Cardiac expression of the cystic fibrosis transmembrane conductance regulator involves novel exon 1 usage to produce a unique amino-terminal protein.

Cystic fibrosis is caused by mutations in the cystic fibrosis transmembrane conductance regulator (CFTR) gene, which encodes a chloride channel present in many cells. In cardiomyocytes, we report that multiple exon 1 usage and alternative splicing produces four CFTR transcripts, with different 5'-untranslated regions, CFTR(TRAD-139), CFTR(-1C/-1A), CFTR(-1C), and CFTR(-1B). CFTR transcripts containing the novel upstream exons (exons -1C, -1B, and -1A) represent more than 90% of cardiac expressed CFTR mRNA. Regulation of cardiac CFTR expression, in response to developmental and pathological stimuli, is exclusively due to the modulation of CFTR(-1C) and CFTR(-1C/-1A) expression. Upstream open reading frames have been identified in the 5'-untranslated regions of all CFTR transcripts that, in conjunction with adjacent stem-loop structures, modulate the efficiency of translation initiation at the AUG codon of the main CFTR coding region in CFTR(TRAD-139) and CFTR(-1C/-1A) transcripts. Exon -1A, only present in CFTR(-1C/-1A) transcripts, encodes an AUG codon that is in-frame with the main CFTR open reading frame, the efficient translation of which produces a novel CFTR protein isoform with a curtailed amino terminus. As the expression of this CFTR transcript parallels the spatial and temporal distribution of the cAMP-activated whole-cell current density in normal and diseased hearts, we suggest that CFTR(-1C/-1A) provides the molecular basis for the cardiac cAMP-activated chloride channel. Our findings provide further insight into the complex nature of in vivo CFTR expression, to which multiple mRNA transcripts, protein isoforms, and post-transcriptional regulatory mechanisms are now added.


Introduction
Cystic fibrosis (CF) is an autosomal recessive disorder that causes severe multisystem disease (1). The cause of CF is mutation of the cystic fibrosis transmembrane conductance regulator (CFTR) gene (2)(3)(4) that encodes a cAMP-activated, PKA-dependent chloride channel (5,6) and is a member of the ABC transporter superfamily of genes (7). CFTR exhibits spatial and temporal regulation of expression (8)(9)(10), accompanied by different transcription start site usage (11) and alternative splicing (12,13). In addition to epithelial tissues, CFTR is expressed in cardiac muscle (14,15) and neuronal tissues (16,17), which may contribute to the pathogenesis of CF.
The expression of CFTR transcripts and cAMP-activated chloride currents in the heart have been demonstrated for humans and simians (18), rabbits (19), guinea pigs (20,21) and cats (22), but are undetectable in murine (23) and canine hearts (24). Previous studies have shown that CFTR transcripts expressed in the heart are alternatively spliced, resulting in the loss of exon 5 (25). We have previously demonstrated that CFTR mRNA is expressed in an epicardial (higher) to endocardial (lower) gradient across the left ventricular free wall (LVFW) of the rabbit heart, coinciding with a 2.5:1 gradient in the cAMP-activated chloride current density in ventricular myocytes (26). Further, this epicardial to endocardial gradient in CFTR expression is developmentally regulated, appearing in the first postnatal week (27), and lost in hypertrophic and failing hearts (28). The cardiac distributions of CFTR mRNA and functional channels are consistent with a role in the maintenance of the normal epicardial to endocardial gradients of ventricular repolarisation and action potential duration (APD) in the heart (14,26). Further, the overall reduction and loss of the gradient of CFTR expression during hypertrophy could contribute to delayed ventricular repolarisation (29) and loss of the gradient of repolarisation in hypertrophied hearts (30), both of which are known to be arrhythmogenic (31). Indeed, it has been shown that some CF patients exhibit an increased risk of ventricular arrhythmia (32,33); however, it is difficult to distinguish between a primary genetic cause due to the loss of CFTR expression and secondary effects due to pathological manifestations in the pulmonary system.
In this study, we investigated the regulation of temporal, regional and pathological changes in CFTR expression in the rabbit heart. We show that the majority of CFTR transcripts expressed in the heart initiate at unique transcription start sites and include novel alternative 5′-exons that replace the traditional CFTR exon 1. These alternative 5′ which results in a CFTR polypeptide with a distinct amino-terminus. Modulation in the levels of these cardiac-specific CFTR transcripts is responsible for the temporal, spatial and pathological changes in CFTR expression observed in the heart. Finally, the distal localisation of these cardiac-specific, alternative 5′-exons upstream of the traditional CFTR exon 1 suggests the presence of a distinct promoter region directing CFTR expression in the heart.

Experimental Procedures
Total RNA source and preparation. Triplicate New Zealand White rabbit tissue samples were collected from heart (embryo day-29; neonate day-7; juvenile week-3; adult year-1, left and right ventricular free wall, sham-operated control, and aortic-banded hypertrophy) and adult year-1 duodenum (control). All atrial tissue was removed from each cardiac sample harvested, except for embryo day-29 where whole hearts were used. Atrial tissue contributes less than 10% of the tissue weight to the embryo day-29 samples and will not significantly affect the comparison of CFTR expression during development. Moderate cardiac hypertrophy was induced in adult rabbits as previously described (34), resulting in an increased heart weight-to-body weight ratio of 26 ± 2%. Sham-operated animals underwent aortic mobilisation, but were not banded. Adult LVFW samples were dissected into 3 equal parts, along both the apical to basal and epicardial to endocardial axes, and the outer portions located at the apical epicardial, basal epicardial, apical endocardial and basal endocardial surfaces were taken for further analysis. All procedures were performed in accordance with the UK Animals (Scientific Procedures) Act 1986. Fresh tissues were homogenised in guanidinium thiocyanate solution (Fluka) and total RNA was isolated according to Chomczynski and Sacchi (35).

5'-Rapid amplification of cDNA ends (5'-RACE). CFTR transcription start sites were
identified by 5′-RACE as previously described (11). Total RNA (2 µg) was reverse transcribed using a CFTR exon 6 reverse primer (Table 1). First-strand cDNAs, tailed with dCTP, were subject to two rounds of hemi-nested PCR using reverse primers to exons 4, 3 or 2 and an anchor primer to the dC-tail. Amplicons were sequenced directly and subcloned into pLITMUS 28 (NEB).
Northern blot analysis. Rabbit CFTR cDNA (courtesy of Professor Burton Horowitz, University of Nevada School of Medicine, USA) was random-primer radiolabelled with [α-32 P] dCTP (3000 Ci mM, Amersham; DECAprime II system, Ambion). Total RNA was electrophoresed on a 1% (w/v) denaturing agarose gel, transferred to a positively charged nylon membrane (GeneScreen Plus, NEN Life Sciences), incubated with labelled probe and then washed as previously described (13). Autoradiography was with Hyperfilm MP X-ray film (Amersham Pharmacia) for 5 days.
Each qPCR reaction consisted of 1 x TaqMan Universal PCR Master Mix (Applied Biosystems), forward and reverse primers (250 nM), probe (50 nM) and cDNA template (100 ng). Triplicate experimental CFTR and endogenous 18S rRNA reactions were performed using three independent cDNA samples from different animals. Reporter dye fluorescence was detected using an Applied Biosystems Prism 7700 Sequence Detector and data were analysed using Sequence Detector software v1.6.3 (Applied Biosystems). Quantification of CFTR gene expression, relative to 18S rRNA, was performed using the comparative threshold cycle (CT) method according to manufacturer's instructions (Applied Biosystems). After 72 h incubation, cells were trypsinised and transferred to a Thermo-Fast 96 black PCR plate (Abgene). Enhanced GFP and eBFP fluorescence was measured using a SPECTRAmax GEMINI-XS spectrofluorometer (36,37). Prior calibration experiments determined the following optimal parameters: eGFP; λ EX = 472 nm, λ EM = 512 nm and λ CUTOFF = 495 nm; and eBFP; λ EX = 378 nm, λ EM = 445 nm and λ CUTOFF = 420 nm. To correct for differences in transfection efficiency, eGFP fluorescence values were normalised to eBFP signal. Background fluorescence (untransfected cells) was subtracted.

Construction of CFTR
Bioinformatic and statistical analyses. Messenger RNA secondary structures were predicted using MFold v3.1 (38,39). CFTR amino-termini hydropathy plots were generated using the Kyte and Doolittle algorithm (40). Protein secondary structures were predicted by PSIPRED v2.4 (41). The charge distribution for the putative CFTR amino-terminal helix was predicted by helical wheel analyses using MacPlasmap Pro v3.01. Statistical comparisons were made using Student's t-tests (unpaired, two-tailed) and a p value of less than 0.05 was taken to indicate a change in transcript level was of statistical significance.

CFTR transcripts expressed in rabbit heart include multiple alternative 5′-exons
The major in vivo transcription start sites of CFTR transcripts expressed in the heart were identified by 5′-RACE. Typical examples of 5′-RACE products obtained from adult LVFW and duodenal (control) tissues are shown in Figure 1A designated exon -1C, exon -1B and exon -1A, respectively (Genbank: AY256886-AY256888). Exon -1C and exon -1B are spliced directly to exon 2, with exon -1A subject to alternative splicing between exons -1C and 2 (CFTR -1C , CFTR -1B and CFTR -1C/-1A transcripts). All splice donor and acceptor sequences satisfied the GT-AG rule (42). Figure  1B shows the genomic structure and alternative splicing of CFTR transcripts expressed in the heart. In all cases, amplification reactions from control untailed cDNA were negative.
Identical CFTR transcription start sites were mapped using three independent samples for all tissue types, confirming the precise location of each identified transcription start site and ensuring artefacts due to partial mRNA degradation were avoided.
Transcripts that included exon -1A contain an AUG codon in-frame with the main CFTR ORF. Translation initiation from this AUG codon would result in CFTR protein with a novel amino-terminal region. Exons -1B and -1C did not contain an AUG codon in-frame with the main CFTR ORF. However, many upstream AUGs (uAUGs) were identified in the 5′-UTR of CFTR transcripts including exons -1C, -1B or -1A and a single uAUG codon was identified in the 5′-UTR of CFTR TRAD-139 . These uAUGs defined the start of several uORFs, which were followed by translation termination codons and thus encoded short peptides. The presence of short uORFs in the 5′-UTRs of eukaryotic mRNA transcripts is indicative of post-transcriptional and translational regulation of gene expression (43).
Database comparisons of each putative short polypeptide sequence revealed no homology to known protein sequences.

Genomic DNA structure and transcriptional control of CFTR expression in the heart
Rabbit genomic DNA (gDNA) spanning CFTR exon -1C to exon -1A, was amplified by long-range PCR and the sequence compared to the human and mouse gDNA sequences (GenBank: AC000111 and AF162137; Figure 1C; also, supplemental data showing dot matrix comparisons of rabbit, mouse and human sequences). We also compared CFTR traditional exon 1 and flanking sequences across the three species. In both exonic and intronic gDNA regions, human and rabbit sequences showed a high level of identity for both the CFTR exon -1C to -1A region and traditional exon 1( Figure 1C). However, only CFTR traditional exon 1 sequences showed any similarity between the mouse and the other species. The human homologues of rabbit CFTR exons -1C to -1A spanned a region located 6 to 9 kb upstream of the traditional human CFTR exon 1. Further, the short uORFs and the rabbit CFTR exon -1A AUG codon, that is in-frame with the main CFTR ORF, are conserved in the human genome.
The presence of homologues of the rabbit alternative exon 1 sequences in the human genome, but not the mouse, is consistent with expression of CFTR in the hearts of many species, including the rabbit ( Figure 3A) (15) and human (18), but not in the mouse (tested by RNA in situ hybridisation and two rounds of reverse transcription-polymerase chain reaction (RT-PCR), unpublished data). Only 2 kb of genomic DNA sequence upstream of rabbit CFTR traditional exon 1 is available and the alternative CFTR exons -1A to -1C lie beyond that.
Long-range PCR between CFTR traditional exon 1 and exons -1A to -1C, using rabbit gDNA template, was unsuccessful. Therefore, the promoter region controlling the expression of CFTR exons -1C, -1B and -1A is likely to be distally located and cardiac specific.

Greater than 90% of CFTR transcripts expressed in the heart initiate at exon -1C or -1B
Total CFTR mRNA expression in rabbit adult LVFW was similar to the level of CFTR expression detected in duodenal epithelium. A 6.5 kb CFTR transcript was detected in both tissues (Figure 2A), which is consistent with known CFTR transcript size (3). Past studies have also shown alternative splicing of exon 5 (25), and changes in total CFTR expression during heart development (27) and in cardiac hypertrophy (28). To investigate the biological relevance of the multiple alternative CFTR transcripts, we measured the relative expression level of individual CFTR transcripts in rabbit heart tissue from embryo day-29, neonate day-7, juvenile week-3 and normal adult left ventricle (LV) and right ventricle (RV).
We used TaqMan quantitative PCR (qPCR) to quantify each distinct CFTR transcript: CFTR -1C, CFTR -1B, CFTR -1C/-1A, CFTR TRAD-139, CFTR EXON5+, and CFTR EXON5-. In all cases, amplicons were less than 150 bp, the standard curve plots showed a very high correlation coefficient (R 2 > 0.99, p < 0.01) and amplification efficiencies were close to 100%. The expression levels of each CFTR transcript were measured in arbitary relative expression units and normalised to 18S rRNA. We compared CFTR expression in whole heart (atria and ventricles) from embryo day-29, and ventricular tissues from neonate day-7, juvenile week-3 and normal adult year-1 left ventricle (LV) and right ventricle (RV). Atrial tissue contributes less than 10% of the tissue weight to the embryo day-29 samples and will not significantly affect the comparison of CFTR expression during development.
The predominantly expressed CFTR transcript in early rabbit cardiac development was CFTR -1B , with lower levels of the CFTR -1C and CFTR -1C/-1A transcripts ( Figure   2B). As development proceeds the expression of CFTR -1B transcripts decreased (p < 0.05), while expression of both CFTR -1C and CFTR -1C/-1A transcripts increased and became the predominant CFTR transcripts in the adult rabbit heart (p < 0.05 and p < 0.05, respectively).
In contrast, the expression of CFTR TRAD-139 transcripts was low and static. In adult cardiac tissue, greater than 90% of total CFTR mRNA transcripts include the novel alternative exons (exon -1C, exon -1B and exon -1C/-1A) described here. This contrasts with duodenum where more than 95% of CFTR transcripts initiate from traditional CFTR exon 1.
Analysis of CFTR exon 5 alternative splicing showed equal levels of exon 5+ and exon 5-CFTR transcripts at embryonic and neonatal stages ( Figure 3C). In hearts from juvenile and adult animals, there were differential increases in expression of both CFTR EXON5+, and CFTR EXON5-isoforms. This resulted in CFTR EXON5-transcripts being 3-fold more abundant than CFTR EXON5+ transcripts in the LVFW of adult hearts (p < 0.05). This also contrasts with CFTR expression in adult duodenum, where over 90% of CFTR transcripts include exon 5.

The expression of CFTR transcripts initiating at exon -1C are primarily responsible for the epicardial to endocardial gradient across the LVFW
Epicardial (higher) to endocardial (lower) gradients across the rabbit LVFW, have been shown for both cAMP-activated chloride currents and CFTR expression (26).
Quantitative analysis of the differential distribution of each CFTR transcript in the adult LVFW shows that CFTR -1C and CFTR -1C/-1A transcripts, as well as alternatively spliced CFTR EXON5-transcripts, are responsible for the epicardial to endocardial gradient of CFTR expression and function in the left ventricle ( Figure 3A (i and iii) and Figure 3B, p < 0.05 in all cases). In contrast, there were no substantial differences in epicardial versus endocardial expression of CFTR -1B , CFTR TRAD-139 or CFTR EXON5+ transcripts ( Figure 3A (ii and iv) and Figure 3B). This work has also identified a second, perpendicular gradient in CFTR expression along the apical to basal axis of the left ventricle ( Figure 3A and 3B). All three cardiacspecific CFTR transcripts (CFTR -1C , CFTR -1B and CFTR -1C/-1A ) and both exon 5 alternatively spliced forms contribute to the apical to basal gradient (p < 0.05 in all cases).
Collectively, these data show that CFTR expression is distributed in a radial pattern across the LVFW: highest at the apical epicardial surface and decreasing radially to the lowest point at the basal endocardial surface.

Cardiac hypertrophy causes the loss of the epicardial to endocardial gradient, but does not alter the apical to basal gradient of CFTR expression
Cardiac ventricular hypertrophy is associated with a loss of repolarising ion currents, including CFTR (28), and prolongation of the ventricular action potential duration (44). Here we show that the loss of the epicardial to endocardial CFTR expression gradient, in hypertrophic hearts, is due to the preferential down-regulation of CFTR -1C and CFTR -1C/-1A transcripts ( Figure 3A; p < 0.05 in all cases), while CFTR -1B and CFTR TRAD-139 transcripts were unaffected ( Figure 3A (ii and iv)). While cardiac hypertrophy resulted in the loss of the epicardial to endocardial CFTR expression gradient, the apical to basal gradient was unaffected.
Cardiac hypertrophy also differentially affected the expression of exon 5 alternatively spliced CFTR transcripts ( Figure 3B). The establishment of cardiac hypertrophy leads to a loss of the epicardial to endocardial gradient of CFTR EXON5-transcripts, whereas CFTR EXON5+ transcripts remain evenly distributed across the epicardial to endocardial axis. Further, both exon 5 alternatively spliced transcripts contribute to the apical to basal gradient in CFTR expression. The overall increase in CFTR expression during development reflects a preferential accumulation of CFTR -1C and CFTR -1C/-1A transcripts and a shift in exon 5 alternative splicing such that CFTR EXON5-transcripts predominate. This process is reversed in cardiac hypertrophy with preferential loss of CFTR -1C, CFTR -1C/-1A and CFTR EXON5-transcripts.

The translation initiation codon in exon -1A supports CFTR protein production
In the absence of traditional CFTR exon 1, Carroll and co-workers (45) (43). Also, small differences in the sequence immediately surrounding the AUG codon, compared to the Kozak consensus, may contribute to the efficiency of translation.

Translation of CFTR -1C/-1A mRNA results in CFTR protein with an unique aminoterminus
As CFTR exon -1A splices directly to exon 2, the 17 amino acids encoded by traditional exon 1 are omitted during translation of CFTR -1C/-1A transcripts and 2 other amino acids, encoded by exon -1A, constitute the new CFTR amino-terminus ( Figure 6A).
The efficiency of translation of both the CFTR TRAD-139 and CFTR -1C/-1A transcripts are indistinguishable ( Figure 4B). As CFTR -1C/-1A transcripts are approximately 6-fold more abundant in adult heart, than CFTR TRAD-139 transcripts, it is likely that the majority of CFTR protein expressed in the heart is the curtailed amino-terminus isoform reported here (CFTR -1C/-1A protein). We have used computational protein secondary structure and hydropathy analysis to predict possible differential functional roles of CFTR -1C/-1A protein.
Hydropathy analyses over the first 150 amino acids of both full-length and truncated CFTR isoforms were performed using the Kyte and Doolittle algorithm (40). The full-length CFTR amino-terminus is predominantly hydrophilic, which is consistent with previous work (3) and a predicted cytoplasmic localisation. However, we report here the identification of a particularly hydrophobic region, corresponding to amino acids 11-26 of the amino-terminal end and encoded across the exon 1/exon 2 boundary. This amino-terminal hydrophobic region is completely absent in the curtailed CFTR -1C/-1A isoform (first described by Davies, W. L., Ph.D. thesis, 2002; Figure 6B) (52). In contrast, the CFTR amino acid region involved in binding to syntaxin 1A (53, 54) was found to be entirely hydrophilic in nature ( Figure 6B, shaded zone), encoded by exons 2 and 3, and so still present in the aminoterminal truncated CFTR -1C/-1A isoform.
CFTR amino-terminal secondary structure was predicted by PSIPRED. Coinciding with the amino-terminal hydrophobic region described above (amino acids 11-26) is a putative helical structure (predicted with a high level of confidence) that spans the boundary between traditional exon 1 and exon 2 ( Figure 6C). Helical wheel analysis of this putative, hydrophobic helix predicts clustering of three positively charged residues ( Figure 6D). The removal of the amino-terminal half of this putative helix, as would be the case in the CFTR -1C/-1A isoform, is accompanied by a marked reduction in the confidence of prediction of the remainder of the helix. Both secondary structure and helical wheel predictions were used by Naren and co-workers (54) in the identification of a helix that has been confirmed to interact with syntaxin 1A and regulate CFTR channel activity.

Discussion
The distribution and control of CFTR gene expression is tightly regulated by temporal, spatial and tissue-specific mechanisms (8)(9)(10). Multiple in vivo transcription start sites have been localised within 2 kb of the AUG translation initiation codon in traditional CFTR exon 1 (3,11,55). This is the first study showing differential CFTR in vivo transcription start site usage in the heart that is regulated by temporal, spatial and pathophysiological stimuli. CFTR transcripts expressed in the heart initiate at three distinct exon 1 sequences (exon -1C, exon -1B and traditional exon 1 initiating at -139 bp), each of which splice directly to exon 2, with exon -1A alternatively spliced between exon -1C and exon 2. In the heart, the only previously identified alternative splicing of CFTR transcripts involved the differential exclusion of exon 5 (12). Exons -1C to -1A are themselves distributed over 2.5 kb of the rabbit genome and are located distal to the traditional CFTR exon 1.
Comparative phylogenetic analysis identified human homologues of both the rabbit exon -1C to -1A and traditional exon 1 CFTR regions, localised 10 kb upstream of human CFTR and at traditional exon 1, respectively. However, comparison with murine gDNA only revealed homology to traditional CFTR exon 1. These findings are consistent with speciesspecific, molecular and functional distributions of CFTR expression in the heart, identified in humans (18) and rabbits (15,25) but not in mice (23). Collectively, these findings suggest that CFTR transcription from exons -1C to -1A is cardiac-specific and controlled by a distinct, and previously unidentified, promoter. Also, this work provides support for the development of a rabbit model of cystic fibrosis, to allow investigation of some pathophysiological features of human CF disease that are not present in murine models, such as cardiac involvement in CF (56).
Spatial, developmental and pathophysiological signals regulate CFTR expression in the heart, with the development of a left ventricular epicardial to endocardial gradient during the late foetal and neonatal periods, and loss of this gradient following hypertrophic stimuli (26)(27)(28). We have also identified an apical (higher) to basal (lower) gradient in CFTR expression that is unaffected by hypertrophy. This defines a radial pattern of CFTR expression across the left ventricle, correlating well with the spread of repolarisation throughout the left ventricle. These findings further support the view that the cAMPstimulated chloride current, encoded by CFTR, contributes to ventricular repolarisation and differential action potential duration throughout the heart. The expression of multiple CFTR mRNA transcripts in the heart raises questions of differential regulation of specific transcripts by various signals.
During cardiac development there is a preferential accumulation of CFTR -1C, CFTR -1C/-1A and CFTR EXON5-transcripts. Concomitantly, there is a preferential loss of CFTR -1B transcripts. Also, the epicardial to endocardial gradient is primarily due to CFTR -1C, CFTR -1C/-1A and CFTR EXON5-transcripts, while all cardiac-specific CFTR transcripts contribute to the apical to basal ventricular CFTR gradient. In contrast, CFTR transcripts initiated at traditional exon 1 are present at very low, static levels (less than 10%). Analysis of left ventricular CFTR gradients in the hypertrophic heart demonstrated a loss of the epicardial to endocardial gradient but not the apical to basal gradient, culminating in an overall decrease in the radial gradient of ventricular CFTR expression. Again, individual CFTR transcripts were differentially regulated, with a preferential loss of CFTR -1C, CFTR -1C/-1A and CFTR EXON5-transcripts, but no change in CFTR -1B and CFTR TRAD-139 expression. The overall increase in CFTR expression during heart development, and loss in cardiac hypertrophy is consistent with the "re-expression of foetal gene program hypothesis" to explain global gene expression changes in cardiac hypertrophy (57,58). However, our findings extend that hypothesis to include differential effects on individual CFTR transcripts.
This demonstration of the differential regulation of individual CFTR transcripts suggests further complexity in the regulation of CFTR expression, with mechanisms that allow transcript-specific interpretation of spatial, temporal and pathologic signals.
We have identified CFTR regulatory mechanisms that involve both alternative splicing, and differential transcription start site and exon 1 usage. An important consequence of differential exon 1 usage is the generation of four CFTR transcripts with distinct 5′-UTRs, all of which encode one or more uORFs. There is increasing evidence that posttranscriptional regulation of gene expression, through modulation of mRNA stability and translation initiation, is achieved through 5′-UTR encoded elements, such as uORFs and stem-loop secondary structures (43,48,59). While the cis-elements controlling posttranscriptional regulation of gene expression were first identified in yeast, similar mechanisms are now known in mammalian cells. To date over two-thirds of identified mammalian genes that encode uORFs in their 5′-UTRs are proto-oncogenes (60)(61)(62)(63). However, uORFs have also been identified in a few genes with functions unrelated to cell growth control, the S-Adenosylmethionine Decarboxylase gene (64,65) and Huntingtin gene (66). To these examples we now add the CFTR gene. Further, the causative mutations in the inherited diseases of familial melanoma (67,68) and thrombocythaemia (69) create or abolish uAUGs that result in dramatic alterations in steady-state mRNA and protein levels.
It is estimated that less than 10% of eukaryotic mRNAs have an uORF in their 5′-UTR, however, very few have been investigated (48,60). In general, uORFs lead to destabilisation of the mRNA of the main ORF, secondary to disruption of ribosome scanning and reduced translation initiation at the main ORF (43). All CFTR transcripts expressed in the heart have the necessary 5′-UTR elements, uORFs with adjacent stem-loop structures, to allow post-transcriptional mechanisms to contribute to the regulation of CFTR expression.
We have shown that the uORF, probably acting in concert with the adjacent stem-loop, encoded in CFTR TRAD-139 transcripts functions to reduce translation initiation efficiency at the downstream AUG of the main CFTR ORF. Also, the uORFs encoded in exon -1C have a similar effect. With reduced translation resulting in mRNA destabilisation (43), it is probable that differential stability of CFTR transcripts in the heart may be a key factor governing spatial, temporal and pathological changes in CFTR expression. Indeed, in cardiac tissues, different levels of CFTR expression were measured over exons 1 to 2 compared to exons 4 to 6 for all CFTR transcripts, suggesting that the 5′-UTR elements involved in modulating translation efficiency from the main AUG codon may also have a role in modulating differential stability along the CFTR transcript. However, the precise control of in vivo CFTR transcript stability remains unclear.
The discovery that uORFs in CFTR 5′-UTRs modulate CFTR protein production has implications beyond the expression of CFTR in the heart, as most mouse tissues express CFTR transcripts that include an uORF (11). Also, the CFTR transcripts expressed in human foetal lung, but not adult lung, include an uORF (11), allowing the possibility that posttranscriptional mechanisms may contribute to the large changes in CFTR expression that occur during lung development. We have identified a conserved stable stem-loop structure in the 5′-UTR of traditional CFTR exon 1 that would be included in the majority of CFTR transcripts expressed in all tissues. Similar stem-loop structures have been shown to alter translation efficiency (49,50). Whether 5′-UTR encoded secondary structures lead to widespread post-transcriptional regulation of CFTR expression is under further investigation.
The cardiac potassium channels encoded by the HERG and KvLQT1 genes are subject to alternative exon 1 usage, resulting in distinct amino-terminal protein isoforms (70).
Similarly, we have shown that CFTR -1C/-1A transcripts contain an AUG codon that is inframe with the main CFTR ORF and directs translation of an unique CFTR protein isoform, bearing a 15 amino acid truncation at the amino terminus. CFTR -1B and CFTR -1C transcripts do not encode an in-frame AUG, however, translation may initiate from downstream AUG codons in exons 3 and 4 (45). The substantially higher levels of CFTR -1C/-1A transcripts suggest that the majority of CFTR protein expressed in the heart is the truncated amino-terminal isoform.
The regulation of CFTR protein trafficking and channel activity has been shown to involve the amino-terminus. Hydrophilic residues of the H3 domain of syntaxin 1A physically interact with a cytoplasmic amino-terminal helix of CFTR, encoded in exons 2 and 3 (71,72). Furthermore, an exon 1 encoded diphenylalanine motif (F16/F17) regulates CFTR membrane trafficking (73). It has been suggested that loss of CFTR exon 1 would remove both motifs (55). However, while the diphenylalanine motif would be removed, the CFTR domain that interacts with syntaxin 1A would remain as it is encoded within exons 2 and 3. It was also suggested that the CFTR domain that interacts with syntaxin 1A is hydrophobic in nature and encoded at the exon 1/2 boundary (55). However, hydrophobicity analysis demonstrates that the syntaxin 1A interacting domain is in fact hydrophilic ( Figure 6B) (52).
Our analysis does indeed identify a CFTR amino-terminal hydrophobic region (residues 11 to 26) encoded at the exon 1/2 boundary, first described by Davies (2002) (52). This hydrophobic region, truncated by loss of exon 1, has a predicted helical structure and may affect CFTR function or subcellular location by its presence or absence. The putative helical structure of the CFTR amino-terminus (residues 11 to 26) is supported by the recent identification, from the crystal structure (4.5 Å), of an alpha helix of similar length and location within the amino-terminus (residues 10 to 21) of the ABC transporter homologue, MsbA, from E. coli (74).
We, and others, have established an absolute correlation between CFTR mRNA expression, and the species-specific distribution of cardiac cAMP-activated chloride currents (14), present in rabbits (15,26), guinea pigs (20,21), monkeys and humans (18). Antisense oligodeoxynucleotide inhibition studies have confirmed CFTR as the molecular basis of the in vivo cAMP-activated chloride current present in epithelial cells and cardiomyocytes.
Primary cultures of sweat duct epithelial cells (75) and ventricular cardiomyocytes (25), as well as pancreatic duct (76) In all cases, sense oligodeoxynucleotides had no effect on the cAMP-activated chloride current and neither sense nor antisense oligodeoxynucleotides had any effect on calcium-activated chloride conductance. Our findings, that CFTR TRAD-139 mRNA accounts for only a small proportion of the CFTR protein-coding transcripts present in the heart, provides an explanation for the partial inhibition of the cAMP-activated chloride current in ventricular myocytes described by Hart et al. (25). The cardiac-specific CFTR -1C/-1A transcript we have identified will not bind the antisense oligodeoxynucleotides used by Hart et al. (25), and their findings support our conclusion that CFTR -1C/-1A mRNA codes the majority of CFTR protein and cAMP-activated chloride current present in the heart. Further, it is specifically the differential expression of the CFTR -1C/-1A transcript that is absolutely correlated with differences in the whole-cell current density of cAMP-activated chloride currents in epicardial versus endocardial ventricular myocytes (26), and the loss of cAMP-activated chloride conductance in cardiac hypertrophy (78).
Overall this work identifies multiple new levels at which CFTR expression is regulated in vivo. This is the first study to show the in vivo post-transcriptional regulation of CFTR expression through modulation of translation initiation efficiency by 5′-UTR encoded elements. Furthermore, this is probably a widespread mechanism regulating CFTR expression. We are also the first to show that through alternative exon 1 usage an unique isoform of CFTR protein is generated and is likely to be the major CFTR isoform present in the heart. This novel curtailed form of CFTR protein, CFTR -1C/-1A, is missing functionally important amino-terminal motifs and is predicted to display unique subcellular localisation and activity.         GC rich region, no tail; Pu.Py stretch, circular tail; and an indel unit, diamond-shaped tail.
Horizontal bars represent a scale of 500 bp.