A Multifunctional RNA Recognition Motif in Poly(A)-specific Ribonuclease with Cap and Poly(A) Binding Properties*

Poly(A)-specific ribonuclease (PARN) is an oligomeric, processive and cap-interacting 3′ exoribonuclease that efficiently degrades mRNA poly(A) tails. Here we show that the RNA recognition motif (RRM) of PARN harbors both poly(A) and cap binding properties, suggesting that the RRM plays an important role for the two critical and unique properties that are tightly associated with PARN activity, i.e. recognition and dependence on both the cap structure and poly(A) tail during poly(A) hydrolysis. We show that PARN and its RRM have micromolar affinity to the cap structure by using fluorescence spectroscopy and nanomolar affinity for poly(A) by using filter binding assay. We have identified one tryptophan residue within the RRM that is essential for cap binding but not required for poly(A) binding, suggesting that the cap- and poly(A)-binding sites associated with the RRM are both structurally and functionally separate from each other. RRM is one of the most commonly occurring RNA-binding domains identified so far, suggesting that other RRMs may have both cap and RNA binding properties just as the RRM of PARN.

(CBC) (6), which in turn influences mRNA splicing (6), 3Ј end formation (7), the nucleocytoplasmic transport (8), and poly(A) degradation (9). In the cytoplasm, CBC is replaced by the cytoplasmic cap-binding protein, also known as eukaryotic translation initiation factor 4E (eIF4E), which together with factors eIF4A and eIF4G, is responsible for initiation of cap-dependent mRNA translation (see Refs. 10 and 11) and references therein). Similarly, the poly(A) tail is recognized in the nucleus by the nuclear poly(A)-binding protein 1 (PABP1), which subsequently is replaced by the cytoplasmic poly(A)-binding protein (PABP) (see Refs. 12 and 13) and references therein). In particular, the cap and the poly(A) tail coordinate and influence protein synthesis and mRNA degradation (reviewed in Refs. 2, 3, 5, and 14). Both structures are recognized during two of the general pathways of eukaryotic mRNA degradation. In the deadenylation-dependent decapping pathway, the cap is recognized and removed after the initial deadenylation step, whereas hydrolysis of the cap is one of the final steps in the deadenylation followed by 3Ј-5Ј degradation pathway.
Several poly(A) degrading activities have been characterized in eukaryotic cells (reviewed in Refs. 2, 3, and 14), e.g. the PAN2/ PAN3 nuclease (15,16), the CCR4⅐Caf-1 complex (17,18), and poly(A)-specific ribonuclease (PARN) (19 -24). Among these, PARN is unique because it interacts directly with both the cap structure and the poly(A) tail during deadenylation (22,23,(25)(26)(27). The cap stimulates the rate of the PARN reaction and thereby amplifies PARN processivity (22,(25)(26)(27). It has been proposed that the cap binding property of PARN could play a critical role in regulating the competition between translation and mRNA degradation (Refs. 2, 5, 9, and 14 and references therein). For example, this property of PARN could abrogate ongoing protein synthesis and target the mRNA for degradation or vice versa could ensure that translation is not initiated on an mRNA already subjected to PARN-mediated degradation.
PARN is an oligomeric enzyme (22), and a recent crystal structure has revealed a dimeric composition (28). The active site of PARN has been characterized both biochemically (29,30) and structurally (28) and is build up by four conserved acidic amino acid residues that coordinate catalytically essential divalent metal ions. In addition to the cap-binding and the active sites, PARN also contains two potential RNA-binding sites. One of these sites belongs to the recently identified R3H class of RNA-binding domains (Refs. 28 and 31 and Protein Data Bank entry 1UG8), whereas the second is a classical RNA recognition motif (RRM) (Protein Data Bank entry 1WHV) (Fig. 1). However, no functional data has yet confirmed that any of these two domains bind RNA; neither has the cap-binding site been identified. It is of utmost importance to identify and characterize the cap-and poly(A)-binding sites of PARN to fully understand how poly(A) specificity is achieved and how PARN integrates its hydrolytic activity with its cap binding properties.
Here we have investigated the cap binding property of PARN, using fluorescence spectroscopy and mutational analysis and identified at least one tryptophan residue within the RRM that is involved in cap binding. We have also characterized the RNA binding properties of the RRM and found that the RRM preferentially binds poly(A) and that the RNA-and cap-binding sites are located in close proximity to each other but not overlapping.
In conclusion, our data show that the RRM of PARN binds both cap and poly(A), suggesting that the RRM is a major element important for proper PARN activity because the RRM is involved in the two unique properties tightly associated with PARN, i.e. poly(A) specificity and cap binding.

EXPERIMENTAL PROCEDURES
Molecular Cloning-A DNA fragment corresponding to amino acids 443-560 in the human PARN sequence, PARN (443-560), was obtained by PCR amplification using plasmid pE 33 PARN (30) as template and the primers (443-560)5Ј containing restriction site for NdeI and (443-560)3Ј containing restriction site for Bpu1102 I. The sequences of used primers are listed in Table S1. The obtained DNA fragment was cloned into pCR 2.1TOPO vector and subsequently subcloned into the pET19b vector between the NdeI and Bpu1102 I sites.
Site-directed Mutagenesis-All of the PARN mutants were generated from pE 33 PARN using a QuikChange site-directed mutagenesis kit (Stratagene) following the manufacturer's protocol. The mutations were introduced by using primers named as the corresponding mutation and with sequences as listed in Table S1. All of the mutations were confirmed by DNA sequencing.
Expression     , and PARN(443-560,W456,475A) polypeptides were expressed from Escherichia coli strain BL21(DE3) as described (32). Soluble recombinant polypeptides were purified using Talon Metal Affinity Resin (Clontech) according to Nilsson and Virtanen (32). The amount of protein was measured using Bio-Rad protein assay kit, and purity was analyzed by SDS-PAGE followed by silver or Coomassie staining. Fluorescence Spectroscopy-Recombinant PARN or PARN mutants expressed in E. coli and purified as previously described (32) (34).
Fluorescence measurements were performed on LS 50 B and LS 55 spectrofluorometers (PerkinElmer Life Sciences) with an automatic correction for the photomultiplier sensitivity, in a thermostatted quartz micro or semi-micro cuvette (Hellma catalog number 104.002F QS and 119.004F QS), at 20.0°C. For all of the measurements, an excitation wavelength of 280 nm was applied (slit 2.5 nm, auto cut-off filter). Titrations based on whole spectra recording (Fig. 2, A and B) were performed for PARN and single Trp 3 Ala mutants at 5 M per monomer by adding 2-l aliquots of m 7 GTP or m 7 GpppG to 400 l of the protein. Each titration consisted of eight emission spectra recorded for different ligand concentrations. The change in the fluorescence intensity was calculated from data points at 320 nm, which provided precise detection of both the intrinsic protein fluorescence quenching and the fluorescence increase caused by the presence of free ligand in the solution toward the end of the titration.
Detailed fluorescence titrations (Fig. 2, C and D) were performed essentially as described previously (35). PARN and its mutants were used at 0.2 and 1 M/monomer, in steady-state conditions provided by preincubation at a given concentration by at least 1 h. Fluorescence changes were monitored continuously with the integration time of 30 s and the gap of 30 s for adding the ligand, with slow magnetic stirring, at a single wavelength of 320 (slit 4 nm, 290 nm cut-off filter) During the gap, the UV xenon flash lamp was switched off to avoid photobleaching of the sample. Aliquots of 1 l of increasing concentrations (10 M to 2 mM) of m 7 GTP or m 7 GpppG were injected manually to 800 l of protein solution. Each titration consisted of 40 data points.
The fluorescence intensities were corrected for the inner filter effect and the dilution of the sample. The values of the dissociation constants, K D , were obtained by fitting the theoretical curve to the experimental data points, F, using the following full equilibrium binding equation, where the actual concentration of the protein⅐ligand complex, [cx], is as follows, and F(0) is the initial fluorescence of the pure protein; [L] is the total ligand concentration; [P] is the protein concentration per monomer; ⌬ is the difference between the fluorescence efficiencies of the apo-protein and the complex; lig-free is the fluorescence efficiency of the ligand. K D and ⌬ were free parameters of the fitting. The values of lig-free were fixed and taken from independent controlling titration experiments in pure buffer. Regressions were performed by means of a nonlinear, least squares method, using PRISM 3.02 (GraphPad Software) or Origin 7 software (OriginLab Corporation). The final K D values were calculated as weighed averages from at least three independent titration series. Circular Dichroism-CD spectra were measured on an Aviv spectropolarimeter (Lakewood, NJ) in 1.00-mm quartz cuvette Preparation of RNA Substrates-L3(A 30 ) RNA substrates with or without m 7 GpppG at their 5Ј ends were synthesized by in vitro transcription as previously described (27). A 5 -A 20 substrates were purchased from Dharmacon Research, Inc. Before usage the substrates were deprotected according to the instructions from the manufacturer. 10 pmol of A 5 -A 20 were 5Ј-labeled with 20 pmol [␥-32 P]ATP (GE Healthcare; catalog number AA0068) using T4 polynucleotide kinase (USB; catalog number 70031Z), and the reactions were incubated in 37°C for 45 min. The labeled oligonucleotides were resolved on 25% acrylamide gels, bands cut out and eluted in water. The final concentrations of labeled oligo(A) were 2.5-25 nM.
PARN Deadenylation Assay-The conditions for the deadenylation assays were 25 mM Hepes-KOH, pH 7.0, 100 mM NaCl, 0.1 g/l bovine serum albumin, and 15 mM MgCl 2 . 10 nM PARN or PARN(E455,W456,475A) monomers were incubated with 50 nM capped or noncapped radioactively labeled L3A 30 RNA substrates. A 20-l reaction was incubated at 30°C, and 1-l aliquots were taken out at indicated time points. The reactions were analyzed, and released AMP products were sep-arated by one-dimensional TLC by spotting 1 l of the reaction on a polyethyleneimine cellulose F plate (Merck; 5579) and using 0.75 M KH 2 PO 4 , pH 3.5 (H 3 PO 4 ), as solvent. The plate was dried, exposed, and scanned by a 400S PhosphorImager (Molecular Dynamics).
Electrophoretic Mobility Shift Assay-10-l reactions were performed in Buffer A (32 mM phosphate buffer, pH 7.0, 0.2 mM dithiothreitol, 100 mM KCl, 0.2 mM EDTA) using 5 nM of oligo(A) RNA and 0.2-8 M of monomeric PARN or PARN mutant. The reactions were incubated for 15 min at room temperature. 5 l of loading dye (8% glycerol, 0.15% bromphenol blue/xylene cyanole) was added to the reaction prior to loading the samples to nondenaturing gel (0.5ϫ TBE, 6% 19:1 acrylamide/bisacrylamide v/v) prerun at 200 V, 5W for 30 min in 4°C. The gels were run for 3 h at 5W in 4°C or using the BioVectis DNA Pointer System for 33 min at constant 30 W at 10°C, dried in a Bio-Rad gel dryer for 1 h, and finally exposed and scanned by a 400S PhosphorImager (Molecular Dynamics).
Filter Binding Assay-Reactions using 32 P-labeled A 5 -A 20 as the RNA were performed as described under "Electrophoretic Mobility Shift Assay." After 15 min of incubation, the entire reaction mixture was bound to a Protran BA 85 Cellulose nitrate membrane (Schleicher & Schuell; catalog number 10 401191) preincubated in ice-cold Buffer A and mounted on a vacuum manifold with no vacuum applied. The membrane was washed with 0.75 ml of ice-cold Buffer A with vacuum applied and dried, and finally the amount of bound protein⅐RNA complex was quantified in a BeckmannCoulter LC6500 scintillation counter. The equilibrium dissociation constant, K D , was obtained by plotting experimental data and fitting curves with nonlinear regression (Origin 7 software, OriginLab Corporation) using the following binding equation, where [P 0 ] and K D were free parameters of the fitting, and [P 0 ] is the active concentration of PARN polypeptides. The dissociation rate constant, k d , of the PARN⅐A 20 complex was determined by setting up reactions as described above. At time 0 a 100-fold excess of cold A 20 was added, and subsequently 10-l aliquots were taken out at different time points and subjected to filter binding assay. The equation was used to calculate the dissociation rate constant, k d . Supplemental Data-Supplemental information includes supplemental Fig. S1 and supplemental Table S1.

RESULTS
The RRM of PARN Binds the Cap-It has been shown that several cap-binding proteins, e.g. the translation initiation factor eIF4E (36 -38), the CBP20 subunit of CBC (39 -41), the DcpS scavenger enzyme (42), influenza A RNA polymerase (43) and, vaccinia virus cap modification enzyme VP39 (44), bind the cap by stacking of the 7-methylguanosine base of the cap structure between aromatic amino acids. PARN contains a large number of aromatic amino acid residues, making it likely that PARN-cap interaction can be investigated by fluorescence spectroscopy, just as it has been done for eIF4E and CBC (35,45). To investigate this, we mixed PARN with m 7 GTP or m 7 GpppG cap analogs and monitored, using fluorescence spectroscopy, the fluorescence emission spectra of PARN after excitation at 280 nm ( Fig. 2A). From the fluorescence spectra we determined, as described under "Experimental Procedures", the equilibrium dissociation constant (K D ) of the PARN-cap interaction ( Table 1). The calculated K D value was in the low micromolar range and in the same range as previously suggested from kinetically determined inhibition constants (27) and at least 10-fold higher than for m 7 GTP or m 7 GpppG binding to CBC or eIF4E (35,45).
A systematic mutational analysis was performed to investigate whether any of the six tryptophan residues in PARN could play a role in cap binding. The tryptophan residues were selected for mutagenesis because only six were present in the PARN sequence, whereas the other aromatic residues were significantly more abundant and because the fluorescence spectra indicated that tryptophan residues could be involved. Six separate PARN mutant polypeptides, wherein one at a time of the tryptophan residues was changed to alanine, were generated, and the cap binding properties of each mutant PARN polypeptide were investigated by affinity to 7-Me-GTP-Sepharose (supplemental Fig. S1) and by fluorescence titrations using m 7 GTP or m 7 GpppG as cap analogs ( Fig. 2 and Table 1). One mutant polypeptide, PARN(W475A), was severely defective in cap binding, whereas a second mutant polypeptide, PARN (W456A), was slightly affected, as visualized by reduced binding to the 7-Me-GTP-Sepharose matrix (supplemental Fig. S1) and by increased calculated PARN-cap K D values ( Fig. 2 and Table 1). In eIF4E, a glutamate residue that immediately follows one of the cap-interacting tryptophans participates together with the tryptophan residue in cap recognition (35)(36)(37). A glutamate residue at position 455 of PARN could potentially play a similar role. Therefore, mutant PARN(E455A) was prepared, and its cap binding properties were investigated. PARN(E455A) showed no defect in its binding to the 7-Me-GTP-Sepharose matrix (data not shown) and a very small increase in the PARNcap K D value ( Table 1). The triple mutant PARN(E455A,W456, 475A) was as expected severely defective in cap binding ( Fig. 2C and Table 1 and supplemental Fig. S1). Taken together these results suggest that Trp 475 plays an essential role in PARN cap recognition, whereas Glu 455 and especially Trp 456 could play auxiliary roles.
The three amino acid residues Glu 455 , Trp 456 , and Trp 475 are all located within a domain of PARN that folds into a classical RRM (Fig. 1) (reviewed in Ref. 46). It was therefore of interest to investigate whether this domain by itself could bind the cap structure. Toward this end we cloned a fragment of PARN (residues 443-560) comprising the RRM. The purified polypeptide PARN(443-560) was capable of binding the cap as revealed by fluorescence titrations and affinity to the 7-Me-GTP-Sepharose matrix with the binding constant, K D , still in the micromolar range (Table 1, Fig. 2, B and C, and supplemental Fig. S1). As expected, the mutant PARN(443-560,W456,475A) polypep-tide was severely defective in cap binding (Table 1, Fig. 2C, and supplemental Fig. S1). Taken together our data reveal that a fragment containing the RRM domain of PARN by itself binds the cap and that mutation of three amino acids within this domain affects cap binding. Our data also suggest that at least one tryptophan residue, i.e. Trp 475 , plays an essential role for cap binding either by interacting directly with the cap structure or participating in conformational changes associated with cap binding.
Cap Binding Is Not Essential for PARN Activity-To investigate whether cap binding and the three amino acids Glu 455 , Trp 456 , and Trp 475 were essential for PARN activity, we compared the deadenylating activities of PARN and PARN (E455A,W456,475A) using capped or noncapped L3(A 30 ) RNA as the substrates. The released AMP products were monitored during a time course experiment, and the results, plotted in Fig. 3, revealed that PARN(E455A,W456,475A) was an active deadenylase showing that the introduced mutations in the RRM domain of PARN did not inactivate PARN activity and suggesting that cap binding, as previously shown (see for example Ref. 27), is not essential for the hydrolytic activity of PARN. Interestingly, recombinant and bacterially produced PARN showed repeatedly a slightly higher activity on capped relative noncapped RNA substrates, whereas such a difference was never observed when using PARN(E455A, W456,475A) as the deadenylase. This suggests that bacterially expressed recombinant PARN to some extent can reconstitute the stimulatory effect the cap structure has on PARN activity (see also Ref. 9), although the effect is not as prominent as it is when using HeLa cell free S100 extracts (26) or PARN from calf thymus extracts and purified to apparent homogeneity (22,25,27). The reason for the inability to fully reconstitute the cap stimulatory effect using recombinant PARN purified from bacteria is not known but could be due to many reasons, e.g. lack of post-translational modifications.
The RRM of PARN Binds Poly(A)-The presence of a classical RRM within PARN suggests that the RRM could bind RNA besides binding the cap. To investigate this, we performed electrophoretic mobility shift assays (EMSA). Fig. 4 shows that both PARN and PARN(443-560) formed stable complexes with 32 Plabeled A 20 oligonucleotides, suggesting that the RRM of PARN binds RNA. The addition of increasing amounts of unlabeled A 20 , A 10 , or A 5 oligonucleotides showed that both complexes were efficiently competed by A 20 , to some extent by A 10 , and not at all by A 5 (Fig. 4). Further competition experiments  showed that the complexes were efficiently competed by poly(A), to some extent by poly(G) and very inefficiently by poly(C) or heteropolymeric single-stranded RNA (Fig. 4). We have not been able to interpret competition experiments using poly(U), because of its poly(A) base pairing property. Finally, we investigated the oligonucleotide length requirement for protein⅐RNA complex formation using EMSA and filter binding assay (see "Experimental Procedures") and found that more than 10 adenosine residues were required for efficient PARN-RNA or PARN(443-560)⅐RNA complex formation using EMSA (Fig. 5, A and C), whereas the filter binding assayed revealed a significant drop in the K D value for the PARN(443-560)-RNA interaction when the oligonucleotide was 10 nucleotides or longer (Fig. 6). To sum up, these data suggest that the RRM of PARN, besides binding the cap, binds RNA with a preference for poly(A) and that at least 10 adenosine residues are required for efficient poly(A) binding. Furthermore, the unaffected RNA binding properties of the PARN(443-560,W456, 475A) mutant polypeptide relative PARN(443-560) (Fig. 5, C and D) also suggest that the introduction of these two mutations does not severely affect the global fold of the RRM. Thus, we predict that the reduced affinity to the cap of the mutated RRM is related to a change of direct, local interactions with the cap or to an indirect dependence on the mutated residues rather then being caused by a major conformational change of the mutated polypeptide. In support of this we did not observe any significant differences in the CD spectra of PARN(443-560) and PARN(443-560,W456,475A) (Fig. 2F).
The Cap and RNA-binding Sites of the RRM Are Structurally and Functionally Separated from Each Other-The capacity of the PARN RRM to bind both cap and poly(A) implies that the two binding sites are located close to each other. However, it is not clear whether they are completely overlapping or whether they are distinct from each other. Furthermore, the two sites may very well be structurally separated but still functionally influence each other. For example, binding of the cap may pre-vent RNA binding or vice versa. To address these issues we performed a number of experiments. First, we investigated the RNA binding properties of the cap binding-deficient RRM mutant PARN(443-560,W456,475A) using EMSA. Fig. 5 and Table 2 show that PARN(443-560,W456,475A) has RNA binding properties similar to those of PARN, PARN(E455A,W456, 475A), or PARN(443-560), suggesting that the tryptophan residues essential for cap binding are not required for RNA binding neither by PARN nor by the PARN RRM. Subsequently, we determined, using a filter binding assay (see "Experimental Procedures"), the K D values for complex formation between A 20 and PARN, PARN(443-560) or cap binding-deficient mutants thereof. The K D values are summarized in Table  2 and show once again that the two tryptophan residues are not required for poly(A) binding. From these experiments we conclude that (i) the two investigated binding sites, i.e. the capbinding and the poly(A)-binding sites, are structurally separated from each other because they are not dependent on the same molecular determinants (i.e. tryptophan residues 456 and 475) and (ii) the cap binding property of the PARN RRM is not a prerequisite for poly(A) binding.
Next we investigated whether the two sites could functionally influence each other. Toward this end we studied whether the cap could influence RRM⅐A 20 complex formation. First, we performed EMSA where the PARN-A 20 or PARN (443-560)⅐A 20 complex was formed in the presence of increasing amount of cap analog (Fig. 7). We could not detect any effect on the RNA⅐protein complex formation that was induced by the cap even if we increased the cap concentration 500-fold above the K D value for the PARN-cap interaction. Subsequently, we investigated, using the filter binding assay, whether the K D values for the PARN-A 20 or PARN(443-560)⅐A 20 complexes were affected in the presence of 50 M cap analogs and found once again no evidence that the cap interfered with RNA binding (Table 2). Thus, none of these two assays could provide any evidence for the cap influencing RNA binding, suggesting that the two binding sites, besides being structurally distinct, are functionally separated from each other. However, the K D value represents the ratio between the dissociation rate constant (k d ) and the association rate constant (k a ). Thus, it remained possible that the cap affects both these rate constants proportionally without influencing the ratio between them. We therefore determined, as described under "Experimental Procedures," the dissociation rate constant k d for the PARN-A 20 or PARN(443-560)⅐A 20 complexes in the absence or presence of 50 M cap analog. The results are summarized in Table 2 and show that the dissociation rate constant was not affected by the presence of cap, providing further support for the proposal that cap and RNA binding of the RRM are functionally separated from each other. Interestingly, the dissociation rate constants for the PARN-A 20 and PARN(443-560)⅐A 20 complexes differed significantly ( Table 2). The basis for this difference is at present not known, but the difference suggests that there must be at least one additional RNA interacting element in PARN that is not present in PARN(443-560) and that contributes to the faster kinetics of both association and dissociation of the PARN⅐A 20 complex, because the equilibrium K D values remain unchanged. The active site and/or the R3H RNA-binding domain are two likely RNA interacting elements that very well could contribute to the enhanced kinetic rate constants of PARN relative to the RRM.

DISCUSSION
In this report we show that the RRM of PARN harbors both poly(A) and cap binding properties, suggesting that the PARN RRM can bind the two boundary marks, i.e. the cap structure and the poly(A)) tail, that define the extreme borders of a eukaryotic mRNA. Furthermore, our analysis strongly suggests that the two binding sites are separate from each other both functionally, because they cannot influence each other activities, and structurally, because they rely on different molecular determinants. Based on our studies, we therefore propose that the RRM of PARN plays a pivotal role in the two critical and unique properties that are tightly associated with PARN activity, i.e. recognition and dependence on both the cap structure and poly(A) tail during poly(A) hydrolysis. Interestingly, it has previously been shown that PARN activity can be recovered from mutant PARN polypeptides lacking the RRM (28), although such polypeptides are significantly less active than full-length PARN. Thus, the RRM of PARN is not the only domain responsible for adenosine specificity and efficient hydrolytic activity, even if our data strongly suggest that the RRM of PARN is a major structural and functional element important for proper PARN activity. In particular, the RRM is required both for cap and poly(A) binding and thereby contributes to the specificity of the enzyme. To fully understand the role of the RRM, it is crucial to solve a structure or structures consisting of PARN and the mRNA substrate with its cap structure in the PARN cap-binding site and a poly(A) tail with its 3Ј end located adenosine residues in the active site of PARN.
The structure of the RRM domain of mouse PARN (mPARN) was recently determined by NMR analysis (Protein Data Bank entry 1WHV), and a schematic drawing of the structure is depicted in Fig. 1A. It is likely that the RRM of human PARN (hPARN) folds into a very similar structure because the corresponding region of hPARN only differs by one amino acid, i.e. glutamine residue 484 of mPARN is replaced by a lysine residue in hPARN (Fig. 1D). The amino acid numbering systems of mPARN and hPARN differ in this region by seven because of an insertion of seven amino acids in the human sequence N-terminally of the RRM domain. Thus, the two tryptophan residues Trp 456 and Trp 475 in hPARN correspond to residues Trp 449 and Trp 468 in mPARN, respectively. Interestingly, these two tryptophans are highly conserved throughout evolution (Fig. 1D). They are even conserved in the insect Anopheles gambiae, which has been shown to harbor cap-dependent PARN activity (23). However, they are not  19   conserved in Arabidopsis thaliana, which may indicate either that A. thaliana PARN (47) lacks cap binding capacity or that the polypeptide assigned as PARN in A. thaliana is only related to PARN in its nuclease domain. Experimental evidence for cap binding by the A. thaliana PARN remains to be obtained.
In the structure of the mPARN RRM, the residues Trp 449 and Trp 468 are located outside the ␤-sheet platform of the RRM (Fig. 1A), suggesting that the cap-binding site of PARN is not fully defined by the ␤-sheet platform. This is surprising because the corresponding ␤-sheet platform in the RRM of CBP20 harbors the cap-binding site of CBP20 (39 -41). In CBP20 the inverted guanosine residue of the cap is stacked between amino acids Tyr 20 and Tyr 43 (Fig. 1B). A corresponding aromatic slot consisting of Phe 475 and His 442 can be found in the mPARN RRM. However, in the case of the mPARN RRM, the side chain of tyrosine residue 505 is stacked between the two side chains of Phe 475 and His 442 , suggesting that this aromatic slot is unavailable for cap binding, given the assumption that the determined NMR structure of the mPARN RRM is correct in this particular aspect.
Sequence-specific recognition of poly(A) by RRM motifs has previously been well described both biochemically and structurally for PABP-poly(A) interaction (Fig. 1C) (Ref. 48; reviewed in Ref. 46). In this case the ␤-sheet platforms of two RRMs forms a continuous binding surface that can encompass the docking of 11 adenosine residues, required for efficient RRM⅐poly(A) complex formation. Interestingly, several lines of evidence suggest that the PARN RRM may also function as an oligomer when interacting with poly(A). Early biochemical studies suggested an oligomeric composition of PARN (22), and the recent crystal structure revealed a dimeric complex consisting of two PARN subunits (28). Unfortunately, the RRM domain was not included in the polypeptide used for crystallization. Nevertheless, the dimeric structure of PARN provides a structural reason for the possibility that the RRM may function as an oligomer. Our current study is also in keeping with the possibility that the PARN RRM may function as an oligomer when binding poly(A). Significantly, Figs. 4 -6 suggest that at least 10 adenosine residues are required for efficient PARN RRM⅐poly(A) complex formation.
One critical question is whether one subunit of the PARN RRM by itself can bind both poly(A) and cap simultaneously. This could very well be the case and would be in agreement with our biochemical and mutagenic data (Figs. 5 and 7 and Table 2). However, it cannot be excluded that the cap and the poly(A) tail bind to different subunits of a multisubunit RRM oligomer. Thus, even if the two binding sites are clearly separate from each other, both structurally and functionally, each of them may still only be functional on one subunit at the time. The location of the two tryptophan residues outside the ␤-sheet platform may provide enough binding surface to encompass binding of both the cap and the poly(A) tail simultaneously. Even if this very well could be the case, this scenario is highly speculative at the moment because it relies on the two critical assumptions that the cap interacts directly with at least one of the tryptophan residues (Trp 456 or Trp 475 ) and the poly(A) binds to the ␤-sheet platform. Although poly(A) binding to the ␤-sheet platform appears likely, because RNA binding to this platform is a general feature observed in determined RRM-RNA structures (46), our data do not unambiguously show that the cap interacts directly with any of the two tryptophans. It is,  for example, possible that the tryptophan residues play a structural role that indirectly affects cap binding. In this case the quenching effect we observe by fluorescence spectroscopy upon cap binding could be the result of essential conformational changes that effect the microenvironment of the tryptophan indol rings. The heterodimerization of the U2AF 35 ⅐ U2AF 65 complex constitutes one such example where the microenvironment of one tryptophan residue within the RRM of U2AF 35 is drastically changed because of docking of the two subunits during heterodimerization (49).
It has been shown that PABP also interacts with both poly(A) and cap and that simultaneous interactions are dependent on one of the RRMs in PABP (50). However, in the case of PABP it has not been shown unambiguously that the RRM by itself can interact with the cap, as is the case for the PARN RRM. Moreover, PABP-cap interaction requires that the cap is linked to a short RNA moiety (50). Nevertheless, the PABP and the PARN cases indicate that many different RRMs may encompass both cap and RNA binding properties. This could be of biological significance because the RRM domain is a very widespread structural element and one of the most commonly occurring RNA-binding domains identified so far (51).   a PARN or mutants thereof were expressed and purified as described under "Experimental Procedures." The indicated polypeptides at ͓P 0 ͔ being 1.5 nM were incubated with A 20 RNA substrate, and the amount of formed complex was monitored. b Filter binding assays were performed, and equilibrium dissociation constants, K D , and dissociation rate constants, k d , were calculated as described under "Experimental Procedures" in the absence or presence of 50 M cap as indicated. The listed values are the mean values Ϯ experimental error of at least three independent experiments, each based on at least seven independent data points.