Effect of PITX2 knockdown on transcriptome of primary human trabecular meshwork cell cultures

Preparation of primary TM cultures

Four primary cultures were developed from donors without history of eye disease aged 25 (male; TM1), 30 (male; TM2), 65 (male; TM4), and 60 (female; TM5) years old at time of death [12]. Cells were maintained as previously described [12]. The nature of the cells was confirmed by demonstrating increased expression of myocilin mRNA and protein upon dexamethasone treatment (data not shown) [12].

siRNA treatment and RNA extraction

Fourth to sixth passage cells were used in all experiments. Each of the four TM cultures was exposed to three siRNA treatments in duplicate wells. Lipofectamine RNAiMAX (Invitrogen, Carlsbad, CA) was used for reverse transfection according to the manufacturer’s instructions. We had previously shown that this treatment has no detectable effect on cell growth (data not shown). Approximately 8×10⁴ cells were added to wells containing siRNA to achieve a final siRNA concentration of 75 nM. These siRNA duplices (Dharmacon Research, Lafayette, CO) were used: PITX2 siRNA-1 (J-017315–05), PITX2 siRNA-2 (J-017315–06), and scrambled siRNA-1 (D-001810–01–20). All TM cultures had two independent treatments with scrambled siRNA (i.e., two treatments, two wells for each) so as to enable testing of technical replication of the array expression assessments.

Forty eight hours after exposure to siRNAs, cells were harvested and cells of duplicate wells were combined and placed in RNX^TM-plus (Cinnagen, Tehran, Iran). Total RNA was isolated according to the manufacturer’s instructions. RNA quality was assessed using density ratio of 28S to 18S rRNA bands. Half of the RNA was set aside for microarray experiments and the other half was used for real time PCR.

Assessment of knockdown by siRNAs

Candidate control genes for assessment of effects of PITX2 siRNAs were ACTB (actin, beta), β2M (beta-2 microglobulin), GAPDH (glyceraldehyde-3-phosphate dehydrogenase), and HPRT1 (hypoxanthine phosphoribosyltransferase 1). cDNA synthesis was done by standard procedures and real time PCR was performed on a Corbett 65H0 machine (Corbett Research, Sidney, Australia) using the QuantiFast SYBR Green PCR Kit (QIAGEN, Germantown, MD). After selection of appropriate control gene using GNorm program [13], PITX2 siRNA knockdown effects on PITX2 were determined by real time PCR. The real time PCR primers were obtained from QIAGEN.

Microarray experiments and data analysis

Array experiments were performed only on TM1 and TM2 cultures using HumanRef-8 V2 Illumina Genome-Wide Expression BeadChips according to the manufacturer’s instructions (Illumina, San Diego, CA). Arrays were scanned with the Illumina BeadArray reader and images were analyzed using Beadstudio VI software. The Beadstudio data were normalized using a quantile based algorithm available in the Beadarray software [14]. Values for all genes were compared between the two replicate control samples of TM1 and between the two replicate control samples of TM2. Similarly, effect of PITX2 siRNA treatments on three housekeeping genes ACTB, β2M, and GAPDH on the different arrays were determined. Subsequently, genes the expressions of which were affected by siRNA treatments were assessed based on three protocols as described below.

First, the Limma package was used [15]. For each of the four assessments (two siRNAs, two TMs), genes were ordered based on the value of their associated B statistic. Genes were then selected on conditions that they had a B statistic value ≥2 in at least two of the four assessments, that both siRNAs were represented in these assessments, and that the direction of change in expression was the same in the assessments.

In the second protocol, expression in presence of each of two siRNAs was compared separately to each of two controls for each TM, producing eight comparisons for the two TMs. Genes were selected on conditions that they showed ≥2 fold change in the same direction in at least four of the eight comparisons and that each siRNA was represented at least once among the selected comparisons.

In the third protocol, essentially the 40 top ranking genes affected by siRNAs of PITX2 based on fold change in expression were selected. Initially, fold changes in the presence of siRNA1 in each TM were assessed by comparison to average of controls for the respective TM, and forty top ranking genes in each TM were identified (Group A genes: siRNA-1 top 40 TM1 list, siRNA-1 top 40 TM2 list). Subsequently, genes affected were sorted by adding rank of genes affected in the two TMs, and selecting the top 40 genes (Group A genes: siRNA-1 top 40 TM1/2 list). This process was repeated for siRNA-2 (Group B genes: siRNA-2 top 40 TM1 list, siRNA-2 top 40 TM2 list, siRNA-2 top 40 TM1/2 list). Finally, the rankings of genes affected by both siRNAs in each TM were added, and then the rankings of both siRNAs in both TMs were added (Group C genes: siRNA-1/2 top 40 TM1 list, siRNA-1/2 top 40 TM2 list, siRNA-1/2 top 40 TM1/2 list). In this protocol, genes were ultimately selected that appeared at least once in each of the three Groups of genes. A computer program written to perform this selection task is available at Joint-rank.

Real time PCR

Real time PCR was performed on a subset of microarray based selected genes on the two TMs that had been used for microarray analysis and also on TM4 and TM5. cDNA synthesis and real time PCR were performed as described above using QIAGEN primers for ACTB, ALDH1A1, CXCL6, DIRAS3 (DIRAS family, GTP-binding RAS-like 3), DKK1 (Dickkopf-1), KCNJ2 (potassium inwardly-rectifying channel, subfamily J, member 2), MEIS2 (meis homeobox 2), PITX2, and SAMD5 (sterile alpha motif domain containing 5). At least four replicate PCRs for each gene in each TM were performed. Detailed information on all primers is provided in Table 1. ACTB was used as control gene in the real time PCR experiments except where indicated otherwise. Fold change in expression of selected genes were obtained using software provided by the Corbett instrument, used for the real time PCR experiments..

Western blot analysis

TM cultured cells were harvested and lysed in Lameli lysis buffer. Protein (60–100 µg) was electrophoresed on 10%–15% denaturing polyacrylamide gels and the resolved proteins were transferred onto nitrocellulose membranes. The ECL Advance Luminol based chemiluminescence detection kit was used for detection of specific proteins after reaction with appropriate antibodies according to the manufacturer’s instructions (GE Healthcare, Salt Lake City, UT). Specific antibodies used were goat polyclonal antibodies (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) against human myocilin (C-15), Lamin B (C-20) and rabbit polyclonal antibodies (Abcam, Cambridge, MA) against ALDH1A1 (ab51028) and CXCL6 (ab9923). Lamin served as control protein. Secondary antibodies were horseradish peroxidase conjugated anti-goat (SC-2768; Santa Cruz Biotechnology) and anti-rabbit (Sigma-Aldrich, Poole, UK) antibodies. Western blot analysis for each protein was performed twice.

Immunofluorescence analysis on cultured cells and histological sections

TM1 cells grown on coverslips in the presence of scrambled siRNA or PITX2 siRNA-1 were fixed with an acetone-methanol mix, blocked with BSA, incubated with the anti-ALDH1A1 antibody described above, and exposed to fluorescent conjugated goat anti-rabbit secondary antibody (AF8035; Razi BioTech, Tehran, Iran). Nuclei were counterstained with DAPI (Invitrogen). Cells were visualized with an Axioplan 2 fluorescence microscope (Carl Zeiss, Jena, Germany). Additionally, expression of both ALDH1A1 and CXCL6 were assessed in cryosections of a globe from a three year old cadaver.

Comparison of genes identified to genes reported in previous glaucoma relevant array studies

Fifteen microarray gene expression studies in which effects of various glaucoma relevant conditions or treatments were studied, including effects of dexamethsone, pressure, TGFβ, and gene knockdown, have been reported [16-30]. PITX2 affected genes identified here were compared to genes identified in the previous studies to identify possible commonalities.

In silico analysis of selected genes

Promoter regions of selected genes were screened for PITX2 binding sites so as to identify genes whose transcriptions are more likely to be directly affected by this TF. PITX2 binds the bicoid sequence element 5ʹ-TAATCC-3ʹ [31,32]. The region between 3,000 bp upstream and 200 bp downstream of transcription initiation sites of potential target genes was screened for presence of the element [33,34]. Finally, genes identified by the microarray analysis were analyzed to identify potentially relevant functional pathways and functional categories or gene ontology terms annotated by Kyoto Encyclopedia of Genes and Genomes (KEGG) [35] and Gene Ontology (GO) [36] using the Database for Annotation, Visualization, and Integrated Discovery (DAVID) bioinformatics tool [37]. A cutoff p value of 0.05 was used for enriched KEGG pathways or GO functions.

Hydrogen peroxide (H₂O₂) and lithium chloride (LiCl) treatments

H₂O₂ and LiCl treatments were used, respectively, as surrogates for analysis of effects of oxidative stress and Wnt signaling [38,39]. For assessment of these agents on expression of PITX2 and ALDH1A1, TM1 cells were exposed to five treatments and each treatment had its own control. The first treatment was exposure to PITX2 siRNA, and control treatment was exposure to scrambled siRNA. In the second, cells were exposed to 600 mM H₂O₂ for 4 h, and in the third treatment cells had the same exposure to H₂O₂ after having been exposed to PITX2 siRNA-1. The fourth treatment was exposure to 20 mM LiCl for 12 h. In the fifth treatment, H₂O₂ was added to cells that had already been exposed to LiCl for 12 h, and incubation was continued for an additional 4 h. After treatments, real time PCR was performed to assess levels of PITX2 and ALDH1A1 expression. β2M was used as control gene because H₂O₂ has been reported to affect actin expression [39]. The effects of the H₂O₂ and LiCl treatments on ALDH1A1 expression were also assessed by western analysis. Statistical analysis was done using Relative Expression Software Tool (REST) software [40].

Microarray data analysis

Protocols 1, 2, and 3 identified, respectively, 26, 14, and 20 genes affected by PITX2 siRNA treatments, and some genes were common between genes selected by different protocols. In total 41 different genes were identified, and these will henceforth be called “PITX2-filtered genes.” Nine genes were identified by two protocols and five by all three of the protocols. The genes identified are listed in Table 3. The five genes identified by all the protocols were DIRAS3, CXCL6, SAMD5, CBFB (core-binding factor, beta subunit), and MEIS2. Known glaucoma causing genes, CYP1B1 (cytochrome P450, family 1, subfamily B, polypeptide 1), MYOC (myocilin), OPTN (optineurin), and WDR36 (WD repeat domain 36) were not among the genes affected by PITX2 [41]. LTBP2 (latent transforming growth factor beta binding protein 2), a primary glaucoma causing gene, was not probed on the arrays [42].

To confirm microarray based assessments, real time PCR was performed on a subset of genes. Five genes affected by PITX2 siRNAs, ALDH1A1, CXCL6, DIRAS3, MEIS2, SAMD5, and two additional candidate genes, DKK1 and KCNJ2, were tested. The latter two were not among the genes identified by the array analysis protocols; they were selected for assessment by real time PCR based on earlier reports suggesting PITX2 may affect their expressions [17,43,44]. Effects of PITX2 siRNAs as assessed by real time PCR corroborated the array data of most genes tested in the sense of the genes being down-regulated or upregulated (Figure 2). The only exception was MEIS2 in TM2. Notably, the effects on the genes observed in TM1 and TM2, were also observed in TM4 and TM5 which originated from globes of older individuals. DKK1 and KCNJ2 that had not been selected by criteria of the array analysis protocols were also shown by real time PCR to be affected by PITX2 knockdown. Although the array data on these two genes did not meet the protocols selection criteria, the array data did in fact indicate some down-regulation consistent with the real time PCR results (Figure 2).

CXCL6 and ALDH1A1, were further analyzed by western blotting, and the results evidenced that PITX2 siRNAs affected decreased expression of CXCL6 and ALDH1A1 at the protein level (Figure 3A). Furthermore, decreased expression of ALDH1A1 protein in PITX2 siRNA treated cells was also shown by immunofluorescence analysis of the treated cells (Figure 3B-G). Immunofluorescence on globe sections showed expression of both genes in the trabecular meshwork, and additionally in the stroma and Descemet membrane (Figure 3H-S). Highest expressions of CXCL6 and ALDH1A1 were observed, respectively in the iris epithelium and corneal epithelium.

Of the 41 PITX2-filtered genes identified by analysis of array data, 11 were also identified in one or more previously reported glaucoma related global gene expression studies (Table 4). Notably, DIRAS3 and CXCL6 that had shown the two highest fold changes in our experiments (3.9X and 3.3×, respectively; Table 3) and that had both been selected on the basis of all three selection protocols, had also been repeatedly reported in other studies [17-20]. ADAMTS5 that was identified by only one of our selection protocols was previously reported in three studies [17,21,22]. DKK1 and KCNJ2 which were shown to be affected by PITX2 siRNAs by real time PCR experiments were each previously reported in one study [16,22].

In silico analysis of selected genes

The PITX2 binding element was identified within regions surrounding the transcription initiation sites of some of the affected genes. Of the 41 PITX2 filtered genes, 18(43.9%) contained two or more of the binding sequences (data not shown). For example, DKK1 and MEIS2 had, respectively, four and three binding sites. Table 5 lists KEGG and GO terms that are enriched among the 41 PITX2 affected genes.

Oxidative stress and Wnt signaling

PITX2 siRNA decreased PITX2 (−9.1×) and ALDH1A1 (−4.3×) expression as already described above (Figure 4A). As ALDH1A1 is known to be involved in the oxidative stress response of the eye, effect of H₂O₂ on PITX2 and ALDH1A1 expression was tested. Under the condition of acute exposure to H₂O₂ used here, real time PCR showed that it decreased both PITX2 (−3.2×) and ALDH1A1 (−3.9×) expression (Figure 4A). Effect of H₂O₂ on ALDH1A1 expression at the protein level was also evident (Figure 4B). Extent of down-regulation of ALDH1A1 upon simultaneous exposure to PITX2 siRNA and H₂O₂ (−5.1×) was comparable to effect of H₂O₂ alone (p=0.566), suggesting that H₂O₂ and PITX2 may affect ALDH1A1 expression by a common pathway (Figure 4A). It is known that Wnt signaling increases PITX2 expression in various organs including the eye [45]. Lithium mimics the Wnt pathway by suppressing GSK-3 activity and LiCl is commonly used in experimental settings to produce this effect [38]. Human TM cells possess a functional Wnt signaling pathway and Wnt signaling is implicated in glaucoma [46]. Here, it was observed that LiCl indeed enhanced PITX2 expression (+2.2×). Furthermore, it seemed to moderately affect increased expression of ALDH1A1 as well, both at RNA (+1.6×; Figure 4A) and protein (Figure 4B) levels. Pretreatment with LiCl partly prevented H₂O₂ induced down-regulation of both PITX2 (−2.3× versus −3.2×; p=0.226) and ALDH1A1 (−1.6× versus −3.9×; p=0.076; Figure 4A). Although the differences did not achieve statistical significance, these preliminary results are suggestive and deserve to be pursued. The observations are consistent with the proposal that ALDH1A1 may at least partly be affected by the Wnt signaling pathway and that effects on ALDH1A1 expression may be via PITX2.