AMPK is a mechano-metabolic sensor linking cell adhesion and mitochondrial dynamics to Myosin II dependent cell migration

doi:10.21203/rs.3.rs-1653312/v1

Download PDF

Article

AMPK is a mechano-metabolic sensor linking cell adhesion and mitochondrial dynamics to Myosin II dependent cell migration

https://doi.org/10.21203/rs.3.rs-1653312/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 22 May, 2023

Read the published version in Nature Communications →

Version 1

posted

You are reading this latest preprint version

Cell migration are crucial for development, immunity and cancer dissemination. We find that AMPK controls cell migration by acting as an adhesion sensing molecular hub. In 3-dimensional (3D) matrices, fast migrating amoeboid cancer cells exert low adhesion/low traction linked to lower energy levels; while their low ATP/AMP ratio leads to AMPK activation. In turn, AMPK plays a dual role controlling both mitochondrial dynamics and cytoskeletal remodelling. High AMPK in low adhering migratory cells, induces mitochondrial fission via MFF phosphorylation, resulting in lower OXPHOS and lower mitochondrial ATP. At the same time, AMPK phosphorylates and inactivates MYPT1, increasing Myosin II dependent amoeboid migration. Reducing adhesion, inducing AMPK activity or inhibiting mitochondrial fusion result in profound cytoskeletal remodelling and efficient rounded-amoeboid 3D migration and invasion. Moreover, AMPK inhibition suppresses in vivo metastatic potential of amoeboid cancer cells. We measure a mitochondrial/AMPK-driven switch in regions of human tumours where amoeboid cells are actively disseminating. We unveil how mitochondrial dynamics control cell migration and suggest that AMPK is a master “mechano-metabolic sensor” at the crossroads between energetics and the actomyosin cytoskeleton.

Adhesion

DDR receptors

AMPK

mitochondria

mitochondrial dynamics

amoeboid migration

MYPT

Myosin II

traction stress

Cell migration is a key process in immune responses, development, wound healing and tumour dissemination. Cells can migrate within complex 3-dimensional (3D) environments as collective groups or as individual cells, the latter falling mainly into two categories: elongated-mesenchymal and rounded-amoeboid migration. Modes of migration are interchangeable in vitro and in vivo as a consequence of physical and chemical insults ^1–4.

Efficient migration requires remodelling of cell-cell contacts and interactions with the extracellular matrix (ECM). The actomyosin cytoskeleton coupled to adhesions form traction points on the ECM to enable cell movement. Elongated-mesenchymal migrating cells rely on integrin-based focal adhesions, while rounded-amoeboid cells are able to adopt both adhesion-dependent and adhesion-independent migration, where the former employs weak adhesions and the latter involves propulsive, pushing movements ^5,6. Most studies on focal adhesions have been conducted in two-dimensional (2D) environments, lacking the pliability essential for cells to adopt a rounded-amoeboid migration. Yet, mechanical signals in 3D matrices could be very different.

Actomyosin contractility is essential for both elongated-mesenchymal and rounded-amoeboid modes of movement ^7,8. Myosin II-driven contractility relies on multiple kinases. For instance, Rho-kinase (ROCK) or Myosin Light Chain Kinase (MLCK) regulate Non-muscle Myosin II activity through the phosphorylation of its light and heavy chains ⁹. In turn, ROCK can also modulate Myosin activation through inactivating Myosin Phosphatase, leading to increased phosphorylated Myosin Light Chain 2 (pMLC2) ^10,11.

On the other hand, cells require energy to migrate, but how metabolism is rewired to fulfil those demands in 3D environments is poorly understood. Mitochondria, considered the powerhouse of the cell, control a number of cellular functions including cell division, apoptosis and cell migration. Mitochondrial activity is intimately linked to their morphology and through mitochondrial dynamics, which include mitochondrial fission and fusion as well as their transport, cells provide high amounts of energy where required ¹². However, the role of mitochondrial dynamics in cells migrating within physiological 3D complex environments employing different modes of movement is poorly defined.

Here, we investigated how cell adhesion to the matrix coupled to ATP levels and mitochondrial dynamics are key to determine the mode of migration that cancer cells adopt in 3D and how this connects to the actomyosin cytoskeleton at the molecular level.

Intrinsic Myosin II cytoskeletal organization and traction stress generation in 3D migrating cells.

Individual migrating cells can adopt either rounded-amoeboid or elongated-mesenchymal mode of movement. A panel of cell lines was used to define intrinsic patterns of cytoskeletal Myosin II organization and their relationship to cell morphology, adhesion and mode of migration on a pliable 3D matrix. The panel was composed of HT1080 fibrosarcoma, MDA-MB-231 breast cancer and 4 melanoma cell lines: paired WM983A/WM983B, with the same genetic background but derived from primary tumour (WM983A) and lymph node metastasis (WM983B) from the same patient; and paired A375P/A375M2 that share the genetic background since A375M2 were generated from A375P after three rounds of lung metastatic colonization in mice ^13–15. When cultured on a pliable collagen I matrix, within the melanoma group, WM983A and A375P melanoma cells showed high proportion of cells with an elongated morphology, lower levels of cortical pMLC2 and higher adhesion to the matrix -measured by number of points of attachment to collagen fibres and faster adhesion (Figs. 1A-F and Supp Figs. 1A-C). Conversely, their paired amoeboid and metastatic counterparts WM983B and A375M2, displayed a rounded morphology with higher levels of cortical Myosin II activity, membrane blebbing and lower adhesion to the matrix (Figs. 1A-F and Supp Figs. 1A-C). The non-melanoma cells HT1080 and MDA-MB-231 adopted an elongated morphology, lower cortical active Myosin II and higher adhesion (Figs. 1A-F and Supp Figs. 1A-C). Importantly, active Myosin II had a more diffuse localisation throughout elongated-mesenchymal cells while it was organised as a thick cortex in rounded-amoeboid cells (Fig. 1D). This data shows that cell morphology, Myosin II activity/localisation and levels of adhesion are good predictors of the mode of migration (Fig. 1G).

To understand how cells using different modes of migration generate traction stress, 3D image stacks of cells expressing LifeAct-GFP embedded in collagen I matrices were taken over time. For comparison to previous work using beads or artificial manipulation of the matrix ^16–18, reflected/backscatter light was collected to image collagen fibres to visualize reorganization induced by traction ^19,20. We then developed a custom-made algorithm for the direct tracking of collagen fibres and direction/angle of displacement in 3D. Integration of strain data, directional information, depth information and stiffness of the matrix allowed the calculation of traction stress (Supp Fig. 1D and Movies 1–3).

We generated traction maps for maximal intensity projections representing 3D reconstruction of migrating HT1080 and A375M2 cells. While elongated-mesenchymal HT1080 cells exerted traction stress within the order of 20 Pa, A375M2 rounded-amoeboid cells exerted much lower magnitude stresses (Fig. 1H), consistent with previously reported magnitudes for adhesion-free migration ⁶. Importantly, inhibition of Myosin II activity with blebbistatin resulted in a reduction of stress exerted by both HT1080 and A375M2 cells (Fig. 1I), revealing that traction stress applied by migrating cells is Myosin II-dependent.

3D map reconstructions showed that elongated-mesenchymal HT1080 cells exerted traction stress localised at protrusions (0–2 min) and around the cell body (4–6 min) (Supp Fig. 1E), while rounded-amoeboid A375M2 exerted directional stress localised in a single region of the cell (Supp Fig. 1F).

In summary, cell morphology, Myosin II organisation and levels of adhesion displayed by cells using different modes of migration determine the magnitude of the traction stress exerted into the matrix (Fig. 1G).

Different modes of migration show distinct mitochondrial metabolism and ATP levels.

How cells obtain the energy required for generating traction stress in 3D systems remains poorly understood. NMR metabolomics on rounded-amoeboid WM983B and A375M2 cells compared with their elongated-mesenchymal counterparts, WM983A and A375P, respectively, revealed that the metabolic profiles of each pair of cells were statistically different using partial least squares discriminant analyses (PLS-DA), and were enriched in a different set of metabolites (Fig. 2A and Supp Figs. 2A-B). The extracellular acidification rate (ECAR) of the media showed that aerobic glycolysis was not specifically associated with either type of migration (Supp Figs. 2C-D) or any specific oncogene/genetic alteration (Supp Fig. 2E). In contrast, higher mitochondrial respiration was associated with elongated-mesenchymal migration (Fig. 2B and Supp Fig. 2F). As such, higher oxygen consumption rates (OCR) and ATP production derived from oxidative phosphorylation (OXPHOS) were detected in elongated-mesenchymal cells when compared with rounded-amoeboid moving cells (Figs. 2B-C and Supp Fig. 2F). Higher OXPHOS engagement for elongated-mesenchymal cells was confirmed in 3D (Fig. 2D and Supp Fig. 2G). The biosensor PercevalHR ²¹ revealed higher ATP:ADP ratios in elongated-mesenchymal cells migrating in 3D (Figs. 2E-F and Supp Fig. 2H), in agreement with OCR data (Fig. 2C). On the other hand, AMP levels were higher in rounded-amoeboid WM983B and A375M2 cells compared to their elongated-mesenchymal pairs (Fig. 2G and Supp Fig. 2B). Collectively, this data suggests that cells exerting strong traction stress require higher levels of energy from oxidative phosphorylation while cells adopting a rounded-amoeboid migration engage less avidly in mitochondrial respiration.

We have shown that the magnitude of the stress that cells exert on the matrix is connected to the adhesion levels to that matrix (Figs. 1E-H). Considering the data shown above, we hypothesised that levels of adhesion could control the energy demands required for different migratory modes. The analysis of ATP levels comparing adherent cells versus floating cells revealed that when cells are challenged with a non-adherent environment, similar to what we observed in weakly adhesive rounded-amoeboid cells, they decreased ATP levels, particularly ATP derived from mitochondrial respiration (Figs. 2H-I). Accordingly, we observed lower mitochondrial membrane potential in cells under floating conditions (Fig. 2J), in correlation with low OXPHOS in weakly-adherent rounded-amoeboid cells (Figs. 2B-D). Cell viability was not compromised under non-adherent conditions (Supp Fig. 2I). Altogether this data suggests that low adhesion in migrating cells could impact their energetic demands.

DDR1 controls adhesion levels, mode of migration and energy demands.

We next hypothesized that adhesion receptor levels in cells could have an impact in mode of migration and energy requirements. Most of the migratory surfaces including collagens, laminins and fibronectin engage with integrin receptors ²² while Discoidin Domain Receptors (DDRs) also participate in cell adhesion to fibrillar collagens ²³. Using mRNA data from Affymetrix microarrays comparing A375M2 with A375P cells on a collagen I matrix or A375M2 cells with lowered Myosin II activity ⁴, we observed that rounded-amoeboid cells expressed good levels of collagen I binding integrins (Fig. 3A). However, DDR1 mRNA levels were low in these cells (Figs. 3A-B). Importantly, DDR1 protein expression directly correlated with adhesion to the matrix (Fig. 3C) and inversely correlated with cytoskeletal features of “amoeboid-ness” (Fig. 1). Silencing DDR1 in elongated-mesenchymal WM983A and A375P cell lines decreased adhesion to collagen I matrix (Fig. 3D), and induced amoeboid traits: led to increased cell rounding and Myosin II activation, increased membrane blebbing and increased 3D invasion (Figs. 3E-H), without affecting cell viability (Supp Fig. 3A). Similar results were observed using three other on-target siRNA sequences in both cell lines (Supp Figs. 3B-E).

On the other hand, HT1080 and MDA-MB-231 cells exhibited higher levels of β1 integrins compared to A375M2 cells (Supp Figs. 4A-B), in accordance with faster adhesion and higher number of points of attachment to the matrix (Figs. 1E-F). Reducing levels of ITGB1 (β1 integrin subunit) in highly adhesive and elongated cells increased cell roundness, while having no effect on viability (Supp Figs. 4C-E and G). Nevertheless, contrary to the observed rounding response upon DDR1 depletion, roundness upon ITGB1 silencing was not accompanied by any increase in Myosin II activity or membrane blebbing (Supp Figs. 4D and F). Indeed, data from the Affymetrix microarray showed higher expression levels of ITGB1 in A375M2 cells compared to A375P cells or A375M2 cells treated with contractility inhibitors (Fig. 3A). In agreement, rounded-amoeboid WM983B and A375M2 cells expressed higher protein levels of β1 integrins compared to their elongated counterparts WM983A and A375P (Supp Fig. 4H). This data suggests that β1 integrins may be required for both modes of migration in 3D.

Overall, our results indicate that elongated-mesenchymal cells require actomyosin contractility coupled to stronger adhesions to migrate, and define DDR1 as a key collagen I receptor for elongated-mesenchymal movement that is dispensable for rounded-amoeboid migration.

Importantly, DDR1 silencing in elongated-mesenchymal cells cultured on a collagen I matrix resulted in decreased OXPHOS and lower ATP levels (Fig. 3I-J), characteristic of rounded-amoeboid migrating cells (Fig. 2), further confirming the crosstalk between adhesion to the matrix and energy demands.

Energy levels control plasticity of cell migration through AMPK signalling.

Next, we investigated how different energy requirements could result in cytoskeletal remodelling. Rounded-amoeboid cells harbour lower ATP/AMP ratios compared to elongated-mesenchymal cells (Figs. 2C, F and G) and an imbalance in the ATP/ADP or ATP/AMP ratio leads to AMPK activation ²⁴. On the other hand, Myosin II activity is regulated by the phosphorylation of MLC and by Myosin Phosphatase inactivation ^9–11. A chemical genetic screen identified MYPT1 and MBS85, regulatory subunits of the Myosin Phosphatase complex, as new direct substrates of AMPK ²⁵. Phosphorylation of MYPT1 and MBS85 at Ser472 and Ser452 respectively inhibits Myosin Phosphatase activity by preventing the binding of MYPT to Myosin II, leading to Myosin II activation ^25,26. We hypothesized that lower levels of adhesion result in changes in ATP levels that could lead to AMPK activation and subsequent MYPT1 inactivation, MLC2 phosphorylation and rounded-amoeboid migration. We first compared AMPK activity levels (as measured by pThr172-AMPK) in all cell lines and found that rounded-amoeboid cells harbour higher intrinsic levels of pAMPK compared to their elongated-mesenchymal counterparts (Figs. 4A-B). Interestingly, AMPK activation was accompanied with higher levels of phosphorylated, and therefore inactivated, MYPT1 (as measured by pSer472-MYPT1) (Figs. 4A-B), in agreement with higher levels of Myosin II activation at the cortex of rounded-amoeboid cells (Fig. 1C-D). We next treated WM983B and A375M2 rounded-amoeboid cells with Compound C ²⁵, an AMPK inhibitor, resulting in decreased Ser472-MYPT1 phosphorylation (Fig. 4C and Supp Figs. 5A-B). This was accompanied by loss of amoeboid features: decreased cell rounding and reduced Myosin II activity (Figs. 4D-F). Equivalent results were obtained when knocking-down AMPK (PRKAA1 and PRKAA2) using RNAi (Supp Figs. 5C-F). Conversely, treatment of WM983A and A375P elongated-mesenchymal cells with the AMPK activator A769662 ²⁵, led to increased pAMPK and pMYPT1 levels (Fig. 4G and Supp Figs. 5G-H), resulting in the induction of rounded-amoeboid behaviour (Figs. 4H-J). Strikingly, silencing of DDR1 in elongated-mesenchymal cells increased AMPK activity and MYPT1 phosphorylation (Figs. 4K-L) and induced rounded-amoeboid behaviour that was prevented by AMPK inhibition (Figs. 4M-O).

Overall, our data suggests that DDR1-driven adhesion/migration in elongated-mesenchymal cells requires high levels of ATP. By lowering DDR1-dependent adhesion, ATP levels decrease and AMPK is activated, inducing a transition towards rounded-amoeboid migration.

Mitochondrial dynamics in cells adopting different modes of migration in 3D.

By remodelling the mitochondrial network, cells provide energy in specific subcellular regions with high demands. Through mitochondrial dynamics, which includes processes of mitochondrial fission and fusion as well as their transport, cells regulate the size, number, morphology and localization of mitochondria ¹². AMPK has been associated to mitochondrial dynamics through the direct phosphorylation of Mitochondrial Fission Factor (MFF) ²⁷. MFF is one of the main receptors that recruit and stabilize Dynamin related protein 1 (DRP1) to the outer mitochondrial membrane, allowing mitochondrial fission to be carried out ²⁸. We corroborated that activation of AMPK using A769662 in elongated-mesenchymal cells resulted in increased phosphorylation of MFF and increased mitochondrial fission (Figs. 5A-B and Supp Figs. 6A-B) while knocking-down AMPK (siPRKAA1/2) in rounded-amoeboid cells resulted in lower MFF phosphorylation levels and fused mitochondria (Fig. 5C-D and Supp Fig. 6C). This data shows that AMPK levels are connected to both mitochondrial and cytoskeletal dynamics. We therefore decided to study the relevance of mitochondrial dynamics in regulating the cytoskeleton and individual cell migration.

We compared mitochondrial organization in the panel of cell lines growing on a collagen I matrix. Mitochondria of rounded-amoeboid cells (WM983B and A375M2) were punctate and disconnected while mitochondria in their elongated-mesenchymal counterparts (WM983A and A375P), and in HT1080 and MDA-MB-231 cells, were long and formed a network (Figs. 5E-H and Movies 4–5). Elongated mitochondria from mesenchymal cells harboured higher mitochondrial membrane potential than fragmented mitochondria from rounded-amoeboid cells (Fig. 5I). This data together with higher OXPHOS activity detected in elongated-mesenchymal cells (Figs. 2B-D), suggested that mitochondria in these cells are more active. Strikingly, silencing DDR1 resulted in spherical disconnected mitochondria, which was prevented by inhibiting AMPK (Fig. 5J). This data strongly suggests a crosstalk between adhesion, energy levels and mitochondrial dynamics. Overall, mitochondrial fusion and metabolism were higher in elongated-mesenchymal moving cells, possibly to fulfil the energetic demands required for higher levels of adhesion and traction stress.

Perturbing mitochondrial dynamics regulates the cytoskeleton and 3D invasion.

We further analysed the relevance of mitochondrial dynamics for cell migratory plasticity. Mitochondrial morphology is modulated by expression levels, activity and localization of mitochondrial fission and fusion proteins. AMPK activation not only induces a transition to a rounded-amoeboid phenotype but also results in MFF phosphorylation, inducing mitochondrial fission (Fig. 5A-D). Interestingly, preventing mitochondrial fission in rounded-amoeboid cells by silencing MFF, DRP1 or both, resulted in elongated mitochondria, lower levels of phosphorylated AMPK and MYPT, lower myosin activity, loss of amoeboid features and decreased invasion into the matrix (Fig. 6A-F and Supp Fig. 7A).

In addition, we found that elongated-mesenchymal WM983A and A375P cells expressed higher levels of mitochondrial fusion proteins when compared to their rounded-amoeboid counterparts WM983B and A375M2 (Fig. 6G), in accordance with their mitochondrial morphology (Fig. 5E, G and H), which could be a long term adaptation to their energy demands. During mitochondrial fusion, mitofusin 1 and 2 (MFN1 and MFN2) form homo- and hetero-oligomeric complexes to tether and fuse the outer membranes of two mitochondria, while OPA1 Mitochondrial Dynamin Like GTPase (OPA1) directs mitochondrial inner membrane fusion ²⁹. Down-regulation of MFN1, MFN2 or both in elongated-mesenchymal WM983A and A375P cells, not only altered their mitochondrial network to spherical disconnected mitochondria (Figs. 6H-I and Supp Figs. 7B-C), but it also strikingly increased cell rounding, cortical Myosin II activity and membrane blebbing (Figs. 6I-L and Supp Fig. 7C-G). This phenotypic switch resulted in increased 3D invasion, and decreased adhesion to collagen I (Figs. 6M-N and Supp Fig. 7H-I). This data was further validated using four alternative on-target siRNA sequences against MFN1 in both elongated-mesenchymal cell lines (Supp Figs. 7J-M). Similar results were obtained via knocking down OPA1 (regulates mitochondrial inner membrane fusion) in both WM983A and A375P cells (Supp Figs. 8A-H). Cell viability was not compromised after any knock-down (Supp Figs. 7N).

Importantly, rounded-amoeboid behaviour induced by preventing mitochondrial fusion in elongated-mesenchymal cells was mediated by increased AMPK activity and MYPT1 phosphorylation, and it was prevented by AMPK inhibition using Compound C (Figs. 6O-S and Supp Figs. 8H-J).

Altogether, this data indicates that mitochondrial dynamics downstream of cell adhesion play a crucial role in the crosstalk between AMPK and cytoskeletal remodelling in migrating cells.

Adhesion, AMPK and mitochondrial dynamics in vivo.

Tumours derived from WM983B and A375M2 cells have higher levels of Myosin II activity (Figs. 7A-B) and cell roundness compared to tumours from WM983A and A375P cells, respectively ¹⁴, indicating that our in vitro 3D systems replicate cell behaviour in vivo. Accordingly, lower levels of DDR1, MFN2 and OPA1 and higher levels of pAMPK were detected in WM983B and A375M2 tumours, compared to WM983A and A375P tumours (Figs. 7A-B), as shown in vitro (Figs. 3C and 6G).

We have previously shown that cancer cells at the invasive front (IF) of tumours have amoeboid features: they are mainly rounded and harbour higher levels of active Myosin II when compared with cells within the tumour body (TB) ^{4,14,20,30,31}. Accordingly, melanoma cells at the IF of human primary tumours showed lower levels of DDR1, MFN2 and OPA1, while harbouring higher levels of pAMPK and higher amoeboid score (Figs. 7C-D and Supp Fig. 9A). Data showed a positive correlation between mitochondrial fusion proteins and DDR1 levels (Fig. 7E and Supp Fig. 9B). This data suggests that cells in the invasive areas of the tumour harbour lower mitochondrial activity and adhesion.

Localisation at the tumour edge facilitates rounded-amoeboid cell invasion into the surrounding tissue, dissemination and colonization of secondary sites. In order to disseminate, cells need to survive under low-adherence conditions in circulation. Metabolic adaptations found in rounded-amoeboid cells (Fig. 2) may provide a survival advantage in the blood stream. After survival in circulation, cancer cells need to extravasate and colonize the secondary site. Amoeboid behaviour has been previously shown to confer an advantage during lung colonization ^{14,15,20,30,31}. Using well-established experimental metastasis assays, we found that AMPK inhibition in A375M2 rounded-amoeboid cells reduced lung metastatic colonization (Fig. 8A and Supp Fig. 9C), showing that the advantage these cancer cells have in colonizing the lung relies in part on their metabolic rewiring. Finally, tumour edges at metastatic sites resemble those of primary tumours (Fig. 8B) ¹⁴. As such, lower levels of DDR1, MFN2 and OPA1 and higher levels of pAMPK and higher amoeboid scores were also found at the IF of human melanoma metastatic lesions (Figs. 8B-D and Supp Fig. 9D).

Overall, our data shows in vitro and in vivo that intrinsic patterns of adhesion, together with mitochondrial structure and function, govern AMPK-MYPT1-Myosin II activation to facilitate tumour dissemination.

Plasticity of cell migration is crucial during development, immunity and cancer dissemination. We show that adhesion levels are key for cells to sense how much traction stress to apply and how to adapt their metabolism and cytoskeleton to generate the required forces. We also show that downstream of adhesive molecules, mitochondrial dynamics control Myosin II activity via AMPK sensing.

Matrix architecture influences cellular energetics ^32–35 and ECM stiffness has been linked to glycolysis ³⁴ and mitochondrial reprogramming ³⁶ in normal epithelial cells. Also, changes in the ECM stiffness alter mitochondrial dynamics and triggers an enhanced redox response ^36,37. We describe here how AMPK and mitochondrial dynamics operate in cancer cells in physiological 3D matrices to provide the cytoskeletal plasticity required for efficient cell migration. We show that mitochondrial ATP rather than glycolysis is a distinctive feature in 3D cancer cell migration. Mitochondria undergoing fusion form tubular and interconnected networks that are associated with maximal ATP production ²⁹. During elongated-mesenchymal migration, high ATP levels required for establishing strong adhesion rely on highly active and fused mitochondria. Here we show that preventing DDR1-mediated adhesion, ATP levels drop, activating the metabolic sensor AMPK which in turn inactivates Myosin Phosphatase, inducing high Myosin II rounded-amoeboid migration. Importantly, AMPK activation also induces mitochondrial fission, reinforcing the ATP/AMP imbalance and further boosting rounded-amoeboid migration. AMPK has been previously associated to mitochondrial dynamics under metabolic stress conditions. AMPK phosphorylation of S442-MFN2 leads to non-degradative ubiquitination, inhibiting mitochondrial fusion ^38,39. Moreover, AMPK phosphorylation of MFF at S155 and S172, primes DRP1 recruitment to the mitochondrial membrane and stimulates mitochondrial fission ²⁷. Therefore in both cases, AMPK activation results in mitochondrial fragmentation. We report here that rounded-amoeboid migrating cells display higher intrinsic AMPK activation and fragmented mitochondria. In our models -without inducing any further metabolic stress- the sole impairment of mitochondrial fusion by silencing MFN2 or induction of mitochondrial fission due to AMPK driven MFF phosphorylation led to mitochondrial fragmentation and rounded-amoeboid migration. Hypoxia has been reported to promote rounded-amoeboid migration in cancer cells ⁴⁰ and cancer cells under hypoxic conditions display fragmented and less active mitochondria ⁴¹. Altogether, these observations suggest that rounded-amoeboid migration may be an efficient and energetically parsimonious mode of dissemination both in normoxia and under low oxygen conditions ⁴². We open up the exciting possibility that mitochondria orchestrate mechano-responsiveness via AMPK-mediated ATP/AMP sensing.

Changes in mitochondrial structure often drive alterations in mitochondrial metabolism that facilitate metabolic heterogeneity ⁴³. A systematic pan-cancer analysis of metabolic gene expression levels suggested low mitochondrial metabolism in patients with metastatic melanoma, among other cancers ⁴⁴ and mitochondrial dysfunction has been associated with increased cell migration ⁴⁵. We show here that cells with fragmented but functional mitochondria engage in lower mitochondrial metabolism, leading to AMPK activation and rounded-amoeboid invasion. Importantly, we observe that the adhesion receptor DDR1 and mitochondrial fusion proteins are down-regulated in invasive areas of human melanoma tissues while there is an increase in AMPK and Myosin II activity. Our data demonstrates that at the tumour edge, cancer cells have metabolic flexibility to invade and disseminate at a much lower energetic cost, thus posing a great danger. Moreover, this strategy would allow them to escape primary tumours where nutrients have become scarce.

Myosin II activity is regulated by a combination of direct phosphorylation of MLC2 and inactivation of Myosin Phosphatase ^9,10. Several kinases can phosphorylate MYPT, the regulatory subunit, to decrease Myosin Phosphatase activity, resulting in increased Myosin II phosphorylation/activation ¹⁰. MYPT subunits were identified as new substrates of AMPK, the master metabolic sensor ²⁵. pSer472-MYPT1 promotes its interaction with 14-3-3, which impairs MYPT1 binding to Myosin II and inhibits Myosin Phosphatase activity ²⁶. In the present study, we show how AMPK activation in low adherent cells results in pSer472-MYPT1 and leads to increased Myosin II-dependent 3D invasion. Most studies have focused on the direct regulation of Myosin II activity by ROCK (via MLC2 phosphorylation at Thr18 and Ser19, and to a lesser extent via phosphorylation of MYPT1 at Thr696 and Thr853) ^9–11. Our observations are important to further understand the different levels of Myosin II regulation in migrating cells. While there is scarce evidence for other kinases regulating MLC2 function via MYPT in migrating cells, we suggest that AMPK phosphorylation of Ser472-MYPT1 may be a key controller of tumour cell migratory plasticity downstream of mechano-metabolic signals.

β1 integrins are essential for both rounded-amoeboid and elongated-mesenchymal modes of movement. Despite the low degree of attachment, rounded-amoeboid melanoma cells appear to have higher levels of β1 integrins than elongated-mesenchymal melanoma cells. β1 integrins have been linked previously to rounded-amoeboid migration ^1,2,46,47. Interestingly, we show that DDR1 controls the balance between individual cell migration modes and their metabolic rewiring. Previous reports showed that DDRs enhance integrin-mediated adhesion to collagen ^23,48, while loss of DDR1 increases contractility in breast tumours ⁴⁹. We propose that low DDR1 coupled to moderate levels of β1 integrins provide the required degree of adhesion to collagen for cancer amoeboid 3D migration.

We have previously shown that Myosin II activity provides pro-survival signals in cancer cells under drug treatments ¹⁵ or under anoikis-inducing conditions ³¹. To metastasize, cells leave the primary tumour and need to survive in circulation. Matrix detachment during dissemination has been linked to metabolic rewiring ⁵⁰ and Myosin II activity protects circulating cancer cells from shear stress in the blood ⁵¹. We show here how cells in suspension, similar to cells in circulation, exhibit lower OXPHOS levels. Furthermore, we describe how rounded-amoeboid migration is also characterised by low levels of OXPHOS and their metastatic colonization abilities rely on AMPK signalling. We speculate that by lowering their mitochondrial metabolism, circulating tumour cells increase their Myosin II activity, therefore promoting cancer cell survival in low-adherence environments. We propose that rounded-amoeboid cells are primed for dissemination and metastasis by entering a low-energy requirement mode that may allow them to survive under stressful conditions.

Our work unveils how mitochondrial dynamics control the cytoskeleton and cell migration and place AMPK as a key “mechano-metabolic sensor” during tumour dissemination.

Experimental model and subject details

Cell lines

HT1080 fibrosarcoma cells were from Prof Chris Marshall (ICR, UK), MDA-MB-231 breast cancer cells were from Prof Clare Isacke (ICR, UK), A375P and A375M2 cells were from Prof Richard Hynes (HHMI, MIT, USA) and WM983A and WM983B were purchased from Coriell Institute (USA). All cell lines were grown at 37^oC and 10% CO₂ in DMEM Supplemented with 10% FBS and 1% penicillin/streptomycin (all from GIBCO). A375P, A375M2, WM983A and WM983B were authenticated using short tandem repeat DNA profiling. All cell lines were routinely tested for mycoplasma contamination. All cell lines were kept in culture for a maximum of three to four passages and cell phenotypes were verified routinely.

For actin imaging in live cells, cells stably transfected with a LifeAct-GFP plasmid (provided by Dr Rikki Eggert, KCL, UK) were used.

Patient derived tissues

Two tissue microarrays including FFPE biopsies of 53 human primary melanomas and 45 metastasis were included in the case series (for clinical information see Supplementary Tables 1 and 2). Each biopsy was represented by two cores (1 mm diameter) from the tumour body (TB) and two cores from the invasive front (IF) areas. Tumours were classified following the most recent World Health Organization criteria. Tumour samples were processed by IRBLleida (PT17/0015/0027) and HUB-ICO-IDIBELL (PT17/0015/0024) Biobanks integrated in the Spanish National Biobank Network and Xarxa de Bancs de Tumors de Catalunya following standard operating procedures with the appropriate approval of the Ethics and Scientific Committee. Samples were collected with specific informed consent, in accordance with the Helsinki Declaration.

Animal studies

All animals were maintained under specific pathogen-free conditions and handled in accordance with the Institutional Committees on Animal Welfare of the UK Home Office (The Home Office Animals Scientific Procedures Act, 1986). All animal experiments were approved by the Ethical Review Process Committees at Barts Cancer Institute and King’s College London, in accordance with the Animals (Scientific Procedures) Act 1986 and according to the guidelines of the Committee of the National Cancer Research Institute.

Animal derived tissues

Tumours from WM983A-, WM983B-, A375P- and A375M2-EGFP cells injected subcutaneously into severe combined immunodeficient mice (SCID; CB17/Icr-Prkdcscid/IcrIcoCrl) were used for immunohistological purposes. Tumours were generated in a previous study ¹⁴.

Lung colonisation assay

For experimental metastasis assays, A375M2 cells pre-treated in vitro for 24 hours with either DMSO or Compound C (2 µM) were labelled with 10 µM CMFDA-Green (C7025, Life Technologies) for 10 min and then trypsinized and counted. 1x10⁶ labelled cells / 0.2 ml PBS along with drugs (same concentration as pre-treatment) were injected into tail vein of NOD/SCID/ IL2Rγ-/- mice (NSG, Charles River). Mice were sacrificed 30 minutes (to confirm that equal numbers arrived at the lung) and 24 hours after tail vein injection. Lungs were extracted, washed with PBS (with calcium/magnesium) twice and fixed with 4% formaldehyde for 16 h at 4°C. Lungs were examined under a Zeiss LSM 710 Meta confocal microscope (Carl Zeiss) with a 20X objective. Data are presented as percentage of field of area covered by fluorescence, and 15 fields per mouse were analysed. n = 7 mice/condition for each experiment (2 mice sacrificed 30 minutes and 5 mice 24 hours after tail vein injection), n = 2 independent experiments.

Chemicals

Chemicals used in this study: Myosin II inhibitor blebbistatin (Calbiochem; resuspended in 95% DMSO; used at 25 µM), antimycin A (Fisher Scientific; resuspended in DMSO; used at 1 µM), oligomycin (Sigma-Aldrich; resuspended in DMSO; used at 1 µM), FCCP (Tocris Bioscience; resuspended in DMSO; used at 0.75 µM), Compound C (Sigma-Aldrich; resuspended in DMSO; used at 2 µM) and A769662 (Tocris Bioscience; resuspended in DMSO; used at 10 µM).

Antibodies

Antibodies and concentrations used: pSer19-MLC2 (#3671; 1:200 immunofluorescence; 1:50 immunohistochemistry), DDR1 (#5583; 1:1000 immunoblot, 1:200 immunohistochemistry), DRP1 (#8570; 1:1000 immunoblot), MFF (#86668; 1:1000 immunoblot), MFN1 (#14739; 1:1000 immunoblot), pThr172-AMPK (#2535; 1:1000 immunoblot; 1:150 immunohistochemistry), AMPK (#2532; 1:1000 immunoblot), MYPT (#8574; 1:1000 immunoblot), from Cell Signaling Technology; pSer172/Ser146-MFF (AF2365, 1:1000 immunoblot), pSer472-MYPT1 (AF3779; 1:1000 immunoblot), from Affinity Biosciences; GAPDH (MAB374; 1:10000 immunoblot) from Merck-Millipore; Tubulin (T6199; 1:10000 immunoblot) from Sigma-Aldrich; Integrin β1 (ab24693; 1:100 immunofluorescence), MFN2 (ab56889; 1:1000 immunoblot, 1:500 immunohistochemistry), CD44 (ab157107; 1:700 immunohistochemistry) from Abcam; OPA1 (612606; 1:1000 immunoblot, 1:500 immunohistochemistry) from BD Transduction Laboratories; Tom20 (sc-17764; 1:800 immunofluorescence) from Santa Cruz Biotechnology.

Cell culture on 3D collagen I matrices

Fibrillar bovine dermal collagen (PureCol, Advanced BioMatrix) was prepared at 1.7 mg/ml in DMEM (300 µl/well for a 24-well plate; 100 µl/well for a 96-well plate). After collagen polymerization (4 hours at 37^oC – 10% CO₂), cells were seeded on top in DMEM 10% FBS, allowed to adhere for 24 hours and treatments added (where appropriate). To embed cells within the collagen matrix, cells were suspended in the collagen and the appropriate volume of the suspension was seeded (100 µl/well for a 96-well plate; 250 µl/well for optical bottom 8-well µ-slide). The collagen matrices containing cells were allowed to polymerize for 4 hours at 37^oC – 10% CO₂. Then DMEM 10% FBS was added on top. All assays were performed with cells seeded on top of a thick layer of collagen unless otherwise mentioned.

siRNA transfection

Reverse transfection was used to transiently down-regulate the indicated genes. Two hundred and fifty thousand cells were seeded in complete media (DMEM 10%FBS) in 6-well plates with a mix containing 20 nM siGENOME SMARTpool or individual On-Target (OT) siRNA oligonucleotides (Dharmacon), Optimem-I and Lipofectamine 2000 (Invitrogen). Non-targeting siRNA were used as control. 24–48 hours after transfection cells were split and seeded for the downstream experiments/treatments. All siRNA sequences were from Dharmacon (Lafayette, USA) and are listed in Supplementary Table 3.

Immunoblotting

Cells were lysed in Laemmli Lysis Buffer and snap frozen. Then, lysates were boiled for 5 minutes, sonicated for 15 seconds and spun down. Cell lysates were resolved by SDS-polyacrylamide gels (SDS-PAGE) in non-reducing conditions and transferred to PVDF filters (0.45 µm, Immobilion™). Membranes were blocked in 5% BSA in 0.1% Tween 20-TBS. Primary antibodies were incubated overnight at 4^oC. ECL Plus or Prime ECL detection System (GE Healthcare) with HRP-conjugated secondary antibodies (GE Healthcare) were used for detection. Bands were quantified using Fiji software (http://fiji.sc).

Collagen imaging

Imaging was performed with a Zeiss LSM 510 Meta confocal microscope (Carl Zeiss) with C-Apochromat X 40/1.2 NA (water) objective lenses and Zen software (Carl Zeiss). Reflectance imaging of the matrix was performed by collecting the backscattered light. Since reflectance is a surface property, extreme caution was taken to ensure the acquisition of images at the same height across all conditions.

Points of attachment

Collagen I was imaged via reflectance microscopy and points of attachment were assessed by counting the points of collagen attachment to the actin cytoskeleton (visual assessment) through consecutive z-slices. The points of attachment were manually counted using the “Cell counter” tool.

Measurement of traction stress

Live, 3D time-lapse videos coupled with image-processing were used to measure stress applied by cells within 3D collagen environments. Cells labelled with LifeAct-GFP were embedded within 3D Collagen I matrices polymerized in tissue culture wells with optical plastic bottom and imaged directly “in-well” using an inverted confocal microscope, for 8 hours. Cells were seeded such that the cellular density was sparse enough such that only one cell was present within an area of at least 250 x 250 x 50 mm in the middle of the 3D collagen matrix. This was done to ensure that the displacements measured within a certain field were only due to stress generated by the cell within the field of view.

Collagen I was imaged by collecting the reflected/backscatter light from the matrix using laser-scanning confocal microscopy ^19,20,52. This approach was selected due to ease of access and lack of alterations/artificial manipulations to the matrix. By comparison of collagen I immunostained with an anti-collagen I antibody, it has been demonstrated that there is negligible detection error from reflectance imaging of Collagen I ¹⁹. The detection error arises from backscatter-negative fibrils in vertical orientation and is below 3%.

Computation of displacements and strains

Videos were generated by acquiring 3D image stacks of 30 mm thickness with step size of 1 mm. The optimal interval for acquisition of images was determined to be intervals of 2 minutes to avoid phototoxicity. The resultant videos allowed for resolution of individual collagen I fibres surrounding each cell within the x-y plane, which represents the migration of LifeAct-GFP expressing cell (mid-plane) within the surrounding matrix over 10 minutes. This set-up allowed for the optimal resolution of collagen fibres within the z-plane. The videos were then processed using a custom-written Particle Image Velocimetry (PIV) based tracking algorithm in Wolfram Mathematica 10. In order to track the matrix, 3D videos were deconstructed and analysed on a frame-by-frame basis. The deconstruction of the videos resulted in a 512x512x90 image stack corresponding to each time point. The 90 positions correspond to the interpolation along the z-axis. Superimposition of a mesh grid on the image allowed for the tracking of the position of each pixel in one frame, with respect to its relative position in the adjacent frame, at the next time point in 2D x-y plane and, in the 3D case, through each slice of the z-stack. The difference between these positions was used to calculate displacement for all three dimensions. In cases where the corresponding point in an adjacent frame could not be found, the region would be labelled zero and would then be linearly interpolated to the nearest 9 neighbouring pixels. After analysis in all three planes (xy, xz, yz) the 3D axes was then rotated and adjusted in order to reconstruct the 3D displacement data. Cell displacements calculated in X, Y and Z corresponding to each pixel in the 512x512x90 frame stacks, were converted to X, Y and Z strain values using the inbuilt derivative filter function ‘DerivativeFilter’, in Mathematica. The conversion of displacements to strains was based on the displacement-gradient technique ⁵³. Strain data was then exported as a MATLAB file for traction stress computation in the form of a zip folder, consisting of 90 files each containing 512 variables, which in turn consisted of 512x512x3 matrices of 3D strain values.

For ease of analysis during the initial phase of development of the method, collagen displacements were measured in 2D images representing 3D projections of cells embedded within matrix with depth information incorporated. This approach allowed for the tracking of collagen displacements in 3D via maximum intensity based projections of cells in 2D.

Cell traction stress computation

In order to calculate stress from strains, the stiffness of the bovine collagen matrix was calculated using AFM microscopy. Stiffness of bovine collagen was ~ 19 Pa, comparable to the stiffness observed in other studies ^19,54. Stress were then calculated by integrating the strain data, the directional information, depth information and stiffness of the matrix. A custom algorithm was written in MATLAB, using a ‘direct Traction Force Microscopy (TFM)’ method ⁵⁵, and stress were directly calculated from computed strains, as opposed to more complicated and computationally extensive methods involving the solution of inverse equations and boundary conditions. The mathematical approach was based on that used by ¹⁸, following the assumption of a linearly isotropic material or substrate. According to this assumption, the Cauchy relation for traction stress, τ, is:

τ = δ.n Eq. (1)

where ‘δ’ is the Cauchy stress tensor and ‘n’ the direction of the normal vector.

The MATLAB code was designed to process each of the aforementioned 512 variables, setting the X, Y and Z strain values as the principal strains (diagonal elements) of a 3D symmetric strain tensor ‘ε’, defined as:

ε = \(\left[\begin{array}{ccc}{\epsilon }_{11}& {\epsilon }_{12}& {\epsilon }_{13}\\ {\epsilon }_{21}& {\epsilon }_{22}& {\epsilon }_{23}\\ {\epsilon }_{31}& {\epsilon }_{32}& {\epsilon }_{33}\end{array}\right]\)Eq. (2)

Equation (3) below was solved in order to obtain the shear modulus ‘µ’, which was then input into Eq. (4), to obtain the stress tensor ‘δ’, for which isotropic linearly elastic properties are assumed ⁵⁴. The constitutive mechanical properties required for these calculations included a Young’s modulus ‘Ε’ of 28 Pa, measured for our collagen matrix, and a Poisson ratio ‘v’ of 0.25 based on that measured by Steinwachs et al (2016).

Ε = 2µ (1 + v) Eq. (3)

δ = 2µε Eq. (4)

The MATLAB function ‘eig’ was then called to solve the eigen Eq. (5) for the three eigenvalues (λ1, λ2, λ3) and eigenvectors (V₁, V₂, V₃) of this stress tensor. The eigenvalues correspond to the traction stress solutions of the stress tensor ‘δ’, whilst the eigenvectors correspond to the direction of the normal ‘n’, relative to the surface on which the traction stress is acting.

Ε = (δ-λI) V Eq. (5)

where ‘Ε’ represents the set of vectors which satisfy (δ - λl) V = 0.

The matrix dot product of the stress eigen values with their correct and corresponding eigen vectors, yields the traction stress ‘τ’, in the Cauchy relation (Eq. 1). This step is more complicated in 3D, compared to 2D, as there are 9 different eigen vector permutations and therefore combinations of the eigen vectors with their respective eigen values.

Appropriate considerations therefore had to be incorporated into the code in order to deal with this difficult step whilst being able to process the data efficiently.

Once the traction stress had been computed in X, Y and Z (Τ_x, Τ_y, Τ_z), the magnitude of these stress was computed using the standard vector combination relation:

\(\left|\text{T}\right| = \sqrt{{T}_{x}^{2}+{T}_{y}^{2}+{T}_{z}^{2}}\) Eq. (6)

By this stage, one data set would consist of 90 individual 512x512 matrices of traction stress magnitudes corresponding to each pixel. Before generating a visual representation, it was important to vertically flip and rotate this matrix anticlockwise by 900 in order to correct for the default image conversion adjustments made by MATLAB and Mathematica software, respectively. This also ensured accurate analysis of cell movements later on, when cell outlines were overlaid on the maps. After processing each data stack, a 512x512 pixel colour map was generated for each slice in the stack using a ‘jet’ LUT, intrinsic to MATLAB. According to this colour scheme, the resulting gradient based stress maps were colour coded such that the red areas of the reaction maps correspond to traction stress while blue regions correspond to minimal or zero traction stress and are found in areas of matrix where no substantial impact due to cell movement was detected.

Cell morphology on collagen

Cell morphology of cells seeded on collagen matrices was quantified on still phase-contrast images using ImageJ. Cell morphology was assessed using the morphology descriptor tool “roundness” after manually drawing around the cell. Values closer to 1 represent rounded morphology; values closer to 0 represent more spindle-shaped cells.

Cell adhesion to collagen I

Collagen matrices were prepared in 96-well plates. Ten thousand cells/well were seeded in duplicate or triplicate in the 96-well plates and incubated at 37^oC for the indicated times. The centre of the well was then imaged with a phase contrast microscope, washed three times with PBS and imaged again with the same settings. Each well was imaged before and after washing. Cells were counted manually using “Cell counter” function in ImageJ and results are presented as percentage of adhered cells, calculated as number of cells after washing divided by the number of cells before washing.

3D invasion

For 3D invasion assays, cells were resuspended in serum-free bovine collagen I solution at 2.3 mg/ml to a final concentration of 14 000 cells per 100 µl of matrix and spun down, in a 96-well plate. After the matrix was polymerized, 10% FBS-containing media was added on top of the matrix. After 24 hours cells were fixed, stained with Hoechst and imaged using a Zeiss LSM 710 confocal microscope. Invasion was calculated as number of invading cells at 50 µm divided by the number of cells at the bottom.

Immunofluorescence and confocal imaging

Cells seeded on top of collagen matrices were fixed with 4% formaldehyde for 15 minutes at room temperature. Then, cells were permeabilised for 20 minutes using 0.3% Triton X-100 in 5% BSA-PBS, blocked in 5% BSA-PBS for 30 minutes and immunostained with primary antibody overnight at 4^oC. Next, samples were incubated with Alexa Fluor 546-phalloidin and Alexa Fluor 488 or 647 secondary antibodies for 2 hours at room temperature. Nuclei were stained with Hoechst prepared in PBS. Antibodies were diluted in 5% BSA-PBS.

Images were taken with a Zeiss LSM 510 Meta confocal microscope with Plan-Apochromat 40x/1.2 NA (water) objective lenses, Zeiss LSM 710 confocal microscope with Plan-Apochromat 40x/1.3 Oil DIC M27 and Zeiss LSM 880 confocal microscope with Airyscan superresolution mode and Plan-Apochromat 63x/1.4 Oil DIC M27 objective lenses (Carl Zeiss, Germany). Zen software was used to acquire images (Carl Zeiss, Germany). Images were analysed using ImageJ software (NIH). For pMLC2 quantification, fluorescence signal was quantified by calculating the area occupied by pMLC staining in single cells relative to the cell area. For pMLC2 distribution, line scan analysis was performed in ImageJ.

Mitochondrial mass and mitochondrial activity

Mitochondrial activity and mitochondrial mass comparing adherent versus floating cells were analysed by flow cytometry. Cells were seeded either on regular cell culture plates or polyhema pre-coated plates (for non-adherent conditions). After 24 hours cells were collected, spun down and resuspended with a solution containing MitoTracker Deep Red (25nM) and TMRE (200nM), at 37^oC for 20 minutes. After, FACS buffer (1% BSA, 2mM EDTA and 0.1% NaN₃ in PBS) was added to samples and immediately analysed on a BD FACS CANTO II flow cytometer. Data was analysed using FlowJo software (Tree Star).

Mitochondrial activity and mitochondrial mass comparing cells on top of collagen were analysed by confocal imaging. Live cells expressing LifeAct-GFP seeded on top of a collagen I matrix were incubated with MitoTracker Deep Red (25 nM) for mitochondria visualisation and with tetramethylrhodamine ethyl ester (TMRE; 200 nM) for mitochondrial activity analysis, at 37^oC for 20 minutes. Then, were kept in FluoBrite DMEM for imaging. Z-stack images were taken with a Nikon Eclipse Ti Inverted equipped with a Yokogawa CSU-X1 Spinning Disk unit 100x oil objective lenses. Images were analysed using ImageJ software (NIH). Mitochondrial activity (TMRE signal) was quantified from maximal intensity projections and normalized by mitochondrial mass (MitoTracker Deep Red signal).

Mitochondrial morphology

For mitochondrial circularity analysis, MitoTracker Deep Red z-stacks images were imported in to Mathematica 11, converted to 3D and decomposed in to frequency bands using difference of Gaussian filters using radii ranging from 1-to-32 in powers of 2. This enabled simultaneous smoothing and background subtraction. The frequency-decomposed images were then combined to create a frequency-encoded image. The images were then Top-hat transformed with a radius of 4 to highlight the mitochondria and reduce uneven intensities. Filtered images where then thresholded and binary objects that were less than ten pixels in size were deleted. The internal Mathematica function “Component Measurements” was then used to calculate the circularity of each object of the image. These were then exported as “csv” files for further analysis.

For the analysis of mitochondrial branches per cells, a semi-automated image-based analysis was performed in ImageJ using Mitochondria Analyzer ⁵⁶. Single plane images of Tom 20 fluorescently-labelled mitochondria in cells grown on a collagen I matrix were pre-processed and converted to 8-bit images. Next, 2D Optimize Threshold command was used to identify the appropriate settings on test samples from each set acquired under similar imaging conditions. Block size of 2.25 µm and C-value of 28 was used for the 2D Analysis in batch mode. Analysis was obtained on a per-cell basis.

NMR metabolomics

The metabolic profiles were assessed using NMR metabolomics. Three million cells were seeded in 14.5 cm dishes 48 hours prior to dual-phase metabolite extraction. Media was removed and cells were washed twice with cold PBS. Then, each dish was placed on dry ice for 2 minutes to quench the metabolome. Afterwards, 2.2 mL of cold methanol was added and cells were scraped and transferred to a pre-cooled 15 mL tube. 2.2 mL of cold chloroform was added to each tube and mixed in a tube rotor for 10 minutes at 4^oC. Then, 2.2 mL of milli-Q water was added to each tube and the contents were mixed by inversion and incubated on ice for 10 minutes to allow the formation of a stable bilayer. Samples were then centrifuged at 4^oC and 1550 rpm for 45 minutes to separate the top layer and a bottom layer, containing metabolites soluble in the polar and apolar fraction, respectively. The polar fraction was dried in an Eppendorf concentrator (Concentrator plus, Eppendorf).

Prior to NMR experiments, samples were resuspended in a 90/10 H₂O/D₂O buffer, with 100 mM Na₂HPO₄, 5 mM TSP and 4 mM NaN₃. NMR spectroscopy was performed on a Bruker Avance NEO 600 MHz NMR spectrometer equipped with a TCI Prodigy CryoProbe (Bruker). Spectra were acquired at 298 K and consisted for each sample of a 1D ¹H PURGE spectrum ⁵⁷, which was then phase corrected, baseline corrected and the chemical shifts were referenced to the TSP peak at 0.0 ppm.

Spectra were aligned by the PAFFT method (peak alignment by fast Fourier transform) and then normalized by probabilistic quotient normalization (PQN) ⁵⁸. To determine whether spectra cluster into groups, PLS-DA with k-fold cross-validation were performed. PLS-DA of the full 1D ¹H spectrum, after alignment, normalization using PQN, and log-scaling of the spectra was done. For multivariate analyses, spectra were scaled by the pareto method, which decreases the influence of high intensity peaks while emphasizing low intensity signals, such that both high and low intensity signals have equal significance in the model ⁵⁹. ¹H-NMR profiling of these cellular extracts allowed the identification of the main compounds present in the mixture. The metabolite assignment was done by comparing peak chemical shifts to those found in literature, in Human Metabolome Database (HMDB) and in the Biological Magnetic Resonance Bank (BMRB). Furthermore, all the metabolite structures were confirmed by 2D NMR experiments. Relative quantification of metabolites was achieved by integration of peaks to calculate peak area, which is proportional to the number of nuclei that give rise to the signal. Mean ± SD of peak areas were calculated for each metabolite.

Metabolic assays

ECAR and OCR measurement

In 2D systems, oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were measured performing a Mitochondrial stress test and a Glycolysis stress test respectively, using a Seahorse XFe96 analyser (Agilent), following manufacturer instructions. Briefly, 5 000 cells were seeded on Seahorse XF96 cell culture microplates in DMEM 10% FBS and incubated overnight at 37^oC and 10% CO₂. Next, media was changed to XF DMEM media, supplemented as indicated by the manufacturer and cells were incubated at 37^oC without CO₂ for 1 hour, before proceeding immediately to perform the OCR or ECAR analysis. OCR was monitored following the sequential addition of oligomycin (1 µM), FCCP (0.25 µM) and rotenone/antimycin A (0.5 µM). ECAR was monitored following the sequential addition of glucose (10 mM), oligomycin (1 µM) and 2-deoxy-D-glucose (2DG, 50 mM).

For the measurement of OCR in 3D cell cultures, MitoXpress Xtra Oxygen Consumption assay (Agilent) was used. Cells were seeded embedded in a 3D collagen matrix as explained above. Twenty-four hours later, media was replaced with 60 µl DMEM 1% FBS containing MitoXpress Xtra reagent. Then, wells were sealed with HS mineral oil and time-resolved fluorescence (TR-F) was measured immediately for 2 hours, using a BMG Omega plate reader, following manufacturer instructions. Glucose oxidase (0.1 mg/ml) was used as a cell-free positive control. Antimycin A (1 µM) and FCCP (0.75 µM) were used as cell-based negative and positive control, respectively. Slope between 30 and 90 minutes was used to quantify OCR, following manufacturer’s protocol.

Cellular ATP measurement

Luminescent ATP detection assay

For the measurement of total ATP in A375P cells, the luminescent ATP detection assay (ab113849, Abcam, Cambridge, UK) was used. Briefly, the day before the assay 5 000 A375P cells/well were seeded on untreated and polyhema-treated (P3932, Sigma-Aldrich) 96 well plates and allow to grow overnight in DMEM 10% FBS, incubated at 37^oC and 10% CO₂. A375P cells cultured on polyhema plates were grown as suspension cells. The day after, ATP was quantified according to the manufacturer’s protocol.

Seahorse XF Real-Time ATP Rate Assay

For the quantification of ATP production rate from both glycolytic and mitochondrial pathways, Agilent Seahorse XF Real-Time ATP Rate Assay was performed according to the manufacturer’s instructions (Agilent Technologies, Santa Clara, CA) for adherent and floating cells. Briefly, the day prior the assay, 5 000 A375P cells/well, previously grown under adherent conditions, were seeded on a Seahorse XF96 cell culture microplate in DMEM 10% FBS, incubated at 37^oC and 10% CO₂. On the day of the experiment, media was changed for XF DMEM (pH = 7.4) supplemented with 1 mM pyruvate, 2 mM glutamine and 10 mM glucose (Agilent, USA). Furthermore, 7 500 A375P cells, previously grown under floating conditions for 24 hours, were seeded on the Seahorse XF96 cell culture microplate in the XF media specified above. Both adherent and floating cells were then incubated at 37^oC without CO₂ for 1 hour. During this incubation period, oligomycin and rotenone/antimycin A (Agilent, USA) were prepared in Seahorse media to achieve final concentrations of 1.5 µM and 0.5 µM respectively when injected. The glycolytic and mitochondrial ATP production rate were calculated according to manufacturer instructions. The total ATP production rate is the sum of the glycolytic and mitochondrial ATP production rates.

Finally, the ratio between mitochondrial ATP and glycolytic ATP production rate as well as the ratio between mitochondrial ATP and total ATP production were represented.

Measurement of ATP:ADP ratio in 3D

PercevalHR lentivirus were generated in HEK293T cells. Media with lentiviruses was added to recipient cell lines HT1080, A375M2 and A375P. PercevalHR expressing A375P cells were then transfected with DDR1 siRNA following the protocol as explained above. After 48 hours cells were trypsinized, spun down and re-suspended in collagen I. Two hundred and fifty microliters of collagen-cell suspension were added per well in optical bottom 8-well µ-slide (Ibidi). Collagen was allowed to polymerize for 4 hours at 37^oC and DMEM 10% FBS was added on top. Twenty-four hours later, media was replaced by FluoBrite DMEM for imaging. Perceval HR expressing cells were sequentially excited using 405 and 488 nm lasers with emission collected at 540 nm for both channels ^21,60. Z-stack images for the entire volume of the cell were taken using Zeiss LSM 880 with Fast Airyscan mode, Plan-Apochromat 63x/1.4 Oil DIC M27 objective lenses and Zen software (Carl Zeiss, Germany). Pixel-by-pixel ratio of ATP bound (488) / ADP bound (405) signals were calculated using ImageJ RatioPlus tool.

Cell viability

Propidium iodide was used to stain dead cells. After spinning down spent media and cells, cells were resuspended in a solution of Propidium iodide (10 µg/ml) in PBS and incubated for 15 minutes on ice. Immediately afterwards, samples were analysed on a BD LSR Fortessa flow cytometer. Data was analysed using FlowJo software.

Immunohistochemistry

Paraffin-fixed sections were sequentially stained for DDR1, MFN2, OPA1, pMLC2, pAMPK and CD44, as previously described ^15,61. Tumour tissues were formalin-fixed and paraffin embedded as per standard protocols. Sections of 4 µm thickness were heated at 60^oC for 1 hour and then incubated in xylene and ethanol series, with 2x 5 minutes H₂O₂/ethanol incubations to block endogenous peroxidase. Antigen retrieval was performed in Antigen Unmasking Solution pH 6 (H-3300, Vector Labs) using a pressure cooker system (110^oC for 10 minutes). Samples were washed in Dako Wash Buffer (S3006), before primary antibody incubation (40 minutes, 1:200 anti-DDR1, CST #5583), diluted in Antibody Diluent Reagent Solution (003218, invitrogen/ThermoFisher Scientific). Samples were washed and incubated with ImmPRESS® polymer secondary goat anti-rabbit antibody (goat anti-rabbit, RTU, Vector Labs, MP-7451) for 45 minutes. The reaction was developed using VIP peroxidase substrate solution (Vector Labs, SK-4600) for 10 minutes. All incubations were carried out at room temperature. Slides were counterstained with haematoxylin and mounted using DPX mounting medium (06522-500ML, sigma). Slides were imaged using the NanoZoomer S210 slide scanner (Hamamatsu, Japan). The next day, slides were processed using the same procedure. Previous staining was stripped through the antigen retrieval step. In the second round, an anti-MFN2 primary antibody was used (40 minutes, 1:500, ab56889) and ImmPRESS® polymer secondary goat-anti mouse antibody (40 minutes, RTU, Vector Labs, MP-7452); developing, mounting and imaging was performed as before. The third round was performed as above, where slides were incubated with an anti-OPA1 antibody (40 minutes, 1:500, BD-612606) and ImmPRESS® polymer secondary goat-anti mouse antibody (40 minutes, RTU, Vector Labs, MP-7452) was used; developing, mounting and imaging was performed as before. In the forth round, anti- pMLC2 (ser19) antibody was added to the panel (40 minutes, 1:50, CST #3671) and ImmPRESS® polymer secondary goat-anti rabbit antibody (40 minutes, RTU, Vector Labs, MP-7451) was used; developing, mounting and imaging was performed as before. In the following fifth round, pAMPK antibody (40 minutes, 1:150, CST#2535) and ImmPRESS® polymer secondary goat-anti rabbit antibody (40 minutes, RTU, Vector Labs, MP-7451) were used; developing, mounting and imaging was performed as before. Finally, the sixth round included CD44 antibody to stain cell membranes which allowed the quantification of cell morphology. Tissue sections were incubated with anti-CD44 (40 minutes, 1:700, ab157107) and ImmPRESS® polymer secondary goat-anti rabbit antibody (40 minutes, RTU, Vector Labs, MP-7451) was used; developing, mounting and imaging was performed as before.

Image Analysis

Staining quantification was performed using QuPath 0.1.2 (Bankhead et al., 2017). For each marker, whole section images (WSI) from mouse tissue and TMAs were analysed performing positive cell detection, and three different thresholds were applied according to the intensity scores (0, 1, 2 and 3). Next, the software was trained by creating a random tree classification algorithm combined with the intensity information, in order to differentiate tumour from stroma, necrosis and immune cells. Then, IHC staining was graded semiquantitatively by considering the percentage and intensity of the staining. A histologic score (Hs) was obtained from each sample, and values ranged from 0 (no immunoreaction) to 300 (maximum immunoreactivity) arbitrary units. The score was obtained by applying the following formula, Hs = 1 × (% light staining) + 2 × (% moderate staining) + 3 × (% strong staining). For co-localization analysis, images for the 5 markers markers (DDR1, MFN2, OPA1, pMLC2 and pAMPK) were aligned in FIJI v1.52p using TrackEM2 module. Next, colour deconvolution was performed using AEC-Haematoxylin vectors and a composite was created using channel-2 (red) for each staining. The composite was adjusted inverting the LUT for each marker and given a pseudocolour. We used QuPath DoG superpixel segmentation to quantify cell morphology. CD44 staining was used to perform an accurate watershed segmentation. Finally, shape measurements were added in QuPath such as rounding in all the detections. To calculate the amoeboid score (As) the same ROI used in CD44 for each case we used with pMLC2 staining and the intensity for such marker was calculated in each detection. The As was calculated combining roundness and pMLC2 intensity for each detection, as shown in the formula: As = roundness (0–1) * Hs pMLC2 (0-300) = 0-300

Values for As range from 0 to 300 were 0 represents a very elongated cell (roundness = 0) with null levels of Myosin II (Hs = 0), while the representative amoeboid cell harbours As of 300. For each case values are represented as mean of all detection per ROI.

Quantification and statistical analysis

Statistical analyses were performed using GraphPad Prism 8 software (GraphPad, San Diego). The following statistical analysis were used: two-sided Student’s t-test, Mann-Whitney, Wilcoxon test, two-sided one-way ANOVA with Tukey’s or Dunnett’s post hoc test and two-way ANOVA with Dunnett’s post hoc test. Data were represented in bar graphs and dot plots as mean ± standard error of the mean (SEM) and box plots as median (centre line), interquartile range (box) and min-max values (whiskers). In general, experiments were carried out at least 3 independent times with 2–3 technical replicates, unless otherwise stated. Significance was defined as p < 0.05. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

MATERIALS AVAILABILITY

The code generated during this study will be available at GitHub.

Supplementary Data 1. Script for 3D traction stress calculation.

All unique/stable reagents generated in this study are available with a completed Materials Transfer Agreement.

ACKNOWLEDGMENTS

The work was supported by Cancer Research UK (CRUK) C33043/A12065 and C33043/A24478 (VSM, ECM, OM, PP); Barts Charity (VSM, VG, JM, RS, SLG); Royal Society RG110591 (VSM); Fundación Ramón Areces (ECM); Word Wide Cancer Research 22-0329 (VSM, VG). We thank the Centre for Biomolecular Spectroscopy for access to biophysical infrastructure. The Centre was funded by the Welcome and British Heart Foundation grants to MRC (ref. 202767/Z/16/Z and IG/16/2/32273 respectively). We thank Rosa M Marti, Xavier Matias-Guiu (Hospital Universitari Arnau de Vilanova, University of Lleida, IRB Lleida) for providing patient samples, and Kairbaan Hodivala-Dilke, Jeremy Carlton and Verónica Torrano for helpful discussions and Fredrik Wallberg for technical advice.

AUTHOR CONTRIBUTIONS

Conceptualization, VSM, ECM; Methodology, VSM, ECM, PP, SM, MH, RAA, ALG, MRC, VG, JS; Investigation, ECM, OM, PP, SM, JM, AL, GC, MH, ALG, RAA, VG, JS, RS, SLG; Validation, ECM, OM, PP; Software, SM, MH; Writing – Original Draft, VSM, ECM; Writing – Review & Editing, VSM, ECM; Funding Acquisition, VSM, ECM; Resources, RAA, MRC, MH; Supervision, VSM.

COMPETING INTERESTS

Authors declare no competing interests.

MATERIALS AND CORRESPONDENCE

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Prof Victoria Sanz-Moreno ([email protected]).

All unique/stable reagents generated in this study are available from the Lead Contact with a completed Materials Transfer Agreement.

Sanz-Moreno, V. et al. Rac activation and inactivation control plasticity of tumor cell movement. Cell 135, 510–523, doi:10.1016/j.cell.2008.09.043 (2008).
Tozluoglu, M. et al. Matrix geometry determines optimal cancer cell migration strategy and modulates response to interventions. Nat Cell Biol 15, 751–762, doi:10.1038/ncb2775 (2013).
Herraiz, C. et al. Reactivation of p53 by a Cytoskeletal Sensor to Control the Balance Between DNA Damage and Tumor Dissemination. J Natl Cancer Inst 108, doi:10.1093/jnci/djv289 (2016).
Sanz-Moreno, V. et al. ROCK and JAK1 signaling cooperate to control actomyosin contractility in tumor cells and stroma. Cancer Cell 20, 229–245, doi:10.1016/j.ccr.2011.06.018 (2011).
Wolf, K. et al. Compensation mechanism in tumor cell migration: mesenchymal-amoeboid transition after blocking of pericellular proteolysis. J Cell Biol 160, 267–277, doi:10.1083/jcb.200209006 (2003).
Bergert, M. et al. Force transmission during adhesion-independent migration. Nat Cell Biol 17, 524–529, doi:10.1038/ncb3134 (2015).
Yamada, K. M. & Sixt, M. Mechanisms of 3D cell migration. Nat Rev Mol Cell Biol 20, 738–752, doi:10.1038/s41580-019-0172-9 (2019).
Pandya, P., Orgaz, J. L. & Sanz-Moreno, V. Modes of invasion during tumour dissemination. Mol Oncol 11, 5–27, doi:10.1002/1878-0261.12019 (2017).
Vicente-Manzanares, M., Ma, X., Adelstein, R. S. & Horwitz, A. R. Non-muscle myosin II takes centre stage in cell adhesion and migration. Nat Rev Mol Cell Biol 10, 778–790, doi:10.1038/nrm2786 (2009).
Ito, M., Nakano, T., Erdodi, F. & Hartshorne, D. J. Myosin phosphatase: structure, regulation and function. Mol Cell Biochem 259, 197–209, doi:10.1023/b:mcbi.0000021373.14288.00 (2004).
Olson, M. F. Applications for ROCK kinase inhibition. Current Opinion in Cell Biology 20, 242–248, doi:https://doi.org/10.1016/j.ceb.2008.01.002 (2008).
Chan, D. C. Mitochondrial Dynamics and Its Involvement in Disease. Annu Rev Pathol, doi:10.1146/annurev-pathmechdis-012419-032711 (2019).
Clark, E. A., Golub, T. R., Lander, E. S. & Hynes, R. O. Genomic analysis of metastasis reveals an essential role for RhoC. Nature 406, 532–535, doi:10.1038/35020106 (2000).
Georgouli, M. et al. Regional Activation of Myosin II in Cancer Cells Drives Tumor Progression via a Secretory Cross-Talk with the Immune Microenvironment. Cell 176, 757–774 e723, doi:10.1016/j.cell.2018.12.038 (2019).
Orgaz, J. L. et al. Myosin II Reactivation and Cytoskeletal Remodeling as a Hallmark and a Vulnerability in Melanoma Therapy Resistance. Cancer Cell 37, 85–103 e109, doi:10.1016/j.ccell.2019.12.003 (2020).
Legant, W. R. et al. Multidimensional traction force microscopy reveals out-of-plane rotational moments about focal adhesions. Proc Natl Acad Sci U S A 110, 881–886, doi:10.1073/pnas.1207997110 (2013).
Koch, T. M., Munster, S., Bonakdar, N., Butler, J. P. & Fabry, B. 3D Traction forces in cancer cell invasion. PLoS One 7, e33476, doi:10.1371/journal.pone.0033476 (2012).
Franck, C., Maskarinec, S. A., Tirrell, D. A. & Ravichandran, G. Three-dimensional traction force microscopy: a new tool for quantifying cell-matrix interactions. PLoS One 6, e17833, doi:10.1371/journal.pone.0017833 (2011).
Wolf, K. et al. Physical limits of cell migration: control by ECM space and nuclear deformation and tuning by proteolysis and traction force. J Cell Biol 201, 1069–1084, doi:10.1083/jcb.201210152 (2013).
Orgaz, J. L. et al. Diverse matrix metalloproteinase functions regulate cancer amoeboid migration. Nat Commun 5, 4255, doi:10.1038/ncomms5255 (2014).
Tantama, M., Martinez-Francois, J. R., Mongeon, R. & Yellen, G. Imaging energy status in live cells with a fluorescent biosensor of the intracellular ATP-to-ADP ratio. Nat Commun 4, 2550, doi:10.1038/ncomms3550 (2013).
Campbell, I. D. & Humphries, M. J. Integrin structure, activation, and interactions. Cold Spring Harb Perspect Biol 3, doi:10.1101/cshperspect.a004994 (2011).
Coelho, N. M., Wang, A. & McCulloch, C. A. Discoidin domain receptor 1 interactions with myosin motors contribute to collagen remodeling and tissue fibrosis. Biochim Biophys Acta Mol Cell Res 1866, 118510, doi:10.1016/j.bbamcr.2019.07.005 (2019).
Gowans, G. J., Hawley, S. A., Ross, F. A. & Hardie, D. G. AMP is a true physiological regulator of AMP-activated protein kinase by both allosteric activation and enhancing net phosphorylation. Cell Metab 18, 556–566, doi:10.1016/j.cmet.2013.08.019 (2013).
Banko, M. R. et al. Chemical genetic screen for AMPKalpha2 substrates uncovers a network of proteins involved in mitosis. Mol Cell 44, 878–892, doi:10.1016/j.molcel.2011.11.005 (2011).
Koga, Y. & Ikebe, M. A novel regulatory mechanism of myosin light chain phosphorylation via binding of 14-3-3 to myosin phosphatase. Mol Biol Cell 19, 1062–1071, doi:10.1091/mbc.e07-07-0668 (2008).
Toyama, E. Q. et al. Metabolism. AMP-activated protein kinase mediates mitochondrial fission in response to energy stress. Science 351, 275–281, doi:10.1126/science.aab4138 (2016).
Boulton, D. P. & Caino, M. C. Mitochondrial Fission and Fusion in Tumor Progression to Metastasis. Front Cell Dev Biol 10, 849962, doi:10.3389/fcell.2022.849962 (2022).
Sprenger, H. G. & Langer, T. The Good and the Bad of Mitochondrial Breakups. Trends Cell Biol 29, 888–900, doi:10.1016/j.tcb.2019.08.003 (2019).
Cantelli, G. et al. TGF-beta-Induced Transcription Sustains Amoeboid Melanoma Migration and Dissemination. Curr Biol 25, 2899–2914, doi:10.1016/j.cub.2015.09.054 (2015).
Rodriguez-Hernandez, I. et al. WNT11-FZD7-DAAM1 signalling supports tumour initiating abilities and melanoma amoeboid invasion. Nat Commun 11, 5315, doi:10.1038/s41467-020-18951-2 (2020).
Zanotelli, M. R. et al. Regulation of ATP utilization during metastatic cell migration by collagen architecture. Mol Biol Cell 29, 1–9, doi:10.1091/mbc.E17-01-0041 (2018).
Papalazarou, V. et al. The creatine–phosphagen system is mechanoresponsive in pancreatic adenocarcinoma and fuels invasion and metastasis. Nature metabolism 2, 62–80, doi:https://doi.org/10.1038/s42255-019-0159-z (2020).
Park, J. S. et al. Mechanical regulation of glycolysis via cytoskeleton architecture. Nature 578, 621–626, doi:10.1038/s41586-020-1998-1 (2020).
Zanotelli, M. R. et al. Highly motile cells are metabolically responsive to collagen density. Proc Natl Acad Sci U S A 119, e2114672119, doi:10.1073/pnas.2114672119 (2022).
Tharp, K. M. et al. Adhesion-mediated mechanosignaling forces mitohormesis. Cell Metab 33, 1322–1341 e1313, doi:10.1016/j.cmet.2021.04.017 (2021).
Romani, P. et al. Mitochondrial fission links ECM mechanotransduction to metabolic redox homeostasis and metastatic chemotherapy resistance. Nat Cell Biol 24, 168–180, doi:10.1038/s41556-022-00843-w (2022).
Chen, Y. & Dorn, G. W., 2nd. PINK1-phosphorylated mitofusin 2 is a Parkin receptor for culling damaged mitochondria. Science 340, 471–475, doi:10.1126/science.1231031 (2013).
Hu, Y. et al. The AMPK-MFN2 axis regulates MAM dynamics and autophagy induced by energy stresses. Autophagy 17, 1142–1156, doi:10.1080/15548627.2020.1749490 (2021).
Lehmann, S. et al. Hypoxia Induces a HIF-1-Dependent Transition from Collective-to-Amoeboid Dissemination in Epithelial Cancer Cells. Curr Biol 27, 392–400, doi:10.1016/j.cub.2016.11.057 (2017).
Kuo, C. W. et al. mtDNA as a Mediator for Expression of Hypoxia-Inducible Factor 1alpha and ROS in Hypoxic Neuroblastoma Cells. Int J Mol Sci 18, doi:10.3390/ijms18061220 (2017).
Te Boekhorst, V. et al. Calpain-2 regulates hypoxia/HIF-induced plasticity toward amoeboid cancer cell migration and metastasis. Curr Biol 32, 412–427 e418, doi:10.1016/j.cub.2021.11.040 (2022).
Roth, K. G., Mambetsariev, I., Kulkarni, P. & Salgia, R. The Mitochondrion as an Emerging Therapeutic Target in Cancer. Trends Mol Med 26, 119–134, doi:10.1016/j.molmed.2019.06.009 (2020).
Gaude, E. & Frezza, C. Tissue-specific and convergent metabolic transformation of cancer correlates with metastatic potential and patient survival. Nat Commun 7, 13041, doi:10.1038/ncomms13041 (2016).
Gaude, E. et al. NADH Shuttling Couples Cytosolic Reductive Carboxylation of Glutamine with Glycolysis in Cells with Mitochondrial Dysfunction. Mol Cell 69, 581–593 e587, doi:10.1016/j.molcel.2018.01.034 (2018).
Lorentzen, A., Bamber, J., Sadok, A., Elson-Schwab, I. & Marshall, C. J. An ezrin-rich, rigid uropod-like structure directs movement of amoeboid blebbing cells. J Cell Sci 124, 1256–1267, doi:10.1242/jcs.074849 (2011).
Guzman, A., Avard, R. C., Devanny, A. J., Kweon, O. S. & Kaufman, L. J. Membrane blebs play a critical role in a hybrid mode of cancer cell invasion in three-dimensional environments. J Cell Sci, doi:10.1242/jcs.236778 (2020).
Bernadskaya, Y. Y., Brahmbhatt, S., Gline, S. E., Wang, W. & Christiaen, L. Discoidin-domain receptor coordinates cell-matrix adhesion and collective polarity in migratory cardiopharyngeal progenitors. Nat Commun 10, 57, doi:10.1038/s41467-018-07976-3 (2019).
Takai, K. et al. Discoidin domain receptor 1 (DDR1) ablation promotes tissue fibrosis and hypoxia to induce aggressive basal-like breast cancers. Genes Dev 32, 244–257, doi:10.1101/gad.301366.117 (2018).
Labuschagne, C. F., Cheung, E. C., Blagih, J., Domart, M. C. & Vousden, K. H. Cell Clustering Promotes a Metabolic Switch that Supports Metastatic Colonization. Cell Metab 30, 720–734 e725, doi:10.1016/j.cmet.2019.07.014 (2019).
Moose, D. L. et al. Cancer Cells Resist Mechanical Destruction in Circulation via RhoA/Actomyosin-Dependent Mechano-Adaptation. Cell Rep 30, 3864–3874 e3866, doi:10.1016/j.celrep.2020.02.080 (2020).
Doyle, A. D., Carvajal, N., Jin, A., Matsumoto, K. & Yamada, K. M. Local 3D matrix microenvironment regulates cell migration through spatiotemporal dynamics of contractility-dependent adhesions. Nat Commun 6, 8720, doi:10.1038/ncomms9720 (2015).
Lanza di Scalea, F., Hong, S.S. & Cloud, G.L. Whole-field strain measurement in a pin-loaded plate by electronic speckle pattern interferometry and the finite element method. Experimental Mechanics 38, 55–60, doi:https://doi.org/10.1007/BF02321268 (1998).
Steinwachs, J. et al. Three-dimensional force microscopy of cells in biopolymer networks. Nat Methods 13, 171–176, doi:10.1038/nmeth.3685 (2016).
Schwarz, U. S. & Soine, J. R. Traction force microscopy on soft elastic substrates: A guide to recent computational advances. Biochim Biophys Acta 1853, 3095–3104, doi:10.1016/j.bbamcr.2015.05.028 (2015).
Chaudhry, A., Shi, R. & Luciani, D. S. A pipeline for multidimensional confocal analysis of mitochondrial morphology, function, and dynamics in pancreatic beta-cells. Am J Physiol Endocrinol Metab 318, E87-E101, doi:10.1152/ajpendo.00457.2019 (2020).
Simpson, A. J. & Brown, S. A. Purge NMR: effective and easy solvent suppression. J Magn Reson 175, 340–346, doi:10.1016/j.jmr.2005.05.008 (2005).
Dieterle, F., Ross, A., Schlotterbeck, G. & Senn, H. Probabilistic quotient normalization as robust method to account for dilution of complex biological mixtures. Application in 1H NMR metabonomics. Anal Chem 78, 4281–4290 (2006).
Worley, B. & Powers, R. Multivariate Analysis in Metabolomics. Curr Metabolomics 1, 92–107 (2013).
Berg, J., Hung, Y. P. & Yellen, G. A genetically encoded fluorescent reporter of ATP:ADP ratio. Nat Methods 6, 161–166, doi:10.1038/nmeth.1288 (2009).
Remark, R. et al. In-depth tissue profiling using multiplexed immunohistochemical consecutive staining on single slide. Sci Immunol 1, aaf6925, doi:10.1126/sciimmunol.aaf6925 (2016).

There is NO Competing Interest.

SupplementaryInformation190422.pdf
Supplementaryinfo
SupplementaryData1.Scriptfor3Dtractionstresscalculation.txt
SupplemenataryData1
Movie1CollagenImatrixdisplacement.mp4
Movie 1. Collagen I matrix displacement associated with elongated-mesenchymal movement.3D time-lapse video of HT1080 cell expressing LifeAct-GFP (F-actin, green) embedded within 3D Collagen I matrix. In grey, collagen I fibres imaged by collecting the reflected/backscatter light from the matrix using laser-scanning confocal microscopy.
Movie2Stressvectors.mp4
Movie 2. Stress vectors associated to matrix displacement from video 1.
Movie3Depthinformation.mp4
Movie 3. Collagen I matrix from video 1 with colour coded depth information.
Movie4Mesenchymalcellshowingelongatedmitochondria.mp4
Movie 4. Elongated-mesenchymal cell showing elongated mitochondria forming a network. A375P cell expressing LifeAct-GFP (F-actin, green) stained with Mitotracker Deep Red (red) and Hoechst (blue). First half of the video shows 3D reconstruction of elongated mitochondria (Mitotracker Deep Red) and nucleus (Hoechst). Second half of the video showing actin structures (LifeAct-GFP expression), in addition to mitochondria and nucleus.
Movie5Amoeboidcellshowingpunctateindividualmitochondria.mp4
Movie 5. Rounded-amoeboid cell showing motile punctate fragmented mitochondria. A375M2 cell expressing LifeAct-GFP (F-actin, green) stained with Mitotracker Deep Red (red) and Hoechst (blue). First half of the video shows 3D reconstruction of highly motile fragmented mitochondria (Mitotracker Deep Red) and nucleus (Hoechst). Second half of the video showing bleb actin structures (LifeAct-GFP expression), in addition to mitochondria and nucleus.

Download PDF

Journal Publication

published 22 May, 2023

Read the published version in Nature Communications →

Version 1

posted

You are reading this latest preprint version

AMPK is a mechano-metabolic sensor linking cell adhesion and mitochondrial dynamics to Myosin II dependent cell migration

Status:

Journal Publication

Version 1

Abstract

Figures

Introduction

Results

Discussion

Methods

Experimental model and subject details

Cell lines

Patient derived tissues

Animal studies

Animal derived tissues

Lung colonisation assay

Chemicals

Antibodies

Cell culture on 3D collagen I matrices

siRNA transfection

Immunoblotting

Collagen imaging

Cell morphology on collagen

Cell adhesion to collagen I

3D invasion

Immunofluorescence and confocal imaging

Mitochondrial mass and mitochondrial activity

Mitochondrial morphology

NMR metabolomics

Metabolic assays

ECAR and OCR measurement

Cellular ATP measurement

Luminescent ATP detection assay

Seahorse XF Real-Time ATP Rate Assay

Measurement of ATP:ADP ratio in 3D

Cell viability

Immunohistochemistry

Quantification and statistical analysis

Declarations

References

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1