Mechanics-driven nuclear localization of YAP can be reversed by N-cadherin ligation in mesenchymal stem cells

Physiological responses to cell-cell interactions can be mimicked by HAVDI immobilized on a hydrogel

To elucidate how N-cadherin affects YAP localization, we developed a dual-peptide functionalized PEG hydrogel analogous to the hyaluronic acid (HA) based hydrogel of Cosgrove et al.³⁹. The PEG hydrogel system presented either HAVDI/RGD (from N-cadherin, representing neighboring cells), or non-functional scrambled HAVDI sequence (Scram) combined with RGD (Scram/RGD) (Fig. 1). 8-arm PEG maleimide (PEG-MAL) modified with mono-thiolated peptides was subsequently crosslinked by 8-arm PEG thiol (PEG-SH) via Michael addition) (Fig. 1a)⁴¹. As with analogous HA hydrogels, PEG hydrogels modified with 1 mM RGD and 1 mM non-functional scrambled HAVDI sequence (Scram/RGD) established only integrin adhesions, while PEG hydrogels modified with 1 mM RGD and 1 mM HAVDI enabled both integrin and N-cadherin mediated interactions (HAVDI/RGD) (Fig. 1b and Supplementary Fig. 1). Concentrations of HAVDI peptides followed Cosgrove et al.³⁹, and were within the range of E-cadherin density at adherens junctions in monolayer epithelial cells of developing Drosophila embryos⁴². The grafted peptide was observed to distribute homogeneously on the surface of hydrogels (Supplementary Fig. 1). The binding efficiency of peptide to the PEG backbone was measured at over 86% (Supplementary Fig. 2). Through the binding efficiency, we calculated the surface concentrations of peptides, which fell into a reasonable range (Supplementary Text).

Hydrogel stiffness could be tuned independently from 2-41 kPa by varying relative concentrations of PEG-SH and PEG-MAL (Fig. 1c). Modification with HAVDI/RGD or Scram/RGD did not affect Young’s moduli of the hydrogel over the range of concentrations of PEG-SH and PEG-MAL used in this study (Fig. 1c). Young’s moduli were obtained from load-indentation curves (Supplementary Fig. 3). When we seeded cells on the hydrogels with different stiffness and peptide, cell area increased with increasing substrate stiffness, in accordance with previous studies^43,44; no significant difference was observed between the Scram/RGD and HAVDI/RGD groups (Supplementary Fig. 4). Results indicated that the HAVDI peptide did not affect cell spreading, and differences in YAP n/c ratio observed between Scram/RGD and HAVDI/RGD hydrogels did not arise from difference in cell area.

We began by using this system to establish whether HAVDI presentation faithfully replicates the N-cadherin clustering and β-catenin recruitment associated with N-cadherin-mediated cell-cell interactions. These two mechanosensing proteins accumulate as force transmission between cells matures adherens junctions via an α-catenin/vinculin dependent pathway⁴⁵. To study this, we cultured hMSCs at prescribed cell densities and levels of confluence on Scram/RGD or HAVDI/RGD hydrogels with a stiffness of 20 kPa for 3 d, then fixed and stained the hMSCs for N-cadherin and β-catenin on their basal planes (Fig. 2a).

Sparse cells that had no contact with their neighbors showed neither N-cadherin clustering nor β-catenin recruitment when cultured on Scram/RGD hydrogels (Fig. 2b). However, N-cadherin clustering and β-catenin co-localization were rescued in sparse cells cultured upon HAVDI/RGD hydrogels (Fig. 2c). Likewise, confluent cells, which formed adherens junctions with their neighbors, showed clustering of N-cadherin as well as recruitment and co-localization of β-catenin on both the HAVDI/RGD and Scram/RGD hydrogels (Fig. 2d). To check whether N-cadherin co-localized with β-catenin, the percentage of N-cadherin area that was also positive for β-catenin was quantified from fluorescence images using a custom MATLAB (R2014a) script that implemented standard cross-correlation techniques from the literature^39,46 (Supplementary Fig. 5). For both sparse and confluent cells, significantly more co-localization (i.e., N-cadherin area positive for β-catenin) was observed for cells cultured on HAVDI/RGD hydrogels than for cells cultured on Scram/RGD hydrogels, indicating that substrates with immobilized HAVDI induced both N-cadherin clustering and β-catenin recruitment.

Treatment with 1 mM soluble HAVDI(-C) peptide, designed to bind with membrane N-cadherin receptors while avoiding conjugation to the PEG-Mal backbone, blocked recruitment and co-localization of N-cadherin and β-catenin in cells cultured on both Scram/RGD and HAVDI/RGD hydrogels (Supplementary Fig. 6). These results suggested that HAVDI/N-cadherin ligation arising in cells cultured on HAVDI/RGD hydrogels served to generate force from substrates triggering recruitment and co-localization β-catenin forming force bearing proteins chain.

HAVDI ligation reverses YAP nuclear localization in hMSCs

HAVDI presentation is known to affect nuclear localization of YAP³⁹. Immunostaining that replicated this result in our PEG system showed nucleus-to-cytoplasm (n/c) ratios of YAP significantly attenuated by HAVDI presentation over an intermediate range of substrate moduli (10~20 kPa), but unaltered for compliant (~1 kPa) or stiff (~40 kPa) substrates (Fig. 2e–f). Consistent with the observations of others^{15,39,47,48,49,50}, our myosin inactivation experiments showed YAP signaling to be highly dependent on the contractile state of the cell (Fig. 2f). After myosin inactivation using blebbistatin, we found that all YAP n/c ratios dropped to the same level for both the Scram/RGD and HAVDI/RGD hydrogels, independent of substrate moduli (2-41 kPa). These results suggested that actomyosin inhibition shielded the mechanosensing of substrates stiffness by cells^48,49,50. In addition, differences of YAP n/c ratio on substrates with different peptides could be attributed to different actomyosin contractility, because myosin inactivation could eliminate the observed HAVDI effect.

To quantify how peptide concentrations affect nuclear localization of YAP, we varied the HAVDI/Scram/RGD peptide concentration ratios on hydrogels while keeping the total effective ligand density constant. Increasing RGD presentation resulted in un-changed YAP nuclear-to-cytoplasmic (n/c) ratios on Scram/RGD hydrogels, indicating that 0.25 mM RGD was sufficient to induce a maximal YAP n/c ratio (Supplementary Fig. 7). Cells on HAVDI/RGD hydrogels showed lower YAP n/c ratios, with the differential becoming more pronounced with increasing HAVDI concentration. These results suggested that the increasing RGD concentration (≥0.25 mM) did not affect YAP nuclear translocation measurably, while increasing HAVDI concentration inhibited it.

We therefore chose two hydrogels with a fixed, intermediate stiffness (Young’s modulus = 20 kPa) and two different peptide combinations (Scram/RGD for 1 mM Scram and 1 mM RGD peptides, HAVDI/RGD for 1 mM HAVDI and 1 mM RGD peptides) to test the hypothesis that N-cadherin ligation can reverse nucleus-to-cytoplasm shuttling of YAP.

Before studying this reversal of nucleus-to-cytoplasm shuttling of YAP, we repeated earlier results showing that sparse hMSCs developed significantly higher YAP nuclear localization when cultured on Scram/RGD hydrogels than when cultured on HAVDI/RGD hydrogels (Supplementary Fig. 8a, upper-left; Supplementary Fig. 8b, left); blocking N-cadherin with soluble HAVDI(-C) peptide (1 mM) restored the YAP n/c ratio in sparse hMSCs cultured on HAVDI/RGD substrates to the level seen in hMSCs cultured on Scram/RGD substrates (Supplementary Fig. 8a, lower-left; Supplementary Fig. 8b, left), consistent with the observations of Cosgrove et al.³⁹. YAP was excluded from the nucleus in hMSCs in confluent cells cultured on either Scram/RGD or HAVDI/RGD substrates (Supplementary Fig. 8a, upper-right; Supplementary Fig. 8b, right), consistent with previous studies^15,18,25 and suggesting that cell-cell interactions inhibited nuclear localization of YAP. Blocking N-cadherin with soluble HAVDI(-C) peptide abrogated this inhibition in confluent cells as well (Supplementary Fig. 8a, lower-right; Supplementary Fig. 8b, right), indicating that N-cadherin mechanoresponsiveness controlled YAP exclusion associated with both real and mimicked cell-cell interactions.

We next tested the hypothesis that YAP nuclear localization, known to arise from priming of cells on stiff tissue culture plastic (TCP, ~3 GPa), could be reversed by N-cadherin signaling from HAVDI/RGD substrates. For this, hMSCs were cultured on TCP for 1-10 d, and then transferred to soft (20 kPa) HAVDI/RGD or Scram/RGD hydrogels for 3 or 10 d (Fig. 3a). As seen by others¹¹, the YAP n/c ratio in hMSCs increased with increasing culture time on TCP (Fig. 3b, “no transfer”). When hMSCs were transferred to Scram/RGD or HAVDI/RGD hydrogels after 1 d on TCP substrate (DT1), the YAP n/c ratio returned to the levels seen in cells that had never been exposed to TCP (“soft control” group), which suggested that YAP nuclear localization was fully reversible over 1 d (DT1 + So1) (Fig. 3c). Longer culture time on soft gels (DT1 + So3) did not change the YAP n/c ratio, suggesting that the kinetics of YAP nuclear localization reached a steady state within 1 d.

This observation was in accordance with previous studies that after substrate softening, the YAP n/c ratio reduced to baseline levels within 1 d and maintained these levels for up to 10 d²¹. Cui et al. found that nuclear accumulation of YAP reached a steady state within 6 h⁵¹, and Elosegui-Artola et al. found that YAP n/c ratio changed within several minutes after nuclear force is applied³¹. To exclude the possible effects of mechanical stimuli generated by cell passaging, we conducted experiments where cells were transferred from TCP to TCP. The YAP n/c ratio showed no difference between groups that had undergone passaging (DT3 + DT0.25 and DT3 + DT3 in Supplementary Fig. 9) and those that had not (DT3.25 and DT6 in Supplementary Fig. 9), indicating that mechanical stimuli associated with transferring substrates had no measurable effect on the MSC mechanosensing or memory.

In hMSCs cultured for 3 d on TCP (DT3) and subsequently transferred to soft HAVDI/RGD hydrogels for 1 d (DT3 + So1), the YAP n/c ratio recovered to the level of “soft control” group. Further increasing the culture time on soft HAVDI/RGD hydrogels to 3 d did not change the YAP n/c ratio. This suggested that YAP nuclear localization was again fully reversed within 1 d (Fig. 3d). In contrast, transferring to Scram/RGD hydrogels led to a higher YAP n/c ratio than that observed in the “soft control” group, suggesting that YAP nuclear localization was only partially reversible under these conditions (Fig. 3d). After 7 d on TCP (DT7), YAP nuclear localization was partially reversible in cells transferred to either Scram/RGD or HAVDI/RGD hydrogels; the YAP n/c ratio of cells transferred to HAVDI/RGD gels dropped more than those of cells transferred to Scram/RGD gels (Supplementary Fig. 10).

In hMSCs cultured for 10 d on TCP (DT10) and then transferred to Scram/RGD hydrogels, the YAP n/c ratio remained the much higher levels in cells in the “soft control” group with 1-10 d culture time. However, transferring to HAVDI/RGD hydrogels led to a drop of the YAP n/c ratio within 1 d to a level that remained constant for up to 10 d culture time. This confirmed that 10 d of culture on TCP made YAP nuclear localization irreversible in cells transferred to soft Scram/RGD hydrogels, as observed in previous studies^11,21, and revealed that YAP nuclear localization was again partially reversible in hMSCs transferred to HAVDI/RGD hydrogels (Fig. 3e). Increasing HAVDI concentration increased the magnitude of the drop in YAP n/c ratio (Supplementary Fig. 11).

Thus, hMSCs integrated and stored nuclear YAP when exposed to stiff TCP. This was increasingly persistent in response to increasing mechanical dosing on TCP, but could be partially or entirely reversed via HAVDI ligation that mimicked N-cadherin signaling in cell-cell interactions.

HAVDI ligation can be used to control stem cell fate

To demonstrate therapeutically useful reversal of YAP nuclear localization in hMSCs, we explored whether the seemingly irreversible, osteogenic commitment of hMSCs¹¹ could be reversed by HAVDI ligation. Yang, et al., demonstrated that priming of hMSCs for 10 d on stiff substrates resulted in nuclear localization of the osteogenic marker RUNX2 even after in-situ softening of the substratum for 10 d¹¹. We found similar results for cells cultured in growth medium for 1-7 d on TCP (~3 GPa) then transferred to soft (20 kPa) Scram/RGD hydrogels for 3 d in osteogenic medium (Fig. 4a), with nuclear localization of RUNX2 increasing with increasing culture time on TCP (Fig. 4b, c). However, in cells that were instead transferred from TCP to soft (20 kPa) HAVDI/RGD hydrogels, a negligible increase of RUNX2 nuclear localization was observed following stiff priming of up to 3 d, and elevation of RUNX2 levels was greatly attenuated for stiff priming of 7 d.

To verify that the observed RUNX2 nuclear localization was indicative of osteogenesis, the late osteogenic markers alkaline phosphatase (ALP) and osteocalcin (OCN)^52,53 were tracked in hMSCs that were exposed to TCP for 1-7 d in growth medium, and that were then transferred onto soft (20 kPa) Scram/RGD or HAVDI/RGD hydrogels for 7 d in osteogenic differentiation medium before being fixed and stained (Fig. 4a). Consistent with the results of Yang, et al.¹¹, ALP and OCN expression increased with increasing culture time on TCP for hMSCs transferred to Scram/RGD hydrogels (Fig. 4d–g). However, HAVDI ligation significantly attenuated both ALP and OCN levels, verifying that HAVDI ligation diminished osteogenic commitment. These findings demonstrated HAVDI ligation as a tool for attenuating nuclear localization of YAP and sustaining stem cell plasticity during in vitro expansion.

HAVDI dependent motor-clutch dynamics affect the mechanical sensation and nuclear translocation of YAP

Growing evidence implicates a motor-clutch machine of myosin, actin filaments, and focal adhesion complexes^54,55 in sensation of extracellular matrix (ECM) stiffness^56,57,58. The forces associated with actomyosin contraction increase with substrate stiffness because of increased resistance to contraction, increased engagement of the motor-clutch mechanism, and increased development of polarized stress fibers. The forces associated with increased traction on the substrate also deform the nucleus and can initiate nuclear-cytoplasmic shuttling of YAP and downstream signaling pathways^31,59. We hypothesized that N-cadherin affected mechanical sensation and the YAP n/c ratio (({R}_{{{{{{rm{NC}}}}}}})) by inhibiting the rate ({k}_{{{{{{rm{on}}}}}}}) of integrin binding, and hence reducing integrin clustering, tractions, and nuclear deformation. To test this hypothesis about the effects of HAVDI ligation, we applied a motor-clutch model (model details in Supplementary Methods and Supplementary Table 1)^56,57,58,60.

Briefly, in the motor-clutch model (Fig. 5a), actin filaments attached to ECM through dynamic bonds at integrin clutches with binding rate (({k}_{{{{{{rm{on}}}}}}})) and unbinding rate (({k}_{{{{{{rm{off}}}}}}})), while myosin motors contracted the actin filaments with traction force (({F}_{{{{{{rm{trac}}}}}}})) at velocity (({v}_{{{{{{rm{f}}}}}}})). The actin filament deformed the nucleus and substrate simultaneously. The nucleus was regarded as viscoelastic, with elastic deformation that recovered transiently with reduced traction force and inelastic deformation that accumulated loading (culture) time on TCP. As expected, cell traction forces increased with substrate stiffness (({E}_{{{{{{rm{sub}}}}}}})) due to the tension-mediated recruitment of integrin, which resulted in a higher ({k}_{{{{{{rm{on}}}}}}}) on stiffer substrates⁵⁷. The range of ({k}_{{{{{{rm{on}}}}}}}) depended upon the integrin density on the membrane⁶⁰. N-cadherin could downregulate traction force and focal adhesion length over an intermediate range of ({E}_{{{{{{rm{sub}}}}}}}) (5–20 kPa)³⁹, due to lower integrin density and lower ({k}_{{{{{{rm{on}}}}}}})⁶⁰. In our model, an Arrhenius relationship was used to relate ({k}_{{{{{{rm{on}}}}}}}) and ({E}_{{{{{{rm{sub}}}}}}}) to the presentation of HAVDI (details in Supplementary Methods) (Fig. 5b). The different ({k}_{{{{{{rm{on}}}}}}}-{E}_{{{{{{rm{sub}}}}}}}) relationships on Scram/RGD and HAVDI/RGD hydrogels resulted in double sigmoidal relationships between ({F}_{{{{{{rm{trac}}}}}}}) and ({E}_{{{{{{rm{sub}}}}}}},) (Fig. 5c). The stiffness threshold for elevating traction force was thus higher on HAVDI/RGD substrates than on Scram/RGD substrates.

Stress fibers alignments are well known to respond dynamically to the mechanics of the cellular microenvironment^61,62,63. To validate model predictions, we first measured the organization of F-actin (quantified by F-actin anisotropy). In accordance with prior observations³⁹, F-actin anisotropy increased with substrate stiffness, and decreased significantly in presence of HAVDI on substrates of intermediate stiffness (20 kPa) (Supplementary Fig. 12). We then performed traction force microscopy for cells cultured on substrates of prescribed peptide mixtures and stiffness. For cells seeded on Scram/RGD hydrogels, the traction increased with increasing hydrogel stiffness (Supplementary Fig. 13). For cells seeded on HAVDI/RGD hydrogels, consistent with observations of Cosgrove et al.³⁹ and our models, tractions were reduced by HAVDI presentation on substrates with an intermediate range of substrate moduli (10–20 kPa), but unaltered for compliant (1 kPa) or stiff (41 kPa) substrates (Fig. 5c). These results verified the prediction that MSC mechanosensing, and the peptide dependence thereof, required actomyosin contractility.

The model predicted that ({R}_{{{{{{rm{NC}}}}}}}) increased with increasing nuclear deformation, due to previously reported increases in the rate of translocation of cytoplasmic YAP to the nucleus³¹. In our model, HAVDI/RGD and Scram/RGD substrates yielded different tractions, nuclear deformation, pore sizes and hence ({R}_{{{{{{rm{NC}}}}}}}) (Fig. 5d). We hypothesized that actomyosin contractility affects nuclear deformation, which in turn affects the YAP n/c ratio. We thus characterized the nuclear flattening (({lambda }_{{{{{{rm{N}}}}}}}), nuclear length/height), a measure of nuclear deformation evidenced from our confocal images, and confirmed that it increased monotonically with increasing ({F}_{{{{{{rm{trac}}}}}}}) (Fig. 5e, Supplementary Fig. 14). As predicted, nuclear flattening was proportional to ({R}_{{{{{{rm{NC}}}}}}}) (Fig. 5f), an effect attributed to the enlargement of nuclear pores in deformed nuclei that promoted YAP import rate without changing the export rate³¹. Thus, ({R}_{{{{{{rm{NC}}}}}}}) served as a proxy for nuclear deformation in our model.

Our predictions of ({R}_{{{{{{rm{NC}}}}}}}) matched trends observed in our experiments, and were consistent with the model that HAVDI and stiffness co-determined ({k}_{{{{{{rm{on}}}}}}}), and that this effect gave rise to the observed double sigmoidal relationships between ({R}_{{{{{{rm{NC}}}}}}}) and ({E}_{{{{{{rm{sub}}}}}}}) on substrates with and without HAVDI, showing attenuated YAP nuclear localization (Fig. 5g). After this verification of the reasonability of our model for ({k}_{{{{{{rm{on}}}}}}}), the model was used to reproduce the phenomena observed in Fig. 3. Consistent with other observations of nuclear deformation, the nucleus was treated as viscoelastic, with deformation having both elastic (reversible) and plastic (irreversible) components^64,65,66. Our model predicted the reductions in ({R}_{{{{{{rm{NC}}}}}}}) over time following transferal of MSCs from TCP to soft hydrogels, and attributed these trends to reductions in tractions associated with release of elastic deformation of the nucleus. The plastic deformation of the nucleus captured the observed increase in ({R}_{{{{{{rm{NC}}}}}}}) with increasing culture time on TCP (Fig. 5h). In cells transferred from TCP to soft gels, the elastic deformation recovered while the plastic deformation remained, reducing ({R}_{{{{{{rm{NC}}}}}}}) over time. ({R}_{{{{{{rm{NC}}}}}}}) in cells transferred to HAVDI/RGD substrates was lower than in those transferred to Scram/RGD substrates because HAVDI reduced traction and thus facilitated elastic recovery. Results suggested that effects of HAVDI on mechanical sensation and nuclear translocation of YAP observed in our culture system could be explained through motor-clutch dynamics.

MSCs in vivo receive mechanical cues from both cell-ECM and cell-cell adhesions, with the former promoting mechanosensing and the latter inhibiting it. We showed that these two antagonistic factors also affect nuclear translocation of YAP, and found that N-cadherin signaling can reverse these effects arising from cell-ECM adhesions. This is important for regenerative therapies¹⁰ involving in vitro expansion of MSCs on TCP and subsequent in vivo transplantation, which must overcome the effects of TCP exposure that bias MSCs toward osteogenic differentiation. Using PEG hydrogels with co-presentation of RGD and HAVDI peptides that we showed to mimic cell-ECM and cell-cell adhesions, we demonstrated that HAVDI signaling can be used to reverse a nuclear translocation of YAP, as well as sustained mechanosensing effects in the form of RUNX2, ALP and OCN activity by altering the motor-clutch dynamics in MSCs. Taken together, results show that N-cadherin signaling arising from cell-cell adhesion reduces long-term retention of the effects of mechanotransduction in MSCs and can be used to engineer cell culture platforms to modulate “mechanical memory” in hMSCs.