Loss of Lymphotoxin Alpha-Expressing Memory B Cells Correlates with Metastasis of Human Primary Melanoma

Activated antigen-experienced B cells play an unexpected complex role in anti-tumor immunity in human melanoma patients. However, correlative studies between B cell infiltration and tumor progression are limited by the lack of distinction between functional B cell subtypes. In this study, we examined a series of 59 primary and metastatic human cutaneous melanoma specimens with B cell infiltration. Using seven-color multiplex immunohistochemistry and automated tissue imaging and analysis, we analyzed the spatiotemporal dynamics of three major antigen-experienced B cell subpopulations expressing lymphotoxin alpha (LTA/TNFSF1) or interleukin-10 (IL-10) outside tertiary lymphoid structures. The expression of both LTA and IL-10 was not restricted to a particular B cell subtype. In primary melanomas, these cells were predominantly found at the invasive tumor-stroma front and, in metastatic melanomas, they were also found in the intratumoral stroma. In primary melanomas, decreased densities of LTA+ memory-like and, to a lesser extent, activated B cells were associated with metastasis. Compared with metastatic primary tumors, B cell infiltrates in melanoma metastases were enriched in both LTA+ memory-like and LTA+ activated B cells, but not in any of the IL-10+ B cell subpopulations. Melanoma disease progression shows distinct dynamics of functional B cell subpopulations, with the regulation of LTA+ B cell numbers being more significant than IL-10+ B cell subpopulations.


Introduction
There is now conclusive and independent evidence that antigen-experienced activated B cells play an unexpected, essential role in anti-tumor immunity in cancer patients, including the regulation of anti-tumor T cell responses (reviewed in [1]).
Melanomas are malignant primary tumors originating from melanocytes located on the skin or mucosal surfaces. These malignancies are often characterized by high aggressiveness and poor prognosis [2]. In human melanoma, B cells have been shown to regulate CD8 + T cell recruitment to tumor sites, promote tumor-associated inflammation  The cohort included 36 patients with primary cutaneous melanoma who were between 31 and 93 years of age at the time of initial diagnosis. Twenty-two patients presented without metastasis within a maximum follow-up of 194 months (mean: 69 months, Table 1). Fourteen patients developed metastasis within a maximum follow-up interval of  Table 2). None of these patients received local or systemic anti-tumor treatment before surgery for the primary tumor.
In addition, we analyzed whole FFPE tissue sections from 23 human melanoma metastases (Supplementary Table S1). These samples were collected at the University Hospital Basel between the years 2015 and 2019. These tumor samples were also collected with the informed consent of the patients and the pathology files were obtained with approval from the local ethics committee (EKNZ vote BASEC 2019-00927). Histologic diagnoses were made by board-certified pathologists at the Institute of Pathology, University Hospital Basel, under the direction of K.G. In lymph node metastases, tumor deposits had completely or almost completely replaced lymph node tissue; in the latter, tumor deposits could clearly be histologically separated from the remains of lymphatic tissue.
Negative controls included the use of isotype instead of primary antibodies and staining without primary antibodies. Single antibody stainings were run in parallel to control for false positive (incomplete stripping of antibody-tyramide complexes) and false negative results (antigen masking by multiple antibodies, "umbrella-effect"), as well as for spillover effects (detection of fluorophores in adjacent channels), as previously described [3,24,25]. Reproducibility was controlled by a reference slide in each run, and antibody batches were not changed in this study.

Automated Acquisition and Quantification of LTA + and IL-10 + B Cell Subpopulations
We scanned multiplexed slides of whole-tumor sections on a Vectra 3 Automated Quantitative Pathology Imaging System (version 3.0.5., Akoya), which were then analyzed after spectral unmixing with the inForm ® Tissue Finder™ (version 2.4.1, Akoya) [3,24,25]. Tumor tissue areas with ulceration or with densely packed clusters of lymphoid cells, e.g., in TLS, did not allow for a clear assignment of fluorophore signals to individual cells and were excluded from our analysis. When present, remnants of lymphatic tissue in nodal metastases were also excluded.
Based on differential expression of CD19, CD20, CD38 and CD27, three B cell phenotypes were determined along the lines we described previously [3]: (i) CD19 + CD20 − CD38 − CD27 + as activated B cells, (ii) CD19 + CD20 + CD38 − CD27 var as memory-like B cells, (iii) CD19 + CD20 − CD38 + as plasmablast-/ plasma cell-like cells or antibody secreting cells and (iv) other cell types. Based on expression signals for LTA or IL-10, these B cell subpopulations were further classified as LTA + or IL-10 + (Figure 1, Supplementary Figure S1). Cancer tissues including melanoma are known for strong plasma positivity for cytokines/chemokines including IL-10 [26,27]. Thus, immunostaining for secreted cytokines/chemokines often shows staining of the tissue surrounding cells with specific cytoplasmic expression. Therefore-as described above-control staining with isotype-matched antibodies and single antibody stainings were used to control for both false-positive results and spillover effects. In addition, we defined background staining as the weak signal obtained for fibroblasts in the tumor microenvironment. These cells are known to produce very low-if any-levels of LTA and IL-10 themselves [26,27], but can bind cytokines via their corresponding receptors on the cell surface, e.g., anchoring heterotrimeric LTA/B via TNFR3. We, therefore, randomly selected 50 stamps from 10 different tissue samples in which we determined the mean and standard deviation of signal intensities for LTA and IL-10 in a total of 250 intratumoral fibroblasts. Only B cells with signal intensities above the upper standard deviation from the mean of fibroblasts were considered positive. As CD27 expression can be downregulated on tumor-infiltrating B cells [28,29], the number of activated B cells may be underrepresented. All phenotyping and subsequent quantifications were performed blinded to sample identity.

Statistical Evaluation
Cell-level data (i.e., intensities per channel for the cell compartments "membrane", "cytoplasm", and "nucleus") were exported from inForm ® Tissue Finder™ (version 2.4.1, Akoya) as text files and processed using R (version 4.0.3). Intensity thresholds were manually adjusted after an inspection of each slide. Based on these thresholds, markers were defined as positive or negative and cells were assigned to specific phenotypes based on the above marker combinations.

Experimental Strategy
B cells do not homogeneously infiltrate primary human melanomas throughout the tumor tissue, but preferentially at the invasive tumor front, and also sometimes intratumorally, in a scattered and patchy manner [3,12,24]. In melanoma metastases, B cells are predominantly found in the intratumoral stroma around tumor nests and in the peritumoral stroma at the invasive tumor front [3]. We, therefore, performed seven-color multiplex immunohistochemistry and automated tissue imaging and analysis on whole-tissue sections from human melanoma samples.
Our previous work on the spatiotemporal distribution of B cell subpopulations in primary human melanoma further showed the presence of activated and memorylike B cells, as well as of plasmablast-like antibody secreting cells, whereas other subpopulations such as plasma cell-like antibody secreting cells, germinal center-like and transitional/regulatory-like B cells were rarely found outside TLS [3]. We, therefore, Diagnostics 2021, 11, 1238 7 of 14 decided to focus our analyses on the detection of the three first-mentioned B cell subpopulations by differential expressions of CD19, CD20, CD27 and CD38. In contrast to primary melanomas, melanoma metastases contain also considerable numbers of plasma cells. Here, the aforementioned markers cannot differentiate between plasmablast-and plasma cell-like antibody secreting cells. To allow for an analysis of the different B cell subpopulations within individual tumor samples, we included only tumor samples in which we could detect at least 50 B cells per mm 2 by CD19 and/or CD20 immunoreactivity (36 primary melanomas, 23 melanoma metastases) in this comparison.
We then determined the densities (cells/mm 2 ) of each of the LTA + and IL-10 + B cell subpopulations in the primary melanomas and their association with the most important categorical clinicopathologic parameters. Due to the sometimes considerable size of the melanoma metastases, we decided to analyze B cell subtypes there in areas with dense immune cell infiltrates, but not to evaluate B cell numbers per entire tissue area. Therefore, we compared the CD19 + and/or CD20 + B cell infiltrates of metastases with those of primary tumors in terms of their composition (relative frequencies) for LTA + and IL-10 + B cell subpopulations.

Metastasis of Primary Human Melanoma Is Associated with a Decrease in the Number of Intratumoral LTA + Activated and Memory-Like B Cells
The expression of LTA and IL-10 was not restricted to a distinct B cell subtype but detected in each of the analyzed B cell subpopulations, namely activated and memory-like B cells and antibody secreting cells (Figure 1, Supplementary Figure S1). These cells were predominantly found at the invasive tumor-stroma front of primary human melanomas, and sometimes associated with a patchy intratumoral infiltration. Thus, their distribution followed the previously described infiltration pattern of CD20 + B cells and of activated and memory-like B cell subpopulations [12,24].
We detected LTA + B cell populations in all 36 primary melanoma samples (100%, Table 3). The cell densities were highest for LTA + antibody secreting cells, followed by LTA + memory-like and activated B cells. IL-10 + B cells were also present in all 36 primary melanoma samples (100%). Again, the densities were highest for IL-10 + antibody secreting cells, followed by IL-10 + memory-like and activated B cells, but generally lower than those of their LTA + counterparts (Table 3).
Primary tumor samples included both primary tumors that metastasized and those that did not. We compared these two tumor subgroups with each other because metastasis is the most important prognostic factor for melanoma patients.
Primary human melanomas that metastasized had significantly fewer LTA + memorylike B cells (mean 0.612 ± 1.032 vs. 7.419 ± 14.398 cells/mm 2 , p = 0.04, CI 95% = 0.1775 to 3.8572, Bonferroni-corrected Wilcoxon Rank Sum test) and some minor trend towards reduced LTA + activated B cells (mean 1.495 ± 2.497 vs. 6.384 ± 11.78, p = 0.19, CI 95% = 0.0318 to 5.0355, Bonferroni-corrected Wilcoxon rank sum test) than primary melanomas without metastasis ( Figure 2, Table 3). Differences were not found for LTA + and IL-10 + antibody secreting cells and IL-10 + activated and memory-like B cells.  We have previously shown that the metastasis of primary melanomas is associated with the decrease in memory-like B cell numbers [24]. To test for a preferential decrease in LTA + memory-like B cells in primary tumors with metastasis, we also compared the relative frequencies of LTA + vs. IL-10 + cells for different dynamics within the memory-like B cell subpopulation. While the relative frequency of IL-10 + cells in primary tumors with metastasis did not change (Bonferroni-corrected p = 0.61, CI 95% −0.0119 to 0.0154, Wilcoxon rank sum test), the relative abundance of LTA + cells rather decreased (Bonferronicorrected p = 0.1, CI 95% 0.0 to 0.1667, Wilcoxon rank sum test). We have previously shown that the metastasis of primary melanomas is associated with the decrease in memory-like B cell numbers [24]. To test for a preferential decrease in LTA + memory-like B cells in primary tumors with metastasis, we also compared the relative frequencies of LTA + vs. IL-10 + cells for different dynamics within the memorylike B cell subpopulation. While the relative frequency of IL-10 + cells in primary tumors with metastasis did not change (Bonferroni-corrected p = 0.61, CI 95% −0.0119 to 0.0154, Wilcoxon rank sum test), the relative abundance of LTA + cells rather decreased (Bonferronicorrected p = 0.1, CI 95% 0.0 to 0.1667, Wilcoxon rank sum test).
When primary melanoma samples were stratified for another four prognostically important categorical clinicopathologic parameters, we observed a significant association of LTA + memory-like B cell numbers with low Breslow depth (p = 0.05, Bonferronicorrected Wilcoxon Rank Sum test), but not with age, sex and ulceration. Differences in the cell densities of the other LTA + or IL-10 + B cell subpopulations were not found (Supplementary Figure S2).
Thus, LTA and IL-10 are not expressed in a distinct population, but in all analyzed B cell subpopulations and metastasis of primary melanoma is associated with a decrease in LTA + memory-like and, to a minor degree, in LTA + activated B cells but not in IL-10 + B cell subpopulations.

Frequencies of LTA + Memory-Like B Cell Subpopulations Change with Melanoma Disease Progression
We have previously shown that the composition of B cell subpopulations also changes with disease progression from primary to metastatic tumor sites [24]. Therefore, we compared primary human melanomas and melanoma metastases for the composition of LTA + and IL-10 + B cell subpopulations. As in primary tumors, neither the expression of LTA nor IL-10 was restricted to a distinct B cell subset here.
LTA + and IL-10 + B cell subpopulations were mainly found in the intratumoral stromal septa and the peritumoral stroma of melanoma metastases. This is consistent with the distribution previously described for CD20 + B cells and of antigen-experienced B cell subpopulations [12,24].
We found LTA + B cell populations in all 23 metastatic melanoma samples (100%). Relative frequencies were highest for LTA + activated B cells, followed by LTA + antibody secreting cells and memory-like B cells. IL-10 + B cells were also present in all 23 metastatic melanoma samples (100%). Here, the relative frequencies were highest for IL-10 + memorylike B cells, followed by IL-10 + antibody secreting cells and activated B cells. As seen in primary tumors, frequencies were much lower than those of their LTA + cell counterparts.
Next, we compared the relative frequencies of LTA + memory-like and activated B cells and antibody secreting cells in melanoma metastases and primary tumors, and only found significant differences in the metastasized primary tumors. Here, the metastases contained higher frequencies of both LTA + memory-like B cells (mean 0.03 ± 0.04 vs. 0.005 ± 0.007, p < 0.01 (adjusted), CI 95% −0.0118 to −0.0044, Bonferroni-corrected Wilcoxon Rank Sum test) and LTA + activated B cells (mean 0.052 ± 0.067 vs. 0.012 ± 0.012, p = 0.01 (adjusted), CI 95% = −0.0379 to −0.0075, Bonferroni-corrected Wilcoxon rank sum test). No significant differences were observed for the relative frequencies of LTA + and IL-10 + antibody secreting cells, as well as IL-10 + activated and memory-like B cells (Figure 3, Supplementary Table S2).
We have previously shown that both TLS density and maturation as well as the composition of B cell subpopulations vary between different metastatic tumor sites, particularly between lymph node and skin [24,25]. When we compared the relative frequencies of LTA + and IL-10 + B cell subpopulations in metastatic lymph nodes and skin samples, we found enrichment of IL-10 + activated B cells at lymph node sites because we could not detect IL-10 + activated B cells at metastatic skin sites (mean 0.005 ± 0.005 vs. 0 ± 0 cells, p = 0.04, CI 95% 3 × 10 −4 to 0.0113, Bonferroni-corrected Wilcoxon Rank Sum test). All other B cell subpopulations showed no significant differences (Figure 4). contained higher frequencies of both LTA + memory-like B cells (mean 0.03 ± 0.04 vs. 0.005 ± 0.007, p < 0.01 (adjusted), CI 95% −0.0118 to −0.0044, Bonferroni-corrected Wilcoxon Rank Sum test) and LTA + activated B cells (mean 0.052 ± 0.067 vs. 0.012 ± 0.012, p = 0.01 (adjusted), CI 95% = −0.0379 to −0.0075, Bonferroni-corrected Wilcoxon rank sum test). No significant differences were observed for the relative frequencies of LTA + and IL-10 + antibody secreting cells, as well as IL-10 + activated and memory-like B cells (Figure 3, Supplementary Table S2). We have previously shown that both TLS density and maturation as well as the composition of B cell subpopulations vary between different metastatic tumor sites, particularly between lymph node and skin [24,25]. When we compared the relative frequencies of LTA + and IL-10 + B cell subpopulations in metastatic lymph nodes and skin samples, we found enrichment of IL-10 + activated B cells at lymph node sites because we could not detect IL-10 + activated B cells at metastatic skin sites (mean 0.005 ± 0.005 vs. 0 ± 0 cells, p = 0.04, CI 95% 3 × 10 −4 to 0.0113, Bonferroni-corrected Wilcoxon Rank Sum test). All other B cell subpopulations showed no significant differences (Figure 4).

Discussion
Using seven-color multiplex immunohistochemistry and an automated tissue imaging and analysis approach, we show (i) that the expression of LTA and IL-10 is not limited to a particular B cell subtype, but can be detected in any of the analyzed antigen-experienced B cell subpopulations, (ii) that metastasis of primary tumors is associated with a decrease in the densities of LTA + memory-like B cells and, to a minor degree, LTA + acti-

Discussion
Using seven-color multiplex immunohistochemistry and an automated tissue imaging and analysis approach, we show (i) that the expression of LTA and IL-10 is not limited to a particular B cell subtype, but can be detected in any of the analyzed antigen-experienced B cell subpopulations, (ii) that metastasis of primary tumors is associated with a decrease in the densities of LTA + memory-like B cells and, to a minor degree, LTA + activated B cells, but not IL-10 + B cell subpopulations, (iii) that metastatic sites are enriched for LTA + memory-like and LTA + activated B cells compared with metastatic primary tumor sites, (iv) that the B cell infiltrates in metastases at lymph node and skin sites have comparable compositions for LTA + and IL-10 + B cell subpopulations, and (v) that the composition of B cell infiltrates for IL-10 + B cell subpopulations does not change significantly with disease progression.
To date, in human melanoma, correlative studies between B cell infiltration and tumor progression are limited by the fact that they do not differentiate between functional B cell subtypes. We have shown that human melanoma cells can induce both B cell receptor and CD40 signaling in activated B cells [3]. Among other mechanisms [30], the balance between B cell receptor and CD40 signaling in activated B cells critically determines the production of immunostimulatory vs. immunoinhibitory cytokines such as LTA vs. IL-10 (reviewed in [17]).
LTA is secreted as a bioactive homotrimer (LTA3) that, like TNFA, binds to TNFR1, TNFR2, HVEM/TNFRSF14 and, unlike TNFA, as a membrane-anchored hetero-trimer with LTB (LTA1B2, LTA2B1) that binds to LTβ-R/TNFR3. Both TNFR1 and TNFR2 are potent pro-inflammatory, pro-apoptotic and, when caspases are inhibited, necrotic-like cell death-inducing receptors, but TNFR1 in particular seems to be responsible for LTAmediated cytotoxicity in vitro (reviewed in [31]). In transgenic murine models, LTA showed a pro-inflammatory function, e.g., through the induction of cell adhesion molecules and chemokines in endothelial cells (reviewed in [32,33]). In syngeneic transplantation models with melanoma cells, LTA supported natural killer (NK) cell-dependent anti-tumor activity [34] and, as a recombinant GD2 scFv-LTA fusion protein, stimulated adaptive T cell responses with clonal expansion of different melanoma-reactive T cell receptors in newly induced TLS [35]. In line with these data, B cell-expressed LTA1B2 controlled the organization of ectopic TLS at homeostasis and, following immune challenge, the activation of TLS for the generation of both adaptive T helper (T H ) 1 and T H 2 immune responses (reviewed in [36]).
We here demonstrate that several B cell subpopulations are a source of LTA in human melanoma and the decreased density of LTA + B cells is associated with primary tumor metastasis. Whether the expression of LTA in B cells promotes tumor cell cytotoxicity, inflammation and induction of functional TLS in human melanoma remains to be established; our initial data, however, point to the presence mainly of early immature but not fully mature TLS in primary human melanomas [25]. Additionally, the association of a decreased density of LTA + B cells with primary tumor metastasis needs to be evaluated for other human cancer types, particularly in light of the reported promotion of androgen-independent tumor growth in a murine prostate cancer model through B cell-derived LTA/B, which activated pro-tumorigenic IKK-alpha and STAT3 signaling in cancer cells [37].
Immunosuppressive IL-10 + B cells may accumulate at inflammatory or tumor sites through increased cell survival via the enhanced expression of transcription factor C-MAF upon restraint of SLAMF5 receptor-mediated signaling [38]. Through the secretion of IL-10, these B cells can attenuate T H 1, T H 2 or T H 17 cell-mediated immune responses in murine models of autoimmune, inflammatory and infectious diseases (reviewed in [36]) and inhibit anti-tumor cytotoxic T cell responses in mouse models of colorectal cancer and melanoma (reviewed in [39,40]). Reported mechanisms include the suppression of IFN-γ and IL-17 production in T H 1, NK and T H 17 cells, respectively; interference with differentiation of T cells into T H 17 cells and proliferation of CD4 + T cells; induction of regulatory T cells and IL-10 + suppressive T cells from effector T cells; suppression of monocyte proliferation and IL-12 production in dendritic cells (reviewed in [40][41][42]).
In parallel with studies in mouse models, IL-10 + B cell numbers have also been associated with human cancer progression. Examples are IL-10 + B cells in tumor tissues of gastric cancer [20], squamous cell carcinoma of the tongue [21] and ascites from ovarian cancer [22]; IgA + CD138 + PD-L1 + IL-10 + B cells in prostate cancer [23]; CD5 high CD24 −/+ CD27 high/+ CD38 dim PD-1 high B cells producing IL-10 in human hepatocellular carcinoma [9]; Granzyme B + CD38 + CD1d + IgM + CD147 + B cells expressing IL-10 and IDO in breast, cervical and ovarian carcinomas [19]. In contrast to these studies, our results now only show low numbers of IL-10 + B cells in human melanoma, numbers which were considerably lower than those of LTA + B cells. An association of IL-10 expression with a specific B cell subpopulation was not found, nor was an association of melanoma metastasis with an increased number of IL-10 + B cells. Given the poorly recognized but still substantial body of data on pro-inflammatory functions of IL-10 and other IL-10 cytokine family members (reviewed in [43,44]), our data support further evaluation of B cell-derived IL-10 in human melanoma.

Conclusions
Taken together, this study expands the knowledge of the spectrum of functional B cell subpopulations and their spatiotemporal dynamics in human melanoma. Interestingly, our data indicate that melanoma progression appears to be associated with changes in the number of LTA + rather than IL-10 + B cell numbers, suggesting a more relevant functional role for LTA + B cells. Moreover, the expression of immunoregulatory cytokines in multiple, rather than just one, B cell subpopulation raises the interesting question of the extent to which one B cell subpopulation could functionally replace the others during melanoma progression. The different dynamics of putative immunostimulatory and immunosuppressive B cell subpopulations during melanoma progression underscores the importance of this type of analysis for other cancer types.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/diagnostics11071238/s1, Figure S1: IL-10 + B cell subpopulations in human melanoma, Figure S2: The density (cells/mm 2 ) of LTA + and IL-10 + B cell subpopulations in primary human melanomas and their association with another four categorical prognostic clinicopathologic parameters, Table S1: Biopsy sites of melanoma metastases, Table S2: Comparison of relative frequencies of LTA + and IL-10 + B cell subpopulations in melanoma metastases to metastasized and non-metastasized primary tumors.

Informed Consent Statement:
The patients/participants provided their written informed consent to tumor sample collection.

Data Availability Statement:
The data presented in this study are available in this article and the supplementary material.