Summary
Our current understanding of cellular mechano-signalling is based on static models, which do not replicate the dynamics of living tissues. Here, we examined the time-dependent response of primary human mesenchymal stem cells (hMSCs) to cyclic tensile strain (CTS). At low-intensity strain (1 hour, 4% CTS at 1 Hz) morphological changes mimicked responses to increased substrate stiffness. As the strain regime was intensified (frequency increased to 5 Hz), we characterised rapid establishment of a broad, structured and reversible protein-level response, even as transcription was apparently downregulated. Protein abundance was quantified coincident with changes to protein conformation and post transcriptional modification. Furthermore, we characterised changes within the linker of nucleo- and cytoskeleton (LINC) complex of proteins that bridges the nuclear envelope, and specifically to levels of SUN domain-containing protein 2 (SUN2). The result of this regulation was to decouple mechano-transmission between the cytoskeleton and the nucleus, thus conferring protection to chromatin.
The stiffness of tissue correlates with its ability to resist mechanical damage, with the structure and integrity of the human body defined by stiff tissues such as skin, muscle, cartilage and bone. Tissue mechanical properties are determined by the extracellular matrix (ECM), in particular by the identities and concentrations of its constitutive proteins1-3. ECM properties are further modulated by protein cross-linking, post-translational modifications (PTMs) and higher-order organisation. Cells resident within tissues maintain mechanical equilibrium with their environments4, 5, and the mechanical properties of cells are also regulated by the identities, concentrations, conformations and PTMs of structural intracellular proteins1, 3, 6, 7. The characteristics of adherent cells can be influenced by physical stimulation from the surrounding ECM, with protein content1, morphology1, 8, motility9, 10 and differentiation potential11, 12 amongst behaviours known to be affected by stiffness. Cells in living tissues experience microenvironments of diverse stiffness5, but are also subject to deformation during activity. Cells sense and respond to mechanical signals through pathways of mechanotransduction13-15, but must also maintain integrity and homeostasis within the tissue. A mismatch between mechanical loading and cellular regulation can contribute to pathology, such as in musculoskeletal and connective tissue disorders16, with ageing being a significant risk factor17.
Here, we sought to compare responses to stiffness and mechanical loading in primary human mesenchymal stem cells (hMSCs), a cell type with important physiological and reparative roles, that have led to investigations of their therapeutic potential in tissues such as muscle18 and heart19. We contrast cellular responses to stiffness and strain cycling, and identify a rapid, reversible and structured regulation of the proteome following high-intensity mechanical loading. Furthermore, we identify SUN domain-containing protein 2 (SUN2) as a strain-induced breakpoint in the linker of nucleo- and cytoskeleton (LINC) complex of proteins that acts as a pathway of intracellular mechano-transmission13, 20, thus enabling the nucleus to ‘decouple’ from the cytoskeleton in response to intense strain.
RESULTS
Cyclic tensile strain (CTS) uncouples the correlation between cellular and nuclear morphology
Primary hMSCs were cultured on stiffness-controlled polyacrylamide hydrogels or silicone elastomer sheets that could be subjected to CTS (both collagen-I coated). hMSCs were found to spread increasingly on stiffer substrates over a physiological range (2 – 50 kPa, cultured for 3 days; Fig. 1a, Supplementary Fig. 1a), as has been reported previously1, 21. Cells subjected to sinusoidal, equiaxial CTS for 1 hour at 1 or 2 Hz (change in strain = 4%) showed significantly increased spreading immediately after loading (p ≤ 0.05), returning to initial spread areas after 24 hours (Fig. 1b, Supplementary Figs. 1b, c). Earlier reports of cell behaviour following strain have described cell alignment relative to the direction of strain22, 23 and reorganization of focal adhesion (FA) complexes and the cytoskeleton24-26. As the strain applied in our system had radial symmetry, no overall alignment was observed, but increased cell spreading was consistent with previous reports describing FA activation27. The increase in spreading of hMSCs following dynamic straining at 1 and 2 Hz was thus similar to that observed with changes in static substrate stiffness.
To explore mechanisms that allow cells to endure more challenging mechanical environments, we increased the frequency of CTS to 5 Hz (change in strain = 3.6%; referred to henceforth as ‘high-intensity CTS’). The increased cell spreading observed at lower frequencies was not seen following 1 hour of high-intensity CTS (Fig. 1c). Cell spreading was significantly decreased 24 hours after treatment (p = 0.05), but cells remained attached to the substrate. Furthermore, neither cell viability nor proliferation were significantly affected (Supplementary Figs. 1d, e).
The nuclear area of hMSCs was found to increase with cell spreading on stiffer substrates (Fig. 1d, Supplementary Fig. 1a). This agrees with findings in earlier works1, 28 and reflects the interconnected nature of the cyto- and nucleoskeleton29, which has been shown to be necessary for mechanotransduction20. However, we found that this correlated behaviour of cell and nuclear spreading was lost in hMSCs subjected to CTS: there were no significant changes in nuclear area following 1 hour of CTS at 1 or 2 Hz (Fig. 1e); and nuclear area was significantly decreased following 1 hour CTS at 5 Hz (p = 0.003; Fig. 1f, Supplementary Fig. 1f), recovering after 24 hours. Under all CTS conditions, ratios of nuclear to cytoplasmic area were significantly decreased immediately following strain (p < 0.05; Supplementary Fig. 1g). Thus CTS was found to decouple the coordinated behaviour of cell and nuclear spreading observed at equilibrium on stiffness-defined substrates, either through failure of the nucleus to match CTS-induced cellular spreading (CTS at 1 and 2 Hz), or through nuclear contraction while cell spreading remained constant (at 5 Hz; Fig. 1g). Dynamic loading was thus accompanied by a disruption of the mechanisms linking the cyto- and nucleo-skeletons.
CTS-induced nuclear contraction requires stretch-activated ion channels
Stretch activated ion-channels can enable rapid response to mechanical stimulation30. To investigate the role of ion channels in CTS-induced nuclear contraction, we combined high-intensity strain with a panel of ion channel inhibitors: GdCl3, a broad-spectrum inhibitor of stretch-activated ion channels31; RN9893, an inhibitor of transient receptor potential cation channel subfamily V member 4 (TRPV4)32; amiloride, an inhibitor of acid sensing ion channels (ASICs)31, and GsMTx4, an inhibitor of piezo channels33 (Fig. 2a, Supplementary Fig. 2a). GdCl3 inhibited nuclear contraction following strain, as did the TRPV4-specific drug RN9893 and, to a lesser extent, amiloride. GsMTx4 did not prevent nuclear contraction, although earlier work has shown it to be effective in inhibiting chromatin condensation under milder loader regimes (3% uniaxial strain at 1 Hz)33. This suggested that activation of different ion channels may be specific to the loading regime. Following all treatments, nuclear area was recovered to control levels 24 hours after straining (Fig. 2b).
High-intensity CTS significantly increased the texture parameter of nuclear DAPI staining (p = 0.006, Fig. 2c, Supplementary Fig. 2b), indicative of chromatin condensation33, 34 (comparable to the effect of divalent ions, Supplementary Figs. 2c, d). Treatment with GdCl3 at its IC50 of 10 μM35 did not prevent changes to DAPI-stain texture following CTS. This contrasted with earlier characterizations of milder loading regimes, where GdCl3 was found to block chromatin condensation, although the drug concentration was higher in this case33. Our finding indicated the robustness of the chromatin condensation response in cells subjected to high-intensity CTS, but also suggested that chromatin condensation and contraction of nuclear area could be caused by different mechanisms.
Cellular responses to high-intensity CTS are driven at the protein level
Using an initial targeted gene approach, we determined that the transcriptional response of hMSCs to 1 hour low-intensity CTS (1 Hz, change in strain = 4.0%), included upregulation of genes associated with cytoskeletal remodelling: intermediate filament vimentin (VIM) and alpha-actin-2 (ACTA2); and with the ECM: collagen alpha-1 (I) chain (COL1A1) and collagen alpha-1 (II) chain (COL2A1) (Supplementary Fig. 3a-d). These results are consistent with observed changes to cell morphology, and earlier characterizations of cellular responses to strain26, 36 and substrate stiffness1, which were proposed to increase robustness to stress. However, we found changes to the transcriptome assessed by RNA-Seq immediately following 1 hour high-intensity CTS to suggest only a narrow degree of change (Gaussian width = 0.21; Fig. 3a). Furthermore, gene ontology (GO) term analysis37, 38 of the genes affected by CTS suggested a general suppression of transcription and metabolism (Fig. 3b). Downregulation of transcriptional activity is consistent with our observations of chromatin condensation and previous reports of histone-methylation mediated gene silencing in endothelial progenitor cells subjected to low frequency (0.1 Hz) strain cycling39. The distribution of changes to gene expression was narrowed 24 hours after CTS (Gaussian width = 0.14; Supplementary Fig. 3e).
In contrast to the analysis of transcript levels, analysis of the intracellular proteome quantified by mass spectrometry (MS), showed greater changes compared to unstrained controls (Fig. 3c, Supplementary Fig. 3f). A Gaussian fit to the distribution of protein fold changes had a width of 0.61 and was displaced to the left, suggesting that for an increased number of proteins, the rate of turnover was greater than the rate of translation. Analysis of the Reactome pathways40-42 significantly affected (false discovery rate (FDR)-corrected p < 0.05) by CTS showed that ontologies relating to metabolism of both protein and RNA, signal transduction and response to external stimuli were downregulated (Fig. 3d). Changes to transcript and proteome following CTS were not correlated (R-squared = 0.002; Fig. 3e), indicating a post-transcriptional regulation of protein levels. The proteome returned towards the control state after 24 hours (Gaussian width = 0.25; Supplementary Figs. 3g-i).
The time-resolved proteomic response to high-intensity CTS was further classified by K-means clustering (Fig. 3f, Supplementary Fig. 3j). Clusters of protein levels were identified with: (i) an immediate but unsustained decrease (cluster 1), enriched for Reactome annotations associated with translation, protein folding and mechanisms of actin and tubulin folding; (ii) an initial but unsustained increase (cluster 2), enriched for an annotation of metallothionein binding (associated with the management of oxidative stress43); and (iii) an immediate and sustained suppression (cluster 3), with enrichment of annotations for translation and regulation of the Slit/Robo signaling pathway (associated with cell polarity and cytoskeletal dynamics44). Taken as a whole, this analysis shows a complex, time-resolved and structured protein-level response to cellular stress management.
CTS causes changes to protein conformation and post-translational modification
As changes in protein conformation are important to mechanotransduction14, 15, MS was performed following protein labeling with monobromobimane (mBBr), which by selectively labeling solvent-exposed Cys acts as an indicator of protein folding (Fig. 4a). MS was used to both identify mBBr-labeled proteins and quantify differential labeling in hMSCs following high-intensity CTS, relative to unstrained controls (Fig. 4b). The histogram of log2-fold changes in mBBr labeling showed a broad distribution of CTS-induced changes to mBBr reactivity, with labeling increased on average immediately following strain. The distribution was narrowed and centred about zero 24 hours after CTS, indicating a recovery of protein folding. Earlier applications of mBBr labeling have been used to identify force-dependent unfolding of domains in spectrin45, 46, cytoskeletal proteins47 and nuclear lamin-A/C (LMNA)1. Labeling of Cys522 in the Ig-folded domain of LMNA was previously used to report on the deformation of isolated nuclei subjected to shear stress in a rheometer1. We found the labeling of Cys522 to be increased 1.1-fold immediately following strain (FDR-corrected p < 0.001). We correlated changes to mBBr-labeled cysteine site occupancy and changes to total quantities of the parent proteins (Fig. 4c). This analysis suggested a systematic link between CTS-induced changes to protein conformation and stability (rates of translation vs. turnover), both on average (Figs. 3c, 4b) and in some cases on a protein-by-protein basis (Fig. 4c).
Changes to endogenous PTMs were quantified by MS in the same experiment. A histogram of log2-fold changes to phospho-site occupancy following high-intensity CTS versus unstrained controls (Fig. 4d) showed increased phosphorylation. Phosphorylation of LMNA has been shown to be regulated in response to changes in substrate stiffness1, 7 and here we detected modest (~1.1-fold) but significant (FDR-corrected p < 0.001) increases in phosphorylation at S22, S390, S392 and S636. A correlation plot between changes of phosphorylation at individual residues and changes in quantities of the phosphorylated protein (Fig. 4e) showed that changed phosphorylation often accompanied modulation of protein quantity. A similar analysis of oxidised peptides showed that oxidation was increased immediately following CTS (Fig. 4f) and in a third of detected sites of oxidation, increased oxidation correlated with decreased protein levels (points in the top left quadrant of Fig. 4g). Previous work has established that cyclic stretching can increase levels of reactive oxygen species (ROS) in a range of cell types through activation of nicotinamide adenine dinucleotide phosphate (NADPH) oxidase systems and mitochondria. ROS have an important role in signal transduction, for example during vascularization, but can contribute to oxidative damage to lipids, proteins and DNA48. This potential for oxidative damage is perhaps consistent with the upregulation of protective metallothioneins (Fig. 3f). Profiles of both phosphorylation and oxidation became similar to controls 24 hours after CTS (Figs. 4d, f).
CTS disrupts the linker of nucleo- and cytoskeleton (LINC) complex
Having established systematic responses to CTS, we sought to identify specific cases where protein regulation modulated mechano-transmission to the nucleus. There is a continuous system of structural proteins that connect cell interactions with the ECM at FAs, through the cytoskeleton and LINC complex to chromatin13, 14, 20. A central feature of this pathway is the LINC complex, which spans the nuclear envelope (NE) and includes nesprin proteins which bind to cytoskeletal components in the cytoplasm, but have Klarsicht, ANC-1, and Syne homology (KASH) domains extending into the nuclear lumen. The KASH domains bind to the Sad1 and UNC-84 (SUN) domains of SUN-domain containing proteins, which in turn bind to the nuclear lamina that lines the inner NE49. The nuclear lamina is composed of intermediate filament lamin proteins that confer structural integrity to the nucleus2, 50, 51 and also interface with chromatin and a range of regulatory and NE associated proteins1, 52. The complete system of protein linkages enables nuclear positioning53 and acts as a conduit for mechanical signals to regulate the genome20, 52.
LINC and NE protein levels were quantified by MS following 1 hour of high-intensity CTS, relative to unstrained controls (Fig. 5a). SUN-domain containing protein 2 (SUN2) was reduced to 52% of control levels (FDR-corrected p < 0.0001). In contrast, LMNA levels were not significantly altered. We also quantified proteins located specifically at the NE using immunofluorescence (IF) imaging (Fig. 5b, c, Supplementary Figs. 4a-e). IF confirmed that SUN2 levels were decreased at the NE following CTS (p = 0.03). Emerin (EMD), which has a role in the mechanical stimulation of the serum response factor (SRF) pathway54 was found to be significantly enriched at the NE (p < 0.0001), consistent with previous reports39. Additionally, we found that treatment with GdCl3 – shown to prevent contraction of nuclear area following CTS (Fig. 2a) – also prevented the loss of SUN2 (Fig. 5c, Supplementary Fig. 4f).
The composition of the nuclear lamina, characterised by the ratio of LMNA to lamin-B1 (LMNA:B1), has been used previously as a readout of nuclear adaption to the mechanical properties of the cellular microenvironment, with increased LMNA:B1 indicative of nuclear stiffening1, 50. We quantified LMNA:B1 by IF after 1 hour high-intensity CTS (Fig. 5d), finding it to be significantly decreased by CTS, relative to controls (p = 0.007); LMNB1 at the NE was significantly increased (p = 0.001). In contrast to the response of hMSCs to stiffer substrates1, the composition of the lamina was not altered to suggest stiffening following CTS.
SUN2 is upstream of chromatin and cytoskeletal regulation
As decreased SUN2 at the NE was identified in response to CTS, we sought to further investigate its role. We quantified the effects of siRNA knockdown (KD) of SUN2 on primary hMSCs by MS proteomics, comparing two siRNAs to increase confidence of identifying on-target effects (Fig. 6a). The siRNAs showed KD of SUN2 to 35% of scrambled controls (Supplementary Figs. 5a, b). A Reactome pathway analysis of both KDs (Fig. 6b) identified significant perturbations to ‘polycomb repressive complex 2 (PRC2) methylates histones and DNA’, ‘Protein lysine methyltransferases (PKMTs) methylate histone lysines’ and ‘Rho GTPases activate IQGAPs’ (all FDR-corrected p-values < 0.05). These annotations suggested that SUN2 could be upstream of both aspects of chromatin regulation, consistent with our earlier observations, and nucleus-to-cytoskeleton (‘inside out’) signalling. A scatter plot of protein fold changes following SUN2 KD versus CTS showed correlation in cytoskeletal proteins such as actin and microtubule-associated protein (Fig. 6c), again suggesting that SUN2 regulation following CTS could be upstream of aspects of cytoskeletal remodelling.
SUN2 modulates transmission of CTS to the nucleus and DNA damage
To investigate the role of SUN2 on mechano-transmission to the nucleus, we examined changes to nuclear morphology in immortalised hMSCs (Y201 line) shown to maintain the multipotency and mechano-responsiveness of primary MSCs55, 56. These were cultured on plastic with siRNA KD or doxycycline-induced overexpression (OE) of SUN2 (Fig. 6d, Supplementary Figs. 5c-h). In contrast to the response to CTS, where loss of SUN2 accompanied nuclear contraction, both KD and OE of SUN2 caused nuclear areas to increase. This suggested that there was a LINC complex composition determined by the cellular microenvironment – in this case, culture on plastic – and that imposed changes to protein levels caused a mismatch between nuclear and cellular properties1. We also found evidence supporting a disruption of nucleus-to-cytoskeleton signalling as both SUN2 KD and OE caused significantly increased cell spreading (Supplementary Figs. 5i, j). Consistent with an interpretation of LMNA as a reporter of a functioning mechanical linkage between the cytoskeleton and the nucleus, SUN2 OE led to loss of LMNA at the NE (Fig. 6e, Supplementary Figs. 5k, l).
Finally, we sought to determine how perturbation of SUN2 could affect cellular responses to CTS. We found that SUN2 KD in primary hMSCs was sufficient to prevent changes to nuclear:cytoplasmic area ratio following 1 hour of high-intensity CTS (Figs. 7a, b, Supplementary Figs. 6a-c). An siRNA treatment was also sufficient to block changes to nuclear texture indicative of chromatin condensation (Fig. 7c, Supplementary Fig. 6d). Likewise, SUN2 OE in immortalised hMSCs blocked the changes to nuclear:cytoplasmic area ratio observed in controls cells following 1 hour of high-intensity CTS (with recovery after 24 hours, Figs. 7d-f). Mechanical strain has previously been shown to cause DNA damage, inducing apoptosis in vascular smooth muscle cells57, and causing accumulation of damage to DNA and chromatin in nuclei subjected to extreme deformation as cells migrate through constricted environments58-60. We were surprised, therefore, to find that CTS here resulted in a small but significant decrease in the intensity of γH2AX staining in primary (p = 0.03) and immortalised MSCs (p = 0.0002), suggestive of a protective effect (Figs. 7g, h, Supplementary Fig. 6e). We found OE of SUN2 to override the decoupling response to CTS in immortalised hMSCs, concomitant with a significant increase in γH2AX staining (p < 0.0001; Fig. 7i). These results indicated that appropriate levels of SUN2 were essential for the mediation of nuclear decoupling in response to dynamic loading and therefore to afford protection to DNA.
DISCUSSION
We have demonstrated in primary cells from multiple donors that hMSCs have a rapid, structured and reversible response to CTS regulated at the protein level. This response was dependent on both functional ion channels and appropriate levels of the LINC complex protein SUN2 (Fig. 8a). Furthermore, CTS was shown to cause changes within the LINC complex (Fig. 8b), in particular to the regulation of SUN2, enabling cells to decouple nuclear and cellular morphological behaviours and conferring protection to DNA. Robustness is increased through cyto- and nucleoskeletal remodeling in cells that have reached a ‘mechanical equilibrium’ state on increasingly stiff substrates1, 61. However, remodeling of the nuclear lamina seemed less important in the rapid response to high-intensity CTS. Although we were able to quantify modest (but significant) responses in LMNA conformation and phosphorylation, changes to the composition of the lamina were not indicative of nuclear stiffening. Mechano-transmission to the nucleus is an important mode of mechanical signaling, but if unregulated, has potential to apply stresses to chromatin. While a number of nuclear stress management mechanisms have been characterised, including chromatin condensation33, 39, chromatin detachment from the NE62, and altered nuclear mechanics1, 7, 63, a mechanism that isolates the nucleus from the cytoskeleton, as demonstrated here through regulation of SUN2, has potential to be both rapid and reversible. A role for SUN proteins in such mechanisms is further supported by analysis of protein turnover rates: SUN1 and 2 were reported to have the shortest half-lives of LINC complex proteins (Supplementary Fig. 7f)64.
This study used large, unbiased ‘-omics’ datasets to identify and focus on regulation of the LINC complex. However, the techniques described here have potential to explore other aspects of the global cellular response to mechanical stress in greater detail. These include regulation of other structural proteins, such as the intermediate filaments65, molecular chaperones and the pathways that manage DNA and oxidative damage. The use of MSCs as a model system to study mechano-responsive processes has been widespread, but these cells are also being assessed for their potential for therapy in heart19 and muscle18 – tissues subject to sustained and high-frequency deformation66. Furthermore, this work may be particularly relevant to understanding how mechanical stress contributes to age-related pathology. Many aspects of the cellular stress response are abrogated in ageing67, but crucially, the NE may be particularly susceptible to misregulation68, 69.
Methods
Methods and any associated references are available in the online version of the paper.
Note: Supplementary Information is available in the online version of the paper.
Author Contributions
Investigation, HTJG, VM, OD, RP, MRJ and JS; Formal Analysis, HTJG, VM and JS; Writing – Original Draft, HTJG; Writing – Visualization, Review & Editing, HTJG, VM, OD, MRJ, RP, APG, SMR and JS; Project Administration and Funding Acquisition, JS.
Competing Interests
The authors declare no competing interests.
Acknowledgements
HTJG and JS were funded by a Biotechnology and Biological Sciences Research Council (BBSRC) David Phillips Fellowship (BB/L024551/1). VM was partially supported by a studentship from the Sir Richard Stapley Educational Trust. OD was supported by a Wellcome Trust Institutional Strategic Support Fund (097820/Z/11/B). Proteomics was carried out at the Wellcome Trust Centre for Cell-Matrix Research (WTCCMR; 203128/Z/16/Z) Biological Mass Spectrometry Core Research Facility; RNA-Seq was performed by the Genomic Technologies Core Facility (GTCF). We thank Professor Paul Genever (University of York, UK) for use of the Y201 immortalised hMSC line; Professor Tim Hardingham (University of Manchester, UK) for useful discussions; Drs. Ronan O’Cualain and David Knight (WTCCMR) for advice on mass spectrometry analysis.