Retinal interneuron survival requires non-cell-autonomous Atrx activity

Abstract

ATRX is a chromatin remodeling protein that is mutated in several intellectual disability disorders including alpha-thalassemia/mental retardation, X-linked (ATR-X) syndrome. We previously reported the prevalence of ophthalmological defects in ATR-X syndrome patients, and accordingly we find morphological and functional visual abnormalities in a mouse model harboring a mutation occurring in ATR-X patients. The visual system abnormalities observed in these mice parallels the Atrx-null retinal phenotype characterized by interneuron defects and selective loss of amacrine and horizontal cells. The mechanisms that underlie selective neuronal vulnerability and neurodegeneration in the central nervous system upon Atrx mutation or deletion are unknown. To interrogate the cellular specificity of Atrx for its retinal neuroprotective functions, we employed a combination of temporal and lineage-restricted conditional ablation strategies to generate five different conditional knockout mouse models, and subsequently identified a non-cell-autonomous requirement for Atrx in bipolar cells for inhibitory interneuron survival in the retina. Atrx-deficient retinal bipolar cells exhibit functional, structural and molecular alterations consistent with impairments in neuronal activity and connectivity. Gene expression changes in the Atrx-null retina indicate defective synaptic structure and neuronal circuitry, suggest excitotoxic mechanisms of neurodegeneration, and demonstrate that common targets of ATRX in the forebrain and retina may contribute to similar neuropathological processes underlying cognitive impairment and visual dysfunction in ATR-X syndrome.

Introduction

Epigenetic regulation plays a critical role in central nervous system (CNS) development, function, and disease (1). The modulation of chromatin structure establishes gene expression programs that determine neuronal identity, enable integration and long-term maintenance within defined circuits, and allow for adaptive changes through synaptic plasticity (2). It is therefore not surprising that chromatin regulators have been implicated in the pathogenesis of a broad range of human neurodevelopmental and neurodegenerative diseases (2–4), further underscoring their significance in CNS regulation throughout life.

ATRX (α-thalassaemia/mental retardation, X-linked) is a chromatin remodeling protein that is mutated in a variety of inherited intellectual disability syndromes as well as in a growing number of cancers (5–14). Mouse models of Atrx dysfunction in neuronal tissues have been developed to provide insight into the neuropathophysiology of ATR-X syndrome, which includes microcephaly, cognitive impairment and visual deficits (8,15). Conditional deletion of Atrx from the brain results in defective neurogenesis and hypocellularity of the cortex and hippocampus due to the loss of proliferating neural progenitors during embryonic stages, resulting in reduced numbers of dentate granule cells, pyramidal cells of upper cortical layers and GABAergic interneurons (16,17). Deletion of Atrx in the developing retina causes the death of post-mitotic neurons, characterized by postnatal loss of the inhibitory interneurons, amacrine and horizontal cells, after they have differentiated (15). Many subtypes of these retinal interneurons exhibit reduced numbers and altered morphology, suggesting widespread requirement of Atrx in functionally diverse populations of cells (15). Examining the importance of Atrx in different types of neurons and at different times will help delineate its role in establishing and maintaining the viability of neural tissues, and will further our understanding of how neuronal dysfunction arises in its absence or deficiency.

The retina is an extension of the brain and serves as a model for the CNS, providing an accessible, tractable and well-defined system to study neuronal biology in health and disease (18). Given the importance of chromatin regulation for the survival of neurons in the retina (19), we sought to determine the effect of Atrx deficiency in the context of visual dysfunction associated with neurological disease. In this study we demonstrate that a mouse model of ATR-X syndrome harboring a mutation identified in affected patients (20) exhibits death and dysfunction of retinal interneurons. This phenotype was consistent with our characterization of mice with conditional retinal deletion of Atrx (15), underscoring the utility of our models for the study of visual and neuronal dysfunction in human disease pathogenesis. To further examine how Atrx loss leads to neuronal death and visual impairment, we aimed to investigate the temporal and spatial requirement of Atrx in specific retinal cell types. Here, we show that Atrx is required postnatally in retinal bipolar cells to maintain amacrine and horizontal cell integrity, indicating that a non-cell-autonomous mechanism underlies Atrx-mediated retinal interneuron survival. The molecular changes observed in Atrx-deficient retinas suggest defects in intercellular communication and synaptic structure, and suggest an excitotoxic mechanism of neurodegeneration. In addition, we observe the dysregulation of genes corresponding to previously described Atrx targets in the brain, particularly those encoding synaptic proteins implicated in visual and/or cognitive pathology, further demonstrating the relevance of our findings to understanding Atrx function across the CNS.

Results

Retinal interneuron defects occur in a patient mutation model of ATR-X syndrome

Tissue-selective conditional Atrx ablation in mice recapitulates clinical features of human ATR-X syndrome (15,16,21–23). In this regard, conditional ablation of Atrx in the retina using the Pax6α-Cre driver line resulted in approximately a 35% reduction in amacrine and horizontal cells while maintaining normal numbers of photoreceptors, bipolar cells and ganglion cells. The loss of these interneurons was associated with an altered scotopic electroretinogram, including reduced b-wave and oscillatory potential (OP) amplitudes (15). However, total ablation of ATRX protein is not observed in ATR-X syndrome patients, who instead have hypomorphic mutations in the ATRX gene that result in reduced levels of the protein accompanied by defects in catalytic activity and/or intracellular mislocalization of the mutant protein (24–27). An Atrx mutant mouse model deleted for exon 2 (Atrx^ΔE2) actually mimics exon 2 truncating mutations identified in patients with mild intellectual disability (20,28,29). To determine if visual abnormalities occur in this disease model that more accurately reflects a genetic defect underlying human ATR-X syndrome, we characterized the retinal phenotype of Atrx^ΔE2 mice. Western blotting of protein extracts derived from Atrx^ΔE2 and wildtype retinas revealed that Atrx levels are reduced by ∼80% in the mutants (Fig. 1A), which is consistent with previous observations made for Atrx^ΔE2 brain tissue and lymphoblast cells from human ATR-X syndrome patients carrying the analogous mutation (20,30). Examination of specific retinal cell types revealed a reduction in the amacrine and horizontal cell populations (Fig. 1B and C), similar to what we observed in the Atrx-null retinas of conditional knockout mice (Pax6-AtrxKO in Table 1) (15). The Atrx^ΔE2 retinas contain fewer Pax6-immunoreactive amacrine cells (Atrx^ΔE2: 261.8 ± 12.1; WT: 297.7 ± 3.0; P =  0.00687, n = 6 eyes; Fig. 1B) and calbindin-immunoreactive horizontal cells (Atrx^ΔE2: 7.5 ± 2.9; WT: 14.0 ± 2.0; P = 4.46 × 10⁻⁹, n = 6 eyes; Fig. 1C) compared with wildtype controls, while bipolar cell number does not change (Atrx^ΔE2: 188.3 ± 7.4; WT: 194.4 ± 2.4; P = 0.443, N = 6 eyes; Fig. 1D). Functional deficits in the Atrx^ΔE2mice also parallel those observed for the retinal Atrx-null mice, with a reduction in the scotopic electroretinogram (ERG) b-wave amplitude by an average of 32% (Fig. 1E;Table 1), indicating a post-receptoral neuronal defect affecting the health of the inner nuclear layer (INL), particularly retinal bipolar cell activity. This was accompanied by reduced amplitudes of the OPs (Fig. 1F;Table 1), which is associated with impaired amacrine cell function. The timing of retinal electrophysiological responses to visual stimulation appeared normal and photoreceptor function was not impaired in the mutants, as b-wave latency and a-wave amplitude and latency remained unaffected (Supplementary Material, Fig. S1). Therefore, mice that have a mutation mimicking ATR-X syndrome patients exhibit retinal interneuron abnormalities and deficits in inner retinal function, which are consistent with the phenotype of mice harboring Atrx-null retinal tissue.

Figure 1

Atrx^ΔE2 mice have retinal interneuron defects similar to those observed in Atrx cKO mice. (A) Atrx protein levels are reduced in Atrx^ΔE2 mouse retinas. Immunoblotting was performed on total retina protein extracts from adult (WT) and Atrx^ΔE2 mice; n = 6 eyes. (B–D) Retinal cryosections from 8-week-old WT and Atrx^ΔE2 mice were immunostained with antibodies against the amacrine cell marker Pax6 (B, red), the horizontal cell marker calbindin (C, green; arrows indicate the location of cell somas), or the bipolar cell marker Chx10/Vsx2 (D, green). Amacrine (B) and horizontal (C) cell numbers are reduced in Atrx^ΔE2 mouse retinas compared with controls, while bipolar cells are maintained (D). Sections were counterstained with DAPI to label cell nuclei. Cell counts and immunohistochemistry were performed on vertical sections taken from central retina; n = 6 eyes. ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer. Scale bar, 50 µm. Scotopic b-wave amplitudes (E) and OPs (F) of Atrx^ΔE2 mice are reduced compared to controls. ERGs for Atrx^ΔE2 mice were performed on n = 30 eyes; WT, n = 22 eyes. Percentage of Atrx protein expression in (A), cell counts in (B, C), and (D), and amplitudes in (E, and F), are represented as mean ± standard error of the mean (SEM). Statistically analysis for (A–D) was performed using a two-tailed, two-sample Student’s t-test with equal variance. For (E–F), a repeated measures ANOVA with a Bonferroni–Dunn correction for multiple comparisons was performed to detect differences between Atrx^ΔE2 and wildtype cohorts. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Table 1

Summary of the retinal phenotype associated with different Atrx transgenic mouse lines

Transgenic Mouse Line	Onset of Atrx Deficiency	Targeted Retinal Cell Types	% Amacrine Cell Loss	% Horizontal Cell Loss	Reduction in b-wave amplitude	Reduction in OP amplitude	Abnormal Bipolar Cell Morphology?	Aberrant Bipolar Cell Marker Expression?
Pax6-AtrxKO/Chx10-AtrxKO	E10.5	Retinal progenitors (pan-retinal)	∼34–41%	∼37–50%	∼30–35%	∼30–40%	Yes	Yes
Ptf1-AtrxKO	E14.5	Amacrine & horizontal cell precursors	None	None	None	None	ND	ND
Atrxflox + CAG-Cre (in vivo electroporation)	P0	Mainly rod photoreceptor & bipolar (few Müller & amacrine) cell precursors	∼33%a	ND	ND	ND	ND	ND
Vsx2-AtrxKO	P3	Bipolar cell precursors	∼24%	∼33%	∼27%	∼37%	Yes	Yes
AtrxΔE2	Constitutive	Retinal progenitors	∼12%	∼46%	∼32%	∼33%	ND	Yes

^aReflects quantification of ChAT-immunoreactive SACs

ND, not determined.

Loss of Atrx induces non-cell-autonomous neurodegeneration

Atrx is expressed in embryonic progenitor cells throughout retinal development and in the mature neuronal cell types generated from them, however its absence leads to the selective death of amacrine and horizontal cells postnatally (15). Since all retinal cell types arise from common multipotent progenitors (31), inactivation of Atrx at early developmental stages through conditional knockout in retinal progenitor cells or in the germline of the hypomorphic Atrx^ΔE2 mice does not restrict Atrx loss to a particular cell lineage that would reveal cell type specificity of the defect. To determine if the death of amacrine and horizontal cells was caused by Atrx deficiency in a common precursor cell population, we generated lineage-restricted Atrx conditional knockout mice using the Ptf1a-Cre driver line. Ptf1a is a bHLH transcription factor expressed in amacrine and horizontal precursor cells, and Cre expression in the Ptf1a-Cre transgenic mouse retinas is detectable beginning at E14.5 (32). Loss of Atrx protein in Ptf1a-Cre^+/-;Atrx^flox/y (Ptf1Cre-Atrx cKO) retinas was restricted to the inhibitory interneuron populations as expected (Fig. 2A and B) and was consistent with the reported function of the Ptf1a-Cre driver line (32–34). To our surprise, these mice did not exhibit a loss of amacrine cells (cKO: 160.7 ± 12.6; WT: 167.8 ± 9.8; P = 0.438, n = 3-4 mice) or horizontal cells (cKO: 13.3 ± 1.2; WT: 14.0 ± 1.1; P = 0.699, n = 3–4 mice) despite a clear reduction in the number of Atrx-immunoreactive cells in the INL (Fig. 2A–C). While Cre activity (and thus Atrx loss) was not observed in all amacrine cells due to incomplete targeting of various amacrine subtype populations, as previously documented for the Ptf1a-Cre driver line (33,34), the lack of any amacrine or horizontal cell loss observed in the retinas of the Ptf1Cre-Atrx cKO mice indicates that Atrx is not required in these cell types for their survival. Moreover, these mice did not exhibit functional deficits upon measurement of scotopic ERG b-wave amplitudes (Fig. 2D) that were previously observed in Pax6α-Cre driver-mediated Atrx cKO animals (Table 1) (15). Taken together, these studies ruled out a cell-autonomous role for Atrx in amacrine and horizontal cell function. This in turn indicates that amacrine and horizontal survival depends on Atrx expression in earlier stage proliferating retinal progenitors, in post-mitotic amacrine and horizontal cell precursors prior to E14.5, or in another retinal cell type.

Figure 2

Conditional deletion of Atrx in horizontal and amacrine cell progenitors using the Ptf1a-Cre driver line does not alter interneuron cell numbers or impair retinal function. (A) Retinal cryosections from adult Ptf1Cre^-/-;Atrx^flox/y (WT) or Ptf1Cre^+/-; Atrx^flox/y (Ptf1-AtrxKO) mice immunostained with antibodies against Atrx (red) and the amacrine cell marker Pax6 (green). (B) WT and Ptf1-AtrxKO retinas immunostained with antibodies against Atrx (red) or the horizontal cell marker calbindin (green). DAPI staining of cell nuclei is shown in blue in the top panels of (B). Higher magnifications of the INL are shown in the bottom panels of (A) and (B). Fewer Atrx-immunoreactive amacrine cells (A) and horizontal cells (B) are detected in the Ptf1-AtrxKO mice, but the overall number of amacrine and horizontal cells remains the same. Similarly, there is no reduction in the thickness of the INL compared with wildtype retinas. Scale bar, 50 µm. (C) The number of Pax6-immunoreactive amacrine cells and calbindin-immunoreactive horizontal cells is unchanged in the Ptf1-AtrxKO (KO) mice relative to WT littermates (n = 3–4 mice). Error bars represent standard error of the mean. Statistical analysis was performed using a two-tailed, two-sample Student’s t-test with equal variance. (D) Electroretinogram showing equivalent scotopic b-wave amplitudes for the Ptf1-AtrxKO mice (n = 13–14 eyes) and Ptf1Cre^-/-;Atrx^flox/y control (WT) littermates (n = 6–9 eyes). Error bars represent standard error of the mean. A repeated measures ANOVA with a Bonferroni–Dunn correction for multiple comparisons was performed to detect differences between Ptf1-AtrxKO and wildtype cohorts.

Post-natal retinal Atrx deletion phenocopies embryonic Atrx loss

The different retinal cell types are generated from a common pool of progenitor cells in a well-defined temporal order (31). During embryogenesis, ganglion cells, cone photoreceptors, horizontal cells, and the bulk of amacrine cells are generated, while the majority of rod photoreceptors, bipolar cells, Müller glia, and the remainder of amacrine cells are generated postnatally (35). To investigate the postnatal requirement of Atrx in the retina, we deleted Atrx at postnatal day 0 (P0) by subretinal injection and in vivo electroporation of a pCAG–CreGFP expression plasmid into the retina of Atrx^flox/flox or Atrx^flox/y (Atrx^flox) mice. This plasmid enables high levels of expression of a Cre–GFP fusion protein (which can be detected by fluorescence) in all cells responsive to the ubiquitous chicken ß-actin promoter. Differential expression within specific retinal cell populations in vivo is achieved based on preferential uptake and distribution of the plasmid to daughter cells by proliferating progenitors at the time of transfection. Previous studies have shown that transfection by electroporation at P0 targets progenitor cells that give rise to primarily rod photoreceptors (80%) and bipolar cells (15%), with 5% of transfectants corresponding to amacrine and Müller glial cells (36). Although Atrx is not expressed in rod photoreceptors, this experiment would reveal the importance of Atrx in other postnatally born cell types. At 4 weeks post-injection, retinal wholemounts were analyzed for CreGFP transgene expression, Atrx deletion, and the viability of cholinergic (choline acetyltransferase (ChAT)-immunoreactive) starburst amacrine cells (SACs), a population of interneurons that was significantly reduced in the transgenic Atrx cKO retinas (15). ChAT-immunoreactivity of SAC dendrites also enables an analysis of stratification defects, which are readily detected in retinal wholemounts, and which were observed in the Pax6α-Cre driver-mediated transgenic Atrx cKO retinas (15), thus serving as a good readout for the retinal AtrxKO phenotype. Upon Cre expression in transfected retinas, we observed loss of Atrx from the distal INL of the Atrx^flox mice, where bipolar cells reside (Fig. 3A, compare top left and right panels). Following ChAT immunostaining, the wildtype injected retinas displayed a normal pattern of SAC distribution and morphology with numerous cell somas labeled in the INL and the ganglion cell layer (GCL), and two labeled dendritic tracks in the inner plexiform layer (IPL), indicating the location of synapse formation with bipolar and ganglion cells (Fig. 3a, bottom left panel). In contrast, the Atrx^flox injected retinas showed a dramatic reduction in cell soma immunoreactivity and a disorganized dendritic stratification pattern (Fig. 3a, bottom right panel). Quantification indicated that P0 deletion of Atrx resulted in a comparable reduction in the number of ChAT+ cells as observed for the transgenic Pax6α-Cre driver-mediated Atrx cKO retinas (Fig. 3B;Table 1), in which Atrx deletion is initiated in embryonic retinal progenitors at E10.5. This also paralleled the overall proportion of approximately one-third of total amacrine cells lost in the Pax6-AtrxKO model (Table 1) (15). Detection of fluorescence of the CreGFP fusion protein verified electroporation of retinal cell populations consistent with previous reports (Supplementary Material, Fig. S2) (36). Since Atrx is not expressed in rod photoreceptors and electroporation at P0 targets minimal numbers of Müller glia or amacrine cells, this data suggests that ChAT+ amacrine cell loss results from a requirement for Atrx in bipolar cells and/or their precursors.

Figure 3

Deletion of Atrx at P0 recapitulates the AtrxKO phenotype. (A) Maximum intensity projections of confocal Z-stacks taken for retinal wholemounts from adult wildtype C57BL/6 or Atrx^flox mice subretinally injected and electroporated in vivo with a pCAG-CreGFP expression construct, and immunostained with antibodies against Atrx or choline acetyltransferase (ChAT), shown in red. DAPI staining of cell nuclei is shown in blue. Electroporated C57BL/6 retinas exhibit robust expression of Atrx in the INL and GCL (top left panel), while injected Atrx^flox retinas show significant loss of Atrx from the distal INL (top right panel, white bracketed region). Injected Atrx^flox retinas also exhibit disrupted ChAT immunoreactivity (bottom right panel) compared with the C57BL/6 retinas (bottom left panel). ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer. Scale bar, 10 µm. (B) The number of ChAT-immunoreactive cells is reduced in the electroporated Atrx^flox (mutant) retinas compared to the injected C57BL/6 (WT) retinas and to a similar extent as observed for the transgenic Pax6-AtrxKO retinas. Number of mice analyzed per group is indicated within the bars on the graph. Error bars represent standard error of the mean. Statistical analysis was performed using a two-tailed, two-sample Student’s t-test with equal variance. *P < 0.05.

Atrx loss in retinal bipolar cells mediates the neurodegeneration

The proximity and synaptic connectivity of retinal bipolar cells to both horizontal and amacrine cells suggest that degeneration of these inhibitory interneurons may arise non-cell-autonomously from the loss of Atrx in bipolar cells. To test this hypothesis, we generated retinal bipolar cell-specific Atrx-null mice (Vsx2Cre-Atrx cKO mice, hereon) using the Vsx2-5.3-PRE-Cre driver line, in which Cre recombinase is active in post-mitotic bipolar cell precursors beginning at P3 (37). The Vsx2Cre-Atrx cKO retinas were analyzed at P17 and exhibited robust loss of Atrx exclusively from bipolar cells, while retaining expression in other cell types within the INL and the GCL (Fig. 4A). More importantly, Vsx2Cre-Atrx cKO retinas contained fewer Pax6+ amacrine cells (cKO: 184.9 ± 11.3; WT: 244.8 ± 13.6; P = 2.15 × 10⁻¹¹, n = 6 eyes; Fig. 4B) and calbindin+ horizontal cells (cKO: 8.0 ± 1.2; WT: 11.9 ± 1.5; P = 4.19 × 10⁻⁹, n = 6 eyes; Fig. 4C) than wildtype littermates. In contrast, Atrx deletion had no effect on bipolar cell survival, as the number of cells stained with antibodies against Chx10, a pan-bipolar cell marker, was equivalent in mutant and wildtype retinas (cKO: 279.2 ± 17.8; WT: 283.7 ± 18.9; P = 0.543, n = 6 eyes; Supplementary Material, Fig. S3). Overall, approximately one-quarter of amacrine cells and one-third of horizontal cells are lost in the mutant retinas, which was comparable to the pan-retinal Atrx cKO retinas (Table 1) (15). We found similar electrophysiological changes upon bipolar cell-specific Atrx deletion, with both scotopic b-wave and OP amplitudes reduced in Vsx2Cre-Atrx cKO mice compared with wildtype control littermates (Fig. 4D, E, G;Table 1). Loss of Atrx specifically impacted the magnitude of the bipolar cell response to visual stimuli while maintaining normal temporal components of the response, since b-wave latency measurements were unchanged (Fig. 4F). Photoreceptor function was also unaffected as indicated by normal amplitude and timing of the scotopic a-wave (Supplementary Material, Fig. S4). Taken together, these studies highlight the importance of Atrx for retinal bipolar cell function and for the non-cell-autonomous survival of amacrine and horizontal cells.

Figure 4

Retinal bipolar cell-specific deletion of Atrx causes the AtrxKO phenotype. (A) Retinal cryosections from postnatal day 17 Vsx2-5.3-PRE-Cre^-/-;Atrx^flox/y (WT) or Vsx2-5.3-PRE-Cre^+/-;Atrx^flox/y (Vsx2-AtrxKO) mice were immunostained with antibodies against Atrx (red), the bipolar marker Chx10 (third panels from the left, green) or Cre recombinase (right panels, green), demonstrating the specificity of Atrx deletion in bipolar cells. White brackets delineate the distal INL where bipolar cells reside. Bottom panels show higher magnification of the INL. ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer. Scale bar, 50 µm. Amacrine (B) and horizontal (C) cell numbers are reduced in Vsx2-AtrxKO mice compared with WT littermates. Vertical sections taken from central retina; cell counts are represented as mean ± SEM, n = 6 eyes. Statistical analysis was performed using a two-tailed, two-sample Student’s t-test with equal variance. Scale bar, 50 µm. Scotopic ERG measurements of b-wave amplitudes (D, E), b-wave latency (F), and OPs (G) of adult Vsx2-AtrxKO mice (n = 18) and WT littermates (n = 20). B-wave and OP amplitudes are reduced in the Vsx2-AtrxKO cohort. Amplitudes in panels E–G are represented as mean ± SEM. A repeated measures ANOVA with a Bonferroni–Dunn correction for multiple comparisons was performed to detect differences between the Vsx2-AtrxKO and wildtype mice. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Atrx-depleted retinal bipolar neurons have structural defects

Since the timing of amacrine cell death coincides with bipolar cell maturation, we reasoned that Atrx loss may be associated with defects in bipolar cell development, including the establishment of connectivity and intercellular communication within the IPL, which in turn could affect amacrine cell survival. As there are 13 bipolar cell subtypes, we analyzed subtype-specific retinal bipolar cell marker gene expression in three mouse models of Atrx deficiency: (1) the Atrx^ΔE2 germline mutant, (2) Chx10Cre-Atrx cKO, a more complete pan-retinal Atrx cKO model and (3) Vsx2Cre-Atrx cKO, the bipolar cell-specific deletion mutant (see Table 1). The Chx10Cre-Atrx cKO transgenic mice were generated using the Chx10-GFP/Cre-IRES-AP driver line, which mediates robust pan-retinal deletion of Atrx due to early (E10.5) and widespread expression of Cre throughout the retina (38), in contrast to the peripheral retina-restricted deletion mediated by the Pax6α-Cre driver line used in our initial study (15,39). Retina tissues isolated from the Chx10Cre-Atrx cKO mice are therefore more appropriate for global gene expression analysis of the Atrx-null state than the Pax6α-Cre driver-mediated Atrx cKO retinas, which were only partially null for Atrx. The Atrx^ΔE2 and Chx10Cre-Atrx cKO retinas exhibited significant dysregulation for many of the bipolar cell subtype-specific genes examined (Supplementary Material, Fig. S5, panel A). While we did not observe any significant changes in bipolar cell-specific gene transcript levels in the Vsx2Cre-Atrx cKO mice relative to wildtype, substantial variability between mutants was noted (Supplementary Material, Fig. S5, panel B). We performed immunohistochemistry with antibodies directed against bipolar subtype-specific markers in order to visualize the cell soma density in the INL and the sublaminar stratification of axons and terminal boutons within the IPL to identify the subtypes affected. In this regard, the following markers were used to stain retinas of Chx10Cre-Atrx cKO and Vsx2Cre-Atrx cKO mice: recoverin (cone bipolar cells, type 2), synaptotagmin (type 2 and 6), HCN4 (type 3a), Cabp5 (type 3 and 5) and PKCα (rod bipolar cells). For all cone bipolar subtypes examined, we observed defects in the density of axon terminals stratifying in characteristic sublaminae within the IPL and reduced immunoreactivity of the terminal boutons in the mutant retinas compared with wildtype littermate control retinas (Fig. 5A–D). Moreover, the analysis of the PKCα-immunoreactive rod bipolar cells revealed defects in the trajectory of the axons, which exhibited tortuosity and dense-staining varicosities along their length in the mutant retinas that were not observed in wildtype tissue (Fig. 5E). The bipolar cells analyzed in both the Chx10Cre-Atrx cKO and Vsx2Cre-Atrx cKO models shared the identical morphological phenotype (Table 1; data not shown).

Figure 5

Morphological axon defects are observed for multiple retinal bipolar cell subtypes in AtrxKO retinas. (A–E) Immunohistochemistry was performed on vertical cryosections taken from the central retina using antibodies for the bipolar subtype marker protein labeled on the top left corner of each panel. Black arrows in (A–D) indicate the positions of axon terminals whose staining pattern differs in Chx10-AtrxKO (and Vsx2-AtrxKO) versus WT retinas. White arrows in (E) indicate tortuous axon trajectories of PKC-immunoreactive rod bipolar cells in the AtrxKO retina, while white arrowheads point to dense-staining varicosities in the axons. Scale bar, 50 µm.

The IPL corresponds to the layer in which retinal bipolar cells, amacrine cells, and ganglion cells form synapses. IPL thickness is thus a measure of synaptic density and reflects the complexity of inner retinal circuitry. We observed that, in addition to the INL, the IPL appeared thinner in the Atrx cKO mutant retinas compared with wildtype littermates (Fig. 5). Immunohistochemical staining with vGlut1, a marker of all retinal bipolar cell synaptic terminals (40,41), revealed a dramatic reduction of these structures and indicated a markedly thinner IPL in the conditional mutants (Fig. 6A–F). Indeed, quantitative analysis of retinal layer thickness (Fig. 6G) confirms that the Atrx cKO mutant retinas contain a significantly thinner INL (Chx10-AtrxKO: 48.1 ± 0.9 µm versus WT: 72.0 ± 1.3, P = 3.73 × 10⁻⁸, n = 6 eyes; Vsx2-AtrxKO: 56.5 ± 1.3 µm versus WT: 73.2 ± 2.5, P = 4.22 × 10⁻⁵, n = 8 eyes) and IPL (Chx10-AtrxKO: 40.5 ± 1.6 µm versus WT: 64.6 ± 1.3, P = 3.37 × 10⁻⁷, n = 6 eyes; Vsx2-AtrxKO: 49.8 ± 1.8 µm versus WT: 67.4 ± 2.6, P = 6.43 × 10⁻⁵, n = 8 eyes) compared with wildtype, while there is no change in the thickness of the ONL (Chx10-AtrxKO: 77.6 ± 1.1 µm versus WT: 81.7 ± 2.3, P = 0.138, n = 6 eyes; Vsx2-AtrxKO: 76.3 ± 0.9 µm versus WT: 78.8 ± 1.3, P = 0.133, n = 8 eyes), consistent with previous photoreceptor counts and with the normal scotopic a-wave responses measured. Co-immunostaining with vGlut1 and antibodies recognizing secretagogin (SCGG; Fig. 6A–C), a cone bipolar cell marker (42), or the rod bipolar cell-specific protein PKCα (Fig. 6D–F) indicated that both cone and rod bipolar cell axonal boutons are deficient in the IPL of both the Chx10Cre-Atrx cKO and Vsx2Cre-Atrx cKO models. Taken together, we conclude that all subtypes of bipolar cells examined show abnormalities in axonal structure, which may interfere with activity-driven signaling or the distribution/exchange of chemical mediators of neuroprotection required by post-synaptic amacrine cells.

Figure 6

AtrxKO retinas display a marked reduction in IPL thickness and bipolar cell synaptic terminal density. Confocal micrographs of vertical retinal cryosections co-immunolabeled with antibodies for vGlut1 (red) to mark glutamatergic bipolar cell synaptic terminals and either (A–C) secretagogin (SCGG; green) to visualize cone bipolar cell axonal boutons (yellow) or (D–F) PKCα (green) to visualize rod bipolar cell terminal boutons (yellow). Scale bar, 20 µm. (G) Quantitative analysis of retinal layer thickness in the AtrxKO retinas. Measurements were obtained from confocal micrographs of DAPI-stained vertical sections of the central retina from mice at postnatal day 17 and are represented as mean thickness ± SEM, n = 6 eyes for Chx10-AtrxKO mice (white bars) and wildtype littermates (black bars), n = 8 eyes for Vsx2-AtrxKO mice (grey bars) and wildtype littermates (black bars). Statistical analysis was performed using a two-tailed, two-sample Student’s t-test with equal variance. ****P < 0.0001.

Retinal Atrx targets subserve neuronal communication and are shared in the brain

To identify potential molecular pathways that mediate retinal bipolar cell development or function and the subsequent loss of interneurons, we used microarray analysis to perform transcriptional profiling of retinas isolated from Chx10Cre-Atrx cKO mice and wildtype control littermates at postnatal day 17, the peak of neurodegeneration (15). Based on a threshold fold-change of 1.5 and P-value <0.05 (Fig. 7A), we identified 109 upregulated genes (Supplementary Material, Fig. S6) and 64 downregulated genes (Supplementary Material, Fig. S7) in the AtrxKO retinas relative to wildtype transcript levels. The upregulated genes did not cluster into any Gene Ontology (GO) categories, whereas the downregulated transcripts were enriched for GO terms indicating neurotransmitter metabolism, ion transport, and cell signaling or communication (Fig. 7B and C). Further examination of the list of genes exhibiting reduced transcription levels reveals an expression profile that is consistent with neuronal excitotoxicity, including alterations in glutamate activity, ion channel regulation, and neuroprotective peptides (Fig. 7D). Quantitative RT-PCR confirmed the altered expression of the excitotoxicity related genes in the AtrxKO retinas (Fig. 7E).

Figure 7

Gene expression alterations following pan-retinal deletion of Atrx. Gene expression analysis of Chx10-AtrxKO versus wildtype retinas isolated at P17 was performed using Affymetrix 2.0 mouse genome microarrays. Thresholds for inclusion criteria include ≥1.5-fold change (up or down) in expression level between AtrxKO and WT retinas with a 95% confidence interval (ANOVA P-value <0.05). (A) Volcano plot of transcript clusters upregulated (red) or downregulated (green) in the KO retinas relative to WT controls. Grey data points represent genes whose expression levels fall outside the filtering parameters. (B, C) Genes involved in neuronal structure and communication are downregulated upon conditional knockout of Atrx in the retina. GOrilla (B) and GoSlim (C) analysis was performed for the list of genes downregulated in KO. (D) Genes for which reduced transcript levels are associated with neuroexcitotoxicity are downregulated in AtrxKO retinas. Color-coded expression levels correspond to Tukey’s bi-weight average of gene level intensity of all the samples in each condition, ranging from high (red) to low (blue). (E) Validation of reduced expression levels for excitotoxicity related genes by qRT-PCR of retinal total RNA extracts from P17 WT (average relative fold-change set to 1; black bars) and Chx10-AtrxKO (white bars) mice, normalized to 18S. Fold-change values are represented as mean ± SEM, n = 6 mice. Statistical analysis was performed using a two-tailed, two-sample Student’s t-test with equal variance. *P < 0.05, **P < 0.01, ***P < 0.001.

Molecular analysis of the pan-retinal Chx10Cre-mediated AtrxKO retinas using microarrays further enabled a comparison of the gene expression signatures characterizing different Atrx-null regions of the CNS, namely retina and brain tissues. The proportion of differentially expressed genes with increased or reduced expression in the retina is consistent with the results obtained for microarray analysis of AtrxKO mouse forebrain tissue, with approximately two-thirds of transcriptional changes corresponding to upregulated probe sets (43). Moreover, we identified four genes, Csf2ra, Cbln4, Syt13 and Nlgn4, implicated in neuronal function and visual or cognitive pathology (44–51), that were common to both downregulated gene sets (43,52). Expression differences were confirmed by qRT-PCR (Fig. 8). These results suggest that Atrx shares common transcriptional targets in CNS tissues and that there may be similar underlying roles of Atrx in axon guidance, synaptic function and network activity that contributes to cognitive function and visual processing, both of which are defective in ATR-X syndrome patients.

Figure 8

Genes that are downregulated in the Atrx-null mouse forebrain are similarly misexpressed in AtrxKO retinas. (A) Microarray results for genes that were previously observed to exhibit reduced expression levels in the forebrain of Atrx cKO mice. Color-coded expression levels correspond to Tukey’s bi-weight average of gene level intensity of all the samples in each condition, ranging from high (red) to low (blue). (B) qRT-PCR results for the genes in (A) as well as two additional genes previously identified as putative Atrx targets in mouse forebrain. Average fold-change relative to wildtype is plotted for the indicated genes in wildtype (set to 1; black bars) and Chx10-AtrxKO (white bars) mouse retina, normalized to 18S. Error bars represent standard error of the mean for n = 6 mice. Statistical analysis was performed using a two-tailed, two-sample Student’s t-test with equal variance. *P < 0.05, **P < 0.01, ***P < 0.001.

Discussion

Mutations in the Atrx gene affect its chromatin remodeling activity and give rise to a severe intellectual disability syndrome (9). Animal models developed to investigate the role of Atrx in neural development have used early neural progenitor deletion strategies that generated insight on the requirements for Atrx in progenitor proliferation but have masked its role in mature neurons (16,17,53). Indeed, inactivation of Atrx in early retinal progenitors led to a complex phenotype that included interneuron cell loss coupled with changes in nuclear morphology and electrophysiological deficits (15). Here, we used a temporal and lineage-specific inactivation strategy to deduce the cellular requirement for retinal Atrx function. In this regard, we made four novel insights: (1) Atrx function in progenitors can be uncoupled from its role in post-mitotic neurons; (2) there is a cell-autonomous requirement for Atrx in retinal bipolar cells that is critical for their morphology, circuitry, and function; (3) unexpectedly, Atrx function in bipolar cells also exerts a non-cell-autonomous effect on interneuron survival; and (4) Atrx patient mutations phenocopy the spectrum of defects associated with complete Atrx inactivation. Collectively, this study highlights the pleiotropic nature of activities regulated by the Atrx protein and the complexity of the pathology associated with patient mutations.

Atrx function in progenitors can be uncoupled from its role in post-mitotic cells

Chromatin remodeling proteins are important for maintaining genome integrity during progenitor cell growth and for establishing the chromatin landscape that defines gene expression profiles and cell fate (54). Studies characterizing mice deleted for Atrx in the cortex, testes and muscle highlighted a requirement for Atrx in progenitor cell survival and development of tissue size (16,21,22). More specifically, Atrx deficiency is associated with replication fork stalling and DNA damage, fragile telomeres and mitotic catastrophe (22,55,56). Surprisingly, within the retina we did not observe a significant increase in progenitor cell death (15). Thus Atrx appears to function in a tissue and cell type-specific manner during development. One key difference in the retina is that progenitor cell expansion is not temporally compressed like it is in the cortex, muscle or testes but occurs over a prolonged developmental time, suggesting that it may be insulated from cumulative replication stress-induced DNA damage (35,57,58). A rapid proliferative phase seems to be critical for the replication infidelity and premature cell death; this is absent in the retina, and pan-retinal deletion at early embryonic times (Chx10-AtrxKO, Pax6-AtrxKO, Atrx^ΔE2) resulted in a similar phenotype to ablation in late progenitors (P0 deletion by in vivo electroporation) or in restricted bipolar progenitors (Vsx2-AtrxKO). Nevertheless, we did observe some differences in bipolar cell gene expression in the retinas with pan-retinal Atrx inactivation and germline Atrx^ΔE2 mutation compared with the Vsx2-AtrxKO line that are most likely attributed to deficiencies in the chromatin signature accrued by inactivating Atrx at earlier times. In support of this, Atrx-containing chromatin remodeling complexes bind repetitive sequences within heterochromatin and regulate histone variant deposition to promote either transcriptional activation or silencing at specific genetic loci (59–61). Differential transcriptional regulation by Atrx may account for gene expression requirements in immature progenitors that are not shared by mature neurons. This may also explain the cell type-specificity of Atrx function, whereby transcriptional dysregulation of post-mitotic retinal bipolar cell-specific genes may influence the synthesis and/or release of survival factors, or alternatively neurodegenerative factors, that act on target neurons. Indeed, unraveling the function of Atrx in post-mitotic neurons will prove useful in identifying dysfunctional pathways that could be amenable to therapeutic intervention in patients.

Atrx mediates retinal bipolar cell maturation and/or function

Within the retina, Atrx function is required postnatally in bipolar neurons for proper morphology and functional neurotransmission through these cells. Unlike the loss of other bipolar genes (e.g. Vsx1, Bhlhb4, Bhlhb5, Irx5 or Irx6) (62–67), we did not observe sub-lineage specification defects but rather more general defects in the development or maintenance of structural integrity within all bipolar cells. In this regard, we observed axon morphology defects, fewer terminal boutons, and the mislocalization of synapses across the IPL strata for all bipolar cell subtypes examined. In support of a developmental role of Atrx in promoting bipolar cell maturity, the microarray results revealed reduced expression of somatostatinergic components (Sst and Sstr2). Somatostatin receptors are upregulated at or near eye-opening, after which complete morphological maturation of bipolar cells takes place (68,69). Deletion of Sstr2 causes retinal bipolar axon terminal developmental defects (70), further suggesting that the altered gene expression and bipolar cell morphology may be the result of impaired bipolar cell maturation. Consistent with a general role for Atrx in neuronal maturation is the observation that hippocampal neurons in the Atrx^ΔE2 mice have fewer mature dendritic spines and learning deficits (20,71). However, we cannot distinguish whether interneuron death in the Atrx mutant retina is secondary to the defects in bipolar cell development or due to light-stimulated neuronal activity initiated upon eye opening because both processes are temporally co-incident. Activity-dependent mechanisms resulting from bipolar cell responses to visual stimulation may generate anterograde axon-derived survival signals directed towards amacrine and horizontal cells in a form of afferent control (72,73). In this regard, light deprivation in rabbits reduces cell density in the INL (74), and dark rearing causes loss and impaired dendritic ramification of cholinergic SACs in mice (75). Furthermore, loss of retinal input due to photoreceptor degeneration leads to morphological remodeling of bipolar cells and death of horizontal and amacrine cells in the dystrophic retinas (76,77). Accordingly, Atrx has been shown to mediate transcriptional changes upon neuronal stimulation (78). Indeed, the relative contribution of developmental and activity-dependent factors to this complex phenotype could be reconciled by assessing the impact of Atrx loss in mature bipolar cells and under conditions of light deprivation. In addition, comprehensive transcriptional profiling of Atrx-null bipolar cells by RNA sequencing and utilizing the specificity conferred by the Vsx2Cre driver line at different timepoints may provide insight into the molecular events that characterize either impaired development or mature functions of retinal bipolar cells with defective chromatin remodeling owing to Atrx deficiency.

Atrx has a non-cell-autonomous effect on interneuron survival

Using a collection of different Cre driver lines we were able to demonstrate that horizontal and amacrine cell survival was dependent on Atrx activity within retinal bipolar cells. The interdependence between these cell types was unanticipated and represents the first report, to our knowledge, of retinal bipolar cell-mediated neuroprotection. In addition, interneuron cell loss was restricted to a 10-day window (P7–P17), did not include bipolar cells themselves yet affected all amacrine cell subtypes examined (15). There may be several possible mechanisms underlying this cell-non-autonomous effect. Gene expression studies identified the dysregulation of numerous genes (Sst, Sstr2, Gria1, Cnih3, Cacng4, Olfm3 and Gad1) involved in glutamate homeostasis that reflects a shift toward excess glutamate production and collectively support a glutamate-mediated excitotoxicity mechanism. Other studies have demonstrated that amacrine and horizontal cells are most vulnerable to excitotoxic cell death while bipolar cells are relatively spared (79,80). This may parallel region-specific differences in sensitivity to excitotoxicity in the brain, in which Calcium/calmodulin-stimulated protein kinase II (CaMKII) activity is implicated (81), which in turn regulates glutamate receptors (82). Developmental abnormalities in bipolar cells ensuing from Atrx deletion may result in erroneous glutamate release, either through leakage from defective or degenerating axons, or via aberrant synaptic transmission. Intrinsic developmental cues or visual stimulation upon eye opening may promote the release of glutamate or survival/growth factors from bipolar cells that function to establish retinal tiling, thereby pruning and organizing receptive fields to foster functional circuit formation. The associated loss of amacrine and horizontal cells may reflect those connections made with bipolar cells that did not project to the correct strata. This is supported by a reduction in the levels of neuropeptides with protective effects in the Atrx-null mice and might also explain why the cell death is stabilized after P17. Regardless, further experiments are required to delineate the complexity of the non-cell-autonomous loss of the interneurons.

Conservation among CNS models of ATR-X syndrome

Mutations in genes involved in chromatin remodeling represent a growing cause of intellectual disability yet the underlying mechanisms remain largely unknown. Atrx loss can contribute to cognitive disorders by impairing progenitor cell expansion and neuron production (16,17,53). This study suggests that the role of Atrx in progenitor cell expansion can be uncoupled from a separate functional requirement for Atrx protein in post-mitotic mature neurons. Moreover, the Atrx^ΔE2 defect, which mimics hypomorphic ATR-X syndrome patient mutations, phenocopies the pan-retinal deletion and further supports a critical role for Atrx in differentiated neurons. Analysis of the Atrx^ΔE2 mice also indicates that conditional ablation of Atrx provides a valid method for modeling disease conditions arising from a clinically relevant mutation. Our identification of genes that are coordinately regulated in the Atrx-null retina and forebrain suggests that there may be similar transcriptional activities of Atrx at specific genetic loci implicated in visual and cognitive processing. Such commonalities suggest conserved roles for Atrx in genetic regulation underlying structural and functional integrity, and ultimately neuroprotection, among CNS tissues. Exploiting Atrx-mediated survival pathways affecting post-mitotic neurons offers the prospect of developing therapeutics for disease intervention and functional restoration in the mature brain.

Materials and Methods

Animals

The following Cre driver transgenic mouse lines were crossed with Atrx^flox/flox mice (16) to generate the retinal Atrx conditional knockout mice: Pax6α-Cre (39), obtained from P. Gruss, Max Planck Institute for Biophysical Chemistry, Göttingen, and maintained on a C57BL/6J background; Chx10-GFP/Cre-IRES-AP (38), obtained from C. Cepko, Harvard Medical School, and maintained on a C57BL/6J background; Ptf1a-Cre (32), obtained from Chris Wright, Vanderbilt University, and maintained on a C57BL/6 background; Vsx2-5.3-PRE-Cre (37), obtained from R. Chow, University of Victoria, and maintained on a 129S1/SvImJ background. Male Atrx^flox/y;Cre^+/- progeny were used as conditional knockout mice while male Atrx^flox/y;Cre^-/- littermates were used as controls. Transgenic Atrx^ΔE2 mice (20) were obtained from Hideyuki Beppu, University of Toyama, and maintained on a C57BL/6J background. Wildtype C57BL/6J mice were used as controls for the Atrx^ΔE2 experiments. Wildtype C57BL/6J and 129S1/SvImJ mice were purchased from Jackson Laboratories (Bar Harbor, ME, USA). Genotyping for the Cre transgene was performed by PCR using ear clip or tail genomic DNA preparations and the following primers: Cre-forward, 5′-ATGCTTCTGTCCGTTTGCCG-3′; Cre-reverse, 5′-CCTGTTTTGCACGTTCACCG-3′. Animals were maintained under a 12-h light-dark cycle with food and water available ad libitum. Euthanasia was performed by asphyxiation with CO₂ and/or cervical dislocation. All experiments were approved by the University of Ottawa’s Animal Care Ethics Committee adhering to the guidelines of the Canadian Council on Animal Care.

In vivo retinal electroporation

Subretinal injection and in vivo electroporation of the pCAG-CreGFP plasmid into newborn (P0) mouse pups was performed as previously described (36), with the following modifications: 0.05 mg/kg buprenorphine was administered subcutaneously to dams ∼2 h prior to injection of the pups as well as several hours post-injection, and once daily for the next 2 days. Mouse pups were anaesthetized with isofluorane/O₂ by inhalation prior to surgery. Following electroporation, the pups were oxygenated and warmed in an incubator at 37 °C until recovery from the anesthetic, and then were returned to the mother. Plasmid DNA was transfected into right eyes only. Transfection efficiencies were monitored by detection of GFP-mediated fluorescence of the Cre-GFP fusion protein and were observed to be consistent with previous reports (36).

Immunohistochemistry

Preparation of retinal wholemounts isolated from 4-week-old mice and cryosections of retinal tissues harvested at postnatal day 17 or adult timepoints for immunohistochemical staining was performed as previously described (15,83), with the following modifications. For retinal cryosection preparation, mouse eyes were fixed for 30 min in 4% PFA/0.1 M phosphate-buffered saline (PBS) pH 7.4 at room temperature either within the head or after enucleation. Posterior eyecups were subsequently prepared by removing the cornea and lens, followed by PBS washes, overnight incubation in a 30% sucrose/PBS solution at 4 °C, and cryopreservation in a 1:1 mixture of 30% sucrose/PBS and OCT. Primary antibodies, their sources, and concentrations used are provided in Supplementary Material, Table S1. AlexaFluor- or DyLight-conjugated secondary antibodies raised in donkey were purchased from Life Technologies, Inc. (Burlington, ON, Canada) or Jackson ImmunoResearch Laboratories, Inc. (West Grove, PA, USA) and were used at a dilution of 1:1000 for retinal cryosections and 1:200 for retinal wholemounts.

Microscopy and image processing

Immunolabeled retinal cryosections were analyzed on a Zeiss Axioplan epifluorescence microscope and digital images were captured using an AxioVision 4.6 (Carl Zeiss Inc., Oberkochen, Germany) camera followed by image processing with Adobe Photoshop CS5 software (Adobe Systems Inc., San Jose, CA, USA). Following immunohistochemical staining, retinal wholemounts were viewed and confocal micrographs acquired using a Zeiss LSM 510 Meta Axioplan 2 laser-scanning confocal microscope (Carl Zeiss Inc., Oberkochen, Germany) equipped with UV (405 nm), argon (488 nm) and helium-neon (546 nm, 633 nm) lasers and Plan-Apochromat 40X/1.4 or 63X/1.4 oil immersion objective lenses. Image reconstruction and processing was performed with ZEN 2009 Software (Carl Zeiss Inc., Oberkochen, Germany) or Imaris software (Bitplane AG, Zurich, Switzerland).

Quantitative histological analysis

Cells were counted and retinal layer dimensions were determined manually using epifluorescent light micrographs or confocal micrographs of immunostained retinal cryosections for tissues sampled at postnatal day 17 or retinal wholemounts harvested at 4 weeks of age. For cell counts from cryosections, all immunopositive cells of interest within a 400-µm-wide region of the INL positioned within 200 µm adjacent to the optic nerve head were enumerated for central retina analyses. Retinal sections located at nasal, middle and temporal positions of the eyes were examined. All cell counts reported correspond to analysis of central retina positions in mid-ocular sections. Cell counts were obtained from three adjacent sections for each eye, and mean counts were calculated for the total number of eyes analyzed. For retinal wholemounts, the total number of ChAT-immunoreactive cell somas located throughout the INL and GCL were counted within a 150-µm² area positioned within a central to mid-peripheral location of the retina, corresponding to electroporated regions as determined by abundant GFP immunostaining of cells within the underlying ONL and distal INL, and equivalent locations in transgenic Pax6-AtrxKO retinas. Retinal layer thickness measurements were obtained from 40× magnification confocal micrographs of the central retina region of mid-ocular sections. DAPI staining was used to define the borders of the ONL and INL, while the proximal INL border and distal GCL border were used to define the width spanned by the IPL. Layer widths were determined by drawing a perpendicular line to join the borders of each layer using the ‘Measurement’ tool in the ZEN 2009 software (Carl Zeiss Inc., Oberkochen, Germany). Measurements were taken from three adjacent sections for each eye, and mean layer thickness was calculated for the total number of eyes examined.

Electroretinography

Full-field scotopic electroretinograms were generated using the ESPION system (Diagnosys LLC, Littleton, MA, USA). 6–8-week-old mice were weighed and dark-adapted overnight prior to ERG analysis, and all subsequent procedures were conducted under safe-light conditions. Mice were anaesthetized with an intraperitoneal injection of 1.25% Avertin at a dose of 0.22–0.25 ml/g. 0.5% proparacaine hydrochloride (Alcaine^®, Alcon Canada Inc.) was applied as topical anesthetic to each eye. Eyes were dilated using both 1% tropicamide (Mydriacil^®, Alcon Canada Inc.) and 2.5% phenylephrine hydrochloride (Mydfrin^®, Alcon Canada Inc.). Mice were placed on a warming source to ensure constant body temperature during ERG recordings. Gold wire loop electrodes were placed on both corneas with a drop of 0.3% hypromellose lubricant eye gel (Genteal^®, Alcon Canada Inc.) to maintain corneal hydration. A gold minidisc reference electrode was placed on the tongue and a ground needle electrode was placed subcutaneously in the tail. The animal’s head was positioned under the center of the Ganzfeld dome. Single flash stimuli (4-ms duration) were presented at nine increasing intensities ranging from 0.0025 to 10 (P) cd s/m². OP measurements were recorded for three stimulus intensities ranging from 0.63 to 10 (P) cd s/m². Five ERG traces were obtained and averaged for each luminance step. The minimum negative deflection occurring between 10 and 40 ms post-stimulus was defined as the a-wave peak. The maximum positive deflection occurring between 40 and 80 ms post-stimulus was defined as the b-wave peak. The a-wave amplitude was measured from the baseline to the a-wave trough, and the b-wave amplitude was measured from the a-wave trough to the b-wave peak. The a-wave and b-wave implicit times (latency) were measured from the stimulus onset to the a-wave trough and b-wave peak, respectively. OP measurements were taken for the first three wavelets superimposed on the ascending phase of the b-wave. OP amplitudes were calculated as peak-to-trough measurements for each successive wavelet. A repeated measures ANOVA with a Bonferroni–Dunn correction for multiple comparisons was used to detect differences between Atrx-mutant and wildtype cohorts.

Quantitative RT-PCR

Pooled retinas harvested from the same animal at postnatal day 17 were manually homogenized and total RNA was extracted using TRIzol reagent (Life Technologies Inc.) using 5 mg/ml glycogen (Ambion) as a carrier. DNase I treatment was performed with the DNA-free^TM DNA Removal Kit (Ambion) prior to cDNA synthesis using 1–2 mg of total retinal RNA and the GoScript^TM Reverse Transcription System (Promega Corp., Madison, WI, USA). Control reactions without reverse transcriptase were prepared in parallel. Quantitative RT-PCR was performed using the SYBR Advantage qPCR Premix (Clontech Laboratories, Inc., Mountainview, CA, USA) and a Stratagene Mx3000P qPCR system (Agilent Technologies, Inc.). Reaction parameters were as follows: 30 s at 95 °C followed by 40 cycles of 5 s at 95 °C, 10 s at 60 °C and 10 s at 72 °C, and a final melting curve generated in increments of 0.5 °C per plate read. Sequence-specific primers used for transcript amplification are provided in Supplementary Material, Table S2. All primers were analyzed by melt curve analysis and agarose gel electrophoresis following amplification to assess amplicon purity and specificity. Standard curves were generated for each primer pair using 3-fold serial dilutions of control cDNA. Relative gene expression fold-change values were calculated using the comparative C(T) method (84). Data were normalized to 18S and/or L32 expression levels. Threshold cycle variance in three to six biological replicates was tested for significance using a two-tailed Student’s t-test with equal variance. Error bars represent the standard error of the mean.

Immunoblotting

Total protein extracts were prepared from frozen adult mouse retinas by manual homogenization in ice-cold RIPA lysis buffer (50 mM Tris–HCl pH 7.4, 150 mM NaCl, 1% NP40, 0.1% SDS, 0.5% sodium deoxycholate, 10 mM NaF, 5 mM sodium citrate, 1.5 mM MgCl₂, 10 µM ZnCl₂) supplemented with a protease inhibitor cocktail (cOmplete Mini, EDTA-free; Roche Diagnostics, Mannheim, Germany). Homogenates were incubated for 2 h at 4 °C with gentle rotation, followed by centrifugation at 15 000×g for 20–30 min at 4 °C to sediment insoluble material. Protein samples were quantitated by the Bradford method, resolved by denaturing SDS-PAGE on 3–8% Tris-acetate polyacrylamide gels (NuPAGE; Life Technologies Inc., Burlington, ON, Canada), and blotted onto PVDF membranes (Immun-Blot; Bio-Rad Laboratories, Hercules, CA, USA) by electrophoretic transfer in Tris-glycine buffer containing 10% methanol and 0.05% SDS. Efficiency of protein transfer was monitored by post-transfer staining of gels with 0.2% Coomassie Blue and staining of the PVDF membranes with 0.1% Ponceau S in 5% acetic acid. Membranes were blocked with 5% skim milk in Tris-buffered saline containing 0.05% Triton X-100 (TBST) for 1–2 h at room temperature, followed by overnight incubation at 4 °C with a 1:6 dilution of mouse anti-Atrx antibodies (F39; gift from Douglas Higgs) or a 1:2500 dilution of rabbit anti-vinculin antibodies (ab129002; Abcam Inc, Toronto ON, Canada), and subsequent incubation for 1–2 h at room temperature in a 1:50 000 dilution of sheep anti-mouse IgG (A5906; Sigma-Aldrich Canada Co., Oakville, ON, Canada) or goat anti-rabbit IgG (A4914; Sigma-Aldrich Canada Co., Oakville, ON, Canada) horseradish peroxidase-conjugated secondary antibodies. Membranes were subjected to three 10-min washes in TBST following primary antibody incubation, and three 10-min washes in TBST followed by a 10-min wash in PBS after incubation with secondary antibodies. Enhanced chemiluminescent detection of immunoblot signals was performed using the SuperSignal™ West Femto Maximum Sensitivity Substrate (Life Technologies Inc., Burlington, ON, Canada). Densitometry of immunoblot signals was performed using ImageJ version 1.49 software (http://imagej.nih.gov/ij/). Immunoblots of retinal protein extracts from three to six mice were used for quantitation of Atrx and vinculin protein levels.

Microarray and bioinformatic analysis

Total RNA was isolated from two pooled retinas of each of three male Chx10Cre-Atrx cKO mice and three wildtype (Cre-) male littermates on postnatal day 17. RNA purification was performed using RNeasy MinElute columns (QIAGEN Inc., Toronto, ON) and assessed for quality on a Bioanalyzer (Agilent Technologies, Inc., Santa Clara, CA, USA). The RNA Integrity Number (RIN) for all samples was >8.0. The RNA samples were used to probe Affymetrix Mouse Gene 2.0 ST Arrays through the McGill University and Génome Québec Innovation Centre (Montreal, QC). Raw data were independently processed and analyzed using the FlexArray version 1.6.3 Software Package (Génome Québec, Montreal, QC) and the Expression Console^TM and Transcriptome Analysis Console^TM Software (Affymetrix, Santa Clara, CA, USA). Normalization of the data was performed using the Robust Multi-array Average (RMA) method. Gene level differential expression analysis was performed using a one-way between-subject ANOVA (unpaired) algorithm. Signal intensities correspond to Tukey’s Bi-weight average of gene level intensity of all the samples in either the AtrxKO or wildtype condition (log2). Linear fold-change values are calculated as 2^[AtrxKO bi-weight average signal (log2)− wildtype bi-weight average signal (log2)]. For hierarchical clustering analysis, distances between clusters of objects are computed using the complete linkage method (maximum distance between a pair of objects in the two clusters). GO analysis was performed using the GOrilla gene ontology enrichment analysis and visualization tool (http://cbl-gorilla.cs.technion.ac.il/) (85,86). Batch analysis of GO annotations for gene function groups was carried out using the CateGOrizer GO terms classifications counter web tool (http://www.animalgenome.org/tools/catego/) (87). Microarray datasets are deposited in NCBI’s Gene Expression Omnibus (GEO accession number GSE78877).

Statistics

Pair-wise comparisons between mean values of protein expression levels, cell counts, retinal layer dimensions and fold-change differences in transcript levels measured by qRT-PCR calculated from datasets representing wildtype and Atrx-mutant conditions were performed in Excel 2013 (Microsoft Corp., Redmond, WA, USA) using a two-tailed, two-sample Student’s t-test with equal variance. 95% confidence intervals were used to determine significance of differences. P values <0.05 were considered statistically significant. Error bars represent standard error of the mean. For ERG analysis, a repeated measure ANOVA with a Bonferroni–Dunn correction for multiple comparisons was performed using GraphPad Prism version 6.04 software (GraphPad Software, Inc., La Jolla, CA, USA) to detect differences between Atrx-mutant and wildtype cohorts. For gene expression analysis using DNA microarrays, normalization of the data was performed using the Robust Multi-array Average method in both FlexArray (Génome Québec, Montreal, QC) and Expression Console^TM (Affymetrix, Santa Clara, CA, USA) software. Gene level differential expression analysis was performed using a one-way between-subject ANOVA (unpaired) algorithm. Signal intensities correspond to Tukey’s Bi-weight average of gene level intensity of all the samples in either the AtrxKO or wildtype condition.

Supplementary Material

Supplementary Material is available at HMG online.

Acknowledgements

We thank P. Gruss, C. Wright, R. Chow, C. Cepko and H. Beppu for providing transgenic mice. F. Müller, F. Haeseleer and R. Chow generously provided antibodies and helpful discussions.

Conflict of Interest statement. None declared.

Funding

This study was supported by operating grants from the Canadian Institutes of Health Research (MOP84412; MOP133586) and the Foundation Fighting Blindness-Canada to D.J.P., and by the Clifford, Gladys and Lorna J. Wood Fellowship to P.S.L.