Matrix stiffening promotes perinuclear clustering of mitochondria

Abstract

Mechanical cues from the tissue microenvironment, such as the stiffness of the extracellular matrix, modulate cellular forms and functions. As numerous studies have shown, this modulation depends on the stiffness-dependent remodeling of cytoskeletal elements. In contrast, very little is known about how the intracellular organelles such as mitochondria respond to matrix stiffness and whether their form, function, and localization change accordingly. Here, we performed an extensive quantitative characterization of mitochondrial morphology, subcellular localization, dynamics, and membrane tension on soft and stiff matrices. This characterization revealed that while matrix stiffness affected all these aspects, matrix stiffening most distinctively led to an increased perinuclear clustering of mitochondria. Subsequently, we could identify the matrix stiffness-sensitive perinuclear localization of filamin as the key factor dictating this perinuclear clustering. The perinuclear and peripheral mitochondrial populations differed in their motility on soft matrix but surprisingly they did not show any difference on stiff matrix. Finally, perinuclear mitochondrial clustering appeared to be crucial for the nuclear localization of RUNX2 and hence for priming human mesenchymal stem cells towards osteogenesis on a stiff matrix. Taken together, we elucidate a dependence of mitochondrial localization on matrix stiffness, which possibly enables a cell to adapt to its microenvironment.

• The tissue microenvironment stiffness is known to influence cellular behavior, but the role of mitochondria in this process is not well understood.

• This study provides novel insights into this phenomenon by demonstrating that matrix stiffness induces perinuclear clustering of mitochondria, regulated by filamin.

• Understanding how matrix stiffness affects mitochondrial dynamics is crucial for tissue engineering and regenerative medicine. It sheds light on cellular adaptation mechanisms and offers potential avenues for therapeutic intervention in diseases characterized by abnormal tissue stiffness, such as fibrosis and cancer.

INTRODUCTION

In biological tissues, there exists a delicate synergy between the extracellular matrix (ECM) and the cells embedded in it. The cells in our body constantly change their internal organization and adapt to various signals emanating from the extracellular environment. Mechanical cues from the ECM elicit an intracellular tensional response that ensures tissue homeostasis. This response manifests itself in terms of cytoskeletal remodeling, altering actomyosin contractility, driving nuclear reprogramming (Dupont et al., 2011), and metabolic adaptations (Park et al., 2020; Romani et al., 2021). Disruptions in these mechanotransduction pathways have implications for cancer progression, aging, wound healing, morphogenesis, stem cell differentiation, fibrosis, and cardiovascular diseases (Engler et al., 2006; Butcher et al., 2009; Jaalouk and Lammerding, 2009; Mammoto and Ingber, 2010; Schwartz, 2010). Till now, in the context of cellular response to mechanical cues, several studies have dissected out the stiffness-sensing mechanisms with a primary focus on the changes in the cytoskeletal elements. The cytoskeletal network, being the primary load-bearing element, acts as a sink for the extracellular forces encountered by tissues. Considering that the intracellular environment is a complex web of proteins and structures where the cytoskeleton interacts with almost every organelle, cytoskeleton-generated, and transmitted forces should distort organelles leading to changes in their structure and function. However, the participation of various organelles in cellular mechanoresponse remains mostly poorly explored.

Among the organelles, the mitochondrion stands out not only due to its crucial role in energy production, but also because of its interactions with the cytoskeleton. Importantly, mitochondria are central players in the complex metabolic processes which have the potential to drive the different layers of mechanotransduction. Mitochondria are extensively chained to other organelles and cytoskeletal entities and are therefore bound to be under tension. Over the years, the idea of mitochondria as simply the “powerhouses” has expanded, and it is now clear that mitochondria act as signaling hubs, buffer calcium, aid in apoptosis, and are sites of reactive oxygen species (ROS) signaling and metabolite generation (McBride et al., 2006; Nunnari and Suomalainen, 2012; Spinelli and Haigis, 2018). Moreover, our understanding of mitochondrial form has evolved dramatically over the last few decades. With improved imaging tools, it is now appreciated that mitochondria are not isolated or static. Rather they are highly dynamic and exhibit many morphologies ranging from long interconnected tubules to fragmented and swollen blobs (Liesa et al., 2009). Mitochondrial dynamics, form, function, and localization are not only interlinked but are also governed by the cellular architecture (Moore and Holzbaur, 2018; Shah et al., 2021). In fact, changes in mitochondrial form or localization during biochemical stimulations, inflammations, cell division, or in highly active cells like neurons, muscles, and secretory cells are well characterized (Chang et al., 2006; Motori et al., 2013; Mishra and Chan, 2014; Wikstrom et al., 2014; Misgeld and Schwarz, 2017). However, the variability in the dynamics and function of mitochondria in response to mechanical cues, especially from the stiffness of the ECM has only now emerged to the forefront and warrants active exploration.

Understanding the interplay between matrix stiffening and mitochondrial form and function is essential for the following reasons. First, there are emerging discoveries highlighting the mechanosensitivity of mitochondria to mechanical forces (Helle et al., 2017; Bartolák-Suki and Suki, 2020; Mahecic et al., 2021). In the tissue microenvironment, matrix stiffness critically influences these forces. Second, both mitochondrial dysregulation and ECM stiffening are hallmarks of cancer and aging (Balaban et al., 2005; Butcher et al., 2009; Wallace, 2012). Finally, there exists a reciprocal relationship between cell mechanics and metabolism (Romani et al., 2021). Consequently, some studies have investigated the effect of matrix stiffness. For example, a recent study involving untransformed cells showed an increase in fragmented mitochondria on a stiff matrix (Tharp et al., 2021). On the contrary, in a study involving transformed cells, a similar phenotype was seen on soft matrix (Romani et al., 2022). Although the mitochondrial form is contradictory, both studies highlight the effect of ECM stiffness on redox homeostasis and oxidant stress responses. However, a complete picture of mitochondrial morphology, subcellular localization, and function upon changes in matrix stiffness remains elusive. Here, we discover matrix stiffness as an important microenvironmental factor that distinctly decides the subcellular localization of mitochondria, in addition to its effect on morphology, motility, and membrane tension.

RESULTS

Matrix stiffness alters morphology and subcellular localization of mitochondria

Cells exert contractile forces onto their substrate and subsequently deform the surrounding ECM. We tested the mechanosensing capabilities of breast cancer derived MCF-7 cells by culturing them on collagen I-coated polyacrylamide hydrogels of elastic modulus 4 and 90 kPa, mimicking soft and stiff tissue conditions, respectively (Engler et al., 2006). MCF-7 cells extended protrusions on soft and stiff ECM indicating mechanical engagement on either stiffness condition. Furthermore, these cells showed distinct actin stress fibers on stiff ECM, indicating mechanosensing of ECM stiffness. Hence, we proceeded to test the effect of matrix stiffness on mitochondrial morphology and localization. We first cultured MCF-7 cells for 24 h on collagen I-coated polyacrylamide hydrogels of elastic modulus 4 and 90 kPa. We then incubated these cells with a mitochondrial potential-dependent dye Mitotracker Green, and imaged mitochondria using live confocal microscopy and super-resolution microscopy. We segmented three-dimensional confocal images of mitochondria and quantified sphericity, which is a morphology descriptor as well as a few parameters related to mitochondrial networks, including branches per mitochondria, total branch length per mitochondria, branch junctions per mitochondria, and branch endpoints per mitochondria. On soft ECM, the number of branches, branch endpoints, and branch junctions per mitochondria were higher indicating highly networked mitochondrial structures as compared with distinct mitochondria on stiff ECM (Figure 1A, Supplemental Figure 1, A and B). On stiff ECM, mitochondrial sphericity was higher indicating highly fragmented and globular mitochondria as compared with the elongated mitochondria on soft ECM (Figure 1, A and B; Supplemental Figure 1A). Super-resolution images also showed highly networked mitochondria on soft ECM and toroidal mitochondria on stiff ECM (Figure 1C). To rule out any artifact caused by live imaging and Mitotracker Green, we fixed MCF-7 cells cultured on soft and stiff ECM, stained them for the alpha subunit of mitochondrial membrane ATP Synthase or Complex V, and imaged mitochondria using confocal microscopy. Again, mitochondria on soft ECM appeared elongated and highly networked whereas mitochondria on stiff ECM appeared fragmented and toroid-shaped (Figure 1D; control points in Supplemental Figure 1F). These consistent results showed that both live cell confocal microscopy with mitochondrial dyes (Mitotracker Green) and fixed cell confocal microscopy of immunostained mitochondria (stained with Complex V) did not affect the observed differences in mitochondrial morphology and network properties on soft and stiff ECM. Along with the stiffness-dependent changes in the mitochondrial morphology, we also observed distinct mitochondrial localization patterns. Mitochondria on soft ECM were homogeneously distributed across the cell area, whereas mitochondria on stiff ECM were predominantly clustered near the nucleus (the perinuclear [PN] space; Figure 1, A and D). Because some cells are reported to spread more on stiff ECM (Engler et al., 2006), we decided to check whether the spreading of MCF-7 cells was sensitive to ECM stiffness cues. MCF-7 cells extended protrusions on soft and stiff ECM indicating mechanical engagement on either stiffness condition (4 and 90 kPa). However, these cells did not show any significant change in cell area when cultured on 4 and 90 kPa substrates, respectively (Supplemental Figure 1C), corroborating previous reports (Gil-Redondo et al., 2022). Therefore, in the background of comparable cell spreading on either stiffness, we further checked whether mitochondrial localization was dependent on cytoplasmic distribution. To this end, we plotted the distribution of cytoplasm and mitochondria (localization distribution function) as a function of distance from the nucleus and further quantified the percent of mitochondria and cytoplasm localized in a defined PN radius (5 µm). The distribution of the cytoplasm as well as the PN fraction of cytoplasm was similar on either stiffness condition (Supplemental Figure 1C). However, the mitochondrial distribution showed a distinct peak in the PN zone on stiff ECM, representing the PN clustering of mitochondria (Figure 1, E and F), confirming that mitochondrial localization was not dependent on cytoplasmic distribution. Taken together, these results showed that matrix stiffening creates distinct mitochondrial populations – homogeneously distributed, elongated, and networked tubules on soft ECM and fragmented PN clusters on stiff ECM. However, the effect of matrix stiffness on mitochondrial localization appeared more distinct than on mitochondrial morphology (Figure 1F). On stiff ECM, mitochondria appeared to be distinctly clustered around the nucleus, and a major region of the cytoplasm lacked mitochondria (Figure 1E). Hence, subsequently we tested whether stiffness-sensitive mitochondrial localization depended on integrin signaling (Figure 1, G and H), probed the motility of PN versus peripheral (PL) mitochondrial populations (Figure 2), asked what caused the differential localization (Figure 3), measured the mitochondrial membrane tension (Figure 4), and finally, elucidated the physiological relevance of PN clustering of mitochondria on stiff ECM to stiffness-mediated change in the stem-cell fate (Figure 5).

FIGURE 1:

Matrix stiffness alters morphology and cellular localization of mitochondria. (A) Three-dimensional stacked confocal images of MCF-7 cells cultured on soft (top panel) and stiff (bottom panel) ECM showing mitochondria stained with Mitotracker Green. From left to right: merged images showing nuclei in cyan and mitochondria in yellow followed by a magnified view of mitochondria in the red box (Scale bar, 2 μm). (B) Scatter dot plot showing weighted sphericity of mitochondria per cell cultured on soft (4 kPa) and stiff (90 kPa) ECM. n = 29 FOVs (4 kPa), 27 FOVs (90 kPa). Each FOV (field of view) contains 10–15 cells. (C) Representative super-resolution images of MCF-7 cells cultured on soft (top panel) and stiff (bottom panel) ECM showing mitochondria (yellow) stained with Mitotracker Green. (D) Immunofluorescence confocal images of MCF-7 cells cultured on soft (top panel) and stiff (bottom panel) ECM showing nuclei stained with DAPI and mitochondria stained with ATP5α. From left to right: merged images showing nuclei in cyan and mitochondria in yellow followed by a magnified view of mitochondria in the red box (Scale bar, 2 μm). (E) Representative binarized images of cell body, nucleus, and mitochondria on soft ECM (top panel) and stiff ECM (bottom panel). From left to right: binarized images showing area of a single cell (in black) followed by nuclear area (in black) and mitochondria positive pixels (in white), graphs showing distribution of cytoplasm (blue) and mitochondria (orange) plotted as a function of distance from the nucleus (localization distribution function). The nuclear, PN, and cytoplasmic zones are indicated on the graphs. (F) Scatter dot plot showing fraction of mitochondria present in the PN zone in MCF-7 cells cultured on soft (4 kPa) ECM and stiff (90 kPa) ECM. n = 47 (4 kPa), 46 (90 kPa) cells. (G) Three-dimensional stacked confocal images of MCF-7 cells cultured on soft (top panel) and stiff (bottom panel) ECM under control (untransfected) and transfected conditions showing mitochondria stained with ATP5α. From left to right: merged images of control (untransfected) MCF-7 cells showing nuclei in cyan and mitochondria in yellow followed by mitochondria in grayscale highlighting differences in localization (also shown in Supplemental Figure 2A), merged images of MCF-7 cells transfected with Vcl-T12 (magenta) showing nuclei in cyan and mitochondria in yellow followed by mitochondria in grayscale showing PN clustering (also shown in Supplemental Figure 2B), merged images of MCF-7 cells transfected with Vcl-T (magenta) showing nuclei in cyan and mitochondria in yellow followed by mitochondria in grayscale showing homogenous distribution (also shown in Supplemental Figure 2C).(H) Scatter dot plot showing fraction of mitochondria present in the PN zone in control (untransfected), Vcl-T12, and Vcl-T transfected cells on soft (4 kPa) ECM and stiff (90 kPa) ECM. n = 47 (Control, 4 kPa), 41 (Vcl-T12, 4 kPa), 40 (Vcl-T, 4 kPa) 46 (Control, 90 kPa), 39 (Vcl-T12, 90 kPa), and 34 (Vcl-T, 90 kPa) cells. Statistical significance was calculated using two-tailed Mann-Whitney test (B) or unpaired t test with Welch’s correction (two-tailed) (F) or Brown- Forsythe Anova test with Welch’s correction (H). Data are given as median and interquartile range (B) or mean ± s.e.m (F and H) taken over three independent replicates. P values are shown in the graphs. White lines denote the cell boundary in (A, D, and G) and cyan ovals depict cell nuclei in (A and G), both drawn manually using DIC images (for wild-type /control cells) or GFP fluorescence (for transfected cells) that were acquired simultaneously with other fluorescence images. Scale bar,10 μm. Soft ECM: 4 kPa; Stiff ECM: 90 kPa.

FIGURE 2:

Matrix stiffness generates distinctly dynamic mitochondrial populations. (A) Montage showing mitochondrial motility at PN and PL regions of MitoDendra2 expressing MCF-7 cells cultured on soft ECM. From left to right: White boxes represent ROIs where Dendra2 expressing mitochondria were photoconverted (t = 00:00) and subsequently the diffusion of the photoconverted molecules (shown in cyan hot) was tracked for 5.5 mins (also see Supplemental Video 1; Supplemental Figure 3C); a spatial motility map showing distance traveled by individual mitochondria after photoconversion (yellow boxes represent ROIs where Dendra2 mitochondria were photoconverted). (B) Scatter dot plots showing motility metrics: distance (μm), speed (μm s^–1), displacement (μm), and velocity (μm s^–1) of photo-converted mitochondria in PN and PL regions of MitoDendra2 MCF-7 cells cultured on soft and stiff ECM. n = 546 (soft PN), 594 (soft PL), 469 (stiff PN), and 675 (Stiff PL) distinct mitochondrial tracks. Statistical significance was calculated using Kruskal-Wallis nonparametric test with uncorrected Dunn’s test. Data are given as median and interquartile range taken over three independent replicates. P values are shown in the graphs. The Y-axes of plots showing distance, speed, displacement, and velocity are on the “log 10” scale. (C) Montage showing fluorescence recovery after photobleaching in PN and PL regions of MCF-7 cells transfected with mCh-KIFC1* and Tom20-GFP and cultured on soft ECM without Biotin (top panel) and with Biotin (bottom panel). White circles represent photo-bleached regions. From left to right: GFP Fluorescence in PN and PL ROIs (white circles) captured at prebleach time point (t = 00:00) followed by bleaching (t = 00:04) and subsequent recovery tracked for 1 min 40 s, FRAP curves showing fluorescence recovery of Tom20-GFP at the indicated ROIs (also see Supplemental Videos 3 and 4). ROI: region of interest. Scale bar, 10 μm. Soft ECM: 4 kPa; Stiff ECM: 90 kPa.

FIGURE 3:

PN accumulation of filamin alters mitochondrial localization. (A) Immunofluorescence images of MCF-7 cells stained for filamin cultured on soft (top panel) and stiff (bottom panel) ECM. Arrowheads show PN enrichment of filamin on stiff ECM. (B) Scatter dot plots depicting total filamin protein levels (total intensity) and fraction of filamin in the PN region (PN filamin %) on soft (4 kPa) and stiff (90 kPa) ECM. n = 102 (4 kPa), 107 (90 kPa) cells. (C) Top panel: Confocal images of MCF-7 cells transiently expressing mAppFLNA cultured on soft ECM followed by a magnified view of the white box showing PN enrichment of filamin (in magenta). Bottom panel: Confocal images of MCF-7 cells transiently expressing dnFLNA cultured on stiff ECM followed by a magnified view of the white box showing reduced PN localization of endogenous filamin (in magenta). N: nucleus. (D) Three-dimensional stacked confocal images of MCF-7 cells cultured on soft (top panel) and stiff (bottom panel) ECM under control (untransfected) and transfected conditions showing mitochondria stained with Mitotracker Green. From left to right: merged images of MCF-7 control (untransfected) cells showing nuclei in cyan and mitochondria in yellow followed by mitochondria in grayscale highlighting differences in localization, merged images of MCF-7 cells transfected with mAppFLNA (magenta) showing nuclei in cyan and mitochondria in yellow followed by mitochondria in grayscale showing PN clustering, merged images of MCF-7 cells transfected with dnFLNA (magenta) showing nuclei in cyan and mitochondria in yellow followed by mitochondria in grayscale showing homogenous distribution. White lines denote the cell boundary and cyan ovals depict cell nuclei, both drawn manually using DIC images (for control cells) or mApple fluorescence (for transfected cells) acquired simultaneously with other fluorescence images. (E) Graphs showing distribution of cytoplasm (in blue) and mitochondria (in orange) in cells shown in (D) plotted as a function of distance from the nucleus. From left to right: control cells, mAppFLNA transfected cells, and dnFLNA transfected cells cultured on soft ECM (top panels) and stiff ECM (bottom panels). The nuclear, PN, and cytoplasmic zones are indicated on the graphs. The respective binarized images are shown in Supplemental Figure 5, C–E. (F) Scatter dot plots showing fraction of mitochondria present in the PN zone in untransfected (Control) and mAppFLNA (left graph) or dnFLNA (right graph) transfected cells cultured on soft (4 kPa) and stiff (90 kPa) ECM. n = 47 (control, 4 kPa), 46 (control, 90 kPa), 42 (mAppFLNA, 4 kPa), 40 (mAppFLNA, 90 kPa), 34 (dnFLNA, 4 kPa), and 36 (dnFLNA, 90 kPa) cells. Statistical significance was calculated using unpaired t test with Welch’s correction (two-tailed) (B) or Brown-Forsythe Anova test with Welch’s correction (F). Data are given as mean ± s.e.m (B, F) taken over three independent replicates. P values are shown in the graphs. Scale bar, 10 μm. Soft ECM: 4 kPa; Stiff ECM: 90 kPa.

FIGURE 4:

Mitochondrial membrane tension is sensitive to matrix stiffness but not to mitochondrial localization. (A) Confocal (in grayscale) and FLIM images of MCF-7 cells cultured on soft (top panel) and stiff (bottom panel) ECM showing mitochondria stained with Mito Flipper-TR. White boxes represent ROIs drawn in the PN and PL regions that were used to calculate fluorescence lifetimes. Magnified view of ROIs (Scale bar, 2 μm) shows higher fluorescence lifetimes on soft ECM. Scale bar,10 μm. (B) Scatter dot plot showing average fluorescence lifetimes calculated in PN and PL regions of cells cultured on soft and stiff ECM. n = 158 (soft PN), 110 (soft PL), 129 (stiff PN), and 104 (stiff PL) ROIs from cells (125 cells on soft ECM, 123 cells on stiff ECM). Data are mean ± s.e.m taken over three independent replicates. (C) Confocal images of MCF-7 cells transfected with mAppFLNA (in red) and cultured on soft ECM (top panel) and with dnFLNA (in red) and cultured on stiff ECM (bottom panel) showing mitochondria (in green) stained with Mito Flipper-TR. (D) FLIM images of MCF-7 cells transfected with mAppFLNA and cultured on soft ECM (top panel); and with dnFLNA and cultured on stiff ECM (bottom panel) showing mitochondria stained with Mito Flipper-TR. White boxes represent ROIs drawn in the PN and PL regions that were used to calculate fluorescence lifetimes. Magnified views of ROIs (Scale bar, 5 μm) are shown for comparison. Scale bar, 20 μm. (E) Scatter dot plot showing average fluorescence lifetimes calculated in PN and PL regions of control (untransfected) and mAppFLNA transfected cells cultured on soft ECM (left graph); and control (untransfected) and dnFLNA transfected cells cultured on stiff ECM (right graph). n = 76 (Control PN), 42 (Control PL), 33 (mAppFLNA PN), 21 (mAppFLNA PL) ROIs from cells (42 control cells and 21 mAppFLNA transfected cells on 1.2 kPa); n = 62 (Control PN), 54 (Control PL), 53 (dnFLNA PN), and 32 (dnFLNA PL) ROIs from cells (47 control cells and 31 dnFLNA transfected cells on 90 kPa). Data are mean ± s.e.m taken over two independent replicates. Statistical significance was calculated using unpaired t test with Welch’s correction (two-tailed) (B and E). Each data point represents the average lifetime in each ROI. P values are shown in the graphs. ROI: region of interest. Soft ECM: 1.2 kPa; Stiff ECM: 90 kPa.

FIGURE 5:

PN mitochondrial clustering prime mesenchymal stem cells towards osteogenesis. (A) Three-dimensional stacked confocal images of hMSCs cultured on soft (top panel) and stiff (bottom panel) ECM showing mitochondria stained with Mitotracker Green. From left to right: merged images showing nuclei in cyan and mitochondria in yellow, grayscale images of mitochondria showing differences in localization, and magnified view of the red boxes (Scale bar, 2 μm) showing differences in mitochondrial morphology. (B) Three-dimensional stacked confocal images of hMSCs under untransfected (control) and transfected (mAppFLNA in magenta) conditions cultured on soft ECM (left panel); and under untransfected (control) and transfected (dnFLNA in magenta) conditions cultured on stiff ECM (right panel) showing mitochondria immunostained with ATP5α. From top to bottom: merged images showing nuclei in cyan and mitochondria in yellow followed by grayscale images of mitochondria showing differences in localization. (C) Confocal immunofluorescence images of hMSCs cultured on stiff ECM under untransfected (control, top panel) and transfected (dnFLNA in magenta, bottom panel) conditions and stained with DAPI, ATP5α, and RUNX2. From left to right: merged images showing nuclei in cyan and mitochondria in yellow, grayscale images of mitochondria showing differences in localization, magnified view of the white boxes showing nuclei in blue, RUNX2 in gray and merged images of nuclei in blue and RUNX2 in green, scatter dot plot showing total intensity of RUNX2 in nuclei of control (untransfected) and dnFLNA transfected cells on 90 kPa. n = 60 (Control), 31 (dnFLNA) cells. (D) Confocal immunofluorescence images of hMSCs cultured on stiff ECM under transfected (mCh-KIF5B* in magenta and Tom20-GFP in yellow) conditions and stained with DAPI and RUNX2.From left to right: merged images showing nuclei in cyan and mitochondria in yellow, grayscale images of mitochondria showing loss of PN clustering, magnified view of the white boxes showing nuclei in blue, RUNX2 in gray and merged images of nuclei in blue and RUNX2 in green, scatter dot plot showing total intensity of RUNX2 in nuclei of control (untransfected) and transfected (KIF5B* + Tom-20) cells on 90 kPa. n = 52 (Control), 26 (KIF5B*+ Tom20) cells. (E) Three-dimensional stacked immunofluorescence images of hMSCs cultured in osteogenic differentiation media (for 14 d) on stiff ECM under untransfected (control, top panel) and transfected (dnFLNA in magenta, bottom panel) conditions and stained with DAPI, ATP5α, and RUNX2. From left to right: merged images showing nuclei in cyan and mitochondria in yellow, grayscale images of mitochondria showing differences in mitochondrial localization, and magnified view of the white boxes showing nuclear RUNX2 in gray. (F) Schematic summarizing the key findings: Matrix stiffening affects mitochondrial localization. Mitochondria on soft ECM are homogeneously distributed across the cytoplasm and are highly networked and elongated (left half of the top cartoon). Mitochondria on stiff ECM are fragmented and clustered near the nucleus (right half of the top cartoon). Stiffness-sensitive differential localization of filamin alters mitochondrial localization. PN enrichment of filamin is responsible for PN clustering of mitochondria (top cartoon). Stiffness-sensitive changes in mitochondrial positioning have physiological implications in stem cell biology. In particular, the PN clustering of mitochondria is important for nuclear localization of RUNX2 which primes hMSCs towards osteogenesis on stiff ECM (bottom cartoon). Statistical significance was calculated using two-tailed Mann-Whitney test (C and D). Data are given as median and interquartile range (C and D) taken over three independent replicates. P values are shown in the graphs. White lines denote the cell boundary and cyan ovals depict cell nuclei, both drawn manually using DIC images (for untransfected cells) or mApple/mCherry fluorescence (for transfected cells) acquired simultaneously with other fluorescence images. Scale bar, 10 μm. Soft ECM: 1.2 kPa; Stiff ECM: 90 kPa.

Stiffness-sensitive mitochondrial localization, but not morphology, depends on integrin signaling

Although MCF-7 cells showed equivalent spreading on either stiffness condition (Supplemental Figure 1C), there were distinct differences observed in the organization of actin. Cells on soft ECM exhibited diffused actin structures, while those on stiff ECM displayed prominent actin stress fibers (Control panels in Supplemental Figure 1E), indicating integrin-based mechano-sensing of ECM stiffness. Hence, to further understand the effect of ECM stiffness on the subcellular localization of mitochondria, we targeted integrin-mediated mechanosignaling, which connects the cytoskeleton to ECM forming a mechanical continuity (Schwartz, 2010). Vinculin is an integral part of integrin signaling, serving as a molecular link between talin and the actin cytoskeleton (Humphries et al., 2007). Vinculin binds to talin via its head domain and to filamentous-actin (F-actin) through its tail domain. However, vinculin typically exists in an autoinhibited state where intramolecular interactions between head and tail domains mask ligand binding sites. Under mechanical tension, vinculin gets activated leading to conformational changes that dissociate the head-tail interaction and expose the binding sites (Mandal et al., 2021; Supplemental Figure 1D). We asked whether mitochondrial morphology and localization might be sensitive to Vinculin-mediated integrin signaling. To this end, we used two vinculin plasmids: a constitutively active form of Vinculin labeled with green fluorescent protein, where the head-tail interaction was abolished (called Vcl-T12) and Vinculin tail domain labeled with green fluorescent protein, which only binds to actin but not talin (called Vcl-T; Supplemental Figure 1D; Mandal et al., 2021). Vcl-T binds to the barbed ends of F-actin and hence prevents actin polymerization, thereby having a dominant negative effect on the interaction between endogenous vinculin and F-actin (Le Clainche et al., 2010; Carisey et al., 2013). We first enhanced integrin mechanosignaling by transient expression of Vcl-T12 in MCF-7 cells cultured on soft ECM. Cells positive for Vcl-T12 showed distinct actin filaments in contrast to diffused actin structures seen in control (untransfected) cells on soft ECM (Supplemental Figure 1E). Conversely, we abrogated integrin signaling by transient expression of Vcl-T in MCF-7 cells cultured on stiff ECM. Cells positive for Vcl-T had diffused actin structures in contrast to distinct stress fibers seen in control (untransfected) cells on stiff ECM (Supplemental Figure 1E). Mitochondrial localization appeared very sensitive to mutant Vinculin-changes. On soft ECM, cells positive for Vcl-T12 showed increased PN clustering of mitochondria in contrast to the homogeneously distributed mitochondria seen in the control (untransfected) cells (Figure 1, G and H; Supplemental Figure 2, A and B). On stiff ECM, cells positive for Vcl-T showed homogenous distribution of mitochondria in contrast to the PN clustering seen in the control (untransfected) cells (Figure 1, G and H; Supplemental Figure 2, A and C). However, Vinculin mutants did not have any significant effect on mitochondrial morphology or network properties on either ECM stiffness (Supplemental Figure 1F). Taken together, these results clearly showed that mitochondrial positioning within a cell critically depends on integrin-mediated sensing of ECM stiffness.

Matrix stiffness generates distinctly dynamic mitochondrial populations

We subsequently characterized the dynamical properties of mitochondria on soft and stiff ECM and asked whether there could be any localization-specific differences among mitochondria on soft and stiff ECM. To this end, we stably expressed MitoDendra2, a photo-convertible fluorescent protein construct targeted to the mitochondrial matrix (Pham et al., 2012), in MCF-7 cells. We cultured these cells on soft and stiff ECM. We then imaged mitochondrial dynamics by two-dimensional time-lapse confocal microscopy and used an automated toolkit for mitochondrial segmentation and tracking, namely Mitometer (Lefebvre et al., 2022), to analyze motility features such as distance, speed, displacement, and velocity. This analysis showed that all the motility features of mitochondria were higher on stiff ECM than on soft ECM (Supplemental Figure 3A), implying mitochondria in general are more mobile on stiff ECM than on soft ECM. Next, considering the effect of ECM stiffness on mitochondrial localization, we asked whether this difference arose due to differences in the motility of specific mitochondrial subpopulations or whether it might be general and location-independent. To test these alternative possibilities, we photo converted a population of MitoDendra2 expressing mitochondria at two locations each: one in the PN region and one away from the nucleus, in the PL region. Such an approach helped us to track the motility of mitochondria at specific locations within the cytoplasm. We thus photo converted subsets of mitochondria at PN and PL regions and monitored the diffusion of the photo-converted molecules using time-lapse imaging techniques. Time lapse images showed greater spread of the photoconverted molecules in PL regions as compared with PN regions on soft ECM (Figure 2A; Supplemental Figure 3C; Supplemental Video 1). However, on stiff ECM, the spread was comparable in both PN and PL regions (Supplemental Figure 3D; Supplemental Video 2). Furthermore, quantitative analysis of photo-converted mitochondria showed that on soft ECM, the velocity, displacement, distance, and speed of PL mitochondria were higher than PN mitochondria, while on stiff ECM the same parameters were comparable between PN and PL mitochondria. (Figure 2B). In short, it was surprising that soft ECM created two mitochondrial populations that were different in their motilities.

Movie S1.

Download video file ^{(299.6KB, mp4)}

Mitochondria on soft ECM (4 kPa) have subpopulations that differ in their motilities. Left: Time‐lapse video showing mitochondrial motility in unconverted mitochondria (green) and photo‐converted mitochondria (red) of a MitoDendra2 MCF‐7 cell cultured on soft ECM. Right: Time‐lapse video showing mitochondrial motility of photoconverted mitochondria (in black). Photoconversion was carried out at PN and PL regions (yellow boxes) and the movement of photo‐converted mitochondria was tracked at an interval of 8.5s for ∼5.5 mins. Diffusion of photoconverted molecules was faster in PL mitochondria when compared to PN mitochondria highlighting differences in mitochondrial motility. Montage is shown in Figure 2A and Supplemental Figure 3C. Movie is shown at 5 frames per second. PN: perinuclear, PL: peripheral. Scale bar,10 μm.

Movie S2.

Download video file ^{(633KB, mp4)}

Mitochondria on stiff ECM (90 kPa) have subpopulations with comparable motilities. Left: Time‐lapse video showing mitochondrial motility in unconverted mitochondria (green) and photo‐converted mitochondria (red) of a MitoDendra2 MCF‐7 cell cultured on stiff ECM. Right: Time‐lapse video showing mitochondrial motility of photoconverted mitochondria (in black). Photoconversion was carried out at PN and PL regions (yellow boxes) and the motility of photo‐converted mitochondria was tracked at an interval of 8.5s for ∼5.5 mins. Diffusion of photoconverted molecules was comparable between PN and PL mitochondria. Montage is shown in Supplemental Figure 3D. Movie is shown at 5 frames per second. PN: perinuclear, PL: peripheral. Scale bar,10 μm.

To further test whether the mitochondrial localization was indeed influencing mitochondrial motility, we used the reversible association with motor proteins (RAMP) system to manipulate organelle positioning within the cell (Guardia et al., 2019). This system consists of coexpressing in cultured cells an organelle-localizing protein fused to streptavidin-binding protein (SBP) and an anterograde or retrograde motor protein fused to streptavidin. The SBP-Streptavidin interaction causes coupling of the organelle to the motor protein, leading to redistribution of the organelle either to the cell periphery or near the nucleus. Importantly, this redistribution can be reversed by adding the vitamin biotin, which outcompetes the SBP-streptavidin interaction, allowing restoration of normal organelle distribution (Supplemental Figure 4B). Therefore, we used this system to localize a major population of mitochondria near the nucleus on soft ECM and studied whether this imposed PN localization would perturb the motility. In particular, we used the retrograde motor protein KIF-C1 fused to streptavidin and mCherry (mCh-KIFC1*) and mitochondria outer membrane protein TOM20 fused to streptavidin binding protein (SBP) and GFP (Tom20-GFP) to generate minus-end (retrograde) directed movement of mitochondria. We transiently coexpressed mCh-KIFC1* and Tom20-GFP to promote PN mitochondrial localization on soft ECM and then studied mitochondrial motility by measuring the fluorescent recovery of the photobleached GFP. We selected two locations for photobleaching: one in the PN region and one away from the nucleus (peripheral or PL). Cells positive for both plasmid constructs indeed had increased PN mitochondria (top panel in Supplemental Figure 4C). Subsequently, fluorescent recovery after photobleaching (FRAP) analysis showed that the mobile fraction in PN and PL mitochondria was comparable (top panel in Figure 2C; Supplemental Figure 4D; Supplemental Video 3). However, upon addition of biotin which reverted mitochondria to their original homogenous distribution (bottom panel in Supplemental Figure 4C), the mobile fraction of PL mitochondria was greater than PN mitochondria (bottom panel in Figure 2C; Supplemental Figure 4D; Supplemental Video 4), reiterating our previous observation that on soft ECM, PL mitochondria are more mobile (Figure 2B). As expected, on stiff ECM, where mitochondria are anyways predominantly localized in the PN region, the mobile fractions of PL and PN mitochondria were comparable before and after biotin treatment (Supplemental Figure 4E).

Movie S3.

Download video file ^{(345.1KB, mp4)}

Perinuclear clustering of mitochondria on soft ECM (4 kPa) abolishes differences in mitochondrial motility. Perinuclear clustering of mitochondria in MCF‐7 cells was achieved by transient transfection of mCh‐KIFC1* and Tom20‐GFP cultured on soft ECM (4 kPa) without the addition of biotin (‐ Biotin) (Supplemental Figure 4, B‐C). Photo‐bleaching was carried out at T= 00:04s in PN and PL regions (white circles) and fluorescence recovery was monitored for 1 min 40 s. The fluorescence recovery in PN and PL mitochondria were comparable. Montage and FRAP curves shown in Figure 2C (top panel) highlight that perinuclear mitochondrial clustering abolishes differences in mitochondrial motility. Movie is shown at 5 frames per second. PN: perinuclear, PL: peripheral. Scale bar,10 μm.

Movie S4.

Download video file ^{(767.1KB, mp4)}

Reversing perinuclear clustering of mitochondria to its original homogenous distribution on soft ECM (4 kPa) reestablishes mitochondrial populations with different motilities. Perinuclear clustering of mitochondria in MCF‐7 cells achieved by transient transfection of mCh‐KIFC1* and Tom20‐GFP cultured on soft ECM (4 kPa) was reversed upon addition of biotin and mitochondria were reverted to their original homogenous distribution within the cell (Supplemental Figure 4, B‐C). After an hour of incubation with biotin (+ Biotin), photo‐bleaching was carried out at T=00:04s in PN and PL regions (white circles) and the recovery of fluorescence was monitored for 1 min 40 s. The fluorescence recovery in PL mitochondria was higher than in PN mitochondria reiterating our observation that soft ECM generates mitochondrial populations that have different motilities. Montage and FRAP curves shown in Figure 2C (bottom panel). Movie is shown at 5 frames per second. PN: perinuclear, PL: peripheral. Scale bar,10 μm.

These results highlight that specific PN clustering of mitochondria on stiff ECM abolishes differences in motility that were seen in homogeneously distributed mitochondria on soft ECM. In fact, both correlation experiments (with MitoDendra2) and causation experiments (with RAMP system) converged to the same conclusion that soft ECM creates two mitochondrial populations that differ in their motilities. Upon PN clustering, this difference gets abrogated. Taken together, these results showed that matrix stiffness affects mitochondrial motility and the differences in motilities are tied to the specific location of mitochondria within the cell.

PN accumulation of filamin alters mitochondrial localization

Having established the stiffness-sensitive changes in mitochondrial positioning, we next asked what factors determined the differential localization of mitochondria on soft and stiff ECM. It is known that ECM stiffness alters the cellular localization of several force-sensitive cytoskeleton-related proteins (Lee et al., 2012). Recently, we have shown that ECM stiffness-sensitive differential localization of an actin crosslinker protein, filamin (filamin A or FLNA), plays a critical role in epithelial defense against cancer (Pothapragada et al., 2022), where PN filamin molecules colocalize with PN actin cytoskeleton on stiff ECM. In this current study, we also found that while total filamin levels on soft and stiff ECM were comparable as quantified by immunostaining and microscopy, a significant fraction of filamin was localized in the PN region on stiff ECM (Figure 3, A and B; Supplemental Figure 5A) corroborating our previous results (Pothapragada et al., 2022). Here we speculated that filamin could be an interesting candidate concerning stiffness-sensitive mitochondrial localization, considering the intricate association of actin structures with mitochondrial dynamics (Hatch et al., 2014; Korobova et al., 2014; Ji et al., 2015; Moore et al., 2016; Chakrabarti et al., 2018; Yang and Svitkina, 2019). In addition, filamin A has been identified as a guanine nucleotide exchange factor, controlling the activity of a mitochondrial fission protein, Drp1, thus influencing mitochondrial fragmentation (Nishimura et al., 2018). We, therefore, hypothesized that stiffness-sensitive differential localization of filamin might be responsible for the stiffness-sensitive differential localization of mitochondria. To test this hypothesis, we transiently overexpressed filamin, labeled with fluorescent protein mApple (called mAppFLNA hereafter), in MCF-7 cells cultured on soft ECM. Filamin overexpression increased filamin distribution throughout the cell, especially in the PN region ([Pothapragada et al., 2022], top panel in Figure 3C). Interestingly, cells overexpressing filamin showed increased PN mitochondrial clustering even on soft ECM in contrast to the homogenous distribution seen in the control (untransfected) cells. However, filamin overexpression did not further affect PN mitochondrial clustering on stiff ECM (Figure 3, D–F; Supplemental Figure 5, C and D), which were already clustered around the nucleus. We next tested the effect of disruption of PN filamin on mitochondrial localization. Considering that refilinB-filamin interaction is important for localizing filamin to the PN cytoskeleton (Gay et al., 2011; Pothapragada et al., 2022), we disrupted this interaction and reduced the PN filamin localization on stiff ECM by transient expression of a dominant negative filamin mutant (called dnFLNA hereafter; bottom panel in Figure 3C). dnFLNA carries only four filamin repeats and is known to have a dominant negative effect on the interaction between endogenous filamin and refilinB (Gay et al., 2011). Cells expressing dnFLNA (labeled with mApple) indeed showed homogeneous distribution of mitochondria on stiff ECM in contrast to the PN clustering seen in the control (untransfected) cells. Interestingly, dnFLNA expression did not affect mitochondrial localization on soft ECM (Figure 3, D–F; Supplemental Figure 5, C and E). MCF-7 cells expressing mApple (mAppC1) alone had mitochondrial distribution similar to untransfected cells cultured on either stiffness condition, thus ruling out any effect of fluorescent protein mApple on mitochondrial distribution (Supplemental Figure 5B). Taken together, these results showed that ECM stiffening-induced PN enrichment of filamin affected mitochondrial localization, and PN filamin aggregation is critical for PN clustering of mitochondria.

Mitochondrial membrane tension is sensitive to matrix stiffness but not to mitochondrial localization.

Considering that filamin is an important cytoskeleton-associated protein, and cellular organelles are known to experience physical cues through their association with the cytoskeleton (Moore and Holzbaur, 2018; Shah et al., 2021), we next asked whether the filamin-dependent mitochondrial localization would also influence the physical properties of mitochondria. Relevantly, mitochondria are tubular structures and bear tension during force loading (Helle et al., 2017; Yang and Svitkina, 2019). Because mitochondrial membrane tension plays a role in mitochondrial dynamics (Mahecic et al., 2021), we examined how stiffness-mediated changes in mitochondrial morphology and localization would affect mitochondrial membrane tension. To measure tension, we used Mito Flipper-TR (Goujon et al., 2019), a mitochondrial-targeted mechanosensitive FliptR probe that reports changes in membrane tension through its fluorescence lifetime. FliptR, a fluorescent lipid tension reporter, is a planarizable push–pull probe composed of two dithienothiophene (DTT) flippers. It senses changes in the organization of lipid bilayer membranes through changes in the twist angle and polarization between the two DTT groups of the mechanophore. The fluorescence lifetime of Flipper-TR depends linearly on membrane tension, enabling easy quantification of membrane tension by fluorescence lifetime imaging microscopy (FLIM). We confirmed the mitochondrial localization of the flipper probe by costaining with an established mitochondrial marker (Mitotracker Red; Supplemental Figure 6A). Upon FLIM analysis, we observed that mitochondrial membrane tension as indicated by fluorescence lifetimes of Mito Flipper-TR probe was lowered on stiff ECM as compared with soft ECM (Figure 4A; Supplemental Figure 6, B and C). Fluorescence lifetimes in PN and PL mitochondrial populations on either stiffness condition were comparable (Figure 4, A and B). To further test the effect of mitochondrial localization on mitochondrial membrane tension, we measured membrane tension in mAppFLNA expressing cells on soft ECM and in dnFLNA expressing cells on stiff ECM (Figure 4C). PN aggregation of mitochondria by mAppFLNA did not alter fluorescence lifetimes in PN or PL mitochondrial populations when compared with the control (untransfected) cells on soft ECM. Conversely, disruption of PN mitochondrial clusters by dnFLNA did not alter fluorescence lifetimes in PN or PL mitochondrial populations when compared with the control (untransfected) cells on stiff ECM (Figure 4, D and E). Taken together, these results showed that the subcellular localization of mitochondria does not affect membrane tension. However, mitochondrial membrane tension is sensitive to matrix stiffness.

PN mitochondrial clustering primes mesenchymal stem cells towards osteogenesis

Finally, we asked what the physiological relevance of PN mitochondrial clustering on stiff ECM might be. It is well known that culturing human mesenchymal stem cells (hMSCs) on stiff ECM promotes osteogenesis, generally marked by increased nuclear expression of an osteogenesis-related protein RUNX2 (Engler et al., 2006). In this respect, at first, examining the mitochondrial morphology and localization in hMSCs, we obtained similar results as before in MCF-7 cells (Figure 5A). Mitochondria in hMSCs cultured on soft ECM were highly networked, elongated, and homogeneously distributed whereas mitochondria in hMSCs cultured on stiff ECM were fragmented and formed PN clusters (Figure 5A; Supplemental Figure 7A). Interestingly, we also observed stiffness sensitive differential localization of filamin. hMSCs on stiff ECM had a distinct PN accumulation of filamin (Supplemental Figure 7B). Importantly, transient expression of mAppFLNA in hMSCs on soft ECM led to enhanced PN filamin (Supplemental Figure 7B) and mitochondrial clustering near the nucleus (Figure 5B), whereas PN filamin disruption using dnFLNA on stiff ECM (Supplemental Figure 7B) led to loss of PN mitochondrial clusters (Figure 5B), further extending the generality of our findings. Given these findings, we hypothesized that whether mitochondrial localization on stiff ECM indeed influenced osteogenic differentiation of hMSCs, perturbing the former would attenuate osteogenesis on stiff ECM. To test this hypothesis, we cultured hMSCs in normal media conditions (undifferentiated state) on soft or stiff ECM for 24 h. Upon immunostaining for RUNX2, we observed that this osteogenic marker was lowly expressed on soft ECM, whereas the expression was detectable in the cell nuclei on stiff ECM (Supplemental Figure 7C). We then disrupted PN mitochondrial clustering on stiff ECM via transient expression of dnFLNA and observed that the nuclear signal of osteogenic marker RUNX2 reduced significantly when compared with control (untransfected) cells (Figure 5C). hMSCs on stiff ECM expressing mApple (mAppC1) alone showed PN mitochondrial clustering and nuclear levels of RUNX2 comparable to untransfected (wildtype) cells, thus ruling out any effect of fluorescent protein mApple on mitochondrial distribution or RUNX2 nuclear levels (Supplemental Figure 7D). To rule out indirect effects of dnFLNA as well as to confirm the specific effects of altering mitochondrial positioning on nuclear localization of RUNX2, we employed other approaches to manipulate mitochondrial distribution. To be specific, we disrupted PN mitochondrial clustering using dominant negative vinculin Vcl-T and microtubule-mediated manipulation using the RAMP system (Guardia et al., 2019). Abrogating integrin signaling using Vcl-T disrupted PN aggregation of mitochondria and also reduced nuclear levels of RUNX2 when compared with control (untransfected) cells on stiff ECM (Supplemental Figure 8A). Additionally, in the RAMP system, we used the anterograde motor protein KIF5B* fused to streptavidin and mCherry (mCh-KIF5B*) and mitochondria outer membrane protein TOM20 fused to SBP and GFP (Tom20-GFP) to generate plus-end (anterograde) directed movement of mitochondria (Supplemental Figure 4B). hMSCs coexpressing mCh-KIF5B* and Tom20-GFP exhibited loss of PN mitochondrial clusters and reduced nuclear levels of RUNX2 when compared with control (untransfected) cells on stiff ECM (Figure 5D; Supplemental Figure 8B). Moreover, cells expressing either mCh-KIF5B* or Tom20-GFP alone had high nuclear levels of RUNX2 (top and middle panels in Supplemental Figure 8B). These results collectively showed that integrin-, filamin-, and microtubule-mediated disruption of PN mitochondrial clustering resulted in a common phenotype of reduced RUNX2 nuclear signal.

Because the subcellular localization of YAP is also dependent on ECM stiffness (Dupont et al., 2011) we also examined the effect of mitochondrial positioning on the nuclear import of YAP on stiff ECM. Interestingly, despite disrupting PN aggregation of mitochondria using dnFLNA or mCh-KIF5B* and Tom20-GFP, the nuclear percentage of YAP in these transfected cells was comparable to control (untransfected) cells (Supplemental Figure 8C). This finding indicated a specific role of PN mitochondrial clustering in the nuclear localization of RUNX2. Our results showed that PN mitochondrial clusters were crucial for the nuclear localization of RUNX2 on stiff ECM and hence primed hMSCs towards osteogenesis, albeit in their undifferentiated state. On stiff ECM, in the presence of osteogenic differentiation media, the process of osteogenesis requires 2 wk or more. We, therefore, cultured hMSCs on stiff ECM in the presence of osteogenic differentiation media and on Day 14 observed clear nuclear localization of RUNX2 (Figure 5E, upper panels) in control (untransfected) cells. Next, using nucleofection, we transfected hMSCs with dnFLNA and then cultured them on stiff ECM in the presence of osteogenic differentiation media for 14 d. However, even after 14 d in differentiation media, cells positive for dnFLNA, which had dispersed mitochondria, showed significantly reduced nuclear levels of RUNX2 as compared with the control (untransfected) cells (Figure 5E). This experiment showed that PN mitochondrial positioning not only primes stem cells towards osteogenesis but is also a crucial determinant of proper osteogenic differentiation. Taken together, our results showed PN mitochondrial clustering to be a crucial determinant of osteogenic differentiation in a stiff microenvironment (Figure 5F).

DISCUSSION

Till now, most studies have focused on the effect of matrix stiffening on cytoskeletal architecture, nuclear mechanics, and metabolism. However, the role of matrix stiffness in the regulation of endomembranes, especially mitochondria, has been mostly overlooked or just started to be unveiled. Matrix stiffening and mitochondrial dysregulation both being hallmarks of cancer, aging, fibrosis, and cardiovascular diseases (Lampi and Reinhart-King, 2018), it is crucial to study the crosstalk between matrix stiffness and mitochondrial form and function. Among the organelles, the mitochondrion is a good starting point not only because of its bioenergetic capabilities but also because it interacts with the cytoskeletal elements and is shown to bear forces. Hence, we chose to focus on the effect of matrix stiffness on mitochondrial morphology and localization. Our study shows that mitochondria in human epithelial cells, MCF-7, and hMSCs are sensitive to mechanical cues. Upon matrix stiffening, they undergo a transition from a homogeneously distributed, filamentous, and highly networked phenotype to a fragmented and less networked phenotype showing PN clustering (Figure 5F). In this respect, we found that the effect of matrix stiffness is more distinct on mitochondrial localization than on mitochondrial morphology. On the effect of matrix stiffness on mitochondrial morphology, while our results are statistically significant for a p < 0.05 and corroborate with the reports of Tharp et al. (2021), the distribution in morphology is very broad. We believe that this wide distribution may lead to opposing conclusions depending on the context and the study design, which might explain the contradicting results from other studies (Chen et al., 2021; Romani et al., 2022). On the contrary, the effect of ECM stiffness on mitochondrial localization is very distinct. To this end, we further show that stiffness-sensitive differential localization of filamin is crucial for determining the subcellular localization of mitochondria in two different cell systems: MCF-7 breast cancer cells and hMSCs. PN enrichment of filamin on stiff ECM leads to PN mitochondrial clustering (Figure 3). Going forward, it will be interesting to understand how this PN clustering of fragmented mitochondria is dynamically achieved on matrix stiffening in vivo. It is possible that during matrix stiffening due to fibrosis, mitochondria undergo fragmentation to increase ATP activity, are transported to the nucleus in a retrograde manner by microtubules and are finally retained near the nucleus by PN actin structures formed by filamin and associated proteins. However, the crosstalk between actin, microtubule, and mitochondrial form and localization on soft and stiff ECM needs further investigation. Relevantly, filamin could be a key molecule connecting both mitochondrial form and localization. In fact, given recent reports on stiffness-sensitive regulation of mitochondrial fission proteins (Chen et al., 2021) and the role of filamin in mitochondrial fission (Nishimura et al., 2018), it would be interesting to look at the effect of filamin on localization-dependent or independent dynamics and functions of mitochondria.

Along with localization, we also studied mitochondrial motility by photoactivation and photobleaching. Subsequently, our experiments involving imaging mitochondrial motility revealed that mitochondria on stiff ECM are more mobile (Supplemental Figure 3A) and we attribute this behavior to fragmented mitochondria seen on stiff ECM (Supplemental Figure 3B). In other words, mitochondria on stiff ECM can traverse their local surroundings more easily due to their fragmented morphology as opposed to mitochondria on soft ECM which are tubular and elongated. We also show that on soft ECM, there exist two distinct mitochondrial populations that differ in their motilities (Figure 2B). PL mitochondria are more mobile than PN mitochondria because of their shorter lengths (minor: major axis) as shown in Supplemental Figure 4A. In essence, specific morphometric features contribute to mitochondrial motility. However, we speculate that the motility difference in PN and PL mitochondrial populations arises due to the difference in the volume constraint of intracellular space near the cell nucleus. The PN region is relatively more organelle-crowded than the PL region. The former, therefore, should impose enhanced restrictions on mitochondrial motility as compared with the latter. As it appears, the interaction of mitochondria with filamin on stiff ECM abolishes this difference. Beyond physical constraints and molecular interactions, it is also likely that the dynamics of mitochondrial subpopulations is eventually influenced by local subcellular demands. This possibility needs to be further investigated. In fact, distinct mitochondrial populations are also reported in the secretory cells of the salivary gland which have central mitochondria that are mobile and rarely fuse, and basolateral mitochondria that are static and frequently fuse (Porat-Shliom et al., 2019). Also, pancreatic acinar cells have PN, perigranular, and subplasmalemmal mitochondria with distinct functions to regulate calcium transport (Park et al., 2001). Exploring specific physiological roles of mitochondria in these subcellular niches could therefore provide interesting insights into the spatiotemporal regulation of mitochondrial structure and function and add to the underappreciated heterogeneity in mitochondrial dynamics and positioning.

Finally, while exploring the physiological implication of PN clustering of mitochondria on a stiff matrix, we discovered that it is crucial for priming stem cells towards osteogenesis. In this regard, a stiff ECM drives osteogenic differentiation of mesenchymal stem cells in a YAP/TAZ-dependent manner (Engler et al., 2006; Dupont et al., 2011). Mitochondrial dynamics and function also play a crucial role in stem cell differentiation (Chen et al., 2008; Prowse et al., 2012; Khacho et al., 2016; Li et al., 2017). Upon induction of differentiation, there is a shift from glycolysis to oxidative phosphorylation, and mitochondria being at the center of the energy generation process becomes very crucial. While a previous study had just observed PN mitochondria in the early phase of osteogenic induction in MSCs (Chen et al., 2008), our study reveals a previously unknown role of organelle positioning in priming stem cells toward osteogenesis. We show that stiffness-sensitive PN clustering of mitochondria is important for the nuclear localization of RUNX2, the master regulator of osteogenesis (Figure 5). RUNX2 activity has epigenetic regulations, and at the same time, certain mitochondrial metabolites that play a key role in epigenetic modifications, are upregulated upon matrix stiffening (Tharp et al., 2021). Hence, it is possible that PN mitochondrial clustering is crucial not only for successful RUNX2 entry into the nucleus but also for increased mito-nuclear crosstalk in general. This crosstalk is then expected to drive several epigenetic modifications affecting gene transcription. However, how the PN mitochondria facilitate the nuclear entry, sequestration, and activity of RUNX2 and metabolites in the nucleus requires further investigation. Given recent reports showing the inhibitory effects of low stiffness on osteogenic differentiation due to mitochondrial dysfunction (Ma et al., 2023), it would also be worthwhile to explore the role of mitochondrial positioning in enhancing the osteogenic potential of hMSCs in microenvironments that are not conducive to osteogenesis. Therefore, we propose that PN mitochondrial clustering in a conducive microenvironment (stiff ECM) is crucial for osteogenic differentiation of mesenchymal stem cells. Taken together, considering the recent discoveries of nucleus-associated mitochondria that help in stress resistance and survival (Desai et al., 2020) and nuclear deformations caused by mitochondrial swelling in cardiomyocytes (Kaasik et al., 2010), we speculate that PN mitochondrial clustering in a stiffer microenvironment is an adaptive response that possibly provides an energetic barrier protecting as well as maintaining nuclear architecture and functions.

MATERIALS AND METHODS

Request a protocol through Bio-protocol.

Cell culture

MCF-7 (NCCS, Pune) cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with GlutaMax (Life Technologies) with 10% fetal bovine serum (tetracycline-free FBS; Takara Bio) and 10 U ml^–1 penicillin and 10 μg ml^–1 streptomycin (Pen-Strep, Invitrogen) in an incubator maintained at 37°C and 5% CO₂. To establish stable cell lines expressing MitoDendra2, MCF-7 cells were transfected with MitoDendra2 plasmid DNA using Lipofectamine 2000 (Invitrogen). Selection pressure was provided by medium (DMEM-GlutaMax with 10% FBS) containing 400 μg ml^–1 geneticin (Invitrogen). Stably expressing fluorescent clones were picked using cloning cylinders (Sigma) following fluorescence confirmation. From this suspension, single cells were seeded via serial dilution in a 96-well plate and their growth was monitored for over 2 wk. Post this, homogeneously fluorescent colonies derived from single cells were scanned under a fluorescent microscope and subcultured into stable cell lines. Subsequent maintenance and passaging of stable cell lines were done in a medium containing 100 μg ml^–1 geneticin.

Transient transfection with plasmids was done either using Lipofectamine 2000 (Invitrogen) or Xfect (Takara) in a sixwell plate following the manufacturer’s protocol. Post 8–12 h of transfection, cells were trypsinized and seeded onto hydrogel substrates. After at least 24 h in culture, cells were either fixed and immuno-stained or imaged live.

hMSCs (Lonza) were cultured in mesenchymal stem cell growth medium (MSCGM) comprising of mesenchymal stem cell basal medium (MSCBM; Lonza) supplemented with mesenchymal cell growth supplement (MCGS; Lonza), L-Glutamine and GA-1000 in an incubator maintained at 37°C and 5% CO2. To induce osteogenesis, hMSCs were first allowed to adhere to the culture surface in MSCGM for 24 h. Osteogenesis was then induced by replacing MSCGM with osteogenic induction medium which was prepared by supplementing osteogenic differentiation basal medium with dexamethasone, L-glutamine, ascorbate, penicillin/streptomycin, beta-glycerophosphate, and MCGS. The hMSCs in culture were replenished every 2–3 d with osteogenic induction medium for 14 d.

Transfection of hMSCs was carried out using an efficient nonviral linear polyethylenimine (LPEI) based transfection method described elsewhere (Ho et al., 2020). Briefly, 1 mg/ml LPEI (PEI MAX, Polyscience) was added to plasmid DNA in serum free Opti-MEM at pDNA: LPEI ratio (1:2.5) in a total volume of 100 μl and incubated at room temperature (RT) for 15 min. The culture media was removed and replaced with this transfection mixture. Cells were incubated in this transfection mixture in 2 ml normal media supplemented with Enhancers at a final concentration of 10 μg/ml DOPE/CHEMS (9:2 M ratio; Polar Avanti Lipid) and 1.25 μM Vorinostat (SAHA, BioVision). Cells were incubated for 24 h before fixation. Nucleofection in hMSCs was carried out using the Amaxa P3 primary cell 4D-Nucleofector X kit (Lonza) according to the manufacturer’s instructions.

Hydrogel preparation for compliant ECM

To provide cells with the compliant ECM having different stiffness, polyacrylamide hydrogels coated with collagen-1 were made as described previously (Engler et al., 2006). 4% (3-Aminopropyl)triethoxysilane(APTES)-treated and 2% glutaraldehyde-activated glass-bottom dishes (Ibidi) were used to cast polyacrylamide (PAA) hydrogel substrates. Hydrogel substrates of varying stiffness with an elastic modulus of 1.2, 4, and 90 kPa were prepared by mixing the desired volume of 40% acrylamide and 2% bisacrylamide as given in Supplemental Table 1). Gel surfaces were functionalized with sulfosuccinimidyl-6-(4′-azido-2′-nitrophenylamino) hexanoate (Sulfo-SANPAH, Thermo Fisher Scientific) and covalently coated with 300 μg ml^–1 collagen-I (Invitrogen) overnight at 4°C to ensure cell attachment. Cells were seeded onto the gel area and cultured for 24 h before fixing or live imaging.

Antibodies, fluorophores, plasmids, and other reagents

Source and dilution information for all primary and secondary antibodies and fluorophores are given in Supplemental Table 2. Details of plasmids used in this study are listed in Supplemental Table 3 with their source. Details of all other reagents used in this study are listed in Supplemental Table 4.

Mitochondrial dyes

For live imaging of mitochondria, cells were incubated with 100 nM Mitotracker Green or Red dye for 15 min before imaging.

Immunofluorescence

Cells were fixed with 4% formaldehyde diluted in 1X phosphate-buffered saline (PBS, pH 7.4) at RT for 15 min, followed by three washes in 1X PBS. The fixed samples were then permeabilized using 0.25% (vol/vol) Triton X-100 (Sigma) diluted in 1X PBS for 10 min at RT followed by three washes for 5 min each in 1X PBS to remove the detergent. To block nonspecific antibody binding, the samples were incubated in 2% bovine serum albumin in PBST (0.1% vol/vol Triton X-100 in 1X PBS) at RT for 45 min. The blocking buffer was removed after 45 min, and the primary antibody dilution prepared in the blocking buffer was added to the samples for 60 min at RT or at 4°C overnight. After this, samples were washed twice with PBST and thrice with 1X PBS. Samples were then incubated with secondary antibody tagged with Alexa Fluor 488/568/594/647 (same dilution as primary antibody) prepared in blocking buffer, for 60 min at RT in dark. Counter-staining cell nuclei with a DNA-binding dye 4′,6-diamidino-2-phenylindole (DAPI, 1 μg ml⁻¹ in PBS), and F-actin with Alexa Fluor dye conjugated phalloidin (Invitrogen) was also done at this step. Finally, samples were washed twice with PBST and thrice with 1X PBS before being imaged using confocal microscopy.

Confocal microscopy

Immunofluorescence images were acquired using 60X water objective (UPLSAPO W, N.A.  =  1.2, Olympus) mounted on an Olympus IX83 inverted microscope equipped with a scanning laser confocal head (Olympus FV3000), Olympus FV31-SW (v2.3.1.198). Time-lapse images of live samples, photoconversion, and photo-bleaching studies were done in the same setup using a live-cell chamber at 37°C. HEPES (Life Technologies) buffer at a final concentration of 50 μM was added to the culture medium to maintain CO₂ levels during live imaging.

Super-resolution microscopy

Images were acquired with a 60X silicone oil objective (UPLSAPO60XS2, N.A = 1.3, Olympus) mounted on an Olympus IX83 inverted microscope supplied with Yokogawa CSU-W1 (SoRa Disk) scanner.

FLIM imaging and analysis

For FLIM imaging with Mito Flipper-TR, cells were incubated with 1 μM of the Mito Flipper-TR (Spirochrome) for 15 min. Media conducive for live imaging conditions (fluorobrite DMEM with 10% FBS, 1 mM Glutamine, and 50 μM HEPES buffer) was added to dilute the probe concentration three times and imaging was done in a live cell chamber at 37°C. Imaging was performed using the FV3000 Olympus microscope equipped with a time-correlated single-photon counting module from PicoQuant. A pulsed 485 nm laser (PicoQuant LDH-D-C-485) operated at 20 MHz was used for excitation. The emission was collected through a 600/50 nm bandpass filter, on a gated PMA hybrid 40 detector and a PicoHarp 300 board (PicoQuant). FLIM data was analyzed using the SymPhoTime 64 software (PicoQuant). Rectangular ROIs of comparable areas were drawn manually in PN and PL regions of cells. The average fluorescence lifetime was calculated in these ROIs. To determine the fluorescence lifetimes, the fluorescence decay data was fit to a double exponential model after deconvolution for the calculated impulse response function. The values reported in the main text are the average lifetime intensity (tau 1).

Mitochondrial segmentation and quantification

The workflow and procedures for image processing and thresholding using MitoAnalyser plug-in are described in detail elsewhere (Chaudhry et al., 2020). In short, three-dimensional image stacks were first acquired and preprocessed using the commands: subtract background, sigma filter plus, enhance local contrast, and gamma correction. To identify mitochondria in the images, “adaptive thresholding” method was used. The settings for thresholding in the Mitochondria Analyser plugin (on FIJI) were adjusted, with “scale max slope through stack” turned on and settings for “block size” and “c-value” adapted per set of images (block size: 2 microns, c:4) with other default settings. The resulting binarized images were postprocessed using the despeckle, remove outliers, and fill 3D holes commands. At this stage, each final thresholded image was compared visually to the original image as a quality control check of object identification and segmentation. Image correction, wherever required, was manually done by deselecting some of the preprocessing or postprocessing commands until the final thresholded image closely resembled the original image. The identified mitochondrial objects were then analyzed in 3D using 3D object counter and “3D particle analyzer” commands to quantify count, volume, surface area, and sphericity. The thresholded objects were also converted into skeletons using “skeletonize (3D)”, and the “analyze skeleton” command was applied to obtain the number of skeletons, number of branches, lengths of branches, and number of branch junctions and endpoints in the 3D network.

Mitochondrial motility using photoconversion and photobleaching

MitoDendra2 MCF-7 cells were used for mitochondrial motility studies. MitoDendra2 cells were seeded and cultured on soft and stiff ECM and incubated with 50 nM SiR Actin with 10 μM Verapamil for 12–16 h before imaging. SiR Actin was used to mark cell boundaries which helped in single cell analysis. Photoconversion was done on a rectangular region-of-interest (ROI) in PN and PL regions of MitoDendra2 MCF-7 cell using a 405-nm laser at 4% intensity with a scan speed of 40 μs/pixel. This was immediately followed by LSM imaging of the green and red channels, and mitochondrial motility was subsequently tracked at an interval of 8.5 s for ∼5.5 mins.

Photobleaching studies were done in MCF-7 cells transiently expressing mCh-KIFC1* and Tom20-GFP (RAMP system) and cultured on soft and stiff ECM. These doubly transfected cells were incubated with 50 μM Biotin for an hour to restore mitochondria to their original distribution. 488 nanometer laser was used at 10% intensity for bleaching a circular ROI in PN and PL regions, iterated or looped over three times with a scan speed of 20 μs/pixel. For LSM imaging, the laser power was attenuated to avoid phototoxicity. Photobleaching was done before and after Biotin treatment in the same setup but in different cells. Images were collected before, immediately after, and for ∼1.5 min following the bleaching. All FRAP data were analyzed and mobile fraction was calculated using Stowers plugin in FIJI.

Analysis of mitochondrial motility using mitometer

Single-cell videos of unconverted mitochondria (green channel) in MitoDendra2 MCF-7 cell line were acquired and fed into the MATLAB plugin of Mitometer for global characterization of mitochondrial motility. Videos of photo-converted mitochondria (red channel) in PN and PL regions were cropped out manually and fed into the Mitometer plugin for location-specific characterization of mitochondrial motility and morphology. The detailed protocol for running Mitometer is described elsewhere (Lefebvre et al., 2022). In brief, the average value of all the motility and morphology parameters obtained for each distinct mitochondrial track was calculated and plotted.

Analysis and quantification of mitochondrial localization

Nucleus and cell body from the respective images were manually segmented and binarized using FIJI. Mitochondria were segmented, thresholded, and binarized using the MitoAnalyser plugin in FIJI. Subsequent analysis of mitochondrial and cytoplasmic distribution within a cell as a function of distance from the nucleus was done using a custom MATLAB and python code. Briefly, the centroid of the nuclear mask was determined and considered to be the center of the circles for subsequent radial distribution analysis. For all the nonzero pixels of the mitochondrial mask, the distance from the nuclear centroid was calculated and written into a “csv” file. After subtracting the nucleus mark from the cell body mask, the distance of all the nonzero pixels of the cell body mask from the nuclear centroid was calculated and written into another “csv” file. The minimum distance obtained here is considered as the nuclear radius. The “PN region” was defined as the region that falls within 5 microns from the nuclear radius. These output files from the MATLAB code were then used for further analysis in Python. Inbuilt functions of the Python seaborn library were used for radial distribution analysis. Normalized frequency distribution of nonzero mitochondrial and cell body pixels with distance from the nucleus was determined using the “distplot” function. In the histogram of nonzero mitochondrial pixels, the area within the PN region is the fraction of PN mitochondrial content. All distances in pixels were converted to microns before analysis. The MATLAB program used for the analysis is available at https://github.com/basilthurakkal/perinuclear-clustering-of-mitochondria.

Image analysis

To determine PN filamin, ROIs were traced out manually using the selection brush tool (fixed at 20-pixel width) in FIJI. Cell-nuclei frame was synced with the filamin frame and used as a reference for tracing the PN region. Whole-cell boundaries were manually traced using DIC or phalloidin frames. Total intensity values were taken for PN and whole cell regions per cell. The fraction of PN filamin localization per cell was quantified as the ratio of the total intensity of the PN region to the total intensity of the whole cell. All fluorescence images were brightness-adjusted and denoised uniformly throughout the whole image for representation purposes only. Denoising was done using the Pure Denoise tool in FIJI.

Statistics and reproducibility

Statistical analyses were carried out in GraphPad Prism 9 (Version 9.2.0). Tests done for statistical significance and respective p values are mentioned in all the figures. p-values < 0.05 were considered to be statistically significant. No statistical methods were used to set the sample size. Quantification was done using data from at least three independent biological replicates. All the experiments with representative images were repeated at least thrice.

Abbreviations used:

ECM

Extracellular matrix

hMSC

human mesenchymal stem cell