Characterization of cellular senescence mechanisms in human corneal endothelial cells

The human cornea is a tri-laminar structure composed of several cell types with substantial mitotic potential. Age-related changes in the cornea are associated with declining visual acuity and the onset of overt age-related corneal diseases. Corneal transplantation is commonly used to restore vision in patients with damaged or diseased corneas. However, the supply of donor tissue is limited, and thus there is considerable interest in the development of tissue-engineered alternatives. A major obstacle to these approaches is the short replicative lifespan of primary human corneal endothelial cells (HCEC). Accordingly, a comprehensive investigation of the signalling pathways and mechanisms underpinning proliferative lifespan and senescence in HCEC was undertaken. The effects of exogenous human telomerase reverse transcriptase expression, p53 knockdown, disruption of the pRb pathway by over-expression of CDK4 and reduced oxygen concentration on the lifespan of primary HCEC were evaluated. We provide proof-of-principle that forced expression of telomerase, when combined with either p53 knockdown or CDK4 over-expression, is sufficient to produce immortalized HCEC lines. The resultant cell lines express an HCEC-specific transcriptional fingerprint, and retain expression of the corneal endothelial temperature-sensitive potassium channel, suggesting that significant dedifferentiation does not occur as a result of these modes of immortalization. Exploiting these insights into proliferative lifespan barriers in HCEC will underpin the development of novel strategies for cell-based therapies in the human cornea.


Introduction
Structurally, the cornea has three layers, each populated by a different cell type with highly distinctive functions and patterns of gene expression (Joyce, 2003). The outermost epithelium is a self-renewing multilayered epithelial sheet. Located beneath the epithelium, the stroma is a structured lattice of collagens and proteoglycans deposited and maintained by specialized keratocytes. Below the stroma, the endothelium is a single layer of cells on the innermost surface that forms the boundary between the cornea and the aqueous humour. Corneal endothelial cells have a number of essential functions including active transport of nutrients from the aqueous humour into the stroma and regulation of stromal hydration via Na + ⁄ K + ATPase pumps to maintain corneal transparency (Fischbarg & Maurice, 2004). In vivo human corneal endothelial cells (HCEC) reside within the highly growth-factor-depleted environment of the aqueous humour, and rarely enter mitosis under normal conditions (Klenkler & Sheardown, 2004). Thus, in many species, the repair of wounds to the corneal endothelium is principally achieved by cell enlargement and migration rather than cell division (Joyce, 2003). This quiescent state is thought to be maintained in vivo by a combination of contact inhibition and antiproliferative autocrine and paracrine TGFb signalling (Joyce, 2003).
Despite their proliferation being severely restricted in vivo, HCEC retain the ability to divide when removed from this environment, and primary cultures can be established from explants of donor corneal tissue (Engelmann et al., 1988;Kipling et al., 2009). However, after a relatively short proliferative lifespan [typically 20-30 population doublings (PD)], HCEC cultures enter senescence. Senescence is a process at the single cell level that acts to limit excessive cell division and halt early neoplastic progression (Adams, 2009). It is characterized by long-term cell cycle arrest accompanied by altered cellular morphology and physiology (Campisi & d'Adda di Fagagna, 2007;Adams, 2009). Senescence is regulated in a cell type-and stimulus-dependent manner by a complex signalling network involving the tumour suppressors, p53 and pRb (Fig. 1). Expression of exogenous human telomerase reverse transcriptase (hTERT) in several human cell types restores telomerase activity and prevents replicationdependent telomere erosion (Bodnar et al., 1998;Vaziri & Benchimol, 1998), thus preventing the telomere-dependent DNA damage signal that induces senescence in certain cell types (d'Adda di Fagagna et al., 2003). As a result, expression of hTERT leads to immortalization with retention of normal physiological function in many human cell types including fibroblasts, bronchial epithelial cells and corneal epithelial cells, and is therefore considered to be a promising tool for boosting cell numbers in bioengineering applications (Robertson et al., 2005;. However, several human cell types including myoblasts (Zhu et al., 2007), skin keratinocytes (Kiyono et al., 1998) and astrocytes (Evans et al., 2003) do not show extension of cellular lifespan following forced expression of hTERT. These data show that substantial cell type specificity exists in the senescence mechanisms and regulators used, and that the mechanisms underlying replicative senescence in any given cell type cannot be assumed but must be characterized on a cell type by cell-type basis .
In vivo studies provide some clues to the potential senescence mechanisms operating within HCEC. Donor age and anatomical location within the corneal endothelium influence the capacity of explanted HCEC to divide ex vivo (Zhu & Joyce, 2004;Enomoto et al., 2006;, suggesting the presence of senescent cells in human corneal endothelium in vivo. The expression of genes associated with senescence, such as the cyclin-dependent kinase inhibitors (CDKIs) p16 INK4a and p21 CIP1 , has been reported to be increased in the corneal endothelium, and primary HCEC cultures derived from them, of donors over the age of fifty years (Enomoto et al., 2006;Song et al., 2008). These increases in CDKIs are reflected in slower kinetics of pRb hyperphosphorylation in primary HCEC from older donors (Enomoto et al., 2006). In addition, SA-bgal staining, a marker of senescence, was absent in corneal endothelium of young donors but present in approximately 50% of peripheral cells and 80% of central cells in corneal endothelium of donors over fifty Song et al., 2008). However, these differences in replicative capacity and expression of senescence markers between younger and older donors, and peripheral vs. central HCEC, were not associated with differences in telomere length. This observation was interpreted as indicating that senescence may be induced in corneal endothelium in vivo by telomere-independent mechanisms (Konomi & Joyce, 2007).
Whilst it is possible that senescent cells may contribute to agerelated ocular pathologies (Faragher et al., 1997b), it is clear that there is a practical need to understand the proliferative lifespan barriers operating within HCEC in vitro to manufacture utile cell lines for both fundamental study and translational use. Accordingly, we have carried out a detailed dissection of the molecular mechanisms that control replicative senescence (Fig. 1) in cultured HCEC, and have established the routes by which these may most efficiently be bypassed so as to support the future development of functional, differentiated endothelial cell lines.

Results
Late passage HCEC show morphological features of senescence and a transcriptome enriched with p53 targets, but lack elevated CDKIs Strains of HCEC will proliferate for approximately 20 PD in culture ( Fig. S1). At the end of this proliferative lifespan, the cultures are overwhelmingly composed of cells that show characteristics typical of replicative senescence. These include an enlarged, flattened morphology and lack of proliferation (as measured by Ki67 staining) when compared with early passage HCEC (data not shown). A comparison of the transcriptomes of these 'senescent' HCEC with their proliferating and quiescent isogenic counterparts showed that a cluster of genes differentially upregulated in senescent HCEC is significantly enriched with p53 transcriptional targets (Table S1, Fig. S2), although p53 itself is not up-regulated at the protein or transcript level (Figs 2 and 3), consistent with previous studies of fibroblast senescence (Webley et al., 2000). An up-regulation of p21 CIP1 (CDKN1A) and p16 INK4A (CDKN2A) at the transcript level is detectable in senescent HCEC (Fig. 2), although marked changes at the protein level at senescence are absent (Fig. 3b, lanes 1 and 2 and data not shown).
Senescence in HCEC is not driven by telomere length or stress signalling but is modulated by extrinsic factors that impact upon oxidative stress Figure 4 shows the effects of stabilization of telomere length in primary HCEC by ectopic expression of hTERT using a pBABE series retroviral vector under standard culture conditions. No extension of replicative lifespan was observed either by mass culture or in individual clones. However, reducing the partial pressure of oxygen in the culture incubator to 3% and adding ascorbic acid 2-phosphate to the growth medium produced a significant increase (12 PD) in the capacity of HCEC to proliferate in vitro. Expression of hTERT under these conditions extends replicative lifespan to 50 PD but does not produce immortalization. Blockade of p38MAP kinase signalling in primary HCEC by treatment with the inhibitor SB203580 did not alter their lifespan, despite control experiments with anisomycin-stimulated HCEC that confirmed that this compound was able to block stress signalling (Fig. S3).
Telomerase expression in combination with the ablation of p53 function will immortalize HCEC Ectopic expression of the E6 oncoprotein of human papilloma virus 16 (HPV16) in HCEC using a pBABE series retroviral vector (Bond et al., 1999;Davis et al., 2003;Evans et al., 2003) reduces p53 protein expression as measured by western blotting (Fig. 3a, Fig. S4) and extends the replicative lifespan of primary HCEC by approximately 30 PD (Fig. 4b).
Similarly, infection of HCEC with a retroviral vector carrying a small hairpin RNA directed against p53 produced a significant extension of lifespan to 37 PD, at which point the culture ceased to expand (Fig. S4, Fig. 4d). Reinfection of E6-or p53 shRNA-expressing cells with a second retroviral vector expressing hTERT prevented this growth arrest and resulted in extensive proliferation (Fig. 4c,d). HCEC cell lines coexpressing E6 and hTERT have now proliferated at least 100 PD beyond the replicative limit of the progenitor primary HCEC and may be considered immortal.  Ectopic expression of CDK4 can also immortalize HCEC in the presence of telomerase Introduction of a wild-type CDK4 transgene into HCEC produced elevated levels of CDK4 protein (Fig. 3b, lanes 3 and 4) and in turn elevated p16 INK4a at both the protein (Fig. 3b) and transcript ( Fig. 2) levels, consistent with a known feedback circuit (Ruas et al., 2007), and elevated transcript abundance (Fig. S5a,b) of E2F targets (Blais & Dynlacht, 2004). Over-expression of CDK4 extended HCEC proliferative lifespan by 28 PD (Fig. 4e) and introduction of hTERT either at the same time as CDK4 (denoted 'Zante'; Fig. S5c) or else subsequent to CDK4 (denoted 'Cardiff'; Fig. 4f) gave further extension of lifespan to produce cell lines that have undergone at least 100 PD more than primary HCEC, and are thus judged to be immortal. pRb, p53 and p16 INK4A remained detectable at both transcript and protein level (Figs 2 and 3 and data not shown) in both cell lines, as was elevated CDK4 expression (Fig. 3b).

Immortalized HCEC retain the transcriptional signatures and characteristics typical of primary cells
Both the Zante and Cardiff cell lines have retained normal growth controls, as judged by retaining their ability to exit from the cell cycle in response to contact inhibition (data not shown). To determine whether any substantial alteration in corneal endothelial phenotype had occurred in the cell lines, the expression profiles of primary HCEC, the CDK4-hTERT Zante cell line and three other primary cell strains (Ek1.Br corneal keratocytes, together with HCA2 and WI38 primary human fibroblasts) were obtained using Affymetrix GeneChip microarrays. Clus- tering analysis of differentially expressed genes, defined using ANOVA, showed that the Zante cell line data clustered closely with primary HCEC and separately from the other three cell strains (Fig. S6a). These data were supplemented using a set of thirty-three genes that exhibit high expression levels in HCEC relative to each of the other cell types (Kipling et al., 2009), and thus constitute an HCEC-specific transcriptional fingerprint. Expression levels of these genes by two cultures of the Zante cell line at different cumulative PD levels was examined and compared with expression of the same transcripts in HCEC, EK1.Br, HCA2 and WI38 ( Fig. S6b and Table S2). HCEC-CDK4-TERT Zante cells at the earlier PD level of 52 retained elevated expression of at least 25 of the 33 genes differentially up-regulated in primary HCEC, thus maintaining a very similar profile to the parent primary HCEC strain. The transcriptome data also revealed that the cell lines retain mRNA-level expression of COL8A2, a major component of Descemet's membrane underlying the corneal endothelium and a marker for HCEC (Engler et al., 2009), and the corneal endothelial cotransporters SLC4A4, SLC4A2, SLC9A6 and NKCC1 (Bonanno, 2011). In addition, whole-cell patch-clamp measurements on the Zante cell line demonstrated the presence of a temperature-sensitive potassium channel (Fig. 5), together with dual expression of the intermediate filament proteins cytokeratin and vimentin (data not shown).

Discussion
Although cellular senescence has been widely studied in numerous cell types, before the current study little was known about pathways to senescence in HCEC. The data we have presented provide the most comprehensive description to date of the regulation of cellular lifespan in this human cell type.
It is clear from our study that HCEC exit from the cell cycle and enter a state of viable proliferative arrest consistent with replicative senescence. Senescence in this cell type is associated with up-regulation of a cluster of genes that is significantly enriched with p53 transcriptional targets. p53 message and protein levels themselves do not change, consistent with the previous studies in other cell types demonstrating that the activation of p53 in senescence occurs primarily via post-translational modifications (Webley et al., 2000). Similarly, although up-regulation of p21 CIP1 and p16 INK4A at the transcript level is detectable in these senescent HCEC, this is not reflected to the same extent at the protein level. This is consistent with a recent report that p21 CIP1 expression is controlled separately at the transcriptional and translational levels in human fibroblast senescence and that elevated p21 CIP1 transcription does not simply translate into enhanced p21 CIP1 protein (Lian & Gallouzi, 2009  The available in vivo data on CDKI expression in HCEC, although limited by small sample numbers, is consistent with an elevation of p16 INK4a at both protein and transcript levels in HCEC from older donors (Enomoto et al., 2006;Song et al., 2008). The situation with p21 CIP1 is less clear, with apparent elevation at the protein (Enomoto et al., 2006) but not transcript (Song et al., 2008) levels in different studies. However, the small sample numbers and innate variation in p21 CIP1 levels in these studies does not render such observations inconsistent with ours. , on the basis of their failure to detect changes in telomere length with HCEC ageing in vivo, proposed that entry into senescence in HCEC is telomere independent. We have conducted an in-terventional test of this hypothesis through ectopic expression of hTERT and have shown that it does not extend lifespan. The Joyce group (Joyce et al., 2011) have also recently shown that the capacity of HCEC to establish primary cultures and proliferate in vitro correlates with the amount of oxidative DNA damage received by the cells in vivo prior to explants, suggesting that the capacity of HCEC to proliferative is significantly limited by oxidative damage to DNA. We have shown that interventions that manipulate the levels of pro-oxidants (e.g., reduced ppO 2 and the addition of ascorbic acid 2-phosphate) lengthen replicative lifespan, but that this does not produce immortalization even when telomere erosion is blocked using ectopic hTERT. However, when telomerase expression is combined with either abrogation of p53 or CDK4 over-expression, we find that HCEC reproducibly immortalize. Given the evidence that innate telomere-based protective responses exist that act to reduce oxidative damage to cells (Lee et al. 2009), and that the catalytic subunit of telomerase may itself be protective against oxidative damage by a mode of action independent of its effects on the telomere (Ahmed et al. 2008), the failure to immortalize with a combination of telomerase and antioxidants appears incompatible with a simple model based on a linear relationship between the accumulation of DNA damage and the onset of replicative senescence in HCEC. The exact nature of the second parallel signal(s) that act to limit HCEC proliferation in concert with telomere erosion is unclear, but could be dissected through additional studies using HCEC cultured under different conditions, and with different proliferative lifespan barriers ablated.
However, from a practical perspective, our key finding is the identification, through interventional approaches, of the role of the p53 pathway, the telomere pathway, and the CDK4 ⁄ pRb axis in the control of senescence in HCEC. Corneal transplants have a high success rate, but the short supply of donor corneas with competent endothelium (Engelmann et al., 2004) and the continued risk of graft failure because of immunological rejection (Chong & Dana, 2008) remain challenges that may be addressed by the development of bioengineered corneal grafts. In addition, standardized in vitro systems containing cultured corneal cells could provide appropriate replacements for routine tests in which animal corneas are commonly used, such as drug permeation studies (Reichl et al., 2004). Hence, there would be great potential utility in the development of human corneal cell lines that are sufficiently differentiated to meet these needs.
Towards this aim, our study demonstrates that it is possible to produce immortalized HCEC that retain differentiated function. Our data show that in addition to retaining the ability to exit the cell cycle in response to contact inhibition, immortalized HCEC-CDK4-hTERT cells retain a HCEC transcriptional profile (Fig. S6), as well as the temperature-sensitive potassium channel that is a key marker for HCEC (Rae & Watsky, 1996), considerably longer than would be required to generate sufficient numbers of cells for most applications. Although the specific cell lines that we have generated would not be good candidates for immediate human in vivo studies, they may have significant utility for research on the biology of this cell type, or use in genotoxicology and drug delivery studies.

Cells and cell culture
Primary HCEC, strain 288 ⁄ 97, were cultured in medium F99 comprising a 1:1 mixture of Ham's F12 (Invitrogen, Paisley, UK) and M199 (Invitrogen) supplemented with 5% foetal calf serum (FCS), 100 lM sodium L-ascorbate (Sigma-Aldrich Company Ltd., Poole, UK), bovine insulin (Sigma-Aldrich Company Ltd.), 2.5 lg mL )1 transferrin (Sigma-Aldrich Company Ltd.), 40 pM sodium selenite (Sigma-Aldrich Company Ltd.) and 10 ng mL )1 basic fibroblast growth factor (Sigma-Aldrich Company Ltd.). Cells were cultured on plastic, coated using a solution of 5 mg mL )1 chondroitin sulphate (Sigma-Aldrich Company Ltd.) and 5 lg mL )1 mouse laminin (Invitrogen). They were maintained at 37°C in a humidified atmosphere containing 5% CO 2 , 95% air unless otherwise specified and re-fed three times a week. Cellular lifespan was expressed in PD, calculated using the standard formula. To study the effects of reducing oxidative stress, HCEC were cultured as above but in 3% O 2 and the sodium ascorbate in the medium was replaced with 400 lg mL )1 L-ascorbic acid 2-phosphate magnesium salt (WAKO Biochemicals, Osaka, Japan). For the p38 MAPK inhibition experiment, experimental and control cultures were re-fed daily with medium F99 containing either 10 lM SB203580 (Tocris Bioscience, Bristol, UK) in DMSO or DMSO only, respectively. Quiescent cultures were obtained for the transcriptional profiling experiment by seeding the cells at high density and allowing them to grow to confluency, followed by replacing the medium with F99 containing 0.5% FCS for 3 days before harvesting. Exit from the cell cycle was determined by loss of Ki67 staining.

Retroviral gene transfer
Amphotropic retrovirus vectors expressing HPV16 E6 genes from a pLXSN construct, packaged in PA317 cells, were kindly provided by Denise Galloway, Seattle, Washington, USA. Human wild-type CDK4 cDNA in pBluescript, kindly given by Gordon Peters, CRUK LRI, London, was recloned into pLXSNneo. hTERT cDNA in pGRN121, a gift from Geron Corp, Menlo Park, CA, USA, was inserted in pBABE retroviral vectors with puro or neo selection. Empty pBABEneo or pBABEpuro vectors packaged in wCrip cells were used as controls. shRNA targeted to p53 was expressed from pMKO.1PS, kindly donated by William Hahn, Harvard Medical School, packaged in wCrip. Empty pMLKO.1PS was used as the puromycin control. HCEC for retroviral infection were plated in 6-cm Petri dishes 24 h prior to infection at a seeding density of 6000 cells cm )2 . Prior to infection, cells were treated with polybrene at 8 lg mL )1 for 1 h. Cells were infected by treating with 2 mL of filtered retroviral supernatants supplemented with 8 lg mL )1 polybrene for 2 h, followed by the addition of 2 mL F99 and incubation for a further 24 h before replacing the supernatant with 5 mL of fresh F99 medium without selection. Forty-eight hours after the start of infection, HCEC were passaged into fresh dishes with selective medium and cultured as described above to produce bulk cultures of mixed clones.

RNA isolation and microarray processing
Cells were rinsed briefly with phosphate-buffered saline and then lysed in situ using TRIzol (Invitrogen) as per the manufacturer's instructions. Integrity of the total RNA was confirmed by spectrophotometry and using an Agilent 2100 Bioanalyzer (Agilent Technologies UK Ltd, Edinburgh, UK). Fifteen microgram of labelled cRNA was prepared, essentially as per the standard Affymetrix protocol, from 10 lg of total RNA, using the Superscript II system (Invitrogen) and BioArray High Yield Kit (Enzo Life Sciences Ltd., Exeter, UK), and then hybridized to U133A GeneChips. Expression values and absent ⁄ present calls were calculated using MAS 5.0 (Affymetrix), using default parameters and a TGT of 100.

Ki67 immunodetection
Cells on coverslips were fixed with methanol ⁄ acetone, 1:1, for 5 min then incubated with a one in 20 dilution of mouse anti-human Ki67 antibody (DAKO UK Ltd., Ely, UK) in 0.1% FCS ⁄ PBS at 4°C overnight. Coverslips were washed and incubated with a one in 30 dilution of rabbit antimouse IgG FITC-conjugated antibody (DAKO UK Ltd.) in 0.1% FCS ⁄ PBS in the dark for 60 min. Nuclei were counterstained 4¢6-diamidino-2-phenylindole (DAPI; Vector Laboratories Ltd., Peterborough, UK). One thousand DAPI-stained cells were counted and the labelling index estimated as the proportion of those cells that stained positively for Ki67.

Electrophysiological measurements
Cells were bathed in HEPES Ringer solution (136 mM NaCl, 2.6 mM CaCl 2 , 2.4 mM KCl, 1.2 mM MgCl 2 , 15 mM HEPES, 10 mM glucose, pH 7.4). The pipette solution was 110 mM KCl, 3.0 mM MgCl 2 , 40 mM HEPES, 3 mM EGTA, pH 7.4. Voltage-gated macroscopic currents were recorded from single cells by a whole-cell patch-clamp technique as described previously (Faragher et al., 1997a). Temperature-sensitive currents were investigated by stepping from a holding potential of )5 mV to between )65 and +75 mV in 10 mV steps for a 500 ms duration at 22 or 34°C.

Supporting Information
Additional supporting information may be found in the online version of this article:

Table S1
Over-represented transcription factors by enrichment of their target genes in a cluster of genes up-regulated at senescence in HCEC. As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer-reviewed and may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.