–specific GO gene clusters were apparent (Fig. 6A, bottom right square); for example, the MCM complex/DNA unwinding and metabolic process GO terms. Because some of the genes in these GO clusters are known to encode lysine acetylation targets, we examined in greater detail the acetylation content of each cluster. This indicated that each ABC subgroup–specific cluster had a high content of lysine acetylation targets, above the average lysine acetylation content of the top 500 genes (Fig. 6B). Thus, in the ABC subgroup, MYD88 expression is associated with an enriched set of genes, which, in addition to their GO term, are related through their high level of lysine acetylation.

MYD88 is an essential cytosolic adaptor protein expressed at high levels in the hematopoietic lineage, required for signaling induced by members of the TLR and IL1R super family, which play an important role in innate and adaptive immunity by regulating the expression of diverse proinflammatory genes through the activation of NFκB (25, 34). Somatic mutation in MYD88 resulting in L265P has been described in a number of human hematologic malignancies, including the ABC subtype of DLBCL and Waldenstrom macroglobulinemia (14, 15). The L265P mutation is believed to provide a gain-of-function, which activates MYD88-dependent TLR signaling, thereby driving NFκB activity and inflammatory gene expression (14).

We identified MYD88 through an unbiased loss-of-function screen aiming to define cellular pathways that are responsible for growth inhibition by HDAC inhibitors, as described previously (16). The screen identifies genes that, when expressed, sensitize cells to drug-induced cell death (8, 20). As MYD88 is a mediator of inflammatory signaling, we assessed the role of cytokines, and found that increased levels of extracellular IL6 are an important downstream effector of HDAC inhibitor cell sensitivity. The fact that neutralization of extracellular IL6 reduced the effect of HDAC inhibitors on cells provides a direct test for the autocrine action of IL6 in mediating cell sensitivity to HDAC inhibitors.

Our results suggest that one role of the pleiotropic cytokine IL6 (26, 27) is to enable cell-cycle progression by allowing transition through the G₂ phase checkpoint (28, 35), which is significant because, at a general level, HDAC inhibitors affect proliferating cells more so than nonproliferating cells (9). Cells expressing MYD88 with an active TLR signaling pathway are therefore likely to exhibit increased sensitivity to HDAC inhibitors (Fig. 6C).

We identified an acetylation site at residue K132. The K132R-mutant derivative exhibited increased protein stability and caused higher levels of IL6 and apoptosis relative to wild-type MYD88, establishing a mechanistic connection between MYD88 acetylation and cytokine signaling. We further identified HDAC6 as a key upstream regulator of MYD88 acetylation, which in turn enables HDAC and HDAC inhibitors, which target HDAC6 to impact on MYD88 activity. The reduced level of MYD88 in HDAC6-depleted cells is entirely consistent with the observed increase in half-life of the hypoacetylated K132R mutant, arguing that deacetylation by HDAC6 endows MYD88 with enhanced signaling capability (Fig. 6C). Furthermore, HDAC inhibitors influence innate immunity and have immune-modulating functions (36–39). Consequently, our results detailing the acetylation of MYD88 provide a possible mechanism that relates HDAC inhibition to immune modulation.

It is of further interest that MYD88 undergoes somatic mutation in DLBCL and Waldenstrom macroglobulinemia (14, 15, 40), and our results suggest that L265P impacts on the sensitivity of the malignant cells to HDAC inhibitors. The increased sensitivity of cells expressing L265P could reflect the ability of IL6 to promote proliferation, as documented in the cell systems investigated in this study, and is consistent with high level production of IL6 by DLBCL and Waldenstrom macroglobulinemia in clinical disease (14, 41). It will be interesting to assess how the ABC subgroup in general, and specifically the L265P subgroup, responds to HDAC inhibitor-based therapeutics. Furthermore, MYD88 has been implicated in other malignancies, for example, metastatic hepatocellular carcinoma (42), where it will be important to establish its role in mediating the effect of HDAC inhibitors in the pathology.

We uncovered a provocative relationship involving MYD88 acetylation, where MYD88 expression in the ABC subgroup correlated with a subgroup-specific cluster of genes, which exhibit distinct ontological relationships relative to the GC subgroup. Most importantly, an analysis of each GO cluster indicated a high representation of genes encoding lysine acetylation targets. Thus MYD88, which is itself subject to acetylation control, is a representative of a broader network of genes that share a propensity to encode proteins subject to acetylation control, perhaps suggestive of a more general role for lysine acetylation in mechanisms relevant to the ABC subgroup.

In conclusion, our study highlights a novel mechanism through which MYD88 and its role as an adaptor protein in TLR signaling regulate cell sensitivity to HDAC inhibitors, in part, mediated through the expression and autocrine action of cytokines like IL6 (Fig. 6C). We suggest that the ability of IL6 to foster cellular proliferation thereby augments cell sensitivity to HDAC inhibitor treatment. Most importantly, the mechanism is likely to be relevant to the established sensitivity of hematologic malignancies to HDAC inhibitors (10) and argues in addition that inflammatory disease where MYD88 expression is high might also be susceptible to HDAC inhibitor–based therapies. Our results therefore establish the importance of TLR signaling and innate immunity, and more generally the inflammatory properties of the extracellular tumor microenvironment, in mediating the effect of HDAC inhibitors on malignant cells.

No potential conflicts of interest were disclosed.

Conception and design: M. New, M. Bekheet, H. Olzscha, S. Fotheringham, N.B. La Thangue

Development of methodology: M. New, S. Sheikh, M. Bekheet, H. Olzscha, S. Fotheringham, N.B. La Thangue

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. New, S. Sheikh, M. Bekheet, H. Olzscha, M.-L. Thezenas, B.M. Kessler, N.B. La Thangue

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. New, S. Sheikh, M. Bekheet, H. Olzscha, M.-L. Thezenas, M.A. Care, R.M. Tooze, B.M. Kessler, N.B. La Thangue

Writing, review, and/or revision of the manuscript: M. New, S. Sheikh, M. Bekheet, H. Olzscha, R.M. Tooze, B.M. Kessler, N.B. La Thangue

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. New, H. Olzscha, N.B. La Thangue

Study supervision: M. New, N.B. La Thangue

The authors thank the MRC, CRUK, and Rosetrees Trust for supporting this work.

This work was supported by the MRC, CRUK (Programme grantC300/A13058), and Rosetrees Trust. M. New was in receipt of a Gordon Pillar studentship from LLR, S. Sheikh an Anya Sturdy Fellowship and a CRUK Clinical Research Fellowship, M. Bekheet a Yousef Jameel Fellowship, and H. Olzscha a long-term EMBO Fellowship.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.