StemBond hydrogels control the mechanical microenvironment for pluripotent stem cells

StemBond hydrogels have controlled stiffness and ECM tethering strength

In order to establish an optimal mechanical microenvironment for pluripotent stem cells, and to enable their stable substrate attachment, we adapted a specific PAAm hydrogel protocol similar to what was suggested in refs. ^28,29. We started from a precursor solution for a standard PAAm hydrogel below 1 kPa, approaching the low stiffness of the pre-implantation embryo from which mESCs are sourced (varying between 200 and 600 Pa³⁰). We added to the precursor solution the co-factor 6-Acrylamidohexanoic acid (AHA), which can bind to acrylamide chains without crosslinking two acrylamide chains together. The terminal carboxyl groups of the AHA serve as anchorage points for covalent ECM protein binding by first forming an amine-reactive ester through a carbodiimide reaction. As with the widely used sulfo-SANPAH, this ester can react with any primary amine of, for example, lysine chains, to tether ECM proteins to the surface (Fig. 1a). However, this method allows comparatively better control over the binding and surface density of the carboxyl groups because the AHA chains polymerise with acrylamide. We thus varied AHA concentration to obtain different levels of ECM tethering (Fig. 1b), estimating that the surface density of ECM anchorage points approximately triples (see the “Methods” section) when AHA concentration is increased from 16 mM (low AHA) to 48 mM (mid AHA) and 80 mM (high AHA) (Fig. 1b). We adapted the ratio of acrylamide to bis-acrylamide to ensure a constant stiffness over the range of AHA concentrations (Fig. 1c).

It is not immediately clear how control over ECM tethering might translate to how cells sense the ECM. One possibility is that ECM protein coverage varies with increasing density of tethering points. However, this is unlikely because we are working with saturating ECM concentrations. As expected, immunofluorescence confirmed that protein coverage levels were similar on soft and stiff substrates over a range of AHA concentration (Fig. 1d and Supplementary Fig. 1a). The other potentially important parameter is how strongly ECM proteins are tethered to the substrate, which varies not only with substrate type but also potentially with density of tethering points. In order to measure ECM tethering strength, we used AFM cantilevers coated with an anti-fibronectin antibody. On both soft and stiff StemBond hydrogels, we detected significant binding events (Supplementary Fig. 1b–e). We did not, however, measure significant binding events on sulfo-SANPAH functionalised stiff hydrogels, primarily because the rupture length was very short on these gels (Supplementary Fig. 1e). The short rupture length on stiff sulfo-SANPAH substrates, which have relatively high ECM tethering density, likely indicates a weak tethering of ECM to the substrate. On soft sulfo-SANPAH substrates, the significant binding events had longer rupture lengths but significantly smaller rupture forces than on soft StemBond substrates (Fig. 1e and Supplementary Fig. 1b–e), also indicative of weaker ECM tethering to the substrate. Plateauing rupture forces on high AHA hydrogels suggest saturation of ECM tethering points. Notably, the rupture forces are very similar for both soft and stiff StemBond hydrogels (Supplementary Fig. 1b and e). Our analysis strongly suggests that the increasing concentration of AHA does not significantly affect ECM protein coverage, but, in line with previous studies²⁹, that it has a potent effect on ECM tethering strength, and thus potentially on ECM stability.

To determine the stability of the ECM layer on cell-seeded hydrogels, we plated mouse embryonic fibroblasts (MEFs), which are a strongly adherent cell type that secrete and remodel ECM, on both StemBond and sulfo-SANPAH hydrogels coated with the same concentration of ECM. We observed that, on StemBond, the ECM layer remained on the surface even after several days; however, significantly, the ECM layer on sulfo-SANPAH treated hydrogels was depleted. This observation strongly suggests that the ECM is more stable on StemBond hydrogels (Fig. 1f), and this stability correlates with the tethering strength. Thus, controlling co-factor concentration on StemBond substrates allowed stable, reproducible, and tuneable ECM tethering, on both soft and stiff hydrogels.

StemBond hydrogels promote strong pluripotent stem cell attachment

We then tested whether strong ECM tethering would improve stem cell attachment. We first seeded mESCs on standard and StemBond PAAm hydrogels. We compared hydrogels of different pore size (varying acrylamide:bis-acrylamide ratio A:B) and sulfo-SANPAH concentrations. We found that prolonged cell attachment on fibronectin-coated StemBond hydrogels was significantly higher than on standard PAAm hydrogels functionalised with sulfo-SANPAH (Supplementary Fig. 2a, b). Notably, after a few days of culture on sulfo-SANPAH functionalised hydrogels, independent of concentration, there were clear floating or barely attached colonies, and these hydrogels ultimately yielded as few attached colonies as the unfunctionalised hydrogels. In contrast, on StemBond hydrogels we observed many large colonies irrespective of A:B ratio (Supplementary Fig. 2a, b).

We further tested whether different levels of AHA would impact cell attachment over the course of 48 h. We found that cell attachment was long-lasting on mid and high AHA hydrogels, irrespective of stiffness, and weaker on the low AHA hydrogels only (Fig. 2a), which had similar attachment numbers to standard PAAm hydrogels. Immunostaining confirmed that markers of focal adhesions, integrin β1 and phospho-paxillin, were present on soft and stiff hydrogels (Supplementary Fig. 2c). Actin stress fibres were also observed on tissue culture plastic (TCP) and stiff hydrogels, but actin was mostly cortical on soft substrates (Supplementary Fig. 2c). Furthermore, cell attachment and viability were comparable to TCP (Supplementary Fig. 2d, e). Proliferation rates assessed by EdU incorporation were approximately the same for all substrates (Supplementary Fig. 2f). Thus, StemBond hydrogels, in contrast to sulfo-SANPAH functionalised hydrogels, provide robust conditions for prolonged culture of mESCs, even on soft substrates.

Morphologically, colonies on soft StemBond hydrogels were rounder, resembling naïve mESCs in 2i culture; on stiff StemBond hydrogels, colonies were flat and spread out, similar to mESCs on TCP (Fig. 2b). ECM tethering density did not dramatically impact morphology, though we observed that colonies were systematically slightly rounder on, particularly, the low and mid AHA soft substrates (Fig. 2b). Similar morphological differences between soft and stiff substrates were found on gelatin, laminin and collagen-coated StemBond hydrogels, although mESCs did not flatten as much on collagen (Supplementary Fig. 3a). We also compared StemBond hydrogels to other types of soft substrates. On soft PDMS substrates, cells spread out to a level similar TCP (Supplementary Fig. 3b). StemBond hydrogels are therefore unique in providing a substrate viable for mESC attachment, and recapitulating naïve state morphology.

To show that StemBond hydrogels are compatible with multiple pluripotent stem cell types, we also established hPSC culture on StemBond hydrogels. We here compared laminin 521 (Lam 521) and vitronectin (Vtn) coating, both routinely used for hPSC culture, on either high AHA StemBond or sulfo-SANPAH functionalised hydrogels. Similar to our examination of mESCs, we found that sulfo-SANPAH does not provide a stable culture environment for hPSCs and leads to rapid colony detachment (Fig. 2c). In contrast, StemBond hydrogels allowed robust colony attachment and growth. Interestingly, we observed that only some combinations of stiffness and ECM show optimal colony attachment. Laminin 521 is highly supportive of colony growth on soft substrates, with colonies appearing round on both soft and stiff StemBond hydrogels, while vitronectin supports colony growth only on stiff substrates (Fig. 2c, d). These results suggest that the interaction between ECM protein and stiffness regulate cell attachment and cell growth.

Soft substrates support naïve pluripotency

Having established that StemBond hydrogels provide robust cell attachment, we next asked if they also support maintenance of pluripotency in mESCs and hPSCs. To do this, we first examined several canonical naïve transcription factors of mESCs, finding that the expression of those factors is systematically higher on soft substrates compared to TCP and stiff substrates. We found, for example, a significant increase in the expression of Esrrb in comparison to stiff substrates (Fig. 3a). At the protein level, ESRRB was also significantly more expressed on soft substrates compared to stiff hydrogels and TCP, though we did not find such a dramatic difference in NANOG (Fig. 3b). In addition, we probed the activity of STAT3 and ERK, which are known regulators of mESC cell state, on soft substrates compared to stiff substrates in serum + LIF conditions. STAT3 activity was higher but not significantly different, while ERK activity was significantly lower, suggesting that soft substrates stabilise naïve pluripotency by reducing ERK-driven differentiation and potentially by increasing self-renewal signals such as STAT3 (Fig. 3c). Finally, comparing gene expression on soft StemBond hydrogels to soft PDMS substrates on both fibronectin- and laminin-coated substrates, we showed that StemBond hydrogels lead to higher expression of naïve factors Esrrb, Tfcp2l1 and Klf4 and lower expression of formative marker Fgf5 than PDMS (Supplementary Fig. 3c). Together, these results confirm that StemBond hydrogels fully support the maintenance of naïve pluripotency in mESCs, with the expression of key pluripotency factors being further increased on soft substrates compared to both stiffer substrates, and other hydrogel substrates such as PDMS.

To determine more closely the population response on StemBond hydrogels, we used a destabilised GFP reporter line for Rex1 (Zfp42), a high-fidelity reporter for naïve pluripotency³¹, and measured fluorescence intensity over several passages. While a bimodal distribution typical of serum + LIF conditions was observed in all conditions, we found that soft substrates significantly increased the proportion of Rex1-positive cells without shifting the peak fluorescence of the naïve cells (Fig. 3d). Additionally, this proportion increased over time in culture on both soft and stiff hydrogels, but not on TCP (Fig. 3d). These data suggest that in serum + LIF conditions StemBond hydrogels—particularly the soft substrates—progressively improve the homogeneity of the mESC population.

In contrast to serum + LIF, the 2i conditions defined by the use of CHIRON (Gsk3β inhibitor) and PD03 (MEK inhibitor) create a homogeneous naïve population from which mESCs can be differentiated into multiple lineages. In 2i conditions, mESCs were homogeneously naïve regardless of substrate (Supplementary Fig. 4a) and mESCs exited from naïve pluripotency on all substrates (Supplementary Fig. 4a).

Given that exit from naïve pluripotency was unaffected by substrate type, we then asked if multilineage specification of mESCs was similarly unaffected. For this, we differentiated mESCs starting from 2i conditions into neuroectoderm and mesoendoderm³². When prompted to differentiate into neuroectoderm, cells upregulated Pax3, Pax6 and Sox1, markers of early neuroectoderm lineage commitment on both soft and stiff hydrogels (Supplementary Fig. 4b, c) within 4 days. When prompted to differentiate towards mesoendoderm, cells showed normal upregulation of lineage markers Eomes, Foxa2 and T/Brachyury within 4 days (Supplementary Fig. 4b, c). Overall, we conclude that, though naïve pluripotency is stabilised on StemBond hydrogels, mESCs are able to differentiate on these substrates, and that both exit from naïve pluripotency and lineage priming proceed normally regardless of substrate condition.

We next verified that StemBond hydrogels support pluripotency in hPSCs. General pluripotency markers (Pou5f1 and Nanog) were similar in hPSCs on TCP and StemBond hydrogels coated with either laminin and vitronectin, despite the weaker attachment on soft vitronectin substrates (Supplementary Fig. 3d). Furthermore, although the ECM protein was shown to make a difference on cell attachment and cell morphology (Fig. 2d), this did not affect pluripotency gene expression in hPSCs. Only expression of Nanog on vitronectin-coated TCP was lower than in other conditions. In naïve hPSCs, gene expression was overall comparable for the naïve markers Klf4, Klf17, and Tfcp2l1 on TCP and StemBond hydrogels for HNES1 and cR-H9 cells (Supplementary Fig. 3e). Thus, we conclude that StemBond hydrogels are compatible with culture and propagation of hPSCs without the requirements for feeder layers or Matrigel.

Pluripotent stem cell self-renewal in minimal media conditions

So far, we have established that StemBond hydrogels support mESC and hPSC self-renewal. In mESCs, the purity of the naïve population increased in serum + LIF conditions, although not to the level observed in 2i conditions. Nevertheless, the small molecule inhibitors and LIF are potent self-renewing factors that could mask any difference emerging from substrate conditions. In order to challenge the system, we then asked if the above-mentioned factors are necessary for mESC self-renewal on StemBond hydrogels by conducting clonogenicity assays in minimal media conditions. In these assays, we removed one or several self-renewing factor(s) for a defined period of time, then replated the cells at very low densities in 2i + LIF medium for 5 days to select for cells that retained their ability to self-renew in naïve conditions. We then counted the number of resulting naïve pluripotent colonies (Fig. 4a(i)). First, we removed LIF (a positive effector of self-renewal) from serum + LIF medium for 5 days on TCP, stiff and soft StemBond substrates. On TCP, removal of LIF led to drastic loss of self-renewal. On stiff substrates, there was a modest loss of naïve pluripotency upon removal of LIF. On soft substrates, however, significantly more cells maintained naïve pluripotency than on stiff substrates and TCP (Fig. 4b). Surprisingly, twice as many cells maintained naïve pluripotency on soft substrates without LIF when compared to the control of undifferentiated cells grown on TCP in serum + LIF. These results suggest that in serum only conditions, StemBond hydrogels at least partially compensate for LIF to sustain self-renewal of mESCs.

We then tested whether soft substrates could compensate for one of the two otherwise essential inhibitors, CHIRON or PD03, and would allow the propagation of mESCs for multiple passages in minimal media conditions. For this, we cultured cells on different substrates for five passages in N2B27 + CHIRON by serial dilution, before replating them at very low density in 2i + LIF to assess how many cells remained in a naïve pluripotent state (Fig. 4a(i)). We found that significantly more cells gave rise to naïve colonies from soft substrates than from stiff substrates (Supplementary Fig. 5a, b). Moreover, cell survival was also significantly higher on soft substrates (Supplementary Fig. 5c). Since the higher survival could be a confounding factor because it influences plating density in a serial dilution experiment, we repeated the experiment keeping constant cell plating density at each passage (Fig. 4a(iii)). After three passages of constant plating density, there was low cell survival on stiff substrates, but survival was still high on soft substrates. Furthermore, we found in these conditions that there were significantly more naïve pluripotent cells on soft substrates than on stiff substrates and TCP (Fig. 4c).

In N2B27 + PD03, cell survival was too low to perform the serial dilution clonogenicity assay. We therefore kept a constant cell plating density over multiple passages (Fig. 4a-iii). This allowed propagation for up to three passages on soft and stiff substrates only, with no cells surviving past passage 2 on TCP (Supplementary Fig. 5d), in line with previous studies¹⁷. Survival did gradually decrease on soft and stiff substrates, with no significant differences in cell numbers, and <10% of cells giving rise to naïve colonies on both substrates (Supplementary Fig. 5e, f). Because sensitivity to MEK/ERK inhibition (but not to GSK3β inhibition) has been shown to depend on cell line¹⁷, we repeated this assay with a different cell line, from a C57BL/6-Agouti background. In this case, we found a greater effect of the substrate stiffness on the number of naïve pluripotent colonies obtained with ~9% on soft substrates but only ~1% on stiff substrates and TCP (Fig. 4d). To ultimately confirm that the cells cultured in N2B27 + PD03 on soft hydrogels were still naïve pluripotent, we injected GFP-targeted mESCs into mouse blastocysts to test their ability to contribute to the germ layers and the germline. We obtained five viable chimaeras from the first injection and seven GFP-expressing pups from the next generation, demonstrating germline transmission (Fig. 4e and Supplementary Table 1). To our knowledge, this is the longest that mESCs have ever been successfully maintained in this highly challenging condition without genetic manipulation. We conclude that soft substrates more robustly provide a microenvironment supportive of self-renewal in N2B27 + PD03.

MEK inhibition is also used to maintain naïve hPSCs (HNES1)^33,34. We thus asked whether soft substrates would also support self-renewal of HNES1 cells in absence of MEK inhibition, as is the case for mESCs. We found that, after 3 days without MEK inhibition, ERK activity was higher on TCP compared to soft but also compared to stiff substrates (Supplementary Fig. 5g), suggesting that ECM–substrate attachment could impact ERK activation. We found a relative higher expression in naïve markers Klf4 and Klf17 after 4 days without MEK inhibition on soft substrates compared to TCP and stiff hydrogels (Supplementary Fig. 5h). We further cultured HNES1 in control conditions or without PD03 for 5–6 days before replating them at clonal density and staining for naïve colonies (as in Fig. 4a(i)). The number of growing replated colonies in control conditions did not vary significantly between substrates, but in absence of PD03, soft StemBond substrates yielded 2.5 times more colonies than stiff substrates and 3 times more than TCP (Fig. 4f), mirroring the above results with mESCs. Taken together, we conclude that both stiffness and substrate type impact pathway activity, ultimately resulting in increased self-renewal of mESCs and hPSCs on soft StemBond hydrogels cell in minimal media.

Soft StemBond hydrogels promote acquisition of naïve pluripotency

To illustrate how StemBond substrates could be used across different stem cell types, we turned to primed mouse epiblast stem cells (mEpiSC), which are known to require stronger cell adhesion than mESCs. mEpiSC can be reprogrammed to induced naïve pluripotent stem cells (iPSC) by the forced expression of a single naïve transcription factor together with signalling cues^35,36,37. We here hypothesised that soft StemBond hydrogels could improve the induction of naïve pluripotency. We employed doxycycline-inducible Esrrb (iEsrrb) as an efficient driver of reprogramming in EpiSCs^38,39,40 with a Rex1::dGFP-IRES-bsd reporter for the naïve identity³¹. mEpiSCs did not stably adhere to sulfo-SANPAH treated substrates and only weakly to mid AHA StemBond hydrogels, but they adhered well to high AHA StemBond hydrogels, which we thus used for reprogramming assays. We plated mEpiSCs on plastic, stiff and soft substrates for 24 h before inducing reprogramming with 2i+doxycycline and then assessed the efficiency of the process at different time points (Fig. 5a). After 24 h, cells were more clustered together on soft substrates than on TCP and stiff hydrogels (Fig. 5b). Both early markers of reprogramming, Tfcp2l1 & Klf2 and the late marker Zfp42 (Rex1) showed a twofold increase in gene expression on soft substrates. The general pluripotency marker Pou5f1 (Oct4), which transiently drops when reprogramming is triggered, was more effectively maintained on soft substrates (Fig. 5c). These results suggested a boost in the naïve pluripotency network during the reprogramming process. Correspondingly, we found a higher end-point efficiency of reprogramming on soft substrates, with twice as many naïve colonies on soft substrates at day 8 (Fig. 5d). Naïve genes were expressed at similar levels in resultant iPSCs demonstrating full reprogramming (Supplementary Fig. 6). Therefore, soft StemBond hydrogels not only support mESC self-renewal but also promote specification of naïve pluripotency during reprogramming.

Substrate stiffness induces genome-wide transcriptomic changes

To demonstrate potential applications of StemBond hydrogels, we used them to gain deeper insight into the potential synergies between adhesion, stiffness, and regulation of naïve pluripotency. We here examined the genome-wide transcriptional changes triggered by different substrate conditions in mESCs. We first determined the impact of substrate stiffness and ECM tethering on mESC gene expression by performing RNA sequencing in serum + LIF conditions (Supplementary Dataset 1). We found that substrate stiffness had a greater influence on gene expression than ECM tethering strength, with 452 differentially expressed genes due to changes in stiffness along with greater overall fold changes in gene expression, against 52 differentially expressed genes due to ECM tethering. The genes that were differentially expressed due for different ECM tethering strengths were primarily observed on soft substrates (Fig. 6a). Principal component analysis (PCA) confirmed that samples primarily cluster by substrate stiffness rather than by substrate adhesiveness (Fig. 6b). Among the genes differentially regulated by substrate stiffness are naïve pluripotency factors, as already observed by qPCR (Fig. 3a). In contrast, the same markers showed no significant differences in gene expression between our low AHA and high AHA substrates (Supplementary Fig. 7a), confirming that ECM tethering is not a significant factor in the regulation of naïve pluripotency. Gene ontology analysis of the differentially expressed genes revealed enrichment for stem cell population maintenance and transcription processes in genes upregulated on soft substrates, whereas cell adhesion and migration as well as some differentiation pathways were enriched in downregulated genes (Supplementary Fig. 7b). We conclude that while sufficient substrate adhesiveness is essential for the prolonged attachment of mESCs, substrate stiffness is the key mechanical property impacting the regulation of naïve pluripotency in mESCs.

To further assess if the observed stiffness sensitivity of naïve mESCs was dependent on LIF or ERK signalling, two signals known to be important for maintenance and differentiation of mESCs respectively, we again used RNA sequencing of mESCs in serum with or without LIF and/or without the MEK/ERK inhibitor, PD03, now focussing on a single adhesiveness condition (mid AHA). We found that, in all media conditions, substrate stiffness had a significant impact on gene expression, with over 210 genes being modulated independently of the presence of LIF or PD03 (in bold in Supplementary Fig. 7c). We first examined the expression of naïve and of recently defined formative genes^31,41, the latter being indicative of cells having exited naïve pluripotency. In detail, we found that the expression of most naïve pluripotency factors, but not of general pluripotency factors, was significantly increased on soft compared to stiff hydrogels and to TCP, whereas the expression of formative genes was significantly decreased on soft substrates (Fig. 6c, Supplementary Fig. 7d–f and Supplementary Dataset 2). Notably, these changes were more pronounced after 24 h in culture as opposed to 48 h (Supplementary Fig. 7d), suggesting that phenotypic differences begin to be masked at high cell density. The changes in expression levels were also observed across different media conditions (Supplementary Fig. 7d, e), and most pronounced in the absence of LIF and PD03. Consistent with these results, PCA revealed that samples separate based on stiffness, more so in absence of LIF (Supplementary Fig. 8a). Taken together, these results suggest that biological processes active in primed cells could be specifically stiffness-sensitive, and that this stiffness-sensitivity is mediated by interactions between substrate stiffness and the LIF/STAT3 and MEK/ERK pathways, which respectively drive naïve self-renewal¹⁸ and early differentiation^42,43.

Gene ontology analysis on the genes differentially regulated by substrate stiffness (in serum ± LIF ± PD03) showed that beyond naïve pluripotency and formative genes, functions related to the cytoskeleton and the extracellular matrix were also differentially enriched, as well as signalling pathways involved in self-renewal and pluripotency (Fig. 6d). The transcription of key mechanosensing structures, such as focal adhesions, cell–cell junctions, and the cytoskeleton, was also differentially regulated on soft and stiff substrates in all conditions (Supplementary Fig. 8b–d). When further scrutinising the enriched pathways, we found that, on average, 74% of differentially regulated components known to be involved in epithelialisation, cell adhesion and cytoskeleton reorganisation were downregulated on soft substrates in serum conditions (Supplementary Fig. 8e, Supplementary Dataset 2).

Since epithelialisation is essential to epiblast progression in vivo ⁴⁴, these results prompted us to ask if the observed stiffness-dependent gene expression might have significance in the context of embryonic development. To do this, we computed the log₂ fold change of the differentially expressed genes that were associated either to the pre- or post-implantation epiblast (previously characterised by sequencing early embryos⁴⁵) and found that among pre-implantation genes, 80% were upregulated on soft substrates, while among the post-implantation genes, 85% were downregulated on soft substrates (Fig. 6e, Supplementary Fig. 8f, and Supplementary Dataset 2). The trend was similar across all media conditions. Thus, mESCs cultured on soft and stiff substrates recapitulated key transcriptional differences between pre-implantation epiblast and post-implantation epiblast, respectively.

Substrate stiffness sensing therefore not only impacts the organisation of the cytoskeleton (as shown in Supplementary Fig. 2c) but also feeds back onto the transcriptional control of cytoskeleton-driven processes. Altogether, our RNA sequencing analysis indicates that biological processes occurring upon post-implantation, among which is epithelialisation, are sensitive to substrate stiffness. We speculate that soft substrates reduce the background occurrences of epithelialisation and cytoskeleton remodelling in serum-rich media, thereby promoting self-renewal in minimal conditions, as we observed above (Fig. 4). Stiff substrates on the other hand promote background differentiation in serum-rich media.

The substrate dependence of ERK activity, a driver of early differentiation in mESCs, was already indicated by protein assays after 24 h in serum + LIF (Fig. 3c). To probe this at the transcriptional level, we identified ERK targets by carrying out RNA-sequencing at early time-points (2 and 12 h) after removing PD03 from 2i + LIF media (Fig. 6f and Supplementary Dataset 3). Note that CHIRON + LIF media is sufficient to maintain naïve pluripotency, so this assay tests specifically for the effects of ERK signalling induction in mESCs independent of changes in pluripotent state. Known ERK targets such as Dual-Serine Phosphatases (Dusps) and Immediate-Early Genes (Fos, Egr1, Jun) were all activated after 2 h on soft and stiff substrates, confirming that PD03 withdrawal led to ERK activation in all conditions. We observed a systematic effect of substrate stiffness on the magnitude of ERK-target regulation (Fig. 6f). A comparison of the expression of all co-regulated genes (i.e. either activated or suppressed by ERK on both substrates) revealed that nearly 90% of ERK-activated genes were downregulated on soft substrates, whereas 85% of ERK-suppressed genes were upregulated on soft substrates compared to stiff ones (Supplementary Fig. 9b). This was already evident 2 h after removing PD03, and significantly more so at 12 h. The absolute fold change was significantly larger for ERK-regulated genes compared to non-target genes (Supplementary Fig. 9c). Additionally, we computed log₂ fold changes of ERK target genes in different media conditions and on different adhesiveness (data sets of Fig. 6a, b and c–e) and found similar trends, whereby ERK-suppressed genes were higher on soft substrates, but ERK-activated genes were lower. Other variables did not have a significant impact (Supplementary Fig. 9d).

In order to see if GSK3/β-catenin activity, which also regulates pluripotency, is similarly affected by substrate stiffness, we performed the same analysis removing CHIRON from 2i + LIF. Both canonical and non-canonical Wnt pathway genes were suppressed on soft and stiff substrates when CHIRON is removed. However, we observed no dependence on substrate-stiffness, even after 12 h (Supplementary Fig. 10 and Supplementary Dataset 4). This suggests that the GSK3/β-catenin pathway is not sensitive to substrate stiffness in mESCs. Taken together, our results suggest that substrate stiffness significantly impacts the activation and transcriptional activity of the ERK pathway, pointing to ERK mechanosensitivity as a primary mechanism for the optimised maintenance of self-renewal on soft substrates.