Histone variant macroH2A1 regulates synchronous firing of replication origins in the inactive X chromosome

Abstract

MacroH2A has been linked to transcriptional silencing, cell identity, and is a hallmark of the inactive X chromosome (Xi). However, it remains unclear whether macroH2A plays a role in DNA replication. Using knockdown/knockout cells for each macroH2A isoform, we show that macroH2A-containing nucleosomes slow down replication progression rate in the Xi reflecting the higher nucleosome stability. Moreover, macroH2A1, but not macroH2A2, regulates the number of nano replication foci in the Xi, and macroH2A1 downregulation increases DNA loop sizes corresponding to replicons. This relates to macroH2A1 regulating replicative helicase loading during G1 by interacting with it. We mapped this interaction to a phenylalanine in macroH2A1 that is not conserved in macroH2A2 and the C-terminus of Mcm3 helicase subunit. We propose that macroH2A1 enhances the licensing of pre-replication complexes via DNA helicase interaction and loading onto the Xi.

Graphical Abstract

Figure 1.

MacroH2A2 depletion results in a lower frequency of Xi replication patterns and faster Xi replication without affecting global cell cycle progression. (A) Diagram depicting the structure and splicing of the gene encoding macroH2A1 isoforms (left). Black boxes represent non-coding exons, while white boxes represent coding exons. MacroH2A1.1 specific exon is in purple and macroH2A1.2 specific exon is in grey. To the right, is a schematic representation of the three macroH2A variants and their domain architecture. (B) Cell cycle analysis in C2C12 mouse myoblast stable knockdown cell lines: the barplots show the quantification of doubling time and S-phase duration. N-numbers (fixed cells): 673, 551, 449 (Doubling time); N-numbers (live-cells): 11 (S-phase duration). Three independent replicates. (C) Xi replication pattern quantification is expressed as a percentage from the total S-phase cell population, in both stable knockdown and knockout cell lines for macroH2A. N-numbers (cells)/Replicates: Scramble 1524/9, mH2A1 KD 1492/7, mH2A2 KD 1578/9, WT 521/4, mH2A1 KO 510/3, mH2A2 KO 523/3. (D) Scheme showing the well-conserved spatiotemporal dynamics of DNA replication and the different replication patterns over the cell cycle (early, mid, and late), distinguishable by EdU or PCNA (replication machinery component) signal. (E) Fluorescence microscopy images of immunofluorescence detection of H3K27me3, EdU replication labeling, and DAPI. Some examples of the different S-phase patterns illustrated in (D) are shown for C2C12 cells, marked with a gray box. A magnified image of a mid-S-phase cell (right) shows replicating Xi chromosomes (black arrowheads) with H3K27me3 accumulation. (F) Living cells expressing mRFP-PCNA (red) as a marker for sites of active DNA replication and MaSat-GFP (green) as a marker of late-replicating mouse constitutive heterochromatin (chromocenters) were imaged at 20 min intervals for several hours using a spinning disk confocal microscope. Total replication times and duration of the Xi replication could be visualized and quantified over time by the appearance of the Xi synchronous replication pattern during the mid-S-phase (yellow boxes). N-numbers (live-cells): 4, 6, 6 (S-phase); 32, 26, 28 (Xi replication). Two independent replicates. Barplots show the average value of the distribution and the whiskers represent the standard error with a 95% confidence interval. Statistical significance was tested with a paired two-sample Wilcoxon test (n.s., not significant, is given for P-values ≥ 0.05; one star (*) for P-values <0.05 and ≥0.005; two stars (**) is given for values <0.005 and ≥0.0005; three stars (***) is given for values <0.0005). N-numbers and P-values are shown in Supplementary Table 8 (Statistics). Scale bars = 5 μm.

Figure 2.

MacroH2A1 and macroH2A2 depletion increases replication fork speed. (A) Scheme showing the fundamentals and the pipeline of the analysis. DNA synthesis speed was measured by calculating the ratio of the total nucleotide signal incorporated on the Xi during a 20-min labeling pulse (as a proxy for the amount of DNA synthesized during the indicated period) to the total Xi PCNA signal (as a marker for the number of replisomes). Representative images and image analysis procedures are shown. (B) Barplots showing the average values of the ratio EdU/PCNA within the Xi for stable knockdowns and knockout cells. N-numbers (cells): Scramble 174, mH2A1 KD 150, mH2A2 KD 174, eight independent replicates; WT 22, mH2A1 KO 32, mH2A2 KO 26, two independent replicates. (C) Barplots showing the average values of EdU/PCNA ratios for the full nucleus. In both (B) and (C), the values of EdU/PCNA ratios were normalized to the average of control cells (Scramble or WT). N-numbers (cells): Scramble 28, mH2A1 KD 23, mH2A2 KD 26, two independent replicates. (D) Scheme and pipeline of the experiment: as a proxy for helicase activity was analyzed by measuring single-stranded binding protein (GFP-RPA) accumulation on the replicating Xi. Cells were double-transfected as indicated 24 h before imaging. RPA accumulation at replicating Xi was calculated as depicted in the formula and normalized to the average of the respective pretreatment control. (E) Line Plots corresponding to the analysis of (D), showing the mean values for RPA accumulation (Cv ± standard deviation in the whiskers). Cells were imaged as depicted. N-numbers (live-cells): Scramble 18, mH2A1 KD 11, mH2A2 KD 12, two independent replicates. (F) A diagram represents the main output of these experiments: macroH2A depletion is directly associated with an increase in replication fork speed, which relates to the higher stability of macroH2A-containing nucleosomes. Barplots show the average value of the distribution and the whiskers represent the standard error with a 95% confidence interval. Statistical significance was tested with a paired two-sample Wilcoxon test (n.s., not significant, is given for P-values ≥ 0.05; one star (*) for P-values <0.05 and ≥0.005; two stars (**) is given for values <0.005 and ≥0.0005; three stars (***) is given for values <0.0005). N-numbers and P-values are shown in Supplementary Table 8 (Statistics). Scale bars = 5 μm.

Figure 3.

MacroH2A1 depletion reduces the number of active Xi replication origins. (A) Pipeline of the experiment and image analysis for nanoRFi using 3D-SIM. Briefly, cells were incubated with EdU for 20 min to label replicating DNA, fixed and stained for H3K27me3, EdU and DAPI for DNA counterstaining. Then, C2C12 stable knockdowns and primary dermal fibroblast knockouts were imaged using super-resolution microscopy (3D-SIM). Images were preprocessed using FIJI, H3K27me3 signal was used for Xi segmentation, and nanoRFi were counted using Volocity software. (B) Results of the image analysis described in (A) for C2C12 stable knockdowns (left) and primary dermal fibroblast knockouts (right). Representative 3D-SIM images of control cells are shown in each case (top), EdU (green), and DAPI (gray). Amplified regions (white boxes) show the nanoRFi on the Barr body (Xi). Below the images, barplots show the mean number of nanoRFi within the Xi quantified in super-resolution microscopy images in stable knockdowns (left) and knockouts (right) cell lines. N-numbers (cells)/replicates: Scramble 18/3, mH2A1 KD 25/3, mH2A2 KD 22/3, WT 4, mH2A1 KO 6, mH2A2 KO 10. (C) Pipeline of the experiment and image analysis for hypotonically resolved nanoRFi. Briefly, after EdU incubation as above, cells were hypotonically treated in KCl solution (75 mM) for 30 min and then cytospined. H3K27me3 or X-FISH was performed for Xi segmentation in C2C12 stable knockdown cells, followed by EdU detection and DNA counterstaining with DAPI. Confocal images were acquired, and FIJI was used for image preprocessing and counting of nanoRFi using the plugin 3D foci picker. (D) Results of the image analysis described in (C) using H3K27me3 for Xi segmentation (left) or X-FISH (right). Representative confocal images of control (Scramble) cells are shown in each case (top), EdU (green), and DAPI (gray). Amplified regions (white boxes) show the nanoRFi on the Barr body (Xi). Below the images, barplots show the mean number of hypotonically resolved nanoRFi within the Xi quantified in confocal microscopy images. N-number (cells)/replicates: Scramble 34/2, mH2A1 KD 30/2, mH2A2 KD 23/2 (H3K27me3); Scramble 14, mH2A1 KD 7, mH2A2 KD 8 (X-FISH). Barplots show the average value of the distribution and the whiskers represent the standard error with a 95% confidence interval. Statistical significance was tested with a paired two-sample Wilcoxon test (n.s., not significant, is given for P-values ≥0.05; one star (*) for P-values <0.05 and ≥0.005; two stars (**) is given for values <0.005 and ≥ 0.0005; three stars (***) is given for values <0.0005). N-numbers and P-values are shown in Supplementary Table 8 (statistics). Scale bars: 5 μm and 1 μm in the amplified region.

Figure 4.

MacroH2A1 depletion switches the replication timing of the Xi to an earlier time in S-phase. (A) Living C2C12 cells expressing mRFP-PCNA (red) as a marker for sites of active DNA replication and MaSat-GFP (green) were imaged at 20-min intervals for several hours using a spinning disk confocal microscope. The start of replication, start, and duration of the Xi replication (light grey boxes highlighted with yellow square), and total replication times were quantified in minutes (number of frames*20 min) and plotted as a barplot. N-numbers (Live-cells): Scramble 33, mH2A1 KD 32, mH2A2 KD 27 (Xi starts); Scramble 16, mH2A1 KD 23, mH2A2 KD 25 (Xi replication time). Three independent replicates. Barplots show the average value of the distribution and the whiskers represent the standard error with a 95% confidence interval. Statistical significance was tested with a paired two-sample Wilcoxon test (n.s., not significant, is given for P-values ≥0.05; one star (*) for P-values <0.05 and ≥0.005; two stars (**) is given for values <0.005 and ≥0.0005; three stars (***) is given for values <0.0005). P-values are shown in Supplementary Table 8 (Statistics). The timing and appearance of the Xi synchronous replication pattern (yellow squares) can be visualized in the representative images in (B), showing the premature Xi replication in macroH2A1 knockdown cells. Dash red lines indicate the start of Xi replication for each condition. See also full Supplementary Movies S1–S3. Scale bars = 5 μm.

Figure 5.

MacroH2A1-depleted cells show wider and irregular DNA Halos, corresponding to an increase in the size of chromatin loops and replicon units. (A) Pipeline of the experimental procedure. Cells were incubated with high-salt extraction buffers and cytospined onto coverslips. Afterward, nuclear DNA (Scaffold) and DNA in the Halos were stained with DAPI, followed by incubation with Halo buffer, washing buffers, and fixation. After imaging the cells using fluorescence microscopy, the image analysis was performed as depicted in (B) Nuclear scaffold and total area (Nuclear scaffold + Halo) were both thresholded, and their areas and circularity were measured. Areas were subtracted (At-As) and then the Halo radius was calculated. Results of these analyses are shown in (C) as boxplots: increase in the Halo radius area for macroH2A1 depleted cells (left), and decrease in circularity compared with control (right). Representative images for each condition are shown below the boxplots. N-numbers (cells): Scramble 56, mH2A1 KD 60, mH2A2 KD 61, three independent replicates. (D) X-FISH image analysis scheme performed by applying line-profile analysis in FIJI. Briefly, X-FISH fluorescence intensities were measured in a line drawn from the nuclear scaffold border (0, blue oval), up to 20 microns over the X chromosome cloud (red). Representative images are shown below the scheme. (E) Line plots showing the results of the quantification explained in (D). Line plots show the average normalized X-FISH intensities at each point/pixel of the line, and their position (in microns) relative to the nuclear scaffold. The error bands show the respective standard deviation. 95% confidence intervals are indicated in the plot as a band. N-numbers (cells): Scramble 81, mH2A1 KD 82, mH2A2 KD 75, two independent replicates. (F) Violin plots showing values of Xi relative area, with a significant increase for macroH2A1 knockdown cells prompting Xi decondensation (G) Scheme representing the relationship between DNA Halos radius and chromatin loops and the role of macroH2A1: its involvement in chromatin loops formation determines the space between replication origins (interorigin distances) affecting the number of replication origins within the Xi. For all boxplots, the box represents 50% of the data, starting in the first quartile (25%) and ending in the third (75%). The line inside represents the median. The whiskers represent the upper and lower quartiles. The violin plot depicts the density curves of the numeric data. Statistical significance was tested with a paired two-sample Wilcoxon test (n.s., not significant, is given for P-values ≥0.05; one star (*) for P-values < 0.05 and ≥ 0.005; two stars (**) is given for values <0.005 and ≥0.0005; three stars (***) is given for values <0.0005). N-numbers and P-values are shown in Supplementary Table 8 (statistics). Scale bars: 5 μm.

Figure 6.

Counterbalance between active origins and replication progression rate. (A) Summary of the mean values obtained during the analysis of Xi replication dynamics on C2C12 stable knockdown cells (S-phase duration, Xi replication duration, Xi replication fork speed, number of nanoRFi, and DNA Halo radius). Relevant changes compared with control (Scramble cells) are highlighted in red. (B) A diagram illustrating the trend of these results is shown: nucleosome stability is anti-correlated with replication fork speed, while macroH2A1 is correlated with replication synchrony. Synchrony and replication fork speed affect Xi′s replication rate. (C) Table summarizing the estimated replication fork speed in control versus macroH2A depleted cells taking into consideration: the size of the X chromosome, the duration of Xi replication, and in a subsequent step the number of active origins and their bi-directionality.

Figure 7.

MacroH2A1.1 and macroH2A1.2, but not macroH2A2, interact with the DNA helicase. (A) Diagram illustrating the fundamentals of the proximity ligation assay (PLA) to analyze the in situ interaction between macroH2A1 and Mcm2, including a pipeline of image analysis. In this technique, small oligonucleotide probes, (+) and (–), conjugated to secondary antibodies specifically recognize the primary antibodies against the proteins of interest. When the two probes are closer than 40 nm, ligation by ligase incubation can occur. This generates circular DNAs that will be amplified by a polymerase incorporating fluorescently labeled nucleotides. Afterward, fluorescent spots can be detected and quantified using microscopy and image analysis, considering each spot an interaction site between the two proteins. The number of spots and their location, the Xi in this case, can be quantified by image analysis using FIJI. Negative control was performed using only one of the primary antibodies. (B) Representative confocal images of the PLA assay between macroH2A1 and Mcm2, showing spots corresponding to protein-protein interaction for control and macroH2A2 knockdown cell lines. (C) Barplots showing mean values for the number of PLA spots per nuclei in knockdown cell lines after performing the image analysis described in (A). N-numbers (cells): Scramble 27, mH2A1 KD 34, mH2A2 KD 33, two independent replicates. (D) In addition to the previous analysis, the density of spots was calculated for the full nuclei and the Xi, dividing the number of spots counted in their respective areas by values of the sum intensity of DAPI (as a proxy for DNA content). In this case, all mean values were normalized by the average value of nuclear density for control (Scramble) cells. Barplots show these quantifications, with a significantly higher density of spots for the Xi. N-numbers (cells): Scramble 21, mH2A1 KD 23, mH2A2 KD 23, two independent replicates. (E) The same experimental pipeline of (C) was used to analyze the interaction between macroH2A1.2, macroH2A2, and Mcm2. In this case, high-content screening microscopy was used for imaging, and quantification of spots was performed using the Harmony software. The results of these analyses are shown in the barplot, where macroH2A1.2/macroH2A2-Mcm2 interaction is compared with the interaction macroH2A1-Mcm2 analyzed before. In addition, macroH2A-Mcm2 interaction is compared with the negative control of the assay and with a different cell line, MEFs. N-numbers (cells)/replicates: 7356/2, 3980/1, 11 707/2, 2816/1, 4265/2, 4762/2, 11 617/2, 6710/1. (F) Representative confocal images of the PLA assay on (E), showing spots corresponding to protein-protein interaction for macroH2A1/macroH2A1.2-Mcm2. (G) Co-immunoprecipitation: C2C12 cells were transfected with EGFP or EGFP-tagged macroH2A1.1, macroH2A1.2, or macroH2A2. Cell extracts were analyzed by immunoprecipitation with immobilized GFP-binding nanobody, followed by detection with antibodies against GFP, Mcm2, Mcm4 and Mcm5 (Supplementary Figure S9A). The cut-outs show input/bound GFP and input/bound Mcm fractions. To the right, the scheme of the DNA helicase shows the two hexamers. The position of Mcm2 subunits on the Mcm hexamers is indicated with a red star. Two additional replicates for these co-immunoprecipitations are shown in Supplementary Figure S9B-C. Barplots show the average value of the distribution and the whiskers represent the standard error with a 95% confidence interval. Statistical significance was tested with a paired two-sample Wilcoxon test (n.s., not significant, is given for P-values ≥0.05; one star (*) for P-values <0.05 and ≥0.005; two stars (**) is given for values <0.005 and ≥0.0005; three stars (***) is given for values < 0.0005). N-numbers and P-values are shown in Supplementary Table 8 (statistics). Scale bars: 5 μm.

Figure 8.

MacroH2A1 isoforms interact with Mcm3 through a conserved Phe residue in their macro domains. (A) Schematic illustrating the protein fragments used for AF modeling (full-length Mcm5, 6 and 7 are not shown). macroH2A1.2_D is identical in sequence to macroH2A1.1_D why it was omitted. (B) Heatmap showing the model confidences obtained for AF structural models. Labels on the x and y axis indicate the paired protein fragments for structural modeling. Gray fields indicate fragment pairs that were not subjected to structural modeling apart from macroH2A1.1 paired with Mcm5 for which structural modeling failed. (C) Superimposition of structural models obtained for the macroH2A domains paired with Mcm3_O2. macroH2A domains are shown in green colors as in A, Mcm3_O2 is shown in different shades of blue. Individual structural models are shown in Supplementary Figure S11. (D) Zoom into the interface between the macro domains of macroH2A1.1 and 1.2 and Mcm3_O2 with key residues shown as sticks. (E) Co-immunoprecipitation experiments: C2C12 cells were transfected with EGFP or EGFP-tagged macroH2A1.1, macroH2A1.2, or macroH2A2. Cell extracts were analyzed by immunoprecipitation with immobilized GFP-binding nanobody, followed by detection with antibodies against GFP and Mcm3. (F) Co-immunoprecipitation experiments were performed as described for (E) but replacing EGFP-tagged macroH2A1.1, macroH2A1.2, or macroH2A2 for their respective mutants: macroH2A1.1-F192V, macroH2A1.2-F192V or macroH2A2-V192F. For (E) and (F), the cut-outs show input/bound GFP and input/bound Mcm fractions. To the right, the scheme of the DNA helicase shows the two hexamers. The position of Mcm3 subunits on the hexamers is indicated with a red star. Two additional replicates for these co-immunoprecipitations are shown in Supplementary Figure S12.

Figure 9.

MacroH2A1 depletion, but not macroH2A2, affects the loading of Mcm2 to pre-RCs in the Xi. (A) Scheme showing a simplified summary of the different steps of origins activation during G1, from the assembly of unlicensed pre-RC (starting with ORC, Cdc6 and Cdt1-Mcm), to the licensing of pre-RC by the loading of the second hexamer of the DNA helicase Mcm, and their activation to pre-IC at the end of G1 and during S-phase. (B) Experimental pipeline and image analysis. C2C12 stable knockdowns were synchronized by mitotic shake-off and seeded onto coverslips. One hour after seeding, eight different time points were collected in one-hour intervals to track G1 progression over time. Cells were extracted and fixed, and afterward immunostained against Mcm2 and H3K27me3. Cells were imaged using confocal microscopy and image analysis was performed with FIJI. H3K27me3 signal was used for Xi ROI selection and calculating Mcm2 loading coefficients (as depicted in the formula). (C) Mcm2 loading coefficients for each G1 time point were normalized by the average of control cells at time 1 (t1) and plotted, obtaining Mcm2 G1 loading curves. N-numbers (cells): Scramble 24–28, mH2A1 KD 15–19, mH2A2 KD 15–25, two independent replicates. Representative images for selected time points are shown in for Scramble cells (D). The full gallery of images is shown in Supplementary Figure S14. (E) Boxplots showing Xi levels of Mcm2 (top) and Mcm2-phosphoS108 (bottom) in extracted and non-synchronized cells imaged using confocal microscopy. Cells replicating the Xi were selected for image analysis as described in (B). Boxplots show relative Mcm2 sum intensity in the replicating Xi, normalized by the average of control cells. N-numbers (cells): Scramble 13, mH2A1 KD 16, mH2A2 KD 15 (Mcm2); 20, 20, 29 (Mcm2-phosphoS108). (F) Table summarizing the changes (in percentage) for Mcm2 and Mcm2phosphoS108 levels in the Xi, and the number of nanoRFi (relative to control cells). A reduction of 25%/40.9% on final Mcm2phosphoS108/Mcm2 loading in the Xi was in line with a reduction of 31.3% in the number of replication nanofoci (aka, active origins) for macroH2A1 knockdown cells. Significant changes are highlighted in red. (G) ORC1 loading coefficients for each G1 time point were normalized by the average of control cells at time 1 (t1) and plotted, obtaining ORC1 G1 loading curves. N-numbers (cells): Scramble 42–49, mH2A1 KD 40–44, mH2A2 KD 42–43, four independent replicates. The same experimental approach was followed for Cdc6 (N-numbers (cells): Scramble 22–29, mH2A1 KD 22–24, mH2A2 KD 22–24) (H), and Cdt1 (N-numbers (cells): Scramble 20–22, mH2A1 KD 19–23, mH2A2 KD 19–20) (I), two independent replicates. All line plots show normalized average fluorescence values, and error bands show the respective standard deviation. 95% confidence intervals are indicated in the plot as a band. For all boxplots, the box represents 50% of the data, starting in the first quartile (25%) and ending in the third (75%). The line inside represents the median. The whiskers represent the upper and lower quartiles. Statistical significance was tested with a paired two-sample Wilcoxon test (n.s., not significant, is given for P-values ≥0.05; one star (*) for P-values <0.05 and ≥0.005; two stars (**) is given for values <0.005 and ≥0.0005; three stars (***) is given for values <0.0005). N-numbers and P-values are shown in Supplementary Table 8 (statistics). Scale bars: 5 μm.

Figure 10.

Model representing the role of macroH2A1 in Xi replication dynamics. Replication dynamics depend on two different factors that together affect replication rate: synchrony or number of active origins at a certain time, and replication fork speed progression. The first one rests on the assembly of different components of the pre-replication complexes (pre-RCs) during G1, part of which is the replicative DNA helicase Mcm. We propose that the isoform-specific association of macroH2A1 with some of the Xi replication origins is a factor regulating the formation of chromatin loops and the accessibility of these origins to pre-RC assembly. Hence, Mcm chromatin loading is reduced after macroH2A1 depletion. The reduced Mcm G1 loading turns into less Mcm2 in the inactive X chromosome during S-phase, and consequently less Mcm2-phosphoS108, the active form of this helicase during replication. Finally, this causes a decrease in the number of active origins, which negatively affects the replication rate. However, the effects of macroH2A1 depletion are in part counterbalanced by the impact of both macroH2A isoforms in replication fork speed, by hampering the progression of the replication fork machinery, as shown in Figure 2.