Since the inception of the fluid mosaic model, cell membranes have come to be recognized as heterogeneous structures composed of discrete protein and lipid domains of various dimensions and biological functions. The structural and biological properties of membrane domains are represented by CDM (cholesterol-dependent membrane) domains, frequently referred to as membrane ‘rafts’. Biological functions attributed to CDMs include signal transduction. In T-cells, CDMs function in the regulation of the Src family kinase Lck (p56lck) by sequestering Lck from its activator CD45. Despite evidence of discrete CDM domains with specific functions, the mechanism by which they form and are maintained within a fluid and dynamic lipid bilayer is not completely understood. In the present chapter, we discuss recent advances showing that the actomyosin cytoskeleton has an integral role in the formation of CDM domains. Using Lck as a model, we also discuss recent findings regarding cytoskeleton-dependent CDM domain functions in protein regulation.
Over the last four decades, our perspective of cell membranes has changed from their structure being a homogenous lipid bilayer sandwich containing embedded proteins, represented by the fluid mosaic model , to one where cell membranes are composed of domains of distinct protein and lipid composition. In the past two decades, many studies of membrane structure have focused on studies of domains composed of proteins and lipids that exhibit a cholesterol-dependent clustering in the plasma membrane. Often termed ‘rafts’, in the present chapter we use the label CDM (cholesterol-dependent membrane) domain for these structures to emphasize the central role that cholesterol has in their formation and maintenance. Early studies of CDM domains exclusively employed detergent extraction to study the domains, an approach that limited interpretation of the biological and physical properties of the domains. This state of the field has since been replaced as sophisticated imaging studies have measured CDM domains in intact membranes with nanoscale precision. Collectively, these studies show that CDM domains occur in most cells that contain cholesterol, and with nanoscale and larger dimensions.
An emerging theme from imaging studies of CDM domains is that clustering of associated reporters is often sensitive to disruption of the cortical cytoskeleton [2–4]. Similarly, recent data show evidence that the actomyosin cytoskeleton establishes ordered lipid environments in the plasma membrane , a physical state that is thought to be important for CDM domain formation. In the present chapter, we review these developments and the impact of these findings on understood roles of CDM domains in protein regulation.
Soapy extracts to patchy membranes: evolution in the discovery of cholesterol-dependent domains
Early identification and study of CDM domains was based on the notion that their composition is represented by a DRM (detergent-resistant membrane) fraction that forms upon lysis of cholesterol-containing cells by mild, non-ionic detergents at low temperatures. Brown and Rose  first showed that DRMs contain cholesterol and sphingolipids, a finding that was significant since it was understood that these lipids can interact to form a discrete Lo (liquid ordered) phase within fluid phase bilayers . Thus, identification of DRMs enriched with cholesterol and sphingolipids was interpreted as evidence of cholesterol-dependent lipid domains in cell membranes, analogous to the Lo phase domains shown in in vitro systems by Sankaram and Thompson , and later to include cholesterol and lipids other than sphingolipids . In this model, membrane domains composed of ordered-state lipids and associated proteins are selectively resistant to solubilization by mild detergents at reduced temperatures, forming intact membrane sheets and vesicles during detergent lysis of cells. Consistent with this interpretation are findings that show that Lo phase lipids are resistant to solubilization in conditions where DRMs form , and that proteins that occur in DRMs are often modified with lipid groups, such as palmitate, that preferentially partition into ordered lipid environments .
As reviewed in Chapter 5 in this volume , preparation of DRMs has been a frequently employed method towards identifying and characterizing presumptive CDM-associated proteins and lipids. Nevertheless, the inherently disruptive nature of detergent lysis, together with evidence that detergents used for DRM preparation can artificially cause domain formation , has led some investigators to contend that DRMs are misrepresentative of domains in intact membranes. Despite an initial over-reliance on detergents to characterize CDM domains, separate approaches show where DRM-associated lipids and proteins exhibit a specific and cholesterol-sensitive clustering in the plasma membrane (e.g. [4,14–16]), properties that we operationally define as CDM domains. Furthermore, imaging studies show CDM domains to have a structural hierarchy based on size (reviewed in ), ranging from nanoscale dimensions to macrodomains that are micrometres in diameter (Figure 1). Examples of functionally distinct CDM macrodomains are the IS (immunological synapse) in stimulated lymphocytes, the leading edge and uropod of motile cells, and integrin-enriched adhesion complexes. Protein association with CDM domains is thought to occur through a combination of protein–lipid and protein–protein interactions .
CDM domain heterogeneity in the plasma membrane
FRETing out lipid domains formed by the cytoskeleton
A finding that emerged from imaging studies of CDM domains was their association with the underlying cortical cytoskeleton, first suggested by observed co-enrichment of actin filaments with CDM macrodomains [19,20] and data showing that DRMs contain actin and proteins and lipids that associate with actin filaments at the plasma membrane (e.g ). More recently, techniques that resolve nanoscale clustering of membrane proteins, such as FRET and electron microscopy, show that disrupting the cytoskeleton using drugs such as latrunculin B can be as effective as reducing plasma membrane cholesterol towards de-clustering CDM-associated proteins [2,3]. Conversely, enriching actin filaments in the cortical cytoskeleton using jasplakinolide can significantly and specifically elevate clustering of protein markers of CDM domains .
Although separate studies provide compelling evidence of associations between the actin cytoskeleton and CDM domains, whether the cytoskeleton participates in domain formation has not been known. This has been an important question, since our understanding of CDM domain formation is constrained by a difficulty in envisioning how the fluid and dynamic state of cell membrane lipids can be sufficiently ordered to form domains sufficient in size and time scale to be biologically meaningful. Accordingly, models were proposed where membrane-associated proteins are a driving force in forming CDM domains, such as by tipping lipids towards an ordered state through interactions with the proteins ; observed interactions of filamentous actin with CDM domains suggested the cytoskeleton as one candidate in protein-mediated domain formation. Evidence of an ordering of lipids by cytoskeletal filaments was shown to occur in a model membrane system containing attached actin filaments . However, whether this was representative of cytoskeletal ordering of lipids in the plasma membrane was not known until recently, when we showed by FRET an ordering of long-chain lipophilic fluorescent probes by the cytoskeleton .
The rationale for the FRET-based approach for measuring lipid ordering in nanodomains is summarized in Figure 2. Specifically, energy transfer is measured between lipophilic donor–acceptor pairs with a high affinity for ordered lipid environments. FRET is highly sensitive to the distance between the donor and acceptor, with FRET efficiency dropping by the inverse 6th power of the distance between the pair. Consequently, deviation of distances near the Förster radius of the donor–acceptor pair, where FRET efficiency is 50% and is typically between 5 and 10 nm, will result in large changes in FRET efficiency. Co-clustering of probes within domains of this size range will therefore cause an elevated FRET signal.
Detection of nanoscopic membrane heterogeneities by FRET
Using this strategy, we demonstrated lipid ordering events in the plasma membrane that were disrupted by either reducing the availability of cholesterol or disrupting the actin cytoskeleton by drug treatment . Related to these findings, the Parmryd group showed a direct correlation between cytoskeleton and plasma lipid ordering by measuring the generalized polarization of lipophilic probes . Collectively, we interpret these results as showing that CDM domains are formed through interactions between the cytoskeleton and the plasma membrane, such that perturbing the cytoskeleton has the same effect as removing available cholesterol towards disrupting the domains.
We also showed that lipid ordering in CDM domains is sensitive to sequestration of phosphoinositides . Similarly, CDM macrodomains formed by cross-linking cholera toxin B subunit co-enrich with PI(4,5)P2 (phosphatidylinositol 4,5-bisphosphate) and its metabolites . PI(4,5)P2, and its phosphoinositide 3-kinase product phosphatidylinositol 3,4,5-trisphosphate, are essential co-factors for interactions between the cytoskeleton and the plasma membrane. Thus, findings showing both phosphoinositide-dependent lipid ordering and association of phosphoinositides with CDM domains are important evidence that the cytoskeleton participates in CDM domain formation. Interestingly, lipid ordering was also sensitive to inhibition of NM II (nonmuscle myosin II) , suggesting that cytoskeletal tension is another important factor in forming CDM domains.
Cytoskeleton mechanics in domain formation
The notion that the cytoskeleton participates in CDM domain formation is attractive, since it combines acknowledged dynamic and plastic properties of the cytoskeleton with understood effects of concentrating specific pools of proteins and lipids in domains for biological functions. Nevertheless, how is the cytoskeleton with associated myosin motor activity able to affect CDM domain formation? One model for how this may occur is illustrated in Figure 3. Our interpretation is based on observed orthogonal compressive forces of actin filament networks on lipid bilayers  and the documented effect of compressive forces on lipid ordering, such as that shown in seminal work by McConnell et al. . We posit that CDM domains occur by orthogonal compression by the cytoskeleton, this triggering cholesterol-dependent lipid ordering in the membrane and resulting in separation of ordered lipid domains within the bilayer. Contributing to the compressive effects will be tension imparted on the actin filaments and bilayer through myosin II activity, thus accounting for observed declustering of CDM domain-associated proteins by inhibiting NM II .
Model for CDM domain formation by cytoskeletal compression on the membrane
Our model evoking mechanical force as a major determinant in domain formation departs from traditional thinking, where lipid domains are thought to form exclusively from chemical properties of the lipid and protein components within the bilayer. Importantly, mechanotransduction of stimuli is understood to be important for cellular processes, such as those relating to cell signalling and adhesion [28–30], and that the cytoskeleton is a primary conductant of mechanical force within the cell . Thus, cytoskeleton-mediated mechanotransduction is pervasive within the cellular compartment, and available to impact cell membrane structure.
CDM domains and the cytoskeleton in Src kinase regulation
An attractive feature of membrane compartmentalization by structures such as CDM domains is that it provides a mechanism for concentrating specific proteins and lipids for spatiotemporal regulation of biological processes. This is particularly important in signal transduction, where cellular responses to stimuli occur as a result of complex interactions between multiple intermediates. Consistent with this interpretation are data that show CDM domains are critical for certain signal transduction pathways to occur .
Importantly, CDM domains also function in the steady-state regulation of cell signalling. This includes sequestering associated effectors from activators that are restricted to membrane outside of the domains. Such domain-mediated regulation of signalling effectors has been evidenced with members of the Src family of non-receptor protein tyrosine kinases, including Lck (p56lck). Lck is expressed in T-cells, where it initiates tyrosine phosphorylation cascades following engagement of the TCR (T-cell receptor) with peptide-MHC . Efficient regulation of Lck is important in preventing T-cell neoplasms . Detergent fractionation and cell imaging experiments establish that Lck associated with CDM domains has a greater phosphorylation of its regulatory Tyr505 and reduced activity relative to non-domain pools of Lck [35,36]. Down-regulation of Lck in CDM domains is due to sequestration from its activator CD45, which is excluded from CDM domains. Altogether, these data show that association of Lck with CDM domains is one mechanism for regulating Lck-dependent signalling.
Given the evidence that the cytoskeleton plays a critical role in CDM domain formation, it follows that regulation of Lck will be compromised in conditions that disrupt the cytoskeleton. We addressed this question by measuring the phosphorylation of Lck Tyr505 in conditions that disrupt the cytoskeleton and associated lipid ordering. These conditions caused dephosphorylation of Lck Tyr505 , which required expression of CD45. In addition to reducing phosphorylation of Lck Tyr505, actin depolymerization also elevated phosphorylation of Tyr394, which reports activation of Lck .
Collectively, these data suggest the model illustrated in Figure 4, where steady-state regulation of Lck is maintained by CDM domains that are formed and maintained by the cytoskeleton. The domains exclude CD45, thereby maintaining domain-associated Lck in a state of attenuated activity through phosphorylation of Tyr505. Not shown in Figure 4 are T-cell signalling proteins that co-associate with Lck in CDM domains and that function in T-cell activation following triggering of the TCR. Examples are CD4, CD8 and LAT (linker for activation of T-cells). Although proteins such as CD4, CD8 and LAT may be proximal to Lck in CDM domains, their expression is not necessary for Lck association with the domains and subsequent sequestration from CD45 . We exclude these additional proteins from our model in Figure 4 to emphasize where interactions between Lck and CD45 are regulated through Lck association with CDM domains, which, in turn, are effected by the cytoskeleton. Interestingly, recent data from our group show that once T-cell activation begins, CD45 is transiently targeted to CDM complexes to become proximal to Lck, where it can down-regulate Lck through dephosphorylation of Tyr394 within the Lck kinase domain . Eventually in stimulated T-cells, Lck is sequestered from CD45 by association of Lck with the central zone of the IS, from which CD45 is excluded.
Cytoskeletal regulation of Lck by CDM domains
Importantly, our findings do not preclude the possibility of cytoskeletal regulation of Lck by lipid-independent mechanisms, such as sequestration of Lck from CD45 through direct interactions with components of the cytoskeleton, and possible effects of the cytoskeleton on Lck self-oligomerization . Although lipid-dependent and -independent mechanisms may be at play in the cytoskeletal regulation of Lck, these need not be mutually exclusive. Thus, further study is not only necessary to identify these alternative mechanisms, but also to determine their contribution to Lck regulation relative to that of the cytoskeleton-dependent CDM domains.
In our model in Figure 4, cellular processes that alter either the F-actin content of cells or actin association with the plasma membrane are predicted to also alter the physical properties of CDM domains, such as their size and composition. Consistent with this interpretation, enrichment of the cortical cytoskeleton with F-actin following cross-linking surface proteins coincides with expansion of CDM domains to form macrodomains composed of relatively ordered lipids [19,24]. Because many cellular events are accompanied by gross changes in the cytoskeleton, cytoskeletal modulation of CDM domain formation may soon be recognized as a frequent mechanism for tuning biological functions in cell membranes.
Descriptions of cell membrane domains related to DRMs have since been corroborated by data from separate and non-invasive methods to study cell membranes. Accordingly, many of the proteins and lipids that occur in DRMs have now been shown to exhibit a specific and cholesterol-dependent clustering into domains that range from nanoscopic dimensions to macroscopic structures micrometres in diameter. Studies show an association of CDM domains with the cytoskeleton across the spectrum of domain sizes and compositions, and that the cytoskeleton is an important determinant in forming CDM domains.
Using the Src family kinase Lck as a model to understand the biological properties of cytoskeleton-dependent CDM domains, Lck phosphoregulation is attenuated in conditions that disrupt the cytoskeleton and associated domains. Although these findings do not preclude lipid-independent mechanisms for the cytoskeletal regulation of effectors such as Lck, they are provocative in that they are entirely consistent with anticipated effects of disrupting the cytoskeleton on CDM domains and subsequent elevated interactions with its activator CD45.
Altogether, our current understanding of CDM domains and their interactions with the cytoskeleton provide an intriguing example of lipid-dependent spatiotemporal regulation of membrane events by a dynamic cytoskeleton network. Actin polymerization and remodelling of the cytoskeleton is a global feature of cell activation, often occurring very soon following initiation of activation stimuli. Based on the discussion here, this remodelling of the cytoskeleton is anticipated to cause reorganization of plasma membrane structure to alter its make-up of CDM domains. These changes in membrane structure are expected to have significant biological outcomes, exemplified by the altered regulation of Lck when the cytoskeleton and associated CDM domains are disrupted.
Although studies to date have documented many of the physical and biological properties of CDM domains, important features of their structure remain elusive. Thus, further research is needed to better understand fundamental physical properties of the domains, such as the lifetime of the nanoscopic structures, which predominate in steady-state conditions, and the movement or dynamics of proteins and lipids between CDM domain and non-domain regions of the plasma membrane. As suggested by our model in Figure 4 regarding Lck regulation through association with CDM domains, such movement of proteins between these separate membrane environments could have significant effects on protein activity. We anticipate that further studies of the cytoskeleton, CDM domains and domain-associated proteins will show more examples of biological properties from the cytoskeleton regulating domain formation.
Looking back on four decades of research following description of the fluid mosaic model, a persistent theme has been the inherent complexity of cell membranes, represented in part by protein and lipid domains with identifiable biological functions. The CDM domains represent only a single variety of domains, and are themselves heterogeneous in composition and function. Many other types of domains are expected to occur, each through an array of protein and lipid interactions, and thus with physical and biological properties distinct from that of the CDM domains. Going forward, perhaps the most significant challenge in the field is a refined discrimination of separate species and sub-species of domains, and the biological functions that are specific to each. Achieving this distinction is certain to provide a more accurate description of membrane structure, as well as an improved understanding of the role of membrane structure in the biology of the cell.
Plasma membrane heterogeneity is exemplified by CDM domains of varying composition, size, and function, and includes domains formed by interactions between the actin cytoskeleton and the membrane.
The dynamics and plasticity of the actin cytoskeleton provides a mechanism for rapid reorganization of CDM domains to meet varying biological demands within the cell.
Compressive forces exerted by the cytoskeleton on the membrane may be one mechanism by which the cytoskeleton contributes to CDM domain formation, yet further study is necessary to better understand this and other biophysical properties associated with CDM domains.
This work was supported by the Oklahoma Center for the Advancement of Science and Technology [grant number HR11-42] (to W.R.).