RbAp46/48LIN-53 and HAT-1 are required for initial CENP-AHCP-3 deposition and de novo holocentromere formation on artificial chromosomes in Caenorhabditis elegans embryos

Abstract Foreign DNA microinjected into the Caenorhabditis elegans syncytial gonad forms episomal extra-chromosomal arrays, or artificial chromosomes (ACs), in embryos. Short, linear DNA fragments injected concatemerize into high molecular weight (HMW) DNA arrays that are visible as punctate DAPI-stained foci in oocytes, and they undergo chromatinization and centromerization in embryos. The inner centromere, inner kinetochore and spindle checkpoint components, including AIR-2, CENP-AHCP-3, Mis18BP1KNL-2 and BUB-1, respectively, assemble onto the nascent ACs during the first mitosis. The DNA replication efficiency of ACs improves over several cell cycles, which correlates with the improvement of kinetochore bi-orientation and proper segregation of ACs. Depletion of condensin II subunits, like CAPG-2 and SMC-4, but not the replicative helicase component, MCM-2, reduces de novo CENP-AHCP-3 level on nascent ACs. Furthermore, H3K9ac, H4K5ac and H4K12ac are highly enriched on newly chromatinized ACs. RbAp46/48LIN-53 and HAT-1, which affect the acetylation of histone H3 and H4, are essential for chromatinization, de novo centromere formation and segregation competency of nascent ACs. RbAp46/48LIN-53 or HAT-1 depletion causes the loss of both CENP-AHCP-3 and Mis18BP1KNL-2 initial deposition at de novo centromeres on ACs. This phenomenon is different from centromere maintenance on endogenous chromosomes, where Mis18BP1KNL-2 functions upstream of RbAp46/48LIN-53.


INTRODUCTION
Histone H3 variant, CENP-A, replaces histone H3 in some of the centromeric nucleosomes and serves as the foundation for building the kinetochore (1), which connects the sister chromatids to opposite spindles and orchestrates chromosome movement in cell division. CENP-Aspecific chaperone deposits CENP-A precisely to the centromeric regions of sister chromatids. Centromere propagation through cell cycles and generations is crucial for accurate chromosome segregation and maintenance of genome integrity.
Ectopic formation of a centromere can result in a dicentric chromosome, which may undergo chromosome breakage-fusion-bridge (BFB) cycle, leading to chromosomal rearrangements, chromosome losses or gains, aneuploidy, and potentially chromosome instability (CIN) and tumorigenesis. Multiple cases of neocentromeres were found in human patients with congenital abnormalities or developmental disorders (2). The mechanism of new centromere formation is still not fully understood, partly because of the technical challenges in tracing the early events of neocentromere formation in patients' cells. However, new centromere formation has been observed in diverse species. For example, tethering CENP-A-specific chaperones to a euchromatin locus or overexpressing CENP-A could cause ectopic CENP-A localization and ectopic centromere formation in fission yeast and human cells (3,4). Besides, transforming or transfecting centromeric DNA into yeast (3,5,6) or human cells (4,7,8) can form artificial chromosomes with de novo centromeres. However, most artificial chromosome (AC) formation in these species often relies on the presence of their endogenous centromeric DNA sequences, which suggests endogenous centromeric DNA sequences are preferred for new centromere formation. The isolation of these ACs require long-term selection, and they have relatively low frequencies of de novo centromere formation, which has limited the study of the early events of new centromere formation.
In Caenorhabditis elegans, injecting foreign DNA, even devoid of C. elegans sequences, into its syncytial gonad could form episomal extra-chromosomal arrays, which we also call artificial chromosomes, in the embryonic cells. Some of these ACs acquire the ability to propagate mitotically and are also inherited through subsequent generations (9,10). These heritable ACs have established a functional holocentromere rather than hitchhiking on the endogenous chromosomes (11). Dissecting the mechanism of de novo centromere establishment on ACs may extend our understanding of the neocentromere formation process on endogenous chromosomes.
In the present study, after injection of short, linear DNA, we investigated the timing of de novo CENP-A HCP-3 deposition on ACs, and demonstrated that CENP-A HCP-3 starts to assemble on ACs after fertilization. Another inner kinetochore protein, Mis18BP1 KNL-2 , and an inner centromere protein, AIR-2, are also recruited to the nascent ACs in the first mitosis. The ACs attempt to segregate in the first cell division, but with anaphase bridges. We also analyzed the histone post-translational modifications (PTMs) that are co-present with de novo CENP-A HCP-3 on nascent ACs in one-cell embryos. Based on the profiles of the enriched histone PTMs on nascent ACs, we depleted the relevant histone modifiers or the associated histone chaperones by RNA interference (RNAi), and analyzed the AC segregation rate by live-cell imaging, and the canonical histone and centromeric protein signals by immunofluorescence analysis. We demonstrated that HAT-1 and RbAp46/48  are required for the enriched H3K9ac, H4K5ac and H4K12ac on nascent ACs in one-cell embryos. Depleting HAT-1, RbAp46/48  or both will reduce ACs' segregation competency and reduce de novo CENP-A HCP-3 deposition on nascent ACs. We show that de novo CENP-A HCP-3 deposition on ACs also requires condensin II subunits, but it is independent of DNA replicative helicase component, MCM-2. These results demonstrate that the mechanism of de novo CENP-A HCP-3 deposition on ACs requires histone acetyltransferase HAT-1, and the CENP-A deposition machinery, including histone chaperone RbAp46/48  , together with M18BP1 KNL-2 and condensin II.

Worm strains and maintenance
Worm strains used in this study are listed in Supplementary  Table S1. All worms were maintained at 22 • C, unless otherwise mentioned, on MYOB plates [all components: 2.0 g NaCl, 0.55 g Tris-HCl, 0.24 g Tris-OH, 3.1 g Bacto peptone, 8 mg cholesterol, 20 g agar; mixed with 1 L water and then autoclaved] seeded with E. coli OP50. The CRISPR/Cas9 transgenic technique described by Dickinson and Goldstein (12) was used to design and generate a GFP-tagged HAT-1 at the endogenous locus. PCR genotyping was performed using primer set: Seq-Hat-1 (Supplementary  Table S2).

Double-stranded RNA (dsRNA) synthesis and RNA interference (RNAi)
PCR primers were designed to amplify a region of target genes from N2 C. elegans genomic DNA or cDNA. T3 promoter (AATTAACCCTCACTAAAGG) or T7 promoter (TAATACGACTCACTATAGG) was added to the 5 end of primers. Primers (Supplementary Table S2) were selected using NCBI-Primer-Blast and were subjected to BLAST search using the C. elegans genome to confirm the primer specificity. PCR was performed using TaKaRa Ex Taq® DNA Polymerase and the PCR products were purified by Qiagen PCR purification kit. Purified PCR products were subjected to in vitro transcription using Ambion T3 and T7 MEGAscript® Kit at 37 • C for 4-6 h. Reaction products were digested with TURBO DNase at 37 • C for 15 min and purified using Ambion MEGAclear TM Kit. Eluates were incubated at 68 • C for 10 min followed by 37 • C for 30 min for complementary RNA annealing. Annealed dsRNA was adjusted to 1 g/l in ddH 2 O for microinjection. For RNAi, L4 stage worms were injected with dsRNA (1 g/l) and recovered at 22 • C for 24 h before further analysis. For RNAi plus AC introduction, L4 stage worms were injected with dsRNA (1 g/l) and recovered at 22 • C for 18 h to reach the young adult stage. The p64xLacO plasmid DNA was linearized by AfaI and purified (known as L64xLacO). The RNAi-treated young adult worms were then injected with L64xLacO in the gonad and recovered at 22 • C for another 5 h before live-cell imaging or immunofluorescence staining.
cDNA synthesis was performed using Maxima H Minus cDNA synthesis kit (Thermo Fisher). About 5 l of cDNA synthesis mix was added to the worm lysate. The final mix contains 1× RT buffer, 0.5 mM each dNTP, 5 M random hexamer, heat labile dsDNAse 1 unit/l, RNAse inhibitor and 20 unit/l reverse transcriptase. The tube was briefly centrifuged, mixed and incubated at 25 • C for 10 min, followed by 55 • C for 30 min and finally 85 • C for 5 min. The cDNA was diluted to 100 l with RNase-free H 2 O and 2 l was used for each PCR reaction for a final volume of 20 l.
Quantitative PCR was performed using StepOnePlus Real-Time PCR System using the Applied Biosystems™ Fast SYBR™ Green Master Mix with the following parameters: 95 • C for 20 s and 40 cycles of 95 • C for 3 s, 60 • C for 30 s. All data were normalized to the act-1 gene. The primers used are listed in Supplementary Table S1.

Live-cell imaging and AC segregation assay
Episomal artificial chromosomes (ACs) were visualized by injecting DNA containing LacO tandem repeats into worm strain OD426 (Supplementary Table S1), as reported previously (11), except that here we used linear DNA (L64xLacO) for microinjection. Injected worms were recovered on OP50-seeded plates for 5-8 h after microinjection. About 3-4 worms were then dissected in 2 l M9 buffer to release the embryos. Embryos were mounted on a freshly prepared 2% agarose pad, and the slide edges were sealed with Vaseline. Live-cell images were taken with a Carl Zeiss LSM710 laser scanning confocal microscope with a 16 AC Plan-Neofluar 40× Oil objective lens and PMT detectors. Stacks with 17 × 1.32 m planes were scanned for each embryo in a 3× zoom, 1-min or 30-s time interval, with 3.15 s pixel dwell and 92 m pinhole. Laser power for 488 nm and 543 nm was set at 5.5% and 6.5%, respectively.
To determine the AC segregation rates, every dividing cell that contains at least one AC was counted as one sample. Each division was categorized as either containing at least a segregating AC or containing all non-segregating AC(s). Segregating ACs were defined as those that aligned with the metaphase plate and segregated with endogenous chromosomes during anaphase. Non-segregating ACs were defined as those that remained in the cytoplasm or nucleus and did not segregate in mitosis. The AC segregation rate was calculated as the number of dividing cells containing at least one segregating ACs over the total number of dividing cells containing ACs. Among the segregating ACs, some may lag and have anaphase bridges, and may not eventually equally divide.

Image signal quantification
Images were processed with Fiji 2.0.0. For immunofluorescence, 31 z-sections were acquired with a spacing of 0.4 m for each embryo. The region of interest (ROI) and the number of z-stacks for each target object were manually selected. A larger area enclosing the whole ROI within the embryo was drawn in each sample (ROI-L). Integrated density (Int-Den) equals to area times mean grey value. For each channel, the integrated density of the ROI and ROI-L from all selected z-stacks containing the target object were summed (ROI IntDen and ROI-L IntDen ). The area between ROI and ROI-L were used for calculating the mean grey value of background following the equation Bg mean = (ROI-L IntDen -ROI IntDen ) / (ROI-L area -ROI area ). The corrected integrated density of each targeted protein, histone modification or EdU was calculated by ROI Corr.IntDen = ROI IntDen -(ROI area x Bg mean ). ROI Corr.IntDen was then normalized with the corrected integrated density of DAPI. The ROI Corr.IntDen of ACs were normalized to the ROI Corr.IntDen of endogenous chromosomes within the same embryo for comparing the signal of targeted proteins, histone modifications or EdU on AC with that on endogenous chromosomes.

Formation of artificial chromosomes (ACs) through chromatinization and de novo centromerization of foreign DNA in C. elegans one-cell embryos
In our previous study, we observed that foreign circular, supercoiled plasmid DNA injected into C. elegans gonad forms ACs in embryonic cells after 4-8 h of microinjection (11). An increasing proportion of ACs acquire segregation competency after they go through several cell divisions (11). We have further tested AC formation by injecting different DNA forms, including linearized plasmid DNA, and linearized plasmid DNA mixed with sheared salmon or enzyme-digested yeast genomic DNA (16). Based on the foci size of GFP::LacI that binds to the injected LacO arrays, larger ACs are formed by concatemerization of linear DNA, as compared to those formed from circular supercoiled DNA (Supplementary Figure S1A and B). This might suggest that it is more efficient to fuse linear foreign DNA fragments than to fuse circular DNA. ACs generated by injecting linear DNA (L64xLacO; (2794 bp) acquires segregation ability faster than those generated by injecting circular (supercoiled) DNA (Supplementary Figure S1C), possibly due to the larger ACs it produces (Supplementary Figure S1D). By injecting a complex DNA mixture from sheared salmon sperm DNA without the LacO repeat sequence, we confirmed that de novo CENP-A HCP-3 deposition can also occur on such 'complex' AC (Supplementary Figure S1E and F). However, the AC segregation rates in repetitive ACs and complex ACs have no significant difference (16).
To follow the fate of foreign DNA injected into the syncytial gonad of C. elegans, the germline, oocytes and embryos were imaged 5 h after injection of linearized 64 copies of LacO arrays (L64xLacO). To identify the timing and location of high molecular weight (HMW) DNA array formation, DAPI (4 ,6-diamidino-2-phenylindole) staining was used to stain DNA, including the concatemerized foreign DNA. Yet, in the syncytial gonad where L64xLacO was injected, no DAPI staining can be observed, suggesting that the DNA has not been concatemerized and cannot be visualized. However, punctate DAPI foci were found in the cytoplasm of the diplotene and diakinesis oocytes, suggesting that the injected DNA has fused into high molecular weight (HMW), extra-chromosomal DNA arrays ( Figure 1A and Supplementary Figure S1G). Based on their smaller size, unpaired morphology and cytoplasmic location, the HMW DNA arrays can be easily distinguished from the six highly compacted endogenous bivalent chromosomes.
To determine the timing of canonical and centromeric nucleosome assembly on these DNA arrays, histone H2B and CENP-A HCP-3 , respectively, were traced by the direct detection of the mCherry::H2B signal or the immunofluorescence signal of CENP-A HCP-3 . These DNA arrays in oocytes lack histone H2B, CENP-A HCP-3 and Mis18BP1 KNL-2 in oocytes ( Figure 1B and Supplementary Figure S1G). However, in fertilized zygotes, these DNA arrays became artificial chromosomes (ACs) that contain detectable histone H2B (11) and CENP-A HCP-3 ( Figure 1C), indicating that chromatinization and de novo centromerization of foreign DNA have begun in one-cell embryos after fertilization. CENP-A HCP-3 signal was observed on ACs as early as in embryos undergoing meiosis I ( Figure 1C). Live-cell imaging showed a newly formed AC aligned at the metaphase plate and attempted to segregate during the first mitosis ( Figure 1D, Supplementary Figure S1H and Supplementary Video S1). The aligned AC was pulled toward opposite poles at anaphase, but formed an anaphase bridge ( Figure 1D and Supplementary Figure S2G). This result suggests that the kinetochores on newly formed ACs were sufficient to attach to the mitotic spindles to move the ACs, but error of attachment, like merotelic attachment, could cause chromosome bridge.

Impaired DNA replication causes centromere disorganization on metaphase ACs, but does not affect de novo CENP-A HCP-3 deposition
Proper sister chromatid segregation depends on the biorientation of kinetochores on sister chromatids, and the capture of microtubules emanating from opposite centrosomes. Un-bi-oriented kinetochore arrangement may facilitate merotelic attachments of microtubules and result in lagging chromosomes. To elucidate the cause of anaphase bridge formation of nascent ACs, we investigate their centromere protein level and orientation. Interestingly, we found that the initially formed centromeres, containing CENP-A HCP-3 and Mis18BP1 KNL-2 , inner centromeric protein, AIR-2, and condensin component SMC-4, were distributed throughout nascent ACs at metaphase ( Figure  2A and B; Supplementary Figure S2A). SMC-4 has been shown to dissociate from the properly segregated endogenous chromatin during telophase (17). In contrast, high levels of SMC-4 still appeared in the center of the AC chromatin bridges at telophase (Supplementary Figure S2B).
Replicative stress induced by hydroxyurea (HU) causes excessive chromatin bridge formation (18). DNA replication has been shown to be needed for chromatin decondensation in C. elegans embryos during anaphase (15), which is coincident with the disassociation of condensin component SMC-4. We found that impairing DNA replication by HU treatment also results in the persistent presence of SMC-4 on the bridging endogenous chromosomes at telophase (Supplementary Figure S2C), which resembles the phenomenon of nascent AC segregation with bridges. We speculated that incomplete DNA replication may cause nascent ACs lagging. We evaluated the DNA replication efficiency on ACs by their 5-ethynyl-2 -deoxyuridine (EdU) incorporation efficiency in different embryonic stages. In one-cell stage embryos, ACs showed only 32% of EdU incorporation when compared to that on endogenous chromosomes ( Figure 2C and D), which suggests that DNA replication is inefficient for nascent ACs in early-stage embryos. DNA replication is supposed to finish in S phase, so we anticipate that the DNA replicative helicase complex, MCM-2-7, to have dissociated from endogenous chromosomes before metaphase. However, live-cell imaging shows that MCM-4::mCherry has a prolonged association with ACs (Supplementary Figure S2D), suggesting that the process of DNA replication on nascent ACs was not yet complete even at metaphase. By injecting worms without GFP::LacI expression, lagging chromatin was still observed in the one-cell embryos produced (Supplementary Figure S2E), suggesting that GFP::LacI-tethering per se on ACs is not the source of the AC segregation problem. The sister chromatids and the kinetochores of nascent ACs have not been resolved to become bi-oriented, which may result in microtubule misattachments, lagging ACs and AC missegregation in earlystage embryos.
In multi-cell embryos, we found that the DNA replication efficiency of ACs has improved, since the EdU incorporation rate on ACs is comparable to that on endogenous chromosomes ( Figure 2C and D). Consistently, more ACs in multi-cell stage embryos (37%) possessed bi-orientated centromeres than in one-cell stage wild-type embryos (0%) MCM-2, a component of the MCM-2-7 replicative helicase, is essential for DNA replication (19), and it is also a histone chaperone for restoring histones to newly synthesized DNA (20). As the formation of the MCM-2-7 complex depends on each of its subunits, depleting MCM-2 prevents MCM-2-7 complex assembly and blocks the process of DNA replication (19). After mcm-2 RNAi treatment, the efficiency of the depletion is confirmed by embryonic lethality (data not shown) and reduced level of mcm-2 transcript by RT-qPCR (Supplementary Figure S5A). However, CENP-A HCP-3 level on endogenous chromosomes is similar in the untreated and mcm-2 RNAi-treated one-cell embryos (Supplementary Figure S2H and I), consistent with previous work that CENP-A HCP-3 deposition is independent on DNA replication (21). MCM-2 depletion did not prevent ACs from aligning at the metaphase plate and segregating with the endogenous chromosomes (Supplementary Figure S2J and K), although some are missegregated. Quantification of the normalized intensity of CENP-A HCP-3 on nascent ACs shows no significant difference between the untreated and mcm-2 RNAi-treated one-cell embryos (Figure 2E and G), suggesting that de novo CENP-A HCP-3 deposition is independent of MCM-2. Moreover, BUB-1, the outer kinetochore and spindle checkpoint component (22), was recruited to nascent ACs in the absence of DNA replication (Supplementary Figure S2L), which suggests that the assembly of kinetochore is independent of DNA replication.

Condensin II facilitates de novo CENP-A HCP-3 deposition on nascent ACs in one-cell embryos
In C. elegans, condensin II complex co-localizes with centromere proteins on metaphase chromosomes (17,23), and is proposed to have a specific function at the centromere in addition to chromatin condensation. However, depletion of condensin I and II component SMC-4, condensin I subunit CAPG-1 or condensin II subunit CAPG-2 has undetectable effect on CENP-A HCP-3 deposition for endogenous chromosomes (Supplementary Figure S3), despite that SMC-4 and CAPG-2 depletion caused lagging chromosomes and chromosome missegregation (17,23) ( Supplementary Fig-ure S3A). In human cells and Xenopus egg extracts, condensin II is required for new CENP-A deposition in mitotic cells and new CENP-A loading in the first mitosis, respectively (24,25). As the SMC-4 signal is positive on nascent ACs, we further determined if condensin contributes to de novo centromere formation on ACs in C. elegans embryos. We depleted CAPG-1, CAPG-2 and SMC-4, respectively. Depletion of either SMC-4 or CAPG-2 reduces the level of de novo CENP-A HCP-3 on nascent ACs (Figure 3). In contrast, depletion of CAPG-1 does not affect de novo CENP-A deposition on ACs ( Figure 3B and D). These results suggest that condensin II may facilitate de novo CENP-A HCP-3 deposition on nascent ACs in C. elegans embryos.

The spectrum of histone post-translational modifications (PTMs) on nascent ACs
We have previously shown that H4, H3K9 acetylation and RNAPII docking facilitate de novo centromere formation in ACs formed by injecting circular, supercoiled p64xLacO plasmid (26). To further identify essential factors in de novo centromere formation, we profiled the histone PTMs on the nascent ACs generated from linear L64xLacO. We hypothesize that histone codes that co-exist with chromatinization and centromerization of newly formed ACs in one-cell embryos may lead to the identification of the required factors. We chose several histone PTMs that have been reported to be associated with centromere function. The spectrum of histone PTMs on newly formed ACs in one-cell embryos by immunofluorescence analysis ( Figure  4A and Supplementary Figure S4) is summarized in Table  1. For PTMs that are associated with transcription activity, we analyzed H3K4me1 (27,28), H3K4me2 (28,29) and H3K4me3 (29,30) on newly formed ACs (Supplementary Figure S4). A medium level of H3K4me1 was present on ACs, even though H3K4me2 and H3K4me3 were lacking, suggesting that ACs might be actively transcribing. Consistently, the signal intensities of H4K5ac (31,32) ( Figure  4B), H4K12ac (31,32) ( Figure 4C), H3K9ac (33) ( Figure  4D) and H4K20me1 (34,35) ( Figure 4E) on nascent ACs were significantly higher than that on endogenous chromosomes. Meanwhile, the DNA replication-associated histone PTM, H3K56ac (20,36), on nascent ACs has dimmer signal intensity as compared to that on endogenous chromosomes (Supplementary Figure S4), consistent with    Table 1. For quantification of PTMs on ACs and endogenous chromosomes, the signal density of each PTM was normalized with that of DAPI. The number of samples (n) analyzed is indicated. The error bars represent SD. Significant differences are analyzed by the Student's t-test (*, P < 0.05; **, P < 0.01).  (27), were undetectable on nascent ACs in onecell embryos (Supplementary Figure S4), consistent with our previous finding that heterochromatin is dispensable for de novo centromere formation (11).

RbAp46/48 LIN-53 is essential for chromatinization of nascent ACs
On endogenous chromosomes, RbAp46/48 LIN-53 also affects H2B level, though not as severe as the effect on CENP-A HCP-3 level (39). As RbAp46/48 LIN-53 is also known to be a H3-H4 chaperone (40)(41)(42), and, we further tested if RbAp46/48 LIN-53 specifically deposits acetylated histone or affects general histone deposition on nascent ACs. We found that RbAp46/48 LIN-53 is essential for total histone H3 and H2B deposition on nascent ACs, but surprisingly not histone H4 ( Figure 5A-C). The immunofluorescence experiments for H4 were repeated with two different antihistone H4 antibodies (ab10158 and ab177840) ( Figure 5B and C). We speculated that in the absence of RbAp46/48 LIN-53 , H3 might be replaced by another major H3 variant, H3.3, which depends on HIRA-1 for deposition in C. elegans (43). To test if depletion of RbAp46/48 LIN-53 affects H3.3 level of ACs, we injected L64xLacO into a strain expressing GFP::HIS-72 (H3.3). Surprisingly, we found that RbAp46/48 LIN-53 depletion also reduced the level of GFP::HIS-72 on nascent ACs to 30% ( Figure 5D), as compared with 11% of H3. It is possible that in the absence of RbAp46/48 LIN-53 , remaining H3 or H3.3, and other H3 variants could form tetramers with H4 on the injected foreign DNA, and does not affect the total level of H4 on nascent ACs. These results indicate that RbAp46/48 LIN-53 is essential for chromatinization of foreign DNA.

The AC segregation and the enrichment of H4K5ac, H4K12ac and H3K9ac on ACs depend on RbAp46/48 LIN-53 and HAT-1
To investigate whether the corresponding histone modifying enzymes of the enriched AC PTMs facilitate de novo centromere formation, we performed RNA interference (RNAi) by injecting dsRNA of candidate histone modifier genes to L4 stage worms expressing GFP::LacI and mCherry::H2B ( Figure 6A). The RNAi efficiency of each gene was confirmed by RT-qPCR ( Supplementary Figure S5A) or live-cell imaging (Supplementary Figure S5B). Eighteen hours after dsRNA injection, L64xLacO was injected to RNAi-treated worms or untreated worms of the same stage. Embryos were dissected from injected worms and mounted for live-cell imaging to measure the AC segregation rate, and for immunofluorescence analysis to compare the CENP-A HCP-3 signal on ACs.
4 knockdown embryos have no significant difference with WT embryos ( Figure 6B). Depletion of HDA-1 or SET-1 also did not affect AC segregation in one-cell embryos ( Figure 6B).
Interestingly, depletion of RbAp46/48 LIN-53 , the CENP-A HCP-3 chaperone (39) and the HAT-1 physical interactor (Supplementary Figure S5D), completely abolished the segregation competency of nascent ACs (Figure 6B and C). Live-cell imaging videos (Supplementary Videos S3 and S4) show examples of an untreated and a lin-53 RNAi-treated embryo, both containing ACs, and they went through the first three consecutive cell divisions from one-cell stage to four-cell stage. A nascent AC in a WT embryo aligned at the metaphase plate with endogenous chromosomes, attempted to segregate, but formed chromosome bridges at the first anaphase in one-to-two cell stage, and has less severe chromosome lagging during the three-to-four cell transition. In contrast, all nascent ACs were passively remained in one of the two daughter cells during each division in lin-53 RNAi-treated embryo. Either RbAp46/48  or HAT-1 depletion, or double depletion significantly reduces the level of H4K5ac, H4K12ac and H3K9ac on ACs ( Figure 6D-F and Supplementary Figure S5E-G). The results suggest that RbAp46/48 LIN-53 and HAT-1 may facilitate histone H4 acetylation at these sites or preferentially deposit acetylated histones. On the other hand, as RbAp46/48 LIN-53 affects total H3 level on ACs in addition to H3K9ac, it is difficult to conclude which step functions in.

RbAp46/48 LIN-53 is required for de novo Mis18BP1 KNL-2 localization on nascent ACs
Mis18BP1 KNL-2 and CENP-A HCP-3 are interdependent for each other's localization on endogenous chromosomes of C. elegans (45). To test if Mis18BP1 KNL-2 is also necessary for de novo CENP-A HCP-3 deposition on nascent ACs, we depleted Mis18BP1 KNL-2 and performed immunofluorescence analysis. knl-2 RNAi almost completely abolished CENP-A HCP-3 signal on nascent ACs ( Figure 7A and E). This indicates that Mis18BP1 KNL-2 is also essential for CENP-A HCP-3 localization on both nascent ACs and endogenous centromeres. Similar to endogenous centromeres, Mis18BP1 KNL-2 localization on nascent ACs also relies on CENP-A HCP-3 ( Figure 8A and D). We simultaneously stained CENP-A HCP-3 and Mis18BP1 KNL-2 on nascent ACs in one-cell embryos, and show that 62% of ACs have both CENP-A HCP-3 and Mis18BP1 KNL-2 , while 38% of ACs have neither of the signals (Supplementary Figure S6A and B). We have not found any ACs that have only CENP-A HCP-3 or only Mis18BP1 KNL-2 , which is consistent with the codependence of Mis18BP1 KNL-2 and CENP-A HCP-3 localization ( Figures 7A, E, 8A and D).
On endogenous chromosomes, RbAp46/48 lin-53 depletion only reduced CENP-A HCP-3 level but did not affect Mis18BP1 KNL-2 level (39). We monitored the M18BP1 KNL-2 signal on nascent ACs in lin-53 RNAi-treated embryos. Surprisingly, at de novo centromeres on nascent ACs, RbAp46/48 LIN-53 depletion also leads to the loss of initial Mis18BP1 KNL-2 deposition ( Figure 8A and B), which suggests that while Mis18BP1 KNL-2 could be a self-directing factor for centromere maintenance in the existing centromeres, it is downstream of RbAp46/48 LIN-53 in de novo centromere establishment ( Figure 9A).
Interestingly, we found that hat-1 RNAi treatment reduces M18BP1 KNL-2 level on nascent ACs to 30% ( Figure  8A and C). In human cells, MYST2 has been described as an interactor of Mis18 complex for regulating CENP-A deposition (46). However, the function of MYS family of acetyltransferases on centromere chromatin has not been reported previously in C. elegans. Since we found that MYS-1 and MYS-2 depletion have additive effect with hat-1 RNAi in abolishing AC segregation, we tested whether double knockdown of mys-1 and mys-2 prevents Mis18BP1 KNL-2 localization on nascent ACs. However, the Mis18BP1 KNL-2 signal on nascent ACs in mys-1 and mys-2 double depleted embryos shows no significant difference as compared with untreated embryos (Figure 8A and D).

DISCUSSION
Chromatinized ACs were formed in fertilized embryos a few hours after microinjection of foreign circular DNA (11) and linear DNA. Histone H2B and centromeric protein CENP-A HCP-3 signals on ACs were detectable in fertilized oocytes and one-cell embryos ( Figure 1B and C). This is consistent with the finding that major sperm proteins trigger nuclear membrane breakdown (47) and release the nuclearlocalized histones and centromeric proteins to allow chromatinization and centromerization of the HMW DNA arrays, which are initially located in the cytoplasm. The whole process of AC formation is summarized in Figure 9B.
Interestingly, in wild-type, the average level of CENP-A HCP-3 (normalized to the amount of DNA based on DAPI staining) on nascent ACs formed from linear plasmid DNA is 1.3-fold higher than that on endogenous chromosomes in one-cell embryos (Supplementary Figure S1I). This observation is slightly different from previous study in which the level of CENP-A HCP-3 on ACs in one-cell stage is lower than that on endogenous chromosomes (26). However, the CENP-A HCP-3 signal on ACs formed from circular DNA increases quickly in the first few cell cycles to become comparable with that in endogenous chromosomes in 17-32 cell stage (26). Such difference in initial of CENP-A HCP-3 level could be due to the difference in size or structure of the ACs from different injected DNA forms (Supplementary Figure  S1B-D).    is required for M18BP1 KNL-2 localization on HMW DNA; Condensin II complex also facilitates CENP-A HCP-3 deposition. Chromatinization and centromerization of the HMW DNA generates nascent ACs. Nascent ACs have DNA replication defects and lack bi-oriented sister kinetochores, which could lead to merotelic attachments to the mitotic spindle and chromosome bridging (in early embryonic cells). 3. Finally, DNA replication efficiency gradually improves on ACs, and ACs 'mature' in late embryonic stage. In matured ACs, bi-oriented sister kinetochores allow amphitelic attachment of spindles and proper segregation.
We found that SMC-4 is present on chromosome bridges upon HU treatment, which is possibly due to the decondensation defects caused by replication stress (15), although the presence of SMC-4 could also be due to the enrichment of condensin I complex on chromatin bridges to prevent cleavage-furrow regression (48). We examined the DNA replication status on nascent ACs in one-cell embryos and found that it is less efficient than on endogenous chromosomes. In C. elegans, DNA replication origins contain H3K4me2 enrichment (49), whereas H3K4me2 is absent on nascent ACs. The underlying non-C. elegans sequences or chromatin environment of ACs could be the reason for the less efficient DNA replication on nascent ACs, but surprisingly DNA replication efficiency improves quickly in few cell cycles together with higher frequency of bi-oriented kinetochores ( Figure 2F) and even segregation (Supplementary Figure S2F and G). Expectedly, mcm-2 depletion perturbed kinetochore bi-orientation on both en-dogenous chromosomes and ACs ( Figure 2E and F). During DNA replication, nucleosomes are disassembled from the parental DNA strand ahead of the DNA replication machinery (50). MCM-2 has been recently proposed to be able to chaperone H3-H4 and CENP-A-H4 dimers, for replenishing them to the sister chromatins behind the replication forks (20). However, mcm-2 RNAi did not prevent the process of de novo CENP-A HCP-3 deposition on nascent ACs ( Figure 2E and G), and did not affect their ability to recruit outer kinetochore and spindle checkpoint component BUB-1 (Supplementary Figure S2J). Although CENP-A HCP-3 was found to be almost completely turned over between two cell cycles (21), it is not clear when the new CENP-A HCP-3 was loaded on the holocentromere during the cell cycle in C. elegans embryos. We postulate that de novo centromere formation on foreign DNA per se is independent of DNA replication. Indeed, the CENP-A HCP-3 levels on nascent ACs ( Figure 2E and G) and endogenous chromosomes (Supplementary Figure S2H), when normalized to the DNA amount (DAPI) or H2B, are similar in wild-type and mcm-2 RNAi-treated embryos, which supports this hypothesis. Interestingly, this is similar to CENP-A replenishment on chromatin in human cells, which occurs before S phase and is independent of DNA replication (51). Yet, it is different from the case in budding yeast, in which CENP-A Cse4 also completely turns over and reloads to sister chromatids in S phase, dependent on DNA replication (52,53).
Condensin II, but not condensin I, is specifically enriched at the centromeres and has been found to promote CENP-A deposition in human cells and Xenopus oocyte extracts (24)(25)54). In human cells, the interaction between condensin II and HJURP is needed for HJURP's centromeric localization, and for depositing new CENP-A (24). On the other hand, depleting condensin II in Xenopus oocyte extracts reduces the CENP-A level at the centromere, but not the HJURP level (25). In C. elegans, condensin II subunits also co-localize with CENP-A HCP-3 starting at prometaphase in embryonic cells (23). Despite that condensin II depletion does not affect loading of CENP-A HCP-3 on endogenous chromosomes (Supplementary Figure S3), we found that depleting condensin I/II component SMC-4 and condensin II-specific component CAPG-2 reduces de novo CENP-A HCP-3 deposition on newly formed ACs ( Figure  3). These findings indicate that condensin II facilitates de novo centromere formation. However, we could not detect physical interaction between condensin II subunit SMC-4 and CENP-A HCP-3 chaperone RbAp46/48 LIN-53 by coimmunoprecipitation (Supplementary Figure S5D). Condensin II might modify or maintain the ideal chromatin structure for de novo CENP-A HCP-3 deposition.
In human and chicken DT-40 cells, H4K20me1 was reported as a histone PTM on CENP-A nucleosomes enriched at centromeres, which is essential for CENP-T localization (34). In C. elegans, H4K20 is monomethylated by methyltransferase SET-1 (44). Though H4K20me1 is enriched on nascent ACs, the AC segregation rate in set-1 RNAi-treated embryos shows no significant difference from that in untreated embryos ( Figure 6B), indicating that H4K20me1 is not essential for de novo CENP-A HCP-3 deposition.
We found that histone acetylation on H3K9, H4K5 and H4K12 was significantly enriched on nascent ACs formed from linear DNA, which is consistent with the previously documented acetylated H3K9 and panH4ac (on K5, 8, 12 or 16) on ACs from circular injected DNA (26). H3K9ac has been reported to be associated with the deposition of newly synthesized histones H3 in Tetrahymena, while the involved acetyltransferase is not clear (55). In human cells, H3K9ac also induces de novo CENP-A deposition on alphoid DNA at ectopic site and is compatible with centromere functioning (33). Our result shows that the level of H3K9ac ( Figure 6F) and H3 ( Figure 5A) were both significantly decreased in lin-53 RNAi-treated embryos, suggesting that RbAp46/48 LIN-53 is involved in both chromatinization and H3K9 acetylation. However, we cannot distinguish if RbAp46/48 LIN-53 specifically deposits H3K9ac to the foreign DNA or its depletion inhibits the overall deposition of H3 including H3K9ac on foreign DNA. Notably, in species as divergent as humans, Drosophila, Tetrahymena and yeast, newly synthesized and newly deposited H4 are acetylated in a conserved pattern at lysines 5 and 12 before its association with DNA (55-57). Pre-nucleosomal H4 is di-acetylated at K5 and K12 by HAT-1 in human cells, DT-40 cells and yeast, through forming a complex with H3-H4 chaperone, RbAp46/48 (40,(58)(59). We observed a significant reduction of H4K5ac and H4K12ac ( Figure  6D and E) on nascent ACs after lin-53 RNAi treatment, while the total H4 level remains unchanged ( Figure 5B and C). RbAp46/48 LIN-53 may preferentially deposit acetylated H4K5,12ac nucleosomes on the foreign DNA.
In lin-53 RNAi-treated embryos, the unchanged level of H4 did not match with the reduced levels of H3 and H3.3 on nascent ACs. H3.3 could be deposited through other histone chaperones, like HIRA ( Figure 9B) (43) and did not completely rely on RbAp46/48   (62). The unmatched level of H4 and H3 variant is observed possibly because we only detected one of the H3.3 isoforms (HIS-72). Other H3.3 isoforms and other H3 variants, might also be able to form tetramer nucleosomes with H4 on foreign DNA. Besides, non-nucleosomal H4, possibly without an H3 partner, might bind to the partially chromatinized DNA, potentially causing more chromosomal instability.
In chicken DT-40 cells, RbAp48 is essential for new CENP-A deposition to the centromere, which cooperates with HAT-1 to acetylate pre-nucleosomal CENP-A-H4 complex at H4K5 and K12 (31). In Drosophila, an in vitro experiment shows that RbAp48 alone is sufficient to assemble CENP-A CID -H4 to the naked DNA (63). New evidences show that HAT-1 interacts directly with CENP-A CID in Drosophila and depletion of HAT-1 reduces the efficiency of new CENP-A CID deposition significantly (64). Nevertheless, in fission yeast, RbAp46/48 Mis16 and HJURP Scm3 are present in the same complex, where RbAp46/48 Mis16 distinguishes CENP-A Cnp1 -H4 from H3-H4 by recognizing HJURP Scm3 and H4 independently (65). The temporal (RbAp46/48 Mis16 -CENP-A Cnp1 -H4-HJURP Scm3 ) complex binds to the centromere-specific Mis18-Eic1 complex through Mis18-HJURP Scm3 interaction, then RbAp46/48 Mis16 switches binding from H4 to Eic1 for CENP-A Cnp1 deposition (66). In HeLa cells, RNAi knockdown of RbAp46/48 reduces ectopic loading of CENP-A S68E mutant on the chromosome arms, which cannot bind to HJURP, suggesting that RbAp46/48 may deposit CENP-A S68E -H4 to ectopic loci in the absence of HJURP (67). On the other hand, in oocyte extracts of Xenopus, RbAp48 depletion does not affect normal CENP-A incorporation to the centromere on the sperm chromatin but causes ectopic CENP-A deposition (25,31). We observed a significant reduction of the AC segregation rate in hat-1 RNAi-, lin-53 RNAi-and lin-53 hat-1 double RNAitreated embryos, indicating that RbAp46/48 LIN-53 and HAT-1 are both involved in de novo centromere formation ( Figure 6B). Indeed, we found that both RbAp46/48  and HAT-1 are critical for de novo CENP-A HCP-3 deposition on C. elegans ACs ( Figure 7A-D).
Similarly, depletion of HAT-1 reduces M18BP1 KNL-2 levels on nascent ACs ( Figure 8A and C), which could be due to the decreased level of CENP-A HCP-3 , or nucleosome, on them ( Figure 7A and B). The centromeric protein localization dependency in C. elegans endogenous chromosomes and in de novo centromere formation on ACs has been summarized and compared ( Figure 9A). It will be of great interest to know if RbAp46/48 LIN-53 and M18BP1 KNL-2 form a complex to deposit pre-nucleosomal CENP-A HCP-3 -H4, in which RbAp46/48 LIN-53 is released from the centromere after the CENP-A HCP-3 deposition, while M18BP1 KNL-2 is retained on the chromatin for CENP-A HCP-3 stabilization and for recruiting outer kinetochore proteins.
The process of de novo centromere formation has been summarized in Figure 9B. Here, we proposed a hypothesis to explain the robust de novo centromere formation on foreign DNA in C. elegans, which may be associated with a single amino acid substitution of CENP-A during the evolution, and the promiscuous CENP-A HCP-3 deposition by RbAp46/48  . Phosphorylation of human CENP-A at Ser68 has been proposed to be important for preventing premature HJURP binding at metaphase in HeLa cells (67), but the function of CENP-A at Ser68 has been controversial (73). However, mutating the Ser68 residue to alanine, as a phospho-dead mutant, causes continuous CENP-A binding to HJURP and ectopic CENP-A deposition (67). Interestingly, this site is evolutionarily conserved in most eukaryotes except in budding yeast and C. elegans, where these CENP-A homologues have an alanine at this position, similar to the human phospho-dead mutant. In budding yeast and C. elegans, the CENP-A loading time may not be dependent on phosphorylation at this site at all. In budding yeast, CENP-A Cse4 propagation relies on HJURP Scm3 (52), and as a result, RbAp46/48 Hat2 can be a non-essential protein in this species (74). On the other hand, C. elegans may have lost HJURP, but CENP-A HCP-3 deposition depends on RbAp46/48 LIN-53 instead. For endogenous chromosome CENP-A HCP-3 propagation, the cue for new CENP-A HCP-3 deposition depends on pre-existing M18BP1 KNL-2 at the centromere. For nascent ACs, RbAp46/48 LIN-53 probably does not rely on pre-existing CENP-A HCP-3 to deposit CENP-A HCP-3 on the foreign DNA, facilitating de novo centromere formation.