Microbial cancer immunotherapy reprograms hematopoiesis to enhance myeloid-driven anti-tumor immunity

Summary:

Mycobacterium bovis Bacillus Calmette-Guérin (BCG) is the vaccine against tuberculosis and an immunotherapy for bladder cancer. When administered intravenously, BCG reprograms bone marrow hematopoietic stem and progenitor cells (HSPCs), leading to heterologous protection against infections. Whether HSPC reprogramming contributes to the anti-tumor effects of BCG administered into the bladder is unknown. We demonstrate that BCG administered in the bladder colonizes the bone marrow and, in both mice and humans, reprograms HSPCs to alter and amplify myelopoiesis. BCG-reprogrammed HSPCs are sufficient to confer augmented anti-tumor immunity through production of neutrophils, monocytes, and dendritic cells that broadly remodel the tumor microenvironment, drive T cell-dependent anti-tumor responses, and synergize with checkpoint blockade. We conclude that bladder BCG acts systemically through hematopoiesis, highlighting the broad potential of HSPC reprogramming to enhance the innate drivers of T cell-dependent tumor immunity.

Graphical Abstract

Introduction

Bacillus Calmette-Guérin (BCG) is an attenuated strain of Mycobacterium bovis used worldwide as a vaccine for tuberculosis. BCG is also the first cancer immunotherapy¹ and the only bacterial therapy for cancer. The instillation of live BCG directly into the bladder is the standard of care for non-muscle-invasive bladder cancers (NMIBC)² due to its ability to reduce recurrence rates when compared with surgical resection alone^3–6. However, approximately 50% of bladder cancer patients fail to maintain durable responses to BCG^7–10, and there are no well-established biomarkers to predict responses in advance of treatment initiation, in part due to an incomplete understanding of the mechanism by which BCG mediates tumor clearance^11,12.

Following bladder instillation, BCG attaches to urothelial cells^13–16, resulting in tumor infiltration of myeloid and lymphoid cells in both humans and mice^17,18. In mouse models, bladder BCG activates tumor-specific CD4⁺ and CD8⁺ T cells, including induction of interferon gamma (IFNγ) production, resulting in tumor rejection^19–22. Complementary evidence from human studies has identified tumor-specific CD4⁺ T cells in BCG-treated patients with NMIBC²³. These effects are presumed to be mediated locally within the bladder, but the upstream events stimulated by BCG that enable tumor-specific immunity remain poorly defined^24–26.

An expanding body of research has established that microbes and their products, including BCG, can lead to epigenetic changes in innate immune cells that confer an increased capacity to respond to homologous and heterologous immune stimuli, a process termed innate immune memory or trained immunity^27–29. The persistence of these phenotypes in short-lived innate immune cells is explained by microbe-induced epigenetic changes in hematopoietic stem and progenitor cells (HSPCs) in the bone marrow^27,30. Epigenetic changes in HSPCs (central innate immune memory) are associated with skewed hematopoiesis toward myeloid cell production (myelopoiesis) and can be inherited by progeny cells^31,32,33. Single-cell analysis of gene expression in human monocytes following BCG vaccination revealed that increased activity of type II versus type I interferon programs is associated with augmented innate immune function^32,34. This augmented innate immune activity can confer heterologous protection against infection, including against respiratory viruses in BCG-vaccinated children or elderly nursing home residents^35–38. Although intravenous (IV) BCG administration generates innate immune memory³⁰, it is unknown whether HSPC-reprogramming occurs upon BCG administration in the bladder and whether this effect contributes to anti-tumor immunity. Recent studies indicate that bladder administration of BCG can increase circulating cytokines³⁹, modify monocyte phenotypes^40,41, and reduce the frequency of upper respiratory infections⁴². The relation of these observations to anti-tumor immunity is unknown, as are the mechanisms involved, including HSPC alterations. We sought to understand how bladder BCG contributes to anti-tumor immunity, including whether it acts systemically to reprogram HSPCs and, if so, how this reprogramming contributes to a functional anti-tumor response.

Results

Bladder BCG induces central innate immune memory

To determine if bladder BCG stimulates innate immune memory in HSPCs of human NMIBC patients, we collected blood samples from two independent longitudinal post-resection cohorts (Memorial Sloan Kettering Cancer Center, n=8; and McGill University, n=13) before initial BCG administration and one week after the 5th administration of bladder BCG (Figure 1A, Table S1).

Figure 1: Bladder BCG reprograms HSPCs and myeloid progeny in human bladder cancer patients.

A) Experimental design: PBMC paired single-nucleus ATAC- and RNA-seq analysis with progenitor enrichment (PBMC-PIE) before and after BCG in bladder cancer patients⁴³.

B) snRNA-seq UMAP: Cell types, with enriched HSPCs highlighted.

C) HSPC gene expression change (Cohort 1): MAST coefficient of change (FDR <0.01) post- versus pre-BCG in HSPCs, monocytes (CD14⁺), and cDCs.

D,E) Gene expression fold change (combined cohorts): HSPCs versus cDCs (D) and HSPCs versus monocytes (E) post- versus pre-BCG.

F) ChromVAR motif accessibility (combined dataset): Differential accessibility in cDCs post- versus pre-BCG. MeanDiff is the difference in average chromVAR score for accessible peaks in cDCs.

G) Transcription factor enrichment (chromVAR): HSPCs, monocytes (CD14⁺, CD16⁺), and cDCs.

H) ATAC-seq accessibility tracks (pseudobulk): HSPCs pre- and post-BCG.

I) Pre- and post-BCG interferon gamma (IFNγ) and antigen presentation module scores in CD14 monocytes and cDCs in individual subjects.

P-values were derived by Student’s t-test. P > 0.05 = ns, P ≤ 0.05 = *, P ≤ 0.01 = **, P ≤ 0.001 = ***, P ≤ 0.0001 = ****.

See also Figure S1 and Table S1.

We interrogated the bladder BCG-induced changes to HSPCs and mature immune cells by employing Peripheral Blood Mononuclear Cell analysis with Progenitor Input Enrichment (PBMC-PIE), a workflow established by our group (Figure 1A). PBMC-PIE allows for combined single-nucleus RNA and ATAC sequencing of human HSPCs via isolation and enrichment of rare circulating CD34⁺ cells in the blood, which faithfully captures the extensive diversity and phenotypes of bone marrow CD34⁺ cells^43,44. Using well-defined HSPC marker genes including MEIS1 alongside a panel of standard immune cell-type markers⁴⁵, we captured a total of 58,652 cells, including 57,543 mature immune cells and 1,118 HSPCs (Figure 1B, Figure S1A).

To identify molecular programs stimulated by bladder BCG, we focused on analysis of the HSPC (MEIS1⁺) and mature myeloid populations, namely CD14⁺ monocytes (CD14⁺, LYZ⁺) and conventional dendritic cells (cDCs) (FCER1A⁺, CST3⁺) independently in each cohort. Analysis of all significant differentially expressed genes pre- and post-BCG therapy in both HSPCs, CD14⁺ monocytes, or cDCs revealed prominent and consistent transcriptional changes across the two cohorts following bladder BCG treatment, including upregulation of genes and pathways associated with antigen presentation (Figure 1C, S1B). An HSPC-specific program was also induced, including upregulation of BST2, a regulator of HSPC activation downstream of IFNγ⁴⁶(Figure S1B).

In HSPCs from Cohort 1, we observed significant post-BCG transcriptional upregulation of genes and pathways associated with antigen presentation, including HLA-C, HLA-DRB5, HLA-DQB1, and B2M (Figure 1C). Analysis of HSPCs from Cohort 2 was largely consistent with Cohort 1, including upregulation of antigen presentation genes HLA-DRA, HLA-DRB1, B2M, and CD74 (invariant chain) as well as genes with other immune-related functions such as BST2 (Figure S1B).

To validate that combining cohorts would not introduce bias into our datasets, we applied an entropy-based measure to evaluate batch effects and found samples well-mixed in RNA local cell neighborhoods (Figure S1C–D). Because our subsequent analysis is based on cell-type clusters derived from the RNA neighbor graph, we did not apply further batch correction methods. After combining datasets, our RNA-based analysis showed differential genes consistent with analysis of the cohorts individually. We thoroughly validated similarity of both cohorts before combining datasets for further analysis (Figure S1B–D). These analyses again highlighted an antigen presentation program shared between HSPCs, cDCs, and CD14⁺ cells, indicating an HSPC derived effect that is passed to monocyte and cDC progeny (Figure S1E). To visualize this shared program from both cohorts, we compared the fold change of significantly differentially expressed genes pre- and post-BCG in HSPCs with the corresponding fold change in either cDCs or monocytes. This visualization highlights key genes associated with interferon-gamma-mediated signaling and antigen presentation (Figure 1D–E), a finding that was confirmed by GO pathway analysis, with top enriched categories “interferon gamma mediated signaling pathway” (p-value: 0.004) and “antigen processing and presentation of exogenous peptide antigen via MHC-I, TAP-independent” (p-value: 0.002) (Fig S1F).

We then explored predicted transcription factor (TF) activity in HSPCs and mature immune cells that may be driving the altered gene expression programs observed above, including those associated with augmented expression of antigen presentation and IFNγ pathways. In cDCs, we observed enrichment of characteristic interferon response factor (IRF) family motif accessibility post-BCG, as well as GATA and KLF motifs, previously associated with cellular survival and activation^47–49(Figure 1F, G, S1G,H). In HSPCs, we observed enrichment for the predicted activity of AP-1 (FOS/JUN), RUNX^50,51, and TAL/ZEB/ETS family members^52–54 (Figure 1G, S1H). AP-1 has been associated with the formation of stem cell innate immune memory^43,55, and the strong association of ETS family members with myelopoiesis indicates that these HSPCs are reprogrammed for increased myeloid output. Pseudobulk ATAC-seq tracks from HSPCs pre- and post-BCG revealed increased accessibility at the promoters of HLA antigen presentation genes (Figure 1H). Analysis of individual transcriptional responses of CD14⁺ monocytes and cDCs to bladder BCG revealed a consistent upregulation of both IFNγ and antigen presentation gene modules in both cell-types, with some interindividual variability (Fig 1I, S1I, S1J).

Collectively, these findings establish a BCG exposure signature in both HSPCs and their mature monocyte and cDC progeny following bladder BCG in humans. Chromatin accessibility and transcriptomic analyses indicate a signature consistent with interferon imprinting. The data indicate that the previously characterized systemic effect of IV BCG on HSPC-driven innate immune memory is also a feature of BCG administered in the bladder as a cancer immunotherapy.

Bladder BCG colonizes the bone marrow and alters HSPC composition

The changes observed in HSPCs after bladder BCG in humans indicate that locally administered BCG has systemic effects, possibly secondary to dissemination to the bone marrow, as has been reported with IV BCG in mice³⁰. To determine whether BCG colonizes the bone marrow after bladder administration and the functional consequences of BCG-induced HSPC reprogramming, we employed the BCG-responsive MB49 murine model of bladder cancer.²¹. We harvested and cultured bone marrow at weekly intervals during a 5-week course of bladder BCG administration and observed live BCG in the bone marrow of all mice that had received 5 doses of bladder BCG, and several of those that had received 3 or 4 doses (Figures 2A, S2A), a finding corroborated by PCR of the bone marrow using BCG-specific primers (Figure S2B).

Figure 2: Bladder BCG directly colonizes the bone marrow and reprograms HSPCs.

A) Cultured BCG in mouse bone marrow after weekly doses of bladder BCG. The proportion of culture positive versus negative mice according to number of weekly BCG treatments received is displayed on the right.

B) Experimental schematic: Bladder PBS, Bladder BCG (5 doses), or intravenous BCG (single dose).

C) Bone marrow HSPC subsets from mice in in Figure 2B were quantified by flow cytometry. Intravenous BCG samples were from an independent reference experiment and were not compared statistically with bladder PBS or bladder BCG samples.

D) Morphologic quantification of colonies of the indicated types was performed from single cell suspensions of bone marrow plated into methocult for 14 days. E – erythroid; G – granulocyte; M – macrophage.

E) Experimental schematic (left panel). Mice were implanted with MB49 administered 3 weekly doses of bladder PBS or bladder BCG. Bone marrow was harvested 7 days after the final administration, sorted for lineage^- cells, and paired snMulti-ome was performed. In right panel, UMAP(right) demonstrates cellular lineages.

F) Density of PBS- or BCG-treated cells projected onto the UMAP.

G) Predicted transcription factor (TF) activity (snATAC-seq): HSC/MPP, monocyte, and neutrophil populations after BCG.

H) Conserved TF activity changes (human vs. mouse): HSPC/HSC-MPP and monocyte populations.

I) Mouse ATAC-seq comparison: bulk LSK vs pseudobulk (snATAC-seq) HSC/MPP for CD74 (top) and H2-Eb1(bottom).

J) Serum Cytokine levels after bladder BCG: IFNγ (top panel), G-CSF (middle panel), and TNF (bottom panel).

Data shown in panels C, D and J represent mean with standard deviation P-values were derived by Student’s t-test. P > 0.05 = ns, P ≤ 0.05 = *, P ≤ 0.01 = **, P ≤ 0.001 = ***, P ≤ 0.0001 = ****.

See also Figures S2 and S3.

Prior studies established that IV BCG induces expansion and epigenetic modification of the lineage^- Sca-1⁺ Kit⁺ (LSK) population of HSPCs linked to the functional enhancement of mature myeloid cells^30,32,56. We treated mice with IV BCG or five doses of bladder BCG (Figure 2B) and performed flow cytometric analysis of HSPC subsets (Fig S2C). With both administration routes we observed expansion of the LSK population, decreases in proportion of long-term hematopoietic stem cells (LT-HSCs) and short-term HSC (ST-HSCs) and increases in multipotent progenitor (MPP) and common myeloid progenitor (CMP) cells (Figure 2C). Colony forming assays revealed that BCG stimulates differentiation toward granulocyte and granulocyte-macrophage lineages (Fig 2D). Consistent with published results³⁰, subcutaneous BCG did not stimulate LSK expansion to the same degree as bladder administration (Figure S2D). IV and bladder BCG were similarly protective against bladder tumors, and there was no synergy when combining the two (Figure S2E, F). Taken together, these results indicate that bladder BCG colonizes the bone marrow and alters hematopoiesis by increasing immune cell production skewed toward the myeloid lineage.

Bladder BCG remodels the chromatin landscape of HSPCs

To determine the bladder BCG-induced cellular and molecular programs of HSPCs and their progeny cells, we performed in-depth single-cell epigenomic and transcriptomic analyses of HSPCs from mice bearing MB49 bladder tumors after 3 doses of bladder BCG (Figure 2E). After identifying HSPC and mature cell subsets based on marker genes (Figure 2E, Figure S3A), we compared the distribution of cells from BCG- and PBS-treated mice and observed a notable increase in the density of cells in the neutrophil progenitor cluster in BCG-treated mice (Figure 2F), indicating that a component of the myeloid skewing induced by bladder BCG includes the neutrophil lineage. Similarly, an analysis of human HSPCs post-BCG revealed an enriched neutrophil signature (Figure S3B). Predicted TF activity analysis revealed enriched IRF and STAT TF activity across stem cells and myeloid progenitors (Figure 2G), strongly indicating that bladder BCG causes a systemic response that exposes HSPCs to interferon signaling, similar to what has been characterized with IV BCG³⁰. Differential gene expression analysis revealed transcriptional changes in HSC/MPP, neutrophil progenitor, and monocyte progenitor populations consistent with an interferon-responsive antigen presentation signature that is passed from HSPCs to progeny myeloid cells (Figure S3C).

We co-visualized myeloid cell TF activity programs from both our mouse and human datasets, plotting the fold change of each orthologous chromVAR score and denoting those that were significantly different in both species (Figure 2H). In HSC/MPP we saw a consistent activation of TFs that regulate myelopoiesis (SPI1, RUNX1, ZEB1), and in monocytes and dendritic cells, we saw a consistent activation of IRF family members and STAT1/3 (Figure 2H). We also observed concordant mouse and human RNA upregulation of antigen presentation and IFNγ responses in both monocytes and dendritic cells (Figure S3D).

To confirm these results in a purified bulk stem cell population, we treated mice with 5 doses of bladder PBS or BCG and performed bulk ATAC-seq on sorted LSKs (Figure S3E). Principal Component Analysis revealed clustering of LSKs from BCG-treated replicates compared to PBS-treated mice, with PC1 capturing chromatin accessibility associated with BCG treatment (Figure S3F). Differential peak accessibility analysis showed an overall upregulation of accessibility after BCG treatment (Figure S3G). Consistent with our findings from human and mouse single-cell analysis, HOMER motif analysis of differential peak accessibility from bulk ATAC-seq revealed increased inferred TF activity for IRF family members, along with NFY, a TF required for MHC enhanceosome formation and transcription of MHC-II genes (Figure S3H)⁵⁷, and PU.1 (SPI1), a master regulator of hematopoiesis crucially important for myeloid cell development⁵⁸. Comparison of pseudobulk tracks from the HSC/MPP snATAC-seq data with bulk LSK ATAC-seq revealed concordance across single-cell annotations of HSC/MPP with sorted LSK, and between these different experiments, with extensive similarities, including at genes involved in antigen presentation, such as CD74, and H2-Eb1 (Fig 2I).

To confirm the systemic inflammatory effects of bladder BCG indicated by the transcriptomics results, we assayed the levels of cytokines in the serum of mice treated with five doses of bladder BCG and found an increase in the levels of key cytokines including IFNγ and G-CSF (Figure 2J), consistent with our findings from the bone marrow showing neutrophil and interferon signatures. Additional cytokines elevated after BCG included TNF, CXCL5, CXCL10, IL1β, and IL12-p70 (Figure 2J, Figure S3I) indicating broad systemic inflammatory effects of BCG administered in the bladder.

BCG-experienced HSPCs are sufficient to limit tumor growth

We next sought to determine the contribution of BCG-induced reprogramming of HSPCs and mature myeloid cells to tumor control. We transplanted bone marrow from bladder or IV BCG-treated CD45.2^+/+ donor mice into naive irradiated CD45.1^+/+ recipients (Figure S4A). Flow cytometric analysis of PBMC subsets from bone marrow chimeric mice at 8-weeks post-transplant confirmed complete immune reconstitution (Figure S4B). To more directly assess the epigenetic programs and phenotypes of defined progenitors, we designed parallel experiments with chimeric animals generated from sorted and in vitro expanded LSK populations rather than total bone marrow (Figure 3A, S4A). IFNγ signaling in bone marrow cells, including lineage^- Kit⁺ cells, has been shown to induce Sca-1 expression, thus confounding flow cytometry gating strategies utilizing this as a marker for murine HSPC populations⁵⁹. We mitigated this bias and heterogeneity of sorted LSK populations by culturing them in media containing polyvinyl alcohol and growth factors optimized for expanding the primitive self-renewing HSC population⁶⁰ for two weeks before transplant. This protocol allows for acute inflammatory programs to resolve and reduces the potential for transfer of live BCG along with LSKs which could induce training in the recipient mouse³⁰. In line with our previous observations of post-BCG myeloid skewing, and highlighting the HSPC origin of these changes, we observed an increase in circulating myeloid cells derived from BCG-experienced donor bone marrow versus PBS controls (Figure 3B).

Figure 3: BCG-induced HSPC reprogramming through interferon gamma encodes tumor immunity.

A) Schematic for LSK transfer chimera. CD45.2^+/+ donor mice were treated bladder PBS or BCG (5 doses), or intravenous BCG (1 dose). Lineage^- KIT⁺ Sca-1⁺ cells were sorted and pooled from the groups of donors and expanded using poly-vinyl alcohol medium for 2 weeks. Mice were challenged with subcutaneous MB49 tumors and growth was measured longitudinally^95,96.

B) Quantification of circulating CD11b⁺ myeloid cell from bladder BCG or PBS bone marrow chimera as depicted in figure S4A.

C) MB49 tumor growth curves from chimeric mice reconstituted with LSKs from bladder PBS-, bladder BCG -, or intravenous BCG-treated donors, as in figure 3A. Statistical comparisons for Days 14 and 16 are displayed (right).

D) B16F10 melanoma challenge in chimeric mice from HSPCs from bladder PBS or BCG as described in Figure 3A.

E) MB49 tumor growth on chimeric mice generated as depicted in in Figure S4A using bulk bone marrow from BCG-treated donor mice that also received +/- neutralizing antibodies to IFNγ or the type I interferon receptor (IFNAR1).

F) Experimental schematic for the generation of mixed bone marrow chimeras. A mix of bulk bone marrow from CD45.2^+/+ donors treated bladder BCG(Group 1) or intravenous BCG (Group 2) and CD45.1^+/- CD45.2^+/- naive donors, into naive irradiated CD45.1^+/+ recipient mice.

G) Quantification of bone marrow HSPC populations from from chimeras generated as described in Figure 3F.

H) Quantification of tumor-infiltrating myeloid cells (top). Values represent fold change of cell frequency in the tumor compared to cell frequency in the spleen within each congenically-marked cell type. Representative flow (bottom). Percentages shown represent the frequency of the parent gate within each congenic marker. Monocytes (left) were gated by CD45 congenic marker, CD11b⁺, F4/80^-, Ly6G^-, and Ly6C⁺. Macrophages (middle) were gated by CD45 congenic marker, CD11b⁺, F4/80⁺, Ly6G^-, and Ly6C^-. Dendritic cells were gated by CD45 congenic marker, F4/80^-, CD11c+, and MHC-II⁺ from Chimeras generated as described in Figure 3F.

Longitudinal and cross-sectional analysis of tumor growth curves was performed using TumGrowth. Pairwise comparisons were performed using a Type II ANOVA. Cross-sectional analysis of individual timepoints was performed utilizing the Wilcoxon rank sum test, and p-values are adjusted using the Holm method.

Data shown in panels B, C, D, G and H represent mean and standard deviation P > 0.05 = ns, P ≤ 0.05 = *, P ≤ 0.01 = **, P ≤ 0.001 = ***, P ≤ 0.0001 = ****.

See also Figure S4.

Chimeric animals generated either from cultured LSKs (Figure 3C) or whole bone marrow (Figure S4B) were challenged with subcutaneous MB49 tumors, as this model is amenable to rapid detection and quantitative monitoring of tumor size⁶¹. BCG experienced donor LSKs conferred enhanced control of tumor growth, with no discernible difference between the two routes of BCG administration (Fig 3C). Challenge of BCG-experienced LSK-reconstituted mice with B16 melanoma tumors revealed a similar effect on tumor growth (Figure 3D) establishing that tumor control conferred by BCG-reprogrammed HSPCs extends to multiple tumor types. As an additional control to confirm that tumor control is not the result of transferring viable BCG, we treated donor LSKs with the anti-mycobacterial antibiotic isoniazid for two weeks prior to transfer into recipient mice and observed similar tumor control (Figure S4C). These results confirm that the BCG-experienced LSK HSPC subset alone, without contribution from mature populations can confer a systemic anti-tumor effect.

BCG-induced, HSPC encoded tumor immunity requires IFNγ but not type I IFN signaling.

To determine whether the interferon pathway upregulation observed in BCG experienced human and mouse HSPCs and their myeloid progeny (Figures 1,2) are important for HSPC encoded tumor control, we neutralized IFNγ⁶² or the interferon alpha receptor (IFNAR1^63–65) during bladder BCG treatment and generated chimeras using bone marrow from these interferon-neutralized donors. Neutralization of IFNAR1 had no effect on BCG-induced HSPC encoded tumor immunity (Figure 3E). In contrast, neutralization of IFNγ in BCG-treated HSPC donors abolished tumor control in recipient mice (Figure 3E), demonstrating that IFNγ pathway upregulation observed in HSPCs is critical for HSPC encoded tumor immunity.

BCG-reprogrammed hematopoietic stem cells confer enhanced tumor infiltration to mature innate immune cells

To characterize the effect of BCG-induced HSPC reprogramming on the bladder tumor microenvironment (TME), we generated groups of congenically marked mixed bone marrow chimeras according to the schematic shown in Figure 3F, wherein bone marrow from CD45.2^+/+ donor mice treated with either bladder or IV BCG was mixed 1:1 with naive CD45.1^+/-CD45.2^+/- bone marrow and transplanted into naive irradiated CD45.1^+/+ recipients (Figure 3F). Bone marrow analysis revealed similar proportions of LSKs from each donor (Figure 3G). However, we observed preferential myeloid progenitor expansion originating from bladder or IV BCG-experienced bone marrow compared with cells of the naive donor origin, consistent with our findings of BCG-stimulated myelopoiesis (Fig 3G, S4D). Specifically, CMPs, common monocyte progenitors (cMoPs), granulocyte-monocyte progenitors (GMPs), and neutrophil progenitors (NPs) originating from BCG-experienced HSPCs were all more abundant than those of naïve origin, with a similar magnitude of enrichment between the two routes of BCG administration (Figure 3G and Figure S4D).

These chimeric animals were then challenged with bladder tumors to determine competitive cell-intrinsic phenotypes within the TME. Since we observed myeloid skewing in BCG-experienced HSPCs, and to control for bone marrow output and estimate cell-intrinsic differences in tumor migration and proliferation, we analyzed the bladder infiltration of immune cells by normalizing to cell abundance in the spleen. We found an increased relative abundance of monocytes, macrophages, neutrophils, and DCs from both the bladder and IV BCG-experienced donor groups (Figure 3H, S4E, S4F, S4G). In contrast, there were no significant differences in abundance of T cell subsets from naive or BCG-experienced origins (Figure S4H), indicating that the effects of BCG on HSPCs do not affect abundance of their T cell progeny in the tumor. Together, these data indicate that reprogramming of HSPCs by BCG confers an augmented cell-intrinsic capacity for tumor infiltration to progeny myeloid cells.

BCG remodels myeloid cell function in the TME

To characterize the full effects of BCG on the bladder TME, we treated bladder MB49 tumors with BCG or PBS and performed single-cell RNA sequencing (scRNA-seq) on sorted tumor infiltrating CD45⁺ immune cells (Figure 4A). After unbiased clustering and cell type annotation of individual clusters (Figure S5A), we observed differential expression of genes related to antigen presentation and interferon pathways in monocytes and neutrophils and an activation signature in T cells consistent with the bone marrow phenotypes (Figure S5B)²¹. We also observed an enhancement of TNF gene expression in tumor infiltrating monocytes (Figure 4B). To validate this finding, we utilized our mixed chimera model (Figure 3F) to compare the proportion of TNF-producing cells in mature immune cell subsets from the tumor or the spleen and found an increase in TNF production in monocytes, neutrophils, and CD4⁺ T cells derived from BCG-reprogrammed HSPCs (Figure 4C, S5C). This result is also consistent with the increased level of TNF seen in the serum of bladder BCG-treated mice (Figure 2J) and studies which identified TNF as the principal serum factor induced by BCG that confers resistance to tumor challenge⁶⁶. To investigate the role of TNF in conferring HSC transplantable tumor control, we treated chimeric animals with a TNF blocking antibody and observed a loss of tumor control conferred by BCG-experienced HSPCs (Figure 4D). Consistent with a direct role for TNF in tumor elimination, we found that MB49 cells express the TNF receptor TNFR1 and are sensitive to TNF-mediated cell death, and that this effect is synergistic with IFNγ (Figure S5D, S5E, S5F), previously shown to be important for BCG-induced tumor elimination by signaling through the interferon gamma receptor on tumor cells²¹.

Figure 4: Tumor control by BCG reprogramming of HSPCs depends on enhanced TNF production and phenotype of tumor neutrophils.

A) Experimental schematic (left). Mice were implanted with MB49-YFP bladder tumors on Day 0 and treated with 3 weekly doses of bladder PBS or BCG on Days 2, 9, and 16. On Day 21 tumors were removed, CD45⁺YFP^- cells were sorted and characterized by scRNA-seq. Right panel: UMAP showing cellular lineages. Cells were annotated using cell type references from SingleR.

B) Density of TNF expression projected onto UMAP of cells from PBS- (left) or BCG- (right) treated mice.

C) Mixed bulk bone marrow chimeras were generated as described in Figure 3F utilizing bulk donor cells from a single dose of intravenous BCG and challenged with bladder tumors after reconstitution. Tumor and spleen cells were cultured for 4 hours in the presence of brefeldin A in the absence of further stimulation, and intracellular flow cytometry was performed to assess production of TNF. Data shown represent mean and standard deviation.

D) Congenically-marked bone marrow chimeras were generated by transferring bulk bone marrow cells from CD45.2^+/+ donor mice treated with a single-dose of intravenous BCG or PBS, into CD45.1^+/+ naive irradiated recipient mice, as described in Figure S4A. A subset of each group was treated with 200 μg of a TNF-blocking antibody. Longitudinal tumor growth is shown with quantitation of days 14 and 16.

E) Mice were given 5 weekly doses of bladder PBS or BCG. Bulk bone marrow was harvested 1 week after the final bladder treatment and bone marrow-derived macrophages (BMDMs) were generated and stimulated with LPS. RNA was extracted and RT-qPCR was performed. Data is plotted relative to GAPDH.

F) Bone marrow chimeras were as described in Figure S4A. After reconstitution, chimeric mice were challenged with subcutaneous MB49 tumors on day 0. A subset of mice from each group received 250ug of Ly6G depleting antibody. Longitudinal tumor growth is shown with Day 16 and 19 time point quantitated.

G) Average gene score of T3 neutrophils⁶⁹ from BCG- or PBS- treated tumors.

H) Dot plot of genes representing T1, T2, or T3 neutrophils

I) Stacked bar plot showing the frequency of T1, T2, or T3 neutrophils (left) and the ratio of T3 to T2 neutrophils in tumors from PBS- and BCG-treated mice (right).

Longitudinal and cross-sectional analysis of tumor growth curves was performed using TumGrowth. Pairwise comparisons were performed using a Type II ANOVA. Cross-sectional analysis of individual timepoints was done utilizing the Wilcoxon rank sum test, and p-values were adjusted using the Holm method.

Analysis of flow cytometry based experiments (Figure 4C, and 4E) was performed using the Students T-Test.

P > 0.05 = ns, P ≤ 0.05 = *, P ≤ 0.01 = **, P ≤ 0.001 = ***, P ≤ 0.0001 = ****.

See also Figure S5.

Bladder BCG activates HSPC-encoded macrophage hyper-responsiveness

We next asked if the population-level and epigenetic changes stimulated by BCG in HSPCs and their myeloid progeny result in functional enhancement by comparing bone marrow-derived macrophages (BMDMs) from mice treated with bladder BCG versus PBS. Consistent with the increased myelopoiesis documented above in BCG-treated mice, we saw a trend toward greater in vitro expansion of BMDMs from mice treated with bladder BCG (Figure S5G. Stimulation of BMDMs with LPS revealed enhancement of transcripts encoding the T cell-recruiting chemokine CXCL10, and the cytokines TNF and IL-6, in macrophages derived from bladder BCG-treated LSKs compared to control macrophages (Fig 4E).

BCG Reprogramming of Neutrophils in the TME through HSPCs

To assay the function of bladder BCG-induced changes in neutrophil development (Figure 2D, 2F, S3B) in the enhanced tumor control conferred by BCG reprogrammed HSPCs, we generated chimeric animals from bladder PBS or BCG-treated donor mice, challenged them with subcutaneous MB49 tumors, and depleted neutrophils⁶⁷. Consistent with published results⁶⁸, neutrophil depletion in control animals enhanced tumor control, suggesting loss of a pro-tumorigenic neutrophil population (Figure 4F). In contrast, enhanced tumor control conferred by BCG-reprogrammed HSPCs was lost when neutrophils were depleted (Figure 4F), indicating that BCG driven conversion of neutrophils from a pro-tumor to anti-tumor function is a critical effector mechanism of BCG-stimulated, HSPC-encoded tumor immunity.

A recent study observed recruitment of pro-tumorigenic neutrophils in bladder cancer⁶⁸, consistent with another recent report characterizing a tumor-enforced program that results in long-lived pro-angiogenic neutrophils termed T3 neutrophils⁶⁹. Analysis of gene expression and subsets of tumor-infiltrating neutrophils from bladder tumors from BCG or PBS treated mice revealed a shift from T3 pro-tumorigenic neutrophils to mature T2 neutrophils (Figure 4G–I). These results indicate that BCG contributes to improved tumor control by reprogramming HSPCs to produce neutrophils that are resistant to the pro-tumor functional reprogramming induced by the TME.

Bladder BCG reprogrammed myeloid cells increase antigen presentation and drive T cell response

We observed prominent upregulation of antigen presentation pathways in myeloid cells after BCG in both mice and humans (Figure S3D), a finding corroborated by analysis of the correlation between antigen presentation pathways in tumor neutrophils, monocytes, and bone marrow progenitor cells (Figure 5A), and in human monocytes post-BCG and mouse BCG-treated tumors (Figure 5B). Flow cytometry confirmed enhanced MHC-II expression in tumor infiltrating neutrophils from BCG-treated tumors (Figure 5C, S6A) and on spleen monocytes and neutrophils in mixed chimeric mice reconstituted with BCG-experienced HSPCs (Figure 5D), all suggesting that BCG enhances antigen presentation across myeloid lineages derived from BCG-experienced HSPCs in a cell-intrinsic manner.

Figure 5: BCG reprogramming of HSPCs augments MHC expression in myeloid cells and improves T cell activation and recruitment.

A) Correlation plots depicting relative expression of RNA transcripts in BCG versus PBS conditions in neutrophil progenitors versus tumor neutrophils (left panel) and in monocyte progenitors versus tumor monocytes (right panel).

B) Correlation plot between human circulating monocytes and mouse tumor monocytes showing shared transcriptional signature in BCG-treated conditions.

C) Mice were implanted with MB49 bladder tumors on Day 0 and treated with 3 weekly doses of BCG or PBS on Days 2, 9, and 16. On Day 21 tumors were removed and single cell suspensions were assessed by flow cytometry. The proportion of tumor neutrophils expressing MHC-II is shown.

D) Expression of MHC II in splenic monocytes and neutrophils from the mixed chimera experiment shown in Figure 3F

E) Averaged gene score from T cells (GO:0042110) in the sc-RNA-seq data from the experiment shown in Figure 4A.

F) Experimental Schematic. Bone marrow chimeras were generated as described in Figure S4A. After reconstitution, chimeric mice were challenged with bladder MB49^OVA bladder tumors, and CD45.1^+/+ OT-I and OT-II T cells were transferred 10 days later. Bladder tumors were harvested 5 days after T cell transfer to assess tumor-specific T cell frequency.

G) Representative histograms depicting the frequency of OT-I T cells in bladder tumors among total CD8⁺ cells in bladder PBS-, bladder BCG-, and intravenous BCG-experienced bone marrow recipients are shown at left. Quantification of OT-I and OT-II T cell frequency for all groups is shown at right.

H) Proportion of proliferated OT-I and OT-II cells out of total tumor OT-I and OT-II cells in the experiment shown in F.

Data shown in panels 5C, 5D, 5E, 5G, 5H represent mean and standard deviation.

Analysis of flow cytometry experiments (Figure 5C, 5D, 5G, 5H) was performed using the Students T-Test.

P > 0.05 = ns, P ≤ 0.05 = *, P ≤ 0.01 = **, P ≤ 0.001 = ***, P ≤ 0.0001 = ****.

See also Figure S6.

To determine if BCG-stimulated expression of antigen presentation machinery in myeloid cells drives enhanced tumor T cell responses, we calculated an average score for each cell in the T cell cluster, utilizing the genes in the GO Category “T Cell Activation”. We observed higher expression of these genes in T cells sorted from BCG-treated tumors (Figure 5E), consistent with our prior data (Figure S4G)²¹. To test whether tumor antigen specific T cell responses were enhanced by BCG-reprogrammed HSPCs, bone marrow chimeric mice reconstituted with LSKs from bladder PBS-, bladder BCG-, and IV BCG-experienced donors were implanted with MB49 tumors expressing the MHC-I and MHC-II epitopes of the model neoantigen ovalbumin (OVA). Seven days after tumor implantation, mice received congenically marked OT-I (CD8⁺) and OT-II (CD4⁺) transgenic T cells (Figure 5F). We observed increased infiltration of OVA-specific CD8⁺ T cells in both bladder and IV BCG-experienced bone marrow chimeras versus the control group (Figure 5G, S6B), as well as enhanced OVA-specific CD4⁺ and CD8⁺ T cell proliferation in animals reconstituted with BCG experienced HSPCs (Figure 5H, S6C).

Myeloid reprogramming from BCG exposed HPSCs drives T cell mediated tumor clearance

Prior data indicates that bladder tumor clearance by BCG is due to tumor-specific T cell immunity²¹. To determine whether HSPC-dependent tumor control also depends on T cells, we generated chimeric animals from either PBS- or Bladder BCG-treated donor mice, challenged them with subcutaneous MB49 tumors, and depleted CD4⁺ and CD8⁺ T cells (Figure 6A). We found that depletion of CD4⁺ and CD8⁺ T cells resulted in a loss of the tumor control phenotype conferred by BCG-experienced bone marrow, confirming a requirement for T cells in tumor control in this setting.

Figure 6: HSPC encoded tumor immunity depends on cDCs and T cells and synergizes with T cell directed immunotherapies.

A) Bone marrow chimeras were generated as described in Figure S4A. After reconstitution, chimeric mice were challenged with subcutaneous MB49 tumors on Day 0. A subset of BCG-experienced bone marrow recipients received CD4 and CD8 depleting antibodies. Tumor growth was measured longitudinally.

B) Bone marrow chimeras were generated by transferring bulk bone marrow from ZBTB46-DTR donor mice treated with one dose of intravenous BCG or PBS as described in Figure S4A. After reconstitution, chimeric mice were challenged with subcutaneous MB49 tumors on Day 0. A subset of each group was injected intraperitoneally with diphtheria toxin 200 ng per mouse on Days -2, 1, 4, 7, 11, 14, 17, and 20. Tumor growth was measured.

C) Bone marrow chimeras were generated by transferring bulk bone marrow from mice treated with 5 weekly doses of bladder BCG or bladder PBS, as described in Figure S4A. Chimeric mice were challenged with subcutaneous or bladder MB49 tumors followed by 5 doses of anti-PD-1 or PBS every 2 days.

D) Survival of mice from the experiment described in Figure 6C.

P-values for bar graphs were derived by Student’s t-test.

Longitudinal and cross-sectional analyses of tumor growth curves were performed using TumGrowth. Pairwise comparisons were done using a Type II ANOVA. Cross-sectional analysis of individual timepoints was performed utilizing the Wilcoxon rank sum test, and p-values are adjusted using the Holm method.

Data shown in panels 6A and 6B represent mean and standard deviation.

P > 0.05 = ns, P ≤ 0.05 = *, P ≤ 0.01 = **, P ≤ 0.001 = ***, P ≤ 0.0001 = ****.

See also Figure S6.

To understand the contribution of the HSPC-dependent enhancement of antigen presentation by DCs on T cell activation and tumor control, we depleted DCs using ZBTB46-DTR⁷⁰. Chimeric animals from either PBS- or BCG-treated donor ZBTB46-DTR mice were challenged with subcutaneous MB49 tumors and treated with DT to deplete DCs (Figure S6D). DC depletion abolished tumor control conferred by BCG-experienced bone marrow but had no effect on tumor control in PBS bone marrow chimeras (Figure 6B), demonstrating a critical role for DCs in tumor control conferred by BCG training of HSPCs.

The broad BCG-induced reprogramming of the myeloid TME stimulates enhanced anti-tumor T cell responses, which are the ultimate mechanism of BCG-induced tumor control. To test whether these effects are synergistic with immunotherapy that directly targets T cells, we challenged bone marrow chimeric mice reconstituted from BCG-experienced or control HSPCs with subcutaneous MB49 tumors and treated them with PD-1 blocking antibody. Mice reconstituted from BCG-experienced bone marrow demonstrated enhanced control of tumors over the PBS control group (Figure 6C, S5I). Control bone marrow chimeras treated with PD-1 blocking antibody demonstrated a similar level of tumor control to BCG-experienced bone marrow chimeras (Figure 6C, S6E), however, the combination of BCG-experienced HSPCs and PD-1 blocking antibody had the smallest tumors and approximately 30% survival, demonstrating a synergistic effect of BCG-induced innate immune memory and checkpoint blockade (Figure 6C, 6D, S6E).

Discussion

The development of BCG as the first cancer immunotherapy emerged from studies of the effects of microbes and microbial products on tumor growth^71–73. Despite its well-established clinical efficacy, the detailed mechanisms by which BCG controls tumors have remained elusive. Recent evidence in mouse models and human patients indicated that BCG-mediated tumor rejection is dependent on tumor-specific T cells^21–23. The accepted model of the anti-tumor effects of BCG, and the rationale for its administration in the bladder, was that it acts as a local immunotherapy at the site of administration to improve anti-tumor T cell priming. The upstream immunologic events stimulated by BCG that enable this immunity were unknown.

Although the fundamental principle of vaccination is the induction of antigen-specific T and B cell immunity, it has become clear that BCG vaccination elicits innate immune responses that cross-protect against antigenically unrelated pathogens^74–79. A prominent and rapidly growing body of work that embodies this phenomenon, termed innate immune memory or trained immunity, details the reprogramming of hematopoietic progenitors in the bone marrow and subsequent skewing of the abundance and function their myeloid progeny^30–32. It was unknown if bladder administration of BCG could induce this progenitor-encoded or “central” innate immune memory, nor if this innate memory contributes to the anti-tumor effects of BCG.

These data indicate that the hematopoietic reprogramming previously implicated in the heterologous protection against infection conferred by early-life intradermal BCG vaccination in humans^29,37,76, and by IV administration in mice^30,33,80, is a shared and intrinsic part of BCG-mediated immunotherapy of cancer. BCG-reprogrammed HSPCs were sufficient to confer enhanced tumor control to recipient mice, indicating a durable and persistent cell-intrinsic memory in progenitor cells that is conveyed through differentiation to mature myeloid cells. Our single-cell ATAC and RNA sequencing data from mice and humans, coupled with functional characterization of mixed bone marrow chimeras, indicate that a broad enhancement of myeloid function contributes to the anti-tumor effects of innate immune memory.

Myeloid cells derived from BCG-reprogrammed HSPCs broadly remodel the TME, including a functional dependence on and reprogramming of neutrophils, a critical role for TNF, as well as enhanced infiltration of tumors with inflammatory monocytes and DCs. Importantly, we show that tumor neutrophils derived from BCG-reprogrammed HSPC were resistant to conversion to pro-tumor, pro-angiogenic T3 neutrophils by the TME⁶⁹, supporting the idea that central trained immunity may also interfere with the ability of the tumor to co-opt neutrophils⁸¹. These findings demonstrate that cancer immunotherapy works in part through creating resilient anti-tumor neutrophil programs. A recent meta-analysis showed that a high pretreatment neutrophil to lymphocyte ratio in the peripheral blood of patients receiving BCG for bladder cancer was associated with worse recurrence free survival, supporting a pro-tumorigenic role for neutrophils in the pre-BCG setting⁸². The exact effector mechanisms in neutrophils that contribute to tumor permissiveness or elimination in our model remain to be determined. Multiple recent studies have identified a variety of neutrophil states and effector functions that contribute to tumor control in various murine tumor models, including direct killing of tumor cells by effector molecules⁸⁸, neutrophil-intrinsic gene expression programs⁸³, antigen presentation⁸⁴, and other mechanisms⁸⁵. Several of these mechanisms overlap with the characteristics of BCG-induced neutrophil reprogramming and the associated factors we identify as critical to BCG-induced, HSPC-encoded tumor immunity. Anti-tumor HSPC reprogramming was dependent on IFNγ and featured prominent epigenetic priming of antigen presentation pathways with augmented expression in mature myeloid progeny cells and driving increased anti-tumor T cell responses. Of substantial clinical significance, the myeloid cell progeny of BCG-experienced HSPCs strongly amplify the response to PD-1 blockade, thereby directly coupling innate immune memory to the anti-tumor T cell response.

Our findings have implications for the use of BCG not only in the context of bladder cancer but also as a broad immunotherapy against other cancers, given its induction of systemic anti-tumor HSPC phenotypes. BCG remains the standard of care for NMIBC, but a substantial minority of treated patients will experience tumor recurrence and there are no reliable pre-treatment predictors of response⁸⁶. It is possible that inter-individual differences in pre-treatment or BCG-induced innate immune memory could predict response to BCG, a hypothesis that can be tested using PBMC-PIE⁴³.

Broadly, our data indicate that augmenting the abundance and function of the myeloid compartment through HSPC-reprogramming is an effective method to improve anti-tumor immunity, especially in combination with immune checkpoint blockade. These concepts are supported by a recent study that revealed a pro-tumor hematopoietic circuit involving IL-4 that could be therapeutically targeted to improve anti-tumor responses⁸⁷, and another that described a bone marrow targeting peptidoglycan that induces HSPC reprogramming, increased myelopoiesis, and improved tumor control⁸⁸. There is substantial evidence that DC abundance and function are important determinants of checkpoint blockade activity^89–93, but harnessing this knowledge for DC-derived therapies is limited by the short lifespan of these cells. Reprogramming DC activity at the level of progenitors may represent an approach to overcome their short lifespan and TME-mediated suppression of DC maturation and antigen presentation^40,41.

The identification of HSPC-reprogramming as a mechanism of a longstanding microbial immunotherapy highlights therapeutic tuning of HSPC as a strategy for more durable alterations in myeloid function to enable successful anti-tumor immune responses across a range of anatomical locations and tumor types.

Study Limitations

Although we demonstrate that BCG traffics to the bone marrow, we do not identify the mechanism by which BCG travels to the bone marrow. A plausible mechanism would be that inflammatory cells in the bladder, recruited by local BCG, traffic BCG to the bone marrow, but further experiments will be required to validate this. Similarly, although BCG clearly appears in the bone marrow and HSPCs are reprogrammed in both mice and humans, we cannot conclude that BCG colonization of the bone marrow is required for the induction of HSPC encoded tumor immunity. While we demonstrate sufficiency of HSPC reprogramming for augmented anti-tumor responses, including via changes in the TME and T cell activation and across different anatomical sites, bladder BCG can also influence the local epithelial environment including through direct effects⁹⁴. Future work will be required to parse the relative contributions of local versus systemic effects. Finally, given the small sample size of our human study, we cannot yet determine whether measurement of innate immune memory can be used as a predictor of clinical outcome in BCG therapy. Future work should address this and extend beyond BCG immunotherapy to address whether variance in, and therapeutic tuning of hematopoietic phenotypes broadly contribute to anti-tumor immunity.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact Dr. Steven Josefowicz (szj2001@med.cornell.edu).

Materials availability

We will share all unique/stable reagents generated in this study upon request from the lead contact with a completed materials transfer agreement (MTA).

Data and code availability

Sequencing data and analysis code are publicly available in Zenodo repository: https://doi.org/10.5281/zenodo.10695064 and in GEO: GSE295309. All software used for analysis are listed in the key resources table. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

STAR Methods

Experimental Model and Study Participant Details

Clinical studies and participants

Memorial Sloan Kettering Cancer Center (MSKCC) cohort:

The cohort consists of 8 patients NMIBC who received induction therapy with bladder BCG. Specimens were collected under MSKCC IRB-approved protocols #06–107 and #12–245 and studied under protocol #19–015. Written informed consent was granted by all participants. Clinicopathological data is provided in Table S3.

McGill University cohort:

The cohort consists of 13 patients NMIBC who received induction therapy with bladder BCG. Specimens were collected under McGill University IRB-approved protocols. Written informed consent was granted by all participants. Clinicopathological data is provided in Table S3.

Animals

Wild-Type C57BL/6 (Strain #: 000664), CD45.1 (Strain #: 002014), OT-I (Strain #: 003831), OT-II (Strain #: 004194), and ZBTB46-DTR (Strain #019506) mice were from The Jackson Laboratory. All mouse strains were bred and housed in MSKCC Research Animal Resource Center under specific pathogen-free conditions. All animal studies were approved by the MSKCC Institutional Animal Care and Use Committee and were compliant with all applicable provisions established by the Animal Welfare Act and the Public Health Services Policy on the Human Care and Use of Laboratory Animals.

Cell lines

MB49 expressing luciferase under G418 selection, was a gift from Yi Luo, University of Iowa. B16 was obtained from Taha Merghoub. MB49 and B16 were grown in RPMI supplemented with 10% FBS, and 2 mM L-glutamine. MB49-YFP was constructed as previously described^19–22. MB49 TNFR1KO tumor cells were generated with help from the Gene Editing and Screening Core at MSKCC as described previously^21,22. Three candidate plasmid constructs were designed using the transient plasmid PX458 [pSpCas9(BB)-2A-GFP; Addgene, Plasmid #48138] containing GFP reporter, CRISPR-Cas9, and single guide RNA sequence targeting the Tnfrsf1a gene. The parental MB49 cell line was transfected with the candidate plasmids using Lipofectamine 2000 (Thermo Fisher) according to the manufacturer’s protocol. The transformed cell lines were sorted for purity based on GFP positivity and TNFR1 negativity. The final cell line used for TNFR1 knockout experiments utilized the gRNA sequence CGGGCCTCCACCGGGGATAT targeting the Tnfrsf1a gene at codons 125360786 to 125360808. Cells were cultured at 37 °C in a humidified atmosphere of 5% CO2. All cell lines used were confirmed to be negative for mycoplasma by annual testing using MycoAlert Plus (Lonza).

Bacteria

BCG Pasteur was grown at 37°C in Middlebrook 7H9 supplemented with 10% albumin/dextrose/saline, 0.5% glycerol, and 0.05% Tween 80. BCG was grown to mid-log phase (OD600 0.4 to 0.6), washed twice in PBS with 0.05% Tween 80, resuspended in PBS with 25% glycerol, aliquoted, and stored at -80°C. The final bacterial titer was determined from serial dilutions cultured on 7H10 agar.

Method Details

MB49 orthotopic tumor implantation

7–8-week-old female mice (The Jackson Laboratory) were anesthetized in an isoflurane chamber. For each mouse, a 24-gauge catheter (Santa Cruz) was inserted into the bladder through the urethra. Next, 100 μL of poly-L-lysine (Sigma) was injected through the catheter, the catheter was capped using an injection plug, and the mice were kept under anesthesia for 30 minutes. After 30 minutes, catheters were removed from one mouse at a time in the same order as they were implanted. The catheter was then flushed with a solution containing 500,000 MB49 cells/mL in RPMI. Each mouse was then removed from the isoflurane chamber, the bladder was manually emptied, and the catheter was re-inserted. 100 μL of the MB49 solution (50,000 cells/mouse, unless otherwise noted) was injected into the bladder and the catheter was re-capped. The mice were kept under anesthesia for 1 additional hour before catheters were removed, and the mice were allowed to recover from anesthesia. Mice were observed daily and were euthanized if they displayed signs of distress, such as dull fur, apathy, or visible signs of growing tumor.

Subcutaneous tumor implantation and measurement

100,000 cells were implanted per mouse and tumor measurements were obtained using a caliper by measuring the longest axis of the tumor first, followed by the perpendicular axis. Tumor area was calculated by multiplying the two axes. Mice were euthanized if tumor measurement surpassed 14mm in any dimension or if tumors were ulcerated⁶¹.

BCG administration

Frozen titered stocks of BCG were thawed and resuspended in PBS for a final concentration of 3×10⁷ colony forming units/mL (CFU/mL). PBS alone was used as a control. Mice were placed under anesthesia in an isoflurane chamber, transferred from the chamber to a nose cone for the procedure, and returned to the chamber for incubation steps. For bladder administration, a 24-gauge catheter was inserted into the bladder through the urethra, 100 μL of BCG (3×10⁶ CFU/mouse) was injected into the bladder, and the catheter was capped using an injection plug. The mice were kept under anesthesia for 2 hours, after which catheters were removed and mice were allowed to recover from anesthesia. Intravenous BCG (3×10⁶ CFU/mouse) was administered to anesthetized mice via retro-orbital injection according to standard techniques.

Flow cytometry

Cell suspensions were analyzed on a LSR Fortessa (BD Biosciences) or Aurora (Cytek), using FACS DiVa software (BD Biosciences) or SpectroFlo (Cytek), respectively. Data analysis was performed using the FlowJo software package (BD). For determination of cytokine production by T cells, single cell suspensions were restimulated with 1X Cell Stimulation Cocktail (plus protein transport inhibitors) (eBioscience) for 6 h at 37°C. Cells were first stained with a fixable viability dye, followed by surface markers, then fixed and permeabilized using Foxp3 Fixation/Permeabilization Buffer (eBioscience) according to the manufacturer’s instructions, and finally stained for intracellular antigens. Antibodies used for this study are detailed in the Key Resources Table.

Bulk bone marrow chimeras

Whole body irradiated mice using a cesium source (9 Gy) received 5×10⁶ bulk bone marrow donor cells in 0.1 mL sterile PBS via retro-orbital injection within 24 hours of irradiation. In mixed chimera experiments, each recipient mouse received 2.5×10⁶ of each congenically marked bone marrow (total of 5×10⁶ cells per animal). All recipient mice were rested for a minimum of 8 weeks before further experimentation and full reconstitution was confirmed by flow cytometric analysis.

LSK culture and expansion

Lineage-negative, KIT⁺, Sca-1⁺ cells were sorted utilizing a BD Symphony S6 sorter and cultured utilizing media optimized for HSC expansion and function⁶⁰. 50,000 cells/donor animal yields around 5 million LSK, allowing for 1 donor mouse per 5 recipient animals. The culture medium is composed of F12 media with 1% insulin-transferrin-selenium-ethanolamine, 1% Penicillin/Streptomycin/Glutamine (P/S/G), 10 mM HEPES, 100 ng/mL mouse recombinant TPO, 10 ng/mL mouse recombinant SCF and 87% hydrolyzed polyvinyl alcohol. Cells were cultured in fibronectin coated 96 well plates. After an initial period of 6 days, cells were fed with fresh media every 2 to 3 days and split when they reached ~80% confluency. LSK chimeras were generated as above with 10⁶ donor LSK cells via retro-orbital injection within 24 hours of irradiation.

T cell isolation and adoptive transfer

CD4⁺ and CD8⁺ T cells were isolated using mouse T Cell Isolation Kits from single-cell suspensions prepared from donor spleens. Recipient mice received 3 to 5 million cells per mouse via retro-orbital injection.

Bone marrow derived macrophage generation

Bone marrow derived macrophages (BMDMs) were generated by harvesting femurs and culturing of bone marrow precursor cells in L929 conditioned media⁹⁷ for 7 days, followed by interleukin-3 (IL-3) and CSF-1 (both at 5 ng/mL) for an additional 3 days in RPMI supplemented with 10% FBS, 1% Penicillin/Streptomycin/Glutamine, 10 mM HEPES, and 50uM Beta-mercaptoethanol (BME). RNA extraction was performed using the RNeasy mini kit with DNAse treatment.

Annexin V assay

In each well of a 6 well plate, 120,000 of parental MB49 or MB49 TNFR1KO were plated and treated with 100 ng/mL IFNγ (Thermo Fisher). After 24 hours, cells were treated with 100 ng/mL TNF (Thermo Fisher). 48h following TNF treatment, cells were harvested by incubation in 0.05% Trypsin, 0.02% EDTA in HBSS without calcium and magnesium and pooled with supernatant. Cells were stained for cell death with 1:4000 eFluor 506 fixable viability dye (Thermo Fisher), then with 5 μL BUV395-Annexin V (Thermo Fisher) in 100 μL Binding Buffer (Abcam).

IncuCyte assay

40,000 parental MB49 or MB49 TNFR1KO cells per well were plated in a 96 well plate and treated with 100 ng/mL IFNγ (Thermo Fisher). After 24 hours, cells were treated with 100 ng/mL TNF (Thermo Fisher) and 0.2 μM YOYO-1 (Thermo Fisher). Plates were incubated in an IncuCyte^® S3 Live-Cell Analysis System with imaging at 1-hour intervals for 48 hours.

PBMC-PIE

For each human sample, we prepared two conical tubes with RPMI (labeled as tube1 and tube2). We thawed frozen PBMCs in a 37°C water bath and transferred them into tube1. Subsequently, 10% of this suspension was transferred to tube2 for genotyping and sorting viable PBMCs into enriched CD34⁺ HSPC. Both tubes were centrifuged at 300g for 10 minutes. The resulting pellets were resuspended in MACS buffer. We combined the pellets from 6–8 samples into one tube and proceeded with CD34⁺ cell enrichment using CD34 microbeads (Miltenyi). To optimize the yield of HSPC, the MACS column was not washed, the CD34- fraction (flowthrough) was re-added to the column and then cells were removed from the magnet and eluted. The enriched cells collected in the conical tube were then centrifuged and resuspended in FACS staining buffer containing the following antibodies: FITC anti-CD34 (Miltenyi), Pacific Blue anti-CD49f (Biolegend, 1:200), PE anti-CD90 (Biolegend, 1:100), PE-Cy7 anti-CD38 (Biolegend, 1:100), APC-Cy7 anti-CD45RA (Biolegend, 1:400), and anti-lineage (cat number). Staining was performed for 30 minutes in the dark. Post-staining, cells in both tube1 (CD34⁺ enriched cells) and tube2 were resuspended in 7-AAD-containing MACS buffer for sorting. Initially, viable lineage-negative cells were sorted into a PCR tube, followed by sorting PBMCs from tube2 into the same PCR tube. The number of PBMCs sorted was determined based on the desired ratio of PBMC and CD34⁺ cells in the data.

Mouse progenitor enrichment

For bone marrow isolations, we harvested bone marrow from tibia and femurs, after RBC lysis, cells were stained for lineage markers (CD3, NK1.1, Gr-1, B220, Ter119), cKit and Sca-1. 90k viable, lineage-negative cells were sorted into a PCR tube, and then 10,000 lineage positive cells were sorted into the same PCR tube, allowing for a small representation of lineage positive cells in our dataset.

10X multiome library generation

Following sorting, we immediately proceeded with the 10X Multiome protocol. Nuclei isolation was conducted following the low-yield nuclei procedures in the appendix of the nuclei isolation protocol provided by 10X Genomics. The rest of the steps were performed as per the manufacturer’s manual. Sequencing libraries for ATAC-seq and RNA-seq were generated and sequenced using Novaseq6000.

Mouse antibody administration

For depleting or blocking experiments mice were treated 2 days before tumor cell challenge, and then every 2–3 days subsequently with 200ug of TNF blocking antibody (BE0058), 250ug of Ly6G depleting antibody (BE00775–1), 250ug of IFNAR1 blocking antibody (BE0241, which broadly inhibit all type I IFN signaling^63,65), 250ug of IFNγ blocking antibody (BE0055)⁶², or 250ug of CD4 depleting antibody (BE0003–1) and 250ug of CD8 depleting antibody (BE0061) per mouse administered IP. All antibodies were purchased from BioXCell.

Diphtheria toxin (DT) treatment

Recipients of bone marrow from ZBTB46-DTR mice were treated with DT 200ng per mouse diluted in PBS every 3 days

Bladder tumor CD45 single-cell RNA sequencing

Single cell suspensions of MB49-YFP bladder tumors were stained with a BUV395-CD45 antibody. Live YFP-negative BUV395-positive cells were sorted on a BD FACSAria Fusion cell sorter.

Bulk ATAC sequencing of mouse LSKs

For ATAC-seq, we followed the Omni-ATAC-seq protocol, working with 50,000 LSK cells sorted directly into PCR tubes.

BCG PCR

Bone marrow was centrifuged at 4000 RCF for 10 minutes, the pellet was resuspended in 1mL of 5% Triton-X100 in PBS and incubated at room temperature for 10 minutes. The sample was centrifuged at 10,000 RCF for 10 minutes. Genomic DNA was extracted, and PCR was performed using primers for the mycobacterial gene pknB.

Colony forming assay

Single cell suspensions were isolated from femurs of mice and counted on a Countess II. Cells were resuspended in 10X working stock of 2.5 ×10^5 cells/ml in IMDM +2% FBS media. 300ul was then added to 3mL complete MethoCult Media (StemCell Technologies). 1.1ml of media was dispensed using 3ml Syringe (StemCell Technologies) into one well of a meniscus free 6 well plate (StemCell Technologies) in technical duplicates. After 12 days colonies were quantified based on their morphology and technical duplicates were averaged and plotted.

Luminex assay

Blood was collected via terminal cardiac puncture, and serum was isolated by centrifugation. Luminex assay was performed on serum using the Mouse 48-plex ProcartaPlex kit (Thermo Fisher) according to manufacturer’s protocol with modifications as described below. Samples were added to the plate containing antibody-linked beads and incubated at 4°C overnight. Following overnight incubation, the plate was incubated at room temperature for 30 minutes with orbital shaking, then subsequent steps were performed per manufacturer’s protocol. Wash buffer was added to wells prior to loading on a Luminex 200 instrument. Each sample was read in duplicates, with a lower bound of 50 beads per sample per analyte.

Preprocessing of single-cell multiome sequenced data

The Cell Ranger ARC 2.0.2 pipeline was used for initial processing (sample demultiplexing, barcode processing, alignment of reads, counting of transcripts, cell filtering) of all human and mouse single-cell multiome data with the hg38 and mm10 reference genome.

Human single-cell multiome data processing and analysis

Starting from initial Cell Ranger filtered cells, we performed additional manual filtering per sample to ensure only high-quality cells remained in our data: iteratively embedding and clustering the data, removing clusters with poor quality-control (QC) metrics, and then re-embedding and clustering. RNA data was processed using Scanpy 1.9.3⁹⁸ (median and log normalization of counts, PCA, and UMAP), and ATAC data was processed using ArchR 1.0.1⁹⁹ (iterative LSI, UMAP). Clustering was run on the respective PCA and LSI matrices using PhenoGraph¹⁰⁰. Cluster QC metrics evaluated included standard Scanpy and ArchR-calculated metrics, DoubletDetection score, and mitochondrial and ribosomal fraction.

Data from separate samples was integrated without additional data harmonization. For the RNA modality, count matrices from each sample were concatenated into a full data matrix, which was then median and log-normalized. Ribosomal and mitochondrial genes were removed. The top 45 PCs were calculated using 1500 highly variable genes (HVG). We ran PhenoGraph clustering and UMAP using 30 nearest neighbors. For the ATAC modality, we applied ArchR’s implementation of iterative LSI to the tile matrix using 100,000 variable features and ran PhenoGraph on the 30 nearest neighbors in LSI embedding coordinates. Cells were annotated based on manual evaluation of PBMC marker gene expression in RNA clusters. These steps were performed first on cohort 1 and cohort 2 samples independently, and then on all samples combined.

Entropy analysis for the evaluation of single-cell multiome batch effect

We calculated entropy as described by¹⁰¹. We first constructed a k-NN(k=30) graph in RNA PCA space using euclidean distance. Given a cell 𝑖 and its 30 nearest neighbors, we computed the fraction of cells from each sample 𝑠 = 1, . . . ,10 as $p_i^s$. We find the Shannon entropy per cell as:

$$H_i=-\sum _{s=1}^{10}{p}_i^s{log}_2\left(p_i^s\right)$$

High entropy indicates that a cell’s neighborhood in RNA space is made up of a well-mixed set of samples, whereas low entropy indicates that nearby cells mostly come from the same sample.

Single-cell multiome/RNA differential gene expression

To determine differentially expressed genes post-BCG treatment, we employed MAST, a hurdle model that accounts for the many zero-counts in scRNA-seq data¹⁰². For each cell type, we fit the MAST model to log-normalized RNA counts of post vs. pre-treatment cells, returning a false discovery rate (fdr) and Natural log fold change, labeled as coefficient (Coef.), per gene per cell type. For mouse single cell RNA-seq analysis, we fit the mast model to long-normalized RNA counts of cells sorted from BCG vs PBS treated animals, returning a FDR and L2FC, per gene per cell type.

Due to an overrepresentation of female cells in our HSPC cluster, we found significant genes in MAST results for that cluster to be dominated by sex-linked genes. For HSPC, we ran MAST a second time excluding the genes XIST, TSIX, and all genes on the Y-chromosome. MAST results including all genes for HSPC can be found in Table S1. Gene set enrichment analysis was performed on notable genesets throughout this project using EnrichR¹⁰³, comparing a given set of genes to the GO Biological Process 2021 reference^104,105.

Chromvar motif accessibility

We constructed a reproducible peak set for each iteration of our human multiome dataset (cohort 1, cohort 2, and combined) using ArchR, grouping cells by ATAC PhenoGraph clusters before calling and merging peaks. We annotated motifs within peaks using the CISBP motif database and determined chromVAR score per cell using ArchR’s addBgdPeaks and addDeviationsMatrix functions.

To evaluate changes in motif accessibility post-treatment, we compared chromVAR scores in all post-treatment vs. pre-treatment cells for each motif and cell type using a Wilcoxon rank-sum test. Statistical significance of motifs was decided based on an p-value cutoff of 0.05. We also calculated the mean difference (MeanDiff) in chromVAR score post v. pre-treatment as the mean of cell scores for a given celltype and motif pre-treatment subtracted from the mean of scores post-treatment.

Bulk ATAC-seq data processing and analysis

ATAC-seq fastq files were processed using an in-house pipeline implemented in NextFlow¹⁰⁶ at https://github.com/michaelbale/jlabflow. Briefly, paired-end reads were trimmed for low-quality base-calls and adapter contamination using the Cutadapt¹⁰⁷ wrapper Trim Galore. Remaining reads were then mapped to mm10 using Bowtie2¹⁰⁸ with parameters “--no-mixed --no-unal –no-discordant –-local -–very-sensitive-local -X 1000 -k 4 –mm” retaining only properly mapped fragments with a MAPQ score of at least 30. Mitochondrial reads and improperly paired reads or secondary alignments were removed with Samtools¹⁰⁹ and Picard¹¹⁰ was used to remove duplicate fragments. Finally, we removed all mapped fragments that were associated with the ENCODE Forbidden list¹¹¹.

Genome Browsing Tracks

Genome browsing tracks were generated as bigwigs files using Deeptools bamCoverage¹¹² with reads per genomic content normalization using an effective genome size of 2648000000. For Figures 1H, 2I, bigwigs from individual replicates were averaged together using Deeptools bigwigAverage.

Analysis of ATAC-Seq Data

Peak calls for individual samples were made using Genrich¹¹³ in ATACseq mode (“-j”). Reproducible peaks within each treatment condition were determined using ChIP-r¹¹⁴ and optimal peak calls between conditions were merged to form an atlas of 25,245 total peaks. Reads in peaks were generated by Deeptools multiBamSummary and read into R v4.3.0 for differential analysis using DESeq2¹¹⁵ v1.40.2. Finally, motif bias analysis was performed using HOMER2 findMotifsGenome.pl¹¹⁶ with input peaks as peaks that were differentially accessible in BCG-treated LSK over PBS-treated (as defined by DESeq2 analysis) using differentially accessible peaks in PBS-treated LSK over BCG-treated as the custom background set (-bg).

Mouse single-cell multiome data processing and analysis

For the analysis of mouse single-nuclei multiome datasets, we employed R packages Seurat and Signac^117,118. We initiated the process by utilizing Cell Ranger outputs for filtered cells to create Seurat objects for each sample. Subsequently, we conducted manual cell filtering for each sample, eliminating cells with either excessively high or low fragment numbers or RNA counts per cell, along with cells exhibiting low TSS enrichment scores. We filtered out cells with RNA count less than 100, ATAC fragment count less than 1000, or cells with unusually high counts.

Next, we utilized the ATAC-seq assay within the Seurat object to call peaks for each sample using MACS2. These peaks were then combined using the ‘reduce’ function of GenomicRanges. Following this step, we generated peak count matrices once again for each sample and created a merged Seurat object. We then applied the standard Signac workflow, including TF-IDF normalization and SVD with default parameters. UMAP embeddings were generated from the first 50 dimensions obtained through the LSI reduction method. Finally, we computed nearest neighbors using the default settings of the Signac package.

Utilizing the RNA-seq assay of Seurat objects for each sample, we created another merged Seurat object, which was then divided into layers by sample using the ‘split’ function. Standard Seurat preprocessing workflow steps such as normalization, scaling, and PCA were carried out. The split layers were integrated using ‘integrateLayers,’ resulting in a new dimensional reduction labeled ‘integrated.cca.’ The layers were subsequently rejoined using the ‘JoinLayers’ function within the Seurat package. We did not further process the RNA-seq dataset for UMAP embedding and clustering analysis.

For motif analysis, we incorporated motif information into the merged object using the ‘AddMotifs’ function in Signac. Motif information for mm10 was obtained from the JASPAR2020 database. Additionally, we added per-cell TF motif activity scores using chromVAR with the ‘RunChromVAR’ function in Signac as a separate assay to the object.

Due to the limited depth of the RNA-seq dataset, we were unable to derive meaningful clusters based on transcriptome data. Consequently, we relied on the cluster information derived from the ATAC-seq assay to identify cell types. When annotating each cluster, we referenced the cell type calling results from SingleR package¹¹⁹ , and with the expression of cluster marker genes and major cell type-specific chromVAR TF activity. We avoided integration and in-depth analyses of snRNA because we observed low read depth in our snRNA libraries, a common feature of bone marrow multiome; communications with 10X Genomics.

To analyze differential TF activity (chromVAR scores) across groups, we utilized the ‘FindMarkers’ function in Seurat, applying the Wilcoxon test. For differential gene expression analysis, the Wilcoxon test was also employed, enabling us to generate volcano plots that highlight significantly differentially expressed genes with adjusted p-values less than 0.05. To correlate differential gene expression with differential chromVAR TF activity, we employed the MAST¹¹⁸ test to identify genes that were significantly differentially expressed

Human-mouse differential gene expression comparison

Human-mouse orthologous genes were matched based on the HGNC Comparison of Orthology Predictions (HCOP) search tool. Not all matches were one-to-one; a number of mouse genes in our data were mapped to multiple human orthologs. For each gene-gene pair, we compared differential expression results from mouse and human single-cell multiome datasets: specifically, comparing MAST Coef. in humans to Log2FC in mouse.

Mouse bladder single-cell RNA processing and analysis

Starting from initial Cell Ranger filtered cells, we filtered heuristically based on distributions of QC metrics per sample, eliminating cells with either excessively high or low RNA counts per cell or excessively high mitochondrial RNA content. RNA from high quality cells was processed, CPM normalized, and log transformed per biological sample with Scanpy 1.9.3. All samples were integrated without additional data harmonization, embedded, and clustered (clusters with high doublet scores were removed). Scanpy was used to identify 2000 highly variable genes, which were used to calculate the top 50 PCs, which were used to calculate the nearest neighbors distance matrix. Cells were clustered and visualized using Scanpy’s implementation of the Leiden algorithm and UMAP from the nearest neighbor’s distance matrix. Cells were then annotated based on manual evaluation of marker gene expression in unsupervised clusters.

T1/T2/T3 neutrophil assignment

Scanpy score genes function was used to score neutrophils on published gene sets for T1, T2, T3 neutrophils⁶⁹. Cells were identified as T1/T2/T3 based on which neutrophil subtype score was highest, and any cell with a gene score below 0.5 for all gene sets was classified as ‘Other’.

Quantification and Statistical Analysis:

Flow cytometry-based analysis

We first performed the Shapiro-Wilks test for normality, and confirmed the populations were normally distributed. For experiments involving two groups (relevant to figures 2C, 2D, 3B, 4C, 4E, 5C, 5D, 5H, S5D). We then performed a Students T-test. For Figures 4C, and 5D we utilized a paired analysis, whereas the rest were unpaired. For experiments with multiple groups (relevant to figures 3G, 3H, 5G, S2D, S4D, S4H, S5E, and S6B, S6E) we performed a one-way Anova. Normality tests and P-values are provided in Table S2.

Serum Cytokine analysis

We first performed the Shapiro-Wilks test for normality, and confirmed the populations were normally distributed (relevant to figures 2J, S3I). We then performed a Students T-test. Normality tests and P-values are provided in Table S2.

TumGrowth statistical analysis of tumor growth curves:

This analysis as developed by the Kroemer Laboratory¹²⁰ utilizes an automatic analysis of breakpoints across the tumor growth curve up to the point of the first mouse death and includes automatic outlier detection using Bonferroni-corrected p-values calculated from residuals. The tumor growth curves are then subjected to type II ANOVA and pairwise comparisons across groups. Additionally, this analysis provides cross-sectional analysis and calculates p-values by the Wilcoxon rank sum test. The p-values reported utilize the Holm method for multiple hypothesis correction testing where multiple groups are tested.

Supplementary Material

1

Table S1: Human single cell multi-ome MAST differential analysis results. Related to Figure 1.

2

Table S3: Clinicopathological information of human cohorts 1 and 2. Related to STAR Methods.

4

Table S2: Supplementary Statistical Analyses. Related to STAR Methods.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
CD103 APC-eFluor780	eBioscience	Cat#47–1031-82
CD103 BV421	BD Bioscience	Cat#562771
CD115 BV605	BD Bioscience	Cat#743640
CD11b PE-Cy7	BD Bioscience	Cat#561098
CD11c PE-Cy5	BioLegend	Cat#117316
CD11c R718	BD Bioscience	Cat#567076
CD127 BV711	BD Bioscience	Cat#565490
CD135 (Flt3) BV421	BD Bioscience	Cat#562898
CD14 APC-Fire750	BioLegend	Cat#123331
CD150 BV786	BD Bioscience	Cat#567518
CD152 (CTLA-4) PE-CF594	BD Bioscience	Cat#564332
CD16/32 - APC	BD Bioscience	Cat#558636
CD172a BV711	BD Bioscience	Cat#740766
CD201 (EPCR) APC-eFluor780	eBioscience	Cat#46–2012-82
CD206 PE	BD Bioscience	Cat#568273
CD27 BV510	BD Bioscience	Cat#563605
CD279 (PD-1) BV786	BD Bioscience	Cat#744548
CD3 BUV615	BD Bioscience	Cat#751418
CD317/BST-2 BV615	BD Bioscience	Cat#751038
CD34 PE	BioLegend	Cat#152204
CD4 BV510	BD Bioscience	Cat#563106
CD44 BUV496	BD Bioscience	Cat#741057
CD45.1 BUV737	BD Bioscience	Cat#612811
CD45.2 BUV395	BD Bioscience	Cat#564616
CD48 - PerCP-Cy5.5	BioLegend	Cat#103422
CD62L - PerCP-Cy5.5	BD Bioscience	Cat#560513
CD69 BV605	BD Bioscience	Cat#563290
CD8 APC	BD Bioscience	Cat#553035
CD80 BV650	BD Bioscience	Cat#563687
CD86 BB700	BD Bioscience	Cat#742145
F4/80 BV510	BD Bioscience	Cat#743280
FoxP3 PE-Cy7	Tonbo	Cat#60–5773-U025
GranzymeB eFluor450	eBioscience	Cat#48–8898-82
IFNg FITC	BD Bioscience	Cat#554411
Ki67 R718	BD Bioscience	Cat#566963
CD117 (c-KIT) PE-Cy7	BioLegend	Cat#105814
Biotin Mouse Lineage Panel (CD3, NK1.1, B220, Ter-119)	BD Bioscience	Cat#559971
BV 510 anti-human Lineage Cocktail (CD3, CD14, CD16, CD19, CD20, CD56)	BioLegend	Cat#348807
Ly-6C BV785	BioLegend	Cat#128041
Ly-6G PE-CF594	BD Bioscience	Cat#562700
MHCII (I-A/I-E) BUV615	BD Bioscience	Cat#751570
Perforin PE	BioLegend	Cat#154306
SCA1 PE-eFluor610	eBioscience	Cat#61–5981-82
Streptavidin FITC	BD Bioscience	Cat#554060
TNFa BV711	BD Bioscience	Cat#563944
XCR1 APC	BioLegend	Cat#148206
CD45 BUV395	BD Bioscience	Cat#563792
CD45 BUV661	BD Bioscience	Cat#612975
CD90.2 FITC	Biolegend	Cat#140304
NK1.1 FITC	Invitrogen	Cat#11–5941-85
CD19 FITC	Invitrogen	Cat#11–0193-82
TCRgd FITC	Invitrogen	Cat#11–5711-82
CD11b APC-Cy-7	BD Bioscience	Cat#557657
Ly6G Percp Cy5.5	Biolegend	Cat#127616
Ly6c BUV605	BD Bioscience	Cat#563011
MHCII BUV737	BD Bioscience	Cat#748845
Siglec F FITC	Biolegend	Cat#155504
CD90.2 BV786	BD Bioscience	Cat#564365
CD11c APCR700	BD Bioscience	Cat#565872
CD34 FITC	Miltenyi	Cat#130–113-178
CD49f Pacific Blue	Biolegend	Cat#313620
CD90 PE	Biolegend	Cat#328110
CD38 PE-Cy7	Biolegend	Cat#303516
CD45RA APC	Biolegend	Cat#304128
Anti-mouse TNF	BioXcell	Cat#BE0058
Anti-mouse Ly6G	BioXcell	Cat#BE00775–1
Anti-mouse IFNAR1	BioXcell	Cat#BE0241
Anti-mouse IFNγ	BioXcell	Cat#BR0055
Anti-mouse CD4	BioXcell	Cat#BE0003–1
Anti-mouse CD8	BioXcell	Cat#BE0061
Bacterial and virus strains
BCG Pasteur	This study	NA
Biological samples
PBMCs from patients with NMIBC at Memorial Sloan Kettering Cancer Center(n=8)	This study	NA
PBMCs from patients with NMIBC at McGill University (n=13)	This study	NA
Chemicals, peptides, and recombinant proteins
G418	Thermo Fisher	Cat#10–131-035
RPMI with 10% FBS and 2mM L-glutamine	Sloan Kettering Institute media preparation facility	NA
0.05 Tripsin, 0.02 EDTA in HBSS	Sloan Kettering Institute media preparation facility	NA
F12 media	Sigma Aldrich	Cat#11–765-062
IMDM media	Thermo Fisher	Cat#31980030
MethoCult media	StemCell Technologies	Cat#M3434
Middlebrook 7H9 broth	Thermo Fisher	Cat#DF0713–17-9
Bovine Serum Albumin	Sigma	Cat#3116964001
Dextrose	Thermo Fisher	Cat#BP350–1
Sodium Chloride	Thermo Fisher	Cat#BP358–10
Glycerol	Thermo Fisher	Cat#AAA16205AP
Tween 80	Sigma	Cat#P8074
Phosphate buffered saline (PBS)	Sloan Kettering Institute media preparation facility	NA
7H10 agar	Thermo Fisher	Cat#DF0627174
Poly-L-lysine	Sigma	Cat#P4707
Cell stimulation cocktail	Thermo Fisher	Cat#00–4975-03
Foxp3 fixation/permeabilization buffer	Thermo Fisher	Cat#50–112-8857
Fixable blue dead stain kit	Thermo Fisher	Cat#L34962
eFluor 506 fixable viability dye	Thermo Fisher	Cat#65–0866-14
DAPI	Thermo Fisher	Cat#D1306
BUV395-Annexin V	Thermo Fisher	Cat#564871
Annexin V binding buffer	Abcam	Cat#ab14084
YOYO-1 Dimeric Cyanine Nucleic Acid Stain	Thermo Fisher	Cat#Y3601
Insulin-Transferrin-Selenium-Ethanolamine	Thermo Fisher	Cat#51500056
Penicillin/streptomycin/glutamine	Thermo Fisher	Cat#10378016
HEPES	Thermo Fisher	Cat#15630080
MACS buffer	Miltenyi	Cat#130–091-376
Mouse recombinant TPO	Thermo Fisher	Cat#315–14
Mouse recombinant SCF	Thermo Fisher	Cat#250–03
87% hydrolyzed polyvinyl alcohol	Sigma	Cat#363081
Fibronectin Coated 96 well plates	Corning	Cat#354409
IL-3	Thermo Fisher	Cat#213–13
CSF-1	Thermo Fisher	Cat#315–02
Beta-mercapto-ethanol	Thermo Fisher	Cat#31350010
Lipofectamine 2000	Thermo Fisher	Cat#11668027
Mouse recombinant IFNγ	Thermo Fisher	Cat#315–05
Mouse recombinant TNF	Thermo Fisher	Cat#315–01A
7-AAD	Thermo Fisher	Cat#A1310
Diphtheria toxin	Sigma	Cat#D0564
Triton X-100	Sigma	Cat#T8787
Critical commercial assays
MycoAlert Plus	Lonza	Cat#LT07–710
Mouse T cell isolation kit	Miltenyi	Cat#130–095-130
CD34 MicroBead Kit, human	Miltenyi	Cat#130–046-702
MACS columns	Miltenyi	Cat#130–042-201
Mouse 48-plex ProcartaPlex kit	Thermo Fisher	Cat#EPX480–20834-901
RNeasy Mini Kit	Qiagen	Cat#74104
RNase-Free DNase Set	Qiagen	Cat#79254
Deposited data
Human PBMC single cell Multiomics cohort 1 sequencing code	This study	https://zenodo.org/records/14046727
Human PBMC single cell Multiomics cohort 2 sequencing code	This study	https://zenodo.org/records/14046727
Mouse bone marrow ATAC-sequencing code	This study	https://zenodo.org/records/14046727
Mouse bone marrow single cell RNA sequencing code	This study	https://zenodo.org/records/14046727
Mouse tumor CD45+ single cell RNA sequencing code	This study	https://zenodo.org/records/14046727
Human PBMC single cell Multiomics cohort 1 sequencing Raw Data	This study	GEO Accession Number: GSE295309
Human PBMC single cell Multiomics cohort 2 sequencing Raw Data	This study	GEO Accession Number: GSE295309
Mouse bone marrow ATAC-sequencing Raw Data	This study	GEO Accession Number: GSE295309
Mouse bone marrow single cell RNA sequencing Raw Data	This study	GEO Accession Number: GSE295309
Mouse tumor CD45+ single cell RNA sequencing Raw Data	This study	GEO Accession Number: GSE295309
Experimental models: Cell lines
MB49-luciferase	Yi Luo, University of Iowa	NA
B16	Taha Merghoub	NA
MB49 TNFR1KO	This study	NA
MB49-luciferase YFP	This study	NA
Experimental models: Organisms/strains
C57BL/6J	The Jackson Laboratory	Cat#6664
B6 CD45.1	The Jackson Laboratory	Cat#2014
OT-I	The Jackson Laboratory	Cat#3831
OT-II	The Jackson Laboratory	Cat#4194
ZBTB46-DTR	The Jackson Laboratory	Cat#19506
Oligonucleotides
TNFR1 gRNA CGGGCCTCCACCGGGGATAT	IDT	NA
pknB_F GGACCAGAGCCAACGATGATG	IDT	NA
pknB_R AAACTGACTGCCGCCGGATTC	IDT	NA
Recombinant DNA
pSpCas9(BB)-2A-GFP (PX458)	Addgene	Cat#48138
Software and algorithms
R studio	Posit	RRID:SCR_000432
Prism	GraphPad	RRID:SCR_002798
Flowjo	BD Bioscience	RRID:SCR_008520
FACS DiVa	BD Bioscience	RRID:SCR_001456
SpectroFlo	Cytek	RRID:SCR_025494
Cell ranger ARC 2.02	10x Genomics	RRID:SCR_023897
Scanpy 1.9.3	NumFOCUS	RRID:SCR_018139
ArchR 1.01	Greenleaf Lab	https://www.archrproject.com/
PhenoGraph	Pe’er Lab	RRID:SCR_016919
EnrichR	Ma’ayan Lab	RRID:SCR_001575
DeSeq2 v1.40.2	Bioconductor	RRID:SCR_015687
MAST	Bioconductor	RRID:SCR_016340
Seurat	Satija Lab	RRID:SCR_016341
Signac	Satija Lab	RRID:SCR_021158
ChromVAR	Greenleaf lab	RRID:SCR_026570
Cutadapt	OMICtools	RRID:SCR_011841
Genrich	John M. Gaspar	https://github.com/jsh58/Genrich
Homer2	NITRC	RRID:SCR_009586
Bowtie2	Debian	RRID:SCR_016368
Samtools	OMICtools	RRID:SCR_002105
Picard	OMICtools	RRID:SCR_006525
Other
LSRFortessa flow cytometer	BD biosciences	NA
Aurora full spectrum analyzer	Cytek	NA
FACSymphony S6 sorter	BD biosciences	NA
FACSAria cell sorter	BD biosciences
NovaSeq 6000 Sequencing System	Illumina	NA
Polyurethane I.V. Catheter 24G x 3/4”	Santa Cruz	Cat#sc-360126
3mL syringe	StemCell Technologies	Cat#28240
6-well plate	StemCell Technologies	Cat#27370