The N Terminus of Saccharomyces cerevisiae Msh6 Is an Unstructured Tether to PCNA

The Msh6 NTR Is Hypersensitive to Proteolysis

We initially identified the NTR/mispair binding domain (MBD) boundary by partial trypsin digestion. In titration experiments, the 140 kDa Msh6 was digested to a 112 kDa fragment under conditions where Msh2 was resistant to proteolysis (Figure 1, upper arrow). Mass spectrometric analysis revealed that this Msh6 product was an N-terminal truncation; the most N-terminal peptide consistently identified by mass spectrometry included residues 252–265, whereas peptides including the C-terminal residue of Msh6 were recovered. No smaller proteolysis products were visible in the gel, indicating extensive digestion of the NTR. The next stable Msh6 proteolysis product (Figure 1, lower arrow) was an N-terminal truncation of the first 441 residues, corresponding to loss of the MBD. Thus, the Msh6 NTR is hypersensitive to cleavage by trypsin.

N-Terminal Deletions of Msh6 Define Two Mutation Classes

The Msh6 NTR contains a canonical PIP box (27-QSSLLSFF-34). Mutation of Phe33 and Phe34 to Ala (msh6 F33F34AA) eliminates coimmunoprecipitation of PCNA but only causes a moderate MMR defect (Flores-Rozas et al., 2000). As residues surrounding the PIP box also contact PCNA in other proteins (Bruning and Shamoo, 2004; Chapados et al., 2004; Gulbis et al., 1996; Kontopidis et al., 2005; Matsumiya et al., 2002; Sakurai et al., 2005), we constructed a series of N-terminal deletions to investigate the contribution of residues surrounding the PIP box as well as the role of the rest of the NTR in vivo. Deletions were tested for complementation of the high mutation rates of an MMR-defective msh3Δ msh6Δ double mutant strain by expression from the native MSH6 promoter on a low copy number plasmid.

Short N-terminal deletions (msh6Δ2-50 and msh6Δ2-123) caused modestly increased mutation rates equivalent to the PIP box mutation msh6 F33F34AA (Table 1; p > 0.1 for all pairwise comparisons, Mann-Whitney test). In contrast, the msh6Δ2-294 allele had a null phenotype indistinguishable from the empty vector control (p > 0.2); GFP tagging this mutant at its C terminus to assess expression by western blotting revealed that it was not present at detectable levels (data not shown). Intermediate deletions (msh6Δ2-143, msh6Δ2-163, msh6Δ2-183, msh6Δ2-201, msh6Δ2-251, and msh6Δ2-271) caused mutation rates substantially higher than msh6 F33F34AA, msh6Δ2-50, and msh6Δ2-123 (p < 0.0001 for all comparisons between the two groups) but lower than the empty vector in at least one assay. Western blot analysis of a GFP-tagged version of msh6Δ2-251 showed that it was present at higher levels than tagged wild-type MSH6, indicating that the defect in this mutant was not due to protein expression or stability problems (data not shown). Thus, mutants fell into two groups: the msh6 F33F34AA, msh6Δ2-50, and msh6Δ2-123 mutants had minor MMR defects, while deletions of 143 or more codons caused substantial MMR defects.

The increased mutation rates caused by deletions extending past codon 123 suggested that the region between residues 123 and 251 has a role in Msh6 function (Table 1). The internal deletion allele msh6Δ51-251, in which PIP box is retained, complemented the msh3Δ msh6Δ strain with only minor defects like the msh6 F33F34AA allele. Thus, the 51–251 region is only important in the absence of residues 2–50. Interestingly, combining the internal deletion with the PIP box mutations resulted in an allele (msh6Δ51-251 F33F34AA) that had a more serious defect than either msh6Δ51-251 or msh6 F33F34AA alone but still complemented better than the N-terminal deletions that extend past residue 123 (p < 0.006 for all comparisons). Given that there are residues in addition to F33 and F34 that participate in PCNA-PIP box interactions (Gulbis et al., 1996), we suspect that the functionality of this mutant may be due to residual interaction with PCNA (see below).

In contrast to the mostly functional msh6Δ51-251 allele, deletion of codons 51–271 (msh6Δ51-271) caused a dramatic increase in mutation rate. To assess whether there was an important region between residues 251 and 271, we made an msh6Δ252-271 mutant, which was similar to wild-type MSH6 (Table 1). Together with results described below, this suggests that the defect caused by msh6Δ51-271 is due to the NTR being too short for productive interaction with PCNA. In contrast to the msh6Δ252-271 allele, deletion of codons 252–284, 252–294, or 252–304 caused severe defects (Table 1 and Figure S2); taken together with the NTR swap experiment described below, this suggests that the boundary between the NTR and the MBD probably lies between residues 272 and 284.

Deletion of the N-Terminal 123 Residues of Msh6 Completely Abrogates Interactions with PCNA

To assess the possibility that the mildness of the defects caused by msh6 F33F34AA, msh6Δ2-50, and msh6Δ2-123 was due to residual interaction of the encoded proteins with PCNA, we tested the binding of purified Msh2-Msh6 F33F34AA and Msh2-Msh6Δ2-123 to immobilized PCNA by surface plasmon resonance. Wild-type Msh2-Msh6 bound PCNA with a K_D of 2.5 nM (Figure S3), consistent with published reports (Lee and Alani, 2006). In contrast, Msh2-Msh6 F33F34AA had greatly reduced binding, and Msh2-Msh6Δ2-123 lacked measurable interaction. Thus, complementation by the alleles encoding these proteins is not due to significant interactions with PCNA, although the low level of interaction seen with the F33F34AA mutant may allow partial function in the context of the msh6Δ51-251 F33F34AA mutant described above.

The Msh6 NTR Is Not Required for Nuclear Localization, DNA Binding, or Mispair Discrimination

To ensure that MMR-defective NTR deletions do not disrupt Msh6 localization, we examined an msh3Δ msh6Δ strain expressing GFP-tagged wild-type Msh6 or Msh6Δ2-251 by fluorescence microscopy. The tagged wild-type Msh6 was functional for MMR in vivo, and both proteins were localized to the nucleus, in contrast to GFP expressed alone from the Msh6 promoter, which was cytoplasmic (data not shown; localization of GFP-tagged MSH6 was initially described by Huh et al. [2003] and http://yeastgfp.ucsf.edu); the defect of Msh6Δ2-251 is therefore not due to loss of nuclear localization. Furthermore, purified wild-type Msh2-Msh6 and Msh2-Msh6Δ2-251 complexes bound similarly to DNA substrates containing a GT mispair or a +1 insertion, and both complexes discriminated mispairs from fully base paired DNA (Figure S4) and dissociated from DNA upon addition of ATP. The defects caused by deletion of residues 2–251 are therefore not due to failure to recognize DNA or discriminate between mispairs and homoduplex DNA.

The Msh3 and Msh6 NTRs Are Interchangeable, and the Msh6 NTR Can Function When Placed on Msh2

Both Msh3 and Msh6 have NTRs with N-terminal PIP boxes. A chimeric Msh3 in which the first 154 residues were replaced with residues 1–304 of Msh6 complemented the msh3Δ msh6Δ strain nearly as well as wild-type MSH3 (Table 2). Similarly, a chimeric Msh6 in which the first 251 residues were replaced with residues 1–154 of Msh3 was functional, as was a version in which the first 271 residues were replaced with 1–154 of Msh3. However, shifting the fusion position in the Msh6 chimera to residue 285, 295, or 305 resulted in strongly MMR-defective phenotypes (Table 2 and Figure S2). Combined with our partial proteolysis and deletion results, the properties of these chimeric constructs suggest that Msh6 residues between 272 and 304 may stabilize Msh6 or be part of the MBD. The functional chimeras demonstrate that the Msh3 and Msh6 NTRs perform similar functions, consistent with the NTRs lacking (or functionally replacing) interactions with the rest of the Msh2-Msh3 and Msh2-Msh6 heterodimers.

If the primary role of the Msh6 NTR is to localize Msh2-Msh6 within the nucleus, an Msh6 NTR-Msh2 fusion might restore function to a complex containing Msh6Δ2-251. A chimeric MSH2 gene encoding residues 1–251 of Msh6 fused onto its N terminus caused a slight increase in mutation rate relative to wild-type MSH2 when expressed from a plasmid along with wild-type MSH6 in an msh2Δ msh3Δ msh6Δ strain, but significantly decreased the mutation rate of a strain expressing Msh6Δ2-251 (Table 2). Thus, the Msh6 NTR is functional when fused to either Msh2 or Msh6.

The Msh6 NTR Is an Unstructured Peptide that Can Interact with PCNA Trimers at a 3:1 Ratio

We cloned and expressed the first 304 residues of Msh6 for biochemical and biophysical characterization. A similar Msh3 construct (residues 1–154) expressed poorly and was not studied further. The isolated Msh6 NTR (34 kDa) showed similar trypsin sensitivity to the NTR in the Msh2-Msh6 heterodimer (Figure 1). It also eluted in size-exclusion chromatography earlier (Stokes radius, 45.7 Å) than predicted for a compact globular protein (Stokes radius, 26.7 Å; Figure 2A), consistent with oligomerization or an extended conformation. The NTR bound PCNA and formed complexes that eluted at one of three retention volumes prior to PCNA or NTR alone (Figure 2B), with the earlier eluting species forming at the higher molar ratios of NTR to PCNA. Thus, each PCNA trimer bound one, two, or three Msh6 NTR molecules.

Statistical analysis of the amino acid composition of the NTR showed that hydrophilic Ser, Thr, and Asn residues were overrepresented and hydrophobic Val, Ile, and Leu residues were underrepresented (in the top and bottom 95% quantiles of all S. cerevisiae proteins, respectively; Brendel et al. [1992]). This amino acid bias is characteristic of intrinsic disorder (Romero et al., 2001; Tompa, 2002), consistent with the protease accessibility and aberrant size exclusion chromatography. Additionally, boiling PCNA for 10 min resulted in formation of a large aggregate (Figure 2C), but the elution profile of NTR was unchanged by boiling (Figure 2D).

SAXS can directly monitor protein folding and solution conformations, so we collected X-ray scattering data for the NTR, PCNA, and the PCNA-NTR complex (Figure 3A). Consistent with gel filtration results, the radius of gyration (R_g) for the NTR was much larger than that observed for similarly sized globular proteins, and the NTR-PCNA complex was larger than PCNA or the NTR alone (Figure 3B).

The Kratky plot measures the overall compactness of proteins. Folded proteins follow Porod's law (I[S] is proportional to S⁻⁴, where S is reciprocal resolution) and have a bell-shaped curve as seen for both PCNA and the complex (Figure 3C). In contrast, the NTR curve was flat, similar to proteins denatured by chemicals or heat (Doniach, 2001). Thus, the NTR does not fold into a compact domain.

SAXS can provide structural information for unfolded polymers by fitting to Equation 1 (see the Experimental Procedures; Figure 3D). The polypeptide length of 1078 Å matched the expected length of 1091 Å (304 residues × 3.78 Å/residue × 0.95, where 0.95 is a factor due to nonlinearity of an extended polypeptide chain [Cantor and Schimmel, 1980]). The fitted parameter b = 18.7 Å agrees well with the observed range of 19–25 Å (Perez et al., 2001), corresponding to a persistence length, the length for which the polypeptide tends to act as a rigid unit, of 9.35 Å or roughly three amino acids. The R_g of 57.2 Å calculated from these parameters (Equation 2, see Experimental Procedures) is quite close to the observed R_g of 56 ± 2 Å calculated by the inverse Fourier transform method (Figure 3B).

PCNA Does Not Induce Substantial Structure in the Msh6 NTR

X-ray scattering of the PCNA-NTR complex suggests that PCNA binding does not induce substantial structure, as can be inferred by comparing the pair distribution functions (P[r]s) calculated from the scattering curves for all of the samples (Figure 3E). The P(r) function is a histogram of all distances within the scattering particle and can be calculated from atomic models. The overall shape of the PCNA P(r) function was similar to those observed for well-folded proteins, and the double peak was due to the presence of a hole in the center of the PCNA ring, which could be reconstructed from the SAXS curve by ab initio calculations (Figure 3F). The “tail” of the P(r) function for the PCNA-NTR complex (r > 100 Å) was consistent with the presence of a long unstructured addition to PCNA. The maximum intraparticle distance in the P(r) function (D_max, where the P[r] function returns to zero) of the PCNA-NTR was 250 Å. The free NTR and free PCNA, on the other hand, had maxima of only 185 and 89 Å, respectively. The difference between the D_max of the complex and the D_max of the NTR, 70 Å, was almost the same as the D_max of PCNA alone. In fact, the P(r) function of the complex closely resembled that of the NTR once the contributions from PCNA, including the double peak with the maximum at 55 Å and the shift of the scattering from the NTR tail by 70 Å, were excluded (Figure 3E). These data indicate that binding PCNA did not dramatically change the overall unfolded nature of most of the length of the NTR.

These SAXS results were supported by partial proteolysis experiments in which trypsin was added to NTR-PCNA complexes or NTR mixed with equivalent amounts of BSA (Figure 4A). The NTR was cleaved by a concentration of trypsin too low to have any visible effect on PCNA, consistent with the results seen for the NTR of full-length Msh6 described above (Figure 1) and with the observed lack of secondary structure. Furthermore, PCNA did not affect the cleavage of the NTR, consistent with the NTR binding PCNA primarily via the PIP box.

Solution Structure of the Msh2-Msh6-PCNA Complex

Since the Msh6 NTR lacks structure alone, when bound to PCNA, or in the context of full-length Msh2-Msh6, we tested the protease sensitivity of the NTR in the Msh2-Msh6-PCNA complex (Figure 4B). PCNA conferred only a very slight degree of protection to the NTR relative to BSA controls and did not alter the proteolytic products produced. The lack of protection is consistent with the NTR remaining unstructured, proteolytically accessible, and not having significant regions of interaction with PCNA outside of the PIP box.

An unstructured NTR could be consistent with Msh2-Msh6-PCNA complexes where folded domains of Msh2-Msh6 and PCNA make direct contacts and the NTR is looped out. To assess this possibility, we collected SAXS data on full-length Msh2-Msh6 alone and in complex with PCNA (Figure 4C). The Msh2-Msh6-PCNA complex was detected by size-exclusion chromatography (data not shown); however, the P(r) function of Msh2-Msh6-PCNA complex differed only subtly from Msh2-Msh6 alone, with the long extended tail (r > 100 Å) in the P(r) function of the complex having a stronger signal. These results are inconsistent with a substantial population of complexes having close contacts between the folded domains of Msh2-Msh6 and PCNA. However, these results could be explained if the domains are connected by with a flexible, ∼1000 Å long NTR, which would average out the positions of the intermolecular vectors between the folded domains of PCNA and Msh2-Msh6.

To confirm this hypothesis, we used the deletion mutant Msh6Δ51-251 to shorten the tether to PCNA and decrease the NTR flexibility, thereby decreasing the conformations that average out the intermolecular components in the P(r) function. We collected SAXS data on Msh2-Msh6Δ51-251 alone and in complex with PCNA (Figure 4D). As predicted, the shorter tethering of the Msh2-Msh6Δ51-251-PCNA complex resulted in larger differences in the tail of the P(r) function when comparing the complex to Msh2-Msh6Δ-51-251 alone. Moreover, SAXS data collected on the Msh2-Msh6Δ51-251 F33F34AA complex with and without PCNA demonstrated that these changes were lost when the Msh6-PCNA interaction was eliminated (Figure 4D). Calculation of the P(r) curves using atomic models of the complex confirmed that no single conformation could account for the observed scattering (Figure 4E), while the average of the scattering from many complexes might approximate the observed data. Thus, in the absence of DNA and MutL homologs, the NTR is a flexible molecular leash between Msh2-Msh6 and PCNA.

Most Functional PIP Boxes in S. cerevisiae Proteins Are Located in Disordered Regions

Our results show that the PIP box-containing NTR of Msh6 is unstructured. Long-range disorder predictions by IUPRED (Dosztanyi et al., 2005) for yeast (Figure 5A) and human (Figure S5) Msh2, Msh3, and Msh6 are consistent with our experimental results and indicate that Msh6 and Msh3 have unstructured NTRs, whereas Msh2 does not. To determine how representative these results are of other functional PIP boxes, we investigated the predicted flexibility of PIP box sequences (Figure 5B). We calculated the IUPRED long-range disorder parameters for all S. cerevisiae proteins with the consensus PIP box sequence Qxx[LIV]xx[FY][FY] as well as known PCNA-interacting proteins with noncanonical PIP boxes such as Rfc1. A PIP box disorder score was assigned as the average disorder tendency of the PIP box and ten residues surrounding it on both sides (if present). The scores were divided into three groups: proteins that bind PCNA, proteins that do not bind PCNA, and proteins with unknown binding status (Burgers, 1991; Clark et al., 2000; Flores-Rozas et al., 2000; Gary et al., 1997; Haracska et al., 2001; Li et al., 1995; Otterlei et al., 1999; Schmidt et al., 2002; Subramanian et al., 2005; Unk et al., 2002; Zhang et al., 2000). The median disorder tendency for the 13 known PCNA binding proteins was high, at 0.4886. This IUPRED prediction suggests that the PIP boxes of many PCNA binding proteins lie in flexible tails, sometimes of substantial length, and are strongly reminiscent of Msh6 and Msh3 (Figure S6). These disorder tendencies were different (p < 0.001, Mann-Whitney test) from those of PIP box regions of proteins known not to bind to PCNA (median 0.0827) and those with unknown binding status (median 0.0836), which were indistinguishable from each other (p > 0.75).

Strains and Plasmids

The msh2Δ msh3Δ msh6Δ S. cerevisiae strain RDKY5963 was made by disrupting MSH2 with HIS3 in RDKY4234 (MATa ura3-52 leu2Δ1 trp1Δ63 his3Δ200 hom3-10 lys2-10A msh3::hisG msh6::hisG). Strains were propagated in YPD or complete synthetic medium (CSM) lacking specified amino acids to select for plasmid markers. Deletion and domain swap alleles were made from MSH2-, MSH3-, or MSH6-bearing plasmids by recombination of linear fragments in yeast or by PCR with primers diverging from the region to be deleted and containing homology tails to allow reassembly in E. coli (Table S1). Point mutations were made by site-directed mutagenesis.

Genetic Assays

MSH2, MSH3, and MSH6 alleles were tested for complementation of RDKY4234 or RDKY5963 by fluctuation analysis (Lea and Coulson, 1948) to determine rates of hom3-10 and lys2-10A reversion (Marsischky et al., 1996; Tran et al., 1997) and CAN1 inactivation. For each strain, at least 14 independent 10 ml cultures were grown in CSM –Leu or CSM –Leu –Ura for 24–30 hr, harvested by centrifugation, washed, resuspended in water, serially diluted, and plated on CSM missing Leu or Leu and Ura for plasmid selection and Thr, Lys, or Arg +60 μg/ml canavanine (Sigma, St. Louis, MO). P values were calculated by two-tailed Mann-Whitney test (http://faculty.vassar.edu/lowry/utest.html).

Protein Overexpression and Purification

The his-tagged NTR expression vector pRDK1136 was constructed by inserting codons 1–304 of S. cerevisiae MSH6 between the NcoI and XhoI sites of pET28b, resulting in the addition of Ala-Leu-Glu-6His after residue 304. Protein was expressed in the E. coli Rosetta(DE3) strain (Novagen, Madison, WI) and purified by sequential chromatography on Ni-NTA (Qiagen, Valencia, CA) and Source 30S (Amersham, Piscataway, NJ) columns (see the Supplemental Data). Wild-type and mutant Msh2-Msh6 complexes and PCNA were purified using modifications of previously published protocols (see the Supplemental Data).

Analytical Gel Filtration

Analytical gel filtration was performed on a SMART machine (Pharmacia) using a Superose 6 column and buffer containing 25 mM Tris (pH 8), 0.1 mM EDTA, 1 mM DTT, and 300 mM NaCl. PCNA-NTR complexes were generated by mixing various molar ratios of the proteins followed by incubation on ice for 1 hr. Stokes radii were calculated with a linear regression of retention time versus Stokes radius using standards as follows: bovine serum albumin (35.5 Å), aldolase (48.1 Å), ferritin (61.0 Å), and thyroglobulin (85.0 Å).

Partial Proteolysis and Mass Spectrometry

Partial proteolysis was carried out in 10 μl reactions containing 1 μg Msh6 NTR or Msh2-Msh6 in 25 mM Tris (pH 8), 100 mM NaCl, 10% glycerol, and 5 mM MgCl₂. When indicated, Msh6 NTR and Msh2-Msh6 were premixed with 2.5 μg and 1 μg, respectively, of PCNA or BSA (Sigma) and incubated for 1 hr on ice before adding trypsin as indicated and shifting the reactions to room temperature for 1 hr before stopping them by adding PMSF to 10 mM. Western blotting was performed with rabbit polyclonal antibodies to S. cerevisiae Msh2 or Msh6 raised in this laboratory. For mass spectrometry analysis, individual bands were excised from silver-stained gels, cut into pieces, and soaked in acetonitrile for 30 min before drying. Gel pieces were rehydrated in 50 mM NH₄HCO₃ containing 7 ng/μL trypsin, incubated overnight at 37°C, and then washed three times for 30 min each with 2% acetic acid 50% acetonitrile. The wash liquid was dried and peptides resuspended in 2% acetic acid and purified by ZipTip (Millipore, Billerica, MA) before analysis by mass spectrometry (Zhou et al., 2004).

SAXS Data Collection and Processing

SAXS data were collected at beamline 12.3.1 at the Advanced Light Source at Lawrence Berkeley National Laboratories by using a silver behenate calibration standard (diffraction maxima 0.107633 and 0.215266 Å⁻¹). Buffer and sample were collected alternately with short exposures bracketing a longer exposure, typically 10 and 100 s in length. Short exposures were compared to ensure that no radiation damage had occurred, and data were merged with the program PRIMUS. Msh2-Msh6 and Msh2-Msh6-PCNA complexes were dialyzed into the following buffer for all SAXS experiments: 50 mM Tris (pH 8.0), 0.5 mM EDTA, 1% glycerol, 5 mM DTT, and 200 mM NaCl. Msh6 NTR and NTR-PCNA complexes were dialyzed into buffer consisting of 25 mM Tris (pH 8), 0.1 mM EDTA, 1 mM DTT, and 100 mM NaCl. For each sample, data were collected at several different protein concentrations, and the scattering was fit by the indirect Fourier transform method (Moore, 1980) as implemented in GNOM (Svergun et al., 1988).

For unfolded polymers with a persistence length of b/2 and a length of L, the observed scattering in the range S < 3/b has the following functional form (Brulet et al., 1996):

where y = S² Lb/6, I(0) = the scattering intensity at zero angle (forward transfer momentum), S = the reciprocal resolution, and R_c = the radius of gyration of the cross-section. Scattering curves from Msh6 NTR were fit to this equation in the S range of [0 Å⁻¹, 0.115 Å⁻¹] using gnuplot (Williams et al., 1990).

Using these parameters, the expected R_g for the unfolded polypeptide can be calculated with the following equation:

where y = L/b.

Computational Msh2-Msh6-PCNA Complex Modeling

Models of bacterial MutS bound via a flexible tether to the PIP box binding site in PCNA were generated by first generating the Cα positions of 304 residue random polypeptide. The geometry of the Cα trace was constrained through random sampling of pseudoangles and pseudotorsion angles (Oldfield and Hubbard, 1994) in five degree increments weighted by their frequency in Nh3D version 3.0, a non-homologous subset of the protein database (Thiruv et al., 2005). Cα-Cα distances of 3.80 Å in adjacent residues and pseudosteric Cα-Cα clashes of 5.0 Å used during the generation of the random polypeptide were derived from statistics calculated from the Nh3D database. The final residues of the p21 PIP box sequence in the PCNA-p21 structure (Protein Data Bank [PDB] ID 1axc; Gulbis et al. [1996]) were superimposed with the first residues of the polypeptide and the first residues of one of the MutS subunits (PDB ID 1fw6; Junop et al. [2001]) were superimposed with the last residues of the polypeptide. The superimposed PCNA and MutS structures were rotated about the overlapping Cα-Cα bond and accepted if conformations with less than 10 Cα-Cα contacts closer than 4.0 Å were obtained.