ABSTRACT
We have measured the processivity of telomeric DNA extension by Euplotes aediculatus telomerase at various concentrations of the nucleotide substrates dGTP and dTTP. The maximum processivity (~3 repeats) was observed at ~100 [mu]M of each dNTP. Processivity decreased as the dNTP concentrations were reduced and, surprisingly, as the concentration of dGTP was increased. Also, the characteristic banding pattern generated by telomerase extension of DNA primers shifted in response to changes in dGTP concentration. One pattern with 8 nt periodicity was predominant at dGTP concentrations <= 16 [mu]M, while at >= 250 [mu]M an 8 nt repeat pattern out-of-phase with the first was observed; at intermediate concentrations the two patterns coexisted. We propose that two different segments of the RNA subunit can serve as the template for repeat synthesis; nt 42-49 at low dGTP concentrations and nt 36-43 at high dGTP concentrations. An alternative model for the low dGTP pattern involves an internal pause site but no pause at the end of the template and is, therefore, considered less likely. Because the effects of dGTP on processivity and banding pattern appear to be distinct from nucleotide binding in the polymerase active site, we propose a second dGTP binding site involved in template selection and processivity.
In most eukaryotes, the DNA sequence at the chromosome termini consists of an array of simple tandem repeats. This repetitive DNA and associated proteins make up a terminal structure called the telomere (for reviews see 1 ,2 ). The telomere provides a protective cap to prevent degradation or terminal fusion of the chromosomes (3 ,4 ). The DNA repeat sequence varies between organisms, but is often 6-8 nt long and rich in G and T nucleotides in the strand running 5' to 3' towards the end of the chromosome (5 ). This holds true for organisms as diverse as ciliated protozoa (T4G4 or T2G4), vertebrates (T2AG3) and plants [T2(T/A)G3 or T3AG3], although exceptions have been found (6 ). These repetitive sequences are maintained during DNA replication by the enzyme telomerase, which can also synthesize new telomeres after chromosome breakage (7 ,8 ) or during programmed development of the macronucleus of the hypotrichous ciliates (for reviews see refs 9 ,10 ).
Telomerase is a ribonucleoprotein enzyme that contains a single RNA subunit and associated protein components. Two protein components of telomerase have been reported for the holotrichous ciliate Tetrahymena thermophila (11 ), one of which has homologs in mammals (12 ,13 ). Two unrelated protein components have been found in the hypotrichous ciliate Euplotes aediculatus (14 ), and one of these (p123) is a reverse-transcriptase related protein proposed to serve as the catalytic subunit (15 ). The yeast homolog of p123, Est2p (Ever Shorter Telomeres) (16 ), is essential for telomerase activity in vivo and in vitro (15 ).
The sequence of the RNA moiety has been determined for many organisms, and each contains a region complementary to the telomeric repeat (17 ). Modifications of the telomere-complementary sequence produce correspondingly altered telomeric repeats, demonstrating that this RNA provides the template for telomere synthesis (18 -20 ). The region of telomere complementarity is ~1.5-2 repeats in length. However, only a segment coding for one repeat appears to serve as the actual template. The adjacent sequence has been shown to align the 3'-end of DNA substrates by base-pairing, so that correct repeats are synthesized (21 ).
The telomerase RNA secondary structure was determined for the ciliates by phylogenetic analysis (22 ,23 ), and while the primary sequence of the RNA varies greatly between species, the overall architecture is very similar. The telomerase RNA from E.aediculatus, studied herein, is shown schematically (Fig. 1 ) with a primer positioned on the template (shown in bold). Also shown is the UGUCA motif that is conserved among the ciliates and is located 2 nt 5' of the template (23 ).
Euplotes aediculatus was grown as described (27 ) under non-sterile conditions in aerated 15 gallon reactors, with Chlorogonium as the food source. Cells were collected on a 15 [mu]m Nytex filter and lysed in the presence of Nonidet P-40 non-ionic detergent. Nuclei were isolated by sucrose cushion centrifugation and a nuclear extract was prepared by Dounce homogenization as previously described (14 ).
Nuclear extracts were partially purified as previously described (25 ). Extracts were fractionated by centrifugation in a 15-40% glycerol gradient. Fractions containing telomerase were identified by a gel-shift assay for the presence of telomerase RNA, using a radiolabeled telomeric primer. Peak fractions were pooled and dialyzed against telomerase reaction buffer (50 mM Tris-HCl pH 7.5, 10 mM MgCl2, 50 mM sodium glutamate, 1 mM dithiothreitol, 10% glycerol) using a Spectra/Por CE membrane with a 100 kDa molecular weight cutoff.
Primers were prepared by standard phosphoramidite synthesis and purified by polyacrylamide gel electrophoresis as previously described (25 ). The non-telomeric competitor primer had the sequence 5'-ATTGAATGACTACGAGATGAA. Telomeric primers all contained 5-iododeoxyuridine (IU) substitutions for thymidine at positions 1 and 3, as they were previously used for cross-linking studies; the IU substitution is isosteric with thymidine and does not disrupt normal base-pairing. All of the telomeric primers are composed of full and partial T4G4 repeats. The sequence of the 21 nt reference primer was 5'-IUTIUTGGGGTTTTGGGGTTTTG. Shorter and longer primers were all altered at the 3'-end such that the 5'-end remained constant. Purified primers were 5'-end-labeled using T4 polynucleotide kinase and [[gamma]-32P]ATP (6000 Ci/mmol; NEN), then purified using Beckman G-25 TE spin columns.
Partially purified telomerase was pre-incubated with the non-telomeric primer (6.5 [mu]M), then 5'-end-labeled telomeric primer was added (0.015 [mu]M) and allowed to bind for 10 min at 25oC. Aliquots of the reaction (10 [mu]l) were mixed with an equal volume of telomerase reaction buffer containing dGTP and dTTP (or ddTTP) and specific competitor primer (5 [mu]M), then incubated at 25oC. The concentrations of the added nucleotides and the incubation times were varied in individual experiments (see below). Control reactions with no dNTP's were included where indicated. The specific competitor primer for each reaction had the same sequence as the labeled primer used in that reaction. Extension reactions were stopped by addition of 100 [mu]l of proteinase K buffer (20 mM Tris-HCl pH 7.9, 10 mM EDTA, 1% sodium dodecyl sulfate) and incubated for 3 min at 85oC. The nucleic acid component was isolated by digestion with proteinase K (80 [mu]g/ml) for 45 min at 45oC, followed by ethanol precipitation. In reactions where the dGTP and dTTP concentrations were varied independently, the proteinase K was pre-mixed with the proteinase K buffer and no 85oC denaturation was performed. Primer extension products were separated by denaturing polyacrylamide gel electrophoresis. Reaction products were visualized and quantitated using a Molecular Dynamics phosphorimager.
To verify the effectiveness of the specific competitor, pre-chased reactions were performed in which the chase and 32P-labeled primers were mixed before incubation with telomerase. To ensure that all endogenous nucleotides were removed during the telomerase purification, a control with no addition of nucleotides was included. In neither of these reactions did significant extension of labeled primer occur. The correct alignment of primers on the telomerase template was verified by substituting ddTTP for dTTP in the reaction. For each of the primers tested, extension was terminated at the first adenosine in the template, except where noted in the Results.
For in vitro telomerase assays, a characteristic pattern of product bands is seen on polyacrylamide gels corresponding to additions of complete repeats to a substrate primer. These bands are believed to arise from primer pausing and/or dissociating upon extension to the end of the template (17 ). Processive extension of DNA primers by telomerase is observed as the addition of multiple repeats without intervening dissociation of the bound primer. We have previously developed a bind-and-chase assay to measure processivity by a quantitative analysis of the telomerase banding pattern (25 ). This assay is used herein to test the effects of nucleotide concentration on processivity.
The banding pattern observed for telomerase primer extension changed as the dNTP concentration was varied. To determine if the change was due specifically to either dGTP or dTTP, one of the nucleotides was held constant at 656 [mu]M while the other was varied from 10 nM to 656 [mu]M (Fig. 4 ). Comparison of the banding pattern in the dTTP lanes to that in the dGTP lanes showed that the banding pattern was shifting in response to the dGTP concentration. The same concentration dependence seen when the nucleotides were varied together was observed in the dGTP reactions. With <= 10 [mu]M dGTP, the major pauses occurred at +1, +9, +17, etc., whereas with >164 [mu]M dGTP the pattern shifted to +7, +15, etc; intermediate concentrations gave a mix of the two patterns.
The specificity of this shift was tested by primer extension using 5 [mu]M dGTP (where the low dGTP banding pattern was expected) supplemented with concentrations of its ribo-analog (GTP) up to 1 mM (data not shown). All of these reactions gave the low dGTP banding pattern, indicating a high level of specificity for dGTP.
The effect of each dNTP on processivity was also different. Low dTTP concentrations limited the processivity primarily because partial extension products accumulated, as one would expect; these products are labeled G4, T1, T2, T3, T4 and G1 on the left of Figure 4 . Assuming correct G4T4 repeats were added (see below), G4 corresponds to the fourth G in the GGGGTTTT repeat, T1 to the first T, etc. Once the build-up of these intermediate products was minimized at concentrations >41 [mu]M dTTP, the processivity did not increase further, nor did it decrease. The dGTP concentration had a different impact on the processivity; as the concentration increased, so did the processivity, but at the two highest concentrations it began to decrease again. Decreased processivity is apparent by visual inspection of Figure 4 , where it can be seen that the intensity of the banding pattern at higher molecular weights falls off more rapidly at the two highest dGTP concentrations, 164 and 656 [mu]M. This was confirmed by quantitation as in Figure 3 . An identical effect was observed when the concentrations of dGTP and dTTP were varied together.
One possibility for the origin of the alternative banding pattern was the misalignment of the primer on the RNA template. This was tested by comparing the extension products for primers that differed at their 3'-ends (Fig. 6 ).
These primers were first used in a processive extension assay with 500 [mu]M dTTP and either 5 [mu]M dGTP (Fig. 6 A) or 500 [mu]M dGTP (Fig. 6 B). The expected low-dGTP and high-dGTP banding patterns were produced for each of the primers (arrows in Fig. 6 A and B, respectively). This result indicated that each of the different primer 3'-ends was properly aligned so that the end of the template occurred at the same location for each. This is the expected result if the banding pattern arises from either pausing or dissociation after primers are extended to the end of the template.
A second test for addition of correct repeats was to substitute ddTTP for dTTP, again under low-dGTP (Fig. 6 C) and high-dGTP (Fig. 6 D) reaction conditions. Incorporation of ddTTP causes chain termination, and will occur the first time that adenosine is encountered in the RNA template. The same pattern of dideoxynucleotide termination was observed at low dGTP as at high dGTP, revealing that the sequence being added to the primer was the same for each. The only significant deviation occurred with the 27mer in the high-dGTP reaction. The 27mer ends with the sequence -GTTT and apparently some misalignment (pairing of the T's with the template positions 45, 46 and 47) allowed the addition of -GGGGddT as well as the expected single nucleotide addition of ddT. This may be due in part to a relatively low efficiency of incorporation of ddTTP as was reported for the Tetrahymena enzyme (28 ).
These results can be explained if telomerase from Euplotes uses different regions of its RNA subunit as the template at low and high dGTP concentrations (Fig. 6 E and F). The region of the RNA used as the template under each condition can be determined from the site of the first major pause when primers of different lengths are extended (Fig. 6 A and B) and from the observation that an 8 nt repeat is added at either concentration. At low dGTP concentrations (Fig. 6 E), translocation occurs after addition of the third G in the G4T4 repeat, indicating that the 8 nt template extends from C49 to C42. At high dGTP concentrations (Fig. 6 F), translocation occurs after addition of the first G in the G4T4 repeat, indicating that the region from C43 to C36 serves as the template.
An alternative explanation for the low dGTP pattern would require pausing or dissociation prior to reaching the end of the C43-C36 template. For this model to explain the data, two conditions would have to be met. First, the addition of the fourth G, templated by C41, must be much less efficient than that of the first three Gs and second, translocation at C36 would have to be very efficient such that it did not result in the accumulation of significant product (Fig. 6 A, 2 nt below the upper arrow). Correspondingly, as the dGTP concentration is increased, the resulting increase in efficiency of G addition at position 41 would have to be accompanied by an almost exactly compensating decrease in the efficiency of translocation. (The relative efficiency of each translocation is evidenced by the total amount of primer extended beyond the corresponding pause.) The first condition is conceivable; the Km for dGTP could for some reason be especially high at the C41 template position, or the rate constant for addition of the fourth G could be lower than that for the first three. The second condition seems unlikely, however, because the processivity of Euplotes telomerase is modest (only ~2 repeats; 25 ), and a vast excess of unlabeled primer in the reaction prevents rebinding after dissociation; thus, some accumulation of product corresponding to polymerization to the end of the template is expected. Therefore, while we cannot rule out this `internal pause site' model for the alternative banding pattern, the template switching model seems simpler because it involves fewer ad hoc assumptions.
Telomerase adds repeats to DNA primers using its RNA subunit as the template. After extension to the end of the template, the primer can translocate without dissociation so that additional repeats can be added. A characteristic banding pattern with a periodicity equal to the telomeric repeat length is observed by gel electrophoresis of DNA primers extended by telomerase. Such a pattern may in principle arise from either pausing or dissociation at the end of the template prior to or during translocation, or at an internal site due to a slow polymerization step or the absence of the next cognate nucleotide.
In primer extension assays with Euplotes telomerase at different dGTP concentrations, two different banding patterns were observed (Figs 2 and 4 ). For both patterns, correct 8 nt repeats are synthesized as judged by the position of stops in the presence of ddTTP (Fig. 6 C and D). The two patterns arise from the same initial population of telomerase-bound primer (Fig. 5 ), indicating that both patterns can be generated by the same enzyme. The change in banding pattern occurs at dGTP concentrations between 40 and 160 [mu]M. This is well above the concentration required to minimize stalling within the repeat which is observed at lower nucleotide concentrations (Figs 2 and 4 , and data not shown). We therefore propose that these two banding patterns are the result of dGTP-dependent template switching, such that nucleotides C49-C42 or nucleotides C43-C36 of the RNA subunit provide the template at low and high dGTP concentrations, respectively (Fig. 7 ).
In ciliates, the telomerase template is preceded by a conserved UCUGA sequence motif (Fig. 1 ) (23 ). Based on site-directed mutagenesis of the Tetrahymena enzyme, it has been proposed that this motif serves to define the 5'-end of the template (30 ). Correspondingly, insertion or deletion of nucleotides within the Tetrahymena telomerase RNA template results in repeats that are longer or shorter, respectively (30 ,33 ). However, in Euplotes a different mechanism may contribute to definition of the template boundaries. Since the two templates are shifted by several nucleotides within the RNA sequence, it does not appear that sequences outside the template region can define both templates. Instead, Euplotes telomerase appears to have the ability to count nucleotides, as the use of either template results in the addition of a complete 8 nt repeat prior to translocation. We therefore propose that telomerase proteins help to define and maintain a fixed template size of 8 nt.
Telomerase has two distinct roles in the life cycle of the hypotrichous ciliates. One role is the maintenance of the chromosomal 3'-overhang during DNA replication by synthesis of telomeric sequences (31 ,32 ), and the second is de novo telomere synthesis during the development of the macronucleus (33 ; reviewed in 10 ). It has been suggested that telomerase may use different template regions for each of these processes (30 ). If primers are aligned exclusively by base-pairing interactions, for Euplotes telomerase there must be at least four primer-template base-pairs formed during primer alignment in order to confer a single unique alignment for each 3'-end. Comparison of the segments of telomerase RNA used for primer alignments under the high-dGTP and low-dGTP conditions reveals that they consist of the telomeric sequence, 5'-CAAAACC-3', and the distinctly non-telomeric sequence 5'-CUUACA-3', respectively. Only the former is phylogenetically conserved among hypotrichous ciliates.
This raises the possibility that the high-G and low-G templates are used for telomere maintenance and macronuclear development, respectively, due to the different base-pairing regions for primer alignment. During DNA replication, the substrate for telomerase presumably has a normal telomeric repeat sequence and may be easily aligned on the high-G template. In contrast, a comparison of the sequences of the micronuclear and macronuclear genomes revealed that there is no substantial telomere homology at the sites of telomere addition (34 ,35 ). More flexibility in the sequences that can be aligned on the low-G template may be allowed by the ability of U to form stable base-pairs with either G or A. Alternatively, telomerase protein-DNA interactions may supplant base-pairing for primer alignment on the low-G template.
Studies in the related organism Euplotes crassus have shown that extension of non-telomeric primers in vitro always begins with four dG residues, leading to the postulation of a default binding register (36 ). Extension from the default register may in fact be occurring not within the canonical template, but rather on the alternative template that is proposed herein to be used at low dGTP conditions. Trans-acting factor(s) that were specifically present in extracts from cells undergoing macronuclear development enhance the use of DNA substrates lacking 3' telomeric sequences in vitro (39 ). Perhaps one of the functions of this factor(s) is to direct telomerase to use the template observed herein at low dGTP concentrations.
The concentration of dGTP not only affected the template usage, but also had a marked effect on the processivity of Euplotes telomerase. The processivity appears to be affected by the nucleotide concentration in two distinct ways. The first is the efficiency of nucleotide incorporation within each repeat, which is expected to be related to the affinity of the active site for nucleotide binding. Indeed, in this study and others (37 ,38 ), increasing nucleotide concentrations improve repeat synthesis and longer products are observed. This is expected for any simple model in which primer extension competes with primer dissociation, with the rate of primer extension being dependent on the nucleotide concentration.
A second and novel effect was observed for dGTP in the experiments presented herein. At nucleotide concentrations above those required to prevent stalling within a repeat, dGTP appears to modulate the number of repeats added. That is, increasing the dGTP concentration first causes an increase in processivity and then at concentrations >100 [mu]M causes a decrease in processivity.
As a polymerase, telomerase clearly must bind dGTP in the active site as a substrate for primer extension, and this could influence which template is used. However, primers for which the next cognate nucleotide is T also use two different templates in a dGTP-dependent manner (Fig. 6 A and B). For these primers the active site must bind dTTP, which would prevent simultaneous binding of dGTP at the same site. These results indicate the presence of at least one additional dGTP binding site.
The mechanism by which dGTP binding at such a secondary site stimulates translocation is unknown. One possibility is that dGTP hydrolysis may be required by a telomerase associated helicase-like activity that opens primer-template base-pairs and allows translocation. Alternatively, a simple bind-and-release of dGTP could activate translocation. Distinguishing between these and other possibilities is complicated by the fact that dGTP is required for telomeric repeat addition. Future testing of non- hydrolyzable dGTP analogs may provide some support for one of these models.
It is not clear whether the effects of dGTP on the processivity and the template switching phenomena are coupled. However, translocation and template switching both require a movement of the RNA relative to the polymerase active site. Identification and further characterization of a second dGTP binding site would provide additional insight into the mechanism of primer translocation by telomerase.
We thank Loren Denton and Brent Dickinson for growing E.aediculatus cells; Cheryl Grosshans and Elaine Podell for oligonucleotide synthesis; Lara Weinstein and Stacie Froelich-Ammon for critical reading of the manuscript; Evelyn Jabri and Lara Weinstein for helpful discussions. This work was supported by NIH grant GM28039. T.R.C. is an investigator of the Howard Hughes Medical Institute and an American Cancer Society Professor.
*To whom correspondence should be addressed. Tel: +1 303 492 8606; Fax +1 303 492 6194; Email: thomas.cech@colorado.edu
+Present address: Mosaic Technologies, Inc., 1106 Commonwealth Avenue, Boston, MA 02215, USA
REFERENCES