Introduction

Granular corneal dystrophy Groenouw type I (CDGG1) [ref. 1; MIM No. 121900] is an autosomal dominant disease of the cornea [2, 3]. There is complete penetrance, i.e. opacities in all carriers of the gene at 5 years of age and an increasing number of opacities through the following years [4, 5]. At around 50 years of age visual acuity is approximately 0.5, and at 70 years most patients have serious visual problems.

The opacities consist of greyish-white granules with sharp borders in the central area of the cornea. The distribution of granules varies between families [4], which may represent different mutant alleles or different loci. A Finnish family, also with autosomal dominant inheritance, but with mildly affected cases and later age at onset (between 15 and 20 years) has been described [4, 6]. A recessive form has also been claimed [7], but these patients may have macular dystrophy of the cornea (Groenouw type II) [5]; Reis-Biicklers’ corneal dystrophy has been proposed as a form of CDGG1 [8].

The granules have been shown to consist of a non-collagenous protein containing tyrosine, tryptophan, arginine and sulphur-containing amino acids, which may suggest a relationship to either amyloid or keratin [9]. Phospholipids [10] and immunoglobulins [11] have been found in the deposits.

In an earlier study, CDGG1 was scored for linkage against the blood group systems ABO, MNS and RH, but with negative results [12]. In a French family, association with an uncommon green eye colour in 6 patients out of 9 in two generations was observed [13]. Re-analysis of this pedigree with LIPED [14] gave a lod score below 1.0, while a study of 13 markers on chromosomes 1, 2, 4, 6, 8, 14 and 20 gave a positive lod score (z = 0.57) with GPT[15].

A linkage analysis with 35 markers in a large Danish family considered here [16] gave a positive lod score (z = 1.04) to the C1R serum protein (chromosome 12). The analysis of this family is here extended by inclusion of PCR markers [17–19]. Highly positive lod scores were found to the microsatellite markers IL9, D5S210, D5S119, D5S396, D5S402, D5S434 and D5S436, suggesting that the disease locus is on chromosome 5q, region 5q22-q33. Similar results were found in two subsequently examined smaller Danish families with a variant of CDGG1.

Material and Methods

Family Material

Blood samples from 124 individuals, all older than 5 years of age and comprising 52 affected and 72 non-affected persons, were collected from one large pedigree (family 1) [20]. Two additional small families independent of the large family and of each other (families 2 and 4) [20] were also collected for study of linkage and possible heterogeneity. Serum, erythrocytes and DNA were isolated from all blood samples of the three families and frozen for later typing [21].

Linkage A nalysis

A total of 35 classical markers was tested and analysed for linkage [16]. Later 10 RFLP and 16 PCR systems were tested until a highly significant positive lod score was obtained using the computer program LIPED [14]. The scores L (θM = θF) of chromosome 5 markers, as shown in table 1, were calculated using the computer program LINKAGE [22], The gene order was assessed both by manual haplotype drawing and ordering, and through multipoint analysis by the computer program LINKMAP.

Table 1 Linkage relations of Groenouw I and chromosome 5 markers: lod (Z) scores at various recombination fractions of (θM = θF) in family 1
Full size table

DNA Amplification

The radioactive PCR was performed in a volume of 10 µl in microtitre plates (Hybaid). Reactants comprised 50 ng of template DNA, 1–3 µM of non-label and 32P-end-labelled primer, 0.1 mM of each dNTP, 1.5 mM MgCl2, and 0.15 U of Taq DNA polymerase in buffer (Promega). The amplification was run in a programmable heating block (Omnigene, Hybaid) for 27 cycles comprising 94 °C for 30 s, 55°C for 0.30 s and 72 °C for 30 s. The amplified fragments were separated by electrophoresis for 1–3 h at 100 W in a 6% denaturing Polyacrylamide sequencing gel (25 × 42 × 0.4 cm). Gels were fixed for 10 min (10% acetic acid), washed for 20 min, dried and exposed to X-ray film (XAR-5 Kodak) overnight.

Results and Discussion

By linkage analysis employing the computer program LINKAGE [22], lod scores of Z = 15.96 (Om = 0.027, θF = 0.00) to IL9, Z = 19.45 (θM = 0.00, θF = 0.065) to D5S210 and Z = 2.29 (θM 0.07, θF = 0.416) to D5S119 were first obtained in the large pedigree (table 1). All markers are located on chromosome 5q, in the area 5q22-q32 for IL9 and 5q31.3-5q33.3 for D5S119 and D5S210 [23].

Only one crossing over was found between the disease locus (CDGG1) and IL9 and three crossing overs between D5S210 and CDGG1. Another marker, D5S211, which maps to the area 5q33.3-5qter gave negative lod scores. For more precise localisation, the additional marker D5S207 [19] and later the markers D5S396, D5S434, D5S436 and D5S402 [18] were considered.

Reference linkage maps of 5q including PCR markers show the following order: IL9-GRL-D5S210/D5S207-CSF1R/D5S119-D5S211 [23] and -D5S396-D5S436/D5S402-D5S434- [18]. Through four-point analysis using ILINK, LINKMAP and haplotype analysis (fig. 1) with IL9 (A), D5S210 (B) and D5S119 (C) the most likely position of CDGG1 was found to be: IL9-CDGG1-D5S210-D5S119. For the purpose of positioning the disease locus, the Weissenbach marker D5S396 was analysed against the two flanking markers IL9 (A) and D5S210 (B), and D5S396 was found to be in the order: D5S396-IL9-D5S210 (fig. 2). We could then position the disease locus into the map of Weissenbach et al. [18] as follows: D5S396-CDGGl-D5S436-D5S434(fig. 3).

Fig. 1
figure 1

LINKMAP scores of CDGG1 against the markers IL9 (A), D5S210 (B) and D5S119 (C). The most likely position of the disease was found to be between A and B, i.e. the order IL9-CDGG1-D5S210-D5S119, with a maximum location lod score difference of Z = 12 against the most likely other possibility.

Full size image
Fig. 2
figure 2

Scores of D5S396 in the fixed marker position IL9-CDGG1-D5S210 (A-G-B). The most likely position of D5S396 was found to be proximal to A, i.e. the order: D5S396-IL9-CDGG1-D5S210.

Full size image
Fig. 3
figure 3

Location of CDGG1 in the fixed marker sequence of D5S396-D5S436-D5S434 (D-E-F) [18]. The most likely position of the disease gene seems to be between D5S396 (D) and D5S436 (E).

Full size image

In an attempt to connect the two established linked groups, the following order was suggested: D5S396-IL9-CDGG1-D5S436-D5S210/D5S207-D5S434-D5S119-D5S211. This order was derived from the following four-point analysis by the program ILINK: D5S396-IL9-CDGG1-D5S436; CDGG1-D5S436-D5S210-D5S434; D5S210-D5S434-D5S119-D5S211 and D5S436-D5S207-D5S434-D5S119. The order of markers in each group was supported over the next most likely order with odds > 1,000:1, regardless of whether likelihoods were calculated assuming equal female:male recombination fractions or allowing for differences in the female:male recombination frequences. We found no recombination between the marker D5S402 against the markers D5S207, D5S210 and D5S436. ILINK analysis of D5S402 gave the order CDGG1-D5S402-D5S434. No recombinations between D5S207 and D5S210 or between D5S402 and D5S436 were found in this or the CEPH material [18].

Heterogeneity

Two small pedigrees (families 2 and 4) in which the mode of inheritance is also autosomal dominant, but with a later age at onset and milder expression, both gave a positive lod score to IL9 [Z = 2.11 in family 2 and Z = 0.903 (θM, θF) = (0,0) in family 4. The severe and the milder forms might represent different alleles at the same locus. The disease in these three Danish families is associated with entirely different haplotypes, which suggest that they are all of different mutational origin. Reis-Bücklers’ disease and granular dystrophy have some common clinical signs and the same locus may be involved [8]. Linkage analysis in Reis-Bückler families might clarify this question.

Disease Gene

About 24 genes (ADRA1B, ADRB2, ARH9, CAMK4, CCA, CD 14, CD49B, CSF2, DHLAG, DTD, DTS, EGR1, FBN2, FGFA, GRL, IL3, IL4, IL5, IFR1, LGMD1, PDEA, SPARC, TCOF1, TCF7) have been mapped in the region of IL9 (5q22-q32). Many of these are endothelial cell growth factors and receptors. The marker GRL (glucocorticoid receptor gene) which maps to 5q31-32, between IL9 and D5S210, although being close to CDGG1, is not an obvious candidate gene, but is important for more precise localization of CDGG1.

The locus for fibrillin 2 is involved in a Marfan-like disease [24]. The eye is also involved in this syndrome, i.e. dislocation of the lens, but the pathology is very different from that of CDGG1, and the fibrillin 2 locus is not therefore a serious candidate. SPARC (osteonectin) is a calcium-binding acid secreted glycoprotein [25], rich in cysteine. SPARC may be considered a candidate gene for CDGG1 because it is rich in sulphur-containing amino acids, and because, through inhibition of calcification or assembly or modification of glycoprotein compounds, it is involved in the production and assembly of extracellular matrix material (in bone and other tissue). So far, no human genetic disorder has been associated with mutations in the SPARC gene. To identify the gene for CDGG1 we plan to screen the region between IL9 and D5S243/D5S210 for cDNAs from a cDNA library using YAC and cosmid contigs from this region.