Low Resolution Structural Study of Two Human HSP40 Chaperones in Solution

Proteins that belong to the heat shock protein (Hsp) 40 family assist Hsp70 in many cellular functions and are important for maintaining cell viability. A knowledge of the structural and functional characteristics of the Hsp40 family is therefore essential for understanding the role of the Hsp70 chaperone system in cells. In this work, we used small angle x-ray scattering and analytical ultracentrifugation to study two representatives of human Hsp40, namely, DjA1 (Hdj2/dj2/HSDJ/Rdj1) from subfamily A and DjB4 (Hlj1/DnaJW) from subfamily B, and to determine their quaternary structure. We also constructed low resolution models for the structure of DjA1-(1–332), a C-terminal-deleted mutant of DjA1 in which dimer formation is prevented. Our results, together with the current structural information of the Hsp40 C-terminal and J-domains, were used to generate models of the internal structural organization of DjA1 and DjB4. The characteristics of these models indicated that DjA1 and DjB4 were both dimers, but with substantial differences in their quaternary structures: whereas DjA1 consisted of a compact dimer in which the N and C termini of the two monomers faced each other, DjB4 formed a dimer in which only the C termini of the two monomers were in contact. The two proteins also differed in their ability to bind unfolded luciferase. Overall, our results indicate that these representatives of subfamilies A and B of human Hsp40 have different quaternary structures and chaperone functions.

otide binding domain (NBD), and by interaction with its co-chaperones (5,7). Hsp40 assists the folding of nascent proteins, and prevents aggregation and the refolding of aggregates by presenting nascent proteins to Hsp70 and stimulating the hydrolysis of ATP (8 -12).
Hsp40, which can also act as a chaperone by itself (13)(14)(15), consists of four conserved functional regions, as determined by genetic and mutational studies in vivo (16 -19) and by biophysical methods (11, 20 -22) (Fig. 1A). The highly conserved ␣-helical N-terminal domain, referred to as the J-domain, is characteristic of proteins in this family, and the binding of this domain to Hsp70 stimulates the ATPase activity of Hsp70 (11,12,14,21,23). Adjacent to the J-domain is a glycine/phenylalanine-rich region (G/F-rich; Fig. 1A) that is disordered and likely to be responsible for flexibility (11,24). The central region consists of a cysteine-rich domain (Cys_rich; Fig. 1A) that includes four repeats of the motif CXXCXGXG (where X is any amino acid) and folds in a zinc-dependent fashion with two repeats bound to one zinc ion (20,25,26). The C-terminal domain (Fig. 1A) forms a ␤-sheet structure involved in the dimerization of Hsp40 (27). The Cys-rich and C-terminal domains are involved in substrate binding and presentation (22,25,28).
Hsp40 proteins occur throughout the cell and show a high diversity in eukaryotic genomes (29,30), with at least 44 genes present in the human genome (31). Based on their architecture and cellular location (4,32,33), Hsp40 proteins are classified in three main subfamilies (A-C, also referred to as types I-III; Fig.  1A) (4,33,34): subfamily A consists of proteins with the four domains described above, subfamily B contains proteins that lack the Cys-rich domain, and subfamily C has only the Jdomain that is not necessarily located at the N terminus (4,33). Hsp40 proteins of subfamily A have autonomous chaperone activity and may therefore work in an Hsp70-dependent or -independent manner. In contrast, Hsp40 proteins of subfamily B have no autonomous chaperone activity and depend on Hsp70 for full activity (18,33,35,36).
Although high resolution structures of the C-terminal, Cysrich, and J-domains are available, there is still little information about how these domains interact with each other, with other domains and also with Hsp70. To obtain further information about relevant structural and functional characteristics of human Hsp40 proteins, we used biophysical methods, including small angle x-ray scattering (SAXS) and analytical ultracentrifugation (AUC), to study DjA1 (also known as Hdj2/dj2/ HSDJ/Rdj1), a representative of subfamily A, its C-terminal deletion mutant (DjA1-(1-332)), and DjB4 (also known as Hlj1/ DnaJW), a representative of subfamily B that shares 65% identity with DjB1/Hdj1.

EXPERIMENTAL PROCEDURES
Cloning, Expression, and Purification-DjA1 was cloned from the human gene DNAJA1 (cDNA clone, GenBank TM accession number AW247277) and DjB4 was cloned from the human gene DNAJB4 (cDNA clone, GenBank TM accession number AA081471). The Hsp40 nomenclatures DjA1 and DjB4 were used as described by Ohtsuka and Hata (34). Two primers were used to amplify the DjA1 cDNA by PCR and to create restriction enzyme sites for NdeI and XhoI: a DjA1 5Ј-primer (5Ј-CCG-GCAGGCTAGCATGGTGAAAGAAACAAC-3Ј) containing an NdeI restriction site and a DjA1 3Ј-primer (5Ј-TGAGTGTTATTCTCGAGTCAT-TAAGAGGTCTG-3Ј) containing an XhoI restriction site. Two primers were used to amplify the DjB4 cDNA by PCR and to create restriction enzyme sites for NdeI and BamHI: a DjB4 5Ј-primer (5Ј-TCAAGGCAT-TCCATATGGGGAAAGACTATTA-3Ј) containing an NdeI restriction site and a DjB4 3Ј-primer (5Ј-GTGTAACAAAGTGGATCCTACTAT-GAGGCAGG-3Ј) containing a BamHI restriction site. The C-terminal deletion of DjA1 was constructed by site-directed mutagenesis using a primer that created a stop codon at residue Phe 333 followed by a restriction site for XhoI (5Ј-TTATCAGGCTCGAGTTAGCCATTCTC-3Ј).
The PCR products were cloned into the pET28A expression vector (Novagen) for His tag purification methods. The correct cloning was confirmed by DNA sequencing using an ABI 377 Prism system (PerkinElmer Life Sciences). These procedures created the vectors pET28aDjA1, pET28aDjB4, and pET28aDjA1-(1-332), which were transformed into the Escherichia coli strain BL21(DE3) for heterologous protein expression by adding isopropyl thio-␤-D-galactoside (0.4 mM) at A 600 ϭ 0.6. The induced cells were grown for 5 h and harvested by centrifugation for 10 min at 2,600 ϫ g. The bacterial pellet was resuspended in lysis buffer (50 mM Tris-HCl, pH 8.0, 500 mM KCl and 10 mM EDTA; 15 ml of buffer/liter of medium) in the presence of 5 units of DNase (Invitrogen, Life Technologies, Inc.) and 30 g of lysozyme/ml (Sigma) for 30 min. The pellet was then disrupted by sonication in an ice bath, and centrifuged as described above. The supernatant was fractionated by metal affinity chromatography in a HiTrap chelating column (Amersham Biosciences) using an AKTA FPLC system (Amersham Biosciences). The proteins were eluted with imidazole (500 mM) and loaded onto a HiLoad Superdex 200-pg column (2.6 cm ϫ 60 cm; Amersham Biosciences) using an AKTA FPLC system. The degree of purification was estimated by SDS-PAGE, and the protein concentration was determined spectrophotometrically using a calculated extinction coefficient for denatured proteins (37,38).
Analytical Molecular Exclusion Chromatography-Analytical molecular exclusion chromatography was done using a Superose 12 HR 10/30 (Amersham Biosciences) column coupled to an AKTA Purifier system (Amersham Biosciences). The column was equilibrated with 25 mM Tris-HCl, pH 7.5, containing 500 mM NaCl, 1% glycerol, and ␤-mercaptoethanol (1-10 mM). The column was washed with two column volumes of this same buffer at a flow rate of 0.5 ml/min, and aliquots of proteins in 100 l were loaded onto the column. The elution profile was determined by monitoring the absorbance at 280 nm. The apparent molecular mass was calculated using a plot of ln of the molecular mass (kDa) of standard proteins (thyroglobulin, 669 kDa; ␥-globulin, 160 kDa; bovine serum albumin, 69 kDa; chicken ovalbumin, 45 kDa; cytochrome c, 12 kDa) versus the retention time.
Circular Dichroism Spectroscopy-Circular dichroism (CD) measurements were done using a Jasco J-810 spectropolarimeter with the temperature controlled by a Peltier-type control system PFD 425S. Hsp40 proteins were resuspended in 25 mM Tris-HCl, pH 7.5, containing 500 mM NaCl and 1 mM ␤-mercaptoethanol at final concentrations of 20 -50 M. The data were collected at a scanning rate of 50 nm/min with a spectral bandwidth of 1 nm using a 0.1-mm path length cell. CDNN deconvolution software (Version 2, Bioinformatik.biochemtech. uni-halle.dee/cdnn) was used for secondary structure prediction. All buffers used were of analytical grade and were filtered before use to avoid light scattering by small particles.
Measurement of Chaperone Activity-The method described by Lu and Cyr (28,35) was used to assess the ability of Hsp40 proteins to interact with unfolded proteins. Briefly, luciferase (Promega) was chemically denatured by diluting in guanidinium-HCl (6 M) for 40 min at room temperature and then diluted 25ϫ in 50 l of 25 mM Tris-HCl, pH 7.5, containing 500 mM NaCl and 5 mM ␤-mercaptoethanol in the absence and presence of His-tagged Hsp40 proteins (see figure captions for details of the concentrations). In another protocol, luciferase (2 M) was thermally denatured by incubation at 42°C for 10 min in the absence and presence of His-tagged Hsp40 proteins (see figure captions for details of the concentrations) and then cooled to room temperature. The solutions were centrifuged for 10 min at 21,000 ϫ g, and 50 l of the supernatant was incubated with 50 l of a 50% slurry of Talon metal chelate resin (Clontech) in 25 mM Tris-HCl, pH 7.5, containing 500 mM NaCl and 5 mM ␤-mercaptoethanol for 1 h at room temperature. The mixture was then centrifuged for 1 min at 10,000 ϫ g and 4°C, and the pellet containing the resin was washed twice with 75 l of the Tris buffer indicated above containing 15 mM imidazole. The protein complexes bound to the resin were eluted with this same Tris buffer containing 150 mM imidazole and then concentrated by precipitation with acetone 80% (Merck) and visualized by SDS-PAGE.
Analytical Ultracentrifugation-The sedimentation velocity and sedimentation equilibrium experiments were done with a Beckman Optima XL-A analytical ultracentrifuge. The proteins DjA1, DjA1-(1-332), and DjB4 were tested at concentrations from 50 to 1000 g/ml in 25 mM Tris-HCl, pH 7.5, containing 500 mM NaCl, 1% glycerol (but not for DjA1-(1-332)) and 1 mM ␤-mercaptoethanol, with no apparent aggregation. The sedimentation velocity experiments were done at 20°C, 25,000 rpm for DjA1 and DjB4, and 30,000 and 40,000 rpm for DjA1-(1-332) (AN-60Ti rotor), and the scan data were acquired at 230 and 238 nm for low and high protein concentrations, respectively. The sedimentation equilibrium experiments were done at 20°C at speeds of 6,000, 8,000, and 10,000 rpm with the AN-60Ti rotor and scan data acquisition at 238 nm. Analysis of the data involved the fitting of a model of absorbance versus cell radius data using nonlinear regression. All fittings were done using the Origin software package (Microcal Software) supplied with the instrument. The van Holde-Weischet (39) (sediment coefficient plot) and the sedimentation time derivative (g(s*) integral distribution) (40) methods were used to analyze the sedimentation velocity experiments. The methods used to analyze the velocity and equilibrium experiments allowed the calculation of the apparent sedimentation coefficient s, the diffusion coefficient D, and the molecular mass M. The ratio of the sedimentation to diffusion coefficients gave the molecular mass in Equation 1, where R is the gas constant and T is the absolute temperature. The Sednterp software (www.jphilo.mailway.com/download.htm) was used to estimate the partial specific volume of the proteins at 20°C (VbarDjA1 ϭ 0.7275 ml/g, VbarDjA1-(1-332) ϭ 0.7309 ml/g, and VbarDjB4 ϭ 0.7302 ml/g), the buffer density ( ϭ 1.02163 g/ml) and the buffer viscosity ( ϭ 1.0851 ϫ 10 Ϫ2 poise). The self-association method was used to analyze the sedimentation equilibrium experiments, and several models of association for DjA1 and DjB4 were used to fit the data. The distribution of each protein throughout the cell, as determined in the equilibrium sedimentation experiments, was fitted with Equation 2 (41), where C is the protein concentration at radial position r, C 0 is the protein concentration at radial position r 0 , and is the centrifugal angular velocity. The Sednterp software was used to estimate the standard sedimentation coefficients (s 20,w ) at each protein concentration and to calculate s 0 20,w by extrapolation to a protein concentration of 0 mg/ml. This procedure corrects for effects of temperature, solution viscosity, and molecular crowding (42).
SAXS-The SAXS experiments were done at the SAS beamline of the LNLS synchrotron radiation facility in Campinas, Brazil (43) under the same conditions as those described by Borges et al. (44). The measurements were done using a monochromatic x-ray beam with a wavelength ϭ 1.488 Å. For a sample-to-detector distance of 840 mm selected for the experiments, the modulus of the photon momentum transfer (q ϭ 4sin/, 2 being the scattering angle) covered a range from q ϭ 0.01 Å Ϫ1 to q ϭ 0.44 Å Ϫ1 . The scattering intensity curves, I(q), were recorded using a gas 1 dimensional position sensitive x-ray detector. To monitor possible effects from radiation damage and synchrotron beam instabilities, the SAXS curves were determined in many short frames (90 s each). The SAXS data obtained were normalized to account for the natural decay in intensity of the synchrotron incident beam and were corrected for non-homogeneous detector responses. Finally, the scattering intensity produced by the buffer was subtracted, and the difference curves were scaled to give equivalent protein concentrations. Three types of samples with different concentrations were studied by SAXS: 1) 3.3 mg/ml DjA1 in 25 mM Tris-HCl, pH 7.5, containing 500 mM NaCl, 1% glycerol and 1 mM dithiothreitol, 2) 2.9 and 6.5 mg/ml of DjA1-(1-332) in 25 mM Tris-HCl, pH 7.5, containing 500 mM NaCl and 1 mM ␤-mercaptoethanol, and 3) 2.5, 4.4, 6.2, and 8.3 mg/ml DjB4 in 25 mM Tris-HCl, pH 7.5, containing 500 mM NaCl and 1 mM ␤-mercaptoethanol.
Reliable structural information about the low resolution structure of proteins in solution can be derived from SAXS data provided that all proteins are in the same (monomeric or oligomeric) state and that the solution is "dilute," i.e. interference effects in scattering amplitudes produced by different proteins are negligible. Under these assumptions, the total scattering amplitude is proportional to the form factor of an isolated (monomeric or oligomeric) protein averaged for all orientations. When samples with a high concentration were available, data merging was done using low and high concentration samples for the high and low q ranges of the scattering intensity, respectively. This procedure reduced the overall statistical error in the scattering curves without introducing unwanted interferences or spatial correlation effects in the small q range.
To establish the molecular masses of DjA1, DjA1-(1-332), and DjB4, the SAXS intensity produced by a solution of bovine serum albumin (69 kDa, 5 and 10 mg/ml) was also determined. Since, for dilute solutions, the normalized intensity extrapolated to q ϭ 0, I(0), is proportional to the molecular mass, the molecular masses (M) of the proteins investigated here were determined from the ratios between the I(0) values corresponding to the different samples and that of the standard bovine serum albumin.
Computer Programs-By applying the indirect Fourier transform program GNOM (45) to the normalized SAXS curves, the distance distribution function p(r) and the radius of gyration, Rg, of the proteins studied were evaluated (46). Prior to GNOM analysis, a constant intensity background was subtracted from the SAXS data to ensure that the intensity at higher angles decayed as q Ϫ4 , according to Porod's law for a two-electron density model (46). The low resolution shapes (or molecular envelopes) of the proteins were restored from the experimental SAXS curves using an ab initio method named DAMMIN (47) and recently described in Borges et al. (44). Briefly, in this method, the low resolution shape of the protein was simulated by a set of small spheres that initially filled another sphere with a diameter equal to the maximum diameter, D max , of the protein being studied (previously determined using GNOM). DAMMIN yielded a structural model containing a fraction of the initial number of dummy atoms whose associated scattering intensity gave the best fit to the SAXS data. The configuration of the different protein domains could be refined by manually rotating or displacing the domains until the discrepancy between the calculated and experimental SAXS curves was minimized. The protein models were displayed using the program WebLab ViewerLite software (www.accelrys.com).
The HydroPro software (48) was used to estimate the translational diffusion coefficient D t , the radii of gyration R g , the sedimentation coefficient s, and the maximum distance (D max ) starting from the ab initio models generated by DAMMIN using SAXS data obtained at 20°C. The HydroPro software was configured with the radius of the atomic elements from the ab initio analysis, with sigma factors from 5 to 8 (as indicated by the supplier) and a minibeads radius from 6 to 2 Å (SIGMIN and SIGMAX), after an initial evaluation of the two extremes. The parameters Vbar, , and were estimated using the software Sednterp as described above. The translational fraction ratio or Perrin factor P, which indicates the relationship between the frictional coefficient of the Hsp40 particles and a sphere of the same molecular mass (f/f 0 ), was estimated using Solpro software (49).

RESULTS AND DISCUSSION
Purification of DjA1 and DjB4 as Folded Dimers and of DjA1-(1-332) as a Folded Monomer-The correct cloning and sequencing of DjA1, DjB4, and DjA1-(1-332) were confirmed by DNA sequencing. All of the proteins were expressed in large quantities and were more than 95% pure as confirmed by SDS-PAGE ( Fig. 2A). The proteins were unstable at low ionic strength but were soluble in the presence of 500 mM NaCl. DjA1 and DjB4 were purified as dimers and the mutant DjA1-(1-332) was purified as a monomer, as confirmed by analytical molecular exclusion chromatography (Fig. 2B). The folding state of the proteins was investigated by CD, and the resulting spectra corresponded to folded proteins with a minimum at 208 nm (Fig. 2C). The analysis using the CDNN deconvolution software indicated that the proteins had similar amounts of secondary structure: 35% ␤-sheet structure and 10% ␣-helices. The high content of ␤-sheet structure agreed with x-ray crystallography and nuclear magnetic resonance data indicating the presence of ␤-sheet structures in the C terminus (22,27) and in the Cys-rich domain (22,26) of Hsp40.
Although the deletion of the last 54 residues of DjA1 prevented the dimerization of this protein, it had no marked effect on the global conformation (Fig. 2B). However, the thermal stability of the deleted mutant was lower than that of the whole protein. DjA1 and DjB4 were heated to 50°C without loss of structure whereas the amount of secondary structure in DjA1-(1-332) decreased when the protein was heated above 38°C (data not shown). These results indicated that the dimeric state was important for the stability of the protein at temperatures well above the body temperature, an important characteristic for proper cell functioning.
Efficiency of DjA1 in Binding Unfolded Luciferase Compared with That of DjB4 -Because the chaperone action of Hsp40 involves hydrophobic interaction with unfolded or partially folded proteins (15,18,19,22), we investigated the ability of Hsp40 proteins to bind unfolded luciferase. His-tagged chaperones have a high affinity for the Talon metal chelate resin and bind to it in the absence of imidazole. Any unfolded luciferase that binds to a His-tagged chaperone co-purifies with the chaperone when the resin is washed with 150 mM imidazole. Fig. 3 shows the SDS-PAGE profiles of washed samples of chemically (Fig. 3A) or thermally (Fig. 3B) denatured luciferase in the absence or presence of increasing concentrations of DjA1, DjA1-(1-332), or DjB4 (see figure caption for details). At 22°C, folded luciferase did not bind to DjA1 or DjB4, whereas at 42°C unfolded luciferase bound to DjA1 but not to DjB4 (Fig. 3B). Similar results were obtained with chemically unfolded luciferase (Fig. 3A), with this protein binding to DjA1 but having a very low affinity for DjB4. The deletion of the C terminus in the mutant DjA1-(1-332) had a major effect on the efficiency of binding (Fig. 3A).
in the luciferase thermal unfolding experiments because it unfolded at temperatures Ͼ38°C. The lower chaperone activity of DjA1-(1-332) indicated that the quaternary structure of Hsp40 was related to its chaperone activity, as also shown for yeast Hsp40 Sis1, in which a C-terminal deletion abolished the ability to assist Hsp70 (27).
Envelope Models Derived from SAXS-The experimental (corrected and normalized) SAXS curves for DjA1, DjA1-(1-332), and DjB4 are shown in Fig. 4A. Molecular masses of the proteins studied were obtained from the quotient between their intensity extrapolated to q ϭ 0 and the I(0) value corresponding to bovine serum albumin, as described above. The molecular masses of DjA1 and DjB4 were determined to be 95 and 90 kDa, respectively, in good agreement with the values calculated from their amino acid sequences (Table I), assuming that they are dimers. In contrast, the SAXS data for DjA1-(1-332) indicated the existence of a monodisperse set of molecules with a molecular mass of 40 kDa, in good agreement with the value calculated from the amino acid sequence (Table I) of the monomer. The GNOM program was subsequently used to determine the distance distribution function and related structural parameters, such as the radius of gyration (R g ) and the maximum diameter (D max ) of the proteins. The distance distribution functions, p(r), of the three proteins studied are shown in Fig.  4B. Since these functions showed no negative values, we concluded that all of the proteins were in a "dilute" state, as required for further analysis of molecular shape. The values of D max and R g determined by GNOM are shown in Table II. DjA1 and its C-terminal deletion mutant, DjA1-(1-332), had similar maximum diameters (D max ϭ 140 Å) (Fig. 4B), thus implying that they had a similar asymmetry. The DjB4 chaperone had a D max of 200 Å (Fig. 4B), despite having a smaller molecular mass than DjA1 (Table I), which suggested a more elongated conformation than for DjA1. Based on the overall trend of the p(r) functions (Fig. 4B), we concluded that DjB4 had a rather elongated shape (51), and that the shapes of DjA1 and DjB4 were clearly different. Indeed, whereas DjA1 had a p(r) function that showed only a smooth, well-defined peak, the p(r) function for DjB4 had two clear shoulders, one above and another below the main peak (Fig. 4B). This complex profile of the p(r) function suggested that DjB4 consisted of well-separated subunits.
The third step of data analysis was to apply the DAMMIN program to the SAXS data to obtain the low resolution molecular shape of the proteins. The dummy atom model (DAM) of the structure was derived from the experimental data by assuming a 2-fold symmetry for the dimers. This symmetrical restriction significantly reduced the number of free parameters of the models. The starting volume for the DjA1 dimer corresponded to a filled sphere with a D max ϭ 140 Å (provided by GNOM) containing 4,196 dummy atoms with a radius (r a ) ϭ 4.5 Å. By using the DAMMIN program, 600 Ϯ 10 dummy atoms were assigned to the resulting final model of the DjA1 dimer. In contrast, the starting volume for DjA1-(1-332) was filled with 5,594 dummy atoms with r a ϭ 3.7 Å within a sphere of D max ϭ 140 Å. Of these, 395 Ϯ 5 dummy atoms were assigned by DAMMIN to the final model of the DjA1-(1-332) monomer. The starting volume for DjB4 was filled with 6,699 dummy atoms with r a ϭ 4.8 Å within a sphere of D max ϭ 200 Å. Of these, 280 Ϯ 10 dummy atoms remained in the final model of the DjB4 dimer. Twenty independent ab initio DAMMIN simulations were done for each protein.
To determine the uniqueness of the resulting shapes, the restorations were done with different starting conditions, all of which yielded similar results. The DAM models obtained were superimposed and averaged using the DAMAVER package software (52) (Fig. 5). The relevant parameters associated with the use of DAMMIN and the values of D max and R g derived from it are reported in Table II. These values agreed well with the same parameters determined by GNOM for the three molecules studied. The DAM-derived hydrodynamic parameters (estimated by HydroPro) are shown in Table III. The maximum concentrations of protein used for the SAXS studies of the DjA1 and DjB4 dimers allowed 30 Å as the maximum resolution of the final models (Table II), even when the data were measured down to 15 Å. This resolution did not permit an unambiguous determination of the spatial positions of the secondary structural elements of DjA1, but allowed us to determine the overall shape of the molecule and the relative positions of their individual domains. Table II shows different resolutions for the GNOM and DAMMIN calculations. For DjA1, the whole SAXS curve determined up to q max ϳ 0.4 Å Ϫ1 was used for GNOM so that the resolution 2/q max was equal to ϳ 15 Å. Within the high q range from q ϭ 0.2 Å Ϫ1 to 0.4 Å Ϫ1 (Fig. 4A, curve 2), the statistical errors were very high, and therefore no significant structural information was obtained from this high q range. To speed up the rather long calculations of the DAMMIN program while avoiding a significant loss of structural information, only SAXS data up to q max ϭ 0.2 Å Ϫ1 were used for this analysis and yielded a resolution of 30 Å. The inherent limitations of shape determination methods such as DAMMIN prevent higher resolution protein envelopes from being obtained from statistically good SAXS data at higher q max values, mainly because these methods assume a uniform electron density that limits the resolution to 20 -30 Å.
GNOM analysis yielded the same D max for DjA1-(1-332) and DjA1, despite the dimer arrangement of the latter. This was a clear indication of the highly asymmetrical shape of the DjA1-(1-332) monomer, a characteristic also deduced from the profile of the p(r) function. Fig. 5A shows the envelope models for DjA1 (left) and DjA1-(1-332) (two monomers, center). The two monomers of DjA1-(1-332) are shown facing each other to illustrate how they may form a dimer, and the monomer envelope placed within the dimer envelope fitted one half of the envelope very well (Fig. 5A, right). The hydrodynamic properties of the two combined monomers of DjA1-(1-332), calculated by the Hydro-Pro software were similar to those determined for the DjA1 dimer model (data not shown). These results confirmed that the DAM model determined for DjA1 was a good description of its low resolution structure. Two perpendicular views of the envelope model of DjB4 are shown in Fig. 5D. Comparison of the shapes of the DjA1 and DjB4 envelopes strongly indicated that these proteins clearly had different quaternary structures.
The Hydrodynamic Parameters Calculated by AUC Corroborated the Models Generated from the SAXS Data-The hydrodynamic properties of DjA1 and DjB4 were established by sedimentation equilibrium (Fig. 6, A and B) and sedimentation velocity (Fig. 7A). In these experiments, the proteins behaved as a monodisperse system and showed no aggregation. The profiles of the sedimentation equilibrium for DjA1 (Fig. 6A) and DjB4 (Fig. 6B) fitted well to a single exponential, as indicated    DjA1-(1-332), and DjB4 in solution These were studied by SAXS, and mathematical parameters were calculated using the programs GNOM and DAMMIN. In all cases, the experimental error was less than 10%. Mathematical  by the analysis of their residuals. The molecular masses of DjA1 and DjB4 derived from the sedimentation equilibrium measurements were 93.0 Ϯ 1.0 kDa and 80.0 Ϯ 1.0 kDa, respectively (Table I). For DjA1-(1-332), the molecular mass was 42.0 Ϯ 2.0 kDa, which was estimated from Equation 1 using the sedimentation coefficient (s 0 20,w ) calculated from the sedimentation velocity analysis (see below) and the diffusion coefficient (D 0 20,w ) calculated from dynamic light scattering experiments (data not shown). The molecular masses determined here for the human Hsp40 proteins agreed well (within experimental error) with the molecular masses calculated from the amino acid sequences and from the SAXS results (Table I). To determine the sedimentation coefficients s of the proteins studied, the sedimentation velocity experiments were done at different concentrations and velocities (Fig. 7). The standard sedimentation coefficients s 20,w obtained were plotted as a function of the protein concentration used to calculate the standard sedimentation coefficient at 0 mg/ml, s 0 20,w , by extrapolation. This approach was used to correct for effects of temperature, solution viscosity and molecular crowding (42). For DjA1, we obtained s 0 20,w ϭ 4.63 S, and for DjB4 s 0 20,w ϭ 3.78 S (Fig. 7B). The sedimentation coefficient for DjA1-(1-332) was velocitydependent, with s 0 20,w ϭ 2.85 S at 30,000 rpm and 3.40 S at 40,000 rpm, both of which were larger than the value predicted by the DjA1-(1-332) DAM model (s 0 20,w ϭ 2.46 S; see Table III). This effect may be caused by aggregation or, more likely in this case, by a highly asymmetrical conformation (53). Some subdomains of Hsp40 proteins may become more flexible in the monomer form, which would reduce the frictional ratio when compared with a rigid body and would distort the measured s values (54). Since the sedimentation velocity measures the time average s of the particles in solution, it is not possible to distinguish a molecule with a high frictional ratio from another with flexible segments (55). In addition, the Perrin factor and the elongated shape of DjA1-(1-332) determined from the SAXS data and its domain structure suggested that this protein was flexible, thereby reinforcing the hypothesis that the apparently high s 0 20,w was caused by molecular flexibility. The hydrodynamic parameters determined from the AUC experiments for DjA1, DjA1-(1-332), and DjB4 were similar to those derived from the DAM models of DjA1 and DjB4 by the HydroPro software (Table III). This finding confirmed the reliability of the structural models obtained from the SAXS results. Together, the SAXS and the AUC data indicated that DjA1 and DjB4 were dimers with different structural features, and that DjA1-(1-332) was a monomer with flexible segments (Tables I and III). The structural parameters obtained from the AUC measurements strongly suggested that the DAM models derived from the SAXS data represented a good, low resolution structure of the real conformation of the proteins studied here.
The results presented in this and in the preceding sections indicate that DjA1, DjA1-(1-332), and DjB4 were produced as folded proteins with the conformational properties of members of the Hsp40 chaperone family. The next step was to use the low resolution models of the proteins determined from the SAXS data, together with information about the crystallographic and NMR structures of Hsp40 domains available in the Protein Data Bank, to obtain a more detailed insight into the structural conformation of DjA1 (dimer and monomer) and DjB4.
Quaternary Structure of DjA1-The DjA1-(1-332) mutant was a monomer with an elongated shape, as determined by analytical molecular exclusion chromatography, AUC and SAXS. The NMR structure of the J-domain (56) displayed in red in Fig. 5B, and the crystallographic structure of the Cysrich and C-terminal domains (22) shown in green in Fig. 5B, were fitted into the DjA1-(1-332) envelope. The final arrangement is shown in Fig. 5B, where some important features can be seen, namely, the location of the C terminus within the thinnest part of the protein envelope and that of the J-domain in the thickest part of the envelope (Fig. 5B). Some portions of the domains in this model were highly flexible (arrows in Fig. 5B) because of the presence of the G/F-rich region (see below). As argued before, this could explain why the hydrodynamic properties predicted by the HydroPro software for the DjA1-(1-332) envelope were not in perfect agreement with the experimental AUC data. The apparently high sedimentation coefficient for DjA1-(1-332) may be justified by the presence of the G/F-rich region, which is believed to function as a flexible spacer between the J-domain and the remaining structure of the protein (11, 24).   DjA1-(1-332), and DjB4 in solution These were calculated from the AUC data and the structural parameters derived from the DAM model generated by the SAXS experiments using the HydroPro software. The partial specific volume, solvent viscosity, and density were determined using Sednterp software. In all cases, the experimental error was less than 2%. In the Cys-rich domain determined by crystallography (22), the subdomains I and II (Fig. 5B) form an L-shaped structure (90 o angle), whereas the same domain determined by NMR forms an angle of about 45 o resulting in a V-shaped structure (26). This indicates that subdomain II may have some flexibility and could account for the apparently high sedimentation coefficient of DjA1-(1-332) calculated in the AUC experiments. In addition, the Hsp40 proteins studied here were highly susceptible to degradation (data not shown) and the high flexibility of the G/F-rich region may be an explanation for this behavior.
The results of the SAXS experiments indicated that DjA1 and DjA1-(1-332) had an elongated shape and similar D max values, which suggested that they probably shared the same axial orientation. The superposition of the two models (Fig. 5A, right) revealed that a dimer composed of two DjA1-(1-332) monomers (Fig. 5A, center) had nearly the same structure as the DjA1 dimer. This finding indicated that the region involved in dimer formation was located mainly in the C terminus of DjA1. The deletion of this region blocked dimer formation and decreased the protein stability, but had no effect on the secondary and tertiary (low resolution) structures of DjA1-(1-332). Based on the observed orientation of the monomers in the DjA1 dimer model, we fitted the high resolution structures of the J-, Cys-rich, and C-terminal domains into the monomer envelope (Fig. 5B), and then built a model to show the arrangement of the Hsp40 domains in the DjA1 dimer (Fig. 5C). The DAM model generated from the SAXS data indicated that DjA1 was a dimer with a bent horseshoe shape, similar to the structure determined for the C-terminal and Cys-rich domains of Ydj1 (subfamily A) (22). Fig. 5C shows that the DjA1 protein was bullet-shaped and that the C termini monomers were located in the thinnest side of the model and were responsible for the dimerization. The N-terminal monomers (J-domains) were located in the thickest part of the model where they could act as a clamp for binding to Hsp70 (Fig. 5C). The C-terminal and Cys-rich domains of Ydj1 (green in Fig. 5C; 1NLT, Ref. 21) were placed on the axial axis, where they fitted very well. The L-shape of this complex was a consequence of the presence of the Cys-rich domain, which was placed in the center to face its counterpart in the other monomer (Fig. 5B).
Quaternary Structure of DjB4 -The data generated by SAXS were used to create an envelope model of the quaternary structure of DjB4 that agreed with the structural parameters measured (Fig. 5D). The envelope model of DjB4 was compatible with an elongated structure, with a D max greater than that for DjA1 (200 Å versus 140 Å). This was an unexpected finding because DjA1 has a higher molecular mass than DjB4, as shown by the AUC experiments (Table I). The similarities between the predicted and the calculated hydrodynamic parameters of the envelope model were a strong indication of the accuracy of the model (Table III). The DjB4 ab initio model generated from the SAXS data using DAMMIN and DAMAVER is shown in Fig. 5D. The high resolution structures of the Hsp40 domains available, namely, the NMR structure of the J-domain (56), which is shown in red in Fig. 5E, and the crystallographic structure of the C terminus of Sis1 (subfamily B) (22), which is shown in green in Fig. 5E, were fitted into the DjB4 envelope in the best possible position. Studies of C-terminal deletions of Hsp40 (22,27, this work) have indicated that this region is responsible for dimerization, which implies that the C-terminal monomers must be located near to each other. The localization of the two C-terminal domains of DjB4 in the center of the model as shown in Fig. 5E fulfilled this requirement. Any other arrangement for these domains would complicate the formation of the dimer, and would be contrary to the strong tendency of Hsp40 proteins to form dimers (22,27). The J-domains were placed at the extremities in the model (Fig. 5E) because they are connected to the C-terminal by a long, flexible region formed by the G/F-rich region (residues 68 -139) plus a sequence of 60 amino acids of unknown function (residues 140 -199). There is no high resolution structure of this region, which makes it difficult to predict the correct arrangement of these regions (residues 68 -199) in the model. This uncertainty is represented by a dotted line (Fig. 5E).
DjA1 and DjB4: Similarities and Differences-Genetic and biochemical studies have shown that the Hsp40 family is structurally and functionally diverse (33). Such diversity means that the interaction of Hsp70 with different Hsp40 generates specialized combinations that facilitate specific actions of the Hsp70 chaperone machinery within the cell (19,33,35). Al- though Hsp40 from subfamilies A and B differ in their chaperone activity, the reasons for these differences are still incompletely understood. Determination of the tertiary structure of some of the domains of Hsp40 proteins from subfamilies A and B (22,26,27,56) has contributed to our understanding of the differences in their chaperone activity. The data on the quaternary structure of the two proteins studied here provides additional information that may help to explain the differences in their binding to substrates and to Hsp70 proteins.
The presence of a J-domain defines a protein as a member of the Hsp40 family. Since the conserved residues in this domain are responsible for stabilizing the anti-parallel coiled coil (56), we suggest that this domain has the same conformation in DjA1 and DjB4. However, the J-domains are located differently in their quaternary structure (because the dimerization sites, at the C terminus, must be in the center of the model): the J-domains from each monomer of DjA1 are arranged facing each other, whereas in DjB4, these domains are placed in the extremities. Since there is still no high resolution structure of the G/F-rich region, which is an important site for the binding of Hsp40 to Hsp70 (20, 21), we represented this region as a dotted line in our models (Fig. 5, B, C, and E). Based on primary sequence homology, the G/F-rich region is larger in DjB4 than in DjA1 (Fig. 1B) and, in combination with a region of about 60 residues for which there is also no available structure, this allowed for the more extended conformation shown for DJB4 (Fig. 5E), in which the J-domains are located in the extremities. The C-terminal regions of both subfamilies A and B are responsible for substrate binding and are structurally similar (22), but differ in their ability to refold luciferase (28,35). Lu and Cyr (35) studied the interaction of Ydj1 (subfamily A) and Sis1 (subfamily B) with two Hsp70s and concluded that these two Hsp40 proteins differed in their chaperone functions, perhaps because of the zinc finger region in Ydj1 which may stabilize the polypeptide binding groove of Hsp40. Fan et al. (19) engineered two Hsp40 chimeras in which the central part of Ydj1 (residues 101-255) was replaced by the central part of Sis1 (residues 108 -257) and vice versa. This exchange altered the substrate binding specificity of the proteins and their ability to stimulate the luciferase refolding activity of Hsp70. These results indicated that selective substrate binding sites are needed for substrate selection in Hsp40 and that the central portion of this protein is important for specifying chaperone activity (19). In agreement with these observations, the central part of DjA1 and DjB4 appeared to be the main region responsible for the differences in the low resolution models described here: the central part of DjA1 (G/F-rich region plus Cys-rich region, residues 68 -207) apparently keeps the J-do-mains close to each other whereas the central part of DjB4 (residues 68 -199) keeps the J-domains far apart from each other. A different global structural conformation rather than specific amino acid interactions could be responsible for the difference in the substrate specificity of the chaperones, as suggested by Lopez et al. (57), who compared the ability of Sis1/Hdj1 (subfamily B) and Ydj1/Hdj2 (subfamily A) Hsp40s to maintain the Rnq1 prion. Hence, the differences in the quaternary structures of DjA1 and DjB4 mentioned above could contribute to the variations in chaperone activity.
The high diversity of Hsp40 proteins may mediate several types of interactions between Hsp70 and Hsp40, and could explain the need for proteins with different quaternary structures in subfamilies A and B. Several lines of evidence suggest the presence of multiple binding sites among Hsp70 and Hsp40 partners: Suh et al. (21) showed the existence of at least two binding sites for DnaK in DnaJ, both located in the J-domain. Linke et al. (36) suggested that a possible contact between the Cys-rich domain may be important for the transfer of substrate from DnaJ to DnaK. Demand et al. (58) and Qian et al. (59) provided evidence that the C terminus of eukaryotic Hsp70 interacts with Hsp40 from subfamily B. Laufen et al. (12) proposed that the interaction between Hsp40 and Hsp70 may involve multiple steps in which the Hsp40 protein "targets" unfolded substrate to Hsp70, as follows: 1) interaction between the J-domain with the Hsp70 NBD-ATP, 2) shift of substrate from Hsp40 to the Hsp70 substrate binding domain (SBD), 3) transfer of the information regarding substrate binding from the Hsp70 SBD to the Hsp70 NBD, resulting in the complete hydrolysis of ATP and closing of the Hsp70 SBD lid and, finally, 4) Hsp40 dissociation from the Hsp70-ADP-substrate complex. Because of the high diversity of Hsp40 and the existence of multiple contacts among different Hsp40-Hsp70 partners (21,36,58,59), the functional cycle described above may consist of additional steps that depend on the Hsp40-Hsp70 partners.
The fact that DnaJ (subfamily A) is unable to bind to the isolated N-and C-domains of DnaK (21), but that Hsp40 from subfamily B does (58,59), is in good agreement with the quaternary models proposed here: the J-domains of DjA1 are arranged in a position that allows binding to both domains of Hsp70 at once, while the J-domains of DjB4 can act independently of each other when binding to Hsp70. The quaternary structure of DjA1 presented here is compatible with the model of interaction between DnaJ and DnaK proposed by Suh et al. into the groove between the monomers of Hsp40. Despite their low resolution, the size and shape of the models presented here based on SAXS are compatible with these views. Fig. 8 provides a diagram suggesting the possible interactions between DjA1 and DjB4 with Hsp70, based on the quaternary structure determined here for DjA1 and DjB4, and on the current view regarding the interaction of Hsp40 from subfamilies A and B with Hsp70 (21,36,58,59,61). In this model, the interaction between DjA1 and Hsp70 may occur by two binding sites located in each J-domain, one interacting with the Hsp70 NBD and the other with a region close to the Hsp70 substrate binding site (21). In the interaction model for DjA1-Hsp70, the Cys-rich domain is in a proper position to interact with Hsp70 and to transfer substrate to it, as proposed by Linke et al. (36). In the interaction model for DjB4-Hsp70, the binding site located in one J-domain interacts with the Hsp70 NBD, and the C terminus of DjB4 is able to interact with the C terminus of Hsp70, as previously suggested (58,59). The size of DjB4 allows the suggested double interaction between the Jdomain and the Hsp70 NBD and between the Hsp40 C terminus and the Hsp70 C terminus to occur in one cycle.
The data presented here are in good agreement with the current view of the interaction between Hsp40 and Hsp70 described above. The conformational arrangement of the Jdomain and G/F-rich region found for DjA1 and DjB4 may confer flexibility to these proteins, allowing them to accommodate and interact with both nucleotide and substrate binding domains of Hsp70 in the same cycle of interaction (see discussion above). Low or high resolution structural studies for other Hsp40 from subfamilies A and B would provide more information about the interactions between Hsp70-Hsp40 partners. Finally, our results shed light on several structural aspects of DjA1 and DjB4 that are crucial to the intracellular chaperone function of these proteins. CONCLUSION Small angle x-ray scattering experiments yielded low resolution structural models (molecular envelopes) for the proteins DjA1 and DjB4. A low resolution model for the structure of a C-terminal-deleted mutant of DjA1, DjA1-(1-332), in which there is no dimer formation, was also determined. The hydrodynamic properties calculated from these protein models agreed well with those determined directly from AUC experiments. This agreement supported and confirmed the structural models derived from the SAXS results. The characteristics of the low resolution models and the values of their molecular masses indicated that DjA1-(1-332) was a monomer whereas DjA1 and DjB4 were both dimers. An additional analysis combining the envelope functions derived from SAXS data, and the known information about the high resolution structure of the Hsp40 C-terminal and J-domains showed that DjA1 had a rather compact structure in which two highly asymmetrical monomers bound in a bullet shape, while in DjB4 the N termini had a completely different arrangement in which they pointed away from the compact C termini core responsible for dimerization. These differences may explain why these two Hsp40 subfamilies have different chaperone activities and contribute to our understanding of the role of Hsp40 in substrate binding, transport, and interaction with Hsp70 proteins.