A productive NADP+ binding mode of ferredoxin–NADP + reductase revealed by protein engineering and crystallographic studies

letters

nature structural biology • volume 6 number 9 • september 1999 847

A productive NADP+ bindingmode of ferredoxin–NADP+

reductase revealed byprotein engineering andcrystallographic studiesZhan Deng1, Alessandro Aliverti2, Giuliana Zanetti2,Adrián K. Arakaki3, Jorgelina Ottado3, Elena G.Orellano3, Nora B. Calcaterra3, Eduardo A. Ceccarelli3,Néstor Carrillo3 and P. Andrew Karplus1,4

1Section of Biochemistry, Molecular and Cell Biology, Cornell University,Ithaca, New York 14853, USA. 2Dipartimento di Fisiologia e BiochimicaGenerali, Università degli Studi di Milano, Via Celoria 26, 20133 Milano,Italy. 3Molecular Biology Division, PROMUBIE, CONICET, Facultad deCiencias Bioquímicas y Farmacéuticas, Universidad Nacional de Rosario,Suipacha 531. 2000 Rosario, Argentina. 4Department of Biochemistry andBiophysics, Oregon State University, Corvallis, Oregon 97331, USA.

The flavoenzyme ferredoxin–NADP+ reductase (FNR) cat-alyzes the production of NADPH during photosynthesis.Whereas the structures of FNRs from spinach leaf and acyanobacterium as well as many of their homologs have beensolved, none of these studies has yielded a productive geome-try of the flavin–nicotinamide interaction. Here, we showthat this failure occurs because nicotinamide binding to wildtype FNR involves the energetically unfavorable displace-ment of the C-terminal Tyr side chain. We used mutants ofthis residue (Tyr 308) of pea FNR to obtain the structures ofproductive NADP+ and NADPH complexes. These structuresreveal a unique NADP+ binding mode in which the nicoti-namide ring is not parallel to the flavin isoalloxazine ring,but lies against it at an angle of ∼ 30°, with the C4 atom 3 Åfrom the flavin N5 atom.

Ferredoxin–NADP+ reductase (FNR, E.C. 1.18.1.2) is an FAD-containing enzyme that catalyzes electron transfer betweenNADP(H) and ferredoxin. In chloroplasts and in vegetative cellsof cyanobacteria, the production of NADPH catalyzed by FNR isthe final step in the photosynthetic electron transport chain.FNRs have also been found in various tissues and organisms inwhich they are involved in non-photosynthetic processes such asnitrogen fixation and steroid hydroxylation1. The three-dimen-sional structures of FNRs from spinach leaf2,3 and the cyanobac-terium Anabaena PCC 71194 have been determined. They bothconsist of two distinct domains, one for binding the prostheticgroup FAD and the other for binding NADP+. This unique two-domain motif has provided a structural prototype for a largefamily of flavoenzymes5. Structurally characterized members ofthis family include flavodoxin reductase6, cytochrome b5 reduc-tase7, the FAD-containing fragment of NADH-dependent nitratereductase8, phthalate dioxygenase reductase (PDR)9, andNADPH-cytochrome P450 reductase10. While all these struc-tures show the characteristic two-domain FNR motif as theirfundamental building block, some members also contain extradomain(s) that can extend their catalytic capability.

How the NADP+ substrate binds to FNR and other familymembers, however, has remained elusive, despite extensiveexperiments on a variety of enzymes and different crystallo-graphic approaches. In the studies of spinach leaf and Anabaena

FNRs, relevant binding information was obtained only for the2'-phospho-AMP (2'-P-AMP) portion of NADP+, whereas thefunctional nicotinamide portion was either not visible2,3, or toofar away from the flavin to be functionally relevant4. Moreover,despite many efforts, no productive NADP+ bound structure hasbeen obtained with any of the other FNR family members.

Why has productive nicotinamide binding been sodifficult to obtain?Thus far, productive pyridine dinucleotide binding structureshave been reported for two flavoenzyme families that are func-tionally and structurally different from the FNR family. One isrepresented by glutathione reductase11,12 and its homolog NADHperoxidase13, and the other is represented by quinone reductase14.In the first two of these structures, the nicotinamide ring stacksparallel to the isoalloxazine ring, with the NADP+ C4 atom adja-

Fig. 1 Spectroscopic characterization of wild type, Y308S and Y308W peaFNRs. a, Visible spectra of the FNRs as purified, wild type (thin line), Y308S(thick line), Y308W (dashed line). b, Difference spectra of wild type (thinsolid line), Y308W (thick dashed line) and Y308S due to NADP+ binding.The two difference spectra of Y308S are based on titrations of the proteinas purified (thick solid line) and the NADP+-deficient protein (thin dottedline). c, Anaerobic photoreduction of wild type and Y308S FNRs (∼ 13 µM)in the presence of an equimolar amount of NADP+. Spectra of partiallyreduced enzymes were recorded after ∼ 180 s of illumination, when thelong-wavelength absorption band was maximal. Wild type FNR, before(thin dotted line) and after illumination (thin solid line); Y308S FNR, before(thick dashed line) and after illumination (thick solid line).

a

b

c

© 1999 Nature America Inc. • http://structbio.nature.com©

199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://s

tru

ctb

io.n

atu

re.c

om

letters

cent to the flavin N5 atom. In the quinone reductase structure,the nicotinamide is also parallel to the isoalloxazine, but unlikethe former enzymes, it is not adjacent to the central isoalloxazinering. These precedents guide current views of flavin–nico-tinamide hydride transfer mechanisms.

In the structures of FNR and other closely related flavoen-zymes, the side chain of an aromatic residue lies close and paral-lel to the re-face of the flavin2–4,6,9. In plant-type FNR this aminoacid is the C-terminal residue and is conserved as a tyrosine. Byanalogy to the known flavoenzyme complexes, and consistentwith the 2'-P-AMP binding mode as well as the stereospecificityof hydride transfer, the nicotinamide is thought to occupy thesame place as the C-terminal tyrosine (Tyr 314 in spinach FNR)in the FNR structure2–4. Thus, NADP+ binding must involvesome structural rearrangements in which the C-terminal Tyrmoves to make way for the nicotinamide ring.

Attempts to obtain productive NADP+ binding to wild typespinach FNR by growing crystals under various pH and salt con-ditions were unsuccessful (C.M. Bruns & P.A.K., unpublished

848 nature structural biology • volume 6 number 9 • september 1999

results), calling into question our hypothesis2 that buffer condi-tions in the crystals or constraints caused by the crystal latticeinterfered with NADP+ binding. Attempts to detect the confor-mational change of the C-terminal Tyr by tetranitromethanemodification in the presence of NADPH were unsuccessful(L. Bayer, D. Schott & P.A.K., unpublished results) as wereattempts to cocrystallize with NADP+. Thus, we developed a newhypothesis that the crystallographic structure of the complextruly reflected the actual binding mode of NADP+ under physio-logical conditions. Specifically, we hypothesized that the ther-modynamics of NADP+ binding (both in solution and in thecrystal) is such that the 2'-P-AMP half binds tightly, whereas thenicotinamide mononucleotide (NMN) half is largely disorderedbecause the energetic cost of displacing the Tyr outweighs theenergetic gain due to nicotinamide binding. We further postulat-ed that productive binding of NADP+ may be enhanced by mak-ing the thermodynamics of NMN binding more favorablethrough the removal of the C-terminal Tyr of FNR. To test theseideas, we chose to study pea FNR because five mutants of the C-

Fig. 2 NADP+ binding mode in Y308S–NADP+

complex. a, Stereo view of the electron densityfor the complete NADP+ molecule in theY308S–NADP+ complex. The 1.8 Å resolution2Fo-Fc map is calculated using phase informa-tion from the final model and contoured at1.8 σrms. b, Ribbon diagram of the pea FNRY308S–NADP+ complex. The N-terminal FADdomain and the C-terminal NADP+ domain arecolored as teal and red, respectively. The pros-thetic group FAD (yellow) and the substrateNADP+ (green) are represented as ball-and-stick models. The signature sequence Gly-richpyrophosphate binding loop2 of the NADP+

domain is stippled. c, Stereo view of the nicoti-namide binding site in the Y308S–NADP+ com-plex. For clarity reasons, only the isoalloxazineand ribityl portion of FAD and the NMN part ofNADP+ are shown.

a

c

b


199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://s

tru

ctb

io.n

atu

re.c

om

letters

terminal Tyr of this enzyme (Tyr 308) had already been generat-ed to probe the role of this residue15,16. Here, we report the struc-tural characterization of the Ser (Y308S) and Trp (Y308W)mutants of pea FNR, and reveal the productive binding mode ofnicotinamide in the FNR active site.

Tyr 308 mutants, especially Y308S, display enhancednicotinamide bindingComparison of visible spectra of the wild type, Y308S andY308W FNRs (Fig. 1a) revealed that the Y308S spectrum wasstrongly red-shifted. Further studies provided indirect evidencethat this spectral shift occurred because the purification protocolyielded Y308S as a complex with NADP+. First, the Y308S spec-trum was not perturbed upon NADP+ addition (Fig. 1b), andsecond, during photoreduction of this mutant, spectral bandstypical of charge-transfer complexes between flavin andNADP(H), that is, FAD–NADPH (CT1) and FADH2–NADP+

(CT2),were observed (Fig. 1c). The bound ligand could be dis-placed from FNR Y308S by gel filtration in the presence of thecompetitive inhibitor 2'-P-AMP, and identified as NADP+ on thebasis of the increase of absorbance at 340 nm upon addition ofyeast glucose-6-phosphate dehydrogenase and glucose-6-phos-phate to the gel filtration fractions. After this removal procedure,


the spectrum of Y308S became similar to that of the wild typeenzyme, and the original spectrum was regenerated by additionof NADP+. These results imply that FNR Y308S binds NADP+

from the Escherichia coli cytosol tightly enough to copurify withthe enzyme.

We have shown that for spinach FNR, NADP+ titration leadsto spectral changes that indicate how NADP+ is binding17. In par-ticular, the positive difference peak around 500 nm is due to thenicotinamide–flavin interaction and its height correlates withthe extent of the nicotinamide ring occupancy in the active site.The difference spectra for the three pea FNR variants in complexwith NADP+ are qualitatively similar, but differ greatly in themagnitude of the peak around 500 nm (Fig. 1b). We interpret thequalitative similarity of the spectra to indicate that all three com-plexes have a similar geometry for the nicotinamide–flavin inter-action, with the larger magnitudes for Y308W and Y308Sindicating increased binding of the nicotinamide moiety. If theY308S difference spectrum is assumed to correspond to 100%occupancy, then those of the Y308W and the wild type enzymescorrespond to 35% and 10% occupancy, respectively. Therefore,the unperturbed geometry and high occupancy of the nicotin-amide binding in Y308S made it a perfect candidate for structur-al analysis.

Fig. 3 NADP(H) binding to pea FNR mutants. a, b, Electron density for the nicotinamide C4 atom. 2Fo-Fc maps for the Y308S–NADP+ complex (a) andthe Y308S–NADPH complex (b) are both contoured at 2.5 σrms. Unrestrained B-factor refinements of Y308S–NADP+ indicate that the weaker densityfor the C4 atom (B = 36 Å2), but not the carboxamide oxygen (B = 21 Å2), appears to be due to higher mobility. Such mobility is consistent with fluc-tuations involving boat-like conformations that would facilitate hydride transfer22. c, The Y308W–NADP+ complex at the nicotinamide binding pock-et. Shown in stereo view are models for both the nicotinamide and the Trp 308 side chain occupying the same site. The 2Fo-Fc omit electron densitymap is calculated using phase information from the final model with both NADP+ and Trp 308 omitted, and contoured at 1.2 σrms. d, Superimpositionof the Y308S–NADP+ and Y308S–NADPH complexes. The Y308S–NADP+ model (dark red) was overlaid onto the Y308S–NADPH model (teal) based ontheir Cα atoms. 2Fo-Fc electron density for Y308S–NADPH contoured at 2.0 σrms is shown as blue cages, to illustrate how the density clearly indicatesthat in Y308S–NADPH structure the flavin tilts so that the C1 ribityl atom is ∼ 0.3 Å farther out of the plane of the flavin. The distances (Å) betweeninteracting atoms are labeled, with their colors matching their own models. The largest structural difference involves the loss of a solvent molecule(WAT 1096) and the shift of the nicotinamide ribose in the Y308S–NADPH structure.

a b

d

c


199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://s

tru

ctb

io.n

atu

re.c

om

letters

Structure of the Y308S–NADP+ complexCrystallization of the Y308S mutant with or without the additionof NADP+ led to isomorphous crystals whose structure wassolved by molecular replacement. Consistent with the differenceobserved in the spectroscopic analyses in solution (Fig. 1a),Y308S crystals have an unusual orange-yellow color instead ofbright yellow. Electron density maps for both crystals clearlyshowed strong density for a complete NADP+ molecule (Fig. 2a),confirming that NADP+ is the copurified ligand, and providingdirect evidence that the nicotinamide moiety does bind at near100% occupancy.

The crystal structure of the Y308S–NADP+ complex has beenrefined at 1.8 Å resolution and includes two molecules in theasymmetric unit that are highly similar in their active site regionsand NADP+ binding geometry. The overall structure of this com-plex is equivalent to those of wild type spinach leaf FNR2,3 andAnabaena FNR4, with no significant changes in the relative ori-entation of the FAD- and NADP+-binding domains (Fig. 2b).The 2'-phosphoryl group of NADP+ interacts with the sidechains of Ser 228, Arg 229, Lys 238 and Tyr 240, and in generalthe 2'-P-AMP portion of NADP+ binds as has already beendescribed for spinach FNR2,3. Here, we will only focus on thenovel information dealing with the binding of the NMN part ofNADP+ and the active center geometry.

A stereo view and key distances associated with the binding ofNMN portion of NADP+ are given in Fig. 2c and Table 1. TheNMN phosphate receives hydrogen bonds from Lys 110 Nζ inthe FAD domain and Thr 166 N on the glycine-rich loop. Thisinteraction with Lys 110 was anticipated by chemical modifica-tion and mutagenesis studies18,19. The ribose adopts a C2'-endoconformation and the glycosidic bond is in the anti conforma-tion20. The nicotinamide binds with its A side facing the re-faceof the central ring of the isoalloxazine, which is consistent withthe stereochemistry of hydride transfer reported for thisenzyme21. The C4 atom of nicotinamide and the N5 atom of FADare 3.0 Å apart, an arrangement that appears reasonable fordirect hydride transfer. In contrast to what was seen in glu-tathione reductase, NADH peroxidase and quinone reductase,the nicotinamide and the isoalloxazine rings are not parallel.Instead, they align at an angle of ∼ 30°, with the C4 atom of thenicotinamide the closest to the isoalloxazine ring, and the N1 ofthe nicotinamide the farthest (∼ 5 Å from the FAD C10 atom).The structure of the Y308S–NADP+ complex also reveals that the


side chains of Ser 90 and Glu 306 (equiva-lent to spinach FNR Ser 96 and Glu 312,respectively) both hydrogen bond with thenitrogen atom of the carboxamide group(Fig. 2c, Table 1). Finally, the sulfhydrylgroup of the well conserved Cys 266(Cys 272 in spinach FNR) is in van derWaals contact with the backside (B side) ofthe nicotinamide C4 atom. One interac-tion that cannot be present in the wildtype FNR is a hydrogen bond between Ser308 Oγ and the 2'-hydroxyl of the nicoti-namide ribose.

An interesting feature of the nicoti-namide electron density is that its C4 atomhas significantly lower density than therest of the ring atoms (Fig. 3a,b).Quantitatively, this was seen in the atomicB-factors. When the B-factor restraints forbonded atoms were turned off during a

test refinement, the C4 had a final B-factor of ∼ 36 Å2 while theother atoms had values in the range of 15–22 Å2. This suggeststhat the C4 atom is the most mobile locus of the nicotinamide,and is consistent with the expectation that during hydride trans-fer reactions that involve NAD(P), the nicotinamide ring under-goes a transitional boat-like conformational change22.

Structure of the Y308W–NADP+ complexThe 2.0 Å resolution structure of Y308W cocrystallized withNADP+ further supports our contention that the NADP+ bind-ing mode in the wild type FNR matches that observed in theY308S–NADP+ structure. Due to the preservation of an aromat-ic residue at its C-terminus, the binding of the nicotinamide por-tion to Y308W is expected to be less favorable than that to Y308S,but closer in geometry to wild type FNR. This is because the Trpside chain is better able to mimic any structural functions of theTyr side chain than is Ser. In the Y308W structure, the electrondensity clearly showed a mixture of two conformations in thenicotinamide binding pocket — one with the Trp side chain par-allel to the isoalloxazine ring, similar to the Tyr side chain in thewild type spinach FNR, and the other with the nicotinamide ringbound exactly as in the Y308S–NADP+ structure (Fig. 3c). Theoccupancy of the nicotinamide ring was refined as 40%, in rea-sonably good agreement with the spectroscopic result of 35%(Fig. 1b). The average distance between the Trp 308 indole ringand the flavin isoalloxazine ring is ∼ 3.3 Å. No other electron den-sity for the Trp 308 side chain that could be ascribed to an alter-nate position was observed. Thus, it appears that when theTrp 308 side chain is displaced it does not adopt a single well-ordered position. The absence of a well-ordered alternate posi-tion for Trp 308 and the similarity of nicotinamideconformations in the Y308S and Y308W complexes provide evi-dence that in wild type pea FNR, Tyr 308 has little influence onthe geometry of the bound nicotinamide. Table 1 lists the keyinteratomic distances seen in this complex in the nicotinamidebinding pocket region.

Structure of the Y308S–NADPH complexY308S–NADP+ crystals soaked aerobically in 100 mM NADPHchanged color from orange-yellow to blue-green. To identify thenature of this blue-green species, spectroscopic studies on wildtype and Y308S FNRs were performed under conditions mimic-king that of crystallization but without ammonium sulfate.

Table 1 Selected distances at the active site1

Y308S–NADP+ Y308S–NADPH Y308W–NADP+

S90 N FAD N5 3.2 (3.3) 3.2 (3.2) 3.2 (3.2)FAD N5 NADP(H) C4 3.0 (3.1) 3.2 (3.3) 2.9 (3.1)C266 Sγ NADP(H) C4 3.5 (3.4) 3.6 (3.5) 3.5 (3.6)S90 Oγ NADP(H) N7 3.0 (3.0) 3.1 (3.2) 2.9 (3.1)E306 Oε2 NADP(H) N7 3.0 (3.0) 3.3 (3.0) 2.9 (3.4)WAT 1148 NADP(H) O7 2.9 (2.9) 2.9 (2.8) 2.9 (2.9)S90 Oγ WAT 1074 2.6 (2.8) 2.9 (3.0) 2.5 (2.8)E306 Oε2 WAT 1074 2.8 (2.7) 2.7 (2.6) 2.6 (2.5)S308 Oγ NADP(H) O2* 3.2 (3.2) 3.2 (3.0) ———WAT 1096 NADP(H) C3* 3.2 (2.9) ——— 3.0 (2.8)K110 Nζ NADP(H) OP1N 3.0 (2.7) 3.3 (2.8) 3.4 (2.7)T166 N NADP(H) OP2N 3.0 (2.9) 3.0 (2.9) 3.0 (2.5)T166 Oγ NADP(H) OP2N 3.5 (3.4) 3.4 (3.4) 3.7 (3.3)

1Values are in Å, and those in parentheses represent distances between the equivalent atom pairsof the second monomer in the asymmetric unit.


199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://s

tru

ctb

io.n

atu

re.c

om

letters

Spectra similar to those obtained in the photoreduction experi-ments (Fig. 1c) were observed. In both cases, long wavelengthabsorbance bands are present, which are due to a mixture ofcharge-transfer complexes between flavin in different redoxstates and NADP(H)17,23. The long wavelength absorbance ismore intense in the mutant than in wild type FNR, indicatingthat the mutant stabilizes a larger fraction of charge-transfercomplexes. Since it was not possible to establish which is the pre-dominant form in the blue-green crystals, we will refer to thisblue-green species as a FAD–NADPH complex (CT1), becauseCT1 is the predominant form in the wild type FNR23.

The structure of the Y308S–NADPH complex at 1.7 Å resolu-tion is almost identical to that of Y308S–NADP+. The only largevariations are the disappearance of a weak solvent site(WAT 1,096) and a change in the nicotinamide-proximal ribose(Fig. 3d). More subtle changes at the active site include a tiltingof the entire isoalloxazine ring and an accompanying increase inthe distance between the NADPH C4 atom and the FAD N5atom from 3.0 Å to 3.2 Å. Two pieces of evidence suggest thatthese variations are genuine. First, the same changes areobserved in both independently refined monomers and second,the flavin shift involves the correlated motion of many atoms. Wespeculate that these shifts are all related to the accommodation ofthe extra hydrogen on the nicotinamide C4 atom (or the flavinN5 atom) as follows. Because of this hydrogen, the N5-contain-ing edge of the isoalloxazine ring tilts away from the nicoti-namide, with the flavin O2 and O4 atoms shifting by ∼ 0.4 Å (Fig.3d). This movement leads to a steric clash between atom O2 andthe already crowded WAT 1096 and expels this water moleculefrom its binding site. Finally, the absence of this water moleculemakes room between the nicotinamide-proximal ribose andFAD, allowing the ribose to relax towards the flavin.

In contrast to the Y308S–NADP+ complex, the nicotinamideC4 atom has stronger electron density, and its refined B-factor iscomparable to those of the other ring atoms (Fig. 3a,b). The fact


that the C4 atom is more ordered is consistent with our specula-tion that in Y308S–NADPH, this locus becomes more compactand has less room for thermal fluctuation due to the extra hydro-gen. Although we have no good estimate of the strain energyinvolved, the distortion of the ribityl C1-atom out of the plane ofthe flavin (Fig. 3d) is another indicator of the steric crowding inthe active site that could contribute to the catalysis of hydridetransfer.

NADP+, a bipartite ligand for FNRThe crystal structures of the pea FNR Y308S and Y308W in com-plex with NADP(H) provide the first view of a productive sub-strate binding mode for FNR and its homologs. Several lines ofevidence support the notion that this is a catalytically proficientconformation of the pyridine nucleotide and that the engineer-ing has improved the stability of the interaction without affect-ing its geometry. First, the overall features of the spectralperturbations induced by NADP+ and NADPH binding to Y308Smatch those of the wild type enzyme (Fig. 1b). Second, the crys-tals are catalytically active for FNR reduction since a fraction ofthe FAD molecules can be reduced by NADPH to produce CT2.Third, the geometries of Y308S–NADP+, Y308W–NADP+, andY308S–NADPH complexes are similar to one another despite thedifferences in redox state and mutations introduced. Four, the Aside of the nicotinamide faces the isoalloxazine ring, consistentwith the known stereochemistry of hydride transfer in FNR21.Finally, the C4 atom of the nicotinamide is located just 3 Å fromthe flavin N5 atom and has a plausible geometry for hydridetransfer (Fig. 2c, Table 1).

Our results indicate that NADP+ binding can be thought toinvolve nearly independent binding of the two halves of the pyri-dine dinucleotide, similar to what has been seen for glutathionereductase11. In wild type FNR, the 2'-P-AMP half of the moleculebinds tightly to provide an anchor for the cofactor (Kd = 2 µMfor spinach FNR)21 and determines the substrate specificity5,

Table 2 Statistics for data collection and refinement1

Diffraction Data Y308S–NADP+ Y308S–NADP+ Y308S–NADPH Y308W–NADP+

(room temp) (low temp) (low temp) (low temp)Unit cell a (Å) 51.3 50.3 50.4 50.4

b (Å) 110.6 110.7 110.0 110.3c (Å) 81.3 81.0 80.9 80.6β (°) 94.1 94.4 93.9 93.8

Resolution (Å) 2.6 1.8 1.7 2.0Measured reflections 75,705 277,996 357,117 224,800Unique reflections 25,834 65,999 92,562 57,761Completeness (%) 92.6 (85.2) 81.0 (66.9) 96.1 (90.6) 97.5 (96.2)Rmeas (%)2 8.7 (38.7) 6.4 (26.1) 5.8 (33.4) 5.9 (31.4)

RefinementResolution limits (Å) 8.0–2.6 8.0–1.8 8.0–1.7 8.0–2.0Rcryst (%) 18.1 19.5 20.1 19.8Rfree (%)3 23.6 25.6 23.5 26.4Overall B (Å2)4 -3.3, 3.9, -0.6, 1.5 14.6, -9.4, -5.3, 1.8 10.1, -6.2, -3.9, 1.1 20.6, -13.3, -7.3, 2.7r.m.s.d. bond lengths(Å) 0.018 0.015 0.019 0.014r.m.s.d. bond angles (°) 2.1 1.8 1.9 1.7Number of waters 116 721 742 730

1Values for the highest resolution bins are shown in parentheses.2Rmeas = Σh (nh/nh-1)1/2 Σi |Ih –Ih,i | / Σh Σi Ih,i (ref. 41)3Rfree was calculated against reflections, 5% of the total, that were not used in the refinements42.4The anisotropic B-factor tensor parameters (B11, B22, B33 and B13) to scale the calculated to the observed structure factors.


199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://s

tru

ctb

io.n

atu

re.c

om

letters

whereas the functional NMN portion is largely (∼ 90%) disor-dered. A simple model to account for this two-step bindingmechanism has been provided by Strickland et al.24.

FNR + NADP+ ↔ FNR–NADP+(A) ↔ FNR–NADP+(AN) (1)

In which FNR–NADP+(A) represents the binary complex withthe dinucleotide bound only through the 2'-P-AMP portion,whereas FNR–NADP+(AN) would involve interactions of theNMN half as well. In this model, mutations at the C-terminal Tyrare expected to change the rate constants of the second equilibri-um, favoring formation of FNR–NADP+(AN). Unfortunately,the interaction between FNR and NADP+ is too fast to beamenable to rapid reaction measurements25, but the levels ofnicotinamide occupancy determined by difference spectroscopyand by crystallographic data for the NADP+ complexes of thevarious FNR forms allow us to calculate the ratios ofFNR–NADP+(A):FNR–NADP+(AN) as 9 (90:10), 1.9 (65:35),and < 0.11 (<10:>90), for wild type, Y308W and Y308S, respectively. These ∼ 5- and 80-fold increases in nicotinamidebinding in the mutants with respect to wild type enzyme corre-spond to binding free energy increases of ∼ 0.9 kcal mol-1 and>2.6 kcal mol-1 for Y308W and Y308S, respectively.

An overall Kd for NADP+ binding that includes both ligandedcomplexes of Eq. 1 was measured by spectrophotometric titra-tions of the FNR forms with NADP+. Kd values for wild type andY308W FNRs of 9 and 5.5 µM, respectively, were obtained. Dueto the much higher affinity of FNR Y308S for NADP+, an accu-rate measure of the Kd by this technique was not feasible — onlyan upper limit value of 0.2 µM could be estimated. Given that thetight binding of NADP+ to the Ser mutant is counterproductiveto catalysis, it is clear that one important function of Tyr 308 incatalysis is to down-modulate the affinity for NADP+ by makingan even stronger interaction with the nicotinamide-bindingpocket. Alternatively, modulation could also occur by decreasingthe affinity for the 2'-P-AMP half of NADP+, however, this wouldhave the disadvantage of lowering the ability to discriminatebetween NADP+ and NAD+.

The bipartite nature of NADP+ binding highlights an ambigu-ity of spectroscopically based assessments of structural interac-tions of extended ligands such as NADP+. The weak differencespectrum for NADP+ binding to wild type FNR is not due to aweak electronic interaction between flavin and the nicotinamideportion of NADP+. It rather reflects an interaction that has alarge intrinsic signal in the difference spectrum but occurs foronly 10% of the molecules. More generally, this result illustratesthat it is not always correct to assume that when an extended lig-and is present at concentrations high enough to saturate a bind-ing site, the entire site will be interacting with all portions of thebound ligand.

Roles of active site residuesFour important active site residues in FNR whose roles havebeen studied by site-directed mutagenesis are (spinach FNRnumbering in parentheses) Ser 90 (Ser 96)26, Cys 266(Cys 272)27, Glu 306 (Glu 312)17,28, and Tyr 308 (Tyr 314)15,16.Results presented here confirm that Ser 90 and Glu 306 are pri-marily involved in nicotinamide binding rather than beingdirectly involved in the hydride transfer reaction. Cys 266 is inclose contact with the nicotinamide C4 atom, and in theNADPH complex it acts to sterically force the nicotinamideagainst the flavin. Also, the presence of Cys 266 as a backstopcould influence the conformational freedom of the nicoti-


namide, strongly favoring the adoption of boat conformationsthat pucker toward the flavin and are productive for hydridetransfer, as postulated for lactate dehydrogenase22. Tyr 308occupies the nicotinamide binding pocket most of the time, andin this position may play a role in protecting the flavin fromreaction with oxygen. In addition, the low activities of Y308Sand Y308G had been interpreted to indicate an active role forthis residue15,29. However, based on our new results, we now sug-gest that the catalytic impairment is likely due to a decrease inthe rate constant for product release. We conclude that Tyr 308modulates the thermodynamics of nicotinamide binding, butdoes not appear to play an active role in influencing the confor-mation or reactivity of the bound nicotinamide. This contrastswith the more active role played by Tyr 197 in human glu-tathione reductase which is also displaced by nicotinamidebinding, but in the ‘out’ position, it acts as a spring to aid catal-ysis by pressing the nicotinamide against the flavin12,30.

Implications for the broader FNR familyIn the FNR superfamily, most members have an aromaticresidue equivalent to Tyr 308 stacked against the flavin.Furthermore, the equivalent position of Ser 90 is conserved asSer or Thr, Glu 306 as Glu or Asp, and Cys 266 as Cys6,9,10,29.Based on these sequence and structural (in known cases) simi-larities, we postulate that all these proteins adopt a very similarnicotinamide-binding mode. In contrast, two FNR superfamilymembers, nitrate reductase and cytochrome b5 reductase, donot have an aromatic residue at the position equivalent toTyr 3087,8. In these structures, a different domain interactionplaces the Gly-rich pyridine nucleotide pyrophosphate-bindingloop (Fig. 2b) much closer to the flavin, in a position that blocksnicotinamide binding. Therefore, functional binding of NADHin these proteins must involve a domain rearrangement.Consistent with this speculation, crystals of nitrate reductasedissolve when soaked in NADH8. For these two enzyme families,our work still provides the best model for nicotinamide binding,but the requirement for a distinct conformational change makesit unclear how many details will carry over.

In addition to providing a model on which to base mechanis-tic studies of FNR superfamily members, the structures report-ed here prove that having the nicotinamide ring strictly parallelto the flavin ring is not the only mode for hydride transfer reac-tion in NAD(P)-dependent flavoenzymes. With two experimen-tally-derived interaction geometries for flavin–nicotinamidehydride transfer now in hand, theoretical studies are needed tofurther assess their respective properties.

MethodsExpression and purification of ferredoxin–NADP+ reductasevariants. A full description of the procedures employed for cloning,mutagenesis and overexpression of the different pea FNR variants hasbeen reported15. Purification from E. coli JM109 was carried out asdescribed31, except that the gel filtration step was replaced by anionexchange chromatography, using a DEAE Macroprep column (2.5 X 50cm, BioRad) equilibrated in 50 mM Tris-HCl, pH 7.6 (Buffer A). The col-umn was extensively washed with the same buffer, and bound FNRwas eluted using a linear gradient from 0 to 0.4 M NaCl in Buffer A.The FNR fractions eluting near 275 mM NaCl were dialyzed against 10mM Tris-HCl, pH 7.6, and were concentrated by a DEAE Macroprep col-umn (0.5 X 10 cm, BioRad, equilibrated in Buffer A) eluted with300 mM NaCl in Buffer A.

To obtain the NADP+-free form of the Y308S mutant, aliquots ofFNR were titrated with 2'-P-AMP. The displaced ligand was removedby gel filtration through a Sephadex G-25 column equilibrated in10 mM Tris-HCl, pH 8.0 (at 4 °C), containing 3 mM 2'-P-AMP. FNR


199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://s

tru

ctb

io.n

atu

re.c

om

letters

fractions were chromatographed on an anion-exchange NeoBarAQ4 FPLC column (Dynochrom AS, Norway) equilibrated in 50 mMTris-HCl, pH 8.0 (at 25 °C). Elution with a linear gradient from 0 to 0.5 M NaCl yielded ligand-free FNR separated from theFNR–NADP+ complex which was retarded. After concentration byultrafiltration, the stripped Y308S was dialyzed against 10 mMTris-HCl, pH 8.0 (at 4 °C).

Spectroscopic analyses of the mutants. Extinction coefficientsof FNR forms were determined as described26. Difference spec-troscopy and photoreduction experiments were carried out asreported17.

Crystallization and data collection. Pea FNRs were stored at4.5 mg ml-1 in 5 mM potassium phosphate, pH 8. Crystals of Y308Swere grown at 25°C in hanging drops formed by mixing equal vol-umes of protein stock with a reservoir solution consisting of 2.1 Mammonium sulfate in 0.1 M Tris-HCl pH 8.0. Crystallization of Y308Swith or without addition of NADP+ gave identical results.Y308W–NADP+ cocrystals were also grown under the same condi-tions in the presence of 10 mM NADP+. These crystals are all isomor-phous and belong to space group P21 (Table 2). Room temperatureX-ray diffraction data were collected on a San Diego MultiwireSystem area detector with a Rigaku RU200 rotating anode (Cu-Kα)X-ray generator operated at 50 kV and 150 mA, and were mergedusing SCALEPACK32. Synchrotron diffraction data were collectedfrom frozen crystals at 100K on beamline F1 at CHESS. Before theywere mounted, the crystals were soaked in an artificial motherliquor (0.1 M Tris-HCl, pH 8.0, 2.9 M ammonium sulfate), and trans-ferred into a series of artificial mother liquors containing increasingconcentrations of glycerol up to 25% (v/v), then quickly frozenunder 100 K nitrogen stream. In order to obtain the blue-greencomplex, a Y308S crystal was soaked in artificial mother liquor con-taining 100 mM NADPH. The crystal gradually changed from yellowto blue-green within ~30 min, and the crystal was transferred intoartificial mother liquor containing glycerol and 100 mM NADPH,and then flash-frozen. The crystal stayed blue-green throughoutdata collection. Diffraction images were integrated using MOS-FLM33, and data were reduced, scaled using SCALA and TRUNCATEof the CCP4 program suite34.

Structure determination and refinement. Molecular replace-ment search was carried out with program AMoRe35 against theroom temperature Y308S data set ranging from 10–4 Å. A partiallyrefined wild type pea FNR monomer, which crystallized under dif-ferent conditions and belongs to a different space group(Z. Deng et al., unpublished results), was employed as the searchmodel. The wild type pea FNR structure showed that Tyr 308 in peaFNR is oriented similarly to Tyr 314 in spinach FNR, confirming therelevance of these experiments. The molecular replacement searchyielded two distinct solutions representing two monomers in theasymmetric unit. After rigid body refinement, the correlation coeffi-cient and R-factor were 0.68 and 0.34 respectively. An initial differ-ence electron density map clearly showed the presence of a wholeNADP+ molecule. All crystallographic refinements were carried outusing X-PLOR36 and included corrections for anisotropic diffraction(Table 2). The final room temperature model (R-free = 0.236, R = 0.181) was further refined against the 1.8 Å resolution low tem-perature data set, without non-crystallographic symmetry (NCS)restraints. Solvent molecules were gradually included into the struc-ture at stereochemically sensible positions and with a differencedensity higher than 3.0 σrms and 2Fo-Fc density higher than 0.8 σrms.In the late stage of refinement, electron density clearly indicatedthat a peptide segment comprised of the first 13 N-terminalresidues of one monomer, whose equivalent was not visible in thespinach FNR structure, stretches out and binds nonspecifically to theinterface of another neighboring crystallographic symmetry-relateddimer. The final model consists of two monomers (monomer A,residues 1–308; monomer B, residues 14–308), two NADP+ mole-cules, a sulfate anion, a phosphate anion, and a total of 721 orderedwater molecules. This structure was then used as the starting model


for Y308S–NADPH as well as Y308W–NADP+ refinements. In the lat-ter case the occupancies of the Trp 308 side chain and the nicoti-namide were refined by varying their relative occupancies and thencarrying out restrained individual B-factor refinement until theygave reasonable B-factors. Data collection and refinement statisticsare shown in Table 2. Pictures were generated with CHAIN37,MOLSCRIPT38, XtalView39 and RASTER3D40.

Coordinates. Coordinates and structure factors have been deposit-ed in the Protein Data Bank with accession codes 1QFY for the lowtemperature Y308S–NADP+ model, 1QFZ for Y308S–NADPH com-plex, and 1QGA for Y308W–NADP+.

AcknowledgmentsWe thank L. Piubelli for performing some of the spectroscopic experiments, S.E.Ealick for the use of his area detector facility, and T.P. Begley and V. Massey forhelpful discussions. This work was supported by grants from the NSF to P.A.K.,from CONICET and FONCYT (Argentina) to E.A.C., and from MURST to G.Z. N.C.was a recipient of a John Simon Guggenheim Fellowship.

Correspondence should be addressed to P.A.K. email: [email protected]

Received 28 April, 1999; 11 June, 1999.

1. Arakaki, A.K., Ceccarelli, E.A. & Carrillo, N. FASEB J. 11, 133–140 (1997).2. Karplus, P.A., Daniels, M.J. & Herriott, J.R. Science 251, 60–66 (1991).3. Bruns, C.M. & Karplus, P.A. J. Mol. Biol. 247, 125–145 (1995).4. Serre, L. et al. J. Mol. Biol. 263, 20–39 (1996).5. Correll, C.C., Ludwig, M.L., Bruns, C.M. & Karplus, P.A. Protein Sci. 2, 2112–2133

(1993).6. Ingelman, M., Bianchi, V. & Eklund, H. J. Mol. Biol. 268, 147–157 (1997).7. Nishida, H. et al. Biochemistry 34, 2763–2767 (1995).8. Lu, G., Campbell, W.H., Schneider, G. & Lindqvist, Y. Structure (London) 2,

809–821 (1994).9. Correll, C.C., Batie, C.J., Ballou, D.P. & Ludwig, M.L. Science 258, 1604–1610

(1992).10. Wang, M. et al. Proc. Natl. Acad. Sci. USA 94, 8411–8416 (1997).11. Pai, E., Karplus, P.A. & Schulz, G.E. Biochemistry 27, 4465–4474 (1988).12. Karplus, P.A. & Schulz, G.E. J. Mol. Biol. 210, 163–180 (1989).13. Stehle, T., Claiborne, A. & Schulz, G.E. Eur. J. Biochem. 211, 221–226 (1993).14. Li, R., Bianchet, M.A., Talalay, P. & Amzel, L.M. Proc. Natl. Acad. Sci. USA 92,

8846–8850 (1995).15. Orellano, E.G., Calcaterra, N.B., Carrillo, N. & Ceccarelli, E.A. J. Biol. Chem. 268,

19267–19273 (1993).16. Calcaterra,N.B. et al. Biochemistry 34, 12842–12848 (1995).17. Aliverti, A., Gadda, G., Ronchi, S., and Zanetti, G. Eur. J. Biochem 198, 21–24

(1991).18. Aliverti, A., Lubberstedt, T., Zanetti, G., Herrmann, R.G., and Curti, B. J. Biol.

Chem. 266, 17760–17763 (1991).19. Aliverti, A. et al. J. Biol. Chem. 273, 34008–34016 (1998).20. Saenger, W. Principles of Nucleic Acid Structure, Springer-Verlag, New York

(1983).21. Ammeraal, R.N., Krakow, G. & Vennesland, B. J. Biol. Chem. 240, 1820–1823

(1965).22. Young, L. & Post, C.B. Biochemistry 35, 15129–15133 (1996).23. Batie, C.J. & Kamin, H. J. Biol. Chem. 261, 11214–11223 (1986).24. Strickland, S., Palmer, G. & Massey, V. J. Biol. Chem. 250, 4048–4052 (1975).25. Batie, C.J. & Kamin, H. J. Biol. Chem. 259, 8832–8839 (1984).26. Aliverti, A. et al. Biochemistry 34, 8371–8379 (1995).27. Aliverti, A. et al. Biochemistry 32, 6374–6380 (1993).28. Medina, M., Martinez-Julvez, M. Hurley, J.K., Tollin, G. & Gomez-Moreno, C.

Biochemistry 37, 2715–2728 (1998).29. Karplus, P.A. & Bruns, C.M. J. Bioenerg. Biomembr. 26, 89–99 (1994).30. Krauth-Siegel, R.L., Arscott, L.D., Schönleben-Janas, A., Schirmer, R.H. & Williams,

C.H., Jr. Biochemistry 37, 13968–13977 (1998).31. Serra, E.C., Carrillo, N., Krapp, A.R. & Ceccarelli, E.A. Protein Express. Purif. 4,

539–546 (1993).32. Otwinowski, Z. & Minor, W. Methods Enzymol. 276, 307–326 (1997).33. Leslie, A.G.W. CCP4 and ESF-EACMB Newsletters on Protein Crystallography 26

(1992).34. Collaborative Computational Project Number 4, Acta Crystallogr. D50, 760–763

(1994).35. Navaza, J. Acta Crystallogr. A50, 157–163 (1994).36. Brünger, A.T. X-PLOR, a system for crystallography and NMR, Version 3.1. Yale

Univ. Press, New Haven, CT (1992).37. Sack, J.S. J. Molec. Graphics 6, 224–225 (1988).38. Kraulis, P. J. Appl. Crystallogr. 24, 946–950 (1991).39. McRee, D.E. J. Molec. Graphics 10, 44–46 (1992).40. Merrit, E.A. & Murphy, M.E.P. Acta Crystallogr. D50, 869–873 (1994).41. Diederichs, K. & Karplus, P.A. Nature Struct. Biol. 4, 269–275 (1997).42. Brünger, A.T. Nature 355, 472–475 (1992).


199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://s

tru

ctb

io.n

atu

re.c

om

A productive NADP+ binding mode of ferredoxin–NADP + reductase revealed by protein engineering and crystallographic studies

Documents