Loss of MeCP2 causes subtle alteration in dendritic arborization of retinal ganglion cells

ABSTRACT Methyl-CpG-binding protein (MeCP2) is highly expressed in neurons. It plays an important role in the development of synapses and the formation of circuits in the central nervous system (CNS). Mutations in MECP2 cause neurodevelopmental disorders and mental retardation in humans. Therefore, it has become important to determine the distribution and function of MeCP2 in vivo. The retina consists of three nuclear cell layers and two layers of synapses; neurons in each layer are connected to form fine circuits necessary for visual signal transduction. Using immunohistochemical analysis, we found that MeCP2 was expressed in all nuclear cell layers, with differences in the levels of MeCP2 expression observed among the layers. To understand the structural defects in the retina due to the loss of MeCP2, we sought to elucidate the organization of the retinal structure in the Mecp2 knockout (KO) mouse. Overall, we found a normal retinal structure in Mecp2 KO mice. However, because Mecp2 mutations have a highly variable effect on neuronal architecture, we analyzed morphological changes in a subset of retinal ganglion cells of Mecp2 KO mice. In Thy1-GFP mice crossed with Mecp2 mutant mice, Sholl intersections analyses showed a subtle increase in number of intersections due to increased branching proximal to the soma in Mecp2 KO mice. Our results demonstrate that the expression of MeCP2 and the effects of Mecp2 mutations are highly specific to tissue and cell types.


Introduction
MeCP2 is a methyl-CpG binding protein capable of suppressing gene expression by recruiting co-repressor complexes to different gene expression regulatory regions (Nan et al. 1996). Recent studies demonstrate that MeCP2 plays a role in regulating multiple transcriptional, post-transcriptional, and post-translational processes (Chen et al. 2003;Chahrour et al. 2008;Li et al. 2013;Lee et al. 2020). Mutations in MECP2 cause the neurodevelopmental disorder Rett syndrome (RTT) in humans (Bienvenu et al. 2000). The representative RTT brain has been characterized as arrested at early developmental stages, with highly variable abnormalities (Christodoulou 2001;Zappella et al. 2003). Neuroanatomical studies in the Mecp2 mutant mouse reveal changes in neuron maturation as well as in axonal and dendritic morphology, axonal mistargeting, poor synaptic formation, and misregulation of neurotransmission (Kishi and Macklis 2004;Smrt et al. 2007). Overall, these studies demonstrate that a loss of MeCP2 causes defects in the formation and maintenance of neuronal circuits.
Even though it is evident that Mecp2 mutations cause abnormality in the nervous system, it is hard to characterize the cellular specific phenotype due to the complexity of in vivo tissue. For these reasons, comparative analysis from Mecp2 mutations has revealed both overlapping and divergent effects in various cell types, including dendritic and axonal morphology, and dendritic spine density/spine morphology (Stuss et al. 2012;Lee et al. 2014). Since Mecp2 mutations could contain any of these abnormalities in specific cell types, the use of precisely defined cellular subtypes is required in comparative analysis.
The vertebrate retina is highly organized, with multiple layers of nerve cell bodies and layers of synapses. The outer nuclear layer (ONL) is enriched with rod and cone cell bodies; the inner nuclear layer (INL) contains bipolar, horizontal, and amacrine cell bodies; and the ganglion cell layer (GCL) contains ganglion cell bodies and displaced amacrine cell bodies (Tian and Copenhagen 2003). In this study, we used the retina as a model to study neuronal circuit formation in a Mecp2 mutation mouse because of the enriched neuron cell types and fine neuronal connections present.
In contrast to previous reports (Jain et al. 2010), we found that all of the retina cell layers express MeCP2, but at different levels in different layers. Upon characterization of the Mecp2 KO mouse at the gross anatomy level, we found normal-sized eyes and a well-organized retinal structure. For further detailed assessment of the morphological changes in retinal ganglion cells (RGC), we crossed Thy1-GFP mice, which express GFP under the control of the Thy-1 promoter, with Mecp2 KO mice. The data demonstrated that loss of MeCP2 results in subtle increases in dendritic branch density in RGCs. Overall, the result supports the view that the expression of MeCP2 and the effects of MeCP2 null are highly context-dependent in vivo.

Materials and methods
Animals: Female heterozygotes with an MeCP2 tm1.1Bird mutation (Guy et al. 2001) were maintained on a 129/ SvEv background (Tacoma). The MeCP2 tm1.1Bird were crossed with mice expressing GFP (Tg(Thy1-EGFP)MJrs/ J) (Feng et al. 2000) in a subset of RGCs and bipolar cells. All mice were housed in an animal facility approved by the University of California Animal Care and Use Committee. All procedures described here are in accordance with the NIH animal use guidelines and institutionapproved animal use protocols.
Preparation of Retina for Immunostaining: Mice were euthanized using CO 2 and enucleated eyes were chilled in dry ice-cooled propane for 1 min. The eyes were transferred into vials containing 97% MeOH + 3% acetic acid and stored at −80°C for 48hr. The eyes were then transferred to −20°C and incubated for 24hr. Following this, the eyes were placed in 100% EtOH 30-60 min and embedded in paraffin.
Preparation of Retinas for GFP imaging: Eyes were enucleated and hemisected following euthanization of the mice. The cornea, iris, lens, and vitreous humor were removed from the eyes and immersed in PBS. The excised retinas were immersed in 4% PFA in PBS for 30 min at room temperature. The retinas were then stored overnight at 4°C in 30% sucrose in PBS. A rabbit GFP antibody (1:500, Molecular Probes) was used to enhance the GFP signal. Fixed retinas were washed in PBS, incubated for 24hr in primary antibodies with 0.1% Triton X-100 at 4°C, and washed again in PBS. Retinas were then incubated in secondary antibody (1:50) with 0.1% Triton X-100, flattened and mounted on Super-Frost Plus slides (Fisher).
Sholl analysis of RGC dendrites: Using the Bonfire program (Langhammer et al. 2010), the Sholl analyses were performed. For statistical analysis of the dendritic morphological change, a t-test between the WT and Mecp2 KO group for all RG A was performed using Graph-Pad InStat3.
Cell Nuclei Analysis: Toluidine blue-Staining of nuclei was performed in the retina to quantify differences in the thickness of nuclear layers. Images in Figure 3(A-F) were acquired with a microscope (Nikon, eclipse E800). Brightness and contrast were only slightly adjusted using Adobe Photoshop CS5. The thickness of cell nuclei in WT in Mecp2 KO was analyzed by t-test using GraphPad InStat3.

MeCP2 is differentially expressed and phosphorylated in different cell types in the retina
We performed localization studies of MeCP2 in the retina of postnatal WT mouse. Sagittal slices through the retina of 4-week-old male mice were stained for MeCP2 ( Figure  1(A)), with a DAPI counterstain (Figure 1(C)) to visualize heterochromatin. MeCP2 was expressed exclusively within the nucleus of all cells. However, different levels of MeCP2 expression were observed between the ONL, INL, and GCL. In the ONL, DAPI (Figure 1(C)) staining showed a highly condensed pattern in the center of the photoreceptor cell nucleus. However, MeCP2 staining showed a relatively low and diffuse expression pattern (Figure 1(A)). In the INL, MeCP2 is highly co-localized with DAPI-positive heterochromatic foci. But, the levels of MeCP2 expression differed when comparing the upper cell layer (arrowhead in Figure 1(A)) and lower cell layer (arrow in Figure 1(A)) of the INL. To enable more detailed analysis, we used double immunolabeling for MeCP2 and Pax6 (Figure 1(B)). Strong Pax6 immunoreactivity was present in a subpopulation of the INL, amacrine, and horizontal cells (Hatakeyama and Kageyama 2004). The co-staining showed that most Pax6 + cells (arrow in Figure 1(B)) express significantly stronger levels of MeCP2 than Pax6cells (arrowhead in Figure 1(B)) at the INL (Figure 1(G)). The results suggest that MeCP2 has different levels of expression in different cell types within the INL. Immunostaining of MeCP2 and DNA (DAPI) in the GCL confirms the colocalization of MeCP2 and DAPI in neuronal nuclei ( Figure 1(A,C)). The speckled nuclear staining of MeCP2 coincides with DAPI foci, but dispersed nuclear MeCP2 staining is reproducibly seen in a subpopulation of RGCs (star in Figure 1(A)). Overall, MeCP2 is significantly more expressed in INL and GCL than in ONL (Figure 1 (G)). This observation is consistent with previous report (Song et al. 2014).
It has been revealed that the dynamic balance between site-specific phosphorylation and dephosphorylation of MeCP2 controls the transcription of specific target genes (Zhou et al. 2006;Tao et al. 2009). In order to explore the subcellular localization of pS80, we used double immunostaining for MeCP2 (Figure 2(A)) with a pS80 MeCP2 specific antibody (Figure 2(B,E)). Interestingly, while weak MeCP2 staining was observed in all DAPI-positive heterochromatic chromocenters, no pS80 MeCP2 staining was detected in the ONL. In the lower cell layer of INL and GCL (arrowhead and arrow in Figure 2(B)), pS80 MeCP2 staining showed a punctate distribution within the heterochromatic region of nuclei. Consequently, we tested for MeCP2 phosphorylation localization at S421 (Figure 2(H,K)). While no visible signals of MeCP2 pS421 were observed in the ONL and the upper layer of the INL, limited cells in the lower layer of INL and GCL showed a punctate distribution of pS421 (star and double stars in Figure 2(H)). In the Mecp2 KO mouse, there was no visible signals detected for MeCP2, pS80, or pS421, demonstrating the specificity of the antibodies (Figure 1(E) and Figure 2(E,K)). In addition, it is known that neuronal activity leads to MeCP2 phosphorylation at S421 (Zhou et al. 2006). Therefore, we compared the intensity and population of MeCP2 pS421-labelled cells between control mice and mice reared in the dark (24hr). No significant change was observed between these mice (data not shown). Taken together, MeCP2 is phosphorylated at pS80 in most nuclei and phosphorylation at pS421 is cell specific.

Loss of MeCP2 causes limited impairment in retina structure
Various studies have been conducted on Rett syndrome, but little is known about the retina. Therefore, we compared the thickness of nuclei in the ONL of animals aged 4 and 8 weeks, with the rationale that Mecp2 deletion causes a degeneration of the ONL and INL. At 4 weeks (Figure 3(A-C,G)), the thickness (59.0 ± 4.64 μm, n = 5) of Mecp2 KO ONL was 96.27% that of control (56.8 ± 4.66 μm, n = 5) (p = 0.61), and 95.7% at 8 weeks (WT; 61.0 ± 2.92 μm, n = 5 vs Mecp2 KO; 58.40 ± 3.20 μm, n = 5) (p = 0.39) (Figure 3(D-F,H)). Overall, the thickness of cells in the ONL of Mecp2 KO retina did not change and was statistically identical to that of WT littermates. We also sought to determine whether loss of MeCP2 in the INL affects the survival of INL cells. Under light microscopy, the thickness of the INL in Mecp2 KO retinas was not obviously different from WT littermates at 4weeks (WT; 43.6 ± 1.81 μm, n = 5 vs Mecp2 KO; 42.0 ± 2.43 μm, n = 5) (p = 0.61) and at 8weeks (WT; 43.8 ± 5.59 μm, n = 5 vs Mecp2 KO; 43.8 ± 2.75 μm, n = 5) (p = 1.0) (Figure 3(A-H)). These data strongly suggest that overall cell viability is not affected in the Mecp2 KO mouse.
The retina consists of multiple layers and has a sophisticated morphology. Misregulation of gene expression can often cause morphological changes in retina structure (Rattner et al. 2001). To determine whether primary cilium formation is defective in the absence of MeCP2, we compared the retinas of WT and Mecp2 KO at 8 weeks. We performed immunostaining for Pcdh21 (Figure 3(K,L)), expressed at the interface between the retina and retinal pigment epithelium (RPE) (Rattner et al. 2001), and rhodopsin, which is most abundantly expressed in the outer segment (OS) (Figure 3(I,J)). Similar expression of Pcdh21 and rhodopsin was observed in retina sagittal sections of Mecp2 KO. The OS of the photoreceptors appeared intact at the outer edge of the ONL in Mecp2 KO when compared to WT. Therefore, we concluded that loss of MeCP2 does not cause impairment in the development of the photoreceptor outer segments.

MeCP2 expression in Mecp2 heterozygous mice
As Mecp2 is located on the X chromosome, Mecp2 heterozygous mice have a mosaic expression of MeCP2. Although Mecp2 heterozygous mice should show random X chromosome inactivation (XCI) ratios, skewed XCI is observed in the heterozygous mouse (Braunschweig et al. 2004). In order to analyze the MeCP2 expression pattern in Mecp2 heterozygous mice, sagittal sections through the retina were immunostained with C-terminal anti-MeCP2. A mosaic pattern of MeCP2 expression was observed in the retina nuclear layer (INL and ONL) in Mecp2 heterozygous mice (Figure 3(O,Q)). However, the mean percentage of MeCP2-negative cells ranged from 20 ± 4.26% to 39.2 ± 5.23% in Mecp2 heterozygous mice (n = 5), and this was lower than the overall expected 50%. In addition, there was some variation in the percentage of MeCP2-negative cells in the region of the retinal nuclear layer. For example, some regions exhibit mosaic expression (arrowhead in Figure 3(O, Q, and R-R")), while others have little MeCP2 expression (arrow in Figure 3(O,Q, and P-P")). These results indicate patchy regional distribution of MeCP2-expressing cells in the retina of Mecp2 heterozygous mouse.
Due to the localization of the MeCP2 gene in the X chromosome, girls with Rett syndrome carry the defective gene in a portion of their retina for their entire lifetime. This impaired neuronal system is susceptible to degenerative disorders in old age. Here, we compared structural changes in the retina of Mecp2 +/-for aged mouse (4 month and 9 month) with control littermates (Figure 3(S)). Histological experiment-based statistical analysis of retinas from Mecp2 heterozygous mice revealed the absence of retinal degeneration in Mecp2 +/-and this was evident up to 9 months (Figure 3(S)). Taken together, these results demonstrate that MeCP2 was not necessary for structural maintenance. in O and Q. The regional MeCP2 expression of Mecp2 heterozygous mice are shown at high magnification (P-P" and R-R"). Scale bar, 30μm in A-F, I-N, P-P" and R-R". (S) Thickness from the optic nerve head was measured at 0, 300, 600, 900, 1200, 1500, 1800, and 2000 μm. Shown are the mean ± standard deviation of retinal thickness; n = 5.

Retinal ganglion cells from Mecp2 KO mice have morphological abnormalities in dendrites
To characterize the effects of MeCP2 loss on individual RGCs, we took advantage of the M-line of Thy1-GFP transgenic mice, which express GFP in a subpopulation of RGCs (Feng et al. 2000). We crossed homozygous Thy1-GFP-M transgenic mice with Mecp2 heterozygous mice. GFP labeled RGCs, from the retinas of the three Mecp2 WT/Thy1-GFP mice and three Mecp2 KO/Thy1-GFP mice, were visualized with confocal microscopy. By using the organization of names provided by Sun et al. (2002) (Sun et al. 2002), 8RGCs evidently classified as RG A were used for statistical analyses to demonstrate the dendritic morphological difference between WT and Mecp2 KO.
Dendrite morphology was analyzed using a Sholl analysis program, Bonfire (Figure 4(A,B); RG A in WT, C and D; RG A in Mecp2 KO) (Langhammer et al. 2010). In order to determine RGC changes in dendritic morphology between WT and Mecp2 KO, we carried out a thorough morphometric analysis. The number of intersections gives a quantitative representation of how neurite density varies spatially. Statistical analysis revealed a significant change between 80 and 125 µm in the soma: roughly a 7.7% increase in number of intersections (Figure 4(E)). A more detailed picture of morphological change was generated using local-level Sholl analysis (Figure 4(F-H)). This analysis identified that the increased number of Sholl intersections was due to an increase in the number of secondary neurites (Figure 4(G)), causing a significant increase in the number of branch points up to 22.8% (p < 0.05) and up to 15.15% (p < 0.05) at terminal points (branch points WT; 26.65 ± 1.485, n = 5 vs Mecp2 KO; 34.50 ± 1.397, n = 3, terminal points WT; 34.12 ± 1.35, n = 5 vs Mecp2 KO; 40.21 ± 0.98, n = 3) (Figure 4(I)). Also, a significant increase was observed in total dendritic length up to 22.9% (p < 0.05, WT; 4720 ± 438.2, n = 5 vs Mecp2 KO; 5800 ± 700 µm, n = 3) (Figure 4(I)). Overall, the morphometric parameters demonstrate more complexity in Mecp2 KO retinas, when compared with the same types of RGCs from adult Thy1-GFP-M mice. We concluded that Mecp2 KO RGCs retain their characteristic morphology, but have finer dendritic geometry.

Discussion
The retina is a highly organized nervous system with several layers of nerve cell bodies and layers of synapses. Given that MeCP2 tends to be highly expressed in neuronal cell types (Kishi and Macklis 2004), the retina will be a suitable model for exploring the role of MeCP2 in vivo. MeCP2 expression was analyzed in detail during the development of retinal structures (Song et al. 2014). Specifically, MeCP2 begins to be expressed at E17 stage in ganglion and amacrine cells, whereas at P6 in bipolar cells and at 2 weeks of age in rod cells. In this study, we examined MeCP2 expression in the retina and whether the loss of MeCP2 leads to a neuropathological phenotype. We found that, among the three nuclear cell layers of the retina, the ONL had relatively low levels of MeCP2 expression, whereas MeCP2 was highly expressed in the cells of the INL and GCL. Using MeCP2 phosphorylation specific antibodies (Zhou et al. 2006;Tao et al. 2009), we found that pS80 MeCP2 was highly co-localized with pan-MeCP2 in the INL and GCL, pS421 MeCP2 was only observed in a subpopulation of cells in the INL and GCL. It has been reported that pS421 MeCP2 is increased in the suprachiasmatic nucleus (SCN) after light exposure (Zhou et al. 2006). However, light-induced neuronal activity did not increase MeCP2 phosphorylation at S421 in the retina. These results suggest that the level of MeCP2 expression varies from cell to cell in the retina, and that the signaling pathway that regulates MeCP2 phosphorylation in SCN may differ in the retina.
Cell subtypes were visualized at the single cell level using the Thy1-GFP line, and precisely defined subtypes of RGC were analyzed for dendritic morphology changes in Mecp2 KO. Loss of MeCP2 induced significant but residual levels of augmented branching in the dendrites of RGCs. The data presented here support and extend similar findings that have recently been reported in Mecp2 KO (Kishi and Macklis 2004;Fukuda et al. 2005;Stuss et al. 2012;Lee et al. 2014). However, we need to take a cautious approach in interpreting the results, because the effects of MeCP2 null are mostly cell typeor tissue type-dependent. For example, while mitral cell dendrites developed normally in Mecp2 KO (Palmer et al. 2012), pyramidal neurons in the motor cortex showed deficient arborization (Stuss et al. 2012). Recent studies have shown that MeCP2 regulates the gene expression through binding to methylated DNA as well as various epigenetic modifications (Mellén et al. 2012;Lee et al. 2020). Considering that different cells have different epigenetic modifications, it suggests that morphological changes in Mecp2 mutants may be caused by cell-specific mechanisms.
As a part of the mechanism to explain the morphological changes, it was reported that Bdnf gene expression is suppressed by MeCP2 binding to the promoter region of the Bdnf gene (Zhou et al. 2006). In Mecp2 KO, the absence of MeCP2 triggers overexpression of the Bdnf gene and causes morphological changes in dendritic / axonal morphology, dendritic spine density, synapse formation and maturation through secondary neuronal structure modulation (Kishi and Macklis 2004;Fukuda et al. 2005;Smrt et al. 2007). However, genome-wide analysis studies showed that, in the absence of Mecp2, a broad number of genes are significantly affected including transcriptionand translation regulatory genes (Chahrour et al. 2008;Li et al. 2013). Here, it is highly likely that other altered gene expressions exacerbated or compensated for morphological or physiological changes.
Various defects have been observed in RTT patients and Mecp2 null mice, but so far few have been reported to have ocular defects in RTT patients (Jain et al. 2010;Townend et al. 2018;de Breet et al. 2019). This study detailed MeCP2 expression in the retina and analyzed structural changes including ganglion cell morphology in the retina of Mecp2 KO. Different cells in the retina exhibit different levels of MeCP2 expression and intranuclear distribution. Therefore, it is necessary to analyze the role of MeCP2 at the single cell level for an in-depth understanding.

Disclosure statement
No potential conflict of interest was reported by the author(s).