Novel TGF-β inhibitors ameliorate oral squamous cell carcinoma progression and improve the anti-tumor immune response of anti-PD-L1 immunotherapy

Abstract

TGF-β is a key regulator of oral squamous cell carcinoma (OSCC) progression and its potential role as a therapeutic target has been investigated with a limited success. This study evaluates two novel TGF-β inhibitors as mono or combinatorial therapy with anti-PD-L1 antibodies (α-PD-L1 Abs) in a murine OSCC model. Immunocompetent C57BL/6 mice bearing malignant oral lesions induced by 4-nitroquinoline 1-oxide (4-NQO) were treated for 4 weeks with TGF-β inhibitors mRER (i.p., 50μg/d) or mmTGF-β2–7m (10μg/d delivered by osmotic pumps) alone or in combination with α-PD-L1 Abs (7x i.p. of 100μg/72h). Tumor progression and body weight was monitored. Levels of bioactive TGF-β in serum were quantified using a TGF-β bioassay. Tissues were analyzed by immunohistology and flow cytometry. Therapy with mRER or mmTGF-β2–7m reduced tumor burden (p<0.05) and decreased body weight loss compared to controls. In inhibitor-treated mice, levels of TGF-β in tumor tissue and serum were reduced (p<0.05), while they increased with tumor progression in controls. Both inhibitors enhanced CD8⁺ T cell infiltration into tumors and mRER reduced levels of myeloid-derived suppressor cells (p<0.001). In combination with α-PD-L1 Abs, tumor burden was not further reduced, however, mmTGF-β2–7m further reduced weight loss (p<0.05). The collagen-rich stroma was reduced by using combinatorial TGF-β/PD-L1 therapies (p<0.05), enabling an accelerated lymphocyte infiltration into tumor tissues. The blockade of TGF-β signaling by mRER or mmTGF-β2–7m ameliorated in vivo progression of established murine OSCC. The inhibitors promoted anti-tumor immune responses, alone and in combination with α-PD-L1 Abs.

Introduction

Oral squamous cell carcinoma (OSCC) is the sixth most common malignancy and a major cause of cancer morbidity and mortality. Every year, approximately 500,000 new cases of oral and pharyngeal cancers are diagnosed worldwide [1]. Current treatment modalities for OSCC include among others chemoradiotherapy, surgery, EGFR and COX-2 inhibitors, and photodynamic therapy [2]. However, the 5-year survival rate has improved only marginally in recent decades, which emphasizes the need for advancements in treatment of OSCC [3].

Most epithelial malignancies, including OSCC, are characterized by overexpression of oncogenes, growth factor receptors, enzymes and various immunosuppressive factors, ultimately driving disease progression [1]. Transforming growth factor-β (TGF-β) is a key regulator of OSCC progression. Although the prognostic relevance of TGF-β expression in OSCC and the correlation with clinicopathological data have been controversial, the generally accepted view is that most oral carcinomas overexpress TGF-β and that TGF-β is essential for tumor progression [4,5]. Various pro-malignant functions of TGF-β have been reported and can be collectively attributed to its abilities to re-shape the tumor microenvironment (TME) [6]. One major effect of TGF-β is the downregulation of immune cell functions which facilitates tumor escape from immune surveillance [7]. It was shown that TGF-β inhibits tumor clearance mediated by CD8⁺ cytotoxic T cells and downregulates natural killer (NK) cell cytotoxicity [8,9]. In addition to these systemic effects, TGF-β modulates the peritumoral stroma, reprogramming the extracellular matrix and promoting the formation of cancer-associated fibroblasts (CAFs) [10]. The reprogrammed fibroblast- and collagen-rich peritumoral stroma prevents infiltration of immune cells into the tumor parenchyma, ultimately restraining anti-tumor immunity [11].

More recently, it was shown that TGF-β signaling might play an important role in patients who receive certain anti-cancer therapies. Zhu et al. demonstrated that chemotherapeutics can stimulate TGF-β production and consequently increase TGF-β signalling in the TME. The authors reported that combinatorial therapy with a TGF-β ligand trap alleviated this unintended side effect and enhanced therapeutic efficacy of chemotherapy [6]. Similar findings were reported by Mariathasan et al., who described enhanced TGF-β signaling in response to anti-PD-L1 antibody (α-PD-L1 Ab) therapy [11]. These findings might explain why some cancer patients do not respond to immune checkpoint inhibitors (ICIs) or manifest hyperprogression [12]. Thus, inhibition of TGF-β signaling emerges as an urgent need in cancers refractory to available therapies. To date, various TGF-β inhibitors have been utilized alone or in combination with current cancer therapies in experimental human clinical trials with a variable success rate [13,14]. However, despite the known importance of TGF-β in promoting cancer progression, no TGF-β inhibitor has been approved by FDA for human use so far.

The 4-NQO oral carcinogenesis orthotopic murine model, which faithfully reproduces the initiation and progression of human OSCC, offers an opportunity for a variety of pre-clinical studies. The model has been used in studies of oral carcinogenesis [15], in cancer prevention research [16] and for screening of therapeutic efficacy of novel drugs [17]. 4-NQO tumors express TGF-β and PD-L1, as previously shown in the literature [18,19]. The aim of this study was to evaluate therapeutic efficacy of two recently generated TGF-β inhibitors alone and in combination with α-PD-L1 Abs in the orthotopic and immunocompetent 4-NQO mouse model.

Materials and Methods

TGF-β inhibitors

TGF-β inhibitors were developed and produced in the laboratory of A.P.H. The trivalent TGF-β ligand trap RER was described previously and consists of the endoglin-like domain of the rat TGF-β co-receptor betaglycan (BG_E, or E) fused to one domain of the human TGF-β type II receptor extracellular domains (RII, or R) on the N- and C-terminus [6,20]. In this study, we used mRER, which is based on an all murine sequence, except for linkers, which are non-natural. Compared to other types of TGF-β inhibitors, mRER has a near picomolar antagonistic potency and a size that is smaller than that of a neutralizing antibody, potentially enabling better penetration of dense tissues, such as the extracellular matrix, and effective sequestration, even at low concentrations of the inhibitor in tissues [6,20]. Daily intraperitoneal injections of mRER (50 μg) were performed starting in week 18 for 4 weeks (Fig. 1A).

Figure 1.

Characterization of the 4-NQO model. (A) A schema is provided for 4-NQO oral administration in water for tumor initiation and for delivery of treatments beginning in week 18. Green, blue and red arrows indicate timing of the treatments. (B) Experimental groups of this study. mRER was injected intraperitoneally daily in the dose of 50μg. mmTGF-β2–7m was delivered by osmotic pumps in the dose of 10μg per day. α-PD-L1 Abs or corresponding isotype control (100μg) were injected intraperitoneally seven times every 72 hours starting in week 18. (C) Representative images of sublingual and lingual tumors on gross observation, harvested at indicated time points and representative image of lingual tumor in situ. (D) Marker expression of TGF-β and PD-L1 during 4-NQO carcinogenesis assessed by immunofluorescent stainings. Both markers are upregulated during the formation of oral carcinomas. Scale bars: 200μm. (1: basal layer; 2: epithelium; dotted line: tumor/stroma border; asterisk: tumor tissue).

The other inhibitor is an engineered TGF-β monomer, designated mmTGF-β2–7m, which lacks the heel helix, a structural motif essential for binding the TGF-β type I receptor (TβRI) but dispensable for binding the other receptor required for TGF-β signaling, the TGF-β type II receptor (TβRII) [21]. mmTGF-β2–7m retains the same affinity and binding to TβRII as TGF-β1 and TGF-β3 dimers but is unable to recruit TβRI and signal. mmTGF-β2–7m has inhibitory activity against TGF-β1, -β2, and -β3 in the nanomolar range (20–60 nM) [21,22]. It has high specificity for TβRII and, therefore, has a high target selectivity (unlike small molecule TGF-β receptor kinase inhibitors) and due to its small size (ca. 10 kDa) has great potential to penetrate dense tissues (like small molecule TGF-β receptor kinase inhibitors, but unlike antibodies) [22]. mmTGF-β2–7m was delivered by osmotic pumps (Alzet® NO 2004–0.25 μl/hr 28 days pump; Cat. #NC0836845; Durect Corporation), which were loaded and implanted subcutaneously in week 18 according to the manufacturer’s recommendations. The delivery rate was 10 μg of mmTGF-β2–7m per day until experimental end-point in week 22 (Fig. 1A).

Antibody treatment

The anti-mouse PD-L1 Ab (clone: 10F.9G2) and matching isotype control were purchased from BioXCell. Mice received 7 intraperitoneal injections with 100 μg of Ab or isotype control every 72hrs starting in week 18 (Fig. 1A).

Murine oral carcinogenesis model

To establish the 4-nitroquinoline 1-oxide (4-NQO) model, female, immunocompetent C57BL/6J mice aged 8 weeks were purchased from Jackson Laboratories. Protocols for animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC) under the reference #18042580. To induce the development of oral carcinomas, mice were administered the carcinogen 4-NQO (Tokyo Chemical Industry CO., LTD.) in drinking water once a week. Therefore, 4-NQO was dissolved in propylene glycol (Sigma-Aldrich) as stock solution (4 mg/mL), which was immediately diluted to a concentration of 0.1 mg/mL in the drinking water. 4-NQO was protected from light at all times. Mice were treated with 4-NQO for 16 weeks followed by the provision of normal drinking water.

Experimental groups

A longitudinal study was performed in the first cohort of mice. Mice were treated with 4-NQO in drinking water for 16 weeks followed by normal drinking water until week 22. Mice were sacrificed at weeks 0, 16, 18, 20 and 22 (n = 4 for each time point) and tongue tissue was photographed and harvested for histopathological analysis.

The second cohort of mice was randomly divided into experimental groups on week 18 and was treated until week 22, which was the experimental end-point (Fig. 1A). The CTRL group either received intraperitoneal injections of PBS (vehicle of mRER and mmTGF-β2–7m) or rat IgG2b isotype control (n = 6). Mice were treated with mRER (n = 6), mmTGF-β2–7m (n = 5) and α-PD-L1 Abs (n = 7) as described above. Additionally, mice received combinatorial therapy of mRER and α-PD-L1 Abs (n = 6) or mmTGF-β2–7m and α-PD-L1 Abs (n = 5). Experimental groups are listed in Fig. 1B.

Pathological analysis and tissue histology

At the experimental end-point on week 22, number of tumors were counted and sizes were measured by caliper. Volumetric caliper readings were cross validated with different formulas as previously described by us and most aligned with the volume of an ellipsoid: 4/3π(d1/2 × d2/2 × d3/2) [15]. Quantification of tumor specimens was assessed blinded by an independent examiner (S.S.Y.).

For tissue histology, oral tumors were dissected, placed in 4 % paraformaldehyde for 24 h, and subsequently in 30 % sucrose (Sigma-Aldrich) for 24 h. Samples were embedded in OCT compound (Thermo Fisher Scientific) and stored at - 80 °C for subsequent sectioning. Cryostat sections (6 μm) were cut and stained with hematoxylin and eosin (HE). Immunofluorescence staining was performed by incubating sections with a rat anti-mouse PD-L1 mAb (1:100, clone: 10F.9G2, BE0101, BioXCell), rabbit anti-mouse TGF-β1 (1:100, ab92486, abcam), rabbit anti-mouse Collagen IV (1:100, ab6586, abcam), rat anti-mouse CD68 (1:200, ab53444, abcam), rat anti-mouse Ly-6G/Ly-6C (Gr-1; 1:100, MAB1037, R&D Systems) or rat anti-mouse CD8a (1:50, 550281, BD Biosciences) overnight at 4 °C. After washing, tissue sections were incubated with Cy™3-conjugated AffiniPure F(ab’)₂ Fragment Donkey Anti-Rabbit IgG (1:400, 711–166-152, Jackson Immuno Research) or donkey anti-rat Alexa Fluor 488 (1:400, A-21208, Invitrogen) for 1 h at RT. Negative controls were stained in parallel with the secondary antibodies alone. Sections were counterstained and mounted with DAPI Fluoromount-G^® (SouthernBiotech) and imaged using an Olympus BX61 microscope and the cellSens Dimension software (Version 1.17, Olympus). For image analysis, the fluorescent signals were captured using a fixed fluorescent scaling and signal intensities were quantified with the ImageJ software.

Flow cytometric analysis of splenocytes

Spleens were harvested from all animals at experimental end-point and placed in RPMI 1640 media (Lonza). Single-cell suspensions were prepared by mechanically pressing the cells through a 70 μm cell-strainer. Red blood cells were lysed using RBC lysis buffer (Roche), and cells were washed twice with PBS. Splenocytes were frozen at −80 °C in RPMI 1640 media supplemented with 20 % (v/v) heat-inactivated FBS (Gibco), 100 U/ml penicillin and 100 μg/ml streptomycin and 10 % (v/v) DMSO (Sigma) until analysis by flow cytometry. Cells were stained with labeled antibodies for 60 min at RT in the dark to analyze the immune cell subpopulations. First, a gate on CD45⁺ cells (CD45-PE-Cy7, 1:50, #25–0451-82, Invitrogen) was used to isolate hematopoietic cells. Further gates were set on CD3⁺ (CD3-APC, 1:50, #47–0031-80, Invitrogen) and CD8a⁺ (CD8a-645, 1:50, #64–0081-80, Invitrogen) cells to study cytotoxic T cells. NK cells were studied by gating on NKp46⁺ cells (NKp46-FITC, 1:50, #560756, BD Bioscience). Myeloid-derived suppressor cells (MDSCs) were defined as CD11b⁺ (CD11b-PE, 1:3, #PNIM2581U, IOTest) and Gr-1⁺ (Gr-1-Alexa Fluor 700, 1:50, #56–5931-80, Invitrogen). Gating strategy is presented in Fig. S1. The data were acquired on LSR Fortessa Flow cytometer (BD Biosciences) and analyzed using the Flow Jo software.

Quantification of bioactive TGF-β serum levels

Blood was drawn from mice by submandibular bleeding before treatment (week 18), during treatment (week 20) and at the experimental end-point (week 22) and centrifuged at 1000 xg for 10 min to obtain serum, which was frozen at −80 °C until further use. A TGF-β bioassay was used to measure serum levels of bioactive TGF-β. Therefore, MFB-F11 reporter cells were used according to the protocols published by Tesseur et al. [23]. Briefly, MFB-F11 cells were seeded at 4 × 10⁴ cells/well in 96-well flat-bottom tissue culture plates (BD Falcon). After overnight incubation, cells were washed twice with PBS and incubated in 50 μl serum-free DMEM supplemented with penicillin/streptomycin for 2 h before serum samples were added in 10 μl volume. For SEAP assay, a SEAP Reporter Gene Assay Chemiluminescent kit (Roche) was used according to the manufacturer’s recommendations.

Statistical analysis

All data were analysed using the GraphPad Prism software (v7.0). Values are expressed as mean ± SEM. Differences between groups were assessed by one-way ANOVA. To isolate differences between pairs of groups, Student-Newman-Keuls post-hoc tests were performed. Differences were considered significant at p < 0.05.

Results

4-NQO-induced oral carcinomas overexpress TGF-β and PD-L1

All mice receiving 4-NQO showed dysplastic lesions on the tongue on week 16 which transformed to invasive carcinomas on week 17 (Fig. 1A and C). During this period, tumor incidence was 100 %. Tumors disrupted the basal layer of the tongue epithelium and infiltrated the neighboring tissue (Fig. 3A). Immunofluorescent staining showed expression of TGF-β and PD-L1 in the tumor tissue. Expression levels of these markers correlated with tumor progression in the 4-NQO model, and advanced carcinomas showed especially strong expression of TGF-β and PD-L1 (Fig. 1D). Interestingly, TGF-β was not only upregulated in tumor tissue: most stromal cells in advanced carcinomas were also positive for TGF-β, as illustrated in Fig. 1D. The 4-NQO model is ideal for evaluation of the effects of a targeted TGF-β and PD-L1 blockade, as it realistically recapitulates TGF-β and PD-L1 expression profiles of human oral carcinomas [4,24,25].

Figure 3.

Histopathologic analysis of 4-NQO tumors harvested from mice treated with TGF-β inhibitors. (A) Representative images of sections of 4-NQO mice. The images show HE-staining and immunofluorescence staining for TGF-β, PD-L1, and collagen IV (Col IV; green fluorescence) with DAPI counterstaining (blue fluorescence) at 10x magnification. Immunofluorescence staining for CD8a, CD68 and Gr-1 (green fluorescence) with DAPI counterstaining (blue fluorescence) is presented in 20x magnification. Scale bars: 100μm. PD-L1 staining presents the tumor tissue. CD68 and Gr-1 staining show surrounding stroma. Asterisks: tumor tissue; dotted lines: tumor/stroma border. (B) Quantitative analysis of TGF-β⁺ signals by immunofluorescent staining. Staining intensity in tumor tissues was quantified by using ImageJ and data are expressed as fold induction compared to CTRL. (C) Quantitative analysis of PD-L1⁺ signals in the tumor tissue. (D) Quantitative analysis of Col IV⁺ signals in the tumor stroma. (E) Numbers of tumor-infiltrating CD8a⁺ cells per region of interest (ROI). The dotted lines in A indicate the tumor borders, and only CD8a⁺ cells within the tumor tissue were counted. (F) Quantification of macrophage numbers in tongue tissues closely located to the 4-NQO tumors. Data are expressed as numbers of CD68⁺ cells per ROI. (G) Quantification of MDSC numbers in the tumor tissue and associated stroma. Data are expressed as numbers of Gr-1⁺ cells per ROI. (H) Analysis of splenocytes by flow cytometry at the experimental end-point. Cytotoxic T cells were defined as CD45⁺CD3⁺CD8⁺. (I) CD45⁺NKp46⁺ splenocytes were analyzed by flow cytometry to quantify NK cells. (J) CD45⁺CD11b⁺Gr-1⁺ splenocytes were analyzed by flow cytometry to quantify MDSCs. All values in this figure represent means ± SEM. *p < 0.05 vs. CTRL; **p < 0.01 vs. CTRL; ***p < 0.001 vs. CTRL.

Blockade of TGF-β effectively inhibits tumor progression in 4-NQO-treated mice

Therapeutic efficacy of the TGF-β inhibitors was evaluated by measuring the number of primary tumors and the total sizes of tumors per animal. Treatment with mRER significantly reduced numbers and sizes of oral carcinomas (p < 0.05; Fig. 2A, B and C). Similarly, mmTGF-β2–7m reduced numbers and sizes of tumors, however, no significant differences compared to CTRL were detected. mmTGF-β2–7m significantly reduced the weight loss of mice from week 20 until the experimental end-point (Fig. 2D). Until week 21, treatment with mRER resulted in no alterations of weight loss; however, at the experimental end-point on week 22, the weight loss of mRER-treated mice was reduced (Fig. 2D). Additionally, serum was collected before treatment (week 18), during treatment (week 20) and after treatment (week 22) and bioactive TGF-β was measured in a TGF-β bioassay. The analysis showed that serum levels of bioactive TGF-β increased during 4-NQO carcinogenesis (Fig. 2E). This increase in TGF-β serum levels was completely blocked in all mice treated with mRER (p < 0.05; Fig. 2E). The treatment with mmTGF-β2–7m blocked the increase of TGF-β serum levels on week 20 (p < 0.05); however, increased values were present on week 22 (p < 0.05; Fig. 2E). Possibly, the delivery of mmTGF-β2–7m by osmotic pumps might have ended a few days before the experimental end-point and thus a significant increase in TGF-β serum levels was observed at the experimental end-point (p < 0.05; Fig. 2E). The use of osmotic pumps could also explain the outlier in each group of mice treated with mmTGF-β2–7m (Fig. 2A and B). All pumps were individually loaded and surgically implanted thus creating a potential for individual variability.

Figure 2.

Effects of TGF-β inhibitors on tumor progression in the 4-NQO model. (A) Number of tumors per mouse at the experimental end-point. (B) Aggregate volumes of tumors in mm³ per mouse as measured in groups of mice receiving indicated treatments at the experimental end-point. (C) Representative images of sublingual and lingual tumors on gross observation, harvested at the experimental end-point. (D) Rates of weight loss during 4-NQO carcinogenesis in mice treated with mRER or mmTGF-β2–7m compared to CTRL. (E) Quantification of bioactive TGF-β in serum of mice before treatment (week 18), during treatment (week 20) and at experimental end-point (week 22) using MFB-F11 reporter cells. All values in this figure represent means ± SEM. *p < 0.05 vs. CTRL.

In 4-NQO-treated mice, the novel TGF-β inhibitors effectively inhibited tumor progression by ameliorating tumor burden, weight loss and TGF-β serum levels.

TGF-β inhibition impacts the peritumoral stroma and alters immune cell infiltration into the tumor

Tissue specimens were collected at the experimental end-point and immunostaining was performed to evaluate density and phenotype of tumor-infiltrating cells. First, sections were stained for TGF-β and PD-L1, and as our results in the longitudinal cohort indicate (Fig. 1D), tumors in the CTRL group overexpressed TGF-β and PD-L1 (Fig. 3A). PD-L1 expression levels were not altered by TGF-β inhibitors (Fig. 3A and C). All mice which received daily injections of mRER showed reduced levels of TGF-β in the tumor tissue (p < 0.01; Fig. 3A and B). Mice treated with mmTGF-β2–7m also showed a significant decrease of TGF-β levels, however, to a smaller extent compared to mice treated with mRER (p < 0.05; Fig. 3A and B). This could be either explained by downregulation of TGF-β in the tumor tissue or blocking of the epitope by mRER. Immunostaining for collagen IV (Col IV) revealed no differences between treated or untreated mice (Fig. 3A and D). Treatment with mRER and mmTGF-β2–7m resulted in increased numbers of tumor-infiltrating CD8⁺ T cells (p < 0.05; Fig. 3A and E). Staining for CD68, a macrophage marker, showed that numbers of macrophages in the tongue tissue close to the tumor borders were reduced in mice receiving treatments with mRER (p < 0.05; Fig. 3A and F). Analogous to our observation for macrophage levels, numbers of MDSCs in the TME were significantly reduced in mice treated with mRER (p < 0.05) and slightly reduced in mice treated with mmTGF-β2–7m (Fig. 3A and G).

These results demonstrate that treatments with mRER and mmTGF-β2–7m promoted the infiltration of lymphocytes into the tumor tissue, but reduced levels of macrophages or MDSCs, which are known to be activated by TGF-β.

TGF-β inhibitors induce systemic immunomodulatory effects in vivo

At the time of sacrifice (week 22), splenocytes were harvested from individual 4-NQO-treated mice and cryopreserved for analysis. Splenocytes were thawed, stained and analyzed for selected immune cell subpopulations by flow cytometry. The analysis of CD8⁺ T cells revealed no significant differences compared to CTRL (Fig. 3H). mRER did not alter NK cell levels, but the treatment with mmTGF-β2–7m resulted in an increase of NK cell frequency (p < 0.05 vs. CTRL; Fig. 3I). Treatment with mRER reduced the frequency of MDSCs (p < 0.001; Fig. 3J). The data suggest that inhibition of TGF-β results in systemic immunoregulatory changes, ultimately creating a favourable immune landscape.

Specific inhibition of PD-L1 reduces tumor burden in 4-NQO-treated mice

Although the number of oral tumors was only slightly affected by α-PD-L1 Abs, the tumor burden was significantly decreased in mice which received PD-L1 monotherapy (p < 0.05; Fig. 4A, B and C). The weight loss of mice was reduced after treatment with α-PD-L1 Abs throughout the observation period from week 19 to 22 (p < 0.05; Fig. 4D). The levels of soluble TGF-β in the serum of mice which received α-PD-L1 Abs were decreased in week 20 (p < 0.05), but increased more than 2-fold in week 22 compared to CTRL (p < 0.01; Fig. 4E).

Figure 4.

Effects of combinatorial PD-L1/TGF-β therapy on tumor progression in the 4-NQO model. (A) Number of tumors per mouse at the experimental end-point. (B) Aggregate volumes of tumors in mm³ per mouse as measured in groups of mice receiving indicated treatments at the experimental end-point. (C) Representative images of sublingual and lingual tumors on gross observation, harvested at the experimental end-point. (D) Rates of weight loss during 4-NQO carcinogenesis in mice treated with α-PD-L1 alone or in combination with mRER or mmTGF-β2–7m compared to CTRL. (E) Quantification of bioactive TGF-β in serum of mice before treatment (week 18), during treatment (week 20) and at experimental end-point (week 22) using MFB-F11 reporter cells. Data of CTRL group is also presented in Figure 2. All values in this figure represent means ± SEM. *p < 0.05 vs. CTRL; ^#p < 0.05 vs. CTRL; ^##p < 0.01 vs. CTRL.

The analysis of tissue specimens which were harvested at the experimental end-point by immunostainings revealed significantly downregulated levels of PD-L1 after treatment with α-PD-L1 Abs (p < 0.01; Fig. 5A and C), either by PD-L1 downregulation in the tumor tissue or by blocking of the epitope by the α-PD-L1 Ab treatment. No alterations of TGF-β levels in response to α-PD-L1 Abs were observed (Fig. 5B). However, the levels of Col IV were increased in the peritumoral stroma (p < 0.05; Fig. 5D), which is considered as one of the TGF-β-mediated effects in the TME. Similar to CTRL, mice which received α-PD-L1 Abs showed only small numbers of tumor-infiltrating CD8⁺ T cells (Fig. 5A and E). The levels of macrophages slightly increased and the levels of MDSCs significantly decreased in mice which received α-PD-L1 Abs (p < 0.05; Fig. 5A, F and G). The analysis of splenocytes revealed increased frequencies of NK cells (p < 0.05), while no differences were observed for cytotoxic T cells and MDSCs (Fig. 6A-C).

Figure 5.

Histopathologic analysis of 4-NQO tumors harvested from mice treated with α-PD-L1 Abs alone or in combination with TGF-β inhibitors. (A) Representative images of sections of 4-NQO mice. The images show HE-staining and immunofluorescence staining for TGF-β, PD-L1, and collagen IV (Col IV; green fluorescence) with DAPI counterstaining (blue fluorescence) at 10x magnification. Immunofluorescence staining for CD8a, CD68 and Gr-1 (green fluorescence) with DAPI counterstaining (blue fluorescence) is presented in 20x magnification. Scale bars: 100μm. PD-L1 staining presents the tumor tissue. CD68 and Gr-1 staining show surrounding stroma. Asterisks: tumor tissue; dotted lines: tumor/stroma border. (B) Quantitative analysis of TGF-β⁺ signals by immunofluorescent staining. Staining intensity in tumor tissues was quantified by using ImageJ and data are expressed as fold induction compared to CTRL. (C) Quantitative analysis of PD-L1⁺ signals in the tumor tissue. (D) Quantitative analysis of Col IV⁺ signals in the tumor stroma. (E) Numbers of tumor-infiltrating CD8a⁺ cells per region of interest (ROI). The dotted lines in A indicate the tumor borders, and only CD8a⁺ cells within the tumor tissue were counted. (F) Quantification of macrophage numbers in tongue tissues closely located to the 4-NQO tumors. Data are expressed as numbers of CD68⁺ cells per ROI. (G) Quantification of MDSC numbers in the tumor tissue and associated stroma. Data are expressed as numbers of Gr-1⁺ cells per ROI. Data of CTRL group is also presented in Figure 3. All values in this figure represent means ± SEM. *p < 0.05 vs. CTRL; **p < 0.01 vs. CTRL; ^#p < 0.05 vs. α-PD-L1.

Figure 6.

Analysis of splenocytes by flow cytometry at the experimental end-point. (A) Cytotoxic T cells were defined as CD45⁺CD3⁺CD8⁺. (B) NK cells were defined as CD45⁺NKp46⁺. (C) Myeloid-derived suppressor cells were defined as CD45⁺CD11b⁺Gr-1⁺. Data of CTRL group is also presented in Figure 3H-J. All values in this figure represent means ± SEM. **p < 0.01 vs. CTRL; ****p < 0.0001 vs. CTRL; ^#p < 0.05 vs. α-PD-L1.

These results suggest, that α-PD-L1 Abs effectively reduce tumor burden in 4-NQO-treated mice. However, the analysis of the peritumoral stroma shows several characteristics of a TGF-β signature, such as increased collagen production and an immune-excluded phenotype. In accordance with these findings, the TGF-β serum levels were significantly increased by α-PD-L1 Abs indicating a potential benefit of combinatorial blockade of PD-L1 and TGF-β.

Combinatorial therapy with α-PD-L1 Abs and TGF-β inhibitors enhances the anti-tumor immune response in 4-NQO-treated mice

To evaluate a potential benefit of a combined TGF-β/PD-L1 blockade, mice were simultaneously treated with α-PD-L1 Abs and novel TGF-β inhibitors as illustrated in Fig. 1A and B. This combination did not improve effects of PD-L1 monotherapy on the tumor burden. However, treatment with α-PD-L1 Abs and mmTGF-β2–7m further reduced the number of tumors (p < 0.05; Fig. 4A) as well as the weight loss of mice (p < 0.05; Fig. 4D). The increased TGF-β serum levels which were observed when mice were treated with α-PD-L1 Abs were significantly reduced when mice also received mRER (Fig. 4E). TGF-β levels were not only reduced in the serum, but also in the tumor tissue, as described above for monotherapy with TGF-β inhibitors (p < 0.05; Fig. 5B). The elevated Col IV levels in mice treated with α-PD-L1 Abs were significantly reduced when mice also received mRER or mmTGF-β2–7m (Fig. 4A and D). In comparison to PD-L1 monotherapy the combined PD-L1/TGF-β blockade resulted in an increased number of infiltrating CD8⁺ T cells (p < 0.05; Fig. 4A and E). The increase of lymphoid cell infiltration might be explained by therapy-induced changes of the collagen-rich peritumoral stroma. The number of macrophages in the tumor stroma was decreased in mice receiving combinatorial therapy (p < 0.05; Fig. 4A and F). The number of MDSCs in the tumor stroma was further reduced when combining α-PD-L1 Abs with mRER (p < 0.05; Fig. 4A and G). The analysis of splenocytes revealed, that mice treated with α-PD-L1 Abs and mRER had elevated NK cell frequencies, even further increased compared to the values of PD-L1 monotherapy (p < 0.05; Fig. 6B).

These results indicate that even without major differences in the tumor burden, the combinatorial PD-L1/TGF-β therapy improves the composition of the peritumoral stroma, as well as the systemic immune cell composition. It increases the infiltration of T cells into the tumor tissue and, therefore, enhances the potential of a response to α-PD-L1 Abs.

Discussion

Accelerated development of TGF-β inhibitors in recent years has led to numerous clinical trials in patients with cancer (NCT03834662, NCT01291784, NCT02452008). So far, no inhibitors have been approved by FDA for use in humans. Among these evaluated TGF-β inhibitors are: small molecule TGF-β receptor kinase inhibitors, antisense oligonucleotides, and synthetic peptides, or protein-based biologics, such as TGF-β or TGF-β receptor neutralizing antibodies, TGF-β ligand traps, or antibodies that interfere with TGF-β maturation [22,26]. In this study, we demonstrated therapeutic efficacy of two novel TGF-β inhibitors, mRER and mmTGF-β2–7m, in an orthotopic and immunocompetent oral carcinoma mouse model. The tumor progression in this model was significantly reduced, and both inhibitors potently reduced levels of soluble TGF-β in the serum. One limitation of our study was the complicated drug delivery of mmTGF-β2–7m by osmotic pumps, which probably resulted in the observed outliers. The promising results of this study indicate that further improvements of the drug delivery should be considered. It might be possible to inject mmTGF-β2–7m either alone or as a conjugate with the Fc domain of an antibody or with albumin to diminish renal filtration and facilitate higher blood circulation half-life [22].

TGF-β was chosen as a target of our therapies because of its ability to modulate the TME and to suppress anti-tumor immune responses [7]. In patients, high TGF-β levels are often associated with the signature of TGF-β signaling in cancer-associated fibroblasts (CAFs) in the fibroblast- and collagen-rich peritumoral stroma, where CD8⁺ T cells that were excluded from the tumour parenchyma are found [11]. This “immune excluded” phenotype has been previously observed in various tumors [27]. Inhibition of TGF-β signaling to enable effector T cells to come in contact with cancer cells is one of the therapeutic objectives expected to overcome this exclusion of T cells from the tumor parenchyma [28]. In this study, we have demonstrated the presence of higher numbers of infiltrating CD8⁺ T cells in mice treated with mRER or mmTGF-β2–7m inhibitors. Additionally, TGF-β is expected to directly inhibit functions of Th1 helper and cytotoxic T cells, suppress NK cell functions and activate MDSCs [29]. The latter needs to be addressed in future studies by performing functional suppression assays to further show that suppression of T cell activation by MDSCs is blocked by mRER or mmTGF-β2–7m. While we were able to show differences in the infiltration of immune cells after treatment with mRER or mmTGF-β2–7m a further characterization of these infiltrating cells would be helpful to understand their functions. For instance, analysis of macrophage polarization towards M1 or M2 as well as analysis of Granzyme B⁺, Perforin-1⁺, or Ki-67⁺ CD8⁺ T cells should be included in future studies. In this study, although we did not observe changes in the numbers of CD8⁺ T cells in the spleen of mice, treatment with mRER or mmTGF-β2–7m altered the frequency of NK cells and MDSCs, indicating a shift to a favourable immune response.

The 4-NQO model is an immunocompetent model of OSCC and thus, naturally, we focused on immune suppression and its reversal by the combination of TGF-β inhibitors and α-PD-L1 Abs. However, TGF-β has effects on various other cancer-associated pathways and the anti-tumor response was unlikely solely to an activated immune response. One example would be the active change of the tumor cell plasticity by these inhibitors, which would make them more susceptible to immune cell lysis. These non-immune-related effects need to be addressed in future studies.

Inhibitors for TGF-β have been developed for decades resulting in promising candidates which have been introduced in recent years. The most advanced TGF-β signaling antagonists are large molecules, such as monoclonal antibodies, which have been developed for cancer therapy or fibrotic disorders [30]. Both of these diseases are characterized by the presence of dense tissues with large amounts of extracellular matrix, which can potentially compromise the infiltration of large inhibitors, such as monoclonal antibodies. mRER and mmTGF-β2–7m are both smaller than typical IgG monoclonal antibodies and may enable better tissue penetration, especially mmTGF-β2–7m which is just 10 kDa in size (mRER is 70 kDa, whereas typical IgG monoclonal antibody is 150 kDa). The smaller size could also potentially facilitate their easier loading into nanoformulations such as nanoemulsions and liposomes. Other potential advantages of these inhibitors include their high potency, especially mRER which has been shown to neutralize TGF-β3 at picomolar concentrations [20], and high specificity. High potency may contribute to the therapeutic efficacy of mRER by enabling it to efficiently sequester TGF-β, even at low tissue concentrations of both the inhibitor and TGF-β. High specificity is an especially significant and unique property of mmTGF-β2–7m compared to kinase inhibitors, as this provides a means of selectivity targeting TβRII, something which has not been possible using receptor kinase inhibitors [20,22]. In the concentrations used, both inhibitors showed antagonistic activity in treated mice by reducing levels of TGF-β in tumor tissue and in sera without any signs of toxicity. However, significant work on drug pharmacokinetic and pathway inhibition threshold using pharmacodynamic biomarkers will be necessary to evaluate potential toxicities at higher doses and over sustained treatment periods [31].

Recent papers by Mariathasan et al. [11] and Tauriello et al. [7] link poor responses to the PD-1/PD-L1 blockade with enhanced TGF-β signaling in the TME and demonstrate that the blockade of TGF-β together with α–PD-L1 Abs reduced TGF-β signaling in stromal cells, facilitated T cell infiltration into the tumor, and provoked vigorous anti-tumor immunity and tumour regression [7,11]. These findings were recently confirmed by showing the beneficial effects of α-PD-L1 Abs combined with different classes of TGF-β inhibitors [32,33]. In this study we showed that monotherapy with α-PD-L1 Abs reduced disease progression in 4-NQO mice. These results are in agreement with other reports, which demonstrate that 4-NQO mice respond to the blockade of PD-1 [16,34]. However, the combinatorial therapy of α-PD-L1 Abs with TGF-β inhibitors was found to be more effective in stimulation of an anti-tumor immune response compared to treatment with α-PD-L1 Abs alone, indicating that the therapeutic efficacy of the PD-1/PD-L1 blockade can be further enhanced. Thus, lymphocyte infiltration into tumors was increased, while MDSCs and macrophage frequency was reduced. However, in terms of tumor burden the combination of α-PD-L1 Abs with mRER was not leading to significant differences compared to α-PD-L1 Ab treatment alone. We showed slightly decreased numbers and sizes of 4-NQO tumors, reduction in TGF-β tumor tissue and serum levels, modulation of the collagen-rich stroma, higher numbers of tumor-infiltrating lymphocytes and a drastic increase in NK cell frequency in the spleens. Further studies will be required to optimize timing to enable an effective therapeutic targeting which might also lead to improvements in tumor burden. Future studies should also focus on the full characterization of the pharmacokinetics and pharmacodynamics of the TGF-β inhibitors and evaluate the most beneficial timing and drug delivery of a combinatorial TGF-β/PD-L1 blockade. Similar to the studies by Mariathasan et al. [11] and Zhu et al. [6], it would be of major interest to identify the convergences of TGF-β effects with other signaling pathways in the TME to ensure the use TGF-β inhibitors in a most efficient way.