The yeast CDP1 gene encodes a triple-helical DNA-binding protein

Abstract

The formation of triple-helical DNA has been implicated in several cellular processes, including transcription, replication and recombination. While there is no direct evidence for triplexes in vivo, cellular proteins that specifically recognize triplex DNA have been described. Using a purine-motif triplex probe and southwestern library screening, we isolated five independent clones expressing the same C-terminal 210 amino acids of the Saccharomyces cerevisiae protein Cdp1p fused with β-galactosidase. In electrophoretic mobility shift assays, recombinant Cdp1pΔ1-867 bound Pu-motif triplex DNAs with high affinity (K_d ~5 nM) and bound Py-motif triplex, duplex and single-stranded DNAs with far lower affinity (0.5–5.0 µM). Genetic analyses revealed that the CDP1 gene product was required for proper chromosome segregation. The possible involvement of triplex DNA in this process is discussed.

INTRODUCTION

Triple-helical (triplex) DNA is a structure characterized by a third pyrimidine-rich (Py triplex) or purine-rich (Pu triplex) DNA strand located within the major groove of a homopurine/homopyrimidine stretch of duplex DNA (1). Stable interaction of the third strand is achieved through either specific Hoogsteen (Py triplex) or reverse Hoogsteen (Pu triplex) hydrogen bonding with the homopurine strand of the duplex. Preferred base triplets include T*AT and C⁺*GC in the pyrimidine motif and G*GC, A*AT and T*AT in the purine motif. Triplexes can be either intermolecular in nature, where the third strand originates from a separate DNA molecule, or intramolecular, where the third strand originates from the same DNA molecule as its duplex acceptor (1,2). Intramolecular triplexes are postulated to form in vivo under suitable conditions (such as high torsional stress) and their involvement has been implicated in several cellular processes, including transcription, replication and recombination (1,2).

If triplexes were to arise in vivo, either as a required intermediate or as an undesired side-product of a necessary process, then it is reasonable to assume that cellular proteins that specifically recognize this form of DNA might also exist. To date, several examples of triplex-binding proteins (3BPs) have been described in the literature. These include two reports of similar human proteins that exhibit binding specificity for Py triplex DNAs (3,4), our finding of several human proteins that specifically recognize a Pu triplex (5) and evidence that the Drosophila GAGA factor can tolerate binding to Py triplexes but not Pu triplexes containing a (GA·TC)₂₂ sequence (6). More recently, we purified the primary Pu triplex-binding activity in Saccharomyces cerevisiae extracts and identified it as Stm1p, a protein thought to be involved in mitosis (7).

To better understand the biological roles of 3BPs in yeast, we undertook to identify additional genes that encode these proteins. Using southwestern methods, in which a membrane-bound protein is identified by its ability to bind a specific labeled nucleic acid, we screened a S.cerevisiae genomic library with a Pu triplex probe to identify clones expressing 3BPs. Here we describe the cloning of one yeast 3BP-encoding gene, the centromere binding factor 1-dependent protein 1 gene, CDP1, which encodes a protein involved in proper nuclear division and chromosome segregation (8).

MATERIALS AND METHODS

Oligonucleotides and DNA probes

The sequences of oligodeoxyribonucleotides used in this study are presented in Table 1. Oligonucleotides whose names begin with a ‘P’ or whose sequence begins with a ‘P∼’ contained a C2-psoralen moiety attached to their 5′-termini. Duplex and triplex probes were made essentially as previously described (5). The structures of the longest, cross-linked Pu triplex (X-Pu-T19) and Py triplex (X-Py-T17) probes are shown in Figure 1. The ‘X’ refers to a species containing a psoralen photo-cross-link and ‘T19’ refers to a 19-base-triplet triplex. Also shown is a schematic representation of the triplex and duplex probes used in southwestern library screening (Fig. 1B). These were composed of either three tandem cross-linked Pu triplexes (X-Pu-T19-3) or the corresponding duplexes (Pu-D-3).

Table 1. Oligonucleotides.

Name	Sequence (5′→3′)a
Triplex-forming oligonucleotides, purine motif
PODN 1	P∼TGGGTGGGGTGGGGTGGGT
ODN 1	TGGGTGGGGTGGGGTGGGT
PODN 1 (15)	P∼TGGGTGGGGTGGGGT
PODN 1 (14)	P∼TGGGTGGGGTGGGG
PODN 1 (12)	P∼TGGGTGGGGTGG
ODN 2	AGGGAGGGGAGGGGAGGGA
Triplex-forming oligonucleotides, pyrimidine motif
PODN 3	P∼TTTCTTTTTCTTTTTTT
Duplex probe oligonucleotides, purine motif
Pu-CT	AGCTATCCCTCCCCTCCCCTCCCTTAGGA
Pu-GA	AGCTTCCTAAGGGAGGGGAGGGGAGGGAT
Duplex probe oligonucleotides, pyrimidine motif
Py-AG	GTGCAGATCTAAAGAAAAAGAAAAAAA
Py-TC	AGCTTTTTTTCTTTTTCTTTAGATCTGCACTGCA

^aP∼ refers to a psoralen-hexyl moiety.

Figure 1.

Schematics of the triplex DNA probes used for characterizing yeast 3BPs. (A) Single Pu triplex probe used in EMSAs. The X-Pu-T19 probe is composed of the Pu duplex and a covalently attached G-rich triplex-forming oligonucleotide PODN 1. P∼, trimethylpsoralen moiety on the 5′-end of PODN 1, which allows generation of photo-cross-linked triplexes. The psoralen cross-linking site (5′-TA-3′) is indicated by an X. Boxed nucleotides are those incorporated by 3′-end filling. (B) Tandem Pu triplex probe (X-Pu-T19-3) used to screen the yeast genomic expression library. Open boxes indicate locations of Pu triplex DNA. Black boxes, psoralen cross-linking sites. H, HindIII sites. (C) Cross-linked Py triplex probe (X-Py-T17) composed of the Py duplex and a covalently attached T-rich triplex-forming oligonucleotide PODN 3.

Southwestern library screening

Southwestern screening of a S.cerevisiae strain YNN 295 genomic library (Clontech, Palo Alto, CA) was performed at 4°C in continuous gentle agitation as follows. Replica filters were treated for 10 min in buffer A (25 mM HEPES–Na⁺ pH 7.9, 50 mM KCl, 10% glycerol, 1 mM dithiothreitol) with 6 M guanidine–HCl and renatured by seven sequential 1:2 dilutions with buffer A, each lasting 5 min. The filters were blocked in 5% non-fat dry milk in buffer A for 30 min followed by 0.25% non-fat dry milk in buffer A for 5 min. Binding was performed in buffer A containing 10 µg/ml salmon sperm DNA, 10 µg/ml herring DNA and 3′-end-labeled Pu-D-3 or X-Pu-T19-3 probes (∼10⁶ c.p.m./ml) for 60 min. Probed filters were washed four times with buffer A for a total of 30 min and then dried and autoradiographed. After three rounds of selection, positive clones (i.e. those specifically recognized by triplex but not duplex probe) were expanded and the phage DNA was directly sequenced.

Subcloning, recombinant protein synthesis and purification

The DNA fragment encoding the C-terminal 210 amino acids of Cdp1p (630 bp) was generated by PCR and cloned into the pET-23a(+) expression vector (Novagen, Madison, WI) in frame with an N-terminal T7 epitope tag (MASMTGGQQMGRGS, T7 epitope tag underlined) and a C-terminal oligohistidine tag (LEHHHHHH, histidine tag underlined). Transformed Escherichia coli BL21(DE3)pLys^S was used to express high levels of the recombinant protein upon induction of T7 RNA polymerase with isopropyl β-d thiogalactopyranoside. Cell extract preparation and protein purification by immobilized metal affinity chromatography were performed by following standard protocols (9). Protein purification was verified by sodium dodecylsulfate–polyacrylamide gel electrophoresis (SDS–PAGE).

Electrophoretic mobility shift assay (EMSA)

To effect triplex binding, purified recombinant Cdp1pΔ1-867 protein (BC1000/100 fraction) was incubated for 15 min at 24°C in a 20 µl vol containing buffer A, 1 mM MgCl₂, 5 µg of bovine serum albumin and 0.5 nM each 3′-end-labeled single duplex or triplex DNA probes. Note that carrier DNAs were not required in the binding reaction, given the lack of non-specific DNA-binding proteins in our purified Cdp1pΔ1-867 preparation. Protein–DNA complexes were resolved by electrophoresis for 2 h at 150 V through a 5% polyacrylamide gel containing 45 mM Tris and 45 mM boric acid. Complex formation was quantitated by densitometry. In competition experiments, additional unlabeled competitor DNAs were present in the binding reactions, as indicated in the figure legends.

RESULTS

Isolation of clones encoding a yeast 3BP

Previously we had found by southwestern blotting that multiple 3BPs exist in HeLa cell extracts (5). Similar results were obtained with yeast whole cell extracts (data not shown). To identify the yeast genes that encode possible 3BPs, southwestern blotting was used to screen a genomic library consisting of mechanically sheared S.cerevisiae DNA inserted in the inducible expression bacteriophage λgt11. Starting with 400 000 recombinant phages, five independent clones were isolated following three rounds of selection with the X-Pu-T19-3 probe. These selected clones encoded triplex-specific binding proteins, as shown by their strong binding to a triplex probe and their inability to bind to the corresponding duplex probe (Fig. 2A). These clones were directly sequenced using the λgt11 forward and reverse primers. As shown in Figure 2B, the inserts in these five clones were found to contain the same 5′-terminus, identified as the EcoRI site at nucleotide 2602 in the yeast open reading frame AOE1045 (10), although each clone had a different 3′-terminus. All five clones contained the open reading frame corresponding to the C-terminal 210 amino acids of the 1077 amino acid S.cerevisiae protein Cdp1p (8) fused in frame with β-galactosidase. This C-terminal amino acid sequence is shown in Figure 2C. Noteworthy in this sequence is the high density of charged residues (59 acidic and 41 basic of 210) and the fact that the sequence was not predominately positively charged, as might be expected for a polypeptide that binds highly negatively charged triplex DNA. Acidic and basic residues alternated along this sequence, with short charged patches (1–2 amino acids) predominating in the first half and larger clusters (5–8 amino acids) predominating towards the C-terminus. Only the stretch of amino acids from 122 to 145 contained mostly basic residues (one acidic and 13 basic of 24). It would be interesting if this region alone was sufficient to confer triplex-specific binding to a protein.

Figure 2.

Isolation of a gene encoding a yeast 3BP. (A) Autoradiograms of nitrocellulose filters onto which bacteriophage plaques from a twice-screened λgt11-S.cerevisiae genomic library were adsorbed and probed with either tandem triplex (X-Pu-T19-3, left) or duplex (Pu-D-3, right) probes. (B) Schematic of inserts in bacteriophage clones isolated after three screenings with the X-Pu-T19-3 probe. Clone name is indicated on the left; insert length (kb) is indicated on the right. Shaded box, region of insert encoding the C-terminal 210 amino acids of yeast Cdp1p. (C) Amino acid sequence of the C-terminal 210 amino acids of Cdp1p.

Triplex binding by recombinant Cdp1p

To demonstrate triplex-specific binding by the C-terminal portion of Cdp1p, we cloned the corresponding sequence into a bacterial expression vector in frame with an N-terminal T7 epitope tag and a C-terminal oligohistidine tail. This construct was used to express large amounts of a recombinant protein, Cdp1pΔ1-867, which was purified to homogeneity by immobilized metal affinity chromatography following a standard protocol (9). Purified Cdp1pΔ1-867 was incubated with either labeled triplex or duplex probes under conditions allowing protein binding. The resulting protein–probe complexes were resolved by non-denaturing PAGE and visualized by autoradiography.

Autoradiograms of Cdp1pΔ1-867 binding to cross-linked Pu triplex and Py triplex probes are shown in Figure 3A and B, respectively. As shown, Cdp1pΔ1-867 formed a single shifted species with a relative electrophoretic mobility (R_f) of 0.35 compared to the unbound triplex when bound to the X-Pu-T19 probe. Detectable complex formation was observed at 80 nM Cdp1pΔ1-867 (Fig. 3A, lane 5). A species with identical mobility was observed with the X-Py-T17 probe, although only at higher Cdp1pΔ1-867 concentrations. Neither duplex probe demonstrated detectable protein binding, even at the highest Cdp1pΔ1-867 concentrations tested (1.2 µM, lane 14). Interestingly, both triplex probes demonstrated evidence for metastable Cdp1pΔ1-867:DNA binding. This was especially evident for the X-Pu-T19 probe, which demonstrated a smear of labeled probe throughout the region R_f = 0.71–1.0 (asterisks in Fig. 3A and B) when intermediate concentrations of Cdp1pΔ1-867 were present in the binding reaction (Fig. 3A, lanes 4–11). A similar result was observed with the X-Py-T17 probe, although most of the radioactivity appeared localized near the unbound probe (Fig. 3B, lanes 5–9). Thus, the stable complexes observed after gel electrophoresis may not be the only species that can form between Cdp1pΔ1-867 and triplex DNAs.

Figure 3.

Recombinant Cdp1pΔ1-867 bound triplex but not duplex DNA. EMSAs were performed on reaction mixtures containing 0.5 nM labeled probe and increasing concentrations of Cdp1pΔ1-867 as follows: 0 nM (lanes 1 and 13), 10 nM (lane 2), 20 nM (lane 3), 40 nM (lane 4), 80 nM (lane 5), 120 nM (lane 6), 160 nM (lane 7), 200 nM (lane 8), 240 nM (lane 9), 300 nM (lane 10), 600 nM (lane 11) and 1.2 µM (lanes 12 and 14). (A) Reaction mixtures contained either X-Pu-T19 (lanes 1–12) or Pu-D (lanes 13 and 14) probes. The locations of the gel well (W), the primary protein–Pu triplex complex (C), metastable protein–Pu triplex complexes (*), the free X-Pu triplex probe (T) and the free Pu duplex probe (D) are indicated on the left. (B) Reaction mixtures contained either X-Py-T17 (lanes 1–12) or Py-D (lanes 13 and 14) probes. The locations of the gel well (W), the primary protein–Py triplex complex (C), metastable protein–Py triplex complexes (*), the free X-Py triplex probe (T) and the free Py duplex probe (D) are indicated on the left. (C and D) Graphical representations of the data shown in (A) and (B). (C) Shown are the percentages of total labeled probe present in protein–Pu triplex (solid squares) and protein–Py triplex (inverted solid triangles) complexes as a function of Cdp1pΔ1-867 concentration. (D) Shown are the percentages of total labeled probe present as unbound X-Pu-T19 (solid circles) or X-Py-T17 (open squares) as a function of Cdp1pΔ1-867 concentration.

Recombinant Cdp1pΔ1-867:triplex DNA complex formation was quantitated by densitometry and the data are presented graphically in Figure 3C. As shown here, the binding curves for both Pu and Py triplex DNAs were both generally sigmoidal, with 50% complex formation occurring at Cdp1pΔ1-867 concentrations of 450 and 1040 nM, respectively. However, a significant fraction of triplex probe did not migrate as either free or bound, thus complicating the analysis. As an alternative assay, we measured the loss of free triplex probe that occurred with increasing Cdp1pΔ1-867 concentration (Fig. 3D). By this measure, 50% binding of X-Pu-T19 and X-Py-T17 probes occurred at 64 and 130 nM, respectively. Thus, it is likely the binding affinity of Cdp1pΔ1-867 for these DNAs is greater in solution than was demonstrated by the survival of their respective protein–triplex complexes after gel electrophoresis.

To better characterize the binding site recognized by Cdp1pΔ1-867, we next investigated Cdp1pΔ1-867 binding to different-length Pu triplex DNAs. Previously we had found that a 12 nucleotide G-rich oligonucleotide was optimal for Pu triplex formation under our standard reaction conditions (11,12), although longer Pu triplexes were apparently more stable after non-denaturing gel electrophoresis (5). Sequential 3′ deletions of the triplex-forming oligonucleotide PODN 1 (Table 1) were synthesized and used with the standard Pu duplex to make cross-linked Pu triplex probes 19, 15, 14 and 12 base triplets long. These probes were incubated with increasing concentrations of Cdp1pΔ1-867 under our standard binding conditions and the resulting complexes were resolved by non-denaturing PAGE and visualized by autoradiography. As seen in Figure 4, the different free triplex probes (T) migrated with only slightly different relative electrophoretic mobilities, with the longest triplex (X-Pu-T19) having an R_f = 0.84 compared with the free duplex probe and the shortest (X-Pu-T12) having an R_f = 0.90. This phenomenon may reflect the considerable duplex nature present in both our standard Pu triplex probe (14 of 33 bp) and the shortest triplex investigated here (21 of 33 bp). Only a single protein–DNA complex was observed for each of these triplex probes at the highest Cdp1pΔ1-867 concentration tested (600 nM), although evidence of metastable complex formation was also found for each. Interestingly, no significant differences in electrophoretic mobility were observed among these complexes, suggesting that their mobilities were primarily dictated by their protein components and not their nucleic acid components. More importantly, the amount of complex formed was directly proportional to the length of triplex present. Thus, the X-Pu-T19 probe demonstrated >74% complex formation in the presence of 600 nM Cdp1pΔ1-867, whereas the X-Pu-T12 probe demonstrated <25% complex formation. This was not the result of different triplex DNA stabilities, as each probe contained a psoralen cross-link. Rather, these data suggest that Cdp1pΔ1-867 binds more stably to a longer Pu triplex. Note that these data do not provide a great deal of information regarding the stoichiometry of the Cdp1pΔ1-867:Pu triplex complex. It may be argued that because there is no significant difference between the mobilities of the different protein–triplex complexes, then they must have similar protein masses. However, we cannot be sure whether this is the result of a single or constant number of protein molecules being bound to the DNA. The existence of metastable species argues for the latter hypothesis, because they may represent species containing suboptimal amounts of Cdp1pΔ1-867.

Figure 4.

Cdp1pΔ1-867 binding to different-length Pu triplexes. EMSAs were performed on reaction mixtures containing 0.5 nM labeled probe; either X-Pu-T19 (lanes 1–3), X-Pu-T15 (lanes 4–6), X-Pu-T14 (lanes 7–9) or X-Pu-T12 (lanes 10–12); and increasing concentrations of Cdp1pΔ1-867 as follows: 0 (lanes 1, 4, 7 and 10), 160 (lanes 2, 5, 8 and 11) and 600 nM (lanes 3, 6, 9 and 12). The locations of the gel well (W), protein–Pu triplex complexes (C), metastable protein–Pu triplex complexes (*), the free X-Pu triplex probes (T) and the free Pu duplex probe (D) are indicated on the left.

Binding specificity of recombinant Cdp1p

To better characterize the binding specificity of Cdp1pΔ1-867, we investigated its binding to a variety of nucleic acids, including different triplex, duplex and single-stranded DNAs. Since we had evidence that some Cdp1pΔ1-867 complexes may not be completely stable during gel electrophoresis, we chose to examine relative binding affinities using competition binding experiments. In each, constant concentrations of labeled X-Pu-T19 (0.5 nM) and Cdp1pΔ1-867 (120 nM) were incubated with increasing concentrations of unlabeled competitor (15–5000 nM) before analysis of the radioactive protein–DNA complexes by EMSA. The sequences and structures of the competitor DNAs are shown in Table 1 and Figure 1. Quantitation from a series of competition binding experiments is shown in Figure 5 and the competitor concentrations necessary to reduce Cdp1pΔ1-867:X-Pu-T19 complex formation 50% (EC₅₀) is provided in Table 2. As shown, a Pu triplex with a G/A-rich third strand (ODN 2) demonstrated the highest apparent affinity for Cdp1pΔ1-867, 4-fold better than for the equivalent Pu triplex with a G/T-rich third strand (ODN 1). Interestingly, a covalent Pu triplex (X-Pu-T19) was a 4-fold better competitor than its non-covalent counterpart (Pu-T19). This raises the interesting question of whether these probes have intact triplexes throughout their entire lengths. We already know that the 3′-end of the triplex third strand is more involved in Pu triplex formation than its 5′-end (11,12). Thus, having a psoralen cross-link on the weaker, 5′-end could drive the entire length into a proper triplex conformation, which is preferentially recognized by Cdp1pΔ1-867. Consistent with our Cdp1pΔ1-867 titration experiment (Fig. 3), Cdp1pΔ1-867 demonstrated 25-fold weaker binding to a Py triplex than to a Pu triplex and had no apparent affinity for a Pu duplex within the range tested. Similar results were found for the alternating homopolymer poly(dI-dC)·poly(dI-dC) (M. Musso, unpublished observation), suggesting that Cdp1pΔ1-867 generally has a very low affinity for duplex DNA. Regarding single-stranded DNAs, Cdp1pΔ1-867 bound oligonucleotides with the following preference order: ODN 2 > Pu-GA > ODN 1 > Pu-CT. This preference order exactly mirrors the propensity of these oligonucleotides to form G-quartet-containing structures under our binding conditions (13,14), suggesting that Cdp1pΔ1-867 does recognize G4 DNAs. However, these short, intramolecular quadruplexes are not the best binding sites for Cdp1pΔ1-867, given its low apparent affinity for these DNAs.

Figure 5.

Binding specificity of Cdp1pΔ1-867 for different DNAs. EMSAs were performed with 0.5 nM labeled X-Pu-T19 probe, 120 nM Cdp1pΔ1-867, and concentrations of unlabeled competitor DNAs as indicated. These included the covalent Pu triplex X-Pu-T19 (solid squares), the non-covalent Pu triplexes Pu-T19 (solid triangles) and Pu-T19(GA) (solid diamonds), the covalent Py triplex X-Py-T17 (solid circles), the Pu duplex (inverted solid triangles) and the oligonucleotides ODN 2 (open diamonds), Pu-GA (inverted open triangles), ODN 1 (open triangles) and Pu-CT (open squares).

Table 2. Relative binding affinity of Cdp1pΔ1-867 for different DNAs.

DNA	EC50a (M)
Triplex DNA
Pu-T19(GA)b	5.4 × 10–9
X-Pu-T19	2.1 × 10–8
Pu-T19	8.9 × 10–8
X-Py-T17	45% at 0.5 µMc
Duplex DNA
Pu-D	1% at 0.5 µM
Single-stranded DNA
ODN 2	1.7 × 10–6
Pu-GA	3.9 × 10–6
ODN 1	27% at 5 µM
Pu-CT	8% at 5 µM

^aEC₅₀ refers to the concentration of competitor DNA that reduces complex formation 50% when 0.5 nM labeled X-Pu-T19 probe and 120 nM Cdp1pΔ1-867 were present in an EMSA binding reaction.

^bPu triplex composed of the Pu duplex and the triplex-forming oligonucleotide ODN 2.

^cPercentage inhibition of complex formation at maximal competitor DNA concentration tested.

DISCUSSION

Using EMSA, we have obtained evidence that the yeast S.cerevisiae contains at least two different 3BPs, one of which is the product of the STM1 gene (7). To identify other genes encoding 3BPs in yeast, we performed a southwestern screening of a S.cerevisiae genomic expression library by using a tandem trimer of psoralen cross-linked Pu triplexes as a probe. After three rounds of screening, we identified five independent clones containing the C-terminal 210 amino acids of the CDP1 gene. Interestingly, no other yeast genes were found, including the previously biochemically identified 3BP gene STM1. This is somewhat surprising, given that Stm1p fulfills all the criteria necessary for a suitable southwestern candidate, i.e. Stm1p renatures efficiently and functionally and N-terminal Stm1p fusion proteins bind Pu triplexes with high affinity and specificity (7). This failure to identify other yeast 3BP genes may reflect limitations of this particular library, e.g. triplex-binding domains that span an intron would not have been identified. It is more likely that this approach is not optimal for identifying Pu triplex-binding proteins, given our limited success in identifying human 3BPs by southwestern library screening (P. Ciotti, M.W. Van Dyke, G. Bianchi-Scarrà and M. Musso, manuscript submitted).

Is Cdp1p a bona fide purine-motif triplex-binding protein? Southwestern screening consistently identified only a small portion of the Cdp1 protein. This may reflect limitations on the renaturation of larger Cdp1p polypeptides into an active, triplex-binding conformation or the requirement of auxiliary proteins for this purpose. Cdp1pΔ1-867 bound a Pu triplex with at least a thousand-fold greater affinity than its corresponding Pu duplex, suggesting that it is a specific Pu triplex-binding protein. However, its apparent binding affinity was almost two orders of magnitude weaker than that found for another yeast 3BP, Stm1p, (7) indicating that Cdp1pΔ1-867 is not a high affinity 3BP. This relatively low affinity may also explain our difficulty in observing an electrophoretic mobility shift corresponding to a complex containing Cdp1pΔ1-867 and a non-covalently bound Pu triplex. However, based on our prior studies with human triplex-binding proteins and Stm1p (5,7), we are confident the DNA species bound by Cdp1pΔ1-867 during electrophoresis is an intact triplex. We were unable to generate full-length Cdp1p in vitro or to successfully express it in either Escherichia coli or the yeast Pischia pastoris (M. Musso, unpublished observation). Thus, we are unable to conclusively state whether the native protein has any affinity for Pu triplexes. However, Cdp1p is not one of the yeast 3BPs we had previously identified by EMSA (7), since a cdp1Δ strain extract yielded an identical pattern as that found with a wild-type yeast extract (M. Musso, unpublished observation).

CDP1, also referred to as AOE1045, CTR9, O0458 and YOL145C, is a large open reading frame lying on S.cerevisiae chromosome XV encoding a protein of 1077 amino acids (8,10). A PROSITE (15) structural analysis of this sequence revealed the following domains: a motif A (P-loop) ATP/GTP binding site (amino acids 65–72), a five leucine zipper (amino acids 714–742), several tetratricopeptide repeats (amino acids 56–89, 183–251, 338–405, 462–534 and 680–801) and a bipartite nuclear localization signal (amino acids 992–1009). This is shown schematically in Figure 6. The CDP1 gene was initially identified in a screen for mutants that could be propagated in the presence but not the absence of the S.cerevisiae centromere binding factor I (CBF1) gene (8). In a wild-type background, deletion of CDP1 resulted in a significantly reduced doubling time at 30°C and almost no growth at 37°C, with most cells exhibiting a large single- or double-budded morphology indicative of an arrest in G₂/M (16; M. Musso, unpublished observation). Mutant cdp1Δ cells shifted to the non-permissive temperature exhibited a very high percentage of anucleated (9%) and multinucleated (35%) cells, 5- and 60-fold greater than in wild-type cells, respectively (8). CDP1 mutants were also found to have a defect in chromosome segregation, with a frequency of chromosome fragment loss 60–230-fold higher than in wild-type cells (8). CDP1 has also been implicated in the control of cell cycle-dependent gene expression in late G₁ and has been found to physically associate with Paf1p and Cdc73p, proteins known to be associated with RNA polymerase II (17). Such a dual role in mitosis and transcriptional regulation is not unknown in yeast. For example, Cbf1p is both a centromere binding protein required for mitosis and a transcription factor regulating methionine biosynthesis genes (18). CDP1 transcripts were found by serial analysis of gene expression (SAGE) to be present at very low levels in both log phase and growth-arrested yeast (19) and no significant pattern of gene expression has been identified by DNA microarray analyses (20). A BLASTP search (21) of the non-redundant protein databases found closely related proteins in other organisms, including the Schizosaccharomyces pombe Tpr1 protein, which is involved in complementing K⁺ transport-defective yeast; the Caenorhabditis elegans hypothetical protein B0464.2; the murine nuclear phosphoprotein p150TSP and the human KIAA0155 open reading frame. The biological roles of these related proteins are not known, but their pronounced similarity suggests that they have conserved, important cellular functions.

Figure 6.

Schematic representation of S.cerevisiae Cdp1p. Functional motifs defined by PROSITE are indicated. Shaded boxes, tetratricopeptide repeats.

What are the possible roles of triplexes and 3BPs in vivo? Cdp1p is a nuclear protein that combines a purine nucleotide binding pocket, several protein–protein interaction domains and a highly charged triplex-binding domain. Its genetic relationship with a centromere-binding protein and its involvement in chromosome segregation are consistent with Cdp1p DNA binding. Likewise, the triplex-tolerant Drosophila GAGA protein preferentially associates with centromeric heterochromatin in mitotic chromosomes and some GAGA mutants have mitotic defects consistent with a role in chromosome condensation and/or segregation (22,23). Most interestingly, antibodies raised against Py triplex DNA were found to recognize centromeric regions of isolated murine metaphase chromosomes (24) and these antibodies have a significant effect on cell growth but not cell viability when incorporated into synchronized cells during late S/early G₂ (25). Taken together, these findings support a model in which transmolecular triplexes are involved in chromatin condensation and decondensation, with condensation being facilitated by cooperative 3BP binding and decondensation promoted by 3BP inactivation. By transmolecular triplexes we mean triple-helical DNA structures in which the third strand originates from a distal portion of the same DNA molecule as its duplex target. Formation of transmolecular triplexes would result in the condensation of DNA, a process that could be facilitated by 3BPs. Reversal of this process would then require dissolution of the 3BP-transmolecular triplex complexes by a change in 3BP-DNA binding properties, possibly through covalent modification of the 3BP itself. In the case of triplex-specific antibodies, this change may not be possible, thus resulting in the observed delay in the cell cycle at mitosis. Further validation of this hypothesis must await additional experiments, e.g. indirect immunofluorescence staining with anti-Cdp1p antibodies and confocal microscopy to determine the quantity and cellular location of this 3BP throughout the cell cycle.