IgA transcytosis and antigen recognition govern ovarian cancer immunity

Abstract

Most ovarian cancers are infiltrated by prognostically relevant activated T cells^1-3, yet exhibit low response rates to immune checkpoint inhibitors⁴. Memory B cell and plasma cell infiltrates have previously been associated with better outcomes in ovarian cancer^5,6, but the nature and functional relevance of these responses are controversial. Here, using 3 independent cohorts that in total comprise 534 patients with high-grade serous ovarian cancer, we show that robust, protective humoral responses are dominated by the production of polyclonal IgA, which binds to polymeric IgA receptors that are universally expressed on ovarian cancer cells. Notably, tumour B-cell-derived IgA redirects myeloid cells against extracellular oncogenic drivers, which causes tumour cell death. In addition, IgA transcytosis through malignant epithelial cells elicits transcriptional changes that antagonize the RAS pathway and sensitize tumour cells to cytolytic killing by T cells, which also contributes to hindering malignant progression. Thus, tumour-antigen-specific and -antigen-independent IgA responses antagonize the growth of ovarian cancer by governing coordinated tumour cell, T cell and B cell responses. These findings provide a platform for identifying targets that are spontaneously recognized by intratumoural B-cell-derived antibodies, and suggest that immunotherapies that augment B cell responses may be more effective than approaches that focus on T cells, particularly for malignancies that are resistant to checkpoint inhibitors.

Fig. 1. IgA–pIgR colocalization is associated with protective immunity in human ovarian cancer.

a, Left, percentage of FACS cell counts of IgA⁺, IgG⁺ or IgM⁺ cells among Ig⁺ B cells or plasmablasts or plasma cells, normalized to 10,000 viable CD45⁺ cells. B cells, CD45⁺CD3⁻CD19⁺CD20⁺ cells; plasmablasts, CD45⁺CD3⁻CD19⁺CD20⁻CD38^high cells; plasma cells, CD45⁺CD3⁻CD19⁺CD20⁻CD138⁺ and CD45⁺CD3⁻CD19⁻CD20⁻CD138⁺ cells. Each dot represents one tumour (n = 29). Details of box plots can be found in Methods. P values were obtained by a two-way analysis of variance (ANOVA) followed by Dunnett’s test for multiple comparisons. Supplementary Table 1 provides further details on statistics. Right, bar graphs representing the percentage of each isotype produced by plasma cells (top) or B cells (bottom) in the same tumours, normalized to 10,000 viable CD45⁺ cells. IC, intracellular. b, IgA-coated CD45⁻EpCAM⁺ tumour epithelial cells (mean ± s.e.m., n = 10) in dissociated HGSOC. c, Expression of pIgR protein in independent HGSOC (n = 27); tumour-free Fallopian tube (n = 3), ovary (n = 5) and omental (n = 4) samples; ovarian tumour cell lines; and K562 leukaemia cells and THP1 monocyte cells (negative controls). Positive control, recombinant human pIgR. Western blots were repeated twice. NSCLC, non-small-cell lung cancer. d, Histograms showing FACS analysis of pIgR, in ovarian surface epithelial (OSE), K562, THP1, wild-type or PIGR-ablated (pIgR^CRISPR) OVCAR3 cells. e, Left, representative (n = 273) combined staining of IgA, pIgR, IgG, PCK and DAPI. Instances with IgA–pIgR colocalization are indicated with arrows. Scale bar, 50 μm (top left), 20 μm (all other panels). Top right, representative (n = 137, IgA–pIgR colocalization ≥ median) dot plot showing IgA–pIgR colocalized signal among DAPI⁺PCK⁺ cells. Bottom right, scattered graph showing number of IgA–pIgR colocalized cells (averaged from duplicated cores) per mm² of PCK⁺ (mean ± s.e.m., n = 273). f, Increased numbers of cells with IgA–pIgR colocalization per PCK⁺ tumour islet area (averaged from duplicated cores) are associated with improved outcome (threshold, median; P = 0.0116, H. Lee Moffett Cancer Centre cohort (MCC) (right); P = 0.0002, New England Case–Control study cohort (NECC) (left)). g, Density of IgA-coated cells (averaged from duplicated cores) in tumour islets (cells per mm² PCK⁺ area) is associated with improved outcome (P = 0.0110 for MCC (right) and P = 0.0054 for NECC (left) cohorts). *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, two-sided log-rank (Mantel–Cox) test. Source data

Fig. 2. Transcytosis of IgA through pIgR⁺ ovarian cancer cells impairs tumour growth and augments cytotoxic killing mediated by T cells.

a, Left, images of APC-labelled IgA binding and internalization in pIgR⁺ OVCAR3 cells (repeated three times). Scale bar, 50 μm (main panels), 10 μm (magnified regions). Right, comparison of antibody internalization signal (mean ± s.e.m.) in different treatment conditions and at different temporal points. Each dot represents quantification from one cell. ***P ≤ 0.001, unpaired two-tailed t-test. Supplementary Table 2 provides details of statistics. b, OVCAR3 cells were incubated with control IgA or IgG for 8 h in the presence of wortmannin, brefeldin A (BFA) or vehicle, and supernatants were subjected to liquid chromatography with tandem mass spectrometry (LC–MS/MS). Heat map of all peptides of the extracellular domain of pIgR (n = 3); scale represents log₂-transformed intensities of pIgR peptide fragments detected in LC–MS/MS. c, Left, co-immunoprecipitates of supernatants from IgA-treated pIgR⁺ or PIGR-ablated OVCAR3 cells (with and without brefeldin A or wortmannin) blotted for the secretory component of pIgR and IgA (input control). Right, LC–MS/MS analysis of the co-immunoprecipitates showing intensities (log₂-transformed) of the secretory component of pIgR and IgA (n = 2). WT, wild type; CR, PIGR-ablated. d, Pre-ranked gene-set enrichment analysis (GSEA), showing the top upregulated gene sets in OVCAR3 cells treated with irrelevant IgA compared to IgG or untreated cells (n = 3), Kolmogorov–Smirnov test. GO, Gene Ontology. e, Progressive increase in DUSP5 and concomitant reduction in phospho-ERK1 and phospho-ERK2 (pERK1/2) after IgA treatment (left) of OVCAR3 cells, but not IgG treatment (right). Experiments were repeated three times. tERK1/2, total ERK1 and total ERK2. f, Left, dose-dependent cytotoxic killing of NY-ESO-1-transduced OVCAR3 cells (n = 3) by NY-ESO-1–TCR-transduced T cells is augmented by co-incubation with IgA, compared to IgG or PBS. Right, IgA treatment also augmented the anti-tumour activity of FSH-targeted chimeric receptor T cells. Mean ± s.e.m. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ordinary one-way ANOVA. Supplementary Table 2 provides details of statistics. g, Cytotoxic killing of primary CD45⁻EpCAM⁺ tumour cells (n = 3) by autologous tumour-infiltrating T cells (1:1 ratio) is augmented by co-incubation with autologous (P < 0.0001) or irrelevant IgA (P = 0.0031), but not autologous IgG (P = 0.1951). Mean ± s.e.m. **P ≤ 0.01, ***P ≤ 0.001, NS, not significant; unpaired two-tailed t-test. Source data

Fig. 3. Tumour-antigen-specific IgA produced in the ovarian cancer microenvironment antagonizes ovarian cancer progression.

a, Schematic of design of experiment shown in b, d, f. Antibody at 100 μg per 20 g body weight or equal volume of vehicle (PBS) was intratumourally injected. KO, knockout. b, Autologous tumour growth curves (left), weight (centre) and volume (right) in tumour-bearing RAG1-knockout mice receiving control (irrelevant IgA (iIgA)) or recombinant dimeric IgA antibodies (labelled Ab1, Ab2 or Ab3) produced with three different matching IgA sequences clonally expanded in two different HGSOC. Respective autologous HGSOC cells were used (tumour no. 1 for Ab1 and tumour no. 2 for Ab2 or Ab3). Supplementary Table 3 provides details of statistics. c, IgA purified from TSPAN7- and BDNF-reactive immortalized B cells recognizes the corresponding recombinant proteins in western blot analysis, along with endogenous TSPAN7 and BDNF expressed in OVCAR3 cells. HEK293T, THP1 and K562 cells were included as negative controls. Experiments were repeated three times. rhTSPAN7, recombinant human TSPAN7; rhBDNF, recombinant human BDNF. d, Tumour growth curves (left), weight (centre) and tumour volume (right) in tumour-bearing RAG1-knockout mice receiving control or tumour-derived antibodies. Supplementary Table 4 provides details of statistics. e, Representative images (n = 10 per group from 2 independent experiments) of central necrosis in tumours from mice receiving IgA from tumour-derived B cells. Scale bars, 4 mm. f, Antibodies used in d were digested with pepsin to remove their Fc domain and resulting F(ab′)₂ fragments used to treat OVCAR3 tumour-bearing RAG1-knockout mice under identical conditions. Tumour growth curves (left), weight (centre) and tumour volume (right). Supplementary Table 5 provides details of statistics. In b, d, f, growth curves and tumour weights were pooled from 2 independent experiments (n = 10 mice per group in total). Data are mean ± s.e.m. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, NS, not significant; paired two-tailed t-test for growth curves or unpaired two-tailed t-test for tumour weights. Source data

Fig. 4. Antigen-specific IgA redirects Fcα/μR-positive myeloid cells against cell-surface antigen-positive tumour cells.

a, Tumour growth curves (left), tumour weight (centre) and volume (right) in tumour-bearing NSG mice receiving control or tumour-derived antibodies. Supplementary Table 6 provides for details of statistics. b, Tumour growth curves (left), tumour weight (centre) and volume (right) in tumour-bearing RAG1-knockout mice receiving control or tumour-derived antibodies ± intraperitoneal injections of anti-NK1.1 or control antibodies. Supplementary Table 7 provides for details of statistics. In a, b, growth curves and tumour weights were pooled from 2 independent experiments (n = 10 mice per group in total). c, Binding of IgA antibodies to splenic CD11b⁺ cells from tumour-bearing RAG1-knockout mice (n = 10), after incubation with Fcα/μR (CD351)-neutralizing antibodies or isotype controls. P < 0.0001. d, Cytotoxic killing of OVCAR3 cells by splenic myeloid cells from tumour-bearing RAG1-knockout mice (1:1 ratio) is augmented by coating the tumour cells with anti-TSPAN7, and inhibited by neutralizing CD351 (anti-CD351^neut) (n = 3). Supplementary Table 8 provides details of statistics. OVCAR3^Luci, OVCAR3 cells transduced with luciferase-expressing vector. e, Increased accumulation of CD351⁺ myeloid cells in xenografts in RAG1-knockout mice treated with intratumoural anti-TSPAN7, compared to irrelevant IgA or vehicle, irrespectively of NK1.1 depletion (n = 5). Supplementary Table 8 provides for details of statistics. f, Tumour growth curves (left), tumour weight (centre) and volume (right) in wild-type pIgR⁺ (WT) or PIGR-ablated (pIgR^CRISPR (CR)) OVCAR3 tumour-bearing RAG1-knockout mice receiving control or tumour-derived antibodies. Two independent experiments were performed with similar results; tumour growth was represented from 1 experiment (n = 5 mice per group); tumour weights were pooled from 2 experiments (n = 10 mice per group in total). Supplementary Table 9 provides details of statistics. Data are mean ± s.e.m. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, NS, not significant; paired two-tailed t-test for growth curves or unpaired two-tailed t-test for tumour weights and other experiments. Source data

Extended Data Fig. 1. Isotype-switched humoral responses are associated with better outcome and denser T cell infiltration in HGSOC.

a, b, Overall survival associated with the presence of CD19⁺ B cells within the total tumour area (P = 0.04) (a) or specifically in the PCK⁺ epithelial tumour islets (P = 0.0085) (b), in HGSOC as assessed by multiplex immunohistochemistry of TMAs corresponding to 534 patients with HGSOC combined from the NECC (n = 180), NHS (n = 261) and MCC (n = 93). B cell infiltration is defined as the presence of CD19⁺ cells on any of the duplicate sections analysed for each tumour. c, Survival outcome associated with the expression of CD19 in 428 annotated HGSOCs in TCGA datasets (P = 0.0091). *P ≤ 0.05, **P ≤ 0.01, two-sided log-rank (Mantel–Cox) test. d, Representative staining of the association between the accumulations of T and B cells at tumour beds (n = 534). Scale bar, 200 μm or 100 μm, as indicated. e, Accumulation of CD8⁺ (top) and CD4⁺ (bottom) T cells in the PCK⁺ tumour islets are associated with the presence of B cells. CD8: with versus without B cell infiltration, P < 0.0001; CD4: with versus without B cell infiltration, P < 0.0001. Data are mean ± s.e.m. ***P ≤ 0.001, unpaired two-tailed t-test. f, Overall survival associated with the presence of intra-epithelial CD19⁺ B cells and total intra-epithelial CD3⁺ T cells (left) or intra-epithelial CD3⁺CD8⁺ T cells (right) within ovarian carcinomas, as assessed by multiplex immunohistochemistry of TMAs corresponding to 534 patients with HGSOC combined from the NECC (n = 180), NHS (n = 261) and MCC (n = 93). With intra-epithelial CD3 T cells and with- intra-epithelial B cells versus with intra-epithelial CD3 T cells and no intra-epithelial B cells, P = 0.0123; With intra-epithelial CD3 T cells and no intra-epithelial B cells versus no intra-epithelial CD3 T cells and no intra-epithelial B cells, P = 0.0572; with intra-epithelial CD8 T cells and with intra-epithelial B cells versus with intra-epithelial CD8 T cells and no intra-epithelial B cells, P = 0.0035; with intra-epithelial CD8 T cells and no intra-epithelial B cells versus no intra-epithelial CD8 T cells and no intra-epithelial B cells, P = 0.4113. B cell and T cell infiltration are defined as the presence of CD19⁺ or CD3⁺ cells on any of the duplicate sections analysed for each tumour. *P ≤ 0.05, **P ≤ 0.01, NS, not significant, two-sided log-rank (Mantel–Cox) test. g, Representative FACS analysis of immunoglobulin isotypes on the surface of B cells infiltrating freshly dissociated human HGSOCs (n = 29). h, Bar graphs representing tumour-wise (n = 29) FACS analysis comparison of percentages of each Ig⁺ cells among total Ig⁺ B cells (CD45⁺CD3⁻CD19⁺CD20⁺), plasma cells (intracellular in CD45⁺CD3⁻CD19⁺CD20⁻CD138⁺ and CD45⁺CD3⁻CD19⁻CD20⁻CD138⁺) and plasmablasts (intracellular in CD45⁺CD3⁻CD19⁺CD20^-CD38^high), normalized to 10,000 viable CD45⁺ cells. i, FACS dot plots from analysis of IgA, IgG and IgM antibodies, and respective isotype controls, in human peripheral blood mononuclear cells, to evaluate the fidelity of the antibodies used. Source data

Extended Data Fig. 2. IgA coating of tumour cells is associated with a better outcome in HGSOC.

a, Bar graphs representing tumour-wise FACS analysis (n = 10) comparison of percentages of each Ig⁺ cells among total Ig⁺ plasmablasts (intracellular in CD45⁺CD3⁻CD19⁺CD20⁻CD38^high) and CD27⁺ plasmablasts (intracellular in CD45⁺CD3⁻CD19⁺CD20⁻CD38^highCD27⁺), normalized to 10,000 viable CD45⁺ cells. b, Relative abundances of IgH chains on the basis of TCGA transcriptional analyses (n = 430) for each immunoglobulin heavy chain gene. Abundances are shown in log₂-transformed reads per kilobase of transcripts (RPKM) values, which corrects for both gene length and sequencing depth. Details of box plots are in Methods. c, Overall survival associated with the presence of CD19⁺CD138⁺ plasma cells within the total tumour area (P = 0.0285) (left) or specifically in the PCK⁺ epithelial tumour islets (P = 0.0053) (right), in HGSOC as assessed by multiplex immunohistochemistry of TMAs corresponding to 534 patients with HGSOC combined from the NECC (n = 180), NHS (n = 261) and MCC (n = 93). Plasma cell infiltration is defined as the presence of CD19⁺CD138⁺ cells on any of the duplicate sections analysed for each tumour. *P ≤ 0.05, **P ≤ 0.01, two-sided log-rank (Mantel–Cox) test. d, FACS analysis showing number (log-transformed) of plasma cells (CD45⁺CD3⁻CD19⁺CD20⁻CD138⁺ and CD45⁺CD3⁻CD19⁻CD20⁻CD138⁺), plasmablasts (CD45⁺CD3⁻CD19⁺CD20⁻CD38^high), B cells (CD45⁺CD3⁻CD19⁺CD20⁺), T cells (CD45⁺CD3⁺) and other leukocytes (CD45⁺CD3⁻CD19⁻CD20⁻CD138⁻) in HGSOC (n = 29). The data are normalized to 10,000 viable CD45⁺ leukocytes. Data are mean ± s.e.m. e, Graphs showing correlations between log count of T cells and plasma cells (left) (Pearson correlation coefficient (r) = 0.5049; two-sided nominal P = 0.0052); and between log count of T cells and plasmablasts (right) (Pearson correlation coefficient (r) = 0.4755; two-sided nominal P = 0.0091). All three cell types represent absolute counts normalized to 10,000 CD45⁺ leukocytes. f, Colocalization of IgA with pIgR⁺ cells (IgA–pIgR co-localization ≥ median) in the PCK⁺ tumour islets is associated with an improved outcome in HGSOC, compared to only pIgR^high samples (≥median) without IgA colocalization (colocalization < median), in MCC (P = 0.0060) and NECC (P = 0.0044) cohorts. **P ≤ 0.01, two-sided log-rank (Mantel–Cox) test. g, Higher number of IgA-coated cells in the PCK⁻ stromal area (average from duplicated cores) is not associated with an improved outcome in HGSOC, analysed using median IgA-coating threshold in MCC (P = 0.8954) and NECC (P = 0.0537) cohorts. NS, not significant; two-sided log-rank (Mantel–Cox) test, no multiple comparison adjustment. h, A higher number of IgG-coated cells in the PCK⁺ tumour islets (average from duplicated cores) is not associated with an improved outcome in HGSOC, analysed using median IgG-coating threshold in MCC (P = 0.6350) and NECC (P = 0.0731) cohorts. NS, not significant; two-sided log-rank (Mantel–Cox) test. Source data

Extended Data Fig. 3. IgA internalization by HGSOC cells is associated with stronger T cell responses, and is dependent on pIgR–Fc interactions.

a, Representative (n = 273) IgA and IgG staining in tumours with a high or low density of CD4⁺ and CD8⁺ T cells. Scale bar, 200 μm or 100 μm, as indicated. CD4⁺ and CD8⁺ T cell accumulation (≥median) is associated with the density of IgA-coated tumour (PCK⁺) cells (IgA: CD4 < median versus CD4 ≥ median, P = 0.0209; IgA: CD8 < median versus CD8 ≥ median, P = 0.0087); but not associated with IgG-coated tumour (PCK⁺) cells (IgG: CD4 < median versus CD4 ≥ median, P = 0.4451; IgG: CD8 < median versus CD8 ≥ median, P = 0.8304). Data are mean ± s.e.m. *P ≤ 0.05, **P ≤ 0.01, NS, not significant; unpaired two-tailed t-test. b, Representative confocal microscopy images of fluorescently (APC) labelled whole or pepsinized irrelevant IgA- or IgG-binding and internalization in pIgR⁺ (WT) or PIGR-ablated (PIGR^CRISPR) OVCAR3 cells after 0 min, 15 min, 1 h or 8 h of incubation. Experiments were repeated three times. Scale bar, 50 μm or 10 μm, as indicated. c, FACS dot plots showing electroporation efficiency in pIgR-CRSIPR-guide (centre) or control-guide electroporated cells (right), compared to nonelectroporated OVCAR3 or primary HGSOC cells (n = 2) (left). Experiments were repeated three times. Western blots confirmed PIGR ablation in OVCAR3 and primary HGSOC cells (n = 2, repeated twice). THP1 cells were used as a negative control and recombinant pIgR (rPIGR) was used as a positive control. WT, wild type (non-electroporated cells). Source data

Extended Data Fig. 4. IgA transcytosis through HGSOC cells has substantial anti-tumour effects and sensitizes tumour cells to cytolytic killing by T cells.

a, Immunoblots showing pIgR co-immunoprecipitation with IgA, using nondenaturing lysates from two HGSOCs, CD45⁺ and CD45⁻ cells isolated from human ovarian cancer ascites, ascitic fluid and OVCAR3 cells (negative control). Inputs were immunoblotted using 1% of the amount of lysate used for the co-immunoprecipitation. The experiments were repeated three times. b, OVCAR3 cells were incubated with 0.5 μg ml⁻¹ of control IgA or IgG for 8 h in serum-free medium in the presence of wortmannin (1 μM), brefeldin A (1 μg ml⁻¹) or vehicle, and supernatants were subjected to mass spectrometry. The amino acids 62–77 fragment of pIgR was found only after incubation with IgA (repeated three times). c–e, OVCAR4 (c), OVCAR5 (d) and primary HGSOC (e) tumour cells were incubated with 0.5 μg ml⁻¹ of irrelevant IgA or IgG for 8 h in serum-free medium in the presence of wortmannin (1 μM), brefeldin A (1 μg ml⁻¹) or vehicle, and supernatants were then subjected to mass spectrometry (left). Right, heat map of all peptides of the extracellular domain of pIgR (n = 3). BFA, brefeldin A. WM, wortmannin. f, g, GSEA enrichment plots (h) and heat maps (g) using normalized gene expression from RNA-seq analysis from OVCAR3 cells with irrelevant IgA (0.5 μg ml⁻¹), IgG (0.5 μg ml⁻¹) or no treatment for 24 h (n = 3 experiments). h, No change in the protein levels of DUSP5, total ERK1/2 or phospho-ERK1/2 after vehicle (PBS) treatment of pIgR⁺ (WT) OVACR3 cells (left) or after IgA treatment of PIGR-ablated OVACR3 cells (right), incubated up to 8 h. The experiments were repeated three times. i, Dose-dependent cytotoxic killing of OVCAR3 cells by FSH-targeted chimeric receptor T cells is augmented by co-incubation with 0.5 μg ml⁻¹ of irrelevant IgA, anti-TSPAN7–IgA or anti-BDNF–IgA compared to IgG, pepsinized irrelevant IgA or PBS. n = 3 per group. **P ≤ 0.01, unpaired two-tailed t-test. j, Cytotoxic killing of autologous CD45⁻EpCAM⁺ tumour cells (with corresponding decrease of annexin V⁻propidium iodide (PI)⁻ viable cells) by autologous T cells (added at 1:1 ratio) is augmented by co-incubation with 0.5 μg ml⁻¹ of autologous IgA or irrelevant IgA but not with autologous IgG, pepsinized autologous or irrelevant IgA, as compared to uncoated cells (n = 3). Annexin V⁺: irrelevant IgA versus pepsinized irrelevant IgA, P < 0.0001; autologous IgA versus pepsinized autologous IgA, P < 0.0001; uncoated versus pepsinized irrelevant IgA, P = 0.3769; uncoated versus pepsinized autologous IgA, P = 0.2208. Annexin V⁻PI⁻ cells: irrelevant IgA versus pepsinized irrelevant IgA, P < 0.0001; autologous IgA versus pepsinized autologous IgA, P < 0.0001; uncoated versus pepsinized irrelevant IgA, P = 0.3329; uncoated versus pepsinized autologous IgA, P = 0.1916. Data are mean ± s.e.m. *** P ≤ 0.001, NS, not significant; unpaired two-tailed t-test. k, Cytotoxic killing of pIgR⁺ OVCAR3 cells, but not pIgR^CRISPR OVCAR3 cells, by FSH-targeted chimeric receptor T cells (added at 1:1 ratio) is augmented by co-incubation with 0.5 μg ml⁻¹ of irrelevant IgA compared to IgG-coated or uncoated cells (n = 6). Wild-type OVCAR3: PBS or irrelevant IgG versus irrelevant IgA, P < 0.0001. pIgR^CRISPR OVCAR3: PBS versus irrelevant IgA, P = 0.7728; irrelevant IgG versus irrelevant IgA, P = 0.9176. ***P ≤ 0.001, NS, not significant; unpaired two-tailed t-test. Source data

Extended Data Fig. 5. TSPAN7- and BDNF-specific antibodies produced at tumour beds delay the progression of established tumours.

a, RAG1-deficient mice inoculated subcutaneously with 10⁷ OVCAR3 cells received 100 μg per 20g body weight of irrelevant (i)IgA or irrelevant (i)IgG peritumourally at days 7, 11, 15, 19 and 23 after tumour inoculation. Tumour growth curves (left), tumour weight (centre) and representative differences in tumour volume (right) are shown. Growth curve statistics: iIgG versus iIgA, P = 0.0130, paired two-tailed t-test. Tumour weight statistics, iIgG versus iIgA, P = 0.043, unpaired two-tailed t-test. Data are mean ± s.e.m. *P ≤ 0.05. b, Tumour growth curves (left), as well as tumour volume (right) and weight (centre) in OVCAR3-tumour-bearing RAG1-deficient mice receiving irrelevant IgG antibodies or vehicle (PBS). Growth curve statistics: iIgG versus PBS, P = 0.1840, paired two-tailed t-test. Tumour weight statistics: iIgG versus PBS, P = 0.8275, (unpaired two-tailed t-test). Data are mean ± s.e.m. NS, not significant. c, Tumour growth curves (left), as well as tumour volume (right) and weight (centre) in OVCAR3-tumour-bearing RAG1-deficient mice receiving full-length or pepsinized (Fc-removed) irrelevant IgG or irrelevant IgA antibodies. Curves and tumour weights were pooled from 2 independent experiments (10 mice per group in total). Growth curve statistics: iIgA versus F(ab′)₂–iIgA, P = 0.0030; iIgA versus iIgG, P = 0.0578; iIgG versus F(ab′)₂–iIgG, P = 0.0547, paired two-tailed t-test. Tumour weight statistics: iIgA versus F(ab′)₂–iIgA, P < 0.0001; iIgG versus F(ab′)₂–iIgG, P = 0.5585; iIgG versus F(ab′)₂–iIgA, P = 0.5382, unpaired two-tailed t-test. Data are mean ± s.e.m. **P ≤ 0.01, *** P ≤ 0.001, NS, not significant. In a–c, growth curves and tumour weights were pooled from 2 independent experiments (n = 10 mice per group in total). d, Schematic of the optimized protocol for separating, immortalizing, characterizing and selecting tumour-reactive B cells from HGSOCs. e, Tetramers spanning the indicated loop in BDNF or the extracellular domain of TSPAN7 were used to sort reactive B cells immortalized from ten independent HGSOCs. f, The reactivity of expanded cells was confirmed using the same tetramers. g, IgA represents the majority of TSPAN7- or BDNF-reactive B cells sorted from HGSOCs. h, Representative TUNEL (Alexa Fluor 647) staining images in xenograft tumours developed in RAG1-knockout mice. Scale bar, 200 μm. i, Estimation of TUNEL⁺ cells normalized to tumour area. iIgG versus anti-TSPAN7, P = 0.0293. Data are mean ± s.e.m. *P ≤ 0.05, unpaired two-tailed t-test. j, Tumour area quantification. iIgG versus iIgA, P = 0.0146; iIgG versus anti-BDNF, P = 0.0002; iIgG versus anti-TSPAN7, P = 0.0002. Data are mean ± s.e.m. *P ≤ 0.05, ***P ≤ 0.001, unpaired two-tailed t-test. k, Quantification of irrelevant IgA and anti-TSPAN7–IgA antibody uptake in OVACR3 xenografts. P = 0.0343. Data are mean ± s.e.m. *P ≤ 0.05, unpaired two-tailed t-test. In h–k, n = 10 mice per group pooled from 2 independent experiments. Source data

Extended Data Fig. 6. Tumour-derived IgA abrogates tumour growth through antigen-dependent redirection of Fcα/μR⁺ myeloid cells and antigen-independent, pIgR-mediated transcytosis.

a, Dot plots showing FACS analysis of splenocytes for NK1.1 depletion in RAG1-KO mice (left). Scatter plot showing CD45⁺NK1.1⁺ cells percentages among viable splenocytes in respective treatment-group mice (n = 10 mice per group pooled from 2 independent experiments) (right). P < 0.0001, ordinary one-way ANOVA. Data are mean ± s.e.m. ***P ≤ 0.001. b, Representative FACS dot plots showing binding of IgA antibodies to splenic CD11b⁺ cells from tumour-bearing RAG1-deficient mice, after incubation with Fcα/μR (CD351)-neutralizing antibodies or isotype controls (n = 10). c, Representative FACS dot plots show CD351⁺ cells among viable CD45⁺CD11b⁺ cells in xenografts in RAG1-deficient mice treated with intratumoural anti-TSPAN7 (with or without NK1.1-depletion), compared to irrelevant IgA or vehicle (PBS) (n = 10 mice per group pooled from 2 independent experiments). d, Tumour growth curves (left), as well as tumour volume (right) and weight (centre) in pIgR^CRISPR OVCAR3-tumour-bearing NSG mice receiving irrelevant IgA antibodies or vehicle (PBS). Paired or unpaired two-tailed t-tests for tumour growth curves or tumour weight comparisons, respectively. Data are mean ± s.e.m. NS, not significant. e, Internalized intensity of antibodies (APC) were quantified and scattered bar graph showing comparison of antibody internalization in different treatment conditions, in which each dot represents quantification from one cell, pooled from three independent experiments. Supplementary Table 11 provides details of statistics. Data are mean ± s.e.m. ***P ≤ 0.001, NS, not significant; unpaired two-tailed t-test. f, Pathway analysis of RNA-seq from OVCAR3 cells treated with irrelevant IgA (0.5 μg ml⁻¹), anti-TSPAN7–IgA (0.5 μg ml⁻¹), anti-BDNF–IgA or no treatment for 24 h (n = 3). g, Autologous tumour growth curves, tumour weight and volume in wild-type pIgR⁺ (WT) or PIGR-ablated (pIgR^CRISPR) autologous tumour-bearing RAG1-deficient mice receiving control or recombinant dimeric IgA antibodies (Ab1, Ab2 or Ab3) produced with three different matching IgA sequences clonally expanded in two different HGSOCs. Respective autologous HGSOC cells were used (tumour no. 1 for Ab1 and tumour no. 2 for Ab2 or Ab3). Data are mean ± s.e.m. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, NS, not significant. Supplementary Table 12 provides details of statistics. In d, g, curves and tumour weights were pooled from 2 independent experiments (n = 10 mice per group in total). Source data

Extended Data Fig. 7. Multiplex immunohistochemistry staining optimization for CD3, CD4, CD8, CD19, CD20, CD138, IgA, IgG, pIgR and PCK in tonsil tissues.

a, b, Multiplex staining of tonsil tissue sections (n = 5) with anti-human CD3, CD4, CD8, CD19, CD20 and CD138 antibodies or respective isotype controls (a), and IgA, IgG, pIgR and PCK antibodies or respective isotype controls (b). Scale bar, 100 μm.

Extended Data Fig. 8. Multiplex immunohistochemistry staining optimization for CD3, CD4, CD8, CD19, CD20, CD138, IgA, IgG, pIgR and PCK in glioblastoma tissues, and for pIgR in kidney tissues.

a, b, Multiplex staining of glioblastoma tissue sections (n = 5) with anti-human CD3, CD4, CD8, CD19, CD20 and CD138 antibodies (a), and IgA, IgG, pIgR and PCK antibodies (b). Scale bar, 100 μm. c, Multiplex staining of healthy kidney tissue sections (n = 4) with anti-human pIgR antibody and respective isotype control. Scale bar, 100 μm.

Extended Data Fig. 9. Flow cytometry gating strategies.

a, b, Gating strategies for FACS analysis shown in Fig. 1a, Extended Data Figs. 1g, h, i, 2a, d, e. c, Gating strategies for FACS analysis shown in Fig. 1b. d, Gating strategies for FACS analysis shown in Fig. 1d. e, Gating strategies for FACS analysis shown in Fig. 1g, Extended Data Fig. 4j, k.

Extended Data Fig. 10. Further flow cytometry gating strategies.

a, Gating strategies for FACS analysis shown in Extended Data Fig. 2a. b, Gating strategies for FACS analysis shown in Extended Data Fig. 3c. c, Gating strategies for FACS analysis shown in Extended Data Fig. 5f. d, Gating strategies for FACS analysis shown in Extended Data Fig. 5g. e, Gating strategies for FACS analysis shown in Extended Data Fig. 6a. f, Gating strategies for FACS analysis shown in Extended Data Fig. 6b. g, Gating strategies for FACS analysis shown in Extended Data Fig. 6c.