|Home | About | Journals | Submit | Contact Us | Français|
Centromeres mediate the conserved process of chromosome segregation, yet centromeric DNA and the centromeric histone, CENP-A, are rapidly evolving. The rapid evolution of Drosophila CENP-A loop 1 (L1) is thought to modulate the DNA-binding preferences of CENP-A to counteract centromere drive, the preferential transmission of chromosomes with expanded centromeric satellites. Consistent with this model, CENP-A from Drosophila bipectinata (bip) cannot localize to Drosophila melanogaster (mel) centromeres. We show that this result is due to the inability of the mel CENP-A chaperone, CAL1, to deposit bip CENP-A into chromatin. Co-expression of bip CENP-A and bip CAL1 in mel cells restores centromeric localization, and similar findings apply to other Drosophila species. We identify two co-evolving regions, CENP-A L1 and the CAL1 N terminus, as critical for lineage-specific CENP-A incorporation. Collectively, our data show that the rapid evolution of L1 modulates CAL1-mediated CENP-A assembly, suggesting an alternative mechanism for the suppression of centromere drive.
Centromeres are essential chromosomal structures to which kinetochore proteins and microtubules are recruited during cell division to mediate the accurate distribution of genetic material. While centromere function is highly conserved, centromere size and structure vary greatly between organisms (Fukagawa and Earnshaw, 2014). In complex eukaryotes, the specific DNA sequences found at centromeres are neither necessary nor sufficient for centromere formation (Choo, 2000; Karpen and All-shire, 1997), and centromeres are epigenetically defined by the presence of a centromere-specific histone H3 variant called CENP-A (also called CID in Drosophila) (Earnshaw and Rothfield, 1985; Karpen and Allshire, 1997).
Accurate CENP-A deposition is mediated by specific CENP-A assembly factors (or chaperones). While yeast and humans harbor CENP-A chaperones with common ancestry (called Scm3 and HJURP, respectively) (Bernad et al., 2011; Camahort et al., 2007; Dunleavy et al., 2009; Foltz et al., 2009; Mizuguchi et al., 2007; Pidoux et al., 2009; Sanchez-Pulido et al., 2009), Drosophila employ an evolutionarily distinct CENP-A chaperone called CAL1 (Chen et al., 2014; Erhardt et al., 2008; Phansalkar et al., 2012).
Despite the universally conserved function of centromeres in maintaining genome integrity, both CENP-A (Cooper and Henikoff, 2004; Finseth et al., 2015; Henikoff et al., 2001; Malik and Henikoff, 2001; Malik et al., 2002; Ravi et al., 2010; Schueler et al., 2010; Talbert et al., 2002; Zedek and Bureš, 2012) and centromeric DNA (Melters et al., 2013) are rapidly evolving. This paradox has been explained by the centromere drive hypothesis, which proposes that CENP-A adaptively evolves to maintain meiotic parity by modulating its DNA-binding preferences to counteract the transmission advantage gained by satellite expansion in female meiosis (Henikoff and Malik, 2002; Malik and Henikoff, 2002). In support of this model, adaptive evolution has been observed in both the N-terminal tail and loop 1 (L1) of CENP-A (Cooper and Henikoff, 2004; Finseth et al., 2015; Henikoff et al., 2001; Malik and Henikoff, 2001; Malik et al., 2002; Ravi et al., 2010; Schueler et al., 2010; Talbert et al., 2002; Zedek and Bureš, 2012), both of which are putative DNA-binding regions (Luger et al., 1997; Malik et al., 2002; Vermaak et al., 2002), in plants and animals. The role of CENP-A chaperones in this evolutionary “arms race” has yet to be explored.
Somewhat surprising is the fact that, while Drosophila CENP-A is adaptively evolving (Malik and Henikoff, 2001; Malik et al., 2002), its chaperone CAL1 is highly conserved across both the N-terminal domain, which interacts with CENP-A, and the C-terminal domain, which interacts with CENP-C (Chen et al., 2014; Phansalkar et al., 2012; Schittenhelm et al., 2010). How CAL1 is able to interact with and deposit rapidly evolving CENP-A orthologs, given their different rates of evolution, is unknown.
While several lines of evidence support the rapid evolution of both centromeric DNA and CENP-A in many species (Melters et al., 2013), and also the influence of centromere expansion on meiotic segregation distortion (Chmátal et al., 2014; Daniel, 2002; Fishman and Saunders, 2008; Fishman and Willis, 2005; Pardo-Manuel de Villena and Sapienza, 2001; Wyttenbach et al., 1998), biological data supporting a direct correlation between the evolution of centromeric DNA and CENP-A (the second step in the centromere drive hypothesis (Malik, 2009; Malik and Henikoff, 2002)) are lacking. However, one striking experimental observation supporting centromere drive is that CENP-A from Drosophila bipectinata (bip) expressed in Drosophila melanogaster (mel) tissue culture cells is unable to localize to mel centromeres (Vermaak et al., 2002). This incompatibility is the result of specific amino acid changes in L1 of CENP-A (Vermaak et al., 2002). Because L1 of histone H3 has been shown to interact with DNA (Luger et al., 1997), it was proposed that L1 of CENP-A is adaptively evolving with centromeric DNA satellites to suppress centromere drive (Malik and Henikoff, 2001; Vermaak et al., 2002). However, recent structural studies of human CENP-A octamers and tetramers suggest that L1 of CENP-A does not interact with DNA, and instead is exposed in the nucleosome particle (Sekulic et al., 2010; Tachiwana et al., 2012). Interestingly, in yeast and humans, a domain encompassing L1 known as the CENP-A targeting domain (CATD) is recognized by the assembly factors Scm3 and HJURP, respectively (Bassett et al., 2012; Cho and Harrison, 2011). The CATD is sufficient to confer centromeric localization to histone H3 in both yeast and humans (Black et al., 2004; Shelby et al., 1997). However, the corresponding region of Drosophila CENP-A is not sufficient for the centromeric localization of histone H3 in flies (Moreno-Moreno et al., 2011). How CAL1 recognizes Drosophila CENP-A is unknown.
Here, we use evolutionary cell biology to investigate the relationship between centromere divergence and CENP-A assembly in Drosophila. We reveal that a functional interplay between CAL1 and L1 of CENP-A is both necessary and sufficient for the deposition of orthologous CENP-A proteins at mel native centromeres as well as for de novo CENP-A recruitment to an ectopic locus. Successful CENP-A incorporation requires that L1 and the CAL1 N terminus are compatible, demonstrating that these two domains evolve in concert. These data challenge previous models of centromere drive involving the adaptive evolution of L1 with centromeric DNA in Drosophila and suggest that the evolution of L1 may instead mediate CENP-A centromeric deposition by CAL1.
Loop 1 (L1) of CENP-A has long been proposed to be adaptively evolving with centromeric DNA in an “arms race” akin to that occurring between viruses and their hosts (Malik and Henikoff, 2001; Vermaak et al., 2002). A previous study tested the ability of CENP-A orthologs from Drosophila simulans (sim), Drosophila erecta (ere), Drosophila lutescens (lut), Drosophila bipectinata (bip), and Drosophila pseudoobscura (pse) to localize to centromeres in D. melanogaster (mel) cultured Kc cells, to identify CENP-A centromere-targeting motifs (Vermaak et al., 2002). While centromeric localization was observed for CENP-A orthologs from most species, bip CENP-A failed to localize to mel centromeres (12 Ma diverged; Figure 1A), despite the fact that the more divergent pse CENP-A (30 Ma diverged; Figure 1A) was able to localize (Vermaak et al., 2002).
To better understand the relationship between centromeric localization of CENP-A orthologs in mel cells and their phylogenetic distance from mel, we tested additional CENP-A orthologs from four evolutionarily intermediate species between mel and bip (Drosophila takahashii [tak], Drosophila rhopolia [rho], Drosophila kikkawai [kik], and Drosophila ananassae [ana]), and from three more distant species (Drosophila miranda [mir], Drosophila willistoni [wil], and Drosophila virilis [vir]; Figure 1A), along with mel, sim, ere, bip, and pse as in the original study (Vermaak et al., 2002) for their ability to localize to mel centromeres (Figures 1B–1D and S1). GFP-tagged CENP-A orthologs from these 11 Drosophila species, as well as mel histone H3.1 as a control, were transiently expressed in mel Schneider 2 (S2) cells (Figure 1B). The localization of GFP-CENP-A orthologs was assessed by immunofluorescence (IF) on interphase S2 cells using anti-GFP and anti-mel CENP-A antibodies, which are specific to mel CENP-A and are used as a marker for mel centromeres (Figures 1C and S2).
Localization to mel centromeres was observed for those CENP-A orthologs that are most closely related to mel, namely sim, ere, tak, and rho. In contrast, bip, wil, and vir CENP-A failed to localize, resulting in diffuse GFP signal (p < 0.0001). kik and ana CENP-A partially localized to mel S2 centromeres, displaying both centromeric and diffuse GFP signal (p = 0.005 for kik and p = 0.0003 for ana). Interestingly, centromeric localization was also observed for CENP-A orthologs from the obscura group (pse and mir; 55% [p = 0.002] and 70% centromeric, respectively; Figures 1C, 1D and S1), which is more divergent from mel than either the montium or ananassae subgroups (Figure 1A). The same localization pattern was observed with hem-agglutinin (HA)-tagged CENP-A orthologs (Figures S3A–S3C), indicating that the presence of the GFP tag does not interfere with centromeric localization. Together, these findings confirm and expand upon previous work (Vermaak et al., 2002), and demonstrate that the CENP-A localization pathway is conserved between the melanogaster and obscura groups, but has diverged in the ananassae subgroup. Additionally, the CENP-A localization pathway is not conserved in more divergent lineages (e.g., wil and vir).
Unlike the rapid degradation of mislocalized mel CENP-A after pulse induction (Heun et al., 2006; Olszak et al., 2011), which requires the F-box protein PPA and the CATD (Moreno-Moreno et al., 2011), bip and wil CENP-A proteins appear to persist stably, resulting in higher protein levels compared with those of centromere-localizing CENP-A orthologs (Figure 1C). Perhaps mel PPA cannot recognize bip and wil CENP-A due to their divergent L1 (Vermaak et al., 2002), which is part of the CATD.
We next asked whether the localization of CENP-A orthologs followed a similar pattern at the centromeres of sim, a species closely related to mel (Figure 1A). Transient transfection with sim, mel, ere, bip, and pse GFP-CENP-A constructs in sim M-19 tissue culture cells was followed by IF with anti-GFP and anti-CENP-C antibodies, which recognize sim CENP-C providing a centromere marker (Figures S3D and S3E). Similar to the localization results in mel cells, mel, ere, and pse CENP-A localize to sim centromeres, while bip CENP-A does not (Figures S3D and S3E). These results show that the centromeric localization of CENP-A orthologs to mel and sim centromeres can only partially be explained by phylogenetic distance and that the branch containing the ananassae subgroup is evolving on a separate evolutionary trajectory from that of other close lineages. Furthermore, these experiments demonstrate that the incompatibility between CENP-A and the centromere is not unique to the bip/mel species pair.
The mislocalization of bip CENP-A in mel cells is due to key amino acid changes between L1 of bip and mel CENP-A (Vermaak et al., 2002). It was originally proposed that this variation in CENP-A L1 is indicative of adaptive evolution with centromeric DNA to suppress drive (Malik and Henikoff, 2001; Vermaak et al., 2002). However, another possibility is that bip CENP-A may be co-evolving (defined here as undergoing coordinated protein evolution) with its loading factor CAL1 (Chen et al., 2014), and that the failure of bip CENP-A to localize to mel centromeres could be due to an incompatibility with mel CAL1.
We investigated this possibility by first determining whether bip CENP-A can physically interact with mel CAL1. Immunoprecipitations (IPs) with anti-CAL1 antibodies coupled to beads were performed in two separate chromatin extracts that contained normalized amounts of mel or bip GFP-CENP-A. Quantification of GFP-CENP-A western blot bands indicated that mel CAL1 pulled down approximately 10% of mel GFP-CENP-A and 20% of bip GFP-CENP-A relative to the respective inputs. These experiments indicate that mel CAL1 can form a complex with bip GFP-CENP-A at least as efficiently as with mel GFP-CENP-A and that there is no incompatibility as far as physical interaction between these two proteins is concerned (Figure 2A).
We next investigated whether the ability of mel CAL1 to interact with bip CENP-A enables its deposition into chromatin. We turned to an ectopic tethering assay, which allows us to interrogate the functional relationship between mel CAL1 and CENP-A from bip and from other representative species without centromeric DNA as a contributing factor. Tethering mel CAL1 via the lac repressor, LacI, at a lacO array stably integrated within a chromosome arm leads to the stable incorporation of mel CENP-A (Chen et al., 2014). We co-expressed HA-tagged mel, ere, pse, bip, or wil CENP-A and an inducible mel CAL1 tagged with GFP and LacI (Figures 2B and 2C). After 24 hr induction of mel CAL1-GFP-LacI, recruitment of HA-CENP-A orthologs to the lacO site was analyzed by IF with anti-HA, anti-GFP (to detect CAL1-GFP-LacI at the lacO site), and anti-mel CENP-A antibodies (to visualize the mel endogenous centromere) on meta-phase chromosomes. This analysis showed that mel CAL1-GFP-LacI successfully recruits sim, ere, and pse CENP-A to the lacO site (Figure 2B). In contrast, bip and wil CENP-A are not recruited to the lacO site and localize all along the chromosome arms in a pattern reminiscent of mel CENP-A overexpression ((Heun et al., 2006); Figure 2B). These experiments suggest that, although mel CAL1 can interact with bip CENP-A (Figure 2A), this interaction is not functional, i.e., mel CAL1 cannot deposit bip CENP-A into chromatin (Figure 2B). Furthermore, they show that the successful ectopic targeting of CENP-A from sim, ere, and pse reflects their competency to localize to mel endogenous centromeres (Figure 1C). These data also demonstrate that the overexpression of mel CAL1-GFP-LacI is not sufficient to promote the centromeric or lacO targeting of bip and wil CENP-A.
While our ectopic targeting assays suggest that mel CAL1 cannot incorporate bip or wil CENP-A into chromatin at the lacO site, they did not allow us to discriminate between defective recruitment by mel CAL1 or an incompatibility between bip or wil CENP-A and DNA sequences present at mel endogenous centromeres. If the failure of bip CENP-A to associate with mel centromeres is solely due to an incompatible assembly factor (mel CAL1), then supplying bip CAL1 should rescue the centromeric localization of bip CENP-A in mel cells (Figure 3A).
To test this, we first needed to determine whether bip CAL1 can localize to mel centromeres, a necessary prerequisite for the deposition of CENP-A at this location (Chen et al., 2014; Erhardt et al., 2008). HA-tagged bip, ere, pse, or wil CAL1 were transiently expressed in S2 cells (Figure 3B). IF with anti-HA and anti-mel CENP-A antibodies showed that all of the HACAL1 orthologs localize to mel centromeres in at least 50% of cells (Figure 3C). Since CAL1 is recruited to centromeres by CENP-C (Chen et al., 2014), these data suggest that the CENP-C/CAL1 interaction is conserved between mel and bip and, more generally, across the Drosophila phylogeny. The observation that the C terminus of CAL1, which interacts with CENP-C (Chen et al., 2014; Schittenhelm et al., 2010), is under purifying selection (Phansalkar et al., 2012) is consistent with this hypothesis.
Next, we tested whether supplying bip CAL1 enables bip CENP-A to localize to mel centromeres by transiently transfecting mel S2 cells with bip GFP-CENP-A and bip HA-CAL1 constructs (Figure 3D). IF with anti-HA, anti-GFP, and anti-mel CENP-A antibodies on metaphase spreads showed that centromeric targeting of bip CENP-A is completely restored in 86% of cells and partially restored in 12% (Figures 3D and 3E). Furthermore, the observation that the centromeric bip GFP-CENP-A IF signal is resistant to salt extraction demonstrates that it is incorporated into chromatin (Figure S4). The centromeric and lacO targeting of bip CENP-A was also obtained with the reversed tags: bip CAL1-GFP-LacI with bip HA-CENP-A (Figures 3F–3H).
To test whether the functional interaction between bip CAL1 and CENP-A is lineage specific or species specific, we co-expressed bip HA-CENP-A with ana CAL1-GFP-LacI in mel lacO cells and assessed the recruitment of bip HA-CENP-A at the lacO site by IF with anti-HA, anti-GFP, and anti-mel CENP-A antibodies on metaphase spreads. We found that ana CAL1 is competent for bip CENP-A deposition at both mel centromeres (78% fully centromeric and 22% partially centromeric) and the lacO site (100%; Figures 3F–3H). We conclude that the presence of a lineage-specific CAL1 partner can also promote the centromeric targeting of bip CENP-A in mel cells.
To determine whether the centromeric localization of bip CENP-A can also occur in sim cells, we co-expressed bip CENP-A and bip CAL1 in M-19 cells and observed bip CENP-A centromeric targeting in 56% of cells (Figures S5A and S5B). These results are consistent with our findings in mel cells (Figures 3D and 3E) and demonstrate that a similar CENP-A loading defect is present between bip CENP-A and sim CAL1.
Next, we investigated whether a similar mechanism underlies the defective localization of the more divergent wil CENP-A to mel centromeres. We transiently co-expressed wil GFP-CENP-A with wil HA-CAL1, and assessed centromeric localization by IF. As with bip CENP-A, we observed exclusively centromeric localization of wil CENP-A in 92% of cells and partial localization in 8% (Figures 3D and 3E). We conclude that even CENP-A from a species almost 40 Ma diverged from mel can localize to mel centromeres as long as a compatible CAL1 partner is present.
Given the ability of the bip CENP-A/CAL1 complex to localize to mel centromeres, we next asked if this complex can initiate mel kinetochore assembly by assessing the recruitment of the outer kinetochore component Ndc80 (Meraldi et al., 2006). Bip CAL1-GFP-LacI was tethered to the lacO array in mel cells expressing bip HA-CENP-A followed by IF with anti-HA, anti-mel CENP-A, and anti-Ndc80 on metaphase spreads. We noticed that the full-length bip CAL1-GFP-LacI construct recruited mel CENP-A to the lacO site in approximately 50% of chromosome spreads (p < 0.0001 compared with mel CAL1-GFP-LacI recruitment of bip CENP-A), suggesting that there is more functional conservation between mel and bip CAL1 than between mel and bip CENP-A. By scoring bip CENP-A-positive lacO sites for both the presence or absence of mel CENP-A and Ndc80, we found that bip CAL1-GFP-LacI can recruit Ndc80 even when mel CENP-A is absent or nearly undetectable (72% compared with 99% when mel CENP-A is present at the lacO site [p = 0.2]; Figure 3I). We conclude that the bip CENP-A/ CAL1 complex can mediate mel kinetochore formation, bypassing the requirement for mel CENP-A. These results may explain why the co-expression of bip CAL1 and bip CENP-A does not negatively affect chromosome segregation, whereas the expression of ere CENP-A does (Figure S6). In plants, too, centromeric localization of CENP-A orthologs is not a predictor of whether they can form functional kinetochores (Ravi et al., 2010).
It has previously been shown that replacing L1 of mel CENP-A with the homologous region of bip CENP-A results in a loss of centromeric localization, while substituting L1 of bip CENP-A with mel L1 results in a gain of centromeric localization (Vermaak et al., 2002). Based on these data and our findings so far, we hypothesized that L1 of CENP-A could mediate the functional interaction with CAL1, and that the divergence of bip L1 (Vermaak et al., 2002) results in the failure of mel CAL1 to properly deposit bip CENP-A into chromatin.
To test this hypothesis, we generated a GFP-tagged mel CENP-A chimera containing L1 from bip CENP-A (mel CENP-AbipL1; Figure 4A) and transiently expressed it in mel S2 cells (Figure 4B) with and without bip CAL1. When expressed alone, mel CENP-AbipL1 is mislocalized in all mitotic chromosome spreads (0% centromeric). However, when mel CENP-AbipL1 is co-expressed with bip CAL1, it becomes centromeric in all spreads (100%; Figure 4C). A similar pattern was observed in interphase cells, where mel CENP-AbipL1 is mislocalized or only partially centromeric (82% and 19% of cells, respectively) when expressed alone, but becomes fully centromeric when co-expressed with bip CAL1 (90%; Figures 4D and 4E). Thus, the mis-localization of the mel CENP-AbipL1 chimera is the result of some sort of dysfunction occurring within the bip CENP-A L1 and the mel CAL1 complex.
If L1 is critical for the function of CENP-A and CAL1 complexes, the recruitment of bip CENP-A to the lacO site is expected to be restored if L1 from bip CENP-A is replaced with L1 from mel CENP-A (bip CENP-AmelL1 chimera; Figure 4F) (Vermaak et al., 2002). To test this prediction, we transiently transfected mel CAL1-GFP-LacI and HA-tagged bip CENP-AmelL1 in S2 lacO cells (Figure 4G), and assessed recruitment to the lacO site by IF on metaphase spreads. In agreement with previous data (Vermaak et al., 2002), bip CENP-AmelL1 chimera localizes to mel centromeres. Furthermore, mel CAL1-GFP-LacI recruits bip CENP-AmelL1 to the lacO array with the same efficiency as mel CENP-A (100%; Figure 4H) (Chen et al., 2014). These data demonstrate that the centromeric localization gained by the addition of mel L1 to bip CENP-A is a result of its restored ability to be incorporated into chromatin by mel CAL1.
Having determined that the divergence between the L1 of mel and bip CENP-A leads to defective centromeric deposition of bip CENP-A by mel CAL1, we sought to identify the corresponding regions of bip CAL1 that may have adaptively evolved with bip CENP-A. Such a region within bip CAL1 could confer mel CAL1 the ability to deposit bip CENP-A if introduced through amino acid swap experiments. Because the CENP-A interaction domain of CAL1 lies within its N terminus (mel residues 1–407 [Chen et al., 2014; Schittenhelm et al., 2010]; corresponding to 1–420 in bip), we focused on this region to create CAL1 N-terminal bip-mel chimeras (Figures 5A and 5B) and interrogated their competency for bip CENP-A recruitment at the lacO site.
Residues 1–160 of mel CAL1 are sufficient for CENP-A nucleosome assembly in vitro (Chen et al., 2014). Therefore, we created an N-terminal CAL1 (1–407) chimera where the first 160 residues of mel CAL1 were replaced by the homologous region of bip CAL1 (mel CAL1bip1–160; Figure 5A) fused to GFP LacI, and determined whether this construct was able to recruit bip CENP-A to the lacO site. After induction of mel CAL1bip1–160-GFP-LacI in lacO cells co-expressing bip HA-CENP-A, IF on metaphase spreads was performed with anti-HA, anti-GFP, and anti-mel CENP-A antibodies (Figure 5C). We found that mel CAL1bip1–160-GFP-LacI successfully recruits bip CENP-A to the lacO site (82%) while it recruits mel CENP-A inefficiently (20%; Figure 5D). These results indicate that replacing the first 160 residues of mel CAL1 with the corresponding region of bip CAL1 is sufficient to enable the incorporation of bip CENP-A into chromatin and that this region is critical for mel CENP-A recruitment. Furthermore, as we previously observed that full-length bip CAL1 can recruit mel CENP-A to the lacO site in approximately 50% of metaphase spreads (Figure 3I), the lower percentage of recruitment of mel CENP-A by the mel CAL1bip1–160 chimera observed here suggests that the full-length bip CAL1 can engage the endogenous centromere/kinetochore assembly pathway, likely via an interaction between its C terminus and mel CENP-C (Chen et al., 2014; Schittenhelm et al., 2010).
CAL1 contains an “Scm3-like” domain at its N terminus (Figure 5E; residues 1–40 [Phansalkar et al., 2012]), which is essential for ectopic CENP-A deposition (Chen et al., 2014). To further narrow down the region of CAL1 required for CENP-A incorporation, we swapped residues 1–40 and 41–160 of mel CAL1 with the corresponding region of bip CAL1 (mel CAL1bip1–40-GFP LacI and mel CAL1bip41–160-GFP-LacI; Figures 5A and 5B). These chimeras were again transiently expressed in S2 lacO cells along with bip HA-CENP-A, followed by IF on metaphase spreads to assess the presence or absence of bip HA-CENP-A at the lacO (Figure 5C).
We found that mel CAL1bip1–40-GFP-LacI successfully recruits both bip CENP-A and mel CENP-A to the lacO (88% and 61%, respectively). In contrast, mel CAL1bip41–160-GFP-LacI does not efficiently recruit bip CENP-A to the lacO site (13%), but still efficiently recruits mel CENP-A (79%; Figure 5D). These results suggest that residues 1–40 of bip CAL1 are co-evolving with bip CENP-A and that the corresponding mel CAL1 residues are responsible for the incompatibility observed between bip CENP-A and mel centromeres (Figures 1 and and5E)5E) (Vermaak et al., 2002). Furthermore, these findings reveal the conservation of CENP-A recognition mechanisms between the non-homologous CAL1 and Scm3/HJURP chaperones, both of which involve the L1 region of CENP-A (Bassett et al., 2012; Cho and Harrison, 2011).
In summary, L1 of CENP-A is evolving adaptively in Drosophila (Malik and Henikoff, 2001) and has diverged in the branch containing the ananassae subgroup (Vermaak et al., 2002). The Scm3-like region of CAL1 (Phansalkar et al., 2012), which is critical for CENP-A recruitment (Chen et al., 2014), recognizes CENP-A through its L1 and co-evolves with it, thereby maintaining its ability to deposit CENP-A in this branch of the phylogeny. The presence of a competent CAL1 assembly factor (bip CAL1 or a mel CAL1bip1–40 chimera) in mel cells is sufficient to deposit bip CENP-A into chromatin (centromeric or otherwise).
Our work sheds light on a puzzling observation in centromere biology: that a CENP-A ortholog is unable localize to the centromeres of a relatively close species (Vermaak et al., 2002). What makes this even more surprising is the report that yeast CENP-A/Cse4 can complement CENP-A knockdown in HeLa cells (Wieland et al., 2004) despite billions of years since these two species last shared a common ancestor. Using coIPs and an ectopic tethering system, we show that mel CAL1 can form a complex with bip CENP-A, but this complex is not competent for bip CENP-A deposition. Centromeric targeting of bip CENP-A can be restored upon co-expression of a functional CAL1 partner in both mel and sim cells. Using CENP-A and CAL1 chimeras we demonstrate that for successful CENP-A deposition into chromatin to occur residues 1–40 of CAL1 and CENP-A L1 must be compatible, suggesting that these regions mediate CAL1/CENP-A function.
Given that Drosophila CENP-A L1 is under positive selection (Malik and Henikoff, 2001), one might predict that its binding partner, CAL1, is also adaptively evolving to maintain centro-mere integrity throughout evolution. While we found no evidence of positive selection on CAL1 using standard methods (Phansalkar et al., 2012), the lineage-specific CENP-A/CAL1 compatibility demonstrates that the “Scm3-like” domain of CAL1 is undergoing coordinated protein evolution with CENP-A L1.
Secondary functions of CAL1 may be suppressing its rate of evolution. For example, CAL1 also interacts with the highly conserved FACT complex (Chen et al., 2015) and localizes to the nucleolus (Chen et al., 2012; Lidsky et al., 2013). We hypothesize that the overall CAL1 sequence is under purifying selection (Phansalkar et al., 2012) to preserve its functional interactions with highly conserved partners, while key residues within the N terminus of CAL1 evolve to maintain the functional interaction with CENP-A.
Our experiments focused of the role of L1 in centromere evolution. However, the CENP-A N terminus is also adaptively evolving (Malik et al., 2002). Since our experiments used the full-length bip CENP-A gene, they demonstrate that the divergent N-terminal tail of bip CENP-A does not hinder the ability of bip CENP-A to bind to mel centromeres when bip CAL1 is present, at least in mitosis, challenging the proposal that the N terminus also evolves in conflict with centromeric DNA. However, we cannot exclude the possibility that the adaptive evolution of the CENP-A N terminus may be a contributing factor in modulating the DNA-binding preferences of CENP-A exclusively during meiosis, as the N terminus of CENP-A has been shown to have meiosis-specific functions in Arabidopsis (Lermontova et al., 2006; Ravi and Chan, 2010).
Since we have not directly assayed the CENP-A-associated DNA sequences of any of these Drosophila species, we are unable to completely rule out the divergence of centromeric DNA as a contributing factor in the adaptive evolution of CENP-A L1. Nonetheless, bip CENP-A can localize to both mel and sim centromeres, suggesting that the presence of a functionally compatible CENP-A chaperone is what determines the ability of CENP-A orthologs to be incorporated at the centromeres of both species. Even the more divergent wil CENP-A can localize to mel centromeres in the presence of its CAL1 partner. It is possible that mel, sim, bip, and wil all share the same centromeric sequences. However, such divergent species (spanning 40 million years of evolution), having experienced no changes in centromeric DNA sequences, would go against the fundamental assumption of centromere drive that centromeric satellites are rapidly evolving. Collectively, our data are inconsistent with positive selection of CENP-A L1 affecting its DNA-binding preferences throughout evolution (Malik and Henikoff, 2001; Vermaak et al., 2002).
The question of why CENP-A is rapidly evolving in Drosophila still remains, and experimental evidence that CENP-A evolution is a direct result of conflict with centromeric DNA is lacking. CAL1 is unlikely to drive this rapid evolution, since it is evolving more slowly than CENP-A (Phansalkar et al., 2012). We propose that, in Drosophila, positive selection of CENP-A L1 modulates the efficiency of its centromeric deposition by CAL1 rather than its DNA-binding specificity, as originally proposed (Vermaak et al., 2002). Our analysis of the extreme example of the incompatible bip CENP-A and mel CAL1 suggests that the degree of functional compatibility between these two proteins during intermediate evolutionary times could influence how much CENP-A is incorporated, in turn affecting CENP-C recruitment and kineto-chore assembly (Chen et al., 2014; Erhardt et al., 2008). Thus, the ability to “tune” how much CENP-A is deposited at the centromere via changes in L1 could be a mechanism to curb the increased “kinetochore strength” resulting from centromere satellite expansion during centromere drive (Figure 6), akin to the long-standing model proposed by Henikoff and Malik (Henikoff and Malik, 2002; Malik and Henikoff, 2002). Although our work focuses on the critical role of these co-evolving domains in mitosis, it is important to note that CAL1 is also essential for CENP-A deposition during meiosis (Dunleavy et al., 2012). Therefore it is conceivable that our proposed model would apply to meiosis, the natural battleground of centromere drive.
Flies and genomic DNA were obtained from the University of California San Diego Drosophila Species Stock Center or from other laboratories (see Table S1). All non-melanogaster CENP-A and CAL1 orthologs were PCR amplified using Phusion High-Fidelity DNA polymerase (New England Biolabs) from genomic DNA using the primers listed in Table S2. See Supplemental Experimental Procedures for details on cloning.
Drosophila melanogaster Schneider 2 (S2) cells were grown as described previously (Chen et al., 2014; Mellone et al., 2011). S2 cells containing stably integrated LacO arrays (pAFS5 [Straight et al., 1996]) were generated as described previously (Chen et al., 2014; Mendiburo et al., 2011). Drosophila simulans ML82-19a (M-19) cells were purchased from the Drosophila Genomics Resource Center. M-19 cells were grown in Schneider's media with 10% fetal bovine serum at 25°C.
Transient and stable transfections in S2 cells were performed using FuGENE HD Transfection Reagent (Promega) as previously described (Chen et al., 2014). For transient transfection in M-19 cells, 2 × 106 cells were plated in six-well plates and transfected with Cellfectin reagent (Invitrogen) and plasmid DNA. Cells were incubated with the transfection complex in serum-free medium for 3 hr before replacing medium with serum-containing medium. Cells were incubated for 3 days before harvesting for IF.
IF on settled interphase cells and metaphase spreads were performed as previously described (Chen et al., 2014). Primary antibodies: anti-CENP-A (chicken, 1:1,500; Blower and Karpen, 2001) or anti-CID (rabbit, 1:500; Abcam), anti-CENP-C (guinea pig, 1:500; Erhardt et al., 2008), anti-Ndc80 (chicken, 1:200; Cane et al., 2013), anti-GFP Alexa 488-conjugated (rabbit, 1:100; Invitrogen), or anti-GFP (chicken, 1:500; Abcam), and anti-HA (mouse, 1:500; Covance).
For salt extractions, settled cells were incubated with PBS-D (0.1% digitonin) with or without 0.5 M NaCl for 30 min (Perpelescu et al., 2009) before 37% formaldehyde was added to the solution to a final concentration of 3.7% followed by 10 min of incubation before proceeding with IF.
Images were acquired on a wide-field fluorescence microscope (PersonalDV; GE Healthcare) equipped with a 60Å~/1.42 NA or a 100 Å~/1.40 NA oil-immersion objective (Olympus) and a CoolSnap HQ2 camera (Photometrics), keeping exposure conditions constant between all samples. Images were acquired and processed in softWoRx (Applied Precision), maintaining the scaling constant between samples, and saved as PSD files. Figures were assembled in Adobe Illustrator. For quantification, see Supplemental Experimental Procedures.
Whole-cell lysates and western blots were prepared as previously described (Chen et al., 2014). Membranes were incubated with either anti-GFP (Goat, 1:150; Rockland), anti-CAL1 (rabbit, 1:000; gift from Aaron Straight), anti-HA (mouse, 1:500; Covance), anti-tubulin (mouse, 1:500; Sigma-Aldrich), anti-fibrillarin (mouse, 1:1,000; Cytoskeleton), or anti-lamin (mouse, 1:1,000; Hybridoma Bank, University of Iowa) primary antibodies. Blots were imaged on an Odyssey Fc (LI-COR Biosciences) using chemiluminescent substrate for detection of horseradish peroxidase-labeled secondary antibodies, or were developed on X-ray films.
IPs were performed from nuclear extracts as previously described (Chen et al., 2012), using 5 μg of anti-CAL1 antibody or 5 μg of anti-immunoglobulin G antibody. For normalization, whole-cell lysates were prepared from 1 × 106 cells expressing either mel or bip GFP-CENP-A, and total GFP protein levels were quantified by western blotting using Image Studio software (LI-COR) and normalized compared with a loading control (lamin). Nuclear extracts were performed from the same cells and were diluted in resuspension buffer (0.29 M sucrose, 0.5 mM Tris-HCl [pH 7.4], 1.5 mM NaCl, 5 mM MgCl2, 1 mM EGTA, 0.04% Triton X-100, 1× EDTA-free protease inhibitors, and 1 mM DTT) so that the levels of GFP-CENP-A in all samples were equal. 150 μl of diluted bip or undiluted mel nuclear extract were loaded onto antibody-conjugated beads for IP. See Supplemental Experimental Procedures for quantification of IPs.
We thank Gary Karpen, Aaron Straight, Patrick Heun, and Tom Maresca for re-agents; Rachel O'Neill and Eric Joyce for critically reading the manuscript; Jason Palladino and Chin-Chi Chen for discussions and suggestions; Harmit Malik for fly genomic DNA; Doris Bachtrog, Rich Meisel, and Andy Clark for flies; Patrick Lenehan, Chin-Chi Chen, and Ankita Chavan for technical help; and Fly-base, the UCSD Stock Center, and the DGRC for fly stocks and other Drosophila resources. L.R. was supported by NSF award MCB1330667; B.G.M. was supported by NSF award MCB1330667 and NIH award GM108829.
B.G.M. and L.R. conceived the project; L.R. conducted experiments; B.G.M. and L.R. wrote the manuscript.
Supplemental Information includes Supplemental Experimental Procedures, six figures, and two tables and can be found with this article online at http://dx.doi.org/10.1016/j.devcel.2016.03.021.