Heterogeneous Ribonucleoprotein K (hnRNP K) Binds miR-122, a Mature Liver-Specific MicroRNA Required for Hepatitis C Virus Replication*

Heterogeneous ribonucleoprotein K (hnRNP K) binds to the 5′ untranslated region of the hepatitis C virus (HCV) and is required for HCV RNA replication. The hnRNP K binding site on HCV RNA overlaps with the sequence recognized by the liver-specific microRNA, miR-122. A proteome chip containing ∼17,000 unique human proteins probed with miR-122 identified hnRNP K as one of the strong binding proteins. In vitro kinetic study showed hnRNP K binds miR-122 with a nanomolar dissociation constant, in which the short pyrimidine-rich residues in the central and 3′ portion of the miR-122 were required for hnRNP K binding. In liver hepatocytes, miR-122 formed a coprecipitable complex with hnRNP K. High throughput Illumina DNA sequencing of the RNAs precipitated with hnRNP K was enriched for mature miR-122. SiRNA knockdown of hnRNP K in human hepatocytes reduced the levels of miR-122. These results show that hnRNP K is a cellular protein that binds and affects the accumulation of miR-122. Its ability to also bind HCV RNA near the miR-122 binding site suggests a role for miR-122 recognition of HCV RNA.

MicroRNAs (miRNAs) are a class of noncoding RNA of ϳ22-nucleotides in length that can regulate gene expression by either targeting RNA for degradation or suppressing their translation through base pairing to the RNAs (1). Since their discovery in 1993 in Caenorhabditis elegans, miRNAs have been found in many species and are involved in the regulation of proliferation, differentiation, apoptosis, and development (1,2). Moreover, miRNAs are also critical factors in the development of cancers, neurodegenerative diseases, and infectious diseases (3).
MiR-122 is a highly abundant RNA in hepatocytes that regulates lipid metabolism, regeneration, and neoplastic transformation (4 -6). In addition, miR-122 is required for the replication of the hepatitis C virus (HCV), a positive-strand RNA virus that infects over 170 million people worldwide (7)(8)(9). MiR-122 binds to a conserved sequence in the 5Ј untranslated region (UTR) of the HCV RNA to increase the stability of the HCV RNA (10). Silencing of miR-122 can abolish HCV RNA accumulation in non-human primates (11). The expression of human miR-122 in non-hepatic cells can confer the ability to replicate HCV RNA (12). MiR-122 is one of the most critical host factors for HCV replication.
We previously reported that the HCV RNA sequence that anneals to miR-122 is recognized by the heterogeneous ribonucleoprotein K (hnRNP K), a multifunctional RNA-binding protein known to be involved in RNA processing, translation, and the replication of several RNA viruses (13)(14)(15). In an unbiased screen for proteins from human proteome chips containing over 17,000 proteins, we identified 40 proteins that bind mature miR-122, including hnRNP K. Recombinant hnRNP K recognizes short pyrimidine sequences in miR-122 in vitro and a similar sequence in the HCV 5Ј UTR. In hepatocytes endogenous hnRNP K can form a coprecipitable complex with miR-122, whether or not the cells contain replicating HCV. HnRNP K is thus a protein that binds a mature microRNA.

EXPERIMENTAL PROCEDURES
Reagents-HnRNP K was expressed and purified from recombinant Sacchromyces cerevisae using a GST-tag as previously de-scribed (15) and stored in aliquots at Ϫ80°C until use. MiR-122 and variants, including those modified with fluorophores, were chemically synthesized (Bioneer, Alameda CA). The human proteome chip was from CDI Laboratories. RNAs were reverse transcribed into DNA using the TruSeq® Small RNA Sample Preparation Kit. Illumina DNA sequencing was performed with the MiSeq reagent kit v3. All of the siRNAs used were commercially available (Santa Cruz Biotechnology, Dallas TX).
Human Proteome Chip Assay-The human protein chip was screened as described previously (15). The protein chip was preblocked with a solution of 3% BSA and 0.1 mg/ml of salmon sperm DNA under a slide coverslip and then incubated with the probe RNA for 1 h at 37°C and 8 rpm using a BioMixer TM II (CapitalBio, Beijing, China). The coverslip was then removed and the chip was washed with 500 ml of phosphate buffered saline (PBS) amended with diethylpyrocarbonate and 0.05% Triton-X100 for 10 min. To quantify the relative amount of each protein spot on the human proteome chip, the chip was further probed with 80 l of 1 g/ml diluted DyLight TM 549-conjugated anti-GST monoclonal antibody (Rockland) in diethylpyrocarbonate-PBS and incubated for 45 min at 37°C, 8 rpm. After two washes, the chip was dried and scanned using LuxScan TM 10K Microarray Scanner (CapitalBio). The detected binding signals were analyzed using the GenePix Pro 6.0 software.
Affinity Measurements by Interferometry-HnRNP K binding to miR-122 or its variants were quantified using an Octet RED96 System (ForteBio, Menlo Park, CA). GST-tagged hnRNP K (Abnova, Taipei Taiwan) was immobilized on anti-GST biosensors (ForteBio). The capture level of anti-GST biosensors was 2.6 Ϯ 0.1 nM. Association and dissociation measurements were carried out in the presence or absence of miR-122 or its mutants serially diluted in PBS. A reference biosensor was included to determine the background signal. During the measurement, the analytes were subject to rotation at 1,000 rpm at 30°C. Kinetic constants (k on and k off ) were calculated from a 1:1 model fitting by using Octet software. Equilibrium dissociation constants (K D ) were calculated according to the Langmuir adsorption model for the steady-state response. Each reported value represents the results of a minimum of three independent experiments. SL1-hnRNPK Crosslinking Assay-HnRNPK binding to SL1 or variants was determined as previously described in Fan et al. (15). Briefly, 100 ng of purified protein was mixed with 1 pmole of the RNA that was radiolabeled at its 5Ј terminus using polynucleotide kinase and [␥ -32 P]ATP. The reactions were irradiated in a CL-1000 UV crosslinker (Stratagene, Santa Clara, CA) at 1,200 mJ/cm 2 (254 nm) for 3 min. The protein-RNA mixtures were subjected to SDS-PAGE, and the protein-RNA complex was quantified by autoradiography using a PhosphorImager (GE Healthcare Biosciences, Pittsburgh, PA). The proteins in the gel were visualized by Coomassie brilliant blue staining.
Crosslinking-Immunoprecipitation Assay-A crosslinking and immunoprecipitation (CLIP) assay was performed to examine the interaction between hnRNP K and miR-122 in Huh7.5 cells (16). Briefly, 5 ϫ 10 6 cells grown in 10 cm plates were washed twice with ice-cold PBS and irradiated on ice with 150 mJ/cm 2 at 254 nm (Stratalinker, Agilent, Santa Clara CA). The cells were harvested by digestion with trypsin, pelleted by centrifugation at 5,000 g for 5 min, and lysed with Nonidet P-40 lysis buffer (150 mM NaCl, 50 mM TrisCl, pH 8.0, 1.0% Nonidet P-40) on ice for 1 h with occasional mixing. The lysate was centrifuged at 16,000 g for 30 min, and the supernatant was collected and incubated at 4°C for 3 h with protein A/G beads conjugated with mouse anti-human hnRNP K antibody (Abcam, Thermo Fisher Scientific, catalogue number Ab39975) equilibrated with Nonidet P-40 lysis buffer or a goat anti-mouse IgG control antibody treated in the same way (Santa Cruz). Materials bound to Protein A/G beads were washed three times with 50 mM Tris saline solution amended with 0.1% Triton X-100 (TBST) followed by addition of 200 l of Laemmli sample buffer. The mixture was heated at 95°C for 2 min and loaded on precast 4 -12% Bis-Tris gel. Gel electrophoresis was for 1 h at 150 V in 1x MES SDS running buffer (ThermoFisher Scientific, Fredricks, MD). Protein-RNA complexes were transferred from the gel to a nitrocellulose membrane using a wet transfer apparatus (BioRad, Hercules, CA) at 100 V for 2.5 h. Proteins on the membrane were stained with 0.1% Ponceau-S for 10 min, and the portion of the membrane immediately above the 100 kDa hnRNP K band was excised using a new razor blade. RNAs from the membrane were released by digestion with proteinase K, extracted with Trizol reagent (Invitrogen), and precipitated with three volumes of ethanol, 0.5 M ammonium acetate, and centrifugation at 16,000 ϫ g for 30 min. The pellet was washed 2X with ice-cold 70% ethanol and dried. The eluted RNA was dissolved in water, polyadenylated with poly(A) polymerase (New England Biolab [NEB], Ipswich MA) and reverse transcribed using M-MulV reverse transcriptase (NEB) and anchored oligo-dT primer. Sequences for the primers used to amplify cDNAs will be made available upon request. High-throughput DNA sequencing was performed using the Illumina MiSeq system. CLIP-Seq was performed as described previously (17,18). Analyses of MiSeq reads were performed using the Galaxy User Interface. The reads were processed with the FASTX-toolkit by trimming the 3Ј bases with a quality score lower than 20 using the FASTQ Quality Trimmer and removing the 3Ј adaptor and discarding any reads shorter than 15-nt using Clip. Reads mapping to the specific RNA sequence were selected using Grep1.0.1, and the final read counts for each RNA were obtained after further analysis with the Compare two Datasets tool from the SharpLabTools Suite available through the Galaxy Tool Shed.
siRNA Knockdown of hnRNP K and Quantitative Reverse Transcription-PCR (qRT-PCR)-Huh7.5 cells (5 ϫ 10 5 cells) or Huh7.5 cells harboring the replicating 1b/Con1 replicon (5 ϫ 10 5 cells) were transfected with siRNA using Lipofectamine TM 2000 according to the manufacturer's instructions (Invitrogen). Aliquots of the transfected cells were assessed for effects on cell proliferation and cytotoxicity using the Wst-1 reagent (Clontech). The total RNA were extracted from the cells with Trizol Reagent and quantified by spectrometric. The RNA (1 g per sample) was reverse transcribed using M-MulV reverse transcriptase (NEB) and 4 M of a randomized 9-nt primer mix. The levels of mRNA for hnRNP K were determined by qRT-PCR using the iQ TM SYBR Green kit (BioRad). Amplifications were performed with an initial incubation at 10 min at 95°C, followed by up to 45 cycles of 15 s at 95°C, 20 s at 60°C, and 60 s at 72°C. The mRNA levels were normalized to GAPDH mRNA level, and the changes were compared with the mock-transfected control as previously reported (18). MiR-122 levels were quantified using the protocol of Livak and Schmittgen (19).

Identification of Human Mature miR-122-Binding Proteins-
The human proteome chip has allowed the global screening and rapid identification of ligands that can bind proteins (20). A version of the chip that contains ϳ17,000 GST-tagged human proteins has previously been used to screen for binding to specific RNAs (15). We used the chip to perform an unbiased screen of the human proteins that can bind to miR-122. A 22-nt miR-122 modified with a Cyanine 5 (Cy5) dye at its 5Ј terminus was used as a probe (Fig. 1A). As a specificity control, a Cy5-labeled 22-nt RNA that contained a scrambled miR-122 sequence named Sc-22 was used to independently probe a chip and identify nonspecific RNA-binding proteins. The amount of each protein within the chips was determined by probing the chips with DyLight TM 549labeled anti-GST antibodies to allow signal normalization (Fig.  1C). In all, 40 proteins were found to bind miR-122 with a mean of 1 S.D. above that of Sc-22. Bioinformatics analysis revealed that 13 of the 40 proteins contain nucleic-acid-binding motifs, and 10 are expressed in the liver ( Fig. 1C and Table  I). HnRNP K is one of the 10 proteins expressed in the liver that preferentially binds miR-122 (Table I).
We focused on hnRNP K and a possible interaction with miR-122 because hnRNP K was previously determined to bind to a stem-loop structure, SL1, within the HCV 5Ј UTR. Intriguingly, SL1 contains the Seed 1 sequence that anneals to miR-122 (15).
HnRNP K Binds to Both miR-122 In Vitro through Short Pyrimidine Sequences-A consensus motif for the RNAs bound by hnRNP K was previously identified in Xenopus laevis by B. Szaro and colleagues (21). The consensus motif contains three discontinuous stretches of enriched for pyrimidines separated by spacer sequences (Fig. 2A). HnRNP K contains three KH domains that individually can bind to single-stranded RNA and DNAs of ca. 4-nt that are enriched for pyrimidines (22). MiR-122 possesses a short pyrimidine-rich sequence at its 3Ј region, perhaps providing a sequence for hnRNP K binding through the KH domain ( Fig. 2A). Interestingly, we note that the Seed 1 sequence within SL1 also contained a short pyrimidine-rich sequence and that Seed 2 contains a direct repeat of the Seed 1 sequence (Fig. 2A).
An interferonmetry assay was used to determine the region within miR-122 that contributes to the binding of hnRNP K. GST-tagged hnRNP K was immobilized onto a biosensor that can bind GST. Wild-type miR-122 or the scrambled RNA sequence, Sc-22, was added in increasing concentrations to the immobilized hnRNP K. The wild-type miR-122 rapidly associated with hnRNP K while Sc-22 did not (Fig. 2B). GST attached to the chip did not bind miR-122, and Sc-22 had very poor binding ( Fig. 2C and data not shown). The binding isotherm for miR-122 was fitted to a 1:1 binding model, and the results suggest that binding had a dissociation constant of 14.9 nM. MiR-122 binding was characterized by rapid on and off rates (Fig. 2C). The binding of Sc-22 could not be modeled based on the concentrations tested.
Four variants of miR-122 that had substitutions along the length of miR-122 were tested for binding to hnRNP K. Mut1, which has three 5Ј terminal guanylates replaced with adenylates, was minimally affected for binding to hnRNP K, with a K D of ϳ26 nM (Fig. 2C). Mut2, which had two uridylates within the purine-rich sequence changed to adenylates, was severely reduced for hnRNP K binding. Mut3 and mut4, which had a concentration of pyrimidines in the central and the 3Ј region of miR-122 substituted with adenylates, were also reduced for hnRNP K binding (Fig. 2C). These results demonstrate that pyrimidines within the central and 3Ј portion of miR-122 both contribute to effective hnRNP K binding.
Changes in the sequence of miR-122 could have impacted hnRNP K binding by altering the structures of the RNAs. To determine whether a correlation to RNA structure exists, we first predicted the structure of miR-122 using the program mfold (23). MiR-122 formed a weakly stable hairpin structure (⌬G of -0.1 kcal/mol) with four base pairs in the stem (Fig. 3A). Mut1, which had only slightly weaker binding to hnRNP K, also formed a weakly stable hairpin. Interestingly, Mut2, which was predicted to form a more stable hairpin with six base pairs (⌬G -2.4 kcal/mol), was defective for hnRNP K binding. These results suggest that hnRNP K binding to miR-122 is likely through a more flexible pyrimidine-rich sequence in miR-122. Mut3 and Mut4, which had substitutions in the pyrimidines and also lacked a predicted stable structure, were reduced for hnRNP K binding. These results suggest that hnRNP K preferentially binds the short pyrimidine-enriched sequences in the flexible portions of miR-122.
HnRNP K Binds to the Seed 1 Sequence in the HCV SL1-HnRNP K binding to the pyrimidine-enriched sequences within miR-122 led us to further examine the motifs in the HCV SL1 required for hnRNP K binding. Similar to miR-122, SL1 has two pyrimidine-rich sequences. One of these is the Seed 1 sequence that binds miR-122 and is predicted by mfold to be unstructured in SL1 (10, Fig. 3B). The second is a stretch of homocytidylate that is predicted by mfold to base pair to a stretch of homoguanylates (15). To independently confirm that this is the case, we tested the hnRNP K binding of both SL1 and SL1 lacking the Seed 1 residues in an RNA named ⌬1. The RNAs were radiolabeled at the 5Ј terminus and comparable amounts of the two RNAs were incubated with hnRNP K and subjected to UV crosslinking. HnRNP K and the RNAs were resolved by denaturing PAGE (Fig. 3C). SL1 formed a

hnRNP K Binds miR-122
complex with hnRNP K, the majority of which corresponded to a monomeric mass of hnRNP K. A lower abundance complex migrated at the position of a dimeric hnRNP K (Fig. 3C). RNA ⌬1, which lacks the Seed 1 sequence, was greatly reduced for interaction with hnRNP K. These results confirm that hnRNP K recognizes pyrimidine-rich Seed1 sequence in RNA.
HnRNP K Can Bind miR-122 in Cells-We seek to determine whether endogenous hnRNP K binds miR-122 in cultured hepatocytes. A protocol that was previously used to examine the interactions between viral proteins with RNA in cells was adapted (17,18) (Fig. 4A). Briefly, Huh7.5 cells were first UV irradiated to generate covalent RNA-protein complexes. The cells were then lysed and immunoprecipitated with antibody specific to hnRNP K. Western blot analysis showed that the hnRNP K was quite specifically precipitated, as a control antibody did not result in detectable hnRNP K (Fig. 4B). The precipitated materials or regions in the membrane corresponding to the location of hnRNP K were digested with protease and the released RNAs were processed for quantitative RT-PCR. RNAs from immunoprecipitations performed with a control antibody did not detect miR-122 signal (data not shown). PCR performed with primers that do not recognize miR-122 and the RNAs did not result in detectable signal (data not shown). In addition, reactions performed without UV crosslinking resulted in only a low amount of detected miR-122. With UV crosslinking, a fourfold increase in miR-122 was observed (Fig. 4C). These results suggest that endogenous hnRNP K can associate with miR-122 in hepatocytes.
We also examined whether hnRNP K could interact with miR-122 in Huh7.5 cells harboring the replicating 1b/Con1 HCV replicon. When compared with the Huh7.5 cells, a higher amount of miR-122 was recovered in HCVϩ Huh 7.5 cells in the absence of UV crosslinking. There was also a modest of at least three independent experiments. Several RNAs did not reach saturation, and the KDs were estimated to be a minimal value based on the concentrations of the tested RNAs. The Student's t test was used to determine the statistical significance of the data sets. Two asterisks (**) denote a p value of Ͻ .01 when the binding parameters for the miR-122 variant were analyzed versus those of the wild-type miR12. Three asterisks (***) denote a p value of Ͻ .001. reduction in the total number of copies of miR-122 recovered by immunoprecipitation after UV crosslinking (Fig. 4C). However, similar to Huh7.5 cells, the level of miR-122 that selectively precipitated with hnRNP K increased in HCV ϩ cells subjected to UV crosslinking. In these HCVϩ cells, we were also able to detect an enrichment of the HCV RNA using RT-PCR (Fig. 4D).
HnRNP K Preferentially Binds Mature miR-122-cDNAs synthesized from the RNAs that coprecipitated with hnRNP K were subjected to high-throughput Illumina DNA sequencing to analyze the form of miR-122 that binds to hnRNP K. Over 20,000 sequences that were readily identified to contain the miR-122 sequence in both the RNAs from Huh7.5 and from HCV ϩ Huh7.5 cells (Fig. 4E). We also identified a low level of the sequence for RNase P, an abundant RNA in cells important for tRNA processing, and two additional miRNAs, miR-378 and miR-221. RNase P, miR378, and miR221 were not enriched with the immunoprecipitation for hnRNP K to the level observed for miR-122, suggesting that hnRNP K prefer-entially recognized miR-122 (Fig. 4E). However, similar levels of miR-122 were found to interact with hnRNP K in Huh7.5 cells harboring the HCV replicon (Fig. 4E), suggesting that HCV RNA replication is not required for hnRNP K interaction with miR-122.
Analysis of the reads that contained miR-122 sequence revealed that the vast majority (Ͼ 94%) contained mature miR-122 sequences of 22-nts in both cells that harbored the HCV replicon and those that did not. Less than 6% of the sequences had one to four nucleotides missing from the two termini of miR-122. Interestingly, the number of reads decreased when the uridylate-rich sequence at the 3Ј region of miR-122 was increasingly truncated (Fig. 4F). This is consistent with our observation that substitutions in the pyrimidine-enriched sequence negatively affected hnRNP K binding (Fig. 2C).
HnRNP K Level Affects miR-122 Accumulation In Vivo-To further define the effects of hnRNP K binding to miR-122, we used siRNAs to selectively knockdown hnRNP K in Huh7.5 cells. SiRNA specific to hnRNP K reduced the hnRNP K FIG. 3. Potential secondary structures of miR-122 and SL1 and the role of the Seed 1 sequence in hnRNP K binding. (A) Predicted secondary structures for miR-122 and two mutants that were affected for binding of hnRNP K. The RNA secondary structures were predicted using the computer program mfold (23) and the change in entropy (⌬G) is from the mfold calculations. Pyrimidine residues in the RNAs are highlighted in bold letters. The residues in Mut1 and Mut2 that were changed from the WT miR-122 are underlined. (B) The predicted secondary structure of SL1. The sequence for Seed 1 is boxed. RNA⌬1 lacks the Seed 1 sequence. (C) The Seed 1 sequence in SL1 is required for optimal binding to hnRNP K. The gel images were from an assay where purified hnRNP K was crosslinked to RNAs radiolabeled at the 5Ј terminus using ␥-32 P-ATP and polynucleotide kinase. The quantities of the probes used are shown. The complexes were then separated by SDS-PAGE, and the locations of the radiolabeled RNAs were identified using a phosphorimager. The bottom images show the recombinant hnRNP K present in the SDS-PAGE and stained with Coomassie brilliant blue. The relative amounts of the monomeric hnRNP K-RNA complexes from three independent experiments are shown below the gel images. mRNA relative to the GAPDH internal control to less than half of the mock-treated control (Fig. 5A). At this level of reduction, the cells exhibited only a model defect in the rate of proliferation, as determined by the Wst-1 assay (Fig. 5A). MiR-122 levels in these cells quantified using qRT-PCR were significantly reduced (Fig. 5B), suggesting that hnRNP K is required for the normal accumulation of miR-122 in Huh7.5 cells. Huh7.5 cells harboring genotype 1b/Con1 replicons had an overall lower level of miR-122 in the absence of anti-hnRNP K siRNA, but the knockdown of hnRNP K did not result in an additional decrease in miR-122 levels (Fig. 5B). It is possible that the presence of the HCV replicon stabilizes miR-122 even when hnRNP K levels are reduced. DISCUSSION This study was motivated by our previous observation that hnRNP K could bind the SL1 sequence in the 5ЈUTR of the HCV RNA that overlapped with the miR-122 binding site. A human proteome chip screened with miR-122 identified that hnRNP K as a miR-122 binding protein. Recombinant hnRNP K was able to bind to miR-122 in vitro through short pyrimidine sequences in miR-122. HnRNP K can also bind a short pyrimidine-rich Seed1 sequence within the HCV RNA SL1 that binds miR-122. HnRNP K binding to miR-122 was confirmed to take place in cultured hepatocytes using a crosslinkingimmunoprecipitation assay coupled to quantitative PCR. DNA sequencing revealed that hnRNP K preferentially binds mature miR-122.
HnRNP K contains three KH domains, each of which can accommodate RNAs and single-stranded DNAs of four nucleotides (22). Within the nucleic acid binding cleft of the KH domain is a hydrophobic pocket that interacts with the bases of single-stranded nucleic acids. The 3Ј terminal 4-nt of miR-122 (5ЈUUUG3Ј) matches sequences preferentially bound by KH domains. However, a typical KH domain binds RNAs with low micromolar dissociation constants (22) and the K D for miR-122 binding to hnRNP K was ϳ 15 nM. It is possible that multiple KH domains within hnRNP K contact pyrimidine-rich Lysed and subjected to immunoprecipitation with antibody specific to hnRNP K, an unrelated IgG, or mock immunoprecipitated. The precipitated material was then subjected to Western blot analysis detecting hnRNP K. (C) The number of RNA copies of miR-122 that precipitated with hnRNP K was measured by using qRT-PCR assay. Both the Huh7.5 cells lacking or harboring the genotype 1b/Con1 HCV replicons (HCV ϩ ) crosslinked using UV irradiation. A mock crosslinked reaction is used as a background control for the amount of miR-122. Each bar contains the results from three independent assays. Statistical analysis of the results were performed with the Student's t test, and two asterisks (**) denote a p value of Ͻ .01. (D) The number of copies of the HCV 1b/Con1 replicon that immunoprecipitated with hnRNP K. HCV RNA was detected by RT-PCR with primers detecting the 5Ј NTR of the HCV replicon (31). (E) A summary of the number of Illumina sequencing reads of four RNAs that were immunoprecipitated with hnRNP K from UV crosslinked Huh7.5 cells or HCV ϩ Huh7.5 cells. (F) Summary of the miR-122 RNAs sequences co-immunoprecipitated with hnRNP K. The sequences were identified by bioinformatics analysis of the Illumina DNA sequencing results. The topmost sequence is that of the wild-type mature miR-122. sequences within miR-122 (24). Within the HCV sequence that binds miR-122, we have demonstrated that the Seed 1 sequence contributes to hnRNP K recognition. Notably, the Seed 2 sequence also contains a sequence rich in pyrimidines (5ЈCACUCC3Ј) (10). It is likely that hnRNP K molecule(s) bind(s) both the Seed 1 and Seed 2 sequences.
Our data suggest that hnRNP K binding to miR-122 was not dependent on HCV RNA replication. However, miR-122 may be partially stabilized by HCV RNA infection when the level of hnRNP K was reduced. The increase in miR-122 stability could be due to sequestration of miR-122 by the HCV RNA (25). Since hnRNP K can bind both the HCV RNA and miR-122, it may have a role in bringing miRNA-122 to the HCV RNA. HnRNP K has been known to mediate the interaction of multiple molecules, including the HCV Core (26), and we have observed that the UV crosslinked hnRNP K-RNA complex appears to form a dimer when it binds RNA (Fig. 3C). We have attempted to determine whether recombinant hnRNP K could mediate the annealing of miR-122 and a target sequence in vitro, but the results were inconclusive. It is possible that factor(s) not present in our assay is/are needed to induce miR-122 to leave hnRNP K and bind to the target sequence. The phosphorylation of hnRNP K has been shown to decrease its binding to homopolymeric cytidylates (27).
MicroRNA biogenesis involves sequential interaction between the precursor and/or mature miRNA with regulatory proteins (28,29). Argonaute 2, which processes miRNAs, has also been reported to increase miRNA stability. Additional factors are necessary for additional processing and stability of the miRNA. GLD-2 is known to selectively increase the stability of mature miRNAs through 3Ј monoadenylation during biogenesis (30). Our study shows that hnRNP K increases the stability of mature miR-122, possibly by protecting a nonbase-paired portion of miRNA. The interaction between hn-RNP K and miR122 will likely modulate miR-122 stability and the replication of HCV RNA.
Acknowledgments-We thank Dr. B. Szaro for helpful discussion on hnRNP K. We thank L. Kao for editing of the manuscript. siRNAs specific to hnRNP K. hnRNP K mRNA was quantified using qRT-PCR. The signal within each sample was normalized to the concentration of GAPDH mRNA in the same cell extract. Each bar represents four independent transfections. Huh7.5 cells do not harbor the HCV replicons. The HCVϩ Huh7.5 cells, denoted by HCV ϩ , harbor the genotype 1b/Con1 HCV replicons. All samples were analyzed in at least triplicates and the mean and one standard deviation are shown. Two asterisks denote that pairs of samples that had p values less than .01 in the Student's t test. Below the bar graph are summaries of the results for effects of the siRNA on cell proliferation and cytotoxicity using the WST-1 assay. All samples were determined in triplicate and the mean and standard deviation of the degree of cell proliferation are shown. (B) Normalized levels of miR-122 in cells knocked down for hnRNP K. The amount of miR-122 was normalized to the amounts of GAPDH in the same cell extract. All samples were performed in at least triplicates and the mean and one standard deviation are shown. The asterisk denotes pairs of samples that pairs of samples had p values that of less than .05.