Using fluorescence correlation spectroscopy, we measured a dissociation constant of 20 nM between EGFP-labeled LcrV from
Yersinia pestis and its cognate membrane-bound protein YopB inserted into a lipid nanodisc. The combination of fluorescence correlation spectroscopy and nanodisc technologies provides a powerful approach to accurately measure binding constants of interactions between membrane bound and soluble proteins in solution. Straightforward sample preparation, acquisition, and analysis procedures make this combined technology attractive for accurately measuring binding kinetics for this important class of protein-protein interactions.Interactions involving membrane proteins are integral to a multitude of cellular processes, including signal transduction, energy production and conversion, cell adhesion, and foreign molecule identification. More than half of all pharmaceutical drugs target membrane proteins, further illustrating their importance in human health (
1). Due to this, there is a continuing high demand for methods that can screen, validate, and quantify interactions involving membrane proteins. Unfortunately, the quantitative techniques available to characterize protein-protein interactions are most often directed toward soluble proteins, and are often difficult to apply to membrane proteins (
2). Recently, advances in lipid nanodisc technologies, often referred to as reconstituted high-density lipoproteins, have enabled biophysical and biochemical studies of solubilized membrane proteins. The nativelike lipid environment of nanodiscs maintains protein functionality, opening a realm of possibilities in analyzing membrane protein function and dynamics in solution (
3).Nanodiscs are discoidal cell membrane mimetics that are 8–20 nm in diameter, consisting of a lipid bilayer stabilized by two peripheral apolipoprotein A-I proteins (
4). These particles provide an excellent alternative to traditional lipid-based platforms (e.g., liposomes) for membrane protein solubilization and interrogation. The diameter of nanodisc can be engineered to accommodate different-sized membrane proteins by varying lipid composition. As such, nanodiscs represent an important platform for expression, isolation, and study of functional membrane proteins and the multiprotein complexes they form. Several groups have successfully reconstituted a variety of membrane proteins, including bacteriorhodopsin (
5), G-protein coupled receptors (
6), and cytochrome P450 (
7) to name a few. To produce solubilized, discrete membrane proteins, we utilized a cell-free expression approach to embed membrane proteins directly into nanodiscs formed in situ. This approach allows for enhanced purification and rapid labeling of proteins of interest where traditional approaches are unsuccessful (
8). Fluorescence correlation spectroscopy (FCS) analysis coupled with nanodisc technology has been successfully used to monitor small ligand binding interactions with membrane proteins (
6,8) and to measure lipid-protein interactions at the single molecule level (
9). Here, we show that FCS can be used to measure interactions between soluble proteins and cognate membrane proteins inserted into nanodiscs produced using cell-free expression methodologies.FCS uses correlation analysis of fluorescence arising from randomly diffusing molecules to measure diffusion constants, and hence hydrodynamic radii, of species in solution. Diffusion times measured by FCS are inversely proportional to diffusion constants. A shift to slower diffusion time indicates an increase in the hydrodynamic radius due to binding. Due to this size sensitivity, FCS has been proposed repeatedly as a method to quantify molecular interactions in solution (
10). However, an important difficulty in FCS arises when characterizing interactions between two proteins of similar size. The diffusion time is inversely proportional to the cube root of the molecular mass
τD ∼ M
−1/3. Doubling the mass results in only a factor-of-1.3 increase in diffusion time, but a factor of 1.6 is required to resolve two species (
11). As we demonstrate, this difficulty is overcome by using FCS to measure the interactions between labeled soluble proteins and membrane proteins supported within a nanodisc. The much larger size of the membrane protein-nanodisc complex (
A) relative to most soluble proteins provides the necessary differential in diffusion times to easily resolve bound and free species by FCS.
Open in a separate window(
A) Model of YopB (
blue) inserted into a 10-nm nanodisc with cognate protein LcrV (
red) labeled with EGFP (
green). The molecular masses used in this model are: LcrV (35 kDa), EGFP (27 kDa), YopB (42 kDa, monomer), and nanodisc-YopB complex (258 kDa). The small LcrV binds to the much larger YopB-nanodisc complex, resulting in a significant shift in the autocorrelation curves to longer diffusion times. (
B) Hypothetical autocorrelation curves expected for a series of titration experiments measured using FCS. Increased binding is measured as an increase in diffusion time. For intermediate binding, diffusion components are detectable as illustrated (
red,
green, and
blue curves). To see this figure in color, go online.Binding kinetics are readily measured using FCS by titrating increasing amounts of unlabeled cognate protein, resulting in a series of autocorrelation curves similar to those modeled in a hypothetical example in
B. For 0% binding (
black curve), only the labeled soluble protein is present, producing a single component correlation curve with a diffusion time of 0.1 ms. On the autocorrelation curve with a single component, the diffusion time is the time at which the autocorrelation amplitude decreases by half. For 100% binding (
cyan curve), the amount of nanodisc with inserted membrane protein is well above the dissociation constant so that all available soluble protein is bound.Again, a single component correlation curve is found, but now with a longer diffusion time (10 ms in this hypothetical example). Intermediate cases have two components, corresponding to free and bound protein. The relative amplitudes of the components obtained by fitting to two component models in these cases can be used to obtain a binding curve. For binding interactions, the amount of ligand in the free and bound state can be separated by their diffusion time and quantified as a function of concentration. The fraction of bound protein is calculated by fitting the correlation to two components. The first component
f0(
τ) is the correlation of the free protein. The second component
g0(
τ) is the correlation when the protein is bound. The total correlation function is
h(
τ) =
αf0(
τ) +
βg0(
τ), (1)where
f0(
τ) = 1/(1 +
τ/
τf) and
g0(
τ) = 1/(1 +
τ/
τb). The values
α and
β are, respectively, the amplitudes of the correlation function for free and bound protein, with the corresponding diffusion times
τf and
τb. The ratio
F =
β/(
α +
β) gives the fraction bound.We investigated by FCS the interaction of two
Yersinia pestis proteins: soluble EGFP-labeled LcrV and membrane-bound YopB (
A). YopB is an effector protein involved in host cell invasion and disarming the cell’s defense. Although the structure of YopB is largely unknown, it is thought to exist as a dimer (
12). Delivery of YopB to the host cell is regulated by the Type III secretion system (T3SS) (i.e., injectisome). The needle tip of the injectisome contains the LcrV protein. When
Y. pestis comes in contact with a cell membrane, the injectisome forms a pore through the membrane to facilitate the diffusion of effector proteins into the host cell. Multiple lines of evidence suggest that pores are formed only through the direct interaction of YopB with LcrV (
13).The FCS diffusion times of free LcrV (
τf;
A,
black curve) and fully bound LcrV (
τb; [YopB] = 10
μM) were found to be 140 and 630
μs, respectively. Because no measurable change in diffusion time was detected above [YopB] = 1
μM (,
green curve), LcrV was assumed to be completely bound at [YopB] = 10
μM. The diffusion times of free and bound LcrV were determined by fitting the autocorrelation curves to a one-component model (
χ2 ∼ 1 for each diffusion time). These diffusion times correspond to average hydrodynamic radii of 2.5 nm for free LcrV and 11.1 nm for the LcrV + YopB-nanodisc complex, as calculated by the Einstein-Stokes equation. No significant change in diffusion time was detected when LcrV was incubated with 2
μM of nanodiscs lacking the YopB protein (
A,
gray curve), suggesting that LcrV alone is unable to interact directly with a lipid bilayer.
Open in a separate window(
A) FCS autocorrelation curves of 1 nM EGFP-labeled LcrV protein in the absence and presence of YopB at increasing concentrations. The diffusion time of freely diffusing LcrV is
τD,free = 140 s (
black curve). At [YopB] = 1
μM where LcrV is completely bound, the FCS curve is dominated by a single component (
green curve). The autocorrelation curves at intermediate concentrations (in particular [YopB] = 25 and 50 nM,
blue and
cyan curves) contain contributions from two diffusing components. A control experiment was performed with 2
μM of nanodisc lacking the YopB protein (
gray curve). No significant change in diffusion time was detected. (
B) The fraction of bound LcrV as a function of YopB concentration (log scale). This fraction is calculated by fitting the FCS curves shown in
A with
Eq. 1. The dissociation constant extracted is
Kd = 20.45 ± 2.0 nM. Error bars represent an average over six FCS measurements. To see this figure in color, go online.To extract the dissociation constant (
KD), a titration was performed over a range of YopB concentrations (10 pM to 1
μM), obtained by varying the concentration of YopB-nanodisc complexed. Having determined the values of the diffusion times for free (
τf) and bound (
τb) protein, we fitted the autocorrelation curves to a two-species correlation function with
Eq. 1. The brightness per LcrV molecule was between 2170 and 2280 photons/(s × molecule) for the entire titration, validating the use of
Eq. 1. The only varying parameters were the correlation amplitudes
α (free protein) and
β (bound protein). Intermediate binding was observed between 10 and 100 nM, because two correlation components are resolved. At [YopB] = 50 nM (
cyan curve), the ratio
β/
α was 1.5, indicating that more than half of LcrV was bound. For the two-component model in these intermediate cases, the
χ2 value was between 1 and 1.3, indicating a good fit. In contrast, fitting the data to a one-component model resulted in a poor fit (
χ2 > 7), justifying the use of the two-species model. Binding curves were fitted to the equation
y =
x/(
KD +
x), where
y is the fraction of bound LcrV,
x is the YopB concentration, and
Kd the dissociation constant.Fitting this model with a least-squares algorithm (
B), we extracted
Kd = 20.5 ± 2.0 nM. To quantify the statistical error, three measurements were recorded for 2 min, and the entire titration was repeated. Importantly, these data provide not only a quantitative binding affinity for the YopB-LcrV interaction, but support the hypothesis that LcrV requires direct interaction with YopB, not just a lipid bilayer, to promote pore formation.In conclusion, the coupling of FCS with nanodisc technology provides a facile yet powerful tool to quantitatively measure interactions involving membrane proteins in solution. FCS can simultaneously detect the presence of both free and bound species without the need for surface immobilization of the cognate proteins. Cell-free coexpression of both membrane protein and apolipoprotein components in the presence of lipids is a facile methodology for producing functional, soluble, nanodisc-supported membrane proteins. Labeling the soluble cognate proteins with EGFP provides a simple path to obtaining fluorescent, single-labeled proteins compatible with FCS. The combination of FCS with nanodisc technology presented here provides not only new key data for modeling the invasion process of
Y. pestis, but can also be generalized to study interactions between most other soluble and membrane proteins. Such methods have been lacking, yet are critical for understanding interaction networks, e.g., signal transduction cascades.See the
Supporting Material for additional detail on the methodology used.
相似文献