Structure of hnRNP D Complexed with Single-stranded Telomere DNA and Unfolding of the Quadruplex by Heterogeneous Nuclear Ribonucleoprotein D*

Heterogeneous nuclear ribonucleoprotein D, also known as AUF1, has two DNA/RNA-binding domains, each of which can specifically bind to single-stranded d(TTAGGG)n, the human telomeric repeat. Here, the structure of the C-terminal-binding domain (BD2) complexed with single-stranded d(TTAGGG) determined by NMR is presented. The structure has revealed that each residue of the d(TAG) segment is recognized by BD2 in a base-specific manner. The interactions deduced from the structure have been confirmed by gel retardation experiments with mutant BD2 and DNA. It is known that single-stranded DNA with the telomeric repeat tends to form a quadruplex and that the quadruplex has an inhibitory effect on telomere elongation by telomerase. This time it is revealed that BD2 unfolds the quadruplex of such DNA upon binding. Moreover, the effect of BD2 on the elongation by telomerase was examined in vitro. These results suggest the possible involvement of heterogeneous nuclear ribonucleoprotein D in maintenance of the telomere 3′-overhang either through protection of a single-stranded DNA or destabilization of the potentially deleterious quadruplex structure for the elongation by telomerase.

Heterogeneous nuclear ribonucleoprotein D, also known as AUF1, has two DNA/RNA-binding domains, each of which can specifically bind to single-stranded d(TTAGGG) n , the human telomeric repeat. Here, the structure of the C-terminal-binding domain (BD2) complexed with single-stranded d(TTAGGG) determined by NMR is presented. The structure has revealed that each residue of the d(TAG) segment is recognized by BD2 in a base-specific manner. The interactions deduced from the structure have been confirmed by gel retardation experiments with mutant BD2 and DNA. It is known that single-stranded DNA with the telomeric repeat tends to form a quadruplex and that the quadruplex has an inhibitory effect on telomere elongation by telomerase. This time it is revealed that BD2 unfolds the quadruplex of such DNA upon binding. Moreover, the effect of BD2 on the elongation by telomerase was examined in vitro. These results suggest the possible involvement of heterogeneous nuclear ribonucleoprotein D in maintenance of the telomere 3-overhang either through protection of a single-stranded DNA or destabilization of the potentially deleterious quadruplex structure for the elongation by telomerase.
Heterogeneous nuclear ribonucleoprotein (hnRNP) 1 D, also known as AUF1, was isolated from a HeLa cell nuclear extract as a protein that binds specifically to single-stranded d(T-TAGGG) n , the human telomeric DNA repeat (1). It also binds to r(UUAGGG) n (1) and the AU-rich element of the 3Ј-untranslated region of mRNA (2). It was shown biochemically that hnRNP D binding to the Gua-rich telomeric strand d(T-TAGGG) n destabilizes intrastrand Gua:Gua pairing and that hnRNP D interacts specifically with telomerase in human cell extracts (3). Thus, the involvement of hnRNP D in the maintenance of telomere DNA has been implied.
hnRNP D consists of 306 amino acid residues and comprises two ribonucleoprotein (RNP)-type DNA/RNA-binding domains (BDs), BD1 (70 -173) and BD2 (174 -256), and a region rich in glycine and arginine residues (257-306) (4). The RNP-type BD is one of the most common eukaryotic protein sequence motifs for DNA/RNA binding (5), being found in hundreds of proteins (6 -9). A single BD of hnRNP D, either BD1 or BD2, is able to bind to DNA and RNA in a sequence-specific manner (4,10,11). The binding affinity of either BD1 or BD2 is comparable with that of the protein having both BD1 and BD2 (4). These results suggest that the basis of the interactions of hnRNP D with DNA and RNA can be addressed by examination with a single BD. We have already reported the structures of BD1 (10) and BD2 (11). We also qualitatively identified the surfaces of BD1 (10) and BD2 (11) interactive with DNA and RNA on chemical shift perturbation analysis, although it should be kept in mind that the perturbation is caused not only through the direct interaction but also thorough the indirect effect of the interaction. Essentially the same surfaces of the BDs are used for the interactions with DNA and RNA.
Here, we present the structure of BD2 complexed with d(T-TAGGG) determined by NMR. The interactions are decisively identified. The identified interactions have been fully confirmed through gel retardation experiments with series of mutant BD2s and mutant DNAs. It is known that the end of the chromosome possesses a 3Ј-overhang of the Gua-rich telomeric strand. The complex structure at atomic resolution reveals how the single-stranded Gua-rich telomeric strand is specifically recognized by hnRNP D.
We have reported that hnRNP D has the ability to inhibit the formation of the RNA quadruplex structure (10). Here, based on CD and NMR data, we demonstrate that BD2 can unfold the quadruplex of Gua-rich telomeric DNA into a single-stranded form upon binding. The effect of hnRNP D on telomerase activity is also examined. The results suggest how hnRNP D is involved in the maintenance of telomere DNA.

EXPERIMENTAL PROCEDURES
Preparation of hnRNP D BD2 and Its Mutants-Non-labeled, 15 Nlabeled, and 13 C, 15 N-labeled BD2s (amino acid residues Lys-176-Glu-261 of hnRNP D) were expressed and purified as reported previously (11). Site-directed mutagenesis was accomplished with a QuikChange mutagenesis kit (Stratagene) according to its protocol. Mutations were confirmed by sequencing. Mutant BD2s with 6 His residues at the * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The N-terminal were expressed and purified with a Ni affinity column (Qiagen).
Preparation of DNA Oligomers-Non-labeled DNA oligomers, synthesized with a DNA synthesizer and purified by reverse phase HPLC, were purchased (Nippon Seihun) and further treated on a Dowex 50 cation exchange column as described previously (12). 13 C, 15 N-labeled d(TTAGGG) and d(GTTAGGGTTA) were synthesized by the hairpin extension method, using a Klenow fragment (3Ј-5Ј exo Ϫ ) (Daiichi Kagaku) with 13 C, 15 N-labeled dNTPs (Nippon Sanso), and purified following the reported method (13).
For CD experiments, d(TTAGGG), d(TTAGGGTTAGGG), and d(T-TAGGGTTAGGGTTAGGGTTAGGG) (15-60 M) were dissolved in a solution comprising 20 mM sodium phosphate (pH 6.0) and 150 mM KCl. BD2 was added step by step to each DNA sample. CD spectra were recorded with a JASCO J-720 spectropolarimeter with a 1-mm cell. From the CD spectrum of the DNA⅐BD2 complex, that for the corresponding protein was subtracted.
Structure Determinations-1474 distance constraints (112 intraresidue, 386 sequential, 233 medium-range, and 561 long-range distance for BD2, 62 intraresidue and 25 sequential distance for d(TTAGGG), and 95 BD2-d(TTAGGG) intermolecular distance) were obtained for the complex from NOESY, 15 N-edited NOESY-HSQC, and 13 C-edited NOESY-HSQC spectra as described previously (10,11,16). 119 backbone dihedral angle constraints were obtained from an HNHA experiment and TALOS (17) based on analysis of 1 H␣, 13 C␣, 13 C␤, and 15 N chemical shifts. Eight dihedral angle constraints for 1 were also obtained, together with stereospecific assignment of the ␤-methylene protons of AMX residues and the ␥-methyl protons of valine residues. Additionally, in the later stage of the calculations, 48 distance constraints for 24 hydrogen bonds were included for slowly exchanging amide protons within the identified secondary structure elements and imino protons of bases.
Structure calculations were carried out using distance and dihedral angle constraints (1649 constraints in total) with a simulated annealing protocol supplied with X-PLOR Version 3.8 (18 -20). A final set of 20 structures was selected from 200 calculations on the basis of the criteria of the smallest residual energy values. None of them violated the distance constraints by more than 0.4 Å or dihedral angle constraints by more than 5°. A mean structure was obtained by averaging the coordinates of the 20 structures, and a restrained energy-minimized mean structure was obtained by energy minimization of the mean structure under the constraints. The quality of the structures was evaluated with PROCHECK (21).
Telomerase Assay-Human telomerase was reconstituted as described previously (22). Human telomerase reverse transcriptase (hTERT) was prepared with a T7-coupled transcription/translation system (Promega). Human telomerase RNA (hTR) was transcribed from template DNA. hTERT and hTR were mixed to yield the human telomerase. Primer extension reaction with the reconstituted telomerase was carried out as described previously (23,24) with and without hnRNP D BD2 in a solution comprising 50 mM Tris-HCl (pH 7.8), 1 mM MgCl 2 , either 0 or 100 mM KCl, 5 mM ␤-mercaptoethanol, 1 mM spermidine, 1 M primer, 1.25 M [␣-32 P]dGTP (800 Ci/mmol), 1 mM dATP, and 1 mM dTTP at 30°C for 90 min. d(GGGATTGGGATTGGGATTGGGTT) was sed as a primer (25). The extension products were purified by phenolchloroform extraction and ethanol precipitation. The products were separated on a 10% polyacrylamide gel containing 8 M urea and detected with a FLA 2000.

Resonance Assignments and Structure Determination of the
Complex-Sequential assignments of the main chain and side chain 1 H, 13 C, and 15 N resonances of hnRNP D BD2 complexed with d(TTAGGG) were made in the same way as reported for BD1 (10) and BD2 (11) in a free state. The assignments of the main chain H N and N resonances, together with those of H ␦2 and N ␦2 of Asn residues, are presented in red in Fig. 1; the assignments in the free state are also shown in black for reference. The BD2 residues that exhibited large chemical shift perturbation on complex formation and intermolecular NOEs to d(TTAGGG) are labeled. Many of these residues are directly involved in the recognition of telomere DNA, as described below. Sequential assignments of the resonances of d(TTAGGG) in the complex were made with the standard method as reported for other DNA (26) and RNA (27). Because of the 10% excess of d(TTAGGG) over BD2 in the solution, exchange cross-peaks between resonances corresponding to the free and complex states were observed for d(TTAGGG) in NOESY and ROESY spectra (data not shown). The resonance assignments of d(T-TAGGG) in the complex were confirmed by these exchange cross-peaks, the resonance assignments of free d(TTAGGG) being accomplished beforehand. On the basis of resonance assignments, many intermolecular NOEs were identified ( Fig. 1, The structure of the complex was calculated on the basis of distance and dihedral angle constraints. The structural statistics are shown in Table I Structure of the Complex- Fig. 2A shows a stereo view of superposition of the 20 final structures of the complex. Fig. 2B shows the restrained energy-minimized mean structure of the complex. The overall structure of BD2 in the complex is essentially the same as that in the free state. BD2 in the complex exhibits a typical RNP-type fold, i.e. a four-stranded antiparallel ␤-sheet (␤1, Lys-183-Gly-188; ␤2, Val-208-Pro-214; ␤3, Phe-225-Phe-230; and ␤4, Ser-250-Lys-255) packed against two ␣-helices (␣1, Glu-195-Phe-205; and ␣2, Glu-233-Glu-241), as observed for BD2 in the free state (10). Additionally, a short ␤-strand (␤4 Ϫ , Tyr-244-Val-247) was identified in loop 5 located between ␣2 and ␤4. A ␤-bulge structure was found in the Val-208-Ser-210 region of ␤2. These features were found in the free state as well. The backbone r.m.s.d. between the structures in the free and complex states of BD2 for the Val-181-Ser-259 region was 2.4 Å.
The complex structure reveals that the Thy-2-Gua-4 segment of single-stranded d(TTAGGG) is specifically recognized by BD2 (Fig. 3). On the other hand, the Thy-1, Gua-5, and Gua-6 residues are located away from the interactive ␤-sheet surface of BD2, with no interaction with BD2. In fact, no intermolecular NOE to BD2 was observed for these residues.
Thy-2 is recognized by three hydrogen bonds with BD2, T2O2-K255H 3 , T2H3-E253O⑀1(O⑀2), and T2O4-Y244H (Fig. 3A). Of the 2 Thy residues, an H3 resonance was observed only for the Thy-2 residue, which is consistent with the formation of the hydrogen bond. Ade-3 is recognized through the stacking interaction with Phe-185, and by the hydrophobic interaction of A3H2 with A257C␤H 3 and M258C␥H 2 (Fig. 3B). Gua-4 is recognized through the stacking interaction with Phe-227 and by two hydrogen bonds, G4H1-M258OЈ and Gua-4O6-Lys-183 H 3 (Fig. 3C). Among the 3 Gua residues, an H1 resonance was observed only for the Gua-4 residue, which is consistent with the formation of the hydrogen bond. It is notable that to make the interactions possible, Gua-4 takes on the syn conformation, which was directly supported by the experimental observation of a very strong G4H8-G4H1Ј NOE. The other DNA residues take on the anti conformation. Additionally, the electrostatic interaction between the 5Ј-phosphate group of Gua-4 and the guanidinium group of Arg-223 was found.
Confirmation of the Interactions through Gel Retardation Experiments and Chemical Shift Perturbation-The interactions identified in the complex were biochemically confirmed through gel retardation experiments using mutant BD2s. First, we recorded either a CD or 1 H-15 N HSQC spectrum of each mutant BD2 and then confirmed that all mutant BD2s retain similar overall folding to that of the wild type BD2. A drastic decrease in the intensity of the band corresponding to the protein⅐DNA complex was observed for the F185A and F227A mutants (Fig. 4), which confirms the importance of the Ade-3-Phe-185 and Gua-4-Phe-227 stacking interactions. Similarly, the large decreases observed for the E253A, A257G, and R223A mutants are consistent with the importance of the hydrogen bonding interaction between Thy-2 and Glu-253, the hydrophobic interaction between Ade-3 and Ala-257, and the electrostatic interaction between Gua-4 and Arg-223. The involvement of Met-258, Lys-183, Tyr-244, and Lys-255 in the interactions was also supported by the decreases observed for their mutants. As a control, Asn-217, which is not involved in the interaction in the complex structure, was mutated to an Ala residue. As expected, no decrease in the intensity was observed for N217A (Fig. 4).
Gel retardation experiments were also carried out with mutant DNAs and wild type BD2. It was expected that replacement of the third Ade residue in the d(TTAGGG) unit of the DNA with a Gua residue would abolish the hydrophobic interactions with A257C␤H 3 and M258C␥H 2   tern of the aromatic resonances in the complex state was distinct from that in the free state. Then, each residue of d(T-TAGGG) was replaced with a Cyt residue, and the spectral pattern of BD2 complexed with the mutant DNA was analyzed. When d(TCAGGG), in which Thy-2 is replaced with a Cyt residue, was added to BD2, a spectral pattern corresponding to free BD2 was observed in addition to one corresponding to complex BD2, the molar ratio of free:complex being ϳ1:2 (data not shown). This indicates that some fraction of BD2 exists in a free state, due to the decrease in the affinity caused by the replacement of Thy-2 with a Cyt residue. The same phenomenon was observed when either Ade-3 or Gua-4 was replaced with a Cyt residue. Thus, the importance of the interactions involving Thy-2, Ade-3, and Gua-4 residues was confirmed. When Thy-1, Gua-5, or Gua-6 was replaced with a Cyt residue, in contrast, the spectral pattern of BD2 complexed with a mutant DNA was the same as that of BD2 complexed with the wild type DNA, no trace of the spectral pattern of free BD2 being detected (data not shown). This indicates that the interactions involving Thy-1, Gua-5, and Gua-6 are dispensable for the complex formation, if any.
Unfolding of the Telomeric DNA Quadruplex on Binding of hnRNP D-It is known that DNA with the telomeric sequence tends to form a quadruplex under physiological ionic conditions. We have reported that hnRNP D has the ability to inhibit the formation of an RNA quadruplex (10). Here, we examined whether hnRNP D unfolds a preformed DNA quadruplex with the telomeric sequence or not. Under physiological ionic conditions, three DNAs, d(TTAGGG) (6-mer), d(TTAGGGTTAGGG) (12-mer), and d(TTAGGGTTAGGGTTAGGGTTAGGG) (24-mer), gave a positive CD peak at either 260 -265 or 290 nm (Fig. 5, A-C), both of which are established marker peaks for a quadruplex (28,29). It was supposed that 6-, 12-, and 24-mer form a tetrameric parallel quadruplex, a dimeric antiparallel quadruplex, and a monomeric antiparallel quadruplex, respectively. Recently, the formation of a dimeric parallel quadruplex (30, 31) and a monomeric parallel quadruplex (12,26,32,33) has been reported for the related sequences. These findings imply the possibility that 12-mer may form a dimeric parallel quadruplex as well and that 24-mer may form a monomeric parallel quadruplex to some extent. Nonetheless, the possible structures are all quadruplexes, and a positive CD peak at either 260 -265 or 290 nm can still be used as a marker peak for quadruplexes. When hnRNP D BD2 was added to the DNA solution (15-60 M), the intensity of these peaks decreased (Fig. 5, A-C). These results reveal that hnRNP D unfolds the preformed quadruplex of telomeric DNA on binding.
For the 6-mer, the intensity of the marker peak at 260 -265 nm decreases to half when the molar ratio of BD2 to DNA is four (Fig. 5A). For the 12-mer, the intensity of the marker peak at 260 -265 nm decreases to half when the molar ratio of BD2 to DNA is eight (Fig. 5B). The 12-mer comprises two d(T-TAGGG) sequences, so when the quadruplex is unfolded, two d(TTAGGG) sequences/12-mer DNA are released for binding of BD2. If this is taken into account, the extent of unfolding of the quadruplex can be regarded as comparable for the 6-and the 12-mer. For the 24-mer, the intensity of the marker peak at 290 nm decreases to half when the molar ratio of BD2 to DNA is 16:32 (Fig. 5C). Because the 24-mer comprises 4 d(TTAGGG) sequences, the extent of unfolding of the 24-mer quadruplex can also be regarded as roughly comparable with that of the 6and 12-mer quadruplexes. Thus, the unfolding of the quadruplex is observed to the same extent for longer DNA.
Unfolding of a telomeric DNA quadruplex by hnRNP D was revealed in other experiments. The chemical shift perturbations of hnRNP D BD2 caused by the addition of d(TTAGGG) in a single-stranded form and a quadruplex form were compared. In the first experiment, the d(TTAGGG) to be added was kept in a solution containing neither Na ϩ nor K ϩ ions, both of which are known to stabilize a quadruplex. The absence of imino proton resonances of the 3 Gua residues confirmed that d(T-TAGGG) takes on a single-stranded form (data not shown). BD2 was also kept in a solution lacking Na ϩ and K ϩ . In the second experiment, d(TTAGGG) to be added was kept in a solution containing 150 mM K ϩ . Formation of the quadruplex under the conditions was confirmed by the observation of imino proton resonances of the 3 Gua residues at 10.5-11.5 ppm (data not shown), which is characteristic of the quadruplex. BD2 was kept in a solution containing 150 mM K ϩ in the second experiment. The HSQC spectrum of BD2 obtained after the addition of a single-stranded form of d(TTAGGG) in the first experiment ( Fig. 6B) was almost identical to that obtained after the addition of a quadruplex form of d(TTAGGG) in the second experiment (Fig. 6C), both HSQC spectra being clearly different from the HSQC spectrum of free BD2 (Fig. 6A). If BD2 bound to the quadruplex form of d(TTAGGG) in the second experiment, the magnetic environment of BD2 must be to some extent different from that bound to the single-stranded d(TTAGGG) in the first experiment. Then, the corresponding HSQC spectra should be different from each other. Thus, the observation of the nearly identical HSQC spectra in the two experiments reveals that the preformed quadruplex DNA with the telomeric sequence is unfolded into a single-stranded form for interaction with BD2 under physiological ionic conditions. Similarly, when the 24mer in either a single-stranded or quadruplex form was added to BD2, many common spectral features were observed in the two resultant HSQC spectra (Fig. 6, D and E). This indicates that the quadruplex of the 24-mer is also unfolded to some extent on interaction with BD2. In this way, unfolding of the It is noticed that in Fig. 6, panels D and E are not completely identical. The difference partly arises from disappearance of peaks for Fig. 6E, although some of the missing peaks can be detected when the level of the plot for Fig. 6E is lowered. Except for the disappearance, the spectral difference is still present between panels D and E. This may suggest the possibility that a small fraction of BD2 binds to the quadruplex of 24-mer, although further certification is needed.
Primer Extension Assay of the Effect of hnRNP D on Telomerase Activity-Primer extension reaction with the reconstituted telomerase was carried out (Fig. 7A). The primer was successfully extended in the absence of K ϩ (lane 1), whereas the extension was inhibited in the presence of K ϩ (lane 2). The formation of the quadruplex in the presence of K ϩ is supposed to be responsible for the inhibition. Then hnRNP D BD2, which can unfold the quadruplex, was added to the solution, but the extension was still inhibited (lane 3).
Next, the effect of hnRNP D BD2 on the extension was examined in the absence of K ϩ (Fig. 7B). It was found that the addition of hnRNP D BD2 led to a reduction in the extension (lane 2).

Comparison with Other
Telomere-binding Proteins-The complex structures of proteins that bind to a single-stranded DNA with the telomeric sequence have been reported for TEBP of Oxytricha nova (34), Cdc13 of Saccharomyces cerevisiae (35), Pot1 of Schizosaccharomyces pombe (36), and human hnRNP A1 (37). The complex structure of hnRNP D BD2 was compared with these four structures. The five complexes share the common feature that single-stranded telomeric DNA binds to the ␤-sheet surface of each protein with either an OB fold or RNP (or RRM) fold. Aromatic rings are characteristically exposed outwards from the ␤-sheet surface of OB and RNP folds, and the rings undergo stacking interactions with DNA bases in all cases. However, great diversity was seen in the mode of specific recognition of telomeric DNA. In the case of TEBP, three OB folds cooperatively recognize DNA. In the case of Cdc13, a greatly expanded interface with the involvement of a large loop is utilized for the recognition. In the case of Pot1, an unusual DNA structure itself plays a critical role in the recognition. In the cases of hnRNP A1 and hnRNP D, the relatively restricted interface of a single binding domain can independently recognize a single d(TTAGGG) unit.
The target sequences of TEBP and Cdc13 in the complexes contain only Gua and Thy residues. Although an Ade residue is present in the target sequence of Pot1, this Ade residue is not recognized by a protein residue but by other DNA residues (36). Thus, the hnRNP D complex and the hnRNP A1 complex are the only cases where the recognition of the Ade residue of DNA by a protein has been demonstrated. In both complexes, there is a stacking interaction between the Ade residue and the Phe residue. The difference is that the Ade residue of DNA is recognized through the hydrophobic interaction in the hnRNP D complex as mentioned above, whereas it is recognized mainly through hydrogen bonds in the hnRNP A1 complex (37).
An interesting difference between hnRNP A1 and hnRNP D is that only the Thy-2-Gua-4 segment of the d(TTAGGG) unit is recognized by hnRNP D, whereas either the Thy-2-Gua-5 or Thy-1-Gua-5 segment is recognized by hnRNP A1 (37). Exclusive recognition of the Thy-2, Ade-3 and Gua-4 residues by hnRNP D was established by three different lines of evidence: 1) The chemical shift perturbation on complex formation with hnRNP D BD2 was remarkable for the 3 residues, whereas it was moderate for the other residues, if any; 2) almost all intermolecular NOEs with BD2 were detected for the 3 residues (24 NOEs for Thy-2, 33 NOEs for Ade-3, and 34 NOEs for Gua-4), whereas almost no NOEs were observed for the others (only 5 NOEs for Thy-1); and 3) a decrease in the binding affinity was found through the detection of free BD2 on NMR when each of the 3 residues was mutated to a Cyt residue, whereas such a decrease was not found when the others were mutated. It should be noted that the non-recognition of Gua-5 is not due to the use of a short DNA, a 6-mer. In fact, when a longer DNA, d(GTTAGGGTTA), was titrated with BD2, the chemical shift perturbation was again remarkable for the TAG segment, whereas it was rather moderate for the Gua residue following this segment (data not shown), as was observed for the 6-mer. Moreover, we have already reported that the chemical shift perturbation of hnRNP D on complex formation with a 24-mer comprising four repeating units is essentially the same as that with a 6-mer (10), which indicated that the length of the 6-mer is enough to examine the interactions.
The difference in recognition found in the complex structures of hnRNP A1 and hnRNP D is consistent with the biochemical finding with dimethyl sulfate (DMS) and P1 nuclease footprints that Gua-4 and Gua-5 of the d(TTAGGG) unit are protected by hnRNP A1 from methylation by DMS, whereas only Gua-4 is protected by hnRNP D from methylation by DMS and cleavage by P1 nuclease (even Gua-4 is not protected under some conditions) (3).
The difference in recognition between hnRNP A1 and hnRNP D may be rationalized on the basis of the difference in the primary sequence. The 5Ј-phosphate of the fifth Gua residue of the second d(TTAGGG) unit of d(TTAGGGTTAGGG) is electrostatically recognized by Arg-140 of hnRNP A1 BD2 (37). The corresponding residue in the primary sequence of hnRNP D BD2 is Asn-217, which cannot undergo the electrostatic interaction. This may account for the difference in the recognition found between the hnRNP A1 and hnRNP D complexes. Moreover, the fifth Gua residue of the first d(TTAGGG) unit of d(TTAGGGTTAGGG) is recognized through the hydrogen bond with Arg-92 of hnRNP A1 BD1 (37). The corresponding residue of hnRNP D BD1 is Met-175, which cannot undergo a similar hydrogen bonding interaction. This may explain the difference in the degree of protection between hnRNP A1 and hnRNP D found in the dimethyl sulfate footprint experiment. The fact that Gua-5 of the d(TTAGGG) unit, in addition to Gua-6, is not covered by hnRNP D but exposed to the solvent may suggest that the access of telomerase RNA and/or other telomere-binding proteins to telomere DNA is comparatively easier for the hnRNP D-bound form than the hnRNP A1-bound form. This idea may imply diversity in the details of the involvement in telomere maintenance of homologous proteins hnRNP D and hnRNP A1.
Biological Implications of Unfolding of the Telomeric DNA Quadruplex by hnRNP D-A Gua-rich strand of minisatellite DNA Pc-1 consists of tandem repeats of d(CAGGG). We found that the Gua-rich strand forms a quadruplex and causes the arrest of DNA synthesis under physiological conditions (38). The link was implied between the arrest and hypermutable features of Pc-1 and other minisatellite DNA with similar repetitive units. We also found that UP1, a proteolytic product of hnRNP A1, unfolds the quadruplex of Pc-1 and abrogates the arrest of DNA synthesis at the d(GGCAG) repeats (39). Thus, the biological significance of unfolding of the quadruplex in DNA synthesis was revealed. Now, it is known that DNA with the telomeric sequence easily forms a quadruplex under physiological ionic conditions and that the formation of the quadruplex interferes with the elongation of telomere DNA by telomerase (40,41). Here, we have found that hnRNP D BD2 unfolds the quadruplex of telomeric DNA. These facts suggest the possible involvement of hnRNP D in maintenance of the telomere DNA. In analogy with the role of UP1 as a destabilizer of the deleterious quadruplex in DNA synthesis, hnRNP D may abrogate the interference with the elongation of telomere DNA through destabilization of the telomere quadruplex. By unfolding the quadruplex into a single-strand, hnRNP D may provide the environment for telomerase to function. In this case, hnRNP D functions as a chaperon for DNA, a DNA chaperon that transforms "improperly folded" DNA into "properly folded" DNA for successful elongation by telomerase.
The effect of hnRNP D BD2 on telomerase activity was experimentally examined by means of a primer extension assay in vitro. Consistent with previous reports (40,41), the quadruplex formation in the presence of K ϩ caused interference with the elongation by telomerase (Fig. 7A, lane 2). We expected hnRNP D BD2 would abrogate the interference by unfolding the quadruplex, but this was not the case (Fig. 7A, lane 3). The extent of unfolding might not be large enough to abrogate the interference under the current assay conditions. Another possibility is that BD2 bound to an unfolded single-stranded form might hinder the access of telomerase to DNA. To examine this point, the effect of hnRNP D BD2 was studied in the absence of K ϩ (Fig. 7B). The quadruplex is not formed under these conditions. When BD2 was added, reduction of the elongation was observed (Fig. 7B, lane 2). This suggests that BD2 bound to single-stranded DNA expelled the telomerase from the DNA and/or prevented the telomerase from gaining access to and proceeding along the DNA, resulting in reduction of the elongation in vitro.
It is instructive to refer to the situation for hnRNP A1. We revealed that UP1 unfolds the quadruplex of telomeric DNA (39), which was later supported by another group (42). It was also found that hnRNP A1 interferes with the telomere elongation by telomerase in vitro (24). Thus, hnRNP D and hnRNP A1 (UP1) share two key features. It has been reported that a mouse cell line deficient in hnRNP A1 expression harbors telomeres that are shorter than those of a related cell line expressing normal levels of hnRNP A1. Restoration of hnRNP A1 expression in hnRNP A1-deficient cells increased the telomere length. Telomere elongation was also observed upon the introduction of exogenous UP1 (43). This report clearly indicated the positive effect of hnRNP A1 on telomere elongation in vivo. The apparent discrepancy between the in vitro and in vivo results may partially be explained by the fact that binding of hnRNP A1 to single-stranded telomere DNA provided protection against nucleolytic degradation (24). This protection may contribute to the positive effect on telomere elongation. The same thing can be assumed for hnRNP D. Another possible interpretation of the apparent discrepancy for hnRNP A1 is that a not yet identified cellular factor, which is not present in the in vitro system, may be required for hnRNP A1 to exert the positive effect on telomere elongation in vivo. For example, a factor that promotes the dissociation of hnRNP A1 from DNA after unfolding of the quadruplex might be present in cells to guarantee the access of telomerase to DNA. This kind of factor may also help hnRNP D to exert the positive effect on telomere elongation in vivo as a DNA chaperon.