PROTEINS: Structure, Function, and Genetics 9:191-206 (1991)

Comparative Modeling of Mammalian Aspartate Transcarbamylase Joshua L. S c d y and David R. Evans Department of Biochemistry, Wayne State University School of Medicine, Detroit, Michigan 48201

ABSTRACT Mammalian aspartate transcarbamylase (ATCase)is part of a 243 kDa multidomain polypeptide, called CAD, that catalyzes the first three steps in de novo pyrimidine biosynthesis. The structural organization of the mammalian enzyme is very different from E. coli ATCase, a dodecameric, monofunctional molecule comprised of six copies of separate catalytic and regulatory chains. Nevertheless, sequence similarities and other properties s u g gested that the mammalian ATCase domain and the E. coli ATCase catalytic chain have the same tertiary fold. A model of mammalian ATCase was built using the X-ray coordinates of the E. coli catalytic chain as a tertiary template. Five small insertions and deletions could be readily accommodated in the model structure. Following energy minimization the RMS difference in the a carbon positions of the mammalian and bacterial proteins was 0.93 A. A comparison of the hydrophobic energies, surface accessibility index, and the distribution of hydrophilic and hydrophobic residues of the CAD ATCase structure with correctly and incorrectly folded proteins and with several X-ray structures supported the validity of the model. The mammalian ATCase domain associates to form a compact globular trimer, a prerequisite for catalysis since the active site is comprised of residues from adjacent subunits. Interactions between the clearly defined aspartate and carbamyl phosphate subdomains of the monomer were largely preserved while there was appreciable remodeling of the trimeric interfaces. Several clusters of basic residues are located on the upper surface of the domain which account in part for the elevated isoelectric point (PI= 9.4) and may represent contact regions with other more acidic domains within the chimeric polypeptide. A long interdomain linker connects the monomer at its upper surface to the remainder of the polypeptide. The configuration of active site residues is virtually identical in the mammalian and bacterial enzymes. While the CAD ATCase domain can undergo the local conformational changes that accompany catalysis in the E. coli enzyme, the high activity, closed conformation is probably more stable in the mammalian enzyme. 0 1991 WILEY-LISS, INC

Key words: CAD, E. coli ATCase, energy minimization, multifunctional proteins, protein domains, sequence homology, evaluation of protein models INTRODUCTION While the reactions for de novo pyrimidine biosynthesis are nearly universal, the structural organization of the pathway enzymes is very different in different In E. coli, a single carbamyl phosphate synthetase serves both pyrimidine and arginine biosynthetic pathways1p2so aspartate transcarbamylase (ATCase, EC 2.1.3.2), which catalyzes the first committed step in pyrimidine biosynthesis, the synthesis of carbamyl aspartate from carbamyl phosphate and aspartate, is the major locus of contr01.~-~ 33.coli aspartate transcarbamylase is an allosteric enzyme which binds both aspartate3,*and carbamyl phosphate5 cooperatively and is inhibited by CTP and activated by ATP.3,4 Reaction of 303 kDa enzyme with mercurials results in its dissociation into two distinct functional units6: catalytic subunits which are enzymatically active but which exhibit no allosteric transitions and regulatory subunits which bind allosteric effectors but are inactive. The catalytic subunit is a trimer of three identical 34 kDa catalytic chains, while the regulatory subunit is a dimer composed of two 17 kDa regulatory

Received February 20, 1990; revision accepted August 9, 1990. Address reprint requests to Dr. David R. Evans, Department of Biochemistry, Wayne State University School of Medicine, 540 E. Canfield, Detroit, MI 48201. Abbreviations: Aspartate domain, formerly called the equitonal domain, residues 150-284 of the E. coli ATCase catalytic polypeptide; ATCase, aspartate transcarbamylase activity or domain; CAD, the multifunctional protein that has glutaminedependent carbamyl phosphate synthetase, aspartate transcarbamylase, and dihydroorotase activities and that catalyzes the first three steps of mammalian de novo pyrimidine biosynthesis; C chain, the catalytic polypeptide of E. coli aspartate transcarbamylase; carbamyl phosphate domain, formerly called the polar domain, residues 1-149 and 285-308 of E. coli ATCase catalytic polypeptide; CG, conjugate gradient energy minimization; CHY, a-chymotrypsin; CPA, bovine carboxypeptidase A; C subunit, the active trimer comprised of E. coli ATCase catalytic polypeptides; ECO, E. coli ATCase catalytic subunit; PALA, N-phosphonacetyl-L-aspartate; R chain, the regulatory polypeptide of E. coli aspartate transcarbamylase; R subunits, a dimer of E. coli ATCase regulatory polypeptides; SDG, steepest descent gradient energy minimization; SA, surface accessibility; TFU', trypsin.

192

J.L. SCULLY AND D.R. EVANS

Lipscomb and his associates have determined the X-ray structure of unliganded E. coli ATCase and of its complexes with CTP and the bisubstrate analog PALA.11-14 Mammalian ATCase is associated with a 243 kDa polypeptide, called CAD, which also has glutaminedependent carbamyl phosphate synthetase (CPSase 11, EC 2.7.2.9) and dihydroorotase (DHOase, EC 3.5.2.3) activities.l5-'' CPSase 11, which is specific for pyrimidine biosynthesis, controls the flux through the pathwayla and ATCase is unregulated. The polypeptide associates to form hexamers and other higher oligomeric form^.^^,^^ Thus the first three steps of the de novo pathway are catalyzed by a 1.45 niDa complex. The CAD polypeptide is separately folded into structural domains each having a different function.20,21We have isolated22 a 36 kDa fragment from controlled proteolytic digests which carries the ATCase activity of CAD. The isolated domain is a 110 kDa trimer with a Stokes radius of 41 A and has kinetic parameters that are very similar to those of the parent m o l e c ~ l e . ~ ~ ~ ~ ~ By sequencing 500 bp a t the 3' end of a nearly full length cDNA molecule, Shigesada et al.24discovered that the ATCase domain is located on the carboxyl end of the polypeptide. We recently sequencedZ3the entire ATCase domain of CAD and showed that the core domain is approximately the same size as the E. coli catalytic chain and has a very similar sequence with few insertions and deletions. The proteolytic fragment has a 20 residue extension on the amino terminus of the core domain which is part of the 132 residue interdomain linker connecting the ATCase domain to the remainder of the polypeptide. The similarity of the mammalian ATCase domain and the E. coli ATCase catalytic chain in sequence, subunit structure, and other structural and kinetic properties, coupled with the many successes in comparative molecular modeling of serine protease^:^-^^ aspartate pro tease^,^',^^ immunoglobul i n ~ , ~~' a, ~ l m~o d u l i n , 3and ~ * human ~~ liver phosphory l a ~ to e mention ~ ~ but a few examples (for a review see ref. 351, prompted us to model the structure of mammalian ATCase using the E. coli catalytic subunit as a template.

METHODS Sequence Alignment The h a r n ~ t e ? ~ and . ~ ~ E. coZi36,37 ATCase sequences were aligned using the BIONET program FASTP and Protein Information Resource program ALIGN. Input parameters were chosen empirically. A ktup = 2 was used for FASTP. Three standard similarity matrices, mutational (gap penalty 301, genetic code (gap penalty 6), and unitary (gap penalty 6), as well as the structural similarity matrix (gap penalty 30), devised by Risler et al.38 were used for the ALIGN calculations. All five alignments were

for the most part the same but placed the insertions and deletions in somewhat different locations. In each case we chose the alignment found most often or the highest scoring alignment which located insertions and deletions outside regions of well defined secondary structure.

Modeling The program QUANTA (Polygen, Inc.), implemented on a Silicon Graphics 4D70T workstation, was used for modeling. The 2.4 A coordinate set (R factor 0.155) for the E. coli aspartate transcarbamylase complexed with PALA, kindly provided by William N. Lipscomb, was used since this crystal form represents the active conformation of the enzyme.13The catalytic chains within the two catalytic subunits are designated C1, C2, C3 and C4, C5, C6.11 The asymmetric unit contains two catalytic and two regulatory chains with only minor violation of noncrystallographic symmetry.13 The atomic coordinates for the C1 catalytic chain served as the template for modeling the mammalian monomer. The program DSSP (Brookhaven National Laboratory) was used in the analysis of the model for calculation of specific bond angles, lengths, and the identification of hydrogen bonds.

Energy Minimization Energy minimization was carried out essentially as described by Brooks et al.39using the program CHARMM (Polygen, Inc.). The covalent or internal energy terms included bond energies, bond angle energy, dihedral energy, and improper torsional angle energy. Noncovalent or external energy terms included electrostatic and van der Waals interaction energy. Hydrogen bond energies were included implicitly in the electrostatic and van der Waals terms. The distance dependent dielectric (RDIE) was used for electrostatic interactions. The step size was 0.2 b, the nonbonded cutoff distance was 8 A, and the nonbonded update frequency was 20. We adopted the following five stage strategy for minimization of the CAD ATC monomer: (1)steepest descent gradient (SDG)minimization (300 steps) with the entire structure constrained except for the inserted residues, and the backbone and side chain of residues flanking insertions and deletions, (2) SDG (200 steps) followed by 100 steps of conjugate gradient (CG) minimization with the conformation of all identical side chains, the core residues, and all backbone atoms within regions of secondary structure constrained; (3) SDG (150 steps)/CG (150 steps) with the constraints on interior residues relaxe& (4) SDG (100 steps)/CG (200 steps) constraining only the secondary structure backbone atoms; and (5) conjugate gradient (300 steps) with all constraints relaxed. To construct the timer, a least-squares tit of the interior backbone residues of the minimized mono-

MAMMALIAN ASPARTATE TRANSCARBAMYLASE STRUCTURE

mer was fit to subunit C1 of the E. coli ATCase catalytic trimer using the program COMPARE. The other subunits were generated by operating around the 3-fold axis. The trimer was minimized initially in 300 steps using steepest descent gradient with the side chain conformations constrained followed by 150 steps of conjugate gradient minimization with no constraints.

Surface Accessibility The surface accessibility of all nonhydrogen atoms was calculated using the Lee and Richards' method.40A 1.4 A probe was used for these computations which were performed using the program ACCESS (Handschumacher and Richards). The accessible surface area of a globular monomeric protein was calculated from the molecular weight using the r e l a t i o n ~ h i p , 4 ~surface - ~ ~ accessibility = 11.12 M:'3. The surface accessibility (SA) index is defined as the ratio of the surface accessibility calculated using the atomic coordinates of the model to the average surface accessibility of a globular protein. The SA index should approximate 1.0 for a correctly folded protein.

Potential Energy Calculations To approximate the energy of the minimized model structures and X-ray structures, the solvent screened electrostatic energy was calculated by multiplying the charge of each solvent accessible atom given in the QUANTA residue topology file by 0.3, which takes into consideration the approximate 10fold difference in dielectric constant between internal and external environment^.^^ The hydrophobic energy was calculated as the solvation free energy as described by E i ~ e n b e r gThe . ~ ~ hydrophobic energy of the model structure was also estimated as the sum of the tabulated hydrophobic energies of each buried hydrophobic residue. Residues included in these calculations were I, L, V, F, Y, W, A, and M. The unfavorable energy contribution of hydrophobic residues located on the surface of the protein was taken into consideration by omitting from the summation any residue with a surface accessibility larger than 10 A". RESULTS AND DISCUSSION Sequence Alignment To obtain the optimal alignment between the CAD ATCase domain and the E. coli catalytic chain primary structures, the sequences of S. typhurium ATCase,46B. subtilis A T c a ~ e , 4 and ~ the ATCase domain of Drosophila m e l a n ~ g a s t e rand ~ ~ Dictyostelium d i s ~ o i d e u m ~ ~ also used in the alignment. were The inclusion of ATCase sequences from other organisms also made it possible to identify highly conserved regions which are likely to have a crucial role in determining the tertiary structure or in catalysis. As shown in the final alignment (Fig. l), the CAD

193

ATC domain and the E. coli catalytic subunit share 44% sequence identity. The sequence similarity becomes 56% when conservative substitutions are taken into consideration. Of the 308 amino acid residues in the CAD domain, 16% are invariant in all six species (Fig. 1).There were three small deletions and two insertions in the CAD sequence. Residues 76-77, 120, and 180 in the E. coli sequence were deleted in the mammalian protein while the CAD sequence had a four amino acid insertion between E. coli residues 37-38 and a single amino acid insertion between residue 235 and 236. The CAD sequence was three residues shorter at the carboxyl end of the polypeptide. None of the insertions and deletions occurred in highly conserved regions or within regions of welldefined secondary structure (Fig. 1).The hydropathic profiles of the CAD domain and E. coli ATCase are superimposable when allowance is made for insertions and deleti0ns.2~Moreover, all of the active site residues identified in the X-ray structure of the E. coli ATCase complex with the bisubstrate analog PALA are conserved in the CAD sequence. Thus, we were confident that we had the correct alignment.

Initial Model Building Since we were especially interested in the active site of the mammalian protein, we chose the 2.4 A E. coli ATCase PALA complex13 as the template for model building. The backbone and identical residues of the CAD sequence were first fit to the E. coli C chain coordinates. Substitutions were made using the QUANTA subprogram MUTATE superimposing the conformation of the substituted side chains as much as possible. The side chain x angles of the CAD residues were adjusted to closely approximate the conformations of the E. coli residue. When the CAD side chain was longer than the E. coli residue, QUANTA'S SPIN function was used to optimize the electrostatic and van der Waals interactions of the additional atoms. To accommodate the inserted amino acids, one residue on each side of the insertion site was removed. The backbone conformation of the inserted segment and the flanking residues was modeled with polyalanine by adjusting the 4 and angles to obtain the allowed conformations of the nonglycine residues. The side chains were then introduced using MUTATE and the unfavorable nonbonded interactions were relieved using the SPIN function. Deletions were modeled in a similar fashion. The residue to be deleted and one flanking residue on each side were removed. Connectivity was reestablished first with dialanine, the backbone conformation was modeled and the conformation of the side chains was then optimized. The uncertainty in the model is of course greatest in the vicinity of the insertions and deletions and we have not yet attempted to refine

194

J.L.SCULLY AND D.R.EVANS 4)

3)

Q

m

'x)

**

*

CAD K H C - R R DCaLIEFSGNV

90 S V Q LGK

110 QT bC I i i V S V I S T V . A

*

* * t

**

VGE S N Q

260

293

**

S

I

A

A

Fig. 1. The alignment of hamster and E. coli ATCase sequences. The hamster ATCase domain and E. mli ATCase catalytic chain sequences were aligned as described in the Methods section. Identities are indicated by shaded boxes, conservative substitution by unshaded boxes, and residues conserved in all

known ATCase sequences are indicated by ('). The secondary structure, including Q helices (single headed arrows), p-pleated sheets (solid lines), and turns (double headed arrows) defined by the X-ray structure of E. coli ATCase" is also shown. The E. coli ATCase C chain numbering system is used.

the conformation of these regions using a loop building a l g ~ r i t h m . ~ O - ~ ~ Using this approach we built a preliminary model of the mammalian ATCase domain. To test our model building procedures and as a guide in evaluating the model structures, we also constructed a model of t r y p ~ i using n ~ ~ chyrnotryp~in ~ temas~the plate and two bogus structures; a carboxypeptidase A model using the E. coli ATCase C chain as a template and the converse, a E. coli C chain model using the carboxypeptidase ~oordinates.'~ All of these proteins are approximately the same size.

votny et al.,57 even the bogus structures converged smoothly to a comparable low energy minimum. Superposition of the backbone of the mammalian model and E. coli ATCase X-ray structures (Fig. 2) indicated good agreement of the a carbon positions for these two structures. The RMS shift of the refined model structure was 0.93 8,comparable to the value obtained when the E. coli X-ray structure was minimized (Table I). As expected, the largest dislocation of the backbone occurred in regions of insertions and deletions with some a carbons displaced by as much as 2.2 A. The 240's loop, defined as residues 230-245 in the E. coli structure, was another region where significant displacement occurred in both the mammalian model structure (RMS shift = 1.21 and the minimized E. coli X-ray structure (RMS shift = 1.53 b). A plot of the mammalian ATCase model structure (Fig. 3) showed that all of the amino acids, including the inserted residue, the four residue inserted loop and those in the vicinity of deletions, had acceptable conformations.

Energy Minimization Energy minimization of the preliminary model structures was carried out using CHARMM following the five-step procedure, described above, in which the structural constraints were gradually relaxed. The results are summarized in Table I. The test model structures were also minimized using this procedure as were the X-ray structures of the E. coli ATCase catalytic subunit, carboxypeptidase A, chymotrypsin, and trypsin. After energy minimization, the covalent and noncovalent terms and the total energies were similar for all of the structures. As previously noted by NO-

++

Evaluation of the Mammalian ATCase Model Since the minimized potential energies and RMS shifts of the model ATCase structure, the trial model

195

MAMMALIAN ASPARTATE TRANSCARBAMYLASE STRUCTURE

TABLE I. Energy Minimization of Mammalian ATCase Monomer and Test Structures* Final energy (kcal)+ Covalent Noncovalent

RMS shift*

Model Template CAD ATC Model Structure CAD ECO

Total

tellns

tellns

(A)

No of atoms5

-4596

1159

-5756

0.93

2949

Test model structures TRP CHY CPA ECO ECO CPA

-3100 -3656 -4108

934 1145 1240

-4034 -4801 -5348

0.915 0.985 0.999

2011 2983 2986

X-ray structures ECO CPA CHY TRP

-4795 -4732 -3011 -3152

1136 1078 901 830

-5931 -5810 -3912 -3982

0.649 0.597 0.841 0.745

2952 2984 2058 2011

*The abbreviations refer to the true sequence (model) or X-ray coordinates (template) of CAD, E. coli ATCase catalytic subunit (ECO), bovine carboxypeptidase A (CPA), a-chymotrypsin (CHY), and trypsin (TFtP). 'Covalent terms include energy terms for bond lengths, bond angles, dihedral angles, and improper torsion angles. Noncovalent terms include van der Waals, electrostatic, and an implicit treatment of hydrogen bond energies.39 $a carbons. 5The polar hydrogens are included in the total number of atoms. CP DOMAIN

ASP DOMAIN

CP DOMAIN

ASP DOMAIN

Fig. 2. Backbone trace of the €. w/iATCase catalytic chain and the CAD ATCase domain. A stereo view of the CAD ATCase domain backbone (heavy lines) is superimposed on the €. w/i ATCase catalytic chain X-ray structure (lighter lines). Residue

numbers correspond to the €. w/iprotein. This orientation, selected because it illustrates the subdomain structure and the relationship of the 80's and 240s loop to the active site, is perpendicular to the molecular three fold axis.

structures and the X-ray structures are all comparable (Table I), other criteria were used to assess the validity of the model (Table 11). The energy minimization does not take hydrophobic interactions into account. The calculated hydrophobic interaction energy and solvent free energy of the CAD ATCase, the trypsin model and all of the X-ray structures were similar while the bogus structures gave appreciably higher values (Table 11) indicative of many more unfavorable hydrophobic interactions. The volume and surface accessible area of the CAD ATCase model were comparable to that of the E. coli C chain. Similarly the surface accessibility index, a parameter related to the compactness of the molecule, was 1.16for the CAD ATCase model, close to the values obtained for the E. coli ATCase, the

other X-ray structures, and the trypsin model. In contrast, the bogus models gave surface accessibility indices of 1.30 and 1.44 indicating a much more open tertiary structure. Another criterion for evaluating model structures is the extent to which hydrophobic residues are buried and hydrophilic residues are accessible to solvent. In the trypsin model and the X-ray structures, 22-31% of the hydrophobic residues and 74-78% of the hydrophilic residues were surface accessible while the bogus structures gave values 44-48% and 52-56%, respectively. The CAD ATCase model had 31% of its hydrophobic residues and 79% of its hydrophilic residues exposed to solvent, as would be expected for a correctly folded protein. These criterion cannot of course reveal whether or not the structure is correct in detail. There may be

196

J.L. SCULLY AND D.R. EVANS

135 1801', I 180 135

,

1-21' 90

-45

I

0

,

1

45

,

I

I

90

135

,

180

PHI (degrees) Fig 3 I++ plot of the CAD ATC domain model structure The open circles represent the +$ angles of one residue on either side of each insertion and deletion and all of the inserted residues in the energy minimized CAD ATCase domain model The angles for all of the other nonglycine residues in this structure are represented by the filled circles

errors especially in the inserted or deleted loops and in regions where there is a small displacement of the backbone of the model relative to the E . coli structure. Nevertheless, the parameters summarized in Table I1 are typical of a molecule with favorable hydrophobic interactions, a compact globular shape, and a normal distribution of hydrophobic and hydrophilic side chains suggesting that our model is a plausible representation of the structure of the mammalian ATCase monomer.

Distribution of Conserved Residues As expected the core of the ATCase domain is highly conserved. Of 108 buried residues defined as those having a surface accessibility of less than 10 A, 66 are identical in the CAD and E . coli sequences and 17 can be considered conservative substitutions. There is much more variation in the 190 surface residues (SA > 10 A,, 70 are identical, 25 are conservative substitutions, while 95 residues are distinctly different (Fig. 4). Of the 48 invariant residues found in all six prokaryotic and eukaryotic ATCases, 11 are involved in substrate binding and catalysis, 9 are located at the interface between the aspartate and carbamyl phosphate subdomains, and 14 are involved in trimer contacts. As shown in Figure 4, the lower surface of the trimer where the active site is located is more highly conserved than the upper surface. The four residue insertion that occurs between E . coli residues 37 and 38 is located on the upper surface of the trimer near the %fold axis.

Interactions Between the Polar and Equatorial Subdomains There are extensive interactions between the equatorial and polar subdomains of the CAD ATCase model structure with 77% of the 64 contacts found in E . coli ATCase conserved in the mamma-

lian protein. Most of the differences found in the interdomain contacts can be accounted for by changes in two regions. One region (Fig. 5 ) involves residues 178-180, the site of a one residue deletion. The loss of five interactions is partially compensated by two new contacts (Phe-24-Leu-175 and Leu-139Leu-175) and the interdomain contacts in this region are in general good. Also nine contacts involving the helical segment (residues 142-149) which connects the two subdomains are missing in the model structure. However, this region is stabilized by the 10 remaining interactions and four new contacts (Leu-139-Leu-175, Ile-142-Met-155, Glu-147Thr- 151, Gly-150-Lys-206).

Trimeric Subunit Contacts The model of the mammalian trimer was generated by rotation of the monomer about the 3-fold axis followed by energy minimization (see Methods). The surface accessibility of the monomer decreases from 13,912 to 11,770 A2 on incorporation into the trimer. Thus 15.4% of the accessible surface area is buried on trimerization, a value comparable to the 17.4% decrease in surface accessibility calculated from the E . coli ATCase X-ray structure. The CAD ATCase trimer so formed is a compact ellipsoid (SA index = 1.00)with the monomers closely juxtaposed throughout the interface region. There are no gaps or ill-fitting regions and extensive interchain contacts are present at the interface. Of the contacts (Cl-C2) formed on trimerization, 43 (28 nonpolar and 15 polar) were found to be identical in the mammalian and E . coli structures. There are 15 contacts in the E . coli structure which have no counterpart in the mammalian ATCase. For the most part the interactions absent in the mammalian structure are dispersed throughout the Cl-C2 interface region. The missing E. coli C1-C2 contacts in the CAD structure are compensated for by 13 interactions found only in the mammalian protein. Nine of these unique contacts are located on the upper surface of the trimer (see Figs. 4, 6) near the 3-fold axis and involve residues in the vicinity of the four residue insertion between amino acids 37 and 38 in the E . coli sequence. The 11 residue loop (residues 36-42), which includes the insertion, is very hydrophilic and has 6 charged residues (1 arginine, 3 lysines, 1 aspartate, and 1 glutamate). Among the probable new interchain contacts is a clearly defined salt link between C1 Glu-37 and C2 Arg-37, (Fig. 6) in the inserted segment. The bulky side chains in this region effectively block the channel which passes along the 3-fold axis between the subunits. There are also changes in the trimeric contacts near the 76-77 deletion in the CAD structure. Small shifts in the backbone and the replacement (Fig. 7A) of the Asn-78 with the smaller Ala side chain in CAD ATCase eliminates the interactions between

197

MAMMALIAN ASPARTATE TRANSCARBAMYLASESTRUCTURE

TABLE 11. Evaluation of the Mammalian ATCase and Test Model Structures Model structures Model template CAD TRP CPA ECO or structure* ECO CHY ECO CPA ECO Potential Energy (kcal)+ -4596 -3100 -3656 -4108 -4795 Electrostatic energy (kcal) Unshielded -4669 -3433 -4014 -4409 -4818 Solvent shielded -2094 -1891 -2185 -1985 -2317 Hydrophobic interaction energy (kcal) Hydrophobic energy -131 - 20 29 51 -135 -153 -140 -115 -116 Solvation free energy -155 Surface accessibility (nm2) x 10' 136 99 168 147 132 Volume (nm3) x 42.9 36.5 45.2 45.0 43.0 SA index 1.16 1.02 1.44 1.30 1.14 % Hydrophobic* 31.4 30.0 48.0 44.0 .22.0 Surface accessible % Hydrophilic* 78.9 74.0 52.0 56.0 78.0 Surface accessible

X-ray structures CPA CHY -4732 -3011 -4746 -2554 - 52 -163 120 42.0 1.02

-3201 -1573 - 24 -142 106 36.5 1.10

TRP -3152 -3289 -1745 23 -136 95 35.8 0.98 -

26.0

31.0

26.0

74.0

75.0

77.0

*Abbreviations given in Table I. 'Calculated from the energy minimized structures. 'Residues with a surface accessibility of 10 A' or greater are defined as surface accessible.

Fig. 4. The distribution of conserved residues in mammalian and E. coli ATCase. Two views of the surface of the CAD ATCase trimer: (A) bottom view along the 3-fold axis and (B) top view along the 3-fold axis, showing amino acid residues that are identical (blue),conservativesubstitutions(light blue), and nonconser-

vative substitutions (red) in the CAD ATCase domain and E. coli ATCase catalytic chain. The four residue insertion (yellow)is also visible in the top view. Note that the bottom of the trimer where the active site is located is more highly conserved than the upper surface.

the C1 Asp-75-C2 Asn-78 and C1 Ser-74-C2 Asn-78 found in the E . coli X-ray structure. However, the absence of these interactions in CAD is partially offset by the replacement of C1 Asp-75 with the longer GLU side chain (Fig. 7B) which brings the carboxyl group of this residue into contact with C2 Ser-74.

tein can be attributed to five additional basic residues qnd five fewer acid residues. Most of the charges (88%)in both E. coli and mammalian proteins are localized on the surface but the distribution of acidic and basic residues differs. The lower surface of the trimer where the active site is located is similar in both proteins (Fig. 8), while the distal face of the mammalian trimer has a much higher concentration of positively charged residues. Of 31 charged residues on the surface of the upper half of the CAD ATCase trimer, 24 are basic whereas the bottom half has about equal numbers of basic (15) and acidic (18) residues. The acidic and basic resi-

Distribution of Charged Residues The E . coli catalytic subunit and mammalian ATCase domains have an identical number of charged side chains (77 residues) but the isoelectric points are distinctly different, 6.2 and 9.4, respectively. The basic character of the mammalian pro-

198

J.L. SCULLY AND D.R.EVANS

Fig. 5. Stereoview of some interactions between the carbamyl phosphate and aspartate domains. New interactions are found between the carbamyl phosphate domain (heavy lines) and the aspartate domains (thin lines) in the CAD ATCase model.

I

Fig. 6. Stereoview of the inserted four residue loop in CAD ATCase. A top view down the 3-fold axis of the CAD ATCase trimer shows the four amino acid insertions (heavy lines) between E. coli residues 37 and 38.The inserted residues are labeled 37, through 37,. The active site is located on the opposite (bottom)face of the trimer. Six polar links (dotted lines), not found in the E. coli structure, may form between the monomers in the mammalian trimer

dues on both upper and lower surfaces are clustered in several charged patches. The equatorial domain is stabilized by a series of seven salt links (Fig. 9) not found in the prokaryotic protein (Arg-185 with Glu-210 and Asp-223; Arg-193 with Glu-211; Glu216 and Glu-217 with Arg-256; Asp-221 and Asp223 with Lys-258). The 133 residue connecting chain segment, which links the ATCase domain to the remainder of the CAD p~lypeptide,'~ emerges from the upper surface globular subunit 38 A from the %fold axis. It is interesting that this chain segment also carries a large positive charge (calculated pZ = 10.4). We have not attempted to model the structure of the linker but suspect that the positive charged patches on the upper surface of the ATCase domain and the basic residues in the linker participate in electrostatic interaction with the remainder of the CAD chimera which carries a net negative charge.

Active Site All but one of the E . coli ATCase residues that bind the bisubstrate analog, PALAl' are conserved in the hamster protein. The sole exception is the

CAD residue Met-267 which replaces leucine in the

E. coli sequence. Unlike all of the other residues that participate in PALA binding, the Met-267 interaction, a hydrogen bond between the backbone carbonyl oxygen and the peptide amide nitrogen of PALA does not involve the side chain. Potential substrate binding interactions in the CAD ATCase structure were examined by building PALA into the three active sites of the trimer model. The PALA conformation and its position and orientation within the site were modeled using the E . coli ATCase PALA complex as the template. The model was then subjected to 50 cycles of energy minimization constraining only the PALA bond angles and lengths. Minimization did not appreciably alter the position of the active site residues; the average RMS shift of active site residues was 0.51 while the average change in interatomic distances between PALA and the active site residues before and after energy minimization was 0.40 A. Comparable shifts were observed when the E.coli ATCase PALA X-ray structure was minimized. The interactions between the protein and bound PALA are virtually identical in the CAD model and

199

MAMMALIAN ASPARTATE TIZANSCARBAMYLASE STRUCTURE

A

B

2478

2474

Fig. 7. Stereoview of the trimeric contacts in the vicinity of the two residue deletion in CAD ATCase. Some of the interchain interactions between monomers found in the €. coli ATCase X-ray structure are shown in A. Residues 76 and 77 are deleted in the

mammalian ATCase domain and Asn-78 is replaced by Ala (B) with the loss of two hydrogen bonds. A compensatory substitution of Asp-75 with Glu in the mammalian protein establishes alternative interdomain interactions.

E. coli X-ray structures (Fig. 10). In identifying the

Honzatko et a1.l' made the surprising discovery that the active sites are shared between catalytic chains. The adjacent monomer contributes Lys-84 and Ser-80 to the constellation of residues which constitute the PALA binding site. &bey and Schachman5' confirmed this observation when they showed that activity was regenerated in hybrid molecules comprised of inactive derivatives in which either Lys-84 or Tyr-165 was modified. Both Lys-84 and Ser-80 are positioned appropriately to form the shared active site in the energy minimized CAD trimer (Fig. 8). Catalysis in E. coli ATCase involves a preferred ordered of substrate addition. The kinetic pathway in which carbamyl phosphate binds first followed by aspartate is favored.60.61The binding of aspartate analogs induces a conformational change that has been implicated in the allosteric transitions but also occurs in the isolated catalytic s u b ~ n i t .Local ~~,~~ structural changes induced in the catalytic subunit by PALA or carbamyl phosphate and the aspartate analog, succinate result in a more compact conformation.

active site residues, Krause et al.13 included potential hydrogen bonds, salt links, and other polar contacts with bond distances less than 3.5 A. All of the CAD contacts (Table 111)are within this range with the exception of one salt link between Lys-84 and the a carboxyl group of PALA. It should be noted that in the minimized E. coli structure the interaction between Lys-84 and the p carboxyl of PALA is also long (this residue shifted the most, 0.64 A, following energy minimization of the X-ray structure). Most of the PALA CAD interatomic distances are within 0.5 A of the corresponding distances found in the energy minimized E. coli structure. Gln-137, which falls just outside the 3.5 cutoff in the E. coli structure, appears to be somewhat closer in the CAD model and may form a hydrogen bond with the phosphonate moiety of PALA. His-134 (NE2), which has been postulated to function as a general base13 or to orient the carbonyl group of carbamyl phosphate5' during catalysis, is located 2.65 A from the carbonyl oxygen of PALA, about the same distance as seen in the E. coli X-ray structure.

Fig 8 The distribution of polar residues in the CAD ATCase trimer Two views of the surface of the E coh ATCase trimer (A) top view down the 3-fold axis and (B) bottom view along the 3-fold axis showing the distribution of basic (red), acidic (blue), and other polar sidechains (pink) Nonpolar residues are shown in gray The corresponding views (C and D)of the mammalian protein show that the upper surface of the CAD trimer is much more basic than the E COB enzyme

The mechanistic details of this conformational change were revealed by a comparison of the X-ray structures of unliganded E. coli ATCase and its PALA complex.'"^':' E . coli ATCase is poised in the low affinity T state in the absence of substrates (Fig. 11A). PALA binding induces a shift in the aspartate binding domain, which moves approximately as a rigid body 3 A toward the carbamyl phosphate domain around a hinge region located near Gly-130, a region which is well conserved in the CAD sequence. Specific interactions with PALA induce the movement of three loop regions bringing residues 52-55, 80, 84, 229, and 231 into the active site region. Domain closure, which accompanies t h e transition to the high affinity R state, thus establishes the active site which completely surrounds the bound substrates (Fig. 10). We postulate that a comparable domain closure also occurs in CAD (Fig. 11B). Domain closure generates new bridging interactions (Fig. 12A) between the two subdomains which stabilizes the high-affinity, high-activity conforma-

tion of the enzyme.':' Glu-50 and Arg-234, which are more than 10 A apart in the unliganded enzyme, move within 3 A to form a new salt link. Arg-167 also moves 2 A into position to interact with Glu-50 as well as PALA. The importance of these bonds in stabilizing domain closure was convincingly dem~ n s t r a t e d ' ~ , using '~ site-directed mutagenesis to replace these residues. Kinetic analysis showed that the mutant catalytic subunits were in a low-affinity, low-activity conformation. In the mammalian ATC model (Fig. 12B), Glu-50 is within 2.7 and 2.8 A of Arg-167 and Arg-234, respectively, suggesting that the same stabilizing interactions can form in the CAD domain. Another network of interactions has been postulated to stabilize the optimal conformation of the E. coli active site63; Lys-164 and Tyr-165 interact with Glu-239 and Asp-162 interacts with the backbone of Tyr-165 (Fig. 12A). These same interactions also occur in the CAD ATCase domain except that Tyr-165 is replaced by a histidine (Fig. 12B). His-165 is close enough to interact with Lys164 but cannot form a bond with Glu-239. However

MAMMALIAN ASPARTATE TRANSCARBAMYLASE STRUCTURE

201

A

Fig. 9. A network of charged residues on the surface of the CAD ATCase domain. The aspartate domain of mammalianATCase (B) is stabilized by a series of 8 salt bridges. Most of these interactions do not occur (A) in the corresponding region of the E. Coli ATCase catalytic subunit x-ray structure.

Fig. 10. Stereoview of the active site of the CAD ATCase domain. The bisubstrate analog PALA (heavy lines) was built into the active site of the CAD ATCase trimer as described in the text. Most of the active residues are contributed by one monomer, while Lys-84 and Ser-80 are located on the adjacent monomer. The dotted lines represent the polar contacts (Table 111) between PALA and the protein.

a new interaction between His-165 and Asp-162 was found in the mammalian model. Replacement of Tyr-165 with Ser67or Phe,6' or Glu-239 with Gln66 in the E . coli catalytic subunit does not abolish activity but results in a significant increase in the K , for aspartate. While a careful comparative kinetic study of the bacterial and mammalian enzymes has not been carried out, it is noteworthy that the mammalian enzyme has a K , for aspartate that is about two-to-four fold higher than the bacterial enzyme.

Unliganded E . coli ATCase exists in the T state conformation (Fig. 1 2 0 with the active sites in an open, low-affinity conformation.62Two T state hydrogen bonds, an intrasubunit interaction between Tyr-240 and Asp-271 and an intersubunit bond between Tyr-165 and Glu-239 are ruptured during domain closure. Interactions between triniers cannot occur in the isolated catalytic subunit but site directed mutagenesis66 of Asp-271 to aspargine produces a mutant catalytic subunit which has an ap-

202

J.L. SCULLY AND D.R. EVANS

TABLE 111. Residues Involved in PALA Binding CAD residue Ser-52 Thr-53 Arg-54 Thr-55 Ser-80 Lys-84 Arg-105 His-134 Gln-137 Arg-167 Arg-229 Gln-231 Met-267+

-

Atom OG N NH1 N OG1 OG1 OG OG NZ NZ NZ NH 1 NH 1 NH2 NE2 NE2 NH2 NE NH1 NH1 OE 1

0

PALA moiety atom Phosphonate 0 2 P Phosphonate 0 2 P Phosphonate 0 2 P Phosphonate 0 2 P Carbonyl 01 Phosphonate 03P Phosphonate 01P Phosphonate 0 2 P Phosphonate 01P a-Carboxylate 0 2 p-Carboxylate 0 5 Phosphonate 03P a-Carboxylate 0 2 Carbonyl 01 Carbonyl 01 Carbonyl 01 a-Carboxylate 0 2 a-Carboxylate 0 3 p-Carboxylate 0 4 p-Carboxylate 0 5 P-Carboxylate 0 4 PeDtide amide N2

Bond type H bond H bond Salt link H bond H bond H bond H bond H bond H bond Salt link Salt link Salt link H bond H bond H bond H bond Salt link H bond H bond H bond H bond H bond

Bond lengths* CAD E . coli 3.45 3.68 3.91 2.75 2.88 2.74 3.91 2.75 2.99 2.86 3.15 2.50 3.20 3.25 3.45 3.68 2.90 2.85 3.86 3.45 3.21 3.16 3.53 2.83 3.44 3.38 2.83 2.82 2.65 3.62 2.65 3.62 2.64 2.68 2.37 2.78 3.27 3.15 2.75 2.85 2.67 3.44 3.07 3.23

'The bond lengths were determined from the energy minimized CAD ATCase model and the E . coli ATCase PALA x-ray structure 'This interaction involves peptide carbonyl oxygen on the protein. Others are side chain interactions.

Fig. 11. Stereoview of the R and T conformations of the ATCase catalytic domain. The catalytic chain backbone derived from the X-ray structure of the E. coli ATCase PALA complex, the R state conformation, (heavy lines) is shown superimposed on the unliganded or T state conformation of E. coli monomer (thin lines) in A. .The number residues are those which move into the active

''

site region during the domain closure that accompanies the transition to the R conformation. Note that the movement of residues 80 and 84 is toward the active site on the adjacent monomer not shown. In 8, the CAD ATCase monomer (heavy lines), modeled using the E. coli ATCase PALA complex as the template, is shown superimposed on the E. coli (thin lines) T state structure.

MAMMALIAN ASPARTATE TRANSCARBAMYLASE STRUCTURE

203

A

C I I

Fig. 12. Stereoview of the interactions which stabilizes the R and T states. The interactions (dotted lines) which have been identified by X-ray studies and site-directed mutagenesis to stabilize the R state conformation of the E. coli catalytic subunit (A) are for the most part also present in the CAD ATCase domain (B). The bisubstrate analog PALA is also shown (heavy lines). The residues which participate in these interactions are conserved in mammalian ATCase with the exception of Tyr-165 which is replaced by His in the mammalian protein. This substitution of His for Tyr abolishes the interaction with Glu-239 and the Asp-162

backbone found in the E. coli X-ray structure but may allow new interactions with Lys-164 and Asp162. In the T state (C), replacement of the €. coliresidueAsp271 with Asn preserves, but weakens the hydrogen bond with Tyr-240, while the replacement of Tyr-165 in E. Coli with His (heavy lines) in the mammalian protein eliminates any possibility of an intersubunit bond to Glu-239. The orientation of the His-165 side chain was derived from a preliminary T state model, although this residue is too far to interact with Glu-239 in any orientation.

preciably higher affinity for aspartate suggesting a shift toward the high affinity R state. Asn-271 occurs naturally in the CAD sequence so that this interaction, which opposes domain closure, is much weaker in mammalian ATCase (Fig. 12C). The model described here corresponds to the R

state but preliminary modeling studies suggest that the CAD ATCase domain can assume a conformation closely resembling the T state conformation. However, the lack of regulatory subunits and T state interactions suggest that the R state conformation of the mammalian enzyme is more stable.

204

J.L. SCULLY AND D.R. EVANS

CONCLUSIONS

,

The tertiary fold of the mammalian and bacterial ATCase catalytic domain is almost identical suggesting that its basic architecture is very ancient. Nevertheless there are significant structural differences in these proteins, which probably reflect differences in the function and the structural organization of the molecule in which the domain is found. While the core of the mammalian domain is highly conserved, its surface has been extensively remodeled. In particular, the unusual distribution of charged residues on the upper face of the trimer, coupled with the high charge density of the linker, suggests that these positively charged regions of the mammalian protein may interact with the other CAD domains which carry a net negative charge at physiological pH. Moreover, the residues which participate in interactions between trimers and between catalytic and regulatory chains in the E . coli protein are poorly conserved in the mammalian protein. We anticipated that the trimeric contacts would be much more highly conserved than the model indicated. However, the functional requirements are more stringent for the allosteric bacterial protein. These interactions must not only maintain the integrity of the trimer and the correct juxtaposition of active site residues on adjacent subunits, a requirement of both bacterial and mammalian proteins, but also must mediate the transmission of allosteric signals between the monomers in the E . coli catalytic subunit. Thus the differences between the mammalian and bacterial protein may highlight residues that are especially important in intrasubunit communication. The additional salt links near the %fold axis on the upper surface of the trimer as well as an extensive network of interactions on the surface of the aspartate subdomain would be expected to confer greater rigidity to the mammalian protein. All of the residues implicated in substrate binding in the E. coli protein are present in the CAD domain and are within reasonable bonding distance of the appropriate atoms of the bisubstrate analog PALA. Nevertheless, the aspartate affinity of the mammalian protein is somewhat weaker than that of the E. coli catalytic subunit, a difference which may be accounted for in part by replacement of the Tyr-165 by histidine, although substitution of asparagine for Asp-271 should have the opposite effect. It is interesting that the K,,, for aspartate of the CAD domain” is comparable to that of the E. coli h ~ l o e n z y m e ,so~ ~that the substrate affinity is approximately the same for the naturally occurring forms of both mammalian and bacterial enzymes. The global changes in quaternary structure associated with the allosteric transition do not occur in the mammalian protein and there are insertions or deletions in the 80s and 240s loops shown to partic-

ipate in the cooperative interactions. In contrast, the local conformation changes within the catalytic domain are necessary for catalysis. Consequently, the interactions between the aspartate and carbamyl phosphate subdomains, the structure of the hinge region, and the contacts which stabilize the highactivity conformation are for the most part preserved in the mammalian protein. The CAD ATCase domain can probably thus assume the structure found in both T and R states of the E. coli holoenzyme. Unlike the E. coli holoenzyme, which must exist in the T state in the asbence of substrates in order to exhibit cooperativity, the equilibrium strongly favors the R state in both the E . coli and mammalian catalytic domains. Provided that both open and closed local conformations are accessible to the mammalian domain, the interactions which stabilize the T would have no functional advantage in a molecule which lacks allosteric control. Consequently it is understandable that a t least some of the constraints that stabilize the T state conformation of the E. coli enzyme are lost in the mammalian protein. In summary, the structural characteristics required for catalysis have remained for the most part unaltered throughout the course of evolution, while many of the interactions which have been implicated in the allosteric control mechanisms are absent in the unregulated mammalian ATCase domain.

ACKNOWLEDGMENTS We wish to thank Drs. William Lipscomb and Eric Goueaux for providing the X-ray coordinates and for helpful discussions and encouragement. This research was supported by Grants CA 27674 from the National Cancer Institute and an Office of Naval Research Contract N00014-87-K-0081. REFERENCES 1. Evans, D.R. CAD, a chimeric protein that initiates de nouo pyrimidine biosynthesis. In: “Higher Eukaryotes in Multidomain Proteindtructure and Evolution.” Hardie, D.G., Coggins, J.R. (eds.). New York: Elsevier, 1986283331. 2. Jones, M.E. Pyrimidine nucleotide biosynthesis in animals: Genes, enzymes and regulation of UMP biosynthesis. Ann. Rev. Biochem. 49:253-279, 1980. 3. Yates, R.A., ,Pardee, A.B. Control of pyrimidine biosynthesis in Escherichia coli by a feedback mechanism. J. Biol. Chem. 221:757-770, 1956. 4. Gerhart, J.C., Pardee, A.B. The enzymology of control by feedback inhibition. J. Biol. Chem. 2372391-896, 1962. 5. Bethell, M.R., Smith, K.E., White, J.S., Jones, M.E. Carbamyl phosphate an allosteric substrate for aspartate transcarbamylase of Escherichia coli. Proc. Natl. Acad. Sci. U S A . 60:1442-1446, 1968. 6. Gerhart, J.C., Schachman, H.K. Distinct subunits for the regulation and catalytic activity of aspartate transcarbamylase. Biochemistry 4:1054-1062, 1965. 7. Wiley, D.C., Lipscomb, W.N. Crystallographic determination of symmetry of aspartate transcarbamylase. Nature (London) 218:1119-1121, 1968. 8. Weber, K. New structure model of E. coli aspartate tran-

MAMMALIAN ASPARTATE TRANSCARBAMYLASE STRUCTURE scarbamylase and the amino acid sequence of the regulatory polypeptide chain. Nature (London) 218:1116-1119, 1968. 9. Meighen, E.A., Pigiet, V., Schachman, H.K. Hybridization of native and chemically modified enzymes, 111. The catalytic subunits of aspartate transcarbamylase. Proc. Natl. Acad. Sci. U.S.A. 65:234-241, 1970. 10. Rosenbusch, J.P., Weber, K. Subunit structure of aspartate transcarbamylase fkom Escherichia coli. J . Biol. Chem. 246~1644-1657,1971. 11. Honzatko, R.B., Crawford, J.L., Monaco, H.L., Ladner, J.E., Edwards, B.F.P., Evans, D.R., Warren, S.G., Wiley, D.C., Lipscomb, W.N. Crystal and molecular structures of native and CTP-liganded aspartate transcarbamylase from Escherichia coli. J. Mol. Biol. 160:219-263,1982. 12. Ke, H-M., Honzatko, R.B., Lipscomb, W.N. Structure of unligated aspartate transcarbamylase of Escherichia coli at 2.6 A resolution. Proc. Natl. Acad. Sci. U.S.A. 81:40374040,1984. 13. Krause, K.L., Volz, K.W., Lipscomb, W.N. 2.5 A structure of aspartate transcarbamylase complexed with the bisubstrate analog N-(phosphonacety1)aspartate. J. Mol. Biol. 193:527-553,1987. 14. Kim, K.H., Pan, Z., Honzatko, R.B., Ke, H-M., Lipscomb, W.N. Structural asymmetry in the CTP-liganded form of aspartate transcarbamylase from Escherichia coli. J. Mol. Biol. 196:853-875,1987. 15. Shoaf, W.T., Jones, M.E. Uridylic acid synthesis in Ehrlich ascites carinoma. Biochemistry 12:4039-4051, 1973. 16. Mori, M., Ishida, H., Tatibana, M. Aggregation state and catalytic properties of the multienzyme complex catalyzing the initial steps of pyrimidine biosynthesis in rat liver. Biochemistry 142622-2630,1975. 17. Coleman, P.F., Suttle, D.P., Stark, G.R. Purification from hamster cells of the multifunctional protein that initiates de nouo synthesis of pyrimidine nucleotides. J . Biol. Chem. 252:6378-6385, 1977. 18. Tatibana, M., Shigesada, K. Control of pyrimidine biosynthesis in mammalian tissues. V. Regulation of glutamine dependent carbamyl phosphate synthetase: Activation by 5-phosphoribosyl 1-pyrophosphate and inhibition by uridine triphosphate. J . Biol. Chem. 72549-560, 1972. 19. Lee, L., Kelly, R.E., Pastra-Landis, S.C., Evans, D.R. Oligomeric structure of the multifunctional protein CAD that initiates pyrimidine biosynthesis in mammalian cells. Proc. Natl. Acad. Sci. U S A . 82:6802-6802, 1985. 20. Mally, M.I., Grayson, D.R., Evans, D.R. Controlled proteolysis of the multifunctional protein that initiates pyrimidine biosynthesis in mammalian cells: Evidence for discrete structural domains. Proc. Natl. Acad. Sci. U.S.A. 78: 6647-6651,1981. 21. Davidson. J.N.. Rumsbv. P.C.. Tamaren. J . Organization of a multifunctional proGin i n pyrimidine biosynthesis. J . Biol. Chem. 256:5220-5225,1981. 22. Grayson, D.R., Evans, D.R. The isolation and characterization of the aspartab transcarbamylase domain of the multifunctional protein, CAD. J . Biol. Chem. 25841234129, 1983. 23. Simmer, J.P., Kelly, R.E., Scully, J.L., Grayson, D.R., Rinker, Jr., A.G., Bergh, S.T., Evans, D.R. Mammalian aspartate transcarbamylase (ATCase): Sequence of the ATCase domain and interdomain linker in the CAD multifunctional polypeptide and properties of the isolated domain. Proc. Natl. Acad. Sci. U S A . 86:4382-4386,1989. 24. Shigesada, K., Stark, G.R., Maley, J.A., Davidson, J.N. Construction of a cDNA to the hamster CAD gene and its application towards defining the domain for aspartate transcarbamylase. Mol. Cell Biol. 1751-7, 1985. 25. Greer, J . Comparative model-building of the mammalian serine proteases. J. Mol. Biol. 153:1027-1042,1981. 26. Murphy, M.E., Moult, J., Bleackley, R.C., Gershenfeld, H., Weissman, I.L., James, M.N.G. Comparative molecular model building of two serine proteases from cytotoxic T lymphocytes. Proteins: Struct. Funct. Genet. 4:190-204, 1988. 27. Read, R.J., Brayer, G.D., Jurasek, L., James, M.N.G. A critical evaluation of comparative model building of Streptomyces griseus trypsin. Biochemistry 23:6570-6575, 1984. 28. Blundell, T., Sibanda, B.L., Pearl, L. Three-dimensional

205

structure, specificity and catalytic mechanism of renin. Nature (London) 304273-275, 1983. 29. Carlson, W., Karplus, M., Haber, E. Construction of a model for the three-dimensional structure of human renal renin. Hypertension 7:13-26, 1985. 30. Chothia, C., Lesk, A.M., Levitt, M., Amit, A.G., Mariuzza, R.A., Phillips, S.E.V., Poljak, R.J. The predicted structure of immunoglobulin D1.3 and its comparison with the crystal structure. Science 233:755-758, 1986. 31. de la Paz, P., Sutton, B.J., Darsley, M.J., Rees, A.R. Modeling of the combining sites of three anti-lysozyme monoclonal antibodies and the complex between one of the antibodies and its epitope. EMBO J. 5415-425, 1986. 32. O”ei1, K.T., DeGrado, W.F. A predicted structure of calmodulin suggests an electrostatic basis for its function. Proc. Natl. Acad. Sci. U.S.A. 82:4954-4958, 1985. 33. Strynadka, N.C.J., James, M.N.G. Two trifluoperazine binding sites on calmodulin predicted from comparative modeling with troponin-C. Proteins Struct. Funct. Genet. 3:l-17, 1988. 34. Rath, V.L., Newgard, C.B., Sprang, S.R., Goldsmith, E.J., Fletterick, R.J. Modeling the biochemical difference between rabbit muscle and human liver phosphorylase. Proteins Struct. Funct. Genet. 2:225-235, 1987. 35. Ripka, W.C. Computer-assisted model building. Nature (London) 321:93-94,1986. 36. Hoover, T.A., Roof, W.D., Foltermann, K.F., O’Donovan, G.A., Benchini, D.A., Wild, J.R. Nucleotide sequence of the structural gene (pyrB) that encodes the catalytic polypeptide of aspartate transcarbamylase of Escherichia coli. Proc. Natl. Acad. Sci. U.S.A. 80:2462-2466,1983. 37. Schachman, H.K., Pauza, C.D., Navre, M., Karels, M.J., Wu, L., Yang, Y.R. Location of amino acid alterations in mutants of aspartate transcarbamylase: Structural aspects of interallelic complementation. Proc. Natl. Acad. Sci. U.S.A. 81:115-119,1984. 38. Risler, J.L., Delorme, M.O., Delacroix, H., Heraut, A Amino acid substitutions in structurally related proteins A pattern recognition approach. J. Mol. Biol. 2M10191029,1988. 39. Brooks, B.R., Bruccoleri, R.E., Olafson, B.D., States, D.J., Swaminathan, S., Karplus, M. CHARMM: A program for macromolecular energy minimization and dynamics calculations. J. Comput. Chem. 4187-217, 1983. 40. Lee, B.,Richards, F.M. The interpretation of protein structure: Estimation of static accessibility. J . Mol. Biol. 55: 379-400, 1971. 41. Chothia, C. Structural invariants of protein folding. Nature (London) 254304-308. 1975. 42. Janin, J . Surface area of globular proteins. J . Mol. Biol. 10513-14, 1976. 43. Teller, D.C. Accessible area, packing volumes, and interaction surface of globular proteins. Nature (London) 230 729-731,1976, 44. Northrup, S.H., Pear, M.R., Morgan, J.D., McCammon, J.A., Karplus, M. Molecular dynamics of ferrocytochrome c. J. Mol. Biol. 153:1087-1109, 1981. 45. Eisenberg, D., McLachlan, A.D. Solvation energy in protein folding and binding. Nature (London) 319:199-203, 1986. 46. Michaels, G., Kellin, R.A., Nargang, F.E. Cloning, nucleotide sequence and expression of the pyrBI operon of Sulmonella typhimurium LT2. Eur. J. Biochem. 166:55-61, 1987. 47. Lerner, C.G., Switzer, R.L. Cloning and structure of the Bucillus subtilis aspartate transcarbamylase gene (pyrB). J. Biol. Chem. 26411156-11165, 1986. 48. Freund, J.N., Jarry, B.P. The rudimentary gene of Drosophilu melanogaster encodes four enzymic functions. J . Mol. Biol. 193:l-13, 1987. 49. Faure, M., Camonis, J.H., Jacquet, M. Molecular characterization of a Dictyostelium dkcoideum gene encoding a multifunctional enzyme of the pyrimidine pathway. Eur. J. Biochem. 179:345-358, 1989. 50. Moult, J., James, M.N.G. An algorithm for determining the conformation of polypeptide segments in proteins by systematic search. Proteins: Struct. Funct. Genet. 1:146163,1986. 51. Fine, R.M., Wang, H., Shenkin, P.S., Yarmush, D.L., Levinthal, C. Predicting antibody hypervariable loop con-

206

J.L. SCULLY AND D.R. EVANS

formations 11. Minimizationand molecular dynamics studies of MCP603 from many randomly generated loop conformations. Proteins struct. Fund. Genet. 1:342-362,

61.

1986. 52. Jones, T.A., Thirup, S. Using known substructures in protein model building and crystallography. EMBO J. 5:819822, 1986. 53. Bruccoleri, R., Karplus, M. Prediction of the folding of 54.

55. 56. 57.

short polypeptide segments by uniform conformational sampling. Biopolymers 26137-168, 1987. Walter, J., Steigemann, W., Singh, T.P., Bartunik, H., Bode, W., Huber, R. On the disorder activation domain in trypsinogn. Chemical labelling and low-temperature crystallography. Acta Crystallogr., Sect. B 38:1462, 1982. Tsukada, H., Blow, D.M. Structure of alpha-chymotrypsin refined a t 1.68 A resolution. J . Mol. Biol. 184:703, 1985. Rees. D.C.. Lewis. M.. Liuscomb. W.N. Refined crvstal struckre of CarboxypeptidAeat 1.54 A resolution. J."Mol. Biol. 168:367, 1983. Novotny, J., Bruccoleri, R.E., Karplus, M. An analysis of incorrectly folded protein models. J. Mol. Biol. 177:787-

818,1984. 58. Kleanthous, C., Wemmer, D.E., Schachman, H.K. The role

62.

63.

64.

65.

66.

of an active site histidine in the catalvtic mechanism of aspartate transcarbamylase. J . Biol. Chem. 263:13062-

carbamylase: Kinetic studies of the catalytic subunit. J. Biol. Chem. 2M1846-1859, 1969. Hsuanyu, Y.,Wedler, F.C. Kinetic mechanism of native Escherichia coli aspartate transcarbamylase. Arch. Biochem. Biophys. 259316-330,1975. Gerhart, J.C., Schachman, H.K. Allosteric interactions in aspartate transcarbamylase. 11. Evidence for different conformational states of the protein in the presence and absence of specific ligands. Biochemistry 7:538-552, 1968. Kirschner, M.W., Schachman, H.K. Conformational changes in proteins as measured by different sedimentation studies. 11. Effect of stereospecific ligands on the catalytic subunit of aspartate transcarbamylase. Biochemistry 10:1919-1926, 1975. Ladjimi, M.M., Middleton, S.A., Kelleher, K.S., Kantrowitz, E.R. Relationship between domain closure and binding, catalysis, and regulation in Escherichiu coli aspartate transcarbamylase. Biochemistry 27:268-276, 1988. Middleton, S.A., Kantrowitz, E.R. Function of arginine 234 and aspartic acid 271 in domain closure, cooperativity, and catalysis in Escherichiu coli aspartate transcarbamylase. Biochemistry 27:8653-8660, 1988. Ladjimi, M.M., Kantrowitz, E.R. A possible model for the concerted allosteric transition in Escherichia coli aspartate transcarbamylase as deduced from site directed mutagenesis studies. Biochemistry 27:276-283, 1988. Robey, E.A., Schachman, H.K. Site directed mutagenesis of aspartate transcarbamylase. J. Biol. Chem. 25911180-

13067, 1988. 59. Robey, E.A., Schachman, H.K. Regeneration of active en-

67.

zyme by formation of hybrids from inactive derivatives: Implications for active sites shared between polypeptide chains of aspartate transcarbamylase. Proc. Natl. Acad. Sci. U S A . 82:361-365, 1985. 60. Porter, R.W., Modebe, M.O., Stark, G.R. Aspartate trans-

11183,1984. 68. Wales, W.E., Hoover, T.A., Wild, J.R. Site specific substitutions of the Tyr 165 residue in the catalytic chain of

aspartate transcarbamylase promotes a T state preference in the holoenzyme. J . Biol. Chem. 263:6109-6114, 1988.