山东大学：《发育生物学 Developmental Biology》课程教学资源（参考资料）Circular RNAs are a large class of animal RNAs with regulatory potency

ARTICLE doi:10.1038/ nature11928 Circular RNas are a large class of animal RNAs with regulatory potency Sebastian Memczak *, Marvin Jens *, Antigoni Elefsinioti*, Francesca Torti*, Janna Krueger?, Agnieszka Rybak, Luisa Maier, Mackowiak, Lea H Greger Circular RNAs(circRNAs) in animals are an enigmatic class of RNa with unknown function. To explore circRNAs systematically, we sequenced and computationally analysed human, mouse and nematode RNA. We detected thousands of well-expressed, stable circRNAs, often showing tissue/developmental-stage-specific expression Sequence analysis indicated important regulatory functions for circRNAs. We found that a human circRNA, antisense to the cerebellar degeneration-related protein 1 transcript(CDRlas), is densely bound by microRNA (miRNA)effector complexes and harbours 63 conserved binding sites for the ancient miRNA miR-7. Further analyses indicated that CDRlas functions to bind miR-7 in neuronal tissues. Human CDRlas expression in zebrafish impaired midbrain development, similar to knocking down miR-7, suggesting that CDRlas is a miRNA antagonist with a miRNA-binding capacity ten times higher than any other known transcript. Together, our data provide evidence that circRNAs form a large class of post-transcriptional regulators. Numerous circRNAs form by head-to-tail splicing of exons, suggesting previously unrecognized regulatory potential of coding sequences Mature messenger RNAS are linear molecules with 5'and 3 termini computational pipeline can find circRNAs in any genomic region that reflect start and stop of the RNa polymerase on the DNA tem- takes advantage of long(100 nucleotides)reads, and predicts the plate In cells, different RNA molecules are sometimes joined together acceptor and donor splice sites used to link the ends of the RNAs by splicing reactions(trans-splicing), but covalent linkage of the ends We do not rely on paired-end sequencing data or known splice sites of a single RNA molecule to form a circular RNA(circRNA)is usually Using published 2526 and our own sequencing data, our method considered a rareevent circRNAs were discovered in plants and shown reported thousands of circRNAs in human and mouse tissues as well to encode subviral agents!. In unicellular organisms, cirCRNAs mostly as in different developmental stages of Caenorhabditis elegans stem from self-splicing introns of pre-ribosomal RNA, but can also Numerous circRNAs appear to be specifically expressed across tissues rise from protein-coding genes in archaea. In the few unambiguously or developmental stages. We validated these data and showed that validated circRNAs in animals, the spliceosome seems to link the 5 most tested circRNAs are well expressed, stable and circularized and downstream 3 ends of exons within the same transcript'-lo. using the predicted splice sites. circRNA sequences were significantly Perhaps the best known circRNA is antisense to the mRNA transcribed enriched in conserved nucleotides, indicating that circRNAs compete from the SRY (sex-determining region Y) locus and is highly expressed with other RNAs for binding by RNa binding proteins(RBPs)or in testes. Evidence from computational analyses of expression data in miRNAs. We combined biochemical, functional and computational Archaea and Mammalia suggests that circRNAs are more prevalent analyses to show that indeed a known human circRNA, CDRI anti- than previously thought. however, it is unknown whether animal sense(CDRlas), can function as a negative regulator of miR-7,a miRNA with perfect sequence conservation from annelids to human. In comparison to circRNAS, miRNAs are extremely well studied. Together, our data provide evidence that circRNAs form an important miRNAs are -21-nucleotide-long non-coding RNAs that guide the class of post-transcriptional regulators ffector protein Argonaute(AGO) to mRNAs of coding genes to press their protein production"-14. In humans, miRNAs directly circRNAs have complex expression patterns regulate expression of most mRNAss-I8 in a diverse range of bio- To comprehensively identify stably expressed circRNAs in animals we ical functions. However, surprisingly little is known about how screened RNA sequencing reads for splice junctions formed by an and if mRNAs can escape regulation by a miRNA. A recently discov- acceptor splice site at the 5 end of an exon and a donor site at a red mechanism for miRNA removal in a sequence-specific manner is downstream 3'end(head-to-tail)(Fig. la). As standard RNA expres- based on target sites acting as decoys or miRNA sponges. RNA with sion profiling enriches for polyadenylated RNAs, we used data gene miRNA binding sites should, if expressed highly enough, sequester rated after ribosomal RNA depletion (ribominus)and random way the miRNA from its target sites. However, all reported mam- priming, Such data were used before to detect scrambled exons in malian miRNA sponges have only one or two binding sites for the same mammals"(see Methods for comparison). However, this approach miRNA and are not highly expressed, limiting their potency was not specifically designed to detect circRNAs and (1)only used To identify circRNAs across animal cells systematically, we screened existing exon-intron annotations, thus missing RNAs transcribed RNA-seq data for circRNAs. Compared to previous approaches our from introns or unannotated transcripts; (2)did not explicitly identi sYstems Biology of Gene Regulatory Elements, Max-Delbrulck-Center for Molecular Medicine, Robert-Rbssle-Strasse 10,13125 Berlin 2Angiogenesis and Cardiovascular Pathology, Max elbrilck-Center for Molecular Medicine, Robert-Rossle-Strasse 10, 13125 Berlin, Germany. RNA Biology and Post-Transcriptional Regulation, Max-Delbriick-Center for Molecular Medicine, Robert- Rdssle-Strasse 10, 13125 Berlin, Germany. signaling Dynamics in Single Cells, Max-Delbriick-Center for Molecular Medicine, Robert-Rbssle-Strasse 10, 13125 Berlin, Germany These authors contributed equally to this work 0 MONTH 2013 VOL 000I NATURE I @2013 Macmillan Publishers Limited. All rights reserved

ARTICLE doi:10.1038/nature11928 Circular RNAs are a large class of animal RNAs with regulatory potency Sebastian Memczak1 *, Marvin Jens1 *, Antigoni Elefsinioti1 *, Francesca Torti1 *, Janna Krueger2 , Agnieszka Rybak1 , Luisa Maier1 , Sebastian D. Mackowiak1 , Lea H. Gregersen3 , Mathias Munschauer3 , Alexander Loewer4 , Ulrike Ziebold1 , Markus Landthaler3 , Christine Kocks1 , Ferdinand le Noble2 & Nikolaus Rajewsky1 Circular RNAs (circRNAs) in animals are an enigmatic class of RNA with unknown function. To explore circRNAs systematically, we sequenced and computationally analysed human, mouse and nematode RNA. We detected thousands of well-expressed, stable circRNAs, often showing tissue/developmental-stage-specific expression. Sequence analysis indicated important regulatory functions for circRNAs. We found that a human circRNA, antisense to the cerebellar degeneration-related protein 1 transcript (CDR1as), is densely bound by microRNA (miRNA) effector complexes and harbours 63 conserved binding sites for the ancient miRNA miR-7. Further analyses indicated that CDR1as functions to bind miR-7 in neuronal tissues. Human CDR1as expression in zebrafish impaired midbrain development, similar to knocking down miR-7, suggesting that CDR1as is a miRNA antagonist with a miRNA-binding capacity ten times higher than any other known transcript. Together, our data provide evidence that circRNAs form a large class of post-transcriptional regulators. Numerous circRNAs form by head-to-tail splicing of exons, suggesting previously unrecognized regulatory potential of coding sequences. Mature messenger RNAs are linear molecules with 59 and 39 termini that reflect start and stop of the RNA polymerase on the DNA tem￾plate. In cells, different RNA molecules are sometimes joined together by splicing reactions (trans-splicing), but covalent linkage of the ends of a single RNA molecule to form a circular RNA (circRNA) is usually considered a rare event. circRNAs were discovered in plants and shown to encode subviral agents1 . In unicellular organisms, circRNAs mostly stem from self-splicing introns of pre-ribosomal RNA2 , but can also arise from protein-coding genes in archaea3 . In the few unambiguously validated circRNAs in animals, the spliceosome seems to link the 59 and downstream 39 ends of exons within the same transcript4–10. Perhaps the best known circRNA is antisense to the mRNA transcribed from the SRY (sex-determining region Y) locus and is highly expressed in testes6 . Evidence from computational analyses of expression data in Archaea and Mammalia suggests that circRNAs are more prevalent than previously thought3,10; however, it is unknown whether animal circRNAs have any biological function. In comparison to circRNAs, miRNAs are extremely well studied. miRNAs are ,21-nucleotide-long non-coding RNAs that guide the effector protein Argonaute (AGO) to mRNAs of coding genes to repress their protein production11–14. In humans, miRNAs directly regulate expression of most mRNAs15–18 in a diverse range of bio￾logical functions. However, surprisingly little is known about how and if mRNAs can escape regulation by a miRNA. A recently discov￾ered mechanism for miRNA removal in a sequence-specific manner is based on target sites acting as decoys or miRNA sponges19,20. RNA with miRNA binding sites should, if expressed highly enough, sequester away the miRNA from its target sites. However, all reported mam￾malian miRNA sponges have only one or two binding sites for the same miRNA and are not highly expressed, limiting their potency21–24. To identify circRNAs across animal cells systematically, we screened RNA-seq data for circRNAs. Compared to previous approaches10 our computational pipeline can find circRNAs in any genomic region, takes advantage of long (,100 nucleotides) reads, and predicts the acceptor and donor splice sites used to link the ends of the RNAs. We do not rely on paired-end sequencing data or known splice sites. Using published10,25,26 and our own sequencing data, our method reported thousands of circRNAs in human and mouse tissues as well as in different developmental stages of Caenorhabditis elegans. Numerous circRNAs appear to be specifically expressed across tissues or developmental stages. We validated these data and showed that most tested circRNAs are well expressed, stable and circularized using the predicted splice sites. circRNA sequences were significantly enriched in conserved nucleotides, indicating that circRNAs compete with other RNAs for binding by RNA binding proteins (RBPs) or miRNAs. We combined biochemical, functional and computational analyses to show that indeed a known human circRNA, CDR1 anti￾sense (CDR1as)9 , can function as a negative regulator of miR-7, a miRNA with perfect sequence conservation from annelids to human. Together, our data provide evidence that circRNAs form an important class of post-transcriptional regulators. circRNAs have complex expression patterns To comprehensively identify stably expressed circRNAs in animals we screened RNA sequencing reads for splice junctions formed by an acceptor splice site at the 59 end of an exon and a donor site at a downstream 39 end (head-to-tail) (Fig. 1a). As standard RNA expres￾sion profiling enriches for polyadenylated RNAs, we used data gene￾rated after ribosomal RNA depletion (ribominus) and random priming. Such data were used before to detect scrambled exons in mammals10 (see Methods for comparison). However, this approach was not specifically designed to detect circRNAs and (1) only used existing exon–intron annotations, thus missing RNAs transcribed from introns or unannotated transcripts; (2) did not explicitly identify *These authors contributed equally to this work. 1 Systems Biology of Gene Regulatory Elements, Max-Delbru¨ ck-Center for Molecular Medicine, Robert-Ro¨ssle-Strasse 10, 13125 Berlin, Germany. 2 Angiogenesis and Cardiovascular Pathology, Max￾Delbru¨ ck-Center for Molecular Medicine, Robert-Ro¨ssle-Strasse 10, 13125 Berlin, Germany. 3 RNA Biology and Post-Transcriptional Regulation, Max-Delbru¨ ck-Center for Molecular Medicine, Robert￾Ro¨ssle-Strasse 10, 13125 Berlin, Germany. 4 Signaling Dynamics in Single Cells, Max-Delbru¨ ck-Center for Molecular Medicine, Robert-Ro¨ssle-Strasse 10, 13125 Berlin, Germany. 00 MONTH 2013 | VOL 000 | NATURE | 1 ©2013 Macmillan Publishers Limited. All rights reserved

RESEARCH ARTICLE alignment and splice-site detection from at least two independent junction-spanning reads(Fig 1b).The expression of genes predicted to give rise to circRNAs was only slightl shifted towards higher expression values( Supplementary Fig. 1d), Linear splicing indicating that circRNAs are not just rare mistakes of the splice- Donor some. We also identified 1, 903 circRNAs in mouse(brains, fetal head, differentiation-induced embryonic stem cells; Supplementary Fig. le)281 of these mapped to human circRNAs(Supplem tary Fig. 1f). To explore whether circRNAs exist in other animal clades, we used sequencing data that we produced from various C. elegan Oocyte 1-cell embryo d Human circRNAs developmental stages(Stoeckius, M et al, manuscript in preparation) Methods)and detected 724 circRNAs, with at least two independent reads(Fig. 1c). Numerous circRNAs seem to be specifically expressed in a cell type or developmental stage(Fig. Ib, c and Supplementary Fig. le). For example, hsa-circRNA 2149 is supported by 13 unique, head-to-tail spanning reads in CD19 leukocytes but is not detected in CD34 leukocytes(which were sequenced at comparable depth; Supplemen K27 tary Table 1), neutrophils or HEK293 cells. Analogously, a number of C elegans(/24) (1.602 inside coding transcripts) nematode circRNAs seem to be expressed in oocytes but absent in f hird codon position I-or 2-cell embryos CDR1 gene catalogue of non-coding RNAs27-29 oing the RefSeq database and a 5% of human circRNAs align Long ncRNA PVT1 sense to known genes. Their splice sites typically span one to five exons(s ntary Fig. 1g) and overlap coding exons(84%),but ZRANBT only in 65% of these cases are both splice sites that participate in the circularization known splice sites(Supplementary Table 2), demon- Consere of\? cos strating the advantage of our strategy. 10% of all circRNAs align antisense to known transcripts, smaller fractions align to UTRS, introns, unannotated regions of the genome(Fig. 1d). Examples of Figure 1 Detection, classification and evolutionary conservation of human circRNAs are shown in Fig. le quentially to the genome for linear( top) but in reversed orientation for head. sea We analysed sequence conservation within circRNAS. As genomic circRNAs. a, The termini of junction-spanning reads(anchors)align quence is subject to different degrees of evolutionary selection, to-tail spliced reads(bottom). Spliced reads must distribute completely to depending on function, we studied three subtypes of circRNAs anchors, flanked by AG/GU (Methods). b, c, circRNAs in human cell types Intergenic and a few intronic circRNAs display a mild but significant (b)and nematode stages(c).d, Genomic origin of human circRNAs. A total of enrichment of conserved nucleotides(Supplementary Fig. Ih, i). 6% of circRNAs overlap known transcripts. e, Examples of human circRNAs. To analyse circRNAs composed of coding sequence and thus high The AFFI intron is spliced out( Supplementary Fig 2e) Sequence conservation: overall conservation, we selected 223 human circRNAs with circular lacental mammals phyloP (Methods), scale bar, 200 nucleotides f a total of 223 human coding sequence circRNAs with mouse orthologues orthologues in mouse(Methods)and entirely composed of coding (green)and controls(black) with matched conservation level(inset: mean sequence Control(linear)exons were randomly selected to match the onservation for each codon position(grey), controls(black); x axis, codon level of conservation observed in first and second codon positions positions y axis, placental mammals phyloP score; see also Methods and (Methods, Fig. If inset and Supplementary Fig. Ik for conservation served(P<4r,i,k). Third codon positions are significantly more of the remaining coding sequence(CDS). circRNAs with conserved ircularization were significantly more conserved in the third codon position than controls, indicating evolutionary constraints at the nuc- the splice sites used for circularization; and (3)assumed that each pair leotide level, in addition to selection at the protein level(Fig. If and of mates in paired-end sequencing derives from the same RNA mole- Supplementary Fig. 1j, k). In summary, we have confidently identified ule. To search in a more unbiased way for circRNAs, we designed a large number of circRNAs with complex expression patterns, which algorithm(Methods) that identifies linear and circular splicing derive often but not always from coding exons Sequence conservation events in ribominus data. First, we filtered out reads that aligned con- suggests that at least a subset contains functional sequence elements tiguously to the genome, retaining the spliced reads. Next, we mapped the terminal parts of each candidate read independently to the genome Characterization of 50 predicted circRNAS to find ue an tions. Finally, we demanded that(1)anchor We experimentally tested our circRNA predictions in HEK293 cells alignments can be extended such that the original read sequence Head-to-tail splicing was assayed by quantitative polymerase chain aligns completely, and( 2)the inferred breakpoint is flanked by GU/ reaction( qPCR) after reverse transcription, with divergent primers AG splice signals. Non-unique mappings and ambiguous breakpoints and Sanger sequencing( Fig. 2a, b). Predicted head-to-tail junctions were discarded. We detected circularization splicing from the reversed of 19 out of 23 randomly chosen circRNAs(83%)could be validated, (head-to-tail) orientation of the anchor alignments(Fig. la). Our demonstrating high accuracy of our predictions(Table 1). In contrast, method also recovered tens of thousands of known linear splicing 5 out of7(71%)candidates exclusively predicted in leukocytes could not events(Methods and Supplementary Fig. la, b). We estimated sen- be detected in HEK293 cells, validating cell-type-specific expression. sitivity(75%)and false-discoveof real sequencing data (ang simulated Head-to-tail splicing could be produced by trans-splicing or geno- eads and various permutations of Methods and mic rearrangements. To rule out these possibilities as well as potentia upplementary Fig. Ic). However, the efficiency of ribominus pro- PCR artefacts, we successfully validated the insensitivity of human tocols to extract and sequence circRNAs is limited, reducing overall circRNA candidates to digestion with RNase R-an exonuclease that degrades linear RNA molecules -by northern blotting with probe We generated ribominus data for HEK293 cells and, combined which span the head-to-tail junctions(Fig. 2c). We quantified RNase with human leukocyte data, detected 1, 950 circRNAs with support Rresistance for 21 candidates with confirmed head-to-tail splicing by 2I NATURE I VOL 00000 MONTH 2013 @2013 Macmillan Publishers Limited. All rights reserved

the splice sites used for circularization; and (3) assumed that each pair of mates in paired-end sequencing derives from the same RNA mole￾cule. To search in a more unbiased way for circRNAs, we designed an algorithm (Methods) that identifies linear and circular splicing events in ribominus data. First, we filtered out reads that aligned con￾tiguously to the genome, retaining the spliced reads. Next, we mapped the terminal parts of each candidate read independently to the genome to find unique anchor positions. Finally, we demanded that (1) anchor alignments can be extended such that the original read sequence aligns completely, and (2) the inferred breakpoint is flanked by GU/ AG splice signals. Non-unique mappings and ambiguous breakpoints were discarded. We detected circularization splicing from the reversed (head-to-tail) orientation of the anchor alignments (Fig. 1a). Our method also recovered tens of thousands of known linear splicing events (Methods and Supplementary Fig. 1a, b). We estimated sen￾sitivity (.75%) and false-discovery rate (FDR ,0.2%) using simulated reads and various permutations of real sequencing data (Methods and Supplementary Fig. 1c). However, the efficiency of ribominus pro￾tocols to extract and sequence circRNAs is limited, reducing overall sensitivity. We generated ribominus data for HEK293 cells and, combined with human leukocyte data10, detected 1,950 circRNAs with support from at least two independent junction-spanning reads (Fig. 1b). The expression of genes predicted to give rise to circRNAs was only slightly shifted towards higher expression values (Supplementary Fig. 1d), indicating that circRNAs are not just rare mistakes of the spliceo￾some. We also identified 1,903 circRNAs in mouse (brains, fetal head, differentiation-induced embryonic stem cells; Supplementary Fig. 1e)25,26; 81 of these mapped to human circRNAs (Supplemen￾tary Fig. 1f). To explore whether circRNAs exist in other animal clades, we used sequencing data that we produced from various C. elegans developmental stages (Stoeckius, M.et al., manuscript in preparation) (Methods) and detected 724 circRNAs, with at least two independent reads (Fig. 1c). Numerous circRNAs seem to be specifically expressed in a cell type or developmental stage (Fig. 1b, c and Supplementary Fig. 1e). For example, hsa-circRNA 2149 is supported by 13 unique, head-to-tail spanning reads in CD191 leukocytes but is not detected in CD341 leukocytes (which were sequenced at comparable depth; Supplemen￾tary Table 1), neutrophils or HEK293 cells. Analogously, a number of nematode circRNAs seem to be expressed in oocytes but absent in 1- or 2-cell embryos. We annotated human circRNAs using the RefSeq database and a catalogue of non-coding RNAs27–29. 85% of human circRNAs align sense to known genes. Their splice sites typically span one to five exons (Supplementary Fig. 1g) and overlap coding exons (84%), but only in 65% of these cases are both splice sites that participate in the circularization known splice sites (Supplementary Table 2), demon￾strating the advantage of our strategy. 10% of all circRNAs align antisense to known transcripts, smaller fractions align to UTRs, introns, unannotated regions of the genome (Fig. 1d). Examples of human circRNAs are shown in Fig. 1e. We analysed sequence conservation within circRNAs. As genomic sequence is subject to different degrees of evolutionary selection, depending on function, we studied three subtypes of circRNAs. Intergenic and a few intronic circRNAs display a mild but significant enrichment of conserved nucleotides (Supplementary Fig. 1h, i). To analyse circRNAs composed of coding sequence and thus high overall conservation, we selected 223 human circRNAs with circular orthologues in mouse (Methods) and entirely composed of coding sequence. Control (linear) exons were randomly selected to match the level of conservation observed in first and second codon positions (Methods, Fig. 1f inset and Supplementary Fig. 1k for conservation of the remaining coding sequence (CDS)). circRNAs with conserved circularization were significantly more conserved in the third codon position than controls, indicating evolutionary constraints at the nuc￾leotide level, in addition to selection at the protein level (Fig. 1f and Supplementary Fig. 1j, k). In summary, we have confidently identified a large number of circRNAs with complex expression patterns, which derive often but not always from coding exons. Sequence conservation suggests that at least a subset contains functional sequence elements. Characterization of 50 predicted circRNAs We experimentally tested our circRNA predictions in HEK293 cells. Head-to-tail splicing was assayed by quantitative polymerase chain reaction (qPCR) after reverse transcription, with divergent primers and Sanger sequencing (Fig. 2a, b). Predicted head-to-tail junctions of 19 out of 23 randomly chosen circRNAs (83%) could be validated, demonstrating high accuracy of our predictions (Table 1). In contrast, 5 out of 7 (71%) candidates exclusively predicted in leukocytes could not be detected in HEK293 cells, validating cell-type-specific expression. Head-to-tail splicing could be produced by trans-splicing or geno￾mic rearrangements. To rule out these possibilities as well as potential PCR artefacts, we successfully validated the insensitivity of human circRNA candidates to digestion with RNase R—an exonuclease that degrades linear RNA molecules30—by northern blotting with probes which span the head-to-tail junctions (Fig. 2c). We quantified RNase R resistance for 21 candidates with confirmed head-to-tail splicing by AG GT Circularization 5′ anchor Acceptor Donor AG Acceptor GT Donor Spliced read a circRNA Linear splicing Anchor alignment and splice-site detection Sperm 2-cell embryo Oocyte 1-cell embryo b c Human (1,950) CD19+ HEK293 CD34+ Neutrophils C. elegans (724) (1,602 inside coding transcripts) Intergenic ncRNA Antisense Human circRNAs 152 183 90 59 51 30 31 22 20 20 16 16 16 13 5 106 28 79 12 22 89 20 25 3 19 21 194 939 60 333 81 CDS exons 1,000 5′UTR 195 21 3′UTR 79 27 Intronic 80 147 63 204 Other 53 d hsa-circRNA 6 hsa-circRNA 2 CDR1as hsa-circRNA 9 hsa-circRNA 1862 ZRANB1 exon1 Long ncRNA PVT1 exon3 CDR1 gene AFF1 exon4,5 GT GT GT GT GT AG AG AG AG AG Control cons. CDS circRNA Third codon position 123 2 1 0 Conservation *** 0 1 Conservation score Cumulative frequency 0 0.5 1 e f 3′ anchor Conservation 0.5 Intergenic chr4:42,212,391-42,214,180 Figure 1 | Detection, classification and evolutionary conservation of circRNAs. a, The termini of junction-spanning reads (anchors) align sequentially to the genome for linear (top) but in reversed orientation for head￾to-tail spliced reads (bottom). Spliced reads must distribute completely to anchors, flanked by AG/GU (Methods). b, c, circRNAs in human cell types (b) and nematode stages (c). d, Genomic origin of human circRNAs. A total of 96% of circRNAs overlap known transcripts. e, Examples of human circRNAs. TheAFF1 intron is spliced out (Supplementary Fig. 2e). Sequence conservation: placental mammals phyloP score (Methods), scale bar, 200 nucleotides. f, A total of 223 human coding sequence circRNAs with mouse orthologues (green) and controls (black) with matched conservation level (inset: mean conservation for each codon position (grey), controls (black); x axis, codon positions; y axis, placental mammals phyloP score; see also Methods and Supplementary Fig. 1j, k). Third codon positions are significantly more conserved (P , 4 3 10210, Mann–Whitney U-test, n 5 223). RESEARCH ARTICLE 2 | NATURE | VOL 000 | 00 MONTH 2013 ©2013 Macmillan Publishers Limited. All rights reserved

ARTICLE RESEARCH onvergent(4← ficantly enriched compared to coding sequences(P3% vinculin 120f21 vertebrate genomes( b)and high base-pairing probability within seed matches c).d, CDRlas RNA is cytoplasmic and disperse(white spots; single-mole RNA FISH; maximum intensity merges of Z-stacks). siSCR, positive; siRNAl Mouse(adult brain Head-to-tail splicing 3 of 3 negative control. Blue, nuclei(DAPI); scale bar, 5 um(see also Supr Fig 10 for uncropped images).e, Northern blotting detects circular but not Expression>1%β~ actin linear CDRlas in HEK293 RNA. Total, HEK293 RNA; circular, head-to-tail 15of20 probe; circ+lin, probe within splice sites; IVT lin, in vitro transcribed, linear CDRlas RNA f, Circular CDRlas is highly expressed (qPCR, error bars pression>1%ei·3d 120f15 indicate standard deviation). g, CDRlas. Blue, seed matches; dark red, AGO PAR-CLIP reads; bright red, crosslinked nucleotide conversions. 00 MONTH 2013 VOL 000 NATURE 3 @2013 Macmillan Publishers Limited. All rights reserved

qPCR. All of these were at least 10-fold more resistant than GAPDH (Fig. 2d and Supplementary Fig. 2a). We reasoned that circRNAs should generally turn over more slowly than mRNAs. Indeed, we found that 24 h after blocking transcription circRNAs were highly stable, exceeding the stability of the housekeeping gene GAPDH31 (Fig. 2e and Supplementary Fig. 2b). We also validated 3 out of 3 tested mouse circRNAs with human orthologues in mouse brains (Supplementary Fig. 2c). In C. elegans 15 out of 20 (75%) of the pre￾dictions from gametes and early embryos were validated in a mixed stage sample (Supplementary Fig. 2d and Supplementary Table 3). circRNA CDR1as is densely bound by AGO Stable transcripts with many miRNA-binding sites could function as miRNA sponges. We intersected our catalogue of circRNAs with transcript annotations, assuming that introns would not occur in mature circRNAs (as observed for 3 out of 3 tested circRNAs, Supplementary Fig. 2e). We screened for occurrences of conserved miRNA family seed matches (Methods). When counting repetitions of conserved matches to the same miRNA family, circRNAs were significantly enriched compared to coding sequences (P , 2.963 10222, Mann–Whitney U-test, n 5 3,873) or 39 UTR sequences (P , 2.76 3 10221, Mann–Whitney U-test, n 5 3,182) (Supplementary Fig. 3a, b). As an extreme case, we discovered that the known human circRNA CDR1as (ref. 9) harboured dozens of conserved miR-7 seed matches. To test whether CDR1as is bound by miRNAs, we analysed bio￾chemical, transcriptome-wide binding-site data for the miRNA effector AGO proteins. We performed four independent PAR-CLIP (photoactivatable-ribonucleoside-enhanced crosslinking and immu￾noprecipitation) experiments for human AGO (Methods) and ana￾lysed them together with published, lower-depth data32. PAR-CLIP32–34 is based on ultraviolet crosslinking of RNA to protein and subsequent sequencing of RNA bound to a RBP of interest. The ,1.5-kilobase (kb) CDR1as locus stood out in density and number of AGO PAR-CLIP reads (Fig. 3a), whereas nine combined PAR-CLIP libraries for other RBPs gave virtually no signal. Of note, there is no PAR-CLIP read mapping to the sense coding transcript of the CDR1 gene, which was originally identified as a target of autoantibodies from patients with paraneoplastic cerebellar degeneration35. Sequence analysis across 32 vertebrate species revealed that miR-7 is the only animal miRNA with conserved seed matches that can explain the AGO binding along the CDR1as transcript (Methods). Human CDR1as harbours 74 miR-7 seed matches of which 63 are Table 1 | Summary of the validation experiments Sample Validation experiment Validation success Human (HEK293) Head-to-tail splicing 19 of 23 Circularity 21 of 21 Expression .3% vinculin 12 of 21 Expression specificity (leukocyte specific) 5 of 7 Mouse (adult brain) Head-to-tail splicing 3 of 3 Circularity 3 of 3 Expression .1% b-actin 2 of 3 C. elegans Head-to-tail splicing 15 of 20 Circularity 13 of 13 Expression .1% eif-3.d 12 of 15 Most experimentally tested circRNAs are validated. 74 miR-7 seed matches AGO PAR-CLIP reads CDR1 antisense (1,485 nt) 5′ 3′ 5′ 3′ Splice junction miR-7 seed miR-7 (nt) 5′ to 3′ Probability 1.0 0.5 UGGAAGA CUAGUGAUUUUGUUGU Base pairing Distance from seed match (nt) –entropy (bits) –1 –1.8 –2.2 –8 GUCUUCCA +8 Seed match conservation miR-7 seed match PAR-CLIP controls (QKI, PUM2, ELAVL1) HEK293 expression relative to 18S (%) 1.0 0.2 0.6 Circular GAPDH CDR1as a b c d e f VCL g RNA gel 28S 18S 5S Circular circ+lin Total IVT lin. Total IVT lin. Total IVT lin. siSCR siRNA1 Mock Figure 3 | The circRNA CDR1as is bound by the miRNA effector protein AGO, and is cytoplasmic. a, CDR1as is densely bound by AGO (red) but not by unrelated proteins (black). Blue boxes indicate miR-7 seed matches. nt, nucleotides. b, c, miR-7 sites display reduced nucleotide variability across 32 vertebrate genomes (b) and high base-pairing probability within seed matches (c). d, CDR1as RNA is cytoplasmic and disperse (white spots; single-molecule RNA FISH; maximum intensity merges of Z-stacks). siSCR, positive; siRNA1, negative control. Blue, nuclei (DAPI); scale bar, 5 mm (see also Supplementary Fig. 10 for uncropped images). e, Northern blotting detects circular but not linear CDR1as in HEK293 RNA. Total, HEK293 RNA; circular, head-to-tail probe; circ1lin, probe within splice sites; IVT lin., in vitro transcribed, linear CDR1as RNA. f, Circular CDR1as is highly expressed (qPCR, error bars indicate standard deviation). g, CDR1as. Blue, seed matches; dark red, AGO PAR-CLIP reads; bright red, crosslinked nucleotide conversions. 5S 18S 28S RNase R exonuclease Agarose gel GAPDH CDR1as hsa-circRNAs 2 3 16 – + Divergent Convergent hsa-circRNA2 hsa-circRNA3 hsa-circRNA16 GAPDH cDNA gDNA gDNA gDNA gDNA 100 200 300 cDNA cDNA cDNA Sanger sequencing ... ... GAPDH p16 Mock 24 h ActD 0.0 0.5 1.0 Rel. expression CDR1as hsa-circRNAs 2 3 6 9 16 ZRANB1 exon1 hsa-circRNA2 ... ... a b c d GAPDH VCL CDR1as hsa-circRNAs 2 3 6 9 16 0.0 0.5 1.0 Rel. expression Mock RNase R e AG GT – + – + – + – + – + ( ) ( ) Figure 2 | CircRNAs are stable transcripts with robust expression. a, Human (hsa) ZRANB1 circRNA exemplifies the validation strategy. Convergent (divergent) primers detect total (circular) RNAs. Sanger sequencing confirms head-to-tail splicing. b, Divergent primers amplify circRNAs in cDNA but not genomic DNA (gDNA). GAPDH, linear control, size marker in base pairs. c, Northern blots of mock (2) and RNase R (1) treated HEK293 total RNA with head-to-tail specific probes for circRNAs. GAPDH, linear control. d, e, circRNAs are at least 10-fold more RNase R resistant than GAPDH mRNA (d) and stable after 24 h transcription block (e) (qPCR; error bars indicate standard deviation). ARTICLE RESEARCH 00 MONTH 2013 | VOL 000 | NATURE | 3 ©2013 Macmillan Publishers Limited. All rights reserved

RESEARCH ARTICLE conserved in at least one other species(Supplementary Fig. 4). CDR1as and miR-7 in mouse tissues Interspaced sequences were less conserved, indicating that miR-7 32gCERs analysis of predicted circRNA-mirna duplexes(Methods)showed 3 reduced base-pairing of miR-7 beyond the seed( Fig 3c). None of the 2 1, 500 miR-7 complementary sites across 32 vertebrate sequences 8 was complementary beyond position 12 of miR-7(only three could form an 11-nucleotide duplex)(Supplementary Table 4). Slicing by …x mammalian Argonaute requires complementarity of positions 10 and I and depends on extended complementarity beyond position 12 -124 control (ref. 36). Thus, CDRlas seems optimized to be densely bound but not liced by miR-7 Single-molecule imaging(Methods) revealed disperse and most ? ytoplasmic CDRlas expression(HEK293 cells), consistent with miRNA sponge function (Fig. 3d and Supplementary Table 5). Figure 4 CDRlas and miR-7 have ove CDRlas circularization was assayed by northern blotting(Fig. 3e). neuronal tissues. a, Among mouse tissues and MIN6 cells(qPCR,relative to Nicking experiments confirmed that CDRlas circRNA can be linea- cerebral cortex expression; error bars indicate standard deviations; see rized and degraded(Supplementary Fig. 5a). In RNA from HEK293 Supplementary Fig 9a for miR-122 control) neuronal tissues co-express miR-7 cells, circularized but no additional linear CDRlas was detected and CDRlas. b, In situ staining of CDRlas and miR-7 in mouse embryo brain (Supplementary Fig. 5b). Circular expression levels were quantified E13.5(U6 and miR-124, positive control; scrambled probe, negative control) by qPCR with divergent primers calibrated by standard curves (Supplementary Table 6). CDRlas was highly expressed(-15%to 20% of GAPDH expression, Fig 3f). Estimating GAPDH mRNA unknown consequences. This problem is circumvented when using zeb- copy number from HEK293 RNA-seq data(-1, 400 molecules per rafish( Danio rerio)as an animal modeL. According to our bioinformatic cell, data not shown)suggests that CDRlas may bind up to -20,000 analyses (not shown )zebrafish has lost the cdrl locus, whereas miR-7 is miR-7 molecules per cell( Fig. 3g) conserved and highly expressed in the embryonic l brain". Thus, we can If CDRlas functions as a miR-7 sponge, its destruction could trigger test whether miR-7 has a loss-of-function phenotype and if this pheno- downregulation of miR-7 targets. We knocked down CDRlas in type can be induced by introduction of mammalian CDRlas rNA We HEK293 cells and monitored of published miR-7 targets injected morpholinos to knock down mature mir-7 expression in zebra by q PCR with externally spiked-in standards(Methods and Supplemen- fish embryos(Methods). At a dose of 9 ng of miR-7 morpholino, the ary Fig 5c, d). All eight miR-7 targets assayed, but also housekeeping embryos did not show overall morphological defects but reproducibly genes, were downregulated Nanostring technology ?7 additionally indi- and in two independent genetic backgrounds(Supplementary Fig 6a-c) ated downregulation of many genes(data not shown). Furthermore, developed brain defects(Fig, 5a, b). In particular, -70% showed a con- stable loss of CDRlas expression by virally delivered small hairpin sistent and clear reduction in midbrain size, and an additional-5% of RNAs led to significantly reduced migration in an in vitro wound clo- animals had almost completely lost their midbrains. Of note, the tel- sure assay(Methods, Supplementary Fig, 5e, f and Supplementary encephalon at the anterior tip of the brain was not affected in size. Brain Table 7). Thus, knockdown of CDRlas affects HEK293 cells, but we volumes were also measured based on confocal three-dimensional stacks could not delineate miR-7-specificeffects, potentially because ofindirect (Fig. 5c and Supplementary Fig. 7). Reduction of the midbrain size or miR-7-independent CDRlas function(see below). correlated with miR-7 inhibition in the respective animals(Supplemen tary Fig. 6d). These data provide evidence that miR-7 loss-of-function Co-expression of miR-7 and CDRlas in brain causes a specific reduction of midbrain size If CDRlas indeed interacts with miR-7, both must be co-expressed. To test whether CDRIas can function as a miR-7 sponge in vivo,we miR-7 is highly expressed in neuronal tissues, pancreas and pituit injected embryos with plasmid DNA that expressed a linear version of glands. apart from HEK293 cells, a cell line probably derived from the full-length human CDRlas sequence(Supplementary Fig 6e, f)or neuronal precursors in embryonic kidney 9, we quantified miR-7 and a plasmid provided by the Kjems laboratory that can produce circular CDRlas expression across mouse tissues and pancreatic- island- CDRIas in human cells(Fig. 5d, e). q PCR analysis detected circular derived MIN6 cells(Methods and Fig 4a). CDRlas and miR-7 were RNA in zebrafish embryos injected with the latter plasmid (Sup both highly expressed in brain tissues, but CDRlas was expressed at plementary Fig 8), which reproducibly and in independent genetic low levels or absent in non-neuronal tissues, including tissues with backgrounds lead to reduced midbrain sizes(Fig. 5g, h). Similarly very high miR-7 expression. qPCR suggested that CDRlas is exclu- animals injected with in vitro-transcribed partial mouse CDRlas sively circular in adult and embryonic mouse brain( Supplemen RNA, but not with RNA from the other strand, showed significant Fig. 5g, h). Thus, CDRlas and miR-7 seem to interact specifically midbrain reduction(Supplementary Fig 6g-i). Thus, the phenotype in neuronal tissues. Indeed, when assaying CDRlas and miR-7 in is probably caused by CDRlas rna and not by an unspecific effect of mouse brains by in situ hybridizations(Methods), we observed RNA or DNA injection. These results provide evidence that human/ cific, similar, but not identical, expression patterns in the brain of mouse CDRlas transcripts are biologically active in vivo and impai mid-gestation(embryonic day 13.5(E13.5)embryos( Fig 4b). Speci- brain development similarly to miR-7 inhibition. The midbrain fically, CDRlas and miR-7 were highly co-expressed in areas of the reduction could be partially rescued by inje developing midbrain(mesencephalon)*04. Thus, CDRlas is highly (Fig. 5f, g), arguing that the biological effect of CDRlas expression expressed, stable, cytoplasmic, not detectable as a linear RNA and is caused at least in part by interaction of CDRlas with miR-7 shares expression domains with miR-7. Together with extensive miR-7 binding within CDRlas, CDRlas has hallmarks of a potent Discussion We have shown that animal genomes express thousands of circRNAs from diverse genomic locations(for example, from coding and non- Effects of miR-7 and CDRlas in zebrafish coding exons, intergenic regions or transcripts antisense to 5 system. However, a knockout would also affect CDRI protein, with specific manner h plex tissue, cell-type- or developmental-stage- It would be informative to knock out CDRlas in an animal model 3 UTRs)in a cor rovided evidence that cDrlas can act as a 4I NATURE I VOL 00000 MONTH 2013 @2013 Macmillan Publishers Limited. All rights reserved

conserved in at least one other species (Supplementary Fig. 4). Interspaced sequences were less conserved, indicating that miR-7 binding sites are probably functional (Fig. 3b). Secondary structure analysis of predicted circRNA–miRNA duplexes (Methods) showed reduced base-pairing of miR-7 beyond the seed (Fig. 3c). None of the ,1,500 miR-7 complementary sites across 32 vertebrate sequences was complementary beyond position 12 of miR-7 (only three could form an 11-nucleotide duplex) (Supplementary Table 4). Slicing by mammalian Argonaute requires complementarity of positions 10 and 11 and depends on extended complementarity beyond position 12 (ref. 36). Thus, CDR1as seems optimized to be densely bound but not sliced by miR-7. Single-molecule imaging (Methods) revealed disperse and mostly cytoplasmic CDR1as expression (HEK293 cells), consistent with miRNA sponge function (Fig. 3d and Supplementary Table 5). CDR1as circularization was assayed by northern blotting (Fig. 3e). Nicking experiments confirmed that CDR1as circRNA can be linea￾rized and degraded (Supplementary Fig. 5a). In RNA from HEK293 cells, circularized but no additional linear CDR1as was detected (Supplementary Fig. 5b). Circular expression levels were quantified by qPCR with divergent primers calibrated by standard curves (Supplementary Table 6). CDR1as was highly expressed (,15% to ,20% of GAPDH expression, Fig. 3f). Estimating GAPDH mRNA copy number from HEK293 RNA-seq data (,1,400 molecules per cell, data not shown) suggests that CDR1as may bind up to ,20,000 miR-7 molecules per cell (Fig. 3g). If CDR1as functions as a miR-7 sponge, its destruction could trigger downregulation of miR-7 targets. We knocked down CDR1as in HEK293 cells and monitored expression of published miR-7 targets by qPCR with externally spiked-in standards (Methods and Supplemen￾tary Fig. 5c, d). All eight miR-7 targets assayed, but also housekeeping genes, were downregulated. Nanostring technology37 additionally indi￾cated downregulation of many genes (data not shown). Furthermore, stable loss of CDR1as expression by virally delivered small hairpin RNAs led to significantly reduced migration in an in vitro wound clo￾sure assay (Methods, Supplementary Fig. 5e, f and Supplementary Table 7). Thus, knockdown of CDR1as affects HEK293 cells, but we could not delineate miR-7-specific effects, potentially because of indirect or miR-7-independent CDR1as function (see below). Co-expression of miR-7 and CDR1as in brain If CDR1as indeed interacts with miR-7, both must be co-expressed. miR-7 is highly expressed in neuronal tissues, pancreas and pituitary gland38. Apart from HEK293 cells, a cell line probably derived from neuronal precursors in embryonic kidney39, we quantified miR-7 and CDR1as expression across mouse tissues and pancreatic-island￾derived MIN6 cells (Methods and Fig. 4a). CDR1as and miR-7 were both highly expressed in brain tissues, but CDR1as was expressed at low levels or absent in non-neuronal tissues, including tissues with very high miR-7 expression. qPCR suggested that CDR1as is exclu￾sively circular in adult and embryonic mouse brain (Supplementary Fig. 5g, h). Thus, CDR1as and miR-7 seem to interact specifically in neuronal tissues. Indeed, when assaying CDR1as and miR-7 in mouse brains by in situ hybridizations (Methods), we observed spe￾cific, similar, but not identical, expression patterns in the brain of mid-gestation (embryonic day 13.5 (E13.5)) embryos (Fig. 4b). Speci￾fically, CDR1as and miR-7 were highly co-expressed in areas of the developing midbrain (mesencephalon)40,41. Thus, CDR1as is highly expressed, stable, cytoplasmic, not detectable as a linear RNA and shares expression domains with miR-7. Together with extensive miR-7 binding within CDR1as, CDR1as has hallmarks of a potent circular miR-7 sponge in neuronal tissues. Effects of miR-7 and CDR1as in zebrafish It would be informative to knock out CDR1as in an animal model system. However, a knockout would also affect CDR1 protein, with unknown consequences. This problem is circumvented when using zeb￾rafish (Danio rerio) as an animal model. According to our bioinformatic analyses (not shown) zebrafish has lost the cdr1 locus, whereas miR-7 is conserved and highly expressed in the embryonic brain42. Thus, we can test whether miR-7 has a loss-of-function phenotype and if this pheno￾type can be induced by introduction of mammalian CDR1as RNA. We injected morpholinos to knock down mature miR-7 expression in zebra￾fish embryos (Methods). At a dose of 9 ng of miR-7 morpholino, the embryos did not show overall morphological defects but reproducibly, and in twoindependent genetic backgrounds (Supplementary Fig. 6a–c), developed brain defects (Fig. 5a, b). In particular, ,70% showed a con￾sistent and clear reduction in midbrain size, and an additional ,5% of animals had almost completely lost their midbrains. Of note, the tel￾encephalon at the anterior tip of the brain was not affected in size. Brain volumes were also measured based on confocal three-dimensional stacks (Fig. 5c and Supplementary Fig. 7). Reduction of the midbrain size correlated with miR-7 inhibition in the respective animals (Supplemen￾tary Fig. 6d). These data provide evidence that miR-7 loss-of-function causes a specific reduction of midbrain size. To test whether CDR1as can function as a miR-7 sponge in vivo, we injected embryos with plasmid DNA that expressed a linear version of the full-length human CDR1as sequence (Supplementary Fig. 6e, f) or a plasmid provided by the Kjems laboratory that can produce circular CDR1as in human cells (Fig. 5d, e). qPCR analysis detected circular RNA in zebrafish embryos injected with the latter plasmid (Sup￾plementary Fig. 8), which reproducibly and in independent genetic backgrounds lead to reduced midbrain sizes (Fig. 5g, h). Similarly, animals injected with in vitro-transcribed partial mouse CDR1as RNA, but not with RNA from the other strand, showed significant midbrain reduction (Supplementary Fig. 6g–i). Thus, the phenotype is probably caused by CDR1as RNA and not by an unspecific effect of RNA or DNA injection. These results provide evidence that human/ mouse CDR1as transcripts are biologically active in vivo and impair brain development similarly to miR-7 inhibition. The midbrain reduction could be partially rescued by injecting miR-7 precursor (Fig. 5f, g), arguing that the biological effect of CDR1as expression is caused at least in part by interaction of CDR1as with miR-7. Discussion We have shown that animal genomes express thousands of circRNAs from diverse genomic locations (for example, from coding and non￾coding exons, intergenic regions or transcripts antisense to 59 and 39 UTRs) in a complex tissue-, cell-type- or developmental-stage￾specific manner. We provided evidence that CDR1as can act as a CDR1as and miR-7 expression in mouse tissues CDR1as expression CDR1as miR-7 400 miR-7 expression 0 1 2 1 10 600 Hippocampus Cer. cortex Cerebellum Forebrain Midbrain Kidney Lung Skel. muscle Spleen Pancreas Liver Colon Heart Pituitary gland MIN6 cells 5 miR-7 U6 control miR-124 control CDR1as Scrambled a b Figure 4 | CDR1as and miR-7 have overlapping and specific expression in neuronal tissues. a, Among mouse tissues and MIN6 cells (qPCR, relative to cerebral cortex expression; error bars indicate standard deviations; see Supplementary Fig. 9a for miR-122 control) neuronal tissues co-express miR-7 and CDR1as. b, In situ staining of CDR1as and miR-7 in mouse embryo brain E13.5 (U6 and miR-124, positive control; scrambled probe, negative control). Scale bar, 1 mm. RESEARCH ARTICLE 4 | NATURE | VOL 000 | 00 MONTH 2013 ©2013 Macmillan Publishers Limited. All rights reserved

ARTICLE RESEARCH Control Mo 15 ng MO g CDRlas. Thus, CDRlas may function to transport miR-7 to subcel- lular locations, where miR-671 could trigger release of its cargo. Known functions of miR-7 targets such as PAKI and FAKI support these peculation The phenotype induced by CDRlas expression in zebrafish was only partially rescued by expressing miR-7, indicating that CDRlas could have functions beyond sequestering miR-7. This idea is sup- ported by in situ hybridization in mouse adult hippocampus(Su ed Me plementary Fig 9b)where areas staining for CDRlas but not miR- were observed What could be additional functions of circrnas IC egfp) yond acting as sponges? As a single-stranded RNA, CDRlas could Circular CDR1as Circular CDRlas +miR-7 for example, bind in trans 3 UTRs of target mRNAs to regul expression. It is even possible that miR-7 binds CDRlas to these trans-acting activities. Alternatively, CDRlas could be in the assembly of larger complexes of RNA or protein, similar to other low-complexity molecules How many other circRNAs exist? In this study, we identifie ximately 2,000 human, 1,900 mouse and 700 nematode from sequencing data, and our validation experiments most of the 50 tested circRNAS. However, we analysed only a few tissues/developmental stages with stringent cutoffs. Thus, the true number of circRNAs is almost certainly much larger. Although 图 h且x星 CDRlas is an extreme cas circRNAs have conserved seed matches. For example, circRNA from the SRY locus has seed sites for murine miRNAs. Therefore, circRNAs probably compete with 圖 22 other RNAs for miRNA binding. Sequence analyses indicated that oding exons serve additional, presumably regulatory functions when expressed within circRNAs, whereas intergenic or intronic circRNAs generally showed only weak conservation. Because we detected thou Mo (ng) CDR1 larization of exons is easy to evolve and may provide a mechanism Figure 5 In zebrafish, knockdown of miR-7 or expression of CDRlas for rapid evolution of stably and well expressed regulatory RNAs Of uses midbrain defects. a, b, Neuronal ter(Tg(huc egp)embryos(top, note, we detected multiple seed matches for viral miRNAs within light microscopy)48 h post fertilization( bottom, representative confoc human circRNAs(not shown). However, there is no reason to think stack projections; blue dashed line, telencephalon(TC)(control); yellow that circRNAs function predominantly to bind miRNAs. As known dashed line, midbrain(MB). Embryos after injection of 9 ng miR-7 in bacteria, the decoy mechanism underlying miRNA sponges could morpholino(MO)(b)display a reduction in midbrain size. Panel a shows a be important also for RBPs"647. Similarly, circRNAs could function to store, sort, or localize RBPs. In summary, our data suggest that vector encoding human circular CDRlas. f, Rescue experiment with miR-7 circRNAs form a class of post-transcriptional regulators which com- dimensional volumetric reconstructions. d, Empty vector control e, Expr precursor g, Phenotype penetrance(% of embryos, miR-7 MO, n=135: pete with other RNAs for binding by miRNAs and RBPs and may 91; linear CDRlas, n=258; circular generally function in modulating the local free concentration of RBPs, CDRlas, n=153; circular CDRlas plus miR-7 precursor, n=217). Phenotype RNAS, or their binding sites distribution derived from at least three independent experiments. Scale bar, Note added in proof: While this paper was under review, circular 0.1 mm. *P<ool: ***p< o0o1 in Students t-test for normal midbrain RNAs in fibroblasts were described reduced midbrain(see also Supplementary Fig. 6).h, Phenotype quantification (Methods). Error bars indicate standard deviation n= 3 per group. METHODS SUMMARY CDRlas is densely bound by miRNA effector molecules;(2)CDRlas f ode e al pipeline for predicting circRNAs from ribominus sequencing post-transcriptional regulator by binding miR-7 in brain tissues:(1) data. a detailed description of the computational methods is given in the is expressed highly, stably and mostly cytoplasmic: (4)CDRlas and T-REx (Life Technologies) were caultured-following standard protocols. Tran. entre Biotechnologies)treatment (3U ug )was performed on total RNA (5)human/mouse CDRlas is circularized in vivo and is not detectable (5 ug)at 37C for 15 min qPCR primers are listed in Supplementary Table 8 as a linear molecule;(6) human/mouse CDRlas sequences, when Single-molecule RNA fluorescence in situ hybridization(smRNA FISH) injected into zebrafish, and miR-7 knock down have similar pheno- Stellaris Oligonucleotide probes complementary to CDRlas were designed using types in brain. While zebrafish circularization of human CDRlas may the Stellaris Probe Designer(Biosearch Technologies). Probe pools were obtained be incomplete, the midbrain phenotype was stronger compared to from Bio Cat GmbH as conjugates coupled to Quasar 670 Probes were hybridized expressing linear CDRlas RNA that lacks circularization splice sites. at 125 nM at 37 C Images were acquired on an inverted Nikon Ti microscope Although the two DNA plasmids used carry identical pre s and Mouse strains and in situ hybridization. In situ hybridization(ISH)was per- that the difference in midbrain phenotype strength may be explained using locked nucleic acid (LNA) probes or RNAs obtained by in vitro transcrip- by other factors. However, because of the observed extreme stability zebrafish methods. Tg(hu C egfp )and Tg(Xia. Tubb: ds RED)transgenic zebrafish of CDRlas and circRNAs in general, our data argue that circrNAs lines were used 9 0. Morpholino antisense oligomers were injected into the yolk of an be used as potent inhibitors of miRNAs or RBPs. Future studies single-cell-stage embryos. Furthermore, two pCS2+ plasmids coding for full nould elucidate how CDRlas can be converted into a linear mole- length linear CDRlas or CDRlas plus upstream and downstream sequence that ule and targeted for degradation. miR-671 can trigger destruction of can express circular CDRlas in human cells(courtesy of the Kjems laboratory) 00 MONTH 2013 VOL 000 NATURE I @2013 Macmillan Publishers Limited. All rights reserved

post-transcriptional regulator by binding miR-7 in brain tissues: (1) CDR1as is densely bound by miRNA effector molecules; (2) CDR1as harbours 74 miR-7 seed matches, often deeply conserved; (3) CDR1as is expressed highly, stably and mostly cytoplasmic; (4) CDR1as and miR-7 share specific expression domains in mouse embryonic brain; (5) human/mouse CDR1as is circularized in vivo and is not detectable as a linear molecule; (6) human/mouse CDR1as sequences, when injected into zebrafish, and miR-7 knock down have similar pheno￾types in brain. While zebrafish circularization of human CDR1as may be incomplete, the midbrain phenotype was stronger compared to expressing linear CDR1as RNA that lacks circularization splice sites. Although the two DNA plasmids used carry identical promoters and were injected in equal concentrations, we cannot rule out the possibility that the difference in midbrain phenotype strength may be explained by other factors. However, because of the observed extreme stability of CDR1as and circRNAs in general, our data argue that circRNAs can be used as potent inhibitors of miRNAs or RBPs. Future studies should elucidate how CDR1as can be converted into a linear mole￾cule and targeted for degradation. miR-671 can trigger destruction of CDR1as9 . Thus, CDR1as may function to transport miR-7 to subcel￾lular locations, where miR-671 could trigger release of its cargo. Known functions of miR-7 targets such as PAK1 and FAK1 support these speculations43,44. The phenotype induced by CDR1as expression in zebrafish was only partially rescued by expressing miR-7, indicating that CDR1as could have functions beyond sequestering miR-7. This idea is sup￾ported by in situ hybridization in mouse adult hippocampus (Sup￾plementary Fig. 9b) where areas staining for CDR1as but not miR-7 were observed. What could be additional functions of circRNAs beyond acting as sponges? As a single-stranded RNA, CDR1as could, for example, bind in trans 39 UTRs of target mRNAs to regulate their expression. It is even possible that miR-7 binds CDR1as to silence these trans-acting activities. Alternatively, CDR1as could be involved in the assembly of larger complexes of RNA or protein, perhaps similar to other low-complexity molecules45. How many other circRNAs exist? In this study, we identified appro￾ximately 2,000 human, 1,900 mouse and 700 nematode circRNAs from sequencing data, and our validation experiments confirmed most of the 50 tested circRNAs. However, we analysed only a few tissues/developmental stages with stringent cutoffs. Thus, the true number of circRNAs is almost certainly much larger. Although CDR1as is an extreme case, many circRNAs have conserved seed matches. For example, circRNA from the SRY locus6 has seed sites for murine miRNAs. Therefore, circRNAs probably compete with other RNAs for miRNA binding. Sequence analyses indicated that coding exons serve additional, presumably regulatory functions when expressed within circRNAs, whereas intergenic or intronic circRNAs generally showed only weak conservation. Because we detected thou￾sands of circRNAs, it is appealing to speculate that occasional circu￾larization of exons is easy to evolve and may provide a mechanism for rapid evolution of stably and well expressed regulatory RNAs. Of note, we detected multiple seed matches for viral miRNAs within human circRNAs (not shown). However, there is no reason to think that circRNAs function predominantly to bind miRNAs. As known in bacteria, the decoy mechanism underlying miRNA sponges could be important also for RBPs46,47. Similarly, circRNAs could function to store, sort, or localize RBPs. In summary, our data suggest that circRNAs form a class of post-transcriptional regulators which com￾pete with other RNAs for binding by miRNAs and RBPs and may generally function in modulating the local free concentration of RBPs, RNAs, or their binding sites. Note added in proof: While this paper was under review, circular RNAs in fibroblasts were described51. METHODS SUMMARY Computational pipeline for predicting circRNAs from ribominus sequencing data. A detailed description of the computational methods is given in the Methods. Cell culture and treatments. HEK293, HEK293TN and HEK293 Flp-In 293 T-REx (Life Technologies) were cultured following standard protocols. Tran￾scription was blocked by adding 2 mg ml21 actinomycin D (Sigma). RNase R (Epicentre Biotechnologies) treatment (3 U mg21 ) was performed on total RNA (5 mg) at 37 uC for 15 min. qPCR primers are listed in Supplementary Table 8. Single-molecule RNA fluorescence in situ hybridization (smRNA FISH). Stellaris Oligonucleotide probes complementary to CDR1as were designed using the Stellaris Probe Designer (Biosearch Technologies). Probe pools were obtained from BioCat GmbH as conjugates coupled to Quasar 670. Probes were hybridized at 125 nM at 37 uC. Images were acquired on an inverted Nikon Ti microscope. Mouse strains and in situ hybridization. In situ hybridization (ISH) was per￾formed on paraffin tissue sections from B6129SF1/J wild-type mice as described48 using locked nucleic acid (LNA) probes or RNAs obtained by in vitro transcrip￾tion on PCR products. Zebrafish methods. Tg(huC:egfp) and Tg(Xia.Tubb:dsRED) transgenic zebrafish lines were used49,50. Morpholino antisense oligomers were injected into the yolk of single-cell-stage embryos. Furthermore, two pCS21 plasmids coding for full￾length linear CDR1as or CDR1as plus upstream and downstream sequence that can express circular CDR1as in human cells (courtesy of the Kjems laboratory) Empty vector Circular CDR1as Circular CDR1as + miR-7 a Control MO 15 ng b c d e MB TC f g 3D reconstruction Control MO miR-7 MO Uninjected Linear CDR1as Circular CDR1as Volume (×106 μm3) 1 2 ** *** ** MB h TC ** Phenotype (%) 0 Empty vector Linear Circular Circular + miR-7 miR-7 (9 ng) Control (15 ng) Uninjected ****** ** ** 100 *** 50 Normal Reduced MB No MBmiR-7 (15 ng) Normal MB Reduced MB miR-7 MO 9 ng MO (ng) CDR1as 3 Tg(huC:egfp) f Circular CDR1as + miR-7 TC Figure 5 | In zebrafish, knockdown of miR-7 or expression of CDR1as causes midbrain defects. a, b, Neuronal reporter (Tg(huC:egfp)) embryos (top, light microscopy) 48 h post fertilization (bottom, representative confocal z-stack projections; blue dashed line, telencephalon (TC) (control); yellow dashed line, midbrain (MB)). Embryos after injection of 9 ng miR-7 morpholino (MO) (b) display a reduction in midbrain size. Panel a shows a representative embryo injected with 15 ng control morpholino. c, Three￾dimensional volumetric reconstructions. d, Empty vector control. e, Expression vector encoding human circular CDR1as. f, Rescue experiment with miR-7 precursor. g, Phenotype penetrance (% of embryos, miR-7 MO, n 5 135; uninjected, n 5 83; empty vector, n 5 91; linear CDR1as, n 5 258; circular CDR1as, n 5 153; circular CDR1as plus miR-7 precursor, n 5 217). Phenotype distribution derived from at least three independent experiments. Scale bar, 0.1 mm. **P , 0.01; ***P , 0.001 in Students t-test for normal midbrain, reduced midbrain (see also Supplementary Fig. 6). h, Phenotype quantification (Methods). Error bars indicate standard deviation n 5 3 per group. ARTICLE RESEARCH 00 MONTH 2013 | VOL 000 | NATURE | 5 ©2013 Macmillan Publishers Limited. All rights reserved

RESEARCH ARTICLE were injected Confocal imaging was performed using Carl Zeiss Microimaging. 31. lwai, Y, Akahane, K Pluznik, D.H. Cohen, R B Ca" ionophore A23187 Reduced midbrain development was defined as >50% smaller than the mean size of controls. Each experimental group was evaluated in at least three independent experiments; a minimum of 80 individual embryos per group was examined the 3-untranslated region. J ImmunoL 15 Full Methods and any associated references are available in the online version of licroRNA target sites by PAR-CuIP Cel 141, 129-141(2010). criptome-wide analysis of regulatory interactions of the 34. Baltz, A G et al. The mRNA-bound lobal occupancy profile Received er 2012; accepted 24 January 2013. protein-coding transcripts. MoL Cell 46, 674-690(2012) 1. Sanger, H L, Klotz, G, Riesner, D, Gross, H.J.& Kleinschmidt, A K Viroids are paired rod-like structures. Proc. Nat/ Acad. Sci. USA 73, 3852-3856 ighly base- 36.Wee, LM.Flores-Jasso,CF,Salomon, WE&Zamore,P.DArgonaute divides its 12) distinct functions and RNA-binding properties. Cel Ce/l precursor is converted to a circular RNA in isolated nuclei of Tetrahymena. 37. Geiss, G.K. et al. Direct multiplexed measure of gene expression with color- anan, M, Schwartz, s, Edelheit, S& Sorek, R T ne-wide discovery of 38. Landgraf, P. et al. A ma roRNA expression atlas based on small RNA (2012) library sequencing Ce// 129, 1401-1414(2007) 4. Ni B, Hetuin, D. Bailleul, B Mis-splicing yields circular 39. Shaw, G, Morse, S M. Graham, F L Preferential transformation of human molecules. FASEBJ. 7, 155-160(1993). 15如即a23 mouse test Cell 73. 1019-1035 (993. stis-determining gene sry in aduit 40. Kaufman. M 1 H. Bard, J B. L The Anatomical Basis of Mouse Development 7. Chao, C W, Chan, D. C Kuo, A& Leder, P. The mouse formin( mn)gene 41. Schambra, U Prenatal Mouse Brain Atlas(Springer, 2008). abundant circular RNAtranscripts and gene-targeted deletion analysis. Mo. Med. 42. Kapsimali, M et al MicroRNAs show a wide diversity profiles in the 8. Burd, C Eet al. Expression of linear and novel circular forms of an INK4/ARF. developing and mature central nervous system. Genome BioL 8, R173(2007) 43. Jacobs, T. et al. Localized activation of p ted kinase controls neuronal sociated non-coding RNA correlates with atherosclerosis risk. PLoS Genet 6, polarity and morphology. J Neurosci. 27, 8604-8615(200 an, J, Gawad, C, Wang, P L, Lacayo, N& Brown, P 0. Circular RNAs are th ant transcript isoform from hundreds of human genes in diverse cell domains form dynamic fibers within hydrogels Cell 149, 753-767(201 types. PLoS ONE 7, e30733(2012) 1. Ambros, V The functions of animal microRNAs. Nature 431, 350-355(2004). oding RNA molecule CsrB. MoL Microbiol. 29, 1321-1330(1998). 12. Baek, D et al. The impact of microRNAs on protein output Nature 455, 64-71 47. Gottesman, S sof Escherichia coli roles and mechanism 13. Selbach,Meta. Widespread changes in protein synthesis induced by microRNAs. 48. Huelsken, J, vogel, R- Erdmann, B, Cotsarelis, G. Birchmeier, W. B-Cat 14. Bartel, D P MicroRNAs: target recognition and regulatory functions. Cell 136, 49. Park, H C et al Analysis of upstream elements in the HuC promoter leads to the 15. Krek, A et al. Combinatorial microRNA target predictions. Nature Genet 37, establishment of transgenic zebrafish with fluorescent neurons. Dev Biol. 227. 79-293(2000 . 6.Lewis,BP, Burge,CB&Bartel, D.P. Conserved seed pairing, often flanked by 50. Peri, F& Nusslein- Volhard, C Live imaging of neuronal degradation by microglia that thousands of human genes are microRNA targets. Cell reveals a role for vo-ATPase al in phagosomal fusion in vivo Ce/l 133, 916-927 edma r c. nsn of several mamm4343835)and3y51)水WR成C四,厂 RNAs are abundant, conserved, and associated with ALU 18mBN19.228 Supplementary Information is available in the online version of the cular human CDRlas for our zebrafish experiments We t ndependent function of gene and pseudogene mRNAs facility. We thankA. lvanov for assisting in bioinformaticanalysis. NRtha 22. Tay. suppressor PTEN by ing funding sources: PhD program endogenous mRNAs Cell 147, 344-357(2011). Mr:RM project 1210182, MIRNAs as therapeutic targets(AE): DFG trog@T M, F.T. LH.G): the 23. Cesana, M. et al A long noncoding RNA controls muscle differentiation b functioning as a competing endogenous RNA Cell 147, 358-369(2011) KF0218(UZ): Helmholtz Association for the ' MDC Systems Biology Network, MSBI 24. 2b R&&-R85ha P A Emerging roles for natural microRNA sponges. CulT. Biol Bertin(K. FIN.). Funding for the group of ML is supported by BMBF-funding for the 25. Vivancos deep sequencing of the transcriptome Genome Res 20, 989-999(2010) Author Contributions S M, MJ, AE and F T contributed equally. S M. performed 26. Huang, R et al. An RNA-Seq strategy to detect the complete co nscriptome including full-length imprinted macro ncRNAs. PLOS ONE 6, man out most of the computation, 27. Kent W.J.etal. The human genome browser at UcSC Genome Res, 12.,996-1006 AG pARC P R, erformed all northem experiments. LH.G. and M. M. designed and carried out the single molecule experi sova, T& Maglott, D R NCBI Reference Sequence(Refseq): a UZ. performed the mouse part together with A L K contributed the zebrafish experiments upervised by FIN N.R. designed and supervised the project N.R. and MJ. wrote the Nucleic Acids Res. 33, D501-D504(2005) 29. Cabili, M. N. et al. Integrative a RNAs reveals global properties and specific subclasses. Genes Dev. 25, Author Information Sequencing data have been de 1915-1927(2011) 30.SuzukiH.etal.CharacterizationofRnaSeR-digestedcellularRnasourcethatwww.nature.coma are no competing financial interests. consists of lariat and circular RNAs from splicing. Nucleic Acids Res 34, Readers are welcome to comment on the online version of the paper Correspondence e63(2006 and requests for materials should be addressed to N.R. (rajewsky@mdc- berlin. de). 6I NATURE I VOL 00000 MONTH 2013 @2013 Macmillan Publishers Limited. All rights reserved

were injected. Confocal imaging was performed using Carl Zeiss MicroImaging. Reduced midbrain development was defined as .50% smaller than the mean size of controls. Each experimental group was evaluated in at least three independent experiments; a minimum of 80 individual embryos per group was examined. Full Methods and any associated references are available in the online version of the paper. Received 11 September 2012; accepted 24 January 2013. Published online 27 February 2013. 1. Sanger, H. L., Klotz, G., Riesner, D., Gross, H. J. & Kleinschmidt, A. K. Viroids are single-stranded covalently closed circular RNA molecules existing as highly base￾paired rod-like structures. Proc. Natl Acad. Sci. USA 73, 3852–3856 (1976). 2. Grabowski, P. J., Zaug, A. J. & Cech, T. R. The intervening sequence of the ribosomal RNA precursor is converted to a circular RNA in isolated nuclei of Tetrahymena. Cell 23, 467–476 (1981). 3. Danan, M., Schwartz, S., Edelheit, S. & Sorek, R. Transcriptome-wide discovery of circular RNAs in Archaea. Nucleic Acids Res. 40, 3131–3142 (2012). 4. Nigro, J. M. et al. Scrambled exons. Cell 64, 607–613 (1991). 5. Cocquerelle, C., Mascrez, B., Hetuin, D. & Bailleul, B. Mis-splicing yields circular RNA molecules. FASEB J. 7, 155–160 (1993). 6. Capel, B. et al. Circular transcripts of the testis-determining gene Sry in adult mouse testis. Cell 73, 1019–1030 (1993). 7. Chao, C. W., Chan, D. C., Kuo, A. & Leder, P. The mouse formin (Fmn) gene: abundant circular RNA transcripts and gene-targeted deletion analysis. Mol. Med. 4, 614–628 (1998). 8. Burd, C. E. et al. Expression of linear and novel circular forms of an INK4/ARF￾associated non-coding RNA correlates with atherosclerosis risk. PLoS Genet. 6, e1001233 (2010). 9. Hansen, T. B. et al. miRNA-dependent gene silencing involving Ago2-mediated cleavage of a circular antisense RNA. EMBO J. 30, 4414–4422 (2011). 10. Salzman, J., Gawad, C., Wang, P. L., Lacayo, N. & Brown, P. O. Circular RNAs are the predominant transcript isoform from hundreds of human genes in diverse cell types. PLoS ONE 7, e30733 (2012). 11. Ambros, V. The functions of animal microRNAs. Nature 431, 350–355 (2004). 12. Baek, D. et al. The impact of microRNAs on protein output. Nature 455, 64–71 (2008). 13. Selbach, M. et al.Widespread changes in protein synthesis induced bymicroRNAs. Nature 455, 58–63 (2008). 14. Bartel, D. P. MicroRNAs: target recognition and regulatory functions. Cell 136, 215–233 (2009). 15. Krek, A. et al. Combinatorial microRNA target predictions. Nature Genet. 37, 495–500 (2005). 16. Lewis, B. P., Burge, C. B. & Bartel, D. P. Conserved seed pairing, often flanked by adenosines, indicates that thousands of human genes are microRNA targets. Cell 120, 15–20 (2005). 17. Xie, X. et al. Systematic discovery of regulatory motifs in human promoters and 39 UTRs by comparison of several mammals. Nature 434, 338–345 (2005). 18. Friedman, R. C., Farh, K. K., Burge, C. B. & Bartel, D. P. Mostmammalian mRNAs are conserved targets of microRNAs. Genome Res. 19, 92–105 (2009). 19. Ebert, M. S., Neilson, J. R. & Sharp, P. A. MicroRNA sponges: competitive inhibitors of small RNAs in mammalian cells. Nature Methods 4, 721–726 (2007). 20. Franco-Zorrilla, J. M. et al. Target mimicry provides a new mechanism for regulation of microRNA activity. Nature Genet. 39, 1033–1037 (2007). 21. Poliseno, L. et al. A coding-independent function of gene and pseudogene mRNAs regulates tumour biology. Nature 465, 1033–1038 (2010). 22. Tay, Y. et al. Coding-independent regulation of the tumor suppressor PTEN by competing endogenous mRNAs. Cell 147, 344–357 (2011). 23. Cesana, M. et al. A long noncoding RNA controls muscle differentiation by functioning as a competing endogenous RNA. Cell 147, 358–369 (2011). 24. Ebert, M. S. & Sharp, P. A. Emerging roles for natural microRNA sponges. Curr. Biol. 20, R858–R861 (2010). 25. Vivancos, A. P., Guell,M., Dohm, J. C., Serrano, L. & Himmelbauer, H. Strand-specific deep sequencing of the transcriptome. Genome Res. 20, 989–999 (2010). 26. Huang, R. et al. An RNA-Seq strategy to detect the complete coding and non￾coding transcriptome including full-length imprinted macro ncRNAs. PLoS ONE 6, e27288 (2011). 27. Kent,W. J. et al. The human genome browser at UCSC. Genome Res. 12, 996–1006 (2002). 28. Pruitt, K. D., Tatusova, T. & Maglott, D. R. NCBI Reference Sequence (RefSeq): a curated non-redundant sequence database of genomes, transcripts and proteins. Nucleic Acids Res. 33, D501–D504 (2005). 29. Cabili, M. N. et al. Integrative annotation of human large intergenic noncoding RNAs reveals global properties and specific subclasses. Genes Dev. 25, 1915–1927 (2011). 30. Suzuki, H. et al. Characterization of RNase R-digested cellular RNA source that consists of lariat and circular RNAs from pre-mRNA splicing. Nucleic Acids Res. 34, e63 (2006). 31. Iwai, Y., Akahane, K., Pluznik, D. H. & Cohen, R. B. Ca21 ionophore A23187- dependent stabilization of granulocyte-macrophage colony-stimulating factor messenger RNA in murine thymoma EL-4 cells is mediated through two distinct regions in the 39-untranslated region. J. Immunol. 150, 4386–4394 (1993). 32. Hafner, M. et al. Transcriptome-wide identification of RNA-binding protein and microRNA target sites by PAR-CLIP. Cell 141, 129–141 (2010). 33. Lebedeva, S. et al. Transcriptome-wide analysis of regulatory interactions of the RNA-binding protein HuR. Mol. Cell 43, 340–352 (2011). 34. Baltz, A. G. et al. The mRNA-bound proteome and its global occupancy profile on protein-coding transcripts. Mol. Cell 46, 674–690 (2012). 35. Dropcho, E. J., Chen, Y. T., Posner, J. B. & Old, L. J. Cloning of a brain protein identified by autoantibodies from a patient with paraneoplastic cerebellar degeneration. Proc. Natl Acad. Sci. USA 84, 4552–4556 (1987). 36. Wee, L. M., Flores-Jasso, C. F., Salomon, W. E. & Zamore, P. D. Argonaute divides its RNA guide into domains with distinct functions and RNA-binding properties. Cell 151, 1055–1067 (2012). 37. Geiss, G. K. et al. Direct multiplexed measurement of gene expression with color￾coded probe pairs. Nature Biotechnol. 26, 317–325 (2008). 38. Landgraf, P. et al. A mammalian microRNA expression atlas based on small RNA library sequencing. Cell 129, 1401–1414 (2007). 39. Shaw, G., Morse, S., Ararat, M. & Graham, F. L. Preferential transformation of human neuronal cells by human adenoviruses and the origin of HEK 293 cells. FASEB J. 16, 869–871 (2002). 40. Kaufman, M. H. & Bard, J. B. L. The Anatomical Basis of Mouse Development (Academic, 1999). 41. Schambra, U. Prenatal Mouse Brain Atlas (Springer, 2008). 42. Kapsimali, M. et al. MicroRNAs show a wide diversity of expression profiles in the developing and mature central nervous system. Genome Biol. 8, R173 (2007). 43. Jacobs, T. et al. Localized activation of p21-activated kinase controls neuronal polarity and morphology. J. Neurosci. 27, 8604–8615 (2007). 44. Chacon, M. R. et al. Focal adhesion kinase regulates actin nucleation and neuronal filopodia formation during axonal growth. Development 139, 3200–3210 (2012). 45. Kato, M. et al. Cell-free formation of RNA granules: low complexity sequence domains form dynamic fibers within hydrogels. Cell 149, 753–767 (2012). 46. Romeo, T. Global regulation by the small RNA-binding protein CsrA and the non￾coding RNA molecule CsrB. Mol. Microbiol. 29, 1321–1330 (1998). 47. Gottesman, S. The small RNA regulators of Escherichia coli: roles and mechanisms. Annu. Rev. Microbiol. 58, 303–328 (2004). 48. Huelsken, J., Vogel, R., Erdmann, B., Cotsarelis, G. & Birchmeier, W. b-Catenin controls hair follicle morphogenesis and stem cell differentiation in the skin. Cell 105, 533–545 (2001). 49. Park, H. C. et al. Analysis of upstream elements in the HuC promoter leads to the establishment of transgenic zebrafish with fluorescent neurons. Dev. Biol. 227, 279–293 (2000). 50. Peri, F. & Nusslein-Volhard, C. Live imaging of neuronal degradation by microglia reveals a role for v0-ATPase a1 in phagosomal fusion in vivo. Cell 133, 916–927 (2008). 51. Jeck, W. R. et al. Circular RNAs are abundant, conserved, and associated with ALU repeats. RNA 19, 1–17 (2013). Supplementary Information is available in the online version of the paper. AcknowledgementsWe thankM. Feldkamp and C. Langnick (laboratory ofW. Chen) for Illumina sequencing runs. We thank J. Kjems for sending us a plasmid encoding circular human CDR1as for our zebrafish experiments. We thank K. Meier for technical assistance with zebrafish experiments and A. Sporbert from the confocal imaging facility.We thank A. Ivanov for assisting in bioinformatic analysis. N.R. thanks E. Westhof for useful discussions. We acknowledge the following funding sources: PhD program of the Max-Delbru¨ck-Center (MDC) (S.M., F.T., L.H.G.); the MDC-NYU exchange program (M.M.); BMBF project 1210182, ‘MiRNAs as therapeutic targets’ (A.E.); DFG for KFO218 (U.Z.); Helmholtz Association for the ‘MDC Systems Biology Network’, MSBN (S.D.M.); BMBF support for the DZHK (F.l.N. and N.R.); Center for Stroke Research Berlin (J.K., F.l.N.). Funding for the group of M.L. is supported by BMBF-funding for the Berlin Institute for Medical Systems Biology (0315362C). Author Contributions S.M., M.J., A.E. and F.T. contributed equally. S.M. performed many experiments, assisted by L.M. M.J. and A.E. carried out most of the computation, with contributions from N.R. and S.D.M. F.T. performed the circRNA validation experiments. A.R. performed all northern experiments. L.H.G. and M.M. contributed AGO PAR-CLIP experiments and HEK293 ribominus data, supervised by M.L. C.K. designed and carried out the single molecule experiments, in part together with A.L. U.Z. performed the mouse experiments. J.K. contributed the zebrafish experiments, supervised by F.l.N. N.R. designed and supervised the project. N.R. and M.J. wrote the paper. Author Information Sequencing data have been deposited at GEO under accession number GSE43574. Reprints and permissions information is available at www.nature.com/reprints. The authors declare no competing financial interests. Readers are welcome to comment on the online version of the paper. Correspondence and requests for materials should be addressed to N.R. (rajewsky@mdc-berlin.de). RESEARCH ARTICLE 6 | NATURE | VOL 000 | 00 MONTH 2013 ©2013 Macmillan Publishers Limited. All rights reserved

ARTICLE RESEARCH METHODS Computational pipeline for predicting circRNAs from ribominus sequencing chrl:15060286-1 data. Reference genomes(human hg19(February 2009, GRCh37), mouse mmg Permutation testing. To test the robustness of the circRNA detection pipeline we of real cing reads in different ways at the step of ownloadedfromtheUcscgenomebrowser(http://genome.ucsc.edu/)27.InaanchorgenerationWe(1)reversedeitheranchor:(2)reversedthecompleteread first step, reads that aligned contiguously and full length to the genomes were (3)randomly reassigned anchors between reads; or(4)reverse complemented the discarded. From the remaining reads we extracted 20mers from both ends and read (as a positive control). Although the reverse complement recovered the same aligned them independently to find unique anchor positions within spliced exons. output as expected, the various permutations led to only very few candidat Anchors that aligned in the reversed orientation (head-to-tail) indicated circRNA predictions, well below 0. 2% of the output with unpermuted reads and in excel- the complete read aligns and the breakpoints were flanked by GU/AG splice sites. HEK293 RNA-seq after rRNA depletion(Ribominus Seq). Total HEK293 RNA Ambiguous breakpoints were discarded. We used the short-read mapper Bowtie was isolated using Trizol as recommended by the manufacturer Ribosomal RI 2(ref. 52). Initially, ribominus reads were aligned in end-to-end mode to the s depleted from total RNA using the Ribominus kit(Invitrogen). A CDNA library was generated from rRNA-depleted RNA according to the Ilumina RNA-seq protocol. The cDNA library was sequenced on an Illumina GAllx by Sbowtie2-pl6--very-sensitive-phred64-mm-M20-score-min-C-15,0-q-x C. elegans oocyte isolation. Oocytes were isolated from wor a strain)and spe-9(partially ovulated oocytes BA671spe-9(hc88ts)I)as described previously. Oocytes were washed at least four times in PBS containing protease d in Fig. la to obtain 20-nucleotide anchors from both ends of the inhibitors(Sigma-Aldrich) to separate from worm debris. Oocyte purity was observed under the dissection scope(Zeiss). Oocytes were extracted from young adults to enrich for non-endomitotic oocytes, which was also checked by fluo- rescence microscopy(Zeiss) with a nuclear dye. Oocytes isolated from femm-I or S samtoolsview-hf4 sample_vs_genome. bam samtools view-Sb->unmapped_ spe-9 mutant background worms are hereafter referred to as fem-1 oocytes and sample. bam C. elegans sperm isolation. Sperm was isolated in principle as described S/unmapped2anchors py unmapped_ sample. bam I gzip >sample_anchors. qfa. iously from male worms obtained from a fog-2(q71)mutant background.Males ere cut in cold PBS containing protease inhibitors( Sigma-Aldrich). Sperm was subsequently purified by filtration(3X 40 um nylon mesh, 2 X 10 um nylon Here is an example of two anchor pairs in the FASTQ format; the original mesh) and a series of differential centrifugations(30 min 300g, 10 min 450g) read was kept as part of the first anchors identifier to simplify downstream and washed in cold PBS. Sperm was subsequently activated by incubation in PBS containing 200 ug ml-I Pronase(Sigma-Aldrich) for 30 min at 25oC s_8_1_0001_qseq_-14_A__NCCCGCCTCACCGGGTCAGTGAAAAAACGA Sperm purity is around 70% spermatids and spermatozoa contaminated with TCAGAGTAGTGGTCTTCTTCCGGCGGCCCCGCGCGCGCCGCGCTGC around 30% primary and secondary spermatocytes, as observed under oil immer- SIon NCCCGCCTCACCGGGTCAGT C elegans isolation of 1-cell-and 2-cell-stage embryos. 1-cell and 2-cell-stage embryos were obtained by fluorescence-activated cell sorting as described previ ously. Microscopic examination of the sorted embryos indicated that the l-cell- BB?AB;=日B;B@58( es_8_1_0001qseq_14_B 2-cell-stage embryo sample was a mixture of 1-cell-stage(40%), 2-cell-stage (55%)and older()>0;,8非 developed for human and mouse samples, but still performs sufficiently to enrich mRNAs up to 30% in C. elegans. Most of the remaining reads mapped to ribo- Next the anchors were aligned individually to the reference, keeping their paired somal RNAs. I ug of total RNa per sample was depleted from rRNAs with the ordering. The resulting alignments were read by another custom script that Ribominus Transcriptome kit(Invitrogen)according to the manufacturer's jointly evaluates consecutive anchor alignments belonging to the same original instructions with the modification that annealing of LNa probes to total RNA ead, performs extensions of the anchor ts, and collects statistics on was performed in a thermocycler(Eppendorf) with a temperature decrease from splice sites. After the run completes, the scri ts all detected splice junctions 70 to 37C at a rate of 1"C per min Depletion of rRN As was validated by capillary (linear and circular) in a UCSC BED-like with extra columns holding gel electrophoresis on a Bioanalyzer(Agilent). The ribominus RNA was then quality statistics, read counts etc. The original full-length reads that support each processed for sequencing library preparation according to the Illumina protocol. anction are written to stderr. Cluster generation and sequencing of C. eleganslibraries Cluster generation as well as sequencing of the prepared libraries was performed on the Illumina duster s bowtie2-P16 -reorder--mm-M20--score-min=C, 15, 0-q-x genome -U station(Illumina)and sequenced on the HiSeq 2000 according to the manufac- anchors.qfa-gz /find_circ- py-S hg19-p sample_-s sample/sites. log> Human gene models. We obtained gene models for RefSeq transcripts(12 December 2011), non-coding RNAs and the rnaGene and tRNA tracks from sites bed 2>> sites. reads the UCSC table browser(23 April 2012)27 Intersection of circRNAs with known transcripts. Our computational screen The resulting BED-like file is readily filtered for minimal quality cutoffs to pro- identifies only the splice sites that lead to circularization but not the internal exon/ duce the reported circRNA candidates In particular, we demanded the following: intron structure of circular RNAs To perform analyses of the sequence content of (1)GU/AG flanking the splice sites(built in); (2)unambiguous breakpoint detec. circRNAs we therefore inferred as much as possible from annotated transcripts. on;(3)a maximum of two mismatches in the extension procedure; (4)the The conservative assumption was that as little as possible should be spliced out. eakpoint cannot reside more than 2 nucleotides inside an anchor;(5)at least On the other hand, coincidence of circRNA splice sites with exonic boundaries two independent reads(each distinct sequence only counted once per sample) inside a transcript were considered as an indicator for relevant agreement and next-best alignment of at least one anchor above 35 points(-more than twoextra sorted all overlapping transcripts hierarchically by (1)splice-site coinidence support the junction; (6)unique anchor alignments with a safety margin to the internal introns appear to be spliced out( Supple high-quality bases); and (7)a genomic distance between the two (2, 1, or O);(2)total amount of exonic sequence between the splice sites; (3)tot splice sites of no more than 100 kb(only a small percentage of the data). As the amount of coding sequence. The latter was used to break ties only and helped the ribosomal DNA cluster is part of the C. elegans genome assembly(ce6)and ribo- annotation process. If one or both splice sites fell into an exon of the best match somal pre-RNAs could give rise to circular RNAs by mechanisms independent ing transcript, the corresponding exon boundary ed. Likewise, if it fell @2013 Macmillan Publishers Limited. All rights reserved

METHODS Computational pipeline for predicting circRNAs from ribominus sequencing data. Reference genomes (human hg19 (February 2009, GRCh37), mouse mm9 (July 2007, NB137/mm9), C. elegans ce6 (May 2008, WormBase v. WS190)) were downloaded from the UCSC genome browser (http://genome.ucsc.edu/)27. In a first step, reads that aligned contiguously and full length to the genomes were discarded. From the remaining reads we extracted 20mers from both ends and aligned them independently to find unique anchor positions within spliced exons. Anchors that aligned in the reversed orientation (head-to-tail) indicated circRNA splicing (compare main Fig. 1a). We extended the anchor alignments such that the complete read aligns and the breakpoints were flanked by GU/AG splice sites. Ambiguous breakpoints were discarded. We used the short-read mapper Bowtie 2 (ref. 52). Initially, ribominus reads were aligned in end-to-end mode to the genome: $ bowtie2 -p16 --very-sensitive --phred64 --mm -M20 --score-min5C, -15, 0 -q -x ,index. -U reads.qfa 2. bowtie2.log j samtools view -hbuS - j samtools sort - sample_vs_genome The unmapped reads were separated and run through a custom script to split the reads as indicated in Fig. 1a to obtain 20-nucleotide anchors from both ends of the read: $ samtools view -hf 4 sample_vs_genome.bam jsamtools view -Sb - . unmapped_ sample.bam $ ./unmapped2anchors.py unmapped_sample.bam j gzip . sample_anchors.qfa. gz Here is an example of two anchor pairs in the FASTQ format; the original read was kept as part of the first anchors identifier to simplify downstream analysis: @s_8_1_0001_qseq_14_A__NCCCGCCTCACCGGGTCAGTGAAAAAACGA TCAGAGTAGTGGTCTTCTTCCGGCGGCCCCGCGCGCGCCGCGCTGC NCCCGCCTCACCGGGTCAGT 1 #BB@?@AB@; 5 @B;B@@58( @s_8_1_0001_qseq_14_B CCCCGCGCGCGCCGCGCTGC 1 ;.;((.).0;.8######## Next the anchors were aligned individually to the reference, keeping their paired ordering. The resulting alignments were read by another custom script that jointly evaluates consecutive anchor alignments belonging to the same original read, performs extensions of the anchor alignments, and collects statistics on splice sites. After the run completes, the script outputs all detected splice junctions (linear and circular) in a UCSC BED-like format with extra columns holding quality statistics, read counts etc. The original full-length reads that support each junction are written to stderr: $ bowtie2 -p16 –reorder --mm -M20 --score-min5C, -15, 0 -q -x genome -U sample_anchors.qfa.gz j ./find_circ.py -S hg19 -p sample_ -s sample/sites.log . sample/sites.bed 2. sample/sites.reads The resulting BED-like file is readily filtered for minimal quality cutoffs to pro￾duce the reported circRNA candidates. In particular, we demanded the following: (1) GU/AG flanking the splice sites (built in); (2) unambiguous breakpoint detec￾tion; (3) a maximum of two mismatches in the extension procedure; (4) the breakpoint cannot reside more than 2 nucleotides inside an anchor; (5) at least two independent reads (each distinct sequence only counted once per sample) support the junction; (6) unique anchor alignments with a safety margin to the next-best alignment of at least one anchor above 35 points (,more than two extra mismatches in high-quality bases); and (7) a genomic distance between the two splice sites of no more than 100 kb (only a small percentage of the data). As the ribosomal DNA cluster is part of the C. elegans genome assembly (ce6) and ribo￾somal pre-RNAs could give rise to circular RNAs by mechanisms independent of the spliceosome, we discarded 130 candidates that mapped to the rDNA cluster on chrI:15,060,286-15,071,020. Permutation testing. To test the robustness of the circRNA detection pipeline we altered the sequence of real sequencing reads in different ways at the step of anchor generation. We (1) reversed either anchor; (2) reversed the complete read; (3) randomly reassigned anchors between reads; or (4) reverse complemented the read (as a positive control). Although the reverse complement recovered the same output as expected, the various permutations led to only very few candidate predictions, well below 0.2% of the output with unpermuted reads and in excel￾lent agreement with the results from simulated reads (Supplementary Fig. 1c). HEK293 RNA-seq after rRNA depletion (RibominusSeq). Total HEK293 RNA was isolated using Trizol as recommended by the manufacturer. Ribosomal RNA was depleted from total RNA using the Ribominus kit (Invitrogen). A cDNA library was generated from rRNA-depleted RNA according to the Illumina RNA-seq protocol. The cDNA library was sequenced on an Illumina GAIIx by a 2 3 76 bp run. C. elegans oocyte isolation. Oocytes were isolated from worms carrying a tem￾perature-sensitive (TS) allele for fem-1 (unovulated oocytes BA17[fem-1(hc17ts)] strain) and spe-9 (partially ovulated oocytes BA671[spe-9(hc88ts)]) as described previously53. Oocytes were washed at least four times in PBS containing protease inhibitors (Sigma-Aldrich) to separate from worm debris. Oocyte purity was observed under the dissection scope (Zeiss). Oocytes were extracted from young adults to enrich for non-endomitotic oocytes, which was also checked by fluo￾rescence microscopy (Zeiss) with a nuclear dye. Oocytes isolated from fem-1 or spe-9 mutant background worms are hereafter referred to as fem-1 oocytes and spe-9 oocytes, respectively. C. elegans sperm isolation. Sperm was isolated in principle as described prev￾iously54 from male worms obtained from a fog-2(q71) mutant background. Males were cut in cold PBS containing protease inhibitors (Sigma-Aldrich). Sperm was subsequently purified by filtration (3 3 40 mm nylon mesh, 2 3 10 mm nylon mesh) and a series of differential centrifugations (30 min 300g, 10 min 450g) and washed twice in cold PBS. Sperm was subsequently activated by incubation in PBS containing 200 mg ml21 Pronase (Sigma-Aldrich) for 30 min at 25 uC. Sperm purity is around 70% spermatids and spermatozoa contaminated with around 30% primary and secondary spermatocytes, as observed under oil immer￾sion microscope. C. elegans isolation of 1-cell- and 2-cell-stage embryos. 1-cell and 2-cell-stage embryos were obtained by fluorescence-activated cell sorting as described previ￾ously55. Microscopic examination of the sorted embryos indicated that the 1-cell￾stage sample was virtually pure (.98% one-cell stage embryos), whereas the 2-cell-stage embryo sample was a mixture of 1-cell-stage (40%), 2-cell-stage (55%) and older (,5%) embryos. Moreover, purity of the stages was further validated by checking for marker gene expression. Ribominus RNA preparation from C. elegans samples. We used a kit that was developed for human and mouse samples, but still performs sufficiently to enrich mRNAs up to 30% in C. elegans. Most of the remaining reads mapped to ribo￾somal RNAs. 1 mg of total RNA per sample was depleted from rRNAs with the Ribominus Transcriptome kit (Invitrogen) according to the manufacturer’s instructions with the modification that annealing of LNA probes to total RNA was performed in a thermocycler (Eppendorf) with a temperature decrease from 70 to 37 uC at a rate of 1 uC per min. Depletion of rRNAs was validated by capillary gel electrophoresis on a Bioanalyzer (Agilent). The ribominus RNA was then processed for sequencing library preparation according to the Illumina protocol. Cluster generation and sequencing of C. eleganslibraries. Cluster generation as well as sequencing of the prepared libraries was performed on the Illumina cluster station (Illumina) and sequenced on the HiSeq2000 according to the manufac￾turer’s protocols (Illumina). Human gene models. We obtained gene models for RefSeq transcripts (12 December 2011), non-coding RNAs29,56 and the rnaGene and tRNA tracks from the UCSC table browser (23 April 2012)27. Intersection of circRNAs with known transcripts. Our computational screen identifies only the splice sites that lead to circularization but not the internal exon/ intron structure of circular RNAs. To perform analyses of the sequence content of circRNAs we therefore inferred as much as possible from annotated transcripts. The conservative assumption was that as little as possible should be spliced out. On the other hand, coincidence of circRNA splice sites with exonic boundaries inside a transcript were considered as an indicator for relevant agreement and internal introns appear to be spliced out (Supplementary Fig. 2e). We therefore sorted all overlapping transcripts hierarchically by (1) splice-site coincidence (2, 1, or 0); (2) total amount of exonic sequence between the splice sites; (3) total amount of coding sequence. The latter was used to break ties only and helped the annotation process. If one or both splice sites fell into an exon of the best match￾ing transcript, the corresponding exon boundary was trimmed. Likewise, if it fell ARTICLE RESEARCH ©2013 Macmillan Publishers Limited. All rights reserved

RESEARCH ARTICLE into an intron or beyond transcript bounds,the closest exon was extended to were dissected and tissue samples were collected directly into ice-cold Trizol for match the circRNA boundaries.circRNA start/end coordinates were never RNA preparation.Caenorhabditis elegans RNA was isolated from about 7,000 altered.If no annotated exons overlapped the circRNA we assumed a single-exon mixed stage worms by two rounds of freeze-thaw lysis in Trizol LS reagent circRNA.The resulting annotation of circRNAs is based on the best matching (Invitrogen)according to the manufacturer's protocol.RNA was extracted from transcript and may in some cases not represent the ideal choice.Changing the aqueous phase with phenol:chloroform (Ambion).RNA was precipitated with annotation rules,however,did not substantially change the numbers in Fig.ld.isopropanol and Glycoblue(Ambion)overnight at-2o Cor for30 min at-80'C, Finding circRNAs conserved between human and mouse.We reasoned that respectively.Reverse transcription was performed using M-MLV(Promega)or when comparing two species,the cutoff of two independent reads in each of them Superscript I I with oligo(dT)primer((all Invitrogen)or random primer(Meta- could be dropped,as orthologous circRNAs would automatically be supported by bion).For assaying mRNA expression level,qRT-PCR was performed using SYBR- two independently produced reads via the intersection.We therefore mapped all Green Fluorescein (Thermo Scientific,Fermentas)and a StepOnePlus PCR System mouse circRNA candidates with less stringent filtering to human genome coor-(Applied Biosystems).Expression data in CDRlas knockdown experiments,tran- dinates using the UCSC liftOver tool.The mapped mouse circRNAs were com- scriptional block and RNase R assays were normalized to C.elegans spike-in RNA. pared with independently identified human circRNAs,yielding 229 circRNAs To this end 5-10%C.elegans total RNA was added to the respective Trizol sample with precisely orthologous splice sites between human and mouse.Of these,223 and qPCR primer for ama-I or eif-3.d were used.Mouse expression data were were composed exclusively of coding exons and were subsequently used for our normalized to Actb.miRNA expression levels were assayed using TaqMan conservation analysis (Fig.1f).When intersecting the reported sets of circRNAs microRNA assays(Applied Biosystems)and normalized to sno-234.Expression supported by two independent reads in each species,we found 81 conserved levels of circRNAs described in this study were measured by qPCR using divergent circRNAs (supported by at least 4 reads in total). primers.A list of primer sequences is available in Supplementary Table 8. Conserved element counting.We downloaded genome-wide human (hg19)PCR amplification and Sanger sequencing.DNA templates were PCR amplified phyloP conservation scores tracks derived from genome alignments of placental using BioRad Mastercyclers and ThermoScientific DreamTaq Green PCR Master mammals from UCSC27.We interrogated the genome-wide profile inside Mix according to the manufacturer's protocol.We performed 35 cycles of PCR. circRNAs in two different ways.(1)Intergenic and intronic circRNAs.We read PCR products were visualized after electrophoresis in 2%ethidium bromide- out the conservation scores along the complete circRNA and searched for blocks stained agarose gel.To confirm the PCR results,the PCR products were purified of at least 6-nucleotide length that exceeded a conservation score of 0.3 for through Agencourt AMPure XP PCR purification kit.Direct PCR product Sanger intergenic and 0.5 for intronic circRNAs.The different cutoffs empirically adjust sequencing was performed by LGC Genomics Ready2 Run services.Primer Pl for the different background levels of conservation and were also used on the was provided for sequencing the product for each candidate. respective controls.For each circRNA,we computed the cumulative length of all Primer design.Divergent primers were designed for each candidate(P1,P2)to such blocks and normalized it by the genomic length of the circRNA.Artefacts of anneal at the distal ends of its sequence.As negative controls we used divergent constant positive conservation scores in the phyloP profile,apparently caused by primers for GAPDH and ACTB linear transcript in HEK293 cells,and elF-3.D in missing alignment data,were removed with an entropy filter (this did not qua- C.elegans.As a further negative control for divergent primers,we used genomic litatively affect the results).circRNAs annotated as intronic by the best-match DNA extracted through Qiagen DNeasy Blood Tissue kit.As positive controls, procedure explained above that had any overlap with exons in alternative tran-we used convergent primers for the corresponding linear transcripts or for house- scripts on either strand (five cases)were removed from the analysis.The resulting keeping genes (elF-3.D for C.elegans). distributions are shown in Supplementary Fig.Ih,i.(2)Coding exon circRNAs.RNase R treatment.HEK293 DNase-treated total RNA (5 ug)was incubated We used the best-match strategy outlined above to construct an estimated 'exon-15 min at 37C with or without 3 U ug of RNase R (Epicentre Bio- chain'for the circRNAs that overlapped exclusively coding sequence.Using this technologies).RNA was subsequently purified by phenol-chloroform extraction, chain we in silico 'spliced'out the corresponding blocks of the conservation score retro-transcribed through Superscript SSIII(Invitrogen)according to the man- profile.We kept track of the frame and sorted the conservation scores into ufacturer's protocol,and used in qPCR. separate bins for each codon position.In addition to this,we also recorded RNA nicking assay.For partial alkaline hydrolysis (nicking)1 ug ul-of conservation scores in the remaining pieces of coding sequence ('outside'the HEK293 total RNA was incubated in 50 mM NaHCO;for 2.5 or 5 min at 90C circRNA)as a control.However,we observed that the level of conservation is or 5 min on ice for controls.After incubation the samples were immediately re- systematically different between internal parts of the coding sequence and the suspended in denaturing RNA sample buffer and analysed on northern blots. amino-or carboxy-terminal parts(not shown).We therefore randomly generated Northern blotting.Total RNA(l0-2oμg)was loaded on a I.2%agarose gel chains of internal exons,mimicking the exon-number distribution of real containing 1%formaldehyde and run for 2-2.5 h in MOPS buffer. circRNAs,as a control.When analysing the circRNAs conserved between human The gel was soaked in 1XTBE for 20 min and transferred to a Hybond-N+ and mouse,it became furthermore apparent that we also needed to adjust for the membrane (GE Healthcare)for 1 h(15 V)using a semi-dry blotting system (Bio- higher level of overall conservation.High expression generally correlates with Rad).Membranes were dried and ultraviolet-crosslinked (at 265 nm)1X at conservation and thus,an expression cutoff was enforced on the transcripts used 200,000 uJ cm2.Pre-hybridization was done at 42C for 1 h and 32p-labelled to generate random controls.This resulted in a good to conservative match with oligonucleotide DNA probes were hybridized overnight.The membranes were the actual circRNAs (Supplementary Fig.1j,k). washed briefly in 2X SSC,0.1%SDS at room temperature and two additional Overlap of identified circRNAs with published circular RNAs.A number of times at 55C for 30 min,followed by two 30-min washes in 0.2X SSC,0.1%SDS studies in human have reported evidence for circRNAs which derive from exons at 50-55C.For data collection,the membrane was exposed to a phosphoimager of DCC',ETSI and a non-coding RNA from the human INK4/ARF locus and screen. the CDRlas locus'.Additionally,circRNAs from exons of the genes CAMSAPI,Genome alignments for detecting miRNA seed complementary sites.Multiple FBXW4,MAN1A2,REXO4,RNF220 and ZKSCANI have been recently experi-species alignments for the genomic intervals,corresponding to circRNAs pre- mentally validated.For the four genes from the latter study,where we had dicted in C.elegans(ce6),human (hg19)and mouse(mm9),were generated via ribominus data from the tissues in which these circRNAs were predicted (leuko- the Galaxy server at UCSCS9-61.In case that a circle was overlapping with an cytes),we recovered validated circRNAs from all of them (ZKSCAN1,CAMSAPI,annotated transcript,the inferred spliced sequence was used for retrieving the FBXW4,MANIA2). alignments. Cell culture and treatments.HEK293(Fig,3f),HEK293TN(for virus production)The alignments included C.elegans,C.briggsae and C.remanei in the first case and HEK293 Flp-In T-REx 293 (Life Technologies,all other experiments)were and Homo sapiens,Mus musculus,Rattus norvegicus,Bos taurus and Canis famili- cultured in Dulbecco's modified Eagle medium GlutaMax(Gibco)4.5 gl-'glucose,aris in the latter two. supplemented with 10%FCS,20U ml-'penicillin and streptomycin (Gibco)at C.elegans human and mouse miRNAs.Fasta files with C.elegans,human and 37C,5%CO2.Whereas CDRlas/GAPDH ratios were within the given range,we mouse miRNAs were obtained from miRBase release 16(ref.62).Only mature observed two-to fivefold variation of CDRlas/vinculin ratios between different miRNAs were considered for the seed analysis.According to miRBase 16 a HEK lines.Transcription was blocked by adding 2 ug ml actinomycin D or mature miRNA is the predominant miRNA between the two species arising from DMSO as a control (Sigma-Aldrich)to the cell culture medium.For in vitro wound the two arms of the precursor hairpin (information that is not included in more healing assays,cells were grown to confluency,the cell layer was disrupted using a recent versions).The miRNAs were grouped into families that share a common 300 ul pipette tip and cells were washed once with medium.Bright-field images of seed (nucleotides 2-7).There are 117,751 and 723 miRNA families for C.elegans, cells were taken using a Axio Observer.Z1 (Zeiss)right after setting the scratch and human and mouse,respectively. 24h later.The relative scratch areas were measured using Image)software. Detecting putative miRNA seed matches.The C.elegans,human and mouse Quantitative PCR.Total RNA from cell lines was isolated using Trizol (Invi-multiple species alignments were scanned for putative conserved miRNA target trogen)extraction following the manufacturer's protocol Adult B6129SF1/]mice sites for each of the mature miRNA families.A putative target site of a miRNA is a ©2013 Macmillan Publishers Limited.All rights reserved

into an intron or beyond transcript bounds, the closest exon was extended to match the circRNA boundaries. circRNA start/end coordinates were never altered. If no annotated exons overlapped the circRNA we assumed a single-exon circRNA. The resulting annotation of circRNAs is based on the best matching transcript and may in some cases not represent the ideal choice. Changing the annotation rules, however, did not substantially change the numbers in Fig. 1d. Finding circRNAs conserved between human and mouse. We reasoned that when comparing two species, the cutoff of two independent reads in each of them could be dropped, as orthologous circRNAs would automatically be supported by two independently produced reads via the intersection. We therefore mapped all mouse circRNA candidates with less stringent filtering to human genome coor￾dinates using the UCSC liftOver tool57. The mapped mouse circRNAs were com￾pared with independently identified human circRNAs, yielding 229 circRNAs with precisely orthologous splice sites between human and mouse. Of these, 223 were composed exclusively of coding exons and were subsequently used for our conservation analysis (Fig. 1f). When intersecting the reported sets of circRNAs supported by two independent reads in each species, we found 81 conserved circRNAs (supported by at least 4 reads in total). Conserved element counting. We downloaded genome-wide human (hg19) phyloP conservation score58 tracks derived from genome alignments of placental mammals from UCSC27. We interrogated the genome-wide profile inside circRNAs in two different ways. (1) Intergenic and intronic circRNAs. We read out the conservation scores along the complete circRNA and searched for blocks of at least 6-nucleotide length that exceeded a conservation score of 0.3 for intergenic and 0.5 for intronic circRNAs. The different cutoffs empirically adjust for the different background levels of conservation and were also used on the respective controls. For each circRNA, we computed the cumulative length of all such blocks and normalized it by the genomic length of the circRNA. Artefacts of constant positive conservation scores in the phyloP profile, apparently caused by missing alignment data, were removed with an entropy filter (this did not qua￾litatively affect the results). circRNAs annotated as intronic by the best-match procedure explained above that had any overlap with exons in alternative tran￾scripts on either strand (five cases) were removed from the analysis. The resulting distributions are shown in Supplementary Fig. 1h, i. (2) Coding exon circRNAs. We used the best-match strategy outlined above to construct an estimated ‘exon￾chain’ for the circRNAs that overlapped exclusively coding sequence. Using this chain we in silico ‘spliced’ out the corresponding blocks of the conservation score profile. We kept track of the frame and sorted the conservation scores into separate bins for each codon position. In addition to this, we also recorded conservation scores in the remaining pieces of coding sequence (‘outside’ the circRNA) as a control. However, we observed that the level of conservation is systematically different between internal parts of the coding sequence and the amino- or carboxy-terminal parts (not shown).We therefore randomly generated chains of internal exons, mimicking the exon-number distribution of real circRNAs, as a control. When analysing the circRNAs conserved between human and mouse, it became furthermore apparent that we also needed to adjust for the higher level of overall conservation. High expression generally correlates with conservation and thus, an expression cutoff was enforced on the transcripts used to generate random controls. This resulted in a good to conservative match with the actual circRNAs (Supplementary Fig. 1j, k). Overlap of identified circRNAs with published circular RNAs. A number of studies in human have reported evidence for circRNAs which derive from exons of DCC4 , ETS15 and a non-coding RNA from the human INK4/ARF locus8 and the CDR1as locus9 . Additionally, circRNAs from exons of the genes CAMSAP1, FBXW4, MAN1A2, REXO4, RNF220 and ZKSCAN1 have been recently experi￾mentally validated10. For the four genes from the latter study, where we had ribominus data from the tissues in which these circRNAs were predicted (leuko￾cytes), we recovered validated circRNAs from all of them (ZKSCAN1, CAMSAP1, FBXW4, MAN1A2). Cell culture and treatments. HEK293 (Fig. 3f), HEK293TN (for virus production) and HEK293 Flp-In T-REx 293 (Life Technologies, all other experiments) were cultured in Dulbecco’s modified Eagle medium GlutaMax (Gibco) 4.5 g l21 glucose, supplemented with 10% FCS, 20 U ml21 penicillin and streptomycin (Gibco) at 37 uC, 5% CO2. Whereas CDR1as/GAPDH ratios were within the given range, we observed two- to fivefold variation of CDR1as/vinculin ratios between different HEK lines. Transcription was blocked by adding 2 mg ml21 actinomycin D or DMSO as a control (Sigma-Aldrich) to the cell culture medium. Forin vitro wound healing assays, cells were grown to confluency, the cell layer was disrupted using a 300 ml pipette tip and cells were washed once with medium. Bright-field images of cells were taken using a Axio Observer.Z1 (Zeiss) right after setting the scratch and 24 h later. The relative scratch areas were measured using ImageJ software. Quantitative PCR. Total RNA from cell lines was isolated using Trizol (Invi￾trogen) extraction following the manufacturer’s protocol. Adult B6129SF1/J mice were dissected and tissue samples were collected directly into ice-cold Trizol for RNA preparation. Caenorhabditis elegans RNA was isolated from about 7,000 mixed stage worms by two rounds of freeze–thaw lysis in Trizol LS reagent (Invitrogen) according to the manufacturer’s protocol. RNA was extracted from aqueous phase with phenol:chloroform (Ambion). RNA was precipitated with isopropanol and Glycoblue (Ambion) overnight at 220 uC or for 30 min at 280 uC, respectively. Reverse transcription was performed using M-MLV (Promega) or Superscript III with oligo(dT) primer (all Invitrogen) or random primer (Meta￾bion). For assaying mRNA expression level, qRT–PCR was performed using SYBR￾Green Fluorescein (Thermo Scientific, Fermentas) and a StepOnePlus PCR System (Applied Biosystems). Expression data in CDR1as knockdown experiments, tran￾scriptional block and RNase R assays were normalized to C. elegans spike-in RNA. To this end 5–10% C. elegans total RNA was added to the respective Trizol sample and qPCR primer for ama-1 or eif-3.d were used. Mouse expression data were normalized to Actb. miRNA expression levels were assayed using TaqMan microRNA assays (Applied Biosystems) and normalized to sno-234. Expression levels of circRNAs described in this study were measured by qPCR using divergent primers. A list of primer sequences is available in Supplementary Table 8. PCR amplification and Sanger sequencing. DNA templates were PCR amplified using BioRad Mastercyclers and ThermoScientific DreamTaq Green PCR Master Mix according to the manufacturer’s protocol. We performed 35 cycles of PCR. PCR products were visualized after electrophoresis in 2% ethidium bromide￾stained agarose gel. To confirm the PCR results, the PCR products were purified through Agencourt AMPure XP PCR purification kit. Direct PCR product Sanger sequencing was performed by LGC Genomics Ready2 Run services. Primer P1 was provided for sequencing the product for each candidate. Primer design. Divergent primers were designed for each candidate (P1, P2) to anneal at the distal ends of its sequence. As negative controls we used divergent primers for GAPDH and ACTB linear transcript in HEK293 cells, and eIF-3.D in C. elegans. As a further negative control for divergent primers, we used genomic DNA extracted through Qiagen DNeasy Blood & Tissue kit. As positive controls, we used convergent primers for the corresponding linear transcripts or for house￾keeping genes (eIF-3.D for C. elegans). RNase R treatment. HEK293 DNase-treated total RNA (5 mg) was incubated 15 min at 37 uC with or without 3 U mg21 of RNase R (Epicentre Bio￾technologies). RNA was subsequently purified by phenol-chloroform extraction, retro-transcribed through Superscript SSIII (Invitrogen) according to the man￾ufacturer’s protocol, and used in qPCR. RNA nicking assay. For partial alkaline hydrolysis (nicking) 1 mg ml 21 of HEK293 total RNA was incubated in 50 mM NaHCO3 for 2.5 or 5 min at 90 uC or 5 min on ice for controls. After incubation the samples were immediately re￾suspended in denaturing RNA sample buffer and analysed on northern blots. Northern blotting. Total RNA (10–20 mg) was loaded on a 1.2% agarose gel containing 1% formaldehyde and run for 2–2.5 h in MOPS buffer. The gel was soaked in 13TBE for 20 min and transferred to a Hybond-N1 membrane (GE Healthcare) for 1 h (15 V) using a semi-dry blotting system (Bio￾Rad). Membranes were dried and ultraviolet-crosslinked (at 265 nm) 13 at 200,000 mJ cm22 . Pre-hybridization was done at 42 uC for 1 h and 32P-labelled oligonucleotide DNA probes were hybridized overnight. The membranes were washed briefly in 23 SSC, 0.1% SDS at room temperature and two additional times at 55 uC for 30 min, followed by two 30-min washes in 0.23 SSC, 0.1% SDS at 50–55 uC. For data collection, the membrane was exposed to a phosphoimager screen. Genome alignments for detecting miRNA seed complementary sites. Multiple species alignments for the genomic intervals, corresponding to circRNAs pre￾dicted in C. elegans (ce6), human (hg19) and mouse (mm9), were generated via the Galaxy server at UCSC59–61. In case that a circle was overlapping with an annotated transcript, the inferred spliced sequence was used for retrieving the alignments. The alignments included C. elegans, C. briggsae and C. remanei in the first case and Homo sapiens, Mus musculus, Rattus norvegicus, Bos taurus and Canis famili￾aris in the latter two. C. elegans human and mouse miRNAs. Fasta files with C. elegans, human and mouse miRNAs were obtained from miRBase release 16 (ref. 62). Only mature miRNAs were considered for the seed analysis. According to miRBase 16 a mature miRNA is the predominant miRNA between the two species arising from the two arms of the precursor hairpin (information that is not included in more recent versions).The miRNAs were grouped into families that share a common seed (nucleotides 2–7). There are 117, 751 and 723 miRNA families for C. elegans, human and mouse, respectively. Detecting putative miRNA seed matches. The C. elegans, human and mouse multiple species alignments were scanned for putative conserved miRNA target sites for each of the mature miRNA families. A putative target site of a miRNA is a RESEARCH ARTICLE ©2013 Macmillan Publishers Limited. All rights reserved

ARTICLE RESEARCH 6-nucleotide-long sequence in the genome that is the reverse complement of Designer version 2.0(Biosearch Technologies)with a masking level of 4 on the nucleotides 2-7 of the mature miRNA sequence.A putative target site is called human genome to achieve high probe specificity (Supplementary Table 8). conserved if it is found in C.elegans,C.briggsae and C.remanei in the first case or Stellaris probe pools were obtained from BioCat GmbH as conjugates coupled in human,mouse,rat,cow and dog in the latter. to Quasar 670(a Cy 5 replacement).Flp-In T-REx 293 cells (Life Technologies) AGO PAR-CLIP.Generation and growth conditions of human embryonic kid-were grown exponentially and seeded into LabTek 4-well chambered coverslips ney(HEK)293 cells and HEK293 stably expressing Flag/HA-AGOI and Flag/(1 to 2 X 105 cells per well).Hybridizations were performed according to the HA-AGO2 were reported previously3.Stably transfected and parental HEK293 manufacturer's instructions with 50 ng ml-DAPI as nuclear counterstain; cells were labelled with 100 uM 4-thiouridine for 16 h.After labelling.procedure Stellaris probes were hybridized at 125 nM concentration with a stringency of followed the PAR-CLIP protocol as described2.Briefly,ultraviolet-irradiated 10%formamide in overnight hybridizations at 37C.Images were acquired on an cells were lysed in NP-40 lysis buffer.Immunoprecipitation was carried out with inverted Nikon Ti microscope with a Hamatsu ORCA R2 CCD camera,a 60X protein G magnetic beads(Invitrogen)coupled to anti-Flag antibody(Sigma)and NA 1.4 oil objective and Nikon NIS-Elements Ar software(version 4),using an to anti-AGO2 antibody from extracts of stably transfected and parental exposure time of 50 ms for DAPI and 1-1.5 s for Quasar 670.Groups of cells for HEK293 cells,respectively,for I h at 4C.Beads were treated with calf intestinal imaging were chosen in the DAPI channel;Z-stacks were acquired in the Quasar phosphatase(NEB)and radioactively end-labelled by T4 polynucleotide kinase670 channel using0.3μn spacing and comprised a total depth of6.5um(5um (Fermentas).The crosslinked protein--RNA complexes were resolved on4-l2%below and 1.5μm above the middle of the nucleus)and merged using maximum NuPAGE gel (Invitrogen),and a labelled protein-RNA complex of close to intensity. 100 kDa was excised.The protein-RNA was isolated by electroelution.RNA Mouse strains and in situ hybridizations.All mice were bred and maintained in was isolated by proteinase K treatment and phenol-chloroform extraction,the animal facility of the Max Delbruck Centrum under specific pathogen-free reverse transcribed and PCR-amplified.The amplified cDNA was sequenced conditions,in plastic cages with regular chow and water ad libitum.All aspects of on a GAllx (Illumina)with 36 cycles. animal care and experimental protocols were approved by the Berlin Animal Human Argonaute PAR-CLIP analysis.We obtained Argonaute PAR-CLIP Review Board (REG 0441/09).B6129SF1/J wild-type adult,newborns(postnatal reads from ref.32.We additionally produced 4 PAR-CLIP libraries.In total,day 1)or pregnant females (plug detection at day 0.5;embryo collection at day we analysed the following PAR-CLIP data sets:AGO1_4su_1 (SRR048973),13.5)were used,as indicated for each experiment,to obtain the tissues needed for AGO3_4su_1 (SRR048976)from ref.32;AGOl_4su_ML_MM_6,AGO2_RNA analysis and in situ hybridizations (ISH).After death,embryos or tissues 4su_ML_MM_7,AGO2_4su_ML_MM_8,and AGO2_4su_3_ML_LG (our were immediately frozen in liquid nitrogen and stored at -70C,or fixed for ISH. own data,published under GEO accession GSE43574). Mouse brain structures were collected and named according to the anatomical Redundant reads were collapsed (such that each distinct read sequence appears guidelines of the Gene Expression Nervous System Atlas of the Rockefeller only once),aligned to the human genome (assembly hg19)using bwa 0.6.1-r104 University (http://www.gensat.org)and the Mouse Brain Atlas (http://www. (ref.65),and analysed by our in-house PAR-CLIP analysis pipeline (Jens,M.mbl.org/mbl_main/atlas.html). et al,unpublished),essentially as described in ref.33.Briefly,reads uniquely For the RNA analysis and to clone CDRlas-specific RNA probes,two adult aligning to the genome are grouped into clusters contiguously covering the ref-1-year-old mice of both sexes were dissected,total RNA prepared and analysed.If erence,assigning each cluster a number of quality scores(T conversions,number embryos or newborns were sectioned,a minimum of two specimens were eval- of independent reads,etc).Clusters with less than 3 reads from 3 of 6 independent uated;in some instances up to 5 specimens were used. AGO PAR-CLIP libraries or lacking T conversions were discarded.Remaining For ISH,samples were fixed in formalin(1×pBS;4%formaldehyde)forl2h clusters are annotated against a comprehensive list of transcript models (see and post-fixed (70%ethanol,18 h)before dehydrating and paraffin-embedding. below)and collected into 'only sense',only antisense'and intergenic/overlapping Next,the organs were perfused with a standard protocol using a Shandon XP transcription'categories based on their annotation.As PAR-CLIP sequencing Hypercentre.For ISH mouse embryos or organs were cut in RNase-free condi- preserves the directionality of RNA fragments we assume 'only antisense'clusters tions at 6 um and ISH was performed as described with digoxigenin (DIG)- to predominantly represent false positives due to mapping artefacts(PAR-CLIP labelled RNA probes.All DIG-RNA probes were hybridized at 58 Covernight.A RNA is mutated and fragments are often short),and choose quality cutoffs for all total of 600 ng of the labelled probes was used per slide. clusters such that the fraction of kept 'only antisense'clusters is reduced to below To amplify Cdrl sense and antisense sequences for ISH probe preparation a 5%.Remaining 'only antisense'clusters were discarded.For Fig.3a,uniquely standard PCR-amplification was performed using mouse cerebellum cDNA. aligning,collapsed reads are shown. Three Cdrlas amplicons were generated,two of which probes are meant for AGO binding sites in C.elegans.Sequencing reads from the Zisoulis Alg-1 the detection of both linear and circular forms using mmuCdrl_If 5'. HITS-Clip data were obtained from http://yeolab.ucsd.edu/yeolab/Papers_files/TGCCAGTACCAAGGTCTTCC-3'and mmuCdrl_Ir 5'-TTTTCTGCTGGA ALGl_MT-WT_raw.tar.gz(ref.66).The raw sequencing data of the wild-type AGATGTCAA-3',as well as mmuCdrl_2f 5'-CCAGACAATCGTGATCT Alg-1 HITS-CLIP was pre-processed and mapped with the mapper module from TCC-3'and mmuCdrl_2r 5'-ATCTTGGCTGGAAGACTTGG-3'.In addition miRDeep2(ref.74).The pre-processed reads were mapped with bowtie version a probe was generated,specific to the circular probe,using the divergent primers 0.12.7(ref.67)to the C.elegans genome (ce6).All reads that overlapped when mmuCdrl_as_7f 5'-CCACATCTTCCAGCATCTTT-3'and mmuCdrl_as_ mapped to the genome were merged into bigger regions (islands).Read counts 7r 5'-TGGATCCCTTGGAAGACAAA-3'(CDR1_as head to tail probe).All were averaged.This resulted in 24,910 islands in the C.elegans genome. ensuing fragments were subcloned into pCR-BluntlI-TOPO(Invitrogen)and Analysis of sequence conservation in CDRlas.Genome alignments of 32 verte-verified by sequencing.Linearized plasmids were amenable for in vitro transcrip- brates were downloaded from the UCSC database (hg19)27 and analysed for the tion using the T7(antisense)or SP6(sense)polymerase and a DIG-label nucleo- CDRIas locus.Primate species other than human were discarded to not bias the tide mixture according to manufacturer's instruction (Roche Applied Science). analyses.The one species (cow)with more than 50%gaps in the CDRIas locus LNA ISHs were performed according to a protocol suggested by the manufac- was also discarded.The alignments for the seed regions were then corrected.turer (Exigon)with minor modifications.For individual LNAs,specific protocols Specifically,bases that would dlearly align with the seed but had been separated were run at 51C (miR-7;38485-15)or 58C(miR-124;88066-15)on an in the alignment by runs of gaps were re-aligned.These corrections were neces-InsituPro VS robot (Intavis).A pre-hybridization step was added,which con- sary in less than 1%of all seed sites. sisted of an incubation of the slides at 15C lower than the hybridization tem- For an in-depth analysis we BLATed the human CDRIas sequence with2o.perature for3 0 min using hybridization buffer.The antibody-blocking step nucleotide flanking region against all vertebrate genomes in the UCSC genome was performed in the presence of 1%mouse blocking reagent (Roche browser and kept only hits that in turn aligned best to the human locus.The 11096176001)and 10%sheep serum.The LNA probes were used at the following resulting sequences were used to build a multiple species alignment with concentrations:miR-7 40 nM;miR-124 20 nM;U6 snRNA InM;scrambled MUSCLE.The same corrections were applied as described above.This align-40 nM,as suggested by miRCURY LNA microRNA ISH Optimization kit ment was also used for Supplementary Fig.4.Entropy was calculated in logz units (Exiqon;90004).Before detection all slides were washed 4X in NTMT including and averaged across all alignment columns bracketing each human seed site by I mM Levamisole.The doubly DIG-labelled LNAs were detected by the alkaline maximally 8 nucleotides. phosphatase using the substrate BM-purple(Roche;11442074001)at 37C. Analysis of miR-7 base-pairing within CDRlas.RNAcofold?o was used to co-siRNA-and shRNA-mediated knock down.CDRlas was knocked down using fold miR-7 with each of the 74 binding regions within CDRlas defined as the custom designed siRNA oligonucleotides (Sigma)and Lipofectamine RNAiMax miR-7 seed match TCTTCC and the next 16 bases upstream. (Invitrogen).210°HEK293 cells were transfected with10 nM siRNA duplex Single-molecule RNA fluorescence in situ hybridization (smRNA FISH).48 following the manufacturer's protocol.After 12-16 h cells were harvested and oligonucleotide probes(20 nucleotides length;spacing 2 nucleotides)comple-subjected to RNA analysis.For stable knock down of CDRlas,293TN cells were mentary to the CDRlas transcript were designed using the Stellaris Probe co-transfected with the packaging plasmids pLP1,pLP2 and the VSV-G plasmid ©2013 Macmillan Publishers Limited.All rights reserved

6-nucleotide-long sequence in the genome that is the reverse complement of nucleotides 2–7 of the mature miRNA sequence. A putative target site is called conserved if it is found in C. elegans, C. briggsae and C. remanei in the first case or in human, mouse, rat, cow and dog in the latter. AGO PAR-CLIP. Generation and growth conditions of human embryonic kid￾ney (HEK) 293 cells and HEK293 stably expressing Flag/HA–AGO1 and Flag/ HA–AGO2 were reported previously63. Stably transfected and parental HEK293 cells were labelled with 100 mM 4-thiouridine for 16 h. After labelling, procedure followed the PAR-CLIP protocol as described32. Briefly, ultraviolet-irradiated cells were lysed in NP-40 lysis buffer. Immunoprecipitation was carried out with protein G magnetic beads (Invitrogen) coupled to anti-Flag antibody (Sigma) and to anti-AGO2 antibody64 from extracts of stably transfected and parental HEK293 cells, respectively, for 1 h at 4 uC. Beads were treated with calf intestinal phosphatase (NEB) and radioactively end-labelled by T4 polynucleotide kinase (Fermentas). The crosslinked protein–RNA complexes were resolved on 4–12% NuPAGE gel (Invitrogen), and a labelled protein–RNA complex of close to 100 kDa was excised. The protein–RNA was isolated by electroelution. RNA was isolated by proteinase K treatment and phenol-chloroform extraction, reverse transcribed and PCR-amplified. The amplified cDNA was sequenced on a GAIIx (Illumina) with 36 cycles. Human Argonaute PAR-CLIP analysis. We obtained Argonaute PAR-CLIP reads from ref. 32. We additionally produced 4 PAR-CLIP libraries. In total, we analysed the following PAR-CLIP data sets: AGO1_4su_1 (SRR048973), AGO3_4su_1 (SRR048976) from ref. 32; AGO1_4su_ML_MM_6, AGO2_ 4su_ML_MM_7, AGO2_4su_ML_MM_8, and AGO2_4su_3_ML_LG (our own data, published under GEO accession GSE43574). Redundant reads were collapsed (such that each distinct read sequence appears only once), aligned to the human genome (assembly hg19) using bwa 0.6.1-r104 (ref. 65), and analysed by our in-house PAR-CLIP analysis pipeline (Jens, M. et al., unpublished), essentially as described in ref. 33. Briefly, reads uniquely aligning to the genome are grouped into clusters contiguously covering the ref￾erence, assigning each cluster a number of quality scores (T conversions, number of independent reads, etc). Clusters with less than 3 reads from 3 of 6 independent AGO PAR-CLIP libraries or lacking T conversions were discarded. Remaining clusters are annotated against a comprehensive list of transcript models (see below) and collected into ‘only sense’, ‘only antisense’ and ‘intergenic/overlapping transcription’ categories based on their annotation. As PAR-CLIP sequencing preserves the directionality of RNA fragments we assume ‘only antisense’ clusters to predominantly represent false positives due to mapping artefacts (PAR-CLIP RNA is mutated and fragments are often short), and choose quality cutoffs for all clusters such that the fraction of kept ‘only antisense’ clusters is reduced to below 5%. Remaining ‘only antisense’ clusters were discarded. For Fig. 3a, uniquely aligning, collapsed reads are shown. AGO binding sites in C. elegans. Sequencing reads from the Zisoulis Alg-1 HITS-Clip data were obtained from http://yeolab.ucsd.edu/yeolab/Papers_files/ ALG1_MT-WT_raw.tar.gz (ref. 66). The raw sequencing data of the wild-type Alg-1 HITS-CLIP was pre-processed and mapped with the mapper module from miRDeep2 (ref. 74). The pre-processed reads were mapped with bowtie version 0.12.7 (ref. 67) to the C. elegans genome (ce6). All reads that overlapped when mapped to the genome were merged into bigger regions (islands). Read counts were averaged. This resulted in 24,910 islands in the C. elegans genome. Analysis of sequence conservation in CDR1as. Genome alignments of 32 verte￾brates were downloaded from the UCSC database (hg19)27 and analysed for the CDR1as locus. Primate species other than human were discarded to not bias the analyses. The one species (cow) with more than 50% gaps in the CDR1as locus was also discarded. The alignments for the seed regions were then corrected. Specifically, bases that would clearly align with the seed but had been separated in the alignment by runs of gaps were re-aligned. These corrections were neces￾sary in less than 1% of all seed sites. For an in-depth analysis we BLATed68 the human CDR1as sequence with 20- nucleotide flanking region against all vertebrate genomes in the UCSC genome browser and kept only hits that in turn aligned best to the human locus. The resulting sequences were used to build a multiple species alignment with MUSCLE69. The same corrections were applied as described above. This align￾ment was also used for Supplementary Fig. 4. Entropy was calculated in log2 units and averaged across all alignment columns bracketing each human seed site by maximally 8 nucleotides. Analysis of miR-7 base-pairing within CDR1as. RNAcofold70 was used to co￾fold miR-7 with each of the 74 binding regions within CDR1as defined as the miR-7 seed match TCTTCC and the next 16 bases upstream. Single-molecule RNA fluorescence in situ hybridization (smRNA FISH). 48 oligonucleotide probes (20 nucleotides length; spacing 2 nucleotides) comple￾mentary to the CDR1as transcript were designed using the Stellaris Probe Designer version 2.0 (Biosearch Technologies) with a masking level of 4 on the human genome to achieve high probe specificity (Supplementary Table 8). Stellaris probe pools were obtained from BioCat GmbH as conjugates coupled to Quasar 670 (a Cy 5 replacement). Flp-In T-REx 293 cells (Life Technologies) were grown exponentially and seeded into LabTek 4-well chambered coverslips (1 to 2 3 105 cells per well). Hybridizations were performed according to the manufacturer’s instructions with 50 ng ml21 DAPI as nuclear counterstain; Stellaris probes were hybridized at 125 nM concentration with a stringency of 10% formamide in overnight hybridizations at 37 uC. Images were acquired on an inverted Nikon Ti microscope with a Hamatsu ORCA R2 CCD camera, a 603 NA 1.4 oil objective and Nikon NIS-Elements Ar software (version 4), using an exposure time of 50 ms for DAPI and 1–1.5 s for Quasar 670. Groups of cells for imaging were chosen in the DAPI channel; Z-stacks were acquired in the Quasar 670 channel using 0.3 mm spacing and comprised a total depth of 6.5 mm (5 mm below and 1.5 mm above the middle of the nucleus) and merged using maximum intensity. Mouse strains and in situ hybridizations. All mice were bred and maintained in the animal facility of the Max Delbru¨ck Centrum under specific pathogen-free conditions, in plastic cages with regular chow and water ad libitum. All aspects of animal care and experimental protocols were approved by the Berlin Animal Review Board (REG 0441/09). B6129SF1/J wild-type adult, newborns (postnatal day 1) or pregnant females (plug detection at day 0.5; embryo collection at day 13.5) were used, as indicated for each experiment, to obtain the tissues needed for RNA analysis and in situ hybridizations (ISH). After death, embryos or tissues were immediately frozen in liquid nitrogen and stored at 270 uC, or fixed for ISH. Mouse brain structures were collected and named according to the anatomical guidelines of the Gene Expression Nervous System Atlas of the Rockefeller University (http://www.gensat.org) and the Mouse Brain Atlas (http://www. mbl.org/mbl_main/atlas.html). For the RNA analysis and to clone CDR1as-specific RNA probes, two adult 1-year-old mice of both sexes were dissected, total RNA prepared and analysed. If embryos or newborns were sectioned, a minimum of two specimens were eval￾uated; in some instances up to 5 specimens were used. For ISH, samples were fixed in formalin (13PBS; 4% formaldehyde) for 12 h and post-fixed (70% ethanol, 18 h) before dehydrating and paraffin-embedding. Next, the organs were perfused with a standard protocol using a Shandon XP Hypercentre. For ISH mouse embryos or organs were cut in RNase-free condi￾tions at 6 mm and ISH was performed as described48 with digoxigenin (DIG)- labelled RNA probes. All DIG–RNA probes were hybridized at 58 uC overnight. A total of 600 ng of the labelled probes was used per slide. To amplify Cdr1 sense and antisense sequences for ISH probe preparation a standard PCR-amplification was performed using mouse cerebellum cDNA. Three Cdr1as amplicons were generated, two of which probes are meant for the detection of both linear and circular forms using mmuCdr1_1f 59- TGCCAGTACCAAGGTCTTCC-39 and mmuCdr1_1r 59-TTTTCTGCTGGA AGATGTCAA-39, as well as mmuCdr1_2f 59-CCAGACAATCGTGATCT TCC-39 and mmuCdr1_2r 59-ATCTTGGCTGGAAGACTTGG-39. In addition a probe was generated, specific to the circular probe, using the divergent primers mmuCdr1_as_7f 59-CCACATCTTCCAGCATCTTT-39 and mmuCdr1_as_ 7r 59-TGGATCCCTTGGAAGACAAA-39 (CDR1_as head to tail probe). All ensuing fragments were subcloned into pCR-BluntII-TOPO (Invitrogen) and verified by sequencing. Linearized plasmids were amenable for in vitro transcrip￾tion using the T7 (antisense) or SP6 (sense) polymerase and a DIG-label nucleo￾tide mixture according to manufacturer’s instruction (Roche Applied Science). LNA ISHs were performed according to a protocol suggested by the manufac￾turer (Exiqon) with minor modifications. For individual LNAs, specific protocols were run at 51 uC (miR-7; 38485-15) or 58 uC (miR-124; 88066-15) on an InsituPro VS robot (Intavis). A pre-hybridization step was added, which con￾sisted of an incubation of the slides at 15 uC lower than the hybridization tem￾perature for 30 min using hybridization buffer. The antibody-blocking step was performed in the presence of 1% mouse blocking reagent (Roche 11096176001) and 10% sheep serum. The LNA probes were used at the following concentrations: miR-7 40 nM; miR-124 20 nM; U6 snRNA 1 nM; scrambled 40 nM, as suggested by miRCURY LNA microRNA ISH Optimization kit (Exiqon; 90004). Before detection all slides were washed 43 in NTMT including 1 mM Levamisole. The doubly DIG-labelled LNAs were detected by the alkaline phosphatase using the substrate BM-purple (Roche; 11442074001) at 37 uC. siRNA- and shRNA-mediated knock down. CDR1as was knocked down using custom designed siRNA oligonucleotides (Sigma) and Lipofectamine RNAiMax (Invitrogen). 2 3 106 HEK293 cells were transfected with 10 nM siRNA duplex following the manufacturer’s protocol. After 12–16 h cells were harvested and subjected to RNA analysis. For stable knock down of CDR1as, 293TN cells were co-transfected with the packaging plasmids pLP1, pLP2 and the VSV-G plasmid ARTICLE RESEARCH ©2013 Macmillan Publishers Limited. All rights reserved

RESEARCH ARTICLE (Invitrogen)and pSicoR constructs"(sequences available in the Supplementary dre B-actin forward primer, 5'-TGCTGTTTTCCCCTCCATTG-3': reverse pri- Table 8)by calcium phosphate transfections. Viral super were harvested mer, 5'-TTCTGTCCCATGCCAACCA-3'; probe sequence FAM-5'-TGGAC after 24 h and 48 h post transfection and filtered through a 0. 44 um filter. For GACCCAGACATCAGGGAGTG-3'-TAMRA. infection the viral supernatants supplemented with fresh medium and 6 ug ml For measuring the expression of dre-miR-7a/b we used Applied Biosystem lybrene was added to target cells. After overnight infection cells were allowed to Taq Man miR assays(IDo00268, ID001088) r 12 h and subjected to a second round of infection. Cells were collected 48-72 h after the first infection. The list of siRNA oligonucleotides is provided in Langmead, B. Salzberg, S L Fast gapped-read alignment with Bowtie 2.Nature supplementary Table 8 Methods9,357-359(2012) 53. Aroian. R V Field liere, G, Kenyon, C. Alberts. B M. Isolation of actin- tutional ethical guidelines. The Tg(huC egfp) and the Tg(Xia. Tubb: ds RED)trans- 54. L'Hernault, S W.& Roberts, T M Cell biology of nematode sperm. Methods Cell oligomers(Gene Tools)were prepared at a stock concentration of I mM accord. 55. Stoeckius M etal Large-scale sorting c eleganse m to the manufacturer's protocoL Sequences: control morpholino, 5'-CTC 56. T ACCTCAGTTACAATTTATA-3(control crits and isoform switching during cell differentiation. Nature geting miR-7, 5-ACAACAAAATCACAAGTCTTCCACA-3(miR-7 morphe techno.28,511-515(2010 no). For titration experiments we used 15 ng of control morpholino and 9 and 57. Hinrichs, A Set al. The UCSC Genome Browser Database: update 2006. Nucleic 15 ng of miR-7 morpholino; for all other experiments we used 9 ng miR-7 mor- cids Res34,D590D598(2006 pholino.3 nl of morpholinos were injected into the yolk of single-cell-stage 58. Pollard, K. S et al. Detection of nonneutral substitution rates on mammalian s9.Bhe0101212010 A 673-nucleotide mouse Cdrlas fragment was amplified from mouse cerebel lists. Curr. Protoc. Mol Biol Ch. 19. Unit 19 trogen). The vector was linearized with Kpnl or Apal (Fermentas)in vitro Genome Res 15. 1451351455(4005 or interactive large-scale genome analysis. lar cDNA and the amplicon was subcloned into a pCR- Blunt II Topo vector 60. Giardine, B et al. Galax anscribed (IVT)using T7 and SP6 RNA polymerases(Promega)and the result- 61. Goecks, J, Nekrutenko, A, Taylor,J&Galaxy, TGalaxy:a comprehensive ing Cdrlas and reverse complement Cdrlas_control products were used for jections(1.5nl of 100 ngn )into the cell of single-cell-stage embryos In a 62. Griffiths.Jones, S The microRNA Registry. Nucleic Acids Res 32,D109-D111 repetition of these experiments the Carlas fragment amplicon was directly used Prin emplate for IVT by exploiting T7-promoter extended forward and reverse 63. Landthaler, M et al.Molecular ribonucleoprotein complexes and their bound target mRNAs RNA 14, Approximately 1.5 nl of a 50 ng ul construct(backbone pCS2+) e human linear or the human circular CDRlas was injected int 14 1244-1253(2008) ingle-cell-stage embryos(provided by the Kjems laboratory). For alignment with Burrows-Wheeler 1.5nl pre-miR-7 precursor(7 HM, pre-miR miRNA precursor ID PM10047 from 66. Zisoulis, D G et al Comprehensive discovery of endogenous Argonaute Applied Biosystems). The negative control was the vector pCS2+ without insert Confocal imaging was performed using a Zeiss LSM 510 microscope( Car! 67(2010) (empty vector, 50 ng ul Zeiss Microimaging) equipped with a 25X objective(NA=0.8). Embryos were (2009) Nucleic Acids Res 38, W695-W699(2010) ZEN software. Midbrain and telencephalon volumes were calculated using 70. Bernhart, S H et al Partition function and base pairing probabilities of RNA Imaris 64x7.6.1 software based resolution three- dimensional stacks heterodimers Algorithms Mol BioL. 1, 3(2006). obtained from To(Xia. Tubb: ds RED)embryos. Reduced midbrain development 71. Ventura. A et was defined as >50%6 smaller than the mean size of controls Proc. Natl Acad. Sci. USA 101, 10380-10385(2004) erference from Each experimental group was evaluated in at least three independent experi- 72. Westerfield, M. The Ze A Guide for the Laboratory Use of Zebrafish ments: a minimum of 80 individual embryos per group were examined. Data are (Brachydanio rerio)2n Oregon Press, 1993). expressed as mean t standard deviation. Statistical analysi performed using 73. Krueger, J et al. FItI udents't-test, and a P<0.05 was considered statistically significant. Expression of miR-7 in zebrafish embryos at 48 hours post fertilization was 74. Friedlander, MR, Mack Li, N. Chen, w ormalized to expression of B-actin. In the miR-7 morpholino group, only accurately indentifies known and hundreds of novel microRNA genes in seven mbryos with a midbrain phenotype were used for the RNA expression analysis nimal clades. Nucleic Acids Res 40, 37-52(2012) @2013 Macmillan Publishers Limited. All rights reserved

(Invitrogen) and pSicoR constructs71 (sequences available in the Supplementary Table 8) by calcium phosphate transfections. Viral supernatants were harvested after 24 h and 48 h post transfection and filtered through a 0.44 mm filter. For infection the viral supernatants supplemented with fresh medium and 6 mg ml21 polybrene was added to target cells. After overnight infection cells were allowed to recover for 12 h and subjected to a second round of infection. Cells were collected 48–72 h after the first infection. The list of siRNA oligonucleotides is provided in Supplementary Table 8. Zebrafish methods. Zebrafish and their embryos were handled according to standard protocols72 and in accordance with Max Delbru¨ck Centrum insti￾tutional ethical guidelines. The Tg(huC:egfp) and the Tg(Xia.Tubb:dsRED) trans￾genic zebrafish lines have been described elsewhere49,50. Morpholino antisense oligomers (Gene Tools) were prepared at a stock concentration of 1 mM accord￾ing to the manufacturer’s protocol. Sequences: control morpholino, 59-CTC TTACCTCAGTTACAATTTATA-39(control morpholino) and morpholino tar￾geting miR-7, 59-ACAACAAAATCACAAGTCTTCCACA-39 (miR-7 morpho￾lino). For titration experiments we used 15 ng of control morpholino and 9 and 15 ng of miR-7 morpholino; for all other experiments we used 9 ng miR-7 mor￾pholino. 3 nl of morpholinos were injected into the yolk of single-cell-stage embryos. A 673-nucleotide mouse Cdr1as fragment was amplified from mouse cerebel￾lar cDNA and the amplicon was subcloned into a pCR-Blunt II Topo vector (Invitrogen). The vector was linearized with KpnI or ApaI (Fermentas) in vitro transcribed (IVT) using T7 and SP6 RNA polymerases (Promega) and the result￾ing Cdr1as and reverse complement Cdr1as_control products were used for injections (1.5 nl of 100 ng nl21 ) into the cell of single-cell-stage embryos. In a repetition of these experiments the Cdr1as fragment amplicon was directly used as a template for IVT by exploiting T7-promoter extended forward and reverse primer. Approximately 1.5 nl of a 50 ng ml 21 construct (backbone pCS21) expressing the human linear or the human circular CDR1as was injected into the cell of single-cell-stage embryos (provided by the Kjems laboratory). For rescue experi￾ments the construct containing the circular CDR1as was injected together with 1.5 nl pre-miR-7 precursor (7 mM, pre-miR miRNA precursor ID PM10047 from Applied Biosystems). The negative control was the vector pCS21 without insert (empty vector, 50 ng ml 21 ). Confocal imaging was performed using a Zeiss LSM 510 microscope (Carl Zeiss MicroImaging) equipped with a 253 objective (NA 5 0.8). Embryos were anaesthetized using 0.1% tricaine and mounted in 1% agarose as described73. Confocal stacks were acquired of the brain region and processed using Zeiss ZEN software. Midbrain and telencephalon volumes were calculated using Imaris 6437.6.1 software based on high-resolution three-dimensional stacks obtained from Tg(Xia.Tubb:dsRED) embryos. Reduced midbrain development was defined as .50% smaller than the mean size of controls. Each experimental group was evaluated in at least three independent experi￾ments; a minimum of 80 individual embryos per group were examined. Data are expressed as mean 6 standard deviation. Statistical analysis was performed using Students’t-test, and a P , 0.05 was considered statistically significant. Expression of miR-7 in zebrafish embryos at 48 hours post fertilization was normalized to expression of b-actin. In the miR-7 morpholino group, only embryos with a midbrain phenotype were used for the RNA expression analysis. dre b-actin forward primer, 59-TGCTGTTTTCCCCTCCATTG-39; reverse pri￾mer, 59-TTCTGTCCCATGCCAACCA-39; probe sequence FAM-59-TGGAC GACCCAGACATCAGGGAGTG-39-TAMRA. For measuring the expression of dre-miR-7a/b we used Applied Biosystems TaqMan miR assays (ID000268, ID001088). 52. Langmead, B. & Salzberg, S. L. Fast gapped-read alignment with Bowtie 2. Nature Methods 9, 357–359 (2012). 53. Aroian, R. V., Field, C., Pruliere, G., Kenyon, C. & Alberts, B. M. Isolation of actin￾associated proteins from Caenorhabditis elegans oocytes and their localization in the early embryo. EMBO J. 16, 1541–1549 (1997). 54. L’Hernault, S. W. & Roberts, T. M. Cell biology of nematode sperm. Methods Cell Biol. 48, 273–301 (1995). 55. Stoeckius, M. et al. Large-scale sorting of C. elegans embryos reveals the dynamics of small RNA expression. Nature Methods 6, 745–751 (2009). 56. Trapnell, C. et al. Transcript assembly and quantification by RNA-Seq reveals unannotated transcripts and isoform switching during cell differentiation. Nature Biotechnol. 28, 511–515 (2010). 57. Hinrichs, A. S. et al. The UCSC Genome Browser Database: update 2006. Nucleic Acids Res. 34, D590–D598 (2006). 58. Pollard, K. S. et al. Detection of nonneutral substitution rates on mammalian phylogenies. Genome Res. 20, 110–121 (2010). 59. Blankenberg, D. et al. Galaxy: a web-based genome analysis tool for experimentalists. Curr. Protoc. Mol. Biol. Ch. 19, Unit 19 10 11-21 (2010). 60. Giardine, B. et al. Galaxy: a platform for interactive large-scale genome analysis. Genome Res. 15, 1451–1455 (2005). 61. Goecks, J., Nekrutenko, A., Taylor, J. & Galaxy, T. Galaxy: a comprehensive approach for supporting accessible, reproducible, and transparent computational research in the life sciences. Genome Biol. 11, R86 (2010). 62. Griffiths-Jones, S. The microRNA Registry. Nucleic Acids Res. 32, D109–D111 (2004). 63. Landthaler, M. et al. Molecular characterization of human Argonaute-containing ribonucleoprotein complexes and their bound target mRNAs. RNA 14, 2580–2596 (2008). 64. Rudel, S., Flatley, A., Weinmann, L., Kremmer, E. & Meister, G. A multifunctional human Argonaute2-specific monoclonal antibody. RNA 14, 1244–1253 (2008). 65. Li, H. & Durbin, R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 25, 1754–1760 (2009). 66. Zisoulis, D. G. et al. Comprehensive discovery of endogenous Argonaute binding sites in Caenorhabditis elegans. Nature Struct. Mol. Biol. 17, 173–179 (2010). 67. Langmead, B., Trapnell, C., Pop, M. & Salzberg, S. L. Ultrafast and memory-efficient alignment of short DNA sequences to the human genome. Genome Biol. 10, R25 (2009). 68. Kent,W. J. BLAT–the BLAST-like alignment tool. Genome Res. 12, 656–664 (2002). 69. Goujon, M. et al. A new bioinformatics analysis tools framework at EMBL-EBI. Nucleic Acids Res. 38, W695–W699 (2010). 70. Bernhart, S. H. et al. Partition function and base pairing probabilities of RNA heterodimers. Algorithms Mol. Biol. 1, 3 (2006). 71. Ventura, A. et al. Cre-lox-regulated conditional RNA interference from transgenes. Proc. Natl Acad. Sci. USA 101, 10380–10385 (2004). 72. Westerfield, M. The Zebrafish Book: A Guide for the Laboratory Use of Zebrafish (Brachydanio rerio) 2nd edn (Univ. Oregon Press, 1993). 73. Krueger, J. et al. Flt1 acts as a negative regulator of tip cell formation and branching morphogenesis in the zebrafish embryo. Development 138, 2111–2120 (2011). 74. Friedla¨nder, M. R., Mackowiak, S. D., Li, N., Chen, W. & Rajewsky, N. miRDeep2 accurately indentifies known and hundreds of novel microRNA genes in seven animal clades. Nucleic Acids Res. 40, 37–52 (2012). RESEARCH ARTICLE ©2013 Macmillan Publishers Limited. All rights reserved