Cancer-Specific Loss of p53 Leads to a Modulation of Myeloid and T Cell Responses

1There is strong evidence that cancer cells have the potential to be recognized by the immune system but that they can mobilize various mechanisms of immune evasion and escape, such as upregulation of immune checkpoint proteins to dampen T cell effector responses. Current human studies have shown durable and complete responses to immune checkpoint blockades in a number of tumors; however, for reasons that are not completely clear, a sizable proportion of cancers fail to respond. It is apparent that the constellation of oncogenic events that leads to full neoplastic transformation can influence the effector function of the immune response in several ways. Oncogenic RAS can promote expression of various cytokines that result in an inflammatory response, which is thought to promote cancer progression. RAS signaling also increases tumor cell expression of PD-L1, thereby suppressing activated T cells, and increases secretion of granulocyte-macrophage colony-stimulating factor (GM-CSF) to promote pancreatic neoplasia. The co-activation of KRAS and MYC in lung tumors restructures macrophage and T cell responses in a CCL9 and interleukin-23 (IL-23)-dependent manner. Loss of PTEN, another common oncogenic event, results in resistance to PD-1 blockade in both melanoma and uterine leiomyosarcoma, while β-catenin signaling in melanoma was shown to limit T cell infiltration.

2p53 is best understood as a tumor suppressor. However, in the immune compartment, p53 also functions as a negative regulator of autoimmunity by supporting regulatory T (Treg) cells, through directly upregulating Foxp3 and promoting STAT5 activity, and restricting STAT3 in the pro-inflammatory helper T cells (T helper 17 [Th17] cells). Expression of p53 in macrophages leads to both an inflammatory response through co-operation with nuclear factor κB (NF-κB) and an anti-inflammatory response through STAT1 inhibition. In the context of cancer, activation of p53 in the tumor stromal compartment has been shown to promote a tumor-restricting immune response. Induction of p53 in hepatic stellate cells (HSCs) results in senescence and the senescent-associated-secretory phenotype (SASP) that drives M1-macrophage polarization and limits cancer progression. Conversely, HSCs lacking p53 induce the differentiation of macrophages toward the tumor-promoting M2 state. Stromal loss of p53 changes the cytokine secretion pattern to promote myeloid-derived suppressor cells (MDSCs), thereby accelerating tumor growth. Interestingly, activation of p53 in the tumor microenvironment using local injection of the MDM2 inhibitor Nutlin selectively eradicated tumors that were rich in leukocytes. This response was dependent on stromal-p53 expression. These studies show that p53 levels in the stroma shape the inflammatory responses that influence tumor progression.

3Despite the clear role of p53 in immune regulation, relatively few studies have examined how p53 status of the cancer cells affects the immune response in vivo. Reactivation of p53 in established liver cancers induced senescence and SASP, which promoted polymorphonuclear (PMN) infiltration and tumor regression. Futher studies showed that the induction of p53-dependent senescence in hepatocellular carcinomas was accompanied by the elimination of the tumor cells through a mechanism dependent on natural killer (NK) cells. In silico correlations between the retention of wild-type (WT) p53 expression and immune infiltration in breast and head and neck cancers have also been noted. However, a recent study of a PTEN-driven prostate cancer model indicated that concomitant loss of p53 enhanced tumor infiltration of CD11b+Gr1+ PMN cells. The recruitment of this myeloid population was through increased CXCL17 secretion by p53-null prostate cancer cells, and their role in promoting tumor development was associated with the expansion of immunosuppressive Treg cells. Similar findings were observed in mouse models of breast cancers, where loss of p53 increased frequencies of circulating and tumor neutrophils through unchecked WNT signaling, resulting in enhanced metastasis.

4In this study, we show that tumor-specific loss of p53 expression in both autochthonous lung and pancreatic tumor models correlates with changes in the tumor microenvironment. Using KRAS-driven pancreastumor-derived cancer cells as a model of p53 loss, we demonstrate that p53 deletion can promote immune tolerance through the recruitment of both myeloid cells and Treg cells. The enrichment of these suppressive populations results in enhanced protection of p53-null cancer cells from immune-mediated elimination. Furthermore, concomitant activation of KRAS and loss of p53 coordinate to promote immune tolerance.

5Tumor growth involves a complex interaction between stromal cells (of mesenchymal and immune origin) and cancer cells. Numerous studies have shown a role for macrophages in supporting cancer progression, and so we examined whether loss of p53 in autochthonous mouse models of pancreatic and lung cancers could influence myeloid cell recruitment to the tumor microenvironment (TME). Immunohistochemistry (IHC) sections were analyzed for F4/80+ immune cells in pancreatic tumors derived at equivalent endpoints from a pancreatic ductal adenocarcinoma cell (PDAC) model driven by pancreas-specific mutations in KRASG12D with either wild-type p53 (KC model; Pdx1-cre; LSL-KrasG12D/+) or one floxed Trp53 allele (KFC model; Pdx1-cre; LSL-KrasG12D/+; Trp53fl/+) (Figure 1A, left). Tumors derived from KFC pancreatic tumors revealed increased macrophage F4/80+ staining (Figure 1A, right). We also examined an epidermal growth factor receptor (EGFR)-driven model of lung cancer with (ROSA26tTA-LSL-EGFRL858R, EL) and without p53 (Rosa26tTA-LSL-EGFRL858R;Trp53fl/fl, EFL). Tumors excised at similar endpoints were digested and stained for flow cytometry to detect CD11b+ and F4/80+ tumor immune infiltrates. Frequencies of CD11b+ myeloid-derived cells were significantly increased in EGFR-driven tumors null for p53, which were also enriched in CD11b+F4/80+ macrophages (Figure 1B).

6To gain insight into the interplay between macrophages and the p53 status of cancer cells, we isolated primary cancer cells from three independent KC and KFC tumors. KC-tumor-derived cells lines retained p53 expression and activity, showing growth inhibition in response to the p53 activator Nutlin, while cell lines derived from KFC tumors did not express p53 and were resistant to Nutlin treatment (Figures S1A and S1B). Previous studies have also shown that KFC tumors undergo loss of heterozygosity and become p53 null during tumor development. Reexpression of p53 in tumors can induce their ability to produce various inflammatory cytokines, prompting us to examine whether conditioned media from the KC- and KFC-tumor derived cells could impact macrophage surface activation markers: major histocompatibility complex (MHC) class I, MHC class II, PD-L1, and CD80. In comparison to untreated bone-marrow-derived macrophages (BMDMs), conditioned media from all PDAC cell lines induced expression of all the activation markers to similar levels (Figure S1C), although this was not affected by p53 status. As both our autochthonous models (pancreas and lung) displayed a p53-associated change in macrophage infiltration, we examined the effect of conditioned medium on BMDM migration and chemotaxis. Interestingly, BMDMs exposed to conditioned media from KFC-derived cell lines displayed increased migration, demonstrated by enhanced wound healing and chemotaxis, compared to conditioned media from KC cell lines (Figures 1C, 1D, S1D, and S1E).

7To explore the basis of this difference in the activity of the conditioned media, we analyzed cytokine secretion from three independently derived p53WT KC and p53-null KFC tumor-derived cell lines. Out of the 35 cytokines tested by a cytokine Luminex array, 20 were detectable and only 5 (CCL11, CXCL1, CXCL5, CCL3, and macrophage colony-stimulating factor [M-CSF]) were elevated in the three KFC cell lines (Figure 1E). While these myeloid-attracting and differentiating cytokines were elevated, T-cell-associated cytokines were not dependent on p53 expression (Figure S1F). Taken together, these results indicate that the KFC tumors secrete cytokines that have a myeloid-macrophage stimulating effect.

8To understand the consequences of the p53 status of our tumor-derived cell lines on the myeloid compartment in vivo, we used a T-cell-deficient CD1nu/nu subcutaneous tumor model. In order to track accurately and in an unbiased fashion in vivo tumor growth, KC and KFC cells were engineered to express near-infrared fluorescent protein (iRFP) (Figure 1F). While the initial GEMM KC and KFC tumors arise at different rates, cells derived from these tumors grew at similar rates in CD1nu/nu recipients, as assessed by in vivo real-time imaging of the tumors (Figures 1G, S1G, and S1H). Flow cytometry analysis of tumor digests at endpoint revealed no changes in total frequency of CD11b+ myeloid infiltration; however, CD11b+F4/80+ macrophages were enhanced in KFC tumors, as seen in the autochthonous mouse models (Figures 1H and S1I). These results demonstrate that p53 ablation in the tumor can influence the TME independently of tumor growth.

9Loss of p53 in the tumor-derived cell lines did not accelerate subcutaneous tumor growth in athymic recipients, despite the increase of CD11b+F4/80+ infiltrates in the TME. We therefore turned to examine the effect of p53 loss in an immunocompetent model. Since our tumor-derived cells originated from mice of a mixed background, we used a tumor rejection model that has been previously employed to explore contributions of tumor-associated genetic alterations on a complete immune response. Mixed strain KC- and KFC-derived cells were injected subcutaneously into pure FVB recipient mice, an immunocompetent MHC-mismatched strain (Figure 2A). The p53WT KC1 cells followed the expected growth kinetics of initial expansion until day 7, followed by progressive rejection. Surprisingly, the p53-null KFC1 cells continued to grow until the point where ethical considerations determined the termination of the study (Figure 2B). A similar pattern of delayed rejection in p53-null cells was seen in two further independently derived KFC cells (Figures S2A and S2B). Further examination of the tumor digests taken from these mice at day 7 showed an increase in F4/80+ macrophages in p53-null (KFC1) tumors (Figure 2C). In order to confirm that the p53-dependent effect on tumor-associated myeloid cells was not site specific, we performed pancreatic orthotopic injections of KC1 and KFC1 cell lines in FVB mice. In line with the genetic tumor models and the subcutaneous models (Figures 1A, 1H, and 2C), p53-null orthotopic tumors displayed increased myeloid infiltration (Figure 2D).

10One potential caveat with tumors arising in genetically engineered mice is that additional genetic alterations acquired during tumor development may also contribute to tumor-immune interactions. To test directly the effect of p53 loss, we used gene editing to delete p53 from two KC-p53WT cell lines (KC1 and KC2), generating matched isogenic cell lines named KC1-p53KO and KC2-p53KO (Figure 2E). BMDM chemotaxis assays using conditioned media from KC1-p53WT and KC1-p53KO PDAC cell lines confirmed our previous findings that KFC-p53KO conditioned media enhanced BMDM migration (Figure 2F). ELISAs for CCL11, CXCL1, CXCL5, CCL3, M-CSF, and MCP1 (Figure 2G) showed that as seen in the tumor-derived KFC cell lines, p53 loss increased production of these cytokines (Figures 2G and S2C). Additionally, we performed an unbiased analysis of 62 cytokines in conditioned media from KC1-p53WT and KC1-p53KO cells, which further revealed increased CXCL10 and 11, and decreased tumor necrosis factor α (TNF-α) secretion (Figure 2H). Of note, CXCL10 and 11 uniquely bind to the CXCR3 receptor. Our results show that cancer cells deficient for p53 increase the production of chemokines involved in myeloid recruitment and macrophage differentiation.

11Turning back to the in vivo rejection model, we found that the isogenic p53KO cells displayed delayed rejection kinetics, similar to PDAC cells that lost p53 during tumor development (Figures 2I and S2D). At an early time point (3 days post-injection), CD11b+ myeloid cells were enriched in p53-deleted tumors in comparison to their parental p53WT controls (Figure 2J), although this difference was lost by day 7 (Figure S2E). Next, we looked more closely for potential in vivo consequences of the changes in cytokine secretion detected in vitro. M-CSF is a key driver of myeloid differentiation and proliferation, and we detected a 2-fold increase of proliferating CD11b+Ki67+ myeloid cells in p53KO tumors (Figures 2K and S2F), correlating with enhanced M-CSF production by these cells. We also probed for myeloid cells expressing CXCR3 (the receptor for CXCL9-11) and CCR2 (one of the receptors for CCL11 and uniquely a receptor for MCP1) within our tumor digests. Interestingly, both CXCR3+ and CCR2+ infiltrating CD11b+ myeloid cells were significantly enhanced within KC1-p53KO tumors (Figures 2L, 2M, and S2G), correlating with increased production of the cytokines for these receptors by the p53-deleted tumor cells. We further confirmed an increase in intra-tumoral CD11b+ cells expressing CXCR3 in the second isogenic tumor pair at day 3 (Figure S2H). Systemic cytokine changes were also observed in tumor-bearing mice. At endpoint (day 14 post-tumor challenge), mice harboring p53-null tumors displayed significantly increased serum levels of MCP1 (Figure 2N). In conclusion, data from both GEMM-derived cells and the isogenic pairs show that loss of p53 in tumor cells promotes the production of cytokines involved in myeloid recruitment and homeostasis.

12Our observation that loss of p53 affects tumor growth in immunocompetent, but not athymic, mice indicated that an intact T cell response is required for the rejection of p53-expressing cells. To test whether the CD11b+ cells found in the KC1-p53KO tumors were immune suppressive, we isolated CD11b+ cells from the tumors at day 3 and co-cultured them with preactivated T cells stained with the v450 proliferation dye. While the myeloid cells derived from both tumors attenuated T cell proliferation, the myeloid cells isolated from KC1-p53KO tumors more robustly inhibited CD4+ and CD8+ T cell proliferation (Figures 3A, 3B, S3A, and S3B). These observations were confirmed in the second pair of isogenic cells (Figures S3C and S3D).

13Cancer cells have been shown to influence the crosstalk between myeloid cells and T lymphocytes, so we examined the effect of conditioned media from KC1-p53WT and KC1-p53KO cells on BMDM-dependent T-cell-activating functions (Figure 3C). Focusing on cytokines involved in CD4+ T helper and CD8+ cytotoxic T lymphocyte (CTL) differentiation, we found only two of the seven cytokines showed significantly changed mRNA expression in response to KC1-p53KO conditioned media; namely, a strong increase in expression of Il6, a T helper 1 (Th1) cell antagonist, and a subtler decrease in expression of Il12, a Th1-cell- and CTL-promoting cytokine (Figure 3D). These changes predict that BMDMs conditioned by KC1-p53KO media would show impaired polarization of CD4+ Th1 cells and CTL activation. To verify this hypothesis, we co-cultured BMDMs pulsed with the model antigen ovalbumin (OVA) with T cell receptor (TCR)-transgenic CD4+ or CD8+ T cells recognizing the OVA peptides (OTII [CD4+] or OTI [CD8+] T cells). BMDMs educated by KC1-p53KO conditioned media and loaded with OVA peptide were less effective in differentiating Th1 cells, as demonstrated by reduced interferon-γ (IFN-γ)+TNF-α+-producing OTII CD4+ T cells (Figure 3E). Furthermore, CTL activation was compromised by KC1-p53KO-instructed BMDMs, as shown by the weaker expression of the activation marker CD44 on CD8+ OTI cells (Figure 3F). Led by our mouse data, we stratified pancreatic cancer patients from The Cancer Genome Atlas (TCGA) dataset according to a designed gene list based on classical MDSC markers, including ITGAM (CD11b) and CXCL10 (Figure S3E). Patients with a high MDSC signature displayed statistically lower overall survival (Figure S3F). Interestingly, patients with a high MDSC gene signature trended toward an enrichment of TP53 mutations compared to the low-MDSC-signature population (Figure S3G). Taken together, these data demonstrate that p53 deletion in cancer cells creates a tumor-promoting environment through remodeling myeloid-T cell crosstalk.

14Encouraged by our in vitro and ex vivo results, we assessed T cell activity in mice bearing GEMM-derived or isogenic p53WT and p53-null tumors. Ex vivo restimulation of the spleen and tumor draining lymph node (dLN) on day 7, when tumors were of equivalent sizes, revealed a reduction in anti-tumor T cell responses (as measured by CD4+ Th1 TNF-α+ and IFN-γ+ cells) in mice bearing p53-null tumors from either isogenic (Figures 3G, 3H, and S3H) or the GEMM-derived lines (Figures S3I and S3J). Furthermore, sorted CD4+CD25− T cells from mice bearing KC1-p53KO tumors expressed less Il2 and Ifng mRNA than CD4+CD25−T cells from KC1-p53WT recipients (Figure 3I), suggesting an overall lack of CD4+ T cell activation. Impaired CD8+ CTL IFN-γ+ production further demonstrated weakened T cell responses in mice carrying p53-null tumors (Figures 3J and S3K–S3M). Overall, our data suggest that p53 deletion in cancer cells undermines T cell effector responses, at least in part through co-opting myeloid cell functions.

15Our in vitro data suggested that the dampened T cell responses seen in mice bearing p53KO tumors may be a consequence of pro-tumorigenic myeloid cells responding to elevated M-CSF from p53-ablated cancer cells (Figure 2G). To test the contribution of M-CSF production by p53-null cells to the modulation of the T cell response in vivo, we treated immunocompetent mice challenged with either p53WT or p53-null tumor cells with neutralizing antibodies against CSF1R, the receptor for M-CSF (Figure 4A). CSF1R specificity was confirmed by showing the expected compensatory increase in serum M-CSF levels in treated mice, with no change in serum granulocyte colony-stimulating factor (G-CSF) (Figures S4A and S4B). Subcutaneous tumors were harvested at day 7, and effector function of infiltrating CD4+ and CD8+ T cells was assessed. Consistent with results from spleen and dLN of tumor-bearing mice, vehicle-treated, p53-null tumors contained less activated effector CD4+ and CD8+ T cells secreting IFN-γ and TNF-α (Figures 4B, 4C, and S4C–S4E). CSF1R blockade restimulated T cell function in p53-null tumors to levels seen in p53-WT tumors (Figures 4B, 4C, and S4C–S4E) but did not change the rate of rejection of tumors of either genotype (Figure 4D). Furthermore, T cell reactivation in p53-null tumors was irrespective of absolute tumor size, as confirmed by in vivo imaging of the tumors (Figure S4F). These data suggest that inhibition of M-CSF signaling that is induced by p53-deleted cells can reactivate T cell function within the tumor but fails to promote rejection when used as a monotherapy.

16The failure of CSF1R blockade to promote tumor rejection prompted us to investigate other mechanisms of immune suppression that may reflect additional effects of tumor cells on T cell function. One essential immunosuppressive population that is at the epicenter of T cell immune tolerance are Treg cells, which are currently being targeted for cancer immunotherapies. Flow cytometry analysis of the original untreated tumors showed an accumulation of Treg cells within p53-null tumors with no changes in CD4+ T cell frequencies (Figure 4E). Tumors deleted for p53 were biased toward Treg cells, as assessed by the Treg cell/non-Treg cell ratio of CD4+ T cell infiltrates (Figures 4E and 4F). The inability of CSF1R blockade to reduce intra-tumoral Treg cell frequencies (Figure 4G) led us to perform a double blockade targeting both CSF1R and Treg cells using αCD25 to eradicate Treg cells. This method is well established to effectively deplete Treg cells, although various studies have shown that CD25 blockade alone does not effectively block the growth of established tumors. Using a combined treatment of αCD25 for 3 days prior to tumor challenge followed by an αCSF1R regime, we were able to attenuate the growth of the p53-null tumors in FVB recipients (Figure 4H). These intervention data suggest that immunotherapy for p53-compromised tumors requires targeting both the myeloid and Treg cell compartments for a positive outcome.

17Finally, we examined the consequences of direct interactions between CD8+ cytotoxic T lymphocytes (CTLs) and cancer cells. CTL cytotoxic function is elicited by IFN-γ and granzyme B (GzmB) production and can be affected by cancer cells. CTLs co-cultured with KC-p53WT and KFC-p53KO cells were equally capable of secreting IFN-γ (Figure 4I). However, co-culture with KFC-p53KO cells clearly impaired CTL GzmB production upon restimulation (Figure 4J).

18Tumor cells can evade CTL responses through downregulation of MHC class I surface expression, although this was not observed in the p53-null PDAC-tumor-derived cells (Figure S4G). In order to assess if loss of p53 in cancer cells influences their ability to present endogenous peptides, we introduced the model antigen, OVA, into the isogenic PDAC cell lines. Co-culturing of peptide-stimulated OTI cells with KC1p53WT-mOVA and KC1p53KO-mOVA increased cell death in both PDAC cell lines, as marked by increased propidium-iodide-positive cells. However, loss of p53 did not lead to an increased resistance to OTI killing by MHC class I (Figure S4H). Taken together, these results indicate that immune evasion in response to loss of p53 reflects both an accumulation of suppressive Treg cells and inhibition of T cell cytotoxic functions.

19Our observation that targeting Treg cells in parallel with CSF1R was critical in reducing tumor growth of p53-null cells prompted us to examine the effect of the p53 status of cancer cells on regulatory T cells more closely. Treg cells have different degrees of suppressive abilities, which can be phenotypically identified through surface expression of CLTA-4, KLRG1, and GITR and high expression of CD25. Mice bearing KFC1 p53-null tumors displayed increased systemic presence of CD4+FOXP3+CTLA-4+ T cells in the spleen and dLN (Figures S5A and S5B), suggesting a greater suppressive immune response. To determine whether the skewing toward more suppressive Treg cells depended on the site of the tumor, we orthotopically implanted KC1 and the KFC1 cell lines into the pancreata of FVB recipients. Tumors of both genotypes were detected in the pancreas at day 7 (Figure S5C) with a strong accumulation of highly suppressive Treg KLRG1+ cells in p53-null pancreatic tumors (Figure S5D).

20To validate the p53 specificity of these Treg cell responses, we turned again to the isogenic KC1-p53WT and KC1-p53KO cells. Identical CD4+ T cell frequencies were present in the tumor dLN of both KC1-p53WT and KC1-p53KO recipients (Figure 5A). However, the proportion of FOXP3+ (Treg) cells present within the CD4+ population was enhanced in mice bearing KC1-p53KO cancer cells (Figure 5B). Phenotypic characterization of these Treg cells revealed an increase in the proportion expressing markers of suppression (high CD25, GITR, and KLRG1 expression) (Figures 5C and S5E). In addition, while p53 status did not affect total CD4+ T cell frequency in the tumor (Figure 5D), KC1-p53KO tumors accumulated more FOXP3+ Treg cells, especially those with KRLG1 surface expression (Figure 5E). Consistent with these data, ex vivo Treg cell suppression assays of CD4+CD25+ Treg cells sorted from tumor-bearing mice showed that Treg cells isolated from KC1-p53KO recipients were more efficient at inhibiting the proliferation of CD4+CD25− T cells in vitro (Figures 5F and S5F). Both our in vivo characterization and ex vivo functional assays highlight the relationship between tumor p53 loss and the local and systemic accumulation of regulatory T cell responses.

21In humans, nearly 90% of pancreatic cancer patients show mutations in KRAS (cBioportal), and our autochthonous models were based on mutations in KRAS or in upstream signaling receptors (EGFR). Recent work has shown that activating KRAS mutations can drive an immunosuppressive response in cancer cells through increased PD-L1 expression. In order to examine the role of KRAS activation in promoting tumor tolerance in the context of p53 loss, we used previously described doxycycline (Dox)-inducible KRASG12D-driven p53-null PDAC mouse cells (Figure 6A). Tumors were initiated for 7 days in FVB recipient mice treated with Dox to drive mutant KRAS expression before either maintaining Dox treatment (iKRASG12D-ON) or removing Dox (iKRASG12D-OFF). Despite the lack of p53 in all of these tumor cells, removal of Dox resulted in complete eradication of the tumor within 7 days (Figure 6B). We evaluated early and late tumor-associated immune infiltrates at 48 h and 5 days post-Dox withdrawal. Within 48 h of Dox removal, iKRASG12D-OFF tumors displayed reduced frequencies of CD11b+ cells, especially those positive for F4/80 and CXCR3 expression (Figure 6C), with a modest influx of CD4+ T cells but no changes in CD8+ T cell frequencies (Figure 6D). At this time point, iKRASG12D-ON and iKRASG12D-OFF tumors showed no difference in CD4+ T effector cells or CD4+FOXP3+KLRG1+ suppressive Treg cells (Figure 6E).

22Evaluation of peripheral T cells in recipient mice 5 days after Dox removal showed a significant increase of pro-inflammatory CD4+ and CD8+ T cells in the spleen and a trend toward an increase in the dLNs under iKRASG12D-OFF conditions (Figures S6A and S6B). There was no overall change in suppressive Treg cell populations (Figure S6C). These data suggest that KRAS inactivation in part reactivates T cell effector immune responses but has no impact on the accumulation of regulatory T cells. To further explore the cooperation between p53 loss and KRAS activation, we turned to a previously described model of isogenic C57Bl6/J ID8 ovarian cancer cells with and without p53, which are WT for KRAS. These cells were injected into the peritoneum of syngeneic C57Bl6/J mice (Figure 6F), and as we previously published, there was a skewing toward PMN cells present in the ascites of mice carrying p53KO tumors. Interestingly, while there was no change in the total frequency of CD4+ T cells (Figures S6D and S6E), the ascites from mice injected with ID8 p53KO cells showed a 2-fold increase of regulatory T cells (Figures 6G and 6H) and a preferential skewing toward Treg cells, as illustrated by the FOXP3+ to FOXP3− ratio (Figure S6F). A major site of ovarian cancer spread is to the omental fat or the omentum. Processing and staining of omental tumors for flow cytometry revealed enriched populations of regulatory T cells present in p53-ablated lesions (Figure 6I). These results further support a model in which p53 loss in tumors promotes Treg cell infiltration.

23In this study, we showed that p53-deficient tumors can re-orchestrate the innate immune response through suppression of effector CD4+ and CD8+ T cells, which reflected an increase in the inhibitory actions of myeloid suppressor cells and regulatory T cells, accompanied by a direct impairment of GzmB in CD8+ T cells. Other studies have shown that mutations in p53 correlate with reduced GzmB in gastric cancers and a CSF1 (M-CSF) response signature in breast cancers. Somatic changes in the tumor can also induce systemic changes beyond the tumor microenvironment, and we showed dampened local and systemic T cell effector function in mice harboring p53-deficient tumors, seen as decreased IFN-γ and TNF-α production (Figure 3). We propose that these weakened T cell responses are partially mediated by suppressive myeloid cells derived from the tumor as well as through Treg cell suppressive functions. Conditioned media from p53-null cancer cells alters the cytokine profile of BMDMs, which in turn influences CTL and Th1 cell differentiation (Figure 3). These observations are consistent with previous publications showing that tumor-derived CD11bGr+ cells promote the de novo differentiation of Treg cells in prostate cancer models and reduce circulating IFN-γ-producing T cells in B cell malignancies. Systemic changes were also seen in p53-deficient breast cancer models, which also showed an increase in MCP1 serum levels similar to those seen in our p53-null tumor-bearing mice.

24It is important to note that unlike the response to reexpression of p53 in p53-null cancers, the p53-expressing cells in our study did not undergo senescence and maintained proliferative ability that was similar to the p53-null cells both in vitro and in vivo (Figure 1). Recent studies have attributed changes in cytokine profiles to oncogenic drivers, such as mutation in KRAS, MYC, and loss of PTEN. Our data suggest that one consequence of p53 loss is enhanced secretion of cytokines involved in myeloid recruitment and proliferation. Our work suggests the loss of p53 in tumors has a complex effect on cytokine secretion of both tumor and macrophage populations, ultimately affecting the T cell response. These observations complement recent work showing that an increase in WNT secretion in response to p53 loss in breast cancer cells can stimulate systemic inflammation and drive metastasis. The complexity of the immune response to p53 loss is further highlighted by a recent study showing that loss of p53 in a prostate cancer model increases expression of CXCL17 and CXCL5 and consequent attraction of tumor-promoting Gr-1+CD11b+ cells. Consistent with our study, loss of p53 in the prostate model also led to an increased expression of CXCL9 and CXCL10. Although we were unable to identify any changes in the intrinsic ability of p53-null cancer cells to present antigen, previous studies have shown p53-dependent alterations in antigen presentation, and it remains possible that cell-intrinsic changes also play a role in the immune response to p53 loss.

25Although blockade of the M-CSF receptor, CSF1R, produced a positive effect in allowing infiltration of T cell effector function in the p53-null tumors, this was not sufficient for rejection. This observation dovetails with a previous study targeting CSF1R with PD-L1 in an orthotopic pancreatic model (with mutations in KRAS and loss of INK) to increase the efficacy of PD-L1 blockade. Blocking CSF1R by chemical targeting in the p53R172H Kras pancreatic mouse model has shown a 2-week increase in survival and, as we also showed, an increase in T cell effector function. In addition, macrophage depletion with clodronate and CSF1R inhibition did not provide therapeutic benefit to primary tumors in the KPC pancreatic cancer model, similar to our results. Given the complexity of the immune modulatory response to loss of p53, it is not surprising that reversal of only one part of the response is insufficient to fully regain tumor rejection, since CD11b+ and Treg cell populations remained enriched in p53-null tumors. By targeting both suppressive populations through double blockade with anti-CSF1-R and anti-CD25, we were able to diminish the growth of p53-null tumors in an MHC-mismatched recipient. Indeed, our data show that the enhanced Treg cell infiltration into p53-null tumors is not prevented by CSF1R blockade, highlighting the importance of a combination immunotherapy.

26MSDCs and Treg cells are two major immune-suppressive populations hijacked by many cancer cells. MDSCs arise and expand under pathological conditions, especially during cancer. Naturally occurring Treg cells are a heterogeneous population, with different surface markers predicting suppressive capacity. In both our subcutaneous and orthotopic models, we detect an enrichment of KLRG1+ and CLTA4+ Treg cells in p53-null tumor-bearing mice. KLRG1+ Treg cells are considered a highly suppressive and terminally differentiated population of Treg cells that arise during tissue repair, such as influenza-induced lung injury, and help in tissue regeneration. Consistently, Treg cells isolated from recipients harboring p53-deficient tumors displayed increased suppressive capacity compared to their WT counterparts, dovetailing with a similar observation in PTEN- and p53-null prostate tumors.

27Pancreatic and lung tumors harboring mutations in KRAS also frequently acquire mutations in p53. While loss of p53 permits the proliferation of mutant-KRAS-expressing cells, which would otherwise undergo senescence, our data also demonstrate cooperation between KRAS and p53 mutations in allowing cancer cells to evade the immune response. Previous studies on pancreatic cancer have focused on the effects of mutant KRAS in pancreatic ductal epithelial cells (PDECs) and in combination with mutant p53 in PDAC models, where GM-CSF was a contributing factor for neutrophil recruitment to the tumor. Interestingly, while neutrophils were enriched in orthotopically implanted KRASG12D PDECs, there was no change in Treg cell infiltration, supporting our observation that p53 plays a more important role in modulating Treg cells in cancer. The selection for suppressive Treg cells seems to depend predominantly on p53 loss rather than the coordinated overexpression of mutant KRAS, as demonstrated by the iKRAS system and the ID8 ovarian cancer model. Moreover, a recent study analyzing lung cancer patients treated with PD-L1 blockade reported that patients harboring both KRAS and TP53 mutations were resistant to immunotherapy through a mechanism that was independent of PD-L1 expression levels. Hence, understanding how different oncogenic drivers interact to promote immune tolerance will be required to understand how best to apply immunomodulatory therapies in cancer.

28Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact Karen H Vousden (karen.vousden@crick.ac.uk). All unique/stable reagents generated in this study are available from the Lead Contact with a completed MTA.

29The mouse models of pancreatic ductal adenocarcinomas were generated using Pdx1-Cre; LSL-KrasG12D/+; Trp53+/+ and Pdx1-Cre; LSL-KrasG12D/+; Trp53fl/+ as previously published. The mouse model of non-small cell lung carcinomas was based on the tetracycline inducible EGFR-L858R [Tg(tet-O-EGFR∗L858R)56Hev] from the Mouse Repository of the National Cancer Institute. The R26tTA and Trp53fl/fl mice were obtained from the Jackson laboratory. Mice were crossed to generate Rosa26tTaLSL tet(O)EGFRL858R and Rosa26tTaLSL tet(O)EGFRL858R Trp53flox/flox mice and all backcrossed to C57Bl6/J background. Adenoviral Cre (Viral Vector Core, University of Iowa, USA) was delivered via intratracheal intubation (single dose, 2.5x107 virus particles in 50 μl). FVB (male and females) and CD1nu/nu (females) were purchased from Charles River, and OTI and OTII mice were purchased from Jax and maintained at the Beatson Institute for Cancer Research and the Francis Crick Institute animal facilities. All animals used ranged from 10-30 weeks of age and littermates of the same sex were randomly assigned to experimental groups.

30Animal experiments were subject to ethical review by the Francis Crick Animal Welfare and Ethical Review Body and regulation by the UK Home Office project license P8AA77917 and P319AE968 or at the BICR reviewed and approved by the University of Glasgow and UK Home Office for the project license (70/8645). All mice were housed under conditions in line with the Home Office guidelines (UK). Mice were housed from 3-5 per cage and were kept in a 12-hour day/night cycle starting at 7:00 until 19:00. Food and water were available ad libitum and rooms were kept at 21°C at 55% humidity. All procedures were performed following the Animals (scientific procedures) Act 1986 and the EU Directive 2010.

31Phoenix-ECO (ATCC® CRL-3214™) were purchased from the ATCC. PDAC cells from Pdx1-Cre; LSL-KrasG12D/+; Trp53+/+ and Pdx1-Cre; LSL-KrasG12D/+; Trp53fl/+ mice were derived as previously described. Cell lines used (KCs and KFCs) were tested for genetic background purity by the Charles River Genetic Testing Services. After a 384 SNP panel batch analysis (MB-160318AJ), cell lines derived from tumors were deemed mismatched (e.g., B6N 73%–78.9%, FVB 46.7%–59.8%, 129S4SvJae 60.3%–71.7%). ID8 isogenic ovarian cancer cells were previously described. iKRAS cell line was gifted by RA DePinho and were maintained in DMEM, 10% FBS and 2μg/ml of doxycycline. All other cell lines were maintained in DMEM, 10% FBS and penicillin-streptomycin and in 37°C, 5% CO2 humidified incubators.

32Primary T cells and bone marrow were derived from C57Bl6/J mice (either male or female) held at the BICR or Francis Crick institute and aged between 10-30 weeks.

33The pancreatic orthotopic model protocol was described by. Briefly, pancreatic cancer cells were surgically implanted into the pancreas of recipient mice (500,000 cells in 25μl of Matrigel, BD-Biosciences). For the ovarian ID8 cancer model, 5x106 cells were injected intraperitoneally (IP) in 6-8-week-old female C57Bl6/J mice (Charles River Laboratories, UK). The development of ascites and other symptoms were diagnosed as previously described.

34PDAC derived cells expressing iRFP were subcutaneously injected into the left flank of FVB mice at 1x106 cells/mouse. Growth was monitored by in vivo imaging and mice were taken at humane endpoint as dictated by the UK Home Office and the animal license. Allograft growth in CD1nu/nu mice was performed by unilateral flank injections of 1x106 per mouse. Growth was measured by in vivo imaging and humane endpoints were respected. All in vivo antibodies were purchased from Bio-X-Cell. Anti-CSF1R (clone AFS98) was used at a concentration of 300μg/mouse and administered twice a week. Anti-CD25 (PC-61.5.3) was used three days prior to tumor challenge and used at 400μg/mouse. Isotype control used was IgG2a (C1.18.4) and 200μg/mouse was used as a control.

35CD8+ T cells and CD4+ T cells were isolated from spleens and peripheral lymph nodes, prepared into single cell suspensions and lysed for red blood cells (10x RBC lysis buffer, Biolegend). Negative isolation kits and positive isolation kits (CD11b+ cells) were purchased from StemCell Technologies and isolation was performed following manufacture’s procedures. T cells were activated and cultured as previously described using plate-bound anti-CD3 and anti-CD28 antibodies.

36Isolation of intratumoral CD11b+ cells was performed on tumors digested using the tumor preparation protocol (see below). CD11b+ cells were isolated from tumor single cell suspensions using positive isolation kits from StemCell Technologies, following the manufacturer’s protocol.

37Macrophages were differentiated from bone marrow flushed using PBS, a 1mL syringe and a 25G needle. Bone marrow was collected, lysed for red blood cells (10x RBC lysis buffer, Biolegend), and plated on non-TC treated plates at 5x106 cells/10cm dish and 20ng/ml of M-CSF (Peprotech).

38Cell culture of pancreatic ductal adenocarcinoma cells (PDACs) and ID8 cell lines were maintained in DMEM supplemented with 10% FBS and Pen/Strep.

39CD4+CD25+ and CD4+CD25- T cells were isolated by performing CD4+ T cell negative isolation by STEMCELL technology kits and sorted by flow cytometry (FACS Aria Sorter) (CD4-FITC and CD25-APC). Briefly, ratios of 1:1 (Treg:Teffector) to 1:32 were generated in the presence of irradiated splenocytes and activated with 1μg/ml of αCD3 for 3 days.

40CD11b+ cells were isolated by positive selection from digested tumors (see above). T cells were stained with V450 dye and plate-bound activated with a-CD3 (5μg/ml) and a-CD28 (2μg/ml) for 24 hours. After 24 hours, T cells were co-cultured with isolated tumor-associated CD11b+ cells at a ratio of 1:4 in a 24 well plate for a further 2 days and proliferation was assessed using a the eBioscience live/dead fixable viability dye, APC-Cy7, and flow cytometry. Isolated CD11b+ intratumoral cells were plated 8 hours prior to co-culture at 0.5x106 cells per 24 well plate (Corning). Plate-bound activated T cells were plated 24 hours post-activation at 0.75x106 cells per 24 well plate.

41Single cell suspensions were stained for surface markers in PBS for 20 minutes at 4°C. Intracellular proteins (i.e., cytokines, FOXP3, and Ki67) were assessed using the FOXP3/Transcription staining buffer set (eBioscience, San Diego, CA) and following manufacturer’s instructions. Cells were permeabilized for 30 minutes and stained for intracellular proteins for 1 hour at 4°C. All fluorochromes were purchased from Biolegened and eBioscienes. Ex vivo re-stimulation was performed using PMA (Sigma-Aldrich), ionomycin (Sigma-Aldrich), and Golgi Stop (BD Biosciences) for 4 hours as previously described. For OTI and OTII peptide re-stimulation, peptides OVA257-264 and OVA323-339, were incubated with transgenic T cells at 10μg/ml with Golgi Stop (BD Biosciences) for 6 hours followed by surface staining and ICS. Dead cells were distinguished using the fixable viability dye efluor780® from eBioscience. Single cell suspensions were fixed and permeabilized using the FOXP3 Transcription staining buffer set. Samples were acquired on the BD LSRFortessa™ and on the BD FACSymphony™. Flow cytometry data was analyzed using FlowJo (TreeStar).

42ELISA kits were purchased from R&D systems for M-CSF, CXCL1, CCL11, and G-CSF. ELISAs for MCP1, CCL3 and CXCL5 were purchased from Life Technologies, Invitrogen. All ELISAs and cytokine arrays were performed on conditioned media (cells were plated at 1x106 cells/10cm dish) harvested after 24 hours. Cytokine arrays were purchased from R&D systems ® (Mouse cytokine panel array A) and performed on conditioned media from cells prepared as for ELISAs. Cytokines were detected following the manufacture’s procedures and using chemiluminescence (read at 450nm). Pixels from cytokine array data were analyzed using ImageJ software. The Luminex cytokine array used in this study was the Invitrogen eBioscience ProcartaPlex Mouse Cytokine & Chemokine Panel 1 (26 plex) with 4 additional cytokines

43Cells were lysed using RIPA buffer supplemented the 1%SDS and phosphatase inhibitors (La Roche Ltd), denatured at 95°C, and resolved on NuPAGE polyacrylamide pre-cast gels (ThermoFischer Scientific). Transfer of gels onto nitrocellulose membranes was performed using the iBlot2 (Invitrogen). Cells were probed for p53 with monoclonal anti-mouse p53 antibody from Cell Signaling Technologies (clone 1C12). Vinculin (H10, Santa Cruz Biotechnologies) was detected as a loading control. Secondary antibodies were purchased from LiCOR IRDye 800CW and 700CW.

44Mice were anesthetized with isoflurane and imaged using the Pearl Imager by LiCOR. iRFP fluorescence was excited using the 685nm laser and emission was detected in the 700nm channel (730nm). Fluorescence was analyzed using the Image Studio v5 from LiCOR.

45Primary T cells were activated with plate-bound anti-CD3 (5μg/ml) and anti-CD28 (2μg/ml) for 24 hours. Activated T cells were then co-cultured with adherent PDAC cells in 24 well plates or 96 well plates. PDACs were plated at 20 000 cells/well with 1x106 T cells (24 well plate) or 10,000 PDAC cells/well with 200,000 T cells (96 well plate) and co-incubated for 2 days. T cells were re-stimulated with PMA, ionomycin and GolgiStop (BD Biosciences) for 4 hours prior to ICS. Intracellular cytokines probed by ICS and flow cytometry were Granzyme B, IFNγ and TNFα.

460.3x106 BMDMs pulsed with 10μg/ml of ovalbumin (Sigma-Aldrich) for 1 hour, washed and naive OTI or OTII cells were added to the culture at 1x106 cells/well. T cells were kept in co-culture for 3-4 days followed by OVA257-264 and OVA323-339 re-stimulation at 10μg/ml in the presence of a Golgi blocker (GolgiStop, BD Bioscience), surface staining, ICS and acquired by flow cytometry.

47Splenocytes from OT1 TCR transgenic mice were stimulated in vitro with 10μg/ml of SIINFEKL (OVA257-264) for 48 hours. Cells were spun down using Lympholyte M (Cedarlane) following the manufacturer’s instructions. Activated CD8+OT1 cells were co-cultured in a 24 well plate (Corning) at 0.5x106 cells to 200,000 PDAC-mOVA isogenic cells. Cells were stained with propidium iodide for cell viability and acquired on the BDLSRFortessa™.

48Tissue was collected and fixed in 10% neutral buffered formalin (NBF). Samples were replaced with 70% ethanol after 48hours, embedded in paraffin blocks, and processed by standard histological techniques. Sections were cut at 5μM. IHC and H&E was performed as previously published. F480 antibody was purchased from eBioscience (rat anti-mouse F4/80 Cat. No 14-4801-82; RRID:AB_467558). Scoring of tumor sections for individual markers was performed by counting 30 fields using QuPath open source digital software and set as an average per field.

49Tumors were carefully excised from the animals and kept in ice cold medium till processing. Tumors were minced into small (about 1mm) pieces and digested in digestion buffer containing: collagenase 0.012%, dispase, 0.1mg/ml DNase I, 1% FBS in Krebs Ringer Bicarbonate Buffer (KRB) for 45-60 min at 37°C with gentle oscillation. The digestion was stopped by the addition of at least 10 volume of ice cold DMEM supplemented with 10% FBS. The solution was filtered through a 100 μm cell strainer and the isolated cells were precipitated by centrifugation at 300 X g for 5 min, washed with PBS before further processing.

50Serum was collected by cardiac puncture into EDTA coated tubes. Blood was spun in 1.5mL Eppendorf tubes at 2000xg for 15 minutes at 4°C. Serum was collected in upper phase and stored at −80°C.

51Total RNA was extracted using RNeasy® columns (QIAGEN) from at least 3 technical replicates per sample according to manufacturer instructions. Genomic DNA was removed using on column DNA digestion (QIAGEN). cDNA was generated using the High-Capacity cDNA reverse transcription kit (Thermofisher) according to manufacturer’s instructions. Power up ™ SYBR® Green Master MIX (Applied Biosystem) was used to perform QPCR with the following primers:

52Gene expression values were calculated according to Pfaffl method and expressed as relative units compared to the control group

53BMDMs were differentiated with 10ng/ml M-CSF and re-plated onto Imagelock 96 well plates. Scratches were made using the Essenbio scratch making tool. Wound closure was monitored in real-time using IncuCyte™ Live-Cell Analysis System and analyzed using IncuCyte S3 software.

54BMDMs at 5000 cells per chamber were plated in the upper chamber of an IncuCyte™ ClearView 96-well Cell Migration plate (Essen BioScience). Conditioned medium from different PDAC cell lines was placed in the bottom chamber. The Incucyte ZOOM® live-cell imaging system was used to measure cell migration.

55Plasmids for iRFP were transfected into Phoenix-ECO (ATCC® CRL-3214) using GeneJuice® Transfection reagents (Merk Milipore). PDAC cell lines were transduced with viral media and pBABE-iRFP selected for by puromycin (Sigma-Aldrich) at a concentration of 2.5μg/ml. Plasmid positive cells were cultured in selection media for 7 days. CRISPR/CAS9 methods were employed for genetic deletion of the Trp53 gene as previously described. Retroviral introduction of ovalbumin sequence for membrane bound ovalbumin (mOVA) was cloned out of the pCl-neo-mOVA plasmid (Addgene No 25099) and into the MSCV-based MIGR1 (Addgene 27490) vector. Phoenix-EC (ATCC ® CRL-3214) were used for virus production and PDAC cells were virally transduced three consecutive days. mOVA expressing cells were selected by cell sorting for GFP expression.

56Survival curves & gene expression: SEM normalized data was downloaded from the TCGA firehose website [ https://gdac.broadinstitute.org/]. Top (high group) and bottom (low group) quartiles pancreatic adenocarcinoma samples for MDSC-signature ranked expression were compared using the survival package version 2.41-3 (https://cran.r-project.org/web/packages/survival/index.html) in R version 3.4.3 (http://www.R-project.org) and the level of statistical significance determined by the log rank test. Expression data from PAAD patients with a mutation in TP53 was compared to those without a mutation using DESEQ2. Boxplots show the log2 expression values of selected genes grouped by their TP53 mutation status, with the adjusted p value determined by the Wald test.

57Data are presented as mean ± SD for technical replicates, or mean ± SEM for biological replicates. Data was analyzed using the unpaired Student t test when comparing two conditions. One-way ANOVA with a Tukey’s multiple comparisons test was performed on comparisons of more than two conditions as well as in vivo growth studies. Two-way Anova was performed on the in vivo tumor growth studies with treatments and different p53 alterations. Statistical significance is indicated in all figures by the following annotations: ∗ p < 0.05; ∗∗p < 0.01;∗∗∗p < 0.005. GraphPad Prism 7 was used for statistical analysis and graph generation.

58This study did not generate any unique datasets or code