PHEN phen˙12016 B Dispatch: April 12, 2013 Journal: PHEN CE:

Journal Name Manuscript No. Author Received: No of pages: 9 TS: Suresh

Physiological Entomology (2013) 0, 0–0 DOI: 10.1111/phen.12016

Transcriptome sequencing as a platform to elucidatemolecular components of the diapause response inthe Asian tiger mosquito Aedes albopictus

M O N I C A F . P O E L C H A U1, J U L I E A . R E Y N O L D S2, D A V I D L .D E N L I N G E R 2, C H R I S T I N E G . E L S I K 3 and P E T E R A .A R M B R U S T E R 1

1Department of Biology, Georgetown University, Washington, District of Columbia, U.S.A., 2Department of Entomology, Ohio

State University, Columbus, Ohio, U.S.A. and 3Divisons of Animal and Plant Sciences, S134-D Animal Sciences Research

Center, University of Missouri, Columbia, Missouri, U.S.A.

Abstract. Diapause has long been recognized as a crucial ecological adaptationto spatio-temporal environmental variation. More recently, rapid evolution of thediapause response has been implicated in response to contemporary global warmingand during the range expansion of invasive species. Although the molecular regulationof diapause remains largely unresolved, rapidly emerging next-generation sequencing(NGS) technologies provide exciting opportunities for addressing this longstandingquestion. In the present study, a new assembly from life-history stages relevant todiapause in the Asian tiger mosquito Aedes albopictus (Skuse) is reported, alongwith unique methods for the analysis of NGS data and transcriptome assembly. Adigital normalization procedure that significantly reduces the computational resourcesrequired for transcriptome assembly is evaluated. Additionally, a method for proteinreference-based and genomic reference-based merged assembly of 454 and Illuminareads is described. Finally, a gene ontology analysis is presented, which createsa platform for identifying the physiological processes associated with diapause.Taken together, these methods provide valuable tools for analyzing the transcriptionalunderpinnings of many complex phenotypes, including diapause, and provide a basisfor determining the molecular regulation of diapause in Ae. albopictus .

Key words. Aedes albopictus , diapause, next-generation sequencing, transcriptome.

Introduction

The annual rotation of the earth around the sun gives rise tovarious forms of recurring seasonal environmental variationthat affects most forms of life on Earth. Conspicuous examplesinclude the annual arrival of winter in temperate habitats, thewet and dry seasons in tropical habitats, and biotic interactionsthat can severely impact opportunities for the survival andreproduction of a wide range of organisms. Photoperiodicdiapause is a crucial ecological adaptation that allows many

Correspondence: Peter A. Armbruster, Department of Biology,Georgetown University, 37th and O Streets NW, Washington,District of Columbia, U.S.A. Tel.: +1 202 687 2567; e-mail:AQ1paa9@georgetown.edu

insects to cope with recurring seasonal environmental variationby using day length (photoperiod) as a token cue to initiatephysiological changes that prepare the insect for persistenceunder adverse conditions. In temperate habitats, the seasonaltiming of diapause initiation exhibits remarkably consistentgeographical trends across latitudinal gradients (Andrewartha,1952; Danks, 1987). The rapid evolution of diapause timinghas been implicated in response to contemporary globalwarming (Bradshaw & Holzapfel, 2001), and also duringrange expansion of invasive species (Bean et al ., 2012;Urbanski et al ., 2012). Thus, studies of diapause address bothbasic questions concerning organismal adaptation to spatio-temporal environmental heterogeneity, as well as more appliedquestions, such as anticipating biological responses to rapid

123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758

123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758

© 2013 The Royal Entomological Society 1

2 M. F. Poelchau et al.

ongoing global warming and the evolution of invasive speciesduring range expansion.

Despite the well-established ecological significance ofphotoperiodic diapause, the molecular regulation of thiscrucial adaptation remains largely unresolved. In part, thisis because diapause is a physiologically complex processinvolving multiple regulatory hierarchies across the trajectoryfrom diapause preparation to diapause maintenance and,eventually, termination (Kostal, 2006). A number of commonphysiological themes are associated with diapause in adiverse range of insects, including reduced metabolism anddevelopmental arrest (Tauber et al ., 1986), increased coldand desiccation resistance (Rinehart et al ., 2007; Benoit,2010), and increased nutrient storage (Hahn & Denlinger,2011). However, our current understanding of the molecularregulation of these physiological processes during diapause ismostly based on studies of one or several genes in a diverserange of insect taxa. The extent to which common molecularregulatory mechanisms underlie the consistent physiologicalthemes of diapause remains unclear. Thus, elucidation ofthe molecular regulation of diapause across the trajectoryfrom preparation to termination remains an elusive goal, withimportant implications for a wide range of both basic andapplied questions in insect science.

Rapid advances in next-generation sequencing (NGS) tech-nology are opening up exciting opportunities for elucidatingthe molecular basis of complex phenotypes, even in non-modelorganisms for which complete genome sequences are unavail-able (Davey et al ., 2011; Martin & Wang, 2011). This is anespecially exciting opportunity for diapause research becausemany insect species in which diapause has been studied thor-oughly from an ecological, physiological or evolutionary per-spective do not currently have a complete genome sequenceavailable (e.g. Wyeomyia smithii , Sarcophaga crassipalpis ,Culex pipiens , Aedes albopictus). Nevertheless, many chal-lenges remain in applying NGS to elucidate the molecularbasis of diapause. The amount of data from a single NGSexperiment can be overwhelming, requiring substantial com-putational resources for analysis. Additionally, working witha non-model organism for which a genome sequence is notavailable requires creative approaches for the assembly, anno-tation and estimation of differential gene expression by readmapping.

The present study describes methods for the developmentof transcriptomic resources from multiple NGS diapauseexperiments in the Asian tiger mosquito Aedes albopictus(Skuse). A digital normalization procedure that significantlyreduces the computational resources required for the assemblyof hundreds of millions of Illumina reads is evaluated.Additionally, a method is described for protein reference-based and genomic reference-based merged assembly of 454and Illumina reads. The resulting merged assembly leads toincreased gene discovery and annotation. Finally, the resultsof a gene ontology (GO) analysis are reported that establish abasis for identifying physiological processes associated withdiapause. The results of the GO analysis are discussed inrelation to previous results on the transcriptional basis ofdiapause in Ae. albopictus . These methods illustrate a valuable

set of analytical tools that provide a basis for determining themolecular regulation of diapause in Ae. albopictus , as wellas the transcriptional underpinnings of many other complexphenotypes.

Materials and methods

Study organism

Aedes albopictus females typically oviposit desiccation-resistant eggs above the water line in a variety of natural(e.g. tree holes) or artificial (e.g. tyres) containers. Onceembryological development is complete, a nondiapause pharatelarva inside the chorion of the egg persists in a quiescent stateand typically hatches within 1 h of submersion in water withreduced oxygen content. By contrast, a diapause pharate larvais refractory to hatching stimuli for up to several months. Intemperate populations, exposure of the maternal pupa and adultto short day lengths induces diapause (Wang, 1966; Mori et al .,1981). RNA was sequenced from pharate larvae (eggs) at threetime points after oviposition to assemble a trascriptome ofgene expression across the trajectory from the early to latedevelopmental arrest phases of diapause and quiescence.

Insect rearing and RNA generation

A laboratory F13 Ae. albopictus strain collected fromManassas, Vancouver, was reared under an LD 16 : 8 hphotocycle at 21 ◦C and approximately 80% relative humidity,as described previously (Armbruster & Hutchinson, 2002;Armbruster & Conn, 2006). Upon pupation, mosquitoes weredivided into eight replicate cohorts of approximately 100mosquitoes per cohort, each in a 9.5-L cage. Four cohorts(biological replicates) were maintained under a diapause-inducing photoperiod (D) (LD 8 : 16 h) and four cohortswere maintained under a nondiapause-inducing photoperiod(ND) (LD 16 : 8 h). Females were allowed to blood-feed ona human host 9–16 days after eclosion, and again 7 and14 days later, so that eggs could be collected over multiplegonotrophic cycles. Three days after the first blood feeding,a small brown jar half-filled with approximately 50 mL ofdeionized water and lined with unbleached seed germinationpaper was placed into each cage 6–7 h after lights on tostimulate oviposition. The seed germination paper containingeggs was removed after 24 h and this procedure was repeatedfor 26 consecutive days. Egg papers were gently air-dried72 h after collection, and then stored at approximately 80%relative humidity under an LD 8 : 16 h photocycle until furthertreatment; short-day photoperiods experienced at the egg stagedo not result in diapause in Ae. albopictus (Mori et al .,1981). A subset of eggs from each replicate cohort wasreserved for diapause incidence and measurements (see below).The remainder was divided and allowed to develop for 11,21 and 40 days post-oviposition. Upon completion of thespecified development period, individual egg samples werecarefully brushed into 2-mL microcentrifuge tubes, snap-frozen

123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758

123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758

© 2013 The Royal Entomological Society, Physiological Entomology, 0, 0–0

Diapause transcriptome sequencing 3

in liquid nitrogen and stored at −80 ◦C. After collectionof all samples, eggs were pooled according to treatment(D and ND) and development period (11, 21 and 41 dayspost-ovoposition), ground in TRI Reagent (Sigma Aldrich,St Louis, Missouri) and RNA was extracted in accordancewith the manufacturer’s instructions. DNA was removed fromeach sample with Turbo-DNA free (Ambion, Austin, Texas).RNA quality was assessed on an RNA chip (Bioanalyzer2100; Agilent Technologies, Santa Clara, California). Threeof the four biological replicates from each treatment anddevelopment period were chosen for sequencing based onRNA quality and quantity, with the exception of 40 days post-oviposition pharate larvae reared under ND conditions, forwhich only two biological replicates showed sufficient RNAquality for high-throughput sequencing. This resulted in 17sequenced RNA libraries (see below). Incubator malfunctioncaused some 40-day eggs to experience an approximately4 ◦C fluctuation on three consecutive days. Eggs scheduled forsnap-freezing on these days were discarded so that all eggshad at least 24 h to recover. Because ND and D eggs werestored together, temperature fluctuations should not result insystematic differences in gene expression between ND and Dtreatments.

For diapause incidence measurements, eggs ranging from2 to 4 weeks of age were hatched, the number of hatchedlarvae recorded, and the egg papers re-dried. This procedurewas repeated after 7 and 14 days. After the last hatch,eggs were counted and bleached as described by Trpis(1970) to record the number of embryonated but unhatched(= diapause) eggs. Diapause incidence was calculated as:number of embryonated unhatched eggs/(number of hatchedeggs + number of embryonated unhatched eggs) (Urbanskiet al ., 2012). Percentage embryonation was calculated as:(number of embryonated unhatched eggs + number of hatchedeggs)/total number of eggs. Counts were pooled acrosscollection dates within each replicate for the final calculations.

Sequencing

Illumina paired-end mRNA-Seq library construction andsequencing was performed by The University of MarylandGenomics Institute in accordance with the TruSeq RNAsample preparation guide (version 2) (Illumina Inc., SanDiego, California). The 17 libraries were bar-coded (Wonget al ., 2013) in accordance with the manufacturer’s instructionsand sequenced on three flow-cell lanes on an IlluminaHiSeq 2000 high-throughput sequencer, where a proportionof each library was sequenced on each lane (average fragmentsize = 320 bp; average read length = 101 bp). Raw reads areavailable in NCBI’s short read archive under BioProjectaccession PRJNA187045.

Assembly and annotation

Read cleaning. Reads with matches to the NCBI UniVecCore database, which contains common vector, adapter, linker

and primer contaminants (ftp://ftp.ncbi.nih.gov/pub/UniVec/;accessed August 10 2010), as well as the Ae. albopictus rRNAsequence (GenBank # L22060.1), and Illumina multiplexed,paired-end mRNA-Seq adapters were identified using sshaha2(Ning et al ., 2001) and removed along with their read pair.Match criteria for removal were set at 95% identity (rRNA,UniVec) or 90% (adapters) and a Smith-Waterman scorelarger than 18 (UniVec). The first 15 bp of each read wastrimmed because Illumina libraries frequently display low k-mer diversity in the first 12–15 bp of each read, suggestingnonrandom priming during the Illumina mRNA-Seq librarypreparation (Poelchau et al ., 2013). The remaining reads werecleaned using SolexaQA, version 1.10 (Cox et al ., 2010) bytrimming regions with a masked contaminant sequence or aphred score equivalent of less than 20 and removing readsshorter than 25 bp.

Digital normalization and reference-free assembly. A novel‘digital normalization’ technique (Brown et al ., 2012) wasused to reduce the computational resources required for contigassembly. The great sequencing depth and large sequencingerror rate of many NGS experiments results in large anderror-prone datasets that are difficult and computationallyexpensive to assemble. To circumvent this problem, the ‘digitalnormalization’ method identifies all k-mers (short DNA-sequences) of a particular length in a dataset, and eliminatesall reads that contain a k-mer over a given abundance. Thismethod effectively removes the majority of erroneous k-mers,at the same time as keeping almost all real k-mers in thedataset, thereby drastically reducing the number of redundantreads and the computational resources required for assembly(Brown et al ., 2012). One round of digital normalizationwas performed on the cleaned reads using parameters almostidentical to Brown et al . (2012) (using k-mer size 20 and acoverage cut-off of 20; the ‘x’ parameter, which determinesmemory usage, was set to 1e10) (software available at:https://github.com/ged-lab/khmer/tree/2012-paper-diginorm).

The reduced read set was assembled into contigs withvelvet (Zerbino & Birney, 2008), using k-mers from 19to 59 in 10 k-mer intervals and a coverage cut-off of 10.The resulting contigs were then further assembled by oases,which clusters contigs generated by velvet into loci, anduses the paired-end information to generate transcript isoforms(Schulz et al ., 2012). Assemblies from separate k-mers weremerged using cd-hit-est (Li & Godzik, 2006), which groupedsequences with 99% identity over the entire sequence length.

To evaluate the efficacy of the digital normalizationtechnique for this dataset, a comparison of test assembliesusing a normalized and non-normalized subset of the datawas performed. Specifically, assemblies were constructed asoutlined above for the D treatment at 11 days post-ovipositionusing either a digitally normalized or complete set of reads.The first measure of assembly quality was the number ofcontigs, where a smaller number generally indicates lowercontig redundancy, and thus an increased assembly quality.Second, assembly completeness was evaluated by mappingthe non-normalized read set back to each assembly using

123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758

123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758

© 2013 The Royal Entomological Society, Physiological Entomology, 0, 0–0

4 M. F. Poelchau et al.

bowtie, version 0.12.7) (Langmead et al ., 2009) based onthe rationale that more of the original reads would map backto the more complete assembly. Third, contig length wascompared between the two assemblies, where longer contigslikely represent more completely reconstructed transcripts.

Previous data. Previously, transcriptomes were generatedfrom Ae. albopictus oocyte (Poelchau et al ., 2011) and embry-onic (Poelchau et al ., 2013) tissue. To generate a compre-hensive Ae. albopictus transcriptome assembly, the presentassembly was merged with the previous assemblies using areference-based approach, similar to Poelchau et al . (2013) andSurget-Groba & Montoya-Burgos (2010). Assembled contigsand unassembled reads from 454 sequencing of oocyte tissuewere derived from Poelchau et al . (2011). Contigs from Illu-mina reads generated from embryonic tissue (Poelchau et al .,2013) and assembled in velvet (Zerbino & Birney, 2008)and oases (Schulz et al ., 2012) followed by cd-hit-est (Li &Godzik, 2006) were also added. Unassembled Illumina readsfrom the embryo tissue were not included because the shortreads of Illumina sequencing limits the utility of these data forthe current merged assembly.

Protein reference-based assembly

A nonredundant dipteran protein set was generated fromAe. aegypti , Culex quinquefasciatus , Anopheles gambiae andDrosophila melanogaster protein sequences, downloaded asorthologue groups from OrthoDB, version 4 (Waterhouse et al .,2011). For each orthologue group, a single protein sequencefrom the taxonomically closest organism to Ae. albopictuswas retained (Ae. aegypti , then Cx. quinquefasciatus , thenAn. gambiae, then D. melanogaster). The final referenceset, which comprised 21 066 proteins, maximized proteinsequence diversity and eliminated redundancy, thereby leadingto increased computational efficiency.

Unassembled reads from the oocytes and all contigs werealigned to the dipteran protein set using fastx (Pearson et al .,2007). The alignment with the lowest e-value ≤ 1e−6 wasretained for further analysis. Contigs assigned to the samereference protein and at least 95% identity of the overlappingsequence were merged in cap3 (Huang & Madan, 1999). Theannotations of the re-assembled contigs were verified by againaligning them with the reference protein set in fastx; onlyalignments with > 70% identity to the best matching referenceprotein were used in the final, annotated assembly. Chimericcontigs were identified by searching for additional alignmentsoutside of the primary alignment that were within 80% of theprimary alignment’s percentage identity. Contigs that met thesecriteria were considered chimeric and discarded.

Genomic reference-based assembly

Contigs and oocyte reads that did not align to thedipteran protein set described above were matched to

Ae. aegypti genomic supercontigs (Nene et al ., 2007)using blastn (e-value < 1e−6), and then aligned inexonerate (Slater & Birney, 2005) (parameters wereest2genome – softmaskquery – bestn 1 – dnahspdropoff 0). Atleast 72.8% identity of alignments was required for furtheranalysis. Contigs were merged in cap3 if they overlappedand re-aligned, as above, retaining only contigs with > 70%identity in their re-alignments. If contigs spanned multiplegene models, they were considered chimeric and discarded.Contigs that matched within 1 kb up- or down-stream ofannotated genes over > 90% of their length were annotatedas potential untranslated regions (UTRs) for those genes. Thefinal, annotated assembly included contigs that matched bothannotated and un-annotated regions of the genome.

Functional characterization of the transcriptome

GO and GO-Slim (Ashburner et al ., 2000) categories weredownloaded for each reference gene model from BioMart(Haider et al ., 2009). The numbers and composition of eachgroup were then tallied. GO-Slim categories were investigatedin addition to GO categories to organize genes into broaderfunctional groups whose biological significance in the contextof diapause can be more easily summarized.

Results

Diapause incidence

Diapause incidence ranged from 87.5% to 100% in biologi-cal replicates of the diapause-inducing photoperiod treatments(D). Percentage embryonation ranged from 82.9% to 98.9%across all replicates, with an average of 90.7%.

Assembly and annotation

The final assembly, as well as a spreadsheet withannotations for each gene model, is available at:http://www.albopictusexpression.org/?q=data.

Read cleaning and digital normalization of sequences fromdiapause and nondiapause eggs

Illumina sequencing of 17 RNA libraries from D and NDeggs (pharate larvae) produced 648 339 954 reads (324 169 167pairs), of which 602 178 150 (92.9%) remained after qualitycontrol. Digital normalization of the quality-filtered readsreduced the total number of reads to 35 833 461 (5.5%of the original reads; Fig. 1), substantially reducing thecomputational resources necessary to complete the assembly(memory usage scales roughly linearly with the numberof reads used: http://listserver.ebi.ac.uk/pipermail/velvet-users/2009-June/000359.html). The digitally normalizedassembly was performed in much less time and required less

123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758

123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758

© 2013 The Royal Entomological Society, Physiological Entomology, 0, 0–0

Diapause transcriptome sequencing 5

Fig. 1. Steps of the assembly and the number of reads, contigs and gene models resulting from each step of Illumina sequencing from RNAlibraries of the Asian tiger mosquito Aedes albopictus .

memory (digitally-normalized assembly, approximately 4.5 h,11 GB maximum memory usage; non-normalized assembly,approximately 43 h, 62 GB maximum memory usage).

Digitally normalized and non-normalized test assemblieson sequences from the day 11 D sample showed thatthe normalized assembly out-performed the non-normalizedassembly in most aspects (Table 1). In the digitally normalizedassembly, 19% fewer contigs were generated and 10%more reads mapped back to the assembly. However, thenormalized assembly had a 9% shorter average contig length.Because the substantially reduced computational requirements,smaller contig number and greater number of mapped readsindicated more advantages than disadvantages to the digitalnormalization procedure, the digitally normalized full datasetwas used in the analysis.

Assembly metrics: reference-free assembly from diapause andnondiapause eggs

velvet assembly (Zerbino & Birney, 2008) of sequencesfrom D and ND eggs (pharate larvae) followed by oases(Schulz et al ., 2012) yielded 311 071 contigs, which werereduced to 176 502 nonredundant contigs using cd-hit-est(Table 2). Combining this assembly with the oocyte (Poelchauet al ., 2011) and embryo (Poelchau et al ., 2013) reference-freeassemblies resulted in 627 154 contigs (Fig. 1).

Assembly metrics: protein-reference assembly

Of all reference-free contigs described above, 43.3% alignedto the dipteran protein set, generating 61 624 quality-filtered,

re-assembled contigs after cap3 re-assembly (Fig. 1 andTable 2). This substantially reduced the number of redundantcontigs annotated as a given gene model. Four thousand twohundred and fifty-four re-assembled contigs were consideredchimeric and discarded. Twelve thousand one hundred andthirty-nine gene models were identified in the protein-basedassembly (Fig. 1).

Assembly metrics: genomic-reference assembly

In total, 24.7% of reference-free contigs were used in thegenomic cap3 re-assembly, resulting in 28 079 quality-filtered,re-assembled contigs that matched to annotated regions ofthe Ae. aegypti genome (25 603) or potential UTRs (2476).An additional 28 460 re-assembled contigs matched to un-annotated regions of the genome. Of the re-assembled contigs,2311 were chimeric and therefore discarded. Re-assembledcontigs in the genome-based assembly recovered 7629 genemodels, of which 1122 were not found in the protein-basedassembly.

Assembly metrics: unaligned contigs

There were 239 091 remaining reference-free contigs thatdid not meet any of the alignment criteria, and thereforewere not re-assembled or included in the annotated assembly(Fig. 1). An inspection of these remaining sequences revealedthat 9777 aligned to the UniProt database (blastx, e-value< 1e−3), and 57 327 sequences derived from the pharate larvalassembly aligned to un-annotated sequences from previousassemblies (Poelchau et al ., 2011, 2013; blastn, e-value< 1e−6; data not shown). These sequences could thus include

123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758

123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758

© 2013 The Royal Entomological Society, Physiological Entomology, 0, 0–0

6 M. F. Poelchau et al.

Table 1. Quality metrics for digitally-normalized and non-normalized test assemblies of next generation sequencing reads from the Asian tiger mosquitoAedes albopictus using the library from diapause (D) conditions at 11 days post-oviposition.

Numberof reads

Number ofcontigs

Minimum contiglength

Average contiglength

Median contiglength

Maximum readsmapped to

Mean %paired assembly

Without normalization 111 574 867 93 622 94 1042.3 619 14 481 68.6Digital normalization 17 351 298 76 018 96 947.2 553 14 480 75.3

Table 2. Quality metrics for consecutive stages of assembly of next generation sequencing data from the Asian tiger mosquito Aedes albopictus .

Numbercontigs length

Mean contiglength

Mediancontig length

Maximumcontig length

Average% GC

Pharate larval assembly, oases 311 071 975.9 518 23 266 44.10Pharate larval assembly, merged by cd-hit-est 176 502 896.8 490 23 266 43.50All contigs in merged assembly 627 154 885.8 502 23 934 45.27Protein-reference assembly 61 624 1961.0 1388 25 247 49.07Genomic-reference assembly 28 079 840.2 513 20 598 44.46Complete annotated assembly 118 163 1426.2 806 25 247 46.53

rapidly diverged sequence specific to Ae. albopictus , andviral or bacterial sequences. The remaining sequences willbe aligned to the Ae. albopictus genome sequence, once itis completed, for further annotation.

Both protein and genomic reference assemblies had highpercentage identity to gene models, similar to the previous Ae.albopictus transcriptome assembly containing only reads fromoocyte and embryo stages (median percentage identity, protein:91.5%, genomic: 83.5%; Fig. 2). A substantial percentage ofthe annotated contigs’ length was contained in the sequencealignment (median percentage of contig in the alignment,protein: 69.1%, genomic: 83.1%; Fig. 2). Similarly, theproportion of reference sequence covered by contigs in proteinalignments was high (median reference coverage: 85.3%;Fig. 2). As a result of the nature of sequence alignmentsusing the genomic reference, genomic reference coveragewas correspondingly low (median reference coverage: 12.4%;Fig. 2), which suggests that the majority of many genomicalignments occurred outside of (or adjacent to) the annotatedsequence on the Ae. aegypti genome. It is also likely that manyun-annotated UTR regions of the Ae. aegypti genome werecontained in the alignments, or it may be possible that Ae.albopictus genes have different gene boundaries than their Ae.aegypti homologues.

The number of gene models identified increased from11 505 in the oocyte assembly (Poelchau et al ., 2011) to12 345 in the merged embryo and oocyte assembly (Poelchauet al ., 2013), and to 13 261 in the final merged assemblyfrom all three experiments reported in the present study.This result affirms that new life-history stages included inthe assembly increase the assembly’s gene content, likely asa result of the addition of genes with unique, stage-specificexpression. A total of 17 356 gene models have been annotatedin the Ae. aegypti genome sequence (Assembly AaegL1.3,Vector Base: https://www.vectorbase.org/organisms/aedes-aegypti/liverpool-lvp/AaegL1.3, accessed 30 December 2012),suggesting that a large proportion of the genes from the

Ae. albopictus genome are represented in the current fullassembly reported in the present study.

Functional characterization

The new assembly added 125 new GO categories andnine new GO-Slim categories relative to the previousassembly (a full list of GO categories is available at:http://www.albopictusexpression.org/?q=data). The new GO-Slim categories added a diversity of functions that likely con-tribute to physiological and developmental processes during thepharate larval stage in Ae. albopictus . Several of these new cat-egories were related to cellular structure and growth (cilium,cell proliferation, cell junction organization, pigmentation),translation (structural constituent of ribosome) and metabolism(protein transporter activity, enzyme regulator activity, GTPaseactivity, nitrogen cycle metabolic process). Many genes in thefull assembly had functions with known relevance to diapause,including 231 genes related to lipid metabolism and 165 relatedto stress response (Table 3). This demonstrates that the assem-bly represents a rich resource for potential diapause-relatedgenes.

Discussion

Temperate populations of Ae. albopictus survive throughthe winter by entering diapause as a pharate larva insidethe chorion of the egg (Wang, 1966; Mori et al ., 1981).Diapause eggs are more cold resistant (Hawley et al ., 1987),desiccation resistant (Sota & Mogi, 1992; Urbanski et al .,2010), larger, and harbour greater lipid reserves (Reynoldset al ., 2012) relative to nondiapause eggs. In addition toenhancing overwinter survival, all of these properties are likelyto contribute to the ability of Ae. albopictus eggs to survivelong distance transport, and therefore may be important factors

123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758

123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758

© 2013 The Royal Entomological Society, Physiological Entomology, 0, 0–0

Diapause transcriptome sequencing 7

Fig. 2. Percentage identity (a), proportion of contig in the alignment (b) and proportion of reference in the alignment (c) of re-assembled contigsof the Asian tiger mosquito Aedes albopictus resulting from alignments to the protein sequence and genomic references.

contributing to the worldwide spread of this invasive mosquito.Diapause timing has evolved rapidly among populations acrossthe invasive U.S.A. range of Ae. albopictus within a 20-yearperiod (Urbanski et al ., 2012), further emphasizing the centralimportance of diapause with respect to the invasion and rangeexpansion of this mosquito.

Rapidly advancing NGS technologies and analysis toolsprovide exciting opportunities for determining the molecu-lar regulation of complex life-history adaptations such asdiapause. For example, new techniques are emerging thatfacilitate transcriptome assembly, which can require pro-hibitively large computational resources (Brown et al ., 2012).The digitally normalized transcriptome assembly reportedin the present study demonstrates that such techniques canbe used to successfully assemble large Illumina datasetswith significant advantages relative to non-normalized data(Table 1). Additionally, previous assemblies of the Ae. albopic-tus transcriptome include sequence data from prediapause andnondiapause oocytes (Poelchau et al ., 2011) and embryos(Poelchau et al ., 2013), encompassing only some of thelife-history stages that a diapause-destined mosquito expe-riences. In the present assembly, transcriptome data frommultiple pharate larval stages are added to generate a com-prehensive transcriptome encompassing all diapause stagesand the nondiapause counterparts. As anticipated, sequencedata from pharate larval stages in the present, comprehen-sive, assembly increase both the number of gene models inthe assembly and the number of functional groups based onGO assignments.

The reference-based assembly technique reported in thepresent study reduces contig redundancy and increases contiglength relative to reference-free assemblies (Table 2), at thesame time as creating contigs with high-confidence annotationsto reference gene sets from other organisms, and identifyingpotential novel UTR sites. Similar to Poelchau et al . (2013),

the protein-based reference assembly has longer contigs thanthe genome-based reference assembly (Table 2). However, thegenome-based reference assembly adds 1122 additional genemodels to the assembly, as well as potential UTRs, whichwill be of future utility for identifying regulatory regions.Therefore, this ‘hybrid’ transcriptome assembly method canbe a useful strategy for other taxa that are closely related toan organism with an available genome sequence.

Previous studies identify several physiological themes thatappear to be shared across the diapause response of multipleinsect species. Diapausing insects demonstrate up-regulationof stress-response genes (Denlinger et al ., 2005; Rinehartet al ., 2007), increased lipid synthesis and storage (Robich &Denlinger, 2005; Reynolds & Hand, 2009; Sim & Denlinger,2009a), changes in insulin signalling (Tatar et al ., 2001;Williams et al ., 2006; Sim & Denlinger, 2008, 2009b; Hahn& Denlinger, 2011), shifts in metabolism (Kukal et al .,1991; Michaud & Denlinger, 2007; Ragland et al ., 2010;Hahn & Denlinger, 2011) and changes in patterns of cellcycle arrest (Tammariello & Denlinger, 1998; Tammariello,2001; Reynolds & Hand, 2009). Analysis of diapause-relatedexpression changes of genes relevant to these physiologicalprocesses can illuminate the specific mechanisms involvedin diapause-driven physiological change (Sim & Denlinger,2008, 2009a; Ragland et al ., 2010, 2011; Reynolds et al .,2012). In the present assembly, many genes in potentiallydiapause-relevant pathways, such as lipid metabolism andstress response, are identified (Table 3). One approach forusing this resource to investigate the transcriptional regulationof diapause-associated physiological processes is to identifycandidate genes that can then be screened by quantitativereverse transcriptase-polymerase chain reaction for differentialexpression (Reynolds et al ., 2012) or functionally evaluatedusing RNA interference knockdown experiments (Sim &Denlinger, 2008, 2009b). Additionally, the original reads

123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758

123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758

© 2013 The Royal Entomological Society, Physiological Entomology, 0, 0–0

8 M. F. Poelchau et al.

Table 3. Gene ontology categories (GO-Slim) of all gene modelsrepresented in the full assembly of illumina sequencing reads from theAsian tiger mosquito Aedes albopictus .

GO-Slim GOAaccession number Description

Number ofgene modelsa

GO:0008150 Biological_process 5768GO:0005575 Cellular_component 4299GO:0005623 Cell 3995GO:0005622 Intracellular 2727GO:0043226 Organelle 1493GO:0034641 Cellular nitrogen compound

metabolic process1063

GO:0006810 Transport 993GO:0005634 Nucleus 892GO:0009058 Biosynthetic process 878GO:0005737 Cytoplasm 795GO:0044281 Small molecule metabolic

process699

GO:0007165 Signal transduction 624GO:0006464 Cellular protein modification

process479

GO:0043234 Protein complex 433GO:0009056 Catabolic process 421GO:0055085 Transmembrane transport 401GO:0005975 Carbohydrate metabolic

process354

GO:0005576 Extracellular region 282GO:0006412 Translation 276GO:0006629 Lipid metabolic process 231GO:0006259 DNA metabolic process 197GO:0034655 Nucleobase-containing

compound catabolic process193

GO:0005840 Ribosome 175GO:0006950 Response to stress 165GO:0005856 Cytoskeleton 152NA Other categories, combined 2215NA No ontology information 6248

a The combined number of gene models in each category exceeds thetotal number of gene models in the assembly because gene models canbe assigned to multiple GO-Slim categories.The number of gene models assigned to the top 25 categories are listed;the remainder are combined. NA, not available.

from the transcriptome sequencing can be mapped backto the transcriptome assembly to quantify gene expression(Poelchau et al ., 2011, 2013). Thus, the Ae. albopictusdiapause transcriptome reported in the present study representsa comprehensive foundation for elucidating the molecular basisof key traits underpinning geographical adaptation and invasionsuccess in this invasive mosquito.

Acknowledgements

We thank Shin Goto, Jim Hardie and Kyeong-Yeoll Lee forthe opportunity to contribute to this special issue and toparticipate in the symposium on New Complexities in theRegulation of Insect Diapause and Cold Hardiness at the 2012International Congress of Entomology in Daegu, South Korea.

This work was supported by the National Institutes of Healthgrant 5R21AI081041-02 and Georgetown University.

References

Andrewartha, H.G. (1952) Diapause in relation to the ecology ofinsects. Biological Reviews , 27, 50–107.

Armbruster, P.A. & Conn, J.E. (2006) Geographic variation of larvalgrowth in North American Aedes albopictus (Diptera: Culicidae).Annals of the Entomological Society of America , 99, 1234–1243.

Armbruster, P.A. & Hutchinson, R.A. (2002) Pupal mass and winglength as indicators of fecundity in Aedes albopictus and Aedesgeniculatus (Diptera: Culicidae). Journal of Medical Entomology ,39, 699–704.

Ashburner, M., Ball, C.A., Blake, J.A. et al . (2000) Gene ontology:tool for the unification of biology. Nature Genetics , 25, 25–29.

Bean, D.W., Dalin, P. & Dudley, T.L. (2012) Evolution of critical daylength for diapause induction enables range expansion of Diorhabdacarinulata , a biological control agent against tamarisk (Tamarixspp.). Evolutionary Applications , 5, 511–523.

Benoit, J.B. (2010) Water management by dormant insects: compar-isons between dehydration resistance during summer aestivation andwinter diapause. Aesivation: Molecular and Physiological Aspects(eds. by C.A. Navas and J.E. Caralho), pp. 209–229. Springer-Verlag, Germany.

Bradshaw, W.E. & Holzapfel, C.M. (2001) Genetic shift in photope-riodic response correlated with global warming. Proceedings of theNational Academy of Sciences of the United States of America , 98,14509–14511.

Brown, C.T., Howe, A., Zhang, Q. et al . (2012) A Reference-FreeAlgorithm for Computational Normalization of Shotgun SequencingData . arXiv:1203.4802.

Cox, M.P., Peterson, D.A. & Biggs, P.J. (2010) SolexaQA: at-a-glancequality assessment of Illumina second-generation sequencing data.BMC Bioinformatics , 11, 485.

Danks, H.V. (1987) Insect Dormancy: An Ecological Perspective.Biological Survey of Canada (Terrestrial Arthropods), Canada.

Davey, J.W., Hohenlohe, P.A., Etter, P.D. et al . (2011) Genome-wide genetic marker discovery and genotyping using next-generationsequencing. Nature Reviews Genetics , 12, 499–510.

Denlinger, D.L., Yocum, G.D. & Rinehart, J.L. (2005) Hormonalcontrol of diapause. (eds. by L. Gilbert, K. Iatrou and S. Gill), pp.615–650. Elsevier Press, The Netherlands. AQ2

Hahn, D.A. & Denlinger, D.L. (2011) Energetics of insect diapause.Annual Review of Entomology , 56, 103–121.

Haider, S., Ballester, B., Smedley, D. et al . (2009) BioMart CentralPortal—unified access to biological data. Nucleic Acids Research ,37, W23–W27.

Hawley, W.A., Reiter, P., Copeland, R.S. et al . (1987) Aedesalbopictus in North America: probable introduction in used tiresfrom northern Asia. Science, 236, 1114–1116.

Huang, X. & Madan, A. (1999) CAP3: a DNA sequence assemblyprogram. Genome Research , 9, 868–877.

Kostal, V. (2006) Eco-physiological phases of insect diapause. Journalof Insect Physiology , 52, 113–127.

Kukal, O., Denlinger, D.L. & Lee, R.E. (1991) Developmentaland metabolic changes induced by anoxia in diapausing andnondiapausing flesh fly pupae. Journal of Comparative PhysiologyB: Biochemical Systemic and Environmental Physiology , 160,683–689.

Langmead, B., Trapnell, C., Pop, M. & Salzberg, S.L. (2009) Ultrafastand memory-efficient alignment of short DNA sequences to thehuman genome. Genome Biology , 10, R25.

123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758

123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758

© 2013 The Royal Entomological Society, Physiological Entomology, 0, 0–0

Diapause transcriptome sequencing 9

Li, W.Z. & Godzik, A. (2006) Cd-hit: a fast program for clusteringand comparing large sets of protein or nucleotide sequences.Bioinformatics , 22, 1658–1659.

Martin, J.A. & Wang, Z. (2011) Next-generation transcriptomeassembly. Nature Reviews Genetics , 12, 671–682.

Michaud, M.R. & Denlinger, D.L. (2007) Shifts in the carbohy-drate, polyol, and amino acid pools during rapid cold-hardeningand diapause-associated cold-hardening in flesh flies (Sarcophagacrassipalpis): a metabolomic comparison. Journal of ComparativePhysiology B: Biochemical Systemic and Environmental Physiology ,177, 753–763.

Mori, A., Oda, T. & Wada, Y. (1981) Studies on the egg diapause andoverwintering of Aedes albopictus in Nagasaki. Tropical Medicine,23, 79–90.

Nene, V., Wortman, J.R., Lawson, D. et al . (2007) Genomesequence of Aedes aegypti , a major arbovirus vector. Science, 316,1718–1723.

Ning, Z.M., Cox, A.J. & Mullikin, J.C. (2001) SSAHA: a fastsearch method for large DNA databases. Genome Research , 11,1725–1729.

Pearson, R.G., Raxworthy, C.J., Nakamura, M. & Peterson, A.T.(2007) Predicting species distributions from small numbers ofoccurrence records: a test case using cryptic geckos in Madagascar.Journal of Biogeography , 34, 102–117.

Poelchau, M.F., Reynolds, J.A., Denlinger, D.L. et al . (2011) A denovo transcriptome of the Asian tiger mosquito, Aedes albopictus ,to identify candidate transcripts for diapause preparation. BMCGenomics , 12, 619.

Poelchau, M.F., Reynolds, J.A., Denlinger, D.L. et al . (2013) Deepsequencing reveals complex mechanisms of diapause preparation inthe invasive mosquito, Aedes albopictus . Proceedings of the RoyalSociety of London Series B: Biological Sciences in press.AQ3

Ragland, G.J., Denlinger, D.L. & Hahn, D.A. (2010) Mechanisms ofsuspended animation are revealed by transcript profiling of diapausein the flesh fly. Proceedings of the National Academy of Sciences ofthe United States of America , 107, 14909–14914.

Ragland, G.J., Egan, S.P., Feder, J.L. et al . (2011) Developmentaltrajectories of gene expression reveal cadidates for diapausetermination: a key life-history transition in the apple maggotfly Rhagoletis pomonella . Journal of Experimental Biology , 214,3948–3959.

Reynolds, J.A. & Hand, S.C. (2009) Embryonic diapause highlightedby differential expression of mRNAs for ecdysteroidogenesis,transcription and lipid sparing in the cricket Allonemobius socius .Journal of Experimental Biology , 212, 2074–2083.

Reynolds, J.A., Poelchau, M.F., Rahman, Z. et al . (2012) Transcriptprofiling reveals mechanisms for lipid conservation during diapausein the mosquito, Aedes albopictus . Journal of Insect Physiology , 58,966–973.

Rinehart, J.P., Li, A., Yocum, G.D. et al . (2007) Up-regulation of heatshock proteins is essential for cold survival during insect diapause.Proceedings of the National Academy of Sciences of the United Statesof America , 104, 11130–11137.

Robich, R.M. & Denlinger, D.L. (2005) Diapause in the mosquitoCulex pipiens evokes a metabolic switch from blood feeding tosugar gluttony. Proceedings of the National Academy of Sciences ofthe United States of America , 102, 15912–15917.

Schulz, M.H., Zerbino, D.R., Vingron, M. & Birney, E. (2012) Oases:robust de novo RNA-seq assembly across the dynamic range ofexpression levels. Bioinformatics , 28, 1086–1092.

Sim, C. & Denlinger, D.L. (2008) Insulin signaling and FOXOregulate the overwintering diapause of the mosquito Culex pipiens .

Proceedings of the National Academy of Sciences of the United Statesof America , 105, 6777–6781.

Sim, C. & Denlinger, D.L. (2009a) A shut-down in expression ofan insulin-like peptide, ILP-1, halts ovarian maturation duringthe overwintering diapause of the mosquito Culex pipiens . InsectMolecular Biology , 18, 325–332.

Sim, C. & Denlinger, D.L. (2009b) Transcription profiling andregulation of fat metabolism genes in diapausing adults of themosquito Culex pipiens . Physiological Genomics , 39, 202–209.

Slater, G. & Birney, E. (2005) Automated generation of heuristics forbiological sequence comparison. BMC Bioinformatics , 6, 31.

Sota, T. & Mogi, M. (1992) Survival time and resistance to desicationof diapause and non-diapause eggs of temperate Aedes (Stegomyia)mosquitoes. Entomologia Experimentalis et Applicata , 63, 155–161.

Surget-Groba, Y. & Montoya-Burgos, J. (2010) Optimization of denovo transcriptome assembly from next-generation sequencing data.Genome Research , 20, 1432–1440.

Tammariello, S.P. (2001) Regulation of the cell cycle during diapause.(eds. by D.L. Denlinger, J. Giebultowicz and D.S. Saunders), pp.173–183. Elsevier Science, B.V., The Netherlands. AQ4

Tammariello, S.P. & Denlinger, D.L. (1998) G0/G1 cell cycle arrestin the brain of Sarcophaga crassipalpis during pupal diapause andthe expression pattern of the cell cycle regulator, proliferating cellnuclear antigen. Insect Biochemistry and Molecular Biology , 28,83–89.

Tatar, M., Kopelman, A., Epstein, D. et al . (2001) A mutantDrosophila insulin receptor homolog that extends life-span andimpairs neuroendocrine function. Science, 292, 107–110.

Tauber, M.J., Tauber, C.A. & Masaki, S. (1986) Seasonal Adaptationsof Insects . Oxford University Press, New York, New York.

Trpis, M. (1970) A new bleaching and decalcifying method for generaluse in zoology. Canadian Journal of Zoology , 48, 892–893.

Urbanski, J.M., Benoit, J.B., Michaud, M.R. et al . (2010) Themolecular basis of increased desiccation resistance during diapausein the Asian tiger mosquito, Aedes albopictus . Proceedings ofthe Royal Society of London Series B: Biological Sciences , 277,2683–2692.

Urbanski, J., Mogi, M., O’Donnell, D. et al . (2012) Rapid adaptiveevolution of photoperiodic response during invasion and rangeexpansion across a climatic gradient. American Naturalist , 179,490–500.

Wang, R.L. (1966) Observations on the influence of photoperiod onegg diapause in Aedes albopictus Skuse. Acta Entomologica Sinica ,15, 75–77.

Waterhouse, R.M., Zdobnov, E.M., Tegenfeldt, F. et al . (2011)OrthoDB: the hierarchical catalog of eukaryotic orthologs in 2011.Nucleic Acids Research , 39, D283–D288.

Williams, K.D., Busto, M., Suster, M.L. et al . (2006) Natural variationin Drosophila melanogaster diapause due to the insulin-regulatedPI3-kinase. Proceedings of the National Academy of Sciences of theUnited States of America , 103, 15911–15915.

Wong, K., Jin, Y. & Moqtaderi, Z. (2013) Multiplex Illuminasequencing using DNA barcoding. Current Protocols in MolecularBiology , 7.11.1–7.11.11.

Zerbino, D.R. & Birney, E. (2008) Velvet: algorithms for de novoshort read assembly using de Bruijn graphs. Genome Research , 18,821–829.

Accepted 18 March 2013

123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758

123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758

© 2013 The Royal Entomological Society, Physiological Entomology, 0, 0–0

QUERIES TO BE ANSWERED BY AUTHOR

IMPORTANT NOTE: Please mark your corrections and answers to these queries directly onto the proof at the relevantplace. DO NOT mark your corrections on this query sheet.

Queries from the Copyeditor:

AQ1. Author: Please provide a zip code for the corresponding author’s address

AQ2. Author: Please provide the book title for Denlinger et al. (2005).

AQ3. Author: If the reference has now been published in print, please add relevant volume number and page range for Poelchauet al. (2013).

AQ4. Author: Please provide the book title for Tammariello (2001).