Oncogenic RAS Signaling Promotes Tumor Immunoresistance by Stabilizing PD-L1 mRNA

SUMMARY
The immunosuppressive protein PD-L1 is upregu- lated in many cancers and contributes to evasion of the host immune system. The relative importance of the tumor microenvironment and cancer cell-intrinsic signaling in the regulation of PD-L1 expression re- mains unclear. We report that oncogenic RAS signaling can upregulate tumor cell PD-L1 expression through a mechanism involving increases in PD-L1 mRNA stability via modulation of the AU-rich element-binding protein tristetraprolin (TTP). TTP negatively regulates PD-L1 expression through AU-rich elements in the 30 UTR of PD-L1 mRNA. MEK signaling downstream of RAS leads to phos- phorylation and inhibition of TTP by the kinase MK2. In human lung and colorectal tumors, RAS pathway activation is associated with elevated PD-L1 expres- sion. In vivo, restoration of TTP expression enhances anti-tumor immunity dependent on degradation of PD-L1 mRNA. We demonstrate that RAS can drive cell-intrinsic PD-L1 expression, thus presenting therapeutic opportunities to reverse the innately im- munoresistant phenotype of RAS mutant cancers.

INTRODUCTION
Therapeutic antibodies blocking the coinhibitory PD-1 pathway by targeting PD-L1 (programmed death 1 ligand 1, also known as B7-H1 or CD274) or its receptor, PD-1, have caused striking regressions in several malignancies in which RAS mutations are frequent driver events, including non-small cell lung cancer (NSCLC) (Herbst et al., 2014; Topalian et al., 2012) and mismatch-repair-deficient colorectal cancer (Le et al., 2015). PD-L1 is critical for limiting autoimmune-related damage to normal tissues in the context of chronic inflammation but is also aberrantly upregulated on cancer cells in order to evade im- mune destruction (Pardoll, 2012). As anti-PD-1 pathway immu- notherapies are effective in only a minority of cancer patients (Topalian et al., 2012), there is a great need for reliable bio- markers of patient response. To what degree tumor PD-L1 expression is prognostic of patient response to PD-1 pathway blockade remains contentious. Recent clinical trials of the anti- PD-1 antibody nivolumab report that tumor cell PD-L1 expres- sion correlates with response to nivolumab in non-squamous but not the squamous subtype of NSCLC (Borghaei et al., 2015; Brahmer et al., 2015). Notably, non-squamous NSCLC patients with KRAS mutations benefited from nivolumab therapy in terms of overall survival, whereas KRAS wild-type patients did not (Borghaei et al., 2015). Response rate and progression- free survival was increased in NSCLC patients treated with pembrolizumab in cases where at least 50% of tumor cells were positive for PD-L1 (Garon et al., 2015). In this patient cohort, KRAS mutant tumors were more frequently PD-L1 posi- tive than KRAS wild-type tumors.The success of immune-checkpoint blockade is dependent on the immunogenicity of the tumor (Gubin et al., 2014; Linnemann et al., 2015; Rizvi et al., 2015), so one possible confounding fac- tor in the use of tumor PD-L1 as a biomarker for response is the uncoupling of tumor PD-L1 expression from tumor immunoge- nicity.

It is therefore critical to understand the signaling pathways that dictate tumor cell PD-L1 expression. The inflammatory cyto- kine IFN-g is the best-characterized stimulus for PD-L1 expres- sion, but several studies suggest that cell-intrinsic oncogenic signaling can also promote PD-L1 expression in cancer cells through epidermal growth factor receptor (EGFR), the transcrip- tion factor MYC, and the kinase AKT (Akbay et al., 2013; Casey et al., 2016; Parsa et al., 2007). Studies performed on melanoma (Jiang et al., 2013) and acute myeloid leukemia (Berthon et al., 2010) have indicated that MEK signaling is involved in upregula- tion of PD-L1 in some tumor cell lines, but the molecular basis of this regulation remains poorly defined.Separately, genetic rearrangements in the 30 UTR of CD274 (encoding PD-L1) have been found in a multitude of different cancers at low frequency and are associated with massivelyImmunity 47, 1–17, December 19, 2017 ª 2017 The Author(s). Published by Elsevier Inc. 1This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).increased expression of tumor PD-L1 (Kataoka et al., 2016). These results imply that control of PD-L1 expression through the CD274 30 UTR might contribute to immune escape in human cancers, although the underlying mechanisms of post-transcrip- tional regulation responsible for this effect are unclear.In this report, we reveal that tumor cell PD-L1 expression can be driven by oncogenic RAS pathway activation by a mechanism involving post-transcriptional regulation of the stability of PD-L1 mRNA. This provides a direct mechanism whereby RAS signaling in tumor cells can provide protection from attack by the immune system.

RESULTS
We tested the potential role of oncogenic RAS signaling in the regulation of PD-L1 expression in human epithelial cells using ER-RASG12V fusion constructs, which allow for the induction of oncogenic RAS activity with 4-hydroxytamoxifen (4-OHT) (Molina-Arcas et al., 2013). As expected, addition of 4-OHT led to the rapid activation of oncogenic KRAS signaling through MEK and PI3K (Figure 1A) and coincided with induction of MYC mRNA and CCND1 mRNA (encoding cyclin D1) in an immortalized human pneumocyte cell line derived from type II cells (Figure 1B; Kemp et al., 2008). PD-L1 mRNA was rapidly increased following stimulation of oncogenic KRAS signaling with 4-OHT, resulting in a 6-fold induction of mRNA expression after 3 hr (Figure 1B). By way of comparison with known regula- tors, stimulation with IFN-g led to increases in PD-L1 mRNA in excess of 10-fold after 3 hr and both KRAS activation and IFN-g stimulation dramatically increased PD-L1 protein expres- sion at the cell surface after 48 hr (Figure 1C). Oncogenic HRAS signaling was also capable of inducing PD-L1 mRNA and protein expression in the immortalized breast epithelial cell line MCF10A and the KRAS wild-type colon carcinoma cell line HKE-3 (Figures S1A and S1B), implying that induction of PD-L1 expression by RAS is not a tissue-specific or RAS-iso- form-specific phenomenon. The induction of PD-L1 protein was most striking in ER-HRASG12V MCF10A cells, perhaps re- flecting the low basal expression of PD-L1. Chronic RAS activa- tion for 4 days led to more profound increases in PD-L1 protein, whereas shorter-term activation resulted in modest inductions of PD-L1 expression (Figure S1B). Importantly, 4-OHT did not induce PD-L1 expression in parental cell lines lacking ER-RAS constructs (Figure S1C).Direct inhibition of KRAS signaling with the KRASG12C-specific inhibitor ARS853 (Lito et al., 2016; Patricelli et al., 2016) in lung and colorectal cancer cell lines harboring KRASG12C mutations led to reductions in PD-L1 mRNA expression, but not in the KRASG12S A549 control lung cancer cell line (Figures 1D and 1E). Moreover, ARS853 treatment led to significant reduc- tions in PD-L1 surface protein expression in the KRAS mutant lung cancer cell line H358 (Figure 1F). To dissect which down- stream effectors of RAS are responsible for regulating PD-L1 expression, we used the specific inhibitors of MEK and pan type I PI3Ks, GSK1120212 (trametinib) and GDC-0941 (pictilisib), respectively (Figure S1D).

Notably, MEK and PI3K inhibitors could block RAS-induced expression of PD-L1 protein in ER-KRASG12V type II pneumocytes, either alone or in combina- tion (Figure 1G). MEK inhibition significantly reversed KRAS- mediated PD-L1 mRNA upregulation (Figure 1H), but PI3K inhibi- tion only reduced PD-L1 protein expression, concordant with evidence for AKT signaling increasing PD-L1 expression predominantly through activating translation of the transcript (Parsa et al., 2007). MEK inhibition, but not PI3K inhibition, reduced PD-L1 mRNA expression in H358 (Figure 1I), H23, and H1792 lung cancer cell lines (Figure S1E). Downstream of MEK, inhibition of ERK1/2 with SCH772984 potently reduced PD-L1 expression in H358 and H23 cells (Figure S1F). Further- more, PMA, a potent chemical activator of MEK-ERK signaling via protein kinase C stimulation, markedly and rapidly increased PD-L1 expression, an effect that was largely reversed with the inhibition of MEK (Figures 1J and S1G). More extensive analysis of PD-L1 surface expression on multiple KRAS mutant lung cancer cell lines, both human and murine, revealed generally consistent PD-L1 downregulation after MEK and PI3K inhibition, suggesting that this regulatory pathway is of broad significance(Figure S1H). Taken together, these results suggest that onco- genic RAS signaling through MEK and PI3K is sufficient to drive PD-L1 expression.Since RAS signaling has been implicated in reducing the expression of genes involved in the presentation of antigens by MHC class I molecules (Ebert et al., 2016; El-Jawhari et al., 2014), we analyzed the expression of antigen processing and antigen presentation machinery following oncogenic RAS acti- vation (Figure S1I). As expected, KRAS G12V signaling led to significant decreases in expression of TAP1, TAPBP, as well as HLA-A, HLA-B, HLA-C, and B2M, suggesting that compro- mised antigen processing and presentation in concert with in- creases in PD-L1 expression may contribute to an augmented state of immunoresistance in RAS mutant tumor cells.To investigate how RAS-MEK signaling regulates PD-L1 expres- sion, we first asked whether RAS regulates PD-L1 via a tran- scriptional mechanism. We generated a series of luciferase reporter constructs containing promoter fragments cloned from the human CD274 locus (Figure S2A).

In all cases, the phys- iological stimulus IFN-g, but not PMA, induced expression of the promoter reporter constructs in H358 cells, a cell line in which endogenous PD-L1 mRNA expression is robustly induced with PMA (Figure 1J). Incorporation of putative enhancer elements (Sumimoto et al., 2016) into the CD274 promoter reporter con- structs also failed to confer sensitivity to MAPK activation (Fig- ure S2A), as did including predicted regulatory regions spanning the 50 of exon 1 (data not shown). Furthermore, none of the re- porters showed evidence of decreased expression when H358 cells were treated with MEK inhibitor (data not shown).Therefore, we investigated possible mechanisms of post-tran- scriptional regulation of PD-L1 expression by RAS. We induced oncogenic KRAS signaling with 4-OHT in ER-KRASG12V type II pneumocytes and concomitantly blocked transcription with actinomycin D. Surprisingly, we found human PD-L1 mRNA to have a short half-life, which was significantly stabilized by the in- duction of oncogenic KRAS signaling (Figure 2A). Moreover, murine PD-L1 mRNA also had a comparably short half-life, and the stability of the transcript in a Kras mutant, p53-deleted mu- rine lung tumor cell line (KPB6), could be reduced further still when MEK was inhibited (Figure 2B), implicating KRAS-MEK signaling in the stabilization of the labile PD-L1 transcript. Consistently, direct inhibition of oncogenic KRAS signalingwith ARS853 also caused reductions in PD-L1 mRNA half-life in H23, H1792, and H358 cells (Figure 2C). However, inhibition of PI3K alone did not result in altered PD-L1 mRNA stability in KPB6 cells (Figure S2B).Common genetic elements conferring mRNA instability include miRNA binding sites and AU-rich elements (AREs) in the 30 UTR of the transcript. The core motif for AREs is an ATTTA pentamer sequence, but functional AREs are often found in an AU-rich context, conforming to the WWATTTAWW nonamer consensus (where W denotes an A or T) (Zubiaga et al., 1995) constituting the binding site for several AU-rich element binding proteins (AUBPs), which can subsequently recruit mRNA decay machinery (Lykke-Andersen and Wagner, 2005). For example, a canonical ARE-regulated transcript is TNF, which contains nine pentamer sequences in the human transcript and eight pentam- ers in the murine transcript. Upon inspection of the 30 UTR of PD-L1 mRNA, we noted a high number of ARE pentamers.

Specifically, out of 14 ATTTA pentamer sequences in the human transcript and 11 in the murine transcript, there were 3 conserved AREs conforming to the nonamer consensus (Figure 2D).We tested the influence of MEK inhibition on the half-life of another unstable transcript, Tusc2 mRNA (tumor suppressor candidate 2, or Fus1), which does not contain AU-rich elements in the 30 UTR but is targeted by multiple miRNAs (Du et al., 2009). Although Tusc2 mRNA had a similar half-life to PD-L1 mRNA, MEK inhibition did not influence the stability of the Tusc2 transcript (Figure S2C), indicating that the observed post-tran- scriptional regulation of PD-L1 by MEK may relate to AU-rich elements in the 30 UTR. Indeed, a transcript containing functional AU-rich elements, Ptgs2 mRNA (Cha et al., 2011), displayed a significant reduction in mRNA half-life in response to MEK inhibi- tion (Figure S2C), reminiscent of PD-L1 mRNA.To directly analyze the functional importance of these AREs, we constructed a luciferase reporter containing a fragment of the 30 UTR of human CD274 containing the last six ATTTA pen- tamers, including the three conserved nonamer sequences. Mutation of ATTTA pentamers to ATGTA has been shown to increase the expression of ARE-containing mRNAs (Rajagopalan et al., 1995; Yang et al., 2004). Consistent with this, mutating the six ATTTA pentamer sequences to ATGTA increased expression of the PD-L1 30 UTR luciferase reporter in ER-HRASG12V MCF10A and H358 cells, suggesting that these AREs are functionally relevant for controlling the expression of PD-L1 (Figures 2E and 2F). Stimulation with 4-OHT inER-HRASG12V MCF10A cells, or PMA in H358 cells, increased expression of the wild-type reporter, whereas the ATGTA mutant reporter was insensitive to these treatments (Figures 2E and 2F). In sum, these data suggest that AREs in the 30 UTR of PD-L1 mRNA can mediate control of PD-L1 expression by RAS-MEK signaling.To assess which AU-rich element binding proteins (AUBPs) could mediate regulation of PD-L1 expression downstream of RAS signaling, we performed a selected siRNA screen of likely candidate genes, AUF1, KSRP, HuR, and TTP (also known as tristetraprolin or ZFP36), in three RAS mutant lung cancer cell lines (Figures 3A–3C).

Knockdown efficiency was verified in each case by qPCR (Figures S3A–S3C). siRNA-mediated knock- down of KSRP and TTP most consistently increased PD-L1 mRNA expression across the cell line panel, with the exception of A427, where knock-down of TTP did not lead to significant increases in PD-L1 mRNA levels. Overexpression of KSRP or TTP was sufficient to significantly decrease PD-L1 expression (Figure 3D) and PD-L1 30 UTR luciferase reporter expression in H358 cells (Figure 3E), corroborating our results from the siRNA screen and confirming that KSRP and TTP impart their negative regulation of PD-L1 expression through the 30 UTR. Overexpres- sion of TTP and KSRP together did not result in additive reduc- tions in PD-L1 expression, suggesting that they may regulate PD-L1 through the same mechanism (Figure S3D). Notably, siRNA-mediated knockdown of TTP family members, BRF-1 and BRF-2, was incapable of increasing PD-L1 expression to the extent achieved by silencing TTP expression (Figures S3E and S3F). We confirmed that TTP protein expression was reduced following knock-down in H23 and H358 cells, but this was less clear in A427 cells, which express lower levels of TTP protein (Figure S3G). Deconvolution of siRNA pools targeting TTP showed that multiple siRNAs increased expression of PD-L1 mRNA in H23 and H358 cells (Figure S3H).We further examined the regulation of PD-L1 mRNA by TTP by using TTP wild-type (WT) and TTP knock-out (KO) MEFs. In the TTP KO MEFs, TTP mRNA is expressed but no functional TTP protein can be made due to the introduction of a premature stop codon at the endogenous locus (Lai et al., 2006; Taylor et al., 1996). Acute activation of TTP expression with serum temporally coincided with a substantial and transient decrease in PD-L1 mRNA in TTP WT MEFs, but not in the TTP KO MEFs(Figure 3F), with PD-L1 levels recovering to near baseline at 6 hr after serum addition. Moreover, the total absence of func- tional TTP protein in the TTP KO MEFs increased the half-life of PD-L1 mRNA relative to TTP WT MEFs (Figure 3G).Finally, we generated a KPB6 lung cancer cell line with a tetra- cycline-inducible TTP transgene (TTP tet-ON). As expected, inducible expression of TTP led to reductions in wild-type PD-L1 30 UTR luciferase reporter expression, but not of the ATGTA mutant 30 UTR reporter (Figure 3H). When combined with MEK inhibition, TTP expression more robustly suppressed expression of the wild-type reporter. In sum, these data provide evidence for the negative regulation of PD-L1 mRNA expression by the AUBPs KSRP and TTP.RAS Regulates PD-L1 Expression through TTP To further investigate whether MEK and TTP regulate PD-L1 via a shared pathway, we silenced TTP expression using siRNAs in the context of MEK inhibition. Knock-down of TTP was largely able to rescue the decrease in PD-L1 expression caused by MEK inhibition (Figure 4A).

However, the knockdown of KSRP could not rescue this phenotype, despite profound silencing of expression (Figure S4A). Furthermore, MEK inhibition signifi- cantly increased TTP mRNA expression (Figure 4A), and chronic activation of oncogenic KRAS signaling significantly decreased TTP mRNA expression (Figure 4B).Next, we tested whether the RAS pathway regulates the activ- ity of TTP and/or KSRP protein. Crucially, we found that endog- enous levels of TTP and KSRP both co-precipitated with PD-L1 mRNA in RNA immunoprecipitation (RNA-IP) reactions from KPB6 mouse lung cancer cells (Figure S4B). TTP also signifi- cantly bound to PD-L1 mRNA in H358 cells (Figure 4C). In all cases, the enrichment for the PD-L1 transcript was far greater than that of a control mRNA, GAPDH, which lacks AREs in the 30 UTR (Figures 4C and S4C). MEK inhibition did not significantly alter the occupancy of TTP or KSRP on PD-L1 mRNA, consistent with RAS regulating the activity of the AUBP, rather than the occupancy on the target mRNA.ERK has been shown to phosphorylate (Taylor et al., 1995) and negatively regulate TTP activity and expression (Bourcier et al., 2011; Deleault et al., 2008; Essafi-Benkhadir et al., 2007; Ha€rdleet al., 2015). Inhibition of MEK decreased phosphorylation of TTPat PXSP (ERK target-site consensus) and RXXS/T (RSK/AKT target-site consensus) motifs (Figures 4D and 4E), confirming that TTP is regulated by phosphorylation downstream of MEK signaling in cancer cells. Mutation of two of the highestFconfidence predicted ERK-target residues on human TTP (S218 and S228) abrogated detection of TTP with the phospho-PXSP motif-specific antibody (Figure 4D), but the phosphosite mutant TTP (S218A 228A) did not show enhanced activity in reducing PD-L1 mRNA expression compared to wild-type TTP (data not shown), implying the involvement of other residues that are not readily detected with this antibody. Furthermore, although AKT signaling has been shown to regulate KSRP activity through phosphorylation of S193 (Dı´az-Moreno et al., 2009), the KSRP S193A phosphosite mutant did not show enhanced activity in reducing PD-L1 mRNA expression compared to wild-type KSRP (Figure S4D).

Equally, the serine/threonine phosphatase PP2A has been implicated in positively regulating TTP function by reversing inhibitory phosphorylation events (Sun et al., 2007). Therefore, we tested whether inhibition of PP2A with okadaic acid (OA) would increase PD-L1 expression. OA rapidly increased PD-L1 mRNA expression in TTP WT MEFs, but not TTP KO MEFs (Figure 4F), demonstrating that PP2A activity decreases PD-L1 expression specifically through modulating TTP activity. To discover which residues are functionally important for regu- lating TTP activity downstream of RAS, we performed mass spectrometry on immunoprecipitated Myc-TTP after PMA, MEK inhibitor, or PMA and MEK inhibitor treatment. We used the Kras mutant, mouse colon carcinoma cell line CT26, based on its immunogenicity and sensitivity to anti-PD-L1 antibody therapy, making it suitable for downstream in vivo experiments. Most notably, mass spectrometry analysis revealed MEK- dependent phosphorylation of S52 and S178; PMA significantly enhanced phosphorylation of these residues, and this effect was reversed with MEK inhibition (Figures 5A and S5A and Table S1). Moreover, MEK inhibition alone was sufficient to reduce phos- phorylation of these residues (Figure 5A).S52 and S178 residues are crucial for the regulation of TTP activity through binding to 14-3-3 proteins following phosphory- lation by MK2 (also known as MAPKAPK2) downstream of p38 (Chrestensen et al., 2004). Consequently, p38 signaling results in decreased TTP activity, partly through reducing the associa- tion with deadenylase machinery (Mahtani et al., 2001; Stoecklin et al., 2004). In parallel, phosphorylation of S52 and S178 stabi-lizes TTP protein (Brook et al., 2006), which is consistent with the observed increase in abundance of total TTP peptides detected in the PMA versus the MEK inhibitor-treated condition (Figure 5A).We reasoned that oncogenic RAS might stimulate p38 signaling through promoting the MEK-dependent accumulation of reactive oxygen species (ROS) (Nicke et al., 2005) and thus inhibit TTP function. Indeed, oncogenic RAS signaling dramati- cally increased intracellular ROS in MCF10A cells, and ROS levels were distinctly correlated with the extent of PD-L1 induc- tion (Figure S5B). Furthermore, the addition of the potent anti- oxidant N-acetyl-L-cysteine (NAC) largely reversed the induction of PD-L1 protein by RAS (Figures 5B and S5B), collectively sug- gesting that ROS induction by oncogenic RAS is functionally important in driving PD-L1 expression.

Specific activation of the p38 pathway using an inducible version of the upstream kinase MEKK3 (DMEKK3-ER) (Figures 5C and S5C; Garner et al., 2002) was sufficient to increase PD-L1 protein expression, albeit to a lesser extent than that achieved by RAS itself. Co-treatment with NAC was consider- ably less effective in reversing PD-L1 induction in this context, consistent with ROS operating upstream of p38 in this pathway (Figure 5C). Moreover, inhibition of MK2 strongly reversed RAS-induced PD-L1 expression in MCF10A and HKE-3 cells (Figure 5D) and PD-L1 expression in CT26 cells, which have endogenous levels of mutant KRAS (Figure 5E). We also observed reductions in expression of PD-L1 mRNA in several NSCLC cell lines with endogenous KRAS mutations following treatment with NAC, reduced glutathione, or MK2 inhibitor III (Figure S5D), although we noted some heterogeneity in response between the four cell lines tested.To directly test the functional significance of the MK2 target residues downstream of MEK pathway activation, we generated TTP knock-out CT26 cell lines using CRISPR/Cas (to obviate functional contributions from endogenous TTP) and reconsti- tuted these cells with either a wild-type (WT) or phosphosite mutant (S52A S178A), tetracycline-inducible TTP transgene. S52 and S178 of mouse TTP are highly conserved, with S52 conforming to the RXXS/T phosphosite motif (Figure 5F). Immu- noprecipitation of Myc-tagged TTP following acute MAPK acti- vation with PMA revealed phosphorylation of WT TTP, but not of the S52A S178A mutant protein at RXXS/T sites (Figure 5G),verifying our findings from mass spectrometry analysis. Crucially, the S52A S178A mutant TTP had significantly enhanced activity in reducing PD-L1 mRNA expression relative to WT TTP (Figures 5H and S5E). In sum, these results suggest that a RAS-ROS-p38 signaling axis contributes to PD-L1 upre- gulation through phosphorylation and inactivation of TTP.To further evaluate the role of oncogenic RAS signaling in regu- lating PD-L1 expression in cancer, we analyzed TCGA gene expression data from patient-derived lung adenocarcinoma (LUAD) or colon adenocarcinoma (COAD) samples.

To account for the effects of alternative oncogenes that can activate down- stream RAS effector pathways such as EGFR, BRAF, and ALK, we used two published gene expression signatures for RAS activation (Loboda et al., 2010; Sweet-Cordero et al., 2005) to segregate patient samples into ‘‘high’’ and ‘‘low’’ RAS pathway activity based on gene expression. As expected, annotation of KRAS mutation status revealed a strong enrichment for KRAS mutant samples in the high RAS activity cohorts in both signa- tures (Figures 6A and S6A). We compared the expression of T cell function-related genes between high and low RAS activity cohorts and found CD274 (encoding PD-L1) expression to be significantly increased in the high RAS pathway activity samples in LUAD (1.42 log2-fold change) and COAD (1.17 log2-fold change) samples, using either signature (Figures 6A, 6B, and S6A). Stromal PD-L1 and tumor PD-L1 expression appear to have independent, suppressive effects on anti-tumor immunity (Lau et al., 2017), but we noted that the expression of the pan- leukocyte marker PTPRC (coding for CD45) and lymphocyte marker CD3E were only modestly increased in the high RAS pathway activity cohort, indicating that the differential in PD-L1 expression is not likely to be solely attributable to a higher degree of leukocyte infiltration in the tumor microenvironment (Figure 6A).Of note, IFNGR1 was also among the most significantly enriched transcripts in the high RAS pathway activity groups. To investigate the possibility that PD-L1 may be upregulated in RAS active tumors due to regulation by IFNGR1, we induced PD-L1 expression with RAS in ER-KRASG12V type II pneumo- cytes and concomitantly blocked IFNGR1 signaling using a depleting antibody for IFN-g or with the JAK1/2 inhibitor ruxoliti- nib. Although both treatments effectively reduced responses to exogenous IFN-g, PD-L1 induction by RAS was unaffected, suggesting independence from IFN-g-IFNGR1 signaling (Figure 6SB).To further explore the in vivo relevance of TTP regulation in human cancer, we compared TTP mRNA expression in normal tissue and tumor samples by using publically available datasets. TTP mRNA was strikingly downregulated in human lung andcolon tumor samples compared to normal tissue (Figure S6C; Selamat et al., 2012; Skrzypczak et al., 2010), confirming that aberrant regulation of TTP expression is relevant in the human disease.

Consistently, in FACS-sorted epithelial cells isolated from normal lung or matched tumor tissue from KrasLSL-G12D/+; Trp53F/F (KP) mice, TTP mRNA expression was reduced in lung tumor tissue (Figure S6D). PD-L1 mRNA expression was generally higher in tumor tissue than in normal lung but not significantly increased; however, PD-L1 protein expression was significantly elevated, perhaps reflecting the contribution from AKT in promoting PD-L1 protein expression (Figure S6E).Next, we set out to directly assess the functional importance of the regulation of PD-L1 expression by TTP in tumor progression. To this end, we generated a series of stable CT26 cell lines ex- pressing Myc-tagged mouse TTP under a tetracycline-inducible promoter (TTP tet-ON), and in addition, constitutively expressing either empty vector or mouse Cd274 cDNA lacking the 30 UTR (PD-L1 D30 UTR). TTP expression was induced upon addition of doxycycline in a dose-dependent manner (Figure 7A), result- ing in decreased PD-L1 protein expression at the cell surface (Figure 7B). Overexpression of PD-L1 D30 UTR rendered total PD-L1 levels effectively insensitive to TTP induction (Figure 7B). TTP transgene expression with doxycycline was also associated with a decrease in PD-L1 mRNA stability, which was comparable to that mediated by MEK inhibition in this system (Figure S7A).To independently verify our findings in another cell line, we used MC38 tumor cells because they are known to exhibit sensitivity to PD-L1 modulation in vivo and show RAS pathway activation (Giannou et al., 2017). As expected, TTP was induced with doxycycline in MC38 (tet-ON) cells, leading to reductions in PD-L1 expression (Figures S7B and S7C).Using these engineered cell lines, we performed subcutane- ous transplantation experiments in mice and monitored tumor progression. Notably, the growth rates of the stable cell lines in vitro did not significantly differ with the overexpression of PD-L1 D30 UTR cDNA or the induction of TTP transgene expres- sion with doxycycline (Figure S7D and S7E).

However, in vivo, doxycycline treatment significantly reduced CT26 and MC38 tumor growth in immune-competent, syngeneic mice (Figures 7C and 7D). Strikingly, the anti-tumor effects mediated by doxy- cycline treatment were absent in immunocompromised nu/nu mice harboring CT26 tumors (Figure 7E) and in mice treated with depleting antibodies against CD8 and CD4, implying an essential contribution from the adaptive immune system to this anti-tumor response (Figure 7F). CT26 tumor cells overexpress- ing PD-L1 D30 UTR grew faster than the empty vector cells in BALB/c mice but had no growth advantage in nu/nu mice. Moreover, expression of PD-L1 D30 UTR was able to rescuemuch of the growth inhibition mediated by doxycycline treatment in BALB/c mice, suggesting that suppression of tumor cell PD-L1 expression is an essential component of the anti-tumor effects mediated by TTP transgene induction (Figure 7C). As expected, CT26 cells expressing a Cd274 cDNA with the full-length, wild- type 30 UTR (PD-L1 WT 30 UTR) had considerably lower expres- sion of PD-L1 protein than the PD-L1 D30 UTR cells, but still responded to TTP induction in terms of reductions in PD-L1 expression (Figure S7F) and control of tumor growth in im- mune-competent mice (Figure S7G).Consistent with a heightened anti-tumor immune response, tumors derived from mice treated with doxycycline had a greater degree of CD3+ lymphocyte infiltration than tumors from mice treated with vehicle, and this corresponding infiltration was abrogated in tumors derived from cells overexpressing PD-L1 D30 UTR (Figures 7G and S7H). Moreover, we found higher CD8+/Treg cell ratios in tumors expressing the TTP transgene and higher levels of IFN-g production by CD8+ tumor-infiltrating lymphocytes (TILs) derived from TTP-expressing tumors, versus PD-L1 D30 UTR tumors expressing TTP (Figure 7H); however, we did not find significant differences in CD4+ TIL populations (data not shown).Collectively, these data highlight the functional importance of the regulation of PD-L1 expression by TTP in tumor progression and demonstrate that this novel regulatory pathway may be exploited for the treatment of Ras mutant cancers. These find- ings support a model whereby
tumor-specific suppression of TTP can foster PD-L1 upregulation, and ultimately, tumor immu- noresistance (Figures 7I and S7I).

DISCUSSION
In this report, we demonstrate that oncogenic RAS signaling can increase tumor cell-intrinsic PD-L1 expression, implying that mutant RAS oncogenes can directly contribute to the evasion of immune destruction in cancer. We revealed that RAS-MEK signaling controlled expression of PD-L1, at least in part, by modulating the stability of the transcript. We showed that the mouse and human PD-L1 mRNAs were labile tran- scripts containing functional AU-rich elements (AREs) in the 30 UTR that permitted regulation of PD-L1 expression by RAS. Our data provide a potential explanation for the genomic structural variations in the CD274 30 UTR observed in human cancer (Kataoka et al., 2016). The simultaneous loss of regulation by miRNAs and AREs is likely to contribute to the high overexpres- sion observed in tumors with complete loss of the 30 UTR. In addition, we provide a molecular basis for the tendency of KRAS mutant NSCLCs to be positive for PD-L1 expression (D’Incecco et al., 2015; Dong et al., 2017; Li et al., 2017; Yang et al., 2017), implying that PD-1-PD-L1 blockade may prove more successful in RAS mutant patients that also harbor a sufficient number of tumor antigens.
We identify TTP as a principle AU-rich element binding protein responsible for negatively regulating PD-L1 expression, consis- tent with a previous report identifying PD-L1 mRNA as one of a number of TTP targets in an RNA immunoprecipitation, microar- ray-based screen in mouse macrophages (Stoecklin et al., 2008). Mechanistically, MEK inhibition reduced PD-L1 mRNA stability, coinciding with an increase in TTP expression and reduction in phosphorylation of TTP at ERK and RSK/AKT consensus motifs. Conversely, activation of RAS and the associ- ated ROS accumulation led to enhanced TTP phosphorylation, notably by MK2 at key inhibitory sites.

TTP transgene expression restrained tumor growth in CT26 and MC38 tumor transplantation models. This anti-tumor effect is predominantly non-cell autonomous, dependent on the adap- tive immune system and suppression of tumor cell PD-L1 expression. We noted only minor reductions in tumor growth rates following TTP transgene induction in cells overexpressing PD-L1 D30 UTR. TTP has been reported to have cell-autonomous tumor-suppressive roles (Rounbehler et al., 2012) and non-cell- autonomous anti-tumor effects through targeting VEGF and COX-2 mRNAs (Cha et al., 2011; Essafi-Benkhadir et al., 2007), which may contribute to some of these ostensibly PD-L1-inde- pendent effects, the magnitude of which are likely to be deter- mined by the level of TTP overexpression in each system.Our data extend the molecular understanding of the regulation of PD-L1 expression in cancer and highlight druggable targets to enhance anti-tumor immunity in tumors that are wild-type for the CD274 30 UTR. We provide evidence that pharmacological targeting of RAS, or RAS effector proteins, may elicit non-cell- autonomous anti-tumor effects in RAS mutant tumors. Recently, MEK inhibitors and PD-1 pathway blockade were shown to combine strongly in a mouse model of Ras mutant colon carcinoma (Ebert et al., 2016; Liu et al., 2015). We anticipate that our findings will inform the development of effective combination Zunsemetinib therapies with immune checkpoint blockade in cancer.