Small RNA in situ hybridization in Caenorhabditis elegans, combined with RNA-seq, identifies germline-enriched microRNAs

Abstract

Over four hundred different microRNAs (miRNAs) have been identified in the genome of the model organism the nematode Caenorhabditis elegans. As the germline is dedicated to the preservation of each species, and almost half of all the cells in an adult nematode are germline, it is likely that regulatory miRNAs are important for germline development and maintenance. In C. elegans the miR35 family has strong maternal effects, contributing to normal embryogenesis and to adult fecundity. To determine whether any particular miRNAs are greatly enriched in the C. elegans germline we used RNA-seq to compare the miRNA populations in several germline-defective strains of adult C. elegans worms, including glp-4(germline proliferation-4), glh-1(germline helicase-1) and dcr-1(dicer-1). Statistical analyses of RNA-seq comparisons identified 13 miRNAs that are germline-enriched, including seven members of the well-studied miR35 family that were reduced as much as 1000-fold in TaqMan qRT PCR miRNA assays. Along with the miR35s, six others: miR-56 (a member of the miR51 family),−70, −244, −260 , −788 and −4813, none of which previously considered as such, were also identified by RNA-seq as germline-enriched candidates. We went on to develop a successful miRNA in situ hybridization protocol for C. elegans, revealing miR35s specifically concentrate during oogenesis in the pachytene region of the gonad, and persist throughout early embryogenesis, while in adult animals neither let-7 nor miR-228 has a germline-bias.

1. Introduction

Although microRNAs (miRNAs) were first described in C. elegans more than twenty years ago with the report of the lin-4 small, temporal(st), non-coding(nc) RNA (Lee et al., 1993), the miRNA field exploded after the report of let-7, a second C. elegans miRNA affecting developmental timing, and one conserved in many other organisms, including humans (Pasquinelli et al., 2000). miRNAs, usually 21–22nts long, are endogenous to most plants and animals and are cleaved in the cytoplasm from hairpin precursor RNAs by the endo-ribonuclease Dicer. Mature miRNAs regulate their mRNA targets through antisense complementarity (often with mismatches), binding mRNAs in their UTRs (untranslated regions) or coding regions. miRNA binding causes post-transcriptional regulation of their targets, usually resulting in lower mRNA and/or protein levels.

For more than a decade many studies have implicated miRNAs as regulators of development and as markers of disease. While lin-4 and let-7, and the C. elegans miRNA lsy-6 (Johnston and Hobert, 2003), were each discovered by classical genetics and display visible mutant phenotypes, targeted deletion of over 80 individual miRNAs or closely linked miRNA families revealed only a few striking abnormalities. Exceptions include deletions in two families, the miR51s or miR35s that each result in embryonic lethality (Miska et al., 2007; Alvarez-Saavedra and Horvitz, 2010). The highly-conserved miR51 family, consisting of miRs-51-56 (miR-100s in humans), is likely the oldest group of animal miRNAs, found in annelids, nematodes, flies and humans. This family is critical for pharyngeal development in the C. elegans embryo (Shaw et al., 2010). The miR35 family consists of eight miRNAs, each containing the same CACCGGC “seed” in their 5′ ends; miR35s are located in two clusters, with miR-35-41 clustered together and miR-42 350 kb away. Microarray and northern analyses revealed the miR35 family is strongly expressed in the oogenic germline (Miska et al., 2007), while additional studies revealed loss of this family results in embryonic death, male sterility and severe reductions in brood size (Alvarez-Saavedra and Horvitz, 2010; McJunkin and Ambros, 2014). Thus, the C. elegans miR35s are important both for embryonic development and to produce a normal adult germline.

In mammals, several recent studies have also considered roles for miRNAs in embryonic and germline development. In humans, the enhancement of somatic fates by the let-7 miRNA family works in opposition to miR-372, which promotes human germ cell fates in embryonic stem cells (Tran et al., 2016), while in mice, sperm, as well as oocytes, transmit miRNAs and endo-siRNAs critical to the developing embryo (Yuan et al., 2016).

Despite elegant studies of miRNAs in C. elegans identifying miRNAs that are differentially expressed during development (Kato et al., 2009) or with ageing (Kato et al., 2011; Lucanic et al., 2013), no studies targeting germline miRNAs have been reported. These studies were likely hindered by the inherent germline silencing of C. elegans transgenes, resulting in an inability of miRNA transgenes to express in the germline (Kelly et al., 1999; Martinez et al., 2008; Pierce et al., 2008; Shaw et al., 2010) and because in situ hybridizations for miRNAs in C. elegans had not yet been published, with the exception of miR-57, the most abundant miRNA in our RNA-seq studies, Supplemental Table S1, localizing to the posterior of a late-stage embryo (Zhao et al., 2010). Therefore, whether many miRNAs show a germline-bias was unknown.

Thus the goals of this study were to identify germline-enriched miRNA candidates and to develop methods to visualize C. elegans germline miRNAs by in situ hybridization. To do so, we carried out RNA-seq, followed by statistical analyses, utilizing the rather unique advantage available in C. elegans of multiple germline-minus and germline-defective mutant strains. With statistical comparisons using worms with and without germlines, 13 miRNAs were identified as germline-enriched. These include seven miR35 family members, along with a single member of the miR51 family, and five additional, individual miRNAs. To visualize the localization pattern of the miR35 family, we also developed a successful in situ hybridization protocol and applied it to several other C. elegans miRNAs, including miR-228 and let-7.

2. Materials and methods

2.1. Strains

Worm strains used for these studies include the following: wild type (N2 variety Bristol); VC178 [glh-1(gk100) I]; SS104 [glp-4(bn2) I]; and PD8753 [dcr-1(ok247) III/ghT2[bli-4(e937) let-?(q782) qIs48] (I;III)] (the ghT2, GFP-tagged chromosome, allows for selecting non-green homozygous dcr-1 worms); and MT14119 [miR-35-41(nDf50) II] (This latter mutant carries a deletion of miRs-35-41, eliminating the polycistronic miR-35-41 transcript (Alvarez-Saavedra and Horvitz, 2010)). All strains used in this study were maintained as described in (Brenner, 1974) and were grown at the optimal growth temperature of 20° unless otherwise noted. In this report strains will be referred to as follows: VC178 as glh-1; SS104 as glp-4; PD8753 as dcr-1, and MT14119 as miR-35-41.

2.2. Harvesting worms

Wild type (N2) and glh-1 worms were placed at the first larval stage (L1) on NGM agar plates seeded with OP50 E. coli, cultured at 20° or 26° and harvested as young adults 1 day beyond the fourth larval stage (L4), when mature oocytes and embryos are first detected. N2s were harvested off approximately one hundred 15 × 100 mm plates and glh-1 animals were harvested off approximately two hundred 15 × 60 mm plates by washing with M9 buffer (22 mM KH₂PO₄, 22 mM Na₂HPO₄, 86 mM NaCl). glp-4 worms were cultured in liquid S media (10% S basal (0.1 M NaCl, 0.05 M potassium phosphate, 5 μg/ml cholesterol, 1 mM potassium citrate, 1 × trace metals solution [5 mM EDTA, 2.5 M FeSO₄, 1 mM MnCl_2, 1 mM ZnSO_4, 0.3 mM MgSO₄, 0.3 mM CaCl₂, 0.03 mM MgSO₄]) containing 7% χ1666 E. coli and incubated on an orbital shaker at 15° for ~6 days until the majority of worms in the culture were young adults full of embryos. The eggs in glp-4 adults were harvested by treatment with hypochlorite and grown until the adult stage at 26°. glp-4 26° adults were visually inspected for clear gonads, which indicates absence of a germline. All worms from a particular strain were combined and bacteria were cleared from the worms’ cuticles and intestines by shaking in M9 for 45 min. This was followed by a wash in 1 × lysis buffer (50 mM HEPES pH 7.5, 1.0 mM EGTA, 3.0 mM MgCl₂, 100 mM KCl, 10% glycerol), with slurries of 1:1 volume of lysis buffer/worms stored in aliquots at −80°. Approximately 10,000 dcr-1(ok427) homozygous (non-green) worms were hand-picked and frozen in M9 and 50 × protease inhibitor (Complete™, mini, EDTA-free (Roche)). Multiple aliquots of dcr-1 homozygous worms, collected over a period of time, were combined and RNA was isolated.

2.3. RNA isolation

Frozen worm lysate was thawed on ice and re-suspended in ~3 volumes of TRI-Reagent (Ambion, Inc). The TRI-Reagent/worm slurry was ground to a fine powder in liquid nitrogen and sonicated. After clearing insoluble debris, lysates were treated with 0.2 volume of chloroform, and the aqueous phase precipitated with equal volumes of isopropanol at −20°. RNA pellets were concentrated by centrifugation, and were stored in 10 mM Tris–HCl, pH7.5. RNA concentrations were determined with a Nanodrop ND1000 spectrophotometer; yields ranged from 2 to 5 μg/μL.

2.4. Small RNA enrichment

Small RNA fractions were enriched from glp-4, glh-1, N2, and dcr-1 total RNA samples using the mirVana™ (Ambion, Inc) small RNA isolation protocol, with slight modifications as in (Gu et al., 2009). 5 μg aliquots were resolved on 2% agarose, MOPS-formaldehyde gels to visualize the RNAs prior to cDNA library construction and TaqMan^® miRNA Assays.

2.5. Small RNA library construction and deep sequencing

The construction of cDNA libraries from small RNA templates and subsequent sequencing on the Illumina HiSeq-2000 was carried out by University of Missouri DNA Core Facility staff. Libraries were constructed from small RNAs from glp-4 and glh-1 mutant and N2 worms grown at 15°, 20° or 26°, and dcr-1 mutants grown at 15°. For each strain and each tested temperature two cDNA libraries, each representing a unique biological replicate, were prepared with the Illumina^® TruSeq™ Small RNA Sample Preparation kit as per the manufacturer's protocol. Each library was made from 2.5 μg of mir-Vana™-enriched small RNA and was individually identified with a short specific linker/adapter. The quality of construction was determined with an Agilent BioAnalyzer 2100 Chip. A plot illustrating this assay is shown in Supplemental Fig. S1. The cDNA libraries were sequenced using Illumina's HiSeq2000.

2.6. Bioinformatics

Sequencing data are available at NCBI in the Bioproject database with accession number PRJNA322183 (http://www.ncbi.nlm.nih.gov/bioproject/PRJNA322183). Data were generated by Illumina sequencing, with reads cleaned to remove low quality bases and trimmed to remove the adaptors. All reads were aligned against the annotated, non-coding RNA transcripts downloaded from Wormbase ftp://ftp.wormbase.org/pub/wormbase/releases/WS242/species/c_elegans/PRJNA13758/c_elegans. PRJNA13758. WS242.ncrna_transcripts.fa.gz, which contains annotations for 223 miRNAs, 345 snoRNAs, 126 snRNA, 176 lincRNAs, 7980 ncRNAs, 15,365 piRNAs and other RNAs, including tRNAs, and rRNAs. The alignment from the above processed short reads to the non-coding transcripts was conducted using BWA, Burrows-Wheeler alignment (Li and Durbin, 2010). To obtain the counts for each miRNA, the HTSeq protocol (Anders et al., 2014) was applied to the mapping results, which are found in Supplemental Table S1.

Statistical tests comparing differential abundance levels between the miRNAs of various populations were carried out using DESeq2, a statistical analysis software available as an R/Bioconductor package (Love et al., 2014). DESeq2 software was used because it reports consistent results even for experiments with small number of replicates (Seyednasrollah et al., 2015). For our sequencing results DESeq2 provided much more conservative estimates of statistical significance than did the software package edgeR (Robinson et al., 2010). The adjusted p-value (q-value) of <0.05 was used as the cutoff for differential expression with statistical significance. q-values are adjusted p-values, used with the Benjamini-Hochberg procedure (Benjamini and Hochberg, 1995). http://www.stat.purdue.edu/~doerge/BIOINFORM.D/FALL06/Benjamini%20and%20Y%20FDR.pdf.

2.7. TaqMan^® miRNA assays

qRT PCR was used to verify the RNAseq results for several candidates. Levels of miR-38, −40, −228, and miR-244 were analyzed in glp-4, glh-1, dcr-1 and wild type worms with TaqMan^® MicroRNA Assays (Applied Biosystems). Quantitative PCR (qPCR) was performed in most cases with three independent biological samples, using triplicate technical replicates of each sample. Template cDNA specific for each miRNA was generated with the TaqMan^® MicroRNA Reverse Transcription (RT) Kit as per the manufacturer's protocol. 10 ng of glp-4 (15° and 26°), glh-1 (26°), N2 (26° and 20°) or dcr-1 (15°) mirVana™-enriched small RNA were used as the starting material for each 15 μL RT reaction. Products from each RT reaction were used as template for qPCR using TaqMan^® Universal PCR Master Mix II (no UNG), with primers supplied by the manufacturer on a 7900HT Fast Real-Time PCR System (Applied Biosystems). Cycling conditions were: 95° for 10 min, and 40 cycles of 95° for 15 s and 60° for 1 min. Relative expression levels were determined according to (Pfaffl, 2001) and the delta-delta CT method (Schmittgen and Livak, 2008), using U18, a small nucleolar RNA transcript, as an endogenous control. Statistical analyses for these qPCRs were performed using T-tests, with the Relative Expression Software Tool (REST) 2009 (www.gene-quantification.de/rest.html).

2.8. in situ hybridizations

in situ hybridizations were carried out using modifications of the protocols reported by Seydoux and Fire (1994) for detecting messenger RNAs in C. elegans and by Pena et al. (2009) for localizing miRNAs in mouse tissue sections. Because the protocol is a new combination of others, it will be described in some detail here. C. elegans wild type and mutant worms were grown at 15°, 20° or 26 °C either to a young adult stage, when worms first begin producing embryos (~1 day beyond the L4, fourth larval stage), or to a later adult stage, at 2–3 days beyond L4, when worms are fully gravid and the plates contain many embryos and newly-hatched larvae. Worms were placed in M9 buffer containing 1 mM levamisole, an anesthetic, and splayed onto poly-lysine-treated slides to provide access for the miRNA probes to the internal tissues. The slides were placed on dry ice to freeze-crack the adult tissues and embryos, using the light pressure of a coverslip. Slides were stored in 100% methanol at −20° for as little as 10 min, or for up to two weeks, before further fixation, hybridization and probe detection. After an additional 5 min in 100% methanol at RT, slides were rehydrated through an alcohol series of 90%, 70% and 50% methanol in TBS (150 mM NaCl, 50 mM Tris-HCl, pH7.5) for 1 min each, and were washed twice in TBS. All subsequent washes were in 1x TBS or TBS-T (TBS plus 0.1% Tween-20) for 5 min, all solutions were made with DEPC (diethylpyrocarbonate)-treated water and incubations were at RT unless noted otherwise. Throughout the fixation process TBS was used instead of PBS to avoid adding phosphate to the tissue, which would interfere with the EDC treatments to follow. The fixed tissue was then treated with 5 μg/ml, a mild dose of proteinase K (Roche) in 1x TBS for 15–20 min at RT, followed by exposure to 2 mg/ml glycine in TBS, and two TBS washes. An additional fixation for 20 min in 3.2% formaldehyde in a buffer of 80 mM HEPES, pH6.9, 16 mM MgSO₄, and 8 mM EGTA in TBS was followed by repeating the above glycine and wash steps. After the formaldehyde fixation, we followed the protocol of Pena et al., (2009) (Supplementary methods; Nat. Methods: doi.1038/nmeth.1294) using EDC (1-ethyl-3-(3-dimethylaminopropyl) carbodiimide), which significantly increases the levels of miRNAs remaining in the tissues by creating phosphoramidate linkages on the 5′ phosphates of the small RNAs. To first remove extraneous residual phosphates, slides were incubated twice, for 10 min, in 0.13 M 1-methylimidazole, 300 mM NaCl, pH8.0. Thereafter EDC was added to a concentration of 0.16 M into fresh 1-methylimidazole solution, placed on each slide and incubated for 2 h, followed by a third glycine treatment and two TBS washes. While the use of EDC was critical to our detection of a colorimetric signal, rather than storing EDC under argon, we purchased fresh aliquots of anhydrous EDC (Life Technologies) and brought this reagent up in the imidazole solution when needed for the in situ protocol. We also did not retain the harsh step of blocking all endogenous enzymatic activity with 0.1 M triethanolamine and 0.5% acetic anhydride that followed the EDC treatment (Pena et al., 2009).

After the EDC fixation, the slides were pre-hybridized for 2 h in a standard hybridization solution of 50% formamide, 5 × SSC, 5 × Denhardt's solution, 250 μg/ml yeast tRNA, 500 μg/ml herring DNA, 2% Blocking Reagent (Roche), and 0.1% CHAPS and 0.5% Tween-20 detergents. The miRNA probes tested were LNAs, locked nucleic acids, which are nuclease resistant; the LNAs we used were complementary to miR-35, miR-38 and miR-40, let-7 and miR-228. Controls included an LNA corresponding to the C. elegans pre-SL-1, the nuclear precursor of spliced leader-1, a small ncRNA involved in trans-splicing (Krause and Hirsh, 1987), as a positive control, and as a negative control, a scrambled LNA that does not correspond to any known C. elegans miRNA sequence; the LNAs were pre-labeled on both 5′ and 3′ ends with Digoxigenin (Exiqon). The MT14119 miR-35–41 deletion strain was used as an additional negative control. Sequences of the LNA probes are found in Supplemental Table S2. Most probes were used for hybridization at a final concentration of 25 nM; however, when combined, the miR-35, −38 and −40 LNAs were each at 8 nM, while miR-228 was at 50 nM. Probes were briefly heat denatured at 68° before added to the hybridization solution. Slides were covered with parafilm and hybridized in a humidified chamber for 13–16 h at 48 °C, a temperature > 20° below the T_m of any of the LNAs used (57° was also successful). After hybridization, 5x SSC was used to float off the parafilm and slides were washed twice for 30 min each in 1 × SSC; 50% formamide, 0.1% Tween-20 at 48°, then washed at RT in 0.2 × SSC for 15 min and finally in TBS-T. This was followed by blocking for endogenous peroxidase activity, with 3% hydrogen peroxide in TBS-T for 30 min, followed by three 1 min TBS-T washes.

To detect the hybridized probes, a antibody/amplification/antibody process was used. Slides were first blocked for 1 h at RT in 0.5% Blocking Reagent (Roche), with 10% heat-inactivated goat serum in TBS-T, followed by the addition of an anti-DIG-FAB peroxidase (POD, Roche) antibody diluted 1:500 in the above blocking solution and incubated for 1h at RT. Amplification of the peroxidase conjugate on the anti-DIG antibody was then achieved with the addition of the tyramide reagent that catalyzes deposits of dinityrophenyl (DNP) (TSA kit, PerkinElmer Renaissance^® TSA™Plus DNP system NEL 747A). First, slides were washed 3 times in TNT buffer (0.1 M Tris-HCL, pH7.5, 0.15 M NaCl, 0.1% Tween-20) and then the tryamide solution was applied and incubated for 30 min, after which slides were washed 3 times in maleate buffer (0.09 M maleic acid, 0.175 M NaOH, 1 M NaCl, and 0.5% Tween-20, pH7.5). The DNP labels were subsequently detected by a second antibody reaction, with an anti-DNP antibody conjugated to Alkaline Phosphatase (AP) diluted 1:500 in the maleate buffer, supplemented with 10% blocking reagent (Roche); this reaction went for 40 min. Subsequent washes were twice in maleate buffer and four times in TMN buffer (0.1 M Tris-base, pH9.5, 0.05 M MgCl₂, 0.5 M NaCl, 0.05% Tween-20, and 2 mM tetramisole hybrochloride). These steps were as per manufacturer's directions.

The detection of signal followed using a colorimetric reaction with a BCIP/NBT solution (Roche Ready Mix tablets) that was allowed to proceed for 30 min, to allow an initial deposition of blue pigment, after which slides were incubated in KTBT (50 mM Tris-HCl, pH7.5, 150 mM NaCl, 10 mM KCl, with 1% Tween-20, a notably high detergent concentration) at 4° for 1–14 days and a second color reaction was carried out for an additional 30 min or more, until a strong signal was seen or the background became visible. This two-step color reaction, with KTBT buffer, was adapted from the GEISHA, chick protocols (Darnell et al., 2006) and made for a clearer, stronger signal. After completing the color reaction, slides were washed three times in a penultimate wash of 0.01 M Tris-HCL, pH7.5, 0.5 M NaCl. 5 mM EDTA, 0.05% Tween-20, followed by a 10 min fixation in a 4% aqueous paraformaldehyde solution, ending with a wash in water. Cover slips were applied using Vectashield mounting medium with DAPI (Vector Laboratories) and sealed with clear nail polish. Images were taken on a Zeiss Auxio Z1 microscope with SlideBox software and processed with Photoshop and Illustrator software. While the protocol used for these experiments was successful in repeated sets of experiments, using a single miR35- member probe was not effective. Future optimization of these in situ techniques for C. elegans may increase the sensitivity of miRNA detection.

3. Results

3.1. Comparing the small RNA transcriptomes of germline-defective C. elegans

To determine if any C. elegans miRNAs are enriched in the germline, we conducted deep sequencing on 14 cDNA libraries prepared from small RNA populations isolated from wild-type (N2) worms, grown at 20 °C and 26 °C, as well as from three germline-defective strains: glp-4(bn2) at 15 °C and 26 °C, glh-1(gk100) at 20 °C and 26 °C, and dcr-1(ok247) homozygous worms raised at 15 °C used as a control. Each library was constructed from a unique biological replicate. glp-4(bn2) is a mis-sense, partial loss of function mutation in vars-2 (Rastogi et al., 2015), a gene coding for valine aminoacyl tRNA synthetase whose mRNA is highly enriched in the germline. glp-4(bn2) worms are fertile when grown at 15 °C, and are sterile when grown at the restrictive temperatures of 25–26 °C, with each hermaphrodite producing only ~12 germ cells compared to ~2000 in wild type (Beanan and Strome, 1992). glh1(gk100) is a non-null deletion allele in glh-1, a germline-specific RNA helicase component of P granules (Gruidl et al., 1996; Spike et al., 2008; Strome and Wood, 1983; Sheth et al., 2010). The truncated glh-1(gk100) product is expressed at ~10% of wild-type levels. glh-1(gk100) animals are fertile at 20 °C and 100% sterile at 26 °C, with under-proliferated germlines 1/5th-1/2 the normal size (Spike et al., 2008). The glh-1 mutant was chosen for this study in part because GLH-1 is a binding partner of Dicer and therefore GLH-1 might have a role in processing miRNAs or other Dicer-dependent small RNAs (Beshore et al., 2011). The dcr-1(ok247) strain is a null mutation in the gene coding for the Dicer protein. dcr-1(ok247)hermaphrodites have fully developed germlines, but produce abnormal, endo-replicating germ cells (Ketting et al., 2001; Knight and Bass, 2001). The dcr-1(ok247) strain is balanced with a GFP-tagged chromosome and non-green, sterile, dcr-1homozygous adults were hand-picked to generate small RNAs for libraries and qRT-PCRs. All libraries were constructed using a 5′ ligation-dependent protocol that is specific for mono-phosphorylated species, which include micro-, piwi(pi)- and 26G RNAs; each library was individually tagged.

Deep sequencing of the multiple cDNA libraries generated from small RNA templates produced a total of 107,728,496 raw reads, 21,045,231 of which uniquely aligned to annotated, transcribed loci in the C. elegans genome. Of the uniquely aligned sequences, 5,064,689 reads mapped to 93 of the 223 annotated miRNA genes in WormBase.

ftp://ftp.wormbase.org/pub/wormbase/releases/WS242/species/c_elegans/PRJNA13758/c_elegans.PRJNA13758.WS242.ncrna_transcripts.fa.gz. In addition 62,364 reads mapped to 96 pre-miRNA genes, and 486,561 reads mapped to piRNAs. The reads mapping to other small RNAs are illustrated in Supplemental Fig. S2. Through use of the analysis pipeline of BWA and HTseq protocols (Li and Durbin, 2010; Anders et al., 2014), the raw mature miRNAs reads detected in all 14 libraries were determined; however, mature miRNA reads in the two dcr-1 libraries were rare and are not included with the other 12 populations seen in Supplemental Table S1. Subsequently, the DESeq2 software program was used to examine the differentially-abundant miRNAs based directly on the read counts (Love et al., 2014). Both biological and technical variations were accommodated. We considered miRNAs to be differentially expressed using the q-value of <0.05 as the cutoff for statistical significance.

3.2. Several miRNAs are significantly reduced when glp-4 worms lack a germline

Table 1A compares miRNA levels in glp-4 mutants grown at 15° (fertile) versus 26° (sterile). Among the 93 miRNAs detected by our RNA-seq, we identified 21 miRNAs whose relative levels significantly decrease, and eight whose relative levels increase in sterile glp-4 worms compared to fertile ones (Table 1A, Fig. 1A and Supplemental Table S3A). Because glp-4 mutants raised at 26° consist mainly of somatic tissues, the former category likely correspond to miRNAs expressed preferentially in the germline, whereas the latter are likely to correspond to miRNAs expressed preferentially in the soma.

Table 1. Statistical analyses of sterile germline-defective mutants compared to fertile worms.

These tables show the miRNAs that significantly change in each comparison. A The DESeq2 analysis of glp-415° miRNAs compared to glp-4 26°. B DESeq2 calls of significance for glp-4 26° vs. N2 26° miRNAs. C The DESeq2 analysis for glh-1 26° vs. N2 26° miRNAs. To view the results of statistical analyses for all detectable miRNAs see Supplemental Table S3A–C.

	log2 fold change	Adj. P-value
A. glp-4 (bn2) 26C vs 15C
miR-4813	−4.89605	9.30E–08
miR-260	−6.36403	1.21E–07
miR-39	−6.26213	2.89E–07
miR-38	−7.40438	1.29E–06
miR-248	2.35392	2.35E–06
miR-35	−6.45842	0.0001017
miR-37	−6.21010	0.000252791
miR-70	−1.64195	0.000267552
miR-238	1.41267	0.001948051
miR-239.1	2.16398	0.003155295
miR-2212	−1.83031	0.004223912
miR-262	−4.15019	0.004887495
miR-240	−1.62272	0.004995897
miR-2214	1.23544	0.00714549
miR-244	−1.09378	0.00714549
miR-42	−2.73965	0.00714549
miR-73	−1.53514	0.00714549
miR-1830	−3.47870	0.00930712
miR-56	−1.24676	0.00930712
miR-1022	−1.38437	0.009435131
miR-53	1.21911	0.014087703
miR-40	−3.66994	0.015520947
miR-36	−4.62137	0.015631642
miR-60	1.63146	0.018428118
miR-4931	−4.38697	0.024686066
miR-71	1.08388	0.024686066
miR-788	−1.80748	0.02719458
miR-4916	1.80619	0.045959487
miR-1832.1	−2.46817	0.046651332
B.glp-4 (bn2) 26C vs N2 26C
miR-248	2.95717	3.85E–12
miR-38	−6.37337	1.95E–11
miR-4813	−4.59489	2.43E–10
miR-70	−2.20808	4.83E–10
miR-39	−5.25596	3.82E–09
miR-37	−5.55568	2.81E–08
miR-35	−5.48695	4.64E–08
miR-56	−1.79510	3.25E–07
miR-788	−2.79181	1.22E–06
miR-1829.3	−2.14947	0.00026497
miR-249	1.68721	0.00029823
miR-260	−3.76763	0.00029823
miR-71	1.35860	0.00035472
miR-239.1	1.96080	0.0006414
miR-238	1.15484	0.00139564
miR-40	−3.35445	0.00226176
miR-42	−2.38211	0.00835347
miR-53	0.88166	0.00835347
miR-58	0.92226	0.00835347
miR-244	−0.85106	0.01037996
miR-43	−1.86624	0.01037996
miR-36	−3.16671	0.01530756
miR-72	0.80125	0.02515828
miR-65	−0.70855	0.03634389
miR-61	−2.13195	0.03755028
miR-60	1.19401	0.04133958
C. glh-1 (gk100) 26C vs N2 26C
miR-244	−1.32081	1.44E–06
miR-4813	−2.32621	1.24E–05
miR-72	0.95901	0.00615274
miR-58	0.91224	0.01032463
miR-237	−0.91340	0.01035653
miR-56	−0.96261	0.0113635
miR-64	0.69839	0.0113635
miR-788	−1.56113	0.0113635
miR-51	1.08469	0.01302394
miR-85	0.68831	0.01302394

Fig. 1. Multiple miRNAs are affected by loss of the germline.

A The miRNAs that significantly change when sterile glp-4 worms grown at 26° are compared to fertile glp-4 worms grown at the permissive temperature of 15° on this volcano plot produced with Prism software. All 93 miRNAs detected in our adult populations are plotted here (black dots). The 21 miRNAs that decrease when the germline is missing are in the left quadrant; seven members of the miR-35 family are indicated in red. In addition, eight miRNAs, seen in the right quadrant, significantly increase in the glp-4 miRNA population when the germline is missing. The 64 miRNAs that showed no change, or changes not meeting the criteria for significance, are clustered in the center of the A plot below the dotted horizontal cutoff line which indicates a FDR(False Discovery Rate), using the q-value of <0.05 for significance. The data plotted in figure is presented in Table 1A–C, for the significantly changing miRNAs and in Supplemental Table S3A–C for all 93 miRNAs detected. B This volcano plot, similar to that in 1A, compares abundance of glp-4 26° miRNAs with those of N2 worms grown at the same temperature. C qRT-PCR verifies the relative abundance of miR-38, miR-40, miR-228 and miR-244. Levels of miR-38,-40,-228 and −244 were assayed using both biological and technical triplicates by qRT-PCR comparing small RNAs from glp-4 26° sterile worms to those of glp-4 15° fertile worms (panel 1); glh-1 26° sterile mutants compared to N2s at 26° (panel 2); and homozygous, sterile dcr-1 15° animals vs. N2s at 20° (panel 3). U18 was used as the endogenous control for each condition and strain tested. The statistical significance of differences in relative levels of miRNAs was determined using the Relative Expression Software Tool (REST) 2009 (www.gene-quantification.de/rest.html). The brackets with an asterisk in C indicate the p-value is statistically significant (p-value <0.0001). Numerical results are given in Table 2.

Since higher temperatures are known to adversely affect the fertility of wild type animals (Spike et al., 2008), we compared N2 worms grown at the optimal growth temperature of 20° to the more restrictive temperature, 26° and found no statistically significant differences among the 96 miRs detected by our RNAseq and the subsequent DESeq1 analysis, Supplemental Table S4.

We then went on to compare glp-4 mutants grown at 26° to wild-type worms grown at 26° (Table 1B, and Fig. 1B). This comparison yielded 26 significantly mis-regulated mRNAs, including 17 down-regulated and nine up-regulated miRNAs. The gene overlap between the two glp-4 comparisons (Table 1A-B and Fig. 1A-B) was 77% (13/17) for the down-regulated category and 56% (5/9) for the up-regulated category. Among the former, are 13 miRNAs including all seven miR35 family members and six other unrelated miRNAs, miR-56, −70, −244, −260, −788 and −4813, which we consider the “best candidates” for being germline-enriched. With regard to the down-regulated miRNAs that don't overlap in the two comparisons, we predict the four miRNAs only reduced in the comparison of glp-4 26° versus wild type animals grown at 26° may reflect specific effects of high temperature on the defective valine tRNA synthetase in the glp-4(bn2) animals, as three of these four miRNAs (miR- 43,−61 and 1829.3) are not reduced in wild type 26° animals, Supplemental Table S4.

One of the original aims of this research was to determine whether GLH-1, a constitutive P-granule component and a binding partner of Dicer (Beshore et al., 2011) affects germline miRNA levels. Therefore, we compared sterile glh-1 mutants grown at 26° to fertile N2 worms grown at 26°. However, RNA-seq with sterile glh-1 worms did not indicate all 13 of the germline-enriched miRNAs are statistically reduced, Table 1C. Four of the five reduced miRNAs, miR-56, −244, −788 and −4813, are among the germline-enriched miRNAs identified in the glp-4 analyses presented above, Table 1A-B. While these findings are consistent with sterile glh-1 mutants having tiny germlines and fewer germ cells (Spike et al., 2008), seven other germline candidates were reduced, but not significantly, and two showed no reductions, Supplemental Table S3C. We also examined the levels of precursor miRNAs in glh-1 mutants and did not find them obviously higher than those of N2 worms, in contrast to the accumulation of precursor miRNAs in dcr-1 worms, Supplemental Table S5. Therefore, our RNA-seq analysis is inconclusive as to whether this glh-1 strain affects the Dicer-dependent processing of all germline miRNAs. Perhaps too few replicates were tested or the limited response is due to this strongest glh-1 mutant not being a null. The GLH-1/DCR-1 germline complex might also function with the C. elegans 22Gs, endo-siRNAs that are critical in the germline for genome integrity and to repress selected protein coding sequences (Gu et al. 2009). In mice oocytes endo-siRNAs are essential for meiotic maturation (Stein et al. 2015). While the 22Gs are, like miRNAs, Dicer-dependent, they are not mono-phosphorylated, and therefore are not in our libraries. Thus, the basis of the DCR-1/GLH-1 relationship (Beshore et al., 2011) remains to be determined.

3.3. Verification of the RNA-seq results using qRT-PCR

To independently verify some of the changes observed by RNA-seq, we carried out qRT-PCR for four miRNAs, three predicted to be preferentially expressed in the germline based on the RNA-seq data and one, miR-228, predicted to be expressed both in the soma (Pierce et al., 2008) and germline (Table 1A–C). As many as nine replicate TaqMan qRT-PCRs compared: (i). glp-4 26° levels against glp-4 15° worms, (ii). glh-1 worms against N2s, both grown at 26°, and as a control (iii). dcr-1, raised at 15°, vs. fertile N2 adults grown at 20°, Fig. 1C and Table 2. Quantification of the qRT-PCR results for glp-4 confirmed the RNA-seq results, panel 1. We observed the most dramatic changes for the two miR35 family members tested, with reductions as high as 1000-fold for miR-35 in glp-4 sterile mutants, Fig. 1C, panel 1 and Table 2. When fertile N2 worms were tested against sterile glh-1 animals, all four probes were significantly reduced in the qRT PCR assays, Fig. 1C, panel 2 and Table 2. Perhaps unexpectedly, both miR-38 and −40, which were not significantly reduced in the DESeq2 analysis of sterile glh-1 worms, Supplemental Table S3C, were significantly reduced with the increased numbers of biological and technical replicates used for qRT PCR, Fig. 1C, panel 2, and Table 2, further supporting the germline-enrichment of the miR35s. As expected, the dcr-1 RNA showed significant reductions for all miRNAs tested, Fig. 1C, panel 3 and Table 2. However, the markedly greater fold changes for the miR35 probes in the germcell-less glp-4 animals than seen for dcr-1 null mutants might be unexpected, Table 2. We note that dcr-1 animals have large, albeit abnormal, germlines, and therefore, could retain higher levels of stable, maternal miRNAs than the germline-minus glp-4 animals. Therefore, we propose this result is likely due to the perdurance of maternal miRNAs in the dcr-1 mutant, a suggestion also made by Drake et al. (2014) when studying Dicer-1 and observing “only modest differences in bulk miRNA levels between WT and dcr-1(0) animals”.

Table 2.

Relative abundance of mature miRNAs in TaqMan^® miRNA assays. qRT-PCR was utilized to assay miR-38, miR-40, miR-228, and miR-244 levels in germline-minus glp-4 26° compared to fertile glp-4 animals @15°, and in glh-1 worms compared to N2s, both grown at 26°. Levels of these miRNAs were also measured in dcr-1 null mutants, where they were expected to decrease. U18 was used as the endogenous control for each condition and strain tested. Statistical analysis with REST software (REST 2009) indicated that all three germline miRNAs tested were significantly decreased in comparisons at 26° of glp-4 and glh-1 worms, and in the 15° dcr-1 mutant.

glp-4 (bn2)at 26C vs glp-4 at 15C
miRNA	Abundance	Std. Error	P-value	Result
miR-38	0.001	0–0.006	< 0.0001	1000-fold down
miR-40	0.003	0.001–0.009	< 0.0001	333.3-fold down
miR-228	0.377	0.190–0.838	< 0.0001	2.7-fold down
miR-244	0.242	0.093–0.601	< 0.0001	4.1-fold down
glh-1(gk100) at 26C vs N2 at 26C
miRNA	Abundance	Std. Error	P-value	Result
miR-38	0.284	0.165–0.517	< 0.0001	3.5-fold down
miR-40	0.135	0.012–0.576	< 0.0001	8.0-fold down
miR-228	0.319	0.128–0.598	< 0.0001	3.1-fold down
miR-244	0.173	0.062–0.326	< 0.0001	5.8-fold down
dcr-1(ok247) at 15C vs N2C
miRNA	Abundance	Std. Error	P-value	Result
miR-38	0.065	0.045–0.114	< 0.0001	15.4-fold down
miR-40	0.018	0.010–0.027	< 0.0001	55.6-fold down
miR-228	0.066	0.043–0.106	< 0.0001	15.2-fold down
miR-244	0.017	0.007–0.042	< 0.0001	58.8-fold down

3.4. in situs indicate miR35s are highly enriched in the adult germline

To visualize the localization of miRNAs in C. elegans, we developed an in situ hybridization protocol for miRNAs based on techniques reported for chick and mouse miRNAs (Darnell et al., 2006; Pena et al., 2009) as detailed in Section 2.

Hybridizations were carried out on adult C. elegans worms that had been splayed open to release their gonads, partially removing the germline (red arrows), as well as the intestinal tissue (black arrows), away from the body wall cavity, Fig. 2. A cartoon of the two symmetric gonad arms as they appear in an intact adult worm illustrates the normal developing germline, Fig. 2A. In Fig. 2B, strong hybridization is seen in the meiotic pachytene region of the gonad where oogenic and maternal RNAs are abundantly transcribed (red arrow), when a combined LNA probe complementary to the mature miRNAs corresponding to miR-35, −38 and −40 was used. In Fig. 2C, the miR35s are also detected with the combined miR35 probe in the oocytes and embryos (white arrow) of this older adult. A five-fold extension of the color reaction time using two probes, miR-38 and −40, revealed staining throughout both completely-removed gonad arms, beginning in the distal region (red arrow) in this young adult animal, Fig. 2D. Even with this extended exposure, no signal was detected in the adjacent, isolated intestinal tissue (or in other somatic tissues), Fig. 2D, despite previous studies with a miR35 promoter-driven GFP transgene indicating GFP expression in the vulva, several anterior neurons and the rectum (Martinez et al., 2008). This difference in findings may reflect a lack of sensitivity of the current in situ protocol or a failure to penetrate the hard cuticle of the adult body wall with the relatively mild treatments used for fixation, Section 2. A DAPI image of Fig. 2D is seen in Fig. 2E. As negative controls, the mutant strain MT14119 miR-35-41(nDf50) was tested, Fig. 2F, with only background hybridization detected with the combined miR35 probe, while a scrambled LNA also produced no signal, Fig. 2G. Both negative controls were carried out for ~1 h, the same exposure used for Figs. 2B-C, and 2H-I, and for the positive SL-1 probe, which revealed concentrated nuclear staining for the SL-1 precursor, Supplemental Fig. S3, in contrast to the uniform, cytoplasmic staining of the mature miRNAs seen here, Fig. 2B-D and H-I. In addition, a let-7 LNA probe produced no signal in N2 adult tissues or wild type embryos (not shown), an expected result based on the rarity of let-7 in our libraries, Supplemental Table S1. However, surprisingly, we did detect let-7 hybridization in the abnormal embryos of the miR-35-41 deletion strain, Fig. 2H; this staining appears to reveal inappropriate embryonic expression of let-7 when the miR35s are missing. This could relate to findings in (Akay et al., 2013) that both miR35 and let7 families genetically interact with GLD-1, a well-studied RNA binding protein. While Brenner and Schedl (2016) present convincing evidence that GLD-1 is not a target of the miR35s, we suggest GLD-1 may indirectly regulate these two miRNA families and that when one miRNA family is missing, the other increases in expression. In Fig. 2I, in contrast to the exclusive germline location of the miR35s seen by in situ hybridizations in Fig. 2B-D, we detected strong signals in both the intestine and in the distal germline with a miR-228 probe. As the C. elegans miR-228 (miR-183 in mammals) was previously seen in multiple somatic tissues by GFP expression (Pierce et al., 2008) and based on our RNA-seq and qRT-PCR results, this dual localization was not unexpected. In summary, these in situ hybridizations are consistent with our RNA-seq results, Table 1A-B, and provide the first visualization of the miR35 microRNAs as highly enriched the adult germline, concentrating in the pachytene region and in maturing oocytes; they also reveal expected localizations for miR-228 and both expected and unexpected patterns for let-7.

Fig. 2. In situ hybridizations indicate miR35 miRNAs are exclusively detected in the adult germline, while miR-228 is present in both germline and somatic tissue.

A provides an illustration of the C. elegans adult bi-lobed gonad from (Smith et al., 2002) revealing the progression of germline development from the mitotic region in the distal gonad (D) beginning at the distal tip cell (DTC), with germ cells transitioning into meiosis (trans. zone) with oogenesis (pachytene and diakinesis) producing mature oocytes (O) in the proximal gonad (P) which contains mature sperm (S) in the spermatheca. After fertilization the embryos (E) develop and are later released, at about the 40-cell stage, through the vulva (V). In B-I and in Supplemental Fig. S3A several C. elegans miRNAs were localized by in situ hybridization to adult worm tissues using LNA probes complementary to miRs-35,−38 and −40, miR-228, a Scrambled LNA (a negative control), let-7, and to an LNA complementary to pre-SL-1 (a positive control). In each case, unless otherwise indicated, the total concentration of individual or combined probes did not exceed 25 nM and all worms were wild type (N2). All images were taken at the same 100X magnification.B is a whole hermaphrodite adult splayed open to expose the intestine and the germline. The worm was hybridized with a combination of miR-35, −38,−40 probes, revealing a signal concentrated in the meiotic pachytene region (the area of bend in the gonad, red arrow), with no signal seen in the gut (black arrow). In all images in this figure and Supplemental Fig. S3C the germline is marked with red, somatic tissue with black, and embryos with white arrows. Unless otherwise indicated, worms were young adults 1–2 days beyond the L4 stage (>L4) and grown at 15°, because the lower growth temperature extends the lifespan and a “young” adult worm state (Kato et al., 2011). The color reactions were carried out in two steps for a total of ~60 min, unless otherwise indicated. C shows an older adult (2–3 days >L4) also hybridized with the three miR-35 family LNAs and again the signal is concentrated in the mid-section of the germline tissue, in the meiotic, pachytene region (red arrow). In this image the combined miR-35s probe also hybridizes throughout all the cells in the extruded embryos (white arrows) of this older adult. D illustrates the use of two miR35 family probes, miR-38 and −40, on the two gonad arms and an intestinal tract that were completely removed from a young N2 adult grown at 26°. Here the final color reaction was allowed to proceed for 5.5 h. It is clear that the positive signal remains concentrated in the germline, although it is now throughout the two separated arms of the bi-lobed gonad (red arrows) and remains absent from the somatic intestine (black arrow). An asterisk marks the extruded pharynx. This longer exposure reveals that miR35 miRNAs, while abundant and concentrated at the pachytene stage of meiosis, are also present at lower levels throughout the germline, including the proliferative distal region (red arrows). E is the DAPI image of panel D. F shows a MT14119 miR-35-41 worm raised at 15°, splayed and hybridized with the miR-35, −38, and −40 combination. While this worm is negative, in some cases the deletion mutant shows weak hybridization, perhaps due the match of the LNAs to the “seed” sequence of miR-42, a family member not deleted in this strain. In G a Scrambled LNA, a second negative control, was used for the hybridization; this probe consistently produced no signal. H shows the let-7 LNA on an MT14119 miR-35-41 older adult grown at 26°. As with N2 adult worms (not shown), the let-7 LNA does not hybridize to adult tissue. At this non-permissive temperature the miR-35-41 deletion strain retains aberrantly developing embryos and it appears the let-7 probe uniformly hybridizes throughout these embryos (white arrows). I shows an in situ hybridization of miR-228 on an N2 worm grown at 26°. miR-228 localizes in the adult germline, especially at the distal region of the gonad (red arrow) and is also detected in the somatic gut (black arrow).

4. Discussion

We have compared the miRNA transcriptomes of wild-type and germline-defective worms and have identified 13 miRNAs that are preferentially expressed in the germline. These include seven of the eight members of the miR35 family, which appear highly enriched in germ cells. In using our newly developed in situ hybridization protocol to detect miRNAs in C. elegans tissues, the miR-35-41 miRNAs appear exclusive to the germline of adult hermaphrodites. These miRNAs are first detected during oogenesis and accumulate at high levels in oocytes and early embryos. Our findings are consistent with the genetic findings of (Alvarez-Saavedra and Horvitz, 2010) who determined that deletion of seven of the eight miR35 family members causes temperature-sensitive, maternal-effect and zygotic lethality, resulting in embryonic or L1 larval arrest, while loss of all members is lethal at all temperatures.

Since many miRNAs have been individually deleted (Miska, et al., 2007), the phenotypes resulting from the loss of some of the germline miRNAs identified here have already have been reported. As described above, deletion of all miR35 family members results in embryonic lethality; however adding back maternal expression of the miR35–41 gene cluster or of each individual miR35 member, results in normal embryonic development (Alvarez-Saavedra and Horvitz, 2010), implying redundancy of the miR35s and a vital role for this family in embryonic development. In addition, McJunkin and Ambros (2014) have demonstrated that maternal, along with zygotic, miR35s act early in development to insure maximal fertility of the adult progeny, while Brenner and Schedl (2016) quantified the marked reduction in proliferation in the C. elegans germline with loss of all miR35 family members. In the miR51 family, miR-56 is the only member we identified as germline-enriched. Deletion of all six miR51 genes results in dead embryos, with any individual miR51 member able to rescue the embryonic lethality (Shaw et al., 2010); however, a specific role for miR-56 in the adult germline had not been reported. There are no deletions reported for miR-788 and −4813; these two more recently-discovered miRNAs were not a part of the large deletion study by the Horvitz group (Miska et al., 2007). There are individual deletions of miR-70, −244 and −260, which revealed no gross abnormalities, supporting the conclusions of others, including (Miska et al., 2007; Alvarez-Saavedra and Horvitz, 2010) that the normal regulatory functions of most individual miRNAs are likely subtle. And while studies in mammals have demonstrated that most germline miRNAs lack function in mouse oocytes or during early embryonic development (Suh et al., 2010), similar studies are incomplete in C. elegans. In addition, recent findings in mammals suggest that individual miRNAs have decisive roles in promoting germline development, as seen for human miR-372 (Tran et al., 2016). Therefore, in addition to the established maternal contributions of the miR35 and miR51 families, we predict additional individual and opposing C. elegans miRNA teams also contribute to germline development.

Nine of the 13 germline miRNA candidates identified here are among the 50+ miRNAs that change in abundance as hermaphrodites age and cease reproduction (Kato et al., 2011; Lucanic et al., 2013). In particular, six members of the miR35 family dramatically increase in abundance in aged, post-reproductive worms, while miR-70, −244, and −788 decrease. The Lucanic studies also used the germline-minus strain glp-1(e2141) to further verify the age-dependent differences in levels of miR-35, −37 and −244; however, whether these germline miRNAs play a role in aging has not yet been established. Interestingly, only one of the germline miRNAs, miR-70, was reported as a miRNA that is abundant in embryos and increases in abundance during larval development, while miR-35, −37 and −40 decrease dramatically following embryogenesis and remain at low levels until adulthood (Kato et al., 2009). Perhaps miR-70 is involved in the establishment of the germline that begins in embryos and progresses in larvae, while the other 12 germline miRNAs function as maternal (or paternal) transcripts critical to embryogenesis, or in maintenance of the adult germline.

In summary, RNA-seq and the differential statistical comparisons in this report have identified 13 miRNAs that are preferentially expressed in the germline of adult hermaphrodites. These studies have further parsed the relationships of individual miRNAs in the miR-35 and −51 families and have taken the C. elegans miRNA field beyond these two very interesting families, with striking embryonic phenotypes and germline connections, to reveal five additional, germline-enriched miRNAs. We anticipate future studies will reveal these five newly-identified miRNAs also play important roles in development, either individually or in concert.