Continuous MYD88 Activation Is Associated With Expansion and Then Transformation of IgM Differentiating Plasma Cells

Activating mutations of MYD88 (MYD88L265P being the far most frequent) are found in most cases of Waldenström macroglobulinemia (WM) as well as in various aggressive B-cell lymphoma entities with features of plasma cell (PC) differentiation, such as activated B-cell type diffuse large B-cell lymphoma (DLBCL). To understand how MYD88 activation exerts its transformation potential, we developed a new mouse model in which the MYD88L252P protein, the murine ortholog of human MYD88L265P, is continuously expressed in CD19 positive B-cells together with the Yellow Fluorescent Protein (Myd88L252P mice). In bone marrow, IgM B and plasma cells were expanded with a CD138 expression continuum from IgMhigh CD138low to IgMlow CD138high cells and the progressive loss of the B220 marker. Serum protein electrophoresis (SPE) longitudinal analysis of 40 Myd88L252P mice (16 to 56 weeks old) demonstrated that ageing was first associated with serum polyclonal hyper gammaglobulinemia (hyper Ig) and followed by a monoclonal immunoglobulin (Ig) peak related to a progressive increase in IgM serum levels. All Myd88L252P mice exhibited spleen enlargement which was directly correlated with the SPE profile and was maximal for monoclonal Ig peaks. Myd88L252P mice exhibited very early increased IgM PC differentiation. Most likely due to an early increase in the Ki67 proliferation index, IgM lymphoplasmacytic (LP) and plasma cells continuously expanded with age being first associated with hyper Ig and then with monoclonal Ig peak. This peak was consistently associated with a spleen LP-like B-cell lymphoma. Clonal expression of both membrane and secreted µ chain isoforms was demonstrated at the mRNA level by high throughput sequencing. The Myd88L252P tumor transcriptomic signature identified both proliferation and canonical NF-κB p65/RelA activation. Comparison with MYD88L265P WM showed that Myd88L252P tumors also shared the typical lymphoplasmacytic transcriptomic signature of WM bone marrow purified tumor B-cells. Altogether these results demonstrate for the first time that continuous MYD88 activation is specifically associated with clonal transformation of differentiating IgM B-cells. Since MYD88L252P targets the IgM PC differentiation continuum, it provides an interesting preclinical model for development of new therapeutic approaches to both WM and aggressive MYD88 associated DLBCLs.

Activating mutations of MYD88 (MYD88 L265P being the far most frequent) are found in most cases of Waldenström macroglobulinemia (WM) as well as in various aggressive Bcell lymphoma entities with features of plasma cell (PC) differentiation, such as activated Bcell type diffuse large B-cell lymphoma (DLBCL). To understand how MYD88 activation exerts its transformation potential, we developed a new mouse model in which the MYD88 L252P protein, the murine ortholog of human MYD88 L265P , is continuously expressed in CD19 positive B-cells together with the Yellow Fluorescent Protein (Myd88 L252P mice). In bone marrow, IgM B and plasma cells were expanded with a CD138 expression continuum from IgM high CD138 low to IgM low CD138 high cells and the progressive loss of the B220 marker. Serum protein electrophoresis (SPE) longitudinal analysis of 40 Myd88 L252P mice (16 to 56 weeks old) demonstrated that ageing was first associated with serum polyclonal hyper gammaglobulinemia (hyper Ig) and followed by a monoclonal immunoglobulin (Ig) peak related to a progressive increase in IgM serum levels. All Myd88 L252P mice exhibited spleen enlargement which was directly correlated with the SPE profile and was maximal for monoclonal Ig peaks. Myd88 L252P mice exhibited very early increased IgM PC differentiation. Most likely due to an early increase in the Ki67 proliferation index, IgM lymphoplasmacytic (LP) and plasma cells continuously expanded with age being first associated with hyper Ig and then with monoclonal Ig peak. This peak was consistently associated with a spleen LP-like B-cell lymphoma. Clonal expression of both membrane and secreted µ chain isoforms was demonstrated at the mRNA level by high throughput sequencing. The Myd88 L252P tumor transcriptomic signature identified both proliferation and canonical NF-kB p65/RelA activation. Comparison with MYD88 L265P WM showed that Myd88 L252P tumors also INTRODUCTION Waldenström's macroglobulinemia (WM) is an incurable indolent B-cell lymphoma of the elderly accounting for less than 5% of B-cell lymphomas with, as unique characteristics, a serum IgM peak and primary medullary localization of lymphoplasmacytic cells that exhibit continuous differentiation from mature B lymphocytes to IgM secretory plasma cells (1). Secondary lymphoid organ infiltration and/or a leukemic phase is found in 20% cases. Other manifestations include neuropathy, cryoglobulinemia, skin rash, cold-agglutinin hemolytic anemia, and amyloidosis (2). The discovery of the activating mutation of MYD88 (MYD88 L265P being the far most frequent) in more than 90% of WM cases contributed to the concept that this entity is genetically distinct from other B-cell lymphomas (3,4). Being present in 50% of IgM monoclonal gammopathies of undetermined significance (MGUS), MYD88 mutations are most likely a primary event in WM (5). Considered as secondary genetic events, activating mutations of CXCR4 (CXCR4 S338X or CXCR4 WHIM ), a receptor implicated in migration and bone marrow (BM) homing of leucocytes, are found in 30% of WM cases (6). Additional mutations of CD79b, ARID1A or TP53 have been reported (7).
Despite these advances, WM pathophysiology is incompletely understood. Its treatment remains challenging and the exact role of MYD88 mutations in the emergence of lymphoplasmacytic Bcell clones is not known (7,8). Indeed, MYD88 mutations are also found in 30% of activated B-cell type diffuse large B-cell lymphomas (ABC-DLBCL), more than half of primary cutaneous DLBCLs, leg type, and many DLBCLs at immuneprivileged sites but not in plasma cell myelomas, even IgM types (9). It should be noted that IgM expression is a surrogate marker of ABC-DLBCLs (10). Moreover, all these aggressive B-cell tumors associated with MYD88, which often exhibit morphological features of plasma cell (PC) differentiation, are all associated with expression of the PC differentiation marker IRF4. MYD88 protein is the canonical adapter for inflammatory signaling pathways to downstream members of the Toll-like receptor (TLR) and interleukin-1 receptor (IL-1R) families. Forming the myddosome complex, MYD88 binds IL-1R or TLR family members to IRAK kinases family. IRAK activation leads to activation of the NF kappa B (NF-kB) transcription factor and interferon 3 and 7 regulatory factors (IRF3 and 7). MYD88 L265P constitutively increases formation of the myddosome complex with downstream NF-kB activation (3,11,12).
Experimentally, MYD88 L265P is essential for survival of ABC-DLBCL and WM cell lines (3,11). A recent publication suggests the involvement of HOIP and LUBAC-dependent NF-kB activation in the transformation potential of MYD88 activation in a mouse model (13). The current published mouse models with continuous MYD88 activation in the B-cell lineage develop aggressive clonal B-cell lymphomas that resemble human ABC-DLBCLs (13)(14)(15). Although discussed by Jo et al. in the HOIP/ LUBAC activation context, no IgM peak was reported in these models. Therefore, the question of a direct role for MYD88 in the development of a lymphoplasmacytic lymphoma with monoclonal IgM secretion is still open. Recently, K Schmidt et al. reported a mouse model in which MYD88 activation was responsible for an indolent lymphoproliferative disorder resembling to IgM monoclonal gammopathy of unknown significance (IgM MGUS), the asymptomatic preclinical stage of WM (16).
Here, we present a new mouse model in which the MYD88 L252P protein, the murine ortholog of human MYD88 L265P , is continuously expressed in the B-cell lineage together with Yellow Fluorescent Protein (YFP) (Myd88 L252P mice). We show that these mice first developed early expansion of CD93 neg IgM PCs with an increase in both IgM secretion and bone marrow relocalization of IgM B-cells. Moreover, these mice also had increased percentages of IgM high CD138 low and IgM low CD138 high cells with a CD138 expression continuum between both cell types. Then, these mice developed an oligoclonal or clonal IgM lymphoplasmacytic-like Bcell lymphoma together with a serum IgM monoclonal peak. These tumors had marked transcriptomic similarities to WM but they were located in the spleen and exhibited significant increased proliferation. Despite differences between Myd88 L252P LP-like Bcell tumors and WM, our results demonstrate that the MYD88 transformation potential is strongly associated with a shift in B-cell peripheral differentiation toward plasma cells with IgM secretion. These results help explain why MYD88 activation is found in most WM and in various aggressive B-cell lymphomas with IgM PC differentiation engagement such as ABC DLBCLs.

Generation of the Transgenic Mouse Line Myd88 L252P
The transgenes (cDNA) Myd88 WT -IRES-Yfp and Myd88 L252P -IRES-Yfp were synthetized (Genecust, Dudelange, Luxembourg) and inserted into the pcDNA3.1 vector. Sequences of these transgenes are given in Supplementary Materials and Methods. The Myd88 L252P -IRES-Yfp insert was cloned into the pROSA26-1 vector (17) containing a LoxP-flanked region, consisting of a stop cassette and the Neomycin gene (18). The transgene Myd88 L252P -IRES-Yfp was inserted downstream from the stop cassette. JM8 embryonic stem (ES) cells were transfected with the targeting vector according to a previously described protocol (19). Targeted ES cells were screened for homologous recombination by PCR. Genomic DNA (gDNA) extraction was performed using an in house protocol and consisted of gDNA precipitation with absolute ethanol. Twenty nanograms gDNA were used for each PCR (primer sequences in Supplementary Materials and Methods) using LongAmp ® Taq DNA Polymerase (New England Biolabs, Ipswich, MA) according to the manufacturer's recommendations. Recombined ES cells were injected into C57BL/6J blastocysts which were transferred into foster mothers to obtain chimeric mice (Myd88 L252P-flSTOP mice).

Mice
Cd19 Cre mice (20) and mice carrying the Myd88 L252P -IRES-Yfp allele were crossed to induce the expression of the transgene in B cells (Myd88 L252P mice). Offspring were routinely screened by PCR using specific primers for insertion of the transgene (Supplementary Materials and Methods). Animals were housed at 21-23°C with a 12-hour light/dark cycle. All procedures were conducted under an approved protocol according to European guidelines for animal experimentation (French national authorization number: 8708503 and French ethics committee registration number APAFIS#14581-2018041009469362 v3).

Sera Analyses
Serum was obtained from blood collected retro-orbitally. Specific ELISA and serum electrophoresis assay were performed as previously described (19,22).

Flow Cytometry and In Vivo Proliferation Assays
In order to collect BM cells of Cd19 Cre and Myd88 L252P mice, femurs from both hind legs were rinsed with PBS and sternum was gently crushed and cells filtered on a nylon meshwork that was rinsed with PBS. Spleen cells from Cd19 Cre and Myd88 L252P were filtered through a sterile nylon membrane. Blood samples were collected retro-orbitally. Red cells were lysed by RBC Lysis Buffer (Biolegend, San Diego, CA). Cell suspensions were resuspended at 4°C in a labeling buffer (PBS, 1% BSA, 2mM EDTA) and labeled with fluorescent conjugated monoclonal antibodies listed in Supplementary Materials and Methods. Labeled cells were analyzed using a BD Fortessa SORP flow cytometer (BD Bioscience France, Le Pont de Claix, France). Results were analyzed using Kaluza Flow Cytometry software 1.2 (Beckman Coulter, Brea, CA).

Gene Expression Profiling
A series of seven mice (three Cd19 Cre and four Myd88 L252P ) was studied in parallel with bone marrow purified tumor B-cells from a series of 11 patients with MYD88 L265P WM (series 1) as well as lymph nodes from a series of 58 patients: 19 MYD88 WT chronic lymphocytic leukemia, 15 MYD88 L265P WM, 12 MYD88 wt Nodal marginal zone lymphoma, 5 MYD88 wt WM, 4 follicular lymphoma and 3 patients with benign follicular hyperplasia (series 2, Supplementary Tables 1 and 2). Approval of this protocol was obtained from the local IRB of the CHRU of Lille (CSTMT043). MYD88 and CXCR4 mutational status was determined as previously described (6). Total mRNA was extracted from whole infiltrated tissues and purified B-cells as reported (24). For humans and mice, RNA amplification and hybridization onto microarrays were performed on an Affymetrix Human Genome U133 Plus 2.0 Array and on an Affymetrix Gene Atlas system ® with the MoGene-2_1-st-v1 Affymetrix chip (Affymetrix, Santa Clara, CA) respectively according to a previously described protocol (25) (GEO accession number GSE138273). Bioinformatic analyses are detailed in Supplementary Materials and Methods.

Repertoire Analysis
RNA was extracted from total spleen, and 1µg was used for sequencing. Transcripts were amplified by 5'RACE PCR using reverse primers hybridizing within either the membrane or secreted exon of the µ or g genes. ProtoScript ® II Reverse Transcriptase (New England Biolabs, Ipswich, MA) was used for reverse transcription and amplicons were obtained using Phusion ® High Fidelity DNA Polymerase (New England Biolabs, Ipswich, MA) according to the manufacturer's instructions. Primers used are listed in Supplementary Materials and Methods. Illumina sequencing adapter sequences were added by primer extension, and resulting amplicons were sequenced on an Illumina MiSeq sequencing system using MiSeq kit Reagent V2 500 cycles. Paired reads were merged using FLASH (26). Repertoire analysis was done using the IMGT/HighV-QUEST online tool (http://www.imgt.org/IMGT_vquest/vquest). The resulting output was parsed by in-house R script.  Figure 1A). To validate this transgene, we checked that it induced expression of both MYD88 L252P and YFP proteins in the murine A20 B-cell line (Supplementary Figures 1B, C) and that it was responsible for constitutive NF-kB activation (Supplementary Figure 1D). The Myd88 L252P -IRES-Yfp insert was cloned into the pROSA26.1 vector (17) (Supplementary Figure 1E). In this construct, the insert was subcloned downstream from a Neomycin-STOP cassette flanked by LoxP sites. Chimeric mice were intercrossed to obtain stable germinal transmission of the Myd88 L252P -IRES-Yfp transgene (Myd88 L252P-flSTOP mice). Myd88 L252P-flSTOP and Cd19 Cre mice were crossed. Mice with both transgenes (Myd88 L252P mice) were then studied, with their age matched Cd19 Cre littermates as controls (Cd19 Cre LMC). Specific B-cell expression of the transgene was found in more than 90% of blood and spleen B cells compared to virtually no expression in the T-cell compartment (Supplementary Figure 1F). As expected and as evidence of NF-kB activation, Myd88 L252P splenocytes over-expressed the NF-kB target gene Tnfaip3 at the mRNA level (Supplementary Figure 1G).

Serum Protein Electrophoresis Profiles Segregate Myd88 L252P Mice According to Age
As a first exploratory step, serum protein electrophoresis (SPE) was systematically performed on a series of 40 Myd88 L252P mice and 26 age matched Cd19 Cre LMCs. As shown in Figure 1A three SPE profiles were seen: normal, polyclonal hyper gammaglobulinemia (hyper Ig) and a monoclonal Ig peak. All Cd19 Cre LMCs exhibited normal SPE regardless of their age. In other words, SPEs with hyper Ig or Ig peaks were found only in Myd88 L252P mice ( Figures 1A, B). Figure 1B shows the relationship between the age of Myd88 L252P mice and the SPE profile. Young Myd88 L252P mice (16-23 weeks) had a normal or hyper Ig SPE profile. In contrast, most mice older than 32 weeks had an Ig peak. In between these two groups, 24 to 31 week old mice (middle age) had a hyper Ig or an Ig peak in 65% and 35% cases respectively (Fisher test, p=2.10 -4 ). ELISA quantification of serum Ig revealed that young Myd88 L252P mice with a normal SPE exhibited a moderate IgM and IgG hyper Ig when compared to their Cd19 Cre LMC ( Figure 1C and Supplementary Figure 2). Serum IgG levels of middle aged and old Myd88 L252P mice were variable when compared to their young counterparts. This was in contrast to serum IgM levels that were significantly increased in middle aged mice and even more so in old mice and correlated with the SPE profile and age ( Figure 1C and Supplementary Figure 2).
These results first indicate that continuous MYD88 activation in B cells was associated with a global increase in Ig secretion. Second, age related occurrence of polyclonal hyper Ig and then monoclonal Ig peaks correlated with the increase in serum IgM levels. This suggests that, after a polyclonal expansion phase, aging of Myd88 L252P mice was associated with clonal restriction of IgM-secreting B-cells, very likely reflecting the MYD88 L252P Bcell transformation power in these cells. Therefore, these first results point to a strong relationship between MYD88 L252P and IgM-secreting B-cells.

Myd88 L252P Bone Marrow IgM Plasma Cell Content Was Increased and Displayed a CD138 Expression Continuum
As shown in Figure 2A, Myd88 L252P bone marrow global B-cell content was comparable to that of Cd19 Cre LMCs in terms of percentages in 16 week old mice with normal SPE. Transgene expression was mainly found in Myd88 L252P CD19 high B-cells. Indeed, with an on/off effect, percentages of YFP positive cells (i.e of LoxP rearranged cells) was directly correlated with CD19 mean fluorescence intensity ( Figure 2B). Comparison of CD19 cre LMC and Myd88 L252P mice did not reveal any significant bone marrow B-cell increase with age. However older Myd88 L252P mice had increased levels of IgM pos CD19 high B-cells ( Figures 3A, B, left panel and Supplementary Figure 3 for the gating strategy). Strikingly, a CD138 expression continuum was clearly evident in a triple parametric B220/CD138/IgM histogram gated on mature B-cells and/or PCs in Myd88 L252P mice only ( Figure 3A, lower panel see in red the decrease in B220 and the increase in CD138 expression). This CD138 expression continuum, that we recently showed to be characteristics of MYD88 L265P WM bone marrow tumor B-cells (Gayet et al, Cytometry B 2021), started from IgM high CD138 low and ended at IgM low CD138 high cells ( Figure  3A, lower panel). This CD138 expression continuum was absent in Cd19 Cre LMCs. Consequently, Myd88 L252P mice showed increased percentages of both IgM high CD138 low B-cells (most likely precursors of IgM PCs) and total bone marrow PCs ( Figure  3B, right panel). Moreover, the proportion of bone marrow IgM PCs was significantly increased in young Myd88 L252P mice and even more in older Myd88 L252P mice ( Figure 3C). Indeed, Myd88 L252P CD19 pos /YFP pos B cells tended to accumulate in the IgM high B-cell compartment when compared to its Myd88 L252P CD19 pos /YFP neg counterpart (Supplementary Figure 4).
Altogether, these results indicate that transgene expression started in a minority of CD19 weak B-cell precursors and was mainly expressed at the latest stages of B-cell lymphopoiesis when CD19 expression was high. Evidence for bone marrow increase in both IgM high CD138 low and PCs with the characteristic CD138 expression continuum may either suggest

MYD88 L252P First Induced Peripheral Early Lymphoplasmacytic and Plasma Cells Expansion and Then B-Cell Transformation Into a Lymphoplasmacytic-Like Lymphoma
Extended white blood cell differential was not altered and mice did not exhibit any palpable/visible peripheral lymphadenopathy (data not shown) regardless of the SPE profile. Young Myd88 L252P mice with normal SPE tended to have spleen enlargement when compared to their age related Cd19 Cre LMCs ( Figure 4). Spleen enlargement was dramatically increased in Myd88 L252P mice with hyper Ig and even more so in those with an IgM peak, a feature that was most likely to related to B-cell transformation. Indeed, the B/T cell ratio was markedly increased in these latter mice (Supplementary Figure 5).
While spleen histology of young Myd88 L252P mice was comparable to that of their Cd19 Cre LMCs, with an apparent normal spleen architecture, Myd88 L252P mice with hyper Ig or an Ig peak exhibited enlarged and congruent lymphoid nodules ( Figures  5A-E). At high magnification, a marked lymphoplasmacytic aspect consisting of a mixture of small to large lymphocytes with numerous lymphoplasmacytic cells (LP cells) and PCs was noted in all Myd88 L252P mice whatever their SPE profile ( Figures 5F-J). This spleen aspect was particularly striking for mice with Ig peaks, and was characterized by massive and diffuse infiltration of lymphoplasmacytic cells that evoked human B-cell lymphomas with features of PC differentiation, further called "LP-like lymphoma" or "LP-like tumors". (see also the cytological imprint in Supplementary Figure 6). Presence of LP cells and PCs in Myd88 L252P spleen was invariably confirmed by immunohistochemistry after intracellular Ig labeling regardless of SPE status, with numerous LP cells and terminally differentiated PCs (cells with intermediate or strong intracytoplasmic Ig labeling respectively). Noteworthy, cell densities were markedly increased in Myd88 L252P mice with hyper Ig or with an Ig peak ( Figures  5K-O).
Based on B-cell expression of CD21 and CD23, frequencies of CD21 pos CD23 high follicular B-cells were not significantly altered in Myd88 L252P mice regardless of their SPE status ( Figure 6A and Supplementary Figure 7). In contrast, a decrease of CD21 high CD23 pos marginal zone B-cells was observed. This cell content nearly disappeared in Myd88 L252P mice with an Ig peak. Only total PCs were increased in these mice ( Figure 6B). However, among total spleen PCs, percentages of IgM PCs were increased in a similar manner in all Myd88 L252P mice no matter what their SPE status was ( Figure 6C    . Moreover IgM PCs were predominantly CD93 neg suggesting that they belonged to the proliferating PC compartment (27) ( Figure 6C and Supplementary Figure 7). Therefore, morphological and immunophenotypic results indicated that continuous MYD88 activation was associated with continuous peripheral IgM PC differentiation very early on and that these LP and PC subsets continuously expanded with age first being associated with hyper Ig and then with an Ig monoclonal peak and a LP-like lymphoma aspect in the spleen.

Proliferation Index of Myd88 L252P LP-Like Tumors Was Moderately Increased in Myd88 L252P Tumors With an LP-Aspect
To better study these Myd88 L252P LP-like tumors, we compared their Ki67 proliferation index to that their of Cd19 Cre LMCs as well as to L.CD40, L.CD40/Lc-MYC mice as controls. L.CD40 mice are a model of marginal zone spleen B cell indolent lymphomas without plasma cell differentiation but with NF-kB activation (28). L.CD40/Lc-MYC mice are a model of ABC-DLBCLs with both c-Myc and NF-kB activation in B-cells (24). Very few Ki67 positive cells were seen outside germinal centers in spleen sections from Cd19 Cre mice ( Figure 7A panel A). The Ki67 index was weak in L.CD40 mice ( Figure 7A  Since Myd88 L252P mice had a global hyper Ig even if predominantly IgM, it was important to assess µ or g chain clonality of tumor surface and secreted immunoglobulins at the molecular level. Six LP-like cases with monoclonal Ig peaks and five Cd19 Cre mice were studied. mRNA reverse transcription followed by RACE PCR with primers specific for either the membrane or secreted forms of mouse µ or g heavy chains was performed, followed by high throughput sequencing (HTS) of the VDJ-Cµ or VDJ-Cg region ( Figure 8A). Figures 8B, C show the relative frequency of the five most abundant mRNA clones for the µ or g heavy chains respectively. RACE PCR and HTS results indicate that Myd88 L252P mice developed clonal or oligoclonal B-cell expansion with expression of both secreted and membrane forms of the µ heavy chain that had the same VDJ-Cµ clonal rearrangement ( Figure 8B), without any bias in terms of V segment usage (not shown). The same RACE PCR technique with primers specific for either the membrane or secreted form of the mouse g heavy chain did not identify any significant B-cell clonal expansion in Myd88 L252P tumors ( Figure 8C).
These results indicate that, despite initial IgM and IgG hyper Ig in young Myd88 L252P mice, MYD88 L252P specifically promoted IgM B-cell lymphomagenesis with clonal expression of both membrane and secreted µ chain isoforms identical VDJ gene rearrangements. These results are in full agreement with the   Table 3). To analyze this set of genes, we combined both K-mean and hierarchical clustering and principal component analysis, as already published (29). Following this methodology, deregulated genes in Myd88 L252P spleen tumors could be segmented into 14 classes of co-regulated genes (Figure 9 with methodological details in Supplementary Materials and Methods, Supplementary Figure 9 and Supplementary Table  4). Consistent with increased B/T ratios in Myd88 L252P LP-like tumors, expression of genes belonging to the T-cell lineage, as well as T-cell signaling and activation signatures was downregulated in Myd88 L252P spleen tumors (Figure 9, see clusters I, K and L). Of note, RelB signature was associated with that of Tcells and was decreased in Myd88 L252P spleen tumors. In contrast, expression of genes belonging to the proliferation, RelA NF-kB activation pathway and plasma cell differentiation signatures were up-regulated ( Figure 9, cluster J mainly as well as cluster B, G and M for proliferation).
To identify MYD88 L252P deregulated genes in common with those of WM patients having the MYD88 L265P mutation, transcriptomes of Myd88 L252P tumors were compared to those of purified WM bone marrow B-cells from a series of 11  Tables 2, 7). All MYD88 L265P WM patients clustered together after unsupervised hierarchical clustering ( Figure 10A, right panel and Supplementary Tables 7, 8). Also belonging to the MYD88 L265P WM cluster were 3/5 (60%) MYD88 wt WM and 2/ 12 (17%) MYD88 wt NMZLs. We used the linear predicting score described by Wright et al. (31) to estimate the informativeness of the Myd88 L252P /WM signature for WM diagnosis. As shown in Figure 10B, as set of 174 genes (84 up and 90 down, Figure 10B) was found to predict WM with over 90% probability (Supplementary Tables 9, 10).
(not shown). In accordance with results published by Hunter et al. (30), expression of genes such as CXCR4, DUSP22, PIM1 and 2 or TRAM1 was increased while expression of SNED1 was decreased. Few genes of the Myd88 L252P /WM signature overlapped with those of ABC/GC DLBCLs published by Wright et al. (31). These overlapping genes, corresponding only to those that are overexpressed in ABC DLBCLs, were IRF4, IGHM, CXCR4, P2RX5, PIM1 and PIM2. In other words, the Myd88 L252P /WM signature did not significantly overlap with that of GCB DLBCLs. Among other deregulated genes were cyclin kinase inhibitors CDKN1B (p27 kip1 ) and CDKN2C (p18/ INK4AC), mutations of the former being found in hairy cell leukemias (32). RASSF3 and KRAS were also up-regulated. RASSF3 belongs to the Ras association domain family (RASSF).
Altogether, the Myd88 L252P tumor signature highlights proliferation as well as canonical NF-kB p65/RelA activation (but not RelB), which is in agreement with the known fact that MYD88 activates the classical NF-kB pathway. The Myd88 L252P tumor signature also strikingly confirms that lymphoplasmacytic differentiation is at the heart of MYD88 related B-cell transformation in mice, a feature shared with WM tumors with the MYD88 L265P mutation. First, we amplified cDNA between the primer specific for the membrane or the secreted form (black arrows) and the 5'RACE oligonucleotide. Amplicons for Illumina sequencing were obtained after two nested PCRs; the first with the 5' Race CAP primer and either membrane (blue arrow) or secretion (green arrow) exon specific primer, and the second with the same 5' primer and a CH1 exon specific primer (grey arrow). For sequencing, forward (grey) and reverse (red) primers used for the second PCR contained adaptors (blue and purple) and a barcode (orange); each barcode sequence was specific for one sample only.

DISCUSSION
Different mouse preclinical models with continuous MYD88 activation in B-cells have been published; All but one demonstrate the B-cell transformation potential of MYD88 but without presenting evidence for a correlation between IgM B-cell LP and PC differentiation (13)(14)(15).. The first report was published by Knittel et al. (14). The authors generated a mouse model that allows B-cell conditional expression of the Myd88 L252P allele from the endogenous Myd88 locus. In this model, expression of Myd88 L252P would be regulated in a manner similar to that of the wild type allele. At least three regulatory levels of MYD88 activity have been reported. The MYD88 regulatory region harbors various transcription binding sites such as those for NF-kB, IRF1, SP1 or STAT factors and it was shown that this gene is regulated by IL-6 (33), which suggests a role for its expression in either plasma cell differentiation or inflammation. An alternative splice variant of MYD88, MYD88s, lacks exon 2 and is unable to activate NF-kB. This variant is also able to form a heterodimer with the full length MYD88 protein, resulting in decreased formation of the myddosome complex (34). MYD88s is increased during sepsis and is thought to ensure robust termination of MYD88 dependent inflammation (35,36). As a third regulation mechanism, hypomethylation and upregulation of MYD88 are important for NF-kB activation (37). MYD88 promoter demethylation is important in glioblastoma and is associated with increased MYD88 protein expression in lung cancers (38,39). Mice from Knittel's model occasionally developed DLBCLs when they aged. Therefore, the Knittel  same activation properties on a mouse B-cell background as in human B-lymphocytes. In the model of Sewastianik et al, the human loxP-flanked-stop-MYD88 L265P transgene was inserted downstream from the mouse Collagen type I alpha 1 chain (Col1A1) gene and MYD88 L265P expression was induced by crossing with Aid Cre mice (15). In addition to focal skin rashes, some WM features were noted such as expansion of lymphoplasmacytic cells and increased IgM serum levels. However, only DLBCL clonal transformation was seen. Because AID is mainly expressed in germinal center B-cells and because the promoter of the Col1A1 gene is highly active in fibroblasts and osteoblasts (41), this model raises the question whether temporality and/or expression pattern could be important in MYD88 driven B-cell lymphomagenesis.
The Rosa26 locus has been solidly proven to be valuable for expression of numerous oncogenes in the B-cell lineage (see reference (42) for discussion as well as the literature of the K Rajewsky's group). By inserting our Myd88 L252P -IRES-Yfp transgene in this locus, we forced continuous expression of the mutated MYD88 protein in a heterozygous-like context while respecting the native MYD88 activation pathways of mouse Bcells. Moreover, we were able to monitor our transgene expression by flow cytometry due to YFP. Thereby, we created a conditional Myd88 L252P mouse model closed to the one published by Jo et al. (13). However, these authors mainly focused their work on the synergy between MYD88 and the catalytic subunit HOIP which increases LUBAC ligase activity that in turns promotes NF-kB canonical activation; only four socalled CD19-cre-MYD88LP have been studied at the tumor stage.
Here, a longitudinal analysis of a series of 40 Myd88 L252P mice compared to their age matched Cd19 Cre LMCs demonstrated that IgM plasma cell expansion is at the heart of MYD88 dependent B-cell transformation. Indeed, by examining clonal restriction of IgM secreting B-cells, we first showed that ageing of Myd88 L252P mice was associated with polyclonal hyper Ig followed by monoclonal Ig peak due to increased serum IgM. Second, we provide evidence indicating that bone marrow relocalization of IgM B-cells, IgM high CD138 low cells and IgM PCs was increased in Myd88 L252P mice with a CD138 expression continuum, that is a characteristic of WM tumor B-cells. Third, analysis of spleen morphology and spleen B-cell subsets by flow cytometry indicated that continuous MYD88 activation was associated very early with peripheral LP cell and CD93 neg IgM PC expansion and that these cell subsets were markedly increased at the time of the Ig monoclonal peak. Fourth, appearance of a monoclonal Ig peak was constantly associated with a B-cell lymphoma with marked features of lymphoplasmacytic differentiation, so-called Myd88 L252P LP-like lymphoma. Fifth, at the molecular level, Myd88 L252P specifically promoted IgM B-cell lymphomagenesis with mRNA clonal expression of both membrane and secreted µ (but not g) chain isoforms. Finally, the Myd88 L252P tumor gene expression profile not only highlights the canonical NF-kB p65/RelA activation pathway and proliferation, but also strikingly shares an Xbp1 centered lymphoplasmacytic B-cell differentiation signature with MYD88 L265P WM. This signature differentiates MYD88 L265P WM from other indolent B-cell tumors including marginal zone lymphomas. Our results contradict the conclusions of Sewastianik et al. (15), and, being in line with those of K Schmidt et al. (16), firmly demonstrate the specific transforming effect of MYD88 activation in IgM PC differentiating B-cells.
To establish the gene expression signature from bulk Myd88 L252P spleen tumors rather than from purified B-cells is matter of discussion. The very obvious disadvantage of working on bulk tumors is certainly that all mRNA species from all cell types present in the tissue are mixed together. Even if massively invaded, both stromal and other residual immune cells (T-cells, macrophages, dendritic cells…) persist constantly in the tumor. In these conditions, specifically assigning a given mRNA expression pattern to tumor cells is always hazardous. However, because tumors were immediately snap-frozen, all mRNA species are supposed to be well preserved without any significant experimental bias. On the other hand, working on purified cells may also induce artefacts since the abundance of different mRNA species may vary during the time of purification which can also stress the cells. Above all, in the specific case of Myd88 L252P spleen tumors, which exhibit an LP aspect with continuous PC differentiation, the key question would have been to choose the right negative selection marker. Indeed, tumor cells are phenotypically heterogeneous with variable expression of B220, surface Ig or CD138 for example. Should we have selected B220 high versus B220 low B-cells or CD138 low versus CD138 high cells? Rather than make wrong or partial choices concerning which tumor cells to purify, we chose to work on bulk Myd88 L252P spleen tumors with well-preserved total mRNA and to compare this bulk signature to that of WM, including purified WM bone marrow tumor B-cells.
One characteristic feature of Myd88 L252P B-cells was the strong reduction of the marginal zone B-cell compartment. While also characterized by continuous activation of NF-kB, the L.CD40 mouse model published by Hömig-Hözel et al. (28), in which B-cells are submitted to continuous CD40 activation signaling, showed expansion of marginal zone B-lymphocytes. This points out the differences between TLRs and CD40 in terms of NF-kB activation. Indeed, CD40 is able to activate both the classical and alternative pathways, i.e. to induce the nuclear translocation of RelA and RelB NF-kB containing complexes while TLRs only activate the classical pathway. The effect of both pathways as well as the strength of NF-kB activation on B-cell fate has been extensively discussed by Pillai et al. (43). In this review, the authors also indicate that BTK activation blocks the Notch signaling pathway that is essential for marginal zone B-cell differentiation. It turns out that Hunter et al. have shown that Myd88 is able to activate BTK in a BCR independent manner (30), which consequently could repress B-cell maturation toward the marginal zone B-cell lineage.
Gene expression profiles of LP-like Myd88 L252P tumors distinctly suggest the involvement of RelA rather than RelB. RelA, but not RelB, is clearly associated with EBV-dependent Bcell immortalization and with EBV-associated DLBCL tumors, which exhibit a phenotype close to that of ABC-DLBCLs (25).
RelA is also essential for development of GC-derived PCs (44) and immunohistochemistry detected nuclear RelA in WM Bcells (45). Indeed, LP-like Myd88 L252P B-cell tumors shared strong overlaps with human WM in terms of gene expression profile. Even if a few genes were in common with the ABC-DBCL signature such as IGHM, CXCR4 or PIM1 and PIM2, the Myd88 L252P/ WM signature points to dysregulation of plasma cell differentiation as the keystone of MYD88 transforming physiopathology. It also suggests that KRAS activation could be important. RASSF3 and KRAS itself were up-regulated in both Myd88 L252P B-cell tumors and human WM. Consistently, and in agreement with results from the group of Treon (30), we also found RASSF6 overexpression in WM patients (NG and JF unpublished results). CXCR4, whose expression is increased in WM, may activate the RAS pathway through RasGAP-associated proteins (46). The most frequent mutations involve KRAS and NRAS genes in multiple myeloma (47). Even if such mutations have not been reported in WM (6), our results highlight the putative role of the Ras activating pathway in WM, which may lead to the design of novel therapies.
Despite similarities between WM and Myd88 L252P LP-like tumors such as serum Ig monoclonal peak, increase in IgM prePCs, PC bone marrow relocalization and marked lymphoplasmacytic differentiation of tumor cells some major differences exist. The first difference, is the predominant site of tumor involvement. Even if splenomegaly is found in 20% of bona fide WM patients, it is largely admitted that bone marrow is the primary tumor site. In contrast, Myd88 L252P LP-like mouse tumors mainly developed in the spleen. Physiologically, IgM PCs tend to reside mainly in the spleen whereas switched IgG PCs migrate to the bone marrow (48). This raises the question of why WM IgM tumor B-cells migrate to bone marrow. In that view, it is largely suspected that CXCR4 mutation could play a role in bone marrow homing. Further studies could also evaluate the transforming potential of Myd88 by adoptive transfer of LP-like tumor B-cells. Whether the transferred tumor will retain the lymphoplasmacytic aspect would also be interesting.
Another significant difference is the presence of large cells and increased proliferation in Myd88 L252P LP-like mouse tumors. The increased proliferation index was an early event since it was also found in young Myd88 L252P mice with normal SPE. As we previously discussed (24), only three mouse models for indolent lymphomas of the spleen have been published, one mimicking TRAF3 inactivation, the second with constitutive expression of BCL10 and the last one with continuous CD40 signaling (the L.CD40 model that we used as a control in Figure 7) (28,49,50). These three models are characterized by increased RelB activation. We previously demonstrated that immune surveillance may influence morphology and proliferation in the L.CD40 model. In this model, immunosuppression led to transformation of small indolent B-cell L.CD40 tumors into large B-cells with increased proliferation. Reactivating the anti-tumor response using anti-PD-L1 immunotherapy led to tumor regression (51,52). These results on this mouse model suggest that the immunologically silent "indolent phenotype" of a B-cell tumor could be related to the immune pressure exerted on tumor B-cells. Whether and how activation of the alternative and canonical NF-kB pathways differently disturb immune surveillance remains to be determined, and comparison of both L.CD40 and Myd88 L252P mouse models could provide answers. However, as in the L.CD40 tumor model and in spleen marginal zone lymphomas (52,53), the PD1/PD-L1 axis is most likely playing a role in the immune escape of aggressive tumor B-cells with MYD88 activation. Using Eµ-MYC transgenic hematopoietic stem cells (HSC) stably transduced with naturally occurring NF-kB mutants to generate various primary mouse lymphomas, Reimann et al. recently showed that MYD88 tumors express high levels of PD-L1 and that anti-PD-1 therapies induce T-cell dependent senescence of tumor cells (54). PD-L1 surface expression is weak or absent on WM tumor B-cells. However soluble PD-L1 serum levels are increased in WM patients and PD-L1 is upregulated by IL6 (55).
In summary, our longitudinal study of Myd88 L252P mice demonstrated that continuous MYD88 activation is able to promote early expansion of IgM LP cells and PCs with, first, serum polyclonal hyper Ig and then a monoclonal Ig peak. Ig peaks were constantly associated with B-cell lymphomas sharing characteristics with WM. Two major differences with WM were the spleen localization of Myd88 L252P tumors and increased proliferation. Here, we showed for the first time that IgM lymphoplasmacytic B-cell differentiation is at the heart of Myd88 L252P transforming potential. Thus, we also provide an interesting preclinical model for development of new therapeutic approaches or to study immune surveillance for example not only in WM but also in others B-cell lymphomas with features of plasma cell differentiation. Indeed, a better understanding of the underlying molecular mechanisms is necessary in order to develop new therapies for these incurable B-cell cancers.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://www.ncbi. nlm.nih.gov/geo/, GSE138273.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by the local IRB of the CHRU of Lille (CSTMT043). The patients/participants provided their written informed consent to participate in this study. The animal study was reviewed and approved by French national authorization number: 8708503 and French ethics committee registration number APAFIS#14581-2018041009469362 v3.

AUTHOR CONTRIBUTIONS
COu and LR contributed equally to this work. COu and LR performed and analyzed experiments. AS helped to perform the repertoire analysis. COb performed and analyzed ELISA. MD and NG performed the transcriptomic experiments. JF, SP, AS and LR performed the bioinformatics analyses. QL and CC performed flow cytometry analysis of bone marrow B-cell subsets. NF participated in the design of the project. KB and ME participated in the development of this study. CV-F created the mouse model, contributed to the experiments and analyzed the results. JF and CV-F directed the study and wrote the manuscript. All authors contributed to the article and approved the submitted version.

FUNDING
The group of JF is supported by grants from the Ligue Nationale Contre le Cancer (Equipe labelliseé Ligue), the ComiteÓ rientation Recherche Cancer (CORC), the France Lymphome Espoir association, the Nouvelle Aquitaine Region and the Haute-Vienne and Corrèze committees of the Ligue Nationale Contre le Cancer. CV-F was supported by the France Lymphome Espoir association of patients. SP is supported by the Septentrion committee of Ligue contre le Cancer. ME is supported by an ANR @RAction grant (ANR-14-ACHN-0008), an ANR JCJC grant (ANR-19-CE15-0019-01), an IDEX Universitéde Paris grant, a Fondation Arthritis grant and a Fondation ARC grant (P JA20181208173). KB is supported by an ANR PRC grant (ANR-17-CE14-0019), an INCa grant (PRT-K 2017) and the Association Saint Louis pour la Recherche sur les Leuceḿies.