1 Introduction

Homeostasis, in general, refers to the state of a dynamic equilibrium. The plasma membrane is proposed to be at a dynamic mechanical equilibrium in live cells which ensures that its tension is actively maintained at a particular level1. Membrane trafficking, osmotic pressure differences, and flattening/unflattening of membrane folds are some of the key processes contributing actively to the membrane tension regulation. While each of these cellular processes is affected majorly by the cytoskeleton, the actomyosin cortex also directly contributes to the generation of surface tension. Herein, we first briefly discuss the historical evolution of the understanding of the structure and dynamics of the membrane and cortex, followed by summarizing the definitions, techniques, key insights, and future challenges.

Several naturalists from the seventeenth to nineteenth centuries helped in the construction of the concept of cells2,3,4. However, it took ~ 100 years after the cell theory was proposed in 18395, for the concept of membranes in every cell to be proved, appreciated, and believed.

Early studies on water–lipid interfaces can be dated to Benjamin Franklin’s experiments with one teaspoon of oil on water/table in 17726. While Pockels in 1891 developed a way to measure the lipid monolayer covered surface at the air–water interface7, Langmuir in 1917 explained the concept of molecular hydrophily. He showed the flexibility of the hydrocarbon chains and talked about amphipathic molecules, even in protein monolayers8. Though, the first ‘membranologist’ Overton in 1899 showed that cell membranes were in lipid structure9; it took many years to calculate the thickness of cell membranes. Fricke was the first to report values of ~ 3.3–4 nm-thick membranes in erythrocyte10 and yeast cells11. However, a major breakthrough in the study of membrane structures was the report by Gorter and Grendel in 1925, who extracted lipids from blood cells and showed that the ratio of lipid covered surface to the total cell surface was 2:1, indicating the presence of membranes as lipid bilayers12. In 1935, Danielli and Davson proposed a Pauci-molecular concept of cell membranes where they suggested that the lipid bilayer was sandwiched between two monolayers of proteins, giving rise to a tri-laminar structure of biological membranes13. Though contradicted in the upcoming years, Robertson’s EM studies identified the constituents of the unit membrane14; one cannot disregard that the Danielli–Davson–Robertson model was the first to talk about asymmetry in the biological membranes15.

Thus, extensive studies in the 1900s15,16,17 led to the proposal and acceptance of first modern theory of the membrane—the Fluid Mosaic Model by Singer and Nicholson in 197218. While the “mosaic” was believed to be due to the varied composition of the membrane, comprising of phospholipids, proteins, and carbohydrates; the “fluid” aspect was added due to the mobility of these molecules. This model has remained fairly unchanged over the years, with certain modifications arising due to studies conducted in the late 1990s19. The presence of detergent-resistant microdomains in the cell membrane has helped the field of membrane biology evolve further.

With the advancement of microscopy, our knowledge about the structure and the composition of both the membrane and the cytoskeleton has grown immensely. Interestingly, a decade before the 1972 model of membranes came into being, the cell cortex was discovered as the first element of the cytoskeleton with the help of electron microscopy of free-living amoeba cells in 196420, and was identified as a dense layer of actin filaments. Several years later, studies on patients with Duchenne muscular dystrophy (DMD) showed that proteins could attach crosslinked actin networks to glycoproteins on the cell membrane in muscle cells21, indicating the presence of membrane–cortex linkages. The ERM (ezrin22–radixin23–moesin24) family of proteins which are hugely credited for being membrane–cytoskeleton linkers was discovered later between 1983 and 1991.

Discovering the dynamic nature of the membrane and the cytoskeleton has brought a paradigm shift in the field by revealing the constant intermingling of various components together maintaining homeostasis. While lipid diffusion was studied in the late 1950s25, protein diffusion on the membrane was discovered in 1970 when lateral diffusion of proteins was observed by immunofluorescent antibodies in heterokaryons26. In 1976, FRAP (fluorescence recovery after photobleaching) experiments clearly demonstrated the concept of membrane fluidity27. The dynamics in the cytoskeleton28 has been studied over many years (1970s–2010s) and spans several properties. These range from actin treadmilling29, branching30, severing31, capping31, and self-organizing into higher order actin structures32 in eukaryotic systems. However, experiments leading to the discovery of myosin-II , the first molecular motor to be found out, can be traced back to 186433. While the occurrence of endocytic pits and vesicles can be dated back to studies in 196434, the presence of membrane trafficking events was seen in 197935, who showed receptor-mediated endocytosis.

As studies probed the ubiquitous role of membrane and cortical mechanics in controlling membrane homeostasis, the ability of light to impart forces36 on objects started gaining importance. With this knowledge, techniques using light-optical trapping (OT)37, micropipette aspiration (MA)38,39, atomic force microscopy (AFM)40, and interference-based methods41 began to be used extensively to measure membrane tension in cells between 1990s and 2000s. These techniques are used till date and often give a major insight into believing the concept of membrane tension homeostasis42 and surface area regulation1 (in late 1990s). Fluctuation-based experiments which measure mechanics in erythrocyte membranes often probe the dynamics of the RBC membrane (RBC flickering43) was discovered back in 1951. Recent work in 2018 with a fluorescent molecular probe (FliptR) which quantifies changes in membrane tension in living cells has been identified44. In addition to direct observations and measurements in cells, reduced systems—supported bilayers, unilamellar vesicles, and vesicles with reconstituted cytoskeleton in them—have had a great impact on our understanding of the basic physical mechanisms involved in membrane homeostasis.

2 Membrane, Cytoskeleton, and Their Linkages

Ranging in thickness between ~ 5 and 8 nm due to the presence of proteins associated with the bilayers45, cell membranes outline the boundaries of the cells and are usually described by the fluid mosaic model46. These membranes are lipid bilayers where their hydrophobic regions in each monolayer see the interior of the bilayer and their hydrophilic head groups interact with the aqueous phases on each side. Depending on the type of interactions they have with the membrane, proteins can be classified into integral, peripheral, or lipid-anchored proteins47. The distribution of lipids in the membrane can differ in the two leaflets, with cell types and due to continuous flow of membranes from one compartment to another45, 48. Reports show that the cholesterol content in the plasma membrane is larger than ~ 20 weight% in most eukaryotes49. We know from studies that while the plasma membrane of myelin sheath in humans contains 30% phospholipids, 30% proteins, and 19% sterol; those in mouse hepatocytes contain 27% phospholipids, 45% proteins, and 25% sterol45. Several membrane proteins (like membrane pumps, membrane channels, curvature sensing proteins, etc.) are believed to be implicated in membrane homeostasis50,51,52,53,54,55,56,57,58. Lipids can also dynamically partition into nanometric platforms because of cholesterol altering the lipid packing around it59,60 and it is aided by short actin filaments61,62.

Though lipids in biological membranes usually exist in a liquid-ordered state below 37 °C, increase in temperature can lead to a liquid-disordered phase in bilayers63. Lipids have appreciable lateral mobility64 and can free diffuse laterally in a bilayer, about 107 times a second65. The lateral diffusion coefficient of phospholipids/cholesterol mixtures has been estimated to be around 1–10 µm2/s66,67,68. The lateral diffusion of membrane proteins can vary45 and are much slower than lipids27,69,70, both due to their size and due to their interactions with the cytoskeleton.

One of the major components of the cytoskeleton is the crosslinked tightly woven meshwork of contractile actomyosin, called cell cortex or cortical actin (Fig. 1), which is found underlying the plasma membrane. Actin treadmilling, the binding and turning-over of motor proteins and crosslinkers—sall together determines the structure of the actomyosin network71. Actin filaments of different lengths are created by the differences in rates of actin polymerization (Fig. 1b) and depolymerization at the two ends, the presence of optimum monomeric actin concentration along with ATP and other regulatory factors72. These filaments (~ 100 nm to microns in length in non-muscle cells73,74) interact with myosin proteins (Fig. 1c), which can themselves form mini-filaments, about ~ 300 nm in length75. The contractility in the network is brought about by the myosin motors walking on multiple actin filaments which induces compaction to the crosslinked (by proteins like α-actinin and filamin) network76. Depending on the method of measurement, degree of cell spreading, stage of cell cycle, and the type of cell, the average thickness of the cortex can vary ~ 50–400 nm74,77,78. The cortex is studied to be to thicker in low spread-out cells77, and the thickness is more in de-adhered cells than in mitotic cells77,78,79. The mesh size of the cortex can also vary from ~ 20 nm in mitotic HeLa cells to ~ 300 nm in some interphase cells71.

Figure 1:
figure 1

Schematic representation of the organization of submembranous actin network under the membrane. Membrane thickness (~ 7 nm), typical cortex thickness (~ 100–1000 nm), and expected gap between cortex and membrane (~ 10–20 nm) are indicated. Box 1 shows how filaments are arranged parallel to the plane of the membrane at the cortex with size of small filaments expected ~ 100–1000 nm. Box 2 shows how a myosin-II minifilament (~ 100–400 nm long) leads to compaction (initial black dashed to green dashed) of the actin network by binding simultaneous to two different actin filaments using its two ends and walking towards barbed ends of both the filaments. Box 3 shows a zoomed-in schematic of extension of the membrane brought in by a new monomer (shown in light green, ~ 2.75 nm) being added on to an existing polymer of actin—a special case where the filament is normal to the plane of the membrane, like at the lamellipodia. Box 4 highlights a typical ERM family protein (Ezrin, here) adding on to the membrane in its phosphorylated form (and diffusing slowly). It quickly transits to an actin-bound state which makes it almost immobile. Box 5 highlights how the cortex may restructure during endocytosis (process that is believed to enhance membrane tension) and allows exocytosis (process that is believed to reduce membrane tension) due to its mesh like structure with typical sizes of vesicles being ~50–100 nm.

Full size image

All actin-binding proteins undergo cycles of attachment and detachment from the meshwork at varying rates and adds to the dynamic nature of the cytoskeleton. With the help of several accessory protein, the actomyosin network can form a variety of structures, showing varied filamentous actin organization, in the cell80. The cortex is attached to the membrane with the help of certain cytoplasmic proteins. The ERM (ezrin, radixin, moesin) class of proteins81,82 are known to attach to the PIP2 domains of the membrane on one hand and to actin filaments on the other hand (Fig. 1d). These proteins contain three conserved domains, an N-terminus FERM (Band 4.1, ezrin, radixin, moesin) domain, a central helical coiled-coil domain, and a CTD (C-terminal domain)83. Activation by PIP2 mediated uncoupling of the N- and C-terminal domains allows the inactive cytosolic ERM proteins to bind with the plasma membrane either directly or indirectly through membrane-associated proteins84,85. The myosin-1 family of proteins also mediate this membrane–cortex linkages by binding to actin filaments by their calmodulin-dependent motor domains and to acidic phospholipids on the membrane86. Studies report that based on the angle of polymerization, the membrane can extend by ~ 2.75 nm (Fig. 1b) with each polymerization event87.

In the next section, we define membrane tension and discuss the different techniques used to measure it, to understand the role of membrane and its underlying cytoskeleton on membrane tension homeostasis.

3 Membrane Tension and Its Measurement

Over the years, it has been seen that several biological processes have been reported to be affected by membrane tension, namely membrane trafficking88 (Fig. 1e), phagocytosis89, cell motility50,90, cell division, and cellular morphological changes91, in cells. Even in vitro reports suggest that changes in membrane tension can affect clathrin binding92 and polymerization of coat proteins93. But what is membrane tension?

The pure lipid bilayer membrane at the molecular scale is a 2D fluid. It is believed to not experience mechanical stress and be ‘tensionless’94. However, as they form closed structures like vesicles (or cells), they start experiencing tension—arising, for instance, due to the osmotic pressure differences between the outside and inside of the vesicle95. Increasing the pressure difference results in membrane deformations and increases tension. Pascal’s law measurement of deformations allow micropipette aspiration-based measurement of tension (Fig. 2a) in lipid vesicles39. Admittedly, cells are more complex structures due to the presence of the cytoskeleton, and measurements of the ‘net’ tension by micropipette aspiration have contributions both from the cytoskeleton and the membrane96. Such measurements are usually complemented with conditions under which the cytoskeleton is disintegrated and can no longer contribute to the net tension97. Typical net tension in the presence of the cortex in fibroblast cells yields values ~ 0.4 pN/nm, while it drops to 0.04 pN/nm in the absence of a contractile cortex97. Similarly, techniques utilizing nano-tether extraction using optical traps (Fig. 2b) or AFM (Fig. 2b) are based on Helfrich’s elastic model of the membrane. They measure the force required to maintain a nano-tether (Fig. 2d) extracted out of vesicles or cells expected to depend on the membrane tension and bending rigidity. When performed on cells, due to the additional cost involved of breaking membrane–cytoskeleton attachments, tension derived from the force is termed as the “apparent membrane tension”91,96,98.

Figure 2:
figure 2

Schematic representation of the invasive techniques used to measure membrane tension. a Diagram of micropipette aspiration where the membrane of a suspended cell is aspirated into the micropipette where pressure (P1) is applied, the pressure in the sample being, say P2. The pressure difference created (ΔP=P2–P1), and the varying radii of the pipette and the object are used to calculate net tension. It is the summation of membrane tension and cortical tension. Membrane tension can only be calculated using this technique in the presence of a cytoskeletal drug which removes the underlying cytoskeleton and nullifies cortical tension. b Diagram of optical trapping where a latex microsphere (red) is used by a laser to manipulate the membrane of cells and extract tethers. The force calibrated from the trap stiffness (ktrap) and bead displacement (Δx) is used over time to measure the force at the plateau phase of tether pulling. This is used to calculate the apparent membrane tension, using a known value of bending modulus. The apparent tension is a summation of membrane tension and the contribution from membrane–cytoskeleton attachments. c Diagram of atomic force microscopy where a cantilever is used to extract tethers from cells. The force in this technique is calibrated by the cantilever stiffness and its displacement (Δx), using the reflected laser’s displacement (ΔX). Apparent tension is calculated as mentioned above. d A zoomed-in view of a tether being pulled by a bead in optical traps or by a cantilever in AFM where the membrane linkages are broken and membrane, devoid of cytoskeleton, is pulled into the tube.

Full size image

In both micropipette aspiration and nano-tether extraction, the tension can be thought as the lateral mechanical tension. To understand it, one can imagine a hypothetical boundary around a local patch of membrane99 and define tension as the “force applied to the unit length of this imaginary boundary by the surrounding membrane in the direction tangential to the membrane plane”. While tension extracted from MA experiments (based on Pascal’s law) is consistent with this description, Helfrich’s model includes tension as the energy cost per unit area of increasing local surface area. Note that Helfrich’s model100 has been extended to predict membrane height fluctuations and forms the basis of measurements utilizing fluctuation spectra to estimate tension101,102,103,104 (Fig. 3a–c). Intuitively, enhanced tension should suppress transverse fluctuations105. It can be distinguished from other factors that also contribute to fluctuations like bending rigidity, cytoplasmic viscosity, and temperature due to the different frequency dependence of their effect on the amplitude of fluctuations106. In addition to lateral tension and fluctuations tension, the recent breakthrough of fluorescence-based measurement (Fig. 3d) of tension reflects the internal tension44. The fluorescent reporters bind to the hydrophobic region of bilayers and undergo changes in their fluorescence lifetime depending on the immediate lipid packing44. With such different possible routes of measuring tension, it is important to understand if these descriptions of tension are comparable with each other.

Figure 3:
figure 3

source suffers reflection due to a difference in refractive indices. A camera is used for capturing interference images over time which helps to measure fluctuations and tension. IRM can create maps of tension and requires measurement over 40–100 s. d A cartoon of a migrating cell with front and rear end showing a gradient of membrane tension due to incorporation of FliptR, a molecular probe. The changes can be captured by fluorescence lifetime imaging (FLIM). The probe can detect tension changes because of the fluorescence lifetimes alterations created due to tension planarizing the fluorescent groups. Detecting tension differences by FliptR is a fast process which can used be used all over the cell membrane.

Schematic representation of some noninvasive techniques used to measure membrane tension. a Flowchart to show that membrane fluctuations (used in techniques discussed in b, c) can be used to calculate membrane tension using power spectral density as suggested in the equation \(PSD\left(f\right)=\frac{4{\eta }_{eff}A{k}_{B}T}{\pi }{\int }_{{q}_{\mathrm{min}}}^{{q}_{\mathrm{max}}}\frac{\mathrm{d}q}{{(4{\eta}_{eff}(2\pi f))}^{2}+{\left[\kappa {q}^{3}+\sigma q+\frac{\gamma }{q}\right]}^{2}}\). b Schematic representation of an experimental design that uses a laser and a QPD to detect fluctuations at cell edges. The measurement is fast and has been reported on cellular blebs. c Schematic of interference reflection microscopy (IRM) where the basal membrane of adherent cells can be imaged due to interference of reflected rays. Light from a lamp

Full size image

Studies address this by measuring the internal tension (dependent on lipid–lipid distance, thus related to real microscopic area occupied by a fixed number of lipids), lateral frame tension (manifested by changes in projected area for a fixed number of lipids), and fluctuation tension (derived from the fluctuation spectra) of the same simulated sheet of a fluctuating membrane107. The study predicts that internal tension deviates from lateral frame tension and fluctuation tension for very low tensions at which fluctuations are significantly high or the real microscopic area is >  ~ 3% of the projected area. The lateral frame and fluctuation tension, however, match for a large (5 decades) window of tension95,107,108.

Unlike lipid vesicles, cells have additional factors influencing measurements—which need to be properly accounted for the correct interpretation of measured tension. The cytoskeleton, per se, affects the membrane in many ways: it confines the membrane, the actin gel, and/or pinning of transmembrane proteins with the membrane increases the effective viscosity—and it can also impart direct forces109. Fluctuation-based techniques directly follow the movement of the membrane—and can hence be utilized to study the different effects of the cytoskeleton on membrane mechanics102,103,109. While direct measurement of membrane tension from tether-based measurement is difficult in cells, fluorescent reporters still are expected to reflect the internal tension. Tensional inhomogeneities are best measured by fluorescent reporters110 or fluctuation-based techniques103, although a recent demonstration of slow flow of membrane tension utilized two-point measurement of tension using optical trap-based tether extraction111.

The next sections compare evidence accumulated from different techniques to summarize how different aspects of the actin cytoskeleton affect membrane tension regulation.

4 Effect of Actin Polymerization

Actin polymerization can impact membrane tension depending on the network organization. At the lamellipodia of keratocytes; for example (Fig. 4a), actin polymerization occurs almost normally to the membrane and thus maximizes the displacement of the membrane87. OT as well as FliptR measurements show tension to be enhanced at the leading edge in motile cells110,112, showing the direct contribution of actin polymerization forces on membrane tension. In non-motile cells, most of the membrane is free of lamellipodia like protrusions and have an intact layered cortex80,113. In the presence of this intact cortex, actin polymerization occurs almost tangential to the membrane and does not contribute majorly to membrane displacement. Proper polymerization rates, however, ensure longer filaments, resulting in a stronger, well-crosslinked, intact cortex.

Figure 4:
figure 4

Schematic representation to show how the cortex can affect membrane tension. a Effect of actin polymerization forces in motile cells: actin polymerization of the cortex at the lamellipodia in motile cells, like keratocytes, occurs normal to the membrane, thereby extending and ‘taut’ ing the membrane. This directly contributes to the enhanced membrane tension at the leading edges of migrating cells. b Effect of actin polymerization forces in non-motile cells: cytochalasin D treatment not only stops actin polymerization, but it also ruptures and clears the cortex due to myosin-II’s contractile forces. Cartoon depicts a case where cell edge has been seen to straighten (and cortex thickness decreased) till the cortex ruptures. After rupturing the cortex is not sharp at the rupture point (dotted line) and thickness increases. c Effect of contractility: depiction of how myosin-II inhibition may cause an increase in apparent membrane tension, by preventing local curvature-based contractions, which clearly demonstrates the role of contractility to reduce tension during actin flow. However, inhibiting myosin-II can decrease hydrostatic pressure and reduce the apparent membrane tension. d Effect of membrane–cortex attachments: Cartoon depicts how reduced linkages may lead to a decrease in tether force, and increased attachments enhance apparent membrane tension.

Full size image

The timescales of polymerization/depolymerization events, however, are much smaller (these are faster events) than the timescales over which the effect builds up and manifests as functional states. Reports using speckle microscopy have shown that the polymerization velocities in cells can vary from 7 nm/s in fibroblasts114 to 170 nm/s in keratocytes115, which amounts to actin monomer addition rate (perpendicular to the membrane) to be ~ 3–63 /s, since the displacement caused by each monomer addition87 is ~ 2.75 nm. The rate of actin turnover, measured by FRAP (fluorescence recovery after photobleaching) studies, ranges from ~ 20 s in keratocytes116 to ~ 500 s in certain endothelial cells117. Reports suggest that cofilin enhances actin depolymerization at the pointed end by a factor of ~ 30118. Since turnover rates are a combination of uncapping rates and depolymerization rates, studies have calculated depolymerization rate to be ~ 3 s for 100 subunits long filament119. At the site of polymerization, the force is instantaneously felt. However, the time for the force to alter tension of the whole front (of a motile cell, for example) is currently debated. While instantaneous equilibration has been long believed120, recent studies report much slower flow of tension111 (with an effective diffusion constant ~ 0.024 μm2/s). Faster measurements of tension in live cells are therefore required to understand how tension is affected by actin dynamics.

To understand its importance in maintaining membrane homeostasis, we next discuss experiments that measure the new membrane tension inhibiting actin polymerization in non-motile cells by drugs like Cytochalasin D (Cyto D) or Latrunculin A/B (Lat A/B). It needs to be noted that often such treatments not only prevent polymerization of actin but also result in a myosin-II-based clearing of the cortex103,121, leaving the membrane bare in certain regions while accumulating actin in others. Cyto D has been shown to cause a straightening of the cell periphery (cortex + membrane) together with thinning of cortex77 at the central section till it ruptures (Fig. 4b). The heterogenous nature of cortex is amplified on drug treatment, such that some regions are weakened more and pulled by myosin-II-based contractile forces. The progress of this process may be different in different cell types.

The apparent tension or net tension has been reported to decrease in drug-treated cells89,97,120. Results of net tension are expected, since the dissolution of the cortex results in a significant reduction of the massive contribution of membrane–cytoskeleton attachments which tether extraction forces sense. However, whether membrane tension also changes is unclear from these studies.

Fluctuation-based force spectroscopy experiments are extremely sensitive to membrane movements. However, even in these studies, conflicting results can be found probably due to the use of three different systems—blebbing M2 cells102, neurites104, and HeLa cells103, which have different effects of these drugs. Although the most consistent reports are those of reduction of tension on lowering polymerization rates, it is important to note that the report of enhanced tension (by Cyto D103) reports fluctuation tension, and pools data from local patches—and is therefore distinct from the other studies. A corresponding increase in amplitude of fluctuations (on Cyto D) found in the same study. Although the rise in mean tension accompanied by rise in mean fluctuations is non-intuitive, the matter is clear when instead of comparing two different systems, one compares regions in Cyto D-treated cells. Once the parameters of same regions are compared, the dependence of fluctuations on tension remains inverse. The study further corroborates this by experiments that use a pre-treatment of Blebbistatin (Blebb.) to stop cortex clearance. On comparisons between Blebb. and Blebb. + Cyto D-treated cells, membrane tension is seen to significantly increase in the latter case. We believe that this is because the pressure load, although reduced by cortex dissolution, is now completely balanced by membrane tension—hence resulting in its enhancement. We hope that future studies implementing other non-invasive and local measurements like using FlipTr will resolve the mechanism.‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬

MA-based measurements report membrane tension only when the cortex is disintegrated, and hence, do not contribute to the debate about the effect of the cytoskeleton’s polymerization. Interestingly, ATP depletion studies also show similar discrepancies between OT-based measurements and fluctuation-based tension measurements. Note that ATP depletion is expected to drastically reduce actin polymerization as well as contractility of the cortex and does not clear the cortex103. While OT-based tension measurements show a huge drop in apparent tension122, IRM on HeLa103, and similar fluctuation-based study in RBCs101,123 show an increase in tension. These studies emphasize that at the homeostatic state, the plasma membrane is actively maintained at a lowered tension state and increases when specific processes such as mitosis occurs124. To thoroughly understand the role of the cytoskeleton, mapping of tension must be further explored. Also, a combination of OT-based methods with fluctuation-based methods is needed to address the differences observed, for example, when ATP is depleted.

It is important to note that even controlled experiments with video microscopy of GUVs (giant unilamellar vesicles) have shown tension to be reduced on activating membrane pumps125 and that micropipette studies on fixing tension have long reported ATP to enhance fluctuations and excess area126. OT-based tension measurements in such systems would resolve the origin of the conflict, because these systems lack membrane–cytoskeleton linkages which hugely offset tether extraction-based tension measurements. Although fluctuation-based studies report an increased tension (on ATP depletion) even after considering the enhanced fluctuations due to activity (factor A, Fig. 3a), whether the enhanced tension in cells is a misinterpretation due to the lack of incorporations of the actual effect of activity (like allowing for frequency dependence of A) should be definitely addressed.

5 Effect of Myosin-II Based Contractile Forces

The contractile nature of the actomyosin network acting on the closed geometry of a cell results in creating hydrostatic pressure that is balanced together by cortical tension and membrane tension99. This pressure leads to the formation of natural blebs in cells when the membrane has appreciably detached from the cortex like in M2 cells102,127, newt blastomeres128, walker carcinoma cells129, Dictyostelium130, and zebrafish germ cells130. Fluctuation-based measurement in M2 cells have revealed that growing blebs have low tension initially, which increases during expansion127. Authors highlight that the effect is probably connected to the newly formed cortex. However, the experiment shows how contractility of the rest of the cell enhances tension of a patch of membrane that has lost its attachment from the cortex. Here, we point out that these high-speed measurements were possible due to the use of fast measurement of fluctuations—either QPD-based102 or imaging-based131.

Studies with RBCs are important to be noted here, where linkages of the membrane with the spectrin network have been reported to provide inward forces132 due to the biconcave morphology and “soften” the membrane101. However, experiments on adherent nucleated cells reveal at first glance conflicting results. In line with results from RBCs, are measurements on migrating keratocytes. Myosin-II’s ability to contract (and create inward forces on an imaginary frame) when inhibited by the drug Blebbistatin increases the tension at rear ends of these cells120—supporting the role of myosin-II in lowering membrane tension (Fig. 4c). However, similar OT-based tether extraction from other non-motile cells have revealed both an increase50 as well as decrease133 in tether forces (proportional to tension) on myosin-II inhibition. The decrease may be explained by the fact that on inhibiting myosin-II decreases the hydrostatic pressure134,135 (without compromising cortex contractility completely) and results in a decrease in both membrane tension as well as cortical tension (Fig. 4c). The increase, on the other hand, may be expected in systems where curvature results in contractility causing local inward forces reducing tension. Inhibiting myosin-II in these cases would increase tension. We think that in cases where the hydrostatic pressure reduction is still shared by the still partly contractile cortex, myosin-II inhibition by Blebbistatin reduces tension. However, when the cortex and its load-sharing capacity is compromised, an increased membrane tension is observed.

Therefore, although conflicting, the kind of cell, the initial state of the cell (state of adhesion and hydrostatic pressure), and the level of inhibition together are expected decide how myosin-II inhibition affects tension.

The cortex can affect the membrane only, because it lies under the membrane. It stays there because of membrane–cortex attachments. We next discuss their role in membrane homeostasis.

6 Effect of Membrane–Cortex Linkages

Membrane–cortex attachments have been believed to be important for membrane tension, especially since they have been suggested to provide resistance to membrane flow into tubes during tether extractions90. In zebrafish progenitor cells, it has been shown that reduction of attachments by Ezrin or myosin 1b depletion results in reduced static tether force136. Studies have also reported that increasing membrane–cortex attachment by ezrin phosphorylation in lymphoblasts137, adding PIP2 in epithelial cells138, or by incorporating synthetic linkers in stem cells139 lead to an increase in the tether force. Recent studies on HEK cells suggest that the reduced tether forces in epithelial cells in comparison to embryonic cells are due to naturally high expression of membrane–cortex attachments in embryonic cells140.

The effects of perturbing the Myosin 1 family of proteins on membrane tension have also been explored. In line with ezrin studies, reducing attachments by myosin 1g in lymphocytes also reduces tether force141. Overexpression of each Myo1a, Myo1b, Myo1c, Myo1d, or Myo1e individually increases the tether force in NIH 3t3 fibroblast cells; however, overexpression of dominant-negative construct Myo1c decreases it142. Enhanced or reduced membrane-cytoskeleton attachments are expected to alter the tether force (and apparent membrane tension) as reported, since tether extraction in cells includes the cost of detaching these membrane–cortex linkages (Fig. 4d).

It remains to be understood if membrane tension is altered in a similar way. Fluctuation-based experiments/fluorescent tension probes may prove to be useful in this context.

7 Challenges and Future Prospects

Besides global changes, cellular processes may locally alter membrane tension—as expected during endo/exocytosis. Mechanochemical feedbacks are reported to be triggered by global change in tension143. Are local perturbations to tension sensed and corrected locally? Tension controls many processes, ranging from cell–cell fusion, membrane–membrane fusion, motility, membrane trafficking, etc. Future studies need to evaluate the correlation of the local tension profile with spatial heterogeneities of actin polymerization, contractility, and membrane pinning, in addition to curvature sensing proteins. This will provide a platform to explore the role of tension in other phenomena like sorting of endocytic pathways, membrane fusion, etc. We believe that interference and fluorescence-based techniques are best suited toward these goals by providing the required spatial resolution in tension measurement and by allowing specific targeting of fluorescent tension probes to different intracellular organelles144.

The challenges faced by fluctuation-based study is mainly to properly account for active forces. Although in certain systems (HeLa, CHO, and C2C12), a window of frequency has been found (0.01–1 Hz) in which activity alters fluctuations similarly across the band103, it still needs to be characterized before interpreting fluctuations using standard Helfrich-based model to extract tension. Recent studies in RBC, for example, show frequency dependence at the above-mentioned band but a frequency-independent effect at frequencies >10 Hz132. The challenge of utilizing fluorescent probes of tension, however, is to account properly for the altered levels of lipid “order”, and not tension in changing the fluorescence lifetime of the probes.

Despite the challenges, a holistic understanding of the role played by the cytoskeleton in membrane homeostasis will require such new directions of investigations—covering regulation of tension at a local level, in intracellular structures as well as in cells embedded in deep tissues.