Specific interaction between human kinetochore protein CENP-C and a nucleolar transcriptional regulator.

CENP-C is a human kinetochore protein that was originally identified as a chromosomal autoantigen in patients with scleroderma spectrum disease. To begin to establish a comprehensive protein map of the human centromere, affinity chromatography was used to identify nuclear proteins that specifically interact with CENP-C. Whereas a number of polypeptides appeared to interact with the full-length CENP-C protein, only a pair of similarly sized proteins of ~100 kDa specifically interacted with the isolated carboxyl-terminal third of the CENP-C protein. Neither protein of the doublet bound to control affinity columns. Affinity purification and microsequence analysis of the proteins in the doublet identified them as the two highly related nucleolar transcription factors, UBF1 and UBF2 (also known as the nucleolar autoantigen NOR-90). Immunoblot analysis confirmed that both proteins also interacted with the full-length CENP-C polypeptide with similar affinities. Double indirect immunofluorescence using monospecific antibodies demonstrated that a subset of CENP-C and UBF/NOR-90 is colocalized at nucleoli of interphase HeLa cells, suggesting that the in vitro interaction detected by affinity chromatography may reflect an interaction that occurs in vivo. We discuss the implications of these findings in terms of the properties of interphase centromeres and the role of the nucleolus in scleroderma autoimmunity.

CENP-C is a human kinetochore protein that was originally identified as a chromosomal autoantigen in patients with scleroderma spectrum disease. To begin to establish a comprehensive protein map of the human centromere, affinity chromatography was used to identify nuclear proteins that specifically interact with CENP-C. Whereas a number of polypeptides appeared to interact with the full-length CENP-C protein, only a pair of similarly sized proteins of ϳ100 kDa specifically interacted with the isolated carboxyl-terminal third of the CENP-C protein. Neither protein of the doublet bound to control affinity columns. Affinity purification and microsequence analysis of the proteins in the doublet identified them as the two highly related nucleolar transcription factors, UBF1 and UBF2 (also known as the nucleolar autoantigen NOR-90). Immunoblot analysis confirmed that both proteins also interacted with the full-length CENP-C polypeptide with similar affinities.

Double indirect immunofluorescence using monospecific antibodies demonstrated that a subset of CENP-C and UBF/NOR-90 is colocalized at nucleoli of interphase HeLa cells, suggesting that the in vitro interaction detected by affinity chromatography may reflect an interaction that occurs in vivo.
We discuss the implications of these findings in terms of the properties of interphase centromeres and the role of the nucleolus in scleroderma autoimmunity.
The genetic material of eukaryotes is packaged into chromosomes through its interactions with both histones and nonhistone chromosomal proteins. Each of the histones is present in vast quantities in the cell and, as a component of nucleosomes, is responsible for the most basic level of chromatin organization. In contrast, individual non-histone chromosomal proteins are considerably less abundant and have more specialized roles, including the structural and functional organization of specific chromosomal elements. One such element is the centromere, the chromosomal region responsible for the precise and accurate segregation of the genetic material during mitosis and meiosis. The centromere directs the segregation of mitotic chromosomes through a differentiated trilaminar structure called the kinetochore, which serves as the binding site for spindle microtubules and for the mechanochemical motors that move chromosomes along those microtubules. The centromere also regulates the separation of sister chromatids at the metaphase-anaphase transition.
A complete understanding of how the centromere coordinates these important functions requires comprehensive knowledge of the proteins involved. The identification of the protein components of the human centromere initially seemed a daunting task until it was discovered that patients with scleroderma spectrum disease produce circulating autoantibodies that recognize several centromeric polypeptides (1,2). cDNAs encoding three centromere proteins, CENP-A, -B, and -C, have now been cloned, and the proteins have been studied in detail. CENP-A is a novel centromere-specific core histone related to histone H3 (3). CENP-B is an ␣-satellite DNA-binding protein that is localized throughout the centromeric heterochromatin located beneath the kinetochore (4 -6). CENP-C is also a DNA-binding protein and is located at the interface between the centromeric heterochromatin and the innermost region of the kinetochore (7).
Despite these advances, several obstacles have hindered the identification of other proteins that compose the human centromere and kinetochore. First, centromere proteins are very minor constituents in the cell. For example, it has been estimated that CENP-B is present in 20,000 -50,000 copies/cell (8,9). Second, like many structural proteins, centromere proteins are relatively insoluble, a property that complicates their isolation and characterization. Finally, there are no good functional assays available to screen for mammalian centromere proteins, nor are there any widely applicable methods for directly purifying these chromosomal structures.
Despite these difficulties, two different approaches have identified a handful of new centromere proteins. One of these involves the visual/empirical survey of known proteins suspected of being involved in centromere function using monospecific antibody probes and immunoelectron microscopy of mitotic cells (10,11). In the other approach, centromere proteins are identified by generating monoclonal antibody probes to chromosomal protein fractions enriched in known centromeric autoantigens. This type of experiment has now been done using human, chicken, and Xenopus cells as starting material and has resulted in the identification of CENP-E, a kinesinrelated microtubule-binding protein (12,13); the INCENP proteins (inner centromere proteins), chromosomal passenger proteins (14,15); and the 3F3/2 kinetochore-localized phosphoepitopes that may be involved in cell cycle signaling (16,17). In addition, routine screening of patient autoimmune sera continues to occasionally yield a novel centromere-associated autoantigen (e.g. CENP-F) (18,19). These approaches are unlikely to have identified all (or even a majority) of the protein components of the human centromere and kinetochore. Furthermore, none of these approaches begins to characterize the protein-protein interactions that functionally and structurally define the centromere and kinetochore.
Mounting evidence suggests that the 140-kDa human centromeric autoantigen, CENP-C, plays a pivotal role in centromere structure and function. First, the demonstration that CENP-C is found only at the active centromere of a stable dicentric chromosome, while CENP-B is found at both the active and inactive centromeres, suggests a direct role for CENP-C in centromere function (20). Second, CENP-C has been immunolocalized to the inner plate of the kinetochore, a structure intimately involved in the attachment of chromosomes to the mitotic spindle (7). Third, evidence from antibody microinjection experiments indicates that CENP-C is involved in the assembly of a structurally sound, morphologically normal kinetochore and is required for the cell's timely transition into anaphase (21).
We have developed a biochemical approach to identify chromosomal proteins involved in centromere function. By exploiting CENP-C as an affinity handle to dissect protein interactions at the centromere, we have identified a specific interaction with two protein components of the nucleolus. Our results are consistent with known properties of interphase centromeres and the observed involvement of the nucleolus in the autoimmune response in scleroderma spectrum disease.

Expression and Purification of CENP-C and Control Baits-CENP-
C-containing affinity baits expressed in bacteria as GST 1 fusion proteins using pGEX2T (Pharmacia Biotech Inc.) were derived from pT-CATG, an expression plasmid containing the entire CENP-C coding region. pTCATG was constructed by ligating the NdeI (site underlined)and SphI-digested polymerase chain reaction product resulting from amplification of CNPC7 (7) with oligonucleotide primers CENPCATG (5Ј-CCGATACATATGGCTGCGTCCGGTCTGG-AT-3Ј) and CENPC2 (5Ј-GGTATTGTAATCCAAGATCTACTGGC-3Ј) to NdeI-and SphI-digested pTCNPC␣ (7). The full-length CENP-C open reading frame was isolated from pTCATG by digestion with NdeI followed by treatment with T4 DNA polymerase and EcoRI digestion and cloned into BamHIdigested and T4 DNA polymerase-treated pGEX2T, which was also digested with EcoRI. The correct reading frame of the resulting construct, pGEX2T/C1-943 (in which the numbers refer to the amino acid residues of CENP-C), was confirmed by DNA sequence analysis.
pGEX2T/C1-943 was digested with BglII and EcoRI, and the resulting 5.9-kilobase fragment containing the entire pGEX2T sequence and the amino-terminal third of the CENP-C open reading frame was gelisolated, treated with T4 DNA polymerase, and self-ligated to create pGEX2T/C1-315. The 1074-base pair BglII-EcoRI fragment resulting from this same digestion was cloned into BamHI-and EcoRI-digested pGEX2T to create pGEX2T/C635-943.
GST-C1-943 failed to bind glutathione-agarose and was therefore partially purified from induced bacteria as follows. Cells were lysed by sonication at 0°C in Solution I (10 mM NaPO 4 , pH 7.0, 1 mM EDTA, 1 mM EGTA, and 0.1% Nonidet P-40) containing 25 mM NaCl, 0.2 mg/ml lysozyme, and protease inhibitors. The supernatant resulting from centrifugation at 12,000 ϫ g for 20 min was brought to 33% ammonium sulfate, rotated for several hours at 4°C, and then centrifuged at 12,000 ϫ g for 10 min. The pellet was resuspended in Solution I and dialyzed extensively against the same buffer lacking detergent and then centrifuged at 12,000 ϫ g for 10 min. The resulting supernatant was adjusted to 10 mM MgCl 2 , incubated on ice for 45 min, and then centrifuged at 3000 ϫ g for 10 min. The resulting pellet was resuspended in Solution I containing 400 mM NaCl, 10 mM dithiothreitol, and protease inhibitors and centrifuged for 10 min at 120 ϫ g. The resulting supernatant was enriched for GST-C1-943. Control baits were purchased from Sigma (lysozyme) or affinity-purified as described above from induced bacteria transformed with pGEX2T (GST).
Preparation of Protein Affinity Columns-Purified baits were coupled with Affi-Gel 10 (Bio-Rad) essentially as described (22) at a concentration of 1-2 mg of protein/ml of Affi-Gel. Columns containing 2-4 ml of coupled Affi-Gel were constructed in plastic disposable syringes.
Nuclear Extracts-HeLa nuclei were typically isolated from 20 liters of cells grown to a density of 5-8 ϫ 10 5 cells/ml in Dulbecco's modified Eagle's medium/Ham's F-12 medium (1:1 mixture) with 2% calf serum and antibiotics. Cells were resuspended in 50 -100 mM PIPES, 1 mM EGTA, 1 mM MgSO 4 , 1 mM dithiothreitol, and protease inhibitors and then lysed by sonication on ice. Lysates were centrifuged at 2800 ϫ g in a Sorvall SS34 rotor for 10 min. The resulting nuclear pellets were resuspended in the same buffer and quick-frozen in liquid nitrogen. For extract preparation, frozen nuclei were thawed quickly in a room temperature water bath, and their volume was estimated. 0.5 volume of cold H 2 O containing protease inhibitors, 5 mg/ml RNase A, and 2 mg/ml nuclease P 1 (Pharmacia Biotech Inc.) was added, and the mixture was incubated on ice for 30 min. Digested nuclei were homogenized by douncing on ice and then aliquotted to 40-ml Sorvall tubes. To each tube (containing 10 ml of nuclei) were added 10 ml of cold 1 ϫ TEE (10 mM triethanolamine HCl, 0.1 mM EDTA, pH 9.0) and 20 ml of cold 2 ϫ lysis buffer (20 mM Tris, pH 9.1, 20 mM EDTA, 20 mM EGTA, 4 M NaCl, 0.2% Ammonyx Lo, and 60 mM 2-mercaptoethanol), both containing protease inhibitors. The contents of each tube were then homogenized by douncing on ice followed by sonication to decrease the viscosity of the mixture and then centrifuged at 16,000 ϫ g for 30 min. Supernatants were pooled, adjusted to 10 -15% glycerol and 1 mg/ml polyethylene glycol 8000, and dialyzed against several changes of RB buffer (10 mM Tris, pH 9.1, 10 mM EDTA, 10 mM EGTA, 150 mM NaCl, 10% glycerol, 0.2% Ammonyx Lo, 2 mg/ml polyethylene glycol 8000, and 30 mM 2-mercaptoethanol) at 4°C. Dialyzed extracts were centrifuged at 10,000 ϫ g for 30 min, and the supernatants were pooled and used for affinity chromatography.
Chromatography and Gels-Affinity columns coupled with protein baits were equilibrated with at least 20 volumes of RB buffer at 4°C and then loaded with nuclear extract, which was recirculated over experimental and control columns overnight in parallel. The following day, columns were washed with at least 20 volumes of RB buffer, and specifically bound proteins were eluted with RB buffer containing a 150 mM to 1 M NaCl gradient. Gradient fractions containing eluted proteins were trichloroacetic acid-precipitated in the presence of tRNA, resolved on 5-15% SDS-polyacrylamide gradient gels, and visualized by silver staining (Bio-Rad).
Affinity Purification of Doublet and Internal Peptide Microsequence Analysis-The ϳ100-kDa doublet was affinity-purified from HeLa nuclear extract (ϳ150 liters of cells) on three different GST-C635-943 affinity columns. Fractions containing the doublet, determined by silver-stained SDS-polyacrylamide gels, were pooled, concentrated by dialysis against Aquacide (Calbiochem), and then precipitated with 20% trichloroacetic acid. The resulting pellet was washed with 90% acetone and 10% HCl, resuspended in SDS sample buffer, resolved on a 6.5% SDS-polyacrylamide gel, and transferred to polyvinylidene difluoride membranes (Bio-Rad). Bands corresponding to each protein of the doublet were visualized by staining the membrane with 0.2% Ponceau S in 1% acetic acid and excised from the membrane. Internal microsequence analysis was performed at the Harvard Microchemistry Facility.
Immunoblotting and Immunofluorescence-Electrophoresis and immunoblotting were performed as described previously (23). Immunodetections using ECL (Amersham Corp.) and horseradish peroxidaseconjugated anti-human and anti-rabbit IgG (Amersham Corp.) were performed according to the manufacturer's instructions. Indirect immunofluorescence was performed essentially as described (24). Images were obtained using a DAGE SIT camera with a Perceptics PixelPipe-line board driven by a modified version of Adobe Photoshop. 2 Human autoimmune S14 serum (25) was obtained from D. Hernandez-Verdun; affinity-purified rabbit polyclonal anti-UBF was obtained from H. Beckmann and R. Tjian. Polyclonal anti-CENP-C antibody (anti-g296 -943) was as described (7).

Preparation of CENP-C Affinity Columns and Nuclear Extracts-
The purpose of this study was to begin to develop a structural map of the human kinetochore, starting with CENP-C, a well characterized component of the inner kinetochore plate. CENP-C is a particularly suitable subject for such studies as it has been previously shown that this protein is important for determining the overall diameter of the human kinetochore plates (21). Four methods have recently come into wide use for the identification of ligands that interact with a protein of interest. 1) Genetics provides a powerful means of detecting interactions between gene products in certain organisms, especially those where the ability to manipulate chromosomal DNA sequences has been developed. 2) The yeast two-hybrid interaction trap provides a way of detecting protein domains that interact with sufficient affinity to generate a transcriptional activator whose activity can be monitored by the expression of reporter constructs (26,27). 3) Co-immunoprecipitation is frequently used to detect associations between various soluble proteins. 4) Finally, affinity chromatography, the least widely used of these methods, provides a way of detecting interacting components and assessing their relative affinities for one another (28).
In the case of CENP-C, the repertoire of methods that can be used to search for interacting proteins is limited. CENP-C has yet to be detected in any genetically tractable organism outside of mammals. Likewise, co-immunoprecipitation is excluded by the fact that CENP-C is a structural protein that is soluble only under denaturing conditions that disrupt most biologically relevant protein-protein interactions. 3 Thus, the yeast interaction trap 4 and affinity chromatography, which forms the subject of this report, are left as the two viable methods to search for proteins that interact with CENP-C.
Affinity chromatography requires the purification of significant quantities of the protein that is to be used as the "bait" and the development of an appropriate extract of soluble cellular proteins, which is the source of potential interacting proteins or ligands (22). Both of these requirements presented unique challenges for the identification of proteins that interact with CENP-C because of the nature of human centromere proteins. Human centromere proteins (and, presumably, the proteins with which they interact) are not only exceedingly minor components in the cell, but are also associated with the chromosome scaffold fraction (23). Chromosome scaffold proteins are distinguished from other chromosomal proteins by their unusual degree of insolubility. Solubilization of these proteins typically requires not only conditions designed to disrupt protein-protein interactions, such as high salt or low salt in the presence of polyanions, but also requires the addition of substantial levels of a reducing agent, such as ␤-mercaptoethanol (29,30).
This combination of low abundance and insolubility precluded the purification of endogenous CENP-C from human cells for use as an affinity bait. Therefore, to obtain the milligram quantities of CENP-C necessary to make useful affinity columns, full-length CENP-C and several portions of the CENP-C polypeptide were affinity-purified as GST fusion pro-teins from induced bacterial cells that had been transformed with the appropriate pGEX/CENP-C construct. Fig. 1 diagrams the CENP-C-derived GST fusion proteins used as baits in our experiments and lists their predicted properties. The hydropathy plot shown in Fig. 1A predicts that GST-CENP-C is a hydrophilic and highly basic (pI 9.2) protein. Two control baits were also included in all affinity chromatography experiments. One of these, lysozyme (pI ϳ10), was included as a control for nonspecific binding of acidic ligands to the three basic CENP-C-derived baits. The second control bait consisted of GST alone since this bacterial peptide was fused to all CENP-C-derived constructs. In Fig. 1C, the expressed CENP-C baits are displayed on Coomassie Blue-stained polyacrylamide gels. All partial CENP-C baits were affinity-purified on GST-agarose. GST-C1-943, which contains the full-length CENP-C open reading frame, failed to bind GST-agarose and was instead partially purified from induced bacterial extracts by standard biochemical fractionation (see "Materials and Methods").
The development of a suitable extract to be used as the source of putative interacting proteins was also dictated by the known properties of human centromere proteins described above. A protein extract derived from purified human mitotic chromosomes would, in theory, be the ideal source for identifying proteins that interact with a component of the inner kinetochore plate. In practice, however, the vast amounts of purified chromosomes required as starting material if sufficient amounts of interacting proteins are to be recovered for subsequent analysis and identification make this impractical. The compromise was to develop a protein extract from crude preparations of interphase nuclei obtained from HeLa cells grown in large chemostat cultures. Interphase nuclei are a reasonable source of proteins that interact with centromere proteins because several of the CENP proteins, including CENP-C, are constitutive components of interphase centromeres (2, 20, 31). While it is certainly possible that some interactions that require mitosis-specific post-translational protein modifications would not be detected using a nuclear extract as a source of interacting proteins, it is also possible that previously unsuspected interactions that are specific to interphase centromeres could be identified by using such an extract.
Among the requirements for a successful affinity chromatography experiment is that the putative interacting ligands exist as soluble proteins in the extract. In practice, this requirement proved difficult to satisfy. To produce a nuclear extract containing soluble proteins, the putative (unknown) interactors were assumed to behave biochemically like known centromere proteins, which can be monitored by immunoblotting. If an initial high salt nuclear lysate was made by solubilizing isolated nuclei in a buffer containing 2 M NaCl and 30 mM 2-mercaptoethanol (Fig. 2), the majority of nuclear proteins detected by Coomassie Blue staining, including CENP-A, -B, and -C, which were detected by immunoblotting, were recovered in the supernatant or soluble fraction. Such a high salt lysate cannot, however, be used to detect ionic interactions by affinity chromatography. It was therefore necessary to reduce the ionic strength of the solubilized nuclear ligands prior to their passage over the affinity columns by either dialysis or dilution. Initially, either step resulted in the abrupt precipitation of the bulk of the chromosomal proteins including all of the CENP antigens. Eventually, buffer conditions were established that permitted the reduction of the salt concentration of the nuclear lysate to near physiological levels (150 mM) while maintaining the solubility of the majority of total protein as well as significant amounts of centromere proteins detectable by immunoblotting (Fig. 2). This soluble low salt extract was used as the source of interacting proteins in subsequent affinity chromatography experiments.
CENP-C Affinity Chromatography-In a typical affinity chromatography experiment, a low salt nuclear extract derived from ϳ2 ϫ 10 10 HeLa cells was pumped simultaneously over four affinity columns, which were set up in parallel, and recirculated overnight at 4°C. Two of these affinity columns were coupled with the control baits, GST and lysozyme. The remaining two affinity columns were each coupled with a different experimental bait consisting of a CENP-C-derived GST fusion protein. Following extensive washing of the affinity columns with running buffer containing 150 mM NaCl to remove nonspecifically associated nuclear proteins, specifically bound proteins were eluted from each column with a linear salt gradient and visualized on silver-stained polyacrylamide gels.
The protein elution profiles from a typical experiment in which the two experimental columns were coupled with GST-C1-943 and GST-C635-943 are shown in Fig. 3. It is immediately obvious that while very few proteins could be eluted from either control column (panels A and B), a unique spectrum of proteins could be eluted from each of the experimental columns (panels C and D). This ranged from no proteins specifically eluted from the column baited with the amino-terminal 315 amino acids of CENP-C (data not shown), to a complex mixture of proteins that was eluted from the column baited with fulllength CENP-C (panel C). That the binding of this complex mixture was specific is suggested by the observation that each ionic strength appeared to elute a specific subset of proteins. The most striking results were obtained with columns baited with the carboxyl-terminal 308 amino acids of CENP-C (panel D). Only three proteins were eluted from this column as detected on a silver-stained gel. Two of these proteins composed a doublet with a relative molecular mass of ϳ100 kDa, which coeluted from the column at 400 -600 mM NaCl. This coelution property suggested either that the two proteins were closely related to each other and thus interacted in similar ways with the bait or that they were in a complex with each other. (The third protein of ϳ60 kDa indicated by the open arrow in panel D was found by immunoblot analysis to be incompletely coupled bait, which eluted from the column at higher salt concentrations (data not shown).) Neither band of the ϳ100-kDa doublet was observed to elute from either control column, confirming that the binding of these proteins to the column was due to specific interactions with the CENP-C-derived portion of the GST-C635-943 fusion protein.
In similar experiments in which the middle third of CENP-C was used as bait, several low molecular mass proteins specifically eluted from this column. 3 These proteins were not studied further. The complex mixture of proteins eluted from the affinity column coupled with full-length CENP-C was unexpected and will clearly require a secondary screen in order to identify those ligands worthy of further characterization. As such a screen, we have raised monoclonal antibodies against the pools of proteins that eluted from this column at low, medium, and high salt. The results of this screen will be presented elsewhere.
Specific Association of UBF1/NOR-90 with the Carboxyl Terminus of CENP-C in Vitro-Because only two proteins eluted from the affinity column baited with the carboxyl-terminal third of CENP-C, we decided to focus our initial efforts on the further characterization of this fraction. The fact that these proteins eluted from the column at a moderately elevated ionic strength (ϳ500 mM NaCl) suggests that their interactions with CENP-C are of reasonably high affinity, thus making them good candidates for further analysis. Therefore, the two proteins composing the ϳ100-kDa doublet were affinity-purified from nuclear extracts derived from ϳ150 liters of HeLa cells, resolved on SDS-polyacrylamide gels, and transferred to polyvinylidene difluoride membranes. The Ponceau S-stained band corresponding to each protein of the doublet was excised from the membrane and processed for microsequence analysis. Amino acid analysis indicated that 10 -15 pmol of each protein had been affinity-purified. Following enzymatic digestion and peptide separation by high performance liquid chromatography, the lower band of the doublet yielded sequenceable amounts of three peptides. Two of those peptides gave sequences 17 and 8 amino acids in length (Fig. 4). A computer homology search of the protein data base with the 17-amino acid peptide sequence indicated a 100% match with two human proteins, UBF1 ( upstream binding factor 1), a ribosomal RNA (nucleolar) transcription factor, and an alternatively spliced form of UBF1 that has a 37-amino acid deletion (residues 221-257) (32). This variant of UBF1, also called NOR-90 ( nucleolus organizer region), was independently identified as a nucleolar autoantigen in scleroderma (33,34). A visual search of the UBF1 peptide sequence then located a region that matched the partially determined 8-amino acid peptide perfectly (residues 628 -635 of UBF1) (Fig. 4). The relative molecular masses of UBF1 and NOR-90, which are 97 and 94 kDa, respectively, are consistent with the observed molecular mass (ϳ100 kDa) of the affinity-purified doublet. Thus, the protein doublet detected by GST-C635-943 affinity chromatography was tentatively identified as the two closely related proteins, UBF1 and NOR-90. This identification was subsequently confirmed by immunoblot analysis using S14, a well characterized human autoimmune serum specific for UBF/NOR-90 (25). In Fig. 5A, the affinity-purified doublet, which was visualized by silver staining in the left panel, reacted with the autoimmune serum (right panel) and produced a signal that comigrated with endogenous UBF/NOR-90 from isolated HeLa nuclei.
Because the bait used for the original identification of the ϳ100-kDa doublet was a partial CENP-C molecule, an important control was to determine whether UBF/NOR-90 bound with similar affinity to a column coupled with the full-length CENP-C polypeptide. Accordingly, immunoblot analysis of protein fractions eluted from the GST-C1-943 affinity column was performed using a polyclonal antibody specific for UBF/NOR-90. The characteristic doublet signal was observed for fractions eluting at an ionic strength similar to that observed for the elution of UBF/NOR-90 from the GST-C635-943 affinity column (Fig. 5B). Therefore, we conclude that the interaction of UBF/NOR-90 with the full-length CENP-C protein is mediated by amino acid residues residing in the carboxyl-terminal third of CENP-C.
A Subset of CENP-C and UBF/NOR-90 Is Colocalized in HeLa Cells-As a first step to confirm that the interaction between CENP-C and UBF/NOR-90 detected in vitro can occur in vivo, we examined the distribution of CENP-C and UBF/ NOR-90 in cultured human cells. We used double-label indirect immunofluorescence to simultaneously localize CENP-C and UBF/NOR-90 in formaldehyde-fixed HeLa cells (Fig. 6). UBF/ NOR-90 was detected with human autoimmune S14 serum. CENP-C was detected with a monospecific polyclonal antibody (7). In mitotic cells (panels C and D), UBF/NOR-90 was concentrated at the chromosomal NOR regions of five pairs of acrocentric chromosomes, which placed these proteins in a position near to, but distinct from, mitotic centromeres detected with the CENP-C antibody. In interphase cells (panels A and B), S14 gave a punctate staining pattern that was confined to interphase nucleoli as previously documented for this serum (25). In the same cells, the CENP-C antibody gave a typical staining pattern for interphase centromeres: punctate nuclear spots, a number of which closely encircle nucleoli and colocalize with UBF. The limited colocalization we observed between CENP-C and UBF/NOR-90 in interphase nuclei is consistent with immunoblot data demonstrating that the amount of UBF/ FIG. 3. Affinity chromatography using full-length and carboxyl-terminal CENP-C fusion protein baits. Specifically bound proteins eluted with a 0.15-1.0 M NaCl gradient from four affinity columns loaded in parallel with the same low salt nuclear extract were trichloroacetic acid-precipitated using tRNA as carrier and displayed on silver-stained SDS-polyacrylamide gradient gels. The bait protein used to affinity purify the eluted proteins is indicated in the upper right-hand corner of each gel. A and B show the profile of proteins eluted from the columns coupled with the control baits GST and lysozyme, respectively. C and D show proteins eluted from the columns coupled with the CENP-C-derived baits GST-C1-943 and GST-C635-943, respectively. Equivalent volumes of consecutive protein fractions were loaded on each gel, with increasing ionic strength of elution proceeding from the left-most fraction to the right-most for each gel. Arrowheads in D indicate the ϳ100-kDa protein doublet, which was isolated for further study; the open arrow indicates incompletely coupled bait, which eluted from this column at higher salt concentrations. Lanes M, molecular mass markers.
FIG. 4. Peptides isolated from the affinity-purified doublet are homologous to UBF1 and NOR-90, two related nucleolar transcription factors. In this schematic representation, the hatched boxes indicate the positions of the high mobility group box motifs, while the black boxes indicate two highly acidic regions in the carboxyl termini (42). UBF1 and NOR-90 are identical except for a 37-amino acid (aa) deletion in NOR-90 that eliminates the second high mobility group box. The sequences and relative positions of the two peptides isolated from the affinity-purified doublet are shown; numbers in parentheses refer to amino acid residues of UBF1.
NOR-90 detectable in the post-affinity column nuclear extract is indistinguishable from that detectable in the pre-affinity column nuclear extract (Fig. 5C). These results bolster the conclusion that a subfraction of CENP-C may interact with UBF/NOR-90 in the interphase nucleus.
We have attempted to address the nature of the interaction between CENP-C and UBF/NOR-90 further by performing direct in vitro protein-protein interaction studies using purified bacterially expressed GST-C635-943 and UBF protein produced by in vitro transcription/translation. While every effort was made to reproduce the conditions used in our original affinity chromatography experiments in which the interaction between CENP-C and UBF/NOR-90 was initially identified, we have been unable to demonstrate an interaction between these two proteins (data not shown). Other efforts to confirm the interaction between CENP-C and UBF/NOR-90 in vivo have to date proven to be technically unfeasible. Co-immunoprecipitation experiments were not practical as we have only been able to immunoprecipitate CENP-C following denaturation in SDS, which efficiently disrupts most protein-protein interactions. The yeast two-hybrid interaction trap screen was also not feasible as we were unable (despite repeated efforts) to clone the UBF1 cDNA into the interaction trap prey vector pJG4-5 (26) in the correct orientation. DISCUSSION We developed a biochemical approach to identify proteins that interact with the human kinetochore protein CENP-C and have found that two related nucleolar proteins specifically bind affinity columns coupled with CENP-C via residues at its carboxyl terminus. Several facts suggest that the interaction we have detected between CENP-C and UBF/NOR-90 by protein affinity chromatography is significant. First, the interaction of UBF/NOR-90 with the CENP-C carboxyl terminus was highly reproducible and allowed the affinity purification of sufficient quantities of the proteins from nuclear extracts for direct microsequence analysis. Second, we have demonstrated that CENP-C and UBF show limited colocalization in interphase nuclei. Finally, it has recently been shown that transient expression of CENP-C amino acid residues 638 -829 (which do not contain the centromere targeting signal) in HeLa cells results in a protein that specifically accumulates in nucleoli (35). Thus, the potential for the carboxyl-terminal region of CENP-C to interact with nucleoli in vivo appears to be regulated in the context of the whole protein.
Both CENP-C and UBF have previously been functionally characterized. CENP-C is a basic protein capable of binding DNA in vitro (7). The DNA-binding domain of CENP-C has been localized to a short region (ϳ116 amino acid) located in the center of the protein (residues 422-537) (35). It is not yet known whether CENP-C binds a specific DNA sequence. The DNA-binding region overlaps with a domain (residues 478 -537) that is necessary and sufficient to target CENP-C to centromeres in transfected HeLa cells (35). No function has yet been identified for the carboxyl-terminal third of CENP-C (residues 635-943), which we have shown here to interact with UBF/NOR-90. However, several observations suggest that the carboxyl terminus of CENP-C is functionally important. First, the carboxyl terminus of CENP-C is the most highly conserved region of the protein between humans and mice (36). Second, this part of CENP-C shares two regions of amino acid similarity (40% identity over 27 amino acids and 28% identity over 52 amino acids) with an essential budding yeast protein, Mif2p (36,37). Properties of MIF2 mutants are consistent with Mif2p being a centromere protein (36). Interestingly, two mutant alleles of MIF2 that display defects in chromosome segregation and genetically interact with mutants in other yeast centromere proteins have base pair changes that lead to amino acid substitutions in the region of high CENP-C homology. Thus, if this region of CENP-C/MIF2 is functionally conserved, one of those functions may involve interactions with other proteins, including UBF/NOR-90. In this regard, it is noteworthy that one of the centromere proteins with which mutant Mif2p interacts genetically is Cpf1p (36), which, in addition to binding the centromere DNA element CDEI, acts (like UBF/NOR-90) as a transcription factor at other loci in the yeast genome (38 -40).
UBF1 is a highly charged protein that binds DNA sequences in the ribosomal RNA promoter via four high mobility group box motifs and activates transcription by interacting with other factors, including SL1 and RNA polymerase I (32,41). UBF1 contains a dimerization domain at its amino terminus and an acidic carboxyl-terminal tail composed of two uninterrupted tracts of glutamic and aspartic acid residues that are 21 and 18 amino acids long (42). At present, we do not know whether UBF/NOR-90 interacts with CENP-C directly or indirectly. Our inability to demonstrate a direct interaction using partially purified GST-C635-943 and UBF may reflect, in part, our lack of success in accurately replicating the conditions that permitted binding in the affinity chromatography experiments. For example, unknown post-translational modifications of UBF, not present on in vitro translated UBF, may be required for interactions with CENP-B. Additionally, while the functional significance of the naturally occurring deletion variant NOR-90 is not clear, it, as well as UBF, may be important for CENP-C interactions. Alternatively, these results may suggest that CENP-C and UBF/NOR-90 do not interact directly, but rather require an additional protein(s) not detectable on our silver-stained gels. If the interaction of UBF/NOR-90 with CENP-C is direct, we suspect that binding is unlikely to involve nonspecific interactions with the UBF1 acidic tail. Nuclei contain many polypeptides with acidic regions that rival or exceed that of UBF/NOR-90 (43). However, only the UBF/NOR-90 doublet was observed to bind to the affinity column baited with the carboxyl-terminal third of CENP-C. Furthermore, we have failed to detect any interaction between CENP-C and the acidic centromere protein CENP-B (5) either by affinity chromatography or in the yeast two-hybrid interaction assay. 3 CENP-C is not the first centromere protein that has been found to interact with a non-centromere protein. The centromere-associated human autoantigen CENP-F was also cloned under another name (mitosin) based on its ability to bind a portion of the retinoblastoma protein in vitro (44). The biological significance of this interaction remains unclear, and the function of CENP-F is unknown. In addition to CENP-F, the retinoblastoma protein has also been found to interact with UBF (45,46), providing an indirect link between a centromere protein and a nucleolar transcription factor. Here we demonstrate a direct biochemical link between CENP-C and UBF.
Another intriguing recent result also strengthens the link between centromeres and nucleoli. A novel nucleolar protein from rat liver called NAP57 was found to co-immunoprecipitate with Nopp140, a shuttling nucleolar phosphoprotein (47). NAP57 shows striking similarity (71% identity and 85% homology) over ϳ82% of its length to the product of the essential budding yeast gene CBF5. Cbf5p is a low affinity centromerebinding protein that interacts genetically with CBF3, a multiprotein, high affinity centromere-binding complex in Saccharomyces cerevisiae (48). Interestingly, Cbf5p is reported to localize to the nucleolus in yeast (47). It has been proposed that NAP57 and its highly conserved homologs may serve as chaperones for newly made ribosomal components (47).
Compelling evidence for an intimate relationship between interphase centromeres and nucleoli comes from studies using isolated human nucleoli (49). Immunofluorescence and immunoelectron microscopy using anti-centromere antibodies from scleroderma patients unambiguously placed centromeric autoantigens within the chromatin surrounding and, at times, embedded in isolated nucleoli. In addition, specific centromere proteins could be identified in immunoblots of protein extracts from isolated nucleoli using human autoimmune sera. A 140-kDa protein recognized by such sera in a nucleolar extract was enriched relative to an equivalent protein load of nuclear extract (see Fig. 7 in Ref. 49). The speculation that this protein is CENP-C appears reasonable considering that CENP-C is the only autoantigen of this size known to be recognized by human anti-centromere antibodies. The results from this study suggest that the presence of CENP-C in isolated nucleoli may be mediated by specific interactions with the nucleolar transcription factor UBF/NOR-90.
The finding that centromere proteins can be localized ultrastructurally and biochemically to nucleoli is relevant for considering mechanisms of autoimmunity. The presence of circulating autoantibodies that are diverse, yet highly specific and unique for a particular disease is the hallmark of many human autoimmune disorders. This has led to the hypothesis that the immune response in systemic autoimmunity is antigen-driven, a mechanism most easily understood if the antigens in question reside together at some point in a common subcellular particle that serves as the dominant autoantigen in the disease (50). The antigens recognized by the immune system in scleroderma spectrum disease all reside, at least transiently, in the nucleolus (49,51). Our demonstration that CENP-C physically interacts either directly or indirectly with UBF/NOR-90 provides a possible biochemical link between the anti-centromere and anti-nucleolar antibodies that characterize scleroderma spectrum disease.
The biological significance of a biochemical interaction between a kinetochore protein and a nucleolar protein is unclear. It is possible that in vivo such an interaction could serve to regulate the function of CENP-C during interphase by sequestering the protein and its bound centromeric DNA, UBF/NOR-90, or both. Alternatively, the association of CENP-C with nucleoli could play an architectural role in the organization of the interphase nucleus. In any case, the observation that only a subset of interphase centromeres is juxtaposed with nucleoli at any point during interphase suggests that such an interaction must be subject to control mechanisms that are not currently understood. The theme of biochemical associations between centromeres and nucleoli promises to lead to interesting insights into novel aspects of centromere function during interphase.