Introduction

Stickler's syndrome is a genetically heterogeneous disorder affecting the heterotypic (composite) fibrillar collagen molecules found in both vitreous and cartilage.1, 2 These macromolecules consist of a core of type XI collagen that regulates fibrillogenesis by limiting the addition of the more abundant type II collagen molecules to the collagen fibril.3 Situated on the outside of the type II/XI collagen fibril is type IX collagen. This fibril-associated collagen with interrupted triple helices (FACIT) not only connects the collagen fibrils to other components of the extracellular matrix but can also influence spacing between collagen fibrils.4, 5, 6, 7

Mutations in the genes encoding the constituent proteins affect both vitreous and cartilage tissues. Stickler's syndrome presents with congenital vitreous abnormalities and variable expressivity of other features, such as myopia, premature osteoarthritis, hearing loss, and craniofacial dysmorphology including midline clefting.2 The most common form of Stickler's syndrome is type 1 (MIM 108300) and is due to mutations in the COL2A1 gene.8 These mutations mostly result in haploinsufficiency of type II collagen due to premature termination of translation, which leads to degradation of the mutant mRNA via the nonsense-mediated decay pathway.9 Other types of mutations in this gene, that result in dominant negative effects, can result in much more severe disorders, such as spondyloepiphyseal dysplasia congenita (SEDC) and Kniest dysplasia.10, 11

Mutations in the genes for type XI collagen result in type 2 (MIM 604841) and type 3 (MIM 184840) Stickler's syndrome.12, 13 The COL11A1 gene is expressed in both the vitreous and cartilage, mutations here result in the type 2 variety of the disorder, whereas COL11A2 is not expressed in the eye and mutations in this gene result in the ‘non-ocular’ or type 3 Stickler's syndrome, which only has skeletal and hearing defects. All of these are dominant disorders, however, there has been one report of a recessive form of Stickler's syndrome where patients were homozygous for a premature stop codon in COL9A1.14 Interestingly, dominant mutations in type IX collagen genes result in multiple epiphyseal dysplasia (MIM 600204, 600969), but patients with this disorder have not been reported to have any eye pathology.15, 16, 17

The vitreous phenotype in Stickler's syndrome is the most consistent clinical feature. Type 1 Stickler's syndrome has a typical vestigial vitreous gel in the retrolental space bounded by a folded membrane.2, 18 In contrast, type 2 Stickler's syndrome is characterised by a vitreous gel filling the posterior segment, but with disorganised gel architecture, including irregularly thickened fibre bundles that have a beaded appearance under slit-lamp examination, which possibly reflects the role of type XI collagen in regulating collagen fibrillogenesis.19 Systemic features are variable and although some have an increased frequency in certain types of Stickler's syndrome (ie cleft palate in type 1 and hearing loss in type 2), these differences are not sufficient to accurately predict the mutant locus based on phenotype alone. The variable systemic phenotype extends both between and within families, and as most cases are due to haploinsufficiency of type II collagen, there must be factors other than the mutation that modifies the resulting clinical phenotype.

Genetic association studies have started to identify common DNA variants of genes that result in a predisposition to develop common disorders, such as osteoarthritis (OA) and myopia.20, 21, 22, 23 It is likely that coinheritance of these genetic predisposition factors is one source of phenotypic variation. For instance, an individual with a mutation causing Stickler's syndrome, who also inherits an OA predisposing gene allele, is more likely to develop this as one of the systemic features than a patient that only has a Stickler mutation and no other predisposing OA alleles. Interestingly, COL2A1, COL11A2, and the type IX collagen genes have been found to be associated with degenerative cartilage disorders and/or myopia.23, 24, 25 Due to the relatively small numbers of affected individuals with Stickler's syndrome who have been genetically characterised, at present it is not possible to correlate coinheritance of these predisposing risk alleles with the severity of any particular phenotypic trait in this disorder. Instead, our own research has started to examine how apparently similar mutations can have different effects on how the pre-mRNA is processed and how variability in mRNA splicing can potentially affect the resulting phenotype.

Splicing and splicing factors as phenotypic modifiers

To increase the information carried by the genome, many genes are alternatively spliced, that is, certain exons are either included in or excluded from the mature mRNA transcripts. So similar but subtly different proteins, which can vary in both biochemical characteristics and function, can be synthesised from the same gene. Both COL2A1 and COL11A1 are alternatively spliced.26, 27 Type II collagen exon 2 is not expressed in mature cartilage but is expressed in the eye. Mutations within this exon are, therefore, naturally removed from the type II collagen transcript in mature cartilage tissue. Patients with these exon 2 mutations typically have a predominantly ocular phenotype with little or no systemic features usually seen in Stickler's syndrome (MIM 609508).28, 29, 30 In the past, this has resulted in confusion with the rarer Wagner's syndrome (MIM 143200) that lacks systemic features and is due to mutations in the CSPG2 gene that codes for the proteoglycan versican. Although versican is expressed in many tissues, it is also alternatively spliced with four variants that either include or exclude exons 7 and 8 that encode glycosaminoglycan attachment domains. To date, all of the mutations described for Wagner's syndrome affect the splice sites of the alternatively spliced exon 8.31, 32, 33

Correct pre-mRNA processing requires the accurate recognition of the donor, branch, and acceptor splice sites by the spliceosome, a large RNA/protein macromolecular assembly that removes intronic sequences from the mRNA. To aid the specificity of this process, other trans-acting splicing factors bind to cis-regulatory sequences in the pre-mRNA. The trans-acting splicing factors can either promote or inhibit formation of the spliceosome at potential splice sites and so the sequences, to which they bind, are known as splicing enhancers or silencers.34 Naturally occurring alternative splicing is regulated by controlling the expression of trans-acting splicing factors that can either promote or inhibit exon inclusion in a developmental and cell-type-specific manner. Alternatively spliced exons often have weak (differ considerably from the consensus sequence) splice sites and are, therefore, more dependant upon trans-acting splicing factors for inclusion into the mature mRNA.

By exploring this potential tissue-specific alternative splicing, we have identified another mechanism resulting in the predominantly ocular variant of Stickler's syndrome. In this case, a mutation altered a donor splice site from Tggtaagc to Tggcaagc.35 Donor splice sites with gc as the first two nucleotides of the intron occur rarely, but naturally in the human genome and are often alternatively spliced. We demonstrated with the use of minigenes that this gc mutant donor splice site could be spliced normally when transfected into cultured cell lines (Figure 1). However, not all cells were capable of splicing the mutant normally, which presumably was due to differences in expression of trans-acting splicing factors necessary for correct splicing of the mutant allele. It seems reasonable to deduce that the predominantly ocular phenotype in this family is due to the ability to splice the mutant allele normally. However, it is not clear whether this is due to differences in splicing efficiency between cartilage and ocular tissue or whether the development of the vitreous is more sensitive to reductions in the synthesis of type II collagen.

Figure 1
figure 1

A minigene with an IVS51 GC mutant donor splice site can be processed normally. Three different mutations (G+1>T, G+1>A, and T+2>C) of the intron 51 donor splice site of COL2A1 were cloned as minigenes and transfected into various cultured cell lines as indicated. Minigene-specific RT-PCR using a reverse primer that could only amplify correctly processed transcripts demonstrated that unlike G+1>T and G+1>A, the T+2>C mutant was capable of being spliced normally. A minigene with the normal splice site (N) and the cloning vector (V) was used as controls. This mutant resulted in the predominantly ocular phenotype.

Other disease-causing mutations can also have a number of effects, not only by altering the amino-acid sequence but also by disrupting or creating splicing regulatory sequences (splicing enhancers or silencers) within the pre-mRNA. The effect of these mutations may vary from tissue to tissue and so apparently similar mutations, such as premature termination codons, can have dramatically different effects depending on whether they also affect pre-mRNA splicing and the degree, and tissue specificity of any missplicing that occurs and whether the transcript remains inframe or not (Figure 2). This nonsense-associated altered splicing (NAS) has been seen in a number of disorders.36, 37 For collagen encoding genes, NAS resulting in exon skipping can have dramatic effects on phenotype, switching mutations that would normally result in haploinsufficiency, via nonsense mediated decay, into ones that have a dominant negative effect. This is because most exons in collagen genes encode complete Gly–Xaa–Yaa amino-acid repeats and so exon skipping leaves the message inframe. Many missense (typically substitutions of obligate glycines within the collagen helix) and exon-skipping mutations in collagen genes have dominant negative effects because the mutant protein co-assembles with protein encoded by the normal allele to form abnormal trimeric collagen molecules. These structurally abnormal collagens are recognised by an endoplasmic reticulum-mediated quality control system,38 retained intracellularly and degraded. For a homotrimer, such as type II collagen, this means only 1/8 of the molecule will be entirely normal with the remaining 7/8 containing 1, 2, or 3 mutant α-II collagen chains. These dominant-negative mutations typically result in severe phenotypes presenting clinically as spondyloepiphyseal dysplasia congenita (SEDC), spondyloepimetaphyseal dysplasia (SEMD), and Kniest dysplasia (MIM no. 183990, 184250, and 156550). Thus, the clinical outcome resulting from apparent nonsense and missense mutations can be modified by only a small difference in abnormal splicing (Figure 3). This mechanism can also explain why the phenotype can vary between individuals with the same mutation depending on differences in the amount of missplicing that occurs between individuals.

Figure 2
figure 2

Possible effects of mutations on pre-mRNA processing. A mutation may result in either normal splicing or frameshifts that will often lead to incorporation of a premature termination codon into the new reading frame, leading to haploinsufficiency. Alternatively, inframe missplicing can result in mutant proteins that can exert a dominant-negative effect.

Figure 3
figure 3

Missplicing can modify the phenotype of the type II collagenopathies. (a) Apparently similar mutations such as PTCs may be modified by nonsense-associated altered splicing (NAS). Even a small percentage of exon skipping resulting from PTCs will exert a dominant negative effect and modify the resulting phenotype, making it more severe. Similarly, missense mutations that interfere with pre-mRNA splicing can be modified either by exon skipping or by causing frameshifts. (b) Mutations of splice sites may result in the use of cryptic splice sites that result in NMD or exon skipping usually with dominant negative effects. The phenotype may vary depending on the proportion of each type of transcript.

We have studied two individuals with Kniest dysplasia and identical mutations, but subtly different vitreous phenotypes. One had disorganised vitreous gel architecture, whereas the other had the membranous vitreous phenotype commonly seen in patients with type II collagenopathies. The mutation has been seen previously in a number of cases of Kniest dysplasia.39, 40 Despite altering the amino acid alanine 102 to valine, the pathogenic mechanism is missplicing, because it creates a de novo donor splice site in exon 14. Using cultured skin fibroblasts to examine illegitimate transcripts, we detected the same 21 bp deletion from amplified cDNA that had been detected by others. However, when we transfected minigenes of the mutant allele back into the patients’ own cells, exon skipping was detected, but only in one of the patient's cell lines (Figure 4). This variation is presumably due to a difference in splicing factors between the two cell lines. Whether this explains different vitreous phenotype is impossible to determine, but the two misspliced RNAs can potentially have different effects. Although the 21 bp deletion leaves the message inframe, it disrupts the repetitive Gly–Xaa–Yaa protein sequence and so even mutant homotrimers will be recognised as abnormal and retained in the endoplasmic reticulum. In contrast, the 54 bp exon skip leaves the repetitive collagen sequence intact and mutant homotrimers, although shorter than normal, can form a collagen triple helix and may be more efficiently secreted into the extracellular matrix.

Figure 4
figure 4

Variable splice products resulting from a mutant minigene. RT-PCR using illegitimate transcripts (a) from two cases of Kniest dysplaisa (Kn1 and Kn2) demonstrated utilisation of a de novo donor splice site and deletion of 21 bp from the amplified cDNA. The A102V mutation cloned as minigenes (b) was transfected into the same patients’ cell lines. Minigene-specific RT-PCR (c) amplified an additional product in Kn1 cells. Purification and sequencing (d) showed that this was due to skipping of exon 14.

Mutations of splice sites occur frequently in Stickler's syndrome;35 however, unlike those that cause exon skipping and result in Kniest dysplasia, the Stickler mutations appear to result in the utilisation of cryptic splice sites causing frameshifts and nonsense-mediated decay (Table 1). Our analysis of these mutations usually examines illegitimate transcripts from skin, but like the two cases of Kniest dysplasia discussed above, other splice products, such as exon skipping, may also result from these splice site mutations but go undetected in the illegitimate transcripts used for analysis. Hence, some of the phenotypic variability in these patients may be due to differences in the proportion of transcripts that are either targeted by the nonsense-mediated decay pathway (cryptic splice sites and frameshifts) or produce dominant negative effects (exon skips). These variations in mRNA splicing may differ both between cell types and also between individuals, and so modify the resulting phenotype (Figure 3b).

Table 1 A selection of splicing mutations resulting in either Kniest dysplasia or Stickler's syndrome

Missense mutations

Although most mutations causing type 1 Stickler's syndrome result in haploinsufficiency, there have been a few examples of missense mutations causing mild Stickler or Stickler-like phenotypes (Table 2). As discussed above, these mutations may have effects other than altering the amino-acid sequence, but it has also been shown that certain substitutions of glycine have less effect on the collagen molecule than others.47 Substitution of two consecutive glycines at positions 267 and 270 both result in mild phenotypes, suggesting that this region of the molecule is less sensitive to disruptions of the collagen helix. Others such as Arg904Cys produce additional phenotypic features, such as brachydactyly44 (Figure 5). This brachydactyly is more reminiscent of Kniest dysplasia and so may indicate that this mutation has additional effects on RNA processing than just the amino-acid change, such as exon skipping that commonly results in Kniest dysplasia. We have analysed some of these mutations (Table 2) either as illegitimate transcripts or as minigene splicing reporters and have not detected any missplicing, suggesting that their effect is through changes to the biophysical characteristics of the collagen. However, the Ala102Val and a silent mutation,42 that also creates a de novo donor splice site, demonstrate how single-base substitutions, including those thought to be missense mutations, may also have other pathogenic mechanisms that can modify the resulting phenotype. Recently, a missense mutation in exon 2 that converted Cys57Tyr has been described.41 The amino-acid substitution may be pathogenic, as it would be expected to alter the structure of the cysteine-rich domain in the N-propeptide; however, the authors also demonstrated that it inhibits inclusion of exon 2 in the mature transcript. Although not expressed in mature cartilage, exon 2 is expressed in prechondrogenic non-ocular tissues early in development48 and this may explain why some individuals with mutations in exon 2 show some systemic features of the disorder, such as cleft palate.28 It also indicates the importance of the cysteine-rich N-propeptide domain in the normal development of the vitreous.

Table 2 Missense mutations in COL2A1 associated with mild phenotypes
Figure 5
figure 5

Variable brachydactyly seen in cases of arginine to cysteine mutations. A R904C mutation shows brachydactyly, as does an example of Kniest dysplasia. However, a R565C mutation does not.

Conclusion

In the early to mid-1990s, the identification of mutations in the collagenopathies especially those of types I, II, and III collagens lead to attempts to correlate mutation (amino-acid change) and clinical phenotype. More recent studies have demonstrated how apparently similar and even identical mutations can have different effects on the processing of pre-mRNA. Along with the identification of common genetic variation that can predispose individuals to disorders, such as osteoporosis, osteoarthritis, and myopia, the concept of a simple genotype/phenotype correlation, that is, amino-acid change and clinical outcome would now appear to be somewhat outdated.