Review Overlapping genes in vertebrate genomes

Computational Biology and Chemistry29(2005)1–12ReviewOverlapping genes in vertebrate genomesIzabela Makalowska a,∗,Chiao-Feng Lin b,Wojciech Makalowski b,ca The Huck Institute of the Life Sciences,The Pennsylvania State University,502Wartik Lab,University Park,PA16802,USAb Institute of Molecular Evolutionary Genetics and Department of Biology,The Pennsylvania State University,512Mueller Lab,University Park,PA16802,USAc Department of Computer Science and Engineering,The Pennsylvania State University,University Park,PA16802,USAReceived1November2004;received in revised form15December2004;accepted15December2004AbstractOverlapping genes in mammalian genomes are unexpected phenomena even though hundreds of pairs of protein coding overlapping genes have been reported so far.Overlapping genes can be divided into different categories based on direction of transcription as well as on sequence segments being shared between overlapping coding regions.The biologic functions of natural antisense transcripts,their involvement in physiological processes and gene regulation in living organisms are not fully understood.Number of documented examples indicates that they may exert control at various levels of gene expression,such as transcription,mRNA processing,splicing,stability,transport,and translation. Similarly,evolutionary origin of such genes is not known,existing hypotheses can explain only selected cases of mammalian gene overlaps which could originate as result of rearrangements,overprinting and/or adoption of signals in the neighboring gene locus.©2004Elsevier Ltd.All rights reserved.Keywords:Overlapping genes;Anti-sense;Anti-transcript;Evolution;Genome organizationContents1.Introduction (2)2.Number and types of overlapping genes (2)putational methods for overlapping genes identification (4)4.Function of antisense transcripts (6)4.1.Transcriptional interference (6)4.2.RNA masking (6)4.3.dsRNA-dependent mechanism and RNA interference (6)5.Overlapping genes evolution (7)6.Overlapping genes and human diseases (9)7.Conclusions (9)Acknowledgement (9)References (9)∗Corresponding author.Tel.:+18148652695;fax:+18148636699.E-mail address:izabelam@(I.Makalowska).1476-9271/$–see front matter©2004Elsevier Ltd.All rights reserved.doi:10.1016/pbiolchem.2004.12.0062I.Makalowska et al./Computational Biology and Chemistry29(2005)1–121.IntroductionViruses have very compact genomes and yet,the discovery of the overlapping genes in bacteriophage phiX174in1976 (Barrell et al.,1976)came as surprise.It took another decade before similar phenomena were noticed in higher eukary-otes.In1998,in the same issue of Nature,Spencer et al. (1986)published information on two overlapping genes in Drosophila and Williams and Fried(1986)reported similar pattern in mouse.Number of reports of overlapping genes in human(Bristow et al.,1993;Cooper et al.,1998;Duhig et al.,1998;Edgar,2003;Kennerson et al.,1997;Kiyosawa and Abe,2002;Laabi et al.,1994;Morel et al.,1989;Nicolaides et al.,1995;Ohinata et al.,2002;Petrukhin et al.,1998;Zhou and Blumberg,2003),mouse(Batshake and Sundelin,1996; Kasper et al.,2002;Liu et al.,1999;Tvrdik et al.,1999),rat (Adelman et al.,1987;Lazar et al.,1989),chicken(Farrell and Lukens,1995;Zuniga Mejia Borja et al.,1993),Xeno-pus(Kimelman and Kirschner,1989),drosophila(Misener and Walker,2000;Misra et al.,2002),Acidia(Swalla and Jeffery,1996),yeast(Malavasic and Elder,1990;Peterson and Myers,1993),Arabidopsis(Glover et al.,1998;Ito et al., 1997;Osato et al.,2003;Quesada et al.,1999),rice(Osato et al.,2003)followed.However,overlapping genes in mammalian genomes are still considered as unexpected phenomenon and regardless of numerous reports about overlapping genes in eukaryotes, until recently overlapping genes were not considered to be important and large scale event in human and other verte-brate rge scale EST and genomic sequence stud-ies(Kiyosawa et al.,2003;Lehner et al.,2002;Shendure and Church,2002;Veeramachaneni et al.,2004;Yelin et al., 2003)let to conclusion that overlapping genes,are commonly present in higher eukaryotes.2.Number and types of overlapping genesA reliable large scale analysis of overlapping genes was not possible until complete genomes became ck of genomic sequences and good genome maps lead to ex-clusion of some possible gene overlaps from analyzed data sets.A good example could be human TPR and PRG4genes which were wrongly mapped to different chromosomes and therefore excluded as candidate overlapping genes(Burke et al.,1998).However,the overlap between these genes was confirmed by studies of PRG4gene involved in the develop-ment of arthropathy-camptodactylly syndrome(Marcelino et al.,1999).Thefirst evidence that antisense transcription is a common feature of eukaryotic genomes came from analysis of available human mRNAs(Lehner et al.,2002)although only87genomic loci encoding natural antisense transcripts were identifiputational analysis of human and mouse EST sequences(Shendure and Church,2002)led to identi-fication of144human and73mouse antisense transcripts. Although an analysis of EST sequences adds a magnitude of input information a significant portion of dbEST is thought to represent artifacts(Sorek and Safer,2003)and may lead to false positive results.Moreover,many EST libraries are non-directional and therefore the orientation of an EST can-not be easily determined.Also,in relatively many cases,the orientation of EST is mis-annotated.On the other hand some events of gene overlaps were impreted as EST artifacts and taken as ambiguous data(Wolfsberg and Landsman,1997).Total number of overlapping genes in human and other nuclear genomes is still unknown(Boi et al.,2004). Veeramachaneni et al.(2004),based on analysis of annotated human and mouse genes,reported774pairs of human over-lapping protein coding genes and578such cases in mouse. Out of these542pairs in human and455in mouse shared not only locus but also exonic sequences.Yelin et al.(2003) reported2667antisense transcripts in human and Kiyosawa et al.(2003)identified2481overlapping gene pairs in mouse. However,these two studies included protein coding as well as non-protein-coding genes but there is a report that about 70%of all gene overlaps include at least one non-protein-coding gene(Kiyosawa et al.,2003).Our recent analysis of seven vertebrate genomes(based on Ensemble annotations) show differences in the fraction of overlapping genes be-tween vertebrate lineages(Table1).In tetrapoda over10% of all genes are involved in some type of overlaps,while in fish only5–7%.The exception is the rat genome where onlyTable1Overlapping genes in various vertebrae genomesNumber of genes Number ofgenes inoverlapsNumberofoverlapsNested genes Embeddedgenes aTail to tail a Head to head a Exon overlaps a CDS involved*Human222912978(13.36%)1766972(55.04%)179(22.92%)414(52.14%)198(24.94%)634(79.85%)417(52.52%) Chimpanzee215062219(10.32%)1276665(52.12%)144(23.57%)320(52.37%)147(24.06%)479(78.40%)317(51.88%) Mouse253833456(13.67%)20531071(52.17%)192(19.55%)465(47.35%)325(33.10%)819(83.40%)565(57.54%) Rat221591080(4.87%)607458(75.45%)45(30.20%)53(35.57%)51(34.23%)102(68.46%)100(67.11%) Chicken177091960(11.07%)1135474(41.76%)226(34.19%)268(40.55%)167(25.26%)511(77.31%)471(71.26%) Fugu20796993(4.77%)556174(31.29%)150(39.27%)118(30.89%)114(29.84%)290(75.92%)290(75.92%) Zebrafish235241625(6.99%)1026767(74.76%)46(17.76%)115(44.40%)98(37.84%)98(37.84%)85(32.82%) Analysis was done based on the September2004Ensemble data.a Percentage calculated after exclusion of nested genes,which are not typical anti-transcribed genes.Nested genes are these which entire sequence lies in the single exon of another gene.I.Makalowska et al./Computational Biology and Chemistry29(2005)1–123Fig.1.Different types of overlapping genes.(a)Genes sharing the same locus on the same strand,however coding for different proteins.(b)Genes sharing promoter region.(c)Nested gene.(d)Embedded gene.(e)Genes on opposite strands with overlapping locus but no overlap in the exonic region.(f)Tail-to-tail overlap in the exonic region.(g)Head-to-head overlap involving3 -UTRs and coding sequence.Dark(red)boxes:coding sequence;light(blue)boxes: untranslated regions;patterned(green)box:promoter region.4.87%of all genes are overlapping;however this is rather due to annotation quality and not some specific features of the rat genome.Obviously,this could be a case forfish genomes as well.Although,these studies give some idea about the magnitude of the phenomena,the real number remains to be estimated.Overlapping genes may be divided into several categories. Fig.1shows some examples of overlapping human genes. Thefirst examples(Fig.1a)presents eukaryotic cellular gene encoding two proteins in different reading frames.Multiple examples of such gene overlap were reported(Gangopadhyay et al.,1997;Jankowski et al.,1986;Sloan et al.,1999; Stallmeyer et al.,1999).In this type of overlap transcripts share the same locus,however both are on the same DNA strain.Another case of overlapping genes(Fig.1b)are genes on opposite strands sharing the promoter region where a promoter functions bi-directionally(Bachman et al.,1999; Heikkila et al.,1993;Ito et al.,2000;Joseph,1998;West et al.,2003).Here,though genes are located on opposite strands the overlap is only in promoter region and transcripts do not share any sequence.Among genes where there is an overlap between transcripts lying on opposite strands we can distin-guish overlaps which do not involve any sharing of exonic sequence(Fig.1c–e)and genes sharing not only locus but also their mature transcripts share fragments of mRNA coding or untranslated regions.Among anti sense overlaps we can dif-ferentiate head-to-head overlaps,genes overlapping by their 5 ends(Fig.1g),and tail-to-tail(Fig.1e and f)when3 ends are involved in the overlap,and embedded genes(Fig.1c–d), a case when one gene lies completely in the area of another one,with its special class of nested genes(Fig.1c).In case of nested genes one gene is not only completely in the area of another one but also is located in a single intron.Nested genes are not necessary always going the opposite direction than their‘host’gene and for a long time was the only cat-egory considered as a common feature of human and other eukaryotes genomes.Very often more than one gene may be nested in the same intron.For instance intron27of hu-man neurofibromatosis type1gene(NF1)harbors three other genes:OMG,EVI2B,and EVI2A.Fraction of various overlap pairs differ among studies,nevertheless,in all reports tail-to-tail overlap represent majority of cases.Veeramachaneni et al. (2004)found that66.42%of antisense transcripts in human and61.54%in mouse belong to this category.The fraction of genes overlapping at5 end was30.81%and36.92%in human and mouse,respectively.Similar results are reported by Yelin et al.(2003).However,Lehner and Shendure identi-fied significantly smaller fractions of head-to-head overlaps, 15%(Lehner et al.,2002)and5.53%(Shendure and Church, 2002),respectively.4I.Makalowska et al./Computational Biology and Chemistry29(2005)1–12Our latest studies(see Table1)show that in various ver-tebrates31.19–75.45%overlapping genes are nested(here understood as genes embedded completely in a single intron of another gene).Among the rest30.89–52.37%(in fugu and chimpanzee,respectively)are genes overlapping at3 ends, and24.06%up to37.84%(in chimpanzee and zebrafish,re-spectively)are head-to-head overlaps.The remaining fraction of overlapping genes(after excluding nested genes)are em-bedded genes,these that are entirely inside another gene,but span more than one intron or share some exonic fragments. The analysis of overlaps types shows that relatively lower percentage of overlapping genes in rat,fugu and zebrafish, 4.87%,4.77%and6.99%,respectively(Table1),may re-sulted from lack of annotated untranslated regions since al-most all exon overlaps include coding sequences while in human,chimpanzee and mouse big fraction of exon overlaps do not involve coding sequence.In fugu all exon overlaps involve coding sequence from both genes.Many overlaps between UTRs or UTRs and coding sequence could be there-fore lost if similarity and model based methods are used in genome annotation and annotated genes very often do not have associated cDNA or EST sequence.This could also be a case of the chicken genome where majority annotation was done based on computational methods and therefore many UTRs may not be predicted.We can see that although per-centage of overlapping genes in the chicken genome is at similar level as in human,mouse and chimpanzee,almost all exon overlaps involve coding sequence.putational methods for overlapping genesidentificationAvailability of complete or near complete genome se-quences for various species and huge collections of expressed sequences,such ESTs or mRNAs,made possible to perform genome wide overlapping genes analysis.Although several large-scale studies on antisense transcripts were published in the last2years the methodology for overlapping genes iden-tification is not fully established and results of that are not compatible.One approach to overlapping genes identification is to search for complementary regions in full-length mRNA ing BLAST algorithm Lehner et al.(2002)iden-tified regions of complementarity between pairs of mR-NAs.They identified372naturally occurring human antitran-scripts.However,after excluding complementarity resulting from the presence of repetitive elements and mapping these transcripts to the human genome sequence only87pairs ap-peared to be real overlapping genes.In80cases each gene in a pair was transcribed from a different chromosomal location. Additional51pairs contained chimeric transcripts(having identity to more than one region of the genome)and possibly being artifacts of cDNA library construction.This approach is very reliable and allows to identify true overlapping genes. However,it returned only a small subset of existing over-lapping genes.Furthermore,all overlapping genes sharing a locus but not exonic sequences(Fig.1a)as well as these predicted or confirmed by EST sequences only were missed.In other studies identification of overlapping genes was done using both,mRNA and EST sequences.Shendure and Church(2002)used EST libraries that were directionally cloned and sequenced to search for UniGene clusters contain-ing a statistically significant number of misoriented ESTs.As a next step ESTs and mRNAs used in UniGene as clustering “seeds”were mapped to their genomic coordinates and eval-uated if putative sense and antisense ESTs in a given Uni-Gene cluster represent distinct mRNA species.Evaluation was done based on the differential intron–exon splicing struc-tures,the locations of poly(A)signals and tails,and pattern of mouse-human sequence conservation.Finally,the subset of predictions was validated by orientation-specific reverse transcription PCR using this approach Shendure and Church identified144human and73mouse UniGene clusters con-taining two distinct oppositely oriented mRNAs but clustered together as a consequence of a bidirectionally transcribed re-gion of overlap.Experimental validation revealed33out of 39gene pairs tested to be truly representing antisense tran-scripts.Yelin et al.(2003)also used ESTs for antitranscripts pre-diction but they used the entire set of available ESTs and not only selected UniGene clusters.First,software called LEADS was used in order to clean sequences from vector,re-peats and highly abundant genes,and then to align sequences to the genome.Overlapping expressed sequences were as-sembled into‘clusters’representing genes.Analysis of all human expressed sequences yield61,048clusters,exclud-ing singletons and doubletons.Next,an‘Antisensor’algo-rithm,developed by the authors,was applied to detect clus-ters with sequences from opposite strands.This algorithm, based on presence of mRNA in a cluster,annotation of se-quence orientation,splice junction consensus,and poly(A) tail sequence,identified2667clusters that contained putative sense–antisense pairs.Each of them was separated by‘Anti-sensor’into two new clusters,each representing a single gene. From these,a random sample of264pairs was selected for analysis by microarrays and154pairs,almost60%,showed signals for both sense and antisense transcripts leading to the estimate that there are about1600pairs of overlapping genes in the human genome.In two other studies done by Kiyosawa et al.(2003)and Veeramachaneni et al.(2004),full-length cDNA data,with-out ESTs,mapped to genomic sequences were used to com-pile datasets of overlapping genes.Kiyosawa,as an input, used∼61,000full-length cDNA sequences from FANTOM2 project(Okazaki et al.,2002)in addition to the mouse mRNA that existed in the GenBank and mapped all of them to the mouse genome.He identified2481pairs of sense–antisense transcripts out of which in more than70%cases at least one of transcripts represented non-coding RNA.Veeramachaneni et al.(2004)used similar approach and analyzed assembled and annotated human and mouse genomic sequences fromI.Makalowska et al./Computational Biology and Chemistry 29(2005)1–125GenBank.Location of each gene,in term of mRNA loca-tion,was stored in the database and each gene coordinates were used in order to identify overlapping loci.Each exon intervals were utilized in order to mark genes sharing not only the same locus but also exonic fragments and therefore representing sense–antisense transcripts.The main difference in presented approaches is in used datasets:ESTs only,mRNAs only,or ESTs and mRNA com-bined,and in using or not using genomic mapping and genes coordinates on ing ESTs gives higher mag-nitude of input information.Unfortunately,a significant num-ber of ESTs represents artifacts (Sorek and Safer,2003)and not true expressed sequences.Also,orientation for many ESTs cannot be easily determined and in some cases the ori-entation of EST is ing mRNA sequences gives higher level of confidence in the results,however,mRNA datasets do not represent all known expressed se-quences and therefore many gene overlaps may be missed.In addition,large set of mRNAs have missing 3 or 5 end which “masks”the presence of overlap.Mapping RNAs and ESTs to genomic sequence helps in increasing the level of confidence and to avoid getting false-positives due to repeats,vectors or the presence of highly abundant and nearly identical gene paralogs.Mapping of ESTs,and all known mRNAs to the genomic sequences seems to be the best approach to identify overlapping genes.It allows not only to avoid false positive predictions but also to predict overlapping genes with non-overlapping exons.These cannot be detected by transcripts analysis only.Our recent studies on annotated genomes of human,mouse,rat,chicken,chimpanzee,zebrafish and fugu (Table 1)show that combination of annotated genes,sequenced and predicted,together with EST contigs may bring reliable re-sults.Mapping all genes and ESTs guarantees taking advan-tage of all possible information,and employing contigs in-stead of raw or just clustered EST helps to reduce number of false positives since assembled contigs should contain much less artifacts.While inspecting our data obtained based on an-notated in Ensemble genes we noticed many instances where orthologs of human overlapping genes were not overlapping but neighboring in other species genomes.To investigate this we aligned TIGR gene indices (Pertea et al.,2003)to genomic fragments non-overlapping containing orthologs of human overlapping genes.In many cases we were able to extend an-notated and/or predicted transcripts and observe gene overlap (Fig.2).This directed us to using annotations of sequenced genes together with predicted genes and EST contigs in future overlapping genes studies.Methods and tools for detecting overlapping genes in higher eukaryotes are not well developed at the moment.Identifying overlapping genes using model-based methods only is impossible and we can only rely on already known expressed sequences.Majority of gene prediction programs still do not allow overlaps between genes.In addition these programs predict coding region only while a lot of overlaps are observed in untranslated ing transcripts in-formation is not free from problems either.ESTs contain a lot of chimeras and artifacts and together with problem with orientation labeling these data may lead to many false posi-tives.Developing ‘Antisensor’algorithm allowed Yelin et al.(2003)to overcome some problems associated with problem-atic ESTs.However,the set of predicted antisense transcripts still contained pairs which could not be experimentally veri-fied as such.Lavorgna et al.(2004a)developed an online antisense de-tection tool ‘AniHunter’,aimed at facilitating the in silico identification of potential antisense expressed sequence tags within a given genomic region of interest.Program takes a genomic sequence and a list of annotated transcripts of the genomic region as input.Then it runs the RepeatMasker in or-der to filter out repetitive elements,performs BLASTN search versus EST database,parses the BLASTN output looking for antisense EST with respect to the annotated genes.As source of orientation information,the AntiHunter uses the database annotation of the sequence,as well as the splice junctions and the presence of a poly(A)tail in ESTs.Unfortunately,the accuracy of this server is still not perfect and many false positives are usually returned.In addition,response time is very long and might take even several hours to get results for 500kb sequence.In conclusion,out of several approaches used to find over-lapping genes,mapping all mRNA sequences together with EST contigs to the genomic sequences and analysis of these genes coordinates seems to be optimal at the moment,be-cause it allows to use all available information and at the same time minimizes the number of falsepositives.Fig.2.Example of using EST contigs to extend annotated on rat genome transcripts leading to gene overlap.In upper box annotated are two Ensemble genes:ENSRNOG00000017532and ENSRNOG00000017485,in lower box EST contigs from TIGR gene indices (Sequin graphical output of BLASTN results saved in ASN.1format).6I.Makalowska et al./Computational Biology and Chemistry29(2005)1–124.Function of antisense transcriptsThe presence of complementary transcripts in prokary-otes have already been showed to have many functional roles (Inouye and Delihas,1988).Although the effects of eukary-otic antisense RNAs on the corresponding sense RNAs have not yet been established,a number of documented exam-ples indicate that they may exert control at various levels of gene expression,such as transcription,mRNA process-ing,splicing,stability,transport,and translation.In many cases the complementary transcript does not encode for a protein product,but regulates expression by hybridizing to the mRNA of the regulated gene.Numerous instances of complementary transcripts have been seen in eukaryotic or-ganisms but only few pairs have been examined experi-mentally and no generalization concerning the mechanism of action exists.Three mechanisms by which these tran-scripts regulate gene expression were recognized:transcrip-tional interference,RNA masking,and double-stranded RNA (dsRNA)-dependent mechanism(Lavorgna et al.,2004b; Munroe,2004;Noguchi et al.,1994).4.1.Transcriptional interferenceTranscription by RNA polymerase II involves both large protein complexes and the unwinding of duplex DNA.It may be unlikely that two overlapping transcriptional units, with or without exon sharing,could be transcribed at the same time.Investigation of GAL10and GAL7rearrangement in Saccharomyces cervisiae(Prescott and Proudfoot,2002) showed that when genes were arranged convergently,but not overlapping,both are transcribed at normal levels.However, when both genes overlap level of the message is reduced.The gene encoding human eIF2alpha protein has an associated antisense RNA and transcription of the sense and antisense transcripts are inversely correlated(Noguchi et al.,1994). These data suggest that the expression of cis-NAT(naturally occurring antisense transcripts)partners could be regulated through competitive transcriptional interference.However, at other loci,sense and antisense transcription is not always observed to be mutually exclusive.For example,studies by Edgar(2003)on human ABHD1and Secl2genes show that there is no inverse correlation between the respective lev-els of both RNA species and there is no indication for RNA interference.ABHD1and Sec12(PREB)mRNAs were ex-pressed in many(all tested by authors)cultured cell types, being found in smooth muscle,fibroblast,endothelial,epithe-lial and blood cell types.The expression infibroblasts was significantly lower than in other cell types.On average,the expression level of ABHD1was1.4%that of Sec12.Sim-ilarly,the histidyl-tRNA synthetase gene and its antisense gene HO3have distinct tissue expression(O’Hanlon et al., 1995)although it is still unclear whether expression interfer-ence occurs or not.There is also no evidence of interaction of antitranscribed RNAs in the P450c21/TNX system(Bristow et al.,1993).Two Drosophila genes,P5cr and Prp18overlap at their3 ends(Misener and Walker,2000).Both genes have housekeeping function,suggesting that the transcripts show no interference despite their complementarity.This is consis-tent with the observation of the luck of phenotypic effect in transgenic Drosophila expressing anti-sense P5cr under heat shock promoter control.4.2.RNA maskingFormation of RNA duplexes between sense and antisense transcripts might mask some regulatory signals within either transcript and inhibit the binding of trans-acting factors.A good example of this way of antisense expression regulation is thymidylate synthase ErbAalpha(THRA in current nomen-clature)the expression of which was inversely correlated with the level of overlapping at3 end thymidylate synthase Rev-ErbAalpha(NR1D1)(Chu and Dolnick,2002).Thymidylate synthase mRNA is cleaved in a site-specific manner suggest-ing that it is down regulated through a natural RNA-based antisense mechanism.The thyroid hormone receptor gene, c-erbAlpha(THRA),overlaps with RevErb(NR1D1)in a tail-to-tail orientation(Lazar et al.,1989;Miyajima et al., 1989).The c-erbAlpha gene has two isoforms and there is evidence that the expression of isoform alpha2is negatively regulated by antisense interaction with the complementary RevErb mRNA(Hastings et al.,1997;Lazar et al.,1990). These results show that an antisense RNA can specifically inhibit the alternative splicing of an mRNA,probably by blocking the accessibility of cis regulatory elements.4.3.dsRNA-dependent mechanism and RNA interferenceThere is an evidence that interaction of antisense gene pairs can also affect gene expression through the activation of dsRNA-dependent pathways.These may include RNA editing,or RNA interference(RNAi)-dependent gene silenc-ing.Interaction between sense and antisense transcripts may result in the formation of rge double-stranded regions may be modified by the RNA-editing machinery. Kimelman and Kirschner(1989)described a small tail-to-tail complementary transcript arising from the opposite strand of the gene encoding Xenopus basic growth factor bFGF. This antisense NUDT6gene encodes for a protein,but also directly interacts with the sense transcript converting half of the hybridized adenine residues to inosine,thus chang-ing the information content of the sense transcript and caus-ing modification that would trigger rapid degeneration of the reacting RNAs.Hyper-edited transcripts could be rec-ognized by proteins,which inhibit their export from the nu-cleus and prevent translation.The antisense transcript of p53 prevented normal translocation of p53mRNA to the cyto-plasm in urine erythroleukemia(Khochbin and Lawrence, 1989).However,in this case the antisense transcript was a non-protein coding RNA.Alternatively,dsRNAs might be digested into small fragments(small interfering RNAs)。