Downstream Effects of ROCK Signaling in Cultured Human Corneal Stromal Cells: Microarray Analysis of Gene Expression

Abstract

purpose. Rho-associated coiled-coil-containing protein kinase (ROCK) is a downstream target of Rho GTPase signaling and regulates the assembly of stress fibers. Previous reports indicate that Rho/ROCK signaling is involved in the regulation of several cellular processes, some of which may be cell-type specific and are probably critical to corneal stromal cell activation. The present study identified ROCK-regulated gene expression in corneal stromal cells.

methods. Corneal stromal cells derived from eyes of three different donors were cultured to yield the following designated phenotypes: baseline fibroblasts (DMEM with 10% serum), activated fibroblasts (10% serum+bFGF+heparin), and myofibroblasts (1% serum+TGF-β1). Cells were exposed to the ROCK inhibitor Y-27632 or vehicle for 12 hours, and transcript levels altered by ROCK inhibition were identified with oligonucleotide microarrays (GeneChips; Affymetrix, Santa Clara, CA).

results. In these phenotypes, Y-27632 caused marked (twofold or more) increases or decreases in 14/4, 12/3, and 15/10 transcripts. In both fibroblast groups Y-27632-treatment increased expression of endothelin receptors and of parathyroid hormone-like hormone. The upregulation of α-smooth muscle actin in myofibroblasts was attenuated by Y-27632. Combining data from all groups identified ROCK-supported (Y-27632 inhibitable) expression of 10 transcripts, including ribonucleotide reductase M2, the cyclin B1-CDC2-CKS2 system, and four mitotic spindle-associated proteins.

conclusions. ROCK inhibition causes broad inhibition of DNA synthesis and mitosis and causes changes that are different between (bFGF-activated) fibroblasts and (TGF-β1–induced) myofibroblasts. Thus, Rho/ROCK signaling regulates both common and distinct downstream events in corneal stromal cells activated (differentiated) to fibroblast or myofibroblast phenotype.

Rho, a homologue of small GTPase Ras, is involved in many actin-associated cellular processes leading to nuclear signaling (for reviews, see Refs. ^{1 2 3 4 5 6 7 8 9 10 11} ). The recent identification and characterization of several Rho-binding proteins illuminates the biochemical mechanisms of specific downstream effects. One of the Rho targets is Rho-associated coiled-coil-containing protein kinase (ROCK), a kinase that is involved in mediating the assembly of focal adhesions and stress fibers. Two isoforms of (ROCK-I/ROKβ and ROCK-II/ROKα) ^{12 13 14 15} have been identified. ROCK phosphorylates a myosin phosphatase and thereby suppresses its phosphatase activity, ^{16 17} which leads to increases in the levels of phosphorylated myosin and induction of actomyosin-based contraction. ROCK has been shown also to phosphorylate and thereby activate LIM kinase. ^{18 19} LIM kinase can phosphorylate and inactivate an actin depolymerizing protein, switching off actin filament disassembly. We have reported that increased expression of ROCK is associated with limbal epithelial transition to corneal epithelium. ¹³ ROCK also influences the assembly of E-cadherin adherence junctions, and downregulation of its activity promotes the assembly of connexin43 gap junctions. ²⁰  

The relatively quiescent keratocytes of the corneal stroma are activated during wound healing to proliferative fibroblasts and then to contractile myofibroblasts. ²¹ TGF-β1 can induce the phenotypic transition of fibroblasts to myofibroblasts. By removing TGF-β1 and providing bFGF and heparin in the medium the myofibroblasts lose expression of α-smooth muscle actin (α-SMA) and regain the fibroblast phenotype. ²² The transition of keratocytes to fibroblasts is associated with increased actin filament and stress fiber assembly, and the transition to myofibroblast is accompanied by synthesis of α-SMA and development of a more robust network of stress fibers. ^{21 23} Rho/ROCK signaling regulates actin filament assembly and is probably a key element in these keratocyte transitions. Moreover, ROCK-I is essential for centrosome positioning, ²⁴ suggesting a direct and stringent relationship between ROCK and mitosis. Interleukin (IL)-1α–dependent expression changes in corneal fibroblasts have been surveyed successfully using DNA microarrays. ²⁵ To determine the gene expression effects of Rho/ROCK signaling, we inhibited ROCK activity in corneal fibroblasts and myofibroblasts and used oligonucleotide microarrays to measure the resultant changes in mRNA levels. 

Methods and Materials

Cell Culture and RNA Isolation

Human donor eyes deemed unsuitable for transplantation were obtained from Center for Organ Recovery and Education (CORE; Pittsburgh, PA), in accordance with the provisions of the Declaration of Helsinki for research involving human tissue. Corneal fibroblasts cultured from stromal tissue explants from the donor corneas ²⁶ were grown in Dulbecco’s modified Eagle’s medium supplemented with Ham’s nutrient mixture F-12 (DMEM-F12; Sigma-Aldrich, St. Louis, MO) with 10% fetal bovine serum (Atlanta Biologicals Inc., Norcross, GA). Confluent cultures in passages 2 to 4 were subcultured in three sets of 60-mm dishes at a density of 5000 cells/cm². After 24 hours in culture, cells were grown in DMEM-F12 with additions as detailed in Table 1 . In serum, growth-inducing factors predominate over growth-suppressing factors, and so, to induce the myofibroblast phenotype, the serum component was decreased to potentiate the effect of TGF-β1. After 12 hours of Y-27632 (Tocris Cookson Inc., Ellisville, MO) or vehicle treatment, the cells in one dish from each experimental set were immunostained for α-SMA (100 cells were examined in three different fields) and the other dish(es) used for RNA extraction. No Y-27632–dependent apoptosis was apparent by microscopy. Total RNA was extracted from these cells (RNeasy; Qiagen, Valencia, CA). Yields were 11 to 14 μg RNA per confluent 60-mm culture dish, and 6 to 15 μg total RNA was used for sample preparation. Across the nine sample pairs, total RNA extracted from Y-27632–treated cells was 99% ± 14% (mean ± SD) of that extracted from matched controls, indicating no cell loss. 

cRNA Preparation and Array Hybridization

cRNA preparation and array hybridization were performed according to the microarray technical literature (Affymetrix, Santa Clara, CA). Briefly, the mRNA component (≥0.5 μg) of extracted total RNA was reverse-transcribed (SuperScript II; Invitrogen-Life Technologies, Gaithersburg, MD) in the presence of a T7-(dT)₂₄ primer (Genset Corp., La Jolla CA). The cDNA product was purified using phenol-chloroform extraction followed by ethanol precipitation of the cDNA from the aqueous phase. The cDNA was transcribed in vitro in the presence of biotin-labeled ribonucleotides (BioArray; Enzo, Farmingdale, NY). The biotinylated RNA product (≥35 μg) was separated from cDNA and other transcription components (RNeasy; Qiagen). 

A portion (20 μg) of the biotinylated RNA was fragmented in a heat- and ion-dependent manner, and the fragments were hybridized overnight to one of the microarray (GeneChip; Affymetrix) formats. The chip was washed and developed by sequential treatment with streptavidin-phycoerythrin, anti-streptavidin antibody (conjugated to biotin), and then streptavidin-phycoerythrin. The GeneChip was scanned (ChipScanner; Agilent, Palo Alto, CA), and the collected data were processed and analyzed (MicroArray Suite [MAS] ver. 5.0; Affymetrix). 

Microarray Samples

Human corneal stromal cells were derived from three separate donors. Each cell preparation was cultured under six different conditions (Table 1) without cross-donor mixing, yielding 18 unique samples. Two preparations of stromal cells (i.e., 12 samples) were analyzed with HU133A chips, and the remaining preparation (6 samples) was analyzed with HU95Av2 chips. Expression levels for each chip were normalized to the mean level (216) for all 18 chips. The HU95Av2 and HU133A formats contain 12,625 panels and 22,283 panels, respectively, with 9,843 panels showing good agreement between formats (Affymetrix technical literature). Panels measure both perfect match (signal) and mismatch (background) components. The MAS v5.0 software designates a transcript as present only if the perfect-match component is significantly (P < 0.05, single-tailed Wilcoxon signed-rank test) greater than the mismatch; otherwise, it is designated absent. Similarly, the software evaluates the significance of apparent changes between Y-27632–treated samples and matched controls, designating an increase or decrease only if P < 0.006. Our selection rules for three comparisons were as follows, with the less stringent requirements for six or nine comparisons in square brackets. For decreases, every control transcript must be present, and every Y-27632–treated sample must decrease relative to its control; every pair-wise ratio (i.e., Y-27632/uninhibited) must be ≤0.67, [≤0.8] and the mean of the pair-wise ratios (Y-27632/uninhibited) for three experiments must be ≤0.5 [≤0.67]. For increases, every Y-27632-treated transcript must be present and increase relative to its control; every pair-wise ratio must be ≥1.5 [≥1.25], and the mean of the pair-wise ratios (Y-27632/uninhibited) for three experiments must be ≥2 [≥1.5]. 

Results and Discussion

There was good correlation between Y-27632–treated cells and matched controls, with R ² = 0.977 ± 0.004 (mean ± SD, n = 3) for U95A chips and R ² = 0.976 ± 0.004 (mean ± SD, n = 6) for U133A chips. These data highlight the relatively small effect of Y-27632 and the excellent chip-to-chip reproducibility when similar samples are analyzed. They are comparable with the R ² (0.95) for chip-to-chip reproducibility previously reported, ²⁵ using corneal fibroblasts derived from a single tissue source. Comparisons between samples from different donor corneas which were treated identically in vitro gave regression coefficients of R ² = 0.939 ± 0.013 (mean ± SD, n = 6: U133A chips). Analysis of corneal epithelial cells on U95A chips (Harvey SAK, SundarRaj N, unpublished data; 2002) gave R ² = 0.970 ± 0.022 (mean ± SD, n = 3) for the Y-27632 effect and 0.900 ± 0.044 (mean ± SD, n = 6) for donor-to-donor variation. Taken together, these data show that donor-to-donor variation is greater than chip-to-chip variation for both chip formats and led us to select candidates on the basis of consistent multiples of change (x-fold) in pair-wise comparisons rather than requiring consistent absolute changes in expression. 

Phenotypic Shift versus Y-27632–Driven Changes in Corneal Stromal Cells

In our dataset, the largest expression change was between activated fibroblasts (2%–5% cells immunostained for α-SMA) and myofibroblasts (>90% cells immunostained for α-SMA). This phenotypic shift was brought about by omission of exogenous bFGF plus heparin, a 10-fold decrease in serum-derived factors, and the addition of TGF-β1 (Table 1) . The shift increased over 389 unique gene transcripts and decreased over 325, altering a relatively large number of transcripts (714/22,283 = 3.2% of those surveyed) over a wide range (205- to 0.007-fold). Preliminary analysis confirmed that this phenotypic shift includes several previously published TGF-β–induced changes (Harvey SA, et al. IOVS 2003;44:ARVO E-Abstract 834). For example, microarray analysis of the effects of TGF-β in human fetal lung fibroblasts yields 146 altered transcripts of the more than 6000 surveyed. ²⁷ Of these 146, 35 had exact matches in our data where chance would have predicted (146 × 714/22,283) or 4.7 matches. There were a further 11 inexact matches (i.e., different isoforms or family members). 

In contrast to the 714 changes associated with phenotypic shift, only 66 unique transcripts (66/22,283 = 0.3%) were altered by Y-27632 treatment, and these changes were of smaller magnitude (6.6- to 0.31-fold). Transcripts decreased by Y-27632 were designated ROCK supported, and those increased by Y-27632 were designated ROCK suppressed. In Tables 2 3 4 and 5 , transcripts that are also altered by the shift from fibroblast to myofibroblast are identified by arrows that show whether the transcript is increased (↑) or decreased (↓) in myofibroblasts. 

Y-27632–Driven Changes in Corneal Stromal Cells

Changes Common to All Stromal Groups.

We identified genes commonly controlled by ROCK signaling in all threestromal groups (Table 2) . For panels represented on both chip formats, combining all groups provided nine pair-wise comparisons and six pair-wise comparisons for panels present only on U133A chips. For these larger groups we used the less stringent ratio requirements, as described earlier. Moreover, we included candidates that had just one outlier (multiples of increase out of range, and/or transcript not present, and/or transcript level not changed). In Table 2 , ROCK-supported transcripts are broadly associated with cell cycle control and cytokinesis. We have hypothesized (see Fig. 8 in Ref. ²⁸ ) that ROCK signaling upregulates the entry of cells into the S-phase of the cell cycle. ROCK supports both HMGB2, a nonhistone protein participating in the G₁-to-S transition ²⁹ and ribonucleotide reductase subunit M2 (RRM2), which is the rate-limiting step in DNA synthesis in the S-phase and plays a crucial role in cell proliferation. ³⁰ Cyclin B both catalyzes entry into mitosis and promotes the S-phase, ³¹ whereas cell division cycle 2/cyclin dependent kinase 1 (CDC2/CDK1) and cyclin kinase subunit, isoforms 1 and 2 (CKS) control entry into the M-phase through their interactions with cyclin B1. ³²  

ROCK-supported control of mitosis occurs by PRC1, encoding a microtubule-associated protein that maintains the spindle midzone ³³ and is a phosphorylation substrate of cyclin B/CDC2, ³⁴ abnormal spindle-like, microcephaly associated/MCPH5/FLJ10517 (ASPM), which is a major determinant of the increased cerebral cortical size in humans, ³⁵ and rabkinesin6 (KIF20A/RAB6 interacting, kinesin-like protein/RB6K), which is essential for cytokinesis. ^{36 37} RACGAP1 (MgcRacGAP for male germ cell RACGAP) identified ³⁸ as a GAP for Rac/Cdc42 shows progressive localization to the nucleus, mitotic spindle, and midbody in interphase, metaphase, and cytokinesis, respectively. RACGAP1is essential for cytokinesis, ³⁹ and when phosphorylated by aurora B kinase, it is functionally converted to a RhoGAP. ⁴⁰ Finally, the actin-associated protein encoded by ENC1 plays a regulatory role in the cytoskeletal reorganization and cell shape change that occur on differentiation from fibroblastic preadipocytes to spherical adipocytes. ⁴¹  

Although 10 μM Y-27632 causes 70% inhibition of thrombin-stimulated DNA synthesis in rat aortic smooth muscle cells, ⁴² Y-27632 has no apparent effect on cell proliferation in FCS-stimulated human umbilical vein endothelial cells (HUVECs) ⁴³ or (until >200 μM) in renal fibroblasts. ⁴⁴ In Swiss 3T3 cells, inhibition of cytokinesis is detectable only at 30 μM Y-27632 or more. Twelve to 14 hours of treatment with 100 μM inhibitor delays G₁-to-S progression considerably, but treated cells “catch up” by 32 hours. ⁴⁵ We saw a similar delay/catch-up pattern using 25 μM Y-27632, which together with the low inhibitor concentration and short exposure time probably explains why ROCK support of S-phase/mitosis transcripts appears weak. 

Of six ROCK-suppressed transcripts, three are associated with cell signaling. Stromal cell derived factor (SDF-1/intercrine alpha/CXCL12), is an important chemotactic and angiogenic factor. Transcripts for both SDF-1 and its cognate receptor CXCR4 (NPYR3) have been detected in human corneal fibroblasts, ⁴⁶ and intact corneas have functional CXCR4 receptors as indicated by high-affinity SDF-1 binding. Cytosolic phospholipase A₂ is a classic signal transduction enzyme that releases arachidonic acid from cell phospholipids and is activated by MAP kinase and by other ligand-receptor-invoked signaling pathways. Finally, type 2 phosphatidic acid phosphatases (PPAP2A) probably act as ectoenzymes to dephosphorylate signaling phospholipids such as lyso-phosphatidate, ceramide-1-phosphate, sphingosine-1-phosphate, and diacylglycerol pyrophosphate. ⁴⁷ PPAP2A expression is decreased in tumor cells ⁴⁸ suggesting that normally, it acts to decrease proliferation. If this is the case, ROCK suppression of the transcript will increase proliferation. 

The cornea contains crystallin components ⁴⁹ (but see Ref. ⁵⁰ ), and both aldo-keto reductase (AKR) and aldehyde dehydrogenase (ALDH) families are candidates for this role. In the present study, the AKR1 family members B1, C1, C2, C3, and B10 all were present at between 5 and 20 times the global mean expression level, whereas the “conventional” lens crystallins (ALDH3 members A1, B1, and B2) were absent. ROCK suppression of AKR1C2 suggests that Rho/ROCK signaling may have a role in regulating these presumptive AKR crystallins. 

Active carbonic anhydrase II (CAII) binds to and doubles the activities of both the plasma membrane chloride/bicarbonate exchanger (AE1), ⁵¹ and the Na⁺/H⁺ exchanger NHE1, ⁵² suggesting that ROCK markedly suppresses the cells’ capacity to transport these ions. Finally, the hypothetical protein FLJ23191 fulfills stringent requirements for each phenotype (Tables 3 4 5) , making it the most important ROCK-associated transcript of unknown function. 

Baseline Fibroblasts: Stromal Cells with no Added Growth Factors.

Table 3 shows that numerically, ROCK suppresses more transcripts than it supports. AKR1C2, PPAP2A, CAII, and FLJ23191 are all affected as previously discussed (Table 2) , whereas cell signaling is substantially modulated, with ROCK-suppressed transcripts encoding six extracellular mediators, a receptor diptych, and two protein kinase/phosphatase enzymes. Fibroblast growth factor 7/keratinocyte growth factor (FGF7/KGF) is a known paracrine mediator in the cornea, ⁵³ expressed in fibroblasts but not in corneal epithelial cells, ⁵⁴ and acting on corneal epithelial cells but not keratocytes through receptor FGFR2/KGFR. ⁵⁴ Consistent with this, KGFR is absent in stromal cells. Moreover, in corneal epithelial cells KGF is absent and KGFR is present (Harvey SAK, Anderson SC, SundarRaj N, unpublished data, 2003). TGF-β2 is one of four transcripts for which the direction of Y-27632–driven change varies with culture conditions. In baseline fibroblasts expression is increased, whereas in myofibroblasts (Table 5) expression is decreased. Latent TGF-β–binding protein 3 (LTBP3) is the second of these four transcripts. It shows moderate downregulation by Y-27632 (0.62-fold: not shown in Table 3 ) in baseline fibroblasts but is upregulated in myofibroblasts (Table 5) . LTBP3 is an extracellular matrix (ECM) protein that binds the small latent complex (SLC) of TGF-β, and so reciprocal expression of LTBP3 and TGF-β2 (the predominant endogenous TGF-β isoform in the cornea) is consistent with homeostatic control of active TGF-β2. 

Both parathyroid hormone-like hormone (PTHLH/PTHrP: parathyroid hormone related peptide) and stanniocalcin 1 (STC1) control calcium and phosphate metabolism. PTHLH, like its prototype parathyroid hormone, stimulates bone resorption, renal tubular calcium reabsorption, 1,25-dihydroxyvitamin D₃ (calcitriol) synthesis, and urinary excretion of phosphate. STC1 reversibly inhibits calcium L-channels ⁵⁵ and is upregulated in response to ischemia in the brain ⁵⁶ and in the failing heart. ⁵⁵ Taken together, these data suggest that STC can be upregulated in response to various types of stress and probably protects against intracellular Ca²⁺ overload. SCDGF-B/PDGF-D is the first known ligand specific for the β isoform of the PDGF receptor. ⁵⁷ It potently transforms NIH/3T3 cells, eliciting stress fiber reorganization and increased proliferation rate. ⁵⁸ CXCL5 (SCYB5; small inducible cytokine subfamily B/ENA-78; epithelial-derived neutrophil-activating peptide 78) is a proinflammatory chemokine upregulated in cultured corneal stromal fibroblasts in response to IL-1α or TNFα. ⁵⁹  

Transcripts are increased for endothelin (ET) receptor subtypes A and B (ETRA and ETRB). ET is the most potent vasoconstrictor currently known, and there is a rapidly growing body of literature implicating ROCK as a component in ET-dependent signaling. For example, Y-27632 inhibits ET-1-induced hypertrophy in neonatal rat cardiac myocytes. ⁶⁰  

ROCK-suppressed signaling kinase/phosphatase enzymes are DYRK1A and PPP2R1B. The DYRK kinase family catalyze tyrosine-directed autophosphorylation but phosphorylate exogenous substrates at serine-threonine residues. ⁶¹ Dynamin, known to be essential in cytokinesis ⁶² is a substrate of DYRK1A. ^{63 64} The protein phosphatase 2 (formerly protein phosphatase 2A) system comprises catalytic α and β isoforms (PPP2CA and PPP2CB), either of which can associate with the regulatory α or β isoform (PPP2R1A, PPP2R1B). The resultant core dimer can then interact with at least five distinct types of secondary regulatory subunits (for review, see Ref. ⁶⁵ ). Of the four catalytic and primary regulatory subunits PPP2R1B is expressed at the lowest level in control cells, so that ROCK suppresses a limiting component of this complex signaling system. 

ROCK-supported cell cycle control is represented by a transcript (HMGA1) encoding a high-mobility group nonhistone protein (HMG-I) that acts as an architectural transcription factor at mammalian promoters. HMG-I is phosphorylated by cdc2/Cdk1 in vivo, ⁶⁶ and this may contribute to the chromatin condensation that occurs at transition from G₂ to mitosis (see HMGB2 in Table 2 ). ROCK-supported actin interactions are represented by smoothelin, which has a truncated actin-binding domain and associates with stress fibers. It was first described as a cytoskeletal protein specific for fully differentiated smooth muscle cells. ⁶⁷  

Increased expression of intercellular adhesion molecule (ICAM-1/CD54/human rhinovirus receptor) in corneal disease ⁶⁸ increases leukocyte recruitment. As a consequence of their ICAM-1–mediated association, ROCK signaling occurs both in leukocytes ⁶⁹ and in endothelial cells. ⁷⁰ Moreover, ROCK-supported (Y-27632 inhibitable) ICAM-1 expression has been reported in cardiac allografts. ⁷¹  

Plasminogen activator inhibitor type (PAI)-1 inhibits the sequential activation of plasminogen and latent collagenase. ROCK support of PAI expression indicates that ROCK signaling inhibits remodeling of the ECM initiated by plasmin and collagenase. This is consistent with ROCK suppression of some ECM components in myofibroblasts (Table 5) . 

Activated Fibroblasts: bFGF+Heparin-Treated Corneal Stromal Cells.

In bFGF-activated fibroblasts (Table 4) , the ROCK-supported transcripts ASPM (mitosis) and ENC1 (actin-associated) found previously (Table 2) are augmented by the mitosis-associated centromere protein F/mitosin (CENPF). As in baseline fibroblasts, ROCK suppresses CA2 and FLJ23191, and the cell-signaling components PTHLH and ETRA/ETRB (Table 3) . Additional signaling components suppressed by ROCK include two members of the TGF-β superfamily, activin A and FLJ21195. Activin A and its cognate receptors are expressed in corneal fibroblasts, and exogenous activin A increases α2-SMA expression. ⁷² ROCK suppression of activin A probably decreases these autocrine effects and favors the fibroblast over the myofibroblast phenotype. Moreover, hypothetical protein FLJ21195 is similar (54% identity plus 14% positives) to gremlin (CKTSF1B1), an endogenous antagonist of bone morphogenetic protein (BMP)-4 (a TGF-β superfamily member) known to be important in renal fibrosis. ⁷³ The last cell signaling target is neuronal pentraxin I (NPTX1), a secreted glycoprotein that plays a role in remodeling of the nervous system. ⁷⁴ In Xenopus, the NPTX1 homologue is upregulated in regenerating epidermis. ⁷⁵ The transcriptional activator MyoD, which is essential to specification of muscle cell lineage, upregulates both NPTX1 and Id3 (inhibitor of DNA binding, a transcription corepressor) in undifferentiated myoblasts. ⁷⁶ Notably, in the present study Id4 was upregulated, suggesting a consistent relationship between NPTX1 and the Id family. These two components may contribute to ROCK suppression of myofibroblast conversion. 

Two additional ROCK-suppressed transcriptional modulators are Nurr1 and NF-HEV. The transcriptional activator Nurr1 (nuclear receptor subfamily 4, group A, member 2: NR4A2/NOT/RNR1/HZF-3/TINUR) is a member of a superfamily that generally bind steroids or retinoids, but Nurr1 has no ligand binding site ⁷⁷ and its activity is probably modulated by other signal-transduction pathways such as MAP kinase. ⁷⁸ In bone cells, PTH induces Nurr1 ⁷⁹ suggesting that Nurr1 upregulation may be an autocrine response to elevated PTHLH. NF-HEV is a nuclear factor highly expressed in the specialized endothelium of high endothelial venules. ⁸⁰ The canine homologue DVS27 is expressed in (IL-1–stimulated) cultured smooth muscle cells, ⁸¹ suggesting that ROCK suppression of this factor plays a role in suppressing myofibroblast conversion. 

Tumor necrosis factor-α–induced protein 6 (TSG-6, TNFAIP6), is a member of the family of hyaluronate binding proteins, closely related to the adhesion receptor CD44 and often associated with extracellular matrix remodeling (for review, see Ref. ⁸² ). 

Myofibroblasts: TGF-β1/1% Serum-Treated Corneal Stromal Cells.

Upregulation of α2-SMA is diagnostic of the shift to myofibroblast phenotype, and in the present study this shift caused an 8.3-fold increase in α-SMA transcript levels, which was attenuated by Y-27632. The induction of α-SMA during renal fibrosis in vivo ⁴⁴ is decreased by Y-27632, which abrogates the TGF-β1–induced formation of stress fibers in PC-3U human prostate carcinoma cells in vitro. ⁸³ In our hands, another SMA isoform (γ2) was significantly downregulated (0.29-fold) by TGF-β1 treatment, but remained susceptible to further downregulation by Y-27632. 

Because two components of the ROCK-supported triad cyclin B1+CDC2+CKS2 (see Table 2 ) appear in Table 5 , we reviewed cyclin B1 and found that Y-27632 reduced its expression to 0.51 ± 0.13 of untreated levels. Thus, ROCK again supports the coordinated expression of all three components. ROCK facilitation of the M-phase includes support of DNA topoisomerase IIα, which catalyzes the adenosine triphosphate (ATP)–dependent passage of one double-stranded DNA molecule through a transient break in another and is critically important for chromosome condensation and segregation. ROCK modulation of cytokinesis may also occur by support of GBP1, a member of the dynamin ⁶² superfamily, which is largely responsible for the anti-proliferative response to inflammatory cytokines. ⁸⁴ Dual specificity phosphatase 6/MAP kinase phosphatase 3 (DUSP6/MPK3) contributes to G₁ cell cycle arrest and apoptosis, ⁸⁵ therefore DUSP6 suppression by ROCK facilitates normal proliferation. 

In myofibroblasts, ROCK supports TGF-β2 expression and suppresses LTBP3 expression, changes opposite to those seen in baseline fibroblasts (Table 3) . Because TGF-β2 is the principal endogenous isoform of TGF-β, ROCK signaling supports autocrine maintenance of the myofibroblast phenotype. Moreover, ROCK suppresses the expression fibroblast growth factor receptor 1 (FGFR1, the receptor for bFGF/FGF2) decreasing sensitivity to bFGF and again maintaining the myofibroblast phenotype. Corneal epithelial and endothelial cells show a mitogenic response to hepatocyte growth factor (HGF), whereas stromal cells do not; however, transcripts for both HGF and its receptor (MET) occur in all three cell types. ⁵³ In our hands, ROCK supported MET and thereby a presumptive nonmitogenic stromal response. 

As in baseline fibroblasts (Table 3) , ROCK suppresses both a member of the phosphatidic acid phosphatase type 2A superfamily, lipid phosphate phosphatase-related protein type 2a/FLJ13055 (LPPR2), and a candidate crystallin (AKR1B10). Myofibroblasts actively elaborate a wide range of ECM components, and ROCK suppresses a subset of these: biglycan, perlecan (heparan sulfate proteoglycan 2), agrin, adlican, collagen VII α1, and E2IG4. ⁸⁶ Finally, NAD(P)H-quinone oxidoreductase 1/cytochrome b5 reductase (NQO1) is widely distributed in the human eye, with strong immunostaining in keratocytes. ⁸⁷ NQO1 is bifunctional in response to oxidative stress. It has a cytoprotective, antioxidant role in vivo ^{88 89} but also stabilizes the tumor suppressor p53, ⁹⁰ suggesting that it may trigger apoptosis in extreme conditions. ROCK suppression of NQO1 in myofibroblasts may decrease their vulnerability to apoptosis, relative to fibroblasts. 

In summary, cytokines and growth factors are involved in corneal wound healing. Because bFGF and TGFβ are expressed by both corneal epithelial cells and fibroblasts, their effects on corneal stromal cells during wound healing are both autocrine and paracrine (for review see Refs. ⁹¹ , ⁹² ). Quiescent stromal cells are activated during wound healing to become fibroblasts and myofibroblasts, the latter expressing α-SMA. Both these phenotypes have robust networks of actin filaments. Although one of the main functions of Rho/ROCK signaling is the assembly of stress fibers, other cellular processes are also regulated. ⁹³ The present study clearly indicates that Rho/ROCK signaling has both global effects on corneal stromal cell gene expression (e.g., cell cycle control and mitosis) and phenotype-specific effects. Given the number of panels that are separately altered by Y-27632 and the phenotypic shift from fibroblast to myofibroblast (66 and 714, respectively), random interaction between these conditions should yield only approximately two panels (66 × 714/22,283), which respond to both. Twenty-seven transcripts (see Tables 2 3 4 5 ) showed interaction between these two conditions, ∼13-fold higher than predicted. This unexpectedly high number underlines the phenotype-specific importance of ROCK. For example, because TGF-β2 is the predominant endogenous TGF-β isoform in the cornea, ⁹⁴ ROCK suppression of TGF-β2 in baseline fibroblasts attenuates autacoid TGF-β–mediated conversion to myofibroblasts. However, if paracrine-mediated conversion occurs, ROCK accelerates it by supporting myofibroblast output of autacoid TGF-β2. During wound healing, endogenous cytokines and growth factors modulate Rho/ROCK signaling. The present study shows that ROCK reciprocates by orchestrating both the cellular synthesis of mediators (TGF-β2, PDGF-D, KGF, PTHLH, SDF-1 and stanniocalcin) and the cellular responses (through changes in their cognate receptors) to FGF, HGF, and ET. 

 Supported by National Eye Institute Grants EY03263 (NS), EY05945-15S1 (Robert L. Hendricks), and CORE Grant EY08098 to the Department of Ophthalmology; the Eye and Ear Foundation of Pittsburgh, and Research to Prevent Blindness.

 Submitted for publication November 7, 2003; revised January 28 and March 11, 2004; accepted March 23, 2004.

 Disclosure: S.A.K. Harvey, None; S.C. Anderson, None; N. SundarRaj, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Table 1.

 

 

Stromal Cell Culture and Treatment

Table 1.

 

Stromal Cell Culture and Treatment

Designation	Serum (%)	Additional Growth Factors*	Inhibitor Y-27632 (μM)
Baseline fibroblast	10	None	0
Baseline fibroblast	10	None	25
Activated fibroblast	10	bFGF (20 ng/mL)+heparin (5 μg/mL)	0
Activated fibroblast	10	bFGF (20 ng/mL)+heparin (5 μg/mL)	25
Myofibroblast	1	rhTGF-β1 (10 ng/mL)	0
Myofibroblast	1	rhTGF-β1 (10 ng/mL)	25

 Cells were cultured in serum and growth factors from 24 to 72 hours. The inhibitor Y-27632 or vehicle was added at 60 hours, i.e., for the last 12 hours of culture.

*  Recombinant FGF (Invitrogen-Gibco), heparin (Sigma-Aldrich), recombinant human TGF-β1 (Roche Diagnostics, Indianapolis, IN).

Table 2.

 

 

Global Y-27632 Effects in Stromal Cells

Table 2.

 

Global Y-27632 Effects in Stromal Cells

Affymetrix	GenBank	LocusLink	HGNC	Mean Change (x-fold)	Mean Control Level	n	Encoded Protein
214710_s_at	BE407516	891	CCNB1	0.57	302	8	Cyclin B1 ↓
203213_at	AL524035	983	CDC2	0.41	164	5	Cyclin-dependent kinase 1 ↓
204170_s_at	NM_001827.1	1164	CKS2	0.57	410	9	CDC28 protein kinase regulatory subunit 2 ↓
219918_s_at	NM_018123.1	259266	ASPM	0.52	193	6	Asp (abnormal spindle)-like, microcephaly-associated ↓
201340_s_at	AF010314.1	8507	ENC1	0.56	160	(8)	ENC1 ectodermal-neural cortex (with BTB-like domain) ↑
208808_s_at	BC000903.1	3148	HMGB2	0.62	319	(9)	High-mobility group box 2
209773_s_at	BC001886.1	6241	RRM2	0.56	579	(9)	Ribonucleotide reductase M2 polypeptide
218009_s_at	NM_003981.1	9055	PRC1	0.57	201	5	Protein regulator of cytokinesis 1
218755_at	NM_005733.1	10112	KIF20A	0.52	169	5	Rabkinesin6 (kinesin family member 20A)
222077_s_at	AU153848	29127	RACGAP1	0.63	184	5	GTPase-activating protein MgcRacGAP
209301_at	M36532.1	760	CA2	2.65	56	9	Carbonic anhydrase II ↓
210946_at	AF014403.1	8611	PPAP2A	1.86	338	8	PPAP2A: phosphatidic acid phosphatase type 2A ↑
209147_s_at	AB000888.1	8611	PPAP2A	1.79	390	(9)	PPAP2A: phosphatidic acid phosphatase type 2A (↑)
209699_x_at	U05598.1	1646	AKR1C2	1.80	1700	5	Aldo-keto reductase family 1, member C2
209687_at	U19495.1	6387	CXCL12	2.10	650	8	SDF: stromal cell-derived factor 1
210145_at	M68874.1	5321	PLA2G4A	1.68	532	8	Cytosolic phospholipase A2, calcium-dependent ↓
219747_at	NM_024574.1	79625	FLJ23191	2.40	228	6	Hypothetical protein FLJ23191

 For all stromal cells, n = 9 are data derived from the 9843 panels which Affymetrix designates as being in good agreement between U133A and U95Av2 chips; n = 8 are data for which a single outlier was omitted. For panels that are not in good agreement between the chip formats, the value 9 or 8 is in parentheses. Panels present only on the U133A chip have n = 6, or n = 5 if a single outlier was omitted. Arrows show direction of change for the fibroblast → myofibroblast transition. HGNC, Human Gene Nomenclature Committee.

Table 3.

 

 

Y-27632 Effects in Baseline Fibroblasts: Stromal Cells Grown in 10% Serum

Table 3.

 

Y-27632 Effects in Baseline Fibroblasts: Stromal Cells Grown in 10% Serum

Affymetrix	GenBank	LocusLink	HGNC	Mean Change (x-fold)	Mean Control Level	n	Encoded Protein
202637_s_at	AI608725	3383	ICAM1	0.44	568	(3)	Intercellular adhesion molecule 1 (CD54) ↓
209427_at	AF064238.3	6525	SMTN	0.45	279	(3)	Smoothelin
210457_x_at	AF176039.1	3159	HMGA1	0.46	173	(3)	High mobility group I (Y) protein: AT-hook 1 ↓
202628_s_at	NM_000602.1	5054	SERPINE1	0.48	3018	3	Plasminogen activator inhibitor type 1
205782_at	NM_002009.1	2252	FGF7	2.10	99	3	Fibroblast growth factor 7
209908_s_at	BF061658	7042	TGFB2	2.58	74	(3)	Transforming growth factor, beta 2
211756_at	BC005961.1	5744	PTHLH	5.19	36	3	Parathyroid hormone-like hormone
206300_s_at	NM_002820.1	5744	PTHLH	3.18	36	(3)	Parathyroid hormone-like hormone
204595_s_at	AI300520	6781	STC1	2.46	771	(3)	Stanniocalcin 1 ↓
204596_s_at	U46768.1	6781	STC1	2.36	343	(3)	Stanniocalcin 1 ↓
219304_s_at	NM_025208.1	80310	SCDGF-B	2.10	460	2	Spinal cord-derived growth factor-B (PDGF-D)
215101_s_at	BG166705	6374	CXCL5	6.62	63	2	Epithelial-derived neutrophil-activating peptide 78 ↓
204273_at	NM_000115.1	1910	EDNRB	2.57	53	3	Endothelin receptor type B, transcript variant 1
204271_s_at	M74921.1	1910	EDNRB	2.13	169	(3)	Endothelin receptor type B, transcript variant 1
204463_s_at	AU118882	1909	EDNRA	2.28	125	(3)	Endothelin receptor type A ↓
211079_s_at	Z25423.1	1859	DYRK1A	2.48	23	2	Dual-specificity tyrosine-(Y)-phosphorylation regulated kinase 1A
222351_at	AW009884	5519	PPP2R1B	2.59	60	(3)	Protein phosphatase 2 (formerly 2A), regulatory subunit A (PR 65), beta isoform (↓)
209147_s_at	AB000888.1	8611	PPAP2A	2.16	506	(3)	PPAP2A: phosphatidic acid phosphatase type 2A (↑)
209301_at	M36532.1	760	CA2	2.10	40	3	Carbonic anhydrase II ↓
209699_x_at	U05598.1	1646	AKR1C2	2.09	2310	2	Aldo-keto reductase family 1, member C2
219747_at	NM_024574.1	79625	FLJ23191	2.67	315	2	Hypothetical protein FLJ23191

 The format is the same as is described in Table 2 , except no outliers are permitted. Thus, n = 3 for panels in good agreement between U133A and U95Av2 chips (3 appears in parentheses for panels that are not in good agreement), and n = 2 for panels present only on the U133A chip.

Table 4.

 

 

Y-27632 Effects in Activated Fibroblasts: Stromal Cells Grown in 10% Serum+bFGF+Heparin

Table 4.

 

Y-27632 Effects in Activated Fibroblasts: Stromal Cells Grown in 10% Serum+bFGF+Heparin

Affymetrix	GenBank	LocusLink	HGNC	Mean Change (x-fold)	Mean Control Level	n	Encoded Protein
201340_s_at	AF010314.1	8507	ENC1	0.41	116	(3)	Ectodermal-neural cortex (with BTB-like domain) ↑
219918_s_at	NM_018123.1	259266	ASPM	0.44	300	2	Asp (abnormal spindle)-like protein ↓
207828_s_at	NM_005196.1	1063	CENPF	0.45	238	(3)	Centromere protein F (350400kD, mitosin)
210355_at	J03580.1	5744	PTHLH	5.80	29	3	Parathyroid hormone-like hormone
204273_at	NM_000115.1	1910	EDNRB	2.79	77	3	Endothelin receptor type B, transcript variant 1
204271_s_at	M74921.1	1910	EDNRB	2.13	247	(3)	Endothelin receptor type B, transcript variant 1
204463_s_at	AU118882	1909	EDNRA	2.10	160	(3)	Endothelin receptor type A ↓
209301_at	M36532.1	760	CA2	2.12	113	3	Carbonic anhydrase II ↓
204926_at	NM_002192.1	3624	INHBA	2.19	338	2	Activin A
220794_at	NM_022469.1	64388	FLJ21195	2.40	280	2	Hypothetical protein FLJ21195 (Gremlin-like) ↓
204684_at	NM_002522.1	4884	NPTX1	2.24	535	3	Neuronal pentraxin I precursor ↓
209291_at	AW157094	3400	ID4	2.24	44	(3)	Inhibitor of DNA binding 4
206025_s_at	NM_007115.1	7130	TNFAIP6	2.12	464	2	TSG-6
216248_s_at	S77154.1	4929	NR4A2	4.22	140	(3)	Nuclear receptor subfamily 4, group A, member 2
209821_at	AB024518.1	90865	C9orf26	2.78	99	3	NF-HEV (DVS27-related protein) ↓
219747_at	NM_024574.1	79625	FLJ23191	2.52	225	2	Hypothetical protein FLJ23191

 The format is the same as is described in Table 2 , except no outliers are permitted. Thus, n = 3 for panels in good agreement between U133A and U95Av2 chips (3 appears in parentheses for panels that are not in good agreement), and n = 2 for panels present only on the U133A chip.

Table 5.

 

 

Y-27632 Effects in Myofibroblasts: Stromal Cells Grown in 1% Serum and TGF-β1

Table 5.

 

Y-27632 Effects in Myofibroblasts: Stromal Cells Grown in 1% Serum and TGF-β1

Affymetrix	GenBank	LocusLink	HGNC	Mean Change (x-fold)	Mean Control Level	n	Encoded Protein
200974_at	NM_001613.1	59	ACTA2	0.49	3795	3	Actin, alpha 2, smooth muscle, aorta ↑
202274_at	NM_001615.2	72	ACTG2	0.49	637	3	Actin, gamma 2, smooth muscle, enteric ↓
203213_at	AL524035	983	CDC2	0.31	63	2	Cell division cycle 2, G1 to S and G2 to M ↓
204170_s_at	NM_001827.1	1164	CKS2	0.45	284	3	CDC28 protein kinase regulatory subunit 2 ↓
201291_s_at	AU159942	7153	TOP2A	0.44	322	(3)	Topoisomerase (DNA) II alpha 170kDa (↓)
202270_at	NM_002053.1	2633	GBP1	0.37	172	3	Guanylate binding protein 1, interferon-inducible, 67kDa
202269_x_at	BC002666.1	2633	GBP1	0.39	312	(3)	Guanylate binding protein 1, interferon-inducible, 67kDa
201467_s_at	AI039874	1728	NQO1	0.46	561	2	NAD(P)H dehydrogenase, quinone 1 (↓)
209460_at	AF237813.1	57416	ABAT (NPD009)	0.49	116	2	4-Aminobutyrate aminotransferase (↑)
203510_at	BG170541	4233	MET	0.49	525	2	Met proto-oncogene (hepatocyte growth factor receptor)
209909_s_at	M19154.1	7042	TGFB2	0.49	89	3	Transforming growth factor, beta 2
219922_s_at	NM_021070.1	4054	LTBP3	2.28	149	2	Latent TGF beta binding protein 3
201262_s_at	NM_001711.1	633	BGN	4.16	1008	2	Biglycan ↑
201654_s_at	AI991033	3339	HSPG2	2.36	90	2	Heparan sulfate proteoglycan 2 (perlecan) (↑)
217419_x_at	AK021586.1	180	AGRN	2.03	187	2	Agrin
209596_at	AF245505.1	25878	DKFZp564I1922	2.03	236	3	Adlican ↑
217312_s_at	L23982	1294	COL7A1	2.03	302	2	Collagen, type VII, alpha 1 (↑)
218245_at	NM_015516.1	25987	E2IG4	3.04	187	2	Hypothetical protein, estradiol-induced (biglycan-like)
212942_s_at	AB033025.1	57214	KIAA1199	2.60	654	3	KIAA1199 protein ↓
206561_s_at	NM_020299.1	57016	AKR1B10	2.28	519	3	Aldo-keto reductase family 1, member B10
210973_s_at	M63889.1	2260	FGFR1	2.23	298	2	Fibroblast growth factor receptor 1
212090_at	AL571424	2907	GRINA	2.12	204	2	Glutamate-binding subunit of NMDA receptor
64899_at	AA209463	64748	LPPR2	2.35	375	2	Lipid phosphate phosphatase-related protein type 2
208892_s_at	BC003143.1	1848	DUSP6	2.02	105	3	Dual specificity phosphatase 6 ↓
217891_at	NM_022744.1	64755	FLJ13868	2.01	105	2	Hypothetical protein FLJ13868
219747_at	NM_024574.1	79625	FLJ23191	2.01	144	2	Hypothetical protein FLJ23191

 The format is the same as is described in Table 2 , except no outliers are permitted. Thus, n = 3 for panels in good agreement between U133A and U95Av2 chips (3 appears in parentheses for panels that are not in good agreement), and n = 2 for panels present only on the U133A chip.

 

The authors thank Jean-Paul Vergnes and Paul R. Kinchington for assistance in initiating the study.