Cystic fibrosis transmembrane conductance
regulator mutations that disrupt nucleotide
binding.
J Logan, … , W J Cook, E J Sorscher
J Clin Invest.
1994;
94(1)
:228-236.
https://doi.org/10.1172/JCI117311
.
Increasing evidence suggests heterogeneity in the molecular pathogenesis of cystic fibrosis
(CF). Mutations such as deletion of phenylalanine at position 508 (delta F508) within the
cystic fibrosis transmembrane conductance regulator (CFTR), for example, appear to cause
disease by abrogating normal biosynthetic processing, a mechanism which results in
retention and degradation of the mutant protein within the endoplasmic reticulum. Other
mutations, such as the relatively common glycine-->aspartic acid replacement at CFTR
position 551 (G551D) appear to be normally processed, and therefore must cause disease
through some other mechanism. Because delta F508 and G551D both occur within a
predicted nucleotide binding domain (NBD) of the CFTR, we tested the influence of these
mutations on nucleotide binding by the protein. We found that G551D and the
corresponding mutation in the CFTR second nucleotide binding domain, G1349D, led to
decreased nucleotide binding by CFTR NBDs, while the delta F508 mutation did not alter
nucleotide binding. These results implicate defective ATP binding as contributing to the
pathogenic mechanism of a relatively common mutation leading to CF, and suggest that
structural integrity of a highly conserved region present in over 30 prokaryotic and
eukaryotic nucleotide binding domains may be critical for normal nucleotide binding.
Research ArticleFind the latest version:
Cystic Fibrosis
Transmembrane Conductance
Regulator Mutations
That Disrupt Nucleotide
Binding
JamesLogan,* David Hiestand,* Paru Daram, ZhenHuang, Donald D.Muccio,§JohnHartman,
Boyd
Haley,*
William J.Cook,1I
and Eric J.Sorscher*DepartmentofBiochemistry University of KentuckyLexington, Kentucky 40536; and Departments of Medicine, Physiologyand
Biophysics, *Pediatrics, HPathology, *Chemistry, andOCenterfor Macromolecular Crystallography, University of Alabama
atBirmingham, Birmingham, Alabama 35294
Abstract
Increasing evidence suggests
heterogeneity
in the molecularpathogenesisof cystic fibrosis (CF). Mutations such as dele-tion ofphenylalanine at position 508 (AF508) within the
cystic fibrosis transmembrane conductance regulator
(CFTR),for example, appear to cause diseasebyabrogating normal biosynthetic processing, a mechanism which results inretention and degradation of the mutant protein within the endoplasmic reticulum. Other mutations, such as the
relatively
commonglycine
-,aspartic
acidreplacement
at CFTR position 551 (G551D) appear to be normallypro-cessed,and therefore must cause disease through some other
mechanism. Because AF508 and G551D both occur within
apredictednucleotidebindingdomain(NBD) of the CFTR, we tested the influence of these mutations on nucleotide
bindingby theprotein.We found thatG551Dand the corre-sponding mutation in the CFTR second nucleotide binding
domain, G1349D, led to decreased nucleotide binding by
CFTRNBDs, while theAF508 mutation didnotalter nucle-otidebinding.These resultsimplicate defectiveATPbinding ascontributingtothepathogenicmechanism ofarelatively commonmutationleadingtoCF,and suggestthat structural
integrity
ofa highly conserved region present in over 30 prokaryotic and eukaryotic nucleotide binding domainsmay becriticalfor normal nucleotidebinding. (J. Clin.
In-vest. 1994. 94:228-236.)Key words:cysticfibrosis*genetics
*pathogenesis *ATP * structure
Introduction
Many of the mutations whichcauseclinicalcysticfibrosis(CF)l occurwithin thetwonucleotidebindingdomains of thecystic
fibrosis transmembrane conductanceregulator (CFTR) (1-3).
While the precise role of the nucleotide binding domains (NBDs) in CFTR function is stillunknown, previous studies
Addresscorrespondence to E. J. Sorscher, Department ofPhysiology
andBiophysics, University ofAlabamaatBirmingham, UABStation,
Birmingham,AL35294-0005.
Receivedfor publication 16 June 1993 and in revisedform 21
December1993.
1.Abbreviations usedinthis paper:ABC, ATPbindingcassette;CF, cystic fibrosis; CFTR,CF transmembraneregulator; GST,
glutathione-s-transferase; NBD, nucleotidebindingdomain.
have shown that the anion channel function of CFTR is depen-dent upon the presence of ATP or other nucleotides (4-8). Studies inourlaboratory (9) as well as by others (10, 11 ) have suggested that AF508, the most common mutation associated with clinical CF, is not associated with disrupted nucleotide binding by folded
CFJlR
polypeptides. Phenylalanine 508 is not highly conserved among the NBDs of more than 30 prokaryotic and eukaryotic members of theATP binding cassette (ABC) gene family which includes CFTR (1, 12, 13); this observation hassuggested that the F508 residue might not directly contributetoATPbinding byCFTR.Inaddition, increasing evidence sug-gests that
CFTR
possessingtheAF508mutation is improperly processed within epithelial cells, and that disease results because the mutant protein fails to insert properly in the apical cell membrane (14-17). Such a pathogenic mechanism is not likelytobedependent upon ATP binding by the NBD.
The second most common mutation leading to clinical cystic fibrosis isaglycinetoaspartic acid replacement within
NBD-1atCFTRposition 551 (G551D)(2, 18, 19). Because (a) both G551D and the corresponding mutation in NBD-2,
G1349D, cause clinical CF; (b) both mutations involve a
glycine residue which is invariant in alignments of>30
nu-cleotidebinding domains from the ATP bindingcassettegene
family(1, 12,13); (c) chloride channel studies from heterol-ogous eukaryotic cells expressing recombinant CFTRhave raised thepossibilityofadefective responseto ATPbyCFTR
with these mutations (5); and(d)bothG551D and G1349D appeartobenormally processedtothecellmembrane,(5, 14,
15),wetestedthe influence of these mutationsonnucleotide
bindingby CFTR NBDs.Wefound that G551D and G1349D, but not AF508, significantly decreased nucleotide binding
by recombinant NBDs. These results directly demonstrate abnormal nucleotidebindingastheresult of CFTRmutations,
and may haveimportantimplications concerning thegeneral structureofnucleotidebindingdomains within the ABC gene
family.
Methods
Recombinant synthesis andpurification ofCFTR nucleotide binding
domains. A series ofpolymerasechain reaction(PCR) primers defining
CFTR
NBD-l
(AA426 to588, 5 'GCGCGAATTCATGACAGCCTC-TTCTTCAG3'and 5'GCCATCAGTTTACAGACACAGAATTCAAA3')orNBD-2(aminoacids1219to1387,5
'GCGCGAATTCAATAC-ACAGAAGGTGGAAATGCC3' and 5
'GTGCAATCAGCAAATGC-TTGTTTTAGAATTCTTC3')wereusedtoamplifytherespective wild-type andmutantNBDs fromCFTRcDNA
plasmid
templatesobtainedfromthe American TissueCultureCollection (Rockville, MD) (wild
type: ATCCno.T16-1andAF508: ATCCno.Cl-1/5)or as agenerous
giftfrom AlanSmithand RichardGregory (Genzyme Corp.,
Cambridge,
MA)(full lengthCFTRcDNAscontainingG551DorG1349D) (4, 5,
15). PCR conditionswere1 minx 94°C denaturation, 2 minX 50°C
J.Clin.Invest.
© TheAmerican Societyfor ClinicalInvestigation,Inc.
0021-9738/94/07/0228/09 $2.00
A
97 _lp
66 E 44
6
33 4_W
21 '
1 2 3 4
>_~~~~~B C
44 _
_ '.'J I',. '.
C
-_- 44
-_-* 33
A B C D E F
Figure 1. Synthesis and purification of recombinantCFTR NBD-l and NBD-2. (A)CFTRNBDs 1 and 2 weresynthesized as fusion proteins with GST and induced using isopropyl13-D-thiogalactopyranoside (Sigma Chemical Co.) (IPTG),according to a previously described protocol (9). Lanes 2and 3 show total E. coli cell lysate prior to and after IPTG induction, respectively (upper arrow, GST/wild typeNBD-l fusion protein). Lane4, following dialysis into a protease cleavage buffer, human thrombin was used to cleave the fusion protein into its two constituents; a 28-kDglutathione-s-transferasefragment(middle arrow) and the 21 -kDwild-type NBD- 1 (lowest arrow). Lane 1 contains molecular mass standards. (B) Cleaved material such as that shown in panel A., lane 4 was loaded on a preparative electrophoresis cell (Bio Rad). Proteins were separated on thebasis of molecularmassand collected in 5-ml fractionsastheyeluted from the bottom of thegel.Lanes B-J are20-gl samplesfrom every fifthfraction. Thepurified NBD-l (21 kD)is seen in lanes D and E. Molecularmassstandards are shown in lane A. (C) Using the same approach, aseriesof wild-type and mutant NBDs werepurified.LaneA, wild-type NBD-1; lane B, AF508NBD-1; lane C, G551D NBD-1; lane D, wild-typeNBD-2; lane E, G1349D NBD-2; lane F, molecular mass standards. Gels in A-C contain 12% acrylamide. GST/NBD fusion proteins run aberrantlyon 12%SDS-PAGE (- 43 kD). This has been noted for all five independently cloned NBD polypeptides used in this study, as well as forfour additional full length NBD mutants (not shown).
annealing, and 3 minx72°Celongation.Primers werechosen to create EcoRI sites at the ends of PCR products to allow cloning into the EcoRI siteofpGEX 2T(PharmaciaFineChemicals, Piscataway,NJ) in frame with glutathione-s-transferase (GST) and immediately 3' to an engi-neered thrombincleavagesite (9).AllPCRproducts obtainedbythis method were of thepredictedsize (524-bp NBD-1, 545-bpNBD-2).In
addition, identityofcorrectrecombinants were verifiedby (a) restriction
mapping using 10-12digestsperrecombinantconstruct;(b) re-ampli-fication ofthe fulllengthNBDinsertusingtherecombinantplasmidas
template; (c) recombinantplasmid-directedsynthesis of fusionprotein
andspecific cleavage byhuman thrombin into thetwopredicted
frag-ments; and(d) amino-terminalsequencing and compositional analysis (13 amino-terminal residues of NBD-1 and 12 amino-terminal residues ofNBD-2, aswell as >30 internal residues obtainedduring peptide mappingmatched the predicted sequence of the recombinantproteins).
Recombinantfusion proteinwascollectedandpartially purifiedas an
inclusion body preparation as described previously (9). After thrombin
treatment (inaproteasecleavage buffer containing 50 mM TrisHC1,
150 mM NaCl, 2.5 mM CaCl2, pH 7.4, and 20 U human thrombin
(SigmaChemical Co., St.Louis, MO)/10 mg recombinantprotein) - 3
mgofthepartially purified, cleavedpreparation(i.e.,containingNBD,
GST,someuncleavedfusion protein and minor bacterial contaminants) wasapplied to the top of apreparativeacrylamide gel (Bio Rad
Labora-tories)and the NBDpurified using electrophoresis through12%
acryl-amide as in Fig. 1 B. The observation that the polyacrylamide gel electrophoresisbasedprotein purification step used hererequired solubi-lization in 0.8%SDSis consistent with a recent report that solubilization of recombinant full length CFTR was refractory to a large series of
detergents, butultimately could be solubilized in 2% SDSwhile still
permitting efficient (20-40%) refolding of the protein intoa native formcapableofsupporting CFTRanion channel activityin theplanar lipid bilayer (6).
TrinitrophenylATPbindingby NBDs. 250 .gper mlpurified NBD-1 (-10
M1M)
in 10 mM Tris, pH 7.4, was incubated at40C withincreasingconcentrations oftrinitrophenylATP(TNP-ATP; Molecular
Probes, Inc., Eugene, OR). TNP-ATPdemonstrates fluorescence en-hancement when present in a nucleotidebindingpocket(10, 11). Ade-ninenucleotide fluorescencewasmeasuredateach datapointfor 1 sin
a 1-ml quartzcuvette usinga Perkin Elmer LS-50 spectrofluorometer (Perkin-Elmer Corp., Norwalk, CT) with excitation 408 nM and emis-sion 538 nM. Fluorescence enhancement was determinedateachdata point bymeasuring(a) thefluorescent signalof bufferplusTNP-ATP
ateach concentrationstudied; (b)the fluorescence ofproteinin buffer alone; and (c) thefluorescent signal ofprotein + TNP-ATP ateach
concentration and using the calculation: fluorescence enhancement
=(fluorescence signalofprotein+TNP-ATP)-(fluorescenceof TNP-ATPalone+fluorescence ofprotein alone). Becausepreliminary exper-iments indicated thatbindingof TNP-ATP to the NBD-1 isnot depen-dent upon divalentcations,Mg2`was notaddedtothe titration solutions.
y32p-2-azidoATPlabeling ofNBDs.2-azido ATPwassynthesized
from 2azido AMP(20) by themethod of Michelsonasmodifiedby
Salvuccietal.(21). [y-32p]2-azido ATP(sp. act. 5-35
mCi/,.mol)
wasthenprepared andpurified as describedpreviously (22).Thepurity[y-32P]2-azidoATPwasdetermined by TLCanalysis.
Photoaffinity labelingwasconductedat4°Cin50-IAreactions
con-taining4 ,ugpurifiedrecombinantNBD, 10 mMtris, 10 mM MgCl2, pH7.1(withoutATPpresent) and the indicatedamountofphotoprobe.
Reactions wereinitiated by the addition ofphotoprobe. After a 30-s
incubation,the reactionswerephotolyzedfor 1 min withahand-held 254-nm UVlamp(I=3,100HIW/cm2).Afterphotolysis,sampleswere
precipitatedwithanequal volume of 7%perchloricacid and resolubi-lized inabuffercontaining10%SDS,3.6 M urea, 162 mMdithiothreitol,
m-Cresol purple as a tracking dye, and 20 mM Tris-HCl, pH. 8.0.
Samples werethen subjected to electrophoresis by an 8-12% SDS-PAGEseparating gel witha4%stacking gel accordingtothe method of Laemmli(23).Gelswerethenstained,fixed,andlabeling quantified
on anAMBIS 4000Imaging System(Ambis, Inc., SanDiego, CA).
Results
We usedprokaryotic over-expressionto generate quantities of
CFTlR NBDs sufficient forbinding and structural studies(Fig. 1). Because recombinant proteins synthesized in bacteria (as
well as ineukaryotic cells) commonly require refolding
3 4 5 6 7 8 9 10 11 Concentration
(ILM)
= wt.
M AF508
_ G551D
B 200
150
M
-4-,
U 100
0
50
2.5 5 10 15 25 50 75 100
Concentration
(,uM)
obtain arefolding sequence relevant to the native CFTR struc-tureandfunction. First,wehavepreviously shown that recombi-nantwild-typeandAF508CFTR NBD-1 derived from bacteria
bind ATPaffinityagarose aswellas32p-8-azido ATP (9). The
data presented below indicatesthat two additional ATP analogs,
trinitrophenyl ATP and 32p-2-azido ATP, also bind the
recom-binant NBD. All four ligands can be displaced by Na2 ATP
with half maximaldisplacement occurring at - 5 mM ATP. (9,
E.J. Sorscher andJ. Logan,unpublished observations). Second,
theproteins generated for thesestudiesdisplay substantial
sec-ondary structuralcontent (- 70%,3 sheet, 10% a helix) when
studied by circular dichroism, a result in agreement with a high
,/
sheetcontentreported for a synthetic peptide correspondingto a67 amino acid segment of the NBD-1 (10). Third,although
invitro studies of recombinant CFTR using cell free membrane patches have demonstrated half maximal ATP stimulation at
- 270
jIM
(4), the ATPbinding affinity constants which wehave observed for NBD-1 are very similar to those reported for
CFTRCl-conductance reported in both native, perfused human sweat ducts and in anon-recombinant human cell line express-ing high levels of CFTR (T84) (7, 24; see also 8). Half maximal CFTR activity in both systems was reported at 3-5 mM, a finding which suggested that cellular ATP levels may represent
animportant regulatory constraintonCFTR anion channel ac-tivity. These observations indicate that the nucleotide binding
domains used in ourstudies possess substantial structure after refolding and bind ATP with an affinity very similar to that observed in macroscopic measurements of full length CFTR Cl- conductance.
We usedtwo independent assays of nucleotide binding to
testthesignificance of theAF508 and G551D mutations (Fig. 2). NBD-1 containing theAF508 mutation demonstrated very similarbindingtothewild typeNBD-1 as measuredusing either
TNP-ATP or32p-2-azido ATP. This result is consistent with previous studies inourlaboratory using other measurements of nucleotide affinity (9), as well as with conclusions reached using fully folded synthetic peptides corresponding to portions of the CFTRNBD-1 with andwithout F508 (11), and with a computer-based model of the NBDderived from asecondary structural alignment containing the known two- and
three-di-mensional structure of adenylate kinase (12, see below). In
contrast, as shown inFig. 2A, the G551D mutation inhibited
-0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6
Concentration
Figure2.(A)TNP-ATPbinding byNBD-l. 250jg/mlpurified NBD-1in 10 mMTris, pH 7.4,wasincubatedat4°Cwithincreasing
concen-trations oftrinitrophenylATP. Eadie Hofstee transformation of the above data indicatedanapparent half maximalbindingat3.1, 3.5,
and 2.9 MM for wild type,AF508,andG551D, respectively;maximal
bindingvalueswere0.92, 1.05,and0.27relativefluorescence units
for the threerespective proteins.Bindingof TNP-ATPbythe wild type NBD-l could be inhibited in thepresenceof5 mMMg2ATP(50%
inhibitionat5 mMMg2 ATP). TNP-ATPbindingwascompletely ab-sentinthepresenceofdenaturingconcentrationsofurea(4 M),
indicat-ingtheimportanceofproperfoldingof the recombinant NBD for nucleo-tidebinding (not shown).n =7 for wild type andG551D,n= 3 for
AF508.Errorbars = SEM.(B)32p-2 azido ATP labeling of NBD-1.
32p-2-azidoATPatthe concentrationsshownwasincubated with
NBD-1 (4Mg)for30sfollowedby1 minofUVphotolysistocovalently label the recombinantproteinswithin the nucleotidebindingsite. The
gelswerestained, fixed, and thelabeling quantifiedon anAmbis 4000
Imaging System (Ambis, Inc.).InC,aLineweaver-Burkplotof the
data isshown;half-maximallabelingwasobtained at 1.3MMfor the
wild-typeNBD-1,2.1 MMfor the /F508NBD-1,andat7.5 MMfor
NBD-1containingthe G551Dmutation. To obtain maximal photolabel-ing, 32p-2azido ATPphotolysis shownin Band Cwasperformedin
the absence of free ATP.
0.9
A
0.8 0.7
0
0
0
0
IL0
0
Cu ._ Ir
0.6 0.5 0.4 0.3
0.2
0.1 1
TNP-ATP fluorescence enhancement by the recombinant
do-main.
The half maximal TNP-ATP binding values observed for wild type, AF508, and GS5lD were 3.1, 3.5, and 2.9
MM,
respectively, whereas the maximal binding occurred at 0.92,1.05, and 0.27 relative fluorescent units. We estimateda5-mM half-maximalATPbinding basedonATPdisplacement of TNP-ATPand32p-2 azido ATP from the CFTRNBD-1.IntheNBD,
therefore, as in many nucleotide binding proteins, TNP-ATP
exhibits several orders of magnitude greater binding affinity
than does ATP. This may be due in part to thehydrophobic
phenyl group present in TNP-ATP and thehydrophobicnature
of many nucleotide binding sites. Forexample, in the gastric
H',
K+-ATPase, the apparent KD for TNP-ATPbinding was <25nM, whereas the apparent half maximalbinding byATPwas 86 tM (25). Inthe Na/KATPasepurified from eel elec-troplax, theKDfor TNP-ATPwasupto580 times greater than theKDfor ATP,dependingontheionic conditions used(26).
Nucleotidebindingbyapeptidefragment ofDNApolymerase Ialso wassubstantially tighter for TNP-nucleotides(kD = 0.5-1.5
MM)
than for thecorresponding nucleotides without phenyl group attached (kD = 230-2,100MM)
(27).Our estimatedbinding affinityfor ATP agreescloselywith
ourpreviously reported estimates ofATPaffinityfor theNBD
(9), withmacroscipic measurements of the ATP dependence of CFTR Cl channel activity using twoindependent systems (7, 24) and reasonably well with displacement of TNP-ATP
from a maltose binding protein/NBD-l fusion protein (28),
and withdisplacement of 32p-2 azidoATPfromthefull length CFTR (29) (half-maximal displacement at 1 mM). How-ever, while our results clearly indicate diminished nucleotide binding mediated by the GSSlD mutation, the micromolar half maximalbinding constantsusing TNP-ATP whichweobserved are notlikely topredict the dependence ofCFTRCl- channel activity onATP orothernonderivitized nucleotide triphosphates (KD values for ATP dependence of
CFTR
chloride channel have ranged between - 25 ,uM and 5 mM, depending onthe particular CFTR Cl- channel preparation used) (4, 5, 7, 24, 30).The binding signal obtained with TNP-ATP reflects both theinstantaneous number of nucleotide binding sites occupied by theligandand thelocalenvironment within each site (which in turn isreflected inthemagnitude of fluorescence enhance-ment).Becausethebinding environments within the nucleotide pockets of the wild type, AFS08andGSSlD
NBD-l
polypep-tides maynotbeidentical, the quantum of fluorescenceenhance-mentobserved for each molecule ofTNP-ATPwhichbinds to
anNBD molecule cannot be assumed to be the same for the threerecombinant NBD-1 proteins. However, evenif the fluo-rescent (quantum) yield for each binding event is different, if the same numberof binding sites exist in solution, the ligand (TNP-ATP) concentration observed at saturation should not vary between the three NBD-1 polypeptides. This is not the
caseinour
experiments
(Fig.2A), whereGSSlDsaturates at 8-10Mm
TNP-ATP, and saturation for the wild-type and theAFS08 NBD occurred at significantly higher concentrations. This results suggests that inaddition to physically disrupting
thenucleotidebinding pocket within the nucleotide binding site
(see
Discussion),
theG-- Dreplacement
atposition
551might
alsodestabilizeoverall protein folding in such a way as to create
anequilibriumbetween NBD folded and unfolded forms. The result would be an apparent decrease in measurable
binding
sites within the NBD preparation and would explain the lower ligand concentration at saturation for GSS1D described in Fig. 2 A. Ultraviolet spectralanalysis of thestructuresofwild-type andmutant NBDs in the presence and absence of nucleotides, andin the presence of denaturants such as urea (9, 11 ) are in progress to evaluateeffects on protein folding stability mediated by GSSlD.
To confirm that less nucleotide was actually bound to CFTR polypeptides containing theGSSlD mutation, we performed an additional series of experiments using an independent nucleo-tidebinding assay. As showninFig. 2,Band C, at
concentra-tions below saturation, 32p-2-azido ATP labeling is diminished intheGSS1D mutant as compared with wild type (half-maximal photoinsertion was [in
1LM]
1.3 [wt]; 2.1 [AF508]; and 7.5 [GSS1D]. Because the two assays used in this study measure complementary features of nucleotide binding environment and ligand occupancy, taken together these results indicate defective nucleotide binding mediated by the GSSlD mutation. While the fluorescence enhancement observed for TNP-ATPbindingdepends primarily upon the integrity and local environment of anucleotidebinding pocket,32p-2 azido ATPlabelingnotonly requires an adequate binding site, but also requires properly oriented residues within the binding site which areaccessible for covalent photolabeling. In our experiments, the functional defect caused byGSS1D ledto anapparent decreasedmaximal binding by TNP-ATP(Fig. 2 A), and a diminished half maximal binding inexperiments using the32p-2 azido ATP (Fig. 2 B). Such anobservation could beexplained byachange in nucleo-tidebinding pockethydrophobicity (causingadecreased fluo-rescence signal from TNP-ATP even at maximal occupancy) together with loss of oneof a small number of sites normally available for covalent photolabeling by 32p-2-azido ATP (an abnormality with might be overcome at increasing
concentra-tions ofphotoaffinity ligand).
Although the glycine corresponding to G551 is invariant among morethan 30 ABCproteins (andan evenlarger numbers of NBDs, since each ABC protein possesses 2 NBDs), the functional significance of this glycine residue and the immedi-atelyadjacent, highly conserved residues found within all mem-bers of the gene family is not known (Other members of the gene family, for example, include the human multidrug resis-tancegene (MDR), the chloroquine resistance gene product of P. malaria, and the recently described human adrenoleukodys-trophy gene [1, 12, 13, 31-34]). An
LSGGQQXQR
motif is highly conserved across the ABC gene family (glycine corre-sponding to G551 underlined) (1, 12, 13), but is not present in other nucleotide binding proteins such as adenylate kinase, for example, or included within the so-called Walker A or Walker B sequences characterizing a broad range of non-ABC family, ATP-binding proteins (35, 36). It has been proposed that inthe histidine permease of S. typhimurium (an ABC pro-tein), the residues near theLSG(JQXQR
motif could serve as a "linker peptide" which conveys conformational changes between segments of the domain (37).I I, I I I
0 1 2 3 4 5 6 7 8 9 10 11 5 10 15 25 37.5 50
Concentration(PM) Concentration
(AiM)
Figure3. Binding of nucleotide by CFTR NBD-2.Aand B show TNP-ATP binding and 32p-2 azidoATPlabeling ofthe wild-type and G1349D
NBD-2,respectively. ExperimentswereperformedasinFig. 2. Maximalbindingof TNP-ATPtowild-typeNBD-2was 2.5 Xgreaterthan that
observedfor the G1349Dmutant. (Using algebraic transformationasdescribed inthe previous figure, forwtNBD-2, maximal binding = 8.9
fluorescence units; half maximal binding =3.8uMTNP-ATP; for G1349D,maximal binding= 3.5fluorescence units, half maximalat5.3
ILM
TNP-ATP). As shown in B, inclusion of G1349D withinNBD-2blockedazido nucleotide labeling of the protein. The magnitude of 32p-2 azido
ATPlabeling is highly dependentuponthe specific activity of the photoaffinityligand used. Thespecificactivity of 32p-2 azido ATP used in Fig.
3 Bwas 10-fold higher than that used in Fig. 2B.When wild-typeNBD-1 andNBD-2weredirectly compared using photolabeling under identical
conditions(and using thesamelot of32p-2 azido ATP),a - 10-fold enhancement of maximal photolabelingwasobserved inNBD-2 compared
with NBD-1 (datanotshown).
TNP-ATPwere8.9 fluorescence units (wild-type NBD-2) and
3.5 fluorescence units(G1349D NBD-2);half-maximalbinding
wasobservedat3.8 and 5.3gcm TNP-ATPfor thewtNBD-2 and G1349Dmutants,respectively.Half-maximalphotoinsertionof the wildtypeNBD-2 with 32p-2 azidoATPwasobservedata
ligand concentration of - 13 um, whereas photolabeling was
substantiallyless for the G1349Dcomparedwith thewild-type
NBD-2 at every ligand concentration studied. These findings
havesignificance, since theysuggest ageneralfunction of the
conservedglycineandsurroundingresiduesmaybe in
nucleo-tidebinding.The resultsareconsistent both withprevious obser-vationsindicatinganabnormal ATPdependenceinCFTR
con-taining mutations in this glycineresidue (5) and with the ob-served loss of function of other ABC proteins containing comparablemutations (38). However, experiments measuring
CFTRorother ABCproteinfunction couldnotdistinguish
be-tweendisruptedATPbinding, hydrolysis, orutilization of
nu-cleotide asthe mechanism of observed abnormalities in
func-tion. Our data support the notion that in addition toabnormal
biosynthetic processing, cysticfibrosismayresult fromdefects in nucleotidebinding bytheCFTlR.It islikelythatanadditional mechanism (i.e.,absent CFTRpolypeptideduetothepresence
ofpremature stopcodons within certainmutantCFTRalleles)
mayalsoaccountfor the defectinsome cases(18, 39).
Discussion
Totestthe influence of CF mutationsonnucleotidebinding,
we constructed and purified the CFTR domains which have
beenimplicatedasbothnecessary and sufficient for nucleotide interactions with CFTR (1-5, 9-13). The highly conserved, functionally independent and even interchangeable nature of
CFTR and other nucleotidebindingdomains(whichspan - 200
aminoacids and share 30% amino acididentity throughoutthe
ABCgenefamily)suggeststhatoneapproachtounderstanding
the molecularpathogenesisof CFTR mutationsmightfocuson
the domainsthemselves,asopposedtoanyparticularATP
bind-ingcassetteproteincontext.Forexample,the NBDs in the LIV-Ineutral branched chain amino acidtransport systemofE.coli and the histidinepermeaseofS.typhimuriumarefoundongenes
separatefrom the remainder of thetransportcomplex (andare
therefore synthesized separately), but assemble with the
re-mainder of the ABC complex at the cell membrane. When CFTR NBD mutations were placedat corresponding residues within these bacterialNBDs, (orwithin the NBDs of theyeast
mating factortransporter, STE 6), certain of these mutations had functional effects consistent with those observed in CFTR
(althoughothers didnot.See refs.37, 38, 40). Furthermore,in
ABCproteins containingboth NBDs withinasingle polypeptide chain, thetwoNBDscanbeseparatedand still maintain func-tion. Forexample, when the STE 6 cDNA, which codes for
both NBDs of theprotein,was severedintotwohalves, (each
of which containedoneof theNBDs), coexpressionof the two
STE 6 truncations led to functional reconstitution of STE 6
mediated a-factortransport(40).
NBDsarealsointerchangeableamongmembers of the ABC
protein family. The NBD-1 of CFTR, for example, can be
grafted into STE 6 in place of the STE 6 NBD-1 to allow
A 5
4
0 0 c
0
D
(a 0 0
0
cc
3
2
1
functional transport ofmatingfactor.Importantly, NBD-1
muta-tions known to disrupt CFTR function also led to disrupted mating factor transport by the chimeric STE6/CF]R protein (41).Finally, although dataconcerningarequirement for ATP hydrolysis in CF`R activity is not yet conclusive, in some members of the ATP binding cassette gene family known to
hydrolyze ATP, isolated NBDs are sufficient to mediate this
hydrolysis (12,42, 43).Takentogether,thesefindings demon-strate that the NBDs within the ATP binding cassette family
are functionally independent, self-contained "cassettes" and emphasize (a) that characterization of the nucleotide binding domains, themselves, (independentofthetransmembrane, regu-latory and otherregions withinATPbindingcassetteproteins) canoffer substantialinsight into the mechanism of actionwithin
this family of gene products, and(b) the relevance of normal
and abnormalfunction of the domainstotheactivityofthefull
length ABCproteins. Indeed, inasimilarapproachto that ap-plied in the current study, binding and structural studies of synthetic peptidescorrespondingtoportionsof the CFTR NBD-1 (10, 11), as well as computer-assisted models of isolated nucleotidebinding domains basedonsecondary structural align-ments with the known crystal structures of adenylate kinase,
p21 ras, and the elongation factor Tu, have previously been usedtopredict thesignificanceofclinically importantmutations
onCFTR nucleotidebinding and function (seebelow, 12, 13). NBDs within theABCfamilyverylikelysubserve similar func-tions, sothatconclusions concerning the CFTR NBD- 1 could beapplicabletoCFTRNBD-2andtoNBDsofother members
ofthegenefamily aswell.
Themaximalbinding ofTNP-ATPby wild type andAF508 NBD-l was 3-4-fold greater than that observed for the G551D
mutant.Defective maximalbindingcapacity ofasimilar magni-tudehas beenproposed to accountforsomeforms of familial hypercholesterolemia (inwhichligandbinding bymutant LDL
receptor is decreased by fourfold) (44), for oxygen affinity
mutants inclinical hemoglobinopathies (45), aswellas inthe twofold differenceinbinding orenzymatic function which ac-countsforenzymatic defects of many congenital diseases (e.g., those inherited in an autosomal dominant fashion) (46). In addition, since in most cases the G -+ Dmutant allele causes
disease inconjunction withaAF508allele, nucleotide binding in acompound heterozygote (which may already be 50% de-creased duetoaberrantprocessing of theAF508geneproduct) might represent only a residual of15-20% of that seen in a wild-typecell.
Although the anion channelactivity of wild-type and mutant
CFTRpolypeptides has become themostuseful endpoint defin-ing normal CFTRfunction, mutations which cause clinical CF in manycases causeonly small alterations in channel activity in vitro. For example, in a series of four NBD-1 and NBD-2 mutations which cause disease (including G1349D and
GSS1S), although 4-8-fold differences in open probability
wereobserved in vitro when compared with wild type, no abnor-malities in maximal single channel currents could be docu-mented (althoughsmalldifferences in the dependence of current
onATPwereobserved for the mutants) (5). In vivo, potential differencemeasurements inrespiratoryepithelia of CF patients differ by less than threefold compared with normal controls (47, 48). Thisdatafurther supports the notion that even subtle
changesin normal CFTR function may have important physio-logic consequences. Finally, measurements of CFTR activity in intact human tissues and cell
monolayers
have indicated thathalf maximal CFTRactivationrequires maximal cellular ATP
levels (- 5 mM), and thatCFTR activity maybe coupled to
physiologic nucleotide levels within the cell(7, 24). The de-crease in bindingof ATP analogs demonstrated here has sig-nificance, therefore,since a similar defect in nucleotidebinding in vivo could resultin mutantCFTRwhich would be activated
onlyatnucleotide levels which are notpresentwithin thecell.
OurstudiesusingrecombinantCFTRNBDsprovideatest
ofrecentpredictions concerning the function and structure of
NBDswithin the ABC genefamily. Forexample, basedupon Michaels-Menten analysesof the ATP dependenceof CFTR anion channelopenprobability, it has beenproposedboththat
thefull length CFTRcontainsonlyone functionalATP binding site(30), and thattwoactive nucleotide bindingsites are
pres-ent within the fulllengthmolecule(5).Our results are the first todirectly demonstratethat both CFTR NBD-1 and NBD-2 are
capable ofbindingnucleotide, and thereforeprovideevidence
for the presence ofmore than onefunctionalCFTR nucleotide
binding site.Inaddition, binding affinity appearstobe different between the two domains. NBD-2 exhibits a modest(ninefold) increase in maximal TNP-ATP fluorescence when compared with NBD-1 (compareFigs. 2Aand 3 A), andanapproximate 10-fold increase inphotolabeling with32p-2azido ATPwhen NBD-2and NBD-1 arecompared in thesameexperiment (data
notshown). Increasedaffinity ofNBD-2 for a nucleotideanalog could support the notion thattwoNBDs of CFTR maynothave
identical roles in the maintenance of CFTRC1-channel activity.
Thepossibility of different functions of thetwoCFTRNBDs
has beensuggestedpreviouslybaseduponstudies of the influ-enceof NBD-1 andNBD-2mutationsonthe CFT`RC1-channel
in isolated membrane patches, where functional differences in
the domains led to speculation thatATPhydrolysis by CFTR was dependentprimarilyon NBD-2rather than NBD-1 (5).
Althoughahigh resolutioncrystalstructureisnotyet avail-able for any of the ABCprotein NBDs, Chou-Fasman based
predicted secondary structure of the CFTR and other NBDs corresponds verycloselytothatofnon-ABC,knownnucleotide binding proteins such as adenylate kinase, p21 ras, and the
elongation factor, Tu. Thestriking alignment of thepredicted
2° structure of the CFTR NBD and the known structure of these other proteins has been used to predict the functional consequences of disease-associated CFTR mutations. For
exam-ple,amodel oftheCFTRNBDbasedon20structuralhomology with adenylate kinase (12) predicted that theAF508mutation wouldbe spatially so distant from the nucleotide binding pocket of CFTR NBD-1 that it would notdisrupt ATP binding. Our resultsareconsistent with this prediction. No specific prediction
was made
concerning
the G-+DmutationsatCFTRpositions
551 and 1349. In an attempt torelate NBD tertiary structural models to our findings concerning nucleotide binding,
assign-ment of the corresponding locations of these residues within the known adenylate kinase tertiary structure wasperformed.
As
shown
in Fig. 4, assignment of the positions of G551and G1349 isstraightforward based on previously determined alignments (12). These residues would be predicted to fall immediately afterthethird/8strandinboth the adenylate kinase and the predicted CFTR NBD structure. In addition, in the
original CFTR NBD model, the ATP binding pocket was
as-signedto a large cleft corresponding to the AMP binding site in adenylate kinase. When this binding site is appropriately adjusted based on the most recent crystallographic and NMR
ade-Figure 4.Positions ofresidues
G551andG1349inatertiary
structuralmodel of the NBD based on adenylate kinase. The similarities betweenpredicted sec-ondary structure of ABC protein NBDs and the known structureof
adenylatekinase, together with the structuralsimilarities of nucleo-tide binding sites within diverse, unrelated proteins outside of the ABCfamily (e.g.,between ade-nylate kinase, the elongation fac-tor Tu andp21 ras), suggest that the tertiary structure of nucleotide
bindingsites in the NBDs of
CFI7Rand other ABC family members mayclosely resemble those of adenylate kinase (12, 13,
35).In anattempt to relate previ-oustertiary structural models of
CFTRto ourfindingsof defective nucleotidebinding by the G551D andG1349D NBDs, we predicted the location of these residues based on previous alignments of CFTR NBDs with adenylate ki-nase. Asshownin the Figure, G551 and G1349 occur in a loop immediatelyfollowingthe third/s
strandof both the adenylate kinase and theCFTR NBD models (49,
50;theprediction alsoconforms
totheCFTRNBDmodel of ref. 12). The Walker A sequence is part ofaconnecting loopbetween the first/3strand andan ahelix. Both the Walker A motif and the residues atG551 orG1349 are predicted to fall within a large cleft suitable for nucleotide bind-ing; this cleft is known both by
NMRandx-raycrystallographyto
bindAMPwithin theadenylate
ki-nasemolecule.Thetwoviews aboveare- 900rotatedfrom each
other;the/3 strandsaregreen,a
helicesareblue,and theloopsare
white. The WalkerAsite and the location ofG551orG1349are
shownin red. A ball andstick modelofAMPis shown ina
nu-cleotidebindingsite. Theatoms arecoloredaccordingtothe
fol-lowingcode:N, blue,0, red, C,
green, P.yellow. Thefigurewas
prepared using the program Rib-bon 2.0(51).
nylatekinase (49, 50), as shown in Fig. 4, G551 and G1349 are onthe wall ofacleft suitable for ATPbinding within the NBD. Another wall ofthecleft is formed by the so-called
p-looporWalkerAmotif(GXXXXGKT/S)which is(a)present innearlyall ATP bindingproteins, (b)known tocontain
resi-dues whichactuallycoordinate withphosphate groupsofATP (suchasthep-looplysinewithinp21ras),and(c)
highly
likelytobe involved in ATP binding bytheCFTRand otherNBDs,
nucleotide binding mediatedby G551D and G1349D, as well
as normal binding by LF508, provide
experimental
evidence supporting predictions of thestructureof the ATPbinding sitewithin the CFTRNBD. The results further support thenotion that the LSGGQXQR motif found within all ABC proteins is
involved in normal nucleotidebinding, and in factmayforma
critical
portion
ofan ATPbinding
cleft. An essential role innucleotidebindingcouldexplainthehighlevelof conservation
ofthis region withinthe ABC gene
family,
and would offeramolecularmechanismbywhichG551D and G1349Dcontribute
to clinical CF.
Acknowledgments
Theauthors wishtothank Dr. RichardGregoryand Dr. AlanSmith for
providingthe G551D and G1349DmutantCFTRcDNAs used in this
study.Wealso wishtothank Dr. Sadis Matalon forhelpingtoestablish the TNP-ATPprotocol,and Dr. M.-D. Tsaifor useful discussions
con-cerning the adenylate kinase structural model. Weare grateful toDr.
TipoLiu and Edward Walthallfor assistance with theFigures,andto
Dr. Shenyi Peng and Yuee Wang for technical assistance. We thank Jennifer Smith and Bonnie Parrott forexcellent secretarial assistance.
This workwassupportedbyNationalInstitutes of Health grantP01
DK38518 andbytheCysticFibrosis Foundation.EricJ.Sorscher isa
Lucille P. MarkeyScholar in theBiomedicalSciences.
References
1. Riordan, J. R., J. M.Rommens, B.-S. Kerem, N. Alon,R. Rozmahel,
Z.Grzelczak,J.Zielenski,S.Lok,N.Plavsic,J.-L.Chou,M. L.Drumm,M.C.
lannuzzi,F. S.Collins,and L.-C. Tsui. 1989. Identification of thecysticfibrosis gene:Cloningand characterization ofcomplementaryDNA.Science(Wash.DC).
245:1066-1073.
2. Tsui, L. -C. 1991. Probing the basic defect in cystic fibrosis. Current
OpinioninGenetics andDevelopment. 1:4-10.
3. Collins, F. S. 1992. Cystic fibrosis: Molecularbiology and therapeutic implications.Science(Wash. DC).256:774-779.
4.Anderson,M. P., H. A.Berger, D. P.Rich,R.J.Gregory, A. E.Smith,
and M. J.Welsh.1991. Nucleotidetriphosphatesarerequiredtoopen the CFTR chloride channel. Cell. 67:775-784.
5.Anderson,M.P.,and M. J. Welsh. 1992.Regulation byATPand ADP of CFTR chloridechannels that containmutantnucleotide-bindingdomains. Science
(Wash. DC).257:1701-1704.
6.Bear,C.E.,C.Li,and N.Kartner,R. J.Bridges,T. J.Jensen,M. Ramjee-singh,and J. R. Riordan. 1992. Purification and functional reconstitution of the
cysticfibrosis transmembrane conductanceregulator(CFTR).Cell.68:809-818. 7.Quinton,P.M.,and M.M.Reddy.1992. Control ofCFTRchloride conductance
byATPlevelsthrough non-hydrolytic binding.Nature(Lond).360:79-81. 8. Wine,J. J.,and S. C. Silverstein. 1992. ATP and chloride conductance. Nature(Lond.). 360:18.
9.Hartman, J.,Z.Huang,T. A.Rado,S.Peng,T.Jilling,D. D.Muccio,and E. J.Sorscher. 1992. Recombinantsynthesis,purification,andnucleotidebinding
characteristics of the first nucleotidebindingdomain of thecysticfibrosis gene
product.J.Biol.Chem. 267:6455-6458.
10.Thomas,P.J.,P.Shenbagamurthi,X.Ysern,andP. L. Pedersen. 1991.
Cystic fibrosis transmembrane conductanceregulator: Nucleotidebindingto a
synthetic peptide.Science(Wash. DC).251:555-557.
11.Thomas,P.J.,J.S.Shenbagamurthi,J. M.Hullihen,and P. L. Pederson. 1992. Effects of themostcommoncystic fibrosis-causingmutationonthe second-arystructureandstabilityofasynthetic peptide.J. Biol.Chem.267:5727-5730. 12.Hyde, S.,P.Emsley,M.J.Hartshorn,M.M.Mimmack,U.Gileadi,S. R.
Pearce, M. P.Gallagher,D. R. Gill,R. E.Hubbard, andC. F.Higgins. 1990. Structural model ofATP-binding proteinsassociated withcystic fibrosis,
multi-drugresistance and bacterial transport. Nature(Lond.).346:362-365. 13.Mimura,C.S.,S. R.Holbrook,and G. F. Ames. 1991. Structural model of thenucleotide-bindingconserved component ofperiplasmicpermeases. Proc. Natl.Acad. Sci. USA. 88:84-88.
14.Cheng, S. H.,R. J.Gregory,J.Marshall,S.Paul, D. W.Souza,G. A.
White,C. R.O'Riordan,and A. E. Smith. 1990. Defective intracellular transport and processing ofCFTR is the molecular basis ofmostcystic fibrosis. Cell. 63:827-834.
15.Gregory,R.J.,D. P.Rich,S. H.Cheng,D. W.Souza,S.Paul,P. Manava-lan, M. P. Anderson, M. J. Welsh, and A. Smith. 1991. Maturation and function
ofcysticfibrosistransmembrane conductanceregulatorvariantsbearingmutations inputative nucleotide-binding domains 1 and 2. Mol. Cell. Biol. 11:3886-3893. 16. Denning, G. M., M. P. Anderson, J. F. Amara, J.Marshall,A. E.Smith,
and M. W. Welsh. 1992. Processingof mutantcystic fibrosis transmembrane conductanceregulatoristemperature-sensitive.Nature(Lond2).358:761-710.
17.Kartner,N.,0.Augustinas,T. J.Jensen,A. L.Naismith,and J. R. Riordan. 1992.Mislocalization of AF508 CFTR in cystic fibrosis sweat gland. Nature Genetics. 1:321-327.
18.Tsui, L-C. 1992.The spectrum ofcystic fibrosis mutations. Trends in Genet. 8:392-398.
19.Richards,B., J. Skoletsky, A. P. Shuber, R. Balfour, R. C. Stem, H. L. Dorkin, R. B.Parad,D.Witt,and K. W.Klinger.1993.MultiplexPCR
amplifica-tion from the CFTR geneusing DNA prepared from buccal brushes/swabs. Hum. Mol. Gen.2:159-163.
20. Kim, H., and B. Haley. 1990. Synthesis and Properties of 2-Azido-NAD+ J.Biol. Chem. 265 P. 3636-3641.
21. Salvucci, M. E., A. J. Chavan, and B. E. Haley. 1992. Identification of peptides from the adenine binding domains of ATP and AMP in adenylate Kinase: isolationofphotoaffinity-labeled peptides by metal chelatechromatography. Bio-chemistry. 31:4479-4487.
22. Potter, R. L., and B. E.Haley.1983.Photoaffinity Labelingof Nucleotide BindingSiteswith8-azidopurine Analogs.MethodsEnzymol. 91:613-663.
23.Laemmli, U. K. 1970. Cleavage of structural proteins during theassembly
of the head ofBacteriophage4. Nature(Lond.). 227:680-685.
24. Bell, C. L., and P. M.Quinton. 1993. Regulation of CFTR Cl conductance in secretionby cellular energy levels. Am. J. Physiol. 264:C925-C931.
25. Faller, L. D. 1989.Competitive bindingof ATP and thefluorescent sub-strate analogue 2 ',3 '-0-
(2.4.6-Trinitropenylcyclohexadienylidine)adenosine5'-Triphosphate to thegastricH',K+ -ATPase: evidence for two classes of nucleo-tide sites.Biochemistry. 28:6771-6778.
26. Moczydlowski, E. G., and P. A. G. Fortes. 1981. Characterization of 2',3'-0-(2,4,6-trinitrocyclohexadienylidine)adenosine 5'-triphosphateas afluorescent probe of the ATP site of sodium andpotassium transport adenosinetriphosphatase.
J. Biol. Chem.256:2346-2356.
27.Mullen,G. P., P.Shenbagamurthi,and A.S. Mildvan. 1989. Substrate and DNAbindingtoa50-residuepeptide fragmentof DNApolymeraseI. J. Biol. Chem. 264:19637-19647.
28.Ko,Y.H.,P. J.Thomas,M. R.Delannoy,and P. L. Pedersen. 1993. The
cysticfibrosis conductanceregulator.Overexpression, purification,and character-ization of wild type andAF508mutantforms of the first nucleotidebindingfold in fusion with themaltose-bindingprotein. J. Biol. Chem. 15:24330-24338.
29.Travis,S. M., M. R.Carson,D. R.Ries,and M. J. Welsh.1993. Interaction ofnucleotides with membrane-associated cystic fibrosis transmembrane conduc-tanceregulator. J. Biol. Chem.268,21:15336-15339.
30.Schultz,B. D., C. J.Venglaric,R. J.Bridges,and R. A. Frizzell. 1993.
Regulation ofCFI'R Cl-channels by adenine nucleotide phosphates. FASEB
(Fed.Am.Soc. Exp. Biol.) J. 7:2464 (Abstr.).
31.Ambudkar,S. V., I. H. Lelong, J. Zhang, C. 0.Cardarelli,M. M. Gottes-man, and I. Pastan. 1992. Partialpurificationand reconstitution of the human
multidrug-resistancepump:Characterization of the drug-stimulatable ATP
hydro-lysis.Proc.Nati.Acad. Sci. USA.89:8472-8476.
32.Gill, D. R., S. C. Hyde, C. F. Higgins, M. A. Valverde, G. M. Mintenig, and F. V. Sepulveda. 1992. Separationof drug transport and chloride channel functions of the humanmultidrugresistance P-glycoprotein. Cell. 71:23-32.
33. Sarkadi, B., E. M. Price, R. C. Boucher, U. A. Germann, and G. A.
Scarborough.1992.Expression of the human multidrug resistance cDNA in insect cells generates a high activitydrug-stimulatedmembrane ATPase. J.BioL Chem. 267:4854-4858.
34. Mosser, J. A-M. Douar, C-O. Sarde, P. Kioschis, R. Feil, H. Moser, A-M.Poustka,J-L.Mandel, and P.Aubourg. 1993. Putative x-linked
adrenoleuko-dystrophy gene shares unexpected homology with ABC transporters. Nature
(Lond.).361:726-730.
35. Saraste, M., P. R. Sibbald, and A. Wittinghofer. 1990. The P-loop-a commonmotif in ATP- and GTP-binding proteins. TIBS (Trends Biochem.Sci.).
15:430-434.
36.Walker,J.E.,M.Saraste,M.J.Runswick, and N. J. Gar. 1982. Distantly related sequences in the a- and13-subunitsof ATP synthase,myosin, kinases and otherATP-requiringenzymes and a common nucleotide binding fold. EMBO (Eur.Mol.BioL Organ.)J.8:945-951.
37. Shyamala, V., V. Baichwal, E.Beall, and G. Ames. 1991. Structure-function analysisofthe histidine permease and comparison with cystic fibrosis mutations. J.Biol.Chem. 266:18714-18719.
38.Gibson,A.L.,L. M.Wagner,F. S.Collins,and D. L. Oxender. 1991. A bacterial system forinvestigatingtransport effects of cysticfibrosis-associated mutations. Science (Wash. DC). 254:109-111.
39.Cutting,G.R.,L. M.Kasch,B.J.Rosenstein, L.-C. Tsui, H. H. Kazazian, Jr., and S. E. Antonarakis. 1990. Two patients with cystic fibrosis, nonsense mutations in eachcysticfibrosis gene, and mild pulmonary disease. N.Engl.J. Med.323:1685-1689.
a-factor transporter STE6, a member of the ATP binding cassette (ABC) superfam-ily. EMBO(Eur. Mol. Biol. Organ.) J. 10:3777-3785.
41.Teem, J. L., H. A. Berger, L. S.Ostedgaard,D. P. Rich, L.-C. Tsui, and M. J. Welsh. 1993. Identification of revertants for the cystic fibrosisAF508 mutationusing STE6-DFTR chimeras in yeast.Cell.73:335-346.
42. Bishop, L., R. Agbayani, Jr., S. V. Ambudkar, P. C. Maloney, and G. F.-L. Ames. 1989. Reconstitution of a bacterial periplasmic permease in proteo-liposomes and demonstration of ATP hydrolysis concomitant with transport. Proc.
NatL.Acad. Sci. USA. 86:6953-6957.
43.Mimack,M.L.,M. P.Gallagher,S. R. Pearce,S. C. Hyde,I. R.Booth,and C. F.Higgins. 1989. Energy coupling toperiplasmicbinding protein-dependent transportsystems: Stoichiometry of ATP hydrolysis during transport in vivo. Proc.
Nadl.Acad. Sci. USA. 86:8257-8261.
44. Lehrman, M. A., J. L. Goldstein, D. W. Russell, and M. S. Brown. 1987. Duplicationof seven exons in LDL receptor gene caused by Alu-Alu recombination in a subjectwith familialhypercholesterolemia. CelL
48:827-835.
45. Bunn, H. F., and B. G. Forget. 1986.Hemoglobinopathydue to abnormal
oxygenbinding. InHemoglobin:Molecular, Genetic and ClinicalAspects,W. B. Saunders, Inc. editors.Philadelphia,PA.595-622.
46.McKusick,V.1989. Genetic map of the human genome autosomes,and,
x, y, and mitochondrial chromosomes. In The Metabolic Basis of Inherited Dis-eases,6th ed. C. R.Scriver,A.L. Beaudet, W. S.Sly,and D.Valle,editors. New York.73-163.
47. Knowles, M., J. Gatzy, and R. Boucher. 1981. Increased bioelectric poten-tialdifference across respiratory epitheliain cystic fibrosis. N. Engl. J.Med. 305:1489-1495.
48. Knowles, M.-R., L. L. Clarke, and R. Boucher. 1991. Activation by
extracellular nucleotides of chloride secretion in the airwayepitheliaofpatients
with cystic fibrosis. N. Engl. J. Med.325:533-538.
49. Diederichs, K., and G. E. Schulz. 1990. Three-dimensional structure of thecomplexbetween themitochondrial matrixadenylatekinase anditssubstrate AMP.Biochemistry.29:8138-8144.
50. Tsai, M.-D.,and H. Yan. 1991. Mechanism ofadenylatekinase: Site-directedmutagenesisversusX-ray and NMR. Perspect. Biochem.30:6806-6818.
51.Carson,M. 1987. Ribbon models of macromolecules. J. Mol. Graphics.