Deposits and Proteoglycan Changes in Primary and Recurrent Granular Dystrophy of the Cornea

Objective  To investigate the origin and distribution of granular deposits in the corneas of 3 patients with granular dystrophy, 1 of whom had previously received a lamellar keratoplasty in which the granular dystrophy had recurred.

Method  Corneal tissue from 2 patients with primary granular dystrophy (patients 1 and 2) and from a patient with recurrent granular dystrophy (patient 3) was examined. Corneal graft tissue was fixed in (1) 3% glutaraldehyde in sodium cacodylate buffer, (2) 2.5.% glutaraldehyde in sodium acetate buffer containing cuprolinic blue, and (3) 4% paraformaldehyde in phosphate-buffered saline.

Results  In patient 1 (aged 48 years), electron-dense granular structures were observed in epithelium, Bowman layer, and throughout the stroma. Bowman layer was absent in several places. Patient 2 (aged 78 years) showed similar features except with more deposits in the stroma. In patient 3 (aged 48 years), granular structures were heavily deposited in the epithelium; there were also some deposits in the posterior (host) stroma, some of which were associated with partially degenerated keratocytes. Bowman layer appeared normal. In all 3 patients, the intracellular or extracellular granular structures were surrounded by fine fibrillar material and abnormal proteoglycans. Electron-lucent spaces within the corneal stroma contained large quantities of abnormal proteoglycan filaments that were attached in part to collagen fibrils.

Conclusions  Results from patient 3 support an epithelial origin for the deposits, presumably from keratoepithelin, aggregated with other proteins. The role of keratocytes is less clear, although the presence of deposits in the stroma of all 3 patients, some associated with keratocytes, suggests that these cells might produce granular material in addition to abnormal proteoglycans.

GRANULAR CORNEAL dystrophy (GD) is a dominant stromal dystrophy with at least 2 clinical phenotypes. An early-onset form, the superficial variant,¹ begins in childhood with confluent subepithelial and superficial stromal changes, frequent recurrent erosive attacks, and early visual loss.^1-7 At least some early-onset cases may represent a homozygous state.^8,9

A milder, late-onset form^10-12 is characterized by multiple, crumblike stromal opacities, slow progression, fewer erosive attacks, less visual disturbance, and less need for corneal grafting. In both forms, the peripheral stroma is clear. Granular corneal dystrophy is characterized by electron-dense, rod-shaped, trapezoid, and fenestrated bodies within the stroma,^6,13-15 although there is morphological overlap between GD and anterior limiting membrane dystrophy (ALMD).¹⁶ Clinical confusion exists between early-onset GD and ALMD.

This article reports the ultrastructural changes and proteoglycan (PG) distribution in GD in the corneal tissue of 3 patients, 1 of whom exhibited a recurrence. Energy-dispersive microanalysis was also performed.

Patients 1 and 3 were from a large pedigree of GD (Figure 1), while patient 2 was of late presentation with no family history other than poor sight in his mother. All had the late-onset GD phenotype.

Patient 1 was a 48-year-old woman with a history of GD treated by left penetrating keratoplasty (PK) 10 years previously. She had a right visual acuity of 20/80 in her ungrafted eye and underwent a right PK (8 mm) in October 1996. Corrected visual acuity 5 months postoperatively was 20/30. Patient 2 was a 78-year-old man diagnosed as having GD at the age of 77 years whose corneas showed typical features of GD (Figure 2). Visual loss, partly due to cataracts, began at the age of 67 years. Corrected visual acuity was 20/300 OS on presentation and he underwent left, combined PK (7.5 mm) and a cataract extraction in June 1996. Corrected visual acuity 5 months postoperatively was 20/40. Patient 3 was a 48-year-old woman with GD who, at the age of 28 years, developed a fine superficial GD. She underwent right lamellar keratoplasty (7.75 mm) in June 1994. Superficial recurrence led to a PK (8.25 mm) in August 1996. Corrected visual acuity was 20/40 1 month postoperatively.

A segment of each corneal button from patients 1 and 3 was processed immediately after removal and cut into 3 pieces: piece 1 was immersed in 3% glutaraldehyde for 2 hours and postfixed in 1% osmium tetroxide for an additional hour. Piece 2 was stained for PGs with 0.05% cuprolinic blue (BDH Ltd, Dorset, England) in a critical electrolyte concentration mode.^17,18 These 2 pieces were then cut into small (1 mm³) cubes, dehydrated through a graded ethanol series, and embedded in Spurr resin. Piece 3 was fixed in 4% paraformaldehyde for 2 hours at 0°C, dehydrated into 70% ethanol, and embedded in resin (LR White; London Resin Co, Basingstoke, England) at 50°C for 24 hours.¹⁹ Material processed in LR White resin was stained with Masson trichrome for light microscopy and examined by energy-dispersive x-ray microanalysis (EDAX).

For ultrastructural studies, 65-nm sections were cut and stained with uranyl actate and lead citrate and observed with a transmission electron microscope (JEOL 1010; JEOL Ltd, Akishima, Japan). For the EDAX, 5 unstained, 1-µm sections on copper grids were studied with a JEOL FX 2000 STEM (JEOL Ltd).

With certain important exceptions, there were similar findings in all specimens. In patients 1 and 2, intracellular and extracellular deposits were present at both light and electron microscopic levels in all corneal layers, excluding Descemet layer and the endothelium. Deposits were stained with Masson trichrome on light microscopy (Figure 3).

Epithelial architecture was normal in many places in tissue from both the 2 primary dystrophies and the recurrent dystrophy. However, there were also distinctive abnormalities in all specimens. In semithin sections (Figure 4, A) or under electron microscopy (Figure 4, B), numerous electron-dense, polygonal intracellular deposits were present of varying size (20-5000 nm), often showing sharp angulations and projecting processes. In the epithelium of both primary and recurrent cases, intracellular deposits were chiefly in the basal cells, particularly basally and around cell nuclei, although they were also sparsely present throughout the cytoplasm and in nonbasal cells (Figure 5, B-D, and Figure 6, A).

Electron-dense particles, granuloamorphous material, and filaments were partially integrated into the flat faces of some larger deposits, appearing to contribute to their composition. In certain places, these were arranged in rows, orientated at right angles to the deposit face (Figure 5, D). The cell depicted in this figure had partial detachment of the plasmalemma from the basal lamina, with apparent expansion of the electron-lucent zone of the lamina and attachment of cells only at the hemidesmosomes.

In the recurrent case (patient 3), numerous extracellular deposits were located deep to the basal aspects of the basal epithelial cells (Figure 5, B). These deposits lay on an irregular, slightly thinned basal lamina. Here and there, in relation to separated plasmalemmae of the basal cells, thin patches of an amorphous material resembling basal lamina could be seen (Figure 5, B), but there were no hemidesmosomes present in these regions. Such deposits appeared to have stripped the basal cells from their associated basal laminae. Extracellular deposits were also seen invaginating the basal aspect of basal cells in the primary dystrophy, overlying extensive deposits within the Bowman layer (Figure 5, E). Here, a well-formed basal lamina was present in direct relation to the basal cells, undulating in a complex fashion over the deposits.

Other epithelial cells (not containing deposits) showed degenerative features, including sparse organelles (mitochondria and rough endoplasmic reticulum) and the presence of electron-lucent vacuoles and multivesicular bodies. Tonofilaments were abundant in relation to the deposits and cell nuclei and, in places, appeared to fuse into clumps (Figure 5, C). An occasional nucleus exhibited a disturbed chromatin pattern with fingerlike extensions from 1 pole; however, this could be an artifact due to the obliquity of the section (Figure 6, A).

In general, extracellular deposits (Figure 5, B) were larger than intracellular deposits (Figure 5, C), but smaller, beadlike intercellular deposits were also found (Figure 5, E).

Scattered, rod-shaped, amorphous, and trapezoid deposits were present throughout the Bowman layer in both primary cases of GD but not in the recurrent case (patient 3). Deposits were also present within a thick, irregular, avascular connective tissue pannus, visible on both light (Figure 7, A) and electron microscopy. They ranged in size from 180 to 7000 nm (Figure 6, E, and Figure 7, D). Some places had gaps in the Bowman layer (Figure 7, A) that were occupied by large, electron-dense aggregates of similar appearance to dense bodies elsewhere in the stroma. These bodies extended superficially into the epithelium, which was elevated, and more deeply into the stroma (Figure 6, E, and Figure 7, B). The overlying basal epithelial cells were often highly vacuolated and some nuclei were hyperchromatic in semithin sections (Figure 7, A). Some deposits were coextensive, running in an irregular sub-Bowman plane (Figure 7, A); there were other discrete sub-Bowman deposits underlying intact Bowman layer (Figure 6, D) and occasional deposits breaching the basal lamina (Figure 6, C).

In addition to granular deposits, aggregations of very fine fibrils, similar in appearance to amyloid fibrils, were observed in the Bowman layer of patients 1 and 2 (Figure 7, B).

In the primary GD stroma, both normal and abnormal keratocytes were demonstrable (Figure 8, C and D, and Figure 9, A and B), and extracellular and intracellular deposits were present between lamellae (Figure 8, C and D). Extracellular deposits, staining positively with Masson trichrome and electron dense on transmission electron microscopy, were found scattered throughout the stroma in the patients with primary dystrophy and were more plentiful and larger in the older patient (patient 2). These were more evident between lamellae than within them and disturbed the lamellar architecture. Deposits were made of granular and rod-shaped aggregates of varying electron density.

Such deposits were present in the posterior (host) stroma in the patient who had received a prior lamellar keratoplasty (patient 3) but not in lamellar (donor) tissue of this patient, where no rod-shaped bodies were found. Some keratocytes within the donor stroma of patient 3 showed a very prominent rough endoplasmic reticulum suggesting activation (Figure 8, C).

A proportion of interlamellar deposits in all patients had a spindle-shaped appearance resembling that of a keratocyte. In the primary dystrophy, deposits appeared to be present within the keratocytes themselves. In the younger patient, such cells exhibited both normal and abnormal organelles. In places, the extracellular deposits were closely associated with extracellular, multivesicular bodies and mitochondria, in the absence of any evidence of a cell membrane, representing perhaps degenerate keratocytes (Figure 8, D, and Figure 9, A). In the older patient, nuclei were demonstrable but other cell organelles such as mitochondria were absent (Figure 7, C).

Near the granular deposits, there was disorganization and increased separation of collagen fibrils (Figure 8, F). In these regions, fine amyloidlike fibrils were also present (Figure 8, E). In places, collagen fibrils were replaced or displaced by very fine fibrillar material or electron-lucent spaces containing PG filaments (Figure 10, A-C).

In the recurrent case, scattered electron-dense granules were seen within the irregular basal lamina of the epithelium in tissue stained with cuprolinic blue (Figure 4, C). Dots of a similar size were also seen in the cytoplasm of neighboring basal epithelial cells without particular concentration around the granular deposits. These particles are assumed to be PGs (Figure 4, C). In the primary cases, they were sparsely present in the epithelial cytoplasm.

In the stroma of corneas from primary dystrophy, in regions where granular deposits were absent, PG size and distribution were similar to that found in normal corneas, with filaments in the region of 60 nm long (Figure 9, F). Patients 1 and 2 showed a buildup of PGs under the Bowman layer within electron-lucent spaces (Figure 9, D), and similar collections of filaments (100-300 nm long) were also found within electron-lucent spaces in the stroma, in both the presence and absence of deposits (Figure 10, C and E). Collagen around these spaces appeared to be abnormal, with no clear banding and with a wider range of fibril diameters (Figure 9, E). In the posterior (host) stroma of the patient with recurrent dystrophy, there were elongated patches of a granuloamorphous material in the region of deposits, which were shown to be mingled with numerous large PG filaments and clusters of electron-dense dots in the cuprolinic blue–stained material (Figure 8, B).

Numerous vacuoles, surrounded by abnormal PG filaments, were present in the anterior and posterior stroma of patients 1 and 2 in regions containing extracellular deposits (Figure 8, D), whereas in patient 3 they were found only in affected posterior (host) stroma. In primary dystrophy tissue, these vacuoles were also observed in regions where the stromal lamellae were of normal appearance and lacked deposits (Figure 9, A).

Affected keratocytes, incompletely membrane bound, in the host stroma of the recurrent case were surrounded by electron-lucent spaces that were rich in large PG aggregates (Figure 9, B and C).

EDAX was used to analyze mineral content within the granular deposits, collagen fibrils, and epithelial cells. The x-ray spectra showed that Ca, S, and Si peaks were markedly higher in granular deposits than in collagen fibrils (Figure 10) and epithelial cells.

In this study, we found electron-dense deposits in the epithelium of both primary and recurrent GD and in the Bowman layer and stroma of primary GD. Deposits in the Bowman layer and stroma were mostly rodlike,^1,20 while those in the epithelium were angular, with fractures or faults suggesting an ordered arrangement. In other material, both primary and recurrent, we observed a banding pattern with a 116-Å periodicity.²¹ Intraepithelial rod-shaped deposits have been observed by other authors.^4,22

Epithelial deposits are intracellular and extracellular and could arise within or outside the epithelium. Corneal epithelium has phagocytic potential.^23,24 Our observations could not distinguish between these possibilities on morphological grounds, but there is evidence that epithelium is an important source.

Garner²⁵ implied that the deposits were of epithelial origin but found no direct evidence. Johnson et al²² noted a close relationship between epithelial deposits and the presence of abundant tonofilaments in affected epithelial cells. Recently, Wollensak and Witschel²⁶ demonstrated cytokeratin 18 and vimentin immunohistochemically in stromal deposits. Cytokeratin 18 is an intermediate filament protein that is expressed by corneal epithelium and, to a lesser extent, endothelium in normal cornea but not by keratocytes. Vimentin, usually associated with cells of mesenchymal origin, has been shown to be present in corneal epithelium.

Furthermore, Rodrigues et al⁴ demonstrated epithelial deposits at the ultrastructural level in a superficial variant of the disorder. This finding is in keeping with the superficial and subepithelial disposition of early stromal deposits, particularly in the superficial form, and would explain the anteroposterior progression of the dystrophy over time. These observations are also supported by the demonstration of deposits confined to the epithelium and subepithelium in recurrent GD, in the absence of stromal changes,²² and by our observations reported herein.

The basis by which an epithelial disorder may give rise to a predominantly stromal disease has been suggested by reports concerning its genetic and molecular aspects. Granular dystrophy, along with lattice and Avellino dystrophies and ALMD, maps to the same region of chromosome 5q.31, which contains a locus for the gene Big-h3 (transforming growth factor β–induced gene in clone H3). Big-h3 encodes the protein keratoepithelin (KE), which appears to be the key molecule determining the phenotype of these dystrophies.

Keratoepithelin is an adhesion protein of 653 amino acids whose mass (about 68 kd) is similar to that of albumin (66.4 kd). Such a size would permit diffusion across the corneal stroma.²⁷ In the normal mature cornea, KE is strongly expressed by corneal epithelial cells and, to a lesser extent, by endothelial cells but not by keratocytes, which do, however, express the protein in the fetal state^28,29 and after injury. It is a secreted protein with a signal sequence and an RGD peptide that could facilitate cell-matrix interactions. Although the protein is of epithelial origin, it is covalently attached to type VI collagen in the corneal stroma³⁰ and will bind to types I, II, and IV collagens.³¹ The turnover of stromally associated KE is unknown. Since keratocytes do not seem to express KE in the normal state, it may be speculated that stromal KE in the adult cornea is, at least in part, of epithelial origin.

Munier et al³² reported missense mutations in the Big-h3 gene in family members with each of the 4 dystrophies mapping to 5q.31. Mutations at codons 124 (arginine to cysteine and arginine to histidine) were associated with lattice and Avellino dystrophy, and mutations at codon 555 (arginine to tryptophan and arginine to glutamine) with GD and ALMD. The implication is that the 4 clinical phenotypes and their ultrastructural counterparts are due to differences in behavior of different forms of mutated KE.

These studies raise questions as to the nature of the deposits in the different dystrophies. At the structural level, Bücklers¹¹ likened the deposits of GD to the epithelial protein keratin (cytokeratin). This observation was followed up by Garner,²⁵ who presented histochemical evidence for the presence of arginine and sulfur-containing amino acids in the deposits, which are found in keratin. Our EDAX results confirm the presence of sulfur and significant levels of calcium. It is possible that calcium ions facilitate the aggregation of proteins within the deposits. Klintworth et al³³ reported the presence of mutated KE protein in GD stroma. However, other non-KE proteins have been demonstrated in these deposits, and it must be presumed that it is the presence of the mutated protein that leads to their deposition. Stromal deposits have been shown to contain immunoglobulin (IgG and κ and λ light chains) by Møller et al³⁴ (but not by Rodrigues et al⁴), cytokeratin 18 and vimentin,²⁶ and microfibrillar protein.^35,36 Amyloid was noted as a nonspecific deposit in GD by Garner²⁵ and is a specific component of deposits in lattice and Avellino dystrophies.^37,38

The presence of amyloid within GD lesions and those of lattice and Avellino dystrophy again suggests that mutated KE has an affinity for an amyloid precursor, leading to its deposition within coaggregates. A similar theory was advanced by Bishop et al³⁹ in relation to lattice dystrophy. They showed that glycoproteins, demonstrated by lectins on normal corneal basal epithelial cells, are translocated to the stroma in the presence of the dystrophy. Bishop et al advanced the hypothesis that, in lattice dystrophy, a glycoprotein was shed from the plasma membranes of epithelial cells and sequestered within the corneal stroma, where it subsequently stimulated amyloid deposition. In the present article, we suggest that, in the dystrophies under discussion, mutated KEs, secreted by the epithelium or perhaps keratocytes, coaggregate with various precursor proteins to produce the characteristic features of the diseases.

Proteoglycans play a major role in the organization of the extracellular matrix.^40-42 Large and abnormal PG filaments were found among the stromal deposits in each specimen examined within the primary dystrophy tissue. Nearby stroma showed a disruption of the collagen organization that is presumed to contribute to light scattering from these parts of the tissue. Regions were also found within the affected primary dystrophy stroma with an undisrupted lamellar organization in which PGs appeared normal; these are presumed to correspond to the clinically clear zones within dystrophic tissue. Normal keratocytes were identified within these latter zones but not within the affected zones. Minor PG abnormalities were identified by cuprolinic blue staining of donor graft stroma in patient 3 and there was some evidence for the presence of activated keratocytes.

We hypothesize that the production of abnormal PG in primary GD is a secondary response to the disease process initiated by mutated KE. Using cuprolinic blue, similar structures have been reported in the stromas of macular corneal dystrophy,⁴³ Hurler syndrome,⁴⁴ Morquio syndrome,⁴⁵ keratoconus,⁴⁶ and keratoconus epikeratoplasty.⁴⁷ It is possible in the present study that early PG changes in donor stroma in the recurrent case are the earliest sign of recurrence detectable in the stroma; however, it cannot be excluded that this may simply be a reactive change induced by the prior surgery of keratoplasty and is unrelated to recurrence.

The present study indicates that stromal deposits are both intracellular and extracellular. Keratocytes showed intact plasma membranes and contained the usual organelles in addition to material identical to that located extracellularly. Degenerate "ghost" cells, assumed to be of keratocyte origin, were also present, exhibiting imperfect remnants of intracellular organelles and lacking a plasma membrane. The origin of these intracellular deposits can only be surmised. Keratoepithelin is apparently not expressed in normal postnatal cornea or by keratocytes in Fuchs dystrophy cornea.⁴⁸ It is thus possible that the intracellular keratocyte deposits are due to phagocytosis of extracellular material. It cannot yet be excluded that they are of keratocyte origin, since messenger RNA expression of the Big-h3 gene has been observed in the healing stroma of the rabbit.³⁰

Accepted for publication November 17, 1998.

This work was supported by the Wellcome Trust, London, England.

Reprints: Keith M. Meek, PhD, The Open University, Oxford Research Unit, Foxcombe Hall, Boars Hill, Oxford OX1 5HR, England (e-mail: k.m.meek@open.ac.uk).