Effect of VCP modulators on gene expression profiles of retinal ganglion cells in an acute injury mouse model

In glaucoma, retinal ganglion cells are damaged, leading to the progressive constriction of the visual field. We have previously shown that the valosin-containing protein (VCP) modulators, Kyoto University Substance (KUS)121 and KUS187, prevent the death of retinal ganglion cells in animal models of glaucoma, including the one generated by N-methyl-D-aspartate (NMDA)-induced neurotoxicity. KUSs appeared to avert endoplasmic reticulum (ER) stress by maintaining ATP levels, resulting in the protection of ganglion cells from cell death. To further elucidate the protective mechanisms of KUSs, we examined gene expression profiles in affected ganglion cells. We first injected KUS-treated mice with NMDA and then isolated the affected retinal ganglion cells using fluorescence-activated cell sorting. Gene expression in the cells was quantified using a next-generation sequencer. Resultantly, we found that KUS121 upregulated several genes involved in energy metabolism. In addition, we observed the upregulation of Zfp667, which has been reported to suppress apoptosis-related genes and prevent cell death. These results further support the suitability of KUS121 as a therapeutic drug in protecting retinal ganglion cells in ophthalmic disorders, such as glaucoma.


Purification of retinal ganglion cells by FACS.
To study the effects of the KUSs on the affected cells, we first isolated the CFP expressing retinal ganglion cells from dissociated whole neural retina of Thy1-CFP mice using FACS. The CFP-positive cells were found to account for 0.04-0.12% of all retinal cells (Fig. 2a, area The relative expression levels of (a) jun proto-oncogene (Jun), (b) FBJ osteosarcoma oncogene (Fos), (c) v-rel reticuloendotheliosis viral oncogene homolog A (Rela), (d) capsase-3 (Casp3), (e) nuclear factor of kappa light polypeptide gene enhancer in B cells inhibitor, alpha (Nfkbia), (f) tumor necrosis factor (Tnf) and (g) interleukin 6 (Il6) mRNA were analysed by qRT-PCR. The ratios of mRNA expression of each gene to that of actin were calculated. The ratios to actin at 2 h, 4 h and 6 h were divided by those at 0 h. 1); no CFP-positive cells were observed in the retina of wild-type mice (Fig. 2b). Retinal cells expressing both brain-specific homeobox/POU domain protein 3A (Brn3a) and paired box protein (Pax) 6 are defined as retinal ganglion cells. By immunostaining the dissociated cells, we confirmed that Brn3a-and Pax6-positive retinal  S1). 88.9-93.4% of the sorted cells were included in area 1 (CFP-positive). (d) The sorted cells were stained with anti-brain-specific homeobox/POU domain protein 3A (Brn3a, green) and anti-paired box genes 6 (Pax6, red) antibodies. Nuclei were counter stained with 4' ,6-diamidino-2-phenylindole (DAPI, blue). Scale bar: 50 µm. (e) mRNA expression of synaptosomal-associated protein 25 (Snap25), tubulin, beta 3 class III (Tubb3), RNA binding protein with multiple splicing (Rbpms) and rhodopsin (Rho) in cells sorted by gate RGCs or gate PR (see Methods and Supplementary, Fig. S1b) were visualised with the Integrated Genome Browser. Snap25, Tubb3 and Rbpms were highly expressed in Gate RGCs and not in Gate PR while Rho was highly expressed in Gate PR and not in Gate RGCs (black arrows). RGC: retinal ganglion cell, PR: photoreceptors. ganglion cells accounted for 0.01-0.12% of the total retinal cells. The percentage of the retinal ganglion cells estimated with the immunocytochemical analysis was almost the same as the FACS analysis.
FACS from 2 retinas of 2 Thy1-CFP mice enabled us to sort 1,494-3,550 CFP-positive cells (see method  and Fig. S1, gate RGCs) in 30-50 min, which was considered sufficiently fast to collect fresh cells to analyze mRNA expression. Re-analysis of the sorted cells showed that 88.9-93.4% of the cells were CFP-positive, which indicates that the FACS sorting effectively collected and concentrated the CFP-positive retinal cells (Fig. 2c). Immunostaining of the sorted cells showed that almost all the sorted cells were Brn3a-and Pax6-positive and were retinal ganglion cells ( Fig. 2d and Fig. S2). To confirm that the collected cells were indeed retinal ganglion cells, their mRNAs were visualised with the Integrated Genome Browser 35 . We confirmed expected mRNA expression profiles: high expression of synaptosomal-associated protein (Snap25) 36 , and tubulin, beta 3 class III (Tubb3), which are expressed in neuronal cells and RNA binding protein with multiple splicing (Rbpms) 37 , which is expressed in retinal ganglion cells and low expression of rhodopsin (Rho), which is expressed in rod photoreceptors. These results further validated our FACS sorting protocol (Fig. 2e, gate RGCs in Fig. S1). Consistent with the above, CFP-negative cells collected by gate PR in Fig. S1, expressed high Rho levels without Snap25, Tubb3, and Rbpms expression, indicating that they contained rod photoreceptors and not ganglion cells (Fig. 2e, gate PR). From these data, we assumed the CFP-positive cells sorted by FACS, successfully enriched retinal ganglion cells and were suitable for the next experiments. mRNAs related to gene expression and energy metabolism were upregulated in KUS-treated retinal ganglion cells. Using analysis of variance (ANOVA), 255 genes showed significant (P < 0.01) expression changes among four conditions; non-treatment (non-treat), vehicle with intravitreous injection of NMDA (NMDA-saline), KUS121 treatment with intravitreous injection of NMDA (KUS121), and KUS187 treatment with intravitreous injection of NMDA (KUS187) (Supplementary Table S1). Hierarchical clustering analyses of the 255 genes distributed samples between the conditions (Fig. 3a,b). Hierarchical clustering analyses of the ANOVA-passed genes revealed that ANOVA successfully selected genes that distinguish each condition (Fig. 3a). Moreover, the samples of the three experimental repeats showed similar patterns of upregulated and downregulated genes on heatmap (Fig. 3b). These results showed that the experimental repeats displayed great reproducibility of the gene expression profiles within each condition. While some genes showed similar expression patterns between the NMDA-saline and KUSs-treated groups, other genes showed clearly differential expression patterns (Fig. 3b). KUSs-treated groups were clearly separated from saline-treated groups. These data indicated the KUSs-treated groups have characteristic gene expression profiles distinct from the non-treat or NMDA-saline groups.
Gene ontology (GO) analysis 38 was performed to annotate genes into biological ontology. After X-means clustering, genes in hyper cluster A and B were analyzed ( Supplementary Fig. S3). Genes in hyper cluster A were found to be associated with 31 GO terms, which included RNA metabolic processes, biosynthetic processes, gene expression and metabolic processes ( Table 1, Supplementary Table S2) while genes in hyper cluster B were not associated with enrichment of any GO terms.
Of the genes studied, we focused on Zfp667, which has been reported to suppress apoptosis-related genes and consequently prevent cell death in ischemia-reperfusion injury 41,42 . Western blot analysis of mouse retinal proteins showed that expression levels of Zfp667 was not significantly different between non-treated retinas and saline-treated NMDA-injected retinas (NMDA-saline as control) of wild-type mice (P = 0.61, Turkey HSD, Fig. 5a,b). In contrast, expression of Zfp667 was significantly increased in KUS121-treated NMDA-injected retinas compared to the NMDA-saline group (P = 0.004, Turkey HSD, Fig. 5a,b). Immunohistochemical analysis with an anti-Zfp667 antibody showed higher expression of Zfp667 predominantly at the retinal ganglion cell layer in KUSs-treated NMDA-injected retinas compared to NMDA-saline injected retinas (Fig. 5c).

Discussion
In this study, we successfully isolated retinal ganglion cells to a high level of purity using FACS and found that KUSs affect the expression of a wide variety of genes. These include genes involved in the regulation of energy metabolism and suppression of apoptosis in the retinal ganglion cells of an NMDA-induced retinal injury model mouse.
There have been studies in which neural cells including retinal ganglion cells were isolated using FACS with retrograde labelling 43,44 or with antibodies 45 . In the current study, we used Thy1-CFP transgenic mice 29,30 , in which CFP is expressed in retinal ganglion cells, to omit the need for antibody reaction or the retrograde labelling step. It enabled us to collect the cells simply and quickly, which were suitable for the analysis of mRNA profiles. www.nature.com/scientificreports www.nature.com/scientificreports/ In the preliminary experiments to decide the timing for evaluating gene expression after intravitreous NMDA injection, neural retina of wild-type mice was used. While intravitreous NMDA injection has been reported to damage RGCs 19 , whole retinas of mice that received intravitreous NMDA injection were used for mRNA evaluation by qRT-PCR 46 . Hence, we used neural retina in the preliminary experiments. Moreover, among the 18 genes evaluated by qRT-PCR using neural retina, Jun and Fos were included in the 255 genes evaluated by next-generation sequencing using sorted RGCs that showed significant changes in expression among the four conditions. The expression of Jun and Fos genes was upregulated by 1.63 and 3.59 times, respectively, in NMDA treated mice RGCs compared to non-treated controls.
Our experiments showed that KUS121, which has already been used in a clinical trial for ischemic retinal disease (UMIN000023979), upregulated gene expression and translation of Zfp667 in the retinal ganglion cells. Zfp667 has been reported to suppress apoptosis-related genes in ischemia-reperfusion injury 41,42 . KUS121 also upregulated the expression of a variety of genes such as Phkb, which is involved in glycogen metabolism 47 ; Ppargc1a, which is a strong activator of mitochondrial function and a regulator of energy metabolism [48][49][50] ; Ptcd2, which is involved in mitochondrial gene expression 51 ; Npm1, which promotes cell survival under stress 52 ; Dusp18, which may dephosphorylate and inactivate mitogen-activated protein kinase (MAPK) 47 ; Peg10, which is anti-apoptotic 53 . On the other hand, genes that were downregulated by KUS121 included Top3a, which adjusts the DNA topological states during transcription 47 . Whether the translation of these genes is also altered remains to be clarified. In addition to reducing ATP consumption, these results revealed the possibility that KUS121 www.nature.com/scientificreports www.nature.com/scientificreports/ prevents retinal ganglion cell death through several mechanisms, including activating energy production and suppression of apoptosis. These mechanisms could be related to the modulation of VCP function or alternatively, KUSs may have additional targets, which could be involved in transcriptional control of cell survival. These possibilities need to be elucidated by further studies.
In conclusion, KUS121 can modulate gene expression profiles in retinal ganglion cells in mice, via mechanisms not yet fully elucidated, which are likely to contribute to protecting the retinal ganglion cells from NMDA-induced neurotoxicity. This study further strengthens the suitability of KUS121 as a therapeutic drug in rescuing retinal ganglion cells in eye diseases that are currently incurable, such as glaucoma.

Methods
Experimental animals. This study was conducted in accordance with the Association for Research in Vision and Ophthalmology (ARVO) Statement for the Use of Animals in Ophthalmic and Vision Research. All protocols were approved by the Institutional Review Board of Kyoto University Graduate School of Medicine (MedKyo 12245, 13221, 14213, 15531, 16501). B6.Cg-Tg(Thy 1-CFP) 23Jrs/J mice were obtained from the Jackson Laboratory (Bar Harbor, ME, USA) and wild-type mice (C57/BL6), which share the genetic background of Thy1-CFP mice, were purchased from Japan SLC, Inc. Mice were kept in a 14 h light/10 h dark cycle and fed ad libitum. Male mice aged 2 to 3 months were used for the experiments. Before intravitreous NMDA injection (5 nmol), mice were anesthetised with intraperitoneal pentobarbital (50 mg/kg) injection and pupils were dilated with tropicamide and phenylephrine eye drops (0.5% each).
Quantitative RT-PCR of neural retinae. Changes in mRNA expression in the neural retina were examined at several time points after intravitreous NMDA injection. NMDA was intravitreally injected into wild-type mice to induce acute damage of retinal ganglion cells 17,19 . Eyeballs were enucleated 2, 4, and 6 h after NMDA injection after pentobarbital overdose. Enucleated eyeballs were immersed in cold Hanks' balanced salt solution www.nature.com/scientificreports www.nature.com/scientificreports/ immediately after enucleation. Incisions were made using pinholes in the corneas, then using the incisions, the sclera was peeled to remove the mixture of the retinal pigment epithelium, choroid and sclera from the neural retina as previously described 16 . The lens and iris were removed. RNA was isolated from the neural retina using the  Table 2. Pathways which were common in NMDA-saline group < non-treat group, KUS121 group > NMDAsaline group and KUS187 group > NMDA-saline group (fold change > 2 or fold change < 0.5, respectively). saline: NMDA-saline group, K121: KUS121 group, K187: KUS187 group. In fold change analysis, >2 indicates more than 2 folds while <0.5 indicates less than 0.5 fold. For pathway analysis, A < B means statistically significant pathways of genes that were downregulated in group A by less than 0.5 fold compared to group B, and A > B means statistically significant pathways of genes that were upregulated in group A by more than 2 folds compared to group B. www.nature.com/scientificreports www.nature.com/scientificreports/ RNeasy Mini Kit (QIAGEN, Venlo, Netherlands). The mRNA was reverse transcribed with the M-MLV reverse transcriptase (Promega, WI, USA) and then complementary DNA was amplified by PCR with SYBR premix Ex Taq polymerase (Takara Bio Inc., Shiga, Japan) and 60 °C as the annealing temperature on the 7300 Real-Time PCR System (Applied Biosystems, CA, USA). Eyes were enucleated before and 2, 4 and 6 h after NMDA injection and each eye was analysed separately.

Administration of KUSs and preparation of cell suspension for flow cytometry. Daily KUS121,
KUS187 (50 mg/kg/day each), or vehicle (5% Cremophor EL (Sigma) in phosphate buffered saline (PBS)) as a control were given orally to Thy1-CFP mice using a feeding tube. Seven days after the start of the medication, NMDA (5 nmol) was intravitreally injected with a 33-gage needle 54 . Our experiment involved one-week pre-treatment with KUSs because we previously showed that it protects against NMDA injury 17 . Four hours after the NMDA injection, the retinas were collected as described in the qRT-PCR section and incubated in 0.2% papain solution (two retinas of two mice each) including glucose (1 mg/mL), DNase 1 (Worthington, 100 U/mL), superoxide dismutase (Worthington, 5 µg/mL) and catalase (Sigma, 5 µg/mL) at 8 °C for 30 min and at 28 °C for 9 min. The solution was centrifuged at 100 G for 5 min at 4 °C, cells were resuspended in a solution containing antipain (Roche, 50 µg/mL), and then centrifuged again at 100G for 5 min at 4 °C 55 . The cell pellet was resuspended in 500 µL of ice-cold Ames' medium (with L-glutamine, without sodium bicarbonate, Sigma).

Collection of retinal ganglion cells by flow cytometry.
The cells in suspended Ames' medium described above were sorted by FACS Aria 2 (BD Biosciences) based on the size and intensity of CFP fluorescence. Immediately after 7-Amino-Actinomycin D (BD Biosciences, Cell Viability Solution, 20 µL) were added,

Description (possible) Function
Zfp667 may play a role in protecting cells against ischemia-reperfusion injury 41,42 Phkb plays a role in glycogen metabolism 47 Ppargc1a activate mitochondrial function [48][49][50] Ptcd2 plays a role in post-transcriptional expression of the mitochondrial genome 51  www.nature.com/scientificreports www.nature.com/scientificreports/ sorting was performed using a 85-micron nozzle into Ames' medium at 4 °C. Forward scatter (FSC) and sideward scatter (SSC) were used to segregate retinal ganglion cells. CFP-positive cells were contained in the high FSC sub-population (area 2 in Supplementary Fig. S1a). To collect retinal ganglion cells with high purity, only cells included in both gate 4 ( Supplementary Fig. S1b), which is a smaller area than area 2, and gate 5 ( Supplementary  Fig. S1c), which was narrower than the area 1 in Fig. 2a-c, were collected (gate RGCs). For comparison, cells included in both gate 6 ( Supplementary Fig. S1b) and gate 7 ( Supplementary Fig. S1c), which were considered to be photoreceptors, were collected in the same way (gate PR). cDNA synthesis, amplification and next-generation sequencing. The sorted cells (1,494-7,848 cells for each sample) were centrifuged at 500G for 5 min at 4 °C, suspended in 100 µL of buffer B (Prelude Direct Lysis Module, NuGEN), centrifuged again at 500G for 5 min at 4 °C, and the pellet was resuspended in 1 µL of buffer A (Prelude Direct Lysis Module, NuGEN). The lysates were then taken forward for cDNA synthesis and amplification using Ovation RNA-Seq System V2 (NuGEN) according to the manufacturer's instructions. The amplified cDNA was purified using MinElute Reaction Cleanup Kit (QIAGEN, Venlo, Netherlands). The quality of the amplified cDNAs was analyzed by a 2100 Bioanalyzer (Agilent) and was high enough for sequencing. The concentration of cDNA as measured by Qubit (Invitrogen) was sufficient (84.7-185.0 ng/µL). The amplified cDNA was then sequenced with a next-generation sequencer (Illumina HiSeq). The experiments were repeated three times.
Analysis of the RNA-sequencing results. The results of RNA-sequencing were mapped on reference sequence (mouse mm10, USCS genome browser) using TopHat2. Reads which formed reasonable pairs (on the same chromosome, two directionally, and distance between pair reads < 500 k b.p.) were used to calculate expression levels (61.1 ± 8.5% of reads). Reads Per Kilobase of exon model per 10 Million mapped reads (RPK10M) www.nature.com/scientificreports www.nature.com/scientificreports/ were calculated for terminal exons with the in-house scripts as the expression values for genes, and transferred into log 2 scale. Genes whose maximum expression values were more than 3 among the conditions were considered as expressed genes and used for the following analyses. Then the expression value was normalized using quantile normalization methods 56 . Using analysis of variance (ANOVA), 255 genes (including isoforms) showed significant changes in expression among the four conditions (non-treat, NMDA-saline, KUS121 and KUS187, p < 0.01). Hierarchical clustering of the 255 genes that passed the ANOVA test with approximately unbiased p-values and bootstrap probability value of mRNA expression was performed with pvclust package in the statistical environment R using Euclidean distances. A heatmap was drawn with Z-value transferred expression values. Expression profiles of ANOVA-passed genes were transferred to Z-scores and clustered with the x-means method. X-means clustering was performed on the statistical environment R. The genes were divided into 12 clusters that were further categorized according to expression changes across conditions. The upregulation and downregulation of gene expression were defined as the difference in the cluster centers between conditions with more than 0.2 and less than −0.2 z-values, respectively ( Supplementary Fig. S3). The clusters with upregulated expression in NMDA-saline compared to non-treat and downregulated expression in KUS121 and KUS187 compared to NMDA-saline were defined as hyper cluster A. The clusters whose expression was upregulated in NMDA-saline compared to non-treat and whose expression was upregulated in KUS121 and KUS187 compared to NMDA-saline, were defined as hyper cluster B. GO analysis of the genes included in hyper cluster A (55 genes) and B (12 genes) was performed based on hypergeometric distribution.
Pathway analysis was performed by Gene Spring14 (TOMY Digital Biology) using Wiki Pathways (http:// www.wikipathways.org/index.php/WikiPathways). Genes that were upregulated (115 genes) or downregulated (57 genes) in the NMDA-saline group compared to the non-treat group, and those that were upregulated (61 genes) or downregulated (73 genes) in the KUS121 group compared to the NMDA-saline group as well as upregulated (50 genes) or downregulated (53 genes) in the KUS187 group compared to the NMDA-saline group (fold change > 2 or fold change < 0.5, respectively) were analyzed. The level of statistical significance was set to P < 0.05.
In fold change analysis, >2 indicates more than 2 folds while <0.5 indicates less than 0.5 fold. For pathway analysis, A < B means statistically significant pathways of genes that were downregulated in group A by less than 0.5 fold compared to group B, and A > B means statistically significant pathways of genes that were upregulated in group A by more than 2 folds compared to group B.

Immunocytological evaluation of cells. Dissociated retinal cells or FACS sorted cells were fixed by
adding an equal amount of 4% paraformaldehyde, centrifuged at 3000 rpm for 15 min at 4 °C. After the extra supernatant was removed, the cell suspension (200 µL) was centrifuged at 1000 rpm for 10 min using Cytospin (Thermo Scientific) to be pasted onto slides. The cells were stained with anti-Brn3a (CHEMICON) and anti-Pax6 (COVANCE) antibodies and imaged under an optical microscope (Axio Imager.A1, Zeiss).
Immunohistological evaluation of retinas. Non-treated eyeballs or NMDA-injected eyeballs of mice treated with KUS121, KUS187 or vehicle (saline) were enucleated after pentobarbital overdose. A marking dye (Davidson) was placed on the edge of the superior conjunctiva to identify the superior portion of the retina as previously described 57 . The eyes were fixed in 4% paraformaldehyde for 24 h at 4 °C, embedded in O.C.T. compound (Sakura Finetek Japan) and frozen. Serial 16 µm O.C.T-embedded sections were cut through the dye and at the point of insertion of the optic nerve. Sections that included the center of the optic nerve head were stained with an anti-ZFP667 antibody (GeneTex) and imaged under an optical microscope (BZ-9000, Keyence) at a distance of 400 µm from the edge of the optic nerve head. The fluorescence intensity of each eye was measured in 40 µm × 300 µm squares including the retinal nerve fiber layer, the ganglion cell layer, and the inner plexiform layer using BZ II Analyzer software (Keyence).
Western blotting of neural retinas. Neural retinas were prepared as described in the qRT-PCR section. Neural retinas and wild-type mouse brain, which was used as a positive control, were analysed with an anti-ZFP667 antibody (GeneTex). Actin was used as a loading control. The ratio of ZFP667 to actin was compared between the KUS121-treated and the control retina using an unpaired t-test.

Data availability
All data generated or analyzed during this study are included in this published article and its Supplementary Information files.