Structure and function of the C-terminal PABCdomain of human poly(A)-binding proteinGuennadi Kozlov*†, Jean-Francois Trempe*†, Kianoush Khaleghpour*‡, Avak Kahvejian*, Irena Ekiel§,and Kalle Gehring*†¶

*Department of Biochemistry, McGill University and †Montreal Joint Center for Structural Biology, 3655 Promenade Sir William Osler, Montreal, QC,Canada H3G 1Y6; and §Pharmaceutical Biotechnology and Montreal Joint Center for Structural Biology, Biotechnology Research Institute,National Research Council of Canada, 6100 Royalmount Avenue, Montreal, QC, Canada H4P 2R2

Edited by Roger D. Kornberg, Stanford University School of Medicine, Stanford, CA, and approved February 7, 2001 (received for review January 16, 2001)

We have determined the solution structure of the C-terminalquarter of human poly(A)-binding protein (hPABP). The proteinfragment contains a protein domain, PABC [for poly(A)-bindingprotein C-terminal domain], which is also found associated withthe HECT family of ubiquitin ligases. By using peptides derivedfrom PABP interacting protein (Paip) 1, Paip2, and eRF3, we showthat PABC functions as a peptide binding domain. We use chemicalshift perturbation analysis to identify the peptide binding site inPABC and the major elements involved in peptide recognition.From comparative sequence analysis of PABC-binding peptides, weformulate a preliminary PABC consensus sequence and identifyhuman ataxin-2, the protein responsible for type 2 spinocerebellarataxia (SCA2), as a potential PABC ligand.

Poly(A)-binding protein (PABP) is a ubiquitous and abundantcytosolic protein involved in mRNA translation. PABP binds

the poly(A) tail of mRNA and mediates mRNA circularizationthrough binding of the eIF4F translation initiation complex,which is associated with the mRNA 59 cap structure (1). Thiscircularization enhances translation by facilitating recycling ofribosomes (2). PABP acts as a general scaffolding protein thatrecruits various translation factors to the translating mRNA. Inaddition to the eIF4G subunit of eIF4F, protein partners ofPABP include PABP-interacting proteins Paip1 and Paip2, andeukaryotic release factor eRF3 (3–5). In plants and yeast,additional partners have been identified as follows: eIF4B,Pbp1p, and Rna15p (6–8).

Structurally, PABP is composed of two parts, an N-terminalportion that contains four RNA recognition motifs (RRM) anda C-terminal portion which contains a conserved poly(A)-binding protein C-terminal domain (PABC; CTC or PABP)sequence of '75 residues. Because of the essential function ofthe RRM domains, they have received more attention than thePABC domain. Early work in yeast established that at least oneRRM domain is required for binding mRNA and cell survival(9). Numerous functional and mutagenesis studies have alsoestablished the importance of the RRM domains in enhancingmRNA translation and stability. Recently, the structure of theRRM1 and RRM2 domains in a complex with poly(A) RNA wasdetermined by x-ray crystallography (10).

In contrast, the structure and function of the C-terminaldomain has remained enigmatic. Early work suggested that theC terminus is important in structuring the mRNA tail and for thecellular localization of PABP. This work was confirmed by morerecent studies that describe PABP dimerization through theC-terminal domain and a requirement of the C terminus forproper nuclear shuttling (11, 12). The C-terminal half of PABPcontains binding sites for eRF3 (5), Paip1 (13), Paip2 (4), Pbp1p(7), and a viral RNA polymerase (14).

Comparative sequence analysis of the C terminus of PABPreveals the phylogenetic conservation in the PABC domain (Fig.1). PABC is highly conserved in eukaryotes and is also foundassociated with a subset of HECT E3 ubiquitin-protein ligases(15). The significance of this association is unknown. In humans,

three different PABP proteins are known: the major formstudied here and two tissue-specific or inducible forms (16, 17).The sequence similarity between these isoforms is lower than forhomologous forms across species, suggesting conserved butdistinct functions for the different families of PABP proteins. Anunrelated nuclear poly(A)-binding protein, referred to as PABII, is also known in humans and yeast.

Here, we determine the solution structure of the last 139residues from human PABP (hPABP) and show that this se-quence contains a well-folded domain of '74 aa. This domain,PABC, functions to bind peptides from a number of proteinsknown to interact with the C terminus of PABP. Analysis ofchemical shift changes on peptide binding allowed us to map thepeptide binding site on PABC and determine the orientation ofthe peptide in the binding site. Finally, we used the set ofpeptides and proteins known to bind to PABC domains toformulate a preliminary PABC recognition sequence.

Materials and MethodsProtein and Peptide Preparation. Human PABP residues 498 to 636were expressed in Escherichia coli as a glutathione S-transferase(GST) fusion protein and purified by affinity chromatography.For NMR analysis, the GST-tag was removed by PreScissionprotease (Amersham Pharmacia) cleavage (leaving a five residueN-terminal extension) and the protein exchanged into NMRbuffer (50 mM KzHPO4y100 mM NaCly1 mM NaN3, pH 6.3) at303 K.

Human Paip2 was expressed as a GST-fusion protein andpurified by affinity chromatography. For NMR, the GST moietywas removed by thrombin cleavage (leaving a four residueN-terminal extension), and Paip2 was exchanged into NMRbuffer. NMR spectroscopy was carried out under the sameconditions as for hPABP (498 to 636). Peptides (Fig. 1B) weresynthesized by fluorenylmethoxycarbonyl (Fmoc) solid-phasepeptide synthesis and purified by reverse phase chromatographyon a C18 Vydac (Hesperia, CA) column. The composition and

This paper was submitted directly (Track II) to the PNAS office.

Abbreviations: PABC, poly(A)-binding protein C-terminal domain; hPABP, human poly(A)-binding protein; Paip, poly(A)-binding protein interacting protein; RRM, RNA recognitionmotif; NOESY, nuclear Overhauser effect spectroscopy; GST, glutathione S-transferase.

Data deposition: The atomic coordinates have been deposited in the Protein Data Bank,www.rcsb.org (PDB ID code 1G9L). The NMR assignments have been deposited in theBioMagResBank, www.bmrb.wisc.edu (accession no. 4915).

See commentary on page 4288.

‡Present address: Pharmaceutical Biotechnology, Biotechnology Research Institute, Na-tional Research Council of Canada, 6100 Royalmount Avenue, Montreal, QC, CanadaH4P 2R2.

¶To whom reprint requests should be addressed at the temporary address: Unite deBiochimie Structurale, Institut Pasteur, Centre National de la Recherche Scientifique, Unitede Recherche Associee 2185, 25 Rue du Dr. Roux, 75724 Paris Cedex 15, France. E-mail:kalle@bri.nrc.ca.

The publication costs of this article were defrayed in part by page charge payment. Thisarticle must therefore be hereby marked “advertisement” in accordance with 18 U.S.C.§1734 solely to indicate this fact.

www.pnas.orgycgiydoiy10.1073ypnas.071024998 PNAS u April 10, 2001 u vol. 98 u no. 8 u 4409–4413

BIO

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

21, 2

020

Fig. 1. Amino acid sequences of PABC domains and ligands. (A) Sequence alignment of PABC domains from human PABP (h, Homo sapiens), fly PABP (d,Drosophila melanogaster), wheat PABP (w, Triticum aestivum), yeast PABP (y, Saccharomyces cerevisiae), and the rat (Rattus norvegicus) 100-kDa HECT E3ubiquitin-protein ligase (100KD). The secondary structure and residue numbering are based on hPABP. Highly conserved residues are shown bold and underlined.(B) Sequence alignment of four PABC-binding peptides and the deduced consensus sequence. Peptides were derived from human Paip2, human Paip1, andhuman release factor RF3. Residues common to all four sequences are shown in red. (C) Putative PABC-sites in ataxin-2, an ataxin-2-related protein fromArabidopsis (MDC16.14), and two plant RRM-containing proteins (aF6N18.17, oRNA-BP). Numbers indicate the total number of amino acids in the protein andthe first amino acid shown in the alignment.

Fig. 2. Paip2 binding to PABC. (A) 15N–1H correlation (heteronuclear single quantum correlation) spectra of 15N-labeled Paip2 showing the small chemicaldispersion characteristic of an unfolded protein. (B) Spectra of 15N-labeled hPABP (498 to 636) in the absence (black) and presence (red) of unlabeled Paip2. (C)Spectra of 15N-labeled hPABP in the absence (black) and presence (blue) of a peptide corresponding to residues 106–127 of Paip2. (D) Magnitude of the chemicalshift changes (uDdu) in ppm of hPABP plotted by residue (red, intact Paip2; blue, Paip2 peptide; and black, difference between Paip2 and peptide). The dashedline indicates the cutoff for residues shown in Fig. 3D.

4410 u www.pnas.orgycgiydoiy10.1073ypnas.071024998 Kozlov et al.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

21, 2

020

purity of the peptides were verified by ion-spray quadrupolemass spectroscopy.

Structure Determination. NMR resonance assignments of freehPABP (498–636) were carried out by using standard triple-resonance techniques on a 13C,15N-labeled sample (18). Assign-ments of the hPABP–Paip2 complexes were based on 15N–1H-edited nuclear Overhauser effect spectroscopy (NOESY)spectra. The 15N–1H heteronuclear NOE was measured at 500MHz on 15N-labeled hPABP (498–636) (19). For the structuredetermination, a set of 1862 NOEs were collected from homo-nuclear and 15N and 13C isotope-edited NOESY spectra ofhPABP (498–636) acquired at 800 and 500 MHz. After deter-mination of the protein fold by using manual NOE assignments(20), automatic peak NOE assignments were made by using ARIA(21) and the structure refined by using standard protocols inCNS V. 0.9 (22). In the final 30 structures, PABC residues 1 to 74,the backbone pairwise heavy atom rms deviation was 0.48 Å, andthe all atom rms deviation was 1.16 Å. PROCHECK showed 65.4%

of residues in the most favored region of Ramachandran plot,30.4% in additionally allowed regions, and 3.6% in generouslyallowed regions (23). The coordinates have been deposited in theRCSB PDB under accession number 1G9L and the NMRassignments under BMRB accession number 4915.

Peptide Titrations. Titrations were carried out with either full-length recombinant Paip2 (residues 1–127) or a chemicallysynthesized peptide, Paip2 (residues 106–127), by using 15N–1Hheteronuclear single quantum correlation (HSQC) spectra at500 MHz, 303 K on a 1-mM sample of 15N-hPABP (498–636).The PABC residues showing the largest chemical shift changes[(D 1H shift)2 1 (D 15N shift 3 0.2)2]1/2 in PPM on binding offull-length Paip2 were K35 (0.9), V68 (0.63), M39 (0.59), G38(0.5), E19 (0.46), M16 (0.43), E64 (0.41), A65 (0.41), L40 (0.38),A67 (0.36), H72 (0.32), and T37 (0.3).

Consensus Peptide Analysis. Sequence searches of potential PABCligands were carried out by using the preliminary PABC binding-site consensus as query (24). Potential hits were screened based

Fig. 3. Structure of the PABC domain and peptide-binding site. (A) Ca trace of hPABP (residues 498 to 636) colored according to residue flexibility (blue, 15N–1Hheteronuclear NOE . 0.5; white, hNOE 5 0; red, hNOE , 20.5). (B) Ca trace colored according to phylogenetic conservation (magenta, .80% identity; white,'50%; red, ,20%) in a BLAST alignment of 40 unique PABC sequences (24). (C) Ca trace colored according the size of the amide chemical shift change (uDdu) onPaip2 binding (green, uDdu . 0.6; white, uDdu ' 0.35; red, uDdu , 0.1). Amide resonances for K35, V68, and M39 showed the largest changes. (D) Structure of thePABC domain. Thirty superimposed structures are shown with backbone rms deviation of 0.48 Å. Green balls represent amide groups whose chemical shiftschange by more than 0.3 ppm. Ball diameter is proportional to the chemical shift change. Sidechains of the hydrophobic core of PABC are represented in lightblue. (E) Secondary structure of PABC. The five helices are colored according to Fig. 1A. The conserved salt bridge between the sidechains of K35 and E42 is shown.(F) Molecular surface of PABC within 5 Å of residues with uDdu . 0.42. The deep hydrophobic cavity directly contacts the backbone amide of K35. Stacking by F22could stabilize the binding of an aromatic ring in the pocket. Figures were generated with MOLSCRIPT (27), GRASP (28), and RENDER (29).

Kozlov et al. PNAS u April 10, 2001 u vol. 98 u no. 8 u 4411

BIO

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

21, 2

020

on the sequence conservation of the PABC-site in relatedproteins, the accessibility of the PABC-site as judged from thepresence of surrounding low complexity sequences, and consid-erations of biological relatedness. A more complete list ofproteins identified is available in Table 1, which is published assupplemental data on the PNAS web site, www.pnas.org.

Results and DiscussionThe C-terminal quarter of human PABP (residues 498–636) wasprepared as an isotopically labeled, recombinant protein frag-ment and subjected to NMR spectroscopy. The protein gaveexcellent 15N–1H correlation spectra, with the dispersion ofresonances typical for a protein with a-helical secondary struc-ture and the possible presence of unfolded residues (Fig. 2B). Todetermine the extent of folded structure, NMR assignmentswere obtained, and the 15N–1H heteronuclear NOE was mea-sured (Fig. 3A). This heteronuclear NOE is a measure of thereorientation rate of the amide nitrogen-hydrogen internuclearvector and varies between 23.6 and 0.82 (at 500 MHz) forunfolded and folded residues (19). For hPABP residues 498 to541 and residues 623 and 636, the hNOE was negative, indicatingthe lack of a folded structure. Between residues 546 and 619, thehNOE was positive and generally above 0.6, indicating slowtumbling of a rigid, folded domain. This region corresponds tothe PABC domain, as identified by comparative sequence anal-ysis (Fig. 3B).

The solution structure of the 74-residue PABC domain wasdetermined by NMR spectroscopy at 500 and 800 MHz. Thestructure is almost three-quarters a-helical, with a well definedhydrophobic core and compact globular structure. PABC con-sists of five a-helices arranged as an arrowhead. Helix 1 consti-tutes the tip, helices 2 and 4 the sides of the arrow, helix 3 iscrossing, and the long C-terminal fifth helix constitutes the shaftof the arrow (Fig. 3E). Helix 2 is bent by the presence of a prolineresidue, P23 (residue 568 of hPABP), and is preceded by ana-helical-like loop (residues 13 to 19 in PABC). Helix 3 isdistorted by the presence of branched Ca amino acids I36, T37and of glycine G38. This destabilization is balanced by a saltbridge in helix 3 between the conserved amino acids K35 andE42.

Recently, a novel PABP interacting protein, Paip2, was iden-tified in a Far-Western screen of a human cDNA expressionlibrary (25). To characterize Paip2 structurally, we preparedrecombinant Paip2 from E. coli as a 15N-labeled GST-tag fusionprotein. The 15N–1H correlation spectrum of this 127-residueprotein showed the limited chemical shift dispersion character-istic of an unfolded protein (Fig. 2 A). Nonetheless, recombinantPaip2 is active in binding PABP in vitro assays (25), so weattempted to detect its binding by NMR spectroscopy. Onaddition of unlabeled Paip2 to a 15N-labeled sample of hPABP(398–636), approximately half of the 15N–1H correlation peaksshifted, indicating the formation of a Paip2–PABP complex (Fig.2A). Deletion mutagenesis of Paip2 and comparative sequenceanalysis had previously identified the 22 C-terminal residues ofPaip2 as responsible for binding to PABP (4). Accordingly, wechemically synthesized this peptide and carried out a secondtitration. The peptide caused the same spectral changes as intactPaip2 (Fig. 2 C and D). Similar results were obtained with ashorter Paip2 peptide and two peptides from other PABP-binding proteins (Fig. 1B). All four peptides bound in slow-exchange, suggesting ,10 mM binding affinity. These resultsestablish that PABC is a peptide binding domain.

Based on the Paip2 titration, we identified K35, V68, and M39of PABC as the residues displaying largest shifts on ligandbinding and likely to be close to the peptide binding site (Fig.3C). The molecular surface around the six residues most per-turbed by Paip2 binding reveals the putative peptide-binding site(Fig. 3F). This nearly continuous surface wraps around the

crosspoint between helices 3 and 5 and up toward helix 2. Themost striking feature of the protein surface is a deep hydropho-bic pocket formed between helices 2 and 3. This pocket isbounded by the amide of residue K35 and the sidechains of F22,I25, A33, and residues 34–38 of helix 3. The magnitude of thechemical shift change of K35 suggests the involvement ofaromatic ring current effects. These effects could result fromeither the insertion of a phenylalanine (F118) from Paip2 intothe hydrophobic pocket or an intramolecular rearrangementinvolving F22 of PABC. The presence of many residues withsmaller but significant chemical shift changes suggests that somestructural changes occur in much of PABC on Paip2 binding.

There is a strong correlation between the residues identifiedby the preceding chemical shift perturbation analysis and thosethat are the most conserved in the family of PABC domains.PABC contains several regions of near 100% sequence conser-vation. The longest stretch is KITGMLLE at position 35 to 42 inhelix 3. This stretch is preceded at position 17 by LGE-LFP inhelix 2 and followed at position 64 by the pair EA in helix 5. Inthe PABC structure, these residues are all in close proximity andappear to be involved in peptide binding. Comparison of fourPABC-binding peptides reveals conserved amino acid positionsover a span of 12 residues (Fig. 1B). Residues L3, N6, A7, andF10 appear to be most important for binding whereas positions2, 5, 8, and 10 are variable. This spacing of conserved andhydrophobic residues is consistent with peptide binding in ahelix-like conformation. Assuming that Paip2 follows the mo-lecular surface of Fig. 3F, we expect it to be oriented verticallywith PABC-binding site residue S1 near helix 5 and P12 nearhelix-like loop preceding helix 2.

By using the preliminary PABC-site consensus sequence as akey, the National Center for Biotechnology Information (NCBI)nonredundant sequence database was searched for potentialPABC ligands. In addition to revealing proteins known tointeract with PABP (Fig. 4), this screen identified ataxin-2, ahuman protein involved in a familial neurodegenerative diseaseand several ataxin-2-related proteins (Fig. 1C). Ataxin-2 and therelated proteins show strong similarity to Pbp1p, which suggests

Fig. 4. Model of the interactions identified and proposed for PABC. PABPconsists of an N-terminal section of four RRMs linked by a long unfoldedregion to PABC. Multiple PABP molecules (shown in gray) bind to the poly(A)tail via their RRM domains to form an RNA protein (RNP) complex. Cyclizationof the mRNA occurs through binding of eIF4F and the mRNA 59 cap to the RRMdomains of PABP. The PABC domain of PABP binds linear peptide sequencesto recruit protein factors to the mRNA RNP complex. Known binding partnersinclude Paip1, Paip2, and eRF3 (GSPT), which themselves act as linkers torecruit eIF4A (4A) and possibly other protein factors (?) to the mRNA (3, 25).The C terminus of PABP has also been reported to be involved in PABPdimerization (11), nuclear shuttling (12), mRNA stability (30), and polyade-nylation (ref. 7; dashed lines). Picornaviral protease 2 cleaves both eIF4F (notshown) and the linker region of PABP to shut off host cell protein synthesis (31,32). A potyviral RNA-dependent RNA polymerase has also been shown to bindPABC from cucumber PABP (14).

4412 u www.pnas.orgycgiydoiy10.1073ypnas.071024998 Kozlov et al.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

21, 2

020

that they may be involved in polyadenylation of mRNA (7).Ataxin-2 has also been found associated with the RRM-containing protein, A2BP1 (26). The screen also identified anumber of small putative RNA binding proteins.

ConclusionThe structural and functional studies described underline therole of PABP as a scaffolding protein that binds the mRNA 39poly(A) tail and a large number of translation factors (Fig. 4).Two types of PABP–protein interactions have been identified:those that occur via the the RRM domains and in close proximityto the mRNA poly(A) tail, and those that occur via the PABPC terminus. The large linker region between the RRM andPABC domains in PABP allows these second types of proteinassociations to occur farther from the mRNA strand and mayrelax steric or dynamic constraints in the structuringyassembly ofthe mRNA RNA protein (RNP) complex.

The identification of the C-terminal PABP peptide bindingdomain, presented here, should help clarify deletion mutagen-esis studies of PABP that have often led to ambiguous inter-pretations about the function and importance of the C terminus.Future structural studies will define more clearly the specificityof different PABC domains and provide a detailed picture of themechanism of peptide recognition.

We thank Nahum Sonenberg for help in the initial stages of this work andthe National High Field NMR Center (NANUC) facility for assistanceand access to their 800 MHz NMR spectrometer. K.G. is a Fonds de larecherche en sante du Quebec Chercheur-Boursier. J.-F.T. is a NaturalSciences and Engineering Research Council (Canada) PostgraduateScholarship-A Fellowship awardee. This work was funded by a CanadianInstitutes of Health Research grant to K.G. National Research Councilpublication 43000.

1. Tarun, S. Z., Jr., & Sachs, A. B. (1996) EMBO J. 15, 7168–7177.2. Jacobson, A. (1996) in Translational Control, eds. Hershey, J. W. B., Mathews,

M. B. & Sonenberg, N. (Cold Spring Harbor Lab. Press, Plainview, NY), pp.451–480.

3. Craig, A. W., Haghighat, A., Yu, A. T. & Sonenberg, N. (1998) Nature (London)392, 520–523.

4. Khaleghpour, K. (2000) Ph.D. thesis (McGill University, Montreal), p. 200.5. Hoshino, S., Imai, M., Kobayashi, T., Uchida, N. & Katada, T. (1999) J. Biol.

Chem. 274, 16677–16680.6. Le, H., Tanguay, R. L., Balasta, M. L., Wei, C. C., Browning, K. S., Metz, A. M.,

Goss, D. J. & Gallie, D. R. (1997) J. Biol. Chem. 272, 16247–16255.7. Mangus, D. A., Amrani, N. & Jacobson, A. (1998) Mol. Cell. Biol. 18,

7383–7396.8. Amrani, N., Minet, M., Le Gouar, M., Lacroute, F. & Wyers, F. (1997) Mol.

Cell. Biol. 17, 3694–3701.9. Sachs, A. B., Davis, R. W. & Kornberg, R. D. (1987) Mol. Cell. Biol. 7,

3268–3276.10. Deo, R. C., Bonanno, J. B., Sonenberg, N. & Burley, S. K. (1999) Cell 98,

835–845.11. Kuhn, U. & Pieler, T. (1996) J. Mol. Biol. 256, 20–30.12. Afonina, E., Stauber, R. & Pavlakis, G. N. (1998) J. Biol. Chem. 273,

13015–13021.13. Gray, N. K., Coller, J. M., Dickson, K. S. & Wickens, M. (2000) EMBO J. 19,

4723–4733.14. Wang, X., Ullah, Z. & Grumet, R. (2000) Virology 275, 433–443.15. Callaghan, M. J., Russell, A. J., Woollatt, E., Sutherland, G. R., Sutherland,

R. L. & Watts, C. K. (1998) Oncogene 17, 3479–3491.

16. Yang, H., Duckett, C. S. & Lindsten, T. (1995) Mol. Cell. Biol. 15, 6770–6776.17. Feral, C., Mattei, M. G., Pawlak, A. & Guellaen, G. (1999) Hum. Genet. 105,

347–353.18. Bax, A. & Grzesiek, S. (1993) Acc. Chem. Res. 26, 131–138.19. Peng, J. W. & Wagner, G. (1994) Methods Enzymol. 239, 563–596.20. Wuthrich, K. (1986) NMR of Proteins and Nucleic Acids (Wiley, New York).21. Nilges, M. (1997) Fold. Des. 2, S53–S57.22. Brunger, A. T., Adams, P. D., Clore, G. M., Gros, P., Grosse-Kuntsleve, R. W.,

Jiang, J.-S., Kuszewski, J., Nilges, M., Pannu, N. S. & Read, R. J. (1998) ActaCrystallogr. D Biol. Crystallogr. 54, 905–921.

23. Laskowski, R. A., MacArthur, M. W., Moss, D. S. & Thornton, J. M. (1993)J. Appl. Crystallogr. 26, 283–291.

24. Altschul, S. F., Madden, T. L., Schaffer, A. A., Zhang, J., Zhang, Z., Miller, W.& Lipman, D. J. (1997) Nucleic Acids Res. 25, 3389–3402.

25. Khaleghpour, K., Svitkin, Y. V., Craig, A. W., DeMaria, C. T., Deo, R. C.,Burley, S. K. & Sonenberg, N. (2001) Mol. Cell 7, 205–216.

26. Shibata, H., Huynh, D. P. & Pulst, S.-M. (2000) Hum. Mol. Genet. 9, 1303–1313.27. Kraulis, P. J. (1991) J. Appl. Crystallogr. 24, 946–950.28. Nicholls, A., Sharp, K. & Honig, B. (1991) Proteins Struct. Funct. Genet. 11,

281–296.29. Merritt, E. A. & Murphy, M. E. P. (1994) Acta Crystallogr. D Biol. Crystallogr.

50, 869–873.30. Wang, Z. & Kiledjian, M. (2000) Mol. Cell. Biol. 20, 6334–6341.31. Kerekatte, V., Keiper, B. D., Badorff, C., Cai, A., Knowlton, K. U. & Rhoads,

R. E. (1999) J. Virol. 73, 709–717.32. Joachims, M., Van Breugel, P. C. & Lloyd, R. E. (1999) J. Virol. 73, 718–727.

Kozlov et al. PNAS u April 10, 2001 u vol. 98 u no. 8 u 4413

BIO

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

21, 2

020