Microarray-based gene expression analysis combined with laser capture microdissection is beneficial in investigating the modes of action of ocular toxicity.

The retina consists of several layers, and drugs can affect the retina and choroid separately. Therefore, investigating the target layers of toxicity can provide useful information pertaining to its modes of action. Herein, we compared gene expression profiles obtained via microarray analyses using samples of target layers collected via laser capture microdissection and samples of the whole globe of the eye of rats treated with N-methyl-N-nitrosourea. Pathway analyses suggested changes in the different pathways between the laser capture microdissection samples and the whole globe samples. Consistent with the histological distribution of glial cells, upregulation of several inflammation-related pathways was noted only in the whole globe samples. Individual gene expression analyses revealed several gene expression changes in the laser capture microdissection samples, such as caspase- and glycolysis-related gene expression changes, which is similar to previous reports regarding N-methyl-N-nitrosourea-treated animals; however, caspase- and glycolysis-related gene expressions did not change or changed unexpectedly in the whole globe samples. Analyses of the laser capture microdissection samples revealed new potential candidate genes involved in the modes of action of N-methyl-N-nitrosourea-induced retinal toxicity. Collectively, our results suggest that specific retinal layers, which may be targeted by specific toxins, are beneficial in identifying genes responsible for drug-induced ocular toxicity. ©2022 The Japanese Society of Toxicologic Pathology.

Introduction

Ocular toxicity, including toxicity in the retina and choroid, is a serious problem in the research and development of medical drugs owing to the associated decreased quality of life among patients. The retina is a light receptor tissue of the eye. It consists of several layers, such as the outer nuclear layer (ONL) and the rod and cone photoreceptor layer (PRL)[1], [2]. Each layer is highly specialized; therefore, each layer can be a specific target for retinal toxicants. For example, ethambutol induces selective degeneration of the ganglion cell layer[3], whereas the primary target of sodium iodate is the retinal pigment epithelium[4], [5]. The choroid, which exists under the retinal pigment epithelium and feeds the retina, is a target of toxicity from drugs, such as hydroxychloroquine[6]. The characteristics of layer-specific toxicity in the retina and choroid can make it difficult to investigate the mechanism of ocular toxicity via analyses using whole globe (WG) or striped retina samples owing to dilution of biologically significant signals.

Laser capture microdissection (LCM) has been used to collect target cells and tissues. This technique can be used to overcome the difficulty of investigating molecular changes in each layer of the eye. Microarray-based global gene expression analysis has been widely used to obtain large amounts of transcriptional information on the mechanism of toxicity. Although a combination of these techniques seems preferable in investigating the mechanism of ocular toxicity, such reports are limited.

The purpose of this study was to confirm whether microarray analysis of retinal samples collected via LCM is beneficial compared with analysis of samples from the WG of the eye with the cornea, lens, and vitreous body removed. We compared gene expression levels between these two different samples obtained from rats treated with N-methyl-N-nitrosourea (MNU), which is known to selectively induce toxicity in the ONL and PRL[7], [8], [9]. We also discuss new insights into the mechanism of MNU-induced retinal toxicity.

Materials and Methods

Chemicals, animals, and experimental design

MNU was purchased from Sigma-Aldrich Chemical Co. (St. Louis, MO, USA). Twenty-four 6-week-old male Crl: CD Sprague Dawley rats were obtained from Charles River Laboratories Japan, Inc. (Kanagawa, Japan) and acclimated for 1 week. All rats were fed a normal rodent diet (CRF-1, Oriental Yeast Co., Ltd., Tokyo, Japan) and provided with tap water ad libitum. At 7 weeks of age, the rats were given either MNU dissolved in saline containing 0.05% acetic acid (60 mg/kg body weight, n=12) or saline containing 0.05% acetic acid (n=12). After MNU or vehicle dosing, the rats were necropsied at 6 h (n=6 for both the MNU- and vehicle-treated rats) or 7 days (n=6 for both the MNU- and vehicle-treated rats). A dose of 60 mg/kg was selected on the basis of previous reports, in which MNU was administered to female Sprague Dawley rats at or near this dose[8], [10]. The time point of 6 h was set on the basis of previous reports, in which MNU-induced apoptosis in the ONL was completely inhibited by nicotinamide treatment at 4 h after MNU dosing but not at 6 h[10]; this suggested that retinal toxicity due to MNU was irreversible, and some gene expression changes were induced at this time point. The time point of 7 days was used to confirm whether apparent morphological changes were noted in the retina. All experimental procedures were performed in accordance with the in-house guidelines of the Institutional Animal Care and Use Committee of Daiichi Sankyo Co., Ltd. (approval number: A1600654).

Collection of eye samples

At the time of necropsy, all rats were euthanized via exsanguination under anesthesia with isoflurane using an anesthesia apparatus for small animals (MK-A110, Muromachi Kikai Co., Ltd., Tokyo, Japan). After euthanasia, the eyes were collected. The right eyes of the rats necropsied at 6 h after MNU or vehicle dosing were embedded in O.C.T. Compound (Sakura Finetek Japan Co., Ltd., Tokyo, Japan) to prepare the LCM samples (n=3 for both the MNU- and vehicle-treated rats) or frozen in liquid nitrogen to be used as the WG samples after macroscopic removal of the cornea, lens, and vitreous body (n=3 for both the MNU- and vehicle-treated rats). The frozen samples were stored at −80°C for subsequent analysis. The left eyes of the rats necropsied at 6 h and the bilateral eyes of the rats necropsied at 7 days (n=6 for each of the four groups) were fixed with Davidson’s fluid and embedded in paraffin for histopathology.

Histopathological examination

Histopathological specimens were prepared from the paraffin-embedded eyes, stained with hematoxylin and eosin, and examined under a light microscope. The noted histopathological changes were graded as 0 (absent), 1 (minimal), 2 (moderate), and 3 (severe) based on the severity and extent of the change, as described below. The reviewing pathologists (K.Y. and A.S.) confirmed the appropriateness of the findings and their grades (grade 0: no apparent changes; grade 1: less than 20% of the ONL and PRL are affected; grade 2: 20% to 80% of the ONL and PRL are affected; and grade 3: more than 80% of the ONL and PRL are affected).

LCM sampling

Frozen tissue specimens were sectioned at a thickness of 8 μm and mounted on RNAse-free foil slides (Leica Microsystems, Tokyo, Japan). After hematoxylin staining, the ONL and PRL were sampled together using a Leica LMD7000 system (Leica Microsystems). Each sample had an area of 8 mm2 or more in total so that microarray analysis could be conducted. To minimize RNA degradation, we limited the time from staining the sections to sampling by LCM to 60 min.

RNA isolation, quantification, and qualification

The WG and LCM samples were homogenized using TRIzol® Reagent (Thermo Fisher Scientific, Carlsbad, CA, USA), and total RNA was isolated. The quality of RNA, including RNA integrity number (RIN), was assessed using a 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). The concentration of RNA was assessed using a NanoDrop system (Thermo Fisher Scientific, Inc., Waltham, MA, USA).

Microarray-based gene expression analysis

Complementary DNA and then biotin-labeled complementary RNA were synthesized from the total RNA using the GeneChip® IVT Express kit (Affymetrix, Inc., Santa Clara, CA, USA), in which an oligo-dT primer was contained, according to the manufacturer’s instructions. Each biotin-labeled cRNA sample (approximately 10 μg) was individually hybridized to a GeneChip® Rat Genome 230 2.0 Array (Affymetrix, Inc.) at 45°C for 16 h, followed by washing and staining with streptavidin-phycoerythrin using the Fluidics Station 400 (Affymetrix, Inc.). The scanned image was analyzed with the MAS5 algorithm using the GCOS software (Affymetrix, Inc.). All MAS5-analyzed data were scaled via global normalization. Genes whose mean signal intensity values were significantly different from those of the concurrent control with a p-value of <0.05 and whose detection calls were not absent were analyzed using QIAGEN’s Ingenuity Pathway Analysis (IPA) software (Qiagen, Tokyo, Japan). Three types of functional analyses using IPA were performed: canonical pathway analysis, upstream regulator analysis, and downstream effector analysis. In the individual gene expression analyses, genes with mean signal intensity values at least 1.2-fold higher/0.83-fold lower than those of the concurrent control were regarded to have significantly changed.

Statistical analysis

In the microarray data analysis, the homogeneity of variance was evaluated using the F test, and the Student’s t-test (for homogeneous data) or the Aspin–Welch t-test (for heterogeneous data) was performed using Microsoft Office Excel 2003 (Microsoft, Redmond, WA, USA). In the canonical pathway, upstream regulator, and downstream effector analyses, the z-score was calculated using IPA. We used conservative criteria with a p-value of <0.05 to judge statistical significance in terms of gene expression and a z-score of ≥2 or ≤−2 to judge significant changes in the IPA functional analyses.

Results

Histopathological examination showed reproducible retinal toxicity by MNU treatment

MNU did not induce any histopathological changes at 6 h after treatment (Table 1 and Fig. 1). However, similar to previous findings[7], [8], [9], severe depletion of the ONL and PRL was evident in all MNU-treated rats at 7 days.

Table 1.

Histopathological Scores of the Retinal Lesions

Fig. 1.

Histopathological appearance of typical lesions in the retina at 6 h or 7 days after MNU administration.

A: The retina of a rat at 6 h after vehicle treatment. B: The retina of a rat at 6 h after MNU treatment. C: The retina of a rat at 7 days after MNU treatment. No apparent histopathological changes were noted in the retina at 6 h after MNU administration, while the outer nuclear layer and rod and cone photoreceptor layer disappeared at 7 days after MNU administration. GCL, ganglion cell layer; IPL, inner plexiform layer; INL, inner nuclear layer; MNU, N-methyl-N-nitrosourea; OPL, outer plexiform layer; ONL, outer nuclear layer; PRL, rod and cone photoreceptor layer; PEL, pigment epithelial cell layer. Hematoxylin and eosin staining; bar=50 μm.

RNA from the WG and LCM samples could be used for microarray-based gene expression analysis

To investigate the degree of RNA degradation, we evaluated the RIN. In the WG samples, the RIN ranged from 8.3 to 8.7 (mean: 8.5), while in the LCM samples, it ranged from 3.4 to 6.5 (mean: 4.84); however, the value could not be calculated for one of the LCM samples. Therefore, RNA from the LCM samples was considered to be partially degraded. To confirm whether RNA degradation from the LCM samples affected the results of the microarray analysis, we evaluated the signal intensities of the probes hybridizing the 5 and 3 ends of glyceraldehyde-3-phosphate dehydrogenase (GAPDH). As shown in Fig. 2, the signal intensities of GAPDH were comparable between the probes at the 5 and 3 ends, and the signal intensity ratio of the probe of the 5 end to that of the 3′ end was 0.8 to 0.9, suggesting that RNA degradation in the LCM samples did not strongly affect the results of the microarray analysis.

Fig. 2.

Signal intensities of the probes hybridizing the 5′ and 3′ ends of GAPDH and their ratio in the samples collected via LCM

Blank and solid bars indicate the signal intensities of the probes hybridizing the 5′ and 3′ ends, respectively, and the values are indicated by the leftmost numbers. Triangles indicate the ratios of the signal intensities of the probes on the 5′ end to those on the 3′ end, and the values are indicated by the rightmost numbers. The numbers in parentheses at the end of each sample indicate the RNA integrity numbers of the samples. GAPDH, glyceraldehyde-3-phosphate dehydrogenase; LCM, laser capture microdissection.

Veh 1 to 3: Individual LCM samples collected from the vehicle-treated rats (n=3); MNU 1 to 3: Individual LCM samples collected from the MNU-treated rats (n=3).

Canonical pathway analysis indicated an inflammatory response in the WG samples

In terms of the number of genes upregulated or downregulated by MNU treatment, the LCM samples showed fewer genes than the WG samples (Table 2). The results of the canonical pathway analysis, which was used to determine which pathways were activated or suppressed in the gene expression profiles by the drug treatment, suggested that the pathways, such as the eukaryotic initiation factor 2 (EIF2) signaling pathway and the ras homolog gene family, member A (RhoA) signaling pathway, were upregulated by MNU treatment in the LCM samples (Table 3). Meanwhile, upregulation of the inflammatory pathways, such as the inducible nitric oxide synthase (iNOS) signaling pathway, triggering receptor expressed on myeloid cells 1 (TREM1) signaling pathway, tumor necrosis factor receptor 1 (TNFR1) signaling pathway, high mobility group box-1 protein (HMGB1) signaling pathway, inflammasome signaling pathway, and a pathway related to Fcγ receptor-mediated phagocytosis in macrophages and monocytes, upregulation of EIF2 signaling, and downregulation of peroxisome proliferator-activated receptor signaling were predicted in the WG samples.

Table 2.

Number of Genes with Expression Level Changes after N-methyl-N-nitrosourea Treatment

Table 3.

Significantly Upregulated and Downregulated Pathways in the Canonical Pathway Analysis after MNU Treatment

Upstream regulator and downstream effector analyses also suggested differential changes between the LCM and WG samples

The upstream regulator analysis, which was used to indicate the upstream regulators responsible for the observed gene expression profiles, showed different upstream regulators between the LCM and WG samples (Table 4). In the LCM samples, the activating effects of regulators, such as X-box binding protein 1, avian myelocytomatosis viral-related oncogene, neuroblastoma-derived, and fms-related receptor tyrosine kinase 1, and the suppressive effect of regulators, such as mitogen-activated protein kinase kinase kinase kinase 4 and estrogen receptor 1, were suggested by MNU treatment. Meanwhile, in the WG samples, inflammatory factors were suggested as upstream regulators. The activating effect of an inflammatory cytokine (interferon gamma), an inflammation-associated growth factor (platelet-derived growth factor BB), an adaptor molecule of toll-like receptors (toll-like receptor adaptor molecule 1), and an inflammatory substance (Salmonella minnesota R595 lipopolysaccharide) and the suppressive effect of an adhesion molecule (alpha catenin), Rho, and an immunosuppressant (sirolimus) were predicted in the WG samples.

Table 4.

Top 10 Significantly Changed Activated or Suppressed Pathways Predicted in the Upstream Regulator Analysis after MNU Treatment

The downstream effector analysis, which was used to predict the biophysiological effectors led by the gene expression profiles, also suggested different results between the LCM and WG samples (Table 5). In the LCM samples, 9 of the top 10 functions predicted to be activated by MNU treatment were related to infection. Cell death, RNA synthesis, and glycolysis were expected to be suppressed. In the WG samples, functions related to cell survival, proliferation, death, and movement were activated, and the quantity of photoreceptors was suppressed. Moreover, there were no overlapping upstream regulators or downstream biophysiological functions between the LCM and WG samples, at least in the top 10 items.

Table 5.

Top 10 Significantly Changed Activated or Suppressed Pathways Predicted in the Downstream Effector Analysis after MNU Treatment

Genes related to phosphoinositide-3-kinase (PI3K)-AKT and calcium signaling may be differentially expressed between the LCM and WG samples

The results of the canonical pathway and upstream regulator analyses suggested that the inflammatory pathways changed selectively in the WG samples. To understand the differences in further detail, we investigated individual gene expression changes induced by MNU treatment with respect to the genes contained in the iNOS, TREM1, TNFR1, and HMGB1 signaling pathways, the inflammasome pathway, and the pathway related to Fcγ receptor-mediated phagocytosis in macrophages and monocytes, from the canonical pathway analysis using IPA. As shown in Table 6, increased expression levels of phosphoinositide-3-kinase regulatory subunit 2 (PI3KR2) and phosphoinositide-3-kinase regulatory subunit 3 (PI3KR3) and decreased expression levels of phosphatidylinositol-4-phosphate-3-kinase catalytic subunit type 2 alpha (PIK3C2A) and AKT serine/threonine kinase 2 (AKT2) were observed in the LCM samples, suggesting the suppression of PI3K-AKT signaling by MNU treatment. In addition, the changes in the mRNA expression levels of cAMP response element binding protein (CREB) and calmodulin 1 suggested that MNU affected calcium signaling. Caspase 3 and caspase 9 were also upregulated in the LCM samples. Meanwhile, inflammation-related genes, such as toll-like receptor 4 (TLR4), interferon gamma receptor 2, intercellular adhesion molecule 1, tumor necrosis factor receptor superfamily member 1 (TNFRSF1) A, tumor necrosis factor receptor superfamily member 11B (TNFRSF11B), interleukin 1 receptor type 1 (IL1R1), myeloid differentiation primary response 88, nuclear factor kappa B subunit 1 (NFKB1), Jun proto-oncogene, AP-1 transcription factor subunit, and Fos proto-oncogene, AP-1 transcription factor subunit were upregulated in the WG samples; meanwhile, most of these genes were unchanged or downregulated in the LCM samples (Tables 6 and 7).

Table 6.

Inflammatory Genes with Significantly Changed Expression Levels in the LCM Sample or Both the LCM Sample and WG Sample after MNU Treatment

Table 7.

Inflammatory Genes with Significantly Changed Expression Levels in the WG Sample after MNU Treatment

Caspases were discriminatingly expressed in the LCM samples

Given the changes in the expression levels of caspase 3 and caspase 9 in the LCM samples, the expression levels of other caspases were also investigated. In the LCM samples, the mRNA expression levels of caspases other than caspase 3 and caspase 9 remained unchanged (Table 8). In contrast, in the WG samples, the expression level of caspase 3 did not change, while the expression level of caspase 9 showed the opposite change compared with that in the LCM samples. In addition, increased expression levels of caspases 1, 4, 6, 8, and 12 were noted.

Table 8.

Caspases with Significantly Changed Expression Levels in the LCM and/or WG Samples after MNU Treatment

Certain genes related to PI3K-AKT and calcium signaling selectively changed in the LCM samples

Since the results of the individual gene expression analyses of the inflammatory genes suggested changes in PI3K-AKT and calcium signaling by MNU treatment in the LCM samples, the genes contained in the signaling pathways from the canonical pathway analysis using IPA were investigated. For the genes related to PI3K-AKT signaling, increased expression levels of phosphatidylinositol-4,5-bisphosphate-3-kinase catalytic subunit alpha (PIK3CA) and nuclear factor kappa B inhibitor beta (IKB) and decreased expression levels of Janus kinase 1, SOS Ras/Rac guanine nucleotide exchange factor 1, SOS Ras/Rac guanine nucleotide exchange factor 2, and mechanistic target of rapamycin kinase, as well as changes in PI3KR2, PI3KR3, PIK3C2A, and AKT2 shown in the previous section, were noted in the LCM samples (Tables 6 and 9). In the WG samples, these genes did not change significantly. Meanwhile, upregulation of NFKB1 and nuclear factor kappa B (NFKB) and downregulation of AKT3, PIK3CA, and the KRAS proto-oncogene, GTPase (KRAS), were observed. Investigation of the mRNA expression levels of the genes in the calcium signaling pathway revealed that upregulation of calcium/calmodulin-dependent protein kinase 1 G (CAMK1G), calcium/calmodulin-dependent protein kinase 2 beta (CAMK2B), calcium/calmodulin-dependent protein kinase 2 gamma (CAMK2G), and cAMP response element binding protein 3 (CREB3) and downregulation of cAMP response element binding protein 1 (CREB1), as well as changes in the expression of CREB binding protein and calmodulin 1 mentioned in the previous section, were noted in the LCM samples (Table 10). In the WG samples, changes in CAMK1G, CAMK2B, CAMK2G, and CREB binding proteins were not observed or were the opposite of those in the LCM samples. In addition, increased expression levels of cAMP response element binding protein 5 (CREB5) and decreased expression levels of calcium/calmodulin-dependent protein kinase 2 delta were noted in the WG samples.

Table 9.

PI3K-AKT Signaling-related Genes with Significantly Changed Expression Levels in the LCM and/or WG Samples after MNU Treatment

Table 10.

Calcium Signaling-related Genes with Significantly Changed Expression Levels in the LCM and/or WG Samples after MNU Treatment

Distinguishable glycolysis-related gene expression changes were observed in the LCM samples

Based on the estimation of the changes in the glycolysis in the LCM samples from the downstream effector analysis, the mRNA expression levels of the genes involved in the glycolysis of cells and the glycolysis of tumor cell lines from the downstream effector analysis were investigated. In the LCM samples, the genes related to glycolysis, such as 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 2 (PFKFB2), hexokinase 2, pyruvate dehydrogenase E1 alpha 1 subunit (PDHA1), GAPDH, phosphoglucomutase 1 (PGM1), glucose-6-phosphate isomerase (GPI), and pyruvate dehydrogenase phosphatase catalytic subunit 1 (PDP1) were upregulated, while glycogen synthase kinase 3 beta (GSK3B) was downregulated (Table 11). Most of these genes were unchanged or downregulated in the WG samples.

Table 11.

Glycolysis-related Genes with significantly Changed Expression Levels in the LCM And/or WG Samples after MNU Treatment

Discussion

Ocular toxicity profoundly worsens patients’ quality of life, and once it occurs, it becomes a serious problem in the research and development of medical drugs. The retina consists of several thin layers, which can be separately affected by drug toxicity[5], [6], [7], [8]; therefore, more useful information can be obtained by investigating target layers compared with the homogenate of the WG of the eye or retina. Because it is difficult to separately collect each layer of the retina manually, except when it comes to the retinal pigment epithelium[11], LCM, which is an established technique for collecting the intended part of the tissues from histopathological sections, is considered to be beneficial when investigating the modes of action (MOA) of ocular toxicities. Herein, we investigated whether microarray analysis of target retinal layers collected via LCM could be used to detect reproducible changes due to MNU toxicity in line with previous studies and obtain additional information to speculate on the MOA of retinal toxicity compared with the analysis of WG samples.

First, the reproducibility of retinal toxicity due to MNU treatment was confirmed. All of the rats in this study showed severe depletion of the ONL and PRL at 7 days after treatment with 60 mg of MNU, which is similar to previous findings[7], [8], [9]. Therefore, significant gene profile changes were considered to be induced in all of the samples at 6 h after the treatment, at which time the samples for the microarray analysis were obtained, as MNU-induced retinal toxicity became irreversible at this time point[10]. Second, the RNA integrity was confirmed by the RIN; however, the RIN ranged from 3.4 to 6.5 (mean: 4.84) in the LCM samples, and RNA degradation was suspected. Hence, we investigated the signal intensities of the different probes for GAPDH with reference to the 3′:5′ assay on GAPDH to confirm the RNA integrity; this is based on the logic that with oligo-dT primer, cDNA can be synthesized only from the mRNA with poly(A), so the possibility of cDNA synthesis from the mRNA sequence near the 3′ end (poly[A] side) should be higher than the 5′ end if mRNA is degraded[12]. As a result, the ratio of the signal intensity of the probe on the 5 end to that on the 3′ end of GAPDH was 0.8 to 0.9 in the LCM samples, suggesting that degradation was unbiased in the RNA sequence and therefore did not affect the relative gene expression intensity among the genes. Indeed, it has been reported that formalin-fixed samples, whose RIN was lower than that of the LCM samples in our study (2.2 to 2.4), showed gene expression profiles similar to those of frozen samples with a higher RIN (n=3, Pearson’s correlation coefficient=0.876–0.904)[13]. Therefore, the RNA degradation noted in the LCM samples was judged to have limited substantial effect on the gene expression profiles, and the microarray-based gene expression analysis was continued.

The comparison of the gene expression changes induced by MNU treatment revealed clearly different profiles between the LCM and WG samples. For example, the canonical pathway analysis, upstream regulator analysis, and biophysiological effector analysis predicted mostly non-overlapping pathways. In particular, inflammatory pathways were notably predicted in the WG samples but not in the LCM samples; therefore, one of the major differences in the gene expression changes between the LCM and WG samples would be expected to be related to inflammatory genes. Indeed, the individual analyses of inflammatory gene expression changes demonstrated that the genes related to PI3K-AKT signaling were upregulated in the LCM samples, while the expression levels of genes, such as TLR4, TNFRSF1, TNFRSF11B, and IL1R1, increased in the WG samples. MNU treatment activates Müller cells, which are glial cells in the retina, and Müller cells express toll-like receptors, such as TLR4, and produce cytokines, such as TNF-α and interleukin-1β[14], [15], [16]. The cytosol of Müller cells exists from the outer limiting membrane to the inner limiting membrane, including the ONL, but not the PRL[14], [15]. In addition, the retinal pigment epithelium produces various cytokines[17], and treatment with ultraviolet-irradiated photoreceptor outer segments and human complement sera results in increased cytokine production[18]. Therefore, it is speculated that the changes in the inflammatory gene expression selectively noted in the WG samples arose from Müller cells and/or the retinal pigment epithelium, which do not exist in the main target layers of MNU toxicity; further, the gene expression changes noted only in the LCM samples were considered to be more useful in speculating on the MOA of toxicity.

Apoptosis in photoreceptor cells has been noted in MNU-treated rats at 24 h and was completed by 7 days after treatment[8], and caspases have been reported to be involved in retinal degeneration caused by MNU[9], [19]. In this study, the gene expression profiles of the LCM and WG samples were different. In the LCM samples, caspases 3 and 9 were upregulated. Caspase 3 is activated by caspase 9[20], [21], and a caspase 3 inhibitor weakens MNU-induced retinal toxicity in rats[9]; therefore, the increased expression levels of caspase 3 and caspase 9 in the ONL and PRL were considered to be related to MNU-induced retinal toxicity. Meanwhile, in the WG samples, there were no apparent changes in caspase 3, while caspase 9 was downregulated. In addition, increased expression levels of caspases 1, 6, 8, and 12, which are related to inflammation and/or microglial activation[21], [22], [23], [24], were also noted. Hence, in the WG samples, changes in the expression of caspases related to inflammatory reactions may have been apparent since these samples contained Müller cells and retinal pigment epithelium, while specific changes, such as increases in the expression of caspases 3 and 9, could not be detected in the target retinal layers.

Individual gene analyses showed changes in the expression of genes related to PI3K-AKT signaling in both the LCM and WG samples. Although the changes in PIK3C and AKT were shared, upregulation of IKB, a suppressor of NFKB[25], was observed only in the LCM samples, while an increased expression level of NFKB was noted only in the WG samples. In the MNU-treated rats in the study by Miki et al., the level of phospho-NFKB significantly decreased, and the level of phosphorylated IKB did not significantly change in the homogenate of the whole retina[26]; therefore, NFKB activity is suspected to contribute to MNU-induced retinal toxicity. Hence, our results showed the possibility that the ONL/PRL-specific decrease in IKB expression contributes to the suppression of NFKB signaling and retinal toxicity caused by MNU.

In addition to PI3K-AKT signaling, calcium signaling was also suggested to be affected by MNU treatment. Upregulation of calmodulin and calcium/calmodulin-dependent protein kinase (CAMK) was observed in the LCM samples, while upregulation of calmodulin and downregulation of CAMK were noted in the WG samples. CAMK2, a kinase activated by calmodulin, leads to an apoptosis response[27]. Glutamate has been shown to decrease cell viability in a differentiated human retinoblastoma-derived cell line and increase CAMK2 activity[28]. Furthermore, a mouse model harboring a mutation in a rod photoreceptor-specific phosphodiesterase gene with rapid degeneration of the cells has been shown to increase CAMK2 activity[29]. Based on these data, it is possible that ONL- and PRL-specific increases in the gene expression levels of calmodulin and CAMK contribute to MNU-induced retinal toxicity.

The IPA downstream effector analysis predicted changes in glycolysis, and the individual gene analysis demonstrated increased expression levels of the glycolysis-related genes, such as HX1, PDHA1, GAPDH, PGM1, GPI, PFKFB2, and PDP1, as well as decreased expression levels of GSK3B in the LCM samples, but not in the WG samples. MNU becomes a DNA adduct to become toxic, and DNA adducts are formed specifically in the ONL in MNU-treated rats[19], [30]. DNA adduct formation induces poly (ADP-ribose) polymerase, which results in adenosine triphosphate (ATP) depletion by using nicotinamide adenine dinucleotide for DNA repairment[19]. Additionally, decreased glycogen and increased glycolysis enzymes, such as hexokinase, glucose-6-phosphatase, and glucose-6-isomerase, are induced by MNU in the brains of mice, suggesting complementary glucose metabolism, including glycolysis, to synthesize ATP[31]. Hence, the upregulation of the glycolysis-related genes selectively noted in the LCM samples was considered to be a reactive change against DNA adduct formation in the ONL by MNU treatment. Meanwhile, the expression of GSK3B, which encodes an inhibitory enzyme of glycogen synthesis[32], [33], was suppressed, while no change in glycogen synthetase expression was noted in the LCM samples. Therefore, decreased GSK3B expression may increase glycogen synthesis by interrupting the complimentary glycolysis against MNU toxicity. Moreover, it has been reported that GSK3B activity in the retina of MNU-treated rats decreased and that GSK3B activator treatment ameliorated MNU-induced retinal toxicity[34]. Collectively, it is speculated that one of the MOA of MNU-induced retinal toxicity is the inhibition of glycolysis for ATP supplementation by increased glycogen synthesis due to decreased GSK3B expression in the ONL and PRL.

Although we evaluated the usefulness of a combination of retinal layer sampling via LCM and microarray analysis, there are alternatives for collecting samples and investigating retinal layer-specific transcriptome changes. Separately collecting each layer of the retina manually under a light microscope immediately after necropsy can obtain more intact RNA than using LCM, as shown by the higher RIN of the WG samples in this study, although it is difficult to perform, except for the retinal pigment epithelium[11]. Single-cell RNA sequencing technologies may provide more detailed and precise bioinformatics information of individual cells in target layers than the method used in our study; however, it takes a longer time to establish optimal methods to collect each cell, optimize and standardize analytical protocols, and interpret large amounts of data[35]. Considering that retinal toxicity is mostly unexpected and rapid investigation would be required, especially when toxicity is induced in studies for developing drugs, the combination of LCM sampling and microarray analysis is considered to be the first selected method owing to the high flexibility of LCM sampling and well-established methods of microarray analysis during almost 20 years of history[35].

As mentioned above, we judged that the RNA quality measured by the RIN was acceptable for microarray analysis; however, there may be several methods to improve the RNA quality of LCM samples obtained from the retina. We performed sampling via LCM at room temperature; thus, the RNA in the section was considered to be degrading during sampling. If this process can be performed in a cold room, RNA degradation should be reduced. In addition, considering the throughput to collect samples at once, the working time from staining sections to collecting samples via LCM was set as 60 min at maximum in our study. Higher-quality samples may be obtained if the maximum working time is shortened. Hematoxylin staining was used as the staining method. This was determined in the in-house preliminary trials comparing the RNA quality of samples stained by hematoxylin to toluidine blue, which is recommended in the manufacturer’s instructions (data not shown); however, there may be room for optimizing the staining method to obtain high-quality RNA samples.

In conclusion, our results suggest that LCM and microarray investigations of selective toxicity of specific retinal layers may be useful in analyzing the MOA of retinal toxicity, although RNA degradation may occur during sample preparation and/or collection via LCM. The MNU-induced gene expression changes noted in the LCM samples, such as those involving caspase- and glycolysis-related genes, were similar to those previously reported in MNU-treated animals. Moreover, the analyses of the LCM samples revealed new potential candidates involved in the MOA of MNU-induced retinal toxicity, such as upregulation of the IKB and CAMK genes, and inhibition of glycolysis by decreasing GSK3B gene expression. Most of these changes were not detected in the WG samples. Collectively, our results strongly suggest that microarray analysis of specific LCM-collected retinal layers, which may be affected by specific toxins, is beneficial and that it is possible to find MOA-related genes, which cannot be detected in the analysis of WG samples.

Disclosure of Potential Conflicts of Interest

The authors have no conflicts of interest to declare.