A novel phenotype-genotype relationship with a TGFBI exon 14 mutation in a pedigree with a unique corneal dystrophy of Bowman’s layer

Navigate by section:

• Introduction
• Methods
• Results
• Discussion

Abstract

Purpose: Corneal dystrophy of Bowman’s layer (CDB) belongs to a group of dystrophies associated with mutations in the transforming growth factor-beta-induced (TGFBI) gene. CDB is further divided into a geographic variant (CDB1/Reis Bücklers, RBCD), and a honeycomb variant (CDB2/Thiel Behnke, TBCD). We undertook mutational analysis of TGFBI in a family with an unusual CDB variant and describe a novel phenotype-genotype association.

Methods: Individuals from a pedigree with CDB underwent extensive phenotyping, including laser scanning in vivo confocal microscopy, and histological examination of four corneal buttons obtained at penetrating keratoplasty. Transmission electron microscopy of an excised allograft cornea from one affected individual was also performed. Following informed consent, DNA samples were collected. Polymerase chain reaction (PCR) and sequencing of all coding exons of TGFBI was performed. Family members were recruited with subsequent phenotyping and genotyping, and paternity testing.

Results: Clinical examination and other phenotypic information confirmed a diagnosis of CDB, with various features either more suggestive of CDB1 or of CDB2. A mutation in exon 14, H626P, segregated with the disease in this pedigree. This mutation was confirmed with NlaIII restriction enzyme digest, and was not seen in 100 control chromosomes.

Conclusions: Within this pedigree, CDB segregates with an H626P mutation, which is previously described occurring in lattice corneal dystrophy. The majority of mutations in TGFBI previously described segregating with CDB1 and CDB2 are R124L and R555Q, respectively. Although a Bowman’s layer dystrophy, the phenotype in this pedigree does not closely conform to the classical diagnostic criteria for either CDB1 or CDB2, and therefore represents a novel phenotype-genotype correlation.

Introduction

Reis-Bücklers corneal dystrophy (RBCD/CDB1, OMIM 608470) was first described by Reis in 1917 as an annular dystrophy (dystrophia anularis) [1]. Bücklers re-examined the same pedigree in 1949 and showed that the dystrophy was present in four consecutive generations [2], and affected members consistently presented with painful corneal erosions in childhood, with geographic opacities seen at the level of Bowman’s layer. Thiel and Behnke described a similar corneal dystrophy in 1967 that presented with recurrent erosions, moderately reduced vision, autosomal dominant inheritance, and honeycomb shaped opacities at the level of Bowman’s layer (TBCD/CDB2, OMIM 602082) [3].

Since the report of Thiel and Behnke, descriptions in the literature have varied in nomenclature and classification, rendering differentiation between autosomal dominant CDB1 and CDB2 difficult [4-7]. In a comprehensive review preceding genetic characterization, Küchle et al. [5] proposed that these dystrophies are indeed two distinct entities that differ markedly in their appearance at the slit-lamp, associated degree of visual loss, histopathological findings, transmission electron microscopy (TEM) features, and prognosis following corneal transplantation. Using Kuchle’s classification - clinically, CDB1 is described as exhibiting confluent geographic opacities at the level of Bowman’s layer and histopathology demonstrates band shaped granular, subepithelial deposits which stain intensely red with Masson trichrome. TEM reveals the presence of “rod shaped bodies” [5]. In contrast CDB2 has honeycomb shaped opacities, a subepithelial fibrocellular layer interposed between the epithelium and stroma, stains less strongly with Masson trichrome and has “curly collagen” fibers present on TEM [5]. While visual loss with CDB1 is described as early and marked [1,2,5], with CDB2 it is later and moderate [3,5]. The recurrence of the dystrophy after corneal transplantation is noted to be early with CDB1 and much later with CDB2 [5].

Following linkage of CDB1 and other corneal dystrophies (Granular, Avellino, and Lattice) to 5q31 [8,9], Munier et al. [10] reported mutations in the TGFBI (transforming growth factor -beta-induced) gene (OMIM 601692) correlating the R555Q mutation with CDB1. However, later, clinical, molecular and ultrastructural findings have demonstrated that it is actually CDB2 that correlates with the mutation R555Q and CDB1 to the mutation R124L [11-14].

TGFBI, initially called BIGH3 [15], encodes for a 683 amino acid protein product, with several nomenclatures including “keratoepithelin” and the now preferred “TGFBI associated protein,” TGFBIp. This extracellular matrix protein is expressed in many tissues including the corneal epithelium [16], and contains four regions of internal homology known as Fasciclin-like (Fas) domains. The protein has high sequence homology to fasciclin1, an insect cell adhesion molecule, and is thought to play a role in cell adhesion [17]. The majority of mutations in TGFBI reported to date have been located either within the codon encoding for amino acid R124, or at the boundary of, or within the Fas domain 4 [4,12], although a few mutations outside these regions are also reported [18-22]. Immunohistological studies have shown TGFBIp to be present in the abnormal corneal deposits seen in certain corneal dystrophies [23], as well as in secondary amyloidosis in corneal disease unrelated to TGFBI mutations [24].

We report a four-generation family with an autosomal dominant dystrophy that was initially diagnosed as CDB1 on the basis of clinical appearance and clinical course. Comprehensive characterization of the disease included expert clinical assessment, in vivo confocal microscopy, histology on four primary corneal buttons and a corneal allograft with disease recurrence, immunohistochemical staining, TEM, and mutational analysis of TGFBI. The results of these investigations suggest a novel genotype-phenotype correlation. Comparison of these data with an extensive review of the literature supports our view that this reported dystrophy, although exhibiting some features of both CDB1 and CDB2, does not meet the diagnostic classification for either, thereby adding to the complexity of the pathogenesis of the TGFBI corneal dystrophies.

Methods

Results

Discussion

Phenotypically this pedigree had some features that overlap with both CDB1 and CDB2. Features typically associated with CDB1 include marked visual loss (BCVA range pre-treatment 6/18⁻²-6/60), a young age at presentation (10.3±1.5 years, mean± SD) and early extensive recurrence observed following PKP in one individual [5]. Slit-lamp biomicroscopy also revealed confluent opacities that appeared geographic in shape as opposed to the honeycomb shaped opacities of CDB2. However, histopathology revealed that the subepithelial deposits did not stain red with Masson trichrome - a feature more characteristic of CDB2, often reported as minimal or questionable staining [5,14,27]. In vivo confocal microscopy images were consistent with reported findings in CDB2 [11]. Of the 5 youngest members of the pedigree examined and tested, three carried the mutation but were apparently clinically unaffected on slit lamp examination. However, they were at an age (3 years) much younger than the mean onset of disease symptoms/signs (10.3±1.5) in this pedigree.

Unfortunately tissue samples were not available on all subjects treated for corneal opacification, decreased vision and recurrent erosions. Although subjects with severe scarring and vascularization (and those treated before 1990) were treated by PKP, in selected members of the pedigree with the primarily superficial corneal lesions seen early in this dystrophy, excimer laser phototherapeutic keratectomy (PTK) was used successfully [28]. Having been shown to be a safe and effective alternative to re-grafting it has recently been used by the authors to treat the recurrence of the dystrophy in subject II:1 OD, with a promising early post operative result [29].

On light microscopy, a variable thick band of hyaline avascular connective tissue and/or fibrocellular tissue was identified beneath the corneal epithelium which did not stain red with Masson’s trichrome stain. (Figure 7) In the allograft specimen, the characteristic rod-shaped bodies observed in CDB1 were clearly not identified on TEM, and high magnification TEM demonstrated the subepithelial deposit immediately underlying degenerating basal epithelial cells. This deposit appears to consist of cell debris with some irregular aggregates, and scattered fibrils. These histopathological observations are not classically characteristic of either CDB1 or CDB2, although perhaps the presence of fibrils is more in keeping with a CDB2 variant [5,14,19]. Immunofluorescence demonstrated TGFBIp within the subepithelial deposit, which is consistent with the findings of Streeten et al. [27], who showed strong specificity of the stain to the curly fibers. However TGFBIp has been demonstrated in secondary corneal amyloidosis (confirmed with Congo red) in non-TGFBI-associated disease [24]. No Congo red staining was observed in any of our 4 corneal specimens, therefore this subepithelial band is unlikely to represent an amyloid deposit, however, the presence of TGFBIp is apparently not sufficient to differentiate a dystrophic deposit from scar tissue [24].

The novel association of the mutation H626P with this atypical CDB phenotype provides further evidence of the complexities of the classification and nomenclature of the TGFBI corneal dystrophies. This genotype-phenotype correlation is unique from two perspectives: First -historically CDB has predominantly been associated with 2 different distinct mutations; R124L with CDB1, and R555Q with CDB2 [13]. Second, the H626P mutation has been reported with a significantly different phenotype - late onset, asymmetric, lattice, dystrophy, (variant LCD, classified by the authors as CDLI/IIIA) [12], characterized by a dense haze associated with lattice lines, demonstrating multiple fusiform amyloid deposits in the stroma of an allograft corneal button. Recently the H626P change was shown to segregate with a superficial dystrophy that also showed the presence of amyloid [30]. A further mutation at this same base pair c.1877A>G resulting in H626R has also been shown to co-segregate with a late onset, asymmetric, form of lattice dystrophy (with previous numbering c.1924) [12,19,31-34]. One family had disease onset at 27 years of age, no recurrent erosions, and PKP necessary in the fifth decade [33]. H626R is postulated as being responsible for up to 75% of LCD seen in a Vietnamese population [32], termed LCDIIIB, and distinctive in that it has a later age of onset, asymmetry is common, and the lattice lines are thicker and more rope-like. Recently this mutation is described in a family with granular corneal dystrophy (GCD) [35].

In a recent paper the phenotype associated with the R124L mutation was well characterized, but as a corneal biopsy in the proband failed to distinguish between CDB1 and CDB2, the authors expound the virtues of genetic testing to determine the diagnosis [36]. These genotype-phenotype correlations (CDB1–R124L and CDBII–R55Q) have been confirmed numerous times with more detailed clinical characterization [13,14,36-38], and a suggested nomenclature for these dystrophies is “Classic” CDB [4].

In contrast CDB occurring with mutations other than the 124 or 555 mutations can be termed “Atypical” CDB; however histology is not recorded in these cases. Rozzo et al. [39] described F540Δ in CDB1, but reclassified the phenotype as late-onset lattice IIIa (CDL3A). Two pedigrees are described with G623D [40,41], one of which was CDB1 with apparent non-penetrance [40], and the other had deposits in Bowman’s layer, anterior stromal lattice lines, with symptomatic onset in the fourth decade [41].

Paternity testing is highly suggestive of a de novo mutation in TGFBI in II.1, by confirming the paternity of his unaffected father, however maternity could not be confirmed as his reportedly unaffected mother had died. Further possibilities which have not been investigated further include parental germ line mosaicism, or non-penetrance from the reportedly unaffected mother, although none of her other relatives were examined. Certainly de novo mutations have been documented in TGFBI-associated disease, as well as germ-line mosaicism [36,42,43]. It would appear that TGFBI does have some mutational “hot-spots,” where identical random mutations occur at an increased frequency at the same site of a gene. Korvatska et al. [44,45] demonstrated different haplotypes segregating with R124L in Japanese families, and in different ethnicities. They suggested that mutations at codons 124 and 555 in particular represent multiple independent occurrences.

With the addition of this report, significant phenotypic variability has now been demonstrated with the histidine at position 626, segregating with CDB, GCD, and LCD IIIB phenotypes. Likewise mutations within close proximity of the 626 residue also appear to show phenotypic variability. Previous modeling of the structure of TGFBIp demonstrates that His626 makes a crucial stabilizing hydrogen bond within the Fas4 domain [17]. This hydrogen link would be abolished by replacement of a proline, and therefore is likely to severely destabilize the domain [17]. So why is there such broad phenotypic variability seen in exon 14 mutations, and of specific relevance here, to mutations involving the His 626? Clout et al. [17]. also modeled His626Arg, which as well as abolishing the hydrogen bond, is likely to make the protein very unlikely to fold because of steric hindrance by the arginine residue

Although Fas4 domain mutations are thought to result in misfolding of the protein, a major role of the Fas domain is to mediate cell adhesion. TGFBIp also polymerizes to form a fibrillar structure and strongly interacts with type I collagen, laminin, and fibronectin [46]. It is unclear how the H626P change results in the subepithelial deposit observed with poor adherence to underlying tissues in this family, however one possibility is the accumulation of mutant misfolded protein is somehow deleterious to the cell, and a further theory is that the mutant TGFBIp cannot participate adequately in its usual binding functions, including polymerization to self and/or to other tissues [17,46,47]. There does however appear to be a corneal specific factor which influences this process, as other epithelial tissues in the body from individuals with a mutation, do not appear to manifest with TGFBIp deposits [47].

Other modifying factors could include co-existent SNPs, although there is no prior evidence that the L217L SNP present in exon 6 of the proband has any functional role. Interestingly, within this pedigree, slit-lamp biomicroscopy of the proband’s cousin (III:8) clinically revealed PPCD. He did not carry the mutation H626P, which could suggest that a mutation in a second gene may be responsible for his phenotype, and perhaps even modify the expression of the H626P mutation in this family.

The association of this pedigree with the phenotype CDB and the mutation H626P is in keeping with the majority of the published TGFBI mutations occurring either at Arg124 or within the Fas4 domain, and the absence of this mutation in our 100 control chromosomes and in previous controls suggests it is pathogenic rather than polymorphic. This novel genotype-phenotype correlation challenges our current knowledge of the “phenotypic specificity” that CDB1 segregates with the R124L genotype, and CDB2 with R555Q. Likewise mutations at the 626 residue segregate with non-lattice phenotypes. To explain the marked difference in phenotype between this pedigree and those described with H626 mutations, it is possible that a modifying genetic or environmental mechanism may be involved, the nature of which is unknown. Certainly further investigation is warranted on the nature of the deposits and the protein’s properties. This H626P CDB phenotype does not fit neatly into either CDB1 or CDB2 categories, and demonstrates the striking phenotypic variability observed with TGFBI mutations. This study and recent molecular evidence therefore highlights the ongoing complexity of the pathogenesis of TGFBI associated corneal dystrophies, and emphasizes the need to review the current nomenclature used to define TGFBI-associated dystrophies of Bowman’s layer.

Acknowledgments

The authors would like to thank: participating family members, Dr Archie Mackillop and other referring ophthalmologists, Judy Loh, Dr Diane Kenwright, Dr Hilary Holloway (BIRU) for technical assistance, Dr Andrew Huang, University of Minnesota for TGFBI antibody, and Patricia Stapleton for assistance with paternity testing. Funding was obtained from Save Sight Society New Zealand, University of Auckland Research Committee, Vice Chancellors’ Development Fund and the Maurice and Phyllis Paykel Research Trust. Portions of this research was previously presented in part at the Association for Research in Vision and Ophthalmology ARVO, Fort Lauderdale FL, May, 2007 (IOVS 2007;ARVO E-Abstract 48;5859).