Journal Pre-proof Energy expenditure during cell spreading influences the cellular response to matrix stiffness

Cells respond to the mechanical properties of the extracellular matrix (ECM) through formation of focal adhesions (FAs), re-organization of the actin cytoskeleton and adjustment of cell contractility. These are energy-demanding processes, but a potential causality between mechanical cues (matrix stiffness) and cellular (energy) metabolism remains largely unexplored. Here, we culture human mesenchymal stem cells (hMSCs) on stiff (20 kPa) or soft (1 kPa) substrate and demonstrate that cytoskeletal reorganization and FA formation spreading on stiff substrates lead to a drop in intracellular ATP levels, correlates with the activation of AMP-activated protein kinase (AMPK). The resulting increase in ATP levels further facilitates cell spreading and reinforces cell tension of the steady state, and coincides with nuclear localization of YAP/TAZ and Runx2. While on soft substrates (1 kPa), lowered ATP levels limit these cellular mechanoresponses. Furthermore, genetic ablation of AMPK lowered cellular ATP levels on stiff substrate and strongly reduced responses to substrate stiffness. Together, these findings reveal a hitherto unidentified relationship between energy expenditure and the cellular mechanoresponse, and point to AMPK as a key mediator of stem cell fate in response to ECM mechanics. RhoA, β -integrin, myosin, pAMPK, YAP/TAZ and Runx2 staining, nonspecific binding sites were blocked in 10% BSA solution for 1 h, followed by incubation with primary antibody anti-vinculin (1:500, AB18058, from Abcam), anti-integrin β 1 (1:500, AB30394, from Abcam), anti-RhoA (1:500, AB187027, from Abcam), anti-myosin IIa (1:500, 150M4764, from Sigma), anti-AMPK alpha 1 (phospho T183) + AMPK alpha 2 (phospho T172) (1:500, AB23875, from Abcam), anti-YAP/TAZ (1;500, D24E4, from Cell Signaling) and anti-Runx2 (1:500, AB76956, from Abcam) for 1 h. Subsequently, stained with DAPI, phalloidin and secondary antibody, Alexa-488 goat anti-mouse (1:1000, A11029, from Thermo Fisher Scientific) or anti-rabbit IgG (1:1000, A10040, from Thermo Fisher Scientific) for 1 h at room temperature.

The physical properties of the extracellular matrix (ECM) have a profound impact on cell behavior and stem cell fate, and a direct relationship between matrix stiffness, cell spreading and lineage selection has been demonstrated [1,2,3,4,5,6]. Forces generated within the actin cytoskeleton and transmitted through FAs, play a major role in the cellular response to biophysical cues [7,8,9,10]. Much attention has been given to how cells respond to mechanical forces. However, less is known how these forces are integrated with other cellular processes including bioenergetics [11,12,13]. In order to understand how mechanical forces shape the cellular phenotype and regulate cell fate, it is necessary to consider the highly dynamic nature of the cellular response, which involves a hitherto overlooked large energy expenditure. Growing evidence indicates that the transduction of external mechanical forces is linked to metabolic signals such as neutral lipid synthesis [13], mitochondrial structural remodeling [14], cellular glucose uptake [11] and cell migration [15]. The cellular response is a well-tuned process that must balance energy supply with energy demand to allow proper actin polymerization, FA formation and cellular contractility buildup. But it remains unclear how cells respond to this energy expenditure, tune energy supply and demand, and maintain energy homeostasis for mechanotransduction.
AMPK, a well-characterized cellular energy sensor [16], plays a key role in the coordination of cell function by controlling intracellular ATP levels [17]. AMPK is Thr172-phosphorylated and thereby activated (yielding pAMPK) when the AMP/ATP ratio increases. This occurs for instance during starvation, hypoxia or cell detachment from the ECM [18,19,20]. Upon AMPK activation, energy homeostasis is restored by altering glucose, protein and lipid metabolism, paralleled by mitochondrial morphology changes [21]. Mechanical cues, chemical stimulations or metabolic stresses [22,23,24] can all lead to alterations in mitochondrial morphology, and AMPK is considered an important regulator of mitochondrial dynamics through phosphorylation of mitochondrial fission factor (Mff), thereby stimulating mitochondrial fission [24,25]. A recent study in MCF10A (human breast epithelial) cells and MDCK II (canine kidney epithelial) cells, revealed that external forces from neighboring cells triggered AMPK activation via E-cadherin to increase cellular ATP levels and enhance actomyosin contractility, thus linking energy homeostasis with cell-cell adhesion mechanotransduction [11]. Actomyosin contractility also plays a major role in the cellular response to the physical properties of the ECM through FA formation and organization of the actin cytoskeleton, which are ATP-demanding processes [11,12,26].

J o u r n a l P r e -p r o o f
In this study, we aim to elucidate the role of energy expenditure in the cellular response to substrate stiffness. When cells were seeded on stiff substrates, we observed an initial drop in ATP levels followed by higher ATP levels at 24 hrs, increased glucose uptake and actomyosin contractility, altered mitochondrial morphology, and sustained AMPK activation, resulting in nuclear localization of YAP/TAZ and Runx2, and osteogenic differentiation. Independent of substrate stiffness, cell fate appeared strongly correlated with the activation or inhibition of AMPK.
Our findings establish a critical role for AMPK in connecting intracellular energy expenditure and cellular mechanotransduction, with downstream effects on stem cell fate.

Materials and Methods
Preparation of polyacrylamide (PAAm) hydrogels. This method was adapted from a previously described procedure [4,5]. 13-mm-diameter borosilicate glass coverslips (VWR) were freshly oxidized by oxygen plasma and then incubated in a solution composed of 0.3% w/v 3-(trimethoxysilyl)propyl methacrylate (Sigma Aldrich) and toluene (Fisher Scientific) overnight.
Slides were washed with ethanol and dried with nitrogen. The stiffness of PAAm gels was regulated through the concentration of acrylamide (AA) and bis-acrylamide (BA). PAAm gel solutions were prepared with AA at final concentrations of 8, 20 and 30% w/v and BA at 0.02, 0.15 and 0.375% w/v. 5 µL of 10% w/v ammonium persulfate (Sigma Aldrich) and 1.5 µL TEMED (Sigma Aldrich) were added to AA/BA solutions to initiate PAAm polymerization. 4 µL of PAAm solution was immediately pipetted onto the pre-treated 13-mm-diamter coverslips and covered with an untreated 20-mm-diameter coverslip. The polymerization was completed in about 2 h, and then the samples were immersed in PBS buffer overnight. The top coverslips were carefully peeled off to obtain the gels adhering to the bottom coverslips. With the different concentrations of AA/BA described above, PAAm gels with different stiffness can be prepared. 8% w/v AA and 0.02% w/v BA, 20% w/v AA and 0.15% w/v BA, 30% w/v AA and 0.375% w/v BA showed stiffness of ~ 1 kPa, 20 kPa and 100 kPa, respectively.  Microscopy image analysis. Fluorescence images were acquired using a Leica SP8 confocal microscope with different objectives and overlaid in Fiji software with Image 5D plugin. All analysis of cells was based on single cells not in contact with other cells. For quantification of fluorescence intensity, images were taken with photon counting mode to make sure all settings were fixed. For quantification of G/F actin ratio, it was performed as our previous studies [28,29].
The integrated fluorescence of F-actin and G-actin z-stack images was measured for individual cells, followed by subtraction of the background fluorescence. For quantification of YAP/TAZ and Runx2 subcellular localization, nuclear localization was defined as a ratio of the average intensity of target proteins in the nucleus compared to the average intensity of target proteins in the cytoplasm was larger than 1.

Mitochondrial morphology was quantified in cells stained with MitoTracker Deep Red using
Image Pro Plus® software (Media Cybernetics, Rockville, MD, USA) and an algorithm described in detail previously [30]. In brief, microscopy images were converted to 8-bit grayscale and Data availability. The authors declare that the data is presented within the paper and its Supplementary Information or will be made available by the corresponding author upon reasonable request.

Feedback between spreading and intracellular energy expenditure
We first determined the influence of substrate stiffness on indicators of energy expenditure such as intracellular ATP levels and glucose uptake rates. We cultured hMSCs on 1 kPa (soft) or 20 kPa Removal of glucose from the culture medium resulted in significantly reduced actin polymerization and FA formation, reflected by a 3.8-fold increase in G/F-actin levels (Figure 1f) after 20 h culture. We further studied the temporal characteristics of intracellular ATP changes upon disruption and re-organization of the actin cytoskeleton. To do this, cells were treated for 1 h with Cytochalasin D (CytoD) [8], which disrupts the actin cytoskeleton, and resulted in cells displaying a small rounded morphology (Figure 1g). After CytoD removal from the medium, cells were allowed to re-spread on the substrate, during which intracellular ATP levels dropped by ~22% (Figure 1g)

AMPK activation correlates with altered mitochondrial morphologies
In stem cells, mitochondria are prime generators of ATP [14,24], these organelles are motile, and continuously fuse and divide, a process influenced by mechanical cues [14,22]. In addition to its role in ATP production, mitochondrial morphology has been coupled to the coordination of self-renewal vs. differentiation of stem cells [33]. Primed by these findings we next determined whether substrate stiffness affected mitochondrial morphology. Visual inspection suggested that, on soft substrates, mitochondria display a filamentous structure at both 5 h and 20 h after seeding.

J o u r n a l P r e -p r o o f
In contrast, mitochondrial morphology appeared more fragmented on stiff substrates (Figure 2a and Supplementary Figure 5). Quantitative analysis [30] revealed that the mitochondrial area (Am, a measure of mitochondrial size) and form factor (F, a combined measure of mitochondrial length and degree of branching), were significantly reduced on stiff relative to soft substrates (Figure 2b, c). Mitochondria in cells on stiff substrates displayed a stronger staining with the fluorescent cation tetramethylrhodamine methyl ester (TMRM), (Supplementary Fig. 6), suggesting that the mitochondrial membrane potential of these mitochondria is more negative.

J o u r n a l P r e -p r o o f
It has been shown that AMPK mediates mitochondrial fragmentation in response to energy stress [24]. In this context, we hypothesized that the observed drop in ATP levels during cell spreading would result in a similar energy stress, leading to AMPK activation and induction of a fragmented mitochondrial phenotype. Western blot analysis revealed increased pAMPK/AMPK ratios in hMSCs cultured for 5 and 20 h on stiff vs. soft substrates (Figure 2d). To gain insight into the chain of events during AMPK phosphorylation we performed western blot and immunostaining analysis at different time points (Supplementary Figure 8). On stiff substrates, virtually no pAMPK was observed during initial cellular adhesion period (0.5 h) (Supplementary Figure 8).
Cells cultured on these substrates for longer than 1 h displayed elevated pAMPK levels as well as nuclear localization. Together, these experiments establish that AMPK phosphorylation is preceded by a drop in ATP levels (Figure 1a and Supplementary Figure 8). Furthermore, a fragmented mitochondrial phenotype was observed 1 h after seeding (Supplementary Figure 9), in agreement with the time point at which increased pAMPK levels were observed.

AMPK plays a central role in mechanotransduction
Previous studies [11]

Manipulating intracellular AMPK activation alters cellular mechanoresponses
To further dissect the role of AMPK, we modulated its phosphorylation status using a cell-permeable AMPK inhibitor (Compound C) or AMPK activator (A-769662) at different concentrations after 30 min of cell spreading on stiff or soft substrates, respectively, and then evaluated changes in cellular response after culturing cells for 20h. It was found that increasing concentrations of AMPK inhibitor resulted in ~60% lower ATP levels on stiff substrates, while addition of AMPK activator on soft substrates led to 20% higher ATP levels (Figure 4a). The inhibition or activation of AMPK also impacted on glucose uptake, inducing a small (~13%) decrease after inhibition of AMPK on stiff substrates, and a 1.2~fold increase after activating AMPK on soft substrates (Figure 4b). Inhibition of AMPK on a stiff substrate induced a shift in mitochondrial morphology from a fragmented into a more filamentous phenotype. However, activation of AMPK on a soft substrate yielded a fragmented mitochondrial phenotype, previously associated with the cells cultured on stiff substrates (Figure 4c and Supplementary Figure 10).
These experiments support a mechanism in which alterations in mitochondrial morphology are evoked by changes in AMPK activity. We next investigated whether inhibition or activation of AMPK was sufficient in regulating FA formation, cytoskeletal organization and YAP/TAZ localization. Inhibition of AMPK clearly influenced FAs, which appeared smaller and confined to the edge of spreading cells (Figure 4d). Furthermore, cytoskeletal organization was less developed, with fewer cross-cell stress fibers with increasing concentrations of AMPK inhibitor (Figure 4d). YAP/TAZ was increasingly localized in the cytoplasm after inhibition of AMPK (nuclear localization declined five-fold at 20 µM inhibitor concentration), while nuclear YAP/TAZ localization in cells on soft substrates increased upon AMPK activation (Figure 4d).
Furthermore, we modulated energy substrate supply by varying the extracellular glucose concentration. YAP/TAZ nuclear localization only became pronounced at glucose concentrations >2.5 g/L and during glucose starvation YAP/TAZ nuclear localization was as low as 21±9% (Supplementary Figure 11). Nuclear YAP/TAZ localization induced by ECM stiffness requires stress fibers and cytoskeletal tension which can stretch nuclear pores and promote J o u r n a l P r e -p r o o f YAP/TAZ translocation. There is a balance of forces between cell adhesion on the outside and myosin II-based contractility on the inside of the cell, and myosin II thus plays an important role in controlling the mechanoresponse [36]. Inhibition or activation of AMPK clearly alters myosin levels (Figure 4e) and the changes mirror trends in YAP/TAZ localization.
J o u r n a l P r e -p r o o f  where cells have less pronounced stress fibers and associated energy demands, and this might explain why the modulation of AMPK does not lead to significant differences.

Discussion
Adhering and spreading cells form focal adhesions and a network of dynamic actomyosin stress fibers that maintains a certain tensional homeostasis adapted to the mechanical properties of the ECM [1,5,39]. These processes require energy, which raises the question of how cells ensure the availability of this energy and whether high energy expenditure limits cellular responses. An J o u r n a l P r e -p r o o f elevated ATP : ADP ratio due to F-actin formation [40], cytoskeleton remodelling [41], and adhesion mediated contractility [42] has been reported to increase glucose uptake during cell migration [43], the relationship between the cell dynamics and increased energy utilization. In this study we observed a clear link between mechanotransduction and the energy expenditure needed for the formation of well-developed FAs, cytoskeletal reorganization into well-defined stress fibers in spread cells, as well as the contractility of the cytoskeleton. Intracellular ATP levels are strongly affected by these energetically costly processes. In order to look into details about the extent of ATP consumption used for actin polymerization and contractility, we selectively inhibited actin polymerization by CytoD and actomyosin by Bleb. and revealed that inhibition of actin, but not myosin, restored intracellular ATP levels at early time points. However, at steady state (after cell area was at equilibrium, or better: steady state), inhibition of myosin, but not actin, restored intracellular ATP levels. Those experiments suggested that cells consume ATP to organize cytoskeleton at early times during spreading, to maintain tension at later time points (at steady state) and approximately 20% of cellular ATP is consumed by these processes.
A further link between energy metabolism and development of the actin cytoskeleton is highlighted by glucose starvation experiments. In the latter, glucose depletion significantly reduces actin polymerization and FAs formation, confirming that cellular mechano-responses consume a significant fraction of intracellular ATP. We demonstrate (Figure 1) that cell spreading on stiff substrates is associated with an initial drop in ATP levels. This drop activates AMPK, a kinase that is triggered by changing AMP/ATP ratios that arise from, for example, starvation, hypoxia or cell detachment from the matrix [18][19][20]. Recent studies also provided other AMPK activation pathways in various cell types, including Ca 2+ influx [44,45] or cell-cell forces [11].
Here we show that AMPK is activated coinciding with the ATP drop and mitochondrial fragmentation ( Figure 2). As reported previously, mitochondrial trafficking to the leading edge of the cell could support the formation of protrusions and focal adhesions, necessary for cell migration and force generation [15,46,47]. Here, we showed that mitochondrial morphology was highly responsive to AMPK activation which is in full agreement with previous studies [24]. This strongly suggests that AMPK activation or inhibition directly affects mitochondrial morphology, indicating that changes in mitochondrial morphology were coupled to energy metabolism. Given the fact that AMPK plays a central role in the coordination of cell metabolism and function [17], its activation likely impacts on nuclear localization of YAP/TAZ. In this context, previous reports J o u r n a l P r e -p r o o f demonstrated that AMPK lowers YAP activity via phosphorylation [20], evoking the cytoplasmic localization of YAP/TAZ. Conversely, we find that YAP/TAZ is predominantly localized in the nucleus (Figure 3f), as typically observed in cells spreading on stiff substrates [3,32]. It should be noted that in our study, cells were seeded at very low density to avoid cell-cell interactions, whereas previous work has shown that cell-cell contacts play a role in AMPK activation [11].
The main finding of our study is that maintaining a high cellular contractility is regulated (indirectly) by AMPK as high ATP levels are necessary to support actin polymerization, focal adhesions formation and actomyosin contractility. These processes set up a feedback loop that couples into YAP/TAZ localization. We present three different lines of evidence for placing YAP/TAZ translocation mechanistically downstream of AMPK activation: i) YAP/TAZ increasingly localized in the cytoplasm with declining ATP levels after inhibition of AMPK by Compound C treatment (Figure 4d). ii) AMPK activation was altered by changing the glucose concentration in the culture medium, and we found that YAP/TAZ nuclear localization only became pronounced at glucose concentrations >2.5 g/L, whereas glucose starvation resulted in YAP/TAZ nuclear localization as low as 21% (Figure 4d). iii) We found that ATP levels and glucose uptake were greatly reduced in a AMPKα-null mouse embryonic fibroblast (MEF) cell line, and YAP/TAZ localization in the nucleus on stiff substrates was also significantly reduced (by 32%). All these results firmly confirmed that AMPK activation can induce YAP/TAZ nuclear localization. We further observed that osteogenic differentiation appears correlated with activation or inhibition of AMPK, and not with the substrate stiffness.
In summary, our findings indicate that AMPK-mediated energy regulation plays an important role in the cellular mechanoresponse, as shown schematically in Figure 5d. The mechanoresponse on stiff substrates is initiated by cell spreading and the concomitant consumption of ATP to establish FAs and remodel the actomyosin network. Soon after initiation of cell spreading (i.e. in the first 0.5 h) ATP expenditure exceeds production, and the drop in intracellular ATP levels triggers activation of AMPK. This is consistent with recent evidence demonstrating that cells adapted their metabolic activity to variable mechanical cues through stress fiber formation and F-actin bundling [48]. After activation of AMPK, we observe a restoration of ATP levels, and ultimately (after 24h) even higher levels than before spreading. At the same time, we observe significant changes in mitochondrial morphologies pointing at increased ATP production. We found that upregulation of ATP production provides the required energy for further spreading and increased cell tension as J o u r n a l P r e -p r o o f the actin cytoskeleton organizes and FAs form, leading to a reinforcing cycle. Those experiments suggest that cells consume ATP to polymerize actin at early time points, and to increase cell tension at later time points. With the increased cell tension, cells that spread on stiff substrates then showed the expected YAP/TAZ nuclear localization and, ultimately, osteogenic differentiation.

Conclusions
Taken together, this work establishes temporal changes in intracellular ATP levels during cell spreading as a new mechanistic link between mechanical forces and molecular responses. In this sense, energy expenditure couples mechanical cues with cellular metabolism via AMPK activation in response to lowered ATP levels. These findings provide an impetus for further studies into the mechanisms that guide energy metabolic remodeling in cells with increased energy demands during their response to environmental cues. Of particular interest in this context is how the remodeled energy metabolism in tumor cells shapes their mechanoresponse to ECM alterations.