Severe Retinal Degeneration Associated with Disruption of Semaphorin 4A

Abstract

purpose. Semaphorin 4A (Sema4A) is a member of the transmembrane class 4 family of semaphorins. It has recently been shown to participate in cell–cell communication in the immune system. High levels of sema4A are also present in brain and eye, but its function in the central nervous system has not been studied. To investigate the function of Sema4A, we generated mice deficient in this transmembrane signaling molecule.

methods. An embryonic stem (ES) cell clone with a retroviral gene-trap insertion in the sema4A gene was used to generate mice lacking this transmembrane semaphorin. Fundus photography, fluorescein angiography, and electroretinography were used to evaluate retinal anatomy and physiology in mice lacking Sema4A. Electron microscopy and immunohistochemistry with cell-type–specific markers were used to characterize retinal development. In situ hybridization with sema4A-specific riboprobes was used to localize expression of this gene in the developing and adult eye.

results. Fundus photography performed at 14 weeks of age revealed severe retinal degeneration, attenuated retinal vessels, and depigmentation in mice lacking Sema4A. At this age, the outer nuclear layer was reduced to a single row of photoreceptor cells, and the outer plexiform layer was thin and disorganized. Disruption of Sema4A also compromised the physiological function of both rod and cone photoreceptors. Developmental studies in Sema4A-deficient mice revealed abnormal morphology of photoreceptor outer segments during the time at which they establish contacts with apical microvilli of the retinal pigment epithelium (RPE). Sema4A is expressed in the inner retina and RPE during the time at which photoreceptor outer segments elongate.

conclusions. These findings identify a previously unknown function of Sema4A in the developing visual system and provide a useful model for understanding cell–cell interactions that occur between photoreceptors and the RPE.

Development of multicellular organisms requires interactions between cells and their local environment and between neighboring cells. These interactions are often mediated by secreted, sometimes diffusible cues that affect adjacent populations of cells through association with membrane-bound proteins. A benchmark of the diversity associated with intercellular communication is the developing nervous system, in which a myriad of cells responds to a variety of cues to enable formation of axonal pathways. ¹ Among those cues known to be important for axon guidance are ephrins, netrins, slits, and semaphorins. ²  

Semaphorins are a large, conserved family of secreted or membrane-associated ligands that function as either repulsive or attractive cues for growth cones. ³ The semaphorin family is divided into subclasses based on functional domains and sequence similarity. Classes 1 and 2 are found in invertebrates, whereas classes 3 to 7 are found in vertebrates. Class 3 semaphorins are secreted ligands that bind to a family of transmembrane proteins known as neuropilins. ⁴ Neuropilins lack signaling motifs in the cytoplasmic tail and form receptor complexes with plexins to stimulate intracellular signaling cascades initiated by secreted semaphorins. Transmembrane class 4 and GPI-linked class 7 semaphorins bind directly to plexins. ^{5 6} Although genetic and in vitro data have established a clear role for semaphorins in axonal outgrowth, biochemical evidence suggests that semaphorins may ultimately affect the organization of the cytoskeletal network. ⁷  

Semaphorins and their receptors also have important roles in angiogenesis, organ development, and immune system functions. ^{8 9 10 11} In fact, several class 4 semaphorins such as Sema4A and Sema4D are involved in interactions between T cells and antigen-presenting cells. ^{12 13} Class 4 semaphorins are also expressed at high levels in the developing and adult brain, but their function in the nervous system is not known. ^{12 14 15} Sema4E has been shown recently to guide facial motor axons to their appropriate targets in zebrafish. ¹⁶ Sema4F induces collapse of retinal ganglion cell axons in vitro, ¹⁷ implying that other class 4 semaphorins have important as yet undiscovered roles in the developing nervous system. 

A strategy based on high-throughput retroviral gene-trapping in embryonic stem (ES) cells has produced mutations in approximately 60% of genes composing the mouse genome. ^{18 19} Using this technology, we generated a mutation in the semaphorin 4A (sema4A) gene that severely disrupts the development of retinal photoreceptors. Although photoreceptors are produced in mice lacking Sema4A, both rods and cones fail to form elongated outer segments during the second postnatal week, and subsequently most of the photoreceptors die before the first month of life. Sema4A is expressed in ganglion cells, inner retinal neurons, and retinal pigment epithelial (RPE) cells during the time at which photoreceptors establish intimate contacts with the RPE. Therefore, loss of photoreceptors in the absence of Sema4A is a non–cell-autonomous defect, suggesting that Sema4A functions as a transmembrane ligand for a receptor present on photoreceptors. 

Materials and Methods

Generation of Sema4A Mutant Mice

The sema4A gene trap allele was generated as part of the OmniBank gene trap database. ^{18 19} The ES cell clone represented by OmniBank Sequence Tag (OST) 393408 was chosen for the generation of Sema4A mutant mice based on sequence identity to the published mouse sema4A sequence (accession number, NM_013658). The precise genomic insertion site of the gene-trap vector was determined by inverse genomic PCR. ²⁰ The gene trap mutation was generated in Lex1 ES cells derived from the 129/SvEv^Brd strain. Mouse lines were generated by microinjection of the OmniBank ES cell clone represented by OST393408 into host blastocysts, using standard methods. ²¹ Chimeric mice resulting from the ES cell injections were bred to C57BL/6J albino mice for germline transmission of the sema4A mutation. Mice used in this study were of mixed genetic background (129/SvEv^Brd and C57BL/6J). All studies were performed in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Genotyping of Omnibank Mice

Oligonucleotide primers (LTRrev, 5′-GAG TGA TTG ACT ACC CGT CAG CG-3′, A, 5′-ACG GCT TGC TGT GGA GTC AGC-3′, and B, 5′-CCG ACT GAA AGC TCA GAA AGG C-3′) were used in a multiplex reaction to amplify corresponding sema4A alleles (Fig. 1A) . Approximately 100 ng of purified tail genomic DNA was used as a template for PCR. Cycling conditions were 94°C (30 seconds), 58°C (1 minute), and 72°C (30 seconds) repeated for 30 cycles. Amplified products were separated on precast 1.2% agarose gels (Amersham, Arlington Heights, IL). 

Expression Analysis of the Sema4A Mutation by RT-PCR

RNA was isolated from wild-type, heterozygous, and homozygous eyes using a bead homogenizer and extraction reagent (RNAzol; Ambion, Austin, TX) according to the manufacturer’s instructions. Reverse transcription (RT) was performed (SuperScript II; Invitrogen, Carlsbad, CA) according to the manufacturer’s instructions. PCR amplification conditions for the expression analysis were 95°C (30 seconds), 59°C, (45 seconds), 70°C (1 minute), repeated for 35 cycles with oligonucleotide primers (5′-CCC TGC TGG TGA AGT CTG GTG TGG AGT ACA-3′ and 5′-TGC AGG TTT CGA ACA GGC TCA GAG TCA G-3′) complementary to sema4A exons flanking the gene trap insertion site (Fig. 1B) . Primers to the mouse beta actin gene (accession number M12481) were used as an internal control. PCR products were verified by sequencing. Forward and reverse primers (5′-AAC CAT CTG GTG ACC ATC TCA GG-3′ and 5′-CCT CAG AAG AAC TCG TCA AGA AGG-3′) were used to amplify the fusion transcript between the endogenous sema4A gene and the selectable marker (neomycin phosphotransferase, NEO) in the gene-trap vector. PCR conditions are 95°C (2.5 minutes), then 95°C (45 seconds), 59°C (45 seconds), and 70°C for 1 minute, for 34 cycles. 

Fundus Photography and Fluorescein Angiography

Eyes of 14-week old wild-type (n = 4), heterozygous (n = 4), and homozygous (n = 8) mice were examined with a slit lamp (Nikon, Tokyo, Japan). Pupils were dilated with a drop of 1% cyclopentolate and 1% atropine (Alcon, Fort Worth, TX). A small animal fundus digital camera (Kowa Genesis, Tokyo, Japan) was used to photograph mouse fundi. ²² The instrument was used in conjunction with a 60- or 90-D condensing lens (Volk, Mentor, OH) mounted between the camera and eye. For retinal angiography, mice were injected intraperitoneally with 25% sodium fluorescein at a dose of 0.01 mL per 5 to 6 g body weight (Akorn Inc., Decatur, IL). The same general fundus photography procedure was used, except that the standard light was replaced with blue light with a barrier filter. The digital imaging system consists of a camera, computer, and software (Komit+ software; Kowa, Tokyo, Japan). 

Electroretinography

Electroretinograms (ERGs) were recorded from wild-type (n = 4) and homozygous (n = 5) mice at 3 weeks of age. Animals were deeply anesthetized with (per milliliter) ketamine (7.5 mg), xylazine (0.38 mg), and acepromazine (0.074 mg), delivered at 10 mL/kg body weight after a 2-hour dark-adaptation period. The left pupil was dilated with 0.1% atropine (Alcon). A platinum recording electrode was placed on the corneal surface of the left eye with a micromanipulator. A reference electrode was placed subcutaneously in the forehead and a ground electrode was placed subcutaneously in the proximal end of the tail. Rod-driven responses to light flashes covering a 3.0-log intensity range (0.0024–2.4 cd/mm²) were recorded by a visual electrodiagnostic system (UTAS-E 3000 with EM for Windows; LKC Technologies, Inc., Gaithersburg, MD). Cone-driven ERG responses to a stimulus intensity of 2.4 cd/mm² were recorded on a rod-saturating background. Signals were amplified 10,000× and filtered with low- and high-frequency cutoffs at 0.3 and 500 Hz, respectively. 

Histology and Immunohistochemistry

Histopathology was conducted on mice at different developmental and adult ages. Mice were euthanatized with CO₂ and eyes were removed and placed in Davidson’s fixative (Poly Scientific, Bay Shore, NY) overnight at room temperature. The following morning, eyes were processed in paraffin, cut at a thickness of 5 μm and stained with hematoxylin and eosin. For immunohistochemistry, mice were deeply anesthetized with (per milliliter) ketamine (7.5 mg), xylazine (0.38 mg), and acepromazine (0.074 mg), delivered at approximately 10 mL/kg body weight followed by perfusion with 4% paraformaldehyde in 0.1 M sodium phosphate-buffered saline (PBS; pH = 7.2). Eyes were removed and incubated in PBS containing 25% sucrose and embedded in tissue-freezing medium (Triangle Biomedical Sciences, Durham, NC). Sections were cut at a thickness of 12 to 16 μm and mounted (Superfrost/Plus slides; Fisher Scientific, Pittsburgh, PA). Primary antibodies were diluted in 5% normal goat serum (Vector Laboratories, Burlingame, CA) in PBS. Dilutions used were mouse anti-rhodopsin at 1:100,000 (Chemicon, Temecula, CA), rabbit anti-protein kinase C (PKC) at 1:5000 (Cambio, Cambridge, UK), mouse anti-calbindin D-28K at 1:5000 (Sigma-Aldrich, St. Louis, MO) and fluorescein-conjugated peanut agglutinin (PNA) at 1:250 (Vector Laboratories). AlexaFluor 594 goat anti-mouse (Molecular Probes, Eugene, OR) diluted at 1:400 was used to detect rhodopsin- and calbindin-bound antibodies, and AlexaFluor 594 goat anti-rabbit was used to detect PKC-bound antibodies. Coverslips were applied with antifade mounting medium containing 4′,6′-diamino-2-phenylindole (DAPI; Vectashield; Vector). Digital images were then obtained (ORCA II cooled CCD camera; Hamamatsu, Hamamatsu City, Japan; mounted on an BX60 microscope; Olympus, Lake Success, NY). Images were acquired in image-analysis software (Photoshop ver. 6.0; Adobe Systems, San Diego, CA) and saved at a resolution of 300 ppi. Figures were assembled in the software (Photoshop; Adobe Systems) and minimal adjustments were made to the figure contrast to obtain the best micrograph. 

Electron Microscopy

Mice at postnatal day (PD)10 and PD14 were anesthetized as described earlier and perfused with 3.5% glutaraldehyde in 0.1 M phosphate buffer. Eyes were removed and retinas were incubated in the same fixative overnight. Tissue was then incubated in 2% osmium tetroxide, stained with 2% uranyl acetate, and embedded in Spurr’s resin. Semithin sections were stained with 1% toluidine blue. Ultrathin sections were obtained and collected in formvar-coated slot grids and stained with lead citrate. Micrographs were taken at magnifications of 5,000× or 12,000×. 

In Situ Hybridization

The method of in situ hybridization analysis was performed as described elsewhere ²³ on 12-μm cryosections of eyes obtained from PD6, PD10, and adult albino mice. Albino mice were use in these studies to enable visualization of hybridization in the RPE, which in pigmented animals is obscured from view in dark-field due to melanin granules. A sema4A-specific cDNA (base pairs 702–1043 of NM_013658) was generated by PCR with primers that incorporate the T7 RNA polymerase promoter sequence into the PCR amplicon. This DNA template was used for in vitro transcription reaction with 80 μCi of αP33- UTP (NEN Life Science Products, Boston, MA). After hybridization at 60°C for 16 hours, sections were treated with RNase and washed in SSC buffer. Slides were dehydrated in a graded ethanol series and exposed to a 50% solution of autoradiographic emulsion type NTB2 (Eastman Kodak Company, Rochester, NY) for 3 to 6 days. Slides were developed using standard protocols, ²⁴ dehydrated, and coverslips were applied (Permount; Fisher Scientific, Pittsburgh, PA). Digital images were acquired (ORCA II; Hamamatsu) using a cooled CCD camera mounted on a BX60 (Olympus) microscope equipped with dark-field optics. 

Results

A high-throughput mutagenesis strategy has been adapted for mammals based on retroviral gene trapping in mouse ES cells. The retrovirus contains heterologous splice acceptor and donor sites that disrupt splicing of the trapped gene. ¹⁸ ES cells with mutations in approximately 60% of the mouse genome have been collected and are stored in an ES cell library called OmniBank. ¹⁹ The approximate insertion site and the identity of the trapped gene are revealed from an OST obtained through 3′ rapid amplification of cDNA ends (RACE) from exons downstream of the retroviral integration. We identified OST393408, which corresponds to a retroviral insertion in the semaphorin 4A (sema4A) gene (NM_013658). Inverse genomic PCR was performed to determine the precise insertion site of the retroviral gene-trap vector. Sequence comparison with the assembled mouse genome revealed that the integration occurred in intron 11 (Fig. 1A) . A genotyping strategy was designed using primers that flank the genomic insertion site of the gene-trap vector and a primer (LTRrev) specific for the gene-trap vector (Fig. 1A) . 

Bioinformatic analysis predicts that sema4A encodes a transmembrane protein comprising 760 amino acids (aa). Sema4A contains a signal peptide preceding a conserved semaphorin domain (aa 64-478), followed by a PSI domain (aa 496-580), an Ig-like domain (aa 570-630), a transmembrane domain (aa 680-702), and a short cytoplasmic tail (aa 703-760). Sequence analysis of RNA isolated from homozygous eyes revealed the presence of the expected fusion transcript between the endogenous sema4A exons upstream of the gene-trap insertion and the selectable marker in the gene-trap vector. ¹⁹ This fusion transcript encodes the Sema4A signal peptide followed by a truncated Sema domain (aa 64-438) lacking the last 40aa because of the presence of a stop codon engineered in the gene-trap vector. Native sema4A transcript was not detected in the brain (not shown) and eye of homozygous animals by RT-PCR using primers complimentary to exons 11 and 12 of the sema4A gene (Fig. 1B) , indicating efficient disruption of endogenous sema4A splicing. Amplification of RNA extracted from heterozygous mice revealed a decrease in the amount of endogenous sema4A transcript in mice with a single copy of the trapped allele. 

Mice deficient in Sema4A appeared healthy and failed to exhibit any phenotypes in tests performed in our neurologic screen. ²⁵ Crosses between F1 heterozygous individuals produced the expected ratios of genotypes in the F2 offspring, indicating that Sema4A is not required for embryonic development. Fundus photography performed at approximately 3 months of age revealed large areas devoid of pigmentation in all homozygous mutant mice examined (n = 6; Fig. 2C ), compared with their wild-type (n = 4) and heterozygous (n = 4) littermates (Figs. 2A 2B , respectively). Angiography revealed attenuated retinal vessels in homozygous mutant mice (Fig. 2F) compared with their wild-type and heterozygous littermates (Figs. 2E 2F , respectively). 

These noninvasive tests suggest that severe retinal degeneration is associated with disruption of Sema4A in adult animals. Electroretinography was performed on wild-type (n = 4) and homozygous mice (n = 5) approximately 3 weeks of age to determine the physiological consequences associated with loss of Sema4A. In wild-type mice, serial recordings from rod-driven circuitry revealed increased responses over a 3-log-unit intensity range, reflected in both the a-and b-wave (Fig. 2G) . In contrast, we failed to detect any responses in mice lacking Sema4A with light flashes up to 2.4 cd/mm², the maximum intensity tested. Cone-dominated responses were also attenuated in mice lacking Sema4A compared with their wild-type littermates at 3 weeks of age (Fig. 2H) . These results suggest that disruption of Sema4A is associated with a severely compromised retinal function. 

To determine the severity of retinal degeneration, histologic sections were taken through the central sagittal plane of eyes at approximately 3 weeks of age. In wild-type animals, the retina contains three cellular layers separated by two synaptic layers (Figs. 3A 3D) . Rod and cone photoreceptor cell bodies are located in the outer nuclear layer (ONL). Horizontal, bipolar, amacrine, and Müller glia cells comprise the inner nuclear layer (INL). Retinal ganglion cells, the sole projection neurons of the retina, and displaced amacrine cells are located in the ganglion cell layer (GCL). The organization of the retina in heterozygous mice was indistinguishable from that in wild-type mice (Figs. 3B 3E) . In contrast, the number of photoreceptors in mice lacking Sema4A was dramatically decreased at 3 weeks of age (Figs. 3C 3F) . Low-magnification images revealed a decrease in the thickness of the ONL in both the central and peripheral retina, with the most severe thinning occurring in the central retina at this age. Clusters of surviving photoreceptors were located immediately adjacent to the optic nerve head, similar to that reported in the RCS rat, which contains a mutation in the Mer receptor tyrosine kinase. ^{26 27} Higher-magnification images obtained approximately 500 μm from the optic nerve head revealed normal histology in retinas obtained from wild-type and heterozygous Sema4A mice (Figs. 3D 3E , respectively). In contrast, the ONL contained only a few rows of photoreceptors, whereas other retinal layers appeared relatively intact in mice lacking Sema4A (Fig. 3F) . Histologic analysis of eyes obtained from homozygous Sema4A mice over 1 year of age revealed a progressive retinal degeneration. At this age, retinas from homozygous mice lacked any detectable photoreceptors and the thickness of both the INL and the inner plexiform layer was decreased compared with either heterozygous or wild-type littermates (compare Fig. 3I with Figs. 3G and 3H ). Anatomy of the retina obtained from heterozygous Sema4A mice (Fig. 3H) was indistinguishable from that in wild-type mice (Fig. 3G) , demonstrating that a single copy of the endogenous sema4A is capable of sustaining retinal morphology. 

Mice contain both rod and cone photoreceptors, with rods comprising most of the photoreceptors. Retinas obtained from 3-week-old animals were incubated with an antibody that recognizes rhodopsin, which is expressed exclusively in rods. In wild-type mice, immunohistochemical staining with the rhodopsin antibody revealed intense staining in rod outer segments (OS) that contain stacked discs specialized for capturing light (Fig. 4A) . Weaker rhodopsin immunoreactivity was observed in rod inner segments (IS). Cones represent approximately 3% to 5% of all photoreceptors in mice, and these cells can be identified with PNA lectin. In retinas obtained from either wild-type (not shown) or heterozygous Sema4A mice, FITC-conjugated PNA bound to cone OS and IS (Fig. 4B) . In contrast, both rhodopsin and PNA staining revealed profound photoreceptor disease in mice lacking Sema4A. For example, rhodopsin immunoreactivity in presumptive rod outer segments was not observed in midperiphery sections taken from Sema4A-homozygous retinas (Fig. 4C) . Although cones were capable of binding PNA in Sema4A-deficient retinas, staining in the OS layer was not obvious and the thickness of the IS layer was dramatically decreased (Fig. 4D) . These results imply that loss of Sema4A results in an early onset, severe impairment in both rod and cone systems. 

Histopathology indicated a severely compromised retinal structure and function in mice at 3 weeks of age, suggesting that Sema4A is essential for photoreceptor development. The majority of photoreceptors in mice are produced during the first two postnatal weeks before the eye opens at approximately PD14. ²⁸ To assess the role of Sema4A in retinal development, light and electron microscopy was performed on retinal sections obtained from mice at PD10 and PD14. Morphogenesis of the mouse retina occurs in a gradient such that the central retina is developmentally more advanced compared with the retinal periphery. No obvious difference in retinal cellular anatomy was observed among wild-type, heterozygous, or homozygous mice at PD10 (Figs. 5A 5B) . The thickness of developing cellular and synaptic layers was comparable between mice lacking Sema4A and their control littermates at this age. 

Ultrastructure analysis of PD10 mice revealed that the initiation stage of photoreceptor outer segment formation occurs normally in wild-type mice and mice lacking Sema4A. For example, connecting cilia and rudimentary outer segment discs (Fig. 5C 5D , arrowheads) were apparent in both wild-type and Sema4A-deficient mice (Figs. 5C 5D , respectively). In wild-type and heterozygous mice at this age, developing outer segments were found adjacent to the RPE. In contrast, developing outer segments in mice lacking Sema4A were found adjacent to a layer (approximately 1 μm in thickness) of processes emerging from the RPE (Fig. 5D , arrows). Ultrastructural characteristics of these processes resembled actin filaments associated with the RPE apical microvilli. ²⁹ These exuberant RPE processes were never observed in either wild-type or heterozygous mice. 

At PD14, the cellular organization of the peripheral retina in Sema4A-deficient mice was indistinguishable from that in wild-type mice (compare Fig. 6A with 6B ). The outer limiting membrane and both synaptic layers were apparent in mice lacking Sema4A. Closer examination of outer segment morphology revealed changes in the organization of these important structures. In wild-type mice, outer segments were neatly aligned toward the RPE. In contrast, outer segments in Sema4A-deficient retinas lacked proper orientation toward the RPE (Fig. 6B , arrowheads), although the photoreceptor cell somas appeared normal. Disorganization of the OS was more apparent in the midcentral and central retina in mice lacking Sema4A compared with wild-type mice (Fig. 6D , arrowheads). In these regions, many cells in the ONL exhibited intensely stained nuclei associated with loss of chromatin organization, suggesting apoptotic cell death (Fig. 6D , arrows). TUNEL histochemistry (not shown) was consistent with photoreceptor apoptosis in retinas obtained from homozygous Sema4A mice. 

Ultrastructure analysis of retinas obtained from wild-type mice at PD14 revealed highly organized outer segments that establish close contacts with RPE apical microvilli (Figs. 7A 7C) . Outer segments were arrayed as stacked discs oriented toward the RPE. No differences in RPE polarity were detected between wild-type and homozygous mutant mice. Basal infoldings and apical microvilli were present in all mice examined, implying that Sema4A does not contribute to overall RPE cell polarity. However, the interface between apical microvilli and outer segments was highly abnormal in Sema4A-deficient mice. At all retinal locations studied in homozygous animals, outer segments were disorganized and appeared as whorls without proper orientation toward the RPE (Fig. 7B) . In Sema4A-deficient retinas, microvilli extended beneath the RPE adjacent to whorls of outer segments (Fig. 7B 7D) . These observations suggest a failure in cell–cell communication between the RPE and photoreceptor outer segments in the absence of Sema4A. 

Isotopic in situ hybridization was performed to determine the cellular expression of sema4A during postnatal retinal development and in adult retina. Sema4A is highly expressed in the retinal ganglion cell layer and in the amacrine cell layer associated with the inner portion of the inner nuclear layer at postnatal day 6 (Fig. 8A) . The sense control probe failed to hybridize in sections from mice at any age (Fig. 8B 8D 8F) . The RPE contains high levels of sema4A transcript, whereas developing photoreceptors did not express this gene. Elongation of photoreceptor outer segments begins during the second postnatal week in the mouse. ³⁰ At PD10, sema4A is highly expressed in the RPE adjacent to photoreceptors, which remain negative for sema4A expression (Fig. 8C) . Cells in the GCL and INL continue to express sema4A. This pattern of sema4A distribution is maintained in adult retinas with high levels of sema4A detected in the GCL, INL, and RPE (Fig. 8E) . Sema4A was not detected in photoreceptors at the ages examined in this study. 

High levels of sema4A expression were present in the GCL and INL, which contains cells that are postsynaptic targets of photoreceptors. To address the possibility that photoreceptor loss occurs primarily due to defects in the inner retina, antibodies recognizing rod bipolar cells and horizontal cells were used to stain retinas obtained from PD8 mice. Rod bipolar cells were identified with PKC immunoreactivity. ³¹ In retinas obtained from wild-type mice, rod bipolar cells were located in the outer portion of the INL (Fig. 9A) . Their dendrites extended into the outer plexiform layer (OPL), where they receive synaptic contact from rod photoreceptors. ³² Rod bipolar cells and their dendrites appeared normal in both Sema4A heterozygous and homozygous mice at PD8 (Figs. 9B 9C , respectively). Regardless of genotype, axons of rod bipolar cells extended toward the GCL, where they establish synaptic contacts with type AII amacrine cells in the IPL. ³³ Horizontal cells receive both rod and cone photoreceptor synaptic input. ^{14 34} In PD8 retinas, horizontal cells were identified using antibodies recognizing the calcium-binding protein, calbindin. ³¹ Horizontal cells were located in the outermost layer of the INL in retinas obtained from wild-type mice, and their processes extended throughout the OPL (Fig. 9D) . In retinas obtained from either heterozygous or homozygous Sema4A mice, horizontal cell density and their projections were indistinguishable from that in wild-type mice (Figs. 9E 9F) . Calbindin also identifies a subset of amacrine cells and their processes distributed in two strata in the IPL. ³¹ The distribution of calbindin-positive cells and their projections was indistinguishable among the genotypes examined. Additional cellular markers of inner retinal cells such as reelin, choline acetyltransferase, and calretinin (data not shown) also revealed comparable anatomy in wild-type, heterozygous, or homozygous Sema4A mice at PD8 or PD12 (not shown). Thus, several early developmental hallmarks of inner retinal cells are normal in the absence of Sema4A. 

Discussion

Semaphorins represent a large family of proteins implicated in a variety of biological mechanisms such as organogenesis, angiogenesis, immune system function, and nervous system development. ^{7 35} Secreted class 3 semaphorins are among the most well-characterized class known to be involved in nervous system development, where they function either as chemoattractants or repellents in a context-dependent manner. ³ Several class 4 semaphorins are expressed at high levels in the developing and adult nervous system, but their function in the nervous system is not known. ¹⁴  

In the present work, we show that homozygous mice carrying a retroviral gene trap insertion in intron 11 of the sema4A gene exhibit severe degeneration of retinal photoreceptors. This insertion disrupts splicing of endogenous sema4A exons that encode the C terminus of the extracellular Sema domain. Disruption of native sema4A transcript in the retina was confirmed by RT-PCR, and sequence analysis of the fusion transcript in homozygous eye samples revealed a premature stop codon in the mutant allele. The gene-trap allele described herein could result in the synthesis of a truncated protein containing the first 427aa of Sema4A. The fact that heterozygous mice exhibit normal retinal anatomy and are indistinguishable from wild-type mice in all other assays in our screen ²⁵ indicates that the degeneration observed in the homozygous mice is due to a loss of function of native Sema4A and not due to a dominant-negative or gain-of-function effect on Sema4A signaling by a truncated Sema4A product. 

Photoreceptor degeneration in Sema4A-deficient mice becomes apparent before the time at which the eyes open, at approximately PD14. This precocious degeneration implies that Sema4A is important for photoreceptor development. Most photoreceptors differentiate by the second postnatal week in mice. ^{28 36} Subsequently, morphologic differentiation of outer segments appears first in the central retina and proceeds toward the retinal periphery. ³⁰ The cellular organization of the retina in wild-type and Sema4A-deficient mice was comparable at PD8 and PD10, implying that photoreceptor production occurs normally in the absence of this transmembrane semaphorin. Moreover, immunohistochemical studies demonstrated that during the first 2 weeks of postnatal retinal development, the organization of rod bipolar and horizontal cells and the stratification pattern of the IPL are preserved in the absence of Sema4A. Although these studies do not formally exclude a molecular role for Sema4A in the development of synaptic connections between inner and outer nuclear layers, structural abnormalities detected in photoreceptor outer segments during the second postnatal week imply that photoreceptor degeneration occurs primarily due to a defect in the developing outer retina. 

Two stages of photoreceptor outer segment formation, initiation, and elongation, have recently been described. ²⁹ Photoreceptor outer segment initiation begins with the appearance of nonmotile connecting cilia, which are readily identifiable at the ultrastructural level. Connecting cilia and rudimentary outer segments were indistinguishable in mice lacking Sema4A compared with either wild-type or heterozygous littermates, indicating that Sema4A is not necessary for outer segment initiation. During the second stage of photoreceptor outer segment development, elongation occurs through the further accumulation and organization of membranous discs such that the adult morphology is attained by 3 weeks of age in mice. ^{29 30 37} At this age in mice lacking Sema4A, outer segments of both rod and cone photoreceptors are severely shortened. This result demonstrates that in the absence of Sema4A, outer segments fail to obtain their adult morphology characteristic of long, organized stacks of discs. 

Accumulating evidence suggests that elongation and organization of photoreceptor outer segments are dependent on the RPE, but the molecules responsible for this interaction are unknown. ^{29 37 38 39} In situ hybridization studies revealed that sema4A is expressed in the inner retina, and the RPE as early as PD6 and during the time at which outer segments attain their adult morphology. During postnatal weeks 2 and 3, apical microvilli of the RPE cells surround developing outer segments. This interaction is believed to support the organization of the membranous discs and to promote elongation of outer segments. ²⁹ At PD10, several characteristic features of RPE cell morphology, such as basal infoldings and melanosomes are apparent in mice lacking Sema4A. However, RPE apical microvilli in Sema4A-deficient mice, which normally project between outer segments, formed a dense layer interposed between the rudimentary outer segments and the RPE. This abnormality became more severe at PD14, with the emergence of large thumbprint-like collections of apical microvilli located above whorls of disorganized photoreceptor outer segments. These results suggest that transmembrane Sema4A, present in the RPE, is necessary for the proper development of photoreceptor outer segments. 

Sema4A was not detected in photoreceptors at the ages examined in this study. Loss of photoreceptors in the absence of Sema4A implies that photoreceptor degeneration occurs in a non–cell-autonomous manner. This observation is consistent with the notion that the initial morphogenesis of outer segments is independent of the RPE, but outer segment organization and elongation are dependent on signals derived from the RPE. ^{29 38 39 40} Sema4A present in the RPE could provide cell–cell communication between the RPE and rod and cone outer segments. This observation raises the interesting question of how Sema4A signals are received by photoreceptors. One possibility is that outer segments express receptors for Sema4A that transduce an organizational signal for developing outer segments. A receptor for Sema4A in the immune system, known as Tim-2 has recently been identified. ¹³ Identification of receptor components in the nervous system will help elucidate Sema4A-mediated signaling mechanisms. Collectively, these findings identify a previously unknown function of Sema4A in the visual system and provide a useful model to understand signaling mechanisms that control development of the outer retina. 

 Submitted for publication January 9, 2004; revised February 26, 2004; accepted March 9, 2004.

 Disclosure: D.S. Rice, Lexicon Genetics Inc. (E); W. Huang, Lexicon Genetics Inc. (E); H.A. Jones, Lexicon Genetics Inc. (E); G. Hansen, Lexicon Genetics Inc. (E); G.-L. Ye, Lexicon Genetics Inc. (E); N. Xu, Lexicon Genetics Inc. (E); E.A. Wilson, Lexicon Genetics Inc. (E); K. Troughton, None; K. Vaddi, Incyte Corporation (E); R.C. Newton, Incyte Corporation (E); B.P. Zambrowicz, Lexicon Genetics Inc. (E); A.T. Sands, Lexicon Genetics Inc. (E)

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Generation of Sema4A-deficient mice. (A) The gene-trap vector insertion in sema4A (accession number NM_013658) was identified in an ES cell clone from the OmniBank library and was used to generate Sema4A-deficient mice. Animals carrying the gene trap insertion in sema4A were identified by PCR. Primers A and B flank the gene trap insertion in intron 11 of sema4A and amplify a PCR product representing the wild-type (+/+) allele. The mutant allele was detected in heterozygous (+/−) and homozygous (−/−) mice with primer A and a vector-specific primer, LTRrev. (B) RT-PCR analysis of sema4A transcript using primers (RT) complimentary to exons 11 and 12. Endogenous sema4A transcript was detected in eyes of wild-type (+/+) and heterozygous (+/−) mice. Relative levels of native transcript were decreased in heterozygous eyes compared with that in wild-type. No endogenous sema4A transcript was detected in homozygous (−/−) eyes. RT-PCR analysis using primers (actin) complimentary to the mouse β-actin gene (accession number M12481) was performed in the same reaction as an internal amplification control.

Figure 2.

 

Fundus photography of Sema4A wild-type (A), heterozygous (B), and homozygous (C) mice at 14 weeks of age. The posterior eye in wild-type and heterozygous mice appeared normal, whereas retinal depigmentation was observed in mice lacking Sema4A. (D–F) Fluorescein angiograms. The retinal vasculature was normal in wild-type (D) and heterozygous (E) mice. (F) In contrast, mice lacking Sema4A exhibited attenuated retinal vessels and window defects, indicating retinal degeneration. (G) Electroretinography was performed on mice approximately 3 weeks of age. Serial rod responses to light increasing in intensity over a 3-log-unit range (maximum intensity is 24 cd-s/mm²) were recorded in wild-type mice (+/+). However, mice lacking Sema4A (−/−) failed to elicit rod-driven ERG activity represented in either a- or b-waves. (H) Cone-dominated responses on a rod saturating background at a stimulus intensity of 2.3 cd/mm² were recorded in wild-type mice. Cone responses were not detected in mice deficient in Sema4A.

Figure 3.

 

Histopathology of Sema4A wild-type, heterozygous, and homozygous mice at 3 weeks of age (A–F) and in mice over 1 year of age (G–I). The ONL, INL, and GCL were normal in wild-type (A) and heterozygous (B) Sema4A mice as shown in these low-magnification images. The optic nerve head is present in the bottom right in (A–C). (C) In mice lacking Sema4A, the ONL was dramatically decreased in thickness in the central and the peripheral retina. Retinal anatomy was comparable in wild-type (D) and heterozygous (E) mice. Photoreceptor outer segments (OS) and inner segments (IS) are indicated in (D). (F) The central retina in mice homozygous for the Sema4A mutation exhibited a profound loss of photoreceptors. (G–I) Histologic analysis of mice older than 1 year of age. Normal retinal anatomy was observed in wild-type (G) and heterozygous (H) Sema4A mice. In contrast, the homozygous retina (I) was very thin and atrophic. Photoreceptors were not apparent, and the INL was decreased in thickness compared with wild-type and heterozygous mice. Scale bar in (I): (A–C) 500 μm; (D–I) 100 μm.

Figure 4.

 

Rod and cone photoreceptors were abnormal in mice lacking Sema4A. Rhodopsin immunostaining of rods (A, C) and PNA labeling of cones (B, D) in retinas obtained from mice approximately 3 weeks of age. (A) Rod outer segments (OS) were heavily labeled with rhodopsin (red) antibodies in wild-type mice. Rod inner segments (IS) were negative at the dilutions used for rhodopsin immunohistochemistry. (C) Rod outer segments were not observed in homozygous Sema4A mice and the ONL was thinner compared with that in wild-type. Cone OS and IS bound to PNA (green) in the heterozygous (B) retina. (D) In contrast, cone OS were not apparent in mice lacking Sema4A, and the length of the IS was reduced. Cryosections were counterstained with DAPI (blue) to reveal nuclei. Scale bar: 50 μm.

Figure 5.

 

Histologic and ultrastructure analysis of eyes obtained from PD10 mice. (A) Semithin, toluidine blue–stained section obtained from a wild-type mouse revealed three nuclear layers (ONL, INL, and GCL) and developing outer segments (arrowhead) located beneath the RPE. Arrow: a dying cell in the INL, as revealed by the intense staining in the nucleus. (B) The cellular organization of the homozygous Sema4A retina was comparable to that in the wild-type shown in (A). Arrowhead: a dying cell in the INL. (C) Ultrastructural appearance of a PD10 retina obtained from a heterozygote mouse at the level of the outer retina. The RPE is identified by the numerous melanosomes located inside the RPE cells. Developing outer segments (arrowheads) were found adjacent to the RPE. The outer limiting membrane and the outer nuclear layer (ONL) are also apparent in this image. (D) The organization of the outer retina was preserved in mice lacking Sema4A. However, an ectopic layer containing numerous filamentous processes (arrows) was present between developing OS (arrowheads) and the RPE. Scale bar in (D): (A, B) 50 μm; (C, D) 4 μm.

Figure 6.

 

Histology of PD14 retina stained with toluidine blue at the level of the retina periphery (A, B) or central (C, D) retina. (A) Sections taken from the periphery revealed three cellular layers (ONL, INL, and GCL) in retinas obtained from wild-type mice. Developing photoreceptor outer segments (OS) and inner segments (IS) were aligned toward the RPE. (B) Cellular anatomy of retinas obtained from homozygous Sema4A mice is similar to that in the wild-type. Although the histology of photoreceptor cells in the ONL appeared normal, OS (arrowheads) in mice lacking Sema4A were disorganized and lacked proper orientation toward the RPE. (C) The OS in the central retina of wild-type mice were longer than that in the periphery (A). (D) In the central retina of mice deficient in Sema4A, many photoreceptor cells (arrows) were in the process of dying, and the OS (arrowheads) were severely disorganized at PD14. Scale bar, 50 μm.

Figure 7.

 

Electron microscopy of the outer retina at PD14. (A) Outer segments ( ) in wild-type mice were oriented toward the RPE, which surrounded the outer segments with apical microvilli (arrow). (B) Outer segments in mice lacking Sema4A appeared as whorls of membranes ( ) that failed to exhibit the normal structure seen in control animals at this age. These whorls accumulated beneath expansive apical microvilli of the RPE (arrow). (C) Intimate contacts formed between apical microvilli (arrow) and outer segments ( ) in the wild-type retina. Stacked discs of outer segments are apparent in this higher magnification image. (D) In mice lacking Sema4A, apical microvilli (arrows) were adjacent to whorls of outer segments ( ), which lacked proper orientation toward the RPE. Scale bar in (D): (A, B) 2 μm; (C, D) 1 μm.

Figure 8.

 

Expression of sema4A in the postnatal and adult eye. (A) Dark-field micrograph of whole eye sections incubated with an antisense riboprobe specific for sema4A. At PD6, high levels of sema4A expression were detected in the GCL, INL, and RPE. (B) No hybridization was obtained with the sense control probe on an adjacent section. (C) At PD10, sema4A was present at high levels in the RPE, INL, and GCL. Expression was not detected in photoreceptors in the ONL. (E) A similar pattern of sema4A expression was observed in adult retina. Sense probe on PD10 and adult eyes failed to hybridize, as shown in (D) and (F), respectively. Scale bar, 100 μm.

Figure 9.

 

Distribution of rod bipolar cells and calbindin-positive neurons in retinas obtained from PD8 wild-type, heterozygous, and homozygous Sema4A mice. Rod bipolar cells were identified with PKC immunoreactivity (A–C). Rod bipolar cells were located in the correct position within the inner nuclear layer (INL) in retinas obtained from wild-type (A), heterozygous (B), and homozygous (C) Sema4A mice. In all cases, dendrites (A, arrow) of rod bipolar cells extended toward the outer nuclear layer (ONL), and their axons projected toward the GCL. Horizontal cells, identified with calbindin immunoreactivity appeared in their appropriate location in the INL of retinas obtained from wild-type (D), heterozygous (E), and homozygous (F) Sema4A mice. Processes of horizontal cell extended into the OPL in all cases. In addition, calbindin immunoreactivity in wild-type mice was apparent in a subpopulation of amacrine cells in the INL, whose processes formed two distinct bands (arrowheads) in the IPL. A similar banding pattern was observed in retinas obtained from heterozygous (E) or homozygous (F) Sema4A mice at PD8. Scale bar, 50 μm.