Mutations in Lama1 Disrupt Retinal Vascular Development and Inner Limiting Membrane Formation*

The Neuromutagenesis Facility at the Jackson Laboratory generated a mouse model of retinal vasculopathy, nmf223, which is characterized clinically by vitreal fibroplasia and vessel tortuosity. nmf223 homozygotes also have reduced electroretinogram responses, which are coupled histologically with a thinning of the inner nuclear layer. The nmf223 locus was mapped to chromosome 17, and a missense mutation was identified in Lama1 that leads to the substitution of cysteine for a tyrosine at amino acid 265 of laminin α1, a basement membrane protein. Despite normal localization of laminin α1 and other components of the inner limiting membrane, a reduced integrity of this structure was suggested by ectopic cells and blood vessels within the vitreous. Immunohistochemical characterization of nmf223 homozygous retinas demonstrated the abnormal migration of retinal astrocytes into the vitreous along with the persistence of hyaloid vasculature. The Y265C mutation significantly reduced laminin N-terminal domain (LN) interactions in a bacterial two-hybrid system. Therefore, this mutation could affect interactions between laminin α1 and other laminin chains. To expand upon these findings, a Lama1 null mutant, Lama1tm1.1Olf, was generated that exhibits a similar but more severe retinal phenotype than that seen in nmf223 homozygotes. The increased severity of the Lama1 null mutant phenotype is probably due to the complete loss of the inner limiting membrane in these mice. This first report of viable Lama1 mouse mutants emphasizes the importance of this gene in retinal development. The data presented herein suggest that hypomorphic mutations in human LAMA1 could lead to retinal disease.

The Neuromutagenesis Facility at the Jackson Laboratory generated a mouse model of retinal vasculopathy, nmf223, which is characterized clinically by vitreal fibroplasia and vessel tortuosity. nmf223 homozygotes also have reduced electroretinogram responses, which are coupled histologically with a thinning of the inner nuclear layer. The nmf223 locus was mapped to chromosome 17, and a missense mutation was identified in Lama1 that leads to the substitution of cysteine for a tyrosine at amino acid 265 of laminin ␣1, a basement membrane protein.

Despite normal localization of laminin ␣1 and other components of the inner limiting membrane, a reduced integrity of this structure was suggested by ectopic cells and blood vessels within the vitreous. Immunohistochemical characterization of nmf223
homozygous retinas demonstrated the abnormal migration of retinal astrocytes into the vitreous along with the persistence of hyaloid vasculature. The Y265C mutation significantly reduced laminin N-terminal domain (LN) interactions in a bacterial twohybrid system. Therefore, this mutation could affect interactions between laminin ␣1 and other laminin chains. To expand upon these findings, a Lama1 null mutant, Lama1 tm1.1Olf , was generated that exhibits a similar but more severe retinal phenotype than that seen in nmf223 homozygotes. The increased severity of the Lama1 null mutant phenotype is probably due to the complete loss of the inner limiting membrane in these mice. This first report of viable Lama1 mouse mutants emphasizes the importance of this gene in retinal development. The data presented herein suggest that hypomorphic mutations in human LAMA1 could lead to retinal disease.
Laminins are a family of heterotrimeric glycoproteins required for basal lamina formation (1,2). The 15 known laminin isoforms each contain one of the five ␣, three ␤, and three ␥ chains and have distinct functions and tissue expression patterns. Although ␣1 is the first ␣ chain to appear during embryonic development and is widely expressed during organogenesis (3,4), its expression in adults is highly restricted, being observed primarily in the central nervous system, eye, kidney, digestive system, and reproductive organs (5). Within the eye, laminin ␣1 is expressed in the inner limiting membrane as well as the ciliary body and lens (5)(6)(7). Furthermore, the gene encoding laminin ␣1, LAMA1, and several peptides derived from laminin ␣1 have been shown to induce angiogenic sprouting in cell culture and angiogenesis in chicken chorioallantoic membrane assays (8,9). It has been difficult to study the functional role of this gene beyond embryogenesis, because all Lama1 mouse mutants generated to date die by embryonic day 7 (10,11). Mutant zebrafish, however, have provided some insights into the function of lama1 (12)(13)(14)(15)(16). Zebrafish produced with lama1 knockdown morpholinos were shown to develop severe lens degeneration by 48 h postfertilization with apparent reduction in eye size by 72 h (16). In separate studies, zebrafish with mutations in lama1 were shown to have defects in axonal guidance (13,14). Another study identified a zebrafish lama1 mutant, lama1 a69 , in a forward genetic screen that carries a cysteine to serine substitution at amino acid 56 (15). At 10 days postfertilization, lama1 a69 zebrafish have a reduced body axis and early lens degeneration as well as defects of retinal ganglion cells and of hyaloid vessels, which provide nutrients to the retina prior to its vascularization. Lama1 zebrafish mutants also exhibit inner limiting membrane disruptions and shortened Müller cell end-feet (12). Although these studies indicate the importance of lama1 in retinal development, the effects of these mutations in the adult retina could not be determined because they are larval lethal. In particular, the potential involvement of lama1 in retinal vascular development could not be discerned because zebrafish models die prior to the initiation of this process.
The present study reports a mouse mutant in which a missense mutation generated by chemical mutagenesis in the Lama1 gene disrupts retinal development. In order to better understand the functional consequences of this mutation, a indirect ophthalmoscopy. Lama1 ⌬ and Lama1 floxed animals at INSERM were maintained under similar conditions with the exception that they were fed a 3.1% fat diet. For all experiments, a minimum of three control and three mutant mice were investigated.
Genetic Mapping and Analysis-B6-nmf223 homozygotes were mated to DBA/2J mice to generate F1 progeny, which were subsequently intercrossed. F2 progeny were phenotyped at 4 weeks of age by indirect ophthalmoscopy. DNA was isolated from tail snips using a modified version of previously described methods (18). A genome-wide scan using simple sequence length polymorphic markers to determine the map position of nmf223 was performed by The Jackson Laboratory Fine Mapping Facility. Additional simple sequence length polymorphic markers and single nucleotide polymorphisms were used to narrow the critical region encompassing the nmf223 mutation. All unaffected recombinant mice were progenytested by mating them to B6-nmf223 homozygotes to determine if they carried the nmf223 mutation. A minimum of 20 offspring from each progeny test were genotyped and phenotyped. PCR amplification was carried out as follows: 94°C for 2 min, 94°C for 30 s, annealing temperature of 55°C for 45 s, and 72°C for 1 min. Steps 2-4 were repeated for 35 cycles, followed by one cycle at 72°C for 2 min. PCR products were separated by electrophoresis on a 4% MetaPhor-agarose (Lonza) gel, stained with ethidium bromide, and visualized under UV light.
For sequencing of candidate genes, total RNA was prepared from three mutant mice and three control, B6 mice. RNA was isolated from snap-frozen eyes using TRIzol (Invitrogen), according to the manufacturer's instructions. The Retroscript kit (Ambion) was used to generate cDNA. All genes within the critical region were amplified from cDNA using PCR assays, and purified products were sequenced by The Jackson Laboratory Allele Typing and Sequencing Service.
Complementation Testing-In order to verify that the mutation identified in Lama1 was the nmf223 disease-causing mutation, a complementation test was performed with the previously described Lama1 ⌬ heterozygotes. In vitro fertilization technology was used to generate offspring from female nmf223 homozygotes and male Lama1 ⌬ heterozygotes. Progeny from the in vitro fertilization procedure were genotyped for the Lama1 ⌬ allele and examined clinically by indirect ophthalmoscopy. A minimum of three mice with each genotype were examined.
Clinical Assessment-Mice were phenotyped by indirect ophthalmoscopy and by fluorescein angiography according to previously described methods (19).
Electroretinogram (ERG) Testing-Two systems were used to examine ERGs in nmf223 and Lama1 ⌬ mice. In both cases, responses of mutant mice were compared with those obtained from WT littermates, and a minimum of five mice per genotype were tested. The ERG system used to test nmf223 mutants has been described previously (19). For Lama1 ⌬ mice, the following procedure was used. Mice were dark-adapted for at least 12 h and subsequently handled under dim red light. Anesthesia was achieved through intramuscular injection of a mixture of ketamine (120 mg/kg) and xylazine (14 mg/kg) diluted in saline solution. The left pupil was dilated with a drop of 1% atropine on the cornea and vibrissae were trimmed. Animals were placed on a heating plate to prevent anesthesia-induced hypothermia. The reference and ground electrodes were placed in the cheek and in the tail, respectively. The measuring gold electrode was positioned onto the cornea with a drop of methylcellulose. ERGs were recorded using VisioSystem and a Ganzfeld stimulator (SIEM Bio-Médicale, Nîmes, France). Signals were low pass-filtered at 300 Hz prior to digitization. For dark-adapted ERGs, 5-12 responses were averaged, depending on the stimulus intensity (12 for 0.001, 0.0032, 0.01, and 0.032 cd s/m 2 and 5 for 0.1, 0.32, 1, and 1.5 cd s/m 2 ). Flashes were separated by 10 s for lower intensities and by 1 min for the higher stimulus intensities. All flashes were 3 ms in duration, except for 1.5 cd s/m 2 , for which flash duration was increased to 5 ms. To extract the aand b-waves, dark-adapted ERG data were band pass-filtered at 1-70 Hz. The a-wave amplitude was measured from the base line to the first negative peak, whereas the b-wave amplitude was measured from the a-wave to the next positive peak. After a 5-min adaptation to a background light of 24 cd/m 2 , light-adapted ERGs were recorded with flash durations of 10 ms (1 and 3.2 cd s/m 2 ) or 30 ms (10 cd s/m 2 ), and five responses were averaged.
Optomotor Testing-Optomotor responses were obtained and analyzed as described previously (20).
To better understand the vascular abnormalities observed, retinal vessel development in WT and Lama1 nmf223 homozygotes was examined longitudinally using marker analysis of retinal whole mounts. The development of retinal astrocytes was also investigated because these glial cells are known to guide endothelial cell migration and retinal vascular development (24). Astrocytes were labeled using anti-GFAP, whereas endo-thelial cells and developing blood vessels were visualized with Griffonia simplicifolia isolectin conjugated with Alexa Fluor 594, which also binds to endothelial cells as well as macrophages. Hyaloid vessels were distinguished from retinal vasculature by the plane of focus, a more intense G. simplicifolia isolectin staining, and staining with anti-endostatin (25,26).
For whole mount experiments, all eyes were fixed in 4% paraformaldehyde for 30 min prior to the dissection of retinas in phosphate-buffered saline. For staining of retinal vessels, G. simplicifolia isolectin (1:200; Molecular Probes) was applied simultaneously with the secondary antibody onto retinal whole mounts. Previously described methods were used for all other histological and immunohistochemical assessments, including electron microscopy (19). Labeling was visualized by fluorescence microscopy using a Leica DMLB microscope equipped with a SPOT TM CCD camera (Diagnostic Instruments ADPase histochemistry, which labels endothelial cells as well as angioblasts and vascular buds, was also used to visualize developing vessels. This was performed using a modification of the method described by Lutty and McLeod (27). Images were taken under dark field using a Leica DFC 300 FX cooled-CCD color camera attached to a Leica Upright DMRXE.
Plasmid Constructions-The mouse laminin ␣1 LN (from amino acid 25 to 274, accession number ENSMUSP00000043957 in the Ensembl data base) was amplified in 30 cycles by PCR using the Platinum Taq polymerase High Fidelity, according to the manufacturer's instruction (Invitrogen), with the primers Lama1 Nt up HindIII (5Ј-AAGCTTGCAGAGAGGCTTGTTCCCTGCC-3Ј) and Lama1 Nt down BamHI (5Ј-GGATCCCCGCCTCCAACG-GAAATGTCTTT-3Ј) using 0.1 ng of plasmid pCIS mouse laminin ␣1 (a kind gift from Dr. Peter Yurchenco, Piscataway, NJ). The mouse laminin ␤1 LN (from amino acid 69 to 316, accession number ENSMUSG00000002900; primers Lamb1 LN up HindIII (5Ј-AAGCTTGCAGGAACCGGAGTTCAGCTAT-3Ј) and Lamb1 LN down BamHI (5Ј-GGATCCCCCCCTCGAACCACCATAT-CATA-3Ј) and the ␥1 LN (from amino acid 42 to 280, accession number ENSMUSG00000026478; primers Lamc1 LN up HindIII (5Ј-AAGCTTGGGGCGGCCGCAGCGCTGCATG-3Ј) and Lamc1 LN down BamHI (5Ј-GGATCCCCGCCCACAGCAAA-GTCTGAGAT03Ј) were amplified by PCR using the same procedure from mouse colon cDNA. The different PCR fragments were cloned in the pCRII-TOPO vector, according to the manufacturer's instruction (Invitrogen), to generate the plasmids pCRII-TOPO Lama1, Lamb1, or Lamc1 LN. The GeneEditor in vitro site-directed mutagenesis system (Promega) with the primer Lama1NT LNnmf223 (5Ј-AGACCTTGACCCCATTGTCACG-CGTCGTTATTGCTATTCGATAAAAGACATTTCC-3Ј) was used to introduce a point mutation changing the tyrosine at position 265 to a cysteine (in order to generate the nmf223 mutation), according to the manufacturer's instructions to generate the plasmid containing the pCRII-TOPO Lama1 LNnmf223 domain. The different plasmids were then digested with HindIII ϩ BamHI restriction enzymes. Inserts were gel-purified using the Ultra agarose spin kit according to the manufacturer's instructions (ABgene, Courtaboeuf, France) and subcloned in gel-purified vector pUT18 or pKNT25 from the bacterial double hybrid (BACTH) system kit (Euromedex) that were restriction-digested with Hin-dIII ϩ BamHI to generate constructs pUT18 Lama1 LN, Lama1 LNnmf223, Lamb1 LN, or Lamc1 LN and pKNT25 Lama1 LN, Lama1 LNnmf223, Lamb1 LN, or Lamc1 LN fragments. All constructs were verified by sequencing at the IGBMC (Illkirch, France) sequencing facility.
BACTH and ␤-Galactosidase Activity Measurements-The BACTH system kit (28) was chosen because it permits an evaluation of the strength between two interacting proteins. The system is based on the interaction-mediated reconstitution of the adenylate cyclase activity in Escherichia coli, which allows cyclic AMP to bind with the catabolite activator protein. Given that the cAMP-catabolite activator protein complex is a pleiotropic regulator of gene transcription in E. coli, its formation turns on the expression of several resident genes, including genes of the lac operon involved in lactose catabolism. The efficiency of complementation between the two-hybrid proteins can then be quantified by assaying the ␤-galactosidase enzymatic activities in bacterial extracts, which is correlated with cAMP levels produced in the bacteria. 0.5 ng of different combinations of pUT18 plus pKNT25derived constructs were electroporated into the bacterial strain BTH101 (FϪ, cya-99, araD139, galE15, galK16, rpsL1, hsdR2, mcrA1, mcrB1) and selected on a Luria-Bertani (LB) medium agar plate containing ampicillin (100 g/ml), kanamycin (50 g/ml), in the presence of 50 l of 2% 5-bromo-4-chloro-3indolyl-␤-D-galactopyranoside (X-gal) and 10 l of 100 mM isopropyl ␤-D-thiogalactopyranoside, at 30°C for 48 -72 h.
Blue colonies (white for the empty vectors and the combination of ␤1 LN plus ␤1 LN or ␥1 LN plus ␥1 LN) were cultured in 3 ml of LB medium containing ampicillin (100 g/ml), kanamycin (50 g/ml), and 0.5 mM isopropyl ␤-D-thiogalactopyranoside overnight at 30°C. The different samples were then diluted to obtain an absorbance at 600 nm of 0.6. After centrifugation at 4000 ϫ g for 15 min at 4°C, the pellets were resuspended in 50 l of lysis buffer from the ␤-galactosidase assay kit (Invitrogen), frozen on dry ice, and thawed at 37°C three times in order to rupture bacterial membranes. After removing insoluble materials by centrifugation at 4000 ϫ g for 10 min at 4°C, the amount of protein present in the supernatant was measured using the protein assay kit (Bio-Rad). ␤-Galactosidase activity was then measured using the ␤-galactosidase assay kit according to the manufacturer's instructions (Invitrogen) and calculated as the nmol of orthonitrophenyl-␤-D-galactopyrannoside hydrolyzed/mg of protein. Four independent electoporation experiments with at least three clones per condition were analyzed. The results are presented as the mean -fold change of the Lama1 LNnmf223 domain compared with the wild type Lama1 LN.
The results were analyzed using GraphPad Prism version 5.0. The Shapiro-Wilk normality test was used to confirm the normality of the data, the difference in variance was analyzed using the F-test, and the statistical difference of the mean was analyzed using Student's unpaired two-tailed t test with Welch's correction in case of unequal variances. The one-way analysis of variance Kruskal-Wallis test following by Dunn's multiple comparison post-test was used for multiple data comparison.

nmf223 Mutants Exhibit Vitreal Fibroplasia and Vascular
Abnormalities-Indirect ophthalmoscopy of nmf223 homozygotes revealed abnormalities, including small white retinal spots, vitreal fibroplasia (which presents as cobweb-like material covering the retina), and vessel tortuosity, that were not seen in WT littermates (Fig. 1, A and B). Fluorescein angiography further demonstrated the vascular tortuosity and also revealed hypervascularization of the retina in nmf223 mutants ( Fig. 1, C and D). These features were observed as early as 3 weeks of age, the earliest time point investigated, and remained unchanged in mice as old as 2 years, indicating that the clinical phenotype of nmf223 homozygotes is stable.
The nmf223 Phenotype Is Caused by a Recessive Mutation in Lama1-When mice bearing the nmf223 phenotype were outcrossed to B6 animals, the resulting F1 mice were phenotypically normal. When F1 progeny were intercrossed ϳ25% of F2 progeny were affected, consistent with a single recessive mutation.
The nmf223 locus was mapped to mouse chromosome 17, between microsatellite markers D17Mit89 and D17Mit160. Identification of critical genetic recombinants and progeny testing of potentially informative, unaffected animals refined the minimal genomic region to 2.6 Mb between single nucleotide polymorphisms RS13483075 and RS33491771 (Fig. 2, A and B). Sequence analysis of Lama1, contained in the critical region, revealed an A to G missense mutation at nucleotide 794 that is predicted to change amino acid 265 from a hydrophilic tyrosine to a hydrophobic cysteine residue (Y265C). This mutation was confirmed by comparing sequence from genomic DNA in affected, control, and F1 animals (Fig. 2C). Both the original nucleotide and the amino acid it encodes are highly conserved across all vertebrate species for which Lama1 sequence is available. The affected amino acid lies in the end of the N-terminal domain (LN) of Lama1 (Fig. 2E). Interestingly, this central Tyr in the YYY motif is among few amino acids conserved in all laminin subunits containing an LN (Fig. 2D).
In order to verify that Lama1 carries the mutation responsible for the nmf223 phenotype, homozygous nmf223 mice were crossed to heterozygous Lama1 ⌬ mice (10). Retinas from nmf223 heterozygotes not carrying the Lama1 ⌬ allele were identical to those of WT mice when viewed by indirect ophthalmoscopy (Fig.  3A). The phenotype of Lama1 nmf223/⌬ compound heterozygotes, however, was comparable with nmf223 mutants with white spots, vitreal fibroplasia, and vessel tortuosity (Fig. 3B), confirming that nmf223 is a mutant allele of Lama1. The nmf223 allele will now be designated as Lama1 nmf223 .
The Localization of Laminin ␣1 and Other Inner Limiting Membrane Components Is Unaltered in the Retinas of Lama1 nmf223 Homozygotes-In order to determine whether the Y265C mutation disrupted the localization of laminin ␣1 in the retina, immunohistochemical studies were performed using both a monoclonal antibody 200 (21, 29) and a polyclonal anti-  body 1057ϩ (directed against the LN). Both anti-laminin ␣1 antibodies primarily stained the inner limiting membrane of the retina (Fig. 4A) with some labeling surrounding the optic nerve in control mice. Staining of retinal sections from homozygous Lama1 nmf223 mutants revealed the mutant laminin ␣1 localized correctly to the inner limiting membrane (Fig. 4B).
Because Lama1 is an integral part of the inner limiting membrane, we sought to determine the effects of the Y265C mutation in laminin ␣1 on this structure. The localization of other inner limiting membrane components, including collagen IV, perlecan, and dystroglycan, were similar in control and Lama1 nmf223 homozygotes (Fig. 4, C-H). In addition, the localization of the laminin ␣1 receptor, integrin ␤1, was unaltered in Lama1 nmf223 homozygotes (Fig. 4, I and J) as was the localization of integrin subunit ␣1 (data not shown). To determine if the Müller cell end-feet, which normally attach to the inner limiting membrane, were affected in Lama1 nmf223 homozygotes, retinal sections were stained with an antibody against glutamine synthetase. Both at 4 weeks (Fig. 4, K and L) and at 6 months (data not shown), Müller cell processes in Lama1 nmf223 homozygous retinas resembled those of age-matched WT retinas, extending through the entire length of the retina. Together, these data suggest that the Y265C point mutation does not affect the inner limiting membrane assembly.
Staining of retinal sections with H/E, however, demonstrated the presence of ectopic cells and vessels within the vitreous of the mutant retina, indicating the possibility of a disruption of the inner limiting membrane (Fig. 4, M and N). Indeed, ultrastructural examination of the inner limiting membrane with electron microscopy revealed structural abnormalities. At postnatal day 2 (P2), the inner limiting membrane was normal in some areas but absent or severely attenuated in others. In addition, although Müller cell end-feet terminated at the inner limiting membrane in some cases, in others they protruded into the vitreous (Fig. 4, O and P). In addition to aberrant blood vessels (Fig. 4N), ectopic cells that appeared morphologically to be astrocytes and retinal ganglion cells were also found in the vitreous (data not shown). Similar inner limiting membrane abnormalities were noted at 4 weeks of age (data not shown).

Peripheral Inner Nuclear Layer Thinning and an Associated Reduction in ERG Response Are Observed in Lama1 nmf223
Homozygotes-The central retinas of Lama1 nmf223 homozygotes remained comparable with those of WT mice as late as 18 months with the exception of sporadic areas of dysplasia, which are likely to correspond to the small white spots observed by indirect ophthalmoscopy. In contrast, between 8 and 12 weeks of age, a reduction in the number of cells in the peripheral inner nuclear layer of Lama1 nmf223 homozygous retinas was observed ( Fig. 5). This loss progressed slowly and stopped after 1 year of age.
Immunohistochemical studies of retinal sections were used to compare the relative number of amacrine, horizontal, and bipolar cells in the inner nuclear layer of 9-month-old Lama1 nmf223 homozygous and WT controls. Labeling with antibodies against calbindin (horizontal cells), VSX2 (formerly CHX10; bipolar cells), paired box 6 (amacrine and horizontal cells), and TH1 (dopaminergic amacrine cells) did not reveal a marked reduction of any one particular cell type (data not shown). At 12 weeks of age, there was also a variable loss of cells in the retinal ganglion cell layer in ϳ20% of Lama1 nmf223 homozygotes. The number of mutants experiencing this loss increased to 80% at 9 months of age (Fig. 5). As with inner nuclear layer thinning, this cell loss affected only the peripheral retina. The peripheral cell loss was not associated with a reduction in vascularization.
ERG results obtained from 1-year-old mice are also summarized in Fig. 5. Fig. 5E depicts representative responses obtained from WT littermates (left) and Lama1 nmf223 homozygotes (right) to strobe stimuli presented to the dark-adapted eye. Lama1 nmf223 homozygotes had ERG waveforms that were comparable with those of WT animals but of reduced amplitude. Fig. 5F plots intensity-response functions for the major ERG components obtained under dark-adapted conditions. The aand b-wave components of Lama1 nmf223 homozygotes were reduced below those of WT animals by a fraction that was consistent across flash intensities. Fig. 5G depicts representative responses obtained from WT (left) and Lama1 nmf223 homozygotes to strobe stimuli superimposed upon a steady adapting field. Although light-adapted ERGs of Lama1 nmf223 homozygotes had waveforms that were comparable with those of WT animals, the overall Lama1 nmf223 mutant response was reduced. Fig. 5H shows that the reduction in cone ERG amplitude of Lama1 nmf223 homozygotes was consistent across flash intensity. Similar ERG results were obtained for 6.5-and 8.5month old Lama1 nmf223 homozygotes (data not shown).
Persistent Hyaloid Remnants Are the Predominant Vessel Type in Lama1 nmf223 Homozygotes-In WT mice, a honeycomb pattern of GFAP-positive astrocytes was formed between P0 and P1 and extended across the developing retina. G. simplicifolia isolectin highlighted the emergence of vessels from the optic disc, which followed the GFAP-labeled astrocytic template (Fig. 6A). During sample processing, the hyaloid vasculature readily detached from the retina of WT mice and was observed to contain many macrophages. At P7, when the primary plexus was near completion, retinal blood vessels were found in close association with the fully formed retinal astrocyte template (Fig. 6C). Retinal capillaries were observable with both G. simplicifolia isolectin and ADPase staining (Fig. 6, C and E). Concomitantly, staining with anti-endostatin and G. simplicifolia isolectin demonstrated that the hyaloid vasculature had begun to regress (data not shown). This regression continued so that by P10 only the primary hyaloid artery remained in the vitreous, as shown by anti-endostatin and G. simplicifolia isolectin labeling (Fig. 6G). At P21, the intermediate plexus was fully developed, completing retinal vascularization (Fig. 6I).
The Lama1 nmf223 mutation significantly altered the development of astrocytes and retinal vessels. Between P0 and P1, GFAP-positive astrocytes were condensed around the optic disc, with only a few processes extending to the periphery in mutant retinas. Many of the astrocytes appeared to lie within the vitreous rather than the retina itself and were associated with hyaloid vessels (Fig. 6B). Retinal blood vessels, when present, were observed only immediately adjacent to the optic disc and were reduced in number compared with those seen in controls. In addition, fewer G. simplicifolia isolectin-positive cells were observed throughout the retina. At P7, the majority of blood vessels were observed in the vitreous rather than the retina. As with younger retinas, double staining with anti-GFAP and G. simplicifolia isolectin demonstrated that mutant astrocytes associated with vitreal vessels, wrapping around them as they would mature retinal vessels (Fig. 6D). In addition, GFAP staining was more intense than that seen in WT retinas, with astrocytes appearing hyperstellated. Removal of the vitreous and hyaloid vessels revealed G. simplicifolia isolectin-positive cells in the underlying retina, particularly in the periphery (data not shown). These cells were associated with retinal astrocytes that formed a less dense version of the WT astrocyte template. A majority of the vessels observed in Lama1 nmf223 homozygous retinas were endostatin-positive, indicating their hyaloid vessel origins (data not shown). ADPase staining of mutant retinas demonstrated that some hyaloid vessels, rather than regressing, were beginning to bud at this stage (Fig. 6F).
At P10, immunohistochemical labeling demonstrated that most of the larger vessels in Lama1 nmf223 homozygous retinas were endostatin-positive and, therefore, of hyaloid origin. At this stage, endostatin-positive blood vessels could be seen branching into the deeper retina, where endostatin expression was absent (Fig. 6H). The larger, vitreal vessels remained endostatin-positive throughout adulthood (data not shown). ADPase staining at P21 confirmed that the larger, primary plexus-like vessels were found within the vitreous rather than the retina (Fig. 6J). This stain also demonstrated that the secondary and intermediate plexi do form in mutant retinas despite the lack of a normal retinal primary plexus. The blood vessels within these deeper layers, however, were disorganized and more abundant with a higher density of capillary branch points than in control retinas. GFAP staining of P21 retinas was identical to that observed at earlier time points with the exception that focal areas completely void of astrocytes and blood vessels were observed (data not shown).
The Y265C Mutation Significantly Affects Laminin ␣1 LN Domain Interactions in Vitro-The BACTH system (28) was used to test the hypothesis that the Y265C mutation leads to a perturbation of laminin chain interactions. The plasmids coding for T18 and T25 fused to the Lama1 LN or LN nmf223 , Lamb1 LN, and Lamc1 LN were introduced into E. coli BTH101, an adenylate cyclase-deficient bacterial strain. ␤-Galactosidase activity was used to evaluate the strength of interaction between two interacting proteins. The BACTH system was first validated by testing interactions between the different combi- ADPase staining of P7 retinas shows a complete primary plexus as well as the formation of capillaries in WT retinas (E), whereas in the Lama1 nmf223 homozygous retina, budding of the vitreal vessels is observed, but no capillaries are present (F). The arrows indicate the hyaloid remnants in WT mice and budding of hyaloid vessels in Lama1 nmf223 homozygotes, whereas the asterisks indicate areas devoid of retinal vessels in Lama1 nmf223 homozygotes. Labeling with anti-endostatin and G. simplicifolia isolectin at P10 demonstrates the failure of hyaloid vessels to regress in Lama1 nmf223 homozygotes (H) compared with controls (G). Only endostatin-positive primary-like vessels are observed in nmf223 retinas. The arrow in G indicates the hyaloid artery in controls, whereas circles in H highlight anastomoses of hyaloid vessels entering the retina. Rather than within the retina, as seen in controls (I), in adult retinas, ADPase staining confirms that primary-like vessels lie in the vitreous of Lama1 nmf223 homozygotes (J).
nations of wild type LNs. Interactions were observed between ␣1 LN and ␣1 LN and between ␤1 LN and ␥1 LN, as well as between ␣1 LN and ␤1 LN and between ␣1 LN and ␥1 LN but not between ␤1 LN and ␤1 LN or between ␥1 LN and ␥1 LN, which had levels of ␤-galactosidase activity similar to that observed with transfectants harboring the empty vector (Fig.  7A). The one-way analysis of variance test indicated that both the ␣1 LN plus ␣1 LN and ␤1 LN plus ␥1 LN combinations were statistically different (p Ͻ 0.001) compared with the ␤1 LN plus ␤1 LN, ␥1 LN plus ␥1 LN, or empty vectors. These data are in agreement with that described in the literature (30,31). The effect of the laminin ␣1 LN nmf223 mutation on interactions between ␣1, ␤1, and ␥1 LN was then investigated. In comparison with wild type, the presence of the ␣1 LN nmf223 mutation significantly decreased the affinity of Lama1 LN with all binding partners (p Ͻ 0.001) (Fig. 7, B-D). Taking into account the basal ␤-galactosidase activity observed in E. coli BTH101 transfected with empty vectors, the ␣1 LN nmf223 mutation leads to a reduction of ϳ50% with the wild type ␣1 LN and ␤1 LN and 40% with the ␥1 LN.

Deletion of Lama1 in the Epiblast by Activation of the Lama1 Floxed Allele Using Tg(Sox2-cre)1Amc
Allows for the Generation of Viable Lama1 Null Mice-In order to bypass the early embryonic lethality observed previously in null Lama1 mice (10, 11), Lama1 ⌬ heterozygotes were crossed to Tg(Sox2cre)1Amc mice, which express cre under the control of a 12-kb Sox2 promoter region. This transgenic line has been reported to induce recombination in all epiblast cells by embryonic day 6.5 with no activity in extraembryonic cell types (17). Progeny from this cross, which carried both the deleted Lama1 allele and the Tg(Sox2-cre)1Amc transgene, were mated to homozygous Lama1 floxed mice. This cross generated viable Lama1 null, Lama1 ⌬/⌬ , Tg(Sox2-cre)1Amc mice. Here, we use the term Lama1 ⌬ homozygote to refer to mice that were also, by necessity, hemizygous for the Tg(Sox2-cre)1Amc transgene. The absence of the WT or Lama1 floxed allele in these mice was verified by performing an analysis of various tissues by PCR, indicating that these mice are not mosaic and should be considered as a true Lama1 knockout mouse strain (Fig. 8A). A number of tissues were examined by immunofluorescence using a specific antibody against laminin ␣1 (21). Laminin ␣1 staining was found in the basal lamina of both the cortical tubules of the kidney and the testis seminiferous epithelium of control mice, whereas no laminin ␣1 staining was observed in Lama1 ⌬ homozygotes (Fig. 8B). The absence of laminin ␣1 staining was also confirmed in the retina (Fig. 9), bladder, mammary gland, intestine, and ovary of Lama1 ⌬ homozygotes (data not shown). The only physical abnormality observed in Lama1 ⌬ homozygotes was a 13% reduction in birth weight when compared with control mice (p Ͻ 0.05, Student's t test, n ϭ 14). By 8 weeks of age, however, this weight difference was no longer observed. Both male and female Lama1 ⌬ homozygotes had normal life spans and fertility.
The Inner Limiting Membrane Is Absent in Lama1 ⌬ Homozygotes-Immunohistochemical staining for inner limiting membrane components confirmed a lack of laminin ␣1 in Lama1 ⌬ homozygote retinas (Fig. 9, A and B). The linear inner limiting membrane staining for collagen IV and perlecan observed in heterozygous mice (Fig. 9, C and E) and  MARCH 5, 2010 • VOLUME 285 • NUMBER 10

JOURNAL OF BIOLOGICAL CHEMISTRY 7705
Lama1 nmf223 mutants (Fig. 4) was punctate in Lama1 ⌬ homozygotes (Fig. 9, D and F). In addition to being punctated, dystroglycan staining was reduced in Lama1 ⌬ homozygotes (Fig. 9H) compared with littermate controls (Fig. 9G). Together, these observations suggest that the inner limiting membrane does not form properly in Lama1 ⌬ homozygotes. Histological assessment of Lama1 ⌬ homozygotes supported this observation, with ectopic cells and blood vessels in the vitreous (Fig. 9, I and J). Labeling with glutamine synthetase revealed many Müller cells lacking end-feet with misguided and shortened processes (Fig. 9, K and L).
Lama1 ⌬ Homozygotes Have a Similar but More Severe Retinal Phenotype than That Observed in Lama1 nmf223 Homozygotes-Fundus examination by indirect ophthalmoscopy revealed that Lama1 ⌬ homozygotes also exhibit vitreal fibroplasia, vessel tortuosity, and white retinal spots (Fig. 10, A  and B). Compared with Lama1 nmf223 homozygotes, the retinal spots were larger and more prominent, resembling patches of retinal degeneration that progressed with age. At 1 year of age, eye diameters of Lama1 ⌬ homozygotes were 11% larger than those of control mice (n ϭ 10 control and 13 mutant mice, p Ͻ 0.001, Student's t test), a phenotype not seen in Lama1 nmf223 mutants (data not shown). ADPase staining of retinal whole mounts demonstrated larger vessels lying within the vitreous with an apparent lack of primary plexus veins and arteries within the retina (Fig. 10, C and D). The aberrant vitreal vessels covered the entire retinal surface and were similar in number to those seen in Lama1 nmf223 homozygotes. As with Lama1 nmf223 homozygotes, the vitreal vessels remained endostatinpositive throughout adulthood, suggesting hyaloid origin. These vessels could be seen branching into the retina as well (data not shown). Although present, the intermediate and deep vascular plexi of the retina did not appear to be as dense as those in Lama1 nmf223 homozygous retinas. Furthermore, the deeper retinal vasculature of Lama1 ⌬ homozygotes penetrated the periphery in some but not all quadrants. Co-staining of Lama1 ⌬ homozygous whole mount retinas with anti-GFAP and G. simplicifolia isolectin revealed astrocytes associating with vitreal vessels, a phenotype similar to that seen in Lama1 nmf223 homozygotes (Fig. 10,  E and F).
Histologically, the central retina of Lama1 ⌬ homozygotes was similar to that of controls at both 4 weeks and 6 months of age with the exception of the observed inner limiting membrane defects (Fig. 11, A-C). As early as 4 weeks of age, however, Lama1 ⌬ homozygotes showed peripheral loss of cells similar to but more severe than that seen in Lama1 nmf223 homozygotes, affecting both nuclear layers of the retina and the inner plexiform layer (Fig. 11, D and E). In addition to an earlier onset, this cell loss extended further into the central retina, unlike Lama1 nmf223 mutants, which primarily affected cells of the peripheral inner nuclear layer. Examination of 6-month-old retinal sections showed that this cell loss was progressive, particularly in the outer nuclear layer (Fig. 11F). Cell loss from the retinal ganglion cell layer of Lama1 ⌬ homozygotes was also more severe than that observed in Lama1 nmf223 homozygotes and was readily apparent by 4 weeks of age, the earliest time point examined. Moreover, the cells in this layer were disorganized, with some areas having extra cells whereas others were thinned with many ectopic cells on the vitreal surface of the retina. Consistent with a more severe histological phenotype, reductions in ERG amplitude were FIGURE 8. Activation of the Lama1 floxed allele with Tg(Sox2-cre)1Amc-transgene generates a Lama1 null mouse. A, genotype analysis of various organs of heterozygous and homozygous Lama1 ⌬ mice. Lama1 alleles were detected by PCR using primer pairs Ln127/Ln135, Ln126/Ln137LPN, and Ln123/Ln127. Tg(Sox2cre)1Amc-transgene was detected using primers AFX2/AFX3. Genomic DNA from a Lama1 floxed/ϩ mouse was used as a control for the floxed allele amplification. Only the deleted allele can be detected in homozygous Lama1 ⌬ mouse tissues. B, immunofluorescence analysis of kidney and testis from a Lama1 ⌬ mutant (right) and a control mouse (left) using anti-laminin ␣1 (green) and anti-collagen IV (red) antibodies. Merged images of laminin ␣1 and anti-collagen IV images together with nuclear 4Ј,6-diamidino-2-phenylindole staining (blue) are shown on the right. Laminin ␣1 staining is found in the basal lamina of both the cortical kidney tubules and the testis seminiferous epithelium of control mice. No laminin ␣1 staining is observed in Lama1 ⌬ mutants. more pronounced in Lama1 ⌬ homozygotes (Fig. 11, G and H) than in Lama1 nmf223 homozygotes (Fig. 5, E-H). Optomotor testing showed that Lama1 ⌬ homozygotes were behaviorally blind at 1 year (supplemental Fig. 1).

DISCUSSION
The present study describes the first mammalian hypomorphic and null alleles of Lama1 that survive into adulthood. The current work supports the role of Lama1 in ocular development previously suggested by zebrafish studies (12,15,16) because both Lama1 nmf223 and Lama1 ⌬ homozygotes develop ocular defects.
Lama1 mutations in zebrafish have been reported to disrupt inner limiting membrane formation in zebrafish (12). Although inner limiting membrane disruptions were observed in Lama1 mutant mice, the severity and consequences of these differ from the zebrafish models and from one another. In Lama1 nmf223 homozygotes, the presence of the primary inner limiting membrane proteins and electron microscopic analysis demonstrates that this structure is formed. However, electron microscopy revealed focal disruptions with cells and vessels protruding into the vitreous. Thus, in Lama1 nmf223 homozygotes, the inner limiting membrane forms, but its integrity is reduced by the Y265C mutation. By contrast, histological assessment of Lama1 ⌬ FIGURE 9. The inner limiting membrane does not form in homozygous Lama1 ⌬ mutants. Shown is immunofluorescence analysis of 4-week-old eyes from littermate controls (A, C, E, and G) and Lama1 ⌬ homozygotes (B, D, F, and H) with anti-laminin ␣1, anti-collagen IV, anti-perlecan, and anti-dystroglycan antibodies. Histological examination with H/E staining demonstrates that the well organized ganglion cell layer and intact inner limiting membrane seen in controls (I) are replaced by a disorderly ganglion cell layer with ectopic cells and vessels within the vitreous in Lama1 ⌬ homozygotes (J). Glutamine synthetase staining shows that, in the control retinas, Mü ller cell end-feet lie at the inner limiting membrane (K). In the Lama1 ⌬ homozygous retinas, however, many Mü ller cell processes fall short, terminating in the inner plexiform layer, and are lacking end-feet (L). Ectopic cells (arrows) are also visible on the vitreal side of the retina in Lama1 ⌬ homozygous retinas. homozygotes revealed an apparent panretinal loss of the inner limiting membrane with numerous ectopic cells and blood vessels within the vitreous. Indeed, three primary inner limiting membrane components, collagen IV, perlecan, and laminin ␣5 (data not shown), were expressed in the Lama1 ⌬ homozygous retina but in a punctate manner unlike the continuous linear distribution seen in controls or Lama1 nmf223 homozygotes. In addition, the intensity of dystroglycan was reduced in the retina of Lama1 ⌬ mutants. Together, these observations suggest that the inner limiting membrane does not form in Lama1 ⌬ mutants.
Although the absence of the inner limiting membrane in the Lama1 ⌬ mutants is not surprising because laminin-111 is a prerequisite for basal membrane formation (10,32,33), the abnormalities in the Lama1 nmf223 mutants are more difficult to explain. Given that cysteines are important for disulfide bridge formation (34,35), the addition of this residue may perturb the three-dimensional structure of laminin ␣1 and subsequent interactions with other laminin chains and receptors. In order to investigate the effects of the Y265C mutation on laminin ␣1 function, bacterial two-hybrid studies were undertaken. The use of this novel method for studying laminin chain interactions was verified by the presence of interactions between ␣1 LN and ␣1 LN and between ␤1 LN-␥1 LN as well as between ␣1 LN and ␤1 LN and between ␣1 LN ␥1 LN but not between ␤1 LN and ␤1 LN or ␥1 LN and ␥1 LN, which is consistent with findings reported in the literature (30,31). All interactions with the ␣1 chain were markedly reduced in strength by introduction of the nmf223 mutation Y265C.
The Lama1 LN has been shown previously to self-interact in vitro with high affinity in the presence of Ca 2ϩ (30) and in the absence of Ca 2ϩ (31,36) to form dimers. Recently Yurchenco and co-workers (37) demonstrated that laminin ␣1 as well as the ␤1 and ␥1 LNs are essential for basement membrane formation. The LN self-interaction may facilitate the polymerization of the different laminin chains. Our BACTH results demonstrate that the tyrosine 265 is important for such interactions. We hypothesize that the nmf223 mutation in Lama1, Y265C, could lead to an abnormal mesh formation in the basement membrane by decreasing the number of ␣1 LN binding partners in vivo, despite the presence of interactions between ␤1 LN and ␥1 LN. In addition, the ability of laminin 111 to act as a chemoattractant for Müller cells (38) may be hindered by the Y265C mutation.
The Y265C mutation may also disrupt other important laminin ␣1 interactions. For example, the binding of laminin ␣1 to receptors on Müller cells may be weakened. Indeed, the LN contains receptor binding sites for integrins ␣1␤1 and ␣2␤1 as well as perlecan (2,39). These changes may also contribute to the decreased integrity of the inner limiting membrane and the abnormal presence of cells and vessels within the vitreous. Support for the hypothesis that binding to the Müller cell receptors is disrupted comes from observations that a similar although more severe phenotype is observed in mice with mutations in Large, which encodes a glycosyltransferase required for the proper glycosylation of ␣-dystroglycan (19). Interactions of other proteins, known to contain an LN, such as netrins (40) may also be disrupted by the Y265C mutation. Although the specific protein interactions disrupted are yet to be elucidated, the current study demonstrates that the tyrosine 265 is crucial for normal LN function. The conservation of this central Tyr in the YYY motif in all laminins containing an LN further supports the importance of the mutated residue.
It is interesting to note that other animal models with point mutations in the LN of laminin ␣ chains exist. This is the case of the zebrafish mutant bal a69 that has a mutation C56S in the lama1 gene, leading to several ocular defects (15). The Lama2 nmf417 carries a C79R mutation in the LN of the mouse Lama2 gene (encoding the laminin ␣2 chain) that causes congenital muscular dystrophy (41). Although basement membranes appear generally normal at the electron microscopy level in Lama2 nmf417 mice, immature Schwann cells fail to assemble this structure and underscore the importance of the LN in this process.
The inner limiting membrane abnormalities noted in lama1deficient zebrafish were associated with retinal ganglion cell layer defects (12,15) and axonal abnormalities (13,14). Similarly, cells in this layer were affected in both Lama1 mouse mutants but not in the same manner. Although retinal ganglion cell development was normal in Lama1 nmf223 homozygotes, a progressive cell loss was observed within this layer in some mice beginning at 12 weeks of age. In Lama1 ⌬ homozygotes, the retinal ganglion cell layer was thickened in some areas and attenuated in others as early as 4 weeks of age, the earliest time point investigated, suggesting possible developmental defects. In neither case was the ganglion cell loss associated with a change in intraocular pressure (data not shown). The ganglion cell layer alterations in lama1-deficient zebrafish were attributed to abnormal inner limiting membrane formation and subsequent aberrant development of Müller cells (12). At the light microscopic level, Müller cell processes of Lama1 nmf223 homozygotes were similar to those of WT mice, terminating with end-feet just below the inner limiting membrane. However, in the Lama1 ⌬ homozygous retina, many Müller cell processes were truncated and lacked end-feet. These abnormalities suggest that laminin ␣1 is important in path searching of Müller glial cells in vivo as has already been demonstrated for laminin-111 in vitro (38). The differences observed between these mouse models in regard to retinal Müller cells and ganglion cells are likely to reflect allelic differences and their effects on laminin ␣1 protein structure and function as well as subsequent protein interactions.
The ERG phenotype observed in Lama1 nmf223 homozygotes is consistent with the modest loss of retinal neurons in these mice. Taking the amplitude reduction into account, the overall waveform was comparable in Lama1 nmf223 homozygotes and WT mice, indicating that the cells lost in the inner nuclear layer were not preferentially bipolar cells, which would present as a b-wave reduction (42). Indeed, staining for VSX2, a bipolar cell marker, showed no marked reductions in Lama1 nmf223 homozygotes. A recent report showed that mice lacking Lamb2 experience disruption in the development of retinal dopaminergic neurons, which are identified by TH1 (43). No difference was observed in TH1 expression in Lama1 nmf223 homozygotes compared with WT mice. Similarly, no other cell type was markedly reduced in mutants, suggesting subtle losses of all inner nuclear layer cell types.
Lama1 ⌬ homozygotes experienced a more pronounced reduction in ERG responses compared with Lama1 nmf223 homozygotes. This was associated with an increase in the severity of retinal degeneration as well as an earlier onset. Cell loss from the inner nuclear layer and ganglion cell layer was observed as early as 4 weeks of age, the earliest time point examined, compared with 12 weeks in Lama1 nmf223 homozygotes. By 6 months of age, the outer nuclear layer and inner plexiform layers were also thinner in Lama1 ⌬ homozygotes. The increased severity of the retinal phenotype observed in Lama1 ⌬ homozygotes suggests that laminin ␣1 and perhaps the inner limiting membrane itself are required for the survival of retinal neurons. This supports previous work in chick retinas implicating the retinal ganglion cell dependence on the inner limiting membrane (44).
Although the zebrafish models die prior to retinal vascular development, they have provided valuable insights into lama1 function in vessel formation and maintenance. In lama1 zebrafish morphants, endothelial cell differentiation is delayed, and capillary formation and blood flow in the hyaloid vasculature are reduced (12,15). Similarly, endothelial cell development was attenuated or at least delayed in Lama1 nmf223 homozygous retinas. Most large vessels in Lama1 nmf223 homozygotes remained endostatin-positive into adulthood, suggesting hyaloid origins (26). Anastomoses from these vessels appeared to enter the retina and form the intermediate and deep retinal vascular plexi, which were comparable with those of WT mice, although less organized with potentially more capillaries. It is possible that the increased number of capillaries in Lama1 nmf223 homozygotes is caused by errors in remodeling of these structures. In adult Lama1 ⌬ homozygous retinas, endostatin-positive hyaloid vessels also appeared to penetrate into the retina, suggesting that a similar compensatory development of retinal vessels occurs in these mice. Interestingly, although the hyaloid vessels of the homozygous Lama1 ⌬ extended over the entire vitreal-retinal surface, the deeper retinal vasculature was unevenly distributed, with some areas containing fewer peripheral blood vessels compared with both littermate controls and Lama1 nmf223 homozygotes. Additional studies are under way to investigate the vascular differences between the Lama1 ⌬ and Lama1 nmf223 mutant retinas. Interestingly, the formation of retinal vessels from branching hyaloid vessels has also been reported in Collagen type XVIII ␣1-targeted mutant mice (25,45). The similarities between the vascular development of Lama1 nmf223 homozygotes and Collagen XVIII mutants suggest that Lama1, like Collagen XVIII, is required for the regression of the hyaloid vessels (25,45). It has been demonstrated in vitro that the C-terminal endostatin domain of collagen XVIII inhibits endothelial cell proliferation (46) and migration (47). Perhaps more importantly with regard to hyaloid regression, endostatin has been shown to induce endothelial cell apoptosis (48). Moreover, endostatin has been shown to bind to the LN of all laminin-111 chains, and the interaction is crucial for endostatin function (46,49,50). Given that the Y265C mutation significantly reduced LN-LN interactions in BACTH studies, it is worth considering that this point mutation may disrupt the binding of laminin-111 to the endostatin domain of collagen XVIII. In doing so, the Y265C mutation would inhibit the apoptotic function of endostatin on hyaloid vessels. Further studies are required to investigate this proposed mechanism.
The Lama1 mutants described herein also share a phenotype described in mice lacking both Collagen XVIII and Collagen XV (45), an abnormal migration of astrocytes into the vitreous. The ectopic astrocytes associate with hyaloid vessels in the vitreous, potentially stabilizing them and preventing their normal regression. These observations suggest that an intact, fully functional inner limiting membrane is required to retain astrocytes within the retina. Alternatively, laminin ␣1 may provide signals that guide astrocyte migration. Interestingly, some astrocytes from Lama1 nmf223 homozygous mutants do enter the retina, producing a honeycomb-like pattern similar to but far less dense than that seen in control retinas. Why this subset of astrocytes is able to migrate normally whereas most astrocytes enter into the vitreous has yet to be determined. What is clear is that these changes in astrocytes, which are required to guide normal endothelial cell differentiation and migration, hinder formation of the retinal primary plexus in Lama1 mutants.
It is interesting to note that some disease features observed in zebrafish morphants with reduced lama1 expression were not encountered in either Lama1 mouse model described here. These include micropthalmia, cataracts, and lens degeneration (15). The anterior segment differences between zebrafish and mice may be due to allelic or species variation. In addition, mammals may form laminin heterotrimers not found in zebrafish that compensate for the loss of functional ␣1.
The present study demonstrates the importance of laminin ␣1 in the development of the mouse retina, particularly with regard to angiogenesis. The following model may explain the phenotypes observed. In Lama1 nmf223 homozygotes, the Y265C point mutation disrupts the binding of laminin ␣1 to one of its interacting partners on retinal Müller cells, whereas in the conditional null mouse described, the loss of Lama1 disrupts the entire assembly of the inner limiting membrane. In both cases, astrocytes migrate into the vitreous, where they associate with hyaloid vessels, preventing their regression. The lack of an astrocyte template affects endothelial cell differentiation and migration, disrupting the retinal primary vascular plexus formation and leaving the retina dependent on the hyaloid vasculature. Around P10, hypoxia stimulates the branching of hyaloid vessels into the retina, allowing the deep and intermediate vascular plexi to form. Although the cause of inner nuclear layer thinning in Lama1 nmf223 homozygotes is unknown, the uneven peripheral retinal vasculature in the Lama1 ⌬ homozygotes probably results in hypoxia and the observed cell loss.
The mouse models described herein suggest that point mutations in human LAMA1 could impact retinal development and function without altering survival. Based on the vitreal fibroplasia and abnormal vascular development observed in Lama1 mouse mutants, it is possible that mutations in LAMA1 could predispose individuals to retinal detachment. Indeed, retinal detachment was observed in six of eight (75%) Lama1 ⌬ homozygotes examined (data not shown). Given the similarities between the Lama1 and the Collagen XVIII mutants as well as the known interactions between laminins and endostatin, it is possible that LAMA1 could also contribute to Knobloch syndrome, in which patients experience persistent fetal vasculature. Mutations in Collagen XVIII have been identified in Knobloch syndrome but do not account for all cases of this condition (51,52). Furthermore, some mutations in Collagen XVIII leading to Knobloch syndrome also reduce affinity toward laminin (51,52). The increased eye diameter observed in Lama1 ⌬ homozygotes also suggests a possible link to myopia, which is associated with increased eye size (53). Interestingly, a locus for high grade myopia-2 maps to a region on human chromosome 18 that contains LAMA1 (54,55). Finally, mutations in LAMB2 have recently been shown to cause Pierson disease, in which patients experience microcoria (narrowing of the pupils) as well as kidney and central nervous system abnormalities clinically characterized by microcoria (fixed narrowing of the pupils) (56). Therefore, this is the second member of the lami-nin gene family to cause eye disease when mutated. Although additional studies will be required to completely define the function of Lama1 in retinal development, the availability of the two models described here will facilitate the discovery process.
In conclusion, this study emphasizes the importance of the inner limiting membrane in retinal development and maintenance and suggests that mutations in genes encoding other components of this structure or binding partners could have similar effects on the retina. These should, therefore, be considered candidate genes for human retinal diseases.