MicroRNAs (miRNAs) play an essential role in the post-transcriptional regulation of animal development and physiology. However, in vivo studies aimed at linking miRNA function to the biology of distinct cell types within complex tissues remain challenging, partly because in vivo miRNA-profiling methods lack cellular resolution. We report microRNome by methylation-dependent sequencing (mime-seq), an in vivo enzymatic small-RNA-tagging approach that enables high-throughput sequencing of tissue- and cell-type-specific miRNAs in animals. The method combines cell-type-specific 3′-terminal 2′-O-methylation of animal miRNAs by a genetically encoded, plant-specific methyltransferase (HEN1), with chemoselective small-RNA cloning and high-throughput sequencing. We show that mime-seq uncovers the miRNomes of specific cells within Caenorhabditis elegans and Drosophila at unprecedented specificity and sensitivity, enabling miRNA profiling with single-cell resolution in whole animals. Mime-seq overcomes current challenges in cell-type-specific small-RNA profiling and provides novel entry points for understanding the function of miRNAs in spatially restricted physiological settings.
The implementation of distinct gene expression profiles is essential for animal development and physiology. Post-transcriptional silencing by small regulatory RNAs such as miRNAs plays essential roles in shaping gene expression during these processes. These 21–24-nt long RNAs act in an RNA-induced silencing complex (RISC) with an Argonaute-family protein, directing repression of target mRNAs through base-pairing interactions typically between the mRNA 3′ UTR and the miRNA 5′ end or seed region1.
The functional contribution of miRNAs, or other gene expression regulators, to the biology of an organism depends on the cells in which they act. Thus, miRNA profiling methods are powerful entry points to dissect the roles of members of this repressor class. Many miRNAs are expressed with high cell-type specificity, sometimes in rare cell populations within complex tissues2,3,4,5,6. This is most noticeable in animal nervous systems, composed of a so-far innumerable cellular diversity, within which miRNA expression can be restricted to specific neuronal subpopulations7. Therefore, understanding endogenous context-dependent miRNA functions requires approaches to profile miRNAs from specific cells within complex mixtures—for example, whole tissues, organs or even organisms when dissection is unfeasible.
Two main strategies for identifying cell-specific miRNAs have been employed. The first is to physically isolate cells of interest from a complex tissue by, for example, fluorescence-activated cell sorting or laser-capture microdissection, followed by high-throughput sequencing methods to uncover their miRNome8,9,10,11,12. Cell isolation can require expensive, specialized equipment and—depending on how rare a cell type is—yields can be low, making well-established small-RNA sequencing protocols challenging. In addition, cell manipulations may cause changes in gene expression or even cell death. Alternatively, cell-specific expression of epitope-tagged RISC components followed by immunopurification and sequencing of associated small RNAs has been employed13,14,15. However, the underlying protocols are more laborious than direct RNA isolation and, like any antibody-based separation method, may result in low-signal-to-noise performance.
To overcome these challenges, we devised mime-seq, a strategy to chemically mark animal small RNAs in a cell-specific manner by transgenic expression of the plant methyltransferase HEN1 from Arabidopsis thaliana (Ath-HEN1). When combined with a methylation-dependent small RNA cloning strategy, we uncover methylated miRNAs from an estimated tissue-contribution as small as 1/100,000. We employed mime-seq to unveil the miRNomes of specific tissues in C. elegans and Drosophila.
An Arabidopsis RNA methyltransferase methylates animal miRNAs
To gain genetic access to cell-type-specific miRNA profiles in a temporally and spatially controlled manner, we chose Ath-HEN1, which methylates small RNAs in plants, a modification that is absent from the bulk of animal miRNAs16,17,18. Ath-HEN1 methylates the 2′ position of the 3′-terminal ribose in RNA duplexes with 2-nt 3′ overhangs and a length of 21–24 nt16,19 (Fig. 1a). Because these features are hallmarks of Dicer products in plants and animals, we reasoned that expression of Ath-HEN1 in animals should lead to miRNA methylation. Ath-HEN1-directed, 3′-terminal 2′-O-methylation prevents periodate-induced oxidation, which otherwise cleaves the terminal, cis-diol-containing ribose in unmodified RNA. Therefore, only methylated small RNAs preserve an adaptor-ligatable 3′ OH for cDNA library preparation (Fig. 1a and Supplementary Fig. 1a), as previously reported20. We reasoned that restricted expression of Ath-HEN1 in animals would then enable detection of cell-specifically methylated miRNAs.
We first conducted in vitro methylation assays using FLAG-tagged Ath-HEN1 purified from Drosophila S2 cells (Supplementary Fig. 1b) and assessed RNA methylation by β-elimination, which removes the 3′-terminal nucleoside of unmethylated but not methylated RNA21, followed by high-resolution PAGE. Consistent with previous reports, Ath-HEN1 acted specifically on double-stranded small RNAs of 21–24 nt in length irrespective of primary sequence16,19 (Supplementary Fig. 1c).
Most animals possess homologs of Ath-HEN1 (hen1 in flies, henn-1 in worms), which predominantly act on small-RNA classes other than miRNAs (i.e. siRNAs and piRNAs)22,23,24,25,26 (Supplementary Fig. 2a). Accordingly, Hen1 depletion does not result in miRNA-related phenotypes in animals, in contrast to plants22,23,24,25,26,27, and does not affect relative miRNA abundance (Supplementary Fig. 2e and ref. 24). Nevertheless, to exclude endogenous Hen1/HENN-1 as potential sources of miRNA methylation, we performed subsequent experiments in hen1/henn-1 loss-of-function backgrounds. In Drosophila S2-hen1KO cells, expression of Ath-HEN1 caused methylation signatures for all tested miRNAs, although to different levels (Supplementary Fig. 1d). In contrast, the abundant 2SrRNA remained unmethylated, consistent with Ath-HEN1 specificity for small-RNA duplexes.
To test orthogonal methylation in vivo, we expressed Ath-HEN1 ubiquitously in C. elegans and in Drosophila. In Drosophila, Ath-HEN1 was driven via the Gal4-UAS system, using transgenic Act5C-Gal4 and UAS-Ath-HEN1 (Supplementary Fig. 1e). Methylation was only detected when both Act5C-Gal4 and UAS-Ath-HEN1 were present, confirming that Ath-HEN1 methylates miRNAs in vivo in flies (Fig. 1b and Supplementary Fig. 1f). In C. elegans, we generated a single-copy integration of rps-5prom::Ath-HEN1 (Supplementary Fig. 1g) in the henn-1(tm4477) background24,28. All tested miRNAs, present in different tissues, showed signatures of methylation by Ath-HEN1 (Fig. 1c and Supplementary Fig. 1h), although to varying degrees. Importantly, expression of Ath-HEN1 did not affect miRNA relative abundance in either Drosophila cells or whole worms (Supplementary Fig. 1i,j). Moreover, worms and flies expressing ubiquitous Ath-HEN1 were viable and fertile and showed no signs of altered miRNA function.
Methylation-dependent small-RNA cloning
To assess the sensitivity and specificity of methylation-dependent small-RNA cloning and sequencing, we diluted total RNA from Drosophila S2-hen1KO cells stably expressing Ath-HEN1 with total RNA from mouse embryonic stem cells (mESC, which lack methylated small RNAs), in ratios spanning five orders of magnitude (1:1 to 1:100,000; Fig. 2a). Small RNAs in each mixture were sequenced before and after oxidation, and the enrichment of fly over mouse miRNAs was assessed in oxidized versus unoxidized libraries (Fig. 2a and Supplementary Table 1). In dilutions up to 100-fold, ≥96% (44/46) of abundantly expressed Drosophila miR strands were recovered, and Drosophila miRNAs made up ≥84% of all miRNA reads after oxidation. Even when diluted 100,000-fold—to a point where Drosophila miRNAs were undetectable in unoxidized samples—oxidation-based cloning retrieved 83% (38/46) of abundantly expressed Drosophila miR species, which made up 11% of all miRNA reads after oxidation (Fig. 2, Supplementary Fig. 3a, Supplementary Table 1). Statistical analysis on an extended set of confidently detected miRNAs (64 Drosophila and 301 mouse miRNAs) revealed a true positive rate (TPR) ≥ 0.80 and a false positive rate (FPR) ≤ 0.06 throughout the dilution series (Fig. 2b and Supplementary Fig. 3a)29.
Consistent with Ath-HEN1 methylating both strands of miRNA duplexes in plants, we also effectively recovered confidently detected (>100 counts per million (cpm) in unoxidized sample) Drosophila miR* strands upon oxidation (Supplementary Fig. 3b). However, TPRs obtained from the analysis of miR* species corresponding to confidently detected miRs (64 Drosophila and 301 mouse miR* species) were significantly lower when compared to miRs (TPR ≥ 0.33), but also maintained low FPRs (≤0.04) (Supplementary Fig. 3c). This lower overall recovery was expected, as miR* strands are typically low in abundance because of their rapid decay upon loading of the partner miR strand onto Argonaute30,31.
Our results suggest that directing Ath-HEN1 expression to specific cells should enable recovery of the most relevant miRNAs expressed in a single cell out of whole C. elegans worms (959 somatic cells) at a TPR of 0.88, and even in single neurons within fly brains (∼105 neurons) at a TPR of 0.80, in both cases with very low FPRs (<0.06).
To test whether Ath-HEN1-directed methylation followed by oxidation-resistant cloning preserves the relative quantity of miRNAs, we compared the abundance of untreated Drosophila S2 miRNAs to that of the Drosophila miRNAs that are significantly enriched after oxidation (P < 0.001), throughout the S2-Ath-HEN1/mESC dilution series. This analysis revealed statistically significant, high correlations (Spearman's correlation rs > 0.82, P < 10−5), indicating that Ath-HEN1-directed methylation followed by chemoselective small-RNA cloning can be considered semiquantitative because it preserves, to a large extent, the relative abundance of methylated miRNAs (Supplementary Fig. 4). Note, however, that for weakly expressed miRNAs at higher dilutions, this correlation is weaker.
To test our ability to retrieve methylated miRNAs from in vivo systems, we sequenced miRNAs from Drosophila S2 cells expressing Ath-HEN1 before and after oxidation. The relative abundances of mature miRNAs in these two samples show a significant, high correlation (rs = 0.88, P < 10−15; Supplementary Fig. 2d), suggesting that Ath-HEN1 efficiently methylates most Drosophila miRNAs, which are effectively retrieved by methylation-dependent cloning. We also sequenced miRNAs from C. elegans larvae ubiquitously expressing Ath-HEN1, before and after oxidation. The relative abundances of mature miRNAs in both conditions display a significant, high correlation (rs = 0.92, P < 10−15; Supplementary Fig. 2f), showing that Ath-HEN1 efficiently methylates the majority of C. elegans miRNAs. However, the relative abundance of 20 miRNAs significantly decreased after oxidation (P < 0.001, log2 fold change < −2; Supplementary Fig. 2f,g and Supplementary Table 2). 15/20 were not enriched in any of the subsequent experiments, in which Ath-HEN1 was driven by different tissue-specific promoters (in blue, Supplementary Fig. 2f,g; Supplementary Table 2), most likely because the precursor duplexes for these miRNAs lack the overhang structure preferred by Ath-HEN116,19. The remaining five, despite inefficient methylation in this experiment, were significantly enriched (log2 fold change > 1, P < 0.001) in one or more subsequent experiments in which Ath-HEN1 was driven by tissue-specific promoters (Supplementary Table 2). Overall, we were able to efficiently recover >90% of all expressed miRNAs in Ath-HEN1-expressing C. elegans.
Mime-seq robustly retrieves neuronal miRNAs in Caenorhabditis elegans
To test the specificity and reproducibility of mime-seq, we set out to obtain the miRNome of the C. elegans nervous system by driving Ath-HEN1 with three pan-neuronal promoters (rgef-1, unc-31 and rab-3), in the henn-1(tm4477) background. The three promoters drive expression in most of the worm's 302 neurons32, albeit at different levels (Supplementary Fig. 5a), thus providing a means to test the sensitivity of mime-seq to Ath-HEN1 dosage. We sequenced small RNAs from synchronized L1-stage worms (222 neurons/558 cells) before and after oxidation and calculated the change in relative abundance upon treatment for every miRNA. Implementing cutoffs of log2 fold change > 1 and P < 0.001 revealed 33–44 enriched miR strands, depending on the driver (Fig. 3a; Supplementary Fig. 5b,d,f; Supplementary Table 3). Reproducibility between biological replicates for each individual neuronal driver was very high (rs ≥ 0.98) (Supplementary Fig. 5c,e,g), and the pairwise correlations of miRNA fold changes between the three pan-neuronal experiments was also remarkably high (rs1 = 0.86, rs2 = 0.83, rs3 = 0.93), despite disparate Ath-HEN1 expression levels (Fig. 3b and Supplementary Fig. 5a). Together, these experiments revealed 49 miRs present in the nervous system, of which 59% were enriched in 3/3, and 84% in ≥2/3 experiments (Fig. 3c and Supplementary Table 3). Additional support for the observed miR strand enrichments was obtained from the analysis of confidently detected miR* strands (Supplementary Fig. 6).
To independently assess the specificity of miRNA expression, we generated transgenic C. elegans lines carrying transcriptional reporters for 40 miRNAs. These reporters were generated from large genomic clones, in fosmid vectors, in which the miRNA of interest was replaced by gfp within 35–40 Kb of endogenous genomic context bearing all relevant cis-regulatory sequences33,34. The GFP-based expression patterns showed remarkable correspondence with enrichments and depletions obtained by mime-seq (highlighted in Fig. 3a,b). Specifically, 13/15 reporters for mime-seq-enriched miRNAs were expressed in neurons (Fig. 3c and Supplementary Table 3). These represent miRNAs expressed in several neurons (miR-124), as well as restricted miRNAs (miR-791 in 3 pairs of sensory neurons35, and lsy-6 in a single sensory neuron34) (Fig. 3d). An additional 13 mime-seq-enriched miRNAs were reported as neuronal in a previous study that employed GFP reporters driven by short (1–2 Kb) sequences upstream of miRNA precursors4 (P < 0.00), Supplementary Table 3). Moreover, the neuronal identity of several miRNAs is supported by neuronal miRISC immunoprecipitation experiments, which identified 16 neuronal miRNAs but did not have the sensitivity to sufficiently enrich for miRNAs like lsy-6 (ref. 14). Furthermore, 23 fosmid-based reporters showed expression exclusively in non-neuronal cells. Out of these, 21 corresponded to miRNAs depleted in all three pan-neuronal experiments. From a total of 40 reporters, only four miRNAs showed neuronal expression but were not significantly enriched by mime-seq (P < 0.001, Supplementary Table 3). Overall, concordance between mime-seq and the transcriptional reporters is very high. The few discrepancies between both may correspond to technical false positives/negatives or, alternatively, may reflect unexplored instances of post-transcriptional control of miRNA abundance.
By applying mime-seq to the C. elegans nervous system, we revealed the most comprehensive neuronal miRNome in worms to date. A few of these miRNAs have already been shown to play roles in neuronal development or physiology35,36,37,38; we provide an extended list of candidates which may have similarly interesting functions.
Mime-seq reveals the miRNomes of diverse tissues
We also applied mime-seq to the pharynx, intestine and body-wall muscle, tissues that differ in size and cellular complexity. We cloned small RNAs before and after oxidation, from synchronized L1-stage larvae expressing Ath-HEN1 in the different tissues. Implementing cutoffs of log2 fold change > 1 and P < 0.001 revealed 40 miRNAs enriched in the pharynx, 38 miRNAs in body-wall muscles and 44 in the intestine, upon oxidation (Fig. 4, Supplementary Fig. 7, Supplementary Table 4). Validation with multiple GFP-based transcriptional reporters confirmed the known tissue-specificity of a number of miRNAs (e.g., miR-1 (ref. 39); Fig. 4b) and revealed the tissue-specific expression of many miRNAs whose expression patterns were unknown (Fig. 4b and Supplementary Table 4).
Comparing the enriched miRNomes from different tissues revealed that each tissue has shared and exclusive miRNAs. For example, 17/49 (35%) of neuronal miRNAs are absent from muscles, pharynx and intestine (Fig. 4c and Supplementary Table 4); this was confirmed in 8/8 GFP-based neuronal reporters (Supplementary Tables 3 and 4). Using mime-seq, we were able to capture miRNAs produced from restricted compartments of C. elegans without the need for cell dissociation and/or sorting.
Mime-seq reveals the miRNome of a single neuron pair in Caenorhabditis elegans
To test the sensitivity of mime-seq in vivo, we expressed Ath-HEN1 in only one pair of neurons (2/558 cells in L1), in henn-1-deficient worms (Supplementary Fig. 8a). We chose the ASE sensory neurons, as we previously identified a number of miRNAs expressed in these cells34,35. Upon oxidation, we retrieved a set of 21 significantly enriched miRNAs (log2 fold change > 1, P < 0.001; Fig. 5a, Supplementary Fig. 8b, Supplementary Table 5). Remarkably, the four miRNAs known to be expressed in the ASEs are among the top five enriched miRNAs, with a log2 fold change > 5 (Fig. 5a,b and Supplementary Table 5). The expression pattern of the top enriched miRNA, miR-1821, was unknown; a new transcriptional reporter showed that it is expressed in the right ASE neuron in addition to a few other neurons (Fig. 5b and Supplementary Table 5). These findings corroborate the sensitivity of the method, which retrieved miRNAs expressed in ∼1/300 cells from a whole organism, with an extraordinary rate of validation, confirming known miRNA expression patterns as well as confidently ascribing new ones.
Mime-seq does not require a specific genetic background
Working in an endogenous Hen1 mutant background eliminates potential unwanted small RNA methylation, but may be undesirable or unfeasible if working with animals where endogenous Hen1 mutants are not available or well-characterized. We therefore repeated the ASE-specific mime-seq from Figure 5a in a wild-type background (Supplementary Fig. 8c). This experiment retrieved all miRNAs previously identified as ASE-specific in the henn-1 mutant, plus a few additional ones, suggesting that endogenous HENN-1 contributed to the methylation of miRNAs. To quantify this background activity, we sequenced small RNAs from wild-type, nontransgenic C. elegans before and after oxidation and found 26 miRNAs that are consistently enriched upon this treatment (Supplementary Table 6). Subtracting this background enrichment from the ASE-specific experiment resulted in a miRNA candidate list with highly significant overlap to that obtained in the henn-1(tm4477) background (P = 1.66 × 10−12 Fig. 5c, Supplementary Fig. 8d, Supplementary Table 5).
We further compared results from wild-type and henn-1(tm4477) backgrounds for the body-wall-muscle experiment (Supplementary Fig. 8e and Supplementary Table 5). Again, upon computational subtraction of the endogenously methylated miRNAs in the wild-type background, we obtained a common set of muscle-enriched miRNAs from both experiments (Supplementary Fig. 8e and Supplementary Table 5).
Mime-seq scales to a more complex animal
To test how mime-seq scales to bigger, more complex animals, we investigated the specific miRNomes from Drosophila muscles by sequencing small RNAs extracted from adult fly bodies expressing Ath-HEN1 from a Mhc-Gal4 driver (all muscles) or an Act88F-Gal4 driver (flight muscles) (Supplementary Fig. 8f and Supplementary Table 5). We retrieved 11 and 13 miRNAs enriched >two-fold, respectively. Out of these, seven are shared between both data sets, including the well-known, muscle-specific miR-140 as well as miR-277, recently shown to play a role in adult fly muscle41(Supplementary Fig. 8f).
Mime-seq overcomes current challenges for in vivo miRNA profiling from specific cell types within complex tissues or whole organisms. It circumvents excessive manipulations such as cell sorting or immunoprecipitation, which reduce yields and increase noise, achieving greater sensitivity. In direct comparison with neuronal and intestinal Argonaute immunoprecipitations in C. elegans13,14, mime-seq identified 3- and 2.6-fold more miRNAs respectively, many of which we validated, including a miRNA expressed in a single neuron out of the whole animal34.
Comparison of the miRNomes obtained by mime-seq with transcriptional reporters for several C. elegans miRNAs showed a high correspondence, both for enriched and depleted miRNAs. However, we also observed a few discrepancies. These may result from technical issues—for example, if a miRNA is poorly methylated (Supplementary Table 2); or from biological reasons—for example, if a miRNA is transcribed but not processed in a certain tissue or cell type42. In that respect, because mime-seq reports on the presence of the mature miRNA, it should provide a more accurate picture of the functional miRNome of a cell type than transcriptional reporters. Mime-seq may therefore additionally reveal instances of post-transcriptional regulation of miRNA maturation and decay.
Mime-seq builds on the rich genetic toolset available for animal model organisms, including Gal4-UAS-based expression system in Drosophila and transgenesis in C. elegans, which enable expression of Ath-HEN1 in cells of interest. We envision that mime-seq is similarly applicable to other genetically accessible animal systems. Note that good descriptions of the drivers of Ath-HEN1 expression are essential for the interpretation of these experiments. For instance, in all pan-neuronal worm experiments we detect miR-252 (Supplementary Table 3) which, based on GFP reporters, is only transcribed in the worm pharyngeal gland cells. These secretory cells share the neuronal secretory machinery and in fact express rab-3, rgef-1 and unc-3132. If applicable, combining mime-seq with partial dissection may overcome some of the limitations posed by imperfect specificity of genetic expression systems.
In setting up mime-seq, we found C. elegans miRNAs that are endogenously methylated by HENN-1. We noticed that 11/26 endogenously methylated miRNAs have precursor stems with only one or two mismatches (Supplementary Fig. 2c and Supplementary Table 6). The nature of the precursor duplex determines the sorting of small RNAs into different Argonaute proteins; while miRNA stems tend to have bulges, siRNAs arise from perfectly complementary duplexes43,44,45,46,47,48. Consistent with this, seven endogenously methylated miRNAs were detected in immunoprecipitations of the Argonaute RDE-1, which typically binds siRNAs49 (Supplementary Table 6). As the specificity of HENN-1 is given by its association with specific Argonaute proteins24,25,26, we suggest that either HENN-1 is able to associate with RDE-1 or some of these miRNAs are loaded onto other Argonaute proteins that interact with HENN-1.
Interestingly, as Ath-HEN1 targets Dicer-cleavage products, the observation that animal miRNAs are also efficiently methylated implies that, similar to plants, miRNA biogenesis and loading onto Argonaute are uncoupled in animals. This is further supported by the recovery of methylated miR* strands from flies and worms.
MicroRNAs provide an additional layer of gene regulation during cell-type specification, and we and others have hypothesized that the number of miRNAs in a tissue may reflect the complexity in its cellular composition. Supporting this, the worm nervous system, with ∼100 different neuron types, has the highest number of exclusively enriched miRNAs. Further dissection of this and other tissues into their different cellular constituents will provide insights into the contribution of miRNAs to cellular diversification. Mime-seq offers a promising entry point toward this.
Primers and probes.
All oligonucleotide sequences used in this study are provided in Supplementary Table 7.
All fly stocks were maintained under standard conditions at 25 °C. All worm strains were grown at 20 °C and generally handled as previously described51. A complete list of strains used in this study is presented in Supplementary Table 8. For all mime-seq experiments using C. elegans, we profiled synchronized L1s obtained through standard bleaching protocol52. Strains generated in this study will be available through the Caenorhabditis Genetics Center (CGC) or the Vienna Drosophila Resource Center (VDRC).
A Drosophila melanogaster codon-optimized version of Ath-HEN1 was designed using the IDT codon optimization tool (https://eu.idtdna.com/CodonOpt) and generated by gene synthesis (IDT). Ath-HEN1 coding sequence was PCR amplified (using primers Ath-HEN1-fwd and –rev), cloned into pENTR/D using TOPO-cloning (Invitrogen), and verified by Sanger Sequencing. A catalytic mutant version of Ath-HEN1 carrying four amino acid exchanges was generated by consecutive site-directed mutagenesis using primers Ath-HEN1-CM-fwd-1 and -rev-1 (E796A), Ath-HEN1-CM-fwd-2 and –rev-2 (E799A/H800A) and Ath-HEN1-CM-fwd-3 and –rev-3 (H860A). A constitutive Drosophila expression vector was derived for Ath-HEN1WT and Ath-HEN1CM by LR cloning into pAFMW, resulting in Actin5C promoter-driven expression of N-terminal FLAG-Myc-tagged Ath-HEN1WT or Ath-HEN1CM.
A PhiC31-integrase-compatible vector for Gal4-driven somatic expression of Ath-HEN1 in vivo in flies (pTFMW-attB) or for germline-enhanced expression (pPFMW-attB) was generated by Gibson assembly using pTFMW (Drosophila Genomic Resource Center, Indiana University) and PCR amplified attB site followed by LR recombination.
Worm-targeting vectors for single-copy transgene insertion on chromosome II were constructed in the pCFJ350-ttTi5605 vector backbone (chromosome II targeting vector28). A C. elegans codon-optimized53 version of Ath-HEN1 was synthesized by IDT, N-terminally tagged with 2xFlag-linker-MYC and cloned under the control of tissue-specific promoters into pCFJ350-ttTi5605. To express Ath-HEN1 pan-neuronally, three drivers were used. First, the rab-3 promoter (1.2 Kb) was subcloned together with Ath-HEN1 ORF and unc-54 3′ UTR (3.9 Kb fragment), amplified using primers containing AvrII restriction sites, and then cloned into pCFJ350-ttTi5605 MCS as an AvrII fragment. The unc-31 promoter (6.8 Kb) was cloned as a Not-I/Msc-I fragment into the recipient targeting mosSCI vector expressing the above mentioned rab-3::Ath-HEN1 sequence and linearized with the same restriction enzymes (in order to replace the rab-3 promoter surrounded by Not-I/Msc-I sites). The third pan-neuronal promoter, rgef-1 (2.7 Kb), was cloned as an Asc-I/FseI fragment into the recipient targeting mosSCI vector expressing the above mentioned rab-3::Ath-HEN1 sequence and linearized with the same restriction enzymes, again in order to swap the rab-3 promoter (surrounded by Sph-I/Msc-I sites) upstream the 2xFlag-linker-MYC-Ath-HEN. The rps-5 promoter (4 Kb) was chosen for ubiquitous expression, cloned as an Sph-I/MscI fragment into the recipient targeting mosSCI vector to replace the rab-3 promoter (excised as an Sph-I/Msc-I fragment) upstream the 2xFlag-linker-MYC-Ath-HEN1. For expression in the pharynx, the myo-2 promoter (2.5 Kb) was subcloned upstream of the Ath-HEN1 ORF and unc-54 3′UTR (3.9 Kb fragment), and the whole transgene was amplified using primers containing AvrII restriction sites and then cloned into pCFJ350-ttTi5605-MCS as an AvrII fragment. For expression in the intestine, the elt-2 promoter (5 Kb) was subcloned upstream of Ath-HEN-1::unc-54 3′UTR and then amplified all together in two PCR reactions (each one ∼4.5 Kb), using outside primers containing AvrII restriction sites, and Gibson overlap in between the two fragments. The three pieces were Gibson ligated into the AvrII-linearized pCFJ350-ttTi5605-MCS. For expression in body wall muscle, the myo-3 promoter (2.5 Kb) was cloned as an AscI-I/Nhe-1 fragment (amplified with the addition of restriction sites from an existing vector) into the AscI-I/Nhe-1 linearized targeting vector expressing 2xFlag-linker-MYC::Ath-HEN1::T2A::mCherry::TY1::H2B (newly generated for further characterization of HEN1 localization). Finally, for ASE expression, a motif including the binding site for the transcription factor che-1 multimerized eight times was amplified from the published vector54 and Gibson cloned into the targeting vector expressing Ath-HEN1::T2A::mCherry::TY1::H2B. All sequences and plasmids used to generate transgenic worms are available upon request, and mosSCI transgenic strains are deposited in CGC. Primers used for cloning are listed in Supplementary Table 7. Plasmids generated in this study will be made available through Addgene (pUASp-FMW-attB-AthHEN1 (104957), pUASt-FMW-attB-AthHEN1 (104958)).
A monoclonal antibody against dmHen1 was raised by the MFPL monoclonal antibody facility (S. Schuechner and E. Ogris). Briefly, a 6× HIS-tagged full-length dmeHen1 was expressed in E. coli BL21 and purified under denaturing conditions by Ni-affinity chromatography. Balb/c mice were immunized subcutaneously three times (every 2–3 weeks) with 50 μg of purified antigen mixed at a ratio of 1:1 with adjuvant, before a final intravenous immunization with 30 μg purified antigen (adjuvant-free). Splenic B cells were fused with X63-Ag8.653 mouse myeloma cells, and clones were tested by immunoblot analysis for the detection of Hen1. Clone 8C2-H4 yielded the best signal-to-noise performance.
In vitro methylation assay.
FLAG-tagged Ath-HEN1 was immunopurified from Drosophila S2 cells stably expressing FLAG-Myc-Ath-HEN1 as described previously55. For RNA substrate preparation, 40 pmole miRNA guide strand (i.e., dme-let-7-5p or dme-miR-34-5p) was 5′ 32P-radiolabeled in a standard kinase reaction using T4-polynucleotide kinase (New England Biolabs), unincorporated nucleotides removed by Sephadex G-25 spin column purification (GE Healthcare), and labeled RNAs polyacrylamide-gel purified. If indicated, guide strands were annealed to 5′ phosphorylated miR* strands (i.e., dme-let-7-3p or dme-miR-34-3p). For RNA substrate sequences see Supplementary Table 7. In the methylation assay, miRNA substrates (5 μM) were incubated with immunopurified FLAG-Ath-HEN1 in a standard RNAi reaction containing 1.2 mM S-adenosylmethionine at 25 °C for 30 min56, followed by Phenol/Chloroform extraction. RNA was then subjected to beta-elimination and 15% denaturing gel electrophoresis. Gels were dried and exposed to a storage phosphor screen.
Generation of hen1 (FBgn0033686) mutant flies (hen1m1-6 allele) by CRISPR–Cas9 genome engineering was performed as described57. Briefly, isogenized w1118 embryos were injected with the plasmid pDCC6 (Addgene) containing a gRNA sequence (Supplementary Table 7) targeting the first exon of the hen1 locus. Hatched flies were crossed to second chromosome balancer flies, and F1 resulting males were screened for frameshift mutations by PCR amplification of the targeted hen1 locus using the primers hen1-fwd and hen1-rev (Supplementary Table 7) followed by Sanger sequencing. Progeny carrying a 5-nt frameshift deletion (allele hen1m1-6) were used for further experiments. Depletion of Hen1 was confirmed by western blotting. An isogenic w1118 stock was used as the wild-type control. Age-matched, 2–5 day old flies were used for experiments.
Transgenic UAST-FLAG-Myc-Ath-HEN1 flies were generated by injecting expression vector pTFMW-attB-Ath-HEN1 into stocks with attP2 landing site on chromosome 3 (ref. 58). Correct integration was confirmed by Sanger sequencing. Ath-HEN1 expression was functionally validated by β-elimination and Northern hybridization.
CRISPR–Cas9 genome editing in Drosophila S2 cells.
For cloning of gRNA-expression vectors targeting the Hen1 locus by CRISPR–Cas9 in Drosophila S2 cells, pairs of gRNA-coding oligonucleotides were annealed (to generate four sgRNAs, #1–4; see Supplementary Table 7) and ligated to BspQI-digested pAc-sgRNA-Cas961. Drosophila S2 cells were transfected with a mixture of four pAc-sgRNA-Cas9 plasmids encoding Cas9 and an sgRNA targeting a 105-bp region in the first exon of DmHen1. Cells were selected for Cas9-transgene expression using puromycin (5 μg/ml) for 9 d, followed by serial dilution under nonselective conditions and expansion of single-cell clones. For clonal selection, cells were then diluted serially 1/4 in a 96-well plate starting with ∼20,000 cells/well in clonal selection medium (80% fresh Schneider Medium (containing 10% FCS), 20% conditioned Schneider medium) and incubated 10–14 d. Single, round-shaped clones of cells emerging from wells with permissive cell density were picked and expanded. Individual clones were tested for successful targeting by interrogating methylation status of siRNAs (i.e., esi-2.1) by β-elimination and northern blotting. Positive candidates were verified by western blotting, and disruptive mutations were confirmed by Sanger sequencing.
Generation of mosSCI transgenic worm strains.
In order to generate transgenic strains, we followed the published MosSCI protocol28. Injection mixes contained: 50 ng/μl specific-insertion template, 50 ng/μl pCFJ601 (Peft-3::Mos1 Transposase), 10 ng/μl pMA122 (Phsp-16.41::peel-1::tbb-2 3´UTR for heat-shock-driven PEEL-1 negative selection), 2 ng/μl Pmyo-2::mCherry (pharyngeal coinjection marker), 5 ng/μl elt-2::DsRed (intestinal coinjection marker), 50 ng/μl ttx3::mCherry (AIY-specific coinjection marker) and 33 ng/μl pBSK (as a carrier). The mixture was microinjected into the gonads of Unc young adults of strain EG6699 (ttTi5605 II; unc-119(ed3) III). After injection, single worms were picked to new plates and expanded at 25 °C until starvation (10–12 d). Plates with rescued non-Unc animals were subjected to heat shock 34 °C for 3 h in a water bath to activate the PEEL-1, a negative-selection marker that kills animals still carrying extrachromosomal arrays. After overnight recovery at 20 °C, plates were visually screened to identify non-Unc animals that survived heat shock and did not express the red fluorescent coinjection markers. Single worms from these plates (about 5–6 each) were picked to establish lines, and the presence of insertion events was verified by PCR using primers listed in Supplementary Table 7.
Five to ten adult flies per 50 μl S2 cell pellet were lysed in 100 μl 1× Lysis-IP-Buffer (30 mM HEPES KOH pH 7.4, 100 mM KOAc, 2 mM MgOAc, 100 mM DTT, 1% NP-40, 5% Glycerol (Sigma), cOmplete EDTA-free proteinase inhibitor cocktail (Roche)). Protein concentration was determined by Bradford Protein Assay (BioRad). Upon addition of 2× sample buffer (50 mM TRIS pH 6.8, 5% SDS, 20% Glycerol, Bromphenolblue) to 30 μg total protein, samples were boiled at 95 °C for 5 min and separated on 4–15% Mini-PROTEAN TGX Gels (BioRad) and transferred to a PVDF-membrane (Merck). The following antibodies were used to detect Ath-HEN1: mouse monoclonal anti-FLAG Antibody, clone M2 (Sigma, F1804) at 1:10,000 dilution and goat-anti-mouse IgG HRP-linked Antibody (Thermo Fisher, G21240) at 1:10,000 dilution. The following antibodies were used for Actin detection: rabbit anti-Actin Antibody (Sigma, A2066) at 1:10,000 dilution and goat-anti-rabbit IgG HRP-linked Antibody (Thermo Fisher, G21234) at 1:10,000 dilution. For detecting dmHen1, mouse anti-dmHen1 (own source) at 1:500 dilution was used. Membranes were developed with Clarity Western ECL reagent (BioRad).
50–100 young larvae worms (L1–L3) were collected in 15 μl of water, vortexed and freeze cracked at least three times to facilitate dissociation of the cuticle. Upon addition of 15 μl of 2× sample buffer, samples were boiled at 100 °C for 4–5 min, separated on 10% Mini-PROTEAN TGX Stain-Free Gels (BioRad) and transferred to nitrocellulose (BioRad). The following antibodies were used to detect MYC-Ath-HEN1: mouse monoclonal anti-MYC tag Antibody, clone 4A6 (catalog number 05-724, Merck Millipore) at 1:2,000 dilution and anti-mouse IgG HRP-linked Antibody (Cell Signaling Technology, #7076) at 1:2,000 dilution. The following antibodies were used for Tubulin detection: rabbit anti-gamma Tubulin Antibody (ab50721, abcam) at 1:1,100 dilution and anti-rabbit IgG HRP-linked antibody (Cell Signaling Technology, #7074) at 1:2,000 dilution. Membranes were developed with ECL reagent (Pierce ECL Plus Western Blotting Substrate, Thermo Fisher).
Fosmid-based reporters were generated as previously described33. Primer sequences used to build all fosmids and resulting constructs described in this work are available upon request. Briefly, all miRNA fosmid reporters shown in this study were generated by replacing the precursor miRNA hairpin by gfp34. All fosmids were injected as complex arrays at 10 ng/μl together with sonicated OP50 genomic DNA at 100 ng/μl and a coinjection marker for screening (typically ttx-3::mCherry PvuI 5 ng/μl). Furthermore, mir-790, -791, -793, -1821 and lsy-6 transcriptional reporters were crossed to a Pche-1::mCherry reporter (otIs232) for easy identification of the ASE neurons.
Differential interference contrast (DIC) and fluorescence imaging of whole worms mounted on agar pads and immobilized using 100 mM sodium azide as a paralytic was performed using a widefield microscope, Axio Imager.Z2 (Zeiss) equipped with sCMOS camera Orca Flash 4.0 (Hammamatsu) running under Metamorph (Molecular Devices). Images were acquired as z-stacks (at 1 μm distance) and further processed with Fiji62 to obtain maximum intensity projections. No image manipulations were performed, except adjustment of brightness and contrast.
Total RNA extraction.
Total RNA from S2 cells and whole flies was extracted using TRIzol Reagent (Ambion). C. elegans larvae (L1) were harvested in 600 μl TRIzol (Invitrogen), freeze cracked three to four times and vortexed to dissolve the cuticle. Upon addition of 120 μl of chloroform, samples were agitated vigorously and centrifuged at full speed (14,000 r.p.m.) for 15 min at 4 °C. Following centrifugation, the mixture separates into lower red, phenolchloroform phase, an interphase and a colorless upper aqueous phase. RNA remains exclusively in the aqueous phase. Therefore, only the upper aqueous phase was carefully transferred without disturbing the interphase into fresh tube, where it underwent an extra cleaning step performed by adding 1 volume of acid phenol chloroform (5:1 solution, pH = 4.5, Ambion). Upon vortexing and a centrifugation step of about 3 min at full speed at 4 °C, again the upper aqueous phase was transferred into a new tube to be subjected to RNA precipitation, achieved by adding 1 volume of isopropyl alcohol and 1 μl of glycogen (20 mg/ml, G1767-1VL Sigma). Samples were incubated at RT for 10 min and centrifuged full speed for at least half an hour. The RNA precipitate, often invisible before centrifugation, forms a gel-like pellet on the side and bottom of the tube. RNA wash was performed by first removing the supernatant and then adding at least 300 μl of 80% cold ethanol. Samples were mixed by vortexing and centrifuged for 10 min at full speed at 4 °C. Upon removal of all leftover ethanol, the RNA pellet was dissolved in 20 μl (or more, depending on the pellet size) of DEPC-treated water (Ambion) by passing the solution a few times through a pipette tip.
At least 20 μg of total RNA were subjected to either high-resolution northern blot experiments or small-RNA library preparation, coupled with deep sequencing.
Small RNA library preparation.
Small RNA libraries were generated from >20 μg total RNA as described previously but including an oxidation step after RNA size selection if indicated55. For oxidation conditions, 18–30 nt size-selected (and 2SrRNA-depleted if applicable) RNA was incubated in the presence or absence of 50 mM freshly prepared sodium periodate (Sigma) in 1× Borate buffer (30 mM borax, 30 mM boric acid, pH 8.6) for 30 min at room temperature. After oxidation, samples were ethanol precipitated and further subjected to small-RNA cloning.
All C. elegans libraries were generated from biological duplicates. For the titration experiment in Figure 2a, libraries were generated on three independent sets of samples.
Small-RNA-library reads were recovered by adaptor clipping—the adaptor was cut once with cutadapt v1.12.0 (ref. 64). Adaptor-derived random 4-mers on the 5′ and 3′ ends of recovered reads were removed with custom scripts. Processed reads were size selected (i.e., 18–30 nt). Mapping of sequencing libraries to Drosophila genome (dm3), the Mus musculus genome (mm10), or the C. elegans genome (WBcel235) was performed as described55. Annotations were derived from FlyBase (r5.57) and mirBase (v21). Reads were assigned to miRNA arms with htseq-count (v0.6.1p1)65. Multimapping reads were counted only as fraction (1/number of mapped locations). Only reads with a tail fraction smaller than 0.12 are considered. For analysis of titration experiments (Fig. 2 and Supplementary Figs. 3 and 4), reads were first aligned to the mouse genome, and remaining reads were aligned to the Drosophila genome. Note that one expressed miRNA (miR184-3p) aligned to both the mouse and the Drosophila genome but was assigned to the Drosophila sample because miR-184 is only spuriously expressed in mESC.
To test the enrichment of Drosophila miRNAs in the titration experiment in Figure 2 (conducted in triplicate), we considered Drosophila and mouse miR strands with an average cpm > 10 over the three replicates for untreated samples with 1:1 dilutions (64 fly and 301 mouse miRNAs) and also their associated miR* strands (without expression cutoff; note that only 62 of 64 fly and 293 of 301 mouse miR*s were detected in at least one replicate in one condition). Then we employed DESeq2 (ref. 29) (v1.16.1) to estimate scaling factors and dispersion parameters using all dilution experiments. Lastly, pairwise comparisons between treated and untreated samples were separately done for each dilution condition, whereby the fold changes and P values were obtained from DESeq2. Small RNAs for which no enrichment score or P value could be determined (either because of absence of reads after normalization of miRNA-mapping reads or because of strong variability in detection) were excluded from the analysis (represented in n values of enrichment plots). True-positive miRNA-recovery rate (TPR) was determined from the number of abundantly expressed fly miRNAs detected in the respective libraries above cutoff (log2 fold change > 1; P value < 10−3) relative to the 64 miRNAs recovered at a read depth of >10 cpm (na = 64) in untreated 1:1 input libraries (or their respective miR*s). False-positive miRNA-recovery rate (FPR) was determined from the number of mouse miRNAs above cutoff (log2 fold change > 1; P value < 10−3).
To identify the tissue- or cell-type-specific miRNAs for both C. elegans and Drosophila, we calculated the enrichment for every miRNA by comparing the oxidation treated with untreated samples using DESeq2 (v1.16.1). For all C. elegans samples, we first removed spuriously expressed miRNAs by requiring an average cpm (counts per million) > 10 over all untreated samples analyzed in this study. All miRNA arms above that cutoff were used for a robust estimation of the dispersion parameters of the negative binomial distributions employed by DESeq2. However, beyond this step, the miRNA arms that showed higher expression in more than half of the samples without oxidation treatment were retained, and the lower expressed miRNA* arms were discarded. By comparing the respective pairs of treated and untreated samples, we obtained the fold changes (in log2 scale) and enrichment scores (i.e. P values) from DESeq2. For experiments conducted in the wild-type background, where the endogenous methyltransferases (Cel-HENN-1 or Dme-HEN1) are still present, we had to subtract the background methylation that occurs on some miRNAs. In order to do this, we included an interaction term between genotypes (e.g., ASE::Ath-HEN1 and nontransgenic N2, and similarly for the muscle experiment done in both backgrounds) and treatments in the design formula of DESeq2 to obtain miRNAs that are not only enriched due to treatment but also show higher enrichment than in the background methylation.
The R script used for these analyses, as well as the pipeline developed for the preceding mapping and counting (see above) can be found at: https://gitlab.com/tburk/smallRNA-meth.
Statistical tests for correlation analysis were performed in Prism v7.0c (GraphPad Software Inc.). The expected number of miRNAs in the overlap between two sets was calculated considering 132 expressed miRNAs in Drosophila and 123 in C. elegans as (#miRNAs in set 1 × #miRNAs in set 2)/total # of expressed miRNAs. The significance of the overlap was calculated using a hypergeometric test in R.
Life Sciences Reporting Summary.
Further information on experimental design is available in the Life Sciences Reporting Summary.
The data sets generated and analyzed during this study are available in GEO (GSE104470). The Nextflow pipeline and the R scripts used for data analysis can be found at https://gitlab.com/tburk/smallRNA-meth.
C. elegans strains are available through the Caenorhabditis Genetics Center (CGC), and Drosophila strains are available through the Vienna Drosophila Resource Center (VDRC).
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Gene Expression Omnibus
We thank Cochella lab members for various fosmid reporters; F. Schnorrer (IBDM, Marseille, France) for muscle-specific Gal4 driver lines; and the CGC (NIH-P40-OD010440) for worm strains. This work was supported by FP7/2007-2013 grants from the European Research Council to L.C. (ERC-StG-337161) and S.L.A. (ERC-StG-338252) and the Austrian Science Fund to L.C. (W-1207-B09 and SFB-F43-23) and S.L.A. (Y-733-B22 START, W-1207-B09 and SFB-F43-22). R.A.M. is a recipient of a DOC fellowship of the Austrian Academy of Sciences at IMBA. Basic research at IMP is supported by Boehringer Ingelheim GmbH.
Integrated supplementary information
Synthetic dilution experiment to determine thesensitivity of mime-seq
A few miRNAs are not efficiently methylated by AthHEN1
Pan-neuronal miRNAs from C. elegans revealed by mime-seq and comparison with transcriptional reporters
Tissue-specific miRNAs from C. elegans revealed by mime-seq
The miRNome of a single neuron class in C. elegans revealed by mime-seq in wild-type and henn-1(0) backgrounds
miRNAs endogenously methylated by Cel-HENN-1
About this article
Investigational New Drugs (2018)