Prognostic Value of B Cells in Cutaneous Melanoma

Genome Med 2019 AI 7 Explanations View Original
Original Paper (PDF)

Unable to display PDF. Download it here or view on PMC.

Plain-English Explanations
Pages 1-2
B Cells as Independent Prognostic Factors in Melanoma

The T cell paradigm and its gaps For decades, research on anti-tumor immunity in melanoma focused on CD8+ T cells. While intratumoral T cell density and function have established prognostic and predictive value in skin cutaneous melanoma (SKCM), T cell density alone is neither a universal prognostic factor nor a sole predictor of immunotherapy response. The role of tumor-infiltrating B cells remained poorly characterized.

What this study found Selitsky and colleagues used a bioinformatics tool (V'DJer) to reconstruct full-length B cell receptor (BCR) sequences from standard bulk RNA-seq data in 473 TCGA SKCM samples. They found that higher clonality of the BCR repertoire (lower species evenness) was a favorable prognostic factor independent of standard clinical variables - and that the absence of an assembled BCR in pre-treatment tumor samples was associated with non-response to anti-CTLA-4 immunotherapy.

Biological context The study focused on immunoglobulin heavy chain gamma (IGHG), the most abundant BCR isotype in SKCM, which is associated with activated, class-switched B cells capable of antibody-dependent cellular cytotoxicity. This suggests the B cell response in melanoma is antigen-driven and mature, rather than the naive responses associated with IGHM-expressing cells.

TL;DR: B cell receptor clonality measured from standard tumor RNA-seq data is an independent favorable prognostic factor in cutaneous melanoma, and absence of B cells is associated with failure to respond to anti-CTLA-4 immunotherapy.
Pages 2-3
BCR Repertoire Profiling from RNA-Seq Data

V'DJer bioinformatics tool The study's key methodological innovation was using V'DJer, a bioinformatics tool that assembles complete BCR VDJ sequences from short-read bulk RNA-seq data. Unlike previous approaches that could only analyze germline BCR regions, V'DJer reconstructs full variable-diversity-joining chain sequences, enabling characterization of clonotype diversity, somatic hypermutation, and BCR isotype distribution from standard sequencing data without additional experiments.

Diversity metrics Multiple BCR diversity metrics were calculated from assembled sequences: total BCR counts (abundance), Shannon entropy and Gini-Simpson (diversity indices capturing species richness and evenness), species evenness (Shannon entropy normalized by species richness), top clone proportion, and mean V-region identity (a surrogate for somatic hypermutation, reflecting antigen-driven B cell maturation).

Machine learning B cell subset classifiers Beyond BCR repertoire metrics, the study applied two machine learning classifiers trained on published gene expression datasets: the B Cell-Associated Gene Signature (BAGS) classifier for five B cell subtypes (naive, centrocytes, centroblasts, memory, plasmablasts) and an IL10-positive B regulatory cell classifier. These classifiers assessed the relative similarity of each bulk tumor sample to known B cell phenotypic states.

TL;DR: The study used V'DJer to reconstruct BCR sequences from 473 bulk tumor RNA-seq samples and combined this with machine learning-based B cell phenotype classifiers to characterize the B cell immune landscape of SKCM at unprecedented resolution.
Pages 4-5
Clonal BCR Restriction Predicts Improved Survival

Species evenness as the key prognostic metric Of all BCR repertoire measurements tested, species evenness was the most strongly and consistently prognostic. Lower species evenness indicates clonal restriction - a small number of dominant BCR clones rather than many equally abundant clones. Lower evenness (more clonal) was associated with significantly better overall survival (HR = 1.38, p = 0.0017 in univariable analysis; HR = 1.45, p = 0.0014 after conditioning on clinical variables including tumor tissue site, sex, age, and stage).

Total BCR abundance also matters Higher total BCR counts (greater B cell infiltration) were independently associated with better OS (HR = 0.72, p = 0.010), and this association persisted after multivariable adjustment (p = 0.010). This suggests that melanomas with both higher B cell infiltration and more clonal BCR repertoires represent a particularly favorable immune context.

Mutation burden is uncorrelated Tumor somatic mutation burden and predicted neoantigen burden showed no significant correlation with any BCR repertoire measurement. This finding suggests that the B cell response in melanoma is directed against antigens beyond tumor-specific neoantigens - possibly cancer testis antigens, melanocyte differentiation antigens, or infectious agents - and that B cell clonality captures immune activation that is independent of mutational landscape.

TL;DR: Greater clonal restriction of the BCR repertoire (lower species evenness) and higher total B cell infiltration are both independently prognostic for better overall survival in SKCM, even after controlling for standard clinical variables and independent of tumor mutation burden.
Pages 5-7
Memory B Cells Are Favorable; Regulatory B Cells Are Adverse

Memory B cell classification predicts survival Using the BAGS machine learning classifier, samples classified as memory B cell-predominant showed significantly better OS after accounting for age, sex, tissue site, and stage (p = 2 x 10-8). Memory B cells represent a mature, antigen-selected population capable of rapid antibody production upon re-encounter with antigen, suggesting an adaptive and durable anti-tumor immune response.

Regulatory B cell classification predicts worse survival Samples classified as IL10+ B regulatory cell-predominant showed significantly worse OS (p = 4 x 10-9). IL10-producing B cells can suppress immune responses by inhibiting T cell function. The study notes, however, that the IL10+ classification may be more broadly indicative of an immunosuppressive tumor microenvironment rather than specifically measuring B regulatory cells.

Unsupervised clustering identifies four immune archetypes Unsupervised hierarchical clustering of all BCR and TCR metrics identified four distinct tumor immune archetypes. Two clusters were characterized by high BCR and TCR abundance, high diversity, low evenness, and enrichment for memory B cells and the TCGA immune-high subtype - favorable prognostic profiles. The other two clusters showed high B regulatory cell scores or low overall immune cell infiltration and were associated with worse outcomes.

TL;DR: Memory B cell-predominant tumors have the best prognosis while IL10-positive regulatory B cell-predominant tumors have worse outcomes, identifying B cell phenotype as a prognostically informative axis that complements T cell characterization.
Pages 7-8
B Cell Absence Predicts Anti-CTLA-4 Non-Response

BCR absence and CTLA-4 non-response In patients treated with anti-CTLA-4 therapy, the absence of an assembled IGHG BCR sequence in pre-treatment tumor samples was significantly enriched in non-responders (p = 0.04, Fisher's exact test). This finding suggests that patients without detectable intratumoral B cells are less likely to benefit from CTLA-4 checkpoint blockade - possibly because B cells contribute to the anti-tumor immune cascade that CTLA-4 inhibition aims to unleash.

IL10+ B cell classification and CTLA-4 non-response Non-responders to anti-CTLA-4 were also enriched for the IL10+ B regulatory cell classification (p = 0.03), consistent with the hypothesis that an immunosuppressive B cell microenvironment, rather than absence of B cells, can similarly blunt checkpoint blockade responses.

No association with anti-PD-1 response Neither BCR absence nor IL10+ B cell classification was associated with anti-PD-1 response. This differential association suggests that B cells contribute to the immune responses unlocked by CTLA-4 blockade through a mechanism distinct from those governing PD-1 response, potentially reflecting the different stage of T cell activation at which each checkpoint operates.

TL;DR: Absence of B cells in pre-treatment tumor samples predicts failure to respond to anti-CTLA-4 but not anti-PD-1 therapy, suggesting that B cell infiltration is a modality-specific predictive biomarker for checkpoint blockade.
Pages 6-7
B and T Cell Responses Are Coordinated in Melanoma

BCR and TCR repertoire metrics are correlated BCR abundance and diversity measures were significantly correlated with the same TCR measures (Spearman rho 0.47-0.55, p < 4 x 10-16), and both were independently associated with overall survival after multivariable adjustment. This concordant activation of B and T cell repertoires suggests that the adaptive immune response in melanoma is broadly coordinated rather than operating through independent mechanisms.

Site-specific differences in BCR diversity B cell diversity was significantly lower in primary tumor and regional subcutaneous metastatic samples compared to other metastatic sites, while somatic hypermutation was lower in primary tumors. This may reflect different maturation states of the B cell response at different anatomic locations and stages of disease progression.

Sex differences in BCR diversity Female melanoma patients had significantly higher BCR diversity (Shannon entropy p = 0.02, Gini-Simpson p = 0.01) and lower top clone proportion (p = 0.01) than males, suggesting sex-based differences in the B cell immune response to melanoma. The mechanism underlying this difference and its clinical significance require further investigation.

TL;DR: BCR and TCR repertoires are coordinately activated in SKCM, and both contribute independent prognostic information, supporting the view that effective anti-tumor immunity in melanoma requires integrated B cell and T cell responses.
Pages 8-9
Limitations and Opportunities for B Cell Biomarker Development

Bulk RNA-seq limitations V'DJer reconstructs BCR sequences from bulk tumor RNA-seq, which cannot link heavy and light chain sequences to individual B cells, resolve BCR structure at single-cell resolution, or distinguish intratumoral B cell subpopulations at the spatial level. Single-cell BCR sequencing and spatial transcriptomics would provide more precise characterization of tumor-infiltrating B cell states and locations.

Immunotherapy cohort size constraints The immunotherapy response analyses were limited by small cohort sizes (approximately 30-60 patients per treatment arm). These sample sizes were insufficient to test some associations (such as TCR absence and non-response) and may have been underpowered to detect weaker associations between BCR metrics and anti-PD-1 response.

Future directions The study identifies several promising research priorities: prospective validation of BCR evenness and total B cell counts as prognostic biomarkers in larger SKCM cohorts; testing whether BCR absence can serve as a clinical criterion for excluding patients unlikely to benefit from CTLA-4 therapy; investigating the tumor antigens driving B cell clonal expansion in melanoma; and characterizing the spatial distribution and function of B cell subsets within tumor tertiary lymphoid structures, which are increasingly recognized as important for anti-tumor immunity.

TL;DR: While bulk RNA-seq limits single-cell resolution and immunotherapy cohorts were small, this study establishes BCR clonality and B cell phenotype as promising prognostic and predictive biomarkers in SKCM that warrant prospective validation.
Citation: Open Access, 2019. Available at: PMC6540526.