Most human cancers have strong correlations between IRF2 and MHC I pathway gene expression
A number of types of human cancers express significantly lower levels of IRF2 transcripts compared to their normal counterparts [6] and similarly almost all categories of human cancers have a subset of cases with low IRF2 levels (Supp. Fig.1). To investigate whether this variation in IRF2 expression potentially had functional consequences, we analyzed the TCGA RNAseq database to determine whether cancer cases that downregulated IRF2 transcripts levels had corresponding reductions in the expression of MHC I pathway genes (β2M, ERAP1/2, HLA-ABC, PDIA3, PSMB8-10 PSME1, TAP1/2, TAPBP and TABPL), as well as a few other IRF2-regulated genes (CASP7, CD274 and GSDMD). Remarkably, there were significant (p<0.05) positive correlations between the levels of IRF2 expression and that of virtually all of the MHC I pathway components (Fig. 1). In other words, cases with low IRF2 transcripts had correspondingly low levels of immunoproteasome subunits, TAP, ERAP1 and other pathway gene transcripts. Since this transcript data came from RNAseq databases, we didn’t have access to the primary samples to further correlate these transcript levels with corresponding protein levels. However, where it has been examined, loss of IRF2 expression from gene knock out consistently reduced MHC I protein expression and conversely IRF2 transfection increased MHC I protein expression (see below and [6]). In other words, IRF2 function may be responsible for limiting for MHC I expression.
A Spearman correlations between transcripts of IRF2 and those of MHC I pathway genes (β2M, Erap1/2, HLA-A, HLA-B and HLA-C, PDIA3, PSMB9-10, PSME1, TAP1/2, TAPBP and TAPBPL) in tumor tissue from patients with the indicated cancer types (TCGA abbreviations). B-D Gene expression correlation between IRF2 vs ERAP1, PSME1, PSMB9, and Casp7 transcripts in primary and/or metastatic melanomas. E Spearman correlations between IRF2 transcripts and IRF2-regulated genes (Casp7, CD274 and GSDMD) and housekeeping genes (CALR, GAPDH, PSMB-7) in tumor tissue from patients with the indicated cancer types (TCGA abbreviations). A-E Data comes from the TCGA RNAseq database and was analyzed with TIMER [35].
A similar positive correlation was observed for IRF2 and Casp7 transcripts (Fig. 1D and E). In contrast, there was no positive correlation between IRF2 expression and that of the housekeeping genes GAPDH and PSMB7-9 (the active site subunits of constitutive proteasomes). Calreticulin, which plays a broad role as a chaperone for many ER proteins including MHC I, was only positively correlated with IRF2 expression in some cancer types. The correlation between IRF2 and Gasdermin D (GSMD) expression was weak or absent in a number of cancers. Different from what was seen in mouse dendritic cells [6], CD274 transcripts were not negatively correlated with IRF2 mRNA (Fig. 1E) but rather had a positive correlation in some cancers. Seeing that IRF2 expression was significantly correlated with expression of all MHC I pathway gene transcripts led us to examine whether these pathway genes had IRF1/2 binding sites. For this purpose, we examined publicly available ChIPseq data and found that all of the correlated MHC I pathway genes, with the exception of PDIA3 (ERP57), as well as CD274 and CASP7, had IRF2 and/or IRF1 bound to their 5’ region of these genes (Supp. Table-1A&B).
From these analyses, we observed that melanomas were one of the human cancers that had a subset of cases that expressed low levels of IRF2 (Supp. Fig.1) and had strong correlations between IRF2 expression and the MHC I pathway genes (Fig. 1B and C). Since melanomas are one of the cancers that can be immunogenic and a target of CD8 T cells, we further analyzed the functional consequences of IRF2 loss in these cancers.
Loss of IRF2 reduces the expression of MHC I pathway components in human and mouse melanomas
We examined a human melanoma patient-derived xenograft (Fig. 2A & B). These primary cells expressed detectable MHC I molecules by immunofluorescence and flow cytometry. These cells were transduced with a lentiviral vector containing Cas9 without (EV) or with IRF2-targeting guides to create IRF2-sufficient and IRF2-deficienct cells that were otherwise isogenic. Cells were then injected in highly immunodeficient NOD scid gamma (NSG) mice to expand these cells. Tumors were collected once they were palpable, enzymatically digested to create single cell suspension and then analyzed for surface MHC I molecules by immunofluorescence and flow cytometry. As show in Fig. 2B, loss of IRF2 dropped MHC I and PDL1 levels substantially (Supp. Fig. 4A). When the expression of MHC I pathway genes was analyzed by qPCR, we found reductions in transcripts of TAP1, TAP2, and ERAP1 genes (other MHC I pathway genes were not examined), which was consistent with our earlier studies with other cancers (Fig. 2C) [6]. The effect of loss of IRF2 dropped the mRNA expression levels of Cas7 and GSDM as well.
Loss of IRF2 reduces the expression of MHC I pathway components in a primary human melanoma. A Diagram of gene editing of a human patient melanoma (AV17) from passage in NSG mice. After editing, NSG mice were injected s.c. with WT and IRF2KO tumors, and once the tumors were palpable, they were harvested and analyzed for: B the expression of MHC I and PDL1 molecules on the tumors was analyzed by flow cytometer (C) mRNA expression of the MHC I pathway components was analyzed by qPCR. Each dot represents a biological replicate. Statistical analysis was calculated by GraphPad Prism, **P < 0.01, ***P < 0.001
To generalize these results, and to do so in a murine system whose immunobiology could be explored in vivo, we performed similar experiments in the B16F0 mouse melanoma cell line. B16FO cells were from a melanoma that arose spontaneously in C57BL6 mice and from which variants arise [15, 16]. For example, the frequently used B16F10 derivative was obtained through 10 serial passages in immunocompetent mice, and during this process lost MHC I expression [17], presumably due to immune-selection. For our studies of the MHC I pathway, we sorted B16F0 cells for uniform high levels of cell surface MHC I molecules (Fig. 3A). These cells were then transduced with a lentiviral vector containing Cas9 without (EV) or with mouse IRF2-targeting guides to create isogenic IRF2-sufficient and IRF2-deficienct cells (Fig. 3A). Similar to primary melanoma cells, loss of IRF2 dropped the surface expression of MHC I and significantly reduced transcripts of TAP2, ERAP1, and PSME1 detected in B16F0 mouse melanoma cells (Fig. 3B).
Loss of IRF2 in the mouse melanoma cell line B16F0 reduces the expression of MHC I pathway components but has no effect on B16F0 tumor growth kinetics in NSG or C57BL/6 mice. A Diagram of the experimental setup and B In vitro mRNA expression levels of MHC I pathway components in B16F0 WT (n=4) vs IRF2KO (n=3) were analyzed by qPCR. C Diagram of experiments testing the in vivo growth of WT vs IRF2KO B16F0 cells in NSG mice (n=10) and C57BL/6 mice (n=10) D & E Tumors from C were collected on day 15 and MHC I expression was analyzed by flow cytometry (D) and mRNA expression of the MHC I pathway components was analyzed by qPCR (E). Each dot represents a biological replicate and the curves on F show mean +SD. Statistical analysis was calculated by GraphPad Prism, **P < 0.01, ***P < 0.001
The effect of IRF2 on B16 melanoma growth in immunocompetent vs immunodeficient mice
When the isogenic pair of B16 melanomas was injected into highly immunodeficient NSG mice and immunocompetent WT mice, they both grew and did so with the same kinetics (Fig. 3C and F and Supp. Fig.2A and 2B). To determine whether the phenotype of the transplanted tumors changed in vivo, we harvested these cells and analyzed them by flow cytometry and qPCR. After growth in the NSG and WT mice, IRF2-deficient cells still had significantly lower levels of MHC I than the IRF2-sufficient tumors (Fig. 3D, Supp. Fig.4B). Interestingly, when analyzed by qPCR, the expression of the MHC I pathway components, while decreased, were not reduced as much as in the original cells (Fig. 3E).
To summarize, IRF2 loss doesn’t affect growth of the mice melanoma cells B16F0. Finding that WT and IRF2-deficient B16 cells grew rapidly in immunocompetent mice set the stage for the next experiments.
Effect of IRF2 loss on response to CPI therapy
Although B16 cells grow aggressively in WT mice, their growth can be slowed in mice treated with CPIs, such as anti-PD1 [18]. Therefore, the immune system still has the potential to detect and slow the growth of this cancer. This situation is similar to what occurs in many melanoma patients that are treated with CPIs. This allowed us to investigate whether loss of IRF2 would lead to resistance to CPI therapy (Fig. 4A).
IRF2-deficient human and mouse melanomas are resistant to CPI therapy. A Diagram of experiments testing the WT (n=14) vs IRF2KO (n=12) B16F0 in vivo tumor growth in C57BL/6 mice after isotype control or αPD1 treatment. B Tumor growth was recorded until the end of the experiment. C & D Tumors were collected on day 17 and MHC I expression was analyzed by flow cytometry (C) and mRNA expression of MHC I pathway components were analyzed using qPCR method (D). E Another group of C57BL/6 mice (n=58) were subcutaneously injected with either WT (n=29) or IRF2KO (n=29) B16F0 cells and tumor growth was recorded for survival analysis. F Diagram of experiments testing the WT (n=14) vs IRF2KO (n=15) the A17 patient-derived human melanoma growth in NSG (n=6) and NSG with HuHSC (n=23) mice after isotype control or αPD1 treatment. G Tumor growth was recorded until the end of the experiment and (C) MHC I expression was analyzed by flow cytometry on day 55 and day 112. C, D & H Each dot represents a biological replicate and the curves on B and G show mean+SD. Statistical analysis was calculated by GraphPad Prism, **P < 0.05, **P < 0.01, ***P < 0.001
Injection of anti-PD1 antibody into mice transplanted with IRF2-sufficient B16, significantly slowed the growth of this tumor and extended survival significantly (Fig. 4B, Supp. Figure2B-2F). In contrast, IRF2-deficient B16 were resistant to anti-PD1 therapy and tumor growth and survival were unaffected relative to untreated IRF2-sufficient tumors (Fig. 4B and E). We again analyzed the harvested cells by flow cytometry and qPCR methods and found that IRF2-deficient cells still had significantly lower cell surface levels of MHC I and PDL1 molecules compared to the IRF2-sufficient tumors (Fig. 4C, Supp. Fig.4C). Similarly, the IRF2-null cells had reduced expression of mRNA for MHC I pathway components, e.g. TAP2 and PSME1 (Fig. 4D). Together these results indicate that loss of IRF2 allows the B16 cells to evade the host immune response leading to treatment failure.
Effect of IRF2 loss on growth and response to CPI therapy in a primary human melanoma
When the isogenic pair of human IRF2-positive and negative primary melanomas were injected into humanized NSG mice, the IRF2-deficient tumors grew more rapidly than their IRF2-positive counterparts (Fig. 4G). This same pattern was observed in NSG mice, indicating that this growth differential was not due to evasion from human anti-tumor adaptive immune responses. Presumably, the loss of IRF2 confers some growth advantage to this melanoma, which is different from what we observed in the mouse melanoma model.
The growth of the WT tumor was significantly slower in humanized NSG mice as compared to NSG animals (Fig. 4G). Therefore, the presence of the human hematopoietic system restrained the WT tumor growth. In contrast, the IRF2-deficient cells evaded this control and grew at the same rapid rate in both NSG and humanized NSG mice (Fig. 4G).
We next investigated how the presence versus absence of IRF2 influenced responses to CPI with anti-PD1. Injection of anti-PD1 antibody into mice transplanted with the IRF2-sufficient human primary melanoma significantly (p ≤0.5) slowed the growth of this tumor (by 1 week). In contrast, IRF2-deficient primary melanomas were resistant to anti-PD1 therapy and tumor growth and survival were unaffected relative to untreated IRF2-sufficient tumors (Fig. 4G). We again analyzed MHC I levels on the harvested tumor cells by flow cytometry. IRF2-deficient cells still had significantly lower cell surface levels of MHC I than the IRF2-sufficient tumors after ICI treatment (Fig. 4H). The MHC I levels in the IRF2-deficient group without ICI were not decreased (Fig. 4H), for reasons that were not clear and likely an outlier as the levels remained low in two other independent experiments (Supp. Fig.2A). In any case, the CPI results indicate that similar to mouse B16 cells, loss of IRF2 allows human primary melanomas to evade the host immune response leading to treatment failure.
Reversing the immune evasion and resistance to therapy from IRF2-loss with IFN/IFN-inducers
We have previously shown that in IRF2null dendritic cells, the immune evasion phenotype from IRF2-loss could be reversed by treating cells with IFNγ or IFNα [6]. This led us to investigate whether treatment with IFNs would be able to reverse the immune evasion phenotype and CPI resistance with IRF2-deficient B16 cells. To test this hypothesis, we treated the B16 EV vs IRF2 KO cells with IFNγ or IFNα in vitro for 24 hours and then analyzed their phenotype (Fig. 5A, Supp. Fig.3). When analyzed by FACS and qPCR, cell surface MHC I levels (Fig. 5A) and the expression of MHC I pathway component transcripts (Fig. 5B) increased in both WT and IRF2KO cells after the IFN treatments. With IRF2KO cells, IFNs increased MHC I levels above those found in the unstimulated control cells but somewhat below the levels in IFN-stimulated control cells (Fig. 5A and B, Supp. Fig.3).
Effect of IFNα on WT and IRF2 KO B16 melanomas. A MHC I levels on IRF2KO (n=3) and WT cells (n=3) after 24 hour stimulation with or without IFNα (histograms) and after withdrawal of IFN (line graph). B mRNA expression of MHC I pathway components in WT (n=2) and IRF2KO cells (n=2) after the IFN treatments. C Diagram of the experiment testing the effects of IFNα+poly(I:C) treatment ± anti-PD1 on WT vs IRF2KO B16 melanoma. D & E Day 15 post tumor injection, tumors were harvested and (D) mRNA expression of the MHC I pathway components was analyzed by qPCR and E the expression of MHC I molecules on the tumors was analyzed by flow cytometer. Each dot represents a biological replicate and the curves on A. and C. show mean+SD. F Another group of C57BL/6 mice (n=40) were injected with WT (n=20) and IRF2KO (n=20) tumors for survival and tumor growth was recorded for survival analysis. Statistical analysis was calculated by GraphPad Prism, **P < 0.01, ***P < 0.001
Given that IFNs could restore MHC I expression, we next investigated whether IFN would reverse the immune evasion phenotype and resistance to treatment with CPI in vivo (Fig. 5C). For this purpose, we injected control EV vs IRF2 KO B16 cells into mice that were treated ± the type I IFN-inducer poly(I:C) and ± anti-PD1. After harvesting the cells, qPCR analysis revealed that the expression of the MHC I pathway component transcripts increased in IRF2 KO tumors (Fig. 5D). We also analyzed the expression of MHC I molecules on the IRF2 KO tumors in the anti-PD1 + poly(I:C)-treated mice, and similarly found that the treatment had increased MHC I levels above those in control isotype-treated and aPD1-treated mice, although these were not quite as high as in the EV B16 in the treated mice (Fig. 5E, Supp. Fig.2). Importantly, the combination of anti-PD1 and poly(I:C) let to a significant prolongation in survival of mice bearing either the EV or IRF2KO melanomas (Fig. 5F). The therapeutic response seen with the IRF2KO tumor was even a bit better than with EV melanoma. Therefore, the treatment with the IFN-inducer completely reversed the resistance of the IRF2null cancer to CPI.
Role of IRF1 in reversing the immune evasion phenotype in IRF2-deficient B16 melanoma cells
The IRF1 and IRF2 transcription factors bind to the same promoter regulatory elements in MHC I pathway genes. IRF2 is constitutively expressed, and IFNs (both type I&II) can modestly increase IRF2 expression, typically by 2-4 fold after 24h treatment [19]. In contrast, IRF1 is often minimally expressed under basal conditions but is markedly induced (often by 4-60-fold) by IFN stimulation [19]. Consistent with this pattern, low levels of IRF1 are detected in B16 melanoma cells by western blot, but IRF1 expression is substantially increased in cells stimulated with IFNα (Fig. 6B, Supp. Fig.5A). To test whether this induction of IRF1 played a role in the IFN-induced reversal of the immune evasion phenotype in IRF2 null B16, we generated and examined IRF2+IRF1 double KO B16 cells (Fig. 6B). The IFN-induced increase in the expression of some MHC I pathway component transcripts that had been seen in IRF2 KO cells in vitro was attenuated in IRF2+IRF1 double KO cell (Fig. 6C). Similarly, the IFN-induced increase in surface MHC I levels in the IRF2KO cells was also attenuated when these cells additionally lacked IRF1 (Fig. 6D&E). These results indicated that IRF1 was participating in the restoration of MHC I expression.
Transcription factor IRF1 substitutes for the loss of IRF2. A Same experimental design as 5C and 5D except, IRF1 mRNA expression was analyzed by qPCR. B-E EV, IRF1KO, IRF2KO and DKO (IRF1+IRF2KO) B16F0 cells were stimulated with IFNα in vitro and analyzed by: B Western blot for IRF1 or ß-actin; C. qPCR expression for MHC I pathway components (n=2/group). D & E Surface MHC I levels after 0 or 100 ng/ml IFNα treatment for 24 h (n=2/group)
To further evaluate the role of IRF1 in vivo, we examined the therapeutic effect of anti-PD1+poly(I:C) treatment on IRF1+IRF2 double knockout melanoma cells (Fig. 7A-D). While a therapeutic effect of anti-PD1+poly(I:C) was still seen with these tumors, they were still significantly more resistant to CPI compared to the IRF2 KO cancer cells (Fig. 7E and F). Together, these results show that after being induced by IFN, IRF1 can partially substitute for the function of IRF2 in driving expression of the MHC I pathway.
B16 melanoma IRF1+IRF2 double KO cells show impaired responses to IFN inducer Poly(I:C) plus CPI. A NSG mice (n=10) and C57BL/6 mice (n=12) were subcutaneously injected with WT, IRF2KO and IRF1+IRF1KO cells and data display tumor growth in individual mice. B Tumors were collected on day 15 and the surface MHC I expression of C57BL/6 tumors was analyzed by flow cytometry. C C57BL/6 mice (n=45), that were treated with poly(I:C) and aPD1 were injected with IFNα treated (10ng/mL, 24h) control WT vs IRF2KO IRF2+IRF1 (double) KO B16 cells and tumor growth was followed. D & E On day 17 post tumor injection, n=5 mice/group were sacrificed for: D mRNA expression analysis of MHC I pathway components using qPCR method and E the tumor cell surface expression analysis of MHC I molecules by flow cytometer. Each dot represents the average measurement of the individual tumors collected from mice. F Survival analysis of WT or IRF2KO or IRF1+IRF2 KO tumors in C57BL/6 mice. This experiment was repeated twice. Tumor growth curve shows mean+SD (C)
Add Comment