Evolution of MicroRNAs and the Diversification of Species

Evolution of MicroRNAs and the Diversi?cation of Species Yong-Hwee E.Loh,Soojin V .Yi,and J.Todd Streelman*

School of Biology and Petit Institute for Bioengineering and Bioscience,Georgia Institute of Technology

*Corresponding author:E-mail:todd.streelman@https://www.wendangku.net/doc/513413426.html,.

Accepted:9December 2010

Abstract

MicroRNAs (miRNAs)are ancient,short noncoding RNA molecules that regulate the transcriptome through post-transcriptional mechanisms.miRNA riboregulation is involved in a diverse range of biological processes,and misregulation is implicated in disease.It is generally thought that miRNAs function to canalize cellular outputs,for instance as ‘‘fail-safe’’repressors of gene misexpression.Genomic surveys in humans have revealed reduced genetic polymorphism and the signature of negative selection for both miRNAs themselves and the target sequences to which they are predicted to bind.We investigated the evolution of miRNAs and their binding sites across cichlid ?shes from Lake Malawi (East Africa),where hundreds of diverse species have evolved in the last million https://www.wendangku.net/doc/513413426.html,ing low-coverage genome sequence data,we identi?ed 100cichlid miRNA genes with mature regions that are highly conserved in other animal species.We computationally predicted target sites on the 3#-untranslated regions (3#-UTRs)of cichlid genes to which miRNAs may bind and found that these sites possessed elevated single nucleotide polymorphism (SNP)densities.Furthermore,polymorphic sites in predicted miRNA targets showed higher minor allele frequencies on average and greater genetic differentiation between Malawi lineages when compared with a neutral expectation and nontarget 3#-UTR SNPs.Our data suggest that divergent selection on miRNA riboregulation may have contributed to the diversi?cation of cichlid species and may similarly play a role in rapid phenotypic evolution of other natural systems.

Key words:cichlid,single nucleotide polymorphism,riboregulation,evolutionary radiation,minor allele frequency.Introduction Ever since King and Wilson compared protein sequence be-tween chimpanzee and human and concluded that there was insuf?cient coding divergence to explain phenotypic differences (King and Wilson 1975),biologists have high-lighted regulatory change in gene expression as a source for adaptive evolution (Wray 2007;Carroll 2008).There is now ample direct evidence that cis -acting mutations cause phenotypic variation among closely related organisms by modulating gene expression (Sucena et al.2003;Miller

et al.2007).These data,coupled with the signature of divergent and positive selection at putative gene regulatory el-ements (Haygood et al.2007;Sethupathy et al.2008),have established the general consensus that 5#promoters act as evolutionary engines of transcriptional change (e.g.,tinker where the tinkering’s good;Rockman and Stern 2008).Plausible scenarios for the evolution of animal diversity hinge on the ever-growing complexity of 5#promoters and the modi?cation of transcriptional regulatory networks (Levine and Tjian 2003).Notably,evolutionary ‘‘tinkering’’with transcription at 5#promoters may have evolved in con-cert with post-transcriptional safeguards encoded at the 3#end of cistrons.Reports suggest that microRNAs (miRNAs),potent agents of riboregulation,are as old as metazoan 5#cis -regulatory logic (Grimson et al.2008;Wheeler et al.2009).miRNAs are short (;22nucleotides),endogenous noncoding RNA molecules that regulate gene expression after transcription.Generally,animal miRNA targeting is achieved by complementary base pairing between the miRNA and the speci?c sequences in the 3#-untranslated region (3#-UTR)of messenger RNAs (mRNAs).Target recog-nition is thought to be determined by perfect Watson–Crick base pairing at a miRNA ‘‘seed’’region (base positions 2–7counting from the 5#end;Lewis et al.2005),although this is not a necessary condition,and targeting may include other determinants (Grimson et al.2007;Barbato et al.2009).Transcript silencing then occurs through inhibition of trans-lation or via mRNA degradation (Bartel 2004).Individual miRNAs may regulate hundreds of loci,and it has been es-timated that a majority of human genes are potential miRNA targets (Lewis et al.2005;Friedman et al.2009).aThe Author(s)2011.Published by Oxford University Press on behalf of the Society for Molecular Biology and Evolution.

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (https://www.wendangku.net/doc/513413426.html,/licenses/by-nc/

2.5),which permits unrestricted non-commercial use,distribution,and reproduction in any medium,provided the original work is properly cited.

GBE

by guest on January 13, 2011

https://www.wendangku.net/doc/513413426.html,

Downloaded from

miRNAs generally act as ‘‘fail-safe’’buffers against gene misexpression in time and/or space,in effect canalizing the transcriptome (Carrington and Ambrose 2003;Stark et al.2005).Consistent with this notion,miRNA misexpression and/or genetic polymorphism in target sequences can cause abnormality and disease (Clop et al.2006;Eberhart et al.2008;Sethupathy and Collins 2008;Menc?′a et al.2009).Likewise,and in contrast to predicted transcription factor binding sites in 5#promoters,human miRNAs and their 3#-UTR target sequences evolve under purifying selection

(Chen and Rajewsky 2006;Saunders et al.2007;Sethupathy

et al.2008).As humans and chimps diverged from a common ances-

tor during the last 5–7My,the East African Rift lakes Tanganyika,Malawi,and Victoria spawned three of the most spectacular evolutionary radiations known to biology (Korn?eld and Smith 2000;Salzburger et al.2005).In Lake Malawi alone,hundreds of cichlid ?sh species have evolved from a common ancestor over the last million years (Won et al.2005).These species are remarkably diverse in size,shape,color,and behavior (Streelman et al.2003;Albertson et al.2005;Carleton et al.2008;Fraser et al.2008;Sylvester et al.2010),yet their genomes are highly similar and share ancestral polymorphism (Moran and Korn?eld 1993;Loh et al.2008).We have shown recently that most of the genome is not genetically differentiated among Malawi species and major lineages;only 2–4%of single nucleotide polymorphism (SNP)loci exhibit the statistical signature of strong evolutionary divergence (Loh et al.2008).Cichlids are models of the mapping of phenotype to genotype;the problem of so many biological species in so little time (Kocher 2004)is equally matched by the problem of rapid diversi?ca-tion and evolutionary novelty (Streelman et al.2007).

We hypothesized that divergence of miRNAs or their tar-

get sequences might be one of the genomic mechanisms contributing to the rapid phenotypic evolution observed in Lake Malawi cichlids.To this end,we analyzed available low-coverage genome sequence and SNP data (Loh et al.2008)and computationally identi?ed 1)putative cichlid miRNAs and 2)the target sequences in 3#-UTRs to which miRNAs may bind.Most studies of miRNA focus on evolu-tionary conservation of the molecules and their target sites (Bartel 2004;Alexiou et al.2009;Barbato et al.2009).Our goal of evaluating the link(s)between miRNAs,polymor-phism in putative miRNA targets and diversity among Lake Malawi cichlid species predicates that we not only consider target sequences conserved for hundreds of millions of years but also those that may have evolved more recently.Such ‘‘nonconserved’’targets are known to be functional and may be generated by single mutations to standing sequence (Farh et al.2005;Clop et al.2006).We observed that predicted cichlid mature miRNAs are strongly conserved in sequence.On the other hand,miRNA targets exhibited greater SNP densities than ?anking sequences and the overall 3#-UTR average.Moreover,poly-morphic sites in target sequences showed higher minor allele frequencies (MAFs)and divergence among Malawi evolutionary lineages when compared against a neutral expectation and nontarget SNPs in the same set of 3#-UTRs.Our data reveal a signature of divergent selection on cichlid miRNA binding sites and suggest an evolutionary role for miRNA riboregulation in the diversi?cation of species.

Materials and Methods

Lake Malawi Genomes

We obtained Lake Malawi cichlid genomic data,consisting of 304,310sequences from ?ve species,25,458multispe-cies alignments,and 32,417SNPs,from a previous study (Loh et al.2008),which applied various criteria to ensure that alignments are allelic and not products of paralogous loci.Sequence data were generated by the Sanger method,allowing the detection of variable sites with an even distri-bution across the data set and with high con?dence (Loh et al.2008).Examination of these data and subsequent gen-otyping revealed very low genetic variation,and the persis-tence of ancestral polymorphism across the Malawi cichlid ?ock.Molecular genetic analyses across multiple cichlid spe-cies are thus highly analogous to within-species polymor-phism studies conducted in other organisms (e.g.,humans,Chen and Rajewsky 2006;Saunders et al.2007).Our use of the term ‘‘SNP’’in this context therefore extends to in-clude variable sites across multiple cichlid species (For more details,see Loh et al.2008).

miRNA Gene Detection

A database of 623known teleost precursor miRNA (pre-miRNA)sequences was downloaded from miRBase release 14.0(Grif?ths-Jones et al.2008).To detect miRNA genes in cichlids,we conducted a BlastN similarity search of these pre-miRNAs against the cichlid genomic sequences as described above,with an E value cutoff of 0.001.The BlastN hits were then manually inspected and compared with their query sequences in order to extract adjacent nucleotides that might form part of the pre-miRNA.RNA secondary structure of the cichlid putative miRNA sequences was pre-dicted using Mfold (Zuker 2003)to ensure proper stem–loop folding,and excess bases were trimmed.A reciprocal BlastN of the putative cichlid miRNAs against known teleost miR-NAs was performed to identify the cichlid miRNA and to assign orthology.Multiple sequence alignments of the pu-tative cichlid miRNAs and their orthologs were then gener-ated using ClustalW (Larkin et al.2007).Mutations in the alignments were marked and counted based on the region (mature miRNA,stem,or loop)where they reside.Loh et al.GBE

by guest on January 13, 2011

https://www.wendangku.net/doc/513413426.html, Downloaded from

3#-UTR Annotation Cichlid genomes have yet to be fully sequenced and anno-tated;therefore,we ?rst annotated cichlid 3#-UTRs from partial genomic sequence.We chose to work with genomic and not transcript sequences because our ultimate goal was to map SNPs to putative miRNA targets found within 3#-UTRs (below);SNP data exist for genome survey sequen-ces (Loh et al.2008)but not for the small number of publicly available cichlid expressed sequence tags (ESTs).Sequences used to annotate cichlid 3#-UTRs include Fugu rubripes ,Tet-raodon nigroviridis ,Oryzias latipes ,Gasterosteus aculeatus ,and Danio rerio proteins (98,037entries)downloaded from Ensembl version 56,all ‘‘Actinopterygii’’proteins (41,746entries)from Refseq release 39,and all ‘‘Eukaryota’’proteins (158,696entries)from UniProtKB/Swiss-Prot release 2010_02databases.We applied the TBlastN algorithm with an E value cutoff of 1?10à10to identify similarity between the protein se-quences above and the cichlid multispecies alignments (Loh et al.2008).High-scoring segment pairs (HSPs)of the TBlastN output with lengths of at least 30amino acids were parsed and retained,and in cases where the end position of a HSP query was found to be within three amino acids from the known 3#end of the full-length query protein,it was deemed that a corresponding cichlid coding region might

also have ended in this region.We further looked within the ±9nucleotide region of the HSP subject (i.e.,cichlid)end to con?rm that a stop codon was indeed present and in frame with codon phase of the HSP .Cichlid 3#-UTRs were thus annotated to begin at the next nucleotide beyond the stop codon and presumed to continue for 500nucleo-tides in length.This approximation of 3#-UTR length was based on a calculation of the mean 3#-UTR length in zebra ?sh (513nucleotides),as annotated by Ensembl.During our work on this project,an additional ;56,000unique ESTs were released for the tilapia cichlid,roughly 10–15My divergent from the Malawi assemblage (Lee et al.2010).Comparing our annotations with these data,we observed that 66%of our predicted 3#-UTRs showed signi?cant sim-ilarity (E value ,1?10à5)to ESTs.miRNA Target Prediction A total of 249unique mature miRNA sequences,consoli-

dated from the 623known pre-miRNAs from Fugu ,Tetrao-don ,and Danio (miRBase)and the 100derived from miRNA loci in cichlids (this study),was used for the prediction of target sites on annotated cichlid 3#-UTRs.The target predic-tion algorithm (hereby termed the SeedMatch algorithm)was written in Perl programming language,implementing the seed-matching requirements similar to that of Target ScanS (Lewis et al.2005):namely,1)a six nucleotide Watson–Crick complementary match between miRNA and mRNA at positions 2–7of the miRNA plus 2)an anchor of either an adenosine at the mRNA target aligned to miR-NA position 1and/or a Watson–Crick match at position 8of the miRNA.

Conservation of predicted cichlid miRNA target sites in other ?sh species was determined by 1)generating multiple sequence alignments (MLAGAN;Brudno et al.2003)of cichlid 3#-UTRs and their orthologs (when determined)in puffer ?shes,medaka,stickleback,and zebra ?sh,2)apply-ing the SeedMatch algorithm separately to each sequence in the multiple alignment to identify target sites,and 3)calling a cichlid target site conserved,when an identical target site was found in at least one other ?sh at a location within 50nucleotide positions along the alignment.We de?ned conservation as such,in contrast to other target prediction strategies requiring strict conservation across multiple species (Alexiou et al.2009;Barbato et al.2009)for two reasons.First,the ?shes with complete genome sequences noted above are all at least 100My divergent from Malawi cichlids.Second,?sh genomes are generally more divergent with greater neutral nucleotide substitution rates compared with mammals (Brunet et al.2006).The latter consideration in?uences the degree of target conservation observed be-tween species and also our initial task of generating robust multiple sequence alignments.

Target SNP Density Calculations

Subsequent to predicting miRNA target sites on 3#-UTRs,we mapped SNPs to these same data (Loh et al.2008).For statistical analysis of observed SNP densities in predicted miRNA targets,we obtained a distribution of randomized target SNP densities by running 1,000simulations that permute the occurrence of SNPs along the 3#-UTRs.In each simulated run,every empirical SNP in the 3#-UTRs was shuf?ed to a random position maintaining the same trinu-cleotide sequence (i.e.,the SNP position itself and the nu-cleotides immediately before and after).For example,a G[A/T]C trinucleotide where [A/T]represents the SNP would be shifted to a random GAC or GTC position.The ‘‘randomized’’target SNP density was then calculated for each run.This simulation strategy controls for neighbor-

dependent mutation rates and has been used previously

to investigate SNP densities in miRNA target sites (Hiard

et al.2010).

3#-UTR Resequencing,Alignment,and Target Prediction

The analyses described above using data from Loh et al.(2008)allow us to identify cichlid miRNAs,their putative targets,and to calculate SNP densities in target sequence.However,because those data do not represent full genomes from the ?ve species sequenced,alignments of orthologous sequence rarely contain more than three species (Loh et al.2008).To better understand,evolutionary processes acting on putative cichlid miRNA target sequences,we Evolution of miRNAs and the Diversi?cation of Species GBE

by guest on January 13, 2011

https://www.wendangku.net/doc/513413426.html, Downloaded from

resequenced annotated 3#-UTRs in a diverse and standard-ized collection of species.Polymerase chain reaction primers were designed (supplementary file 3,Supplemen-tary Material online)and used for ampli?cation and se-quencing of a subset of annotated 3#-UTRs from the genomic DNA of eight individuals:Labeotropheus fuellebor-ni (LF),Melanochromis auratus (MA),and Maylandia zebra (MZ)are members of the rock-dwelling mbuna lineage;Tyrannochromis maculiceps (TM),Docimodus evelynae (DE),Nimbochromis polystigma (NP),and Mchenga cono-phoros (MC)belong to a sister lineage of pelagic and sand-dwelling species (henceforth termed non-mbuna);and Rhamphochromis esox (RE)represents an early diverg-ing,deepwater group within the radiation (pictures at https://www.wendangku.net/doc/513413426.html, ).The individuals of LF ,MA,MZ,MC,and RE were those survey sequenced by the JGI (Loh et al.2008).Sequences were aligned using ClustalW (Larkin

et al.2007),from which polymorphic positions were iden-ti?ed at locations exhibiting at least seven species depth of

coverage (supplementary file 5,Supplementary Material online).We applied the target site prediction algorithms and SNP density calculations to these data as described above.We also carried out additional analyses,described below,with these resequenced data.MAFs of SNPs in Resequenced 3#-UTRs We calculated the MAF of each SNP (in and out of putative miRNA targets)identi?ed in the resequenced data set.We then compared these MAF distributions with a neutral expectation.From a set of 70nongenic SNPs typed across a diverse mix (183individuals,62species)of Lake Malawi cichlids (Cichlid Genome Consortium,Broad Institute),we randomly sampled eight individuals to match our rese-quenced 3#-UTR data set (three mbuna,four non-mbuna,and one deepwater species)and calculated the allele frequency distribution of the sample.This process was re-peated 1,000times to approximate a neutral distribution of allele frequencies and the 95%con?dence intervals at each allele frequency.Because we sequenced and re-sampled eight individuals or 16total alleles,the empirical and simulated allele frequency data are largely discrete,with the majority of observations falling around multiples of 1/8(0.125).Therefore,bins were set around multiples of 0.125,and bin edges fall at the midpoint of consecutive bins;for example,the ?rst bin edge (0.1675)is the midpoint be-tween 0.125and 0.25.Z -tests were implemented within each allele frequency bin to detect signi?cant shifts in the

proportion of SNPs exhibiting that particular range of MAFs between empirical and resampled neutral distributions.Genetic Differentiation of High-MAF Target SNPs in Resequenced 3#-UTRs We observed that SNPs in predicted targets exhibited higher MAFs than expected under neutrality.To test,whether these high-MAF (31.25,MAF ,50%)miRNA target SNPs exhibited greater genetic differentiation among Malawi lin-eages than expected under neutrality,we generated 1,000sets of matching ‘‘neutral’’genotype data using the same nongenic SNP data set and sampling strategy as described above.For each set of genotypic data,we calculated for each SNP the 1)overall population,2)mbuna,and 3)non-mbuna allele frequencies,where each allele frequency value lies between 0and 1.We de?ned an SNP as displaying clear lineage-speci?c differentiation when the difference in mbuna and non-mbuna allele frequencies was equal or greater than 0.75,and hence calculated the proportion of high-MAF SNPs that were well differentiated between lin-eages.Values were aggregated for the 1,000data sets to obtain a distribution from which a Z -test was used to deter-mine the statistical signi?cance of our observed data.Results

miRNA Prediction

We used a reference set of 623known teleost pre-miRNA sequences from Fugu ,Tetraodon ,and Danio ,obtained from miRBase release 14.0(Grif?ths-Jones et al.2008),in a similarity search (see Materials and Methods)against a da-tabase of 304,310cichlid genomic sequences (Loh et al.

2008).We manually curated the similarity hits to extract

putative cichlid pre-miRNAs and con?rmed that they were able to fold into the secondary stem–loop structure neces-sary for miRNA biogenesis (Bartel 2004).This resulted in the identi?cation of 100distinct cichlid pre-miRNA genes (sup-plementary file 2,Supplementary Material online)that pro-duce 87unique mature miRNAs.

We compared cichlid pre-miRNA loci with their ortho-logues in other ?sh species and found a total of 1002of 6422nucleotide positions where substitutions had oc-curred.This results in an overall nucleotide divergence of 0.156(variable sites/nucleotide positions).When the pre-miRNAs were divided into mature miRNA,stem,and loop regions (?g.1A ),we observed nucleotide divergences of 0.015,0.172,and 0.485,respectively (?g.1B ),with no mutations found in the miRNA seeds.A similar trend of region-speci?c variation holds for the subset of substitu-tions,where cichlids exhibit a different nucleotide than all other species;a divergence of 0.008,0.060,and 0.185at the mature miRNA,stem,and loop regions,respectively (?g.1B ).

Polymorphism in Cichlid miRNA Targets

To study genetic variation in putative cichlid miRNA targets,we mapped SNPs (Loh et al.2008)to target sequences pre-dicted to fall within 3#-UTRs.We ?rst annotated 731cichlid

3#-UTRs (supplementary file 4,Supplementary Material on-

line)that contained 367SNPs (0.28%SNP density).To direct our computational prediction of targets,we used 249Loh et al.GBE

by guest on January 13, 2011

https://www.wendangku.net/doc/513413426.html, Downloaded from

unique mature miRNAs,derived from miRNA loci in cichlids (above)as well as known miRNAs from other ?sh species Fugu ,Tetraodon ,and Danio .These miRNAs are highly con-served among vertebrates;86%are in miRNA families that extend outside of ?shes.Note that the 100cichlid miRNAs we identi?ed here (above)possess identical seed sequences to their ?sh orthologues;this justi?es our use of additional ?sh miRNAs,conserved among vertebrates but not yet iden-

ti?ed in cichlids (see below),to facilitate target prediction.

Putative miRNA binding sites in 3#-UTR sequences were

predicted using a Perl script written to implement a ‘‘Seed-

Match’’algorithm incorporating rules similar to those of

TargetScanS (Lewis et al.2005).Brie?y,7-and 8-mer target

sites were identi?ed that had exact Watson–Crick base pair

matches at seed sequences (positions 2–7counting from the

miRNA 5#end)plus a corresponding base anchor at position

1and/or 8(see Materials and Methods).

Considering all putative 3#-UTRs identi?ed from the Loh

et al.(2008)data,we detected 6,299miRNA target sites on

719of 7313#-UTR sequences (an average of 8.62miRNA

target sites per 3#-UTR;table 1).As expected,we observed

overlaps among predicted target sites for multiple miRNAs;

13.0%of the total 3#-UTR length (39,660nucleotides)were

predicted to be bound by one or more miRNA(s),similar to

results reported in human and mouse (Hiard et al.2010).

Seventy-eight SNPs mapped within 17,607informative ba-ses of miRNA target sites.Thus,the SNP density for miRNA

target sites is 0.44%,higher than 1)the average 3#-UTR SNP density (0.28%),2)the SNP densities of target-?anking sequence (0.21–0.28%),and 3)the average randomized target SNP density of 0.28%(Z -test,P 52.41?10à6;?g.2A ).For reference,the SNP densities of synonymous and replacement coding sites in the same set of data are 0.42%and 0.20%,respectively (Loh et al.2008).

Enforcing a criterion of target site conservation reduced

the size of our data set considerably (see Materials and

Methods and below;table 1).We assigned orthology to

single genes in other ?sh genomes for 481of 731predicted cichlid 3#-UTRs.Other predicted 3#-UTRs showed similarity to members of gene families or to speci?c pairs of dupli-cated loci,but we could not specify reciprocal orthology with con?dence.Conserved sites accounted for 21%of cichlid miRNA targets (875of 4,182),similar to previous study (Friedman et al.2009;Hiard et al.2010),and

covered

F I

G .1.—Evolutionary divergence in pre-miRNA sequences.(A )An example of predicted stem–loop secondary structure for a cichlid miRNA

(lfu-mir-199-1shown here),classi?ed into separate regions for analysis.Nucleotide symbols are colored red for the mature miRNA region,blue for the loop region,and gray for the stem region excluding the mature miRNA.Vertical bars represent Watson–Crick or G:U base-pairing matches.(B )Distribution of divergence across different regions of the pre-miRNA.Bar colors correspond to the regions de?ned in A ,with black representing the divergence over the entire molecule.Solid-colored bars are calculated from all observed variable sites.Shaded bars

are calculated from variable sites where cichlids displayed a different nucleotide than all other

species.F IG .2.—SNP densities within computationally predicted miRNA target sites and their ?anking regions on data from (A )all predicted 3#-UTRs and

(B )select resequenced 3#-UTRs.Flanking regions 1–2on both 5#and 3#ends of target represent successive,nonoverlapping windows of sizes equal to that of the target sites.Dotted lines show the average 3#-UTR SNP density.Filled circle with error bars represents the mean and 95%con?dence intervals of SNP densities calculated from 1,000simulated replicates of randomized SNP shuf?ing.Asterisk symbols indicate signi?cant deviation from simulated distributions (Z -test,*P ,10à5,**P ,10à9).

Evolution of miRNAs and the Diversi?cation of Species GBE

by guest on January 13, 2011

https://www.wendangku.net/doc/513413426.html, Downloaded from

only 2.7%of nucleotides in these 4813#-UTRs.The SNP density in conserved target sites was 0.29%,similar to the average SNP density for ?anking and overall 3#-UTRs and within the 95%con?dence interval of randomized tar-get SNP densities (supplementary fig.S1,Supplementary Material online).MAFs and Genetic Differentiation of ‘‘Target’’SNPs in Resequenced 3#-UTRs We resequenced a set of 1303#-UTRs in eight individuals of Malawi cichlid species spanning a range of morphologies and behaviors,representing the three major evolutionary lineages in the lake (Won et al.2006;Loh et al.2008).Our rationale here was 2-fold.First,we reasoned that 3#-UTR sequence variation across samples,in and out of putative miRNA target sites,could be examined for the evo-lutionary signature of natural selection (Chen and Rajewsky 2006;Saunders et al.2007).Second,in order to better val-idate predicted miRNA–mRNA interactions against previous literature,we chose certain gene subsets whose molecular functions have been well characterized for interactions with miRNAs (e.g.,development,Plasterk 2006;immunity,Xiao and Rajewsky 2009).From 48,114base positions of multiple sequence align-ments (supplementary file 5,Supplementary Material on-line),we identi?ed 160SNPs,an overall SNP density of 0.33%.We then applied the SeedMatch algorithm to these data.SeedMatch targets covered 6,602total bases,within which we mapped 40SNPs (table 1).This resulted in an SNP

density in predicted targets of 0.606%,higher than the overall average in resequenced data (0.33%),target-?anking sequence (0.12–0.31%),and randomized target SNP densities (0.28%;Z -test,P 54.88?10à10;?g.2B ).Similar to the analysis of all putative 3#-UTRs (above),en-forcing a strong conservation criterion for target sites re-duced the size of the data set (only 4.8%of 3#-UTR bases are covered by conserved target sites).Conserved sites accounted for 36%of all targets on 124cichlid 3#-UTRs;the empirical SNP density in conserved targets was 0.32%,el-evated from ?anking sequence but similar to the overall 3#-UTR and randomized target SNP densities (supplemen-tary fig.S1,Supplementary Material online).

Next,we examined the allele frequency distribution of

SNPs in predicted miRNA target sites in relation to 3#-UTR

nontarget sites compared against a neutral expectation.We approximated a neutral distribution by subsampling from a data set of 70randomly chosen,nongenic SNPs typed in a diverse mix of Lake Malawi cichlids.Signi?cant departure from a neutral distribution of allele frequencies might be in-dicative of natural selection (Chen and Rajewsky 2006;Drake et al.2006;Sethupathy et al.2008).Notably,allele frequen-cies at nontarget 3#-UTR SNPs did not depart from the neutral distribution (nearly 80%of polymorphisms exhibit minor al-leles that are relatively rare)but predicted target SNPs differed signi?cantly,with a bias toward high MAFs (?g.3;supple-mentary fig.S2,Supplementary Material online).

We asked if high-MAF SNPs in predicted miRNA targets were differentiated among lineages (i.e.,mbuna vs.non-mbuna)to a degree beyond expectation under neutrality.We found that a signi?cantly elevated proportion (86%)of high-MAF (31.25,MAF ,50%)target SNPs exhibit ge-netic differentiation between Malawi evolutionary groups (Z -test,P 59.32?10à7).Predicted miRNA–gene interac-tions,highlighting evolutionarily differentiated SNPs,are shown in ?gure 4and as discussed below.

Discussion

Lake Malawi cichlids have evolved in a brief evolutionary window.Their genomes are highly similar and segregate an-cestral polymorphism.For comparison,nucleotide diversity across the ?ock (0.26%,Loh et al.2008)is less than that observed among laboratory strains of the zebra ?sh (0.48%,Guryev et al.2006),comparable to that of chimpanzees (0.24%,Fischer et al.2004)and humans (0.11%,Table 1.

miRNA Target Prediction Results on All Putative and Select Resequenced 3#-UTRs

All 731Putative 3#-UTRs 130Resequenced 3#-UTRs

All (731UTRs)Conserved Targets (481UTRs)Nonconserved Targets (481UTRs)All (130UTRs)Conserved Targets (124UTRs)Nonconserved

Targets

(124UTRs)

Number of targets predicted 6,2998753,3071,296360639Number of targets (per 3#-UTR)8.62 1.82 6.889.97 2.90 5.15Total coverage of targets (nt)39,6605,50521,1576,6022,1594,0893#UTR coverage by targets (%)13.0 2.710.513.7 4.769.01Informative sites within targets a (nt)

17,6072,7619,3556,6022,1594,089Number of SNPs in targets 7884040729SNP density in targets (%)0.4430.2900.4280.6060.3240.709

a Only a subset of 3#-UTR positions had multispecies sequence data to determine polymorphism.

Loh et al.GBE

by guest on January 13, 2011

https://www.wendangku.net/doc/513413426.html, Downloaded from

International Hapmap Consortium 2007),and contrasts against the ;1.2%divergence between chimps and hu-mans (King and Wilson 1975;Chen and Li 2001).It is no-

table,then,that the range of variation across Malawi species for many phenotypes (body size,tooth,and taste bud number)spans an order of magnitude and that the di-versity of other traits (color pattern,feeding and breeding biology,brain organization)is comparable to that observed in other vertebrate taxonomic orders.The cichlid system is thus a model of the genotype to phenotype mapping func-tion (Streelman et al.2007),with speculation revolving around the rapid evolution of novelty.Here,we test the hy-pothesis that evolutionary divergence of miRNAs and/or their binding sites may have contributed to the diversi?ca-tion of species (Plasterk 2006).

Cichlid miRNA Target Sites Exhibit Elevated SNP Densities

We identi?ed 100distinct miRNA loci in the genomes of cichlid ?shes.The mature miRNAs encoded by these loci are highly conserved among ?shes (?g.1B ).The trend of higher divergence in stems and loops (vs.the mature miR-NA)has been observed in other species (Hertel et al.2006)and may be indicative of purifying selection against change to the functional component of the miRNA molecule (and/or a relaxation of constraint at stems and loops).The

number

F I

G .4.—Multiple sequence alignments of several miRNA targets containing differentiated SNPs,discussed in the text.Red blocks indicate SNP minor alleles.Dashes represent gaps in sequence (indel in osr2).miRNAs predicted to bind to the target are shown,with the seed region in red font.Vertical bars represent Watson-Crick base pairing and colons represent G:U base pairing.Raised and lowered nucleotides illustrate bulges in the predicted miRNA binding.TM,Tyrannochromis maculiceps ;DE,Docimodus evelynae ;NP ,Nimbochromis polystigma ;MC,Mchenga conophoros ;LF ,Labeotropheus fuelleborni ;MA,Melanochromis auratus ;MZ,Maylandia zebra ;and RE,Rhamphochromis esox .Yellow,green,and blue boxes over abbreviated species names represent non-mbuna,mbuna,and deepwater lineages,

respectively.

F I

G .3.—Comparison of MAF distributions.3#-UTR miRNA target SNPs are colored in red,nontarget SNPs in blue,and nongenic (neutral)SNPs in black.Error bars represent the 95%con?dence interval of the neutral expectation.Asterisk symbols indicate signi?cant deviation from neutral expectation within each bin (Z -test,*P ,10à4).

Evolution of miRNAs and the Diversi?cation of Species GBE

by guest on January 13, 2011

https://www.wendangku.net/doc/513413426.html, Downloaded from

of miRNAs we identi?ed is likely to be an incomplete count,as the available cichlid genomic resources used here com-prise only about 32%coverage of the cichlid genome (Loh et al.2008).As a reference,there are 360zebra ?sh (characterized from an assembled genome and by deep RNA sequencing;Wienholds et al.2005;Soares et al.2009)and 132puffer ?sh miRNAs in miRBase.Predicted miRNA target sites,located in the 3#-UTRs of cichlid genes,showed elevated SNP densities when com-pared with ?anking regions,the overall 3#-UTR average and randomized simulations that account for nucleotide composition (table 1and ?g.2).For a more restricted set of evolutionarily conserved targets,SNP densities were not distinguishable from those in ?anks,the overall 3#-UTR average and simulation values.This trend held in both the genome-wide 3#-UTR data set and the directed set of resequenced 3#-UTRs.Our observation of elevated or equiv-alent SNP densities in both conserved and nonconserved miRNA targets runs counter to results from previous study within humans,where average SNP density in predicted tar-get sites (both conserved and nonconserved)was reduced compared with ?anking regions (Chen and Rajewsky 2006;Saunders et al.2007).miRNA Target Sites Show the Signature of Divergent Natural Selection The observation of increased SNP density at predicted miR-NA target sites does not provide conclusive information about the evolutionary forces shaping this pattern;for in-stance,even though the SNP density of predicted targets is high within the context of 3#-UTR sequence,minor alleles at variable sites could be rare.We therefore resequenced a collection of 3#-UTRs in a standard set of species and designed a test to evaluate the allele frequency distribution of 1)SNPs predicted in miRNA binding sites and 2)other 3#-UTR nontarget SNPs,against a neutral expectation.This test is conceptually similar to the derived allele frequency approach (Chen and Rajewsky 2006;Drake et al.2006;Sethupathy et al.2008).However,because Lake Malawi cichlid ?shes retain ancestral polymorphism that may pre-date the species ?ock (Loh et al.2008)we have not attemp-ted to designate ancestral versus derived alleles.We found that whereas the allele frequency distribution of nontarget SNPs in 3#-UTRs was not different than the neutral expectation,the distribution of predicted miRNA tar-get SNPs was biased toward high MAFs (?g.3).In addition,we observed that 86%of putative miRNA target SNPs with high MAFs showed a clear pattern of evolutionary diver-gence between major Malawi lineages (?g.4and below).To put this in greater context,we have previously observed that ,5%of haphazardly chosen SNPs are outliers for genetic differentiation in a large sample of mbuna versus non-mbuna (Loh et al.2008).The alternative that the dif-ferentiated polymorphisms we highlight in ?gure 4are not in fact in miRNA targets but are each physically linked to other,as yet unidenti?ed nucleotide sites,is unlikely be-cause it would require that we happened upon these un-identi?ed sites in six independent loci through the sole discovery operation of searching for miRNA targets.Taken together,our observations of 1)elevated SNP den-sities,2)a bias toward high MAFs,and 3)the pattern of ge-netic differentiation among lineages for high-MAF SNPs suggest that select miRNA binding sites have experienced divergent selection during the evolution of the Lake Malawi species ?ock.Differentiated SNPs in miRNA Targets are Biologically Relevant A secondary goal of our resequencing project was to inves-tigate putative miRNA binding site polymorphism in gene sets whose molecular functions have been well-studied vis-a `-vis miRNAs.We reasoned that such data would add biological plausibility to our computational predictions and population genetic analyses.Figure 4displays examples of high-MAF SNPs,genetically differentiated among Malawi cichlid lineages,mapped to miRNA target sites in 3#-UTRs.These examples represent miRNA–gene pairs supported by previous research in humans and other model organisms.The interplay between miRNAs and Hox gene riboregu-

lation is well known (Yekta et al.2008).We predict an as-

sociation between two miRNAs,miR-181c and miR-23a,which share a target site SNP in the cichlid hoxa103#-UTR (?g.4A );this target site in hoxa10is conserved between cichlid and stickleback.The SNP differentiates non-mbuna predators (TM,DE,NP)from other species.miR-181is known to target mouse Hoxa11(a Hox cluster family mem-ber of hoxa10)during muscle differentiation (Naguibneva et al.2006);?sh hoxa10genes are expressed in paired ?ns and associated musculature (Ahn and Ho 2008).Recently,it has been shown that miR-181is upregulated whereas miR-23is downregulated in mouse leg muscle during endurance exercise (Safdar et al.2009).These data raise the possibility that a single SNP modulates the miRNA riboregulation of Hox-mediated ?n muscle development and regeneration in Lake Malawi predators.We highlight two miRNA–gene pairs that may modify sensory modalities among Lake Malawi cichlids.We predict differential binding of miR-34to cichlid crb1(?g.4B ),a member of the Crumbs protein complex.crb1contributes to photoreceptor morphogenesis and sensitivity,mutations cause retinal degeneration in humans,mice,and ?ies (Bulgakova and Knust 2009).miR-34is expressed in neural tissue (including the optic tectum)of larval and adult zebra ?sh (Kapsimali et al.2007),also in the retina of embryonic and adult mice (Arora et al.2010).This association is of particular interest given the vast literature implicating the role of vision in Malawi cichlid ecology,mate choice,and evolution (Carleton et al.2008).Next,we predict that the Loh et al.GBE

by guest on January 13, 2011

https://www.wendangku.net/doc/513413426.html, Downloaded from

TRIO and F-actin binding protein (triobp )is differentially bound by miR-200a (?g.4C ).triobp functions in the hair cell cilia of the inner ear (Kitajiri et al.2010),mutations result in nonsyndromic hearing loss (Shahin et al.2006).miR-200a is expressed in sensory epithelia,including those of the inner ear of zebra ?sh,chicken,and mouse (Soukup 2009).Re-cent reports have linked hearing to mate choice and com-munication in East African cichlids (Simo ?es et al.2008;Verzijden et al.2010).Two SNPs are predicted to affect binding of miRNAs to genes involved in immune response (Xiao and Rajewsky 2009).fbxw5(?g.4D )is a F-box protein with a role in inter-leukin signaling (Minoda et al.2009);a T 4C SNP differen-tiated among Malawi cichlids is predicted to modulate binding of miR-122,a liver-speci?c miRNA (Sarasin-Filipowicz et al.2009;Soares et al.2009).The miR-122binding site in fbxw5is conserved between cichlid and medaka.Second,tfec (?g.4E )is a macrophage-restricted basic helix-loop-helix transcription factor,also involved in interleukin signaling (Rehli et al.2005).We predict that a differentiated G 4A SNP modi?es binding of miR-155,a well-known regulator of immune function (O’Connell et al.2009).Finally,our data may be useful to identify new interac-tions between miRNAs and genes of interest.For example,we predict that an indel in the 3#-UTR of cichlid osr2should differentially regulate binding of miR-740in mbuna cichlids (LF ,MA,MZ)versus others (?g.4F ).Osr2restricts the teeth of mice to a single row (Zhang et al.2009),among other functions in the craniofacial skeleton.Tooth row number is highly variable among cichlid species (Fraser et al.2008).miR-740is poorly understood (Kloosterman et al.2006);our data suggest it may play a role in craniofacial development.Conclusion Biologists recognize that 5#cis -acting mutations regulate gene expression and contribute to phenotypic evolution (King and Wilson 1975;Wray 2007;Carroll 2008).Cor-respondingly,studies have reported the signature of di-versifying selection on population genetic variants in computationally predicted 5#promoter elements (Haygood et al.2007;Sethupathy et al.2008).The situation is differ-ent for 3#-UTRs.miRNAs and their binding sites collaborate as post-transcriptional capacitors to canalize the transcrip-tome (Carrington and Ambrose 2003;Stark et al.2005).Ev-idence suggests that both miRNAs and their target sequences in 3#-UTRs evolve under purifying selection (Chen and Rajewsky 2006;Saunders et al.2007).Metazoan cis-trons may therefore have evolved for transcriptional explo-ration at 5#promoters,with post-transcriptional safeguards encoded at the back.We provide evidence that the evolution of miRNA binding sites may play a role in evolutionary diversi?cation.We dem-onstrate that 1)computationally predicted miRNA targets in cichlid 3#-UTRs harbor elevated SNP densities,2)a greater frequency of polymorphic sites in predicted targets have high MAFs compared with a neutral expectation,and 3)these sites are often genetically differentiated among Malawi lineages.It has been argued that polymorphisms in miRNA target sites are deleterious within species because even single base mismatches (especially to the seed)can abrogate binding and disrupt riboregulation (Clop et al.2006;Sethupathy et al.2008;Menc?′a et al.2009).We suggest that mutations in 3#-UTRs where miRNAs may bind,whether breaking tran-scriptome canalization or introducing new regulation,may contribute to phenotypic differentiation among rapidly evolving lineages.Further analyses,with fully annotated and assembled cichlid genomes (https://www.wendangku.net/doc/513413426.html,/10002154),deeper genotyping,next-generation miRNA,and miRNA target prediction algorithms (Chaudhuri and Chatterjee 2007;Barbato et al.2009),and experimental val-idation of predicted miRNAs and their interactions with tar-get genes (Kuhn et al.2008;Sethupathy and Collins 2008)will reveal additional intricacies of miRNA riboregulation and evolution.Supplementary Material Supplementary ?les 1–5are available at Genome Biology and Evolution online (https://www.wendangku.net/doc/513413426.html,/).Acknowledgments We thank Craig Albertson,Greg Gibson and two anony-mous reviewers for comments on previous drafts of this manuscript and Michael Norsworthy for experimental assis-

tance.This work was supported by the National Science

Foundation (IOS 0546423)and the National Institutes of

Health (R01DE019637).

Literature Cited Ahn D,Ho RK.2008.Tri-phasic expression of posterior Hox genes during development of pectoral ?ns in zebra?sh:implications for the evolution of vertebrate paired appendages.Dev Biol.322(1):220–233.Albertson RC,Streelman JT,Kocher TD,Yelick PC.2005.Integration and evolution of the cichlid mandible:the molecular basis of alternative feeding strategies.Proc Natl Acad Sci U S A.102(45):16287–16292.

Alexiou P ,Maragkakis M,Papadopoulos GL,Reczko M,

Hatzigeorgiou AG.2009.Lost in translation:an assessment and perspective for computational microRNA target identi?cation.Bioinformatics.25(23):3049–3055.Arora A,et al.2010.Prediction of microRNAs affecting mRNA expression during retinal development.BMC Dev Biol.10:1.Barbato C,et https://www.wendangku.net/doc/513413426.html,putational challenges in miRNA target predictions:to be or not to be a true target?J Biomed Biotechnol.2009(803069):1–9.

Bartel DP .2004.MicroRNAs:genomics,biogenesis,mechanism and

function.Cell.116(2):281–297.Evolution of miRNAs and the Diversi?cation of Species GBE

by guest on January 13, 2011

https://www.wendangku.net/doc/513413426.html, Downloaded from

Brudno M,et https://www.wendangku.net/doc/513413426.html,GAN and Multi-LAGAN:ef?cient tools for large-scale multiple alignment of genomic DNA.Genome Res.13(4):721–731.Brunet FG,et al.2006.Gene loss and evolutionary rates following whole-genome duplication in teleost ?shes.Mol Biol Evol.23(9):1808–1816.Bulgakova NA,Knust E.2009.The Crumbs complex:from epithelial-cell polarity to retinal degeneration.J Cell Sci.122(Pt 15):2587–2596.Carleton KL,et al.2008.Visual sensitivities tuned by heterochronic shifts in opsin gene expression.BMC Biol.6:22.Carrington JC,Ambrose V .2003.Role of microRNAs in plant and animal development.Science.301(5631):336–338.Carroll SB.2008.Evo-devo and an expanding evolutionary synthesis:a genetic theory of morphological evolution.Cell.134(1):25–36.Chaudhuri K,Chatterjee R.2007.MicroRNA detection and target prediction:integration of computational and experimental ap-proaches.DNA Cell Biol.26(5):321–337.Chen FC,Li WH.2001.Genomic divergences between humans and homonoids and the effective population size of the common ancestor of humans and chimpanzees.Am J Hum Genet.68(2):

444–456.Chen K,Rajewsky N.2006.Natural selection on human microRNA binding sites inferred from SNP data.Nat Genet.38(12):1452–1456.Clop A,et al.2006.A mutation creating a potential illegitimate microRNA target site in the myostatin gene affects muscularity in sheep.Nat Genet.38(7):813–818.Drake JA,et al.2006.Conserved noncoding sequences are selectively constrained and not mutation cold spots.Nat Genet.38(2):223–227.Eberhart JK,et al.2008.MicroRNA mirn140modulates Pdgf signalling during palatogenesis.Nat Genet.40(3):290–298.Farh KK,et al.2005.The widespread impact of mammalian MicroRNAs on mRNA repression and evolution.Science.310(5755):1817–1821.Fischer A,Wiebe V ,Paabo S,Przeworski M.2004.Evidence for a complex demographic history of chimpanzees.Mol Biol Evol.21(5):799–808.Fraser GJ,Bloomquist RF ,Streelman JT.2008.A periodic pattern generator for dental diversity.BMC Biol.6:32.Friedman RC,Farh KK,Burge CB,Bartel DP .2009.Most mammalian mRNAs are conserved targets of microRNAs.Genome Res.19(1):92–105.Grif?ths-Jones S,Saini HK,van Dongen S,Enright AJ.2008.miRBase:tools for microRNA genomics.Nucleic Acids Res.36:D154–D158(Database issue).Grimson A,et al.2007.MicroRNA targeting speci?city in mammals:determinants beyond seed pairing.Mol Cell.27(1):91–105.Grimson A,et al.2008.Early origins and evolution of microRNAs and Piwi-interacting RNAs in animals.Nature.455(7217):1193–1197.Guryev V ,et al.2006.Genetic variation in the zebra?sh.Genome Res.16(4):491–497.Haygood R,Fedrigo O,Hanson B,Yokoyama KD,Wray GA.2007.Promoter regions of many neural-and nutrition-related genes have experienced positive selection during human evolution.Nat Genet.39(9):1140–1144.Hertel J,et al.2006.The expansion of the metazoan microRNA repertoire.BMC Genomics.7:25.Hiard S,Charlier C,Coppieters W,Georges M,Baurain D.2010.Patrocles:a database of polymorphic miRNA-mediated gene regulation in vertebrates.Nucleic Acids Res.38:D640–D651.International HapMap Consortium.2007.A second generation human haplotype map of over 3.1million SNPs.Nature.449(7164):851–861.Kapsimali M,et al.2007.MicroRNAs show a wide diversity of expression

pro?les in the developing and mature central nervous system.Genome Biol.8(8):R173.King MC,Wilson AC.1975.Evolution at two levels in humans and

chimpanzees.Science.188(4184):107–116.Kitajiri S,et al.2010.Actin-bundling protein TRIOBP forms resilient rootlets of hair cell stereocilia essential for hearing.Cell.141(5):

786–798.Kloosterman WP ,et al.2006.Cloning and expression of new microRNAs

from zebra?sh.Nucleic Acids Res.34(9):2558–2569.Kocher TD.2004.Adaptive evolution and explosive speciation:the

cichlid ?sh model.Nat Rev Genet.5(4):288–298.Korn?eld I,Smith PF .2000.African cichlid ?shes:model systems for

evolutionary biology.Ann Rev Ecol Evol Syst.31:163–196.Kuhn DE,et al.2008.Experimental validation of miRNA targets.Methods.44(1):47–https://www.wendangku.net/doc/513413426.html,rkin MA,et al.2007.ClustalW and ClustalX version 2.Bioinformatics.23(21):2947–2948.

Lee BY,et al.2010.An EST resource for tilapia based on 17normalized libraries and assembly of 116,899sequence tags.BMC Genomics.11:278.

Levine M,Tjian R.2003.Transcription regulation and animal diversity.Nature.424(6945):147–151.

Lewis BP ,Burge CB,Bartel DP .2005.Conserved seed pairing,often ?anked by adenosines,indicates that thousands of human genes are microRNA targets.Cell.120(1):15–20.

Loh YH,et https://www.wendangku.net/doc/513413426.html,parative analysis reveals signatures of differentiation amid genomic polymorphism in Lake Malawi cichlids.Genome Biol.9(7):R113.

Menc?′a A,et al.2009.Mutations in the seed region of human miR-96are responsible for nonsyndromic progressive hearing loss.Nat

Genet.41(5):609–https://www.wendangku.net/doc/513413426.html,ler CT,et al.2007.cis-Regulatory changes in Kit ligand expression

and parallel evolution of pigmentation in sticklebacks and humans.Cell.131(6):1179–1189.Minoda Y ,Sakurai H,Kobayashi T ,Yoshimura A,T akaesu G.2009.An F-

box protein,FBXW5,negatively regulates TAK1MAP3K in the IL-1beta signaling pathway.Biochem Biophys Res Commun.381(3):412–417.

Moran P ,Korn?eld I.1993.Retention of ancestral polymorphism in the Mbuna species ?ock of Lake Malawi.Mol Biol Evol.10:1015–1029.Naguibneva I,et al.2006.The microRNA miR-181targets the

homeobox protein Hox-A11during mammalian myoblast differen-tiation.Nat Cell Biol.8(3):278–284.O’Connell RM,Chaudhuri AA,Rao DS,Baltimore D.2009.Inositol

phosphatase SHIP1is a primary target of miR-155.Proc Natl Acad Sci U S A.106(17):7113–7118.

Plasterk RH.2006.Micro RNAs in animal development.Cell.124(5):877–881.

Rehli M,et al.2005.Transcription factor Tfec contributes to the IL-4-inducible expression of a small group of genes in mouse macro-

phages including the granulocyte colony-stimulating factor receptor.J Immunol.174(11):7711–7722.Rockman MV ,Stern DL.2008.Tinker where the tinkering’s good.Trends Genet.24(7):317–319.

Safdar A,Abadi A,Akhtar M,Hettinga BP ,Tarnopolsky MA.2009.miRNA in the regulation of skeletal muscle adaptation to acute

endurance exercise in C57BI/6J male mice.PLoS One.4(5):e5610.Salzburger W,Mack T,Verheyen E,Meyer A.2005.Out of Tanganyika:genesis,explosive speciation,key-innovations and phylogeography

of the haplochromine cichlid ?shes.BMC Evol Biol.5(1):17.Loh et al.GBE

by guest on January 13, 2011

https://www.wendangku.net/doc/513413426.html, Downloaded from

Sarasin-Filipowicz M,Krol J,Markiewicz I,Heim MH,Filipowicz W.2009.Decreased levels of microRNA miR-122in individuals with hepatitis C responding poorly to interferon therapy.Nat Med.15(1):31–33.Saunders MA,Liang H,Li WH.2007.Human polymorphism at microRNAs and microRNA target sites.Proc Natl Acad Sci U S A.104(9):3300–3305.Sethupathy P ,Collins FS.2008.MicroRNA target site polymorphisms and human disease.Trends Genet.24(10):489–497.Sethupathy P ,Giang H,Plotkin JB,Hannenhalli S.2008.Genome-wide analysis of natural selection on human cis-elements.PLoS One.3(9):e3137.Shahin H,et al.2006.Mutations in a novel isoform of TRIOBP that encodes a ?lamentous-actin binding protein are responsible for DFNB28re-cessive nonsyndromic hearing loss.Am J Hum Genet.78(1):144–152.Simo ?es JM,Duarte IS,Fonseca PJ,Turner GF ,Amorim MC.2008.Courtship and agonistic sounds by the cichlid ?sh Pseudotropheus zebra.J Acoust Soc Am.124(2):1332–1338.Soares AR,et al.2009.Parallel DNA pyrosequencing unveils new zebra?sh microRNAs.BMC Genomics.10:195.Soukup GA.2009.Little but loud:small RNAs have a resounding affect on ear development.Brain Res.1277:104–114.Stark A,Brennecke J,Bushati N,Russell RB,Cohen SM.2005.Animal microRNA confer robustness to gene expression and have a signif-icant impact on 3#UTR evolution.Cell.123(6):1133–1146.Streelman JT,Albertson RC,Kocher TD.2003.Genome mapping of the orange blotch colour pattern in cichlid ?shes.Mol Ecol.12(9):2465–2471.Streelman JT,Peichel CL,Parichy DM.2007.Developmental genetics of adaptation in ?shes:the case for novelty.Ann Rev Ecol Evol Syst.38:655–681.Sucena E,Delon I,Jones I,Payre F ,Stern DL.2003.Regulatory evolution

of shavenbaby/ovo underlies multiple cases of morphological parallelism.Nature.424(6951):935–938.Sylvester JB,et al.2010.Brain diversity evolves via differences in

patterning.Proc Natl Acad Sci U S A.107(21):9718–9723.Verzijden MN,et al.2010.Sounds of male Lake Victoria cichlids vary within and between species and affect female mate preferences.

Behav Ecol.21(3):548–555.Wheeler BM,et al.2009.The deep evolution of metazoan microRNAs.

Evol Dev.11(1):50–68.Wienholds E,et al.2005.MicroRNA expression in zebra?sh embryonic development.Science.309(5732):310–311.

Won YJ,Sivasundar A,Wang Y,Hey J.2005.On the origin of Lake Malawi cichlid species:a population genetic analysis of divergence.Proc Natl Acad Sci U S A.102(1Suppl):6581–6586.

Won YJ,Wang Y,Sivasundar A,Raincrow J,Hey J.2006.Nuclear gene variation and molecular dating of the cichlid species ?ock of Lake Malawi.Mol Biol Evol.23(4):828–837.

Wray GA.2007.The evolutionary signi?cance of cis-regulatory mutations.Nat Rev Genet.8(3):206–216.

Xiao C,Rajewsky K.2009.MicroRNA control in the immune system:basic principles.Cell.136(1):26–36.

Yekta S,Tabin CJ,Bartel DP .2008.MicroRNAs in the Hox network:an

apparent link to posterior prevalence.Nat Rev Genet.9:789–796.Zhang Z,Lan Y,Chai Y,Jiang R.2009.Antagonistic actions of Msx1and Osr2pattern mammalian teeth into a single row.Science.

323(5918):1232–1234.Zuker M.2003.Mfold web server for nucleic acid folding and hybridization prediction.Nucleic Acids Res.31(13):3406–3415.

Associate editor :Wen-Hsiung Li Evolution of miRNAs and the Diversi?cation of Species GBE

by guest on January 13, https://www.wendangku.net/doc/513413426.html, Downloaded from