HMG Advance Access published September 30, 2014 1

CFTR mRNA expression is regulated by an upstream open reading frame and RNA secondary structure in its 5´ untranslated region

© The Author 2014. Published by Oxford University Press. All rights reserved. For Permissions, please email: [email protected]

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

Samuel W. Lukowski1,2,#,*, Joseph A. Rothnagel1, Ann E. O. Trezise1,2 1 School of Chemistry and Molecular Biosciences, The University of Queensland, Brisbane, QLD 4072, Australia 2 Australian Equine Genetics Research Centre, The University of Queensland, Queensland, 4072, Australia * Corresponding author: Samuel W. Lukowski, Department of Genetic Medicine and Development, University Medical School of Geneva, Geneva 1211, Switzerland, Email: [email protected]; Tel: +41 22 379 5535; Fax: +41 22 379 5706 # Current Address: Department of Genetic Medicine and Development, University Medical School of Geneva, Geneva 1211, Switzerland

2 Abstract Post-transcriptional regulation of gene expression through 5´UTR-encoded cis-acting elements is an important mechanism for the control of protein expression levels. Through controlling specific aspects of translation initiation, expression can be tightly regulated whilst remaining responsive to cellular requirements. With respect to Cystic Fibrosis, the overexpression of CFTR protein trafficking mutants, such as delta-F508, is of great biological and clinical interest. By understanding the post-transcriptional mechanisms that regulate CFTR expression, new procedures can be developed to enhance CFTR expression

negative regulatory mechanism that is encoded within the human CFTR 5´UTR and show how these elements act in combination to restrict CFTR gene expression to a consistently low level in a transcriptspecific manner. This study shows, for the first time, that endogenous human CFTR expression is posttranscriptionally regulated through a 5´UTR-mediated mechanism. We show that the very low levels of endogenous CFTR expression, compared with other low expression genes, are maintained through the cooperative inhibitory effects of an upstream open reading frame and a thermodynamically stable RNA secondary structure.

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

in homozygous delta-F508 cystic fibrosis patients. We have identified the key elements of a complex

3

Introduction CFTR is expressed at very low levels in a tightly controlled, tissue-specific manner. One of the main challenges in Cystic Fibrosis (CF; OMIM #219700) research is to understand the regulatory mechanisms that underlie the low expression of endogenous CFTR expression, so that it may be manipulated for therapeutic benefit. There is evidence that marginally increasing CFTR expression, even in delta-F508 genotypes, can

increase CFTR mRNA and protein levels through the transfer of complete CFTR cDNA sequences are promising (2), however since the identification of CFTR in 1989, a CFTR promoter has not been completely defined and, as such, it is difficult to completely resolve the transcriptional mechanisms of CFTR regulation.

Non-coding regions, including the 5´ and 3´ untranslated regions (UTRs), are known to contain both cis- and trans-acting elements that contribute to the regulation of gene expression (3). Elements such as upstream open reading frames (uORFs) and internal ribosomal entry sites (IRES) (4) and RNA G-quadruplexes (5) regulate translation, whereas microRNAs (miRNAs) (6) and trans-acting RNA-binding proteins (3) modulate transcript stability and steady-state levels. Kozak (7) demonstrated that stable RNA secondary structures in the 5´UTR inhibit translation initiation by blocking the progress of scanning ribosomes and may decrease steady-state mRNA levels and transcript stability.

While CFTR expression is under the control of multiple, tissue-specific transcription start sites, two major transcription start sites, -132 and -69, are the most frequently used in human epithelial tissues. The CFTR132 5´UTR contains multiple cis-regulatory elements, including an upstream open reading frame (uORF) and a predicted RNA stem-loop. We previously analysed the human CFTR-132 and -69 5´UTRs using mFold (8, 9), which revealed that an RNA stem-loop is predicted to form immediately downstream of the CFTR-132 uORF stop codon, similar to that described by Vilela et al (10). The CFTR-69 5´UTR does not encode the uORF, but rather its transcription start site begins at the stop codon (UAG) of the CFTR-132 uORF. Analysis of the CFTR-69 5´UTR shows that a similar, stable RNA secondary structure is predicted to form at the same location as the one identified in the CFTR-132 5´UTR (8).

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

reduce the severity of symptoms associated with Cystic Fibrosis and CFTR related diseases (1). Attempts to

4 There is evidence that CFTR gene expression is regulated by post-transcriptional mechanisms. We have previously described a mutation in the CFTR 5´UTR of a patient with Disseminated Bronchiectasis, which creates a uORF that overlaps, and is out-of-frame with, the main CFTR translation initiation codon. We demonstrated that this CFTR uORF mutation was sufficient to decrease gene expression up to 99% by reducing translation efficiency and mRNA stability (8). This was the first study to demonstrate a CFTR regulatory mutation, associated with a CFTR-related disease, acting at the post-transcriptional level.

stem-loop regulate translation efficiency and stability of the CFTR mRNA. We provide evidence for a novel mechanism of post-transcriptional regulation of endogenous human CFTR expression and show that individual cis-regulatory elements independently, and synergistically, repress endogenous CFTR expression.

Results Computational analysis of the CFTR-132 5´UTR mRNA secondary structure Initial RNA secondary structure predictions were performed on the entire CFTR-132 or CFTR-69 5´UTR sequence using established software (9, 11) (Figure 1A). To determine whether the predicted secondary structure formed due to the length of the 5´UTR sequence, thus occurring by chance, or whether the region of the predicted CFTR 5´UTR stem-loop was the most thermodynamically favorable for this structure, we computationally examined the CFTR-132 5´UTR using a sliding window technique, previously described by Jenkins et al (12). The sequence was interrogated one nucleotide at a time, in windows ranging from 30 to 100 nucleotides in length using the RNAfold software (13), and the minimum free energy (MFE) and ensemble free energy (EFE) structures were determined. Analysis of the ensemble free energy per nucleotide (Figure 1B) shows a region of lower free energy between nucleotides -64 and -32 (relative to the start codon of the main CFTR protein-coding ORF (pAUG), where the A is +1) in nucleotide windows 64 to 100, indicating that this region has a greater propensity to form a stable secondary structure. Conversely, the free energy associated with the 5´ end of the sequence is much closer to zero, suggesting an absence of secondary structure in this region. The MFE and EFE plots of dinucleotide shuffled sequences based on the sequence of the CFTR-132 5´UTR show similar free energy patterns to the real CFTR 5´UTR sequence (Figure 1B-C), indicating that the nucleotide composition of each window, rather than the nucleotide order,

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

In this study, we show that a naturally occurring upstream open reading frame and a predicted, stable RNA

5 is likely to be responsible for the formation of stable secondary structure in the -64 to -30 region. Notably, the nucleotide composition of the predicted stem-loop region shows a slightly increased G/C content, compared to the CFTR-132 5´UTR in its entirety (Table 1), and is indicative of secondary structure formation in this area.

Segment scores (14) were calculated to assess the statistical significance of structured regions in the CFTR-

than +1.96 suggest statistically significant folding of RNA. Using the minimum and ensemble free energies, we determined the segment scores of the entire CFTR-132 5´UTR (MFE = +1.98; EFE = +1.89), suggesting the CFTR-132 5´UTR is inherently structured. We then determined the likelihood of base-pairing between individual nucleotides with probability >0.5 (most likely to pair) for each 30 nucleotide window using the RNAfold centroid structure and the minimum free energy of the ensemble structures. The number of paired bases with probability >0.5 was expressed as a percentage of the total number of paired bases in each 30nt window and shows increased probability of base-pairing in the stem-loop region (Figure 1D: graph and 1E: centroid structure). In the region of the stable stem (-64 to -30nt), we found the base-pairing probabilities were generally between 0.75 and 0.99 (median = 0.85). The increased base-pairing probability correlates with the regions of lowest free energy identified in the EFE and MFE plots (Figure 1B-C). Taken together, this data indicates a propensity for stable secondary structure formation between nucleotides -64 and -30 in the CFTR-132 5´UTR.

Analysis of the CFTR 5´UTR upstream open reading frame and its effect on downstream translation initiation Previously, we reported the difference between the CFTR-132 and the CFTR-69 5'UTRs in their capacity to drive translation initiation at the main CFTR AUG codon (8). We showed that the baseline reporter activity of a luciferase construct containing the CFTR-69 5'UTR was approximately 40% of a construct containing the CFTR-132 5'UTR in HT29 and HEK293 cells.

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

132 5´UTR compared to shuffled sequences. Segment scores that are generally less than -1.96 or greater

6 Reporter activity directed by the CFTR-132 and the CFTR-69 5´UTRs, both of which contain thermodynamically stable RNA secondary structures, were compared to luc2 activity driven by the human beta-globin 5´UTR (HBB), which is short (50nt) and efficiently initiates translation (15). As predicted, luciferase activity under the control of the HBB 5´UTR was significantly higher than either of the CFTR constructs: 263% in HT29 cells (P < 0.0001) and 165% (P = 0.0041) in HEK293 cells, compared to the CFTR-132 5´UTR (Figure 2A). A HBB negative control, in which the pAUG was mutated to a stop codon

control produced no luc2 reporter activity, indicating that any luc2 activity observed using the CFTR constructs or the HBB positive control was due exclusively to translation initiation at the pAUG codon. With the normal luc2 activity directed by both wildtype CFTR 5´UTRs established, changes in activity due to CFTR uORF-based mechanisms were investigated by mutating the CFTR-132 5´UTR using site-directed mutagenesis. Firstly, the uORF start codon (only present in the CFTR-132 5´UTR) was converted to a stop codon (AUG>UAG). Contrary to expectations, repressing translation initiation at the uORF-AUG had little effect on the reporter activity in both HT29 and HEK293 cell lines compared to wildtype levels (Figure 2A: uORF AUG>UAG). A slightly increased, but not significant level of reporter activity was noted in both cell types.

Efficiency of translation initiation at an AUG codon is also influenced by the nucleotide sequence immediately surrounding the start codon. Since the uORF-AUG is surrounded by a ‘poor’ Kozak consensus (0/2 critical nucleotides), the likelihood of scanning ribosomes consistently initiating translation at this AUG codon is less than if the uORF-AUG was in a more favorable Kozak consensus (16). With this in mind, we tested the effect of altering the Kozak consensus of the uORF-AUG to sub-optimal (1/2 critical nucleotides) and optimal (2/2 critical nucleotides), both of which are known to increase the probability of translation initiation occurring at the target AUG codon, on translation initiation at the CFTR pAUG. Following the conversion from 0/2 to 1/2 critical nucleotides, luciferase activity reflecting the rate of translation initiation at the CFTR pAUG, increased by 55% (P = 0.0009) in HT29 cells and 98% (P = 0.0129) in HEK293 cells (Figure 2A: uORF Kozak 1/2). When the Kozak consensus of the uORF-AUG was converted from 0/2 to 2/2 critical nucleotides, luciferase activity remained at levels similar to wildtype (Figure 2A: uORF Kozak

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

(UAG), was designed to eliminate reporter activity arising from the pAUG. As predicted, the HBB negative

7 2/2). This result was unexpected, given that the conventional effect of translation initiating at uAUGs typically reduces translation initiating from downstream AUGs (17).

The unexpected increase in luciferase activity associated with the optimization of the uORF Kozak consensus sequence prompted an investigation into the effect of Kozak nucleotides on the translation efficiency of the downstream CFTR-AUG. The Kozak context of the native CFTR-AUG is 1/2 critical

CFTR-132 and CFTR-69 5´UTRs exhibited patterns of luciferase activity consistent with the current understanding of the effects of Kozak context on translation initiation efficiency. In reporter assays featuring the CFTR-132 5´UTR, translation efficiency was slightly below wildtype in a poor Kozak context (n.s; Figure 2B: -132 Koz 0/2), while an optimized Kozak context (2/2) showed an increase in translation efficiency of nearly two-fold (HT29 - 192%, P = 0.03; HEK293 - 203%, P = 0.0257; Figure 2B: -132 Koz 2/2). Accordingly, the CFTR-69 5´UTR constructs displayed similar patterns of translation efficiency to the CFTR-132 constructs (Figure 2B). Compared to the CFTR-69 wildtype pAUG, changing the Kozak context to 0/2 in the CFTR-69 5´UTR reduced reporter activity by 67% (P = 0.0022) in HT29 and HEK293 (P = 0.005), while improving it to 2/2 increased activity by 81% (P = 0.0318) and 49% (n.s) in HEK293 cells.

To assess the capacity of the uORF-AUG itself to initiate translation, the uORF-AUG and uORF second codon were linked directly to the third codon of luc2. This modification removed all intervening sequence as well as the downstream CFTR-AUG codon, allowing the uORF-AUG to be tested in isolation. The uORFluc2 fusion construct exhibited an increase of 1.3- and 1.2-fold in luciferase activity compared to the wildtype CFTR-132 5´UTR construct (P = 0.0096; Figure 2B: uORF:luc fusion).

Direct measurement of changes to uAUG mediated translation initiation demonstrated that the CFTR -132 uAUG actively supports translation initiation. Altering the Kozak sequence of the uAUG led to changes in uORF translation efficiency, shown by the effects on pAUG translation when the uAUG Kozak context was 0/2, 1/2 or 2/2 critical nucleotides. This data is supported by previous work by Wang and Rothnagel (18) demonstrating that increased uORF translation leads to reduced pAUG translation. However, the effects of the 0/2, 1/2 or 2/2 Kozak context mutations on CFTR-132 uAUG translation initiation differed from the

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

nucleotides, therefore, the Kozak context was mutated from 1/2 to 0/2 (poor) and 1/2 to 2/2 (optimal). Both

8 predicted effects of surrounding Kozak sequence, suggesting that additional factors may be involved. With this in mind, further investigations into the activity of additional cis-acting elements present within the CFTR 5´UTR were undertaken.

The effect of RNA secondary structure on translation efficiency at the CFTR-AUG

mRNA stability in eukaryotes. Scanning ribosomes are often blocked or detached from an mRNA upon encountering stable stem-loop (SL) structures. This results in decreased translation of downstream coding regions as well as mRNA instability. The predicted, conserved stable hairpin structure present in the CFTR 5´UTR is located immediately 3´ of the uORF stop codon in CFTR-132 mRNA and begins at the third nucleotide from the transcription start site of CFTR-69 mRNA. We initially predicted this structure using mFold (9), and this was confirmed using the Vienna RNAfold tool (13). The specific location and evolutionary conservation (data not shown) of the predicted hairpin suggested a role in the posttranscriptional regulation of CFTR. To address this issue, destabilization of the stem-loop structure was achieved by mutating a number of nucleotides in the double-stranded stem region, such that stem formation via complementary base pairing would be inhibited. Several stem-loop mutants were created (Table 2) and their effect on CFTR pAUG-luc2 translation initiation efficiency was measured using our CFTR 5´UTRCFTR pAUG luc2 reporter system. Compared to the wildtype CFTR-132 5´UTR (folded structures: Figure 3A and C), the mutation that produced the greatest increase in reporter activity was in stem-loop mutant 2, (132-SL2; GG>CC; folded structures: Figure 3B and D), which produced a 7-fold increase in luciferase activity in HT29 cells (P = 0.0002) and a 1.5-fold increase in HEK293 cells (P = 0.003; Figure 4).

The effect of the same stem-loop structure located at the distal 5´ end of the CFTR-69 5´UTR was assessed using the same mutation strategy as the CFTR-132 5´UTR. Here, -69-SL2 (folded structures: Figure 3F and H) had a similar effect on translation efficiency as -132-SL2, compared to the wildtype CFTR-69 5´UTR (folded structures: Figure 3E and G), resulting in a 4-fold increase in luciferase activity in HT29 cells (P = 0.0222) and a 2.5-fold increase in HEK293 cells (P = 0.0411; Figure 4).

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

RNA secondary structure within a 5´UTR has been shown to influence both translation efficiency and

9 Having defined the CFTR 5´UTR as a potent modulator of translation initiation, an additional construct was generated, combining the uORF-AUG>UAG and -132-SL2 mutations in a single construct. By eliminating the effects of the uORF and stem-loop on downstream translation, a substantial, 14-fold increase in luciferase activity was observed in HT29 cells (P < 0.0001; Figure 4). We also observed a 4- to 5-fold increase in reporter activity in HEK293 cells (Figure 4), compared to the wildtype 5´UTR (P = 0.0009). These data suggest a post-transcriptional regulatory mechanism comprised of two distinct, cis-acting

132 5´UTR.

uORF and stem-loop mutations alter CFTR mRNA abundance and stability Luciferase reporter activity of wildtype or mutant CFTR 5´UTR-luc2 constructs suggested that changes in luc2 activity might be due to altered translation initiation efficiency at the CFTR pAUG. To distinguish the effects of the different constructs, and determine their impact upon luc2 mRNA stability and translation initiation efficiency, steady-state mRNA levels and the stability of CFTR-luc2 mRNA transcripts were measured using quantitative real-time PCR. Molecular Beacon qPCR probes were designed to bind specifically to the luc2 coding region. This allowed accurate quantitation of steady-state luc2 mRNA levels the kinetics of mRNA decay (measured at 0, 3, 6 and 9 hours following treatment with Actinomycin D).

Relative steady-state expression profiles of luc2 transcripts linked to the CFTR-132 and CFTR-69 wildtype 5´UTRs, were determined in HT29 cells 48h post-transfection and compared to the luc2 mRNA expressed from the unmodified pGL4.13-luc2 vector. qPCR analysis showed a 36% reduction (Supplementary Figure 1) of CFTR-69-luc2 mRNA compared to luc2 transcripts containing the CFTR-132 5´UTR sequence. mRNA decay rates were determined following the inhibition of transcription with Actinomycin D (5µg/mL). No significant difference was observed over a 9hr time period between the CFTR-132-luc2 and the CFTR-69luc2 mRNA half-lives (t1/2), which were measured as 5.2 and 4.9h respectively (Figure 5A: P = 0.335). Relative steady-state abundance of pGL4.13-luc2 transcripts (vector only) was measured as 27% of the wildtype CFTR-132-luc2 mRNA level (P = 0.0259), and the half-life of the pGL4.13-luc2 mRNA was not significantly different from the CFTR-132-luc2 mRNA half-life (t1/2 = 6.4h: P = 0.0712).

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

elements (an upstream ORF and a stable RNA secondary structure) acting combinatorially within the CFTR-

10

To test the effect of translation of the native uORF on mRNA stability, both the transcript half-life and the steady-state mRNA level of CFTR-132-uAUG>UAG-luc2 mRNA was compared with that of the wildtype CFTR-132-luc2 mRNA. Replacement of the uAUG codon with a UAG codon in the CFTR-132 5´UTR had no significant effect on either steady-state mRNA levels or transcript half-lives (Figure 5B). The half-life of the wildtype CFTR-132-luc2 mRNA was 5.2h and the half-life of the CFTR-132-uAUG>UAG-luc2 mRNA

A 7-fold increase in reporter activity was observed following the destabilization of the predicted stable RNA secondary structure that is present in the CFTR-132 wildtype 5´UTR. To determine whether this increase was due to increased efficiency of translation initiation at the CFTR-AUG codon and/or altered transcript stability, the steady-state level and rate of decay of CFTR-132-SL2-luc2 mRNA were measured. qPCR analysis of the CFTR-132-SL2-luc2 transcript revealed a 3.5-fold increase in mRNA stability (t1/2 = 18.2h; P = 0.0314; Figure 5C), compared to the CFTR-132-luc2 wildtype 5´UTR, but also a 71% reduction in steady-state mRNA quantity (Supplementary Figure 1; P = 0.0269). The increased mRNA stability of the CFTR-132-SL2-luc2 transcripts suggests that the overall increase in luc2 activity produced by destabilization of the CFTR 5´UTR stem-loop is a result of both more efficient translation initiation at the CFTR pAUG codon and increased mRNA stability.

The 14-fold increase that we observed when we combined the uAUG>UAG and SL2 mutations suggested that the native uORF and RNA secondary structure might work in concert to maintain low levels of endogenous CFTR expression. To examine the relative contributions of translation and mRNA stability in a co-operative mechanism, a CFTR-132-luc2 construct containing both uORF-AUG>UAG and SL2 mutations (CFTR-132-uAUG>UAG-SL2-luc2) was analyzed. Quantitation of steady-state mRNA showed a return to levels similar to that of wildtype CFTR-132-luc2 (P = 0.141), and an 82% increase compared to the steadystate levels of CFTR-132-SL2-luc2 mRNA (P = 0.0427). The half-life of the CFTR-132-uAUG>UAG-SL2luc2 mRNA was substantially higher than the wildtype (t1/2 = 15.8h; P = 0.0114; Figure 5D), representing a 3-fold increase in transcript stability. The difference in mRNA stability between the CFTR-132-SL2-luc2 and the CFTR-132-uAUG>UAG-SL2-luc2 transcripts was minimal (P = 0.418; Figure 5E).

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

was 5.0h (P = 0.439).

11

To assess whether the presence of the RNA secondary structure in the CFTR-69 wildtype 5´UTR modulates the abundance and/or stability of the mRNA despite its proximity to the transcription start site, the hairpin was mutated using the same strategy and analyzed using qPCR. Steady-state mRNA levels arising from the transfected CFTR-69-SL2-luc2 construct were increased 3.4-fold compared to the wildtype CFTR-69-luc2 construct (P = 0.0015), and the half-life of the CFTR-69-SL2-luc2 mRNA was also 1.6-fold higher (t1/2 =

Discussion CFTR is expressed at very low levels across a wide variety of epithelial cells, and in some non- epithelial cell types. Tightly controlled, tissue specific expression is a hallmark of CFTR (19-25) and fluctuations in the number of CFTR ion channels at the cell surface are harmful to both the cell and the organism. Remarkably low levels of CFTR gene expression, approximately 6% that of the beta-actin gene (26), are due to a critical and evolutionarily conserved mechanism about which little is known. To understand the mechanisms that maintain CFTR expression levels within critical limits, whilst maintaining a tightly regulated, tissue-specific level of gene expression, is an intriguing problem. CFTR expression requires strict regulation since abnormally low levels result in the Cystic Fibrosis disease state, whereas over-expression can lead to excessive inflammatory responses (27). The data presented here provide evidence for a previously undescribed mechanism regulating CFTR, and illustrate how a low, yet essential, level of gene expression is maintained.

The involvement of putative regulatory elements present in the CFTR 5´UTR was explored using a series of luciferase reporter constructs containing a variety of wildtype and mutant 5´UTRs from the two predominantly expressed CFTR mRNA transcripts. Of these transcripts, the 132nt 5´UTR (CFTR-132) is expressed in human fetal lung. The second 5´UTR is -69nt in length (CFTR-69) and is expressed mainly in adult lung and small intestine (25).

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

7.7h; P = 0.027; Figure 5F).

12 CFTR-132 and CFTR-69 5´UTRs support translation initiation in a manner consistent with a developmentally regulated switch Wildtype levels of CFTR 5´UTR-mediated luciferase translation were assessed in two cell lines – HT29 (human intestinal epithelial) cells that express CFTR at comparatively high levels, and HEK293 (human embryonic kidney) cells that do not express CFTR. Luciferase reporter activity measured using either the wildtype CFTR-132 or the CFTR-69 5´UTR to drive luc2 expression showed that the CFTR-69 5´UTR

These results are consistent with observations made in vivo by White and colleagues (25) that steady-state CFTR mRNA levels are considerably higher in fetal lung than in adult lung. White et al (25) found that the major transcription start site used in fetal lung is the CFTR-132, whereas CFTR-69 is used in post-natal and adult lung, and intestinal epithelia. Their data supports the reduced luc2 activity that was measured in the CFTR-69-luc2 constructs. They also reported that the use of the CFTR-132 and the CFTR-69 transcription start sites is developmentally regulated, and that the utilization of the CFTR-132 or the CFTR-69 transcription start sites is distinctly correlated with a developmental switch in CFTR regulation in the lung. The key characteristic of this developmental switch is the conversion of the lung epithelium from absorptive to secretive after birth.

The CFTR-132 uORF acts as a ‘roadblock’ to scanning ribosomes Upstream open reading frames are widely accepted as repressors of translation initiation at downstream AUG codons. To determine the effect of the CFTR 5´UTR uORF on translation initiation at the CFTR pAUG, the uORF start codon (uORF-AUG) was converted to a stop codon (uAUG>uAUG) to prevent translation initiation occurring upstream of the main CFTR start codon (CFTR pAUG). Unexpectedly, mutation of the uAUG to UAG had no significant effect on steady-state transcript levels, mRNA half-life or translation initiation efficiency at the CFTR pAUG.

The nucleotide sequence surrounding a translation initiation codon (the Kozak sequence) influences the rate of recognition of the AUG codon by scanning ribosomes. The Kozak consensus sequence is recognized as a potent modulator of ribosome recognition and has been shown to significantly alter the translational

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

supports translation at a significantly reduced rate (~40%), compared to the CFTR-132 5´UTR (P & 0.001).

13 efficiency of a given AUG codon. The native Kozak context of the CFTR uORF-AUG is 0/2 critical nucleotides. The Kozak context can be used as an indicator for identifying an infrequently recognized start codon. The minor effect of the CFTR uAUG>UAG mutation on translation initiation at the CFTR pAUG might be explained by a low rate of recognition of the uORF-AUG codon by scanning ribosomes, due to the Kozak context of 0/2 nucleotides. This would reduce translation initiation at the uAUG codon and increase the likelihood of ribosomes encountering, and initiating translation at, the CFTR pAUG. Improving the

+4 positions, yielded puzzling results. In general, promoting uAUG translation initiation would be expected to decrease downstream translation initiation. Here, improvement of the native uORF-AUG Kozak context (0/2) by conversion to 1/2 critical nucleotides increased the level of luciferase activity, correlating with an increase in translational efficiency from the CFTR pAUG. It is possible that improving the Kozak context of the uORF-AUG (1/2) promotes uORF translation, but also promotes post-termination ribosome re-initiation – a process that would allow a higher proportion of ribosomes to re-initiate and continue translation from the downstream CFTR-AUG codon.

For constructs in which the Kozak context of the uORF-AUG was improved through conversion to 2/2 critical nucleotides, a slight decrease in reporter activity was observed compared to the wildtype CFTR-132 5´UTR. These results are consistent with increased uORF-AUG recognition by scanning ribosomes and indicate that leaky scanning through the uORF to the main ORF is inhibited if the uORF-AUG is in an optimized Kozak context.

To confirm that the Kozak rules apply in the context of the CFTR mRNA, the same mutations were applied to the Kozak sequence surrounding the native CFTR pAUG codon. Observed reporter activity correlated with CFTR pAUG translation initiation was as expected (16, 18), with both the CFTR-132 and CFTR-69 5´UTRs displaying a pattern of increasing activity from 0/2 through to 2/2 critical nucleotides.

Changes in luciferase reporter activity produced by 5´UTR mutants may also be due to altered transcriptional processes (28, 29), miRNA binding sites (30, 31) and RNA-protein interactions (32-35). To confirm the results of earlier experiments, the capacity of the uORF-AUG to support translation initiation in the absence

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

Kozak context of the uORF-AUG, by converting the 0/2 status to 1/2 or 2/2 critical nucleotides at the -3 and

14 of other AUG codons was assessed. We generated a construct in which the sequence encoding the uORF stop codon through to the first two codons of the main coding region was deleted, allowing the uORF coding sequence to be linked directly to the third codon of luc2. Any luc2 reporter activity was deemed due to translation initiation at the uORF-AUG. By maintaining the majority of the uORF sequence, it was possible to test the Kozak context of the uORF-AUG and its contribution to uORF translation initiation under native and mutant conditions. Compared to the wildtype CFTR-132 5´UTR, the modified uORF-luc2 fusion 5´UTR

sequence of such short length upstream of the first AUG codon, approaches the minimum sequence requirements for successful translation initiation (18). Luciferase reporter activity from this fusion construct showed that the uORF-AUG drives luc2 translation at 150% that of the wildtype CFTR-132 5´UTR in HT29 cells and 110% in HEK293, supporting the notion that translation of the uAUG reduces the availability of ribosomes at the main CFTR pAUG and likely contributes to the overall regulation of CFTR translation.

RNA secondary structure in the CFTR 5´UTR limits ribosome access to downstream start codons Constructs containing the CFTR-132 and CFTR-69 5´UTRs were mutated to destabilize the stem of the predicted RNA secondary structure. Previous work has demonstrated the effect of RNA secondary structure on translation initiation and re-initiation, as well as mRNA stability, which can occur via stem-loop-mediated ribosome stalling and shunting, or detachment of the ribosome from the mRNA, which may lead to mRNA degradation (10, 36-38). In particular, Vilela et al showed that translation initiation and mRNA stability can be severely reduced in the yeast gene Yap1p, if an RNA stem-loop is inserted immediately downstream of a benign uORF (10, 38, 39). As such, the presence of a stable stem-loop in the CFTR 5´UTR suggests a similar post-transcriptional mechanism regulates CFTR gene expression.

To test this hypothesis, the stem of the hairpin structure was destabilized in the CFTR-132 5´UTR using the 132-SL2 mutation (GG>CC) which produced a 7-fold increase in luc2 activity. The half-life of those mRNAs containing the -132-SL2 mutation was increased following destabilization of the stem-loop structure (t1/2 = 18.2h), while the steady-state mRNA levels were reduced by 71% (Supplementary Figure 1), suggesting an altered rate of transcription associated with the SL2 mutation. However, when the same

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

was 63nt in length, however the distance from the 5´cap to the uAUG codon was only 15nt. A leader

15 mutation is introduced into the CFTR-132-uAUG>UAG-SL2 transcript, mRNA abundance increased and the rate of decay was marginally lower (t1/2 = 15.8h). The 3.5-fold increase in stability observed in CFTR-132SL2 transcripts, in conjunction with increased reporter activity, implies that in the wildtype state, both CFTR protein expression and mRNA stability are heavily repressed in the presence of the RNA hairpin structure. Recently, two independent groups showed that impaired mRNA decay mechanisms in yeast are compensated by decreased mRNA synthesis via a global RNA exonuclease (Xrn1)-mediated mechanism, and that the

humans may underlie the reduced steady state mRNA level we observed for the CFTR-132-SL2 mutant. For this transcript, the rate of decay was substantially increased compared to the wildtype, and moderately increased compared to the CFTR-132-uAUG>UAG-SL2 transcript, while also exhibiting a reduced steady state mRNA level, which raises the possibility of transcriptional buffering.

In the CFTR-132 5´UTR, the stem-loop is located immediately 3´ of the uORF termination codon and may impede linear scanning by ribosomes that did not initiate translation at the uORF-AUG. Indeed, this could interfere with ribosomes actively translating the uORF by preventing termination-dependent re-initiation or promoting ribosome stalling at the uORF stop codon (37, 42, 43). Ribosomes would still be able to initiate translation at a downstream start codon were they able to shunt across the region of the stem-loop and migrate in a non-linear fashion (36), however this process is dependent upon proper ribosome termination, termination codon sequence context and uORF-peptide release (37). Destabilization of the RNA secondary structure in the CFTR 5´UTR is thought to allow a much greater proportion of scanning ribosomes to initiate translation at a downstream AUG due to (i) less structural inhibition of ribosomes and (ii) a significantly enhanced mRNA half-life.

The synergistic effect of the uORF/stem-loop arrangement in the CFTR-132 5´UTR represses CFTR translation initiation and mRNA stability The possibility that a complex mechanism, involving the uORF and stable RNA secondary structure acting in concert in the CFTR-132 5´UTR, might repress endogenous, wildtype CFTR gene expression and transcript stability, was addressed using the CFTR-132 5´UTR containing a double mutant (CFTR-132-

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

mechanisms of mRNA decay and synthesis are coordinated and circular (40, 41). A similar mechanism in

16 uAUG>UAG-SL2). The CFTR-132-uAUG>UAG-SL2 construct contains the uORF with a UAG-mutated uAUG codon, as well as the -132-SL2-destabilised RNA stem-loop, shown previously to increase luc2 activity 7-fold.

The combination of the uAUG>UAG and -132-SL2 mutations increased luc2 activity 15-fold in reporter assays, and produced a 3-fold increase in mRNA stability. Importantly, there was also a 162% increase in

likely due to the increased mRNA half-life. Compared to the steady-state mRNA levels of the CFTR-132SL2 construct, the CFTR-132-uAUG>UAG-SL2 transcript levels increased by 82% (Supplementary Figure 1) and may be attributed to the addition of the uAUG>UAG mutation. The increased steady-state mRNA level of the CFTR-132-uAUG>UAG-SL2 construct suggests that converting the uORF-AUG to a stop codon prevents uORF translation and could allow scanning ribosomes to shunt across the base of the stem-loop and continue migrating until the next AUG is encountered. This data implies a shared regulatory role operating between the uORF and stem-loop, such as that described in YAP1-type uORFs (10). Destabilization of the stem-loop alone contributed a 7-fold increase in luc2 reporter activity, a 3.5-fold increase in mRNA stability and a reduced level of steady-state mRNA, while the uORF-AUG>UAG mutation alone had negligible effect. The substantial increase in reporter activity and the rescue of steady-state mRNA to wildtype levels arising from the CFTR-132-uAUG>UAG-SL2 construct indicates a co-operative 5´UTR-mediated mechanism involved in the regulation of CFTR gene expression.

The post-transcriptional mechanism regulating CFTR gene expression discussed here closely resembles the YAP1-type uORF mechanism described by Vilela et al (10, 38). The CFTR-132 uAUG is located 13nt downstream of the CFTR-132 transcription start site and has an initiation context of 0/2 critical nucleotides. A 5´ leader sequence of 8-12nt is required for efficient translation initiation (18, 44) and the location of this uAUG, coupled with the poor Kozak context, is positioned such that translation initiation at the CFTR-132 uAUG does not impact greatly on translation initiation from the CFTR pAUG. However, reporter assays measuring luc2 activity from uORF-uAUG-luc2 fusion constructs revealed that the CFTR-132 uAUG can drive translation at wildtype CFTR levels if it is recognized by scanning ribosomes, even in a poor Kozak context. In addition, impeding translation initiation at the uORF-AUG codon did not significantly affect

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

steady-state mRNA abundance compared to the CFTR-132 wildtype (Supplementary Figure 1), which is

17 reporter activity, steady-state mRNA levels or transcript stability, indicating that active translation of the uORF alone does not influence CFTR expression.

In this study, we have identified the key elements of a complex, negative regulatory mechanism that is encoded within the human CFTR 5´UTR and examined how these elements act in combination to restrict CFTR gene expression to a consistently low level in a transcript-specific manner. These data show, for the

mediated mechanism. The data reported here show that the very low levels of CFTR expression, compared with other low expression genes, are maintained through the combinatorial inhibitory effects of the CFTR132 uORF and the thermodynamically stable RNA secondary structure located immediately 3´ of the uORF termination codon. Similar post-transcriptional regulatory mechanisms have been identified in humans (45, 46), however neither of these examples use this specific mechanism. It is possible that there are other human disease-causing genes that are regulated this way and are amenable to modulation of their expression through its perturbation. This multifaceted post-transcriptional mechanism may provide unique methods and therapeutic targets by which CFTR expression can be manipulated to over-express CFTR proteins in homozygous delta-F508 patients with the potential for minimizing the symptoms associated with cystic fibrosis and CF-related disorders.

Materials and Methods Reporter constructs - CFTR 5´UTR-luciferase reporter constructs Plasmid constructs containing the CFTR 5´UTR (RefSeq: NM_000492) were generated in a pGL4.13[luc2/SV40] vector (Promega) using a combination of restriction endonuclease digestion and PCR. The full-length CFTR-132 5´UTR plus the first two codons of the CFTR coding region was amplified by PCR using genomic DNA extracted from HEK293 cells. PCR reactions consisted of 100ng of HEK293 genomic DNA template, 10mM dNTPs, 1U Phusion DNA Polymerase (Finnzymes), 1X Phusion HF buffer and 500nM each of the CFTR-specific primers in a final volume of 50µL. CFTR-specific primers incorporated 5´ tails complementary to the SV40 promoter region, to include a HindIII restriction site (forward primer) and the luc2 coding region, excluding the first two luc2 codons (reverse primer), of the vector. The PCR protocol comprised a 30s denaturation step at 98oC, followed by 35 cycles of denaturation

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

first time, that endogenous human CFTR expression is post-transcriptionally regulated through a 5´UTR-

18 (98oC, 10s), annealing (60oC, 30s) and extension (72oC, 60s), and a final extension step at 72oC for 10min. Next, the luc2 region was amplified from the pGL4.13 vector using the same method. The forward primer incorporated a 5´ tail complementary to the CFTR 5´UTR plus codons 1 and 2, while excluding the first two codons of the luc2 region. The reverse primer included a 5´ tail incorporating the XbaI restriction site downstream of the luc2 coding region. To create the CFTR 5´UTR-luc2 fusion, both amplicons were added in equal quantities to a single PCR reaction as described. The forward and reverse primers were the SV40-

amplicon were individually digested with HindIII and XbaI to create compatible cohesive restriction sites. The vector and fusion insert were ligated and cloned as described (section 2.1). The new construct was verified by sequencing across the full length of the plasmid in both forward and reverse directions. The CFTR-132-5´UTR-pGL4.13 construct created in these steps was the basis for all further CFTR-5´UTR-based constructs used in luciferase reporter assays.

HBB 5´UTR-luciferase reporter constructs Control constructs were generated in the same manner as described above, except the HBB 5´UTR was used in place of the CFTR-132 5´UTR. Primers were designed such that the forward and reverse primers included 5´ tails complementary to the SV40 promoter and the luc2 coding region, excluding the first two luc2 codons, of the pGL4.13 vector respectively. The luc2 region was amplified as described, however, the forward primer included a 5´ tail complementary to the HBB 5´UTR sequence plus codons 1 and 2 of the HBB coding region. The negative control, HBB(neg)-5´UTR-pGL4.13, was created using the HBB-5´UTRpGL4.13 positive control plasmid. The HBB-AUG codon in this construct was converted to a stop codon (HBB-UAG) using site-directed mutagenesis.

Mutagenised CFTR and HBB 5´UTR-luciferase constructs All mutant constructs were generated from either the CFTR-132-5´UTR-pGL4.13 or HBB-5´UTR-pGL4.13 plasmids. Typically, back-to-back primers were designed to mutate the region of interest using PCR-based site-directed mutagenesis (SDM). The SDM reaction consisted of 1ng CFTR-132-5´UTR-pGL4.13 plasmid

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

CFTR-5´UTR (forward) and the luc2-XbaI (reverse). The pGL4.13 vector and the CFTR 5´UTR-luc2 fusion

19 DNA template, 10mM dNTPs, 1U Phusion DNA Polymerase (Finnzymes), 1X Phusion HF buffer and 500nM each of the CFTR-69 mutagenic primer in a final volume of 50µL. The SDM PCR protocol comprised a 30s denaturation step at 98oC, followed by 35 cycles of denaturation (98oC, 10s), annealing (6071oC, 30s) and extension (72oC, 90s), and a final extension step at 72oC for 10min. All mutants were cloned, verified by sequencing and prepared using the same method as for the recombinant CFTR-132-5´UTRpGL4.13 plasmid.

HT29 (human intestinal epithelia) cells were grown in RPMI 1640 media plus L-Glutamine (Gibco), while HEK293 (human embryonic kidney) cells were cultured in Dulbecco's modified Eagle's medium (DMEM) plus L-Glutamine and Sodium Pyruvate. Both cell lines were supplemented with 10% (v/v) Newborn Calf Serum (NCS) (Gibco) and 1% (v/v) Penicillin/Streptomycin (Gibco) and grown at 37oC in a 5% CO2 atmosphere. Culture media was changed 2-3 times per week and passaged 1:3 or 1:5 when required.

Transient transfection HT29 and HEK293 cells were plated out at 70% confluency in 96-well plates and allowed to adhere overnight. The following day, cells were washed with PBS and overlaid with antibiotic-free OptiMEM (Gibco) prior to transfection with FuGENE HD (Roche) as per the manufacturers instructions. Briefly, construct-specific transfection mastermixes were prepared at a ratio of 10:2 (µL FuGENE HD:µg DNA) with OptiMEM media in a final volume of 110µL. Each mastermix was mixed by pipetting and incubated at room temperature for 15min to allow the formation of FuGENE:DNA complexes. Each construct-specific complex was added directly to the cells in the 96- or 24-well plates in triplicate and overlaid with OptiMEM media. Each plate was gently rocked for 20s to evenly distribute the transfection complexes. 4-6h post-transfection, the media was withdrawn and replaced with maintenance media containing antibiotic supplements. Cells were incubated at 37oC, 5% CO2 for 24-48h prior to harvesting.

Luciferase reporter assays Reporters were transfected at equal concentrations and did not exceed 100ng per well. Transfection efficiency was controlled using a transfected Renilla luciferase reporter (pHRL-TK; Promega). At 48h post-

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

Cell lines

20 transfection, Firefly luciferase activity was determined using the Dual-Glo Luciferase Reporter Assay kit (Promega) and a BMG Fluostar Optima luminometer (BMG Labtech, Germany) as per the manufacturers instructions. Differences in transfection efficiency were corrected for by normalizing Firefly luciferase activity to that of Renilla luciferase.

Actinomycin D timecourse

by measuring the amount of luciferase mRNA at selected intervals: 0 (control), 3, 6 and 9h, following the addition of 5µg/mL Actinomycin D 48hrs post-transfection (Sigma-Aldrich). HT29 cells were transfected as described above. Time course intervals were chosen based on manufacturers data of luc2 mRNA half-life (approximately 2h).

Preparation of cDNA cDNA used in real-time quantitative PCR reactions was prepared from cell cultures containing approximately 2.5x106 cells in 24-well tissue culture plates (Nunc). Cells were washed with ice- cold PBS and RNA was extracted from cells using the innuPREP RNA Mini Kit (Analytik Jena). 10µL of each RNA preparation was used as template in the High-Capacity Reverse Transcription cDNA kit (Applied Biosystems) as per the manufacturers instructions. Following reverse transcription, cDNA was stored at 20oC until required.

Quantitative real-time PCR GAPDH was used as the reference gene in all qPCR experiments. qPCR was performed on an MX3000P real-time PCR machine (Stratagene). Reactions were assembled using Brilliant II qPCR Mastermix (Stratagene) in a total volume of 25µL. 2% of cDNA template was added to each qPCR reaction. Typically, primer concentrations were 900nM each, whereas probes were added at 400nM (final concentration). Data was analyzed using the MxPro software (ver 4.0).

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

The stability of luciferase mRNA transcripts from constructs used in earlier reporter assays was determined

21 Molecular Beacons probe design Molecular Beacons (MBs) were designed such that the region of complementarity was between 21 and 29nt. The annealing temperature of the probe:template region was typically 59-62oC, while stem region melting temperatures were ±2oC compared to the probe:template region to ensure the two melting temperatures were approximately 7oC above the annealing temperature of the PCR reaction. Additionally, each stem region was designed to ensure the extreme 5´ nucleotide was not a guanine (G) residue. The 5´ fluorophore was 6-

conjunction with MBs were specific to each probe and individual amplicon size did not exceed 230nt. Melting temperatures for each primer pair were closely matched and were typically ~60oC, allowing a consistent annealing temperature for the majority of qPCR experiments. All primer pairs were tested for target specificity in silico in either the UCSC In Silico PCR program (http://genome.ucsc.edu/cgi-bin/hgPcr) or the NCBI Primer-BLAST program, which uses the mRNA RefSeq database (http://www.ncbi.nlm.nih.gov/tools/primer-blast/). Primer:probe targets were designed to measure mRNA stability at selected intervals along the length of each mRNA.

Sequence databases and bioinformatic tools All nucleotide sequences were obtained from the National Centre for Biotechnology (NCBI) Entrez sequence database (http://www.ncbi.nlm.nih.gov). A number of web-based utilities were used to analyze the CFTR 5´UTR for post-transcriptional regulatory motifs: NCBI ORF Finder (http://www.ncbi.nlm.nih.gov/gorf/gorf.html), UTResource (http://www.ba.itb.cnr.it/UTR) (47) and the Cystic Fibrosis Mutation Database (http://www.genet.sickkids.on.ca/cftr/app).

RNA secondary structure prediction Preliminary RNA secondary structures were predicted using mFold 3.2 (9, 11) (http://mfold.rna.albany.edu) and Vienna RNAfold (http://rna.tbi.univie.ac.at/cgi-bin/RNAfold.cgi) (48). All folding computations were performed using default prediction parameters for 37oC. For the sliding window analysis, a 5'UTR sequence was processed into 30nt bins. Scrambled control sequences were created by shuffling the 5'UTR sequence from the original 30nt bins 100 times each. Dinucleotide distribution was preserved. Shuffling was performed using esl-shuffle from the HMMER 3.0 command-line package (hmmer.org). Each line of

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

Carboxyfluorescein (FAM), while the 3´ quencher was a Black-hole Quencher (BHQ). Primer pairs used in

22 sequence was processed using the command-line version of RNAfold (ViennaRNA-1.8.4; https://www.tbi.univie.ac.at/RNA/index.html). The predicted MFE and EFE structures for each sequence and the base-pairing probabilities of each nucleotide in the analyzed sequence were analyzed in Excel.

Statistical analysis Experimental data was analyzed using GraphPad Prism software (v6.0e). Statistical analysis was performed

error bars are expressed as mean ± standard error of the mean (SEM). Funding This work was supported by the University of Queensland. SWL was the recipient of a University of Queensland Research Scholarship.

Acknowledgments The authors wish to thank Dr. Richard J. Fish (The University of Geneva, Switzerland) and Dr. Prudence Donovan (EPFL, Switzerland) for critical reading of the manuscript and valuable discussions.

Conflict of Interest Statement The authors declare no conflict of interest

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

using unpaired Student’s t-test and linear regression. P values less than 0.05 were considered significant. All

23

References

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

1 Romey, M.-C., Pallares-Ruiz, N., Mange, A., Mettling, C., Peytavi, R., Demaille, J. and Claustres, M. (2000) A Naturally Occurring Sequence Variation That Creates a YY1 Element Is Associated with Increased Cystic Fibrosis Transmembrane Conductance Regulator Gene Expression. J. Biol. Chem., 275, 3561-3567. 2 Padegimas, L., Kowalczyk, T.H., Adams, S., Gedeon, C.R., Oette, S.M., Dines, K., Hyatt, S.L., Sesenoglu-Laird, O., Tyr, O., Moen, R.C. et al. (2011) Optimization of hCFTR Lung Expression in Mice Using DNA Nanoparticles. Mol. Ther., 20, 63-72. 3 Gebauer, F. and Hentze, M.W. (2004) Molecular mechanisms of translational control. Nat. Rev. Mol. Cell Biol., 5, 827-835. 4 Pichon, X., Wilson, L.A., Stoneley, M., Bastide, A., King, H.A., Somers, J. and Willis, A.E.E. (2012) RNA Binding Protein/RNA Element Interactions and the Control of Translation. Curr. Protein Pept. Sci., 13, 294-304. 5 Bugaut, A. and Balasubramanian, S. (2012) 5'-UTR RNA G-quadruplexes: translation regulation and targeting. Nucl. Acids Res., 40, 4727-4741. 6 Avraham, R. and Yarden, Y. (2012) Regulation of signalling by microRNAs. Biochem. Soc. Trans., 40, 26-30. 7 Kozak, M. (1989) Circumstances and mechanisms of inhibition of translation by secondary structure in eucaryotic mRNAs. Mol. Cell. Biol., 9, 5134-5142. 8 Lukowski, S.W., Bombieri, C. and Trezise, A.E.O. (2011) Disrupted post-transcriptional regulation of the cystic fibrosis transmembrane conductance regulator (CFTR) by a 5'UTR mutation is associated with a CFTR-related disease. Hum. Mutat., 32, E2266-2282. 9 Zuker, M. (2003) Mfold web server for nucleic acid folding and hybridization prediction. Nucl. Acids Res., 31, 3406-3415. 10 Vilela, C., Ramirez, C.V., Linz, B., Rodrigues-Pousada, C. and McCarthy, J.E. (1999) Posttermination ribosome interactions with the 5'UTR modulate yeast mRNA stability. EMBO J., 18, 3139-3152. 11 Mathews, D.H., Sabina, J., Zuker, M. and Turner, D.H. (1999) Expanded sequence dependence of thermodynamic parameters improves prediction of RNA secondary structure. J. Mol. Biol., 288, 911-940. 12 Jenkins, R.H., Bennagi, R., Martin, J., Phillips, A.O., Redman, J.E. and Fraser, D.J. (2010) A conserved stem loop motif in the 5'untranslated region regulates transforming growth factor-β(1) translation. PLoS ONE, 5, e12283. 13 Gruber, A.R., Lorenz, R., Bernhart, S.H., Neubock, R. and Hofacker, I.L. (2008) The Vienna RNA Websuite. Nucl. Acids Res., 36, W70-W74. 14 Le, S.Y. and Maizel, J.V. (1989) A method for assessing the statistical significance of RNA folding. J. Theor. Biol., 138, 495-510. 15 Shakin, S.H. and Liebhaber, S.A. (1986) Translational profiles of alpha 1-, alpha 2-, and beta-globin messenger ribonucleic acids in human reticulocytes. J. Clin. Invest., 78, 1125-1129. 16 Kozak, M. (1989) Context effects and inefficient initiation at non-AUG codons in eucaryotic cellfree translation systems. Mol. Cell. Biol., 9, 5073-5080. 17 Kozak, M. (1989) The scanning model for translation: an update. J. Cell Biol., 108, 229-241. 18 Wang, X.Q. and Rothnagel, J.A. (2004) 5'-untranslated regions with multiple upstream AUG codons can support low-level translation via leaky scanning and reinitiation. Nucl. Acids Res., 32, 1382-1391. 19 Trezise, A.E. and Buchwald, M. (1991) In vivo cell-specific expression of the cystic fibrosis transmembrane conductance regulator. Nature, 353, 434-437. 20 Trezise, A.E., Buchwald, M. and Higgins, C.F. (1993) Testis-specific, alternative splicing of rodent CFTR mRNA. Hum. Mol. Genet., 2, 801-802. 21 Trezise, A.E., Chambers, J.A., Wardle, C.J., Gould, S. and Harris, A. (1993) Expression of the cystic fibrosis gene in human foetal tissues. Hum. Mol. Genet., 2, 213-218. 22 Trezise, A.E., Linder, C.C., Grieger, D., Thompson, E.W., Meunier, H., Griswold, M.D. and Buchwald, M. (1993) CFTR expression is regulated during both the cycle of the seminiferous epithelium and the oestrous cycle of rodents. Nat. Genet., 3, 157-164. 23 Trezise, A.E., Ratcliff, R., Hawkins, T.E., Evans, M.J., Freeman, T.C., Romano, P.R., Higgins, C.F. and Colledge, W.H. (1997) Co-ordinate regulation of the cystic fibrosis and multidrug resistance genes in cystic fibrosis knockout mice. Hum. Mol. Genet., 6, 527-537.

24

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

24 Trezise, A.E., Romano, P.R., Gill, D.R., Hyde, S.C., Sepulveda, F.V., Buchwald, M. and Higgins, C.F. (1992) The multidrug resistance and cystic fibrosis genes have complementary patterns of epithelial expression. EMBO J., 11, 4291-4303. 25 White, N.L., Higgins, C.F. and Trezise, A.E. (1998) Tissue-specific in vivo transcription start sites of the human and murine cystic fibrosis genes. Hum. Mol. Genet., 7, 363-369. 26 Yoshimura, K., Nakamura, H., Trapnell, B.C., Dalemans, W., Pavirani, A., Lecocq, J.P. and Crystal, R.G. (1991) The cystic fibrosis gene has a "housekeeping"-type promoter and is expressed at low levels in cells of epithelial origin. J. Biol. Chem., 266, 9140-9144. 27 Chroneos, Z.C., Wert, S.E., Livingston, J.L., Hassett, D.J. and Whitsett, J.A. (2000) Role of cystic fibrosis transmembrane conductance regulator in pulmonary clearance of Pseudomonas aeruginosa in vivo. J. Immunol., 165, 3941-3950. 28 Elfakess, R. and Dikstein, R. (2008) A translation initiation element specific to mRNAs with very short 5'UTR that also regulates transcription. PLoS ONE, 3, e3094. 29 Vostrov, A.A., Taheny, M.J., Izkhakov, N. and Quitschke, W.W. (2010) A nuclear factor-binding domain in the 5'-untranslated region of the amyloid precursor protein promoter: Implications for the regulation of gene expression. BMC Res. Notes, 3, 4. 30 Simon, D., Laloo, B., Barillot, M., Barnetche, T., Blanchard, C., Rooryck, C., Marche, M., Burgelin, I., Coupry, I., Chassaing, N. et al. (2010) A mutation in the 3'-UTR of the HDAC6 gene abolishing the posttranscriptional regulation mediated by hsa-miR-433 is linked to a new form of dominant X-linked chondrodysplasia. Hum. Mol. Genet., 19, 2015-2027. 31 Tsai, N.P., Lin, Y.L. and Wei, L.N. (2009) MicroRNA mir-346 targets the 5'-untranslated region of receptor-interacting protein 140 (RIP140) mRNA and up-regulates its protein expression. Biochem. J., 424, 411-418. 32 Galban, S., Kuwano, Y., Pullmann, R., Martindale, J.L., Kim, H.H., Lal, A., Abdelmohsen, K., Yang, X., Dang, Y., Liu, J.O. et al. (2007) RNA-Binding Proteins HuR and PTB Promote the Translation of Hypoxia-Inducible Factor-1 alpha. Mol. Cell. Biol., in press. 33 Leontieva, O.V. and Ionov, Y. (2009) RNA-binding motif protein 35A is a novel tumor suppressor for colorectal cancer. Cell Cycle, 8, 490-497. 34 Pisareva, V.P., Pisarev, A.V., Komar, A.A., Hellen, C.U. and Pestova, T.V. (2008) Translation initiation on mammalian mRNAs with structured 5'UTRs requires DExH-box protein DHX29. Cell, 135, 1237-1250. 35 Yeh, C., Hung, L., Hsu, C., Le, S., Lee, P., Liao, W., Lin, Y., Chang, W. and Tseng, J.T. (2008) RNA-binding Protein HuR Interacts with Thrombomodulin 5'Untranslated Region and Represses Internal Ribosome Entry Site-mediated Translation under IL-1-beta Treatment. Mol. Biol. Cell, 19, 3812. 36 Hemmings-Mieszczak, M. and Hohn, T. (1999) A stable hairpin preceded by a short open reading frame promotes nonlinear ribosome migration on a synthetic mRNA leader. RNA, 5, 1149-1157. 37 Hemmings-Mieszczak, M., Hohn, T. and Preiss, T. (2000) Termination and peptide release at the upstream open reading frame are required for downstream translation on synthetic shunt-competent mRNA leaders. Mol. Cell. Biol., 20, 6212-6223. 38 Vilela, C., Linz, B., Rodrigues-Pousada, C. and McCarthy, J.E. (1998) The yeast transcription factor genes YAP1 and YAP2 are subject to differential control at the levels of both translation and mRNA stability. Nucl. Acids Res., 26, 1150-1159. 39 Vilela, C. and McCarthy, J.E. (2003) Regulation of fungal gene expression via short open reading frames in the mRNA 5'untranslated region. Mol. Microbiol., 49, 859-867. 40 Haimovich, G., Medina, D.A., Causse, S.Z., Garber, M., Millán-Zambrano, G., Barkai, O., Chávez, S., Pérez-Ortín, J.E., Darzacq, X. and Choder, M. (2013) Gene expression is circular: factors for mRNA degradation also foster mRNA synthesis. Cell, 153, 1000-1011. 41 Sun, M., Schwalb, B., Pirkl, N., Maier, K.C., Schenk, A., Failmezger, H., Tresch, A. and Cramer, P. (2013) Global analysis of eukaryotic mRNA degradation reveals Xrn1-dependent buffering of transcript levels. Mol. Cell, 52, 52-62. 42 McCarthy, J.E. (1998) Posttranscriptional control of gene expression in yeast. Microbiol. Mol. Biol. Rev., 62, 1492-1553. 43 Rajkowitsch, L., Vilela, C., Berthelot, K., Ramirez, C.V. and McCarthy, J.E. (2004) Reinitiation and recycling are distinct processes occurring downstream of translation termination in yeast. J. Mol. Biol., 335, 71-85. 44 Kozak, M. (1991) Structural features in eukaryotic mRNAs that modulate the initiation of translation. J. Biol. Chem., 266, 19867-19870.

25 45 Horvath, P., Suganuma, A., Inaba, M., Pan, Y.B. and Gupta, K.C. (1995) Multiple elements in the 5' untranslated region down-regulate c-sis messenger RNA translation. Cell Growth Differ., 6, 1103-1110. 46 Zhao, X., Chen, J., Lei, L., Hu, G., Xiong, Y., Xu, J., Li, Q., Yang, X., Chang, C.C., Song, B. et al. (2009) The optional long 5'-untranslated region of human ACAT1 mRNAs impairs the production of ACAT1 protein by promoting its mRNA decay. Acta Biochim. Biophys. Sin., 41, 30-41. 47 Pesole, G., Liuni, S., Grillo, G., Licciulli, F., Mignone, F., Gissi, C. and Saccone, C. (2002) UTRdb and UTRsite: specialized databases of sequences and functional elements of 5' and 3' untranslated regions of eukaryotic mRNAs. Update 2002. Nucl. Acids Res., 30, 335-340. 48 Hofacker, I.L. (2003) Vienna RNA secondary structure server. Nucl. Acids Res., 31, 3429-3431.

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

26 Legends to figures

Figure 1: CFTR 5´UTR schematic and folding energy analysis of the CFTR-132 5´UTR (A) Schematic diagram of the CFTR 5´UTR and part of the coding sequence showing the two major transcription start sites (bent-arrows), an upstream open reading frame and predicted stem-loop structure. The stem-loop structure abuts the double stop-codon of the uORF. The main translation initiation codon (M)

Ensemble free energy (EFE)(C) profiles of the CFTR-132 5'UTR compared to a scrambled sequence showing a region of lower free energy between the -64 to -30 nucleotides (window positions 68-102; window size = 30nt). (D) The base-pairing probabilities of each nucleotide in each 30nt window compared to a scrambled sequence reveal a higher probability of base-pairing between the -64 to -30 nucleotides. (E) The centroid structure determined by RNAfold for the region surrounding the predicted stem-loop. The hairpin is predicted to form immediately after the double stop codon of the CFTR-132 uORF and the transcription start site of the CFTR-69 5´UTR.

Figure 2: (A) CFTR-132 uORF activity in HT29 and HEK293 cells The luciferase activity of each construct (shown at left) was measured in HT29 and HEK293 cells (right). Compared to the CFTR-132 5'UTR, the CFTR-69 5'UTR was reduced by approximately 60% in both cell types (P < 0.0001). CFTR uORF mutant constructs were placed upstream of a luciferase coding region (luc2) in pGL4.13 plasmids to investigate the effects of the CFTR uORF upon translation initiation at the CFTR pAUG. Conversion of the uORF-AUG to a stop codon did not significantly affect luc2 reporter activity. Unexpectedly, improving the Kozak sequence of the uAUG to 1/2 critical nucleotides (from 0/2) increased luc2 activity compared to the wildtype CFTR-132 5´UTR in HT29 (155%; P = 0.0009) and HEK293 (198%; P = 0.0129) cells. Improving the Kozak sequence of the uAUG to 2/2 critical nucleotides had no effect on luc2 activity compared to the wildtype CFTR-132 5´UTR in either cell line. For each construct schematic, a single or double inverted arrowhead located at either the CFTR-uAUG or CFTRpAUG corresponds to a Kozak sequence with 1/2 or 2/2 optimised nucleotides respectively. The absence of

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

at the CFTR CDS is marked by a downward-pointing arrow. (B-C) Minimum free energy (MFE)(B) and

27 an arrowhead at these locations indicates a poor Kozak sequence with 0/2 optimised nucleotides. The ‘X’ indicates mutation of an AUG codon to UAG.

(B) Effects of CFTR pAUG Kozak context on translation initiation 5 ́ UTR constructs with mutated Kozak contexts were used to determine the impact of wildtype (1/2nt) vs. poor (0/2nt) or optimal (2/2nt) Kozak context on the translational efficiency of the CFTR pAUG codon.

variants were compared to the CFTR-69 wildtype construct. Conversion of the CFTR-132 pAUG Kozak context to 0/2 (poor) did not significantly affect luc2 activity, however, improving the CFTR pAUG context to 2/2 (optimal) increased luc2 reporter activity 176% (P = 0.03) and 203% (P = 0.0257) in HT29 and HEK293 cells respectively, compared to the wildtype CFTR-132 construct. Conversion of the CFTR-69 pAUG Kozak context to 0/2 (poor) reduced luc2 activity 4-fold in both HT29 (P = 0.0022) and HEK293 cells (P = 0.005) compared to the wildtype CFTR-69 pAUG construct. Improving the CFTR-69 pAUG Kozak context to 2/2nt increased luc2 activity 1.8- and 1.5-fold in both cell lines (HT29: P = 0.0318) compared to the wildtype CFTR-69 pAUG construct. The ability of the CFTR-132 uORF-AUG to initiate translation was tested by linking the majority of the uORF (excluding the stop codon) directly to the third codon of the luc2 reporter coding sequence. The uORF:luc2 construct directed an increase in translation initiation of 1.3-fold in HT29 cells (P = 0.0096) and 1.2-fold in HEK293 cells (n.s) compared to the wildtype CFTR-132 5´UTR control. For each construct schematic, a single or double inverted arrowhead located at either the CFTR-uAUG or CFTR-pAUG corresponds to a Kozak sequence with 1/2 or 2/2 optimised nucleotides respectively. The absence of an arrowhead at these locations indicates a poor Kozak sequence with 0/2 optimised nucleotides.

Figure 3: Predicted thermodynamics and stability of CFTR-132 and CFTR-69 5 ́ UTR wildtype and stem-loop mutant mRNA The mFold software (Mathews et al, 1999; Zuker, 2003) was used to predict the formation of secondary structures in the CFTR-132 and CFTR-69 5´UTR. These representative structures show the destabilization of

Downloaded from http://hmg.oxfordjournals.org/ at Washington University School of Medicine Library on October 7, 2014

CFTR pAUG Kozak variants were compared to the CFTR-132 wildtype construct, whereas CFTR-69

28 the stem-loop in the CFTR-132 wildtype (A) and the -132-SL2 (GG>CC) mutant 5´UTRs (B). Panel C and D show the effect of the stem-loop mutation in CFTR-69 wildtype (C) and the 69-SL2 mutant (D) 5´UTRs. The blue highlighted regions show the location of the mutation target in each 5´UTR while the boxed regions represent the area of the predicted inhibitory hairpins before and after mutagenesis. Arrows show the 5´-3´ direction of each 5´UTR.

the post-transcriptional level Changing the nucleotide sequence of the stem-loop (SL2; GG>CC; Table 2) significantly increased the reporter activity in HT29 (7-fold; P = 0.0002) and HEK293 (1.6-fold; P = 0.003). A double mutant (CFTR132/uAUG>UAG/SL2) construct tested the combined the effects of -132-uAUG>UAG and -132-SL2 mutant constructs on translation initiation at the CFTR pAUG. In HT29 cells, luc2 reporter activity increased 14fold (P < 0.0001) and 3.6-fold (P = 0.0009) in HEK293 cells, compared to the wildtype CFTR-132 5´UTR, suggesting a complex regulatory mechanism involving both the uORF and RNA secondary structure. For each construct schematic, a single or double inverted arrowhead located at either the CFTR-uAUG or CFTRpAUG corresponds to a Kozak sequence with 1/2 or 2/2 optimised nucleotides respectively. The absence of an arrowhead at these locations indicates a poor Kozak sequence with 0/2 optimised nucleotides. The ‘X’ indicates mutation of an AUG codon to UAG. * - P

CFTR mRNA expression is regulated by an upstream open reading frame and RNA secondary structure in its 5' untranslated region.

Post-transcriptional regulation of gene expression through 5' untranslated region (5'UTR)-encoded cis-acting elements is an important mechanism for th...
687KB Sizes 1 Downloads 6 Views