A Role for MeCP2 in Switching Gene Activity via Chromatin Unfolding and HP1γ Displacement

Methyl-CpG-binding protein 2 (MeCP2) is generally considered to act as a transcriptional repressor, whereas recent studies suggest that MeCP2 is also involved in transcription activation. To gain insight into this dual function of MeCP2, we assessed the impact of MeCP2 on higher-order chromatin structure in living cells using mammalian cell systems harbouring a lactose operator and reporter gene-containing chromosomal domain to assess the effect of lactose repressor-tagged MeCP2 (and separate MeCP2 domains) binding in living cells. Our data reveal that targeted binding of MeCP2 elicits extensive chromatin unfolding. MeCP2-induced chromatin unfolding is triggered independently of the methyl-cytosine-binding domain. Interestingly, MeCP2 binding triggers the loss of HP1γ at the chromosomal domain and an increased HP1γ mobility, which is not observed for HP1α and HP1β. Surprisingly, MeCP2-induced chromatin unfolding is not associated with transcriptional activation. Our study suggests a novel role for MeCP2 in reorganizing chromatin to facilitate a switch in gene activity.


Introduction
Gene activity is governed by the interplay between various proteins that modulate the epigenetic composition of chromatin (e.g. DNA methylation, histone modifications) [1]. Histone modifications and DNA methylation are linked by CpG-binding proteins such as methyl-CpG-binding protein 2 (MeCP2) [2] through, for instance, cross-talk between MeCP2 and heterochromatin protein 1 (HP1) isoforms [3]. MeCP2 is ubiquitously expressed in human tissues and particularly enriched at pericentromeric heterochromatin domains in brain cells [4,5]. MeCP2 plays a role in neuronal maturation and impaired MeCP2 function results in neurodevelopmental disorders such as Rett syndrome [6,7]. HP1 is a chromatin-binding protein that bridges H3K9methylated histones with other chromatin-associated proteins thereby advancing the 'spreading' of heterochromatin [8,9]. Both the clustering of pericentromeric heterochromatin domains and the relocalization of HP1 (in particular HP1c) occur during myogenic differentiation when the level of methyl-CpG-binding proteins is up-regulated [3,10].
MeCP2 was originally found to bind methylated DNA and to act as a transcriptional repressor [11][12][13]. More recent work demonstrated that MeCP2 also binds at actively transcribed genes and promotes activation of DNA-methylated genes, suggesting a role as a transcriptional activator [14][15][16][17][18]. Currently, MeCP2 is considered a multifunctional protein [19], i.e. MeCP2 is known (i) to bind methylated DNA [11,12,13], (ii) to recruit a wide range of proteins (e.g. chromatin-remodeling proteins Brahma, ATRX) [20][21][22][23][24][25][26], (iii) to induce the formation of repressive chromatin [5,27,28] and change the number and size of pericentromeric heterochromatin domains [29], (iv) to be involved in histone H1 displacement [21,[30][31][32], (v) to play a key role in neurological disease (e.g. Rett syndrome) involving both gene activation and repression [30], (vi) to be implicated in the regulation of imprinted genes [33]. To unambiguously assess how MeCP2 contributes to epigenetic gene regulation within the context of the mammalian genome, we targeted MeCP2, an MeCP2 Rett mutant (R133C) or separate MeCP2 domains as EGFP-lac repressor (lacR)-tagged fusions in cells harbouring a lac operator (lacO) and reporter gene-containing genomic domain [34]. Using this methodology, we previously showed that HP1 targeting is sufficient to induce local chromatin condensation and recruitment of histone methyltransferase SETDB1, concomitant with increased tri-methylation of H3K9 [35]. Here we show that MeCP2 targeting causes extensive chromatin decondensation of the targeted genomic domain, which occurs independently of the MeCP2 methyl-cytosine-binding domain (MBD) and results in eviction of the HP1c isoform without an alteration in the transcriptional activity of the targeted chromatin. The images show an individual optical section of an AO3_1 cell highly overexpressing tagged MeCP2 (5,3 times higher than representative cells shown in (A). Representative cells are the cells that allow the Argos image analysis programme to select the lacO array from nuclear background, i.e. the cells on which we performed our quantitative measurements (see Figure 4 and S1). This analysis of MeCP2 over-expression illustrates that in cells exhibiting overexpressed MeCP2 levels a visually unfolded lacO chromatin array is observed similar as in the cells expressing representative MeCP2 levels. (E) The images show AO3_1 cells transfected with EGFP-lacR (control) and EGFP-lacR tagged MeCP2 (green signal), FISH-labeled with a fluorescent lacO probe (red signal) and DAPI-stained (blue signal). The images represent individual optical sections. Bars = 5 mm. doi:10.1371/journal.pone.0069347.g001 Spring Harbor Laboratory, New York (USA) [35,37]) were used. The AO3_1 and RRE_B1 clone are derivatives of CHO DG44 cells and contain an integrated amplified chromosomal region consisting of the dihydrofolate reductase (DHFR) cDNA transgene, 256 lac operator repeats and flanking DNA, representing a compact chromatin structure and an unusually extended fibrillar chromatin conformation, respectively. The 2-6-3 clone is a U2OSderived clone containing a multicopy inducible transgene consisting of 256 lac operator repeats, a tetracycline-inducible reporter gene encoding cyano fluorescent protein with a peroxisomal targeting signal, 24 repeats of the MS2 bacteriophage translational operator, a splicing cassette and the 3' UTR from the rabbit b globin gene [35,37]. U2OS and NIH/3T3 cells were cultured in Dulbecco minimal essential medium containing 10% fetal bovine serum, 1% pen/strep. (Gibco) The AO3_1 and RRE_B1 clone were cultured in Ham's F-12 medium without hypoxanthine or thymidine supplemented with 10% dialyzed fetal bovine serum (HyClone Labs, Logan, Utah), 1% pen/strep and methotrexate up to a final concentration of 0.03 mM or 10 mM, respectively. The 2-6-3 clone was cultured in high glucose Glutamax (Gibco) with 10% tetracycline free FBS (Clontech), 1% pen/strep and 100 mg/ ml hygromycinB (Gibco). All cells were cultured at 37uC in a 5% CO 2 atmosphere.
Transfection was performed with Lipofectamine 2000 (Invitrogen) or SAINT mix (Synvolux Therapeutics, Groningen, The Netherlands) in their respective media without Pen/Strep. For microscopy experiments cells were seeded on cover slips coated with Alcian Blue or 35 mm glass bottom dishes (MaTek). After 24-48 hours, cells were imaged directly or fixed in 4% paraformaldehyde for 15 minutes at 4uC and embedded in Vectashield (Brunschwig, Burlingame, CA) with DAPI (Vector Laboratories, Burlingame, CA).
Transfections for luciferase reporter assays were done with 8x lacO pGL3 luciferase vector, lacZ construct as an internal reference reporter and EGFP-lacR-tagged effector plasmids. Cells were harvested and lysed at 48 hours post-transfection. Luciferase reporter gene-targeted repression assay was performed as described previously [35]. Briefly, luciferase reporter assay transfections were done with 500 ng of the 8x lac operator containing pGL3 luciferase vector, 500 ng of pSV/b-Gal

FACS sorting and quantitative PCR
Expressing cells were sorted by flow cytometry (Mo Flo XPD Cell sorter Beckman Coulter, Woerden, The Netherlands). Extracted mRNA was converted to cDNA using the Fermentas RevertAid TM First strand cDNA synthesis kit. Quantitative PCR amplifications were performed on an ABI Prism 7900HT Sequence Detection System. All PCR reactions were carried out in triplicate using TaqmanH gene expression assay Mm00515662_m1 for DHFR and Mm99999915_g1 for GapdH (Applied biosystems). Relative quantification of gene expression was calculated based on the comparative cycle threshold (Ct) method.

Image acquisition
Cells were imaged using a Zeiss LSM 510 (Zeiss, Jena, Germany) confocal laser scanning microscope equipped with a Zeiss Plan-Apochromat 636/1.4 oil immersion objective or a Zeiss plan neofluar 636/1.25 NA oil objective. We used multitrack scanning, employing a UV laser (364 nm), an argon laser (488 and 514 nm) and a helium-neon laser (543 nm) to excite DAPI staining and green/yellow and red fluorochromes. Emitted fluorescence was detected with BP 385-470, BP 505-550 and 560 LP filters. Three-dimensional (3D) images were scanned at 512 by 512 pixels averaging 4 times using a voxel size of 200 nm axial and 60 nm lateral.
To detect the fluorescence intensity levels of the endogenously immunolabeled MeCP2 and exogenous transfected fluorescently tagged MeCP2, a tile scan was made of 64 tiles of 512 by 512 pixels/tile. Image analysis that were transfected (green signal) with EGFP-lacR (control) or EGFP-lacR-tagged full-length MeCP2, -VP16 and -MeCP2 separate domains (i.e. MBD, DC-terminus, TRD, C-terminus and R133C Rett syndrome mutation). The images show a typical representation of cells transfected for 48 hours. For quantitative analysis 30 nuclei per condition were measured with comparable microscopical set-up. The images represent examples of individual optical sections. Bar = 5 mm. (B) The figure shows the changes in lacO array large-scale chromatin structure measured with a 3D image analysis parameter, i.e. the surface factor. The surface factor determines the surface of a given chromosomal domain/object normalized to the surface of a sphere with an equal volume [43]. The distribution of the surface factor measurements is plotted as a box-plot. The second and third quartiles of the observed values are within the box, the median value is shown by the white horizontal line, the whiskers show the first and fourth quartiles of the observed values, dots are the outliers. The dotted line represents the ,20% of cells in the EGFP-lacR control population that exhibit a mildly decondensed array (see also [69]). Considering a population (20%) of control cells exhibiting a decondensed configuration as the threshold of MeCP2induced decondensation, the following percentages are found for lacR control 18%, VP16 68%, MeCP2 full length 47%, MBD 23%, DC-terminus 23%, C-terminus 47% and R133C 50%). We scored the chromatin structure based on the surface factor in the full cell population. Based on our statistical analysis the EGFP-lacR tagged VP16, MeCP2 (full length) and the R133C population are significantly different from EGFP-lacR control, whereas EGFP-lacR tagged MBD, TRD, C-terminus and DC-terminus, are not significantly different from EGFP-lacR (for p values see Table 1)   For quantitative analyses approximately 30 nuclei were imaged with comparable microscopical set-up. To quantitatively analyze changes in large-scale chromatin structure, we applied 3D image analysis tools (the Huygens system 2 software package; Scientific Volume Imaging, Hilversum, The Netherlands) as described previously [35,47]. Briefly, 3D images of the amplified chromosome region are acquired after which the EGFP-lacR-labeled chromosome region is automatically identified. Specific features of the LacO array (3D structure, volume and intensity) are subsequently calculated using Huygens software. To assess the 3D chromatin structure, we used a parameter termed the surface factor, which represents the surface of a given chromosomal domain/object normalized to the surface of a sphere with an equal volume [43]. A surface factor of 1 therefore represents a perfectly spherical structure, whereas a lower value represents a more furrowed structure, such as a decondensed chromatin domain. The distribution of the calculated surface factors (,30 nuclei per variable) is plotted in a box-plot. The second and third quartiles of the surface factor values are within the box, the median value is shown by the thick horizontal line, the vertical small lines show the first and fourth quartiles of the observed values. Since our surface factor data is not normally distributed (as tested with Shapiro Wilktest rejecting the null hypothesis that the data are normally distributed with p,0.05, p values see S1) and exhibits an unequal variance (Variance Equivalence test), we used the Wilcoxon nonparametric test [48] corrected for multiple testing (Bonferroni, p = 0.07) to calculate the P-value giving the probability that the control population (cells expressing EGFP-lacR) and the test population (cells expressing EGFP-lacR tagged MeCP2, separate MeCP2 domains or VP16 protein) are significantly different from each other. For the expression level measurements, the intensity of the transfected constructs at the lacO array was detected and normalized to the gain and offset settings of the PMT using a standard curve for the used parameters. The Pearson correlation coefficient was calculated to detect the correlation between the surface factor and the relative normalized intensities.
To depict variable expression levels of fluorescently tagged MeCP2, expression levels were measured using imageJ on both single scanned cells as well as tile-scanned images. Cells were masked and the nuclear counterstain intensities normalized.

Photobleaching experiments
For FLIP and FRAP experiments, microscopes were equipped with an objective heater and cells were examined in microscopy medium (137 mM NaCl, 5.4 mM KCl, 1,8 mM CaCl 2 , 0.8 mM MgSO 4 , 20 mM D-glucose and 20 mM HEPES) at 37uC. FRAP and FLIP analysis was performed as described previously [49]. FRAP was used to measure the mobility of GFP-HP1cin-and outside of the array visualized by mCherry-LacR. Briefly, images were taken at 5126512 pixels (0.0460.04 mm), 1.60 ms per frame, zoom 7. After 10 images, a square of 56656 pixels was bleached for 10 scans (total time = 1.1 s) and recovery was measured for at least 60 images at a 2-second time interval. The data was normalized to the original intensity before the bleach pulse by using the equation: I FRAP = (I strip t = t -I background t = t )/(I strip t = 0 -I background t = 0 ), where I strip t = t and I strip t = 0 represent the intensity within the strip at t = t and the intensity before the bleach pulse (t = 0), respectively. For graphical representation, recovery plots were normalized between 0 and 1. FLIP analysis was used to measure the residence times of GFP-HP1c on chromatin. Briefly, images of 5126512 pixels were acquired with a scan time of 1.60 ms (1x average/frame) at zoom 7 (1 pixel is 0.0460.04 mm). After 10 images, a region of 2756150 pixels, occupying an area of 1/3 of the nucleus (excluding the lac operator array), was continuously bleached with maximal 488 nm and 514 nm laser intensity (AOTF 100%). EGFP-HP1c fluorescence was monitored with low laser intensity for at least 80 images with a 2-second time interval between images. The loss of fluorescence in the unbleached part of the nucleus was quantified. All values were background corrected and normalized to 1 by using the equation: I FLIP = (I spot t = t -I background t = t )/(I spot t = 0 -I background t = 0 ). Curve fitting was performed according to N1*e (l1*t) +N2*e (l ? 2*t) .

MeCP2 targeting causes local chromatin unfolding
To investigate the effect of MeCP2 targeting on chromatin folding we used cell lines that enable targeting of (EGFP-lacRtagged) MeCP2 and separate MeCP2 domains to an integrated lacO genomic region, that is present in a highly amplified chromosomal domain in hamster cells (the AO3_1 and RRE_B1 clones) or as a multicopy genomic integration in human cells (the 2-6-3 clone). While the AO3_1, RRE_B1 as well as the 2-6-3 cells allow measurements of visual changes in 3-D chromatin folding upon targeting to the lacO array, it should be noted that the AO3_1 and RRE_B1 cells harbor a much larger array (80 Mpb) [35,36,50,51] compared to the 2-6-3 cells (4 Mbp). Therefore, the impact of induced chromatin changes in 3-D folding are more striking in the AO3_1 and RRE_B1 cells. Importantly, the 2-6-3 clone exhibits less genomic instability then the AO3_1 and RRE_B1 clones, which makes it more suitable to study cell-cycledependent chromatin folding. Moreover, the 2-6-3 clone harbors a reporter gene containing 24 tandem MS2 repeats allowing visualization and measurement of changes in the transcript levels using YFP-tagged MS2 coat protein.
We generated EGFP-lacR-tagged full-length MeCP2 or separate MeCP2 domain fusion proteins and expressed these fusions in the different cell lines harboring lacO arrays. Analysis of 3-D confocal images revealed that MeCP2 targeting induced extensive unfolding of the lacO array compared to targeting lacR in AO3_1, RRE_B1 and 2-6-3 cells ( Figure 1A). MeCP2 targeting induced unfolding of the lacO array in AO3_1 and 2-6-3 cells to the same extent as targeting of the viral activator VP16, which is known to cause extensive chromatin unfolding ( Figure 1A) [35,36,50,51]. We measured the expression level of exogenously expressed MeCP2 compared to endogenous MeCP2 to verify the effect of that physiological levels of MeCP2 can induce chromatin unfolding. Fluorescent immunolabeling using MeCP2 specific antibodies that recognize both exogenous and endogenous MeCP2 showed that mCherry-lacR tagged MeCP2 contained a ,25% higher MeCP2 level compared to the endogenous MeCP2 level in non-transfected cells ( Figure 1B, C). Moreover, we determined the effect of over-expression of exogenous MeCP2 on chromatin folding. This analysis shows that cells with highly over-expressed MeCP2 levels visually exhibit a comparable level of chromatin unfolding ( Figure 1D) compared with cells that express intermediate levels of lacR-MeCP2 (( Figure 1B, C).
FISH labeling of the lacO array in AO3_1 cells using lacO probes confirmed that the MeCP2-induced unfolded structure overlapped completely with the unfolded lacO array ( Figure 1E), showing that the EGFP-lacR visualized unfolded structure resembles the decondensed lacO array. Since the lacO chromo- somal domain exhibits an extended fibrillar chromatin conformation in RRE_B1 cells and a compact chromatin (heterochromatic) structure in the AO3_1 and 2-6-3 cells, we used the AO3_1 and 2-6-3 cells for investigation of the effect of MeCP2 on epigenetic regulation [35,47].
To study whether MeCP2-induced chromatin unfolding is cellcycle dependent, we analyzed cells in S-phase based on the known accumulation of proliferating cell nuclear antigen (PCNA) into foci in replicating cells. These experiments were performed in human U2OS 2-6-3 cells containing a 200-copy lacO array [37]. Our data show that MeCP2-induced unfolding occurred as frequently in Sphase cells as it did in non S-phase cells based on transfections with mCherry-PCNA or staining for endogenous PCNA. These results illustrate that the MeCP2-induced chromatin structural changes are independent of the cell-cycle stage (Figure 2A, B). [44].

MeCP2 chromatin unfolding acts independently of the MBD domain
MeCP2 harbors a methyl-binding domain (MBD), a transcription-repression domain (TRD) and a poorly characterized Cterminal domain [31,52,53]. We tagged various MeCP2 subdomains and regions spanning the MBD, TRD, the C-terminus and MeCP2 lacking its C-terminus (DC-terminus) or mutant MeCP2 R133C (Rett Syndrome point mutation in the MBD domain) [54] to EGFP-lacR to identify which MeCP2 subdomain is responsible for unfolding of the lacO chromosomal domain upon targeting ( Figure 3A). First we tested the nuclear localization of the EGFP-lacR-tagged MeCP2 full-length and MeCP2 separate domain constructs in mouse fibroblasts lacking a lacO array. Similar to native full-length MeCP2, all lacR-tagged MeCP2 fusion proteins localized to pericentromeric heterochromatin in the mouse fibroblasts, except for the lacR-tagged Cterminus, which was homogeneously distributed in the nucleus ( Figure 3B). Next, we analyzed the effect of targeting EGFP-lacR tagged MeCP2 full-length, separate MeCP2 domains or VP16 versus EGFP-lacR (control) in the lacO-containing AO3_1 cells (Figure 4). Targeting of lacR-tagged MBD, TRD or DC-terminus domains did not cause chromatin unfolding into an extended fibrillar stucture as observed with full-length MeCP2 targeting, while targeting of MeCP2 R133C did result into mild chromatin structural changes. Targeting of the lacR-tagged MeCP2 Cterminus caused chromatin fibrillar unfolding but also to a much lesser extent than targeting of full-length MeCP2. Figure 4A shows typical representations of cells transfected for 48 hours with the respective constructs. For quantitative analysis approximately 30 nuclei per condition were imaged with a comparable microscopical set-up ( Figure 4B, C, Table S1). We quantitatively assessed the change in chromatin structure by measuring the surface of the lacO chromatin domain relative to surface of a sphere with equal volume. A perfectly spherical structure, has a designated surface factor of 1, whereas a fibrillar unfolded chromatin structure will have a lower surface factor, due to its furrowed surface with equal volume. The quantitative measurements mirrored our visual observations: the degree of chromatin unfolding as observed in full-length MeCP2 (and VP16) was most pronounced in the cells targeted by separate MeCP2 domains containing the MeCP2 domains downstream of the MBD (i.e. MeCP2 R133C , C-terminus and TRD), while the MBD and DCterminus show almost a similar chromatin structure as observed upon EGFP-lacR control targeting ( Figure 4B, Table 1, Table  S1). To verify whether the amount of expressed MeCP2 has an impact on chromatin unfolding we measured the fluorescence intensity of the transfected constructs in the lacO chromosomal array and we compared them with the surface factor measure-ments. Our intensity measurements show that there is no significant correlation between the amount of transfected construct and the extent of chromatin unfolding as measured by the surface factor ( Figure 4C, Table S1). Our findings reveal that the MeCP2 regions downstream of the MBD are involved in MeCP2induced chromatin unfolding whereas the MBD itself is dispensable for this phenomenon.

MeCP2-associated factors: MeCP2 targeting interferes with HP1c binding
Through targeting to the lacO array in AO3_1 cells, we subsequently analyzed the accumulation or displacement of a variety of previously reported MeCP2-associated factors including proteins and epigenetic marks related to a transcriptionally active chromatin state (i.e. the Brahma subunit of the SWI/SNF complex [22,24,55], TFIIB, CREB1 [15,56], RNA polymerase II, RNA splicing factor SC35 [57], H3K4 di-methylation and H4K16 acetylation and CpG methylation) ( Table 2). While the distribution of most of these factors or epigenetic marks was not altered by targeting MeCP2, we confirmed that MeCP2 targeting interferes with chromatin binding of linker histone H1 (FRAP data not shown), which is compatible with the observed chromatin unfolding [19,[30][31][32]. Strikingly, we detected MeCP2-induced changes in the distribution of HP1c and decided to study this in more detail.
We recently identified an interaction between MeCP2 and the HP1 proteins in mouse myoblast cells [3]. Co-transfection of mCherry-lacR and EGFP-tagged HP1a, b or c in A03_1 cells showed enrichment of all HP1 isoforms at the lacO array ( Figure 5A-C). Indeed, FRAP measurements on the mobility of HP1c at the lacO array (visualized by mCherry-LacR) or elsewhere in the nucleus showed that HP1c has a slower exchange rate at the array ( Figure 5D). The measured binding kinetics of HP1c at the lacO array is in accordance with previous findings on the binding dynamics of HP1 at pericentromeric heterochromatin [58]. Therefore, our findings imply that HP1c binding at the lacO array in the AO3_1 cells reflects binding at heterochromatin, which is in agreement with the heterochromatic nature of the array ( Figure 5A-D). Strikingly, while HP1a and b remained bound ( Figure 5A, B), HP1c accumulation at the lacO array was lost upon MeCP2 targeting ( Figure 5C). We determined the HP1c exchange rates using fluorescence loss in photobleaching (FLIP) (Figure 5E), which identified two dynamic HP1c pools corresponding to a freely diffusing (t 1/2 ,3 s) or a transiently chromatin-bound population (t 1/2 ,50 s) ( Figure 5E). LacRtransfected cells displayed reduced HP1c mobility at the array (37%, t 1/2 = 52 s), indicating efficient retention of HP1c at the array. Strikingly, the proportion of the freely diffusing HP1c population went up to 97% (t 1/2 = 1.5 s) after targeting of lacRtagged MeCP2. These results confirm that MeCP2 effectively antagonizes binding of HP1c to chromatin.
To gain more insight into the MeCP2-induced HP1cdissociation from the lacO array upon chromatin unfolding, we analyzed the ability of MeCP2 to displace two separate HP1c domains, the chromodomain, CD (1-75), which mediates HP1 binding to H3K9 tri-methylation, and the chromoshadow domain, CSD (92-173), which mediates HP1 dimerization and binding to a number of other proteins [59]. The HP1c CSD domain localized at the lacO array upon targeting mCherry-tagged lacR whereas MeCP2 targeting triggered the displacement of the HP1c CSD from the lacO array ( Figure 6B, C). The distribution of the HP1c CD was not affected upon MeCP2 targeting. FLIP analysis confirmed these findings and revealed that the mobility of the HP1c CSD at the lacO array in control cells was similar to the full-length HP1c as freely diffusing (73% t 1/2 , 8.77 s) and chromatin-bound (27%, t 1/ 2 ,57.8 s) pools could be detected. Upon targeting mCherry-lacR tagged MeCP2, a striking shift towards freely diffusing HP1c CSD molecules (92%) could be measured, while only a small fraction remained chromatin bound (8%; Figure 6E). These experiments suggest that MeCP2-induced displacement of HP1c is mediated through its CSD.

MeCP2 targeting does not change the genomic transcriptional state
To test whether lacR-tagged MeCP2 and MeCP2 subdomains modulate gene activity at the promoter level, we measured gene expression levels of a transfected plasmid containing a luciferase reporter gene fused to 8 lacO copies [35]. Targeting LacR-tagged MeCP2, MeCP2 R133C and DC-terminus to a transiently expressed lacO-containing luciferase reporter gene in U2OS cells significantly repressed luciferase expression (60-70%) compared to targeting only lacR ( Figure 7A). In contrast, targeting the Cterminal domain of MeCP2 caused repression to a moderate extent (,25%) ( Figure 7A). Taken together, these results demonstrate that the lacR-tagged MeCP2 protein is able to repress gene activity of a transiently expressed luciferase reporter gene plasmid. To address the effect of MeCP2 on gene activity at an integrated genomic locus, we assessed whether MeCP2 influences transcription of genes embedded in the lacO array integrated in the genome of AO3_1 cells. The MeCP2-induced unfolded chromatin structure in AO3_1 cells did not show enrichment of BrUTP-labeled transcripts from the DHFR selection gene compared to lacR-targeted control cells ( Figure 7B). In contrast, a considerable increase in BrUTPlabeled transcripts was observed at the VP16-induced unfolded chromatin structure of the lacO chromosomal domain [50,51] ( Figure 7B). Corroborating these findings, we performed RT-qPCR analysis of the DHFR reporter gene upon MeCP2 targeting and subsequent FACS sorting of the transfected AO3_1 cells. Our results show that the DHFR reporter gene expression was not significantly altered compared to lacR-targeted control cells, whereas VP16 targeting resulted in significantly enhanced expression levels of the reporter gene ( Figure 7C). In addition, we employed the 2-6-3 clone to visualize MeCP2-induced changes in transcript levels [37]. Expression of MS2-YFP to visualize nascent transcripts showed that MeCP2 targeting did not activate transcription at the locus although significant unfolding of the chromosomal array was observed. Conversely, VP16 targeting induced significant accumulation of MS2-YFP-bound transcripts at the chromosomal array compared with MeCP2 targeting or lacR control targeting ( Figure 7D). Taken together, we show that decondensation of the amplified chromosomal array upon MeCP2 targeting does not coincide with a change in gene activation.

Discussion
Regulation of mammalian gene expression is a tightly controlled process. Mistakes in gene expression control can have far-reaching consequences, such as the manifestation of various developmental disorders or cancer. Mutations in the epigenetic regulatory protein MeCP2 underlies the developmental disorder known as Rett Syndrome [60]. Still, the molecular mechanisms underlying MeCP2-induced context-dependent functioning are largely unresolved. Post-translational MeCP2 modifications, changes in the MeCP2 genomic binding sites or regulation through MeCP2 cofactors likely influences whether MeCP2 acts as a transcriptional repressor or activator [60].
Here we provide evidence that direct targeting of MeCP2 as an EGFP-lacR tagged protein to a lacO-containing chromosomal domain induces extensive chromatin unfolding. Previous studies have shown that the lacR-lacO targeting system is a very powerful method to define the induced effect of (epigenetic) regulatory proteins on genomic behavior [35][36][37][38]. Engineered targeting systems allow to systematically unravel the cause-effect chain of the epigenetic state, chromatin folding, nuclear location and gene activity advancing our understanding of the principles of functional genome organization. We show that MeCP2 binding to native chromatin in intact cells triggers extensive chromatin unfolding and that the MBD is not required for this effect. In contrast to our findings, previous in vitro studies showed that binding of the C-terminal domain of MeCP2 to reconstituted nucleosomal arrays results in chromatin compaction [20,28]. It should be noted that reconstituted nucleosomal arrays lack higherorder chromatin structure and could therefore respond differently to MeCP2 binding than chromatin embedded in the nucleus. In intact mouse myoblasts, overexpression of MeCP2 is shown to induce the MBD-dependent clustering of chromocenters during myogenic differentiation [10]. MeCP2 is a striking example of an intrinsically unstructured protein having a large number of regions predicted to acquire structure when complexed with binding partners [61]. The C-terminal portion of MeCP2 is known to be required for chromatin interactions, it harbors the Group II WW domain-binding motif required for splicing factor binding and the SPxK DNA-binding motif found in histone H1 [19]. Most likely, MeCP2 functioning depends on the type of chromatin and the MeCP2 domain involved. MeCP2 and histone H1 have been shown to compete for chromatin binding sites in vitro and in vivo [19,32]. It has been suggested that a complex competitive equilibrium between MeCP2 and histone H1 for nucleosome and chromatin binding exists and that other competing chromatin binding proteins can affect this histone H1-MeCP2 binding equilibrium [21]. Such context-dependent in vivo functioning of MeCP2 is further underscored by a recent study demonstrating the unique physical properties and interaction domains of MeCP2 [21].
Several studies hint at a relationship between DNA methylation levels, the presence of methyl-CpG-binding proteins and changes in chromatin structure. For instance, a genome-wide loss of H3K9 di-methylation and a progressive increase in H3K9 acetylation, as well as increased chromocenter clustering was observed in mouse embryonic stem cells lacking DNA methyltransferases Dnmt3a and Dnmt3b. Moreover, during myogenic differentiation, overexpression of methyl-CpG-binding proteins induced increased DNA methylation levels and chromocenter clustering, independent of the H3K9 histone methylation pathway and requiring the MBD domain of MeCP2. We detected in situ CpG methylation both at the MeCP2-induced unfolded chromatin and the non-MeCP2 induced compact chromatin which is in line with our observations indicating that the chromatin unfolding is independent of the MeCP2 MBD domain.
Interestingly, we show that MeCP2 interferes with HP1c chromatin binding. Our FRAP analysis shows similar kinetics of HP1c at the lacO chromosomal domain as previously measured at mouse heterochromatic sites [58,62]. The MeCP2-bindinginduced release of HP1c is reflected by the loss of local HP1c accumulation at the lacO chromosomal domain. This MeCP2induced interference with HP1c is also observed with the HP1c CSD but not with the HP1c CD, indicating that the local (chromatin) protein composition influences the ability of MeCP2 to change the HP1c binding kinetics. Moreover, we show that the transcriptional activator VP16 is also able to interfere with HP1c binding upon chromatin unfolding and transcriptional activation, illustrating that the chromatin unfolding and HP1c displacement is not restricted to the changes induced by MeCP2. However, while Janicki et al. showed that VP16-mediated unfolding triggered displacement of the HP1a isoform [37], we find MeCP2-induced chromatin unfolding to result in the selective removal of the HP1c isoform. In breast cancer cells displacement of HP1c is shown to precede transcriptional activation of an integrated luciferase reporter gene [63]. In this study, hormonal signaling triggered phosphorylation of H3S10, displacement of HP1c and ATP-dependent chromatin remodeling resulting in an open, transcriptionally competent chromatin structure. It is tempting to speculate that the MeCP2-mediated chromatin unfolding and eviction of HP1c are part of a similar mechanism to render chromatin amenable to subsequent transcriptional changes.
Our finding that MeCP2 mediates extensive chromatin unfolding, while maintaining transcriptional silencing of genes embedded within the unfolded chromatin structure, is quite surprising. Recent evidence suggests that proteins that typically accumulate at pericentromeric heterochromatin such as HP1 may function as transcriptional activators, in addition to their role as transcriptional silencer [18,64,65]. Such findings would argue that the canonical view in which open chromatin is transcriptionally active and closed chromatin is transcriptionally inert is too simplistic. Our data might indicate that MeCP2-induced chromatin unfolding prepares chromatin for subsequent transcriptional regulation. Examples of changes in chromatin structure preceding transcriptional activation have been reported for the HoxB and MHC locus [66,67]. Moreover, biochemical analyses revealed that transcriptionally inactive sites can occur both in compact and less compact chromatin [68]. We propose that MeCP2-mediated chromatin unfolding reflects an indeterminate state that facilitates a switch in gene activity. In this scenario, the action of subsequent regulatory factors determines the transcriptional fate of genes embedded within chromatin that has been rendered permissive by MeCP2. Such a role of MeCP2 fits well with recent observations in neuronal cells, where MeCP2 is abundantly present and proposed to act as a versatile global transcriptional regulator in concert with other regulatory proteins [5]. Elucidating this global role of MeCP2 in restructuring chromatin in vivo is intriguing and may aid in understanding the pathophysiology of neurodevelopmental disorders, such as Rett syndrome.

Supporting Information
Table S1 Quantitative data MeCP2-induced chromatin unfolding. Our measurements of the surface factor and the testing for a normal distribution of the surface factor data (Shapiro Wilktest), the microscopical gain and offset settings and the measurements of volume and intensity of the lacO chromosomal array are provided. AO3_1 cells (CHO-derived clone containing an amplified chromosomal region consisting of the DHFR cDNA transgene and 256 lac operator repeats) were transfected with EGFP-lacR (control) or EGFP-lacR-tagged full-length MeCP2, -VP16 and -MeCP2 domains (i.e. MBD, DC-terminus, TRD, C-terminus and R133C Rett syndrome mutation) and 30 nuclei per transfected construct were measured with comparable microscopical set-up. We applied a 3D image analysis parameter (the Huygens system 2 software package; Scientific Volume Imaging, Hilversum, The Netherlands) as described previously [35,47]. Specific features of the LacO array (3D structure, volume and intensity) are calculated using Huygens software. The EGFP-lacR-labeled chromosome region is automatically identified in the acquired 3D images, given as the volume. Changes in lacO array large-scale chromatin structure are measured with a 3D image analysis parameter, i.e. the surface factor. The surface factor determines the surface of a given chromosomal domain/object normalized to the surface of a sphere with an equal volume [43]. We tested with Shapiro Wilktest whether the surface factor data is normally distributed. The intensity of the transfected constructs at the lacO array is detected, i.e. total array intensity and normalized to the gain and offset settings of the PMT using a standard curve for the used parameters thereby providing the relative normalized intensity of the transfected constructs. (DOCX)