Reports
`
`The Predicted Structure of Immunoglobulin D 1.3 and
`Its. Comparison with the Crystal Structure
`
`CYRUS CHOTHIA, ARTHUR M. LESK, MICHA.EL LEVITT,
`ADOLFO G. AMIT, Roy A. MARrnzzA, SIMONE. V. PHILLIPS,
`ROBERTO J. POLJAK
`
`Predictions of the structures of the antigen-binding domains of an antibody, recorded
`before its experimental structure determination and tested subsequently, were based
`on comparative analysis of known antibody structures or on conformational energy
`calculations. The framework, the relative positions of the hypcrvariable regions, and
`the folds of four of the hypervariable loops were prewcted correctly. This portion
`includes all residues in contact with the antigen, in this case hen egg white lysozyme,
`implying that the main chain conformation of the antibody combining site docs not
`change upon ligation. The conformations of three residues in each of the other two
`hypcrvariable loops are different in the predicted models and the experimental
`structure.
`
`domains and the V L -V H packing, and hy­
`pervariable loops in the regions that link 13
`sheets (5-ll). In the five known VL struc­
`tures, 69 residues in two 13 sheets have the
`same conformation. The root-mean-square
`differences in position (A) of their main
`
`chain atoms are between 0.5 and 1.6 A for
`different pairs of V L domains. The 13-sheet
`residues adjacent to the hypcrvariable re·
`gions differ in position by less than 1.5 A
`(2). Similarly, 79 residues in the 13 sheets of
`d1ree V H domains have the same conforma­
`tions. The relative geometry of the V L -V H
`packing is also conserved (12). This con­
`served framework structure is produced by
`conservation of residues buried between the
`sheets and at the VL-VH interface. In the
`monoclonal antibody IgG Dl.3 (13), 56 of
`the 62 buried framework residues (Table 2)
`are identical to those in known structures;
`the other six differ in size by no more than a
`methyl group. We predicted the framework
`structure of Dl.3 to be the same as that of
`the known structures.
`The antigen-binding site contains three
`hypervariable loops from V L and three from
`VH, denoted Ll, L2, L3, and Hl, H2, H3
`(11). The residues, nun1bered according to
`Kabat et al. (5), are:
`Vi..
`26-32
`Residues
`50-52 H2
`H3
`91-96
`
`Region
`HI
`
`VH
`26-32
`Residues
`53-55
`96-101
`
`Region
`Ll
`
`L2
`L3
`
`NTIBODIES ACHIEVE THEIR RANGE
`
`A
`
`of specificity in antigen recognition
`through the modulation of the con­
`formation of specific loops by changes in the
`an1ino acid sequence. An understanding of
`how the conformations of these loops arc
`determined is essential for an appreciation of
`the basic mechanism of antibody affinity and
`specificity, and for the design of modifica­
`tions of immunoglobulin structures.
`We describe here the prediction of the
`conformation of the antigen-binding do­
`mains of immunoglobulin G (IgG) Dl.3,
`and the comparison of the predicted models
`with the e,xperimental structure determined
`from the 2.8 A resolution electron density
`map (1). Two methods of prediction were
`used. One was based on analysis of the
`known structures of the variable (V) do­
`mains of light (L) and heavy (H) chains (V L
`and V H) (2) of the myeloma proteins
`NEWM, McPC603, KOL, REI, and RHE
`(Table 1). From these structures, we identi­
`fied the relatively few residues that, through
`packing, hydrogen bonds, or ability to as­
`sume unusual conformations (Gly, Pro,
`Asn, Asp), are primarily responsible for the
`conformations of the regions in these do­
`mains involved in antigen binding: the hy·
`pervariable loops. In contrast, de la Paz et al.
`have based predictions on the length and
`overall sequence homology of the hypervar­
`iablc regions (3). Our other predictions
`were based on conformational energy calcu­
`lations ( 4).
`Comparisons of the atomic structures of
`variable domains confirm and refine the
`distinction between a conserved framework
`consisting of the 13 sheers of the individual
`
`15 AUGUST 1986
`
`L
`
`fab' NEWM >·J
`
`Fab KOL
`Vi,_REI
`V1..RHE
`
`K
`>.I
`
`in immunoglobulin variable domains.
`
`At the interfuce between the rwo sheets in V 1..
`domains:
`
`These residues are outside the common 13·
`sheet framework of the molecules of known
`structure. [The complementarity-determin­
`Table 1. Immunoglobulin variable domains
`ing regions (CDR's) defined by Kabat et al.
`known to atomic resolution from x·ray crystal
`(5) on the basis of sequence comparisons are
`structure analysis. A fab is a fragment of an
`more extensive.] The homology of the hy­
`immunoglobulin containing variable domains V 1..
`and V H and coDstant domains Ci,_ and CH 1•
`pcrvariable loops of Dl.3 with those of
`known structures of the same class is shown
`Chain
`Reso·
`in Table 3. Fewer than 10 ofd1e 32 residues
`lution Refer-
`Protein
`cype
`in the hypervariable regions oflgG D 1.3 are
`encc
`(AJ
`2.0 (6)
`the same as those in homologous positions
`H
`fab McPC603 K
`2.7 (7)
`in any single known structure.
`>.I 'Y{ll 1.9 (8)
`'YI!
`2.0 (9)
`We analyzed the determinants of the con­
`'Yl
`1.6 (10)
`formation of the hypcrvariable loops.
`L l: The structure of the L l region is
`characteristic of the class of the domain (14).
`The V" class includes the V L domains of
`Table 2. Residues buried in sheet-sheet packings
`residues long and 7 in REI. The common
`REI, McPC603, and the target structure
`Dl.3 (Table 1). In McPC603, Ll is 13
`residues have the same main chain confor­
`4,6,19,21,23,25,33,35,37,47,48,62,64,
`mation: for 26 to 29 and 32, A = 0.5 A.
`71, 73, 75, 82, 84, 86, 88, 90, 97, 99, 101,
`102, 104
`C. Olothia, MRC Laboratory of Molecular Biology,
`
`
`HiUs Road, Cambridge CB2 2QH, England, and Chris­
`
`topher Ingold Laboratory, University CoUcge London,
`Hills Road, Cambridge CB2 2QH, England, and Fair·
`At the interface between the two sheets in V H
`4,6,18,20,22,24,34,36,38,48,49,51,69,
`20 Gordon Street, London WCIH OAJ, England.
`A. M. Lcsk, MRC Laboratory of Molecular Biology,
`78, 80, 82, 86, 88, 90, 92, 104, 106, 107,
`domains:
`109
`
`
`leigh Dickinson University, Teaneck-Hackensack Cam­
`pus, Teaneck, NJ 07666.
`A.G. Amit, R. A. Mariuzza, R. J. Poljak,
`M. Levitt, Department
`
`of Chemical Physics, Weizmann
`Vi..: 36, 38,44, 87, 98
`
`
`lnStinitc of Science, 76100 Rehovot, Israel.
`At the interface between the V 1. and V H domains:
`Dcpartement
`V H: 37, 39, 45, 47, 91, 93, 103
`
`
`
`d'lmmunologie Sm1cnirale, Institute Pasteur, 25-28,
`of Leeds, Leeds LS2 9J1', England.
`rue du Docteur Roux, 75724 Paris, France.
`
`S. E. V. Phillips, Astburv Department of Biophysics,
`REPORTS 755
`University
`
`1 of 4
`
`BI Exhibit 1078
`
`

`

`the total volume of Phc 29 and Met 34 in
`KOL and McPC603.
`H2: In KOL, NEWM, and McPC603,
`H2 differs in size and conformation. In
`NEWM, the three residues in H2 are part of
`a seven-residue hairpin rum wid1 a Gly in
`the fourth position. Such turns have the
`same conformation whenever they occur in
`known structures (JS). As H2 in Dl.3 also
`has three residues and a Gly at this position,
`we predicted it to have the NEWM confor·
`mation.
`H3: The length and confonnation of H3
`differ in KOL, NEWM, and McPC603.
`Medium and large H3 loops extend from
`the surface of the domain. In McPC603 the
`
`confonnarion of the stem of H3 is deter·
`mined primarily by the packing ofTyr lOOb
`into a cavity between the V 1.. and V H do­
`mains, its OH group remaining accessible.
`l11e homologous residue in Dl .3, Arg 99, is
`also large. We predicted it would pack in the
`same manner ro produce a similar H3 con­
`formation.
`We began the construction of the model
`based on the analysis of known strucrures
`with the framework regions of V L REI and
`VH KOL. Of those domains of known
`strucrure of the same classes as those of
`DI.3, these had been determined at the
`highest resolution. The overall sequence ho·
`mologics with D 1.3 arc 66 percent (V 1..)
`
`26
`··-.(._
`l1
`
`32
`
`L2
`
`L3
`
`l0�.
`s3(\ss t� .
`32.;;_
`
`···
`
`.. /
`...
`!
`
`i'
`
`.. f
`26
`
`H1
`
`94 <,
`'\ I'
`�102
`H3
`H2
`Fig. L The main chain atoms of the hypervariable regions in rhc cxperimcncaJ structure (---) and in
`the predicted model based on analysis of the known sm1Cturc' <--). Jn a 2.8 A electron density map,
`the exacr orientations of peptides outside secondary structures may nor be weU determined. This figure
`concaiJ1S independent pictures of these regions, each shown in ,1 viewpoint perpendicular to the mean
`plane of rhe loop.
`
`They adopt an extended confonnarion,
`across the rop of the Vt. domain. The six
`additional residues in McPC603 fonn a loop
`extending away from the domain. The L l
`confom1ation is determined primarily by
`packing of residue 29 against residues 2, 25,
`33, and 71 in the inrerior of the framework.
`In Dl.3, Ll has the same length as in REI,
`and the residues packing against the frame­
`work are conserved:
`
`Residue:
`01.3
`REI
`McPC603
`
`25
`2
`Ile Ala
`Uc Ala
`Ile
`Ser
`
`29
`Uc
`Ile
`lie
`
`33
`7l
`Leu Tyr
`Leu Tyr
`Leu Tyr
`
`We predicted Ll of Dl.3 to have the REI
`conformation.
`L2: In four V L structures L2 is a three·
`residue rurn (in NEWM, L2 is deleted).
`Their structures arc similar, as confonna­
`tions of short rums are restricted by the
`allowed values of the main chain torsion
`angles (15). We predicted that L2 in 01.3
`would also have this common strucrure.
`L3: The L3
`regions
`in REI and
`McPC603 are each six residues long and
`have the same structure: fl = 0.4 A. The
`conformations usually observed for such
`hairpin loops would involve main chain
`hydrogen bonds between residues 92 and
`95. The residue Pro 95, present in most V.
`domains including REI, McPC603, and
`D 1.3, precludes this confom1ation (Fig. l ).
`In REI and McPC603, Pro 95 has a cis
`peptide, and the side chains of Asn 90
`(REI) and Gin 90 (McPC603) fonn hydro­
`gen bonds ro L3 main chain atoms. In D 1.3,
`L3 is also six residues long and includes
`Pro 95 and His 90. If His 90 could form the
`same hydrogen bonds as Gin or Asn do in
`REI and McPC603, we should expect a
`similar L3 conformation in D 1.3.
`H l: In KOL and McPC603 the seven
`residues have similar conformations (fl =
`0.4 A). ln KOL and McPC603, Gly 26
`produces a sharp rum. Residues 28 to 32 are
`a helix with Phe 29 packing in a cavity
`formed by the side chains of residues 34 and
`94 and the main chain of 72 and 77. In
`D 1.3, H l also has seven residues, and the
`important side chains arc similar bur nor all
`identical:
`
`' '
`
`' '
`
`L2
`
`H1
`
`H2
`
`Antibody
`
`26
`
`01.3
`Gly
`KOL
`Gly
`McPC603 Gly
`
`Residue
`
`27
`
`Phc
`Phe
`Phe
`
`29
`
`34
`
`94
`
`Leu Val
`Tyr
`Phc Mer Tyr
`Phe Met Tyr
`
`.. ___ ... \
`\ \
`
`-c: L3
`
`In particular, two important residues arc
`both smaller in D 1.3. l11e coral volume of
`Leu 29 and Val 34 in DI. 3 is 65 A 3 Jess than
`756
`
`Fig. 2. The main chain atoms of rhe pan of the binding sire that makes conractwich the antigen (1), in
`the experimental srrucrurc (--) and in the predicted model based on analysis of the known structures
`(--). The acnaal relative posirions of the regions in the srrucrnrc are shown from a viewpoint
`roughly perpendicular ro the plane of the anrigen·amibody interface.
`
`SClllNCE, VOL. 233
`
`2 of 4
`
`BI Exhibit 1078
`
`

`

`and 49 percent (Vtt)· To combine the do­
`mains we superposed V L REI onto V L
`KOL by fitting the conserved V L residues of
`the domain-domain interface. We took L l
`and L2 from V L REI and H2 from V H
`NEWM. For L3, noting the conserved Pro,
`and believing on the basis of model building
`that His 90 could make the same hydrogen
`bonds as Gin, we used the loop from REI.
`In the H l region of Dl.3, two crucial
`residues are smaller than those in the known
`structures. Nevertheless, we did not expect
`this loop to change its conformation radical­
`ly, and took Hl from V H KOL.
`No known structure has an H3 region of
`the same length as D 1.3. Faced with alterna­
`tive conformations in known structures, we
`first chose to build a loop based on NEWM
`H3
`(for reasons we now believe to be
`incorrect). Informed that this initial predic­
`tion was qualitatively incorrect, we then
`based our prediction on McPC603, which
`proved to be the proper model. To correct
`the length we removed three residues from
`the apex of the loop-these were chosen to
`preserve the pattern of hydrogen bonding­
`and closed the chain using the interactive
`graphics program FRODO
`(16). Side
`chains ofDl.3 were substituted wherever its
`sequence differed from the parent structures,
`and bad stcric contacts relieved by changing
`their confom1acions.
`Two other models were constructed inde­
`pendently witl1 the use of conformational
`energy calculations. One was based exclu­
`sively on McPC603, the molecule of known
`structure having the highest homology with
`Dl.3 in both V regions. The side chains of
`McPC603 were replaced witl1 their Dl.3
`homologs, and insertions and deletions
`made without, initially, preserving the con­
`tinuity of the backbone. Equivalent atoms
`were taken from the McPC603 structure
`whenever possible; 125 missing atoms (out
`of 1977) were placed at random within 1 A
`of the acorn of known position to which
`they arc bonded. Energy minimization
`closed the gaps in the chain; eliminated close
`contacts; fom1cd strong hydrogen bonds;
`and brought all bond lengths, bond angles,
`and torsion angles close to their standard
`values
`(4). After 600 conjugate gradient
`cycles, the total energy was -1655 kcal/mo!.
`A second model merged features from
`each of the three known Fab structures, by
`selecting seven stretches of residues locally
`most homologous in sequence to Dl.3:
`From McPC603 V L> residues 1 to 17 and
`27to 107; KOL VL, 18 to26; NEWM Vtt,
`1to67 and 90 to 101; and KOL Vtt, 68 to
`89 and 102 to 115. Only 100 atoms were
`missing. Minimization reduced the total en­
`ergy to -1681 kcal/mo!.
`We compared the predicted strucrurcs
`
`IS AUGUST 1986
`
`with the experimental structure of D 1.3
`derived from the 2.8 A electron density map
`of a crystal containing the Fab fragment of
`Dl.3 binding tl1e antigen hen egg white
`Iysozymc (1). In Fig. 1, the experimental
`hypervariable regions and those predicted
`from structural analysis arc superposed. The
`observed antigen-binding site (J) and the
`homologous regions of this predicted struc­
`ture are shown in Fig. 2. Deviations in
`atomic positions of observed and predicted
`structures are shown in Tables 4 ro 7.
`Results are given for the framework struc­
`tures, the residues adjacent co tl1c hypervar­
`iable regions, the hypervariablc regions
`themselves, and a maximal set of well-fitting
`residues that includes tl1c lysozymc binding
`site. Atoms included in superposing the
`individual loops are N, Ca, C, and CJ3. The
`
`carbonyl 0 is omitted because in a 2.8 A
`electron density map the orientations of
`peptides outside secondary structures may
`not be well determined. The Cp is included
`to evaluate ilie general positioning of side
`chains.
`ilie
`In th.is model, the framework and
`folds of Ll, L2, H2, and H3 were predicted
`correctly (Table 6). Three residues in the
`middle of L3 are different. In ilie current
`interpretation of the electron density map,
`the peptide of Pro 95 has a trans confom1a­
`tion; in the predicted structure it is cis (Fig.
`I). Although the trans conformation is fa­
`vored by the current map, a cis conformation
`cannot be ruled out. Three residues at ilic
`center ofH 1 arc very different; in particular,
`in ilie experimental soucrure the side chain
`of Leu 29 does not lie in il ie cavity occupied
`
`Table 3. The homology of the hypervariable loops oflgG Dl.3 and other immunoglobulins of known
`structure and the same class. References to the structures arc given in Table 1. Residue numbers for the
`hypcrvariable regions arc given in the ccxr.
`
`Ll R!>I
`
`Ser Gln Asp Ile Ile Lys
`
`Tyr
`
`McPC603
`
`Ser Gln Ser Len Leu Asn Ser Gly ASn Gln Lys Asn Phc
`
`Dl.3
`
`Ser Gly Asn Ile
`
`fiis Asn
`
`Tyr
`
`REI
`
`Glu Ala Ser
`
`L2 Mcl'C603
`
`Gly Ala Ser
`
`Dl..3
`
`Tyr Thr Thr
`
`REI
`
`Tyr Gln Ser Leu Pro Tyr
`
`L3 Mcl'C603
`
`Asp His Ser Tyr Pro Leu
`
`Dl.3
`
`Phe Trp Ser Thr Pro Arg
`
`Ill KOL
`
`NEW
`
`Gly Phe Ile Phc Ser Ser Tyr
`
`Gly Thr Ser Phe Asp Asp Tyr
`
`McPC603
`
`Gly Phe Thr Phe Ser Asp Phe
`
`Dl.3
`
`Gly Phe Ser Leu Thr Gly Tyr
`
`112 KOL
`
`Asp Asp
`
`Gly Ser
`
`NEW
`
`Tyr His
`
`Gly
`
`McPC603
`
`Asn Gly Lys Asn Lys
`
`Dl.3
`
`Gly Asp
`
`Gly
`
`113 KOL
`
`Gly Gly Eis Gly Phe Cys Ser Ser Ala Ser Cys Phe Gly Pro Asp
`
`NEW
`
`Leu Ile Ala
`
`.McPC603
`
`Tyr Tyr (./ly Ser
`
`Dl.3
`
`lug Asp
`
`Gly Cys Ile Asp
`
`Thr Trp Tyr Phe Asp
`
`Tyr Arg Leu Asp
`
`REl'ORTS 757
`
`3 of 4
`
`BI Exhibit 1078
`
`

`

`by Phe 29 in KOL and McPC603, but is
`external. Although chc experimental scn1c­
`ture may have to be modified, as d1e d1ree
`residues have main chain conformations
`usually disallowed because of sceric hin­
`drance, the predicted structure is incompati·
`ble with d1e electron density. In retrospect,
`it appears diat die reduced volume of two
`
`Table 4. Fit of VL-VH framework. Comparisons
`of predicted and observed structures. [n Tables 4
`to 7, che model SA is based on structure analysis,
`and model CE on conformational energy calcula­
`tions starting from a structure derived initially
`from McPC603. 6. is the root-mean-square devi­
`ation in angstroms after superposition.
`
`6.
`(N, Ca, C, and
`0 atoms)
`SA
`CE
`
`1.0
`
`l.O
`
`Residues
`
`VL: 4-6, 19-25, 33-48,
`52-54, 61-76,
`84-90, 97-107
`V H 3-12, 17-25, 33-52,
`56-60, 68-82,
`88-95, 102-113
`
`Table 5. Shifts of framework residues adjacent to
`hypcrvariable regions. Shift is the difference ( ang­
`stroms) in position of residue after superposition
`of the framework.
`
`Residues
`
`VL: 25
`48
`90
`VH: 25
`52
`95
`
`33
`52
`97
`33
`56
`102
`
`Shift
`
`SA
`
`0.6
`0.6
`0.8
`0.9
`1.5
`1.2
`
`0.3
`0.8
`0.4
`1.8
`1.8
`1.0
`
`CE
`0.8
`0.6
`0.8
`1.2
`0.7
`0.6
`
`2.5
`0.2
`0.9
`0.6
`1.7
`1.3
`
`important side chains (29 and 34) may have
`changed the packing to produce a different
`fold.
`None of the residues at which the model
`differs from the observed structure is in­
`volved in antigen contacts. The main chain
`conformation of the binding site itself is
`predicted accurately (Fig. 2).
`Both the models based on conformational
`energy calculations correctly predicted the
`framework and the folds of L2, H2, and
`H3. Loop Ll is very different from the
`experimental structure because it was built
`from the extended part of McPC603 Ll
`rather d1an from the part common co od1er
`V. chains. L3 and Hl arc similar to those in
`the model based on structure analysis.
`The comparisons show that the predic­
`tion of the main chain conformation of
`Dl.3 was largely successful (Tables 4 to 7
`and Figs. l and 2). They support the prem­
`ise that d1e binding site conformation is
`determined principally by specific interac­
`tions of a few residues, and dlat these resi­
`dues can be identified and used to formulate
`rules valid for structure prediction. A struc­
`ture ofDl.3 at higher resolution will pennit
`a more detailed evaluation of d1e prcdk­
`tions, as well as d1c refinement and exten­
`sion of the rules.
`The predictions treated an isolated V L -
`V H dimer. The observed D 1.3 strucn1re is
`contained in a Fab-antigen complex. The
`dose similarity of the observed and predict­
`ed D 1.3 binding site and framework implies
`
`that the association with antigen docs not
`significantly alter the main chain conforma­
`tion of the antibody.
`
`REFERENCES AND NO TES
`I. A. G. Amir, R. A. Mariuz?.a, S. E. V. PhiJ.ijps, R.
`
`
`Poljak, Science 233, 747 (1986).
`2. C. Chothia and A. M. Lcsk, in preparation.
`3. P. de la Paz, B. ). Sutton, M. J. Darslcy, A. R.. Recs,
`EMBO J. 5, 415 (1985).
`4. M. uvitt,J.Mol. Biol. 170, 723 (1983).
`5. E. A. Kabat, T. T. Wu, H. Bilofsky, M. Rcid­
`Mib1cr, H. Perry, Eds., Sequences ofl'roteins oflmm11-
`
`1wh/gienl 1111erest (U.S. Public Health Service, Wash­
`ington, DC, 1983).
`6. F. Saul, L. Amzcl, R. Poljak,/. Biol. Chem. 253, 585
`(1978).
`7. i:>. Segal tt al., Proc. N111l. Acad. Sti. U.S.A. 71,
`4298 (1974).
`
`8. M. Marquart,). Dciscnhofcr, R. Huber, W. Palm,/.
`Mol. Biol. 141, 369 (1980).
`9. 0. Epp, E. Larham, M. Schiffer,
`R. Huber, W.
`Pa�n, Biochemi.!try 14, 4943 (1975).
`10. W. Furey, B. C. Wang, C. S. Yoo, M. Sax,f. Mol.
`
`Biol. 167, 661 (1983).
`I I. ). Novomj etal.,/. Biol; Cl1tm. 258, 1:1433 (1983).
`12. C. Chodua, ). Novotny, R.. Bruccoleri, M. Karplus,
`/. Mol. Biol. 186, 651 (1985).
`
`13. M. Vcrhocycn, unpubJjshcd.
`14. A. M. Lesk and C. Chothia,]. Mo/. Biol. 160, 325
`(1982).
`15. B. L. Sibanda and J.M. Thornton, Nature (Lo111i<J11)
`316, 170 (1985}; C. Chothia and A. M. Lesk,
`unpublished results.
`16. T. A. Jones, in Co111p 11rnrio11al
`D.
`Crys111l/09nrphy,
`
`
`Sayre, Ed. (Clarendon, Oxford, 1982), pp. 303-
`317.
`17. We thank Dr. M. Vcrhocycn for making the anlino
`
`
`acid sequence available; Ors. A. Feinstein, C. Mil­
`
`srein, and G. W inter for discussion; and the Royal
`Society, the U.S. National Science Foundation
`
`
`(PCM83-20171), the National Inscirure of General
`Medical Sciences (GM25435), and tJ1e European
`
`Molecular Biology Organization (ASTF 4475) for
`support.
`
`
`J 6 April 1986; accepted 30 J unc 1986
`
`Cambrian River Terraces and Ridgetops in Central
`Australia: Oldest Persisting Landforms?
`
`Table 6. Fies of hypcrvariablc loops.
`
`A. J. STEWART, D. H. BLAKE, c. D. OLLIER*
`
`cut in the Ashburton surface of the Davenport
`Pluvial sediments in paleovallcys
`province of central Australia form terrace remnants that appear to retain their original
`depositional tops and have probably existed as subaerial landforms since their
`Marine fossils in sediments conformable with the fiuvial sediments near the
`
`inception.
`
`southeast margin of the province give a Cambrian age for the terraces; the Ashburton
`
`
`surface forming the ridgetops between the paJeovalleys is Cambrian or older.
`
`I NLAND AusrRALlA IS WELL KNOWN
`for die flatness of its landscape and the
`antiquity of its erosion surfaces (1, 2).
`The highest surface in the Tennant Creek
`region of central Australia, the Ashburton
`surface, has long been thought to be Creta­
`ceous or older (2). We report evidence from
`the Davenport province of the Tennant
`Creek region (Fig. 1) that suggests diat,
`during d1C Cambrian, fluvial sediments were
`deposited in valleys between ridges whose
`cops are remnants of the Ashburcon surface.
`Subsequent dissection of d1c sediments
`
`fom1ed terraces and mesas, but where pre­
`served, the relation of the terraces to the
`adjoining ridges indicates that the terraces
`and mcsatops have existed as subaerial land­
`forms since d1e Cambrian sedimentation.
`The Ashburron surface itself is therefore
`Cambrian or older.
`The Davenport province (3} is a broad
`
`Bureau of Mineral Resources, Canberra, 2601, Aumali3.
`
`
`
`
`
`•Present address: Department of Geography, University
`
`of New England, Armid:tlc, 2350, Austr:tlia.
`SCIENCE, VOL. 233
`
`Residues
`
`Ll 26-32
`L2 49-51
`L3 91-96
`Hl 26-32
`H2 53-55
`H3 99-104
`
`6.
`(N, Ca, C, and Cl3 atoms)
`CE
`SA
`3.76
`0.85
`0.47
`0.63
`l.10
`0.97
`2.07
`l.68
`0.50
`0.89
`0.86
`0.87
`
`Table 7. Fir of maximal well-fitting portion,
`including framework and all lysozymc contact
`residues (mode.I SA).
`
`Residues
`
`6.
`(N, Ca, and C
`atoms)
`
`L: 2-6, 19-65, 69-92, )
`
`LO
`
`V .
`
`96-104
`v H: 3-7, 14-27, 30-59,
`
`67-75, 78-81,84-113
`
`758
`
`4 of 4
`
`BI Exhibit 1078
`
`