Differences in Protein Structure of Xanthine Dehydrogenase and Xanthine Oxidase Revealed by Reconstitution with Flavin Active Site Probes*

The native flavin, FAD, was removed from chicken liver xanthine dehydrogenase and milk xanthine oxidase by incubation with CaC12. The deflavoenzymes, still retaining their molybdopterin and iron-sulfur prosthetic groups, were reconstituted with a series of FAD derivatives containing chemically reactive or en- vironmentally sensitive substituents in the isoalloxa-zine ring system. The reconstituted enzymes contain- ing these artificial flavins were all catalytically active. With both the chicken liver dehydrogenase and the milk oxidase, the flavin 8-position was found to be freely accessible to solvent. The flavin 6-position was also freely accessible to solvent in milk xanthine oxidase, but was significantly less exposed to solvent in the chicken liver dehydrogenase. Pronounced differences in protein structure sur-rounding the bound flavin were indicated by the spec- tral properties of the two enzymes reconstituted with flavins containing ionizable “ O H or “SH substit- uents at the flavin 6- or 8-positions. Milk xanthine oxidase either displayed no preference for binding of the neutral or anionic flavin (8-OH-FAD) or a slight preference for the anionic form of the flavin (6-hy- droxy-FAD, 6-mercapto-FAD,


Differences in Protein Structure of Xanthine Dehydrogenase and Xanthine Oxidase Revealed by Reconstitution with Flavin
Active Site Probes* (Received for publication, November 22, 1988) Vincent MasseyS and Lawrence M. Schopfer The native flavin, FAD, was removed from chicken liver xanthine dehydrogenase and milk xanthine oxidase by incubation with CaC12. The deflavoenzymes, still retaining their molybdopterin and iron-sulfur prosthetic groups, were reconstituted with a series of FAD derivatives containing chemically reactive or environmentally sensitive substituents in the isoalloxazine ring system. The reconstituted enzymes containing these artificial flavins were all catalytically active. With both the chicken liver dehydrogenase and the milk oxidase, the flavin 8-position was found to be freely accessible to solvent. The flavin 6-position was also freely accessible to solvent in milk xanthine oxidase, but was significantly less exposed to solvent in the chicken liver dehydrogenase.
Pronounced differences in protein structure surrounding the bound flavin were indicated by the spectral properties of the two enzymes reconstituted with flavins containing ionizable " O H o r " S H substituents at the flavin 6-or 8-positions. Milk xanthine oxidase either displayed no preference for binding of the neutral or anionic flavin (8-OH-FAD) or a slight preference for the anionic form of the flavin (6-hydroxy-FAD, 6-mercapto-FAD, and possibly 8-mercapto-FAD). On the other hand, the chicken liver dehydrogenase had a dramatic preference for binding the neutral (protonated) forms of all four flavins, perturbing the pK of the ionizable substituent 24 pH units. These results imply the existence of a strong negative charge in the flavin binding site of the dehydrogenase, which is absent in the oxidase.
In the first paper of this series, we have described rapid reaction kinetics studies on the reduction of chicken liver xanthine dehydrogenase by xanthine and reduced pyridine nucleotides, as well as the reoxidation of the reduced enzyme by oxidized pyridine nucleotides (Schopfer et d . , 1988). In the second paper, we concentrated mainly on the catalytic reaction of the enzyme with molecular oxygen as acceptor, includ-DMB-88500291 (to V. M.), by Grant-in-aid 62480135 for scientific * This work was supported by National Science Foundation Grant research from the Japenese Ministry of Education, Science and Culture of Japan (to Takeshi Nishino), and by a research grant for intractable diseases from the Japanese Ministry of Health and Welfare (to Takeshi Nishino). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisernent" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ To whom correspondence should be addressed.
ing both rapid reaction and steady state kinetics studies (Nishino et d . , 1989a). In both of these publications, we drew attention to the similarities and differences found between the chicken liver xanthine dehydrogenase and the much studied milk xanthine oxidase (see Bray, 1975 andCoughlan, 1980, for extensive reviews on the properties of these enzymes).
In an attempt to define better the structural differences between the two proteins, which despite having similar molecular weights and identical cofactor composition (one molybdopterin, one FAD, and two Fe2/S2 centerslsubunit) have very different patterns of reactivity with pyridine nucleotides and molecular oxygen, we decided to employ the strategy of removing the native flavin, FAD, and replacing it with a series of artificial flavins with different redox potentials, as well as with flavins known to be useful as active site probes. The previous paper described the effects of such flavin replacements on catalytic activity, and in some instances, also described the changes in rapid reaction kinetics brought about by the flavin replacements (Nishino et d., 1989b). Similar studies have been carried out in the past with the milk oxidase (Hille et QL, 1981;Hille and Massey, 1985) and provided valuable experimental evidence in support of the rapid equilibrium hypothesis of Olson (Olson et d . , 1974a) that reducing equivalents taken into the enzyme (at the molybdenum center) from xanthine, are, after dissociation of the product, urate, rapidly distributed among the various redox centers on the basis of their reactive redox potentials. Similarly, it had also been proposed that in the reaction of the reduced enzyme with 02, it was the reduced flavin which reacted with 0 2 , and that the other redox centers served largely as electron sinks in rapid equilibrium with the flavin (Olson et d . , 1974b). The results which we have obtained with chicken liver xanthine dehydrogenase are also fully consistent with the rapid equilibrium concept, and add further experimental support to it.
One suggestive finding in the previous paper was that 6-OH-FAD binds to the chicken liver enzyme at pH 7.8 as the neutral flavin species (Nishino et d . , 1989b). In free solution, 6-OH-FAD has a pK of 7.1 (Mayhew et d . , 1974), so it was clear that the pK of the flavin was perturbed on binding to the protein, with stabilization of the neutral flavin species. In a previous study of 6-OH-FAD bound to milk xanthine oxidase (Hille et d . , 1981), the spectrum of the bound flavin was that of the anionic form. Since the 6-OH-FAD enzyme was studied only at pH 8.5, the finding of an anionic flavin spectrum was not surprising. However, it did raise the possibility of quite different protein environments around the flavin in the two enzymes. Thus it was of considerable interest to study the spectral properties of several ionizable flavins bound to chicken liver xanthine dehydrogenase, and to cornpare these with the same flavins bound to milk xanthine oxidase. It is found that while the milk oxidase has a slight preference for binding the anionic or benzoquinoid forms of 6-OH-FAD, 6-mercapto-FAD, 8-OH-FAD, and 8-mercapto-FAD, the chicken liver enzyme stabilizes strongly the neutral (protonated) forms of these flavins.
MATERIALS AND METHODS AND RESULTS AND DISCUSSION' CONCLUSIONS Milk xanthine oxidase and chicken liver xanthine dehydrogenase share many properties (see Coughlan 1980, for a review). Thus, they have similar molecular weights and each contains the same set of four redox centers, a molybdopterin, two separate Fe& centers, and FAD. Previous papers in this series have examined the rapid reaction kinetics of the chicken liver enzyme (Schopfer et al., 1988), the differences in O2 reactivity of the two enzymes (Nishino et al., 1989a) and the properties of chicken liver xanthine dehydrogenase substituted with a series of modified flavins of different redox potentials (Nishino et al., 1989b). The present paper describes the properties of both enzymes reconstituted with a series of chemically reactive flavins, or ones which are sensitive to the nature of the protein environment in which the flavin is embedded. The results show dramatic differences between the two enzymes which must reflect the nature of the flavinprotein interactions. In both enzymes the flavin 8-position appears to be exposed to solvent, since 8-mercapto-FAD forms of both react readily with iodoacetamide and MMTS.' In the case of milk xanthine oxidase, the flavin 6-position also appears to be quite accessible to solvent, since the 6-SCN-FAD enzyme is converted very rapidly to the 6-mercapto-FAD form with DTT, and since the 6-mercapto-FAD enzyme reacts readily with iodoacetamide, MMTS, H202, and Nethylmaleimide, and since the 6-S-S-CH3-FAD enzyme reacts rapidly with DTT to regenerate the 6-mercapto-FAD enzyme. On the other hand, the flavin 6-position in chicken liver xanthine dehydrogenase appears to be significantly less exposed. The 6-SCN-FAD enzyme does not react with DTT, and the 6-mercapto-FAD enzyme is markedly less reactive with iodoacetamide and the other thiol reagents tested than is 6-mercapto-FAD milk xanthine oxidase, and the 6-S-S-CH,-FAD enzyme reacts only very slowly with DTT. Since the 8-mercapto-FAD forms of both milk xanthine oxidase and chicken liver xanthine dehydrogenase react readily with iodoacetamide and MMTS, and since the 8-mercapto-FAD enzymes are regenerated rapidly from the 8-S-S-CHs-FAD forms with DTT, it follows that the different results found for the 6-substituted FAD enzymes are due to different accessibility of the flavin 6-position in milk xanthine oxidase and chicken liver xanthine dehydrogenase, rather than to different reactivities due to stabilization of the neutral flavin in one case (chicken liver xanthine dehydrogenase) and the anion in the other (milk xanthine oxidase).
By far the most dramatic differences between the two enzymes are shown with the FAD derivatives containing an ' Portions of this paper (including "Materials and Methods," "Results and Discussion," Figs. 1-11, Tables 1-111, and Scheme I) are presented in miniprint at the end of this paper. Miniprint is easily read with the aid of a standard magnifying glass. Full size photocopies are included in the microfilm edition of the Journal that is available from Waverly Press.
The abbreviations used are: MMTS, methylmethanethiosulfonate; XDH, chicken liver xanthine dehydrogenase; MXO, milk xanthine oxidase; DTT, dithiothreitol. ionizable " O H or " S H residue. While the milk oxidase has either no preference (eg. 8-OH-FAD and possibly 8-mercapto-FAD) or a slight preference for the anionic form (6-OH-FAD and 6-mercapto-FAD), the chicken liver dehydrogenase has a very strong preference for binding the neutral (protonated) forms of all four flavins, with perturbations of the pK by 2 4 p H units. The only other case that we are aware of where the neutral forms of these flavins are stabilized strongly is with hen egg white riboflavin-binding protein Massey, 1980;Ghisla et al., 1986;Ghisla and Massey, 1986). While there are several ways in which such stabilization might be achieved, the simplest explanation would be the existence of a negatively charged region in the protein, which prevented binding of the anionic flavin. This would appear the likely explanation with riboflavin-binding protein, which has been found to possess a cluster of negative charges (13/14 residues) in the amino acid sequence 186-199 (White and Merrill, 1988). With all four flavins, the negative charge of the anion can be distributed throughout the whole flavin ring system, with a predominant resonance form having the negative charge in the flavin N(l)-C(2=0) locus (Mayhew et al., 1974;Ghisla and Mayhew, 1976;Massey et al., 1979;Ghisla et al., 1986). As the flavin 8-position is exposed to solvent in chicken liver xanthine dehydrogenase, it is unlikely that there can be any negatively charged residue of the protein in this region; it is therefore more likely that the negative charge of the protein is located near the flavin N(1)-position in the dehydrogenase, and that this charge is absent in the oxidase. In the case of 8-mercapto-FAD milk xanthine oxidase, the wavelength maximum of the bound flavin (X,.,-579 nm) indicates that it is stabilized as the benzoquinoid anion ) (see Scheme I). Based on the position of this maximum, the extent of stabilization appears to be weaker than with some enzymes where a strong positive charge is thought to be located in the region of the flavin N(1)-position (eg. lactate oxidase, X , , , = 607 nm ; D-amino acid oxidase, A,,,,,= = 595 nm (Fitzpatrick and Massey, 1983)), but is similar to that found for 8-mercapto-FAD p-hydroxybenzoate hydroxylase, where the Amax is at 547 nm in the absence of substrate and 565 nm in the presence ofp-hydroxybenzoate . The stabilization of the anionic flavin forms of p-hydroxybenzoate hydroxylase has bee? correlated with a partial positive charge from the N-terminal portion of an a-helix of the protein directed toward the flavin N(1)position (Hofsteenge et al., 1980) and suggests a better orientation of this helix toward the flavin N(1)-position in the enzyme-substrate complex than in substrate-free enzyme. The present results with 8-mercapto-FAD milk xanthine oxidase suggest a similar interaction.
Differences in structure of chicken liver xanthine dehydrogenase and milk xanthine oxidase in the reduced state with the native FAD-containing enzymes is also evident from their different reactivities with iodoacetamide. Xanthine-reduced milk xanthine oxidase has been shown to react with iodoacetamide, resulting in alkylation of the reduced flavin at the C(4a)-position, yielding enzyme containing the catalytically inactive C(4a)-acetamido FAD (Komai and Massey, 1971). In contrast, chicken liver xanthine dehydrogenase reacted under the same conditions as those giving rapid inactivation with milk xanthine oxidase (0.1 M phosphate, pH 6.3, 1 mM xanthine or excess Na2S204, 1 mM iodoacetamide, 25 "C, under anaerobiosis) gave only marginal loss of activity over 70 min and no indication of alkylation of the enzyme flavin, as judged by the return of the original absorption spectrum on reoxid a t i~n .~ These results clearly imply that the flavin C(4a)-T. Nishino and V. Massey, unpublished results. position in substrate-reduced milk xanthine oxidase is considerably more accessible than in reduced chicken liver xanthine dehydrogenase (substrate-reduced or chemically-reduced).
The results with modified flavin forms of milk xanthine oxidase and chicken liver xanthine dehydrogenase would predict that in the reversible interconversion of xanthine dehydrogenase and xanthine oxidase, which can occur, for example, with the enzyme from rat liver (Stirpe and Della Corte, 1969;Rajagopalan, 1976a, 1976b) by oxidation and reduction of suitably positioned thiol residues (Della Corte and Stirpe, 1972;Waud and Rajagopalan, 1976b;Saito and Nishino, 1989), substantial changes in protein conformation must occur. Thus, the dehydrogenase form, with intact NADbinding site, and a strong negative charge located near the flavin N(1)-position, would change its conformation on oxidation of the crucial thiol residues in such a manner that the NAD-binding site was lost at the same time as the negatively charged protein residue was removed from the vicinity of the flavin N(l)-position, and replaced instead by a partial positive charge, such as could be provided from amide nitrogens or the N-terminal region of an a-helix. Evidence that such conformational changes may indeed occur in a reversible fashion with rat liver xanthine dehydrogenase has been obtained in a similar study using the same flavin derivatives as reported here, and will be described in full in a future p~blication.~  CaCI, 4H20 very sallsfactory. For stable, reproducible preparaliom of deflavo-nX0. it was also found desirable to employ enzyme prepared wilhoul the use Or pancrealin.
Wllh rnosl of the fiavlns employed in lhis study, the free f l a w has signlrlcanl long wavelenglh abSOrpt10n. dlfferent In the neutral and anlonlc forms The contrlbullon 01 the enzyme-bound flavin Io the ab5orp11on speclrum or Ihe reconstlluled enzyme was determined as foliows A small excess of flavln (-1.2-1.5 mol rlavinlmol dellavoenzyme) was incubaled wllh the dellavoenzyme and spectra were followed wlth tlme unlll no further chayes occurred. The mlxture was lhen centrifuged through an Amlcon CenlriCOn 30 ultraflllration rnembraw The spectrum or the filtrate was recorded, and sublracled from thal of the rnlxluro before ullrafillrallon, lo yleld the speclrm of Ihe reconstlluted enzyme Extinttion coefflcienl5 were then calculated based on the spectrum of Ihe startlng deflavoenzyme. u s l q a value of E465=25.2 mW1 cm" for deflavo-XDH (Nlshlm et. al., 1989a) Table I

Protein Structural Differences between XDH and MXO
aeflavo-XDH is that occurring WI M 8-mercapto-FAD. This is illustrated in Figwe 7, which mlxlng. the 535 m absorbance or rree 8-mercaplo-FAD disappears reasonably rast ft I/z shows the spectra or the two separate compownts. deflavo-XMI and 8-mercapto-FAD. On greater than lhat or the mltial deflavoonzp. The difrerewe spectrum between zmlm at 4' C), so that the final abs0rbBnCe In the 500-650 M) region IS only sllghtly reconstituted and deflavoomyme has a hmax at 456 m (see Table I) anb has the characterisilcs 01 wulral (protonated) 8-5H-FAD (noore et. al., 1979). The DK 01 the free flavin is 3.8 ). yet when bound to XDH at pH 7.8 the llavln IS clearly stlii fully protonated, k. the pK 28.3. a s h i r t of 2 4.5 units.
On anaerobic tllration with Xanthlw. (Ftgure 8) there is rapid bleaching Of both the flavin and F d 5 chromophores, wlth a pronounced accumulation of rlavln semiqulm. as judged by Ihe lmreased absorbawe in the 600 m regiw which disappears only on the addltion or dithJonttQ (Flgwe 8, inset). The s~m u l t~o u s loss in A550 and iwrease in A540 means that the redox ootentlal 01 the hlghest potential center, FeiS il, must be close to that o i the 8-SH-FAD&-SH-FADH couple. The ract that tho Increased absorbance at 590 m persists untlt the addlllon 01 dithionile suggests that the potentlai or the 8-5H-FAD+ /8-SH-FADH2 couple is ouite tow. possibly lower even than that of the low potential Iron center Fe/S I. No  in the previous w k . he spectral characteristics of the 8-mercapto-FAD onzyw were determinpd by reaction or the 8-CI-FAD enzyme wlth Na25 (Schopler et. al., 1981;Hille et. 81.. 1981) In the present study. WQ have reconstiluted deflavoenzyme with preformed 8-mercapto-FAD. aM determined the extinction coefficients or tk recwtiluted e n z w b!j the Cmtricon uitrafiltration method described tn 'naterlals and flethods". The results are idPntical with those previously published. with a At579 01 SI mn" cm-I In the present study nE580 or 30.5 mn-1 cm" reported by  Table I).
The spectral charactertstics of the bwnd llavln (hmau 444 MI. E=209 m"! Cm") are typical of those or the w t r a l flavin (Ghista m Mayhew. 1976) In free solution. 8-OH-FAD has a DK or 4.8 (Ghisla and flqhew. 1976); when bound to XDH the PK is 2 8.5 Thus. as In the case or 8-mercaplo-FAD. there is a rema-kable StaDlliZatiOn or the wutral form of the rtavmn with t e pK beirq perturbed by at least 3.5 pH wits.
Xanthiw-NAD redllctase actlvlty which IS *30% that of M t w e enzyme but very tow xanlhiw-As described in the prevlous paper (Nishino et. ai.. 198%). the 8-OH-FAD enzyme has O2 reductase activily (~10% thal of Mtive enzyme). In kewing wllh Ihls. the aeroblc addition of xanthine results in a partially rehlced SPeCtrum whlch stay5 at a c m t a n l Steady state level until the xanthinp 1s exhausted. There was phl obvtow presme of ftavln semlquiwne in such experimonts; I k dirrerence spectrum between the oxldized enzyme and aeroblc sleady state can be accomled lor by reduction Or 12 FelS ChomOphOreS and approx:mately 30% reduction or the enzyme-bound ftavm iresuits rwt shown). The neutral sem!quinow form of 8-OH-flavin has a high extwlion at 580 m, (€e8 M" cm" (GhlSla and nayhew. 1976)). and so should have been srgnirlcanl contrtbutor to thp steady state s~ectrum Heme. we can cMCiUde, as we have dow in Ihe previous papers (Nishlno el. ai.. 1989a and1989b) lhat the low xanthmm-02 reduclase acllvity or XDH IS not wcessarily associated wlth stabillzallon or Ihe llavln semiquinom. as proposed by previous workers (Barber et al, 1976(Barber et al, . 1980 B-HIJ~~OIIY-FAD nxo.     nost of the enzyme forms h a v~ substantla1 x a n t h i~-o x~~ reductase activlly. As has a redox potential equal lo or higher than that Of FAD. wyf low actlvlty IS shown when the flavin has a s~s t~t l a t l y lower potential. This has been ascribed lo a ~~n called 'tnermodywmlc control' (Palmer and Olson. 1980) and follows from the rapld equilibrlm hypoIhe5is or Olson et. at. (1974). A c c o r w q to thhs hypolhesls. a5 IhQ flavin polentlal 15 made more mqattve than that of tho ~l~~m center. r & u c q equivalents will associate wlfh the Mo rather than the flavin so that e m p e form5 conlainlng FAD and nolV will Dredomimte In turmver Instead of forms conlainlng noVl and FAN, As reduclng equivalents from xanthine require IM ~1~

Catalylic Activity ol Ruco~tilutod Enzpes
to be in the noVl slate and as FAOH? is ttm rlavln form reactlng rapidly with 02, the catalytlc actlvlty should therefore aecrease as ~~e n t l y 5erIm discrEQancq to thss COPrept with 4-thio-FAD enzyme, which instead of the flavin potential becows lower than that of the PloVlIflolV cowte. There IS ow having an activlty s~mllar to that of native enzyme. has only 30% the expected acllvity. It must be emphasized, however, that the values listed In Table 111 are those obtalwd only Concentrallons will be requlred before this polnt can be addressed properly.
i i & StWard assay condilioffi. and that fwll kiretic analyses varyiq both xanthine W O2 One notable feature of the results shown In Table 111 IS that both with 6-mercaplo-FAD nXO and 8-mrcapto-FAD NXO, the catalytic aclivtty is increased signmcanlly on reaction wlth iodoacetamlde and methylmethawthio~ulfomte. These reagents change the nature or the flavln substituent at lne 6-and 8-po51tioffi. W the lwrease in catalytic activlty is due Io these changes. Tho smngte exposed thiol resldue oi hative xanthine oxidaw inassey et. al.. 1970) reacts readily wlth methylmethanethiosUIfonate, but has no effect on the catalytlc activlty (results mt shown).

Catalytic Activities of ~c~t l t u t e a
Xanlhiw Oxidase a) Table 111 Enzyme recomlituted wlth: 'i) H~~e f l c h , P personal c o~n l c~t~o~ 0) Gnlsla and Mayhew, 1980 COnverSlon of the flavin to an unidentified product (noore e1 ai., 1979).