Aldose reductase activity and basement membrane thickening

Aldose reductase activity and basement membrane thickening

Aldose Reductase Activity and Basement Membrane Thickening Robert N. Frank Rats fed a high-galactose diet develop marked thickening of their retina...

689KB Sizes 0 Downloads 125 Views

Aldose Reductase

Activity and Basement Membrane Thickening Robert N. Frank

Rats fed a high-galactose

diet develop marked thickening of their retinal capillary basement membranes. The effect is prevented if the animals also receive the aldose reductase inhibitor sorbinil. The effect does not appear to be due to aldose reductase itself, since immunoreactive aldose reductase has not been found in the retinal microvasculature of the rat but rather to a related enzyme with similar substrate specificity. The detailed biochemical mechanism for basement membrane thickening is obscure, involving an alteration of the extracellular matrix, where aldose reductase and similar enzymes have not been described; osmotic damage to the microvascular cells, such as has been described following aldose reductaseinduced sugar alcohol accumulation in lens epithelial cells, is not apparent in diabetic or galactosemic animals. It is possible that concentrations of intracellular sugar alcohols that do not substantially change the osmolarity of the cell cytosol alter intracellular enzyme activities. This, in turn, could affect the biosynthesis of extracellular matrix macromolecules. as suggested, for example, by the hypothesis of Rohrbach et al, based on studies of a basement membrane-producing tumor implanted in diabetic mice, which proposes that the hyperglycemia of diabetes mellitus causes a reduced synthesis of the heparan sulfate BM-1 proteoglycan with a subsequent overproduction of type IV collagen. This and other hypotheses of basement membrane thickening can be tested in diabetic or galactosemic rats, some of which receive aldose reductase inhibitors, or in retinal microvascular pericytes and endothelial cells grown in culture. ~11986 by Grune & Stratton, Inc.

C

ATARACT was the first complication of diabetes whose pathogenesis was attributed to the accumulation of SOIbitol as a result of aldose reductase activity. The argument, buttressed by an impressive body of experimental data, was that the very high concentration of intracellular sorbitol 101 of galactitol in the analogous galactosemic model) produ:ed a hyperosmotic intracellular environment. leading to imbibition of water by the cell, followed by cell swelling and lez kage of various intracellular contents through the plasma membrane. When a sufficient number of lens epithelial cells ,trc so affected, a cataract results.‘.* Subsequent experimenral and clinical studies have suggested that other complica:ions of diabetes may also result from the activity of aldose reductase or of similar enzymes3-* Could a similar osmotic mechanism be responsible? The answer to that question would appear to be no. The eplthelial cells of the lens, at least in species that develop diabetic and galactosemic cataracts, have very high aldosereductase activities per unit weight of protein compared with most other types of cells. Conversely, their glycolytic and glucose oxidative activities are relatively low, so that these pathways saturate at relatively low intracellular levels of ghrcose. Of the glucose-oxidizing activity, however, the pentoxe shunt makes up a relatively large proportion, thereby supplying ample reduced triphosphopyridine nucleotide (KADPH) to serve as a cofactor for the aldose reductase.ld Finally, the epithelial cells of the lens lack an insulindependent plasma membrane pump to regulate transport of glucose into the ce1l.4 Thus, the intracellular environment cannot be protected against high extracellular glucose (or galactose) levels. Other cells that suffer damage in diabetes dc not share these properties. For example, the pericytes and sndothelial cells of the retinal microvessels have higher glucose-oxidizing activity than do the epithelial cells of the lens, compared with their aldose reductase-like activity.‘The retinal pericytes, which are specifically damaged early in di,ibetic retinopathy.” are insulin-dependent for at least some of their functions.” Although the presence of an insulin-requiring glucose transport system has not been

Metabolism. Vol 35, No 4. Suppl 1 (April), 1986: pp 35-40

demonstrated in pericytes, its presence may at least be suspected from these experiments. The osmotic hypothesis satisfactorily explains diabetic and galactosemic cataractogenesis and may also be an explanation for those complications of diabetes in which severe damage to specific types of cells (eg, the pericytes of the retinal capillaries in diabetic retinopathy”’ and, perhaps, the Schwann cells in diabetic peripheral neuropathy3v4) occurs. Recently, however, evidence was presented that aldose reductase may also be responsible for certain complications of diabetes that involve the extracellular matrix. WeI and Robison et alI3 have observed that rats fed a high-galactose diet over prolonged periods (28 or 44 weeks on a 50% galactose diet in the Robison study, 15 to 21 months on a 30% galactose diet in our work) developed marked thickening of retinal capillary basement membranes and other basement membrane abnormalities resembling those of diabetes. The abnormalities did not develop if the animals received sorbinil, 250 mg/kg diet (Fig I, Table 1). Similar abnormalities, also preventable with an aldosereductase inhibitor, have been reported in the retinal microvascular and glomerular basement membrane of rats made diabetic with streptozotocini4 and prevention of retinal capillary basement membrane thickening has been described

From the Kresge Eye Institute of Wayne State University School

of Medicine,Detroit. Presented at the P$zer Sorbinil Symposiunl---The Effects of Sorbinil on the Pathophysiology of Diabetic Complications. Dorado, Puerto Rico. May 30-June 2. 1985. Supported in part by Grants EY-01857 and Ef’O2566fkom the National Eye Institute. grants from the Juvenile Diabetes Foundation to RNF, and by a departmental unrestricted grant from Research lo Prevent Blindness. Address reprint requests to Robert iv. Frank, MD, Kresge Eve Institute. 3994 John R Street. Detroit, MI 48201. v I986 by Crune & Stratton, Inc. 0026-0495/86/3504~1008$03.00/0

35

36

ROBERT N. FRANK

Fig 1. (A) Electron micrograph of a retinal capillary from a female Wistar-Kyoto rat fed for 21 months on a diet containing 30% galactose, by weight, substituted for other dietary carbohydrate. Note marked thickening of the basement membrane with vacuolization (V) and deposition of fibrillar collagen (C). A pericyte nucleus is indicated by (P). (B) A retinal capillary from a female Wistar-Kyoto rat maintained also for 21 months on a diet that was identical to that fed the animal in Fig 1 A, except that 250 mg sorbinil was added to each kg of diet. Arrows indicate basement membrane outcroppings that were not included in the thickness measurements. A pericyte nucleus is designated by (PI.

in galactosemic rats treated with an aldose-reductase inhibitor that differs structurally from sorbinil.‘s More recently, effective clinical treatment with sorbinil of another abnormality of extracellular matrix in diabetic patients has been reported. Eaton et al8 described three diabetic patients with a syndrome of joint immobility and flexion contractures of the hands whose signs and symptoms were dramatically relieved following 3 months of sorbinil therapy at a dose of 250 mg/d. (These data are presented in this supplement.) They suggested these results might be

explainable if the stiff joint syndrome were caused by excessive hydration of collagen, resulting from a high-sugar alcohol content. Yet, while collagen molecules are normally glycosylated at the hydroxylysine residues by glucose and galactose,16 and they may also undergo nonenzymatic glycosylation in hyperglycemic states,” sugar alcohols have never been reported to participate in these glycosylation reactions. Indeed, they cannot do so, since the glycosylation reactions require participation of an aldehyde group, which is lost when an aldose is reduced to a sugar alcohol by the action of

37

BASEMENT MEMBRANE THICKENING

Table 1. Basement

Membrane

Thickness

Counts in Electron Micrographs

(BMTI and Nuclear

of Rat Retinal Capillaries

ABMT (nm) + SD

MBMT (nm) * S

PN

EN

VvKY control (7)

1,604

i- 312

1,016

+ 172

9

25

VlKY + 30% gal (7)

2.033

+ 376

1,223

i

181

9

30

sorbinil (9)

1,594

k 352

969 A 202

10

25

SIHR control (7)

1,594

+ 361

9822

7

26

2,307

k 623

10

23

1,543

k 435

\RKY + 30% gal +

SHR

t

30% gal (6)

195

1,362+325

SIHR + 30% gal + sorbinil (8)

973 k 252

Total

11

36

56

165

neoplastic tissue, which may not respond to induced diabetes or to galactosemia in a way analogous to the response of normal retinal blood vessels, renal glomeruli. or other tissues in which basement membrane thickening is observed in diabetes. Morphological evidence of basement membrane thickening in EHS tumors implanted into diabetic animals has not been presented. This is important to establish, since not all basement membranes in diabetic animals or humans undergo thickening,22 and biochemical changes in the extracellular matrix related to diabetes or galactosemia may be meaningless in terms of the pathogenesis of disease processes unless they can be correlated with the presence of significantly thickened basement membranes in these diseases.

4bbrewatlons: ABMT. average (mean) BMT: MBMT, minimum BMT; PN. pericyte nuclei; EN, endothelial nuclei. lumbers of animals in each group are given in parentheses after the ‘description of the group in the first column. Further subdivision of each ~rcup by age is gwen

in the

caption

of Fig 2.

Reprinted

CRITERIA

FOR ESTABLISHING

BASEMENT

MEMBRANE

MECHANISMS

OF

THICKENING

with

permission.”

dldose reductase. There is no reason to believe that excessive glycosylation by any sugar moiety of a macromolecule in the extracellular matrix should lead to its increased hydration. Our own observations of capillaries with greatly thickened basement membranes in galactosemic rats have never demonstrated evidence of osmotic damage such as swelling or disruption of organelles, and we have not observed pericyte dropout” or other evidence of cellular necrosis. Hence, an explanation for the role of aldose reductase in the extracellular matrix abnormalities of diabetes and galactosemia that is analogous lo the osmotic hypothesis for diabetic and galactosenic cataractogenesis seems implausible. How can we explain these observations? At this point, only hypotheses can be offered but experimental systems are at hand by which they can be tested. One challenging hypothesi:, proposed by Rohrbach et al,‘* is that basement membrane thickening in diabetes results not from a primary overproduction of basement membrane macromolecules but, in tially, from a decrease in the synthesis of the major proteoglycan known as the BM-1 proteoglycan of the basem:nt membrane. Although this proteoglycan composes only a small fraction of the dry weight of the basement membranes, I9 it contains a high net negative charge as a result of it: high content of heparan sulfate and may serve as an el:ctrical charge barrier to decrease permeability of the b: sement membrane to charged molecules.20 According to the hypothesis.” reduced production of the BM-1 proteoglycon in some manner secondarily increases production of other basement membrane components (type IV collagen, laminin, and, perhaps, other proteins such as fibronectin and “cntactin”) leading to the observed basement membrane thickening. This hypothesis has been proposed on the basis of experiments on a tumor, the Engelbreth-Holm-Swarm (EHS) sarcoma, implanted into normal mice and mice made diabctic with streptozotocin. The tumor produces copious amounts of basement membrane macromolecules. However, other investigator? have not obtained the same results as Rohrbach et al. Equally important, the EHS tumor is

Minimal criteria for establishing that a biochemical mechanism is responsible for diabetes-induced basement membrane thickening are fairly straightforward: (1) The biochemical mechanism must be demonstrated in a tissue (preferably a tissue taken from normal animals or humans, ie, non-neoplastic) that has been shown to undergo basement membrane thickening in diabetes or galactosemia and the morphological changes of which were prevented or reversed with an aldose-reductase inhibitor. (2) The biochemical changes must be produced both by hyperglycemia and by galactosemia or by their in vitro equivalent, increased glucose or galactose levels in the culture medium, and they must be prevented by administration of an aldose-reductase inhibitor. These statements have several corollaries. First, aldose reductase-like activity must be demonstrable in the tissue using biochemical assays but it is not necessary to demonstrate by immunocytochemical methods that the enzyme is aldose reductase itself. A variety of aldehyde reductases exist in tissues with properties at least somewhat similar to aldose reductase. The best known of these is L-hexonate dehydrogenase (glucuronate reductase; L-gulonate: NADP+ loxidoreductase; EC 1.l. I .19). These enzymes may also be inhibited by drugs designed as “aldose reductase” inhibitors.23 In our laboratory (Kennedy A, Frank RN, unpublished results), we have observed biphasic Eadie-Hofstee plots of glyceraldehyde and glucuronate-reducing activity in the presence of NADPH when homogenates of whole brain and retina were used as the enzyme source. Such kinetics are usually taken to indicate the presence of two enzymes. with different K,‘s for the substrate. In the presence of an aldose reductase inhibitor, both branches of the curve showed inhibition. Ludvigson and Sorenson2” were unable to demonstrate immunocytochemically the presence of aldose reductase activity in retinal microvessels of the rat, and Kern and Engerman found no aldose reductase immunoreactivity in retinal microvessels of the dog. In each case, the antigen was highly purified aldose reductase from tissues of the species under investigation. Yet, we and Robison et al have demonstrated galactose-induced basement membrane thickening in retinal microvessels of the rat that is preventable by an aldose reductase inhibitor, and Kern and Engerman2” have found

38

ROBERT N. FRANK

sorbitol accumulation in canine retinal microvessels incubated in high-glucose medium while we*’ have found aldose reductase-like enzyme activity and sorbitol accumulation in cultured canine retinal microvessel endothelial cells. Secondly, mechanisms other than the osmotic effect of very high intracellular sugar alcohol concentrations may account for the results of the basement membrane studies that we and others have reported. For example, sugar alcohols in concentrations much too low to exert an osmotic influence on cells may alter activities of important biosynthetic or degradative enzymes with a role in the metabolism of specific components of the basement membrane. IN VITRO AND MECHANISMS

IN WV0

SYSTEMS

OF BASEMENT

effects of galactosemia in the presence and absence of an aldose reductase inhibitor on renal glomerular basement membrane morphology in such animals have not been done. Similar studies of basement-membrane morphology upon the lens capsule and cerebral microvessels of diabetic and galactosemic rats need to be done as a basis for the biochemical experiments to be outlined. Biosynthesis of new basement membrane collagen may be measured by injection of radio-labeled proline; of sulfated glycosaminoglycan by the use of [ “S]-sulfate; and of nonsulfated glycosaminoglycan by the use of [‘HI-glucosamine. A problem with the latter compound, however, is that it may be diluted by unlabeled glucosamine produced from the excess glucose present in diabetes,2’ producing a false impression of reduced glycosaminoglycan synthesis. This is a specific example of a general problem that is present in all such biosynthesis experiments involving radio-labeled precursors. Reduced labeling of the product under study may indicate reduced synthesis, increased degradation, or increased size of the unlabeled pool of precursor molecules, with resultant decrease of the specific activity of the radio-labeled precursor that is added by the investigator. Appropriate experiments should be carried out to distinguish among these possibilities. When they are successfully conducted, such experiments permit comparison of collagen with noncollagen protein synthesis from the ratio of radioactive counts in double-label studies in which animals are injected with

FOR STUDYING

MEMBRANE

THICKENING

Ideally, one should study the mechanism in an animal known to develop basement membrane thickening in diabetes and galactosemia. Biochemical studies of the retinal microvascular basement membrane of rats are difficult, however, because of the minute quantities of material that may be isolated. This problem does not apply to the basement membranes of the renal glomerulus, the cerebral microvessels, or the lens capsule, each of which can be isolated easily in amounts sufficient for study. Renal glomerular basement membrane thickening is known to occur in diabetes in the rat,** and its prevention in these animals with an aldosereductase inhibitor has also been reported.14 Studies of the

1

2

3

4

5

6

7

(I

9

10

11

2 B)

--

3

“4, w

1

6 _@

7

Collagens produced by retinal microvascular and lens epithelial cells in culture. The cells were labeled with [‘HI-L-proline and Fig 2. collagens were extracted from the medium and cell layer by limited pepsinization and differential salt precipitation. Because of the small amounts of collagen present in the cultures, type I collagen prepared from bovine aortic media was added as carrier. (A) illustrates Coomassie blue stained SDS-7.5% polyacrylamide gels from these experiments. Lanes 1 to 3 were taken from cultures of neonatal canine retinal microvessel endothelial cells; lanes 4 and 5 were taken from cultures of bovine lens epithelial cells: and lanes 6 and 7 were taken from cultures of bovine retinal microvessel pericytes. The @ region and a,(l) and a,(l) bands of the type I collagen carrier are the most prominent bands seen in these gels (read from top to bottom). Lanes 8 and 9 are collagens extracted from fresh bovine lens capsule. The arrows indicate bands that. in other experiments using immunoblotting techniques, we have identified as type IV collagen. Lane 10 contains type V collagen from bovine cornea1 stroma, and lane 11 is type I collagen from bovine aortic media. IB) illustrates autoradiograms of the labeled collagens from lanes 2 to 7 of A. Note the prominent o,(I) and cr,(l) bands in the pericyte and endothelial cell cultures that are lacking in the lens-cell cultures. No beta bands are present because the cells were cultured in the presence of j3#‘-aminodiproprionitrile, which prevents interchain crosslinking. Radiolabeled bands are present in all cultures (arrows. lane 4) corresponding to the position of the bands identified as type IV collagen in the lens capsule collagen gel (A. lanes 8 and 9).

BASEMENT

MEMBRANE

39

THICKENING

[‘HI-proline (which is incorporated into both collagen and noncollagen proteins) and [‘4C]-tryptophan (which is incorporated only into noncollagen proteins; cf”). Finally, quantitation of other basement membrane proteins, such as laminin or Tibronectin, may be carried out by immunoprecipitation witl specific antibodies.‘8*2’ Studies other than quantitative comparison of the synthesis of various macromolecules may also be carried out. These might include investigations of the degradation of these molecules and of their posttranslational modifications by glycosylation. sulfation, and other mechanisms. Such experiments are not technically unduly difficult but carried out in vivo, they suffer from the disadvantage that they require injecting numbers of animals with relatively large amounts of radioactivity. Another approach is to study the synthesis of basement membrane macromolecules in vitro. For retinal microvessel basement membrane, this is possible because the microvessel cells can be grown in culrure,‘7,30 32where they have been shown to produce collagers and glycosaminoglycans33-)5 (Figs 2 and 3). These macromolecules include type IV collagen and heparan sulfat{:. which are known to be present in basement membranes.

1

2

3

4

5

6

7 Fig 3. Autoradiogram of a series of cellulose acetate strips on which 3%-labeled glycosaminoglycans from a culture of bovine retinal capillary pericytes ware electrophoresed. Material from the combined medium and cell layer was used in this experiment. The material in lane 1 was first digested with chondroitinase ABC; in lane 2, with chondroitinase AC; in lane 3. with heparinase: in lane 4, with heparitinese; in lane 5, with Sfreptomyces hyaluronidase; in lane 6, with nitrous acid; and in lane 7. no prior enzyme or HNO, treatment was used. This experiment identifies the major band, to the right, as composed of chondroitin sulfates and the minor (trailing. leftmost) band as heparan sulfate.

Using microvessel cell cultures and conducting experiments with radio-labeled collagen and glycosaminoglycan precursors, we should be able to investigate in detail the mechanisms by which excess glucose and galactose lead to basement membrane thickening, whereas aldose reductase inhibitors prevent this effect.

REFERENCES I. Kinoshita JH: Cataracts in galactosemia. Invest Ophthalmol Vis Sci 4:786-799, 1965 2. Kinoshita JH, Fukushi S, Kador P, et al: Aldose reductase in diabetic complications of the eye. Metabolism 28:462-469, 1979 (SUPPI 1) 3. Gabbay KH: The sorbitol pathway and the complications of diabetes. N Engl J Med 28X:831&835, 1973 4. Gabbay KH: Hyperglycemia, polyol metabolism, and complications of diabetes mellitus. Ann Rev Med 26:521-~536, 1975 5. Young RJ, Ewing DJ, Clark BF: A controlled trial of sorbinil, an aldose reductase inhibitor, in chronic painful diabetic neuropathy. Diabetes 32:938-942, 1983 6. Jaspan J, Maselli R, Herold K, et al: Treatment of severely painful diabetic neuropathy with an aldose reductase inhibitor: Relief of pain and improved somatic and autonomic nerve function. Lancet 2:758-762, 1983 7. Engerman RL, Kern TS: Experimental galactosemia produces diabetic-like retinopathy. Diabetes 33:97- 100, 1984 8. Eaton RP, Sibbitt WL Jr, Harsh A: The elrect of an aldose reductase inhibiting agent on limited joint mobility in diabetic patients. JAMA 253:1437-1440, 1985 9. Kennedy A, Frank RN. Varma SD: Aldose reductase activity in retinal and cerebral microvessels and cultured vascular cells, Invest Ophthalmol Vis Sci 24:1250-l 258, 1983 10. Cogan DC, Toussaint D, Kuwabara T. Retinal vascular patterns. IV. Diabetic retinopathy. Arch Ophthalmol 66:366- 378. 1961 Il. King GL, Buzney SM. Kahn CR, et al: Differential responsiveness to insulin of endothelial and support cells from micro- and macrovessels. J Clin Invest 71:974-997. 1983 12. Frank RN, Keirn RJ, Kennedy A. et al: Galactose-induced retinal capillary basement membrane thickening: Prevention by sorbinil. Invest Ophthalmol Vis Sci 24: 15 19- 1524, 1983 13. Robison WG Jr. Kador PF, Kinoshita JH: Retinal capillaries: Basement membrane thickening by galactosemia prevented with aldose reductase inhibitor. Science 221:1177--l 179. 1983 14. Chandler ML, Shannon WA, DeSantis L: Prevention of retinal capillary basement membrane thickening in diabetic rats by aldose reductase inhibitors. Invest Ophthalmol Vis Sci 25:159. 1984 (SUPPI) 15. Robison WG Jr, Kador PF. Akagi Y. et al: Basement membrane thickening in ocular vessels of galactosemic rats prevented with aldose reductase inhibitors. Invest Ophthalmol Vis Sci 25:66, 1984 (Suppl) 16. Miller EJ: Comparison of basement membrane collagens with interstitial collagens, in Kefalides NA (ed): Biology and Chemistry of Basement Membranes. Orlando, Fla. Academic, 1978. pp 265278 17. Cohen MP, Urdanivia E. Surma M, et al: Nonenzymatic glycosylation of basement membrane. In vitro studies. Diabetes 30:367-371, 1981 18. Rohrbach DH. Wagner CW, Star VL, et al: Reduced synthesis of basement membrane heparan sulfate proteoglycan in streptozotocin-induced diabetic mice. J Biol Chem 25X:1 1672-I 1677, 1983

ROBERT N. FRANK

40

19. Kanwar YS, Farquhar MG: Isolation of glycosaminoglycans (heparan sulfate) from glomerular basement membranes. Proc Nat Acad Sci USA 76:44934497, 1979 20. Farquhar MG, Courtoy PJ, Lemkin MD, et al: Current knowledge of the functional architecture of the glomerular basement membrane, in Kuehn K, Schoene H-H, Timpl R (eds): New Trends in Basement Membrane Research. New York, Raven, 1982, pp 9-30 21. Pihlajaniemi T, Myllyla R, Kivirikko KI, et al: Efects of streptozotocin diabetes, glucose, and insulin on the metabolism of type IV collagen and proteoglycan in murine basement membraneforming EHS tumor tussue. J Biol Chem 257:14914-14920,1982 22. Engerman RL, Colquhoun PJ: Epithelial and mesothelial basement membranes in diabetic patients and dogs. Diabetologia 23:521-524,1982 23. Srivastava SK, Petrash JM, Sadana IJ, et al: Susceptibility of aldehyde and aldose reductases of human tissues to aldose reductase inhibitors. Curr Eye Res 2:407-410, 1982/1983 24. Ludvigson MA, Sorenson RL: Immunohistochemical localization of aldose reductase. II. Rat eye and kidney. Diabetes 29:450-459, 1980 25. Kern TS, Engerman RL: Distribution of aldose reductase in ocular tissues. Exp Eye Res 33: 175-l 82, 198 I 26. Kern TS, Engerman RL: Hexitol production by canine retinal microvessels. Invest Ophthalmol Vis Sci 26:382-384, 1985 27. Buzney SM, Frank RN, et al: Retinal capillaries: Proliferation of mural cells in vitro. Science 190:985-986, 1975

28. Steffes MW, Brown DM, Basgen JM, et al: Glomerular basement membrane thickness following islet transplantation in the diabetic rat. Lab Invest 41:116-l 18, 1979 29. White R, Carlson EC, Brendel K, et al: Basement membrane biosynthesis by isolated bovine retinal vessels: Incorporation of precursors into extracellular matrix. Microvasc Res 18: 185-208, 1979 30. Buzney SM, Massicotte SJ: Retinal endothelium in vitro. Invest Ophthalmol 1979

vessels: Proliferation of Vis Sci 18:1191-l 195.

3 I. Frank RN, Kinsey VE, Frank KW, et al: In vitro proliferation of endothelial cells from kitten retinal capillaries. Invest Ophthalmol Vis Sci 18:1196-1200,1979 32. Gitlin JD, D’Amore PA: Culture of retinal capillary using selective growth media. Microvasc Res 26:74-80, 1983

cells

33. Cohen MP, Frank RN, Khalifa AA: Collagen production by cultured retinal capillary pericytes. Invest Ophthalmol Vis Sci 19:90-94, 1890 34. Kennedy A, Frank RN: Retinal capillary pericytes and endothelial cells in culture produce similar collagens. Invest Ophthalmol Vis Sci 24:158, 1983 (Suppl) 35. Kennedy A. Frank RN, Mancini MA: Glycosaminoglycans of the retinal microvessel basement membrane and of cultured retinal vascular cells. Invest Ophthalmol Vis Sci 26, 1985, (suppl), (in press)