Molecular Vision 2016; 22:1468-1489
Received 13 January 2016 | Accepted 22 December 2016 | Published 24 December 2016
1College of Optometry, University of Houston, Houston, TX; 2Department of Biology and Biochemistry, University of Houston, Houston, TX; 3Department of Natural Sciences, University of Houston-Downtown; Houston, TX; 4Department of Pharmacology and Pharmaceutical Science, University of Houston, Houston, TX
Correspondence to: Dr. Shawnta Y Chaney, University of California San Francisco, Department of Ophthalmology, 675 Nelson Rising Lane, Rm535, San Francisco, CA 94134 Phone: 415-502-7350 email: Shawnta.Chaney@ucsf.edu
Dr. A. Giddabasappa is now at: Pfizer Inc., 10724 Science Center Dr. San Diego, CA 92121 Dr. W.R. Hamilton is now at: Solenis, 3 Beaver Valley Rd. Suite 500 Wilmington, DE 19803 Dr. S. Mukherjee is now at: UT Southwestern Medical Center Department of Molecular Biology, 5901 Forest Park Rd, Dallas, TX 75390
Purpose: Studies of neuronal development in the retina often examine the stages of proliferation, differentiation, and synaptic development, albeit independently. Our goal was to determine if a known neurotoxicant insult to a population of retinal progenitor cells (RPCs) would affect their eventual differentiation and synaptic development. To that end, we used our previously published human equivalent murine model of low-level gestational lead exposure (GLE). Children and animals with GLE exhibit increased scotopic electroretinogram a- and b-waves. Adult mice with GLE exhibit an increased number of late-born RPCs, a prolonged period of RPC proliferation, and an increased number of late-born rod photoreceptors and rod and cone bipolar cells (BCs), with no change in the number of late-born Müller glial cells or early-born neurons. The specific aims of this study were to determine whether increased and prolonged RPC proliferation alters the spatiotemporal differentiation and synaptic development of rods and BCs in early postnatal GLE retinas compared to control retinas.
Methods: C57BL/6N mouse pups were exposed to lead acetate via drinking water throughout gestation and until postnatal day 10, which is equivalent to the human gestation period for retinal neurogenesis. RT-qPCR, immunohistochemical analysis, and western blots of well-characterized, cell-specific genes and proteins were performed at embryonic and early postnatal ages to assess rod and cone photoreceptor differentiation, rod and BC differentiation and synaptic development, and Müller glial cell differentiation.
Results: Real-time quantitative PCR (RT-qPCR) with the rod-specific transcription factors Nrl, Nr2e3, and Crx and the rod-specific functional gene Rho, along with central retinal confocal studies with anti-recoverin and anti-rhodopsin antibodies, revealed a two-day delay in the differentiation of rod photoreceptors in GLE retinas. Rhodopsin immunoblots supported this conclusion. No changes in glutamine synthetase gene or protein expression, a marker for late-born Müller glial cells, were observed in the developing retinas. In the retinas from the GLE mice, anti-PKCα, -Chx10 (Vsx2) and -secretagogin antibodies revealed a two- to three-day delay in the differentiation of rod and cone BCs, whereas the expression of the proneural and BC genes Otx2 and Chx10, respectively, increased. In addition, confocal studies of proteins associated with functional synapses (e.g., vesicular glutamate transporter 1 [VGluT1], plasma membrane calcium ATPase [PMCA], transient receptor potential channel M1 [TRPM1], and synaptic vesicle glycoprotein 2B [SV2B]) revealed a two-day delay in the formation of the outer and inner plexiform layers of the GLE retinas. Moreover, several markers revealed that the initiation of the differentiation and intensity of the labeling of early-born cells in the retinal ganglion cell and inner plexiform layers were not different in the control retinas.
Conclusions: Our combined gene, confocal, and immunoblot findings revealed that the onset of rod and BC differentiation and their subsequent synaptic development is delayed by two to three days in GLE retinas. These results suggest that perturbations during the early proliferative stages of late-born RPCs fated to be rods and BCs ultimately alter the coordinated time-dependent progression of rod and BC differentiation and synaptic development. These GLE effects were selective for late-born neurons. Although the molecular mechanisms are unknown, alterations in soluble neurotrophic factors and/or their receptors are likely to play a role. Since neurodevelopmental delays and altered synaptic connectivity are associated with neuropsychiatric and behavioral disorders as well as cognitive deficits, future work is needed to determine if similar effects occur in the brains of GLE mice and whether children with GLE experience similar delays in retinal and brain neuronal differentiation and synaptic development.
The adult retina consists of six different types of neurons (i.e., rod and cone photoreceptors and horizontal, bipolar, amacrine, and ganglion cells) and one glial cell (i.e., Müller glia). These seven cell types develop in a specific histogenic order that is conserved across species [1,2]. In mice, early-born neurons (i.e., cone photoreceptors and horizontal, amacrine, and ganglion cells) exhibit peak neurogenesis before birth, while late-born neurons (i.e., rod photoreceptors and bipolar cells [BCs]) and Müller glia peak after birth . Despite the significant time span it takes for retinal cells to develop (i.e., embryonic day [E] = 10.5; postnatal day [PN] = 10 in mice), all these cells originate from one common pool of multipotent retinal progenitor cells (RPCs) [4,5]. To ensure the production of the correct number of cell types, the developing retina must maintain a constant regulation of mitosis, proliferation, and differentiation throughout development .
In the later stages of retinal development, differentiated retinal neurons undergo a regulated sequence of synaptic development in the outer and inner plexiform layers (OPLs and IPLs). For example, the developmental sequence of synapses in the OPL, as determined by vesicular glutamate transporter 1 (VGluT1) protein expression, is as follows: 1) cone pedicles, 2) horizontal synapses, 3) rod spherules, and 4) BC innervation . In the IPL, the OFF pathway develops before the ON pathway . Much of what is known about functional synaptic development in the retina is in the context of mediated activity-dependent mechanisms [8-10].
Other than the coordinated regulation that occurs within the individual stages of the proliferation, differentiation, and functional development of neurons in the young retina, little is known about the interdependence of each phase within populations of RPCs. Many studies have examined the effects of various genes (i.e., cyclin-dependent kinases, transcription factors, cell signaling proteins, etc.) [11-14], soluble neurotrophic factors , and/or neurotransmitters  on cell proliferation, the cell cycle, and/or cell fate decisions. The goal of the current study was to determine how the increased and prolonged proliferation of RPCs would affect the subsequent structural and functional maturation of the mouse retina.
Gestational lead exposure (GLE) in children, monkeys, and rodents results in dose-dependent supernormal scotopic electroretinograms [17-20]. Consistent with these results, we observed a dose-dependent increase in the number of RPCs, prolongation of RPC proliferation, and selective increase in the number of rods and BCs in adult mice  and Long-Evans hooded rats . This is in marked contrast to postnatal or adult lead exposure, at similar blood lead levels, which produces rod-selective apoptosis [22,23], photoreceptor synaptic degeneration , and hippocampal granule cell apoptosis [25,26], synaptic dysfunction, and degeneration .
We hypothesized that the increased number of RPCs accelerates the development of rods and BCs and increases the abundance of the functional synaptic markers of these cell types as they differentiate and mature. We used a murine model of GLE. Unexpectedly, we found a two- to three-day delay in the initial expression of proteins associated with rod and BC differentiation, synaptic development and maturation.
All experimental and animal care procedures complied with the National Institutes of Health (NIH) Public Health Service Policy on Humane Care and Use of Laboratory Animals (NIH 2002), were approved by the Institutional Animal Care and Use Committee of the University of Houston, and adhered to the Association for Research in Vision and Ophthalmology (ARVO) Statement for the Use of Animals in Research. Data was collected from litters bred at our facility and maintained on a 14:10 light:dark cycle (100 lux: maximum cage luminance) with food and water available ad libitum [21,28]. Wild-type C57BL/6N mice (Harlan Sprague-Dawley, Indianapolis, IN) were used for all the experimental data shown herein. We observed the same phenotypic effects of GLE in C57BL/6J mice  and Long-Evans hooded rats . Despite the rd8 mutation [30,31], we did not observe dysplasia, irregularities in retinal layer thickness, or the displacement of photoreceptors at any experimental time point in this or in our previous studies [21,24,29,32]. Briefly, naïve females were given either tap (control) or 55 parts per million (ppm) lead acetate containing drinking water (GLE group) for two weeks before mating, during pregnancy, and until PN10, after which the lead was replaced with tap water. For the E16.5 and E18.5 retinas, the dams were mated overnight and checked for vaginal plugs in the morning, and the pups were used for timed-pregnant experiments. Mice were sacrificed by decapitation between 1,000 and 1,200 h. The Institutional Animal Care and Use Committee of the University of Houston approved decapitation without anesthetic for all ages so we have not included information of this type. As published, the control and GLE groups had peak blood lead concentrations at PN0 and/or PN10 of 0.75 ± 0.06 and 22.11 ± 1.05 μg/dl, respectively. In the E14, the blood lead concentration of the dams was similar to that of the PN0 pups; at PN30, the blood lead concentration in the GLE mice was not different from the control mice . There were no statistical differences between the control and GLE groups for any dam measure, litter measure, or bodyweight in the males or females . No age- or sex-dependent differences were found in the analysis; therefore, all the data were combined for presentation herein.
Real-time quantitative PCR (RT-qPCR) studies were conducted using retinas from E16.5, E18.5, PN2, PN6, and PN10 control and GLE mice. E16.5 and E18.5, entire litters were used for control and GLE mice (N=4-5 litters). For the postnatal ages, four to five mice from the control and GLE litters were used. All the supplies and equipment as well as the surgical and laboratory procedures followed RNase-free conditions essentially as described . Briefly, retinas were dissected and immediately placed into TRIzol (Thermo Fisher Scientific, Waltham, MA) and vortexed; chloroform was added and the retinas were vortexed again, then centrifuged. Next, 2-propanol was added, and the retinas were vortexed and centrifuged. The pellet was saved at −80 °C until the RNA was isolated.
Total RNA was synthesized into cDNA as described . Briefly, 1 µg of total RNA was used for first-strand cDNA synthesis and added to oligo dT and random hexamer primers according to Bio-Rad specifications. All RT-qPCR experiments were run on the Bio-Rad iCycler platform (Bio-Rad Laboratories, Hercules, CA). The RT-qPCR primers were designed and validated within the following parameters: GC content = 50%–60%, melt temperature = 55–65 °C, and no secondary structures, primer dimers, or homodimers. A primer quality analysis was conducted using IDT DNA’s OligoAnalyzer. All the primers designed were intron spanning, and they were selected from the Roche Applied Science Universal Probe Library and Assay Design Center database (https://lifescience.roche.com/shop/products/universal-probelibrary-system-assay-design). The primers were tested for alternative sites of homology with NCBI’s BLAST and the UCSC Genome Browser. The primer sequences are shown in Table 1. The primers were validated by a melt-curve analysis for a single peak of fluorescence.
All the RT-qPCR experiments were performed in triplicate using SYBR green (Bio-Rad Laboratories, Hercules, CA) with four to five independent biologic samples per age per treatment. β-Actin was used as the internal control. The PCR mixture consisted of 12.7 µl of Bio-Rad supermix, 1 µl of cDNA template, and 1.5 µl of gene specific forward and reverse primers, all combined in a 0.5 ml PCR tube on a 96-well plate. A no template control, water control, and air [empty well] control were run for each plate. The PCR cycle ran at 95 °C for 3 min, followed by 40 cycles of 95 °C for 30 s and 60 °C for 30 s. A melt-curve analysis was run at the end of each plate to ensure the proper performance of the primer pairs starting at 60 °C increased by 0.5 °C per min to 95 °C. The threshold for the Ct values was manually adjusted for each plate to coincide with entrance into the exponential growth phase of the PCR. The Ct values were then exported to Microsoft Excel for further analysis. The level of gene expression relative to the controls was determined using the ∆∆Ct method : ∆∆Ct = ∆Ct of the gene of interest - ∆Ct of β-actin. 2∆∆Ct was used to compare the fold change in the gene expression between the GLE and control retinas.
All tissue processing and immunohistochemistry techniques were performed as described [19,21,33]. Briefly, developing or adult mice were decapitated, and their eyes were quickly enucleated and placed in ice-cold phosphate-buffered saline (PBS) (144 mM NaCl, 7.7 mM Na2HPO4, 2.7 mM NaH2PO4-H2O, 305 mOsm, pH 7.40), in which the corneas were gently slit. The eyes were immersion fixed at room temperature, buffered with 4% paraformaldehyde (Ladd Research, Williston VT) for 30 min, washed in PBS 3X for 10 min, and cryoprotected in 30% w/v sucrose for 48 h. The anterior segments were removed, and the eyes were embedded in Tissue-Tek optimum cutting temperature (OCT) Compound (Electron Microscopy Services, Fort Washington, PA), flash frozen in liquid nitrogen, and stored at −80 °C. Three nonadjacent fixed and frozen transverse sections (10 μm) were taken from the vertical meridians of four to nine independent biological samples per age per treatment. The retinal sections were mounted on Superfrost Plus glass slides (Thermo Fisher Scientific, San Francisco, CA) and stored at −20 °C.
All sections were taken from the central retina 200–400 μm from the optic nerve. The slides were thawed slowly, and the sections were post-fixed with 4% paraformaldehyde. They were washed and briefly immersed in a 1% sodium-borohydride (Sigma-Aldrich, St. Louis, MO) solution to reduce double bonds, to allow for proper epitope access for antibodies, and to decrease auto-fluorescence. The sections were washed and incubated for 2 h with a blocking buffer containing 10% normal goat serum (Jackson ImmunoResearch Labs Inc., West Grove, PA) and 0.3% Triton-X100 in PBS. Primary antibodies were applied for 2 days at 4 °C, after which the slides were washed with PBS and blocked for 30 min; then, secondary antibodies were applied. Table 2 lists the well-characterized and commercially available primary antibodies that were used. For the double-labeling experiments, primary antibodies from different host animals were applied simultaneously. Dilutions of Alexa 488, 555, and 647 (1:400) conjugated secondary antibodies were made in a blocking buffer and applied to the slides, which were left to incubate in the dark at room temperature for 1 h. The slides were washed with PBS, double-distilled H2O, dried, and mounted with Vectashield Mounting Medium (Vector Laboratories Inc., Burlinghame, CA) and a No. 1 coverslip; then, they were stored at 4 °C. For the DAPI-labeling experiments, the sections were incubated with DAPI for 1 h in the dark after the secondary incubation. All antibodies were titrated through a broad range of working dilutions (with three orders of magnitude) to determine the optimal working dilutions. Immunolabeling specificity was confirmed by processing the retinal sections in the absence of a primary antibody and/or with blocking antibodies. Outer nuclear layer (ONL) thickness was measured from slides using a calibrated Filar micrometer eyepiece at a total magnification of 1000x (Reichert Scientific Instruments, Buffalo, NY) . The ONL thickness from three nonadjacent coded retinal sections from control and GLE mice were analyzed (N=4-5 litters).
Immunoblotting for rhodopsin was done as described [21,29]. The blots were probed with anti-rhodopsin and -glyceraldehyde 3-phosphate dehydrogenase (GAPDH) antibodies (Table 2), followed by incubation with an HRP-conjugated secondary antibody. The blots were visualized using enhanced chemiluminescence (ECL) Plus (Thermo Fisher Scientific). Densitometry measurements were obtained from non-saturated blots using NIH-Image J. Four mice from different litters per age per treatment group were used.
For all data, animals from the same litter represented only one data point. The data are presented as the mean ± Standard Error of the Mean (SEM) for four to nine animals per treatment group. Data were analyzed using a two-way ANOVA (treatment by age) followed by a post-hoc analysis using the Fisher’s Least Significant Difference Test or the Student t test when only two means were compared. Values of p < 0.05 were considered significantly different from the controls and were noted in the figures by asterisks where appropriate. In the text, values of p < 0.05 were noted as significantly different from the controls. IHC images were compiled and presented using Adobe Photoshop CS (Adobe Systems Inc., Mountain View, CA). Graphs were generated using KaleidaGraph (Synergy Software, Reading, PA).
Figure 1 shows the relative gene expression of the rod-specific transcription factors Nrl (Figure 1A), Nr2e3 (Figure 1B), Crx (Figure 1C) and the rod-specific functional gene Rho (Figure 1D) from E16.5 to PN10. The RT-qPCR data were normalized with β-actin so as not to reflect the overall increased number of cells associated with the GLE model . In both the control and GLE retinas, the developmental gene expression pattern was similar for all four genes (i.e., a continuous increase from E18.5 to PN10). In the GLE retinas, all four of the rod-associated genes decreased significantly at PN2. In addition, Nrl expression decreased significantly at PN6. These data reveal an overall decrease in rod-specific gene expression in the GLE retinas at PN2, suggesting a possible delay in development.
To determine the functional development of rod photoreceptors, we analyzed the spatiotemporal appearance of the photoreceptor specific proteins rhodopsin and recoverin by immunohistochemistry. Rhodopsin is a rod-specific photopigment that initiates the phototransduction cascade in the presence of light [35,36]. Recoverin, a calcium-sensing protein present in all photoreceptors, regulates phototransduction by deactivating rhodopsin kinase [37,38]. Rhodopsin and recoverin protein expression were determined in the control and GLE retinas at E16.5, E18.5, PN1, PN3, PN5, PN7, PN10, and PN60 (Figure 2 and Figure 3). Recoverin-immunoreactive (IR) cells represent cone photoreceptors, while rods were double labeled with recoverin and rhodopsin. In the control retinas, rhodopsin-IR cell somas were first observed in the developing ONL at PN1 (Figure 2C). This is consistent with the published data for rats . Rhodopsin expression increased throughout development, consistent with the increased differentiation of the RPCs into rods . By PN10 in control retinas (Figure 3C), rhodopsin-IR was seen in the outer segments of rod photoreceptors. In the GLE retinas, rhodopsin-IR cells were observed first in the ventricular zone at PN3, two days after controls (Figure 2P; Table 3). Rhodopsin expression in the GLE retinas increased throughout development, and the ONLs of the GLE retinas were consistently thicker than the age-matched controls beginning at PN5 (compare Figure 3A-M). Moreover, at PN10, rhodopsin-IR in the outer segments was not seen in the GLE retinas (Figure 3O) as it was in the controls (Figure 3C). In contrast, the onset of and developmental increase in recoverin-IR (using an antibody at a low concentration of 1:1500 and with a short fixation time to selectively label photoreceptors ) was similar in both the control and GLE retinas. Recoverin-IR cells were first observed at E18.5 in the control and GLE mice (Figure 2B,N), consistent with the published data in C57BL/6 mice . Although a few double-labeled recoverin-IR and rhodopsin-IR photoreceptors were seen in the PN1 control retinas (Figure 2K), these were not observed until PN3 in the GLE retinas (Figure 2X). We were unable to detect M or S cone opsin-IR cells at the embryonic and early postnatal ages tested. This is in harmony with what is known about M and S cone opsin expression, which does not occur until PN5 . Since the double-labeled rhodopsin- and recoverin-IR rods did not appear until PN1 and PN3 in the control and GLE retinas, respectively, and since recoverin-IR cells are observed in both the control and GLE retinas at E18.5, we take this to possibly suggest that the cones differentiate at similar time points. Our Affymetrix data (Fox et al, unpublished) from E16.5 to PN10 for the control and GLE mice show that the gene expression of Opn1sw (S cones) and Opnmw1 (M cones) were not significantly different in the control and GLE retinas at any age. Together, these data indicate that the rods were functionally delayed by approximately two days in the GLE retinas.
To more precisely determine the timing of the delay in the appearance of rhodopsin-IR in the GLE retinas, we conducted western blot experiments using PN1, PN2, and PN3 control and GLE retinas. Figure 4 shows that rhodopsin was detectable in whole homogenized control retinas at PN1 and PN2, consistent with published data . Rhodopsin was detected at PN1 and PN2 in the GLE retinas but significantly reduced by ≥ 20%. By PN3, the rhodopsin concentration was ≥ 10% higher in GLE than in the control retinas. The rhodopsin concentration increased approximately 40% between PN1 and PN2, and between PN2 and PN3 in the control and GLE mice. Together, these immunohistochemistry and immunoblot data indicate that there is a two-day delay in relative rhodopsin expression in GLE mice retinas.
Figure 5 reveals that the developmental profile of GS protein expression was similar in the control (Figure 5A-E) and GLE (Figure 5F-J) retinas throughout development. Figure 5K shows that GS gene expression was similar and low from E16.5 to PN6, then significantly increased by PN10 in the control and GLE retinas. These GS results are consistent with the birth-dating and adult retinal immunohistochemistry data for GSs, which shows that GLE did not alter the initiation or pattern of gliogenesis or the number of Müller glial cells .
Figure 6 shows the relative gene expression of the rod and bipolar proneural genes Otx2 (Figure 6A) and Chx10 (Vsx2; Figure 6B) in developing retinas. In the control retinas, Otx2 expression increased continually from E18.5 to PN6 and peaked at PN6, when most late-born retinal neurons were differentiated . This is consistent with in situ hybridization data for similar ages . The pattern of Otx2 expression in GLE retinas followed that in the control retinas; however, gene expression in the GLE retinas increased significantly at PN2 and PN6. In the developing control retinas, Chx10 expression declined from E16.5 to PN2 and increased from PN2 to PN10. In contrast, in the GLE retinas, a decline in Chx10 expression was never observed; rather, the pattern of expression continually increased, and PN2 and PN6 expression levels were significantly increased relative to those in the controls. Thus, the Otx2 and Chx10 expression patterns in PN2 to PN6 GLE retinas were consistent with the increased number of rods and BCs observed in the GLE retinas .
To examine the differentiation of rod BCs, we analyzed the spatiotemporal protein expression of Chx10 and PKCα in PN3, PN5, PN7, PN10, and PN60 control and GLE retinas (Figure 7). We also colabeled Chx10 with CCND1 at PN1 and PN5 in the control retinas to illustrate the difference between RPC-associated and BC-associated Chx10 (Figure 8). Chx10 is a homeobox domain transcription factor that is initially expressed in RPCs during the earliest stages of neuroretinal development . Chx10 expression persists in RPCs until the early postnatal ages, after which its expression permanently localizes to rod and cone BCs [46,47]. Immunohistochemically, this switch from neuronal- to bipolar-associated Chx10-IR can be determined by the appearance of the Chx10 antibody label in BC somas. In the PN3 control retinas, we observed Chx10-IR in the inner segments (ISs), inner nuclear layer (INL), and ganglion cell layer (GCL) (Figure 7A). In the INL, cytoplasmic and nuclear labeling was seen in RPCs and immature BCs, respectively. Furthermore, the strongly labeled Chx10-IR cells localized only to the center of the INL. The appearance of Chx10-IR BCs at PN3 correlates with the peak of BC differentiation [21,48]. In the PN5 controls, Chx10-IR was restricted to BCs and the GCL, and the diffuse label associated with RPCs was no longer present (Figure 7B). Figure 8 illustrates the difference between RPC-associated and BC-associated Chx10 in the control retinas. At PN1, Chx10-IR and CCND1-IR colabeled (Figure 8C), whereas at PN5, they did not (Figure 8F). In the PN7 and PN10 controls (Figure 7C,D, respectively), the appearance of Chx10-IR BCs was similar to that at PN60 (Figure 7E ). In the PN3 GLE retinas, Chx10 labeled the ISs, INL, and GCL (Figure 7K). However, the strongly observed Chx10-IR BCs in the controls were not present in the GLE retinas, and Chx10-IR in the INL was restricted to proliferative cells. However, the Chx10-IR in the ‘inner plexiform layer (IPL) and GCL of the control and GLE retinas was similar at all ages (Figure 7A-E,K-O), confirming that the effect of GLE was selective for late-born neurons. In the PN5 GLE retinas, immature BCs were strongly IR for Chx10 as in the age-matched controls; however, the cells appeared as spindles and were elongated (Figure 7L; Table 3). By PN7, the appearance of Chx10-IR BCs in the GLE retinas was similar to that in the control retinas (Figure 7 M). At PN10 and PN60, an increased number of Chx10-IR BCs consistent with the GLE phenotype was observed (Figure 7N,O, respectively) .
Figure 7 also shows the spatiotemporal profile of PKCα in relation to Chx10-IR. PKCα, a calcium-activated serine/threonine protein kinase, selectively labels the soma, dendrites, axons, and synaptic terminals of adult rod BCs in mouse retinas . In the PN3 control retinas, PKCα minimally labeled the INL and strongly labeled the developing ISs, IPL, and GCL (Figure 7F). In the PN5 control retinas, PKCα-IR increased in the ISs, IPL, and GCL and appeared in the OPL (Figure 7G). By PN7, PKCα-IR was restricted to rod BC somas, dendrites, axons, and bipolar cell terminals (Figure 7H). For comparison to the GLE retinas, the adult phenotype of control PKCα-IR BCs is shown at PN60 (Figure 7J). In the PN3 GLE retinas, PKCα-IR was less intense and more diffuse than in the age-matched controls (Figure 7P). The GLE retinas were similar to the controls at PN5 and PN7 (Figures 7Q,R, respectively), although the OPL was less developed and fewer PKCα-IR rod BCs were present. Similar to Chx10-IR, PKCα-IR in the IPL and GCL of the PN3 and PN5 control and GLE retinas was similar, again showing that the effect of GLE was selective for late-born neurons. At PN10 and PN60 in the GLE retinas (Figure 7S,T, respectively), there was an increased number of PKCα-IR rod BCs relative to the controls (Figure 7I,J, respectively).
To examine the differentiation of the cone BCs, we analyzed the spatiotemporal protein expression of secretagogin (SCGN) in the control and GLE retinas (Figure 9). SCGN is an EF-hand calcium-binding protein that localizes to subpopulations of ON and OFF cone BCs, but not rod BCs, in mammalian retinas . Figure 9 shows the spatiotemporal SCGN-IR for PN5, PN7, PN10, and PN60 retinas in control (Figure 9A-L) and GLE mice (Figure 9M-X) in relation to PKCα-IR, the adult rod BC marker . In the control mice, SCGN-IR cone BCs were first observed in the INL at PN5 (Figure 9A). These SCGN-IR cells were elongated, and their projections extended distally to the IS and proximally toward the GCL. By PN7, most of the SCGN-IR was restricted to the somas of the INL, dendritic processes of the OPL, and axon terminals in IPL sublamina a (IPL-a) and b (IPL-b; Figure 9B). SCGN labeled more intensely in IPL-a (the OFF layer) than in IPL-b (the ON layer). In the PN10 controls, the pattern of SCGN-IR in the INL changed from moderately labeled linear projections to intensely labeled soma. Furthermore, the dendrites and axon terminals of these SCGN-IR cone BCs exhibited punctate and well-defined labeling in the OPL and IPL, respectively (Figure 9C). In the PN10 controls, the appearance of SCGN-IR somas, dendrites, and axon terminals closely resembled the appearance seen at PN60 (Figure 9C,D, respectively). These results are consistent with the images . The control retinas double labeled for PKCα and SCGN showed no colocalization at any age (Figure 9I-L).
In the GLE retinas, SCGN expression was delayed at PN5 (Figure 9M; Table 3). By PN7, however, there was an increase in the number of SCGN-IR cone BCs when compared to the age-matched controls (Figure 9N). Interestingly, the SCGN-IR pattern and intensity in PN10 GLE retinas (Figure 9O) appeared more similar to the PN7 than the PN60 retinas, unlike that seen in the controls. That is, the punctate dendritic and axon terminals observed in the PN10 control retinas were not present in the age-matched GLE retinas. Together, these data indicate that BC differentiation and synaptic development are delayed two to three days in GLE retinas.
Since the initial expression of rod-specific genes (Figure 1) and proteins (Figure 2 and Figure 4) was delayed, we examined the expression of proteins associated with the functional synapse in the OPL of rod photoreceptors to determine if the entire rod was similarly and coordinately affected. Figure 10A-J shows the spatiotemporal expression of VGluT1 in the control and GLE retinas at ages PN3, PN5, PN7, PN10, and PN60. In the control retinas, sporadic VGluT1-IR puncta were first observed below the ventricular zone at PN3 (Figure 10A). At PN5, VGluT1 labeling in the control retinas localized to the developing OPL (Figure 10B). In the control retinas, VGluT1-IR increased in the OPL at both PN7 and PN10 as the rods developed and matured (Figure 10C,D, respectively) toward the adult phenotype (Figure 10E). In the GLE retinas, VGLUT-IR puncta were first observed in the OPL at PN3 (Figure 10F), similar to the controls. However, in the PN5 GLE retinas, the extent of VGluT1-IR in the OPL decreased (Figure 10G; Table 3). In the PN7 GLE retinas, VGluT1-IR in the OPL decreased; although it was detected in the PN7 retinas, it still decreased (Figure 10H; Table 3). By PN10, VGluT1-IR in the OPL of GLE was greater than that in the controls (Figure 10I; Figure 10D). In the GLE retinas at PN60, VGluT1-IR in the OPL and IPL (Figure 10J) was greater than in the controls, as reported . Thus, GLE caused a two to three day in the appearance and development of the age-matched extent and intensity of VGluT1-IR in the OPL.
Neurons use four different plasma membrane calcium ATPase (PMCA) isoforms to return to or maintain the basal levels of intracellular calcium after the stimulation of voltage-operated calcium channels and the subsequent release of neurotransmitters [33,50]. In the OPL of adult mammalian retinas, a commercially available pan-PMCA antibody (Table 2) is able to recognize all four isoforms and label both rod spherules and the apical portion of cone pedicles; however, intense and widespread PMCA-IR is associated with rod spherules . Figure 10K-T shows the labeling profile of PMCA-IR at ages PN3, PN5, PN7, PN10, and PN60 in the control and GLE retinas. In the PN3 control retina (Figure 10K), but not the GLE retina (Figure 10P), the PMCA-IR was in the ISs, supporting the delay in rhodopsin appearance. The PMCA-IR in the IPL and GCL of PN3 was similar in the control and GLE retinas, confirming that the effect of GLE was selective for late-born neurons. At PN5, the controls exhibited strong PMCA-IR, which was first observed in the OPL (Figure 10L). In the PN5 GLE retinas, the extent and intensity of PMCA–IR in the OPL decreased (Figure 10Q; Table 3). At PN7 and PN10, the pattern and intensity of PMCA-IR appeared similar in the control (Figure 10M,N) and GLE retinas (Figure 10R,S). However, at PN60, the increased number of rod synapses in GLE retinas was evidenced by a higher number of PMCA-IR spherules (Figure 10T) relative to the controls (Figure 10O).
To further examine the development of the proteins associated with glutamatergic synaptic transmission in photoreceptors, we analyzed the spatiotemporal profile of the synaptic vesicle protein 2B (SV2B) in the control and GLE retinas at PN3, PN5, PN7, and PN10 (Figure 11). SV2B is one of three SV2 isoforms (SV2A, SV2B, and SV2C) present on presynaptic vesicles that regulates excitatory calcium-dependent vesicle release . In mouse retinas, SV2B is expressed in the OPL and the IPL [52,53] and regulates rod photoreceptor neurotransmitter release . In the control retinas, SV2B exhibited an age-dependent increase in extent and intensity from PN3 to PN10 (Figure 11A-D). It was detectable in the OPL at PN3 (Figure 11A). In the OPLs of GLE mice, SV2B expression was delayed and decreased such that it was barely visible at PN3 and PN5 (Figure 11M,N, respectively; Table 3). At PN7, the SV2B-IR in the GLE retinas (Figure 11O) appeared similar to that of the PN5 controls (Figure 11B). At PN10, the SV2B-IR pattern and intensity were similar in the GLE (Figure 11P) and control (Figure 11D) retinas.
The structural development of BC dendrites in the control and GLE retinas was observed using an antibody against the transient receptor potential melastatin 1-long isoform (TRPM1-L) protein (Figure 11). In the retina, TRPM1 is required in the light response of ON BCs; in PN14, PN21, and adult retinas, it characteristically labels ON BC dendrites, somas, and axons in the IPL and in retinal ganglion cells [54-56]. In wild-type control mice, Trpm1 gene expression is relatively low from PN2 to PN8, increases slightly at PN10 and PN14, and then markedly increases at PN21 to a steady-state level seen at PN90 . In the control retinas, TRPM1-IR first appeared in the OPL at PN5 (Figure 11F) and increased its pattern and intensity of expression at PN7 (Figure 11G) and PN10 (Figure 11H). Figures 11E-H also reveals that TRPM1-IR was observed in the photoreceptor ISs, IPL, and GCL throughout early development. In the GLE retinas, TRPM1-L-IR first appeared in the OPL at PN7 (Figure 11S), although its intensity was less than that in the age-matched controls (Table 3). The pattern in PN10 GLE retinas (Figure 11T) was not markedly different than that at in the PN7 GLE retinas. These results reveal that the maturation of the ON bipolar dendrites in the GLE retinas was delayed and decreased.
To determine the structural development of BC axonal synapses, we analyzed the patterns of VGluT1 (Figure 10A-J) and SV2B (Figure 11A-D, M-P) in the IPL. In the control retinas, VGluT1 was first visible in the IPL at PN5 (Figure 10B) and exhibited an age-dependent increase in extent and intensity in IPL-a and IPL-b, consistent with the published data . In the GLE retinas, the initial VGluT1 expression in the IPL was delayed until PN7 (Figure 10H; Table 3). Moreover, and in contrast to the PN10 controls (Figure 10D), VGluT1 was not expressed in the IPL-b of GLE retinas (Figure 10I), and it was lower in IPL-a. By PN60, VGluT1-IR increased in the IPL of the GLE retina (Figure 10J) relative to the controls (Figure 10E). In the controls, SV2B was first visible in the IPL at PN5 (Figure 11B) and exhibited an age-dependent increase in extent and intensity in IPL-a and IPL-b. IPL-b developed before IPL-a, consistent with the published data . In the GLE retinas, the initial SV2B expression in the IPL was delayed until PN7 (Figure 11O; Table 3). At PN10, SV2B-IR was similar in the control (Figure 11D) and GLE (Figure 11P) retinas.
The overall goal of this study was to examine the differentiation and development of rods and BCs, which increased in the adult mouse and rat GLE retinas following increased and prolonged RPC proliferation [19,21]. There were six main findings in this study. First, GLE delayed the differentiation of rod photoreceptors, as indicated by the developmental profiles of several photoreceptor-specific genes and proteins, the latter via immunocytochemistry and immunoblot studies. Second, the differentiation of rod and cone BCs was delayed, as indicated by the delayed spatiotemporal patterns of Chx10, PKCα, and SCGN protein expression. Third, GLE delayed the expression of proteins and SV2B that are associated with functional synaptic development in rod photoreceptors, such as PMCA, VGluT1. Fourth, GLE delayed the expression of several essential proteins that are associated with the dendritic and synaptic structure and function of ON and OFF BC subtypes, such as VGluT1, SV2B and TRPM1-L. Fifth, there were no treatment-related changes in the gene or protein expression of GS associated with late-born Müller glial cells. Sixth, there were no treatment-related changes in the protein expression of early-born cells in the IPL or GCL. Together, these data reveal that the development of late-born neurons, but not Müller glial cells, is delayed in GLE retinas by two to three days. We suggest that this delay in differentiation is a direct downstream consequence of an increase in and prolonged period of RPC proliferation. Moreover, the delay in the markers of functional synaptic development in rods (OPL) and BCs (OPL and IPL) reveals that the coordinated differentiation and maturation of neuronal processes are linked with cellular differentiation and are therefore similarly delayed.
These results were unexpected for three reasons. First, because we observed the increased proliferation of RPCs in GLE retinas , we expected to see the accelerated and/or increased, not delayed, differentiation of rods, as seen in developing retinas exposed to different mitogenic soluble factors [reviewed in 15,57] or brains exposed to drugs . Second, our birth-dating experiments revealed an increased number of rods and BCs that underwent terminal mitosis during postnatal development without a phase shift in the timing . Third, most delays in retinal and brain development associated with chemical or toxicant exposure (e.g., ethanol, lead, polychlorinated biphenyls, etc.) are the result of decreased neurogenesis or increased apoptosis [26,59-61]. Based on our findings, as well as on the previously published work in wild-type mice retinas described below, we conclude that RPC terminal mitosis and neuronal differentiation are separate spatiotemporal events under both normal and pathophysiological conditions. For example, in vivo and in vitro studies demonstrate that the early-born rod photoreceptors possess an intrinsically determined ability to postpone rhodopsin expression after terminal mitosis [40,62]. That is, in developing rat retinas, there are two phases of rod differentiation: 1) an early phase, where rods born before E19 exhibit a delay of 8.5–12.5 days before rhodopsin expression and 2) a late phase, where rods born after E19 exhibit a delay of only 5.5–6.5 days before rhodopsin expression . Although the postnatal microenvironment cannot prematurely induce early-born rods to express rhodopsin [40,62], it is unknown whether early-born rods can further delay their expression of rhodopsin. Since the initial delay in rhodopsin expression occurred at PN2 in GLE retinas, it is likely that GLE affected the early cohort of rod photoreceptors, suggesting that the terminal commitment to rod differentiation in this early cohort of rod photoreceptors is plastic and can be delayed further.
The cellular mechanisms underlying the delay in rod and BC commitment are unknown. However, the delay in rod differentiation seen in GLE retinas is consistent with studies showing that various soluble neurotropic ligands can alter the commitment of post-mitotic late-born retinal neurons to rod differentiation. For example, in vivo and in vitro experiments show that the ciliary neurotrophic factor can delay rod differentiation and rhodopsin expression for 3–4 days [63-65]. Similarly, the Müller glial cell-derived leukemia-induced factor (LIF) can temporarily inhibit rod differentiation [63,64,66]. Ex vivo and in vitro experiments also show that mitogen retinoic acid can increase proliferation and delay rod differentiation [57,67]. Our Affymetrix data (Fox et al, unpublished) from E16.5 to PN10 control and GLE mice shows that the gene expression of the LIF receptor Lifr and its transducing subunit gp130 are transiently upregulated by 28% at E16.5 and E18.5 in the GLE retinas relative to the controls, indicating one possible mechanism of action for delay. To date, one observational report notes a delay in BC development. Wu and Chiao  demonstrated that dark-rearing rabbits delayed the morphological maturation of BCs by two to three days, as characterized by histology and confocal microscopy.
Although we found a decrease in rod-specific genes at PN2 in the GLE retinas, both Otx2 and Chx10 gene expression increased at this age. The Otx2 gene regulates rod photoreceptor cell fate determination, and the conditional knockout of Otx2 causes post-mitotic rod precursors to become amacrine-like cells [69,70]. The upregulation of the Otx2 gene in the retinal progenitor cell causes the cell to become a postmitotic rod or BC precursor . Otx2 gene expression remains elevated in BCs but decreases in rods . Therefore, the increased expression in Otx2 observed in GLE mouse retinas is likely associated with an increased number of post-mitotic rod and BC precursors at PN2.
Chx10 is also duplicitous in its expression profile. The pattern of gene expression we observed is consistent with the idea that Chx10 protein expression associates with all proliferative cells during the embryonic and early postnatal stages of retinal development and becomes specific for BCs, as their numbers increase during late postnatal ages [71,72] (Figure 5 B). Therefore, an increase in Chx10 gene expression at PN2 in GLE retinas is likely a consequence of the increase in retinal progenitor cells at this age. Interestingly, the control retinas exhibited a decreased expression in the Chx10 gene at PN2 when compared to other ages. This was not observed in GLE retinas. We postulate that this represents a general switch from a RPC-dominated retina to a microenvironment conducive for differentiation and functional development.
Finally, we showed that GLE decreased the expression of several functional proteins associated with synaptic function in both the OPL and the IPL. The photoreceptor ribbon synapse is a complex structure that is specialized for fast and sustained neurotransmitter release . Critical proteins associated with proper photoreceptor synaptic function include calcium channels, ion pumps, cytoskeletal scaffolds, presynaptic endocytotic and exocytotic proteins, and postsynaptic receptors . Specifically, we assessed the proteins involved in neurotransmitter uptake into VGluT1, PMCA, SV2B, and TRPM1 rather than into the scaffolding proteins structurally associated with the synaptic ribbon . We showed that, although both the GLE and control animals initially expressed VGluT1 at PN3, GLE animals exhibited the decreased expression of VGluT1 in the developing OPL at PN7. Furthermore, PMCA-IR decreased in the GLE retinas at PN5. At PN3, VGluT1-IR in the OPL is indicative of cone synapses forming their first glutamatergic connections with horizontal cells, while the rod photoreceptors first form connections around PN8 in mice (Figure 2) . Therefore, the pattern of the VGluT1 label in the OPL of the developing GLE retinas indicates that the cone photoreceptor terminals developed normally, whereas rod terminals were delayed in their ability to package glutamate into the vesicles. This was confirmed by the decrease in PMCA labeling observed in the OPL at PN5. PMCA localizes to the photoreceptor terminals during the first postnatal week in rats . In adult mouse retinas, most PMCA is associated with rod spherules rather than cone pedicles [33,55]. Therefore, limited PMCA-IR in the OPL of PN5 GLE retinas may be a consequence of immature rod synapses. Interestingly, this delay is not seen at PN7 as with VGluT1 labeling. The fact that the delay in the PMCA label occurs two days before the major delay in the OPL VGluT1 label suggests that, in the normal maturation of the rod synapse, the mechanisms for calcium extrusion develop before those of the glutamate packaging. Therefore, the photoreceptor terminal primarily establishes a way to maintain calcium homeostasis before glutamate release is possible.
In summary, the novel GLE model enabled us to more precisely decipher and understand the coordinated and time-dependent patterns of RPC proliferation, differentiation, and maturation. Although the GLE retinas eventually caught up to the controls, the long-term retinal effects as well as the effects on central visual connectivity and processing are unknown. Whether similar changes occur in the different populations of brain cells in GLE mice is also unknown. Although the clinical and public health relevance of our findings in GLE mice has yet to be determined, neurodevelopmental delays and altered synaptic development in the brain development of children are associated with various neuropsychiatric and behavioral disorders as well as cognitive deficits [77-81].
We thank Dr. Laura J. Frishman for her support throughout this project, Dr. Weimin Xiao for his technical assistance, Dr. Roger Janz for the SV2B antibody, Dr. Anand Swaroop for help and support with the Affymetrix study, and Dr. David R. Copenhagen for his suggestions and comments on an earlier draft of this manuscript. This work was partially supported by NIH Grants RO1ES012482 to DAF, P30EY07551, T32EY07024 for SYC, an UH AGEP Grant to SYC, and an UHD ORC Grant 206400004B0303 to JEJ.