Primer- and template-independent RNA polymerization by terminal nucleotidyltransferase TENT4B Authors: Neha Nagpal, Albert K Tai, Yick W Fong, Suneet Agarwal Published: 2026 ABSTRACT AbstractRNA synthesis by eukaryotic polymerases requires existing polynucleotides to serve as templates or primers. Here, we describe primer- and template-free RNA generation by human terminal nucleotidyltransferase 4B (TENT4B) via de novo polymerization of free nucleotides. We observed that recombinant TENT4B (rTENT4B) consumes ATP to yield inorganic pyrophosphate in the absence of a primer or template, concurrent with the appearance of oligomeric poly-adenosine RNA products. Remarkably, 5′ labels on γ-phosphate-modified ATP or GTP are retained during polymerization in the presence of unlabeled nucleotide triphosphates (NTPs). These polymers are created at a similar efficiency irrespective of the inclusion of a primer, indicating robust RNA synthesis by rTENT4B from free NTPs. While canonical purine NTPs are favored, nucleotide diphosphates (NDPs) can also serve as substrates for rTENT4B-mediated de novo RNA polymerization. rTENT4B-mediated RNA synthesis using free adenosine nucleotides shows high processivity to generate 1000s-mers, whereas guanosine nucleotide polymerization is strongly and uniformly self-limited and yields a 3′-exonuclease-resistant oligonucleotide. Interrogation of other RNA polymerases reveals potential capacity for de novo polymerization using free ATP, albeit at significantly higher substrate concentrations and lower efficiency compared with rTENT4B. Our data provide definitive evidence of efficient template-free de novo RNA synthesis by a eukaryotic polymerase. Introduction RNA polymerases use nucleotide triphosphates (NTPs) to generate polymers using pre-existing nucleic acids in two non-mutually exclusive ways: (i) template-dependent polymerization to create a product complementary to a DNA or RNA sequence; and (ii) primer-dependent polymerization via 3′ extension of another RNA species. Template-dependent RNA polymers are mainly formed de novo starting with two free NTPs by mechanisms well characterized at the biochemical and structural levels for diverse polymerases ranging from eukaryotic RNA polymerase II to T7 bacteriophage RNA polymerase, which is widely exploited in biotechnology [1–4]. In most cases, template-dependent polymers begin with a free purine NTP (ATP or GTP) opposite a templating pyrimidine with synthesis of a short RNA, followed by extensive coordinated structural changes of the transcription complex into an efficient template-directed elongation machinery [5]. Template-directed RNA polymerization by viruses can also be initiated from the free hydroxyl group of a pre-existing nucleic acid or protein [6, 7], but generally, primer-dependent RNA polymerization refers to template-independent polynucleotide addition to existing short or long coding and non-coding RNAs (ncRNAs) by ribonucleotidyltransferases [8]. Amongst the best characterized of these are canonical poly(A) polymerases (PAPs) that modify mRNAs in a termination-coupled manner downstream of endonucleolytic cleavage at the AAUAAA sequence motif, thus promoting mRNA stability, nuclear export, and translation [9, 10]. The CCA-adding enzyme is a notable ncRNA-dependent RNA polymerase that undergoes structural changes to ensure addition of a precise, untemplated trinucleotide to achieve functional tRNA maturation [11]. Despite these diverse properties, there are no examples of known polymerases using neither a primer nor a template for RNA synthesis. Distinct from canonical PAPs, non-canonical template-independent RNA polymerases in eukaryotes are grouped under the category of terminal nucleotidyltransferases (TENTs) and serve varying RNA quality control and maturation functions [8, 12, 13]. Eleven TENTs have been identified in humans, sharing homology in the nucleotidyltransferase domains with PAPs and each other, but with varying NTP and RNA substrate preferences [14]. These include the terminal uridyltransferase TUT1 (aka TENT1) which modifies U6 small nuclear RNA (snRNA) to promote its biogenesis, and TUT4 (aka TENT3A) and TUT7 (aka TENT3B) which oligouridylate mRNAs, microRNAs (miRNAs), and other ncRNAs to cause RNA degradation [15, 16]. A similar role in miRNA regulation is played by TENT2 but via oligoadenylation [17]. MTPAP (aka TENT6) encodes the mitochondrial PAP which oligoadenylates and regulates the function of mitochondrial mRNAs and tRNAs [18–20]. The roles of the four TENT5 family members are emerging as diverse, and include innate immunity, gametogenesis, and secretory functions in different cell types [21–24]. The remaining family members, TENT4A and TENT4B, are unusual amongst TENTs in displaying a relaxed substrate specificity, primarily utilizing ATP but also GTP, followed by CTP and UTP [25–29]. This promiscuity may be relevant in vivo for mixed adenylyl and guanylyl tailing in the stabilization of mRNAs [30]. Interestingly, TENT4A and TENT4B have been found to play critical roles in polynucleotide tailing of RNAs from viruses including hepatitis B virus (HBV) and hepatitis A virus (HAV) [31–35]. Distinct roles for TENT4B have also been shown in oligoadenylation of several human ncRNAs including telomerase RNA (TERC), small nucleolar RNAs (snoRNAs), rRNAs, and miRNAs [27, 36–41]. TENT4B has emerged as a therapeutic target due to its role in the regulation of nascent TERC and HBV surface antigen (HbSAg) RNA. A phenotypic screen to reduce HbSAg expression from hepatocytes, a hallmark of chronic infection, yielded potent small molecule dual inhibitors of TENT4A and TENT4B, revealing their unexpected role in viral mRNA formation and thus providing new targets for chronic HBV infection [32, 42, 43]. Independently, we and others identified a role for TENT4B in destabilizing TERC, the non-coding template for telomerase that maintains telomere length and stem cell self-renewal capacity [25, 36, 44–46]. TENT4B oligoadenylates and targets nascent TERC transcripts for degradation by the nuclear exosome, a process that is critically opposed by the exonuclease poly(A) ribonuclease PARN [27, 36, 37, 47, 48]. In genetic forms of human telomere biology disorders (TBDs), PARN loss of function leads to unopposed TERC destabilization by TENT4B, and in turn reduced TERC RNA, telomerase insufficiency, and myriad diseases due to premature stem cell exhaustion. These findings led us to screen for small molecule inhibitors of recombinant TENT4B (rTENT4B) as a potential pharmacological approach to restore telomerase function in individuals with genetic TBDs [25]. Using TENT4B inhibitors from our screen as well as those identified in HbSAg studies, we showed restoration of TERC RNA steady-state levels and telomere length in vitro in patient-derived induced pluripotent stem cells and in vivo in human hematopoietic stem cells in a mouse xenotransplantation model of TBD [25, 31, 43]. While performing the high-throughput inhibitor screen, we unexpectedly identified novel activities of rTENT4B in RNA synthesis. Here, we demonstrate that rTENT4B can initiate RNA polymerization from free ATP or GTP in the absence of a primer or template. rTENT4B shows highly processive de novo RNA polymer formation from free NTPs, similar to using an oligonucleotide primer. We find that purine nucleotide diphosphates (NDPs) can also be used as substrates for de novo chain synthesis by rTENT4B, albeit at lower efficiency. Targeted evaluation of other canonical and non-canonical polymerases shows their variable capacity for primer- and template-independent RNA synthesis, but at higher ATP concentrations than rTENT4B and without definitive proof of primer-free initiation. Our data define novel primer- and template-free RNA synthesis capabilities of a eukaryotic RNA polymerase. These results promise to broaden our understanding of RNA polymerase function and evolution, and may prove to be useful in biotechnology applications. Materials and methods Purification of recombinant proteins For rTENT4B [full length, wild type (WT)], cDNA (GenBank CCB84642.1) was cloned into a modified pMtac-His6 vector [49] and purified as described [25]. For rTENT4B (full length, catalytic site mutant), cDNA (GenBank CCB84642.1) was cloned into a modified pMtac-His6 vector, and two aspartate residues in the catalytic domain were exchanged to alanine [26]; the resulting construct was purified as for WT full-length TENT4B. For recombinant TENT2 (full length) purification, the codon-optimized TENT2 (NCBI NP_001107865.1) sequence was introduced into a modified pMtac-His6 vector and the purification was performed as described [25]. For recombinant PARN (full length) purification, the codon-optimized PARN (NCBI_NP_002573.1) sequence was introduced in a modified pMtac-His6 vector, in frame with a C-terminal FLAG tag, and purification was performed as described [25]. For the recombinant TENT4B C-terminal truncation construct (TENT4B-delC) encoding amino acids 1–511 (deletion of amino acids 511–631) of GenBank CCB84642.1, cDNA was cloned into a modified pMtac-His6 vector, and expressed and purified by Ni-NTA affinity chromatography as described [25], except that proteins were eluted in elution buffer with 0.1 M NaCl. Peak Ni-NTA fractions were pooled and applied to Q Sepharose Fast Flow pre-equilibrated in elution buffer [50 mM HEPES-KOH, pH 7.6, 0.1 M NaCl, 2 mM MgCl2, 0.05% NP-40, and 10% glycerol containing 2 mM dithiothreitol (DTT), 1 mM benzamidine, and 0.5 mM phenylmethylsulfonyl fluoride (PMSF)]. The Q Sepharose flow-through fraction contains active TENT4B-delC and was dialyzed against elution buffer containing 0.4 M NaCl. For recombinant TENT4B and TENT4A catalytic cores (for TENT4B, amino acids 115–453 of GenBank CCB84642.1; for TENT4A, amino acids 225–560 of NCBI_NP_008930.2), cDNA was cloned into a modified pMtac-His6 vector, expressed, and purified by Ni-NTA affinity chromatography as described [25], except that proteins were eluted in elution buffer with 0.125 M NaCl. Peak Ni-NTA fractions were pooled and applied to Q Sepharose Fast Flow pre-equilibrated in elution buffer (50 mM HEPES-KOH, pH 7.6, 0.125 M NaCl, 2 mM MgCl2, 0.05% NP-40 and 10% glycerol containing 2 mM DTT, 1 mM benzamidine, and 0.5 mM PMSF). The Q Sepharose flow-through fraction contains active TENT4B-catalytic core and was dialyzed against elution buffer containing 0.4 M NaCl. Luciferase assay ATP consumption was measured using Kinase-Glo Max reagent (Promega, V6074). Enzymatic reactions with rTENT4B and ATP were performed in a buffer containing 25 mM Tris–HCl, 5 mM MgCl2, and 50 mM KCl, in the presence or absence of 1 pmol of an RNA oligo (oligo A15) as substrate, in 10 µl in 384-well format (Corning, 3820). After 1 h (or at the times indicated in the figures), reactions were stopped with 5 µl of Kinase-Glo reagent followed by centrifugation at 2000 rpm for 1 min. Luciferase signal was read using an Envision plate reader (Perkin Elmer). Each reaction was performed in n = 3 biological replicates. RNA oligonucleotides RNA nucleotides were synthesized by IDT and purified by RNase-free HPLC. oligo1, /5′-6-FAM/CUGCCUGCCUGCCUGCCUGC; oligo2, CUGCCUGCCUGCCUGCCUGC; oligo A15, AAAAAAAAAAAAAAA; and oligo3, /5′-6-FAM/CUGCCUGCCUGCCUGAAAAA. Nucleotide substrates ATP (N0450), GTP (N0450), CTP (N0450), and UTP (N0450) were obtained from New England Biolabs, ADP (NU-1198), Cy5-ATP [γ-(6-aminohexyl)-ATP-Cy5 (NU-833-CY5)] and Cy5-GTP [γ-(6-aminohexyl)-GTP-Cy5 (NU-834-CY5)] were from Jena Bioscience, and AMP (A1752),GDP (G7127), UDP (94 330), and CDP (C9755) were from Sigma. Polymerase reactions For RNA substrate extension assays, reactions were performed using various NTPs or NDPs with rTENT4B or other TENTs. Briefly, 2.5 pmol of rTENT4B (or TENTs) were added per 10 µl reaction containing 1 pmol of RNA oligo1 and 25 mM Tris–HCl (pH 7.4), 50 mM KCl, 5 mM MgCl2, plus 100 µM ATP or 1 mM NDP/ NTP (or the concentrations indicated in the figures). Reactions were incubated at room temperature for 1 h (or at the time indicated in the figures), stopped using formamide loading buffer (10 mM EDTA and 83.3% formamide), and resolved using denaturing polyacrylamide gels (15% Criterion™ TBE–urea polyacrylamide gel, 26 well, 15 µl, Bio-Rad, 3 450 093). Gels were imaged using the FAM channel of the GE Amersham Typhoon biomolecular imager. De novo synthesis or no-substrate reactions were performed as above but without an RNA oligo. The reactions products were resolved using denaturing gels as above, synthesized RNA products were detected using Sybr Gold Nucleic Acid stain (ThermoFisher Scientific, S11494), and gels were imaged using the Cy3 channel of the imager. Reactions with CID1 or poly(U) polymerase (New England Biolabs, M0337S) were performed in enzyme-specific buffer and 1 U of enzyme was added per 10 µl of the reaction mix. Reactions were incubated at 37°C for 1 h, stopped using formamide loading buffer, and resolved and imaged as above. Cy5-labeled NTP reactions were performed using similar buffer conditions to those above. Briefly, 250 µM ATP (or at the concentrations indicated in the figures) or other NTPs or NDPs (at the concentrations indicated in the figures) were incubated with 100 µM Cy5-ATP or Cy5-GTP and 2.5 pmol of rTENT4B or other TENTs. Reactions were performed at room temperature for 30 min (or at the times indicated in the figures) and stopped using formamide loading buffer. Reaction products were column purified (Oligo Clean and Concentrator, D4060) to remove unincorporated nucleotides or Cy5 dyes and resolved using denaturing polyacrylamide gels (15%). Gels were imaged using the Cy5 channel of the imager. Post-imaging, to visualize non-Cy5-labeled nucleic acid products, the gels were stained using Sybr Gold and imaged using the Cy3 channel of the imager. Reactions with yeast poly(A) polymerase (YPAP; Jena Biosciences, RNT-006) were performed using the manufacturer’s buffer. Briefly, 300 U of YPAP were added per 10 µl reaction containing 100 µM Cy5-ATP as above, and 1000 µM ATP (unlabeled), and incubated at 37°C for 30 min. Reactions were stopped using formamide loading buffer, and column purified and resolved as described above. Post-imaging, to visualize non-fluorescent RNA products, the gels were stained using Sybr Gold and imaged using the Cy3 channel of the imager. RG7834 treatment For RG7834’s effects on de novo synthesis, 100 µM (final concentration) of RG7834 (Medkoo Biosciences, 563 793) solution in dimethylsulfoxide (DMSO) was added to 10 µl of reaction mix containing 100 µM Cy5-ATP with 2.5 pmol of rTENT4B in the presence/absence of 100 µM ATP, in the absence of an RNA substrate. Reactions were performed at room temperature for 30 min and stopped using formamide loading buffer. Reaction products were column purified (Oligo Clean and Concentrator, D4060) to remove unincorporated nucleotides or fluorescent dyes, and resolved using denaturing polyacrylamide gels (15%). Gels were imaged using the Cy5 channel of the imager. Pyrophosphate analysis To detect the pyrophosphate (PPi) released during the polymerase reactions, reactions were performed in 96-well plates (Costar 96-well Flat-Bottom EIA plate, BioRad, 2 240 096). ATP at 100–1000 µM (or at the concentrations indicated in the figures) or 1000 µM GTP and 50 pmol of rTENT4B were incubated in a 100 µl reaction containing buffer, dye, and kit components (EnzChek Pyrophosphate Assay kit, Thermofisher Scientific, E6645), with or without 25 pmol of oligo2. Reactions were performed at room temperature for 30 min (or at the times indicated in the figures), and absorbance was measured at 360 nm using the CLARIOstar BMG Labtech microplate reader. For reactions with YPAP and Escherichia coli poly(A) polymerase (EPAP), 3000 U of YPAP or 25 U of EPAP were added per 100 µl of reaction buffer and other kit components (EnzChek Pyrophosphate Assay kit), reactions were incubated at 37°C, and absorbance was measured as above. Differential scanning fluorimetry Direct binding of NTPs to rTENT4B was detected using the indicator dye SYPRO orange (ThermoFisher Scientific, S6651) diluted 1:5000 in 20 µl of buffer containing 20 µM rTENT4B alone or with either 1 pmol RNA oligo A15, or 100 µM NTP, 25 mM Tris–HCl, 5 mM MgCl2, 50 mM KCl. Dye–buffer–protein mixtures were heated from 10 to 95°C at a rate of 1°C min−1, and fluorescence signals were monitored by 7500 Fast Real-Time PCR Systems (Applied Biosystems). Each curve represents an average of three measurements; Thermal Shift software (ThermoFisher Scientific, 4 466 038) was used for raw data analysis, and GraphPad Prism was used for plotting the melt curves. Ribonuclease treatment For polymerase reactions using GTP and FAM-labeled RNA oligo: reactions were performed using 1000 µM GTP with 2.5 pmol of rTENT4B and 1 pmol of RNA oligo 3. Reactions were incubated at room temperature for 1 h and stopped using formamide loading buffer. Reaction products were column purified (Oligo Clean and Concentrator, D4060) to remove unincorporated nucleotides, and reaction products were subjected to exonuclease treatments as below. For rPARN, 0.001 pmol of rPARN was incubated with purified reaction products in buffer containing 25 mM Tris–HCl (pH 7.4), 50 mM KCl, 5 mM MgCl2, 0.1mg ml−1 bovine serum albumin (BSA), 0.02% NP-40, 0.01 mM EDTA, and 1 mM DTT in a 10 µl reaction. Reactions were incubated at room temperature for 20 min. For exonuclease T (New England Biolabs, M0265), 5 U of enzyme was used in enzyme-specific buffer in a 10 µl reaction and incubated at room temperature for 30 min. Reactions were stopped using formamide loading buffer and resolved using 20% TBE–urea polyacrylamide gels. Gels were imaged using a GE Amersham Typhoon biomolecular imager in the FAM channel. For polymerase reactions using Cy5-labeled NTPs: reactions were performed using 100 µM Cy5-ATP and 10 µM ATP or 1000 µM GTP with 5 pmol of rTENT4B (catalytic core), without any RNA oligo. Reactions were incubated at room temperature for 30 min and stopped using formamide loading buffer. Reaction products were column purified (Oligo Clean and Concentrator, D4060) to remove unincorporated nucleotides and fluorescent dye components. Polymerase reaction products were subjected to exonuclease treatments with rPARN and exonuclease T as described above. For RNase R (Lucigen, NC1163864) treatments, polymerase reaction products were treated with 10 U of RNase R enzyme with enzyme-specific buffer in a 10 µl reaction at 37°C for 10 min. All the reactions were stopped using formamide loading buffer, resolved using 20% TBE–urea polyacrylamide gels, and imaged in the Cy5 channel of the imager. Quantification and statistical analysis Error bars presented the mean with the standard error (SE). P-values were calculated based on Student’s t-test or analysis of variance (ANOVA) (see the figure legends) and P < 0.05 was defined as significant. Statistical analyses were done using GraphPad Prism 10 (V10.6.1) software. Results Primer-independent ATP polymerization by rTENT4B During our high-throughput screen to identify TENT4B inhibitors, we observed that ATP was consumed by rTENT4B in the absence of an RNA oligonucleotide substrate. Specifically, our screen measured residual ATP substrate in an rTENT4B primer elongation reaction via luciferase-mediated light intensity, an elevation in which would indicate enzyme inhibition. Interestingly, we observed a similar ATP concentration-dependent decrease in luciferase signal over time irrespective of oligonucleotide addition (Fig. 1A). To determine the basis of the loss of luciferase signal, we assessed whether ATP hydrolysis products PPi versus inorganic phosphate (Pi) were being generated, as measured using a Pi assay with and without incubation with pyrophosphatase (PPA). We found that the majority of products accompanying ATP depletion by rTENT4B were PPi irrespective of oligonucleotide addition (Fig. 1B), and that a catalytically dead rTENT4B did not consume ATP (Supplementary Fig. S1A), arguing against a contaminating ATPase activity in the rTENT4B purification process that would be expected to generate Pi. We next asked whether contaminating RNAs in trace amounts in the rTENT4B preparation could be serving as primers to underlie the ATP consumption and PPi generation. To assess this, we treated 100 pmol of rTENT4B (100× the amount of RNA oligo used for comparison) with proteinase K, purified the residual material under conditions for isolating RNA, and subjected it to 5′ end labeling with γ-32P-ATP and polynucleotide kinase. One pmol of a 15-mer homopolymeric adenosine RNA oligonucleotide (oligo A15) was also end-labeled and used as a control. However, we were unable to detect labeled RNA in this manner, suggesting that significant quantities of contaminating RNA species in the rTENT4B prep do not serve as substrates to account for the observation of primer-independent ATP consumption by rTENT4B (Supplementary Fig. S1B). These data indicated an unusual convergence of both primer- and template-independent PPi production from ATP by a polymerase. De novo ATP polymerization by rTENT4B leads to RNA synthesis and RNA substrate 3′ extension. (A) Upper: schematics of ATP consumption by rTENT4B in the presence/absence of an RNA oligo as substrate and generation of residual ATP. Lower: luciferase-based detection of ATP consumption by rTENT4B in the presence of an RNA substrate oligo compared with reactions with no RNA substrate added. ATP alone in the absence of rTENT4B was used as control (no protein control). n = 3 biological replicates. The mean ± SD is shown, two-way ANOVA, **P < 0.01, ****P < 0.001. (B) Upper: schematics of ATP conversion and detection via PPi assay, depending on the presence of PPA, PPi or Pi would be released. Lower: consumption of ATP by rTENT4B via analysis of PPi yield, in the presence/absence of PPA as a measure of ATP hydrolysis (ADP+ Pi) versus 3′ extension of an RNA substrate oligo yielding AMP + PPi. n = 2 biological replicates. The mean ± SD is shown, two-way ANOVA, ns, not significant, ***P < 0.005. (C) De novo ATP polymerization using rTENT4B. rTENT4B was incubated with ATP (100 µM) in the absence of an RNA substrate. Generation of de novo synthesized nucleic acid products was detected using a Bioanalyzer (Agilent picoChip analysis). (D) Upper: schema of replacing the RNA substrate oligo with 5′-Cy5-blocked ATP as an initiator nucleotide in the RNA substrate extension assay. Lower: 5′-Cy5-labeled ATP was subjected to 3′ adenylation by rTENT4B, showing generation of Cy5-fluorescent RNA chains of varying length, elongating over time. n = 3, one representative is shown. (E) Comparison of 3′ adenylation by rTENT4B using 5′-Cy5-labeled ATP in the presence/absence of 100 µM RG7834 (rTENT4B inhibitor), showing generation of Cy5-fluorescent RNA chains of varying length, which disappeared in the presence of RG7834. n = 3, one representative is shown. We next asked whether it was possible that rTENT4B was mediating de novo nucleotide polymerization from free ATP molecules. To test this, we performed a time-course of rTENT4B reactions with varying concentrations of ATP and analyzed the products by polyacrylamide gel electrophoresis (PAGE). In this manner, we detected concentration- and time-dependent increases in high molecular weight nucleic acid species, beginning at 10 µM ATP within 15 min (Supplementary Fig. S1C). Polymers > 6000 nt were detected within 15 min at higher ATP concentration (100 µM), indicating high processivity (Fig. 1C). ATP consumption in the luciferase assay correlated with the appearance of these products (Supplementary Fig. S1D). To definitively evaluate for de novo synthesis from a free ATP molecule, we next asked whether rTENT4B could initiate a polymer using a fluorescent γ-phosphate-labeled ATP plus unlabeled ATP. In these experiments, if γ-phosphate-labeled ATP is utilized for initiation of an RNA polymer, the label would be retained and thus detectable in progressively higher molecular weight species. Conversely, if a contaminating primer is initiating polymers rather than ATP, the γ-phosphate would be lost and labeled high molecular weight species would not be detectable. Remarkably, using γ-[(6-aminohexyl)-imido]ATP-Cy5 ([γ-Cy5]ATP) in the presence of rTENT4B, we were able to detect abundant fluorescently labeled polymers of increasing length as a function of unlabeled ATP concentration and time (Fig. 1D; Supplementary Fig. S1E, F). These data definitively demonstrate de novo primer- and template-independent polymerization of free ATP molecules by a eukaryotic polymerase, TENT4B. We next verified that the core catalytic activity of rTENT4B is required for de novo polymerization of ATP. We have previously shown that pharmacological inhibition of rTENT4B prevents RNA oligo 3′ adenylation. Thus we tested RG7834, a potent TENT4B inhibitor discovered in HbSAg screens [31, 42, 43], for effects on its de novo nucleotide polymerization activity. We detected newly synthesized fluorescent chains with Cy5-labeled-ATP in the presence of rTENT4B, but not after RG7834 treatment (Fig. 1E). These data support de novo RNA synthesis as a catalytic property of rTENT4B that is amenable to pharmacological inhibition. Next, based on reports that a conserved motif of basic amino acids at the C-terminus of rTENT4B is required for RNA substrate binding [26], we tested the ability of a C-terminal truncation construct (rTENT4B-delC) to mediate primer-free RNA polymerization. Again, we found that fluorescent RNA chains could be formed by rTENT4B-delC using [γ-Cy5]ATP as the initiating nucleotide (Supplementary Fig. S1G). Finally, we performed a phylogenetic analysis of the mammalian TENT4B catalytic core and found ≥ 98.8% identity, with common variation at two amino acid positions. We generated an rTENT4B construct containing these two amino acid changes and found that the catalytic core containing these variants retained robust capacity for primer-independent RNA polymerization using [γ-Cy5]ATP, equivalent to the human rTENT4B core (Supplementary Fig. S1H). Taken together, these results demonstrate that the core catalytic function of rTENT4B includes a capacity for primer- and template-independent polymerization of free ATP molecules. High efficiency of de novo ATP polymerization by rTENT4B We next interrogated the relative efficiency of de novo versus primer-dependent nucleotide polymerization by rTENT4B. To do this, we quantified PPi production from ATP by rTENT4B in the presence or absence of an RNA oligonucleotide. We found that the generation of PPi by rTENT4B increases in an ATP concentration- and time-dependent manner (Fig. 2A, B). However, PPi generation is not significantly different in the presence or absence of 25 pmol of oligonucleotide primer, at concentrations as low as 5 µM ATP and reaction durations from 30 min to plateau at 180 min (Fig. 2A, B). Direct comparison of kinetic parameters showed similar TENT4B initial velocities of PPi generation at a range of ATP concentrations with or without an RNA oligonucleotide primer (Fig. 2C). We next asked whether other RNA polymerases showed the capacity for de novo NTP polymerization. We began by testing canonical RNA PAPs and observed no yield of PPi when incubating either eukaryotic (yeast) PAP or prokaryotic (E. coli) PAP with 100 µM ATP in the absence of an RNA oligonucleotide (Fig. 2D). In contrast, we observed a significant increase in PPi yield with YPAP at 1000 µM ATP without an oligonucleotide, whereas EPAP failed to yield PPi at 1000 µM ATP without an oligonucleotide (Fig. 2E; Supplementary Fig. S2A, B). No significant difference was observed with 1000 µM versus 100 µM ATP using rTENT4B, with or without an oligonucleotide (Supplementary Fig. S2C). No Cy5-labeled high molecular weight polymers were observed with yeast PAP when tested using [γ-Cy5]ATP and 1000 µM unlabeled ATP (Fig. 2F). However, high molecular weight fluorescent signals indicating nucleic acid polymers were detected using Sybr Gold intercalating dye on the same denaturing gels (Fig. 2F). This result potentially indicates that YPAP cannot utilize [γ-Cy5]ATP as an initiating NTP as opposed to ATP. Alternatively, we cannot rule out that the YPAP preparation might contain contaminating RNA that is acting as a primer. Taken together, these data indicate a highly efficient and potentially unique capacity for TENT4B to polymerize [γ-Cy5]ATP and unlabeled ATP relative to canonical PAPs. Rate of de novo ATP polymerization by rTENT4B compared with PAPs. (A) Analysis of PPi generated by rTENT4B in response to increasing concentrations of ATP, in the presence of an RNA oligo as substrate, compared with reactions with no RNA substrate added, and ATP alone in the absence of rTENT4B (no protein control). n = 2 biological replicates. Mean ± SD is shown, two-way ANOVA, ns, not significant. (B) Analysis of PPi yield generated by rTENT4B in the presence/absence of an RNA oligo as substrate as a measure of ATP consumption over time. n = 2 biological replicates. Mean ± SD is shown. (C) Michaelis–Menten plots depicting the initial reaction velocities of rTENT4B in no RNA substrate reactions compared with RNA substrate (oligo) extension reactions. (D) Comparison of PPi yield generated by canonical PAPs from yeast and bacteria (E. coli PAP) via using 100 µM ATP as substrate in the presence/absence of an RNA oligo as substrate in 30 min. n = 2 biological replicates. Mean ± SD is shown, two-way ANOVA, **P < 0.01, ***P < 0.005. (E) Comparison of PPi yield generated by canonical PAPs from yeast and bacteria (E. coli PAP) via using 100 µM and 1000 µM ATP as substrate in reactions with no RNA oligo added in 30 min. n = 2 biological replicates. Mean ± SD is shown, two-way ANOVA, ns, not significant, ***P < 0.005. (F) Comparison of ATP polymerization using rTENT4B and canonical PAP from yeast using 5′-Cy5-labeled ATP as substrate along with 1000 µM ATP. ATP incorporation was confirmed via detecting generation of Cy5-labeled RNA chains after 30 min in reactions with no RNA oligo added. Unlabeled nucleic acid products generated during the reaction were detected via staining with Sybr Gold. n = 3, one representative shown. We next tested other non-canonical PAPs including rTENT4A, rTENT2, and CID1 for their ability to catalyze primer-independent de novo RNA synthesis. As expected, all recombinant enzymes could extend RNA oligonucleotide primers (Supplementary Fig. S2D). In all cases, high molecular weight fluorescent signals using Sybr Gold in the absence of RNA oligonucleotides could be observed using these non-canonical PAPs at 1000 µM ATP (Supplementary Fig. S2E), albeit with varying product yields that may be reflective of the recombinant protein preparation and/or inherent differences in their efficiency of primer-independent ATP polymerization (Supplementary Fig. S2F, G). Taken together, we find that other non-canonical polymerases might also be able perform de novo synthesis of RNA polymers using ATP as a substrate. De novo, self-limiting polymerization by rTENT4B using GTP TENT4B has been shown to utilize NTPs other than ATP in vitro and in cells [25, 26, 28, 30]. By differential scanning fluorimetry, we found that rTENT4B has the capacity to bind to and thus be stabilized by any of the four canonical NTPs (Supplementary Fig. S3A). We next sought to assess the NTP utilization by rTENT4B in de novo RNA synthesis. We first used [γ-Cy5]ATP with unlabeled NTPs and analyzed the reaction products for synthesis of fluorescent RNA polymers. We found that rTENT4B could generate polymers starting with [γ-Cy5]ATP at high concentrations of GTP (1–3.3 mM) but not CTP or UTP (Fig. 3A). Next, we tested de novo GTP polymerization using rTENT4B in the absence of [γ-Cy5]ATP. Remarkably, we detected strong and discrete oligonucleotide products by Sybr Gold staining at GTP concentrations (1.5–5 mM) similar to those observed in [γ-Cy5]ATP experiments (Fig. 3B). We verified that the polymerization of free GTP nucleotides by rTENT4B was accompanied by PPi generation (Fig. 3C). Interestingly, the polymers created using GTP were restricted to ≤ 20 nt, versus > 6000 nt for ATP at similar concentrations and time frames (Figs 1C and 3A, B; see below). Next, we tested whether de novo chains could be initiated with γ-[(6-aminohexyl)]GTP-Cy5 ([γ-Cy5]GTP) rather than [γ-Cy5]ATP, using unlabeled ATP. We found robust formation of high molecular weight species retaining a 5′ Cy5 label when we added unlabeled ATP, that increased in a concentration-dependent manner (Fig. 3D). Finally, we analyzed the capacity of rTENT4B to utilize any NTP at high concentrations in the absence of an RNA primer and analyzed the results with Sybr Gold staining. This experiment confirmed de novo synthesis of nucleic acid polymers using ATP or GTP but not CTP or UTP. Again, chain formation was self-limiting with GTP (Fig. 3E). These data demonstrate de novo polymerization of GTP by rTENT4B, in a self-limiting manner compared with ATP. De novo NTP polymerization by rTENT4B. (A) rTENT4B plus 5′-Cy5-labeled ATP as substrate reacted with each of four canonical NTPs: ATP (A), CTP (C), UTP (U), and GTP (G). NTP incorporation was confirmed via detecting generation of Cy5-labeled RNA chains after 30 min in reactions with no RNA substrate added. n = 3, one representative shown. (B) De novo GTP polymerization using rTENT4B. rTENT4B along with increasing concentrations of GTP was subjected to RNA extension assays in reactions with no RNA substrate added, and generation of de novo synthesized nucleic acid products was detected using a denaturing TBE–urea polyacrylamide gel followed by Sybr Gold staining. n = 3, one representative shown. (C) rTENT4B along with GTP was subjected to RNA substrate extension assay in the absence of an RNA substrate, and GTP polymerization was detected via PPi analysis. n = 4 biological replicates as shown. Mean ± SD, two-tailed unpaired t-test ****P < 0.001. (D) 5′-Cy5-labeled GTP was subjected to 3′ adenylation using increasing concentrations of ATP and rTENT4B, showing generation of fluorescent (Cy5)-labeled RNA chains of varying length. n = 3, one representative shown. (E) De novo NTP polymerization using rTENT4B. rTENT4B along with increasing concentrations of four canonical NTPs was subjected to RNA substrate extension assay in no RNA substrate added reactions, and generation of de novo synthesized RNA products was detected using a denaturing TBE–urea polyacrylamide gel followed by Sybr Gold staining. n = 3, one representative shown. (F) 3′ exonuclease degradation assay using exonuclease T (ExoT), recombinant purified PARN, or single strand-specific RNase R (RnR), for de novo synthesized Cy5 fluorescent RNA products by rTENT4B, in reactions with no RNA substrate added, via using Cy5-labeled ATP alone versus Cy5-labeled ATP plus unlabeled ATP or unlabeled GTP. n = 2, one representative shown. Earlier we demonstrated that rTENT4B can utilize 2′-O-methyl-ATP to form a self-limiting polymer on RNA 3′ ends, in a manner that both constrained further nucleotide addition and also conferred 3′ exonuclease resistance [50]. Homopolymeric poly(G) sequences are poorly soluble and may form stable secondary structures (e.g. G-quadruplexes). Thus the self-limiting nature of oligo(G) polymer formation might be a reflection of intrinsic limitations of the enzyme and/or formation of a secondary structure or inaccessible product. If the latter, we hypothesized that the poly(G) products would be resistant not only to further enzymatic polymerization but also to enzymatic degradation. In keeping with this, we found that an RNA oligonucleotide extended using rTENT4B and GTP formed discrete high molecular weight species that unlike the unmodified oligonucleotide were completely resistant to 3′ exonucleases, exonuclease T, and recombinant PARN (rPARN) (Supplementary Fig 3B). Next, we tested whether the de novo synthesized RNA products produced by rTENT4B from [γ-Cy5]ATP plus unlabeled ATP versus unlabeled GTP were differentially sensitive to exonucleases. Indeed, we found that γ-Cy5-labeled polymers formed by rTENT4B using GTP were resistant to exoribonucleases compared with those generated via using ATP (Fig 3F). RNase R (RnR) is a 3′ exonuclease with a characteristic capacity to degrade highly structured RNAs. We found that GTP-formed oligonucleotides were sensitive to RnR compared with exonuclease T and rPARN (Fig 3F), in keeping with the notion that a secondary structure underlies their nuclease resistance. Collectively, these results demonstrate the ability of rTENT4B to utilize not only free ATP but also GTP in de novo RNA synthesis, but in a self-limiting manner yielding exonuclease-resistant polymers. Purine nucleotide diphosphates serve as substrates for rTENT4B We have previously shown that rTENT4B could utilize ADP in RNA oligonucleotide extension reactions (Supplementary Fig. S4A) [50]. We thus asked whether other NDPs could serve as substrates for rTENT4B in RNA polymerization. In a primer elongation assay, we first determined that rTENT4B could utilize GDP but not UDP or CDP, and only at higher concentrations than ADP (1 mM versus 0.1 mM) (Fig. 4A). As with GTP, short, self-limiting polymers were observed with GDP compared with those generated using ADP (Fig. 4A). We next tested de novo RNA synthesis via using NDPs and rTENT4B in primer-free reactions with [γ-Cy5]ATP as a substrate, and observed robust γ-Cy5 polymer formation using ADP and GDP. As expected, no fluorescent polymers were generated by rTENT4B when incubated with CDP or UDP (Fig. 4B). When free ADP was utilized as a substrate for rTENT4B without an RNA primer or [γ-Cy5]ATP, high molecular weight RNA polymers could be detected using Sybr Gold within 1 h, albeit at high concentrations ≥ 3.3 mM (Fig. 4C). When using free ADP as a substrate, Pi was generated instead of PPi, as expected (Fig. 4D). Finally, we tested rTENT4B for utilization of ADP in the presence of increasing concentrations of Pi. We found that there was no pronounced inhibition of product extension or polymer degradation even when the molar concentration of Pi was greater than that of ADP (Supplementary Fig. S4B). In contrast, when we tested recombinant bacterial polynucleotide phosphorylase (PNPase), which carries out NDP polymerization versus 3′–5′ exoribonuclease activity as a function of Pi concentration [51], RNA oligo degradation was observed with increasing concentrations of Pi (Supplementary Fig. S4C). These data collectively suggest that rTENT4B can utilize purine diphosphate nucleotides in primer- and template-independent RNA synthesis. Purine NDP polymerization by rTENT4B. (A) Utilization of NDPs by rTENT4B. rTENT4B along with a 20-mer RNA substrate oligo (oligo2) was incubated in the presence of each of four canonical NDPs in RNA substrate extension assays, and reaction products were resolved using a denaturing TBE–urea polyacrylamide gel followed by Sybr Gold staining. (B) rTENT4B was incubated with each of four canonical NDPs plus 5′-Cy5-labeled ATP as substrate. NDP incorporation was confirmed via detecting generation of Cy5-labeled RNA chains after 30 min in reactions with no RNA substate added. (C) De novo ADP polymerization using rTENT4B. rTENT4B with increasing concentrations of ADP was subjected to RNA substrate extension assay in the absence of an RNA substrate, and generation of de novo synthesized RNA products was detected using a denaturing TBE–urea polyacrylamide gel followed by Sybr Gold staining. (D) Consumption of ADP by rTENT4B via analysis of Pi yield in the presence/absence of PPA, with or without an RNA substrate oligo, compared with to that of rTENT4B alone versus rTENT4B plus ATP. The Pi yield for ADP reactions is provided based on the PPi standard curve. n = 4 biological replicates. Mean ± SD is shown, two-way ANOVA, ns, not significant, ****P < 0.001. Discussion Our data show primer- and template-free polymerization of RNA from free nucleotides by TENT4B, a property that has not been definitively ascribed to a polymerase in the past. Qβ replicase is an RNA-dependent RNA polymerase from E. coli bacteriophage Qβ previously proposed to be capable of catalyzing de novo synthesis of various RNA oligonucleotides from free NTPs, i.e. in a primer- and template-independent manner [52]. While initial concerns about potential RNA primer contamination [53] were addressed using more rigorous purification methods [54], it was ultimately found that Qβ replicases utilize short host cell-derived RNA templates in a reaction that becomes autocatalytic [55, 56]. Thus Qβ replicase conforms to classification as a primer-independent but template-directed RNA polymerase. In 1955, PNPase from Azotobacter vinelandii was shown to polymerize RNA from NDPs in vitro, enabling synthesis of polyribonucleotides used to crack the genetic code [57, 58]. While PNPase participates in RNA synthesis in prokaryotic cells under certain circumstances [51], the primary role of PNPase across diverse species is RNA degradation via phosphorolysis [59]. To our knowledge, this report is the first of primer- and template-free RNA synthesis by a eukaryotic polymerase. Our experimental demonstration of de novo RNA polymerization using rTENT4B exploited its ability to utilize [γ-Cy5]ATP or [γ-Cy5]GTP as the initiating nucleotide. In retaining the 5′ label, single nucleotide additions were readily visualized using free purine nucleotides. We did not detect a lag in reaction kinetics as measured by PPi production without an oligonucleotide, indicating that engagement of a free nucleotide by rTENT4B in the “product” (P)-site of the polymerase active site appears to be as efficient as that of an RNA primer, although the precise quantitation of this remains to be established. This is further indicated by robust activity of a C-terminal mutant of rTENT4B that does not contain a putative RNA-binding site. Once initiated, peak PPi generation and the extent of product formation were similar with or without an RNA oligonucleotide primer. These data and the 1000s-mer reaction products rapidly generated using ATP indicate that rTENT4B-mediated RNA polymerization is highly processive in vitro and limited only by the available ATP pool. The physiological relevance of the novel properties and capabilities of rTENT4B we have characterized here is not known. TENT4B is the human ortholog of yeast Trf4p, which forms the TRAMP complex with Air2 and Mtr4p involved in RNA quality control [60–63]. Trf4p oligoadenylates nascent nuclear RNAs that are potentially misfolded with an available 3′ end. Via the TRAMP complex, these modified RNAs are targeted for rapid degradation by the nuclear RNA exosome. Similarly, human TENT4B forms multi-protein complexes with orthologs of Air2 and Mtr4, ZCCHC7 and MTR4, respectively [64, 65], catalyzing oligoadenylation-mediated exosome targeting of incomplete and cryptic promoter upstream transcripts and other RNAs including TERC and snoRNAs [47, 66–68]. In vitro, yeast Mtr4p limits polyadenylation by Trf4p by stimulating its dissociation from the substrate after addition of 3–4 adenosines [69]. In human cells, RNA targets such as TERC and scaRNA13 display oligoadenylation (<10 nt) rather than polyadenylation when TENT4B activity is unopposed due to PARN disruption [70], also suggesting that TENT4B processivity is more constrained in a physiological setting than what we have observed in vitro. However, a recent report implicates TENT4B in untemplated 3′ tailing of rRNA fragments of several hundred nucleotides long (so-called starRNA) in the setting of amyloidogenesis [71], suggesting that under certain conditions TENT4B may display higher processivity in cells. We have previously demonstrated that complete inactivation of TENT4B by CRISPR/Cas9 is tolerated in human cells without adverse effects on cell viability or growth [25, 72]. Nevertheless, a role for de novo or processive NTP polymerization by TENT4B in vivo remains to be determined. Despite showing a preference for purines, rTENT4B is capable of condensing all four canonical nucleotides into polymers [25, 26], an observation with potential relevance in biotechnology. Complete chemical synthesis of short RNA species has enabled novel therapeutics such as antisense oligonucleotides (ASOs) and short interfering RNAs (siRNAs) [73], but has limitations in terms of polymer length and synthetic capacity, including environmental concerns at large scale due to use of hazardous chemicals [74]. Recent studies demonstrate enzymatic synthesis of custom short RNA polymers using mutant forms of CID1 and reversible terminator NTPs, but this approach relied on a pre-existing oligonucleotide of at least 10 nt in length followed by an endonucleolytic cleavage step [75]. Given TENT4B’s homology to CID1 as members of the superfamily of polymerase beta-like nucleotidyltransferases [76], and its inherent NTP promiscuity relative to other nucleotidyltransferases for canonical and modified ribonucleotides [26, 50], it is tempting to speculate that engineered forms of TENT4B could be used for de novo, template-independent enzymatic synthesis of RNAs in aqueous solution, without relying on an initiating oligonucleotide. As such, further studies of rTENT4B could advance the emerging field of complete enzymatic RNA synthesis, which promises to overcome barriers including polymer length and composition, scale, and environmental safety that limit the expansion of RNA biotechnology and therapeutics. Supplementary Material Acknowledgements The authors thank K. Arnett for help with kinetic analyses. This content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Author contributions N.N. and S.A. conceived this study and designed the experiments. N.N. performed in vitro and biochemical analyses. A.K.T. performed computational analyses. Y.W.F. purified recombinant proteins. S.A. and N.N. wrote the manuscript. All authors edited the manuscript. Contributor Information Neha Nagpal, Division of Hematology/Oncology and Stem Cell Program, Boston Children’s Hospital, Boston, MA 02115, United States; Pediatric Oncology, Dana-Farber Cancer Institute, Boston, MA, United States; Harvard Stem Cell Institute, Cambridge, MA, United States; Department of Pediatrics, Harvard Medical School, Boston, MA, United States; Manton Center for Orphan Disease Research, Boston, MA, United States; Harvard Initiative for RNA Medicine, Boston, MA, United States. Albert K Tai, Department of Immunology, Tufts University School of Medicine, Boston, MA, United States. Yick W Fong, Harvard Stem Cell Institute, Cambridge, MA, United States; Brigham Regenerative Medicine Center, Cardiovascular Division, Brigham and Women’s Hospital, Boston, MA, United States; Department of Medicine, Harvard Medical School, Boston, MA, United States. Suneet Agarwal, Division of Hematology/Oncology and Stem Cell Program, Boston Children’s Hospital, Boston, MA 02115, United States; Pediatric Oncology, Dana-Farber Cancer Institute, Boston, MA, United States; Harvard Stem Cell Institute, Cambridge, MA, United States; Department of Pediatrics, Harvard Medical School, Boston, MA, United States; Manton Center for Orphan Disease Research, Boston, MA, United States; Harvard Initiative for RNA Medicine, Boston, MA, United States. Supplementary data Supplementary data are available at NAR online. Conflict of interest N.N. and S.A. are listed as co-inventors on provisional patent applications that include de novo nucleotide polymerization work described in this manuscript. Funding Boston Children’s Hospital (BCH) Department of Pediatrics and Manton Center for Orphan Disease Research, Uplifting Athletes Young Investigator Draft, and Team Telomere [to N.N.]; the National Instiutes of Health (NIH) [R01GM152463 to Y.M.F. and R01DK107716 to S.A.]; the Tufts University Core Facility Genomics Core [RRID:SCR_012501 to A.K.T.]; BCH Translational Research Program [to S.A.]. Source of Open Access funding: Institutional discretionary funds. Material availability Request for materials, reagents, and resources used in this study should be directed to the corresponding author, Suneet Agarwal (suneet.agarwal@childrens.harvard.edu). Requests for certain materials will require a completed Materials Transfer Agreement. Data availability The data that support the findings of this study are available within the manuscript and associated Supplementary data files. Unprocessed data will be provided on request. There are no restrictions on data availability from this study. References Associated Data Supplementary Materials Data Availability Statement The data that support the findings of this study are available within the manuscript and associated Supplementary data files. Unprocessed data will be provided on request. There are no restrictions on data availability from this study.