` on October 7, 2015
` on October 7, 2015
` on October 7, 2015
`
`www.sciencemag.org
`www.sciencemag.org
`www.sciencemag.org
`www.sciencemag.org
`
`Downloaded from
`Downloaded from
`Downloaded from
`Downloaded from
`
`REPORTS
`
`17. M. Y. Yang, J. D. Armstrong, I. Vilinsky, N. Straus-
`field, K. Kaiser, Neuron 15, 45 (1995).
`18. K. O’Dell, J. Armstrong, M. Y. Yang, K. Kaiser, ibid.,
`p. 55.
`19. Pan-neural expression was driven by P-GAL4 1407
`[generated by J. Urban and G. Technau; described
`in L. Luo, Y. Liao, L. Jan, Y. Jan, Genes Dev. 8, 1787
`(1994)].
`20. The P-GAL4 C309 line was crossed to UAS-lacZ
`and UAS-lacZ;UAS-G␣
`s* lines. Brains from the re-
`sultant progeny were dissected, and -Gal activity
`was detected as described [S. T. Sweeney, K.
`Broadie, J. Keane, H. Niemann, C. J. O’Kane, Neu-
`ron 14, 341 (1995)].
`21. L. Levin et al., Cell 68, 479 (1992).
`22. D. Cooper, N. Mons, J. Karpen, Nature 374, 421
`(1995).
`23. D. Clapman, Annu. Rev. Neurosci. 17, 441 (1994);
`
`W. Schreibmayer et al., Nature 380, 624 (1996).
`24. K. Kaiser, unpublished data.
`25. S. Boynton and T. Tully, Genetics 131, 655 (1992).
`26. R. Sokal and F. Rohlf, in Biometry (Freeman, New
`York, 1981), pp. 229 –240.
`27. E. Yeh, K. Gustafson, G. Boulianne, Proc. Natl.
`Acad. Sci. U.S.A. 92, 7036 (1995).
`28. L. Luo, T. Tully, K. White, Neuron 9, 595 (1992).
`29. We thank K. Moffat, J. Keane, K. Sto¨ rtkuhl, R.
`Greenspan, M. Yang, G. Boulianne, and A. Brand
`for reagents. Supported by grants from the Well-
`come Trust and the Science and Engineering Re-
`search Council (GR/F94989) (to C.J.O’K.), the Hu-
`man Frontiers Science program (to C.J.O’K., K.K.,
`and M.F.), and NIH grant HD 32245 (to T.T.).
`
`3 June 1996; accepted 21 October 1996
`
`A Mechanism of Drug Action Revealed by
`Structural Studies of Enoyl Reductase
`Clair Baldock, John B. Rafferty, Svetlana E. Sedelnikova,
`Patrick J. Baker, Antoine R. Stuitje, Antoni R. Slabas,
`Timothy R. Hawkes, David W. Rice*
`
`Enoyl reductase (ENR), an enzyme involved in fatty acid biosynthesis, is the target for
`antibacterial diazaborines and the front-line antituberculosis drug isoniazid. Analysis of
`the structures of complexes of Escherichia coli ENR with nicotinamide adenine dinu-
`cleotide and either thienodiazaborine or benzodiazaborine revealed the formation of a
`covalent bond between the 2⬘ hydroxyl of the nicotinamide ribose and a boron atom in
`the drugs to generate a tight, noncovalently bound bisubstrate analog. This analysis has
`implications for the structure-based design of inhibitors of ENR, and similarities to other
`oxidoreductases suggest that mimicking this molecular linkage may have generic
`applications in other areas of medicinal chemistry.
`
`ENR catalyzes the final reaction of the fatty
`acid synthase cycle: the reduction of a car-
`bon-carbon double bond in an enoyl moiety
`that is covalently linked to an acyl carrier
`protein. Recent studies have identified ENR
`as the target for a number of therapeutic
`agents against Mycobacterium tuberculosis (1)
`and Escherichia coli (2). Mycobacterium tuber-
`culosis ENR is the target for a metabolite of
`isoniazid, a potent drug that is used in the
`front-line chemotherapeutic treatment of tu-
`berculosis. However, strains of M. tuberculo-
`sis are emerging that are resistant to isoniazid
`(3), with consequent problems in treatment.
`Escherichia coli ENR is inhibited by a range of
`
`C. Baldock, J. B. Rafferty, S. E. Sedelnikova, P. J. Baker,
`D. W. Rice, Krebs Institute for Biomolecular Research,
`Department of Molecular Biology and Biotechnology,
`University of Sheffield, Sheffield S10 2TN, UK.
`A. R. Stuitje, Department of Genetics, Institute of Molec-
`ular Biological Studies (IMBW), Vrije Universiteit, Bio-
`center Amsterdam, De Boelelaan 1087, 1081 HV Am-
`sterdam, Netherlands.
`A. R. Slabas, Department of Biological Sciences, Univer-
`sity of Durham, Durham DH1 3LE, UK.
`T. R. Hawkes, Department of Exploratory Plant Sciences,
`Zeneca Agrochemicals, Jealott’s Hill Research Station,
`Bracknell, Berkshire RG12 6EY, UK.
`* To whom correspondence should be addressed.
`E-mail: D.Rice@sheffield.ac.uk
`SCIENCE 䡠 VOL. 274 䡠 20 DECEMBER 1996
`
`diazaborines, heterocyclic boron-containing
`compounds whose action is thought to lead
`to the inhibition of cell growth by prevent-
`ing lipopolysaccharide synthesis (4). Bio-
`chemical studies on E. coli ENR have shown
`that nicotinamide adenine dinucleotide
`(NAD⫹) is required for diazaborine binding;
`this finding has led to the suggestion that the
`drug binds
`to ENR in association with
`NAD⫹ or that NAD⫹ converts the drug to
`an active form (5). To obtain a molecular
`explanation for the inhibitory activities of
`this class of antibacterial agents, we deter-
`mined and analyzed the structure of E. coli
`ENR in complexes with NAD⫹ and either
`thienodiazaborine or benzodiazaborine as
`well as with NAD⫹ alone.
`The structure of the ENR-NAD⫹ com-
`plex was solved to 2.1 Å (6); those of the
`ENR-NAD⫹-thienodiazaborine and ENR-
`NAD⫹-benzodiazaborine complexes were
`solved to 2.2 and 2.5 Å, respectively (7)
`(Table 1 and Fig. 1A). In the final map of
`the ENR-NAD⫹ complex, the electron den-
`sity is of high quality for most of the protein
`atoms. However, there is a break in the
`density for a stretch of 10 amino acid resi-
`dues; these 10 residues form a loop, between
`
`2107
`
`2. D. Byers, R. Davis, J. Kiger, Nature 289, 79 (1981);
`M. Livingstone, P. Sziber, W. Quinn, Cell 37, 205
`(1984); P. Drain, E. Folkers, W. Quinn, Neuron 6, 71
`(1991); Z. L. Wu et al., Proc. Natl. Acad. Sci. U.S.A.
`92, 220 (1995); W. Li, T. Tully, D. Kalderon, Learn.
`Mem. 2, 320 (1996).
`3. E. Skoulakis, D. Kalderon, R. Davis, Neuron 11, 197
`(1993).
`4. A. Nighorn, M. Healy, R. Davis, ibid. 6, 455 (1991); P.
`Han, L. Levin, R. Reed, R. Davis, ibid. 9, 619 (1992).
`5. M. Heisenberg, A. Borst, S. Wagner, D. Byers,
`J. Neurogenet. 2, 1 (1985); J. S. de Belle and M.
`Heisenberg, Proc. Natl. Acad. Sci. U.S.A. 93, 9875
`(1996).
`6. J. S. de Belle and M. Heisenberg, Science 263, 692
`(1994).
`7. A. Brand and N. Perrimon, Development 118, 401
`(1993).
`8. F. Quan, W. Wolfgang, M. Forte, Proc. Natl. Acad.
`Sci. U.S.A. 86, 4321 (1989); W. Wolfgang et al.,
`J. Neurosci. 10, 1014 (1990).
`9. F. Quan, L. Thomas, M. Forte, Proc. Natl. Acad. Sci.
`U.S.A. 88, 1898 (1991).
`10. M. Simon, M. Strathmann, N. Gautam, Science 252,
`802 (1991); H. Bourne, D. Sanders, F. McCormick, Na-
`ture 349, 117 (1991); W. Tang and A. Gilman, Cell 70,
`869 (1992); A. Spiegel et al., inG Proteins (Molecular
`Biology Intelligence Unit, R. G. Landes Co., Austin, TX,
`1994), pp. 19–20; E. Neer, Cell 60, 249 (1995).
`11. Complementary DNAs encoding the short forms of
`sWT and G␣
`G␣
`s* (9) were subcloned into pUAST (7)
`and used to transform flies [A. Spradling, in Drosoph-
`ila: A Practical Approach, D. Roberts, Ed. (IRL Press,
`Oxford, 1986), pp. 75 –197]. The UAS-G␣
`s lines
`were verified by Msc I digestion of polymerase chain
`reaction–amplified G␣
`s transgenes to detect the
`presence of a site found in G␣
`sWT but lacking as a
`result of the Q215L mutation.
`12. 238Y and 201Y are described in (17) and C232 in
`(18). C747 shows identical expression to that of
`C772 (17) (24). C309 is described by J. D. Armstrong
`[thesis, University of Glasgow, Scotland (1995)]. The
`O’Kane laboratory screened 500 P-GAL4 lines to
`identify OK66, OK86, OK62, OK107, OK348, and
`OK415.
`13. T. Tully and W. Quinn, J. Comp. Physiol. 157, 263
`(1985); T. Tully, T. Preat, S. Boynton, M. Del Vecchio,
`Cell 79, 35 (1994).
`14. In total, we analyzed behaviorally 16 P-GAL4 inser-
`tions that expressed to some degree in MBs. Flies
`with four insertions (OK66, OK86, KL65, and KL107)
`showed reduced learning in the absence of expres-
`sion of constitutively activated G␣
`s*, most likely be-
`cause of genetic background differences. These
`were not tested further. Flies with eight insertions
`showed defects in olfactory acuity (30Y, C35, OK62,
`OK107, C302, OK415, and C532) or shock reactivity
`(C772), as transheterozygotes with UAS-G␣
`s*.
`These were eliminated from further behavioral test-
`ing. Flies with the remaining four P-GAL4 insertions
`(201Y, 238Y, C309, and C747) showed disrupted
`learning but normal olfactory acuity and shock reac-
`tivity, as transheterozygotes with at least one UAS-
`G␣
`s* insertion. We characterized a total of two P-
`GAL4 insertions (C232 and OK348) that expressed
`to some degree in the CC.
`15. We suggest that olfactory learning results from Gs-
`dependent convergence of the conditional stimulus
`(CS) and unconditional stimulus (US) in the MBs. This
`belief is reinforced by the observation that expres-
`sion of G␣
`s* in the CC does not affect learning. For-
`mally, however, we cannot exclude the possibility
`that other sites of convergence of the CS and US
`contribute to the conditioned behavior, as in honey-
`bees [M. Hammer and R. Menzel, J. Neurosci. 15,
`1617 (1995)].
`16. Direct immunolocalization of transgenic G␣
`s forms
`proved impossible, as endogenous G␣
`s is ex-
`pressed throughout the central nervous system (8).
`We have not observed any significant differences in
`qualitative expression between UAS-regulated inser-
`tions [see also (18) and D. Lin and C. Goodman,
`Neuron 13, 507 (1994)]. Thus, P-GAL4 – driven re-
`porter gene expression should reflect genuine ex-
`pression patterns of G␣
`s* under UAS regulation.
`
`CFAD v. Anacor, IPR2015-01776
`ANACOR EX. 2005 - 1/4
`
`
`
`strand 6 and helix ␣6, that borders the
`nucleotide-binding site. In the structures
`of the ENR-NAD⫹-diazaborine complex-
`es, this loop is well defined and provides
`two residues whose side chains are in van
`der Waals contact with the non–boron-
`containing five- and six-membered rings
`of
`thienodiazaborine and benzodiazabo-
`rine, respectively.
`
`The quality of the electron density map
`for both diazaborines enabled the unambig-
`uous positioning of the bicyclic thienodi-
`azaborine or benzodiazaborine rings, the sul-
`fonyl groups, and the respective propyl or
`tosyl substituents (Fig. 1, B and C). Both
`diazaborine compounds bind in a closely
`related manner, adjacent to the nicotin-
`amide ring of the cofactor,
`in a pocket
`
`A
`
`C
`
`B
`
`D
`
`Folate
`
`NAD(P)
`
`Thieno-(cid:13)
`diazaborine
`
`Fig. 1. The E. coli ENR tetramer is made up of four
`subunits, each consisting of a single domain of
`approximate dimensions 55 by 45 by 45 Å com-
`posed of a parallel  sheet of seven strands (1 to 7), flanked on one side by helices ␣1, ␣2, and ␣7
`and on the other by helices ␣3 to ␣5, with a further helix, ␣6, lying along the top of the  sheet. (A)
`Schematic diagram of a single subunit of the ENR-NAD⫹-thienodiazaborine complex. The ribbon trace
`of E. coli ENR is shown in red; NAD⫹ (blue) and diazaborine (cyan) are shown in an all-atom represen-
`tation. The loop that orders on diazaborine binding is highlighted in green. [Produced using MIDAS (25).]
`(B and C) Initial Fourier maps of the NAD⫹-thienodiazaborine complex at 2.2 Å resolution (B) and of the
`NAD⫹-benzodiazaborine complex at 2.5 Å resolution (C) with the final refined structures superimposed.
`兩
`
`The density (contoured at 1.2 and 0.9, respectively) was calculated with coefficients 2兩Fobs兩 ⫺ 兩Fcalc
`and phases that were calculated from the refined structure from the molecular replacement solution that
`had been generated with the model of the E. coli ENR-NAD⫹ complex, which contained no information
`about the inhibitor. [Produced using BOBSCRIPT (26), a modified version of MOLSCRIPT (27).] (D) The
`superposition (based on the nicotinamide and its associated ribose) of the nucleotide-inhibitor complex
`of ENR into the active site of the nucleotide-substrate complex of DHFR [PDB entry 7DFR (13)]. The C␣
`backbone trace for DHFR is shown in green, with bound NADP and folate colored turquoise and by
`atom, respectively; the superimposed NAD⫹ and thienodiazaborine of ENR are shown in red and all
`atom colors, respectively (red, oxygen; white, carbon; blue, nitrogen; yellow, sulfur; green, boron). The
`covalent bond between the 2⬘ hydroxyl of the nicotinamide ribose and the boron of the diazaborine in
`ENR is represented by a dotted yellow line. [Produced using MIDAS (25).] When the NAD⫹-thienodi-
`azaborine complex is fitted into the active site of DHFR, there are some steric clashes between the
`sulfonyl group and the propyl tail of the diazaborine with parts of the enzyme surface. Nonetheless, there
`is sufficient space around the 2⬘OH of the nicotinamide ribose to envisage the formation of a linker
`between the ribose and a folate analog.
`
`2108
`
`SCIENCE 䡠 VOL. 274 䡠 20 DECEMBER 1996
`
`formed by the side chains of Tyr146, Tyr156,
`Met159, Ile200, Phe203, Leu100, Lys163, and
`the main-chain peptide between Gly93 and
`Ala95. The bicyclic rings of the diazaborines
`form a face-to-face interaction with the nic-
`otinamide ring, allowing the formation of
`extensive - stacking interactions with
`additional van der Waals interactions be-
`tween the rings and the side chains of
`Tyr156, Tyr146, Phe203, and Ile200. The ma-
`jor difference between the binding of the
`two diazaborines is that their respective to-
`syl and propyl groups occupy subtly differ-
`ent positions. The tosyl moiety lies perpen-
`dicular to the bicyclic ring and interacts
`with the main chain peptide between Gly93
`and Ala95 and the side chain of Leu100,
`whereas the propyl moiety folds back onto
`the planar bicyclic ring system in a manner
`reminiscent of a scorpion’s tail and forms
`interactions with the side chain of Met159
`and Ile200 and the main chain peptide of
`both Gly93 and Phe94. Additional interac-
`tions made by both drugs include hydrogen
`bonds between the boron hydroxyl and the
`phenolic hydroxyl of Tyr156 (Fig. 2) and
`between a nitrogen in the boron-containing
`ring and an ordered solvent molecule.
`Analysis of the drug complex showed that
`the distance between the boron atom of the
`diazaborine and the 2⬘ hydroxyl of the nic-
`otinamide ribose was ⬃1.7 Å, comparable
`with a B–O covalent bond length of 1.6 Å.
`The quality of the electron density map
`(particularly for the thienodiazaborine com-
`plex at 2.2 Å) implies that the errors in
`coordinates are very small, and thus we can
`be confident that the interaction between
`these two atoms is covalent. This conclusion
`is further supported by the unambiguous
`identification of the position of the hydroxyl
`oxygen to which the boron is linked, which
`can be seen to form part of a tetrahedral
`rather than a trigonal arrangement, as re-
`quired if the boron forms four covalent bonds
`(Fig. 1, B and C). This finding provides a
`clear explanation for the strong inhibitory
`properties of the diazaborines and for the
`requirement of NAD⫹ for diazaborine bind-
`ing. The formation of the covalent bond
`further explains why the replacement of the
`B-N group in the diazaborine by an isoelec-
`tronic C-C unit in an isoquinoline analog
`showed no biological activity (8). The for-
`mation of the covalent bond in this case
`resembles the manner in which the boronic
`acid inhibitors of
`serine proteases act
`through the chemical modification of the
`active-site serine to give a covalently bound
`tetrahedral adduct (9).
`The structure of the E. coli enzyme is
`closely related to that of its mycobacterial
`counterpart [Protein Data Bank (PDB) en-
`try 1ENY (1); 40% sequence identity; 199
`C␣ atoms superimposed, root-mean-square
`
`ANACOR EX. 2005 - 2/4
`
`
`
`deviation (rmsd) 1.1 Å], particularly in the
`region of the active site, and also to that of
`Brassica napus ENR [PDB entry 1ENO (10);
`33% sequence identity; 208 C␣ atoms su-
`perimposed, rmsd 1.4 Å]. Examination of
`the structures of the E. coli and M. tubercu-
`losis ENRs shows that their respective Gly93
`3 Ser (G93S) and Ser94 3 Ala mutations,
`which lead to resistance to diazaborine (11)
`and isoniazid (1), respectively, map to re-
`gions close to the nucleotide-binding site.
`Modeling studies show that in the absence
`of changes to the main-chain torsion angles
`in the G93S mutant of E. coli ENR, the C
`atom of the serine side chain would be
`unacceptably close to the two oxygens of
`the sulfonyl group of
`the diazaborine.
`Therefore, resistance to diazaborine is prob-
`ably explained by the serine side chain of
`the G93S mutant encroaching into the
`drug-binding site and causing severe steric
`hindrance.
`The position of the aromatic bicyclic
`ring of the diazaborine above the nicotin-
`amide ring strongly resembles the model
`proposed for the binding of the enoyl sub-
`strate on the basis of studies on B. napus
`ENR (10), with the proposed position for
`the negatively charged oxygen of the eno-
`late anion of the substrate close to that of
`the boron atom in the drug. Thus, the
`formation of a covalent bond between
`ENR and diazaborine generates a tight,
`noncovalently bound bisubstrate analog.
`No such good bisubstrate inhibitors have
`previously been described for NAD(P)-
`dependent oxidoreductases. Examples of
`this type designed to inhibit 3-hydroxy-3-
`methyl glutaryl (HMG) coenzyme A re-
`ductase were only very weak inhibitors of
`cholesterol biosynthesis, possibly because
`of the lack of a moiety to mimic the
`adenosine diphosphoribose, but more like-
`ly because of steric problems in the active
`site associated with the nature of the link-
`age that used the C-4 atom of the nico-
`tinamide ring (12). In contrast, our study
`provides clear evidence for the type of
`linkage that may need to be created in
`order to synthesize a bisubstrate analog
`with the necessary geometry to occupy the
`active-site cleft. This may prove crucial in
`the design of a new generation of antibac-
`terial agents against a range of drug-resis-
`tant organisms, including M. tuberculosis.
`However, given the inherent toxic poten-
`tial of boron (8), the development of a
`range of inhibitors with minimal side ef-
`fects will require the design of a preformed
`bisubstrate analog in which another atom
`is substituted for boron.
`More widely, a number of NAD(P)-de-
`pendent oxidoreductases are known to be
`drug targets, including dihydrofolate reduc-
`tase (DHFR), the target for the anticancer
`
`agent methotrexate (13); steroid 5␣-re-
`ductase, the target for finasteride, used to
`treat benign prostatic hyperplasia (14);
`and inosine monophosphate dehydroge-
`nase (IMPDH), the target for the immu-
`nosuppressant mycophenolic acid (MPA)
`(15). Structural analysis of members of the
`family of NAD(P)-dependent oxidoreduc-
`tases has shown that the relative orienta-
`tion of the nicotinamide ring, its associat-
`ed ribose, and the enzyme active site are
`often closely related. Moreover, the cata-
`lytic oxidation-reduction cycle in these
`enzymes necessarily leads to a situation
`where the electron system of a substrate
`approaches the face of the nicotinamide
`ring. Thus, for the subset of dehydroge-
`nases where there is sufficient space in the
`structure around the 2⬘OH group of the
`nicotinamide ribose, there is an excellent
`opportunity to mimic the chemistry seen
`in the NAD⫹-diazaborine complex in the
`
`REPORTS
`synthesis of new enzyme inhibitors. For
`example, the superposition of the nucle-
`otide-inhibitor complex of ENR into the
`active site of
`the nucleotide-substrate
`complex of DHFR indicates that linking a
`folate analog to the nicotinamide ribose is
`a distinct possibility, and such a strategy
`might be used in the design of new anti-
`cancer agents (Fig. 1D). Moreover, the
`similarity between the ENR-NAD⫹-diaza-
`borine complex and the orientation of the
`inosine-5⬘-monophosphate
`thioimidate
`intermediate and the inhibitor MPA in
`IMPDH (15) suggests that the 2⬘ hydroxyl
`of the inosine might be linked to the
`inhibitor in a similar manner; however,
`coordinates of the latter are not yet avail-
`able for a more detailed comparison.
`These examples suggest that the use of the
`ribose hydroxyl to create a bisubstrate an-
`alog may find important applications in
`other areas of medicinal chemistry.
`
`Table 1. X-ray data collection and phasing statistics. See (6) and (7) for descriptions of data sets.
`
`Data set
`
`Form A crystals
`
`Form B crystals
`
`Native-1
`
`Native-2
`
`Merged
`
`Hg
`
`Thieno
`
`Benzo
`
`2.5
`114,615
`27,520
`78
`7.1
`
`2.1
`86,550
`49,465
`89
`3.8
`
`2.1
`113,658
`51,902
`93
`5.8
`
`hkl
`
`Resolution (Å)
`No. of observed reflections
`No. of unique reflections
`Completeness (%)
`Rmerge* (%)
`Mean fractional isomorphous
`difference†
`No. of heavy atom sites
`Phasing power (acentric/centric)‡
`RCullis (acentric/centric)§
`兩/兺
`兩Ii
`⫽ 兺
`⫺ Im
`*Rmerge
`hklIm, where Ii and Im are the observed intensity and mean intensity of related reflections,
`
`
`
`†Mean fractional isomorphous difference ⫽ 兺 兩兩FPH兩 ⫺ 兩FP兩兩/兺 兩FP兩, where FPH and FP are the structure
`respectively.
`‡Phasing power ⫽ 具FH/lack of closure典.
`factor amplitudes for derivative and native crystals, respectively.
`§RCullis
`⫽ 具lack of closure典/具isomorphous difference典.
`
`2.2
`76,176
`31,179
`95
`5.8
`
`2.5
`45,376
`20,393
`95
`5.7
`
`3.0
`20,674
`14,042
`73
`7.7
`0.24
`
`6
`1.4/1.0
`0.76/0.68
`
`Fig. 2. Schematic repre-
`sentation of the interac-
`tions made by the NAD⫹-
`thienodiazaborine com-
`plex with the enzyme sur-
`face and ordered solvent
`molecules. For NAD⫹,
`only the nicotinamide ring
`and the nicotinamide ri-
`bose are shown. Hydro-
`gen bonds are represent-
`ed by dashed lines, hy-
`drophobic contacts are
`shown as semicircular
`arcs, and Wat 1 and Wat
`2 are two ordered solvent
`molecules.
`[Produced
`using LIGPLOT (28).]
`
`SCIENCE 䡠 VOL. 274 䡠 20 DECEMBER 1996
`
`2109
`
`ANACOR EX. 2005 - 3/4
`
`
`
`REFERENCES AND NOTES
`___________________________
`1. A. Banerjee et al., Science 263, 227 (1994); A. Des-
`sen, A. Que´ mard, J. S. Blanchard, W. R. Jacobs Jr.,
`J. C. Sacchettini, ibid. 267, 1638 (1995).
`2. H. Bergler et al., J. Biol. Chem. 269, 5493 (1994).
`3. S. T. Cole, Trends Microbiol. 2, 411 (1994).
`4. G. Ho¨ genauer and M. Woisetschla¨ ger, Nature 293,
`662 (1981).
`5. M. M. Kater, G. M. Koningstein, H. J. J. Nijkamp, A.
`R. Stuitje, Plant Mol. Biol. 25, 771 (1994).
`6. Escherichia coli ENR is a homotetramer (Mr ⬃ 28,000
`per subunit) that was prepared from an overexpress-
`ing E. coli strain (2, 5). Crystals of the ENR-NAD⫹
`complex (crystal form A) belong to space group P21
`and have unit cell dimensions of a ⫽ 74.0 Å, b ⫽ 81.2
`Å, c ⫽ 79.0 Å, and  ⫽ 92.9° with a tetramer in the
`asymmetric unit (16). Data were collected to 2.5 Å
`( Table 1, data set Native-1) on a twin San Diego Mul-
`tiwire Systems (SDMS) area detector with a Rigaku
`RU-200 rotating anode source, and the data set was
`processed with SDMS software (17). Data were also
`collected to 2.1 Å ( Table 1, data set Native-2) at the
`CLRC Daresbury Synchrotron and processed with
`the MOSFLM package (18), and the 2.1 and 2.5 Å
`data sets were then scaled and merged with CCP4
`software (19). Initially, a model of B. napus ENR (10)
`was used as a basis for a molecular replacement
`solution of the structure, but the map, calculated after
`the model was refined with the program TNT (20), was
`not of sufficient quality to confidently assign residues
`in regions of structural differences between the B.
`napus and E. coli enzymes. Therefore, to solve the
`structure, we obtained a heavy-atom derivative by
`soaking an ENR-NAD⫹ (form A) crystal for 1 hour in
`0.1 mM ethylmercuriphosphate, 10 mM NAD⫹, 20%
`(w/v) polyethylene glycol (molecular weight 400), and
`100 mM acetate (pH 5.0). Derivative data were col-
`lected at the CLRC Daresbury Synchrotron to a reso-
`lution of 3 Å ( Table 1, data set Hg) and were pro-
`cessed as above. The positions of the heavy atoms in
`this derivative were revealed by difference Fourier
`methods with the use of the approximate phases pro-
`vided by the molecular replacement solution. The
`heavy-atom parameters were refined with the pro-
`gram MLPHARE (21) and resulted in a phase set with
`an overall mean figure of merit of 0.34 to 3 Å resolu-
`tion. Using a map derived from these phases, we
`generated molecular masks for the molecule with the
`program MAMA (22) and performed 50 cycles of sol-
`vent flattening and fourfold molecular averaging with
`the program DM (19, 23). In the resultant electron
`density map, calculated from the averaged phases,
`we were able to find clear density for all but the first
`residue, the last four residues, and 10 residues from
`the loop joining 6 and ␣6; using the graphics pro-
`gram FRODO (24), we were able to build with confi-
`dence a model comprising 247 of the 262 amino ac-
`ids of E. coli ENR. Several cycles of rebuilding and
`refinement gave a final R factor for the model of 0.157
`(52,346 reflections in the range 10 to 2.1 Å, 7836
`atoms including 324 water molecules), with an rmsd
`of 0.017 Å for bonds and 2.92° for angles [R ⫽
`⌺(cid:239)((cid:239)F obs(cid:239) ⫺(cid:239) Fcalc
`)(cid:239) /⌺((cid:239) Fobs
`), where (cid:239)F obs
`and
`(cid:239) are the observed and calculated structure fac-
`(cid:239)F calc
`tor amplitudes, respectively]. The average B factor
`for the tetramer is 30 Å2 (24 Å2 for main-chain
`atoms), where B ⫽ 82( 2) and
`is the mean
`square displacement of the atomic vibration.
`the ENR-NAD⫹-diazaborine complex
`7. Crystals of
`(crystal form B) belong to space group P6122 and
`have unit cell dimensions of a ⫽ b ⫽ 80.9 Å, c ⫽
`328.3 Å, ␣ ⫽  ⫽ 90°, and ␥ ⫽ 120° for the thieno-
`diazaborine complex, and a ⫽ b ⫽ 80.6 Å, c ⫽ 325.3
`Å, ␣ ⫽  ⫽ 90°, and ␥ ⫽ 120° for the benzodiaza-
`borine complex with a dimer in the asymmetric unit
`(16). Data sets were collected on the ENR-NAD⫹-
`thienodiazaborine complex to 2.2 Å and on the ENR-
`NAD⫹-benzodiazaborine complex to 2.5 Å ( Table 1,
`data sets Thieno and Benzo) at the CLRC Daresbury
`Synchrotron and were processed as above. The
`structures of both ENR-NAD⫹-diazaborine complex-
`es were solved independently by molecular replace-
`ment with the use of an appropriate dimer from the E.
`coli ENR-NAD⫹ structure and were refined against
`
`2110
`
`their respective data sets with the program TNT (20).
`The initial electron density maps were readily inter-
`pretable and unambiguous density could be ob-
`served for the location of the diazaborine com-
`pounds, which were then incorporated into the re-
`finement. Clear density could be found for all but the
`first residue and the last four residues. Refinement of
`the thienodiazaborine complex gave a final R factor
`of 0.191 (30,825 reflections in the range 10 to 2.2 Å,
`3936 atoms), with an rmsd of 0.012 Å for bonds
`and 2.9° for angles. The average B factor for the
`dimer is 27 Å2 (22 Å2 for main-chain atoms, 20 Å2
`for diazaborine atoms). Refinement of the benzodi-
`azaborine complex gave a final R factor of 0.169
`(20,204 reflections in the range 10 to 2.5 Å, 3930
`atoms), with an rmsd of 0.013 Å for bonds and 2.7°
`for angles. The average B factor for the dimer is 24
`Å2 (20 Å2 for main-chain atoms, 20 Å2 for diazabo-
`rine atoms). For the ENR-NAD⫹ complex and the
`ENR-NAD⫹-thienodiazaborine complex, 244 C␣
`atoms superimpose with an rmsd of 0.3 Å, whereas
`for the two ENR-NAD⫹-diazaborine complexes,
`256 C␣ atoms superimpose with an rmsd of 0.2 Å.
`8. M. A. Grassberger, F. Turnowsky, J. Hildebrandt,
`J. Med. Chem. 27, 947 (1984).
`9. S. Zhong, F. Jordan, C. Kettner, L. Polgar, J. Am.
`Chem. Soc. 113, 9429 (1991).
`10. J. B. Rafferty et al., Structure 3, 927 (1995).
`11. H. Bergler, G. Ho¨ genauer, F. Turnowsky, J. Gen.
`Microbiol. 138, 2093 (1992).
`12. C. Taillefumier, D. de Fornel, Y. Chapleur, Bioorg.
`Med. Chem. Lett. 6, 615 (1996).
`13. J. T. Bolin, D. J. Filman, D. A. Matthews, R. C. Ham-
`lin, J. Kraut, J. Biol. Chem. 257, 13650 (1982); C.
`Bystroff, S. J. Oatley, J. Kraut, Biochemistry 29,
`3263 (1990).
`14. H. G. Bull et al., J. Am. Chem. Soc. 118, 2359
`(1996).
`15. M. D. Sintchak et al., Cell 85, 921 (1996).
`
`16. C. Baldock et al., Acta Crystallogr., in press.
`17. R. Hamlin, Methods Enzymol. 114, 416 (1985); A. J.
`Howard, C. Nielson, N. H. Xuong, ibid., p. 452; N. H.
`Xuong, C. Nielson, R. Hamlin, D. Anderson, J. Appl.
`Crystallogr. 18, 342 (1985).
`18. A. G. W. Leslie, Joint CCP4 and ESF-EACBM News-
`letter on Protein Crystallography No. 26 (SERC
`Daresbury Laboratory, Warrington, UK, 1992).
`19. Collaborative Computational Project No. 4, Acta
`Crystallogr. D50, 760 (1994).
`20. D. E. Tronrud, L. F. Ten Eyck, B. W. Matthews, ibid.
`A43, 489 (1987).
`21. Z. Otwinowski, in Proceedings of the CCP4 Study
`Weekend, W. Wolf, P. R. Evans, A. G. W. Leslie, Eds.
`(SERC Daresbury Laboratory, Warrington, UK,
`1991), p. 80.
`22. G. J. Kleywegt and T. A. Jones, ESF/CCP4 News-
`letter No. 28 (1993), p. 56.
`23. K. Cowtan, ESF/CCP4 Newsletter No. 31 (1994), p.
`34.
`24. T. A. Jones, J. Appl. Crystallogr. 11, 268 (1978).
`25. T. E. Ferrin, C. C. Huang, L. E. Jarvis, R. Langridge,
`J. Mol. Graphics 6, 13 (1988).
`26. R. Esnouf, personal communication.
`27. P. J. Kraulis, J. Appl. Crystallogr. 24, 946 (1991).
`28. A. C. Wallace, R. A. Laskowski, J. M. Thornton,
`Protein Eng. 8, 127 (1995).
`29. We thank the support staff at the Synchrotron Radi-
`ation Source at Daresbury Laboratory for assistance
`with station alignment. Supported by grants from the
`UK Biotechnology and Biological Sciences Research
`Council
`(BBSRC) and Medical Research Council
`(D.W.R. and A.R.S.). C.B. is funded by a Zeneca
`Agrochemicals–supported CASE award. J.B.R. is a
`BBSRC David Phillips Research Fellow. The Krebs
`Institute is a designated BBSRC Biomolecular Sci-
`ence Centre.
`
`2 August 1996; accepted 15 October 1996
`
`Orientation Maps of Subjective Contours in
`Visual Cortex
`Bhavin R. Sheth, Jitendra Sharma, S. Chenchal Rao,
`Mriganka Sur*
`
`Responses to subjective contours in visual cortical areas V1 and V2 in adult cats were
`investigated by optical imaging of intrinsic signals and single-unit recording. Both V1 and
`V2 contain maps of the orientation of subjective gratings that have their basis in specific
`kinds of neuronal responses to subjective orientations. A greater proportion of neurons
`in V2 than in V1 show a robust response to subjective edges. Through the use of
`subjective stimuli in which the orientation of the luminance component is invariant, an
`unmasked V1 response to subjective edges alone can be demonstrated. The data
`indicate that the processing of subjective contours begins as early as V1 and continues
`progressively in higher cortical areas.
`
`Contours that are perceived under stimulus
`configurations in which the stimulus lacks
`any physical discontinuity (such as a lumi-
`nance border) are termed subjective or illu-
`sory contours. Subjective contours and sub-
`jective shapes can be perceived in a manner
`analogous to the perception of luminance
`
`Department of Brain and Cognitive Sciences, Massachu-
`setts Institute of Technology (MIT ), Cambridge, MA
`02139, USA.
`* To whom correspondence should be addressed at De-
`partment of Brain and Cognitive Sciences, MIT, E25-235
`Cambridge, MA 02139, USA. E-mail: msur@wccf.mit.
`edu
`SCIENCE 䡠 VOL. 274 䡠 20 DECEMBER 1996
`
`contours and shapes (1), suggesting that
`subjective and luminance contours might
`be processed in similar manner, perhaps by
`similar neural substrates. A subset of neu-
`rons in V2 of monkeys has been shown to
`respond to subjective edges (2). In addition,
`cells in monkey V1 have been reported to
`respond to subjective edges (3); however,
`the stimuli used were different from classi-
`cal subjective stimuli because they had a
`luminance gradient across
`the putative
`“subjective” edge. Thus, the question of
`whether or not V1 neurons respond to sub-
`jective stimuli remains unresolved. Further-
`
`ANACOR EX. 2005 - 4/4
`
`(cid:239)
`(cid:239)
`(cid:239)