Pan-TGFβ inhibition by SAR439459 relieves immunosuppression and improves antitumor efficacy of PD-1 blockade

Abstract

TGFβ is a pleiotropic cytokine that may have both tumor inhibiting and tumor promoting properties, depending on tissue and cellular context. Emerging data support a role for TGFβ in suppression of antitumor immunity. Here we show that SAR439459, a pan-TGFβ neutralizing antibody, inhibits all active isoforms of human and murine TGFβ, blocks TGFβ-mediated pSMAD signaling, and TGFβ-mediated suppression of T cells and NK cells. In vitro, SAR439459 synergized with anti-PD1 to enhance T cell responsiveness. In syngeneic tumor models, SAR439459 treatment impaired tumor growth, while the combination of SAR439459 with anti-PD-1 resulted in complete tumor regression and a prolonged antitumor immunity. Mechanistically, we found that TGFβ inhibition with PD-1 blockade augmented intratumoral CD8+ T cell proliferation, reduced exhaustion, evoked proinflammatory cytokines, and promoted tumor-specific CD8+ T cell responses. Together, these data support the hypothesis that TGFβ neutralization using SAR439459 synergizes with PD-1 blockade to promote antitumor immunity and formed the basis for the ongoing clinical investigation of SAR439459 in patients with cancer (NCT03192345).

Keywords: PD1 blockade; T cell response; TGFβ; Tumor microenvironment; combination immunotherapy.

© 2020 The Author(s). Published with license by Taylor & Francis Group, LLC.

Figures

Figure 1.

Higher TGFβ pathway activation is associated with worse objective response and lower overall survival under anti–PD-1/PD-L1 therapies. a–c. High TGFβ pathway activation correlates with resistance to PD-1/PD-L1 therapies and with reduced overall survival. Box plots show TGFβ activation signatures in nonresponders (NR) and responders (R) in (a), melanoma data from Hugo et al.; (b), melanoma data from Sade-Feldman; and (c), urothelial cancer data. Enrichment scores are obtained via a regulated Kolmogorov-Smirnov analysis using the RNA sequencing whole-transcriptome profiles available in each study. P values are from Wilcoxon tests. (d), Overall survival for the groups of patients with urothelial cancer who had TGFβ-low pathway activation (n = 243; enrichment score ≥1) and TGFβ-high pathway activation (n = 55; enrichment score <1). TGFβ-low and TGFβ-high data from urothelial cancer was further subdivided on the basis of the respective cytotoxic T lymphocyte (CTL) scores, and patient survival was compared in each case. (e), Survival curves are shown for the TGFβ-low patients, and (f), for the TGFβ-high patients. Box plots show CTL scores in TGFβ-high and TGFβ-low groups (high versus low TGFβ pathway activation), which were not different (g).

Figure 2.

SAR439459 binds to various TGFβ isoforms and inhibits TGFβ-mediated signaling similar to fresolimumab. SPR Biacore data shows similar binding properties of SAR439459 and fresolimumab against human TGFβ1, 2, and 3 proteins, (a). Murine colon carcinoma, MC38; mouse breast cancer cells, EMT6 and human colorectal carcinoma cells HCT116-overexpressing TGFβRII were cultured with human TGFβ1 (1 ng/mL) in presence of SAR439459, fresolimumab or isotype control. Total and phosphorylated SMAD2/3 protein levels were assessed by ELISA. Graph shows ratio of phospho-/total SMAD2/3 for SAR439459 and fresolimumab and isotype control, (b). P value <.0001 using two-way ANOVA test. SBE reporter cells were cultured in the presence of human TGFβ isoforms −1, −2, and −3 with various concentrations of SAR439459, fresolimumab or isotype control and luminescence activity was measured. Graphs show ability of SAR439459 and fresolimumab to prevent TGFβ1, 2 and 3-mediated SMAD activation in the reporter cell line, (c). IC50 values for SAR439459 were 0.008, 1.22, 0.45, and for fresolimumab were 0.02, 1.86, 0.61 nM for TGFβ1, TGFβ2, and TGFβ3 respectively. Each graph shows mean ± SEM and represents one of three independent experiments.

Figure 3.

SAR439459 restores TGFβ-mediated suppression of primary immune cell function. Enriched CTV-labeled human CD8+ T cells were co-cultured with B-LCL cells in the presence of TGFβ1 (1 ng/mL) and, to analyze effect on CD8+ T cell proliferation and IFNγ, measured by FACS. Cells were gated on live CD8+ T cell population followed by CTV-low and IFNγ-positive cells. FACS plots show proliferation (CTV-low) of CD8+ T cells producing IFNγ under various treatment conditions, (a). Bar graphs show mean ± SEM percentage of CD8+ T cells that were functional CTV-low–IFNγ+, (b). Data are representative of 3 independent experiments with CD8+ T cells from 2 human donors. P values: ***<0.005, **<0.05, ***<0.001, ****<0.0001 for SAR439459 versus isotype at doses (µg/mL) 12.5, 25, 50, and 100, respectively. Enriched NK cells from human donors were cultured in the presence of IL-2 for 2–3 days and with various concentrations of human TGFβ1 (0.1, 1.0, and 10 ng/mL) as shown in the FACS plots. NK proliferation was examined by Ki-67 staining. FACS plots show percentages of live Ki-67-positive NK cells in the presence of varying TGFβ1 doses, (c). FACS plots show NK cell proliferation in presence of TGFβ1 (1 ng/mL) and its blockade by SAR439459, (d). Bar graph shows IL-2–induced NK cell proliferation in presence of various doses of TGFβ (0.1, 1.0, and 10 ng/mL) and treated with SAR439459 (50 nM), human IgG4, or untreated control, (e). P values: ***<0.01, ****<0.0001, and ***<0.05 at 0.1, 1.0, and 10 ng/mL TGFβ1, respectively. FACS plots show percentages of K562 cells positive for cleaved granzyme B (GZB) peptides delivered by NK activity under various conditions, (f). Bar graph shows calculated percentage of cleaved granzyme B peptides in K562 cells (co-cultured with NK cells) in the presence of 1 ng/mL TGFβ1 under various treatment conditions, with or without SAR439459 (50 nM) or human IgG4, (g). P values: **<0.005, ****<0.0001 at 0.1 and 1.0 ng/mL TGFβ1, respectively. One of three experiments is shown. APC, antigen-presenting cell; FITC, fluorescein isothiocyanate. NS, not significant. Graphs show dose dependent induction of GZB, perforin, TNFa and MIP1a in the culture supernatants of K562: NK cell co-culture experiments, (h). Data represents one of two experiments. Bar graphs show IFNγ produced in a human T cell: DC co-culture MLR assay with SAR439459 or fresolimumab treatments or isotype control, (i). P values: SAR439459 vs. untreated = 0.02; fresolimumab vs. untreated = 0.007. SAR439459 vs fresolimumab, NS.

Figure 4.

PD-1 blockade-mediated T cell response is further enhanced by SAR439459. Total T cells and monocyte-derived dendritic cells from healthy human donors were co-cultured in a ratio of 10:1 with anti–PD-1 (10 nM) alone, SAR439459 (50 nM) alone, their combination, or isotype controls. Bar graph shows mean ± SEM level of IFNγ under various treatments, (a). Assay was set up with or without addition of exogenous TGFβ1 to the cultures. P values are shown between isotype control vs SAR439459 for all treatment conditions: *<0.01, **<0.008, a = 0.08, b = 0.03. Experiment was performed in 2 different donors and repeated at least four times. Graph shows mean ± SEM luciferase activity from NFAT reporter Jurkat assay under similar treatment conditions as detailed in Materials and Methods section, (b). Data represent at least 3 independent experiments. P value: ****<0.0001. RLU, relative light units. In MLR assay data demonstrates production of IL-2, GZB, Perforin, IL-6, IL-10 and TNF by TGFβ and PD1 co-inhibition, (c). P values: Combination vs. anti-PD1 or SAR439459 alone, IL2 < 0.0004 and <0.0001, GZB 0.0003 and 0.0001, Perforin, 0.01 and 0.08, IL-6 0.001 and 0.01, IL-10 < 0.0001 both and TNFα 0.03 both.

Figure 5.

SAR439459 improves antitumor efficacy of PD-1 blockade in mouse tumor models and combination establishes long-lasting immunity. Antitumor efficacy of SAR439459 in combination with PD-1 antibody was examined in mouse colon and breast carcinomas, MC38 and EMT-6, respectively. Graph shows time dependent MC38 tumor volumes over time after receiving various treatments as detailed in Materials and Methods (a). Survival curves of mice from MC38 study over several days. P values: *<0.05, **<0.005 (b). Graph depicts EMT-6 breast cancer tumor volumes over time in mice after receiving various treatments (c). Survival data from EMT-6 tumor model at various days (d). Mice surviving EMT-6 experiment in combination and SAR439459 monotherapy group were re-challenged with EMT-6 tumors, (e). Curves show that tumors were immediately rejected in these mice. Tumor growth inhibition in Lovo model by SAR439459 and fresolimumab as a monotherapy at day 31 and 35 post tumor implantation, (f). Data was compared with two-way ANOVA and no difference was observed between SAR439459 vs. fresolimumab.

Figure 6.

SAR439459 in combination with PD-1 blockade elicits proinflammatory cytokine response, tumor-specific CD8+ T cell response. Mice were implanted with various types of tumors and treated with three doses of SAR439459 alone, PD-1 antibody alone, or their combination. Control groups received either vehicle (PBS) or similar doses of isotype antibodies in combination. Intratumoral cytokines were evaluated. Bar graphs show mean ± SEM intratumoral levels of A, IFNγ, combo versus PBS, *P = .01; combo versus isotype, P = .07; B, TNFα, combo versus PBS, *P = .06; C, IL-6, combo versus PBS, *P = .04; and D, TGFβ, combo versus PBS, ***P = .0004; combo versus isotype, **P = .001; combo vs anti–PD-1, ***P = .0003; SAR439459 versus PBS, **P = .001; SAR439459 versus isotype, **P = .005; SAR439459 versus anti–PD-1, **P = .003. n = 5 mice/each group, and data are representative of two or more experiments. Mice implanted with EMT6 tumors were treated with three doses of anti-TGFβ (1D11), anti-PD1, their combination or vehicle (n = 5 each group) and intratumoral levels of various cytokines IFNγ, TNFα, IL-6 and TGFβ are shown, (e-h). P values; PBS vs. combination, 0.06 for IFNγ, 0.3 for TNFα, 0.1 IL-6. For TGFβ, PBS vs. combination = 0.0001 and anti-TGFβ = 0.0004, and anti-PD1 vs. combination = 0.0004 and anti-TGFβ = 0.002. In addition, time course evaluation of IFNγ was performed after various treatments in EMT6 tumor model, (i). Two-way ANOVA analysis show interaction P value 0.0009 and column factor 0.0013. Intratumoral levels of TGFβ were analyzed in Lovo model after treatment with different doses of SAR439459 or fresolimumab, (j). Mice were implanted with MC38OVA tumors and divided into various treatment groups. Tumors and draining lymph nodes were examined for presence of total and OVA-specific CD8+ T cells using FACS. Dot plots show dextramer staining on tumor samples among various groups. Graph shows counts of OVA-specific CD8+ T cells normalized to tumor weight, (k). P value combination versus isotype: *<0.03. Graph shows mean percentages of total and OVA-specific CD8+ T cells in tumor-draining lymph nodes among various treatment and control groups (l) & (m) respectively. P value, **0.009 combination versus isotype control. TDLN, tumor-draining lymph node.

Figure 7.

Single-cell RNA sequencing profiling in MC38 tumor model indicates that SAR439459 induces proliferative phenotype in intratumoral CD8+ T cells. Dimensional reduction and tree inference by a graphical clustering method on a set of 3,476 CD8+ T cells generated from MC38 tumor models in 12 mice assigned to 4 treatment groups (all 4 treatment groups shown together) clusters the T cells into 5 distinct “states” (see Methods). Cells are displayed in accordance to their DDRTree coordinates; numeric membership per state is indicated next to each label, (a). Bar plots showing fractional membership in states 1 and 5 for CD8+ T cells as a function of treatment. ANOVA P values (n = 12) are indicated in the plot headers. The frequency of cells in state 1 increases under the sequence of treatments (isotype-control < anti-PD-1 < SAR439459 < SAR439459+ anti-PD-1), from about 35% to 45%, while the frequency of cells in state 5 markedly decreases in the same order, from about 45% to 15%. No such trends are seen for the states 2, 3 and 4. States 1 and 5 accounted for 75% of the CD8+ T cells, while state 4 accounted for about 15% and state 2 and 3 for approximately 10% of the total. P values are shown (b). Heat map of genes selected for specificity to each of the 5 CD8+ T cell states. The top 50 genes ranked by specificity were selected for each state (some overlap occurs between gene sets, resulting in only 227 genes selected overall). Colors represent Z-transformed log2 (transcripts per 100,000 + 1) values. While state 1 shows marked expression of genes for cell cycle and cellular proliferation (inset shows representative genes in this group), state 5 shows marked expression of genes for cytotoxic or checkpoint proteins (see inset). State 4 is enriched in genes induced by type I interferons, but its membership did not significantly change with treatment (c). Gene set enrichment analysis confirms engagement of cellular proliferation pathways in many of the state 1 CD8+ T cells, and in none of the state 5 CD8+ T cells; thus, a signature consisting of G2/M transition genes scores positively for many cells in state 1, and for none in state 5, (d). Comparison of 2 signature based CD8+ T cell categories, “primed/activated” versus “exhausted,” shows that state 1 has about twice as many primed/activated cells than state 5, with respective proportions 40% and 20%, (e).