Architecture of human telomerase RNA

Abstract

Telomerase is a unique reverse transcriptase that catalyzes the addition of telomere DNA repeats onto the 3′ ends of linear chromosomes and plays a critical role in maintaining genome stability. Unlike other reverse transcriptases, telomerase is unique in that it is a ribonucleoprotein complex, where the RNA component [telomerase RNA (TR)] not only provides the template for the synthesis of telomere DNA repeats but also plays essential roles in catalysis, accumulation, TR 3′-end processing, localization, and holoenzyme assembly. Biochemical studies have identified TR elements essential for catalysis that share remarkably conserved secondary structures across different species as well as species-specific domains for other functions, paving the way for high-resolution structure determination of TRs. Over the past decade, structures of key elements from the core, conserved regions 4 and 5, and small Cajal body specific RNA domains of human TR have emerged, providing significant insights into the roles of these RNA elements in telomerase function. Structures of all helical elements of the core domain have been recently reported, providing the basis for a high-resolution model of the complete core domain. We review this progress to determine the overall architecture of human telomerase RNA.

Telomerase is a large, multisubunit ribonucleoprotein (RNP) that replicates the 3′ end of linear chromosomes by processive synthesis of telomere DNA repeats. Telomeres, the physical ends of linear chromosomes, generally comprise dsDNA with a short repeating species-specific sequence ending in a 3′ single-stranded overhang of variable length plus associated telomere binding proteins, called shelterin in humans (1, 2). Telomeres protect the integrity of linear chromosomes by allowing the cellular DNA repair machinery to distinguish them from double-strand breaks, thus playing critical roles in maintaining genome stability in eukaryotes (1, 2). Shortening of telomeres below a critical length because of inherent incomplete replication of DNA ends ultimately leads to telomere fusions and cell senescence (3–6). The 3′ ends of telomeres are replicated by telomerase, a unique reverse transcriptase discovered almost three decades ago (7), which catalyzes the addition of telomere DNA repeats onto the ends of linear chromosomes using an embedded RNA as the template (8, 9). Although telomerase has a low or undetectable level of activity in most somatic cells, it is active in some germline, epithelial, and hematopoietic cells, and it is highly active in the majority (∼90%) of cancer cell lines (10–12). Telomerase deficiency because of mutations in human telomerase RNA (hTR) has also been linked to several inherited human diseases, such as dyskeratosis congenita, aplastic anemia, myelodysplasia, and idiopathic pulmonary fibrosis (13–26).

The telomerase holoenzyme includes a unique reverse transcriptase [telomerase reverse transcriptase (TERT)], an essential RNA (TR), and several species-specific proteins required for proper function in vivo (27–29). The protein TERT is highly conserved across different species, and it usually contains four major functional domains: the TERT N-terminal domain (TEN), the TERT RNA binding domain (TRBD), the reverse transcriptase domain (RT), and the C-terminal extension (27, 30–32). The TEN domain interacts with the ss telomere DNA repeats, the TRBD domain binds multiple sites of TR, and the RT and C-terminal extension domains bind the RNA/DNA hybrid and catalyze the addition of DNA repeats onto the 3′ end (27, 31–34). Although no structures of TERT with all four domains have yet been reported for any species, structures of the TEN and TRBD domains from Tetrahymena thermophila telomerase (33, 35) and the full-length Tribolium castaneum TERT that lacks the TEN domain have been reported (36, 37). These high-resolution structures have significantly advanced our understanding of how TERT catalyzes the reverse transcription of telomere DNA and how TERT could potentially interact with the ss telomere DNA and the template RNA/DNA hybrids. Excellent reviews can be found elsewhere that describe the current state of knowledge about the structure and function of TERT (31, 32).

Although the essential templating function of TR was discovered more than 20 y ago (38), the TR contains more than a template. To date, the TR sequences of 28 ciliates, 43 vertebrates, and 25 yeasts have been determined (39–41). In contrast to the relatively conserved TERT, TRs differ greatly not only in sequence but also in length, ranging from 147 to 205 nt in ciliates (38, 42–45), from 312 to 559 nt in vertebrates (46, 47), and from 779 to >2,030 nt in yeasts (40, 41). Despite significant challenges in identifying common functional TR elements because of such divergence, phylogenetic and mutational studies have revealed a conserved secondary structure found in common among TRs across different species, which includes a large loop containing the template, a 5′ template boundary element, a pseudoknot, a loop-closing helix, and a stem terminus element (STE) (9, 15, 46, 48). The conservation of secondary structures rather than sequences suggests a role for these RNA structural motifs in telomerase function, and indeed, these regions of TR are essential for synthesis of telomere repeats (49–55). Other regions of TR are involved in species-specific roles in telomerase biogenesis, RNA processing, localization, and accumulation (56–61).

A conserved secondary structure for vertebrate TRs was first determined on the basis of 35 sequences (46), with four proposed conserved structural domains: (i) the pseudoknot, which includes the template, and (ii) the conserved regions 4 and 5 (CR4/CR5), which together comprise the catalytic core of the TR; (iii) the box H/ACA; and (iv) the CR7. An updated secondary structure of hTR (47, 62) is shown in Fig. 1 with known associated holoenzyme proteins. All of the secondary structure elements found in common among TRs across different species are in the 5′ region of hTR, where the STE is the CR4/CR5 domain (nucleotides ∼243 to ∼326), and the remaining motifs comprise the core domain, also known as the template/pseudoknot domain or the pseudoknot/core domain (nucleotide 33–191) (46, 47, 63). These two highly conserved structural domains independently bind TERT (50, 64) and are the only required hTR elements for in vitro reconstitution of catalytically active telomerase with hTERT (49, 50). The 3′ end of hTR has been identified as an H/ACA small Cajal body (CB) -specific RNA (scaRNA; nucleotides ∼211 to ∼237 and ∼334 to 451, the upper boundary of the 5′ hairpin has not been determined) (59, 62, 65) (Fig. 1), and it plays essential roles in biogenesis and regulation of telomerase holoenzyme in vivo, including accumulation, 3′-end processing, and localization of hTR (58–60, 66). The hTR scaRNA binds two sets of the four evolutionary conserved H/ACA RNP proteins (dyskerin, Gar1, Nop10, and Nhp2) to form an H/ACA RNP (Fig. 1) (67). Structures of single hairpin H/ACA RNPs from archaea have provided insight into the likely placement of proteins on each hairpin of the hTR scaRNA domain (66, 68–71). The hTR scaRNA also contains a conserved Cajal body localization element (CAB box) (59) in its 3′ terminal hairpin loop in the region identified as the CR7 domain, identifying it as a scaRNP. The protein telomerase Cajal body protein 1 (TCAB1)/WD-repeat domain 79 (WDR79) that binds the CAB box and drives localization of hTR into Cajal body was recently identified (72, 73).

Fig. 1.

Secondary structure and known protein components of the human telomerase holoenzyme. The human telomerase RNA (hTR) contains three major structural and functional domains, the core domain, the CR4/CR5 domain, and the H/ACA scaRNA domain (46, 59, 62). The hTR core and CR4/CR5 domains independently bind the hTERT (blue ellipse) (50, 64). The hTR scaRNA domain binds two sets of the four H/ACA RNP proteins: dyskerin (green), Gar1 (cyan), Nop10 (magenta), and Nhp2 (orange) (67). The protein TCAB1/WDR79 (purple) binds both the dyskerin and the CAB box located at the CR7 region within the H/ACA scaRNA domain (72, 73).

To date, there are no crystal structures of telomerase RNA or telomerase protein–RNA complexes; however, over the last several years, structures of several key hTR elements have been determined by NMR spectroscopy (52, 54, 55, 60, 74–76). This review summarizes what NMR structural and dynamics studies, combined with biochemical and mutational analysis, have revealed about the functional roles of hTR domains and the overall architecture of hTR.

Core Domain of hTR

The core domain has been the main focus of biochemical and biophysical studies of hTR, because it contains most of the conserved nucleotides and most of the disease-linked mutations (21–24, 26, 48, 62). It is the largest functional RNA domain at the 5′ end of hTR. Biochemical characterizations have identified nucleotides between residues 33 and 191 as the minimal hTR core domain required for telomerase catalytic activity in vitro (63). This region can be further divided into three major segments, a large loop containing the template, the P1 helix as the loop-closing helix, and a full-length P2/P3 pseudoknot (Fig. 2A). Within the P2/P3 pseudoknot, all four helices (P2a.1, P2a, P2b, and P3) have been shown to be required for telomerase activity, and disease-linked mutations disrupting these helical structures severely impair telomerase catalysis (63). Between these helical regions are internal loop regions. Except for loop J2b/3 and the 3′ portion of loop J2a/3, which are critical for the formation of P2b-P3 pseudoknot, most of these loop regions are not conserved, and their nucleotide identities are not essential for telomerase activity. The P2b-P3 pseudoknot is linked to the P1 helix through three nucleotides at its 3′ end and to the P2a helix through an asymmetric bulge J2a/b at its 5′ end (Fig. 2A). On the other side of the P2a helix are P2a.1 and the adjacent internal loop J2a.1, which are a mammalian-specific extension to P2a helix (Fig. 2A). The 5′ end of the P2a.1 helix and 3′ end of the P1 helix are linked by the large template-containing loop. These secondary structure features suggest a unique and overall compact architecture adopted by the hTR core domain, because the ends of the ∼43-bp P2/P3 pseudoknot are constrained by the intervening 24-nt template-containing loop.

Fig. 2.

Structures of subdomains of the hTR core domain. (A) Sequence and secondary structure of the hTR core domain, which can be divided into four subdomains: P2a-J2a.1-P2a.1 (blue-gray-gold), P2a-J2a/b-P2b (gold-green-red), P2b-P3 pseudoknot (red-pink), and P1 (dark green) (46). (B) NMR solution structures of the P2b-P3 pseudoknot, where PKDU (PDB ID 2K96) and PKWT (PDB ID 2K96) are structures for ΔU177 mutant and WT constructs, respectively (54). (C) NMR solution structure of P2a-J2a/b-P2b (PDB ID 2L3E) (55). (D) Structural model of the P2a-J2a.1-P2a.1 determined by the RDC-MC-Sym approach (55). All 3D structures are color-coded like the secondary structure in A, and nonnative residues are colored light gray.

Structure and Dynamics of the P2b-P3 Pseudoknot.

The P2b-P3 pseudoknot contains almost all of the highly conserved nucleotides within the hTR core domain (46). Early studies of minimal pseudoknots containing these conserved nucleotides showed that the pseudoknot unfolds to a P2b hairpin containing an unusual run of U-U and U-C base pair (77) and that a two-base mutation in the P3 found in some patients with dyskeratosis congenita (13) destabilizes the pseudoknot. Although a conformational switch between the completely folded pseudoknot and this partially folded hairpin was proposed as a molecular switch for template translocation (77, 78), both the solution structure of the minimal pseudoknot (discussed below) and mutational analysis suggest that a completely folded pseudoknot is required for optimal activity of telomerase (52, 79). The solution structure of the minimal hTR pseudoknot is a compact H-type pseudoknot with extensive tertiary interactions between the loop and stem nucleotides, and it revealed an essential triple helix in the pseudoknot (52). The functional importance of the tertiary interactions, described below, was shown by comparison of thermodynamic stabilities of pseudoknots with mutations and compensatory mutations and the effects of the same nucleotide substitutions incorporated into the full-length TR on telomerase activity in vitro (52). The tertiary structure of the minimal hTR pseudoknot also provided a structural explanation for the effect of disease mutations in this region of hTR (52, 54). Mutational studies and modeling of yeast telomerase pseudoknot domains (53, 80) have provided additional evidence for an essential role for a conserved triple helix in the pseudoknot domain. The putative pseudoknots in ciliates are much smaller than those pseudoknots in yeasts and vertebrates. Although no structures of ciliate pseudoknots have yet been reported, sequence analysis and modeling of 28 ciliate TRs have predicted one (and in one case, two) potential base triples in the ciliate pseudoknots (81).

In the minimal hTR pseudoknot (PKDU), the bulge U177 was removed to stabilize the pseudoknot formation. However, deletion of this residue in hTR resulted in a 2- to 10-fold decrease in telomerase activity (52, 78), and computational modeling of the P2b-P3 pseudoknot suggested that deletion of U177 alters tertiary interactions within the pseudoknot (82). In addition, substitution of the 2′-OH of residue A176, which is adjacent to residue U177, with 2′-OMe or 2′-H results in about a twofold decrease in activity, leading to the proposal that the A176 2′-OH may contribute directly to telomerase catalysis (53). To provide structural insight into the WT P2b-P3 pseudoknot (PKWT) and the functional role of U177, the solution structure of PKWT was determined, the original structure of PKDU was further refined with an extensive set of NMR residual dipolar couplings (RDCs) (Fig. 2B), and systematic structural and dynamical comparisons were made between PKWT and PKDU (54). These structural characterizations revealed that PKWT folds into the same H-type pseudoknot conformation as PKDU with almost identical tertiary interactions (Figs. 2B and 3 A and B). In both PKWT and PKDU, the P2a and P2b helices are stacked on either side of a junction that is a Hoogsteen base pair formed between the first nucleotide (U99) in the J2b/3 loop and the last nucleotide (A173) in the J2a/3 loop, and both pseudoknots are stabilized by the same base triples flanking this junction base pair (Figs. 2B and 3B). The A-rich J2a/3 loop forms two minor groove base triples with the P2b helix, and the U-rich J2b/3 loop forms three major groove U-A⋅U Hoogsteen base triples with P3 helix (Fig. 3C). However, PKDU is also stabilized by an additional C112⋅G178-U103 triple that is disrupted in PKWT because of the presence of U177, which creates a large roll and tilt between the flanking base pairs (Fig. 3 A and D). In PKWT, U177 is flipped out of the P3 helix into the minor groove and is located on the opposite side of the helix from the major groove triple helical interactions. Based on the structure, the base of U177 sits right over the 2′-OH of residue A176 and would sterically block its accessibility during telomerase catalysis (Fig. 3D). Dynamic characterization by NMR spin relaxation measurements revealed that U177 is intrinsically highly flexible, which suggests that, rather than blocking A176, U177 may serve as a hinge to provide additional backbone flexibility for residue A176 to facilitate the catalysis (54).

Fig. 3.

Tertiary interactions in the pseudoknot. (A) The structure of the triple helical region of PKWT (54). (B) Schematic representations of the tertiary interactions in PKWT. In PKDU, U103 forms a Hoogsteen base pair with G178. (C) U·A-U Hoogsteen base triple. (D) Detailed view of the base pairs surrounding U177. The base of U177 stacks over the 2′OH of A176. The nucleotides are in CPK colors, except that U177 is colored magenta.

The structure determination of PKWT also revealed why it is thermodynamically less stable than PKDU (52). Comparison of NMR imino proton spectra of PKWT and PKDU as a function of temperature revealed an unanticipated difference in the late folding/early unfolding pathway. In PKDU, as expected, when temperature increases, the Hoogsteen base pairs between the J2b/3 loop and P3 helix begin to melt before the Watson–Crick base pairs in P3 helix. However, in PKWT, the presence of bulge U177 results in partial unstacking of the three Watson–Crick U-A base pairs from the rest of the helix below the bulge U177 (Fig. 3A). As a consequence, the three U-A base pairs above and the two base pairs below the bulge U are thermodynamically less stable than the tertiary (Hoogsteen base pair) interactions, and they begin melting before the Hoogsteen U-A base pairs. Thus, the presence of U177 in PKWT results in an altered late folding/early unfolding pathway in the pseudoknot, which may be relevant to RNA assembly. In addition, for both pseudoknots, the melting studies revealed that the junction J2b/3 loop–J2a/3 loop Hoogsteen U-A base pair is remarkable stable. Thus, this base pair plays a critical role in overall folding of the pseudoknot (54).

Structure, Dynamics, and Function of the P2a-J2a/b-P2b.

Adjacent to the highly conserved P2b-P3 pseudoknot is a 5-nt bulge loop, J2a/b, which serves as a bridge between the pseudoknot and the P2a helix. This pyrimidine-rich J2a/b bulge loop is not highly conserved in sequence, except for a relatively conserved G at the 5′ end (61% in vertebrate and 83% in mammalian TRs) (39). Initial studies showed that swapping the bulge sequence 5′→3′ or replacing the hTR sequence with the mouse TR sequence had little effect on human telomerase catalytic activity (63, 83). These results, along with the lack of sequence conservation, argued against an important role of the J2a/b bulge loop in telomerase activity. However, the location of the J2a/b bulge loop is conserved in all vertebrate TRs, and its length is usually 5 nt in mammalian TRs (46, 47). The solution structure of the J2a/b bulge loop together with flanking P2a and P2b helices showed that J2a/b introduces a large bend (89 ± 3°) between P2a and P2b across the major groove (Fig. 2C) (55). The 5′-end residue of the bulge loop, G84, stacks above P2a, and the 3′-end residue of the bulge loop, C88, stacks below P2b. A change in the backbone direction occurs at the center bulge residue, U86, which leads to an overall S-shape conformation of J2a/b. Remarkably, the S-shape structure of the J2a/b results in almost no twist between P2a and P2b (−10 ± 10°) (Fig. 2C). A search for other 5-nt bulge structures, using the FRABASE program (84), uncovered only one other 5-nt bulge among all RNA structures solved to date from the hepatitis C virus (HCV) internal ribosome entry site (IRES) domain II (85). Even more surprisingly, this HCV IRES domain II bulge adopts the same S-shaped structure, despite the fact that it has a different sequence from the hTR J2a/b. A somewhat broader search that also allowed for noncanonical closing base pairs and swapping the strand with the 5-nt bulge uncovered two additional sequences, and these sequences have similar S-shaped internal loops (86, 87). Thus, the J2a/b bulge represents a rare structural motif.

Systematic dynamic characterizations by NMR ¹⁵N spin relaxation (88) and NMR RDCs (89, 90) also revealed another unusual feature of the J2a/b bulge loop. It was initially expected that the J2a/b bulge loop would be highly flexible, but although J2a/b exhibits some flexibility, its motion is remarkably limited on the nanosecond to millisecond time scale compared with the 3-nt bulge loop in HIV-1 transactivation response element (TAR) RNA. Based on a cone motional model, P2a and P2b move relative to each other with a motional amplitude of ∼39°. In contrast, HIV-1 TAR RNA has a cone-model motional amplitude of ∼55° (91).

The rare structure and unexpected dynamics of J2a/b suggested that it might have an important functional role in telomerase RNA topology and function. To test this hypothesis, systematic mutations were carried out to investigate the effects of the length, strand location, and sequence of J2a/b on telomerase function (55). The results showed not only that the directional bending defined by J2a/b is required for overall telomerase activity but also that the intrinsic flexibility across J2a/b is important for processive catalysis by telomerase. The results also suggest the consensus sequence G₈₃Y₇₈Y₈₇Y₉₆Y₈₇ for the 5-nt bulge in mammalian TRs may have evolved for both nucleotide addition and template translocation during telomerase catalysis. Finally, the large bend at the J2a/b bulge would be expected to play a significant role in determining the overall topology of the core domain, which discussed more below.

Structure and Dynamics of the Mammalian-Specific P2a.1-J2a.1-P2a.

Located on the other side of hTR P2/P3 pseudoknot is the P2a.1-J2a.1-P2a domain, where P2a.1 is a mammalian-specific extension to the P2a helix through an asymmetric internal loop J2a.1. Biochemical studies have shown that some nucleotide substitutions in this region and the C72G mutation associated with aplastic anemia result in a decrease in telomerase activity (62, 63, 83). In NMR spectra of a P2a.1-J2a.1-P2a construct, most of the nonexchangeable protons from the J2a.1 loop and surrounding nucleotides were exchange-broadened and in many cases, not visible. This finding not only indicated the presence of significant conformational exchange in J2a.1, apparently because of the bases in the internal loop adopting more than one conformation involving alternative base pairs, but also hindered high-resolution structure determination of this region of hTR by solution NMR (55). To overcome these difficulties, the RDC-MC-Sym approach was developed for structure determination of nucleic acids (55). This approach uses a combination of computational modeling by program MC-Sym (92) and experimentally derived restraints by NMR RDCs. The structural analysis revealed that P2a.1 is essentially a linear extension of the P2a helix (Fig. 2D). The interhelical bend between P2a.1 and P2a is only 6 ± 3°, and the twist between the two helices is 135 ± 14°, which agrees well with the amount of twist for an ∼4-bp irregular helix formed by the asymmetric loop J2a.1. Although J2a.1 undergoes significant conformational exchange, the interhelical motion between P2a.1 and P2a is remarkably restricted (∼26° cone motions) based on dynamic characterization by NMR RDCs. The stability between helices is consistent with the observations that disruption of base pairs flanking J2a.1 decreases telomerase activity and that nonmammalian vertebrate TRs have a single long P2a helix without a J2a.1 (46, 47).

High-Resolution Model of the hTR Core Domain.

The structures and dynamic analysis of subdomains of the hTR core domain, in particular, the conserved pseudoknot, have provided great insights into their functional roles in telomerase catalysis. However, they do not provide an overview of how the P2/P3 pseudoknot folds and how its architecture imposes conformational constraints that position the template into the active site of telomerase. A low-resolution (6.5–8.0 Å) FRET model of the hTR core domain has been reported (PDB ID 2INA), where distances derived from FRET between fluorescently labeled peptide nucleic acids were used in structural modeling (93). In this approach, although these distances were obtained in the full-length hTR, they provide only indirect constraints on the positions of the helical regions of the hTR core domain, because the peptide nucleic acids were hybridized onto three single-stranded regions, the 5′ end of hTR (nucleotides 1–13), the template region (nucleotides 44–56), and the J2a/3 region (nucleotides 146–158). One caveat is that the inherent flexibility of the single-strand regions, in particular, at the 5′ end of the hTR, was not considered in the modeling. Recently, by applying a computational modeling approach that incorporates the Assisted Model Building with Energy Refinement (AMBER) force field and short-range NOE restraints from individual structures, a high-resolution structure model of the P2/P3 pseudoknot was determined that contains all nucleotides except for the nonconserved single-strand region of J2a/3 (Fig. 4A) (55). There are significant differences between the FRET- and NMR-based structure models in both the local structures and relative helical orientations, but both models do exhibit an open architecture of the hTR core domain and a major interhelical bend occurring across the J2a/b bulge (55, 93).

Fig. 4.

Models of the hTR core domain and interaction with TERT. (A) NMR-based model of the hTR core domain including a DNA primer bound to the template (55). (B) The hTR P2/P3 pseudoknot positioned onto the T. castaneum TERT in two possible orientations, where the hTR P2/P3 pseudoknot lies either parallel (Left) or perpendicular (Right) to the T. castaneum TERT-telomeric RNA/DNA complex (PDB ID 3KYL) (37). The color scheme for the hTR P2/P3 pseudoknot domain is the same as in Fig. 2. Domains of the T. castaneum TERT are colored and labeled as shown, and the RNA template and telomeric DNA are colored in purple and cyan, respectively. A comparison of the domain structures of the hTERT and T. castaneum TERT is also shown.

The NMR-based structure model revealed that the full-length P2/P3 pseudoknot of the hTR core domain folds into an overall V-shape conformation that is defined by the ∼90° bend across the J2a/b bulge. The P2b-P3 pseudoknot and P2a.1-J2a.1-P2a domain are located on either side of the J2a/b bulge, and they are each ∼50Å long. Thus, the bend across J2a/b creates an ∼70 Å end to end distance between P2b-P3 and P2a.1, which agrees well with the length of the intervening 24-nt single strand containing the template without significant stretching. Because of the negligible twist between the P2a and P2b induced by the J2a/b bulge loop, the structural model also revealed that the conserved bulge residue U177 (52, 78) and the catalytically important 2′OH of residue A176 (53) are both positioned on the inner surface of the core domain and face to the 5′ end of the template, where the active site should be located. The structural model is also consistent with the close proximity of the end of the pseudoknot to the template needed to obtain maximal activity in an engineered cis-telomerase (94). In addition, the dynamic characterization showed that interdomain motion between P2a and P2b has an amplitude of ∼39° assuming a cone motional model, which can be translated into a displacement of ∼28 Å between the two ends in the full-length P2/P3 pseudoknot. This distance agrees well with the ∼17 Å that the template must translocate through the active site during the synthesis of each TTAGGG telomere repeat. Although the binding sites of TERT on the core domain have not been identified, the overall shape of the P2/P3 pseudoknot fits nicely into the crystal structure of the TERT from T. castaneum (36, 37). The pseudoknot can be modeled onto the TERT either in parallel or perpendicular to where the template/primer fits in the donut hole of the TERT, such that flexing of J2a/b would allow the template to be pulled through the active site during telomere synthesis (Fig. 4B). This finding provides a testable model for how the core domain interacts with TERT.

STE of hTR

The STE is the other catalytically essential TR element conserved across ciliates, vertebrates, and yeasts (9, 50, 95, 96). In hTR, the CR4/CR5 domain is the STE, and it interacts directly with hTERT independently from the core domain. The hTR CR4/CR5 domain contains a three-way junction (Fig. 5A), which has also been proposed to form in yeast telomerase RNAs (97). Like the core domain, although the CR4/CR5 secondary structure is conserved, most conserved residues are localized in one particular region, the P6.1 hairpin. Additional highly conserved residues are located at the large internal loop that forms a three-way junction between P6, P6.1, and P5. Although various Watson–Crick base pairs can be drawn between nucleotides in the three-way junction, substitutions and compensatory mutations indicate that it is the sequence rather than the base pairs that are important for telomerase activity (98). In addition, a minimal CR4/CR5 domain construct, which includes P6.1, all of P6a-P6b except the terminal hairpin, and flanking nucleotides but does not include the P5 helix, is sufficient to reconstitute telomerase activity in trans with the core domain and TERT (Fig. 5A) (76). Thus, it is not clear whether this region forms a three-way junction in the context of its association with TERT. To date, no structural information on the CR4/CR5 three-way junction has been reported, but the structures of two helical regions of the hTR STE have been solved (74–76). These structures are the essential P6.1 hairpin and the central portion of P6 surrounding the J6 internal loop.

Fig. 5.

Structures of subdomains of the hTR CR4/CR5 domain. (A) Sequence and secondary structure of the hTR CR4/CR5 domain (46). The minimal CR4/CR5 required for reconstitution in vitro of active telomerase is highlighted by a dashed box (76). (B) NMR solution structures of P6.1 (PDB ID 1OQ0) (75) and Ψ4P6.1 (PDB ID 2KYE) (76). The two U and three G letters in the P6.1 loop are colored in green and yellow, respectively. The two pseudouridines (Ψs) in the P6.1 loop are colored in red. (C) NMR solution structure and secondary structure of a P6 hairpin construct (PDB ID 1Z31) (74). The internal loop residues (C266-C267 and A289-U291) are colored in green, and the bulge residue C262 is colored in purple.

Structures of P6.1 and Pseudouridylated P6.1

The P6.1 is a short hairpin off of the three-way junction that has a helix of four Watson–Crick base pairs capped by a 5-nt loop. The 13-nt sequence of P6.1 is highly conserved; in addition to U307 and G309 in the loop, all 8 nt forming Watson–Crick base pairs are 100% conserved in vertebrates. The formation of base pairs in P6.1 has been shown to be essential for TERT binding and telomerase activity (50, 98). Both the length of the P6.1 helix and the two conserved loop residues are critical for telomerase activity but not for binding of TR to TERT (64, 98). The structure of P6.1 showed that the first (U306) and the last (G310) residues of the 5-nt loop form a U-G wobble pair on top of the A-form helix (Fig. 5B) (75). The remaining three residues from the loop, U307, G309, and G310, are exposed to solvent, with U307 and G309 located on the minor groove side of the loop and G308 located on the major groove with partial stacking on U306. The small loop formed by these three residues has a well-defined conformation, and this architecture of the loop has been proposed to allow the essential loop nucleotides to interact with TERT or TR (75). Intriguingly, gel shift and cross-linking experiments have shown that isolated hTR fragments of P6.1 and the template can directly interact with contacts between the P6.1 loop residues and the two ends of the template region (99), although the biological relevance of this interaction in the context of hTERT remains to be established.

Most noncoding RNAs contain modified nucleotides, and hTR seems to be no exception. Six potential pseudouridine (Ψ) modification sites have been identified within the hTR, two of which are located in the P6.1 loop (76). Pseudouridine is the most abundant posttranscriptional modified RNA nucleotide, and it is found in all species (100). The locations of pseudouridines are usually highly conserved, and they play important roles in biological functions. To investigate the potential functional roles of these Ψ-modifications in the essential P6.1 loop, the solution structure of a pseudouridine-modified P6.1 (Ψ-P6.1) was determined (Fig. 5B) (76). Furthermore, both the thermodynamic stability of the Ψ-P6.1 vs. unmodified P6.1 and the effect of pseudouridylation on telomerase activity in vitro were characterized. The loop structure of the Ψ-P6.1 is significantly different from the loop structure of P6.1. The stem closing Ψ306⋅G310 base pair has a single hydrogen bond between the imino proton of G310 and the O4 of Ψ306, which is different from the canonical U⋅G wobble pair found in the unmodified P6.1. The bases of Ψ306 and Ψ307 are located in the major groove of the loop and are stacked on G305 and Ψ306, respectively. These two pseudouridines are spatially positioned to form water-mediated hydrogen bonds between the imino protons and their 5′ phosphate oxygens. Only one base, the nonconserved G308, is flipped out of the loop, and it is on the minor groove side. The phosphate backbone turns between G308 and G309, and G309 in the syn conformation stacks on the sugar of G310. These structural features revealed that the Ψ-modifications at the loop increase both base-stacking and hydrogen-bonding interactions in the Ψ-P6.1 relative to the unmodified P6.1, resulting in higher thermodynamic stability that is characterized by UV melting studies. Interestingly, the pseudouridine modification at 306 and 307 in the P6.1 loop decreased the telomerase activity in vitro by approximately threefold but slightly increased the telomere addition processivity (∼20%), indicating that pseudouridylation may have a subtle but significant effect on telomerase activity.

Structure of the P6 Hairpin.

The P6 hairpin is located adjacent to the 5′ end of P6.1. It has two helical regions, P6a and P6b, that flank a small asymmetric internal loop J6, and it is capped by a UCCG hairpin (Fig. 5A). Deletion of the hairpin loop has no effect on telomerase activity, but additional deletions that include the J6 internal loop have been shown to inhibit interactions between CR4/CR5 and TERT and abolish telomerase activity (50).

The solution structure of an RNA derived from P6, including the J6 internal loop, has been determined (Fig. 5C) (74). P6a was predicted on the basis of phylogenetic analysis to have a C262-A295 base pair next to bulge U261, but in the structure, a U261-A295 base pair forms and C262 is bulged out. The J6 internal loop is well-defined, with a potential triple formed between C267 through water-mediated hydrogen bonds to the G268-C288 base pair. Interestingly, similar to the P2a.1-J2a.1-P2a, the interhelical motion between P6a and P6b across this asymmetric bulge is very limited with amplitude <10°, which was revealed by RDC analysis. These structural and dynamical properties of J6 result in a unique conformation of the P6 region. In addition to a 20° interhelical angle and ∼3 Å deflection between P6a and P6b, the J6 internal loop forms an unusual solvent-accessible opening, leading Leeper et al. (74) to speculate that this internal loop may be important for TERT or TR interactions. However, a variety of base substitutions in this internal loop do not abrogate activity (50), and therefore, it is not clear at the current stage that this structure is unique.

H/ACA scaRNA Domain of hTR

In vivo biogenesis and regulation of telomerase holoenzyme require additional telomerase RNA motifs. In vertebrates, the 3′-half of TR comprises an H/ACA scaRNA domain, which includes a telomerase-specific CR7 domain that forms a 3′-terminal hairpin. All H/ACA small nucleolar RNAs (snoRNAs) and scaRNAs form RNPs with evolutionary conserved H/ACA RNP proteins and generally function to direct site-specific pseudouridylation of ribosomal RNAs and small nuclear RNAs, respectively (101–104), but in the case of telomerase, the domain seems to have been co-opted to help localize telomerase to Cajal bodies (56, 58, 59). The H/ACA motif has a hairpin-hinge-hairpin-tail secondary structure, where the conserved H box is located at the hinge region and the ACA box is located at the 3′-end tail (66). In addition to the H/ACA motif, scaRNAs share another common motif, known as the CAB box, which has a consensus ugAG sequence and serves as a CB localization signal (105). Remarkably, the CR7 domain of vertebrate TRs, as discussed below, contains not only a CB localization signal but also another signal for the accumulation and processing of TRs (60). Although the H/ACA and CR7 domains are not required for telomerase activity in vitro, they form a scaRNA domain that is essential for in vivo accumulation, 3′-end processing, and localization of vertebrate TRs.

Although no high-resolution structures of the H/ACA domain from any vertebrate TRs with or without H/ACA RNP proteins have been determined to date, a combined structural and biochemical study, including localization studies using FISH, on the hTR has delineated the sequence and structural elements of the two independent signals, located in CR7, for 3′-end processing and accumulation and localization of hTR to Cajal bodies (Fig. 6) (60). The CR7 hairpin loop has four key structural features. First, within the 8-nt loop, there is 1 bp observed between the first and the second to last nucleotides of the loop, a U411⋅G417 wobble pair. Second, because of the U⋅G pair in the loop and stable Watson–Crick base pairs in the helix, the last nucleotide of the loop, U418, adopts an unpaired conformation. Third, the first 3 loop nt that are also the first 3 nt of the CAB box are stacked over each other, forming A-form backbone geometry. Finally, although the last CAB box residue (G414) does not stack over other residues and is quite flexible, a change in the backbone direction occurs between G414 and C415, positioning all CAB box residues on the 5′ side of the loop. To determine which of these structural features is important for the 3′-end processing and localization signals, parallel structural determinations were performed on the terminal loop construct derived from the human U64 H/ACA snoRNA and human U85 C/D-H/ACA scaRNA. It had been previously shown the U64 snoRNA replacement of the hTR CR7 in the full-length hTR allows hTR accumulation and processing, but the RNA localizes to nucleoli instead of Cajal bodies. In contrast, replacement of the CR7 hairpin with the 3′ hairpin of U85 scaRNA abrogated processing, and the hTR remained at the site of transcription. Comparison of the terminal loop structures of the hTR CR7 and the 5′ hairpin of human U64 H/ACA snoRNA revealed that the first two structural features described above are shared in common between these two RNAs, and additional mutational analysis confirmed that the processing signal for 3′-end maturation of hTR comprises a U-G base pair between the first nucleotide of the loop and the penultimate nucleotide of the loop, an unpaired U418 at the last position of the loop, and some sequence specificity for the top 2 bp in the helix (Fig. 6 C and D). Comparison of the structures of the hTR CR7 hairpin loop and the human U85 C/D-H/ACA scaRNA, combined with characterization of the processing, localization, and accumulation of various mutant and Wt RNAs, revealed that the CB localization signal comprises the ugAG CAB box nucleotides as well as a base-paired helix, with some dependence of the sequence identity of the top 2 bp (Fig. 6 E and F). Recently, the protein responsible for binding to the CB localization signal and driving telomerase and other H/ACA scaRNAs to Cajal bodies has been identified as TCAB1/WDR79 (72, 73). It is interesting to note that, although they are both in the same hairpin, the CB localization signal and the 3′-end processing signal could potentially independently and simultaneously bind TCAB1/WDR79 and the unidentified protein that is required for TR accumulation and 3′-end processing. Furthermore, the 3′ processing signal is the same site where the H/ACA RNP protein Nhp2 would be expected to bind (60, 106).

Fig. 6.

Structure and function of the hTR CR7 domain (60). (A) Sequence and secondary structure of the hTR CR7 hairpin and flanking base pairs. Boxed residues are the CAB box sequence. Residues with >95% and >85–95% conservation among all vertebrate species are shown in capital letters and bold fonts, respectively. Nonnative residues used in structure determination are the base pairs below the dashed line. (B) NMR solution structure of the CR7 hairpin (PDB ID 2QH2). Residues are colored by type: orange, A; green, U; blue, G; red, C. (C and D) Residues comprising the hTR-specific processing signal are highlighted in magenta in the secondary structure (C) and colored as in B in the surface representation of the CR7 structure (D). (E and F) Residues comprising the CB localization signal are highlighted in dark blue (CAB box) and light blue in the secondary structure (E) and colored as in B in the surface representation of the CR7 structure (F).

Summary

During the past decade, combined use of secondary structure predictions, mutational analysis, structure determination of subdomains, and structure-based modeling of hTR in conjunction with biochemical characterization have revealed essential elements of hTR structure and function beyond the role in templating the telomere DNA. Important insights include the identification and characterization of an essential triple helix in the conserved pseudoknot in the core domain, a structurally conserved bulge loop in the core domain that directs the overall topology, and two independent signals, for Cajal body localization and hTR 3′-end processing, in the 3′-terminal hairpin of hTR. Recent structural studies of the core domain have provided a testable model structure of the core domain that is consistent with all current biochemical data. However, much remains to be learned, particularly about how and where hTERT interacts with the core domain and how the core domain and the STE interact with hTERT and potentially each other to facilitate catalysis. The structure of the H/ACA scaRNA domain and how two sets of H/ACA RNP proteins bind two consecutive hairpin-bulge-hairpin sequences remain to be determined. Combined use of NMR, X-ray crystallography, small-angle X-ray scattering, single molecule techniques, and EM will likely reveal these structures within the next few years.