Expansion of tumor-associated Treg cells upon disruption of a CTLA-4-dependent feedback loop

Abstract

Foxp3⁺ T regulatory (Treg) cells promote immunological tumor tolerance, but how their immune-suppressive function is regulated in the tumor microenvironment (TME) remains unknown. Here, we used intravital microscopy to characterize the cellular interactions that provide tumor-infiltrating Treg cells with critical activation signals. We found that the polyclonal Treg cell repertoire is pre-enriched to recognize antigens presented by tumor-associated conventional dendritic cells (cDCs). Unstable cDC contacts sufficed to sustain Treg cell function, whereas T helper cells were activated during stable interactions. Contact instability resulted from CTLA-4-dependent downregulation of co-stimulatory B7-family proteins on cDCs, mediated by Treg cells themselves. CTLA-4-blockade triggered CD28-dependent Treg cell hyper-proliferation in the TME, and concomitant Treg cell inactivation was required to achieve tumor rejection. Therefore, Treg cells self-regulate through a CTLA-4- and CD28-dependent feedback loop that adjusts their population size to the amount of local co-stimulation. Its disruption through CTLA-4-blockade may off-set therapeutic benefits in cancer patients.

Keywords: CD28; CTLA-4; MP-IVM; NFAT; T regulatory cell; Treg cell; cytotoxic T lymphocyte-associated protein 4; multiphoton intravital microscopy; nuclear factor of activated T cells; tumor tolerance.

Figure 1.. The polyclonal TCR repertoire of Treg cells is enriched for tumor reactivity

(A) Experimental protocol. (B) Phenotype of adoptively transferred T cells in the TME. (C and D) MP-IVM micrographs (top panels) and individual cell images (middle panels) depicting the nucleo-cytoplasmic distribution of NFAT-GFP in an exemplary Treg (C) and Th cell (D) in the TME. Bottom panels show false-color representations of the NFAT signaling index (SI) (see Figures S1F–S1L for details). Numbers indicate min:s. Migratory tracks indicate 1-min intervals. Scale bars, 20 μm. (E–G) 3D-instantaneous migratory velocities (E), arrest coefficient (F), and instantaneous NFAT SI frequency distribution (G) of Treg (n = 1,088 from 25 cells) and Th cells (n = 1,213 from 31 cells) from 3 individual movies each recorded in 2 independent experiments. Numbers in (G) refer to percentages of NFAT SI >0.5. (H) Fraction of time in which NFAT SI >0.5 in individual cell tracks. Values above graphs refer to frequencies of cells in which NFAT >0.5 for greater than 20% of time. (I) Instantaneous cell velocities with NFAT inactive (NFAT SI ≤0.5) or active (NFAT SI >0.5). (J) Instantaneous NFAT SI as a function of distance to nearest APC. Numbers in grid indicate quadrant frequencies. Light blue symbols indicate instances of NFAT activation following a state of inactivity. p values calculated by Mann-Whitney U test. See also Videos S1 and S2.

Figure 2.. cDCs activate Treg cells locally within the TME

(A) Experimental protocol. (B) MP-IVM micrographs of Treg cells in MC38 tumors in zDC^DTR bone marrow chimeras. Images on right show representative migratory tracks and false-color-coded NFAT SI. Scale bar, 50 μm. (C) Distribution of instantaneous Treg cell NFAT SI in the TME in presence (black, n = 933 from 15 cells) or following ablation (red, n = 824 from 28 cells) of cDC. Gate frequencies refer to NFAT SI >0.5. (D) The fraction of time in which NFAT SI >0.5 in individual Treg cells. Percentages above graphs refer to frequencies of cells in which NFAT >0.5 for greater than 20% of time. Data in (C) and (D) are pooled from 4 recordings per condition. (E) Experimental protocol. (F and G) Frequency (gated on CD45⁺ cells) (F) and activation markers (G) on Treg cells in the TME following ablation of cDC. n = 8 mice per group pooled from two independent experiments. (H) Experimental protocol. (I and J) Frequency of Treg cells in the TME and expression of Foxp3 (I), and of activation markers (J) after CnB depletion. For Treg cell quantification, n = 12–16 in 3 independent experiments. For marker expression analysis, n = 7–11 from two separate experiments. Bars depict medians, symbols individual mice. MFIs normalized by the mean of control groups. p values calculated by Mann-Whitney U test. Fold change as compared to control groups in parentheses. See also Figures S2 and S3 and Video S3.

Figure 3.. Treg cells destabilize intratumoral APC contacts with Th cells

(A) Experimental protocol. (B) MP-IVM micrographs illustrating cell motility during NFAT activation in representative HA-Treg (top), HA-Th (middle), and HA-Th cells in presence of non-visualized HA-Treg cells (bottom), in the CT26HA TME. Numbers indicate min:s. Migratory tracks indicate 1-min intervals. Scale bars, 20 μm. (C) Distribution of instantaneous NFAT SI of HA-Treg (n = 2,259), HA-Th (n = 2,177), and HA-Th + HA-Treg (n = 3,322). Numbers refer to fraction of NFAT SI >0.5. (D) Gating on signaling segments within tracks for motility analysis shown in (E)–(K). (E and F) Duration of signaling segments (E) and percentage of segments ≤20 min (F). n = 16 (HA-Treg), 13 (HA-Th), and 22 segments (HA-Th + HA-Treg) recorded in 2 independent experiments. (G) Instantaneous 3D velocities in signaling segments. n = 582 (HA-Treg), 608 (HA-Th), and 860 (HA-Th + HA-Treg). (H) Arrest coefficients in signaling segments. n = 16 (HA-Treg), 13 (HA-Th), and 22 (HA-Th + HA-Treg). (I) Schematic depiction of the 10-min displacement metric. (J) 10-min displacement indices. n = 305 (HA-Treg), 372 (HA-Th), and 482 (HA-Th + HA-Treg). (K) Migratory tracks of non-overlapping 10-min signaling segments with normalized starting coordinates. n = 36 (HA-Treg), 34 (HA-Th), and 52 (HA-Th in presence of HA-Treg). Data in (C)–(K) were collected from 6 (HA-Treg), 6 (HA-Th), and 5 (HA-Th + HA-Treg) independent recordings from two independent experiments. Bars represent medians and white symbols individual segments. Values in parenthesis above graphs indicate fold change of medians. p values calculated by Mann-Whitney U test. See also Figure S4 and Video S4.

Figure 4.. CTLA-4 blockade increases the density of co-stimulatory ligands on tumor-associated cDC

(A) Fixed/permeabilized splenic Treg cells were treated with αCD28 mAb 37.51 to prevent binding of CD80/CD86 to CD28, αCTLA-4 mAb 4F10 added at indicated concentrations, and binding of CD80-hIgG1Fc, CD86-hIgG1Fc, or 4F10 to unbound CTLA-4 was determined. Dotted lines represent background staining (secondary Ab without CD80/CD86-hIgG1Fc or isotype control mAb). (B) Tumor-bearing mice received 750 μg of 4F10 or isotype control i.v. and 12 h later, unbound cell surface CTLA-4 on tumor-infiltrating T cells was stained ex vivo with fluorescent 4F10. The difference in residual ex vivo binding of fluorescent αCTLA-4 (after substraction of isotype staining MFI) was used to estimate functional in vivo CTLA-4 blockade. (C) Percentage of Treg cells in tumor tissue 24 h after injection of αCTLA-4 clone 4F10 (n = 21), clone 9H10 (n = 9), or isotype control (n = 19) mAb (pooled from five independent experiments). (D) Experimental protocol. (E) Gating strategy to quantify expression of CD80/CD86 on cDC1, cDC2, and TAMs in tumor tissue. Numbers indicate gate frequencies. (F) CD80/CD86 on cDC1, cDC2, and TAMs 24 h after CTLA-4 blockade or ablation of Treg cells. n = 10 (αCTLA-4 clone 4F10), n = 10 (isotype), n = 5 (DT treatment in Foxp3^DTR), and n = 10 (DT treatment in WT) (pooled from two separate experiments). (G and H) Frequency (G) and numbers (H) of the indicated cell types in tumors after treatment with αCTLA-4 mAb clone 4F10 (n = 8) or isotype-matched control mAb (n = 10) (pooled from two independent experiments). (I and J) Expression of proliferation (I) and activation markers (J) in tumor-associated Treg cells after CTLA-4 blockade. n = 9 (αCTLA-4) and n = 7 (isotype), pooled from two separate experiments. Bar graphs indicate medians, each symbol an individual animal. MFI are normalized to means of the control group, values in parenthesis indicate fold change relative to medians of the control group, and p values are calculated by Mann-Whitney U test. See also Figure S5.

Figure 5.. Treg cells self-regulate their intratumoral population size via CTLA-4 in trans

(A) Experimental protocol. (B) CTLA-4 expression in CD45.2⁺ Ctla4^f/f or control Treg cells in the TME following tamoxifen treatment. (C and D) Expression of Ki67 (C) and frequency (D) of CTLA-4-sufficient (n = 8) or CTLA-4-deficient (n = 7) CD45.2⁺ Treg cells in the TME (pooled from two experiments). Fold change as compared to control groups indicated in parentheses. MFI was normalized by the mean of control groups. p values calculated by Mann-Whitney U test. See also Figure S6.

Figure 6.. CTLA-4 blockade expands tumor-associated Treg cells through enhanced CD28 co-stimulation

(A) Experimental protocol. (B) Frequency of tumor-associated Treg and of Ki67⁺ Treg cells after blockade of CTLA-4 and/or IL-2. n = 10–11 per group in two independent experiments. (C) Experimental protocol. (D) Tamoxifen-induced downregulation of CD28 on Treg cells in the TME of Foxp3^creERT2 control (left) or Foxp3^creERT2 × CD28^f/f (right) mice treated with αCTLA-4 (4F10) or isotype mAbs. Gates defining CD28^lo cells are based on CD28 expression in Foxp3^creERT2 control animals. (E) Ki67 expression in CD28^hi and CD28^lo Treg cells in Foxp3^creERT2 control (left, no CD28^lo cells detected) or Foxp3^creERT2 × CD28^f/f mice (right) following αCTLA-4 or isotype mAb treatment. Numbers depict cell percentages. Bars represent medians, each symbol represents an animal. Fold change as compared to control groups indicated in parentheses. MFI normalized to the mean of controls; p values calculated by Mann-Whitney U test. See also Figure S7.

Figure 7.. Treg cells limit the therapeutic efficacy of CBI even when their CTLA-4-dependent function is disabled

(A) Experimental protocol. (B) MP-IVM micrographs illustrating the dynamic physical contact zone (white) between a Treg cell (green, arrowhead in first frame) and a CD11c⁺ APC (magenta) in the TME, as determined by their overlapping fluorescent signals. Numbers indicate min:s. Migratory tracks indicate 1-min intervals. Scale bar, 20 μm. (C) MP-IVM micrographs recorded 18 h after αCTLA-4 4F10 or isotype control mAb treatment. Note increased Treg cell-APC contacts (white) upon CTLA-4 blockade. Scale bar, 50 μm. (D) Cumulative Treg cell contact time per hour for individual APCs in the same region of interest of the tumor at various time points. Symbols represent individual APCs and bars means. Numbers above graphs indicate percentages of APCs interacting with Treg cells for >10 min (dashed line) per hour. (E–G) Ratios of mean cumulative Treg cell contact time per APC (E) mean number of Treg cell contacts per APC (F) and median Treg cell-APC contact duration (G) between 18 h after and before injection of either αCTLA-4 or isotype mAbs. Bars depict means, symbols individual experiments. p value calculated by Student’s t test. (H) Median track speed of Treg cells within a 20 μm radius of selected APCs identifiable before and 18 h after αCTLA-4 (n = 6) or isotype mAbs (n = 8) administration. Bars represent means, paired symbols individual areas. p values calculated by ratio paired Student’s t test. In (E) to (H), numbers in parentheses represent the fold change from the control group. (I) Experimental protocol. (J) Growth curves of MC38 tumors upon 4F10-mediated blockade of CTLA-4 and/or Treg cell-specific deletion of CnB. n = 10–13 per group. The fraction of mice that rejected tumors is indicated in each graph (pooled from three independent experiments). The rates of tumor rejection were compared using χ² test. Tumor growth rates were compared using type II ANOVA followed by Holm-corrected pairwise comparisons. Table shows Holm-adjusted p values. See also Figure S7 and Video S5.