AP-2α Regulates S-Phase and Is a Marker for Sensitivity to PI3K Inhibitor Buparlisib in Colon Cancer

Colon cancer is one of the most common cancers with approximately 1.4 million new cases diagnosed each year and is responsible annually for 639,000 deaths (1). The incidence of colon cancer has decreased in certain countries, including the United States, largely due to routine screening and removal of precancerous lesions; however, the incidence is increasing in other countries and, despite the decreasing incidence in the United States, colon cancer remains the third leading cause of cancer death (2). FOLFOX is the first-line chemotherapy regimen for colon cancer and has been shown to be efficacious and successful at improving survival when used as adjuvant therapy (3, 4); however, chemoresistance is still thought to be the cause of treatment failure in over 90% of patients with metastatic cancer (5). Hence, molecular characterization of tumors including next-generation sequencing to identify gene mutations that alter response to chemotherapeutic drugs continues to be of great interest to optimize targeted, personalized therapies (6).

The use of gene expression profiling in colon cancer has been validated to predict recurrence and estimate prognosis, but is thus far unable to reliably predict response to chemotherapy (7). Currently, the National Comprehensive Cancer Network does not feel there is sufficient evidence regarding next-generation sequencing to direct therapy and recommends testing only in select cases. For example, testing for mismatch repair gene status, HER2 amplification, KRAS and BRAF mutations in certain patients, such as those with metastatic disease, is encouraged to guide second line chemotherapeutic agents that offer clear benefit to patients with these mutations (8). These advancements have significantly improved survival in a subset of patients with colorectal cancer and have avoided unnecessary therapy in patients likely to be resistant (9); however, the majority of patients currently have no targeted therapy options (10). Therefore, expanding our knowledge of genetic markers associated with sensitivity to specific chemotherapeutic agents is of utmost importance to improve outcomes and prevent chemoresistance.

Transcription factor activating protein 2 alpha (AP-2α), encoded by the TFAP2A gene, is one of five members of the family of AP-2 factors which are highly expressed early in differentiation of the ectoderm (11). The AP-2 family regulates the molecular phenotype of multiple cancer types, including breast, thyroid, melanoma, and colon cancer (12–17). AP-2α functions as a tumor suppressor in many cancer types (18–20) and influences response to chemotherapy (21, 22). In colon cancer, although the rate of somatic mutation is quite low, AP-2α has been shown to be involved in tumorigenesis (23, 24), cell growth (14), and acts as a tumor suppressor (23, 24), potentially through interactions with p53 (25). Overexpression of TFAP2A inhibited tumorigenesis in the colon cancer cell line SW480 (24) and in Apc(min) mice, in vivo gene delivery of AP-2α inhibited intestinal polyp formation (23). Given the importance of expanding our knowledge regarding gene expression profiling to direct chemotherapy and the growing body of evidence indicating AP-2α as an important regulator in colon cancer, we sought to examine the molecular changes in colon cancer when AP-2α is lost with a complete knockout (KO) and its effect on responsiveness to chemotherapeutic agents.

The cell lines HCT116 (RRID:CVCL_0291), LoVo (RRID:CVCL_0399), SW48 (RRID:CVCL_1724), LS180 (RRID:CVCL_0397), and MDA-MB-231 (RRID:CVCL_0062) were purchased from the ATCC and were used at a low passage number (<10) with no further testing. HCT116 was maintained in McCoy's 5A media, LoVo in F-12K media, SW48 in DMEM/F12, LS180 in EMEM, and MDA-MB-231 in DMEM. Cell lines were tested for Mycoplasma using LONZA MycoAlert Mycoplasma Detection Kit. All media were supplemented with 10% FBS and 1% Plasmocin (InvivoGen, catalog no. ant-mpp).

To minimize off-target effect, we used pCas9D10A-GFP plasmid for coexpression mutant CasD10A nickase (Addgene plasmid # 44720; http://n2t.net/addgene:44720; RRID:Addgene_44720) and two guide RNAs (gRNA) cloned as gBlocks into pCR4-TOPO Vector (Thermo Fisher Scientific; Guide A: ATCAAACTGTAATTAAGAA and Guide B: TTCTACATGCTGCAACAAA) within exon 3 of TFAP2A. Lipofectamine 2000 was used for DNA transfection of HCT 116 CCL-247 cells accordingly to the manufacturer protocol (molar ratio Cas9D10A/gRNAa/gRNAb was 1/5/5). After 24 hours, GFP-positive cells were sorted as a single cell per well in 96-well plates. Colonies were screened after 14 days for the presence of genomic alterations (insertion/deletion) using a pair of primers TFAP2A_EX3_F1_ctrl AGC TAG CCT GTT GGC ATT ACC and TFAP2A_EX3_R1_ctrl CCT CTT GAG TTG CAA AGC CC. Disruption of ORF was confirmed by sequencing and Western blot analysis with AP-2α antibody.

Cells were transfected using siRNA directed toward nontargeting (NT) #2 (Ambion by Life Technologies, catalog no. 4390846), TFAP2A (Stealth siRNA by Invitrogen ID:82654707) and TGM2 (Ambion by Life Technologies, catalog no. 4390824) with Lipofectamine RNAiMAX reagent (Thermo Fisher Scientific, catalog no. 13778150), as per manufacturer's instructions. After 72 hours of incubation, cells were immediately analyzed or used in subsequent experiments.

Knockdown of TFAP2A was completed in HCT116, LoVo, SW48, LS180, and MDA-MB-231 with lentiviral delivered short hairpin (sh)RNA vector (Sigma-Aldrich, catalog no. TRCN0000004926) at a multiplicity of infection (MOI) of 0.5. The same cell lines were transduced with NT short hairpin (shRNA) lentivirus (Sigma-Aldrich, catalog no. SHC016) at the same MOI to serve as controls. Cells were cultured in Opti-MEM media (Thermo Fisher Scientific, catalog no. 31985-062) for 24 hours with the respective lentivirus and then put under puromycin dihydrochloride (Sigma-Aldrich, catalog no. P9620) selection at 1 μg/mL for 24 hours. Knockdown of TFAP2A was confirmed with Western blot analysis.

Protein was isolated in RIPA lysis buffer (Millipore, catalog no. 20-188), supplemented with protease inhibitor (Roche, catalog no. 11836170001) and PhosSTOP (Roche, catalog no. 4906845001). The following primary antibodies were used according to the manufacturer's recommendations: AP-2α (Abcam, catalog no. ab108311), TGM2 (Abcam, catalog no. ab2386), CDK1+2+3 (Abcam, catalog no. ab32384), RAD51 (Santa Cruz Biotechnology, catalog no. sc-8349), CALB2 (Abnova, catalog no. MAB2741), GMNN (DSHB, catalog no. CPTC-GMNN-1-s), ACSS2 (Santa Cruz Biotechnology, catalog no. sc-398559), pan AKT (Cell Signaling Technology, catalog no. 4691), pAKT Thr308 (Cell Signaling Technology, catalog no. D25E6), and pAKT1/2/3 Ser473 (Santa Cruz Biotechnology, catalog no. sc-7985 R). GAPDH (Santa Cruz Biotechnology, catalog no. sc-47724) was used as a loading control. Secondary antibodies were used according to the manufacturer's specification: anti-rabbit horseradish peroxidase (HRP; Cell Signaling Technology, catalog no. 7074) and anti-mouse HRP (Cell Signaling Technology, catalog no. 7076). Protein was visualized with SuperSignal West Femto maximum sensitivity substrate (TFS, catalog no. 34095).

Experimental setup and analyses were performed in accordance to ENCODE Guidelines and Best Practices for RNA sequencing (RNA-seq). RNA was harvested using the RNeasy Mini Plus Kit (Qiagen, catalog no. 74134) at a concentration of 100–200 ng/μL and biologic triplicates of each condition were sent to the University of Nebraska (Omaha, NE) for further processing. The RNA quality was confirmed by the receiving facility and was subsequently sequenced with specifications for gene differentiation (50 bp, single-end reads). The RNA-seq analysis of the raw data was performed using the Galaxy web platform (RRID:SCR_006281) at usegalaxy.org with the built-in tools: HISAT2 (RRID:SCR_015530) and Cufflinks (RRID:SCR_014597). For gene expression comparisons, genes with significant expression changes as determined by Cufflinks data analysis were included, using an arbitrary log₂fold change cutoff to allow for a reliable analysis of consistently altered gene expression.

RNA-seq data are available at the Gene Expression Omnibus (GEO) database (National Center for Biotechnology Information, Bethesda, MD) under accession number GSE155169.

All obtained single-cell KO TFAP2A clones and maternal HCT 116 cell line were analyzed for the presence of potential heterozygosity c.3140A>G p.H1047R in Exon 20 of PIK3CA. Full sequence of Exon 20 was amplified using a pair of primers PIK3CA ex20_F1 GCC ACA CAC TAC ATC AGT GG and PIK3CA_ex20_R1 AGA GAT TGG CAT GCT GTC GAA. Obtained PCR product of 372 bp was cloned in pCR4-TOPO Vector and at least 12 individual colonies were sequenced per each single-cell clone or maternal cell line.

Chromatin immunoprecipitation sequencing (ChIP-seq) was accomplished with AP-2α 3B5 antibody (Santa Cruz Biotechnology, catalog no. sc-12726X) as described previously (26). AP-2γ (Santa Cruz Biotechnology, catalog no. sc-12762X) was used as the negative control. ChIP-Seq data are available in the GEO database under accession number GSE155169.

For cell-cycle analysis, cells in culture were synchronized with 2.5 μg/mL Aphidicolin (Sigma-Aldrich, catalog no. A4487) in media supplemented with serum as described previously (26). After 36 hours of incubation, the Aphidicolin containing media was removed, cells were washed with PBS and 10% FBS containing media without Aphidicolin was replaced. Cells were harvested immediately after release from synchronization then at 4 and 8 hours after release and were fixed immediately with 70% ethanol. Fixed cells were stained within 24 hours of fixation with propidium iodide (PI; Life Technologies, catalog no. P3566) and DNA quantification was performed on a FACS Becton Dickinson LSR II machine.

For the apoptosis assay, cells were seeded in 6-well plates (Corning, COSTAR, 3516) and allowed to attach overnight in full media. After attachment, cells were washed with Dulbecco's Phosphate-Buffered Saline (DPBS) (Gibco) three times and cell were incubated in serum-free medium for 24 hours. Apoptosis was measured 24 and 48 hours after the cells were released from G₀ cell arrest by adding complete medium. The FITC-Annexin V Apoptosis Detection Kit (BD Biosciences) was used according to the manufacturer's protocol. In brief, floating and harvested cells were mixed, washed twice with cold PBS (DPBS) and resuspended in binding buffer at a final density of 10⁶ cells/mL. FITC-annexin (5 μL) and PI (5 μL) were added to 100 μL of the cell suspension containing 10⁵ cells. The cell suspensions were mixed by gently vortexing and incubated for 15 minutes at room temperature in the dark. Subsequently, 400 μL of binding buffer were added and cells were analyzed by flow cytometry using FACS analysis.

Cells were synchronized as above and harvested for protein 8 hours after release from synchronization. Protein quantification was completed using the Cell Cycle Phospho Antibody Array (Full Moon Biosystems, catalog no. PCC076) using the Antibody Array Assay Kit (Full Moon Biosystems, catalog no. KAS02) according to manufacturer's instructions. Cy3-streptavidin (Invitrogen, catalog no. 434315) was used to allow for fluorescent imaging quantification. Levels of expression of the protein were normalized to the background.

Using the Premo FUCCI Cell Cycle Sensor BacMam 2.0 (Thermo Fisher Scientific, catalog no. P36237) according to manufacturer's instructions, unsynchronized cells in media supplemented with 10% FBS were incubated for 16–24 hours prior to imaging. An Olympus IX81 Inverted Microscope was used to obtain time lapse images for time lapse and levels of fluorescence. For each condition within the experiment, six randomly chosen locations with similar levels of cellular confluency were chosen for imaging.

Cells were plated in 96-well plates in technical triplicates and were incubated with MTT (Thermo Fisher Scientific, catalog no. M6494) according to manufacturer's recommendations. Measurements to reflect viability were obtained on an Infinity 200 Pro (Tecan) plate reader at an absorbance of 570 nm. Measurements were normalized to the initial read 24 hours after plating to reflect changes in viability.

A clonogenicity assay was performed as previously described by Franken and colleagues (27). Each condition was completed as a biologic triplicate. Wells were imaged for analysis using a 55 mm lens Canon camera.

Colon adenocarcinoma cases and their corresponding “best tumor block” were identified in the diagnostic files of the University of Iowa Hospitals and Clinics. Tissue microarrays (TMA) were constructed from 34 cases with sufficient paraffin-embedded tumor. Four-μm-thick sections of these TMA blocks were cut and used for IHC. IHC was performed to detect expression of transcription factor AP-2α [rabbit mAb, clone EPR2688 (2), RRID:AB_867683, AbCam, 1:100 dilution] and TMG2 (mouse mAb, clone CUB 7402, RRID:AB_2287299, AbCam, 1:200 dilution). Expression for each marker was evaluated for extent (0%–100%) and intensity (0–3+) in each TMA core, and an overall H-score (mean extent × intensity) was calculated for by each tumor.

Invasion assay was completed as described previously (28). Cells were treated within an invasion chamber with either DMSO or 250 nmol/L NVP-BKM120 (Selleck Chemicals, catalog no. S2247) immediately after plating. Cells were incubated for 24 hours prior to fixation and quantification of invasion.

Following University of Iowa Institutional Animal Care and Use Committee approval, 12 male and 12 female nu/J mice (RRID:IMSR_JAX:002019, The Jackson Laboratory) were flank injected with 1 × 10⁶ cells of either parental HCT116 or sub-clone KO-3 suspended in media/Matrigel (BD Biosciences) in a 1:1 ratio. As described previously (29), tumors were allowed to establish for 7 days prior to starting treatments. Mice were randomized to a treatment group prior to flank injection so that each condition and respective treatment group had an equal number of male and female mice.

They received either 40 mg/kg NVP-BKM120 (Selleck Chemicals, catalog no. S2247) or control (water) for 10 consecutive days by oral gavage. Tumor growth was monitored by calipers and tumor volume was calculated in mm³ using the width, length, and height of tumors (16). Animals were euthanized when tumors reached 1,000 mm³. Tumors were harvested, fixed in formalin, and embedded in paraffin. IHC staining for phosphorylated histone H3 (PHH3; Cell Marque, catalog no. 369A-14) was performed and quantified on all harvested tumors.

GraphPad Prism 8.0 (RRID:SCR_002798) was used to complete statistical analysis. Statistical comparisons between length of cell cycle, clonogenicity, protein expression, invasion, and tumor volume were represented as mean ± SEM and compared by Dunn multiple comparison test. Overall survival curves were compared by log-rank test.

Using CRISPR/Cas9 genome editing, clones of HCT116 colon carcinoma cells were generated with complete loss of AP-2α following disruption of exon 3 of the TFAP2A gene. We characterized three clones with KO of TFAP2A, after confirming complete absence of the AP-2α protein by Western blot analysis (Fig. 1A). To determine changes in gene expression following loss of AP-2α, RNA was isolated from the parental HCT116 cell line, KO-1, KO-2, and KO-3, and was analyzed by RNA-seq, comparing gene expression in each TFAP2A KO clone to the parental HCT116 cell line. Differential expression analysis revealed 654 differentially expressed genes in KO-1, 394 differentially expressed genes in KO-2, and 1437 differentially expressed genes in KO-3 when compared with the parental HCT116 cell line. The significantly changed genes from all KO clones were compared and 131 commonly regulated genes were identified (Fig. 1A). Differential expression of selected genes was confirmed by Western blot analysis (Fig. 1A).

Loss of AP-2α alters gene expression through direct chromatin occupancy of regulatory regions of target genes and secondarily through downstream effects. To identify primary AP-2α target genes, ChIP-seq analysis was performed on HCT116, KO-2, and KO-3. Of the 131 genes consistently differentially expressed following complete KO of TFAP2A, 37 genes were identified using peak calling software that had AP-2α occupancy in the HCT116 parental cell line, but no AP-2α occupancy in TFAP2A KO clones (Fig. 1B and C).

Ingenuity Pathway Analysis (IPA) software identified “Cell Cycle Control of Chromosomal Replication” as the most significantly altered pathway with loss of AP-2α (Fig. 1D). To confirm that loss of AP-2α alters cell-cycle progression, parental HCT116 cells, and TFAP2A KO cells both transfected with siRNA to TFAP2A or NT were synchronized in G₁-phase with aphidicolin and allowed to progress through the cell cycle following drug withdrawal (Fig. 2A). Appropriate reduction of AP-2α expression with siTFAP2A was confirmed by Western blot analysis (Supplementary Fig. S1). Parental HCT116 cells expressing AP-2α progressed more rapidly through S-phase and into G₂–M-phase compared with conditions with either knockdown or complete loss of AP-2α (Fig. 2A). In cells with transient knockdown of TFAP2A 4 hours after release from synchronization, only 7% of cells had progressed to in G2–M compared with 22% of cells in HCT116 transfected with NT siRNA. Similarly, in KO-3 clone with KO of TFAP2A transfected with NT siRNA only 14% of cells had progressed to G₂–M-phase. As an additional control, treatment of KO-3 with siRNA targeting TFAP2A had no significant effect on cell-cycle progression (Fig. 2A). As cells were synchronized at the end of G₁-phase, these results were consistent with a prolongation in S-phase after loss of AP-2α.

The prolongation of S-phase with loss of AP-2α was further confirmed on unsynchronized cells through imaging fluorescence. Parental HCT116 and all three KO clones were transduced with the Premo FUCCI Cell Cycle Sensor, which utilizes CDT1-RFP and GMNN-GFP to indicate the start of G₁-phase and S-phase, respectively. Parental HCT116 cells spent a mean of 12.6 ± 7.2 hours in G₁-phase and 11.2 ± 5.4 hours progressing through S-, G₂- and M-phase. All KO clones spent a mean of 8.5 ± 6.0 hours in G₁-phase and 17.3 ± 7.4 hours in S- and G₂-phase until reaching M-phase and dividing confirming that, as predicted, the KO clones spent a significantly longer period of time progressing through S- and G₂-phase compared with parental HCT116 cells expressing AP-2α (P = 0.004; Fig. 2B–D).

The overall length of the cell cycle after loss of AP-2α did not significantly differ from the parental HCT116 cells despite the prolonged duration of S- and G₂-phase, which is likely due to a compensatory shortened G₁-phase (Supplementary Video S1 and S2). To confirm this conclusion, we examined the growth and clonogenicity TFAP2A KO clones compared with parental HCT116. We found no differences in baseline cell growth or clonogenicity after KO of TFAP2A (Supplementary Fig. S2). This was consistent with our findings identifying no difference in the overall length of the cell cycle after loss of AP-2α. In addition, we confirmed that no difference existed in the level of early or late apoptosis after loss of AP-2α as KO clones progressed through cell cycle, as compared with parental HCT116 (Supplementary Fig. S3).

We sought to identify the mechanism by which loss of AP-2α prolongs S-phase. Parental HCT116 and KO-3 clone were compared for alterations in phosphorylation of cell-cycle regulatory proteins. Protein was harvested from parental HCT116 and KO-3 cells 8 hours after release from synchronization of G₁-phase and a phospho-protein microarray examining common cell-cycle regulators was performed. The level of protein and phosphorylated forms of these proteins were analyzed (Fig. 3A). We determined that AKT expression was unaltered by KO of TFAP2A; however, the level of phosphorylation of AKT at Ser473 and Thr308, the two critical sites of phosphorylation activating AKT, were significantly reduced with KO of TFAP2A (Fig. 3B and C). Likewise, phosphorylation of Ser124 of AKT reduced, but did not reach statistical significance (Fig. 3C). In addition, one of the downstream targets of AKT, CDC25A, showed reduced phosphorylation at Ser124 with KO of TFAP2A, consistent with decreased activity downstream of the AKT cascade after loss of AP-2α. To further confirm these findings, Western blot analysis of unsynchronized parental HCT116 and TFAP2A KO clones revealed similar overall expression of AKT, but decreased phosphorylation of AKT at Ser473 and Thr308 with KO of TFAP2A (Fig. 3D). A similar decreased level of phosphorylation of AKT was observed with transient knockdown of TFAP2A in the colon cancer cell line, LoVo, in unsynchronized cells (Fig. 3E). AKT is phosphorylated at Thr308 by PDK1 through the activation of the PI3K cascade (30) and at Ser473 by mTOR (31). The decreased level of phosphorylation after loss of AP-2α, both at baseline and as cells actively progress through the cell cycle, suggests a disruption of the PI3K/mTOR/AKT pathway.

The activation of the PI3K/AKT cascade is often driven by the presence of mutations in PIK3CA, encoding the catalytic subunit of PI3K. The colon cancer cell line HCT116 has a PIK3CA gene mutation c.3140A>G causing the missense mutation H1047R leading to its constitutive activation (32). The mutation status of all three AP-2α KO clones was examined and was unchanged compared with the parental cell line (data not shown). We subsequently sought to identify further mechanisms for the alteration of PI3K/AKT cascade activity.

We identified the protein tissue transglutaminase (TGM2) as a potential mechanism for how AP-2α affects the activity of the PI3K/AKT cascade. TGM2, which is known to form a complex with PI3K, leading to its activation (33–35), was one of the 131 common genes whose level of expression was altered after loss of AP-2α. The level of TGM2 RNA and protein was decreased in KO-1 and KO-2 compared parental HCT116 (Fig. 1) and ChIP-seq identified direct promoter occupancy by AP-2α in regulatory regions of the TGM2 gene (Fig. 4A), confirmed by the presence of multiple AP-2α consensus sequences at the sites of observed AP-2α occupancy (Supplementary Fig. S4).

We characterized the effect of TGM2 on cell cycle (Fig. 4B–D). TGM2 was transiently knocked down with siRNA in parental HCT116 and KO clone KO-2. Unsynchronized cells at similar levels of confluency were transduced with the Premo FUCCI Cell Cycle Sensor, which utilizes CDT1-RFP and GMNN-GFP to fluorescently indicate the current phase of cell cycle.

Cells in G₁-phase were identified by the sole expression of CDT1-RFP and cells in S/G₂–M-phase expressed either GMNN-GFP alone or GMNN-GFP + CDT1-RFP. The proportion of cells in S/G₂–M-phase compared with G₁-phase was significantly higher after transient knockdown of TGM2 in parental HCT116 cells compared with cells transfected with siRNA to nontargeting (siNT; Fig. 4C and D). As expected, there was no effect in the KO clone KO-2 with knockdown of TGM2. This supports the hypothesis that activation of TGM2 may be one of the mechanisms through which AP-2α regulates the activity of the PI3K/AKT cascade and subsequently cell-cycle progression (Fig. 4E). To determine whether a correlation exists between TFAP2A and TGM2 expression in clinical cancer samples, expression of these genes was examined in published databases. A positive correlation was observed in the GDC Pan-Cancer database between TGM2 and TFAP2A expression (Fig. 4F). This correlation was supported using a colon cancer TMA, which identified 32% of tumors as expressing AP-2α, with a positive correlative trend between AP-2α and TGM2 expression in AP-2α–expressing tumors (Fig. 4G and H; Supplementary Fig. S5).

Given the alteration in PI3K/AKT activity observed after loss of AP-2α expression, we predicted that KO of TFAP2A will alter the response to a PI3K inhibitor. To test this hypothesis, we utilized the highly selective PI3K inhibitor, BKM120, also known as Buparlisib. The effect of BKM120 on the invasiveness of cells was tested in vitro using an invasion assay. After 24 hours of in vitro treatment, the parental HCT116 cells exposed to 250 nmol/L BKM120 exhibited significantly less invasion compared with vehicle-treated cells (Fig. 5A and B). All three TFAP2A KO clones did not demonstrate any significant change in invasion with 250 nmol/L BKM120 treatment compared with vehicle, suggesting that loss of AP-2α in colon cancer may confer resistance to this highly selective PI3K inhibitor.

We confirmed the effect of loss of AP-2α expression on response to BKM120 in two additional conditions. First, we performed an invasion assay which confirmed that stable knockdown of TFAP2A in HCT116 with shRNA conferred resistance to BKM120 (Fig. 5C and D). Second, we analyzed the effect of stable knockdown of TFAP2A using shRNA transduction in a panel of colon cancer cell lines. On invasion assay, BKM120-treated cells transduced with short hairpin RNA nontargeting (shNT) demonstrated decreased invasion compared with those treated with vehicle, whereas loss of AP-2α in the panel of colon cancer cell lines conferred resistance to BKM120 (Fig. 5C and D).

To further test the efficacy of BKM120 after loss of AP-2α, xenografts were established by inoculating immunocompromised Nu/J mice with either parental HCT116 cells or KO-3 in the subcutaneous tissue of their flanks. Tumors were allowed to establish over 7 days and mice were randomly assigned to be gavaged daily for 10 days with either vehicle or 40 mg/kg of BKM120. Tumors in mice with parental HCT116 xenografts treated with BKM120 demonstrated significantly reduced tumor growth (Fig. 6A) and mice had longer disease-specific survival (Fig. 6B) compared with vehicle-treated mice. Interestingly, mice inoculated with KO-3 had no difference in either tumor growth or tumor-specific survival when treated with BKM120 as compared with vehicle-treated mice (Fig. 6A and C). Tumors from all mice were analyzed for mitotic activity via IHC for PHH3, which is known to decrease in vivo with PI3K inhibition (30). The number of PHH3-positive cells was significantly lower in BKM120-treated parental HCT116 tumors compared with vehicle-treated tumors; however, there were no differences in the number of cells expressing PHH3 in KO-3 tumors that were treated with BKM120 as compared with vehicle-treated tumors (Fig. 6D and E). It is worth commenting that KO-3 xenografts did grow at a significantly slower rate than parental HCT116 and had a lower number of PHH3-positive cells (Fig. 6A and D). This potentially may be reflective of the baseline change due to decreased PI3K/AKT activity after loss of AP-2α and that treatment with BKM120 may decrease the level of PI3K/AKT activity to that of the KO clone. These in vivo data further support that expression of AP-2α is a marker of BKM120 sensitivity.

Consistent with our findings, there is further evidence that TFAP2A expression correlates with sensitivity to BKM120 treatment in several cancer cell types. Analysis of RNA expression data in BKM120-sensitive versus BKM120-resistant cells from published GEO Databases (GSE69405, GSE98824, GSE49416), we found a statistically significant relationship between TFAP2A expression and response to BKM120 in lung adenocarcinoma (Supplementary Fig. S6A). Furthermore, expression of TGM2 was also significantly predictive of response to BKM120. Similarly, we noted a trend in the relationship between TFAP2A and TGM2 expression and sensitivity to BKM120 in triple-negative breast cancer, though this relationship did not reach statistical significance (Supplementary Fig. S6A). We performed a transient knockdown of TFAP2A in a triple-negative breast cancer cell line which demonstrated a loss of BKM120 sensitivity with knockdown of TFAP2A expression (Supplementary Fig. S6B). Similar trends in the relationship between TFAP2A and TGM2 expression and BKM120 sensitivity were also seen in glioblastoma (Supplementary Fig. S6A).

Previous studies have characterized changes in cell-cycle progression after overexpression of AP-2α (25, 36, 37), and our findings both support these findings and add to our understanding of how AP-2α effects cell-cycle progression. We identify a novel pathway by which AP-2α regulates the cell cycle and we have shown that the effect of AP-2α upon cell-cycle regulation through the PI3K/AKT cascade alters the response to the highly specific PI3K inhibitor, BKM120.

Previous studies in the HCT116 colon cancer cell line demonstrated that AP-2α induced cell-cycle arrest through p53-dependent activation of CDKN1A/p21 (25, 38), and may also function in a p53-independent mechanism (38). Similar mechanisms of CDKN1A/p21 gene regulation by AP-2α have been demonstrated in breast, cervical carcinoma, hepatoblastoma, lung carcinoma, and other colon cancer cell lines (14, 36, 38, 39). Overexpression of AP-2α is also capable of inducing growth arrest and apoptosis characterized by hyperphosphorylation of Rb and cleavage of PARP (38). Although it is unlikely to be the sole pathway through which AP-2α alters cell-cycle progression, the current study identified TGM2 as a primary target of AP-2α and a potential mechanism through which AP-2α alters the PI3K/AKT cascade. The previous studies and our current data support the finding that AP-2α plays a clear role in the progression of cancer cells through S-phase.

The role of AP-2α in the regulation of the PI3K/AKT cascade is relatively unexplored, although there are previous data that support this relationship. A study by Fertig and colleagues utilized the LINCS (library of integrated network-based cellular signatures) database and identified that AP-2α gene expression signatures were increased with the inhibition of MEK, PI3K, and mTOR pathways indicating a positive feedback mechanism (40). In addition, in a previous study knockdown of TFAP2C, encoding the AP-2γ family member, resulted in a decrease in AKT activation in breast cancer indicating that other AP-2 factors may have similar activity (41). Retinoic acid has been shown to repress the PI3K/AKT pathway in several cell types (42–44), and can induce expression of TFAP2A (45, 46); however, the role of AP-2α in response to retinoic acid was not examined in these studies. In addition, a study in mouse muscle myoblasts reported that AP-2α can repress activation of AKT through miR-25-3p, suggesting competing effects in different cell types that may be influenced by miR expression (47).

Recent data on the use of BKM120 in clinical trials in patients with colorectal cancer has highlighted the need for the identification of biomarkers to predict response (48). Identifying biomarkers for sensitivity to BKM120 will expand and improve patient directed therapy. This will allow for the early identification of patients who will receive the most benefit from BKM120 and avoid unnecessary toxicity in patients unlikely to demonstrate a favorable response. This in turn will help to establish the true efficacy of BKM120 in treating colon cancer. Previous studies have identified that, although BKM120 was efficacious in multiple colon cancer cell lines, it exhibited a greater level of apoptosis in those cell lines with a PI3KCA mutation (49). PIK3CA is one of the most commonly mutated genes in cancer (5% of cancers) and individualized treatment based on mutational profiles and in vitro drug testing are becoming more common methods to guide therapy (50). Our study identified that loss of AP-2α conferred resistance to PI3K inhibition in the presence of a PIK3CA mutation and expression of AP-2α may serve as a biomarker for response in other cancer types.

In conclusion, we identified that loss of AP-2α results in prolongation of S-phase, which is mediated through the PI3K cascade and activation of AKT. In addition, TGM2 is an AP-2α target gene that influences activation of the PI3K/AKT pathway. The role of AP-2α in regulating the PI3K/AKT cascade in colon cancer is of particular importance as we identified that AP-2α is a marker of sensitivity to the highly selective PI3K inhibitor, BKM120, which is currently being used in multiple clinical trials. Therefore, AP-2α may be utilized as a biomarker for identifying patients likely to benefit from BKM120 therapy.

A.C. Beck reports grants from NIH during the conduct of the study. E. Cho reports grants from T32 NIH during the conduct of the study. D.T. Thompson reports grants from NIH during the conduct of the study. C. Franke reports grants from NIH during the conduct of the study. R.J. Weigel reports grants from NIH during the conduct of the study. No disclosures were reported by the other authors.

A.C. Beck: Conceptualization, formal analysis, supervision, investigation, writing–original draft, writing–review and editing. E. Cho: Validation, investigation, writing–review and editing. J.R. White: Conceptualization, formal analysis, validation, investigation, writing–review and editing. L. Paemka: Conceptualization, validation, investigation, writing–review and editing. T. Li: Software, formal analysis, methodology, writing–review and editing. V.W. Gu: Formal analysis, investigation, writing–review and editing. D.T. Thompson: Software, formal analysis, validation, investigation, writing–review and editing. K.E. Koch: Formal analysis, investigation, writing–review and editing. C. Franke: Validation, investigation, writing–review and editing. M. Gosse: Resources, formal analysis, investigation, visualization, methodology, writing–review and editing. V.T. Wu: Formal analysis, validation, investigation, writing–review and editing. S.R. Landers: Investigation, writing–review and editing. A.J. Pamatmat: Investigation, writing–review and editing. M.V. Kulak: Conceptualization, formal analysis, supervision, validation, investigation, writing–review and editing. R.J. Weigel: Conceptualization, resources, formal analysis, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing

The data presented herein were obtained at the Flow Cytometry Facility, which is a Carver College of Medicine/Holden Comprehensive Cancer Center core research facility at the University of Iowa (Iowa City, IA). The facility is funded through user fees and the generous financial support of the Carver College of Medicine, Holden Comprehensive Cancer Center, and Iowa City Veteran's Administration Medical Center. We thank Alejandro A Pezzulo at the University of Iowa (Iowa City, IA) for his assistance in analysis allowing for quantification of fluorescent imaging, and Andrew Bellizzi and Anand Rajan for assistance with IHC staining. In addition, the authors would like to acknowledge use of the University of Iowa Central Microscopy Research Facility, a core resource supported by the University of Iowa Vice President for Research, and the Carver College of Medicine. This work was supported by the NIH grants R01CA183702 (PI: R.J. Weigel) and T32CA148062 (PI: R.J. Weigel). A.C. Beck, E. Cho, D.T. Thompson, K.E. Koch, and V.T. Wu were supported by the NIH grant T32CA148062.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.