FAK regulates antigen processing and presentation pathways

FAK kinase activity is elevated in human PDAC14 and FAK inhibitors, either alone or in combination with immunotherapies, can impair tumour growth in mouse models of PDAC.14 22 23 Clinical trials are now testing FAK inhibitors in combination with immune checkpoint inhibitors in patients with advanced pancreatic cancer (ClinicalTrials.gov; NCT02546531, NCT02758587, NCT03727880). However, our understanding of FAK as an immune modulator in PDAC is limited to the effects of FAK inhibitors on the tumour stroma and immune microenvironment in murine models of pancreatic cancer.14 23 Therefore, we set out to better understand how FAK regulates the antitumour immune response with particular focus on cancer cell-intrinsic mechanisms of immune evasion. We first used CRISPR-Cas9 gene editing to delete ptk2 (FAK gene) expression in Panc47 cells, a syngeneic cell line isolated from PDAC arising in fully back-crossed C57BL/6 KPC mice, and reconstituted wild-type FAK (FAK-wt) expression into a clonal Panc47 FAK-/- (herein termed FAK-/-) cell line (figure 1A). No increase in Pyk2 expression or phosphorylation on tyrosine-402 were observed following FAK loss. 0.5×10⁶ FAK-wt or FAK-/- cells were then implanted into the pancreas of C57BL/6 mice and tumours harvested and weighed after 2 weeks. FAK loss resulted in a tumour growth delay (figure 1B). We have previously shown using a murine model of skin squamous cell carcinoma that cancer cell-intrinsic deletion of FAK expression can promote an antitumour CD8 T-cell response via modulation of the effector CD8 T-cell: regulatory T-cell ratio in tumours.24 To determine whether the observed delay in FAK-/- tumour growth was also dependent on CD8 T-cells, C57BL/6 mice were treated with either isotype control or anti-CD8 T-cell-depleting antibodies and 0.5×10⁶ FAK-wt or FAK-/- cells were implanted into the pancreas. Tumours were harvested and weighed 2 weeks postimplantation. CD8 T-cell depletion restored a significant proportion of the delay in FAK-/- tumour growth, but had no effect on the growth of FAK-wt tumours (figure 1C). Thus, FAK-loss was sufficient to promote an antitumour CD8 T-cell response that could restrain tumour growth.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

FAK regulates antigen processing and presentation pathways in Kras^+/G12Dp53^+/R172HPDXcre pancreatic cancer cells. (A) Representative anti-FAK, FAK pY397, Pyk2 and Pyk2 pY402 western blot using whole cell lysates from parental Panc47 cells and FAK-wt and FAK-/- clonal cell lines. Anti-tubulin antibody used as a loading control. (B) Fold-change in tumour weight relative to FAK-wt tumours 2 weeks postimplantation into the pancreas of C57BL/6 mice. n=13–16 tumours per group. (C) Fold-change in tumour weight relative to FAK-wt tumours 2 weeks postimplantation into the pancreas of C57BL/6 mice. Mice were either treated with isotype control or anti-CD8 T-cell-depleting antibodies. n=4–6 mice per group. (D) Flow cytometry analysis of IFNγ expression in FAK-wt and FAK-/- tumours. Left, quantification of the frequency of CD45⁺IFNγ⁺ cells as a percentage of live cells; right, representative histograms of IFNγ staining in CD45^-, CD45⁺ and CD45⁺CD3⁺ cells. FMO=fully stained sample minus IFNγ antibody. n=4. (E) Cluster analysis of NanoString nCounter gene expression data acquired using the mouse PanCancer Immune Profiling panel. Clusters 1–3 are annotated with the top 20 most over-expressed genes in IFNγ-stimulated FAK-/- cells compared with FAK-wt cells and further annotated with immunoproteasome and antigen processing and presentation components identified in (F, G). (F) KEGG pathway enrichment analysis of genes in clusters 1–3 in E. (G) Functional association network analysis of hits in the top two most enriched pathways in (F). Network edges (connecting lines) represent reported physical (dark orange) or predicted (light orange) interactions. Pathway membership is delineated in grey. Nodes (circles) representing immunoproteasome components have thick node borders. (H) Relative quantification of Psmb8, Psmb9 and Psmb10 gene expression using qRT-PCR following stimulation with IFNγ. n=3. (I) Flow cytometry quantification of MHC-I surface expression on FAK-wt and FAK-/- cells following IFNγ stimulation. n=3. (J) Fold-change in tumour weight relative to untreated FAK-wt tumours 2 weeks postimplantation into the pancreas of C57BL/6 mice. Mice were treated daily with PBS or IFNγ by intraperitoneal injection from day 8 until day 14. n=7–14 mice per group. (K) Flow cytometry quantification of the frequency of CD45^-MHC-I⁺ cells in FAK-wt and FAK-/- tumours±IFNγ as a percentage of live cells. n=4–8 tumours per group. (L) Flow cytometry quantification of the median fluorescence intensity of MHC-I expression in CD45^-MHC-I⁺ cells in FAK-wt and FAK-/- tumours±IFNγ. n=4–8 tumours per group. IFNγ treatment was 10 ng/mL for 24 hours (E–H) or 72 hours (I). Unless otherwise stated, all data are represented as mean±SEM. Statistical significance in B, D, H and I was calculated using an unpaired t-test. Statistical significance in (C, J, K, L) was calculated using a one-way ANOVA with Tukey’s multiple comparison test. *p≤0.05, **p≤0.01, ***p≤0.001, ****p≤0.0001. ANOVA, analysis of variance; FAK, focal adhesion kinase; MHC-I, Major Histocompatibility Complex class-I.

IFNγ signalling plays an important role in promoting T-cell tumour recognition,16 and IFNγ is secreted by multiple cell types present within the tumour microenvironment (TME).17 We, therefore, sought to determine whether IFNγ was present within the TME of FAK-wt and FAK-/- tumours, and whether FAK deletion altered the response of PDAC cells to this important proinflammatory cytokine. 0.5×10⁶ FAK-wt or FAK-/- cells were implanted into the pancreas of C57BL/6 mice and tumours harvested 2 weeks later for flow cytometry analysis. FAK-/- tumours exhibited a significant increase in the number of cells positive for IFNγ expression when compared with FAK-wt tumours (figure 1D, left), with almost all of the IFNγ being secreted by immune (CD45⁺) cells (figure 1D, right). Interestingly, while T-cells were a source of IFNγ, they did not appear to be the major source in these tumours. To investigate whether FAK expression had any impact on the cancer cell response to IFNγ, we treated FAK-wt and FAK-/- cells in vitro with 10 ng/mL IFNγ and measured (1) the secretion of chemokines and cytokines, and (2) the expression of 770 immune-related genes. Forward-phase array profiling of tumour cell-secreted proteins identified a number of IFNγ-induced chemokines and cytokines likely to mediate paracrine signalling with the TME, several of which were downregulated in response to FAK loss and are protumourigenic, such as interleukin-6 (IL-6),25 26 in the context of pancreatic cancer (online supplemental figure 1A). We performed NanoString nCounter digital gene expression analysis using the PanCancer Immune Profiling Panel, which identified subsets of IFNγ-induced genes that were upregulated in FAK-/- cells compared with FAK-wt cells (figure 1E and online supplemental table 1). Pathway enrichment analysis of these gene clusters identified the proteasome and antigen processing and presentation as the two most significantly enriched pathways (figure 1F). Network analysis of these pathway hits showed direct association of the immunoproteasome components Psmb8, 9 and 10 with the antigen presentation module and enrichment in IFNγ-treated FAK-/- cells of several key pathway components, including Psmb8, Psmb9, Tap1 and H2-K1 (figure 1G and online supplemental figure 1B). Similar findings were observed when FAK-wt and FAK-/- cells were cultured in conditioned media from activated CD8 T-cells (online supplemental figure 1C and D and online supplemental table 2). qRT-PCR confirmed the upregulation of IFNγ-induced Psmb8 and Psmb9 expression in FAK-/- cells when compared with FAK-wt cells (figure 1H). Expression of Psmb10 was not regulated by FAK. Flow cytometry analysis confirmed the upregulation of IFNγ-induced MHC-I (including H2-K) expression on FAK-/- cells when compared with FAK-wt cells, with both an increase in the proportion of cells positive for expression of MHC-I and the level of MHC-I expression (figure 1I). To determine whether similar modulation of the IFNγ response also occurred in vivo, 0.5×10⁶ FAK-wt or FAK-/- cells were implanted into the pancreas of C57BL/6 mice and tumours allowed to develop for 7 days. On day 8, half of each cohort was dosed by intraperitoneal injection with vehicle (PBS) and the other half with IFNγ. Treatment was administered daily for a total of 7 days, and on day 14 tumours were harvested, weighed and analysed by flow cytometry. IFNγ treatment resulted in a small increase in the average weight of FAK-wt tumours; however, this was not statistically significant (figure 1J). In contrast, IFNγ treatment resulted in a significant decrease in the growth of FAK-/- tumours (figure 1J). Flow cytometry analysis of MHC-I expression showed a greater proportion of CD45^- cells (including tumour cells) positive for expression of MHC-I in both FAK-wt and FAK-/- tumours in response to IFNγ treatment (figure 1K). However, the median fluorescence intensity (MFI) of MHC-I-positive staining was significantly lower in FAK-wt tumours when compared with FAK-/- tumours, irrespective of IFNγ treatment (figure 1L), supporting in vitro findings that FAK loss upregulates MHC-I expression on the cell surface. Overall, these findings imply that FAK deletion reprogrammes the response of PDAC cells to IFNγ, resulting in increased expression of pathways important for T-cell tumour recognition.

FAK-dependent regulation of antigen processing and presentation is kinase-independent but requires FAK nuclear translocation

A number of FAK kinase inhibitors are currently in clinical development and are being tested in combination with immunotherapies for the treatment of PDAC.27 Therefore, to determine whether regulation of antigen processing and presentation pathways was dependent on FAK kinase activity we used three different FAK kinase inhibitors: GSK2256098, VS4718 and Defactinib. To identify the lowest dose required to achieve maximum inhibition of FAK phosphorylation on tyrosine-397, the autophosphorylation site used as a surrogate readout of FAK kinase activity, we first treated FAK-wt cells with a range of drug concentrations. This identified 100 nM of GSK2256098 and 500 nM of both VS4718 and Defactinib as optimal concentrations (figure 2A). We next treated FAK-wt cells with these concentrations of each inhibitor for either 2 days or 14 days and then stimulated cells with IFNγ in the presence of inhibitor. qRT-PCR showed that neither regulation of Psmb8, Psmb9 nor H2-Kb transcription was dependent on FAK kinase activity (figure 2B,C). Flow cytometry analysis also confirmed that regulation of MHC-I surface expression on FAK-wt cells was unchanged following treatment with FAK kinase inhibitors (figure 2D). To further support these findings, we also re-expressed a FAK kinase-deficient mutant, FAK-G431,28 into FAK-/- cells at similar levels to FAK-wt cells (figure 2E—left). FAK-wt, FAK-/- and FAK-G431 cells were stimulated with IFNγ and qRT-PCR used to quantify Psmb8, Psmb9 and H2-Kb gene expression (figure 2E—right). FAK-G431 cells expressed comparable levels of Psmb8, Psmb9 and H2-Kb to FAK-wt cells. Thus, FAK-dependent regulation of Psmb8, Psmb9 and H2-Kb is independent of FAK kinase activity.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

FAK-dependent suppression of antigen processing and presentation is independent of kinase activity but requires nuclear translocation. (A) Anti-FAK, FAK pY397, Pyk2 and Pyk2 pY402 western blots of FAK-wt cells treated with a range of concentrations of GSK2256098, VS4718 and Defactinib. Anti-tubulin used as a loading control. (B, C) Relative quantification of Psmb8, Psmb9 and H2-Kb expression in IFNγ stimulated FAK-wt cells using qRT-PCR following either 2 days or 14 days of treatment with FAK kinase inhibitors. n=3. (D) Relative quantification of the percentage of cells positive for MHC-I expression and the median fluorescence intensity (MFI) of MHC-I expression using flow cytometry. FAK-wt cells were treated with FAK kinase inhibitors for 2 days prior to IFNγ stimulation in the presence of inhibitor. n=3. (E) Left—Representative anti-FAK and anti-FAK pY397 western blot using whole cell lysates isolated from FAK-wt, FAK-/- and FAK-G431 cells. Anti-tubulin used as a loading control. Right—Relative quantification of Psmb8, Psmb9 and H2-Kb expression using qRT-PCR following IFNγ stimulation. n=3. (F) Left—Representative anti-FAK and anti-FAK pY397 western blot from whole cell and nuclear lysates isolated from FAK-wt, FAK-/- and FAK-NLS cells. Right—Relative quantification of Psmb8, Psmb9 and H2-Kb expression using qRT-PCR following IFNγ stimulation. n=3. IFNγ treatment was 10 ng/mL for 24 hours (B, C, E, F) or 72 hours (D). Data in (B–F) represented as mean±SEM. Statistical significance in (B–F) was calculated using a one-way ANOVA with Tukey’s multiple comparison test. **p≤0.01, ***p≤0.001, ****p≤0.0001. ANOVA, analysis of variance; FAK, focal adhesion kinase.

We have previously shown that FAK can localise to the nucleus where it can interact with transcription factors and transcriptional regulators to control chemokine and cytokine expression.24 We therefore re-expressed a FAK mutant deficient in nuclear translocation (FAK-NLS) into FAK-/- cells at comparable levels to FAK-wt cells (figure 2F—left) and stimulated FAK-wt, FAK-/- and FAK-NLS cells with IFNγ. qRT-PCR showed that FAK-dependent suppression of Psmb8, Psmb9 and H2-Kb transcription required nuclear FAK (figure 2F—right). Thus, FAK-dependent regulation of immunoproteasome and MHC-I genes is independent of FAK kinase activity but requires FAK nuclear translocation.

Psmb8 deletion restores FAK-/- tumour growth

Psmb8 is a key member of the immunoproteasome and is essential for maturation of the preproteasome containing Psmb9 and Psmb10, and subsequent acquisition of catalytic activity.29 30 Little is known about the immunoproteasome in the context of pancreatic cancer and whether it can contribute to the induction of antitumour T-cell responses. We, therefore, used CRISPR-Cas9 gene editing to delete Psmb8 expression in FAK-/- cells, generating two independent Psmb8 knockout clones termed C23 and C34 (figure 3A). Western blotting of whole cell lysates from IFNγ-stimulated FAK-wt, FAK-/-, FAK-/-Psmb8-/-C23 and FAK-/-Psmb8-/-C34 cells identified that Psmb8 knockout also resulted in loss of Psmb9 expression, but had no effect on Psmb10 (figure 3B). qRT-PCR further confirmed these results (figure 3C,D), suggesting that loss of Psmb9 expression was due to transcriptional downregulation. Psmb8 knockout also resulted in downregulation of MHC-I expression (figure 3E,F), implying that the activity of Psmb8 was important for sustaining the elevated MHC-I expression observed in FAK-/- cells. To determine whether Psmb8 upregulation contributed to the growth defect of FAK-/- tumours, 0.5×10⁶ FAK-wt, FAK-/-, FAK-/-Psmb8-/-C23 or FAK-/-Psmb8-/-C34 cells were implanted into the pancreas of C57BL/6 mice and tumours harvested and weighed after 2 weeks. Psmb8 knockout was sufficient to rescue the FAK-/- tumour growth delay (figure 3G), suggesting that Psmb8 upregulation was critical in restraining FAK-/- tumour growth. These findings were further validated using two independent shRNAs to deplete Psmb8 expression in FAK-/- cells (online supplemental figure 2A), both of which promoted the growth of FAK-/- tumours (online supplemental figure 2B).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Psmb8 deletion restores FAK-/- tumour growth. (A) Relative quantification of Psmb8 expression using qRT-PCR following IFNγ stimulation. n=3. (B) Representative western blot showing expression of Psmb9, Psmb10, FAK pY397 and FAK in FAK-wt, FAK-/-, FAK-/-Psmb8-/-C23 and FAK-/-Psmb8-/-C34 cells following IFNγ stimulation. Anti-tubulin was used as a loading control. (C) Relative quantification of Psmb9 expression using qRT-PCR following IFNγ stimulation. n=3. (D) Relative quantification of Psmb10 expression using qRT-PCR following IFNγ stimulation. n=3. (E) Flow cytometry quantification of the frequency of cells positive for MHC-I expression as a percentage of live cells following IFNγ stimulation. n=3. (F) Flow cytometry quantification of the median fluorescence intensity of MHC-I expression following IFNγ stimulation. n=3. (G) Fold change in tumour weight relative to FAK-wt tumours 2 weeks postimplantation into the pancreas of C57BL/6 mice. n=6–9 tumours per group. IFNγ treatment was 10 ng/mL for 24 hours (A–D) or 72 hours (E and F). All data represented as mean±SEM. Statistical significance was calculated using a one-way ANOVA with Tukey’s multiple comparison test. *p≤0.05, **p≤0.01, ***p≤0.001, ****p≤0.0001. ANOVA, analysis of variance; FAK, focal adhesion kinase; MHC-I, Major Histocompatibility Complex class-I.

In addition to the constitutive and immunoproteasomes, two further proteasomes have been identified and termed intermediate proteasomes.19 These contain either Psmb8 or Psmb8 and Psmb9 from the immunoproteasome, with the remaining catalytic subunits coming from the constitutive proteasome. Therefore, we also investigated the requirement for Psmb9 in regulating the growth of FAK-/- tumours. CRISPR-Cas9 gene editing was used to delete Psmb9 expression, generating two Psmb9 knockout clones termed C2 and C6 (online supplemental figure 3A). Psmb9 deletion resulted in a reduction in Psmb8 expression. However, Psmb8 expression was still significantly higher in FAK-/-Psmb9-/-C2 and C6 cells when compared with FAK-wt cells (online supplemental figure 3B). The percentage of cells positive for expression of MHC-I was variable between Psmb9-/- clones but remained significantly higher than FAK-wt cells (online supplemental figure 3C). A small decrease in the MFI of MHC-I expression was observed in FAK-/-Psmb9-/-C2 and C6 cells when compared with FAK-/- cells; however, MHC-I expression was still significantly higher than in FAK-wt cells (online supplemental figure 3D). To determine whether Psmb9 upregulation in FAK-/- cells contributed to the defect in tumour growth, 0.5×10⁶ FAK-wt, FAK-/-, FAK-/-Psmb9C2 or FAK-/-Psmb9C6 cells were implanted into the pancreas of C57BL/6 mice and tumours harvested and weighed after 2 weeks (online supplemental figure 3E). Psmb9 depletion did not rescue FAK-/- tumour growth, implying that an intermediate proteasome containing Psmb8 is sufficient to restrain FAK-/- tumour growth.

FAK regulates the antigen repertoire in a Psmb8-dependent manner

To better understand how FAK and Psmb8 might act to restrain tumour growth, we next profiled the antigen repertoire using mass spectrometry (MS)-based immunopeptidomics. 1×10⁸ FAK-wt, FAK-/-, FAK-/-Psmb8-/-C23 and FAK-/-Psmb8-/-C34 cells were stimulated with IFNγ for 24 hours, then lysed as detailed in materials and methods. MHC-I (specifically, H2-Kb) was immunoprecipitated using 20 mg of protein lysate and bound peptides eluted for analysis by MS. A total of 144 peptide antigens were identified from FAK-wt cells, 613 from FAK-/- cells, 116 from FAK-/-Psmb8-/-C23 cells and 68 from FAK-/-Psmb8-/-C34 cells. Comparison of peptide antigens from FAK-wt and FAK-/- cells identified 62 peptides specific to FAK-wt cells, 82 common to both FAK-wt and FAK-/- cells and 531 peptides specific to FAK-/- cells (figure 4A and online supplemental table 3). Further comparisons between FAK-/- and FAK-/-Psmb8-/-C23 and C34 cells suggested that a substantial proportion of the unique peptides presented by FAK-/- cells were dependent on Psmb8 expression (figure 4B and online supplemental table 3). Thus, FAK deletion increases the diversity of the antigen repertoire presented by Panc47 cells in a Psmb8-dependent manner.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

FAK regulates the antigen repertoire in a Psmb8-dependent manner. (A) Venn diagram showing common and unique antigen peptides identified from IFNγ treated FAK-wt and FAK-/- cells using MS. (B) Venn diagram showing common and unique peptides from IFNγ treated FAK-/- vs FAK-/-Psmb8-/-C23 cells (left) and FAK-/- vs FAK-/-Psmb8-/-C34 cells (right). (C) The percentage of peptide antigens predicted to bind strongly to MHC-I molecules normalised to sample size. (D) Analysis of amino acid frequency and hydrophobicity of the C-terminal residue of each peptide antigen identified from FAK-wt, FAK-/-, FAK-/-Psmb8-/-C23 and FAK-/-Psmb8-/-C34 cells. (E) Bar chart showing the frequency of peptides containing a hydrophobic C-terminal residue as a proportion of the total peptide number. (F) Weighted average hydrophobicity score representing all hydrophobic residues in each sample. (G) Log₂ fold change in the expression of antigen peptides common between FAK-wt and FAK-/- cells grouped by protein of origin. Blue, enriched in FAK-wt cells; red, enriched in FAK-/- cells. IFNγ treatment was 10 ng/mL for 24 hours. FAK, focal adhesion kinase.

Assembly of the immunoproteasome results in increased tryptic and chymotryptic activity, with a concomitant decrease in caspase-like activity. As a consequence, there is a preference for C-terminal cleavage at basic and hydrophobic residues. MHC-I preferentially binds peptides with basic or hydrophobic C-termini, suggesting that the immunoproteasome likely yields peptides with a higher affinity for binding MHC-I.19 Our data identifying an important role for Psmb8 but not Psmb9 in regulating FAK-/- tumour growth implied that an intermediate proteasome containing β1, β2 and Psmb8 may be sufficient to increase antigen diversity. However, this intermediate proteasome retains caspase-like activity via the β1 subunit, rendering it difficult to predict whether Psmb8 expression could influence the MHC-I binding affinity of the antigen repertoire presented by FAK-/- cells. To understand the binding affinity landscape in this model, we used NetMHCpan-4.0 to predict the binding affinity of each peptide in the immunopeptidomics profiles gathered. Peptides were classified based on their predicted percentage rank,31 which ranks stronger binders in lower percentiles. Strong binders are those peptides in the top 0.5% percentile, while weak binders are in the top 2%. The number of strong binders was then normalised to the number of peptides in the sample. Forty percent of all peptides identified from FAK-/- cells were predicted to bind H2-Kb strongly, compared with fewer than 10% of all peptides identified from FAK-wt, FAK-/-Psmb8-/-C23 and FAK-/-Psmb8-/-C34 (figure 4C and online supplemental table 4). Hence, Psmb8 expression in response to FAK loss resulted in an antigen repertoire predicted to bind H2-Kb with higher affinity.

To better understand why peptide antigens presented by FAK-/- cells might have higher affinity binding to H2-Kb, we next looked at important physicochemical properties of the peptides. Typical peptide binding lengths for MHC-I are 8–9 amino acids.32 Peptides identified from FAK-/- cells were consistently shorter than those from FAK-wt, FAK-/-Psmb8-/-C23 and FAK-/-Psmb8-/-C34 cells when comparing their length distributions (pairwise permutation test, 95% CI, p<0.05). Indeed, 53.5% of all peptides identified from FAK-/- cells were 8 amino acids long and 22.5% were 9 amino acids long (online supplemental figure 4 and online supplemental table 3). In contrast, only 37.5% of all peptides identified from FAK-wt cells were 8 amino acids long and 15.3% were 9 amino acids long. An even more pronounced reduction in the frequency of 8 and 9mer peptides was evident in FAK-/-Psmb8-/-C23 and C34 cells, suggesting that Psmb8 expression was important in the generation of peptide antigens with optimal length for binding MHC-I.

MHC-I preferentially binds peptides with either hydrophobic or basic C-termini.19 We, therefore, calculated the hydrophobicity score of the C-terminal amino acid for each peptide according to the Kyte-Doolittle scale.33 The C-terminal hydrophobicity of peptides differed significantly between samples (pairwise permutation test, 95% CI, p<0.05). Peptides identified from FAK-/- cells displayed a greater proportion of hydrophobic residues at their C-termini than those identified from any of the other samples, with leucine and valine residues dominating (53.3% and 10.3%, respectively) (figure 4D–F and online supplemental table 5). Notably, these are among the most hydrophobic of amino acids according to the assigned hydrophobicity score. While leucine was the most common C-terminal residue for peptides in all samples, the proportion of peptides containing a C-terminal leucine residue was lower in all other samples when compared with FAK-/-. Furthermore, the second most common hydrophobic residue in the FAK-/- sample was valine with a hydrophobicity score of 4.2, while in all other samples it was either methionine or alanine which have notably lower hydrophobicity scores than valine (4.2 vs 1.9 for methionine and 1.8 for alanine). These samples also showed an increase in peptides containing hydrophilic/polar C-terminal residues when compared with FAK-/- (figure 4D and online supplemental table 5). Hydrophobic C-terminal residues are generated through chymotryptic activity, a function of Psmb8.19 These data suggest that increased expression of Psmb8 as a consequence of FAK loss results in a preference for cleavage at hydrophobic C-terminal residues, further optimising peptide antigens for high affinity binding to MHC-I.

Having assessed potential differences in the physicochemical properties of the antigen peptides present in each sample, we next evaluated differences in the expression of those peptides that were commonly expressed between FAK-/- and FAK-wt cells. The number of unique spectra mapping onto each peptide was calculated and normalised as described in the methods section (online supplemental figure 5 and online supplemental table 6). There were significant differences in the expression of peptides common to the two samples (Welch’s two-sample t-test, 95% CI, p<0.05), with 58 out of 82 peptides being upregulated in the FAK-/- sample, 6 downregulated and 18 remaining unchanged in expression, taking 1.5-fold (log₂ fold change=0.58) as the cut-off for the upregulated peptides and 0.5-fold (log₂ fold change = −1) as the cut-off for downregulated peptides. In some cases, multiple peptides were found to have originated from the same protein, resulting in a total of 61 proteins being represented by the peptides common to both samples. We, therefore, also grouped peptides based on the protein from which they originated, and summed the unique spectra mapping onto each peptide as a total measure of protein expression (figure 4G and online supplemental table 7). In total, 45 proteins were upregulated in the FAK-/- sample, 4 were downregulated and 12 were unchanged using the same cut-offs as those applied to the peptide analysis.

In order to determine whether any of the peptides detected were neoantigens, whole genome sequencing datasets derived from FAK-/- cells were converted to protein FASTA files (online supplemental table 8) and used to search for the presence of mutated peptides in the immunopeptidomics datasets. No mutated peptides were identified.

FAK and STAT3 impair STAT1-dependent expression of Psmb8 and MHC-I

To explore the mechanism underpinning FAK-dependent regulation of these pathways, we first used flow cytometry to quantify IFNγ receptor 1 and 2 (IFNγR1 and IFNγR2) cell-surface expression on FAK-wt and FAK-/- cells untreated or treated with 10 ng/mL IFNγ for 24 hours (online supplemental figure 6). No difference in the expression of IFNγR1 or IFNγR2 was observed, suggesting that regulation of IFNγ receptor expression was not contributing to the observed phenotype following FAK loss. Downstream of IFNγ receptor activation, the Signal Transducer and Activation of Transcription family member STAT1 plays an important role in driving gene expression. Reciprocal regulation of STAT1 by STAT3 can impair STAT1 activity, suggesting that STAT3 controls the balance between activation of these transcription factors and subsequent downstream signalling.34 Tyrosine phosphorylation of STATs is crucial for IFN-mediated signalling and translocation to the nucleus.35 We, therefore, asked whether FAK could regulate STAT1/3 expression or tyrosine phosphorylation as a mechanism of controlling MHC-I and Psmb8 expression. Western blotting using whole cell lysates from FAK-wt and FAK-/- cells either untreated or treated with 10 ng/mL IFNγ identified that STAT1 was not constitutively expressed in either cell type, but that expression and phosphorylation on tyrosine-701 was induced in response to IFNγ. Little to no difference was observed in either the expression or phosphorylation of STAT1 when comparing FAK-wt and FAK-/- cells (figure 5A). In contrast, STAT3 was constitutively expressed and phosphorylated on tyrosine-705 in both FAK-wt and FAK-/- cells, with both expression and phosphorylation increasing in response to IFNγ. No difference in either expression or phosphorylation of STAT3 was observed between FAK-wt and FAK-/- cells. Thus, FAK does not regulate STAT1/3 expression or phosphorylation in these cells. To investigate whether FAK might regulate STAT1 or STAT3 subcellular localisation we performed confocal immunofluorescence studies in FAK-wt and FAK-/- cells treated with IFNγ (figure 5B). STAT3 localisation was restricted to the nucleus whereas STAT1 showed both nuclear and cytoplasmic staining. Quantification of STAT1 nuclear staining did not identify any difference between FAK-wt and FAK-/- cells (figure 5C). These findings show that FAK does not regulate the subcellular localisation of either STAT1 or STAT3.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

FAK and STAT3 impair STAT1-dependent expression of Psmb8 and H2-Kb. (A) Representative western blot of STAT1 pY701, STAT1, STAT3 pY705, STAT3 and FAK expression in FAK-wt and FAK-/- cells±stimulation with IFNγ. Anti-tubulin used as a loading control. (B) Representative confocal fluorescence images of IFNγ treated FAK-wt and FAK-/- cells stained with DAPI, anti-STAT3 antibody and anti-STAT1 antibody. Scale bar=100 µm. (C) Quantification of nuclear anti-STAT1 immunofluorescence staining from 5000 FAK-wt and FAK-/- cells treated with IFNγ. (D) Representative western blot showing anti-FAK IP blotted with either anti-STAT1, anti-STAT3 or anti-FAK antibodies. Anti-tubulin used as a loading control. Cells were stimulated with IFNγ. (E) Representative western blot showing anti-STAT1 and anti-STAT3 IPs blotted with either anti-STAT1 or anti-STAT3 antibodies. Anti-FAK and anti-FAK pY397 were used to confirm FAK expression and activation, respectively, and anti-tubulin used as a loading control. Cells±stimulation with IFNγ. (F) Relative quantification of Psmb8 gene expression using qRT-PCR following stimulation with IFNγ. n=3. (G) Relative quantification of H2-Kb gene expression using qRT-PCR following stimulation with IFNγ. n=3. (H) Relative quantification of STAT1-Psmb8 promoter binding using anti-STAT1 chromatin immunoprecipitation (CHIP) and Psmb8 promoter-specific qRT-PCR from cells treated with IFNγ. n=3. (I) Representative western blot showing anti-FAK IP blotted with anti-STAT1 antibody in whole cell lysates from cells treated with IFNγ. (J) Representative western blot showing anti-STAT1 IP blotted with an anti-FAK antibody from whole cell lysates isolated from cells treated with IFNγ. IFNγ treatment was 10 ng/mL for 24 hours. Data in (C, F, G, H) represented as mean±SEM. Statistical significance in (F, G, H) calculated using a one-way ANOVA with Tukey’s multiple comparison test. ***p≤0.001, ****p≤0.0001. Statistical significance in C calculated using an unpaired two-tailed t-test. ANOVA, analysis of variance; FAK, focal adhesion kinase; IP, immunoprecipitation.

We next checked whether FAK could interact with STAT1 and/or STAT3 as a potential mechanism of regulation. Immunoprecipitation (IP) of FAK from IFNγ-stimulated FAK-wt or FAK-/- whole cell lysates, followed by western blotting identified that FAK was in complex with both STAT1 and STAT3 (figure 5D). STAT1 and STAT3 are known to form a heterodimer and STAT3 can impair STAT1 transcriptional activity.34 35 We, therefore, performed further IP studies to determine whether FAK-expression status could influence STAT1/3 heterodimer formation. Both STAT1 and STAT3 IPs confirmed the presence of a STAT1/3 heterodimer following IFNγ-stimulation and showed a reduction in heterodimer formation in FAK-/- cells when compared with FAK-wt cells (figure 5E). Collectively, these findings imply that FAK stabilises the STAT1/3 heterodimer in the nucleus.

To investigate the role of STAT1 and STAT3 in regulating expression of MHC-I and Psmb8, we next used shRNA to deplete either STAT1, STAT3 or STAT1 and STAT3 in FAK-wt and FAK-/- cells. Successful knockdown of expression was confirmed using western blotting (online supplemental figure 7). The resulting panel of cell lines were stimulated with IFNγ for 24 hours and RNA extracted for qRT-PCR analysis. qRT-PCR using a primer set specific for Psmb8 identified a significant STAT1-dependent induction of Psmb8 expression only in response to co-depletion of FAK and STAT3 (figure 5F). Similar findings were observed for H2-Kb expression (figure 5G). Therefore, while the induction of these genes in response to FAK depletion alone was not fully dependent on STAT1, the co-depletion of FAK and STAT3 appeared to hyperactivate both Psmb8 and H2-Kb expression in a STAT1-dependent manner. Given that STAT1-dependent hyperactivation of these pathways only occurred in the absence of FAK and STAT3 we next performed chromatin IP using an anti-STAT1 antibody and qRT-PCR using primers specific to the promoter of Psmb8.36 STAT1 was observed to bind to the Psmb8 promoter in all cell lines tested when compared with IgG control and neither FAK nor STAT3 expression status significantly altered the extent of promoter binding (figure 5H). Therefore, changes in STAT1-Psmb8 promoter binding did not explain the STAT1-dependent hyperactivation observed. We next asked whether the FAK-STAT1 interaction was dependent on STAT3. IPs using both anti-FAK and anti-STAT1 antibodies identified that FAKs interaction with STAT1 was independent of STAT3 (figure 5I,J). Thus, the only scenario in which STAT1 was not in complex with FAK and / or STAT3 was in the FAK-/- STAT3 shRNA cells, implying that both FAK and STAT3 can exert inhibitory effects on the transcriptional function of STAT1, at least in the context of Psmb8 expression.

Given that depletion of STAT1 alone or in combination with STAT3 in FAK-/- cells did not revert Psmb8 and H2-Kb expression to levels comparable with FAK-wt cells, we next investigated other potential regulators of MHC-I gene expression. The promoters of MHC-I genes contain a number of cis-regulatory elements including an enhancer A element which contains nuclear factor-κ B (NF-κB) binding sites, an interferon-stimulated response element which contains binding sites for IRF1 and an SXY module that is required for assembly of the NLRC5-enhanceosome.37 Further analysis of NanoString gene expression data from figure 1E identified that the NF-κB genes Nfkb1 and Nfkb2 were expressed in FAK-wt and FAK-/- cells, but poorly stimulated by IFNγ (figure 6A). In contrast, Irf1 was highly stimulated by IFNγ in both FAK-wt and FAK-/- cells (figure 6B), leading us to focus on IRF1 as a candidate transcription factor responsible for expression of both Psmb8 and H2-Kb. Western blotting of protein lysates from FAK-wt and FAK-/- cells±IFNγ confirmed a substantial induction of IRF1 protein expression following IFNγ stimulation, but no difference was observed between FAK-wt and FAK-/- cells (figure 6C). Additionally, CRISPR-Cas9 mediated deletion of IRF1 in FAK-wt and FAK-/- cells identified that expression of STAT1 in response to IFNγ was partially dependent on IRF1 in both cell types, but that STAT3 expression was independent of IRF1 (figure 6C). To investigate whether IRF1 was important for Psmb8 and H2-Kb expression we next treated FAK-wt, FAK-/-, FAK-wt IRF-/- and FAK-/- IRF1-/- cells with IFNγ and quantified Psmb8 and H2-Kb expression using qRT-PCR (figure 6D). IRF1-deletion in both FAK-wt and FAK-/- cells completely abolished Psmb8 expression and significantly reduced the expression of H2-Kb. IRF1-deletion also abolished the STAT1-dependent hyperactivation of Psmb8 and H2-Kb in FAK-/- STAT3 shRNA cells (figure 6E,F). Collectively, these findings imply that IRF1 expression is critical for downstream expression of Psmb8 and H2-Kb. To better understand how FAK may impact IRF1-dependent transcription, we next performed chromatin IP using an anti-IRF1 antibody and qRT-PCR using primers specific for the promoter of Psmb8 (figure 6G). IRF1 was observed to bind the Psmb8 promoter at similar levels in both FAK-wt and FAK-/- cells, suggesting a mechanism based on enhancing IRF1 transcriptional function rather than differential promoter binding. Further analysis of NanoString gene expression data from figure 1E also identified expression of several genes encoding proteins belonging to the NLRC5-enhanceosome. Creb1 and Atf1, two transcription factors belonging to the NLRC5-enhancesome, were both expressed in FAK-wt and FAK-/- cells, but not stimulated by IFNγ or regulated by FAK (figure 6H,I). In contrast, Nlrc5 was only expressed in response to IFNγ stimulation and was upregulated in FAK-/- cells when compared with FAK-wt cells (figure 6J,K). NLRC5 is a transcriptional co-activator that has been shown to induce the expression of MHC-I genes, including H2-K, and also MHC-I accessory genes including Psmb8, Psmb9 and Tap 1.37 38 Thus, FAK likely regulates IRF1-dependent transcription of MHC-I and MHC-I accessory genes through regulation of the NLRC5-enhanceosome in the context of IFNγ stimulation.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6

Expression of Psmb8 and H2-Kb requires IRF1. (A, B) NanoString nCounter gene expression data showing the expression of Nfkb1, Nfkb2 and Irf1 in FAK-wt and FAK-/- cells±IFNγ stimulation. (C) Anti-IRF1, STAT1, STAT3 and FAK western blot from whole cell lysates isolated from FAK-wt and FAK-/- cells±IRF1 CRISPR-Cas9±IFNγ stimulation. Anti-tubulin used as a loading control. (D, E) Relative quantification of Psmb8 and H2-Kb expression using qRT-PCR following IFNγ stimulation. n=3. (F) Relative quantification of H2-Kb expression using qRT-PCR following IFNγ stimulation. (G) Relative quantification of IRF1-Psmb8 promoter binding using anti-IRF1 chromatin immunoprecipitation (CHIP) and Psmb8 promoter-specific qRT-PCR from cells treated with IFNγ. n=3. (H, I, J) NanoString nCounter gene expression data showing the expression of Creb1, Atf1 and Nlrc5 in FAK-wt and FAK-/- cells±IFNγ stimulation. (K) Relative quantification of Nlrc5 expression using qRT-PCR following IFNγ stimulation. Cells were treated with 10 ng/mL IFNγ for 24 hours. n=3. Data in (D–G, K) represented as mean±SEM. Statistical significance was calculated using a one-way ANOVA with Tukey’s multiple comparison test. *p≤0.05, **p≤0.01, ***p≤0.001, ****p≤0.0001. ANOVA, analysis of variance; FAK, focal adhesion kinase.

Co-depletion of FAK and STAT3 further impairs tumour growth

Having identified that Psmb8 was important for regulating the growth of FAK-/- tumours (figure 3G) and that the FAK-/- tumour growth defect was CD8 T-cell-dependent (figure 1C), we postulated that increased expression of both Psmb8 and H2-Kb in response to co-depletion of FAK and STAT3 might further restrain tumour growth and potentiate the antitumour CD8 T-cell response. To address this, 0.5×10⁶ FAK-/- or FAK-/-STAT3shRNA cells were implanted into the pancreas of C57BL/6 mice and mice sacrificed either 2 weeks or 4 weeks later (figure 7A,B). At both time points, mice bearing FAK-/-STAT3shRNA tumours were found to have a significant reduction in tumour burden when compared with those bearing FAK-/- tumours. Flow cytometry analysis of dissociated tissue from 4-week-old tumours also confirmed in vitro observations, identifying an increase in MHC-I expression on CD45^- cells in FAK-/-STAT3shRNA tumours when compared with FAK-/- tumours (figure 7C and online supplemental figure 8). The frequency of immune cells (figure 7D and online supplemental figure 8) and effector CD8 T-cells (effCD8) (figure 7E and online supplemental figure 8) infiltrating into FAK-/-STAT3shRNA tumours was also significantly increased. Immunohistochemical staining of FAK-/- and FAK-/-STAT3shRNA tumour sections using an anti-CD8 antibody also confirmed the infiltration of CD8 T-cells within the tumour mass (figure 7F—left and J), while an anti-granzyme-B antibody showed positive intratumoural staining indicative of ongoing CD8 T-cell activity (figure 7F—right and K). Further, a greater proportion of effCD8 T-cells in FAK-/-STAT3shRNA tumours were positive for expression of the receptor programmed death receptor 1 (PD-1) (figure 7G and online supplemental figure 8), which has been shown to be a marker of tumour-reactive T-cells that have encountered antigen.39 Thus, co-depletion of FAK and STAT3 promoted extensive CD8 T-cell infiltration and further restrained pancreatic tumour growth, likely via increased CD8 T-cell engagement. In contrast, STAT3-depletion in FAK-wt cells had no impact on tumour growth (figure 7H) despite driving a notable influx of CD8 T-cells into the tumour mass (figure 7I—left and J). This was associated with a reduction in granzyme-B positive intratumoural staining when compared with FAK-/- and FAK-/- STAT3 shRNA tumours (figure 7I—right and K), implying a lack of effective CD8 T-cell engagement in the context of FAK-wt tumours. To better understand how STAT3 could promote CD8 T-cell infiltration independent of FAK, we next profiled chemokine secretion from FAK-wt, FAK-/-, FAK-wt STAT3shRNA and FAK-/-STAT3shRNA cells±IFNγ using Proteome Profiler Chemokine Antibody Arrays (figure 7L). In the presence of IFNγ, which we have shown is secreted by immune cells within the FAK-wt and FAK-/- TMEs (figure 1D), STAT3-depletion resulted in upregulation of Cxcl9 independent of FAK expression status. CXCL9 is a ligand for the receptor CXCR3 and has been shown to promote the infiltration of CD8 T-cells into tumours.40 In addition, the blood plasma levels of CXCL9 in PDAC patients receiving chemotherapy has been shown to correlate with longer overall survival and a longer time to progression.41 Thus, STAT3-depletion reprograms the chemokine secretory profile of FAK-wt and FAK-/- cells in favour of CD8 T-cell infiltration, complementing FAK-dependent reprogramming of antigen processing and presentation pathways required to promote T-cell tumour recognition.

• Download figure
• Open in new tab
• Download powerpoint

Figure 7

Co-depletion of FAK and STAT3 potentiates antitumour immunity. (A) Fold change in tumour weight relative to FAK-/- tumours 2 weeks postimplantation into the pancreas of C57BL/6 mice. n=6 tumours per group. (B) Fold-change in tumour weight relative to FAK-/- tumours 4 weeks postimplantation into the pancreas of C57BL/6 mice. n=6 tumours per group. (C) Left, flow cytometry quantification of MHC-I-expressing cells as a percentage of CD45^- cells from FAK-/- and FAK-/-STAT3shRNA tumours. Right, flow cytometry quantification of the median fluorescence intensity of MHC-I expression in CD45^- cells from FAK-/- and FAK-/-STAT3shRNA tumours. n=5 tumours per group. (D) Flow cytometry quantification of CD45⁺ cells as a percentage of live cells from FAK-/- and FAK-/-STAT3shRNA tumours. n=5 tumours per group. (E) Flow cytometry quantification of effCD8⁺ T-cells represented as number of cells per mg of tumour tissue (left) and as a percentage of CD45⁺ cells (right) from FAK-/- and FAK-/-STAT3shRNA tumours. n=5 tumours per group. (F) Representative images of FAK-/- and FAK-/- STAT3shRNA tumour sections stained with anti-CD8 or anti-granzyme-B antibodies. (G) Flow cytometry quantification of PD-1⁺ cells as a percentage of effCD8⁺ T-cells from FAK-/- and FAK-/-STAT3shRNA tumours. n=5 tumours per group. (H) Fold change in tumour weight relative to FAK-wt tumours 2 weeks postimplantation into the pancreas of C57BL/6 mice. n=6 tumours per group. (I) Representative images of FAK-wt and FAK-wt STAT3shRNA tumour sections stained with anti-CD8 or anti-granzyme-B antibodies. (J) Quantification of the number of CD8 T-cells per field of view (FOV) from F and I. (K) Quantification of the relative area positive for Granzyme B staining per FOV from (F, I). (L) Proteome profiler array analysis of chemokine secretion from FAK-wt, FAK-/-, FAK-wt STAT3shRNA and FAK-/- STAT3shRNA cells±10 ng/mL IFNγ stimulation for 24 hours. Data in (A–E, G, H, J, K) represented as mean±SEM. Statistical significance in (A–E, G and H) calculated using a two-tailed unpaired t-test. Statistical significance in J and K calculated using a Kruskal-Wallis test with Dunn’s multiple comparison. *p≤0.05, **p≤0.01, ***p≤0.001, ****p≤0.0001. FAK, focal adhesion kinase; MHC-I, Major Histocompatibility Complex class-I.

PDAC cell heterogeneity impacts FAK function

Transcriptomic analysis of human pancreatic cancer has identified two major molecular subtypes, here termed ‘classical’ and ‘squamous’.42–44 However, single-cell RNA sequencing (scRNAseq) studies have identified an intermediate transitional phenotype,45 suggesting that the terms ‘classical’ and ‘squamous’ should be considered as two poles of a continuum rather than as a binary classification. The process of generating FAK-/- cells using CRISPR-Cas9 gene editing required single-cell cloning, resulting in the isolation of a further five FAK-knockout clonal cell populations (figure 8A). Two of these additional cell clones originated from parental Panc47 cells and three from an independently established additional parental cell line from KPC tumours termed Panc117. Therefore, we sought to use these to better understand whether cancer cell heterogeneity could impact FAK function with respect to regulation of Psmb8 and MHC-I. All five FAK-knockout cell lines were reconstituted with FAK-wt and expression confirmed using western blotting (figure 8B). Cells were treated with IFNγ for 24 hours and RNA isolated for NanoString nCounter digital gene expression analysis using the PanCancer Immune Profiling Panel. H2-Kb expression was upregulated in 47-4-3 FAK-/-, 117-6-9 FAK-/- and 117-6-4 FAK-/- cells (figure 8C, left), with all other cell lines showing either unaltered or reduced expression when compared with FAK-wt counterparts. Psmb8 expression was also upregulated in 47-4-3 FAK-/-, 117-6-9 FAK-/- and 117-6-4 FAK-/- cells when compared with FAK-wt counterparts, although the fold-increase in expression was less than that observed in FAK-/- cells (figure 8C, right). No change in Psmb8 expression was observed in 117-4-7 FAK-/- and 47-6-11 FAK-/- cells when compared with FAK-wt counterparts. Thus, FAK-dependent regulation of Psmb8 and MHC-I is not universal across PDAC cell clones, even from the same tumour, suggesting that pancreatic cancer cell heterogeneity has the potential to impact FAK function. Notably, Nlrc5 expression mirrored that of Psmb8 and H2-Kb across the clones, further supporting a role for the Nlrc5 enhanceosome in the regulation of these genes (online supplemental figure 9).

• Download figure
• Open in new tab
• Download powerpoint

Figure 8

PDAC heterogeneity and transcriptional subtype impact FAK function. (A) Graphical summary showing that single-cell cloning can result in a panel of cell lines broadly representing the heterogeneity of the original parental population. (B) Representative anti-FAK western blots showing FAK re-expression in a panel of clonal cell lines in which CRISPR-Cas9 has been used to delete FAK expression. (C) NanoString nCounter gene expression data from IFNγ stimulated cells generated using a mouse PanCancer Immune Profiling panel. Data represented as fold change in gene expression relative to FAK-wt counterpart. (D) Heat map of FAK-/- GSVA scores for expression of genes associated with classical and squamous subtypes of pancreatic cancer. (E) Representative anti-GATA6, HNF1A, HNF4A, FOXA2 and PDX1 western blot from whole cell lysates isolated from six FAK-/- clonal cell lines. Anti-GAPDH used as a loading control. (F) Western blot showing CRISPR-Cas9 deletion of HNF1A, GATA6 and HNF1A+GATA6 in FAK-wt and FAK-/- cells. Anti-GAPDH used as a loading control. (G) Relative quantification of Psmb8 and H2-Kb gene expression using qRT-PCR following stimulation with IFNγ. n=3. (H) Relative quantification of Psmb8 and H2-Kb expression using qRT-PCR in FAK-wt and FAK-/- cells±treatment with TGFβ1 for 9 days. Cells were stimulated for 24 hours with IFNγ ± TGFβ1 prior to RNA extraction. n=3 (I) Relative quantification of Psmb8 and H2-Kb expression using qRT-PCR in FAK-wt and FAK-/- cells±treatment with TGFβ inhibitor for 2 weeks. Cells were stimulated for 24 hours with IFNγ ± TGFβ inhibitor prior to RNA extraction. n=3. IFNγ treatment was 10 ng/mL for 24 hours (C, G, H, I). Data in G–I represented as mean±SEM. Statistical significance calculated using a one-way ANOVA with Tukey’s multiple comparison test. *p≤0.05, **p≤0.01. ANOVA, analysis of variance; FAK, focal adhesion kinase; GSVA, gene set variation analysis; PDAC, pancreatic ductal adenocarcinoma.

To characterise features that define pancreatic cancer cells in which FAK loss leads to upregulation of Psmb8 and MHC-I expression in response to IFNγ, we next performed RNA sequencing of all six FAK-/- cell populations and compared this with characterised gene sets that define classical and squamous molecular subtypes42 using gene set variation analysis (GSVA) (figure 8D). Based on GSVA scores, FAK-/- cells were the most enriched in genes associated with the classical subtype, with the 47-4-3 FAK-/-, 117-6-9 FAK-/- and 117-6-4 FAK-/- cells showing some enrichment of both classical-associated and squamous-associated genes, perhaps suggesting that these cell clones have an intermediate transitional phenotype. 117-4-7 FAK-/- and 47-6-11 FAK-/- cells, in which FAK loss did not positively regulate Psmb8 and MHC-I expression, were found to be enriched in genes associated with the squamous subtype, implying that FAK-dependent regulation of Psmb8 and MHC-I is lost as cells differentiate towards a more extreme squamous phenotype.

Emerging evidence suggests that PDAC cells can interconvert between subtypes, implying a degree of plasticity.45 Therefore, to further determine whether transition between subtypes could impact FAK-dependent regulation of Psmb8 and MHC-I, we next set out to switch FAK-/- cells (the most classical-like clone) towards a squamous phenotype. Loss of GATA6 expression together with HNF1A and HNF4A has been shown to drive differentiation towards the squamous subtype.46 Notably, western blotting using lysates from all 6 FAK-/- cell clones identified downregulation of GATA6 and HNF1A in both 117-4-7 FAK-/- and 47-6-11 FAK-/- cells when compared with the more classical-like cell clones, FAK-/- or 47-4-3 FAK-/- (figure 8E). We, therefore, used CRISPR-Cas9 gene editing to delete expression of HNF1A, GATA6 or HNF1A and GATA6 (figure 8F) and determined the effects on Psmb8 and H2-Kb expression using qRT-PCR (figure 8G). Co-deletion of GATA6 and HNF1A attenuated the upregulation of Psmb8 expression in FAK-/- cells when compared with mock transfected controls, but had no impact on the regulation of H2-Kb expression by FAK. Therefore, the loss of GATA6 and HNF1A during subtype transition may contribute to altering FAK function but is not sufficient. TGFβ45 and expression of the transcription factor Gli2,47 a TGFβ-induced gene,48 have also been reported to drive squamous differentiation in culture. Treatment of FAK-wt and FAK-/- cells with TGFβ1 for 9 days almost completely abolished the upregulation of both Psmb8 and H2-Kb expression in response to FAK loss (figure 8H), suggesting an important role for TGFβ driven squamous differentiation in the regulation of FAK function. Given the profound impact of TGFβ treatment, we next treated the most squamous-like cell clone, 47-6-11, with a TGFβ inhibitor (TGFβi) for 2 weeks to promote differentiation towards a more classical-like phenotype. Treatment of 47-6-11 FAK-wt and FAK-/- cells with TGFβi restored FAK-dependent regulation of both Psmb8 and H2-Kb (figure 8I), further supporting the conclusion that TGFβ signalling plays an important role in regulating FAK function in PDAC.

FAK regulates antigen processing and presentation in human PDCLs

Having established a role for FAK in regulating antigen processing and presentation in mouse models of PDAC, we next sought to determine whether similar pathways were also regulated by FAK in human pancreatic cancer cells. To address this, we used a panel of 13 (Capan-1, TKCC22, TKCC02, Mayo4636, Mayo5289, PacaDD137, Mayo4666, TKCC07, TKCC26, TKCC09, Panc1, Psn1 and TKCC10) human patient-derived pancreatic cancer cell lines (PDCLs) for which genomic and transcriptomic characterisation is available, ensuring that a broad range of both the classical and squamous subtypes were represented. To maintain the heterogeneity of the parental cell populations, FAK expression was deleted in PDCLs using CRISPR-Cas9 gene editing and 4 days later protein lysates were isolated for analysis using mass spectrometry (figure 9A and online supplemental figure 10A). To confirm the molecular subtype of each cell line, proteomic and transcriptomic data relating to genes previously reported to define the classical and squamous subtypes47 49 was used to cluster the cell lines, confirming two main clusters representing classical and squamous (online supplemental figure 10B,C). Based on this classification, we next used western blotting to determine the expression of common classical markers including GATA6, HNF1A, HNF4A, FOXA2 and PDX1 in protein lysates from 8 of the cell lines (online supplemental figure 10D). While Capan, TKCC02 and TKCC22 cells were positive for expression of all five markers, TKCC10 cells were negative for all markers suggesting that these cells represented the most extreme of each subtype. Notably, TKCC07, TKCC09, Panc1 and Psn1 cells retained expression of at least some classical makers implying that these represent different points of progression from classical to squamous. The list of significantly differentially expressed proteins (q-value≤0.05; online supplemental table 9) from each cell line was then analysed using the WEB-based Gene Set Analysis Toolkit (www.webgesalt.org) to identify Gene Ontology (GO) biological processes overrepresented within each dataset. GO biological processes enriched in ≥10 cell lines are shown in figure 9B. Notably, only 3 cell lines from 13 lacked significant overrepresentation of antigen processing and presentation, 2 of which were classified as the most extreme squamous from the panel of cell lines analysed. Such findings are in broad agreement with our observations from mouse models suggesting that FAK's role in regulating antigen processing and presentation is diminished as cells differentiate towards an extreme squamous phenotype. Network analysis of proteins relating specifically to the antigen processing and presentation pathways regulated by FAK in our mouse models identified significant (q-value≤0.05) regulation of proteins including Human Leucocyte Antigen (HLA)-A, HLA-B, HLA-C, HLA-E, HLA-F, beta-2-microglobulin (B2M), TAP1, TAP2, tapasin (TAPBP), ERAP1 and the immunoproteasome in response to FAK loss (figure 9C) in the majority of human PDCLs tested, including the Mayo4666 which did not show significant overrepresentation of antigen processing and presentation in GO analysis. Notably, regulation of these proteins was not evident in TKCC10 and TKCC26 cells, further confirming our observations linking the extreme squamous phenotype with loss of FAK function in relation to the regulation of antigen processing and presentation pathways.

• Download figure
• Open in new tab
• Download powerpoint

Figure 9

FAK regulates antigen processing and presentation in human PDCLs. (A) Graphical summary of experimental setup used for proteomic analysis of FAK function in human PDCLs. (B) Gene ontology biological process enrichment analysis of differentially expressed proteins in 13 human PDCLs following FAK-deletion using CRISPR-Cas9. (C) Functional association network analysis of proteins related to antigen processing and presentation from proteomics expression analysis of human PDCLs in which FAK expression has been deleted using CRISPR-Cas9. FAK, focal adhesion kinase; FDR, false discovery rate; GO, gene ontology; PDCLs, patient-derived cell lines.

To further interrogate the link between FAK expression and regulation of antigen processing and presentation we next analysed publicly available bulk RNAseq datasets from human PDAC, acknowledging that such analysis may be confounded by the expression of genes within the TME. Using RNA sequencing data from Panc47 FAK-wt and FAK-/- murine cell clones, we first identified a unique gene signature relating to FAK loss in the most classical-like cell clone (figure 10A). We next subdivided the International Cancer Genome Consortium (ICGC)42 and TCGA50 datasets into classical and squamous transcriptional subtypes, and investigated the association of this gene signature with three gene signatures (detailed in Methods) based on genes we had observed to be regulated by FAK in our mouse models: (1) a refined gene set based on the antigen processing and presentation Kegg pathway, (2) a refined gene set based on the IFNγ Reactome pathway and (3) the immunoproteasome genes Psmb8, Psmb9 and Psmb10. In classical tumours from both the ICGC and TCGA datasets, the gene signature relating to FAK loss showed a statistically significant positive correlation with all three other gene signatures, with the exception of the immunoproteasome signature in the TCGA dataset (figure 10B-D). In squamous tumours from the ICGC dataset no significant correlation was identified. In contrast, a statistically significant positive correlation between FAK loss and both antigen processing and presentation and IFNγ was observed in squamous tumours from the TCGA dataset, implying potential differences between the squamous tumours represented within each dataset. To better understand the reason for these differences we next plotted the classical and squamous score for each tumour from both datasets (figure 10E). Comparison of the score distribution suggested that the ICGC dataset contained more tumours of an extreme squamous phenotype. Plotting the score for each individual squamous tumour further confirmed this observation (figure 10F), showing that the ICGC dataset was skewed towards tumours of the extreme squamous phenotype when compared with TCGA. Thus, the differences in correlation observed are likely the consequence of the squamous datasets representing different points of differentiation within the squamous subtype. This observation further supports our conclusions from both murine and human cell line models implying that FAK-dependent regulation of antigen processing and presentation is diminished in cells and tumours of the extreme squamous phenotype.

• Download figure
• Open in new tab
• Download powerpoint

Figure 10

A FAK loss gene signature positively correlates with antigen processing and presentation gene signatures in human PDAC tumours. (A) Cluster analysis showing a gene signature of FAK loss specific to the most classical FAK-/- clonal cell line. (B–D) Analysis of ICGC and TCGA bulk RNA sequencing datasets from human PDAC examining the correlation between a gene signature of FAK loss and gene signatures associated with antigen processing and presentation (B), interferon-γ signalling (C) and the immunoproteasome (D). (E) Box and whiskers plot of classical and squamous scores for all tumours in ICGC and TCGA datasets. (F) Graph representing the squamous score for each tumour classified as squamous in ICGC and TCGA datasets. FAK, focal adhesion kinase; ICGC, International Cancer Genome Consortium; PDAC, pancreatic ductal adenocarcinoma; TCGA, The Cancer Genome Atlas.