Forces and Flows at Cell Surfaces

Graphical Abstract

Introduction by Damien Thévenin, Associate Editor

I am delighted and honored to introduce the following essay by Dr. Aurelia Honerkamp Smith, an Assistant Professor in the Department of Physics at Lehigh University, in this issue’s ‘Up-and-Coming Scientist’ section. Dr. Aurelia Honerkamp-Smith earned a B.S. in Chemistry from the University of Oregon in Eugene, OR, and a Ph.D. in Physical Chemistry from the University of Washington in Seattle, WA, where she used fluorescence microscopy to study lipid phase transitions with Prof. Sarah L. Keller. As a postdoctoral researcher in Prof. Raymond E. Goldstein’s laboratory in the Dept. of Applied Mathematics and Theoretical Physics at the University of Cambridge, she directed her interests in microfluidics, micromanipulation, and light sheet microscopy toward making mechanical measurements on lipid membranes and cell sheets. In this essay, Dr. Honerkamp-Smith discusses recent developments in membrane hydro-dynamics, protein transport, and flow mechanosensing by living organisms to make the case that lipid mechanics can contribute to flow sensing. While it is well-established that flow affects cellular functions, the molecular mechanisms that cells use to sense it have remained surprisingly opaque. The first steps in flow detection are likely to occur at the plasma membrane, an organized, two-dimensional molecular array with solid and fluid properties. The mobility of membrane proteins and lipids is constrained by complex interactions with the cytoskeletal protein network that supports the membrane. Physiologically relevant flows can only generate tiny forces on individual proteins, smaller than those from thermal noise. However, Dr. Honerkamp Smith proposes that cells could overcome this problem by sensing micron-scale concentration gradients of membrane proteins sorted laterally by flow over the extracellular surface of the plasma membrane. In this context, Dr. Honerkamp-Smith also discusses the implications of her work on lipid responses to flow and how proximity to rigid surfaces modifies those responses (see for example Anthony et al. 2022, Ratajczak et al. 2023, and Miller et al. 2023, referenced in the current essay).

Flow is an Important Sensory Input for Cells

Fluid flow is a ubiquitous feature of the surroundings of living cells. Flow can result from circulation of fluids inside an organism, from movement of tissue such as muscle contraction, or it can be the result of motility of the cell itself. All of these flows reveal information about a cell’s environment. Cellular responses to shear flow regulate diverse physiological processes such as maintenance of bone density, inflammation, and angiogenesis during development and cancer (Wittkowske et al. 2016; Tarbell et al. 2014; Wragg et al. 2014).

Individual cell responses to flow have been observed in a variety of contexts. For example, marine dinoflagellates bioluminescence in response to fluid flow, when shear stress is strong enough to deform their cell walls (Jalaal et al. 2020). Mammalian fibroblasts show increased proliferation, migration, and alignment under flow (Shi and Tarbell 2011; Dan et al. 2010), and the ciliated protist Spirostomum ambiguum contracts in response to external flows that increase its membrane tension (Mathijssen et al. 2019).

In some cases, individual flow responses coordinate to modify tissue organization. Living bone is constantly remodeled by osteoblasts, which secrete extracellular matrix which is subsequently mineralized and osteocytes, which resorb bone matrix and regulate the activity of osteoblasts. Both cell types respond to the flows of interstitial fluid that occur when a bone is repeatedly deformed under mechanical load. Experiments reveal an array of specific responses to low and high shear, as well as to oscillating or steady flow, which coordinate to direct efficient allocation of bone density (Wittkowske et al. 2016).

In humans, blood flow generates pulsatile shear stress of 1–4 Pa at arterial walls (Tanaka et al. 2021). Endothelial cells form a monolayer on the inner surface of blood vessels and their responses to flow regulate blood pressure, control immune system activation, and direct the organization and remodeling of the vascular system itself. Multiple cell structures and proteins interact to modulate their extensive palette of physiological responses to flow: endothelial cells can distinguish between different flow speeds, directions, and pulsation frequencies (Aitken et al. 2023; Wang et al. 2013; Ostrowski et al. 2014).

Flow mechanosensing, particularly in the context of endothelial cell sensing and regulation blood flow, has been studied extensively. Possible flow sensors identified in endothelial cells include primary cilia, mechanosensitive ion channels, several types of protein complex located at intercellular junctions, and components of the glycocalyx (Fig. 1) (Tarbell et al. 2014). Among these candidates, the most direct evidence points to primary cilia; however, these are only found on cells in locations which experience very low shear stress (Aitken et al. 2023); larger shear causes them to be resorbed. Therefore, open questions about how endothelial cells detect flow magnitude, direction, and pulsatility in high-shear regions remain.

Fig. 1.

A cross-section through an artery shows a single layer of endothelial cells (orange) lining the inner walls. Blood flow applies a pulsatile shear stress of approximately 2–4 Pa. Multiple avenues for flow sensing have been identified in endothelial cells; flow can apply forces to the glycocalyx or surface coat of proteoglycans (top right); flow may apply force to activate proteins in cell junction protein complexes, such as integrins (bottom right); and flow forces can stretch the plasma membrane in order to open mechanosensitive ion channels (bottom left)

The identity of the primary flow sensor in endothelial cells remains unclear (Tanaka et al. 2021), but it is likely that the first steps in flow sensing take place at the outer or “plasma” membrane. The biophysical properties of lipid membranes determine the reactivity and mobility of membrane proteins. Thus, understanding lipid membrane bio-physics has the potential to yield useful insights into the mechanisms behind biological flow sensing. Due to their molecular structure, biological membranes have counterintuitive mechanical properties that resemble both fluids and solids and respond to flow in surprising ways.

Cell Membranes are Fluid

Lipid bilayers resemble a nanometer-thick sandwich in which the hydrophobic tails of the lipid molecules are separated from the surrounding water by the double layer of polar head groups. At low temperatures, membranes can enter a soft solid state resembling a sheet of wax two molecules thick. Warming a solid membrane melts it: area per lipid and diffusion constants increase, and the membrane becomes a quasi-two-dimensional, nearly incompressible fluid. In fluid membranes, lipid molecules have freedom to move laterally in the plane of the membrane, but their movement perpendicular to the membrane plane is tightly constrained. The membrane can be described as a two-dimensional fluid because lipids flow and circulate in response to parallel flows in the surrounding water, even while the membrane sheet itself remains stationary (Honerkamp-Smith et al. 2013; Sakuma et al. 2020). A lipid membrane interrupts fluid flow in the direction perpendicular to it, although individual water molecules can diffuse through.

There is abundant evidence that the membranes of living cells are in a fluid state. Experiments such as fluorescence recovery after photobleaching (FRAP) indicate fast diffusion of lipids and proteins: typical molecular diffusion constants in living cells are 1–5 μm²/s, so that a single molecule can explore an entire cell within tens of seconds (Koppel et al. 1976; Wawrezinieck et al. 2005). This freedom means that highly mobile membrane protein complexes can quickly form or dissipate in response to external or internal stimuli. Experiments in which a “tether” or nanometer-sized membrane tube is pulled from cells (Hochmuth et al. 1982, 1996) demonstrate that lipids can flow. The ability to flow allows lipid membranes to self-heal after rupturing (Chabanon et al. 2017), to bend when forming vesicles or caveolae, and to flex in order to cover cell protrusions, like filopodia or lamellipodia.

Lipid Composition Complexity in Cell Membranes

Not all fluids are alike: imagine the difference when pouring honey versus acetone out of a bottle. Cell membranes contain a complex mixture of hundreds of different lipid molecules. By adjusting the relative concentrations of lipids with different structures, a cell can change the mechanical features of its membrane: its viscosity, thickness, area per lipid, and bending stiffness. These properties impact protein activation both directly (in the case of proteins that bind individual lipids as ligands) and indirectly via nonspecific mechanical interactions with proteins (Levental and Lyman 2023). Control of membrane lipid composition is crucial for cell function, demonstrated by the fact that cells modify their lipid content over both long and short timescales. Bacteria and zebrafish cells alter their lipid makeup when grown at different temperatures (Sinensky 1974; Burns et al. 2017), and the lipids found in comb jelly membranes are correlated with the temperature and pressure of the organisms’ living environment (Winnikoff et al. 2021). Mammals adjust their membrane lipid and cholesterol content within hours of ingesting unsaturated fatty acids (Levental et al. 2020). One goal of lipid regulation may be to maintain membrane viscosity at an optimal value, since viscosity controls the rates of reactions that involve membrane proteins, such as electron transfer, photosynthesis, and infection resistance (Budin et al. 2018; Ford and Barber 1983; Heaton and Randall 2011). Membrane viscosity and thickness set the scale of physical forces acting on membrane proteins, so understanding how they depend on lipid content should help illuminate how membrane lipid composition regulates membrane protein function.

Model Membrane Experiments Reveal Fundamental Membrane Properties

The material properties of living membranes are challenging to determine directly. The enormous number of different lipid constituents makes it difficult to assign a specific function to each one. In addition, the two leaflets of living membranes are highly asymmetric in both lipid composition and lipid packing (Lorent et al. 2020; Doktorova et al. 2020). Fortunately, lipid molecules self-assemble in water to form stable, spherical membrane sheets called vesicles (ranging from “small” vesicles 10–100 nm in diameter, to “giant” vesicles which can be 200 μm in diameter) (Dimova and Marques 2018). Isolated lipid mixtures containing only a few components and surrounded only by water can serve as a minimally complex “model” membrane system. In studying model membranes, some detail and physiological activity is sacrificed in exchange for increased control and tractability.

Basic physical properties such as viscosity and bending stiffness which are well understood in macroscopic materials are surprisingly complex in membranes because of their molecular dimensions and unique structure. This makes it challenging to complete quantitative measurements. For example, current research finds values for membrane viscosity that span several orders of magnitude (Fitzgerald et al. 2023).

Diffusion in Membranes

Even a single-component membrane cannot be understood in isolation, movement in the membrane is coupled to movement of the surrounding fluid. These interactions make diffusion in fluid membranes more complex than diffusion through a three-dimensional liquid. They are captured in the theoretical framework developed by Saffman and Delbruck (Saffman and Delbruck 1975). Their theory considers the entire membrane as a two-dimensional fluid with a membrane viscosity η_m, which is larger than the water viscosity η_w, and considers diffusion of a cylindrical inclusion which spans the membrane; for instance, a small transmembrane protein. The ratio of the two viscosities $\frac{{\eta }_m{\eta }_w}{}$ has units of length and is often called the hydrodynamic length l_H. For objects with radius r smaller than the hydrodynamic length, diffusion depends more weakly on radius r than in the usual Stokes–Einstein diffusion for three-dimensional fluids $(D\propto ln(\frac{1}{r}\text{rather than to}\frac{1}{r})$. If the object is large or membrane viscosity is very small, flows in the surrounding fluid dominate over flows inside the membrane and an object’s motion is instead described using the theory developed by Hughes, Pailthorpe, and White (Hughes et al. 1981). Typical hydrodynamic lengths for lipid membranes are in the range of 1–10 micrometers, much larger than single proteins (Stanich et al. 2013; Honerkamp-Smith et al. 2012). Observations of diffusion of small and large inclusions in giant vesicles confirm that their motion is consistent with either Saffman–Delbruck or Hughes–Pailthorpe–White theory, as appropriate (Cicuta et al. 2007; Petrov et al. 2012; Noruzifar et al. 2014).

Membrane Viscosity

Even though diffusion laws are observed to be consistent with theoretical predictions, quantitative measurements of membrane viscosity are not trivial. In membranes, measured viscosity values depend to a remarkable extent on the method of measurement. One category of measurements observes microscopic-scale membrane dynamics during equilibrium fluctuations, electrostatic deformations, micro-falling-ball viscometry, diffusion of large fluid domains, or in-plane circulation; these experiments generally result in viscosity values in the range of 1 – 60 · 10⁻⁹ Pa m s (Honerkamp-Smith et al. 2013; Faizi et al. 2022; Dimova et al. 1999). On the other hand, fluorescence measurements of viscosity using molecular rotors or diffusion of nanometer-sized inclusions, as well as careful simulations, yield viscosity values in the lower range of 10 – 20 · 10⁻¹¹ Pa m s (Fitzgerald et al. 2023). Future experimental and theoretical advances are required to explain the discrepancy between different results.

Interleaflet Friction

When a membrane bends, the two leaflets must slide past each other. The interleaflet friction coefficient describes how easily and how quickly they can do this. This coefficient is needed to understand membrane bending and fluctuation dynamics. It has been measured experimentally by pulling membrane tethers, observing microscopic or nanoscopic membrane fluctuations, or using microfluidic shear stress to slide the upper and lower membrane leaflets past each other (Evans and Yeung 1994; Pfeiffer et al. 1993; Pott and Meleard 2002; Jonsson et al. 2009). However, the resulting estimates of this coefficient span an unexpectedly large range (Kelley and Heberle 2022). Recently, we used a microfluidics-based method to compare interleaflet friction in a series of different membranes, demonstrating how cholesterol content modifies interleaflet friction (Anthony et al. 2022). Additional consistent, systematic measurements of this parameter will provide insight into lipid interactions taking place at the membrane midplane; for instance, simulations predict that friction arises from interdigitation of lipid acyl chains from opposite leaflets and will increase with the fraction of lipids with different length chains (den Otter and Shkulipa 2007; Kelley et al. 2023).

Supported Bilayers

A common experimental strategy is to convert lipid vesicles into surface-supported lipid bilayers, which are flat, stable under solution changes, and permit experiments with angstrom resolution-like atomic force microscopy. When membranes are supported on solid substrates, parallel fluid flow cannot generate lipid circulation. Instead, it exerts shear stress on the membrane. Friction between the substrate and membrane alters lipid packing and mobility, even causing shifts in lipid phase transition temperatures (Charrier and Thibaudau 2005; Gunderson and Honerkamp-Smith 2018). The presence of the substrate creates asymmetry, so that Saffman–Delbruck theory no longer accurately predicts diffusion laws. In order to do so, additional details of membrane structure must be included. Instead of treating the membrane as a single two-dimensional fluid, each leaflet is considered as a separate fluid layer. The upper leaflet interacts with the bulk water; the lower leaflet is coupled both to the upper leaflet and to the thin layer of water underneath the membrane (Fig. 2). Camley and Brown developed a computational method called Regularized Stokelets which accurately predicts motion of lipids and lipid domains in supported bilayers (Camley and Brown 2013, 2019).

Fig. 2.

Left: vesicles are formed from lipid bilayers surrounded on both sides by water or cyto-plasm. Right: solid-supported membranes are asymmetric because the water layer between the membrane and surface is very small. Lipids in the upper leaflet have different mobility and packing than those in the lower leaflet

The layer of water between the membrane and its solid support is approximately one nanometer thick (Johnson et al. 1991). Some experiments indicate that the effective viscosity of this water and other water confined in nanometer-sized spaces, increases dramatically relative to its usual value in bulk water (Kim et al. 2006; Bhattacharyya 2003). For instance, Camley and Brown found excellent agreement with diffusion constants measured in a supported membrane by increasing the viscosity of the water under the membrane by 75x (they also note that that the layer is so thin that the concept of continuum viscosity may become inappropriate) (Camley and Brown 2013; Knight et al. 2010). This prediction also depends on the value of the interleaflet friction coefficient, so more accurate knowledge of membrane viscosity and interleaflet friction will help fully determine the properties of sub-membrane and other nano-confined water.

Cytoskeleton Drag on Lipid Motion

Mammalian outer cell membranes are closely associated with the cytoskeleton, an active, elastic network of actin filaments and associated proteins. In some ways, cell plasma membranes appear to be in a fluid state: FRAP experiments show that plasma membranes of living cells diffuse quickly (Koppel et al. 1976; Wawrezinieck et al. 2005). However, lipids cannot flow freely: lipids cannot not slide rapidly over the cytoskeleton in response to changes in membrane tension (Shi et al. 2018; Gomis Perez et al. 2022). These dynamics resemble observations of lipid bilayers supported on rigid surfaces. In solid-supported membranes, collective lipid flow is prevented by friction with the substrate. As a result, the membrane is in a quasi-fluid state in which individual lipid diffusion constants are high, but typical fluid phenomena like droplet coalescence and circulation in response to flow are prevented (Przybylo et al. 2006; Scomparin et al. 2009; Garg et al. 2007; Stottrup et al. 2004; Blosser et al. 2015). Membrane-cytoskeleton coupling is specific to particular cell types; however, in neurons which are specialized for rapid membrane lipid exchange at the synapse, cytoskeleton-membrane friction was low enough to allow lipid flow to equalize tension (Gomis Perez et al. 2022).

Diffusion of lipids and proteins in living cells is observed to be confined by interactions with actin, although the detailed organization of these interactions remains under discussion (Ritchie et al. 2003; Morone et al. 2006; Chein et al. 2019). Some transmembrane proteins appear to have a specific attachment to actin; after being displaced by forces applied to nanoparticles, they returned elastically to their original positions (Turkcan et al. 2013). One consequence of the fact that motion in membranes is confined to two dimensions is that even a relatively small fraction of immobile obstacles can interrupt lipid flow dramatically, while leaving single-molecule diffusion constants relatively unchanged (Cohen and Shi 2020). A recent experiment demonstrated that while cell membrane tension is localized, it is also actively managed. In neutrophil-like cells, localized perturbations to the cytoskeleton via both mechanical and optogenetic methods resulted in a rapid alteration to membrane tension on the opposite side (De Belly et al. 2023). Close and highly regulated association with the cytoskeleton means that most cell plasma membranes are highly unlikely to circulate under flow, as a vesicle would. However, they can still react to flow; in particular, flow can create lateral transport of membrane proteins. The next section collects experimental results that demonstrate how flow modifies membrane protein distributions.

Lipid Flow in Cell Plasma Membranes

In living cells, plasma membrane lipids are constantly removed and replenished by internal membrane trafficking. If endocytosis and exocytosis rates are different in a particular location, lipid flow must occur to rectify the imbalance. Recent experiments demonstrate that at the growing tip of a fission yeast cell, exocytosis generates lipid flow away from the tip with an approximate velocity of 3 nm/s. This flow carries membrane proteins along with it. Transport is more effective for proteins that spend a longer time bound to the membrane than for those that rapidly cycle on and off, so membrane flow differentially sorts proteins according to their membrane residence time (Gerganova et al. 2021) (Fig. 3). Similar flows are likely to be generated at the growing tips of pollen tubes, root hairs, and neuronal growth cones (Gerganova and Martin 2023). Persistent membrane flow can even move entire cells: by directing asymmetric lipid traffic in macrophages, O’Neill et al. created treadmilling of the membrane relative to the cytoplasm, allowing the cells to migrate without forming adhesions (O’Neill et al. 2018).

Fig. 3.

Top: Localized exocytosis in yeast cells results in “fountain” membrane flow away from the growing tip. Experiments show that membrane-associated proteins with fast diffusion and high rates of membrane dissociation are not affected by membrane flow (RitC-sfGFP monomer), whereas those with slower diffusion constants and longer dwell time on the membrane (RitC-sfGFP tandem repeat) are depleted from the cell tip. Bottom: blood flow over endothelial cells applies a small force to glycoproteins, which can result in long-range transport of extracellular lipid-anchored glycoproteins such as glypican-1 but not of actin-associated transmembrane proteins, such as syndecan-1. Neither protein can cross-intracellular junctions

Flow-Mediated Protein Sorting

Protein sorting can also occur in cases where the fluid surrounding the cell flows faster than the membrane lipids. Lipid-anchored proteins, which are attached to only one leaflet of the membrane, can travel past any cytoskeletal obstacles and traverse large distances. In this case, proteins are sorted according to the size of their aqueous domain. Microfluidic experiments have demonstrated that protein mixtures on supported bilayers can be separated using this method (Jonsson and Jonsson 2015).

Trypanosomes are single-celled parasites which use a single flagellum to swim through their host organisms’ blood-stream. Their swimming motion may be more important for self-defense than for transport. Engstler et al. showed in 2007 that host organism immunoglobulins recognize and bind to GPI-anchored coat proteins covering the trypanosome surface. The flows generated by swimming sort the protruding IG-coat protein complexes to the posterior end, where they are endocytosed and destroyed (Engstler et al. 2007). This helps the parasite evade the host immune system.

A recent review notes that detection of typical arterial flows poses a challenge because the shear stresses applied directly to the artery surface are so small, exerting only femtonewton-sized forces on an area the size of a single protein (Tanaka et al. 2021). This is smaller than available energy from thermal noise, about 4.1 piconewton-nm (Gomez et al. 2021) and much smaller than the forces known to change the conformation of mechanosensitive proteins, such as integrins and talins, which extend under 1–5 piconewtons of force (Sun et al. 2019). Similarly, mechanosensitive ion channels like Piezo1 open in response to membrane tension changes of approximately 1 piconewton/nm (Lewis and Grandl 2015). Cells could circumvent this difficulty by sensing flow using micron-scale spatial distributions rather than conformational changes in individual proteins.

We recently tracked the motion of lipid-anchored proteins across supported lipid bilayers in microfluidic channels. Our experiments confirmed that forces on the order of 0.1 to 0.9 femtonewtons, generated by 1-5 Pa of shear stress, are sufficient to create optically resolvable concentration gradients (Ratajczak et al. 2023). The Peclet number is a dimension-less constant that indicates how important advective transport is, relative to diffusive transport. When it is larger than one, advection is dominant; when it is much smaller than one, transport is mostly diffusive. We measured the diffusion constant for streptavidin molecules bound to biotinylated lipids of approximately 0.4 μm²/s and found that applying 2 Pa of shear stress resulted in an average drift velocity of 30 nm/s. The Peclet number for streptavidin over protein-sized length scales were very small, approximately 0.0006 if L = 5 nm. However, the Peclet number exceeds one for a length scale of 50 μm. Thus, weak hydrodynamic forces can generate significant concentration gradients of mobile proteins over distances of tens of micrometers, but not over nanometers.

There is some experimental evidence that endothelial cell proteins move under flow. The glycocalyx, or surface coating of glycoproteins that covers endothelial cells, is made up of a mixture of transmembrane and GPI-anchored glycoproteins. Glycoproteins bind various oligosaccharides, including heparan sulfate. A few minutes after flow was applied to endothelial cells in vitro, heparan sulfate was observed to collect at the downstream edge of the cells, likely due to advective transport of the GPI-anchored glycoproteins (Zeng et al. 2013; Ebong et al. 2014) (Fig. 3). However, whether this transport is related to flow signaling remains unproven.

Supported Bilayer Lipid Packing is Modified by Shear Flow

Finally, there is evidence that flow itself alters lipid organization. Fluorescence anisotropy experiments show that plasma membrane viscosity in endothelial cells is quickly and reversibly lowered in response to shear stress (Yamamoto and Ando 2013; Haidekker et al. 2000). In living cells, this can be explained by active processes, such as release of a membrane reservoir via exocytosis, unfolding of membrane invaginations, or active modification of lipid concentrations. However, a simpler mechanical explanation is possible. Under flow, single-phase membrane patches supported on glass exhibit a substantial and partially reversible increase in their apparent area. The physical origin of this increase has not yet been identified; its magnitude is independent of lipid order parameter, shear stress magnitude, and membrane patch size. Neutron reflectivity measurements are consistent with a simultaneous decrease in supported membrane thickness during flow (Miller et al. 2023). Further experiments are needed to determine whether this flow response is unique to membranes supported on a rigid substrate. If it is a general phenomenon, it may help explain changes to membrane tension and packing observed in living cells.

Perspective

Although flow sensing impacts many aspects of physiology, its mechanistic details remain surprisingly obscure. Fluid shear stress from blood flow poses a particular conundrum, since cells are known to respond to forces that appear to be smaller than reasonable estimates for lower limit of single-molecule force detection. The cytoskeleton-membrane complex provides an organized, soft scaffold for protein movement which constrains membrane proteins to the two-dimensional surface but allows them freedom to move laterally across it. Femtonewton-sized forces can rearrange lipid-anchored proteins, potentially providing a role for protein spatial organization in flow sensing.

Additional work remains to be done in order to elucidate the relationship between flow transport of membrane proteins and physiological responses by cells. But it seems clear that the architecture and fluidity of the lipid membrane-cytoskeleton complex is tuned to permit communication, transport, and sorting of membrane proteins. This complex structure is required to provide a unique physical environment for membrane proteins, one that is particularly responsive to flow.