|Home | About | Journals | Submit | Contact Us | Français|
The severity of disease in patients with retinitis pigmentosa (RP) can vary significantly, even among patients with the same primary mutations. It is hypothesized that modifier genes play important roles in determining the severity of RP, including the retinitis pigmentosa 1 (RP1) form of disease. To investigate the basis of variation in disease expression for RP1 disease, the authors generated congenic mice with a gene-targeted retinitis pigmentosa 1 homolog (Rp1h) allele (Rp1htm1Eap) on several different genetic backgrounds and analyzed their retinal phenotypes.
The Rp1htm1Eap allele was placed onto the C57BL/6J, DBA1/J, and A/J backgrounds. Retinal function of the resultant congenic mice was evaluated using electroretino-graphic analyses. Retinal structure and ultrastructure were evaluated using light and electron microscopy. Rp1h protein location was determined with immunofluorescence microscopy.
Analysis of the retinal phenotype of incipient congenic (N6) B6.129S-Rp1h+/tm1Eap, DBA.129S(B6)-Rp1h+/tm1Eap, and A.129S(B6)-Rp1h+/tm1Eap mice at 1 year of age showed retinal degeneration only in the A.129S(B6)-Rp1h+/tm1Eap mice. Further analyses revealed that the photoreceptors of the fully congenic A.129S(B6)-Rp1h+/tm1Eap mice show evidence of degeneration at 6 months of age and are almost completely lost by 18 months of age. In contrast, the photoreceptor cells in the fully congenic B6.129S-Rp1h+/tm1Eap mice remain healthy up to 18 months.
The severity of the retinal degeneration caused by the Rp1htm1Eap allele is notably dependent on genetic background. The development and characterization of the B6.129S-Rp1h+/tm1Eap and A.129S(B6)-Rp1h+/tm1Eap congenic mouse lines will facilitate identification of sequence alterations in genes that modify the severity of RP1 disease.
Inherited retinal degenerations are important causes of blindness.1–3 For the most part, these disorders are thought to be caused by mutations in single genes. Great progress has been made identifying the genes that harbor mutations that cause inherited retinal degenerations; more than 150 disease genes have been identified to date.4 Although the identified mutations are clearly pathogenic, a growing body of evidence suggests that mutations or sequence variants in genes other than the primary disease gene make important contributions to disease phenotypes.5,6 For example, the severity of retinal disease has been reported to vary significantly in patients who share primary mutations in a number of other retinal degeneration disease genes, including CRX, PRPH2, RHO, and RPGR.7–13 Variation in disease expression is especially evident in patients with retinitis pigmentosa 1 (RP1) disease. Mutations in RP1 are a common cause of dominant RP.14–18 There are also reports of mutations in RP1 causing recessive and simplex RP.19–21 Clinical characterization of patients with RP1 disease showed notable differences in visual field diameters and electroretinographic (ERG) amplitudes in patients of the same age who shared the same primary mutations.18,22
In general, variation in the phenotypes of genetic disorders can be attributed to allelic heterogeneity, environmental factors, and genetic modifiers.5,6 For RP1 disease, genetic modifiers are thought to be especially important because variation in the severity of disease is observed in patients with the same primary mutation and because environmental exposures are likely to be relatively similar within families.18,22 Identification of sequence variations in genes that modify disease severity is an important goal because the modifier genes may provide insight into the pathogenesis of disease. Modifier genes may also provide additional targets for therapeutic interventions.5,6
Although they are important factors in determining disease severity, few human modifier genes have been identified. One of the best examples comes from Bardet Biedl syndrome (BBS), a pleiotropic cilia disorder that includes retinal degeneration.23,24 BBS is an autosomal recessive disorder, but much of the observed phenotypic variability in patients with BBS cannot be explained by mutations at a single locus. It has been observed that some BBS patients have mutations at two BBS loci, indicating that inheritance of this disorder can be oligogenic.25–27 Further, a hypomorphic sequence variant in the CCDC28B gene has been observed to modify the severity of BBS caused by primary mutations in other BBS genes.28 Given that mutations in CCDC28B itself were not found to cause BBS, this is a BBS modifier gene.28 Similarly, cases of digenic RP caused by mutations in the PRPH2 and ROM1 genes have been reported.29 Sequence variants of the ROM1 gene serve as the modifiers. Several modifier loci for inherited retinal degenerations in mice have been identified. This includes quantitative trait loci for the Rd3, Nr2e3, and Rs1 genes and for age-related retinal degeneration.30–33
To investigate the basis of the variation in disease expression for RP1 disease, we generated congenic mice with the retinitis pigmentosa 1 homolog (Rp1h)-myc allele on several different genetic backgrounds.34 The official designation for this allele is Rp1htm1Eap (http://www.informatics.jax.org/). Rp1htm1Eap is a gene-targeted allele in which the mouse Rp1h gene is truncated to mimic most human RP1 mutations, which are either nonsense or frameshift mutations predicted to result in truncation of the RP1 protein after the N-terminal one-third.18,34 Homozygous Rp1htm1Eap/tm1Eap mice on a mixed 129/B6 background experience a relatively rapid photoreceptor degeneration, characterized by loss of organization of outer segment discs. In contrast, heterozygous Rp1h+/tm1Eap mice on the mixed 129/B6 background do not show signs of retinal degeneration up to 1 year of age.34 We generated several strains of Rp1h+/tm1Eap congenic mice and analyzed their retinal phenotypes. Results show that A.129S(B6)-Rp1h+/tm1Eap congenic mice, with the heterozygous mutant allele transferred to the A/J background, experienced photoreceptor degeneration whereas pigmented B6.129S-Rp1h+/tm1Eap congenic mice did not.
This research was conducted in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and the University of Pennsylvania Guidelines for Animal Care and Use. Rp1htm1Eap mice were generated by gene targeting, as described.34 Wild-type C57BL/6J, DBA1/J, and A/J mice were purchased from the Jackson Laboratories (Bar Harbor, ME). Heterozygous Rp1h+/tm1Eap mice on a mixed 129-B6 background were crossed with C57BL/6J, DBA1/J, and A/J mice to initiate the production of congenic Rp1htm1Eap strains. N1 heterozygous Rp1h+/tm1Eap offspring from each of these outcrosses were repeatedly backcrossed with wild-type C57BL/6J, DBA1/J, and A/J mice to generate the B6.129S-Rp1htm1Eap, DBA.129S(B6)-Rp1htm1Eap, and A.129S(B6)-Rp1htm1Eap lines, respectively. The notation used for describing these congenic mice follows the Rules and Guidelines for Nomenclature of Mouse and Rat Strains, which are available at the Mouse Genome Informatics Web site (http://www.informatics.jax.org/mgihome/nomen/strains.shtml). Lighting in the animal facility was maintained on a 12-hour light/12-hour dark cycle, with 5 to 10 lux of dim white light at the cage surface. Mice were genotyped by PCR amplification of a 612-bp product specific for the Rp1htm1Eap allele using primers 5′-GTGCCACAACAATGCTTTACCGT-CAAC-3′ and 5′-GGACTGAGGGGCCTGAAATGA-3′.34 To identify the Rpe65450 allele, total RNA was reverse transcribed, and a 400-bp fragment containing codon 450 was amplified using primers 5′-GTCA-CACTGCCCCATACAACT-3′ and 5′-CAGCCCTGGCAATTTCACTCAA-3′. PCR products were purified (QIAquick kit; Qiagen, Valencia, CA) and sequenced.35
Eyes from wild-type or heterozygous mice from the B6.129S-Rp1htm1Eap and A.129S(B6)-Rp1htm1Eap congenic lines were processed for immunostaining experiments, as described previously.34 Briefly, the eyes were enucleated after cardiac perfusion and fixed in 4% paraformaldehyde for 3 hours, then embedded in OCT freezing media and cryosectioned at 10 μm. Frozen sections were immunostained with anti–Rp1h and anti–myc antibodies, followed by Alexa 488- and Alexa 555-conjugated secondary antibodies (Invitrogen, Carlsbad, CA). Stained sections were viewed with a confocal microscope (LSM 510 Meta; Zeiss, Thornwood, NY), and the images were processed with Zeiss software (Meta 510).
Eyecups from the desired age and genetic background of heterozygous Rp1h+/tm1Eap mice and wild-type littermate controls were fixed in 2% paraformaldehyde + 2% glutaraldehyde in 0.2 M sodium cacodylate buffer (pH 7.4) for 4 hours. Eyecups were trimmed and postfixed in 1% osmium tetroxide. Tissues were then dehydrated in a graded ethanol series, infiltrated, and embedded in Epon resin (EMbed812; Electron Microscopy Sciences, Hatfield, PA) according to the manufacturer’s instructions. Sections of 1-μm thickness were cut and stained with alkaline toluidine blue for light microscopy. Ultrathin sections of 60- to 80-nm thickness were cut, stained with lead citrate and uranyl acetate, and examined using a transmission electron microscope.34,36
The thickness of the outer nuclear layer in 1-μm thickness of toluidine blue-stained retinal sections from B6.129S-Rp1htm1Eap and A.129S(B6)-Rp1htm1Eap congenic mice was measured using image processing software (Image-pro Plus 6.0; Media Cybernetics, Bethesda, MD) essentially as described.37 ONL thickness measurements from the Rp1h+/tm1Eap mice were compared with those observed in littermate control mice using Student’s t-test.
Electroretinography was performed as previously described.34,38 Briefly, mice were dark adapted for a minimum of 12 hours before the ERG experiments. Preparations of the animals for recordings were made under dim red light. Pupils were dilated with 1% tropicamide. Full-field electroretinograms were recorded in a Ganzfeld dome on dark-adapted, anesthetized mice, with care taken to maintain 37°C body temperature at all times. Retinal responses were detected with platinum electrodes embedded in contact lenses on the cornea and recorded using custom software. Rod a-wave, rod b-wave, and cone b-wave amplitudes measured in the Rp1h+/tm1Eap mice were compared with those observed in the littermate control mice using Student’s t-test.
We followed the recommendations of the Banbury Conference on phenotypic analyses of gene-targeted mouse strains to analyze the effect of genetic background on the phenotype of the Rp1htm1Eap allele.39 The phenotypes of all three lines of Rp1h+/tm1Eap mice were analyzed at the incipient congenic (N6) stage. Backcrosses of the B6.129S-Rp1htm1Eap and A.129S(B6)-Rp1htm1Eap lines were continued to generate fully congenic (N10) lines.
To examine the retinal function of mice with the Rp1htm1Eap allele on different genetic backgrounds, we measured full-field ERG responses of 12-month-old Rp1h+/tm1Eap mice and litter-mate controls from all three lines at the incipient congenic (N6) stage. The retinal function of heterozygous B6.129S-Rp1h+/tm1Eap and DBA.129S(B6)-Rp1h+/tm1Eap mice was the same as that observed in wild-type littermate controls (Fig. 1). In contrast, rod and cone function was decreased in the A.129S(B6)-Rp1h+/tm1Eap mice compared with controls (Fig. 1). Specifically, the saturating amplitude of the rod a-wave(amax) of A.129S(B6)-Rp1h +/tm1Eap mice was reduced by approximately 36% at 12 months of age (P < 0.01). Rod b-wave amplitudes were significantly reduced in the A.129S(B6)-Rp1h+/tm1Eap mice compared with wild-type littermate controls. Cone b-wave amplitudes were also reduced by approximately 28%, though this difference was not statistically significant (Fig. 1). Of note, the amplitudes of the rod and cone responses in the control albino mice were less than those observed in the control pigmented mice of the same age, consistent with the age-related retinal degeneration described in these mice.31
We also assessed the retinal function of fully congenic (N10) B6.129S-Rp1htm1Eap and A.129S(B6)-Rp1htm1Eap mice. As with the N6 mice, the retinal function of the N10 B6.129S-Rp1h+/tm1Eap mice was the same as that observed in wild-type littermate controls at 6 and 12 months of age (Fig. 2). Rod a-wave amplitudes of the 6-month-old N10 A.129S(B6)-Rp1h+/tm1Eap mice were significantly decreased, in this case by 23% (P < 0.05). By 12 months of age, the retinal function of the A.129S(B6)-Rp1h+/tm1Eap congenic mice was decreased further, similar to the results obtained from the N6 mice (Fig. 2).
To evaluate the extent of photoreceptor degeneration in the A.129S(B6)-Rp1h+/tm1Eap congenic mice, we examined the thickness of the outer nuclear layer (ONL) in semithin retinal sections from heterozygous mice and wild-type littermate controls at 6, 12, and 18 months of age. Eyes from three to four Rp1h+/tm1Eap and Rp1h+/+ mice were evaluated at each time point. At the light microscopic level, the A.129S(B6)-Rp1h+/tm1Eap mice demonstrated progressive degeneration of photoreceptor cells (Fig. 3). At 6 months of age, the ONL of retinas from the A.129S(B6)-Rp1h+/tm1Eap mice was one to two rows of nuclei thinner than controls. By 12 month of age, the ONL of the A.129S(B6)-Rp1h+/tm1Eap mice decreased to five to six rows of nuclei, which was two to three rows thinner than the ONL of wild-type littermate controls. This difference in ONL thickness was present throughout the retina (Fig. 4) and was statistically significant (P < 0.0001). Retinal degeneration was more severe in 18-month-old A.129S(B6)-Rp1h+/tm1Eap mice, with only two to three rows of photoreceptor nuclei remaining. Although some age-related degeneration was observed in the control albino mice, five to six rows of photoreceptor nuclei were still present in the control retinas at 18 months of age (Fig. 3). In contrast, the retinal morphology in 6-, 12-, and 18-month-old heterozygous B6.129S-Rp1h+/tm1Eap mice was similar to that of age-matched control animals, without evidence of photoreceptor cell loss (Fig. 3).
The ultrastructure of photoreceptors in Rp1htm1Eap congenic mice was evaluated by electron microscopy. On the A/J genetic background, photoreceptor outer segment length and organization decreased progressively in the heterozygous Rp1h+/tm1Eap mice. At 12 months of age, the photoreceptor outer segments of A.129S(B6)-Rp1h+/tm1Eap mice were shorter and demonstrated early signs of disorganization (Figs. 5I, 5J). By 18 months of age, the outer segments of the A.129S(B6)-Rp1h+/tm1Eap mice were notably shorter and more disorganized than those in controls. Small packets of enlarged and incorrectly oriented discs were seen in place of intact outer segments (Figs. 5K, 5L). Ultrastructural analyses of retinas from the B6.129S-Rp1h+/tm1Eap mice demonstrated normal photoreceptor outer segment structure up to 18 months of age (Figs. 5C, 5D).
Locations of Rp1h proteins in the A/J and C57BL/6J congenic strains were evaluated using immunofluorescence analyses. The truncated Rp1h-myc protein produced from the Rp1htm1Eap allele was detected using antibodies to the myc tag at the C-terminal end of this protein, and the full-length Rp1h protein was detected using antibodies to the C-terminal portion of the protein, which is not present in the truncated Rp1h-myc protein.34,40 Immunofluorescence analyses using frozen retinal sections from B6.129S-Rp1h+/tm1Eap and A.129S(B6)-Rp1h+/tm1Eap mice showed that the Rp1h-myc protein was correctly located in the axoneme of photoreceptor outer segments, as reported previously (Fig. 6).34 The Rp1h-myc protein colocalized with the full-length Rp1h protein in both Rp1h+/tm1Eap congenic lines (Fig. 6). Signal intensities of the wild-type Rp1h and mutant Rp1h-myc proteins were similar in the B6.129S- Rp1h+/tm1Eap and A.129S(B6)-Rp1h+/tm1Eap mice.
Previous studies have shown that a sequence variation in the Rpe65 gene can act as a genetic modifier of inherited retinal degeneration in mice. C57BL/6J mice have a variant in their Rpe65 gene (Rpe65450Met) that is protective against light dam-age and retinal degeneration caused by the VPP rhodopsin transgene.41–43 The Rpe65450Met variant leads to decreased levels of Rpe65 protein, a decreased rate of rhodopsin regeneration, and protection against photoreceptor damage.35,41,42 As for most other mouse strains, A/J mice have Leu at position 450 of their Rpe65 protein. Although no mechanistic connection is evident between the RP1 and RPE65 proteins, we investigated the potential contribution of the Rpe65-Leu450Met variation on the phenotype of the A.129S(B6)-Rp1h+/tm1Eap mice by evaluating the retinal function of albino mice generated in the N2 cross during development of the A.129S(B6)-Rp1htm1Eap congenic line. At 1 year of age, rod and cone ERG amplitudes were not significantly different between the Rpe65450Leu/Leu mice (n = 13) and the Rpe65450Met/Leu mice heterozygous for the Rp1htm1Eap allele (n = 7).
The results described show that the severity of the retinal degeneration caused by the Rp1htm1Eap allele is notably dependent on genetic background. Photoreceptors of the A.129S(B6)-Rp1h+/tm1Eap congenic mice show evidence of degeneration as early as 6 months of age and are almost completely lost by 18 months of age. In contrast, photoreceptor cells in the B6.129S-Rp1h+/tm1Eap congenic mice remain healthy up to 18 months of age. Incipient congenic N6 DBA.129S(B6)-Rp1h+/tm1Eap mice also retain normal retinal function. These results establish the basis for using the A.129S(B6)-Rp1h+/tm1Eap and B6.129S-Rp1h+/tm1Eap congenic mice to map and identify sequence alterations in genes that modify the severity of RP1 disease.18,22 As described, identification of a modifier gene or genes for RP1 would be valuable for several reasons, including improved understanding of disease pathogenesis and insight into potential therapies.6 Genes that harbor sequence alterations that modify the severity of RP1 disease may also be disease genes in their own right. In addition, genetic modifiers of RP1 disease severity may affect other inherited retinal degenerations; thus, their identification will help improve our understanding of the genetic variations that cause and modulate inherited retinal diseases.
Identification of modifier loci in patient populations can be difficult because of the large number of patients required.5,44 Mouse models of genetic disorders can facilitate modifier screens by providing genetically homogeneous populations and controlled environments. Further, inbred mouse strains replicate much of the genetic diversity of the human population, and several modifiers identified in mice have been found to be relevant to human disease. For example, polymorphisms in the SORCS1 (sortilin-related VPS10 domain containing receptor 1) gene have recently been linked to susceptibility to diabetes in patients; this modifier gene was originally identified in mice.45
The finding that truncation of the RP1 protein can cause dominant disease in mice is also important because this is the primary mode of inheritance of RP1 disease in humans.14–17,19–21 This finding confirms the value of gene-targeted mice as models of RP1 disease. It also confirms that loss of the C-terminal two-thirds of the RP1 protein leads to loss or alteration of protein function.14,34 Based on the dominant mode of inheritance in humans, it has been suggested that the truncated RP1 protein has a toxic or dominant-negative effect on photoreceptor cells.14 The observation that dominant disease also occurs in mice is consistent with this hypothesis, but additional studies are needed to determine whether RP1 disease is caused by loss or gain of function.
The observation that the retinal degeneration caused by the Rp1htm1Eap allele is more severe in albino mice raises the possibility that the lack of pigment in these mice or their increased susceptibility to light-induced retinal degeneration contributes to the Rp1htm1Eap phenotype.46 A/J mice are albino because of a mutation (C103S) in the tyrosinase (Tyr) gene.47 In humans, defects in tyrosinase result in oculocutaneous albinism type 1 (OCA1). Patients with OCA1 have several ocular abnormalities, including iris transillumination, foveal hypoplasia, and misrouting of ganglion cell axons in the optic chiasm.48 It is not known why diminished melanin pigment formation leads to defects in axon routing or foveal development. Albino mice share many of these defects and have been reported to have approximately 25% fewer rod photoreceptors than C57BL/6 mice.49 A/J mice have also been reported to experience faster age-related decline in photoreceptor cell numbers than C57BL/6J mice.31 It is therefore possible that the more severe Rp1 phenotype observed in the A/J congenics is related to lack of pigmentation. What is not known, however, is how the amount of melanin in RPE cells could influence the RP1 protein in the axoneme of photoreceptor cells.34,36 Further, the difference in the rates of age-related retinal degeneration between the A/J and C57BL/6J mice is not the result of pigmentation because C57BL/6J-c2J mice, which are albino but otherwise identical to C57BL/6J mice, have the same relatively slow rate of age-dependent retinal degeneration as C57BL/6 mice.31
Lack of pigmentation in albino mice contributes to their increased susceptibility to light-induced retinal degeneration compared with pigmented mice, in part because of lack of pigment in the irides of albino mice, which permits increased transmission of light to the retina.46 It is not likely that light-induced damage played a role in the degeneration observed in the A.129S(B6)-Rp1h+/tm1Eap congenic mice because all mice used in these studies were raised in dim light, with approximately 5 to 10 lux of light at the cage surface. This level of light is too low to cause light-induced degeneration.50
An additional issue related to the more severe retinal degeneration observed in the A.129S(B6)-Rp1h+/tm1Eap congenic mice compared with the pigmented B6.129S-Rp1h+/tm1Eap congenic mice relates to the Rpe65 gene. C57BL/6J mice have a variant in the Rpe65 gene (Rpe65450Met) that is protective against light damage. The Rpe65450Met variant leads to decreased levels of Rpe65 protein, decreased rhodopsin regeneration, and consequent protection against light-induced damage.35,41,42 The Rpe65450Met variant is also protective against retinal degeneration caused by the VPP rhodopsin transgene and thus can function as a modifier allele for retinal degenerations caused by mutations in genes associated with photo-transduction.43 Like most other strains of mice, the A/J strain has the nonprotective Rpe65450Leu variant of the Rpe65 gene, raising the possibility that this could contribute the increased severity of the Rp1 phenotype in the A/J congenic mice. Several results from the studies described indicate that this is not the case. First, no degeneration was observed in the DBA incipient congenic mice, which have the Rpe65450Leu allele. Second, we compared the retinal function of albino mice with Rpe65450Met/Met and Rpe65450Met/Leu alleles generated in the N2 cross during development of the A.129S(B6)-Rp1htm1Eap congenic line and did not observe any difference in retinal function in Rp1h+/tm1Eap mice that were Rpe65450Met/Met compared with Rpe65450Met/Leu.
Mice heterozygous at position 450 of Rpe65, with one Met and one Leu allele, have an intermediate sensitivity to light damage because of intermediate levels of Rpe65 and a twofold reduction in the rate of rhodopsin regeneration compared with Rpe65450Leu/Leu mice. 35,42 These results suggest that the retinal degeneration observed in the A.129S(B6)-Rp1htm1Eap congenic mice is not caused by the loss of the protective Rpe65 allele in the albino mice.
Based on the data presented here, we conclude that sequence variations in one or more modifier genes in the B6 or A/J strains are responsible for the observed difference in phenotypes between the A.129S(B6)-Rp1h+/tm1Eap and the B6.129S-Rp1h+/tm1Eap congenic mice. An analysis of 8.27 million SNPs in 16 inbred strains of mice showed that C57BL/6J and A/J were discordant at 14% (approximately 1 million) of the 7.47 million SNPs successfully tested in the two strains.51 We hypothesize that this genetic variation underlies the phenotypic variation between the albino and black congenic Rp1htm1Eap mice. Among the SNPs that were discordant between the B6 and A/J strains were 28 SNPs that resulted in premature stop codons, 9 SNPs that altered stop codons, 16 SNPs that altered start codons, and 42 SNPs that altered splice sites in the A/J strain.51 One of the genes that harbors a premature stop codon in the A/J strain is Gpr98. Mutations in human GPR98 cause Usher syndrome 2C, in addition to familial febrile seizures.52,53 The A/J allele of Gpr98 results in an Arg4296Ter mutation (GenBank NM_054053.4; NP_473394.3). Because this mutation is in exon 64 of 90, the mutant allele would be expected to be null secondary to non–sense-mediated decay.54 The Gpr98 protein has been localized to the transition zones (connecting cilia) of photoreceptor cells, which is adjacent to the location of the Rp1h protein in the axoneme of photoreceptor cells.36,40,55 These proteins are thus part of the photoreceptor sensory cilium, which encompasses the outer segments, transition zone, and basal bodies of photoreceptor cells.56 The mutant allele of Gpr98 in the A/J mice raises the possibility that the Gpr98 and Rp1 proteins interact or participate in the same cell biological pathway. The A.129S(B6)-Rp1h+/tm1Eap and the B6.129S-Rp1h+/tm1Eap congenic mouse lines provide the opportunity to identify sequence differences between the B6 and A/J strains that modify the Rp1 phenotype in mice.
Supported by the National Institutes of Health (Grant EY12910), F. M. Kirby Foundation, Foundation Fighting Blindness, Research to Prevent Blindness, Rosanne Silbermann Foundation, and the Mackall Foundation Trust.
The authors thank Ray Meade and Biao Zuo in the Biomedical Imaging Core of the University of Pennsylvania School of Medicine for their assistance with preparing samples for histologic and ultrastructural analyses; Yun Liu for his technical assistance; and Edward Pugh for his thoughtful comments on the manuscript.
Disclosure: Q. Liu, None; A. Saveliev, None; E.A. Pierce, None