Generation of a PAX6 knockout glioblastoma cell line with changes in cell cycle distribution and sensitivity to oxidative stress

The U251 N cell line was a kind gift from Dr. Hrvoje Miletic, University of Bergen, Norway [25]. The U251 N cells, the CRISPR-Cas9 cells generated from U251 N (this paper), and the HEK Phoenix cells (ATTC #CRL-3213) were all cultured in Dulbecco’s Modified Eagle’s Medium - high glucose (cat#D5796 Sigma) supplemented with 10% FCS (Biochrom AG, Berlin, Germany) and 1% penicillin/streptomycin (cat#P0781, Sigma). Cells were transfected using Lipofectamine 2000 (cat#11668–019, Invitrogen, Calsbad, CA). Authentication of the original U251 N cell line and generated cells, were performed at the accredited (ISO/IEC 17025) Centre of Forensic Genetics, University of Tromsø – The Arctic University of Norway. No ethics approval and informed consent were required to use any of the above mentioned cell lines in this study.

Oligos for guide RNAs were selected with the help of an online CRISPR-Cas9 Design tool, crispr.​mit.​edu/​. We chose three of the guide RNA sequences suggested by the program and added 5’CACCG-3′ to the five prime end of the oligo. Oligos for guide RNAs are displayed in Table 1.

The guide RNA sequences chosen had the least number of potential off-target sites calculated by the CRISPR Cas9 Design tool. Annealed guide oligos were cloned into the CRISPR-Cas9 expression vector pSpCas9 (BB)-2A-GFP (PX458) (#48138 Addgene plasmid [26]. The vector was linearized by BbsI (cat#R0539S New England BioLabs), and guide oligos were cloned into the vector by T4 DNA ligase (cat#15224–041, Invitrogen). The three different plasmids were separately transfected into U251 N cells. The next day, single cells were sorted into 96-well plates by FACS. Two weeks later colonies emerged and single colonies were expanded into 6 well plates. PAX6 knockout cells and PAX6 control cells were selected by western blot. Successful knockout was verified by genome sequencing: The N-terminal part of PAX6 was PCR amplified by use of the PAX6 oligos (Table 2). The PCR product was by Zero Blunt PCR cloning kit (Invitrogen cat#44–0302) and transformed into E.coli DH5-alpha. Colonies were sequenced by M13 primers. Four PAX6 positive single cell clones were verified by sequencing and kept as controls. Potential off-target sites were also sequenced. A list of potential off-target sites were generated by the online CRISPR Cas9 Design tool, crispr.​mit.​edu/. We chose the two potential off target-sites that were located in an exon or splice sites. For guide 2, these sites were located in NOL6 (Nucleolar Protein 6) and PPP1R9A (Protein Phosphatase 1 Regulatory Subunit 9A). Primers used for amplification of the genomic areas are displayed in Table 2, and cloning and sequencing were done as described above The cells were used in experiments for a maximum of five passages after pooling single cells clones, to avoid the potential of one clone dominating the pool.

Cell lysates were prepared by washing 90% confluent cells in 6 well plates with PBS, detaching and scraping in 100 μl 2xSDS PAGE buffer (100 mM Tris-HCl, pH 6.8, 4% SDS, 20% glycerol, 0.2% bromphenolblue and freshly added DTT to 0.2 M). Cells were boiled at 100 °C for 5 min. Cell lysates were sonicated 2.5 min in a icecold water bath, using the Bioruptor (Diagenode) with settings “High”, 30 s on/off cycles. Samples were briefly centrifuged before being loaded in premade gels (cat# NW04122BOX Invitrogen). See Blue Plus2 Prestained Standard (cat#LC5925 Invitrogen), Super Signal Molecular Weight protein ladder (cat#84785, Thermofisher) and MagicMark XP Western Protein Standard (cat#LC5602 Invitrogen) were used as molecular weight markers. Gels were run 35 min, at 200 V, 120 mAmpere. Proteins were blotted onto Li-Cor Odyssey nitrocellulose membranes (cat #926–31,092). Rabbit anti-PAX6 antibody (cat#AB2237, Millipore) at 1:1000 and mouse anti-Pax6 antibody(Cat# pax6, RRID:AB_528427, Developmental Studies Hybridoma Bank, University of Iowa) at a 1:37 dilution was used for screening single cell clones, and rabbit anti-Actin (cat#A2066, Sigma-Aldrich) at a 1:1000 dilution as a loading control. Anti-rabbit 680LT (cat#926–68,023), anti-rabbit 800CW (cat#926–32,213) or anti-mouse 800 CW (cat#926–32,212) IRDye secondary antibodies (LI-COR Biosciences) were used at a dilution of 1:10000 in TBST. Images were acquired on Odyssey Sa (LI-COR Biosciences).

Cells (50000) were seeded at sterile fibronectin coated coverslips in 24 well plates, and left in a cell incubator overnight (ON). The cells were fixed in 4% paraformaldehyde, 20 min at room temperature (RT), washed in 1xPBS and permeabilized by Methanol, 5 min, RT. Cells on coverslips were washed in PBS and blocked using 5% BSA, 20 min, RT. PAX6 primary antibody was added (Cat# pax6, RRID:AB_528427, Developmental Studies Hybridoma Bank, University of Iowa) at a dilution of 1: 37 in PBS, 1% BSA, 1 h, RT. The coverslips were washed in PBS, 1% BSA containing DAPI, and subsequently incubated in secondary antibody (Alexa Fluor® 555-conjugated goat anti-mouse IgG, Cat# A-32727 (1:1000). Life Technologies and Alexa Fluor® 647-conjugated goat anti-rabbit IgG (Cat# A32733 (1:5000)). One hour later the coverslips were washed repeatedly in PBS, 1% BSA and then in H₂O. The coverslips were attached to object glass using Mowiol mounting media and left to dry ON at RT. The next day the cells were photographed by Zeiss LSM 780 inverted confocal microscope with a 63 × 1.4 plan-Apochromat NA objective at different wavelengths of light to visualize EGFP and secondary antibody tags.

Cells were seeded in 6 well plates in triplicates of either 200, 500 or 1000 cells/well. The cells were incubated for 12 or 16 days in culture medium. At day 12 or 16, cells were fixed 5 min in 3:1 methanol: acetic acid and then stained 15 min in 0.005% crystal violet. Colonies were counted and the percent of cells forming colonies compared to plated cells (100%) were calculated. Morphology of clones and cells in general, were investigated by use of Zeizz Axiovert S100 microscope and the NIS-Elements Documentation software (Nikon).

Cells were seeded in Ibdi 2 well silicone inserts (Ibdi cat# 81176) in 24 well plates; 20,000 cells per insert well to obtain 95–100% confluency the next day. To inhibit further cell proliferation, the cells were treated with 50 μg/ml Mitomycin C (cat# M4287, Sigma) in complete culture medium for 2 h. The cells were washed twice with complete medium, and inserts were removed. The cell gaps were photographed (timepoint 0). Cells were returned to the tissue culture incubator and photographed again after 20, 48 and 72 h, at the same positions along the gap. The distance of cell migration was calculated by digitally drawing ten evenly spaced horizontal lines from edge to edge of the wounds at 0 and 20 h using NIS Elements BR 2.3 (Nikon). The mean value was calculated for each scratch. For each experiment, 4–10 scratches were measured per cell type. Three independent experiments were analyzed. To determine the statistical relevance of differences in migration between the different cell lines, Student’s t-test was performed.

Cells were seeded in 6 well plates, 50,000 cells per well, done in triplicate at Day 0, and counted on Day 1, 2 and 3 by use of a hematocytometer. Alternatively, 2200 or 4400 cells were seeded per well in 96 well dishes and analysed by Cell Titer Glo (G7571, Promega) according to the manufacturers instruction. This assay detect ATP, and thus the amount of viable cells in a culture. The luminescence was recorded using a CLARIOstar microplate reader (BMG Labtech).

Cells at 70–90% confluency were detached by trypsin and washed twice in PBS. PBS was removed, leaving 200 μl for resuspension. While carefully vortexing cells, 2 ml cold 70% EtOH was added dropwise. Cells were incubated at 4 °C for a minimum of 1 h. After washing twice in PBS, cells were resuspended in 200 μl 50 μg/ml Propidium iodide (cat#P4170 Sigma) diluted in dH₂O. 50 μl 100 μg/μl RNase A (cat#EN0531 Thermo Scientific) was added, and cells were left in the dark, at room temperature for 30 min, before analyzing by flow cytometry in a LSR Fortessa (BD).

Cells were seeded to 50% confluency in complete medium. The next day medium was replaced with 300 μM hydrogen peroxide containing medium, and left for 24 h. Medium was collected and cells were washed in PBS that was collected. Cells were trypsin treated, and the previously collected medium was used to inactivate the trypsin. Cells, collected medium and PBS were centrifuged at 1200 rpm, 3 min. Cells were further treated according to protocol of the FITC Annexin V Apoptosis Detection Kit I (cat#556546 BD Pharmigen), and analyzed by flow cytometry at PI and FITC channels in a LSR Fortessa (BD).

At 50% confluency, cell growth medium containing chemotherapeutic agents were added to the cells for a 72 h incubation. DMSO was used as a solvent. The concentrations were as follows: Temozolomide (TMZ) (cat#T2577, Sigma Aldrich) 250 μM, Withaferin A (WA) (cat#W4394, Sigma Aldrich) 1.5 μM, Sulforaphane (SFN) (cat#574215, EMD Millipore) 10 μM. Cells were also incubated in growth medium containing 0.2% DMSO, or in growth medium. Cells were photographed and apoptosis assays were performed as described above.

RNA purification, cDNA synthesis and qPCR reactions are previously described in [27]. Housekeeping genes for standardizations were TFRC and GAPDH. The primer pairs for the housekeeping genes and for MMP2 were purchased as KiCqStart primers from SIGMA (PrimerPair ID: H_TFRC_1, H_GAPDH_2 and H_MMP2_1). Other primers were designed by use of Primer3 [28]. Primer3 designed primer pairs for qPCR are shown in Table 3.

The CRISPR-Cas9 technology was used to knockout PAX6 from the U251 N glioblastoma cell-line. Three different guide RNAs were chosen (crispr.​mit.​edu/​). The guide RNAs were designed to create mutations close to the 5’end of the PAX6 gene, between positions + 25 to + 58 relative to the transcription start site (TSS) (Additional file 1: Figure S1). The PAX6 protein expression in expanded single cell clones were evaluated by western blot using anti-PAX6 antibody. One of the guide RNAs produced several PAX6 knockout (KO) clones. In addition, some of the single cell clones had a marked reduction in PAX6 expression compared to wild type (WT) U251 N cells, indicative of heterozygous knockout of PAX6 (data not shown). Sequencing of PAX6 KO cells showed that cell lines originating from single cell clones had a variety of mutations creating stop codons downstream of the TSS (Additional file 2: Figure S2), therefore it will not be correct to label them as clones and they will hereby be referred to as cell lines or simply cells. We selected three PAX6 KO cell lines obtained by use of guide 2, namely 2.10, 2A.3 and 2A.28. As control cells, we used four single cell clones that had been through the CRISPR-Cas9 protocol of transfection and selection but remained PAX6 positive, namely 1.7, 1A.9, 1A.26 and 1A.67. By performing western blot (Fig. 1a) and immunocytochemistry (ICC) (Fig. 1b) using anti-PAX6 antibody, we verified the absence of PAX6 in the knockout cells. All the PAX6 KO cell lines were further transduced with a retroviral vector containing a Doxycycline (Dox) inducible EGFP-PAX6 gene to create cells that are PAX6 expression “rescued”. We observed a concentration dependent induction of PAX6 expression when Dox concentrations between 10 ng/ml and 500 ng/ml were tested (data not shown). A concentration of 10 ng/ml Dox was sufficient to induce a PAX6 expression similar to that of WT cells (data not shown), and this concentration was therefore used for further experiments with the PAX6 Rescue cells. Figure 1C shows a western blot of the three different PAX6 Rescue cell lines with and without Dox. The uninduced cells do not show EGFP-PAX6 expression (nor PAX6 expression). However it should be noted that for some experiments we observed weak EGFP-PAX6 expression without the addition of Dox, indicating leakiness of the inducible system. Leakiness of inducible expression systems is not an unusual phenomenon [29, 30]. The presence of PAX6 in the Rescue cells were confirmed by comparing ICC of the pooled PAX6 KO cells and pooled PAX6 Rescue cells. In addition to the green fluorescence mediated by the EGFP-tagged PAX6 protein, cells were DAPI stained, and immune-stained by PAX6 antibody (Fig. 1d). The intensity and location of the green fluorescence and the PAX6 immunostaining coincided and thereby confirmed induced expression of the EGFP-PAX6 protein. However, it also showed significant variability in the amount of PAX6 protein expressed in individual cells upon Dox induction. The variability in PAX6 expression amongst cells cannot be detected by western blot. Although it should be considered when performing experiments on the Rescue cells as PAX6 is known to have dose dependent effects.

It was apparent that there were morphological varieties in the PAX6 KO cells compared to the U251 N WT cells (Fig. 2). U251 cells are a mixture of morphological diverse cells that also form different types of colonies [31]. The WT cells consists of elongated neuron-like cells and some shorter cells. The U251 CRISPR-Cas9 PAX6 control 1.7 cells retains morphology as U251 N WT, whereas the PAX6 KO 2.10 cells have shorter triangular cobblestone-like morphology (Fig. 2a). The PAX6 KO cell lines 2A.3 and 2A.28 show similar morphology to WT, with 2A.28 having elements of the cobblestone morphology of the 2.10 cells (Fig. 2a). However, all three KO cell lines consist of varieties of cell shapes and sizes that are all present more or less in U251 N WT. However, each of them have the dominating cell morphologies described above. The three KO cell lines (2.10, 2A.3 and 2A.28) were pooled to diminish potential clonal effects in further experiments. We also pooled four different CRISPR-Cas9 PAX6 control cell lines, amongst these the 1.7 cells. Even though the pooled KO cells consisted of morphological divergent cells, the triangular cobblestone morphology dominated, while the pooled PAX6 control cells had similar morphology to the WT and the 1.7 cells (Fig. 2b) The morphology of the pooled Rescue cell lines are similar to pooled PAX6 KO, but with more dominance of elongated cells (Fig. 2b). The pooled cells were kept for five passages for experiments, with the described variety in morphology being constant during this time. We showed that knocking out PAX6 alters the cell morphology, with the main feature being a reduction in the elongated neural morphology of the U251 N WT cells.

To investigate the cells ability to maintain cell growth and form colonies independent of contact with surrounding cells, we performed colony formation assays. It has previously been shown that mir-335 affect the colony forming abilities of glioma cells, and that regulation of PAX6 expression contribute to this [32]. We performed successful colony forming assays for the KO 2.10 cells, the PAX6 control 1.7 cells and the WT cells. The variety of morphology between the individual KO cell lines was reflected in the various forms of colonies observed in the assay. We observed three different types of colonies, tight, intermediate and loose, like described for U251 WT by Cao and colleagues [31]. Sixteen days after seeding, the WT cells had formed mainly large colonies. The colonies were loose or intermediate with a center of tightly packed cells. In addition, WT formed tight colonies after 12 days and 16 days (Fig. 3a, b, e). The PAX6 control 1.7 cells formed colonies similar to WT, while the PAX6 KO 2.10 cells formed mainly smaller tighter colonies, often with cells on top of each other or tightly packed (Fig. 3a). The PAX6 KO 2.10 cells covered a smaller surface area than the colonies formed by the WT- and the PAX6 control 1.7 cells, but they seemed to contain the same number (or even a higher number) of cells.

The three different PAX6 KO cells lines, and the four different PAX6 control cell lines, were pooled and used in colony-formation assays. The colonies formed by the pooled KO cells displayed more variety than what was observed for the 2.10 KO cells alone (Fig. 3b), probably reflecting the differences in morphologies observed for the individual KO cell lines (Fig. 3a). EGFP-PAX6 Rescue cells were also used in colony formation assays, and the colonies formed were smaller compared to the PAX6 KO cells, but had the same morphologies of tight, intermediate and loose (Fig. 3c). The colonies formed by PAX6 KO cells were denser and more tightly packed than the colonies formed by the WT and the PAX6 control cells, and thus strongly stained by crystal violet (Fig. 3d). Colonies were counted if they contained more than fifty cells. However, due to the tight packing of the colonies formed by the PAX6 KO cells, some colonies having more than fifty cells may have been overlooked during counting. As an illustration, arrows in Fig. 3d indicate PAX6 KO colonies estimated to have more than fifty cells, and WT colonies indicated to have less than fifty cells. Initial experiments clearly showed that KO cells had increased colony formation capabilities compared to the wild type U251 N cell line. Further, experiments with pooled Rescue cells indicated reduced colony formation compared to pooled KO cells after Dox stimulation, but the standard deviations were too large to be considered statistical relevant (data not shown). To clarify the colony forming ability of the individual cell lines, the three KO cell lines (2.10 KO, 2A.3 KO and 2A.28 KO) and the Rescue cell lines derived from these (2.10 R, 2A.3 R and 2A.28 R) were used in colony formation assays together with WT U251 N cells. For the Rescue cells, both Dox stimulated and unstimulated cells were used. The Dox stimulated Rescue cells were FACS sorted for EGFP-PAX6 expression 16–24 h after stimulation and allowed to recover for 1 day before they were seeded for this assay. The KO cell lines and the U251 N cell line were also grown in media containing Dox, so that potential differences observed are the result of induced PAX6 expression and not the presence of Dox. The experiment was done three times in triplicate, and a representative result is shown in Fig. 3e. This shows that the different KO clones have huge variations in their capability to form colonies, with the 2A.3 KO cell line being equally bad as the U251 N cell line in colony formation (less than 1% of seeded cells formed colonies), while the 2.10 KO and 2A.28 KO cell lines were far better, with efficiencies ranging from 7 to 22% for 2.10 KO and 32–51%for the 2A.28 KO in three experiments (Table 4). The ability of the Rescue cells to generate colonies were similar to their original KO counterpart, but especially the 2A.3 R cell line was better than its KO counterpart in colony formation. However, Dox stimulation and FACS sorting of EGFP-PAX6 expressing Rescue cells did not lead to changes in the ability to form colonies in any of the three Rescue cell lines (Fig. 3e).

To summarize, we have shown that PAX6 influences colony morphology, and that the ability to form colonies is enhanced in two of three PAX6 KO cell lines compared to the U251 N cells.

The observed morphology of individual cells and cell colonies could indicate differences in mobility between the PAX6 KO cells and the PAX6 control (or WT) cells. To test this, migration assays were performed on pooled cells. The cells were treated with Mitomycin C to abolish proliferation. At 20 and 48 h after removal of well inserts, the PAX6 KO cells covered a larger area than WT and PAX6 control cells did (Fig. 4a). After 72 h, the entire region (originally cell-free at 0 h) was covered by the PAX6 KO cells. The WT and the control cells did not migrate at the same speed and there were still cell-free areas at 72 h after removal of well inserts. The differences in migration distance between the three cell lines are shown in Fig. 4b. These results are statistically relevant and show that the PAX6 KO cells have increased migration capacity compared to the PAX6 expressing cells; even though the morphology of WT cells (elongated, forming loose colonies) indicated that they would be more mobile. The results are in line with what others have found for PAX6 with regard to migration and invasion, e.g. in murine astrocytes reported by Sakurai et al. [33], and in glioblastoma cell lines, shown by Pavlakis et al. [34], Cheng et al. [32] and Mayes et al. [20].

In order to investigate the proliferation rate of the PAX6 KO cells compared to the WT cells, proliferation assays were performed by seeding cells and counting them at different time points. Regulation of PAX6 expression levels by miR-335 indicates that PAX6 restrict cell proliferation in glioblastoma cell lines [32]. The PAX6 KO 2.10 cells clearly proliferated faster than WT cells and the PAX6 control 1.7 cells (Fig. 5a). Pooled cells displayed the same pattern (Fig. 5b). The reintroduction of EGFP-PAX6 into the KO cells (generating Rescue cells) caused a clear reduction in proliferation when Dox stimulated and FACS sorted before use (Fig. 5c). PAX6 has been shown to be involved in cell cycle regulation in many cells and tissues, however it affects cell cycle differently in various cell types [17, 35‐37]. We therefore wanted to investigate if there were changes in cell-cycle distribution in the PAX6 KO cells compared to the PAX6 expressing cells. The results showed that by knocking out PAX6 the percentage of cells in G0/G1 decreased more than 50% (Fig. 5d and e). This indicates that PAX6 keeps cells in the G0/G1 phase of the cell cycle. PAX6 KO cells showed a strong increase of cells in the G2/M phase (50% increase compared to the WT and the PAX6 control cells), with a strong reduction of cells in G0/G1- and S phase. For the PAX6 Rescue cells, the number of cells in the G0/G1 phase increased, while the number of cells in the G2/M phase decreased (Fig. 5d and e), creating a slight shift in the cell-cycle distribution for the PAX6 KO Rescue cells towards the WT distribution. To investigate if prolonged incubation with Dox would result in increased PAX6 expression (and thus less heterozygosity), the incubation time and concentration with Dox was increased to 6 days and 25 or 50 ng/ml. However, the results maintained the same. The difference in number of cells in G1/G0-phase for PAX6 KO cells compared to Rescue was small but significant, with a p-value < 0.05. Use of Dox stimulated and FACS sorted Rescue cells did not change this result (results not shown). Cyclin D1 and p27 have both been shown to be involved in PAX6 regulation of cell cycle in other cell lines [37, 38]. We therefore compared the expression of these genes in the PAX6 KO cells and WT cells, by use of RT-qPCR. We observed minor fold changes (− 1.5) between PAX6 KO and WT cells with regard to p27 expression. However, for cyclin D1 there was a fold change of − 2.4 in the PAX6 KO cells compared to the WT cells (Table 4). If cell cycle regulation by PAX6 was mediated by regulation of cyclin D1 expression, the opposite would be expected, that is; lower level of cyclin D1 in the WT cells that were arrested in the G1/G0 phase of the cell cycle. An alternative explanation for the low level of Cyclin D1 in the PAX6 KO cells is the fact that a higher proportion of these cells are in the G2 phase of the cell cycle, and thus would be expected to have lower level of Cyclin D1 expression since the expression varies through the cell cycle. Hence, by harvesting and comparing mRNA from cell populations with striking differences in their G1 to G2 phase distribution one would expect differences in the level of Cyclin D1 mRNA.

To summarize, PAX6 KO cells proliferate at a higher rate, and display altered cell cycle distribution with accumulation in the G2/M phase compared to the WT and PAX6 control cells.

There are reports linking PAX6 to oxidative stress in glioblastoma cells [23]. H₂O₂ is naturally produced in large amounts by tumor cells [39‐42]. To assess whether PAX6 KO, WT and control cells were differently affected by oxidative stress we exposed the PAX6 2.10 KO cells, the WT and the PAX6 1.7 control cells to various concentrations of H₂O₂ for 24 h. At the H₂O₂ concentration of 60 μM the PAX6 2.10 KO cells displayed minor changes, with a small proportion of the cells rounded up. However most of the cells still looked vital and attached. Similareffect was obtained using a 3 times lower concentration of H₂O₂ (20 μM) on the WT and the PAX6 1.7 control cells. At 60 μM H₂O₂ most of the WT and PAX6 1.7 control cells were rounded up or detached from the surface (data not shown). This confirmed that the presence of PAX6 renders the cells more sensitive to oxidative stress.

To investigate this further, we incubated pooled PAX6 KO cells, WT, PAX6 control and PAX6 Rescue cells in 300 μM H₂O₂ for 24 h. The concentration of 300 μM was chosen as Chang and colleges found that PAX6 increased glioma cell sensitivity to detachment induced oxidative stress at ROS levels comparable with what 300 μM H₂O₂ induce [23]. Cells were stained with PI and anti-Annexin V FITC, and analyzed by FACS. The results strongly support that the PAX6 KO cells have a higher tolerance to H₂O₂ induced oxidative stress than WT cells (Fig. 6a and b). While 50% of the H₂O₂ treated KO cells were healthy and intact, less than 20% of the WT and PAX6 control cells were viable. The highest percentage of late apoptotic and dead cells (79%) were found in the sample of PAX6 control cells, closely followed by the WT cells (72.9%). For PAX6 KO cells only 41.5% were found to be late apoptotic or dead. In line with this, the PAX6 Rescue cells displayed less viable cells and increased the percentage of apoptotic cells upon H₂O₂ treatment compared to the PAX6 KO cells (Fig. 6a). This was observed for the cells even without Dox treatment, and we believe that is due to the leakiness of the Dox inducible system. However, the difference was more pronounced for Dox induced cells (Fig. 6b). While Dox in combination with H₂O₂ had no effect on the amount of live PAX6 KO cells (52.7% vs 50.1% in Dox stimulated vs unstimulated cells), the combination of Dox and H₂O₂ caused even more reduction in live Rescue cells (26.6% vs 31.1% in Dox stimulated vs unstimulated cells) (compare respective panels in Fig. 6a and b). This indicates that the increased PAX6 expression in the cells make them more sensitive to H₂O₂. To summarize, these results show that depletion of PAX6 affects cell morphology, cell proliferation, cell cycle distribution and sensitivity to oxidative stress – and these changes are reversed by re-introduction of PAX6 using a Dox inducible PAX6 expression vector. Furthermore, PAX6 depletion also highly enhanced the formation of colonies in a colony formation assay. However, this was not reversed by reintroduction of PAX6.

To investigate if PAX6 and its ability to render cells more sensitive to oxidative stress would play a role in treatment of glioblastoma, we treated the PAX6 KO cells and controls with various chemotherapeutic agents; Temozolomide (TMZ), Withaferin A (WA) and Sulforaphane (SFN). TMZ treatment, 250 μM for 72 h, reduced the cell number of both the WT and the PAX6 KO cells compared to untreated cells as observed in the microscopy. FACS was used to investigate the proportion of live and dead cells. As shown in Fig. 7, the TMZ treated PAX6 KO cells had more than double the amount of cells in the category late apoptotic/necrotic/dead cells compared to the WT cells. This clearly shows that PAX6 expression is beneficial for cell survival after TMZ treatment. WA treatment, 1.5 μM for 48 h, showed no differences in the percentage of cells in late apoptosis/necrosis for the WT (49%) and KO (48.4%). However, there were less live cells, and more early apoptotic cells for the WT (40.2 and 10.8% respectively) than for the PAX6 KO cells, where the comparable numbers for live and early apoptotic cells were 47.9 and 3.7%, respectively (Fig. 7). WA induces oxidative stress in cancer cells [43] and that may be the reason for WT cells being more sensitive to the treatment compared to PAX6 KO cells. Treatment with SFN, 10 μM for 48 h, seemed to strongly reduced proliferation in both PAX6 KO and control cells when observed in the microscopy. PAX6 KO cells displayed a slightly higher percentage of vital cells (84.3%) compared to the WT cells (76.4%), which had more apoptotic cells in the FACS analyses (Fig. 7). This strongly support that PAX6 expression sensitizes U251 N cells to chemotherapy treatment.

Changes in gene expression between PAX6 KO cells and WT cells were studied by RT-qPCRs for selected genes (MMP2, CAV1, Cyclin D1, p27, and Nrf2) (Table 4) known to be regulated by PAX6, or known to participate in proliferation, migration and/or redox regulation. MMP2 is involved in migration and is a known PAX6 target [20]. To our surprise, the PAX6 KO cells showed a 5-fold reduction of MMP2 compared to WT cells (Table 5). This is contradictory to what others have shown [20]. Another gene involved in migration and proliferation is CAV1. RT-qPCR showed a 2-fold downregulation of CAV1 in the PAX6 KO cells, which is in line with what we have observed when PAX6 was knocked down in the pancreatic adenocarcinoma cell line HPAFII (Forsdahl et al: in preparation). In the HPAFII cell line, knockdown of PAX6 caused enhanced proliferation and migration, in addition to the reduced expression of CAV1 (results not shown). Cyclin D1 and p27 are known target genes of PAX6 [44, 45], and important molecules regarding the regulation of cell cycle progression. PAX6 KO cells had a 2.4 fold downregulation of cyclin D1 compared to WT, while p27 expression was not significantly affected by PAX6 KO (Table 5). High levels of Cyclin D1 is required for the progression from G1 to S in the cell cycle. The expression of cyclin D1 is expected to be low or absent in most of the other cell cycle phases, although cancer cells may show aberrant regulation. A reduction in PAX6 KO cells in the G0/G1 phase compared to WT cells (as shown in Fig. 5D), might reflect the difference in cell cycle distribution between the PAX6 KO and WT cells, and not the PAX6 regulation of Cyclin D1 expression. Nfr2 is a key regulator of the oxidative stress defense mechanisms, and controls the basal and the induced expression of an array of antioxidant response element–dependent genes [46]. PAX6 KO cells has a 3.7 fold downregulation of Nrf2 compared to WT cells (Table 5), which is interesting since we have shown that these cells are in fact more resistant to oxidative stress by H₂O₂. PAX6 expressing cells are thus more sensitive to oxidative stress despite having higher expression of Nrf2 than the PAX6 KO cells. This urge that the effect of PAX6 on other regulators of oxidative stress responses should be included in further studies.