Alteration of the tumor stroma using a consensus DNA vaccine targeting Fibroblast Activation Protein (FAP) synergizes with anti-tumor vaccine therapy in mice

Abstract

Purpose

Fibroblast activation protein (FAP) is over-expressed in cancer-associated fibroblasts and is an interesting target for cancer immune therapy, with prior studies indicating a potential to impact the tumor stroma. Our aim was to extend this earlier work through development of a novel FAP immunogen with improved capacity to break tolerance for use in combination with tumor antigen vaccines.

Experimental Design

We used a synthetic consensus (SynCon) sequence approach to provide MHC class II help to support breaking of tolerance. We evaluated immune responses and anti-tumor activity of this novel FAP vaccine in pre-clinical studies, and correlated these findings to patient data.

Results

This SynCon FAP DNA vaccine was capable of breaking tolerance and inducing both CD8+ and CD4+ immune responses. In genetically diverse, outbred mice, the SynCon FAP DNA vaccine was superior at breaking tolerance compared to a native mouse FAP immunogen. In several tumor models, the SynCon FAP DNA vaccine synergized with other tumor-antigen specific DNA vaccines to enhance anti-tumor immunity. Evaluation of the tumor microenvironment showed increased CD8+ T cell infiltration and a decreased macrophage infiltration driven by FAP immunization. We extended this to patient data from the Cancer Genome Atlas, where we find high FAP expression correlates with high macrophage and low CD8+ T cell infiltration.

Conclusions

These results suggest that immune therapy targeting tumor antigens in combination with a micro-consensus FAP vaccine provides a two fisted punch inducing responses that target both the tumor microenvironment and tumor cells directly.

Introduction

For many solid tumors, the immune suppressive tumor stroma excludes T cell infiltration or activation(1). Major reasons for this include a lack of chemokine secretion, a lack of innate immune cell activation, aberrant blood vessel maturation and a dense extracellular matrix rich tumor stroma(1,2). CD8+ T cell function may be suppressed through expression of inhibitory factors by the tumor cells, T cell exhaustion due to constant antigen stimulation, and the presence of regulatory T cells that prevent proper CD8+ T cell killing(1).

Cancer associated fibroblasts (CAFs) are known to promote tumor immune evasion. Micro-environmental factors within the tumor, such as stress, growth factors, hypoxia, and cytokines, promote fibroblast activation within the tumor(2). These activated fibroblasts then actively remodel the matrix, secrete pro-inflammatory cytokines and proliferate rapidly. This ultimately results in an inflamed, suppressive tumor microenvironment that limits adaptive CD8+ responses. Accordingly, CAFs represent important targets for immune therapy.

Fibroblast activation protein (FAP) is a membrane-bound enzyme that is up-regulated in CAFs in over 90% of human carcinomas(3). Expression of FAP is up-regulated in activated fibroblasts associated with wound healing and cancer, but is low or absent in normal somatic tissues, supporting that it represents an ideal universal stromal antigen. FAP expression is high in breast cancer, pancreatic adenocarcinoma, ovarian carcinoma, lung carcinoma, prostate carcinoma, melanoma and others(4–8). FAP-expressing stromal cells are known to suppress anti-tumor immunity, since ablation of these cells from transgenic mice synergizes with some anti-cancer immune therapies(9,10).

Several immune-based therapies have been developed to deplete FAP-expressing cells. These include chimeric antigen receptor (CAR) therapy, vector-based vaccines, cell-based vaccines, DNA vaccines and immunotoxins(11). While chimeric antigen receptor (CAR) therapy for FAP is effective at slowing tumor progression, these CARs can cause lethal toxicity due to killing of multipotent bone marrow cells that express low levels of FAP(12–14). Alternative vaccine based approaches targeting FAP are capable of altering the tumor microenvironment and synergizing with tumor antigen targeted vaccines without these toxic effects (15–18).

Recent advances in DNA vaccine technology may make DNA vaccines an important complement or alternative to other immune targeted approaches because they 1) elicit immunity to multiple epitopes instead of relying on a single polyvalent binding site, 2) are not restricted by a particular HLA type, 3) are well tolerated in humans, 4) are stable and cost-effective to produce, 5) can be boosted without anti-vector immunity, and 6) do not integrate into the host DNA. While initial DNA vaccine formulation and delivery methods led to poor clinical immunogenicity, several recent optimizations have led to robust immune responses in non-human primates and humans(19). These optimizations include codon/RNA optimization for efficient translation, the addition of synthetic immunoglobulin leader sequences for improved protein secretion, and use of adaptive electroporation to enhance plasmid transfection(20–22). In addition, we recently described a synthetic consensus (SynCon) approach in which gene sequences are compared across species to generate a consensus sequence with 95% homology to native gene sequences(23). This approach maintains protein structure while introducing new neo-epitopes that can help break tolerance, thus improving on original DNA technology. Here, we utilize these important DNA vaccine optimizations to develop a novel DNA vaccine targeting FAP that drives both CD4 and CD8 T cells that increase the effectiveness of tumor immune therapy in vivo.

Methods

Animal Immunization

C57Bl/6, Balb/c and CD-1 outbred mice were purchased from Jackson Laboratory. All animal procedures were done in accordance with the guidelines from the Wistar Institute Animal Care and Use Committee (IACUC) and the NIH. Mice were immunized by injecting 30μL of DNA (μg quantities of DNA are indicated in figure legends) into the tibialis interior (TA) muscle, followed by delivery of two 0.1 Amp electric constant current square-wave pulses using the CELLECTRA^®-3P device (Inovio Pharmaceuticals). The vaccine schedule is indicated in each figure.

Tumor challenge studies

For tumor challenge studies, 50,000 TC-1 cells, 1,000,000 TRAMP-C2 cells, 200,000 Brpkp110 cells or 250,000 TSA cells were implanted subcutaneously into the right flanks of female C57Bl/6 mice (TC-1 and Brpkp110), male C57Bl/6 mice (TRAMP-C2) or female Balb/c mice (TSA). One week (for TC-1, Brpkp110 and TSA implantation) or four days (for TRAMP-C2 implantation) after implantation, mice were randomized into treatment groups. Mice were then immunized once weekly for a total of four immunizations. Tumors were monitored twice weekly, and measured using electronic calipers. Tumor volume was calculated using the formula: Volume=(π/6)*(height)*(width²). Mice were euthanized when tumor diameters exceeded 1.5cm.

Splenocyte and tumor infiltrating lymphocyte (TIL) isolation

Spleens from immunized mice were harvested in RPMI medium supplemented with 10% FBS. Splenocytes were dissociated using a stomacher, filtered and Red Blood Cells were lysed using ACK Lysis Buffer (LifeTechnologies). Cells were filtered through a 40μm filter, and counted and plated for staining or for ELISpots. Tumors were mechanically dissociated using a scalpel, and then incubated in a mixture of Collagenase I, II and IV (170mg/L, ThermoFisher), DNAseI (12.5mg/L, Roche), Elastase (25mg/L, Worthington) in a 50/50 mixture of Hyclone L-15 Leibowitz medium (ThermoFisher) and RPMI + 10% FBS + 1% Penicillin/Streptomycin. Dissociated cells were then filtered twice through a 40μm filter, and plated for stimulation and staining.

ELISpot assay

ELISpot assays were performed using the MABTECH Mouse IFN-γ ELISpot^PLUS plates. Briefly, 200,000 splenocytes were plated per well, and stimulated for 24 hours in the presence of peptides (15-mer peptides overlapping by 9 amino acids). Cells were stimulated with 5μg/mL of each peptide in RPMI + 10% FBS media. Spots were developed and quantified according to the manufacturer’s instructions. Media alone and Concanavalin A stimulated cells were used as negative and positive controls, respectively. Spot forming units (SFU) per million cells was calculated by subtracting the media alone wells from the peptide stimulated wells. Spots were quantified using an ImmunoSpot CTL reader.

Intracellular Cytokine Staining and Flow cytometry

Splenocytes or TILs were stimulated with native mouse FAP peptides for 5 hours with Protein Transport Inhibitor Cocktail (eBioscience). Cell stimulation cocktail (plus protein transport inhibitors) and complete media (R10) were used as positive and negative controls, respectively. During stimulation, cells were incubated with FITC α-mouse CD107a (clone 1D4B, Biolegend) to detect degranulation. After stimulation, cells were incubated with LIVE/DEAD violet to detect viability. Cells were then incubated with surface stain for 30 minutes at room temperature. Cells were then fixed and permeabilized using the FoxP3/transcription factor fixation/permeabilization kit (eBioscience). Cells were then incubated in intracellular stain for 1 hour at 4°C. A list of antibodies used is included in the Supplementary Methods. All samples were run on a 14- or 18-color LSRII flow cytometer (BD Bioscience), and analyzed using FlowJo software.

The Cancer Genome Atlas data analysis

Normalized RSEM RNAseq counts were downloaded through the GDAC data portal for the following human tumors (https://gdac.broadinstitute.org/): BLCA, BRCA, CESC, COADREAD, ESCA, GBM, HNSC, KIRC, KIRP, LIHC, LUAD, LUSC, OV, PAAD, PRAD, SKCM, STAD, THCA, and UCEC. We filtered out all samples marked as matched normal, and analyzed the RNAseq data using the CIBERSORT analytical tool, using the LM22 gene signature (https://cibersort.stanford.edu/)(24). We then correlated relative immune cell abundance with FAP RNA expression levels by dividing the samples into high FAP expression (top 25%) and low FAP expression (bottom 25%). We downloaded matched overall survival and progression free survival data from these datasets using the CGDS-R package from the cBioPortal.

Statistical Analysis

For TCGA survival data, the Hazard ratio comparing patients with high FAP expression (FAPhi, top 25%) to low FAP expression (FAPlo, bottom 25%) was calculated using the Mantel Haneszel approach. For survival data, the p value was calculated using the Gehan-Brelow-Wilcoxon method. For CIBERSORT data, the p value was calculated comparing FAPhi and FAPlo samples using a two-tailed student’s t-test. For human tumor staining analysis, correlation analysis was performed using linear regression.

For animal experiments, error bars represent the mean ± standard error of the mean (SEM). For experiments with more than two experimental groups, statistical significance was determined by one- or two-way ANOVA, followed by Tukey’s post-hoc HSD test. For animal experiments with only two groups, significance was determined using a two-tailed student’s t-test. For mouse tumor survival studies, significance was determined by Gehan-Breslow-Wilcoxon test.

Results

Design and in vitro expression of SynCon FAP DNA vaccine

FAP is a membrane-bound enzyme with a large extracellular domain and a small cytoplasmic tail and transmembrane domain. For our vaccine design, we constructed a plasmid that contains only the extracellular domain of FAP (amino acids 26-761) fused to an immunoglobulin E (IgE) leader sequence at the N terminus (Supplementary Figure 1A). We mutated the serine at position 624 to alanine to block both dipeptidyl peptidase and gelatinolytic activities (25–27) (Supplementary Figure 1A,B). To facilitate breaking tolerance to the native mouse FAP (mFAP) sequence, we designed a SynCon sequence using sequence alignment from various related species in the NCBI database, which provides some sequence diversity, but conserves structure. This SynCon sequence shares 95.1% homology to the mFAP sequence (Supplementary Figure 1B). We similarly generated a mFAP plasmid for comparison purposes which shares 100% homology to the mFAP sequence, and otherwise contains the same sequence optimizations. We were able to detect expression of both the native and SynCon FAP plasmids in vitro in the lysates of transfected 293T cells (Supplementary Figure 1C).

Immunogenicity of SynCon FAP DNA vaccine in C57Bl/6 mice

In order to determine whether the SynCon FAP DNA vaccine was immunogenic and capable of breaking tolerance in mice, we immunized C57Bl/6 mice with different doses of SynCon FAP DNA vaccine (5μg, 10μg, 25μg and 50μg) by intramuscular injection with EP (Figure 1A). Mice were immunized three times at two week intervals, and splenocytes were harvested for analysis one week after final immunization (Figure 1A). We performed interferon γ (IFN-γ) ELISpots using peptides exactly matched to the vaccine sequence (SynCon peptides), or peptides matched to the mFAP sequence (native peptides). C57Bl/6 mice generated robust IFN-γ ELISpots to both native FAP and SynCon FAP peptides, indicating that the SynCon FAP vaccine is capable of breaking tolerance in mice (Figure 1B,C). There was a dose-dependent effect of the vaccine against SynCon peptides; however, the dose-dependence for the native peptides reached maximum responses at the 10μg dose (Figure 1B,C). We therefore used a 10μg dose, and only show responses to mFAP peptides (which match mouse FAP 100%) for the remaining experiments.

Figure 1. Immunogenicity of SynCon mouse FAP vaccine in C57Bl/6 mice.

(A) Experimental setup. Mice were immunized three times at two-week intervals, and were sacrificed one week following final vaccination. Splenocytes were analyzed to examine T cell responses. (B,C) IFN-γ ELISpot responses to native mouse FAP peptides (B) or SynCon peptides matched to the vaccine sequence (C). (D,E) Intracellular cytokine staining of CD8+ (D) and CD4+ (E) T cells following stimulation with native mouse FAP peptides for 5 hours. The 10μg dose of FAP vaccine was used for this study. Significance was determined by a student’s t-test for panels D and E. *p<0.05, **p<0.01, ***p<0.001. N=5 mice per group, shown is a representative of two independent experiments.

To further evaluate the CD8+ and CD4+ cytokine responses generated against native FAP peptides, we performed intracellular cytokine staining on stimulated splenocytes (Figure 1D,E). We observed a significant increase in IFN-γ and TNF-α production in CD8+ T cells in SynCon FAP immunized mice compared to naïve control mice (Figure 1D). We next evaluated cytolytic potential of the CD8+ T cells generated by the SynCon FAP vaccine using the degranulation marker CD107a and the transcription factor T-bet, which is expressed in activated T cells (Figure 1D). We found a significant increase in CD8+ T cells that were simultaneously positive for IFN-γ, CD107a and T-bet in SynCon FAP vaccinated mice compared to naïve control mice, indicating that this vaccine induces production of effector T cells with cytolytic killing potential (Figure 1D). We also observed a significant increase in TNF-α production in CD4+ T cells in SynCon FAP immunized mice compared to naïve control mice (Figure 1E). There was a trend towards increased IFN-γ production in CD4+ T cells as well (Figure 1E).

Consensus DNA vaccine design is superior to the native FAP DNA in breaking tolerance and generating a CD8+ T cell response in genetically diverse mice

We next evaluated the capacity of this SynCon FAP DNA vaccine to generate immune responses in outbred mice (CD-1 ICR “Swiss” mice) in comparison to a mFAP DNA vaccine (Figure 2). These genetically diverse mice were used as an important indication of immune potency in a more relevant tolerance model for extrapolation to outbred populations such as humans. We immunized mice with 10μg of mFAP DNA vaccine or SynCon FAP DNA vaccine according to the schedule in Figure 1A, and evaluated immune responses by IFN-γ ELISpot (Figure 2A). While variability was observed between the mice, the overall immune response was higher for the SynCon FAP immunized group compared to the native FAP immunized group (Figure 2A,B). Overall, 14/15 mice in the SynCon FAP group, compared to 9/15 mice in the native FAP group, generated an immune response above 100 SFU/million splenocytes (Figure 2A). Responses observed in outbred mice (average of 407 SFU) were more diverse and higher than those observed in C57Bl/6 mice (average 195 SFU) (Figure 1B, Figure 2B).

Figure 2. Comparison of Native and SynCon FAP vaccines in CD-1 outbred mice.

(A) IFN-γ ELISpot responses to native mouse FAP peptides from individual CD-1 outbred mice in naïve control group (top), native mouse FAP vaccine group (middle) or SynCon mouse FAP vaccine group (bottom). Immunized mice received 10μg of DNA plasmid. The immunization schedule for these mice was the same as in Figure 2. (B) Total IFN-γ ELISpot responses from the mice immunized in panel (A), not separated by pool. (C) Endpoint binding titers from the mice in (A) against the native FAP protein (extracellular domain). Significance was determined by two-way ANOVA followed by Tukey’s HSD test for panel B. *p<0.05, **p<0.01, ***p<0.001. 10 mice were used in the naïve group, and 15 mice each were used in the Native FAP and SynCon FAP groups.

We also evaluated antibody responses in these mice using mFAP protein corresponding to the extracellular domain (amino acids 26-761) by ELISA (Figure 2C). Interestingly, the majority of the mice immunized with either native FAP vaccine or SynCon FAP vaccine generated robust antibody responses (Figure 2C). The percentage of mice in the SynCon FAP group that generated antibody responses was higher compared to the native FAP group (11/15 mice compared to 9/15 mice). However, the difference was not statistically significant.

Consensus FAP DNA vaccine comparison to native FAP DNA vaccine in C57Bl/6 and Balb/c mice

We next examined the difference in immune responses generated from the SynCon FAP vaccine compared to the native FAP vaccine in the commonly used mouse strains C57Bl/6 and Balb/c mice. We performed the same immunization schedule and vaccine dose in these mice (Supplementary Figure 2A, 3A).

In the C57Bl/6 strain, which tend to generate better Th1 responses over Th2 responses, we found that the SynCon FAP vaccine generated a similar IFN-γ ELISpot response compared to the native FAP vaccine (Supplementary Figure 2B). We found that these mice also generated similar IFN-γ, TNF-α and IFN-γ/T-bet/CD107a triple-positive CD8+ T cell responses to both native and SynCon FAP vaccines (Supplementary Figure 2C). However, the C57Bl/6 mice did generate improved IFN-γ and TNF-α CD4+ T cell responses to the SynCon vaccine compared to the native vaccine (Supplementary Figure 2D). In C57Bl/6 mice, the SynCon FAP vaccine did not improve antibody responses compared to native FAP vaccine (Supplementary Figure 2E). In fact, native FAP trended towards better antibody responses; however, this trend was not statistically significant.

In the Balb/c strain, which tend to generate better Th2 responses over Th1 responses, we found that the SynCon FAP vaccine generated superior IFN-γ ELISpot responses compared to the native FAP vaccine (Supplementary Figure 3B). We found that Balb/c mice generated better IFN-γ and TNF-α responses in both CD8+ and CD4+ T cells (however, this was only statistically significant for IFN-γ production in CD8+ T cells) (Supplementary Figure 3C,D). In addition, Balb/c mice generated more robust IFN-γ/T-bet/CD107a triple-positive CD8+ T cells upon immunization with SynCon FAP DNA vaccine compared to native FAP DNA vaccine (Supplementary Figure 3C). Strikingly, in Balb/c mice the native FAP vaccine did not generate any detectable antibody titers, while the SynCon FAP vaccine generated robust antibody levels in 4/5 mice (Supplementary Figure 3E).

These results indicate that the commonly used strains of mice may skew the results of immune based studies, and that use of a genetically diverse population will be important for clinical application of an immune therapy. Overall, the SynCon vaccine showed improvements in some immune aspect of breaking tolerance to native FAP antigen compared to the native FAP vaccine, both in the more immune tolerant Balb/c model and the immune responsive C57Bl/6 model.

Consensus FAP DNA vaccine synergizes with tumor antigen DNA vaccines in multiple tumor models

After establishing that robust IFN-γ and TNF-α immune responses are generated with increased frequency against native antigen for the SynCon FAP DNA vaccine, we next evaluated therapeutic efficacy of SynCon FAP in conjunction with a tumor associated antigen vaccine in tumor challenge models. We tested combination therapies with two vaccines that have been previously studied that target the tumor antigens PSMA or TERT (Figure 3, Supplementary Figure 4)(28,29). We implanted female C57Bl/6 mice with the lung tumor cell line TC-1 (Figure 3A), and began immunizations on day 7 after tumor implantation. Mice were either immunized with SynCon FAP DNA vaccine alone, mTERT (mouse TERT) vaccine alone, or a combination of SynCon FAP and mTERT tumor antigen vaccine, injected into the same leg. Mice were immunized once weekly for a total of four immunizations. For the TC-1 tumor model, the combination of FAP and mTERT generated the most robust anti-tumor activity and improvement in mouse survival compared to either vaccine alone (Figure 3B,C). To verify these results in a different tumor model, we performed a similar experiment using the SynCon FAP vaccine in combination with a PSMA vaccine in the TRAMPC2 prostate tumor model (Supplementary Figure 4A–C). We implanted male C57Bl/6 mice with TRAMPC2 tumor cells, and began immunizations on day 4 after tumor implantation (Supplementary Figure 4A). For the TRAMPC2 tumor model, SynCon FAP DNA vaccine alone had no impact on tumor growth, while the PSMA vaccine alone decreased tumor volume (Supplementary Figure 4B). However, the combination of PSMA and FAP decreased tumor volume and improved tumor survival more than the PSMA vaccine alone, indicating synergy between the two vaccines (Supplementary Figure 4B,C).

Figure 3. Efficacy of FAP vaccine and combination therapy in therapeutic lung tumor model.

(A) Experimental setup. Mice were implanted with TC-1 cells on day 0, randomized on day 7 and immunized once weekly for a total of 4 immunizations. 10μg of SynCon FAP DNA and 25μg of SynCon mouse TERT DNA was used. (B) Tumor volume measurements over time for indicated vaccination regimen for mice implanted with TC-1. (C) Mouse survival over time for indicated vaccination regimen for mice implanted with TC-1. (D) Western blot expression of mouse FAP in the mouse tumor cell lines TC-1 and TRAMP-C2. 293T cells transfected with native mouse FAP plasmid were used as a positive control. Significance for tumor volume measurements was determined by two-way ANOVA followed by Tukey’s HSD test. Significance for mouse survival was determined by Gehan-Breslow-Wilcoxon test. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. N=10 mice per group for TC-1 study. Shown is a representative of two independent experiments.

We confirmed that our SynCon FAP vaccine would only target cancer-associated fibroblasts by probing for expression of FAP in both the TC-1 and TRAMP-C2 cell lines (Figure 3D). We verified that these two cell lines do not express FAP.

We additionally tested the efficacy of our FAP vaccine alone in two stromally driven breast tumor models: Brpkp110 and TSA(30,31). Both of these tumor models responded to the FAP vaccine alone, resulting in reduced tumor burden, and, for mice implanted with the TSA tumor model, prolonged survival (Supplementary Figure 4D–F).

We next tested whether FAP specific immune responses were necessary for the synergy observed with the TERT DNA vaccines, or if this same effect could be observed with breakdown of the tumor extracellular matrix by other means. To do this, we tested combination therapy of our TERT vaccine with hyaluronidase treatment at the tumor site in TC-1 tumors (Supplementary Figure 5A). Hyaluronidase alone was able to modestly slow tumor growth, similar to the effect of the TERT vaccine alone. However, the combination therapy did not significantly further slow tumor growth compared to either vaccine alone or hyaluronidase alone, and did not prolong mouse survival (Supplementary Figure 5B,C). Thus, depletion of FAP expressing cells is more effective at synergizing with DNA vaccines compared to extracellular matrix breakdown alone.

Consensus FAP DNA vaccine induces FAP-specific tumor infiltrating lymphocytes

We next examined immune responses systemically and in the tumors of FAP-immunized tumor bearing mice. We implanted mice with TC-1 tumors, and began immunizations 7 days after tumor implant (Figure 4A). We immunized the mice twice at one-week intervals, and then sacrificed the mice one week after the final immunization (on day 21). For the immune analysis in tumor-bearing mice, we sacrificed the mice at an early time point to ensure that the mice would not be sick from tumor burden. We analyzed antigen-specific immune responses in both splenocytes and in tumor-infiltrating lymphocytes from these mice. Despite giving the mice fewer immunizations over a shorter period of time, the mice exhibited superior CD8+ IFN-γ and TNF-α production, as well as robust co-expression of CD107a and IFN-γ (Figure 4B). Furthermore, we observed robust FAP-specific T cell responses in tumor infiltrating lymphocytes as well, with a significant increase in IFN-γ, TNF-α and IFN-γ/CD107a co-production in CD8+ T cells within the tumor (Figure 4C). Furthermore, when we examined the relative proportion of CD8+ T cells and regulatory T cells (CD3+/CD4+/CD25+/FoxP3+ cells), we observed an increase in CD8+ T cells and a decrease in Tregs upon FAP immunization (Figure 4D).

Figure 4. FAP vaccine induces FAP-specific TILs.

(A) Experimental setup. Mice were implanted with TC-1 tumor cells on day 0, randomized on day 7 and immunized once weekly for a total of 2 immunizations. 10μg of SynCon FAP DNA was used. Mice were sacrificed on day 21, and splenocytes and TILs were harvested. (B) Intracellular cytokine staining of CD8+ T cells in the spleen following stimulation with native mouse FAP peptides for 5 hours. (C) Intracellular cytokine staining of tumor infiltrating lymphocytes (TILs) that were stimulated with native mouse FAP peptides for 5 hours. (D) Frequency of CD8+ T cells and CD4+/CD25+/FoxP3+ Tregs in each tumor, as a percentage of CD45+/CD3+ lymphocytes, assessed by flow cytometry staining. Significance was determined by a student’s t-test for panels B–D. *p<0.05, **p<0.01, ***p<0.001. N=9–10 mice per group, shown is a representative of two independent experiments.

We also compared immune responses in TC-1 tumor bearing mice receiving treatment with FAP vaccine alone, mTERT vaccine alone, or the combination therapy (Supplementary Figure 6A–D). As expected, mice receiving either FAP vaccine alone or mTERT vaccine alone induce robust CD8+ IFN-γ and TNF-α responses in both the spleen and tumor to FAP peptides or mTERT peptides, respectively (Supplementary Figure 6A–D). Interestingly, in mice receiving combination therapy with mTERT and FAP simultaneously, the responses were diminished compared to mice receiving each vaccine alone, suggesting antigen interference (Supplementary Figure 6A–D). Despite this antigen interference, there was still improvement in anti-tumor responses in the combination therapy group compared to each vaccine alone, suggesting that dual-targeting of fibroblasts and tumor cells is an important strategy for cancer immune therapy.

SynCon FAP DNA vaccine alters the immune microenvironment of TC-1 tumors, increasing the proportion of CD8+ T cells and reducing the proportion of macrophages in the tumor

We next examined the tumor microenvironment of TC-1 tumors upon immunization with SynCon FAP using both immunohistochemical approaches as well as flow cytometry (Figure 5, Supplementary Figure 7). We observed a decrease in the area of tumor sections covered by FAP-expressing cells and the amount of hyaluronan, an extracellular matrix glycosaminoglycan secreted by both fibroblasts and tumor cells, upon vaccination with SynCon FAP DNA (Figure 5A–D). We also observed a decrease in F4/80+ macrophage infiltration and an increase in CD8+ T cell infiltration upon FAP vaccination (Figure 5E–H). We also observed a decrease in the frequency of F4/80+/CD11b+ macrophages per tumor upon FAP vaccination by flow cytometry, but did not observe any change in the frequency of B cells, NK cells or dendritic cells upon FAP vaccination (Supplementary Figure 7A–D). In order to distinguish between the relative proportions of M1 polarized and M2 polarized macrophages upon FAP vaccination, we examined the expression of Arg1, MHCII, CD68, CD80 and CD86 in tumor infiltrating macrophages. We did not observe any differences in marker expression, suggesting that there was no skewing in macrophage polarization upon vaccination with SynCon FAP vaccine (Supplementary Figure 7E–I).

Figure 5. FAP vaccine alters the tumor microenvironment.

(A) Representative immunohistochemical staining of tissues from control mice or SynCon mouse FAP immunized mice for FAP expression. (B) Quantification of the percentage of area in the tumor covered by FAP-expressing cells. (C) Representative immunofluorescent images of tissues from control mice or SynCon mouse FAP immunized mice for hyaluronan expression. (D) Quantification of the percentage of area in the tumor covered by hyaluronan. (E) Representative immunofluorescent image of tissues from control mice or SynCon mouse FAP immunized mice for F4/80 and EpCAM expression. (F) Quantification of the percentage of area in the tumor covered by F4/80 expressing cells. (G) Representative immunofluorescent image of tissues from control mice or SynCon mouse FAP immunized mice for CD8α and EpCAM expression. (H) Quantification of the percentage of area in the tumor covered by CD8α expressing cells. N=6–8 mice per group. Image quantification was performed for at least 5 images per mouse. Significance was determined by a student’s t-test for panels B–D. *p<0.05, **p<0.01, ***p<0.001. Scale bar= 100μm.

FAP expression in patient samples correlates with poor patient outcome as well as enhanced macrophage infiltration and decreased CD8+ T cell infiltration

We sought to extend this analysis of FAP expression to data from patient samples. We analyzed human patient RNA-seq and clinical data available from the Cancer Genome Atlas (TCGA). We included datasets from 19 different tumor types, each containing at least 150 patient samples, from the BROAD GDAC. We correlated FAP RNA expression with progression-free survival and overall survival in these datasets. We compared patients in the top quartile of FAP expression (FAPhi) to the bottom quartile of FAP expression (FAPlo). Low FAP expression predicted significant prolonged overall survival for 6 of the 19 tumor types: GBM, LUSC, HNSC, STAD, KIRP and KIRC (Figure 6A). Low FAP expression predicted significant prolonged progression free survival for 5 of the 19 tumor types: KIRC, KIRP, PRAD, OV and STAD (Figure 6B).

Figure 6. FAP expression correlates with poor patient survival and altered immune cell infiltration.

Samples from 19 different tumor types in The Cancer Genome Atlas (TCGA) were examined for correlation of high FAP expression (top 25%) or low FAP expression (bottom 25%) with overall survival (A), progression free survival (B) and relative ratio of innate immune cell subsets, calculated using CIBERSORT (C). The Hazard Ratio in (A) and (B) was calculated comparing patients with high FAP expression to low FAP expression using the Mantel Haneszel approach. For survival data, the p value was calculated using the Gehan-Brelow-Wilcoxon method. For CIBERSORT data, the p value was calculated comparing FAPhi and FAPlo samples using a two-tailed student’s t-test. (D,E) Representative immunohistochemical staining of human squamous cell carcinoma patient tissue for FAP, CD163 and CD8. (F,G) Quantification of the percentage of tumor section covered by the indicated staining in (D) and (E), respectively. P values were calculated using linear regression. Scale bar= 50μm.

Next, we examined the impact of FAP expression on immune cell infiltration in the tumor microenvironment using the CIBERSORT tool for characterizing composition of tumor tissues from gene expression profiling data(24). Using these 19 datasets, we compared the relative proportion of immune cell subsets defined in the leukocyte signature matrix LM22 in FAPhi patients compared to FAPlo patients(24). We found that there was an increase in the proportion of macrophages in tumors with high FAP compared to low FAP expression across almost all tumor types examined (Figure 6C). This was true for all subsets of macrophages (M0, M1 and M2). High FAP expression significantly correlated with increased M0, M1 and M2 macrophage infiltration in 10/19, 7/19 and 15/19 tumor types, respectively (Figure 6C). We also observed that there was a decrease in the proportion of CD8+ T cells and monocytes in tumors with high FAP expression compared to tumors with low FAP expression (Figure 6C). High FAP expression significantly correlated with decreased CD8+ T cell infiltration in 8/19 cases, and with decreased monocyte infiltration in 9/19 cases (Figure 6C). High FAP only correlated with increased CD8+ T cell infiltration for one tumor type (OV) and decreased M1 macrophage infiltration for one tumor type (KIRC) (Figure 6C).

We also confirmed these findings in a human lung cancer patient cohort using immunohistochemical staining for FAP, CD8 and CD163 (Figure 6D–G). We found that FAP expression at the protein level negatively correlated with CD8 staining and positively correlated with CD163 staining, trending towards statistical significance (Figure 6D–G). These results are consistent with our analysis in mice, where we observed decreased macrophage infiltration and increased CD8+ T cell infiltration upon FAP immunization (Figure 5).

We did not observe any clear trend in the relative proportion of CD4+ T cells, resting NK cells or resting DCs in our TCGA dataset (Supplementary Figure 8A–D). However, we did observe that high FAP levels correlated with decreased T follicular helper cells, activated NK cells and activated DCs (Supplementary Figure 8A–D). This suggests that FAP may have additional influences on other immune cells in the tumor microenvironment. These findings may be interesting to explore in future studies.

Discussion

There has been a surge in interest in developing immune therapies targeting FAP-expressing cells(11). Here, we developed a DNA vaccine targeting FAP that incorporated novel improvements, including the use of SynCon sequences to help break tolerance. We previously demonstrated for the tumor antigen WT1, that a SynCon vaccine was superior at breaking tolerance and generating anti-tumor immunity in C57Bl/6 mice(23). In this study, we expanded on this finding using a surface antigen instead of an intracellular antigen. For the FAP immunogen, both humoral and cellular immune responses are important for anti-tumor immunity. We show that the SynCon strategy is effective at enhancing humoral immune responses for FAP in Balb/c mice compared to the native antigen. We further extended this concept using genetically diverse outbred mice to demonstrate that this consensus vaccine for FAP is superior to the native sequence in generating higher and more diverse cellular immune responses.

Importantly, the SynCon FAP DNA vaccine that we developed synergized with both TERT and PSMA tumor-targeting vaccines in generating more robust anti-tumor immunity than each vaccine alone. However, there were differences in the efficacy of SynCon FAP DNA vaccine as mono-therapy depending on the tumor model tested. The SynCon FAP DNA vaccine showed robust anti-tumor activity in the TC-1, Brpkp110 and TSA tumor models, but did not show any anti-tumor immunity in the TRAMP-C2 tumor model. This may reflect the inherent immunogenicity of the tumor cell line used for these studies. The related TRAMP-C1 is known to have very few neo-epitopes, indicating that this cell line may be poorly immunogenic on its own; however, the SynCon FAP vaccine may alter the milieu of the tumor microenvironment to allow the PSMA vaccine to have a more robust anti-tumor effect(32). The TC-1 tumor cell line whole genome sequence has not been published; however, it is known to express the E6 and E7 HPV16 oncoproteins, and thus may be more naturally immunogenic than the TRAMP-C2 tumor model. This is consistent with a previous report that showed greater efficacy of FAP CAR T cells for the TC-1 tumor model compared to other tumor models, including LKR, CT26, 4T1 and AE17.ova(12). Less is known about the potential immunogenicity of the Brpkp110 and TSA breast tumor models used; however, breast tumors are known to be stromally driven, and these tumor models have shown some responses to immune therapy in other studies(30,33,34). For future studies, it will be interesting to test these vaccine combinations in autochthonous tumor models with very low mutational burden to better mimic the physiology of spontaneously occurring human tumors.

We observed a decrease in the relative abundance of F4/80+ macrophages in the tumor microenvironment upon FAP vaccination. Secretion of pro-inflammatory cytokines and chemokines by CAFs is known to promote macrophage infiltration into tumors as well as skew the macrophage profile from M1 classically activated inflammatory macrophages to M2 alternatively activated cancer promoting macrophages(35–38). While we did observe a decrease in overall macrophage infiltration, examination of several M1 and M2 macrophage markers revealed no skewing towards M1 or M2 macrophages after immunization. This was also consistent in TCGA data, which showed that high FAP expression correlated with increased infiltration of all sub-types of macrophages, including M0, M1 and M2 macrophages, across a variety of carcinomas. Interestingly, studies have shown that subsets of tumor-associated macrophages, specifically a small subpopulation of F4/80+/CCR2+/CD206+ M2 macrophages, also express FAP(4,39). Both CAFs and FAP-expressing M2 macrophages were shown to promote tumor growth, since ablation of each cell type independently decreased tumor volume to a similar extent(39). The FAP+ macrophages only constituted approximately 10% of all macrophages in this mouse study; thus, it is unlikely that the decreased macrophage infiltration that we observe is a direct result of depletion of FAP+ macrophages, but rather the result of reduced macrophage recruitment by CAFs(39).

We have shown using TCGA data that expression of FAP correlates with poor prognosis or poor survival for several tumor types, including glioblastoma, lung, head and neck, stomach, kidney, prostate and ovarian carcinoma. Other studies using other datasets or protein expression data have also shown a correlation of FAP expression with poor outcome in pancreatic adenocarcinoma and colon adenocarcinoma(5,6,40). We did not observe a statistical correlation for these tumor types in TCGA data; however, this may be due to differences in the number of patients, type of cancer patients, or difference in RNA compared to FAP protein expression data. Importantly, no patients had a negative correlation with disease free or progression free survival with FAP expression in our study. This is not the case for a different marker for myofibroblasts in pancreatic cancer: α-smooth muscle actin (α-SMA)(41). α-SMA expression correlates with improved patient prognosis in pancreatic cancer patients(41). However, the opposite was true for esophageal carcinoma, hepatocellular carcinoma and breast carcinoma(42–44). Thus, this may be a reflection of the specific tumor microenvironment of pancreatic carcinoma patients, which has been shown to have distinct populations of fibroblasts that contribute to different aspects of tumor growth and the tumor microenvironment(45). These data highlight the heterogeneity in the CAF population and show the importance of examining the roles of specific CAF sub-populations in specific tumor types.

We and others have thus demonstrated that FAP is a viable therapeutic vaccine target for cancer immunotherapy, and shows particular efficacy when used in combination with tumor antigen vaccine therapy. Future improvements to this vaccine platform to further enhance anti-tumor immunity may include addition of immune plasmid adjuvants and/or combination therapies with immune checkpoint inhibitors.

Translational Relevance.

Fibroblast activation protein (FAP) is over-expressed in cancer-associated fibroblasts in 90% of all solid tumors and is thus considered a universal stromal antigen. Here, we evaluated a novel DNA vaccine targeting FAP in pre-clinical mouse models. We show that our micro-consensus vaccine design strategy improves immune responses compared to immunization with native FAP, and we show that our vaccine synergizes with anti-cancer immune therapy targeting the tumor antigens TERT and PSMA in lung and prostate tumor models. This vaccine used in combination may be effective for patients with high FAP expression in the tumor stroma.