A noncanonical Pol III-dependent, Microprocessor-independent biogenesis pathway generates a germline-enriched miRNA family

In this study, Sakhawala et al. describe a noncanonical miRNA biogenesis pathway in which a germline-enriched, functional family of miRNAs—derived from independent RNA polymerase III transcribed transcripts—is processed in a manner dependent on Dicer but not on Microprocessor or other miRNA processing factors. Such differently processed miRNAs were identified in C. elegans and human data sets, suggesting that these nonclassical, young, de novo miRNAs may be representative of the evolution of conserved miRNA genes.

Abstract

MicroRNAs (miRNAs) are short RNAs that post-transcriptionally regulate gene expression. In canonical miRNA biogenesis, primary miRNAs are transcribed from intergenic loci or intronic regions by RNA polymerase II and sequentially cleaved by the Microprocessor complex and Dicer, and the resulting mature miRNAs are loaded into Argonaute to repress target mRNAs. A minority of miRNAs are generated via noncanonical biogenesis pathways that bypass the Microprocessor complex and/or Dicer. Here, we describe a new Pol III-dependent, Microprocessor-independent, and Dicer-dependent biogenesis pathway exemplified by the mir-1829 family in Caenorhabditis elegans. Although the mir-1829 family loci reside in intronic regions of protein-coding genes, we show that the miRNAs are derived from independent Pol III transcripts. Unlike other Pol III-dependent miRNAs, the mir-1829 family small RNAs are the dominant species derived from their loci rather than fragments of a larger functional noncoding RNA. These germline-enriched miRNAs are loaded in multiple miRNA Argonautes, including the recently characterized germline Argonaute ALG-5, which we demonstrated is repressive when tethered to a reporter transcript. We extend these findings, identifying additional Pol III transcribed and noncanonical small RNAs in C. elegans and human data sets, including human miR-4521. These young, noncanonical miRNAs may represent an early snapshot in the evolution of de novo miRNA genes.

MicroRNAs (miRNAs) are small, ∼22 nt regulatory RNAs that selectively repress gene expression during development and differentiation. In canonical miRNA biogenesis, primary miRNAs are transcribed from intergenic loci or intronic regions by RNA polymerase II (Ha and Kim 2014). These transcripts are then processed by Microprocessor (MP), a complex that consists of the catalytic RNase III enzyme Drosha and a homodimer of the RNA-binding protein DGCR8 (known as DRSH-1 and PASH-1 in Caenorhabditis elegans, respectively) (Lee et al. 2003; Denli et al. 2004; Gregory et al. 2004; Han et al. 2004). MP cleaves the stem of hairpin structures of primary miRNAs (pri-miRNAs) to produce miRNA precursors, which are further processed into small RNA duplexes by the RNase III enzyme Dicer (DCR-1 in C. elegans) (Grishok et al. 2001; Hutvágner et al. 2001; Knight and Bass 2001). One strand of the resulting duplex is preferentially loaded into an Argonaute (Ago) protein, forming the miRNA-induced silencing complex (miRISC), which can target mRNAs via base pairing to their 3′ UTR to repress translation and/or mediate the recruitment of factors that cause mRNA decay (Gebert and MacRae 2019).

Beyond canonical miRNA biogenesis, multiple alternative biogenesis pathways bypass MP and/or Dicer. One of the best-studied examples is mirtron biogenesis, in which a short intron takes on a hairpin-like structure after debranching; this hairpin is then processed as a miRNA precursor by Dicer, thus bypassing the requirement for MP (Berezikov et al. 2007; Okamura et al. 2007; Ruby et al. 2007). Some mirtron-like molecules, termed “Agotrons,” bypass both MP and Dicer and associate with Agos as full-length introns (Hansen et al. 2016). Short, 5′ 7-methylguanine (m⁷G)-capped Pol II transcripts can also bypass MP, being directly processed by Dicer before Ago loading (Babiarz et al. 2008; Xie et al. 2013; Zamudio et al. 2014; Sheng et al. 2018). Dicer-independent biogenesis is exemplified by vertebrate miR-451, which is cleaved by MP but then relies on the catalytic activity of Ago to further its maturation (Cheloufi et al. 2010; Cifuentes et al. 2010; Yang et al. 2010).

In addition to these Pol II transcripts that undergo noncanonical processing, Pol III-derived transcripts can also be fragmented (generally in a MP-independent, Dicer-dependent manner), giving rise to Ago-loaded small RNAs. Regions of tRNAs that form hairpins can be further processed by Dicer to produce miRNA-like small RNAs (Babiarz et al. 2008; Maute et al. 2013; Martinez et al. 2017; Kuscu et al. 2018). Similarly, specific snoRNAs and the small NF90-associated RNA A (snaR-A) serve as sources for distinct Dicer-dependent small RNA species (Ender et al. 2008; Lemus-Diaz et al. 2020; Stribling et al. 2021). Dicer can also generate small RNAs from hairpins formed by transcribed SINEs (Babiarz et al. 2008). Overall, diverse Pol III transcripts give rise to less abundant small RNA products that can function as miRNAs.

In this study, we present a new noncanonical miRNA biogenesis pathway exemplified by the mir-1829 family in C. elegans. Although they reside within long introns of Pol II-derived host genes, the mir-1829 family members are derived from independent transcriptional units transcribed by Pol III. The mir-1829 family is MP-independent but Dicer-dependent. The mir-1829 family is enriched in the germline and, in particular, enriched in ALG-5, a germline-specific Ago protein; we show via a tethering assay that ALG-5 is repressive and thus likely functions similarly to other miRNA Agos. Unlike previously identified Pol III-dependent miRNAs, the mir-1829 family small RNAs appear to be the primary product of these loci, not derivatives of a more abundant functional species. We further expanded our search for additional noncanonical and Pol III-derived miRNAs by cross-referencing data sets in which MP, Dicer, and Pol III are depleted, identifying multiple novel candidates in C. elegans and human samples. We confirmed the Pol III dependence and noncanonical biogenesis of human miR-4521. The mir-1829 family, miR-4521, and additional candidate miRNAs are evolutionarily young; thus, these miRNAs may represent a snapshot of miRNA loci at an early stage of their emergence.

Results

The mir-1829 family is Microprocessor-independent

While investigating miRNA decay in C. elegans adults using a temperature-sensitive (ts) allele of the DGCR8 ortholog pash-1 (Lehrbach et al. 2012; Vieux et al. 2021), we noticed a small portion of miRNAs that appeared to be insensitive to pash-1 inactivation (Fig. 1A, X-axis). These pash-1-insensitive miRNAs included both known MP-independent mirtrons (Ruby et al. 2007; Chung et al. 2011; Jan et al. 2011) and uncharacterized miRNAs such as the mir-1829 family. To determine whether the mir-1829 family is also MP-independent, we compared the pash-1(ts) data with another data set in which RNAi and the auxin-induced degron system were used together (RNAiD) to stringently deplete MP in embryos (Fig. 1A, Y-axis; Dexheimer et al. 2020). When comparing the two data sets, we observed that most miRNAs are sensitive to MP inactivation in both data sets, as shown by their decreased abundance (Fig. 1A). Multiple miRNAs decrease upon MP inactivation only in the adult data set; many of these are substrates of EBAX-1-dependent miRNA decay (likely target-directed miRNA degradation) in adults but not embryos (Supplemental Fig. S1A; Kotagama et al. 2024; Stubna et al. 2025); thus, their differential stability (higher in embryos) could explain their differential depletion upon MP inactivation. Other miRNAs that show differential MP sensitivity could also represent instances of differential stability or could be due to differences in the experimental systems [embryo RNAiD vs. adult pash-1(ts)]. miRNAs that are insensitive to MP inactivation in both data sets included MP-independent mirtrons and uncharacterized miRNAs. Among the novel MP-insensitive miRNAs, the vast majority are members of the mir-1829 family. The mir-1829 family is comprised of four miRNA loci—mir-1829a, mir-1829b, mir-1829c, and mir-4812—with highly similar sequences, including a shared seed sequence in the 3p-derived strand. Both 5p and 3p miRNAs of these four loci are MP-independent. (A fifth locus, F39B1.3, bears high similarity to those of the mir-1829 family but does not appear to be expressed.) Thus, we hypothesized that the mir-1829 family is representative of a new noncanonical MP-independent class of miRNAs.

Figure 1.

The mir-1829 family is Microprocessor-independent. (A) Comparison of log₂(fold change) of miRNAs 24 h after upshift to restrictive temperature in pash-1(ts) versus wild type (X-axis) (Vieux et al. 2021) to those in RNAiD depletion of the Microprocessor (Y-axis) (Dexheimer et al. 2020). miRNAs are annotated as having a “short hairpin” if the predicted structure in MirGeneDB has <32 bp (Fromm et al. 2020). A small number of miRNAs are insensitive to Microprocessor depletion in both data sets, including mirtrons and the mir-1829 family. Only miRNAs with baseMean ≥ 10 in the pash-1(ts) data set are plotted. (B) Schematics of miRNA structures emphasizing the long basal stem of canonical miRNAs and the lack thereof in mirtrons and the mir-1829 family. (C) Predicted secondary structures of pri-mir-1829b and pri-mir-1829c. (D) Cleavage efficiency scores of Drosha (X-axis) or Microprocessor (Y-axis) were calculated as log₂(∑N_P + 0.1) − log₂(∑N_S + 0.1), where N_P and N_S are normalized counts of cleaved products and pri-miRNA substrates, respectively (Nguyen et al. 2023). The mir-1829 family is among the least favorable substrates. (E) Genome browser tracks showing raw sequence reads from bias-minimized small RNA cloning from Stubna et al. (2025). (F) Northern blot of total RNA from adult animals exposed to auxin for 24 h starting at the L4 stage at 20°C. mir-1829b-5p, mir-1829c-5p, and U6 probes were 5′ end-labeled, whereas the mir-1 probe was labeled according to the Starfire method (Behlke et al. 2000). High stringency conditions were used for probing; see the Materials and Methods for details. (G) The same samples used for Northern blotting were assayed by RT-qPCR. miRNA expression was normalized to a small RNA control (sn2429) and then further normalized to wild-type samples. Mean and standard deviation of six biological replicates are shown. Two-way ANOVA; (****) P < 0.0001, (***) P < 0.001, (**) P < 0.01.

Canonical miRNAs are characterized by an ∼35 bp/∼87 nt stem–loop hairpin that serves as an optimal substrate for MP cleavage (Fig. 1B; Auyeung et al. 2013; Fang and Bartel 2015; Jin et al. 2020; Partin et al. 2020; Kang et al. 2021; Kim et al. 2021; Nguyen et al. 2023). Mirtrons, which are derived from splicing, bypass MP and serve as substrates for Dicer cleavage. Mirtrons generally have shorter hairpins than canonical miRNAs, lacking a basal stem that is characteristic of MP substrates (Fig. 1B; Westholm and Lai 2011). To determine whether newly identified MP-independent miRNA candidates also lack a basal stem, we examined the distribution of miRNAs derived from primary transcripts containing only a “short hairpin” (<32 bp) among the pash-1(ts) and MP RNAiD data. Although “short hairpins” were present among MP-sensitive miRNAs, the MP-insensitive miRNAs were highly enriched for “short hairpins,” highlighting both known mirtrons and novel MP-independent candidates like the mir-1829 family (Fig. 1A,B).

The structure of the mir-1829 family, along with their MP insensitivity, suggests that they are poor MP substrates (Fig. 1A–C). Consistent with this, the mir-1829 family was among the least efficiently cleaved pri-miRNAs in in vitro processing assays performed using purified C. elegans Drosha (Fig. 1D, X-axis) or the whole MP complex (Fig. 1D, Y-axis; Nguyen et al. 2023). Despite their apparently MP-independent biogenesis, the mir-1829 family members exhibit sequencing reads from both the 5p and 3p strands and are expressed at a level comparable with canonical miRNAs (Fig. 1E; Supplemental Fig. S1B).

We further confirmed these published sequencing data sets by examining miRNA levels via Northern blotting and RT-qPCR. We used strains in which PASH-1 and/or DRSH-1 are tagged with the auxin-inducible degron (AID) to deplete components of the MP complex for 24 h (Supplemental Fig. S1C; Dexheimer et al. 2020). Using stringent Northern blot conditions, we observed depletion of a canonical miRNA (mir-1) in samples depleted of PASH-1 and/or DRSH-1, demonstrating efficient inactivation of MP (Fig. 1F). (Note that high stringency conditions were used here to distinguish between mir-1829b-5p and mir-1829c-5p. See the Materials and Methods for details of high stringency conditions.) In contrast, mature mir-1829b-5p and mir-1829c-5p were not diminished in samples depleted of MP; in fact, they were modestly upregulated (Fig. 1F). (Among the mir-1829 family, only the 5p guide strands of mir-1829b and mir-1829c were detectable by Northern. We similarly focused on mir-1829b-5p and mir-1829c-5p for all RT-qPCR experiments because they were also the most robustly detectable mir-1829 family members in these assays.) Similar to Northern blots, RT-qPCR showed slightly higher levels of mir-1829b-5p and mir-1829c-5p in the same samples (Fig. 1G). The increased accumulation of noncanonical miRNAs upon MP inactivation has been noted previously (Babiarz et al. 2008; Kim et al. 2016; Sheng et al. 2018) and likely reflects reduced competition from more abundant canonical miRNAs for Ago loading and/or Dicer-mediated processing. Overall, these data confirm deep sequencing data indicating that MP inactivation does not impair mir-1829 family biogenesis. The caveat that the lack of sensitivity to MP depletion could be attributed to an unusually long half-life of the mir-1829 family is addressed below in experiments demonstrating Dicer dependence.

The mir-1829 family is germline-enriched

C. elegans has 19 Ago proteins, including three that are primarily dedicated to loading miRNAs: Argonaute-like gene 1 (ALG-1), ALG-2, and ALG-5 (Seroussi et al. 2023). Both ALG-1 and ALG-2 are expressed ubiquitously in the soma, and ALG-2 is also expressed in the germline and early embryos (Vasquez-Rifo et al. 2012; Aalto et al. 2018). ALG-5 expression is restricted to the germline (Brown et al. 2017; Seroussi et al. 2023). To examine potential downstream functions of the MP-independent mir-1829 family, we examined its Ago loading profile and tissue of expression. According to small RNA sequencing of immunoprecipitations of C. elegans Argonautes (IP-sRNA-seq), all members of the mir-1829 family are highly enriched in ALG-5 IPs and modestly enriched in ALG-2 IPs (Brown et al. 2017; Seroussi et al. 2023). Knockout of ALG-1 decreases levels of multiple mir-1829 family members, whereas ALG-2 knockout and ALG-5 knockout display depletion of one or two of the miRNA family strands, respectively (Supplemental Fig. S2A; Seroussi et al. 2023). Building on these data, we sought to determine the relative distribution of the mir-1829 family in each miRNA Argonaute via IP-RT-qPCR from strains in which ALG-1, ALG-2, or ALG-5 is tagged at its endogenous locus with a 3xFLAG epitope (as well as GFP or mKate2) (Brown et al. 2017; Aalto et al. 2018). We observed that mir-1829b-5p is evenly distributed between ALG-2 and ALG-5, whereas mir-1829c-5p is evenly distributed between all three miRNA Argonautes (Supplemental Fig. S2B).

We reasoned that our IP-RT-qPCR data could be consistent with IP-sRNA-seq data (in which higher amplitude of enrichment was observed in ALG-5 IPs than in ALG-1/2 IPs) if low abundance of ALG-5 imparts higher dynamic range to ALG-5 IP enrichment scores than those for ALG-1 and ALG-2. We therefore determined the relative expression of ALG-1, ALG-2, and ALG-5 in whole-animal adult samples; because each Argonaute is endogenously tagged with 3xFLAG, the anti-FLAG signal can be compared across strains to assess relative Argonaute abundance (Supplemental Fig. S2C, input lanes). Accordingly, ALG-1 makes up ∼60% of the miRNA Argonaute pool, whereas ALG-2 makes up ∼37% and ALG-5 makes up a mere ∼3% in whole-adult hermaphrodite samples (Supplemental Fig. S2D). Therefore, the dynamic range of enrichment from whole-animal samples is greater for ALG-5 IPs than for those of ALG-1 and ALG-2, explaining the even distribution of mir-1829b/c across multiple Argonautes. Although mir-1829 family members can be detected in some somatic tissues (Alberti et al. 2018; Brosnan et al. 2021), the exclusive loading of mir-1829b-5p in ALG-5 and ALG-2 suggests that its expression is very strongly germline-enriched, limiting association with ALG-1 (which is not expressed in the germline). The even distribution of mir-1829c-5p across ALG-1/2/5 suggests that its expression is strongly enriched in the germline (given that ∼33% of the miRNA is loaded into the 3% of the Argonaute pool comprised of ALG-5), along with some expression in the soma. Additionally, germline-specific profiling of miRNAs shows that all mir-1829 family members are highly expressed in the distal germline during the mitosis-to-meiosis transition (Diag et al. 2018).

Because of its germline-enriched expression, we sought to determine whether the mir-1829 family plays a role in germline function, specifically in fertility. Using CRISPR–Cas9, we generated strains that contain various combinations of mir-1829 knockouts including single knockouts of each family member, a double mutant of mir-1829b and mir-1829c, a triple mutant of mir-1829b/c and mir-4812, and a quadruple knockout of all four family members (Supplemental Fig. S2E). Reduced brood size and embryonic lethality were initially observed in the single mir-1829a deletion [mir-1829a(del)] and the full quadruple knockout of all members of the mir-1829 family [mir-1829 family(del)] at elevated temperatures (Supplemental Fig. S2E). However, these phenotypes were not rescued by a second round of CRISPR that restored the mir-1829a sequence [mir-1829a(rescue)] and thus are due to off-target effects of CRISPR (Supplemental Fig. S2E). The mir-1829 family members reside in introns of protein-coding genes (see also below), and the host gene of mir-1829a, gas-1, is required for normal fertility (Morgan and Sedensky 1994). We therefore redesigned the mir-1829a deletion to minimize effects on gas-1 expression by deleting the whole mir-1829a host intron and fusing the flanking exons together [mir-1829a(intron del)]; this strain did not show abnormal brood or embryonic viability phenotypes (Supplemental Fig. S2E). Finally, we used mir-1829a(intron del) to generate another quadruple mutant strain in which all four family members are deleted [mir-1829 family(null)], and this was also superficially wild type. Therefore, the mir-1829 family is dispensable for normal fitness in laboratory conditions (Supplemental Fig. S2E).

ALG-5 has a demonstrated role in promoting the mitosis-to-meiosis transition in the C. elegans germline, as evidenced by strong enhancement of the glp-1(gf) tumorous phenotype by alg-5(null) (Supplemental Fig. S2F; Brenner et al. 2022). Because of the expression of the mir-1829 family in the distal germline and its loading in ALG-5, we examined its role in this context. We observed that the quadruple knockout of the mir-1829 family does not modify the glp-1(gf) phenotype (Supplemental Fig. S2F); thus, it is not the primary miRNA guiding ALG-5's function in this context. Overall, the biological function of the mir-1829 family remains to be determined.

ALG-5 represses expression of a tethered mRNA

Because of its shared miRNA binding repertoire with ALG-1/2, ALG-5 likely acts as a repressor of its targets, though it has also been proposed to activate targets (Liontis et al. 2023). Because ALG-5's targets remain to be determined, we tested the functionality of ALG-5 using a tethering assay. Briefly, a single-copy GFP-histone transgene containing three box B RNA hairpins in its 3′ UTR is expressed throughout the germline (Aoki et al. 2021). In this background, we fused a 3xFLAG tag and λN22 peptide to the N terminus of ALG-5 by CRISPR (or 3xFLAG alone as a negative control background). Because the reporter is prone to silencing, we simultaneously inactivated mut-2 by CRISPR in both strains. Due to the strong interaction of the λN22 peptide with box B hairpins, 3xFLAG-λN22::ALG-5 but not 3xFLAG::ALG-5 should be tethered to the reporter transcript, allowing us to assess the impact of ALG-5 association with an mRNA by comparing the two strains. Imaging of these two strains shows lower expression of the GFP-histone reporter in 3xFLAG-λN22::ALG-5 than in 3xFLAG::ALG-5 in the distal germline of L4 and adult animals (Fig. 2A–C). Thus, this in vivo tethering assay suggests that ALG-5 acts similarly to other miRNA Agos by repressing associated transcripts.

Figure 2.

Both ALG-5-mediated reporter repression and mir-1829 family expression peak in the distal germline. (A) Measured GFP fluorescence intensity from five nuclei per region (distal, loop, and proximal) across nine L4 stage animals, grown at 20°C, expressing a germline single-copy GFP::histone transgene with three box B RNA hairpins in its 3′ UTR. CRISPR-generated animals expressing ALG-5 fusion proteins were tagged with either 3xFLAG and λN22 (experimental) or 3xFLAG alone (control) in a mut-2(null) background. (B) Measured relative GFP fluorescence intensity in day 1 adult worms (same strains as in A). Average fluorescence intensity was measured from six regions of interest (shown in the right schematic). Zones 1–3 correspond to regions of the distal gonad across from the −1, −2, and −3 oocytes, respectively. Five nuclei were measured in each zone in the distal germline per nine adult worms. (A,B) Mean and SD are shown. Two-way ANOVA; (****) P < 0.0001, (**) P < 0.01, (*) P < 0.05. (C) Representative images from A and B. (D) Schematic of relative expression levels of the mir-1829 family and its host genes in the C. elegans germline. Data from the Spatial Caenorhabditis elegans Germline Expression of mRNA and miRNA (SPACEGERM) Atlas (Diag et al. 2018; https://shiny.mdc-berlin.de/spacegerm) indicate that the mir-1829 family (pink) peaks in expression in the distal germline, whereas its host genes (black) peak in expression in the proximal germline. (E,F) Prominent Pol III ChIP-seq peaks (Araya et al. 2014) reside at the mir-1829 loci (pink). Arrowheads indicate 5′ RACE reads cloned from RNA of the indicated stage. Putative Pol III promoter sequence motifs upstream of the miRNA precursors are indicated.

The mir-1829 family is differentially transcribed from its host genes

The four mir-1829 family loci reside in the introns of three protein-coding genes of apparently unrelated function (gas-1, piki-1, and F20D1.3). As mentioned above, data from the Spatial Caenorhabditis elegans Germline Expression of mRNA and miRNA (SPACEGERM) Atlas (https://shiny.mdc-berlin.de/spacegerm) indicate that the mir-1829 family peaks in expression in the distal germline (Fig. 2D; Diag et al. 2018). Curiously, the host genes peak in expression in the proximal gonad, suggesting differential expression of the miRNAs and their host genes (Fig. 2D). Published ChIP-seq data of Pol III subunit RPC-1 further support the idea that the mir-1829 family is transcribed independently from its host genes and by Pol III (Araya et al. 2014); the mir-1829 loci (Fig. 2E,F, mir-1829 precursors shown in pink) are marked by prominent Pol III ChIP peaks, and these peaks are just as prominent as those at tRNA loci (Fig. 2E,F, purple boxes).

To investigate whether the differential expression between host gene and miRNA is due to differential transcription, we performed 5′ rapid amplification of cDNA ends (RACE) on RNA isolated from embryo, L4, and adult tissues using primers that bind in the apical loop of the miRNA hairpins. For each of the mir-1829 family loci, multiple 5′ RACE reads mapped within the host intron of the miRNA, suggesting alternative proximal transcription start sites of transcripts independent of the protein-coding host genes (Fig. 2E,F; Supplemental Table S3). For three of the four members of the mir-1829 family (mir-1829a–c), multiple 5′ RACE reads mapped consistently 101–103 bp upstream of the precursor, which further suggests that the mir-1829 family members share conserved regulatory elements that drive their independent transcription.

The mir-1829 family is transcribed by Pol III

Because results of 5′ RACE together with published Pol III ChIP suggest that the mir-1829 family may arise from independent Pol III transcriptional units, we sought additional evidence of their Pol III dependence. Type II Pol III transcripts (e.g., tRNAs) are defined by having box A and box B motif promoter elements located downstream from the transcription start sites (TSSs) (Fig. 3A; Dieci et al. 2007). For each mir-1829 family member, we searched upstream of the precursor for the box A and box B consensus sequence motifs previously defined for C. elegans (Ikegami and Lieb 2013; Stutzman et al. 2020). We determined that all four members have a putative box A and box B motif ∼79–87 and ∼35–42 bp upstream of the precursor, respectively. The positions of the motifs are downstream from the prominent −102 bp TSSs identified by 5′ RACE for mir-1829a, mir-1829b, and mir-1829c (Fig. 3A). Additionally, all the mir-1829 family loci have 5Ts immediately downstream from the precursor (3–4 bp downstream from the 3p arm), and another stretch of 4Ts–5Ts further downstream. These features further support our hypothesis that these are Pol III transcripts because typical type II Pol III genes have terminator sequences of 5Ts immediately downstream from the gene body (Ikegami and Lieb 2013).

Figure 3.

mir-1829 family members are transcribed by Pol III. (A) MUSCLE alignment of the mir-1829 family putative Pol III promoter sequence motifs. (B) Schematic of CRISPR alleles generated at the mir-1829b locus (left) and qPCR of the resulting miRNA and host gene expression (right) from young adult samples in wild-type background, grown at 25°C. (C) Schematic of genomic sources of mir-1829b (top left), including reintegration of a 510 bp minimal mir-1829b transcriptional unit in a mir-1829 family(null) mutant background (bottom left). (Right) qPCR of mir-1829b expression from young adult samples grown at 25°C. (D) qPCR of miRNAs and host gene transcripts in an AID-tagged RPC-1 strain with ubiquitous TIR1 (MCJ666) or TIR1 alone (MLC1040). Auxin treatment was for 24 h at 20°C, beginning at the L4 stage. (E) Schematics of promoter deletions for two host genes of the mir-1829 family. (F) qPCR in host gene promoter deletion samples from young adult samples grown at 25°C. (G) Schematic of CRISPR-generated mir-1829b alleles reintroduced into mir-1829 family(null) (left) and qPCR of the resulting miRNA (right) from young adults, grown at 25°C. (B–D,F–G) miRNA expression was normalized to a small RNA control (sn2429) and then further normalized to wild-type levels. Host gene expression was normalized to GAPDH (gpd-1) and then to wild-type levels. Mean and SD are shown. Two-way ANOVA; (****) P < 0.0001, (*) P < 0.05.

To determine whether the box A/B motifs are required for mir-1829 biogenesis, we performed promoter bashing experiments for mir-1829b. As a control, we used the mir-1829 family(null) mutant in which all the family members are deleted; in this background, a 312 bp deletion that removes mir-1829b abolishes expression of the miRNA without disrupting the host gene transcript, F20D1.3 (Fig. 3B). The mir-1829b locus contains a putative box A motif, a second degenerate box A motif, and a putative box B motif. We first deleted all the promoter motifs, an 82 bp deletion, keeping the TSS intact (box A/B del). This mutant resulted in abolished expression of the miRNA without disrupting the host gene transcript. To ensure that any changes in expression observed were not due to disruption of the secondary structure, we designed a new mutant (box A/B scramble) in which we scrambled the sequence within a 62 bp window, removing the box A and box B sequence motifs while maintaining secondary structure of the transcript as predicted by RNAfold (Gruber et al. 2008; Lorenz et al. 2011). Similar to the box A/B del mutant, we observed abrogated miRNA expression without disruption of the host gene transcript (Fig. 3B).

We next mutated each motif individually. To maintain structure for the box A scramble mutant, we introduced additional mutations into the degenerate box A motif, because they form a hairpin according to the RNAfold prediction. In this Box A scramble mutant, we observed abrogated mir-1829b-5p expression (Fig. 3B). We separately mutated only the degenerate box A motif in a manner that is not predicted to maintain secondary structure (second box A scramble); this resulted in no change in expression of mir-1829b-5p (Fig. 3B). Last, in the box B scramble mutant, we maintained the predicted structure while disrupting the sequence of the box B motif and observed loss of mir-1829b-5p expression. Together, these results indicate that the box A and box B motifs, individually, are each required for mir-1829b expression, whereas the degenerate box A motif is not.

Building on the promoter bashing experiments, we attempted to express the predicted mir-1829b locus from an ectopic location. We reintegrated a 510 bp minimal transcriptional unit (spanning 125 bp upstream of the TSS to 222 bp downstream from the precursor) into the mir-1829 family(null) background at a known safe harbor locus (the position of MosSCI locus ttTi5605 in chromosome II). Not only did we observe expression of mir-1829b-5p in this rescue strain, but expression was increased nearly eightfold compared with wild type (Fig. 3C). Northern blot confirmed that the knock-in locus generated a distinct mir-1829b-5p small RNA species (Supplemental Fig. S3); because germline specificity of this species has not been tested, broader tissue expression or reduced interference from host gene transcription could underlie the higher expression levels (see also the Discussion). Overall, the observed expression of the minimal transcriptional unit demonstrates that the 510 bp fragment contains all the required regulatory elements necessary for mir-1829b expression.

To further test the model that the mir-1829 family is a Pol III transcript, we depleted Pol III using the AID system. To do this, we tagged the Pol III catalytic subunit RPC-1 with a 3xFLAG::AID tag using CRISPR in a strain that expresses TIR1 ubiquitously, resulting in efficient RPC-1 depletion upon auxin treatment (Supplemental Fig. S4). We assessed expression of mir-1829b-5p, mir-1829c-5p, and their respective host gene transcripts after 24 h of auxin treatment using RT-qPCR. Depletion of Pol III by auxin treatment of the AID-tagged RPC-1 strain did not impact levels of the Pol II transcribed host gene transcripts but strongly abrogated expression of mir-1829 family members (Fig. 3D). In contrast, mir-1829b/c levels were unaffected in RPC-1 degron strains lacking auxin treatment, signifying that tagging RPC-1 alone does not disrupt its function (Fig. 3D). Furthermore, the significant depletion of mir-1829b/c was not due to off-target effects of auxin treatment or TIR1 expression, as observed by the lack of miRNA depletion in strains that do not contain tagged RPC-1 (Fig. 3D, “TIR1 + auxin”). Overall, depletion of RPC-1 along with the mutation of box A/B motifs supports the model that the mir-1829 family is transcribed by Pol III.

To determine whether Pol II also plays a role in the transcription of the mir-1829 family, we deleted the promoter regions of the host genes with the rationale that this would remove Pol II occupancy from these loci. We deleted 1.1 and 1.7 kb upstream of the start codon for F20D1.3 and piki-1, respectively (Fig. 3E). We observed strong reduction of host gene expression levels in their respective mutants via RT-qPCR, suggesting that we effectively removed Pol II occupancy (Fig. 3F). In the piki-1 promoter deletion mutant, there was no change in mir-1829c-5p levels. In the F20D1.3 promoter deletion mutant, expression of mir-1829b-5p was slightly increased (Fig. 3F). We hypothesize that there may be slight competition between Pol II and Pol III for the F20D1.3/mir-1829b locus, because deletion of the F20D1.3 promoter region (Fig. 3F) or ectopically expressing mir-1829b from another locus (Fig. 3C) results in increased expression of mir-1829b-5p. Together, these results validate the model that the mir-1829 family is primarily transcribed by Pol III with no detectable contribution from Pol II-driven host gene transcription.

Having demonstrated the dependence of mir-1829b on Pol III and its promoter elements, we returned to the mir-1829b locus to assess the role of additional cis elements. For these experiments, variants of the mir-1829b gene were knocked back in to its genomic locus in the context of the mir-1829 family(null) background; this approach facilitated genotyping for isolation of very small sequence changes. All mutations were designed to preserve the predicted secondary structure of pri-mir-1829b. First, we tested whether 4Ts present upstream of the precursor suppress productive mir-1829b transcription by causing early termination. This is not the case, because mutating this run of Ts decreased mir-1829b expression (Fig. 3G, upstream UUUU scramble). Next, we tested whether the regions flanking the precursor in the primary transcript are important for mir-1829b biogenesis. This was true of the 33 nt 5′-flanking sequence whose mutation reduced mir-1829b level but not of the 3 nt 3′ flank whose mutation had an insignificant effect (Fig. 3G, 5′ and 3′ flank scramble). Finally, we tested the importance of the first run of Ts after the precursor; mutating these five residues reduced mir-1829b expression but did not abolish it (Fig. 3G, terminator scramble). This suggests that these 5Ts play a role in biogenesis, likely by promoting Pol III termination, and the intact downstream 4T tract may support the residual mir-1829b expression. The decreased expression in this mutant may suggest that a very short flank 3′ of the precursor promotes biogenesis.

mir-1829 family biogenesis requires Dicer but not DRH-1, RDE-4, or NSUN-2

Having demonstrated that the mir-1829 family is Pol III transcribed and MP-independent, we sought to further elucidate its biogenesis. To do so, we used the AID system to determine the dependence of mir-1829 biogenesis on Dicer (DCR-1). DCR-1 was tagged with 3xFLAG-AID at its endogenous locus and crossed into a background expressing TIR1 ubiquitously or only in the germline, resulting in dramatic or moderate depletion of overall DCR-1, as expected (Supplemental Fig. S4). We first performed Northern blotting to examine the abundance of the mature miRNA species and the accumulation of the precursor. As a control, we probed for the canonical miRNA mir-1, which is primarily expressed in the pharynx and muscle (Simon et al. 2008; Brosnan et al. 2021; Gutiérrez-Pérez et al. 2021; Xu et al. 2024). We observed reduction of mature mir-1 and accumulation of its precursor in the samples where Dicer is depleted ubiquitously but not in the case of germline-only Dicer depletion (Fig. 4A). To achieve maximal signal for mir-1829b/c, low stringency conditions were used that do not distinguish between mir-1829b-5p and mir-1829c-5p when using a mir-1829b-5p probe, as demonstrated by detection of pure synthetic miRNA species corresponding to both sequences (Fig. 4A, right panels; see the Materials and Methods for low stringency conditions). We observed decreased mir-1829b/c expression in all the Dicer-depleted samples, including those in which Dicer is depleted only in the germline, consistent with strong germline enrichment of the mir-1829 family (Fig. 4A). Despite low stringency probing conditions, the mir-1829 family is still near the limit of detection (Fig. 4A). This low signal may contribute to our failure to detect accumulation of the precursor species of mir-1829b/c as expected upon Dicer depletion.

Figure 4.

The mir-1829 family is Dicer-dependent. (A–C) All samples were exposed to auxin for 24 h starting at the L4 stage and then harvested as adults at 20°C. (A) Northern blot using low stringency conditions (see the Materials and Methods for details of low stringency conditions). The right panels show lanes loaded with synthetic RNA oligos, demonstrating recognition of both mir-1829b-5p and mir-1829c-5p with the mir-1829b-5p probe under these conditions. (B) miRNA qPCR was normalized to a small RNA control (sn2429) and then further normalized to levels in a control that only carries the Dicer AID tag (UY212). Mean and SD are shown. Two-way ANOVA; (****) P < 0.0001 (C) Small RNA-seq. MA plot showing abundance in control (DCR-1::AID tag alone; UY212) on the X-axis and log₂(fold change) in DCR-1::AID; ubiquitous TIR1 (MCJ387) compared with UY212 on the Y-axis. miRNAs showing significant sensitivity to Dicer depletion are shown in blue–green, and the mir-1829 family is highlighted with red open circles. Three biological replicates of each genotype were analyzed using DESeq2.

Due to the limited detection of the mir-1829 family by Northern blotting, we further assessed miRNA levels using RT-qPCR. We observed significant reduction of both mir-1829b-5p and mir-1829c-5p in samples in which Dicer is ubiquitously depleted (Fig. 4B). Furthermore, the level of reduction in mir-1829b-5p in germline-only Dicer depletion samples nearly mirrors the level seen in the ubiquitously depleted samples (Fig. 4B), further supporting germline-restricted expression of mir-1829b-5p. We observed milder depletion of mir-1829c-5p in the germline Dicer depletion samples compared with depletion in the whole worm (Fig. 4B). This result further supports that mir-1829c is expressed in the both the germline and the soma as noted above (Supplemental Fig. S2B).

To further assess Dicer dependence, we performed sRNA-seq of samples in which Dicer is depleted ubiquitously or in the germline (Supplemental Tables S6, S7). Overall, almost all miRNAs appear to be Dicer-dependent, and many of these are statistically significant after spike-in normalization (P_adj < 0.05) (Fig. 4C; Supplemental Table S6). The mir-1829 family behaves similarly to canonical miRNAs in that it is sensitive to Dicer depletion (Fig. 4C, red circles). Consistent with all results discussed above, we also observed depletion of the mir-1829 family members when Dicer was depleted in the germline alone, along with depletion of a small set of germline-enriched canonical miRNAs (Supplemental Fig. S5; Supplemental Table S7).

To process long dsRNA substrates, C. elegans Dicer works in concert with two cofactors: double-stranded RNA (dsRNA) binding protein RDE-4 and RIG-I-like receptor DRH-1 (Tabara et al. 2002; Duchaine et al. 2006; Lu et al. 2009; Consalvo et al. 2024). Because the mir-1829 family is noncanonically processed but requires Dicer, we tested the role of RDE-4 and DRH-1 in mir-1829 biogenesis using knockout strains. From RT-qPCR experiments, we observed no change in mir-1829b/c in the rde-4 and drh-1 mutant strains compared with wild type (Supplemental Fig. S6A). Thus, Dicer functions in mir-1829 maturation independent of RDE-4 and DRH-1.

We also investigated the role of nsun-2 and multiple cellular nucleases in mir-1829 biogenesis. Vault RNAs can be further processed by Dicer to produce small RNA fragments (svRNAs) that function in RNAi when loaded into Ago (Persson et al. 2009). In cell lines, NSun2, an RNA methyltransferase that modifies cytosine, methylates vault RNA to prime them for dicing (Hussain et al. 2013). To determine whether the C. elegans homolog NSUN-2 plays a role in mir-1829 biogenesis, we introduced a 3xFLAG-AID tag to the C terminus of NSUN-2 by CRISPR. After 24 h auxin treatment in a ubiquitous TIR1 background, although NSUN-2 levels decreased ∼90% (Supplemental Fig. S4), mir-1829b/c levels remained unchanged (Supplemental Fig. S6B). Thus, nsun-2 does not appear to play a role in mir-1829 family biogenesis. To determine the role of exonucleases in mir-1829 family biogenesis, we reanalyzed data from a study in which most core cellular nucleases were knocked down by RNAi, followed by sRNA-seq (Vieux et al. 2021). None of these RNAi conditions altered mir-1829 family member abundance, suggesting that these exonucleases are also dispensable for mir-1829 family biogenesis (Supplemental Fig. S6C).

The 3′ end of the mir-1829 family precursor is likely generated in the nucleus

Although they differ in biogenesis of their precursor, mirtrons are similar to the mir-1829 class of miRNAs in their MP independence and Dicer dependence. Mirtron biogenesis is counteracted by untemplated 3′ end uridylation of the precursor species, which destabilizes these intermediates (Bortolamiol-Becet et al. 2015; Reimão-Pinto et al. 2015). Activity of terminal nucleotidyl transferases on miRNA precursor substrates can be detected through sequencing of mature miRNA populations because the untemplated nucleotide additions (tails) can persist after Dicer-mediated maturation. In particular, tailing that is strongly skewed toward 3p-derived miRNA strands is likely a signature of an enzyme that acts on miRNA precursor substrates; such activity would not result in tailing of 5p-derived miRNA strands whose 3′ ends are not available for modification until after Dicer-mediated precursor cleavage. We previously observed that miRNA A-tailing in C. elegans is strongly skewed toward 3p-derived miRNAs, suggesting that an adenyltransferase acts on miRNA precursors (Vieux et al. 2021). To determine whether mir-1829 family biogenesis is regulated by tailing of the precursor similar to mirtrons, we examined the sequences of the mature strands for evidence of tailing (which may have been deposited on either precursor or mature miRNA species). We observed various tailed species of mir-1829 family members with a clear enrichment of adenylation on the 3p-derived strands, suggestive of adenylation of the miRNA precursors (Fig. 5A). Sequencing reads of mir-1829b/c-3p (which share an identical reference sequence and displayed the highest level of adenylation) show 3′ trimmed isoforms and reveal that adenylation occurs on an isoform that is 1 nt shorter than the reference (Fig. 5B). This adenylation is not predicted to change the structure of the miRNA precursor from the reference secondary structure; either an unmodified precursor or a precursor in which the terminal base is substituted with A is predicted to have 1 nt unpaired at the 5′ end and 3 nt unpaired at the 3′ end (see Fig. 1C). Although the trimmed 3p isoforms suggest that an exonuclease modifies or matures the 3′ end of the precursor, no exonuclease has yet been identified to play a role in biogenesis (Supplemental Fig. S6C).

Figure 5.

The mir-1829 family primary transcripts are likely cleaved in the nucleus. (A) Percentage of untemplated 3′ nucleotide additions (tailing) of the mir-1829 family members in wild type. (B) Raw read counts across four biological replicates showing the top six isoforms of mir-1829b/c-3p. (C, left) Percentage of A-tailing in wild type and gldr-2(null). (Right) The abundance of each mature species is shown. (A,C) All mir-1829 family member strands with more than one RPM average abundance are shown. Raw data were reanalyzed from Vieux et al. (2021). (D) Representative images of 3xFLAG::AID::GFP::GLDR-2. (E) Model for a new miRNA biogenesis pathway of the mir-1829 family in contrast to canonical miRNA biogenesis.

The 5p-derived strands also displayed tailed isoforms (which must arise from modification of mature miRNAs), including uridylation of mir-1829a-5p and mir-1829c-5p, cytidylation of mir-4812-5p, and guanylation of mir-1829b-5p (Fig. 5A). Terminal guanylation and cytidylation of miRNAs are rarely observed in animals; uridylation is the most frequently observed modification of C. elegans mature miRNAs (Vieux et al. 2021). A functional role for tailing of mature miRNAs has not yet been established in C. elegans, and because the modifications of 5p strands differ across the family members, we did not explore a potential role in biogenesis.

Because GLDR-2 is responsible for A-tailing of miRNA precursors in C. elegans, we reanalyzed published data to examine whether GLDR-2 is also responsible for tailing of the mir-1829 family (Vieux et al. 2021). We observed loss of A-tailing on 3p arms of mir-1829 family members in gldr-2(null) mutants (Fig. 5B,C; Vieux et al. 2021). To determine whether A-tailing is involved in mir-1829 biogenesis, we evaluated mature miRNA levels in the presence and absence of GLDR-2. We observed no change in the levels of expression, suggesting that although mir-1829 precursors can be adenylated by GLDR-2, A-tailing is dispensable for mir-1829 biogenesis (Fig. 5C; Vieux et al. 2021).

The A-tailing of the 3p arms of the mir-1829 family suggests that a precursor-like intermediate (in which the 3′ end of the 3p arm is available for tailing) precedes Dicer-mediated maturation of the mature duplex despite our inability to detect this intermediate on Northern blots. We reasoned that localization of GLDR-2 may further inform the order of events of mir-1829 biogenesis. A GFP tag fused to GLDR-2 at its endogenous locus reveals that GLDR-2 specifically localizes to the nucleus (Fig. 5D). This suggests that the cleavage events that give rise to the miRNA precursor (or at least maturation of the precursor 3′ end) occur in the nucleus prior to Dicer-mediated cleavage in the cytoplasm (Fig. 5E). Thus, we have delineated a new miRNA biogenesis pathway, as exemplified by the mir-1829 family, that involves the transcription of an atypical primary miRNA transcript by Pol III and initial MP-independent maturation steps in the nucleus, followed by Dicer-mediated cleavage in the cytoplasm (Fig. 5E).

Noncanonical miRNA biogenesis extends beyond the mir-1829 family

Having defined atypical features of the mir-1829 family's biogenesis, we sought to extend these findings by identifying additional noncanonical miRNAs. First, we re-examined data sets in which MP is inactivated (Dexheimer et al. 2020; Vieux et al. 2021). For these analyses, we used uncurated miRBase annotations (Kozomara et al. 2019) rather than MirGeneDB, which stringently filters for features of canonical miRNAs and evolutionary conservation (Fromm et al. 2020). Among MP-insensitive miRNAs, we observed many annotated mirtrons, as expected (Fig. 6A; Supplemental Tables S10, S11; Ruby et al. 2007; Chung et al. 2011; Jan et al. 2011). In both MP inactivation data sets, we also observed that mir-8196a-3p, mir-8196b-3p, and mir-8199-5p were not depleted (Fig. 6A; Supplemental Tables S10, S11; Dexheimer et al. 2020; Vieux et al. 2021). In contrast, all of these were decreased upon Dicer depletion over the same time scale (24 h), demonstrating that a long half-life for these small RNAs is not responsible for their perdurance in the setting of MP inactivation (Fig. 6C; Supplemental Table S10). Based on its sequence and loading into ERGO-1 (Seroussi et al. 2023), mir-8199-5p appears to be a misannotated 26G class endo-siRNA. mir-8196a/b-3p are more intriguing because they are bound in part by ALG-1 and ALG-2 (Seroussi et al. 2023) and therefore may function as noncanonical miRNAs. mir-8196b lies 65 bp downstream from a recently annotated 21U piRNA on the X chromosome (Supplemental Fig. S5A; Seroussi et al. 2023). Similarly, mir-8196a lies 71 bp downstream from an unannotated 21U piRNA, also on the X chromosome; the upstream RNA meets the requirements of a 21U based on its sequence, loading in PRG-1, and depletion in prg-1^−/− (Fig. 6B). This consistent proximity to a 21U locus raises the possibility that this relative position may contribute to mir-8196a/b transcription or biogenesis. Consistent with this, mir-8196a/b are germline-enriched, as evidenced by their significant reduction in samples in which Dicer is depleted only in the germline (Supplemental Table S12).

Figure 6.

The noncanonical miRNA biogenesis pathway extends beyond the mir-1829 family. (A,C,D) Small RNA-seq results in samples depleted for 24 h of functional PASH-1 (A), Dicer (DCR-1) (C), or RPC-1 (D). MA plots show log₂(fold change) in experimental versus control on the Y-axis, with normalized control abundance on the X-axis. Controls are wild type (N2) (A), auxin-treated DCR-1::AID tag alone (UY212) (C), and auxin-treated TIR1 (MLC1040) (D). All miRBase annotations are plotted, and relevant noncanonical small RNAs are highlighted. (B) Genome browser tracks showing the proximity of mir-8196a to an upstream unannotated 21U RNA. 5′-independent small RNA cloning following polyphosphatase treatment also shows 22G RNAs generated from the locus. mir-8196a but not the 21U shows Dicer dependence. (E) Genome browser track showing RPC-1 (a Pol III subunit) ChIP at the mir-4937 locus.

Multiple Dicer-independent small RNAs were also detected (Fig. 6C; Supplemental Tables S8, S10, S11). These include misannotated 21U piRNAs that persist in mirBase that were recently noted (Seroussi et al. 2023). Thus, these small RNAs act as positive controls for Dicer independence, similarly to mirtrons for MP independence. We also found that mir-5549-3p is an additional misannotated 21U RNA encoded on chromosome III; it is 21 nt in length with U in the 5′ position, bound by PRG-1, and depleted in prg-1^−/− (Supplemental Fig. S5B; Seroussi et al. 2023). Finally, we identified mir-4937 as a Dicer-independent small RNA (see more below).

We also sought to further define Pol III-dependent miRNAs. To this end, we performed sRNA-seq on samples in which RPC-1 was depleted using the AID system and overlapped these data with published Pol III ChIP-seq data to identify bona fide Pol III transcribed miRNAs (Fig. 6D; Supplemental Tables S9–S11). The strongest candidate is again mir-4937, which is RPC-1-dependent and has prominent Pol III ChIP peaks (Fig. 6E). Consistent with these regulatory elements and Pol III dependence, a previous study noted homology between mir-4937 and a tRNA locus, mostly within putative box A and box B motifs (Roberts et al. 2013). Although mir-4937 did not meet our stringent criteria for defining novel MP-independent small RNAs, its depletion in pash-1(ts) is marginal and not called as significant (Supplemental Table S10), and its predicted secondary structure suggests that its primary transcript would make a poor MP substrate (Supplemental Fig. S7C). Thus, mir-4937 is likely a Pol III transcribed MP- and Dicer-independent small RNA. mir-4937 was not robustly immunoprecipitated with any of the C. elegans Argonautes according to Seroussi et al. (2023) (all average IP < 0.2 reads per million); thus, its functional class remains unclear. Nonetheless, previous studies identified it as a small RNA that increases during C. elegans aging (De Lencastre et al. 2010).

Human miR-4521 biogenesis is noncanonical and Pol III-dependent

To determine whether miRNAs can derive from independent Pol III transcripts outside of C. elegans, we also examined ChIP data sets of Pol III subunits in three human cell lines: HeLa (Oler et al. 2010), K562 (Dunham et al. 2012), and HEK293 (Gerber et al. 2020). Manual curation of HEK293 RPC62 ChIP confirmed that miR-4521 and miR-4638 had strong Pol III ChIP signals (Fig. 7A). Whereas miR-4638 is too lowly expressed to assess its biogenesis requirements, miR-4521 shows evidence of Drosha and Dicer independence in HEK293 cells (Supplemental Fig. S7D; Rybak-Wolf et al. 2014).

Figure 7.

Human small RNA miR-4521 is Pol III-dependent and DGCR8- and Dicer-independent. (A) Genome browser tracks showing RPC62 (Pol III) ChIP in HEK293 cells at the mir-4638 and mir-4521 loci. (B) Small RNA-seq results from doxycycline-inducible Cas9 (iCas9) RKO cells treated with Pol III inhibitor ML-60218 for 24 h compared with vehicle-treated (DMSO) control. (C,D) Small RNA-seq results from cells grown for 4 days after induction of guide RNAs targeting DGCR8 (C) or Dicer (D), compared with control iCas9 RKO cells. miRNA reads were normalized to spike-ins. (B–D) miRNAs with baseMean ≥ 1 in control samples are shown.

To further assess the noncanonical transcription and biogenesis of miR-4521, we examined a colon carcinoma cell line (RKO) that expresses high amounts of miR-4521 (but not miR-4638, which was not further examined). A variant of the cell line with doxycycline-inducible Cas9 (iCas9) was used for all experiments (de Almeida et al. 2021). To determine whether miR-4521 is dependent on Pol III, RKO cells were treated with Pol III inhibitor ML-60218, followed by sRNA-seq. miR-4521 was significantly depleted after 24 h of Pol III inhibition (Fig. 7B). A few other miRNAs also showed Pol III dependence; these include both strands of miR-1277 and guide strands miR-33a-5p, miR-153-3p, and miR-32-5p (Fig. 7B; Supplemental Table S13). Like miR-4521, the miR-33a-5p locus also exhibits Pol III ChIP (Ozsolak et al. 2008). Next, we examined DGCR8 and Dicer dependence by introducing dual-guide RNAs (dgRNAs) targeting these loci (de Almeida et al. 2021). Cas9 expression was induced for 4 days, followed by harvest and Western blot (which confirmed knockout of DGCR8 or Dicer) (Supplemental Fig. S7E) and sRNA-seq. Knockout of DGCR8 or Dicer resulted in strong or moderate global depletion of miRNAs, respectively, when normalized to spike-ins (Fig. 7C,D; Supplemental Tables S14, S15). Notably, miR-4521 expression was not affected by either DGCR8 or Dicer knockout (Fig. 7C,D). Thus, evolutionarily young small RNAs may also arise from Pol III transcription in humans.

Discussion

In this study, we identified a new noncanonical miRNA biogenesis pathway that circumvents cleavage by MP but relies on Dicer processing (Figs. 1, 4). Using a variety of techniques, we have determined that the mir-1829 family in C. elegans is derived from independent transcripts that are transcribed by Pol III (Figs. 2, 3). Although these miRNA loci reside in the long introns of three Pol II transcribed host genes, there is no detectable contribution of Pol II in mir-1829 transcription (Fig. 3F). Many classes of Pol III transcripts (including snaR-A, tRNAs, and SINEs) generate miRNAs via cleavage of a larger functional transcript (Babiarz et al. 2008; Maute et al. 2013; Martinez et al. 2017; Stribling et al. 2021). Unlike these transcripts, the mir-1829 family exists exclusively as independent miRNA transcripts. The small RNAs generated by these loci are the only species detectable by Northern blot, demonstrating that the mir-1829 family members are the primary product of these loci and thus differentiating this unique biogenesis pathway from previously identified Pol III transcribed miRNAs.

Given that the Pol III transcribed mir-1829 family resides in the long introns of its Pol II transcribed host genes, we speculate that there may be competition between Pol II and Pol III for the F20D1.3/mir-1829b locus. When the mir-1829b locus is expressed from an ectopic location in the mir-1829 family(null) background, mir-1829b-5p expression dramatically increases in comparison with levels in wild type (Fig. 3C). Additionally, inhibiting F20D1.3 transcription by removing its promoter region—and thereby removing Pol II occupancy—results in a modest increase of mir-1829b-5p expression (Fig. 3F). These results suggest that there is transcription interference of Pol III by Pol II at this locus. Previous studies have shown that Pol II can repress the transcription of tRNA genes (Lukoszek et al. 2013; Gerber et al. 2020). Because the mir-1829 family is considered to be under the same class of Pol III transcripts as tRNAs (i.e., type II), we hypothesize that Pol II has an inhibitory effect on mir-1829b transcription. Additionally, it is plausible that elongating Pol II physically occludes assembly of Pol III transcriptional machinery, limiting the transcriptional output of the mir-1829 family (Corbin and Maniatis 1989; Martens et al. 2004; Bird et al. 2006; Petruk et al. 2006). However, the relationship between Pol II and Pol III is complex, as emerging reports suggest that Pol II can promote Pol III transcription and vice versa (Listerman et al. 2007; Gerber et al. 2020; Jiang et al. 2022; Rajendra et al. 2024). Pol II occupancy may also play a role in the differential spatial expression of the mir-1829 family and its host genes. This could be driven by chromatin remodeling during germ cell maturation, allowing for transcription of the miRNA but not the host gene in the distal germline (Yague-Sanz et al. 2023).

Although we have demonstrated that the mir-1829 family miRNAs require Pol III and Dicer for their biogenesis, additional required factors remain unidentified. We propose that a nuclear exonuclease may mature the 3′ end of the precursor by trimming this flank. Mutation of the proximal Pol III terminator reduces mir-1829b-5p production, suggesting the necessity of a very short flank 3′ of the precursor (Fig. 3G); length heterogeneity at the 3′ end of 3p strands of mir-1829 further suggests exonucleolytic trimming of the short trailing flank. The 3p strands also display some adenylated isoforms that are dependent on the enzyme GLDR-2. Our model (Fig. 5E) is that the 3′ end of the miRNA precursor is matured and adenylated in the nucleus because (1) GLDR-2's substrates show a strong 3p bias genome-wide beyond just the mir-1829 family, suggesting that its substrates are miRNA precursors (Vieux et al. 2021), and (2) GLDR-2 is localized to the nucleus (Fig. 5D). Other interpretations in which GLDR-2 acts on the mature miRNA species are also possible but less parsimonious because they would require Dicer action in the nucleus or GLDR-2 activity in the cytoplasm. Importantly, GLDR-2 only modifies a small fraction of mir-1829-3p reads and is not required for biogenesis because its knockout does not perturb mir-1829 family abundance (Fig. 5C). If cleavage that generates the 5′ end of the precursor also occurs in the nucleus, then the precursor would be indistinguishable from a canonical miRNA precursor, enabling nuclear export. Spatiotemporal control of the unidentified nuclease(s) in mir-1829 biogenesis could contribute to the localized enrichment of mir-1829 in the distal germline.

Building on previous work that characterized the expression patterns and binding complements of C. elegans Argonaute proteins, we show that the mir-1829 family's expression is strongly germline-enriched. Multiple lines of evidence suggest that mir-1829b-5p is very highly germline-enriched, whereas a larger portion of mir-1829c-5p is expressed in the soma. An unusually large portion of the mir-1829 family is loaded in the newly characterized germline-specific Argonaute ALG-5. Although two roles of ALG-5 in germline biology have been demonstrated, the targets of ALG-5 in these processes have yet to be discovered. Here we demonstrated that ALG-5 represses a tethered RNA (Fig. 2); thus, ALG-5 likely acts redundantly with ALG-1/2 on a shared set of target RNAs via their shared small RNA cargo. Given that ALG-5 functionally represses a tethered mRNA in the distal germline of adult hermaphrodites, which coincides with peak mir-1829 expression, stability of the mir-1829 family may be maintained within the spatial domain of ALG-5 activity. Furthermore, the rapid clearance of mir-1829 in Dicer knockdown conditions suggests rapid turnover of mir-1829 miRNAs that may determine or reinforce their spatial expression domain.

What is the contribution of the mir-1829 family to animal physiology? Despite their enrichment in the germline, this miRNA family is not required for viability, germline function, or the robustness of the mitosis-to-meiosis transition. Although these miRNAs are C. elegans-specific, they are preserved in the numerous wild isolates that have been collected around the world, suggesting that they may provide a selective advantage (Crombie et al. 2024). An interesting aspect of their future study will be to determine whether the 5p or 3p strands are the biologically important species. Although the 5p strands are more abundant, the 3p strands have greater sequence conservation, including across the functional seed region.

We expanded our search for additional noncanonical and Pol III-dependent miRNAs by comparing data sets in which MP, Dicer, or Pol III function is inactivated. In C. elegans, we identified (1) misannotated and unannotated 26G and 21U RNAs, (2) a pair of identical MP-independent small RNAs (mir-8196a/b) whose biogenesis may rely on their close proximity to a 21U RNA, and (3) a Pol III-dependent small RNA (mir-4937) that does not require either MP or Dicer for its biogenesis. In human cells, we identified multiple candidates for Pol III transcribed small RNAs, including miR-32-5p, miR-33a-5p, miR-153-3p, miR-1277-3p, mir-1277-5p, and miR-4521. miR-4521 is also DGCR8- and Dicer-independent. Whether miR-4521 functions as a miRNA is currently debated because it has shown activity in luciferase assays in some contexts but also displays low Ago association in one study (Feng et al. 2019; Sun et al. 2021; Xing et al. 2021; Kuthethur et al. 2023; Orang et al. 2025). Notably, all of the newly identified noncanonical small RNAs are recently evolved. Because miRNA loci are predicted to gradually evolve toward perfect biogenesis substrates (Liu et al. 2008; Berezikov 2011), these small RNAs and the mir-1829 family may be substrates for further evolution to canonical miRNAs.

De novo miRNAs are likely to target many mRNAs, having potentially deleterious effects and preventing their retention in the genome. Chen and Rajewsky (2007) suggested a model for miRNA gene evolution in which miRNAs first gain a low level of expression, possibly in a tissue-restricted manner. Then, deleterious target sites can be purged, and finally, miRNA level and domain of expression can increase, providing a selective advantage by repressing the remaining target mRNAs. The mir-1829 family and other identified noncanonical small RNAs like mir-8196a/b and miR-4521 may be in the early stages of this trajectory, given their low level and tissue-restricted expression. Overall, these miRNAs fit multiple theoretical predictions of the characteristics of young de novo miRNA loci, thus representing a snapshot along the potential evolutionary trajectory to conserved canonical miRNA genes.

Materials and methods

C. elegans growth and maintenance

C. elegans were maintained on NGM plates seeded with OP50 at 20°C except where noted otherwise (Stiernagle 2006). For all experiments, embryos were harvested from gravid adults by hypochlorite treatment and washed thoroughly with M9 before hatching overnight in M9 supplemented with 1 mM cholesterol at 20°C. Synchronized L1s were grown at 20°C to the appropriate stage as indicated. Strains and alleles used in this study are listed in Supplemental Tables S1 and S2.

For experiments using the auxin-inducible degron, synchronized L1s were grown for 48 h on NGM plates at 20°C and then washed onto 4 mM auxin NGM plates at the L4 stage and grown for an additional 24 h at 20°C before harvesting.

CRISPR/Cas9-mediated genome editing

All strains generated in this study were made by CRISPR-mediated editing as described previously (Yang et al. 2020). Deletions of the mir-1829 family members and ALG-5 tethering reporter strains were generated by simultaneous CRISPR of two loci, in which one or both repair templates were integrated as indicated in Supplemental Table S2.

Brood size and embryonic viability assays

Brood size assays were conducted at 25°C as described previously (Kotagama et al. 2024).

For glp-1(ar202) experiments, all strains were maintained at 15°C. A synchronized egg lay was performed for 3 h to isolate age-matched cohorts; these embryos were allowed to hatch at 15°C and then upshifted to 20°C 24 h after the synchronized egg lay. The worms were then kept at 20°C, and single worms were followed to begin brood counts as above.

Immunoprecipitation

Synchronized L1s were grown for ∼ 65 h (to first day of adulthood) at 20°C prior to pellet collection. Frozen pellets were resuspended with a volume of 2× lysis buffer corresponding to the volume of the pellet. Lysis buffer (1×; 30 mM HEPES at pH 7.4, 50 mM KCl, 0.1% Triton X-100, 2 mM MgCl₂, 10% glycerol, 2 mM DTT, 80 U/mL RNase inhibitor, 1× cOmplete mini protease inhibitor cocktail [Roche]) was added to bring the volume of all the samples to 1 mL. Samples were subjected to four to six rounds of sonication (10 cycles of 30 sec on and 30 sec off per round) in a Bioruptor Pico and then clarified by centrifugation at >15,000g for 20 min at 4°C. Pierce reducing agent-compatible BCA protein assay was used to quantify protein prior to IP. IP inputs were 1 mL of lysate at 5 mg/mL for ALG-5 IP or 2 mg/mL for all other samples. Inputs were combined with 50 µL of washed anti-FLAG M2 magnetic bead slurry (Millipore Sigma) and incubated with gentle agitation overnight at 4°C. After three washes (500 µL of 30 mM HEPES at pH 7.4, 100 mM KCl, 0.1% Triton X-100, 2 mM MgCl₂, 10% glycerol, 2 mM DTT, 80 U/mL RNase inhibitor, 1× cOmplete mini protease inhibitor), beads were resuspended in 1 mL of 1× lysis buffer. Input, supernatant, and IP samples were saved in 50 µL aliquots, which were either run on 4%–20% TGX gels (Bio-Rad) or used for RNA extraction (see below).

RNA isolation

Frozen samples were resuspended with four pellet volumes of TRIzol reagent (Fisher Scientific) and then subjected to three freeze–thaw cycles on dry ice, followed by vortexing for 15 min at room temperature. RNA isolation was then performed according to the TRIzol manufacturer's instructions.

For IP-qPCR samples, extra care was taken to recover an equal proportion of RNA from each sample from IP workflows (input, supernatant, and IP). To this end, 82 µL of dH₂O and 2.5 µL of 10 mg/mL yeast RNA (Invitrogen) were added to 50 µL aliquots of input, supernatant, or IP. Four-hundred microliters of TRIzol LS (Thermo Fisher Scientific) was added to each sample, followed by 15 min of vortexing at room temperature. Chloroform (106.25 µL) was then added to each sample prior to centrifugation at >15,000g for 10 min. A consistent volume (200 µL) of the aqueous phase was transferred to a new tube and mixed with an equal volume of phenol:chloroform:isoamyl alcohol (25:24:1 at pH 4.5). After centrifugation at >15,000g for 10 min, 100 µL of the aqueous phase was transferred to a new tube, and RNA was precipitated with 1 vol (100 µL) of isopropanol overnight at −20°C. Pellets were washed with 75% ethanol and resuspended in 28 µL of dH₂O.

Rapid amplification of cDNA ends (RACE)

5′ ends of primary mir-1829 family transcripts were determined using 5′/3′ RACE, second-generation kit (Sigma-Aldrich). For cDNA synthesis, 1 µL of ∼1 µg/µL RNA sample was used. The amplification steps were carried out using Q5 high-fidelity PCR kit (NEB). For a list of the primers used, see Supplemental Table S2. Purified PCR products were cloned into pCR-Blunt II-TOPO vector using a Zero Blunt TOPO PCR cloning kit (Invitrogen).

TaqMan miRNA qPCR

For each sample, 1.66 µL of total RNA at 200 ng/µL was used per 5 µL of reverse transcription reaction using the TaqMan microRNA reverse transcription kit (Thermo Fisher). Completed RT reactions were diluted 1:4 with dH₂O. Next, 1.33 µL of diluted product was used in a 5 µL qPCR reaction with TaqMan miRNA probes and TaqMan Universal master mix II with UNG (Thermo Fisher) and run in triplicate on an Applied Biosystems QuantStudio Pro 6.

Reverse transcription quantitative polymerase chain reaction (RT-qPCR) for mRNAs

RT-qPCR of mRNA was performed using KAPA SYBR Fast one-step qRT-PCR master mix (2X) kit (Fisher Scientific) and gene-specific primers on an Applied Biosystems QuantStudio Pro 6. The primer sequences are listed in Supplemental Table S2. For each reaction, 1.25 µL of total RNA at 10 ng/µL was used in 5 µL qPCR reactions, and technical triplicates were performed.

Immunoblotting

C. elegans protein samples were prepared and quantified as described previously (Kotagama et al. 2024). Samples were run on Mini-Protean TGX gels (Bio-Rad). All blots were blocked in TBST containing 5% BSA (for DCR-1::AID experiments) or 5% milk for 10 min to 1 h at room temperature. Membranes were probed overnight at 4°C with primary antibody, followed by F(ab′)2-goat antimouse IgG (H+L) secondary antibody and HRP (Thermo Fisher) diluted 1:5000. Primary antibodies used were anti-FLAG M2 antibody (Sigma) diluted 1:1000, anti-Dicer (Cell Signaling 3363) diluted 1:500, anti-DGCR8 (GenuIN Biotech 62316) diluted 1:500, and anti-α-Tubulin antibody DM1A (Abcam ab7291) diluted 1:5000 for AID experiments or AA4.3 (Developmental Studies Hybridoma Bank) diluted 1:5000 for IP experiments. ProtoGlow ECL (National Diagnostics) was used for signal detection of FLAG-tagged proteins, whereas Pierce ECL Western blotting substrate (Thermo Scientific) was used for signal detection of tubulin and human proteins. Signal was measured using an Amersham Imager 680 (GE) and quantified using Image Studio Lite. For reprobing, membranes were stripped for 45 min at 50°C with stripping buffer (2% SDS, 62.5 mM Tris HCl at pH 6.8, 0.8% β-mercaptoethanol).

Northern blot

Total RNA was separated on 15% urea gel (1.5 mm) and transferred to Amersham Hybond-NX membranes. Membranes were cross-linked using 0.16 M l-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC) (Sigma) in 0.13 M 1-methylimidazole (pH 8) for 30 min at 60°C as described previously (Pall et al. 2007). DNA oligos complementary to miRNAs of interest and U6 were 5′ end-labeled with ATP-γ-[³²P] unless noted otherwise. For MP depletion experiments, 25 µg of RNA was loaded, and high stringency conditions were used. High stringency conditions consisted of probing overnight in hybridization buffer (7% SDS, 0.2 M NaPO₄ at pH 7.2) at 42°C, followed by washing four times for 10 min at 42°C (twice with 2× SSPE + 0.1% SDS, once with 1× SSPE + 0.1% SDS, and once with 0.5× SSPE + 0.1% SDS) and then exposing blots. For Dicer depletion experiments, 10 µg of RNA was loaded, and low stringency conditions were used. Low stringency conditions consisted of probing overnight in ExpressHyb hybridization solution (Takara Bio USA) at 32°C, followed by washing once with 2× SSC + 0.1% SDS for 10 min at 32°C and then once with 1× SSC + 0.1% SDS for another 10 min at 32°C before exposing blots. In between each probe, membranes were stripped three times by incubating in a solution of boiling 0.04% SDS for 20 min at room temperature with shaking.

RNA structure predictions

The structures of the primary and precursor sequences of mir-1829 family members were predicted using the minimum free energy (MFE) structure on RNAfold (Gruber et al. 2008; Lorenz et al. 2011). The Vienna format output for the primary structure was loaded into RNAcanvas (Johnson and Simon 2023), and the predicted substructure of the precursor was applied when this region differed in the primary transcript structure prediction.

Small RNA-seq—C. elegans samples

Metadata of sequenced samples are shown in Supplemental Table S4. Library preparation was performed using the NEBNext small RNA library preparation kit for Illumina as described previously (Donnelly et al. 2022). Pooled samples were sequenced on a NextSeq 2000. Raw sequence data are accessible under GEO accession number GSE282009. Briefly, sequence analysis was performed on the National Institutes of Health High-Performance Computing Cluster. Libraries were trimmed using Cutadapt 4.4 (Martin 2011) and mapped using Bowtie2 2.5.1 with settings “–no-unal –end-to-end –sensitive” (Langmead and Salzberg 2012), and the resulting BAM files were sorted and indexed using SAMtools 1.19 (Danecek et al. 2021). Reads were assigned to miRNAs using HTSeq 2.0.4 with settings “‐‐mode union ‐‐nonunique fraction -a 0” (Anders et al. 2015). These raw read counts are shown in Supplemental Table S5. Analysis of miRNA expression using DESeq2 was performed similarly to that described previously (Kotagama et al. 2024), with the modification that size factors were calculated by dividing spike-in reads in each sample by the mean spike-in reads per sample. Published small RNA data sets were analyzed similarly, except that reads were trimmed and filtered for quality using TrimGalore 0.6.7 (https://zenodo.org/records/5127899) with default settings.

Small RNA-seq—cell culture samples

Small RNA libraries were prepped as described previously with minor modifications (see the spike-in sequence and concentrations outlined in Supplemental Table S2; Reichholf et al. 2019; Mandlbauer et al. 2024). Library preparations were pooled and then sequenced on NEBNext 2000.

Libraries were adaptor-trimmed and then demultiplexed with Cutadapt 5.0. UMIs were trimmed with UMI-tools 11.5 before mapping to the Homo_sapiens_NCBI_GRCh38 genome using Bowtie2 2.5.3. The resulting BAM files were sorted and indexed using SAMtools 1.21. Reads were assigned to miRNAs using HTSeq 2.04. Analysis of miRNA expression was performed with DESeq2 including spike-in size factors as above.

Fluorescence imaging

GLDR-2 adult worms raised at 15°C were mounted in 15 µL of M9 with 5 mM levamisole on a 2% agarose pad. ALG-5 tethering reporter strains were maintained at 20°C and mounted with M9 supplemented with 50 mM sodium azide. Slides were coverslipped and then imaged on a Nikon C2 confocal microscope using a 20× or 60× water immersion objective.

Pol III ChIP analysis

C. elegans Pol III ChIP peaks were accessed from ENCODE (accession file ENCFF924VCM) (Araya et al. 2014). For human ChIP data analysis, low-quality bases and adaptors were trimmed from raw sequence reads with Cutadapt v2.7 using -q 20 -a AGATCGGAAGAGC -minimum-length 25. Trimmed reads were aligned to the human GRCh38.p14 genome assembly using Bowtie2 v2.3.5 (Langmead and Salzberg 2012). Multimapping reads were removed using SAMtools v1.9 (Li et al. 2009) with the “view” subcommand and the additional argument “-q 20,” and duplicate reads were removed using the Picard v2.21.4 (http://broadinstitute.github.io/picard/index.html) “mark duplicates” tool. Peaks were called for each replicate individually using MACS v2.2.7.1 (Zhang et al. 2008). Pooled peaks were called by providing multiple BAMs of all the replicates to MACS2. miRNAs containing prominent Pol III peaks were called using BEDTools intersect with the miRBase annotation file for the respective species (Quinlan and Hall 2010).

Cell culture

iCas9-RKO (sex unspecified; American Type Culture Collection CRL-2577) cells were cultured in RPMI 1640 (Fisher Scientific) supplemented with 10% FBS (Takara Bio), 4 mM L-glutamine (Thermo Fisher Scientific), 1 mM sodium pyruvate (Sigma-Aldrich), and 100 U/mL penicillin/100 U/mL streptomycin (Sigma-Aldrich) (de Almeida et al. 2021). Lenti-293T lentiviral packaging cells (female; Clontech 632180) were cultured in Dulbecco's modified Eagle's medium (Sigma-Aldrich) supplemented with 10% FBS, 4 mM L-glutamine, 1 mM sodium pyruvate, and 100 U/mL penicillin/100 µg/mL streptomycin. Cell lines were maintained at 37°C with 5% CO₂ and routinely tested for mycoplasma contamination (InvivoGen).

Lentivirus production and infection

Semiconfluent Lenti-X cells were transfected with lentiviral plasmids pCMVR8.74 helper (Addgene 22036), pCMV-VSV-G (Addgene 8454), and Dicer or DGCR8-targeting dgRNAs (pLentiV1-dual-CRISPR-hU6-mU6-EF1as-mCherry-P2A-Puro) using polyethylenimine (PEI) transfection (molecular weight 25,000; Polysciences) (de Almeida et al. 2021). After 4 days, virus-containing supernatant was harvested and then concentrated with Lenti-X concentration (Takara Bio) according to the manufacturer's instructions.

iCas9-RKO cells were transduced with Dicer or DGCR8 targeting dgRNAs at ≥0.4 MOI. Cells containing Dicer or DGCR8 dgRNA were selected using 2 µg/mL puromycin (Thermo Fisher Scientific).

Polymerase III inhibitor and Dicer and DGCR8 knockout sample collection

Semiconfluent iCas9-RKO cells were treated with DMSO control or 50 µM polymerase III inhibitor (MilliporeSigma CAS 577784-91-9). At 24 h, cells were washed once with 1× PBS and then harvested in TRIzol.

Semiconfluent Dicer or DGCR8 dgRNA containing iCas9-RKO cells were induced for Cas9 expression using 500 ng/µL doxycycline (Sigma-Aldrich). After 4 days, cells were harvested for protein and RNA extraction.

For protein extraction, cell pellets were lysed on ice with NP40 lysis buffer (50 mM Tris-Cl at pH 7.4, 150 mM NaCl, 1% NP40, 1 mM EDTA, 1 mM EGTA) supplemented with cOmplete protease inhibitor cocktail (MilliporeSigma). After 30 min, lysates were centrifugated at 17,000g for 15 min. Supernatant was collected and mixed with one-quarter volume 4× Laemmli buffer (Bio-Rad) with 50 mM DTT (Bio-Rad). Protein lysates were boiled for 5 min at 95°C and stored at −20°C until Western blotting.

For RNA extraction, cell pellets were resuspended in TRIzol (Fisher Scientific) and then immediately frozen in dry ice and stored at −80°C until extraction. Once all samples were collected, RNA extraction was performed according to the manufacturer's instructions with the modification of supplementing isopropanol and 80% ethanol with 0.2 mM DTT. Finally, purified RNA was resuspended in water supplemented with 0.1 mM DTT.

Data availability

All primary deep sequencing data can be accessed at NCBI GEO under accession number GSE282009.

Competing interest statement

The authors declare no competing interests.