Identification of GW182 and its novel isoform TNGW1 as translational repressors in Ago2-mediated silencing

RNA interference (RNAi) is a potent post-transcriptional regulation mechanism for gene expression. It is triggered by small-molecule RNAs, including small interfering RNA (siRNA) and microRNA (miRNA), and then executed by the RNA-induced silencing complex (RISC) wherein targeted mRNA is degraded through the 5′→3′ RNA decay pathway. GW bodies (GWBs), also known as mammalian processing bodies (and also called dcp bodies or P-bodies in yeast) were found closely associated with RNAi and its related RNA-turnover activities (Jakymiw et al., 2007; Eulalio et al., 2007). Increase in siRNA-mediated activities induced GWB formation (Lian et al., 2007), whereas inhibition of the miRNA pathway led to disassembly of GWBs (Pauley et al., 2006; Eulalio et al., 2007). Blocking the 5′→3′ RNA decay before its initiation diminished GWBs (Cougot et al., 2004), whereas blocking after its initiation increased the size and number of GWBs (Sheth and Parker, 2003; Cougot et al., 2004; Andrei et al., 2005). A recent report demonstrated that the formation of GWBs was the consequence of RNAi activities in Drosophila melanogaster (Eulalio et al., 2007).

Trinucleotide repeat containing 6A (TNRC6A, hereafter referred to as GW182), one of the marker proteins of GWBs, was first identified in 2002 as a target protein of autoantibodies from a patient suffering from motor and sensory neuropathies (Eystathioy et al., 2002). It is an 182-kDa protein characterized by multiple glycine-tryptophan (GW) repeats. GW182 is important for GWB formation and RNA-induced gene silencing function even though it had no known enzymatic activity. Our earlier studies showed that knockdown of GW182 significantly disassembled GWBs (Yang et al., 2004; Jakymiw et al., 2005) and impaired the efficiency of siRNA functions (Jakymiw et al., 2005). Other investigators reported that GW182 was more important in miRNA function and closely associated with translational repression (Liu et al., 2005; Chu and Rana, 2006). The formation of GWBs from the sub-microscopic to microscopic level may occur during the process (Franks and Lykke-Andersen, 2007). The functional role of GW182 in this process has been shown to have at least three key facets. First, GW182 tightly interacted with Argonaute proteins in human (Jakymiw et al., 2005; Liu et al., 2005) and other species including Drosophila (Behm-Ansmant et al., 2006), Caenorhabditis elegans (Ding et al., 2005) and Arabidopsis (El-Shami et al., 2007). In addition, the interaction between GW182 and Ago2 (also known as EIF2C2 in eukaryotes) was RNA-independent (Liu et al., 2005) and proposed to depend on evolutionally conserved WG and/or GW motifs (El-Shami et al., 2007; Till et al., 2007). A recent report has shown that the GW182-Ago2 interaction is essential for miRNA-induced gene silencing and its subsequent mRNA decay in Drosophila (Eulalio et al., 2008b). Second, the translational repression effect is impaired when GW182 was knocked down (Liu et al., 2005; Chu and Rana, 2006). Behm-Ansmant et al. also showed in Drosophila that tethering GW182 to the 3′-UTR of the reporter luciferase mRNA, which bypassed the requirement of the antisense miRNA, led to translational repression (Behm-Ansmant et al., 2006). Third, GW182 is important for GWB formation as it may form an optimal microenvironment for recruiting the RNA decay factors to GWBs (Jakymiw et al., 2005). Tethering GW182 to reporter mRNA induced RNA degradation that required 5′→3′ RNA-degradation factors in Drosophila (Behm-Ansmant et al., 2006). In the absence of detectable GW182, even the formation of microscopic GWBs cannot be detected. Nevertheless, the inter-dependence among these proteins during translational repression remains unclear in the mammalian system. With the elucidation of additional GWB components, questions arose about how GW182, Ago2 and RNA-decay factors contribute to the RNA-induced gene silencing.

Interestingly, the GenBank database predicts another isoform of GW182 that we have provisionally named trinucleotide GW1 (TNGW1) because it contains trinucleotide repeats (TNRs) in its mRNA. Expansion of TNRs is known to be related to a set of diseases, most notably those with neurologic features, such as Huntington disease (Margolis et al., 1997). TNRC6A is one of the TNR-containing genes in the human genome, but to date it has not been related to trinucleotide expansion diseases. In clinical studies, we have shown that autoantibodies against GW182 and/or GWBs were associated with Sjögren's syndrome, mixed motor-sensory neuropathies, ataxia and systematic lupus erythematosus (Eystathioy et al., 2003b; Bhanji et al., 2007). However, to date there are no published reports describing the expression of TNGW1 and, therefore, we examine its expression and potential effect on translational repression.

Human TNRC6A (the gene encoding GW182) is located on chromosome 16p11.2. The first reported protein isoform of this gene, GW182, has distinct regions enriched in glycine (G) and tryptophan (W) repeats referred to as the GW-rich regions (Eystathioy et al., 2002). GW182 also has a glutamine-asparagine (QN)-rich region (Decker et al., 2007) in the middle and a classic RNA recognition motif (RRM) near the C-terminus. The predicted novel isoform TNGW1 (Fig. 1A), is a protein of ∼210 kDa containing a TNR Q-repeat domain (aa 93-127) in its N-terminus. The mRNA of TNGW1 contains five additional exons upstream of the putative AUG start codon of GW182 (Fig. 1B). The TNR Q-repeat domain is encoded by the fifth exon of TNGW1 mRNA and its corresponding nucleotide and amino acid sequences are shown in Fig. 1C. Interestingly, on the basis of genomic sequence analysis, using the University of California Santa Cruz Genome Browser software, the translation initiation sites of these two isoforms are predicted to be about 60 kb apart (Fig. 1B). Sequence alignment analysis showed that the N-terminus of TNGW1 was conserved among human, rat and mouse with some degree of diversity in the Q-repeat region (Fig. 1D). Similar domains were not identified in Drosophila or C. elegans, or the two other human homologues TNRC6B and TNRC6C.

On the basis of the predicted mRNA sequence of TNGW1 in the NCBI GenBank database, a reverse transcriptase (RT)-PCR assay was designed to verify the existence of TNGW1 mRNA by using primer sets that flanked the unique region of TNGW1 (nucleotides 248-1492). The anticipated 1.2 kb PCR bands were amplified from cDNA samples of HeLa, HEp-2, HepG2 cells and human testis (Fig. 2). The 1.2 kb PCR products obtained from HeLa, HEp-2 and human testis extracts were gel-purified, submitted for direct DNA sequencing and verified to be identical to the NCBI reference sequence (NM_014494.2). These results demonstrated that the TNGW1 mRNA, which contains an in-frame junction between the novel 5′ exons of TNGW1 and GW182, could be detected in at least two human cancer cell types and one normal adult tissue.

After having experimentally verified that mRNA of TNGW1 was expressed, we were interested to determine whether GW182 and TNGW1 proteins were both expressed. Antibodies specifically recognizing TNGW1 were developed to complement the previously generated anti-GW182 antibodies (Eystathioy et al., 2002; Eystathioy et al., 2003a). A recombinant polypeptide containing the TNR Q-repeat domain (rTNR, aa 1-204) was generated to immunize two rabbits (rabbit 6225 and rabbit 6226). The polyclonal antibodies isolated from both rabbits showed strong reactivity to rTNR in an addressable laser beads immunoassay (ALBIA; Fig. 3A) or in western blots (Fig. 3B) but did not crossreact with GW182 (Fig. 3B). Pre-immune antibodies from rabbits 6225 and 6226, as well as the two rabbit polyclonal antibodies against GW182, named 5182 and 6642, did not show reactivity to rTNR (Fig. 3A,B). Furthermore, three mouse monoclonal antibodies (mAbs) against rTNR, named 2E11, 2F11 and 5C8, were generated after initial screening and subsequent subcloning. As demonstrated by ALBIA, all three anti-rTNR mAbs showed a high number of median fluorescent units (MFUs), indicating strong reactivity to rTNR (Fig. 3C). By contrast, mAb against GW182 (GW182 4B6) and mAb against TNRC6B (GW2 25) had remarkably low numbers of MFU, which was comparable with culture supernatant controls (Fig. 3C). Since there are more than 20 TNR-containing genes in the human genome (Margolis et al., 1997), antibodies generated against rTNR might potentially crossreact with Q-repeat sequences of other TNR-containing genes. Therefore, additional studies were performed to determine the specificity of each anti-rTNR antibody, by using arrays of synthetic peptides that span the first 300 amino acids of TNGW1 (Fig. 3D). All three sera (rabbit 6225 and 6226, and human anti-GWB 18033) recognized multiple 15-mer peptides including those in the TNR Q-repeat domain. By contrast, all three mouse anti-rTNR mAbs recognized a relatively narrow set of the peptides (peptides 9 to 11, supplementary material Table S1) that reside outside of TNR Q-repeat domain. In summary, our data demonstrated that all the generated anti-rTNR Abs recognized a specific sequence within the N-terminus of TNGW1. The mouse mAb was highly specific to TNGW1, with a lower risk of crossreacting with other TNR-containing gene products.

Detection of endogenous GW182 and TNGW1 proteins by standard western blot was challenging because the protein levels were usually very low. Hence, separating and distinguishing these high-molecular mass proteins required careful optimization. To demonstrate the specific expression of both proteins in HeLa cells, combined immunoprecipitation and western blot (IP-WB) analysis was carried out by using human serum 18033, an anti-GWB serum known to contain antibodies against GW182, Ago2 and enhancer of mRNA decapping 4 (EDC4, hereafter referred to as Ge-1), to enrich these protein complexes from HeLa cell lysates prior to western blot detection. If TNGW1 were present together with GW182 in HeLa cells, as was predicted from the RT-PCR data (Fig. 2), anti-GW182 antibodies should recognize two bands because TNGW1 includes the entire sequence of GW182. As expected, the two rabbit polyclonal antibodies against GW182 (5182 and 6642) and the mouse mAb against GW182 (4B6) recognized both forms of GW182 proteins, and GW182, detected as the faster-migrating band, was the predominant isoform (Fig. 4A, left panel). In support of this conclusion, the anti-rTNR mAbs 2E11 (Fig. 4A) and 5C8 (Fig. 4B), as well as rabbit anti-rTNR sera 6225 and 6226 (Fig. 4B) recognized only the slower-migrating TNGW1. These data confirmed that both TNGW1 and GW182 proteins were expressed in HeLa cells.

There are three possibilities as to how TNGW1 and GW182 are expressed in cells. The first is that TNGW1 and GW182 are independently translated from two individual mRNA transcripts derived from chromosome 16 with different transcriptional start sites separated by ∼60 kb. The second possibility is that TNGW1 and GW182 are translated from the same mRNA with two different AUG start sites governed by respective Kozak consensus sequences (Kozak, 1991). The third possibility is that GW182 is a post-translationally processed product of TNGW1. To address these possibilities, we designed siRNA specifically targeting TNGW1 mRNA (siTNR) to examine the effect of suppressing TNGW1 mRNA on the expression of TNGW1 and GW182 protein. The knockdown effect of siTNR was initially validated by demonstrating its repression upon the expression of co-transfected EGFP-TNGW1 but not having an effect on the expression of co-transfected EGFP-GW182 in HeLa cells (data not shown). Forty-eight hours after siTNR transfection, western blot analysis of cell lysates showed that only the TNGW1 band disappeared, whereas the GW182 band remained the same (Fig. 4C, lane 1) when compared with untreated (Fig. 4C, lane 4) or mock-transfected controls (Fig. 4C, lane 3). These observations indicated that TNGW1 was derived from its own unique mRNA. However, if GW182 were to be post-translationally processed from TNGW1, as considered above, the GW182 band in siTNR-transfected cells (Fig. 4B, lane 1) would represent stable processed products under these experimental conditions. In contrast to this possibility, transfection of siRNA targeting the common regions of TNGW1 and GW182 (siGW182) resulted in the disappearance of both isoforms (Fig. 4B, lane 2). These data supported the conclusion that GW182 was not processed from TNGW1, and that TNGW1 and GW182 were transcribed independently. In summary, our data demonstrated that TNGW1 and GW182 are distinct at both transcription and translation levels.

GW182 is one of the accepted marker proteins of GWBs (Eystathioy et al., 2002) and has been shown to be important for GWB formation (Jakymiw et al., 2005; Lian et al., 2007). Since TNGW1 shares the same amino acid sequence as GW182 – except for the N-terminal domain including the TNR Q-repeat region – we were interested to examine the intracellular location of TNGW1 and its role in GWB formation. The immunofluorescence staining of mouse monoclonal anti-rTNR 2F11 on HEp-2 cells showed GWB staining that was also recognized by the human anti-GWB serum 18033 (Fig. 5A, arrows). Notably, both mouse anti-rTNR 2F11 (Fig. 5A) and rabbit anti-rTNR 6226 (supplementary material Fig. S1A) stained a subset of ∼30% GWBs. By contrast, anti-GW182 mAb 4B6 stained more, although not all, GWBs recognized by serum 18033 (supplementary material Fig. S1B). These data imply that the amount of TNGW1 or GW182 varies in different GWBs, even when they are both present in the same GWBs. We therefore examined this hypothesis by performing dual staining using anti-rTNR 2F11 and anti-GW182 4B6 in the same HEp-2 cells (Fig. 5B). Visualized by IgG-subclass-specific secondary antibodies, anti-GW182 4B6 stained apparently more GWBs than anti-rTNR 2F11, whereas all 2F11 staining colocalized with that of 4B6. The results support the hypothesis that TNGW1 is absent or, at least, in very low abundance in a subset of GWBs, where only GW182 is present as the predominant isoform. However, given the limitation of antibodies, we could not rule out the possibility that the immunoreactive region of these isoform were obscured by the presence of one or more additional GWB component. Nevertheless, our data indicated that TNGW1 resided in a subset of GWBs and this clearly demonstrated heterogeneity in GWBs contributed by GW182 gene products.

GW182 was once proposed to be a matrix protein in GWBs because it was required for the assembly of these foci (Yang et al., 2004), which also harbored multiple proteins including Ago2, Dcp1a, Ge-1 and LSM14A (also known as and hereafter referred to as RAP55) (reviewed by Jakymiw et al., 2007). Therefore, we decided to determine whether the extra N-terminal domain in TNGW1 affects the localization of other GWB components to GWBs. Immunofluorescence assay was performed on HeLa cells into which EGFP-Ago2 had been co-transfected with either GST-GW182 or GST-TNGW1. Fig. 6A shows that both GST-GW182 and GST-TNGW1 are enriched in cytoplasmic foci – together with EGFP-Ago2, indicating that the extra N-terminal region of TNGW1 does not interfere with the localization of TNGW1 or Ago2 to GWBs. Furthermore, Fig. 6B shows that, in cells transfected with EGFP-TNGW1 alone the EGFP-labeled GWBs were also co-stained using anti-Dcp1a antibody and human serum IC6 containing antibody against Ge-1 and RAP55 (Bloch et al., 2006). Notably, transfected cells with either low or high expression of EGFP-TNGW1 did apparently not affect the localization of endogenous Dcp1a and Ge-1 and/or RAP55 to the TNGW1-containing foci, suggesting that TNGW1 can efficiently substitute for putative GW182 functions, such as recruitment of Ago2 and the formation of foci enriched in RNA-decay factors.

As shown in previous studies, GW182 is essential for the formation of microscopically visible GWBs (Young et al., 2004). However, since the existence of TNGW1 was not appreciated at that time, previous conclusions were based on the knockdown of both TNGW1 and GW182. Since we showed that siTNR achieved almost complete knockdown of TNGW1 without affecting the level of GW182 (Fig. 4B), we could determine whether TNGW1 knockdown affects the formation of GWBs. siTNR was transfected into HeLa cells and the changes of GWBs were monitored at days 0, 1, 2 and 3. siGW182 was transfected side by side into HeLa cells as a control, and the formation of GWBs was monitored by co-staining with rabbit anti-Dcp1a and human anti-GWB serum 18033. Consistent with the western blot data (Fig. 4C), 2 days after transfection the siGW182 transfection led to the disassembly of most GWBs, presumably due to knockdown of both TNGW1 and GW182 (Fig. 7). By contrast, 2 and 3 days after siTNR transfection, microscopic GWBs were still detected by both anti-Dcp1a and human serum 18033. This observation demonstrated that TNGW1 is not important for GWB formation under these experimental conditions.

To explore the effect of translational repression through TNGW1 compared with that of GW182 or Ago2, we adopted the tethering assay from the work of Pillai et al. (Pillai et al., 2004). An N-terminal λN-hemagglutinin (NHA) polypeptide tag was fused to TNGW1, GW182 or Ago2. The NHAtag binds the 5BoxB secondary structures harbored in the 3′-UTR of the firefly luciferase (FL)-5BoxB mRNA, resulting in a tethering effect of the tagged protein to the 3′-UTR. The repression effect in HEK 293 cells was evaluated by comparing the FL-5BoxB activities among different experimental groups relative to the untargeted Renilla luciferase (RL) reporter activities (supplementary material Fig. S2A) using the method described in a previous study (Lytle et al., 2007). After 48 hours of transfection, FL-5BoxB activity was repressed by 46% when Ago2 was tethered to the reporter. Interestingly, tethered TNGW1 or GW182 induced strong repression on reporter (67.6% and 65.3%, respectively) which was 46.9% or 41.3% stronger than that induced by Ago2, respectively (Fig. 8A). Comparison of the corresponding FL-5BoxB and RL mRNA levels by quantitative RT-PCR showed that both tethered TNGW1 and GW182 were accompanied by a reduction in FL-5BoxB mRNA (by 23.7% and 24.5%, respectively), whereas tethered Ago2 showed associated 50.8% reduction in the levels of FL-5BoxB mRNA (Fig. 8B). Therefore, the analysis of translation efficiencies of FL-5BoxB mRNA in each experimental group calculated using the formula described in previous study (Lytle et al., 2007) showed that the tethered TNGW1 and GW182 reduced the translational efficiency of FL-5BoxB mRNA (to 42.5% and 46.0%, respectively) to a significantly greater extent than tethered Ago2 (no reduction, 109.7%), which repressed the FL-5BoxB activity with a lower abundance of reporter mRNA at 48 hours (Fig. 8C). Previously, Pillai et al. have shown that tethering Ago2 to a RL reporter with 5BoxB in 3′-UTR mainly induced translational repression in HeLa cells (Pillai et al., 2004). Our observations, however, indicated that tethered Ago2-repressed FL-5BoxB activity was accompanied by a reduced level of its reporter mRNA in HEK 293 cells. The discrepancy observed in RNA levels might be caused by the use of different cell lines (HEK 293 vs HeLa cells) or a difference in mRNA turnover between the FL and RL reporters.

With the interesting finding that TNGW1 and GW182 exerted a stronger translational repression effect than Ago2, the next issue was to address the inter-dependence of these proteins in translational repression. The same tethering assay was performed using HeLa cells, by transfecting siRNA in order to knock down either Ago2 or GW182 before tethering protein and reporter constructs were co-transfected. Surprisingly, the repression effect exerted by NHA-Ago2 was totally abolished in GW182-knockdown cells (Fig. 9A). By contrast, the repression effect of tethered GW182 or TNGW1 was not affected by knockdown of Ago2 (Fig. 9B). The observations that tethered Ago2 is required GW182/TNGW1 for translational suppression whereas tethered GW182 does not require Ago2 to suppress translation was reproducible when either FL-5BoxB or RL-5BoxB were used as reporters (supplementary material Fig. S7). Our data implies that both TNGW1 and GW182 have a direct effect on repression, more than that of Ago2, and that they are both required for Ago2-mediated translational silencing.

In this study, we have identified TNGW1 as a novel 210-kDa isoform of GW182 with both proteins highly enriched in GWBs. As predicted in the NCBI GenBank database, the amino acid sequences of TNGW1 and GW182 are identical, with the exception that TNGW1 has an extra 253-aa polypeptide in its N-terminus. TNGW1 was expressed in several human cancer cell lines and human testis together with the reported short isoform GW182. Our data demonstrated that TNGW1 and GW182 are expressed independently, and GW182 was the predominant gene product. The expression of their mRNAs is possibly due to alternative splicing, alternative promoters or transcriptional start sites. The expression levels of each isoform under different cellular conditions with potentially different RNAi activities and the factors determining their expression need further investigation.

Both TNGW1 and GW182 were highly enriched in GWBs, but TNGW1 was detected only in about one-third of endogenous GWBs. Knockdown of TNGW1 did not noticeably disrupt GWB formation indicating that TNGW1 was not required for the formation of many GWBs. Both TNGW1 and GW182 exerted a similar translational repression effect in the tethering assay, and these data implied that TNGW1 might be functionally redundant and that GW182 has a more-important role in the formation of GWBs. However, this interpretation needs to be solidified by determining the effect of TNGW1 on GWB formation in the absence of GW182. Since an efficient method to knock down GW182 without affecting TNGW1 is not currently available, it remains unclear whether TNGW1 alone can substitute for all putative function for GW182. Given that TNGW1 contains the whole amino acid sequence of GW182, it is likely that TNGW1 is capable of most functional characters of GW182. With the extended N-terminal region containing the TNR Q-repeat domain, the full functional characteristics for TNGW1, in addition to GW182, remain to be determined.

In mammalian systems, Ago2 (EIF2C2) is considered to be the most important factor in the RNA-induced silencing complex (RISC) because it binds siRNA and miRNA as well as being the only factor harboring the slicing activity responsible for siRNA-induced silencing (Liu et al., 2004; Yuan et al., 2005). However, the mechanism for miRNA-mediated repression remains unclear. In the presence of incomplete complementarity between anti-sense strand miRNA and its target mRNA, the `slicing' function of Ago2 is interfered, where Ago2 may need to recruit multiple factors to secure its repression effect on the targeted mRNA and possibly induce the subsequent mRNA degradation. GW182 has been reported as more important in gene silencing when slicing activity is limited, for example, in miRNA-mediated silencing (Liu et al., 2005; Chu and Rana, 2006). In the present study, we extended the understanding of their interdependence by tethering TNGW1, GW182 or Ago2 to the 3′-UTR of a luciferase reporter mRNA and comparing their relative potentials in translational repression. Our data demonstrated that both TNGW1 and GW182 exerted stronger translational repression than Ago2. Furthermore, the repression effect of tethered Ago2 was sensitive to the presence of TNGW1 and/or GW182, whereas repression by tethered TNGW1 or GW182 did not require Ago2. These observations suggested that either TNGW1 or GW182 has a more direct impact on translational repression than Ago2. Although it is possible that other Argonaute proteins could substitute for Ago2 when it was knocked down, the functional importance of TNGW1 and/or GW182 was obvious even though human cells may also have two homologues of GW182, TNRC6B and TNRC6C. In miRNA-mediated translational silencing, the miRNA-loaded Ago2 can direct GW182 to their targeting mRNA and repress its translation. Hence, when TNGW1 or GW182 was tethered to the reporter mRNA via the λN tag, it may bypass the requirement of miRNA-Ago2 guidance. Consistent with this observation, Eulalio et al. have shown that the interaction between GW182 and Argonaute proteins is essential for miRNA-mediated translational repression and mRNA decay in Drosophila (Eulalio et al., 2008b). However, whether the repression effect is caused directly by TNGW1 and/or GW182, or the additional factors recruited in later stages of the process requires further investigation.

In summation, our data suggest the following scenario with a putative order of events. With the guidance of miRNA, Ago2 is able to target the 3′-UTR of a specific mRNA. GW182 and/or TNGW1 are subsequently enriched at the 3′-UTR owing to its interaction with Ago2. Once GW182 and/or TNGW1 are brought to the 3′-UTR, the translational suppression is triggered either directly by GW182 and/or TNGW1 or by other factors further recruited to the complex. In the tethering assay, tethered GW182 and/or TNGW1 did not require miRNA and Ago2 for translational suppression because their functions is substituted by the interaction of the λN tag and BoxB sequence. As more studies of translational repression are reported (reviewed by Filipowicz et al., 2008; Eulalio et al., 2008a), the importance of GW182 in each step may be further characterized in future studies.

It is possible that TNGW1 and GW182 are redundant protein products of the human TNRC6A gene in some aspects of translational repression process. Our data showed that they both formed cytoplasmic foci that colocalized with other RNAi-related factors and mRNA decay factors. As demonstrated by the functional assays tested in the current study, both isoforms induced translational repression and mRNA degradation to a similar extent. However, the hypothetical redundancy between TNGW1 and GW182 may be because of our limited understanding of their functions. A similar case could also be made for the human Argonaut protein family, the eukaryotic translation initiation factors 2c, which comprises at least four Argonaute-like proteins (EIF2C1, EIF2C2, EIF2C3 and EIF2C4, also known as AGO1, AGO2, AGO3 and AGO4, respectively) that share over 90% sequence similarity. Except for EIF2C2, known as the catalytic engine of RISC (Liu et al., 2004), the biological functions and significance of EIF2C1, EIF2C3 and EIF2C4 are not well understood.

One distinguishing feature of TNGW1 is that it is only localized to a subset of GWBs in HEp-2 cells. This specific localization of TNGW1 is probably related to its unique N-terminal polypeptide domain that is not found in GW182. This N-terminal domain might be responsible for interacting with or recruiting protein factors to help with translational suppression. It is also possible that this unique N-terminal domain affects the protein folding of TNGW1 and somehow interferes with its interaction with Argonaute proteins. The fact that TNGW1 functional capacity is similar to that of GW182 makes it important to further investigate the potential differences in their functions.

Our data confirmed the heterogeneity in GWBs in terms of TNGW1 and GW182 distribution, which is consistent with the observations from recent reports (Jakymiw et al., 2007; Moser et al., 2007). The importance of this heterogeneity could be a reflection of different stages of GWB assembly. However, it could also reflect the diversity in the functional status of GWBs, which are closely related to miRNA-mediated function. When targeted by miRNA, most mRNAs may enter the accelerated turnover process whereas some may not. A recent report showed that two luciferase reporter mRNAs that carry different 3′-UTRs were degraded at different rates when both were translationally repressed (Eulalio et al., 2008b). Under stress conditions, the translational efficiency of some miRNA-targeted mRNAs have been reported to be upregulated (Bhattacharyya et al., 2006; Vasudevan et al., 2007), which required these targeted mRNAs to remain stable during miRNA-targeting. Interestingly, the determining factors in the turnover of targeted mRNAs seemed not only to depend on the miRNA-targeting sequence but also on the proteins that were recruited during the silencing process (Bhattacharyya et al., 2006). Since the formation of GWBs has been shown to be a consequence of miRNA activity (Pauley et al., 2006; Eulalio et al., 2007), it is not surprising that there are multiple functional roles for these cytoplasmic foci. The identification of the TNGW1 as a novel isoform of GW182, and the heterogeneous distribution of TNGW1, GW182 and other RNAi factors in GWBs (Jakymiw et al., 2007), will provide us with a better understanding of the RNAi process at molecular cell biology level.

To identify TNGW1 mRNA, PCR was performed on cDNA from HeLa, HEp-2 and HepG2 cell lines (ATCC, Manassas, VA), and normal adult human testis (BioChain, Hayward, CA) using primer TNRC-1 (5′-ATAATGCCAAGCGAGCTACAG-3′; nucleotides 248-268) and TNRC-2 (5′-AAGGGAAGTGCCATTCATACC-3′; nucleotides 1512-1492). PCR reactions were carried out using SureStart™ Taq DNA polymerase (Stratagene, Cedar Creek, TX) following manufacturer's protocol. Annealing temperature for PCR amplification was 54°C. Complete nucleotide sequences of the PCR products were determined in both strands by using BigDye terminator sequencing at the University of Florida Interdisciplinary Center for Biotechnology Research Sequencing Core Laboratory.

To construct full-length TNGW1, PCR amplification was conducted on the human testis cDNA (BioChain) using primer TNRC-5a [5′-TTTGGAAGATCTATGAGAGAATTGGAAGCTAAAGCT-3′ containing a synthetic BglII site sequence (underlined) immediately upstream of the ATG translational start site of TNGW1], and primer TNRC-2, which is downstream of an internal KpnI site (nucleotide 1252). The 1.5 kb PCR product was purified and digested with BglII and KpnI to generate a 1.2 kb fragment that was used to replace the 5′ 500-bp BamHI-KpnI fragment in full-length GW182 cloned in the pENTR vector. The BamHI restriction site was from the 5′ linker sequence of the pENTR vector. Both the 1.2 kb BglII-KpnI fragment and the BamHI-KpnI linearized plasmid of pENTR-GW182 were gel-purified and then ligated at 16°C overnight to generate pENTR-TNGW1 with an 8.4 kb insert. Expression vectors, enhanced green fluorescence protein (EGFP)-tagged and glutathione-S-transferase (GST)-tagged TNGW1 were generated using pENTR-TNGW1 and respective pDEST vectors via recombination using LR Clonase® II (Invitrogen, Carlsbad, CA) following the manufacturer's protocol.

To generate a construct expressing a recombinant polypeptide containing the TNR (rTNR; aa 1-204), pENTR-TNGW1 was first digested with BamHI (nucleotide 610) and NotI (3′ end linker) to release a 6.5 kb fragment containing GW182. The overhangs of the vector that encode the N-terminus of TNGW1 was filled in and then ligated at room temperature for 1 hour to generate the deletion construct pENTR-rTNR. Expression vectors for rTNR in pDEST17 (Invitrogen) and pDEST-EGFP were generated to produce 6×His-rTNR in Escherichia coli, and EGFP-rTNR for expression in mammalian cells, respectively.

Tethering assay plasmids, including pClneo-NHA vector, NHA-Ago2, firefly luciferase (FL) reporter containing five BoxB structures (FL-5BoxB), Renilla luciferase containing five BoxB structures (RL-5BoxB), FL and RL reporters were gifts from Witold Filipowicz (Friedrich Miescher Institute for Biomedical Research, Basel, Switzerland) (Pillai et al., 2004). To generate NHA-GW182 and N-HA-TNGW1, the pCIneo-NHA vector was converted to a gateway-destination vector by using the Gateway Vector Conversion System (Invitrogen). Then, TNGW1 and GW182 were moved from corresponding pENTR vectors to the pCIneo-NHA gateway vector, respectively, by recombination. All DNA constructs were confirmed by direct DNA sequencing.

The expression of recombinant 6×His-rTNR protein in BL21 (DE3) E. coli and purification by Ni²⁺-affinity chromatography was performed using Qiagen's protocol as previously described (Eystathioy et al., 2002). Two New Zealand White rabbits (rabbits 6225 and 6226) were used to generate polyclonal antibodies following the standard protocol by Lampire Biological Laboratories, Pipersville, PA. Pre-immune blood samples, as well as samples collected after initial and booster injections, were harvested and analyzed for reactivity. For the production of monoclonal antibodies (mAbs), hyperimmunized BALB/c mice were used to generate hybridomas (carried out by the University of Florida Interdisciplinary Center for Biotechnology Research Hybridoma Core Laboratory). Three mouse mAbs (2E11, 5C8 and 2F11) were on the basis of enzyme-linked immunosorbent assay and indirect immunofluorescence screening. All three mouse mAbs were identified as IgG2a,κ antibodies.

HeLa and HEK 293 cells (from ATCC) were cultured in DMEM containing 10% fetal bovine serum at 37°C under 5% CO₂. Lipofectamine 2000 (Invitrogen) was used for transient siRNA and DNA-plasmid transfection following manufacturer's protocol. Briefly, cultured cells were grown to 40-50% confluence and transfected with 100 nM siRNA. Cells were fixed or lysed 2 days after the transfection. In 3-day-long experiments, cells were fixed at days 1, 2 and 3 after transfection. The siRNA targeting the TNR region of TNGW1 (siTNR) was designed by using the Dharmacon online tool. The corresponding sequences are sense (5′-UCGGUAUCCUCGUGAAGUATT-3′) and antisense (5′-UACUUCACGAGGAUACCGATT-3′). The sequences of siRNA targeting EGFP (siGFP), Ago2 (siAgo2) and GW182 (siGW182) have been reported previously (Lian et al., 2007). In DNA transfection experiments, cells were maintained at 70∼90% confluence for transfection, harvested 24 or 48 hours after transfection and analyzed by immunofluorescence and/or western blotting.

Combined immunoprecipitation and western blot (IP-WB) analysis was performed as described in detail previously (Moser et al., 2007). In brief, for the IP step human anti-GWB antibodies from the prototype serum 18033 (Mitogen Advanced Diagnostics Laboratory, University of Calgary, Calgary, AB, Canada) were chemically crosslinked to proteinA-Sepharose beads to prevent elution in the subsequent SDS-PAGE step. IP samples were resolved by a 6.5% SDS-PAGE with the low-molecular-mass proteins (<75 kDa) run off the gel in order to achieve optimal separation of the two GW182 isoforms. The latter were then electrophoretically transferred to nitrocellulose membranes (Bio-Rad Laboratories, Hercules, CA). Primary antibodies used in the western blot step included mouse antibodies anti-rTNR 2E11 (undiluted), 5C8 (undiluted), anti-GW182 4B6 (1:10) (Eystathioy et al., 2003a), anti-hemagglutinin (HA) (1:1000, Covance, Emeryville, CA), and rabbit antibodies, anti-rTNR 6225 (1:200), 6226 (1:200), anti-GW182 5182 (1:200), 6642 (1:200) and anti-EGFP (1:1000, Invitrogen). Secondary antibodies included either horseradish peroxidase (HRP)-goat anti-human Ig (1:20,000; Sigma, St Louis, MO), goat anti-rabbit IgG (1:20,000; Jackson ImmunoResearch, West Grove, PA), or goat anti-mouse IgG (1:2,000; Santa Cruz Biotechnology, Santa Cruz, CA). Bands were detected by using the Enhanced Chemiluminescence kit (Amersham Biosciences, Piscataway, NJ) or the Supersignal Chemiluminescent system (Pierce Chemical, Rockford, IL). When necessary, nitrocellulose membranes were stripped using stripping buffer (100 mM, 2-mercaptoethanol, 2% SDS, 62.5 mM Tris pH 6.7) for 30 minutes at 65°C for further probing.

A set of addressable beads bearing laser reactive dyes (Luminex, Austin, TX) were coupled to purified rTNR polypeptide and analyzed for antibody reactivity as previously described (Eystathioy et al., 2003a). Mouse mAb and rabbit sera were diluted in QUANTA Plex diluent (INOVA, San Diego, CA) to a final concentration of 1:100. Thirty microliters of QUANTA Plex diluent was added to each well followed by 10 μl of the diluted sample and then incubated on an orbital shaker for 30 minutes at room temperature. This was followed by the addition of 40 μl of phycoerythrin-conjugated species specific anti-IgG (Jackson ImmunoResearch, West Grove, PA; diluted 1:50) to each well, incubated on the orbital shaker for an additional 30 minutes. The reactivity of the antigen-coated beads was determined on a Luminex 100 dual-laser flow cytometer (Luminex, Austin, TX). Each assay included negative and positive controls, and results were expressed as median fluorescent units (MFUs).

To characterize the specific reactivity of the generated anti-rTNR antibodies, membranes containing in situ synthesized sequential 15-mer peptides that were offset by five amino acids and represent the region-terminal domain of the TNGW1 protein (supplementary material Table S1) were prepared (Eve Technologies, Calgary, AB, Canada) as previously described (Selak et al., 2003; Eystathioy et al., 2003a). The dehydrated membranes were prepared for immunoblotting by an initial 10-minute incubation in 100% ethanol followed by rehydration in Tris-buffered saline (TBS; 10 mM Tris-HCl pH 7.6, 150 mM NaCl) for 10 minutes at room temperature. The membranes were blocked in 2% milk-TBS overnight at 4°C and incubated with various primary antibodies at the appropriate dilution for 1.5 hours at room temperature on a shaker. Following three washes of 5 minutes each with 2% milk-TBS, appropriate HRP-conjugated secondary antibodies diluted in 2% milk-TBS as described above were incubated with the membranes for 45 minutes at room temperature on a shaker. Membranes were washed three times with TBS for 2 minutes each and the bound antibodies were detected using the enhanced chemiluminescence kit (Amersham Biosciences). The same stripping method was used for the peptide membrane as described in the western blot section.

HEp-2 slides (ImmunoConcepts, Sacramento, CA) or HeLa cells grown as a monolayer were used to perform indirect immunofluorescence assay as described (Jakymiw et al., 2005). Primary antibodies used included: mouse anti-rTNR 2F11 (undiluted culture supernatant), rabbit anti-rTNR 6226 (1:200), rabbit anti-GST (1:1000) (provided by Peter Sayeski, University of Florida), rabbit anti-Dcp1a (1:500) (provided by Jens Lykke-Andersen, University of Colorado), human anti-GWB sera 18033 (1:6000) and IC6 (1:1000, Mitogen Advanced Diagnostics Laboratory, University of Calgary). Secondary antibodies include Alexa-Fluor-488 (1:400, Invitrogen), Alexa-Fluor-568 (1:400, Invitrogen), and Cy5 (1:100, Jackson ImmunoResearch Laboratories, West Grove, PA) conjugated goat antibodies to IgG of corresponding species (human, rabbit or mouse). Goat anti-mouse IgG2a TRITC (1:50, Southern Biotech, Birmingham, AL) and goat anti-mouse IgG1 488 (1:400, Invitrogen) were used specifically in 2F11 and 4B6 dual staining assay. The slides were mounted using VECTASHIELD mounting medium with 4′,6-diamidino-2-phenylindole (DAPI; Vector Laboratories, Southfield, MI). Fluorescent images were captured with a Zeiss Axiovert 200M microscope (Carl Zeiss, Jena, Germany and processed using Adobe Photoshop (Adobe Systems, San Jose, CA). Magenta was used as a pseudo-color in images when needed.

HEK 293 cells seeded at 75-80% confluence were transfected with 600 ng DNA plasmid of NHA-tag-expressing vector, NHA-Ago2, NHA-GW182 or NHA-TNGW1 plus targeted luciferase (either 150 ng FL-5BoxB or 10 ng RL-5BoxB) and control luciferase (50 ng RL or 100 ng FL) plasmid using Lipofectamine 2000. Cells were harvested 48 hours after transfection and the FL and RL activities were measured using Dual-Luciferase® reporter assay system (Promega, Madison, WI) following the manufacturer's protocol. Relative luciferase activities (ratio of targeted luciferase activities over control luciferase activities) were first calculated as described in the laboratory of Witold Filipowicz (Pillai et al., 2004) and the translational repression was calculated based on a recent study (Lytle et al., 2007). Briefly, FL-5BoxB/RL activity in NHA vector transfected (control) group was regarded as 0% translational repression. The repression levels of other experimental groups were calculated by the percentage reduction of relative luciferase activities compared to that in NHA control group. All data were collected from three to six independent experiments for statistical analysis. The expressions of all NHA constructs were monitored by western blot as shown in supplementary material Fig. S2B.

Total RNA samples from tethering assays were extracted from HEK 293 cell lysates by using RNeasy Mini Kit (Qiagen, Valencia, CA). RNase-free DNase Set (Qiagen) was applied to eliminate the potential DNA contamination. Samples were analyzed in duplicate by quantitative RT-PCR using SYBR-Green Master mix (Applied Biosystems, Foster City, CA). The relative mRNA levels of FL-5BoxB/RL were calculated by ΔΔCt method. The melting curve in each individual measurement was monitored to guard against non-specific amplification. The FL-5BoxB/RL mRNA levels of NHA-Ago2, NHA-GW182 and NHA-TNGW1 were compared to the mRNA level in NHA vector transfected control group, which was defined as 0% mRNA degradation, and calculated the corresponding mRNA degradation. Sequences of primers for RL were: forward 5′-TCCTACGAGCACCAAGACAAGA-3′, reverse 5′-GATCACGTCCACGACACTCTCA-3′. Sequences of primers for FL were: forward 5′-GCGACCAACGCCTTGATT-3′, reverse 5′-TCCCAGTAAGCTATGTCTCCAGAA-3′. In siRNA tethering assay, the mRNA levels of Ago2, and TNGW1 and/or GW182 were measured using TaqMan® Fast Universal Master Mix (Applied Biosystems) with the corresponding TaqMan® gene expression assay (Ago2, Hs00293044_m1; TNRC6A, Hs00379422_m1 and 18S rRNA, 4310893E, Applied Biosystems).

We thank Witold Filipowicz (Friedrich Miescher Institute for Biomedical Research, Basel, Switzerland) for tethering assay plasmids, Jens Lykke-Andersen (Molecular, Cellular and Developmental Biology, University of Colorado) for providing the rabbit anti-Dcp1 antibody, Tom Hobman (Department of Cell Biology, University of Alberta) for providing the hAgo2 rabbit antibody, and Peter Sayeski (Department of Physiology and Functional Genomics, University of Florida) for rabbit anti-GST antibody. We appreciate help from Hideko Kasahara (Department of Physiology and Functional Genomics, University of Florida) in the design of SYBR-Green quantitative real time PCR and also help from Diane Duke (UF Hybridoma Core Laboratory, Gainesville, FL) for the generation of hybridomas, subcloning and her generous advice during the process. This work was supported in part by the Canadian Institutes for Health Research Grant MOP-38034, NIH Grant AI47859, and the Andrew J. Semesco Foundation, Ocala, FL. M.J.F. holds the Arthritis Society Chair. S.L. and S.L. are supported by NIDCR oral biology training grant T32 DE007200. J.J.M. is supported by a CIHR Doctoral Research Award in the Area of Clinical Research and by an Alberta Heritage Foundation for Medical Research Studentship Award.