Mutations that alter RNA splicing of the human HPRT gene: a review of the spectrum

Abstract

The human HPRT gene contains spans approximately 42,000 base pairs in genomic DNA, has a mRNA of approximately 900 bases and a protein coding sequence of 657 bases (initiation codon AUG to termination codon UAA). This coding sequence is distributed into 9 exons ranging from 18 (exon 5) to 184 (exon 3) base pairs. Intron sizes range from 170 (intron 7) to 13,075 (intron 1) base pairs. In a database of human HPRT mutations, 277 of 2224 (12.5%) mutations result in alterations in splicing of the mRNA as analyzed by both reverse transcriptase mediated production of a cDNA followed by PCR amplification and cDNA sequencing and by genomic DNA PCR amplification and sequencing. Mutations have been found in all eight 5′ (donor) and 3′ (acceptor) splice sequences. Mutations in the 5′ splice sequences of introns 1 and 5 result in intron inclusion in the cDNA due to the use of cryptic donor splice sequences within the introns; mutations in the other six 5′ sites result in simple exon exclusion. Mutations in the 3′ splice sequences of introns 1, 3, 7 and 8 result in partial exon exclusion due to the use of cryptic acceptor splice sequences within the exons; mutations in the other four 3′ sites result in simple exon exclusion. A base substitution in exon 3 (209G→T) creates a new 5′ (donor) splice site which results in the exclusion of 110 bases of exon 3 from the cDNA. Two base substitutions in intron 8 (IVS8-16G→A and IVS8-3T→G) result in the inclusion of intron 8 sequences in the cDNA due to the creation of new 3′ (acceptor) splice sites. Base substitution within exons 1, 3, 4, 6 and 8 also result in splice alterations in cDNA. Those in exons 1 and 6 are at the 3′ end of the exon and may directly affect splicing. Those within exons 3 and 4 may be the result of the creation of nonsense codons, while those in exon 8 cannot be explained by this mechanism. Lastly, many mutations that affect splicing of the HPRT mRNA have pleiotropic effects in that multiple cDNA products are found.

Introduction

Most genes are divided into coding regions (exons) and intervening non-coding regions (introns). Transcription of genomic DNA creates large pre-RNAs from which the intervening non-coding regions must be precisely removed (spliced) to create functional mRNAs. Splicing is performed by what has been called the spliceosome of which 5 small nuclear RNA (snRNAs) [U1, U2, U4, U5 and U6] are the major components (reviewed in Refs. 1, 2, 3, 4, 5). A number of proteins are also important in this process including the serine-rich proteins (SR proteins) (reviewed in Ref. [4]) and proteins that bind heterogenous nuclear RNA (hnRNP proteins) (reviewed in Ref. [6]). SR proteins contain an amino-terminal RNA recognition motif which is widely present in RNA binding proteins and an arginine-serine dipeptide repeat region (RS domain) at their C terminus which is considered diagnostic of splicing proteins. Two examples are SF2/ASF and SC35. While these proteins were originally thought to be non-specific as they could be switched between organisms, more recent work has shown specific binding and regulation of different genes by different SR proteins [7]. The SR proteins bind to exon sequences, especially those 25 bases upstream of the 3′ end of the exon [5]. The hnRNP are found only in organisms with large introns and contain an RNP domain and protein–protein recognition domains [6]. They appear to recognize intronic sequences and could be involved in exon juxtaposition [8].

The snRNAs and proteins recognize consensus sequences on the DNA at or near the splice junctions. These consensus sequences will be discussed in detail below but the most important elements are: (1) a gt as the first 2 bases of the intron (the 5′ splice or splice donor site), (2) an ag as the last two bases of the intron (the 3′ splice or splice acceptor site), (3) a pyrimidine-rich tract usually just upstream of the acceptor site and (4) an `a' upstream of the pyrimidine-rich tract (branch site).

Briefly, splicing proceeds by ordered binding of 4 complexes (E, A, B, C) at the splice sites. Initial binding of U1 occurs at 5′ splice sites (E complex) while U2 binds to the branch point region (A complex) by complementary base pairing with the help of U2 auxiliary factor (U2AF) bound to the pyrimidine rich sequence. U1 is replaced by U4–U5–U6 with complementary base-pairing of U6 with the 5′ splice site. Complementary base-pairing of U2 and U6 bridges across the intron (B complex) with release of U4 [9]. Splicing commences by U2 facilitating a bulging out of the branch A and the 2′ hydroxyl of A's `attack' of the phosphodiester bond between the last base of the proceeding exon and the first base of the intron (C complex). This causes a lariat to be formed with the branch A attached to the first base of the intron. The `free' 3′ OH of the 5′ exon, held by U5 [10], then `attacks' the last base of the intron, attaching the 2 exons together and releasing the intron and associated snRNPs and proteins as a lariat (I complex).

At least one way that the SR proteins appear to facilitate these steps is by strengthening binding of the complexes to the RNA, especially by recruiting U1. They are especially crucial when the 5′ and 3′ site sequences are less than optimal. In fact, U1 is apparently dispensable if sufficient SR proteins are present 11, 12. Heterogenous ribonucleoprotein particles (hnRNPs) appear to compete with SR binding and are important in regulating alternative splicing (see below). For example, adjusting relative levels of SF2/ASF and hnRNP A1 promotes exon inclusion or exclusion, respectively 13, 14. SR proteins can be phosphorylated at the serines in the RS domains and this may play an important part in their function, regulation or specificity [4]. Polypyrimidine tract binding protein (PTB) binds in vitro to sequences that are very similar to the polypyrimidine tract in the alternatively spliced exons of rat α and β tropomyosin and can apparently suppress U2AF binding to pyrimidine sequences [15].

When introns were first discovered, discussion focused on the intron as a recognized unit that is spliced out; however, there are a number of difficulties with this model. Introns can be hundreds to thousands of basepairs long, and to scan down the DNA from one end to find the next is not a feasible plan. Also, this model would predict that mutations of consensus splice sites would cause inclusion of introns in the final mutant mRNA because they cannot be spliced out. This is rarely observed. Instead, the most common result of a splice site mutation is exclusion of the adjacent exon from the mRNA (i.e., the upstream exon for 5′ or donor site mutations and the downstream exon for 3′ or acceptor site mutations). In mammalian genes, splice mutations occur at these frequencies: exon skipping—51%, splicing at a nearby spot—32%, creation of a pseudo exon in an intron—11% and intron retention—only 6% [16]. Also, in in vitro studies, mutations in the donor site affect splicing of both the previous intron and the intron in which they occur [17]. Lastly, surveys of exons have demonstrated that exons have a clear maximum size of about 300 bp; very few exons are larger than this (95% are less than 300 bp) 8, 18. In vitro studies of constructs with increasing exon lengths show that splicing either is diminished or cryptic sites within the exons are utilized when the exon reaches 300 bp 8, 19. These studies led investigators to develop the exon definition model of splicing whereby binding of splicing factors at either end of the exon defines the exons; these exons are then spliced together to form the mRNA (exon juxtaposition) 8, 17, 19. If either one of the splice sites at the ends of the exon do not function, then the exon is not recognized—it is just considered part of the new large intron made by joining what were the two smaller introns on either side of that exon.

The exon definition model is supported by evidence showing that SR proteins bind to exon sequences, apparently bridging the acceptor and donor binding snRNPs across the exon [20]. In addition, in vitro experiments show that deletion of certain exon sequences abrogates splicing in of the exon [21]. These sites are called exon recognition sequences (ERS) or exonic splicing enhancers or exon splicing elements (ESE) 22, 23, 24, 26. Splicing enhancers are genomic sequences in exons or nearby in introns that facilitate splicing (Table 1). These sequences are generally purine rich.

Poison sequences (the opposite of enhancers) also exist 28, 29. Del Gatto et al. [28]found that the exonic sequence TAGG in the alternative K-SAM exon of fibroblast-growth factor receptor-2 inhibits splicing of that exon; this sequence inhibits splicing in other genes as well, possibly because it is the consensus 3′ splice sequence (as discussed below). Furdon and Kole [29]found that a 5′ region of the rabbit globin intron 2 inhibited splicing when downstream of the 3′ splice site.

Some genes undergo different (alternative) splicing patterns in different tissues or during development. The classic example is of tra and dsx in Drosophila, which are spliced differently in the two sexes, thereby controlling sexual development (reviewed in Ref. [30]). For tra, choice of an upstream splice site leads to inclusion of an exon with a stop codon in the mRNA causing a non-functional protein and male development (default splice). Use of a more 3′ splice site excludes this exon and the resulting protein controls female development. The gene Sex-lethal (Sxl), which is active only in females, codes for a snRNP that binds to the U-rich pyrimidine sequence at the upstream splice sequence and suppresses splicing. Sxl is also alternatively spliced in males and females through autoregulation. Tra and tra-2 regulate the doublesex gene (dsx) by binding to 13 base repeats in the fourth exon and causing the female alternative splicing pathway. A purine rich sequence can substitute for this 13 base repeat. These examples show both positive and negative regulation of splicing.

Consensus sequences for the 5′, 3′ (includes pyrimidine tract) and branch sequences have been determined and a number of groups have developed methods to score and/or search for sites within genes or DNA sequences 31, 32, 33, 34, 35, 36, 37, 38, 39. In one model, the 5′ consensus DNA sequence is: AGgt(a/g)agt and the 3′ consensus DNA sequence is: 10(t/c)n(t/c)agG. Different groups suggest different lengths for the acceptor associated pyrimidine tract 31, 33, 35. The most important feature appears to be the number of pyrimidines in the stretch; consecutive T's strengthen the acceptor while purines weaken it 40, 41. Chu et al. [42]describe a polymorphism in the pyrimidine tract of exon 9 of the cystic fibrosis gene which affected the level of exon 9-transcript. In order to rank potential splice sites and determine possible cryptic sites, Senapathy et al. [31]used tables of base frequencies at each position to develop ad hoc measures of the respective `strength' of 3′ and 5′ sites (a Senapathy score).

Stephens and Schneider [35]have discussed the features of splice sites in human genes and reviewed the models offered for this process. Based on the sequence of 1799 donor sites and 1744 acceptor sites, they have developed an information theory-based model of splicing in which information curves define the relative constraints on the specific base in that sequence 34, 37, 38, 39. This study argues that previous models for splice sites do not comprehensively define functional splice sites and does not include many functional sites with different sequences. The information in bits for a splice site (R_i) is defined as the dot product of a weight matrix derived from the nucleotide frequencies at each position of the splice site from the database and the vector of a particular splice junction sequence. According to information theory and the second law of thermodynamics, sites with R_i<0 cannot be recognized or bound by the splicing proteins. In practice, sites with R_i<2.36 bits are not utilized. The sequence logo graphically illustrates the average information content (in bits) flanking each site of the exon.

Using this model to study 1799 donor and 1744 acceptor sites, Stephens and Schneider [35]found donor sites had an average value of 8.4 bits and showed a significant sequence specificity in the last 3 bases in the exon and the first 6 bases of the intron and the acceptor splice sequence had an average value of 9.3 bits and included the last 25 bases of the intron and the first 2 bases of the exon. Since the information content of the acceptor excedes that of the donor; this supports the exon definition model of the spliceosome, in which the acceptor site is recognized first prior to finding an appropriate donor site downstream. Once 2 exons have been located, the intervening intron is removed. (This model does not explain how the first exon's donor site is recognized or how the last exon's acceptor site is spliced.) The donor and acceptor sequence logos contain the common sequence CAG/GT which may have been the sequence of a proto-splice junction that evolved early. The idea of a single splice junction recognition sequence is attractive because it would maximize the information required for recognition on the intronic component of both splice sites allowing largely unrestricted codon choices throughout exons.

Further analysis of 130 mutations in 42 genes showed significant losses of information (−4.0 bits for acceptors and −6.7 bits for donors) with more severe mutations showing more significant decreases [38]. By contrast, polymorphic variants do not significantly change the R_i value of a site. The consensus donor and acceptor sequences have the maximum possible R_i values of 13.1 and 21.4 bits, respectively while actual donor and acceptor sequences have average R_i values of 7.1 and 9.4 bits, respectively. Cryptic sites were also detected and generally have R_i values comparable to the natural splice sequences.

Whereas a large number of donor (5′) and acceptor (3′) splice sites have been sequenced and collated, the branch site sequence containing the branch whose `attack' forms the lariat has only rarely been experimentally determined. The branch consensus sites given in the literature are: ctray or tnctrac (where a is the branch A described above) 32, 43. Nelson and Green [43]list 31 branch sites; however, a number of these are from in vitro constructs, lower organisms or viruses and 8 of the remaining 14 are from one family of genes (globin). Table 2 lists 25 vertebrate branch sites found in the literature. At the 20 constitutive sites, the branch A is located 18 to 38 bases from the acceptor.

A special feature of several alternatively spliced exons is that the branch site is located far away from the acceptor (up to 200 bases) (Table 2). These branch sites have a very long pyrimidine tract located just downstream 43, 44, 45, 46, 47, 48, 49. While studies of acceptor sites have generally assumed that the pyrimidine tract is associated with the 3′ acceptor, these cases suggest instead that the pyrimidine tract is in reality associated with the branch point and that a very long pyrimidine tract can allow usage of a distant branch site. Reed [41]also performed in vitro experiments showing that a branch point 116 bases away but with a long pyrimidine stretch was used over a closer branch point that had only a purine stretch.

The importance of the different sequences (5′, 3′, branch and pyrimidine tract) is debated. Clearly, weaknesses in one can be made up for by strengths in the other 40, 50. One hypothesis is that the 3′ site is simply the first ag downstream of the branch point although if a stronger ag is close (<30 bases further down) then that will be chosen (sites are rated cag>tag>aag>gag) [51]. This is supported by the fact that ag's are not commonly found 10–20 bases upstream of acceptor sites [36]and when present they are associated with weaker sites. However, other work where the branch points are mutated shows that a new branch point (always an A) is chosen 22 to 37 bases upstream of the acceptor implying that the 3′ site is the more important site [52]. Senapathy et al. [31]state that the 5′ (donor) site must be recognized first, however, based on information content, as stated previously, Stephens and Schneider [35]suggest just the opposite.

As discussed above, internal exons have a maximum size of about 300–400 bp with a mean of 245 bases and a peak of about 125 bases 8, 18; however, they also have a minimum size of approximately 50 bases [53]. In vitro experiments with constructs have shown that when exons drop below this size, they are skipped 27, 29. Small exons can be spliced however, and exons as small as 3 bases are known 8, 47. These small exons apparently all require specific enhancer sequences in order not to be skipped. For example in the cardiac troponin gene the 30 base exon 5 has a purine rich enhancer [47]. In vitro, small exons that are skipped can be spliced in if purines are removed from the pyrimidine tract [53].

The above size constraints concern only internal exons. The average size of the 5′ (first) exon is 200 bases with the highest percentage less than 100 bases and ranging up to 1 kb. Deletions in the first exon do not appear to affect splicing of exon 2 [29]. The 3′ (last) exon averages 649 bases with a peak at 300 bases and a large percentage over 900 bases [18]. These features imply that the splicing signals for the first and last exon must be different than for internal exons. For example, Liu and Mertz [54]found that an acceptor signal that worked well at an internal exon did not work for the last exon unless the pyrimidine tract was strengthened or sequences were deleted in the last exon. First exons apparently require the presence of the 5′ 7-methyl-guanosine cap (and its binding proteins) which is present on all polymerase II transcripts 8, 55. Last (3′) exons end at the poly A tail and studies indicate that splicing and polyadenylation factors interact (i.e., mutation of the last exon inhibits polyadenylation and mutation of the adenylation site inhibits last exon splicing) 8, 56.

A minimum intron size has also been determined 33, 49, 52, 57. Smith and Nadal-Ginard [49]found a minimum 5′ splice site to branch distance of 51–59 bases in the α-tropomyosin gene while Wieringa et al. [57]found a minimum intron size of 80 bases (did not function at 69 bases or less) in the rabbit β-globin gene and their survey of the literature found no introns of less than 50 bases. An unusual mutation in the donor site of exon 8 of the COL1A1 gene [58]causes not only skipping of exon 8 but inclusion of 96 bases on the 5′ end of intron 7 into the mRNA. Looking at the genomic DNA, the new donor site (Ggtaaga, Senapathy score—86.5) is better than the old site (Tgtgagt, Senapathy score—75.8); the reason that the better site must not normally be used is because this would reduce the size of intron 7 to only 63 bases (too small for an intron); however, in the mutant where exon 8 is not recognized then `intron 7' extends until the beginning of exon 9 (280 bases).

There is also information that suggests that some exons are recognized as a cluster or cooperative unit 59, 60, 61. Sterner and Berget [59]found that a mini-exon requires the preceding exon to be spliced in; mutation of the 5′ splice site of the upstream exon leads to intron inclusion. Ge et al. [60]found cooperation of two 3′ sites in order to splice a 48 bp exon. Both exon 8 (44 bp) and exon 9 are skipped with a mutation in the exon 9 donor site in the p67-PJOX gene [61].

Splicing can also be influenced by pre-mRNA secondary structure 46, 48, 62. In the human growth hormone gene, stabilization of a stem loop causes shifting to the alternative splice site [46]while a single nucleotide polymorphism in exon 2 of the episialin gene [62]causes use of different splice acceptor sites; this polymorphism is hypothesized to cause the loss of a hairpin loop. In the tropomyosin gene, a large hairpin encompassing the alternative exon 6B causes it to be spliced out. Mutations that destabilize the loop cause the exon to be spliced in Ref. [48].

Mutations leading to splicing errors are a significant proportion of mutations leading to human disease (i.e., 101 of a total of 659 point mutations causing human diseases are in splice junctions [63]). Krawczak et al. [63]and Nakai and Sakamoto [16]review 101 human and 209 mammalian splicing mutations, respectively. Krawczak et al. [63]report 62 donor (5′) splice site mutations, 26 acceptor (3′) splice site mutations and 13 creation of novel sites. For the donor (5′) splice site mutations, 60% involve the invariant gt with increased mutations at IVS+1 and IVS+2. For the acceptor (3′) splice site, 87% involve the invariant ag with an excess of mutations at IVS-2. Of interest, 12 of 13 of the new created novel sites are upstream of the normal site and only 4 of 13 have lower Senapathy scores than the original site. For the 5′ splice site mutations, 16 lead to skipping, 7 to cryptic utilization and 5 to both. For the 3′ splice mutants, 4 result in exon skipping and 6 in cryptic site utilization. For both sites, purines are most likely to be introduced in the mutation. Nakai and Sakamoto [16]found that 92% of mutations occur at consensus sites, 97 at 5′ sites and 55 at 3′ sites with 15% causing novel sites. For the 5′ site mutations, 48 lead to skipping and 33 to cryptic and for the 3′ site 29 lead to skipping and 21 to cryptic. They also found that novel sites are all upstream of the normal site. As indicated above, intron retention is rare. Among the intron retention mutants, 4 are very short introns and 3 are terminal introns; however, 3 are large internal introns.

Mutations at splice sites often lead to the observation of multiple cDNA species after RT/PCR analysis of intracellular mRNA. Sometimes utilization of a cryptic site is incomplete with some normal message observed or both skipped and normal or skipped and cryptic site mRNAs seen. These indicate competition between splice sites. In addition, sometimes mRNAs with multiple skipped exons are observed. Nonsense mutations in exons have also been found to lead to both low levels of mRNA and/or mRNAs (sometimes multiple) usually excluding the exon containing the nonsense mutation (64, 65, 66, 67, 68, 69, 70, 71, 72; reviewed in Ref. [73]). It has been suggested that the premature termination of translation caused by the nonsense mutation causes mRNA instability and, thus, low levels of mRNA. This is supported by the observation that nonsense mutants do not affect mRNA levels if they occur at the very 5′ end of the gene (re-initiation of transcription possible) or at the very 3′ end (transcription goes long enough that the mRNA is not unstable). Often, if the exon containing the nonsense mutation is skipped and an inframe deletion results, premature termination is avoided and that mRNA will be stable. In general, it is not suggested that the nonsense mutation causes the exon skipping, but merely that it stabilizes what is normally a low level product that is seen in the mutant only because the major full length (nonsense containing) product is degraded. The exon skipped product seen in nonsense mutants has been demonstrated at a low level in normal cells for some genes 68, 69. However, some groups suggest that the nonsense mutation does directly affect splicing because changes in mRNA are observed in the nucleus and translation effects must be cytoplasmic 64, 70. It is difficult to model how a nonsense mutation could be recognized by the spliceosome as this requires recognition of mRNA frame in the nucleus and many `nonsense codons' obviously exist out of frame in normal mRNAs. Some of these nonsense/exon skipping mutations could occur because the nonsense mutation just happens to be within an enhancer which is disrupted [71]. Models have been proposed whereby a link exists between splicing and translation; the ribosome pulls the hnRNA through the nuclear pore driving splicing [65]. If the ribosome releases the RNA, splicing ceases. Caenorhabditis elegans with mutations in their smg genes do not have nonsense mutation reductions in mRNA in other genes [72]implying that specific genes exist to degrade prematurely terminated mRNAs.

The X-linked hypoxanthine-guanine phosphoribosyltransferase (HPRT) gene is widely used in mutation studies because of the ease of selection of mutant cells. HPRTase phosphoribosylates hypoxanthine and guanine in the purine salvage pathway. Although constitutively produced, the HPRT enzyme product is non-essential as purines can also be created de novo. HPRTase deficient mutants can be selected in the presence of purine analogues such as 6-thioguanine (TG) which kill normal, wild type cells; only cells containing a mutation in their HPRT gene are able to survive and proliferate. Although HPRT is dispensable at the cellular level, germinal HPRT deficiency does lead to human disease, either Lesch–Nyhan syndrome (complete deficiency) or X-linked gout (partial deficiency), apparently caused by the high levels of purine metabolites such as uric acid in the blood. The symptoms of Lesch–Nyhan disease include an auto-mutilation syndrome in hemizygous males. A reversion assay for selection against HPRT mutations also exists, which is the ability to grow in hypoxanthine-aminopterin-thymine (HAT) medium. Studies using the HPRT gene include: in vitro analyses of radiation and chemical effects in human lymphoblastoid cell lines or peripheral T-cells, in vivo studies of mutations induced in humans after radiation or chemical exposures, in vivo studies of individuals with DNA repair defects and analysis of mutations in Lesch–Nyhan individuals. Numerous in vivo and in vitro studies have also been performed using the hprt gene in animals and animal cells including mice, rats and CHO cells.

The human HPRT gene is located at Xq26 and its 9 exons cover 43 Kb (Fig. 1). The mRNA is 900 bp with a coding region of 657 bp. Of note is that HPRT has two large introns (introns 1 and 3) and one relatively small one (intron 7). Also, exon 5 is very small (only 18 bases); this is well below the `minimum' exon size of 50 bases and suggests that exon 5 should possess splice sites with high scores and probably also an enhancer. Exon 7 is also small (47 bases). A 57 kb region containing the entire gene has been sequenced and linked DNA markers have been mapped, allowing for detailed molecular studies 74, 75. A world-wide database and international repository has been developed describing human HPRT mutation molecular spectra [76]. Currently this database (release 6) includes 2224 mutations, 638 of which (28.7%) are listed as `splice' mutations. This is the largest collection of splicing mutations from a single gene. This paper reviews mRNA splicing at the HPRT gene including the sequences and structures of the splice sites, predicted and actual cryptic sites, the locations and frequencies of splice mutations and the production of multiple mRNA products. We will especially focus on mutations with unusual or unexpected consequences which reveal information on splicing mechanisms.