INCENP Centromere and Spindle Targeting: Identification of Essential Conserved Motifs and Involvement of Heterochromatin Protein HP1

Unless otherwise stated, all chemicals were obtained from Sigma (Dorset, UK) and all restriction and modifying enzymes used for DNA cloning were obtained from New England Biolabs (Beverly, MA). TEN buffer is 10 mM Tris-HCl, pH 7.7, 2 mM EDTA, 50 mM NaCl; KB buffer is 10 mM Tris-HCl, pH 7.7, 150 mM NaCl, 0.1% BSA; TE buffer is 10 mM Tris-HCl, pH 8.0, 1 mM EDTA; RSB buffer is 10 mM Tris-HCl, pH 7.4, 10 mM NaCl, 5 mM MgCl₂; and glutathione S-transferase (GST) binding buffer is TEN buffer containing 0.05% Nonidet P-40.

Aequoria victoria GFP (clone TU 65, provided by M. Chalfie, Columbia University, New York, NY; Chalfie et al., 1994) was cloned into the Kpn and EcoRI sites within the multiple cloning site of the pECE vector (Ellis et al., 1986) downstream of INCENP_1-839 cDNA, previously cloned into the BglII site (Mackay et al., 1993). This generated an intermediate construct, INCENP_1-839:GFPi, which contained stop codons at the 3′ ends of both the INCENP and GFP cDNAs. INCENP_1-839:GFPi was linearized using an AflII site 1212 bp from the start codon in INCENP, blunt-ended with T4 polymerase, then digested with BglII in the pECE multiple cloning site and blunt-ended using Mung bean nuclease. The vector was then ligated to create INCENP_1-405:GFP. All GFP chimeras used in these studies contain the S65T point mutation (Heim et al., 1995). Carboxyl-terminal deletions of INCENP were then made by digesting INCENP_1-405:GFP with HindIII and KpnI and subsequently with ExoIII nuclease (Promega, Madison, WI) digestion (Henikoff, 1984). In frame INCENP:GFP fusions were sequenced, CsCl purified (Ausubel et al., 1991), and electroporated into either HeLa or LLCPK cells as described previously (Mackay et al., 1993).

Cells were fixed 16 h after transfection with 4% paraformaldehyde using standard conditions (Mackay et al., 1993). Microtubules were labeled with the mouse mAb Tu27B (Caceres et al., 1983), a gift from Lester Binder (Northwestern University, Chicago, IL). DNA was visualized by staining with 4,6-diamidino-2-phenylindole (DAPI, 0.5 mg/ml; Calbiochem, Nottingham, UK). Coverslips were mounted in Vectashield (Vector Laboratories, Inc., Burlingame, CA). Images were acquired on a Zeiss Axioplan II epifluorescence microscope equipped with a Princeton Instruments Micromax cooled CCD camera, driven by IP Lab Spectrum software.

We used site-directed mutagenesis to generate a unique restriction site at a position 141 bp (47 aa) from the start of the coding region in INCENP_1-405:GFP. This site lies between two highly conserved sequences (aa 32–44 and 52–62). Together with flanking restriction sites, this created a cloning cassette that allowed us to delete each sequence independently or to randomize the amino acids using a mutagenic oligonucleotide in which the amino acid composition of this region was retained, but the amino acids were organized in random order (see Fig. 5 B). Each sequence motif was manipulated separately.

Chromosome spreads were made from transfected cells as described previously (Earnshaw and Migeon, 1985; Mackay et al., 1998). In brief, cells were plated onto poly-lysine–coated coverslips in individual 35-mm dishes and 16 h after transfection were incubated with 6 nM vinblastine for an additional 2 h. Cells were incubated with 3 ml of RSB for 10 min at room temperature. RSB was removed by aspiration and replaced with 500 μl of fresh RSB. Dishes were placed on foam cushions in the bottom of the bucket holders of a Heraeus Megafuge and centrifuged at 4,000 rpm for 10 min at room temperature. Cells were then fixed with 4% paraformaldehyde, permeabilized with 0.1% Triton X-100, and immunostained for centromeres using human autoimmune serum GS, which recognizes CENPs A, B, and C (Earnshaw and Rothfield, 1985) at a dilution of 1:5,000. After immunostaining, DNA was visualized with DAPI. Three-dimensional data sets of selected mitotic cells were collected using a DeltaVision microscope (Applied Precision, Issaquah, WA), based on an Olympus IX-70 inverted microscope with a Photometrics PXL cooled CCD camera. Data sets were deconvolved, projected onto a single plane, and the contrast was adjusted before export as TIFF files to Adobe Photoshop.

We used vectors and yeast strains described by Brent and co-workers (Gyuris et al., 1993). Full-length INCENP and truncations thereof were cloned from pBluescript in frame into the pEG202 bait vector, tested for activation in the absence of an interactor plasmid, and further tested in the presence of HP1^Hsα in pJG4-5 (interactor plasmid) and pSH18-34 (reporter plasmid). pRFHM 1 (encoding a bicoid protein) and pSH17-4 were used as negative and positive activation controls, respectively, as described (Gyuris et al., 1993). An initial screen of 500 recombinant clones from a HeLa cDNA library (Gyuris et al., 1993) led to the isolation of one clone of HP1^Hsα and two clones of HP1^Hsγ. Two independent yeast colonies from each construct listed in Fig. 1 were tested for galactose-dependent growth in the absence of leucine, blue/white selection, and quantitation in a β-galactosidase liquid assay. For blue/white selection, yeast colonies were plated onto Gal-U-H-W-L plates, overlaid with top agar containing NaPO₄, pH 7.0, 2% N-lauroylsarcosine sodium salt, 0.04% β-mercaptoethanol, 0.02% X-gal, incubated at 30°C, and checked for blue colonies. β-galactosidase activity of yeast transformants grown in liquid culture was determined using o-nitrophenyl β-d-galactopyranoside (ONPG) as a substrate as described (Reynolds and Lubland, 1989). All plasmids used for the two-hybrid screen were obtained from the R. Brent laboratory (Harvard Medical School, Boston, MA).

PCR was used to generate GST:HP1^Hsα_1-68, GST:HP1^Hsα_68-111, and GST: HP1^Hsα_110-190 by designing sequence specific primers with unique restriction sites at the 5′ end. These products were subcloned in-frame into the pGex vector (Pharmacia, Herts, UK) and fully sequenced. GST:HP1^Hsα _1-110 and GST:HP1^Hsα_68-190 were both subcloned using standard methods. GST:HP1^Hsα_1-190 was digested with BamHI and EcoRV to release the GST:HP1^Hsα_1-110 insert and this fragment was subcloned into an empty pGex 2T vector previously digested with EcoRI, blunt-ended using Mung bean nuclease, and then digested with BamHI. GST:HP1^Hsα_68-190 was generated by ApoI digestion of GST:HP1^Hsα_1-190 to release the insert and the fragment was subcloned into an EcoRI-digested and phosphatased pGex 1 vector. All cloning junctions were sequenced to confirm that open reading frames were maintained.

Full-length INCENP_1-839 was subcloned into pBluescript (Stratagene, Cambridge, UK) using standard methods and then truncated by restriction enzyme digestion to create INCENP_1-405, INCENP_43-839, and INCENP_43-405 as described (Mackay et al., 1993; Mackay et al., 1998). cDNAs were then transcribed and translated using the TNT T3 system (Promega) in the presence of [³⁵S]methionine (ICN, Thame, UK) for 1 h at 30°C. Labeled proteins were diluted to 500 μl in GST binding buffer and concentrated by centrifugation using a 10,000 molecular weight cutoff filter (Millipore Corp., Bedford, MA) for 15 min at 4°C (Heraeus, Brentwood, UK) followed by two additional washes with binding buffer.

HP1^Hsα cDNAs (Saunders et al., 1993) were subcloned (see above), transformed into Escherichia coli CAG456 cells, and expression induced with 0.5 mM IPTG for 5 h, 30°C. GST:HP1^Hsα chimeras were purified by repeated rounds of fast freeze and thaw, followed by the addition of 4 mg/ml egg white lysozyme, sonication, and centrifugation at 4,000 rpm, 10 min at 4°C in a Heraeus Megafuge. Translated INCENP protein (20 μl) was added to 150 μl GST binding buffer, incubated with glutathione agarose conjugated to 3–5 mg of GST alone or GST:HP1^Hsα chimeras, and incubated on a rotating mixer for 1 h at 4°C. After incubation, samples were washed three times with GST binding buffer. Bound fractions were solubilized by boiling in 1× sample buffer for 3 min and analyzed on a 12% SDS-polyacrylamide gel followed by Coomassie blue staining and autoradiography. Synthesis was confirmed before each experiment.

INCENP_1-405:GFP and INCENP_1-68:GFP were created in the pECE vector as described above and subcloned into pBluescript (Stratagene) for in vitro translation. These proteins were tested as above with full-length HP1^Hsα. GFP and an empty vector were used as controls.

Primer T3 and sequence specific PCR primers (Oswel, Southampton, UK) with a 3′ SacII site were used to generate the full-length HP1 (5′GCAGGAATTCTCCGCGGGGCTCTTTGCTGT3′) or the chromo domain alone (5′GCAGGAATTCTCCGCGGGCATAAATTCAGA3′). The PCR products and the CENP-B_1-158:INCENP_45-839 vector (Eckley et al., 1997) were digested with BglII and SacII and the fragments were subcloned using standard methods. The resultant constructs were purified by CsCl density gradients (Ausubel et al., 1991) and transfected into HeLa cells as described previously (Mackay et al., 1993) onto duplicate coverslips. Cells were fixed 16 h after transfection, for 5 min at −20°C in methanol. Immunofluorescence was performed as described above, using a chicken-specific anti-INCENP antibody (RabC) (Mackay et al., 1993) at a dilution of 1:500. Both full-length INCENP and INCENP_43-839 transfectants were used as controls.

HeLa cells were transfected with constructs encoding INCENP_1-839, INCENP_43-839, or HP1^Hsα_1-68:INCENP_45-839 as described and plated onto one 100-mm petri dish containing a 16-mm coverslip. The coverslip was removed 16 h after transfection, the cells were fixed for 5 min at −20°C in methanol, and stained for INCENP (RabC) and DNA (DAPI) as described above. The transfection efficiency was calculated by dividing the number of transfectants by the total number of cells stained with DAPI. Five fields were counted per sample for a total of ∼100 cells. The cells remaining in the petri dish were trypsinized, washed twice with PBS, and counted in a hemacytometer. Cell pellets were resuspended in 1× sample buffer, boiled for 3 min and sonicated. Equivalent cell numbers were loaded per lane. Blots were probed using an INCENP antibody (RabC) and visualized using ECL reagents (Amersham, Buckinghamshire, UK).

INCENP_1-839 in the pECE vector was digested with Eagl and AflII, blunt-ended using Mung bean nuclease, and the vector was religated to generate INCENP_1-839^(Δ69-404). After transfection, immunofluorescence was performed using the INCENP antibody (RabC) as described above. In parallel, INCENP_1-839^(Δ69-404) was subcloned into pBluescript, transcribed and translated (Promega), and tested for in vitro binding to GST or GST: HP1^Hsα. The samples were resolved by SDS-PAGE and visualized by autoradiography. INCENP_1-839 was used as a control.

During prophase and early prometaphase, INCENP is distributed along the chromosome arms; however, the protein becomes concentrated at centromeres by metaphase (Earnshaw and Cooke, 1991). Other centromere proteins such as CENPs A–C and E all bind directly to the centromere with no prior binding to the chromosome arms (Earnshaw and Rothfield, 1985). Therefore, we refer to this distinctive pattern of INCENP movements as indirect centromere targeting.

Previous protein truncation studies demonstrated that indirect centromere targeting by INCENP requires the amino-terminal 42-aa residues of the protein (Mackay et al., 1998). To delineate the minimal region of the protein required to carry out this novel movement, we generated a series of progressively shorter amino-terminal INCENP fragments, each containing GFP fused at its carboxyl terminus. The localization of these proteins on metaphase chromosomes was confirmed in both cycling and metaphase-arrested cells. We began by confirming that an INCENP:GFP chimeric protein underwent indirect centromere targeting (Fig. 2). INCENP_1-405:GFP was distributed all along the chromosomes at prophase, becoming concentrated at centromeres during metaphase. This protein could transfer to the spindle at anaphase, albeit less efficiently than wild-type INCENP (Fig. 2 D; see also Earnshaw and Cooke, 1991; Mackay et al., 1998). Observation of living cells (data not shown) revealed that the targeting of INCENP_1-405:GFP to centromeres is not due to loss of the noncentromeric INCENP from chromosomes, but instead represents an accumulation of the protein at centromeres.

INCENP_1-405:GFP (Fig. 3,A′) and the progressively smaller truncated molecules, INCENP_1-308:GFP and INCENP_1-226: GFP, targeted to centromeres in mitosis (Fig. 3, B′ and C′). Ultimately, we found that a polypeptide containing only the amino-terminal 68 aa of INCENP was sufficient to target GFP to centromeres (Fig. 3 D′). In control experiments, transfected GFP that was not fused to other proteins localized diffusely in the cytoplasm during interphase and was never found associated with the chromatin (Rizzuto et al., 1995; and data not shown). Thus, INCENP contains a relatively compact module at its amino terminus that is necessary and sufficient to target the protein to centromeres during the early stages of mitosis.

Further analysis of the distribution of the amino-terminal INCENP:GFP fusion proteins in cycling mitotic cells revealed a surprising result. Every truncated INCENP molecule capable of targeting to centromeres was also able to transfer to the anaphase spindle and ultimately to the midbody (see below). Thus, the minimal centromere-targeting module also functions as a midbody-targeting module.

The observation that chicken INCENP targets normally to centromeres in mammalian cells (Mackay et al., 1998) suggests that the mechanism of centromere targeting by INCENP is highly conserved. Therefore, we looked for regions of amino acid conservation within the INCENP minimal centromere-targeting module. Comparison of the first 68 aa of the chicken, Xenopus, and mouse INCENP proteins revealed two highly conserved regions corresponding to aa 32–44 (69% identity) and 52–62 (100% identity) of the chicken INCENP polypeptide (Fig. 4 A).

Two approaches were taken to examine the involvement of the conserved region spanning aa 32–44 in indirect centromere targeting by INCENP. First, we deleted this region of the protein. The mutated protein, INCENP _1-405^(Δ27-47):GFP, bound to chromosome arms, but failed to target to centromeres (data not shown). Second, we used a mutagenic oligonucleotide to randomize the amino acid sequence as shown in Fig. 4,B. This also abolished centromere targeting: INCENP_1-405^(32-44r):GFP “painted” the chromosome arms (Fig. 3 F). Therefore, the conserved motif encompassing aa 32–44 is essential for targeting INCENP to centromeres during mitosis. To determine whether aa 32–44 are sufficient to target INCENP to centromeres, we prepared a deletion construct (INCENP_27-839) encoding a protein missing aa 1–26. This protein failed to target centromeres (data not shown), behaving in all respects like INCENP_45-839 (Mackay et al., 1993). This suggests that sequences in the first 26 aa of INCENP are also essential for centromere targeting, although effects of the deletion on protein structure cannot be excluded.

We next examined the role of the highly conserved sequence (aa 52–62) in centromere targeting. Deletion of aa 47–69 (Fig. 1) had no effect on centromere targeting. In agreement with this result, randomization of the conserved 11-aa sequence as shown in Fig. 4,B also had no effect on indirect centromere targeting: INCENP_1-405^(52-62r): GFP located normally to centromeres (Fig. 3 E′). Thus, INCENP residues 52–62 are not involved in targeting the protein to centromeres.

Instead, INCENP residues 52–62 appear to be involved in targeting INCENP to the midbody. Although INCENP _1-405^(52-62r):GFP was unimpaired in its ability to target to the centromere (Fig. 3,E′), the protein was subsequently unable to target to the central spindle or midbody (Fig. 5 F). In telophase cells, INCENP_1-405^(52-62r):GFP colocalized with the chromatin as seen by the overlap of blue and green. These data reveal for the first time that INCENP movements to centromeres and to the spindle involve distinct peptide motifs and perhaps different ligands.

In other studies, INCENP molecules in which the sequence of aa 32–44 had been randomized not only failed to target to centromeres, but also remained associated with chromatin throughout mitosis and did not transfer to the anaphase spindle (Fig. 5 E). Although we cannot formally exclude that aa 32–44 encodes a separate determinant required for midbody targeting, the simplest interpretation of this result is that, as suggested by our earlier observations with amino-terminal deletion proteins INCENP_43-839, INCENP_43-405 (Mackay et al., 1998) and INCENP_27-839, INCENP must first target to centromeres if it is to transfer to the midbody later in mitosis.

Given the essential role played by the amino-terminal region of INCENP in centromere targeting and spindle transfer, we predicted that this region might make important contacts with other chromosomal proteins. To identify such interacting factors, we performed a yeast two-hybrid screen of a HeLa cell cDNA library using INCENP _1-405 fused to the LexA DNA-binding domain as bait (Gyuris et al., 1993). An initial screen led to the isolation of seven clones that grew on the appropriate media and were blue in the presence of β-galactosidase. None of the seven proteins interacted with control bait proteins, including the LexA DNA-binding domain alone, bicoid, and GFP alone (Fig. 6 A and data not shown). In addition, all seven interacted both with INCENP_1-405:GFP and with full-length INCENP_1-839. Subsequent DNA sequence analysis identified three classes of putative INCENP-interacting proteins: known heterochromatin proteins (three clones), β-tubulin (one clone), and novel polypeptides (three clones). The remainder of this report is concerned with our analysis of the interaction of INCENP with members of the first class, which comprised one isolate of human heterochromatin protein 1 alpha, HP1^Hsα (Saunders et al., 1993), and two independent isolates of HP1^Hsγ (Ye and Worman, 1996).

We performed a number of control experiments using a quantitative liquid assay with ONPG as the substrate in order to better characterize the two-hybrid interaction between HP1^Hsα and INCENP (Fig. 6,A). Neither INCENP _382-839, which fails to associate with chromosomes in vivo (Mackay et al., 1998), nor INCENP_48-85, which is encoded by a single conserved exon in the human INCENP sequence (Eckley, D.M., and W.C. Earnshaw, unpublished observations) and which contains the spindle targeting motif (residues 52–62), interacted with HP1^Hsα in the two-hybrid assay (Fig. 6,A). Further studies using a series of truncated INCENP molecules localized the HP1-binding site to a region between INCENP residues 135 and 270, and possibly between residues 135 and 200 (Fig. 6 B).

We next used an in vitro binding assay to confirm that HP1^Hsα and INCENP are capable of direct biochemical interaction. GST-tagged HP1^Hsα protein was expressed in E. coli, bound to glutathione agarose, and incubated with [³⁵S]methionine-labeled INCENP proteins that had been translated in vitro using reticulocyte lysates. Under these conditions, GST:HP1^Hsα bound selectively to INCENP_1-839 (Fig. 7, lane 2), INCENP_1-405 (lane 4), and INCENP_1-405: GFP (lane 12). In control experiments, these INCENP molecules showed no interaction with GST alone (Fig. 7, lanes 1, 3, and 11). The interaction between INCENP and HP1 in vitro was stable at physiological ionic strength, but was disrupted by higher salt concentrations (data not shown).

Because of the well known association of HP1^Hsα with centromeres (Nicol and Jeppesen, 1994; Wreggett et al., 1994), we had initially expected that the INCENP:HP1^Hsα interaction would be involved in targeting INCENP to centromeres. However, further experiments revealed that this is unlikely. INCENP_1-68:GFP, which defines the minimal INCENP centromere-targeting module (Fig. 3,D), failed to interact with HP1^Hsα in the two-hybrid and in vitro binding assays (Fig. 6 and Fig. 7, lanes 13 and 14). In addition, both INCENP_43-839 and INCENP_43-405, which associate with chromosomes but do not target to centromeres, showed significant interaction with HP1^Hsα (Fig. 6 and Fig. 7, lanes 6 and 8).

Although the two-hybrid and in vitro binding analyses suggest that HP1 is not involved in targeting INCENP to centromeres, it is possible that the INCENP:HP1 interaction at centromeres might be required for some other aspect of INCENP function in mitosis. To attempt to identify such a role of HP1 binding in INCENP function, we created a deletion construct, INCENP_1-839^(Δ69-404), from which the HP1-binding site was deleted. This protein failed to interact with HP1 in the in vitro binding assay (Fig. 8, A and B), yet it showed surprisingly normal behavior in vivo: it was nuclear in interphase, centromeric at metaphase, and located to the spindle midzone and midbody at ana/telophase (Fig. 8, C–F). Thus, HP1 cannot be solely responsible for directing any of the obvious aspects of INCENP behavior, at least for INCENP molecules that are capable of targeting to centromeres.

Therefore, we were surprised when a second experiment revealed that the interaction between INCENP and HP1 might after all be one component of a redundant system promoting the release of INCENP from chromosomes and targeting to the spindle midzone. Initially to examine the potential role of HP1 in centromere targeting by INCENP, we constructed two chimeric cDNAs that encoded proteins in which either the chromo domain (residues 1–68 of HP1^Hsα) or full-length HP1^Hsα was fused to the amino terminus of INCENP_45-839, a form of INCENP that we have shown previously to be unable to target to centromeres or to transfer to the spindle (Mackay et al., 1998). In a previous study, we found that a similar chimeric protein consisting of a fusion between the centromere-binding domain of CENP-B and INCENP_45-839 could target to centromeres, where it remained bound throughout the cell cycle (Eckley et al., 1997).

Analysis of transfected cells expressing either HP1^Hsα: INCENP_45-839 (data not shown) or HP1^Hsα_1-68:INCENP _45-839 (Fig. 9,D′) during mitosis revealed a striking result. Both chimeric polypeptides associated with mitotic chromosomes, but were unable to target to centromeres. However, despite this inability to target to centromeres, both chimeric proteins transferred efficiently to the spindle in anaphase (Fig. 9 D′′), whereas control protein INCENP _45-839 did not. Immunoblots of transfected cells confirmed that polypeptides of the predicted size could be observed, and that these and the control proteins accumulated to roughly equal amounts in vivo. This result suggests that the INCENP:HP1 interaction may be one component of a redundant mechanism promoting the transfer of INCENP to the spindle at anaphase.

Sequence analysis and functional experiments in vivo reveal that HP1 has a modular domain organization. The amino- and carboxyl-terminal portions of the molecule have been termed the chromo domain and the chromo shadow domain, respectively (Paro and Hogness, 1991; Aasland and Stewart, 1995). Both the Drosophila melanogaster HP1 chromo domain and chromo shadow domains could target the euchromatic protein Polycomb to heterochromatin when swapped for the Polycomb chromo domain (Platero et al., 1995), although only the chromo shadow domain could target to heterochromatin on its own (Powers and Eissenberg, 1993). Recent studies have identified a number of cellular proteins that interact with HP1s (α, β, and γ) (reviewed in Cavalli and Paro, 1998; Lamond and Earnshaw, 1998). Most of these interactions occur via recognition of the chromo shadow domain. Occasionally, the interaction requires the additional presence of the central so-called hinge region.

To gain further insight into the nature of the interactions between INCENP and HP1, we performed in vitro binding experiments (as described above) using a series of truncated HP1 proteins (Fig. 10,A) and full-length INCENP. These studies mapped the region in HP1^Hsα that interacts with INCENP to the hinge region separating the chromo and chromo shadow domains (Fig. 10). Deletion constructs expressing either the chromo domain (HP1^Hsα _1-68), or chromo shadow domain (HP1^Hsα_110-190), did not bind to INCENP_1-839 (Fig. 10, B and C, lanes 2 and 6). However, when either of these fragments was extended to include the hinge region (Fig. 10, lanes 3 and 5) or when the hinge region (HP1^Hsα_68-110) was expressed on its own (Fig. 10, lane 4), specific interactions were seen. This is the first demonstration of interaction of a chromatin protein exclusively with the hinge region of HP1.

Unlike other centromeric proteins, including CENPs A–C, E, and F, Mad2, and Bub1, which target directly to the centromere/kinetochore (Earnshaw and Rothfield, 1985; Yen et al., 1991; Casiano et al., 1993; Rattner et al., 1993; Zhu et al., 1995; Chen et al., 1996; Taylor and McKeon, 1997), INCENP binds first to noncentromeric sites along the chromosome arms, then moves to centromeres only during prometaphase/metaphase. We refer to this behavior as indirect centromere targeting.

Indirect centromere targeting by INCENP is directed by an autonomous module located in INCENP residues 1–68 (Fig. 11) and including the conserved sequence WLEEIHEEAARMF (residues 32–44, Fig. 11, motif C). This sequence is not related to previously determined centromere-targeting motifs (Pluta et al., 1992; Yang et al., 1996). An INCENP molecule in which the 13-aa centromere-targeting motif was mutagenized (INCENP_1-405^(32-44r): GFP), in addition to being defective in centromere targeting, was also defective in transfer to the anaphase spindle and midbody. This supports our previous suggestion (Mackay et al., 1998) that INCENP targeting to centromeres is an obligatory event upstream of its transfer to the spindle in ana/telophase.

During this analysis we unexpectedly discovered that the minimal centromere-targeting module also functions as an autonomous spindle midzone/midbody-targeting unit during ana/telophase. Site-directed mutagenesis studies revealed that a conserved sequence motif spanning residues 52–62 (PELMPKTPSQK), which was not required for indirect centromere targeting, was essential for spindle/midbody targeting by INCENP_1-405:GFP (Fig. 11, motif S). However, in addition to this amino-terminal motif, INCENP must also have a second spindle transfer determinant elsewhere, between residues 405 and 839: full-length INCENP_1-839 carrying the (52-62r) mutation also transfers to the spindle during anaphase (data not shown). However, both INCENP_1-405:GFP and INCENP_1-839^(52-62r) transferred to the central spindle abnormally late, in mid-anaphase, and with reduced efficiency (some mutant INCENP typically remained associated with the chromosomes, Fig. 2 and data not shown). Thus, although either the amino-terminal or carboxyl-terminal motif is sufficient to target centromere-associated INCENP to the midbody, the presence of both motifs appears to be essential for timely, complete, and efficient transfer to occur. These observations reveal that INCENP uses redundant mechanisms to target to the spindle midzone, a theme that is emphasized in our studies of the interaction of INCENP with HP1 (see below).

INCENP_1-405:GFP is released from the centromeres back onto the chromosome arms at anaphase onset, yet still transfers to the spindle during mid-anaphase (Fig. 2). A similar release of INCENP back onto the chromosome arms occurs in rare cells in which INCENP transfer to the spindle is spontaneously delayed (Earnshaw, W.C., unpublished observations). However, in our various studies, in observations of many thousands of cells, we have never seen a late anaphase/telophase cell in which transfer of endogenous INCENP to the spindle had failed to occur. These observations reveal a paradox. Why do INCENP _43-405, INCENP_43-839, and INCENP_27-839, all of which occupy similar locations on the chromosome arms during anaphase and which contain an intact spindle/midbody-targeting motif (aa 52–62), fail to transfer to the central spindle at any time during ana/telophase? The most obvious difference between these molecules is that like endogenous INCENP, INCENP_1-405:GFP had previously targeted centromeres during metaphase, whereas the other mutant proteins had not.

We suggest that INCENP may undergo some sort of priming event at centromeres that renders it capable of targeting to the spindle midzone and midbody even if it leaves the centromeres before its exit from the chromosome. This priming event could be a posttranslational modification such as phosphorylation by a centromere-associated kinase such as active ERK (Shapiro et al., 1998; Zecevic et al., 1998) and/or an interaction with a centromeric protein that promotes INCENP release from the chromosomes at anaphase. To explore this second possibility, we have screened for proteins with which INCENP might interact at centromeres.

A yeast two-hybrid screen revealed a specific interaction between INCENP and heterochromatin proteins HP1^Hsα and HP1^Hsγ. HP1s are thought to be structural adapter proteins involved in the assembly of chromatin complexes (reviewed in Lamond and Earnshaw, 1998). HP1s interact with a number of proteins including HP1 itself (Ye et al., 1996), the lamin B receptor (Ye and Worman, 1996), Orc1 and Orc2 (Pak et al., 1997), and SP100, a major component of PML nuclear bodies (Lehming et al., 1998; Seeler et al., 1998). The INCENP:HP1 interaction, which was confirmed in in vitro binding studies, involves residues 135–270 of INCENP and the central, so-called hinge region of HP1. Although inclusion of the hinge region has been reported to increase the binding of some proteins in vitro to HP1 (Ye et al., 1996), interaction solely with this sequence has not been reported previously. It is possible that INCENP binding to HP1 might leave the chromo and chromo shadow domains available for further interactions and possible ternary complex formation.

For the interaction between HP1^Hsα and INCENP to be functionally significant, then the two proteins should colocalize in vivo, at least during some portion of the mitotic cycle. In fact, indirect immunofluorescence on metaphase chromosome spreads has revealed that both INCENP and HP1^Hsα are distributed throughout the central domain of the mammalian centromere (Nicol and Jeppesen, 1994; Wreggett et al., 1994; Mackay et al., 1998), consistent with an association between the two proteins in vivo.

The observed interaction between HP1^Hsα and INCENP suggests two possibilities. First, binding to HP1 might be responsible for targeting INCENP to centromeres. This model is effectively ruled out by our two-hybrid, in vitro binding and in vivo transfection studies. Second, interaction with HP1 could be a component of the priming event that occurs at centromeres and which is required for INCENP function during late mitosis. The interaction of HP1 with INCENP cannot be solely responsible for priming the transfer of INCENP to the spindle, since two proteins that do not bind HP1 (INCENP_1-68 and INCENP_1-839^[Δ69-404]) can transfer to the spindle and one that does bind HP1 (INCENP_1-839^[52-62r]) does not transfer. However, both of the proteins that do transfer also target to centromeres and could have received a second priming event there.

Other results suggest that the INCENP:HP1 interaction could have a role in spindle transfer. Thus, fusion of either the chromo domain (residues 1–68 of HP1^Hsα) (Aasland and Stewart, 1995) or full-length HP1^Hsα to INCENP truncation mutant INCENP_45-839 could bypass the need for the protein to target to centromeres. Both chimeric proteins transferred efficiently to the spindle during anaphase (Fig. 9), despite the fact that neither was able to target to centromeres. Although results obtained with such chimeric fusions must be interpreted with caution, the fact that two different constructs both restored the spindle transfer function is consistent with the possibility that INCENP interaction with HP1 is one component of a complex priming event that INCENP normally undergoes at centromeres. Thus, these results suggest an unexpected model in which the heterochromatin protein HP1 may promote INCENP exit from the chromosomes during anaphase.

The experiments reported here have begun to characterize the mechanisms of INCENP movements in mitosis. The identification of discrete motifs required for centromere targeting and spindle transfer will be particularly useful in future screens to identify the relevant ligands that interact with INCENP at these disparate cellular locations. Evidence obtained thus far suggests that INCENP targeting to centromeres may be relatively straightforward, requiring a compact module located within the first 68 aa, but that transfer of the protein to the spindle during anaphase may be more complex. Spindle/midbody transfer is contributed to by sequences at both ends of the protein, and appears to involve substantial functional redundancy. The characterization of the INCENP:HP1 interaction and the suggestion that this interaction might be one factor involved in transfer of INCENP to the spindle have begun to reveal the mechanism by which the various movements of INCENP are regulated.

We thank Roger Brent for the gift of the materials necessary for the two-hybrid screen, Lester Binder for sanctioning the gift of TU27b mAb, and Martie Chalfie for the gift of GFP. We thank our colleagues Yasuhisa Adachi, Richard Adams, Jeremy Brown, Jeff Craig, Hiro Ohkura, Paola Vagnarelli, and Sally Wheatley for their helpful comments on the manuscript.

aa

amino acids

DAPI

4,6-diamidino-2-phenylindole

GFP

green fluorescent protein

GST

glutathione S-transferase

HP1^Hsα and HP1^Hsγ

human heterochromatin protein α and γ

INCENP

inner centromere protein

ONPG

o-nitrophenyl β-d-galactopyranoside