Engineering a Prototypic P-type ATPase Listeria monocytogenes Ca²⁺-ATPase 1 for Single-Molecule FRET Studies

Abstract

Approximately 30% of the ATP generated in the living cell is utilized by P-type ATPase primary active transporters to generate and maintain electrochemical gradients across biological membranes. P-type ATPases undergo large conformational changes during their functional cycle to couple ATP hydrolysis in the cytoplasmic domains to ion transport across the membrane. The Ca²⁺-ATPase from Listeria monocytogenes, LMCA1, was found to be a suitable model of P-type ATPases and was engineered to facilitate single-molecule FRET studies of transport-related structural changes. Mutational analyses of the endogenous cysteine residues in LMCA1 were performed to reduce background labeling without compromising activity. Pairs of cysteines were introduced into the optimized low-reactivity background, and labeled with maleimide derivatives of Cy3 and Cy5 resulting in site-specifically double-labeled protein with moderate activity. Ensemble and confocal single-molecule FRET studies revealed changes in FRET distribution related to structural changes during the transport cycle, consistent with those observed by X-ray crystallography for the sarco/endoplasmic reticulum Ca²⁺ ATPase (SERCA). Notably, the cytosolic headpiece of LMCA1 was found to be distinctly more compact in the E1 state than in the E2 state. Thus, the established experimental system should allow future real-time FRET studies of the structural dynamics of LMCA1 as a representative P-type ATPase.

Graphical Abstract

INTRODUCTION

Phosphorylation-type (P-type) ATPases undergo large-scale conformational changes when pumping typically cations across membranes to generate and maintain concentration gradients. The transport of cations against an electrochemical gradient is tightly coupled to ATP hydrolysis via formation and breakdown of a phosphoenzyme intermediate. All P-type ATPases contain a transmembrane (TM) domain linked to three cytosolic domains: the nucleotide-binding (N) domain, the actuator (A) domain, and the phosphorylation (P) domain.

The sarco/endoplasmic reticulum Ca²⁺-ATPase (SERCA), belonging to subclass P_IIA, is one of the best characterized P-type ATPases, from both a biochemical and a structural point of view. In animals, Ca²⁺ is released from the sarco/endoplasmic reticulum (SR) to induce, e.g., muscle contraction. Subsequent termination of an SR-induced Ca²⁺ signal such as in muscle relaxation requires the removal of Ca²⁺ from the cytosol, which is primarily done by re-sequestration to the SR by the action of SERCA.¹

In most bacteria, Ca²⁺-levels are maintained in the submicromolar range by a variety of secondary and primary transporters, including P-type ATPases.^2,3 In the pathogenic bacteria Listeria monocytogenes and Streptococcus pneumoniae, as well as in fungal pathogens, P-type ATPases are associated with virulence and survival at high extracellular Ca²⁺ concentrations present in infected host cells.^4–6 Recently, a L. monocytogenes P-type ATPase, LMCA1, was identified and characterized.^7,8 LMCA1 shows 38% sequence identity with SERCA and differs from its eukaryotic counterpart by displaying a lower Ca²⁺ affinity and transporting only one Ca²⁺ ion and one H⁺ counterion per cycle. Furthermore, LMCA1 exhibits a higher pH optimum and is up-regulated at the transcriptional level upon exposure to alkaline pH.⁹

So far, only a preliminary structural analysis has been performed for LMCA1 in the Ca²⁺-free state stabilized by AlF₄⁻, representing an occluded E2-P* intermediate state of dephosphorylation with a fold similar to that observed for SERCA under identical conditions.¹⁰ In contrast, SERCA has been captured in several conformations along its functional cycle and subjected to structural characterization by X-ray crystallography.^1,11–13 Also, other P-type ATPases have been analyzed^14–16 and show a similar structural architecture despite low overall sequence similarity.

The majority of P-type ATPases, including Ca²⁺-ATPases, possesses ten TM helices. In SERCA, two Ca²⁺ binding sites (I and II) are located between helices M4, M5, M6, and M8,¹¹ while only LMCA1 site II is conserved and functional.⁷ The TM domain is connected to the cytoplasmic domains (A, N, and P) via extended helices and linkers, which allow the coupling of conformational changes in the cytoplasmic domains to the actual transport of the ions in the TM domain.

The structural conservation of P-type ATPases suggests a common reaction mechanism based on the alteration between two major conformational regimes, namely, the E1 and E2 states. In the E1 state, the TM domain of the pump exhibits high affinity for the primary substrate (i.e., Ca²⁺ for LMCA1 and SERCA). Following Ca²⁺ binding, a series of conformational changes lead to the occlusion of the cytosolic ion pathway as well as phosphorylation of a conserved Asp residue in the P domain via transfer of the γ-phosphate of ATP present at the interface with the N domain. This leads to a conformational change resulting in the phosphorylated E2 state (E2P) now with an outward-open pathway of the TM domain, where the bound Ca²⁺ ion(s) are exchanged for H⁺ counterion(s). Dephosphorylation is stimulated by the TGES motif of the A domain, which undergoes large movements during the phosphorylation/dephosphorylation cycle to actuate the ion transport through the TM domain.¹³

X-ray crystallography in combination with molecular dynamics simulations and biochemical data have provided a model of the conformational changes accompanying the functional cycle of SERCA.^1,17 However, crystallography favors compact, well-ordered structures, while transient, functionally important conformations may be overlooked. In addition, some crystallized states may be influenced by crystallization conditions (e.g., detergents, lipids, additives, inhibitors). Thus, complementary high-resolution methods are needed to elucidate the dynamics of P-type ATPases during their functional cycle. Previous studies of SERCA tagged with GFP variants on the intracellular domains have shown that FRET can probe the structural changes accompanying the functional cycle.^18–20 However, due to the large size of fluorescent proteins and their relatively poor photostability, it is desirable to use small, extrinsic organic fluorophores, which have recently undergone a great leap in stability and brightness.²¹ These enable the detection of dynamics at the single-molecule level with high resolution in time and space.²²

In this study, LMCA1 was biochemically validated to be a prototypic P-type ATPase that can be engineered to allow single-molecule studies of the dynamics during functional cycling. An LMCA1 variant optimized for maleimide labeling was developed, and pairs of cysteines were introduced to permit labeling, while maintaining functional activity and reducing background labeling. Ensemble and confocal single-molecule FRET studies enabled the observation of distinct FRET states related to structurally well-characterized conformations consistent with those observed for SERCA. Interestingly, the cytosolic headpiece of LMCA1 was found to become more compact upon Ca²⁺ binding.

RESULTS AND DISCUSSION

Validation of LMCA1 as a Prototypic P-type ATPase

LMCA1 has fewer cysteines than its mammalian P-type ATPase homologues and therefore presents a more accessible target for labeling with thiol-reactive fluorophores. Before embarking on the labeling of LMCA1, we validated LMCA1 as an appropriate model system to study general features of P-type ATPases. Traditionally, fluoride analogues of phosphate, such as BeF_x, AlF_x, and MgF_x, have been used as general inhibitors of P-type ATPases that trap the pumps in conformational states resembling enzyme phosphoforms (EP). These inhibitors have proven very useful in structural studies of SERCA and other P-type ATPases by stabilizing analogues of the E2·Pi product state (E2·MgF₄²⁻),²³ the E2-P* transition state (E2·AlF₄⁻),^13,24 and the E2P ground state (E2·BeF₃⁻).^13,24 Here, the capacities of these complexes as inhibitors of LMCA1 were characterized in ATPase assays.

NaF was found to inhibit LMCA1 with a high IC₅₀ value of 2.3 mM (Figure 1A). This inhibitory effect is arguably caused by formation of a MgF_x complex from NaF and 1 mM MgCl₂ present in the buffer to sustain LMCA1 activity. The Hill coefficient was equal to −2.4 (Table 1), i.e., almost identical to the value reported for Na,K-ATPase under similar conditions.²⁵

Figure 1.

Inhibition of LMCA1 activity by metal fluoride complexes was assessed in ATPase assays. A. Inhibition of LMCA1 by NaF in the presence of 1 mM MgCl₂. B. Inhibition of LMCA1 by BeSO₄ at a constant NaF concentration of 0.5 mM (solid symbols), or in the absence of NaF (open symbols). C. Inhibition of LMCA1 by AlCl₃ at a constant NaF concentration of 0.5 mM (solid symbols), or in the absence of NaF (open symbols). Error bars represent standard deviation of three measurements. The data were fitted by a Hill equation (solid and dashed lines). The resulting Hill coefficients and IC₅₀ values are listed in Table 1.

Table 1.

Inhibition Constants for Metal Fluorides

	IC50 [µM]	Hill coefficient
NaF (1 mM MgCl2)	2300 ± 200	−2.4 ± 0.4
BeSO4 (0.5 mM NaF)	1.7 ± 0.2	−1.1 ± 0.1
BeSO4 (H2O)	27 ± 5	−0.72 ± 0.08
AlCl3 (0.5 mM NaF)	6.9 ± 0.8	−1.4 ± 0.2
AlCl3 (H2O)	77 ± 4	−1.7 ± 0.2

BeF_x and AlF_x complexes were obtained by mixing 0.5 mM NaF (a concentration where no inhibition due to formation of MgF_x has yet been observed) with BeSO₄ or AlCl₃, respectively. Both BeF_x and AlF_x complexes inhibited LMCA1 more efficiently than MgF_x. BeF_x was the most potent inhibitor, with an IC₅₀ value of 1.7 μM (Figure 1B), compared to an IC₅₀ value of 6.9 μM in the case of AlF_x (Figure 1C), in the presence of 0.5 mM NaF. In the absence of NaF, BeSO₄ and AlCl₃ were still capable of inhibiting LMCA1, albeit with higher IC₅₀ values of 27 and 77 μM, respectively (Table 1). These data are in agreement with earlier studies of the Na,K-ATPase showing an inhibition of the ATPase by chloride complexes of beryllium or aluminum in the micromolar range.^26,27 The two Hill coefficients obtained upon inhibition by BeSO₄ were approximately −1, suggesting the binding of one metal ion at the phosphorylation site, while inhibition by AlCl₃ resulted in Hill coefficients of approximately −1.5. In a structure of SERCA crystallized in the E2-P* transition state stabilized by AlF₄⁻,^13,24 one Mg²⁺ ion binds at the phosphorylation site and coordinates the fluoride. It is likely that at high AlCl₃ concentrations, the second aluminum ion may bind and replace the Mg²⁺ ion, as has also been proposed for the Na,K-ATPase.²⁵

The effects of AlF_x and BeF_x on LMCA1 activity were found to be similar to their effects on SERCA, and thus, it is expected that the same specific structural states can be trapped with these metal fluorides for LMCA1. Therefore, we considered it sensible to apply LMCA1 as a model system in FRET studies of P-type ATPases.

Rational Design of a Cysteine-Deprived LMCA1 Mutant

LMCA1 contains five native cysteines, of which two are located in the N domain (C380 and C458), one in the P domain (C332), and two in the transmembrane region (C251 and C827) (Figure 2A). Estimates of the distance changes of these cysteines during the functional cycle based on equivalent positions of SERCA revealed only minor changes between the different conformational states. Furthermore, the intrinsic cysteines did not seem to be completely solvent-exposed. Hence, none of these cysteines were suited for labeling for FRET studies. Instead, they were mutated to other amino acids to avoid their contribution to background labeling. Analysis of the conservation of the five intrinsic cysteines based on a multiple sequence alignment of over 1000 bacterial P_IIA ATPases (Figure 2B) indicated the possibility of introducing the mutations C251A, C332A, C380A, C458S, and C827S, corresponding to positions C268, C349, C420, C525, and S940 in rabbit SERCA1a, with only minor effects on activity.

Figure 2.

Intrinsic cysteines of LMCA1. A. Homology model of LMCA1 highlighting the five intrinsic cysteines represented as black spheres. The N, P, and A domains are shown as red, blue, and yellow cartoons, respectively. Transmembrane helices M1-M2, M3-M4, and M5-M10 are shown as pink, wheat, and cyan cartoons, respectively. B. Conservation of intrinsic cysteines calculated as the percentage of amino acids occurring in a multiple sequence alignment at the position corresponding to the cysteines in the sequence of LMCA1. A dash (–) corresponds to an amino acid deletion.

The previously described (His)₆-tagged LMCA1 construct allowing expression in E. coli and subsequent purification by a two-step immobilized metal affinity chromatography (IMAC)⁷ was improved by the addition of four extra histidines. This enabled the use of higher imidazole concentrations during binding and washing steps of the purification, resulting in significantly improved purity after a single-step IMAC (Figure S1), as has been observed for other targets.²⁸

Single-point mutants of the cytoplasmic cysteines C332A, C380A, and C458S were purified (Figure 3A), and their ATPase activities were compared to that of wild-type LMCA1 (LMCA1_WT). All of the tested LMCA1 variants showed a linear, Ca²⁺-dependent production of inorganic phosphate (Pi) within the time course of the experiment (Figure 3B). Mutation of the cysteines located in the N domain (C380A and C458S) only had minor effects on the activity of LMCA1, while mutation of C332 resulted in a 50% decrease in specific activity, from 3.6 to 1.7 μmol/(min mg) (Figure 3C), corresponding to a decrease from approximately an average of 6 to 3 turnovers per second. This activity is lower than previously reported by Faxen et al.,⁷ primarily due to the lower concentration of Mg²⁺ and ATP applied in the present study. C332 is located in the P domain and is conserved in over 80% of all P_IIA ATPases. In canine SERCA2a, mutation of the corresponding cysteine (C349) to alanine leads to a 50% loss of activity²⁹ in agreement with the results obtained here for LMCA1. The importance of this cysteine is arguably caused by its proximity to the catalytic site D334 (D351 in rabbit SERCA1a), a part of the P-type ATPase hallmark DKTG phosphorylation motif. Since the C380A and C458S mutations did not considerably decrease the activity of LMCA1, these mutations were included in all subsequent constructs to reduce background labeling.

Figure 3.

Effects of cytoplasmic cysteine mutations on the activity of LMCA1. A. SDS-PAGE analysis of purified LMCA1 variants mutated at the indicated positions. The band corresponding to LMCA1 is marked with a red arrow. B. Time-course ATPase activity measurements of LMCA1 mutants. Open symbols at 15 min represent a control measurement without Ca²⁺. Duplicates were made for each condition tested. The points represent mean values and the error bars represent standard deviation. The linear part of every curve was subjected to linear regression (black lines) to provide a measure of activity. C. Specific ATPase activity of the LMCA1 mutants as obtained by linear regression of the time course in B. Error bars correspond to 95% confidence intervals of the fit.

Next, four mutants harboring different combinations of intrinsic cysteines, henceforth referred to as “cysteine backgrounds”, were designed: a mutant completely devoid of intrinsic cysteines denoted LMCA1_NC, a mutant harboring only C332 denoted LMCA1_C332, a mutant carrying only the two transmembrane cysteines, C251 and C827, denoted LMCA1_TM and a mutant containing three cysteines, C332 and the two transmembrane cysteines, denoted LMCA1_3C (Figure 4A). For all four mutants, ATPase activities (Figure 4B) and background labeling (Figure 4C) were compared to those of LMCA1_WT. The removal of all intrinsic cysteines in LMCA1_NC caused a severe loss of activity, down to 15% of LMCA1_WT. As expected, LMCA1_NC showed a very low background labeling, ~4-fold lower than LMCA1_WT. The reintroduction of the most conserved cysteine in LMCA1_C332 increased the activity to ~20% of LMCA1_WT, while the background labeling of LMCA1_C332 was similar to LMCA1_WT, suggesting that C332 is by far the most reactive native cysteine. The background labeling of LMCA1_TM was similar to that of LMCA1_NC, while the ATPase activity corresponded to ~50% of LMCA1_WT. In LMCA1_3C, C322 was reintroduced into LMCA1_TM leading to an expected increase in activity (~70% of LMCA1_WT). However, the background labeling of LMCA1_3C was as high as LMCA1_WT.

Figure 4.

Characterization of LMCA1 cysteine backgrounds. A. Homology models illustrating the different LMCA1 cysteine backgrounds. Cysteines are represented as black spheres. LMCA1 coloring and representation is analogous to Figure 2A. B. Comparison of ATPase activity of LMCA1 cysteine backgrounds. The relative activity is presented as the percentage of the wild-type (WT) activity. Error bars correspond to standard deviations of two (LMCA1_C332 and LMCA1_3C) or three (LMCA1_NC and LMCA1_TM) biological replicates. C. Comparison of maleimide-based background labeling of LMCA1 cysteine variants relative to LMCA1_WT. LMCA1_NC and LMCA1_TM are average values of two, and LMCA1_WT of three biological replicates. Absolute labeling efficiency values varied between replicas (approximately 2–8% for LMCA1_WT, depending on the conditions), but the ratios of mutant-to-LMCA1_WT were constant. Error bars show standard deviation of labeling with Cy3 and Cy5 dyes.

In an attempt to improve the activity of LMCA1_TM, alternative substitutions at position 332 were endeavored: C332L, C332S, and C332D. Leucine was chosen as the third most common amino acid at position 332 (Figure 2B), serine was a structurally conservative mutation, while aspartate was chosen to mimic the negatively charged thiolate ion of cysteine, which reacts with maleimide. The high labeling efficiency of C332 indicated that the latter strategy might be fruitful. Disappointingly, all mutants showed a very low ATPase activity (Figure S2).

Thus, the LMCA1_TM mutant containing an alanine at position 332 displayed the best compromise between low background labeling (20% of LMCA1_WT) and high activity (~50% of LMCA1_WT) and was chosen for the introduction of pairs of cysteines for labeling.

Design of Cysteine Labeling Sites in LMCA1 and Optimization of Labeling

An automated script was written and used to identify residues in LMCA1_TM homology models (made in MODELLER³⁰ based on structures of SERCA) appropriate for the introduction of cysteine pairs for labeling (Supporting Information). Pairs of residues in the cytoplasmic A, P, and N domains showing the greatest distance changes during the conformational cycle were identified (Figure 5A). In addition, ideal pairs should not be conserved (Figure 5B) and should be solvent exposed. Reactivity is highly variable even between fully surface exposed cysteines,³¹ and ideally this should be taken into account. However, due to the large uncertainty in cysteine pK_a predictions,³² we chose to approach reactivity empirically. Based on these rationales, two combinations of labeling sites were identified (Figure 5). One cysteine was introduced into the A domain (at residue 18 or 24), which undergoes the largest relative movements during the functional cycle. The second cysteine was introduced into either the N domain at position 413 or into the P domain at position 530. The resulting mutants were denoted LMCA1_TM-A/N (labeled at positions 18 and 413) and LMCA1_TM-A/P (labeled at positions 24 and 530). Since both the A and N domains show significant dynamics, the first mutant was expected to be a very potent FRET reporter, but might at the same time be more difficult to interpret, as the observed distance changes result from the movements of two mobile domains. In the second mutant, the measured distance change mainly reflects the rotation of the A domain during the functional cycle.

Figure 5.

Labeling sites of LMCA1. A. Homology models of two LMCA1 variants designed for labeling, denoted LMCA1_TM-A/N and LMCA1_TM-A/P, are shown as two rows, each displaying six different functional states (based on the following SERCA PDB entries from left to right: 1SU4, 4H1W, 1T5T, 3B9B, 3B9R, 2C88). Each of them is labeled with the corresponding Post-Albers state and additives used for crystallization (SLN – sarcolipin, TG – thapsigargin) on top of the picture. Three E1 states are shown on a blue background, whereas three E2 states are shown on a gray one. Amino acids constituting the labeling sites are shown as colored spheres, and are additionally labeled with amino acid name, number, and the domain it belongs to in every leftmost homology model. The distances between labeling sites are shown as black lines with distances in Å shown above. All labeling sites are shown in the TM cysteine background, hence cysteines C251 and C827 are shown as black spheres. Coloring is analogous to Figure 2A. B. Conservation of labeling sites calculated analogously to Figure 2B.

Dual labeling with maleimide-derivatives of Cy3 and Cy5 was optimized with respect to reaction time and molar excess of dye over protein (Figure S3). Initially, 12 and 15 times molar excess of Cy3 and Cy5, respectively, was applied for the mutant LMCA1_WT-A/P during 15, 30, or 60 min. Unreacted dye was removed on a desalting column and the first fraction containing pure LMCA1 was used for determination of labeling efficiency and ATPase activity. The labeling efficiency remained constant at approximately 50% during the chosen time span (Figure S3A). Under these conditions, the mean background labeling of LMCA1_WT was ~8%, while the background labeling of LMCA1_NC was only ~2%. The effect of introducing labeling sites and labeling on the ATPase activity of LMCA1_WT-A/P was found to be minimal (Figure S3B).

Subsequently, the dye-to-protein ratio was optimized to increase the labeling efficiency. Labeling of LMCA1_NC-A/P was evaluated at three different dye-to-protein ratios: 4/5, 8/10, and 16/20 (molecular excess of Cy3/Cy5, respectively, over protein). Cy3 and Cy5 were always used in a 4:5 ratio to decrease the fraction of LMCA1 molecules double-labeled with Cy3, causing zero-FRET background peaks. The labeling was performed during 60 min to ensure that time was not a limiting factor at lower dye-to-protein ratios. The labeling efficiency increased steadily to approximately 50% at 16/20-fold excess without reaching a plateau (Figure S3C). Previous studies have reached much higher degrees of double labeling for soluble proteins.^33,34 The modest yield and the need for a large excess of labeling reagent stems from the relatively low protein concentration (10 μM), where bimolecular maleimide–thiol reactions compete with the unimolecular inactivation of the reactive maleimide. Since higher dye and protein concentrations will favor bimolecular labeling reactions over unimolecular reactions, it is thus likely that even higher concentrations would lead to a higher labeling efficiency. However, a labeling efficiency of 50% was judged to be sufficient for single-molecule studies, and therefore higher dye or protein concentrations were not pursued to avoid undesirable effects from high concentrations of DMSO used to solvate the dyes or protein aggregation. The level of background labeling of both LMCA1_WT and LMCA1_NC was low, and did not increase at higher dye-to-protein ratios for LMCA1_NC. Therefore, all subsequent labeling reactions were performed for 60 min at a 16/20 molecular excess of Cy3/Cy5 dyes over LMCA1.

When size exclusion chromatography (SEC) was employed to remove unbound dyes after labeling, labeled LMCA1 was separated into a monomeric fraction and a fraction containing oligomers and aggregates. LMCA1 is present in both peaks, but only the monomeric peak demonstrates Ca²⁺-dependent activity (Figure S4). In addition, differential background labeling of the monomeric and oligomeric peak was observed, especially in LMCA1_WT and LMCA1_TM, in which the monomeric fraction was essentially nonlabeled, as opposed to the oligomeric one. Thus, SEC was introduced as an additional purification step after labeling to isolate the monomeric fraction of labeled LMCA1 and to remove unreacted dyes.

Applying the optimized labeling strategy, LMCA1_TM-A/N and LMCA1_TM-A/P were labeled to approximately 50% (Figure 6A). The labeling efficiencies of the negative controls, LMCA1_WT and LMCA1_TM, were ~2% and ~0.5%, respectively, indicating a lower background labeling of the monomeric fraction as compared to the background labeling without the SEC purification step. No significant effects on ATPase activity of introducing cysteines for labeling were detected (Figure 6B). The activities after the labeling were also measured and found to be unaffected for LMCA1_WT, LMCA1_TM and LMCA1_TM-A/N, whereas the activity of LMCA1_TM-A/P was reduced by 50% after labeling (Figure 6B). This was surprising given that no decrease in activity was observed after the labeling of LMCA1_WT-A/P (Figure S3B), suggesting that it was not necessarily the labeling of the cysteines at positions 24 and 530 (A/P) that had a marked effect on the activity.

Figure 6.

Maleimide-based labeling and activity of LMCA1_TM-A/N and LMCA1_TM-A/P. A. Labeling efficiency after 60 min of labeling with 16× molar excess of Cy3 and 20× of Cy5, calculated as a corrected ratio of Cy3 or Cy5 absorbance to LMCA1 absorbance. Error bars correspond to standard deviation of two independent labeling reactions. B. ATPase activity of the indicated LMCA1 variants before and after a 60 min labeling reaction. Time course measurements performed in duplicate were subjected to linear regression to provide a measure of activity (slope of the line). Error bars correspond to 95% confidence intervals of the fit.

Fluorescence Characteristics of Labeled LMCA1_TM Mutants

In order to determine intramolecular distances from FRET efficiencies, fluorophores must be freely rotating on the time scale of imaging (κ² = 2/3) and their quantum yield must remain relatively constant over the imaging period.³⁵ Thus, fluorescence anisotropy (r) measurements were conducted to evaluate the extent of rotational freedom of the donor fluorophore attached to LMCA1. The anisotropy was assessed individually at both sites of labeling in the Cy3-labeled LMCA1_TM-A/N and LMCA1_TM-A/P mutants under four buffer conditions including saturating concentrations of different ligands intended for use in subsequent FRET experiments (Figure 7A). The measured anisotropy values (r ~ 0.30) were only slightly higher than anisotropy values of the free Cy3 (r ~ 0.25). The high anisotropy of free Cy3 is in good agreement with previously reported anisotropy values for carbocyanines³⁶ and consistent with the short Cy3 fluorescence lifetime.³⁷ Importantly, no changes in anisotropy were observed in the presence of different ligands. In both pairs of labeling sites, anisotropy values were independent of the labeling position, suggesting that similar FRET signals will result irrespective of the specific position to which each dye is randomly attached in the dual-labeled LMCA1_TM-A/N and LMCA1_TM-A/P mutants.

Figure 7.

Fluorescence characteristics of labeled LMCA1_TM mutants. A. Fluorescence anisotropy (r) was measured for Cy3 free in solution or attached to individual cysteines in LMCA1_TM-18, -413, -24, and -530. Error bars represent standard deviation of measurements from 10 nm emission range at three excitation wavelengths. B. Quantum yield (Φ_f) was measured relative to Cy3 free in solution for Cy3 attached to individual cysteines in LMCA1_TM-18, -413, -24, and -530. Error bars represent standard deviation of measurements performed at four excitation wavelengths.

To exclude the possibility that potential changes in FRET efficiency could (partially) reflect changes in fluorescence quantum yield, relative quantum yield measurements of Cy3-labeled LMCA1_TM-18, -413, -24, or -530 were performed under two extreme ligand conditions including either 10 mM CaCl₂ or 1 mM EGTA with 0.1 mM BeF_x (Figure 7B). None of the mutants exhibited any changes in quantum yields under these two conditions. The relative quantum yield values of free Cy3 were lower than for the Cy3-labeled mutants, consistent with the model in which the activation energy for photoisomerization of Cy3, which is responsible for the low fluorescence quantum yield and short lifetime of free Cy3, increases upon binding to a macromolecule.³⁷

Thus, the results of fluorescence anisotropy and quantum yield analyses validate the suitability of the LMCA1_TM-A/N and LMCA1_TM-A/P mutants for FRET measurements of relative distance changes induced by different ligand conditions. The differences in quantum yields and anisotropies between the labeling sites, however, suggest that direct comparisons of the FRET efficiencies between the two labeling schemes are not feasible in ensemble and confocal single-molecule FRET experiments.

Ensemble and Confocal Single-Molecule FRET Studies

The presence of a FRET signal from the LMCA1_TM-A/N and LMCA1_TM-A/P mutants labeled with a mixture of photostabilized Cy3 and Cy5 dyes (LD550 and LD650)³⁸ was confirmed by excitation of Cy3 at 530 nm and concomitant recording of emission spectra (Figure 8). Expectedly, an emission peak at ~570 nm caused by Cy3 emission was observed. The other emission peak emerging at ~670 nm was indicative of sensitized Cy5 emission, and confirmed the fluorescence resonance energy transfer from Cy3 to Cy5 for both LMCA1_TM-A/N and LMCA1_TM-A/P.

Figure 8.

Ensemble FRET imaging of LMCA1. Emission spectra of dual-labeled LMCA1_TM-A/N and LMCA1_TM-A/P mutants in the absence of ATP and Ca²⁺ (apo state). Cy3 was excited at 530 nm and the resulting emission spectra were recorded and normalized to Cy3 maximum emission (at ~570 nm).

The labeled LMCA1_TM-A/N and LMCA1_TM-A/P mutants were further investigated with the aid of confocal-based smFRET to test the eligibility of these mutants for single-molecule FRET investigations. The results are presented as the proximity ratio, also known as “relative FRET”, E_rel. Proximity ratio is equivalent to FRET efficiency, but without the correction for detection efficiency and quantum yield. When there are site-specific differences in quantum yield, the correction factor would be different between different stochastically labeled species, and single-event determination of the correction factor is thus necessary for determination of true FRET efficiencies. Single-event correction is not possible due to the short observation times in the confocal setup, but would be possible in future TIRF based studies. Thus, only relative distance changes can be inferred from these FRET measurements.

The results clearly demonstrated the presence of FRET signals in both mutants (Figure 9), which changed in response to distinct buffer compositions. Under all of the applied conditions, a zero FRET peak was present, which stems from molecules missing an active acceptor, primarily due to the stochastic labeling. In principle the presence of the zero peak could obscure very low FRET states, but the distances predicted from crystal structures should correspond to FRET values that are distinguishable from the zero peak.

Figure 9.

Confocal smFRET imaging of LMCA1. Proximity ratio (E_rel) histograms of LMCA1_TM-A/N (left) and LMCA1_TM-A/P (right) in the presence of (top to bottom): 10 mM CaCl₂, 0.5 mM CaCl₂, 0.5 mM EGTA (apo), 0.5 mM EGTA + 0.1 mM BeSO₄ + 1 mM NaF (BeF_x), or 0.5 mM EGTA + 0.1 mM AlCl₃ + 1 mM NaF (AlF_x). The histograms were fitted to a sum of three (or two in case of apo LMCA1_TM-A/N) Gaussians with the mean E_rel shown above each peak. Proximity ratio (E_rel) is defined analogously to FRET, with the exception that the detection efficiencies are not corrected (gamma-factor correction).

The remaining distribution accounts for sensitized FRET signal and was best modeled by two Gaussians in all conditions except for apo LMCA1_TM-A/N, where a single Gaussian peak fitted the data best (Figure 9). This suggests the presence of at least two populations of conformations, which in the LMCA1_TM-A/P mutant correspond well with the high-FRET E1 and low-FRET E2 states predicted by the structural model (Figure 4A). In the presence of 10 mM or 0.5 mM CaCl₂, the high-FRET peak predominates (E_rel ~ 0.74 or 0.72, respectively), in agreement with Ca²⁺ stabilizing the E1 conformations. In the apo condition, the LMCA1_TM-A/P mutant displayed a lower high-FRET peak (E_rel ~ 0.64) with a broad distribution, suggesting either a novel conformational state or a dynamic equilibrium between E1 and E2 states. Also the LMCA1_TM-A/N mutant exhibited a single, broad FRET distribution (E_rel ~ 0.44) in the apo condition, consistent with a dynamic equilibrium between E1 and E2 states as proposed for SERCA.¹⁸ The phosphate analogues BeF_x and AlF_x were used to stabilize the E2P and E2-P* states, respectively, in the absence of Ca²⁺. Both mutants responded to these inhibitors by shifting the distribution toward lower FRET, in accordance with the structural models.

There are two different classes of crystal structures of the Ca²⁺-bound state of SERCA. The first structure was determined in the absence of nucleotides and showed a wide open cytoplasmic headpiece.¹¹ Subsequent structures of SERCA in complex with Ca²⁺ and ATP-analogues have been much more compact.¹² Thus, the functional relevance of the open structure has been controversial.³⁹ The two types of Ca²⁺ states can be directly probed in smFRET studies by monitoring the interdye distance in LMCA1_TM-A/N. We observed that in the presence of both 10 mM CaCl₂ (high concentration used to crystallize SERCA with an open cytoplasmic headpiece) and 0.5 mM CaCl₂ (where maximum activity of LMCA1 was observed), a high-FRET peak (E_rel ~ 0.57 and 0.58, respectively) predominated. These FRET values are higher than in the apo condition or in the presence of metal fluorides, which suggests a more compact conformation of the cytoplasmic headpiece of LMCA1 to prevail upon Ca²⁺ binding in solution. Thus, the compact headpiece structure seems to be physiologically relevant, while the open structure may represent a transient structure that can be captured by crystallization at high concentrations of Ca²⁺. This conclusion is in concordance with a previous proposal based on FRET studies of a SERCA-CFP fusion protein.¹⁸

In conclusion, an experimental system for single-molecule FRET measurements was established, which should allow high-resolution, real-time FRET studies of the conformational dynamics of a functional cycle of LMCA1 in the future. The high degree of structural conservation among P-type ATPases suggests that mechanistic conclusions drawn from LMCA1 can be generalized to large parts of the family, which justifies the use of a convenient bacterial homologue as a model system for studying general features of the pumping cycle. Furthermore, inhibition studies confirmed LMCA1 to be a representative P-type ATPase, which responds to metal fluorides in the same manner as SERCA. The states trapped by BeF_x and AlF_x were demonstrated to be structurally distinct via confocal smFRET measurements. Additionally, our smFRET data suggest that the cytoplasmic headpiece of LMCA1 becomes more compact after Ca²⁺ binding. As opposed to previous FRET studies of SERCA engineered with fluorescent proteins,^18–20 our approach is based on the site-specific labeling with small organic fluorophores characterized by remarkable stability and brightness.²¹ This strategy potentially allows the detection of dynamics at a single-molecule level by total internal reflection fluorescence (TIRF) microscopy.

MATERIALS AND METHODS

Sequence Alignment, Homology Models, and Calculation of Intramolecular Distances

1121 bacterial ${\mathrm{P}}_{{\text{IIA}}^{-}}$ and ${\mathrm{P}}_{0^{-}}$ (unclassified) ATPases from the UNIPROT database were found using the PUMPKIN P-type ATPase database: http://octo3.bioxray.au.dk/pump-classifier/p-type-atpase-database/. A multiple-sequence alignment of those sequences was constructed using MUSCLE.⁴⁰ When plotting conservation scores, a column in the alignment corresponding to a given residue in LMCA1 was plotted as amino acid frequencies in MATLAB (MathWorks). PDB entries 4H1W, 1SU4, 1T5T, 3B9B, 3B9R, and 2C88 were used as template structures in MODELLER³⁰ to model LMCA1 in the inward-open E1 state, the nucleotide-free E1 state, the calcium-bound E1~P state, the phosphorylated E2P ground state, the dephosphorylated E2-P* transition-like state, and the calcium-free E2 state, respectively. Homology models were made using an online version of the program provided by the Bioinformatics Toolkit from Max-Planck Institute for Developmental Biology (http://toolkit.tuebingen.mpg.de/modeller). A sequence alignment of LMCA1 and SERCA1a performed in MUSCLE⁴⁰ was used as an input file together with the aforementioned PDB structures. Structural figures were prepared in PyMol (v 1.7, Schrödinger LLC, http://www.pymol.org).

Intramolecular distances in LMCA1 were evaluated using a script written in Tool Command Language (Tcl) and designed to work under Tk console in VMD.⁴¹ It enabled loading of multiple structures of the LMCA1 homology model, choosing two selections in the protein and calculating the distances between all residues within these selections and within all the structures loaded. In this way, pairs of residues suitable for reporting distance changes using FRET were identified. The code is provided in Supporting Information.

Site-Directed Mutagenesis, Expression, and Purification

A pET-22b plasmid (Novagen) containing the LMCA1 gene followed by a nucleotide sequence encoding a C-terminal linker (DYDIPTT sequence), a Tobacco Etch Virus (TEV) protease site (ENLYFG sequence), an XhoI restriction site (CTCGAG sequence), and a six histidine tag (6x CAC sequence), previously described in Faxen et al.,⁷ was used as a template for introducing four additional histidines into the histidine tag using the QuikChange mutagenesis kit (Agilent Technologies). The resulting construct, pET-22b:LMCA1–10xHis, was used as the template for the introduction of mutations removing or introducing cysteines for labeling purposes. All mutations were verified by Sanger sequencing (Eurofins, MWG).

The applied protocol for LMCA1 expression and purification was adapted from Faxen et al.⁷ Solubilized membranes were loaded on a 5 mL HisTrap HP column (GE Healthcare) equilibrated in buffer C (20 mM Tris-HCl pH 7.6, 200 mM KCl, 20% v/v glycerol, 1 mM MgCl₂, 5 mM BME, 0.25 mg/mL C₁₂E₈) with 50 mM imidazole. After washing of the column in the same buffer, bound protein was eluted with buffer C containing 150 mM imidazole.

Determination of ATPase Activity

The ATPase activity of LMCA1 was measured by determining the liberation of inorganic phosphate by the Baginski method.⁴² 1 μL of purified LMCA1 at a concentration of 0.1–5 μg/μL was added to a final volume of 50 μL in reaction buffer (20 mM Tris-HCl pH 7.6, 200 mM KCl, 20% v/v glycerol, 1 mM MgCl₂, 1 mM CaCl₂, 0.4 mM NaMoO₄, 10 mM NaN₃, 40 mM KNO₃, 5 mM BME, 0.25 mg/mL C₁₂E₈) in a 96-well microtiter plate at room temperature. Reactions were initiated by the addition of ATP to a final concentration of 3 mM. The reactions were stopped after 1–30 min by addition of 50 μL ascorbic acid solution (140 mM ascorbic acid, 5 mM ammonium heptamolybdate, 0.1% (w/v) SDS and 0.4 M HCl) and incubated for 15 min. The ascorbic acid solution was freshly prepared and stored on ice. The color of the reduced heteropolymolybdate-phosphate complex (molybdenum blue) was stabilized by the addition of 75 μL arsenate solution (150 mM sodium arsenate, 70 mM sodium citrate, and 350 mM acetic acid). The absorbance of molybdenum blue was measured at 860 nm in a VICTOR X plate reader (PerkinElmer).

Maleimide-Based Labeling of Cysteines

LMCA1 was dialyzed against the labeling buffer (50 mM MOPS-KOH pH 6.8, 200 mM KCl, 20% glycerol, 1 mM MgCl₂, 0.2 mM tris(2-carboxyethyl)phosphine (TCEP), 0.25 mg/mL C₁₂E₈). Cy3 and Cy5 (~0.25 mg) (GE Healthcare) or LD550 and LD 650 dyes (Lumidyne Technologies) were dissolved in 31 μL DMSO, degassed, and the concentrations of the dyes were determined by measuring the absorbance of 4000× diluted dyes in methanol. 4–16 molar excess of Cy3 and 5–20 molar excess of Cy5 were added to the protein. Typically, the LMCA1 concentration was 10–20 μM and the labeling reaction volume was 500 μL. Labeling was performed for 15–60 min at room temperature and stopped by the addition of 140 mM BME. LMCA1 was separated from the unbound dyes either by dialysis followed by passage over a PD-10 desalting column (GE Healthcare) or by size exclusion chromatography on a Superdex 200 Increase 10/300 GL column (GE Healthcare). SDS-PAGE of fractions collected from the PD-10 column was run and in-gel fluorescence was visualized using a Typhoon gel imager (Amersham Biosciences). Molar concentrations of the dyes and of the protein were determined by absorption measurements on a NanoDrop (Thermo Scientific) or directly during size exclusion chromatography, if it was used to separate unreacted dyes. Absorption was measured at 550, 650, and 280 nm with extinction coefficients of 150 000, 250 000, and 48 100 M⁻¹ cm⁻¹ for Cy3, Cy5, and LMCA1, respectively. Labeling efficiency was calculated as a dye to LMCA1 molar ratio, corrected for absorption of Cy3 at 280 nm (0.08 × Cy3 absorption at 550 nm), absorption of Cy5 at 280 nm (0.05 × Cy5 absorption at 650 nm), and by absorption of Cy5 at 550 nm (0.06 × Cy5 absorption at 650 nm).

Ensemble Anisotropy Measurements

Anisotropy measurements were performed using a Fluoromax 3 spectrofluorometer (HORIBA Jobin Yvon) with polarization filters. 50 nM solutions of Cy3-labeled LMCA1_TM variants in the imaging buffer (50 mM Tris-HCl, pH 7.6, 200 mM KCl, 20% glycerol, 1 mM MgCl₂, 0.2 mM TCEP, 0.25 mg/mL C₁₂E₈) were used for measurements in a 60 μL quartz cuvette. Data were acquired using three excitation wavelengths of 530, 540, and 550 nm and a 10 nm emission range with 1 nm increments around the emission peak (~570 nm), and presented as an average and standard deviation from the emission range upon three excitations. Fluorescence anisotropy (r) was calculated as

$$r=\frac{I_{\mathrm{VV}}-G\times I_{\mathrm{VH}}}{I_{\mathrm{VV}}+2G\times I_{\mathrm{VH}}}$$

where I_VV and I_VH are the fluorescence intensities with polarizations vertical and horizontal to the vertically polarized excitation beam, and G is a grating factor defined as G = I_HV/I_HH, where I_HV and I_HH are the fluorescence intensities with polarizations vertical and horizontal to the horizontally polarized excitation beam.

Relative Quantum Yield Measurements

The relative fluorescence quantum yield of Cy3-labeled LMCA1_TM variants in the imaging buffer (50 mM Tris-HCl, pH 7.6, 200 mM KCl, 20% glycerol, 1 mM MgCl₂, 0.2 mM TCEP, 0.25 mg/mL C₁₂E₈ supplemented either with 10 mM CaCl₂ or 1 mM EGTA, 0.1 mM BeSO₄, and 1 mM NaF) was determined using absorption and fluorescence measurements in a quartz cuvette with 1 cm path length. The measurements were performed at two sample concentrations and the absorbance values were kept below 0.1. The relative fluorescence quantum yield (Φ_f) was calculated as

$${\mathrm{\Phi }}_f=\frac{I}{\mathrm{Abs}}$$

where Abs is the corrected absorbance at the fluorescence excitation wavelength and I is the integrated area under the corrected fluorescence spectrum. The data were presented as an average and standard deviation of measurements at four excitation wavelengths: 520, 530, 540, and 550 nm.

Ensemble FRET Measurements

FRET measurements in solution were performed using a Fluoromax 3 spectrofluorometer (HORIBA Jobin Yvon). 50 nM solutions of dual-labeled LMCA1_TM-A/N or LMCA1_TM-A/P in the imaging buffer (50 mM Tris-HCl, pH 7.6, 200 mM KCl, 20% glycerol, 1 mM MgCl₂, 0.2 mM TCEP, 0.25 mg/mL (~0.46 mM) C₁₂E₈) were used for measurements in a 60 μL quartz cuvette. Emission spectra were recorded in the range of 540–720 nm after excitation at 530 nm.

Confocal Single-Molecule FRET Measurements

Dual-labeled LMCA1_TM-A/N or LMCA1_TM-A/P was diluted to a final concentration of 40 pM in 50 mM Tris-HCl, pH 7.6, 200 mM KCl, 20% glycerol, 1 mM MgCl₂, 0.2 mM TCEP, 0.25 mg/mL C₁₂E₈ additionally containing either 10 mM CaCl₂, 0.5 mM CaCl₂, 0.5 mM EGTA, or 0.5 mM EGTA with 1 mM NaF and with 0.1 mM BeSO₄ or 0.1 mM AlCl₃. Single-molecule measurements were recorded at 20 °C on a custom built instrument: A collimated Gaussian laser beam (532 nm, final power 100 μW) was directed to the back port of a Nikon TiE inverted microscope, and focused 10 μm into a sealed solution containing the protein solution. Fluorescence emission was collected through an oil-immersion objective (Apochromat 60 NA 1.40, Nikon), directed to the detection pathway by a bandpass dichroic (Chroma: zt532dcrb-UF1), and focused onto a 50 μm pinhole. Donor and acceptor fluorescence was separated with a dichroic mirror (Chroma: ZT633rdc-UF1) and filtered by wavelength (Donor channel: Chroma: ET600/75m, Acceptor channel: Chroma ET720/150m, Semrock: LP02–647) before being focused onto avalanche photodiodes (SPCM-AQR-14, PerkinElmer). Photon counts were binned into 1 ms bins and recorded for 30 min for each sample. Event-selection based on a simple SUM threshold of 50 photons and correction for autofluorescence and cross-talk were done using pyFRET.⁴³ Proximity ratio, also known as “relative FRET” (E_rel) was calculated as E_rel = I_A/(I_A+A_D) without correction for differences in relative quantum yields of the two dyes or for the relative detection efficiency of the two channels (γ-factor of 1). The proximity ratio histograms were fitted with a sum of two or three Gaussians in GraphPad Prism 7.0. The appropriate model was chosen based on the extra sum-of-squares F test with the P value set to 0.05.