CFTR (ABCC7) is a phospho-regulated chloride channel that is found in the apical membranes of epithelial cells, is gated by ATP and the activity of the protein is crucial in the homeostasis of the extracellular liquid layer in many organs [Annu. Rev. Biochem. (2008) 77, 701–726; Science (1989) 245, 1066–1073]. Mutations in CFTR cause the inherited disease cystic fibrosis (CF), the most common inherited condition in humans of European descent [Science (1989) 245, 1066–1073; Pflugers Arch. (2007) 453, 555–567]. The structural basis of CF will be discussed in this article.
CFTR is a member of the ATP-binding cassette (ABC) family of proteins [1–3]. As such, CFTR has the standard four-domain structure consisting of two transmembrane domains and two soluble nucleotide-binding domains (NBDs) . The ABC family members are predominantly active transporters. In accord with the Jardetzky model for transporter action , the ATP is employed to drive the transition of the transmembrane domains from an inward-facing to an outward-facing conformation (or vice versa for ABC importers). However, for CFTR it is thought that ATP drives a conformational change that opens a channel in the transmembrane domains, through which chloride ions can diffuse passively . CFTR is the only member of the ABC family that has this channel-type functionality, although some members do act as regulatory switches for other channels [7,8].
Initial CFTR structural data
Initial structural studies were carried out over a decade ago on the soluble CFTR domains studied in isolation from the transmembrane portions [9,10]. In these early studies, the first soluble ATP-binding domain (nucleotide-binding domain 1, NBD1) containing the most common CF-causing mutation (F508 deletion) was crystallised in the presence and absence of this phenylalanine residue. The F508 deletion was found to have almost no effect on the structure of the isolated domain apart from some local rearrangements around the F508 position. This was unexpected: prior studies of the mutation in various cells had been interpreted in terms of its protein folding, with the F508del mutation discussed as a misfolding mutation that led to its recognition by the endoplasmic reticulum quality control machinery [11–14]. However, crystallisation of the domain, as well as the second NBD (NBD2), required an exhaustive search for mutations that promoted solubility and stability. In the case of NBD1, these same mutations were later demonstrated to partially rescue the defect caused by F508 deletion . Isolated domains were also studied by NMR although, as above, mutations that increased the protein solubility were needed to assign the NMR resonances; however, later studies of the effects of F508 deletion were with NBDs with background wild-type sequences [16–19].
NMR was also employed to study the interactions of the third soluble ‘domain’ in CFTR, the 200 residue-long Regulatory-region or R-region [18,20]. These interactions were studied with several potential partner proteins, including the two soluble NBDs. Whether this region was intrinsically disordered (as suggested by the NMR data and predictions based on its amino acid sequence), or a structured domain, or some intermediate between the two, remained a moot point until recently. When cryo-EM data for single CFTR particles emerged [21–24], the data implied that the R-region existed in an intermediate state in the absence of phosphorylation — i.e. it displayed weak, but localised, Coulomb-scattering density in the cryo-EM maps. This weak density was observed between the two NBDs and in one map it protruded into the aqueous compartment generated by the two transmembrane domains in their inward-facing configuration (, see Figure 1). Upon phosphorylation, the weak density for the R-region could no longer be observed [22,24], implying that the addition of ∼7–9 negative charges to Serine and Threonine residues in this region caused its dissociation from the CFTR core domains and resulted in an increase in disorder throughout this portion of the protein. Such a phosphorylation dependence on R-region/NBD association was observed by the NMR interaction studies using the isolated domains .
Context of CF-causing mutations and the importance of surface charges.
New structural data
There have been several cryo-EM-derived structures of CFTR published recently, with the studies based on zebrafish, chicken and human versions of CFTR expressed in mammalian cells [21–24]. Reflecting the major improvement in cryo-EM technology, all these structures contain structural data with a resolution considerably better than was previously possible . However, there are features of the new structures that are not readily explained by prior models of CFTR structure–function relationships [6,26]. For example, dephosphorylated CFTR (studied without ATP) and phosphorylated CFTR (studied with ATP) both displayed an inward-facing configuration in the structures of chicken CFTR described by Fay et al. . Minimal changes in conformation were reported upon phosphorylation and nucleotide binding, a major departure from the prior models. However, zebrafish CFTR in the dephosphorylated/ATP-free condition was shown by cryo-EM to be inward-facing , while after phosphorylation and addition of ATP there was a major conformational shift to the outward-facing state . This latter conformation was obtained with a version of the protein where the catalytic Walker B glutamate residue (E1372) was mutated to glutamine. This mutation's effects can be rationalised in terms of prior models in that prevention of ATP hydrolysis should prolong channel opening and the outward-facing state. However, the phosphorylated/ATP bound chicken CFTR structure which was in the inward-facing state was also obtained with mutations that stabilised the open state of the channel . Hence the new structures may demand a new model of CFTR structure–function relationships that can accommodate all the current structural data.
The multiple cryo-EM-derived structures for CFTR at medium resolution (3–9 Å resolution) represent structures equivalent to the wild-type protein as well as structures with mutations that were incorporated to stabilise the protein or to favour a particular conformation [21–25]. The structural effects of mutations that are cystic fibrosis (CF)-causing in a significant number of patients (such as F508del) still remain to be studied using cryo-EM, although small-angle X-ray scattering (SAXS) has been employed to study this most prevalent disease-causing mutation . The structural data so far imply that F508 deletion causes significant changes in the overall structure of the purified full-length protein as detected by SAXS. This is in accord with the biochemical data that show that F508 deletion has a major effect on the stability of the protein [28,29]. F508 sits in a buried position in NBD1 at its interface with the 2nd intracellular loop of transmembrane domain 2 (Figure 1). This loop links transmembrane helices 10 and 11, which cross over from the opposite side of the molecule in a domain swap-type arrangement. Hinge-like movements of these transmembrane helices are closely associated with the transition from inward- to outward-facing conformations and this may explain why F508del CFTR has poor channel activity. However, the position of F508 at a crucial and exquisitely conserved domain–domain interface may explain why it has a large effect on the overall stability of the structure of the protein (Figure 1A).
The second most common missense mutation that causes CF is G551D, which results in a predominantly closed channel. The open probability of fully phosphorylated G551D CFTR channels in the presence of milliMolar ATP concentrations is 10-fold lower (Po = 0.04) compared with the WT channel (Po = 0.40) . G551 is part of the ‘signature’ sequence of NBD1 which jointly binds ATP in association with the Walker A and Walker B residues from NBD2 (Figure 1B). The signature sequence is a characteristic of all ABC proteins [30–32]. The incorporation of an additional negative charge at this position as well as the –CH2–COO− side-chain of aspartate is likely to impede ATP binding and make the formation of the NBD1–NBD2 sandwich dimer much less likely. Not only are the negatively charged phosphate groups of the ATP in close proximity to G551, but also the E1371 and D1370 residues in the Walker B region of NBD2. A drug to treat G551D patients was developed over the last decade (Ivacaftor [33,34]) and this therapeutic compound greatly increases the open probability of the channel to close to WT levels . Ivacaftor has been shown to be effective at increasing the channel function of many, but not all CF-causing mutations studied in vitro , and these mutations are distributed throughout the 3D structure of the protein. Of note is that many of the mutations that Ivacaftor can work on involve a change in the charge of the mutated residue  akin to G551D. Hence elucidation of the mode of action of the drug is challenging but we can infer that charge effects appear to be implicated. Examination of the distribution of charged residues across the water-exposed surface of CFTR allows the calculation of the Coulombic surface potential, which is displayed using colour-coding in Figure 1C. This shows that the inner surfaces of the intracytoplasmic loops are strongly positively charged, a feature of significance for the above discussion since these surfaces must be pressed together in order to reach the outward-facing state and hence open the channel.
Integration of the biochemical and structural data for CFTR
We propose a new model to explain these observations (Figure 2). This model assumes that nucleotide will remain bound to CFTR for most of its lifetime in the cell, with a slow rate of ATP hydrolysis. Nucleotide-free CFTR will never exist unless ATP levels drop in the cell due to cell death, but note that this state can be produced and studied experimentally (e.g. in inside-out membrane patches ). For CFTR activity to be regulated, phosphorylation by PKA and dephosphorylation by protein phosphatase will be needed (step 1). Once phosphorylation of the R-region occurs, then this region can dissociate from its interfering position between the NBDs and take up a more disordered structure. In this new situation, inward-facing CFTR with no channel activity can alternate with an outward-facing conformation that can form a channel for chloride ions (step 2). Production of the outward-facing state may have a significant energy barrier, possibly because of the long-range influence of electrostatic forces due to the positively charged residues that cluster in the inward-facing surface of the intracytoplasmic loops as well as the negative charge of the ATP molecules in the NBDs (Figure 1B,C). The ‘heavy-lifting’ needed to produce the outward-facing state leads to the naming of this conceptualisation as the ‘Grunt model’. Once the outward-facing state is formed, short-range interactions (H-bonding, van der Waals) are likely to dominate and stabilise it. Mutations such as G551D that reduce the open probability of the channel are likely to affect the equilibrium at step 2 as well as binding of ATP at step 5, and we predict that Ivacaftor also acts at step 2, but in this case it will promote the outward-facing state. We expect the drug will be acting in a similar fashion to ABC transporter allocrites (transported substrates), which are known to stimulate ATPase activity when they bind and hence may consequently lead to the outward-facing state. For exporters in the same family as CFTR, allocrites bind to the inward-facing state with higher affinity  and for some transporters, the binding sites have been characterised by mutagenesis and structural studies. For example in P-glycoprotein (ABCB1), which can transport xenobiotics and drugs out of the cell, there are two drug-binding sites at the apex of the inner vestibule formed by the two transmembrane domains in the inward-facing conformation [37,38]. It is known that Ivacaftor is one of the many transported substrates of P-glycoprotein .
‘Grunt’ model for CFTR structure–function relationships.
ATP hydrolysis is likely to occur at step 3 and the model implies that catalysis requires the NBDs to be dimerised, as they exist in the outward-facing conformation, for this hydrolysis to be efficient. Mutations that affect this step (such as those in the Walker B region) may have some influence on step 2 as well. Indeed, mutation E1371Q in CFTR was used to generate a sufficient proportion of outward-facing CFTR molecules to allow that state to be studied by single-particle cryo-EM . Steps 4 and 5 are typical of the subset of ABC transporter proteins where one of the ATP-binding sites has evolved to be catalytically inactive. There are many ABC transporters that show this feature, including CFTR where the site formed between the NBD1 Walker motifs and the NBD2 signature motif is non-consensus . Mutations that affect these steps may also influence the balance at step 2 for the same reasons as above.
Both inward-facing and outward-facing conformations of CFTR in the presence of ATP have recently been reported for phosphorylated protein [18,22]. The inward-facing conformation was also observed in a study of CFTR interacting with the NHERF1-PDZ1 domain, which is known to bind at the CFTR C-terminus . Moreover, a detailed study of negatively stained single particles of two purified ABC transporters found a spectrum of conformations in the presence of ATP, with 1–2% in the outward-facing configuration even in the absence of ATP . Similarly, structure-based calculations suggest that the free energy changes needed to affect NBD dimerisation may be relatively small , and surprisingly, CFTR channel activity can be detected in constructs completely lacking NBD2 . Hence there seems to be a remarkable plasticity in ABC protein conformations and structures that are emerging. The ‘Grunt’ model we propose seems to be consistent with the (otherwise perplexing) diversity of CFTR structures as well as observations concerning channel-disrupting and ATPase-disrupting mutations  and the broad mutation specificity of potentiators such as Ivacaftor. Future structural studies will highlight the effects of disease-causing mutations and the most significant of these will be structures that include information about the binding site(s) of the new therapeutics.
R.C.F. wrote the paper. X.M., J.C., A.D.C. and E.R.M. discussed and edited the manuscript and contributed to the concepts therein.
The authors acknowledge the Cystic Fibrosis Foundation [CFF FORD13XX0] and the Cystic Fibrosis Trust [F508del CFTR SRC] for funding.
We thank Alessandro Barbieri, Nopnithi Thonghin and Talha Shafi for useful discussions.
R.C.F. consulted for Vertex Inc. in 2016 and 2017.