Competition between RNA-binding proteins CELF1 and HuR modulates MYC translation and intestinal epithelium renewal

Fasting increases CELF1 and lowers MYC levels in small intestinal mucosa

To determine the involvement of CELF1 in the regulation of intestinal mucosal growth, we used a mouse fasting model in this study, because it represents a physiological model of intestinal mucosal atrophy (Ito et al., 2010; Lalles and David, 2011). As reported previously (Xiao et al., 2013), fasting for 48 h inhibited small mucosal growth, as indicated by a reduction in the proliferating crypt cell population marked by incorporation of BrdU (and thus representing S-phase cells) (Figure 1A, a) and a decrease in the lengths of villi and crypts (Supplemental Figure 1). Importantly, the inhibition of small intestinal mucosal growth in fasted mice was associated with a significant increase in the levels of CELF1. CELF1 immunostaining (Figure 1A, b) and protein levels (Figure 1B) increased substantially in the small intestinal mucosa in fasted mice compared with control mice. CELF1 increase by fasting was paralleled by a decrease in the levels of MYC protein, although fasting only marginally reduced Myc mRNA levels (Figure 1C). In particular, fasting-induced intestinal mucosal atrophy was associated with an increase in CELF1 binding to Myc mRNA, as measured by ribonucleoprotein (RNP) immunoprecipitation (IP) assays using anti-CELF1 antibody under conditions that preserved RNP integrity (Figure 1D). The interaction of Myc mRNA with CELF1 was examined by isolating RNA from the immunoprecipitated material and subjecting it to reverse transcription followed by real-time quantitative PCR (RT-qPCR) analysis. The induction in levels of the [CELF1/Myc mRNA] complex occurred 24 h after fasting and remained elevated 48 h thereafter. We also examined changes in CELF1 association with Occludin mRNA, a known CELF1 target transcript (Yu et al., 2013), and found that fasting for 24 or 48 h did not increase the levels of the [CELF1/Occludin mRNA] complex (Supplemental Figure 2A). On the other hand, HuR association with Myc mRNA decreased significantly after fasting (Supplemental Figure 2B). These findings suggest that fasting increases CELF1 abundance in small intestinal mucosa and that induced [CELF1/Myc mRNA] association, along with reduction of the [HuR/Myc mRNA] complex, plays a role in MYC repression and subsequent mucosal atrophy.

Myc 3′-UTR is a direct target of CELF1

There are several computationally predicted hits of the CELF1 motif in the Myc 3′-UTR based on the reported CELF1-binding sequences (Tsuda et al., 2009; Rattenbacher et al., 2010), suggesting that CELF1 interacts with Myc mRNA via its 3′-UTR. Consistent with the findings obtained from small intestinal mucosal tissue (Figure 1D), CELF1 was also found to bind to Myc mRNA in cultured IEC-6 cells (Figure 2A). Myc PCR products were highly enriched in CELF1 samples compared with control immunoglobulin G1 (IgG1) samples. The enrichment of CDK4 PCR product was also examined and served as a positive control (unpublished data), since Cdk4 mRNA is a known target of CELF1 (Xiao et al., 2011), while the amplification of Gapdh PCR products, a nonspecific contaminating transcript (not a target of CELF1) encoding the housekeeping protein GAPDH, served to monitor the evenness of sample input, as reported previously (Zhang et al., 2009). [CELF1-Myc mRNA] associations were further tested by using biotinylated transcripts that spanned the Myc 5′-UTR, CR, or 3′-UTR (Figure 2B, schematic). Following incubation with cytoplasmic lysates, the interaction between the biotinylated Myc transcripts and CELF1 was examined by biotin pull down followed by Western blot analysis (Abdelmohsen et al., 2007; Liu et al., 2009). The Myc 3′-UTR transcripts readily associated with CELF1 (Figure 2B, bottom), but the Myc 5′-UTR or CR did not. In addition, HuR also formed complexes with the Myc 3′-UTR but not with the 5′-UTR and CR as reported previously (Liu et al., 2009). On the other hand, none of the Myc partial transcripts (5′-UTR, CR, 3′-UTR) was found to interact with β-actin, included here as a negative control. For further definition of the specific CELF1-binding regions in the Myc 3′-UTR, various partial biotinylated transcripts spanning the Myc 3′-UTR (spanning positions 1898–2355) were prepared (Figure 2C, schematic). As shown, CELF1 predominantly bound to the F2 (spanning positions 2044–2212, with predicted CELF1 binding sites, Supplemental Table 1) of the Myc 3′-UTR, but there was no detectable binding of CELF1 to fragments F1 (spanning positions 1899–2071) or F3 (spanning positions 2205–2355). Interestingly, HuR also predominantly interacted with fragment F2, while it only marginally associated with F3. These results indicate that CELF1 interacts with Myc mRNA via specific RNA segments within the F2 of the 3′-UTR.

CELF1 inhibits MYC translation by interacting with the Myc 3′-UTR

To examine the functional consequences of CELF1 interactions with Myc mRNA, we first determined the effect of overexpressing the wild-type Celf1 gene on MYC abundance in IECs. As shown in Figure 3A, transient transfection with the CELF1 expression vector increased CELF1 protein expression levels approximately fivefold relative to cells transfected with the empty vector. Increased CELF1 abundance was associated with a potent inhibition of MYC expression; the reduction in MYC levels likely occurred at the translation level, since ectopic CELF1 overexpression did not decrease the levels of total Myc mRNA (Figure 3B) but repressed the rate of nascent MYC protein (Figure 3C). To further define the role of CELF1 in the regulation of MYC translation, we examined the relative distribution of Myc mRNA in individual fractions from polyribosome gradients after CELF1 overexpression. Although increasing the levels of CELF1 did not affect global polysomal profiles (Figure 3D, top), the abundance of Myc mRNA associated with actively translating components of the gradient (fractions 7–10) decreased dramatically in CELF1-transfected cells, where a significant shift of Myc mRNA was observed toward low-translating parts of the gradient (fraction 5; Figure 3D, bottom). In contrast, Gapdh mRNA, which is not a target of CELF1 and encodes the housekeeping protein GAPDH, distributed similarly in both groups. To examine whether the translational effect of CELF1 on Myc mRNA was exerted through GREs, we used a firefly luciferase (FL) reporter gene construct containing the GRE present in the Myc 3′-UTR, and negative control vector pGL3-Luc (Figure 3E, schematic). A plasmid expressing Renilla luciferase (RL) was also cotransfected as an internal control for normalization of firefly luciferase. For distinguishing translational output from changes in mRNA turnover, the luciferase activities were normalized to luciferase mRNA levels (RL mRNA, FL mRNA) to assess the translational efficiency (the “translation index”). Ectopic CELF1 overexpression was found to decrease the levels of luciferase reporter gene activity when cells were transfected with the FL-Luc (containing full-length Myc 3′-UTR) or F2-Luc but not with the F3-Luc. When cells were transfected with the F1-Luc, increasing the levels of CELF1 just slightly reduced the reporter gene activity.

Our results further showed that CELF1 silencing by transfection with small interfering RNA (siRNA) targeting the Celf1 mRNA (siCELF1) resulted in an increase in MYC expression. These specific siCELF1 nucleotides were designed to reduce Celf1 mRNA with high specificity and efficacy and low toxicity (Xiao et al., 2011; Cui et al., 2012). The levels of CELF1 protein decreased by > 80% at 48 h after the transfection with siCELF1, whereas the levels of MYC protein increased by approximately twofold compared with those in cells transfected with C-siRNA (Figure 4A). Decreasing the levels of endogenous CELF1 by siCELF1 induced MYC expression by enhancing its translation, because CELF1 silencing did not alter the levels of total Myc mRNA (Figure 4B), but it enhanced nascent synthesis of MYC protein (Figure 4C) and increased the Myc 3′-UTR luciferase reporter gene activity (Figure 4D). These results indicate that CELF1 inhibits MYC translation by directly interacting with its 3′-UTR.

HuR competes with CELF1 to bind Myc mRNA

We recently demonstrated that HuR enhances MYC translation (Liu et al., 2009), whereas target deletion of HuR in IECs causes small intestinal mucosal atrophy as a result of inactivation of the Wnt signaling pathway (Liu et al., 2014). As both HuR and CELF1 predominantly bound to the F2 of the Myc 3′-UTR (Figure 2C), we postulated that HuR and CELF1 might compete for interaction with Myc mRNA, thus jointly regulating MYC translation. To test this possibility, we first examined the effect of altering HuR levels on CELF1 binding to Myc mRNA. As shown in Figure 5A, HuR silencing by transfection with siRNA targeting the HuR mRNA (siHuR) decreased the levels of [HuR/Myc mRNA] complex, but it increased the amount of Myc mRNA associated with CELF1, although it had no effect on total Myc mRNA levels as reported previously (Liu et al., 2009; unpublished data). In contrast, CELF1 silencing not only decreased the amount of Myc mRNA associated with CELF1 but also increased the levels of Myc mRNA bound to HuR (Figure 5B). Second, we investigated the competitive binding of HuR and CELF1 to Myc mRNA by examining the effect of adding purified glutathione S-transferase (GST)-HuR (Figure 5C, a) or GST-CELF1 fusion proteins to in vitro binding reactions of Myc 3′-UTR with HuR and CELF1. When increasing concentrations of GST-HuR were added to the binding reaction (Figure 5C, b), its interaction with CELF1 was reduced (Figure 5C, c). Neither HuR nor CELF1 binding to the Myc 3′-UTR was affected by GST being added to the binding reaction (unpublished data). Conversely, increasing the concentrations of GST-CELF1 in the binding reaction mixture decreased, in a concentration-dependent manner, Myc 3′-UTR association with HuR (Figure 5D). Third, we determined the role of competitive binding of HuR and CELF1 in the regulation of MYC translation. Our results showed that CELF1 silencing and ectopic HuR overexpression together stimulated MYC translation synergistically, since the levels of MYC protein expression in cells cotransfected by siCELF1 and HuR expression vector were higher than those observed in cells transfected with siCELF1 alone (Figure 6A). Moreover, ectopic HuR overexpression in CELF1-silenced cells induced the abundance of Myc mRNA associated with actively translating fractions (fractions 8–9) in the polyribosome gradients (Figure 6B) and also resulted in additional increases in the levels of Myc 3′-UTR luciferase reporter gene activity (Figure 6C). Together these results indicate that HuR and CELF1 competitively bind to the Myc 3′-UTR and modulate MYC translation in opposite directions; HuR promoted MYC translation, CELF1 repressed it.

Polyamines regulate MYC abundance by altering the levels of CELF1

Polyamines (spermidine and spermine and their precursor putrescine) are organic cations found in all eukaryotic cells (Casero and Marton, 2007; Pegg and Casero, 2011) and are essential for maintenance of gut epithelial integrity (Wang and Johnson, 1994; Wang et al., 2010; Guo et al., 2002; Liu et al., 2003, 2005). The supply of polyamines to cells dividing in the crypts is essential for normal intestinal mucosal growth, as it increases MYC expression, whereas polyamine depletion inhibits epithelial renewal through down-regulation of MYC. In this study, we examined whether polyamines regulate MYC expression by altering CELF1. Consistent with our previous studies (Xiao et al., 2007; Zou et al., 2006), inhibition of ornithine decarboxylase (ODC, a key enzyme for polyamine biosynthesis) by treatment with DFMO (d, l-α-difluoromethylornithine) for 4 d almost totally depleted polyamines, as putrescine and spermidine were undetectable and spermine was decreased by ∼50% (unpublished data). Polyamine depletion by DFMO increased CELF1 abundance, and this induction was completely prevented by exogenous putrescine given together with DFMO (Figure 7A, top). The levels of [CELF1-Myc mRNA] complex were also increased in polyamine-deficient cells (Figure 7B), which was associated with a decrease in MYC expression (Figure 7A, middle). Furthermore, CELF1 silencing by transfection with siCELF1 in polyamine-deficient cells (Figure 7C) not only prevented the increased CELF1-Myc mRNA association (Figure 7D), but it also promoted MYC translation, as shown by an increase in the levels of MYC protein (Figure 7E) and its 3′-UTR luciferase reporter gene activity (Figure 7F). On the other hand, CELF1 silencing also enhanced formation of the [HuR-Myc mRNA] complex in polyamine-deficient cells (Supplemental Figure 3). These findings strongly suggest that polyamines regulate CELF1 negatively and that elevated CELF1 contributes to repression of MYC translation in polyamine-depleted cells.

Induced CELF1 causes G1-phase growth arrest in IECs

To study the in vitro functions of CUGBP1, we examined changes in IEC proliferation after ectopic CELF1 overexpression or CELF1 silencing. Ectopic overexpression of CELF1 using an expression vector resulted in G1-phase enrichment, with an increase in G1-phase cells, a reduction in S-phase cells, and a reduction in the total number of cells overexpressing CELF1 (Figure 8, A–C). In contrast, CELF1 silencing by transfection of siCELF1 enhanced cell proliferation, with a decrease in G1-phase cells, an increase in S-phase cells, and elevated cell numbers (Figure 8, D–F). Increasing or decreasing CELF1 did not directly affect cell death, as there were no apparent differences in cell viability, morphology, or caspase-3 cleavage after modulation of CELF1 levels (unpublished data). Together these findings indicate that CELF1 functions as a repressor of IEC proliferation and is implicated in the regulation of intestinal epithelial homeostasis.

Animal studies

C57BL/6J mice were obtained from the Jackson Laboratory and housed and handled in a specific pathogen-free animal facility at the Baltimore VA Medical Center. All experiments were performed according to animal experimental ethics committee guidelines approved by the Institutional Animal Care and Use Committee of University of Maryland School of Medicine and Baltimore VA hospital. Animals were deprived of food but allowed free access to tap water for 24 or 48 h in the fasting model. A 4-cm intestinal segment taken 0.5 cm distal to the ligament of Trietz was removed, and the mucosa was scraped from the underlying muscle with a glass microscope slide as described previously (Liu et al., 2014) and used for various measurements of the levels of mRNA and protein expression and CELF1 association with given mRNAs.

Chemicals and cell culture

Tissue culture medium and dialyzed fetal bovine serum (FBS) were from Invitrogen (Carlsbad, CA), and biochemicals were from Sigma-Aldrich (St. Louis, MO). The antibodies recognizing CELF1, HuR, Myc, GAPDH, and β-actin were obtained from Santa Cruz Biotechnology (Santa Cruz, CA) and BD Biosciences. The secondary antibody conjugated to horseradish peroxidase was purchased from Sigma-Aldrich. The IEC-6 cell line (derived from normal rat intestinal crypt cells) was purchased from the American Type Culture Collection (ATCC) at passage 13 and was maintained in T-150 flasks in DMEM supplemented with 5% heat-inactivated FBS. Passages 15–20 were used in experiments, and there were no significant changes of biological function and characterization of IEC-6 cells at passages 15–20 (Li et al., 2001; Zhang et al., 2004; Wang et al., 2010). Caco-2 cells (a human colon carcinoma cell line) were also purchased from ATCC and cultured similarly to the IEC-6 cells.

Plasmid construction

CELF1 expression vector was purchased from Origene (Rockville, MD), and the HuR expression vector was described previously (Xiao et al., 2011). The chimeric firefly luciferase reporter construct containing the Myc 3′-UTR was generated as described previously (Liu et al., 2009). The full-length Myc 5′-UTR, CR, 3′-UTR, and different GRE fragments from the Myc 3′-UTR were amplified and subcloned into the pGL3-Luc plasmid (Promega) to generate the chimeric pGL3-Luc-Myc-5′-UTR, Myc-CR, or Myc 3′-UTR reporter constructs. The sequence and orientation of the fragment in the luciferase reporter were confirmed by DNA sequencing and enzyme digestion. Transient transfections were performed as recommended by the manufacturer (Invitrogen). The luciferase reporter constructs were transfected into cells along with phRL-null, a Renilla luciferase control reporter vector from Promega, to monitor transfection efficiencies as described previously (Ouyang et al., 2015). Forty-eight hours after transfection, luciferase activity was measured using the Dual Luciferease Assay System following the manufacturer's instructions. For measurement of translational changes, the ratio of firefly luciferase to Renilla luciferase was further normalized to the levels of Firefly and Renilla mRNAs in every experiment.

RNA interference

CELF1 or HuR was silenced by transfection with a specific siRNA as described previously (Zou et al., 2010; Xiao et al., 2011). The siRNAs specifically targeting mRNAs encoding CELF1 (siCUGBP1) or HuR (siHuR) and control-siRNA (C-siRNA) were purchased from Santa Cruz. For each 60-mm cell culture dish, 15 μl of the 20 μM stock duplex siCELF1, siHuR, or C-siRNA was used. Forty-eight hours after transfection using Lipofectamine, cells were harvested for analysis.

RT-qPCR analysis

Total RNA was isolated from cells after different treatments by using the RNeasy mini kit (Qiagen, Valencia, CA) and used in reverse transcription and PCR amplifications as described previously (Guo et al., 2002). The levels of β-actin PCR product were assessed to monitor the even RNA input in RT-qPCR samples. RT-qPCR was performed using 7500-Fast Real-Time PCR Systems (Applied Biosystems, Foster City, CA) with specific primers (Rn01328012 for CELF1; Rn01519412 for Myc), probes, and software (Applied Biosystems) by following the manufacturer's instructions.

Western blot analysis

Whole-cell lysates were prepared using 2% SDS, sonicated, and centrifuged (12,000 rpm) at 4°C for 15 min. The supernatants were boiled for 5 min and size-fractionated by SDS–PAGE (7.5% acrylamide). After proteins were transferred onto nitrocellulose filters, the blots were incubated with primary antibodies recognizing Myc, CELF1, or HuR; following incubations with secondary antibodies, immunocomplexes were developed by using chemiluminescence.

Analysis of newly translated protein

New synthesis of MYC protein was measured by l-[³⁵S]methionine and l-[³⁵S]cysteine incorporation assays as described previously (Liu et al., 2009). Cells were incubated with 1 mCi (1 Ci = 37 GBq) of l-[³⁵S]methionine and l-[³⁵S]cysteine per 60-mm plate for 20 min, whereupon cells were lysed using radio-IP assay buffer. IPs were carried out for 1 h at 4°C by using either a polyclonal antibody recognizing MYC or IgG1 (BD Biosciences PharMingen, San Diego, CA). After extensive washes in TNN buffer (50 mM Tris-HCl, pH 7.5, 250 mM NaCl, 5 mM EDTA, and 0.5% NP-40), the immunoprecipitated material was resolved by 10% SDS–PAGE, transferred onto polyvinylidene difluoride filters, and visualized with a PhosphorImager (GE Healthcare, Little Chalfont, Buckinghamshire, UK).

Polysome analysis was performed as described (Chen et al., 2008). Briefly, cells at ∼70% confluence were incubated for 15 min in 0.1 mg/ml cycloheximide, lifted by scraping in 1 ml of polysome extraction buffer, and lysed on ice for 10 min. Nuclei were pelleted, and the resulting supernatant was fractionated through a 10–50% linear sucrose gradient to fractionate cytoplasmic components according to their molecular weights. The eluted fractions were prepared with a fraction collector (Brandel, Gaithersburg, MD), and their quality was monitored at 254 nm using a UV-6 detector (ISCO, Louisville, KY). After the RNA in each fraction was extracted, the levels of each individual mRNA were quantified by RT-qPCR in each of the fractions.

Biotin pull-down assays

For synthesis of RNA transcripts, cDNA from IEC-6 cells was used as a template for PCR amplification of the CR and 3′-UTR of Myc mRNA. The 5′ primers contained the T7 RNA polymerase promoter sequence (T7): 5′-CCAAGCTTCTAATACGACTC-ACTATAGGGAGA-3′. For preparation of the Myc 5′-UTR template (spanning positions 55–536), oligonucleotides 5′-GCCTCCTGCCTCCAAAAG-3′ and GCTTCAAATAACGCGAGGAG-3′ were used. For preparation of the Myc CR template (spanning positions 537–1898), oligonucleotides (T7)5′-TCTGCGACGAGGAAGAGAAT-3′ and 5′-TGCTCATCTGCTTGAACGGA-3′ were used. For preparation of the Myc 3′-UTR template (spanning positions 1899–2355), oligonucleotides (T7)5′-ACTTACTGAGGAAACGGCGA-3′ and 5′-TAAGAGAAGGCTCAATTATATTT-3′ were used. For preparation of the Myc 3′-UTR fragment-1 (F1) template (spanning positions 1899–2071), oligonucleotides (T7)5′-TGCATAAACTGACCGGAAGTGAGGA-3′ and 5′- AGTTC­TTTTATGCCTTAACTTTGAGGCA-3′ were used. For preparation of the Myc 3′-UTR fragment-2 (F2) template (spanning positions 2044–2212), oligonucleotides (T7)5′-TGCCTCAAAGTTAAGGCATAAAAGAACT-3′ and 5′-ATCTTGTATAACTGTTATAA-ACGTTTTATTAAAGT-3′ were used. For preparation of the Myc 3′-UTR fragment-3 (F3) template (spanning positions 2205–2355), oligonucleotides (T7)5′-TACAAGATTTTAAGACATGTATG-ATAAACCATAA-3′ and 5′-TAAGAGTTGGCTCAATTATATTTTTTCCA-3′ were used. PCR-amplified products were used as templates to transcribe biotinylated RNAs by using T7 polymerase in the presence of biotin-cytidine 5′-triphosphate as previously described (Liu et al., 2009). For biotin pull-down assays, biotinylated transcripts (6 μg) were incubated with 120 μg cytoplasmic lysate for 30 min at room temperature. Complexes were isolated with paramagnetic streptavidin-conjugated Dynabeads (Dynal, Oslo, Norway) and analyzed by Western blotting.

RNP IP assays

For assessment of the association of endogenous CELF1 or HuR with endogenous Myc mRNA, IP of RNP complexes was performed as previously described (Zou et al., 2006; Yu et al., 2013). Twenty million cells were collected per sample, and lysates were used for IP for 4 h at room temperature in the presence of excess (30 μg) IP antibody against CELF1 or HuR, or IgG1 (negative control). RNA in the IP materials was used in RT followed by RT-PCR and RT-qPCR analyses to detect the presence of Myc and Gapdh mRNAs.

Assay for ODC enzyme activity and polyamine analysis

ODC activity was determined by a radiometric technique in which the amount of ¹⁴CO₂ liberated from L-[1-¹⁴C]ornithine was estimated. Sample collection and the assay procedure were carried out as described in our previous publications (Wang and Johnson, 1994). Enzymatic activity was expressed as picomoles of CO₂ per milligram of protein per hour. The cellular polyamine content was analyzed by high-performance liquid chromatography (HPLC) analysis as previously described (Liu et al., 2006). Briefly, after 0.5 M perchloric acid was added, the cells were frozen at −80°C until ready for extraction, dansylation, and HPLC analysis. The standard curve encompassed 0.31–10 μM. Values that fell > 25% below the curve were considered undetectable. The results are expressed as nanomoles of polyamines per milligram of protein.

Statistics

Values are means ± SEM from three to six samples. Autoradiographic results and studies of immunofluorescence staining were repeated three times. The significance of the difference between means was determined by analysis of variance (ANOVA). The level of significance was determined by using Duncan's multiple-range test (Harter, 1960).