Atypical B Cells Promote Cancer Progression and Poor Response to Bacillus Calmette-Guérin in Non–Muscle Invasive Bladder Cancer

Bladder cancer is the most common urological malignancy known worldwide with 573,278 cases diagnosed in 2020 (1). The majority of patients with bladder cancer (75%–85%) are diagnosed with early-stage non–muscle invasive bladder cancer (NMIBC; ref. 2). Exposure to environmental and occupational carcinogens, tobacco smoking, recurrent urinary tract infections, and aging can be attributed as the major risk factors for bladder cancer (3). Although more common in males (2), females with bladder cancer generally present with advanced-stage tumors and experience shorter recurrence- and progression-free survival (4–7). Following transurethral resection of bladder tumor (TURBT), patients predicted to be at a high risk of recurrence or progression to muscle-invasive disease are treated with intravesical Bacillus Calmette-Guérin (BCG) immunotherapy (8). The treatment regimen involves an induction phase of 6 weekly instillations of BCG after tumor resection, followed by a maintenance phase of 3 weekly instillations every 3 to 6 months for more than 1 to 3 years depending on the risk stratification (9).

Despite its proven efficacy, recurrence post-BCG treatment occurs in more than 50% of patients, necessitating repeated treatments, continuous surveillance, and cystectomy upon progression to the muscle-invasive stage (10, 11).

Our previous study discovered a significant association between increased density of intratumoral B cells and poor outcomes in patients with NMIBC (12). Such an inverse association has been also observed in patients treated with BCG immunotherapy (bioRxiv 2023.10.19.562817). Tumor-infiltrating B cells are mainly localized within stromal aggregates of immune cells, also commonly referred to as tertiary lymphoid structures (TLSs; ref. 13). TLS/bladder mucosa-associated lymphoid tissue formation in the bladder mucosa is induced due to chronic inflammation (14), such as persistent urinary tract infection (15, 16), exposure to carcinogens or smoking, immunomodulatory therapy (13), or age-related increase in systemic levels of tumor necrosis factor-α (TNFα; refs. 17, 18). It is established that TLSs act as inductive sites that sample both exogenous and endogenous mucosal antigens. Indeed, our previous report demonstrated an increased density of TLSs in tumors from advanced stages of bladder cancer compared to those tumors at the early stage (19). Furthermore, we also showed that aging-associated TLS formation in the bladder mucosa occurs in a sex-differential manner (18).

Simultaneous to biological aging–associated decline in the periphery, B‐cell proportions significantly increase at mucosal sites, predominantly within mucosa-associated TLSs (17, 20–24). Biological aging, repeated vaccination, and chronic inflammation also lead to the expansion of a circulating B‐cell subset known as atypical B cells (ABCs). ABCs are observed at a significantly higher frequency in females compared to males, exhibit features of high self-reactivity, and are equipped with specialized antigen-presenting ability (24–28). In addition, ABCs are known to produce high levels of IL6 and TNF-α and augment the suppression of B‐cell lymphopoiesis (29–31). In aging mice, ABCs increase in the bone marrow with a more pronounced expansion in the spleen, produce autoantibodies and respond uniquely to innate/microbial stimuli (26, 32). A key feature of ABCs is their proliferative response to stimuli that activate endosomal nucleic acid sensing pathways (toll-like receptors; TLR7/9) leading to the production of immunosuppressive cytokines such as IL10 (26, 28, 33–36).

Following intravesical instillation, BCG antigens are internalized by residual cancer cells, urothelial cells, and immune cells, leading to immune activation via various cytosolic and cell-surface receptors such as TLR2/9 (36–38). We hypothesized that repeated local exposure to BCG during the induction phase of the treatment potentiates systemic expansion of ABCs and recruitment to the bladder, specifically in those who exhibit a higher immune infiltration or tumor-adjacent (TA) TLSs in their pre-BCG tumors. The local immunosuppression mediated by ABCs potentially contributes to disease progression. As B cells/ABCs exhibit a sex- and age-dependent expansion and response to persistent immune activation, we investigated the role of B cells, with a focus on ABCs, in aging wild-type C57BL/6 and four-core genotype (FCG) mice exposed to the carcinogen N-butyl-N-(4-hydroxybutyl) nitrosamine (BBN), which is closely related to a carcinogen present in tobacco smoke, and treated with BCG. We then characterized TLSs in pretreatment tumors from patients, categorized as BCG responders and non-responders, using multiplex immunofluorescence (mIF) and NanoString GeoMx digital spatial technology (DSP). Overall, our study demonstrates that the expansion of ABCs following chronic carcinogen exposure and repeated BCG immunotherapy is one of the underlying factors for poor response to BCG immunotherapy, specifically, in females.

All murine experiments were in accordance with the guidelines provided by Queen’s University Animal Care Committee and approved in an animal utilization protocol. The B6.Cg-Tg(Sry)2Ei Srydl1Rlb/ArnoJ mice were kindly provided by Dr. Manu Rangachari (University of Laval) and also purchased from The Jackson Laboratory (stock number 010905, RRID:SCR_002187). Breeding of the FCG mice was conducted at the Queen’s University animal care facility to derive the four genotypes (gonadally female mice and male mice with XX and XY chromosomal complement each). The FCG transgenic mouse model (on a C57BL/6 background) allows for the generation of sex-reversed XX males and XY females by breeding mice with a loss-of-function mutation in the Y chromosome-linked testes-determining Sry gene with transgenic mice carrying an Sry rescue transgene on an autosome (39). This model permits determination of independent effects of gonadal and sex chromosome–specific features on bladder cancer progression–associated immunologic profiles. Mice were aged for 12 months at the facility. Wild-type C57BL/6 female and male mice (12 months old) were purchased from Charles River Laboratory. All mice were fed sterilized conventional diet and water and maintained under specific pathogen-free environment.

Female and male C57BL/6 mice (12-month-old) and FCG mice (12-month-old) were administered 0.05% BBN (Cat. # B0938, TCI America) ad libitum in drinking water (protected from light) once a week for 7 to 12 weeks as previously described (18). Our previous findings have demonstrated the formation of lymphocyte aggregates within the lamina propria of bladder beginning at week 4 post-BBN exposure as a result of BBN-induced DNA damage (18). To determine the influence of B cells on disease progression, we initiated B‐cell depletion starting week 3 post-BBN exposure. Briefly, a cocktail of monoclonal antibodies at 150 μg/mouse—rat anti-mouse CD19, rat anti-mouse B220, and mouse anti-mouse CD22—was injected via intraperitoneal (i.p.) injection, as described previously (40). After 48 hours, mice were injected with mouse anti-rat κ secondary antibody at 150 μg/mouse. To maintain the depletion of B cells in vivo, mice were injected with CD20-targeting antibody. B‐cell depletion cocktail injections were repeated three times every 10 days to ensure the complete depletion of systemic B cells, which was confirmed with flow cytometry post last injection. To determine the effect of B‐cell depletion on response to BCG in BBN-exposed mice, the depletion protocol was initiated 1 week prior to first BCG instillation followed by weekly anti-CD20 injections for 3 weeks during BCG treatment. Isotype controls for each antibody mentioned above were also used. Details of the antibodies used in the above experiments are provided in Supplementary Table S1.

Lyophilized TICE strain of BCG (Merck) was obtained from the pharmacy at Kingston Health Sciences Centre (KHSC). Each 50 mg vial contained 1 to 8 × 10⁸ colony forming units (CFU) of BCG, as tested by the manufacturer. BCG was resuspended in sterile saline for all experiments. At 7 weeks post-BBN exposure, mice were subjected to intravesical treatment with three weekly instillations of BCG (1 mg/mouse in 50 μL of saline). BCG was delivered via transurethral catheterization with a 24-gauge catheter (3/4″, Cat. # 4053, Jelco) for female mice and PE-10 tubing in male mice under isoflurane anesthesia. Urethral VASCU-STATT Midi clamps (Cat. # 1001-501, Scanlan) were placed to hold BCG within the bladder for 1 hour. Whole bladders were collected from all treatment groups post first (8 weeks post-BBN exposure) and third BCG (10 weeks post-BBN exposure) instillation, post first anti-CD20 treatment after first BCG instillation (8 weeks post-BBN exposure) and post third anti-CD20 treatment after third BCG (10 weeks post-BBN exposure) instillation and fixed in 10% formalin for histopathological evaluation and spatial immune profiling via mIF. Single-cell suspensions from bone marrow and spleen were subjected to profiling using multispectral flow cytometry. Blood was collected from the submandibular vein for plasma cytokine profiling.

Dissected formalin-fixed paraffin-embedded (FFPE) bladder tissue samples from mice were sectioned at 5 µm and hematoxylin and eosin (H&E) staining was performed using the standard protocol by the histology service of Queen’s Laboratory for Molecular Pathology (QLMP) in Queen’s University. The histopathological evaluation of H&E sections of bladder tissues from each mouse was quantified as per the scoring system mentioned in Supplementary Table S2. The scoring system was adapted from previous reports (41, 42), based on quantification of BBN-induced inflammation, hyperplasia, carcinoma in situ (CIS), and invasion. Inflammation characterized by presence of immune cells in previous report was scored as immune infiltration (41, 42). CIS in a mouse BBN-induced urothelial carcinoma model was defined as a carcinoma confined to the urothelium where the malignant urothelial (transitional) cells have loss of cell polarity, present cellular atypia, have increased number of mitotic figures, and large irregular nuclei with a high nuclear to cytoplasmic ratio.

Whole bladders collected from all treatment groups, at the indicated time points, were cut into two halves (sagittal plane), fixed in 10% formalin, and paraffin-embedded for H&E staining as described in above section. Five micrometers of unstained FFPE sections were subjected to spatial immune profiling using mIF staining to determine the infiltration patterns of selected immune markers. Antibodies to identify CD11b⁺ myeloid cells, CD3⁺ total T cells, CD8⁺ cytotoxic T cells, Pax5⁺ B cells, CD19⁺CD21⁻CD11c⁺ ABCs, PD-L1⁺ cells (Supplementary Table S3) were used in mIF staining at the Molecular and Cellular Immunology Core (MCIC) facility, BC Cancer Agency as per previously reported methods (18). Multiplex IF-stained images were acquired using the Vectra 3 multichannel imaging system. Images were then imported to Phenochart (11.0, Akoya Biosciences) and Inform Viewer (2.5.0, Akoya Biosciences) software for annotation followed by capturing at high magnification. The Inform Viewer was used to visualize a detailed composite image of each section with the flexibility of spectral separation of all the fluorophores from each panel. StarDist (https://github.com/stardist/stardist), a deep learning tool extension available in the QuPath software (https://qupath.github.io, v0.40), was used to further analyze the expression of markers within a region of interest (ROI; ref. 43). Positive staining thresholds were manually defined for each marker following confirmation using the DAPI nuclear stain channel. Immunecell infiltration was evaluated in 10 random annotated ROI adjacent to the basal layer of the urothelium in the bladders for each treatment group. Composite object classifier tool was used to identify co-expression of markers. Measurements were calculated per mm² and exported to GraphPad Prism (v9.5.1) to analyze differences between sexes and treatment groups.

Unstained sections from the FFPE whole bladder tissues were acquired for mIF staining. Sections were first deparaffinized then boiled for 14 minutes in citrate-based antigen retrieval buffer. Tissues were then blocked using 5% skim milk diluted in PBS-T (Tween 20, 0.05%) buffer for 30 minutes. Samples were stained with antibodies against mouse CD11c (Cat. # 14-0114-81, Thermofisher), B220 (Cat. # 14-0452-81, Thermofisher), CD21 (Cat. # MA5-32227, Thermofisher), and IgM (Cat. # A90-101A, Fortis Life Sciences), diluted in 0.5% skim milk, and incubated >12 hours at 4°C. Sections were washed in PBS-T, then stained with secondary antibodies conjugated to fluorochromes with different emission wavelength (Supplementary Table S4) and further incubated at room temperature (RT) for 1 hour. Imaging of stained sections was performed using Leica SP8 confocal, laser scanning microscope (Leica Microsystems).

Single-cell suspensions from spleens were prepared by mechanical dissociation. Splenocytes were digested in RPMI-1640 media (Cat. # 350-000-CL, Wisent Bioproducts) containing 10 μg/mL DNase I (Cat. # 04716728001, Millipore Sigma) and 1 mg/mL collagenase IV (Cat. # 11088866001, Millipore Sigma) for 30 minutes at 37°C. Bone marrow was aspirated with RPMI-1640 media using a 26-gauge needle and passing it through a 40 μm cell strainer followed by red blood cell (RBC) lysis in 1× working concentration of RBC lysis buffer (10×: NH₄Cl 8.02 g, Cat. # 12125-02-9, Bioshop), NaHCO₃ (0.84 g, Cat. # S233-3, Fisher Scientific), and EDTA (0.37 g, Cat. # E6758, Sigma-Aldrich) in 100 mL distilled water. Cells were washed in PBS and counted, wherein 0.5 × 10⁶ cells were seeded/well in a 96-well plate and utilized for staining. Nonspecific Fc receptors were blocked using 1 μg of TruStain FcX anti-mouse CD16/32 antibody (Cat. # 10319, BioLegend) for 30 minutes on ice. Single-cell suspensions were stained (all antibodies were purchased from BioLegend unless otherwise stated, Supplementary Table S5) for immune-cell populations of interest using one of three antibody panels: the ABC (CD45, CD19, CD21, B220, CD11c, IgD), T-cell (CD45, CD3, CD4 , CD8, PD-1, FoxP3, NK1.1) and myeloid cell panel (CD45, CD11b, CD11c, F4/80, PD-L1) for 30 minutes at 4°C. Stained cells were then washed using FACS buffer (Cat. # 00422226, Invitrogen), followed by fixation of the cells using 100 μL of fixation buffer (Cat. # 00822249, Invitrogen) for 30 minutes at 4°C. Fixed cells were washed with permeabilization buffer (Cat. # 00833356, Invitrogen) twice and stained with FoxP3 antibody (intracellular) for 30 minutes at 4°C. After two washes with permeabilization buffer, cells were resuspended in 200 μL FACS staining buffer and analyzed on the CytoFlex-S Flow Cytometer (Beckman Coulter). Single color positive controls, unstained and fluorescence minus one negative control were used for each antibody and their respective panels to aid in downstream analysis of immune cell populations. Gating strategies for identifying ABCs, myeloid cells, and T cells are shown in Supplementary Figs. S1 and S2. Analysis of the flow cytometry data was performed using FlowJo software v10.9 (BD Biosciences).

Gene expression profiling was performed on total RNA isolated from fresh frozen whole bladders using a custom designed nCounter gene panel (Supplementary Table S6) that included genes associated with immune function, various cancer-related chemokines and cytokines, and housekeeping genes (NanoString Technologies). Because of the pronounced changes observed in urothelium of female bladders following B‐cell depletion during BCG treatment, this analysis was only performed in bladders from female mice. Total RNA from fresh frozen bladders (as indicated) was isolated using the RNeasy Mini Kit (Cat. # 74004, Qiagen) as per the manufacturer’s instructions. The purity and concentration of isolated RNA was assessed using the NanoDrop ND-100 spectrophotometer (NanoDrop Technologies). Total RNA (150 ng) was used as a template for digital multiplexed profiling at Ontario Institute for Cancer Research as per previously established protocols (44). The nSolver software (NanoString Technologies) was used to normalize the raw nCounter NanoString counts using built-in positive controls. Normalization was performed using housekeeping genes and overall assay efficiency was calculated using geometric mean of each control. Normalized nCounter gene counts between different treatment groups was further analyzed in GraphPad Prism.

Whole bladders were collected from 12-month-old FCG mice exposed to BBN for 7 weeks and snap frozen in liquid nitrogen. Total RNA was isolated using TRI reagent (Cat. # T9424, Millipore Sigma). Qualitative and quantitative assessments of RNA were performed using a Nanodrop spectrophotometer ND-100 and a bioanalyzer (Agilent). Samples that met purity requirement (RNA integrity number >8.0) were subjected to bulk RNA sequencing (Genome Quebec). Analysis of the raw FASTQ files was done using R (v4.3.1). RNA sequencing reads were mapped to the mouse reference genome (GRCm39, GENCODE) and aligned using HISAT2 pipeline (http://daehwankimlab.github.io/hisat2/manual/). Raw read counts were obtained from the aligned transcriptome using the featureCounts function from the Bioconductor Rsubread package (v2.0.3). Transcripts per million normalization was performed, followed by log₂ transformation for visualization of the data. Expression data for genes of interest (Aicda, Btla, Cd163, Cd274, Cd4, Cd40lg, Cd8a, Cr2, Cxcl13, Cxcr3, Fcrl5, Icos, Il21, Pax5, Tnf) were extracted and visualized using the ggplot2 (v3.5.0) package. Kruskal–Wallis tests were conducted to compare median expression levels between different FCG genotypes. All statistical analyses were performed using R, with a significance threshold of P < 0.05.

Female and male mice from each treatment group (n = 4–5 mice per group) were subjected to blood collection from the submandibular (facial) vein. Separated plasma was aliquoted as per the requirements for multiplex cytokine analysis and stored at −80°C until further profiling. Plasma was diluted using PBS (1:1) and subjected to multiplex cytokine analysis using the commercially available 32-Plex Mouse Cytokine Discovery Array (MD-31; Eve Technologies).

Plasma Ig were analyzed using the LEGENDplex Mouse Immunoglobulin Isotyping panel (6-plex) by flow cytometry (Cat. # 740493, BioLegend). Plasma sample and standards were prepared using assay buffer as per the manufacturer’s instruction. Data were analyzed using the cloud-specific LEGENDplex Data Analysis Software Suite Qognit (v.2023.02.15). Statistically significant differences in the levels of Ig isotypes between the groups were determined using GraphPad Prism.

Twelve-month-old female and male mice (n = 3–4 per group) were exposed to BBN carcinogen as described above. Splenic B cells were isolated after 7 weeks of BBN exposure using the EasySep Mouse Pan B‐cell negative isolation kit (Cat. # 19844, StemCell Technologies). The purity and viability of enriched B cells was >90% as assessed by anti-B220 and anti-CD19 staining using flow cytometry. B cells were resuspended in a base B‐cell culture media comprising RPMI-1640 media supplemented with 2 mmol/L L-glutamine, 10% FBS (Cat. # F1051, Millipore Sigma), 1% penicillin/streptomycin (Cat. # 450-201-EL, Wisent Bioproducts), 1× sodium pyruvate (Cat. # 11360070, Gibco), 50 µmol/L 2-mercaptoethanol (Cat. # 21985023, Gibco), 1× nonessential amino acid (Cat. # 11140-050, Gibco), and 10 mmol/L HEPES (Cat. # 7365-45-9, Millipore Sigma). A total of 1.5 × 10⁶ cells/mL were seeded in a 96-well plate and treated with IFNγ (20 ng/mL, Cat. # 315-05, Peprotech), IL21 (20 ng/mL, Cat. # 210-21, Peprotech), and the TLR7 agonist Imiquimod (R837; 2.5 µmol/L, Cat. # tlrl-imqs, Invivogen) for either 6 or 18 hours at 37°C and 5% CO₂. Cells were also infected with BCG TICE (Merck) along with IFNγ and IL21 to compare the expansion of ABCs. Lyophilized BCG was reconstituted in B‐cell culture media at a concentration of 1 mg/mL, resulting in approximately 0.2 to 1.6 × 10⁷ CFU per ml. Following an 18-via hour incubation, the supernatant was collected, and cells were subjected to analysis via flow cytometry using the ABC antibody panel.

Total RNA was extracted from the cultured splenic B cells, after 6 hours of in vitro treatment/BCG infection, using the RNeasy Mini Kit (Qiagen) as per the manufacturer’s instruction. The quality and concentration of the RNA was assessed by the NanoDrop ND-100 spectrophotometer. Then, 150 ng of total RNA per sample was used to synthesize cDNA using LunaScript RT SuperMix kit (Cat. # E3010L, NewEngland BioLabs) in a 20 µL reaction volume. Quantitative real-time PCR was performed using a ViiA 7 Real-Time PCR System (Applied Biosystems). Briefly, the 20 µL reaction volume consisted of 10 µL of TaqMan Fast Advanced Master Mix (Cat. # 4444556, Applied Biosystems), 2 µL of cDNA, 1 µL of TaqMan primers (Supplementary Table S7), and 7 µL of nuclease-free water. Relative gene expression was calculated using 2^−ΔΔCt method and Ubc was the housekeeping gene.

Supernatants collected from the cultured splenic B cells after 18 hours of treatment were aliquoted in separate tubes and stored at −80°C until further profiling. Supernatant was subjected to multiplex cytokine analysis using a commercially available Mouse Cytokine Proinflammatory Focused 10-plex Discovery Array (Cat. # MD-F10; Eve Technologies).

Splenic B cells, treated with IFNγ, IL21, TLR7, and infection with BCG for 6 and 18 hours were subjected to immunocytochemistry. The treated cells were then transferred to cover slips precoated with a 0.1% poly-L-lysine solution (Cat. # P8920, Millipore Sigma) and incubated at 37°C with 5% CO₂ for 1 hour. Next, the cells were fixed with 4% paraformaldehyde (Cat. # 047392-9M, ThermoFisher) for 30 minutes at RT and blocked with 5% skim milk for 30 to 45 minutes at RT. Subsequently, the cells were incubated with primary antibodies—anti-CD11c and anti-LAM—overnight at 4°C. The cells were then incubated with the corresponding secondary antibodies for 1 hour at RT. All antibodies used in this protocol are listed in Supplementary Table S4. All intermediate washes were performed using PBS-T buffer. Finally, ProLong Gold Antifade Mountant with DNA stain DAPI (Cat. # P36941, Invitrogen) was used to prevent photobleaching, and the cells were transferred onto slides. The slides were manually visualized using a 100× objective lens in Zeiss Axio Imager M2 epifluorescence microscope (Zeiss). Raw TIFF image files were exported to QuPath software for visualization and qualitative analysis.

All studies involving the use of patient materials were approved by the Human Ethics Review Board at Queen’s University. The cohort comprised 28 patients (25 males and 3 females) with NMIBC who underwent TURBT surgery at KHSC between March 2016 and November 2017. FFPE tumor specimens were retrieved from the Department of Pathology at Queen’s University. Patient clinical details were retrieved via electronic chart review (Supplementary Tables S8 and S9). All patients in the cohort received adequate induction of BCG immunotherapy (>5 out of six doses) following index TURBT surgery. Response to BCG was defined as the time from each patient’s earliest TURBT resection to next malignant diagnosis. Operative notes were reviewed to exclude re-resections as recurrences. Patients who had recurrence-free survival greater than 2 years were classified as BCG responders (n = 15). Patients who had recurrence-free survival less than 1 year were classified as BCG non-responders (n = 13).

We employed mIF to characterize the organization and maturity of TLSs found in tumors from NMIBC patients. Using our selected archival FFPE samples, triplicate 5 µm sections were cut and mounted at QLMP. One section per sample was utilized for H&E staining. Stained sections were subjected to high-resolution scanning (Olympus VS120 Scanner) and the scans were uploaded for viewing on the web-based HALO image hosting service. Immune cell aggregates, mature TLSs, and other areas of interest were electronically annotated and reviewed for accuracy by a trained pathologist.

An unstained section from each of the samples was stained via automated mIF staining at the Molecular and Cellular Immunology Core facility (Deeley Research Centre). Samples were stained with fluorophore-conjugated primary antibodies against cell-surface proteins to identify B cells, T cells, dendritic cells (DCs), and high endothelial venules (Supplementary Table S10). Stained sections were scanned using the Vectra multispectral imaging system and viewed using PhenoChart 1.1.0 Software (Akoya BioSciences). Aggregates/TLSs were classified into primary (immature) or secondary (mature) follicle as previously reported (19).

For the quantification of different markers of TLS used in mIF, sections from H&E staining were evaluated (in HALO imaging platform) and matched mIF staining was used to identify regions to annotate using QuPath software annotation tool. Given the inherent heterogeneity and varying size of tumor sections obtained from TURBT, blinded manual scoring method was performed on 28 whole tumor sections. TLS evaluation based on H&E staining using HALO imaging platform allowed identification of lymphoid aggregates containing >50 immune cells. From this initial selection, we further evaluated only those TLSs that exhibited expression of CD79a⁺ B cells, CD3⁺ T cells, and CD8⁺ cytotoxic T cells in mIF staining. Each annotated region was meticulously scored based on expression of PNAd (high endothelial venules), CD21 (follicular DCs), and CD208 (mature follicular DCs) to identify different subsets of TLSs. QuPath software was used to analyze the expression of markers within a ROI (43). Positive staining thresholds were manually defined for each marker following confirmation using the DAPI nuclear stain channel. Composite object classifier tool was used to identify co-expression of markers. Measurements were calculated per mm² and exported to GraphPad Prism to analyze differences and perform statistical test between treatment groups.

For the quantification of B cells based on expression of CD79a and CD21, TA-TLS from both responders (n = 14) and non-responders (n = 11), with a total of 85 and 73 annotated regions respectively, were examined. TA-TLS containing >50 cells, specifically within a 1 mm distance of the tumor as per previously established criteria (45, 46), were used for the analysis. HALO imaging platform was used for initial cell count and distance measurement. Matched mIF stained sections of selected TA-TLS were annotated for cell detection using QuPath software (Supplementary Fig. S3A). Sections that failed these criteria were excluded from this analysis (1 from “responder” and 2 from “non-responder”). Positive staining thresholds were manually defined for each marker following confirmation using the DAPI nuclear stain channel. Composite object classifier tool was used to identify co-expression of markers. Measurements were calculated per TA-TLS and exported to GraphPad Prism to analyze differences and perform statistical test between treatment groups.

We employed NanoString GeoMx DSP technology (Supplementary Fig. S3B) to spatially characterize the expression profiles of 49 proteins (Supplementary Table S11) in TLSs and epithelial regions within a subset of 12 tumors (Supplementary Table S9; six responders and six non-responders) from the cohort of 28 tumors subjected to mIF profiling. This unique, high throughput multiplexed technique uses 60-nucleotide indexing oligonucleotides via a photocleavable linker region to quantify the spatially resolved abundance of protein or RNA from FFPE tissue sections. Unstained fresh-cut 5 µm sections from the selected samples were processed on the NanoString GeoMx DSP platform at the Imaging Centre (University of Minnesota). In addition to these oligo-linked antibodies, sections were also stained with fluorophore-conjugated primary antibodies against markers of tissue morphology. These fluorescent antibodies allowed for the sections to be pseudo-marked for CD45 (marker of immune cells; clone 2B11 + PD7/26, Novus, NBP2-34528), pan-cytokeratin (PanCK; marker of cancer cells; clone AE1 + AE3, catalog # NB2-33200), Syto13 (nuclear stain), and CD20 (marker of B cells; clone IGEL/773, catalog # NBP2-47840). Following pseudo-marking, remote ROI selection was performed, and a total of 180 ROIs were designated across our 12 samples. ROIs were characterized as purely PanCK⁺, purely PanCK⁻, or a mixture of PanCK^+/− staining (Supplementary Fig. S3C). Purely positive and purely negative stained ROIs were considered to be a single area of interest (AOI). ROIs with a mixture of PanCK^+/− staining were artificially segmented into tumor and tumor microenvironment areas via automated computer algorithms based on the presence or absence of PanCK staining. Such segmented regions were counted as two AOI. In total, we selected 232 AOI across our 12 samples (Supplementary Fig. S3C). Following selection, each AOI was individually subjected to focalized ultraviolet (UV) light. This highly precise UV exposure cleaved only the photocleavable linkers in the specified AOI, releasing the indexing oligonucleotides. These oligonucleotides were aspirated and quantified using the automated nCounter hybridization assay (NanoString). Protein expression was subsequently normalized and compared between AOIs.

Data processing and normalization were done using either the NanoString DSPDA software v2.2 or v2.3. Quality control (QC) analysis was performed using the default parameters in the GeoMx DSP Analysis Suite, wherein all 232 samples passed the QC. GAPDH and S6 expression values were used to normalize the data. Subsequently, 70 and 111 AOIs were selected for TLS and epithelium tissue, respectively (Supplementary Fig. S3C). The AOIs were marked based on four morphological markers which are Syto13, PanCK, CD45, and CD20. Three protein modules i.e., immune cell profiling core module (B2M, CD11c, CD20, CD3, CD4, CD45, CD56, CD68, and CD8), IO drug target module (4-1BB, LAG3, OX40L, Tim-3, VISTA, ARG1, B7-H3, IDO1, STING, and GITR), immune activation status module (CD127, CD25, CD80, ICOS, PD-L2, CD40, CD44, and CD27) were used to study the differences between TLS and adjacent epithelium tissue (Supplementary Table S11). Additionally, linear mixed effect models in conjunction with Benjamini–Hochberg multiple-correction testing was used to calculate differentially expressed genes between cancer and TLS with scan and ROI as random effects. To enable comparisons with other groups (responders vs. non-responders within TLS and epithelial regions), the sample size was insufficient for employing a linear mixed model. Consequently, the average normalized expression for both TLS and cancer within each sample (averaging across segment) was calculated. Subsequently, Mann–Whitney U tests (nonparametric t test) were conducted to assess the differences between groups. The differences were plotted as boxplots using in-house data processing script (http://github.com/dataanalysis).

Publicly available tumor whole transcriptome profiles of 283 pre-BCG Ta/T1 high grade index TURBT tumors (227 males, 56 females), recently reported by de Jong and colleagues (47) were used to determine BCG response associated expression patterns of TLS imprint genes (48), immune exhaustion-associated genes (PDCD1, CTLA4, and TIGIT), and ABC-associated genes (MZB1, CD19, FCRL5, ITGAX, CXCR3, and TBX21). Preprocessed and normalized bulk RNA sequencing data were used for comparative analysis of the BCG response subtype (BRS1, BRS2, and BRS3)-associated expression patterns. The BRS subtypes were previously established by de Jong and colleagues (47). Tumors within the BRS3 subtype were enriched in patients who experienced early recurrence/BCG failure. Differential gene expression analysis was performed using the nonparametric Kruskal–Wallis test in R (v. 4.3.1). Box plots were generated for visual representation of the results.

Statistical analysis was performed using GraphPad Prism. Results are expressed as mean ± SD. Comparisons between two groups with one independent variable was performed using nonparametric t test (Mann–Whitney). Three or more groups with one independent variable were analyzed using one-way analysis of variance (ANOVA) with Tukey’s post-hoc test. Three or more groups with two or more independent variables were analyzed using two-way ANOVA with Tukey’s multiple comparison test. Statistical significance is indicated by ^∗, P < 0.05; ^∗∗, P < 0.01; ^∗∗∗, P < 0.001; ^∗∗∗∗, P < 0.001; ns, not significant.

Data generated in this study are available in the manuscript and its supplementary files. All materials and any other data are available from the corresponding author upon reasonable request. The whole transcriptome data used in this study are publicly available through the published article (47).

To confirm the previously reported age/sex-related shifts in B‐cell subsets (17), we first determined the profiles of total B cells and ABCs in the spleen and bone marrow of healthy 12-month-old (reflecting immune profiles of middle-aged humans) C57BL/6 female and male mice. We did not observe any significant differences in the frequency of total B cells (B220⁺CD19⁺) between male and female mice in the spleen and bone marrow (Fig. 1A). However, the density of splenic ABCs (CD21^−/loCD11c⁺) was significantly higher in 12-month-old female mice compared to their male counterparts (Fig. 1B and C). Optimization of the B‐cell–depletion protocol, as per previous reports (41), was performed prior to conducting studies in the carcinogen-exposed mice.

To determine whether long-term BBN exposure–induced genotoxicity and local inflammation leads to increased recruitment of ABCs in the bladder microenvironment, we exposed 12-month-old aging female and male mice to BBN carcinogen for 12 weeks with or without concurrent B‐cell–depletion treatment as per the previously reported protocols (49). Multiparametric flow cytometry-based profiling of spleen and bone marrow revealed a significantly decreased frequency of total B cells within the depleted group (Fig. 1D; Supplementary Fig. S4A). To understand the cause and effect, we first assessed whether B‐cell depletion during BBN exposure attenuates inflammation in the bladder and delays cancer progression. Results from systemic immune profiling revealed clear sex differences in ABC expansion during BBN exposure after B‐cell depletion. Despite undergoing depletion, female mice showed relatively higher splenic ABC frequency, in contrast to their male counterparts (Fig. 1E; bone-marrow profiles for ABCs are shown in Supplementary Fig. S4B). On the other hand, untreated female mice exposed to BBN for 12 weeks, showed lower splenic ABC frequency compared to untreated males (Fig. 1E) indicating potential trafficking to the bladder. Consistent with this hypothesis, spatial immune profiling at this time point using mIF indicated higher ABC density in the bladder microenvironment of female mice not depleted of B cells compared to their male counterparts (Fig. 1F). Bladders from the B‐cell–depleted group, collected 1 week post-depletion (12 weeks post-BBN initiation), revealed significantly fewer numbers of total B cells and ABC compared to the non-depleted group (Fig. 1F and G). Histopathological assessment of whole bladder sections from both the groups revealed a benign or close to normal histology of the bladder urothelium in the B‐cell–depleted female mice (Supplementary Fig. S5A). Reactive atypia/dysplasia was present in the majority of B‐cell–depleted male and BBN-exposed female and male mice that did not undergo any depletion (Supplementary Fig. S5A). These findings suggest a sex-differential recruitment (or differentiation) of B cells to the bladder microenvironment due to chronic carcinogen–induced mucosal inflammation.

To determine the role of B cells in treatment response, we treated BBN-exposed mice with 3 weekly intravesical instillations of BCG (Fig. 2A) to mimic the state of repeated instillations at weekly intervals as previously established for in vivo studies in murine models. Histopathological evaluation using H&E staining revealed increased infiltration of immune cells and TLSs in the bladder lamina propria of BBN-exposed female mice compared to that of healthy female mice and BBN-exposed male mice, after 7 weeks of BBN initiation (Fig. 2B and C; Supplementary Fig. S5D and S5E; Supplementary Table S12). Following the first dose of BCG, histopathological assessment revealed the urothelium of BBN-exposed female mice to be in its benign morphological states (Fig. 2D; Supplementary Fig. S5D and S5E; Supplementary Table S12), in contrast to the saline treated female mice where dysplasia/CIS-like features were still present at this time point (Supplementary Fig. S5B). However, at 1-, 6-, and 8-week post-completion of all three doses of BCG, we observed urothelial dysplasia and progressive focal to widespread CIS, invasive carcinoma-like changes, or papillary tumors despite removal of BBN exposure at the time of BCG treatment initiation (Fig. 2E–G). This effect was pronounced in female mice only. Three weekly repeated BCG treatments also led to the formation of TLSs in the bladder microenvironment of male and female mice, as confirmed by mIF staining (Fig. 2J), in contrast to the bladders from saline-treated female mice where tumor growth was observed with no evidence of TLS formation (Supplementary Fig. S5B). An increase in splenic ABCs at 1 week post first BCG, compared to the pre-BCG levels, was observed in both female and male mice (Fig. 2H and I). The magnitude of increase was significantly higher in female mice (Fig. 2I). A significant decline in the proportion of splenic ABCs post third BCG in both sexes was observed (Fig. 2I) suggestive of potential trafficking of these cells to the site of BCG instillation. Additionally, histopathological evaluation (Fig. 2E; Supplementary Fig. S5E; Supplementary Table S12) and mIF (Fig. 2J and K) of whole bladder sections revealed increased density of ABCs and lymphoid aggregates, an effect more pronounced in BCG-treated female mice, compared to the pre-BCG bladder microenvironment. Indeed, removal of BBN exposure at the time of treatment initiation reduced the frequency of splenic ABCs and CD21^−/loCD11c⁺ B cells in female mice at 1 week after the third BCG instillation (Supplementary Fig. S6). Overall, these findings suggest a combination of carcinogen and BCG treatment induced expansion of B cells and ABCs that occurs at a higher magnitude in females. Disease progression post-repeated BCG instillations, despite the removal of BBN, potentially reflects mucosal immune exhaustion.

The X chromosome harbors multiple B‐cell function–associated genes such as CD40L, CXCR3, TLR7, IL9R, IL13RA1/2, and BTK (50). Furthermore, TLR7 (known to escape X inactivation) and IL13RA are also known to play key roles in the expansion of ABCs. We thus addressed the fundamental question of whether sex-associated ABC expansion in BBN-exposed and BCG-treated mice is influenced by gonadal hormones, sex chromosome complement, or a combination of both factors. We employed 12-month-old aging FCG mice to investigate these interactions (Fig. 3A). While no significant differences in the total splenic B cells (Supplementary Fig. S4C) were observed between the four genotypes; XXF (chromosomal and gonadal females), XYF (gonadal females), XXM (chromosomal and gonadal males), and XYM (gonadal males), the XXF mice exhibited significantly higher ABCs compared to all other genotypes (Fig. 3B and C). Such genotype-associated differences were not observed in bone marrow ABC profiles (Fig. 3C). We characterized the bladder immune microenvironment of FCG mice exposed to BBN for 7 weeks (time point for BCG treatment initiation) using bulk RNA sequencing and mIF. These analyses revealed significantly increased transcript levels of immune exhaustion-, ABC- and TLS-associated genes (Aicda, Btla, Cd274, Cd4, Cd79a, Cd8a, Cd163, Cr2, Cxc113, Fcrl5, Cxcr3, Icos, Il21, Pax5, and Tnf), and higher immune infiltration and TLSs in XXF mice compared to all other genotypes (Fig. 3D and E). We then investigated the sex-associated expansion of ABCs in BBN-exposed BCG-treated FCG mice. Splenic ABC proportions significantly decreased following the third BCG instillation in the BBN-exposed XXF mice compared to the BBN-exposed untreated and saline-treated XXF groups (Fig. 3B; Supplementary Table S13). This shift was accompanied by a simultaneous increase in the bone marrow frequency of ABCs at this time point, with a significantly higher frequency in XXF mice compared to the XYF, XYM, and XXM mice (Fig. 3C). The BBN exposure– and BCG treatment–associated significant shifts were not observed within any other genotype of the FCG mice indicating that ABC expansion is a plausible combined effect of both X chromosome–associated and hormonal factors. Histopathological evaluation revealed the presence of a higher density of lymphoid aggregates in the bladder lamina propria of XXF mice at this time point compared to XYF mice (Fig. 3F). Similarly, H&E staining and mIF revealed an overall higher immune-cell infiltration in the bladders of XXM mice compared to the XYM mice (Fig. 3E and F). These findings confirmed a biological sex– and age-associated expansion of ABCs in response to BCG treatment.

To further determine the response of B cells from female and male mice to ABC differentiation–inducing stimuli, we treated splenic B cells (Fig. 4A and B) isolated from 12-month-old female and male mice at 7-week post-BBN exposure with the ABC-inducing cytokines IFNγ and IL21 in the presence or absence of BCG bacteria (Fig. 4C). A significant increase in the frequency of ABCs (CD21⁻CD11c⁺) in IL21-treated B cells isolated from female mice was confirmed via multiparametric flow cytometry at 18 hours post-treatment (Fig. 4C; Supplementary Fig. S7A). B cells isolated from male mice did not show a similar increase in response to IL21 treatment (Fig. 4C). However, CD11c⁺ B cells from male mice exhibited a higher proportion of CD21^lo/+ cells compared to the female mice (Supplementary Fig. S7B and S7C). The higher expression of CD21 in B cells isolated from male mice potentially suggests a delayed expansion of ABCs following treatment with different ABC phenotype–inducing stimuli. Furthermore, stimulation with a TLR7 agonist (Imiquimod) significantly increased differentiation of splenic B cells to ABCs for B cells isolated from both female and male mice (Supplementary Fig. S7D). Gene expression analysis, using quantitative real time (qRT)-PCR, revealed significant alterations in the expression of the ABC-associated genes Itgax, Tbx21, and Fcrl5 at 6 hours post-stimulation (Fig. 4D–F). Treatment of splenic B cells from male mice with IL21 alone led to a significant increase in the expression of Itgax (Fig. 4D), indicative of higher ABC differentiation. BCG infection elevated Itgax expression in B cells from female mice only when added in combination with IFNγ (Fig. 4D). Treatment with IFNγ alone led to similar increase in expression of Tbx21, only in B cells from females (Fig. 4E). However, in the presence of both BCG and IFNγ, increase in Tbx21 expression was similar in both females and males (Fig. 4E). In contrast, the combination of BCG infection with IL21 and IFNγ treatment enhanced Fcrl5 expression only in splenic B cells of male mice (Fig. 4F). These findings indicate that BCG infection drives a higher differentiation of ABCs in B cells isolated from female mice. Similarly, multiplex cytokine analysis revealed significantly higher levels of TNFα, IL10, GM-CSF, IL6, and MCP-1 in the supernatants collected at 18 hours post-treatment of B cells from female mice compared to those from males (Fig. 4G–K). Internalization of BCG by B cells was confirmed via immunofluorescence assay using an anti-LAM (Supplementary Fig. S7E). These findings underscore the importance of considering sex-specific responses in B‐cell differentiation to ABCs following treatment with IFNγ and IL21 in the presence or absence of BCG bacteria.

Our findings showing ABC expansion during disease progression prompted us to further investigate whether B‐cell depletion alters the response to BCG immunotherapy in mice with chronic exposure to BBN carcinogen. We found that depletion of B cells prior to and during BCG treatment led to the return of urothelium to its benign morphological state in female mice (Fig. 5; Supplementary Fig. S5E; Supplementary Table S12). B‐cell depletion also led to reduced inflammation in the lamina propria of female mice (Fig. 5A) compared to male mice (Fig. 5C). In contrast, bladders from BCG-treated B‐cell–intact female (Fig. 5B) and male (Fig. 5D) mice showed dysplasia and focal CIS-like changes at 1 week following completion of treatment with three doses of BCG. TLS formation was rare or absent in the BCG + B‐cell–depleted mice (Fig. 5E; Supplementary Fig. S5E; Supplementary Table S12) compared to B‐cell–intact female mice (Fig. 5F; Supplementary Fig. S5E; Supplementary Table S12). Exposure to BBN carcinogen was maintained throughout the course of BCG treatment and B‐cell depletion to mimic smoking behavior during BCG treatment. Indeed, removal of the BBN carcinogen at the time of BCG treatment initiation in combination with B‐cell depletion revealed decreased invasion into the submucosal region (Supplementary Fig. S5C). To assess the long-term benefits of B‐cell depletion, histopathological assessment and spatial immune profiling was conducted on bladder whole sections at 8 weeks post third BCG instillation. Bladders from B‐cell–depleted female mice exhibited reduced TLS formation, invasion and multifocal hyperplasia compared to those of males (Supplementary Fig. S5F) and the female B‐cell–intact group (Supplementary Fig. S5F). Spatial profiling–based ABC quantification in the bladder microenvironment confirmed the sex-differential effects of B‐cell depletion (Fig. 5G and H; Supplementary Fig. S8).

Given the pronounced effect of depletion, specifically on the urothelium of female mice under both BBN removal and continued exposure conditions, we next examined the profiles of genes involved in immune-cell recruitment and TLS formation using a custom NanoString gene panel. The BCG-treated B‐cell–depleted female group demonstrated an increased expression of genes associated with an active effector immune state and antigen processing, such as Cd8A, Cxcl10, Cxcl9, H2eb1, H2d1, H2k1, Stat1, and Batf3 relative to the BCG-treated B‐cell–intact female group (Fig. 5I; Supplementary Fig. S9). On the other hand, bladders from female mice that underwent repeated BCG treatment without B‐cell depletion showed increased expression of genes associated with the expansion of ABCs and TLS formation, such as Tnf, Il6, Il17A, Cxcl13, Cd80, Cxcr4, Cxcr5, Cxcr1, Il1b, Ilr1, Tlr4, E2-2, Cd39, and Tmem173, relative to the BCG-treated B‐cell–depleted group (Fig. 5I; Supplementary Fig. S9). These findings suggest that in female mice, repeated weekly instillations of BCG lead to the systemic expansion and trafficking of ABCs to the bladder microenvironment accompanied by TLS formation.

Bladder-resident and recruited populations of myeloid cells are key to BCG response. In this study we observed that BCG treatment led to significant alterations in the profiles of CD11b⁺ myeloid cells systemically in both female and male mice. Overall, female mice depicted a higher proportion of myeloid cells in the spleen compared to males. A small but significant increase in splenic total CD11b⁺ myeloid cells, at 1 week post first BCG dose, was only observed in the female mice compared to the male mice (Fig. 6A; Supplementary Fig. S10A). Following completion of treatment with all three doses of BCG, in both female and male mice, there was a significant decline in bone marrow CD11b⁺ cells in B-cell–intact mice, which was not observed in mice undergoing B‐cell depletion during BCG (Fig. 6A). A significant increase in splenic PD-L1 immune checkpoint–expressing myeloid cells was observed only in female mice following the first BCG dose (Fig. 6E). The proportion of PD-L1–expressing myeloid cells was significantly higher following the third BCG dose in the spleen and bone marrow of both female and male mice that underwent B‐cell depletion during BCG treatment (Fig. 6E and F). This was also confirmed by spatial immune profiling of bladders from the same mice where increased infiltration of macrophages was observed (Fig. 6B and C). Spatial immune profiling of bladders revealed that the BCG + isotype treated female mice had increased infiltration of PD-L1⁺ cells relative to the B‐cell–depleted group (Fig. 6B–D). These findings suggest that the combination of BBN exposure and repeated weekly BCG treatment induces sustained local inflammation, which could potentially accelerate the exhaustion of systemic and bladder local antigen-presenting cells (APCs).

Given that BCG primarily induces T helper type 1 (Th1) responses via activation of CD4⁺ T helper cells, we explored the influence of B‐cell depletion on the profiles of T‐cell subsets. We did not observe any significant sex differences in splenic total CD3⁺ T‐cell frequency at both early and late (post third BCG) time points (Fig. 6G; Supplementary Fig. S10B). Interestingly, the frequency of total CD3⁺ T cells in the bone marrow was significantly higher in female mice post third BCG compared to male mice (Fig. 6G). Notably, in the B‐cell–depleted group there was a significant increase in total splenic T‐cell frequency compared to the B‐cell–intact group in both sexes at this time point (Fig. 6G). Splenic immune profiles of saline-treated mice reflected catheterization-induced sterile inflammation, an effect that is well established in both humans and mice (51, 52). A BCG-specific immune response was revealed by the changes in the urothelium (Supplementary Fig. S5) and the T‐cell profiles in the bone marrow (Fig. 6G) where a significant increase in CD3⁺ T‐cell proportions was seen only in the female mice that received three doses of BCG with or without B‐cell depletion. A decline in splenic CD4⁺ T helper cells and CD8⁺ cytotoxic T cells following all three BCG instillations hint towards potential recruitment to the bladder (Fig. 6H and I; Supplementary Fig. S10C). Presence of high numbers of CD8⁺ T cells were observed in the bladder by mIF at the same time point (Supplementary Fig. S8A). Overall, these findings suggest that the administration of BCG via the intravesical (mucosal) route may elicit a host response, specifically a BCG-specific T helper cell–associated response, potentially enhancing a cancer cell specific cytotoxic immune response.

Multiplex cytokine profiling of plasma collected following the completion of three doses of BCG revealed significant differences in pro-inflammatory and immune cell–activating cytokines key to mediating Th1 and Th2 responses. In female mice undergoing B‐cell depletion during BCG treatment, plasma levels of the Th1 cytokines; IL2, IFNγ, GM-CSF, MIP-1α, and MIP-1β were significantly higher compared to female mice that did not undergo B‐cell depletion during BCG treatment (Fig. 6L–P). Similarly, significantly increased levels of Th2 cytokines such as IL9 and IL13 were observed in female mice subjected to B‐cell depletion during BCG treatment (Fig. 6J and K). Healthy female mice displayed higher plasma levels of IgM, IgG2a, and IgG2b, compared to the age-matched male mice (Supplementary Fig. S11A–S11D). However, the levels of IgG2a declined at 7-week post-BBN and 1 week post third BCG in female mice (Supplementary Fig. S11C). In the female mice, the levels of IgG2b increased from their pre-BCG levels reaching up to similar levels to those in age-matched healthy controls, despite the depletion of B cells during BCG treatment (Supplementary Fig. S11D). In contrast, IgG2b levels decreased significantly in B‐cell–depleted male mice (Supplementary Fig. S11D). The elevated levels of IgG subtypes (Supplementary Fig. S11B–S11F) and Th1/Th2 cytokines, following BCG treatment and B‐cell depletion suggest a robust and effective immune activation in female mice compared to males. These findings imply a critical role of B‐cell subsets in inhibiting Th1 and Th2 responses triggered by repeated mucosal BCG administration in female mice.

Based on our previous observations indicating a correlation between high stromal B‐cell density and early recurrence and progression (bioRxiv 2023.10.19.562817), we investigated the spatial immune cell profiles and activation state of TA-TLSs. Whole tumor sections from 28 patients (clinical details in Supplementary Table S8) were analyzed. TLSs were identified via mIF analysis using antibodies specific for TLS-resident immune cells. Histopathological evaluation revealed presence of both small and large immune-cell aggregates within the lamina propria, indicative of chronic inflammation in the bladder mucosa (Fig. 7A and B). mIF staining showed varying degrees of TLS organization and maturation stages in tumors from both BCG responders and non-responders (Fig. 7B). Overall, the total number of TA-TLSs was higher in whole tumor sections from BCG non-responders compared to responders (Supplementary Fig. S3D).

Quantitative analysis of B‐cell subsets and follicular DCs in TA-TLS (identified as TLS within 1 mm distance of tumor), revealed significantly higher density of total CD79a⁺ B cells, CD79a⁺CD21⁺ mature B cells, and CD21⁺ follicular DCs in the non-responder group (Supplementary Fig. S3E).

Further, we analyzed the expression of 49 immune function–related proteins (Supplementary Table S11) in TLSs from a subset of 12 specimens (Fig. 7C; clinical details in Supplementary Table S9) using NanoString GeoMx DSP technology (Supplementary Fig. S3B and S3C). The area of mature TLSs was higher compared to all other lymphoid aggregates in the lamina propria (Supplementary Fig. S3F). We did not find any statistically significant differences in the expression levels of proteins associated with immune regulatory functions between TLSs from the responders and non-responders. However, when compared to the tumor epithelial compartment, TLSs from both responders and non-responders exhibited statistically significant higher expression of (i) immune-regulatory proteins such as CD25, ICOS, TIM3, VISTA, PD-L2, IDO-1; (ii) antigen presentation–associated proteins such as B2M, CD11c, CD40; and (iii) immune cell phenotype–associated proteins, CD45, CD20, CD3, CD4, CD56, and CD68 (Fig. 7D). These findings showing increased expression of immune-exhaustion and regulatory proteins are indicative of chronic carcinogen-induced mucosal inflammation characterized by exhausted populations of immune cells at the time of TURBT.

We further validated these findings using the whole tumor transcriptomic profiles of an independent cohort of 283 pre-BCG Ta/T1 high-grade tumors (47). This analysis focused on exploring the differences in expression levels of TLS imprint, ABC-associated and immune exhaustion–related genes, among the three distinct BCG response subtypes: BRS1, BRS2, and BRS3. The findings from this analysis (Fig. 7E), revealed significantly higher expression of TLS-associated (48), immune exhaustion–related (PDCD1, CTLA4, and TIGIT), and ABC-associated genes (FCRL5, TBX21) in tumors from patients experiencing early recurrence/BCG failure (BRS3 subtype). These findings further validate our results from spatial immune profiling.

B cells and TLSs have emerged as biomarkers of favorable prognosis and response to chemotherapy or immune checkpoint inhibitor therapy in some cancers including muscle-invasive bladder cancer (refs. 53, 54). Recent reports indicate that increased density and proximity of TA-TLSs is associated with improved survival in patients with MIBC receiving adjuvant chemotherapy or immune checkpoint inhibitors (55, 56). Furthermore, TA-TLSs are associated with favorable prognosis in patients with melanoma, renal, gastric, and lung cancers (57, 58). Contrary to this, in colorectal, breast, and hepatocellular carcinoma (59), a negative correlation between TA-TLSs and patient outcomes has been demonstrated (60, 61). Similar to these observations and our previously reported findings from pretreatment tumors of patients with NMIBC and MIBC, the results of the current study suggest a negative association between TA-TLS and BCG response (12). This association of TA-TLSs with progression following BCG treatment may primarily be attributed to the difference in local immunomodulation and downstream responses induced by live BCG bacteria compared to systemically administered immune checkpoint inhibitors or cytotoxic chemotherapy. The discrepancy underscores the complex relationship between TA-TLSs and patient outcomes, influenced by several parameters, including functional states of immune cells within TLSs, persistence of carcinogen, chronic/recurrent urinary tract infections, treatment type, age/sex, and disease stage. While our findings indicate high expression of immune cell exhaustion–associated proteins in chronically evolved pre-BCG TA-TLSs and elevated expression of TLS imprint signatures, immune exhaustion, and ABC-associated genes in the transcriptomic profile of tumors from patients with poor response to BCG, it is plausible that the induction of acute TLSs that form post-BCG would be associated with positive clinical outcomes due to BCG’s ability to induce granuloma formation (62). However, this association may be reversed depending on the carcinogen persistence and recruitment of exhausted immune-cell populations, from the systemic circulation following repeated exposure to live BCG bacteria (as observed in our in vivo studies).

While TLSs are known to act as inductive sites for sampling local antigenic stimuli, they may also play a pathological role by promoting auto-antibody production and harboring immunosuppressive and dysfunctional immune cell populations due to sustained local inflammation (63). Indeed, increased expression of JCHAIN transcripts and C1QA complement component in the BRS3 subtype tumors are suggestive of such local alterations. Previous reports in NMIBC have also showed significantly increased levels of circulating IgG antibodies in patients exhibiting early recurrence post-BCG treatment (64). This duality helps elucidate why TLSs have been strongly associated with both positive and negative outcomes across different cancer types. In this study, mIF-based characterization of TLSs did not reveal significant differences in the TLS density between tumors from patients who responded to BCG treatment and those who exhibited early recurrence post-BCG. However, the location and maturity of TLSs and its proximity to tumor epithelium/invasive margin were indicative of an aggressive tumor behavior accompanied by an exhausted local immune response. Despite the small cohort size (28 patients; 89% male), our findings provide insights into TLS biology. Moreover, validation of TLS-ABC gene signature in a larger independent cohort comprising 283 high risk patients highlights the importance of future investigations into the architectural details and functional states of TLSs pre- and post-BCG treatment.

BCG immunotherapy in NMIBC is unique with respect to its mucosal route of administration and dosing schedule that requires repeated intravesical instillations at weekly intervals. Such repeated exposure to BCG antigens, specifically during the induction phase, may cause an expansion of exhausted immune cell populations in the effector memory phase in some patients. Such an expansion may occur specifically in BCG unresponsive patients with a history of chronic or persistent carcinogen exposure, smokers, patients with autoimmune conditions or female patients with a history of recurrent urinary tract infections. A recent report by Strandgaard and colleagues (65) demonstrated T‐cell exhaustion as a major factor underlying BCG failure. It is plausible that in some patients, repeated BCG-induced immune activation will also lead to increase in exhausted memory B cells and their recruitment to the bladder microenvironment leading to less-effective antitumor immune responses. This effect will be more pronounced in older patients or those who exhibit higher pre-BCG intratumoral density of B cells within TA-TLSs, reflective of their aging/chronic carcinogen exposure associated profiles. It is tempting to speculate that analogous to this T-bet transcription factor associated fate in both T and B cells, expansion of ABCs (during the effector memory phase) may result from repeated encounter with BCG bacteria in the induction phase, specifically, in unresponsive patients. This effect may potentially also be more pronounced in female patients who generally experience early recurrence and progression. Indeed, our previous report suggests that both female and male patients with high pretreatment intratumoral B-cell density often experience early recurrence post-BCG (12).

Patient age is an independent risk factor associated with poor response to BCG immunotherapy in NMIBC (66). Immunosenescence-associated B‐cell alterations, governed by hormonal and sex chromosome–associated factors, accompany aging in a sex-differential manner (23). Indeed, immune profiling of healthy aging mice confirmed the higher frequency of ABCs in females compared to their male counterparts. Moreover, using the FCG mouse model, we established that such sex differences in BCG-induced systemic immune profiles as well as bladder-associated lymphoid aggregate formation, are influenced by both gonadal hormones and X chromosome–linked factors. Based on our in vivo findings, we reason that BCG internalized by ABCs in the bladder microenvironment/TLSs, activates the TLR9 pathway following the binding of bacterial CpG DNA (35). Such an amplified TLR9 response could lead to the production of immunosuppressive IL10, which may further inhibit antitumor T‐cell responses and polarize macrophages to M2-like suppressive states. In vitro infection of B cells using BCG bacteria confirmed such an increase in IL10. Further, significant decrease in PD-L1 expression in bladders B‐cell–depleted BCG-treated female mice reflects a favorable antitumor immune response.

Validation to the central role of B cells, in mediating responses to BCG, is reinforced by observations from our study. The depletion of B cells during BCG treatment reduced inflammation in the urothelium and helped accelerate its recovery despite the presence of carcinogen specifically in female mice. Additionally, the recovery of urothelium was accompanied by significant increases in circulating levels of the cytokines such as IL2, IFNγ, IL9, and IL13 that favor a Th1 and Th2 response. These cytokines play crucial roles in enhancing the immune system’s ability to fight against cancer cells and facilitate the resolution of inflammation. Furthermore, increased expression of Cd8A, Cxcl10, Cxcl9, H2-Eb1, H2-d1, and Batf3 was observed in bladders from mice undergoing B‐cell depletion during BCG treatment. A higher expression of Cd8A in the depleted group indicated a higher T‐cell density, which is also supported by our findings from mIF of whole bladder sections from the depleted group. Conversely, the upregulation of Batf3 suggests either increased DCs or the suppression of T regulatory cells, in the depleted group. On the other hand, the BCG-treated group exhibited higher expression of genes linked to ABC expansion and TLS formation, such as Tnf, Il6, Il17, Cxcl13, Cxcr4, Cxcr5, Cxcr1, Cd80, and Il1B. These findings align with our hypothesis that ABC expansion and recruitment following repeated BCG treatment could contribute to an exhaustive bladder microenvironment, potentially leading to early disease recurrence. These gene expression patterns further corroborate the differential immune responses observed between the two treatment groups, emphasizing the pivotal role of B cells in shaping the response to BCG immunotherapy. Despite the depletion of B cells, comparable levels of IgM, IgG2a, and IgG2b between the healthy control and B‐cell–depleted female mice is suggestive of a memory B‐cell response triggered by BCG infection given the enhanced rejuvenation of lymphopoiesis in females. It is known that the cytokine IL21 (inhibits CD8⁺ T‐cell cytolytic function), which governs ABC expansion, plays opposite roles compared to those of IL2 (promotes cytotoxic function of CD8⁺ T cells; ref. 67). B‐cell depletion prior to initiation of BCG treatment may thus ablate such IL21-mediated suppression of T‐cell function thus enhancing antitumor immunity especially in females or males with high TA-TLSs.

Our study is not without limitations. We did not establish the precise in vivo cellular trafficking associated events following BCG. However, our findings of significant systemic expansion of CD3⁺CD8⁺ T cells post third BCG in both B‐cell–depleted and nondepleted mice, suggest that BCG bacteria travel to the bone marrow where further cellular priming and activation may occur. While our study demonstrates the critical role of B cells and ABCs in mediating response to BCG, the mechanistic basis of how these cells suppress antitumor cytotoxic response remains unclear and needs further investigation. In this study, we administered only three doses of BCG, as our primary focus was on examining the effects of repeated weekly BCG exposure. The events observed in female mice following three doses could potentially manifest in males with further doses. Future studies should investigate the dose-dependent response, taking into consideration the inherent differences in immune physiology between sexes. Additionally, the spatial immune profiling conducted on a small cohort of 12 high-grade tumor sections (six responders and six non-responders) predominantly included samples from male patients (83.33%), limiting gender-based analyses, and comparison within other clinical features such as tumor grade. Conducting an in-depth characterization of the TA-TLS in a larger cohort of patients could provide mechanistic insights into the TLS biology and their potential as predictive biomarkers of response to BCG. Nonetheless, the results from the comprehensive analysis of the whole transcriptome in a large independent cohort (283 high-risk tumors) validate our findings, as they align with the immune exhaustion signatures obtained from spatial profiling of TA-TLS. It is established that B‐cell–associated CD21 (complement receptor 2) binds to autoantigens or immune complexes (68), and interacts with TLR9. Moreover, CD21/CR2 has been shown to be critical for the internalization of BCG bacteria by B cells (69). The role of the complement cascade involving antibody-dependent cell-mediated cytotoxicity in BCG response in NMIBC has not been explored and needs further investigation. While we propose peritumoral/TA-TLSs to have evolved due to chronic carcinogenesis-associated inflammation, the possible role of recurrent urinary tract infection cannot be ruled out given their prevalence with increasing age.

Overall, this study not only advances the current understanding of the BCG mode of action in the context of bladder mucosal immunity but also provides the first evidence for a critical role of B‐cell exhaustion, reflected by higher pre-BCG TA-TLSs and expansion of ABCs, in disease progression and treatment response. While mouse models do not reflect the disease complexity observed in human scenarios, our study considers aging aspects of the immune system, unique aspects of the mucosal immune responses and sex differences via the inclusion of mice from both sexes and the FCG mice. Taken together, the findings from this study partially explain the poor outcomes in patients with high intratumoral B cells and TA-TLSs. Therapeutic targeting of ABC-associated markers in combination with optimal sequencing of PD-L1 immune checkpoint blockade could be a novel approach for treatment of patients identified early based on their systemic and bladder-associated immune profiles.

M. Rangachari reports grants from Remedy Pharmaceuticals and personal fees from Novartis Canada outside the submitted work. W. Kassouf reports personal fees from Janssen, Bayer, Ferring, Photocure, BMS, EMD Serono, and AstraZeneca outside the submitted work. D.M. Berman reports personal fees from Acrivon Therapeutics outside the submitted work. No disclosures were reported by the other authors.

P. Yolmo: Data curation, software, formal analysis, validation, investigation, visualization, methodology, writing original draft, writing, review and editing. S. Rahimi: Validation, investigation, visualization, methodology. S. Chenard: Investigation, visualization, methodology. G. Conseil: Validation, investigation, visualization, methodology, writing, review and editing. D. Jenkins: Investigation, visualization. K. Sachdeva: Software, validation, writing, review and editing. I. Emon: Investigation. J. Hamilton: Software, formal analysis, investigation. M. Xu: Formal analysis, validation. M. Rangachari: Resources, writing, review and editing. E. Michaud: Resources, writing, review and editing. J.J. Mansure: Data curation, writing, review and editing. W. Kassouf: Resources, writing–review and editing. D.M. Berman: Resources, data curation, writing, review and editing. D. R. Siemens: Resources, data curation, formal analysis, supervision, validation, writing, review and editing. M. Koti: Conceptualization, resources, formal analysis, supervision, funding acquisition, validation, investigation, methodology, writing–original draft, project administration, writing, review and editing.

M. Koti conceptualized and designed the study. P. Yolmo, S. Rahimi, S. Chenard, G. Conseil and M. Koti performed experiments included in this study, analyzed data, and contributed to manuscript writing and reviewing. I. Emon helped with BBN experiments. K. Sachdeva and J. Hamilton analyzed the NanoString GeoMx DSP data. D. Jenkins helped with chart review and identification of BCG responders and non-responders. M. Xu helped with reviewing histopathological features and retrieving the archival tumor tissue specimens. D.R. Siemens and D.M. Berman helped with clinical classifications of BCG-treated patient specimens. E. Michaud and J.J. Mansure designed the Nanostring nCounter panel in consultation with W. Kassouf. All authors reviewed the manuscript. This study was supported by research operating grants from Bladder Cancer Canada, Cancer Research Society, Ontario Ministry of Research Innovation and Science: Early Researcher Award, and Canada Foundation for Innovation to M. Koti. We thank Lee Boudreau at QLMP for his assistance with embedding of FFPE blocks, sectioning, and H&E staining. We thank Shakeel Virk at the QLMP for his assistance with imaging and Halo software and Katy Milne at BC Cancer’s Molecular and Cellular Immunology Core (MCIC) for immunofluorescence staining of TMAs. Dr. Patricia Lima at the Queen’s Cardiopulmonary Unit helped with optimization of ABC panel and confocal microscopy-based imaging of multiplex IF stained whole bladder sections. We thank Dr. Andrew Winterborn and the staff at the Queen’s animal care facility for all the support with animal studies.