Atlas of breast cancer infiltrated B-lymphocytes revealed by paired single-cell RNA-sequencing and antigen receptor profiling

To gain mechanistic insights into the functions and developmental dynamics of tumor-infiltrated immune cells, especially B-lymphocytes, here we combine single-cell RNA-sequencing and antigen receptor lineage analysis to characterize a large number of triple-negative breast cancer infiltrated immune cells and report a comprehensive atlas of tumor-infiltrated B-lymphocytes. The single-cell transcriptional profiles reveal significant heterogeneity in tumor-infiltrated B-cell subgroups. The single-cell antigen receptor analyses demonstrate that compared with those in peripheral blood, tumor-infiltrated B-cells have more mature and memory B-cell characteristics, higher clonality, more class switching recombination and somatic hypermutations. Combined analyses suggest local differentiation of infiltrated memory B-cells within breast tumors. The B-cell signatures based on the single-cell RNA-sequencing results are significantly associated with improved survival in breast tumor patients. Functional analyses of tumor-infiltrated B-cell populations suggest that mechanistically, B-cell subgroups may contribute to immunosurveillance through various pathways. Further dissection of tumor-infiltrated B-cell populations will provide valuable clues for tumor immunotherapy.

Analysis of the single productive TCR rearrangements from 10,972 infiltrated T cells in TNBCs and 5,737 T cells from their corresponding PBMCs did not reveal significantly different VH and JH gene usage for majority of them (Supplementary Fig. 29a-b).
As shown in Supplementary Fig. 29c,

TNBC-infiltrated T cells contained higher percentages of T cell clones than
PBMC samples (70% vs 34%), which were also larger than the clones in PBMCs ( Supplementary Fig. 29c-d).
To demonstrate the intra-and inter-cluster distributions of T cell clones, the overlap of TCR clonotypes between each two clusters was shown in a heat map after normalized by their cell numbers ( Supplementary Fig. 30). In tumor tissues, we found that CD8 Exhausted (C3) had the most intra-cluster clonotype overlap, suggesting the highest clonality for this group ( Supplementary Fig. 30a). CD8 Exhausted (C3) and CD4/CD8 (C12) clusters had the most inter-cluster clonotype overlap. In PBMC, significantly less clonotype overlap was observed than in TNBC samples ( Supplementary Fig.   30b).

T cell clustering analysis for TNBC2-6 samples
After

Paired TCR and single-cell RNA-seq data analyses
To demonstrate the intra-and inter-cluster distributions of T cell clones, we counted the overlap of clonotypes between two T cell clusters. T cell clonotype was defined as cells with the same CDR3 nucleotide sequences for TCRα and TCRβ pair. For any two cells with the same clonotype, the overlap number between the two T cell clusters containing those two cells was added with 1.
Then there would be a sum number of overlaps for each two T cell clusters. To compare the levels of the overlap among T cell clusters with different cell numbers, the counts of overlap were normalized by the formula (2) = × × 10,000 in the heatmap (Supplementary Fig. 30a-b).
N normalized is the normalized value shown in Supplementary Fig. 30a     Both TNBC and PBMC samples contributed similarly to each cluster.

(b) The t-SNE projections of IGH non-switched B cells (left panel) and IGH class-switched B cells (right panel). IGH class-switched and non-switched B cells could be well separated
in the t-SNE plots.
(c) Percentages of cells from each sample for each B cell cluster. Cells from different samples contributed similarly to each B cell cluster (also see Supplementary Table 6). C12 germinal center B cells were observed mainly in tumor samples and in 4 of the 5 patients (also see Supplementary Table 6). These multiplex immunofluorescence stainings were conducted for TNBC1 and TNBC5.
Similar results were observed, and representative images from TNBC5 were showed.
These multiplex immunofluorescence stainings were conducted for TNBC1 and TNBC5.
Similar results were observed, and representative images from TNBC5 were showed.