A Shared Docking Motif in TRF1 and TRF2 Used for Differential Recruitment of Telomeric Proteins

See allHide authors and affiliations

Science  22 Feb 2008:
Vol. 319, Issue 5866, pp. 1092-1096
DOI: 10.1126/science.1151804


Mammalian telomeres are protected by a six-protein complex: shelterin. Shelterin contains two closely related proteins (TRF1 and TRF2), which recruit various proteins to telomeres. We dissect the interactions of TRF1 and TRF2 with their shared binding partner (TIN2) and other shelterin accessory factors. TRF1 recognizes TIN2 using a conserved molecular surface in its TRF homology (TRFH) domain. However, this same surface does not act as a TIN2 binding site in TRF2, and TIN2 binding to TRF2 is mediated by a region outside the TRFH domain. Instead, the TRFH docking site of TRF2 binds a shelterin accessory factor (Apollo), which does not interact with the TRFH domain of TRF1. Conversely, the TRFH domain of TRF1, but not of TRF2, interacts with another shelterin-associated factor: PinX1.

Shelterin acts in conjunction with many associated factors (16). Most of the shelterin-associated proteins are recruited to telomeres through interactions with TRF1 or TRF2 (26). However, the molecular mechanism of these TRF1- and TRF2-mediated interactions remains unknown. TRF1 and TRF2 share the same molecular architecture, characterized by a C-terminal Myb/SANT DNA binding domain (7, 8) and an N-terminal TRFH domain (9). The TRFH domains (TRF1TRFH and TRF2TRFH) mediate homodimerization and are required for telomeric DNA binding by TRF1 and TRF2 (10, 11). Several different protein interactions have been mapped to the TRFH domains of TRF1 and TRF2 (2, 1214). The TRFH domains have almost identical three-dimensional structures (11); therefore, it is difficult to explain how TRF1 and TRF2 can interact with different proteins.

TRF1 and TRF2 both bind to another shelterin protein: TIN2 (12, 15, 16). The TRF1-TIN2 interaction was mediated by TRF1TRFH and the C terminus of TIN2 (12). Further mapping revealed that a peptide of TIN2—denoted as TIN2256–276 [TIN2TBM: TIN2–TRFH binding motif (TBM)]—retains the TRF1TRFH binding activity with a binding affinity of 314 nM (figs. S1 and S2 and Fig. 1A). To understand how TIN2TBM is recognized by TRF1TRFH, we crystallized the TRF1TRFH-TIN2TBM complex and solved its structure at 2.0 Å resolution (table S1) (17). The electron density map shows that residues 257 to 268 of TIN2TBM assume a well-defined conformation (fig. S3). TRF1TRFH forms homodimers, and each TRF1TRFH interacts with one TIN2TBM peptide (Fig. 1B). TRF1TRFH exhibits essentially the same conformation as unliganded TRF1TRFH except for loop L34 (Fig. 1C) (11). Loop L34 is partially disordered in the peptide-free structure (Fig. 1C). However, once TIN2TBM is bound, loop L34 folds back upon helices α3 and α4, sandwiched between the helices and TIN2TBM (Fig. 1C).

Fig. 1.

Structure of the TRF1TRFH-TIN2TBM complex. (A) In vitro ITC measurement of the interaction of TRF1TRFH with the TIN2TBM peptide. The inset shows the ITC titration data. (B) Overall structure of the dimeric TRF1TRFH-TIN2TBM complex. TRF1TRFH and TIN2TBM are colored in green and yellow, respectively, in one complex, and dark green and orange, respectively, in the other. (C) Superposition of the TRF1TRFH-TIN2TBM complex on the unliganded structure of TRF1TRFH. Loop L34 in the complex is in red and that of unliganded TRF1TRFH is in cyan, whereas the rest of TRF1TRFH is in green (TIN2TBM-bound) or gray (peptide-free).

The structure of the complex reveals two adjacent but structurally distinct interaction modes. The N terminus of TIN2TBM [His257-Phe-Asn-Leu-Ala-Pro262 (H257-F-N-L-A-P262)] (18) adopts an extended conformation stabilized by an extensive intermolecular hydrogen-bonding network (Fig. 2A and fig. S4). The side chain of L260 is therefore positioned into a deep hydrophobic pocket of TRF1TRFH (Fig. 2, B and C). In addition, F258 and P262 also make hydrophobic contacts with TRF1TRFH: F258 sits on a concave hydrophobic surface, whereas P262 stacks with TRF1-F142 (italics are used here for individual residues of TRF1 and TRF2) (Fig. 2, A to C, and fig. S4). In contrast, the C terminus of TIN2TBM (L263-G-R-R-R-V268) is positioned on the surface of loop L34 through formation of an antiparallel β sheet with D139-A-Q141 of TRF1TRFH (Fig. 2A and fig. S4) so that R265-R-R267 of TIN2TBM contacts TRF1TRFH through electrostatic interactions (Fig. 2C). In particular, R266 is nested within an acidic depression on the surface of loop L34 through a network of salt bridges and hydrogen bonds (Fig. 2, A and C, and fig. S4).

Fig. 2.

The TRF1TRFH-TIN2TBM interface. (A) Schematic depiction of the TRF1TRFH-TIN2TBM interaction. The main-chain atoms of TIN2TBM are shown as circles [carbon in yellow (Cα in orange), oxygen in red, and nitrogen in blue]. Residues of TRF1TRFH are shown as green ovals (side-chain interaction) and square boxes (main-chain interaction). Hydrophilic and hydrophobic interactions are shown as straight magenta lines and curved red lines, respectively. The pale yellow arrows denote the intermolecular β sheet. (B) The shape of the hydrophobic pocket of TRF1 (green mesh) complements the side chain of TIN2-L260 well. (C) Electrostatic surface potential of the TIN2TBM binding site of TRF1TRFH. Positive potential, blue; negative potential, red. (D) In vitro ITC binding data of wild-type and mutant TRF1TRFH-TIN2TBM interactions. Kd, equilibrium dissociation constant; nd, not detectable by ITC. (E) Co-IP of the same sets of mutant TRF1-TIN2 interactions (except the TRF1-TIN2 F258A interaction) as in (D). Lanes marked “In” represent 2.5% of input cell lysate used for the immunoprecipitation.

To investigate the importance of the TRF1-TIN2 interaction, we first measured the binding of different mutant TIN2TBM peptides to TRF1TRFH by isothermal titration calorimetry (ITC). Substitution of L260 with either an alanine or a glutamate abolished the binding (Fig. 2D). Similarly, mutant TIN2TBM–F258→A258 (TIN2TBM-F258A) substantially impaired the interaction (Fig. 2D). By contrast, mutant TIN2TBM-P262A, designed to eliminate a stacking interaction with TRF1-F142, had a wild-type binding affinity, indicating that loss of this interaction is not essential for binding (Fig. 2D). However, substitution of TRF1-F142 with an alanine completely abrogated the binding to TIN2TBM (Fig. 2D). We then tested the interactions of mutant proteins transiently expressed in human embryonic kidney 293T cells, and the coimmunoprecipitation (Co-IP) results are consistent with the in vitro ITC measurements (Fig. 2E). We therefore conclude that the TRFH interaction motif in TRF1 is necessary for the TRF1-TIN2 interaction both in vitro and in vivo.

Given the sequence and structural similarities of the TRFH domains of TRF1 and TRF2, we expected that TRF2 would also bind to TIN2 through the TRFH domain (figs. S5 and S6). However, Co-IP studies of a specific mutant in TRF2 (TRF2-F120A, where TRF2-F120 is structurally equivalent to TRF1-F142), which was predicted to abolish TIN2 binding to TRF2TRFH, did not have the expected effect (Fig. 3A). Therefore, TRF2TRFH is not required for the stable association with TIN2 in vivo. In order to define the actual TIN2 binding site, we tested an array of glutathione S-transferase–TRF2 fusion fragments in a Far-Western assay for their ability to interact with TIN2. The result showed that a short peptide of TRF2 (TRF2352–365) can mediate an efficient interaction with TIN2 (Fig. 3B). In addition, purified TRF2350–366 comigrated with TIN21–220 in gel-filtration chromatographic analysis, indicating that TIN21–220 is sufficient for binding (fig. S7). Furthermore, Co-IP data showed that a deletion mutant of TRF2 (TRF2-Δ352–367) that retains the entire TRFH domain but lacks the TIN2 binding site failed to associate with TIN2 (Fig. 3A). Therefore, TRF2TRFH does not mediate a stable interaction with TIN2 in vivo. Collectively, we conclude that, although TRF1 binds TIN2 through its TRFH domain, TRF2 interacts with TIN2 through a short motif in its C terminus.

Fig. 3.

The TRF2-TIN2 interaction. (A) Co-IP of TIN2 with cotransfected wild-type and mutant TRF2. (B) Far-Western analysis of the TIN2 binding region of TRF2 (FL, full-length; TRF2-ΔB, TRF2-Δ1–42). (C) Superposition of the TIN2TBM binding sites in the TRF1TRFH-TIN2TBM and TRF2TRFH-TIN2TBM complexes. TRF1TRFH and TRF2TRFH are in green and cyan, respectively. The TIN2TBM peptides bound to TRF1TRFH and TRF2TRFH are shown in stick model format and in yellow and magenta, respectively. (D) TIN2-F258 interacts less efficiently with TRF2 than with TRF1. The F258 binding surfaces of TRF1TRFH (top panel) and TRF2TRFH (bottom panel) are shown in magenta (hydrophobic patch) and blue (hydrophilic patch). The rest of TRF1TRFH and TRF2TRFH is in green and cyan, respectively.

The distinctive specificity of the TRFH domains of TRF1 and TRF2 suggested that subtle structural differences are responsible for the ability of TIN2 to distinguish between these two paralogous proteins. ITC measurement showed that TRF2TRFH interacts with TIN2TBM in vitro, but with a much lower affinity (6.49 μM) (fig. S8A). To understand this binding specificity, we solved the crystal structure of the TRF2TRFH-TIN2TBM complex at 2.15 Å resolution (fig. S8B and table S1). Although the overall conformations of TIN2TBM bound to TRF1TRFH and TRF2TRFH are very similar (Fig. 3C), subtle differences can explain the difference in affinities of the two complexes (Fig. 3D and fig. S8, C and D). In the TRF1TRFH-TIN2TBM complex, TIN2-F258 sits snugly on a hydrophobic surface of TRF1TRFH (Fig. 3D). In contrast, F258 rotates away from the interface and packs less efficiently with TRF2TRFH, because the edge of the interaction surface is partially occupied by polar residues S98 and R102 (Fig. 3D). In addition, TRF1-E192, which is key for TIN2TBM binding, is replaced by a lysine residue in TRF2 (K173), resulting in loss of two ion-pairing interactions and an electrostatically unfavorable contact between TIN2-R266 and TRF2-K173 (figs. S6 and S9).

These results suggested that TRF2 might use its TRFH domain peptide docking site to recruit one or more of the shelterin accessory factors (2, 13, 19). TRF2TRFH is known to interact with Apollo, which functions together with TRF2 in protecting telomeres during S phase (2, 13). TRF2TRFH directly binds to the C terminus of Apollo (Apollo496–532) (fig. S10A) (13). We confirmed this interaction using the ITC binding assay (Fig. 4A). Under the same conditions, no binding enthalpy was measurable between Apollo496–532 and TRF1TRFH, indicating that Apollo496–532 binding is specific for TRF2 (Fig. 4A). To understand how TRF2 recognizes Apollo, we determined the crystal structure of the TRF2TRFH-Apollo496–532 complex at 2.5 Å resolution (Fig. 4B and table S1). The structure clearly shows electron density corresponding to the 12 N-terminal residues of Apollo496–532 (amino acids 498 to 509), referred to as ApolloTBM (fig. S10, A and B). The structure reveals that ApolloTBM interacts with TRF2TRFH through the same molecular surface as in the TRF1TRFH-TIN2TBM complex (Fig. 4C). Overlay of the two complexes reveals many similarities between the C terminus of ApolloTBM (Y504-L-L-T-P-V509) and the N terminus of TIN2TBM (F258-N-L-A-P-G265). First, two peptides are almost identical in overall conformation (Fig. 4C and fig. S10, C and D). Second, most of the hydrogen bonds in the TRF2TRFH-ApolloTBM complex are conserved in TRF1TRFH-TIN2TBM (fig. S10, C and E). Third, L506 and P508 of Apollo interact with TRF2TRFH in the same fashion as do their counterparts of TIN2TBM (Fig. 4C and fig. S10D). It is noteworthy that the TBMs of TIN2 and Apollo share the sequence Y/F-X-L-X-P (where X is any amino acid).

Fig. 4.

The TRF2-Apollo interaction. (A) ITC measurement of the interactions of TRF1TRFH (red) and TRF2TRFH (blue) with the ApolloTBM peptide. (B) Overall structure of the dimeric TRF2TRFH-ApolloTBM complex. (C) Superposition of ApolloTBM (orange) and TIN2TBM (yellow) reveals a shared F/Y-X-L-X-P motif. (D) Superposition of the TRF2TRFH-ApolloTBM and the TRF2TRFH-TIN2TBM complexes in the vicinity of the Apollo helix. The TRF2TRFH molecules are colored in cyan (ApolloTBM-bound) and gray (TIN2TBM-bound), respectively. (E) ApolloTBM binding is TRF2TRFH-specific. The surface representations show that there is no room for Apollo L500 and Y504 to fit into the peptide binding site of TRF1TRFH. (F) In vitro ITC binding data of wild-type and mutant TRF2TRFH-ApolloTBM interactions. (G) Co-IP data show that Apollo double-mutant L504E/P506 and TRF2 single-mutant F120A disrupt the in vivo TRF2-Apollo interaction. (H) Localization of retrovirally expressed HA-tagged wild type and L506E/P508A double mutant of Apollo in BJ-hTERT cells.

Despite the high degree of similarity between the TRF1TRFH-TIN2TBM and TRF2TRFH-ApolloTBM interactions, substantial structural variations are evident outside the Y/F-X-L-X-P motif. Unlike TIN2TBM, the Y-X-L-X-P motif resides at the C terminus of ApolloTBM, and ApolloTBM lacks a C-terminal polyarginine tail (Fig. 4C). Instead, it has a six-residue extension preceding the Y/F-X-L-X-P motif, which adopts a short helical conformation (Fig. 4, C and D) and packs on loop L23 and helices α2 and α3 of TRF2TRFH through hydrophobic contacts (Fig. 4D and fig. S10, C and E). Apollo-Y504 rotates ∼ 90° relative to TIN2-F258 in the TRF1TRFH-TIN2TBM complex to fit into a hydrophobic cleft formed by L101 and R102 of TRF2 (Fig. 4D). This reorientation of Y504 is coupled with a partial refolding of loop L23 of TRF2: TRF2-E94 rotates ∼ 180° relative to its position in the peptide-free conformation and makes two electrostatic interactions with K503 and Y504 of Apollo (Fig. 4D and fig. S10E). These marked conformational differences suggest that a tyrosine residue is preferred at the N-terminal position of the F/Y-X-L-X-P motif for efficient binding to TRF2TRFH, whereas a phenylalanine is preferred for TRF1TRFH. Furthermore, superposition of the TRF1TRFH-TIN2TBM and the TRF2TRFH-ApolloTBM complexes shows that the space occupied by L500 and Y504 of ApolloTBM is occluded in TRF1TRFH, which explains why ApolloTBM binding is TRF2TRFH-specific (Fig. 4E and fig. S11). Given the close similarity of the TRFH domains of TRF1 and TRF2, these structural variations emphasize that the TRFH domain is a versatile framework for interactions with different proteins.

The crystal structure of the TRF2TRFH-ApolloTBM complex is corroborated by mutagenesis. Mutations of the conserved hydrophobic residues of Apollo (F504, L506, and P508) or TRF2 (F120) completely abolished the interaction both in vitro and in vivo (Fig. 4, F and G). We further assayed the cellular localization of wild-type and mutant Apollo by expressing hemagglutinin (HA)–tagged proteins in human telomerase reverse transcriptase (hTERT)–immortalized human BJ fibroblasts. Although wild-type Apollo showed the expected telomere localization, the L506E/P508A double mutant was distributed throughout the nucleoplasm with no obvious accumulation at telomeres (Fig. 4H). This result confirms the structural information and indicates that the binding of Apollo to the TRFH domain of TRF2 is required for the telomeric localization of Apollo.

We next asked whether other shelterin-associated proteins might contain the F/Y-X-L-X-P motif suggestive of an interaction with the TRFH domain of TRF1 or TRF2. We identified this motif in PinX1, originally identified as a TRF1-interacting protein in a yeast two-hybrid screen (6). An 11-residue fragment of PinX1 (R287-D-F-T-L-K-P-K-K-R-R297), referred to as PinX1TBM, closely resembles TIN2TBM (fig. S12A), suggesting that it may bind to TRF1TRFH in the same fashion as does TIN2TBM. ITC data confirmed the TRF1TRFH-PinX1TBM interaction, whereas no measurable interaction was observed between TRF2TRFH and PinX1TBM (fig. S12B). Mutagenesis studies showed that PinX1-L291 and TRF1-F142 are critical for the interaction, whereas PinX1-P293 is not (fig. S12C). These results are consistent with those of the TRF1TRFH-TIN2TBM interaction (Fig. 2D) and indicate that PinX1, like TIN2, binds the TRFH domain of TRF1 but not TRF2. Protein sequence database searches showed many instances of telomere-associated proteins containing the F/Y-X-L-X-P motif (fig. S13). Future studies are needed to address whether this motif mediates the TRF1/TRF2 binding of these telomere-associated proteins in vivo.

Our results indicate that binding to the TRFH docking site involves the sequence F/Y-X-L-X-P in shelterin-associated proteins, which contacts the same molecular recognition surface of the TRFH domains of TRF1 and TRF2 with distinct specificities. Because TRF1 and TRF2 play different roles in telomere length homeostasis and telomere protection (1), we propose that the TRFH domains of TRF1 and TRF2 function as telomeric protein docking sites that recruit different shelterin-associated factors with distinct functions to the chromosome ends.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S14

Table S1


References and Notes

View Abstract

Navigate This Article