We analyzed patients with IgM MGUS and asymptomatic WM who were diagnosed between 2021 and 2022 in Hospital Clínic de Barcelona. The diagnosis was established according to the International Consensus criteria (2). All patients had bone marrow aspirates available, and only WM patients underwent a bone marrow biopsy. WM patients were not exposed to previous treatments. All patients signed an informed consent to collect BM samples in the biobank of Hospital Clínic de Barcelona, in accordance with the Declaration of Helsinki. The study was approved by the institutional review board.

The methodology followed to process the samples is described in the Supplementary Material . In short, BM samples were collected in EDTA tubes for further characterization of B cells and plasma cells using multiparametric flow cytometry (MFC) ( Supplementary Table 1 ), detection of the MYD88 L265P by ddPCR, targeted next generation sequencing (NGS) and mRNA expression analyses. Regarding the MFC gating strategy, CD4+, CD8+ and NK cells were quantified relative to the total CD3+ T-cell population or events. B cells and plasma cells were quantified relative to the total number of events (bone marrow cellularity). NGS was performed after CD19+ isolation by immunomagnetic beads (Milteny Biotec, Bergisch Gladbach, Germany). RNA was extracted from whole BM samples after three steps of erythrocyte lysis, centrifugation followed with the supernatant discarded to capture the signal from all immune cells.

DNA from the CD19+ enriched samples was used for NGS using a panel of 54 target genes ( Supplementary Table 2 ) that are recurrently mutated in mature B cell neoplasms (SOPHiA Lymphoma Solution, SOPHiA Genetics, Rolle, Switzerland). A minimum of 100 ng of DNA was used for library preparation, following the recommendations of SOPHiA Genetics. Libraries were then sequenced using 150 bp paired end reads on a NextSeq 2000 P1 instrument (Illumina, San Diego, CA, USA), with a mean coverage of 1500x. The analysis of the sequencing data is described in the Supplementary Material .

The isolated RNA was quantified using Qubit 2.0 Fluorometer (Thermo Fisher). At least 300 ng of RNA per sample was used in the nCounter instrument (NanoString Technologies, Inc. Seattle, WA). The panel used was the PanCancer ImmuneProfiling which consisted of 730 genes that are involved in adaptive and innate immune response to cancer and 40 housekeeping genes. The analysis of the mRNA expression is described in the Supplementary Material .

The chi-square test was used to compare proportions across clinical categories. The Kruskal-Wallis test was used to compare quantitative features according to diagnosis. Pearson correlation or Spearman tests were used to obtain a correlation index between numerical values. The log-rank test was used to assess the significance between conditions in PFS, and the Kaplan-Meier method to plot survival curves. For differential gene expression (DGE) analysis, a false discovery rate (q-value) of less than 0.25 was considered significant. All downstream analyses were performed using R (v. 4.4.0, R Foundation for Statistical Computing, Vienna, Austria).

A total of 36 patients (19 IgM MGUS and 17 WM) were evaluated. Table 1 summarizes their clinical characteristics. WM patients exhibited higher levels of both serum IgM and bone marrow lymphocytic infiltration, along with lower serum albumin levels compared to those with IgM MGUS. In addition, WM patients had a trend towards lower serum IgG and higher β2-microglobulin levels than IgM MGUS. The majority of IgM MGUS and WM patients had kappa-light chain restriction, and no differences were observed regarding the plasma cell bone marrow infiltration between the two groups. All patients with WM had a clonal B-cell population detected in the BM except for one who was diagnosed with lymphoplasmacytic lymphoma in the lymph nodes. On the contrary, only 7 (36.8%) patients with IgM MGUS had a B-cell clone in the BM. The size of the B-cell clones measured by MFC was higher in WM compared to IgM MGUS (24.5% vs. 85.5%; p<0.001). We also assessed the expression of antigen markers on B-cell clones that are commonly associated with IgM gammopathy. In fact, the presence of CD22^low (36.8% vs. 94.1%, p<0.001) and CD25+ (36.8% vs. 76.5%, p=0.017) were detected in both IgM MGUS and WM patients. Two patients with IgM MGUS had only clonal plasma cells without any detectable B cell clone by MFC. Regarding the MYD88 L265P detection using ddPCR or targeted NGS, the mutation was detected in 15 (78.9%) and 14 (82.4%) patients with IgM MGUS and WM, respectively. The mean MYD88 mutation burden assessed by ddPCR was 1.22% and 8.63% in patients with IgM MGUS and WM (p<0.001), respectively. The MYD88 mutation burden was partially correlated with the proportion of clonal B cells (r=0.54, p<0.001) and plasma cells (r=0.40, p=0.024) in the BM detected by MFC. Seven (36.8%) MYD88 mutated IgM MGUS patients did not have any detectable clonal B or plasma cells by MFC. Figures 1A, B summarizes the processing of samples and associations between the MYD88 mutation status and MFC. According to the Spanish risk model, 78.9% and 82.4% of IgM MGUS and WM patients were categorized as intermediate or high risk, respectively. The MYD88 mutation was prevalent in 12 (80.0%) out of 15 intermediate and high-risk IgM MGUS, and 11 (78.6%) out of 14 intermediate and high-risk WM. In total, six high-risk and 3 intermediate-risk patients progressed. On the other hand, the DFCI risk model classified most patients as low and intermediate risk, with 3 high-risk, 4 intermediate-risk and 2 low-risk patients experiencing disease progression ( Figure 1C ). These findings led us to use the Spanish model as the framework for further analyses. After a median follow-up of 2.62 years (interquartile range [IQR] 2.32 – 3.10), high-risk patients had an increased cumulative probability of progression compared to intermediate- (p=0.013) and low-risk patients (p=0.010) ( Figure 1D ). Seven WM and 2 IgM MGUS patients experienced disease progression.

Regarding the MYD88 status, there was no association between risk groups and the mutation prevalence (85.7%, 80.0%, 77% for low-, intermediate-, and high-risk, respectively; p=0.920) assessed either by ddPCR or NGS. The mean MYD88 mutation burden by ddPCR was 6.0%, 1.80% and 9.9% for low-, intermediate-, and high-risk, respectively (p=0.065). The Pearson correlation between the MYD88 L265P VAF detected by NGS and the mutation burden by ddPCR was 0.82 (p<0.001) ( Supplementary Figure 1A ). A total of 93 mutations were detected in 34 (94.4%) patients using NGS ( Supplementary Table 3 ). Two patients without detectable mutations by NGS were one IgM MGUS and one WM, and both were categorized as intermediate risk. The first one had a MYD88 L265P mutation burden by ddPCR of 0.12%, while the second one did not have the mutation by ddPCR. Based on diagnosis, 14 (73.7%) IgM MGUS and 14 (82.4%) WM patients had the MYD88 L265P mutation by NGS (n=36). The mean VAF of the MYD88 L265P by NGS was 7.81% and 29.3% in IgM MGUS and WM, respectively.

The second most frequent mutation involved was CXCR4, in which 4 (21.1%) IgM MGUS and 7 (41.1%) WM patients had at least one mutation. The mean VAF of CXCR4 mutations was 1.7% and 12.0% in IgM MGUS and WM, respectively. KMT2D mutations were prevalent in 4 (11.1%) IgM MGUS and 3 (8.3%) WM patients. Other less frequent mutations had an uneven distribution. For instance, mutations in BIRC3, BCL2, CIITA, KRAS, MEF2B, NFKBIE and NOTCH2 were only detected in IgM MGUS ( Supplementary Figure 1B ). The mutations in CXCR4 were detected only in intermediate- and high-risk groups, harboring also the MYD88 mutation. CXCR4 mutations encompassed frameshift and nonsense alterations such as c.1025C>A; p.Ser342* or c.1012C>T; p.Arg338*, the latter exclusively found in WM. One WM patient who was categorized as intermediate-risk presented a mutation in KMT2D and had disease progression, while 3 had IgM MGUS and were intermediate-risk without presenting progression. ARID1A was altered in 5 (13.8%) patients, being all categorized as intermediate and high risk. Three patients who had ARID1A mutations presented disease progression. EP300 was the fifth gene most frequently altered, and the mutations were only present in intermediate and high-risk groups. EP300 c.2091T>G, previously reported in diffuse large B cell lymphoma (DLBCL) (20), was found in one IgM MGUS and one WM. No recurrent somatic mutations were found in KMT2D, ARID1A or EP300 in patients who progressed ( Table 2 ). Regarding the number of mutations, high-risk patients had a median of 4 somatic mutations, whereas intermediate- and low-risk patients had only 3 and 2 median somatic mutations (p=0.014) ( Figure 2 ).

We next examined the immune cell landscape. We used a previously validated gene signature to deconvolute cell types from our targeted gene expression panel, similar to other studies in solid tumors and DLBCL (21, 22). The proportion of CD8+ (r=0.36; p=0.033), NK (r=0.42; p=0.010) and CD4+ T cells (r=0.37; p=0.030) inferred from gene expression were positively correlated with that obtained from MFC ( Supplementary Figures 2A-C ). We excluded the inferred proportion of B cells and plasma cells to better illustrate the role of immune cells, and merged specific cell types (mast cells, macrophages, NK cells, dendritic cells and CD4 T cells) that contained multiple subtypes to increase the correlation of canonical genes from our targeted mRNA assay and the reference. We found high heterogeneity of the immune cell infiltration in each risk group, characterized by a high proportion of myeloid cells. Meanwhile, the T-cell subpopulations represented a few percentage of the immune cell landscape in each patient. For instance, the mean infiltration rates for CD8+ T cells, monocytes, mast cells and neutrophils were 1.4%, 26.1%, 21.3% and 29.7%, respectively. Among all cell types, the monocyte infiltration showed a consecutive significant decrease from low towards high-risk patients (p=0.040). Of note, mast cells had an increasing trend towards high risk and represented the third most abundant immune cell type (mean of low-, intermediate-, and high-risk groups were 17.1%, 21.5% and 24.4%, respectively) ( Figures 3A, B ). According to diagnosis, the proportion of monocytes, neutrophils and mast cells were similar in both IgM MGUS and WM. Although present in low counts, naïve CD4+ T cells (p=0.022) and dendritic cells (p=0.037) were increased in IgM MGUS patients. We further investigated whether the presence of the MYD88 L265P or CXCR4 mutations were associated with a distinct immune repertoire. The frequency of CD8+ T cells (p=0.017), NK cells (p=0.048) and mast cells (p=0.015) were increased in the MYD88 mutated samples, whereas the neutrophils (p=0.049) proportion was decreased ( Supplementary Figures 2D, E ). CXCR4 mutations was not associated with any particular cell type (data not shown).

In total, 110 genes were differentially expressed between high and low risk and 4 genes between high and intermediate risk ( Figures 4A, B ). The most significant were B-cell related genes, given the increase in the B-cell (mean 9.0%, range 0 to 42.1%) and plasma-cell (mean 2.1%, range 0 to 10%) compartments inferred from gene expression. Only the plasma cells were positively correlated between gene expression and MFC (r=0.40; p=0.034) ( Supplementary Figure 2F ). Genes that showed an increasing gene expression pattern from low towards high-risk groups were BLK, TNFRSF13C, DUSP4, CD28, IL10 and CD44. On the contrary, TOLLIP, CEACAM1 and CR1 were upregulated in low-risk patients ( Figures 4C-K ). Other molecules of interest such as the IGF1R and IGF2R were downregulated in the high risk compared to the low-risk group. The expression of the immune checkpoints CTLA4, TIGIT, LAG3, PD1 and HAVCR2 did not show any differences ( Supplementary Figures 3A-E ). Considering the possibility of capturing gene expression signatures associated with the diagnosis rather than the risk categories, we further performed DGE analysis based on diagnosis. We did not find significant differences, except for the upregulation of ITGA6 in IgM MGUS compared to WM (log2FC 0.85, p<0.001, q-value=0.063), that was also observed in low-risk patients in contrast to high risk (log2FC 1.00, p=0.003, q-value=0.17). No differential expressed genes were found according to MYD88 L265P or CXCR4, KMT2D and ARID1A mutations.

We then performed enrichment analysis by merging low- and intermediate-risks into a single group, in contrast to high-risk. We observed that high-risk patients had a significant association with the NF-κB signaling and an inflammatory response mediated by IFN-gamma and the IL-2 mediated STAT5 activation ( Figure 5A ). Other signatures associated with the high-risk group were related to an inflammatory response mediated by the NF-κB and an activation of Th1 cytotoxic cells. Meanwhile, the low-risk group was characterized by an increase of myeloid-related compartment signatures ( Figure 5B ), reflecting the increased proportion of monocytes and neutrophils. Using the top upregulated genes in high and low/intermediate risk groups to infer protein interactions, we found that the high-risk was associated with the BCR signaling and expression of CD22, the B cell-activating factor receptor (BAFFR or TNFRSF13C), and key components of the BCR (CD79A/B). The low and intermediate risk networks were characterized by B-cell activation based on SYK (CR1, Fc gamma receptor IIa/IIIa, LILRB2, CXCL12, IL6R) ( Figures 5C, D ). SYK and MAPK1 showed a continuous increase from high to low risk ( Supplementary Figures 4A, B ). Along with B cell markers, the intermediate and low risk groups had a strong association with CD9 and CD36 involved in macrophage phagocytosis and inflammation (23, 24). To find the unique differentially expressed genes across risk groups, we obtained the significant genes (p<0.05) specific for each category. High- and low-risk groups had a higher number of unique expressed genes compared to the intermediate group ( Figure 6 ).