Non-redundant functions of H2A.Z.1 and H2A.Z.2 in chromosome segregation and cell cycle progression

H2A.Z is a H2A-type histone variant essential for many aspects of cell biology ranging from gene expression to genome stability. From deuterostomes, H2A.Z evolved into two paralogues H2A.Z.1 and H2A.Z.2 that differ by only three amino acids and are encoded by different genes (H2AFZ and H2AFV respectively). Despite the importance of this histone variant in development and cellular homeostasis, very little is known about the individual functions of each paralogue in mammals. Here we investigated the distinct roles of two paralogues in cell cycle regulation. Using a specific siRNA approach for each paralogue in human cells, we unveiled non-redundant functions for H2A.Z.1 and H2A.Z.2 in cell division: H2A.Z.1 regulates the expression of important cell cycle genes (including Myc and Ki-67) and its depletion leads to a G1 arrest, whereas H2A.Z.2 is essential for centromere integrity and function, thus playing a key role in chromosome segregation.


SUMMARY
H2A.Z is a H2A-type histone variant essential for many aspects of cell biology ranging from gene expression to genome stability. From deuterostomes, H2A.Z evolved into two paralogues H2A.Z.1 and H2A.Z.2 that differ by only three amino acids and are Although H2A.Z knockdown leads to early embryonic lethality in Drosophila 10 and mice 11 , depletion of the H2A.Z orthologue in S. cerevisiae, HTZ1, is not lethal 12 , pointing out to a possible difference on the role of H2A.Z among species.
H2A.Z is also linked to heterochromatin regulation. Recent evidence suggests that H2A.Z and H3K9me3, a known marker for heterochromatin, can cooperate to enhance the binding of Heterochromatin Protein 1 alpha (HP1α) to chromatin in vitro 22 24 25 .
However, the specific contribution of each paralogue towards these quite different aspects of chromatin biology is currently unknown. Although H2A.Z.1 and H2A.Z.2 are distributed similarly around the nucleus and are subjected to comparable posttranslational modifications, their 3D structure, genome localisation and tissue distribution appears to be quite different 7,8 . In fact, H2A.Z.2 does not compensate for the loss of H2A.Z.1 in vivo, as H2A.Z.1 knock-out is lethal in mouse 11 .
To date, very few studies have attempted to investigate and differentiate the specific roles of these two paralogues in vertebrates. In DT40 chicken cells, knock-out of H2A.Z.2 results in a slower cell proliferation rate compared to the wild type and H2A.Z.1 knock-out cells 26 , while in humans, Floating-Harbor syndrome 27 and malignant melanoma 28 have been specifically linked to H2A.Z.2 27 . In addition, both variants seem to play independent roles in the transcription of some genes involved in the response to neuronal activity 29 . Moreover, very recently, it was shown that the differences between H2A.Z.1 and H2A.Z.2 on transcription regulation seems to depend more on the relative level of the two paralogues rather than on their chromatin localisation 30 . However, we are still missing a full understanding of the role of each variant in human cells.
As the role of histone variants in genome organisation and regulation becomes more appreciated and emerging studies have linked H2A.Z to cancer, it is important to investigate and clarify the possible differential roles of the H2A.Z paralogues and splice variants for cell cycle regulation in vivo. This will not only provide a better understanding of their function, but it will also clarify their relative contribution to divergent aspects of chromatin biology. In this study we used siRNA to specifically

H2A.Z.2, but not H2A.Z.1, is necessary for heterochromatin maintenance in human cells
Several studies have linked H2A.Z to heterochromatin in different systems but it is still unclear if a subset of histone variants is particularly enriched at these genomic regions.
In order to study the chromatin flavour of HP1-bound nucleosomes in an unbiased manner, we used a TAG-Proteogenomic approach recently developed in the lab 31 ( Supplementary Figure 1 A). Briefly, HeLa nucleosomes prepared from SILAC-light labelled cells were incubated with recombinant GST:HP1 or GST alone. After elution and SDS-PAGE, the histone fractions were analysed by mass spectrometry. The SILAC heavy-labelled nucleosome fraction was used to normalise the abundance of the different histones before the binding to the GST column (Figure 1 A). Next, we analysed the abundance of different histone variants present in the GST:HP1-bound fraction and compared it to the input (I) and GST alone. We used another chromatin- It is well established that HP1 binds to H3K9me3 to form a repressive chromatin environment 34,35 , but recent work indicates that H2A.Z is also able to enhance in vitro HP1 binding to nucleosome arrays that do not contain H3K9me3 24 25 . If an interplay between H2A.Z, H3K9me3 and HP1 exists in vivo, it is not known. However, studies have shown that in mouse cells lacking Ki-67 and presenting a significant decrease in H3K9me3, HP1 maintains its localisation; this suggests that alternative mechanisms exist to maintain HP1-binding at most foci, but the nature of this parallel system is not known. We therefore asked whether H2A.Z could impact on H3K9me3 levels in vivo, as this has never been reported. We analysed the levels of H3K9me3 by In conclusion, our results suggest that the H2A.Z.2.1 variant is playing a key role in chromosome segregation, and that this is also mediated by its post-translational modification status.

H2A.Z.2 regulates sister chromatids cohesion
We next aimed to understand the molecular mechanisms underlying the micronuclei Reduced CENP-A levels could be at the basis of a dysfunctional kinetochore and lead to the decrease in cohesin and Aurora B; alternatively, the two pathways could be independent. To better understand this aspect, we analysed the correlation between Sgo1 and CENP-A levels at centromeres of H2A.Z.2-depleted cells: a component on the dependency seems to correlate with CENP-A levels but we cannot exclude that other factors are also playing a role in the amount of Sgo1, particularly at the higher levels ( Microtubules are yet other major effectors impacting on nuclear morphology with a number of studies reporting their importance for normal nuclear shape in interphase 49,50 . We analysed -tubulin distribution in the H2A.Z.1-depleted cells and observed an accumulation of tubulin at the points of invaginations (Figure 4 E, yellow arrows).
Although more investigations will be required to unveil the underlying mechanisms, this data suggests that unbalanced mechanical forces possibly account for the nuclear deformity in the H2A.Z.1-knockdown cells.

H2A.Z.2 is essential for chromosome segregation
Using an unbiased proteogenomic approach, we have shown that HP1 preferentially is still unknown but being an alanine that cannot be phosphorylated, (compared to the threonine in the paralogue counterpart), it could suggest that a differential phosphorylation between the two forms may play a critical role, especially during mitosis. In our experiments all these mutants were efficiently incorporated into the chromosomes, suggesting that the lack of rescue can be linked to either the ability of generating a different chromatin conformation or to the recruitment of specific effectors.
We have also investigated the molecular mechanisms behind the chromosome segregation defects. H2A.Z.2-depleted cells presented impaired loading of CENP-A, Sgo1 (protector of centromeric cohesion) and Aurora B (member of the CPC).
Interestingly, two H2A.Z specific peptides (GDEELDSLIK and ATIAGGGVIPH -note that these sequences are present in both H2A.Z.1 and H2A.Z.2) were enriched in the CENP-A nucleosome pull-downs when compared with those containing histone H3.1 59 . This suggests that a specific centromeric CENP-A containing chromatin could favour the maintenance or deposition of CENP-A. Interestingly, the majority of H2A.Z deposition at pericentric heterochromatin occurs in G1 60 , at the same time as CENP-A. Treatment with 5'-aza 2'-deoxycytidine (5-Aza) leads to an increased incorporation of both CENP-A and H2A.Z at pericentric heterochromatin that has been interpreted as a consequence of a disruption in the heterochromatin at the centromere 60 . However, our data seems to suggest that H2A.Z.2 incorporation is upstream and can affect CENP-A deposition or stability. This is a novel and interesting aspect that will be further investigated.
We have also observed a decrease in Aurora B levels at centromeres of mitotic Interestingly, a few studies have pointed at the possibility that it is the balance between the two paralogues that allows for a normal cell proliferation. In this respect, as we have shown that H2A.Z.2 depletion resulted in chromosome instability while H2A.Z.1 supports the expression of master cell cycle genes, we can envisage that slight changes in their balance could easily produce an aberrant cancer prone phenotype, on one hand by supporting proliferation and on the other by increasing genome instability, both beneficial for a tumour cell.
In conclusion, we have demonstrated non-redundant roles for H2A.Z.1 and H2A.Z.2 in different aspects of cell biology and gene expression, highlighting the importance of studying these variants independently. More studies will require detailed analyses of the downstream effectors of these paralogues, the chaperones and machineries dedicated to the specific deposition of these variants. This is particularly important for therapeutic purposes, as only targeting the correct variant would offer a proper intervention.
The YL1 sequence was obtained by PCR using the primers in Table 1 and cloned into GFP:LacI 31 by HindIII/BamHI. GFP was replaced by RFP using NheI/BglII.
The primers used for the study were acquired from Eurofins Genomics (Germany) and all the restriction enzymes from New England Biolabs (UK).

Immunofluorescence microscopy
Cells were fixed in 4% PFA and processed as previously described (Vagnarelli et al., 2006). Primary and secondary antibodies were used as in Table 2 For quantification of enrichment at the LacO locus, four circles were designed: one around the LacI spot, two within the nucleus and one outside the cell. Signals intensities were extracted and the mean of the two circles within the nucleus calculated. The outside circle was used as background and subtracted from both the mean nuclear and the LacI spot, then the signal intensity from the LacI was normalized relative to the intensity of the nuclear signal.
Violin plots were generated using the ggplot2 package in R.

Preparation of HP1-bound nucleosomes and mass spectrometry
Preparation of HP1-bound nucleosomes and mass spectrometry were performed as For the first search, peptide tolerance was set to 20 ppm while for the main search peptide tolerance was set to 4.5 pm. Isotope mass tolerance was 2 ppm and maximum charge to 7. Digestion mode was set to specific with trypsin allowing maximum of two missed cleavages. Carbamidomethylation of cysteine was set as fixed modification.
Oxidation of methionine and acetylation of the N-terminal were set as variable modifications. Multiplicity was set to 2 and for heavy labels Arginine 10 and Lysine 8 were selected. Peptide and protein identifications were filtered to 1% FDR. The Perseus computational platform 85 version 1.6.0.2 was used for the statistical analysis of the MaxQuant-generated datasets and for the creation of the volcano plots.
The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD018719.

FISH
FISH was performed in HeLa cells as previously described 31 using a probe against chromosome 17 86 .
For the centromere position analyses, a single plane containing each spot was selected. The distance of the spots from the periphery was measured and represented as a fraction of the radius on which each spot belongs.

Cell cycle analysis with flow cytometry
Cells were trypsinised, resuspended and incubated at room temperature for 30 min in If two or more H2A.Z regions overlap, they were considered as one region. For each set of genes, the H2A.Z regions that were not within +/-5 kb of any of the genes in that set were excluded, and the remainder were classified into 3 categories: regions that overlap and/or are within +/-3 kb of one or more of the TSS associated with a gene whose expression changes after RNAi knockdown; regions that overlap the loci of genes whose expression changes after RNAi knockdown but not any of the TSS associated with those genes; and regions that do not overlap the TSS nor the loci of genes whose expression changed after RNAi knockdown. The number of H2A.Z regions in each category is divided by the total number of regions for that gene set to give the proportion of H2A.Z regions in each category.

Differentially expressed genes frequency analyses
To calculate the expected and observed frequency of differentially expressed genes on each chromosome, we used the gene numbers present on each human chromosome and used the chromosome copy number in HeLa cells integrating both the published datasets (10.1371/journal.pone.0029225) and our own experimental dataset for chromosomes 17 (this study) and chromosomes 13 and 14 31 .

Statistical analyses
Statistical analyses were performed either in Excel (Chi-square test), or in R (using the Wilcoxon rank test function, differential expression, lowess smoothing).