In a previous study, we showed that tryptophan substitutions I201W and I422W at the dimerization interface of ClC-ec1 (Figure 1—figure supplement 1) yield a functionally folded, monomeric form of the transporter in lipid bilayers (Robertson et al., 2010). To set up the system for fluorescence studies, we moved a partially buried cysteine to a more accessible position, C85A/H234C that allows for quantitative labeling by Cy5-maleimide without impacting stability (Figure 1—figure supplement 2). For simplicity, we will refer to this single-exposed cysteine construct as WT, and tryptophan substitutions as W (I422W) and WW (I201W/I422W).

To measure the dimerization reaction of ClC-ec1 in lipid bilayers (Figure 1A), we reconstituted Cy5-labeled protein in 2:1 POPE/POPG lipids at different mole fractions (χ subunit/lipid) and freeze/thawed the proteoliposomes to produce ~10 µm diameter multilamellar vesicles (MLVs) (Pozo Navas et al., 2005). This creates a model of an infinite bilayer where subunits can exchange with one another, even at low dilutions (Figure 1B). After equilibration, the membranes are fractionated by extrusion (Figure 1C), forming small liposomes that are then imaged on a TIRF microscope (Figure 1D) for single-molecule photobleaching analysis (Figure 1—figure supplement 3D). With single-molecule sensitivity, we can count the number of subunits captured into each liposome (Figure 1E) and determine the probability distribution of fluorescent protein occupancy in liposomes (Figure 1F). This approach extends another single-vesicle fluorescence method that examines the behavior of membrane proteins in individual liposomes (Mathiasen et al., 2014). However, in our case, we are not studying the state of the protein in the final proteoliposome. For example, a spot that bleaches in two steps could represent a dimer or two independent monomers trapped in the same vesicle. Our approach ignores this ambiguity as it measures the probability that two fluorescent subunits were captured in the same vesicle, reporting the proximity of subunits at the point of extrusion of the large membranes. Therefore, the liposome extrusion step captures the monomer-dimer equilibrium in the prior MLV membrane state and ignores any changes in protein density or lipid composition that might arise during the extrusion process.

Figure 1 with 3 supplements see all

Download asset Open asset

Figure 1—source data 1

Excel file including data and statistical analysis presented in Figure 1 and Figure 1—figure supplements 1–3 including Ellman’s cysteine reactivity data, Cy5 labeling yields, Cl⁻ transport rates, functional F_0,Cl, subunit/lipid mole fraction quantification, Fraction of protein co-localized with liposomes and F₀, the fraction of unoccupied liposomes measured from co-localization microscopy.

https://doi.org/10.7554/eLife.17438.004

Download elife-17438-fig1-data1-v2.xlsx

Using two-color TIRF microscopy on ClC-ec1-Cy5 reconstituted into fluorescent liposomes (Figure 1—figure supplement 3), we find that nearly all Cy5 spots co-localize with liposomes, demonstrating the fidelity of reconstitution. We then counted photobleaching steps in all the imaged Cy5 spots. While the high signal and low background allows for counting of up to nine discrete steps, we found that single-, double- and ≧ three-step photobleaching probabilities, ${\displaystyle P_1}$, ${\displaystyle P_2}$ and ${\displaystyle P_{3+}}$ are sufficient to describe the monomer vs. dimer population in the large MLVs:

Under ideal experimental conditions, i.e. dilute conditions and 100% fluorescent labeling yield, a single or a double photobleaching step corresponds to a monomer or dimer respectively. In reality, labeling is imperfect, so and ${\displaystyle P_2}$ include additional states such as singly labeled dimers or two monomers, respectively. In addition, to quantify the dimerization reaction, one must examine the liposome occupancy as a function of increasing protein density in the membrane, a condition that increases the chance of randomly trapping independent subunits in the same liposome, whether associated as dimers or not. The probability for fluorescent subunit capture follows an apparent Poisson distribution that depends on: (i) the size distribution of the liposomes, (ii) the mole fraction subunit density in the lipid bilayer, (iii) Cy5 labeling yields and (iv) the monomer-dimer equilibrium in the membrane at the time of extrusion. The first three factors must be known and corrected for in order to properly extract information about the monomer-dimer reaction across a wide range of mole fraction densities.

To calculate this correction, we used the cryo-EM liposome size distribution reported by Walden et al. (Figure 2—figure supplement 1E) (Walden et al., 2007) and spectrophotometrically determined Cy5 labeling yields (Figure 2—figure supplement 1A) to simulate the capture of non-interacting monomer into the extruded liposome population, i.e. the ideal monomer photobleaching probability distribution P^M_n (Figure 2A), where n refers to the number of observed photobleaching steps. Measurement of the subunit/lipid mole fraction after freeze/thaw and extrusion show that the experimental mole fraction is 50% of the original reconstituted (Figure 1—figure supplement 2E,F). From here on, χ refers to the observed mole fraction, and it is this value that is considered in all of the simulations. At dilute conditions, i.e. densities less than ${\displaystyle \chi }$ = 2 × 10^–6 subunit/lipid (0.2 μg/mg reconstitution density, see Table 1), P^M₁ is constant and close to one, reflecting single subunit occupancy in the liposomes. The simulation also predicts a non-zero P^M₂ that arises from the small amount of non-specific labeling at sites other than the cysteine, resulting in the occasional double-labeled subunit. For ${\displaystyle \chi }$ > 2 × 10^–6 to 3.8 × 10^–4, P^M₁ decays to 0 as P^M₃₊ increases to 1, reflecting the increase in random co-encapsulation of monomers into the same liposome. This has been described before as 'artifactual togetherness' (Tanford and Reynolds, 1976; Fleming et al., 1997; Stanley and Fleming, 2005) and must be corrected for to extract the true dimerization reaction. Next, we simulated a population of non-interacting dimers to obtain the ideal dimer photobleaching probability distribution P^D_n (Figure 2B). P^D₁ and P^D₂ are comparable across the entire protein density range, arising from the fact that our experimental labeling yield is ~70%, leading to an equal probability of dimers labeled with one Cy5 vs. two Cy5 in the single-molecule range (Figure 2—figure supplement 1D). For ${\displaystyle \chi }$ > 7.5 × 10^–6, P^D₁ and P^D₂ decrease as the protein density increases reflecting the growing probability of liposomes with more than two dimers, described by P^D₃₊. To assess the dynamic range for this approach, we performed a chi-squared analysis of the monomer vs. dimer probability distributions at each mole fraction value. The distributions were statistically different (p ≦ 0.0001) for all mole fraction values except for the highest measured density, ${\displaystyle \chi }$ = 3.8 × 10⁻⁴ or 50 μg/mg (p > 0.05). Therefore, the photobleaching probability distributions can distinguish between the monomer vs. dimer state across 5 orders of magnitude: ${\displaystyle \chi }$ = 7.5 × 10^–10 to 7.5 × 10^–5.

Figure 2 with 1 supplement see all

Download asset Open asset

Figure 2—source data 1

Excel file including data and statistical analysis presented in Figure 2 and Figure 2—figure supplement 1 including the ideal monomer and ideal dimer photobleaching distributions (walden liposomes, P_Cy5 = 0.72, P_ns = 0.14, bias = 4) distribution, chi-squared analysis, pooled protein labeling data.

https://doi.org/10.7554/eLife.17438.009

Download elife-17438-fig2-data1-v2.xlsx

The single-molecule photobleaching method allows us to explore extremely dilute densities within the membrane with no loss of signal. At lower densities, the number of fluorescent spots in a field of view decreases, which can easily be compensated for by increasing the number of imaged fields to maintain similar counting statistics. We used this approach to investigate whether experimentally measured ${\displaystyle P_1}$, and ${\displaystyle P_{3+}}$ reflect the equilibrium population of ClC-ec1 monomers and dimers in the membrane. If the system is in dynamic equilibrium, then the fraction of dimers out of all subunits will depend on the mole fraction density, χ (subunit/lipid), but not the path followed to reach this density. We tested this by comparing the photobleaching data obtained using two distinct methods of setting the final mole fraction, one that starts with the protein in monomeric form, and another that starts with the protein as a dimer.

In the first method, monomeric W-Cy5 was reconstituted at ${\displaystyle \chi }$ = 7.5 × 10^–8, 7.5 × 10^–7 and 7.5 × 10^–6 subunit/lipid by dialysis (Figure 3A). Size exclusion chromatography of W in n-Decyl-β-D-Maltopyranoside (DM) micelles shows a mixture of monomers and dimers upon purification (Robertson et al., 2010); however, the protein rapidly dissociates to the monomeric form as revealed by an immediate re-run of the eluted protein (Figure 1—figure supplement 1D). Photobleaching analysis on extruded liposomes derived from these membranes (Figure 3C) shows that at the lowest density, the probability distribution resembles the ideal monomer distribution (Figure 2A) and as χ increases, the distributions approach the ideal dimer distribution (Figure 2B) with ${\displaystyle P_1}$ decreasing as ${\displaystyle P_2}$ and ${\displaystyle P_{3+}}$ increase.

Figure 3 with 1 supplement see all

Download asset Open asset

Figure 3—source data 1

Excel file including data and statistical analysis presented in Figure 3 and Figure 3—figure supplement 1 including WT and W dilution data, chi-squared analysis between reconstituted and diluted samples, and FRET data.

https://doi.org/10.7554/eLife.17438.013

Download elife-17438-fig3-data1-v2.xlsx

In the second method, we started with W-Cy5 already reconstituted in lipid bilayers at high density, ${\displaystyle \chi }$ = 7.5 × 10^–6 subunit/lipid (Figure 3B), a condition where W exists almost entirely as dimers. This state was confirmed by macroscopic FRET measurements in MLVs showing that W-Cy3 + W-Cy5 and WT-Cy3/Cy5 yield the same dimer FRET signal (Figure 3—figure supplement 1). The high-density W-Cy5 membranes were then diluted 100X and 10X by freeze/thaw-mediated fusion with empty vesicles. At 1 day after freeze/thaw, P_1-3+ (Figure 3D) shows no significant difference with the probability distributions obtained with W-Cy5 samples directly reconstituted at the corresponding ${\displaystyle \chi }$, 7.5 × 10^–8 and 7.5 × 10^–7 subunit/lipid, respectively (Figure 3C). Measurement of the 100X diluted photobleaching probabilities as a function of time after freeze/thaw shows no change up to 27 days at room temperature (Figure 3E). The agreement between the two methods demonstrates that the reaction is reversible and that equilibrium subunit exchange of W-Cy5 occurs during the freeze/thaw fusion step.

Since the equilibration kinetics may be different for different constructs, we conducted the same experiment for WT-Cy5. So far, WT has only been observed to exist as dimers in lipid bilayers, but all experiments conducted have examined the protein at relatively high mole fraction densities ${\displaystyle \chi }$ ≧ 7.5 × 10^–7 subunit/lipid (i.e. ρ ≧ 0.1 μg/mg). We reconstituted WT-Cy5 at ${\displaystyle \chi }$ = 7.5 × 10^–9 to 7.5 × 10^–6 subunit/lipid, up to 100-fold lower than the lowest density studied so far. At the lowest density, there is a significant ${\displaystyle P_1}$ probability (Figure 3F) indicating that WT exists as monomers in the lipid bilayer at low dilutions. Similar to the W-Cy5 data, as χ is increased, ${\displaystyle P_1}$ decreases while ${\displaystyle P_2}$ and ${\displaystyle P_{3+}}$ increase indicating a conversion from monomers to dimers. Next, high-density samples of WT-Cy5 reconstituted at ${\displaystyle \chi }$ = 7.5 × 10^–6 were diluted 1000X, 100X and 10X with empty vesicles by freeze/thaw-mediated fusion (Figure 3G). The diluted probability distributions show no significant difference compared to the reconstituted distributions except for the 1000X diluted sample corresponding to ${\displaystyle \chi }$ = 7.5 × 10^–9 subunit/lipid. Since these data were collected one day after the freeze/thaw process, we continued to image the 1000X diluted WT-Cy5 samples for up to 25 days after freeze/thaw, incubating the samples at room temperature. The photobleaching probabilities of the diluted sample slowly converged to the reconstituted ${\displaystyle \chi }$ = 7.5 × 10^–9 subunit/lipid probabilities with a mid-point of 13 days (Figure 3H). This demonstrates that the reconstituted WT samples reflect the protein population in a dynamic equilibrium, albeit with higher kinetic stability compared to W, a finding that has been observed with other membrane proteins such as Diacylglycerol kinase (Jefferson et al., 2013).

Next, we measured P_1-3+ for the three ClC-ec1 constructs reconstituted across a wide range of mole fraction densities: ${\displaystyle \chi }$ = 7.5 × 10^–10 to 3.8 × 10^–4 subunit/lipid. The experimental photobleaching distribution for WW-Cy5, P^WW_1-3+ (Figure 4A) resembles the ideal monomer distribution P^M_1-3+ (Figure 2A). For W-Cy5 (Figure 4B) and WT-Cy5 (Figure 4C), the experimental photobleaching distributions P^W_1-3+ and P^WT_1-3+ exhibit three phases: (i) at low densities they mimic the monomer distribution, (ii) as the density increases, there is a gradual approach to the ideal dimer distribution, and finally (iii) at higher densities, ${\displaystyle P_1}$ and ${\displaystyle P_2}$ decrease, while ${\displaystyle P_{3+}}$ increases indicating the capture of multiple subunits in each liposome. WT and W both demonstrate a monomer to dimer transition; however, the reaction of W is shifted along the χ subunit/lipid axis, indicating the weaker stability of the W dimer.

Figure 4 with 1 supplement see all

Download asset Open asset

Figure 4—source data 1

Excel file including data and statistical analysis presented in Figure 4 including raw and averaged data for WT, W and WW photobleaching distributions.

https://doi.org/10.7554/eLife.17438.016

Download elife-17438-fig4-data1-v2.xlsx

Using the data in Figure 4, we calculated the fraction of dimer (F_Dimer) across the range of experimental χ subunit/lipid densities. To take into account that the protein inserts into membranes in two orientations, but only similarly oriented protein participates in the dimerization reaction, we use the reactive mole fraction scale, ${\displaystyle {\chi }^{\ast }=\chi /2}$. For each value of ${\displaystyle {\chi }^{\ast }}$, we carry out a least-squares fit of the residual sum of squares (R²) of the experimental P_n probabilities to the F_Dimer weighted linear combination of the ideal monomer P^M_n and ideal dimer P^D_nprobability distributions from Figure 2 (Figure 5—figure supplement 1). The F_dimer vs. χ* data was fit to an equilibrium dimerization isotherm (Figure 5) to determine the mole fraction equilibrium constants and the mole fraction standard state free energies, using a standard state density of χ° = 1 subunit/lipid. The data for the single tryptophan mutant, W, shows a complete reaction from monomer to dimer, yielding Kχ^W$$ = 3.7 ± 1.6 × 10⁶ (best-fit ± standard error (SE)) and ΔG°_W = −9.0 ± 0.3 kcal/mole in 2:1 POPE/POPG lipid bilayers at room temperature. Note that as the density increases beyond ${\displaystyle {\chi }^{\ast }}$ = 1.9 × 10^–6, R² increases indicating that the experimental probability distributions deviate from P^D_n (Figure 5—figure supplement 2), due to the observation of more liposomes bleaching in 3+ steps than expected from theory (Figure 5—figure supplement 1E). It is expected that the theoretical calculations at high densities will be sensitive to inaccuracies in the liposome size distribution, especially if larger liposomes or multi-lamellarity have been omitted from the distribution. Another possibility is that there is a small amount of non-specific oligomerization at high mole fractions. To account for these deviations, the fits are weighted by 1/R², allowing us to define an observed dynamic range of = 7.5 × 10^–10 to 3.8 × 10^–6 subunit/lipid, or 0.0001 to 0.5 μg/mg ClC-ec1/lipid. While WW only shows the initial part of the reaction, this rise occurs within this dynamic range and is thus likely reflecting the onset of dimerization. Fitting the data to a dimerization isotherm yields Kχ^WW = 2.1 ± 0.8 × 10⁵ lipids/subunit and ΔG°_WW = −7.3 ± 0.2 kcal/mole. Fitting of WT data yields leads to Kχ^WT$$= 2.1 ± 0.5 × 10⁸ and ΔG°_WT = −11.4 ± 0.1 kcal/mole, although the baseline must be constrained (Y₀ = 0.07) since it is not possible to go to lower dilutions where the all-monomer state is expected to be observed in the membrane. With this, the results demonstrate that substitution at I422W (W) destabilizes ClC-ec1 dimerization by ΔΔG_W-WT = 2.4 ± 0.3 kcal/mole, while the additional tryptophan at I201W (WW) destabilizes the dimer by an additional ΔΔG_WW-W = 1.7 ± 0.4 kcal/mole, and ΔΔG_WW-WT = 4.1 ± 0.2 kcal/mole overall.

Figure 5 with 2 supplements see all

Download asset Open asset

Figure 5—source data 1

Excel file including data and statistical analysis presented in Figure 5 including raw and averaged F_Dimer data, minimum R² values, standard error of the estimate (SEE) and 1/R² weights on the fit.

Results from fitting to the dimerization isotherm are presented in Table 2.

https://doi.org/10.7554/eLife.17438.019

Download elife-17438-fig5-data1-v2.xlsx

All isoforms were inserted into a pASK vector containing a hexa-histidine tag at the C-terminus. Site-directed mutagenesis was carried out by QuickChange (Agilent, Santa Clara, CA) followed by DNA sequencing of the full gene. List of experimental constructs for site-specific labeling: C85A/H234C (WT) MW = 51,997 g/mole, ε = 46,020 M^-1 cm^-1; C85A/H234C/I422W (W) MW = 52,070 g/mole, ε = 51,700 M^-1 cm^-1; C85A/H234C/I201W/I422W (WW) MW = 52,146 g/mole, ε = 57,410 M^-1 cm^-1. List of constructs used to calculate the non-specific labeling: C85A (WT_non-specific) MW = 52,031 g/mole, ε = 45,900 M^-1 cm^-1; C85A/I422W (W_non-specific) MW = 52,104 g/mole, ε = 51,590 M^-1 cm^-1; and C85A/I201W/I422W (WW_non-specific) MW = 52,177 g/mole, ε = 57,280 M^-1 cm^-1. Molecular weight and extinction coefficients calculated using the Peptide Property Calculator at http://biotools/nubic.northwestern.edu/proteincalc.html.

Expression and purification of ClC-ec1 was carried out as previously described (Maduke et al., 1999; Robertson et al., 2010). BL21-AI E. coli competent cells (Thermo Fisher Scientific, Waltham, MA) were transformed with the plasmid and then 2 L Terrific Broth supplemented with ampicillin was inoculated and grown at 37°C. Protein expression was induced with anhydro-tetracycline at OD₆₀₀ = 1.0. After 3 hr of induction, cells were harvested, then lysed by sonication in buffer supplemented with 5 mM reducing agent TCEP (Tris(2-carboxyethyl)phosphine; Soltec Bioscience, Beverly, MA) and pH adjusted to 7.5. Protein extraction was carried out with 2% n-Decyl-β-D-Maltopyranoside (DM; Anatrace, Maumee OH) for 3 hr at room temperature. Cell debris was pelleted down and the supernatant was run on a 2 mL column volume (CV) TALON cobalt affinity resin (Clontech Laboratories, Mountain View, CA) equilibrated in CoWB/TCEP: 100 mM NaCl, 20 mM Tris, 1 mM TCEP, pH 7.5 with NaOH, 5 mM DM. After binding, the column was washed with 15 CVs of CoWB/TCEP followed by a low imidazole wash of CoWB/TCEP containing 20 mM imidazole (Sigma-Aldrich, St. Louis, MO). ClC-ec1 was eluted with CoWB/TCEP containing 400 mM imidazole, then concentrated in a 30 kDa NMWL centrifugal filters (Amicon, EMD Millipore) to ~500 μL and injected on a Superdex 200 10/30 GL size exclusion column (GE Healthcare, Little Chalfont, UK) equilibrated in size exclusion buffer (SEB): 150 mM NaCl, 20 mM MOPS pH 7.5, 5 mM analytical-grade DM, attached to a medium pressure chromatography system (NGC, Bio-Rad).

Wild-type ClC-ec1 contains three endogenous cysteines: C85, C302 and C347 (Figure 1 – supplementary 1B). While these cysteines can be mutated to yield a ‘cys-less’ form of ClC-ec1 (C85A/C302A/C347S) that maintain transport function (Nguitragool and Miller, 2007), we found that the ‘cys-less’ substitution on I201W/I422W expresses but results in aggregated protein upon purification. Examining the structure, C85 is partially accessible to the aqueous solution while C302 and C347 are buried within the protein core. We tested whether substituting C85 with alanine alone would be sufficient to minimize background labeling for our fluorescent experiments. We made a construct C85A/H234C, which introduces an aqueous solvent exposed cysteine near the dimerization interface, for specific labeling by Cy5-maleimide (Figure 1—figure supplement 1B). A cysteine accessibility assay was used to measure the reactivity of –SH groups present in ClC-ec1 (Ellman, 1959; Riddles et al., 1983). A 10 mM master stock of Ellman’s reagent (DNTB, 5,5’-Dithio-bis(2-nitrobenzoic acid); Sigma-Aldrich) was freshly prepared in reaction buffer (0.1 M sodium phosphate, 1 mM EDTA, pH 8.0) then diluted to 5 mM with SEB. Reaction of the thiolate anion with DTNB produces 2-nitro-5-thiobenzoate (TNB^-) that ionizes to TNB^2- and absorbs light at 412 nm. A₄₁₂ was monitored by UV-VIS spectroscopy (Nanodrop 2000c, Thermo-Fisher Scientific) for the protein at 10 μM (300 μL) for 5 min to establish a baseline. Following this, 20 μl of the Ellman’s reagent working stock was added ([protein] = 9.4 μM, [DNTB] = 313 μM), and the reaction monitored for ten minutes. To estimate the total number of cysteines in the protein, 40 μl of 2% SDS in SEB was added to denature the protein, exposing the buried cysteines C302 and C347 ([protein] = 8.3 μM, [DNTB] = 278 μM, 0.2% SDS), and the reaction was monitored for 45 min, until a steady saturation of A_{412 nm} was reached. Absorbance at 412 nm was background subtracted by the absorbance at 750 nm to correct for baseline drift during the measurement. An extinction coefficient of 14,150 M^-1 cm^-1 was used to calculate [TNB^2-]. Addition of Ellman’s reagent to C85A/H234C in 5 mM DM shows an instantaneous increase in A_412nm signal (Figure 1—figure supplement 1C) indicating rapid and specific conjugation with H234C. Addition of 0.2% SDS shows that the internal cysteines are reactive in the SDS denatured state. We calculated the molar ratio of TNB^2- produced per ClC-ec1 subunit in DM, and found that there is little reactivity in C85A. All constructs on the C85A/H234C background shows a single reactive thiol in 5 mM DM and a total of three reactive thiols in 5 mM DM + 0.2% SDS, indicating that introduction of tryptophan substitutions do no affect the fold of C85A/H234C in DM micelles (Figure 1—figure supplement 1D).

Cy5-maleimide dye was obtained as lyophilized powder as either 1 mg (GE Healthcare) or 50 mg (Lumiprobe, Hannover, Germany), stored as 10 mM master stocks (50 μl each) in anhydrous DMSO (Thermo Fisher Scientific) at -80°C. Single-use 5 μl working stocks of 10 mM strength were also prepared and stored at -80°C to avoid multiple freeze/thaw cycles of the fluorophores. Both master and working stocks were stored in boxes in the presence of anhydrous CaSO₄ (Drierite, W A Hammond Drierite Co Ltd., Xenia, OH). The fluorophore conjugation reaction was carried out in SEB with 10 μM ClC-ec1 subunits and 50 μM Cy5-maleimide for 12–15 min at room temperature in dark. At the end of the reaction, 100-fold molar excess of cysteine was added to quench the maleimide reaction (from freshly prepared 100 mM stock in SEB, pH adjusted to ~7.5). The ‘free’ dye was separated from the labeled protein by binding the reaction mixture to a 250 μL cobalt affinity resin column equilibrated with 15 CV CoWB (no TCEP) in a Micro-Bio spin chromatography column (Bio-Rad Laboratories, Hercules CA), washed 15 CV with CoWB and then eluted with 400 mM imidazole in CoWB, manually collecting only the fluorescently labeled protein. To remove the interfering absorbance of imidazole at 280 nm, the labeled protein was added to a 3 mL Sephadex G50 size exclusion column (Sigma-Aldrich) equilibrated in CoWB (no TCEP). The fluorescently labeled protein was eluted after addition 2–2.5 mL of CoWB to the column. The concentration and labeling efficiency of protein calculated from the UV-VIS absorbance spectrum of the sample and λ_max of ClC-ec1 (280 nm) and Cy5 (655 nm) as follows:

where ε_subunit (M^-1 cm^-1) is the molar extinction coefficient for the ClC-ec1 isoforms, ${\displaystyle A_{fluorophore}}$ is the absorbance of Cy3 or Cy5 at λ_max, ${\displaystyle C{F}_{fluorophore}}$ is the correction factor for fluorophore absorbance at 280 nm (CF_Cy3 = 0.08 CF_Cy5=0.02) and ${\displaystyle {ϵ}_{fluorophore}}$ is the molar extinction coefficients for the fluorophore (ε_Cy3 = 1.5 × 10⁵ M^-1 cm^-1 at 565 nm, and ε_Cy5 = 2.5 × 10⁵ M^-1 cm^-1 at 665 nm).

There was no significant difference in labeling yields of the C85A/H234C constructs (Figure 1—figure supplement 1G) allowing for the calculation of a pooled average labeling yield of P_Cy5 = 0.72 ± 0.08 (n = 10). For C85A constructs lacking the reactive H234C, we found that there was no significant difference in labeling and measured a pooled average of P_non-specific = 0.14 ± 0.04 (n = 11), representing the non-specific labeling yield. We tested the various C85A/H234C constructs in 5 mM DM for monomer vs. dimer behavior by size exclusion chromatography (Figure 1—figure supplement 1C) and found similar behavior compared to the tryptophan substitutions on the wild-type background (Robertson et al., 2010): C85A/H234C elutes as a dimer, C85A/H234C/I201W/I422W elutes as a monomer, while C85A/H234C/I422W yields a mixture of monomers and dimers upon initial purification, which quickly dissociates into monomers as shown by re-injecting the eluted protein on the size exclusion column. The profiles of un-labelled protein, and protein labeled with Cy5 do not show any significant difference (Figure 1—figure supplement 2).

Alexa-Fluor 488 (AF488) SDP ester (Thermo Fisher Scientific) was used to label the primary amine on POPE (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine; Avanti Polar Lipids, Alabaster, AL). AF488 was selected as it is spectrally removed from the protein labeling Cy5 channel, and it does not partition into membranes (Hughes et al., 2014). Dye stocks were prepared in the same manner described for the cyanine-maleimide stocks. Liposomes were prepared from a 2:1 mixture of POPE and POPG (1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-(1'-rac-glycerol); Avanti Polar Lipids) as a synthetic mimic of the major phospholipid composition of E. coli polar lipid extract. Briefly, a 2:1 mixture of POPE and POPG in chloroform (25 mg/mL) was dried under a continuous stream of N₂ gas, then resuspended in 0.5–1 mL pentane and dried again. Labeling buffer (LB): 300 mM KCl, 100 mM NaHCO₃ pH 8.3 was added to the lipids for a final concentration of 20 mg/mL (27 mM total lipids: 18 mM POPE, 9 mM POPG). The lipid solution was sonicated in a cylindrical bath sonicator (Avanti Polar Lipids) for 15 min until turbid, then 35 mM CHAPS (Sol-grade; Anatrace) was added and sonication continued until the solution was transparent (30–60 min). AF488 SDP ester was added to the lipid-CHAPS suspension at a final 0.3% mole fraction of total lipids, an amount that allows for visualization of all liposomes as measured by protein co-localization as a function of dye mole fraction. Roughly, the smallest liposomes in the population (r = 10 nm) are estimated to have five dyes assuming a labeling efficiency of 50%. The reaction proceeded at room temperature for 4 hr and then was stopped with 1.5 M Tris (pH 8.0). The lipids were stored at room temperature, in the dark, until fluorescently labeled protein was ready for reconstitution (1–3 hr), and then combined with the Cy5-labeled protein as described in the next section. AF488 fluorescence in the spent dialysis buffer was measured in the fluorometer at λ_ex = 485 nm, showing no detectable free fluorophore at the end of 48 hr.

For experiments that did not require fluorescent labeling, lipids were resuspended in either Reconstitution Buffer for functional or microscopy studies (RB-F): 300 mM KCl, 20 mM Citrate pH 4.5 with NaOH, or Reconstitution Buffer for FRET (RB-FRET): 150 mM NaCl, 20 mM Citrate, 10 mM MES, 20 mM Hepes, pH 7.0 with NaOH, with the change to NaCl allowing for the addition of SDS to samples. CHAPS (35 mM) solubilized lipids were combined with protein from 0.0001 to 50 μg ClC-ec1 per 1 mg of lipids, corresponding to χ = 7.5 × 10^–10 to 3.8 × 10^–4 protein/lipid mole fraction. The protein-lipid-detergent mixture was dialyzed in cassettes (NMWL 10 kDa; ThermoFisher Scientific) at 4°C against 4 L of the appropriate reconstitution buffer for 48 hr with buffer changes every 8–12 hr. After completion of dialysis, the proteoliposomes were harvested from the cassettes, freeze/thawed (see 'TIRF microscopy of proteoliposomes') then stored at room temperature, in the dark until further use.

Quantification of Cy5-labeled protein in proteoliposomes was performed by solubilizing 10 μL of vortexed sample into 190 μL of SEB supplemented with 35 mM CHAPS detergent (Anatrace) and 10% CTAB (Sigma-Aldrich). A standard curve was created alongside each set of samples by mixing serially diluted purified Cy5-labeled ClC-ec1 in the above buffer volume. Protein fluorescence was quantified in a 96-well plate using a Typhoon FLA 9500 Scanner (632nm laser/LPR emission filter permitting light >665nm). In general, the protein labeling yields show little variability between preps, allowing for the determination of amount of protein in each sample well compared to the standard curve.

Quantification of lipids was performed by ashing proteoliposome samples in 8.9 N Sulfuric Acid (Sigma-Aldrich) at 200–215°C for 25 min. Samples were then digested further with concentrated hydrogen peroxide (Sigma-Aldrich) for 30 min at 200–215°C. To each sample milliQ, 2.5% Ammonium Molybdate (Sigma-Aldrich), and 10% Ascorbic Acid (Macron, Center Valley, PA) was added. Color was developed by heating the samples at 100°C for 7 min. Samples were placed into a 96-well plate and the absorbance of each sample at 820 nm in a UV-VIS spectrophotometer plate reader. The samples are compared to a standard curve prepared with sodium phosphate dibasic (RPI, Mount Prospect, IL) along side each batch of measurements to determine the molar amount of phosphate (Fiske and Subbarow, 1925).

To measure Cl^- transport, un-labeled or fluorescent ClC-ec1 isoforms were reconstituted into liposomes in high chloride RB-F at 1 µg/mg, ${\displaystyle \chi }$ = 7.5 × 10^–6. These proteo-liposomes were freeze-thawed seven times, then extruded 21 times through a 0.4 µm polycarbonate membrane (Whatman Nuclepore Track-Etched Membranes). The external buffer was exchanged by passing 100 µL of the liposome sample through a 2.5 mL Sephadex G-50 size exclusion column equilibrated in low-chloride buffer (ExB): 150 mM K₂SO₄, 1 mM KCl, 20 mM Citrate at pH 4.5 with NaOH. This sets up a Cl⁻ gradient, however, efflux of Cl^- ions does not occur because of the opposing potential driving force. A potentiometer to measure Cl^- efflux was setup using silver chloride electrodes (Walden et al., 2007). Efflux was measured by the potentiometer in reference to 1 M KCl. For the recording, 1.8 mL of ExB was added to the measurement cell, followed by 15 µL of 10 mM KCl for calibration of the signal. Liposomes (~200 µL) in ExB were added to the measurement cell, and transport was initiated by addition of 1 µM K⁺ ionophore valinomycin and 2 µM of protonophore FCCP (Sigma-Aldrich). To normalize and compare Cl^- efflux across samples, total Cl^- concentration was measured by breaking the liposomes by adding 40 µl of 1.5 M ß-OG, releasing the remaining chloride trapped inside unoccupied liposomes. The traces were normalized for total Cl⁻, then fit to a two-component exponential relaxation function to determine k_ClC and F_Cl,0 (Walden et al., 2007). Leak from empty 2:1 POPE/POPG liposomes was measured to determine k_leak.

Förster Resonance Energy Transfer (FRET) was measured for upon W-Cy3 + W-Cy5 mixed during reconstitution or co-labeled WT-Cy3/Cy5 in the freeze/thawed MLV state. Fluorescence was measured using a fluorometer with double monochromators on excitation and emission sides to detect fluorescence in highly scattering MLVs (Fluorolog 3–22, Horiba Jobin Yvon, Edison, NJ). Fluorescence emission spectra were collected while exciting the donor (${\displaystyle {\lambda }_{Ex}}$ = 530 nm, ${\displaystyle {\lambda }_{Em}}$ = 540–850 nm, 2 nm slit width, 0.1 s integration time and averaged from 8 to 128 independent sequential scans) or acceptor Cy5 (${\displaystyle {\lambda }_{Ex}}$ = 640 nm, ${\displaystyle {\lambda }_{Em}}$ = 650–850 nm, 2 nm slit width, 0.5 s integration time). FRET was measured by correcting the donor emission spectrum for background fluorescence arising from the buffer and membrane, and for direct excitation of Cy5 when ${\displaystyle {\lambda }_{Ex}}$ = 530 nm. A correction factor for direction excitation of Cy5 was determined as 0.12 ± 0.01 (n = 3) by measuring Cy5-only samples excited at ${\displaystyle {\lambda }_{Ex}}$ = 530 nm vs. ${\displaystyle {\lambda }_{Ex}}$ = 640 nm. The reported FRET signal is the area normalization of the FRET-specific Cy5 emission, I^Cy5-FRET, over the total emission, I^Cy3 and I^Cy5-FRET:

The FRET signal was plotted as a function of increasing acceptor to donor ratio (P_Cy5/P_Cy3), fit to a single site binding curve, then normalized by the maximum FRET and fit to the oligomeric model function as described in Fung et al. (Fung et al., 2009):

where R = P_Cy5/P_Cy3 and n is the number of subunits in the oligomer.

A multi-wavelength single molecule total internal reflection fluorescence microscope was built following the CoSMoS design (Friedman et al., 2006; Larson et al., 2014). The microscope is equipped with 488 and 637 nm excitation lasers (OBIS, Coherent Inc., Santa Clara, CA) each with variable attenuators for control of laser power. The lasers were focused onto the back focal aperture of a high numerical aperture 60X objective lens (Olympus TIRF 60X 1.49 NA, Olympus Corporation, Tokyo, Japan) via a 3 mm diameter micro-mirror (MicroMirrorTIRF platform from Mad City Labs Inc., Madison WI). The illuminated area on the coverslip is ~2500 µm². Emitted fluorescence was filtered and focused onto the CCD chip of iXon3 888 EM/CCD camera (Andor Technology Ltd., Belfast, Nothern Ireland).

Glass slides (Gold-seal, 24 × 60 mm no.1.5 thickness) and coverslips (Gold-seal, 25 × 25 mm no. 1.0 thickness) were prepared as follows: slides and cover slips were placed in slide mailers, 5 at a time, then sonicated in presence of 0.1% Micro-90 detergent (Cole-Parmer, Vernon Hills, IL) for 30 min. After sonication, slides and coverslips were rinsed with MilliQ water 10 times to remove all traces of detergent, and then sonicated in presence of absolute ethanol for 30 min followed by rinsing 10 times with MilliQ water. Finally, glass coverslips and slides were sonicated in 0.2 M KOH for 5 min to render the surface hydrophilic. After final rinsing with water the coverslips and slides were stored in MilliQ water in a clean air hood for a period of 7–10 days.

Proteoliposomes harvested after dialysis were freeze-thawed seven times by incubating the samples in a dry ice/ethanol bath for 15 min, followed by incubation at room temperature for 15 min. At this step, the membranes fuse to form large multilamellar vesicles (MLVs) on the order of 10 µm in diameter (Pozo Navas et al., 2005), i.e. 1 × 10⁹ lipids per lamella. The samples were stored at room temperature with 0.02% NaN₃ until imaging, 1–95 days post freeze/thaw (see Figure 5—figure supplement 2 for specific time points of imaging). The formation of large membranes allows for the investigation of the protein state across a wide span of protein density ( = 7.5 × 10^–10 to 3.8 × 10^–4 subunits/lipid), with the lowest value representing roughly two subunits in a liposome lamella of ~300 µm² – i.e., close to infinite dilution. Prior to imaging, the membranes were extruded (LiposoFast-Basic; Avestin, Ottawa, Canada) 21 times through a 0.4 µm Nuclepore membrane to form liposomes for TIRF microscopy with a defined size distribution (Figure 2—figure supplement 1E). Between samples, the extruder apparatus was dismantled, flushed profusely with hot tap water, then sonicated in presence of dilute detergent (0.5% Dial dish soap) followed by sonication in deionized H₂O water to maintain extremely low background. After washing, buffer was passed through the extruder and loaded onto the microscope to confirm that there was no contamination between samples. Liposomes were diluted in low-adhesion tubes and allowed to passively bind to glass slide (Johnson et al., 2002). The density of fluorescent spots on the glass slide was maintained at 0.02 to 0.09 spots/µm² to minimize overloading (~50–200 spots per field). For experiments at low protein density, the slide was loaded with a high number of liposomes since the probability of protein occupied liposomes was rare. At higher mole fractions, liposomes were serially diluted before loading onto coverslips. Images were acquired at the rate of ~1 frame per second (fps), EM gain set to 300 and laser incident power typically set to 15 µW for AF488 imaging and 240 µW for Cy5 imaging in order to obtain long photo-bleaching traces while maintaining good signal to noise for single molecule spots. In each field, Cy5 imaging was carried out before AF488 imaging as we observed the 488 nm laser convert Cy5 to a dark state. After initial liposome-protein co-localization experiments to measure F₀ and reconstitution efficiency, samples for protein photobleaching experiments were typically reconstituted with regular POPE/POPG mixture without any fluorescent label on the lipids. All imaging was carried out in RB-F buffer, filtered three times through a 0.22 µm filter (Millex-GS MCE, Merck Millipore, Billerica, MA).

Cy5 blinking (Ha and Tinnefeld, 2012) was avoided by imaging the proteoliposomes in the absence of oxygen scavengers. Depletion of oxygen results in longer dwell time in the triplet state which often leads to fluorescent blinking. Triplet-state quenchers, such as Trolox, are often added to the imaging solutions that contain oxygen scavengers to reduce blinking behavior. However, Trolox has been shown to alter lipid bilayer properties by partitioning into membranes (Alejo et al., 2013). Instead, we found that avoiding oxygen scavengers altogether allowed us to observe long-lived photobleaching intensity traces of Cy5 without blinking behavior. In addition, the removal of the oxygen scavengers helps to reduce background contamination as we are working with non-passivated clean slides, and we have observed components of the glucose oxidase/catalase oxygen scavenger system demonstrate autofluorescence when bound to the slide. No filtering of images was necessary, as laser power was adjusted for optimal signal to noise while ensuring relatively long photobleaching traces. Photobleaching data was collected in both unlabeled and AF488-labeled liposomes, showing no significant difference in probabilities (Figure 4—figure supplement 1). Image files were analyzed in a MATLAB-based CoSMoS analysis program (Friedman and Gelles, 2015). Fluorescent spots were auto-detected based on intensity thresholds selecting 4 × 4 pixel areas of interest (AOIs) around the peak fluorescence. The AOIs were integrated over time; typically 300 frames were acquired at 1 frame per second (fps). Integrated intensity trajectories extracted from these AOIs were manually classified by counting the number of steps before complete photo-destruction of Cy5. To test for subjectivity of counting, 2–3 individuals analyzed the same data set in a blinded approach (Figure 4—figure supplement 1). For probabilities calculated from an individual population, we report fraction ± standard deviation (SD), calculated as the binary uncertainty ${\displaystyle \sqrt{p(1-p)N}/N}$, where p is the probability of a Cy5 spot bleaching in n steps and N is the total number of Cy5 spots in the sample, typically 200–500. In most cases, the probabilities are calculated over 2–6 samples and 2–3 manual counters, with values reported as mean ± standard error (SE) across both samples and counters.

The capture of ClC-ec1-Cy5 into extruded liposomes is a Poisson process (Maduke et al., 1999; Walden et al., 2007). Therefore, the probability of observing a certain number of Cy5 fluorophores, N_Cy5 = 1, 2, 3,... in a liposome is related to the Poisson distribution. However, under experimentally realistic conditions, the actual distribution will differ due to the following reasons: (1) extruded liposome sizes are heterogenous, (2) microscopy visualizes Cy5 and not the protein itself, so un-labeled subunits or subunits with bleached fluorophores are not counted, and (3) there are multiple labeling possibilities for monomers and dimers due to site-specific and non-specific labeling. As long as the liposome size distribution and fluorescent labeling yields are determined experimentally, the expected ${\displaystyle P_1}$ and ${\displaystyle P_2}$ photobleaching probabilities for an ideal monomer or dimer can be calculated. We created a MATLAB script (Mathworks, Natick, MA) that simulates the random process of subunit encapsulation given a certain number of subunits, ${\displaystyle N_{subunits}}$, and liposomes, ${\displaystyle N_{liposomes}}$, and the cryo-EM determined liposome size distribution, ${\displaystyle P_{radius}}$, for 0.4 µm extruded liposomes from freeze/thawed E. coli polar lipid membranes (composition ~2:1 POPE/POPG) (Walden et al., 2007). The experimental mole fraction, χ, is calculated using the following equation:

using a surface area per lipid, ${\displaystyle S{A}_{lipid}}$, equal to 0.6 nm² (Murzyn et al., 2005). The simulation follows by creating a matrix of ${\displaystyle N_{liposomes}}$ for each sub-population with radius r, and randomly inserting each protein species into these liposomes. In our case, we consider only the all-monomer and all-dimer condition, where ${\displaystyle N_{protein,M}=N_{subunits}}$ and ${\displaystyle N_{protein,D}=N_{subunits}/2}$. We now discuss how to calculate (i) ${\displaystyle P_{label,M}(N_{Cy5})}$ and ${\displaystyle Plabel{,}_D(NCy5)}$ – the distribution of the fluorescent labeling states for the monomer or dimer, (ii) ${\displaystyle N_{liposomes,M}(r)}$ and ${\displaystyle N_{liposomes,D}(r)}$ – the number of liposomes that are accessible to monomers or dimers, and (iii) ${\displaystyle N_{protein,M}(r)}$ and ${\displaystyle N_{protein,D}(r)}$ – the total number of monomers or dimers to be inserted into ${\displaystyle N_{liposomes,M}(r)}$ and .

A brief description of the derivation of the dimerization equilibrium isotherm is provided here (Wyman and Gill, 1990). The equilibrium dimerization reaction scheme can be written as follows for monomers, M, and dimers, D:

Membrane proteins react in a two-dimensional lipid bilayer and so the scale of the reaction coordinate must be appropriately selected. As discussed in the literature (White and Wimley, 1994; 1999; Fleming, 2002; Zhang and Lazaridis, 2006), the subunit/lipid mole fraction scale, χ, is a conventional choice for studying membrane protein association, as it can apply to reactions in both detergent micelles and lipids. For reactions in lipid bilayers, the area density scale, ρ_Area (subunit/nm²) is intuitive and often used (Hong et al., 2010). In the end, both scales are imperfect and require further corrections to account for the differences in the sizes of subunit and lipid solvent molecules (White and Wimley, 1994; Fleming, 2002). For consistency with the current literature, we report the equilibrium constant Kχ, dissociation constant Kd and standard state free energy ΔG° on the subunit/lipid mole fraction scale (Table 2), but also include the subunit/nm² the area density scale for reference. The area density scale is calculated from the mole fraction scale using the following equation:

where ${\displaystyle S{A}_{lipid}}$ is the surface area per lipid in nm². Assuming ideal dilute conditions, the mole fraction equilibrium constant for dimerization is:

where ${\displaystyle {\chi }^{\ast }}$ is the total reactive mole fraction, equivalent to ${\displaystyle \chi /2}$, that subunits are randomly incorporated into the membrane (Matulef and Maduke, 2005) and that the reaction only occurs between oriented subunits. ${\displaystyle {{\chi }^{\ast }}_D}$ is the dimer/lipid mole fraction, ${\displaystyle {{\chi }^{\ast }}_M}$ is the monomer/lipid mole fraction, and ${\displaystyle {K_{\chi }}^{\ast }}$ is the dimerization equilibrium constant in inverse mole fraction units (i.e. lipid/subunit) (Fleming, 2002). The total mole fraction of subunits in the membrane is:

The fraction of protein in the dimer state (${\displaystyle F_{Dimer}}$) is derived by substituting (26) into (25) to determine ${\displaystyle {\chi }_D}$:

To determine ${\displaystyle F_{Dimer}}$ from the photobleaching analysis, we used the theoretical calculations of the ideal monomer probabilities, ${\displaystyle {P^D}_{n=1-5+}}$, and ideal dimer probabilities, ${\displaystyle {P^D}_{n=1-5+}}$ as the respective all-monomer and all-dimer signals. Least-squares analysis was carried out on the sum of squared residuals (R²) between the experimental data, ${\displaystyle {P^{expt}}_n}$, and a linear combination of ${\displaystyle {P^M}_{n=1-5+}}$, and, ${\displaystyle {P^D}_{n=1-5+}}$, weighted by ${\displaystyle F_{Dimer}}$

The minimum R² value corresponds to the predicted F_Dimer for a given photobleaching distribution (Figure 5—figure supplement 1). ${\displaystyle F_{Dimer}}$ vs. ${\displaystyle {\chi }^{\ast }}$ was fit to the equilibrium dimerization isotherm above with the addition of a baseline offset ${\displaystyle Y_0}$:

Analysis of the R² as a function of mole fraction density shows that the quality of the fits of the experimental distributions to the theoretical distributions deviate for ${\displaystyle {\chi }^{\ast }}$ >1.9 × 10^–6subunits/lipid (Figure 5—figure supplement 2), which could arise from inaccuracies in the liposome size distribution by exclusion of larger liposomes or multilamellar liposomes, or a small proportion of non-specific oligomerization. To account for this uncertainty in ${\displaystyle F_{Dimer}}$ estimation at high densities, we weighted the fits by 1/R², essentially limiting the data to a dynamic range of ${\displaystyle {\chi }^{\ast }}$ = 3.8 × 10^–6 to 1.9 × 10^–6. The weighted non-linear fits were carried out using the non-linear regression function in MATLAB.

The mole fraction standard state free energy is calculated as:

where R is the gas constant (1.987 2036 cal/mole K), T is the temperature (~298 K) and χ° is the standard state mole fraction density 1 subunit/lipid. Note that this standard state is not physically relevant, but provides a normalization point for other measurements to be compared. On the area density scale, the standard state density ρ° = 1 subunit/nm² is used. For comparison, the typical standard state of 1 M on the molar scale is equivalent to 1 subunit/1.6 nm³ or 1 subunit/54 water molecules assuming a molecular volume of water of 30 ų (White and Wimley, 1999)

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

We are extremely grateful to Christopher Miller for constructive discussions throughout the project, as well as Nicholas Last, and the Miller and Gelles laboratories for feedback early on. We would also like to thank Laura Hughes, Bob Rawle and Steven Boxer for advice on fluorophores and introduction to the ‘splat’ approach, and to Benoît Roux for useful feedback about the manuscript. The research was supported by NIH/NIGMS grant R00GM101016 and a Roy J. Carver Trust Foundation Early Investigator Award.

1. Received: May 1, 2016
2. Accepted: July 29, 2016
3. Accepted Manuscript published: August 3, 2016 (version 1)
4. Version of Record published: September 2, 2016 (version 2)

© 2016, Chadda et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.