Timing of assembly and turnover of CENP-A at centromeres using the SNAP tag
The SNAP tag, a modified variant of the suicide enzyme O6
-alkylguanine-DNA alkyltransferase, whose normal function is in DNA repair, has been extensively engineered to covalently and irreversibly modify (and inactivate) itself through acceptance of the cell-permeable guanine derivative O6
-benzylguanine (BG; or fluorescent derivatives thereof). In effect, this allows labeling of SNAP fusion proteins at will in vivo (Keppler et al., 2003
, 2006). We applied pulse labeling with this methodology to determining CENP-A turnover specifically at centromeres () as well as quench-chase-pulse labeling to follow the fate of newly synthesized CENP-A (). We established cell lines stably expressing centromere-localized CENP-A–SNAP at near endogenous levels in HeLa cells ( and Fig. S1, available at http://www.jcb.org/cgi/content/full/jcb.200701066/DC1
Figure 1. Principle of SNAP tag–based pulse labeling. (A and B) Schematic of labeling strategies for pulse-chase labeling (A) or quench-chase-pulse labeling (B) of CENP-A–SNAP fusion protein with BG (BG-block; quench) or TMR-Star (pulse). (C) The (more ...)
Multiple lines of evidence indicated that CENP-A with the SNAP tag substituted functionally for CENP-A in centromere maintenance. We have previously reported that transgene-encoded CENP-A expression leads to reduction of the endogenous CENP-A pool through competition at the protein level (Foltz et al., 2006
). Here, a similar reduction in endogenous CENP-A in response to CENP-A–SNAP expression resulted in an unchanged overall CENP-A pool, the majority of which was SNAP tagged (Fig. S1 A; line 23 is used for all pulse-labeling experiments). Because chronic reduction of CENP-A to <50% is a cell-autonomous lethal event (Black et al., 2007b
), the SNAP-tagged CENP-A pool in the stable CENP-A–SNAP cell lines not only competed for assembly at centromeres with authentic CENP-A but also provided essential aspects of CENP-A function in centromere maintenance. (Retention of substantial CENP-A function by CENP-A–SNAP [219-aa tag] is in agreement with what has been shown for N-terminally YFP-tagged [240 aa] or C-terminally tandem affinity–tagged [TAP; 172 aa] CENP-A, which, respectively, rescue CENP-A lethality and incorporate into bona fide centromeric nucleosomes that are associated with a six-member complex of centromere components [CENP-A–TAP; Foltz et al., 2006
; Black et al., 2007b
].) 15-min pulse labeling with the tetramethylrhodamine (TMR)-conjugated SNAP substrate, TMR-Star
, specifically identified CENP-A–SNAP already assembled into centromeric chromatin (, top). Preincubation of CENP-A–SNAP–expressing cells with the nonfluorescent SNAP substrate (BG-block) led to complete quenching of SNAP and rendered CENP-A undetectable with TMR-Star
CENP-A is stably associated with centromeres across the cell cycle
To determine turnover of CENP-A at centromeres, cells were synchronized at the G1–S boundary by tandem treatments with thymidine. CENP-A bound to unreplicated centromeres was pulse labeled with TMR-Star
and chased for up to two cell cycles ( and ). Consistent with earlier reports on total CENP-A levels that had indicated slow protein turnover (Shelby et al., 2000
; Regnier et al., 2005
), centromere duplication in the initial round of DNA synthesis produced a 60 ± 14% reduction in intensity of TMR-Star
–labeled CENP-A–SNAP at individual centromeres by the first mitosis and through the subsequent G1 (). After a second cycle of DNA replication, the previously labeled, centromere-bound CENP-A–SNAP was diminished to 25 ± 5% of its initial level, whereas the total number of fluorescent centromeres positive per cell remained unchanged throughout the experiment (). Thus, despite continued synthesis of both SNAP-tagged and endogenous CENP-A, CENP-A already loaded into centromeric chromatin by late G1 is redistributed to, and retained by, daughter centromeres.
Figure 2. CENP-A turnover at centromeres. (A) Outline of cell synchronization and labeling regimen for CENP-A turnover. Cells were synchronized and labeled as depicted followed by fixation and immunofluorescence with anti-HA. Representative images for each time (more ...)
Loading of newly synthesized CENP-A initiates in telophase
CENP-A must be replenished at centromeres after DNA replication to complete duplication of new centromeres. To determine the timing of CENP-A incorporation into chromatin of newly replicated centromeres, cells stably expressing CENP-A–SNAP were synchronized at the G1–S boundary by double thymidine block, and both centromere-associated and any free pool of CENP-A–SNAP were quenched with nonfluorescent BG ( and ). The cells were then released into S phase for 6.5 h, nascent CENP-A–SNAP was pulse labeled for 15 min by reaction with TMR-Star
, and incorporation of the fluorescent CENP-A was examined in late S, G2, M, and the subsequent G1. CENP-A synthesized during S phase was diffusely localized in the nucleus ( and Fig. S2, available at http://www.jcb.org/cgi/content/full/jcb.200701066/DC1
) but did not appear at centromeres during any phase of G2 or M. Rather, only after passage through mitosis and entry into the G1 phase of the next cell cycle did CENP-A assemble at centromeres (). Quantification of the number of cells positive for CENP-A loading confirmed that the initially synchronized cell population did not load substantial levels of CENP-A before ~11 h after release from thymidine and concomitant with entry into G1 ().
Figure 3. CENP-A loading initiates in telophase/early G1. (A) Schematic of cell synchronization and labeling protocol. Cell cycle stages are estimates based on time elapsed after release from double thymidine–induced arrest at G1–S. Representative (more ...)
Close examination of cells just before and after mitotic entry revealed that the earliest time CENP-A loading could be detected at centromeres was concomitant with nuclear envelope reformation and completion of furrow ingression as indicated by midbody formation in late telophase/early G1 ( and Fig. S3, available at http://www.jcb.org/cgi/content/full/jcb.200701066/DC1
). No newly made CENP-A was observed at centromeres at any time during mitosis before telophase. The absence of CENP-A at centromeres at these earlier time points cannot be attributed to a pool of newly synthesized CENP-A too small to be detected because the prelabeled pool size is the same for all time points. The pattern of loading restricted to late mitosis/early G1 was not a result of thymidine treatment per se because randomly cycling cells were also dependent on mitotic progression to permit CENP-A loading (Fig. S1).
To time the arrival of CENP-A–SNAP at the centromere more accurately, we followed live cells containing a pool of TMR-Star
–labeled but nonassembled CENP-A–SNAP from metaphase through early G1 (). TMR-Star
labeling of live cells resulted in the nonspecific retention near the cell periphery (presumably in internal membranes) of a proportion of the dye, a proportion that is removed during normal fixation and washing conditions. Nevertheless, no TMR-Star
signal could be detected specifically at centromeres in metaphase (). However, assembly of nascent CENP-A–SNAP could be detected as early as ~50 min after anaphase onset and continued for several hours in early G1 ( and Video 1, available at http://www.jcb.org/cgi/content/full/jcb.200701066/DC1
CENP-A assembly occurs exclusively during G1
Next, we determined whether loading of CENP-A is unique to the early hours of G1 or whether loading is also permissive at any other point in the cell cycle, including the possibility of a secondary CENP-A loading stage, as has been suggested in fission yeast (Takahashi et al., 2005
). CENP-A–SNAP in mitotic cells arrested with nocodazole treatment was initially quenched with nonfluorescent BG, and a G1 phase cell population was generated by release from nocodazole arrest. The CENP-A–SNAP pool produced during mid-to-late G1 was labeled and monitored for timing of centromeric deposition (). DNA content was assayed by FACS to verify cell cycle position (unpublished data). No assembly of CENP-A–SNAP–TMR-Star
was detectable at centromeres during the subsequent S, G2, or M phases (). However, fluorescent CENP-A from the prior G1 assembled into the new daughter centromeres after exit from this subsequent mitosis (). Thus, despite the presence of a stable noncentromeric CENP-A pool, no loading occurred at any stage of the cell cycle before the following G1 phase (10–18 h after CENP-A synthesis and labeling).
Figure 4. Nascent CENP-A loads exclusively during G1. (A) Schematic of cell synchronization and late G1 labeling protocol. (B) Representative images at indicated time points. Percentages of cells that have loaded CENP-A are indicated below. Note that the 11% of (more ...)
Although SNAP-tagged CENP-A faithfully tracks to centromeres and provides an essential function of CENP-A in centromere maintenance, there remained the possibility that the SNAP tag or the cell synchronization methods we have used would interfere with the timing of CENP-A loading. If CENP-A loading is normally restricted to early G1, levels of CENP-A (tagged or endogenous) on individual centromeres should double from early mitosis to when loading is completed, in late G1. On the other hand, if CENP-A loading occurs before mitosis, as previously proposed (Shelby et al., 2000
), CENP-A levels in mitosis and G1 would be similar. Examination using indirect immunofluorescence to track endogenous CENP-A or direct fluorescence measurement of a cell line stably expressing YFP–CENP-A revealed that in both cases CENP-A levels increased from M to late G1 (~3.4- and ~2.5-fold, respectively; ), findings only consistent with CENP-A loading in G1 rather than before mitosis.
Figure 5. Centromeric levels of endogenous CENP-A increases during G1 phase. (A) HeLa cells stably expressing YFP–CENP-A (Kops et al., 2004) or parental HeLa cells were synchronized by tandem treatments with thymidine and released, and cells in mitosis (more ...)
Passage trough mitosis is critical for CENP-A assembly in early G1
The discrete, abrupt onset of CENP-A assembly as cells exit from mitosis suggested that passage through mitosis is a prerequisite for CENP-A assembly. Alternatively, entering the G1 cell cycle state may be triggering CENP-A assembly without any mechanistic involvement of mitosis per se. To distinguish these possibilities, the G1 cell cycle phase was disconnected from mitotic passage by combining the SNAP-based CENP-A assembly assay with a classic cell–cell fusion approach (Rao and Johnson, 1970
). Heterophasic heterokaryons were generated by fusing G1 cells with G2 cells, each expressing CENP-A–SNAP and each uniquely marked by stable expression of CFP-tagged histone H3 or tubulin, respectively, to mark nuclei or microtubules (). The two differentially marked CENP-A–SNAP cell populations were synchronized by double thymidine treatment. Previously deposited CENP-A–SNAP was quenched, and each population was released for differing lengths of time so as to produce two synchronized populations, one of which was at mitosis/early G1 (H3-CFP cells) and the other at late S/early G2 phase (CFP-tubulin cells). The two populations were mixed, and cell fusion was induced with polyethylene glycol (PEG). Nocodazole was added to prevent any further passage through mitosis. 4 h after fusion, TMR-Star
labeling was used to assay in both nuclei of the heterokaryons for assembly at centromeres of the unloaded, newly synthesized CENP-A–SNAP pool that was present in all nuclei ().
Figure 6. Passage through mitosis is critical for CENP-A loading in early G1. (A) Schematic of cell synchronization, labeling, and PEG-mediated cell–cell fusion protocol. CENP-A–SNAP cells marked with H3-CFP or CFP-tubulin were sequentially released (more ...)
After cell–cell fusions, nuclei originating from G2 and G1 cells share the same cytoplasm and, in principle, the same cell cycle state. Control fusions revealed that, as expected, loading of CENP-A in both nuclei occurred exclusively in G1 cell to G1 cell fusions. In contrast, no binucleate heterokaryons derived from fusion of two G2 populations could be found in which both nuclei loaded CENP-A (), and the vast majority (86%) loaded it in neither nucleus.
It should be noted that because of the short time cells spend in mitosis (~1 h) and the inherent spread in synchrony as cells transverse across the cell cycle, an early G1 phase cell population will invariably contain a fraction of cells that are in G2. (In this case, 33% of the H3-CFP G1 population had in fact not yet reached G1 by the time cells were fused.) Therefore, in all fusions, a spread of heterokaryons loading CENP-A–SNAP at centromeres at one, both, or neither of the nuclei is expected. Nevertheless, despite this inherently imperfect synchrony, a striking finding was that in binucleate heterokaryons derived from fusion between cell populations enriched in G1 and G2, most (66%) G1 cell–derived nuclei (H3-CFP marked) recruited CENP-A–SNAP to centromeres to levels indistinguishable from surrounding nonfused G1 cells. In contrast, no heterokaryons were found that had assembled CENP-A in both nuclei, indicating that G2-derived nuclei, although sharing the same cytoplasm with a CENP-A–assembling G1-derived nucleus, did not assemble CENP-A (), despite the presence of fluorescently labeled CENP-A–SNAP. The frequency of heterokaryons loading CENP-A–SNAP in one or neither nucleus corresponded to the frequency of H3-CFP G1 and G2 cells at the time of fusion (, arrows), indicating that in heterokaryons the G1- and G2-derived nuclei are neither inducing nor inhibiting CENP-A assembly in the other nucleus. Therefore, the early G1 cell cycle state that is permissive for CENP-A assembly does not directly dictate the ability to load CENP-A. Rather, passage through mitosis is crucial to allow CENP-A assembly as cells enter G1.
Microtubule attachment is not required for CENP-A assembly in G1
Our experiments suggest that mitosis is a key cell cycle determinant in initiating CENP-A loading. To exclude the possibility that proficiency for CENP-A loading is determined by a “timing” mechanism rather than actual mitotic passage and G1 entry, cells were arrested using nocodazole to produce a nascent unloaded pool of CENP-A–SNAP in mitosis. Nocodazole-treated cells never assembled CENP-A–SNAP, even by the time 94% of control cells had reentered G1 and loaded CENP-A–SNAP (), reaffirming the notion that exit from mitosis is required for CENP-A loading.
Multiple processes occur during mitosis that might act to trigger new CENP-A nucleosome recruitment. These include chromatin stretching, which occurs during metaphase and has been proposed as a mechanism for the exchange of histone H3 for CENP-A–containing nucleosomes (Ahmad and Henikoff, 2002
; Mellone and Allshire, 2003
; Carroll and Straight, 2006
). Although the concept of functional reinforcement of centromere location that is part of this model is appealing, no experimental evidence has been generated in support for such a mechanism. Alternatively, DNA decondensation or the presence of other mitotic kinetochore components may be integral to triggering the process of centromere assembly.
To test the tension-dependent CENP-A loading model, cells were produced that completed mitosis in the absence of microtubule attachment (and therefore microtubule-mediated chromatin stretching). To do this, cells were depleted of BubR1 with transcription-mediated short hairpin RNA and treated with nocodazole to block microtubule assembly, and CENP-A loading was assessed (). Under these conditions, cells enter mitosis without spindle assembly or kinetochore attachment, but quickly exit without the BubR1-dependent mitotic checkpoint (Kops et al., 2004
Figure 7. Microtubule attachment in mitosis is not required for CENP-A assembly at centromeres. (A) Schematic of cell synchronization, transfection, and labeling protocol. (B) Representative image of cells transiently transfected with a transcription-mediated BubR1 (more ...)
Depletion of BubRI alone did not affect the ability of cells to load CENP-A, whereas nocodazole treatment of cells with normal BubR1 levels prevented mitotic exit and any loading (). Nocodazole treatment of cells depleted of BubR1 () or another mitotic checkpoint component Mad2 (not depicted) produced CENP-A loading to levels comparable to that seen in untreated cells, along with normal interphase nuclei with twice the number of resolved centromeres. This was indicative of a successful mitotic exit where sister chromatids had disjoined but failed to segregate and complete cytokinesis because of the absence of microtubules. Conversely, cells that did not load CENP-A were either arrested in mitosis (i.e., not depleted in BubRI) or had not yet entered mitosis (; as indicated by a centromere number consistent with unresolved sister centromeres, as is the case in G2 phase). Thus, passage through mitosis is critical for CENP-A loading, but microtubule attachment or microtubule-generated tension across centromeric chromatin is not.