FAK regulates IL-33 expression via chromatin accessibility and c-Jun

Focal adhesion kinase (FAK) localises to focal adhesions and is overexpressed in many cancers. FAK can also translocate to the nucleus where it binds to, and regulates, several transcription factors including MBD2, p53 and IL-33 to control gene expression by unknown mechanisms. We have used ATAC-seq to reveal that FAK controls chromatin accessibility at a subset of regulated genes. Integration of ATAC-seq and RNA-seq data showed that FAK-dependent chromatin accessibility is linked to differential gene expression, including of the FAK-regulated cytokine and transcriptional regulator interleukin-33 (Il33), which controls anti-tumour immunity. Analysis of the accessibility peaks on the Il33 gene promoter/enhancer regions revealed sequences for several transcription factors, including ETS and AP-1 motifs, and we show that c-Jun, a component of AP-1, regulates Il33 gene expression by binding to its enhancer in a FAK kinase-dependent manner. This work provides the first demonstration that FAK controls transcription via chromatin accessibility, identifying a novel mechanism by which nuclear FAK regulates biologically-important gene expression.


Introduction
Focal adhesion kinase (FAK) is a non-receptor tyrosine kinase that is overexpressed in many cancers, including squamous cell carcinoma (SCC) (1), breast, colorectal (2) and pancreatic cancer (3). In addition to more traditional localisation at integrin-mediated cell-matrix adhesion sites (focal adhesions), FAK is also known to localise to the nucleus, where it binds a number of transcription factors including p53 (4), Gata4 (5) and Runx1 (6) to regulate the expression of Cdkn1a (which encodes p21), Vcam1 and Igfbp3, respectively. By binding to these transcription factors, FAK has been linked to cancer-associated processes such as inflammation (5), proliferation (6) and survival (4).
Our previous work demonstrated that, in mutant H-Ras-driven murine SCC cells, nuclear FAK controls expression of cytokines and chemokines, for example Ccl5, to drive recruitment of regulatory T cells into the tumour microenvironment, resulting in suppression of the antitumor CD8 + T cell response and escape from anti-tumour immunity (7). FAK regulates biologically important chemokines by interacting with a network of transcription factors and transcriptional regulators in the nucleus (7), which includes transcription factors may be candidate mediators of FAK-dependent gene expression, we used ATACseq to analyse active (accessible) chromatin in FAK-deficient SCC cells or cells expressing different FAK mutant proteins. Specifically, we compared SCC cells in which FAK had been depleted by Crelox-mediated Ptk2 (which encodes FAK) gene deletion (FAK -/-) with the same cells re-expressing wildtype FAK (FAK-WT) or FAK mutants that were impaired in nuclear localisation (FAK-nls) or deficient in kinase activity (FAK-kd) (cell series described previously in 7, 10). These permitted the identification of FAK-dependent, FAK nuclear localisation-dependent and FAK kinase-dependent alterations in chromatin accessibility and transcription factor motif enrichment in accessible regions of chromatin across the genome.
Chromatin accessibility analysis of ATAC-seq data detected 20,000-60,000 peaks (accessible regions) per sample (Appendix Fig S1A; see Appendix Table S1 for further details of ATAC-seq statistics).
The standard peak number in ATAC-seq experiments can vary depending on cell type, species, context and variations in the ATAC-seq protocol. Importantly, the peak number reported in our study is in the medium-to-high range for an ATAC-seq experiment performed in mouse cells (see additional file 2 in 11).
The majority of peaks in FAK-WT-, FAK-nls-, FAK-kd-expressing as well as FAK -/cell lines were located 0-100 kb from the transcriptional start sites (TSS) (Appendix Fig S1B). The distance of ATACseq peaks from the TSS suggests that the accessible regions were predominantly located in likely enhancer (12) and promoter (13) regions.
We identified differentially accessible gene regions using the DiffBind package (14). Differential peak calling was performed for each pairwise comparison for which FAK-WT samples were compared with each of the FAK knockout (FAK -/-) and FAK mutant (FAK-nls and FAK-kd) cell lines. From this analysis, it was apparent that a subset of genes are regulated by FAK-dependent changes in chromatin accessibility (discussed further below).
We next analysed the transcription factor motif sequences present in FAK-dependent differentially accessible peaks (hereafter termed motif-enrichment analysis). Motif-enrichment analysis allowed us to predict which transcription factors were regulating genes across the genome by analysing the motif sequences in ATAC-seq peaks. We used the HOMER tool (15) to identify motif binding sites (i.e. genomic regions that match known transcription factor motifs) in the differentially accessible peaks In the motif-enrichment analyses of FAK-WT-expressing cells vs FAK -/cells and FAK-WT-vs FAKnls-expressing cells, the two most highly enriched transcription factor motifs were for Jun-AP-1 and Fosl2 (all Benjamini-Hochberg-corrected P = 0), which exhibited an almost two-fold enrichment in motifs in the target (% target, FAK-WT-expressing cells) compared to the motifs identified in the background (% background, FAK -/cells or FAK-nls-expressing cells) (Fig 1A). The top two hits in the motif-enrichment analysis of FAK-WT-vs FAK-kd-expressing cells were motifs for Ets1 and Etv1 (all Benjamini-Hochberg-corrected P = 0), which likewise revealed a two-fold enrichment in these motifs in the target (% target, FAK-WT-expressing cells) compared to the background motifs (% background, FAK-kd-expressing cells) (Fig 1A). These data imply that FAK and specific FAK functions (kinase activity and nuclear localisation) robustly regulate enrichment of particular AP-1 and ETS motifs within accessible chromatin regions in SCC cells.  ) were compared to all the motifs identified in peaks called in either the FAK-nls-expressing cells, FAK-kd-expressing cells or FAK -/cells (% background). The name of the transcription factor motifs are reported next to the consensus sequence (images from HOMER, 15), followed by the proportion of target sequences with that motif (% target) and the proportion of background sequences with that motif (% background). P ≤ 1 × 10 −31 and q = 0 for all motifs shown. (B) Motif-enrichment data for FAK-WT-expressing cells vs FAK -/cells, FAK-nls-expressing cells or FAK-kdexpressing cells were filtered by set analysis to identify transcription factors enriched in FAK-WT-expressing cells but not enriched in FAK -/-, FAK-nls and FAK-kd motif-enrichment analyses (intersection set, red circle). (C) Protein domain enrichment analysis of transcription factors associated with motifs that are enriched in FAK-WT-expressing cells. The full FAK-WT-expressing cells vs FAK -/cells motifenrichment analysis (transcription factors predicted from both significant and non-significant motifs) was used as the background list. All terms with Benjamini-Hochberg-corrected P ≤ 0.05 are displayed (−log 10 -transformed corrected P-values are shown). The full domain name is reported in parentheses next to the corresponding SMART domain term. (D) Transcription factors that have enriched motifs in FAK-WT-expressing cells (intersection set in (B), red circle) were used to construct a functional association network using Ingenuity Pathway Analysis. Only direct, mammalian interactions are shown. Edge (line) style represents type of physical or functional connection. Node (circle) size indicates the connectivity of the node (number of associations that node has within the network). Node borders for transcription factors from the AP-1 family are red and for the ETS transcription factors are purple. The network was structured using the yFiles hierarchical layout algorithm. n = 2 biological replicates.   (Fig 1C). This analysis indicated that there was an over-representation of transcription factors known to bind motifs containing ETS (term SM00413:ETS, Benjamini-Hochberg-corrected P = 1.02 × 10 −10 ) and PNT domains (term SM00251:SAM_PNT, Benjamini-Hochberg-corrected P = 1.06 × 10 −5 ; Fig 1C), including the ETS transcription factor family members Fli1, Elf3, Elf5, Gabpa, Spdef, Erg, Ehf and Ets1. Furthermore, there was also an enrichment for transcription factors known to bind motifs that contain basic-leucine zipper domains (term SM00338:BRLZ, Benjamini-Hochbergcorrected P = 0.0033; Fig 1C), including members of the AP-1 family, such as c-Jun, JunB, Fosl1, Fosl2 and Atf3. Thus, our analyses revealed FAK-dependent enrichment of a set of sequence motifs known to bind AP-1 and ETS transcription factors.
To understand better the likely transcription factors responsible for FAK-dependent gene expression, we performed interactome analysis to determine putative connections between transcription factors known to associate with FAK-regulated motifs. We constructed a functional association network, incorporating curated protein-protein and protein-DNA interactions, of the transcription factors whose motifs were enriched in FAK-WT-expressing cells (Fig 1D). The network analysis revealed that transcription factors known to bind FAK-regulated motifs have a large number of connections with other transcription factors known to bind FAK-regulated motifs (Fig 1D). The most highly connected transcription factor was the AP-1 member c-Jun (outlined in red in Fig 1D), and network topology implied that c-Jun is a key signal integrator between all the other transcription factors in the network.
Other well-connected nodes in the network were members of the AP-1 family, including JunB, Atf3 and Fosl1 (outlined in red in Fig 1D). In addition, certain members of the ETS transcription factor family had many physical and functional connections within the network, namely Ets1 and Spi1 (outlined in purple in Fig 1D). Collectively, these data suggest that FAK regulates motif enrichment in accessible regions of chromatin, in particular sequences known to bind to the AP-1 and ETS transcription factor family members.

FAK regulates chromatin accessibility at a subset of genes, including Il33
Differential peak calling analysis identified chromatin accessibility changes that were dependent on FAK, as well as FAK kinase activity and its nuclear localisation (Fig 2A). All ATAC-seq peaks were set to 500 bp to allow comparison between peaks in the SCC cell lines used in this study, and we reported distances from the ATAC-seq peak centre as a heatmap (red indicates high read count (highly accessible region) in Fig 2A). This analysis revealed ATAC-seq peaks across the genome with differential accessibility (varied read count) when comparing FAK-WT SCC cells to FAK -/-, FAK-nlsexpressing and FAK-kd-expressing SCC cells, identifying changes in a subset of genes that varied depending on FAK status (Fig 2A). These data implied that FAK regulates the chromatin accessibility at a subset of genes.
We next identified which genes were regulated by FAK-dependent changes in chromatin accessibility using comparisons between the cell lines that varied only in FAK status. We wanted to determine which genes were associated with the ATAC-seq peaks enriched in FAK-WT-expressing cells (as identified by differential peak calling) to understand which genes are regulated by FAK-dependent accessibility changes. To assign each ATAC-seq peak to genes, we used ChIPseeker (17), which links each peak to the closest TSS using data from the University of California, Santa Cruz, genome browser annotation database (https://genome.ucsc.edu/). We used FAK RNA-seq data to confirm whether the genes that were regulated by FAK-dependent changes in chromatin accessibility were also differentially expressed in a FAK-and FAK kinase-dependent manner (Fig 2B and FAK RNA-seq dataset reported in Dataset EV2). Set analysis identified genes that were either up-or down-regulated in a FAK-or FAK kinasedependent manner, and also those genes whose FAK-dependent changes were associated with chromatin accessibility changes (intersection sets in upper panels in Fig 2B). We found 36 genes whose expression and chromatin accessibility profiles were both regulated by FAK and its kinase activity (intersection sets in lower panel in Fig 2B). Comparison of the FAK-nls mutant chromatin accessibility   17 18 Enhancer Promoter data to this subset of genes revealed that most of these are also dependent on FAK's ability to localise to the nucleus (asterisks in lower panel in Fig 2B).
As an exemplar, we next focussed on one of these genes, Il33, because we had previously reported it as a FAK-regulated cytokine of biological significance in mediating FAK-dependent anti-tumour immunity (9). Using ATAC-seq data to investigate whether chromatin accessibility was one mechanism by which FAK regulates Il33, we found that there were ATAC-seq peaks in FAK-WT-expressing SCC cells on Il33 enhancer (−3199 from TSS) and promoter (+821 from TSS) regions (Fig 2C). Moreover, these peaks were absent in the FAK-kd-and FAK-nls-expressing cells and reduced in FAK -/cells, which had no detectable peak on the promoter region and a suppressed ATAC-seq peak on the enhancer region. However, we note that the suppressed ATAC-seq peak on one replicate of the FAK -/cells (FAK -/-2) did not have sufficient read count to be identified as an ATAC-seq peak, and therefore the peak was not called. We conclude that FAK regulates chromatin accessibility at a subset of gene promoters, and some of these are differentially expressed in a FAK-dependent manner, as exemplified by Il33. This suggests that FAK-regulated, biologically important gene expression alterations may be controlled by FAK-dependent chromatin accessibility changes.

FAK regulates IL-33 expression via chromatin accessibility at the c-Jun motif in the Il33 enhancer
In order to define key transcription factors that drive FAK-dependent Il33 expression in mouse SCC cells, we performed motif-enrichment analysis on the ATAC-seq peaks proximal to the Il33 promoter and enhancer regions in FAK-WT-expressing cells using HOMER (using data depicted in Fig 2C).
Analysis of the raw peak calling data revealed that there was a number of peaks upstream of the Il33 It is well established that in order for a transcriptional event to occur, transcription factors often need to form complexes with other transcription factors in the same or different families. For example, it is well known that c-Jun homodimerises, as well as heterodimerises with c-Fos and Fra or Atf family members, to regulate the expression of AP-1-dependent genes (18). Furthermore, the transcription factor Nr4a1 have been shown to bind and co-operate with c-Jun to regulate the transcription of the Star gene (19).
Therefore, we addressed whether the transcription factors predicted to regulate Il33 expression in FAK-WT-expressing cells can bind to and/or regulate each other. In addition, we generated an Il33 transcription factor regulatory network for FAK-WT-expressing cells, which indicated that the transcription factors known to bind FAK-regulated motifs at the Il33 gene have multiple functional connections (Fig 3B). Indeed, the most highly connected transcription factor was the AP-1 member c-Jun (largest node in Fig 3B), suggesting it may be a key node in the FAK-dependent Il33 transcription factor network.
We next examined nuclear FAK binding partners (described previously in 7) and contextualised these with regard to transcription factors that may bind to the identified sequence motifs in the Il33 gene where accessibility is FAK-regulated. This predicted upstream connections between nuclear FAK binding proteins and transcription factors likely to access sequences in Il33 promoter/enhancer regions in a FAK-dependent manner (Fig 3C). The resulting network indicated that the transcription factors with motif sequences on the Il33 promoter/enhancer have varying numbers of functional associations with putative nuclear FAK-interactors (indicated by node size in Fig 3C). The transcription factors with the most links to nuclear FAK-binding proteins were c-Jun and Nr4a1 (Fig 3C). This implied that there were likely interesting connections between FAK and the transcription factors known to access motifs in the Il33 promoter in a FAK-dependent manner. Interestingly, c-Jun was detected previously as a potential nuclear FAK interaction partner by proteomics (7). However, attempts to validate the FAKc-Jun interaction by co-immunoprecipitation were unsuccessful, suggesting that c-Jun is not a robust  . Nodes for transcription factors predicted to bind to the Il33 promoter/enhancer regions in FAK-WT-expressing SCC cell lines are coloured in yellow; all potential FAK interactors that bind to predicted Il33 motifs are coloured purple (node labels omitted in (C) for clarity). Red node borders indicate proteins identified as FAK interactors by previous validation experiments (7,9). Node size indicates the connectivity of the node. The yFiles organic layout algorithm was applied to the network. n = 2 biological replicates for the ATAC-seq dataset (B-D); n = 3 biological replicates for the proteomics dataset (C and D).

D C
or high-stoichiometry interaction partner of FAK. Nonetheless, our network analysis did imply that c-Jun interacts with a number of FAK binders identified previously (7), such as Pin1, which has been shown to bind to c-Jun and increase its transcriptional activity (20). The FAK binding partner and transcription factor Sp-1 (9) has been reported to bind both the two most highly connected nodes in the network, c-Jun (21; left in Fig 3D) and Nr4a1 (22; right in Fig 3D). Other nodes that had connections with validated FAK binders included Tbp, which binds to the FAK binding protein Taf9 (7; Appendix   Fig S2) to form the TFIID component of the basal transcription factor complex (23). Therefore, our interactome analysis indicates that FAK is functionally well connected to transcription factors known to bind to sequence motifs on the Il33 enhancer whose accessibility is FAK-regulated.

c-Jun regulates IL-33 expression by binding to Il33 enhancer regions
Our nuclear FAK interactome analysis showed that c-Jun formed a hub (i.e. a highly connected node) in the Il33 transcription factor network (Fig 3C). c-Jun is a component of the AP-1 family of transcription factors, and it is an important regulator of skin inflammation (24). For example, c-Jun proteins are known to be important in for the expression of the cytokine CCL5 (25), which we have shown to be regulated by FAK and IL-33, and loss of Jun proteins can lead to the onset of a chronic psoriasis-like disease (26). Therefore, we hypothesised that c-Jun may be a likely regulator of inflammatory gene expression in SCC cells (which originate from skin keratinocytes) used in our studies. We performed siRNA-mediated depletion of Jun mRNA (which encodes c-Jun) (Fig 4A), which led to a parallel significant downregulation of Il33 mRNA (Fig 4B) and reduced IL-33 protein expression (Fig 4C). In addition, the FAK and IL-33 target gene in SCC cells, Cxcl10, also showed reduced mRNA levels as a result of Jun knockdown (Fig 4D). Taken together, these data imply that c-Jun is likely an important regulator of IL-33 expression and of FAK-and IL-33-dependent target genes.  Jun transcription factors generally dimerise with other Jun family members, and c-Fos or Fra1, to bind to AP-1 sites, whereas CRE sites are known to be preferentially bound by c-Jun together with Atf complexes (18). We therefore used ChIP to show that c-Jun binds to the AP-1 and CRE motifs at the Il33 enhancer in a FAK-dependent manner via accessibility. Our attempts to perform c-Jun ChIP experiments at the AP-1 site were unsuccessful, but we performed c-Jun ChIP at the alternative c-Jun binding site to establish whether it binds to the CRE site at the Il33 enhancer. Primers were designed around the region containing the CRE sequence motif in the Il33 enhancer and an unrelated region upstream of this site to control for background binding (depicted in Fig 4E). We used an anti-c-Jun ChIP-grade antibody to pull down DNA in formaldehyde-crosslinked, sonicated chromatin preparations from FAK-WT and FAK-kd cells, since loss of FAK's kinase activity displayed the most striking loss of chromatin accessibility at the Il33 enhancer (Fig 2C). Following immunoprecipitation, the DNA was purified and a qPCR was performed, whereby the Il33 enhancer region and an upstream background region were amplified. We used the % input method to normalise the ChIP-qPCR data for potential sources of variability, including the starting chromatin amount in the chromatin extract, immunoprecipitation efficiency and the amount of DNA recovered (see Materials and Methods). Using ChIP, we found that c-Jun bound to the Il33 enhancer in the FAK-WT-expressing cells (Fig 4F) and there was a significant loss of c-Jun binding in FAK-kd-expressing cells in comparison to FAK-WTexpressing cells (Fig 4F). These data are consistent with FAK kinase activity regulating chromatin accessibility at the enhancer region upstream of the Il33 gene at the predicted c-Jun binding site.
Thus, we conclude that FAK, which is classically thought to be primarily an integrin adhesion protein, can function in the nucleus to control chromatin accessibility at specific gene promoters/enhancers. In turn, this leads to FAK-dependent transcription of specific genes, an example of which is the cytokine Il33. FAK/IL-33 downstream effectors significantly influence tumour biology (9).

Discussion
In this study, we have discovered an undescribed function of nuclear FAK as a key regulator of chromatin accessibility changes and transcription factor binding. Furthermore, we have confirmed that nuclear FAK regulates c-Jun binding at the Il33 enhancer region via chromatin accessibility changes to control Il33 expression. As IL-33 is an important regulator of cytokine expression and tumour growth (8), FAK-dependent c-Jun regulation of IL-33 expression would be predicted to influence cancer cell biology, such as that we described previously (9). It is perhaps not surprising that FAK can regulate c-Jun, since cytoplasmic-localised FAK is known to transduce signals through pathways such as MAPK (27) and Wnt (28,29), which are known to control c-Jun expression and its transcriptional activity (18,30); however, what is surprising is the more direct link we have uncovered here between nuclear FAK function and its regulation of c-Jun transcriptional activity at the Il33 enhancer via chromatin accessibility. Consistent with the links between nuclear FAK and c-Jun activity being more common, nuclear FAK is reported to regulate the expression of Jun (which encodes c-Jun) in response to 'stretch' in cardiac myocytes by binding to, and enhancing, the transcriptional activity of MEF2 (31).
Focal adhesion proteins other than FAK have been detected in the nucleus, such as Lpp (32) and Hic-5 (33), which are believed to function as transcription factor co-regulators (32,34). Furthermore, the focal adhesion protein paxillin can also translocate to the nucleus (35), where it contributes to proliferation (36), and we believe that there are other integrin-linked adhesion proteins capable of translocating to the nucleus and functioning at the nuclear membrane or inside the nucleus (Byron et al., submitted).
Relevant to the work we present here, a number of consensus adhesome components containing LIM (Lin11-Isl1-Mec3) domains have been directly linked to the regulation of chromatin accessibility and dynamics. For example, Hic-5 can inhibit the binding of the glucocorticoid receptor to the chromatin remodellers chromodomain-helicase DNA-binding protein 9 (also known as ATP-dependent helicase CHD9) and brahma (also known as ATP-dependent helicase SMARCA2), resulting in a closed chromatin conformation and transcriptional repression of a subset of glucocorticoid receptor target genes (34,37). Also, paxillin can regulate proliferation-associated gene expression by controlling promoter-enhancer looping via nuclear interactions with the cohesin and mediator complex (36). Taken together, these reports suggest that focal adhesion proteins in the nucleus are capable of scaffolding chromatin remodelling complexes to regulate chromatin structure and gene expression changes.
An unanswered question is the mechanism by which FAK controls chromatin accessibility at regulated genes. In this regard, our previous nuclear interactome proteomics revealed that FAK can interact with proteins known to regulate chromatin accessibility (7). These include the Smarcc2 and Actl6a components of the BRG1/BRM-associated factor (BAF) complex (7), which have been shown to be recruited to target gene enhancers by AP-1 to regulate chromatin accessibility (38). IL-33 is required for the chromatin recruitment of the Wdr82 component of the chromatin-modifying protein serine/threonine phosphatase (PTW/PP1 phosphatase) complex (9,39). IL-33 binds to the Brd4 (bromodomain-containing protein 4) transcriptional coactivator (9), which is known to recruit the BAF complex to target genes (40). The nuclear FAK binding protein Sp-1 also interacts with the BAF complex to facilitate its recruitment to specific promoters (41). Therefore, there is abundant evidence of connections between FAK or FAK binding proteins (i.e. FAK-Sp-1, FAK-IL-33) and FAKregulated transcription factors (e.g. AP-1) to chromatin accessibility factors, such as the BAF complex and PTW/PP1 phosphatase complex. It is likely that FAK, and proteins to which it binds, scaffold chromatin remodelling proteins at target genes, such as Il33 we describe here, in order to determine the state of chromatin accessibility, the binding of transcription factors like AP-1 and transcription ( Fig   4G).
In summary, we have discovered a completely new paradigm for how FAK may regulate transcription in the nucleus, i.e. as a critical regulator of chromatin accessibility changes at biologically important target genes, such as Il33 we show here. Translocation of FAK to the nucleus, where it can bind to factors that control chromatin accessibility, can therefore communicate extracellular cues to the gene transcription machinery in the nucleus by this route.

FAK SCC cell line generation
Generation of the FAK SCC cell model has been described previously (10). Briefly, K14CreER

ATAC-seq
ATAC-seq samples were prepared as described previously (42). ATAC-seq data were aligned to the Mus musculus reference genome mm10 using the bcbio ATAC-seq pipeline (43). Accessible regions (i.e. ATAC-seq peaks) were called from the BAM files using the MACS2 algorithm (44) with the following parameters: -B -broad -q 0.05 -nomodel -shift -100 -extsize 200 -g 1.87e9. Differentially accessible regions between the FAK-WT cells and the FAK -/-, FAK-nls and FAK-kd cells were identified by differential peak calling using the R/Bioconductor package DiffBind (14), where significantly different peaks were defined as those with a false discovery rate (FDR) of below or equal to 0.05. Motif-enrichment analysis was performed using HOMER (15) following default parameters.
ATAC-seq peaks were assigned to genes using ChIPseeker (17)  Transfections were carried out using Lipofectamine RNAiMAX transfection reagent (Invitrogen) following manufacturer's instructions. Cells were incubated in transfection mixes for 48 hours before harvesting for RNA extraction or whole cell lysate preparation.

Chromatin immunoprecipitation (ChIP)-qPCR
The ChIP-qPCR experiments were performed as described previously (46,47). FAK-WT-and FAKkd-expressing cells (4 × 10 6 ) were plated on 10-cm dishes (Corning) and then, the following day, were formaldehyde crosslinked and fractionated as described in 46. Sonication was carried out using a

RT-qPCR
RNA extraction was performed using an RNeasy Mini kit (Qiagen) following manufacturer's instructions. cDNA synthesis was performed using the SuperScript First-strand Synthesis System (Invitrogen) following the manufacturer's random hexamers protocol. qRT-PCR analysis was performed using SYBR green mastermix (Thermo Scientific) following manufacturer's instructions.

Cell lysis and immunoblotting
Cells were washed twice with ice-cold PBS before scraping in 1× RIPA buffer (150 mM NaCl, 1% Signaling Technology).

RNA-seq
RNA was extracted from FAK-WT, FAK -/and FAK-kd SCC cells using an RNeasy kit (Qiagen) following manufacturer's instructions. To verify sample purity, the samples were run on a 2100 Bioanalyzer using the Bioanalyzer RNA 6000 pico assay (both Agilent). Samples that achieved an RNA integrity number (RIN) of 8 or above were considered suitable purity for sequencing. Samples were prepared for sequencing using the TruSeq RNA Library Prep Kit v2 (low-sample protocol) (Illumina) and paired-end sequenced using a HiSeq 4000 platform (Illumina) at BGI.
Transcript abundance was determined using the pseudoalignment software kallisto (48) on the mouse transcriptome database acquired using the kallisto index, implementing default parameters. Quality control was performed on the kallisto output using MultiQC software (https://github.com/ewels/MultiQC). Transcript abundance was summarised to gene level and imported into the differential expression analysis R package DESeq2 (49) using the R package tximport (50).
Genes which had zero read counts were removed prior to differential expression analysis.

Protein domain enrichment analysis
Protein domain enrichment analysis was performed for SMART domains using DAVID (52,53). All terms that acquired a Benjamini-Hochberg corrected P-value of below 0.05 were considered statistically significant.
The following parameters were used for network construction: database sources (Ingenuity expert information, protein-protein interaction database, BioGrid, IntAct), direct interaction, experimentally observed, protein-protein and functional interactions, mammalian interactions only. All networks were exported into Cytoscape (54), and the NetworkAnalyzer plugin (55) was used to visualise the most connected nodes in the networks before applying yFile layout algorithms (yWorks).