Sensitivity of Peripheral Membrane Proteins to the Membrane Context: A Case Study of Phosphatidylserine and the TIM Proteins

Abstract

There is a diverse class of peripheral membrane-binding proteins that specifically bind phosphatidylserine (PS), a lipid that signals apoptosis or cell fusion depending on the membrane context of its presentation. PS-receptors are specialized for particular PS-presenting pathways, indicating that they might be sensitive to the membrane context. In this review, we describe a combination of thermodynamic, structural, and computational techniques that can be used to investigate the mechanisms underlying this sensitivity. As an example, we focus on three PS-receptors of the T-cell Immunoglobulin and Mucin containing (TIM) protein family, which we have previously shown to differ in their sensitivity to PS surface density.

1. Introduction

Protein-lipid interactions underlie several biological phenomena such as the firing of a neuron, catalysis of chemical reactions, or transport of cargo across membranes. As opposed to membrane-embedded proteins which associate with lipids in plane, peripheral membrane-binding proteins are water-soluble and directly bind or adsorb to lipids. Once associated with the membrane, these proteins serve as structural scaffolds, induce changes in local membrane properties, and mediate cellular-scale processes such as phagocytosis, clot formation, and neurotransmitter release. Binding to a lipid membrane can cause downstream gene activation or repression, clustering of surface receptors, or membrane permeabilization. The regulation of these processes is governed, in part, by the specificity of these proteins for particular lipids and membrane conditions [1–3].

Cell membranes are comprised of thousands of distinct lipid species. The lipid composition of cell membranes significantly varies among different cell types, organelles, and organisms [4,5]. Thus, the two-dimensional lipid membrane effectively encodes the identity of a cell or organelle. Changes in the lipid composition, asymmetry, or morphology of cell membranes reflect changes in the cell’s internal state [6–8], allowing the membrane to transmit dynamic and contextual information to other cells through the recognition of peripheral membrane-binding proteins. Peripheral membrane-binding proteins recognize particular membranes through one or more of the following methods: The burial of solvent-exposed amino acids, neutralization of exposed surface charges, post-translational modifications such as palmitoylation, and specific lipid coordination. The first three mechanisms depend on bulk membrane properties such as fluidity, curvature, or charge, while the last relies on the presence of a specifically recognized lipid ligand [9]. Understanding how these mechanisms mediate recognition of specific membranes would lead to numerous applications, for example drug delivery platforms that target particular cells based on their distinct membranes.

1.1. PS Receptors Constitute a Large Class of Peripheral Membrane-Binding Proteins

Several lipid-binding proteins specifically recognize phosphatidylserine (PS) [10], which normally resides only in the inner leaflet. Upon exposure to the outer leaflet of cell membranes, PS serves as a signal for multiple pathways, most prominent of which is apoptosis [11–17]. The clearance of apoptotic cells by PS-recognizing phagocytes prevent damage to surrounding cells and tissue [18]. During apoptosis, PS exposure is accompanied by changes in the local extracellular and intracellular calcium concentrations, as well as membrane fluidity, which do not necessarily accompany PS exposure in other signaling contexts [19–22]. Constitutive exposure of PS is not sufficient to induce phagocytosis [23], suggesting that the membrane context of PS exposure is a significant component of the signal. PS exposure in apoptosis is significantly higher and less transient than non-apoptotic PS signaling [24] implicating PS surface density as an important contextual signal [14,25]. Annexin V, a PS-receptor suspected to inhibit blood coagulation by outcompeting coagulation factors with C2 and Gla PS-binding domains, binds cooperatively to PS and calcium, indicating a preference for membranes with high PS surface density [25,26]. Similarly, Protein Kinase C (PKC) family members also bind PS cooperatively [27] and the coagulation factor Protein Z binds both PS and phosphatidylethanolamine (PE), another lipid normally enriched on the inner leaflet that flips out during apoptosis [28]. Considering that PS receptors are each specialized for particular functions, such as the recognition of apoptosis, signal transduction, exocytosis, phosphorylation of phosphoinositide, and coagulation [10], it is likely that PS-binding proteins recognize both PS and the membrane context of PS presentation. The diversity of PS presentation seemingly requires a complementary diversity of PS-receptors.

2. Methodology for Determining Protein’s Sensitivity to the Membrane Context

The Enzyme-Linked Immunosorbent Assays (ELISA), Surface Plasmon Resonance (SPR), Langmuir monolayer systems, lipid microarrays, liposome microarrays, fluorescence spectroscopy, and Isothermal Titration Calorimetry (ITC) have been integral in measuring the binding affinity of peripheral membrane-binding proteins for lipids [29–32]. Fluorescence spectroscopy and ITC have the advantage of measuring the binding of liposomes, soluble unilamellar lipid vesicles that require no membrane immobilization and serve as models for natural membrane structures of various sizes such as vesicles, organelle membranes, or cell membranes [33]. In contrast to ELISA assays, SPR, ITC and fluorescence spectroscopy do not require the binding of secondary biochemical agents to measure affinity. However, fluorescence spectroscopy requires that proteins bury a tryptophan residue when bound to the membrane [34]. With the exception of lipid microarrays all of these techniques can be used to measure the contribution of the membrane context by conducting experiments with varying membrane and buffer conditions. These properties can be varied while maintaining the same lipid composition for Langmuir monolayer assays [35], however liposomes are more restricted since they are self-assembled structures. The influence of bulk membrane properties such as fluidity and charge density can be probed by varying the lipid composition of liposomes.

Structural techniques which, at best, only yield the protein structure complexed with a single lipid cannot identify the residues responsible for conferring sensitivity to the membrane context. Techniques such as lipid cubic phase crystallography and cryo-electron microscopy (cryo-EM) of 2D crystals, Nuclear Magnetic Resonance (NMR), and Electron Paramagnetic Resonance (EPR) have been utilized to determine bound-structures of peripheral membrane-binding proteins to short-chained lipids or mimetic bilayers such as bicelles, micelles, or lipid nanodiscs [30,36–39] However, these mimetic membranes can significantly differ from bilayer or monolayer membranes and may affect the protein’s configuration. Fourier-Transform Infrared Spectroscopy (FTIR) approaches have been used with monolayer and bilayer membranes to obtain membrane-bound protein structures [40]. While these approaches sometimes identify the membrane burial of specific residues, they do not easily yield the overall surface burial of the protein [36,41]. The overall insertion of the protein is a highly informative structural parameter that correlates with the interaction strength between the protein and the membrane. The stronger the interaction, the more energy that is available for the protein to displace lipids in the membrane.

Reflectivity techniques that utilize x-rays or neutrons have been used to probe the orientation of proteins bound to lipid monolayers [33,42–51]. Reflections are scattering events caused by changes in electron density or scattering length density (in the case of neutrons) normal to the interface. As a result of this scattering geometry, in-plane structural information is not directly obtainable by reflectivity experiments. Reflectivity is often used to characterize systems with homogeneous or periodic in-plane structure. However, globular proteins do not form a homogeneous, adsorbed layer, instead they patchily cover the lipid monolayer and are randomly oriented in-plane. The one-dimensional electron density corresponding to such a system is an average over regions of the monolayer with proteins in random in-plane orientations and regions lacking proteins altogether. Thus, unlike crystallographic structures, reflectivity-derived protein structures cannot be directly “solved” but must be inferred from a maximum likelihood fitting procedure that compares the measured reflectivity to simulated reflectivity produced from a structural model of the electron density. This model assumes a rigid configuration of the bound protein and thus requires an initial protein structure obtained from crystallography, NMR, or molecular dynamics (MD) to represent the bound configuration. This approach was successfully applied to PKCα [42,43,52] and then later to TIM4 [53], yielding resolved membrane-bound structures for both proteins. However, if the protein structure is unknown, reflectivity can only estimate the average length of the adsorbed protein layer normal to the membrane, this information can still be used to eliminate candidate bound-states that result in adsorbed layers which are too short or too long. Proteins with a unique binding face or pocket that determines a consistent, well-defined ensemble of binding orientations are systems best-suited to be studied by reflectivity methods. If the membrane-bound state consists of multiple binding configurations, this approach is highly limited in its ability to extract orientation parameters.

2.1. A Combination of Approaches Determines the Sensitivity of Protein Binding to the Membrane Context

Uncovering the mechanisms by which proteins can “read” the membrane context requires the use of structural and thermodynamic approaches in tandem. For example, we have employed a specific suite of techniques to measure the sensitivity of PS-receptors to PS surface density. We used tryptophan fluorescence to measure the binding of PS receptors to liposomes of various lipid compositions, x-ray reflectivity to determine the membrane-bound structure of PS receptors, and MD to both corroborate the stability of the membrane-bound state and implicate residues that confer sensitivity to PS surface density. These techniques form a feedback loop, depicted in Fig. 1, where the results of one technique inform and inspire experiments with the others. The residues implicated by reflectivity and MD inspire tryptophan fluorescence experiments with PS-receptors bearing alanine substitutions of these residues. In turn, the membrane conditions that maximize binding in tryptophan fluorescence techniques are chosen for reflectivity experiments.

Figure 1.

The combination of tryptophan fluorescence, x-ray reflectivity and MD simulations characterize the sensitivity to the membrane context of PS-receptors. Fluorescence experiments measure the thermodynamics of the membrane context. A) Tryptophan fluorescence spectra of the unbound and bound states differ in amplitude and wavelength of maximum emission corresponding to solvent-exposed and membrane-inserted tryptophan respectively. Partially bound states are fit with linear combinations of the bound and unbound spectra, yielding the bound fraction. The results of these experiments provide the membrane and buffer conditions to maximize binding in reflectivity experiments and MD. B) X-ray reflectivity determines the bound orientation of the protein, implicating residues that confer sensitivity to the membrane context as measured in fluorescence experiments, and providing initial conditions for MD. C) MD simulations corroborate the orientation yielded from reflectivity by demonstrating stability of the bound state and further implicate residues involved in sensing the membrane context. Additionally, MD simulations can provide a refined structure to improve analysis of reflectivity data. Figure adapted from Tietjen et al. (2014) [53].

A major benefit of combining MD simulations and reflectivity is that MD-equilibrated structures can be used to fit reflectivity data when crystal structures are not representative of the membrane-bound structure. Typically, studies seek correspondence of membrane bound structures yielded by MD and reflectivity or another structural technique such as NMR [54–58]. We believe reflectivity and MD are more effectively coupled when the MD structure serves as the model for analysis, as was done for direct calculation of scattering length profiles for neutron reflectivity [59]. There are many ways to couple MD simulations to orientation fitting of x-ray reflectivity data. An initial analysis of the reflectivity using the crystal structure yields orientation parameters that can then be used as initial conditions for a simulation of the protein docked to a membrane. The equilibrated protein structure resulting from this simulation can subsequently be used to reanalyze the reflectivity data and improve the fit quality. Alternatively, MD approaches that simulate the binding process and do not rely on an initial docked orientation for the protein can be utilized to determine the bound state for reflectivity analysis.

Indeed, we found structures resolved from simulations employing Highly Mobile Membrane Mimetic (HMMM) models provide better fits than those resulting from docked simulations. In the HMMM model, a large fraction of the hydrophobic core of the bilayer is replaced with dichloroethane (dCIE), a hydrophobic solvent on which the short-tailed lipids diffuse much more freely, [60–62]. Using this membrane with lipid mobility enhanced by 1–2 orders of magnitude, the protein of interest can be initially placed at a sufficient distance away from the bilayer and then allowed to bind the membrane in an unbiased manner within hundreds of nanoseconds of simulation time. Once equilibrium of the bound state is established, the dCIE solvent is replaced with the full lipid tails and the system is further equilibrated. Compared to docked simulations, the HMMM model is far less likely to result in a bound state that is kinetically trapped in a metastable configuration.

Taken together, these three techniques can be used to characterize the sensitivity of PS-receptors to the membrane context, complementing one another. To illustrate the value of this combination of approaches, we choose to focus on its application to study three PS-receptors from the T-cell Immunoglobulin and Mucin domain containing (TIM) protein family. The three TIM proteins have distinct physiological roles and thus serve as a microcosm for the wider diversity of PS-receptors.

3. TIM Proteins

There are three human proteins in the TIM family and four murine proteins, of which TIM1, TIM3, and TIM4 are homologous to the human protein family members [63]. The three human TIM proteins specifically recognize a single PS coordinated by exactly one calcium ion in a conserved binding pocket located in the immunoglobulin domain [64–67]. This pocket is preserved in the immunoglobulin domains of the murine homologues, the subjects of this review. The TIM proteins are expressed on several cell populations; TIM1 and TIM3 are predominantly expressed on T cells and B cells [68,69] while TIM4 is mostly expressed on macrophages [64,70,71]. All of these cells are believed to use apoptotic cells as a signal for mounting an immune response in the form of tolerance induction, inflammation, or phagocytosis. It is thus natural to question why all cells do not use the same, one TIM protein or even just one PS-receptor. Instead, there are three distinct PS-binding TIM proteins – the same three conserved in mice, rats, and humans – and even more PS-receptors in general. Since non-apoptotic cells expose PS at lower surface densities than apoptotic cells, it would be reasonable to hypothesize that PS-receptors involved in phagocytosis would be sensitive to PS surface density. For example, TIM4’s selectivity for high PS surface densities might ensure phagocytosis of apoptotic cells rather than otherwise fully functional, PS-presenting cells.

Solution crystal structures of the three TIM proteins have been solved [65,66,72] and each TIM protein contains a tryptophan residue near the PS-binding pocket. The previously described combination of tryptophan fluorescence, x-ray reflectivity, and MD are thus well-suited to characterize the sensitivity of the TIM proteins to PS surface density. Fluorescence experiments with liposomes of varying PS surface densities showed that in addition to their PS specificity, the TIM proteins are sensitive to the surface density of PS in a membrane. Analysis of x-ray reflectivity experiments of TIM4 and TIM1 yielded structures with the PS-binding pockets embedded in the membrane. MD simulations of TIM4 implicated interactions of four positively charged residues with adjacent PS in the membrane. Alanine substitutions of these residues showed a reduction in sensitivity, corroborating the x-ray reflectivity resolved structures.

3.1. TIM Proteins Are Sensitive to PS Surface Density

Thermodynamic characterization of the membrane context requires preparations of membranes with varying compositions and properties. Liposomes prepared with varying amounts of PS can be used to probe the influence of PS surface density on PS-receptor binding. Figure 2A shows the result of experiments performed with liposomes containing 30 mol% PS and 10 mol% PS, titrated as a function of total PS concentration. At the same concentration of total PS, the 30 mol% PS liposomes bound more TIM4 than did the 10 mol% PS liposomes as illustrated in Fig. 2B. If the TIM proteins followed a single-site binding model for PS, the binding curve would be hyperbolic. Instead, we observed pronounced sigmoidicity for TIM4 as seen in Fig. 2C. TIM1 and TIM3 displayed slight sigmoidal character in their binding curves compared to TIM4 and the sigmoidicity of each binding curve was quantified by a Hill model for cooperative binding. The values of the Hill coefficient for each TIM protein are shown in Fig. 2D. Since the Hill coefficients are above 1, the binding of the three TIM proteins are cooperative and thus sensitive to PS surface density [53]. Unlike the cooperativity measured for Annexin V, the PKC family, or Protein Z, the cooperativity of the TIM proteins cannot be explained by multiple PS or Ca binding sites [25,26,28]. Alternative mechanisms can underly cooperative binding such as the protein’s non-specific electrostatic interactions with neighboring PS. These interactions would contribute more to the free energy of binding with higher PS surface density, presenting as cooperative binding. However, these contributions do not necessarily correspond to structural “binding sites” on the protein that are specific for PS. The TIM proteins are only peripherally sensitive to neighboring PS with the PS surface density serving as a membrane-contextual cue, especially in the case of TIM4.

Figure 2.

TIM4 is sensitive to PS surface density. A). At the same overall concentration of PS, TIM4 has a higher affinity for 30 mol% PS-containing vesicles than 10 mol% PS-containing vesicles. B) Cartoon depicting result in A. Green circles represent PS while gray circles represent phosphatidylcholine (PC), a neutral zwitterionic lipid that the TIM proteins do not bind. The two cells have the same number of PS lipids but the 30 mol% sample on the right has them distributed on fewer vesicles. The two cells also have the same number of protein molecules but the 30 mol% sample has more of them bound to vesicles than the 10 mol% sample on the left. C) Titration series of the mol % of PS-containing vesicles where the overall lipid concentration of PC + PS is held constant. The TIM proteins all have sigmoidal dependence on the PS surface density of vesicles. D) TIM4 has a higher Hill coefficient than TIM1 or TIM3. TIM4 and PS have a 1:3 stoichiometry while TIM1 and TIM3 each have a 1:2 stoichiometry with PS. Figure adapted from Tietjen etal. (2014) [53].

3.2. MD Simulations Help Resolve Membrane-Bound State of TIM1

While TIM4 was co-crystalized with PS [65], TIM1 was not [72]. X-ray reflectivity analysis of TIM4 resulted in a membrane-bound structure which sensibly embeds the PS-binding pocket into the headgroup region and the hydrophobic residues into the tail group region (Fig. 3A). Unlike TIM4, TIM1’s crystal structure is in a “closed” conformation (Fig. 3B) which results in a sensible angular orientation but insufficient insertion of the hydrophobic residues when fit to the x-ray reflectivity data. TIM1’s binding pocket is obstructed by adjacent hydrophobic residues which should mediate the protein’s insertion (Fig. 3C). The closed conformation of TIM1’s crystal structure is evidently not a good representative of the membrane-bound structure, another structure must be used instead to obtain a more sensible protein insertion depth. Here, we compare membrane-bound structures of TIM1 resolved from docked and HMMM simulations, as described in section 2.1. Figure 3D depicts an overlay of the docked MD-equilibrated TIM1 with the original crystal structure. The equilibrated docked state’s hydrophobic residues flipped out of the PS pocket, allowing it to bind PS. HMMM simulations of TIM1 similarly flipped out these hydrophobic residues but also resulted in two distinct bound states independent of initial conditions (Fig. 3E). These states are best characterized by their engagement with the bound PS. The first state coordinates both the phosphate and serine groups in the pocket while the second state coordinates only the serine group. R53 is peripherally associated with the membrane in state 1 while this arginine is far removed from the membrane in state 2 due to a difference in the orientation [73].

Figure 3.

The structural basis of TIM protein’s sensitivity to PS surface density. A) Membrane-bound orientations of TIM4 (purple) and B) TIM1 (orange) crystal structures fit to x-ray reflectivity experiments. C) Overlay of TIM4 (purple) and TIM1 (gray) crystal structures (PDB IDs: 3BIB and 2OR8 respectively). TIM1’s PS binding pocket is blocked by a phenylalanine residue, which is extended in the PS-bound TIM4 structure. D) MD simulations using the parameters yielded from the fit of TIM1’s crystal structure result in an equilibrated structure in which the PS binding pocket is open (orange) as compared with the TIM1 closed crystal structure (gray). E) HMMM MD of TIM1 yielded two different bound states characterized by the engagement of R53 and the coordination of the phosphate group of the bound PS. F) From left to right, fits of the docked MD, HMMM state 1, and HMMM state 2 structures to the x-ray data. HMMM state 1 is the only bound structure that buries the leading hydrophobic residues. G) A comparison of the χ² values for each model over all of the data sets normalized to each data set’s χ² value for the crystal structure. Stars represent differences with significance of p < 0.05. The docked MD structure results in an improved quality of fit over the crystal structure but HMMM state 1 results in a significantly improved fit over all other models. H) A comparison of the insertion depths of each model overall of the data sets. HMMM state 1 results in the most insertion. Figure adapted from Tietjen et al. (2014) [53] and Tietjen et al. (2017) [73].

Figure 3F shows the best fit orientations of the docked MD and the two HMMM-derived states to the x-ray data. HMMM state 1 yielded a much-improved fit for the x-ray data compared to the docked MD and crystal structure fits (Fig. 3G). State 2 yielded a poorer fit than all other structures and moreover the orientation did not match the orientation in the simulations (Fig. 3E right panel). Rather than dismissing the simulations of state 2 as non-physical, we speculate that the simulations sampled an infrequently populated microstate which are unrepresentative of the ensemble average bound state measured by x-ray reflectivity. In addition to fitting the reflectivity data better than all of the other models considered, state 1 was the only model that inserted the hydrophobic residues on the FG loop into the tail group region similarly to TIM4 (Fig. 3H). The x-ray reflectivity data thus allows us to distinguish multiple MD-refined equilibrium structures and, in doing so, make testable predictions by implicating residues that associate with the membrane.

3.3. Membrane-Bound Structures of TIM Proteins Reveal Basis of Sensitivity to PS Surface Density

While TIM4 and TIM1 bind the membrane in roughly the same angular orientation, they differ in their insertion depths and number of peripheral basic residues present at the interface, as shown in Fig. 4A and Fig. 4B, respectively. TIM4 and TIM1 both contain a tryptophan residue and a phenylalanine residue on the most inserted loop. The best fit reflectivity derived membrane-bound structures have these residues protruding into the tail region of the membrane for TIM4 (Fig. 3A) and TIM1 (Fig. 3F middle panel), suggesting that hydrophobic burial is a major component of the interaction. Previous studies that used ELISA showed that double alanine substitution of these residues in TIM1 and TIM4 almost entirely eliminated PS binding, reinforcing that the insertion of these residues is integral to the stability of the bound state [67]. In addition to the energy gained from the insertion of the hydrophobic residues, the burial of these residues likely enhances the affinity of the central binding pocket for PS. The anchoring of the hydrophobic residues allows the binding pocket to penetrate deeply enough into the membrane to effectively coordinate with PS and minimize the degree to which the bound PS is pulled out of the membrane into the solvent.

Figure 4.

The membrane-bound structures implicate peripheral residues that might confer sensitivity to PS surface density. A) TIM1’s membrane-bound structure shows insertion of the PHE and TRP residues into the tail group region and embedding of a single, peripheral ARG into the head group region of the membrane. B) The membrane-bound TIM-4 structure shows similar burial of these hydrophobic residues but also contains insertion of two ARG and two LYS residues into the headgroup region of the membrane. Red: ARG, Green: LYS, Gray: Hydrophobic, Yellow: Polar. C) Alanine mutants of the peripheral basic residues depicted in B result in diminished sigmoidicity of TIM4’s binding curve. D) The alanine mutants reduce the Hill coefficient of TIM4 substantially indicating these residues are important for sensing PS surface density. Figure adapted from Tietjen et al. (2014) [53] and Tietjen et al. (2017) [73].

The structures also show that TIM4 has a larger surface area burying into the membrane with two arginine and two lysine residues embedded in the head group region of the film. TIM1 has only a single arginine residue buried in the head group region and one arginine and one lysine close to the membrane, which are relatively more withdrawn than the equivalent residues in TIM4. These positively charged, peripheral residues presumably interact with adjacent PS to stabilize the bound state and account for the difference in PS surface density sensitivity between TIM1 and TIM4. Additionally, these differences between TIM1 and TIM4 support the proposed positive correlation of the extent of surface burial and the number of associated peripheral residues. For low PS surface densities, the increased surface burial of TIM4 incurs an energetic penalty and TIM4’s peripheral residues are less likely to be interacting with adjacent PS under these conditions. However, at high PS surface densities, TIM4’s peripheral residues are likely to be interacting with an adjacent PS and contributing more than enough energy to compensate for TIM4’s increased surface burial.

The membrane-bound structures implicate specific residues on TIM4 and TIM1 that could explain how TIM4 is much more sensitive to PS surface density than TIM1. Figures 4C and 4D show the binding curves of four TIM4 alanine substitution mutations at the two lysine and arginine residues described above. These mutants have reduced Hill coefficients compared to the wild type, confirming that these residues confer a sensitivity to PS surface density. The elimination of a single peripheral basic residue in TIM4 reduced the Hill coefficient from 2.4 to as low as 1.6. Presumably double mutants would have reduced Hill coefficients comparable to TIM1’s value of 1.35 while a quadruple mutant would be expected to exhibit single site binding [53]. Peripheral residues surrounding the binding sites of other PS-receptors may also confer a sensitivity to PS exposure distinct from multi-site cooperativity.

3.4. Physiological Relevance of PS Surface Density for TIM Proteins

The immunology literature for the TIM proteins paints an incomplete picture thus far, making it difficult to contextualize the exact regulatory role PS sensitivity plays. For example, the TIM genes were first identified from alleles of the TIM genes that enhanced experimental induction of airway hyperreactivity (AHR) [74], but a more recent study shows that induction of AHR in TIM gene knockout mice is relatively unchanged [75], Moreover, the TIMs have several other suspected binding partners besides PS [63], including potentially each other [76] and additional proteins. How these additional binding partners intersect with PS binding is unknown and potentially complex. The binding of PS is demonstrably important for the physiological function of the TIM proteins and in the case of TIM4 is clearly a signal for phagocytosis. However, it is unclear what TIM1 and TIM3 are “reading” from PS exposure, as the predominant cell types on which they are expressed do not phagocytose [66]. Perhaps binding of apoptotic cells by TIM1 or TIM3 acts as a signal for immune activation or suppression. However, their PS sensitivity suggests they do not distinguish high PS surface density from low PS surface density as much as TIM4 does, indicating that TIM1 and TIM3 do not necessarily have a preference for apoptotic membranes over other PS-presenting membranes. Further investigation of bulk membrane properties may elucidate the membrane specificity of TIM1 and TIM3 and their functional consequences.

4. Concluding Remarks

In this review we have summarized techniques and approaches that allow for characterization of peripheral membrane binding proteins and have used the TIM proteins to illustrate how the approach can help decipher their binding in the context of membrane properties. The combination of these techniques can be used to characterize the sensitivity of the TIM proteins to PS surface density more completely than any one technique could on its own. The identification of specific residues that confer membrane-contextual sensitivity clearly demonstrates the advantages of such a combination. The approach presented can easily be generalized to other lipid-binding proteins, as well as other membrane compositions. In addition to PS surface density, the role of fluidity, membrane charge density, and ionic strength can be investigated by varying membrane lipid compositions and buffer conditions. Such investigations have been carried out for the binding of pleckstrin homology (PH) protein domains to phosphoinositide lipids in the context of various lipid compositions [32]. Calcium concentration is another important parameter since calcium is required for many PS-receptors to bind PS and it also affects the membrane packing of charged lipids [77–80]. The sensitivity to PS surface density of TIM4 approximates a “switch,” whereby binding is largely increased above a threshold amount of PS exposure. These other membrane properties might similarly function as a set of “switches” to enhance or deter binding of the PS-receptors proteins to PS. The membranes that maximize the binding affinity of PS receptors likely resemble the cell membranes they bind in vivo. Our work with the TIM proteins provides a proof of concept that combinations of structural, thermodynamic, and computational approaches can be used to uncover the sensitivity of peripheral membrane-binding proteins to the membrane context.