Structural stability of purified human CFTR is systematically improved by mutations in nucleotide binding domain 1

Abstract

The Cystic Fibrosis Transmembrane Conductance Regulator (CFTR) is an ABC transporter containing two transmembrane domains forming a chloride ion channel, and two nucleotide binding domains (NBD1 and NBD2). CFTR has presented a formidable challenge to obtain monodisperse, biophysically stable protein. Here we report a comprehensive study comparing effects of single and multiple NBD1 mutations on stability of both the NBD1 domain alone and on purified full length human CFTR. Single mutations S492P, A534P, I539T acted additively, and when combined with M470V, S495P, and R555K cumulatively yielded an NBD1 with highly improved structural stability. Strategic combinations of these mutations strongly stabilized the domain to attain a calorimetric T_m > 70 °C. Replica exchange molecular dynamics simulations on the most stable 6SS-NBD1 variant implicated fluctuations, electrostatic interactions and side chain packing as potential contributors to improved stability. Progressive stabilization of NBD1 directly correlated with enhanced structural stability of full-length CFTR protein. Thermal unfolding of the stabilized CFTR mutants, monitored by changes in intrinsic fluorescence, demonstrated that Tm could be shifted as high as 67.4 °C in 6SS-CFTR, more than 20 °C higher than wild-type. H1402S, an NBD2 mutation, conferred CFTR with additional thermal stability, possibly by stabilizing an NBD-dimerized conformation. CFTR variants with NBD1-stabilizing mutations were expressed at the cell surface in mammalian cells, exhibited ATPase and channel activity, and retained these functions to higher temperatures. The capability to produce enzymatically active CFTR with improved structural stability amenable to biophysical and structural studies will advance mechanistic investigations and future cystic fibrosis drug development.

1. Introduction

CFTR functions as a chloride- and bicarbonate-selective ion channel in the plasma membrane [1], [2], [3]. Genetic defects in CFTR cause cystic fibrosis (CF), a life-threatening disease manifested by dysregulation of epithelial fluid in the lungs, pancreas and other organs. More than 2000 mutations have been reported in the CFTR gene, of which at least 280 mutations are confirmed to cause the disease (http://cftr2.org) [4]. Depending on location within the CFTR gene, disease-causing mutations vary in severity, and have been categorized into six classes according to their phenotypes in reducing protein expression, function, and/or stability [5], [6], [7].

Personalized medicine treating a CF patient’s underlying defect has become a high priority, and is actively endorsed by the Cystic Fibrosis Foundation [8]. A prominent success is the recently FDA-approved channel potentiator ivacaftor (Kalydeco, VX-770), that improves channel activity in G551D and other gating mutations (about 8% of the patient population) [6,7,9,10]. Yet the majority of patients still have few treatment options. The most prevalent disease-causing mutation is a single amino acid deletion, ΔF508, of which at least one allele is found in 85–90% of CF patients [4]. ΔF508 severely compromises CFTR folding, thus resulting in faulty maturation, poor localization to the cell surface, and reduced channel function by the few molecules which do reach the cell surface [11,12]. Recently tested combination therapies for patients with ΔF508-CFTR that target these multiple defects (Orkambi, or three next-generation CFTR modulator therapies with ivacaftor and tezacaftor (VX-661)) improved function by only a modest 3–15% [13], [14], [15], [16]. Thus, there is a pressing need to better understand CFTR folding, structure and function, not only to address how these are affected in patients with CF, but also to comprehend the effects of potential drug candidates.

CFTR is an atypical ATP-binding cassette (ABC) transporter with two transmembrane domains that harbor the ion channel, and two non-identical nucleotide binding domains (NBD1 and NBD2) that bind and hydrolyze ATP to gate the channel; gating is regulated by phosphorylation of a unique R-region [2,3,17]. Complex folding and domain assembly are limiting steps in the biogenesis and trafficking of CFTR [18], and these are compromised by many CF-causing mutations such as ΔF508 in NBD1, resulting in protein misfolding and degradation by cellular quality control systems [11]. Due to intensive efforts to understand the consequences of F508 deletion, high resolution X-ray structures of bacterially expressed isolated NBD1 domain have been solved for wild-type (WT) and ΔF508 mutants from human and mouse [19], [20], [21], [22]. While ΔF508 alters NBD1 structure very little, it severely affects the folding kinetics, biogenesis, and the stability and dynamics of the domain [23], [24], [25], [26], [27]. F508 is situated on the surface of NBD1 near the interface with the intracellular loops (ICLs) that connect to the transmembrane helices. This interfacial position, akin to a ball and socket joint, further complicates the folding and stability of NBD1 when in the context of the full-length protein, and may explain why a single amino acid deletion causes multiple defects that ultimately lead to defective CFTR biogenesis and compromised function at the cell surface [11,18,25,28,29].

Recently, the first structures of full-length CFTR from human and zebrafish (55% sequence identity to human) in a dephosphorylated, inactive state have been solved by cryoelectron microscopy (cryo-EM) at a nominal resolution of 3.7 to 3.9 Å (PDB 5uar, 5uak) [30,31]. These structures revealed details of the transmembrane α-helical domains and the channel forming pore, with NBD1 and NBD2 widely separated and intercalated by portions of the regulatory R-region; but structures of the NBDs were of low resolution and most of the R-region, the regulatory insertion (RI), the fourth extracellular loop and the C-terminus could not be assigned, perhaps due to their disordered nature. A subsequent 3.4 Å resolution structure of phosphorylated zebrafish CFTR showed two well-ordered NBDs in an asymmetric sandwich dimer conformation with two bound, non-hydrolyzed ATP molecules, with the channel open to the cytoplasm and the extracellular gate closed (PDB 5W81) [32]. The structure revealed insights into the gating mechanism and likely presents a “pre-open” closed channel state [33,34]. Despite these recent breakthroughs, higher resolution structures of human WT CFTR and of CFTR with at least some of the CF-causing mutations, and in conformationally distinct functional states, are still needed to better reveal the underlying structural defect for the many types of disease-causing mutations, and to advance therapeutic strategies. For such studies, it will be critical to have accompanying biochemical and biophysical evidence of perturbations caused by the different mutant theratypes. A stable and monodisperse protein is the key to higher resolution structures whether by cryo-EM or crystallography, and also for biochemical and biophysical studies to probe the energetics, dynamics and mechanism of action that cannot be revealed by structure alone. However, the low stability of CFTR in detergent solution, and even more so ΔF508-CFTR and other folding mutations, has impeded progress [35].

A validated approach to improve CFTR stability comes from the body of literature describing ‘second site’ mutations of ΔF508-CFTR [36], [37], [38]. Those studies utilized sequence alignments and yeast screens of chimeric ABC transporters to identify mutations in NBD1 that rescued biogenesis of full-length ΔF508-CFTR [39], [40], [41], [42], [43]. Structural and biophysical studies of NBD1 domain constructs containing strategic mutations improved F508- and ΔF508-NBD1 solution properties and thermal stability [20], [21], [22]. It was further demonstrated that certain of these ΔF508-stabilizing mutations increased biogenesis of WT CFTR in mammalian cells [29,44], and some improved thermal stability of CFTR channel function [37,45]. Several recent reports have shown that select mutations which thermostabilized NBD1 also enhanced thermal stability of purified CFTR [45], [46], [47], [48], [49]. From these cited studies, a tacit correlation has emerged suggesting that mutations that enhance NBD1 domain structural stability will also improve the structural stability ² of the full length CFTR.

In order to broaden the utility of this implicit correlation, we have conducted a systematic analysis of NBD1 stabilization by a panel of mutations, singly or in combinations. We then investigated their consequences on the stability of purified full-length CFTR. Two hypotheses were tested: 1) that specific combinations of single substitutions can be identified which will cumulatively yield an NBD1 domain with highly improved structural stability; and 2) that stabilizing NBD1 in full-length CFTR will enhance structural stability of purified CFTR in detergent solution. The study serves the overarching goal of producing full-length CFTR proteins with highly improved structural stability. Stabilized human CFTR may be valuable for biophysical and structural studies that advance CF drug development, and future mechanistic and structural studies of rare CF-causing mutations.

2. Results

2.1. NBD1 stabilization by mutagenesis

Fig. 1 shows the locations of the known NBD1 stabilizing mutations. The mutations are color-coded based on their history of discovery; pink: [37,42,43], red: [37], gray: [20], blue: [39], [40], [41]. Most of these known mutations are located in the α-subdomain. As we will show, M470V located in the first helix in the α/β-subdomain (orange in Fig. 1) is also a stabilizing mutation. The M470 polymorphism is nearly always present with the ΔF508 mutation, compared to 50% concurrence of either M or V alleles with WT CFTR [50]. M470-CFTR protein reportedly matures more slowly, and exhibits a 1.7-fold increased intrinsic chloride channel activity compared to V470-CFTR [51], [52], [53].

Fig. 1.

Stabilizing mutations in NBD1. Crystal structure of human NBD1 (2PZE), rendered in Pymol, showing the location of mutations studied. Mutations are color-coded (pink, red, blue, gray) based on their history of discovery (see Results). Helix 5 (blue) contains the conserved Signature sequence. Bound Mg²⁺ and ATP are shown as green ball and lime-green sticks, respectively.

2.2. Single NBD1 mutations increase calorimetric Tm

We have previously applied DSC to investigate the thermodynamics and kinetics of unfolding of isolated NBD1 proteins [24,54,55]. Here we have determined the calorimetric unfolding temperature (T_m^cal) of individual mutations in the isolated NBD1 domain, then assessed the effects of multiple NBD1 mutations (Fig. 2). We used the WT human NBD1 carrying a deletion of the 32 amino acid regulatory insertion (ΔRI-NBD1, Δ405–436) which is not present in other ABC transporters [21]. ΔRI-NBD1 shows favorable solution properties and enabled the structure determinations of WT and ΔF508-NBD1 [21]. We found that the M470V mutation (V470 polymorphism) strongly stabilized ΔRI-NBD1 (Tm of 56.7 ± 0.3 °C, n < 22), with a ΔT_m^cal of +4.6 °C (Tm of 61.3 ± 0.1 °C, orange in Fig. 2), an effect equal to or larger than the majority of stabilizing α-subdomain mutations. S495P gave the highest ΔT_m^cal of +5.8 °C (Tm of 62.5 ± 0.2 °C, red in Fig. 2). S492P, A534P, and I539T all showed modest stabilization of NBD1 by +1.2 to 2.5 °C. S495P is located in the Q-loop of the α-subdomain, a flexible loop that plays an important role in ATP binding and interacts with transmembrane domains [37]. S495P, S492P (in the same Q-loop), A534P, and I539T (pink in Fig. 1) were originally identified from sequence alignments of CFTR orthologs [43]. A534P and I539T are in the structurally diverse region, a flexible loop that diverges in CFTR compared to other human ABC-C subfamily members (Supplemental Fig. S1) [22].

Fig. 2.

Stabilization of ΔRI-NBD1 by individual and multiple mutations. Point mutations were introduced into ΔRI-NBD1 containing the WT CFTR sequence with the M470 polymorphism (named “none”), and the T_m^cal determined as described in Methods. The naming was based on conventions in the CFTR field. S492P/A534P/I539T was previously named 2PT (references [37,42,43,45]), and G550E/R553M/R555K were discovered by Teem et al. (references [39], [40], [41]). New mutation combinations 5SS, 6SS, and 7SS discovered in this study were named according to the number of stabilizing substitutions (SS) introduced on top of the ΔRI (deletion of 405–436) NBD1 background. Shown are average ± range of 2 to 3 independent experiments, with T_m^Cal ranges among replicates ≤0.6 °C. For some data points the error bars were so small that the bar is not seen. Data for these and additional combinations appear in Fig. S2.

The well-studied Teem mutations G550E, R553M and R555K [39], [40], [41] are located in Helix 5 that overlaps the Signature motif LSGGQ (blue in Fig. 1; see also Supplemental Fig. S1). These mutations stabilized NBD1 to varying degrees, with R555K imparting the greatest ΔT_m^cal of +3.5 °C (Tm of 60.22 ± 0.01 °C, blue bars, Fig. 2). Introducing the interface mutation V510D into the flexible loop that contains F508 [24] improved NBD1 thermal stability by +2.6 °C. The so-called solubilizing mutations identified from X-ray crystallography [20,22], F494 N in the Q-loop and Q637R near the unstructured C-terminus (gray in Fig. 1), produced only small ΔT_m^cal in ΔRI-NBD1 (Fig. 2).

2.3. T_m^cal of combinations of known NBD1 mutations

A limited number of mutation combinations were previously shown to increase thermal stability of NBD1 in an additive manner, and to improve biogenesis of full-length ΔF508-CFTR [37]. For example, combining the three mutations S492P/A534P/I539T (2PT) with G550E/R553M/R555K (Teem), restored maturation of ΔF508-CFTR to ~90% of the WT [37]. The T_m^cal of ΔRI-NBD1 containing these 6 mutations was +8.0 °C higher than ΔRI-NBD1 (Tm of 64.7 ± 0.3 °C, 2PT/Teem in Fig. 2). Summing individual ΔT_m^cal for single mutations predicted reasonably well the stabilization effects for combinations of up to three mutations in ΔRI-NBD1, and somewhat overestimated the ΔT_m^cal for larger combinations (Fig. S2). Summation would predict a ΔTm of +8.9 °C for the six-mutation 2PT/Teem combination, a fairly good estimation of the measured ΔT_m^cal of 8.0 °C. Interestingly, when 2PT/Teem was introduced into ΔF508-NBD1, the observed ΔT_m^cal of +10.7 °C was even larger than predicted by summation (data not shown). Therefore, we examined whether strategic mutations with strong stabilizing effects may be combined to maximally stabilize NBD1.

2.4. Maximizing thermal stability of NBD1 by strategic combinations of mutations

Building on the success of ΔRI/2PT/Teem, we attempted to generate a “super-stabilized” NBD by adding two other Q-loop mutations, F494N and S495P, and omitting the ineffective R553M (7SS with a total of seven mutations, red in Fig. 2). 7SS-NBD1 (2PT/F494N/S495P/G550E/R555K) exhibited a much improved T_m^cal of 71.1 ± 0.05 °C (+14.4 °C over ΔRI-NBD1 and +6.4 °C over 2PT/Teem). Additionally, in order to maximally improve thermal stability with the fewest mutations, ΔRI-NBD1 was progressively and systematically modified with two or more of S492P, A534P, or I539T along with the strongest single mutations M470V, S495P or R555K (Fig. S2). This process led to 5SS- and 6SS-NBD1 containing five and six of the most stabilizing single mutations, respectively. The 5SS-NBD1 (2PT/M470V/S495P) registered a T_m^cal of 71.8 ± 0.05 °C (ΔT_m^cal of +15.1 °C, Fig. 2). 6SS-NBD1 (2PT/M470V/S495P/R555K) was the most stable combination of all, with a T_m^cal of 74.2 ± 0.2 °C (ΔT_m^cal of +17.5 °C, Fig. 2). Thus our strategic and iterative testing of combinations of mutations succeeded in identifying several NBD1 variants with greatly improved thermal stability.

2.5. Mammalian cells produce mature full-length CFTR variants with stabilizing mutations

Based on previous studies [37,45], the stabilized NBD1 variants identified above may be expected to favor folding and trafficking of full-length CFTR to the plasma membrane in a mammalian expression system. We introduced a series of mutation combinations shown to increase NBD1 stability into a previously described human CFTR construct, His₁₀-SUMO*-CFTR^FLAG-EGFP [56], and also deleted RI (Δ405–436). Positioning the FLAG epitope into the fourth extracellular loop of CFTR enabled reliable cell-surface compartmentalization by flow cytometry as we demonstrated in [56]. The RI deletion is known to rescue maturation, stability and function of full-length ΔF508-CFTR, and to enhance channel activity of CFTR at the cell membrane [43]. All combination mutations that were expressed in full-length CFTR are summarized in Table 1, together with their short name and cell line designation. Comparisons of expression levels among cell lines suggested that surface CFTR expression increased in the ΔRI-CFTR and ΔRI/2PT-CFTR mutations (cell lines D992 and D994, Fig. S3) compared to WT CFTR (D1044), in agreement with previous reports [43]. However, when ΔRI/2PT-CFTR was combined with additional mutations as in 5SS-(D1012), 6SS-(D1013) or 6SS/H1402S-CFTR (D1028), surface CFTR expression was not further increased. On the other hand, CFTR expression could be increased by other tet-on strategies, see supplementary information (Table S1 and Figs. S3 and S4). Importantly, all cell lines produced fully mature and glycosylated CFTR (Fig. 3A, band C), suggesting that, even in the most highly expressing cell lines (D1013 and D1028, Fig. S3), the majority of CFTR protein passed quality control in the endoplasmic reticulum, and trafficked normally to the plasma membrane. Thus WT and stabilized CFTR variants provided excellent source material for production of purified CFTR.

Table 1:

Mutations that improve NBD1 stability also improve thermal stability of the ATPase activity.

		Cell line 1)	TmTrp (oC)			Tmfunc (oC)			P value 3)
	short name		mean	range	n	mean	sd	n	vs WT	vs 2PT	vs 7SS
WT (HEK)		D165 2)				21.55	0.53	3	ns
WT (CHO)		D421				22.23	0.55	3
NBD1 mutations:
ΔRI (HEK)		D359				24.64	1.34	3	0.0449	0.0311
ΔRI/S492P/A534P/I539T (CHO)	2PT	D727	53.2	0.2	2	27.41	0.61	3	0.0004
ΔRI/2PT/M470V (CHO)		D851	56.1	0.4	2	28.27	1.17	3	0.0013	ns
ΔRI/2PT/M470V/S495P (CHO)	5SS	D1012	66.0	1.3	3	33.50	1.23	4	<.0001	0.0006	0.0369
ΔRI/2PT/M470V/S495P/R555K (CHO)	6SS	D1013	67.4	0.7	3	35.72	0.75	3	<.0001	0.0001	ns
ΔRI/2PT/F494N/S495P/G550E/R555K	7SS	D744	62.8	0.6	4	35.65	0.46	3	<.0001	<.0001
NBD2 mutations:
ΔRI/2PT/H1402S (CHO)		D869	58.2	0.1	2	Inactive 4)
ΔRI/2PT/M470V/H1402S (CHO)		D872	61.3	1.6	8	Inactive 4)
ΔRI/6SS/H1402S (CHO)	6SS/H1402S	D1028	72.2	1.1	3	inactive 4)
ΔRI/2PT/S1359A (CHO)		D805	52.5	1.2	3	27.24	0.10	3	0.0001	ns	<.0001
ΔRI/7SS/S1359A (CHO)		D804				31.32	1.00	3	0.0002	0.0044	0.0024
ΔRI/2PT/S1255L/K1334G/S1359A/Q1411D (CHO)	2PT/quad	D742	53.0	1.1	3	28.07	0.18	3	<.0001	ns
ΔRI/7SS/S1255L/K1334G/S1359A/Q1411D (CHO)	7SS/quad	D743				low activity

¹⁾Cell line designation is given to facilitate future sample distribution. Request for cell lines should be address to John.Kappes@uab.edu

²⁾D165 is the HEK cell line expressing WT CFTR (Hildebrandt et al., 2015).

³⁾T_m^func values were compared using the two-tailed student T-test. ns: P> 0.05

⁴⁾H1402S variants showed less than 2% of WT CFTR activity, see Fig. 3B.

Fig. 3.

Cellular maturation and activity of full-length CFTR with NBD1-stabilizing mutations. A) Purified, phosphorylated WT CFTR and mutants were resolved by SDS-PAGE and silver stained. Insertion of the FLAG epitope at position 901 disrupts one of two glycosylation sites, and thus mature glycosylated (band C) and core glycosylated (band B, indicated by arrows) forms do separate but only poorly on a 7% polyacrylamide SDS-gel. B) Purified proteins were supplemented with destabilized lipid, and ATP hydrolysis was measured at subsaturating 0.3 mM [α³²P]-ATP as described in Methods. D359 is an HEK cell line expressing ΔRI-CFTR. Averages ± standard deviations of at least 3 experiments are shown.

2.6. Stabilized CFTR variants retain ATPase function

We previously demonstrated that phosphorylated, highly purified CFTR from HEK or CHO cells and supplemented with phosphatidylserine hydrolyzes ATP at rates comparable to other ABC transporters [49,57]. Not unexpectedly, rates of ATP hydrolysis were altered when multiple NBD1 mutations were introduced (Fig. 3B). ATPase activity of ΔRI/2PT-CFTR (pink) increased 2-fold over WT, and the highest activity was exhibited by the ΔRI/2PT/M470V-CFTR variant (orange). In many cases the activity was reduced but still readily measurable. Although activities of 5SS-, 6SS- and 7SS-CFTR variants (red) were similar to WT, they were substantially reduced compared to ΔRI/2PT/M470V-CFTR, probably due to the presence of additional mutations in the Q-loop (S495P) or the Signature sequence (G550E and R555K, Fig. 1). In contrast, activity was abolished by negative control mutations of NBD2 residue H1402 (Fig. 4, D869, D872), the switch histidine which interacts with MgATP and bound H₂O at the catalytic center [17]. Mutations of H1402 also abolish ATP-dependent channel gating by locking CFTR in an open channel conformations [45].

Fig. 4.

Stabilization of NBD1 in CFTR increases the inactivation threshold for ATP hydrolysis, T_m^func. Purified CFTR was pre-incubated for 30 min in the absence of lipid or nucleotide over a range of temperatures, then ATPase activity remaining was measured as detailed in Methods. In each experiment, the least squares sigmoidal fit was used to derive the inflection point defined as the functional T_m. T_m^func was measured 3 to 4 times for each mutant, and a statistical comparison is given in Table 1. This plot shows averages ± standard deviations of data combined from the replicate experiments.

2.7. NBD1-stabilization translates to functional stabilization of CFTR

To evaluate the biophysical properties of the purified protein in detergent solution, we determined a functional half-life at different temperatures, and global unfolding of the protein by differential scanning fluorimetry (see next paragraph). A 30 min heat treatment of each purified protein was used to determine the transition temperature for the inactivation of ATP hydrolytic activity as shown in Fig. 4, which we have defined as the functional T_m (or half-life referred to here as T_m^func) [49]. T_m^func was 22 °C for WT CFTR purified from either CHO or HEK cells (Fig. 4B, black circles, Table 1). Deletion of the RI (D359) produced a small but significant thermal stabilization (Table 1). ΔRI/2PT mutations (D727) or ΔRI/M470V/2PT (D851) further shifted the T_m^func to 28 °C. Notably, although M470V significantly stabilized the T_m^cal of the isolated ΔRI-NBD1 and ΔRI/2PT-NBD1 mutant combinations tested by DSC by ~5 °C (Fig. 2), a shift in the T_m^func of less than 1 °C was observed in full-length CFTR (p = 0.322, Table 1). The most significant stabilization was observed with the 6SS (D1013) and 7SS (D744) that shifted T_m^func by +12 °C to 35.7 °C (Fig. 4 and Table 1). T_m^cal correlated strongly with T_m^func as shown in Fig. 5A. The data show that progressive stabilization of the NBD1 domain translates into thermal stabilization of NBD function in the purified full-length protein.

Fig. 5.

NBD1 thermal stability directly correlates with full-length CFTR functional and thermal stabilities. For each combination of mutations, (A) T_m^func or (B) T_m^trp for unfolding of purified full length CFTR in detergent solution by intrinsic fluorescence (Fig. S5) is plotted against T_m^cal for NBD1 unfolding obtained by DSC (Fig. 2). Pearson correlation coefficients were 0.972 for the 6 constructs containing only the NBD1 mutations (solid line), and 0.986 for the 3 constructs containing NBD1 mutations plus H1402S (dashed line). C) Correlation between T_m^trp for CFTR unfolding (Fig. S5) and T_m^func for ATPase inactivation (Fig. 4). The Pearson correlation coefficient was 0.966. See Table 1 for mutation nomenclature. Note that the T_m^func is lower than T_m^Trp likely because of the long exposure time of 30 min at a fixed temperature in the absence of protective MgATP compared to a continuous ramp of 1 °C/min used in Trp unfolding experiments, see Methods.

2.8. NBD1-stabilizing mutations also thermally stabilize full-length CFTR

Tryptophan (Trp) fluorescence is sensitive to environment, and protein unfolding can often be monitored by changes in fluorescence intensity and sometimes changes in the emission maximum wavelength (λ_max) [55]. In the folded state at 10 °C, CFTR showed a λ_max of 334 nm (data not shown), consistent with the majority of detected Trps being in a hydrophobic environment. A representative trace of intrinsic fluorescence vs. temperature is shown in Fig. S5. ΔRI/2PT-CFTR and other stabilized variants all exhibited a single sigmoidal unfolding transition in fluorescence intensity at a characteristic temperature (Fig. 5). A single transition is not unusual for multidomain proteins that cooperatively unfold [58].

The T_m^trp was 53.2 ± 0.1 °C for ΔRI/2PT-CFTR, increasing to 56.2 ± 0.2 °C°C in the ΔRI/2PT/M470V (D851) mutant and to 67.4 ± 0.4 °C for the most stable 6SS. The H1402S mutation, which abolished ATP hydrolysis (Fig. 3), in combination with ΔRI/2PT or with ΔRI/2PT/M470V or 6SS, consistently improved CFTR thermal stability by +5.0 ± 0.2 °C (Fig. 5, green solid triangles). This finding is in agreement with proposed NBD dimerization due to occlusion of putatively non-hydrolyzed MgATP, which imposes a locked-open channel conformation [45].

For CFTR mutants containing only the NBD1-stabilizing mutations, T_m^trp of full length CFTR correlated strongly with the T_m^cal of isolated NBD1 obtained by DSC (Fig. 5B, solid line). The same positive correlation was observed in the presence of the additional increment in CFTR stability from H1402S in NBD2 (Fig. 5B, gray line). The data are consistent with NBD1 stability being an important determinant of CFTR stability, with other regions of the protein also contributing to protein stability. T_m^func also correlated strongly with T_m^Trp (Fig. 5C, solid line) with overall lower values seen for T_m^func, because the longer incubation time (30 min) at each temperature had the same effect as an extremely slow heating rate, which results in lower apparent T_m [59]. However, among the superstabilized CFTR variants, we noted a difference in rank order of stability between T_m^trp (7SS < 5SS < 6SS) and T_m^func (5SS < 7SS = 6SS, Table 1), suggesting that loss of enzymatic activity might occur prior to cooperative unfolding.

2.9. Stabilized CFTR variants retain channel function

CFTR is an ATP-gated anion channel whose function is to transport chloride and bicarbonate across the membrane [3]. Thus, the most decisive functional assessment is anion transport [3,60,61]. To analyze whether and how the properties of the stabilized CFTR channels might have been altered, we recorded single channels in planar lipid bilayers [62]. Recordings of selected CFTR variants, driven by the same concentration of 0.3 mM ATP and at a temperature of 33 °C comparable to conditions of the ATPase assays, and ion channels parameters are shown on Fig. 6A. WT CFTR containing the three SUMO, 901-Flag, and EGFP tags expressed in CHO cells exhibited single channel conductance γ of 12.5 ± 0.3 pS (n = 6) with an open probability P_O of 0.29 ± 0.05 (n = 6) (Fig. 6A upper trace) similar to WT CFTR containing the same tags expressed in HEK cells [56]. Introducing the ΔRI/2PT/M470V-CFTR variant containing the M470V polymorphism (Fig. 6A, middle trace) had a single channel conductance γ of 12.7 ± 0.4 pS (n = 5) and probability P_O of 0.34 ± 0.08 (n = 5) not statistically different from WT (Table 2). This CFTR variant showed a somewhat reduced the dwell-time for the mean open (τ_o = 114 ± ms) and the mean closed time (τ_c = 172 ms), and so shortened τ_cycle to 286 ms compared to WT CFTR (Fig. 6B). The most stable 6SS-CFTR variant, expressed in CHO cells, also displayed gating properties very similar to the WT CFTR (γ of 12.6 ± 0.4 pS (n = 5), P_O of 0.27 ± 0.03), except that both the dwell-times for the mean open (τ_o = 318 ms) and closed times (τ_c = 709 ms) were somewhat prolonged, leading to a prolonged gating cycle duration (Fig. 6B).

Fig. 6.

Single channel recordings of stabilized CFTR mutants. A) Examples of single-channel current recordings in membranes isolated from CHO cells expressing either WT CFTR (D507, top trace), or CFTR containing four (ΔRI/2PT/M470V, D851, middle trace) or six stabilizing mutations (6SS, D1013, bottom trace). Channels were recorded at 33 °C with 0.3 mM ATP after phosphorylation with protein kinase A. All-point histograms (plots at left) were used to calculate single channel conductance γ as a distance between peaks, and Po as a ratio of the area under the peals. The values of mean open (τo) and mean closed times (τc) for the particular examples shown were calculated from the dwell-time histograms (plots at the right). Experiments were repeated 6 times with a total recording time of 48 min for WT, 5 times and 42 min for 2PT/M470V, or 4 times and 52 min for 6SS-CFTR. Averages and SEM of unitary channel parameters are given in Table 2. B) Bar graph of gating cycle duration (τ_o + τ_c). C) Single channel function recorded for 1 min after 9 min exposure at elevated temperature. Under these conditions, the WT ceased gating and remained closed, while 6SS-CFTR retained stable gating.

Table 2.

Single channel conductance parameters

	Conductance (pSi)	Po	n	P value	τo (ms)	τc (ms)	τo + τc (ms)	P value
WT CFTR	12.5 ± 0.3	0.29 ± 0.05	6		162 ± 11	365 ± 23	527 ± 25
ΔRI/2PT/M470V	12.7 ± 0.4	0.34 ± 0.08	5	ns	114 ± 10	172 ± 21	286 ± 23	<0.0001
6SS-CFTR	12.6 ± 0.4	0.27 ± 0.03	4	ns	318 ± 32	709 ± 47	1027 ± 57	<0.0001

To assess the thermal sensitivity, single channels were recorded at 45 °C, a temperature at which we previously demonstrated that WT CFTR was rapidly inactivated within 2 to 3 min, the time required to reach uniform temperature across all compartments of the bilayer apparatus as previously described [45]. The top panel in Fig. 6C shows a trace of WT-CFTR microsomal membranes fused with the bilayer after a 5 min incubation at 45 °C which virtually abolished channel gating. In contrast, 6SS-CFTR actively gated for at least 20 min. The stable channel activity of 6SS-CFTR in lipid membranes thus corroborates the increased thermal (T_m^Trp) and functional stability (T_m^func) of this protein in detergent solution.

2.10. Select NBD2-stabilizing mutations did not stabilize full-length CFTR

We also explored mutations in NBD2 (S1359A, and the quadruplet mutations S1255L/K1334G/S1359A/Q1411D that were recently discovered to significantly improve NBD2 thermal stability [63]). Introducing these NBD2-stabilizing mutations into either ΔRI/2PT (cell line D805, D742) or the 7SS (D804, D743) had no effect on T_m^Trp (Fig. 5B, green crosses). These mutants exhibited slower but (in some cases) still measurable rates of ATP hydrolysis (Fig. 3, green). S1359A had no effect on T_m^func in the context of ΔRI/2PT, but substantially reduced T_m^func in the context of 7SS (Table 1), suggesting the loss of a stabilizing interaction in the latter case. The data are in agreement with expression and processing data reported previously [63], and suggest that full length CFTR thermal stability is not limited by the stability of the NBD2 domain, but instead contacts with NBD2 residues may play a role.

2.11. Molecular dynamics simulations suggest lower fluctuations in the stabilized 6SS-NBD1

Four of the NBD1 residues mutated in this study are located in structured positions (alpha-helices, turns, and beta-sheets), and seven are in less structured positions (e.g. coil regions) (Fig. 1). We attempted to predict the stabilizing effect of single mutations as well as for the combination of mutations present in the 6SS construct using descriptors such as Solvent Accessible Surface Area (SASA), B-factor and calculated ΔΔG values, but correlations with our experimentally determined data were weak (Supplemental Fig. S6). On the other hand, we previously demonstrated a correlation between root mean squared fluctuations (RMSF) profiles derived from replica exchange molecular dynamics (REMD) simulations and thermal stabilities for a set of NBD1 mutants [27]. To test whether this correlation extends to the present set of mutations, we conducted REMD simulations for the most thermally stable NBD1 construct in this work, 6SS (Fig. 7). In these simulations the mutant construct (red) exhibited reduced fluctuations in several regions, particularly in the 528–550 region containing A534P and I539T (in the SDR). In addition, small reductions in fluctuation were consistently observed (n = 3) at positions 470 and 495 (Q-loop) corresponding to the two most stabilizing mutations (M470V, ΔT_m^cal = 4.98 °C; S495P, ΔT_m^cal = 6.1 °C). Of note, visual inspection of the crystal structure suggests that Val sits in this position more “comfortably” than Met, i.e. with less steric hindrance. Lastly, we observed no differences in fluctuation for the third most stabilizing mutation, R555K. However, for this position we compared the average distance between the side chain center of mass and that of residue D529 in the MD trajectories. We found this distance to be smaller for K555 in 6SS (0.51 ± 0.05 nm) than for the R555 in WT (0.57 ± 0.05 nm) indicating better electrostatic interactions in the mutant, in agreement with experimentally observed stabilization. Thus, RMSF again proved to be a reliable predictor of experimentally observed stabilizing effects when applied to this new set of NBD1 mutations.

Fig. 7.

REMD simulations reveal reduced fluctuations in 6SS-NBD1. Root mean square fluctuation profiles for ΔRI-NBD1 (black, three repeats) and 6SS-NBD1 (red, three repeats). Residues 405 to 436 are absent in these constructs. Inset: expansion of residues 460 to 505. The close similarity between the RMSF profiles emerging from different simulations is indicative of simulation convergence.

3. Discussion

Protein stability, or the lack thereof, represents a substantial challenge to biochemical and biophysical analyses of many proteins, especially membrane proteins that require solubilization in detergents that may destabilize extra- as well as intramembranous domains [54,64]. The multidomain complexity and highly dynamic nature of the channel, together with CFTR’s intrinsic instability, particularly in detergents, make its study especially challenging [35,45,54,57].

Recent work applying similar approaches to unfolding of CFTR domains [23,24,37,55,63] significantly advanced our mechanistic understanding of the disease-causing defect of ΔF508. In the current study, we used DSC to analyze in detail the ability of site-specific substitutions to increase thermal stability (T_m^cal) of ΔRI-NBD1. We systematically and iteratively combined mutations to demonstrate progressive stabilization of ΔRI/2PT by introducing M470V, S495P and R555K mutations, to ultimately yield an NBD1 domain with highly improved stability. Indeed, the most stable combinations 6SS-NBD1 and 7SS-NBD1 exhibited T_m^cal values of 74 °C and 71 °C, respectively, representing an increase of more than 15 °C over ΔRI-NBD1, and outperforming previously published ΔRI/2PT/Teem NBD1 [37,43] by 7 to10 °C (Fig. 2).

For the set of single site stabilizing mutations considered in this work, we did not observe a strong correlation between the experimental T_m^cal values and either local structural parameters (B-factors, or solvent accessible surface areas) or FoldX-calculated ΔΔG values. On the other hand, REMD simulations suggested reduced fluctuations at specific positions in the more flexible regions of stabilized 6SS-NBD1, providing a strategy to engineer further mutations in CFTR to generate a protein with even greater stability and potentially paving the way towards NBD1 (and consequently CFTR) stabilization via drug like molecules. Of note, MD simulations have recently been used to probe the unfolding of four NBD1 constructs pointing to several residues (e.g., S492, Q493, T465, A554-Y563, L571, S573, P574, F575, K584, I586) whose mutation was suggested to enhance the folding efficiency of ΔF508-NBD1 [65]. Of these, the S492P and R555K as well as the close-by F494N and S495P mutations were shown in the present work to stabilize the WT domain.

Several earlier studies have suggested that NBD1 structural stability is one of the determining factors in CFTR stability [11,18,[23], [24], [25],28,29,37,54]. In particular, the work by Mendoza et al. and Rabeh et al. recognized the correlation of NBD1 in vitro stabilization with the improved de novo folding efficiency and expression of the full length CFTR [28,29]. Results with the current, more extensive panel of CFTR mutants fully support this idea. Increased thermal stability of the NBD1 domain (T_m^cal) correlated directly with improved thermal (T_m^trp) and functional (T_m^func) stability of full length CFTR (Fig. 5). The combinations of mutations that were best for stabilizing NBD1, 6SS and 7SS, increased T_m^func from 22 °C for WT CFTR to 36 °C (Table 1). ATP binding and hydrolysis in CFTR, as in all ABC transporters, depends on the two NBDs interacting in a head-to-tail dimer to sandwich two MgATP molecules between the Walker A/B motifs of one NBD and the Signature motif of the other [32,66,67]. CFTR NBD1 has a degenerate Walker B motif and lacks the switch histidine (equivalent to H1402), making this composite site inactive [17,68]. Therefore, ATP hydrolysis is contingent on the second composite site and requires key catalytic residues contributed by NBD2. Our data suggest that progressive structural stabilization of NBD1 raises T_m^func by concomitantly protecting the hydrolytic NBD2 domain against thermal inactivation, and is consistent with this requirement for interdomain contacts at the composite site where hydrolysis takes place.

In a typical ABC transporter, ATP binding induces NBD dimerization, while ATP hydrolysis is thought to dissociate the dimer [2,3,17]. A current view is that NBD dimerization and dissociation transmit conformational changes via the ICLs to the transmembrane domains to gate the channel [32], [33], [34],69]. Interestingly, we observed large effects on T_m^func and T_m^trp by mutations in the Q-loop (S492P, S495P) thought to transmit signals to the transmembrane domains (for example 5SS, Table 1) [70]. Potentially, structural stabilization of NBD1 may also enhance these connections to the ICLs and so contribute to stabilization of the full-length CFTR protein. Thus, T_m^func may not only track the stability of the NBDs but also the ICL connections. Importantly, the thermally stabilized 6SS-CFTR channel was active at 33 °C, and remained active for prolonged times at elevated temperatures as high as 45 °C where WT CFTR rapidly lost function.

Conformational state is an additional component of structural stability. T_m^Trp detected the stabilizing effect of the switch histidine mutation H1402S, which abolished ATP hydrolysis and would be presumed, as in other ABC transporters, to trap non-hydrolyzed ATP in an NBD1-NBD2 dimer conformation [71], and to render an open channel conformation [45]. A similar +5 °C shift in T_m^Trp due to H1402S was consistently observed in ΔRI/2PT/H1402S, ΔRI/2PT/M470V/H1402S, and 6SS/H1402S, suggesting that the shift in conformational population enhanced structural stability of the protein. Indeed, a similar stabilizing effect was observed in P-glycoprotein with equivalent mutations that promoted the occluded NBD dimer conformation with trapped MgATP [72,73]. A large body of literature reports on conformational stabilization due to specific mutations and combinations of mutations in G-protein coupled receptors (GPCRs). In those cases, conformational stabilization was essential to obtain crystal structures in different states [74], [75], [76], [77]. Thermal unfolding studies with P-glycoprotein and other membrane proteins have shown that unfolding of α-helical transmembrane domains is unlikely to be detected by Trp fluorescence over the temperature range of the present study [55,73,78,79]. In CFTR, 11 of the 23 intrinsic Trp residues are situated in transmembrane domains. However, several Trps in NBD1 (W496, W401) and its connecting ICL4 loop (W1063), as well as in NBD2 (W1274, W1282, W1310, W1316) and its connecting ICL2 loop (W278) are buried within the domains or at the domain interface and eligible to report tertiary unfolding of these cytoplasmic domains [58,80]. Because fluorescence emission is an average of all Trps in different local environments, the detected single transition for CFTR suggests that unfolding of these domains is highly cooperative. We feel the most straightforward interpretation of the Trp fluorescence data for CFTR is that it represents unfolding of cytosplasmic portions of the protein in a way that is responsive to protein conformational state.

In mammalian cells, the new stabilized variants of CFTR all exhibited strong localization of mature glycosylated protein to the cell surface, were purifiable under mild conditions, and demonstrated significantly improved stability. In conclusion, CFTR variants with a highly stabilized NBD1 domain, imparting greatly enhanced structural and functional stability, together with conformational stabilization by the H1402S mutation, will facilitate future biochemical and biophysical studies, and may open the door to higher resolution structures by Cryo-EM or crystallography. The stabilized CFTR variants may enable structural studies of folding mutations like ΔF508, for which protein production and purification have been problematic. They might also facilitate other ongoing lines of investigation, such as domain assembly, the mechanisms of gating and channel regulation, and insights into the natural mutations that cause cystic fibrosis.

4. Methods

4.1. NBD1 purification

CFTR NBD1 (residues 387–646, [Δ405–436]) containing the wild type sequence with the M470 polymorphism, referred herein as ΔRI-NBD1, was expressed in E. coli, and the protein purified in 150 mM NaCl, 20 mM HEPES pH 7.5, 10% glycerol, 10% ethylene glycol, 1 mM tris-(2-carboxymethyl) phosphine, 2 mM ATP, 3 mM MgCl₂ as previously described [21,23,24,54]. Proteins were >98% pure as judged by Coomassie Blue staining of SDS-PAGE gels (Supplemental Fig. S2), showed no evidence of aggregation and ran as monomers during gel filtration. Protein concentration was determined with the Pierce BCA 660 nm assay in microtiter plate format, using Bacillus subtilis NAD synthetase as a standard. Proteins were stored at −80 °C.

4.2. Differential scanning calorimetry

Differential scanning calorimetry (DSC) was carried out on the VP-Capillary DSC System (MicroCal, Malvern Instruments), in 0.13 ml cells with 0.5 mg/ml proteins, at a heating rate of 2 °C/min, and an external pressure of 2.0 atm to prevent possible solution degassing upon heating. DSC data were analyzed with the MicroCal Origin 7.0 software (OriginLab Corp.), from which the unfolding temperature (T_m^Cal) was obtained. The average T_m^Cal of ΔRI-NBD1 was 56.7 ± 0.3 °C from 22 DSC runs, a stabilized Tm approximately 14°C higher than that of native NBD1 containing the RI sequence. We previously showed that introducing stabilizing mutations (or the destabilizing ΔF508) into different isolated NBD1 backgrounds (ΔRI, or with RI) affected the ΔTm^cal to a similar degree (see [23,24]). DSC for each mutant was repeated 2–3 times, with T_m^Cal ranges among replicates ≤0.6 °C. A heating rate of 1 °C/min gave slightly lower Tm values by about 2 °C, i.e. the ΔRI-NBD1 unfolded at 54.4 ± 0.1 °C, while the 6SS-NBD1 unfolded at 71.9 ± 0.1 °C; nevertheless the nominal difference “ΔTm” was the same.

4.3. Expression of full-length CFTR

We previously described the D165 HEK293 cell line for protein expression, wherein human CFTR was modified with His₁₀-SUMO* and ⁹⁰¹Flag affinity purification tags and C-terminally fused with the enhanced green fluorescent protein (EGFP) in a lentiviral vector [56]. The recombinant CFTR protein with a molecular mass of 212 kDa is referred to herein as WT. We had also generated a similar CFTR construct with the stabilizing mutations ΔRI′/2PT (kind gift of Jack Riordan [45,81]) containing a deletion in RI′ encoding residues 404–435, and the NBD1 mutations S492P, A534P, and I539T (2PT) [49]. Building on this ΔRI′/2PT construct, single substitutions M470V, S495P, or combinations such as the 7SS variant (ΔRI′/2PT/F494N/S495P/G550E/R555K) were added by PCR mutagenesis using Q5 polymerase (New England Biolabs). We also generated ΔRI-CFTR matching the ΔRI-NBD1 deletion Δ405–436 (one amino acid difference), and added the 2PT mutations to build ΔRI/2PT. The latter served to build the 5SS (ΔRI/2PT/S495P/M470V) and 6SS (ΔRI/2PT/S495P/M470V/R555K) constructs. The open reading frames were placed under transcriptional control of either the tetracycline response element TRE-tight (TRE.t) or TRE second generation (TRE.2), as specified in Table S1. As previously described [82], the vectors were packaged, pseudotyped with vesicular stomatitis virus G protein, and used to transduce CHO-S cells (Invitrogen) that had been modified to constitutively express the conventional reverse tet transactivator (rtTA) or 3G–matched rtTA, as specified in Table S1 [83]. Transduced cells were adapted to suspension culture, and CFTR expression was induced with 1 μg/ml doxycycline at 37 °C.

To assess CFTR expression at the cell surface, the cells were live-stained 24 h after induction using mouse anti-Flag monoclonal antibody (Sigma F3165). Surface CFTR staining was quantified by flow cytometric analysis of at least 5000 cells. All CFTR-expressing cell lines were analyzed in parallel, and the experiment was repeated at least four times. LinearFlow® fluorescently labeled polystyrene beads (Molecular Probes) were included in each experiment to ensure identical conditions for flow cytometry data collection. For each experiment, median fluorescent intensity (MFI) of the antiFlag-stained cell population was determined using FlowJo software (FlowJo, LLC).

4.4. Purification of CFTR

Microsomal membrane fractions were prepared from doxycyclin-induced cells expressing recombinant CFTR constructs described above. For ATPase assays, proteins were PKA-phosphorylated and purified to homogeneity using affinity chromatography on Ni-NTA and anti-Flag resins in the presence of 0.05% decyl maltose neopentyl glycol (MNG10) as described [57]. Purified mutant proteins were resolved on 8% polyacrylamide SDS-gels and quantitated by densitometry of in-gel EGFP fluorescence with bracketing amounts of P-glycoprotein-EGFP as external standard [84]. Examples of human WT CFTR purified from HEK or CHO are shown in supplemental Fig. S5A. For tryptophan-unfolding experiments, purified proteins of similar quality were obtained in a one-step purification using anti-Flag resin in the presence of 0.05% 3α,7α,12α-tri-((O-ß-D-maltopyranosyl)-ethyloxy)-cholane (FA-4, Avanti Polar Lipids). The progress of the purification is shown in Supplemental Fig. S5B.

4.5. ATPase activity

Purified CFTR was supplemented with 0.4 mg/ml phospholipid (POPE/brain PS/egg PC/cholesterol 5:3:1:1 w/w, and destabilized with 1/4th wt. C₁₂E₈), and ATP hydrolysis was measured with 0.3 mM α-[³²P]-ATP, 1.5 mM MgCl₂, pH 7.5, at 33 °C for 2 h using thin-layer chromatography for detection of hydrolyzed ADP as described [57]. Background measured with reagent blanks containing all components except CFTR was subtracted. Mutant combinations with H1402S lacked detectable ATPase activity and served as negative controls, confirming that the assay detects CFTR-dependent ATP hydrolysis activity.

4.6. Functional Tm, the threshold temperature for inactivation of ATP hydrolysis

The approach has been previously described [49]. Briefly, aliquots of CFTR (about 0.1 μg in 20 μl) were pretreated 30 min at varying temperatures, then supplemented with phospholipid and assayed for ATPase activity remaining at 33 °C under the conditions stated above. Results for enzymatic activity remaining vs. pretreatment temperature were fit to a three-parameter sigmoidal equation in SigmaPlot. The derived inflection point gives T_m^func. Values were compared using Student’s two-tailed t-test.

4.7. Protein unfolding monitored by intrinsic fluorescence

Effects of mutations on thermal stability of full-length CFTR were compared using Trp fluorescence, which required much less protein than DSC. T_m^Trp was determined with 30 μl of 1–2 mg/ml protein in 50 mM Tris HCl pH 7.5, 150 mM NaCl, 1 mM dithiothreitol, 10% glycerol, 3 mM MgCl₂, 2 mM ATP, and 0.05% w/v FA-4, in a PolarStar Optima fluorescence plate reader (BMG Labtech, USA), in 30 μl volumes overlaid with pure mineral oil (Fisher Scientific) in a 384 well plate. Samples were heated at 1 °C/min using an in-house fabricated heating block while monitoring fluorescence at λex = 290 ± 5 nm, λem = 330 ± 5 nm. CFTR unfolding was irreversible as indicated by the lack of an unfolding transition in rescans of the samples. T_m^Trp is defined as the temperature corresponding to 50% maximal change in fluorescence. Each mutant was determined 2–3 times, and replicates varied within ≤1 °C. The minimum protein concentration required to detect the transition was 1 mg/ml. The purification tags, SUMO* and GFP, by themselves didn’t exhibited an unfolding transition when heated in the same temperature range. Therefore, the observed changes in fluorescence were attributed to CFTR Trps.

4.8. Single-channel analysis

Microsomal membrane fraction, prepared from CHO cells expressing the stabilized CFTR variants, were fused to planar lipid bilayers, and single-channel currents were recorded following published procedures [45,56,60,62]. Briefly, planar lipid bilayers were formed with a 3:1 mixture of 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine and 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoserine (Avanti Polar Lipids). The lipid bilayer separated 1.0 ml of solution (cis side) from 5 ml solution (trans side). Both chambers were magnetically stirred and thermally insulated. CFTR ion channels were transferred into the preformed lipid bilayer by spontaneous fusion of membrane vesicles containing naturally folded CFTR constructs. To maintain uniform orientation and functional activity of CFTR channels transferred into the bilayer, MgATP and Protein Kinase A were added in the cis compartment only. Single channel currents were measured at −75 mV in symmetrical salt solution (300 mM Tris–HCl, pH 7.2, 3 mM MgCl2 and 1 mM EGTA) under voltage-clamp conditions using an Axopatch 200B amplifier (Molecular Devices). The membrane voltage potential of −75 mV is the difference between cis and trans (ground) compartments. The output signal was filtered with an 8-pole Bessel low-pass filter LPBF-48DG (NPI Electronic) with cut-off frequency of 50 Hz to eliminate all closures lasting less than 20 ms including intraburst closings with characteristic life-time less than 5 ms. For kinetic analysis, the signal was digitized by Digidata 1322 (Molecular Device) with a sampling rate of 500 Hz and analyzed using pCLAMP 9.2 (Molecular Device) software. Dwell-time histograms for the open and closed states were plotted in the logarithmic binning mode and fitted by a single exponential function. The validity of the reduced two states kinetic model is evident from the single exponential fit seen for both closed and open dwell time histograms. Origin Pro 7.5 (OriginLab) software was used to fit all-points histograms by multi-peak Gaussians. Single-channel current was defined as the distance between peaks on the fitting curve and used for the calculation of the single-channel conductance (□). The probability of the single channel being open (Po) was calculated as a ratio of the area under the peak for the open state to the total area under both peaks on the fitting curve.

Thermal stability of the channels was measured at 45 °C; temperatures above 50 °C affect the fluidity of the lipids used and compromise the integrity of the planar bilayer [45]. Heating and temperature control were established by a temperature control system TC2BIP (Cell Micro Controls, Norfolk, VA, USA) with heating element covering the “trans” compartment outer side surfaces and bottom surface. Because of the bulky “trans” compartment it takes about 2–3 min to achieve a uniform temperature distribution across all compartments after the first indication of the expected temperature by the local temperature sensor in the chamber. We consider the thermal stability of the CFTR constructs in terms of its ability to support single channel function with stable open state conductance for the next 5 min after initial 5 min incubation at the temperature of interest [45].

4.9. Replica-exchange molecular dynamics

REMD simulations were performed for human 6SS-NBD1 starting from the X-ray structure of ΔRI (ΔRI-NBD1 comprising residues 387–646 excluding a deletion of residues 405–436; PDB code 2PZE). Importantly, this is the same contract as used for the experimental work. Results for the WT construct were taken from [27], and the same simulation protocol was followed. Prior to simulation, the 6SS mutations were introduced into WT-NBD1 using the mutation and the side-chain refinement protocols as implemented in Discovery Studio (Dassault Systèmes BIOVIA, Discovery studio modeling environment, Release 4.5, San Diego, 2015) and the resulting construct was processed by the Prepare Protein protocol as implemented in Discovery Studio to set the correct protonation states for all residues at pH = 7.0. All simulations were performed using the Gromacs Molecular Dynamics package version 4.5.5 [85,86] with the OPLS/AA force field [87]. Each REMD simulation consisted of 32 replicas with each replica running for 10 ns, so that the construct was simulated for a total of 960 ns (32 replicas × 10 ns × 3 times). Simulations covered a temperature range of 300 to 349.38 K (26.85 to 76.23 °C), a range selected to give the same acceptance probability between all adjacent pairs over the entire temperature range based on known energy distributions of solvated proteins [88]. Figs. S8, S9, Table S2 and the accompanying text provide additional details on simulation setup and convergence.

Highlights.

• CFTR presents a challenge to obtain monodisperse, stable protein.

• Strategic combinations of mutations were introduced to stabilize NBD1 domain.

• NBD1 stabilization improved structural and functional stability of full length CFTR.

• Thermal unfolding showed Tm could be shifted >20° above wild type.

• Switch histidine mutation stabilized an additional 5°.

Abbreviations

CF

cystic fibrosis

CFTR

cystic fibrosis transmembrane conductance regulator

cryo-EM

cryoelectron microscopy

EGFP

enhanced green fluorescent protein

NBD

nucleotide binding domain

RI

regulatory insertion (405–436)