Upregulation of IL4‐induced gene 1 enzyme by B2 cells during melanoma progression impairs their antitumor properties

B cells present in human cutaneous melanoma have been associated with protective or detrimental effects on disease progression according to their phenotype. By using the RET model of spontaneous melanoma and adoptive transfer of B16 melanoma cells, we show that immature and follicular B2 (B2‐FO) cells exert a protective effect on melanoma progression by promoting the generation of effector memory T cells and limiting the recruitment of polymorphonuclear myeloid‐derived suppressor cells. Unfortunately, this beneficial effect progressively wanes as a consequence of enhanced expression of the IL4‐induced gene 1 (IL4I1) enzyme by immature B cells and B2‐FO cells. Endogenous IL4I1 selectively decreases CXCR5 expression in splenic immature B cells, subverting their trafficking to primary tumors and enhancing the production of IL‐10 by B2 cells, thereby promoting an immunosuppressive microenvironment. Accordingly, B2 cells from RET IL4I1KO mice more efficiently controlled B16 melanoma growth than B2 cells from IL4I1‐competent RET mice. Collectively, immature B cells and B2‐FO cells are key actors in the control of melanoma growth, but their mobility and functions are differently impaired by IL4I1 overexpression during melanoma progression. Thus, our present data strongly urge us to associate an IL4I1 antagonist with current immunotherapy to improve the treatment of metastatic melanoma.


Introduction
The increasing incidence of cutaneous melanoma represents a major public health problem [1].In-depth characterization of tumor-infiltrating T-cell subsets has allowed the development of immunotherapy based on immune checkpoint inhibitors (i.e.antagonists of CTLA-4, PD1, and PD-L1), currently given as firstline treatment for metastatic melanoma.B cells recently emerged as key markers of therapeutic efficacy [2].Indeed, responder patients showed a higher density of infiltrating CD20 + B cells, switched memory B cells, and tertiary lymphoid structures [3][4][5] as well as high levels of total and melanoma-specific IgG [6,7].By contrast, other findings failed to reveal an association between melanoma-infiltrating B-cell density and response to anti-PD-1 treatment [8].Finally, enrichment of immunosuppressive IL10 + regulatory B cells has been reported in nonresponder patients to anti-CTLA-4 therapy [9].Thus, tumor-specific antibodies and particular B-cell subsets might exert pro-or antitumor functions in human melanoma.
Preferential accumulation of B cells in primary tumors and in nodular over visceral metastases has been reported in patients with cutaneous melanoma.A high density of B cells was associated with a better 5-year survival, arguing for their antitumor role during the early stages of the disease [10,11].Accumulation of plasma cells also correlated with longer overall survival in metastatic melanoma [12,13].In contrast, CD5 + B1 cells and IL10-expressing B cells present in primary tumors and cutaneous metastases are thought to exert protumor functions [14,15].Thus, innate-like B1 and conventional B2-cell subsets might differently contribute to immune surveillance in melanoma during tumor progression [16].
We and others have previously demonstrated that the secreted L-amino acid oxidase IL4-induced gene 1 (IL4I1) enzyme is produced by immune cells, in particular, myeloid antigen-presenting cells, B cells, and induced regulatory CD4 + T cells [17][18][19][20].It exerts immunosuppressive functions and contributes to tumor escape from the immune response [17,[21][22][23][24].In agreement with our observations, recent single-cell transcriptomic analyses identified expression of the IL4I1 gene in immunosuppressive macrophage and dendritic cell subsets of the human tumor microenvironment [25][26][27].In a recent study of primary human cutaneous melanomas, we detected a high density of IL4I1 + cells in patients with an ulcerated tumor or with a positive sentinel lymph node [28].Interestingly, IL4I1-rich areas were poor in cytotoxic CD8 + T cells and the proportions of IL4I1 + cells correlated positively with those of regulatory CD4 + T cells, most often associated with a shorter overall survival.Accordingly, IL4I1 activity was associated with tumor aggressiveness in the spontaneous model of RET melanoma (mice transgenic for a constitutively active form of the human RET oncogene) [29].In this model, modifications of the tumor microenvironment induced by IL4I1 included reduced B-cell infiltration.Interestingly, our recent data highlighted the role of IL4I1 in B-cell biology [30] leading us to further investigate it in the tumor context.
The present work aims at deciphering the role of B-cell subsets in the course of melanoma development and evaluating the impact of IL4I1 on their functions and trafficking.We used RET mice, either competent or deficient (RET IL4I1 KO mice) for IL4I1 and set up a new model in which IL4I1 positive cells are fluorescent (RET IL4I1 YFP mice).We demonstrate that immature B cells and mature follicular B2 (B2-FO) cells infiltrating the primary tumor exert anti-tumor properties, a function progressively subverted by the upregulation of IL4I1 expression in RET mice.First, IL4I1 expression regulates B-cell recruitment through CXCR5dependent (immature B cells) and CXCR5-independent (B2-FO cells) mechanisms.Second, IL4I1 affects the costimulatory functions of B2 cells and thus immune responses to melanoma.

B2 cells control RET melanoma aggressiveness
To investigate the role of B cells in tumor control, RET mice were treated from day 4 postbirth with a depleting anti-mouse CD20 antibody or its isotype control (Fig. 1A).The incidence of primary tumor was similar in both groups (Fig. 1B).In contrast, cutaneous metastases were detected about 20 days earlier (37 days vs. 56 days) in 50% of B-cell depleted mice (Fig. 1C) and were more numerous at autopsy in anti-CD20-treated mice compared with control mice (Fig. 1D).Thus, B cells delayed metastatic dissemination.
At day 60, B-cell depletion reached 90% and 60% in primary tumors and cutaneous metastases, respectively (Fig. 1E), and 99% in the spleen and cervical lymph nodes which drain the primary tumor (TdLN; Fig. 1F).B-cell subsets (see gating strategy in Supporting information Fig. S1A) were differently affected by this treatment.Indeed, mature follicular B2 cells (CD19 + CD21 + CD23 + , referred to as B2-FO cells thereafter) which constituted the main tumor-infiltrating B-cell subset, were particularly affected by the anti-CD20 at all sites (Fig. 1G-J and Supporting information Fig. S2A and B).Due to their low CD20 expression, plasma blasts (PB, CD138 + CD19 + ) and plasma cells (PC, CD138 + CD19 low ) were less affected by the treatment, in particular in the tumor microenvironment (Fig. 1H and Supporting information Fig. S2A).Compared with B2-FO cells, immature B cells (CD19 + CD93 + CD43 − ) and B1 cells (CD19 + CD93 − CD43 + ) from the primary tumor and metastatic sites were less sensitive to the treatment (Fig. 1G and H).Consequently, these cell types became largely predominant over B2-FO cells in primary tumors posttreatment (Fig. 1G, pies).In cutaneous metastases, the depletion of B2-FO cells was as efficient as in primary tumors, immature B cells were slightly more depleted than in primary tumors (79% vs. 62%) but B1-cell depletion was negligible (Fig. 1H).
In secondary lymphoid organs, B-cell depletion induced significant modifications of the T-cell and myeloid-cell populations.A decrease in the proportion of effector memory (CD44 high ) cells affecting both CD4 + and CD8 + T cells, including the TH1 (CXCR3 + T-bet + ) cell subset was observed (Fig. 1K and gating strategy in Supporting information Fig. S1B).In contrast, PMN-MDSC increased significantly in TdLN and spleen (Fig. 1L left) and tended to increase at the primary tumor site (Fig. 1L right).
Thus, our results indicate that B2-FO cells from the tumor microenvironment protect RET mice from tumor progression.These cells may be crucial for optimal TH1 and CD8 + T-cell responses and for restricting PMN-MDSC expansion.

Partial loss of protective B2 cells in primary tumors during RET melanoma progression
Having established the protective role of B2 cells during melanoma progression, we further characterized B-cell subsets present within the primary tumor, cutaneous metastases, spleen, and TdLN of two groups of RET mice representative of early and advanced stages of the disease (Fig. 2A).In the first group referred to as early disease (ED), mice had less than four cutaneous metastases (mainly located in the face and neck) and less than 65 million of total cells (dotted line in Fig. 2B) in cutaneous metastases.In the second group referred to as advanced disease (AD), mice had at least four metastases including cutaneous and distant ones and/or more than 65 million cells within cutaneous metastases.The total cell number did not differ at the primary tumor site between both groups (Fig. 2B).
Whereas similar proportions and densities (number of cells for 1 million of total cells) of hematopoietic cells (CD45 + ) cells were detected in primary tumors of ED and AD mice, B cells were significantly less abundant in AD (5.1% ± 4.8; 3.6 × 10 4 cells) than in ED RET mice (9.2% ± 5.6; 7.0 × 10 4 cells; Fig. 2C).Although the proportion of B2-FO cells among B cells remained unchanged (around 70 % of the B-cell population, pie charts), their density diminished nearly by half at an advanced stage (2.2 × 10 3 cells in AD vs. 4.0 × 10 3 cells in ED for 1 million of total cells).The proportion of immature B cells among B cells increased significantly in primary tumors of AD mice (pie charts), but their density remained stable (Fig. 2C).IgM memory B cells diminished in AD RET mice, whereas IgG1 memory B cells were only detectable in a fraction of them (7/28).The proportion and density of PB and PC were similar in the primary tumors of both groups of RET mice.Thus, our results suggest that the tumor-infiltrating B-cell pool is enriched in immature B cells at the expense of B2-FO cells and IgM memory B cells, during melanoma progression.
In contrast to primary tumors, no significant modulation of total B cells was detected in metastases (Fig. 2D), although more advanced melanomas tended to be associated with reduced densities in nearly all B-cell subsets, which was only significant for PB and PC.Such a varying pattern of B-cell subset density reduction in primary uveal melanoma versus cutaneous metastases may be due to a distinct chemokine microenvironment or immune cell composition within the two tumor microenvironments.
In TdLN and spleen of RET mice, total B cells were more abundant in AD compared with ED RET mice (Table 1).Immature B cells and B2-FO cells are the two main populations in secondary lymphoid organs and their proportion was unaffected by melanoma stage.In contrast to what was observed in primary tumors, the density of B2-FO cells significantly increased with stage in both TdLN and spleen.Other notable changes in AD mice included an increase of splenic B1 cells, and a decrease of IgM + memory B cells with a parallel increase of switched IgG1 + memory B cells in both the TdLN and spleen.These observations suggest the induction of a B-cell response to tumor-associated antigens at these sites.Finally, proportions of PB and PC did not significantly (1.6-fold in TdLN) vary during melanoma progression whatever the compartment studied, but were higher in primary tumors than in spleen or TdLN (Table 1).
Altogether, our findings show that the development of spontaneous melanoma is accompanied by site-specific changes in B-cell subsets.In particular, a strong decrease of B2-FO cells in the primary tumor contrasts with their accumulation in the TdLN and spleen.

RET melanoma progression is accompanied by upregulated or higher IL4I1 expression in immature and B2-FO B cells
Recently, we have shown that splenic B2-FO cells intrinsically produce the phenylalanine oxidase IL4I1 that inhibits B2-cell response to T-dependent antigens in C57BL/6 mice [30].In this study, we also observed that IL4I1 negatively regulates B-cell egress from the bone marrow suggesting a role for IL4I1 on B-cell mobility that has never been explored yet.To better characterize the dynamic of IL4I1 expression in B cells during tumor progression, we generated a novel mouse model in which the yellow fluorescent protein (YFP) is expressed under the control of the IL4I1 promoter (IL4I1 YFP ; Supporting information Fig. S3A and  B).As expected, the MFI of YFP significantly increased in splenic B cells from IL4I1 YFP after in vitro stimulation with recombinant IL4 compared with unstimulated B cells (Supporting information Fig. S3C).This reporter mouse was crossed with RET mice to obtain RET IL4I1 YFP mice.The similar transcriptional IL4I1 expression in RET and RET IL4I1 YFP mice both in spleen and tumors indicated that the insertion of the YFP construct did not influence the enzyme expression (Supporting information Figs.S3D and E).Next, we measured IL4I1 protein expression by comparing YFP MFI in cells from RET IL4I1 YFP mice and YFP − RET mice (background fluorescence) as illustrated for splenic total B cells (Supporting information Fig. S3F).A progressive increase of IL4I1 expression was observed in the primary tumor and cutaneous metastases during melanoma progression, both in total B cells and B2-FO cells (Fig. 3A).Unfortunately, tumor-infiltrating immature B cells were too few to allow a proper study of IL4I1 expression.Total B cells expressed IL4I1 in both spleen and TdLN at early and advanced stages of melanoma (Fig. 3B).IL4I1 expression was significantly upregulated in splenic and nodal immature B cells (6.7fold in spleen and threefold in TdLN) and B2-FO cells (twofold at both sites) in AD compared with ED RET mice.In AD RET mice, the strongest expression of IL4I1 was found in immature splenic B cells, a key stage of B-cell development where these B cells are     exposed to multiple selection signals before their local maturation into B2-FO cells [31].Altogether, our data show a strong upregulation of IL4I1 expression in B2-FO cells and immature B cells during melanoma progression that might reduce their mobility or antitumor properties.

IL4I1 impairs the recruitment of B2 cells in primary tumors during melanoma progression
Extending our prior data supporting that IL4I1 exerts a global detrimental effect on tumor progression [29], total cells in cutaneous metastases were less numerous in ED and AD RET mice lacking IL4I1 compared with their RET counterparts (Fig. 4A).
In absence of IL4I1, immature B cells were significantly more abundant in primary tumors, independently of the aggressiveness, whereas B2-FO cells only accumulated at an advanced stage of the disease (Fig. 4A).Although the cell amount in ED cutaneous metastases in RET IL4I1 KO mice was insufficient to perform immunophenotyping, we characterized B-cell subsets in AD tumors.No difference in immature B cells and B2-FO cell densities was observed between metastases from RET IL4I1 KO and RET mice (Fig. 4A).IL4I1 deficiency also resulted in a significant accumulation of splenic B2 cells mainly due to the early doubling of immature B cells and a more delayed increase of B2-FO cells (Fig. 4A).Whereas a significant increase in immature B cells was also observed in TdLN at both stages of the disease, the increase in total nodal B cells reached significance only at advanced stage (Supporting information Fig. S4 for TdLN).
Collectively, our results suggest that IL4I1 inactivation favors a more potent influx of immature B cells and B2-FO cells within the primary melanoma from the early stages of the disease.Although immature B cells also strongly accumulate in the spleen and TdLN of ED and AD RET IL4I1 KO mice, B2-FO cells significantly accumulate in the spleen only at advanced stages of the disease.According to B-cell depletion experiments, a rapid accumulation of B2 cells likely contributes to delayed tumor progression, with its beneficial effect being antagonized in RET mice by progressive upregulation of IL4I1 expression.

CXCR5 expression increases on splenic immature B cells in the absence of IL4I1
As CXCL13/CXCR5 and CXCL12/CXCR4 are the main chemokine receptor pairs that orchestrate B-cell trafficking, we wondered whether IL4I1 expression interfered with the expression of these molecules during the progression of the RET melanoma.CXCL12 levels neither changed with disease stage or IL4I1 expression status in the aqueous humor (in which primary tumor bathes) nor in the serum of RET mice (Fig. 4B).In contrast, CXCL13 levels increased with stage in both primary tumors and serum, with similar levels in RET and RET IL4I1 KO mice (Fig. 4B).
Comparing splenic B cells from RET IL4I1 KO mice and RET mice, we found no difference in the relative MFI of surface CXCR5 on B2-FO cells (Fig. 4C), but a tendance to higher levels on immature B cells derived from RET IL4I1 KO mice.This tendency reached significance when investigating CXCR5 expression on splenic immature B-cell subsets (see gating strategy in Supporting information Fig. S1A), as MFI was higher on T2 B cells of ED RET IL4I1 KO mice as well as on T1 and T3 B cells of AD RET IL4I1 KO mice compared with RET mice (Supporting information Fig. S4C).In contrast, in the primary tumor, immature B cells and B2-FO cells of RET and RET IL4I1 KO mice expressed similar levels of CXCR5 (Supporting information Fig. S5A).They also expressed similar levels of CXCR4, both in the spleen and in the primary tumor (Supporting information Fig. S5B).
Collectively, our data show a higher CXCR5 expression by splenic immature B cells in the absence of IL4I1, which likely contributes to their preferential accumulation in the spleen of RET IL4I1 KO mice compared with RET mice.Higher CXCR5 expression by immature B cells may also subsequently facilitate their migration to the tumor site despite no difference in CXCL13 production being observed between RET IL4I1 KO and RET mice.

Intrinsic IL4I1 expression by B2 cells impairs their capacity to control melanoma
To directly investigate the homing and antitumor properties of IL4I1-deficient B cells in vivo, B2 cells from RET and RET IL4I1 KO ED mice were adoptively transferred into μMT mice, a strain lacking B cells, prior to subcutaneous transplantation of B16 melanoma cells (Fig. 5A).Animals were monitored for tumor growth, then sacrificed before analyzing the immune composition of the tumor, TdLN, and spleen.Tumor growth was significantly delayed in mice receiving B2 cells from RET mice and even more in mice receiving B2 cells from RET IL4I1 KO mice (Fig. 5B).
The adoptive transfer of B2 cells strongly altered the tumor immune microenvironment, with an increase of effector memory T cells, TH1 CD4 + T cells, CD8 + T cells expressing T-bet (Tc1) or Granzyme B (GrB), and GrB + NK cells accompanied by a decrease of regulatory T cells -both FoxP3 + and FoxP3 − CD25 high OX-40 + LAG3 + ICOS + GITR + subsets -and PMN-MDSC (Fig. 5C and D; Supporting information Fig. S6 for T-cell subset gating strategy).All these modifications were more drastic when B2-cells came from RET IL4I1 KO mice.
We next compared the antitumor properties of B2 cells derived from ED and AD RET mice, which express low and high levels of IL4I1, respectively (Fig. 3).B2 cells from AD mice controlled B16 melanoma growth less efficiently than those derived from ED mice (Fig. 5E).This was associated with reduced proportions of effector memory CD4 + and CD8 + T cells within the tumor microenvironment (Fig. 5F) and increased proportions of PMN-MDSC (Fig. 5G).No differences in the proportions of regulatory T cells were observed (not shown).
Finally, we compared in vitro the capacity of B2-FO cells from RET and RET IL4I1 KO mice to activate naïve CD4 + T cells.In the presence of staphylococcal enterotoxin B (SEB), IL4I1-deficient B2-FO cells generated more effector memory T cells (Fig. 6A) and induced more intense T-cell proliferation (Fig. 6B) than IL4I1expressing B2-FO cells.Although the production of GM-CSF was similar in both conditions, that of IL10 was lower in the presence of RET IL4I1 KO -derived B2-FO cells (Fig. 6C).We also compared the capacity of purified B2-FO cells derived from RET neg., RET, and RET IL4I1 KO mice to produce cytokines in response to CpG.Compared with RET neg.littermates, we detected higher production of IL6 by B2-FO cells from both RET mice and RET IL4I1 KO mice and higher production of IL-10 and GM-CSF by B2-FO cells from RET mice only (Fig. 6D).Notably, CpG-stimulated IL4I1-deficient RET B2-FO cells significantly produced less IL-10 than their IL4I1-competent counterpart (Fig. 6D).
Collectively, these data show that intrinsic expression of IL4I1 by B2 cells promotes the production of IL10 and IL6, the IL10/IL6 ratio being consistent with an immunosuppressive microenvironment in RET mice.IL4I1 expression also diminishes the capacity of B2 cells to costimulate T cells, and, most importantly, is associated with the establishment of a protumor immune microenvironment.

Discussion
The role of B cells in cancer has recently emerged [32][33][34][35].These cells act both as antigen-presenting and effector cells that produce cytokines and antibodies [36,37].Despite conventional (B2) and innate-like (B1) B cells being present in uninflamed and inflamed skin, their respective contribution to melanoma aggressiveness  remains poorly investigated.The present study deciphers the role of B2 cells in the RET model.First, we have shown that these cells delayed metastatic dissemination by promoting CD4 + and CD8 + T-cell functions during the early phase of the disease.These results are consistent with the importance of B cells in presenting tumor antigens to CD4 + and CD8 + T cells in the B16 melanoma model [38].Our study further evidences that the antitumor role of B cells relies mainly on B2 cells.Our observation of a significant increase in IgG1 + switched memory B cells, associated with a reduction of IgM + B cells in secondary lymphoid organs during tumor development, is consistent with our previous detection of antibodies specific to melanoma cells in the RET model [29].These results argue for the induction of a B-cell response specific to tumor antigens in the TdLN and spleen with possible relocation of PB and PC at tumor sites.These antigens remain to be characterized in early and advanced disease as well as their capacity to possibly elicit the production of different antibody isotypes as described in melanoma patients [39,40].In human melanoma, low IgG1 titers, high IgG4 titers, or oligoclonal IgA correlate with reduced overall survival [12,41], suggesting opposite roles of specific antibodies in tumor escape.
Previous data in the B16 melanoma model established that the release of IL10 by B1 cells reduces TH1 cytokine production by tumor-infiltrating CD8 + T cells and drives the polarization of tumor-associated macrophages into an M2 phenotype [42,43].Here, we showed that splenic B2 cells from RET mice compared with WT (RET Neg) mice are able in vitro to produce more IL10 and GM-CSF, likely facilitating the recruitment of regulatory T cells and PMN-MDSC, two key immunosuppressive actors involved in melanoma aggressiveness [44,45].Altogether, our study supports that B2 cells may interfere with anti-melanoma immunity.
Our previous work has shown that total B cells infiltrating primary tumors were significantly more frequent in 3-month-old RET IL4I1 KO mice compared with age-matched RET mice [29].Here, we evidenced that a more intense infiltration of B2-FO cells in the primary tumor of RET IL4I1 KO mice contributes to this phenotype at an advanced stage of the disease.Immature B cells were more frequent at the early and advanced stages of the disease in the primary tumor of RET IL4I1 KO mice, but also in secondary lymphoid organs.Thus, the genetic inactivation of IL4I1 in the melanoma context might endow immature B cells with an enhanced ability to migrate into lymphoid organs, as it occurs in C57/BL6 mice (physiological conditions) due to their accelerated egress from bone marrow [30].Consistently, the present study highlights higher CXCR5 expression by splenic immature B cells of RET IL4I1 KO mice compared with RET mice that might contribute to their accumulation into the spleen from the early stage and synergizes with higher serum CXCL13 levels during melanoma progression.However, early accumulation of immature IL4I1 KO B cells in primary tumors might depend on alternative mechanisms since neither CXCL13 levels nor CXCR5 expression are upregulated at this site.The precise mechanism involved in the accumulation of B2-FO cells in primary tumors and spleen of RET IL4I1 KO mice at a late stage of the disease also deserves further investigation.In the spleen, it might be a direct consequence of more immature B cells being locally available for differentiation into mature B2-FO cells.These cells might secondarily migrate in higher numbers toward primary tumors and TdLN.Increasing levels of CXCL13 during the course of melanoma progression likely facilitate this recruitment.However, CXCL13 levels are similar in the primary tumor and serum of RET IL4I1 KO and RET mice, arguing for IL4I1-independent mechanisms regulating CXCL13 production in this model.Thus, IL4I1 exerts a key role in immature B-cell trafficking in healthy [30] and RET mice.
Our data further show that IL4I1 is expressed by immature and B2-FO cells and is upregulated during melanoma progression.This expression stimulates their capacity to produce IL10 and GM-CSF and likely thwarts their ability to generate effector memory T cells.Recent data support the key role of CXCL13 and tumor-infiltrating B cells in the generation of the tumorspecific immune response, the prolonged survival of mice transplanted with B16 melanomas treated with an anti-PD-L1 anti-body and the improved overall survival of melanoma patients [46].In our model, high tumor and circulating levels of CXCL13 at advanced stages of the disease are consistent with enhanced immune activation and/or humoral response to tumor antigens in secondary lymphoid organs [47,48].This result is reminiscent of CXCL13/CXCR5 interactions that guide the formation of mature tertiary lymphoid structures in human metastatic melanoma [49] where CXCL13 is abnormally produced by exhausted CD8 + T cells [50].
Collectively, our data indicate that IL4I1 reduces the infiltration and antitumor properties of immature B cells and B2-FO cells at the tumor site and in secondary lymphoid organs.The deleterious impact of IL4I1 strengthens the rationale for the development of adjuvant therapy inhibiting IL4I1 to improve B-cell response to melanoma and current anti-PD1 treatment.

Mice and melanoma cell line
MT/ret +/− transgenic mice (designated as RET), generated on the C57BL/6J background, heterozygously express the human RET constitutively activated oncogene.They were used at early and advanced stages of melanoma, and age-matched nontransgenic MT/ret −/− littermates (designated as RET negative littermates, RET neg.) were used as control.Mice deficient in the IL4I1 gene (IL4I1 KO ) were purchased from Taconic.Mice exhibiting fluorescent IL4I1 + cells (designated as IL4I1 YFP ) were set up in the lab in collaboration with CIPHE (Marseille Luminy) to study the protein expression of IL4I1 by B-cell subsets.Genetic engineering of the IL4I1 YFP mouse and genotyping are illustrated in Supporting information Fig. S2A and B.Then, IL4I1 YFP mice were crossed with RET mice to obtain RET IL4I1 YFP mice and qRT-PCR were performed to ensure that the insertion of the YFP construct did not affect the regulatory region of the IL4I1 gene (Supporting information Fig. S2D and E).RET IL4I1 KO mice have been already described [29].B cell-deficient mice (μMT mice [51] kindly provided by Pr.C.A. Reynaud (Institut Necker-Enfants-Malades)) were used for adoptive transfer.
All mice were maintained in our Cochin Institute facilities under specific pathogen-free conditions.Sample sizes were chosen to ensure the reproducibility of the experiments in accordance with the replacement, reduction, and refinement principles of animal ethics regulation.

In vivo depletion of B cells
RET mice were injected i.p with endotoxin-free anti-mouse CD20 (clone 5D2, IgG2a, MTA Genentech) or isotype-matched control mAb (#400282, clone MOPC-173, BioLegend) according to the following protocol: 125 μg at the 4th and 10th day post-birth, then 250 μg from the 20th to the 45th day.The control group included 11 mice (6 males and 5 females) and the anti-CD20treated group included 10 mice (5 males and 5 females).

Single-cell suspension procedures
Lymph nodes and spleen were mechanically dissociated, homogenized, and passed through a nylon cell strainer (70 μm) in a complete medium.Tumors were dissociated after digestion with 1 mg/mL of collagenase A (#11088793001, Roche) and 0.1 mg/mL of DNase I (#11284932001, Roche) for 20 min at 37°C.Treatment was blocked by the addition of FBS and washing.Red blood cells were lysed from cell suspensions using ammonium-chloride-potassium (#A10492-01, Gibco).Cells were resuspended in staining buffer PBS, (#10010-015, Gibco) supplemented with 5% FBS and 2.5 mM EDTA (#15575-038, Life Technologies) and counted on FACScan (BD Biosciences).

Flow cytometry and antibodies
6 × 10 6 cells from cervical lymph nodes, spleen, and tumors were dispensed into 96-well V-bottom plates (#651101, Greiner).After washing with PBS, single-cell suspensions from tumors were incubated with LIVE/DEAD Far Red staining kit (#L10120, Life Technologies) or Zombie NIR viability kit (#423106, BioLegend), according to the manufacturer protocol.After blocking Fc receptors using 5 μg/mL of anti-mouse CD16/CD32 mAb (#BE0307, clone 2.4G2, BioXCell), cells were washed in staining buffer, then stained for 30 min at 4°C with the appropriate combination of antibodies (Supporting information Table S1).After two washes, cells were fixed in 2% PFA for 10 min at 4°C.For intra-cellular staining, the transcription factor staining buffer set (#00-5523-00, Invitrogen) was used and antibodies were incubated for 1 h (for GrB) or overnight (for T-bet and FOXP3) before extensive washing.Data were acquired on LSR II or Fortessa (BD Biosciences) and analyzed with FACSDiva software (BD Biosciences).Histogram overlays were realized on FlowJo software (BD Biosciences).The authors have adhered to the "Guidelines for the use of flow cytometry and cell sorting in immunological studies".

In vitro stimulation
Single-cell suspensions from the spleen (2 × 10 6 cells/mL) were cultured in a complete medium with or without 20 ng/mL of recombinant IL4 (#404-ML, R&D Systems).On day 2, the IL4I1 YFP signal in total B cells was assessed by flow cytometry.

Quantitative RT-PCR
Tumors and spleen samples were collected in RNAlater (Sigma).RNA was extracted using the Macherey-Nagel nucleospin RNA kit.cDNA was reverse transcribed (Thermoscientific Maxima First Stand c DNA) from 1000 ng RNA.IL4I1 expression was measured by qRT-PCR with LightCycler 480 SYBR Green I Master Roche and specific primers against the IL4I1 gene [23] and was normalized against GAPDH expression (primers 5 -TGCGACTTCAACAGCAACTC-3 and 5 -ATGTAGGCCATGAGGTCCAC-3 ).Real-time PCR cycling conditions were as follows: 95°C for 5 min; 45 cycles of 95°C for 10 s, 55°C for 10 s, and 72°C for 10 s.

Quantitative determination of chemokine concentrations
For aqueous humor collection, primary tumors were mechanically dissociated in 150 μL of PBS and lysates were kept at −80°C until use.CXCL13 and CXCL12 were quantified using Quantikine ELISA kit (#MCX130, and MCX120, R&D SYSTEMS, respectively), according to supplier instructions.Plates were read with CLARIOstar (BMG Labtech) plate reader and data were analyzed using MARS data analysis software (BMG Labtech).Cut-off values were 2.84 pg/mL for CXCL13 and 69 pg/mL for CXCL12.

B2 cell isolation, culture, and cytokine production study
B cells (90%-95%) were purified from the spleen using a negative B-cell isolation kit (Dynabeads Mouse CD43 Untouched B cells, #11422D, Invitrogen), according to supplier instructions corresponded to B2 cells.Splenic B2 cells were kept in ice-cold PBS prior to adoptive transfer into μMT mice (see below).
Alternatively, splenic B2 cells (4 × 10 5 /well of 96-well plates) were cultured in complete medium alone or with 5 μg/mL CpG-B (#tlrl-1826, InvivoGen) at 37°C with 5% CO 2 .Each culture condition was run in duplicates.On day 3, cells were stained with LIVE/DEAD and analyzed by flow cytometry.Culture supernatants were frozen until use.Levels of GM-CSF, IL10, and IL6 were measured in culture supernatants using the U-PLEX Biomarker Group 1 Assay (#K15069L-1, Meso Scale Discovery), according to supplier instructions.

B-cell adoptive transfer and tumor-cell inoculation
μMT mice received 15 × 10 6 B2 cells intraperitoneally.One day later, 10 6 B16 melanoma cells were injected subcutaneously into the right shaved flank.Mice were checked for the presence of a palpable tumor, and tumor growth was measured with a caliper for 14 days.Tumor size was calculated as the product of bisecting tumor diameters.Mice bearing a tumor with a diameter over 10 mm were killed according to guidelines for animal experimentation.

Antigen-presenting assay
Naïve CD4 + T cells were purified from lymph nodes by negative selection using Dynabeads Untouched mouse CD4 cells kit (#11415D, Invitrogen) according to supplier instructions.10 6 cells/mL were labeled with 2.5 μM CFSE (Molecular Probes) in complete medium at room temperature for 7 min.5 × 10 4 naive CFSE-labelled CD4 + T cells were distributed in 96-well plates and co-cultured with 5 × 10 4 purified B2 cells, with or without 1 μg/mL SEB.Each culture condition was run in duplicates.On day 5, cells were stained with LIVE/DEAD and analyzed by flow cytometry.Culture supernatants were frozen until use.

Statistics
Data are expressed as mean ± SD or as median.For comparison between the two groups, a Mann-Whitney t-test was used.For multigroup comparisons, a two-way ANOVA test was used.For survival curves, a Mantel-Cox test was used.All statistical analyses were performed with Prism 5 software (GraphPad software).

Figure 1 .
Figure 1.Accelerated tumor cell dissemination and reduced T-cell activation in RET mice treated with anti-CD20 antibody.(A) Illustration of the experimental design.Mice were treated with the anti-CD20 antibody or isotype control at days indicated with arrows and monitored weekly from the weaning.(B and C) Time courses of primary tumor and cutaneous metastases onset, respectively.Control treated with the isotype antibody (n = 11) and mice treated with anti-CD20 (n = 10) are represented in black and red lines respectively.(D) Average number of macroscopic cutaneous metastasis per mouse at day 60.(E and F) Proportions of total B cells in primary tumors, cutaneous metastases (E), tumor-draining lymph nodes (TdLN), and spleen (F) of control (black circles) and anti-CD20 (red circles) treated mice.(G) Mean proportions of immature (IMM.)B cells (CD19 + CD93 + CD43 -), mature B2-FO (CD21 + CD23 + ), PB and PC (CD138 + CD19 + and CD19 -respectively), B1 (CD43 + CD23 -), and a pool of other B cells (including memory at all sites) in primary tumors (left) and metastases (right), respectively.(H) Densities (Absolute numbers for 10 6 total cells) of IMM.B cells, B2-FO, PB+PC, and B1 cells in primary tumors and cutaneous metastases of control and anti-CD20 treated mice.(I) Mean proportions of IMM, B2-FO, PB+PC, B1, and a pool of other B cells (including memory at all sites and B2-MZ in spleen) in TdLN (left) and spleen (right), respectively.(J) Densities of IMM.B cells, B2-FO, PB+PC, and B1 cells in TdLN and spleen of control and anti-CD20 treated mice.(K) Proportions of effector memory CD44 + cells among CD8 + and CD4 + T cells and of TH1 T-bet + CXCR3 + cells among CD44 + CD4 + T cells in TdLN and spleen.(L) Proportions of PMN-MDSC (CD11b + CD11c -Ly6G + Ly6C int ) among CD45 + cells in TdLN and spleen (left) and primary melanoma (right).Black circles symbolize control (n = 24) and red circles represent anti-CD20 treated mice (n = 23).Data were obtained from at least three independent experiments.Results are expressed as median (H and J) or as mean ± SD (other panels).Statistical analysis was performed using a Mantel-Cox test for (B and C) and a Mann-Whitney t-test for other panels (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001).

Figure 2 .
Figure 2. Remodeling of B-cell subsets during melanoma progression in RET mice: decrease of B2-FO cells within the primary tumor.(A) Illustration of early and advanced melanoma stages in 3-to 6-month-old RET mice.Clinical diagnosis was performed at autopsy for the occurrence of primary melanoma, cutaneous and distant (lymph nodes, lung, bladder) metastases.(B) Absolute numbers of total cells at primary and metastatic tumor sites in ED (empty circles) and AD RET mice (full circles).ED RET mice exhibit less than 65 million of cells (dotted line) at cutaneous metastatic site and no distant metastasis at autopsy.Most AD mice exhibited more than 65 million cutaneous tumor cells.All mice displaying at least one distant metastasis were considered AD mice.(C, dot plots, and D, dot plots) Proportions and densities of CD45 + cells, total B cells (CD19 + ), and B-cell subsets (IMM), mature B2-FO, PB, and PC (CD138 + CD19 + and CD19 -respectively), IgM (IgM + IgG1 -), and IgG1 memory (IgG1 + IgD -) cells) within the primary tumor (C) and cutaneous metastases (D) of ED and AD mice.(C, pies, and D, pies) Proportions of IMM.B cells, B2-FO cells, PB and PC, and other B cells among total B cells within the primary tumor (C) and cutaneous metastases (D).Empty and full circles represent ED (n = 21) and AD RET mice (n = 44), respectively.Data were obtained from at least three independent experiments.Statistical analysis was performed using a Mann-Whitney t test (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001).

Figure 3 .
Figure 3. Increased IL4I1 expression in B2 cells during melanoma progression.(A) IL4I1 YFP mean fluorescence intensity (MFI) in total B cells and B2-FO cells of primary tumor and metastases from RET IL4I1 YFP ED and AD mice.(B) IL4I1 YFP MFI in total B cells, immature B cells, and B2-FO cells of spleen and tumor-draining lymph nodes (TdLN) from RET IL4I1 YFP ED and AD mice.The dotted line represents the threshold of the YFP signal in RET ED + AD mice (n = 15).Empty and full diamonds represent ED (n = 4) and AD RET mice (n = 11), respectively.Empty and full circles represent ED (n = 5) and AD RET IL4I1 YFP mice (n = 35), respectively.Data were obtained from at least three independent experiments.Results are expressed as a median.Statistical analysis was performed using a Mann-Whitney t-test (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001).

Figure 4 .
Figure 4. CXCR5/CXCL13 pair in the shaping of B-cell subsets during melanoma progression in RET mice deficient for IL4I1.(A) Absolute number of total cells and densities of immature (IMM) B cells and mature B2-FO cells within the primary tumor, metastases, and spleen of ED and AD RET and RET IL4I1 KO mice.Empty and full circles represent ED (n = 21) and AD (n = 44) RET mice.Empty and full squares represent ED (n = 25) and AD (n = 22) RET IL4I1 KO mice, respectively.(B) CXCL12 and CXCL13 concentrations in aqueous humor and serum of RET and RET IL4I1 KO ED vs. AD mice.Values are reported in pg/mL.Circles and squares represent RET and RET IL4I1 KO mice, respectively.(C) Relative mean fluorescence intensity (MFI) for CXCR5 from splenic B2-FO, total IMM, and IMM.B-cell subsets (T1, T2, T3) derived from ED and AD RET (n = 4 and n = 14, respectively) and RET IL4I1 KO (n = 9 and n = 10, respectively) mice.Data were obtained from at least three independent experiments.Results are expressed as a median.Box-and-whisker plots display the minimum, median, and maximum values.Statistical analysis was performed using a Mann-Whitney t-test (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001).

Figure 6 .
Figure 6.Intrinsic IL4I1 expression by B2 cells impairs their capacity to activate T cells and promotes their IL10 production.(A) Representative dot-plots of CD4 + CD44 + T-cell proportions obtained by flow cytometry and proportions of effector memory CD44 + CD4 + T cells stimulated with or without SEB, in presence of B2 cells from RET and RET IL4I1 KO mice.(B) Representative histograms of CFSE dilution obtained by flow cytometry and proportions of proliferating CD4 + T cells stimulated with or without SEB, in presence of B2 cells from RET and RET IL4I1 KO mice.(C) Levels of IL10 and GM-CSF produced in culture supernatants.Light and dark grey histograms illustrate the addition of B2 cells from RET (n = 5) or RET IL4I1 KO (n = 6) mice, respectively.ND for not detected.(D) Levels of IL10, GM-CSF, and IL6 produced by B2 cells from RET neg.(n = 3, empty triangles), RET (n = 6, full circles), and RET IL4I1 KO (n = 6, full squares) mice in response to CpG.Values for nonstimulated conditions are comprised between 1.5-2 pg/mL for IL10, 0.05 pg/mL for GM-CSF, and 2.2-2.7 pg/mL for IL6.Results are expressed as median (D) or as mean ± SD (other panels).Statistical analysis was performed using a Mann-Whitney t-test (*p < 0.05; **p < 0.01).

Table 1 .
Densities and proportions of B cells present in tumor-draining lymph nodes and spleen from ED vs. AD RET mice.