Alternative Splicing of the Aryl Hydrocarbon Receptor Nuclear Translocator (ARNT) Is Regulated by RBFOX2 in Lymphoid Malignancies

ABSTRACT

Aberrant alternative splicing (AS) of pre-mRNAs promotes the development and proliferation of cancerous cells. Accordingly, we had previously observed higher levels of the aryl hydrocarbon receptor nuclear translocator (ARNT) spliced variant isoform 1 in human lymphoid malignancies compared to that in normal lymphoid cells, which is a consequence of increased inclusion of alternative exon 5. ARNT is a transcription factor that has been implicated in the survival of various cancers. Notably, we found that ARNT isoform 1 promoted the growth and survival of lymphoid malignancies, but the regulatory mechanism controlling ARNT AS is unclear. Here, we report cis- and trans-regulatory elements which are important for the inclusion of ARNT exon 5. Specifically, we identified recognition motifs for the RNA-binding protein RBFOX2, which are required for RBFOX2-mediated exon 5 inclusion. RBFOX2 upregulation was observed in lymphoid malignancies, correlating with the observed increase in ARNT exon 5 inclusion. Moreover, suppression of RBFOX2 significantly reduced ARNT isoform 1 levels and cell growth. These observations reveal RBFOX2 as a critical regulator of ARNT AS in lymphoid malignancies and suggest that blocking the ARNT-specific RBFOX2 motifs to decrease ARNT isoform 1 levels is a viable option for targeting the growth of lymphoid malignancies.

INTRODUCTION

The aryl hydrocarbon receptor nuclear translocator (ARNT), also known as hypoxia-inducible factor-1β (HIF-1β), is a member of the basic-helix-loop-helix/Per-ARNT-Sim superfamily of transcription factors, which act as sensors of environmental or physiological insults and are associated with the initiation and progression of cancerous cells (1–3). Specifically, ARNT has been implicated in chemotherapeutic resistance in cancer cells through the depletion of reactive oxygen species (4–7). Conversely, ARNT has also been reported to hinder tumor migration and invasion (8). These observations suggest that targeting ARNT has potential for treating drug-resistant cancer but could lead to undesirable effects such as increased metastasis. Thus, reconciling these divergent functions of ARNT in cancer pathogenesis will be critical for developing ARNT-based therapeutic strategies. In consideration of the different roles ascribed to ARNT, it is important to note that ARNT is alternatively spliced to produce two main isoforms, ARNT isoform 1 and ARNT isoform 3, which differ by the inclusion or exclusion of exon 5, respectively (9). Moreover, we have previously reported that increased levels of ARNT isoform 1 in diverse lymphoid malignancies correspond to better growth and higher thresholds for killing with doxorubicin (10). Interestingly, suppression of total ARNT (i.e., both isoforms) conveyed a greater growth rate compared to that of cells with high ARNT isoform 1 levels (10). Thus, targeting the specific ARNT isoforms may lead to a more promising therapeutic approach. While we have illuminated the differential roles of the ARNT isoforms in regulating the growth and survival of lymphoid malignancies, there is a lack of understanding on the mechanism(s) controlling ARNT alternative splicing (AS).

AS is regulated by trans-acting factors (RNA-binding proteins) binding to cis-regulatory sequences (RNA-binding motifs), which in turn mediate alternative splicing, and yields multiple mature RNA isoforms from the same pre-mRNA molecule (11). AS is essential for maintaining many physiological processes, but its dysregulation can contribute to cancer pathogenesis (12, 13). In fact, the Cancer Genome Atlas reports that malignant tissues display approximately 30% more alternative splicing events compared to non-malignant tissues (14), which have been attributed to mutations in, and/or expression levels of, genes encoding splicing regulators (15, 16). For instance, members of the RNA-binding Fox (RBFOX) protein family, consisting of RBFOX1, RBFOX2, and RBFOX3, are master regulators of AS due to their interactions with the intronic GCAUG sequence motifs which flank alternative exons in pre-mRNA (17, 18). Although each family member of the RBFOX family recognizes the same RNA motif, they differ in their tissue specific expression patterns. RBFOX1 is predominately expressed in the brain, heart, and ovaries, whereas RBFOX3 is expressed almost exclusively in the brain (19–23). RBFOX2 is more widely expressed than other family members and is upregulated in diverse cancers (24–26). Specifically, RBFOX2 contributes to cellular invasion, is highly expressed in cells undergoing epithelial-mesenchymal transition, and appears to be important for maintaining a mesenchymal phenotype (27–29). In addition, RBFOX2 is highly expressed in various B and T cell malignancies (30).

Given our observations of ARNT isoform 1-dependent proliferation of lymphoid malignancies, and the possible correlation with higher RBFOX2 expression in malignant cells, we hypothesized that RBFOX2 regulates ARNT AS. In this study, we identified RBFOX2 binding motifs in ARNT introns 4 and 5, which flank exon 5. We demonstrate, using an ARNT-specific minigene, that RBFOX2 recognizes both a proximal and a distal motif in ARNT intron 5 to promote exon 5 inclusion. Overall, our data indicate that RBFOX2, an RNA-binding protein commonly upregulated in malignant cells, contributes to the aberrant AS of ARNT in lymphoid malignancies.

RESULTS

RBFOX2 binding sites in ARNT introns 4 and 5 correlate with RBFOX2-mediated ARNT exon 5 inclusion.

To gain insight into the regulatory mechanism(s) controlling ARNT AS, we analyzed the intronic sequences flanking exon 5 for RNA binding motifs and observed multiple GCAUG motifs which are recognized by the RBFOX protein family (Fig. 1A). Next, we used RBFOX2-specific enhanced cross-linking and immunoprecipitation (eCLIP) data from K652 cells, a chronic myelogenous leukemia cell line, found within the Encyclopedia of DNA Elements (ENCODE) database. In alignment with the observed GCAUG motifs, we identified RBFOX2 binding peaks in ARNT introns 4 and 5, which flank exon 5 (Fig. 1B) (31). To determine whether RBFOX2 regulates ARNT AS, we depleted RBFOX2 in human embryonic kidney (HEK)293 cells. As a control, other RNA-binding proteins were also suppressed, including PTBP1/2 and CELF1, which do not display binding sites within ARNT introns 4 and 5. Notably, suppression of RBFOX2 significantly affected the inclusion of ARNT exon 5 (Fig. 1C), consistent with the presence of RBFOX2 binding sites in the introns flanking alternative exon 5; however, suppression of PTBP1/2 or CELF1 did not affect ARNT AS. We further assessed RBFOX2-mediated regulation of ARNT AS during murine heart development, because RBFOX2 protein levels are low at birth but higher in adulthood (32). During murine heart development, inclusion of ARNT exon 5 correlated with expected RBFOX2 levels (Fig. 1D). Together, these observations from human cells and mouse tissue point to RBFOX2 as a critical regulator of ARNT AS.

FIG 1.

RBFOX2 binding sites in ARNT introns 4 and 5 correlate with RBFOX2-mediated ARNT exon 5 inclusion. (A) Schematic showing RBFOX2 binding motifs “GCAUG” in the upstream and downstream introns of ARNT exon 5. (B) RBFOX2 binding peaks in introns flanking ARNT exon 5 from two independent eCLIP experiments (rep1 and rep2), as found in the ENCODE database (GEO: GSE92030). (C) Representative reverse-transcription PCR (RT-PCR) analysis of ARNT AS in HEK293 cells after introduction of the indicated siRNAs (top), and combined densitometry analysis to determine percent spliced in (PSI) of three independent experiments performed in duplicate (bottom). (D) Representative RT-PCR analysis of ARNT AS in murine heart tissue at the indicated developmental stage (top), and combined densitometry analysis to determine PSI of n = 3 (E18 and Newborn) or n = 4 (Adult) (bottom). Data in graphs are presented as means ± standard error of the mean (SEM). E5 = exon 5. P values are derived using a two-tailed unpaired Student’s t test. *, P < 0.05; **, P < 0.01; ****, P < 0.0001.

RBFOX2 promotes ARNT exon 5 inclusion through recognition of cis-regulatory elements in ARNT intron 5.

Both the upstream and downstream introns of ARNT exon 5 harbor multiple GCAUG motifs, and eCLIP validated RBFOX2 binding clusters within these regions. The position activity profile of RBFOX2 proteins, obtained through mechanism-based studies, identified that binding upstream from an exon promotes exclusion, whereas downstream binding promotes exon inclusion (18, 33). Thus, we focused on the GCAUG motifs in the downstream intron, given our observed RBFOX2-dependent inclusion of ARNT exon 5 (Fig. 1C and D). To map the probable RBFOX2 binding motifs necessary for exon 5 inclusion, we constructed an ARNT minigene splicing vector consisting of ARNT exon 5 and flanking introns. Next, we deleted the proximal RBFOX2-recognition motif (ARNTΔp), the distal motif (ARNTΔd), or both the proximal and distal motifs together (ARNTΔpd) (Fig. 2A). We then transfected the wild-type (WT) and the three mutant ARNT splicing vectors, with or without an RBFOX2-expressing plasmid, into HEK293T cells. As a negative control, we transfected the WT ARNT splicing vector together with a nonfunctional (NF) RBFOX2-expressing plasmid which was mutated at the RNA recognition motif so that RBFOX2 could not bind to RNA targets. As expected, we observed that the co-transfection of the WT ARNT minigene with the RBFOX2-expressing plasmid promoted significant inclusion of exon 5, resulting in a percent spliced in (PSI) of 65% compared to 15% with an empty vector or NF RBFOX2-expressing plasmid (Fig. 2B and C). However, deletion of the proximal binding motif, ARNTΔp, in the presence of RBFOX2 resulted in a significant decrease in ARNT exon 5 inclusion (∼30% PSI), compared to the WT ARNT splicing vector. Similarly, deletion of the distal binding motif, ARNTΔd, also significantly hindered RBFOX2-mediated inclusion of ARNT exon 5 (∼20% PSI), compared to the WT ARNT splicing vector (Fig. 2B). Notably, deleting both RBFOX2 binding motifs, as in ARNTΔpd, resulted in a significant 50% decrease of exon 5 PSI compared to the WT ARNT splicing vector, suggesting that the function of RBFOX2 toward exon 5 inclusion is additive (Fig. 2B). In fact, the ARNTΔpd mutation brought exon 5 inclusion to background levels, suggesting that both the proximal and distal GCAUG motifs in ARNT intron 5 are essential for RBFOX2-mediated inclusion of ARNT exon 5 into the final mRNA.

FIG 2.

RBFOX2 promotes ARNT exon 5 inclusion through recognition of cis-regulatory elements in ARNT intron 5. (A) Schematic of ARNT-specific splicing vector with the indicated GCATG motifs that were targeted for deletion. (B) HEK293T cells were transfected with wild-type (WT) ARNT splicing vector or a GCATG deletion mutant ARNT splicing vector (ARNTΔp, ARNTΔd, or ARNTΔpd) plus empty vector control or WT RBFOX2. A nonfunctional (NF) RBFOX2 mutant was used as a negative control. Total RNA was isolated at 72 h posttransfection and analyzed by RT-PCR for splicing changes; a representative image is shown below the graph. The densitometry analysis to determine percent spliced in (PSI) results presented in the graph are means ± SEM of three independent experiments performed in triplicate. P values are derived using a two-tailed unpaired Student’s t test. *, P < 0.05; ***, P < 0.001; ****, P < 0.0001 compared to empty vector control; †††, P < 0.001; ††††, P < 0.0001 compared to WT ARNT + WT RBFOX2. E5 = exon 5. (C) Immunoblot analysis of cells from panel B, showing equal levels of ectopically expressed FLAG-RBFOX2 between samples.

RBFOX2 promotes the inclusion of ARNT exon 5, and higher ARNT isoform 1 levels, in lymphoid malignancies.

To investigate whether a direct relationship exists between RBFOX2 and the increased ARNT isoform 1 levels that we previously observed in lymphoid malignancies (10), we first compared RBFOX2 expression between primary human T cells and two malignant human T cell lines: Karpas 299 cells, an anaplastic large cell lymphoma (ALCL) cell line; and Jurkat cells, an acute T cell leukemia cell line. Remarkably, RBFOX2 expression was upregulated in the malignant T cells compared to that in primary T cells, as assessed by real-time quantitative PCR (RT-qPCR) and reverse-transcription PCR (RT-PCR) analysis (Fig. 3A). Next, we suppressed RBFOX2 in both malignant T cells and observed a significant decrease in ARNT exon 5 inclusion at the transcription and protein levels (Fig. 3B and C). Importantly, the loss of RBFOX2 significantly reduced the growth of Karpas 299 cells in proportion to the loss of ARNT isoform 1 (Fig. 3D). These observations demonstrate that RBFOX2 is upregulated in malignant T cells and acts as a positive mediator of ARNT exon 5 inclusion, resulting in the predominant generation of ARNT isoform 1, which imparts an increased growth advantage.

FIG 3.

RBFOX2 promotes the inclusion of ARNT exon 5, and higher ARNT isoform 1 levels, for optimal growth of lymphoid malignancies. (A) RT-qPCR analysis of RBFOX2 expression in primary T cells, Karpas 299 cells, and Jurkat cells (left), with representative RT-PCR images to show relative expression levels of ARNT and RBFOX2 (right). (B) Validation of RNAi-mediated RBFOX2 suppression by RT-qPCR analysis (left) and immunoblotting with antibodies against RBFOX2 and β-actin (right), in Karpas 299 and Jurkat cells. RT-qPCR data are means ± SEM of three independent experiments performed in triplicate. P values are derived using a two-tailed unpaired Student’s t test. **, P < 0.01; ***, P < 0.001. (C) Representative RT-PCR analysis of ARNT AS in Karpas 299 and Jurkat cells after introduction of the indicated siRNAs (top), percent spliced in (PSI) as determined by combined densitometry analysis (data are means ± SEM of three independent experiments) (middle), and immunoblot analysis using antibodies against ARNT and β-actin (bottom). E5 = exon 5. (D) Twenty-four hours post-electroporation with siControl, siRBFOX2, or siARNT-1, Karpas 299 cells were normalized to 1.7 × 10⁵ cells/mL and counted every 24 h for a total of 72 h (left). P values are derived using two-way analysis of variance. *, P < 0.05; **, P < 0.01; ***, P < 0.001 compared to siControl. †, P < 0.05; ††, P < 0.01 compared to siRBFOX2. Immunoblot analysis was performed at 48 h posttransfection using antibodies against RBFOX2, ARNT, and β-actin (right).

DISCUSSION

Aberrant AS plays a significant role in promoting tumorigenesis and proliferation of cancerous cells. The transcription factor ARNT has been implicated in the development of cancerous cells (34, 35), and we have previously shown that, compared to normal lymphoid cells, ARNT is alternatively spliced in lymphoid malignancies so that ARNT isoform 1 is the predominant isoform (10, 34, 35). Subsequently, ARNT isoform 1 promotes the growth of the lymphoid cancers and resistance to the chemotherapeutic drug doxorubicin (10). In the current study, our goal was to elucidate the regulatory factor(s) responsible for promoting ARNT AS. Accordingly, by utilizing available eCLIP data from the ENCODE database, analyzing the cis-regulatory motifs in the intronic regions flanking ARNT exon 5, and suppressing certain RNA-binding proteins via RNA interference (RNAi), we were able to identify RBFOX2 as a key regulator of ARNT AS. Moreover, we confirmed that RBFOX2 is upregulated in human T cell-derived ALCL and leukemia cells and promotes the inclusion of ARNT exon 5 through recognition of proximal and distal cis-regulatory motifs in ARNT intron 5. Combined, our data support a cell-proliferative role for RBFOX2 through the promotion of increased ARNT isoform 1 levels.

Commonly, AS studies focus within the 500-nucleotide intronic regions flanking alternative exons to analyze the cis-regulatory motifs and trans-acting factors which bind to these sequences (36). However, studies have shown that RNA-binding proteins have important binding sites located much further beyond this arbitrary 500-nucleotide boundary (37) and are recruited closer to the alternative exon through an “RNA bridge” to enhance spliceosome activity (38). RNA bridges are created through RNA secondary-structure changes which dramatically shorten the distance of a distal cis-regulatory motif to its target exon. Thus, we predict that the proximal and distal GCAUG motifs downstream from ARNT exon 5 are brought into proximity to allow for optimal RBFOX2-mediated ARNT exon 5 inclusion, possibly through an RNA bridge, as previously demonstrated for RBFOX2 (38). Our data further suggest that analyzing cis-regulatory sequences outside of the commonly used 500-nucleotide intronic regions is an important consideration when investigating exon usage regulation in AS studies.

Although this study focused specifically on RBFOX2 regulation of ARNT AS in lymphoid malignancies, RBFOX2 is known to be upregulated in many other cancer cell types, such as breast and ovarian cancers (29). Interestingly, RBFOX2 expression is associated with promoting the progression of epithelial to mesenchymal cell transition by regulating the AS of genes important for tissue invasion (27, 29). Notably, ARNT AS was found to be more affected in mesenchymal tissues versus epithelial tissues, correlating with RBFOX2 expression levels (29). Thus, the ARNT isoforms may play a pivotal role in regulating the growth and invasion of non-lymphoid cancerous cells. Together, our observations suggest that targeting the splicing of ARNT by blocking RBFOX2 from binding to the ARNT pre-mRNA provides a more specific strategy for cancer therapeutics.

MATERIALS AND METHODS

Cell culture.

HEK293 and HEK293T cells were cultured in Dulbecco’s Modified Eagle medium (DMEM; Corning, 15-018-CV) complete with 10% fetal bovine serum (FBS; Atlanta Biologicals, S11550) and 2 mM GlutaMAX (GIBCO, 35050-061) at 37°C, 5% CO₂. Karpas 299 and Jurkat cells were cultured in RPMI 1640 medium (Corning, 15-041-CV) complete with 10% FBS and 2 mM GlutaMAX at 37°C, 5% CO₂.

Mice.

All animal experiments were approved by the University of Texas Medical Branch Institutional Animal Care and Use Committee (protocol no. 1101001) and were conducted in accordance with NIH guidelines. For analysis of ARNT AS during development, C57BL6 (Jackson Laboratory) mice were purchased and timed matings were conducted. At embryonic day 18 (E18), the day of birth (Newborn), and 6 months of age (Adult), mice were terminated, and heart tissue was collected for RNA extraction using TRIzol Reagent (Invitrogen, 15596).

Plasmids.

The p3×FLAG-RBFOX2 plasmid has been previously described (39). The control empty vector was the p3×FLAG plasmid (Millipore Sigma, E7408). The ARNT minigene splicing reporter was constructed as follows: pcDNA5-GFP-IL7R splicing reporter (40) was digested with ApaI and BamHI (New England BioLabs [NEB]) and purified with a PCR purification kit (Qiagen). The genomic sequence of ARNT exon 5 and the upstream and downstream introns were ligated into the vector using T4 DNA ligase (NEB). Site-directed mutagenesis was performed to delete the indicated GCATG motifs using a Quick-change Lightning mutagenesis kit (Agilent, 210518). Primers were designed using the Agilent primer design tool and the sequences used were as follows: ARNTΔp, sense 5′-GTA GGA GAA CAG TGT CTT TTG AAG CAG ATG ATG CTG CT-3′ and nonsense 5′-AG CAG CAT CAT CTG CTT CAA AAG ACA CTG TTC TCC TAC-3′; ARNTΔd, sense 5′-GAG TAG AAG CAG CTC CTT GAG GCA TCC-3′ and nonsense 5′-GGA TGC CTC AAG GAG CTG CTT CTA CTC-3′. Resultant deletions were confirmed by DNA sequencing.

Transfections.

HEK293 cells were transfected with 10 μM Control scrambled siRNA (Invitrogen AM4611), a pool of 10 μM CELF1-targeting siRNAs (Invitrogen siRNA ID no. s168616 and s168617), 10 μM human PTBP1/PTBP2-targeting siRNAs (Qiagen, SI02649206 and SI04255146), or a RBFOX2-specific siRNA (Invitrogen siRNA IDnumber s96620), using Lipofectamine RNAiMAX (Invitrogen, 13778100), and were harvested 72 h later. HEK293T cells were transfected with 100 ng ARNT minigene splicing vector, and 250 ng of either empty vector or FLAG-tagged RBFOX2, using a standard calcium-phosphate transfection protocol. DMEM was replaced 7 h later, and cells were harvested after 72 h. Karpas 299 and Jurkat cells were transfected with 4 μM RBFOX2-targeting siRNA (5′-GUU CGU AAC UUU CGA GAA U-3′; Millipore Sigma), 4 μM Control scrambled siRNA (5′-CAU GCC UUG CUU UAC GCA U-3′; Millipore Sigma), or 4 μM ARNT isoform 1-targeting siRNA (10) using a Bio-Rad Gene Pulser Xcell electroporator set on infinite resistance, 950 microfarads, and 300 V for Karpas 299 cells or 250 V for Jurkat cells. Sixteen hours posttransfection, dead cells were removed with Ficoll-Paque (GE Healthcare, 17-1440-02) by centrifugation at 400 × g for 20 min. The live cell layer was collected, washed with phosphate-buffered saline, resuspended in RPMI medium at 0.5 × 10⁶ cells/mL, and propagated for an additional 32 h.

RT-PCR analysis of RNA from human cell culture.

Total RNA was extracted using a QIAshredder (Qiagen, 79656) and RNeasy columns (Qiagen, 74104). Two μg of total RNA was converted to cDNA using the iScript cDNA synthesis kit (Bio-Rad, 1706691). PCR primers were designed to detect inclusion/exclusion (185-bp product and 140-bp product, respectively) of human ARNT exon 5: forward 5′-GCG ATG ATG ACC AGA TGT GT-3′ and reverse 5′-GGC CAT GCG TAA GAT GGT TA-3′. The PCR (20 μL total) included 5 μL of cDNA, 25 μM dNTPs, and 100 ng of each ARNT-specific forward and reverse primer with Biolase Taq polymerase (Bioline BIO-21042). PCR conditions for each reaction were as follows: 95°C, 45 sec; 59°C, 45 sec; 72°C, 1 min for 25 cycles. PCR products were resolved on 5% non-denaturing polyacrylamide gels and stained with ethidium bromide. The bands were imaged using the ChemiDoc MP Imaging system (Bio-Rad). Unsaturated band intensities were quantified after normalizing the ethidium staining to the size of the included and excluded bands, and percent exon inclusion was calculated (i.e., percent spliced in [PSI]) using Image Lab software (Bio-Rad) with the following equation: PSI = ([exon inclusion band]/[exon inclusion band + exon exclusion band]*100).

RT-PCR analysis of RNA from mouse heart tissue.

One μg of total RNA was incubated with 125 ng of oligo(dT) primer at 65°C for 10 min followed by cDNA synthesis using AMV reverse transcriptase (20 units/μg; Life Biosciences, AMV 007-1) in the presence of 10 μM dNTPs (20-μL reaction) at 42°C for 1 h. PCR primers were designed to detect inclusion/exclusion (221-bp product versus 176-bp product, respectively) of murine ARNT exon 5: forward 5′-GCG ATG ATG ACC AGA TGT GT-3′ and reverse 5′-GGC CAT GCG TAA GAT GGT TA-3′. The PCR (20 μL total) included 5 μL of cDNA, 25 μM dNTPs, 100 ng of each ARNT-specific forward and reverse primer, and Biolase Taq polymerase (Bioline BIO-21042). PCR conditions for each reaction were as follows: 94°C, 45 sec; 59°C, 45 sec; 72°C, 1 min for 25 cycles. PCR products were resolved on 5% non-denaturing polyacrylamide gels and stained with ethidium bromide. The DNA bands were imaged and analyzed as described above.

Real-time quantitative PCR analysis and cell growth assay.

For this analysis, 300 ng of total RNA was converted to cDNA using the iScript cDNA synthesis kit (Bio-Rad, 1706691). The CFX96 Real Time System (Bio-Rad) was used to perform quantitative PCR (qPCR). The qPCR (20 μL total) included 1 μL of cDNA, SsoAdvanced Universal Probes Supermix (Bio-Rad, 1725284), and the Homo sapiens primer-probe assay RBFOX2 (Bio-Rad, qHsa CIP0027070) or GAPDH (Bio-Rad, qHsa CEP0041396) as the internal reference gene used to determine fold changes with the threshold cycle (ΔC_T) method. The cell growth assay was performed as previously described (10).

Antibodies and immunoblot analysis.

Whole-cell lysates from HEK293T, Karpas 299, and Jurkat cells were extracted by incubating cells with 1 × radioimmunoprecipitation assay (RIPA) lysis buffer (Cell Signaling, 9806S) complete with phosphatase inhibitor cocktail 2 (Millipore Sigma, P5726), phosphatase inhibitor cocktail 3 (Millipore Sigma, P0044), complete mini protease inhibitor tablets (Roche, 11836170001), and 50 mM sodium fluoride (Millipore Sigma, S1504-110G) on ice for 20 min. Protein samples were normalized using the Bradford assay, resolved on NuPAGE 4 to 12% Bis-Tris Gels (Invitrogen, NP0322BOX), and transferred onto nitrocellulose membranes (Bio-Rad, 1620215). Membranes were blocked with 1% Casein Blocker (Bio-Rad, 1610782) supplemented with 0.1% Tween 20. Membranes were incubated with the indicated antibodies, washed 4 times with Tris-buffered saline supplemented with 0.1% Tween 20, incubated with the relevant horseradish peroxidase (HRP)-conjugated secondary antibodies (GE Healthcare, Mouse-NA931 or Rabbit-NA934), and then washed 4 more times. Lastly, ECL substrate (Bio-Rad, 170-5060) was added to the membrane for 5 min and HRP activity was imaged on the ChemiDoc MP Imaging System (Bio-Rad). Antibodies used are as follows: ARNT (BD, 611079), RBFOX2 (Bethyl Laboratories, A300-864A), and β-actin (Millipore Sigma, A5316).

Statistical analysis.

All data were graphed as mean ± standard error of the mean (SEM) and repeated at least three times unless otherwise indicated. Statistical differences between two individual groups were assessed by performing a two-tailed unpaired Student’s t test using Prism software. Statistical differences between multiple groups were assessed by performing a two-way analysis of variance with Šídák’s multiple-comparison test.