chk-α and PD-L1 expression are interactively related α
To identify the interactive relationship between chk-α and PD-L1, we used siRNA to downregulate chk-α and PD-L1 in TNBC MDA-MB-231 and SUM-149 cells, and in Pa09C and Pa20C human PDAC cells (32). Untreated cells and cells treated with scrambled or luciferase siRNA were used as controls. We used a previously validated siRNA sequence (40) to downregulate chk-α, and two separate siRNA sequences, labeled PD-L1 #1 and PD-L1 #2, to downregulate PD-L1. Changes in mRNA levels of chk-α and PD-L1 in siRNA treated cells, compared to untreated cells, are shown in Figure 1. A significant reduction of chk-α and PD-L1 mRNA levels was detected following treatment with the target-specific siRNAs, given singly or combined, in the two TNBC cell lines (Fig. 1A,B) and in Pa09C cells (Fig. 1C). In Pa20C cells, downregulation of chk-α was less pronounced (<50%) and there was no decrease in PD-L1 mRNA levels with PD-L1 siRNA treatment (Fig. 1D). Importantly, we found an inverse correlation between chk-α and PD-L1 mRNA levels. In MDA-MB-231, SUM-149 and Pa09C cells, a greater than 80% increase in PD-L1 expression was observed following chk-α downregulation.
Conversely, in MDA-MB-231 and SUM-149 cells, a significant increase in chk-α expression of more than 50% was observed following PD-L1 downregulation (Fig. 1A, B). In the PDAC cells, only Pa09C cells showed a small but significant increase in chk-α mRNA expression with PD-L1 downregulation (Fig. 1C). Since PD-L1 was not downregulated in Pa20C cells, we did not detect an increase of chk-α mRNA in these cells. This inverse dependence was lost when cells were treated with a combination of chk-α and PD-L1 siRNA. To ensure that the loss of inverse dependence was not due to the lower siRNA concentration, we treated MDA-MB-231 cells with 50 nM chk-α or PD-L1#1 siRNA alone and observed similar changes as cells treated with 100 nM siRNA (data not shown). To further establish this inverse relationship, we analyzed the correlation between chk-α and PD-L1 mRNA expression levels in all four cell lines treated with single siRNA (Supplementary Fig. 1). We found a strong inverse correlation in MDA-MB-231 (P<0.0001, r=-0.763) and SUM-149 (P<0.0001, r=-0.738) TNBC cells. Pa09C PDAC cells also showed a significant inverse correlation, although weaker than that observed in the TNBC cell lines (P=0.001, r=-0.676). Taken together, these data clearly identified an interdependence between chk-α and PD-L1 at the genomic level. This interdependence was not observed when both genes were downregulated by siRNA.
Changes in chk-α and PD-L1 mRNA levels are reflected in protein changes
To determine whether changes in mRNA translated to changes in protein expression, protein levels of chk-α and PD-L1 were measured in MDA-MB-231 cells by immunobloting. As shown in Figure 2A, proteins were harvested at 24h, 48h and 72 h post transfection with chk-α and PD-L1 siRNA either alone or in combination. chk-α and PD-L1 siRNA treatment resulted in an effective decrease of the targeted proteins when used alone or in combination. Immunoblots showed increased PD-L1 levels at 72h following chk-α siRNA treatment, in good agreement with previous studies that reported a delayed increase in PD-L1 protein levels compared with mRNA levels (41). chk-α increased at 24h following PD-L1 siRNA treatment. This chk-α/PD-L1 interdependence was not observed when cells were treated with both chk-α and PD-L1 siRNA. Overall, the protein expression patterns were similar to the mRNA patterns.
To evaluate whether changes in PD-L1 mRNA and protein levels resulted in changes at the cell surface, we performed flow cytometry analysis of siRNA treated MDA-MB-231 cells (Fig. 2B and 2C and Supplementary Fig. 2). Treatment with PD-L1 #1 siRNA, either alone or in combination, was effective in reducing both the percentage of PD-L1 positive cells (Fig. 2B) and the Mean Fluorescence Intensity (MFI, Fig. 2C), which reflects the total amount of PD-L1 on the cell surface. Treatment with chk-α siRNA resulted in a small, but statistically significant increase in the percentage of PD-L1 positive cells when compared to cells treated with either scrambled or luciferase siRNA. This was also observed in MFI data when comparing chk-α siRNA treated cells with untreated or scrambled siRNA treated cells.
Taken together, these results confirmed that changes in mRNA resulted in changes in PD-L1 protein expression, and its localization on the cell surface. The magnitude of changes in PD-L1 cell surface expression was lower than anticipated based on the changes in mRNA and total level of proteins. This is likely due to the already high cell surface expression of PD-L1 in MDA-MB-231 cells, as reflected by the 99% of PD-L1 positive cells found in untreated cells, making any increase difficult to detect. However, a small but statistically significant increase was observed in the total amount of PD-L1 on the surface, as detected by MFI, in cells treated with chk-α siRNA.
Consequences of chk-α and PD-L1 downregulation on metabolites
To assess whether changes in chk-α or PD-L1 with siRNA treatment altered metabolites including choline containing compounds, we used high resolution 1H MRS to analyze the aqueous and lipid phases of MDA-MB-231 cell extracts. Significant differences in several metabolites were observed in MDA-MB-231 cells treated with PD-L1 or chk-α siRNA individually, as shown in the representative aqueous phase spectra in Figure 3A and the data summarized in Figure 3B and Supplementary Table 1. These metabolic changes were mostly eliminated when both targets were downregulated. Data from untreated and luciferase siRNA treated cells were combined into a single control group for clearer presentation, since the metabolic profiles from these cells were comparable. With PD-L1 downregulation, a significant increase of PC was observed, consistent with the increase of chk-α mRNA and protein found with PD-L1 downregulation. PD-L1 downregulation also resulted in a significant increase of glutamate, arginine, lactate, creatine, glutathione (GSH), oxidized glutathione (GSSG), and ATP. Treatment with chk-α siRNA resulted in a significant decrease of PC demonstrating the functional effects of chk-α downregulation that were in good agreement with previous results (42). chk-α downregulation resulted in a decrease of acetate, and a significant increase of glutamine, glutamate, aspartate, arginine, pyruvate, lactate, glycerophosphocholine (GPC), creatine, myo-inositol, taurine, GSH, GSSG, NADP, ATP, adenosine and S-methyl-5′-thioadenosine (MTA). Finally, when both chk-α and PD-L1 were downregulated together, most of these metabolic changes were not observed with the exception of changes in GSH and myo-inositol.
Consequences of chk-α and PD-L1 downregulation on lipids
High resolution 1H MRS of the lipid phase of MDA-MB-231 cell extracts detected significant changes in MRS detectable lipids following chk-α or PD-L1 downregulation. Representative lipid phase spectra are shown in Figure 4A with the changes in the lipid profile summarized in Figure 4B and Supplementary Table 2. Untreated and luciferase siRNA treated cells were combined into a single control group. chk-α downregulation resulted in a significant decrease of the total lipid content, as represented by the methylene signal of fatty acids (-CH3) and the methylene groups at the b position of the carboxylic function (OOC-CH2-CH2). We also detected significantly decreased levels of arachidonic acid (AA) and eicosapentaenoic acid (EPA), docosahexaenoic acid and linoleic acid. PD-L1 downregulation, consistent with the resultant increase of chk-α, increased PtdCho, phosphatidylethanolamine (PtdEA), and the total level of lipids, as represented by the methylene groups at the a position of the carboxylic function (OOC-CH2). PD-L1 downregulation also caused an increase in the total level of unsaturated lipids, as represented by the fatty acid double bond signal (CH=CH), the methylene groups at the a position of a double bond (CH=CH-CH2) and diallylic methylene protons (CH=CH-CH2-CH=CH)n. We also detected significant increases in linoleic acid, glycerol, sphingomyelin, and docosahexaenoic acid. Changes in lipids induced with PD-L1 downregulation were mainly eliminated when both chk-α and PD-L1 were downregulated, with the exception of sphingomyelin, linoleic acid and unsaturated lipids as represented by CH=CH, and CH=CH-CH2, suggesting that these changes were mediated through chk-α.
Inflammation and the chk-α/PD-L1 interdependence
To further understand the mechanisms underlying the chk-α/PD-L1 interdependence, we evaluated the role of inflammation, COX-2, in this relationship. COX-2 and its metabolite prostaglandin E2 (PGE2) play roles in inflammation, cancer development and adaptation to changing microenvironments (43). More recently, COX-2 and PGE2 have been implicated in cancer immunosuppression (44), and COX-2 and PD-L1 expressions were found to be correlated in melanomas (45) and lung adenocarcinomas (46). chk-α downregulation increased COX-2 mRNA expression by approximately 4-fold in MDA-MB-231 cells compared to untreated cells, but not compared to cells treated with scrambled siRNA (Fig. 5A). On the other hand, PD-L1 downregulation resulted in an over 40-fold increase of COX-2 mRNA, alone or in combination with chk-α downregulation. Consistent with the relatively small increase of COX-2 with chk-α downregulation, COX-2 protein levels (Fig. 5B), and PGE2 concentrations (Fig. 5C) did not increase with chk-α siRNA treatment, whereas the 40-fold increase of COX-2 mRNA following treatment with PD-L1 siRNA resulted in an increase of COX-2 protein and PGE2 concentrations in the cell culture media. To further understand the role of COX-2 in the PD-L1/chk-α dependence, we performed chk-α and PD-L1 downregulation studies in MDA-MB-231 cells with COX-2 silenced using COX-2 short hairpin (sh)RNA (shCOX2-MDA-MB-231). We found that although chk-α and PD-L1 siRNA downregulated the target genes, the increase of PD-L1 with chk-α downregulation, and the increase of chk-α with PD-L1 downregulation was eliminated in cells with COX-2 silenced, at the mRNA (Fig. 5D) and protein levels (Fig. 5E). These data indicate that COX-2 is required for the PD-L1 and chk-α interdependence.
TGF-b, an inducer of COX-2 (47), has been directly associated with suppression of the host antitumor immune response (48) and resistance to immune therapies by increasing tumor cell plasticity (49). We therefore analyzed changes in TGF-b expression in response to chk-α and PD-L1 downregulation. As shown in Figure 6, chk-α downregulation significantly decreased TGF-b mRNA expression in TNBC MDA-MB-231, shCOX-2-MDA-MB-231, and SUM-149 cells. In PDAC cells, Pa09C showed a less pronounced decrease in TGF-b, while Pa20C showed the smallest decrease. These alterations matched the range of changes in the increase of PD-L1 observed with chk-α downregulation, where TNBC cells showed the largest increase of PD-L1, followed by Pa09C cells, and Pa20C cells showed no increase. Conversely, downregulating PD-L1 resulted in an increase of TGF-b levels, that was most pronounced in TNBC, lesser in Pa09C cells, and none in Pa20C cells, matching the levels of PD-L1 downregulation and the corresponding increase of chk-α. The increase of TGF-b in shCOX-2-MDA-MB-231 cells treated with PD-L1 siRNA was clearly attenuated compared to wild type cells. When both chk-α and PD-L1 were downregulated, no consistent pattern was observed.
chk-α and PD-L1 interdependence confirmed in human cancers
To independently confirm the inverse correlation between chk-α and PD-L1 in human cancers, we analyzed the relationship between the tumoral expression of chk-α and PD-L1 in TCGA. We calculated mean chk-α and PD-L1 expression in primary tumor samples from 31 different cancer types (Supplementary Table 3), comprising of more than 9,000 samples. We found an inverse linear correlation between PD-L1 and chk-α mRNA levels (Fig. 7, P= 0.001, r=-0.562). When we analyzed individual tumor values, irrespective of the tumor type (Supplementary Fig. 3), we also found a significant inverse correlation (P< 0.001, r=-0.358). When we examined this relationship within individual tumor types, we found a statistically significant inverse correlation (p<0.001) for 21 out of the 31 tumor types analyzed (supplementary Table 4). When categorizing receptor status in the breast cancer group, we found that TNBC showed the most significant inverse correlation (P=0.003, r=-0.31), whereas the correlation was weaker or not statistically significant in the other breast cancers.
We further analyzed the TCGA samples by ranging primary tumors according to the mRNA levels of chk-α and PD-L1 independently of the tumor type. Tumors with 10% lowest mRNA levels for chk-α had significantly higher levels of PD-L1 (Supplementary Fig. 4A) compared with tumors with 10% highest mRNA levels for chk-α. Similarly, tumors with 10% lowest mRNA levels for PD-L1 had significantly higher levels of chk-α (Supplementary Fig. 4B) compared with tumors with 10% highest mRNA levels for PD-L1. These results are in good agreement with our experimental data showing an increase in PD-L1 with chk-α downregulation and an increase in chk-α with PD-L1 downregulation, further confirming this relationship in human cancers.