Proc. Nati. Acad. Sci. USAVol. 90, pp. 750-754, January 1993Biochemistry

Crystal structure of lignin peroxidaseSTEVEN L. EDWARDS*t, REETTA RAAG*, HIROYUKI WARHSHII, MICHAEL H. GOLDt,AND THOMAS L. POULOS*§¶*Center for Advanced Research in Biotechnology, 9600 Gudelsky Drive, Rockville, MD 20850; *Department of Chemical and Biological Sciences, OregonGraduate Institute of Science and Technology, Beaverton, OR 97006-1999; and §Departments of Molecular Biology & Biochemistry and Physiology &Biophysics, University of California, Irvine, CA 92717

Communicated by I. C. Gunsalus, September 18, 1992

ABSTRACT The crystal structure of lignin peroxidase(LIP) from the basidiomycete Phanerochaete chrysosporiumhas been determined to 2.6 A resolution by using multipleisomorphous replacement methods and simulated analigrefinement. Of the 343 residues, residues 3-335 have beenaccounted for in the electron density map, induding fourdisulfide bonds. The overall three-dimensional structure is verysimilar to the only other peroxidase in this group for which ahigh-resolution crystal structure is available, cytochrome cperoxidase, despite the fact that the sequence identity is only-20%, LiP has four disulfde bonds, while cytochrome cperoxidase has none, and LIP is larger (343 vs. 294 residues).The basic helical fold and connectivity defined by 11 helicalsegments with the heme sandwiched between the distal andproximal helices found in cytochrome c peroxidase is main-tained in LIP. Both enzymes have a histidine as a proximalheme ligand, which is hydrogen bonded to a buried asparticacid side chain. The distal or peroxide binding pocket also issimilar, including the distal arginine and histidine. The moststriking difference is that, whereas cytochrome c peroxidasehas tryptophans contacting the distal and proximal hemesurfaces, LIP has phenylalanines. This in part explains why, inthe reaction with peroxides, cytochrome c peroxidase forms anamino acid-centered free radical, whereas LIP forms a por-phyrin ir cation radical.

Lignin is the most abundant aromatic polymer on earth (1).It comprises 15-30% of woody plant cell walls, forming amatrix surrounding the cellulose. This matrix significantlyretards the microbial depolymerization of cellulose and,hence, lignin plays a key role in the earth's carbon cycle'(2,3). White rot basidiomycete fungi are the only known orga-nisms that are capable of degrading lignin to CO2 and H20 (2,3). The best-studied white rot fungus, Phanerochaete chry-sosporium, secretes two extracellular heme peroxidases,lignin peroxidase (LiP) and manganese peroxidase, which aremajor components of its lignin degradative system (2-5). LIPis a glycoprotein with a molecular mass of =41 kDa, contains1 mol of iron protoporphyrin IX per mol of enzyme, andexists as a series of isozymes of pI 3.2-4.0 (2, 3, 5, 6). LiPcatalyzes the H202-dependent oxidation ofa variety of ligninmodel compounds in the following multistep reaction se-quence:

LiP(Fe3+)P + H202 -- LiP-I(Fe4+-O)F + H20.

In this scheme, R is substrate and P is porphyrin. LiPcompound I (LiP-I) carries both oxidizing equivalents ofH202, one as an oxyferryl (Fe4+-O) center and one as aporphyrin ir cation radical (P ), whereas LiP compound II(LiP-I) carries only one oxidizing equivalent. The substrateR is oxidized by compound I to an aryl cation radical withsubsequent nonenzymatic reactions yielding the final prod-ucts (2, 3, 7-9). P. chrysosporium also is able to degrade avariety of aromatic pollutants (10-12). Overall, the LiPreaction cycle is very similar to that of horseradish peroxi-dase (13).

Despite the similarities and differences between LiP andother peroxidases, important questions regarding the mech-anism of LiP and its exact role in lignin degradation remainunanswered. For example, the unique capacity of LIP tooxidize nonphenolic compounds with high redox potentials(2, 3, 7-9), the basis for the low pH optimum near pH 3.0(14-17), and the sensitivity to excess H202 (18, 19) are notunderstood in detail. Moreover, the mechanism of degrada-tion by LiP of polymeric lignin (2, 3, 20) either directly orpossibly indirectly via a mediator also is not clearly under-stood. To elucidate the detailed molecular structure andmechanism of LiP, we have determined the crystal structureof this peroxidase. II

METHODS AND MATERIALSCrystalization. One of the major isozymes of LIP (LiP-2)

from P. chrysosporium strain OGC101 was purified accord-ing to earlier procedures (14, 19). Crystals were grown bymicroseeding in hanging drops using polyethylene glycol asthe precipitant at pH 4.5 (E. L. Winborne, S.L.E., M.H.G.,and T.L.P., unpublished data). The crystals belong tospace group P21 with a = 44.7 A, b = 77.5 A, c = 100.0 A, and(3 = 1010 with two LiP molecules per asymmetric unit.Intensity data were collected with a Siemens area detectorand Rigaku rotating anode x-ray source.

Structure Determination. The structure was solved byconventional multiple isomorphous replacement (MIR) pro-cedures, solvent leveling, and noncrystallographic averag-ing. A total of five heavy atom derivatives using threedifferent reagents was used. A single crystal was used foreach data set and all were >90%o complete to 3.0 A (Table 1).The initial MIR electron density at 3.0 A was not ofvery goodquality. We subsequently found that this was due to incom-plete identification of heavy atom sites. Nevertheless, non-crystallographic averaging (21) using a suite of programs

LiP-I(Fe4+-O)P + R -- LiP-II(Fe4+-O)P + RI

LiP-II(Fe4+-O)P + R + 2H+ -- LIP(Fe3+)P + R7 + H20

Abbreviations: LiP, lignin peroxidase; CCP, cytochrome c peroxi-dase; MIR, multiple isomorphous replacement; F. and Fc, observedand calculated structure factors, respectively.tPresent address: National Institutes of Health, Building 6, Room114, Bethesda, MD 20892.ITo whom reprint requests should be addressed at the University ofCalifornia.IThe atomic coordinates and structure factors have been depositedin the Protein Data Bank, Chemistry Department, BrookhavenNational Laboratory, Upton, NY 11973 (reference 1LGA).

750

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Dow

nloa

ded

by g

uest

on

Janu

ary

1, 2

020

Proc. Natl. Acad. Sci. USA 90 (1993) 751

Table 1. Summary of structure determinationI/al at

Maximum maximum Total Unique % RmergeData set resolution resolution observed data complete intensity

Native 2.03 2.2 123,543 38,748 98 0.074Uranyl 1 3.0 3.0 24,081 8,538 98 0.14Uranyl 2 2.5 2.6 26,440 18,252 80 0.08Platinum 1 2.2 1.9 79,357 26,829 90 0.09Platinum 2 2.7 2.8 51,322 26,215 82 0.07Mercury 3.0 3.0 52,782 13,619 99 0.12

written by J. K. Mohan Rao (Frederick Cancer ResearchInstitute) did improve the map. An estimate of how the twomolecules are related in the asymmetric unit was made fromthe relationship of heavy atom sites.The averaged MIR map allowed for a preliminary chain

tracing by using the published sequence (22). It becameevident during this stage that the overall fold of LiP closelyfollows that of cytochrome c peroxidase (CCP), the onlyother heme peroxidase for which a high-resolution structureis available (23) even though the sequence homology is <20o(24). Nevertheless, once it became clear that the helices inboth structures were similarly arranged, model building andrefinement proceeded more smoothly. The model was sub-jected to several rounds of restrained least-squares refine-ment using the fast Fourier transform version of PROLSQ (25,26). Sim weighting (27) was used to combine the probabilitydistributions of the MIR phases and calculated phases (28). Anew map using figure-of-merit weighted Fos and combinedcentroid phases then was used for the next round of modelbuilding. In the early stages ofphase combination, only thosereflections for which there were MIR phases were combined.

This procedure was continued for four rounds of modelbuilding, refinement, and phase combination. Nevertheless,refinement did not converge and several regions remaineddifficult to interpret. In an attempt to improve the initial MIRmap, derivative-parent difference Fouriers were reexaminedby using calculated phases based on the model, whichincluded -80%o of the expected protein atoms. As antici-pated, these maps confirmed the location of heavy atom sitesbut also indicated two more sites in the uranyl 1 and three inthe mercury derivatives. The occupancy of the new uranylsite was -0.28 after refinement to be compared with the twomain sites, which exhibited occupancies of0.53 and 0.56. Thenew mercury sites also yielded refined occupancies abouthalf that of the two main mercury sites. MIR refinement wasrepeated with the PHASES package of programs** with theseadditional sites leading to a higher figure of merit at 2.6 Aresolution (Table 2). This MIR map was considerably betterthan the original map. The newer MIR map was subjected tosolvent leveling with a 30% solvent cutoff resulting in anaverage change in acentric phase of 230. The leveled map wasnext averaged about the noncrystallographic twofold axis.For refinement of the noncrystallographic rotation/translation matrix, a 20-A box of density centered on the ironatoms was used since this region represented the most highlyordered and clearest parts of the MIR map. The correlationcoefficient after refinement was 0.56.The resulting orientation matrix then was used to average

volumes of densities corresponding to both LiP molecules.The model was rebuilt using the new MIR, leveled/averaged,and combined phased maps. Despite the low solvent cutoffdue to the relatively low solvent content of the crystals,solvent leveling did improve the map and allowed for suc-cessful interpretation of difficult regions. Molecule 1 in the

asymmetric unit was extensively rebuilt and the noncrystal-lographic rotation matrix was used to generate molecule 2.The model was subjected to 10 more rounds of restrainedleast-squares refinement, giving an R factor of 0.40.At this stage simulated annealing refinement using X-PLOR

(29) was used. One round of simulated annealing using theslow-cooling protocol recommended by Brunger (30) loweredR from 0.42 to 0.27. The procedure was repeated with astarting temperature of4000 K instead of3000 K, followed byrefinement ofan overall isotropic temperature factor. X-PLORwas most useful in fitting molecule 2 in the asymmetric unitand in obtaining a better orientation matrix that relatesmolecule 2 to molecule 1. As a result, model building becameless time consuming since only one molecule need be fit to theelectron density map, with the second generated by theorientation matrix, and X-PLOR was used to "force" a betterfit for molecule 2. The overall major advantage in usingX-PLOR was in the reduction of time spent fitting to theelectron density map.The final R factor is 0.25 for 19,356 reflections between 8

and 2.6 A and .-2a. Individual temperature factors have notbeen refined nor has solvent been included in the refinement.The "annealed" model exhibited excellent geometry withdeviations ofbond distances of 0.027 A and the resulting 2FO- Fc map was of excellent quality. The quality ofthe electrondensity of the molecule 1 heme is shown in Fig. 1.

RESULTS AND DISCUSSIONMolecular models of CCP and LiP shown in Figs. 2 and 3illustrate the striking similarities between the two peroxi-dases despite the fact that the sequence identity is <20%o (24).Most striking is the similar arrangement of 11 helical seg-ments and a limited amount of, structure in the proximal(lower) domain. The extra residues in LiP not present in CCPextend from an extra C-terminal segment that traverses overthe surface of the enzyme with no regular extended elements

Table 2. Heavy atom refinement and phasingPhase Phase power

No. of No. of power at maximumDerivative sites reflections RClijS overall resolution

Uranyl 1 8 8,331 0.59 1.78 1.45 at 2.6 AUranyl 2 6 13,739 0.55 1.83 1.50 at 2.7 APlatinum 1 2 15,502 0.64 1.46 1.21 at 2.7 APlatinum 2 2 12,900 0.68 1.44 1.34 at 2.9 AMercury 8 10,118 0.64 1.49 1.11 at 3.oA

Derivatives were prepared by soaking crystals in U03(NO3)2,Pt(NH3)(NO2)CI2, or Hg(CH3CO2)2. Two data sets were obtained forthe uranyl and platinum soaks and were treated as separate deriva-tives. Reflections were accepted for MIR phasing only if at least fourderivatives contributed and had amplitudes at least 6 times abovebackground level (i.e., .6of) to a maximum resolution of 2.6 A for allderivatives. The final figure of merit for 10,287 phased reflections to2.6 A was 0.68. RcuuiS, R factor for centric reflections, XIIFpHobsI -Fpobs + FHcalcII/XjjFpHobs - Fpobsl. Phase power, mean heavyatom structure factor amplitude divided by mean residual lack ofclosure error, I(FH)/I(FpHobs - FpHcalc).

**Furey, W. & Swaminathan, S., 14th American CrystallographyAssociation Meeting, 1990, New Orleans, abstr. PA33.

Biochemistry: Edwards et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

1, 2

020

752 Biochemistry: Edwards et al.

MIR map

2Fo-Fc omit map

FIG. 1. Fit of molecule 1 heme to electron density before (A) andafter (B) refinement. Both the MIR and 2F. - Fc maps werecontoured at la, to 2.6 A. Heme was excluded from the phasecalculation in generating the 2F. - Fc omit map. (C) Heme is viewededge on with the 2F. - Fc map (dashed lines) contoured at la, andthe F. - Fc map (solid lines) contoured at 3cr. The large lobe ofpositive difference density clearly indicates the presence of waterdirectly over the iron atom. The F. - F_ map was generated with theheme included in the structure factor calculation.

FIG. 2. LiP and CCP molecules with helices colored red. Disul-fide bridges in LiP are indicated and are paired as follows: 3:15,14:285, 34:120, and 249:317.

of secondary structure. The last 6 residues have not beenincluded in the current LiP model. LiP differs from CCP inhaving 4 disulfide bridges (Fig. 2).

In both peroxidases, the heme is sandwiched betweenhelices B (distal) and F (proximal; Fig. 3). In both proteins,one heme edge is situated at the bottom of a crevice formedby the surfaces of both domains. Nevertheless, in LiP thiscrevice is smaller, owing to side-chain interactions, whereasthe same pocket in CCP is more open. This structuraldifference was anticipated from the work ofDePillis et al. (31,32), who found that the heme edge is available for modifica-tion with aryl hydrazines in CCP (31) but not in LiP (32). TheLiP structure now shows that the meso position of the hemefacing the open end of the crevice, which is attacked byhydrazines in other peroxidases, is not readily accessible inLiP, whereas in CCP this region remains open. This alsoraises an interesting question concerninghow substrates suchas veratryl alcohol interact with LiP. This small openingconnecting the surface of the enzyme to the distal pocket isthe best candidate.The distal pocket is very similar in both proteins. Each

contains a distal histidine and arginine, which have beenpostulated to operate in the formation of CCP compound I.F0 - F, difference maps clearly indicate the presence ofactive site water molecules. One of these waters sits directlyover the heme iron atom (Fig. 1C). The average distance ofthis water from the iron atom in the two LiP molecules in theasymmetric unit is 2.5 A compared to 2.4 A in CCP. Thisdistance is consistent with a variety of spectroscopic studies,which indicate that, like CCP, LiP is predominantly high spinand pentacoordinate (33-35). The proximal pocket also isvery similar in that in both LiP and CCP the proximalhistidine ligand is hydrogen bonded to a buried aspartic acidside chain. We have postulated elsewhere that this hydrogenbond imparts a greater anionic character to the histidineligand than in other heme proteins, such as the globins, andtherefore provides additional stability to the ferryl (Fe4+) ironin compound I (36, 37). Overall, these similarities support theview that the detailed mechanism of compound I formationproposed for CCP (36, 37) operates in LiP as well.The most striking difference in the active site was antici-

pated from sequence comparisons-that is, the substitutionof tryptophan residues in both the proximal and distal pock-ets of CCP with phenylalanine in LiP (Fig. 4). This probablyexplains in part why LiP compound I has a porphyrin ircationradical and in CCP compound I the radical is centered on anamino acid side chain. Moreover, CCP is relatively rich inamino acids that can be oxidized (7 tryptophans, 14 tyrosines,5 methionines, 1 cysteine), whereas LiP has no tyrosines, 3tryptophans, 8 methionines, and no free cysteines. OnlyMet-172 actually contacts the heme in LiP, while both Trp-51and -191 contact the heme in CCP.

Proc. Natl. Acad Sci. USA 90 (1993)

Dow

nloa

ded

by g

uest

on

Janu

ary

1, 2

020

Proc. Natl. Acad. Sci. USA 90 (1993) 753

LIP LIP

CCP CCP

LiP also exhibits some unusual properties not shared byCCP, including the types of substrates oxidized by compoundI. LiP can oxidize nonphenolic aromatics such as veratrylalcohol (3,4-dimethoxybenzyl alcohol) and, although it is notknown whether LiP directly attacks lignin, LiP very likely isdirectly involved in lignin degradation (2, 3, 20). In sharpcontrast, CCP is not very effective at oxidizing small aro-matic molecules and appears to be the only peroxidase in thisgroup to be specifically designed to interact with anothermacromolecule, cytochrome c. LiP also exhibits a very lowpH maximum (near pH 3), which controls the reduction butnot the formation of compound 1 (14, 17, 38). Such a low pKfor activity would suggest a carboxylic acid side chainsomehow participating in the reaction (17, 39). While wecannot rule out this possibility, there are no carboxylates

His52

\ ; Xnrg48 184

~~ / ~~~Asnli84

Asp 183

His176

CCP

FIG. 3. Stereo Ca backbonemodels of LiP (Upper) and CCP(Lower). Helices are labeled A-Jaccording to the labeling schemeused earlier for CCP (22). Notethat one of the helices is labeled b'to be consistent with the CCP no-menclature. Because helix G issomewhat obscured in this view,this helix is not labeled. This viewis looking down into the openchannel, which connects the sur-face of the enzymes with the hemedistal pocket.

within the heme pocket, although, as with CCP, the proximalhistidine ligand hydrogen bonds with a buried aspartic acidside chain (Fig. 4). The distal helix (helix B) does contain ahighly conserved aspartic acid (Asp-48; refs. 22 and 24),which immediately follows the distal His-47, yet the Asp-48side chain points away from the active site and does notinteract with the distal His-47. However, LiP does have anunusual carboxylate-carboxylate hydrogen bond not presentin CCP, which might lead to some subtle conformationaland/or stability changes that could indirectly affect electrontransfer from or binding to small substrates as the pH israised. One heme propionate is hydrogen bonded to a surfaceaspartic acid residue (Asp-183; Fig. 4), whereas in CCP theanalogous propionate is hydrogen bonded to Asn-184. Wealso note that aspartic acid at this position is unique to the LiP

His47

Arg43

FIG. 4. Comparison ofLiP andCCP active sites. Note that whereLiP has phenylalanine residuescontacting the heme and proximalhistidine ligand, CCP has tryp-tophan. Note, too, that whereCCP has Asn-184 hydrogen bond-ing with a heme propionate, Asp-183 serves this role in LiP. Thismay explain in part the low pHoptimum of LiP since disruptionof the aspartic acid-propionate hy-drogen bond would be expected todestabilize the heme pocket.

Phe 193

LIP

Biochemistry: Edwards et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

1, 2

020

754 Biochemistry: Edwards et al.

family of peroxidases (24). Disruption of this hydrogen bondat elevated pH could adversely affect activity by causingstructural perturbations. Such perturbations, however, couldnot be so large as to alter compound I formation since the rateof compound I formation is, as with CCP, insensitive to pHfrom pH 3-8 (14, 38, 40). Alternatively, the pH dependencecould have as much to do with the control of the LiPcompound I redox potential. In both CCP (40) and horserad-ish peroxidase (41), the redox potential of compound Iincreases with decreasing pH making compound I a betteroxidizing agent at low pH. If the same pattern holds for LiPcompound I, then the low pH optimum for LiP may simplyreflect the higher redox potential needed to oxidize nonphe-nolic substrates.As this manuscript was being reviewed, the refinement of

LiP continued and the current R factor is 0.15 at 2 A. Whilemany more details are now visible, the primary conclusionsbased on the 2.6-A structure described in this paper areconsistent with the higher-resolution refined structure. Adetailed description of the higher-resolution model will ap-pear elsewhere (42).

We are indebted to Dr. J. K. Mohann Rao (Frederick CancerResearch Institute) for providing us with the suite of programs fornoncrystallographic averaging and to Michael Tung for adapting thesoftware to be compatible with the PHASES package of crystallo-graphic programs and local workstations. This work was supportedin part by National Science Foundation Grants DMB 9104960(T.L.P.) and DMB 890438 (M.H.G.) and by Grant DE-FG-08-87-ER13715 (M.H.G.) from the Division of Energy Biosciences, U.S.Department of Energy.

1. Sarkanen, K. V. & Ludwig, C. H. (1971) Lignins: Occurrence,Formation, Structure, and Reactions (Wiley-Interscience,New York).

2. Gold, M. H., Wariishi, H. & Valli, K. (1989) in Biocatalysisin Agricultural Biotechnology, ACS Symposium Series, eds.Whitaker, J. R. & Sonnet, P. (Am. Chem. Soc., Washington),Vol. 389, pp. 127-140.

3. Kirk, T. K. & Farrell, R. L. (1987) Annu. Rev. Microbiol. 41,465-505.

4. Buswell, J. A. & Odier, E. (1987) CRC Crit. Rev. Biotechnol.6, 1-60.

5. Renganathan, V., Miki, K. & Gold, M. H. (1985) Arch.Biochem. Biophys. 241, 304-314.

6. Leisola, M. S. A., Kozulic, B., Meussdoerfer, F. & Fiechter,A. (1987) J. Biol. Chem. 262, 419-424.

7. Higuchi, T. (1990) Wood Sci. Technol. 24, 23-63.8. Shoemaker, H. E. (1990) Recl. Trav. Chim. Pays-Bas 109,

255-272.9. Miki, K., Kondo, R., Renganathan, V., Mayfield, M. & Gold,

M. H. (1988) Biochemistry 27, 4787-4794.10. Hammel, K. E. (1989) Enzyme Microb. Technol. 11, 776-777.11. Valli, K. & Gold, M. H. (1991) J. Bacteriol. 173, 345-352.

12. Valli, K., Wariishi, H. & Gold, M. H. (1992) J. Bacteriol. 174,2131-2137.

13. Renganathan, V. & Gold, M. H. (1986) Biochemistry 25, 1626-1631.

14. Marquez, L., Wariishi, H., Dunford, H. B. & Gold, M. H.(1988) J. Biol. Chem. 263, 10549-10552.

15. Tien, M., Kirk, T. K., Bull, C. & Fee, J. A. (1986) J. Biol.Chem. 261, 1687-1693.

16. Renganathan, V., Miki, K. & Gold, M. H. (1987) Biochemistry26, 5127-5132.

17. Wariishi, H., Huang, J., Dunford, H. B. & Gold, M. H. (1991)J. Biol. Chem. 266, 20694-20699.

18. Valli, K., Wariishi, H. & Gold, M. H. (1990) Biochemistry 29,8535-8539.

19. Wariishi, H. & Gold, M. H. (1990) J. Biol. Chem. 265, 2070-2077.

20. Hammel, K. E. & Moen, M. A. (1991) Enzyme Microb. Tech-nol. 13, 15-18.

21. Bricogne, G. (1976) Acta Crystallogr. Sect. A 32, 832-845.22. Ritch, T. G., Jr., Nipper, V. J., Akileswaran, L., Smith, A.,

Pribnow, D. & Gold, M. H. (1991) Gene 107, 119-126.23. Finzel, B. C., Poulos, T. L. & Kraut, J. (1984) J. Biol. Chem.

259, 13027-13036.24. Henrissat, B., Saloheimo, M., Lavaitte, S. & Knowles,

J. K. C. (1990) Proteins 8, 251-257.25. Hendrickson, W. A. & Konnert, J. H. (1980) in Computing in

Crystallography, eds. Diamond, R., Ramaseshan, S. & Ven-katesan, K. (Indian Inst. of Sci., Bangalore), pp. 13.01-13.23.

26. Finzel, B. C. (1987) J. Appl. Crystallogr. 20, 53-55.27. Sim, G. A. (1960) Acta Crystallogr. 5, 535-542.28. Hendrickson, W. A. & Lattman, E. E. (1970)Acta Crystallogr.

Sect. B 26, 136-143.29. Brunger, A. T., Krukowski, A. & Erickson, J. W. (1989) Acta

Crystallogr. Sect. A 46, 585-593.30. Brunger, A. T. & Krukowski, A. (1990) Acta Crystallogr. Sect.

A 46, 585-593.31. DePillis, G. D., Sishta, B. P., Mauk, A. G. & Ortiz de Mon-

tellano, P. R. (1991) J. Biol. Chem. 266, 19334-19341.32. DePillis, G. D., Wariishi, H., Gold, M. H. & Ortiz de Montel-

lano, P. R. (1990) Arch. Biochem. Biophys. 280, 217-223.33. Andersson, L. A., Renganathan, V., Chiu, A. A., Loehr,

T. M. & Gold, M. H. (1985) J. Biol. Chem. 260, 6080-6087.34. Kuila, D., Tien, M., Fee, J. A. & Ondrias, M. R. (1985)

Biochemistry 24, 3394-3397.35. Andersson, L. A., Renganathan, V., Loehr, T. M. & Gold,

M. H. (1987) Biochemistry 26, 2258-2263.36. Poulos, T. L. & Finzel, B. C. (1984) in Peptide and Protein

Review (Dekker, New York), Vol. 4, pp. 115-177.37. Poulos, T. L. (1987) Adv. Inorgan. Biochem. 7, 1-36.38. Andrawis, A., Johnson, K. A. & Tien, M. (1988) J. Biol. Chem.

263, 1195-1198.39. Cai, D. & Tien, M. (1991) J. Biol. Chem. 266, 14464-14469.40. Purcell, W. L. & Erman, J. E. (1976) J. Am. Chem. Soc. 98,

7033-7037.41. Hayashi, Y. & Yamazaki, I. (1979) J. Biol. Chem. 254, 9101-

9106.42. Bernstein, F. C., Koetzle, T. F., Williams, G. J. B., Meyer,

E. F., Jr., Brice, M. D., Rogers, J. R., Kennard, O., Shima-nouchi, T. & Tasumi, M. (1977) J. Mol. Biol. 112, 535-542.

Proc. Natl. Acad. Sci. USA 90 (1993)

Dow

nloa

ded

by g

uest

on

Janu

ary

1, 2

020