CRISPR/Cas9 Deletion of SOX2 Regulatory Region 2 (SRR2) Decreases SOX2 Malignant Activity in Glioblastoma

Simple Summary Understanding how SOX2, a major driver of cancer stem cells, is regulated in cancer cells is relevant to tackle tumorigenesis. In this study, we deleted the SRR2 regulatory region in glioblastoma cells. Our data confirm that the SRR2 enhancer regulates SOX2 expression in cancer and reveal that SRR2 deletion halts malignant activity of SOX2. Abstract SOX2 is a transcription factor associated with stem cell activity in several tissues. In cancer, SOX2 expression is increased in samples from several malignancies, including glioblastoma, and high SOX2 levels are associated with the population of tumor-initiating cells and with poor patient outcome. Therefore, understanding how SOX2 is regulated in cancer cells is relevant to tackle tumorigenesis. The SOX2 regulatory region 2 (SRR2) is located downstream of the SOX2 coding region and mediates SOX2 expression in embryonic and adult stem cells. In this study, we deleted SRR2 using CRISPR/Cas9 in glioblastoma cells. Importantly, SRR2-deleted glioblastoma cells presented reduced SOX2 expression and decreased proliferative activity and self-renewal capacity in vitro. In line with these results, SRR2-deleted glioblastoma cells displayed decreased tumor initiation and growth in vivo. These effects correlated with an elevation of p21CIP1 cell cycle and p27KIP1 quiescence regulators. In conclusion, our data reveal that SRR2 deletion halts malignant activity of SOX2 and confirms that the SRR2 enhancer regulates SOX2 expression in cancer.


Introduction
SOX2 is a member of the SOX family of high mobility group (HMG) box transcription factors, and plays an important role in the maintenance of embryonic stem cell pluripotency and progenitor identity during embryogenesis and in adulthood [1]. In the embryonic central nervous system, SOX2 is expressed throughout the neural tube during early stages of neurogenesis and in progenitors as development progresses. It is also required for neural stem cell (NSC) activity in the adult. Several studies describe how SOX2 expression is elevated in a large proportion of cancers, including brain tumors such as glioblastoma (GBM) [2][3][4][5][6].
GBM is the most common malignant primary brain tumor in adults. The average survival of patients diagnosed with GBM is around 15 months [7]. GBMs are notorious for their resistance to therapy, which has been linked to genetic and cellular heterogeneity [8].
Among these, the glioma stem cell (GSC) subpopulation displays stem cell characteristics such as unlimited self-renewal capacity, quiescence and multilineage differentiation potential, being responsible for tumor initiation, progression, therapy resistance and tumor recurrence [9]. Several studies reported overexpression of SOX2 in GSC subpopulations and in patient samples, and the high levels are associated with tumor aggressiveness and worse prognosis. Furthermore, downregulation of SOX2 in GSCs impairs proliferation, self-renewal and their ability to form tumors in vivo, whereas SOX2 overexpression promotes cell proliferation, self-renewal capacity and tumorigenesis [3,[10][11][12][13][14][15]. These results show SOX2 as a relevant driver in GBM progression.
Transcriptional regulation of SOX2 is complex and not fully understood. The expression of the SOX2 gene in embryonic stem cells (ESCs) and multipotent neural progenitor cells in the ventricular zone of embryonic brains is supported in major part by a set of regulatory regions: SRR1, SRR2, SRR18, SRR107 and SRR111 [16][17][18]. In particular, SRR2 is located~4 kb downstream of the SOX2 coding region and has a regulatory core sequence comprising octamer and SOX2 binding sites. In ESCs and NSCs, SRR2 mediates SOX2 activation by recruiting OCT3-4-SOX2 or OCT6-SOX2, respectively, and, on their differentiation, SRR2 recruits p130-E2F4-SIN3A repressive complexes [17,19]. This is in part mediated by the association of p21 CIP1 and p27 KIP1 to SRR2. Indeed, when p27 KIP1 binds to SRR2, together with the p130-E2F4-SIN3A complex, this contributes to the repression of SOX2 and leads to differentiation programs in ESCs, mouse embryonic fibroblasts (MEFs) and teratoma cells [20]. Similarly, p21 CIP1 controls the proliferation of NSCs in the adult brain via SRR2 binding [21]. In breast cancer, high transcriptional activation of SOX2, monitored by an SRR2 reporter, has been associated with higher tumorigenic capacity in a small subset of cancer stem cell phenotypes (CD44+/CD24-) in triple-negative breast cancer cell lines and patient samples, showing the potential relevance of SRR2 in tumorigenesis [22,23].
Thus, in the present study, we used CRISPR/Cas9 technology to delete SRR2 in GBM cells. We observed a reduction of SOX2 expression, which is accompanied by impairment of proliferative activity and reduction of self-renewal capacity in vitro. Additionally, SRR2deleted cells lose the capacity to form tumors in vivo. These effects correlated with an elevation of the p21 CIP1 and p27 KIP1 cell cycle and quiescence regulators. These results show that SRR2 is required for SOX2-driven oncogenic activity in GBM cells.

Genotyping SRR2 Enhancer and Sequencing of Clones 2 and 5
Genomic DNA was extracted from U373MG cells and deleted SRR2 regulatory region and clones and control were identified by PCR genotyping (Table S2). The primer pair SRR2-F and SRR2-R amplified a 724 bp product in wild type cells and a product of around 200-400 bp cells carrying the SRR2 deletion when we used all guides. Sequencing revealed a difference of 1 nucleotide between both clones; clone2 deletion was 479 bp whilst clone5 deletion was 480 bp. Detailed information of sequencing of clones is presented in Figure S1. Sequencing was carried out in a 3130xl Genetic Analyzer (Applied Biosystems, Carlsbad, CA, USA)).
Neurosphere isolation was performed as previously described [25] with some modifications. Briefly, animals were killed by raising the concentration of CO 2 , and the brain was put into Phosphate-Buffered Saline (PBS) 1X (Gibco, St. Louis, MO, USA). The subventricular zone (SVZ) was isolated and centrifuged at 230× g for 5 min. After removing the supernatant, the tissue was mechanically disrupted by pipetting up and down and enzymatically disaggregated using 5 mL DMEM/F12 with 10% papain (Roche), previously activated for 30 min at 37 • C and with 25 U of DNase (Promega, Madison, Wisconsin, USA). Cells were maintained at 37 • C for 30 min in rotation (180 rpm), before being centrifuged at 230× g for 5 min. After centrifugation, cells were washed with PBS and re-suspended in 2 mL of complete DMEM/F12 supplemented with 40 ng/mL Epidermal Growth Factor (EGF) and 40 ng/mL basic Fibroblast Growth Factor (bFGF). The culture was grown as spheres in suspension and the medium was renewed twice a week. For differentiation assays, 50 neurospheres were seeded in chamber slides (Lab-Tek II Chambers, 1023-4121, Thermo Scientific, Budapest, Hungary) pretreated with 10 µg/mL laminin (L2020, Sigma-Aldrich, St. Louis, MO, USA) at 37 • C for 3 h. Cells were maintained in culture for 7 days in complete DMEM/F12 without EGF and bFGF. Half of the medium was replaced every two days.

Lentiviral Infections
For stable SOX2 overexpression, lentiviral infections were performed as previously described [15]. We used a pLM-mCitrine-SOX2 construct (SOX2) and a pWPXL-GFP construct (GFP) as controls. Cells were infected at a multiplicity of infection of 10 for 6 h.

Western Blot Analysis
Immunoblots were performed following standard procedures [26].

RNA Analysis
Total RNA extraction was performed by Trizol (Life Technologies, Budapest, Hungary). Reverse transcription was performed using random priming and Maxima First Strand cDNA Synthesis Kit (Thermo Fisher, Waltham, MA, USA), according to the manufacturer's guidelines. To analyze gene expression, quantitative real-time polymerase chain reaction (qRT-PCR) with 20 ng of cDNA was performed by Absolute SYBR Green mix (Thermo Scientific, Budapest, Hungary) in a LightCycler 96 thermo-cycler (BioRad, Hercules, CA, USA). Transcript levels were normalized to GAPDH and measured using the ∆∆Ct relative quantification method.

Proliferation and Colony Formation Assays
For cell count experiments, a total of 2.5 × 10 4 cells per well were seeded in duplicate in 6-well treated plates. The total number of cells for each condition was determined 1, 3 and 5 days after culture using a Neubauer counting chamber and a light microscope. For colony formation experiments, a total of 500 cells per well were seeded in duplicate in 6-well treated plates. After 10 days, cells were fixed with 4% paraformaldehyde for 30 min and stained with Giemsa (Sigma, St. Louis, MO, USA) 5% in PBS for another 30 min.

Oncosphere Formation Assay
For the oncosphere formation assay, 10,000 cells/well were seeded in non-treated 6-well flat-bottom plates. Primary oncospheres (1 ry GSCs) were grown for 7-10 days in GSC medium, and after quantification, spheres were mechanically and enzymatically disaggregated with accutase (Gibco, Waltham, MA, USA). Then, they were seeded for secondary oncosphere (2 ry GSCs) formation and maintained for another 7-10 days. Fresh medium was added every 2-3 days to the plate.

In Vivo Carcinogenesis Assay
All animal handling and protocols were approved by the animal care ethic committee of the Biodonostia Institute (PRO-AE-SS-100). For subcutaneous xenografts, U373MG SRR2 del control, clone2 and clone5 cells were harvested with trypsin/EDTA and resuspended in PBS before injection. A tumor initiation assay was performed by injecting 5 × 10 5 and 5 × 10 4 cells into both flanks of 8-week-old Foxn1 nu /Foxn1 nu nude mice. In the assay, external calipers were used to measure tumor size twice a week. From these measurements, tumor volume was estimated by V = L × W 2 × 0.5; where L is tumor length and W is the tumor width.

Statistical Analysis
All measurements were represented as mean values ± standard error (SEM) and analyzed by Prism v.8.0c (GraphPad Software, San Diego, CA, USA). Data between two groups were compared by an unpaired t-test, while those among multiple groups were tested using one-way analysis of variance (ANOVA), followed by Tukey's post hoc test. For Extreme Limiting Dilution analysis (ELDA), we used a chi-square test (χ 2 ). Standard significance levels were used: * p ≤ 0.05; ** p ≤ 0.01; and *** p ≤ 0.001.

Characterization of SOX2 Expression and Activation of SRR2 Enhancer in Glioma Cells
We analyzed the expression of SOX2 in the TCGA cohort of human GBM samples and compared them with healthy brain tissue. The expression of SOX2 was upregulated in GBM samples ( Figure 1A). Similarly, Western blotting revealed very high expression of SOX2 in freshly derived glioma stem cell (GSC) cultures from human patients and in U251MG and U373MG cell lines, while A172, T98G and U87MG expressed low levels ( Figure 1B,C). The levels in U251MG and U373MG cells are within the range of expression observed in tumor biopsies [15], suggesting that those levels are of biological relevance.
Since the association of p21 CIP1 and p27 KIP1 partly mediates the effect of SRR2 on SOX2 repression in several cell types [20], we studied the expression of p21 CIP1 and p27 KIP1 in glioma cell lines and biopsies. The levels of p21 CIP1 were elevated in GBM from the TCGA cohort, whilst p27 KIP1 ones were slightly reduced ( Figure 1D,E). The levels of p27 KIP1 were also generally low in GSCs and glioma cells lines ( Figure 1B,C). Glioma cell lines cultured under stem cell media grow as oncospheres, which express a striking elevation of SOX2 levels [15]. To characterize the impact of SRR2 enhancer in SOX2 expression on glioma cells, we performed ChIP assays in U87MG and T98G parental and U87MG oncosphere populations, studying the presence or absence of p27 KIP1 in this region. Interestingly, ChIP assays revealed that p27 KIP1 is highly enriched in SRR2 in U87MG and T98G parental cells and significantly reduced in U87MG oncosphere subpopulations ( Figure 1F). We also studied the presence or absence of p27 KIP1 in SRR2, in both normal and differenti-ated NSCs. For this, neurospheres generated in vitro from freshly dissected brains were used. Differentiated neurospheres showed an enrichment of p27 KIP1 on SRR2 compared to non-differentiated neurospheres ( Figure 1G). p27 KIP1 was recruited to SRR2 in differentiated mouse embryonic stem cells (mESCs) cultured as embryoid bodies ( Figure 1G), corroborating previous studies [20]. Altogether, these results suggest that SRR2 is involved in SOX2 activation in stem cells in physiological and pathological conditions in the brain, at least in part through p27 KIP1 binding.

SRR2 Regulates SOX2 Expression in GBM Cells
Next, we engineered a deletion of SRR2 using CRISPR/Cas9 technology in U373MG cells (Figure 2A), which express high SOX2 levels ( Figure 1C). We confirmed the deletion of SRR2 in U373MG SRR2 del clone2 and clone5 after transfection with guides and Cas9 ( Figure 2B and Figure S1). We then measured the expression of SOX2 in the different cell types. Immunofluorescence showed that most U373MG cells and U373MG CRISPR control cells (control) were positive for SOX2, in contrast with U87MG cells, which were used as a negative control ( Figure 2C). Notably, U373MG SRR2 del clone2 and clone5 showed a reduction in the number of SOX2-positive cells ( Figure 2C,D). In line with this, analysis with Western blots showed a reduction of SOX2 levels in SRR2 del clone2 and clone5 cells ( Figure 2E). These results indicate that SRR2 deletion leads to a reduction in SOX2 expression, and thus demonstrate that SRR2 enhancer activity is required for SOX2 expression in U373MG cells.

SRR2 Deletion Results in Impairment of Cell Proliferation and Stem Cell Activity in Glioma Cells
Next, we evaluated the effect of SRR2 deletion on the tumorigenic capacity of U373MG cells in vitro. By cell counting, we observed less cell growth in U373MG SRR2 del clone2 and clone5 cells compared to CRISPR control cells (control) and U373MG parental cells (U373) ( Figure 3A). Furthermore, SRR2-deleted cells presented a statistically significantly lower number of proliferative, phospho-Histone H3 (pH3)-positive cells ( Figure 3B). Indeed, clone2 and clone5 displayed 1.35 and 1.06 positive cells, respectively, compared to 2.53 in control cells and 2.67 in U373 cells. To investigate the role of SRR2 in stem cell activity, U373, control, clone2 and clone5 cells were seeded in a selective medium for enrichment of the stem cell population. We observed a significant reduction in primary and secondary oncosphere formation efficiency in SRR2-deleted GBM cells compared to control and U373 cells, suggesting reduced stem cell activity ( Figure 3C,D). In line with this, a reduced colony formation capacity was observed in cells deleted for SRR2 ( Figure 3E). No significant differences were observed between U373 and control cells, which acted similarly in all tests. Moreover, the effect in clone5 was more intense than in clone2 in all experiments. These results confirm that SRR2 regulates SOX2-driven malignant properties in GBM cells.

SOX2 Overexpression Rescues SRR2 Deletion Effects in U373MG Cells
To further determine whether deletion of SRR2 is necessary for SOX2 expression and oncogenic activity, we next tested whether SOX2 re-activation could restore the impaired phenotypes described in SRR2 del cells. Immunofluorescence and Western blotting showed a significant increase in the number of SOX2-positive cells and expression levels in SOX2 control cells, as well as restoration of SOX2 in clone2 and clone5 cells infected with ectopic SOX2 (Figure 4A-C). As expected and previously described [15], forced expression of SOX2 significantly increased proliferation and stem cell properties in control U373MG cells ( Figure 4D-H). Interestingly, clone2 and clone5 cells with ectopic expression of SOX2 also showed higher cell growth and a larger number of pH3+ cells compared to their respective controls ( Figure 4D,E). Similarly, the ability to form primary and secondary oncospheres was increased in clone2 and clone5 cells with SOX2 restoration, reaching the numbers of control cells ( Figure 4F,G) In agreement with this, clone2 and clone5 cells with SOX2 restoration displayed larger and bigger spheres ( Figure S2), as well as increased capability to form colonies at low density compared to their respective controls ( Figure 4H). These results show that ectopic expression of SOX2 rescues several phenotypes of SRR2 del cells, indicating that the expression and the oncogenic activity of SOX2 is, at least in part, mediated by the SRR2 regulatory region.   3.5. Expression of p27 KIP1 and p21 CIP1 Is Altered in SRR2-Deleted Cells p27 KIP1 and p21 CIP1 , regulating quiescence and acting as cell cycle brakes, both repress SOX2 expression through SRR2, consequently leading to the reduction of proliferation and/or to differentiation programs in the adult brain [20]. At the same time, SOX2 controls the expression of both proteins in different cell types, forming a molecular loop [21,[27][28][29].
To study the effect of the SRR2 deletion on those two proteins, we performed Western blots of p21 CIP1 and p27 KIP1 in U373MG control and SRR2 del cells cultured in normal medium ( Figure 5A) and as 2 ry oncospheres ( Figure 5B). We observed that SRR2 del clone2 and clone5 cells presented higher levels of p27 KIP1 and p21 CIP1 than control cells in normal conditions ( Figure 5A). Furthermore, the protein expression level of p27 KIP1 and p21 CIP1 in oncospheres was also higher than in controls ( Figure 5B). Moreover, SRR2-deleted cells presented increased p21 CIP1 and p27 KIP1 mRNA expression compared to control cells ( Figure 5C,D). Additionally, SRR2 del clone5 cells also showed higher protein levels of p53, the upstream target of p21 CIP1 ( Figure S3). On the contrary, SRR2 del cells presented decreased mRNA and protein levels of well-known stem cell markers such as BMI1, NANOG, NESTIN, OCT4, ID1 and OLIG2 ( Figure 5E,F), supporting the idea that SRR2 is necessary for the SOX2 regulation of stemness pathways.

SRR2 Deletion Reduces Tumor Initiation and Progression
Finally, in order to assess the effect of SRR2 deletion on the tumorigenic capacity of glioma cells in vivo, we performed a tumor initiation assay in immunodeficient mice by injecting different concentrations of U373MG control, SRR2 del clone2 and clone5 cells. One hundred percent of animals injected with 100,000 control cells developed tumors versus 25% in those injected with the same numbers of clone2 cells, and 0% in those receiving clone5 cells ( Figure 6A). Using the ELDA tool, we calculated the frequency of tumor-initiating cells. This revealed a significantly lower frequency in clone2 and clone5 cells, respectively, of 1/1.9.10 6 cells and no tumor-initiating cells (1/∞), compared to 1/4.10 5 in control cells ( Figure 6B). Of note, clone5 cells were unable to form tumors, further showing a stronger effect than clone2. Moreover, tumors derived from SRR2deleted clone2 cells were significantly smaller compared to controls ( Figure 6C). Indeed, subcutaneous tumors derived from clone2 cells reached a volume of less than 25 mm 3 after 140 days, compared to tumors from control cells that grew to an average of 170 mm 3 during the same time ( Figure 6C,D). In line with these results, the weight of clone2 cell-derived tumors were lower than tumors from control cells ( Figure 6E). We then performed immunohistochemistry with SOX2, Ki67, BMI1, p27 KIP1 and p21 CIP1 . Tumors generated from clone2 displayed lower expression of SOX2, Ki67 and BMI1 compared to the tumors formed by control cells (Figure 6F,G), validating the in vitro observations of impaired proliferation and stemness of SRR2 del cells. In contrast, clone2-derived tumors exhibited higher levels of p27 KIP1 and p21 CIP1 (Figure 6F,G). In summary, these results reveal that SRR2 is required for SOX2 tumorigenic activity.

Discussion
SOX2 expression involves a complex network of transcriptional, post-transcriptional and post-translational regulators. The downstream SOX2 regulatory region 2 (SRR2) exerts relevant control on SOX2 expression. In this regard, SRR2 is a region involved in the expression of SOX2 in pluripotent stem cells during development and in the maintenance of stem characteristics in various adult stem cell populations [16]. In this study, we dissected the influence of SRR2 on SOX2 expression in glioma cells and demonstrated that the deletion of this regulatory region significantly reduces SOX2 levels, and hence activity in these cancer cells.
Our results confirm that SRR2 mediates repression of SOX2 expression in differentiation conditions in GSCs and NSCs. This effect might be mediated via different mechanisms. We reveal that it happens through p27 KIP binding. p21 CIP1 has also been identified as a regulator of SOX2 expression in NSCs via binding on SRR2 [21]. Additionally, OCT3/4 and SOX2 complexes can bind to SRR2 and promote SOX2 expression [30]. Our results extend the understanding and the relevance of p27 KIP1 repressive action for modulation of SOX2 levels because conclusions in previous studies were limited to physiological conditions. In this regard, we previously showed that p27 KIP1 recruits a repressor complex on SRR2 to repress SOX2 expression during embryonic stem cell differentiation [20].
We prove that SRR2 deletion leads to a reduction in the expression of SOX2 in glioma cells. Previous studies showed that SOX2 expression depends on the epigenetic status of SRR2 in NSCs [31]. Indeed, increasing CpG methylation in SRR2 coincides with lower expression of SOX2 and differentiation into neurons [31]. SRR2 contains a functional SOX-binding site whose mutations result in complete loss of transcriptional stimulating activity, demonstrating the positive feedback of SOX2 on its own expression [19]. Our results highlight that SRR2 controls SOX2 expression in brain stem cells not only under physiological but also under pathological conditions. Indeed, our findings experimentally demonstrate the relevance of SRR2 as an enhancer of SOX2 expression in cancer cells because deletion of SRR2 in GBM cells leads to a reduction of SOX2 expression.
The de-repression of SOX2 has been associated with a wide range of cancer types, including GBM [3]. In the latter, high levels of SOX2 are associated with GSCs and lower patient survival and its activity is required for tumor maintenance and progression, because silencing of the gene in GBM cells reduces the capacity to proliferate, migrate and self-renew, whereas SOX2 overexpression promotes an increase in cell proliferation, migration and self-renewal capacity [3,[10][11][12][13][14][15]. Here, we report that deletion of SRR2 in glioma cells leads to a reduction in proliferation and stem cell activity in vitro. Notably, in vivo, our xenograft models revealed that the lack of SRR2 results in an impairment of tumor initiation and growth. Furthermore, ectopic SOX2 overexpression in SRR2 enhancer-deleted cells restored SOX2 expression as well as proliferation and stemness capacities. These results reveal that the SRR2 regulatory region is required for SOX2 expression and hence tumorigenic activity in cancer cells.
To further characterize proteins involved in the regulation of SOX2 expression, we studied p27 KIP1 and p21 CIP1 . We analyzed these proteins because (i) SOX2's proliferative function in cancer cells is exerted, at least in part, through direct repression of p21 CIP1 and p27 KIP1 expression [27,28]; (ii) there is an interaction between SOX2 and p27 KIP1 in mice, in which we have just demonstrated that Srr2 plays a role [20,29]. We observed an increased expression of both p21 CIP1 and p27 KIP1 in SRR2-deleted glioma cells. Together with the information presented above, these results indicate the existence of repressive, cross-regulatory interactions between SOX2, p21 CIP1 and p27 KIP1 that play important roles in cancer cells. In summary, our results highlight the central role of SRR2 to maintain high levels of SOX2 expression, and hence malignant properties of glioma cells establishing the requirement for this regulatory region for SOX2-driven tumorigenic activities.

Conclusions
SRR2 deletion by CRISPR/Cas9 technology leads to a reduction in SOX2 expression in glioma cells. Consequently, cell growth and proliferation are reduced in these cells, while self-renewal capacity is impaired in GSCs. This is associated with an increase in the expression of the negative regulators of the cell cycle, p21 CIP1 and p27 KIP1 . Furthermore, deletion of SRR2 impairs tumor initiation and progression capacity of cancer cells in vivo.
Our results indicate that prevention of SRR2 activity leads to a reduction of oncogenic properties promoted by SOX2, suggesting that SRR2 may represent a novel therapeutic target in the treatment of GBM, to tackle GSCs, and hence relapse of these lethal tumors.