Ibrutinib Therapy Releases Leukemic Surface IgM from Antigen Drive in Chronic Lymphocytic Leukemia Patients

B-cell receptor (BCR)–associated Bruton tyrosine kinase (BTK) inhibitor ibrutinib has become a milestone for therapeutic success in chronic lymphocytic leukemia (CLL) and has been rapidly shifting patient treatment algorithms away from chemotherapy-based regimens (1, 2). Inhibition of BTK by ibrutinib affects BCR signaling but also cell adhesion and migration (3–5), and induces a rapid and prominent lymph node reduction and redistribution of CLL cells into the peripheral blood (2, 6, 7). The prolonged CLL cell redistribution into the peripheral blood offers a unique way to interpret consequences of continuous separation of CL surface receptors from their ligands in the tissue environment of patients, without need of models in vitro or in animals.

The BCR is the essential functional unit for most normal and neoplastic B cells (8, 9). In CLL, BCR signaling is key to survival and proliferation, by affecting a multitude of cell functional components from signal transduction to translation and cell growth (8, 10, 11).

Analysis of the tumor BCR immunoglobulin (Ig) indicates that CLL consists of two major subsets with different origin and clinical behavior. The subset with unmutated (U) Ig gene heavy-chain variable regions (IGHV) has arisen from pregerminal center CD5⁺ B cells and has a more aggressive clinical behavior, whereas the subset with mutated (M) IGHV has arisen from postfollicular CD5⁺ B cells and generally is more indolent (12–16).

The circulating CLL cells from both subsets are characterized by low but variable surface IgM (sIgM) levels and signaling capacity (17). Although sIgM levels and signaling capacity are typically higher in U-CLL than in M-CLL (17–19), variability is evident in both subsets (17). This appears clinically important because cases with relatively higher sIgM levels/signaling capacity have a more rapid progression than those with lower sIgM levels/signaling capacity, (17) likely due to a larger proliferative component at tissue sites (20, 21).

Downmodulation of sIgM, but not of surface IgD (sIgD), levels and signaling capacity is a defining feature of antigen engagement (22). Chronic antigen engagement leads to anergy and is associated with increased basal phosphorylation levels of BCR signal–associated kinases, including spleen tyrosine kinase (SYK) and ERK1/2 (22). This has been documented in transgenic mouse models (23, 24), and similar phenotypic and functional features have been described in human autoreactive B cells (25, 26). We and others have hypothesized that part of the variably reduced sIgM levels and signaling capacity in the circulating CLL cells is a consequence of chronic exposure to putative antigens (8). In support of this hypothesis, we have documented that downmodulation of sIgM, but not of sIgD, is reversible in CLL cells when they are incubated in antigen-free media in vitro and that the recovery of sIgM levels associates with increased IgM-induced phosphorylation of SYK (18, 19). However, antigen-independent “autonomous” signaling by CLL BCRs has also been described in vitro (27). Although its relevance in vivo, where soluble Ig can compete, is unclear, this has opened the question whether chronic BCR stimulation relies on tissue stimuli.

In this study, we examined the dynamics of sIgM levels and signaling capacity upstream to BTK of the tumor cells during their circulation in the blood stream of patients for the first 3 months of ibrutinib therapy. We document recovery of sIgM, but not of sIgD, which is the first in vivo evidence of direct antigen engagement specifically at tissue sites in CLL.

Peripheral blood mononuclear cells (PBMC) were collected from 13 patients with CLL requiring treatment (12 previously treated, and 1 previously untreated) prior to (pre-), and at week 1, month 1, and month 3 following the initiation of single-agent ibrutinib therapy. None of the patients had received any (immuno)chemotherapy or steroids for the 6 months prior to ibrutinib commencement. Following pre-ibrutinib sample collection, all patients received 420 mg ibrutinib once daily each morning, and blood collection was performed no later than 2 hours from the daily ibrutinib administration. The study was approved by the Institutional Review Boards at the University of Southampton (Southampton, United Kingdom; REC: H228/02/t). All patients provided written informed consent.

Viable PBMCs were isolated by density gradient centrifugation and cryopreserved in FBS with 10% DMSO. Prior to each assay, PBMCs were thawed, washed, and allowed to recover for 1 hour at 37°C in complete RPMI1640 medium (supplemented with 10% FBS, 2 mmol/L glutamine, and 1% penicillin/streptomycin). Tumor IGHV usage and mutational status, surface IgM and IgD levels and signaling capacity as measured by intracellular calcium mobilization assay, phenotypic CD38, CD49d, CXCR4 and ZAP70 expression, and FISH characteristics according to Döhner classification at baseline were determined as described previously (7, 18, 28, 29).

PBMCs (5 × 10⁶ cells/mL) were washed, resuspended in FACS buffer (1% BSA, 4 mmol/L EDTA, and 0.15 mmol/L NaN₃ in PBS) and stained for markers (5 × 10⁵ cells/100 μL) on ice for 30 minutes protected from light. The following antibody panels and corresponding isotype controls were used: PerCP/Cy5.5-conjugated anti-CD5 (UCHT2), Pacific Blue-conjugated anti-CD19 (HIB19) or APC/Cy7-conjugated anti-CD19 (HIB19), PE-conjugated anti-IgM (MHM-88), APC-conjugated anti-CXCR4 (12G5) or anti-CD32b (FcγRIIb. 6G11; kind gift from Prof. Mark Cragg, CSU, Southampton, United Kingdom) and FITC conjugated anti-IgD (IA6-2), or anti-CD38 (HB-7), or anti-CD49d (9F10), or anti-CD20 (2H7; all from BioLegend). Following incubation, cells were washed, resuspended in FACS buffer, and 1 × 10⁴ lymphocytes were acquired on a FACSCanto (BD Biosciences). Expression was determined as geometric mean fluorescence intensity (MFI) of test sample subtracted by the relevant isotype control MFI on CD19⁺CD5⁺ CLL cells, using FlowJo software (FlowJo, LLC). To compare results at different time points, when CLL cell size was different (Supplementary Fig. S1), density of surface markers was measured as the ratio between MFI and forward scatter squared (FSC²; ref. 30).

Recovered PBMCs (1 × 10⁷/mL) were stimulated with polyclonal goat F(ab′)₂ anti-human IgM, soluble polyclonal control (20 μg/mL; Southern Biotech) or left unstimulated at 37°C for 10 minutes. For immunoblotting, cells were washed twice in ice-cold PBS and incubated for 30 minutes on ice in RIPA buffer [150 mmol/L NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, 50 mmol/L tris(hydroxymethyl)aminomethane hydrochloride (Tris-HCL), pH 8.0], containing 1× protease inhibitor and phosphatase inhibitor cocktail 2 and 3 (Sigma-Aldrich). Protein concentration was determined using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). Equivalent protein amounts were boiled for 5 minutes in Red Loading buffer and dithiothreitol (DTT; Cell Signaling Technology), separated by SDS 10% polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to nitrocellulose membranes (GE Healthcare). After blocking with 5% BSA in Tris-buffered saline, the membranes were probed with the required primary antibody and subsequently stained with horseradish peroxidase–conjugated secondary antibodies (Dako, Agilent Technologies). Following enhancement by chemiluminescence reagents (Pierce ECL and SuperSignal West Femto, Thermo Fisher Scientific), proteins were visualized using the ChemiDoc-It imaging system (UVP). Primary antibodies used were anti-phosphoBTK (Tyr223), anti-BTK (D3H5), anti-phosphoSYK (Tyr525/526; C87C1), anti-SYK, anti-LC3B (D11), anti-Bim, anti-Bcl-xL (54H6; Cell Signaling Technology), anti–Mcl-1 (Santa Cruz Biotechnology), anti–Bcl-2 (124; Dako), anti–β-actin (AC-15; Sigma), and anti-GAPDH (6C5; Ambion). Band optical density (OD) was quantified using ImageJ and normalized to the loading control β-actin or GAPDH, the levels of phospho-proteins were further normalized on the levels of the respective total proteins. “Basal” and “induced” phosphoprotein levels were defined as (test phosphoprotein OD/test total protein OD) in the absence and in the presence of stimulation by anti-IgM in vitro, respectively. “Inducibility” was defined as the ratio of induced/basal phosphoprotein levels at each time point.

Cell surface proteins were biotinylated and isolated using the Cell Surface Protein Isolation kit (Pierce, Thermo Fisher Scientific), as described previously (31). Cell surface glycosylation patterns were determined by digestion with endoglycosidase H (EndoH; New England Biolabs), which removes only the mannosylated N-glycans characteristic of the immature IgM form, but does not affect the mature IgM form, or peptide:N-glycosidase F (PNGase; New England Biolabs), which removes all attached N-glycans (31). Biotinylated proteins were analyzed by immunoblotting using the primary antibody anti-μ (Jackson ImmunoResearch Laboratories, Stratech Scientific).

Continuous variables were compared by Wilcoxon signed rank nonparametric test. For statistical correlation between two variables, the nonparametric Spearman rank test was used. All statistical tests were two-sided. Statistical significance was defined as P < 0.05. Analyses were performed with GraphPad Prism 6 software.

Twelve patients with CLL were investigated for phenotypic and functional changes prior to (pre-), and at week 1, at month 1, and at month 3 following the initiation of single-agent ibrutinib therapy (Supplementary Table S1). Prior to ibrutinib commencement, mean sIgM levels and signaling capacity were relatively high and similar to those in patients with progressive CLL in 11 of 12 patients investigated (MFI mean 102, range 24–221 and IgM iCa²⁺ % mean 40, range 16–77; ref. 17). The one case (489) that had low levels/signaling capacity was an exception. Time points chosen for analysis posttreatment initiation were week 1, when blood lymphocyte counts were increasing in all patients (fold increase range from pretherapy 1.4–3.5, mean 2.1), month 1 and month 3 (Supplementary Table S2). Basal BTK phosphorylation was decreased at all time points during ibrutinib therapy from week 1 to month 3 compared with pretherapy (Supplementary Fig. S2A–S2C). Following stimulation with anti-IgM, BTK phosphorylation levels increased significantly prior to ibrutinib commencement, whereas no response was detectable during therapy in all patients (Supplementary Fig. S2A–S2C). The reduced basal BTK phosphorylation and the inhibition of anti-IgM–induced BTK phosphorylation capacity at week 1 and at subsequent time points confirmed that ibrutinib was operating in vivo at the time of blood collection.

Expression of molecules involved in BCR signaling (IgM, IgD, CD19, CD32b) or adhesion (CD38, CD49d) and of CD20 was determined on the circulating CLL cells before and during ibrutinib therapy.

At week 1, a significant increase of sIgM expression was observed on the tumor cells of 11 of 12 cases (Fig. 1A). The only exception was case 489, which had very low sIgM levels and signaling capacity since before treatment commencement and failed to increase sIgM expression. The increase occurred in both U-CLL and M-CLL (Fig. 1A). A representative case is shown in Fig. 1B. The mean MFI of the 11 of 12 responders was 64% higher at week 1 than pretherapy. In contrast, expression of other markers, including sIgD, did not change or slightly decreased, while downmodulation of CD20 was already particularly evident at this early time point (Fig. 1A and B).

To compare the changes of surface molecule expression at different time points, when CLL cell size was different, we normalized expression by dividing receptor MFI with FSC² at each time point. In fact, by using forward scatter (FSC) as an indicator of cell size (32–34), we observed that CLL cells significantly and gradually reduced in size at months 1 and 3 of therapy in all cases (P ≤ 0.001; Supplementary Fig. S1A). Reduction of CLL cell size was accompanied by a significant increase of LC3BII (P = 0.02) at month 3 of therapy, to indicate autophagosome formation (Supplementary Fig. S1B; ref. 35). At this time point, there was also specific increase of proapoptotic Bim-EL levels (P = 0.02), while Mcl-1, Bcl-xL, and Bcl-2 levels remained stable in the circulating CLL cells in vivo (Supplementary Fig. S1C). This suggested promotion of specific Bim-mediated apoptosis events (36, 37).

Despite reduction of cell size, sIgM density at months 1 and 3 remained higher than baseline values observed prior to therapy (Fig. 2; Supplementary Fig. S3). In marked contrast, all the other receptors examined significantly reduced expression at these later time points (P < 0.01; Fig. 2; Supplementary Fig. S3). This reflected a real reduction in surface density in the size-contracted cells. The reduction included sIgD and appeared particularly pronounced for CD20.

The specific early dissociation of trends between sIgM levels, which increased during therapy, and sIgD levels, which reduced in a fashion similar to the other receptors measured, was highly suggestive of a distinct influence operating specifically on the sIgM, potentially involving antigen disengagement (22, 25, 26).

We focused our analyses on sIgM at week 1. The later time points were not investigated further because of the significant changes of cell size with ongoing autophagy/apoptosis (Supplementary Fig. S1), which makes biological interpretation of sIgM changes more complex.

It has been reported that CLL cells recently egressed from tissue sites have diminished expression of CXCR4 and relatively high expression of CD5 (CXCR4^dimCD5^hi), whereas those that have been persisting in the circulation are CXCR4^brightCD5^low (38). We determined sIgM expression in both fractions at pretherapy and week 1. Both the CXCR4^dimCD5^hi and the CXCR4^brightCD5^low absolute cell counts increased in the peripheral blood. On average, the relative proportion of the recently egressed fraction was higher (mean percentage 33% of the total CLL population pretherapy, 39% at week 1), whereas the CXCR4^brightCD5^low fraction was lower (mean percentage, 28% pretherapy, 23% at week 1; Fig. 3A and B), although the differences were not significant. SIgM expression was higher in the recently egressed cells than in the persistently circulating cells. However, sIgM increase was significant and similar in both the CXCR4^dimCD5^hi cells and the CXCR4^brightCD5^low cells (Fig. 3C, approximately 40% increase from pretherapy in both fractions, and Supplementary Fig. S4). These data indicated that sIgM increased in all fractions during ibrutinib therapy. They suggested that the increase was not simply a reflection of the redistribution of subpopulations with different sIgM levels (39).

We have previously shown that sIgM, but not sIgD, expression and signaling capacity recover spontaneously in vitro (18). Those observations were obtained from peripheral blood CLL cells that had not been inhibited by ibrutinib and hence that had been exposed to tissue interactions (40, 41). Here, we determined sIgM expression changes in vitro in 4 CLL samples (409, 531, 551, 601) prior to or at week 1 of ibrutinib therapy (Fig. 4). From analysis of the samples obtained pretherapy, sIgM (but not sIgD) levels on the CLL cells increased following culture in vitro for 48 hours in all 4 patients (531, 35%; 551, 42%; 409, 103%; and 601, 194% increase, Fig. 4B and C), as expected (18). The increase in sIgM expression occurred both when ibrutinib was added and when ibrutinib was not added in vitro (Fig. 4B). The sIgM increases after 48 hours culture of the pretherapy samples in vitro were similar to (cases 409, 531, and 601) or lower (case 551) than those observed at week 1 of therapy from the same patient. From analysis of the samples obtained at week 1 of ibrutinib therapy, no further significant increase in sIgM expression could be achieved following culture in vitro for 48 hours (Fig. 4B). To investigate a potentially direct effect of ibrutinib on expression of sIgM, we exposed a B-cell line to ibrutinib at different concentrations in vitro. The results document that ibrutinib per se does not lead to an increase in sIgM expression, but may actually reduce expression, especially at the higher concentration where toxicity is more likely (Supplementary Fig. S5). This indicated that ibrutinib therapy had allowed recovery of sIgM expression. Because the increase was occurring in vivo, on cells persisting in the blood circulation, and the drug had no direct increasing effect in vitro, we reasoned that the specific sIgM increase was due to prolonged disengagement from tissue antigen.

We have shown previously that the N-glycosylation status of the glycans in the IgM heavy chain constant region varies in CLL cells from the blood (31). This reflects the level of activation of the cell, with immature glycan predominating in the blood, but mature glycan returning with recovery of sIgM expression in vitro. An increasing ratio of mature/immature IgM glycoform is therefore an index of antigen-free recovery (31).

We assessed the N-glycosylation status of the sIgM μ-chains in samples from 3 patients, 2 of which had increased sIgM levels on ibrutinib (551 and 343) and the single outlier, which had maintained similarly very low sIgM levels pre- and on ibrutinib therapy (489; Fig. 5). The ratio of mature to immature glycoforms increased in both the patients with increased sIgM, whereas it did not change in the patient that did not increase sIgM levels. In fact, the sample that had the greatest increase in sIgM expression (551) also had the greatest increase in the expression of the mature glycoform. These data are consistent with the established recovery of the mature fully N-glycosylated form of sIgM during incubation in vitro (31), and support the hypothesis that ibrutinib therapy is mimicking this situation by preventing access of the cells to putative antigen in vivo (18, 31). Case 489 presents an interesting exception where, still in the absence of the most common recurrent genomic changes (Supplementary Table S1), there appears to be a profile of chronic and severe anergy, which is not recoverable either by blocking BTK in vivo or by incubation in vitro.

Constitutive activation of SYK and of downstream kinases is a hallmark of chronic stimulation by antigen (42). We examined our 12 cases for changes in the activation status of SYK following ibrutinib commencement. We found that basal phosphorylation of SYK during ibrutinib therapy was significantly lower than prior to treatment in all the 12 cases investigated. Three representative cases are shown in Fig. 6A (P = 0.001). This indicated loss of stimulation by antigen when the tumor cells are exiled in the blood circulation by continuous ibrutinib therapy.

The induced phosphorylation of SYK, which is upstream of BTK, was measured to determine whether the elevated levels of sIgM on the CLL cells were functional during ibrutinib. The ability of anti-IgM to induce SYK phosphorylation was measured as the ratio of sIgM-mediated pSYK (OD): basal pSYK (OD) levels (pSYK inducibility). sIgM-mediated inducibility of SYK at week 1 was greater than pretherapy (P = 0.01; Fig. 6B and C). Inducibility correlated both with overall sIgM levels at week 1 (r = 0.64, P = 0.03; Fig. 6D) and with the differential increase of sIgM levels (Fig. 6E). However, the greater inducibility was a reflection of the fall in basal levels of phosphorylated SYK during ibrutinib therapy, while maximal total phosphorylation capacity was overall similar pretherapy and during ibrutinib. These results demonstrate maintained capacity of anti-IgM to induce SYK while on ibrutinib therapy and indicate that the increased sIgM observed is functional.

The sIgM is a known influence on CLL behavior, and its level or signaling capacity correlates with disease progression (17). Clinical responses to drugs such as the BTK inhibitor ibrutinib, which acts on BCR-associated pathways, are consistent with this. However, the effects of ibrutinib are not exclusive to this pathway, with known inhibition of chemokine-mediated migration and of adhesion, features important in mediating redistribution of CLL cells in the peripheral blood (3–5).

We have investigated the status of sIgM on circulating CLL cells of patients during the initial months of ibrutinib therapy, when lymphocytosis was present. We observed a significant increase in sIgM expression in the first week of therapy. The expression remained higher than pretherapy levels at subsequent time points, despite a significant decrease in CLL cell size. This contrasted with the reduction of expression of all the other surface receptors investigated, including sIgD. Although the parallel decreases in several cell surface markers may simply reflect a contracting cell size with the ongoing autophagocytic events, downregulation of CD20 appeared more significant and has been explained to be caused by the prevention of access of tumor lymphocytes to chemokine-producing stromal cells in tissue sites by ibrutinib therapy (43).

The distinctive behavior of sIgM as compared with sIgD is highly suggestive of a specific role for antigen disengagement, which then allows sIgM recovery (23, 24). This discriminatory effect on sIgM is seen in normal human B cells anergized by chronic antigen exposure (25, 26). The degree of recovery of expression and the presence of the mature glycoform of sIgM in CLL cells from ibrutinib-treated patients point to a complete reexpression of sIgM to levels similar to that obtained by culturing pre-therapy cells under exogenous “antigen-free” conditions (18). Our findings now add a new direct tier of evidence that antigen engagement occurs at specific tissue sites.

By using ibrutinib as a “tool” in vivo to inhibit redistribution in tissue sites (e.g., lymph node or bone marrow), our analysis revealed that the CLL cells within the patients had the capacity to fully recover and maintain expression of sIgM. Recent studies in vitro have raised the hypothesis that BCR levels and signaling are regulated by antigen-independent mechanisms, including “autonomous” signaling or distinct homotypic BCR interactions (27, 44). Either hypotheses imply heterogeneous, but tumor-specific structure characteristics intrinsic and unique to the CLL cell BCRs. However, we found that the Ig structures appear no dissimilar from those of the normal B-cell repertoire (13), and soluble Ig in the blood can compete in vivo. Also, the specific level and signaling recovery that affect IgM, but not IgD, in CLL circulating cells would be difficult to explain with autonomous signaling, which would operate in all tumor compartments. Instead, the modulation could be explained by the evidence in vivo (mouse models) provided by the same investigators of selection and expansion of the CLL clone by pathogen-specific BCRs that cross-react with one or more self-antigens (45). Our current study highlights for the first time in patients that CLL IgM modulation and maturation is regulated by tissue-based antigens.

Ibrutinib is known to prevent the return of circulating CLL cells to tissue sites, while egress will continue (46). The circulating cells at week 1 will therefore include two fractions, identified as the CXCR4^dimCD5^hi recently egressed fraction, possibly with higher levels of CD20 (43), and the CXCR4^brightCD5^low persisting fraction, respectively (38). The proportions are difficult to estimate and the phenotype is likely to change over time in vitro and in vivo (39, 43). When the recently egressed and the persistently circulating fractions were identified by means of these markers, we observed that sIgM was higher in the recently egressed CLL cells. Interactions of CLL cells with tissue environmental cells will affect levels of functional BCR IgM/D complex (47). We and others have speculated that the higher sIgM might be a consequence of expression enhancement by tissue-derived IL4 (48, 49). However, both fractions increased their relative sIgM levels during therapy, suggesting that the reversible downregulation of sIgM due to antigen is a feature of both recent tissue emigrants and of those circulating in the peripheral blood, where the imprint of antigen engagement is still partially evident (18).

In terms of functional capacity of the recovered sIgM, there is a clear blockade of BTK activity in drug-treated patients but the ability of anti-IgM to phosphorylate SYK, which lies upstream of BTK, is still present. Basal levels of pSYK can be detected in CLL cells at the pretreatment stage, which can be explained by antigen engagement in vivo (18). These levels decline on ibrutinib treatment, again pointing to a block on access to tissue-based stimuli (50).

The data on sIgM expression therefore reveal that the effect of ibrutinib on CLL cells in vivo is best explained by the prevention of access of circulating cells to tissue sites, probably by inhibiting chemokine-mediated migration (3–5). The lack of a nourishing signal from (auto)antigen then consigns the CLL cells to a circulating limbo where they survive for some time, possibly due to maintained antiapoptotic strategies, but slowly contract in size. The recovery of sIgM in the absence of antigen engagement argues that tissue-localized antigen is the likely driver of proliferation. One question is whether the CLL cells in limbo, which are presumably prevented from tissue access by ibrutinib, but which retain proximal signaling function, should be removed by a second level of therapeutic attack.

P.W. Johnson reports receiving commercial research grants from Janssen and Epizyme, is a consultant/advisory board member for Takeda, Bristol-Myers Squibb, Novartis, Celgene, Boehringher Ingelheim, Kite, Genmab, and Incyte, and reports other remuneration from Janssen. L. Trentin reports receiving commercial research grants from Janssen and Gilead, and speakers bureau honoraria from Roche and Abbvie. A.J. Steele is an employee of Gilead, reports receiving commercial research grants from Janssen and Portola Pharmaceuticals, and holds ownership interest (including patents) in Portola Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S. Drennan, G. Chiodin, F.K. Stevenson, F. Forconi

Development of methodology: S. Drennan, G. Chiodin, A. D'Avola, F.K. Stevenson, F. Forconi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Drennan, G. Chiodin, A. D'Avola, I. Tracy, F. Forconi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Drennan, G. Chiodin, F.K. Stevenson, F. Forconi

Writing, review, and/or revision of the manuscript: S. Drennan, G. Chiodin, P.W. Johnson, A.J. Steele, G. Packham, F.K. Stevenson, F. Forconi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. D'Avola, F. Forconi

Study supervision: P.W. Johnson, L. Trentin, G. Packham, F. Forconi

The authors would like to thank Isla Henderson (supported by the ECMC C24563/A15581 grant), Dr. Kathy Potter, and Carina Mundy (supported by the Hairy Cell Leukemia Foundation) for sample identification, collection, and storage in the South Coast Tissue Bank (Bloodwise grant 16003). This study was supported by Cancer Research UK (CRUK Centre grant C34999/A18087), the Southampton Cancer Research UK and NIHR Experimental Cancer Medicine Centres at the University of Southampton, Bloodwise (grants 18009, 16003, 14037, and 12021), the Keanu Eyles Haematology Fellowship for the Cancer Immunology Centre, the Hairy Cell Leukemia Foundation, and the Gilead UK & Ireland Oncology Fellowship Programme 2016.

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.