Zcchc11 is a member of the DNA polymerase β-like (pol β) nucleotidyltransferase superfamily ⁵ and it interacts with adaptor proteins in the Toll-like receptor (TLR) signaling pathway ⁶. Overexpression of a truncated form of Zcchc11 (containing amino acids 191 to 625 of the 1644 amino acid protein, including only the first 2 of the 7 domains of interest identified in Fig. 1a) decreased expression of the cytokines IL-6 and TNF-α ⁶. Because the degrees to which this truncation mutant conferred gain or loss of function were unclear, such an approach indicates that Zcchc11 mediates cytokine expression without illuminating the mechanism(s) by which it performed this function. The Zcchc11 protein contains domains capable of interacting with RNA, suggesting that the mechanism(s) for regulating cytokine expression may extend beyond its interaction with adaptors. As depicted in Figure 1a, these domains include a CCHH zinc finger, typically interacting with RNA or DNA; two poly(A) polymerase/2′–5′ oligoadenylate synthetase (PAP/25A) associated domains, common in RNA modifying enzymes ⁷; three CCHC zinc knuckles, often RNA-binding domains ^{8, 9}; and a pol β domain, usually mediating nucleotidyltransferase activity ⁵. The putative pol β catalytic domain of Zcchc11 is highly conserved across genomes from a wide spectrum of eukaryotes (Fig. 1b). Altogether, this primary structure suggested that Zcchc11 might be a ribonucleotidyltransferase.

Several proteins with homology to Zcchc11 are categorized as non-canonical poly(A) polymerases, including the Caenorhabditis elegans protein defective in germ line development 2 (GLD-2) ¹⁰ and the Schizosaccharomyces pombe proteins caffeine-induced death protein 1 (Cid1) and Cid13 ^{11, 12}. To determine whether Zcchc11 possessed such activity, we generated recombinant epitope-tagged wildtype Zcchc11 as well as a mutant form (Zcchc11 DADA) in which two aspartate residues predicted to be essential to catalysis ¹³ were mutated to alanines (Fig. 1b). Recombinant epitope-tagged Zcchc11 exhibited nucleotidyltransferase activity for all RNA substrates examined (Fig. 1c). In contrast, the DADA mutant exhibited no transfer of adenosine to RNA, implicating the pol β domain as responsible (Fig. 1c). Endogenous Zcchc11 immunoprecipitated from a human lung epithelial cell line (A549) also exhibited nucleotidyltransferase activity (Supplementary Fig. S1a). The radiolabeled product was sensitive to RNase digestion, confirming RNA as the substrate of the enzyme (Supplementary Fig. S1b). These data show that Zcchc11 is a ribonucleotidyltransferase enzyme whose activity is dependent on a conserved catalytic site.

Some pol β family members were recently identified as poly(U) polymerases ^{14, 15}. Given this, we examined nucleotides other than ATP in Zcchc11 assays. Both endogenous (Fig. 1d) and recombinant epitope-tagged (Supplementary Fig. S1c) forms of Zcchc11 displayed the greatest nucleotidyltransferase activity with UTP, followed by CTP, and then ATP and GTP. The inclusion of an ATP regenerating system, designed to circumvent ATP loss from potentially co-precipitating ATPases, did not enhance adenosine incorporation (data not shown). As with adenylation, the uridylation activity of Zcchc11 did not display RNA substrate specificity, evidenced by the labeling of diverse RNA moieties including even an RNA ladder (Fig. 1e). These results indicate that Zcchc11 catalyzes the non-selective, template-independent uridylation of RNA.

Although Zcchc11 mRNA was ubiquitous among mouse tissues (data not shown), Zcchc11 protein was tissue-specific, most prominently expressed in the thymus, spleen, testes, and lung (Supplementary Fig. S1d). Zcchc11 was also expressed in human and mouse lung alveolar epithelial cell lines (A549 and MLE-15, respectively: Supplementary Fig. S1e). Expression of Zcchc11 in immune and mucosal cells pointed again to potential roles in cytokine regulation. Effective knockdown in A549 and MLE-15 cells by Zcchc11-targeted siRNA compared to control, non-targeting siRNA was confirmed by immunoblotting using different siRNAs specific to human or mouse Zcchc11 (Fig. 2a). Cell supernatants were collected from control or Zcchc11 siRNA-treated A549 cells stimulated with TNF-α, and cytokine expression was analyzed. A subset of cytokines, namely IL-6, VEGF, TGF-α, and RANTES, was substantially decreased by loss of Zcchc11 (Fig. 2b). In contrast, fractalkine, IP-10, MCP-1, G-CSF, and GM-CSF levels were not significantly affected, and IL-8 concentrations were increased (Fig. 2b). These cytokine-specific effects were not unique to A549 cells, but were also observed in HEK 293T cells treated with control non-targeting or Zcchc11 targeting siRNA prior to TNF-α treatment (data not shown). The effects of Zcchc11 knockdown on IL-6 elaboration by A549 cells was examined over a time-course after TNF-α. Zcchc11 depletion led to decreased IL-6 at all time points examined, with differences between groups increasing over time (Fig. 2c). Zcchc11-dependent effects on IL-6 expression were not unique to TNF-α stimulation, as Zcchc11 depletion similarly diminished IL-6 production elicited by Eschericia coli (Supplementary Fig. S2a). As with results obtained in human A549 cells, IL-6 expression was decreased in murine MLE-15 cells by knockdown using a different siRNA targeting mouse Zcchc11 (Fig. 2d). Thus, these results were not specific to a given siRNA sequence and were consistent across species and stimuli.

To determine whether Zcchc11 influenced cytokine expression at the mRNA level, we first measured steady-state transcripts. IL-6 mRNA was decreased by Zcchc11 knockdown (Fig. 2e). Cytokine transcript stability was subsequently assessed, showing mRNA half-life for IL-6 to be decreased by Zcchc11 knockdown (Fig. 2f). The effects of Zcchc11 knockdown on the stability of transcripts for other cytokines generally mirrored the effects of Zcchc11 knockdown on cytokine protein content (Supplementary Fig. S2b). These data suggest that Zcchc11 mediates cytokine expression at the posttranscriptional level by influencing mRNA stability.

The poly(A) tail length of eukaryotic mRNAs is a critical determinant of transcript stability and translational efficiency ^{16, 17}. By any of three independent measurement methods ¹⁸, the poly(A) tails on IL-6 mRNA were shortened due to Zcchc11 knockdown (Figs. 2g and Supplementary Fig. S2c,d). Other Zcchc11-dependent cytokines examined also demonstrated shorter poly(A) tails due to Zcchc11 knockdown (Supplementary Fig. S2e) whereas tail lengths for cytokines not decreased at the protein or mRNA stability levels were not shortened by Zcchc11 knockdown (Supplementary Fig. S2f). In summary, decreasing Zcchc11 results in shorter poly(A) tails on only those cytokines whose transcript stability and maximal protein expression require Zcchc11. Poly(A) tail length therefore appears to be the step at which Zcchc11 directly or indirectly mediates cytokine expression.

Because of its uridyltransferase activity, we considered whether Zcchc11 might uridylate the 3′ polyadenylate terminus of mRNAs, as was observed for the Zcchc6 homolog of Zcchc11 ¹⁴. IL-6 transcripts were isolated, cloned, and sequenced from control and Zcchc11 knockdown A549 cells, but no terminal uridines were observed (Supplementary Fig. S3). Given these results and the enzymatic activity of Zcchc11, we considered alternative RNA substrates that could influence transcript poly(A) tail length and stability.

mRNA deadenylation and degradation can be stimulated by miRNAs ^{19, 20}, leading us to reason that the RNA-modifying Zcchc11 enzyme may alter cytokine expression through an miRNA-dependent mechanism. Support for this hypothesis was precedented by observations that miRNAs often contain untemplated nucleotides on their 3′ ends, primarily adenosines or uridines ^21–23. The cytoplasmic poly(A) polymerase GLD-2 ²⁴, a relative of Zcchc11, is capable of adenylating the 3′ end of mature miRNA, but no enzyme has yet been identified as responsible for the uridylation of mature miRNA. Analyses using TargetScan ^25–27 revealed that the 3′ untranslated region (UTR) of IL-6 mRNA contains a highly conserved region potentially targeted by the mir-26 family of microRNAs (Fig. 3a), lying in close proximity to an AU-rich region and accessible within the predicted secondary structure of IL-6 mRNA, characteristics positively associating with miRNA-mediated repression ²⁸. Prior studies indicated that mir-26b in naïve T cells contains non-genomic uridines on the 3′ end ²⁹, supporting the postulate that miRNAs in this family including mir-26b in particular may be modified in the proposed fashion. Northern analyses demonstrated that mir-26b is expressed in A549 cells, with no effect of Zcchc11 knockdown on its content (Fig. 3b). Transfection of A549 cells with mir-26b mimetics decreased TNF-induced IL-6 expression (Fig. 3c), demonstrating that this miRNA can repress IL-6. Complementarily, blocking endogenous mir-26b using a mir-26b-antisense inhibitor ^{30, 31} increased IL-6 expression (Fig. 3d). Blockade of endogenous mir-26a also increased IL-6 expression (Supplementary Fig. S4a). Taken together, these gain-of-function and loss-of-function approaches demonstrate that this family of miRNAs targets IL-6. Because the mir-26 mimetics showed less of a repressive effect on IL-6 than did Zcchc11 knockdown, mir-26 may be but one of multiple miRNAs potentially modified by Zcchc11 to regulate IL-6 expression.

Zcchc11 was capable of uridylating mir-26b in vitro (Figure 3e). The enzyme showed a preference for the single-stranded RNA in comparison with the duplex, and this activity was abolished by point mutations in the nucleotidyltransferase domain (Figure 3e). Thus, Zcchc11 directly uridylates mature, single-stranded microRNAs including mir-26b. To our knowledge, there is no prior evidence that sequence variations in the 3′ ends of miRNAs impact their capability to repress targeted transcripts. Interestingly, we observed that 3′ end uridylation of mir-26b mimetics abrogated repression of IL-6 by this miRNA (Figure 3f). Uridylation of mir-26a mimetics also exhibited a decreased ability to repress IL-6 expression (Supplementary Fig. S4b). To test whether this was unique to the mir-26 targeting of IL-6, we transfected cells with uridylated mir-16 mimetics and assessed VEGF concentrations in the supernatant after TNF-α stimulation. We observed strong repression of VEGF by mir-16, as previously reported³², but this repression was not attenuated by mir-16 uridylation (Supplementary Fig. S4c). Additionally we sought to determine whether oligouridine tracts at miRNA 3′ termini are capable of increasing IL-6 expression independent of sequence specificity, such as by acting as ligands for TLR7 and TLR8 ³³. We found no differences in IL-6 production in the supernatants of TNF-stimulated cells treated with uridylated species of mir-16 (Supplementary Fig. S4d). Taken together, these data suggest that uridylation of IL-6 mRNA-targeting miRNAs may be the mechanism by which Zcchc11 influences IL-6 expression.

We next determined whether the influences of this enzyme and miRNA on IL-6 expression involved direct interactions with the 3′-UTR of the IL-6 transcript. The IL-6 3′-UTR was sufficient to reduce protein expression in a luciferase reporter assay (Figure 4a). Importantly, this repressive effect of the IL-6 3′-UTR was exacerbated by Zcchc11 knockdown (Figure 4a). Further, the overexpression of Zcchc11 increased expression of the IL-6 3′-UTR-associated gene (Figure 4b). Mutating the nucleotidyltransferase domain of Zcchc11 significantly diminished these effects (Figure 4b). Thus, the ability of Zcchc11 to enhance IL-6 expression is mediated by the 3′-UTR and is dependent upon its nucleotidyltransferase activity.

mir-26b gain of function further diminished expression associated with the IL-6 3′UTR (Figure 4c). The addition of terminal uridines to this miRNA decreased its repressive activity for the IL-6 3′ UTR (Figure 4c). Importantly, both the decreased expression driven by mir-26b and the counteracting increased expression resulting from its uridylation were abrogated by mutation of the predicted miRNA target site within the 3′-UTR (Figure 4d). Overall these results demonstrate that the effects observed are direct results of miRNA interactions with the 3′UTR.

Altogether, our data strongly suggest that Zcchc11 influences IL-6 expression by the untemplated uridylation of mir-26b or related IL-6 3′-UTR-targeting miRNAs. We thus deep sequenced small RNAs isolated from control and Zcchc11 knockdown cells to assess 3′ terminal modifications of mir-26 family members. In A549 cells, mir-26a sequences were far more abundant than mir-26b (7,068 and 235 sequences recovered, respectively). Of the mir-26a sequences collected from control cells, the majority (74%) were the 22-nucleotide sequence reported in Figure 3a, ending in a single uridine which could be genomically encoded. An additional 4% of sequences contained 2 or 3 uridines rather than the single uridine at the end, which must necessarily represent untemplated uridylation (Fig. 4e and Supplementary Fig. S5).

The Zcchc11 knockdown cells demonstrated remarkably different results in that out of 10,306 mir-26a sequences collected, none (0%) were the sequence reported in Figure 3a. Instead, the majority (67%) were 21 nucleotides in length and missing the terminal uridine (Fig. 4e), suggesting dicer cleavage of the mir-26a pre-miRNA at this nucleotide followed by Zcchc11-mediated uridylation in the control cells. For mir-26a sequences with definitively non-genomic uridines in the Zcchc11 knockdown cells, only 2 (0.02%) sequences ended in UU and none (0%) in UUU. These data indicate that Zcchc11 mediates the addition of 1 to 3 uridines to the 3′ terminus of mir-26a, defining a role for Zcchc11 in miRNA end modification.

Most (33%) of the remaining mir-26a sequences in the knockdown cells ended in UA (Fig. 4e), also observed in the control cells albeit with lesser frequency (8%). One possible explanation for such a pattern could be an enzyme which adds UA residues to microRNAs, potentially competing with Zcchc11 for mir-26a modification. Alternatively, since adenylation of miRNAs increases miRNA stability ²⁴, it is conceivable that mir-26a that had been uridylated by Zcchc11 and then adenylated by GLD-2 or some other enzyme, but not non-adenylated forms of uridylated Zcchc11, may persist from before Zcchc11 was knocked down.

Not only were mir-26b sequences less common than mir-26a, but the former also showed a distinct pattern of non-genomic end modification, with UA ends in 99% and 96% of mir-26b sequences from control and knockdown cells, respectively (Supplementary Fig. S6). Based on prior studies showing non-genomic uridines but not UA additions on mir-26b in T cells ²⁹, we suggest that the modifications of miRNAs is context-dependent. The identity of putative UA-adding enzyme or adenylating enzymes in A549 cells and the significance of these nucleotide additions for mir-26b function remain to be determined.

The addition of non-templated nucleotides, most prominently uridine and adenine, at the 3′ terminal ends of miRNAs has been observed in multiple species ^{34, 21, 35, 22, 23, 36, 37}. The mechanisms determining which miRNAs are modified by Zcchc11, and under which circumstances, remain important avenues for future study. To our knowledge, these studies provide the first identification of an enzyme responsible for the untemplated uridylation of mature miRNA. Such miRNA modifications are not simply by-products of sloppy processing, but rather they impart functional differences that attenuate miRNA-targeted repression. We further demonstrate this to be a means of post-transcriptionally regulating cytokine expression, potentially relevant to inflammation and immunity.

1

2

3

4

5

6

7