Rett Syndrome-causing Mutations in Human MeCP2 Result in Diverse Structural Changes That Impact Folding and DNA Interactions*

Most cases of Rett syndrome (RTT) are caused by mutations in the methylated DNA-binding protein, MeCP2. Here, we have shown that frequent RTT-causing missense mutations (R106W, R133C, F155S, T158M) located in the methylated DNA-binding domain (MBD) of MeCP2 have profound and diverse effects on its structure, stability, and DNA-binding properties. Fluorescence spectroscopy, which reports on the single tryptophan in the MBD, indicated that this residue is strongly protected from the aqueous environment in the wild type but is more exposed in the R133C and F155S mutations. In the mutant proteins R133C, F155S, and T158M, the thermal stability of the domain was strongly reduced. Thermal stability of the wild-type protein was increased in the presence of unmethylated DNA and was further enhanced by DNA methylation. DNA-induced thermal stability was also seen, but to a lesser extent, in each of the mutant proteins. Circular dichroism (CD) of the MBD revealed differences in the secondary structure of the four mutants. Upon binding to methylated DNA, the wild type showed a subtle but reproducible increase in α-helical structure, whereas the F155S and R106W did not acquire secondary structure with DNA. Each of the mutant proteins studied is unique in terms of the properties of the MBD and the structural changes induced by DNA binding. For each mutation, we examined the extent to which the magnitude of these differences correlated with the severity of RTT patient symptoms.

A key epigenetic signal in vertebrates is the symmetrical methylation of CpG dinucleotides, which may be passed on to subsequent generations by the action of hemi-methylases on newly replicated DNA (reviewed in Ref. 1). Screening for proteins that bind preferentially to methylated CpGs has revealed a family of methylated DNA-binding proteins, the founding member of which is the conserved and highly basic 52-kDa methylated DNA-binding protein 2, MeCP2 2 (reviewed in Ref. 2). The portion of MeCP2 responsible for binding methylated DNA is known as the MBD (methylated DNA-binding domain), which extends from residues ϳ75 to ϳ164 (3). NMR and x-ray studies (4,5) have shown the MBD to be ϳ60% structured, with segments of ␣-helix, ␤-strand, and ␤-turn forming a wedge-shaped structure (Fig. 1b). In contrast, the N-and C-terminal portions of MeCP2 are predicted to be largely unstructured (6).
Signals encoded in methylated CpGs frequently lead to transcriptional repression, which appears to be a prominent consequence of MeCP2 binding (7). One model of the mechanism that leads from MeCP2 binding to transcriptional repression involves the recruitment of Sin 3A and histone deacetylase followed by local histone modification (8,9). However, recent evidence suggests that MeCP2 locations in chromatin are not confined to sites of methylated DNA and that MeCP2 occupancy does not necessarily lead to transcriptional repression (10). Moreover, it is now clear that MeCP2 has a wide range of potential functions (reviewed in Refs. 11 and 12), including an involvement in RNA processing (13) and a regulatory role in several human cancers (14 -16). MeCP2 misregulation has also been found associated with autism spectrum disorders (17). Our previous work on the in vitro interactions between MeCP2 and chromatin suggests that it is a potent inducer of compaction and thus may contribute to transcriptional repression via conformational changes to chromatin (18 -21).
An important landmark in understanding the in vivo role of MeCP2 was the finding that most cases of Rett syndrome (RTT), a severe X-linked neurodevelopmental disorder in humans, are attributable to sporadic mutations in MeCP2 (22). Females hemizygous for mutated MeCP2 develop normally for 6 -9 months and then show progressive neurological dysfunctions that appear to be associated with a loss of dendritic complexity and reduction in brain size (reviewed in Ref. 23). There is a wide variation in the range and severity of RTT symptoms, which appears to be related, at least in part, to the specific mutation present and to local X inactivation skewing. MeCP2 mutations in males produce a range of symptoms ranging from mild mental retardation to severe neonatal encephalopathy (reviewed in Ref. 24). Further, deletion of MeCP2 in mice gives rise to RTT-like symptoms, which can be reversed upon expression of an introduced copy of the MeCP2 gene (25). The clear association between MeCP2 and neurological diseases provides an opportunity to understand the mechanism by which MeCP2 selectively interacts with the genome and transmits methylated DNA signals.
The majority of RTT cases result either from C-terminal truncations of varying length (ϳ40% of cases) or missense mutations within the MBD (ϳ45% of cases). Common MBD mutations include T158M (ϳ10% of patients), R106W (ϳ4%), and R133C (ϳ4%) (26,27) (www.mecp2.org.uk/). The x-ray structure of an MeCP2 MBD-DNA complex reveals that Arg-133 is involved in the DNA interaction surface (5), a result consistent with an earlier NMR study using the homologous domain of the methyl-binding MBD1 protein (28). Other frequent RTTcausing MBD mutations are not directly associated with the DNA interaction interface, and the causes of their loss of function appear to involve changes in inter-residue interactions (5).
Our recent studies of the interactions between MeCP2 and chromatin (19,20) revealed multiple DNA binding sites that extend beyond the MBD and involve both methylated and unmethylated DNA. We have also shown that different RTTcausing mutations have a surprisingly diverse impact on DNA and chromatin binding and compaction, suggesting that a comparative study of the effects of the different mutants on MeCP2 structure and substrate binding would be highly informative. We have therefore embarked on a detailed examination of the contributions of the different domains of wild type and mutant and the impact of DNA binding on the conformation of the protein. We chose first to study the MBD because the preponderance of missense mutations found in RTT patients occurs there. We selected four RTT-causing missense MBD mutations ( Fig. 1) for detailed study, including the three that occur most frequently. The impact of these mutations on the secondary and tertiary structure of the MBD and full-length protein were compared using circular dichroism (CD) and fluorescence spectroscopy. We show that each of these MBD mutations has unique structural consequences relevant to their loss of function. Within the intrinsically unstructured MeCP2 protein, protease-protected domains can be identified beyond the structured MBD (6), and an important finding of the present study is evidence of coupling between domains.
Fragment-(75-209) (MBD plus C-terminal flanking 45 amino acids) was constructed using the MBD forward primer and the following reverse primer: 5Ј-GTTAGAGAATTCCAC-CTGCACACCCTCTGACGTGGC-3Ј. The restriction fragments generated by double digestion of the amplicons with NdeI ϩ EcoRI, were then cloned into pTYB1 (New England Biolabs) vector by standard ligation procedures.
Fluorescence Spectroscopy-Fluorescence emission spectra were obtained using a PTI QM1 spectrofluorimeter over a 95 nm window from 305 to 400 nm using 2 nm emission and excitation slits with an integration time of 0.3 s. Three independent scans were averaged for each condition. For thermal unfolding, sample temperature was controlled using a Peltier unit, and samples were held at each temperature for 5 min before collection of data. Spectra were collected at 2.5°C intervals between 5 or 10°C and 85°C The midpoints of the melting transitions (T m ) were obtained from least square fits of plots of relative fluorescence intensity at max versus temperature and from the peaks of first derivative plots. Both approaches yielded very similar T m values. Relative fluorescence intensity was calculated by setting the intensity at the starting temperature to 100% and the ending temperature to 0% and interpolating values for intermediate temperatures.
Experiments using protein ϩ DNA were done in duplicate and, with protein only, in triplicate. Experiments with MeCP2-DNA complexes were carried out with protein:DNA molar ratios of 1:1 and 1:2 with protein concentrations between 2.5 and 5.0 M. Both input ratios gave essentially identical results.
Circular Dichroism-Proteins were first dialyzed extensively in 10 mM phosphate buffer, pH 7.6, containing 100 mM NaF. CD measurements were carried out at 22°C with a J715CD spectropolarimeter (Jasco Inc.) at a bandwidth of 1 nm and spectral resolution of 0.5 nm using a 0.1-cm path length stoppered quartz cuvette (Sterna Inc.). The scanning rate was 20 nm/s in continuous scanning mode with a response time of 4 s. Five spectra were collected per protein, averaged, and buffer-sub-tracted. The mean residue ellipticity [] (degrees cm 2 dmol Ϫ1 residue Ϫ1 ) was obtained by normalization of the raw data using where [] is the mean residue ellipticity (degrees cm 2 dmol Ϫ1 residue Ϫ1 ), obs is the ellipticity measured in degrees, M is the protein mean residue molecular weight, l is the optical path length of the cuvette in cm, and C is the concentration of the protein in mg/ml (29).
In some cases, the proportions of the various types of secondary structure were estimated using CONTINLL supported by DICHROWEB, an online server for protein secondary structure analyses (30,31). For estimations using LINCOMB, data were fitted to the Greenfield-Fasman standard curve set of 17 proteins using constrained least squares minimization as described (32). For estimations using CONTINLL (33,34), reference set 7 on Dichroweb, optimized for the wavelength range of 190 to 240 and containing 32 proteins from SELCON3, six from Ref. 33, five from Ref. 35, and five denatured, was used. The estimated composition of the MBD calculated using LIN-COMB, CONTINLL was also compared with the known NMRderived secondary structure of the MBD (Protein Data Bank code 1QK9 (4)) using STRIDE (36).
Molecular Modeling-Models of human MeCP2 MBD residues 77-165 were built and visualized with Insight II and Biopolymer software (Accelrys, San Diego, CA) and UCSF Chimera (University of California, San Francisco). The representative structure from the most highly populated ensemble cluster (37) of the NMR suite (4) (Protein Data Bank code 1QK9) was selected for this purpose. Amino acid substitutions made at the sites of the selected RTT mutations (R106W, R133C, F155S and T158M) were generated using the Richardson rotamer library (38) and energy-minimized with MMTK (Molecular Modelling Toolkit) (39) with the AMBER ff99 force fields (40). Solvent-accessible surface areas were calculated with GETAREA (41).

RTT-causing Mutations Alter the Structure, Stability, and DNA-binding Properties of MeCP2-Wild-type
MeCP2 has a single tryptophan at position 104 in the MBD, which is known from the NMR structure to be buried (4) and thus expected to have minimal solvent exposure. In a hydrophobic environment, tryptophan has a fluorescence emission maximum ( max ) of ϳ330 nm. Loss of hydrophobic packing and exposure to the aqueous environment, as occurs upon complete thermal melting, results in a red shift of max to ϳ355 nm accompanied by a JULY 18, 2008 • VOLUME 283 • NUMBER 29 large quenching of fluorescence intensity (42). Fig. 2a shows data collected at 25 and 75°C for the full-length wild-type MeCP2 and proteins with four RTT-causing MBD mutations ( Fig. 1), revealing the striking differences between the WT and the mutant proteins (Table 1). All except F155S and R106W had a max of ϳ328 nm, consistent with strong hydrophobic packing of Trp-104. As expected, R106W had the highest fluorescence intensity, because it has two tryptophans and a max of 330 nm. Of more interest was the significant quenching of fluorescence in R133C and F155S, most likely the result of increased solvent exposure. F155S also showed an increase in max to ϳ335 nm, suggesting that it is partially unfolded even at 25°C. Indeed, spectra of F155S recorded at 5°C showed a max close to the WT and a higher fluorescence intensity (not shown).

Mutation-induced Changes in MeCP2
When the proteins were fully "melted" at 75°C and Trp-104 maximally exposed, all showed a max of ϳ350 nm and a much reduced intensity (Fig. 2a). The residual intensity of the 2-tryptophan R106W mutant at 75°C was approximately twice that of the WT and other mutants (Fig. 2a). After heating to 85°C and recooling, all proteins recovered ϳ90% of the native fluorescence intensity at max (not shown), showing that heating did not induce aggregation and that the changes accompanying thermal melting were largely reversible.
To examine the thermal stability of the wild-type and mutant proteins, fluorescence emission was recorded at 2.5°C intervals from 5 to 85°C. Plots of relative fluorescence intensity versus temperature yielded sigmoidal curves (Fig. 2b) indicative of cooperative unfolding. Temperatures resulting in 50% unfolding (T m ) obtained both from least square square fits of these curves or from the peak of the first derivative (Fig. 3a, inset) were essentially identical.
Compared with WT MeCP2, the T158M, R133C, and F155S mutant proteins all showed a striking reduction in T m , indicating an increased susceptibility to thermally induced unfolding (Fig. 2b, Table 1). The mutants F155S and R106W also showed considerable broadening of the slope of the melting transition (Fig. 2b), suggesting a decrease in the cooperativity of unfolding.
To examine the effect of DNA binding on thermal, stability we selected a 45-bp segment of BDNF promoter DNA containing a single CpG and a nearby AT run known to bind MeCP2 in vivo (43,44) and that can be methylated as desired (45). By comparing thermal unfolding in the presence and absence of DNA, it was possible to determine the effect of DNA on the stability of the protein. When exposed to methylated DNA, the wild type showed a dramatic 18.6°C increase in T m (Fig. 3a). A smaller increase (11.2°C) in T m was also observed with unmethylated DNA (Fig. 3a), in agreement with earlier findings of MeCP2-DNA binding in the absence of CpG methylation (6,19). The R133C, T158M, and F155S mutant proteins showed much smaller methylated DNA-induced increases in T m as compared with the wild type (Fig. 3, b-d, Table 1). Thus, the binding of this small segment of DNA strongly stabilizes the wild-type MeCP2 against thermal melting but has a significantly weaker effect on the mutants, indicating a weaker binding affinity.
The Structure of the MBD Is Altered in RTT-causing Mutants-We next examined the structural characteristics of the MBD alone. As shown in Fig. 4d, the WT MBD had a max similar to that of the full-length protein but a 35% reduction in fluorescence intensity at max , suggesting a strong coupling between the MBD and other portions of MeCP2 (see below). The WT MBD had a T m almost identical to the full-length   Fig.  4a), but the spectrum of the R133C mutation was almost identical to that of the WT MBD. Interestingly, molecular modeling of the MBD (see "Experimental Procedures") predicted a substantially increased exposure of Trp-104 in the F155S mutation, but not for R133C or T158M, corresponding closely to the experimentally determined fluorescence emission values (Fig. 4a). All three mutant MBDs showed a reduction in thermal stability compared with the WT (Table 1, Fig. 4c) but to a smaller extent than their full-length counterparts (Fig. 2b).
To gain further insight into the structure of MeCP2, we employed circular dichroism, which yields characteristic signatures of the different forms of secondary structure. To identify the most appropriate deconvolution parameters for predicting MBD secondary structures, we compared the output of various CD deconvolution algorithms. LIN-COMB (32) and CONTINLL (33,34) both yielded very similar results; here we have shown only the CON-TINLL output. A test of the efficacy of deconvolution algorithms for a specific polypeptide is to compare the known NMR or x-ray structure with the CD deconvolution output. The published NMR structure of  Table 1). Error bars denote the standard error.

Mutation-induced Changes in MeCP2
the wild-type MBD (4) was analyzed with the STRIDE algorithm (36), which calculates the proportions of the different types of secondary structure from atomic coordinates. For the MBD, the actual secondary structure was almost identical to the CONTINLL output from our data, providing confidence in the utility of deconvolution and the choice of reference proteins for these data (see "Experimental Procedures"). Fig. 5a shows the striking differences in CD spectra between the largely unstructured full-length wild-type MeCP2 and the more structured MBD reported earlier (6). The full-length MeCP2 showed low ellipticity below ϳ210 nm and a pronounced negative trough at ϳ200 nm, indicative of a protein with substantial unstructured regions (46,47). Deconvolution predicted that the full-length WT MeCP2 is 65% unstructured, similar to the 59% reported earlier (6).
The full-length wild-type protein and the four full-length RTT-causing MeCP2 mutants gave essentially identical CD spectra (Fig. 5b). Because the CD signal represents the added contributions of all segments of the protein both within and beyond the MBD, local secondary structure changes are likely to be obscured in the full-length protein.
When the spectra of the MBDs alone were compared, there were marked changes between wild type and the mutants F155S and R106W (Fig. 5d). Inspection of the F155S spectrum revealed the signature depressions at 208 and 222 nm characteristic of ␣-helix, and deconvolution indicated an increased proportion of ␣-helix over the WT (from ϳ15 to ϳ22%) and concomitant decrease in ␤-strand (from ϳ27 to ϳ21%; Table 2). In contrast, R106W (Fig. 5d, Table 2) revealed the opposite tendency with an ϳ2-fold reduction in ␣-helix (from ϳ15 to ϳ9%) and increase in ␤-strand (from ϳ27 to ϳ33%). For R133C (Fig. 5e) and T158M (Fig. 5f), the CD spectra and deconvolution results were very similar to the wild type ( Table 2), indicating that these mutations lead to no significant changes in MBD secondary structure.
To investigate the effect of DNA binding on MBD secondary structure, we allowed the WT and mutant domains to interact with methylated or unmethylated DNA. The addition of protein to DNA (methylated as well as unmethylated) at input ratios resulting in the fully bound state did not alter the CD FIGURE 5. Circular dichroism of the MBD reveals marked differences between wild-type and mutant secondary structures in DNA-induced changes. a, compared with the full-length protein (white circles), the MBD (black circles) contains much more secondary structure. Full-length MeCP2 acquires secondary structure when bound to DNA (white squares). b, full-length WT and mutant MeCP2 give essentially identical CD spectra. c, addition of methylated DNA to the wild-type MBD (white circles) shows changes typical of acquisition of ␣-helix, whereas with unmethylated DNA (white squares) there is no increase at ϳ197 nm. Black-filled circles denote the WT MBD. d, the F55S MBD (white circles) and R106W MBD (half-black circles) show very different CD spectra from the wild-type MBD (black circles). e, addition of methylated DNA (white circles) or unmethylated DNA (white squares) to R133C results in an increased peak at ϳ197 nm. f, the T158M MBD (black circles) acquires secondary structure upon addition of methylated DNA (white circles) and to a lesser extent with unmethylated DNA (white squares). g and h, the CD spectra of the F155S and R106W mutant MBDs are unchanged by the addition of methylated DNA (white circles) or unmethylated DNA (white squares). spectra in the near UV region (250 -300 nm), indicating that DNA does not contribute to the change in CD spectra observed with DNA-protein complexes. The addition of equimolar amounts of DNA to the WT MBD produced subtle but reproducible changes in secondary structure (Fig. 5c, Table 2). Most striking was the increased magnitude of the peak at ϳ197 nm, which was confined to the methylated substrate. Deconvolution of these data predicts that in the presence of methylated DNA, the MBD acquires additional ␣-helix (from ϳ15 to ϳ20%; Table 2), whereas with unmethylated DNA the ␣-helix increase is smaller. When bound to DNA, the T158M mutant (Fig. 5f) gave a CD spectrum similar to the WT. In contrast, the MBDs from F155S and R106W showed no appreciable change of structure in the presence of either methylated or unmethylated DNA (Fig. 5, g and h; Table 2).
In the presence of DNA, full-length WT MeCP2 showed a marked decrease in the ϳ200 nm CD trough (Fig. 5a). This change, which was largely methylation-independent, is predicted by CONTINLL to result from a ϳ10% gain in secondary structure.
DNA Binding Capacity of MBD Mutants Correlates with Their Structural Properties-We previously used EMSAs to demonstrate the binding of full-length MeCP2 to DNA and chromatin (19). These data demonstrated that in the presence of excess nonmethylated competitor DNA, methylated target DNA showed a robust shift with full-length wild-type MeCP2, whereas unmethylated DNA produced a smaller shift. The R133C mutation also resulted in a shift, but there was no methylation-dependent component, and with the F155S mutation both the methylation-dependent and independent shifts were much smaller than wild type. R106W showed the most dramatic change, producing neither methylation-dependent nor independent interactions.
In the present study, we focused on the MBD and allowed it to interact with DNA consisting of twelve 207-bp DNA seg-   (20). Fig. 6 shows the results of an experiment in which a 10:1 molar ratio of MBD:target 207-bp segment (i.e. 1 MBD/1.8 CpGs) was used. With the wild-type MBD, there is a small methylation-independent shift and a strong methylationdependent enhancement. T158M gave results that were almost identical to wild type, suggesting that its MBD retains normal binding properties, including a functional methylation-specific DNA binding site. This agrees with the CD data, which revealed T158M as a mutation with a secondary structure similar to wild type both in the absence and presence of DNA (Fig. 5f). The other mutations, R133C, F155S, and R106W, all produced very weak gel shifts with unmethylated substrate and showed very little methylation-dependent enhancement. Domains Beyond the MBD Contribute to Its Structure-The dramatic difference in fluorescence intensity between fulllength MeCP2 and the MBD alone suggested that portion(s) of the protein beyond the MBD contributed to the protection of Trp-104 from the aqueous environment. To explore this possibility further, we created constructs that included sequences adjacent to the 75-164-residue MBD. The fragment MeCp2-(1-164) comprises the complete N-terminal region, and MeCP2-(75-210) includes a short region between the MBD and transcriptional repression domain that exhibits strong methylation-independent binding to DNA. 3 We also examined the RTT-causing C-terminal truncation (MeCP2-(1-294)), which lacks almost all of the C-terminal domain (19). Each of these constructs resulted in peak emission values intermediate between the full-length protein and the MBD (Fig. 4d). Thus, in the full-length protein, the MBD is involved in multiple associations with the other portions of the molecule that contribute to the protection of Trp-104 from the aqueous environment.

DNA Binding Impacts Thermal Stability and Secondary Structure of Wild-type MeCP2
Our fluorescence spectroscopy studies showed that in wildtype MeCP2, Trp-104 is fully protected from the aqueous environment at low temperature (10 -25°C), and after thermal melting at 75°C, it becomes exposed to solvent (Fig. 2a). DNA binding has a powerful stabilizing effect on MeCP2. The fulllength wild-type protein has a T m of 44.5°C (Fig. 3a, Table 1), but in the presence of saturating amounts of a 45-bp segment of BDNF promoter containing a single methylated CpG, the T m increases dramatically to 18.6°C (Fig. 3a). In the absence of DNA, raising the temperature to the T m of the DNA-protein complex (63°C) would result in more than 80% unfolding (Fig.   2b). In the absence of methylation, DNA binding has a significant stabilizing effect, evident in the smaller (11.2°C) increase in T m .
In contrast to full-length MeCP2, the MBD shows DNAinduced stabilization that is almost totally methylationdependent (Fig. 4b, Table 1). Importantly, for both the fulllength protein and the MBD, the difference in T m between methylated and unmethylated DNA is very similar (7.4 versus 7.7°C). These results confirm and extend suggestions that DNAbindingbyfull-lengthMeCP2comprisesbothmethylationdependent and -independent events (19). We have now shown that the methylation-dependent component is mediated by the MBD and weakened by MBD mutations, whereas the methylation-independent component(s) are mediated by other MeCP2 domain(s).
The impact of DNA binding also involves increases in secondary structure, as the CD data clearly demonstrate. When exposed to methylated DNA, the wild-type MBD showed a strong increase in the ellipticity peak at ϳ197 nm, whereas unmethylated DNA showed only minor changes in this region of the spectrum (Fig. 5c). Deconvolution predicted that the methylation-specific changes were largely attributed to a modest increase in ␣-helix from ϳ15 to ϳ20% ( Table 2).
The acquisition of additional ␣-helical structure in the presence of DNA has also been observed with other DNA-binding proteins, including the bZIP family (48,49) and GCN4 (50). Also, a portion of the C-terminal region of mouse histone H1-4 containing residues 173-197 was shown to have no ␣-helical component in phosphate buffer but gains ␣-helicity upon binding DNA (51), and Vila et al. (52) observed a substantial DNAinduced increase in the fraction of ␣-helix of the N-terminal region of H1.

Interdomain Interactions Are Key Features of MeCP2
The fluorescence spectra of various truncations of full-length MeCP2 showed reduced fluorescence (Fig. 4d), indicating that altered or missing interdomain contacts are necessary for masking the solvent exposure of Trp-104. The reductions in emission intensity at max were progressive, from the fulllength protein (100%) to the Rett-causing MeCP2-(1-294) mutant (90%) to the N-and C-terminal MBD extensions MeCP2-(1-164) and MeCP2-(75-210) (both 82%) to the MBD alone (64%). Thus, MeCP2 domains other than the MBD contribute to the local environment of the Trp-104 residue. Our results also indicated that the interactions between the MBD and other MeCP2 domains are modulated differently by RTTcausing mutations. For example, the full-length protein carrying the T158M mutation has fluorescence intensity very similar to the WT, whereas the R133C mutant shows a 27% reduction 3 R. P. Ghosh and C. L. Woodcock, unpublished observation.  (64): most hydrophobic ϭ red, neutral ϭ white, least ϭ blue. a and b, the location of Thr-158 is indicated by yellow arrows on the WT surface. c, in the T158M mutation, the substituted Met is more hydrophobic than the WT Thr, and it has a more pronounced topography and greater solvent-accessible surface exposure. Residues 150 -165 are shown in color according to hydrophobicity. d and e, Arg-133 (yellow arrows) is a large topographical feature with extensive surface exposure in a homogeneously charged surface region composed of the DNA-interacting residues (colored) as described in the crystal structure (5) and labeled in e. In R133C (f), the cysteine substitution creates a dramatically different environment. Arg-106 (blue (g)) is located in a pocket in the WT surface. It is part of a continuous ring of uniform charge and is elevated from the surrounding residues (shown by arrows in h and i). In the R106W mutation, the substituted Trp (arrows) does not protrude from the base of the pocket and interrupts the otherwise continuous charge. Views i and k are rotated 90°on the y axis of the overview (g) and the magnified views (h and j).
at 25°C (Fig. 2a, Table 1). In contrast, with the MBD only, the situation is reversed, with R133C having the WT level of fluorescence intensity and T158M showing a 16% reduction (Fig.  4a, Table 1).
A comparison of the stability of the full-length protein with that of the MBD alone further supports the importance of interdomain interactions. The 4.5 nm red shift and 5°C T m difference exhibited by the F155S MBD compared with the 7 nm change in red shift and 10.9°C ⌬T m of the full-length mutant protein suggests a strong perturbation of regions beyond the MBD in this mutant. Both the T158M and R133C mutants show similar trends, indicating that all three mutations have global impacts on MeCP2 structure and supporting the concept of coupling between the MBD and other domains.
When transfected into cells, the MBD alone has a much higher intranuclear mobility than full-length MeCP2 (53), underscoring the importance of domains beyond the MBD in chromatin binding. Understanding the contributions of domains other than the MBD to nonspecific DNA binding, and the influence of RTT-causing mutations on these interdomain interactions, is an important future goal. MeCP2 is predicted to contain short regions of ␤-structure in a number of domains (6), and interactions between these and the MBD ␤-structures may contribute to these interdomain associations.

Unique Properties of the Four MBD Mutants and Their Functional Implications
T158M-RTT patients with the T158M mutation experience relatively severe symptoms, and in vivo the mutated protein shows a greatly increased intranuclear mobility (54) suggestive of major dysfunction. Our findings suggest that the reduction in thermal stability (Fig. 2b) and inability of the T158M mutant to acquire the wild-type level of stability upon DNA binding (Fig.  3d) may be the basis of its dysfunction. The MBD of T158M shows both a reduced fluorescence intensity (Fig. 4a) and thermal stability compared with the WT MBD (Fig. 4c). Moreover, the changes in surface topography and properties of the local charge environment predicted by molecular modeling of the MBD (Fig. 7, a-c) are consistent with dysfunction resulting from less stable interdomain interactions. Also disruption of the 3-10 helix at residues 151-154 that interact with the AT run in the x-ray structure (5) may account for the reduction in DNA-induced thermal stability.
In some respects, the T158M mutant appears to be similar to the WT, as seen from the CD spectra in the presence and absence of DNA (Fig. 5e) and from the EMSA data reported here (Fig. 6) and in previous studies (55,56). Ho et al. (5), however, recently reported EMSA data showing a T158M MBD construct to be deficient in binding methylated DNA. Thus, at present, EMSA results with the T158M MBD have been inconsistent. It is possible that the differences are attributable to the minor differences in the choice of the N and C termini of the MBD constructs, especially as T158M occurs very close to the C terminus of the MBD.
R133C-This mutation clearly gives rise to structural changes in the isolated MBD. The fluorescence intensity of the full-length protein at max is only 73% of wild type (Fig. 2a,  Table 1), although for the MBD alone, the intensity of WT and mutant are identical. Both the full-length mutant and the MBD show reduced thermal stability compared with the WT (Table  1), and the EMSA data (Fig. 6) indicate a much reduced binding to methylated DNA. However, the impact of the Arg 3 Cys mutation differs between the MBD and full-length MeCP2, possibly as a result of the loss of interdomain interactions in the whole protein. Molecular modeling predicts a major change in surface topography (Fig. 7, d-f), suggesting a rationale for its inability to bind methylated DNA and consistent with the central role of Arg-133 in interacting with 5-methyl cytosine (5,42).
F155S-The F155S mutation is unique in its sensitivity to thermal unfolding; even at 25C, it is partially unfolded (Fig. 2a). This was predicted by its inability to be resolved in NMR studies (55) and on the basis of a comparison of EMSA data (54,55). Deconvolution of the CD spectrum of the F155S MBD predicts an increased proportion of ␣-helix and reduced ␤-turn compared with wild type. Its unaltered secondary structure in the presence of both methylated and unmethylated DNA (Fig. 5g) is consistent with its very weak gel shift in the presence of DNA (Fig. 6). Molecular modeling of the F155S MBD (not shown) reveals disruptions that affect the Asx-turn-Ser-Thr motif shown to be important in the MBD-DNA complex (5). Using molecular dynamics simulations to determine the residues that are involved in the CD-predicted reduction in ␤-turn, where many of the known DNA recognition and binding interactions have been identified, is an important future goal.
R106W-Based on the CD spectra of the MBD, R106W is predicted to have a reduced level of ␣-helix and increased ␤-sheet (Fig. 5h, Table 2). As with F155S, it showed no changes upon exposure to methylated DNA, consistent with the reduced DNA binding seen with EMSA (Fig. 6). Modeling the Arg 3 Trp mutation suggests that the increased fluorescence emission, compared with the WT (Fig. 2a), results from a small increase in exposure of Trp-104 concomitant with a significant exposure of the second tryptophan at residue Trp-106. Examination of the modeled surface of the MBD strongly suggests that Arg-106 is located in an interdomain or interprotein interaction pocket (Fig. 7, g and h) and that the R106W mutation disrupts the local topography and charge environment. Thus, the DNA binding failure of R106W may be related to changes in water-mediated methyl group recognition by Tyr-123. As in the case of F155S, dynamics simulations of the MBD-DNA interaction may shed further light on how the CD-predicted secondary structure changes affect its interaction with DNA.

Correlation with Other Experimental Data and Disease Severity
Previous studies comparing the DNA binding capabilities of full-length mutant and wild-type MeCP2 (55,57) are generally consistent with our structural data, and we recently showed that for R133C, F155S, and R106W, this impairment extends to interactions with nucleosomal arrays (19). Moreover, in an in vivo system, Kudo et al. (58) noted that R106W and F155S fail to show the wild-type localization in pericentromeric heterochromatin when transfected into mouse L929 cells. These mutants were also ineffective in causing repression of a reporter gene in Drosophila SL2 cells. The R133C mutation retained both local-ization and repression functionality, whereas T158M was intermediate in both tests. Recently, Kumar et al. (54) reported large changes in the intranuclear mobility of MeCP2 mutant proteins. The R106W, T158M, and F155S mutations resulted in a dramatic ϳ15-fold increase in mobility within pericentromeric foci, whereas with R133C, mobility was about double that of the WT (Table 2). An increase over the WT in intranuclear mobility of transfected MeCP2 carrying the R106W mutation was also noted by Marchi et al. (53).
The remarkable diversity revealed by our comparative studies of the wild type and four missense MBD mutants of MeCP2 is to some extent mirrored by the range and diversity of symptoms in RTT patients. An analysis of symptoms in which the severity of 22 phenotypic symptoms in over 300 patients was scored as mild (0), moderate (1), or severe (2) ranked R106W as the most debilitating of the mutants studied in the present work (mean score 1.1) followed by T158M (0.83) and R133C (0.67). Further, the 15 patients with a mean severity score of 1.25 or greater included two instances each of R106W and T158M. No F155S cases were included in the severity ranking, but the population of F155C mutations had a severity index of 0.82. Interestingly, modeling of the F155C mutation (not shown) predicts a ϳ3-fold increase in surface area exposure of Trp-104 as a result of the Phe 3 Cys substitution.

Rett Syndrome and the Intrinsic Disorder of MeCP2
Our fluorescence and CD data suggest that a primary difference between the wild-type protein and RTT-causing mutations is the failure to adopt the native secondary structure and/or undergo stabilization in the presence of methylated DNA. Structural changes induced by ligand binding are characteristic of intrinsically disordered proteins, the disordered state being a prerequisite for molecular recognition (59). Thus, DNA-induced disorder-to-order transitions may be important in MeCP2 function. Interestingly, the extensively studied p53 protein, which is mutated in ϳ50% of human cancers, shares several features with MeCP2. As shown here for MeCP2, mutation-induced changes include loss of thermal stability, weakened DNA binding capacity (60), and disruption of binding partner interactions (61). Such mutations are considered potential targets for drugs that "rescue" the native structure (e.g. 60,62,63), and a similar strategy may be applicable to MeCP2.