Drosophila CENH3 Is Sufficient for Centromere Formation

See allHide authors and affiliations

Science  04 Nov 2011:
Vol. 334, Issue 6056, pp. 686-690
DOI: 10.1126/science.1206880


CENH3 is a centromere-specific histone H3 variant essential for kinetochore assembly. Despite its central role in centromere function, there has been no conclusive evidence supporting CENH3 as sufficient to determine centromere identity. To address this question, we artificially targeted Drosophila CENH3 (CENP-A/CID) as a CID-GFP-LacI fusion protein to stably integrated lac operator (lacO) arrays. This ectopic CID focus assembles a functional kinetochore and directs incorporation of CID molecules without the LacI-anchor, providing evidence for the self-propagation of the epigenetic mark. CID-GFP-LacI–bound extrachromosomal lacO plasmids can assemble kinetochore proteins and bind microtubules, resulting in their stable transmission for several cell generations even after eliminating CID-GFP-LacI. We conclude that CID is both necessary and sufficient to serve as an epigenetic centromere mark and nucleate heritable centromere function.

The centromere ensures correct segregation of chromosomes during mitosis by providing the site for kinetochore assembly and microtubule attachment. Most eukaryotic organisms contain only one centromere per chromosome, which is specifically positioned and faithfully propagated through each cell cycle. With the exception of Saccharomyces cerevisiae, DNA sequence is considered neither necessary nor sufficient to mark centromeres in most eukaryotes, suggesting that centromere identity in many organisms is determined epigenetically (1). CENH3 (CENP-A in mammals, CID in Drosophila) is a centromere-specific histone H3 variant that replaces canonical histone H3 in centromeric nucleosomes. It is essential for centromere function and kinetochore assembly (25) and thus a prime candidate epigenetic mark for determining centromere identity.

Global misincorporation of CID into chromosome arms leads to the formation of functional ectopic kinetochores only in a small subset of sites (6, 7), hindering a direct correlation between CID presence and kinetochore formation. To determine whether CID is sufficient for directing kinetochore assembly and nucleate centromere identity, we targeted a CID-GFP-Lac Repressor (LacI) fusion protein to an array of lac operator (lacO) sequences stably integrated in Drosophila Schneider S2 cells (Fig. 1A and fig. S1B). Inducible CID-GFP-LacI or GFP-LacI fusion proteins are efficiently targeted to lacO sequences only upon pulse induction (Fig. 1B), whereas low leaky expression of CID-GFP-LacI in uninduced cells correlates with an exclusive centromere localization (Fig. 1B and fig. S1A).

Fig. 1

CID-GFP-LacI is efficiently targeted to lacO sequences and induces kinetochore assembly. (A) Schematic representation of the experimental setup. (B) Examples of the lacO-containing chromosome 3 from fixed mitotic CID-GFP-LacI and GFP-LacI cells before and after induction with CuSO4 are shown. (C) Fixed mitotic chromosomes of CID-GFP-LacI cells were processed 1 day after pulse induction for immunofluorescence (IF) with antibodies to GFP and CENP-C or NDC80. The lacO-containing chromosome is indicated by an asterisk and shown magnified in the right panels. (D) Same as on (B), with GFP-LacI cells 1 day after the pulse induction. (E) Quantification of the percentage of cells showing colocalization between GFP and different kinetochore proteins in the lacO for the CID-GFP-LacI and GFP-LacI cells similar to (A), (B), and fig. S2, A and B. Values are average ± SD of at least three independent experiments. CENP-C, nCID-GFP-LacI = 33, nGFP-LacI = 43. MAD2, nCID-GFP-LacI = 38, nGFP-LacI = 37. POLO, nCID-GFP-LacI = 36, nGFP-LacI = 56. NDC80, nCID-GFP-LacI = 33, nGFP-LacI = 52. *, P < 0.05; **, P < 0.005; ***, P < 0.0005 (two-sided Fisher’s exact test). White arrows indicate endogenous centromeres and green arrows the lacO integration in all panels. Scale bars, 3 μm.

CENH3 has been shown to adopt a specialized nucleosomal structure, which is proposed to mark centromeric chromatin (814). To determine whether CID-GFP-LacI is incorporated in nucleosomes at the lacO, mononucleosomes from GFP-LacI and CID-GFP-LacI cells were separated on a sucrose gradient. Analysis of the fractions from both cell lines revealed that lacO DNA comigrates with fractions containing H3-nucleosomes, suggesting that they are chromatinized (fig. S2, A and B). GFP-LacI did not comigrate with nucleosomes, likely due to its release from lacO binding after micrococcal nuclease treatment. In contrast, CID-GFP-LacI was found in nucleosome-containing fractions, indicating that it can interact with chromatin independently of the LacI binding. Nucleosome incorporation was confirmed by coimmunoprecipitation of CID-GFP-LacI with histone H2A in the nucleosome fraction (fig. S2C). To exclude that CID-GFP-LacI migration pattern is only due to the contribution of protein localized to endogenous centromeres, we created a mutant CID fused to GFP-LacI that does not target to centromeres and is therefore present only at the lacO array (fig. S2E). In this mutated protein, three amino acids [G175S, L177V, and L178M (CIDsvm)] in the CENP-A targeting domain (CATD) of CID are replaced by the corresponding residues of histone H3.1, as shown previously for human CENP-A (15)(fig. S2D). The migration pattern of CIDsvm-GFP-LacI in sucrose gradients resembles that of CID-GFP-LacI (fig. S2A), supporting the notion that CID-GFP-LacI is in nucleosomes also at lacO regions.

Targeting of CID-GFP-LacI, but not GFP-LacI, to the lacO recruited kinetochore proteins, such as CENP-C and the microtubule-binding protein NDC80/HEC1, 1 day after pulse induction (Fig. 1, C to E) (13, 14, 16). Probing for the spindle assembly checkpoint proteins POLO kinase and MAD2 revealed the same pattern, suggesting that ectopically targeted CID is sufficient to direct kinetochore assembly (Fig. 1E and fig. S1, C and D).

The formation of a functional ectopic kinetochore at the lacO is expected to generate a dicentric chromosome causing chromosome breakage and anaphase defects (12). Indeed, we observe a higher number of aberrant mitotic chromosome configurations in cells expressing CID-GFP-LacI compared with control GFP-lacI cells and containing the lacO array (Fig. 2).

Fig. 2

The lacO-containing chromosome is involved in mitotic defects and fragmentation. (A) IF fluorescence in situ hybridization (IF-FISH) on fixed mitotic chromosomes of CID-GFP-LacI cells 1 day after induction reveals a GFP-positive lacO-containing fragment (indicated by an asterisk and magnified in the insets). (B) Example of a mitotic chromosome breaking within the lacO sequence upon CID-GFP-LacI targeting (green arrow). The white arrow indicates the endogenous centromere. (C) Lagging chromosome containing CID-GFP-LacI colocalizing with lacO sequences (white arrow). Grayscale images of individual channels are shown below. Scale bars, 3 μm. (D) Frequency of anaphase figures with defects involving lacO sequences as shown in (C). Horizontal bars are average of two independent experiments. nCID-GFP-LacI = 48, nGFP-LacI = 109. (E) Fragmented lacO-containing chromosomes in CID-GFP-LacI and GFP-LacI cells as shown in (A) and (B). Values are average ± SD of at least two independent experiments. nCID-GFP-LacI = 42, nGFP-LacI = 47. *, P < 0.05; **, P < 0.005 (two-sided Fisher’s exact test).

The epigenetic model for centromere identity predicts that a functional centromere self-directs the loading of new centromeric marks after each cell division (1). If CID-GFP-LacI identifies the lacO as a centromere, CID molecules without a LacI fusion should also be recruited to lacO regions. To investigate this, we created a stable cell line containing both an inducible CID-GFP-LacI and a constitutively expressed CID-hemagglutinin (HA) HA construct. Upon pulse induction and targeting of CID-GFP-LacI, we found a low initial CID-HA recruitment to the lacO (9.7%) (Fig. 3A), increasing by two-fold 7 days later (24%) (Fig. 3B). Higher-resolution analysis using stretched chromatin fibers revealed that at initial time points CID-HA localization is restricted to the lacO region (fig. S3A), whereas 7 days after pulse induction, CID-HA (Fig. 3C) or CID (fig. S3B) is also found spreading into adjacent regions (Fig. 3D). CID-HA or CID were never found in lacO sequences in the presence of GFP-LacI (fig. S3C).

Fig. 3

CID tethered to lacO sites creates a self-propagating epigenetic mark. CID-GFP-LacI expression was induced in the presence of constitutively low levels of CID-HA. Representative images of mitotic chromosomes 1 day (A) and 1 week (B) after pulse induction are shown. White arrows indicate endogenous centromeres, green arrows the lacO site. The percentages of cells where CID-GFP-LacI sites colocalize with CID-HA are shown. n1day = 143, n7days = 29. (C) IF-FISH on a stretched chromatin fiber 1 week after pulse induction. White arrows indicate spreading CID-HA. Dashed white lines delimit the lacO sequence. Scale bars, 3 μm. (D) The distance between the most distal HA signal relative to the lacO was measured and normalized to the size of the lacO FISH signal to correct for differential stretching. n1day = 11, n7days = 19. *, P < 0.05 (two-tailed Student’s t test).

CID-GFP-LacI targeting to an ectopic chromosomal locus results in mitotically unstable dicentric chromosomes, hindering the analysis of long-term inheritance of the centromere function. To directly test de novo centromere heritability, we analyzed whether CID-GFP-LacI targeting confers mitotic stability to a centromere-devoid episomal DNA element carrying lacO sequences and a G418-resistance cassette. This plasmid was transfected into S2 cell lines constitutively expressing low levels of either CID-GFP-LacI or GFP-LacI and kept under selection pressure for 28 days (Fig. 4A) (17, 18). The initial transfection efficiency was comparable in both cell lines (Fig. 4B, –DpnI). However, we observed a decline in the amount of replicated plasmids (+DpnI) in GFP-LacI–expressing cells, whereas CID-GFP-LacI cells maintained the plasmid until the end of the experiment (Fig. 4B, +DpnI). Efficient replication and proper segregation are the two major factors defining the stability of plasmids (18). For the first 7 days, both cell lines replicated the plasmid with similar efficiency (Fig. 4B and fig. S5, F and G), indicating that correct segregation is the main factor contributing to its maintenance in CID-GFP-LacI cells. This was also observed when selection was removed after 16 days or never applied (fig. S5E). Furthermore, we find CENP-C, NDC80, and microtubule connections localizing to CID-GFP-LacI–bound lacO plasmids in mitosis (Fig. 4, C and D, and fig. S4, A and C). These plasmids separate earlier than endogenous chromosomes and generally display a symmetric distribution on the spindle poles (fig. S4, B and C). GFP-LacI cells grew poorly after transfection, and the few mitotic cells detected showed no colocalization of GFP-positive lacO plasmids with either CENP-C or NDC80 (fig. S4D). Further characterization of recovered plasmids showed that plasmid size remained stable except for variability in the repetitive lacO array and did not acquire endogenous Drosophila sequences (fig. S5, A to C). A minor fraction of concatenated plasmids did not correlate with efficient recruitment of kinetochore proteins (fig. S5D).

Fig. 4

CID-GFP-LacI confers stability to episomal lacO plasmids and allows epigenetic inheritance of centric chromatin after its elimination. (A) Schematic representation of the plasmid rescue assay. (B) The average of three independent plasmid rescue experiments is shown ± SEM. tc, transfection control; utc, untransfected control. K = ×1000. (C) Immunostainings of CID-GFP-LacI cells in metaphase 9 days after transfection with a lacO plasmid. (D) Similar to (C), showing microtubule-plasmid attachments. (E) Cells described in (A) were cotransfected with a CID-HA expression construct and the lacO plasmid and subjected to αHA and αGFP–chromatin immunoprecipitation experiments 6 days later. (F) Quantification of the percentage of extrachromosomal CID/CENP-C foci in mitotic cells after transient cotransfection of lacO and CID-GFP-LacI constructs. Black: GFP-negative CID/CENP-C foci coming from GFP-negative cells. Green: Foci from cells that contain at least one GFP-positive focus (GFP-positive cells). Scale bars, 3 μm.

The centromeric mark can self-propagate in the lacO plasmid, as shown by the recruitment of CID-HA to targeted CID-GFP-LacI plasmids and the spreading of both into backbone sequences (Fig. 4E and figs. S5C and S6A). To determine whether centromeric chromatin is inherited even after removal of CID-GFP-LacI targeting, we transiently cotransfected the lacO plasmid and a CID-GFP-LacI expression construct to deliver an initial expression pulse that is progressively lost. Episomal lacO plasmids were detected as nonchromosomal CID/CENP-C foci, with or without a GFP signal. Early after transfection, we found cells that contained both GFP-positive and -negative foci (green bars in Fig. 4F) and a smaller cell population that had already lost CID-GFP-LacI expression but contained nonchromosomal CID/CENP-C foci (black bars in Fig. 4F). The percentage of these GFP-negative plasmids increased until day 27, when all detected CID/CENP-C foci were devoid of any GFP signal (Fig. 4F and fig. S6, B to E). We conclude that an initial targeting of CID-GFP-LacI to the lacO plasmid enables the nucleation of centromere function, which is maintained independently of CID-GFP-LacI expression and targeting.

In recent years, functional biosynthetic kinetochores have been constructed by targeting kinetochore components (1921) and chromatin modifiers (22, 23) to minichromosomes. Even though the possibility of bypassing CENH3 chromatin for the assembly of a kinetochore suggested that CENH3 is dispensable for this function (21), no evidence of centromere heritability was provided by these studies. Recent reports showing CENP-A–dependent kinetochore formation in vitro (24) or by artificially targeting the CENP-A loading factor in humans (HJURP) (25) further stress the importance of CENH3 in kinetochore assembly. The results presented here show that CENH3 behaves as a true centromeric epigenetic mark not only by being sufficient for the recruitment of kinetochore proteins during mitosis but also for directing its own incorporation and maintaining centromere function through each cell division.

Supporting Online Material

Materials and Methods

Figs. S1 to S6

References (26, 27)

References and Notes

  1. Acknowledgments: We thank G. H. Karpen, A. Straight, T. Maresca, E. Salmon, C. Sunkel, S. Heidmann, and C. Lehner for reagents. We also thank L. Jansen for critical reading of the manuscript and the members of the Grosschedl Department, particularly S. Chlamydas for pioneering experiments and A. Olszak, H.-J. Schwarz, D. Venegas, S. Daujat, S. Saccani, and D. van Essen for reagents and helpful discussions. A.S. is supported by the Sonderforschungsbereich (SFB) 646, SFB/TR05, and Schwerpunktprogramm 1230 of the Deutsche Forschungsgemeinschaft. P.H. is funded by the Max Planck Society, the Human Frontier Science Program (Career Development Award; CDA), and the Excellence Initiative of the German Federal and State Governments (EXC 294). M.-J.M. and J.P. are grateful for an International Max Planck Research School (IMPRS) fellowship provided by the Max Planck Institute of Immunobiology and Epigenetics.
View Abstract

Stay Connected to Science

Navigate This Article