Comparative genomics of CXCR4^MUT and CXCR4^WT single cells in Waldenström’s macroglobulinemia

MYD88 and CXCR4 somatic mutations are the most common alterations in Waldenström’s macroglobulinemia (WM), affecting 95% to 97% and 30% to 40% of patients, respectively.¹ CXCR4 mutations occur in the C-terminal domain and are often subclonal to mutated MYD88 with a median allele burden of 35%, suggesting that these mutations occur after acquisition of the MYD88 mutation. These activating mutations in CXCR4 are typically nonsense or frameshift mutations that impact clinical presentation. The nonsense variants in particular are associated with higher bone marrow disease burden, high serum immunoglobulin M levels, symptomatic hyperviscosity, and earlier time to first treatment. CXCR4 mutations also impact ibrutinib response, including depth and time to major response and progression-free survival.^2,3  Differences in response rates are minor when using a combination of ibrutinib and rituximab, but delayed responses are still observed in CXCR4^MUT patients.^4,5  The acquisition of BTK^Cys481 mutations that underlie ibrutinib resistance also appears more common in CXCR4-mutated patients, suggesting an underlying genomic predisposition. As such, we sought to examine the CXCR4 mutant subclone at a single-cell level to identify the alterations that may explain these unique features.

We selected 1 untreated patient with the clinicopathologic diagnosis of WM, with mutated MYD88 and CXCR4 for this proof-of-concept study. Single-cell sorting was performed on CD19⁺immunoglobulin M⁺ bone marrow mononuclear cells, followed by whole-genome amplification (WGA) of the DNA on each isolated cell with the REPLI-g single cell kit (Qiagen, Valencia, CA). MYD88 and CXCR4 mutational status was assessed by Sanger sequencing.⁶  A total of 22 single cells, distributed in 13 MYD88^MUT/CXCR4^MUT and 9 MYD88^MUT/CXCR4^WT, together with the bulk tumor CD19⁺ fraction and the bulk CD19⁻ germline sample (non–whole-genome amplified) from the same patient, were sent for whole-genome sequencing (WGS) to the Broad Institute of Massachusetts Institute of Technology and Harvard (Cambridge, MA). Data were analyzed following the Genome Analysis Toolkit Best Practice Guidelines (Broad Institute).⁷  Sequencing reads were aligned to the human reference genome GRCh37/HG19 using Burrows-Wheeler Aligner. Small variants and indels were called using Strelka⁸  and Ensembl Variant Effect Predictor,⁹  and copy number alterations (CNA) were analyzed using Control-FREEC (Boeva Laboratory, Institut Curie, Paris, France)¹⁰  and Genome Analysis Toolkit Copy Number Variation (Broad Institute).¹¹  Further analyses, including nonnegative matrix factorization,¹²  Fisher's exact test, and differential gene expression analysis with voom from the edgeR/limma Bioconductor packages,¹³  were conducted in R (R Foundation for Statistical Computing, Vienna, Austria). The pipeline followed is detailed in previous studies of the group.^1,14  All validations were carried out by Sanger sequencing.

Because this was the first experiment of this kind carried out in WM, and considering the potential limitations of the WGA (allele dropout, false positives, and sequence-dependent bias),¹⁵  we were able to establish an appropriate workflow to carry out single-cell interrogation (Figure 1). Briefly, we first performed an unsupervised clustering of the 22 single cells based on the variants of the bulk sample. Then, we applied the Fisher's exact test to identify variants significantly different in the single cells according to CXCR4 status and validated them by Sanger sequencing on the bulk sample and on 22 single cells. Next, we searched for mutations on these genes in a series of patients to see whether they were predominant in CXCR4-mutated or wild-type cases, and we conducted differential gene expression analysis based on CXCR4 in an independent set of patients. Finally, copy number alterations were compared between both groups of cells.

Our first approach was to perform an unsupervised clustering of the 22 single cells based on the variants of the bulk sample by using nonnegative matrix factorization,¹²  but the samples did not cluster according to the CXCR4 status, possibly because of biased amplification or allele dropout of the variants in the single cells. Therefore, we decided to find variants from the bulk sample that were enriched in CXCR4-mutated vs CXCR4 wild-type single cells. Fifty-three single nucleotide variants and 10 indels corresponding to 59 genes were identified (supplemental Table 1). Most of these variants (48 of 63; 76%) were predominant in CXCR4^MUT cells. We selected the variants belonging to genes that are expressed in WM or in healthy donors B cells¹⁶  and located near transcribed genes, ending up with 14 mutations (Table 1). Among the affected genes were MACROD2 and CCSER1, which are associated with chromosome instability,^17,18  and UVRAG, which is involved in DNA damage repair.¹⁹  There were also tumor suppressors, such as BTG2²⁰  and DAB2,²¹  a regulator of the cell cycle (SCAPER),²²  and genes responsible for posttranslational protein modifications (LNX1 and DCUN1D4). The remaining genes (TMEM14B, transmembrane protein; LRMP, lymphocyte protein; SPON1, cell adhesion protein; OSGEPL1, endopeptidase; VTA1, protein involved in vesicle trafficking; EXOC6B, part of the exocyst complex) have a less known role.

We picked up half of the variants for validation in the bulk tumor sample by Sanger sequencing and confirmed all of them. Validation in the 22 single cells was also performed, although in this case, results were hampered by the biased amplification and allele drop out associated with the WGA process.¹⁵  However, LRMP kept the significantly different distribution observed in the single-cell WGS results, being detected almost exclusively in CXCR4^MUT cells (supplemental Table 2). This gene encodes a lymphoid-restricted membrane protein involved in antigen receptor assembly and trafficking during lymphocyte development.²³ 

The following step was to translate these findings to a cohort of WM patients. We searched for mutations in the 14 genes in a series of 46 patients from a previous WGS study¹⁴  to see whether they were predominant in CXCR4-mutated or CXCR4 wild-type WM. Characteristics of this cohort are summarized in supplemental Table 3. Results did not show any significant difference in the distribution of the mutations on these genes between both groups of patients from this cohort (supplemental Table 4). As most mutations were noncoding, we hypothesized that they may influence gene expression. Using our previously published RNASeq cohort of 57 WM patients, we checked to make sure the candidate gene was expressed in healthy donor B cells and/or WM.¹⁶  Then we used differential gene expression analysis between the CXCR4^MUT and CXCR4^WT cohorts to look for evidence of transcriptional dysregulation. Interestingly, from our list of 14 genes, LRMP was significant for differential expression (adjusted P = .003). We then analyzed a larger cohort of patients (n = 284) from an ongoing project, and LRMP was within the top 10 most differentially expressed genes (sorted by adjusted P value), with a fold change of −0.6 (supplemental Table 5). Finally, we analyzed the CNA to look for alterations specific to the CXCR4^MUT subclone, but this patient did not present any CNA of sufficient size to analyze at the single-cell level.

In summary, this is the first single-cell study carried out in WM to characterize the clonal diversity of the disease. Our results have highlighted several alterations associated with the CXCR4^MUT clone in DNA repairing genes and tumor suppressors, suggesting that CXCR4 mutations could be related to the alteration of certain mechanisms rather than to specific genes. The findings may therefore help guide future studies to determine the role of CXCR4 in the clonal evolution of WM. Different treatment approaches may be needed for these patients according to the underlying pathogenic mechanisms associated with the presence of CXCR4 mutation. The analysis tools and workflow provided in this paper will help set up the basis for future studies on single-cell interrogation in WM.

The authors thank the Orszag Family Fund for WM Research, Peter S. Bing, the International Waldenström’s Macroglobulinemia Foundation, and the Leukemia and Lymphoma Society. Z.R.H. was supported by an American Society of Hematology Scholar Award. C.J. is supported by a grant from the Fundación Española de Hematología y Hemoterapia-Fundación CRIS.

Contribution: C.J., N.T., M.G.D., A.K., and L.X. performed the experiments; C.J., G.G.C. and Z.R.H. performed the data analysis; C.J. and Z.R.H. wrote the paper; S.P.T. and Z.R.H. conceived and designed the experiments; X.L., M.M., M.L.G., J.G.C., C.J.P., and G.Y. prepared samples; and J.J.C. and S.P.T. provided patient care, obtained consent, and were responsible for sample collection.

Conflict-of-interest disclosure: J.J.C. has received honoraria and/or research funds from AbbVie, Beigene, Janssen, Millennium, Pharmacyclics, and TG Therapeutics. Z.R.H. has received honoraria from Janssen and Pharmacyclics. S.P.T. has received honoraria and/or research funds from Beigene, Janssen, Pharmacyclics, and Bristol-Myers Squibb. The remaining authors declare no competing financial interests.

Correspondence: Zachary R. Hunter, Bing Center for Waldenström’s Macroglobulinemia, Dana-Farber Cancer Institute, Harvard Medical School, Mayer 544, 450 Brookline Ave, Boston, MA 02215; e-mail: zachary_hunter@dfci.harvard.edu.