hnRNP A1 associates with telomere ends and stimulates telomerase activity

RESULTS

UP1 binding to single-stranded human telomeric DNA in solution

The crystal structure of a complex between UP1 and a single-stranded oligonucleotide with two human telomeric repeats (TR2) shows that a UP1 dimer binds to two antiparallel single strands of DNA, and two telomeric repeats function as a minimum binding unit (Ding et al. 1999). We first tested this model in solution using an electrophoretic mobility shift assay (EMSA) with human telomeric DNA oligonucleotides. Consistent with the crystal structure, UP1 or hnRNP A1 formed a single complex with TR2 (data not shown; Fig. 1A). Similar experiments using mutant telomeric DNA repeats and proteins showed that the binding of UP1 to TR2 is sequence-specific and requires both RRMs of UP1 and a minimum of two telomeric repeats, and that similar complexes are formed with RNA and DNA versions of the telomeric repeats (Fig. 2). We also used photochemical cross-linking and confirmed that UP1 binds to TR2 as a dimer in solution, consistent with the crystal structure (see Supplemental Fig. S1 at http://katahdin.cshl.org:9331/papers/Zhang2006RNA).

Another oligonucleotide, TR4, which comprises four repeats, or two minimum binding units, formed two distinct complexes with UP1, depending on the molar ratio of protein to DNA (Fig. 1B). Quantitative analysis showed that there was no synergistic effect observed with these two binding units (see Supplemental Fig. S2 at http://katahdin.cshl.org:9331/papers/Zhang2006RNA).

UP1 binds to and disrupts structured human telomeric DNA

UP1 can unfold G-quadruplexes formed by G-rich mammalian minisatellite DNA, inducing a characteristic conformational change detectable by CD spectroscopy (Fukuda et al. 2002). We tested whether UP1 can similarly unwind higher-order structures formed by human telomeric DNA. We prepared G-quadruplexes in a potassium-containing buffer, according to Sen and Gilbert (1990), who showed that a G-rich sequence with a 5′ extension, when treated to form G4 DNA, yields a product of slower electrophoretic mobility. When oligonucleotides consisting of two telomeric repeats from human, Tetrahymena, or Oxytricha were incubated at high concentrations (400–600 μM) and moderate pH (7.0–8.0), a distinct slow mobility band was observed in each case by native gel electrophoresis (Fig. 3A, lanes 1,2,4,5,7,8). Presumptive G-quadruplex formation by Tetrahymena or Oxytricha telomeric DNA proceeded rapidly, with >90% yield at 55°C (pH 8.0), and 1M K⁺. Under the same conditions, human telomeric DNA formed presumptive G-quadruplexes slowly and in low yield, probably due to its lower G content. Nevertheless, one common feature shared by all three of these higher-order structures is their high stability: Once formed, they were not disrupted by incubation at 60°C or by dilution with TE to a low concentration (10 μM) and physiological salt condition (150 mM K⁺).

We used a mixed-oligonucleotide strategy (Sen and Gilbert 1990) to test whether the observed higher-order structures were indeed G-quadruplexes. The combined oligonucleotides TE1 and TE2, both of which comprise two repeats of Tetrahymena telomeric DNA but differ in the length of their 5′ extension, formed five distinct slowly migrating bands, as expected for tetramolecular G-quadruplexes with four molecules stacking with each other (Sen and Gilbert 1990; Fig. 3A, lanes 4–6). Oxytricha telomeric DNA formed a type of quadruplex similar to that obtained with Tetrahymena telomeric DNA (Sen and Gilbert 1990; Fig. 3A, lanes 7–9). Notably, human telomeric DNA differed from its ciliate counterparts by forming a distinct type of higher-order structure. A mixture of oligonucleotides HO1 and HO2 formed a more compact structure with faster electrophoretic mobility than the ciliate tetramolecular G-quadruplexes (Fig. 3A, lanes 1–3). This structure is unlikely to be a G-G hairpin (Shippen-Lentz and Blackburn 1990; Jarstfer and Cech 2002) or a simple structure involving conventional base pairs, given its resistance to heat and low salt. Crystallographic studies demonstrated that human telomeric DNA can form a dimeric G-quadruplex, in which one molecule with two telomeric repeats folds back to stack with itself (Parkinson et al. 2002). The compact structure we observed with the human telomeric repeats is therefore likely to be a dimeric G-quadruplex.

Next, we tested whether recombinant UP1 could unwind the higher-order structure formed by HO2. The unstructured HO2 formed a single complex with UP1, designated A (Fig. 3B, lanes 1–6). When HO2 was preincubated under conditions leading to partial formation of presumptive G4 DNA, subsequent incubation with increasing amounts of UP1 gave rise to a complex designated B, which runs more slowly than A (Fig. 3B, lanes 7–12). At a low UP1/DNA ratio, both A and B complexes were present, whereas at higher UP1/DNA ratios, B gradually dissociated, apparently changing into A. This result, together with the previous report that UP1 can unfold mouse minisatellite G-quadruplex DNA (Fukuda et al. 2002), indicates that UP1 can bind to, and unwind, the presumptive G-quadruplex formed by human telomeric DNA repeats, suggesting a potential role of UP1 in changing the conformation of mammalian telomeres.

Depletion and add-back of hnRNP A/B proteins in telomerase extracts

Given the ability of UP1 to unwind higher-order telomeric DNA, we next tested whether hnRNP A/B proteins can affect telomerase activity in vitro. We prepared telomerase extracts from HEK293 cells, and tested whether depleting hnRNP A/B proteins interferes with the telomerase reaction (Fig. 4). To this end, we used immobilized, biotinylated single-stranded telomeric DNA repeats to selectively remove these proteins from the extracts (Zhu et al. 2001) and tested the telomerase activity. Because of the use of crude extracts, which may have telomerase inhibitors, and the scale required for depletion/add-back experiments, we could not achieve the sensitivity required to detect telomere extension by a direct telomerase assay (Morin 1989; data not shown). Instead, we relied on a widely used telomeric repeat amplification protocol (TRAP) in which a primer, TS, is elongated by telomerase and the products are amplified by PCR, together with an internal control (Kim and Wu 1997). As confirmed by Western blotting (Fig. 4A), both hnRNP A1 and A2 were undetectable after two rounds of depletion. As a specificity control, we probed for the hnRNP C1/C2 proteins, which were only minimally depleted. Staining for total proteins in the original and depleted extracts showed no substantial changes in the major protein components of the extract (Fig. 4B). In addition, to test whether the telomerase RNP was also depleted, we used RT-PCR to measure the human telomerase RNA (hTR) and found that its abundance did not change substantially before and after depletion (Fig. 4C, lanes 2,3) and no hTR bound to the beads (lane 4). Next, we monitored telomerase activity in the depleted extract using the TRAP assay, including RNase A control treatments to verify hTR-dependence (Fig. 4D). The depletion dramatically reduced telomerase activity (Fig. 4D, lanes 2–5), consistent with a role of hnRNP A/B proteins in promoting telomerase elongation. The presence of similar amounts of PCR products from the internal control excluded the possibility that the depletion had an inhibitory effect on the PCR reaction itself.

To rule out the possibility that limiting factors were codepleted with the hnRNP A/B proteins, we tested whether we could recover telomerase activity by addition of various recombinant hnRNP A/B proteins. First, based on the concentration of hnRNP A1 in the extract (Fig. 4A), we added back a comparable amount of purified recombinant hnRNP A/B proteins to the depleted extract, and measured telomerase activity by the TRAP assay (Fig. 4D, lanes 6–11). Recombinant hnRNP A1 fully restored the telomerase activity to the depleted extract. hnRNP A2, which has 68% amino acid identity with hnRNP A1, gave a similar effect, whereas UP1 had slightly weaker activity. To test the specificity of the effect of these RNA-binding proteins on telomerase activity, we also tested SF2/ASF, an abundant nuclear pre-mRNA splicing factor with two RRMs, or Gene 32 Protein (G32P), a single-stranded-DNA-binding protein from bacteriophage T4. Neither protein restored telomerase activity when added to the depleted extract (Fig. 4D, lanes 12–15). Moreover, the addition of hnRNP A1 to the PCR step of the TRAP assay after the telomerase reaction had no effect, indicating that hnRNP A1 stimulates the telomerase reaction, rather than the PCR amplification step (see Supplemental Fig. S3 at http://katahdin.cshl.org:9331/papers/Zhang2006RNA). Titration of the recombinant protein further showed that hnRNP A1 promotes the telomerase reaction in a concentration-dependent manner (Fig. 4E). Taken together, the hnRNP A/B protein depletion and add-back experiments demonstrate that these proteins can potently stimulate the telomerase-dependent telomere elongation reaction.

The fact that UP1 has substantial telomerase-stimulating activity indicates that the C-terminal glycine-rich domain of hnRNP A1 is not strictly required for this function, although it is essential for some of this versatile protein's other functions, such as the modulation of alternative splicing (Mayeda et al. 1994). Using the affinity depletion/add-back assay, we further tested the domain requirements for the telomerase-stimulatory activity of hnRNP A1. Three RRM mutants of UP1 (Mayeda et al. 1994), in which both phenylalanines of the RNP-1 submotif in RRM1 and/or RRM2 are replaced with aspartates to disrupt nucleic acid binding, lost the ability to restore telomerase activity in the depleted extract (Fig. 5A), correlating with their failure to bind telomeric repeats (Fig. 2D). In contrast, another mutant, UP1-KYAA, in which Lys166 and Tyr167 were simultaneously replaced by alanines, maintained telomere-binding activity by EMSA (data not shown) and retained the ability to restore telomerase activity to the same level as wild-type UP1 (Fig. 5A). Thus, the telomeric DNA-binding activity mediated by both RRMs correlates with the telomerase-stimulating activity of hnRNP A1.

We then tested whether hnRNP A1’s stimulatory activity can counteract the effect of a G-quadruplex-stabilizing compound. The porphyrin derivative N-methyl mesoporphyrin IX (NMM) binds G4 DNA, and specifically inhibits its unwinding by RecQ-family helicases (Huber et al. 2002). NMM inhibited telomerase activity in the HEK293 extract (Fig. 5B, lane 3). However, an excess of hnRNP A1 could completely overcome this inhibition by NMM (Fig. 5B, lane 4). This observation suggests that NMM is unable to block G4 unwinding by hnRNP A1, in marked contrast to the effect of this compound on the RecQ helicases.

hnRNP A1 associates with telomeres in vivo

The above data reveal a stimulatory role of hnRNP A1 in telomerase-dependent telomere elongation. However, it is not known whether hnRNP A1 can access telomere ends in vivo. hnRNP A1 is an abundant protein that localizes in a diffuse nuclear pattern, and thus if a small subpopulation of hnRNP A1 is associated with telomeres, it would be difficult to detect it by immunofluorescence. As telomerase was successfully detected at telomeres by chromatin immunoprecipitation (ChIP) in mammalian cells (Sharma et al. 2003), we carried out ChIP experiments in asynchronous HeLa cells to examine whether hnRNP A1 is physically associated with telomeres in vivo (Fig. 6). Formaldehyde-cross-linked chromatin was sheared by sonication and immunoprecipitated by antibodies against hnRNP A1, TRF2 (positive control), or SF2/ASF (negative control).

Two approaches were used to analyze the enrichment of telomeric DNA repeats in the immunoprecipitated DNA. First, slot-blot analysis was performed, and the presence of telomeric DNA was detected by hybridization with a ³²P-labeled telomeric probe (Fig. 6A). We observed hnRNP A1-dependent and TRF2-dependent, but not SF2/ASF-dependent, immunoprecipitation of telomeric DNA, indicating that hnRNP A1 associates with telomeres in vivo. When a control probe corresponding to human total genomic DNA was used, only the input extract gave a strong positive signal. Another control probe corresponding to centromeric DNA repeats confirmed that the overall DNA amount in each slot of immunoprecipitated DNA was the same.

The slot-blot signals were then quantitated (Fig. 6B). Compared to the SF2/ASF negative control, anti-hnRNP A1 antibody resulted in a four- to sixfold enrichment of telomeric DNA, although this enrichment was not as high as that obtained with the anti-TRF2 antibody (20- to 22-fold). A stronger signal for TRF2 is expected, given that it binds all along the telomeric repeats, whereas hnRNP A1 binding is likely limited to the single-stranded overhangs.

We next used a PCR approach to confirm the above observation. To avoid amplifying the simple repeat sequence of telomere tracts, we designed subtelomeric PCR primers overlapping with the last repeat of the internal end of the telomere tract in human Chromosome 4p16.3 (Fig. 6C). The sonicated chromatin was heterogeneous, ranging from 0.5 to 8 kb, with an average size of 2 kb. The majority of terminal restriction fragment (TRF) lengths in HeLa cells range from 3 to 7 kb (data not shown) including up to 5 kb of nontelomeric sequence. This distribution suggests that the mean length of simple telomeric repeats in HeLa cells is only a few kilobases. Therefore, it is possible to use PCR detection of telomeric signals in immunoprecipitated DNA using subtelomeric primers, even though hnRNP A1 might function along with telomerase at the extreme end of telomeres. A pair of control primers was chosen, corresponding to the same chromosome but 22 kb away from the end of the telomere tract. In addition, we also included as a control an antibody against the SNAP190 subunit of the snRNA-activating protein complex (SNAPc), which is required for both Pol II and Pol III transcription from snRNA promoters (Schramm et al. 2000).

DNA immunoprecipitated by the four antibodies was normalized for equal input based on the slot-blot assay with the centromeric DNA probe. After reversal of the cross-links, DNA from non-cross-linked (Fig. 6D, lanes 1–5) and cross-linked (Fig. 6D, lanes 6–10) samples was amplified with the various primer sets to assess the cross-linking-dependent enrichment and specificity of the protein-bound DNA fragments recovered with each antibody. SNAP190 associated with the U6 core promoter, but not with a region upstream of the U6 promoter, as expected (Schramm et al. 2000), or with the subtelomeric and control regions of Chromosome 4 (Fig. 6D). Both hnRNP A1 and TRF2 associated with DNA fragments spanning the subtelomeric region of human Chromosome 4, but not with the control DNA farther toward the centromere on the same chromosome or with the U6 core promoter and U6 upstream region DNAs. In contrast, SF2/ASF did not show specific association with any of these DNA regions. We conclude that hnRNP A1 is a bona fide telomere-associated protein in vivo.