Recombinant Synthesis, Purification, and Nucleotide Binding Characteristics of the First Nucleotide Binding Domain of the Cystic Fibrosis Gene Product*

The majority of mutations which lead to clinical cys- tic fibrosis are located within the two predicted nucleotide binding domains of the cystic fibrosis gene product. We have used a prokaryotic expression system to synthesize and purify the first nucleotide binding domain (NBD-1, amino acids 426-588) with and with- out the most common mutation associated with the disease (the deletion of phenylalanine at position 508, AF608). Both wild type and AF508 NBD-1 bind ATP-agarose in a quantitatively comparable manner; this binding was inhibited by excess Na2ATP, trinitro-phenol-ATP, or 8-azido-ATP. Irreversible NBD- 1 labeling by an ATP analog was demonstrated using [32P]8-azido-ATP. This covalent labeling was inhibited by preincubation with Na2ATP, with half-maximal inhibition for NazATP occurring at approximately 5 mM for both the wild type and AF508 nucleotide binding domain. These experiments are among the first to con- firm the expectation that the cystic fibrosis transmembrane conductance regulator NBD- 1 binds nucleotide. Since, under the conditions used in our study, NBD-1 on a 12% acrylamide gel. After overnight exposure on film, labeling of some samples was analyzed by densitometry (E-C Appa-ratus, St. Petersburg, FL).

gene product (termed the cystic fibrosis transmembrane conductance regulator (CFTR)) functions at least in part as a C1-channel (4)(5)(6). CFTR has a predicted structure which includes two putative nucleotide binding domains (NBDs) (2, 7,8). A central role for nucleotide binding in normal CFTR function is anticipated, since the majority of mutations which lead to clinical disease occur within these domains (2,3,(7)(8)(9). Furthermore, given the prevalence of clinically relevant mutations in the nucleotide binding domains, it is reasonable to imagine that defects in nucleotide binding or ATP hydrolysis may represent underlying biochemical mechanisms in some forms of the disease.
Nucleotide binding by a full-length CFTR NBD-1 has not been previously demonstrated. Recently, a 67-amino acid polypeptide corresponding to a portion of NBD-1 of the wild type CFTR was synthesized and shown to bind the ATP analog trinitrophenol ATP (8). In the current study, we used a prokaryotic expression system to synthesize and isolate the wild type CFTR NBD-1 and to study its nucleotide binding properties. The most common mutation leading to clinical CF is the deletion of phenylalanine at CFTR position 508 (AF508). In order to test the influence of this AF508 mutation on nucleotide binding by NBD-1, we purified and characterized the corresponding NBD-1 without the phenylalanine at CFTR position 508.

MATERIALS AND METHODS
Synthesis and Isolation of CFTR NBD-1-Synthesis and isolation of CFTR NBD-1 was performed as follows. The nucleotide binding domain was defined by two primers, 5'GCGCGAATTCATGA CAGCCTCTTCTTCAG3' and 5'GCCATCAGTTTACAGA CACAGAATTCAAAB', and amplified using polymerase chain reaction (PCR) in a 100-pl reaction mixture which included 2.5 units of Taq polymerase, 100 p1 of each primer, 1 ng of template cDNA, and 200 mM of each dNTP in a buffer containing 50 mM KC1 and 10 mM Tris-C1 (pH 8.3), 1.5 mM MgC12, and 0.01% gelatin (weight/volume) (94 "C X 1 min denaturation, 50 "C X 2 min annealing, 72 "C X 3 min elongation). Template for the wild type NBD-1 amplification was American Tissue Culture Collection (ATCC) plasmid T16-1; the phenylalanine deleted (AF508) NBD-1 template was ATCC plasmid C1-1/5 (2). Both PCR products had the predicted size of approximately 520-523 base pairs. Products were digested with EcoRI (to generate ends compatible with in-frame insertion into the EcoRI site of pGEX-2T) (10, 11) and gel-purified (Geneclean, BiolOl). pGEX 2T (Pharmacia LKB Biotechnology Inc.) was linearized with EcoRI and dephosphorylated (calf intestinal alkaline phosphatase, GIBCOI BRL) and then ligated with NBD-1 cDNA to generate a plasmid coding for an in-frame glutathione S-transferase nucleotide binding domain-1 fusion protein. The design of pGEX-2T includes a thrombin cleavage site; in this case the cleavage site was engineered between the glutathione-S-transferase and NBD-1 sequences. Bacteria transfected with ligation product were grown on NCZYM agar plates supplemented with ampicillin (0.1 mg/ml). Correct recombinants were identified by multiple restriction digests and reamplification of a full-length insert from recombinant plasmid (but not from pGEX-2T without insert). In addition, internal primers which introduced a ClaI site into the PCR product from cDNA template coding for AF508 but not into the PCR product from wild type NBD-1 cDNA template were used to confirm the presence or absence of the phenylalanine deletion in recombinant plasmid as described previously (12). Growth and induction of recombinant protein synthesis in Escherichia coli were performed according to the manufacturer's protocol (Pharmaphosphate-buffered saline/Triton X-100 (150 mM NaC1, 16 mM cia), and bacteria expressing recombinant protein were sonicated in NaH2P04, 4 mM Na2HP04, pH 7.4, 1% Triton X-100). Insoluble inclusion bodies containing glutathione S-transferase NBD-1 were harvested using a slow speed centrifugation (2000 x g x 5 min).
Inclusion bodies were then solubilized in 8 M urea and dialyzed against 50 mM Tris-HCI, pH 7.4 (13). After 5 changes of 50 X sample volume, NaCl and CaCIZ were added to 150 and 2.5 mM, respectively. Human thrombin (Sigma) was added to a final concentration of 20 units/lO mg of fusion protein and allowed to incubate at 25 "C for 3 h. Products of the cleavage reaction were then loaded on a preparative acrylamide gel, and the 21-kDa NBD-1 polypeptide was electroeluted into a buffer containing 200 mM glycine, 25 mM Tris-HCI, pH 7.4. The isolated polypeptide was then dialyzed against 6 M urea, with 1% Dowex resin to absorb residual sodium dodecyl sulfate (6 X 200-ml dialysis changes). Finally, the sample was dialyzed (5 X 1 liter) into 10 mM Tris-HCI, pH 7.4, and studied using the nucleotide binding protocols described below. Utilizing this approach, 200 ml of E. coli starting material yielded approximately 100-200 pg of isolated, purified NBD-1. Amino acid sequence and compositional analyses were performed on a Beckman 7300 AAA instrument at the Yale University Protein and Nucleic Acid Sequencing Facility, New Haven, CT.
Circular dichroism (CD) spectra were recorded with an AVIV 62DS spectropolarimeter interfaced to a personal computer (80386) and measured every 0.5 nm with 2.0 s averaging per point using a 4-nm bandwidth. A 0.01-cm pathlength cell was used for far-UV spectra.
Spectra were signal-averaged by adding at least 40 scans (-5 p M NBD-1 in 10 mM Tris-HC1, pH 7.4, at 5 "C), and the base line was corrected by subtracting a spectrum for the buffer obtained in an identical manner and then smoothed with a third order polynomial fit. The secondary structure of the protein was determined by PRO-SEC, a software analysis program based on the method of Yang and co-workers (14).
ATP Affinity Column Binding-Twenty pl of ATP-agarose (Pharmacia; type 11, 111, or IV) was added to a microcentrifuge tube and washed in 10 mM Tris, pH 7.4 (3 X 1.5-ml washes). Pure recombinant NBD-1 (120 pl, 20 pg/ml) (wild type or AF508) in 10 mM Tris was then added to the agarose, and the mixture was rotated at 4 "C X 1 h. After this treatment the agarose was washed (10 mM Tris, pH 7.4, 3 X 1.5-ml washes) and boiled in 30 pl of Laemmli sample buffer to release bound NBD-1. Released protein was electrophoresed through a 12% acrylamide gel. For binding competition experiments (see "Results"), NazATP, trinitrophenol-ATP, or 8-azido-ATP was added to the ATP-agarose prior to the 1-h incubation with NBD-1. Gels were stained with Coomassie Blue or silver (15). In some experiments, images of the silver-stained gels were collected with a NuBus series 200 cooled CCD camera (Photometrics, Tucson, AZ) and relayed to a Macintosh IIci computer. The integrated intensities of protein bands were determined by Image 1.22 software, a public domain program available for image processing, and used to measure the average density of grayscale digitized images. p2P]8-Azido-ATP Binding Experiments-0.8 pg of NBD-1 polypeptide in 50 pl of 8 mM Tris, pH 7.4, was covalently labeled using a hand-held UV light (254 nM) at 3 cm X 3 min in the presence of 10 p~ [32P]8-azido-ATP (ICN, 2-10 Ci/mmol). The experiments were carried out at 4 "C in the presence of varying concentrations of NazATP (see "Results"). Laemmli sample buffer was added directly t o the reaction mixture after photolabeling and the labeled product run on a 12% acrylamide gel. After overnight exposure on film, labeling of some samples was analyzed by densitometry (E-C Apparatus, St. Petersburg, FL).

Nucleotide Binding Domain Synthesis and Purifkatwn-
The pGEX vector system was utilized to purify the CFTR NBD-1 (amino acids 426-588) (Fig. 1). After synthesis and partial purification of a glutathione S-transferase NBD-1 fusion protein (lanes B-D), the sample was dialyzed into cutting buffer (lanes E and F). Cleavage with human thrombin resulted in the appearance of two predicted polypeptide fragments, a 28-kDa glutathione S-transferase segment and a 21-kDa CFTR NBD-1 ( l a n e G ) . The NBD-1 was electroeluted and renatured as described under "Materials and Methods'' ( l a n e H). Amino acid compositional analysis and aminoterminal sequencing of the first 13 residues of the 21-kDa polypeptide confirmed identity with that predicted for CFTR CD spectra of wild type and AF508 NBD-1 revealed highly is consistent with the report of a predominantly p sheet secondary structure for a synthetic 67-amino acid peptide corresponding to the first portion of the wild type CFTR NBD-1 (8).
.--, --. An attempt to label NBD-1 after electroelution but prior to renaturation (see "Materials and Methods") was unsuccessful ( l a n e F ) . As shown in lanes G-I, labeling required the presence of UV light and was only obtained when both NBD-1 and [32P]8-azido-ATP ligand were present at the time of UV excitation. Lane J shows the result of an experiment performed at room temperature in which no labeling was observed. Temperature-dependent photolabeling of this type is typically observed with [32P]8-azido-ATP (16). Fig. 3 shows the binding of the wild type CFTR NBD-1 to ATP affinity-agarose. This binding could be inhibited by Na2-ATP, 8-azido-ATP, or trinitrophenol-ATP (lanes C-E, respectively). Both the wild type and AF508 NBD-1 demonstrated higher affinity for ATP-agarose in which ATP was attached to matrix via C8 of the adenine ring (Pharmacia type I11 agarose, lane B) than when attachment was either via adenine N6 (type 11, lane F ) or via ribose (type IV, lane G). These observations may indicate a steric preference for ATP binding by the NBD-1.
Comparison of ATP Binding by Wild Type and AF508 CFTR NBD-1- Fig. 4A shows the inhibition of [32P]8-azido-ATP labeling by Na2ATP at concentrations between 0 and 10 mM. Half-maximal inhibition for both wild type and AF508 polypeptides occurs at about 5 mM Na2ATP. The AF508 NBD-1 also exhibited an ATP affinity-agarose binding profile identical to that seen in Fig. 3 for the wild type NBD-1. Binding of wild type and AF508 polypeptides to ATP-agarose was quantitatively comparable (2 pg per 20 pl of type I11 ATPagarose under conditions such as those shown in Fig. 3, lane  B ) . In addition, orientation-specific binding of wild type and mutant polypeptides appears similar (Fig. 4B), based on differential binding to types 11, 111, and IV ATP-agarose. Furthermore, recent preliminary experiments in our laboratory that increasing concentrations of Na2ATP (0-10 mM) were preincubated with NBD-1 prior to photolabeling. Half-maximal inhibition of binding occurs at 4-6 mM for both NBD-1 preparations. The magnitude of 32[P]8-azido-ATP labeling of wild type and AF508 NBD-1 was quantitatively comparable between the two polypeptides in multiple experiments. Panel B, NBD-1 binding to ATP affinity-agarose. NBD-1 was bound to ATP-agarose (Pharmacia types 11,111, and IV) as in Fig. 3. Gels were silver-stained and subjected to densitometric analysis. The relative binding to ATP-agarose type I1 (attachment of ATP to the matrix via N6), type I11 (attachment via C8), and type IV (attachment via ribose) are shown (dark hatching, wild type NBD-1; light hatching, A508 NBD-1). Density measurements were normalized to a 2 0 4 sample of wild type or AF508 starting material run on the same gel and analyzed in parallel. utilizing a fluorescent ATP analog (trinitrophenol-ATP) have indicated quantitatively comparable nucleotide binding by the normal and AF508 mutant NBD-1 (17).

DISCUSSION
Using two independent protocols, we have demonstrated nucleotide binding by the recombinant CFTR NBD-1. This binding can be displaced by excess Na2ATP. The mechanism by which this displacement occurs (competitive uersus noncompetitive) has not yet been determined. However, under the conditions described here, half-maximal inhibition of labeling occurs at about 5 mM Na2ATP using a photoligand displacement assay. In addition, recombinant NBD-1 displays a steric preference for ATP immobilized on agarose via C8 of the adenine ring. In contrast, substantially less binding of the NBD-1 was observed when ATP was covalently attached to agarose by either ribose or by N6 of the nucleic acid base.
These data indicate that the full-length CFTR NBD-1 binds ATP and other nucleotides, an assumption which formerly has been based upon nucleic acid sequence homology of this domain of CFTR with members of an ATP binding cassette superfamily of gene products (7). Previously, a 67amino acid synthetic polypeptide corresponding to a portion of the CFTR NBD-1 has been shown to bind trinitrophenol-ATP (8). Trinitrophenol-ATP binding to this polypeptide could be displaced by excess ATP with half-maximal displacement of binding occurring at about 300 p~. The discrepancy in ATP binding that we report here could relate to added constraints in the binding of ATP by the complete NBD-1 (174 amino acids), compared with a smaller, 67-amino acid portion of it. Under the conditions used in our study, NBD-1 without phenylalanine 508 displays very similar nucleotide binding characteristics to wild type NBD-1. Our findings, therefore, are compatible with previous computer models of CFTR NBD-1 (7), particularly concerning the predicted affects of the phenylalanine 508 deletion on nucleotide binding (7,8,18). However, while our results support the notion that AF508-related CF does not result from an alteration in nucleotide binding, detailed studies with purified, full-length (170 kDa) CFTR will be required to formally exclude this possibility.
The precise mechanism by which deletion of the phenylalanine at CFTR position 508 leads to clinical disease is not known. It has been proposed that CFTR possessing the AF508 mutation is recognized as abnormal during cellular processing and that disease results because the mutant protein is retained and prematurely degraded within the endoplasmic reticulum (19). Improper cellular processing of this type does not preclude normal nucleotide binding (or even normal anion transport function) by isolated, purified CFTR polypeptides in cellfree systems. On the other hand, since many of the NBD-1 mutations that cause CF do not appear to disrupt CFTR cellular processing, one can argue that the molecular pathogenesis of CF may be heterogeneous and that some NBD-1 mutations (particularly those which occur within the so-called "Walker" sequences (2, 7,20) which are likely to be critical in the NBD-1-ATP interaction) might act by abolishing anticipated NBD-1 functions such as nucleotide binding or ATP hydrolysis. The experimental design described here provides a means of testing the influence of diverse CFTR NBD-1 mutations on nucleotide binding and perhaps also an ap-proach for further delineating heterogeneity in CF pathogenesis.