Mechanical memory based on chromatin and metabolism remodeling promotes proliferation and smooth muscle differentiation in mesenchymal stem cells

Stem cells respond and remember mechanical cues from the microenvironment, which modulates their therapeutic effects. Chromatin organization and energy metabolism regulate the stem cell fate induced by mechanical cues. However, the mechanism of mechanical memory is still unclear. This study aimed to investigate the effects of mechanical amplitude, frequency, duration, and stretch cycle on mechanical memory in mesenchymal stem cells. It showed that the amplitude was the dominant parameter to the persistence of cell alignment. F‐actin, paxillin, and nuclear deformation are more prone to be remolded than cell alignment. Stretching induces transcriptional memory, resulting in greater transcription upon subsequent reloading. Cell metabolism displays mechanical memory with sustained mitochondrial fusion and increased ATP production. The mechanical memory of chromatin condensation is mediated by histone H3 lysine 27 trimethylation, leading to much higher smooth muscle differentiation efficiency. Interestingly, mechanical memory can be transmitted based on direct cell–cell interaction, and stretched cells can remodel the metabolic homeostasis of static cells. Our results provide insight into the underlying mechanism of mechanical memory and its potential benefits for stem cell therapy.

which can persist by several orders of magnitude longer than the timescales of cell signaling. 3Initially, a soft substrate was used to mimic the muscle elasticity to culture muscle stem cells, significantly promoting muscle regeneration after transplantation into mice. 4Subsequently, a previous study reported that the substrate stiffness of mechanical dosing formed mechanical memory to increase osteogenesis in MSCs, and the YAP/TAZ acted as an intracellular mechanical rheostat to regulate cell fate. 5When MSCs were cultured on a stiff substrate, the microRNA miR-21 served as a long-term memory keeper of the osteogenic or fibrogenic program. 6,7A recent study reported that mechanical memory also existed in the protein diffusivity of chromatin and nucleoplasm, which was regulated by nuclear pore complexes. 8ynamic stretch is associated with both physiological and pathological microenvironments for various cell types.It is widely used as a crucial pretreatment procedure in tissue engineering.For instance, studies have shown that preconditioning cells with cyclic strain enhances the contractility of engineered skeletal muscle and promotes cell-based ligament regeneration. 9,10However, research on the optimal mechanical dosing and parameters of cyclic stretch to establish mechanical memory is limited.Therefore, some fundamental questions regarding dynamic stretch need to be addressed, which may help develop an effective strategy for selecting mechanical stretch to regulate stem cell behaviors in clinical treatment.These questions are as follows: What is the key parameter of mechanical memory, magnitude, frequency, duration, or loading mode?How long can mechanical memory be maintained in cells?What are the mechanisms responsible for the mechanical memory of lineage commitment in stem cells?Answering these questions may provide valuable insights into the role of mechanical cues in regenerative medicine and tissue engineering.
Subcellular structures such as the cytoskeleton 11 and nucleus 12 can serve as mechanotransducers to regulate cell fate and behavior.Forces are transmitted through integrinmediated focal adhesions (FAs), which are connected to the cytoskeleton within cells. 13For instance, cyclic stretch induces the reassembly of the actin stress fiber-adhesion system perpendicular to the direction of stretch, leading to a reorientation of the cells in the same direction. 14The constriction of cell nuclei induced by micro-topographies or micropillar patterns, promotes epigenetic and transcriptional reprogramming to determine quiescent-like phenotype or osteogenic differentiation in MSCs. 15,16In addition, the stretch-induced nuclear stiffening is attributed to the remodeling of the Lamin A/C and increased heterochromatin content. 17Forces can directly stretch chromatin and enhance gene transcription. 18The gene transcription continues even after the mechanical force is withdrawn. 19e selection of physical and mechanical experimental parameters is crucial for engineering cell fate and behavior.However, the mechanisms behind the mechanical memory for biophysical cues are less understood, and the role of mechanical memory in driving stem cell differentiation is still far from being elucidated.
Energy metabolism not only provides energy and biosynthetic substrates but also plays a crucial role in stem cell fate. 20,21Undifferentiated MSCs are highly glycolytic. 22witching between glycolysis and oxidative phosphorylation is based on the differentiation of MSC lineage. 23oreover, mitochondria are the main organelles involved in energy metabolism.They are elongated and interconnected during osteogenesis and adipogenesis in MSCs, but fragmented during chondrogenesis in MSCs. 24Besides, emerging evidence has shown that energy metabolism can respond to mechanical cues.For example, shear stress and stretch can promote mitochondrial biogenesis and ATP production. 25,26However, the mechanical memory of energy metabolism is not well understood.
In this study, we found that intermittent stretch-induced reorientation memory was mainly affected by strain amplitude.In addition, both the cytoskeleton and the nucleus underwent remodeling, contributing to mechanical memory.Intermittent stretch led to transcriptional memory and maintained active metabolism.The mechanical memory was regulated by the epigenetic remodeling of chromatin through nuclear deformation, which promoted smooth muscle cell (SMC) differentiation in MSCs.We also preliminarily explored the transmission of mechanical memory between cells.It was observed to be transmitted via direct cell-cell interaction and was accompanied by active metabolism.

| Cell culture
Primary MSCs were obtained from the femurs of 4-week-old Sprague-Dawley male rats (Peking University Laboratory Animal Center, Beijing, China).This study was approved by the Biological and Medical Ethics Committee of Beihang University (IRB No.: BM 20180014).The bone marrow was flushed out and separated by gradient centrifugation for 10 min on 1.073 g/mL Percoll (YT-0727, YTHX Biotechnology).Cells were maintained in Dulbecco's modified Eagle medium-low glucose (DMEM-LG, Gibco) supplemented with 20% fetal bovine serum (FBS) and 1% penicillin/streptomycin at 37°C with 5% CO 2 .After the cells were adherent to the dish for 48 h, the DMEM-LG containing 10% FBS was used for medium change, and nonadherent cells were discarded.The medium was changed every three days.After reaching 80%-90% confluency, the cells were passaged with 0.25% trypsin and 0.02% EDTA.Cells from passages 2-4 were used in the experiments.The isolated cells were conformed with positive markers for CD29 and CD44, and negative markers CD31, CD34, and CD45. 27,28

| Application of uniaxial stretch
The uniaxial stretch device was driven by a stepper motor (35H4R-05-A10, Haydon Kerk), and consisted of a control unit and culture chamber.The stretching device allowed to application of reproducible cyclic tension in the frequency (0-5 Hz) and magnitude (0%-20%).The cells were seeded on an elastic silicone membrane (Specialty Manufacturing Inc.), and it was coated with 10 μg/mL fibronectin (356008, Corning) overnight at 4°C.The cells were subjected to uniaxial stretch after adhering to the elastic silicone membrane for 24 h.During the uniaxial stretch, the device was placed in an incubator at 37°C with 5% CO 2 .The control group was placed in the same incubation condition without tension.

| Quantification of cell orientation
Phase contrast images were collected by an inverted microscope (Olympus) at each time point.The cell orientation was quantitated as described previously. 29The stretch direction and the long axis of the cell were orientation angle α.For the control group, orientation angle α was the long axis of the cell concerning the x-axis of the images.The orientation parameter 29 was defined as: When the S = 0, if the cells are randomly oriented, S = 1 if the cells are parallel oriented, and S = −1 if the cells are perpendicular to the stretch direction.The orientation parameter was calculated from at least 1000 cells per group.The experiment was repeated three times.

| Measurement of chromatin condensation parameter
The nuclei were stained using DAPI and the staining procedure was performed as described above.Images of the nuclear midsection were acquired using a confocal microscope (Leica SP8).To assess the chromatin condensation parameter (CCP), an edge map was produced with a gradient-based Sobel edge detection algorithm. 30The strong edge lines are reduced into single-pixel-thickness entities by a thinning morphological algorithm.The area beyond the nuclear edge is measured, and the ratio of this to the cross-sectional area of the nucleus is the CCP.

| RNA-sequencing
Total RNA was isolated by Trizol (Invitrogen), and RNA quality control was assessed with an Agilent 2100 bioanalyzer (Agilent Technologies).The sequencing library was built using the NEBNext Ultra™ RNA Library Prep Kit (New England Biolabs) from two biological replicates of each group by Novogene (Beijing, China).RNA sequencing was carried out on Illumina NovaSeq 6000.Reads were aligned to the reference genome using HISAT2 software (version 2.0.5) and counted using feature counts (version 1.5.0-p3).Differential gene expression was analyzed by the DESeq2 package (version 1.16.1) and Bioconductor package edgeR (version 3.18.1).Enrichment analysis was performed with the cluster profile (version 3.4.4).

| Quantitative real-time PCR
Total RNA was extracted with Trizol, and reversetranscribed into complementary DNA (cDNA) using a PrimeScript RT reagent kit (Takara Bio).TB Green quantitative real-time PCR (qRT-PCR) was performed in triplicates by QuantStudio 1 (Thermo Fisher Scientific) with a cDNA template.The relative gene expression was calculated by the 2 −ΔΔCt method, and normalized to the glyceraldehyde-3-phosphate-dehydrogenase (GAPDH).The primer sequences (Sangon Biotech) are detailed in Table S1.

| siRNA transfections
The siRNAs were synthesized by GenePharma (GenePharma), and transfected by lipofectamine RNAiMAX (13778150, Thermo Fisher Scientific).Briefly, the 10 nM siRNA was diluted in Opti-MEM reducedserum medium (31985062, Thermo Fisher Scientific), and mixed gently with lipofectamine RNAiMAX.Then, the transfection reagent was aliquoted into each well.After 48 h of transfection, the cells were used for subsequent experiments.The information on siRNAs was listed in Table S2.

| ATP measurement
The intracellular ATP was detected using a luciferinluciferase-based ATP assay kit (G7570, Promega).Briefly, a 100 μL sample containing 4000 cells was mixed with 100 μL of luciferase reagent in an opaque 96-well plate and allowed to stand at room temperature for 10 min.A standard curve was constructed using 0 to 1 μM ATP solutions.The relative luminescence unit (RLU) was detected by a luminescence reader (Thermo Fisher Scientific), and the ATP concentration was calculated based on the standard curve.The experiment was repeated three times.

| Analysis of mitochondrial morphology
At the end of the stretch, cells were cultured with 300 nM TMRM (I34361, Thermo Fisher Scientific) in the dark for 20 min at 37°C, and then living cells were immediately imaged by confocal microscope.For immunofluorescence, cells were fixated and incubated with the primary antibody TOM20 (186735, Abcam) to label mitochondria.Mitochondrial length and network branches were analyzed by Fiji software (MiNA and Fiji plugin).The classification of mitochondrial morphology was quantified by machine learning as described previously. 31The mitochondrial morphology was classified as punctate, intermediate, and filamentous, and then calculated the percentage area of each category in each image.

| EdU assay
The cell proliferation was detected with the iClick EdU Andy Fluor 488 Imaging Kit (Gene Copoeia Inc.) by following the manufacturer's instructions.Cells were subjected to uniaxial stretch and then incubated statically for 12 h, followed by incubation with EdU reagent for 4 h.Cells were fixed with 3.7% paraformaldehyde for 30 min and permeabilized with 0.5% Triton X-100 for 20 min in PBS at room temperature.The EdU-positive cells were stained using an iClick reaction cocktail for 30 min in the dark.The nucleus was stained with 5 μg/mL of Hoechst 33342 for 15 min.Images were taken using a confocal microscope, and EdU-positive cells were quantified by Image J.

| Smooth muscle differentiation
For smooth muscle differentiation, the culture medium was DMEM-LG with 1% FBS and 5 ng/mL TGF-β1 (CA59, Novoprotein).Cells were cultured in 6-well plates with differentiation medium for 7 days after loading intermittent stretch, and the medium was changed every third day.

| STED imaging and analysis
For STED imaging, the secondary antibody of Abberior Star Orange (561 nm, Abberior, NanoTag, 1:150) was used to minimize photobleaching.Super-resolution imaging was acquired using an Abberior STED system (Abberior Instruments) with an Olympus 100×, NA 1.45 objective lens.To identify the clusters of proteins, open-source rap-idSTORM software was used to obtain the reconstructed images and localization files.Then, the open-source SR-Tesseler software was used to identify the localization files and display the single-molecule detection images. 32The software computed voronoï diagrams on the detection dataset generated the tesselations and obtained the local density of proteins automatically.The voronoï diagram subdivided the single-molecule detection images into polygons centered on local molecules, and the local density was defined as the reciprocal of the area of the voronoï unit polygon.
For indirect co-culture, cells were cultured in a 6-well Transwell plate with a 0.3-mm polyester membrane.The static cells were seeded to the upper compartment of the chambers, and the two-cycle stretched (2ST) cells were seeded in the lower chambers.The ratios of cells at 50%:50% for indirect co-culture in a density of 10 000/cm 2 .

| CCK-8 assay
Cell proliferation was detected by Cell Counting Kit-8 (C0038, Beyotime).The cells were exposed to 10% stretch for 3 h at 0.5 Hz with two-cycle stretch (2ST), and then mixed with static (con) cells and seeded on 96-well plates or 12-well Transwell plates in a density of 6000/well.After 24 h incubation, the medium was removed and incubated with CCK-8 solutions at 37°C for 1 h.Absorbance (450 nm) was measured using a varioskan flash multimode reader (Thermo Fisher Scientific).

| Mechanical memory in cell reorientation in response to intermittent stretch
MSCs were subjected to intermittent uniaxial stretch at various amplitudes (5%, 10%, or 15%) and frequencies (0.05, 0.1, 0.5, or 1 Hz).The stretching protocol involved a 3-h stretch period (Stretch 1), followed by a 3-h static culture period (Relax 1), another 3-h stretch period (Stretch 2), and finally another 3-h static culture period (Relax 2) (Figure 1A).We used order parameters to evaluate the call alignment, where a value of −1 indicated perpendicular alignment, a value of 1 indicated alignment parallel to the stretch axis, and a value of 0 indicated random orientation.Initially, MSCs exhibited random alignment on the elastic silicone membrane.However, the order parameter decreased after the first stretch with a 5%, 10%, or 15% amplitude at 0.5 Hz, indicating that cells demonstrated pronounced perpendicular alignment to the stretch direction (Figure 1B).The order parameter increased after Relax1, but further reduced after Stretch 2 and remained lowered even after Relax 2 at 5% and 10% amplitude (Figure 1B).However, it was not preserved after Stretch 2 at 15% (Figure 1B).This indicated that the amplitude of stretch was essential to the persistence of cell alignment.Next, we determined the effect of different frequencies (0.05, 0.1, 0.5, and 1 Hz) on cell alignment.We found that the cell alignment was maintained with 0.1-1 Hz at 5% stretch and 0.05-1 Hz at 10% stretch (Figures 1B and S1A,B).However, the cell alignment was not maintained by any frequency of 15% stretch after Relax 2 (Figures 1B and S1C).Our results showed that the stretch amplitude played a dominant role compared with frequency in maintaining the persistence of cell alignment.Frequency was found to be a minor factor that harmonized with amplitude.The cells were subjected to intermittent stretch for 1 or 6 h to investigate whether the duration of stretch affected the maintenance of cell alignment.Similarly, the cells were oriented perpendicular to the main axis of the strain after 10% amplitude at 0.5 Hz for either 1 or 6 h (Figure 1C).Despite no significant difference in the first period for either 1 or 6 h, the order parameter decreased by a longer duration of stretch (Figure 1C).This result indicated the effect of the duration of stretch on the mechanical memory of cell alignment.When the stretch cycle increased to four cycles, the order parameter was preserved starting from the second cycle (Figure 1D).It indicated that the constituted stretch training contributed to the persistence of cell alignment.We further observed how long the stretch cycle-induced alignment could be maintained.The cells were exposed to two cycles of intermittent stretch and subsequently unloaded for 72 h.The order parameter of the cells gradually decreased and the cells returned to a random orientation after 24 h of relaxation (Figure 1E).Collectively, our data demonstrated that the threshold of an amplitudedependent mechanical memory of cell reorientation was below 10%, whereas the stretch frequency or duration was the minor parameter affecting the mechanical memory.The two cycles of stretch-induced alignment could be maintained for 24 h.

| Intermittent stretch-induced cytoskeleton and nuclear remodeling to contribute to mechanical memory
The mechanical cues could rearrange the cytoskeleton which is the main element for cell alignment. 33We also investigated the F-actin distribution according to different stretch magnitudes and frequencies.Our results indicated that the F-actin stress fibers tended to align perpendicular to the main axis of stretch and the interquartile range was less diverse (Figure 2A-C).One cycle of stretch could maintain the distribution of F-actin by 5%-15% of a stretch at 0.05-1 Hz (Figure S2A-O).The alignment of F-actin stress fibers in response to mechanical stresses was mediated by FA.The FA number decreased on stretching (10% and 0.5 Hz), and one cycle of stretch was sufficient to maintain less adhesion as well as the second cycle (Figure 2D,E).The order parameter of FA was also decreased by intermittent stretch (Figure 2F), which was similar to cell and stress fiber orientation.
Next, we detected whether the nuclear shape demonstrated the mechanical memory for the intermittent stretch.The nuclear aspect ratio (NAR) increased after Stretch 1 with a 10% amplitude at 0.05, 0.1, 0.5, or 1 Hz, indicating elongated nuclei (Figures 2G,H and S3).The NAR barely changed at Relax 1, whereas it was shifted to higher Stretch 2 and preserved after Relax 2 (Figures 2G,H  and S3).A similar trend, with an increase and preservation in NAR in response to intermittent stretch, was observed for both 5% and 15% amplitudes at 0.05, 0.1, 0.5, or 1 Hz (Figure S3).The nuclear height decreased after the first stretch and remained reduced during subsequent relaxation and stretch (Figure 2G,I).It indicated that stretch once could elongate and flatten the nucleus.The remodeled nuclear shape could be maintained for at least 12 h.
The cytoskeleton networks connect the cell membrane and the nucleus through the linker of the nucleoskeleton and cytoskeleton (LINC) complex, allowing the transmission of forces directly to the nucleus.We knocked down nesprin-1, a bridge between the LINC complex and nuclear membrane, to investigate whether stretch-induced nuclear deformation was mediated by the LINC complex.The preservation of intermittent stretch-induced cell reorientation was disrupted by the knockdown of nesprin-1 (Figure S4A), whereas the stress fiber reorientation and the increased NAR were preserved in the absence of nesprin-1 (Figure S4B,C).Furthermore, the nuclear height was not decreased by intermittent stretch with siNesprin-1 (Figure 2J).These findings suggested that the nucleus was sensitive to mechanical cues, and intermittent stretchinduced structural nuclear memory through the cytoskeleton and nucleoskeleton.

| Intermittent stretch-generated transcriptional memory in MSCs
A global transcriptome profile of cells was performed using RNA sequencing after loading with intermittent F I G U R E 1 Intermittent stretch-induced mechanical memory in cell reorientation.(A) Scheme of the intermittent stretch set-up (Stretch 1: loading stretch for 3 h, Relax 1: static culture for 3 h, Stretch 2: loading stretch for 3 h, Relax 2: static culture for 3 h).(B) Left, phase-contrast images of MSCs were acquired with and without stretch at each time point (amplitude of 5%, 10%, 15% at 0.5 Hz).Right, quantification of orientation parameter S = <cos2α> (S = 0, if the cells are randomly oriented, S = 1 if the cells are parallel oriented, and S = −1 if the cells are perpendicular to the stretch direction).(C) Left, at each time point, phase-contrast images of cells were acquired with and without stretch at each time point (amplitude of 10% at 0.5 Hz for 1 or 6 h).Right, quantification of orientation parameter S = <cos2α>.(D) Left, phase-contrast images of cells were acquired with and without stretch at each time point (amplitude of 10% at 0.5 Hz).Right, quantification of orientation parameter S = <cos2α>.(E) Left, phase-contrast images of cells were acquired with and without stretch and cultured statically for 60 h after loading intermittent stretch at each time point (amplitude of 10% at 0.5 Hz).Right, quantification of orientation parameter S = <cos2α>.Data from at least three independent experiments.All graphs showed mean ± SEM.NS, not significant, *p < .05,**p < .01 or ***p < .001.stretch for 3 h to investigate the relationship between intermittent stretch and changes in gene expression.Compared with Stretch 1, Stretch 2 resulted in a larger number of upregulated genes, which remained upregulated even after relaxation (Figure 3A).Among these, 93 upregulated genes displayed a memory response (Figure 3B).The functional enrichment analysis of these 93 upregulated genes revealed enrichment in Gene Ontology categories related to histone methyltransferase activity and chromosome region (Figure 3C).
GO enrichment analysis was performed for the whole genome to obtain a comprehensive view of differentially expressed genes (DEGs).The upregulated DEGs were mainly enriched in six GO terms after Stretch 1, such as RNA modification, signaling receptor binding, and receptor regulator activity (Figures 3D and S5).Only one GO term, nucleus, remained upregulated after Relax 1 compared with Stretch 1 (Figure 3E).However, 78 GO terms were enriched from upregulated genes after Stretch 2, such as nucleus, mitochondrial outer membrane, and RNA processing (Figures 3F and S5).In addition, 60 GO terms were upregulated after Relax 2 compared with Stretch 2, which was primarily related to ATP metabolic process and mitochondrial structure and function (Figures 3G and S5).These results indicate that intermittent stretch generated transcriptional memory and repeated stretch promoted a larger variety of gene expression patterns.

| Intermittent stretch regulated the persistence of chromatin condensation
Given the enrichment of transcriptional memory of genes in GO terms related to the chromosome region and histone methyltransferase activity, we next assessed the chromatin status and histone methylation.An image-based technique was used to calculate the chromatin condensation parameter (CCP). 30The increased CCP indicated that chromatin condensation increased immediately after Stretch 1 and remained stable during subsequent relaxion and stretching (Figure 4A).Chromatin condensation typically corresponds to the enrichment of various epigenetic marks, such as H3 lysine 27 trimethylation (H3K27me3)-marked facultative heterochromatin and H3 lysine 9 trimethylation (H3K9me3)-marked constitutive heterochromatin. 34he nuclear intensity of H3K27me3 increased after Stretch 1 and remained high during the subsequent relaxion and stretching (Figure 4B).However, the nuclear intensity of H3K9me3 was barely affected by intermittent stretch (Figure S6).Furthermore, the knockdown of nesprin-1 inhibited the intermittent stretch-induced increase in CCP (Figure 4C).These results suggested that intermittent stretch mediated the persistence of chromatin condensation, and mechanical memory of chromatin condensation was more sensitive than the reorientation of cells or F-actin stress fibers.

| Intermittent stretch-induced the persistence of increased metabolic activity in MSCs
Emerging evidence indicates that the metabolic change can be induced by mechanical cues. 35,36The RNA sequencing results showed that energy metabolism and mitochondria-related functions were upregulated by intermittent stretch during a 3-h stretch period (Figure 3F,G).We observed an altered mitochondrial morphology when the cells were exposed to intermittent stretch.The mitochondrial length and the number of network branches increased immediately after Stretch 1 and remained fused in the subsequent relaxion and stretching (Figure 5A-C).The intermittent stretch-induced filamentous mitochondria were also confirmed through unbiased machine learning algorithms (Figure 5A,D).The cellular ATP levels were also promoted by intermittent stretch, which remained high after Relax 1 and Relax 2 (Figure 5E).Furthermore, RNA sequencing analysis showed that the glycolysis-related genes were upregulated after Relax 1 but downregulated after Relax 2 (Figure 5F).The tricarboxylic acid cycle genes were not prominently changed by intermittent stretch (Figure 5G).The mitochondrial electron transport chain (ETC) genes were upregulated after Relax 1 but downregulated after Stretch 2 and returned to the same level as the control after Relax 2 (Figure 5H-K).These likely reflected the upregulation of metabolic genes to compensate for the production of energy after Stretch 1. Overall, these results suggested that the stretch promoted cell metabolism.Consequently, the cells maintained high metabolic activity and generated a metabolic memory.

| H3K27me3 preserved mechanical memory-mediated smooth muscle differentiation
Next, we examined the regulation of stem cell behavior through intermittent stretch-mediated mechanical memory.The cells were cultured for 12 h after one or two cycles of stretch, and the effect of intermittent stretch on the proliferation rate of MSCs was measured using the EdU assay (Figure 6A).The proliferation of MSCs was barely affected after one-cycle stretch but was promoted by two-cycle stretch (Figure 6B).Furthermore, the gene expression of CNN1 and TAGLN, both smooth muscle cell markers, was immediately increased by intermittent stretch (Figure 6A,C,D).These markers continued to exhibit high expression even after 2 days of relaxation following a two-cycle stretch, but not after a one-cycle stretch (Figure 6A,C,D).These findings showed that intermittent stretch promoted and preserved the MSCs to differentiate into SMCs.
We then investigated how mechanical memory could serve as a relevant input for controlling MSC differentiation.The nuclear height was decreased by an intermittent stretch on day 0, and preserved by a two-cycle stretch after 2 days of relaxation, rather than a one-cycle stretch (Figure 6E).Likewise, the CCP remained at a high level and was accompanied by the expression of H3K27me3 after 2 days of relaxation following two-cycle stretch (Figure 6F,G).MSCs were subjected to a two-cycle stretch and cultured in an SMC differentiation medium.
The gene expression of CNN1 increased even after 7 days of relaxation (Figure 6H), and the protein expression of SMMHC, α-SMA, and SM22-α was also increased (Figure 6I,J).Super-resolution microscopy was used to quantify chromatin organization.Image segmentation revealed higher H3K27me3 cluster density after 7 days of relaxation following a two-cycle stretch, which represented the formation of denser chromatin (Figure 6K).The regulation of H3K27me3 was mediated by an enhancer of zeste homolog 2 (EZH2), which was inhibited using a histone methyltransferase inhibitor GSK343.The gene expression of CNN1 was suppressed by the GSK343

F I G U R E 6
Mechanical memory persists in smooth muscle differentiation of MSCs via H3K27me3.(A) Scheme of the EdU assay and detection of differentiation.MSCs were subjected to intermittent stretch and cultured statically for 12 h, and then the proliferation was measured by EdU assay.MSCs were subjected to intermittent stretch and cultured statically, and then the genes of SMC differentiation were measured by qPCR on days 0, 1, and 2. (One-cycle: loading stretch for 3 h and static culture for 3 h; Two-cycle: loading stretch for 3 h and static culture for 3 h following loading stretch for 3 h and static culture for after a two-cycle stretch (Figure 6L) and the protein expression of SMMHC, α-SMA, and SM22-α was similarly inhibited by GSK343 (Figure 6M-O).These results suggested that the intermittent stretch-induced SMC differentiation was mediated by mechanical memory encoded with H3K27me3 in MSCs.

| Intermittent stretch-induced memory could be transmitted based on direct cell-cell interaction
To investigate whether mechanical memory is communicated between cells, we directly mixed static MSCs with two-cycle stretched MSCs in different ratios (1:0, 1:1, 3:1, 9:1, and 0:1), and an indirect co-culture system in which these cells were cultured in Transwells in the ratios of 1:0, 1:1, and 0:1 (Figure 7A).Interestingly, the proliferation of static MSCs increased for all cell ratios by direct co-culture, but it was not affected by indirect coculture (Figure 7B,C).However, the markers of smooth muscle cells, α-SMA and SM22-α, were not pronouncedly affected by direct or indirect co-culture for 7 days (Figure 7D,E).We pre-labeled the two-cycle stretched MSCs (red) and static MSCs (green) with dyes for immunofluorescence staining.The mitochondrial length and network branches of static MSCs increased in all groups by direct co-culture, and the mitochondria also displayed (Figure However, the mitochondrial morphology of static MSCs was barely changed by indirect co-culture (Figure S7A-D).Furthermore, the cellular ATP levels also increased in all groups through direct co-culture of two-cycle stretched MSCs with static MSCs for 24 h (Figure 7J).The promoted metabolism of stretched cells could communicate with static cells to remodel their metabolic homeostasis.Collectively, these data indicated that intermittent stretch-induced metabolic memory could be transmitted based on direct cellcell interaction.

| DISCUSSION
Mechanical memory has been observed in various mechanical environments and cell lines, 5,8,37,38 yet the understanding of the key mechanisms that explain this biological phenomenon is limited.Our work demonstrates that MSCs exhibited mechanical memory in response to intermittent stretch, as evidenced by cell reorientation, alignment of F-actin, nuclear deformation, transcriptional activity, condensed chromatin, and active metabolism.Furthermore, we found that the mechanical memory associated with MSC differentiation into smooth muscle was encoded in chromatin remodeling involving H3K27me3.Moreover, the promoted metabolism of stretched cells communicated with static cells to remodel their metabolic homeostasis by direct cell-cell interaction.These findings deepened our fundamental understanding of cellular mechanotransduction and mechanical memory.
Physiological or pathological tensions are often simulated and used to regulate cell fate and behavior in tissue engineering.On exposure to the cyclic stretch of the underlying substrate, the cells reoriented themselves to an angle or perpendicular alignment to the direction of stretch. 29,39This alignment was maintained even after multiple cycles of stretch. 40Similarly, MSCs tended to be perpendicular to the uniaxial stretch direction and formed a cellular orientation memory through the intermittent stretch.The reorientation of cells was influenced by the amplitude and frequency of stretch, as well as by the boundary conditions of the substrate. 29,41Our results showed that the memory of MSC reorientation depended on the stretching amplitude.Specifically, the stretching amplitudes of 5% and 10% were more conducive to the formation of mechanical memory compared with the amplitude of 15%.The cellular reorientation memory was barely affected by the duration of the stretch, reaching saturation after two repetitive loadings.Intermittent stretch-induced cell alignment returned to random distribution after 24 h of relaxation.A recent study also reported that the cell alignment was partially recovered after 3 h of stretch followed by 2 h of relaxation in lung alveolar epithelial cells. 42These findings suggested that maintaining mechanical memory of cellular reorientation required repetitive stretching, which was maintained in the short term.This also prompted further exploration of factors critical for maintaining a more stable and persistent mechanical memory.
Furthermore, at the subcellular level, the polarization and reorientation of F-actin stress fibers generate internal contractile forces, which apparently preceded cell reorientation. 43Our data showed that F-actin stress fibers immediately reoriented and aligned themselves perpendicular to the direction of stretch after loading once, and this alignment was maintained during subsequent relaxion and stretching.The alignment of F-actin stress fibers was not affected by the amplitude and frequency of stretch.The dynamic behavior of F-actin stress fibers is mediated by the stability of FA. 44 The number of FAs decreased with intermittent stretch, indicating that they were in a dynamic assembly process to mediate the reconstruction of F-actin stress fibers.This suggested that F-actin reorientation and FA dynamics were more responsive to mechanical stimuli compared with cell reorientation.
Mechanical cues are transduced to the nucleus through the LINC complex, which connects the cytoskeleton and the nuclear membrane.The nucleus behaves as a viscoelastic material, implying that it can flow and undergo irreversible deformation when exposed to forces. 45In our study, nuclear deformation occurred after the initial stretch and persisted even during subsequent relaxation or stretching.This suggests that the recovery of nuclear morphology was relatively slow compared with cell reorientation, highlighting the viscoelastic characteristics of the nucleus.Furthermore, these findings indicated that the nucleus had a higher sensitivity to mechanical cues as a mechanosensory organelle. 17,46Emerging evidence suggests that nuclear deformation regulates various cellular and nuclear functions. 47These functions are thought to be downstream of cytoplasmic signaling pathways and determined by gene expression.Our findings indicated that repeated stretch loading promoted increased gene transcription, which remained highly even expressed after relaxation.This phenomenon, known as transcriptional memory, was primarily associated with histone methyltransferase and chromatin region.Increased extracellular stiffness promoted nuclear deformation, leading to chromatin condensation and enhanced histone deacetylase activity in fibroblasts. 48Moreover, repeated interferoninduced faster and greater transcription in fibroblasts, thus establishing transcriptional memory. 49A previous study reported that the transcriptional activity could be transmitted from mother to daughter cells to maintain multigenerational transcriptional 50 was also critical for preserving of mechanical memory.Furthermore, we observed that chromatin condensation was promptly enhanced following stretch loading with an increased H3K27me3-marked heterochromatin.This condensation persisted even after the force was withdrawn, which was consistent with the findings of a previous study. 51The stretch-induced nuclear flattening and chromatin condensation were impaired by the suppression of mechanotransduction on the LINC complex.This suggested that nuclear deformation mediated persistent chromatin condensation, which was maintained through epigenetic remodeling.These findings further supported that the nucleus was the central player in mechanical memory.
Likewise, the metabolic activity of MSCs was instantly elevated in response to stretch, resulting in a rapid increase in mitochondrial fusion and ATP production.The GO analysis revealed that both the second stretch and relaxation resulted in a higher expression of genes compared with the first stretch and relaxation, and these genes were related to mitochondrial and ATP metabolism.These results suggested that stretch-induced mechanical memory provoked continuous mitochondria elongation and ATP production, creating an active metabolic environment to coordinate the quick response to the reloading of stretch.The mitochondrial networks are permanently remodeled by replication activation, and the asymmetric segregation of mitochondria ensures cellular memory for hematopoietic stem cell replication. 52This supports our finding that the metabolic activity of mitochondria remained high even after the stretch was stopped.Replication and partitioning of the mitochondrial genome are necessary for functional mitochondrial inheritance, ensuring the transmission of complete mtDNA molecules to the next generation. 53These complementary findings suggest that metabolism is remodeled and maintained by intermittent stretch, contributing to the formation of mechanical memory.
Previous studies demonstrated that proper mechanical dosing induced mechanical memory to regulate MSC fate. 5,7,37Our study revealed that repetitive stretch stimulation promoted proliferation and SMC differentiation in MSCs.The nuclear deformation was preserved to maintain chromatin condensation accompanied by an increase in H3K27me3 during prolonged differentiation.A previous study reported that MSCs exhibited increased nuclear deformation and H3K27me3 on a higher-generation dendrimer surface, which induced them to switch to the muscle lineage. 54H3K27me3, an epigenetic mark, was transmitted to the next generation in proliferating cells. 55,56We found that the inhibition of H3K27me3 eliminated the memory of intermittent stretch-induced SMC differentiation in MSCs.The H3K27me3-marked chromatin was crucial in the early differentiation status of cardiomyocytes 57 or multipotent hematopoietic progenitors. 58In addition, a recent study found that chromatin remodeling, specifically the structural remodeling of H3K9me3, was associated with the mechanical memory-induced chondrocyte phenotype. 37hese findings suggested that epigenetic-modified chromatin was a critical mediator of mechanical memory for MSC differentiation.
Moreover, whether mechanical memory can be communicated between cells is currently unknown.In a previous study, myoblasts and MSCs were directly cocultured and subjected to simultaneous stretch, which enhanced the potential of MSCs for myogenic differentiation. 59We observed that the direct co-culture with stretched cells promoted the proliferation of static cells but had a minimal effect on the SMC differentiation of static cells.In addition, the proliferation and differentiation of static cells were not affected by indirect co-culture with stretched cells.Furthermore, static cells exhibited fused mitochondria and increased ATP production when directly co-cultured with stretched MSCs.A study indicated that the direct co-culture of MSCs and myofibroblasts contributed to protection against cardiac fibrosis, with their interaction occurring through physical contact and tubular structures. 60Increasing evidence shows that the therapeutic potential of MSCs is not only due to cell replacement and paracrine effects but also through the transfer of mitochondria into damaged tissues or cells to regulate cellular metabolism. 61The mitochondrial transfer can rescue the dysfunctional mitochondria in recipient cells and reprogram differentiated cells. 62,63he transfer of mitochondria between mammalian cells was first detected through tunneling nanotubes, 64 which showed promise in achieving energy synchronization in different cells and transmitting cellular mechanical memory.Nevertheless, further evidence is needed to confirm these findings.
In summary, we discovered that intermittent stretchinduced amplitude-dependent memory of cell reorientation.Cytoskeleton and nuclear remodeling mediate the intermittent stretch-induced mechanical memory.Chromatin condensation, transcriptional memory, and active metabolism were also preserved by intermittent stretch.Furthermore, we identified that H3K27me3 modification of chromatin condensation ultimately determined their smooth muscle differentiation in MSCs.We also observed the transmission of mechanical memory that accompanied active metabolism.Our findings suggested that harnessing optimal mechanical memory might be a valuable approach for achieving desired MSC phenotypes

F I G U R E 2
Intermittent stretch-induced mechanical memory via cytoskeleton and nuclear remodeling.(A-C) The angular distribution of F-actin stress fibers.Representative images (top row) and filament orientation (bottom row) of F-actin stress fibers in MSCs after loading intermittent stretch with amplitude of 10% at 0.5 Hz (3-h stretch period) (A), the method of calculation interquartile range (B), and quantification of F-actin stress fibers' interquartile range (C).(D-F) Representative images of paxillin (top) and binary images (bottom) in MSCs after loading intermittent stretch with amplitude of 10% at 0.5 Hz (D), quantification of FA number (E) per cell, and quantification of FA order parameter S = <cos2α> (F).(G-I) Representative images of the nucleus (top) and reconstructed DAPI images (bottom) after loading intermittent stretch with amplitude of 10% at 0.5 Hz (G), quantification of nuclear aspect ratio (NAR, H), and nuclear height (I).(J) Left, representative images of reconstructed DAPI images with knockdown of Nesprin-1 (siNesprin-1) in MSCs.Right, quantification of nuclear height.Data from at least three independent experiments.All graphs showed mean ± SEM.NS, not significant, *p < .05,**p < .01 or ***p < .001.F I G U R E 3 Transcriptional memory of MSCs is established by intermittent stretch.(A) The number of significant upregulated genes were derived from RNA-seq experiments in MSCs after loading intermittent stretch with an amplitude of 10% at 0.5 Hz (3-h stretch period).(B, C) A heatmap of transcriptional memory genes (B), GO categories analysis of transcriptional memory genes (C).(D-G) Representative GO analysis of upregulated genes in Stretch 1 versus Control (D), Relax 1 versus Stretch 1 (E), Stretch 2 versus Control (F), and Relax 2 versus Stretch 2 (G).

F I G U R E 4
Intermittent stretch induces chromatin condensation.(A) Left, representative images of the nucleus with DAPI nuclei in MSCs after loading intermittent stretch with amplitude of 10% at 0.5 Hz (3-h stretch period) (top row), and corresponding edge detection of chromatin condensation parameter (CCP, bottom row).Right, Quantification of CCP.(B) Left, representative images of H3K27me3.Right, Quantification of the H3K27me3 intensity.(C) Top, representative images of the nucleus with DAPI nuclei with knockdown of nesprin-1 (siNesprin-1) in MSCs and corresponding edge detection of CCP.Bottom, quantification of CCP.Data from at least three independent experiments.All graphs showed mean ± SEM.NS, not significant, *p < .05,**p < .01 or ***p < .001.F I G U R E 5 Metabolic activity of MSCs is increased and maintained by intermittent stretch.(A-D) Representative images of mitochondria (top) and classification of mitochondrial morphology (bottom) in MSCs after loading intermittent stretch with amplitude of 10% at 0.5 Hz (3-h stretch period) (A).Quantification of mitochondrial length (B) and mitochondrial branches (C), and punctate, intermediate, and filamentous mitochondria were analyzed (D).(E) Relative ATP levels of MSCs after loading intermittent stretch with amplitude of 10% at 0.5 Hz. (F-K) Heatmaps were derived from RNA-seq experiments in MSCs after loading intermittent stretch with an amplitude of 10% at 0.5 Hz.Glycolytic genes (F), tricarboxylic acid (TCA) cycle genes (G), and mitochondrial electron transport chain (ETC) genes (H-K).Data from at least three independent experiments.All graphs showed mean ± SEM.NS, not significant, *p < .05,**p < .01 or ***p < .001.

F I G U R E 7
Mechanical memory could be transmitted, and accompanied by activated metabolism.(A) Scheme of the direct and indirect co-culture.The static MSCs were direct or indirect co-cultured with MSCs that suffered to two-cycle stretch.(B, C) Quantification of the proliferation of cells in direct (B) and indirect (C) co-culture systems (Con: static MSCs, 2ST: two-cycle stretched MSCs).(D, E) Immunoblots of α-SMA and SM22-α and quantification of their expression on day 7 for direct (D) or indirect (E) co-culturing static MSCs with twocycle stretched MSCs.(F-I) The mitochondrial morphology was detected after direct co-culturing static MSCs with two-cycle stretched MSCs for 24 h.To differentiate between two types of cells (F), cells were labeled with green (Con) or red (2ST) Cell View stains (first row).Representative images of mitochondria (second row) and classification of mitochondrial morphology (third row).Quantification of mitochondrial length (G), mitochondrial branches (H), and classification of mitochondria (I).(J) Relative ATP levels for direct co-culturing static MSCs with two-cycle stretched MSCs for 24 h.Data from at least three independent experiments.All graphs showed mean ± SEM. *p < .05,**p < .01 or ***p < .001.| 15 of 19 NA et al.
Whether the data conformed to the normal distribution was assessed by the Shapiro-Wilk test.Whenever the data conformed to a normal distribution, a one-way analysis of variance (ANOVA) with the least significant difference (LSD) test was used.The data with non-normally distributed were analyzed with the Mann-Whitney U-test.Statistical significance was indicated by *p < .05,**p < .01 or ***p < .001.NS indicates no significant difference.
SPSS software (version 25.0) was used to evaluate statistical significance.The graphs were performed using GraphPad Prism (version 9.0.0).All experiments were performed in at least three independent experiments.All data represent mean ± standard error of the mean (SEM).