The H3 histone chaperone NASP escorts CenH3 in Arabidopsis

Samuel Le Goff , Burcu Nur Kec eli , Hana Je r abkov a , Stefan Heckmann , Twan Rutten , Sylviane Cotterell , Veit Schubert , Elisabeth Roitinger , Karl Mechtler , F. Christopher H. Franklin , Christophe Tatout , Andreas Houben , Danny Geelen , Aline V. Probst* and Inna Lermontova* GReD, Universit e Clermont Auvergne, CNRS, INSERM, BP 38, 63001, Clermont-Ferrand, France, Department of Plants and Crops, Unit HortiCell, Faculty of Bioscience Engineering, Ghent University, Coupure links, 653, 9000, Ghent, Belgium, The Czech Academy of Sciences, Institute of Experimental Botany (IEB), Centre of the Region Han a for Biotechnological and Agricultural Research, Slechtitel u 31, 78 371, Olomouc, Czech Republic, Leibniz Institute of Plant Genetics and Crop Plant Research (IPK) Gatersleben, Corrensstrasse 3, D-06466, Seeland, Germany, Institute of Molecular Pathology (IMP), Vienna BioCenter (VBC), Vienna 1030, Austria, Institute of Molecular Biotechnology (IMBA), Austrian Academy of Sciences,Vienna BioCenter (VBC), Vienna 1030, Austria, Gregor Mendel Institute (GMI), Austrian Academy of Sciences, Vienna BioCenter (VBC), Vienna 1030, Austria, School of Biosciences, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK, and Mendel Centre for Plant Genomics and Proteomics, CEITEC, Masaryk University, Brno CZ-62500, Czech Republic


INTRODUCTION
Eukaryotic DNA is organized into chromatin using histones as components of its building blocks, the nucleosomes. The incorporation of different histone variants into the nucleosome determines nucleosome stability, DNA accessibility and higher order chromatin organization. The histone variant CenH3 is unique in that its incorporation is limited to the centromeres (Talbert et al., 2002;Lermontova et al., 2015;Rosin and Mellone, 2017). CenH3 deposition is a pre-requisite for centromere formation and in turn kinetochore assembly, ensuring equal partitioning of genetic material between daughter cells during cell division.
Escort, deposition and eviction of the different histone variants depend on histone chaperones that play an important role in defining discrete chromatin landscapes important for genome function, stability and cell identity (Hammond et al., 2017). These different histone chaperones, defined 'as factors that associate with histones and stimulate a reaction involving histone transfer without being part of the final product' (De Koning et al., 2007), execute distinct functions in an interaction network, which has been extensively characterized in animals and yeast. Recently, 22 and 25 genes encoding histone chaperones in Arabidopsis and rice, respectively, have been identified (Tripathi et al., 2015).
Here we identify NASP SIM3 as a binding partner of Arabidopsis CenH3. Highly expressed in actively dividing tissues, with a similar pattern as CenH3, NASP SIM3 does not tightly interact with chromatin suggesting that it plays a role in escorting non-nucleosomal CenH3 histones. In line with this role, reducing NASP SIM3 expression negatively affects CenH3 deposition at centromeres.

NASP SIM3 binds CenH3
To screen for yet unknown interactors of Arabidopsis CenH3, we performed immunoprecipitation coupled to mass spectrometry. To this aim we used inflorescences of plants expressing CenH3 as an EYFP fusion (the genomic coding region of CenH3 fused to the EYFP coding sequence) and precipitated gCenH3ÀEYFP with an anti-GFP antibody before submitting the co-immunoprecipitated proteins to mass spectrometry analysis. Plants expressing EYFP alone served as a negative control. In four independent gCenH3ÀEYFP mass spectrometry samples we identified between six and 14 peptides (coverage 15.65-35.57%) that corresponded to NASP SIM3 (AT4G37210) and which were not detected in the EYFP control samples (Figures 1a and S1a). Altogether 52 peptides matching NASP SIM3 with a combined coverage of 41.06% were detected, strongly suggesting that NASP SIM3 can be found in a complex with CenH3 in vivo. In an independent approach to identify CenH3 binding proteins, tandem affinity purification (TAP) experiments coupled to mass spectrometry analysis were performed using GSRhinoÀCenH3 and GSRhinoÀH3.1 proteins expressed under control of the 35S promoter in actively dividing Arabidopsis cell culture suspensions (Figure 1a and Supporting Information Figure S1a). In four independent pull down samples of both CenH3 and H3.1 eight to 10 peptides (coverage 20.9-26.6%) and nine or 10 peptides (coverage 20.1-30.1%), respectively, corresponding to NASP SIM3 have been identified (Figure S1b), while none was identified in the TAP assays with the GSRhino tag alone as bait. These results consolidate that CenH3 interacts in vivo with NASP SIM3 .
To determine whether NASP SIM3 interacts directly with CenH3 or is part of a complex including CenH3, we analyzed interactions of NASP SIM3 with different histone proteins using yeast-two-hybrid (Y2H) assay. Yeast strains expressing NASP SIM3 as bait, and H3.1, H3.3 or CenH3 as prey grew on selective medium, indicating that NASP SIM3 binds CenH3 (Figure 1b). CenH3 binding is specific to NASP SIM3 as the histone chaperone ASF1A, known in mammals to bind H3-H4 dimers (Natsume et al., 2007), interacts with H3.1 and H3.3, but not with CenH3 (Figure 1a). Neither NASP SIM3 nor ASF1A-expressing strains grew on selective medium when co-expressed with an empty prey vector or with a vector expressing the H2A histone variant H2A.W.6 or the H1 variant H1.3 corroborating their specificity for H3 variants. A drop dilution test revealed a strong interaction of NASP SIM3 with H3.3, followed by CenH3 and H3.1 (Figure 1c).

NASP SIM3 interacts with both C-and N-terminal parts of CenH3 in planta
To confirm the direct interaction between NASP SIM3 and CenH3 in planta, bimolecular fluorescence complementation (BiFC) assays were performed on young Agrobacterium-infiltrated tobacco (Nicotiana benthamiana) leaves (Walter et al., 2004) containing mitotic cells. We first analyzed the nuclear distribution of NASP SIM3 and CenH3 proteins when transiently expressed in tobacco. Arabidopsis NASP SIM3 fused to EYFP localized to the nucleus, showing a uniform distribution excluding nucleoli (Figure 2a). When CenH3 was expressed with an N-terminal EYFP tag, two types of nuclei could be detected: either with an exclusive localization of the EYFPÀCenH3 fusion protein in restricted spots corresponding to centromeres (Figure 2b) or both in the nucleoplasm and at centromeres (Figure 2c) showing that Arabidopsis CenH3 could be incorporated at N. benthamiana centromeres under these conditions. When NASP SIM3 fused to the C-and CenH3 to the N-terminal YFP fragments, and vice versa were co-expressed in tobacco leaves, EYFP was successfully reconstituted in nuclei of the transformed cells with both combinations demonstrating interaction between NASP SIM3 and CenH3 in planta (Figure 2d,e). BiFC signals were diffuse throughout the nucleoplasm, and no specific enrichment was observed in defined spots.
We have shown previously that CenH3DN, lacking the Nterminal domain, localizes to centromeres (Lermontova et al., 2006), indicating that the CenH3 histone fold domain is sufficient for its targeting to centromeres. Therefore, we tested which part of the CenH3 protein interacts with NASP SIM3 . To this end, the N-terminal tail (amino acids 1-70) or the histone fold domain (amino acids 71-178) of CenH3 were fused to the N-and C-terminal parts of YFP respectively, and co-expressed with full-length NASP SIM3 . BiFC signals could be detected in both cases (Figure 2f,g), showing that NASP SIM3 can interact both with the Control experiments, in which NASP SIM3 -nYFP was coexpressed with free cYFP or CenH3-cYFP with free nYFP as well as combinations of N-and C-terminal parts of CenH3 fused with N-or C-terminal parts of EYFP, respectively, did not show any fluorescence excluding non-specific interactions of NASP SIM3 or CenH3 with YFP ( Figure 2h).

Arabidopsis NASP SIM3 expressed in actively dividing tissues
Expression of Arabidopsis CenH3 is regulated by E2F, a transcription factor active in dividing tissues (Heckmann et al., 2011). As the promoter of NASP SIM3 contains a putative E2F binding site ( Figure 3a) , we speculated that it is expressed in dividing cells similar to CenH3.
To test whether Arabidopsis NASP SIM3 is preferentially expressed and the corresponding protein is accumulated in dividing tissues, we generated transgenic lines expressing a pNASP:NASP SIM3 ÀEGFPÀGUS (for b-glucuronidase) reporter gene construct (Figure 3a), and analyzed the GUS staining pattern in these plants as a readout for NASP SIM3 gene expression and protein accumulation. Although GUS staining intensity varied between independent transgenic lines, all positive lines showed a similar staining pattern. Five representative lines were chosen for detailed analysis. In 4-day-old seedlings, GUS staining intensity was highest in root tips, shoot apical meristems and cotyledons (Figure 3b). In older plantlets (7-day-old, Figure 3c; 14-day-old, Figure 3d), NASP SIM3 expression was also detected in young leaves containing actively dividing cells as previously reported (Maksimov et al., 2016). In 4-week-old plants only young, but not old fully developed leaves showed GUS staining (Figure 3e-g). In inflorescence meristems, young inflorescences and developing flower buds GUS activity was almost absent (Figure 3h, upper part) or very weak (Figure 3h, lower part) but GFP signal from the NASP SIM3 ÀEGFPÀGUS fusion protein could be detected in petals of flower buds and in the developing male gametophyte ( Figure S2a-d). However, older inflorescences displayed GUS activity of variable intensity in anthers and anther filaments (Figure 3h,i). Further data mining of some available RNA-seq datasets from different tissues and the eFP genome browser confirmed high NASP SIM3 expression in tissues comprising dividing cells such as the root meristem and the shoot apex, but also found NASP SIM3 transcripts to be abundant in flower buds  (Figure 4a, right) due to four nucleotides difference in length of exon 5, resulting in a frame shift generating a premature stop codon in exon 6. The corresponding cDNAs: CP002687.1 À At4g37210.1 and NM_202970.1 À At4g37210.2 were found in the Arabidopsis sequence database (https://www.ncbi. nlm.nih.gov/). To confirm the presence of both NASP SIM3 transcripts of Arabidopsis, we performed RT-PCR on RNA extracted from seedlings and flower buds with an isoformspecific reverse primer ( Figure S3a,b). At4g37210.1 transcripts could be detected in both seedlings and flower buds, while At4g37210.2 transcripts were detected only in floral tissues ( Figure S3b) confirming the existence of the short splice variant and suggesting tissue-specific expression.
To study the subcellular localization of the two predicted NASP SIM3 proteins we generated constructs expressing NASP.1 SIM3.1 and NASP.2 SIM3.2 cDNA fragments fused to EYFP under control of the CaMV 35S promoter. The two constructs were transiently expressed in leaves of N. benthamiana and for both NASP SIM3 isoforms EYFP fluorescence was observed in the nucleoplasm of epidermal nuclei, whereas EYFP alone was localized in the nucleus and cytoplasm ( Figure S3c). Both constructs were used to transform A. thaliana and the NASP.1 SIM3.1 ÀEYFP fusion protein was localized in the nucleoplasm of root tip nuclei ( Figure S3d), in agreement with previous reports (Maksimov et al., 2016). In contrast, no transgenic plants could be recovered on selective media for the second, shorter isoform, in three independent transformation experiments. As expression of the NASP.2 SIM3.2 isoform under the 35S promoter might result in overexpression and affect growth, genomic NASP.2 SIM3.2 fragments including endogenous promoters were cloned to generate a translational fusion with GFP and GUS reporter genes as it was described above for the NASP.1 SIM3.1 variant ( Figure 3a). We showed that the longer isoform localizes to the nucleoplasm of root tip nuclei (Figure 4b,e) as previously reported (Maksimov et al., 2016). For the shorter isoform, only a few transgenic lines showing GFP signals were obtained. Despite lacking part of the coiled-coil domain, the shorter isoform showed a similar nuclear localization to that of the longer isoform ( Figure 4c). The weak expression of NASP.2 SIM3.2 in our transgenic lines is in agreement with the mass spectrometry experiments, in which no peptide was identified that could be specifically assigned to the NASP.2 SIM3.2 isoform.
De novo incorporation of the CenH3 histone variant takes place during G2 phase of the mitotic cell cycle (Lermontova et al., 2006(Lermontova et al., , 2007. We therefore wanted to investigate whether the subcellular localization of the predominant, longer isoform NASP.1 SIM3.1 would change during the mitotic cell cycle. The NASP.1 SIM3.1 -GFP fusion is present as the same diffused nuclear stain in root tip meristems throughout G1, S and even in G2, when newly synthesized CenH3 is loaded (Figure 4e and Video S1). As during transient expression in N. benthamiana (Figure 2), we never observed a pronounced accumulation of NASP.1 SIM3.1 next to chromocenters that could indicate a direct role of NASP.1 SIM3.1 in CenH3 loading.

NASP SIM3 is highly mobile
To study the nuclear localization of Arabidopsis NASP.1 SIM3.1 in relation to chromatin, we performed immunostaining with anti-GFP antibodies on nuclei of pNASP:NASP.1 SIM3.1 -EGFP-GUS transformants and applied structured illumination microscopy. At this higher resolution, NASP.1 SIM3.1 partly co-localizes with chromatin fibers, but is excluded from chromocenters and the nucleolus (Figure 5a). However, it does not completely follow the chromatin pattern and also occurs in between chromatin fibers. Super-resolution microscopy indicates that the NASP.1 SIM3.1 -EGFP-GUS fusion protein is not associated with mitotic chromosomes (Figure 5b). To further test whether Arabidopsis NASP.1 SIM3.1 was associated with chromatin, we used BY-2 cells expressing CenH3ÀEGFP and NASP.1 SIM3.1 ÀmCherry. In these highly dividing cells, CenH3ÀEGFP localized to centromeres and NASP.1 SIM3.1 ÀmCherry showed a diffuse nuclear staining coherent with our previous observations ( Figure S4). When we permeabilized these cells using detergent to release non-chromatin bound proteins from the nucleus, Cen-H3ÀEGFP signal was still distinctly detectable in the nucleus and mostly at centromeres ( Figure S4), while the NASP.1 SIM3.1 ÀmCherry signal was visibly reduced compared to the non-treated condition. This indicates that most of the NASP SIM3 proteins were part of the soluble nuclear protein pool and did not stably associate with chromatin. To confirm this observation with a different assay, we probed the dynamics of NASP.1 SIM3.1 by fluorescence recovery after photobleaching (FRAP) and fluorescence loss in photobleaching (FLIP) (Figure 5c

NASP SIM3 interaction network
To identify other binding factors of Arabidopsis NASP SIM3 that may reinforce its role as CenH3 chaperone, we carried out a Y2H screen using the long NASP.1 SIM3.1 isoform as bait. Among the putative binding partners of NASP.1 SIM3.1 we identified three candidates with a clear link to chromatin: the histone variant H3.3, the WD40 repeat-containing protein MSI3 as well as the H2AÀH2B histone chaperone NAP1;2 (NUCLEOSOME ASSEMBLY PROTEIN 1) (Liu et al., 2009). MSI3 is a homolog of MSI1, which is a subunit of the CHROMATIN ASSEMBLY FACTOR 1 (CAF-1) complex and which has previously been identified by immunoaffinity purification of YFP-tagged NASP SIM3 (Maksimov et al., 2016) (Figure 6a). We further identified TSK-ASSOCIATING PROTEIN (TSA1) (Suzuki et al., 2005) that had been previously shown to bind to Gamma-tubulin Complex Protein 3-interacting proteins (GIPs) that in turn also interacted with CenH3 (Janski et al., 2012;Batzenschlager et al., 2013). This therefore suggests an additional link with the centromere. To validate the interaction of NASP SIM3 with NAP1;2 and TSA1, we cloned the corresponding full-length cDNAs and confirmed that NASP SIM3 , but not ASF1, interacted both with NAP1;2 and TSA1, while the interaction was stronger with NAP1;2 (Figure 6b,c).

NASP SIM3 knockdown impairs CenH3 deposition
To test whether NASP SIM3 is important to maintain CenH3 levels or its deposition, we aimed to analyze NASP SIM3 loss of function phenotypes. As T-DNA insertion lines are not available we turned to existing RNAi lines (Maksimov et al., 2016) and generated transgenic lines expressing an artificial miRNA construct directed against NASP SIM3 . We identified few lines with reduced NASP SIM3 mRNA levels, out of which we selected two lines (RNAi line 4 and amiRNA line 22) for further analysis (Figure 7a). These plants did not show any obvious phenotypic abnormalities during vegetative growth, but demonstrated slightly reduced seed setting and an increased number of aborted seeds ( Figure S5).
To investigate whether reduced NASP SIM3 expression would affect nuclear CenH3 levels and its deposition, we determined CenH3 levels relative to H4 in nuclear extracts of 10-day-old plantlets from two independent wild-type seed batches and the two transgenic lines with reduced NASP SIM3 expression. We reproducibly found a moderate but significant reduction of nuclear CenH3 levels in these two lines compared to the wild-type (WT) (Figure 7b,c). To investigate whether this reflected global CenH3 levels or might also impact nucleosomal CenH3 at centromeres, we sorted 4C nuclei isolated from 3-day-old seedlings of the RNAi line 4 and wild-type plants and carried out immunostaining with anti-CenH3 antibodies. Fluorescence image stacks of each genotype were acquired using structured illumination microscopy and the sum of the fluorescence intensities from the centromere (CenH3) signals was calculated. The level of CenH3 at the centromeres of the NASP SIM3 RNAi line was reduced~30% compared with the wild-type (Figure 7d,e).
Taken together, these results indicated that reduced NASP SIM3 expression negatively affects CenH3 levels. We assumed that reduced loading of CenH3 in plant lines with lower NASP SIM3 levels might result in generation of haploids in crosses with wild-type plants as it was described for plants with altered expression of CenH3 (Ravi and Chan, 2010). To test this hypothesis, the RNAi line 4 was crossed with wild-type. The ploidy level of nuclei from 105 F1 seeds was analyzed by flow cytometry, however, all nuclei were found to be diploid ( Figure S6).

DISCUSSION
The chromosomal location where kinetochores assemble during mitosis and meiosis is in most organisms not defined by DNA sequence but by a specific chromatin organization demarcated by incorporation of the histone variant CenH3. Controlled storage and transport of CenH3 histones, as well as its deposition only at specific genomic locations, are therefore critically important for the balanced segregation of the chromosomes to daughter cells (Lacoste et al., 2014;M€ uller and Almouzni, 2017). Despite its conserved function in eukaryotes, CenH3 molecules are highly divergent between species. In particular the N-terminus of CenH3, longer than that of H3.1 or H3.3, shows variability in length and sequence composition between different phylogenetic groups and even within a genus such as Drosophila (Rosin and Mellone, 2017). In Arabidopsis the N-terminus of CenH3 is required for CenH3 loading to centromeres of meiotic, but not of mitotic chromosomes Ravi et al., 2011). It was suggested that in the absence of the N-terminus, which may be involved in protein-protein interaction like its homolog in yeast (Chen et al., 2000), CenH3 cannot be recognized by meiosis-specific chaperones, and therefore cannot be loaded onto centromeres. At present, the existence of mitosis and meiosis-specific mechanisms of CenH3 loading to centromeres remains unanswered. In metazoans structurally different protein complexes are involved in CenH3 deposition such as HJURP in humans and CAL1 in Drosophila, but no CenH3 transport or deposition factor had so far been identified in plants.

NASP SIM3histone interaction
By searching for Arabidopsis CenH3 interactors in two independent experimental setups, we identified NASP SIM3 as an in vivo CenH3 binding partner and confirmed interaction between NASP SIM3 and CenH3 using two independent approaches such as Y2H assays and BiFC. NASP SIM3 is an evolutionary highly conserved protein that was likely to be already present in the first eukaryotic ancestor (Nabeel-Shah et al., 2014). NASP SIM3 comprises three canonical tetratricopeptide repeat (TPR) motifs and a putative TPR motif interrupted by a large acidic region in its N-terminus and a predicted coiled-coil domain in its C-terminus. Arabidopsis NASP SIM3 has previously been shown to bind with similar affinity to both H3.1 and H3.3 (Maksimov et al., 2016) likely through a conserved heptapeptide motif LA-IRG in the C-terminal region of histone H3 that is sufficient for the interaction with the TPR motifs of human and Arabidopsis NASP SIM3 (Bowman et al., 2016). This heptapeptide is highly conserved in different H3 variants including human CenH3 CENP-A , therefore suggesting that the interaction between Arabidopsis NASP SIM3 and CenH3 would take place via the CenH3 HFD domain. However, while the LA-IRG motif is conserved in Arabidopsis H3.1 and H3.3, this motif differs in Arabidopsis CenH3 (namely LA-LGG). This therefore suggests the existence of additional contacts between CenH3 and NASP SIM3 . Indeed, while interaction of NASP with the H3 C-terminus was lost in a human sNASP binding mutant, this did not completely abolish binding to full-length H3 (Bowman et al., 2016) and in our BiFC assays NASP SIM3 interacts both with the HFD of CenH3 and its Nterminus. A possible supplementary interaction motif could be the proline-rich GRANT motif recently defined in the fission yeast CenH3 N-terminus that is required for interaction with Sim3 (Tan et al., 2018). However, our analysis of the N-termini of different plant CenH3 proteins did not reveal a conserved GRANT motif in plants, suggesting yet another mode and domain of interaction between Arabidopsis CenH3 and NASP SIM3 . Such multivalent proteinprotein interactions between histones and chaperone proteins as observed for NASP SIM3 and CenH3 might allow handing over CenH3 to other transport or assembly factors or histone modifiers. Alternatively, it may reflect different binding modes of CenH3 to either monomeric CenH3 or CenH3-H4 dimers. Despite the fact that it was reported that the mammalian NASP forms also complexes with H1 linker histones through the acidic patch present in TPR2, which is conserved in Arabidopsis NASP SIM3 (Richardson et al., 2000;Cook et al., 2011;Wang et al., 2012), we did not find evidence that Arabidopsis NASP SIM3 binds H1 or H2A histones implying specificity for H3 histones.
CenH3 is predominantly expressed in cells that are actively dividing, an expression pattern that can be explained by specific binding sites for the E2F transcription factor in its promoter (Heckmann et al., 2011). We find that the NASP SIM3 promoter also contains an E2F binding site. Accordingly, NASP SIM3 shows elevated (a) NASP SIM3 transcript levels in seedling from two independent wild-type (WT) (Col 0) seed batches, a line expressing an RNAi construct (RNAi-4) or an artificial miRNA (amiRNA-22) directed against NASP SIM3 transcripts. Relative NASP SIM3 transcript levels normalized to MON1 are shown, one sample of the three biological replicates of Col0 seed batch 2 was set to 1. (b) CenH3 protein levels in nuclear extracts from seedlings revealed by western blotting using an anti-CenH3 antibody. CenH3 levels were normalized to H4 levels. (c) Quantification of CenH3 band intensities relative to H4 from at least three independent biological replicates. Relative CenH3 levels in seedling from the Col0 seed batch 2 were set to one in each independent experiment. Error bars correspond to SEM. Student's t-test, **P < 0.01; *P < 0.05; •P < 0.07. promoter activity and transcript levels in dividing cells such as shoot and root meristems, similar to mammalian NASP that parallels histone expression (Richardson et al., 2000). For most tissues, NASP SIM3 transcript levels were also reflected at the level of the NASP SIM3 -EGFP-GUS fusion protein except for young flower buds, which showed weak GUS staining (Figure 3h) and revealed low GFP fluorescence in petals and the developing male gametophyte ( Figure S2a-d). This observation is potentially echoing the weak correlation between transcript and protein levels as frequently observed (Jiang et al., 2007;Nakaminami et al., 2014). Nevertheless, the high expression in root meristems for example, points to a role in escorting CenH3, while not excluding a role for NASP SIM3 in chaperoning H3 variants during DNA replication when histone pools are particularly dynamic. Indeed, a recent study indicated that the inheritance of pre-existing CenH3 CENP-A nucleosomes at the replication fork requires dedicated machinery that differs from H3.1 and H3.3 maintenance and involves HJURP recruitment (Zasadzi nska et al., 2018).

NASP SIM3 splice variants
In most vertebrate species, functional diversity of NASP proteins is generated through different splice variants. In mammals, alternative splicing generates two different isoforms, tNASP, highly expressed in testis, and sNASP that is rather ubiquitously expressed (Richardson et al., 2000). In Arabidopsis, there is evidence for expression of a shorter NASP.2 SIM3.2 isoform with a truncated coiled-coil domain. While absent from most RNA-seq datasets, suggesting low abundance, we confirmed the presence of this alternative transcript in flower buds and showed that the corresponding protein as a GFP fusion was weakly expressed with similar nuclear localization than the longer isoform. BLAST search against NASP.1 SIM3.1 cDNA of Arabidopsis revealed two splicing variants in cotton Gossypium raimondii: XP 012491717.1, corresponding to the longer isoform, and XP 012491718.1, corresponding the shorter isoform, while in other plant species only the NASP.1 SIM3.1 variant could be identified. It is very likely that the NASP.2 SIM3.2 isoform is not represented in transcriptomes of most species due to its low abundance and limitations of the transcriptome assembly algorithms.
If a NASP SIM3 knock-out were available, it would be interesting to investigate whether the short isoform complements the mutant phenotypes or whether a functional coiled-coil domain is essential for NASP SIM3 function, e.g. by mediating critically important protein-protein interactions essential for H3 or CenH3 handling. The fact that we never recovered plant lines with the 35S::NASP.2 SIM3.2 -GFP construct that might function as a dominant negative may point in this direction.

Impact of reduced NASP SIM3 expression
In contrast with mammals or Drosophila, in which loss of histone chaperones such as CAF-1, HIRA, ASF-1 and NASP causes early embryonic lethality, Arabidopsis plants lacking CAF-1, HIRA and both ASF-1 orthologues are viable (Kaya et al., 2001;Zhu et al., 2011;Duc et al., 2015) indicating an important plasticity in histone handling in plants. Because we did not find lines with strongly reduced NASP SIM3 expression and several attempts to generate nasp mutant plants with CRISPR-Cas9 technology have been unsuccessful, NASP SIM3 is likely to play an essential role in histone H3 handling in plants. Whether CenH3 could be bound by other factors such as the FACT complex (Okada et al., 2006) remains to be determined.

NASP SIM3 interaction network
Previous studies have shown that NASP SIM3 interacts with the heat shock 70 kDa protein HSC70-1 and the WD-40 repeat-containing protein MSI1 (Maksimov et al., 2016), the latter being one of the three subunits of the CAF-1 complex as well as of various remodeling complexes (Kaya et al., 2001;Hennig et al., 2005). By screening an Y2H library we have identified here MSI3 as an additional NASP SIM3 binding partner. MSI proteins are conserved histone binding proteins that bind non-nucleosomal histones H4. Interaction with NASP SIM3 could therefore be direct or occur via histones. Our screen further revealed interaction with histones as expected and with one of the four NAP homologs NAP1;2, a H2A-H2B chaperone. While predominantly cytoplasmic, a small fraction of NAP1;2 was found as well in the nucleus and shown to bind chromatin (Liu et al., 2009). Finally, we detected interaction with TSK-associating protein 1 (TSA1) that is preferentially expressed in shoot apexes similar to NASP SIM3 (Suzuki et al., 2005) and localizes in ER body-like structures and at the nuclear envelope (Batzenschlager et al., 2013) where centromeres are preferentially localized (Fang and Spector, 2005). TSA1 interacts with the small GIP proteins, which also bind CenH3. Similar to NASP SIM3 , loss of GIP proteins results in reduced CenH3 loading (Batzenschlager et al., 2015), establishing an additional indirect link of NASP SIM3 in centromere function. While NASP SIM3 is detected predominantly in the nucleus in interphase, the interaction with two proteins (NAP1;2 and TSA1), which are reported to be predominantly cytoplasmic, could also indicate existence of a small cytoplasmic fraction of NASP SIM3 in plant cells.

Role of NASP SIM3 in CenH3 escort
Reduced NASP SIM3 expression affected CenH3 levels and deposition that, however, had no effect on chromosome segregation and did not lead to the generation of haploids in crosses with wild-type ( Figure S6) as it was described for the cenh3 mutant complemented by a GFP tagged version of CenH3, in which the N-terminal tail was replaced by the one from histone H3.3 (Ravi and Chan, 2010). This could be explained by the fact that only plants with weakly affected NASP SIM3 expression and in consequence onlỹ 30% reduced CenH3 levels were obtained. While we cannot exclude a more direct role for NASP SIM3 in CenH3 deposition or inheritance at the replication fork, most of our evidence points to a role in CenH3 escort like its yeast homolog (Dunleavy et al., 2007). Confocal or high-resolution imaging of independent NASP SIM3 ÀGFP, NASP SIM3 ÀEGFPÀGUS and NASP SIM3 ÀmCherry fusions showed a diffuse nuclear staining excluding nucleoli, coherent with previous observations (Maksimov et al., 2016) and in agreement with a role for NASP SIM3 in chaperoning not only CenH3 but also H3.1 and H3.3. FRAP analysis together with the loss of nuclear NASP SIM3 by detergent extraction further revealed that the majority of Arabidopsis NASP SIM3 is not tightly associated with chromatin. In addition, no enrichment of BiFC signals was observed at centromeres, which may have indicated NASP SIM3 -CenH3 interaction during CenH3 deposition, despite the correct loading of CenH3 in these actively dividing cells. It is therefore likely that NASP SIM3 is not directly involved in CenH3 deposition, but rather binds CenH3 in its soluble, not chromatin bound, state. This is in agreement with the fact that in mammalian cells NASP depletion mostly affects the soluble H3-H4 pool and the observations that Arabidopsis NASP SIM3 preferentially binds to monomeric H3.1 and H3.3 in vivo (Maksimov et al., 2016). NASP SIM3 may thereby ensure CenH3 supply by maintaining the nuclear CenH3 pool and escort CenH3 to the corresponding assembly factor as suggested for Sim3 (Dunleavy et al., 2007). Further identification of additional NASP SIM3 binding partners, may allow us to determine the factors or complexes to which NASP SIM3 hands over CenH3 histones for chromatin assembly.
To generate gCenH3ÀEYFP expressing plants, the CenH3 genomic locus was amplified using primer P1_CenH3 and P2_CenH3 as well as P3_CenH3 and P4_CenH3 (Table S1). Using primer TAG_ for and TAG_rev (Table S1), EYFP CDS was amplified from pBluntÀEYFPÀTAG. The three resulting amplicons were merged into one product in a subsequent PCR reaction using primer P1_universal_TT and P2_universal_TT (Table S1), inserted via SfiI in vector p35S-Nos-BM (dna-cloning-service.com), and confirmed by sequencing. Resulting expression cassette (CenH3 genomic locus expressing CenH3 internally in frame fused via linker sequences with EYFP) was subcloned via SfiI into pLH7000 (http:// www.dna-cloning.com).
To generate plants with reduced NASP SIM3 levels, artificial miR-NAs were designed with the web microRNA designer WMD3 tool (Ossowski Stephan, Fitz Joffrey, Schwab Rebecca, Riester Markus and Weigel Detlef, personal communication) and cloned into pRS300 (Schwab et al., 2006) with subsequent adaptation for Gateway cloning in the pMDC32 expression vector, harboring a dual 35S promoter.
Arabidopsis thaliana plants were transformed according to the flower dip method (Clough and Bent, 1998). T1 transformants were selected on Murashige and Skoog medium containing 50 mg L À1 kanamycin and 50 mg L À1 hygromycin, by kanamycin only for the transformants expressing the pKGWFS7.0 vector based constructs or hygromycin only for pMDC32-based constructs. The plants were propagated under short-or long-day conditions in a cultivation room at 8 h light/20°C:16 h dark/18°C and 16 h light/20°C:8 h dark/18°C, respectively.

Yeast-two-hybrid assays
The full-length cDNAs encoding Arabidopsis NASP.1 SIM3.1 and ASF1A were cloned into the pGBKT7 vector as baits and the sequences encoding full-length histones H3.1, H3.3, CenH3, H2A.W.6 and H1.3 were cloned into the pGADT7 vector as prey. Vectors pGBKT7 and pGADT7 were respectively transformed into Saccharomyces cerevisiae Y187 and AH109 strains based on the manufacturer's instructions of the MatchMaker III GAL4 two-hybrid system (Clontech; TaKaRa, https://www.takarabio.com). Screening and interaction studies between preys and baits were performed by mating compatible yeast strains following Clontech's instructions. Interactions between NASP SIM3 , ASF1A and different histones were determined by growing transformants on medium YNB without Leu and Trp (PM: permissive medium), without Leu, Trp and His (LSM: low stringency medium) and without Leu, Trp, His and Ade (HSM: high stringency medium). Interaction efficiencies were recorded using drop tests on PM, LSM or HSM medium, with serial dilutions of a given strain grown in medium and incubated at 30°C.
To identify yet unknown interaction partners of NASP.1 SIM3.1 , the Walker two-hybrid cDNA library CD4-10 was screened using the pGBKT7-NASP.1 SIM3.1 vector as bait. Positive clones were identified by growth on high stringency medium. Among about 3 9 10 5 transformants, 10 positive clones were identified and sequenced. Among the potential candidates, only two proteins (NAP1;2: AT2G19480 and TSA1: AT1G52410) were subsequently analyzed, after cloning full-length CDS in the pGADT7 vector, to produce NAP1;2 and TSA1 prey constructs.

Bimolecular fluorescence complementation
The binary BiFC plant transformation vectors pSPYNE-35SGW and pSPYCE-35SGW, containing the N-or C-terminus of YFP, respectively, were kindly provided by Klaus Harter (University of T€ ubingen, Germany). The entire coding regions of NASP.1 SIM3.1 and CenH3 or regions coding for the N-or C-terminal parts of CenH3 (Lermontova et al., 2006), respectively were subcloned from the corresponding pDONR221 vector into the BiFC vectors in frame with the split YFP. BiFC was performed in N. benthamiana plants (4 weeks after sowing) after Agrobacterium-mediated transient transformation of very young leaves containing mitotic cells according to Walter et al. (2004).

Proteomic analysis of Arabidopsis inflorescences expressing gCenH3ÀEYFP or EYFP only
Co-IP experiments were conducted on four independent inflorescent protein extracts each from gCenH3ÀEYFP and EYFP only expressing plants using GFP-Trap_MA Kit (Chromotek, https:// www.chromotek.com) according manufacturer's instructions with minor modifications. Here, 150-200 inflorescences per sample were ground to a fine powder in liquid nitrogen and resuspended in ice-cold lysis or RIPA buffer (supplemented with DNase I, MgCl 2 , Halt TM protease and phosphatase inhibitor cocktail; Thermo Fisher Scientific #78440). Extracts were mixed with ice-cold dilution buffer (supplemented with Halt TM Protease and phosphatase inhibitor cocktail) and pre-cleared for 45 0 at 4°C with equilibrated blocked magnetic-agarose beads (Chromotek). Co-IPs were performed by incubating pre-cleared extracts for 90 min to 2 h with equilibrated GFP-trap beads (Chromotek) at 4°C followed by washes to remove non-specific proteins (twice in ice-cold dilution buffer, once in ice-cold wash buffer '175 mM') (10 mM Tris/HCl pH 7.5, 0.5 mM EDTA, 175 mM NaCl) and '150 mM' (10 mM Tris/HCl pH 7.5, 0.5 mM EDTA, 150 mM NaCl) and six times in ice-cold wash buffer '100 mM' (10 mM Tris/HCl pH 7.5, 0.5 mM EDTA, 100 mM NaCl). Mass spectrometry (MS) analysis was performed as described in Data S1.

Microscopy
For time-lapse microscopy, seedlings were grown in coverslip chambers (Nalge Nunc International, https://www.thermofisher. com/nalgene) for 7-10 days and analyzed with a LSM 510 META confocal laser-scanning microscope (Carl Zeiss GmbH). EYFP was excited with a 488nm laser line and the specific fluorescence recorded with a 505-550 nm band-pass filter.
Bleaching experiments were performed with the same microscope. Nuclei were observed using a 63x/1.4 Oil Plan-Apochromat objective (4x zoom, image size 512 9 186-232 pixels). For imaging (pre-and post-bleaching) 1.1-1.8% of a 488nm line from a 100 W argon ion laser running at 50% power was used. The emission was registered with a 505-550 band-pass filter with a maximal opened detector pinhole. Three data points were acquired to measure the pre-bleaching intensity. A small 2 9 2 lm area was photobleached using 100% of a 488nm line with five to seven iterations. The bleaching scans lasted from 50 msec (reiterated bleaching), to 250-350 msec for single and repeated FRAP experiments. Imaging scans were performed with 350 msec intervals.
To analyze the ultrastructure of immunosignals and chromatin beyond the classical Abbe/Raleigh limit at a lateral resolution of 120 nm (super-resolution, achieved with a 488 nm laser) spatial structured illumination microscopy (3D-SIM) was applied using a 63x/1.4 Oil Plan-Apochromat objective of an Elyra PS.1 microscope system and the software ZENblack (Carl Zeiss GmbH, https://www.zeiss.com). Images were captured separately for each fluorochrome using the 561, 488 and 405 nm laser lines for excitation and appropriate emission filters (Weisshart et al., 2016). Maximum intensity projections of whole nuclei were calculated via the ZEN software. Enlarged image sections were presented as single slices to indicate the subnuclear chromatin and protein structures at the super-resolution level. Imaris 8.0 (Bitplane, http://www.bitplane.com) was applied to measure the amount of CenH3 in wild-type and NASP.1 SIM3.1 RNAi interphase nuclei via the sum of voxel intensities.

DATA AVAILABILITY STATEMENT
The authors responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the Instructions for Authors are: Aline Probst (aline.probst@uca.fr) and Inna Lermontova (lermonto@ipk-gatersleben.de).

SUPPORTING INFORMATION
Additional Supporting Information may be found in the online version of this article. Figure S1. NASP SIM3 protein sequence. Figure S2. Expression profile of CenH3 and NASP SIM3 . Figure S3. The NASP SIM3 gene codes for two different splice variants. Figure S4. NASP SIM3 is a soluble protein. Figure S5. Seed set is slightly reduced in plants with reduced NASP SIM3 expression. Figure S6. Flow cytometry analysis of nuclei isolate from F1 seeds generated by crossing of NASP SIM3 RNAi plants with wild-type. Video S1. NASP.1 SIM3.1 dynamics in Arabidopsis thaliana root tip meristem. Table S1. Primers used in this study.