The C-terminal domain of transcription co-activator PC4, PC4_CTD, is a potent ssDNA binding domain (K_d up to 0.7 nM) [Werten et al. 1997, Kretschmar et al. 1994, Ge and Roeder 1994]. PC4_CTD preferentially binds two strands of ssDNA running in opposite directions, e.g. internally melted duplexes. The high resolution native crystal structure of PC4_CTD shows that PC4_CTD molecules form intimate dimer structures containing two grooves, one in each monomer, running perpendicular over a curved floor of anti-parallel b-strands [Brandsen et al. 1997]. The grooves are separated by a ridge, termed b-ridge, at the interface between the two monomers comprising a dimer. The grooves are lined by aromatic and positively charged residues and were shown to be reminiscent of the ssDNA binding channels in replication protein A (RPA) [Bochkarev et al. 1997]. Based on this striking resemblance, the PC4_CTD grooves were predicted to interact with ssDNA by a mechanism similar to that observed in the RPA-ssDNA crystals [Brandsen et al. 1997]. Two ssDNA binding grooves are observed in a PC4_CTD dimer. These grooves run in opposite directions and explain fully the preferential binding of heteroduplex or "bubble" DNA.

Two structures of single stranded DNA binding protein (ssDBP) in complex with ssDNA have been solved to date by X-ray crystallography: the replication fork single stranded DNA binding protein (T4 gp32) [Shamoo 1995] and the single stranded DNA binding domain of replication protein A (RPA) [Bochkarev 1997]. Only in the RPA crystal structure the ssDNA was clearly visible and allowed a detailed analysis of residues involved in complex formation. Protein-ssDNA interactions consisted of electrostatic contacts between the phosphate backbone and positively charged residues, and stacking and hydrogen bond interactions between DNA bases and aromatic residues of the protein.

To study the binding of ssDNA to PC4_CTD in more detail, crystallographic studies were performed on PC4_CTD in complex with oligonucleotides. Here we report the low-resolution structure of PC4_CTD in complex with the hexadeoxythymine, (dT)₆.

The C-terminal domain of PC4 was prepared by Werten [Werten et al. 1997] and concentrated to a solution containing 7 mg/ml PC4_CTD, 10 mM Hepes buffer (pH 7.0) and 200 mM sodium chloride. The oligonucleotides used in the co-crystallization experiments ranged in length from 6 to 16 (dT) bases and were synthesized in a Pharmacia Gene Assembler Plus and freeze dried from a DNA-water solution. The protein-DNA complexes were formed by incubating equimolar solutions at 4°C for 45 minutes. Crystals were grown at 4°C by vapor diffusion in hanging drops using several protein-DNA crystallization screens [Aggarwal 1990]. All crystallization experiments were performed by mixing an equal volume of reservoir solution with the protein-DNA solution. Crystalline precipitant was observed at several conditions with all constructs used. The (dT)₆-construct gave the most promising results judged from the size and shape of the crystals. Crystallization conditions similar to the native crystallization conditions, i.e. MPD (30-40% v/v), sodium chloride (100 mM) and acetate buffer (100 mM, pH 5.0) with additional calcium chloride (20-150 mM) yielded very small crystals of identical morphology to the native PC4_CTD crystals. Crystals of a different morphology, i.e. more bipyramidal like (figure 1), were crystallized using a reservoir solution containing 50 mM acetate buffer (pH 5.0), 15% w/v poly-ethylene glycol 3400 (PEG3400) and 500 mM sodium chloride. Formation and growth of these crystals proceeded very slowly. Diffraction data were collected from bipyramidal crystals (size 0.1 x 0.1 x 0.05 mm³) equilibrated for more than a year.

Figure 1: Photograph of a PC4_CTD-(dT)₆ complex crystal (middle right) showing a different morphology from the native PC4_CTD crystal (see chapter 5, figure 1). Dimensions of the crystal are about 0.1 x 0.1 x 0.05 mm³

For data collection at cryogenic temperature, crystals were transferred into a reservoir solution containing an additional 20% v/v PEG400. Diffraction data were collected from a single crystal maintained at 120 K on an in-house McScience DIP-2020 imaging plate mounted on a Nonius FR-571 rotating anode operated at 45kV and 95mA. The data were integrated with DENZO and intensities scaled with SCALEPACK [Otwinowski 1993]. Despite the difference in morphology with respect to the native PC4_CTD crystals, the complex crystals also belong to the P1 space group, though with slightly different cell dimensions: a = 49.3, b = 63.5, c = 68.3 Å, a = 96.2, b = 90.0, g = 90.1°, in comparison to a = 41.3, b = 67.8, c = 67.2 Å, a = 87.7, b = 84.4, g = 85.8° for the native crystals. The complex crystals did not diffract very well as can be seen from the resolution limit (3.5 Å) and the merging statistics (R_merge=13.6%). As was the case for the native data, a higher symmetry lattice could not be excluded in advance. A likely candidate is P2₁ with cell dimensions a = 63.5, b = 49.3, c = 68.3 Å and b = 96.2°. Reduction of the diffraction data in this space group gave rise to R_merge values of 16.6% compared to 13.6% for P1 and systematic absences exceeding 2s in intensity. Together with the fact that the structure could not be refined in this space group this indicates the incorrectness of the P2₁ space group. A summary of the data collection and reduction statistics is given in table 1.

 ^{Data completeness for the complete resolution range measured. Completeness for the high-resolution shell (3.58-3.50 Å) is given between parentheses.}

^{H R_merge = S | I(obs) - < I(obs) > | / S I(obs).}

^¶ For comparison the corresponding values for the native crystal are given in parentheses.

Initial positional refinement using the high resolution model coordinates failed and therefore the structure was solved by molecular replacement. The molecular replacement calculations were carried out with the AMORE program [Navaza 1994] using a complete PC4_CTD dimer search model derived from the high resolution native structure. Rotation and translation searches using diffraction data between 12 and 4 Å and an integration sphere of 21.6 Å (75% of the maximal distance from the center of mass) lead to a clear solution comprising four dimers with a correlation coefficient of 55.6% and an R-factor of 40.5% after rigid body refinement, table 2.

  a, b, g Eularian angles.

^‡ X, Y, Z fractional coordinates.

^¶ Accumulated correlation coefficient and R-factor in %.

^¥ Data after rigid body refinement as implemented in AMORE.

Inspection of the AMoRe solution using the graphics program O [Jones 1991] showed that the PC4_CTD-(dT)₆ complex forms a similar "tetramer of dimers" arrangement as was previously observed in the native PC4_CTD structure. The structure was refined with the REFMAC [Murshudov et al. 1997] program (version 3.2) using data between 20 and 3.5 Å. To allow cross validation of the refinement steps, the data set was divided into a working set (94% or 8921 reflections) and a test set (6% or 575 reflections). During the refinement strict NCS constraints were applied on a monomer, i.e. all monomers were kept identical. Anisotropic overall data scaling and bulk solvent correction were applied and the temperature factors were isotropically refined. REFMAC refinement without manual rebuilding reduced the R-factor to 32.4% and the R_free factor to 31.1%, table 3. The resulting phase set was used to calculate s_A-weighted electron density maps. Final maps were calculated by 8-fold averaging using the MAMA and AVE programs [Kleywegt and Jones 1994]. Visual inspection of these density maps did not reveal any region in the protein that needed manual rebuilding.

Protein structure

The overall conformation of the protein structure as well as the packing in the crystal are very similar to that of native PC4_CTD, figure 2. A detailed discussion of these is given in chapter 5, section 5.5. The root-mean-square coordinate difference after superpositioning PC4_CTD-dimers from the native and the complex structure on Ca-atoms is 0.73 Å, figure 2 b and c. The largest differences can be found in the loop regions between b2 and b3 (Gly 79 - Lys 80) and the loop region between b3 and b4' (Asp 91 - Lys 97).

The current model consists of all 66 amino acids and shows clear continuous electron density for the complete main chain. Some side chains only show limited density, as can be seen from corresponding high temperature factors along the peptide chain, figure 3 a and b. Both plots show the same trends as were previously found for the high resolution native structure (chapter 5, figure 12 a and b).

The Ramachandran plot of the model (data not shown) shows that all 56 non-glycine and non-proline residues have their (j ,y ) combinations in the allowed (76.8%) and additionally allowed (23.2%) region.

Figure 2 Comparison between native and PC4_CTD-DNA complex. (a) Stereo plot of the packing arrangement ("tetramer of dimers") found in native PC4_CTD (black) and the PC4_CTD-(dT)₆ complex (grey). (b) Ca overlay of both dimer structures. Native PC4_CTD shown in black and the PC4_CTD-(dT)₆ complex shown in grey. The area’s showing the largest differences, b2-b3 and b3-b4', are indicated. (c) Distance plot between corresponding Ca atoms in the native and complex structure after superpositioning. è

Figure 3 Plot of the B-factor distribution per-residue for (a) the main chain, (b) the side chain atoms.

DNA binding site

The final 2 |F_o| - |F_c| density map did not show any clear continuous density that would allow fitting of an atomic model for a (dT)₆ oligonucleotide. The |F_o| - |F_c| difference map (figure 4), however, showed large "blobs" of density in the quarter-pipe region, which could correspond to ssDNA.

Figure 4 Stereo plot of the averaged |F_o| - |F_c| difference density map around the quarter-pipe region showing the extra density present in this region.

Interestingly, most of the pieces of density lie in direct contact or in proximity of positively charged or aromatic residues: Tyr A71, Arg A75, Phe A77, Arg A86, Tyr A88, Trp B89, Arg B100 and Lys B101. Six of these residues, Arg A75, Phe A77, Arg A86, Tyr A88, Trp B89 and Arg B100 have been previously predicted to be involved in ssDNA binding from the superpositioning of PC4_CTD onto RPA [Brandsen et al. 1997]. Seven of these eight residues are either identical (5) or conserved (2) in the sequence of the homologous domain of the yeast transcriptional cofactor Sub1/Tsp1, only Tyr A71 is replaced by an arginine in yeast protein.

The low resolution crystal structure of the complex between PC4_CTD and (dT)₆ presented here does not provide sufficient structural information for a detailed description of the interaction between PC4_CTD and ssDNA. The structure as observed in the complex and native crystal are by-and-large identical. The major difference is the additional electron density located in the quarter pipe region observed in the |F_o| - |F_c| map of the complex. This density cannot be identified unambiguously.

The weak electron density for the DNA might be a result of the flexible way of binding ssDNA to PC4_CTD. The same flexibility was previously observed in both crystal structures of ssDBP-ssDNA complexes. In the T4 gp32 complex structure this disorder also resulted in only weak electron density for the ssDNA. In the RPA complex structure the two almost identical subdomains involved in DNA binding make considerably different contacts with the DNA, suggesting some plasticity in the binding of ssDNA.

This flexibility in the way of binding ssDNA is probably functionally relevant for this class of proteins. It allows them to bind ssDNA with little sequence specificity and to function simultaneously in a variety of different processes like transcription control, DNA replication, recombination and repair [Folmer et al. 1997]. At the same time this flexibility forms an obstacle to obtain suitable crystals for X-ray analysis, ratified by the fact that many different ssDBP structures have been solved, but only two in complex with ssDNA.

The weak electron density in the PC4_CTD complex structure might also be related to the (dT)₆ construct used in the co-crystallization. The binding cleft is thought to be able to accommodate eight ssDNA bases [Werten et al. 1997], thus allowing the six bases of (dT)₆ some (extra) freedom of motion.

From this low resolution study the protein does not seem to undergo major structural rearrangements upon binding of ssDNA. The presence and positioning of extra density is in full agreement with the previously predicted involvement of the PC4_CTD-groove in the binding ssDNA. The importance of the grooves for ssDNA binding was further confirmed by NMR studies and mutagenesis of residues predicted to interact with DNA (Werten, personal communication). The fact that the complex crystallizes in the same "tetramer of dimers" arrangement as the native structure precludes the binding of ssDNA in a heteroduplex fashion. Crystal contacts between successive dimers, i.e. the association of loop b3-b4' into a small intermolecular b-sheet (chapter5, section 5.5), physically prevent this.

The weak electron density observed in the PC4_CTD-(dT)₆ structure together with the flexible and non-standard nature of ssDNA do not allow to construct a detailed model for ssDNA bound to PC4_CTD. The data are very suggestive, however, as they are in full agreement with the general idea on protein-DNA interactions for this class of ssDNA binding proteins[Suck 1997]. They predict simultaneous involvement of aromatic and positively charged residues which are exposed on b-sheet surfaces; aromatic residues which can participate in hydrophobic contacts and base-stacking with DNA bases, and charged residues which can be involved in electrostatic interactions with the phosphate backbone.

More detailed structural information on the interacting residues and for example the directionality of the ssDNA when bound to PC4_CTD awaits further high resolution studies from better diffracting crystals.