CircRUNX1 Drives Immune Evasion and anti-PD1 Immunotherapy Resistance in Lung Adenocarcinoma by the miR-4739/PCSK9/MHC I Axis

Background: Lung adenocarcinoma (LUAD) is the commonest pathological subtype of lung-derived malignant tumor. Blocking the immune checkpoints interaction between programmed death-1 (PD-1) and programmed death ligand 1 (PD-L1) via naturalizing PD-1 or PD-L1 monoclonal antibody is proved to be a effective therapeutic method for drive gene negative LUAD. Dysregulation of circular RNA (circRNA) is a hallmark of cancer and has a critical role in the progression of cancers, including LUAD. However, the exact role of circular RUNX family transcription factor 1 (circRUNX1, hsa_circ_0002360) in LUAD remains unknown. Methods: The expression levels of circRUNX1, miR-4739, and proprotein convertase subtilisin/kexin type 9 (PCSK9) mRNA were examined by quantitative real-time polymerase chain reaction (qRT-PCR). PCSK9, histocompatibility complex I (MHC I), CD8, and Foxp3 protein expression was examined by immunohistochemistry (IHC) or western blot analysis. In vivo assay was carried out to determine the biological function of circRUNX1 on LUAD immune evasion in vivo. The binding relationship between miR-4739 and circRUNX1 or PCSK9 was predicted and veried by Starbase V3.0 online databases, dual-luciferase reporter assay, and RNA-pull down assay. Results: CircRUNX1 expression was frequently increased in LUAD tissues, and forced circRUNX1 expression was associated with poor prognosis, big tumor size, advanced stage, and resistance to anti-PD1 immunotherapy in LUAD patients. Functionally, circRUNX1 overexpression promoted LUAD immune evasion in vivo. In addition, our results demonstrated that circRUNX1 could sponge miR-4739, subsequently, upregulating the PCSK9 expression and promoting to LUAD immune evasion. Conclusion: circRUNX1 promoted cancer immune evasion and resistance to anti-PD1 immunotherapy by regulate miR-4739/PCSK9/MHC circRUNX1 is overexpressed in LUAD and its upregulation is correlated with poor prognosis. Mechanism, circRUNX1 promotes LUAD immune evasion by acting as a sponge for miR-4739, which decreases the activity of PCSK9/MHC I axis. These results demonstrate that circRUNX1 is a promising novel biomarker and therapeutic target in LUAD. or mutant pLG3-circRUNX1 in the T cells after cotransfection with miR-4739 mimic or negative control. g The relative level of circRUNX1 was detected by qRT-PCR in PC-9 and HCC-827 cell lines transfected with miR-4739 mimic or negative control. h The relative level of circRUNX1 was detected by qRT-PCR in NCI-H1975 and Clau-3 cell lines transfected with miR-4739 siRNA or negative control. i The relative level of miR-4739 was detected by qRT-PCR in NCI-H1975 and Clau-3 cell lines transfected with circRUNX1 or mock control. i The relative level of miR-4739 was detected by qRT-PCR in PC-9 and HCC-827 cell lines transfected with circRUNX1 shRNA or negative control. Data are presented as the means ± SD of three independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001. NS, no signicant.

PD1 immunotherapy in LUAD patients. Functionally, circRUNX1 overexpression promoted LUAD immune evasion in vivo. In addition, our results demonstrated that circRUNX1 could sponge miR-4739, subsequently, upregulating the PCSK9 expression and promoting to LUAD immune evasion.
Conclusion: circRUNX1 promoted cancer immune evasion and resistance to anti-PD1 immunotherapy by regulate miR-4739/PCSK9/MHC I axis in LUAD. Taken together, our ndings indicate that circRUNX1 may be a promising therapeutic target for improve the anti-PD1 therapy effect in LUAD patients.

Background
Co-operation of innate and adaptive immune systems is indispensable for cancer immune surveillance to eradicate cancer cells [1]. The emergence of anti-PD-1 immunotherapy for the treatment of non-small cell lung cancer (NSCLC) has effectively improved the prognosis of the patients [2]. However, anti-PD-1-based immunotherapy only bene ts 30-40% of the NSCLC patients, and many of the patients will generate acquired resistance during the process of the therapy [3,4]. Lung adenocarcinoma (LUAD) is the most histopathological subtype of NSCLC [5]. Despite the advanced multidisciplinary treatment, the overall survival rate of LUAD patients remains frustrating. During the process of tumor progression, cancer cells have evolved complicated mechanisms to decrease major histocompatibility complex I (MHC I) expression and resistance to cytotoxic T lymphocytes-mediated immune attack [6]. Decreased MHC I expression has been con rmed in a variety of cancers, including LUAD [7,8]. Furthermore, decreased MHC I expression has been identi ed as a critical mechanism of resistance to anti-PD-1-based immunotherapy [9]. Nonetheless, the effect of decreased MHC I expression on the sensitivity of LUAD to anti-PD-1 immunotherapy has not been explored.
With the rapid development of high-throughput sequencing technology, circular RNA (circRNA) as a class of novel non-coding RNA with an important role in multiple pathophysiological processes, therefore, have been gradually become a hotspot area attracting of investigator's attention. Recently, it has been con rmed that dysregulation of circRNAs play a critical regulatory role in cancer progression and immune evasion [10][11][12]. For example, circFGFR1 promotes the expression of C-X-C motif chemokine receptor 4 (CXCR4) by sponging miR-381-3p and participates in the invasion, migration, and immune evasion of NSCLC [10]. To date, a large number of circRNAs have been veri ed to exist in body uids, including exosomes [13]. In recent study, we have been reported that cancer cell-derived exosomal circUHRF1 upregulates the expression of hepatitis A virus cellular receptor 2 (TIM3) by sponging miR-449c-5p and contributes to immunosuppression of NK cells [12]. However, research on circRNA-induced immune evasion in lung adenocarcinoma (LUAD) has been relatively scarce despite mounting evidence indicating that they have critical roles in malignancy.
Recently, studies have reported that proprotein convertase subtilisin/kexin type 9 (PCSK9) expression is dysregulated in multiple cancers and participates in promoting cancer progression [14]. Importantly, PCSK9 can promote MHC I degradation in lysosome [15]. Above research results indicate that inhibiting PCSK9 might be a promising method to enhance treatment e ciency for cancers immunotherapy. Here, we reported a circRNA derived from the RUNX1 gene, and named circRUNX1 (hsa_circ_0002360).
CircRUNX1 was veri ed to be signi cantly upregulated in LUAD. Furthermore, we con rmed that circRUNX1 plays a critical role in the immune evasion of LUAD by sponging miR-4739 to in uence expression of MHC I molecular via PCSK9-dependent manner. The underlying biological molecular mechanisms and functions of circRUNX1 were indicate that circRUNX1 may represent a promising biomarker and a potential therapeutic target for immunotherapy in LUAD. samples were analyzed by immunohistochemistry (IHC), quantitative real-time polymerase chain reaction (qRT-PCR), and enzyme linked immunosorbent assay (ELISA). The clinicopathological information was collected from 1 January 2012 to 31 December 2019. The Ethics Committee of the Zhongshan Hospital of Fudan University provided ethical approval, and informed consent for collecting and preserving samples and documenting details was obtained from all of the patients.
IHC, qRT-PCR, western blotting, circRNA precipitation (circRIP), RNA immunoprecipitation (RIP), luciferase reporter assay, CD8 + T cell isolation, and uorescence in situ hybridization (FISH) assays IHC, qRT-PCR, western blot, In vivo circRNA precipitation (circRIP), RNA immunoprecipitation (RIP), luciferase reporter assay, CD8 + T cell isolation, and uorescence in situ hybridization (FISH) assays were carried out as described in our previous studies [10,12,16]. The qRT-PCR primers used in this study are listed in Supplementary  Establishment of stable PCSK9-knockout cells CRISPR/Cas9 double-vector lentivirus was used to establish stable PCSK9-knockout NCI-H1975 cells. The lentivirus was purchased from the Genomeditech Co. and transfected as described in reference [10]. Generation of humanized NSG (HuNSG) mice and xenograft models NSG mice aged 4-6 weeks were purchased from Shanghai Model Organisms. The animals were housed and treated in speci c pathogen-free conditions. The method performed to generate the huNSG mice were previously established and described. [17] The animal experiments were carried out in accordance with the guidelines approved by the Shanghai Medical Experimental Animal Care Commission. Mice xenograft anti-PD-1 therapy study were carried out as described in our previous studies [10]. In brief, the experiments in the huNSG mice were approved by the Animal Experimentation Ethics Committee of Zhongshan hospital, Fudan University. A total of 2 × 10 6 cells (LUAD cells with or without increased circRUNX1 expression) were implanted subcutaneously in the left ank of the huNSG mice. When tumors reached a volume approximately 100 mm 3 , the mice were injected in the tail vein with a PD-1 monoclonal IgG antibody (Nivolumab) or its isotype control at 100 µg per dose three times a week for two weeks.
Animals were euthanized when tumors reached a volume 1500 mm 3 (n = 6).

Statistical analysis
Statistical analysis was performed with SPSS software (19.0; SPSS, Inc., Chicago, IL) as in our previous study and described in supplementary methods [10]. The cutoff criterion for high and low expression of the circRUNX1 was the median value. P < 0.05 was considered statistically signi cant.

CircRUNX1 was upregulated in LUAD tissues
Recently, a circular RNA pro ling analysis has reported that circRUNX1 (hsa_circ_0002360) was upregulated in human LUAD tissues [18]. However, the biological functions of circRUNX1 on LUAD remains unclear. To further understand the clinical characteristics and signi cance of the circRUNX1 expression in LUAD, qRT-PCR was used to detect the expression level of circRUNX1 in 110 tumor tissues and paired adjacent non-tumor tissues of LUAD patients. Results indicate that the levels of circRUNX1 was signi cantly increased in tumor tissue compared with that in paired adjacent normal tissues (Fig.   1a). Next, we analyzed the relationship between circRUNX1 expression and the clinicopathological characteristics of 110 LUAD patients, as listed in Table 1. The results demonstrate that the expression levels of circRUNX1 in TNM stage III was signi cantly higher than that in TNM stage I-II (Fig. 1b).
Additionally, tumor size bigger than 5 cm had higher circRUNX1 expression than those tumor size smaller than 5 cm (Fig. 1c). Subsequently, we investigated the prognostic implications of circRUNX1 expression in LUAD patients. Our results demonstrated that patients with higher circRUNX1 expression had a signi cantly poor prognosis than those with low levels of circRUNX1 ( Fig. 1d and e). Multivariate analysis results indicated that circRUNX1 expression is an independent predictor for decreased overall survival (OS) rate and enhanced postoperative relapse rate (Tables 2 and 3). Taken together, our results indicate that circRUNX1 was signi cantly forced in LUAD tissues and correlated with bad prognosis of LUAD patients.   Previous studies have been reported that circRNAs dysregulation authorize tumor immune evasion [10,12], therefore, we wondered whether circRUNX1 overexpression might promote immunosuppressive microenvironment. Then, we explored whether forced circRUNX1 expression exhibited immune evasion signatures by qRT-PCR and IHC analysis in 30 LUAD patient tumor tissues. Interestingly, our results indicated that higher circRUNX1 expression was associated with decreased CD8 + T cells in ltration, while increased Foxp3 + cells in ltration (Fig. 2a). Scatter plot analysis demonstrated a negative correlation between circRUNX1 expression and CD8 + T cells in ltration in the LUAD tissues (Fig. 2b). Moreover, the results from scatter plot analysis also indicated a negative correlation between circRUNX1 expression and Foxp3 + T cells in ltration in the LUAD tissues (Fig. 2c). To determine whether increased circRUNX1 expression induces CD8 + T cell dysregulation, CD8 + T cells from 30 case LUAD patient tumor tissues were selected. Compared with CD8 + T cells in circRUNX1 Low tissues, CD8 + T cells in circRUNX1 High tissues demonstrated signi cant impairment of Granzyme B and perforin secretion when subsequently stimulate with a CD3 antibody ( Fig. 2d and e).
Higher level of circRUNX1 expression is correlated with resistance to anti-PD-1 LUAD therapy To determine the biological function of circRUNX1 in LUAD sensitivity to anti-PD1 therapy, we then explored retrospective data from 18 recurrent LUAD patients receiving anti-PD-1 therapy who had undergone primary tumor resection 2-24 months before the immunotherapy. Then, circRUNX1 expression in LUAD tissues were measured using qRT-PCR (Fig. 3a), and Kaplan-Meier survival analysis demonstrated that the progression-free survival for the circRUNX1-high group was much shorter than that for the circRUNX1-low group (Fig. 3b). To further determine the effects of circRUNX1 expression on the sensitivity of LUAD to anti-PD-1 immunotherapy, we analyzed circRUNX1 expression in ve human LUAD cell lines using qRT-PCR ( Supplementary Fig. 1a). To verify the biological functions of circRUNX1 in LUAD cells, we designed an overexpression lentivirus (circRUNX1-OE) transfected them into the NCI-H1975 and Clau-3 cell lines to upregulate circRUNX1 expression. The qRT-PCR results indicated that transfection of overexpressing circRUNX1 lentivirus vector resulted in signi cant forced circRUNX1 expression levels, while the expression levels of RUNX1 mRNA is not changed (Supplementary Fig. 1b and c). In addition, we also designed two shRNAs lentivirus vector targeting the back splicing site of circRUNX1 and transfected them into the PC-9 and HCC-827 cell lines to knockdown circRUNX1 ( Supplementary Fig. 1d and e). Then, we explored the anti-tumor effects of the PD-1 antibody in humanized NSG mice that received LUAD cells with forced circRUNX1 expression or the respective mock cells (NCI-H1975-Mock, NCI-H1975-circRUNX1). Compared to that of the NCI-H1975-Mock cell group, the tumor growth in the NCI-H1975-circRUNX1 cell recipient xenograft mice showed a signi cantly phenotype of resistance to anti-PD-1 immunotherapy, and the xenograft mice had a decreased survival time (Fig. 3c-e). To further explore the molecular mechanism of circRUNX1-mediated LUAD resistance to anti-PD-1 immunotherapy, we have examined the effect of circRUNX1 on the PD-L1 expression on LUAD cells. Our results showed that the expression of PD-L1 was not changed signi cantly when circRUNX1 was overexpressed or knockdown in the LUAD cell lines ( Fig. 3f and g). Above results indicate that circRUNX1-mediated LUAD resistance to anti-PD-1 immunotherapy in a PD-1/PD-L1-independent manner. CircRUNX1 acts as a sponge for miR-4739 in LUAD cells To reveal the molecular mechanisms of circRUNX1-promoting LUAD immune evasion, we predicted the potential interaction between circRUNX1 and miRNA using the Star Base v3.0. We found that circRUNX1 can bind to 9 potential miRNAs based on StarBase v3.0. To determine the critical miRNA sponging circRUNX1, we performed a pulldown assay with a biotinylated circRUNX1 probe. As shown in Fig. 4a, we selected miR-4739 was the signi cantly captured miRNA for circRUNX1. To determine whether circRUNX1 and miR-4739 are co-localized, the FISH assay was used to NCI-H1975 and HCC-827 cells, and the results showed that circRUNX1 and miR-4739 were co-located major in the cytoplasm (Fig. 4b). AGO2 antibodymediated RIP assay showed that circRUNX1 and miR-4739, but not circANRIL (a circRNA veri ed to not bind to AGO2 protein [19]), were signi cantly captured in NCI-H1975 and HCC-827 cells (Fig. 4c). To further verify the predicted interaction between circRUNX1 and miR-4739, we performed RNA-pull down assays with a biotinylated miR-4739 mimics. In both NCI-H1975 and HCC-827 cells, circRUNX1 was signi cantly captured (Fig. 4d). In addition, we performed a luciferase assay using miR-4739 mimics or negative control (NC) mimics co-transfected with pLG3 luciferase reporter plasmids (which contained a wild-type or miR-4739-target mutant circRUNX1 sequence) into HEK-293T cells. Compared with the NC mimics, miR-4739 inhibited the luciferase reporter activity signi cantly in the cells with the wild-type circRUNX1 sequence but not the cells with either the wild-type-or the miR-4739-target mutant circRUNX1 sequence ( Fig. 4e and f). Importantly, the levels of circRUNX1 did not show obvious changes after miR-4739 expression was either upregulated or knocked down, and the levels of miR-4739 did not show obvious changes after circRUNX1 expression was either was overexpressed or knocked down (Fig. 4g-j; Supplementary Fig. 2a and b). These results demonstrate that miR-4739 and circRUNX1 are likely not degraded by each other.

PCSK9 was a direct target of miR-4739
To further reveal the underlying biological mechanisms of miR-4739 and its down-stream critical molecular, then, Targetscan was used to predict the potential target mRNAs of miR-4739. Interestingly, PCSK9, an immune-related gene is a potential target mRNA of miR-4739 (Fig. 5a). Dual-luciferase reporter assay results demonstrated that the relative luciferase activity of PCSK9 was obviously inhibited, suggesting that the PCSK9 mRNA is a target of miR-4739 (Fig. 5b). Then, the expression of PCSK9 mRNA and protein was detected in miR-4739 overexpression LUAD cell lines. The results indicated that the expression of PCSK9 mRNA and protein was signi cantly decreased in miR-4739-overexpressing cell lines ( Fig. 5c and d). Next, the expression of PCSK9 mRNA and protein was examined in circRUNX1 overexpression and knockdown LUAD cell lines, respectively. Our results showed that the expression of PCSK9 mRNA and protein was signi cantly increased in circRUNX1-overexpressing cell lines, while decreased in circRUNX1 knockdown cell lines (Fig. 5e-h). Next, a miR-4739 target sequence mutant circRUNX1 overexpression lentivirus (circRUNX1-mOE) transfected into the NCI-H1975 and Clau-3 cell lines to further determined that circRUNX1-mediated PCSK9 upregulation was in a miR-4739-depedent manner ( Supplementary Fig. 3a and b). Our results showed that the expression of PCSK9 mRNA and protein was not signi cantly increased in mutant circRUNX1-overexpressing cell lines ( Fig. 5i and j).
Knocking out PCSK9 reverses circRUNX1-induced LUAD resistance to anti-PD-1 immunotherapy Recently, it has been reported that higher levels of PCSK9 downregulates the expression of MHC I proteins on cancer cells and induces tumor resistance to anti-PD1 immunotherapy [15]. Therefore, we conjectured that circRUNX1 promotes LUAD resistance to anti-PD-1 immunotherapy via the miR-4739/PCSK9 axis. To further determine whether circRUNX1-induced resistance to anti-PD-1 immunotherapy of LUAD through the miR-4739/PCSK9 axis, we established PCSK9-knockout NCI-H1975 cells (NCI-H1975-KOP) (Fig. 6a). The PCSK9 protein did not detected after circRUNX1 expression was upregulated in the NCI-H1975 cells (Fig. 6b). Furthermore, we analyzed the anti-tumor effects of the PD-1 antibody in humanized NSG mice that received NCI-H1975-KOP cells with forced circRUNX1 expression or the respective mock cells (NCI-H1975-KOP-Mock, NCI-H1975-KOP-circRUNX1). Compared to that of the NCI-H1975-KOP-Mock cell group, the tumor growth in the NCI-H1975-KOP-circRUNX1 cell recipient xenograft mice showed no obviously phenotype of resistance to anti-PD1 immunotherapy (Fig. 6c-e). In addition, our results showed that the levels of MHC I was signi cantly decreased when circRUNX1 was overexpressed in the LUAD cells (Fig.   6f). To further explore the relationship between the circRUNX1 and immune evasion, we measured the MHC I expression in tissues in the above 30 cases of LUAD tissues (Fig. 6g). The results from a scatter plot analysis showed a negative correlation between circRUNX1 expression and MHC I expression in the LUAD tissues (Fig. 6h). These results indicate that circRUNX1 may exert its immunosuppressive effects by circRUNX1/ miR-4739/PCSK9/MHC I axis.
Discussion PD-1 monoclonal antibody-induced anti-tumor effect relies on restoration of the cancer-immunity system, initiated by immune effector cells including cytotoxic T lymphocytes, natural killer cells, et al [20,21]. Importantly, e cacious MHC I-mediated presentation of tumor-speci c antigens one of the critical factors for anti-PD-1-based immunotherapy curative effect [22]. Anti-PD-1-based immunotherapy have veri ed to be a critical therapeutic approach for LUAD. Mechanistically, anti-PD-1 immunotherapy acts by breaking the interaction between PD-1 and PD-L1, which are responsible for the promotion of cancer immune evasion [23]. However, the application of anti-PD-1 immunotherapy in LUAD is limited by the primary or acquired drug resistance [23,24]. Recently, circRNAs have been reported to play vital roles in the tumorigenesis and progression of tumors [25,26]. Previous studies have identi ed several irregular expressed circRNAs in LUAD that are dysregulated in LUAD tissues and function as tumor suppressor genes or oncogenes. For example, circRNA-ENO1 can promote glycolysis and tumor progression by binding to miR-22-3p and upregulating the expression of ENO1 in LUAD [27]. CircDCUN1D4 can upregulate thioredoxin-interacting protein (TXNIP) expression in LUAD by facilitating the interaction between HuR protein and TXNIP mRNA, thus inhibiting cell metastasis and glycolysis [28]. Wang et al. found that forced circRNA-002178 expression promotes T-cell exhaustion via regulation of the miR-34/PD-L1 signaling pathway [29]. Here, our results demonstrate that circRUNX1 is a critical regulator of immune evasion in LUAD cells.
Recent studies have con rmed that circRNAs are participate in the progression and immune evasion of various cancers via acting as miRNA sponges, therefore, modulating protein translation [11,25]. In this study, we used StarBase v3.0 to predict the potential miRNAs interaction with circRUNX1, then con rmed by biotin-coupled probe pull-down, AGO2-mediated RIP, and dual luciferase reporter assays. Our results showed that circRUNX1 was able to bind with StarBase v3.0. Furthermore, FISH assay con rmed the colocalization between circRUNX1 and miR-4739 in LUAD cells. We next attempted to explore the biological role and molecular mechanism of miR-4739 in LUAD cells and revealed that miR-4739 executed functions in LUAD immune evasion by targeting PCSK9. To further verify whether the biological function of h circRUNX1 in LUAD cells was associated with miR-4739/PCSK9 pathway, we carried out a block experiment using CRISPR-Cas9 to knock out PCSK9 expression. Our results demonstrated that circRUNX1 promotes LUAD immune evasion by upregulating PCSK9 expression via sponging miR-4739.
Decreased MHC I molecular expression is an important mechanism of cancer immune evasion [8].
Importantly, deleted MHC I expression induced cancer resistance to PD-1/PD-L1-based immunotherapy [30]. Inhibiting PCSK9 expression or blocking PCSK9 by antibody increases the expression of MHC I proteins on the cancer cell surface, thus, promoting robust intratumoral in ltration of cytotoxic T cells [15]. Interestingly, we found that PCSK9 protein levels were signi cantly increased in circRUNX1 overexpression LUAD cells. Here, our data reported the critical role of circRUNX1 in tumor resistance to anti-PD-1 immunotherapy and inhibition of MHC I signaling pathway in LUAD immune evasion.

Conclusions
Taken together, our study indicates that circRUNX1 is overexpressed in LUAD and its upregulation is correlated with poor prognosis. Mechanism, circRUNX1 promotes LUAD immune evasion by acting as a sponge for miR-4739, which decreases the activity of PCSK9/MHC I axis. These results demonstrate that circRUNX1 is a promising novel biomarker and therapeutic target in LUAD.

Declarations
• Ethics approval and consent to participate Ethical approval was obtained from the Zhongshan Hospital Research Ethics Committee.
• Consent for publication Not applicable.
• Availability of data and materials All data generated or analyzed in this study are included in this article and its supplementary information les.

• Competing interests
The authors declare no con icts of interest.

• Funding
This study was funded by the National Natural Science Foundation of China (82072575). All authors read and approved the nal manuscript.

Figure 6
Knocking out PCSK9 reverses the cancer-promoting function of circRUNX1 in the LUAD. a PCSK9 expression in NCI-H1975 cells was detected by western blot. b PCSK9 expression in NCI-H1975 cells with PCSK9 knocking out was modi ed by lentivirus-mediated circRUNX1transfection. c NCI-H1975-KOP-Mock or NCI-H1975-KOP-circRUNX1 cells were subcutaneously injected into huNSG mice, and when tumors volume reached about 100 mm3, and mice were treated with IgG or PD-1 antibody (The data are