Histone deacetylase expression patterns in developing murine optic nerve

Acetylation and methylation in the murine optic nerve

The presence of acetylated (K18) and methylated (K9) histone 3, representing active and repressed chromatin within the genome respectively, was determined using triple-label immunofluorescence on sections through developing mouse optic nerve. Three different stages of development were chosen for these studies; embryonic day 16 (E16), which is a period when astrocyte precursors are found in the nascent optic nerve, postnatal day 5 (P5) when immature astrocytes are found throughout the optic nerve, including the optic nerve head and developing lamina and P30 when the glia within the lamina, as well as the optic nerve astrocytes and oligodendrocytes are mature [26–28].

Sections through the optic nerve head and adjacent optic nerve were triple-labeled with antibodies against SOX2, which labels glia in the peripapillary, laminar and optic nerve regions, and antibodies against methylated histone H3 (MeH3K9) and acetylated histone H3 (AcH3K18) as markers of closed and open chromatin structure respectively. Within the optic nerve in the low magnification image of a triple-stained P30 section (Figure 1B), SOX2-immuno positive (SOX2+) glial cells located within 100 microns of the optic globe showed the characteristic elongation across the optic nerve that typifies the laminar region; glial nuclei located beyond the lamina were typically rounded or elongated in parallel to the optic nerve [29].

Within the optic nerve, the majority of SOX2+ nuclei were immunopositive for both MeH3K9 and AcH3K18. Along the external surface of the optic nerve, the majority of cells were SOX2 (-) and MeH3K9 (+). Within the population of extraocular SOX2 (-) cells, AcH3K18 staining was co-localized with MeH3K9 in a subset of cells. The SOX2(-), MeH3K9+ cells located furthest from the optic nerve head were most likely non-glial cells associated with the outer surface of the optic nerve that were revealed by the oblique angle of the section through the optic nerve.

At higher magnification, the majority of SOX2+ nuclei within the developing optic nerve at E16 (Figure 1D-G) and within the laminar area at P5 (Figure 1H-K) and P30 (Figure 1L-O) were co-labeled by antibodies for AcH3K18 and MeH3K9. The nuclei of many of the immature astrocytes at E16 were elongated in parallel to the optic nerve, consistent with the migration of astrocyte precursors towards the optic nerve head and retinal nerve fiber layer where they will differentiate into retinal astrocytes (Figure 1G). As development proceeded, nuclei of SOX2+ glial cells located in the optic nerve immediately posterior to the globe manifested the cross-wise elongation characteristic of the developing glial lamina (Figure 1K, O).

mRNA expression levels in the murine optic nerve glia

The presence of acetylated and methylated histones in optic nerve nuclei, coupled with published reports that glial cell fate decisions require deacetylation, led us to ask which HDACs might be involved in this process [30, 31]. To determine which members of the large family of Hdacs were expressed in the optic nerve, reverse transcription quantitative polymerase chain reaction (RT-qPCR) was performed on murine optic nerve RNA. RT-qPCR clearly indicated that mRNAs encoding Hdacs 1–11 were present at E16, P5, and P30 in the murine optic nerve (Figure 2). Each of the Hdac expression levels from E16, P5 and P30 were normalized to the geometric mean of β-2 microglobulin (β2m), signal recognition particle 14kDa (Srp14) and succinate dehydrogenase complex, subunit A (Sdha) and subsequently standardized to E16 levels for comparison (Figure 2A, B). Four expression level patterns emerged following analysis; 1) very little change in levels from E16 to P30 (Hdac1), 2) P5 and P30 levels that decreased relative to E16 (Hdacs 2, 6, and 10), P5 and P30 levels that increased relative to E16 (Hdacs 3, 4, 5, 7, 8, 9), and levels that decreased at P5 but increased at P30 (Hdac11).

SOX2 positive cells in the developing murine optic nerve are astrocytes

In order to investigate the nature of SOX2+ glial cells within the developing murine optic nerve, sections were double labeled using known markers for astrocytes (PAX2) (Figure 3) and oligodendrocytes (NKX2.2) (Figure 4) [32, 33]. The majority of SOX2+ nuclei in the developing optic nerve at E16 (Figure 3A-C), P5 (Figure 3D-F), and P30 (Figure 3G-I) showed co-localization of SOX2 and the type I astrocyte marker, PAX2. Quantification of the cells at all three stages confirmed that a large majority of SOX2+ cells co-expressed PAX2; at E16 95% (±3%) were co-labeled, while 96% (±2%) and 88% (±2.5%) were co-labeled at P5 and P30 respectively (Figure 3J). Counts of the number of co-expressing cells in the proximal nerve (200μm from the exit of the nerve from the eye) at E16, P5 and P30 confirm the qualitative analysis of the sections (Figure 3J) and are consistent with observations that optic nerve astrocytes primarily populate the nerve at locations comparable to the lamina cribrosa in humans.

To examine the presence of oligodendrocyte precursors within the developing murine optic nerve, double-label immunohistochemistry was performed using antibodies that recognize SOX2 and oligodendrocyte precursor marker NKX2.2 [34]. For these experiments, we examined two regions for co-expression of SOX2 and NKX2.2 at P30: proximal optic nerve (up to 200μm from the exit of the nerve from the eye) and distal optic nerve (over 200μm from the optic nerve exit). At E16 (Figure 4A-C) and P5 (Figure 4D-F) no expression of NKX2.2 was detectable in the proximal or distal optic nerve. At P30, NKX2.2+ cells were not observed in the proximal optic nerve (Figure 4G-I) but were present in the distal optic nerve (Figure 4J-L). Quantification of the SOX2+ and PAX2+ cells revealed that 16.2% of the cells within the distal end of the optic nerve at P30 were NKX2.2+ and all of those cells showed co-localization with SOX2 (Figure 4M). These results are consistent with previous publications showing that oligodendrocyte precursors migrate into the rodent optic nerve postnatally [35].

Patterns of Protein Localization in the developing optic nerve

To examine the temporal expression patterns and cellular localization of each HDAC in the proximal optic nerve, western blot and double-label immunohistochemistry were performed. The specificity of each antibody was tested initially by western blot analysis of proteins isolated from mouse optic nerve (Figure 5). Each antibody recognized a single band of the predicted molecular weight. The bands were then quantified by densitometry using β-TUBULIN as a loading control, and the levels of HDACs at each stage were graphed relative to E16 levels. Densitometric analysis indicated that the levels of a majority of HDACs were higher at E16, with falling levels at P5 and P30 (Figure 5). Protein levels showed temporal changes in expression corresponding to mRNA levels for most of the HDACs. Two exceptions to this observation were HDAC3 and 11. In the case of HDAC3, the mRNA decreased with developmental stage, whereas the protein levels increased. HDAC11 RNA increased substantially at P30, whereas the protein levels decreased. In addition, because RT-qPCR detected Hdac7 mRNA in the optic nerve, we attempted to analyze HDAC7 protein expression in these studies. Unfortunately, none of the 5 antibodies tested were specific by western blot, precluding meaningful analysis of HDAC7 protein expression.

To determine the cellular distribution of each HDAC in the developing optic nerve glia, sections of murine optic nerve and nerve head were double-labeled with SOX2 and antibodies specific for each HDAC. The analysis of HDAC immunoreactivity in the optic nerve revealed 3 basic patterns: 1) predominantly nuclear, 2) nuclear and cytoplasmic, or 3) cytoplasmic (see Table 1). Examples of each of the localization patterns are shown in Figures 6, 7, and 8, whereas the rest of the localization patterns are included as supplemental figures (Additional file 1: Figure S1, Additional file 2: Figure S2, Additional file 3: Figure S3, Additional file 4: Figure S4, Additional file 5: Figure S5, Additional file 6: Figure S6 and Additional file 7: Figure S7).

Stages	Predominantly nuclear	Nuclear and cytoplasmic	Predominantly cytoplasmic
E16	1, 2	3, 5	4, 6, 8, 9, 10, 11
P5	1, 2	5, 6, 8, 11	3, 4, 9, 10
P30	1, 2	5, 8	3, 4, 6, 9, 10, 11

HDACs 1 and 2 were predominantly nuclear in localization at all timepoints examined (Figure 6 and Additional file 1: Figure S1). In both cases, there was a high degree of overlap between SOX2+ nuclei and HDAC1 or HDAC2 immunoreactivity. However, at E16, there were also clearly SOX2+ cells at the periphery of the nerve that had little or no HDAC1 labeling (arrowheads, Figure 6E). There were also HDAC1 and HDAC2 + cells present within the optic nerve that were SOX2-negative. Since SOX2 is a marker of glial cells, HDAC+/SOX2-negative cells within the optic nerve itself were presumably endothelial cells or pericytes that are associated with blood vessels within the optic nerve.

HDACs 3, 5, 6, 8 and 11 were predominantly cytoplasmic, but showed nuclear co-labeling with SOX2 at one or more stages of development (Figure 7 and Additional file 1: Figure S2, Additional file 3: Figure S3 and Additional file 4: Figure S4 and Additional file 7: Figure S7). HDAC3 was co-localized with SOX2 in nearly all nuclei within the laminar region at E16, but only a few SOX2+ nuclei were also HDAC3+ at P5 and none of the SOX2+ nuclei were HDAC3+ at P30 (Figure 7). HDAC5 (Additional file 2: Figure S2) and HDAC 11 (Additional file 7: Figure S7) were rarely detected SOX2+ nuclei at P5 and those that were detected were located primarily near the outermost edge of the optic nerve. HDAC6 was co-localized with SOX2 in the laminar glia at P5 (Additional file 3: Figure S3), but was not detected in glia at other timepoints. HDAC8 (Additional file 4: Figure S4) was detected in a few SOX2+ nuclei located near the junction between the lamina and the more proximal regions of the optic nerve at P5 and P30. There was HDAC immunoreactivity in the optic nerve that was not localized to the nuclei and was therefore presumed to be cytoplasmic. Given the extensive non-nuclear staining for HDACs 3, 5, 6, 8 and 11, it is possible that some of the apparent co-localization was a result of overlap between their cytoplasmic staining and the nuclear SOX2 immunoreactivity. However, confocal imaging showed co-localization of SOX2 and the HDACs in both the x/z and y/z planes supporting at least some nuclear localization.

HDACs 4, 9, and 10 were predominantly cytoplasmic, with little to no nuclear label evident at any stage of development (Figure 8 and Additional file 5: Figure S5 and Additional file 6: Figure S6). For HDAC10, the apparent double labeling observed in low magnification image (Additional file 6: Figure S6) was an artifact that was not supported when examined at higher magnification using confocal imaging.

Summary

To summarize the findings of these studies: 1) HDACs 1–11 were expressed in the optic nerve glial cells at all three stages of optic nerve development in the mouse, 2) HDACs 1 and 2 were predominantly nuclear throughout optic nerve development and glial cell maturation, 3) HDACs 3, 5, 6, 8,11 were predominantly cytoplasmic but showed nuclear localization in at least one stage of optic nerve development, 4) HDACs 4, 9 and10 were predominantly cytoplasmic, with little to no nuclear expression at any time during the developmental stages examined, and 5) nuclei co-labeled for AcH3K18 and MeH3K9 were present at similar levels throughout optic nerve development. In interpreting these data, we propose that both acetylation and deacetylation are required for astrocyte development within the optic nerve head and glial lamina. In light of the observation that HDACs 1, 2, 3, 5, 6, 8, and 11 were all co-localized to SOX2+ nuclei at some stages of development, these HDACs would be good candidates for further study to determine which interact with nuclear histones to regulate chromatin organization.

Our finding that HDAC2 is predominately nuclear in cells within the optic nerve differs from a recent report showing immunoreactivity for HDAC2 throughout the optic nerve in the nuclei of resident cells and the cytoplasm of the axons [36]. A major difference between the two studies is the ages of animals analyzed: in our developmental analysis, P30 was the latest age analyzed, whereas the other study analyzed expression in 12 month old mice. There were also regional differences in the portions of the optic nerve analyzed. In the present study, the area of the optic nerve head and glial lamina was analyzed, whereas in the previous report, the morphology of the nuclei would be more consistent with the non-laminar regions of the optic nerve closer to the optic chiasm. The non-laminar portions of the optic nerve contain the progeny of the oligodendrocyte precursors, whereas the laminar astrocytes are thought to derive from the intrinsic PAX2 expressing astrocyte precursors [5, 29]. Therefore, the differences in HDAC2 expression could reflect the different lineages of the glial cells present in these distinct regions of the optic nerve. In addition, different antibodies were used, which could contribute to different patterns of immunoreactivity. It remains to be determined if these differences represent age-related or lineage-related changes in distribution of HDAC2. Interestingly, the patterns of HDAC1 and HDAC3 in the aging optic nerve appeared similar to that observed at P30 in our study, despite the differences in age, location within the optic nerve and in the primary antibodies used.

Many of the HDACs also appear to be localized to the cytoplasm in addition to or instead of the nucleus. The nuclear import of HDACs is blocked by post-translational modification or cleavage [37–44], therefore, the cytoplasmic localization of some HDACs reflects regulated protein trafficking between cellular compartments. In addition, HDACs have also been shown to deacetylate non-histone proteins resulting in changes in their structural and functional characteristics. For instance, HDAC6 interacts with cytoplasmic proteins such as tubulin and chaperones such as heat shock proteins (Hsp90), causing changes in the contractile functions of the muscles. Similarly, HDAC8 can deacetylate heat shock protein 20 (Hsp20) in the myometrium [45, 46]. Hence the cytoplasmic localization of some HDACs may reflect an additional role for HDACs in deacetylation of cytoplasmic proteins during glial differentiation and maturation.

Protein levels showed temporal changes in expression that corresponded to mRNA levels for most of the HDACs consistent with general regulation at the transcriptional level. Two exceptions to this observation were HDACs 3 and 11, which showed increased mRNA levels at timepoints where protein levels were substantially decreased. Many factors can influence the conversion of mRNA into proteins including mRNA and/or protein half-life, post transcriptional regulation by micro RNAs [47, 48]. HDACs are well-known targets of post-translational modifications that can affect their stability and activity, such as proteasome degradation via ubiquitinases or caspase mediated cleavage [49–52]. The E3 ubiquitin ligase SIAH2 interacts directly with HDAC3 and targets it for ubiquitin-dependent proteosomal degradation [53]. Although mechanisms regulating HDAC11 expression have not been characterized, there are reports of non-congruent changes in mRNA and protein levels in the murine brainstem and diencephalon at later stages of postnatal development [54]. Interestingly, in at least one study, HDAC4 mRNA was shown to have a much shorter half-life in comparison to the protein, leading differential increases in protein expression relative to mRNA levels [49]. Together, these findings are consistent with regulation of at least some HDACs at both transcriptional and post-transcriptional levels.

Cells in the developing optic nerve do not appear to be globally acetylated or deacetylated

Several studies have concluded that the genome of differentiating neurons is globally acetylated [30, 55], whereas other studies have proposed that acetylation is the key to the differentiation of oligodendrocytes [54, 56–61]. In non-neuronal cells including glia, recruitment of HDACs by the RE-1 silencing transcription factor (REST) to the promoter of neuronal genes is critical to repress these genes and is thought to be essential for the development of glial cells [62–64]. Some of our data is consistent with the idea that deacetylation contributes to differentiation of astrocytes within the optic nerve head and laminar regions of the optic nerve; we found seven of the classic HDACs (1, 2, 3, 5, 6, 8, 11) are present in the nuclei of SOX2 positive glia in at least one or more stages during the astrocyte differentiation, consistent with a potential role in chromatin regulation. We did not examine the localization of HDAC7 due to lack of a sufficiently specific antibody, or expression of the sirtunins, the Class III HDACs. Therefore additional HDACs may be expressed in the glial precursors and differentiated cells that are not described in this study.

Many of the previous studies have focused on HDACs 1, 2, or 11 during differentiation of oligodendrocytes [54, 59, 65]. In contrast, very few studies have studied the role of epigenetics in astrocyte differentiation, but those that have been done appear to have conflicting results. Several studies using siRNA and morpholino antisense have shown that knocking down HDACs, in particular HDAC1 and 2, alters differentiation of both oligodendrocytes and neurons, without changing the number of astrocytes [30, 31, 61, 66, 67]. A region-specific role for HDACs has been identified in the telencephalon, inhibition of HDAC activity resulted in a BMP mediated decrease in neurogenesis and increase in astrogliogenesis by progenitors arising from the ganglionic eminences of the telencephalon in vivo and in vitro[68]. Further supporting a role in gliogenesis is the discovery that acetylation of histones associated with the STAT3 binding element in the GFAP promoter, a gene expressed typically in mature astrocytes, allows the LIF/STAT3 pathway to upregulate expression of GFAP [69]. Likewise, microarray analysis of neurosphere culture of forebrain cells treated with trichostatin A (to inhibit histone deacetylation) and BMP- (to promote astroglial differentiation) in vitro showed that STAT3 and phospho-STAT3, downstream members of the LIF and BMP pathways, were upregulated [70, 71]. However, in the developing cortex in rats, neurogenesis was enhanced by HDAC inhibition in vivo, showing that there are regional and possibly lineage-specific differences in the functional contributions of HDAC activity to the neuronal/glial switch [68]. Some of the disparities in the studies of HDACs and astrocytes may be attributable to the fact that both in vitro and in vivo experiments are being compared. Epigenetic factors are regulated by cellular microenvironment and placing the cells in vitro has traditionally been a means of testing the effects of changing microenvironment on the cells. Therefore it should not come as surprise that epigenetic factors change when the cells are removed from the “normal” in vivo microenvironment.

The study presented here provides an excellent starting point for investigating the role of histone acetylation and deacetylation in optic nerve development and disease. Our data are consistent with contributions by not only HDAC1 and 2, but other HDACs that have not yet been actively studied. In addition to deacetylation, we also infer that acetylation is an intrinsic component to the differentiation of the laminar astrocytes. This is based on our finding that at all the three stages examined, antibodies specific for acetylated histone 3 immunolabeled increasing numbers of cells with increasing intensity across development. Whereas the studies herein are consistent with HDACs and HATs acting in combination to bring about macroglial differentiation, future studies will examine the roles of various HDACs and HATs at the genome level to ascertain their function at specific loci.

Interestingly, several studies have reported that systemic inhibition of HDAC activity can reduce RGC loss following injuries such as optic nerve crush [72, 73], ischemia [36] and in the Dba/2J mouse model of pigment dispersion glaucoma [74]. However, an as yet unanswered question is how much of the rescue reflects direct effects on histone acetylation in retinal ganglion cells and how much is attributable to effects on the glial populations in the retina and/or optic nerve. Future studies will focus on testing the role of specific HDACs in glial cell development within the optic nerve and retina, as well as investigating potential changes in expression or activity levels in retinal diseases.

Animal husbandry and tissue collection

Animals in this study were obtained and cared for in accordance with the Association for Research in Vision and Ophthalmology (ARVO) statement for the use of animals in ophthalmic and vision research and were housed in an AAALAC accredited housing facility in the School of Science on a 12 hour (hr) light/dark cycle. The ethical treatment of animals used for this study was approved by the IUPUI Science Animal Resource Center (protocol SC194R).

C57BL/6J mice were obtained from Jackson Labs (Bar Harbor, ME). For embryonic stages, mice were mated overnight and the females checked for vaginal plugs the next morning. The date the vaginal plug was detected was designated 0.5 days post coitum. At embryonic day 16 (E16), heads were removed and fixed in 4% paraformaldehyde in 0.1M phosphate buffer, pH 7.4 at 4°C for 24 hr. For postnatal stages, eyes were dissected and fixed in 4% paraformaldehyde in 0.1M phosphate buffer, pH7.4 at 4°C for 2 hrs. Following fixation, eyes were rinsed twice in 1X phosphate buffered saline (PBS; potassium chloride 200 mg/L, potassium phosphate 200 mg/L, sodium chloride 8000 mg/L, and sodium phosphate 1150 mg/L), pH 7.5 and placed in a graded series of sucrose in 0.1M phosphate buffer, starting at 5% and ending with 20%. Eyes were frozen in a 3:1 ratio of 20% sucrose in 0.1M phosphate buffer to Optimal Cutting Temperature (OCT; Sakura FineTek USA, Inc.) solution and stored at -80°C until sectioned.

Immunohistochemistry

Slides were then washed the next day in 1XPBS (2 × 10 min) at room temperature and were then incubated with secondary antibodies [Alexa-fluor conjugated secondaries (Donkey anti rabbit 549 and Donkey anti goat 488) from Invitrogen and DyLight conjugated secondaries (Donkey anti mouse 546) from Jackson ImmunoResearch] dissolved in 1XPBS at 1:800 dilution for 1 hr in dark at room temperature. Secondary antibodies were removed by washing the slides in 1XPBS (2 × 10 min). Sections used to label NKX2.2, the marker used for oligodendrocytes, were treated with biotinylated secondary antibody (BA9200, Vector laboratories) for 1 hr and then washed with 1XPBS (2X10 min). Incubation with Streptavidin 549 (SA-5549, Vector Laboratories) was then performed for 1 hr. Sections were mounted with anti-fade mounting medium containing DAPI (Prolong gold, Life Technologies). In some experiments, antibodies were preabsorbed with 32 μg/ml of the peptide or protein immunogen that was used to generate the antibodies for 1 hr prior to incubation with sections and immunofluorescence performed as usual. Additional sections were incubated with 0.25 μg/ml of IgG in blocking solution in place of the primary antibody and immunofluorescence performed as usual. Negative controls peptides corresponding to each HDAC are shown in Table 2. For all immunohistochemistry analysis, at least 3 independent samples were examined for antibody at each stage. Immunolabeling was imaged using an Olympus Fluoview FV 1000 confocal microscope. For Z stack images, an average of 22 to 25 optical sections were taken and merged together.

Tissue analysis using cell counts

Confocal images of sections co labeled with SOX2 and PAX2, SOX2 and NKX2.2 were taken using FV1000 Olympus Fluoview microscope. Cells showing positive label in the optic nerve were counted over an area of 200 μm at all three stages (E16, P5 and P30) using retina on both sides as a reference. Cells were counted in every third section of through the optic nerve and three different nerves were analyzed for each timepoint.

RT-qPCR

SDS-PAGE and western analysis

Isolated optic nerve tissue samples were lysed using lysis buffer (5M NaCl, 1M Tris pH 8.0, 0.5 M EDTA, 5% Triton-X 100 supplemented with protease inhibitor cocktail (Cat # P50700, RPI Corp) and 1 mM PMSF (Cat # 10837091001, Roche) on ice. Tissue lysates were centrifuged at 14000 rpm for 10 min at 4°C and supernatant was collected. Total protein concentration was determined using BCA protein Assay reagent (ThermoScientific). Proteins were resolved using SDS-PAGE (4-20% gradient gel), transferred to PVDF membranes, and blocked using 5% milk or protein-free tween 20 buffer (cat # 37571; ThermoScientific) and immunoblotted using specific HDAC antibodies. The concentrations of antibodies employed are specified in Table 2. The blots were stripped and re-probed for β-tubulin to check for equal protein loading. Densitometric analyses of the blots were performed using ImageJ 1.43u software Statistics was performed on the densitometric data using a one-way ANOVA with F-test and Scheffe procedure used for post hoc comparisons, with p < 0.05 considered statistically significant (IBM SPSS Statistics 20 software).