Gene Expression Profiling Reveals Aberrant T-cell Marker Expression on Tumor Cells of Waldenström's Macroglobulinemia

Waldenström's macroglobulinemia (WM) is a rare chronic lymphoproliferative disorder characterized by immunoglobulin M (IgM) paraproteinemia and bone marrow infiltration by lymphoplasmacytoid cells (1, 2). The World Health Organization (WHO) classification defined it as one of the subtypes of non-Hodgkin B lymphoma (NHL; ref. 3). Although WM itself is clinically heterogeneous, ranging from largely asymptomatic disease to the presence of symptoms related to monoclonal IgM deposition and tissue infiltration (4, 5), the major clinical manifestations include cytopenia, paraprotein-related cryoglobulinemia, cold agglutinin syndrome, demyelinating neuropathy, and symptomatic hyperviscosity (6).

Unlike the related gammopathies, multiple myeloma, and non-IgM monoclonal gammopathy of undetermined significance (MGUS), WM is characterized by a conspicuous absence of DNA ploidy or involving the immunoglobulin heavy-chain locus (IgH) 14q translocations (7). Deletion of chromosome 6q, seen in approximately 30% of cases, represents the most recurrent genetic lesion identified (8), but the clinicopathogenetic consequences of this abnormality are not yet understood. Recently, a major breakthrough was that Treon and colleagues discovered highly recurrent MYD88 L265P (92%) and CXCR4 (30 ∼ 40%) somatic mutations in WM using whole-genome sequencing (9). Similar to the WM, the IgM MGUS, the precursor of WM (10), shows a high frequency of MYD88 mutation but much less structural variants (11). Although a few articles using global gene expression profiling (GEP) or deep-sequencing have been reported and gained insight into the enigmatic genetics of this disease, a precision targeting of these mutations and signaling pathways are still challenging in WM clinical treatment.

WM is characterized immunophenotypically as consisting of cells expressing markers of mature B cells with varying degrees of plasma cell differentiation (12). Given this heterogeneity, we purified both lymphoid and plasmacytoid components from WM bone marrow and compared the expression profiles of these cells with those of normal peripheral blood B cells, tonsil B cells, tonsil plasma cells, bone marrow plasma cells, and multiple myeloma plasma cells. Here, we report on the molecular signatures of WM in the context of both normal B cells and plasma cells, and multiple myeloma. Our discovery that T-cell antigens express in WM-BC may explain well for drug resistance and develop a novel treatment strategy in WM.

Deidentified clinical bone marrow aspirates were obtained from patients with WM at the University of Arkansas for Medical Sciences (UAMS, Little Rock, AR), and the Institute of Hematology & Blood Diseases Hospital, Chinese Academy of Medical Science & Peking Union Medical College (Tianjin, China). Studies were approved by the Institutional Review Boards at each institution. Informed consent was obtained in accordance with the Declaration of Helsinki. Mononuclear cells were obtained from bone marrow aspirates from 19 patients with a new diagnosis of WM. Expression of CD20 and CD138 on lymphoid cells was examined by flow cytometry to determine the dominant phenotype of the tumor cell population and the choice of antibody for cell enrichment. Tumor cells expressing B-cell (CD19 or CD20) or plasma cell (CD138) markers were isolated with the use of immunomagnetic beads according to the manufacturer's guidelines (Miltenyi Biotec). WM-BC and WM-PC purities of >90% homogeneity were confirmed via two-color flow cytometry using CD38⁺/CD45⁻ and CD20⁺/CD45⁺ staining criteria, respectively (Becton Dickinson). Cell isolation with immunomagnetic beads and analysis of the other cell types used in this study have been previously reported (13). Of the 19 cases analyzed, both CD19 and CD138 cells were isolated from 2 cases by separating the sample as two parts, only CD19 cells from 10 cases, and only CD138 cells from 7 cases. The sample cohort studied consisted of CD19-selected peripheral blood B cells (PB-BC; n = 7), tonsil B cells (T-BC; n = 7), bone marrow B cells from WM (WM-BC; n = 12), tonsil plasma cells (T-PC; n = 9), bone marrow plasma cells from healthy donors (N-PC; n = 10), WM plasma cells (WM-PC, n = 10), and multiple myeloma plasma cells (MM-PC; n = 10; ref. 14). Tonsils were obtained from patients undergoing tonsillectomy for chronic tonsillitis (14).

Detailed protocols for RNA purification, cDNA synthesis, cRNA preparation, and hybridization to the Human Genome U95Av2 GeneChip Microarray (Affymetrix) have been described previously (13–15). The Gene Expression Omnibus database accession number performed in this study is GSE9656.

All data used in our analyses were derived from Affymetrix Microarray Suite 5.0 output files. Expression values are notated (i) as a signal representing the difference between the intensities of the sequence-specific perfect match and the mismatch probe set, or (ii) as a signal indicating a “present,” “marginal,” or “absent” call as determined by the GeneChip 5.0 algorithm. Gene arrays were scaled to an average signal of 1,500 and then analyzed independently. Signal calls were transformed by the log base 2.

The statistical software package SPSS 12.0 (SPSS Inc.) and the significance analysis of microarrays (SAM) method were used to analyze the data (16). Genes were selected for analysis on the basis of detection and FDR. In each comparison, genes having “present” detection calls in more than one-third of the samples in the overexpressed gene group were retained for statistical analysis. For two-class and multiclass supervised analyses, the SAM method was used with sample-label permutations to evaluate statistical significance.

Hierarchical clustering of average linkage with the centered correlation metric was employed (17).

The modified “spike” method was used to define unique, highly overexpressed genes in WM (13). After the elimination of control genes and the filtering out of genes with a maximum signal of <500 across all samples, the median expression value (log base 2) of each gene across all samples, including PB-BC, T-BC or T-PC, N-PC, WM-BC or WM-PC, or MM-PC, was determined. The value of the i^th gene was called medgene (i). The i^th gene was called a “spike” gene if it had at least three (>14%) sample expression values >2.8+ medgene (i) in WM. All genes having low to undetectable expression in the all-normal controls and multiple myeloma were defined as specifically expressed genes in WM.

Mononuclear cells were obtained from bone marrow aspirates from 5 newly diagnosed patients with WM. About 0.5–1 × 10⁷ mononuclear cells were suspended in 100 μL of flow cytometry buffer and WM-BCs were gated with double positive CD45 and CD19. The strong CD45⁺CD19⁺ cells further analyzed the expression of CD3 and CD8 antigens by flow cytometry to determine T-cell marker expression. All analyses were performed on an LSRII flow cytometer (Becton Dickinson).

FACS was performed to sort CD45⁺CD19⁺ cells from above 5 WM samples. Cytospin slides were prepared and immunofluorescence was carried out using mouse anti-human CD3 (1:500) and rabbit anti-human CD19 (1:50) antibodies (Novocastra Lab Ltd). Cell nucleus was stained with DAPI on WM-BCs. Normal CD3⁺ and CD19⁺ cells were isolated from peripheral blood of healthy donors as positive controls.

WM morphology is marked by small lymphocytes with a variable degree of plasma cell maturation with an immunophenotype of CD19⁺/CD20⁺/CD5⁻/CD10⁻/Ig⁺/sIgM⁺⁺ and distinct histopathology (18–20). Gene expression levels for CD45, CD20, CD79B, CD52, CD19, CD22, CD83, and CD72 showed no differences in comparisons between WM-BCs, PB-BCs, and T-BCs (Supplementary Table S1). Expression levels of CD5, CD23, and CD1, which are high in chronic lymphocytic leukemia, showed a low intensity in WM-BCs (data not shown). Interestingly, although selected with CD19, the plasma cell–associated markers CD138, CD38, and CD27 were significantly higher in WM-BCs than in other B-cell samples (Fig. 1A). Similarly, expression levels of CD45, CD19, and CD20 were significantly higher in WM-PCs than in other plasma cell samples, whereas plasma cell antigens CD38 and CD138 were significantly lower (Fig. 1A). CD27, a soluble protein increased in WM serum and linked to hemoglobin levels in patients with WM (21, 22), was expressed at similar high levels in WM-PCs, T-PCs, and N-PCs, while this gene was expressed at a much lower level in MM-PCs.

Plasma cell differentiation is controlled by the differential activation and inactivation of several transcription factors (23). The interferon-regulatory protein 4 (IRF4) and X-box–binding protein 1 (XBP1) genes are activated, whereas PAX5 and BCL6 are inactivated. XBP1 showed low levels in WM-BCs but a gradual increase in expression from PB-BCs to T-BCs to WM-BCs and similar high levels in the plasma cell samples. PRDM1 expression was lowest in PB-BCs; expression was higher and at similar levels in T-BCs and WM-BCs. PAX5 expression was lowest in WM-PCs, showing similar expression levels to those of T-PCs, N-PCs, and MM-PCs. IRF4 was expressed at higher levels in WM-BCs than in other B-cell samples, but its expression was higher still in WM-PCs, whose expression level was similar to that of T-PCs; N-PCs, and MM-PCs expressed IRF4 at still higher levels. BSAP/PAX5 showed a progressive loss in expression from PB-BCs to T-BCs to WM-BCs. Although much lower than the B-cell samples, expression levels of both WM-PCs and T-PCs were similar and much higher than those observed in N-PCs and MM-PCs (Fig. 1B).

In an attempt to better define molecular signatures of WM-BCs, we performed an unsupervised hierarchical cluster analysis of 6,504 genes exhibiting the greatest variation across all B-cell samples (SD > 0.6; Fig. 2A). This produced a sample dendrogram with two major branches, one containing the PB-BC cases and the other having the T-BC and WM-BC cases on two subbranches. These data showed that WM-BCs have a unique gene expression signature relative to other normal B-cell populations that were also selected on the basis of the expression of CD19. A similar strategy was employed to evaluate global differences in expression patterns in the plasma cell samples (Fig. 2B). In this case, the unsupervised cluster analysis produced two major branches one containing the T-PCs and the other having the bone marrow–derived N-PCs and MM-PCs. Unlike with the WM-BCs, the WM-PCs failed to segregate into a separate sub-branch; instead, these samples were either dispersed within the T-PC cases (n = 5) or the MM-PC cases (n = 4). These data show that although derived from the bone marrow, CD138-selected PCs from WM are not normal and have molecular features of immature plasma cells (consistent with data in Fig. 1) or multiple myeloma. The clinical implications of these results are not known; however, they may indicate that at least two types of WM exist and a different therapeutic strategy may be needed for each type.

To identify genes uniquely expressed in WM-BCs, we applied SAM analysis with a 0.1% FDR to the B-cell samples. This analysis revealed 55 genes that were differentially expressed by greater than 2-fold in WM-BCs, including 31 upregulated and 24 downregulated genes common to the WM-BC samples (Supplementary Table S2). A colorgram of the gene expression patterns of these 55 genes in the B-cell samples clearly demonstrated significant differences in expression across the samples (Fig. 3A). Genes exhibiting greater than 10-fold overexpression in WM-BCs included the chemokine receptor CCR2, the proto-oncogene MYB, and the early B-cell genes IGLL1, RAG1, and RAG2. Although not as uniformly overexpressed, the T-cell gene fibrinogen-like protein 2 (FGL2) was highly overexpressed in a subset of WM cases.

An identical SAM analysis of WM-PCs, T-PCs, and N-PCs revealed 242 differentially expressed genes, 46 of which were differentially expressed by more than 2-fold (Supplementary Table S3). As would be expected, the IGHM gene was uniquely overexpressed in WM-PCs. Most notably, while IL6R was overexpressed in T-PCs and WM-PCs, WM-PCs also significantly overexpressed the IL6 gene. These data suggest that hyperactivation of both the ligand and the receptor of this important plasma cell survival–signaling cascade may play a pathologic role in WM. Protein kinase C, beta 1 (PRKCB1), which plays a major role in early B-cell survival and function, was also upregulated. Among other transcripts overexpressed in WM-PC were genes involved in cell-cycle regulation (PAK2 and DDIT1); the transcription activator (NFATC4); the transcription repressor proto-oncogene (BCL6), which normally functions to block B-cell terminal differentiation; the gene for DNA methyltransferase 1 (DNMT1); WHSC1/MMSET, which is aberrantly activated in MM by the t(4;14)(p16;q32) translocation; the insulin-like growth factor receptor (IGF2R); an iron transporter (TRFC); and a few genes of unknown function. Of the 30 underexpressed genes, the cytoskeleton-related gene for actin-binding protein (SNL) was the most significantly underexpressed gene (4.9-fold); two other cytoskeleton-related genes, P4HA1 and P4HA2, were also included in this group. Three metabolism-related genes, including PLTP, MGLL, and ENO2, were underexpressed 2.1–3.4-fold. Also significantly underexpressed in WM-PCs were the transcription factors FOXO3A and SP3; the cell adhesion genes TM4SF7 and CD6; the transporter-related genes ABCG2, VAPB, ATP5D, and ACATN; the metalloproteinase gene TIMP2; an amyloid-related gene ITM2B; and the autoantigen gene RALY.

To further define the differentially expressed genes in WM-PCs, the 46 most significantly differentially expressed genes relative to N-PCs were further analyzed by examination of their expression levels in T-PCs. Genes encoding B-cell- and plasma cell markers, including CD45, CD20, CD79B, CD52w, CD138, and CD38, that were found to be up- or downregulated in WM-PCs relative to N-PCs were not differentially expressed in a comparison between WM-PCs and T-PCs (data not shown). Only 17 (37%) of 46 genes were found to be significantly differentially expressed in the comparison between WM-PCs and T-PCs. This finding suggests that although derived from marrow, WM-PCs are more similar to T-PCs than N-PCs, a feature consistent with data presented in Fig. 3B.

To explore the relationship between malignant plasma cells from two related gammopathies, we compared the gene expression profiles of WM-PCs and MM-PCs. Among the 46 genes differentially expressed in WM-PCs relative to N-PCs, 17 (37%) were also identified as dysregulated in multiple myeloma. Twenty-nine genes (63%) showed no changes in WM-PCs relative to MM-PCs (Supplementary Table S3). The commonly upregulated genes included two cytokine genes (IGF2R and IL6R) and a cell-cycle–related gene (DDIT1). The commonly downregulated genes included adhesion- and cytoskeleton-related genes (SNL, TM4SF7, and CD6). These genes may be more important for tumor cell proliferation and survival. Interestingly, genes encoding some B-cell CD markers (Supplementary Table S1) were identified as significantly overexpressed in WM-PCs relative to MM-PCs. Examples include CD45 (9.6-fold), CD20 (2.4-fold), CD79B (3.0-fold), CD52w (15.6-fold), CD19 (3.2-fold), and CD22 (2.8-fold). These findings suggest that WM-PCs, expressing CD138 and other plasma cell differentiation markers, continue the expression of genes associated with earlier B-cell development. It is also noteworthy that no transcription factor associated with plasma cell differentiation, including XBP1, BLIMP1/PRDF1, and IRF4, was found to be significantly differentially expressed in the comparison between WM-PCs and MM-PCs.

Multiple myeloma is characterized by “spiked” gene expression patterns for many genes in unique subsets of patients. Expression of CCND3, CCND1, MAF, MAFB, and FGFR3/MMSET are undetectable in normal bone marrow plasma cells, and low (CCND1 only) to undetectable in most multiple myeloma cases. However, as a result of chromosomal translocations between these genes and the IGH locus, these genes exhibit a characteristic “spiked” pattern in specific subsets of tumors. Although WM has been shown to lack translocations involving the IGH locus (24), we nevertheless applied to WM samples a method used to identify spiked gene expression in MM. This strategy was employed to identify genes that may not be uniformly overexpressed across the entire group but might nevertheless be highly overexpressed in subsets of patients and therefore may point to subtypes of disease. Table 1 lists 40 genes with expression values >2.8+ medgene (i) (log₂ of the signal) in at least three WM cases. PDGFRA, encoding receptor tyrosine kinase, was upregulated in 15 of 21 WM samples and expressed at very high levels (Affymetrix signal >10,000) in 9 of 21 samples. CD24, SOX4, MYB, and WAV, mapping to chromosome 6, represented the largest group of genes mapping to one chromosome. Several signaling-related genes were in this list, including IL7R, SCYA4, SCYA5, CX3CR, HCK, and PTGER2 (a prostaglandin E₂ receptor and COX2 downstream gene). Also represented were Annexin 1 (ANNX1), a cell surface receptor linked to caspase-2 apoptosis and calcium metabolism, and TYROBP, which encodes tyrosine kinase–binding protein, which acts as a docking site for Syk or ZAP-70 protein tyrosine kinases in natural killer cells (25).

Consistent with the finding of overexpression of the T-cell marker Fgl2 in WM-BCs (see Fig. 3A), this analysis revealed that a panel of T-cell–related genes were overexpressed in a subset of WM cases. These genes included TCRA, CD2, CD3D, CD3G, CD3Z, and CD8A. WM cell expression of T-cell markers was also accompanied by expression of CTL-associated serine esterase 3 (GZMA), granzyme 3 (GZMK), granzyme (GZMM), a cell adhesion gene (NK4), and cytokines (IL7R and SCYA5). In 7 patients (34%), genes associated with overexpression of T-cell markers were identified as being coactivated in WM-BC (Fig. 4A).

To validate whether T-cell markers identified by GEP are truly expressed on WM-BCs, flow cytometry and immunofluorescence were used to detect the expression of CD3 and/or CD8 antigens on CD19⁺CD45⁺ cells collected from WM patient samples. Five WM patient samples were collected from bone marrow aspirate and determined CD19⁺CD20⁺ tumor cells (Supplementary Table S4). As shown in Fig. 4B, the strong CD19⁺CD45⁺ cells were sorted by FACS and further analyzed for CD3 and CD8 expression. The results clearly showed both CD3 and CD8 to be expressed in subsets of cells (mean: 4.9%; range: 0.1%–13.4%) within in the CD19⁺CD45⁺ tumor cells. We also examined CD3 expression in CD19⁺ WM cells from the same five patient samples using immunofluorescence. Both anti-CD3 and anti-CD19 antibodies were tested in the normal T cells and B cells, respectively (Supplementary Fig. S1), representing positive controls. Double staining with CD3 and CD19 confirmed that a subset of CD19⁺ cells expressed CD3 (Fig. 4C).

WM is a distinct clinical entity characterized by IgM monoclonal gammopathy and infiltration of the bone marrow by lymphoplasmacytoid tumor cells. The expression profiles of WM compared with other B-cell–malignant as well as normal peripheral blood B cells and bone marrow plasma cells have recently been published (26, 27). However, none of them identified WM-related genes differentially expressed during different stages of BC development. In this study, we show that comparative microarray profiling of CD19-enriched tonsil B cells, peripheral blood B cells as well as CD138-enriched PCs from tonsil and bone marrow allowed to gain more comprehensive insights into the biology of WM. GEP studies in this type of lymphoma are challenging, as the malignant clone demonstrates significant intraclonal heterogeneity with respect to plasmacytoid differentiation. WM is a low-grade B-cell or plasma cell tumor. Evidence has accumulated, indicating that malignant B cells display a variety of physiologic activities and requirements possessed by normal cells at similar stages of differentiation (24). We show that WM-PCs differ from other plasma cells in the overexpression of B-cell markers, such as CD45, CD20, CD79B, and CD52. Likewise, WM-BCs exhibit much higher expression levels of the plasma cell markers CD138, CD38, and CD63. The high expression of BCR in both WM-BCs and WM-PCs indicates that antigen stimulation may be important to maintain tumor cell viability and the immature status of WM-PCs. In line with these results, WM-PC cases show a similar gene expression pattern to that of immature T-PCs. Unfortunately, we could not collect age-matched normal bone marrow B cells as a control in this study, which will be performed in the future.

Several of the genes identified as overexpressed in WM-BCs or WM-PCs in this study may provide novel insights into WM disease biology, as well as potential targets for therapeutic intervention. For example, this analysis shows that the oncogene MYB is commonly dysregulated in WM, with a greater than 100-fold increase in expression over normal counterparts. MYB activation, in particular, may be important, as activation of this gene has been shown to induce plasmacytoid lymphosarcomas in mice and B-cell lymphomas in chickens (28, 29). Therefore, further study is needed to understand the basis and consequences of MYB dysregulation in WM. BCL6 is another candidate oncogene that may be dysregulated in these tumors. Dysregulation of BCL6, an important transcription repressor that blocks plasma cell differentiation (30), may also be linked to the observed overexpression of IGLL1, which has been studied for its BCL6-rearranging properties in lymphoma (31). Another candidate oncogene is BLIMP1/PRDMI, which encodes a zinc-finger transcription factor that drives the maturation of B lymphocytes to immunoglobulin-secreting plasma cells in vitro and in vivo (32, 33). Importantly, the dysregulation of oncogenes such as BCL6 may also be related to prognosis for WM, as in other lymphomas (34). Consistent with previous reports, the src family (HCK) and the gp130 cytokine family (IL6 and IL6R) were significantly overexpressed in WM. HCK belongs to the src family of tyrosine kinases, which has also been implicated in B-cell proliferation and malignancies, as well as a direct target of BTK inhibitor (35, 36). IL6 is known to play a critical role in normal and malignant plasma cell growth and survival (26, 27, 37). The upregulation of both IL6 and its receptor on WM cells creates the possibility of an autocrine signaling cascade in this disease. Interestingly, it was reported that the IL6/IL6R signaling pathway could enhance HCK kinase activity including multiple myeloma (38, 39).

This analysis also identified dysregulation of key genes in several signaling pathways not previously implicated in WM. These genes include the protein kinase C family (PRKCB1) and the receptor tyrosine kinases (PDGFRA). Protein kinase C-β plays a critical role in signaling through the B-cell receptor and in early B-cell survival. Upregulation of PRKCB1 in WM as observed in this study, along with data from other mature B-cell tumors (40), suggests dysregulation of this pathway in mature B-cell/plasma cell malignancies as well. Indeed, overexpression of PRKCB1 represents a poor prognostic signature in large B-cell lymphomas in one study (34). PDGFRA, encoding receptor tyrosine kinase, was found to be upregulated in most patients with WM. PDGFRA has previously been reported to enhance medulloblastoma migration and activation of downstream Ras/MAPK pathway (41).

Although the expression of T-cell markers has been considered to be exclusive to the T-cell linage and has not been found on normal B cells, a panel of T-cell markers and CTL-associated genes, including TCRA, CD2, CD3D, CD3G, CD3Z, CD8A, GZMA, GZMK, and GZMM, were found to be differentially upregulated in a subset of patients with WM. To our knowledge, this is the first indication that T-cell markers are involved in WM. Other genes involved in T-cell biology, namely TYROBP, FGL2, IL7R, and SCYA5, were also identified in this study as being overexpressed in WM, which suggests that WM cells may be responsive to cytokine signaling through T-cell receptors. The expression of p56^lck, one of the first molecules to be activated downstream of the T-cell receptor in CD4⁺ and CD8⁺ thymocytes (42), was highly correlated with CD8A, suggesting that functional CD8A signaling and possibly other T-cell–specific stimuli may affect WM biology. The increased expression of T-cell markers exists only in a small subset of B cells of WM, suggesting this population may have dedifferentiated during tumor development. It will be interesting to examine whether the CD19⁺CD3⁺ WM cells have “stemness” features, such as self-renewal and increased drug resistance in the future. If this is true, targeting T-cell signaling in WM may eliminate drug resistant–stem cell-like tumor cells. This discovery is not unique in WM-BCs, because aberrant T-cell marker expression was also reported in plasmablastic transformation of plasma cell myeloma, chronic lymphocytic leukemia (CLL), diffuse large B-cell lymphoma (DLBCL), and ALK-positive large B-cell lymphoma (43–47).

The data presented here have implications for the development of novel therapies for WM. For example, inhibitors of signaling pathways found to be dysregulated in WM (protein kinase C-β and src kinases) are being developed and should be considered. Imatinib, targeting PDGFRA, or other similar therapeutic agents may prove useful. Approaches to inhibiting IL6 signaling may also be of value in WM. N-acetylcysteine (NAC) targeting IL7R signaling pathways has demonstrated therapeutic efficacy in T-ALL (48, 49). We believe that these therapeutic approaches are best tested in the context of patients with known abnormalities in the respective signaling pathways. Furthermore, studies of single-cell sequencing in WM may improve understanding of disease heterogeneity and optimal application of targeted therapies.

J.D. Shaughnessy Jr holds ownership interest (including patents) in Esoterix, Miragen Therapeutics, and Quest Diagnostics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: G. Tricot, J.D. Shaughnessy Jr, F. Zhan

Development of methodology: M. Hao, J.D. Shaughnessy Jr, F. Zhan

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Hao, B. Barlogie, L. Liu, L. Qiu, J.D. Shaughnessy Jr, F. Zhan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Hao, G. Tricot, L. Liu, J.D. Shaughnessy Jr, F. Zhan

Writing, review, and/or revision of the manuscript: M. Hao, G. Tricot, J.D. Shaughnessy Jr, F. Zhan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Liu, L. Qiu, J.D. Shaughnessy Jr, F. Zhan

Study supervision: G. Tricot, L. Qiu, J.D. Shaughnessy Jr, F. Zhan

This work was supported by NIH (grant no. CA55819, to J.D. Shaughnessy Jr, F. Zhan, G. Tricot, and B. Barlogie), the Leukemia & Lymphoma Society TRP (6549-18), the Multiple Myeloma Research Foundation (to F. Zhan.), the NIH Lymphoma Spore grant P50 CA097274, Cancer Center Support Grant NIH grant P30 CA086862; institutional start-up funds from the Department of Internal Medicine, Carver College of Medicine, University of Iowa (to F. Zhan); Cancer Center Support Grant NIH grant no. P30 CA086862, and the National Natural Science Foundation of China (81570181, to M.Hao; 81630007, to L. Qiu.).

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.