A miRNA catalogue and ncRNA annotation of the short-living fish Nothobranchius furzeri
- 1.3k Downloads
The short-lived fish Nothobranchius furzeri is the shortest-lived vertebrate that can be cultured in captivity and was recently established as a model organism for aging research. Small non-coding RNAs, especially miRNAs, are implicated in age dependent control of gene expression.
Here, we present a comprehensive catalogue of miRNAs and several other non-coding RNA classes (ncRNAs) for Nothobranchius furzeri. Analyzing multiple small RNA-Seq libraries, we show most of these identified miRNAs are expressed in at least one of seven Nothobranchius species. Additionally, duplication and clustering of N. furzeri miRNAs was analyzed and compared to the four fish species Danio rerio, Oryzias latipes, Gasterosteus aculeatus and Takifugu rubripes. A peculiar characteristic of N. furzeri, as compared to other teleosts, was a duplication of the miR-29 cluster.
The completeness of the catalogue we provide is comparable to that of the zebrafish. This catalogue represents a basis to investigate the role of miRNAs in aging and development in this species.
KeywordsmiRNome Fish miRNA evolution Nothobranchius furzeri ncRNA
- A. striatum
- D. rerio
- G. aculeatus
- H. sapiens
long non-coding RNA
- M. musculus
- N. furzeri
- N. kadleci
- N. korthausae
- N. kunthae
- N. pienaari
- N. rachovii
- O. latipes
Principal component analysis
- T. rubripes
The annual teleost Nothobranchius furzeri is a recent experimental animal model in biomedical research. In the wild, this fish inhabits ephemeral pools in semi-arid bushveld of Southern Mozambique. It has adapted to the seasonal drying of its natural environment by producing desiccation-resistant eggs, which can remain dormant in the dry mud for one and maybe more years by entering into diapause. Due to the very short duration of the rainy season in its habitat, the natural lifespan of these animals is limited to a few months. They represent the vertebrate species with the shortest captive lifespan of only 4–12 months and also with the fastest maturation. In addition, they express a series of conserved aging markers and are amenable to genetic manipulations, making them an attractive model system for aging research (for a review, see [11, 49]). A striking characteristic of N. furzeri is the existence of laboratory strains differing in lifespan and expression of aging phenotypes [15, 60]: an extremely short-lived strain (GRZ: median lifespan 3–4 months) and several longer-lived strains (e.g., MZM-04/10; median lifespan 7–9 months). The molecular basis for this striking difference in aging is unknown. A previous miRNA-Seq study of brain aging that predated genome sequencing and used homology to miRBase to annotate N. furzeri miRNAs revealed that the two strains have different global patterns of miRNA expression .
Here, we provide a comprehensive microRNA (miRNA) catalogue for N. furzeri. MiRNAs are abundant non-coding RNAs between 18 and 24 nucleotides in length that are produced in a complex biosynthetic process starting from longer transcripts and are established as key players in the post-transcriptional regulation of gene expression. MiRNA genes can be hosted within an intron of a protein-coding gene (and their transcriptional regulation follows that of the hosting gene) or can arise from primary transcripts that are regulated independently of any protein-coding RNA. Several miRNAs are grouped in genomic clusters containing mostly two to six individual miRNAs with an intra-miRNA distance of less than 10 kb, which are co-transcribed. However, unusually large clusters were also found in some species, like the miR-430 cluster in zebrafish, consisting of 57 miRNAs [41, 61, 68]. The advantage of this accumulation is unclear. It could be possible that multiple loci are required to increase the copy-number and therefore the expression level of specific miRNAs in particular conditions, like miR-430 in the maternal-zygotic transition in zebrafish (Danio rerio) . MiRNA genes are grouped into families based on sequence homology and can be defined as a collection of miRNAs that are derived from a common ancestor . On the contrary, miRNA clusters may contain miRNAs belonging to different miRNA families, but are located in relative close proximity to each other. Both the evolutionary conservation of some miRNA families and the innovations leading to appearance of novel miRNAs are well-described. An expansion of the miRNA inventory due to genome duplications in early vertebrates and in ancestral teleosts has already been described .
MiRNAs bind target mRNAs, due to sequence complementarity in the seed region (nucleotides 2–7), mostly in the 3′ untranslated region, thereby silencing expression of the gene product via translational repression and/or transcript degradation. Up to now, several thousands of miRNAs have been predicted and identified in animals, plants and viruses, and one single species can express more than one thousand miRNAs . They frequently represent the central nodes of regulatory networks and may act as “rheostat” to provide stability and fine-tuning to gene expression networks [47, 53]. Before a sequence of the N. furzeri genome assembly became available , we could show by use of the Danio rerio reference from miRBase that aging in the N. furzeri brain displays evolutionary conserved miRNA regulation, converging in a regulatory network centred on the antagonistic actions of the oncogenic MYC and tumor-suppressor TP53 , and the expression of miR-15a and the miR-17/92 cluster is mainly localized in neurogenetic regions of the adult brain . Two draft genome sequences for N. furzeri were recently produced [50, 67]. In this paper, we now provide a comprehensive annotation of the N. furzeri miRNome based on a combination of Illumina-based small RNA-Seq data, different in silico prediction methods on the genome assembly and a final manual curation. Using the newly created miRNA reference, we analyzed a large dataset of 162 small RNA-Seq libraries and report tissue-specific miRNA expression of conserved and non-conserved miRNAs in N. furzeri. We further used the N. furzeri reference to analyze the miRNA expression in other Nothobranchius species and one closely-related non-annual killifish species, which were previously used to analyze positive selection  to identify when in the evolutionary history of N. furzeri non-conserved miRNAs arose.
Results and discussion
Small RNA-Seq libraries
Number of annotated ncRNAs
For each of the six killifish species, A. striatum, N. kadleci, N. rachovii, N. pienaari, N. kunthae and N. korthausae, we generated two biological replicates of small RNA-Seq libraries from the brain of mature animals. The average size per library was 24.5 million reads with a minimum of 16.8 million and a maximum of 36.1 million reads (for more details about the small RNA-Seq libraries, see Supplement Table 1).
Annotation of ncRNAs
We could identify more than 750 non-coding RNA (ncRNA) genes in the N. furzeri genome based on small RNA-Seq reads, including editing signals, RNA elements located in the UTRs of mRNAs either controlling localization or regulation and conserved lncRNA element (see Table 1, Additional file 1 and Supplement Table 5). In line with other eukaryotes, we identified multiple gene copies of rRNAs, tRNAs, several major spliceosomal RNAs, signal recognition particle (SRP) RNAs and one copy of a minor spliceosomal RNA set. Further housekeeping RNA genes, such as RNase P, RNase MRP, and the 7SK RNA, are found, as expected, once in the entire genome. We annotated the widely distributed TPP riboswitch, capable of binding thiamine pyrophosphate and thereby regulating genes that are in charge of the thiamine balance. We could also identify more RNA elements located in the UTRs of mRNAs, being directly involved in the regulation of gene expression (3 copies of IRE – controlling iron responsive proteins, CAESAR – controlling tissue growth factor CTGF, DPB – controlling DNA polymerase β), localization of mRNAs (Vimentin3), DNA replication (four copies of the Y RNA gene, and Telomerase RNA TERC) or of unknown function (12 vault RNAs). Additionally, ncRNAs responsible for editing certain mRNAs have also been found (two copies of Antizyme FSE, one U1A polyadenylation inhibition element (PIE), 26 Potassium channel RNA editing signals (KRES), and six copies of GABA3). Two promising candidate long non-coding RNAs (lncRNAs), SPYR-IT1 and MEG8, were also included in the annotation, even though we were not able to identify all of their exons. Two vague candidates for XIST and MALAT can be viewed in the supplemental material. The MALAT-derived masc and men RNA gene was clearly detected in 42 copies throughout the genome of N. furzeri.
Mapping and miRNA prediction results
Amount of annotated miRNAs, identified miRNA clusters and the number of miRNAs in clusters, as well as known conserved and non-conserved miRNA families in N. furzeri (Nfu), D. rerio (Dre), O. latipes (Ola), G. aculeatus (Gac) and T. rubripes (Fru)
#miRNAs in clusters
The age-dependent expression of the following miRNAs was previously demonstrated by qPCR: tni-miR-15a, tni-miR-101a, tni-miR-101b, dre-miR-145, hsa-miR 29c-1 (100% identical to dre-miR-29a), hsa-let-7a-5p, hsa-miR-124a-1, hsa-miR-1-2, olamiR-21, ola-miR-183-5p and, from cluster dre-miR-17a/18a/19a, and dre-miR-20a (the used primers were Qiagen miScript primer). Expression changes detected by sequencing were validated on an independent set of specimens. All 13 miRNAs showed concordant changes in their expression, of which six reached statistical significance . The expression of the following miRNAs in the brain was confirmed by in situ hybridisation using LNA probes (Exiqon): miR-9, miR-124  and miR-15a, miR-20a .
Target prediction of the miRNA candidates
In order to get a first insight of the potential regulatory functions of our putative miRNA genes, we performed a target prediction based on the miRNA seed regions and the aligned homologous 3′-UTR mRNA regions of N. furzeri and D. rerio. Additionally, we repeated this target prediction analysis, including homologous 3′- UTR mRNA regions of M. musculus and H. sapiens to have a more conservative target list for each miRNA candidate, since in silico miRNA target predictions tend to have a high number of false positive results . Using only the two fish 3′-UTR alignments, we predicted for 438 of our miRNA candidates potential mRNA targets with a median of 47 putative targets per miRNA. With our more conservative approach, still 419 miRNA candidates showed targeting potential with a median of 25 putative targets per miRNA. To further examine these potential targets, we calculated enrichment scores of miRNA binding sites in already known sets of down-regulated genes in the brain of N. furzeri during aging  and in different tissues between young and very old N. furzeri individuals . In the first study, both clusters, containing genes with decreasing activity during aging, show a significant enrichment of miRNA targets (cluster1: p = 8.67−25; cluster5: p = 1.78−5). For all three investigated tissues in the second study, we also found a significant enrichment of miRNA target sites within the downregulated genes (brain: p = 6.19−32; liver: p = 7.72−17; skin: p = 1.49−9). Additionally, we identified single miRNA candidates, whose targets were enriched in any of the above-mentioned gene sets (for details, see online supplement section miRNA target prediction). We found e.g., miR-10, miR-29 and miR-92 showing potential to be significantly involved in the down-regulation of genes in the aging brain of N. furzeri, like cell cycle regulators (ccne2 , nek6 , cdk13 ) or cancer related genes (mycn [8, 12], vav2 [13, 28]), both processes involved in aging.
Effects of tissue and age on global miRNA expression
miRNA expression in closely related killifish
SmallRNA-Seq samples from Nothobranchius strains generated in this study. * – unknown; # – number of replicates; + − two weeks post-fertilization plus diapause
N. furzeri MZM
N. furzeri MZM
brain, liver, skin
5, 12, 20, 27, 32, 39
N. furzeri GRZ
brain, liver, skin
5, 7, 10, 12, 14
miRNA clusters and gene duplications
In all investigated fish species but T. rupripes, the largest cluster is the miR-430 cluster (containing 7 to 55 miRNAs; see Fig. 6c). This cluster is extremely divergent and evolving relatively quickly in each lineage. Not only the number of miR-430 copies within each cluster varies greatly but also the number and organization of the members of this miRNA family. Whereas miR-430a and miR-430c can be found in all five fish species, miR-430b and miR-430d seem to occur only in D. rerio and O. latipes, respectively. Additionally, no structural similarities or shared repetition patterns can be observed for this miRNA cluster, which is an additional indication of the low purifying selection on this specific gene cluster. However, a clear duplication pattern can be observed for the miR-430 cluster in D. rerio (the order miR-430c/b/a is repeated with only a few exceptions) and N. furzeri (the order miR-430c/a/a/c/a/a/a is repeated). For O. latipes and G. aculeatus, the order of miR-430 variants appears to be more random, and T. rubripes has too few copies to show any repeated pattern.
Figure 6b depicting the miR-17-92 cluster shows an example of the other extreme: in all five investigated fish species, two perfectly conserved clusters can be found. These represent a duplication of an ancestral cluster present in all vertebrates, and the order of the different members is perfectly conserved. It is known that the miR-17-92 cluster is transcribed polycystronically and acts in oncogenic and tumor suppressor pathways [23, 57]. Furthermore, up to two smaller and lesser conserved clusters, containing at least two miRNAs of the miR-17 or miR-92 family, were identified per fish species, similar to what is known for mammals. Having correctly identified this highly conserved cluster in N. furzeri is again good evidence for the high quality of its newly assembled genome and completeness of our miRNA catalogue.
Another example for an evolutionary conserved miRNA cluster is the miR-29 cluster depicted in Fig. 6d. Mir-29 family members are up-regulated during aging in a variety of different tissues including muscle, skin, brain and aorta [2, 18, 46, 54, 56, 66] and appear to be key regulators of age-dependent gene expression [6, 51]. This cluster consists of miR-29a (which is identical to the mammalian miR-29c) and its variant miR-29b and is duplicated at least once. In some fish species, an additional variant miR-29c is known, which is identical to the miR-29a in mammals, with one nucleotide being different outside the seed region . As from RFAM (version 12.1) and miRBase (release 21), miR-29 genes are mainly identified in vertebrates as well as one Hemichordata and one Arthropoda, so we can only speculate that the original cluster duplication event arose in the early metazoa lineage. In O. latipes and T. rubripes, both miR-29 clusters are still present, whereas D. rerio seems to has lost one copy of the miR-29a gene. For G. aculeatus, we were only able to identify one miR-29 cluster. However, since its genome assembly is incomplete, we assume that the second cluster may not be lost but is missing in the current version of its miRNA annotation. Interestingly, in N. furzeri, we identified an additional miR-29a/b pair and a fourth single copy of miR-29b. Assuming a complete genome assembly, different scenarios could explain this finding: (1) both original miR-29 clusters were individually duplicated once more, and the fourth miR-29a gene was later lost, (2) one of the two clusters was duplicated as a whole, whereas in the other only miR-29b was copied or (3) both original clusters were duplicated during the same event, and again one of the miR-29a genes was later lost.
About the same amount of different miRBase miRNA families could be identified for all five fish species, despite their big differences in the number of identified miRNA genes. All miRNA genes not matching any known mirRBase family were clustered based on their sequence identity in order to estimate the amount of miRNA ‘families’ not covered by the miRBase database (see Table 3 and Supplement Table 4).
This study involves a multitude of small RNA-Seq libraries from several tissues, ages, strains and embryos of N. furzeri and closely related species. The aim was the characterization of the N. furzeri miRNome and a detailed annotation in the recently published genome . The inclusion of other killifish species allowed us to analyze the occurrence of novel miRNAs in the group of annual fish. Due to the fact that we identified roughly the same number of miRNAs in N. furzeri as known in D. rerio and both fish species share almost equal amounts of miRBase families and unknown miRNA families, we assume that our miRNA catalogue is comparable to the one of the model organism D. rerio.
Animal maintenance was performed as described [59, 60]. To avoid effects of circadian rhythms and feeding, the animals were always sacrificed at 10 a.m. in a fasted state. Animals were sacrificed by an overdose of anesthetics in accordance with the Annex IV of the EU directive 2010/63. They were placed approx. 5–10 min in a methanesulfonate (MS222) solution at a concentration of 1 mg/ml in buffered ethyl 3-aminobenzoate methanesulfonate without prior sedation and observed until no vital signs (body and operculum movement, righting reflex) appeared. At death, animals were transferred on crushed ice, decapitated and organs were harvested. The protocols of animal maintenance and experiments were approved by the local authority in the State of Thuringia (Veterinaer- und Lebensmittelueberwachungsamt). Total RNA was extracted as described . The RNA quality and amount was determined using the Agilent Bioanalyzer 2100 and the RNA 6000 Nano Kit (Agilent Technologies).
Small RNA library preparation and sequencing
The library preparation and sequencing was done using Illumina‘s NGS platform . One μg of total RNA was used for the library preparation, using Illumina‘s TruSeq small RNA sample preparation kit, following the manufacturer‘s instruction. The libraries were quantified on the Agilent DNA 1000 chip and subjected to sequencing-by-synthesis on an Illumina HiSeq2500 in high-output, 50 bp single-read mode. Sequencing chemistry v3 was used. The read data were extracted in FastQ format, using the Illumina supported tool bcl2fastq (v1.8.3 and v1.8.4). The only exceptions were three of the N. furzeri embryo samples, which were sequenced on an Illumina HiSeq2000 in 50 bp single-read mode and where read data was extracted in FastQ format using the tool CASAVA (v1.8.2). The sequencing resulted in around 4–50 million reads per sample with pooling eight samples per lane.
In total, 169 small RNA-Seq libraries from seven different killifish species were created. 157 of them were obtained from N. furzeri strains GRZ and MZM-0410 at several ages from the three tissues brain, liver and skin. The remaining RNA-Seq libraries obtained from Aphyosemion striatum, N. kadleci, N. rachovii, N. pienaari, N. kunthae and N. korthausae were used to identify expression patterns at predicted miRNA locations in N. furzeri and miRbase pre-mature miRNA sequences. For details see Table 4, Supplement Table 1 and Supplement Table 2.
Small RNA-Seq library processing and mapping
In-house scripts were used to cut the RA3 adapter of the TruSeq small RNA preparation kit (5′-TGGAATTCTCGGGTGCCAAGG) from the reads. Additionally, PRINSEQ (v0.20.3)  was used to trim the reads from both sides in order that the read bases had a minimum quality of 20 and reads were at least 15 bases long. The mapping onto the N. furzeri genome was performed with segemehl (v0.2.0)  using the -H 1 option, allowing single reads to be mapped to multiple best fitting locations. The visualization of mapped reads was done using IGV (v2.0.34) . Since Bowtie (v1.0.0)  is the built-in method in miRDeep* for mapping, it was also used for the genomes of N. furzeri, D. rerio, O. latipes and T. rubripes.
Genomes and annotations
The recently published high-quality draft genome assembly and annotation of N. furzeri and the small RNA-Seq libraries described above were used for mapping as well as for miRNA and other ncRNA predictions and annotations . Additionally, these RNASeq libraries were also mapped on the following fish genomes: Danio rerio (GRCz10) , Oryzias latipes (HdrR)  and Takifugu rubripes (FUGU5) . For the annotation comparison, the latest complete genomic information of those three fish and of Gasterosteus aculeatus (BROAD S1)  were downloaded from the ensembl database  Additionally, for miRNA target prediction, the recent genomes and annotations of Homo sapiens (GRCh38) and Mus musculus (GRCm38) from the ensemble database were used.
ncRNA and miRNA annotation
Already characterized and conserved non-coding RNAs were annotated with GoRAP 2.0, which is based on the RFAM database, currently holding 2450 ncRNA families (v12.0) . For an initial prediction of candidate miRNAs, a combination of five tools was used, each of them following a different annotation strategy: miRDeep* (v32) , Infernal (v1.1) , BLAST (v2.2.30) , GoRAP (v2.0, unpublished) and CID-miRNA (version from April 2015) . A detailed description of the individual searches can be found below. All results were merged and putative miRNAs overlapping with genes of the recently published N. furzeri annotation were removed. The expression profiles of the remaining non-redundant candidate miRNA genes were analyzed automatically using Blockbuster (v1)  and in-house scripts in order to mark candidates that did not exhibit a typical miRNA expression profile (according to [30, 36]). All candidates were additionally manually examined and filtered by carefully checking the features of the potential hairpin secondary structure as well as the precise mapping of reads supporting the predicted precursor miRNA, leading to the final set of miRNA predictions.
Mappings of 39 MZM brain, 15 GRZ brain, 25 GRZ liver, 28 MZM liver, 3 MZM skin and 7 MZM embryo small RNA-Seq libraries were used on four different fish genomes (N. furzeri, D. rerio, O. latipes, T. rubripes) as input for miRDeep* (for a detailed list of used libraries, see Supplement Table 1). Predictions from all 117 mappings were pooled together in order to obtain a comprehensive representation of the miRDeep* results. To each predicted miRNA hairpin sequence, we assigned the average of the miRDeep* score computed across the multiple samples were the sequence was found. The merged non-redundant list of identified miRNA sequences was re-mapped with BLAT  on the N. furzeri genome, and only gap-free alignments were accepted. These loci underwent further filtering steps: (i) a hairpin sequence was considered reliable if it showed a BLAT hit (one mismatch allowed) in miRbase (release 20)  or a miRDeep* score equal or more than 7 and (ii) overlapping hairpin loci (i.e., within 100 nt) were discarded, and the sequence with the highest score was kept. Predictions where no hits in miRBase could be obtained were further analyzed based on their secondary structure. Therefore, corresponding sequences were extended by 50 nt on either side and were compared with Rfam using Infernal. All predicted loci that had a significant hit to a known miRNA secondary structure or no hit at all were kept, while loci hitting other ncRNAs were discarded.
In order to identify candidates from the most conserved miRNA families, blastn was used with all mature and pre-mature miRNA sequences available on miRBase (release 21) . Only non-redundant hits were kept if they spanned the complete sequences of their corresponding input miRNAs to at least 90% with no gaps allowed. To further reduce false positive hits, a stringent cut-off of p < 10−7 was chosen.
Being based on a stochastic context-free grammar model to identify possible pre-miRNAs, CID-miRNA follows a similar approach as Infernals covariance models. The N. furzeri genome was given as input with the following thresholds: putative miRNAs have a length between 60 bp and 120 bp, and the grammar and structural cut-off were set to the recommended values of −0.609999 and 23, respectively.
miRNA target prediction
We thank Sabine Matz, Ivonne Heinze and Ivonne Goerlich for excellent technical assistance.
This work has been partially financed by Carl Zeiss Stiftung (Manja Marz; data analysis and interpretation; writing the manuscript) and was funded by the Thuringian country programme ProExzellenz of the Thuringian Ministry for Research (TMWWDG; RegenerAging – FSU-I-03/14; Emanuel Barth; data analysis and interpretation; writing the manuscript). This work was partially supported by the German Ministry for Education and Research (JenAge; BMBF support code: 0315581, Alessandro Cellerino and Mario Baumgart; design of the study; data collection), by the German Research Foundation (DFG support code BA 5576/1–1; Mario Baumgart; writing the manuscript) and by intramural grant of Scuola Normale Superiore di Pisa (Alessandor Cellerino; writing the manuscript).
Availability of data and materials
Supplementary material can be found online at http://www.rna.uni-jena.de/en/supplements/nothobranchius-furzeri-mirnome/.
The data were deposited in GEO with the accession number GSE92854.
The presented miRNome annotation is accessible and downloadable via the NFINgb Nothobranchius furzeri Genome Browser (http://nfingb.leibniz-fli.de/).
MB, MP and AC conceived and designed the study; MB performed the experiments; MG performed the RNA-Seq; IA, AP, EB, AS, PK and MM analyzed and interpreted the data; AC, MB, EB and MM wrote the manuscript with contributions from all other authors. All authors read and approved the final manuscript.
Ethics approval and consent to participate
The protocols of animal maintenance and experiments were approved by the local authority in the State of Thuringia (Veterinaer-und Lebensmittelueberwachungsamt) complying with the EU directive 2010/63 and the german Animal Welfare Act §4(3) on the protection of animals used for scientific purposes.
Animals were bred and maintained for generations in-house in the institute’s fish facility (Leibniz-Institute on Aging - Fritz-Lipmann-institute, Beutenbergstr. 11, 07745 Jena, German) under animal husbandry license J-SHK-2684-04-08/11.
Consent for publication
The authors declare that they have no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
- 4.Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J Royal Stat Soc Ser B (Methodol). 1995;57(1):289–300.Google Scholar
- 10.Cellerino A. Regulation of microRNA expression in the neuronal stem cell niches during aging of the short-lived annual fish Nothobranchius Furzeri. Regulatory RNAs in the Nervous System, 8 vol. 2014.Google Scholar
- 16.Dorn A, Musilov’a Z, Platzer M, Reichwald K, Cellerino A. The strange case of east african annual fishes: aridification correlates with diversification for a savannah aquatic group? BMC Evol Biol. 2014;14(1):1.Google Scholar
- 36.Langenberger D, Bermudez-Santana C, Stadler PF, Hoffmann S. Identification and classification of small RNAs in transcriptome sequence data. Pac Symp Biocomput. 2010;15:80–87. World Scientific.Google Scholar
- 42.Marqu’es F, Moreau J-L, Peaucellier G, Lozano J-C, Schatt P, Picard A, Callebaut I, Perret E, Genevi’ere A-M. A new subfamily of high molecular mass cdc2-related kinases with pitai/vre motifs. Biochem. Biophys. Res. Commun. 2000;279(3):832–7.Google Scholar
- 48.Pinz’on N, Li B, Martinez L, Sergeeva A, Presumey J, Apparailly F, Seitz H. microrna target prediction programs predict many false positives. Genome Res. 2017;27(2):234–45.Google Scholar
- 51.Ripa R, Dolfi L, Terrigno M, Pandolfini L, Arcucci V, Groth M, Tozzini ET, Baumgart M, Cellerino A. Microrna mir-29 controls a compensatory response to limit neuronal iron accumulation during adult life and aging. BMC Biol. 2017;15.1:9.Google Scholar
- 53.Siciliano V, Garzilli I, Fracassi C, Criscuolo S, Ventre S, di Bernardo D. MiRNAs confer phenotypic robustness to gene networks by suppressing biological noise. Nat Commun. 2013;4Google Scholar
- 55.Suzuki R, Shimodaira H. Pvclust: an r package for assessing the uncertainty in hierarchical clustering. Bioinformatics. 2006;22(12):1540–2.Google Scholar
- 61.Thatcher EJ, Bond J, Paydar I, Patton JG. Genomic organization of zebrafish microRNAs. BMC Genomics. 2008;9(1):1.Google Scholar
- 62.Thorvaldsd’ottir H, Robinson JT, Mesirov JP. Integrative Genomics Viewer (IGV): high-performance genomics data visualization and exploration. Brief Bioinform. 2013;14.2:178–92.Google Scholar
- 64.Tozzini ET, Savino A, Ripa R, Battistoni G, Baumgart M, Cellerino A. Regulation of microrna expression in the neuronal stem cell niches during aging of the short-lived annual fish nothobranchius furzeri. Front Cell Neurosci. 2014;8Google Scholar
- 68.Wang Y, Luo J, Zhang H, Lu J. microRNAs in the same clusters evolve to coordinately regulate functionally related genes. Mol Biol Evol. 2016;33.9:2232–47.Google Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.