Long-Term Maintenance of Human Pluripotent Stem Cells on cRGDfK-Presenting Synthetic Surfaces

Synthetic human pluripotent stem cell (hPSC) culture surfaces with defined physical and chemical properties will facilitate improved research and therapeutic applications of hPSCs. In this study, synthetic surfaces for hPSC culture in E8 medium were produced for screening by modifying two polymer brush coatings [poly(acrylamide-co-acrylic acid) (PAAA) and poly(acrylamide-co-propargyl acrylamide) (PAPA)] to present single peptides. Adhesion of hPSC colonies was more consistently observed on surfaces modified with cRGDfK compared to surfaces modified with other peptide sequences tested. PAPA-coated polystyrene flasks with coupled cRGDfK (cRGDfK-PAPA) were then used for long-term studies of three hPSC lines (H9, hiPS-NHF1.3, Genea-02). Cell lines maintained for ten passages on cRGDfK-PAPA were assessed for colony morphology, proliferation rate, maintenance of OCT4 expression, cell viability at harvest, teratoma formation potential, and global gene expression as assessed by the PluriTest™ assay. cRGDfK-PAPA and control cultures maintained on Geltrex™ produced comparable results in most assays. No karyotypic abnormalities were detected in cultures maintained on cRGDfK-PAPA, while abnormalities were detected in cultures maintained on Geltrex™, StemAdhere™ or Synthemax™. This is the first report of long term maintenance of hPSC cultures on the scalable, stable, and cost-effective cRGDfK-PAPA coating.

Human pluripotent stem cells (hPSCs) show potential for drug discovery, studying disease mechanisms, and for clinical applications including the generation of differentiated cell types for transplantation. hPSC cultures were originally maintained in co-culture with mitotically inactivated feeder cells. Although surfaces coated with feeder cells can provide the necessary supporting factors for hPSC maintenance, their use inherently results in variable culture conditions. A number of synthetic hPSC culture surfaces have since become commercially available, yet they are not commonly used. The widespread use of synthetic, chemically defined culture conditions would improve reliability in cell manufacture, reduce inter-laboratory variation in hPSC cultures, improve the consistency of experimental results and facilitate the derivation and application of clinical grade hPSC lines. This paper describes the screening of synthetic surfaces for adhesion and maintenance of hPSC. Synthetic peptide-presenting polymer coatings, able to support adhesion and maintenance of hPSC cultures, were used to screen for peptide ligands that were individually capable of mediating hPSC adhesion. The polymers poly(acrylamide-co-acrylic acid) (PAAA) and poly(acrylamide-co-propargyl acrylamide) (PAPA) were modified to present peptides with previously reported roles in cell adhesion and then subjected to a hPSC adhesion assay. Long term hPSC culture studies were then performed, comparing the lead surface to commercially available, synthetic hPSC culture surfaces (StemAdhere ™ and Synthemax ™ ) and to Geltrex ™ -coated surfaces. The lead surface presented cRGDfK, a cyclic peptide that has been optimised as a potent and selective inhibitor of the αvβ3 integrin, which is able to bind both αvβ3 and αvβ5 integrin 1 .
In this study we used a type of initiator-free, surface-initiated polymer grafting method that is relatively unknown [2][3][4] , but which can be broadly applied to polymer, organic and inorganic surfaces 3 . For the coatings presented here, a methodology was optimised which relied on multiple passes of the multiwell plates or flasks beneath a high intensity UV light in the presence of monomer solution and in the absence of oxygen. This method H9-OCT4 2AChryIM/w adhesion assay for screening peptide-modified polymer coatings. The approach outlined in Fig. 2 was used to screen for hPSC adhesion to 23 peptide-modified PAAA coatings, which had been prepared using 40 passes under a high intensity UV light source (PAAA-40UV) and to 14 peptide-modified PAPA coatings that had been synthesised with 30 UV passes (PAPA-30UV) 5,6 . The full list of peptides is provided in Supporting Information Table S1, with chemical properties regarding solubility described in Supporting Information Table S2. PAPA coatings were used for lysine-containing peptides, since the presence of lysine residues would interfere with the carbodiimide coupling approach used with PAAA coatings. H9-OCT4 2AChryIM/w cells were observed to adhere to coatings that had been modified with the cRGDfK peptide (cRGDfK-PAAA and cRGDfK-PAPA) as well as peptides 20 (pep20-PAPA), 31 (pep31-PAPA), 34 (pep34-PAAA) and 35 (pep35-PAAA), which represented 14% (5/36) of all peptides tested. More colonies were observed to adhere to wells coated with Geltrex ™ or cRGDfK-modified surfaces than to polymer-coated wells that had been modified with the other peptide (Supporting Information Figure S1).
HPSCs were often observed to form rounded clumps on pep34-PAAA or pep35-PAAA surfaces, and appeared to be only loosely attached to the surface (Supporting Information Figure S1). To further examine the pep34-PAAA and pep35-PAAA coatings, we tested thinner PAAA-25UV coatings and increased concentrations of peptide solutions up to 600 μM (Supporting Information Figure S2A). Although these conditions did slightly improve each substrates efficiency, they could not match the cRGDfK-PAAA coating (Supporting Information Figure S2B). This was repeated with another hPSC line, hiPS-NHF1.3, and found to be consistent (Supporting Information Figure S2C).
The only peptide-modified polymer coatings that appeared to support hPSC cultures in E8 medium were those modified with cRGDfK. Although cRGDfK is commonly considered to interact with αvβ3 integrins 1 , reports of the presence of αvβ3-integrins on hPSC cell surfaces are inconsistent and cRGDfK was optimised to mimic vitronectin, which also binds αvβ5-integrins 13,14 . In order to assess whether hPSC adhesion to cRGDfK-PAPA in the present study was mediated by integrins, the surfaces of hPSCs were stained for αvβ3and αvβ5-integrins. The presence of αvβ3 and αvβ5 integrins were detected by flow cytometry on the surface of MDA-MB-435 control cells. In contrast, αvβ5 but not αvβ3 integrin was detected on the surface of H9 hPSCs (Supporting Information Figure S4).
Due to the low peptide concentrations required to produce cRGDfK-PAPA coatings that reliably allowed attachment and proliferation of hPSCs (data not shown), the associated cost-reduction, the reported preference of hPSCs for stiffer substrates 15  Characterisation of synthetic hPSC culture surfaces. To verify the presence of coatings of the expected composition and chemistry, X-ray photoelectron spectroscopic (XPS) analysis was performed on the surface of modified culture flasks. The surface composition and high-resolution carbon 1 s spectra obtained from both uncoated and PAPA-30UV coated flasks are presented in Fig. 4. The atomic composition and high-resolution C 1 s spectra obtained from the analysis gave clear evidence of the presence of a PAPA coating on the surface of a TCPS substrate; significant increases were seen in the nitrogen and oxygen content of the surface compared to TCPS, which contained lower O atomic percentages and a very small N atomic percent. Furthermore, fitting of the high-resolution C 1 s spectra gave a component fit that was consistent with the presence of a polymer coating containing acrylamide species. Strongly indicative of acrylamide species is the presence of an intense C 4 component at a binding energy of approximately 288 eV, compared to TCPS. The presence of an over-layer coating is confirmed by a reduction of the intensity of components (C 6 -C 8 ) specific to the underlying TCPS.
Surface analysis of Geltrex ™ , Synthemax ™ and StemAdhere ™ coatings was also carried out (results not shown). Briefly, the analysed results obtained from Geltrex ™ and StemAdhere ™ coatings were consistent with the presence of protein coatings of varying thickness, while the Synthemax ™ coating was consistent with the The upper line shows the wild type OCT4 locus with exons marked in grey. The relative position of the OCT4 promoter (P) and the point within the 3′ UTR against which specific TALENs were directed is indicated. The targeting vector (middle line) included a 5.4 kb 5′ homology arm that joined sequences encoding a T2A peptide (2A) and mCherry (Chry) in frame with the OCT4 coding sequences. Selection of correctly targeted clones was facilitated by an internal ribosomal entry site (IRES) preceding a Neomycin resistance gene optimised for expression in mammalian cells (Meo). The three translation products of the targeted allele are shown at the bottom. The gel electrophoresis image shows that the correct size fragment (3.6 kb) was detected by PCR screening in 5 of the 6 clones screened. (B) Validation of H9-OCT4 2AChryIM/w hESC reporter fidelity using intra-cellular flow cytometry for OCT4 expression. At the day of passaging from maintenance culture (day 0) 99% of undifferentiated cells were mCherry pos (left panel). Following 5 days differentiation, 20% of cells continued to express mCherry. mCherry pos and mCherry neg cells were sorted at day 5 and each fraction stained for OCT4 protein expression using intracellular flow cytometry. This analysis showed that 84% of mCherry pos cell retained OCT4 protein expression whilst only 9% of cells in the mCherry neg fraction expressed OCT4. OCT4 pos mCherry pos cells could be readily distinguished from the complementary OCT4 neg mCherry neg population. Appearance and proliferation rates of hPSC cultures on synthetic surfaces. The potential of cRGDfK-PAPA coatings as hPSC culture surfaces was assessed using the H9, hiPS-NHF1.3 and Genea-02 hPSC lines. Starter cultures were thawed from banks of validated hPSC lines, which had been maintained for ten passages in E8 medium on Geltrex ™ -coated flasks to adapt cells to control conditions (Supporting Information Figure S5). Thawed cultures were passaged twice on Geltrex ™ to allow recovery from the thawing process before being split into parallel cultures on each of the test surfaces.
Cell lines for experimentation were then maintained over ten passages in flasks coated with either cRGDfK-PAPA, StemAdhere ™ , Synthemax ™ or Geltrex ™ , all hPSC cultures consistently formed tightly-packed colonies which displayed a typical range of hPSC morphologies (Fig. 5). Generally, flasks coated with cRGDfK-PAPA or Synthemax ™ contained hPSCs that were morphologically indistinguishable from control cells maintained in Geltrex ™ -coated flasks. However, cultures maintained in StemAdhere ™ -coated flasks were observed to seed as single cells or smaller colonies than on the other surfaces and then to form colonies that appeared to be flatter and less tightly-packed (Fig. 5). These observations were consistent with observations made for H9-OCT4 2AChryIM/w cells maintained for 3 days on each surface (Supporting Information Figure S6).

Figure 2.
Screening approach feeding into long term experimental plan. A schematic diagram illustrates the screening process used to identify peptides that, when chemically bound to PAAA or PAPA coatings, produced a surface able to bind and maintain short-term culture of hPSCs. Three batches of peptide-coated plates were prepared and triplicate wells in randomised locations of each plate and modified with either the cRGDfK peptide, the non-binding negative control cRADfK peptide or a test peptide. Plates were seeded with H9-OCT4 2AChryIM/w hPSCs at a density of 15 000 cells/cm 2 . Geltrex ™ (GX)-coated control wells were seeded in parallel at equal and one-third (5 000 cells/cm 2 ) density. Colony number and morphology was assessed at 48 hours post-seeding and growth and maintenance of mCherry were assessed at day 4 post-seeding. HPSC cultures were then maintained for ten passages on surfaces coated with the novel candidate hPSC culture surface (PAPA-cRGDfK). hPSCs were characterised before and after this culture period, and comparisons were made to cultures maintained in parallel on Geltrex ™ , StemAdhere ™ and Synthemax ™ . Proliferation rates appeared to be more affected by cell-line variation than by surface type, with the passage duration of Genea-02 cultures being particularly variable between and within cultures maintained on different surfaces (Fig. 6Ai). When Genea-02 cells were initially seeded onto StemAdhere ™ -coated flasks, a typical mix of single cells and small colonies were observed to adhere to the surface. However, from day 4 many of these small colonies and single cells detached, reaching an estimated nadir around day 7 (Supporting Information Figure S7). As few as 10 colonies per flask were observed to survive at this point, yet the colonies continued to expand. This initial adaptation of Genea-02 hPSCs to culture on StemAdhere ™ was repeated a further three times with consistent results (data not shown). When the Genea-02 culture on StemAdhere ™ reached day 21 the colonies were very large, with dense centres surrounded by borders of morphologically normal hPSCs (Supporting Information Figure S7D,H,L), yet they were only estimated to cover 5% of the surface of the flask. At this point the cultures were harvested using EDTA (0.5 mM) and wholly transferred to a single StemAdhere ™ -coated flask, from which the culture recovered and ultimately survived for ten passages (Fig. 6A). No significant differences were observed between hPSC cultures maintained on different culture surfaces in the number of days required for hPSC cultures to recover from each passage or in the cellular yield (Fig. 6Aii). Subtle changes in passage duration were also mirrored by changes in yield, and all cultures were harvested with consistently high cell viability ( Fig. 6Aii-iv), which further indicated that hPSC proliferation was consistent across all culture surfaces.
Cultures of hPSCs maintained on StemAdhere ™ -coated flasks were also observed to contain regions where the cells were highly compacted and colonies appeared to pile up on top of each other, as well as regions with well-defined borders where cells did not adhere (data not shown). Such phenomena were not observed in hPSC cultures maintained on other surfaces. While cultures maintained on StemAdhere ™ lifted readily after shorter incubations in EDTA than were required to remove hPSCs from other surfaces, cell scraping was occasionally required to remove low numbers of cells from cRGDfK-PAPA and Geltrex ™ -coated surfaces during EDTA-mediated harvesting. However, hPSCs were more difficult to dissociate from Synthemax ™ , with cell scraping commonly required during harvesting with EDTA or TrypLE Express (Fig. 6B).
Genetic stability of hPSCs maintained on synthetic culture surfaces. Genetic stability of all hPSC cultures was assessed by G-banding karyotyping (Fig. 6C,i-ii) and using the "transcriptional karyotype" feature of the PluriTest ™ assay (Fig. 6D). The transcriptional karyotype is a heat map that presents gene expression levels arranged by genetic locus and illustrates whether the genes located within each cytoband are up-or down-regulated compared to the reference data set.
All hPSC cultures that had been maintained on cRGDfK-PAPA were observed to be karyotypically normal by G-banding karyotyping (Fig. 6C). However, trisomy of chromosome 12 was detected at a moderate frequency in Genea-02 (11/20 cells) and H9 (7/20 cells) cultures maintained on StemAdhere ™ . Trisomy-12 was also detected in the Genea-02 line by PluriTest ™ (Fig. 6D, arrow), but the matching abnormality was not detected in the H9 population. This indicates that transcriptional karyotyping may be suitable for identifying highly abnormal populations of cells, but it may not be sensitive enough for early detection of low-frequency abnormalities. It was also interesting to observe genetic loci that were differentially regulated between cultures in this study and the reference data sets, probably resulting from the use of a different culture media to that used to originally populate the PluriTest ™ dataset (Fig. 6D, arrowheads).
Additional low-frequency chromosomal abnormalities were only detected by G-banding (Fig. 6C,i-ii) and were predominantly amplifications or duplications of chromosomes 12 and 20, which have been commonly observed in a range of hPSC cultures 16 Figure S8). These findings were inconsistent with an earlier report that OCT4 protein was lost from hPSCs maintained on cRGDfK-presenting culture surfaces 10 Furthermore, each hPSC culture passed teratoma assays for the in vivo formation of tissues representing the three embryonic germ layers (Fig. 6B, Supporting Information Figure S9) and the PluriTest ™ pluripotency assay; all samples produced Novelty Scores and Pluripotency Scores typical of hPSC populations ( Fig. 7C). Minor variations between the Pluritest ™ data generated from this experiment and the reference data The PluriTest ™ assay also generated a phylogenetic tree, which indicated that most hPSC lines maintained on cRGDfK-PAPA were more closely related to control culture maintained on Geltrex ™ (Supporting Information Figure S10) than to other cell lines.

Discussion
The results presented herein show that our method was capable of generating surface coatings that were well-defined in terms of their composition, and to which peptides could be readily covalently attached via at least two well-understood coupling chemistries [water soluble carbodiimide and Cu(I)-mediated click coupling]. The coatings were stable, could be sterilised using industry standard methodologies, and were readily applicable to maintaining hPSC cultures. E8 culture media incubated for 66 h in PAAA-and PAPA-coated wells did not display cytotoxic properties when applied to fresh hPSC cultures (data not shown). Also, following appropriate peptide modification, the coatings readily supported both attachment and maintenance of hPSCs cultures.
It was found that hPSCs bound more efficiently to cRGDfK-modified PAAA and PAPA coatings than they did to coatings modified with other test peptides. A culture system comprising E8 medium and cRGDfK-PAPA coatings was then shown to support the long-term maintenance of undifferentiated hPSC cultures similar to control cultures on Geltrex ™ . hPSC cultures were karyotypically stable over ten passages when maintained on cRGDfK-PAPA-coated surfaces, while cultures on Geltrex ™ , StemAdhere ™ and Synthemax ™ all displayed karyotipic abnormalities to varying degrees. These findings will contribute to the ongoing development of peptide-presenting surfaces for the maintenance of hPSC cultures.
Of the 37 peptide-presenting polymer coatings assessed, only cRGDfK-modified PAAA and PAPA coatings appeared to be feasible for use in hPSC culture. This finding was surprising, since maintenance of hPSC cultures has not previously been reported on surfaces that present cRGDfK-alone, despite the widespread use of cRGDfK in various applications including in vitro adhesion of osteoblasts 18 , chondrocytes 19,20 and melanoma cells 14 . Only one other publication had reported an attempt to maintain hPSC cultures on such a surface 10 .
The culture surface produced by Klim et al. 10 presented biotinylated cRGDfK peptides from streptavidincoated tissue culture-treated polystyrene. Although two hESC lines were reported to adhere to and proliferate on this surface for 3 passages, abnormal morphology, substantial morphological differentiation and loss of OCT4 were observed. On the other hand, the cRGDfK-modified coatings reported herein were consistently observed to support adhesion of large numbers of both hESC and hiPSC colonies and were able to maintain karyotypically normal OCT4-positive hPSC cultures for at least ten passages.  It was interesting to note that, in this study, the only peptide-modified polymer coatings that appeared to be feasible for use as hPSC-supportive culture surfaces were those modified with cRGDfk. This finding differs from a report by Klim et al. 10 , who reported that cRGDfk was not a reliable hPSC culture surface. A possible reason for the poor performance of the cRGDfK-presenting surface reported by Klim et al. 10 is instability of the underlying substrate. Although biotin-streptavidin interactions are some of the strongest non-covalent bonds (bond-dissociation energy of −18 kcal/mol), carbon-carbon covalent bonds are much stronger (−83 kcal/mol) 21 . αv-containing integrins have been observed to selectively localise with high force regions within focal adhesions when bound to linear RGD 22 , and to break biotin-avidin bonds when bound to cRGDfK 23 . Although no loss of cellular attachment was observed in the study by Jurchenko et al. 23 , the experiments were performed over a shorter time frame and likely with a lower starting ligand density. It is therefore conceivable that integrin-mediated removal of peptide ligands could have affected cell adhesion in the study by Klim et al., while the more diffuse HSPGs may have exerted less force and not damaged the surface 10 .
The requirement of Y-27632 supplementation for hPSC culture on Klim's cRGDfK-presenting surface may reflect the role of Y-27632 in inhibiting myosin contractility 10,24 , which would inhibit cell-mediated destruction of the surface. Conversely, PAAA and PAPA are covalently bonded from the peptide ligand through to the underlying polystyrene and so physical removal of cRGDfK peptide ligands by cellular forces in the current study is highly unlikely.
When coatings were modified with three of the remaining four RGD-containing ECMP-derived test peptides (11, osteopontin; 34, bone sialoprotein; 35, vitronectin), low numbers of hPSCs were observed to bind (Supporting Information Figures S4 and S5). Poor adhesion of hPSCs to pep34-PAAA and pep35-PAAA coatings was particularly surprising, since these peptides have been presented on surfaces previously reported to support hPSC culture, including Corning Synthemax ™ 25,26 .
During the development of Corning Synthemax ™ , acrylate coatings were modified with peptides via a carbodiimide chemistry and using peptide concentrations similar to those used to modify PAAA coatings in the present study 26 . A reduction in hPSC adhesion efficiency to peptide acrylate coatings was observed when the surfaces were modified with solutions containing concentrations of peptide between 250 and 500 μM in the Melkoumian study 26 . However, increasing the peptide concentration used to modify pep34-PAAA and pep35-PAAA surfaces to 600 μM did not result in levels of hPSC adhesion equivalent to cRGDfK-PAAA surfaces that had been modified with solutions containing only 200 μM of peptide (Supporting Information Figure S2A). Another recent study reported similar input peptide requirements and stated that it was "necessary to use high concentration of olig-oVN [peptide 34]" to mediate hPSC adhesion 27 . Furthermore, a recent study found that culture media spiked with cyclic RGDfC and RGDfV peptides blocked hPSC adhesion more effectively than media spiked with linear RGD peptide 28 . Peptide-cell interactions can be affected by ligand-CAM affinity and/or integrin subtype specificity. The RGD motif has been reported to bind eight of the 24 known integrin subtypes, including αvβ3, αvβ5 and α5β1 29,30 . The CAM-ligand affinity and integrin subtype specificity of the RGD motif are conformation-dependent and can be affected in vivo by flanking sequences, tertiary structure and post-translational modifications, although the latter two are missing from the short, linear, synthetic peptides used in this study 31 . It is therefore somewhat unsurprising that surfaces modified with the synthetically-derived cRGDfK, for which integrin affinity has been optimised in vitro in the absence of post-translational modifications or a surrounding protein structure, bound hPSCs more efficiently than RGD-containing peptides that have been isolated from the in vivo environment. Essentially, the cyclic structure of the cRGDfK peptide mimics the tertiary structure in the native vitronectin protein as well as providing chemical stability.
It has previously been reported that cyclic peptides containing the RGDf sequence specifically interact with αvβ3 integrins 1 , that undifferentiated hPSC cultures attach to but are unable to be maintained on cRGDfK-presenting surfaces 10 and that hPSCs are unable to bind to surfaces that only present αvβ3-specific ligands, but can adhere to vitronectin-coated surfaces (αvβ3-and αvβ5-binding) 14 . Unfortunately cRGDfK-modified surfaces were not included in the latter study. However, reports of hPSC integrin expression profiles are inconsistent, and the specificity of cRGDfK-αvβ3 integrin interactions has not been demonstrated directly [32][33][34] . The inability to detect αvβ3 integrin on undifferentiated hPSCs by flow cytometry (Supporting Information Figure S4) was consistent with the report by Klim et al. 14 , but not with results from an earlier study Synthemax ™ ). Means are presented with scale bars representing standard deviations. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001 by unpaired two-tailed t-tests. (C) G-banding karyotyping assessment of hPSCs maintained on defined culture surfaces is summarised. The type and frequency of karyotypic abnormalities observed are plotted and arranged. Karyotypes are represented by the following colours: black: partial duplication of chromosome 20 [add (20)], black stripes on grey: loss of one sex chromosome (45, X), dark grey: includes additional unidentifiable chromosome (47, + mar), grey crosshatch on pale grey: trisomy of chromosome 12 (+12), light grey: normal karyotype (i) by surface type and (ii) by surface type and then by cell line. Full karyotypes are shown in Supplementary Figures 11 and 12 (D) The PluriTest ™ assay was applied to passage 10 cultures and generated a transcriptional karyotype, which compared the expression levels of groups of genes in sample data sets to homologous genes in the pluripotent reference data sets. A heat map is formed displaying cytobands in which overall gene expression levels are higher (darker red indicates higher expression) or lower (blue, with darker blue indicating lower expression) in samples than is predicted by the pluripotency model.   13 . Subsequent studies using antibody-blocking experiments demonstrated that hPSC interactions with the Synthemax ™ culture surface are predominantly mediated by αvβ5 integrins, not αvβ3 28,35 and it has also been reported that blocking αvβ3 integrins had no observable effect on hPSC adhesion to Matrigel ™ 36 . Our findings that αvβ5 integrin is detected, and αvβ3 integrin is not, by flow cytometry on the surface of H9 hESCs are entirely consistent with these results (Supporting Information Figure S4). These results collectively indicate that hPSC adhesion to cRGDfK-presenting polymer substrates is likely to be mediated by αvβ5 integrins rather than αvβ3 integrins.
Regardless, hPSC adhesion efficiency to cRGDfK-modified PAAA and PAPA-coated surfaces was greater than surfaces modified with other test peptides. These results indicate that cRGDfK is an ideal peptide for mediating hPSC adhesion to peptide-presenting polymer coatings.
Eleven polymer coatings were modified with HSPG-binding peptides. Of these, hPSCs were only observed to inefficiently and unstably bind to pep20-PAPA and pep31-PAPA. Such a poor result was particularly interesting for pep31-PAPA, since peptide 31 is a vitronectin-derived peptide that has been included in a range of hPSC culture surfaces [10][11][12]37 . Nevertheless, hPSC adhesion to pep31-PAPA remained low despite attempts to optimise reaction conditions (Fig. 3).
The observation that hPSCs bound more efficiently to surfaces modified with RGD-containing peptides (34, 35 and cRGDfK) than to surfaces modified with HSPG-binding peptides is inconsistent with a previous report of hPSC adhesion to self-assembled monolayer (SAM) of alkanethiols that had been modified at a high density with any of 9 different peptides 10 . It was observed that a lower density of peptide 31 was required to mediate hPSC adhesion to the SAMs than of other (integrin-or HSPG-binding) peptides. This "low" peptide density was still 1000-fold higher than the threshold surface peptide density for hPSC attachment to cRGDfK-PAPA (data not shown). It is interesting to note that every report of hPSC culture on surfaces that present peptide 31 or other HSPG-binding peptides has used culture medium supplemented with Y-27632 or other Rho pathway inhibitors [10][11][12]15,[38][39][40] .
The downstream effects of Y-27632 are unclear 41 and may include effects on cell fate bias 42 and increasing cell-cell interactions and adherence 43 . Surfaces presenting a combination of HSPG-and integrin-binding ligands have been reported to obviate the requirement for Y-27632 supplementation 10,13 . However the present study is the first in which surfaces modified with cRGDfK-alone have been reported to mediate adhesion of hPSCs in the absence of Y-27632 supplementation, let alone outperform coatings modified with HSPG-binding peptides. The formation of multilayered aggregates, detachment and loss of OCT4 observed in hPSCs cultured on pep31-PAPA (Supporting Information Figure S3) resembled observations made about hPSCs binding to the cRGDfK-biotin-streptavidin surface developed by Klim et al. 10 So Y-27632 supplementation may compensate for inadequate hPSC culture surfaces in a ligand-and receptor-independent fashion.
Other peptides that have previously been reported to feature in hPSC-binding surfaces but failed to do so in the current study included randomly generated peptides (4 and 5) that were identified from a phage display library 40 . They may have been affected by the use of conditioned medium generated from feeder culture of mitotically inactivated mouse embryonic fibroblast and/or by cross-contamination between adjacent array elements 44 . Conducting the current screen in individual wells instead of on a micro-patterned surface reduced the risk of such contamination. Surfaces modified with laminin-derived peptides (16, 17 and 30) were only observed to mediate hPSC adhesion when applied in combination with BSA-treated surfaces 13 . This is the first follow-up publication using surfaces modified with these peptides and the results support the hypothesis that laminin-derived peptides are individually inadequate for supporting maintenance of undifferentiated hPSC cultures.
Since hPSCs bound more efficiently to surfaces modified with cRGDfK than those modified with any other peptide, a long-term study was conducted using cRGDfK-PAPA. Overall, hPSC cultures that had been maintained for at least ten passages on the cRGDfK-PAPA coating were comparable to cultures maintained in parallel on the commercially available culture surfaces for the presence of OCT4 (Fig. 7A), gene expression (Fig. 7C) or in vivo differentiation potential (Fig. 7B).
Aside from the quality of the cell product, factors worth considering when selecting a scalable surface for maintaining hPSC cultures include cost, preparation, shelf life and logistical challenges involved in passaging cells (e.g. cell scraping). The four test surfaces have been ranked for these parameters in Table 1 (1) to least (4) preferred with regards to cost, preparation, shelf life and the requirement for scraping to harvest cells. For example, Geltrex is the cheapest and is therefore ranked 1 (most preferred) for cost, while StemAdhere TM requires the least scraping and is therefore ranked 1 for scraping. Cost was ranked according to a recently published comparison 57 , preparation was ranked according to the number of steps involved in the coating procedures at the point of passage (commercially available cRGDfK-PAPA would be prepared as a complete coating onto which cells would be seeded directly), the reported shelf lives of commercially available surfaces were compared to the tested shelf life of cRGDfK-PAPA (data not shown) while the cell scraping rankings are a summary of the results presented in Fig. 6B. coating approach is readily scalable in theory, cRGDfK-PAPA-coated flasks have only actually been produced at relatively small scale, so the commercial cost of cRGDfK-PAPA could not be calculated. Nevertheless, efficient hPSC adhesion has been consistently observed on cRGDfK-PAPA coatings modified with concentrations of cRGDfK as low as 1 μM (data not shown), which reflects a potential saving of peptide cost as great as 500-fold and indicates that cRGDfK-PAPA is likely to be less expensive to produce than Synthemax ™ 26 .
The commercial benefits of low-cost surfaces that require little preparation and have a long shelf life are self-evident. Although cell scraping is included in some hPSC passaging protocols [45][46][47] , such manual harvesting is not easily scalable and could not be applied to multilayer cell stackers, cell factories, or three-dimensional cell culture systems including microcarriers and hydrogels 48 .
Aside from the logistical issues involved in large scale hPSC cultures, the use of cell scraping is also undesirable because the surface coating is likely to be harvested along with the cells, particularly for non-covalently bonded products such as StemAdhere ™ and Synthemax ™ . The implications for incorporation of any coating-derived contaminants would therefore need to be assessed prior to the application of cells cultured in these systems to cell-based diagnostic assays and therapies. In the current study, hPSCs maintained in cRGDfK-PAPA-coated or Geltrex ™ -coated flasks were observed to require harvest by cell scraping significantly less often than cultures on Synthemax ™ -coated flasks, while hPSCs maintained in StemAdhere ™ -coated flasks lifted readily after a shorter incubation in EDTA. The ease of harvesting hPSCs from StemAdhere ™ -coated flasks, the associated improvement in viability (Fig. 6A), and the short shelf life compared to other coatings (Table 1) may reflect a reduced coating stability.

Conclusions
PAAA and PAPA coatings were modified with 36 peptides which had previously reported roles in cell adhesion and then screened for adhesion of hPSCs. HPSCs were observed to bind more effectively to cRGDfK-modified surfaces than surfaces modified with any of 36 other peptides with previously reported roles in cell adhesion. A greater number of hPSCs were consistently observed to adhere to cRGDfK-PAAA and cRGDfK-PAPA coatings than to coatings modified with any other peptide, including those used in the commercially available hPSC culture surface Synthemax ™ . HPSCs did not attach to surfaces modified with HSPG-binding peptides, which may have been due to the absence of ROCK inhibition. The chemically defined and relatively affordable cRGDfK-PAPA coating has been demonstrated to maintain high quality hPSC cultures over long-term culture. Cultures of three hPSC lines that were maintained in cRGDfK-PAPA-coated flasks remained comparable to control cultures maintained in parallel in Geltrex ™ -coated flasks in terms of growth rate and maintenance of pluripotency as assessed by cell morphology, gene expression and teratoma formation. Cultures maintained in parallel on the commercially available chemically defined surfaces StemAdhere ™ and Synthemax ™ were also observed to remain pluripotent by these measures. Detailed analysis of proliferation rates was obscured by cell line variation, particularly in the Genea-02 cell line. Importantly, hPSC cultures maintained on cRGDfK-coated surfaces demonstrated less karyotypic abnormality than cultures on the other surfaces tested, although more detailed analysis would be required to confirm this observation. Trisomy of chromosome 12 arose frequently and only in hPSC cultures maintained on StemAdhere ™ -coated surfaces, indicating that this surface may predispose hPSC cultures to in vitro selection pressures and possible genetic instability. Cultures maintained on Synthemax ™ in this study would not have been suitable for large scale production due to the need for physical harvesting of cells by cell scraping, although this may be ameliorated in future by novel dissociation reagents.
The results of this study collectively indicate that cRGDfK-PAPA has the potential to provide a viable, commercially scalable surface for the support of human pluripotent stem cells.

Experimental Section
Cell culture. hPSCs. The hESC-WA09 (H9) 49 , hESC-GENEA-02 50 , and hiPS-NHF1.3 51 cell lines were respectively provided under materials transfer agreements by the Wisconsin Alumni Research Foundation (WARF) and Genea, and kindly by Prof. Ernst Wolvetang (Australian Institute for Bioengineering & Nanotechnology). All hPSCs were maintained in Essential 8 ™ (E8) medium (Life Technologies) on Geltrex ™coated tissue culture polystyrene at 37 °C in an atmosphere containing 5% CO 2 and passaged using EDTA. Cells that did not lift from pipetting with medium after incubation in EDTA (0.5 mM) were manually removed from the surface using a cell scraper. HPSC cultures for these experiments were thawed from banks of cryopreserved H9, hiPS-NHF1.3 and Genea-02 hPSCs that had been adapted to culture in E8 medium on Geltrex ™ -coated surfaces for at least ten passages. All work using hPSCs was carried out in accordance with approvals from Monash University (Project ID 2963) and the CSIRO Human Research Ethics Offices. Generation of H9-OCT4 2AChryIM/w reporter cell line. An OCT4 2AChryIM/w targeting vector was constructed by amplifying 5.4 kb (left homology arm) and 3.5 kb (right homology arm) of genomic DNA 5′ and 3′ of the OCT4 stop codon respectively. These fragments were cloned sequentially into pCR ™ -Blunt II-TOPO ™ (Life Technologies) such that each arm was separated by a single AscI restriction site. Conventional cloning was then used to insert a cassette comprising sequences encoding a T2A peptide fused to mCherry followed by an internal ribosomal entry site located upstream of a synthetic neomycin resistance (NeoR) gene -a version of NeoR that had been optimised for expression in mammalian cells (designated Meo). The structure of the final vector is shown in Fig. 1 immediately 3′ of the OCT4 stop codon (Cellectis) were co-electroporated with linearised OCT4-mCherry vector into H9 hESCs using previously published protocols 52 . G418 resistant colonies were isolated as described previously 52 and screened for gene targeting events using a PCR based approach that utilised a forward primer directed against sequences within Meo in conjunction with a primer homologous to GAPDH genomic sequences downstream of the 3′ most sequences contained within the targeting vector. PCR of correctly targeted clones yielded a PCR fragment of 3.6 kb. Five of six mCherry pos colonies screened were positive for the diagnostic PCR product.
Synthesis of PAAA and PAPA coatings on TCPS. Acrylic acid was purified by short-path distillation to remove dimers and inhibitors, and then stored at −20 °C until use. Propargyl acrylamide was prepared according to a published method 53 . Monomer solutions and polymer coatings were prepared in an oxygen-free environment. To prepare PAAA coatings, co-monomer solutions containing 7.5% w/v acrylamide (60 mole %) and acrylic acid (40 mole %) were prepared in MQ H 2 O. To prepare PAPA coatings, co-monomer solutions containing 10% w/v acrylamide (70 mole %) and propargyl acrylamide (30 mole %) were prepared in MQ H 2 O. Tissue culture-treated polystyrene vessels (either multi-well or tissue culture flasks) were coated with 150 μl/cm 3 of the appropriate mixed monomer solutions and transferred to bags that were subsequently vacuum-sealed. Surface initiated polymerisation from the TCPS substrate materials, covered by monomer solutions and sealed inside a low-oxygen environment, was induced with repeated exposure to high power UV light generated by dual Light Hammer ® 6 UV lamps (Fusion UV) at set at 70% power. Following UV treatment, polymer-coated plates were Peptide screening approach. Unless stated otherwise, the following methods were used to prepare peptide-modified polymer coatings in 24-well plates for peptide screening.
Peptide modification of PAAA wells for peptide screening. The carboxylic acid functional groups present in PAAA coated plates, prepared using methods described above, were activated with an aqueous solution containing 125 mM 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC) and 125 mM N-hydroxysuccinimide NHS, incubated overnight in a 200 μM solution of test peptide in DPBS-and washed thoroughly in DPBS-.
Peptide modification of PAPA wells for peptide screening. A surface-based CuAAC peptide conjugation method was developed based on previous work 54 . All reactions were performed at room temperature in aqueous solutions prepared with MQ H 2 O unless otherwise stated. CuAAC reaction solutions were prepared under atmospheric conditions and contained 100 μM copper sulphate, 272 μM sodium ascorbate and 200 μM azide-conjugated peptide in 1 M HEPES buffer unless otherwise stated. Polymer-coated surfaces were exposed to 400 µL/well of CuAAC reaction solution. Plates were sealed in airtight plastic pockets and incubated at room temperature for 24 hours while shaking at 100 rpm. Peptide-coupled wells were washed thoroughly in DPBS-, treated with 0.1 M Na 2 EDTA overnight and again thoroughly washed in DPBS-. Peptide-modified surfaces were typically stored for no more than three months at 4 °C under DPBS-containing 50 U/mL penicillin/streptomycin and wrapped in parafilm.
Fmoc-deprotection: a mixture of piperidine (Sigma Aldrich) and DMF (20:80, v/v) was added to the resin. The resin was shaken for 5 minutes and filtered and the fresh portion of the mixture of piperidine and DMF (20:80, v/v) was added to the resin. The resin was shaken for 15 minutes, filtered and washed with DMF (3 × 5 minutes), Dichloromethane (DCM; Sigma Aldrich, 3 × 5 minutes) and DMF (3 × 5 minutes).
Cleavage protocol: Peptide resin was treated with 5 mL of cleavage mixture: Trifluoroacetic acid (TFA; Sigma Aldrich)/ Triisopropylsilane (TIS; Sigma Aldrich)/H 2 O (95/2.5/2.5, v/v/v); 90 minutes, RT. The cleavage solution was separated from the resin by filtration and peptides were precipitated by addition of chilled diethyl ether (Sigma Aldrich), centrifuged, and decanted. The product peptides were washed twice with the same solvent. Samples were then left under vacuum for 30 minutes. Peptides were dissolved in MQ H 2 O and lyophilised.
Analysis and Purification: Analytical reversed-phase high performance liquid chromatography (HPLC) was performed as described above. Preparative HPLC was performed on C18 column (21.2 × 250 mm, 10 μm, Phenomenex, Torrance, CA) in a model Agilent 1260 Infinity system (Agilent Technologies, Santa Clara, CA). Solvents A and B were 0.1% TFA (v/v) in MQ H 2 O and acetonitrile, respectively, and elution was with 5-70% linear gradients of solvent B into A over 25 minutes, at 15 mL/minute flow rate, with UV detection at 210 nm. Preparative fractions of satisfactory purity ( ≥ 95%) by analytical HPLC were pooled and lyophilised. All peptides were characterised for identity by HPLC analysis and mass spectrometry.
Polymer-coated 24-well plates were modified with test peptides, the lead cRGDfK peptide or (for PAAA plates only, the non-binding control cRADfK peptide. H9-OCT4 2AChryIM/w cells harvested from Geltrex ™ -coated maintenance flasks using EDTA were seeded in all polymer-coated wells at a density of 15 000 cells/cm 2 in 400 µl of E8 media. Geltrex ™ -coated control wells seeded at equal and one-third density were included on a separate 24-well plate. Forty-eight hours after cell seeding, cultures were visually assessed under phase contrast microscopy and adherent colonies were counted. Daily media changes were gently performed from day 2 and cultures were observed daily. Colony growth was roughly assessed four days after cell seeding by re-scanning wells. Samples and controls were included in triplicate wells in randomised locations on each plate. Experiments were repeated three times using polymer coatings synthesised and modified independently and seeded with hPSCs harvested from different cultures. Peptides were considered to have failed the screen if no hPSCs were observed to bind to any of the wells modified with a peptide in either of the first two experiments. Colony counts from surfaces modified with each peptide were compared to counts from the inbuilt cRGDfK-modified controls.
Preparation of hPSC culture surfaces for long term maintenance. A surface-based CuAAC peptide conjugation method was developed based on previous work 54  Teratoma assay. Teratomas were generated as previously described 56 . For each experiment, a cell/Matrigel ™ suspension was injected into both testes of three mice. Tumours were histologically scored for the presence of tissue types representing endoderm, mesoderm and ectoderm. Test populations were deemed pluripotent if the three germ layers were collectively represented in any tumour or tumours derived from that population. Protocols and use of animals in this project were undertaken with approval of the Monash University Animal Welfare Committee following the 2004 Australian Code of Practice for the Care and Use of Animals for Scientific Purposes and the Victorian Prevention of Cruelty to Animals Act and Regulations legislation.