NMR Structure of the N-terminal J Domain of Murine Polyomavirus T Antigens

The NMR structure of the N-terminal, DnaJ-like domain of murine polyomavirus tumor antigens (PyJ) has been determined to high precision, with root mean square deviations to the mean structure of 0.38 Å for backbone atoms and 0.94 Å for all heavy atoms of ordered residues 5–41 and 50–69. PyJ possesses a three-helix fold, in which anti-parallel helices II and III are bridged by helix I, similar to the four-helix fold of the J domains of DnaJ and human DnaJ-1. PyJ differs significantly in the lengths of N terminus, helix I, and helix III. The universally conserved HPD motif appears to form a His-Pro C-cap of helix II. Helix I features a stabilizing Schellman C-cap that is probably conserved universally among J domains. On the helix II surface where positive charges of other J domains have been implicated in binding of hsp70s, PyJ contains glutamine residues. Nonetheless, chimeras that replace the J domain of DnaJ with PyJ function like wild-type DnaJ in promoting growth of Escherichia coli. This activity can be modulated by mutations of at least one of these glutamines. T antigen mutations reported to impair cellular transformation by the virus, presumably via interactions with PP2A, cluster in the hydrophobic folding core and at the extreme N terminus, remote from the HPD loop.

Numerous biological activities and biochemical properties critical for viral replication and cellular transformation have been genetically mapped to the N-terminal domain shared by polyomavirus tumor antigens, as reviewed (1,2). The N terminus of large T Ag 1 is necessary for cellular immortalization (3) and for overcoming control of cell cycle progression by retinoblastoma-type tumor suppressors (4). The N terminus of Py middle T Ag is required for transformation and for binding to PP2A and pp60 c-Src (5)(6)(7). The N terminus of small T Ag promotes cell cycle progression and binding to PP2A and subsequent induction of cyclin A (8 -10). Analogous experiments with SV40 T antigens confirm the importance of an intact N-terminal domain (1,(11)(12)(13). The N-terminal domain of SV40 large T Ag inactivates the growth-suppressive functions of pRb-related proteins p130 and p107 and decreases the level of p130 phosphorylation (2,14). The domain is needed for Py large T Ag to promote dephosphorylation of p130 at Thr-986 (15).
Recognition of homology between the N-terminal domains of T antigens and Escherichia coli DnaJ, including the invariant HPD motif, led to the hypothesis that T antigens possess the function of J domains, which help recruit chaperones to assist T antigens with their various functions (16,17). Confirming this hypothesis, hsp70 chaperones were found to be associated with the T antigens (4,12,18,19). The N-terminal domain of murine polyomavirus T antigens displays sequence identity to the N-terminal J domain of E. coli DnaJ and human HDJ-1 of only 18 and 19%, respectively (see data below). The N termini of SV40 and polyomavirus T antigens can nonetheless substitute for the E. coli J domain to form functional chimeric DnaJ proteins (20) (and data below). The J domain of the human DnaJ homologs, HSJ1 or DNAJ2, can substitute for the N terminus of SV40 T Ag to support viral replication (12). The N termini of the T antigens of polyomavirus and SV40 share the capacity of DnaJ-like chaperones to stimulate the ATPase activity of the hsp70 class of chaperones (21,22). Since the 79-residue N-terminal domain of polyomavirus T antigens, encoded by exon I, has properties consistent with a J domain, it is referred to as polyomavirus J domain (PyJ).
The invariant HPD motif of the J domains is essential to their stimulation of ATP hydrolysis and substrate binding by hsp70 chaperones (23). In fact, allele-specific suppressors of a mutation of the HPD motif indicate that the J domain of E. coli DnaJ makes contact with DnaK at a cleft within its ATPase domain (24). However, HPD mutations of Py middle T Ag have no effect on cellular transformation, whereas several other N-terminal mutations do (6,7). Chimeras of SV40 large T Ag with N terminus replaced by yeast or E. coli J domains fail to support viral DNA replication, transformation, and liberation of E2F from p130, whereas a chimera using the J domain of JC virus is functional (25). These observations suggest that virusspecific activities, independent of the HPD motif, reside within the N-terminal J domain of T antigens. Where might these virus-specific functions map in the structure of PyJ relative to regions critical for recognition and stimulation of hsp70s? Locating sites of lesions in the PyJ structure that impair functions unique to T antigens, such as cell transformation, can suggest features in the tertiary structure that support these viral functions. Such structural features of PyJ may help account for T Ag recognition and inhibition of PP2A. Comparison of the high precision NMR structure of PyJ, divergent in sequence, with the NMR structures of the J domains of E. coli DnaJ and human HDJ-1 (26 -28) also suggests a minimum of structural elements needed for J domain interactions with hsp70s.

EXPERIMENTAL PROCEDURES
J Domain Complementation Assay-We employed the in vivo complementation assay developed by Kelley and Georgopoulos (20), who generously provided the reporter strain WKG190 and the E. coli DnaJ (residues 1-376) expression plasmid pWKG90 modified to H71T. The E. coli J domain (residues 1-70) was replaced with an NcoI/KpnI PCR fragment encoding the PyJ (residues 1-79) using ctagcaccatggatagagttctgagc and ccggtaccgttcctcgctccgccgtt as primers. The reverse primer added to the C-terminal end six residues unique to tiny T Ag, i.e. NGGARN, for the full 85 residues of tiny T Ag (21). The point mutants V4D, L13V, Q32E, and Q32K were made with GeneEditor (Promega). Expression of J chimeras was induced with 33 mM arabinose. The permissive temperature used was 28°C, and the restrictive temperatures used were 14 and 42°C.
Alteration of Codon Usage to Optimize Expression in E. coli-A synthetic PyJ gene sequence, i.e. for residues 1-79, was designed using codons frequently used in highly expressed E. coli genes (29). Sequences flanking the coding region of exon I of PyJ were added to fuse a six-histidine tag to the C terminus and NcoI or BamHI restriction sites with CG tails to assist subcloning. The resulting sequence is as follows: 1 gcggatccat ggaccgtgtt ctgtcccgtg ctgacaaaga acgtctgctg gaactgctga aactgccgcg tcagctgtgg ggtgacttcg gtcgtatgca gcaggcttac aaacagcagt ccctgctgct gcacccggac aaaggtggtt cccacgctct gatgcaggaa ctgaactccc tgtggggtac cttcaaaacc gaagtttaca acctgcgtat gaacctgggt ggtaccggtt tccagcacca ccaccaccac cactgaggat ccgc 274. This synthetic gene coding for PyJ was constructed in a one-step PCR using four template/primer oligonucleotides at 20 nM and two standard primers to drive the reaction (30). Oligonucleotides were chosen that are composed of approximately onefourth of the coding sequence, with overlaps of 18 -19 nucleotides to allow a stringent annealing temperature of 55°C during PCR. The primers are composed of (i) nucleotides 1-77, (ii) reverse complement of nucleotides 60 -146, (iii) nucleotides 129 -216, (iv) reverse complement of nucleotides 198 -274, (v) nucleotides 1-19, and (vi) the reverse complement of nucleotides 256 -274. The conditions used for the high fidelity PCR with using Taq polymerase (PerkinElmer Life Sciences) were low MgCl 2 concentration of ϳ1.5 mM, short cycling steps, i.e. 94°C for 30 s, 55°C for 30 s, and 72°C for 60 s, and a relatively small number of 20 -25 cycles. The PCR fragment was subsequently cloned into NcoI/ BamHI sites of pET15b (Novagen) and sequenced to confirm accuracy.
Expression-PyJ was expressed in E. coli BL21 (DE3) host using the T7 expression system (31). We used the following protocol to reduce the plasmid instability known to be a problem of this expression system. Prior to the day of overproduction, 10 l of frozen host bacteria were diluted into 50 ml of M9 medium, supplemented as described below, containing 200 g/ml fresh ampicillin. Cultures were incubated overnight at room temperature without shaking to avoid reaching saturation and maintain selective pressure. The next day the E. coli were pelleted, and the supernatant medium containing ␤-lactamase was discarded. The bacteria were resuspended and inoculated into 1 liter of fresh, supplemented M9 medium. The E. coli cultures were grown to an A 600 of ϳ0.25 where an additional 400 g/ml ampicillin was added. Cells were then grown to an A 600 of ϳ0.9 and induced with 0.2 mM isopropyl-1-thio-␤-D-galactopyranoside for about 1.5 h. Bacteria were harvested by centrifugation and frozen at Ϫ70°C until further processed.
Purification-Bacterial pellets were resuspended in 40 ml of lysis buffer (50 mM NaPO 4 , pH 7.2, 100 mM NaCl) with 2 mM imidazole and 1 mM phenylmethylsulfonyl fluoride and were disrupted using a French press operating at 16,000 pounds/square inch. The extract was partially cleared by centrifugation twice at 10,000 ϫ g for 10 min followed by 30,000 ϫ g for 2 h. The His-tagged PyJ domain was then affinity purified from the extract by adding 2 ml of Ni-NTA-agarose (Qiagen) equilibrated with lysis buffer to the supernatant and gently mixing overnight at 4°C. This prolonged mixing, near the binding capacity of the affinity resin, improved the yield and purity of PyJ. The agarose was placed in a disposable column (Pierce), rinsed with 20 volumes lysis buffer, washed with 20 volumes of lysis buffer, pH 6.5, containing 50 mM imidazole, and eluted in lysis buffer, pH 6.5, with 500 mM imidazole. Purity of PyJ was verified using Tricine gels (21). Once dialyzed and concentrated, PyJ was sufficiently pure for analysis by NMR but was susceptible to gradual proteolysis after a month. To improve long term stability (Ͼ6 months), fractions containing PyJ were subjected to gel filtration using a 200-ml S220HR Sephacryl column (Amersham Pharmacia Biotech) or HiLoad 16/60 Superdex 75 (Amersham Pharmacia Biotech) run with 10 mM phosphate buffer (pH) and 25 mM NaCl. Fractions containing PyJ were thoroughly lyophilized, redissolved by adding water (e.g. 5 ml), dialyzed three times against 10 mM phosphate, pH 6, 100 mM KCl, and concentrated to the desired volume (e.g. 2.0 ml) with Centricon YM3 (Amicon). The PyJ solution was adjusted to pH 6 and dialyzed less than 1 h against 10 mM phosphate, pH 6.0, 10 mM KCl, and either 7 or 99% (v/v) D 2 O. To forestall protein precipitation, glycine-d 5 (Cambridge Isotope Laboratories) was then added to 100 mM. PyJ was further concentrated with a Centricon YM3 (Amicon) to 0.7 ml for NMR analysis. Addition of the protease inhibitor mixture Complete (Roche Molecular Biochemicals) was critical to prevent gradual proteolysis.
NMR Structural Restraints-PyJ structure calculations were based on NOE-derived distance restraints, , , and 1 torsion angle restraints, and 13 C␣ and 13 C␤ chemical shifts. The 846 NOE restraints were obtained from the 15 N-and 13 C-separated NOESY spectra mentioned above, as well as from a novel three-dimensional NOESY-CT-HSQC ( m ϭ 200 ms) spectrum of 3.9 mM 13 C/ 15 N PyJ in 93% H 2 O. The spectrum has high 13 C resolution by way of the 13 C constant time period of the sequence. Its excellent water suppression was achieved using gradient-based coherence selection. Its longer NOESY mixing time of 200 ms was chosen for regions of slower NOE buildup, due to shorter internal correlation times, such as the HPD loop and the nearby ends of helices II and III. Demonstration of this experiment will be described elsewhere. NOE intensities were classified as weak, medium, and strong, corresponding to upper bounds of 3.0, 4.0, and 5.0 Å, respectively. The sum of the van der Waals radii of two protons was used as the lower bound of each NOE restraint.
Torsion Angle and Chemical Shift Restraints-44 angle restraints were obtained from 3 J HNHa values measured from a three-dimensional HNCA-J spectrum (45). These angles were restrained to Ϫ65 Ϯ 25°f or 3 J HNHa Ͻ5 Hz and to Ϫ65 Ϯ 30°for 5 Hz Ͻ 3 J HNHa Ͻ5.5 Hz. No 3 J HNHa values higher than 5.5 Hz were observed. Restraints on 58 torsion angles and an additional 29 torsion angles were derived from TALOS interpretation of 13 C␣, 13 C␤, 13 CЈ, 15 N, 1 HN chemical shifts of PyJ, using a data base of chemical shifts and structural motifs from a set of proteins characterized at high resolution both by x-ray and NMR methods (46). The torsion angle restraints from TALOS were used with upper and lower bounds of at least 2.5 standard deviations when predictions were clear. The 13 C␣ and 13 C␤ secondary chemical shifts for all 79 native residues of PyJ were used in refining the structure because of their relationship to and torsion angles (47).
Hydrogen Bonding-Determination of the secondary structure of PyJ enabled identification of hydrogen bonding partners in regions of regular ␣-helix. Amide groups were considered to be donors of hydrogen bonds if their signals were observed in 15  Structure Calculations and Analysis-Structure calculations in CNS 1.0 (50) employed torsion angle dynamics (TAD) to generate each starting structure (51) followed by conventional simulated annealing (SA) in Cartesian space to refine each structure (52). TAD at high temperature excels in exploration of conformational space and in boosting convergence rate (51). Distances to methyl protons, aromatic ring protons, and non-stereospecifically assigned methylene protons were computed using the r Ϫ6 summation method (53). The default force constants of the TAD/SA protocol implemented in CNS were used. In order to improve convergence rate, the numbers of steps in each of the three stages of molecular dynamics of the TAD/SA protocol were increased as follows: high temperature (T ϭ 50,000 K) TAD used 4000 steps; the TAD cooling stage used 2000 steps; and the SA cooling stage used 6000 steps of Cartesian dynamics. Of 100 structures computed, 47 were accepted that meet the following criteria: no NOE violations larger than 0.5 Å, no dihedral angle violations larger than 5°, total energies less than 205 kcal/mol, and van der Waals energies (using quartic repulsive potential) less than 60 kcal/mol. The quality of the structures was analyzed with PROCHECK-NMR (54). Electrostatic surface properties of PyJ and other J domains were investigated with GRASP (55). r.m.s.d., secondary structure, surface exposure, hydrogen bonding, and other parameters of protein structure were measured and illustrated using MOL-MOL (56). Atomic coordinates and NMR structural restraints were deposited at the Rutgers Protein Data Bank under accession code 1faf.

RESULTS
PyJ Substitutes for E. coli J Domain Function-Chimeras of the E. coli DnaJ chaperone were constructed by substituting the wild-type PyJ domain, i.e. all 85 residues of tiny T Ag or various mutant PyJ domains for the E. coli J domain (residues 1-70). The ability of the chimeras to promote viability of a hypersensitive E. coli strain with DnaJ disrupted was tested at extremes of high and low growth temperature (20). The chimera with wild-type PyJ domain can restore growth at the restrictive temperatures of 42 and 14°C equally well as the native E. coli J domain (Table I). A chimera substituting residues 1-79 shared by all Py T antigens gave similar results. The chimera with the V4D mutant PyJ domain supports modestly reduced growth of the E. coli at low temperature and dramatically reduced growth at high temperature. The Q32K PyJ mutant chimera, substituting the consensus positive charge, confers growth equivalent to the native Dna J control and wild-type PyJ chimera (Table I). The Q32E mutant chimera, substituting charge opposite the consensus, dramatically reduces growth at both high and low temperature. The L13V PyJ mutant chimera is almost as deficient in restoring growth as the negative control lacking any source of DnaJ. Similar amounts of DnaJ chimeras are expressed in all cases except L13V, which expressed less than 10% of wild-type level detected by Western blotting (data not shown).
Expression and Purification of PyJ-Initial attempts to over-produce polyomavirus J domain (PyJ) as a glutathione S-transferase fusion protein produced at best only 1.0 mg of PyJ per liter of culture. Expression of eukaryotic gene products in E. coli often can be impeded by mismatch of codon usage with the pools of tRNA available in E. coli. For example, the aga codon for arginine, corresponding to a low abundance tRNA in E. coli, is used by polyomavirus DNA and is inefficiently translated (57)(58)(59)(60). Therefore, we constructed a synthetic gene coding for PyJ to optimize codon usage. We also fused a C-terminal sequence of six histidines to expedite purification with NTAnickel affinity chromatography. A 10-kDa protein was expressed after induction (Fig. 1, lane 1) and was found primarily in the soluble fraction (lane 3). The second centrifugation (lane 4) reduced the amount of broken cell debris before loading onto the affinity column. NTA-nickel affinity chromatography efficiently removed the His 6 -tagged PyJ from the extract (Fig. 1, lane 5). PyJ eluted with greater than 95% purity (lanes 7 and 8). The typical yield was 5-10 mg of protein per liter of M9 medium supplemented with 10% (v/v) Celtone, an improvement of 5-10-fold over the expression of the native gene. NMR Peak Assignments of PyJ-The backbone 1 H, 15 N, and 13 C␣ resonances were assigned for all residues except the four C-terminal residues of the six-histidine tag and the amides of Asp-2 and His-49. Side chain proton and carbon chemical assignments are complete for all residues except the six-histidine tag and aromatic groups of His-49 and Phe-78. Stereospecific assignments of H␤ protons were obtained for 30 residues. Stereo-specific assignments of all leucine and valine methyl groups were obtained, except for Leu-16 and Leu-40. The 1 H, 15 N, and 13 C chemical shifts for PyJ are available at the Bi-oMagResBank under accession number 4403. Analysis of sequential and medium-range NOEs patterns (61), 3 J HNHa (45,62) and 13 C␣ chemical shifts (63) revealed three helical regions in PyJ as follows: residues 7-16, 27-41, and 49 -70.
Initially, 557 NOE restraints were obtained from the conventional 15 N-and 13 C-separated NOESY spectra. Further NOE analysis was complicated by severe spectral overlap of the peaks of leucine methyl groups. To overcome this overlap, a three-dimensional NOESY-CT-HSQC experiment with high 13 C resolution was implemented on 3.9 mM PyJ in 93% H 2 O buffer. This high resolution spectrum provided 289 additional NOE assignments, bringing the total number of NOEs to 846. Structure calculations employed a total of 1183 experimentally derived restraints (Table II). Most inter-residue NOEs were found from residues 3 to 73. The lack of NOEs at the N terminus and the C-terminal region (residues 74 -79, plus His tag) can be explained by the high mobility of these regions evident from 15 N spin relaxation (data not shown). Quality of Structure of PyJ-Structure calculations based on an average of 10.6 inter-residue NOEs per ordered residue of PyJ, including an average of three long-range NOEs per residue, yielded a high precision structure of excellent stereochemical quality (Table II). The ensemble of 47 conformers (Fig. 2) has r.m.s.d. to the mean structure of 0.38 Ϯ 0.10 Å for backbone atoms (N, C␣, and CЈ) and 0.94 Ϯ 0.09 Å for the heavy atoms of ordered residues 5-41 and 50 -69 (Table II). The largest violations of any distance restraint and any dihedral angle re-straint are 0.24 Å and 3.78°, respectively. The maximum violations in each structure average 0.21 Ϯ 0.01 Å for distance restraints and 2.73 Ϯ 0.44°for dihedral angle restraints. 87.8% of the residues from the ensemble of PyJ structures have and dihedral angles in most favored regions of the Ramachandran plot. The other 12.2% of the residues occupy additionally allowed regions. The high quality of the Ramachandran plot appears to be a result of the number and quality of distance restraints and backbone torsion angle restraints coupled with the high helical content; no conformational data base potential was used.
Locations of Helices and Their Caps-PyJ is a highly ␣-helical protein, dominated by a hairpin of anti-parallel helices II and III (Fig. 2). Helices II and III are bridged by helix I. The loop connecting helices II and III appears to contain a turn of 3-10 helix from conserved Pro-43 through Lys-45, in 61% of the structures of the ensemble. This suggests a transient sampling of the 3-10 helical conformation in this conserved loop enriched in fast internal motions (data not shown).
Helix II, lacking N-cap, runs from Phe-27 to Leu-41 ( Fig. 2A). Helix II appears to be terminated by a variant of the proline C-cap with His-42 being the residue in the Ccap position. The pattern of dihedral angles of conserved His-42 and Pro-43 of PyJ are present in the His-Pro C-capped helices of x-ray structures with accession codes 1smd and 2ohx (67). The histidine C-cap residues in these two structures have angles of Ϫ71 to Ϫ78°, angle of 139°, and 1 angles of Ϫ163 to Ϫ174°that are virtually the same as observed for PyJ His-42 that has angle of Ϫ80.5 Ϯ 9.5°, angle of 137 Ϯ 6°, and 1 of Ϫ160 Ϯ 6°. The proline CЈ residues of the structures 1smd and 2ohx have angles of Ϫ21 to Ϫ29°, equivalent to Pro-43 of PyJ that has angle of Ϫ21 Ϯ 12°. The His-42 and Pro-43 side chains appear to stack on each other (Fig. 4) in a fashion similar to the His-Pro motifs of crystal structures 1smd and 2ohx. Conserved Asp-44 has angle of Ϫ81 Ϯ 9°and angle of Ϫ20 Ϯ 5°, well within the ranges observed at the CЉ position of proline C-caps (65). Several close contacts between the Leu-39 (C3 position) and conserved Lys-45 (Cٞ) are consistent with the presence of a C-cap. A hydrogen bond between His-42 CϭO (Ccap) and Lys-45 HN (Cٞ) is another characteristic of the proline C-cap (Fig. 4) (65).
Helix III may also have an N-capping box (Fig. 2A). The / dihedral angles of Gly-47 and Ser-48 fall in the reported ranges  for NЈ and N-cap residues (65). The 13 C␣/ 13 C␤-chemical shifts of Ser-48 (57.0/64.5) are also consistent with Ser-48 being an N-cap residue (66). A medium intensity NOE observed between H␤1 of Ser-48 and HN of Leu-51 suggests a potential N-capstabilizing hydrogen bond between the amide of Leu-51 and the hydroxyl of the Ser-48 side chain. Helix III then extends from His-49 to about Arg-70 ( Fig. 2A). The end of helix III is not well defined and continues to Asn-72 in 19% of PyJ structures, suggestive of fraying of the C-terminal end of helix III in solution. The C-terminal region of PyJ from Gly-74 to Gln-79 is highly disordered, judging from its sharp NMR lines and absence of inter-residue NOEs. Much of this very mobile C-terminal linker region is omitted from the figures.
Stabilizing Elements of PyJ Structure-Tight hydrophobic core packing, buried hydrogen bonds, and electrostatic interactions are considered the most important elements that define and stabilize a protein structure. Participation of residues in the hydrophobic core of PyJ was investigated by measuring surface exposure to solvent, long range NOE contacts, and the number of atoms that lie within 3 Å of each side chain two or more apart in sequence. The hydrophobic core has 16 residues with surface exposure less than 10% of the surface area of the residue as follows: Leu-13, Leu-14, Leu-17, Leu-19, Met-30, Ala-33, Tyr-34, Gln-37, Ser-38, Leu-41, Leu-55, Leu-58, Trp-59, Thr-61, Phe-62, and Val-66 (Fig. 2B). Patterns of long-range NOEs and the number of neighboring atoms suggested that the following partially exposed residues could also stabilize the packing of PyJ as follows: Leu-5, Asp-9, Lys-10, Leu-16, Leu-23, Trp-24, Leu-51, Met-52, and Lys-63. The side chain of Leu-41 at the C-terminal end of helix II makes hydrophobic contacts with Leu-51 and Met-52 of helix III which may stabilize these respective ends of helices II and III, adjacent to the HPD loop.
From H-D exchange experiments, hydrogen bonds were initially identified for amide groups of the following residues: Glu-11, Leu-14, Leu-16, Lys-18, Leu-19, Gln-31, Gln-32, Ala-FIG. 2. Ensemble of 47 conformers of PyJ. A, this "front" view has helix I in the foreground. The ensemble was colored by secondary structure as follows: blue for ␣-helices, red for 3-10 helix, green for Ncap residues of helices, orange for C-cap residues of helices, and black for loops and termini. The residues at the ends of helices are identified. N and C termini and helix numbers are labeled. b, this "back" view colors sites of T Ag mutations and plots the most buried and well defined side chains. Sites of mutations of polyomavirus (Py) middle T Ag which impair transformation are colored red. Sites of mutations of Py and SV40 T antigens that impair hsc70-dependent functions, such as DNA replication and inactivation of pRb family members, are highlighted in green. The buried side chains of other residues are shown in black. 33, Tyr-34, Glu-36, Leu-55, Asn-56, Trp-59, and Gly-60. Except for Lys-18 and Leu-19, these are part of ␣-helices and were restrained with distance restraints representing their ␣-helical hydrogen bonds in structure calculations. The conserved residues Leu-13, Leu-17, Met-30, and Phe-62 may also be protected from hydrogen exchange, although amide spectral overlap renders this ambiguous. Distances of 2.5 Ϯ 0.5 Å between an acceptor and an amide proton donor were used to identify 43 potential hydrogen bonds in the ensemble of structures. Networks of hydrogen bonds between residues four apart in sequence were found for backbone amides of residues 10 -16, 30 -37, 52-70, and 72. Some of these bonds, especially in the second half of helix III, are predicted to be weak and were not detected in amide H-D exchange experiments. The N-cap of helix I may be stabilized in part by hydrogen bonds between the NH of Ala-8 and side chain oxygen of Ser-6 and between the Ser-6 carbonyl and the Lys-10 backbone amide. Distances suggest that backbone amides of loop residues Trp-24, Asp-26, Leu-41, His-42, and Lys-45 may be hydrogen-bonded to i Ϫ 3 carbonyl acceptors.
Potential electrostatic interactions within PyJ were consid-ered for positive and negative charges approaching closer than 5 Å. This was complicated by high r.m.s.d. values of surfaceexposed charged side chains. The following putative electrostatic interactions may enhance the stability of PyJ. The charged atoms of Glu-11 and Arg-21 are less than 5.0 Å apart in 70% of accepted PyJ models, suggesting that they may form a salt bridge. 51% of the structural models have side chain atoms of Arg-12 and Glu-54 closer than 5 Å, suggesting another putative salt bridge. Close contacts between Asp-26 and Arg-29 ( Fig. 5A) may provide electrostatic stabilization of the beginning of helix II. Surface Properties of PyJ-Mutagenesis of PyJ residue Gln-32 is consistent with the importance of helix II in hsp70 recognition. A view of the surface of the helix II side of PyJ is colored to represent its electrostatic and hydrophobic features (Fig. 5A). The negative charge of Asp-26 immediately precedes helix II (at top of Fig. 5A). The only positive charges on helix II of PyJ are Arg-29 and Lys-35. Lys-45 immediately follows helix II and the HPD motif (bottom of Fig. 5A). The partly exposed side chains of Leu-39, Leu-40, and Leu-51 form a hydrophobic "belt" just above HPD loop. The hydrophobic residues Val-4, Leu-73, and Phe-78 are somewhat exposed at the disordered N and C termini of PyJ (not shown). The helix III side of PyJ carries a net negative charge among Asp-9, Asp-54, Lys-63, and Glu-65 (not shown). DISCUSSION The results of the in vivo complementation assay (Table I) confirm our previous findings that the N-terminal domain of polyomavirus T antigens (PyJ) can act as a fully functional J domain (21). The low sequence identity with other J domains and small number of positive charges on helix II for recognizing hsc70 raises the question of how the structure of PyJ supports its vigorous J domain function. The high precision NMR structure of PyJ suggests a minimal set of structural elements necessary for J domain function. The PyJ structure also suggests the structural origins of phenotypes for several mutations  (26,28). Residues similar or identical among these four sequences are marked with white lettering on black background. Residues similar or identical among three sequences are marked with white lettering on dark gray background. Residues similar or identical between two sequences are marked with a light gray background. Where two pairs of similar residues are present at one site, a very light gray background is also used to distinguish the pairs. Dots indicate gaps in the alignment. PyJ sequence numbering is given below the PyJ sequence. Nomenclatures for capping (NЈ-N3 and C4-CЈЈЈЈ) and for coiled-coil heptad repeats (a-g) occupy lines below. The figure was prepared using ALSCRIPT (76).  (Table III). The fourth helix of E. coli and human J domains that is absent from PyJ was omitted from the comparison. The residues used for comparison (Table III footnote a) are a modification of those suggested by the Dali server (68) and include residues in and near ␣-helices I-III of the J domains. The structural and functional similarities of PyJ to HDJ-1 and DnaJ suggest the PyJ tertiary structure can be considered a member of J domain family, albeit with features unique to the mammalian viral T antigens. PyJ is most similar to the J domain of HDJ-1. The fold of PyJ is more similar to active DnaJ (residues 1-104) than to inactive DnaJ (residues 1-78).
The dominant feature of all reported structures of J domains including PyJ is a hairpin of two anti-parallel helices II and III bridged by helix I. The angle between helices II and III in the mean structure of PyJ is Ϫ167°, similar to the angles reported to be Ϫ164°in inactive DnaJ and Ϫ169°in active DnaJ (28). Contacts between helices II and III are organized in a coiledcoil manner (69) as reported earlier for the J domain of DnaJ (26). The residue labeling of coiled-coil heptad repeat positions a through g (69) can be applied to helices II and III of PyJ, where positions a and d pack in the interior between helices II and III. Helix II residues can be labeled with Phe-27 as a1, Met-30 as d1, Tyr-34 as a2, Gln-37 as d2, and Leu-41 as a3 (Fig. 3). Helix III residues can be labeled with Met-52 as a1Ј, Leu-55 as d1Ј, Trp-59 as a2Ј, Phe-62 as d2Ј, Val-66 as a3Ј, and Leu-69 as d3Ј (Fig. 3). These buried residues form an extensive network of contacts and are quite similar to residues of other J domains in these positions (Fig. 2B). Hydrophobic residues from helix I and the I-II loop are an integral part of the hydrophobic core as well. This hydrophobic packing may confer the rigid order of the I-II loop. The PyJ residues protected from hydrogen-deuterium exchange in helix I, the C-cap of helix I, the first two-thirds of helix II, and the third and fourth turns of helix III (see "Results") coincide in large part with the hydrophobic core residues. Because the slow hydrogen exchange core marks the protein folding core (70,71), these protected residues of PyJ are very likely to contribute key tertiary folding interactions, possibly early in the folding process. The equivalent residues of the J domain of E. coli DnaJ are protected from hydrogen exchange as well as other residues throughout helix II (26). The similarity of hydrogen exchange protection would suggest that a similar tertiary folding mechanism may prevail for diverse J domains.
Like PyJ, the J domain of E. coli DnaJ (26) also features N-capping of its helices I and III. The PyJ also shares with the J domains of DnaJ and HDJ-1 a Schellman C-cap of helix I, unrecognized in the other J domains until now. The dihedral angles and stabilization mechanism of the C-cap in PyJ are consistent with a non-Gly Schellman motif (see "Results"). The other J domains substitute Gly in the CЈ position of the cap and exhibit the torsion angles of a regular Schellman C-cap, quite similar to the non-Gly Schellman motif of PyJ. The structures of the J domains of HDJ-1 and DnaJ (Protein Data Bank accession codes 1hdj, 1bq0, and 1bqz) possess the C3 to CЉ hydrophobic contacts, C3 carbonyl to CЉ amide hydrogen bond, and C2 carbonyl to CЈ amide hydrogen bond, all diagnostic of Schellman C-capping motifs. These Schellman motifs orient the hydrophobic C3 and CЉ residues into the hydrophobic core (65) of the J domains and may stabilize the tertiary structure of all J domains. The C-cap coincides with residues whose hydrogen exchange protection and burial suggest the importance of the region in tertiary folding.
Comparison of the Helix II Surface of PyJ, HDJ-1, and DnaJ-Electrostatic attraction of the positively charged surface of helix II of J domains with a negatively charged surface of hsp70 was implicated in interaction between hsp40 and hsp70 family members (72). Among the residues of the E. coli J domain affected in NMR spectra by interaction with DnaK (72) are positively charged residues at e1, f1, b2, c2, and g2 coiledcoil positions of helix II (Fig. 3). Positively charged residues are also present in these positions of helix II of HDJ-1 (Figs. 3 and  5). Intriguingly, PyJ differs from the other J domains in that it lacks arginine and lysine residues at all these positions except for b2, which is Lys-35 in PyJ. Consistent with this positive charge conservation at b2, the equivalent DnaJ b2 residue Lys-26 (Fig. 3) is the residue of DnaJ with amide NMR peak that experienced the greatest shift when titrated with DnaK constructs. The conservation and NMR titration data suggest the positively charged residue in the b2 position could be the To assess further any common patterns of charge and hydrophobicity on or near the helix II surface, NMR structures of J domains of human and E. coli hsp40s were superimposed with PyJ (Table III and (72). Analysis of J domain sequences (16) reveals interchangeability or co-variation of positively charged and hydrophobic residues at the b2 and f2 positions, evident when comparing the SV40 and Py sequences (Fig. 3). Substitution of a positively charged b2 residue by a hydrophobic one is almost always accompanied by substitution of f2 hydrophobic residue to positively charged residue, particularly among the polyomaviruses. The proximity of the b2 and f2 side chains is evident from the Leu-39 side chain of PyJ lying within 7 Å of the side chain of Lys-35. The sequence conservation and NMR data suggest a positive charge at either the b2 or f2 position of helix II may be required for J domain interactions with hsp70.
Notably, the positive charges at the e1, f1, c2, and g2 positions of helix II of most known J domains are absent from PyJ ( Fig. 3). At these sites, PyJ substitutes glutamine residues that superimpose with the corresponding Arg and Lys residues of the J domains of DnaJ and HDJ-1 (Fig. 5C). If these positive charges were essential for J domain function, PyJ would be unable to interact with and stimulate hsp70s. However, the results of the E. coli complementation assay (see above) clearly demonstrate that PyJ is quite competent in replacing the J domain of DnaJ in its interactions with the E. coli hsp70, i.e. DnaK (Table I). The importance of positive charge at the f1 in the PyJ-DnaJ chimera was investigated further. Substitution of native Q32 (f1) with glutamate, known in large T Ag to disrupt Rb-dependent transactivation of E2F promoters (4), also cripples chimeric DnaJ function. Within the uncertainty of the measurements, the Q32K chimera scarcely differs from the wild-type chimera (Table I). It is quite possible that PyJ Gln-32 interacts with one or more carboxylates of the hsp70 such that the Q32E substitution introduces electrostatic repulsion; alternatively, Q32E may offset the key charge of neighboring Lys-35 located within 6 Å. The functional equivalence of glutamine to the usual lysine or arginine at this f1 site suggests that the ability of glutamine to donate hydrogen bonds to carboxylate acceptors may suffice for the interaction with the hsp70. Likewise, Gln-31 at position e1 and Gln-36 at position c2 in helix II of PyJ, where positive charges are present in other domains (Figs. 3 and 5), can be imagined to donate hydrogen bonds to the same hsp70 carboxyl groups that could salt bridge to Arg or Lys residues of other J domains.
Comparison of the HPD Loops of PyJ, HDJ-1, and DnaJ-The side chain of the absolutely conserved aspartate of the HPD loop and the key positively charged residue in the b2 position (Lys-35 in PyJ, Arg-24 in HDJ-1, and Lys-25 in DnaJ) both face the same direction in PyJ, HDJ-1 (Fig. 5C), and active DnaJ J domains. However, in inactive DnaJ, this essential aspartate points in the opposite direction, presumably away from the DnaK interface. This absolutely conserved aspartate of HPD motif, i.e. Asp-35 of DnaJ, was implicated in recognition of conserved Arg-167 of DnaK (24). The orientation of the aspartate of the HPD motif relative to the key positively charged residue of helix II (at b2 or f3 positions) could be a key steric feature determining functional activity of a J domain.
The high precision of the PyJ structure further allowed identification of a turn of 3-10 helix through conserved residues Pro-43, Asp-44, and Lys-45 known to be absolutely required for J domain function (23). 3-10 helix is also now evident in the same region in the average structure of the active J domain of DnaJ. The environment of these 3-10 helix residues in DnaJ was shown to be altered by interaction with DnaK (72).
Features Unique to the Tumor Viral J Domain of PyJ-Although PyJ is similar in structure to other J domains, the PyJ possesses some unique structural features. Helix I of Py-J is about 2-fold longer than helix I in HDJ-1 and DnaJ. Seven additional residues are present at the extreme N terminus of the T antigens. Mutations of these residues suggest their importance in the mammalian host cell transformation (see below), an additional role unique to the J domains of T antigens. Mutagenesis data (5)(6)(7) are consistent with the two-residue insertion in the I-II loop of T Ag J domains and other aspects of the I-II loop conserved in polyomaviruses being necessary to support transformation. Just beyond the HPD loop, helix III of PyJ is one turn shorter than helix III of DnaJ. The C-terminal end of helix III of PyJ continues at least two turns (and bends) beyond where helix III of DnaJ and HDJ-1 end. The 22-residue helix III of PyJ is accompanied by a lack of the fourth helix present in DnaJ and HDJ-1. Consequently, the RK motif at the  1-104).
b Mean structures of HDJ-1, inactive DnaJ, and active DnaJ were calculated using backbone atoms C␣, N, and CЈ in the helical regions to fit the conformers.
beginning of helix IV, for specific recognition of hsp70 (73), is absent from PyJ.
Sites of Mutations in the N-terminal J Domain of T Antigens-Two general classes of mutations have been reported in the N-terminal domain of T antigens from murine polyomavirus and from SV40. One class of mutations toward the N terminus severely compromises transformation mediated by polyomavirus middle T Ag, which depends upon activation of pp60 c-Src and upon inhibition of PP2A. Such mutations include single and double point mutations of residues Asp-2, Val-4, Leu-5, Arg-7, Leu-13, Leu-14, Leu-16, Leu-17, Leu-19, Trp-24, and Gly-25 (5,6,11) colored red in Fig. 2B. The other class of mutations severely compromises DNA replication and cell cycle progression, which depends on large T Ag recruitment of hsc70, interaction with the Rb family of tumor suppressors, and elevation of E2F-dependent gene transactivation. Such mutations include point mutations of Gln-32, Ala-33, Tyr-34, His-42, Pro-43, Asp-44, Gly-46, and Gly-47 (green in Fig. 2B) in SV40 and/or polyomavirus large T Ag (4,12,74).
The substitutions blocking transformation cluster in two regions in the PyJ structure (Fig. 2B). The cluster of extreme N-terminal substitutions, i.e. Asp-2, Val-4, Leu-5, and Arg-7, might seem unlikely to have global effects on the PyJ structure, considering their surface location and the higher inherent mobility of residues Asp-2 to Val-4. However, the very limited activity of the V4D-substituted chimera at 42°C (Table I) might be a consequence of lower thermal stability, perhaps due to electrostatic repulsion from the Glu-65 in close contact with Val-4. Leu-5, being buried, may have importance in maintaining the structure of the domain. Specifically, mutation of Leu-5 may disrupt its packing with Asp-9 and Lys-10, thought to stabilize the N-capping box of helix I. Direct participation of the N-terminal five residues of PyJ and polyomavirus middle T Ag in binding PP2A seems a plausible hypothesis, considering the importance of these residues in immunoprecipitating PP2A (6) and their location in the structure. Substitutions of buried residues Leu-13, Leu-14, Leu-16, Leu-17, Trp-24, as well as Gly-25 may disrupt the hydrophobic structural core of PyJ, destabilizing the N-terminal domain of T antigens. The likelihood of such structural destabilization can be inferred from the low expression and activity of the PyJ-L13V-DnaJ chimera (see "Results" and Table I) and the instability in mammalian host cells of large T Ag having L13V or Leu-17 substitutions (4,75). Mutation of the hydrophobic core could possibly deform the PP2A recognition surface and/or disturb the packing of PyJ with its neighboring domains of the T antigens.
The T Ag substitutions that interfere with hsc70-dependent DNA replication also form two recognizable clusters (Fig. 2B). The cluster of residues Gln/Lys-32-Ala-33-Tyr-34-Lys-35 falls in the heart of helix II, which in E. coli DnaJ was implicated by NMR in recognition of the ATPase domain of DnaK (72). Mutations of Gln/Lys-32 and Lys-35 probably directly alter the hsc70 recognition surface. Mutations of Ala-33 and Tyr-34 may alter the hsc70 recognition surface indirectly by disrupting packing of the core. The other large T Ag substitutions interfering with hsc70-dependent DNA replication map to the universally conserved loop residues His-42-Pro-43-Asp-44 as well as to Gly-46 -Gly-47 which are conserved among T antigens. Like Lys-26 of DnaJ, universally conserved Asp-35 of DnaJ (equivalent to Asp-44 of PyJ) experienced a large NMR peak shift in the titration with DnaK (72). The much smaller chemical shift changes of DnaJ His-33 and Pro-34 upon DnaK titration suggest that they might not interact with DnaK as directly as do Asp-35 and Lys-26. Since in PyJ (and in HDJ-1), the conserved histidine appears to act as the C-cap and the conserved proline appears to act as the CЈ residue, a His-Pro C-cap of helix II may be a feature of all J domains. This His-Pro C-cap may be essential for proper positioning of the conserved aspartate for interaction with the conserved arginine of the hsp70 co-chaperone. The five-membered rings of conserved His-42 and Pro-43 of PyJ appear to stack against each other in a unique manner that other amino acids cannot duplicate (Fig.  4). We hypothesize that substitutions of the histidine or proline of the HPD motif could interfere in part by disrupting this stacking, the unique C-capping, and consequently the positioning of the aspartate for interaction with hsp70.