MicroRNA-155 influences B-cell receptor signaling and associates with aggressive disease in chronic lymphocytic leukemia

Signaling via the B-cell receptor (BCR) is thought to play a role in the pathogenesis and/or progression of chronic lymphocytic leukemia (CLL).^1,2  BCR signaling is triggered by phosphorylation of immunoreceptor tyrosine-based activation motifs of CD79a/b and induces a cascade of downstream signaling pathways.^3,4  The importance of this cascade in CLL biology appears underscored by clinical trials demonstrating clinical activity of small-molecule kinase inhibitors that can block BCR signaling.⁵ 

The biology of CLL also can be influenced by microRNAs.^6-10  In particular, the evolutionarily conserved microRNA miR-155 is a critical regulator of posttranscriptional gene expression in B cells.⁶  It is encoded within a region known as the B-cell integration cluster (BIC, miR155HG), which originally was identified as being a frequent integration site for avian leucosis virus.^11-13  This microRNA is overexpressed in numerous B-cell lymphomas,^14-16  including CLL,^8,17-20  suggesting that miR-155 contributes to lymphoma development.¹⁹  Consistent with this notion is the observation that mice made transgenic for miR-155 under a B-cell–specific promoter developed pre–B-cell lymphomas.²¹  Specifically, high levels of miR-155 also are present in diffuse large B-cell lymphomas with an activated B-cell phenotype, which is associated with a relatively poor clinical prognosis.¹⁴  Moreover, relatively high-level expression of miR-155 in CLL has been associated with expression of adverse prognostic markers, such as the ζ-chain associated protein of 70 kD (ZAP-70), unmutated immunoglobulin heavy chain variable region genes (IGHV), and/or deletions in 17p or 11q.^22-25  Overexpression of miR-155 in transgenic mice induces polyclonal B-cell expansion, suggesting that miR-155 could enhance B-cell proliferation.²¹ 

One identified target of miR-155 is the Src homology-2 domain-containing inositol 5-phosphatase 1 (SHIP1), which is encoded by INPP5D.^26,27  SHIP1 is a phosphatase that acts in opposition to kinases, which are integral to many signal transduction pathways.²⁸  Such inhibitory phosphatases may suppress surface immunoglobulin and BCR signaling.²⁹  Such signaling appears most prominent in lymph nodes (LNs), where CLL cells apparently become activated in response to signals from the LN microenvironment.³⁰  Moreover, CLL cells in the LNs have upregulated BCR gene expression signatures and higher proportions of cells undergoing proliferation than CLL cells in the blood.³⁰  Conceivably, the expression of miR-155 could influence the relative expression of SHIP1 in CLL, which then could influence the relative activation of signaling pathways triggered by ligation of the BCR by self- or environmental antigen(s). We hypothesize that high-level expression of miR-155 in CLL can repress expression of SHIP1 and increase the responsiveness to BCR ligation, thus possibly accounting for its association with adverse clinical outcome in patients with CLL.

Blood samples were collected from consenting patients who satisfied diagnostic criteria for CLL and enrolled in University of California San Diego Moores Cancer Center biorepository per a protocol approved by the institutional review board (080918). At the time of sample collection, patients had not received prior therapy. IGHV mutation status and ZAP-70 expression were assessed according to established criteria.³¹  We used Ficoll-Hypaque density-gradient centrifugation to obtain mononuclear cells, of which ≥95% were CD5⁺CD19⁺ cells. Descriptions of cell–cell transfection, measurement of intracellular calcium flux, real-time polymerase chain reaction (PCR), flow cytometry, and statistical analyses are provided in the supplemental methods on the Blood Web site. This study was conducted in accordance with the Declaration of Helsinki.

We studied the relationship between the relative leukemia cell expression of miR-155 and treatment-free survival (TFS), or overall survival (OS), in a cohort of 86 CLL patients (Table 1; supplemental Table 1), for which we assayed miR-155 expression levels using absolute real-time PCR. Forty-three of the samples used mutated IGHV and the other 43 used unmutated IGHV; 41 of the samples lacked expression of ZAP-70, whereas 45 expressed ZAP-70. We used the profile-likelihood method in a Cox regression model of TFS to determine the optimal threshold level of miR-155 that might segregate these patients into 2 subgroups with disparate progression tendencies. Thirty-one patients were stratified into a “miR-155-Low” subgroup and 55 patients were assigned to a “miR-155-High” subgroup, defining a cutoff point of 2553 copies per CLL cell. The subgroup of patients in the miR-155-High subgroup had a median TFS of 4.4 years, whereas those patients in the miR-155-Low subgroup had a median TFS of 11.4 years (Figure 1A). The median OS for patients in the miR-155-High subgroup (11.3 years) also was shorter than that of patients in the miR-155-Low subgroup (>20 years) (supplemental Figure 1). Triplicate analyses using relative real-time PCR (miR-155 vs RUN6B) demonstrated that the CLL cells in the miR-155-High subgroup had 2.4-fold higher mean level of miR-155 than did the CLL cells in the miR-155-Low subgroup (6.6 ± 0.5 vs 2.8 ± 0.3; supplemental Figure 2).

Although there was a significant association between expression of high-level miR-155 and ZAP-70 or use of unmutated IGHV (supplemental Figure 3), these associations were not absolute. Some of the CLL samples in the miR-155-High subgroup did not express ZAP-70 (n = 24) or used mutated IGHV (n = 17); such cases had a mean level of miR-155 that did not differ significantly from that of cases in the miR-155-High subgroup that were ZAP-70 positive or used unmutated IGHV. Similarly, some of the CLL samples in the miR-155-Low subgroup expressed ZAP-70 (n = 14) or used unmutated IGHV (n = 10) (Figure 1B-C).

We used an independent data set of 181 additional patients, who had expression levels of the miR-155 precursor measured by microarray (GSE13159, GSE13164).³²  The median follow-up time for this validation cohort was 7.6 years, with 64.1% (n = 116) of patients receiving therapy and 24.3% patients deceased (n = 44), similar to the original cohort. In the original collection of 86 samples, the mature miR-155 measured by quantitative PCR was well correlated its precursor measured by microarray (r = 0.6, P < .001; supplemental Figure 4). We thus calibrated the cut-point established from the quantitative PCR assay to the microarray assay using linear regression on these original 86 samples (supplemental Figure 4). We then validated the association of high-level expression of miR-155 (as measured by the microarray) with reduced TFS and OS in the 181 new subjects, using the optimal cut-point established through analysis on the original 86 subjects. From the analysis of these new, independent data, we observed that high-level expression of miR-155 retained its strong predictive effect for TFS (Figure 1D), both when considered alone (hazard ratio [HR] 1.7; 95% confidence interval [CI] 1.1-2.4) or in a model that contained both IGHV mutational status and ZAP-70 status (Table 2). In addition, greater OS was observed for the miR-155-High group, although this did not attain statistical significance in univariate analysis (supplemental Figure 1B). In multivariate Cox models, we confirmed the independent prognostic information of miR-155 over and above the established risk factors IGHV and ZAP-70 (Table 2), both for TFS (HR 1.52; 95% CI 1.03-2.23) and for OS (HR 2.22; 95% CI 1.08-4.54), using the validation cohort. Moreover, among patients with CLL cells that lacked expression of ZAP-70, those with high levels of miR-155 (n = 56) had a shorter observed median TFS (6.0 years) than patients with low-level expression of miR-155 (n = 54, 9.6 years, Figure 1E). Among patients with CLL cells that expressed mutated IGHV genes, high levels of miR-155 (n = 48) were associated with significantly shorter median TFS (6.7 years) than those with low-level expression of miR-155 (n = 52, 18.3 years, Figure 1F).

We examined CLL cells for expression of SHIP1 by flow cytometry. In other cell types, miR-155 has been noted to target the gene encoding SHIP1, namely INPP5D.^26,27  However, miR-155 had not been reported to influence expression of SHIP1 in CLL cells. We found that CLL cells in the “High miR-155” subgroup (n = 21) had a median mean fluorescence intensity ratio (MFIR) when stained for SHIP1 of 6.4 (± 0.3), which was significantly lower than that for CLL cells in the “Low miR-155” subgroup (8.0 ± 0.4, n = 24, P < .01) (Figure 2A; supplemental Figure 5A).

To further evaluate the relationship between expression of SHIP1 and miR-155 in primary CLL cells, we transfected CLL cells with a mimetic of miR-155. Transfection of CLL cells with the miR-155 mimetic reduced expression of SHIP1, as assessed by immunoblot or flow cytometric analysis (n = 5, Figure 2B, top; supplemental Figure 5B,C). Moreover, the median MFIR of samples stained for SHIP1 was significantly lower in cells transfected with the miR-155 mimetic than in control-transfected cells (4.4 ± 1.3 vs 6.5 ± 1.2, n = 3, P < .05) (Figure 2B, bottom). Conversely, we transfected CLL cells with inhibitors of miR-155. CLL cells transfected with a miR-155 inhibitor were found to express higher levels of SHIP1, as assessed by immunoblot analysis or flow cytometry (5.7 ± 0.8 vs 4.0 ± 0.9, n = 3, P < .05) (Figure 2C). Collectively, these studies demonstrate that expression of SHIP1 in primary CLL cells is influenced by the relative expression of miR-155.

CLL cells were stimulated with anti-µ, which rapidly induced a calcium flux that could be assessed via flow cytometry. The median increase in fluorescence intensity of calcium flux induced by BCR ligation inversely correlated with SHIP1 expression in CLL cells (supplemental Figure 6). The median increase in calcium flux induced by anti-µ was significantly greater for CLL cells in the miR-155-High subgroup (3.4 ± 0.6) than in CLL cells in the miR-155-Low subgroup (1.2 ± 0.3, n = 10, P < .01) (Figure 3A). Transfection of the cells with a miR-155 mimetic enhanced the magnitude of calcium flux induced by treatment with anti-µ (Figure 3B). The median increase in fluorescence intensity of calcium induced by treatment of CLL cells with anti-µ was significantly greater in miR-155-transfected cells than in control-transfected CLL cells (Figure 3B) (3.4 ± 0.5 vs 1.9 ± 0.4, n = 7, P < .01). Conversely, CLL cells had reduced calcium flux in response to anti-µ after transfection with an inhibitor of miR-155. The median increase in fluorescence intensity of calcium flux induced by anti-µ was significantly lower in cells transfected with inhibitor of miR-155 than in control-treated CLL cells (3.7 ± 0.4 vs 4.7 ± 0.5, n = 6, P < .01) (Figure 3C). Furthermore, we found that the median increase in calcium flux induced by treatment with anti-µ was significantly greater in ZAP-70–negative CLL cells with a high-level expression of miR-155 than in ZAP-70–negative CLL cells with a low-level expression of miR-155 (2.5 ± 0.6 vs 1.0 ± 0.4 n = 9, P < .05) (Figure 3D).

We interrogated the PubMed GEO database on CLL cells from LN, bone marrow, or peripheral blood. We found that expression of the precursor to miR-155, namely BIC (miR155HG), was significantly greater in CLL cells isolated from LN than in CLL cells isolated from the bone marrow or peripheral blood of the same patient (supplemental Figure 5). On the other hand, the expression level of INPP5D, encoding SHIP1, appeared higher in CLL cells isolated from the blood than in CLL from the LNs (supplemental Figure 6).

Recent studies have identified markers that can distinguish CLL cells that recently have exited from the LNs from those that have been in the circulation longer and that presumably are poised to reenter the LN tissue compartments.³³  The former CLL cells express low levels of CXCR4 (CXCR4^dim), but higher levels of CD5 (CD5^bright), relative to the latter, which are instead CXCR4^brightCD5^dim CLL cells.³³  Prior studies demonstrated that CXCR4^dimCD5^bright CLL cells have higher proportions of cells that stain for the proliferation marker Ki-67 than CXCR4^brightCD5^dim CLL cells. Furthermore, gene expression analyses of isolated CLL subpopulations have revealed that CXCR4^dimCD5^bright CLL cells have higher levels of pro-proliferation and antiapoptotic genes than CXCR4^brightCD5^dim CLL cells of the same patient.³³ 

We isolated blood CXCR4^dimCD5^bright CLL cells and CXCR4^brightCD5^dim CLL cells from the same patient and examined each subset for expression of miR-155 and SHIP1 and their capacity to respond to BCR ligation (Figure 4A). Expression of miR-155 was significantly higher in CXCR4^dimCD5^bright CLL cells than that in CXCR4^brightCD5^dim CLL cells (8.0 ± 4.5 vs 1.0 ± 0, n = 4, P < .05) (Figure 4B). On the other hand, expression of SHIP1 was significantly lower in CXCR4^dimCD5^bright CLL cells than that in CXCR4^brightCD5^dim CLL cells (Figure 4C). Furthermore, the median increase of calcium fluorescence intensity induced by anti-µ was significantly greater in CXCR4^dimCD5^bright CLL cells than that in CXCR4^brightCD5^dim CLL cells (2.4 ± 0.4 vs 0.4 ± 0.1, n = 10, P < .01) (Figure 4D).

In LNs, CLL cells interact with supportive cells, such as activated T cells or Nurse-like cells, which can provide survival- or proliferation-inducing signals via the expression of members of the tumor necrosis factor family of proteins (eg, CD40-ligand or CD154), B-cell–activating factor (BAFF or CD257), or a proliferation inducing ligand (APRIL or CD256).^34,35  We examined whether CD154 or BAFF could upregulate CLL cell expression of miR-155 and thereby influence the expression of SHIP1 and sensitivity to BCR ligation. CLL cells were stimulated with CD154 or BAFF for 24 or 48 hours in vitro and then examined for miR-155, SHIP1, and responsiveness to anti-µ. Treatment with CD154 or BAFF significantly upregulated expression of miR-155, namely 2.7 ± 0.6 vs 1.0 ± 0 (n = 3, P < .05) (Figure 5A) or 1.5 ± 0.1 vs 1.0 ± 0 (n = 3, P < .05) (supplemental Figure 7A), respectively. Conversely, expression of SHIP1 was significantly lower in CLL cells treated with CD154 than in control cells (4.6 ± 0.7 vs 8.8 ± 1.2, n = 6, P < .05) (Figure 5B) or in CLL cells treated with BAFF than in control cells (5.9 ± 1.0 vs 8.8 ± 1.2, n = 6, P < .05) (supplemental Figure 7B). Furthermore, CD154 or BAFF enhanced the responsiveness of CLL cells to BCR ligation (Figure 5C; supplemental Figure 7C); the median increase of calcium fluorescence intensity induced by anti-µ was significantly greater in CLL cells treated with CD154 than in untreated cells (5.0 ± 1.1 vs 2.4 ± 0.4, n = 9, P < .05) (Figure 5C) or in CLL cells treated with BAFF than in untreated cells (4.1 ± 2.2 vs 2.5 ± 1.9, n = 4, P < .05) (supplemental Figure 7C). To examine whether the enhanced responsiveness to BCR ligation induced by CD154 was dependent upon induced expression of miR-155, CLL cells were transfected with miR-155 inhibitor before treatment with CD154. We found that the median increase of anti-µ calcium fluorescence intensity induced by treatment with CD154 was significantly lower in CLL cells transfected with the miR-155 inhibitor than in cells transfected with control miRNAs (2.9 ± 1.3 vs 4.3 ± 1.2, n = 4, P < .05) (Figure 5D).

We evaluated whether stimulation with CD154 could upregulate expression of miR-155 and thereby influence the expression of SHIP1 and responsiveness to BCR ligation in normal B cells. Blood B cells were isolated from normal donors (n = 3) and stimulated with CD154. After 24 or 48 hours of stimulation, we examined for changes in expression levels of miR-155 and SHIP1, and in responsiveness to anti-µ. As noted for CLL cells, treatment with CD154 upregulated expression of miR-155, downregulated expression of SHIP1, and enhanced their responsiveness to BCR ligation (supplemental Figure 8).

Although prior studies identified an association between high-level expression of miR-155 and expression of ZAP-70 or use of unmutated IGHV,^{8,22,23,36,37}  we found that this association was not absolute and that high-level expression of miR-155 in CLL has additional independent prognostic value. We determined the optimal threshold for using miR-155 (∼2550 copies CLL cell) to stratify patients in a training cohort of 86 patients into 2 subgroups with disparate progression tendencies by performing profile-likelihood analysis.³⁸  We observed independent prognostic value of high-level miR-155, over and above ZAP-70 and IGHV status. This observation was confirmed in an independent validation dataset of 181 patients. Although our threshold for defining high-level expression of miR-155 does not take into account intraclonal variation, it nonetheless could be useful for stratifying samples from patients for whom clinical outcome data might not be available. Importantly, defining high-level expression of miR-155 in this manner allowed us to interrogate CLL cells with high vs low levels of miR-155 for phenotypic and functional differences that may account in part for differences in clinical outcome.

We found that CLL cells with high levels of miR-155 expressed significantly lower levels of SHIP1 protein. SHIP1 is a phosphatase that is encoded by INPP5D, which had been found to be a target for miR-155 in cell types other than CLL.²⁶  Although INPP5D encoding SHIP-1 is a known target gene for miR-155, it has never been demonstrated that INPP5D levels inversely correlate with expression of miR-155 in CLL cells. This is important because each miRNA can have multiple different targets, which vary depending upon the cell type in which a given miRNA is expressed. High-level expression of an irrelevant gene that also was targeted by miR-155, for example, could compete with INPP5D for its ability to be regulated by miR-155. Indeed, prior studies by Loosner and colleagues identified a number of genes that could be targeted by miR-155 in HEK293T cells.³⁹  However, when they compared their results with previously identified miR-155 target proteins in other cell lines, they found major differences, demonstrating the cell-line specificity of microRNAs. Consistent with the notion that miR-155 regulates expression of INPP5D in CLL cells, we demonstrated through transfection studies that enhanced expression of miR-155 induces lower expression of INPP5D and that inhibition of miR-155 induced higher expression of INPP5D.

We also found that CLL cells with high levels of miR-155 also were more responsive to surface-µ ligation than CLL cells with low levels of miR-155. Prior studies found that CLL cells that lacked ZAP-70 generally experienced lower levels of activation after treatment with anti-µ.^8,17,18  However, there were atypical cases that lacked expression of ZAP-70, but could be well-stimulated by treatment with anti-µ, suggesting that other factors could contribute to differences in sensitivity to BCR ligation. In the present study, we examined samples that did not express ZAP-70, but that had high vs low expression levels of miR-155. In such cases, the BCR signaling appeared to be associated with the expression level of miR-155, suggesting that the relative expression of this microRNA could influence the relative proficiency of BCR signaling in primary CLL cells. Transfecting CLL cells with miR-155 reduced expression of SHIP1 and enhanced the cells’ sensitivity to surface-µ ligation. Conversely, transfection of CLL cells with a miR-155 inhibitor had the opposite effects. Because enhanced signaling via the BCR has been implicated to contribute to disease progression in CLL,³⁰  the capacity of miR-155 to influence the responsiveness to BCR ligation may explain the noted association between high levels of miR-155 and adverse outcome of patients with this disease.

The microRNA miR-155 also can regulate expression of other mRNAs encoding other proteins, such as the suppressor of cytokine signaling (SOCS)1,^40,41  transforming growth factor (TGF) β-1, TGF β-2, or TGF β-receptor type 2, which may play a role in tumor invasion and/or metastasis.⁴²  However, other than INPP5D, none of the known targets of miR-155 encode proteins that can directly influence BCR signaling. Similarly, SHIP1 expressed in hematopoietic cells could be repressed by overexpression of miR-155, which was associated with increased activation of the kinase AKT during the cellular response to lipopolysaccharide.²⁶  In any case, our results here indicate that the relative levels of miR-155 could influence the proficiency of BCR signaling in CLL.

Although the average leukemia cell expression levels of miR-155 can be used to segregate patients at different relative risk for disease progression, the levels of miR-155 can vary between CLL cells of the same patient. Prior studies by Wang and colleagues observed high-level expression of miR155HG (BIC), the precursor to miR-155, in prolymphocytes and presumably CLL cells within the proliferation centers (PCs), which had high proportions of cells staining with the proliferation marker Ki-67.¹⁹  They noted that the cells within the PCs expressed high levels of BIC/pre-miR-155 relative to the lymphocytes outside the PC via RNA in situ hybridization. The relative number BIC/pre-miR-155–positive cells varied from patient to patient, depending on the relative size and number of PCs present in the LN. We interrogated the GEO database (GSE21029) and noted that CLL cells in lymphoid tissues expressed higher levels of miR155HG, and apparently lower levels of INPP5D, than CLL cells in the peripheral blood. These data suggest that the levels of miR-155 can vary depending on where the CLL cells reside.

However, CLL cells are not fixed in their anatomic distribution, but rather recirculate between blood, lymphoid, and marrow compartments.⁴³  Prior studies indicated that a small proportion of CLL cells in the blood likely represent cells that have reentered the circulation after exiting the lymphoid tissues.³³  In particular, CD5^brightCXCR4^dim CLL cells in the blood represent newly released cells from the tissue microenvironment, whereas the CD5^dimCXCR4^bright CLL cells represent cells that may be about to reenter the tissue microenvironment.³³  In the present study, we found that the subgroup of blood CD5^brightCXCR4^dim CLL cells expressed higher levels of miR-155 and lower levels of SHIP1 protein, and were more sensitive to surface-µ ligation than the “resting” subgroup of blood CD5^dimCXCR4^bright CLL cells (Figure 6). As such, this is the first study to define differences in the expression of miR-155 by CLL cells of the same patient and to find that this can result in functional differences in the leukemia cells’ responsiveness to BCR ligation.

We hypothesized that survival/stimulatory signals from accessory cells present in the leukemia cell microenvironment, such as in the PCs of the LNs, might contribute to the upregulation of miR-155 (Figure 6). To test this hypothesis, we stimulated CLL cells with CD154 or BAFF, each of which could serve to activate nuclear factor κB (NF-κB).^44,45  Prior studies found activation of NF-κB could induce expression of BIC, the precursor to miR-155.⁴⁶  However, in some B-cell lymphomas or B-cell lines (eg, Ramos), NF-κB–induced expression of BIC did not necessarily cause increased expression of miR-155, indicating that there might be 2 levels of regulation for generating mature miR-155: 1 at the transcriptional level involving NF-κB and 1 at the processing level.⁴⁶  The studies reported here demonstrate that exposure of CLL cells to factors that can activate NF-κB significantly upregulated expression of mature miR-155, suggesting that in CLL the control of miR-155 may predominately be at the level of BIC transcription.

In any case, we found that the changes induced in the expression levels of miR-155 upon treatment with CD154 or BAFF had functional significance; such treatment significantly reduced expression of SHIP1 and enhanced the responsiveness to BCR ligation, effects that could be mitigated by transfection of CLL cells with an inhibitor to miR-155. Similar effects of miR-155 also were noted for normal B cells that were stimulated via CD40 ligation with CD154, indicating that miR-155 might play a physiologic role in regulating the B-cell response to BCR ligation. Within the lymphoid tissue microenvironment, T-cells or accessory cells may express factors, such as CD154 or BAFF/APRIL, which can enhance expression of miR-155. This in turn may serve as a rheostat that can enhance the signaling derived through BCR ligation (Figure 6). Conversely, attenuation in expression of miR-155 in CLL or normal B cells when they circulate in the blood may allow for increased expression of SHIP1 and decreased responsiveness to BCR ligation. This could mitigate the risk for inadvertent stimulation by self- or foreign antigen(s) of blood B cells outside of lymphoid tissue compartments, where they would not have the microenvironment and accessory cells necessary for antigen-stimulated B cells to differentiate in an immune response to antigen. As such, miR-155 might play an important role in the physiology or normal B cells and/or the pathophysiology of neoplastic B cells.

These studies also have implications for development of new therapies that disrupt the capacity of CLL cells to home to tissue microenvironments or that inhibit the signaling from accessory cells within the microenvironment, thereby potentially interfering with the capacity of the microenvironment to upregulate miR-155. Agents that block the capacity of CLL cells to home or to engage with accessory cells in PCs might be expected to decrease the levels of miR-155 and increase expression of INPP5D.¹⁶  Monitoring for these changes in patients who are treated with such agents might become a useful surrogate marker for the capacity of the drug to cause diminished leukemia cell stimulation in the lymphoid compartment. Also, agents that reduce expression of miR-155 or that upregulate expression of SHIP1 might have therapeutic value in patients with CLL, particularly for those patients who have CLL cells with high levels of miR-155. Moreover, agents that inhibit leukemia cell expression of miR-155 may have clinical activity in patients with this disease.

This work was supported in part by the National Institutes of Health, National Cancer Institute grants (R37-CA049870) (T.J.K.) and (PO1-CA81534 of the CLL Research Consortium) (L.Z.R., C.M.C., and T.J.K.), the Blood Cancer Research Fund, the South Moravian Programme for Distinguished Researchers (M.M) cofinanced by the EU and the South-Moravian Region, and the EHA Research Fellowship award (granted by the European Hematology Association).

Contribution: B.C. designed research, performed research, analyzed data, and wrote the paper; L.C. designed research, performed research, analyzed data, and wrote the paper; S.Z., M.M., J.-F.F., J. Y., E.M.G., L.Z., and L.Z.R. performed research; L.B. and K.M. analyzed data and contributed to the manuscript; G.A.C. and C.M.C. made scientific contributions; and T.J.K. designed research, analyzed data, and wrote the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Thomas J. Kipps, 3855 Health Sciences Dr, Rm 4307, San Diego, CA 92093-0820; e-mail: tkipps@ucsd.edu.