CpA-TAP has reduced affinity for CENP-C

CpA-TAP consists of a C-terminal 18 kD modified Tandem Affinity Purification (TAP) tag that is made up of S-protein (one inactive component of ribonuclease S [38]-, a Tobacco Etch Virus (TEV) cleavage site (recognition peptide: E-N-LY-F-Q—S/G/A/M/C/H [39], and a minimal Staphylococcus aureus Protein A fragment with the calmodulin-binding peptide (10) (Fig. 1a). Stably expressed CpA-TAP chromatin was purified from human cells, and enriched with kinetochore components including CENP-B, CENP-H, CENP-N, CENP-T, and CENP-U when compared to H3.1-TAP chromatin [40]. These data support the interpretation that CpA-TAP successfully serves as a powerful biochemical tool to purify CENP-A associated complexes, which are otherwise present at low levels.

Fig. 1

CpA-TAP has poor affinity for CENP-C and altered post-translational modifications (PTM) signature. a) Native CpA protein and fusion CpA-TAP protein consisting of CpA + S protein + TEV cleavage site + Protein A. b) CoIF of HeLa cells transiently expressing CpA-TAP with native CENP-C. Scale bar = 5 μm. c) Immuno-precipitation of native CpA versus CpA-TAP (S-tag IP) (see Fig. S1a) followed by Western detection for CENP-C, and quantification of the ratio of CENP-C enrichment normalized against CpA ChIP. Error bar represent SEM. HC = heavy chain. d) Long TAU (L-TAU) Western comparing control HeLa cells and cells with transiently transfected CpA-TAP (merged panel below). e) AFM heights for IP’ed native CpA versus CpA-TAP nucleosomes. AFM measurements were done in air mode. Bulk = extracted input chromatin and α-S ChIP = immuno-precipitated CpA-TAP chromatin. Scale bar = 50 nm

CENP-C directly interacts with CENP-A at the C-terminus [41, 42] and is required to bridge the connection between the centromere and kinetochore [43]. Interestingly, the TAP-tag is located at the C-terminal end of CENP-A. We, and others, have demonstrated the importance of the unstructured C-terminus of CENP-A in correctly recruiting, binding, and rigidifying CENP-A upon CENP-C binding [12, 15, 42, 44]. Indeed, swapping the C-terminus of CENP-A with that of histone H3 results not just in loss of CENP-C binding, but also in abrogation of CENP-A function [45]. Thus, we were curious whether the C-terminal tag impacts CENP-C binding, relative to native CENP-A.

First, we performed IF on transfected cells on control (untransfected) or transiently transfected + CpA-TAP HeLa cells, using anti-CENP-A or anti-S-protein (S-tag) antibodies, respectively. IF revealed poor colocalization between CpA-TAP with endogenous CENP-C (Fig. 1b).

Second, Chromatin Immuno-Precipitation (ChIP) followed by Western blots confirmed native CpA IP from control cells were enriched for CENP-C, but CENP-C was not enriched in the CpA-TAP (ChIP against S-tag protein) fraction (Fig. 1c and Fig. S1a). Similarly to Bailey et al. [46], we also observed that when CpA-TAP is co-expressed in HeLa cells, native CpA levels were reduced (Fig. S1a, b) and that CpA-TAP levels are four-fold reduced compared to native CpA (Fig. S1c). Native CpA was not co-purified with the S-tag IP, suggesting CpA-TAP does not completely occupy the same native CpA domains.

CpA-TAP PTM signature is different from that of native CpA

Elucidating native CENP-A PTMs has been a challenging task, particularly due to its low abundance and difficulty in achieving complete peptide coverage during mass spectrometric.

(MS/MS) analyses. Previously, C-terminally FLAG-tagged CENP-A was immuno-precipitated and found to be ubiquitylated on lysine residue 124, which is important for centromeric deposition [24, 47]. More recent PTM analysis revealed a series of modifications that reside within the N-terminus of purified localization and affinity purification (LAP) tagged CENP-A, but modifications on CENP-A lysine residue 124 were absent [48]. How two CENP-A proteins with different tags can yield different PTM results remains unclear.

The use of alternative methods to successfully resolve different modified species of histones followed by mass spectrometry confirmation, have employed Triton Acid Urea (TAU) electrophoretic gel chemistry to successfully resolve proteins based on charge, hydrophobicity, and size [32, 49, 50] For example, we previously reported that native CENP-A from HeLa cells extracted with hydroxylapatite and high salt, followed by separation on a Long Triton Acid Urea (L-TAU) gel and analyzed by MS/MS, were acetylated on K124 [11, 12]. Other modifications at lower confidence levels were also detected along the N-terminus and throughout the histone fold domain (Bui, Nuccio, Nita-Lazar and Dalal, unpub). TAU gel electrophoresis remains a valuable qualitative tool to distinguish differing PTM signatures among two similar histones -in this case, native CpA versus CpA-TAP. The more CENP-A modified species exist, the greater number of bands or smears are expected on a L-TAU gel, as phosphorylated or acetylated residues cause protein bands to shift upwards [51]. Therefore, we used this method to compare CpA-TAP and native CpA purified from HeLa cells on L-TAU gels. Our results indicate that control HeLa cells exhibit at least four post-translationally modified forms of native CpA. To our surprise, only one distinct CpA-TAP species dominates and partially represses native CpA levels in + CpA-TAP transfected cells (Fig. 1d and Fig. S1d). These data suggest despite both proteins being CENP-A, the tagged version does not share the same PTM signature as native CpA.

CpA-TAP and native CpA nucleosomes are indistinguishable in height

It has been previously reported that CENP-A nucleosomes undergo height transitions during replication [11]. We were curious whether adding a tag would alter nucleosomal heights in unsynchronized cells. Both native CENP-A and anti-S (for CpA-TAP) ChIP followed by Atomic Force Microscopy (AFM) measurements revealed that both types of nucleosomes were indistinguishable in height (Fig. 1e; Table 1), suggesting nucleosomal heights are dictated by the internal histone fold domain.

Table 1 AFM measurements of various nucleosomal structuresCpA-TAP deposition does not coincide with native CpA sites in the genome

In previous works, we, and others, have reported that CENP-A in certain cancer cells accumulates at ectopic or non-centromeric sites in the genome [31], and that this non-native pathway exploits H3.3 chaperones [52]. We observed that CpA-TAP is stably bound to chromatin but appears depleted for CENP-C (Fig. 1c). Therefore, we wanted to explore where CpA-TAP deposits in the genome. To achieve this, we performed either native CpA (in untransfected HeLa background) or S-tag (for CpA-TAP enrichment) ChIP, followed by deep sequencing (ChIPseq).

A total of 792 common hotspots between native CpA and CpA-TAP were identified, making up 18% of total native CpA and 66% of CpA-TAP (Fig. 2a). When the data was separated into centromeric versus non-centromeric identities, an interesting pattern emerged. Consistent with our prior analyses in HeLa cells (31), a vast majority of native CpA sites (82% (3,635/4,458)) in these cells are centromeric. In contrast, only 33% (388/1192) of CpA-TAP were enriched at centromeres. Thus, native CpA has the propensity for centromeric deposition greater than twice that of CpA-TAP (Fig. 2a). In the non-centromeric or ectopic fraction, both native CpA and CpA-TAP share more than 50% common hotspots (428/823 and 428/804, respectively). These data suggest that native and tagged CENP-A share more commonalities in their ectopic “off pathway” fraction than accurate HJURP-mediated deposition at centromeres.

When the hotspots are categorically separated into intergenic (sites not classified as either TSS, TTS, exon, 5’ UTR, 3’ UTR, and intron, but includes centromeres), exon, intron, promoters, and other types of domains, the differences between the two proteins are magnified. Native CpA makes up 93.6% of intergenic domains, while CpA-TAP only 66.8% (Fig. 2b). CpA-TAP makes up greater than 10-fold enrichment at exons (0.3% for native CpA versus 3.9% for CpA-TAP, respectively) and promoters (1.4% versus 10.6%, respectively), 5-fold enrichment at other/uncategorized domains (0.9% versus 4.9%, respectively), and 3-fold enrichment at introns (4.0% versus 13.8%, respectively) (Fig. 2b).

Fig. 2

ChIPseq analysis of native CpA an CpA-TAP in control HeLa cells. a) Venn diagram depicting native CpA versus CpA-TAP total, centromeric, and non-centromeric hotspots. b) Categorical dissection of incorporated sites for native CpA versus CpA-TAP. c) Heat map of promoter occupancy for native CpA unique, CpA-TAP unique, and common sites

Global hotspot occupancy analysis reveals a stark contrast between native CpA and CpA-TAP. CpA-TAP occupancy at unique native CpA sites is lost, while CpA-TAP is enriched at sites that are specific to CpA-TAP (Fig. 2c). However, there are common sites between native CpA and CpA-TAP that remain unchanged (Fig. 2c).

We observed that CpA-TAP is stably bound to chromatin but depleted for CENP-C (Fig. 1c) and has a higher rate of deposition at non-centromeric domains (Fig. 2a-b). In previous works, we and others reported that CENP-A accumulates at non-centromeric sites in cancer cells [31, 53], and that this non-native pathway exploits the H3.3 chaperone, DAXX [52, 54, 55]. Therefore, we sought to determine whether CpA-TAP utilizes the DAXX-mediated pathway to deposit to these ectopic regions.

DAXX promotes non-centromeric deposition of both native and tagged CpA

Centromeric CENP-A relies on the chaperone HJURP to deposit at centromeres [20,21,22]. The predominant 67% (804/1192) non-centromeric deposition of CpA-TAP (Fig. 2a) led us to speculate that addition of the C-terminal TAP-tag can elicit the recruitment of an alternative chaperone. Previous reports suggest that ectopic CENP-A can be deposited by DAXX [52], and that its mis-localization is determined by the sensitive balance among chaperones HJURP, DAXX, and HIRA [54]. Westerns against S-tag (CpA-TAP) and DAXX were performed with HeLa histones (extracted with hydroxlyapatite and high salt), recombinant DAXX (rDAXX), and CpA-TAP transfected into both Control and DAXX KO cell lines confirmed DAXX was not expressed in the DAXX KO cell line (Fig. 3a).

In the DAXX KO cells, co-IF show partial colocalization between CpA-TAP and CENP-C on few centromeres outside of mitosis (Fig. 3b). Similarly, native CpA ChIP was enriched with CENP-C, while CpA-TAP did not pull-down detectable levels of CENP-C (Fig. 3b), consistent with the previous result that CpA-TAP has a lower affinity for CENP-C (Fig. 1c). To determine whether CpA-TAP utilizes DAXX as an alternative chaperone for the 67% sites that are non-centromeric (Fig. 2a), we performed a similar ChIPseq experiment after purifying native CpA or CpA-TAP but using the HeLa DAXX KO cell line. Native CpA (in the DAXX KO background) was further enriched at centromeres from 82% (in untransfected control) to 99% (21,984/22,448), while CpA-TAP at centromeres acquired a moderate increase from 32 to 47%.

(2,119/4,500) in the DAXX KO cell line (Fig. 3c). Noncentromeric deposition of native CpA decreased from 18% (823/4,458) to 2% (464/22,448) and CpA-TAP from 67% (804/1,192) to 53% (2,381/4,500) in the DAXX KO cell line (Fig. 3c), indicating DAXX plays a role in ectopic deposition of both native CpA and CpA-TAP.

Fig. 3

ChIPseq analysis of native CpA-TAP in DAXX KO HeLa cells. a) Western confirmation that DAXX is knock-out and that CpA-TAP is esxpressed (rDAXX: recombinant DAXX protein, AbCam cat #ab131785). b) CoIF of CpA-TAP and native CENP-C during interphase and mitosis (left panel), and native CpA versus S-tag ChIP followed by CENP-C Western (right panel). Scale bar = 5 μm. c) Venn diagram detailing total, centromeric, and non-centromeric hotspots for native CpA and CpA-TAP. d) Categorical dissection of native CpA versus CpA-TAP incorporated sites

Categorical dissection of the different occupied native CpA versus CpA-TAP sites in the DAXX KO cells add another intriguing layer of dynamics between the two. Native CpA now occupies 99% of intergenic sites (which includes centromeric regions) in the DAXX KO (Fig. 3d), compared to 93.6% (Fig. 2b). In the case of CpA-TAP, there is a 5- fold reduction from 3.9 to 0.8% at exons, and 10-fold reduction from 10.6 to 0.9% at promoters (Fig. 3c). One interesting point is that CpA-TAP at introns increased from 13.8% (Fig. 2b) to 20.3% (Fig. 3c) in the DAXX KO cell line, suggesting an alternative chaperone such as HIRA may be taking DAXX’s place, and that DAXX was repressing HIRA’s function at introns.

Native CpA and CpA-TAP deposition at centromeres is enhanced and partially restored upon DAXX KO, respectively

With the exception of chromosome 5, native CpA (green) in control cells is predominantly centromere specific across all chromosomes (Fig. 4). However, CpA-TAP (red) is essentially void at centromeres on chromosomes 2–3, 6–18, 20–22, and X; moderately reduced at the centromere on chromosome 4; and mildly diminished on chromosomes 1, 5, and 19 in control cells when compared to native CpA (Fig. 4).

When DAXX is knocked-out, native CpA domains (blue) are noticeably enriched at centromeres on chromosomes such as chromosomes 17 and 18 (Fig. 4), suggesting DAXX knockout can lead to centromeric expansion. The most fascinating observation is that CpA-TAP returns to centromeres on all chromosomes in the DAXX KO cell line (Fig. 4). The DAXX KO cell line revealed that though both native CpA and CpA-TAP are enriched at centromeres, knocking out DAXX may simultaneously increase ectopic deposition for both proteins on most chromosomes (Fig. 4).

Fig. 4

Karyoplot analysis of native CpA versus CpA-TAP deposition in both control (untransfected) HeLa and DAXX KO cells

Introducing tagged CpA disrupts and redistributes native CpA in control HeLa and DAXX KO cells

It is not known whether introduction of foreign CpA-TAP can disrupt native CpA deposition within the genome. To ascertain the impact (if any), we transfected CpA-TAP to both HeLa and HeLa cells where DAXX is knocked out, followed by first serial depletion of CpA-TAP with S-tag ChIP and then native CpA ChIP. Much to our surprise, CpA-TAP introduction led to shrinkage of the native CpA centromeric domain with simultaneous expansion of the ectopic domains (red native CpA+ CpA−TAP) when compared to native CpA (green) under control HeLa conditions (Fig. 5). Centromeric deposition of native CpA was either gained or loss depending on the chromosome (blue), but non-centromeric domains were significantly expanded when CpA-TAP was introduced and DAXX was knocked-out (blue native CpA+ CpA−TAP +DAXX KO) compared to native CpA (green) (Fig. 5). A summary of observations for native CpA versus CpA-TAP among various conditions can be found in Table 2.

Fig. 5

Karyoplot analysis of native CpA in control HeLa cells, versus in the presence of CpA-TAP, and versus in the presence of CpA-TAP + DAXX KO.

Table 2 Assessing centromeric versus non-centromeric deposition of native CpA versus CpA-TAP under various conditions. + to +++ : range of deposition levels; - : no deposition observed; N/A : not applicable; … : range of varying degrees of deposition, depending on chromosomeCRISPR knock-in of CpA-TAP does not recapitulate native CpA function

To determine whether CpA-TAP can fully replace native CpA in vivo, we performed a knock-in of the C-terminal TAP-tag to the endogenous CpA locus in both control HeLa and HeLa cells where DAXX has been knocked out. Six colonies from each cell line were isolated and expanded for further downstream applications. In the case of HeLa control cells, CpA-TAP knock-in led to all six colonies not surviving past 7 days (Table 3). For HeLa cells where DAXX was knocked out, only colonies #2, 4, 5, and 6 survived, continued to divide (Table 3), and were confirmed heterozygous (Fig. S2a). Colonies #1 and #3 failed to divide and perished after 1 month.

Table 3 CRISPR knock-in of CpA-TAP to recapitulate native CpA function. - : non-viable colonies; + : colonies that grew and maintained under puromycin selection

After 5 months, the same heterozygous colonies in the DAXX KO background were assessed for CpA-TAP protein expression, and intriguingly, all colonies that formerly expressed CpA-TAP no longer did (Fig. S2a-b), suggesting cells were preferentially expressing native CpA while silencing CpA-TAP.

N-terminal SNAP-CpA has reduced de novo early G1 phase deposition

HJURP-dependent CENP-A deposition at centromeres occurs during G1 phase (20, 21). To determine whether a tagged version of CENP-A impacts deposition during early G1 phase, we utilized a similar approach to Jansen, et al. but cloned an N-terminally SNAP-tag CENP-A downstream to the CMV promoter [56]. The transiently transfected cells were synchronized with a double thymidine block, pulse-chase labeled with TMR Star (red), and coIF with anti- native CpA (green). Though we typically observe > 90% transfection efficiency with electroporation, among the 200 cells with native CpA IF centromeric signals observed, < 50% of those SNAP-CpA containing cells showed colocalization with native CpA (co-IF with native CpA in green) (Fig. S3 and S4).

Native CpA, CpA-TAP, and GFP-CpA vary in centromeric deposition

Much of the study thus far has relied on a single C-terminally tagged CpA-TAP construct. We were curious whether we would observe similar differences if we were to utilize an N-terminally.

tagged GFP-CpA, which has been previously reported in IF studies [11, 12]. GFP-CpA is 40 kD, slightly larger than CpA-TAP (Fig. S1). While total sites detected by GFP-CpA are fewer than native CpA and CpA-TAP, all its centromeric sites coincide with native CpA and the majority of its non-centromeric sites are shared between native CpA and/or CpA-TAP (Fig. 6a). When examining each protein’s centromeric affinity, native CpA tops at 81%, followed by GFP-CpA (54%), and CpA-TAP (32%) (Fig. 6b).

Further inspection of the different loci provides details of how the three proteins behave. For instance, at the centromere and one of the TTS, native CpA is significantly enriched while CpA-TAP and GFP-CpA are nearly void (Fig. 6c). On the flip side, CpA-TAP and GFP-CpA (to a lesser degree) have higher affinity for certain exons, promoters, and TSS compared to native CpA (Fig. 6c).

Fig. 6

Native CpA have differing deposition profiles compared to N-terminally tagged GFP-CpA and C-terminally tagged TAP. a) Triple Venn diagram highlighting overlapping and non-overlapping sites among native CpA, CpA-TAP, and GFP-CpA. b) Percentage of centromeric versus non-centromeric sites of native CpA, CpA-TAP, and GFP-CpA under various treatments. c) Peak snapshots of several genic regions for native CpA, CpA-TAP, and GFP-CpA

Tagging histone variants disrupt nucleosomal protein-protein interactions

It is not known whether introduction of a tagged histone variant can impact protein-protein interactions within the nucleosomal context. Of the several available native versus untagged ChIPseq datasets for histone variants, linker histone H1.5 was readily available [57, 58] and its interaction with CENP-A previously deemed unlikely [59]. To determine whether H1.5 (GSM5076929) and tagged H1.5-HA (GSM1197474) histones had altered genomic deposition, we turned to previously reported ChIPseq results [57, 58] and reanalyzed both sequencing datasets for total and centromeric sites. Out of the total number of sites for native H1.5, 664/3,163 (21%) colocalized with total native CpA (Fig. 7a); Whereas 27/98,943 (0.03%) H1.5-HA sites colocalized with total native CpA (Fig. 7a). When we narrowed down our search to overlapping centromeric sites with native CpA, 308/1,763 (17%) of native H1.5 are found with native CpA; However, no centromeric and overlapping native CpA sites were found when examining H1.5-HA (Fig. 7a).

Fig. 7

Tagged and untagged histone variants differ in genome wide distribution and nucleosomal interactions. a) Previously reported H1.5 and H1.5-HA ChIPseq sites were compared to native CpA ChIPseq sites from this study. b) ChIP performed against native CpA, GFP-CpA, and HA-CpA mono-nucleosomes and probed for histone H1.5 (Invitrogen Cat #711,912). HC = heavy chain. c) Ratios of H1.5/native or tagged CpA, normalized against H1.5/native CpA, from 2–5 independent experiments. Error bars = SEM

Previously, it was reported that epitope tagged CENP-A failed to interact with histone H1.5 [59].We hypothesized that tagging CENP-A could disrupt the H1.5 interaction. To test this, we performed native CpA ChIP on mono-nucleosomes from untransfected control HeLa versus GFP-CpA and HA-CpA ChIP in transfected HeLa cells, followed by Western analysis. Similar to previous results when adding a TAP-tag to CpA disrupts CENP-C binding (Fig. 1c), our data indicates tagging CpA may disrupt histone H1.5 interactions, depending on the type of engineered tag utilized (Fig. 7b-c and Fig. S3a-c). Sequential native CpA followed by native H3 ChIP revealed H3 binds H1.5 with a ~ 13-fold higher affinity than native CpA (Fig. 7c).