Retinal dysplasia in mice lacking p56lck - Nature

Oncogene (1998) 16, 2351 ± 2356 ã 1998 Stockton Press All rights reserved 0950 ± 9232/98 $12.00 http://www.stockton-press.co.uk/onc

Retinal dysplasia in mice lacking p56lck B Omri1,3, C Blancher1, B Neron1, MC Marty1, J Rutin1, TJ Molina2, B Pessac1 and P Crisanti1 1

CNRS UPR 9035, 75270 Paris cedex 06; 2Department of Pathology, Faculte de meÂdecine Broussais - Hotel Dieu, 75270 Paris cedex 06, France; 3Universite Pierre et Marie Curie

The product of the proto-oncogene p56lck is a nonreceptor tyrosine kinase member of the Src family. It is found in T cells (Marth et al., 1985, 1988) and in the mouse brain (Omri et al., 1996; Van Tan et al., 1996). In this report, we describe experiments showing that Lck is present in the mouse retina neurons. Lck gene expression was identi®ed after isolating and sequencing the speci®c 5' and 3' part of the cDNA obtained by RT ± PCR. In adult retina Lck immunoreactivity was most abundant in photoreceptor cells and within the outer plexiform layers. Staining was also observed in the inner nuclear and plexiform layers. In transgenic mice, the disruption of the Lck gene had serious consequences on the organization of the retina causing retinal dysplasia. These mice have partial retinal detachment with infolding and rosette formation in the photoreceptor sheet. These retinal abnormalities observed in Lck de®cient mice lead to the loss of normal architecture of the photoreceptor and the inner nuclear layers, and provide an important role of Lck protein in the retina development. The lack of the Lck protein produces a spectrum of retinal pathology that resembles human retinopathy of prematurity (ROP). Keywords: Lck; neuroretina; dysplasia

Introduction Protein tyrosine kinases can be divided in two general classes. One, the receptor-type kinases, are receptors for a number of hormones, mitogens and growth factors (Yarden and Ullrich, 1988). The other class, a non receptor-type kinases, includes the family of Src proto-oncogene-related kinases. Several members of this family including c-Src, Fyn, Yrk c-Yes and Lck have been described in developing and regenerating neurons (Grant et al., 1992; Ignelzi et al., 1994; Yagi 1994; Omri et al., 1996; Van Tan et al., 1996) but their functions in these cells have not been clearly elucitated. p60src has been fully characterized in the neural retina (Sorge et al., 1984). The localization of Src and fyn in the mature post mitotic retinal neurons has been reported by Maness et al., (1988) and Ingraham et al., (1992). In the thymus, p56lck and p60fyn are the two major tyrosine kinases of the c-src family, p56lck being involved strongly in early thymocyte development (Molina et al., 1992). Therefore it was interesting to determine whether the Lck is also expressed, like Fyn and Src proteins, in retinal neurons. If it is so, these two tyrosine kinases could play critical roles in the development of mice retina. Correspondence: P Crisanti Received 24 July 1997; revised 1 December 1997; accepted 1 December 1997

We report here that Lck gene is expressed at the onset of terminal di€erentiation in post mitotic neurons and in the plexiform layers. of normal mouse neuroretina. Futhermore, in mice lacking Lck, generated by homologous recombination after Lck disruption gene (Molina et al., 1992), histological examination revealed a retinal abnormal morphogy with retinal folds, partial retinal detachment and multiple areas where normal retinal architecture was lost and rosette formation was present.

Results Lck mRNA is found in adult mouse retina RT ± PCR analysis Various sequences from mouse retina and thymus Lck mRNA were ampli®ed by RT ± PCR using primers as described in the experimental procedures. A fragment of 1476 bp, corresponding to the Lck coding sequence, was present in retina (Figure 1A1) and thymus (Figure 1A2) samples. A speci®c fragment of 164 bp was ampli®ed from the 5' domain of the Lck sequence, from both retina (Figure 1B1) and thymus (Figure 1B2) extracts. Furthermore the ampli®cation of the 5' region coding for the SH3 domain and a large part of the SH2 domain, produced, as expected, fragments of 434 bp from both retina (Figure 1C1) and thymus (Figure 1C2). We con®rmed that the sequence of the ampli®cation products of 164 bp, the Lck speci®c sequence, and the 279 bp of the 3' coding region were identical to the thymus Lck sequence. Thus the Lck gene is expressed in adult mouse retina. Identi®cation of retina Lck protein by immunoblot analysis We tested whether the Lck protein is present in the adult mouse retina. Retina cell lysates were immunoblotted with two antisera, one directed against the amino-terminal region of Lck and the another against epitope in the carboxy-terminal region of Lck (Figure 2F). In both cases immunoblots showed an immunoreactive band in adult retina extract (Figure 2Fb). The apparent molecular weight of the protein (Figure 2Fa), 56 kDa, was similar to that of the Lck product detected in thymus extracts. Although the intensity of the band in retina samples was about tenfold lower than in the thymus extract, it was consistently reproduced with both antisera. To con®rm that the retina protein recognized by these antisera to Lck was indeed Lck, we also tested retina extracts from mice lacking the Lck gene (Molina et al., 1992) (Figure 2Fc). No speci®c Lck immunoreactivity was detected with any anti-Lck antiserum. These data clearly demon-

Retinal dysplasia in mice lacking p56lck B Omri et al

2352

strate the presence of Lck protein in the normal mouse retina. Localization of Lck gene expression (a) Localization of Lck mRNA in mouse retina The distribution of Lck mRNA in the adult mouse retina was investigated by in situ hybridization using the antisense probes described in Materials and methods. On sections an intense hybridization signal was observed in neurons throughout the entire retina (Figure 2A). In general, the extent of labeling appeared to parallel local neuronal packing density. However labeling was not even in the di€erent neuronal layers. All the photoreceptor cells were stained but, in some regions of the inner nuclear layer, the signal seemed to be restricted to cell subsets. The ganglion cell layers were clearly labeled. No hybridizing signals were observed in control experiments with adjacent sections hybridized with the corresponding sense probes (not shown). Similarly antisense probes did not label transgenic mice in which the Lck gene was disrupted (Figure 2B). In situ hybridization analysis thus con®rmed the production of Lck mRNA in mouse retina. (b) Immunohistochemistry The distribution of Lck was analysed in adult mouse retina by immunofluorescence staining on cryosections (10 mm). The highest levels of p56lck immunoreactivity were seen throughout the outer plexiform layer and in photoreceptor cells particularly in the inner segment and in the inner nuclear layer. Bipolar, amacrine and ganglion cells and the inner plexiform layer were also immunoreactive but

more weakly (Figure 2C). As shown (Figure 2D) no staining was oberved when lck antiserum was preincubated with its corresponding peptide used for the antibody production. Since the irregular staining seen in the inner nuclear layer (Figure 2A) would not allow us to con®rm that was exclusively a neuronal staining and to determine whether, MuÈller cells were also stained, we carried out immunocytochemistry experiments on primary cultures of mouse retina of 16 Ed old after three weeks of culture. (Figure 2E). Results shows that essentially cells with a typical neuronal morphology with their processes were stained and not signi®cant labeling was observed in large ¯at MuÈller cells. (c) Histology of the retina in Lck-de®cient mice We examined the anatomical e€ect of the disruption of the

ONL OPL INL

GCL

INL

ONL GCL OPL

A

B

C

D

))

1476 bp

434 bp 279 bp

164 bp

1 2

1 2

1 2

1 2

Figure 1 Southern hybridization of reverse transcriptase PCR products. (A) Expected fragment of 1476 bp corresponding to the coding sequence from the middle of exon 1' to exon 12 in retina (1) and in thymus (2). (B) 164 bp fragment corresponding to the speci®c unique domain of Lck sequence in retina (1) and in thymus (2). (C) Ampli®cation of the 5' unique domain plus the region coding for the SH3 and a large part of SH2 domains corresponding to a fragment of 434 bp in retina (1) and in thymus (2). (D) The 3' coding region fragment of 279 bp in retina (1) and in thymus (2)

Figure 2 (a) Localization of Lck mRNA and protein in adult mouse retina sections: in situ hybridization. The labeling can be observed: ± throughout the outer nuclear layer (photoreceptor cells). ± in only a subset of cells of the inner nuclear layer and ± in the ganglion cell layer. (b) No labeling is observed in the retina of the Lck minus mouse. Immunocytochemistry with anti Lck serum (Santa Cruz SC-13 N8 2102). (c) In the adult retina intense labeling is observed in the outer plexiform layer, in the outer nuclear layer, and in the externe segments of the photoreceptor cells. Weaker staining is also observed in the inner nuclear layer, in ganglion cells, and in the inner plexiforn layer. (d) Negative control performed with the anti-Lck serum preincubated with its corresponding peptide control as described in Omri et al., (1996). (e) Primary cultures from ED16 retinas were immunolabeled after 3 weeks with anti-Lck serum. Somas and processes of neuronal cells are clearly stained. No labeling is observed in MuÈller cells. Bar: 50 mm for a and b (f) Detection of the Lck protein in mouse retina. Protein extracts were prepared from normal thymus (a) and adult (b) mice and from the retina of a mouse lacking Lck gene product (c) After electrotransfer to nitrocellulose membranes, proteins were probed with an antiserum to the Lck protein (Santa Cruz SC-13 N8 2102)

Retinal dysplasia in mice lacking p56lck B Omri et al

Lck gene on the retina of transgenic lck7/7 and lck+/7 mice at two stages: 3 weeks and 2 months after birth. Because of the poor development of the homozygote mice we could only study few individuals. For the homozygote mice six 3 week old retinas and ten 2 month old retinas after birth were studied. For the heterozygote mice, 16 (3 week, 2 and 3 month) old retinas were examined At these stages retina is developed and di€erentiated. In 3 week old homozygote mice, only partial retinal detachment and retinal infolding was observed. (Figure 3b and d). At this early stage of the dysplastic changes, the cell bodies of the photoreceptors include the external segment of the photoreceptors and the pigmented cells of the retinal pigmented epithelium. Rosette formation was not observed at this stage (Figure 3b and d). In 2 month old homozygote mices, we observed rosettes formation (Figure 4b, c and d). Rosette formation started from the photoreceptor layer. After formation, the rosettes appeared to migrate into the inner nuclear layer with loss of normal retinal architecture. Retinal folds were numerous at the anterior pole of the eye whereas rosettes were more numerous at the posterior pole although the number of rosettes was variable. Some retinas showed groups of numerous small rosettes and others larger and more widely spaced rosettes. In all cases the changes were in distinct areas and the remainder of the retina was entirely normal. Pathological, retinal infolding and rosettes seemed to appear at the end of development. Conversly, in heterozygote mice, no retinal infolding or rosette formation was detected and all the retinas observed were morphologically normal. To test whether the pathology was associated with the mouse strain per

se, the retina morphology of two other normal mouse strains C57 black and Balb3T3 was also studied. All the retinas examined were completely normal. Discussion In this report we demonstrate expression of the Lck gene in the adult mouse retina. This retina tyrosine kinase is indistinguishable from that in the thymus. In adult neuroretina Lck was mostly localized in photoreceptor cell bodies comprising the outer nuclear layer and in the outer plexiform layer. Lck immunoreactivity was also found in amacrine and ganglion cells and in the inner plexiform layer. This localization is similar but not identical to that reported for c-Src (Sorge et al., 1984) and Fyn (Ingraham et al., 1992) tyrosine kinases. The role of these tyrosine kinases in the central nervous system has not been clearly elucidated. Most of these enzymes are produced in post mitotic retinal neurons, and may thus have functions other than mitogenic signal transduction. They may be involved in cell-cell communication and transfer of information within neural networks. To understand the role of these non-receptor tyrosine kinases, transgenic mice obtained by gene knockout techniques have been studied. Transgenic mice in which the Fyn gene has been disrupted show impairment of both LTP and spatial learning and abnormal hippocampal development (Grant et al., 1992). In Src or Yes (Yagi 1994) de®cient mice, no apparent neurological abnormalities have been reported. Although, no morphological

Figure 3 Semithin sections (1 mm thick) (a) and (c) controls. (b) and (c) fold of outer aspect of the retina (arrow) of Lck de®cient mice

2353

Retinal dysplasia in mice lacking p56lck B Omri et al

2354

ONL OPL INL

IPL

GCL

Figure 4 Cryosections (15 mm thick) (a) control. (b), (c) and (d) retina with rosettes composed of photoreceptors cells in Lck de®cient mice and Figure 3 Bar: 15 mm Figure 4d Bar: 50 mm

studies have been reported for the retina. In Lck de®cient mice we show that the morphology of the retina is altered particularly in the photoreceptor cells and in the inner nuclear layer. There were two major abnormalities induced by the absence of Lck protein. First, partial retinal detachment with retinal infoldings and second, rosette formation. There are two major examples of multigene disorders causing retinal disease and which lead to rosette formation. Retinal dysplasia with rosette formation has previously been associated with the loss or absence of retinal pigment epithelium (RPE) (Silverstein et al., 1971) or RPE function. Dysplasia and rosettes have been reported as features of oxygen-induced retinopathy in the rat model for retinopathy of prematurity (ROP) (Penn et al., 1992) and by photosensitization induced retinopathy in the newborn Beagle (Sadda et al., 1994). Rosettes have also been decribed in the retinoblastoma (Flexner 1981; Ts'o 1969). Retinoblastoma is the most common ocular malignancy in newborn babies and infants. It originates during retinal development and the primitive phenotype is consistent with the appearance of the cells in the 6 to 10 weeks fetal retina (Popo€ and Ellsworth 1969). The histogenesis of retinoblastoma has been attributed to neuronal, glial, and/or primitive precursor cells. In our study rosettes form from photoreceptor layer (Figure 4b, c and d). The homogeneity of rosette type cells di€erentiate these rosettes from those of the retinoblastoma, consistent

with Lck being a protooncogene. Furthermore in Lck de®cient mouse retina, no signi®cant increase in cell increase in cell proliferation was observed. It would be interesting to determine whether Lck is strongly expressed in retinoblastoma. In our model, the lack of Lck protein resulted in retinal injury similar to that of ROP frequently seen in human infants and to that resulting from a photosensitization mechanism by which oxygen and light can interact to produce oxygen-derived free radicals, such as superoxide. ROP is a multifactorial disease and numerous risk factors in addition to light and oxygen such as xanthine administration, vitamin E de®ciency, gestational agebirth weight (Zierler 1988; Hammer et al., 1986; Sira et al., 1985) have been reported. A number of these suggested risk factors are consistent with an oxidative mechanism of injury; the lack of sucient antioxydant, such as vitamin E (Penn et al., 1992), would be expected to exacerbate the severity of oxidative retinal damage. Indeed the prophylatic administration of vitamin E, an oxygen reactive product, reduces the incidence and severity of ROP (Quinn et al., 1990) Lck tyrosine kinase may prevent excess reactive oxygen production and thus, may be necessary for normal retinal physiology. The fact that the pathological manifestations appeared at the end of retinal development implicates Lck in the maintenance of the physiology of the photoreceptor cells. It would therefore be interesting to investigate fyn expression

Retinal dysplasia in mice lacking p56lck B Omri et al

and localization in mice lacking Lck, as this tyrosine kinase can partially substitute for Lck (Groves et al., 1996; Van Oers et al., 1996) although, this substitution is not sucient to compensate completely for the loss of the Lck protein.

Materials and methods Mice Mice mutant for lck (lck7/7) have been previously described (ref). All mice were kept in accordance with institutional guidelines. PCR detection Total RNA was isolated from PBS perfused mice and a random primed single strand cDNA was synthesized with reverse transcriptase and ampli®ed using oligonucleotide primers. Primers 5'-CCTGAAGATGACTGGATGGA-3' (exon1') and 5'-CAAGACCTACTGAAGAAGTGTCGG3' (exonl2) (TM=588C) were used to amplify almost the entire coding sequence. Primers 5'-GGATCATGG GCTGTGTCTC-3' (exon1') and 5'-GGTGACCAGTGGATACTCC-3' (exon2) (TM=588C) and primers: 5'-CCT GAAGATGACTGGATGGAGAA-3' (exonl') and 3'-CAC GCATGGATCCTTCCTGAT-5' (exon5) were used to amplify the 5' end of the mouse Lck sequence. Two primers set: 5' - GCACCAGAAGCCATTAACTATG - 3' (exon1) and 3' - GGCTGTGTGAAGAAGTCATCCA GAAC-5' (exon 12) ( TM=568C) were used to amplify the 3' part of the coding sequence. For each ampli®cation, we obtained a PCR product with the expected size of: 1476 bp, 164 bp, 434 bp or 279 bp respectively. Primers were synthesized by GENSET. Ampli®cations were performed by 40 cycles in a Perkin-Elmer/cetus DNA thermal cycler using a cycle of denaturation for 1 min at 948C, annealing for 1 min at 588C or 568C and extension at 728C for 1 min. PCR products were separated electrophoretically in 1.3% agarose gel and transferred to a nylon membrane using the alkaline transfer method (Chomczynski and Asba, 1984). [32P]dATP oligonucleotides used as probes to detect the correct PCR products were as follows: 5'-CCTGAAGATGACTGGATGGA-3' for the ®rst and the third ampli®cation; 5'-CCACTGGTCACCTA TGAGG-3' for the second ampli®cation; and 5'-CCAC GGTCGAATCCCTTACC-3' for the fourth ampli®cation. In situ hybridization Mouse retinas were ®xed overnight in 4% paraformaldehyde (in PBS) and then incubated overnight in 30% sucrose. Sections of 15 mm were freeze-cut. The probes were an antisense 20 mer (3'-TCCCTCATAGGTGACCAGTGG-5') corresponding to the 5' speci®c part of the Lck gene, 24 mer (3' - G GCTGTGAAGAAGTCATCCAGAAC - 5') corresponding to the 3' part of the translated sequence and a 29 mer corresponding to the 3' untranslated sequence (3' - C CAAGGCT TAGACTCACGTGCTC CACAGG - 5'). Three corresponding sense oligonucleotides (5'-CCACTGGTCACCTATGAAGGG-3'), (5'-GTTCTGGATGACTTCTTCACAGCC-3') and (ACCTGTGGAGCA CGTGAGTCTAAGCCTTGGA) were used as controls. Oligonucleotides was tailed with digoxigenin-labeled nucleotide (DIG-dUTP) as described in the Boehringer kit (oligonucleotide tailing kit 14172). Hybridization of neuroretina sections were performed overnight at 378C with 10 ml of probe (10 ± 20 ng/ml) in 50% formamide 56SSC, SDS 0,02%, blocking reagent 5%, N lauryl sarcosine 0,1%. Washing was performed at RT in 26SSC 30 min.; 16SSC

30 min; 0.56SSC 30 min and the tissue was treated quickly in tris HCl 100 mM, NaCl 150 mM pH 7.5 and then incubated in the Boehringer blocking reagent. A conjugated antibody (anti-digoxigenin alkaline phosphatase conjugate [DIG]AP) was then used to detect the probe and a subsequent phosphatase alkaline-catalyzed color reaction with X phosphate and Nitroblue tetrazolium salt produce a precipitate (Nucleic acid detection kit, Boehringer). Protein analysis Retina and thymus extracts were prepared from PBS perfused 2 month old mice. Tissues were homogenized with Te¯on pestle in 20 vol (w/v) of bu€er A: 50 mM Tris Hcl pH 7.5,1 mM EDTA (ethylene diaminetetraacetic acid), 20 mg/ml leupeptin, 1 mM PMSF(phenyl methyl sulfonyl ¯uoride), 1% NP40, 10% glycerol, and 5 mM b-mercaptoethanol. After 30 min at 48C, homogenates were cleared by centrifugation at 10 000 g for 15 min. The supernatants were immediately used for immunoblot analysis on SDS ± PAGE as described in (Omri et al., 1994) and for tyrosine protein kinase determinations. Two di€erent polyclonal antibodies raised against Lck were used for immunoblot analysis. (1) An anity-puri®ed rabbit polyclonal antibody raised against a peptide identical to amino acids 476 ± 505 in the carboxy terminal domain of human Lck (from Santa Cruz Biotechnology). This antibody recognizes both human and mouse Lck and has no detectable crossreactivity with other members of the Src protein kinase family. (2) An anity-puri®ed rabbit polyclonal antibody raised against a peptide identical to amino acids 38 ± 62 the unique NH-region of the Lck protein (Fischer et al., 1987). Immunohistochemistry In vivo Mice were sacri®ced by an overdose of ether and transcardially perfused with 4% paraformaldehyde in 0.1 M phosphate bu€ered saline (PBS, pH 7.4). Then the eyes were removed and post®xed in the same ®xative for 2 h at 48C. After cryoprotection by immersion in 20% sucrose, 8 mm thick cryostat sagittal sections were collected on `siliconated' slides and kept at 7208C. For immunolabeling the retina sections were air-dried and incubated for 2 min in a solution of 1% Triton X100 in PBS. After washing in PBS, the sections were incubated overnight with the anti-Lck (Santa Cruz SC-13 N8 2102) serum diluted 1/ 30 and then with a 1/50 dilution of F(ab')2 FITC goat antirabbit serum (CALTAG code n8 L43001) for 2 h. All incubations were at room temperature. In vitro Mice neuroretinas were explanted from ED16 and incubated with 0.25% of trypsin in Versen during 30 min at 378C. After gentle dissociation, cells were cultured during 3 weeks in Neurobasal Medium (Gibco BRL) with 5% of horse serum (Sigma), 0.5X B27 supplement (Gibco BRL), and 1% HEPES. Medium was renewed twice a week. Immunohistochemistry was performed as described above. Light Microscopy The eyes were enucleated and ®xed overnight with 4% paraformaldehyde in PBS. They were cryoprotected in 30% sucrose in PBS and cryosections of 16 mm were prepared. Sections were stained by the hematoxylin-eosin technique. Ultrathin sections of 1 mm were obtained by the technique described by Rajaofetra et al., (1992) and stained with azur III and methylene blue. Acknowledgements This work was supported by Centre National de la Recherche Scienti®que and Agence Nationale de Re-

2355

Retinal dysplasia in mice lacking p56lck B Omri et al

2356

cherches sur le SIDA and Grants from Association de Recherche sur le Cancer, and Association Retina France.

We thank Michel Louette for photographic work and Suzanne Barrey for secretarial assistance.

References Chomczynski PQ and Asba PK. (1984). Biochem. Biophs. Res. Comm., 122, 340. Fischer S, Fagard R, Boulet I and Gesquiere JC. (1987). Biochem. Biophys. Res. Comm., 143, 819 ± 826. Flexner RA. (1981). Johns Hopkins Hosp. Bull, 2, 115 ± 119. Grant SGN, O'Dell TJ, Karl KA, Stein PL, Soriano P and Kandel ER. (1992). Science, 258, 1903 ± 1910. Groves T, Smiley MP, Forbush K, Perlmutter RM and Guidos CJ. (1996). Immunity, 5, 417 ± 428. Hammer ME, Mullen PW, Ferguson JG, Pai S, Cosby C and Jackson KL. (1986). Am. J. Ophtalmol., 102, 1 ± 6. Ignelzi Jr MA, Miller DR, Soriano P and Maness PF. (1994). Neuron, 12, 873 ± 884. Ingraham CA, Cooke MP, Chuang YN, Perlmutter RM and Maness PF. (1992). Oncogene, 7, 95 ± 100. Maness PF, Aubry M, Shores CG, Frame L and Pfenninger KH. (1988). Proc. Natl. Acad. Sci. USA, 85, 5001 ± 5005 Marth JD, Cooper JA, King CS, Ziegler SF, Tinker DA, Overall RW, Krebs EG and Perlmutter RM. (1988). Mol. Cell. Biol., 8, 540 ± 550. Marth JD, Peet R, Krebs EG and Perlmutter RM. (1985). Cell, 43, 393 ± 404. Molina TJ, Kishihara K, Siderovski DP, Van Ewijk W, Narendran A, Timms E, Wakeham A, Paige CJ, Hartmann KU, Veillette A, Davidson D and Mak TW. (1992). Nature, 357, 161 ± 164. Omri B, Crisanti P, Alliot F, Marty MC, Rutin J, Levalleis C, Privat A and Pessac B. (1994). Int. Immunol. 6, 377 ± 385.

Omri B, Crisanti P, Marty MC, Alliot F, Fagard R, Molina T and Pessac B. (1996). J. Neurochemistry, 67, 1360 ± 1364. Penn JS, Tolman BL, Lowery LA and Koutz CA. (1992). Curr. Eye Res., 11, 939 ± 953. Popo€ N and Ellsworth RM. (1969). J. Ultrastruct. Res., 29, 535 ± 549. Quinn GE, Johnson L, Otis C, Scha€er DB and Bowen FW. (1990). Doc. Ophtalmol., 74, 223 ± 228. Rajaofetra N, Poulat P, Marlier L, Ge€ard M and Privat A. (1992). Dev. Brain. Res., 67, 237. Sadda SR, Yu YS, De Juan Jr E, Rencs EV, Green WR and Gottsch JD. (1994). Invest. Ophtalmol., 35, 1202 ± 1211. Silverstein AM, Osburn BI and Prendergast RA. (1971). Am. J. Ophtalmol., 72, 13 ± 21. Sira BS, Nissenkorn I and Kremer I. (1985). Surv. Ophtalmol., 33, 1 ± 16. Sorge LK, Levy BT and Maness PF. (1984). Cell, 36, 249 ± 257. Ts'o MOM. (1969). Arch. Path. 88, 664 ± 671. Van Oers NSC, Lowin-Kropf B, Finlay D, Conolly K and Weiss A. (1996). Immunity, 5, 429 ± 439. Van Tan H, Allee G, Benes C, Viammey Barnier J, Vincent JD and Fagard R. (1996). J. Neurochem., 67, 2306 ± 2315. Yagi T. (1994). Develop. Growth and Di€er., 36, 543 ± 550. Yarden Y and Ullrich A. (1988). Ann. Rev. Biochem., 57, 443 ± 478. Zierler S. (1988). Retinopathy of prematurity: Problem and Challenge. Flynn JT. and Phelps DL. (Eds) Alan R. Liss, New York. pp. 23 ± 33.