Reprogramming intratumoral T_reg cells by morpholino-mediated splicing of FOXP3 for cancer immunotherapy

Abstract

Regulatory T (T_reg) cells represent a primary barrier to the development of effective anti-tumor immunity. Here, we report that reprogramming T_reg cells by shifting the expression of FOXP3 from its full-length isoform (FOXP3^FL) to a short isoform with exon 2 skipped (FOXP3^dE2) promotes CD8 T cell-mediated antitumor immunity. FOXP3^dE2 mRNA expression in triple negative breast cancer tissue positively correlated with overall patient survival. Mice expressing only the FOXP3^dE2 isoform were resistant to the development of multiple types of tumors. Tumor-infiltrating T_reg cells expressing the FOXP3^dE2 isoform exhibited lower immunosuppressive activity and promoted CD8 T cell activation. Additionally, we designed a morpholino oligo to induce FOXP3 exon 2 skipping, which similarly enhanced antitumor activity in mouse tumor models and the killing capacity of autologous tumor-infiltrating T cells against patient-derived tumor organoids. Our results suggest that promoting FOXP3^dE2 expression reprograms T_reg cells to T helper-like cells, thereby enhancing antitumor immunity.

One sentence summary:

T_reg cells expressing a short FOXP3 isoform exhibit a Th-like phenotype with reduced suppressive function and mediate antitumor immunity.

INTRODUCTION

The emergence of cancer immunotherapy, including immune checkpoint inhibitors and adoptive cell transfer, has revolutionized cancer treatment in the past decade. The success, however, is limited to a relatively small subset of patients and cancer types (1–3). As a major type of immune suppressive cells, regulatory T (T_reg) cells play a central role in maintaining immune homeostasis and negatively regulate immune-mediated inflammation such as autoimmune disease, asthma and allergy. However, tumor-infiltrating T_reg cells also suppress effective immunity against tumors. Identification of factors that influence T_reg homeostasis and function is important for the development of T_reg-targeting therapies to facilitate antitumor immunity (4, 5).

Strategies for depleting intratumoral T_reg cells have shown promising antitumor results, but systemic T_reg cell ablation also results in severe and even lethal autoimmunity in patients (6–10). Although previous studies have shown feasibility of selectively targeting intratumoral T_reg cells while leaving the T_reg cells in the periphery functionally intact (11–13), such approaches have not been translated into clinical practice. FOXP3 is a master regulatory gene that governs the molecular and functional features of T_reg cells. Whereas mouse Foxp3 gene only encodes the FOXP3^FL isoform, the human FOXP3 gene encodes two major isoforms by alternative splicing: a full-length isoform (FOXP3^FL) and a shorter isoform lacking exon 2 (FOXP3^dE2) (Fig. 1A), with FOXP3^FL representing about half of the total FOXP3 expressed (14). Both FOXP3 isoforms are effective in directing T_reg cell differentiation and function in vitro under forced overexpression (15–18). However, the ratio of these two isoforms is skewed towards FOXP3^dE2 in autoimmune diseases (19–22), suggesting FOXP3^dE2 may have a bona fide pro-inflammatory function rather than the immunosuppressive activities of its counterpart FOXP3^FL.

Figure 1. FOXP3^dE2 mRNA expression correlates with survival and tumor immune responses in patients with triple negative breast cancer.

(A) Human FOXP3 gene encodes two major isoforms through alternative splicing: a longer isoform (full length FOXP3, FOXP3^FL) contains all 11 coding exons and a shorter isoform lacking exon 2 (FOXP3^dE2). Black box: translated region; Grey box: untranslated region. (B) Kaplan-Meier survival curves of overall survival for patient groups with high (> median) or low (≤ median) FOXP3^dE2 mRNA expression using TCGA datasets. p values were determined by log-rank test. (C) Kaplan-Meier survival curves of overall survival for patient groups with high (> median) or low (≤ median) FOXP3^FL mRNA expression using TCGA datasets. p values were determined by log-rank test. (D) Volcano plot of differentially expressed genes in TNBC samples with high (> median) versus low (≤ median) FOXP3^dE2 mRNA expression. Blue: genes with down-regulated expression, Red: genes with up-regulated expression. The analysis cutoff was set to p-value < 0.05 and |log₂FC| > 0.25. (E) Gene set enrichment analysis of pathways in TNBC samples with high (> median) versus low (≤ median) FOXP3^dE2 mRNA expression. NES, normalized enrichment score. Fisher exact test was used to calculate p-values. (F) The relative proportions of immune cells in human TNBC tumors were evaluated using bulk RNA sequencing data of TNBC samples from TCGA and analyzed by MCP-counter. Statistical significance was determined by unpaired two-tailed student’s t test. (G) The correlation of the infiltration of cytotoxic lymphocytes (determined by MCP-counter) and the expression of FOXP3^dE2 mRNA in TCGA TNBC cohort.

Genetically engineered mouse models expressing only the FOXP3^dE2 isoform were recently generated to study the function of FOXP3^dE2. Foxp3^dE2 mice are viable and fertile (23, 24) but can develop a mild systemic lupus erythematosus (SLE)-like disease (23). By contrast, Foxp3 deletion or loss-of-function mutant mice develop a fatal lymphoproliferative autoimmune disease shortly after birth (25). Furthermore, it was noted that several rare patients expressing only the FOXP3^dE2 isoform are relatively healthy or develop autoimmunity with varying severity (23, 26). Transfer of FOXP3^dE2-expressing T_reg (hereafter dE2 T_reg) cells into Tcrb^−/− mice that have B cells but not T helper cells is sufficient to induce autoantibodies, suggesting that the dE2 T_reg cells may function as T helpers to help autoantibody-producing B cells (23). These preliminary findings suggest indispensable functions of the FOXP3 exon 2 region. The fundamental differences between FOXP3^FL- and FOXP3^dE2-expressing T_reg cells indicate a functional reprogramming of T_reg cells when they only express the FOXP3^dE2 isoform. In the clinic, targeting T_reg cells by FOXP3 silencing or complete cell depletion has been challenging due to the severe toxicity caused by the resulting imbalanced immune homeostasis. In this study, we reason that reprogramming T_reg cells, by shifting FOXP3^FL to FOXP3^dE2 expression, may provide an approach that reduces these adverse effects while promoting the tumor-killing activity of T effectors.

RESULTS

FOXP3^dE2 mRNA expression in breast cancer correlates with better patient survival

FOXP3 directs gene expression responsible for the development and maintenance of immunosuppressive functions of T_reg cells. To explore the function of FOXP3^dE2 in breast cancer immunology, we analyzed clinical data from triple-negative breast cancer (TNBC) patients in The Cancer Genome Atlas (TCGA) (27). TNBC patients with high (greater than the median) expression level of FOXP3^dE2 mRNA in the tumors displayed significantly (p= 0.031) better overall survival compared to those with low (less than or equal to the median) levels of FOXP3^dE2 (Fig. 1B), while levels of FOXP3^FL mRNA did not correlate with overall survival (Fig. 1C). The FOXP3^dE2-high group displayed increased levels of expression in genes related to tumor immune responses and reduced expression of risk factors (Fig. 1D). Gene set enrichment analysis (GSEA) demonstrated enrichment of upregulated genes in cytokine–cytokine receptor interaction, autoimmune thyroid disease, antigen processing and presentation, NK cell-mediated cytotoxicity, JAK-STAT signaling, and TCR/BCR signaling, whereas down-regulated genes were associated with oxidative phosphorylation, a FOXP3-regulated function important for T_reg proliferation and immune suppression in the tumor microenvironment (TME) (28, 29) (Fig. 1E). To further characterize the tumor immune microenvironment, Microenvironment Cell Populations-counter (MCP-counter) (30) was applied to quantify eight immune and two stromal cell subpopulations from the TCGA TNBC transcriptomic data. We observed higher putative tumor infiltration of B lineage cells, CD8 T cells, cytotoxic lymphocytes, monocytic lineage cells, myeloid dendritic cells, NK cells, and T cells in the FOXP3^dE2-high group (Fig. 1F). Moreover, the computed proportion of cytotoxic lymphocytes positively correlated with the expression of FOXP3^dE2 (Fig. 1G). These data suggest that FOXP3^dE2 expression is associated with effective anti-tumor immunity.

To further determine whether FOXP3^dE2 mRNA expression predicts favorable prognosis across cancers, we analyzed survival data from 27 TCGA cancer types. Overall, FOXP3^dE2 mRNA correlated with better survival in 17 cancer types and showed significant protection (p < 0.05) in 6 of them (Supplementary Fig. 1A). Conversely, FOXP3^FL expression was associated with better survival in 14 cancer types and showed significant protection in 2 of them (Supplementary Fig. 1B). We also observed that FOXP3 expression can be a risk factor: FOXP3^dE2 expression in 10 cancer types and FOXP3^FL expression in 13 cancer types was associated with worse patient survival (Supplementary Fig. 1).

Mice bearing the Foxp3^dE2 allele reject EO771 triple negative breast cancer

To study the anti-tumor function of the FOXP3^dE2 isoform, we used a lab-generated mouse strain in which exon 2 of the Foxp3 gene was deleted (Foxp3^dE2). These mice are viable, morphologically normal, and exhibit normal thymic T_reg development (23). However, their T_reg cells express higher levels of effector cytokines, are less stable, and can transdifferentiate into helper T cells, leading to lupus-like autoimmunity. Compared to WT mice, the Foxp3^dE2 mice rejected EO771 triple negative breast tumor orthotopically implanted into the 4th mammary fat pad (Fig. 2A). We harvested tumors 12 days after inoculation to characterize tumor-infiltrating lymphocytes (TIL) by flow cytometry. In tumors from WT mice, less than 10% of live CD45⁺ cells were CD8 T cells, while about 30% of CD45⁺ cells were CD8 T cells in tumors from Foxp3^dE2 mice (Fig. 2B). CD4 T cell frequencies were similar between groups, but the proportion of tumor-infiltrating T_reg cells was lower in Foxp3^dE2 mice, with reduced CD25 expression (Fig. 2C). The percentage of dE2 T_reg cells expressing IFNγ and IL-2 was higher than tumor-infiltrating T_reg cells in WT mice (Fig. 2D). Interestingly, a lower proportion of dE2 T_reg cells than WT T_reg cells expressed KLRG1 (Fig. 2E), a marker linked to potent suppressive activity and enhanced inhibition of antitumor immune responses (31–34). The majority of IFNγ⁺ dE2 T_reg cells were KLRG1⁻ (Fig. 2E). Likewise, a higher proportion of conventional CD4 T cells infiltrating EO771 tumors in Foxp3^dE2 mice expressed IFNγ (Fig. 2F), as were CD8 T cells expressing IFNγ and granzyme B (GZMB) (Fig. 2G). Similar results were also obtained in mouse models of Lewis lung carcinoma (LL/2) and colorectal cancer (MC38) (Supplementary Fig. 2 and 3), indicating that the FOXP3^dE2 isoform provides an antitumor function in multiple types of cancer.

Figure 2. Foxp3^dE2 mice reject orthotopically implanted EO771 breast tumors.

EO771 cells (1×10⁶ cells) were orthotopically implanted into the fourth mammary fat pad of mice. Since the tumors completely disappeared in Foxp3^dE2 mice 2 weeks after tumor cell implantation, immunophenotyping of tumor-infiltrating lymphocytes was performed 7–10 days after tumor inoculation. (A) Tumor growth curve in wildtype (WT) mice or Foxp3^dE2 mice. (B) Frequencies of tumor-infiltrating CD4 and CD8 T cells. (C) Percentage of T_reg cells in tumor-infiltrating CD4 T cells and expression of CD25 in T_reg cells. (D) Expression of IFNγ and IL-2 in tumor-infiltrating T_reg cells. (E) KLRG1 expression and its relation to IFNγ expression in tumor-infiltrating T_reg cells. (F) IFNγ expression in tumor-infiltrating CD4 T cells. (G) IFNγ and granzyme B (GZMB) expression in tumor-infiltrating CD8 T cells. Data represent mean ± SD (n = 4 mice per group) from one of three or more independent experiments. ns: not significant; *: p < 0.05; **: p < 0.01; ****: p < 0.0001 by two-way ANOVA with Sidak’s test for multiple comparisons (A) or two-tailed unpaired t-test (B – G).

dE2 T_reg cells promote CD8 T cell-mediated antitumor immunity

Our immunophenotyping results suggested that tumor-infiltrating dE2 T_reg cells are dysfunctional, allowing more efficient antitumor immunity in Foxp3^dE2 mice. Given that KLRG1 expression is associated with T_reg suppressive function (32, 34) and fewer tumor-infiltrating dE2 T_reg cells express KLRG1 (Fig. 2E, Supplementary Fig. 2E and 3E), we performed in vitro suppressive assays with tumor-infiltrating T_reg cells, as previously described (11). Since Foxp3^dE2 mice completely rejected EO771 mouse TNBC and we were not able to obtain enough tumor-infiltrating T_reg cells, we purified tumor-infiltrating T_reg cells from subcutaneously implanted MC38 colorectal tumors. We previously showed that both WT and dE2 T_reg cells from lymphoid tissues suppressed ~80% of cell proliferation at 1:1 T_reg-to-responder ratio (23). Conversely, we found that tumor-infiltrating WT T_reg cells had superior suppressive activity and at 1:2 T_reg-to-responder ratio suppressed proliferation of congenically marked (Thy1.1) CD4 T cells by 80% (Fig. 3A). However, tumor-infiltrating dE2 T_reg cells only suppressed 55% of responder cell proliferation at the same ratio (Fig. 3A). The data collectively demonstrate reduced immunosuppressive activity of dE2 T_reg cells compared to WT T_reg cells.

Figure 3. dE2 T_reg cells promote CD8 T cell-mediated antitumor immunity.

(A) In vitro suppressive activity of tumor-infiltrating T_reg cells. Since EO771 tumors did not grow in Foxp3^dE2 mice, T_reg cells were purified from MC38 colorectal tumors implanted into WT mice (WT T_regs) and Foxp3^dE2 mice (dE2 T_regs). (B) CD8 T cell depletion reversed dE2 T_reg mediated antitumor immunity. Foxp3^dE2 mice orthotopically implanted with EO771 cells (1×10⁶ cells, n=4–5 mice per group). Mice were administrated i.p. of isotype control, 10 mg/kg anti-CD4, 10 mg/kg anti-CD8, 12.5 mg/kg anti-NK1.1, or a combination of 15 mg/kg anti-CD19 and 15 mg/kg anti-CD20 antibodies beginning two days before tumor inoculation and thereafter twice weekly for a total of 6 doses. (C) Effects of WT and dE2 T_reg cells on CD8 T cell-mediated antitumor immunity. Tcrb^−/− mice were injected intravenously with PBS, or CD8 T cells (2×10⁶ cells) isolated from wildtype mice, or a combination of T_reg cells (5×10⁵ cells) isolated from WT or Foxp3^dE2 mice. Mice then were orthotopically implanted with EO771 cells (1×10⁶ cells, n=5 mice per group). (D) The effect of dE2 T_reg cells on antitumor immunity in WT mice. BoyJ mice (CD45.1) were orthotopically implanted with EO771 cells (1×10⁶ cells, n=5 mice per group). 12 days after tumor inoculation (arrow), purified WT T_reg cells or dE2 T_reg cells (3×10⁵ cells, CD45.2) were injected into the tumors in BoyJ mice. (E) Frequency of tumor-infiltrating CD4 and CD8 T cells. (F) IFNγ and GZMB expression in tumor-infiltrating CD8 T cells. Data represent mean ± SD from one of two independent experiments. ns: not significant; *: p < 0.05; **: p < 0.01; ***: p < 0.001; ****: p < 0.0001 by two-way ANOVA with Sidak’s multiple comparisons test (A & E) or Turkey’s multiple comparisons test (B - D), or one-way ANOVA with Turkey’s multiple comparisons test (F).

In addition to reduced KLRG1 expression and diminished suppressive function, tumor-infiltrating dE2 T_reg cells expressed higher levels of IFNγ and IL-2. We speculate that dE2 T_reg cells in the TME might promote antitumor immunity. To determine which immune cell type is essential for the antitumor effect in Foxp3^dE2 mice, we selectively depleted mice of CD8 T cells, CD4 T cells, B cells, or NK cells using depleting antibodies administered intraperitoneally before tumor inoculation and every 3 days thereafter until tumors reached 1,000 mm³ or mice became moribund. Effective depletion was confirmed by flow cytometry (Supplementary Fig. 4A&B). The antitumor effect in Foxp3^dE2 mice was substantially compromised upon CD8 T depletion, but not with depletion of CD4 T cells, B cells, or NK cells (Fig. 3B), suggesting that CD8 T cells are critical for dE2 Treg mediated antitumor immunity. Next, to determine whether dE2 T_reg cells promote antitumor immunity, we isolated CD8 T cells from WT mice and T_reg cells from either WT or Foxp3^dE2 mice, intravenously infused them into Tcrb^−/− mice that lack αβ T cells before EO771 tumor inoculation. While CD8 T cell transfer alone reduced tumor burden, co-transfer of WT T_reg cells restored tumor growth to control levels (Fig. 3C). Strikingly, co-transfer of dE2 T_reg cells with CD8⁺ T cells further reduced tumor burden compared to CD8⁺ T cells alone, indicating that dE2 T_reg cells enhance rather than suppress CD8⁺ T cell–mediated antitumor responses (Fig. 3C).

To further verify the antitumor effect of dE2 T_reg cells, we established EO771 tumors in WT BoyJ mice (CD45.1) and intratumorally injected 0.3 million WT or dE2 T_reg cells (CD45.2), purified from the spleens of WT or Foxp3^dE2 mice, 12 days post tumor inoculation. Mice receiving dE2 T_reg cells developed smaller tumors compared to those receiving PBS or WT T_reg cells (Fig. 3D). Although congenic markers were used to distinguish donor (CD45.2⁺) from host T_reg cells, donor T_reg cells were largely undetectable in tumors harvested two weeks after transfer. The small population of CD45.2⁺ cells shown (Supplementary Fig. 5A) was from one of only two tumors with detectable donor cells, limiting our ability to phenotype them. Nevertheless, there were still more CD8 T cells in the tumors that received dE2 T_reg cells (Fig. 3E) and significantly higher percentages of CD8 T cells infiltrating tumors that received dE2 T_reg cells expressed IFNγ and GZMB compared to the PBS control group (Fig. 3F). Among tumor-infiltrating T_reg cells, which were predominantly host-derived WT T_reg cells in all groups, no significant differences were observed in CD25, IFNγ, or KLRG1 expression (Supplementary Fig. 5B). Likewise, no differences in IFNγ expression were detected in tumor-infiltrating conventional CD4 T cells (Supplementary Fig. 5C). Together, these findings indicate that dE2 T_reg cells promote activation and accumulation of tumor-infiltrating CD8 T cell to enhance antitumor immunity.

Tumor-infiltrating dE2 T_reg cells exhibit T helper-like phenotypes

To investigate how expression of the FOXP3^dE2 isoform in T_reg cells promotes antitumor immunity, we isolated tumor-infiltrating T cells from EO771 tumors and performed single-cell RNA sequencing (scRNA-seq) analysis to evaluate the intrinsic and extrinsic impact of FOXP3^dE2 isoform on T_reg cells and other T cells in the TME. Clustering based on differentially expressed genes and UMAP visualization separated tumor-infiltrating T cells into 13 clusters that were annotated manually based on their expression of canonical marker genes. (Fig. 4A). Tumors from Foxp3^dE2 mice had more Th-like T_reg cells and effector CD8 T cells, but much fewer effector T_reg cells and γδ T cells (Fig. 4B). To understand the differences between Th-like T_reg cells and effector T_reg cells in the tumors, we analyzed the gene expression of several sets of markers, including those for T_reg cells (Foxp3, Il7r, Il2ra, Ctla4, Il10), fragile T_reg cells (Pdcd1, Ifngr1, Icos), Th1 cells (Il2, Ifng, Tnf, Stat1, Stat4, Tbx1, Cxcr3, Ccr5), Th2 cells (Il4, Il5, Il13, Gata3, Stat6, Cxcr4), Th17 cells (Il17, Rorc, Stat3) and other genes of interest (Bcl6, Cd4, Eomes, Klrg1, Tigit, Tnfrsf4, Tnfrsf9, Tnfrsf18, Tcf7, Ikzf2, Entpd1, Sell, Lag3, Havcr2, Cd40lg, Ptgdr2, Gzmb) (Fig. 4C&D). Th2 features were also observed in Th-like T_reg cells such as increased expression of Il4, Il5, Il13, Gata3, Stat6, and Cxcr4. Meanwhile, Th-like T_reg cells also partially lost the features of suppressive T_reg cells such as decreased expression of Il2ra, Il10, Klrg1, Lag3, Havcr2, and Gzmb (Fig. 4C&D). Notably, the levels of Foxp3 mRNA in Th-like T_reg cells and effector T_reg cells were comparable (Fig. 4C), indicating that the Th-like feature of dE2 T_reg cells is not due to the loss of total FOXP3 expression.

Figure 4. dE2 T_reg cells acquire a T helper-like phenotype in the tumor microenvironment.

(A) Uniform manifold approximation and projection (UMAP) clustering of total CD3⁺ T cells based upon single-cell RNA sequencing of tumor-infiltrating T cells harvested from the EO771 tumors 10 days after tumor inoculation. (B) Percentages of identified tumor-infiltrating T cells in each group. Data represent mean ± SD (3 WT mice and 4 Foxp3^dE2 mice). ns: not significant; *: p < 0.05; ***: p < 0.001; ****: p < 0.0001 by two-way ANOVA with Sidak’s multiple comparisons test. (C) Expression of selected genes in effector T_reg cells and Th-like T_reg cells. (D) Volcano plot showing differential gene expression in Th-like T_reg cells vs effector T_reg cells. (E) Gene set enrichment analysis of pathways in Th-like T_reg cells versus effector T_reg cells. NES, normalized enrichment score. Fisher exact test was used to calculate p-values. (F) RNA velocity analysis on CD4⁺ T cells with cell states shown on UMAP plots. Vector arrows represent predicted direction of future transcriptome (upper two panels). Overlay of phenotypic clusters on the 6 states of CD4⁺ T cells (lower two panels).

We used flow cytometry to examine the expression of selected genes in the tumor-infiltrating T_reg cells based upon the scRNA-seq results (Supplementary Fig. 6). dE2 T_reg cells infiltrating EO771 tumors had reduced expression of CD25 and KLRG1 but increased expression of IFNγ compared to tumor-infiltrating WT T_reg cells (Supplementary Fig. 6A). Furthermore, dE2 T_reg cells also showed decreased expression of ICOS, GZMB and IL-10 (Supplementary Fig. 6B) but increased expression of PD-1 (Supplementary Fig. 6C). Some of these phenotypic changes in dE2 T_reg cells such as increased IFNγ and PD-1 but decreased ICOS and IL-10 were also observed in fragile T_reg cells (35). However, there were differences between Th-like dE2 T_reg cells and fragile T_reg cells. Despite having higher frequency of IFNγ expression, tumor-infiltrating dE2 T_reg cells had similar T-bet expression to WT T_reg cells (Supplementary Fig. 6D). Unlike fragile T_reg cells showing reduced expression of neuropilin-1 (NRP1), CD73 and Helios (35) that were associated with Treg stability, tumor-infiltrating dE2 T_reg cells had increased expression of NRP1 and CD73 but similar Helios expression compared with WT T_reg cells (Supplementary Fig. 6E). GSEA analysis showed that in comparison with effector T_reg cells, the NF-kB signaling, Th1 and Th2 cell differentiation, TNF signaling, MAPK signaling, and IL-17 signaling pathway were significantly upregulated in Th-like T_reg cells, while antigen processing and presentation, pyruvate metabolism, carbon metabolism, and endocytosis pathways were downregulated (Fig. 4E).

To determine the relationship among effector T_reg cells, Th-like T_reg cells and T helper cells in the TME, we performed RNA velocity analysis on CD4 T cells in our scRNA-seq data. Re-clustering CD4 T cells confirmed that tumor-infiltrating dE2 T_reg cells were mostly Th-like T_reg cells with few effector T_reg cells while WT T_reg cells were mostly effector T_reg cells with few Th-like T_reg cells (Fig. 4F). There was also a distinct T_reg population, especially in Foxp3^dE2 mice, which resembled fragile T_reg cells (11, 35) with upregulated Pdcd1, Ifngr1, Ifng and Il2 but downregulated Icos and Il10 expression (Supplementary Fig. 7). We compared expression of selected genes that are important for T_reg stability and function to further define the differences among effector T_reg cells, Th-like T_reg cells, and fragile T_reg cells. Four categories of genes were identified based upon the scRNA-seq data: 1) upregulated in both Th-like T_reg cells and fragile T_reg cells (Pdcd1, Ifngr1, and Il1rl1, Supplementary Fig. 7A); 2) upregulated in Th-like T_reg cells and further increased in fragile T_reg cells (Tbx21, Ifng, and Il2, Supplementary Fig. 7B); 3) downregulated in both Th-like T_reg cells and further decreased in fragile T_reg cells (Il2ra, Klrg1, Tnfrsf4 and Tnfrsf9, Supplementary Fig. 7C); and 4) downregulated in fragile T_reg cells (Foxp3, Icos, Il10, Tnfrsf18 and Ikzf2, Supplementary Fig. 7D). RNA velocity analysis also suggested that the tumor-infiltrating Th-like T_reg cells in Foxp3^dE2 mice were more likely to differentiate towards fragile T_reg cells, which was not a prominent trajectory from tumor-infiltrating effector T_reg cells in WT mice (Fig. 4F). Taken together, our data suggest that dE2 T_reg cells are less stable and become Th-like cells in the TME.

Tumor-infiltrating CD8 T cells in Foxp3^dE2 mice are more activated

In our scRNA-seq data, we observed an increased population of effector CD8 T and a trend of increased exhausted CD8 T cells in the tumors of Foxp3^dE2 mice compared to those in the tumors of WT mice (Fig. 4B). To examine the differences between effector and exhausted CD8 T cells, we analyzed a panel of genes that are important for the function of CD8 T cells. Effector CD8 T cells exhibited more cytotoxicity markers (Gzma and Gzmk) and less exhaustion markers (Pdcd1, Lag3, and Havcr2) compared to exhausted CD8 T cells (Fig. 5A). We simultaneously performed single cell TCR-seq analysis alongside scRNA-seq and observed that there were more hyper-expanded effector (cluster 4) and exhausted (cluster 5) CD8 T cells in tumors from Foxp3^dE2 mice compared to those from WT mice (Fig. 5B). The hyper-expanded effector CD8 T cells acquired features associated with the exhausted cells, including higher expression of Lag3, Pdcd1, Tox, Gzmb, Gzmk, and Ifng (Fig. 5C), suggesting that the hyper-expanded portion of effector CD8 T cells may skew towards exhausted CD8 T cells. GSEA analysis showed that the hyper-expanded clones in effector CD8 T cells downregulated oxidative phosphorylation, MAPK, FOXO, PI3K-Akt and Wnt signaling pathways, while upregulating glycolysis, carbon metabolism, lysosome, and NK cell-mediated cytotoxic pathways (Fig. 5D). Further analysis of the CD8 T cell TCR repertoire revealed that effector (Teff), exhausted (Tex), effector memory (Tem) and central memory (Tcm) cells shared TCRs with each other in the tumor from Foxp3^dE2 mice, suggesting more activated and expanding CD8 T cells. In comparison, different subsets of tumor-infiltrating CD8 T cells in WT mice had remarkably divergent repertoires, suggesting a compromised CD8 T cell immune response (Fig. 5E). Our findings demonstrate that dE2 T_reg cells promote activation and expansion of CD8 T cells in the TME.

Figure 5. dE2 T_reg cells promote CD8 T cell activation in the tumor microenvironment.

(A) Expression of selected genes in effector CD8 T cells versus exhausted CD8 T. Data derived from the scRNA-seq analysis in Fig. 4. (B) T cell receptor clonotypes of each cluster in the tumors from WT and Foxp3^dE2 mice. (C) Volcano plot of differentially expressed genes in effector CD8 T cells: hyper/large clones versus all other clones. (D) Gene set enrichment analysis of pathways in effector CD8 T cells among hyper and large clones versus all other clones. (E) TCR overlap analysis of tumor-infiltrating CD8 T cells. Teff: effector CD8 T cells; Tex: exhausted CD8 T cells; Tem : CD8 effector memory cells; Tcm: CD8 effector memory cells.

Morpholino-mediated FOXP3 exon 2 skipping induces T helper-like phenotypes in T_reg cells

Our finding that dE2 T_reg cells are less suppressive, become Th-like cells, and promote antitumor immunity suggests modulation of FOXP3 mRNA alternative splicing could be a promising approach for cancer immunotherapy. A morpholino (MO) is a type of sequence-specific oligomer that can recognize and bind to mRNA, leading to the knockdown of the gene or exon skipping (36). We designed MO oligos that bind to the junction of the 2^nd exon and the 3^rd intron of human FOXP3 or mouse Foxp3 gene (Supplementary Fig. 8A) to induce the skipping of the 2^nd exon of FOXP3 in mRNA splicing (referred to dE2 MO thereafter). Both the murine and human versions of dE2 MO efficiently shift the FOXP3 expression from FOXP3^FL isoform to FOXP3^dE2 isoform in a dose-dependent manner (Supplementary Fig. 8B). To further determine the efficiency of dE2 MO-mediated isoform shift at the transcript level, we performed RNA-seq using T_reg cells cultured with control MO or dE2 MO. Percent Spliced-In (PSI) of the Foxp3 exon 2 changed from 35.48% in control MO treated T_reg cells to 0.79% in dE2 MO treated T_reg cells (Supplementary Fig. 8C).

We next tested how MO-mediated FOXP3 isoform shift impacts T_reg phenotypes in vitro. We first compared cytokines released from cultured WT and dE2 T_reg cells using a mouse cytokine array. Out of the 62 cytokines tested, the secretion of IFNγ, CCL3, IL4, IL5 was strongly increased by dE2 T_reg cells, in support of the notion that T_reg cells expressing FOXP3^dE2 isoform became Th-like T_reg cells (Supplementary Fig. 9A). We next cultured 2 × 10⁶ T_reg cells purified from spleens of WT mice in the presence of control MO or dE2 MO and stimulated with anti-CD3/28 beads for 2 days. Cytokine array analysis of the culture supernatant showed that, similar to dE2 T_reg cells, WT T_reg cells treated with dE2 MO also had increased production of IFNγ, IL4 and CCL3 (Supplementary Fig. 9B). To further confirm that the expression of FOXP3^dE2 reprogram T_reg cells to helper-like T cells, we cultured total splenocytes from WT mice in the presence of dE2 MO or control MO for 3 days. Percentage of IFNγ⁺FOXP3⁻ CD4 T cells and IFNγ⁺FOXP3⁺ T_reg cells were increased in dE2 MO treated samples. Similarly, IFNγ⁺CD8⁺ cells were also increased in dE2 MO treated samples compared to those in control MO treated samples (Supplementary Fig. 9C). To determine whether in vitro dE2 MO-mediated FOXP3 isoform shift would impact T_reg suppressive activity, as seen with tumor-infiltrating dE2 T_reg cells, we cultured total splenocytes with 2.5 mM control MO or dE2 MO for three days. Live T_reg cells were FACS sorted from cultured splenocytes and assayed their suppressive activity as described (11). While control MO treated T_reg cells were less suppressive compared with tumor-infiltrating T_reg cells (Fig. 3A), dE2 MO treatment further decreased suppressive activity of T_reg cells at a 1:4 T_reg-to-responder ratio (Supplementary Fig. 9D).

To understand the impact of dE2 MO mediated FOXP3 isoform shift on global transcriptomic changes in tumor-infiltrating T_reg cells and its impact on CD4 and CD8 T cell function, we isolated total T cells from human breast cancer (BC) and colorectal cancer (CRC) TILs, treated with 2.5 μM control and dE2 MO for 3 days, and performed scRNA-seq. Clustering and UMAP analysis based on differentially expressed genes clearly separated the tumor-infiltrating T cells into 11 clusters (Supplementary Fig. 9E). As expected, dE2 MO treatment downregulated IL2RA (CD25) as well as genes involved in T_reg immune suppression, while promoting the expression of IFNG and TNF in T_reg cells (Supplementary Fig. 9F). GSEA analysis revealed that NF-kB signaling pathway, cytokine-cytokine receptor interaction, IL17 signaling pathway, and T cell receptor signaling pathway were upregulated in T_reg cells treated with dE2 MO (Supplementary Fig. 9G), suggesting that shifting FOXP3 expression to the FOXP3^dE2 isoform induced global changes in transcription of T_reg cells, leading to the reprogramming of T_reg cells to Th-like T_reg cells.

Morpholino-mediated FOXP3 exon 2 skipping enhances antitumor immunity in mice

Both WT and Foxp3^dE2 mice express only one isoform of FOXP3, which does not reflect co-expression of both FOXP3^FL and FOXP3^dE2 isoforms in each human T_reg cell. To develop a system modeling human FOXP3 isoform expression for preclinical testing of the efficacy of our dE2 MO in cancer immunotherapy, we generated a mouse model with a humanized exon 2 by replacing the mouse Foxp3 exon 2 region with part of the human intron 2 (including the branch point and polypyrimidine tract) and an engineered human FOXP3 exon 2 to ensure the replacement does not change the amino-acid sequence of murine FOXP3 exon 2 region (Supplementary Fig. 10A). The mice with humanized Foxp3 exon 2 (Foxp3^huE2) are viable and fertile with no signs of autoimmunity. As expected, T_reg cells in the Foxp3^huE2 mice (hereafter referred to as huE2 T_reg cells) expressed total FOXP3 (stained by antibody 3G3 recognizing exon 1 region common to all FOXP3 isoforms) at levels similar to WT T_reg cells but levels of FOXP3 exon 2 intermediate between WT (all positive for exon 2) and dE2 T_reg cells (all negative for exon 2) (Supplementary Fig. 10B). Thus, FOXP3^FL and FOXP3^dE2 isoforms are co-expressed in huE2 T_reg cells and recapitulate the splicing features of human FOXP3. Consistent with a requirement for FOXP3 exon 2 for CD25 expression (23), huE2 T_reg cells have an intermediate expression of CD25 between WT and dE2 T_reg cells (Supplementary Fig. 10C). Interestingly, only IFNγ but not IL-2 and IL-17A expression in huE2 T_reg cells were significantly higher than WT T_reg cells, although the level of IFNγ in huE2 T_reg cells was much lower than that in dE2 T_reg cells (Supplementary Fig. 10D). Detailed immunophenotyping demonstrated that, while Foxp3^dE2 mice showed a more activated immune system including increased percentage of CD4⁺, CD8⁺, CD4⁺CD44⁺, Tfh, CD4⁺IFNγ⁺, CD4⁺IL-17A⁺, CD8⁺IFNγ⁺, CD8⁺GZMB⁺, and germinal center B cells than WT mice, Foxp3^huE2 mice showed similar immunophenotypes to WT mice (Supplementary Fig. 10E – H).

To assess the effect of dE2 MO-mediated T_reg reprogramming on tumor growth in vivo, we established EO771 tumors in the 4^th mammary fat pads of WT and Foxp3^huE2 mice. 14 days after tumor inoculation, we intratumorally injected 5 mg/kg control or dE2 MO every 3 days for 5 times total. Control MO treated Foxp3^huE2 mice, similar to dE2 MO treated WT mice, had significantly lower tumor burden than control MO treated WT mice, indicating partial FOXP3^dE2 expression in huE2 T_reg cells due to alternative splicing (Foxp3^huE2 mice) is sufficient to promote antitumor immunity to the extent similar to dE2 MO treated WT mice (Fig. 6A). dE2 MO treatment in Foxp3^huE2 mice further reduced tumor burden compared to control MO treated Foxp3^huE2 mice (Fig. 6A). Intratumoral dE2 MO treatment shifted FOXP3 to dE2 isoform and reduced CD25 expression in tumor-infiltrating huE2 T_reg cells (Fig. 6B). Consistent with the results in Foxp3^dE2 mice, dE2 MO treated tumors in Foxp3^huE2 mice had increased frequency of IFNγ⁺ and IL-2⁺ T_reg cells (Fig. 6C) but reduced frequency of KLRG1⁺ T_reg cells than control MO treated tumors, and most IFNγ⁺ T_reg cells were KLRG1⁻ (Fig. 6D). Comparing to those in control MO treated tumors, we saw a trend (not significant) of increased frequency of IFNγ⁺ conventional CD4 T cells (Fig. 6E) but significantly increased frequency of IFNγ⁺ CD8 T cells (p < 0.05) (Fig. 6F) in dE2 MO treated tumors from Foxp3^huE2 mice.

Figure 6. Morpholino-mediated FOXP3 exon 2 skipping enhances antitumor responses in mice.

(A) Growth curve analysis of EO771 tumors in WT and Foxp3^huE2 mice intratumorally treated with 5 mg/kg control or dE2 MO (n=5 mice per group). Arrows indicate intratumoral MO treatment. (B) Flow cytometric analysis of FOXP3 isoform and CD25 expression in tumor-infiltrating T_reg cells in MO treated Foxp3^huE2 mice. (C) IFNγ and IL-2 expression in tumor-infiltrating T_reg cells in MO treated Foxp3^huE2 mice. (D) Expression of KLRG1 and IFNγ in tumor-infiltrating T_reg cells in MO treated Foxp3^huE2 mice. (E) Expression of IFNγ in tumor-infiltrating conventional CD4 T cells in MO treated Foxp3^huE2 mice. (F) Expression of IFNγ and GZMB in tumor-infiltrating CD8 T cells in MO treated Foxp3^huE2 mice. Data represent mean ± SD (n = 5 mice per group) from one of three independent experiments. n.s: not significant; *: p < 0.05; **: p < 0.01; ***: p < 0.001; ****: p < 0.0001 by two-way ANOVA with Turkey’s multiple comparisons test (A) or unpaired two-tailed t-test (B – H).

Next, we examined the toxicity of systemic administration of dE2 MO in Foxp3^huE2 mice. 5 mg/kg dE2 MO and control MO were administered i.v. 3 times/week for a total of 10 doses. A PBS treated group was also included as negative controls. Mice were sacrificed 3 days after the last dose for immunophenotyping and histopathological assessment. These 10 doses of systemic dE2 MO at 5 mg/kg were not able to shift FOXP3 alternative splicing in splenic T_reg cells, thus similar levels of FOXP3 isoforms and CD25 in T_reg cells were detected (Supplementary Fig. 11A), and comparable percentage of T_reg cells expressing IFNγ and IL-2 were observed in all three treatment groups (Supplementary Fig. 11B). Likewise, no significant differences in CD4⁺, CD8⁺, and B cell compartment were detected among treatments (Supplementary Fig. 11C – F). Major organs including both left and right kidneys, heart, liver, lung and spleen were harvested and fixed for histology. Hematoxylin and eosin (H&E)-stained slides were assessed by pathologists, and no major treatment-related lesions were observed in this study (Supplementary Fig. 11G). Together, these findings indicate that intratumoral dE2 MO can promote anti-tumor responses but further investigation is needed to determine whether systemic administration is feasible.

Morpholino treated TILs show enhanced killing against autologous patient-derived tumor organoids

We next tested whether treatment of dE2 MO would also enhance the cytotoxicity of autologous TILs isolated from human BC and CRC tumor tissues against patient-derived tumor organoids (PDOs). One part of fresh tumor tissue was dissected into small pieces, digested into single cells, and seeded in Matrigel basement membrane matrix to generate PDOs. Total TILs were isolated from another part of the same tumor tissue and expanded in the presence of 2.5 μM control or dE2 MO for 3 days. The PDOs were then co-cultured with or without expanded TILs for 48 h. A total of 9 BC and 15 CRC tumor samples were harvested from respective human patients (one sample per patient). We found that dE2 MO enhanced the cytotoxicity of autologous TILs in 6/9 of BC (Fig. 7A & Supplementary Fig. 12A) and 8/15 CRC samples (Fig. 7B & Supplementary Fig. 13A), with an overall effectivity rate around 53%–67%. In some samples, where remaining cells were enough for cell profiling analysis, we analyzed the initial cell composition of TILs before MO treatment (Supplementary Fig 12B & 13B). In the limited number of samples, the percentage of CD4 subpopulation in the samples does not seem to be associated with the activity of dE2 MO on cytotoxicity of autologous T cells, suggesting that reprogramming of small but specific population of T_reg cells may lead to activation of CD8 T cells in the TILs.

Figure 7. Morpholino-treated tumor-infiltrating lymphocytes display enhanced killing of autologous patient-derived tumor organoids.

(A&B) Human breast cancer (A) and colorectal cancer (B) patient-derived organoids (PDO) were co-cultured with autologous T cells that were pre-treated with control or dE2 MO (2.5 μM) for 3 days. The cytotoxicity of T cells was assessed by measuring the size of the tumor organoids. Statistical analysis was conducted by one-way ANOVA with Fisher’s LSD multiple comparisons test. n.s: not significant; *: p < 0.05; **: p < 0.01; ****: p < 0.0001.

Discussion

In tumor immunity, T_reg cells can promote the development and progression of tumors by suppressing effective antitumor immune responses in tumor-bearing hosts. High infiltration of T_reg cells into TME is associated with poor patient survival in various types of cancer. Therefore, targeting T_reg cells has been considered a promising anticancer strategy. As a master regulator, FOXP3 governs the transcriptional program and immune functions in T_reg cells. Human FOXP3 gene encodes two major isoforms through alternative splicing, FOXP3^FL and short isoform FOXP3^dE2 lacking exon 2. However, study of FOXP3^dE2 in vivo has been limited due to lack of expression in mice. We previously generated a Foxp3^dE2 mouse model expressing only the FOXP3^dE2 isoform and found that T_reg cells expressing this isoform were less stable and could induce a mild lupus-like autoimmunity (23). In this study, we found that these mice are resistant to a variety of types of mouse tumor formation. We also developed a mouse model Foxp3^huE2 with engineered Foxp3 exon 2 region allowing alternative splicing as in humans and found these mice expressing both FOXP3^dE2 and FOXP3^FL isoforms have reduced EO771 tumor burden compared to WT mice expressing only the FOXP3^FL isoform. Shifting FOXP3^FL isoform to FOXP3^dE2 isoform by intratumoral injection of a morpholino oligo (MO) further reduces tumor burden. Our study suggests that the MO-mediated FOXP3 isoform shifting is a potentially promising immunotherapeutic approach for cancer treatment.

Mechanistically, tumor-infiltrating dE2 T_reg cells are less suppressive in the TME than WT T_reg cells. It has been shown that KLRG1⁺ T_reg cells amass in the TME in various tumor models and might directly correlate with tumor progression (37). Extensive IL-2R signaling is essential for the development of KLRG1⁺ T_reg cells (34) that express higher levels of suppressive molecules and have superior suppressive activity in vitro over KLRG1⁻ counterparts (32, 34). Similarly, expression of CD25, the IL-2 receptor alpha chain, is critical for T_reg lineage stability and suppressive function, especially under inflammatory conditions (38). The lower expression of CD25 we observed in tumor-infiltrating dE2 T_reg cells might explain the low KLRG1⁺/KLRG1^- ratio of dE2 T_reg cells compared to tumor-infiltrating WT T_reg cells. Considering lower CD25 and KLRG1 expression, it is not surprising that tumor-infiltrating dE2 T_reg cells have lower inhibitory potential on the proliferation of responder CD4 T cells than WT T_reg cells. Together with lower production of the cytokine IL-10 in dE2 T_reg cells and a lower percentage of T_reg cells among tumor CD4⁺ cells, Foxp3^dE2 mice generate a less suppressive TME that enables more efficient antitumor immunity than WT mice. In line with our finding that T_reg cells expressing FOXP3^dE2 isoform are less suppressive, a recent study demonstrated that FOXP3 exon 2 expressing (FOXP3E2⁺) T_reg cells derived from patients with breast cancer show higher immunosuppressive capacity (39).

However, lower suppressive activity of dE2 T_reg cells is not likely to be the only mechanism by which the FOXP3^dE2 isoform boosts antitumor immunity. dE2 T_reg cells infiltrating EO771, MC38 and LLC mouse tumors produced higher levels of IFNγ and IL-2 than tumor-infiltrating T_reg cells in WT mice. Our scRNA-seq results identified a large proportion of dE2 T_reg cells displaying T helper-like phenotypes despite comparable total Foxp3 mRNA expression between dE2 T_reg cells and WT T_reg cells in the EO771 TME. The expression pattern of selected genes, including increased IFNγ and PD-1 but decreased ICOS and IL-10 expression, is reminiscent of fragile T_reg cells that promote antitumor immunity and response to immunotherapy (35). We previously showed that dE2 T_reg cells are less stable and transdifferentiate into Tfh cells to induce germinal center B cells and autoantibodies following adoptive transfer into Tcrb^−/− recipient mice (23). In this study, compared to the group that received only CD8 T cells, adoptive transfer of both CD8 T cells and dE2 T_reg cells into Tcrb^−/− recipient mice reduced tumor burden, while transfer of both CD8 T cells and WT T_reg cells increased tumor burden. Furthermore, intratumoral injection of dE2 T_reg cells reduced tumor burden and was associated with increased total CD8 T, IFNγ⁺ CD8 T, and GZMB⁺ CD8 T cells in the TME, despite minimal effect on tumor-infiltrating host T_reg cells. Together, these data strongly suggest that tumor-infiltrating dE2 T_reg cells not only are less suppressive but also act like helper cells able to promote CD8 T cell-mediated immunity against tumors.

To explore the possibility of targeting FOXP3 RNA splicing to treat breast cancer and other solid tumors, we designed morpholino oligoes that efficiently shifted human and mouse FOXP3^FL to the FOXP3^dE2 isoform in vitro. However, 10 doses of 5 mg/kg systemic dE2 MO failed to shift FOXP3 isoform expression in splenic T_reg cells. When injected intratumorally, we found that 5 mg/kg dE2 MO shifted FOXP3 isoform expression and substantially suppressed EO771 tumor growth, especially in Foxp3^huE2 mice that bear a Foxp3 gene encoding both isoforms. An obvious drawback for intratumoral administration is that not all tumors in human patients are readily accessible. Thus, further preclinical studies optimizing delivery of dE2 MO into tumors are needed prior to clinical translation.

Adoptive cell therapy (ACT) with tumor-infiltrating T cells (TILs) has shown promising therapeutic potential for the treatment of unresectable or metastatic solid tumors. However, response rates and long-term cancer remission vary in different ACT trials (40). TIL expansion in vitro uses a high dose of IL-2 that also promotes T_reg proliferation and has been used in T_reg expansion in clinical trials (41–43). In this study, we showed that dE2 T_reg cells not only have reduced suppressive activity but also act like T helpers to promote antitumor immunity and dE2 MO efficiently shifts FOXP3^FL isoform to FOXP3^dE2 isoform in vitro. We predict that shifting FOXP3 isoform expression in T_reg cells with dE2 MO during TIL expansion would relieve T_reg-mediated suppression on and even promote the expansion of effector CD4 and CD8 cells, thus boosting the efficacy of TIL therapy. We used human breast cancer and colorectal cancer tissues to establish PDO models to examine the efficacy of dE2 MO in enhancing the antitumor activity of expanded autologous TILs ex vivo. Recently, tumor organoids have emerged as an important tool for predicting clinical responses in cancer therapy (44). In this study, we used a standardized protocol to establish a tumor organoid-T cell system (45) with breast and colorectal tumor organoids and expanded tumor-specific T cells to facilitate the study of efficacy of dE2 on the antitumor immunity. This approach could inform strategies to enhance the efficacy of adoptive TIL transfer as personalized immunotherapy for human breast cancer as well as other solid tumors.

There are limitations in our study as our research primarily focuses on mouse models to study the role of FOXP3^dE2 isoform in antitumor immunity. The safety of morpholino per se has been tested, and two morpholino-based drugs Eteplirsen and Golodirsen were approved by FDA in 2016 and 2019 respectively for the treatment (intravenous infusion) of Duchenne muscular dystrophy by inducing exon skipping in the dystrophin gene (46). However, given the critical role of T_reg cells in maintaining immune homeostasis and preventing autoimmunity, off-target effects are always a concern when targeting FOXP3 expression. Most patients carrying deletion mutations in the FOXP3 exon 2 developed severe IPEX syndrome, despite detectable FOXP3^dE2 isoform in T_reg cells (23). There are substantial differences between patients expressing FOXP3^dE2 isoform and mice with germline exon 2 deletion or MO-mediated FOXP3 isoform shifting. In addition to expressing FOXP3^dE2 isoform, these patients with deletion mutations in the FOXP3 exon 2 also express a truncated and frame-shifted FOXP3^FL isoform, which likely interferes with the function of FOXP3^dE2 isoform in regulating T_reg development and function. In contrast, germline exon 2 deletion in mice and MO-mediated FOXP3 isoform shifting do not face this complication. Two independently developed Foxp3^dE2 mice are mostly healthy with mild phenotypes (23, 24), indicating that dE2 T_reg cells are capable of maintaining immune homeostasis over time. A reported case of a healthy 70+ year-old male patient carrying a deletion mutation in the FOXP3 exon 2 region (26) reenforces our notion that using dE2 MO in cancer immunotherapy would be a strategy with greater safety and tolerability than FOXP3 silencing. Nevertheless, additional studies using human samples are essential to determine the impact of dE2 MO-mediated FOXP3 isoform shifts on T_reg phenotype and function.

MATERIALS AND METHODS

Study design

The aim of this study was to define the role and mechanism of the FOXP3 dE2 isoform in antitumor immunity and to explore the potential of a Morpholino oligo (MO) that induces FOXP3 exon 2 skipping for cancer immunotherapy. Using Foxp3^dE2 and WT mice, we examined the growth of orthotopically implanted EO771 tumors and subcutaneously implanted MC38 and LLC tumors, performed flow cytometry to immunophenotype intratumoral T_reg cells and in vitro cultures to examine their suppressive function. Using Foxp3^huE2 mice whose Foxp3 gene encodes both FOXP3 FL and FOXP3 dE2 isoforms through alternative splicing as in humans, we examined the growth of orthotopically implanted EO771 tumors in Foxp3^huE2 and WT mice, as well as following intratumoral treatment with a MO to induce FOXP3 exon 2 skipping.

Mice

Mouse colonies were bred and maintained at Laboratory Animal Resource Center of IUSM. All experiments involving animals were approved by IACUC of IUSM. As the manuscript focuses on breast cancer, only 6–10 week old female mice were used. C57BL/6 and Tcrb^−/− mice were obtained from the Jackson laboratory. BoyJ mice were obtained from the In Vivo Therapeutics Core of the Indiana University Simon Cancer Center. Foxp3 exon2 deletion (Foxp3^dE2) mice were generated with CRISPR-Cas9 system as described previously (23). Foxp3 humanized exon 2 (Foxp3^huE2) mice was generated with the help of Biocytogen (Boston, MA) as described in Supplementary Fig. 10.

Cell culture and reagents

Murine cancer cell lines EO771, LLC1, and MC38 were purchased from the American Type Culture Collection. All the cell lines were maintained under standard conditions specified by the manufacturer and were tested negative for mycoplasma contamination. Morpholinos were synthesized by Gene Tools LLC. The sequence of the morpholino for murine Foxp3 exon 2 skipping is 5’- AGCCTGCTCCGATTCCATACCTGAT -3’, and the sequence of the morpholino for human FOXP3 exon 2 skipping is 5’- TGCCCATTCACCGTCCATACCTGGT -3’.

Flow cytometry

Tumors were dissociated using Miltenyi Tissue Dissociation Kit. For cytokine staining, cells were stimulated with Cell Activation Cocktail with Brefeldin A (Biolegend) for 4 hours, washed once with cold PBS and preincubated with anti-mouse CD16/CD32 (for Fc blocking) for 15 min at room temperature before staining. Cells were then washed once and incubated with the desired antibodies for 30 min at 4°C. Cells were immediately stained with fixable viability dye for 5 min, washed once with the staining buffer, and stained with cell surface markers for 30 min at 4°C. For intracellular staining, the cells were fixed and permeabilized using the True-Nuclear^™ Transcription Factor Buffer Set kit (Biolegend). Antibodies used in this study are listed in Supplementary Table 1 and were diluted 1:100 in staining buffer (PBS containing 0.5% BSA).

Acquisition was performed using BD flow cytometers (BD Biosciences) or Cytek Aurora spectral flow cytometer.

Triple negative breast cancer transcriptomics data analysis

For the bioinformatics analysis, we utilized the RNA-seq datasets from the TCGA database. FOXP3 splicing variant expression data was derived using Kallisto method (47). Further analysis and visualizations of the processed data were performed in R and Bioconductor. Overall survival analysis was performed exclusively on patients with both survival data and gene expression data available. Samples were divided into two groups based on the TPM values (high expression (> median) and low expression (≤ median)). Kaplan-Meier survival plots were created using the R packages “survival” and “survminer” and compared using the log-rank test. To perform differential expression analysis for 189 TNBC patients from TCGA-BRCA dataset (27), the raw count data were normalized using the size factor normalization method in DESeq2 and an absolute log₂ fold change > 0.5 and p < 0.01 (48). KEGG pathway analysis was performed using the enrichKEGG and gseKEGG functions from the clusterProfiler R package, with significance defined as adjusted p < 0.05. Next, the relative proportions of immune cells in human TNBC tumors were evaluated using bulk RNA-seq data of TNBC samples from TCGA and analyzed by MCP-counter (49).

In vitro T_reg suppressive assay

Splenic Naïve CD4⁺ (CD62L^hiCD44^low) cells were sorted from Thy1.1 mice as responder cells and labelled with Tag-it Violet^™ Proliferation and Cell Tracking Dye (BioLegend) according to the manufacturer’s protocol. Spleen cells isolated from C57BL/6 mice were CD3 depleted using CD3ε MicroBead Kit (Miltenyi) and treated with 50 μg/mL mitomycin C at 37°C for 30 min. Cells were washed with PBS and used as antigen presenting cells (APCs). Intratumoral T_reg cells (CD25^hi CD127^low) were sorted from MC38 tumors harvested from Foxp3^dE2 and WT mice. For suppression assay, responder cells (2.5 × 10⁴), APCs (5 × 10⁴), and different concentrations of T_reg cells (1:2–1:8 T_reg:Teff ratio, 3125–12500 T_reg cells) were co-cultured and activated with 2 μg/mL anti-CD3 (Biolegend) in 100μl RPMI for 3 days. Suppression was then calculated with the formula (T₀-T_[i])/T₀, where T₀ is percentage of proliferated cells without T_reg cells (responders only); T_[i] is percentage of proliferated cells at a given T_reg: responder ratio. For the morpholino-treated T_reg suppression assay, spleen cells were isolated from WT mice and were cultured for 3 days with 2.5μM dE2 or Control MO. T_reg cells (CD25^hi CD127^low) were FACS sorted used for suppression assay.

Single-cell RNA-seq and TCR-seq

Tumors were harvested 10 days after inoculation, digested into single cells, stained (PE-CD11b, APC-CD45, PE/Cy7-CD3, viability dye), and sorted for viable CD11b⁻CD45⁺CD3⁺ T cells. A total of 100,000 T cells from each sample were subjected to single-cell sequencing. For human TILs, fresh patient samples were digested to single cells, treated with PBS control or 1 μM of dE2 MO for 2 weeks. Cells were then stained (PE-CD11b, APC-CD45, PE/Cy7-CD3, FITC-CD4, APC/Cy7-CD8, viability dye) and sorted for 135,000 viable CD4⁺ T cells and 15,000 CD8⁺ T cells for each sample, which were pooled for the subsequent single-cell sequencing.

Illumina sequencing of 10X scRNA-seq libraries was performed on a NovaSeq platform at the Center of Medical Genetics, Indiana University School of Medicine. Raw sequencing data were demultiplexed into FASTQs using the bcl2fastq software (Illumina, San Diego, CA, USA). Single-cell RNA sequencing data were aligned to GRCm38 reference genome using CellRanger (v.6.1.2). Downstream analysis including data QC, cell clustering, differential gene detection was performed using Seurat package. 1), Low quality cells with <200 or >5000 features, >5% mitochondrial UMI counts were discarded; 2), We used the resolution as 0.6 to discriminate the computed clusters; 3), Small clusters with less than 500 cells were discarded from further analysis. UMAP-projected clusters were manually annotated based on their expression of canonical marker genes. Differentially expressed genes between cell types and conditions were identified using the FindMarker function in Seurat. Cell lineage trajectories were inferred using Monocle DDRTree methods according to previous literature (50). RNA velocity analysis was performed with scVelo to predict the directionality of gene expression transitions between clusters (51). For TCR-seq data, reads were aligned to the GRCm38 reference genome, and TCR annotation was performed using Cellranger VDJ (10x Genomics, v6.1.2). Clonal proportion and diversity analyses of TCR sequences were conducted using the scRepertoire package (52) in R.

Cytokine Array

The cytokine array (# AAM-CYT-3–2, Raybiotech) was performed per manufacturer’s instructions. Briefly, T_reg cells were isolated from the spleen of wildtype or Foxp3^dE2 mice using T_reg isolation kit (Miltenyi). 2 million of T_reg cells from wildtype or Foxp3^dE2 mice were cultured in 1.5 mL RPMI medium containing 10% FBS and 2,000 IU/mL human IL-2 with or without CD3/28 beads. After 36h stimulation, 1mL supernatant was used for cytokine analysis and the dot intensity was analyzed with Protein Array Analyzer for ImageJ.

Human autologous tumor-infiltrating T cells expansion

Tumor samples were cut into pieces with diameters of 2–4 mm in cold PBS and subsequently transferred to 24-well plates (1–2 pieces/well) containing Optimizer^™ CTS^™ complete medium. The recipe for the complete medium is as follow: Optimizer^™ CTS^™ T-Cell Expansion basal medium with supplement (Thermo Fisher), 1,000 U/mL of human IL-2 (Peprotech), 2mM L-Glutamine (Thermo Fisher), and 5% CTS^™ Immune Cell SR (Thermo Fisher). Autologous T cells recovered for 3 days and Dynabeads^™ Human T-Expander CD3/CD28 (Thermo Fisher) were added to wells for further expansion.

Patient-derived tumor organoid (PDO) generation and T cell cytotoxicity

Human breast and colorectal cancer PDOs were generated following our previously published protocol (53). Briefly, surgical tumor specimens were obtained from the Tissue Procurement and Distribution Core at the Indiana University Simon Cancer Center. Tumor tissues were minced and enzymatically digested into single-cell suspensions for culture. Adherent cells were collected and used for organoid generation. Organoids were isolated using cell strainers with pore sizes ranging from 70 to 150 μm and subsequently co-cultured with autologous T cells expanded in the presence of 2.5 μM control or dE2 MO for 48 h in 24-well microplates. After co-culture, the samples were imaged using a LEICA DMi1 microscope with a 4x objective lens. At least six images were captured per sample for analysis. T cell-mediated cytotoxicity was assessed based on changes in organoid size, quantified using ImageJ software (version 4.50e).

Statistical analysis

Data plot and statistical analysis were performed with GraphPad Prism and presented as means ± SD. Sample size and the number of times an experiment was independently performed were indicated in figure legends. Unpaired two-tailed Student’s t tests, and one-way or two-way analysis of variance (ANOVA) analysis with multiple comparisons test were used in data analysis. p < 0.05 was considered statistically significant. ns (not significant): p > 0.05; *p < 0.05; **p < 0.01; ***p < 0.001; and ****p < 0.0001.

Data and materials availability:

Single cell RNA-seq and single cell TCR-seq datasets are deposited to GEO and available under accession number GSE286047. The Foxp3^dE2 and Foxp3^huE2 mice will be shared following the Material Transfer Agreement. The recipient should not distribute the animals further without the consent from the corresponding authors and animals will not be used for commercial purposes. Tabulated data underlying the figures is provided in data file S1. All other data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.