SIM2l attenuates resistance to hypoxia and tumor growth by transcriptional suppression of HIF1A in uterine cervical squamous cell carcinoma

Despite chemoradiotherapy being one of the most important modalities in advanced cervical cancer, there is a lack of both usable biomarkers to predict treatment outcome and of knowledge about the mechanism of refractoriness to the therapy. Here we identified a transcriptional factor Single-minded homolog 2 (SIM2) as an independent predictive biomarker for uterine cervical squamous cell carcinoma (CvSCC). The retrospective study showed that high expression level of SIM2 was correlated to good survival in CvSCC patients. SIM2 knockdown in CvSCC cell lines showed resistance to hypoxia with increased expression of HIF1A and its target genes. Loss of SIM2 also caused growth promotion, resistance to ROS, and radiation in 3D culture. Furthermore, SIM2 knockdown suppressed tumor growth with increased HIF-1α expression and angiogenesis in vivo. On the other hand, SIM2 long isoform (SIM2l)-overexpressed cells had contrary results, indicating the long isoform plays a key role for maintenance of these phenotypes. These data indicated that SIM2l has a potential to be precision medicine for CvSCC patients and that anti-angiogenesis therapy might be usable for SIM2lLow poor survivors.


Results
Decreased expression of SIM2 is a worse prognosis factor in cervical squamous cell carcinoma (CvSCC) patients. SIM2 has been reported to be over-expressed in prostate cancer 20,21 , whereas its expression is decreased in esophageal squamous cell carcinoma compared to normal tissue 26 , indicating SIM2 regulation is distinct by cancer types. About this notion, we first investigated the SIM2 mRNA level in CvSCCs among various types of cancer by RNA sequencing dataset that consisted of 8449 cancer patients from The Cancer Genome Atlas (TCGA) database (Fig. 1a). SIM2 mRNA expression was the highest in prostate cancer, and CvSCC was the fifth highest in all cancers. In CvSCC, SIM2 mRNA levels varied by cases, indicating the existence of distinctive patients with low or high SIM2 mRNA expression. We next analyzed SIM2l protein expression of CvSCC and normal cervix by immunohistochemistry (Fig. 1b). SIM2l was not expressed in the squamocolumnar junction (SCJ) epithelia that are the HPV-related and preneoplastic epithelia but expressed in parabasal cell layers of normal cervix. Although a major subset of CvSCCs expressed no/low SIM2l protein, we found a subset that expressed SIM2l protein aberrantly and highly. Consistent with these observations, SIM2 mRNA expression was lower than normal cervix by expression analysis using Gene Expression Omnibus database (GEO) (Supplementary Figure S1a). We then evaluated the relationship between SIM2 mRNA expression and clinical outcome using another TCGA dataset including RNA sequencing data of 248 CvSCC patients. In this data, 7.9% (20/253) cases were classified into a "SIM2 High" group whose SIM2 expression is more than two-fold higher than the average (Supplementary Figure S1b). Kaplan-Meier analysis revealed that overall survival (OS) of cases with the "SIM2 High" in SIM2 mRNA levels is significantly longer than others (Fig. 1c). In contrast, there was almost no aberrant expression of ARNT, which encodes a key dimerization partner of SIM2 necessary for its transcriptional repressor function (Supplementary Figure S1c). These data suggested that SIM2 plays an independent good prognosis factor in CvSCC patients. To reveal how SIM2 affects a prognosis in CvSCC patients, the relationship between SIM2 mRNA levels and clinical features was also examined by the same TCGA dataset (Table 1). SIM2 mRNA levels had no correlation to age, distal-or lymph node-metastasis, and clinical stage. Furthermore, there was no significant difference of SIM2 mRNA expression between Stage IB1/IIA1 (tumor size <4 cm) and Stage IB2/IIA2 (tumor size >4 cm), indicating that SIM2 contributes not to tumor growth. In contrast, high SIM2 mRNA expression significantly correlated to a good response to primary therapy. Since there are two splicing isoforms of SIM2 mRNA (SIM2s and SIM2l) and their biological function is different as mentioned above, we validated which isoform (or both) affects the radio-sensitivity in CvSCC patients. Using another GEO database, expression levels of SIM2s or SIM2l between pre-and post-radiotherapy were compared in 20 or 18 patients, respectively. The expression level of SIM2l was significantly decreased after radiotherapy, while that of SIM2s was increased at post-therapy ( Fig. 1d and Supplementary Figure S1d). These data suggested that SIM2l-expressed cancer cells can be selectively eradicated by radiotherapy, thereby residual tumor mainly consisted of cancer cells with no/low SIM2l expression. Therefore, a SIM2l-expressed tumor is sensitive to radiotherapy in CvSCC patients and becomes a good prognostic factor. Since all of these analyses were focused on SIM2 mRNA, relationship between SIM2 protein expression and clinical outcome remains unclear. However, there was some correlation between mRNA and protein expression levels in CvSCC cell lines (Supplementary Figure S1e), suggesting SIM2 protein expression level also reflects survival and sensitivity to radiotherapy in CvSCC patients.   Loss of SIM2 upregulates HIF-1α and contributes to apoptosis resistance under hypoxia. We then searched SIM2-target genes in CvSCC, because SIM2 has been reported to act as a transcriptional repressor. Using siRNA-based transient knockdown of SIM2 in five CvSCC cell lines (BOKU, SKG-Ι, SKG-ΙΙΙa, HCS-2, and CaSki), we searched the target from 7 candidate genes (HIF1A, TP63, PTGS2, EGFR, ERBB2, TGFB1, ITGB3) which have been reported as markers sensitive to radiotherapy in CvSCC 27 . By semi-quantitative RT-PCR analysis, HIF1A was only a gene in which mRNA expression was increased by SIM2 knockdown in more than one CvSCC cell line ( Fig. 2a and Supplementary Figure S2). Among 6 CvSCC cell lines, we used 2 cell lines (SKG-Ι and SKG-ΙΙΙa) for following analyses, because HIF1A expression was highly induced by SIM2 knockdown. In SKG-ΙΙΙa, quantitative real-time RT-PCR (qRT-PCR) revealed that SIM2 knockdown resulted in an increase of not only HIF1A but also its direct target genes including VEGFA, HK2, SLC2A1, LDHA, BNIP3, and PDGFB ( Fig. 2b), indicating that SIM2 controls these genes through suppression of HIF1A expression. Consistent with these observations, expression of HIF-1α was increased by SIM2 knockdown (Fig. 2c, Supplementary Figures S3,  S4a and S4b). Next, we tested whether increased expression of HIF-1α by SIM2 knockdown is a consequent of upregulation of four genes (EGLN1, EGLN2, EGLN3, and HIF1AN) that are known to degradate HIF-1α protein.
The mRNA expression levels of these genes were not strongly affected by SIM2 knockdown (Supplementary Figure S4c), indicating that SIM2 directly suppresses HIF1A expression. It is well-known that HIF-1α is dramatically increased under hypoxia and contributes to apoptosis resistance. Therefore, we examined HIF-1α expression by SIM2 knockdown under both normoxia and hypoxia, which showed that, under hypoxia, HIF-1α was notably increased ( Fig. 2d and Supplementary Figure S5). SIM2 knockdown also significantly reduced apoptotic cells by treatment of deferoxamine (DFO), which can mimic hypoxia, whereas the knockdown did not affect apoptosis under normoxia (Fig. 2e). These results suggest that SIM2 acts as a transcriptional suppressor to HIF1A and contributes to the escape from apoptosis of CvSCC cells under hypoxic stress. SIM2 loss enhances cell growth and resistance to oxidative and radiation stresses under 3D environment. As demonstrated above, an anti-apoptotic effect by SIM2 knockdown was detected only under hypoxia even though HIF-1α was upregulated under both normoxia and hypoxia. This finding suggested that SIM2 loss did not affect a cancer phenotype under normoxia. On the other hand, actual tumors are exposed by various stresses including ROS, low-level oxygen availability, and under-nutrition, all of which cannot accurately be recapitulated by a 2D plate culture. In this regard, we validated various cancer phenotypes under a long-term 3D culture, which somewhat mimics such tumor microenvironments as cell-cell interaction and hypoxia. To validate the SIM2-knockdown effect under a 3D culture, we first established two SIM2 shRNA-expressing SKG-ΙΙΙa clones (#13 and #15) ( Fig. 3a and Supplementary Figure S6). Although there was no difference between control shRNA-expressing mixed clones and the two SIM2 shRNA-expressing clones under a 2D culture (Supplementary Figure S7a), both of the two SIM2-knockdown clones significantly promoted spheroid growth in low-adhesive 3D culture plates (Fig. 3b). Furthermore, the spheroid formation ability of SIM2-knockdown clones was significantly higher than that of the control mixed clones (Supplementary Figure S7b), suggesting that SIM2-knockdown CvSCC cells have an advantage of growth promotion in vivo. Notably, induction of HIF1A mRNA by SIM2 knockdown was strongly enhanced under the 3D culture compared to the 2D culture (Fig. 3c). On the other hand, SIM2 knockdown did not induce MDR1 which have been reported to a key drug resistant gene regulated by HIF-1α 28 . Consistent with our results of the transient SIM2-siRNA transfection, stable SIM2-knockdown clones suppressed hypoxia-induced apoptosis under a hypoxia mimic condition, and immunocytology showed that HIF-1α induction was higher in the SIM2-knockdown clones than in the control shRNA-expressing mixed clones (Fig. 3d). Next, we investigated whether sensitivity to a cytotoxic drug, cisplatin (CDDP), oxidative stress, or radiation is decreased by SIM2 knockdown, because the CvSCC patients with low SIM2 expression showed a worse prognosis for both chemoradiotherapy and radiotherapy ( Fig. 1c and Supplementary Figure S1c). Although there was no difference in the sensitivities to CDDP between control shRNA-expressing mixed clones and SIM2-knockdown clones under a 3D culture (Supplementary Figure S3c), the SIM2-knockdown clones showed increased resistance to both H 2 O 2 and γ-ray treatment (Fig. 3d). These data supported our hypothesis that loss of SIM2 protects oxidative and radiation stress, resulting in a poor outcome for irradiation-based therapies in CvSCC patients.

SIM2l acts as a tumor suppressor gene via suppression of angiogenesis and induction of apoptosis in vivo.
We also verified whether our findings from the above in vitro studies are also observed in a mouse xenograft model. SIM2-knockdown SKG-ΙΙΙa cell lines (SIM2 shRNA #13 and #15) were subcutaneously grafted to immunodeficient mice and monitored for tumor progression. Consistent with the in vitro studies, tumor growth of the SIM2-knockdown clones was significantly higher than that of the control shRNA clones ( Fig. 4a and b, and Supplementary Table S1). Immunohistochemical analysis revealed that xenografted tumors of SIM2-knockdown clones increased HIF-1α expression ( Fig. 4c and Supplementary Figure S9) and showed induction of angiogenesis (CD31 staining in Fig. 4c) and suppression of apoptosis (TUNEL staining in Fig. 4c). As shown in Fig. 1d and Supplementary Figure S1e, gene expression profiles from microarray database  Table S2). Immunohistochemical analyses also revealed that xenografted tumors of SIM2l-overexpressing cells showed low angiogenesis with decreased HIF-1α expression and also showed increased apoptosis ( Fig. 5f and Supplementary Figure S10). These data strongly indicate that SIM2l independently acts as a tumor suppressor gene via suppression of HIF-1α-induced angiogenesis and of hypoxia resistance in CvSCC patients who received radiotherapy (Fig. 5g).

Discussion
Although chemoradiotherapy is an important modality for CvSCC, the relapse rate has been reported to be limited (28-64% in FIGO stages IIb-IVa) 28 . Therefore, predicting the response to chemoradiotherapy and presenting therapeutic options for a poor response group are of utmost significance for patients. Great efforts to identify such predictive biomarkers have been made by many researchers. However, only a few biological variances exist for cervical cancer patients who received chemoradiotherapy 29 . In this study, we demonstrate that SIM2 has a potential to be an independent prognostic marker for CvSCC patients. High SIM2 expression was positively correlated to OS and response to radiotherapy in CvSCC patients ( Fig. 1c and d). It has been reported that SIM2 dimerizes with ARNT, binds to CME, and transcriptionally represses expression of target genes. A recent ChIP-sequencing   study reported that 22 genes (e.g. Otx2, Arid1b, and Syngr1) are candidates for a mouse Sim2 target in embryonic stem cells 30

. However, little is known about its direct target genes in human cells including cancer cells.
Since SIM2l mRNA expression was decreased after treatment of radiotherapy (Fig. 1d), we hypothesized that SIM2 can downregulate genes that are involved in radio-resistance. From currently-reported radio-resistant genes in cervical cancer 27 , we identified HIF1A encoding HIF-1α as a target by SIM2 (Fig. 2a and Supplementary Figure S4a). As with SIM2, HIF-1α is a bHLH-PAS family member and activates more than 100 genes related to hypoxic response (i.e., anti-apoptosis, angiogenesis, metabolism, and proliferation) upon binding to ARNT 31 . It is well-known that SIM2 can indirectly suppress HIF-1α function through deprivation of ARNT 32 . However, no report about the direct relationship between SIM2 and HIF1A mRNA expression exists. Here, we provided the first evidence that SIM2 represses transcription of HIF1A. Notably, tumor hypoxia has been reported to have a major impact on the outcome of definitive radiotherapy and chemoradiotherapy among these variances, since its niche is thought to be an abolished oxygen enhancement effect 29,33 . Especially, increased expression of HIF-1α and its targets (e.g., vascular endothelial growth factor, hexokinases 2, and GLUT-1) have been reported to correlate with a worse prognosis in cervical cancer patients who received radiotherapy or chemoradiotherapy [34][35][36][37] .
Considering that HIF-1α is an important transcriptional factor in adapting to severe hypoxia (e.g. cell cycle arrest, anti-apoptosis, and angiogenesis), it can also be a key contributor to radio resistance 38 . As expected, knockdown of SIM2 increased not only HIF1A but also its target genes (i.e. VEGFA, HK2, and BNIP3) (Fig. 2b). In consistent with the result that SIM2 knockdown did not contribute to chemoresistance (Supplementary Figure S7c), it did not alter MDR1 expression (Fig. 3c). Therefore, HIF-1α-MDR1 transcriptional pathway may be not involved in CDDP resistance. Although SIM2 knockdown induced HIF1A expression, SIM2 did not suppress activity of the −1311 to +281 HIF1A promoter (not shown). Identification of the distal promoter is needed to reveal the mechanism of SIM2-mediated HIF1A downregulation. In addition to resistance to hypoxia in normal culture, SIM2 knockdown significantly promoted cell proliferation and spheroid formation under a 3D culture (Figs 2e and 3b, and Supplementary Figure S7b), which can mimic hypoxia 39 . Despite SIM2 knockdown induced HIF1A expression, it did not promote cell growth under normoxic 2D culture. These facts may be attributed to rapid degradation of HIF-1α by the ubiquitin-proteasome under normoxia. We also found that SIM2 knockdown reduced sensitivity to both oxidative stress and radiation under a 3D culture (Fig. 3e). These results are supported by previous reports that HIF-1α attenuates radio sensitivity and oxidative stress [40][41][42] . In agreement with those observations in a 3D culture, SIM2 knockdown promoted tumor growth with enhancement of HIF-1α expression and angiogenesis in vivo (Fig. 4). These results also underscored that SIM2 Low CvSCC patients face a poor prognosis (Fig. 1c) and more importantly indicate that targeting angiogenesis (i.e., chemotherapy by anti-VEGF antibody, Bevacizumab) may be a good therapeutic strategy for them. Moreover, our meta-analyses indicated that radio-sensitivity is ascribed not to the SIM2 short isoform but to the long one ( Fig. 1d and Supplementary Figure S1d). SIM2l-overexpressing CvSCC cells showed both cell growth inhibition and sensitivity to ROS in a 3D culture, and SIM2l-overexpressing tumors suppressed HIF-1α expression, angiogenesis, and tumor progression in vivo (Fig. 5b,c,e and f). Although SIM2l overexpression suppressed HIF-1α (Fig. 5a), both suppression of angiogenesis and induction of apoptosis was not observed in vitro (not shown). The inconsistent result between in vitro and in vivo may be attributed to functional change of HIF-1α depending on environment (e.g. hypoxia). Therefore, it is of important to reveal what kind of tumor niche contributes to enhancement of the SIM2l-HIF-1α axis in CvSCC.
In conclusion, SIM2l attenuates HIF-1α-mediated hypoxia-and radio-resistance; thus it has a potential not only as a radio-sensitive marker for CvSCC patients but as a way to provide a new therapeutic strategy for an SIM2 negative-radio-resistant one.

Materials and Methods
External data analysis. SIM2 mRNA expression data of 309 cervical cancer patients was downloaded as z-scores (RNA Seq V2 RSEM) from the cBioPortal (http://www.cbioportal.org/). The full clinical dataset, including age, gender, histological type, disease stage, treatment history, and overall survival status/period, were also downloaded from the TCGA portal and linked with genetic data. After excluding samples which histological types are not squamous cell carcinoma (n = 256), overall survival of SIM2 High and SIM2 Low quadrant group was analyzed using the Kaplan-Meier method, and differences between two survival curves were tested by the log-rank test in the different groups. SIM2s and SIM2l mRNA expression data was also downloaded from ArrayExpress (GEOD-27678, https://www.ebi.ac.uk/arrayexpress/) which dataset consists of cervical cancer biopsy samples from patients before/after receiving radiotherapy or CRT. Scientific DharmaFECT Transfection Reagents (Thermo Fisher Scientific, MA, USA) following the procedure recommended by the manufacturer. At 48 or 72 hours after siRNA transfection, cells were analyzed by RT-PCR or western blotting, respectively. At 24 hours after siRNA transfection, cells were exposed to normoxia (21% O 2 and 5% CO 2 ) or hypoxia (1% O 2 and 5% CO 2 ) for 48 hours. shRNA-based gene knockdown. 2.5 × 10 4 cells were seeded in a 24-well plate and incubated at 37 °C overnight. The cells were incubated in a serum free medium containing 5 μg/ml polybrene (Santa Cruz Biochemistry) at 37 °C for 4 hours. After incubation, 5 × 10 3 or 2.5 × 10 4 IFU of control and human SIM2 shRNA lentiviral particles (Santa Cruz Biochemistry) were added to the cells respectively and cultured in a penicillin-streptomycin free medium at 37 °C overnight. The medium with lentiviral particles was removed and incubated in normal medium for 2 days. The shRNA-expressing cells were selected by culture with selection medium containing 0.5 μg/ml puromycin (Thermo Fisher Scientific). The cells were isolated into single cell and cultured in a 96-well plate to obtain clone cells. For quantitative real-time RT-PCR, 3 × 10 6 cells were seeded per 3.5 cm EZSPHERE Dish (IWAKI, Chiba, Japan) and incubated for 9 days.

RT-PCR and quantitative real-time RT-PCR. Total RNA was isolated from cells in an ISOGEN
Plasmid construction and transfection. 2.5 × 10 4 cells were seeded in a 24-well plate and incubated at 37 °C overnight. The cells were incubated in serum free medium containing 5 μg/ml polybrene (Santa Cruz Biochemistry) at 37 °C for 4 hours. After incubation, cells were transfected with 1 μg of pCDH-CMV-mock or pCDH-CMV-SIM2l and cultured in a penicillin-streptomycin free medium at 37 °C overnight. The medium was removed and incubated in normal medium for 2 days. The cells were selected by culture with selection medium containing 0.5 μg/ml puromycin (Thermo Fisher Scientific).
Cell viability assays. 2D culture system: cells were seeded at 1 × 10 5 cells per well in 6-well plates. After 24 hours or 48 hours of incubation, cells were trypsinized and the viable cells were counted every day.
3D culture system: cells were seeded at 2 × 10 4 cells per well in a 96-well EZSPHERE plate (IWAKI, Chiba, Japan) and incubated for 9 days to estimate 3D viability. To estimate H 2 O 2 and γ-ray sensitivity, cells were incubated for 2 days before exposure. For estimating H 2 O 2 sensitivity, cells were exposed to H 2 O 2 (0, 500 μM, 1 mM) for 24 hours. For estimating γ-ray sensitivity, cells were irradiated with γ-ray (0, 5Gy) and incubated for 7 days. Viable cells were detected by CellTiter-Glo 3D Cell Viability assay (Promega, Madison, WI, USA) according to the manufacturer's instructions.
Colony formation assay. Pre-chilled 6-well plates were coated with 500 μl of Matrigel (BD Bioscince) per well and incubated at 37 °C for 30 min. 2 × 10 5 cells were pelleted by centrifugation, resuspended into 1 ml of medium and plated onto the coated surface. After mixing medium and 10% of Matrigel (BD Bioscince), 1ml of Matrigel-medium mixture was added to the plated culture. Cells were cultured for 96 hours, and the areas of spheres were quantified by ImageJ. Experiments were performed according to the protocol published by Bissell 43 .
Immunofluorescence analysis. Cells were cultured on glass chamber slides at 37 °C overnight and exposed to deferoxamine (0, 100 μM) for 24 hours. Cells were then fixed with 4% paraformaldehyde for 15 minutes, permeabilized with −20 °C methanol and 0.5% Triton X-100/PBS, and blocked with 10% fetal bovine serum and 2% bovine serum albumin in PBS. Cells were incubated with primary antibody for HIF-1α (#ab51608, Abcam, Cambridge, UK, 1:500) at 4 °C overnight. After washing by PBS twice, cells were incubated with Alexa 488-conjugated anti-rabbit IgG antibody (Invitrogen, CA, USA, 1:200) at room temperature for 10 minutes. After washing by PBS twice, the nuclei were stained with DAPI. Mouse xenograft model. Six-week-old female BALB/c-nu/nu mice were purchased from Charles River Laboratories (Beijing, China) and bred at a room temperature with a 12 hours' light/dark daily cycle. The mice were maintained under specific pathogen-free conditions and provided sterile food, water, and cages. 3 × 10 6 or 6 × 10 6 of cancer cells were suspended in a 3:2 mixture of PBS and Matrigel (BD bioscience) and then transplanted subcutaneously in the back of the mice by use of a 26 1/2-gauge needle. Body weight and tumor volume of the mice were measured weekly. Tumor volume was calculated using the following formula: tumor volume = D/2 × (d/2) 2 × 4/3π, in which D and d refer to the long and short tumor diameter. The mice were euthanized by anesthetic overdose at 6 weeks after the transplantation. We did not observe any pain behaviors and symptoms (e.g. impaired mobility, anemia, severe weight loss, or excess tumor growth) in all mice during the period. All experiments were conducted in accordance with the ethical guidelines of the International Association for the Study of Pain and were approved by the Committee for Ethics in Animal Experimentation of the National Cancer Center. Efforts were made to minimize the numbers and any suffering of animals used in the experiments.
Patients' samples. The tissues of cervical squamous cell carcinoma and normal cervix were obtained from patients (4 and 2 cases, respectively) who underwent surgery at Keio University Hospital (Tokyo, Japan). All patients provided written, informed consent, and the study protocol (No. 2007-0081) was approved by the ethics committee of Keio University. Experiments with these samples were performed in accordance with the approved guidelines.
Immunohistochemistry. Tissues were fixed in formalin and embedded in paraffin. Tissue blocks were sliced into 4 μm sections. After deparaffinization and rehydration, antigen retrieval was performed by autoclaving at 90 °C for 30 minutes in Tris-EDTA buffer (pH 9.0). For staining CD31, antigen retrieval was performed by autoclaving at 90 °C for 30 minutes in 0.5 M Tris buffer. After blocking endogenous peroxidase activity, tissue sections were blocked by PBS with 10% fetal bovine serum for 30 minutes and stained with primary antibodies against SIM2l (#sc-8716, Santa Cruz, CA, USA, 1:50), HIF-1α (#ab51608, Abcam, Cambridge, UK, 1:50), and CD31 (#ab28364, Abcam, Cambridge, UK, 1:100). After incubation overnight at 4 °C, tissue sections were incubated with anti-goat secondary antibody for staining SIM2l and EnVision+ Dual Link System-HRP (Dako, Carpinteria, CA, USA) for staining other targets following coloring by DAB (Dako, Carpinteria, CA, USA). All samples were counterstained with Mayer hematoxylin. TUNEL staining was performed by in situ Apoptosis Detection Kit (TaKaRa, Shiga, Japan) according to the manufacturer's instructions.
Statistical analysis. All data were expressed as the mean ± SD obtained from 3 independent experiments and p-values were calculated using unpaired t-test. In clinical data, overall survival (OS) was estimated by the Kaplan-Meier method and p-values were calculated by log rank test using GraphPad Prism version7 (GraphPad Software, California, USA). Values of p < 0.05 were considered significant (*p < 0.05, **p < 0.01, and ***p < 0.001).