∆133p53 isoform promotes tumour invasion and metastasis via interleukin-6 activation of JAK-STAT and RhoA-ROCK signalling

∆122p53 mice (a model of ∆133p53 isoform) are tumour-prone, have extensive inflammation and elevated serum IL-6. To investigate the role of IL-6 we crossed ∆122p53 mice with IL-6 null mice. Here we show that loss of IL-6 reduced JAK-STAT signalling, tumour incidence and metastasis. We also show that ∆122p53 activates RhoA-ROCK signalling leading to tumour cell invasion, which is IL-6-dependent and can be reduced by inhibition of JAK-STAT and RhoA-ROCK pathways. Similarly, we show that Δ133p53 activates these pathways, resulting in invasive and migratory phenotypes in colorectal cancer cells. Gene expression analysis of colorectal tumours showed enrichment of GPCR signalling associated with ∆133TP53 mRNA. Patients with elevated ∆133TP53 mRNA levels had a shorter disease-free survival. Our results suggest that ∆133p53 promotes tumour invasion by activation of the JAK-STAT and RhoA-ROCK pathways, and that patients whose tumours have high ∆133TP53 may benefit from therapies targeting these pathways.

T he TP53 gene encodes 12 isoforms 1-5 that moderate p53 activities 1,6,7 as well as having independent functions. Several isoforms are dysregulated in human tumours leading to the suggestion that they promote tumour formation. For example, elevated levels of Δ40p53 are found in melanomas and in breast tumours, particularly the aggressive triple-negative cancer subtype 8,9 . Aberrant expression of Δ133p53 has also been reported in renal cell cancer 10 , head and neck tumours 11 , colon carcinomas 7 , lung cancer 12 and again in breast cancer 8,13 , where it is associated with increased risk of recurrence 13 . Thus, these isoforms appear to have oncogenic activities. In addition, transgenic mice modelling Δ133p53 (designated Δ122p53) are highly tumour-prone, display chronic inflammation and autoimmunity 14,15 , and have elevated levels of pro-inflammatory cytokines, notably interleukin-6 (IL-6) 14,15 . Δ133p53 and Δ122p53 promote migration in scratch wound and Transwell assays, which are associated with changes in the polarity of actin fibres 16 and Δ122p53 promotes invasion of pancreatic tumour cells through a three-dimensional collagen matrix. IL-6 was found to be a key mediator of these migratory phenotypes 16 .
In this paper, we test the importance of IL-6 on the oncogenic and inflammatory phenotypes of Δ133p53 by crossing Δ122p53 mice with IL-6 null mice. We show that IL-6 loss reduces tumour incidence and metastasis and that Δ122p53 and Δ133p53 upregulate JAK-STAT3 and RhoA-ROCK signalling pathways that contribute to cell migration and invasion. Thus, IL-6 is an important mediator of the oncogenic activity of the Δ133p53 isoform. We also show that human colorectal cancers with elevated Δ133TP53 mRNA are more aggressive, being associated with a shorter disease-free survival.

Results
Loss of IL-6 reduces JAK-STAT signalling. We previously reported that Δ122p53 mice exhibit widespread inflammation accompanied by 1.5-to 2-fold elevated levels of several proinflammatory cytokines in the serum compared to wt p53 (wildtype) mice, with the exception of IL-6, which was 12-fold higher 15 . These results suggest that IL-6 is an important contributor to the pro-inflammatory and oncogenic phenotypes in Δ122p53 mice. To test this, double-mutant Δ122p53/Δ122p53 (designated Δ122/Δ122) or Δ122p53/p53+ (designated Δ122/+) mice were generated by crossing Δ122p53 mice with IL-6-deficient mice 17 . IL-6 functions by activating the Janus kinases (JAKs), recruiting the signal transducer and activator of transcription (STAT)-3 to induce transcription of genes involved in inflammation 14 . Thus, mice deficient in IL-6 should show an overall reduction in downstream cytokine levels. The sera of 7-to 9-week-old Δ122/Δ122 IL-6+/+, Δ122/Δ122 IL-6−/− and wt p53/ IL-6−/− mice were thus used to quantitate the levels of cytokines and chemokines by Bioplex array. At this time, these mice show no pathology 15 . Results are shown in Fig. 1 and in Supplementary  Fig. 1. A total of 20 molecules were assayed of which 17 molecules are in the JAK-STAT pathway. Of these, 12 were significantly lower in the serum of Δ122/Δ122 IL-6−/− mice (p < 0.05, onetailed unpaired t-test), ranging from a 4-fold difference for Mcp-1 (Ccl-2), to 10-fold for interferon (IFN)-γ and 27-fold for IL-17. Sera from wt p53/IL-6−/− control mice generally showed similar levels of these molecules to Δ122/Δ122 IL-6+/+ mice or were lower (e.g., Tnf-α, Ccl-2 and IL-17 (p < 0.05, one-tailed unpaired t-test)) as we found previously 15 . Thus, the JAK-STAT3 pathway is active in Δ122p53 mice and appears to be largely regulated by IL-6.
Previously, we also reported that Δ122/Δ122 mice display splenomegaly prior to tumour onset indicative of cellular hyperproliferation 15 . To determine if this required IL-6 we measured the wet weights of the spleens in the mouse cohorts. Supplementary Fig. 3a shows the ratio of spleen weight to body weight of Δ122/Δ122 mice (median ratio = 0.02, p = 0.003, twotailed unpaired t-test) was reduced to the levels of wt p53 mice by retention of 1 wt p53 allele or the loss of 1 or 2 IL-6 alleles (median ratio~0.004). Similar results were obtained for liver ( Supplementary Fig. 3b).
Despite the significant reduction in tumour incidence and metastasis in mice that had lost one or both IL-6 alleles, the number of animals studied were not sufficient to ascertain a significant difference in survival between the genotypes, although there was a trend towards increased survival for the heterozygous mice ( Supplementary Fig. 4a and 4b). There was also no difference in the time at which tumours were detected (and mice culled) across the genotypes (Supplementary Fig. 5 and  Supplementary Table 1). Of note, p53+/+ IL-6−/− mice did not develop tumours and were all culled after 500 days ( Supplementary Fig. 4a) mainly due to a distended abdomen.
In summary, the loss of IL-6 in Δ122/Δ122 mice resulted in decreased levels of serum cytokines and chemokines involved in the JAK-STAT pathway and significantly decreased the incidence of T-cell lymphomas with extensive spread. There were also no co-malignancies and no metastatic sarcomas in these mice. Loss of IL-6 in the Δ122/+ mice resulted in a decreased incidence of Tcell lymphomas as well as a significant reduction in sarcomas that underwent metastasis. These results suggest that IL-6 is an important contributor to the pro-inflammatory and tumorigenic phenotypes of Δ122p53 mice. m. Heteroscedastic one-tailed unpaired t-test was performed and p < 0.05 was considered significant. *p < 0.05, **p < 0.005 and ***p < 0.0005 are designated, respectively. Δ/Δ = Δ122/Δ122 Δ122p53 activates RhoA-ROCK signalling. As well as activating inflammatory gene transcription 20 via the JAK-STAT3 pathway, IL-6 also induces transcription of genes involved in migration 21 , angiogenesis [22][23][24] , metastasis 25 and tumour progression 26 . We previously demonstrated that Δ122p53 drives a migratory phenotype associated with actin polarisation 16 . To test whether this is reduced in the absence of IL-6, mouse embryonic fibroblasts (MEFs) were used in scratch wound closure assays and the proportion of cells showing actin polarisation was assessed. Examples of actin polarisation are shown in Fig. 4a and the quantitation is shown in Fig. 4b. Actin polarisation was reduced by 40-60% in the absence of one or both IL-6 alleles (Fig. 4b). In parallel, as a control, we found that loss of IL-6 reduced cell migration two-to four-fold in Transwell assays (Fig. 4c). These results suggest that IL-6 is required for actin polarisation induced by Δ122p53, a process driven by GTPase RhoA and associated kinase ROCK 19,27,28 . These results suggest that RhoA-ROCK signalling is active in Δ122p53-expressing cells and this is dependent on IL-6. Next, we tested whether inhibiting the JAK-STAT and RhoA-ROCK pathways could reduce invasion caused by Δ122p53. To do this we used mouse pancreatic adenocarcinoma (PDAC) cells that had been stably transduced to express Δ122p53 (PDAC/ Δ122) 16 in invasion assays through collagen matrices [29][30][31] . Cells were also incubated with the Pan-JAK inhibitor P6 28 , the ROCK inhibitors Y-27632 and H1152 28 , and the RhoA inhibitor TAT-C3 (C3) 29 . As previously demonstrated 16 , PDAC/Δ122 (Δ) cells showed a greater degree of invasion (2.4-fold) than PDAC/vector cells (Vo; Fig. 5a). The results also show that the three inhibitors reduced invasion of PDAC/Δ122 cells to varying degrees. Thus, at least one mechanism of inhibiting the Δ122p53-driven invasion is by inhibiting the JAK-STAT3/RhoA-ROCK pathway. Further to this point, the 'invading cells' were stained with an antibody to phosphorylated STAT3 (pSTAT3) as a marker of JAK-STAT activation. Results (Fig. 5b) show that~2-fold more PDAC/Δ122 cells express pSTAT3 compared to PDAC/vector cells, which was abolished with the JAK, ROCK and RhoA inhibitors. In other experiments, we stained for phosphorylated myosin phosphatase target subunit 1 (pMYPT1) as a marker of ROCK activity and actin contractility 19 . pMYPT1 staining was at background levels in PDAC/vector cells but 6-to 7-fold elevated in PDAC/Δ122 cells (Fig. 5c), which was reduced 2-to 4-fold with the JAK, ROCK and RhoA inhibitors. The non-invading and whole population of PDAC/Δ122 cells were also treated with inhibitors and stained for pSTAT3 and pMYPT1 ( Supplementary Fig. 6). Again, pSTAT3 was elevated in Δ122p53 cells, which was abolished with all inhibitors. For pMYPT1, the non-invading cells and the whole PDAC/Δ122 cell population also showed elevated staining compared to PDAC/vector cells, but the inhibitors had minimal effects. Y-27632 reduced staining in the non-invading cells but overall had no significant effect, whereas C3 treatment reduced staining of the whole population by 40% in PDAC/Δ122 cells. These results demonstrate that inhibition of JAK-STAT and RhoA-ROCK pathways reduces Δ122p53induced invasion, a process requiring IL-6.
RhoA-ROCK signalling is upregulated by Δ133p53 isoforms. To confirm that the RhoA-ROCK signalling pathways is upregulated by human Δ133p53 isoforms as implied by the above data, HCT116 cells were transfected with the three Δ133p53 isoforms (Δ133p53α, Δ133p53β and Δ133p53γ) and allowed to invade through Matrigel, in the presence and absence of the Y-27632 ROCK inhibitor. All three isoforms invaded the Matrigel, which was blocked by Y-27632 (Fig. 6a). Recently, it was reported that the two homologous ROCK isoforms are not functionally equivalent 32,33 . To determine whether both ROCK isoforms have identical effects on the Δ133p53-dependent 'rounded' phenotype important for migration, we measured expression of ROCK1 and ROCK2 in transfected cells (Fig. 6b, d). Expression of ROCK1 was slightly increased in adherent cells, as compared to controls, but was decreased in the rounded, non-adherent cells ( Fig. 6c for an example of rounded cells-green; Fig. 6b for quantitation of ROCK1 levels). However, ROCK2 expression level was high in all Percentage of tumour types identified by histopathological analysis displayed as pie charts for each genotype. n = cohort size. All cohorts of mice were followed for up to 600 days and mice were culled when signs of disease or distressed were observed Summary of χ 2 -statistic and p-value for no evidence of malignancy (NEM) between the genotypes *p < 0.05 was considered to be significant ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-02408-0 situations (Fig. 6d). This suggests that ROCK2 is preferentially required for the rounded cell movement. ROCK-driven actin reorganisation during motility is largely dependent on RhoA 27,34 . We therefore asked if Δ133p53 isoforms exert their effects on ROCK signalling via RhoA by testing whether the active GTPbound form of RhoA is elevated in cells transfected with the isoforms. Control levels of RhoA were found in adherent HCT116 cells, but this was markedly increased in rounded cells, particularly those transfected with Δ133p53β, which showed 2-3 times more GTP-RhoA than cells transfected with Δ133p53α or Δ133p53γ (Fig. 6e).
Elevated Δ133p53 is associated with higher cancer recurrence. The above studies suggest that elevated levels of the Δ133p53 isoform(s) may cause a more aggressive disease. To test this, we analysed 35, mainly stage II, human colorectal tumours (CRC) for Δ133TP53 and FLTP53 mRNA levels and compared these with patient disease-free survival. Colorectal tumours are known to be driven by inflammation 35 and importantly, mutant p53 has been demonstrated to sustain inflammation in colorectal cancer 36 . Our analysis showed that Δ133TP53 and TP53β transcript variants were abundantly expressed in a subset of these tumours compared to the FLTP53 transcript. Interestingly, the relative Δ133TP53 transcripts correlated with TP53β transcript (R 2 = 0.99, p = 0.0, two-tailed t-test; Fig. 7a) but not with FLTP53 (R 2 = 0.06, p = 0.2, two-tailed t-test; Fig. 7b). This suggests that the predominant Δ133TP53 family transcript is Δ133TP53β. Our data also show that there was no significant difference in disease- Summary of χ 2 -statistic and p-value for local vs. extensive organ involvement by T lymphomas between the genotypes and number of metastasis of sarcomas between the genotypes *p < 0.05 was considered to be significant NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-02408-0 ARTICLE NATURE COMMUNICATIONS | (2018) 9:254 | DOI: 10.1038/s41467-017-02408-0 | www.nature.com/naturecommunications free survival of patients with tumours expressing the highest level of the FLTP53 mRNA ( Fig. 7c; p = 0.67, log rank test), but those expressing the highest levels of the Δ133TP53 mRNA (>80th percentile) were associated with a significantly shorter disease-free survival ( Fig. 7d; p = 0.046, log rank test). The tumours with elevated levels of Δ133TP53 mRNA all had a wild-type TP53 gene (based on sequencing exons 5-8).
In addition to tumour stage, decreased lymphocytic infiltration has been reported to be associated with poor prognosis for CRC patients 37,38 . Analysis of our cohort showed that tumours with low lymphocytic infiltration had significantly increased levels of Δ133TP53β mRNA expression ( Fig. 7f; p = 0.037, two-tailed unpaired t-test), however, this was not the case for the FLTP53 transcript ( Fig. 7e; p = 0.37, two-tailed unpaired t-test). Moreover, lymphocytic infiltration alone failed to significantly discriminate patients with shorter disease-free survival (p = 0.452, log rank test).
Next, to identify pathways associated with Δ133TP53 mRNA expression in CRCs we performed Spearman's correlation analyses combining relative Δ133TP53 mRNA expression, determined by quantitative reverse transcription-PCR (RT-qPCR), with expression of genes from the Affymetrix Human Exon 1.0 ST arrays. We identified that Δ133TP53 mRNA was associated with the expression of 364 genes (Spearman's correlation coefficient ρ ≥ 0.35; Supplementary Data 1). Of the GO biological processes over-represented by genes associated with Δ133TP53 mRNA, leukocyte activation (GO:0045321 Bonferroni-corrected p = 0.015) and G-protein coupled receptor (GPCR) signalling pathway (GO:0007186 Bonferroni-corrected p = 0.01) were common. It is of note that GPCR signalling associated with Δ133TP53 mRNA expression in these tumours is upstream of the JAK-STAT and RhoA-ROCK pathways.

Discussion
Recent years have seen p53 implicated in many aspects of immune regulation; particularly those involved in the immune response to infection by viruses and other pathogens (reviewed in ref. 39 ). Multiple Toll-like receptor genes are transcriptionally upregulated by p53 40 , as well as several cytokine and chemokine genes, including IFN-α, IFN-β and CCL2 41,42 . In contrast, in other studies, p53 has been shown to inhibit STAT1, the transcription factor required to transactivate IFN-inducible genes and pro-inflammatory cytokines (reviewed in ref. 43 ). p53 also represses transcription of IL-6, IL-12 and TNF-α genes and other nuclear factor κB (NFκB)-regulated promoters (reviewed in ref. 44 ), but it has also been reported to cooperate with NFκB 45 . p53 plays a direct role in the removal of damaged host cells in the innate immune response, by upregulating the DD1α gene required for clearance of apoptotic cells 46 . p53 upregulates the immune checkpoint molecules programmed cell death-1 (PD-1) and its ligand PDL-1 46 but p53 also upregulates mir-34, which binds the 3′-untranslated region of PDL-1 to downregulate expression 47 . In addition, p53 may directly affect the adaptive immune response as antigen-presenting (dendritic) cells derived from p53 null cells fail to elicit a cytotoxic T-cell response due to a failure to produce IL-12 48 . Thus, although p53 appears to regulate multiple components of the immune response, its roles are often contradictory.
One way perhaps to rationalise these differences is to consider the role of p53 isoforms. As outlined in the introduction, Δ122p53 mice show profound inflammatory phenotypes and elevated cytokine levels 15 and MEF cells derived from these mice secrete higher amounts of IL-6 and the chemokines CCL2, CCL3 and CCL4 than cells with wt p53 16 . Furthermore, splenocytes from Δ122p53 mice show increased transcription of genes in the The average % of phalloidin-stained cells with elongated actin fibres was quantitated. This was done with 'blinding' by two independent researchers. c The average number of migrated MEFs/field 4 h after seeding into Transwell inserts. The bars represent the mean and error bars represent ±s.e.m. of 15 independent fields. A p-value was obtained after performing a heteroscedastic, two-tailed unpaired t-test between different groups. p < 0.05 was considered significant. *p < 0.05 and ***p < 0.0005 are designated, respectively IFN-γ pathway, and cells expressing human Δ133p53 isoform were found to regulate the STAT1 gene 14 . In addition, infection of gastric epithelial cells with Helicobacter pylori activated the proinflammatory NFκB pathway, and this required the Δ133p53 isoform 49 . Thus, the positive effects of p53 on immune regulation, for example, in response to virus infection, might be attributed to the p53 isoform, whereas the repressive effects may be due to fulllength p53. Depending on the activating signal one or other of these two p53 species will dominate. However, if there is sustained expression of the isoform as occurs in the Δ122p53 mice, then inflammatory pathologies may ensue.   Given this context, we were interested to ask whether IL-6 is an important contributor to the phenotypes of Δ122p53 mice. To do this we generated Δ122p53 IL-6 null mice. We found that loss of IL-6 resulted in reduced levels of multiple serum cytokines and chemokines that are part of the JAK-STAT3 signalling pathway and a reduction in tumour incidence; and where there were sufficient tumours to analyse, mice with loss of IL-6 showed reduced spread to other organs. We also found that IL-6 is required for cell migration, by altering the polarisation of actin fibres, a process driven by RhoA-ROCK signalling, which we found to be upregulated in cells expressing Δ122p53 or overexpressing Δ133p53. Moreover, cell invasion induced by Δ122p53 or human Δ133p53 was prevented with inhibitors of JAK-STAT and RhoA-ROCK signalling.
To link the findings from our mouse studies to human malignancies we analysed a cohort of 35 colorectal cancers for expression of Δ133TP53 mRNA, associated signalling pathways and pathologies. Of interest, we found that the tumours with elevated Δ133TP53 mRNA were enriched for GPCR signalling upstream of JAK-STAT and RhoA-ROCK consistent with the Δ122p53 mouse studies. Moreover, these same tumours had significantly lower levels of infiltrating lymphocytes and the corresponding patients had a shorter disease-free survival 50 . There is growing evidence that the presence of immune cells in tumours has a profound effect on tumour progression 51,52 and this is well documented for colorectal cancers where a low 'immunoscore' is associated with poor outcome 37,38 ; consistent with our observations. Thus, given our data, it is tempting to speculate that the Δ133p53 isoform controls the expression of different cytokines dependent on tissue context, to influence the migration of various immune cells in the tumour microenvironment.
Taken together, we propose that some human cancers will have increased levels of the Δ133p53 isoform. As elevated Δ133p53 is only observed in cells with wt TP53 it seems likely that it is the activation of p53 that transactivates Δ133TP53 isoform 53,54 . Increased Δ133p53 (by some mechanism) elevates the levels of IL-6 and other pro-inflammatory cytokines. IL-6 is secreted and then binds its receptor to activate the JAK-STAT3 and RhoA-ROCK pathways 28 . This results in activation of the NFκB pathway and the generation of multiple pro-inflammatory chemokines that contribute to migration of the tumour cells, as well as to alterations in the actin cytoskeleton, that promote an invasive phenotype. These cytokines will also contribute to the migration of immune cells into the tumour. This process is outlined in Fig. 8. Finally, our data suggest that patients whose cancers have elevated Δ133p53 may benefit from treatment with inhibitors of the JAK-STAT and RhoA-ROCK signalling pathways.

Methods
Animal cohorts. Δ122p53 mice were generated by mating mΔpro mice 55 with the CMV-Cre recombinase mice to delete the regions between the LoxP sites (exons 3 and 4) 15 . To generate Δ122p53 and IL-6 null genotypes, Δ122/Δ122 IL-6+/+ mice were crossed with IL-6−/− mice 17 . Female mice without a functional p53 allele, suffer significant decreases in embryonic implantation, pregnancy rate and often fail to produce offspring 56 . This is also true for homozygous female Δ122p53 mice 15 . Therefore, in order to successfully breed all the desired genotype combinations, all female breeders needed to have at least one wt p53 allele. Additionally, male mice without a wt p53 allele had a reduced lifespan thus limiting their ability to be productive breeders. Thus, male Δ122/Δ122 IL-6+/+ were crossed with female p53+/− IL-6−/− to generate Δ122/+ IL-6+/− mice. Subsequently, these were mated together repeatedly to generate all the genotype combinations, although as few females were generated in this cross (Supplementary Table 2), many breeding pairs were required to ensure sufficient females were obtained. In addition, to increase the numbers of Δ122/Δ122 IL-6−/−, a number of male Δ122/ Δ122 IL-6+/− mice were mated with Δ122/+ IL-6+/− mice, which reduced the numbers Δ122/Δ122 IL-6+/− for the study. Animal survival and pathological analyses. Mice were aged for 600 days and terminated either when visible tumour was apparent, when they lost 20% of body weight, or showed other signs of distress, whichever was earliest. Histological examination was carried out to identify pathology, and tumours were collected and fixed in 10% neutral buffered formalin for at least 24 h followed by standard histological processing and wax embedding. Four micrometre sections were cut followed by hematoxylin and eosin (H&E) staining. Immunohistochemistry analysis was performed on tumour sections to confirm the histopathologic examination. CD3 antibody (#ab5690, 1:400, Abcam, UK) was used to detect T lymphocytes; and B220 (#clone RA3-6B2, 1:100, BD Pharmingen, USA) and paired box transcription 5 (PAX5, # 24/Pax-5, 1:100, BD Pharmingen) antibodies were used to detect B lymphocytes, respectively. All animal experiments were carried out under the respective institutional ethical approval University of Otago (AEC 03/12) and Children's Medical Research Institute (AEC C257).
Patient samples. Primary tumour samples were collected from 35 colorectal cancer patients undergoing elective surgery at Dunedin Hospital. A database of clinico-pathological information matched to each of the samples was collected prospectively over a follow-up period of at least 5 years (Supplementary Table 3). These samples were prospectively collected between 1996 and 2007. Informed patient consent was obtained for the use of these tissues and data. Ethical approval had been granted for use of these tissues and data (New Zealand Health & Disability Ethics Committee Reference #14/NTA/33).
RNA extraction and RT-qPCR. Total RNA was isolated from snap-frozen tumours as per the manufacturer's protocol using the PureLink RNA mini kit (Invitrogen, USA). RNA (2 µg) extracted from the tumour tissue was reverse transcribed using qScript cDNA SuperMix (Quanta Biosciences, USA), according to the manufacturer's instructions. Primers were designed for specific TP53 transcript subclasses (FL, Δ133TP53 TP53α and β) 18 . Three reference genes GAPDH, HPRT and  Fig. 8 Model of how the Δ133p53 isoform regulates the JAK-STAT3 and RhoA-ROCK pathways to promote inflammation and invasion. Wt p53 is activated by some (as yet undefined) stress signal, which turns on Δ133p53. Δ133p53 increases IL-6, which, then binds to its receptor and signals through the JAK-STAT pathway to amplify the IL-6 signal further. The JAK-STAT pathway, now constitutively activated by Δ133p53, drives actin polarisation, cell migration via RhoA-ROCK signalling and alters migration of immune cells qPCR was calculated using the equation: Amplification efficiency^À ðgeometric mean reference gene threshold Ct ÀTP53 threshold CtÞ Survival analysis and correlation analysis. Colorectal tumours were divided into two groups based on the 80th percentile cut-off of the relative expression of either the FLTP53 or Δ133TP53 transcript determined by RT-qPCR. Difference in disease-free survival between the two groups (>80th percentile vs. <80th percentile) of FLTP53 and Δ133TP53 mRNA expression was assessed using Kaplan-Meier analysis followed by log rank test and Cox proportional hazard analyses. This analysis was performed using the survival library in R. A log rank test p < 0.05 was considered significant.
Exon array analyses. RNA quality extracted from the colorectal tumours was assessed using RNA 6000 NanoChips with the Agilent 2100 Bioanalyzer (Agilent, USA). Samples with a RIN value of >7 were used for the arrays. Exon array hybridisation was performed as per the manufacturer's protocol. In brief, biotinlabelled target was prepared using 1 μg of total RNA from the tumours. Initially, a rRNA removal procedure was performed with the RiboMinus Human/Mouse Transcriptome Isolation Kit (Invitrogen) followed by cDNA synthesis using the GeneChip WT (Whole Transcript) Sense Target Labelling and Control Reagents kit as described by the manufacturer (Affymetrix). The sense cDNA was then fragmented by uracil DNA glycosylase and apurinic/apyrimidic endonuclease 1 and biotin-labelled with terminal deoxynucleotidyl transferase using the GeneChip WT Terminal labelling kit (Affymetrix). Hybridisation was performed using 5 µg of biotinylated target, which was incubated with the GeneChip Human Exon 1.0 ST array (Affymetrix) at 45°C for 16 h. Following hybridisation, arrays were washed and stained using a Fluidics Station 450 (Affymetrix) and the Affymetrix wash and stain kit following the manufacturer's instructions using Affymetrix fluidics protocol FS450_0001. Nonspecifically bound material was removed by washing and specifically bound target was stained using the anti-streptavidin-phycoerythrin stain provided in the wash and stain kit. The arrays were scanned using the GeneChip Scanner 3000 7 G (Affymetrix) and raw data were extracted from the scanned images and analysed with the Affymetrix Power Tools software package (Affymetrix).
After the microarrays were scanned, quality checks were performed using the Affymetrix Expression Console Software (2012, available from http://www. affymetrix.com/estore/browse/level_seven_software_products_only.jsp? productId=131414&categoryId=35623#1_1), the intensity data were subjected to quantile normalisation and summarisation using the RMA algorithm in the R 'affy' 57 and 'aroma.affymetrix' 58 packages. Analysis was then performed on transcript-level probesets using the core annotations. At the transcript level, 18 707 probesets represented different transcripts.
Δ133TP53 mRNA expression determined by RT-qPCR was combined with the transcript-level Affymetrix exon array data for each CRC. Spearman's correlation was then performed using the cor() function in R with the method set to 'spearman' 59 . Enrichment of gene sets within the input list was determined using the overrepresentation test with default settings using Bonferroni correction in Pantherdb 60 .
Transwell assays. MEFs were serum-starved in medium with 0.5% FBS for 24 h then harvested, resuspended in serum-deficient medium and seeded into Transwell inserts. After 4 h, cells were fixed using 4% paraformaldehyde, stained with 3% crystal violet, imaged using an Olympus DP71 microscope and analysed using ImageJ software. Data shown are from three biological replicates, each with six technical replicates.
Staining of actin filaments with phalloidin. Actin filaments were stained using phalloidin. Cells were seeded into a 24-well plate and a scratch wound introduced using a p200 tip when the cells were confluent. Cells were allowed to migrate to close the wound for 4 h, then fixed in 2% paraformaldehyde in phosphate-buffered saline, blocked and permeabilised in 5% bovine serum albumin and 0.2% Triton, then stained with phallodin-568 (A12380; Alexa Flour; 1:500; Molecular Probes, USA) to label actin filaments and Hoescht 33342 (H1399; 1:1000; Molecular Probes) to label the DNA. Cells were imaged using the Olympus IX71 inverted microscope (Tokyo, Japan) at ×200 magnification and images were merged using Adobe Photoshop. Images of DNA stained with Hoescht were taken first, and then of the filamentous actin stained with phalloidin. This ensured consistency of imaging and no bias towards intensely polarised areas of actin. Quantitation was done with 'blinding' to genotypes by two independent researchers. The amount of actin elongation as a percentage was determined by dividing the number of polarised actin filaments counted by the total nuclei counted. Fifteen images per slide were taken at ×200 magnification, with an average of 20 nuclei per image, making on average 300 nuclei counted for each treatment.
Organotypic invasion assays. Invasion of PDAC/Δ122 and PDAC/vector cells through collagen-I matrices was performed 29,30 . Collagen-I was extracted from rat tails and purified to a concentration of 2.5 mg/ml. TIFs were embedded in collagen (8 × 10 4 /matrix) and spontaneously remodelled collagen for 12 days in DMEM, supplemented with 10% FBS. Medium was renewed at day 6 of contraction. In all, 4 × 10 4 PDAC cells (transduced with either empty vector or Δ122p53) were then seeded on to the contracted matrices, grown for 4 days before being transferred to an air-liquid interface and allowed to invade towards a chemotactic gradient for 12 days. During invasion, PDAC cells were treated with the Pan-JAK inhibitor P6 5 μM 28 , the ROCK inhibitors Y-27632 (10 µM) and H1152 (5 µM) 30 , and the RhoA inhibitor TAT-C3 (0.25 µg/ml) 29 . Following invasion, samples were fixed in 10% neutral buffered formalin for 24 h and embedded in paraffin for histological analyses 30 . Histological staining was performed on 4 μm sections deparaffinised in xylene and rehydrated using graded ethanol washes. Antigen retrieval was performed for 30 min in a water bath (93°C for organotypic samples and 100°C for mouse tissues). For immunohistochemistry staining, samples were quenched in 3% H 2 O 2 prior to application of protein block (Dako, Agilent). Primary antibodies were incubated (pan-cytokeratin 1:50 C-11 Leica Novocastra, UK; pMYPT1 (Thr696) 1:100 #ABS45 Millipore, USA; pSTAT3(Tyr705) 1:400 #9145 Cell Signaling Technology, USA) and secondary antibodies (Envision, Agilent) coupled to horseradish peroxidase (HRP) were applied. Detection was performed with diaminobenzidine. H&E staining and counterstaining were operated on a Leica autostainer. Invasion index was calculated as per the formula: invasive index = (invasive cells)/(non-invading cells + invasive cells). Data shown are from three technical replicates with three biological replicates each.
Invasion assays. Quantification of cell invasion was done in Transwell chambers containing fluorescence-blocking polycarbonate porous membrane inserts (Fluoroblock; #351152; BD Biosciences, USA; pore size 8 μm). A volume of 100 μl of 2 mg/ml Matrigel with reduced growth factors (a commercially prepared reconstituted BM from Englebreth-Holm-Swarm tumours, #354230; BD Biosciences) were prepared in a Transwell. Cells were transfected and treated with Y-27632 (10 µM, Calbiochem, USA) as monolayers before trypsinisation and plating (1 × 10 5 ) in 2% FBS containing medium on top of a thick layer (around 500 μm) of Matrigel contained within the upper chamber of a Transwell. Controls were left untreated. The upper and lower chambers were then filled with 2% FBS containing media and media with 10% FBS, respectively, thus establishing a gradient of chemo-attractant that permits cell invasion throughout the Matrigel. Cells were allowed to invade at 37°C, 5% CO 2 through the gel before fixing for 15 min in 3.7% formaldehyde. Cells that had invaded through the Matrigel were detected on the lower side of the filter by green fluorescent protein fluorescence and counted. Each assay was performed twice in triplicate for each condition.
Cell extracts and western blotting. Medium containing rounded cells was centrifuged at 1200 r.p.m. for 5 min and the pellet containing the rounded invasive cells lysed. The remaining adherent cells were gently scraped in lysis buffer. The two populations of cells were analysed separately. Total protein concentration was determined using the BCA kit (Promega, USA). Protein samples were electrophoretically separated on 8% SDS-PAGE gels and an equal amount of total protein (30 µg) was loaded into each lane. Proteins were transferred onto nitrocellulose NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-02408-0 ARTICLE NATURE COMMUNICATIONS | (2018) 9:254 | DOI: 10.1038/s41467-017-02408-0 | www.nature.com/naturecommunications membranes. Membranes were blocked with TBS/0.1% Tween 20 containing 3% milk for 1 h and then incubated overnight with the primary antibodies (mouse anti-ROCK I at 1:400 mouse anti-ROCK II at 1:2500, BD Transduction Laboratories, USA; anti-myc at 1:500 9E10, Thermofisher Scientific, USA) diluted in TBS/ 0.1% Tween 20 containing 3% milk. After several washes in TBS/Tween, membranes were incubated with anti-mouse Ig antibodies linked to HRP. Membranes were developed with ECL according to the manufacturer's instructions. Scanned autoradiographs were quantified using AIDA/2D densitometry software.