Open Collections

UBC Faculty Research and Publications

A high-density genetic map reveals variation in recombination rate across the genome of Daphnia magna Dukić, Marinela; Berner, Daniel; Roesti, Marius; Haag, Christoph R; Ebert, Dieter Oct 13, 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


52383-12863_2016_Article_445.pdf [ 1.24MB ]
JSON: 52383-1.0362015.json
JSON-LD: 52383-1.0362015-ld.json
RDF/XML (Pretty): 52383-1.0362015-rdf.xml
RDF/JSON: 52383-1.0362015-rdf.json
Turtle: 52383-1.0362015-turtle.txt
N-Triples: 52383-1.0362015-rdf-ntriples.txt
Original Record: 52383-1.0362015-source.json
Full Text

Full Text

RESEARCH ARTICLE Open AccessA high-density genetic map revealsvariation in recombination rate across thegenome of Daphnia magnaMarinela Dukić1*, Daniel Berner1, Marius Roesti1,4, Christoph R. Haag2,3† and Dieter Ebert1†AbstractBackground: Recombination rate is an essential parameter for many genetic analyses. Recombination rates arehighly variable across species, populations, individuals and different genomic regions. Due to the profoundinfluence that recombination can have on intraspecific diversity and interspecific divergence, characterization ofrecombination rate variation emerges as a key resource for population genomic studies and emphasises theimportance of high-density genetic maps as tools for studying genome biology. Here we present such a high-densitygenetic map for Daphnia magna, and analyse patterns of recombination rate across the genome.Results: A F2 intercross panel was genotyped by Restriction-site Associated DNA sequencing to construct thethird-generation linkage map of D. magna. The resulting high-density map included 4037 markers covering813 scaffolds and contigs that sum up to 77 % of the currently available genome draft sequence (v2.4) and55 % of the estimated genome size (238 Mb). Total genetic length of the map presented here is 1614.5 cMand the genome-wide recombination rate is estimated to 6.78 cM/Mb. Merging genetic and physical information weconsistently found that recombination rate estimates are high towards the peripheral parts of the chromosomes, whilechromosome centres, harbouring centromeres in D. magna, show very low recombination rate estimates.Conclusions: Due to its high-density, the third-generation linkage map for D. magna can be coupled with the draftgenome assembly, providing an essential tool for genome investigation in this model organism. Thus, our linkage mapcan be used for the on-going improvements of the genome assembly, but more importantly, it has enabled us tocharacterize variation in recombination rate across the genome of D. magna for the first time. These new insights canprovide a valuable assistance in future studies of the genome evolution, mapping of quantitative traits and populationgenetic studies.Keywords: Crossover, Daphnia magna, Linkage map, RAD-sequencing, Recombination breakpoints, Recombination rateBackgroundMeiotic recombination is an essential process in sexuallyreproducing eukaryotes since it is involved in the main-tenance of genome stability, in proper segregation ofchromosomes into haploid gametes, and in shaping pat-terns of genetic variation among offspring individuals[1]. Mechanistically, recombination between homolo-gous chromosomes is crucial for accurate repair of DNAdouble strand breaks that are induced in a highlycontrolled manner during early meiotic prophase I(reviewed in [1, 2]). Such homology-based repair ensuresthe maintenance of genome integrity, but also often repre-sents a physical bond between homologous chromosomes,critical for their positioning and proper segregation intothe gamete cells [2, 3]. In proceeding meiotic processes,physical connection between homologs will lead to recip-rocal (crossover; CO) or unidirectional (gene conversion)exchange of DNA between paternal and maternal chro-mosomes. It has been shown in several organisms thatmore abundant and uniformly distributed gene conver-sions have a limited influence on inherited genetic vari-ation as they affect small genomic regions (350–2000 bp;[4]) On the other hand, CO involves reciprocal allelic* Correspondence:†Equal contributors1University of Basel, Zoological Institute, Vesalgasse 1, Basel CH-4051,SwitzerlandFull list of author information is available at the end of the article© 2016 The Author(s). Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0International License (, which permits unrestricted use, distribution, andreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link tothe Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver( applies to the data made available in this article, unless otherwise stated.Dukić et al. BMC Genetics  (2016) 17:137 DOI 10.1186/s12863-016-0445-7exchange across longer chromosomal segments resultingin recombination of genetic variation that can be readilydetected following the inheritance of genetic markers inlarge pedigrees or experimental crosses. Consequently, re-combination rate is traditionally approximated as the ob-served frequency of COs (i.e. neglecting gene conversions)per unit of physical distance (e.g. cM/Mb).Over the last decade, advancements in sequencingtechniques have enabled studies of recombination rate atunprecedented resolution in many different species[5–8]. Importantly, there is accumulating evidence forlarge amounts of variation in recombination rateacross species, populations, individuals, and differentgenomic regions [5, 7, 9]. This is especially interestingfrom an evolutionary perspective since the distribution ofrecombination events across the genome defines the sizeof genomic fragments that will be incorporated into hap-lotypes exposed to selection. When recombination is rare,selection wields its influence across long genomic tractsthat may contain multiple loci with differing fitnesseffects. Theory predicts that genetic linkage between mul-tiple sites under selection leads to a reduction of the over-all efficiency of selection [10, 11] and high levels of COrecombination are considered favourable for breaking upassociation between loci subjected to contrasting selectivepressures [12]. In addition, genomic regions with low re-combination are expected to have lower levels of neutralpolymorphism than genomic regions with high recombin-ation rates because of positive (hitch-hiking) or negative(background) selection on sites in their physical neigh-bourhood [13]. Considering the profound influence thatthe recombination landscape can have on genome-widegenetic variation and diversity, analysis of the recombin-ation rate emerge as a key resource for population andevolutionary genomics studies, emphasising the import-ance of high-density genetic maps as essential tools forstudying many features of genome biology.Waterfleas of the genus Daphnia have emerged as awell-suited model system for studying genetics of fitnessrelated traits in environmental contexts, due to the ex-tensive knowledge of their ecology, a life-cycle includingclonal and sexual reproduction, and the development ofgenomic resources [14–16]. However, to take full advan-tage of this model-system, a better understanding of thegenome architecture of Daphnia is needed, as well as ofthe mechanisms that are shaping it. In the present study,we use a standard F2 intercross panel and Restrictionsite Associated DNA (RAD) sequencing for the con-struction of a high-density genetic map of D. magna,one of the best known and most widely used study spe-cies of the genus. Including more than 4000 markersacross the 238 Mb genome [17], we provide the firstcharacterization of recombination landscape along thechromosomes of D. magna. Our data clearly show highlevels of recombination towards chromosomal peripher-ies with chromosomal centres being almost deprived ofCOs. We discuss these findings in comparison withother organisms and address possible mechanismsunderlying the observed patterns of recombination ratevariation across the genome of this species.MethodsDesign of genetic crosses and DNA isolationD. magna individuals used in this study were obtainedby asexually (clonally) propagating lines selected from anF2 intercross panel that had already been used for theconstruction of microsatellite and SNP-array based gen-etic maps [17, 18] as well as for QTL mapping [19, 20].Details about the crossing design can be found in resultingpapers. Briefly, the F2 panel was established by first cross-ing two parental individuals obtained from two inbredclonal lines of D. magna (Xinb3 and Iinb1, hereafter re-ferred to as “parental lines”) to produce an F1. One of theparental lines (Xinb3) was a third-generation inbred off-spring (three rounds of within-clone mating, each roundbeing genetically equivalent to self-fertilization) of an indi-vidual from Southern Finland, the other (Iinb1) was afirst-generation inbred offspring of a German individual.A female from the Xinb3 (Finnish mother) and a malefrom the Iinb1 (German father) parental line were crossedto obtain the F1 hybrid line (called IXF1). By mass-matinggenetically identical brothers and sisters of the IXF1 line,F2 offspring were generated, with each initial offspring in-dividual (hatchling from a sexually produced egg) being afounder of a clonal F2 line that was maintained via asexualreproduction as a part of our F2 panel. The Xinb3 line isalso the clone on which the D. magna reference genomeis based (Daphnia Genomics Consortium). The draft gen-ome sequence version 2.4 was used in the present study.Finally, two to three females from each parental line, theIXF1 line, and 66 randomly chosen F2 lines were used toestablish asexually propagated sub-lines that were used forDNA extractions (pooling nine individuals for each line).Prior to DNA extractions, all individuals were cleanedby an antibiotic and starvation treatment to minimizealgal and bacterial contamination in the sample of gen-omic DNA. Animals were kept without food during3 days in a medium containing Ampicillin (Sigma),Streptomycin (Sigma) and Tetracycline (Sigma) at a con-centration of 50 mg/L each, and transferred daily tofresh antibiotic medium. To enforce the cleansing of gutcontents, a small amount of superfine Sephadex ® G-25(Sigma-Aldrich) was added frequently to the antibioticmedium, making dextran beads accessible to Daphniafor ingestion and gut evacuation. Animals with clear in-testine were sampled and used for DNA extractions. Inthe majority of cases, DNA was isolated immediatelyafter sampling, but in some instances, animals wereDukić et al. BMC Genetics  (2016) 17:137 Page 2 of 13stored in 70 % ethanol at −20 °C until extraction. DNAextraction was done using the DNeasy Blood and Tissuekit (Qiagen) including the RNaseA (100 mg/ml; Sigma)digestion step.RAD library preparation and sequencingWe prepared libraries for RAD-sequencing [21] adoptingthe protocol of Etter et al. [22] with modifications ac-cording to Roesti et al. [8]. Specifically, 1 μg of genomicDNA from each sample was digested with the PstI HFrestriction enzyme (NEB) in 50 μl reaction volume, for90 min. at 37 °C and then heat-inactivated following themanufacturer’s manual. A P1 sequencing adapter (5 μlof 100 nM stock solution), containing a unique 5-bp bar-code, was ligated to each sample using T4 DNA-ligase(NEB, 0.5 μl of 2,000,000 units/mL stock solution) in60 μl reaction volume for 45 min at room temperaturefollowed by heat-inactivation for 20 min at 65 °C. Thetotal of 70 samples (Xinb3, Iinb1, IXF1 and 66 F2 lines,with one F2 individual replicated twice) were then com-bined into two pools (one with 30 and one with 40 sam-ples) and sheared using a Bioruptor (Diagenode). Therationale of combining fewer individuals into the firstpool, which included the parental, IXF1 and 26 F2 lines(“parental” library), was to ensure higher sequencingdepth and genotyping quality for the founder individualsof the F2 panel, thus facilitating the robust identificationof informative SNPs for genetic mapping. The secondlibrary contained F2 lines exclusively.DNA fragments in a range of 250–500 bp were se-lected using agarose gel electrophoresis (1.25 %, 0.5XTBE), purified and blunt-ended (Quick Blunting Kit,NEB). Klenow fragment exo− (NEB) was used to adddA-overhangs, followed by P2 adapter ligation (1 μl of10 mM stock solution). Products were purified and PCRamplification was done using Phusion High-FidelityDNA polymerase (NEB). To minimize the probability ofPCR error, master mixes for each library were dividedinto six separate 12.5 μl reactions for amplification (30 sat 98 °C, 17 cycles of 98 °C 10 s, 65 °C 30 s, 72 °C 30 s,then a final extension for 5 min at 72 °C).The enriched RAD libraries were sequenced on separateIllumina HiSeq2000 lanes using 100 bp single-end sequen-cing (Quantitative Genomics Facility service platform,Deep Sequencing Unit Department of Biosystems Scienceand Engineering, ETH-Zurich in Basel, Switzerland).Defining genetic markers for linkage mappingIn total 259,580,561 raw 100 bp reads were generatedby sequencing (120,336,323 and 139,244,238 reads inthe first and the second library, respectively). Overallread quality was inspected using FastQC (BabrahamBioinformatics, The Babraham Institute) confirmingthat per-base quality score exceeded 30 (with theexception of the last ten bases of parental library). Acustom script (available upon request) coded in R[23] was used to sort raw reads according to uniquebarcodes into individual samples. Reads containingambiguous bases and reads that did not feature validbarcode or restriction-site sequence were discardedfrom further analysis (23 % of the total raw reads).Moreover, the last ten bases were trimmed from theremaining reads due to a decrease in base-callingquality. The cleaned and individually sorted 85 bpreads were aligned to the reference draft genome as-sembly v2.4 of D. magna using Novoalign v2.07( We allowed on average onehigh-quality mismatch or indel per 14 bases and ac-cepted only reads that aligned to unique locationwithin the reference genome. Eight F2 samples werediscarded because they were sequenced at substantiallylower depths compared to the other samples within thesame library. In summary, we achieved a mean coverageof 68-fold among the individuals from the parental library(including 22 F2 lines), and 40-fold among the final 37 F2lines from the second library.Stacks v1.08 [24] was used for identification of puta-tive marker loci and for genotyping. The samples fromthe two libraries were analysed separately, taking the dif-ferences in sequencing depth into account. In bothcases, individual SAM files were imported in Stacks andanalysed with the pipeline. Parental lineswere used to construct a “catalog” of loci (3 mismatchesallowed between reads mapping to the same locus, op-tion –n in The minimum coverage depth(option -m) was set to 25 (parental library) and 15(lower-coverage library) to call a stack (group of identi-cal reads). Stacks uses error-bounded model for SNPidentification however, since the prior information onsequencing error rate was not available, a lower andupper bound for the error rate were not specified (de-fault between 0 and 1). Default chi-square significancelevel (0.05) required to call a heterozyogote or homozy-gote was used. Custom MySQL scripts were used formerging the results from both libraries. Deleveraged loci(see [24]), loci with more than three SNPs and loci withmore than 2 alleles were excluded from the analysis. Inaddition, we were only interested in loci that werehomozygous for alternative alleles (aa, bb) in the paren-tal lines and heterozygous (ab) in the F1 hybrids. Intotal, haplotype and genotype data for 7183 putativemarkers were retrieved from the Stacks analysis.We inspected the distribution of missing values (perF2 line and per marker) among the obtained genotypes,since they are potential source of errors during linkagemap construction. This resulted in the removal of six F2lines from further analysis because they had more than30 % of missing genotypes (in comparison, the remainingDukić et al. BMC Genetics  (2016) 17:137 Page 3 of 1352 F2 lines had on average less than 14 % of missingvalues). Furthermore, we removed markers exhibitingmore than 20 % missing data across the F2 lines, as sug-gested by Catchen et al. [24] and Davey et al. [25]. Theresulting dataset comprised 52 F2 lines and 4849 geneticmarkers in total.Linkage analysisJoinMap 4.0 [26, 27] was used as the main software forgenetic map construction. However, several additionalsteps were taken (Additional file 1: Figure S1) tomaximize the number of markers that could be mappedand to avoid the reduction in mapping accuracy that isexpected in large datasets (>1000 markers; [28]). Firstwe selected a subset of 253 “anchor” markers (one ortwo markers per large scaffold that were successfully ge-notyped in more than 90 % of F2 lines) representing the211 largest scaffolds of the D. magna draft genome as-sembly (v2.4). Using the regression mapping algorithmwith default parameters in JoinMap, these markers weregrouped into 10 preliminary linkage groups (LGs) atLOD = 3. Assuming no assembly error at this point ofthe analysis, all other markers on the same scaffoldswere attributed to the same preliminary LG as the re-spective anchor marker. We then continued to expandthe preliminary LGs by performing contingency tableanalyses of segregation patterns. More precisely, wecompared terminal markers of scaffolds that were attrib-uted to one of the preliminary LGs, against the dataset ofso-far un-mapped markers. Only markers with very simi-lar segregation patterns (< 3 different genotypes amongthe 52 F2 lines) were assigned to the same preliminary LG(Pearson’s χ2 test, cut-off threshold P < 0.0001), whereasmarkers showing ambiguous association to two or moreLGs were discarded at this point of the analysis. Markerswith the extreme segregation ratio distortion (SRD) thatcontradicted surrounding markers within the same scaf-fold were removed. Following this procedure, 4045markers were assigned to one of the ten preliminary LGs,75 markers were removed while 729 markers remainedunattributed.Many markers included in preliminary map showedidentical segregation patterns across all F2 individuals(i.e. they did not show any evidence of CO recombin-ation). In total, 756 segregation patterns could be distin-guished within our preliminary dataset and the groupsof co-segregating markers are hereafter referred to as“bins” (1 to 384 markers per bin). One of the markersexhibiting the lowest number of missing genotypes (i.e.,successfully genotyped in the largest number of F2 lines)from each bin was denoted as “frame marker” (uniquesegregation pattern within the framework map) and wasused for creating a framework map, a non-redundantrepresentation of all detected segregation patternssuitable for further analysis with JoinMap. The groupingof frame markers into 10 LGs was confirmed at LOD = 7(maximum likelihood, ML, option and otherwise defaultparameter values of the program). We then continuedby iteratively adding sets of the remaining, unattributedmarkers to the preliminary map using same settings inJoinMap. After each round, newly grouped markers wereinspected and designated as frame or non-framemarkers, depending on whether their genotypes matchedone of the previously defined bins. Non-frame markerswere continuously omitted from the framework mapand kept separately for later construction of a compositemap. After several iterations, we managed to include atotal of 4761 markers in the composite map while 13markers did not map to any of the ten LGs and conse-quently, were omitted from the final dataset. Once allmarkers were included, the composite map wasinspected visually, and the ordering of the markerswithin the LGs was corrected, based on the available in-formation of physical position within the scaffolds(mostly applying to markers within the same bin, theposition of which could not be determined based on seg-regation patterns). Dubious genotypes were corrected,making the assumption that the vast majority of single-tons reflect genotyping errors rather than double COwithin short physical distance (i.e. between the focalmarker and the adjacent markers on both sides). Thus, ifthe genotype was not observed in at least 3 adjacentmarkers within the same scaffold, it was replaced with amissing value [29]. We also checked marker pairs ob-tained from sister RAD-tags (i.e. markers obtained fromRAD loci flanking the same PstI restriction site, hencewith a distance of <200 bp) and removed one marker ofthe pair as redundant. If both sister RAD-tags werehighly reliable markers (up to three missing values), theconsensus segregation pattern was kept (thus reducingthe number of missing genotypes in the data set). Ifthe RAD-tag pair showed inconsistent genotypeswithin the same F2 individual, these instances werereplaced with missing values as it is highly improb-able that a recombination event happened within sucha short distance.The final composite map comprised 4037 markers, outof which 952 were defined as frame markers (952 binswith 1 to 354 markers). Grouping and ordering ofmarkers within the framework map was confirmed usingthe ML algorithm at the LOD = 7 (JoinMap, default set-tings). Afterwards, each LG was analysed individually inJoinMap, with markers in fixed order and genetic dis-tances were calculated using the Kosambi mapping func-tion (Additional file 1). Furthermore, the mappingquality of the framework map was validated through anindependent approach using the CheckMatrix program(ć et al. BMC Genetics  (2016) 17:137 Page 4 of 13Estimating physical distances between markersThe current version (v2.4) of the D. magna genome is astill unfinished draft version. Hence, we used the follow-ing procedures to estimate the physical distances be-tween markers and the cumulative physical length ofeach LG: (i) Mapped scaffolds were considered orientedif they had two or more markers separated by at leastone recombination event (so the orientation of the scaf-fold ends could be estimated). Within oriented scaffolds,the distances between markers were known from theiralignment position while distances between two terminalmarkers of adjacent scaffolds were calculated based onthe position of markers within their scaffolds and thenumber of remaining base pairs up to the scaffold’s ends.Note that this assumes no gaps between adjacent scaffolds(see below). (ii) Scaffolds and contigs with only onemarker or without detected recombination events weredesignated as un-oriented. Physical lengths of un-orientedregions were estimated based on the sum of the totallengths of scaffolds and contigs included in those regions.Distances between markers within the non-recombiningregion were attributed an average value (estimated phys-ical length divided by the number of segments defined bymarkers). This was done because it was unknown whichend segment was adjacent to the next oriented scaffold.(iii) When small contigs mapped inside a longer, orientedscaffolds, their size was not considered, as it was assumedthat these contigs mapped to the region of uncertain nu-cleotides (Ns) inside the scaffold. Such regions are presenton all scaffolds due to paired-end sequencing with long,un-sequenced inserts.Analysis of recombination ratesR/qtl (countXO function, [30]) was used to count the re-combination breakpoints observed in each F2 for eachLG. Recombination breakpoints were detected as achange in a genotype along the LGs. More precisely, ob-served genotype transitions A→H, H→A, B→H orH→ B were counted as a single recombination break-point, while double breakpoints between successivemarkers would appear as A→ B or B→A genotypetransitions (“A” being homozygote for the alleles fromthe German father clone, “B” is homozygote for the al-leles from the Finnish mother clone while “H” annotatesheterozygote genotype). The mean number of recombin-ation breakpoints observed in F2 offspring correspondsto the expected mean number of COs during meiosis,averaged across males and females.Genome-wide recombination rate (GWRR) was calcu-lated by summing cumulative genetic distances of allLGs and dividing it by the most recent estimate of thetotal length of the D. magna genome (238 Mb; [17]). Anaverage recombination rate for each LG (chromosomalrecombination rate) was estimated in the same way butwe used the physical length that was based only on scaf-folds included in our map (see above). We calculated theintra-chromosomal (local) recombination rate betweeneach pair of adjacent markers as the ratio of genetic dis-tance and estimated physical distance between thosemarkers (cM/Mb; Additional file 2). Marey maps wereused to plot genetic distance (in cM) against physicaldistance (in Mb) and to visualise variation in recombin-ation rates along LGs [31]. In addition, local recombin-ation rates were plotted against the physical midpoints ofmarker intervals, and LOESS (locally weighted scatterplotsmoothing) was used for smoothing the estimated values(polynomial degree = 1, α value was adjusted to the dens-ity of markers in each linkage group to cover approxi-mately 2 Mb windows). It is important to note here thatthe chromosomal and the intra-chromosomal recombin-ation rates are probably overestimates because themapped scaffolds of the reference genome assembly donot cover the full genomic sequence of D. magna (only131 Mb in total). This effect is likely to be particularlystrong in repeat-rich regions which are not yet assembled.Therefore, the physical distances used here have to beconsidered as minimum estimates.GC content analysisTo test whether the sequence composition is associatedwith the recombination landscape in D. magna, we in-vestigated how GC content correlates with differences inrecombination rate. All analyses of the GC content weredone using the available reference genome sequence(v.2.4). At the chromosomal scale, we tested for differ-ences in sequence composition of scaffolds found in re-combining vs. non-recombining regions: We comparedthe average GC content of all scaffolds mapping to re-gions of low recombination with the ones mapping toregions with high recombination, omitting scaffoldsfound at the borders of these regions. Furthermore, toassess whether the magnitude of recombination rate cor-relates with GC content in more discrete intervals, werestricted our analysis to recombining regions only. Forthis, the two longest scaffolds of each LG were selectedand the GC content was extracted for each intervalbetween two markers for which local recombination ratewas estimated (interval size between 5 and 100 kb,depending on the spacing between markers).ResultsLinkage mapThe genetic map of D. magna constructed in this studyincludes 4037 markers (Additional file 1), assigned toten LGs that correspond to the ten chromosomes of D.magna (n = 10). 952 clusters of co-segregating markers(bins) were identified, and only one marker from eachcluster was used to assemble a framework map (“Frame”Dukić et al. BMC Genetics  (2016) 17:137 Page 5 of 13markers; Table 1.). The cumulative genetic lengths(Kosambi corrected) estimated for each LG ranged from205.4 cM for LG1 to 131.4 cM for LG10, with the totalmap spanning 1614.5 cM (Table 1). LGs were numberedaccording to their genetic length estimated in this study(from largest to smallest); not exactly corresponding tothe previously published D. magna linkage maps [17, 18].We also note that the terms LG and chromosome areused synonymously throughout the manuscript eventhough cytogenetic mapping and numbering of chromo-somes is not available for D. magna. The average geneticdistance between frame markers was 1.7 cM with 78 % ofthe distances being under 2 cM and the largest gap being14.5 cM (LG3, Fig. 1a), possibly corresponding to a regionwith a large assembly gap. The independent validation ofthe framework map is shown as a heatplot (Fig. 1b) withclearly visible LG borders and a red diagonal area, whichis generally considered as a sign of high mapping quality( regions showing significant segregation ratiodistortion (SRD) were identified. A region spanning0.77 Mb within LG5 has been described previously, andis due to an allele responsible for the “Unviable Eggs”phenotype [18]. Homozygotes for the alleles from theFinnish mother individual (hereafter B alleles) are highlyunderrepresented in this region, with complete defi-ciency located at 80.01 cM (within scaffold00084). An-other region, carrying the infertility allele responsible forthe “Red Dwarf” phenotype [18] also displayed SRD inour analysis. This region spans approximately 0.69 Mbwithin LG10 and shows complete deficiency of homozy-gotes for the alleles originating from the German fatherindividual (A alleles) at 72.29 cM (within scaffold01036).In addition to these previously described regions, wealso found a relatively small region with SRD, spanning0.15 Mb on LG7 (at 81.92 cM). However SRD in thisregion was weaker than in the two above regions asnone of the two homozygotes was completely absent.Nevertheless, the strong deficiency of BB homozygotesin this region (4 % genotype frequency among F2 off-spring) suggests the presence of a strongly deleterious,recessive allele in the Finnish mother clone.Genome coverage and scaffold mappingThe total size of the D. magna genome is estimated at238 Mb [17]. The draft genome assembly used in thisstudy (v2.4) comprises 40,356 scaffolds and contigs sum-ming up to 131,266,987 bp of genomic sequence (55 %of the estimated genome size). 813 scaffolds and contigswere incorporated in this map (Table 2); this fraction,however, represents 77 % (100,609,459 bp) of the se-quence currently assembled and 42 % of the estimatedgenome size. The high density of markers enabled us todetermine the orientation of 97 scaffolds (representing63,321,641 bp, i.e. 48 % of the reference genome;Table 2). We found only five scaffolds exhibiting incon-sistency between the physical position of markers in thecurrent assembly and their segregation pattern. In all in-stances, these scaffolds comprised two fragments map-ping to separate LGs or to different regions of the sameLG (Table 3), while the ordering of markers within thesefragments remained consistent. These few discrepancieslikely indicate errors in the reference genome assembly.Nevertheless, the small portion of scaffolds displayingputative assembly mismatches indicates an overall highquality of the draft genome assembly used here. Inaddition, scaffold01409 and scaffold01036, spanningparts of the SRD region on LG10 (see above), showedpartial overlap, probably due to our inability to preciselymap the markers within the region showing SRD.Recombination rate estimatesA total of 1564 recombination breakpoints were de-tected across all F2 individuals and across all LGs. Thenumber of detected recombination breakpoints per indi-vidual and LG mainly lies between zero and six, with anTable 1 Linkage map summary. The physical lengths refer to the cumulative length of the scaffolds mapped in each linkage groupLinkage group Number of markers Number of “Frame” markers Genetic length (cM) Physical length (Mb) Recombination rate (cM/Mb)1 441 124 205.38 13.57 15.142 706 112 177.51 15.97 11.123 312 62 175.77 10.15 17.324 449 97 170.68 9.84 17.355 426 104 168.02 8.77 19.166 407 100 165.91 9.04 18.357 377 85 139.98 9.23 15.168 362 98 139.96 9.41 14.879 319 91 139.89 7.41 18.8810 238 79 131.38 7.23 18.18total/average 4037 952 1614.48 100.61 16.55Dukić et al. BMC Genetics  (2016) 17:137 Page 6 of 13average of three and a maximum of 14 (Fig. 2). Thesecounts represent the number of recombination break-points observed in F2 offspring, the mean of which alsoestimates the minimum number of COs that occurredduring meiosis in F1, averaged across male and femalemeiosis. However, the variance in F2 recombinationbreakpoints and CO numbers during F1 meiosis is notthe same, as can be seen from the following consider-ation: If each chromosome pair undergoes exactly 1 COper meiosis, 50 % of the resulting gametes will have onerecombination breakpoint and the other 50 % will havezero. If these gametes are randomly combined to formF2 individuals, the number of recombination break-points in F2 individuals is the sum of those on the twogametes. Hence 25 % of the F2 individuals would havetwo recombination breakpoints (if each of the two gam-etes has one), 50 % would have one and 25 % wouldhave zero. Hence, the observation that no recombinationFig. 1 a Linkage length and marker distribution of the framework linkage map. The linkage groups (LGs) are ordered from LG1 to LG10 bydecreasing genetic length. Only “Frame” markers are shown with grey lines. Black arrows indicate regions with segregation ratio distortion (seetext). b Heatplot as graphical representation of the quality of the linkage map. The image is produced with CheckMatrix ( to validate the quality of mapping using REC score (low-left diagonal) and BIT score (top-right diagonal). Red colour represents tightlinkage and green to blue colour indicates no linkage. Borders of the LGs are indicated by interruptions (black arrowheads) of the red diagonalwhich confirms the quality of ordering markers within the LGDukić et al. BMC Genetics  (2016) 17:137 Page 7 of 13breakpoint was observed on some LGs in some individ-uals (see Fig. 2) does not imply that zero CO occurredin F1 meiosis during gamete formation that gave rise tothese individuals.A genome-wide recombination rate (GWRR) of6.78 cM/Mb was calculated based on the ratio of thetotal cumulative genetic map length (1614.48 cM) andthe estimated genome size of D. magna (238 Mb; [17]).For an estimation of the GWRR based on the genomelength that was effectively covered by our markers, weused the total length of the current genome assembly(131 Mb) and accordingly obtained a substantially higherestimate of 12.32 cM/Mb. Due to the gaps within thegenome assembly, the later GWRR value has to beregarded as an overestimate.Nevertheless, assuming that the missing genomic se-quence is not randomly dispersed within the genome,but rather uniformly distributed among chromosomes(largely as heterochromatic regions), we can make com-parisons between recombination rates estimated for eachLG (i.e., chromosome). Genetic length increases linearlywith the estimated physical length of each LG (Fig. 3a)with an intercept larger than zero, indicating that eventhe smallest chromosomes harbour at least one CO.Consequently, smaller chromosomes display more re-combination per unit of physical distance resulting instrong negative correlation between recombinationrate and the estimated physical length of LGs (Fig. 3b;Pearson’s correlation; R = -0.839; n = 10; P < 0.002).Recombination rate varied extensively within LGs(Fig. 4). In each of the 10 LGs, we detected one largeregion (two in the case of LG3) where recombinationwas rare or apparently absent. These low-recombinationregions are situated mainly in the chromosomal centresand comprise up to 40 % of the mapped genomic se-quence. In all cases (except only one of the two regionsof LG3), these regions span the map position of thecentromere [32]. In each LG the low-recombinationregions are flanked by regions of high recombination.Furthermore, we observed a drop in recombination ratestowards the very ends of the LGs. However, due to thecurrent state of the genome assembly and the generationTable 2 Summary of scaffolds and contigs included in the linkage mapLinkage group No. of mapped scaffolds& contigsNo. of scaffolds Scaffold bases No. of contigs Contig bases Oriented scaffolds Oriented scaffold bases1 84 63 13938534 21 24389 11 104194012 112 84 16116820 28 30759 7 90153163 81 61 11176544 20 16793 11 29941364 73 60 9026485 13 14395 5 66214375 102 77 8909170 25 28609 13 53608826 81 62 8721446 19 21200 10 64214487 72 55 9304851 17 13654 5 60196978 72 52 8742755 20 20271 11 59137149 77 54 7544014 23 28315 12 549729110 59 47 7119613 12 10754 12 5058319TOTAL 813 615 100600232 198 209139 97 63321641Table 3 Scaffolds of Daphnia magna genome assembly v2.4 whose markers map to different linkage groups. An exception isscaffold02227 which is divided into two fragments mapped to different parts of the LG8Misassembled scaffolds Total length (bp) No. markers; position within scaffold Linkage groupscaffold00093 237880 1 marker; 28344 bp 8scaffold00093 4 markers; 135580–210844 bp 2scaffold03387 219786 6 markers; 11029–100849 bp 7scaffold03387 1 marker; 133916 bp 5scaffold02486 541490 20 markers; 35514–304827 bp 2scaffold02486 7 markers; 400767–528088 bp 4scaffold00233 263417 3 markers; 10570–50454 bp 1scaffold00233 3 markers; 82320–131292 bp 4scaffold02227 412371 7 markers; 311258–396970 bp 8scaffold02227 9 markers; 11932–261034 bp 8Dukić et al. BMC Genetics  (2016) 17:137 Page 8 of 13of markers sensitive to sequence motifs (RAD), theseterminal regions were difficult to study in more detail.GC content analysisWe found no difference in the mean GC content betweenthe scaffolds mapping to low-recombination regions andthe ones located in regions with high recombination(Paired t-test; n = 10; P = 0.97). Focusing only on scaffoldsin highly recombining regions, we found a weak positivecorrelation between GC content and recombinationrate (Pearson’s correlation; r = 0.184; n = 907 markerintervals; P < 10−8).DiscussionWe present a high-density genetic map for D. magnathat can be coupled with the draft genome assembly,thus providing a valuable resource for genomic investi-gation and QTL mapping. In contrast to the previouslypublished maps for D. magna, all large scaffolds in ourmap are covered by multiple markers, enabling us to de-termine their orientation within the chromosome (unlesssituated in a non-recombining region) and the linkage toother genome segments, which were not previouslyknown. Thus, the linkage map constructed in this waycan be used for the on-going D. magna genome assem-bly. Co-segregating markers were used to confirm thatthe observed patterns of segregation are true biologicalevents rather than methodological artefacts. Hence,although a relatively small number of F2 lines was in-cluded in our study, the accuracy of final ordering ofmarkers within and between scaffolds is likely high,much higher compared to previous maps, which werebased on few microsatellites [18] or an error-proneSNP-array [17]. In addition to increased reliably, thethird-generation linkage map presented here, enablesmerging of genetic and physical information, and there-fore addressing the variation in recombination rateacross the genome of D. magna for the first time. This isalso the most comprehensive study of recombinationlandscape for any crustacean species reported so far.Genome-wide and chromosomal recombination rateThe genome-wide recombination rate (GWRR) of D.magna as estimated in the present study is 6.8 cM/Mb,Fig. 2 Bubble plot of recombination breakpoints count in each F2 line and LG. Circle area corresponds to the number of F2 lines with a specific countFig. 3 Relationship between estimated physical length of chromosomesand a genetic length and b chromosomal recombination rateDukić et al. BMC Genetics  (2016) 17:137 Page 9 of 13which is slightly higher than the value of 6.2 cM/Mbassessed from the previously published SNP-based map[17]. Similarly, the GWRR of the related species D. pulexis estimated at 7.2 cM/Mb [33], suggesting conservedlevels of recombination among Daphnia species. Muchlower GWRRs were reported for a handful of crustaceanspecies for which genetic maps and genome size esti-mates are available (mean = 1.2 cM/Mb; [34–37]). Alsocompared to other animal taxa, GWRR of Daphnia ishigh, similar to some Hymenoptera and Lepidopteranspecies [38]. It has been hypothesized that the elevatedGWRRs are favoured in systems with reducedopportunity for sex and recombination including haplo-diploidy, cyclic parthenogenesis or species where recom-bination is restricted to one sex [38]. However, manyexceptions from this pattern [36, 38, 39] indicate thatpeculiar life-cycles per se are likely not the only explan-ation for high recombination rates.More consistently, it has been shown that recombin-ation rate scales negatively with genome size in many or-ganisms, mainly due to the fact that majority of specieshave at least one COs per chromosome, even on thesmallest chromosomes [40]. Consistent with this, wefound that the positive linear relationship betweenFig. 4 Recombination rate along the 10 chromosomes of Daphnia magna. Dots indicate genetic position of markers in centimorgans (referring tothe left axis), plotted against their estimated physical position in the genome (in megabases). An average recombination rate (cM/Mb) wascalculated for intervals between adjacent markers and plotted against their physical midpoint (Mb). Data points are not shown but the grey curvesindicate data smoothed by LOESS, with the polynomial degree of one and the sampling portion adjusted for each LG according to the density ofmarkers to obtain a constant smoothing resolution across the panels (moving average of 2 Mb)Dukić et al. BMC Genetics  (2016) 17:137 Page 10 of 13genetic distance and physical distance of chromosomesin D. magna has a positive y-intercept, and, hence,smaller chromosomes experience more recombinationper physical distance when compared to larger ones.The mean numbers of observed recombination break-points in F2 individuals, as well as the estimated geneticlength per chromosome indicate that the different chro-mosomes of D. magna undergo on average between 2.6and 4.1 CO per meiosis (one expected CO correspondsto 50 cM of genetic map length). It is also interesting tonotice that the number of detected recombinationbreakpoints varies considerably between F2 offspringand individual chromosomes. In 4.8 % of all cases, no re-combination breakpoints were detected along an entireLG within a given individual. These instances likely rep-resent the chance union of two gametes that were non-recombinant for this LG. Such gametes occur even inmeioses with one or several COs and therefore do notrepresent evidence for meioses without CO. Further-more, we may have missed some breakpoints when theywere too closely spaced or when they occurred in theterminal chromosome regions, i.e. peripheral to the lastmarker. However, we believe that this would only ex-plain a small part of the cases without any detectedbreakpoints. On the other extreme, a few individualshad very high number of recombination breakpointsalong a given LG (up to 14). These may suggest a ratherhigh variance in the number of COs per meiosis, or, al-ternatively, they may partly be explained by genotypingerrors. Overall, these instances (in both directions) are,however, rare and hence it is unlikely that they signifi-cantly influence the summary statistics on the overallgenetic map length presented here.Local recombination ratesThe genetic map of D. magna described in this studyrevealed major intra-chromosomal variation in recom-bination rates. The determinants of non-random COpatterning are not yet clear, though several lines of evidenceindicate that the hierarchical combination of multiple fac-tors plays a role in shaping the recombination landscapeacross genomes. These factors include chromosomal sizeand structural properties, large subchromosomal domains,chromatin structure and the local nucleotide composition[41, 42]. In D. magna we found that, CO recombination ismore likely to occur in the peripheral parts of the chromo-some, while large regions of low or no recombination occurnear the central parts of all chromosomes. As for many ani-mal and plant species that were studied earlier [6, 43–46],these regions of extremely reduced recombination coincidewith centromeres of D. magna [32]. LG3 is an exceptionbecause two regions without recombination were detected,though only one of these two regions (the one at 96 cM,also containing a centromere) was also found by Svendsenet al. [32]. The second non-recombining region on this LGmight be the result of an inversion or a large indel sup-pressing recombination specifically in the inter-populationcross used for the present study. It is also interesting tonote that regions without recombination probably extendto the pericentromeric heterochromatin regions becausecentromeric regions are usually not included in geneticmaps due to the repetitive nature of their sequence(tandem sequence repeats).Along with the structural confines on recombinationlandscape, in the majority of animal species that werestudied hitherto it has been shown that recombinationrates covary with the local nucleotide composition [8, 39,47–50]. High GC content is considered as a predictor ofregions with high recombination rate due to involvementof GC-rich elements in the process of recombination (rec-ognition sites of DNA binding proteins) or, conversely,high recombination rates can lead to high GC content dueto GC-biased gene conversions that accompany CO events[51]. In D. magna there is no difference in GC content be-tween recombining and non-recombining regions. Withinthe recombining regions, we found that GC content indeedcorrelates positively with recombination rate, although thedetected correlation is weak. These findings are not sur-prising considering that the correlation between nucleotidecomposition and recombination rate occurs at very smallphysical scales, so testing for this association is stronglydependent on the interval size used and on the precision atwhich recombination hotspots can be identified.ConclusionsDue to the high density of markers included, the geneticmap presented here has enabled us to investigate howCO varies in frequency and distribution along the chro-mosomes of D. magna. We have identified large regionsof low or no recombination in the chromosomal centrescovering approximately 40 % of the mapped genome.These regions also contain the centromeres, but likelyextend much beyond the actual centromeric regions. Incontrast, CO recombination occurs mainly towards thechromosomal peripheries. These insights into therecombination landscape of D. magna can provide avaluable assistance in future studies of the genomearchitecture, mapping of quantitative traits and popula-tion genetic studies. Following improvements in genomeannotation, it will be important to understand how genedensity correlates with variation in recombination rate.Both the density of loci potentially under selection andthe variation in CO rate across the genome can bias gen-omic analyses and should be considered as importantfactors in QTL mapping protocols [52] or populationgenetic studies [13, 51] aiming to understand the effectsof selection on genetic variation within and betweenpopulations of D. magna.Dukić et al. BMC Genetics  (2016) 17:137 Page 11 of 13Additional filesAdditional file 1: Figure S1. A flow chart of the genetic mapconstruction process. For a detailed description, see Methods section“Linkage analysis”. (PDF 84 kb)Additional file 2: The framework and composite map of Daphniamagna. Listed information include the names of RAD markers; markeralignment position to D. magna genome assembly v2.4; marker assignmentas a representative of a segregation pattern (“FRAME” marker), significancelevel of the segregation ratio distortion (SRD) for each marker based on thep-value of Chi-square test for a difference between the observed and theexpected Mendelian ratio (p < 0.1 *, p < 0.05 **, p < 0.01 ***, p < 0.005 ****,p < 0.001 *****, p < 0.0005 ******, p < 0.0001 *******); the number of thelinkage group; the position of the marker within the linkage group (in cM,Kosambi corrected); genotype data used for the construction of geneticmap. (XLSX 1401 kb)Additional file 3: Estimated physical (in bp) and genetic (in cM)positions of all 4037 markers and the average recombination rates(cM/Mb) estimated for intervals between adjacent markers. (XLSX 281 kb)AbbreviationsCO: Crossover; LG: Linkage group; QTL: Quantitative trait loci; RAD: Restrictionsite associated DNAAcknowledgementsWe would like to express our gratitude to Walter Salzburger for kindlysharing laboratory resources and infrastructure. We thank Peter D. Fields andRoberto Arbore for reading the previous versions of the manuscript andfruitful discussions. Jürgen Hottinger, Urs Stiefel and Nicolas Boileaufacilitated laboratory work while Lukas Zimmerman and sciCORE team(University of Basel) provided technical support for the data analysis. Illuminasequencing of RAD libraries was done by Christian Beisel and Ina Niessen atthe Quantitative Genomics Facility service platform, Deep Sequencing UnitDepartment of Biosystems Science and Engineering, ETH-Zurich.FundingThis work was funded by Swiss National Science Foundation (SNSF). MD alsoreceived 6 month fellowship from the Burckhardt-Bürgin-Stiftung (Universityof Basel, Switzerland).Availability of data and materialsThe datasets supporting the conclusions of this article are included withinthe article and its additional files (Additional files 2 and 3).All demultiplexed read data used for genotyping have been submitted tothe National Center for Biotechnology Information Sequence Read Archive(BioProject ID: PRJNA343627; study accession SRP090240) at’ contributionsMD, CRH and DE conceived the study. DE designed the F2 panel. MD andMR prepared the RAD libraries. MD, CRH and DB analysed the data. MD andCRH wrote the manuscript with an input from DE, DB and MR. All authorshave read and approved the final manuscript.Competing interestsThe authors declare they have no competing interests.Consent for publication“Not applicable”.Ethics approval and consent to participate“Not applicable”.Author details1University of Basel, Zoological Institute, Vesalgasse 1, Basel CH-4051,Switzerland. 2Centre d’Ecologie Fonctionnelle et Evolutive – CEFE UMR 5175,CNRS – Université de Montpellier – Université Paul-Valéry Montpellier –EPHE, campus CNRS, 1919, route de Mende, 34293 Montpellier Cedex 5,France. 3Department of Biology, Ecology and Evolution, University ofFribourg, Chemin du Muśee 10, 1700 Fribourg, Switzerland. 4BiodiversityResearch Centre and Zoology Department, University of British Columbia,Vancouver, BC V6T 1Z4, Canada.Received: 21 March 2016 Accepted: 4 October 2016References1. Baudat F, Imai Y, de Massy B. Meiotic recombination in mammals:localization and regulation. Nat Rev Genet. 2013;14:794–806.2. Gerton JL, Hawley RS. Homologous chromosome interactions in meiosis:diversity amidst conservation. Nat Rev Genet. 2005;6:477–87.3. Zickler D, Kleckner N. Recombination, Pairing, and Synapsis of Homologsduring Meiosis. Cold Spring Harb Perspect Biol. 2015;7:a016626.4. Padhukasahasram B, Rannala B. Meiotic gene-conversion rate and tractlength variation in the human genome. Eur. J. Hum. Genet. 2013;1–8. McVean GAT, Myers SR, Hunt S, Deloukas P, Bentley DR, Donnelly P. Thefine-scale structure of recombination rate variation in the human genome.Science. 2004;304:581–4.6. Jensen-Seaman MI, Furey TS, Payseur BA, Lu Y, Roskin KM, Chen C-F, et al.Comparative recombination rates in the rat, mouse, and human genomes.Genome Res. 2004;14:528–38.7. Chan AH, Jenkins PA, Song YS. Genome-wide fine-scale recombination ratevariation in Drosophila melanogaster. PLoS Genet. 2012;8:e1003090.8. Roesti M, Moser D, Berner D. Recombination in the threespine sticklebackgenome–patterns and consequences. Mol Ecol. 2013;22:3014–27.9. Smukowski CS, Noor MAF. Recombination rate variation in closely relatedspecies. Heredity. 2011;107:496–508.10. Hill WG, Robertson A. The effect of linkage on limits to artificial selection.Genet Res. 1966;8:269–94.11. Comeron JM, Williford A, Kliman RM. The Hill-Robertson effect: evolutionaryconsequences of weak selection and linkage in finite populations. Heredity.2008;100:19–31.12. Barton NH. Mutation and the evolution of recombination. Philos Trans R SocB Biol Sci. 2010;365:1281–94.13. Cutter AD, Payseur BA. Genomic signatures of selection at linked sites:unifying the disparity among species. Nat Rev Genet. 2013;14:262–74.14. Eads BD, Andrews J, Colbourne JK. Ecological genomics in Daphnia: stressresponses and environmental sex determination. Heredity. 2008;100:184–90.15. Colbourne JK, Pfrender ME, Gilbert D, Thomas WK, Tucker A, Oakley TH, etal. The ecoresponsive genome of Daphnia pulex. Science. 2011;331:555–61.16. Ebert D. A genome for the environment. Science. 2011;331:539.17. Routtu J, Hall MD, Albere B, Beisel C, Bergeron RD, Chaturvedi A, et al.An SNP-based second-generation genetic map of Daphnia magna andits application to QTL analysis of phenotypic traits. BMC Genomics.2014;15:1033.18. Routtu J, Jansen B, Colson I, De Meester L, Ebert D. The first-generationDaphnia magna linkage map. BMC Genomics. 2010;11:508.19. Roulin AC, Routtu J, Hall MD, Janicke T, Colson I, Haag CR, et al. Localadaptation of sex induction in a facultative sexual crustacean: Insights fromQTL mapping and natural populations of Daphnia magna. Mol Ecol. 2013;22:3567–79.20. Routtu J, Ebert D. Genetic architecture of resistance in Daphnia hostsagainst two species of host-specific parasites. Heredity. 2015;114(2):241-8.doi:10.1038/hdy.2014.97.21. Baird NA, Etter PD, Atwood TS, Currey MC, Shiver AL, Lewis ZA, et al. RapidSNP discovery and genetic mapping using sequenced RAD markers. PLoSOne. 2008;3:e3376.22. Etter PD, Bassham S, Hohenlohe PA, Johnson EA, Cresko WA. SNP discoveryand genotyping for evolutionary genetics using RAD sequencing. In:Orgogozo V, Rockman M V., editors. Mol. Methods Evol. Genet. 2011 p. 157–78.23. Gentleman RC, Carey VJ, Bates DM, Bolstad B, Dettling M, Dudoit S, et al.Bioconductor: open software development for computational biology andbioinformatics. Genome Biol. 2004;5:R80.24. Catchen J, Hohenlohe P. Stacks: an analysis tool set for populationgenomics. Mol Ecol. 2013;22:3124–40.25. Davey JW, Cezard T, Fuentes-Utrilla P, Eland C, Gharbi K, Blaxter ML. Specialfeatures of RAD Sequencing data: implications for genotyping. Mol Ecol.2013;22:3151–64.Dukić et al. BMC Genetics  (2016) 17:137 Page 12 of 1326. Van Ooijen JW. JoinMap® 4, Software for the calculation of gene tic linkagemaps in experimental populations. Kyazma B.V., Wageningen, Netherlands.Kyazma B.V., Wageningen, Netherlands.; 2006.27. Stam P. Construction of integrated genetic linkage maps by means of anew computer package: Join Map. Plant J. 1993;3:739–44.28. Cheema J, Dicks J. Computational approaches and software tools forgenetic linkage map estimation in plants. Brief Bioinform. 2009;10:595–608.29. Isidore E, Van Os H, Andrzejewski S, Bakker J, Barrena I, Bryan GJ, et al.Toward a marker-dense meiotic map of the potato genome : lessons fromlinkage group I. Genetics. 2003;165:2107–16.30. Broman KW, Sen S. A Guide to QTL Mapping with R/qtl. New York City:Springer; 2009.31. Rezvoy C, Charif D, Guéguen L, Marais GAB. MareyMap: An R-based toolwith graphical interface for estimating recombination rates. Bioinformatics.2007;23:2188–9.32. Svendsen N, Reisser CMO, Dukić M, Thuillier V, Liautard-Haag C, Fasel D, etal. Uncovering cryptic asexuality in Daphnia magna by RAD-sequencing.Genetics. 2015;201:1143–55.33. Xu S, Ackerman MS, Long H, Bright L, Spitze K, Ramsdell JS, et al. Amale-specific genetic map of the microcrustacean Daphnia pulex basedon single sperm whole-genome sequencing. Genetics. 2015;201:31–8.34. Du Z-Q, Ciobanu DC, Onteru SK, Gorbach D, Mileham AJ, Jaramillo G, et al.A gene-based SNP linkage map for pacific white shrimp, Litopenaeusvannamei. Anim Genet. 2010;41:286–94.35. You E-M, Liu K-F, Huang S-W, Chen M, Groumellec ML, Fann S-J, et al.Construction of integrated genetic linkage maps of the tiger shrimp(Penaeus monodon) using microsatellite and AFLP markers. Anim Genet.2010;41:365–76.36. Foley BR, Rose CG, Rundle DE, Leong W, Moy GW, Burton RS, et al. Agene-based SNP resource and linkage map for the copepod Tigriopuscalifornicus. BMC Genomics. 2011;12:568.37. De Vos S, Bossier P, Van Stappen G, Vercauteren I, Sorgeloos P, Vuylsteke M.A first AFLP-based genetic linkage map for brine shrimp Artemia franciscanaand its application in mapping the sex locus. PLoS One. 2013;8:e57585.38. Wilfert L, Gadau J, Schmid-Hempel P. Variation in genomic recombination ratesamong animal taxa and the case of social insects. Heredity. 2007;98:189–97.39. Niehuis O, Gibson JD, Rosenberg MS, Pannebakker BA, Koevoets T, JudsonAK, et al. Recombination and its impact on the genome of the haplodiploidparasitoid wasp Nasonia. PLoS One. 2010;5:e8597.40. Lynch M. The origins of eukaryotic gene structure. Mol Biol Evol. 2006;23:450–68.41. Zhu X, Keeney S. Zip it up to shut it down. Cell Cycle. 2014;13:2157–8.42. Libuda DE, Uzawa S, Meyer BJ, Villeneuve AM. Meiotic chromosomestructures constrain and respond to designation of crossover sites. Nature.2013;502:703–6.43. Salomé PA, Bomblies K, Fitz J, Laitinen RAE, Warthmann N, Yant L, et al. Therecombination landscape in Arabidopsis thaliana F2 populations. Heredity.2012;108:447–55.44. Farré M, Micheletti D, Ruiz-Herrera A. Recombination rates and genomicshuffling in human and chimpanzee - A new twist in the chromosomalspeciation theory. Mol Biol Evol. 2013;30:853–64.45. Davey JW, Hohenlohe PA, Etter PD, Boone JQ, Catchen JM, Blaxter ML.Genome-wide genetic marker discovery and genotyping using next-generationsequencing. Nat Rev Genet. 2011;12:499–510.46. Wong AK, Ruhe AL, Dumont BL, Robertson KR, Guerrero G, Shull SM, et al. Acomprehensive linkage map of the dog genome. Genetics. 2010;184:595–605.47. Nachman MW. Variation in recombination rate across the genome:Evidence and implications. Curr Opin Genet Dev. 2002;12:657–63.48. Tortereau F, Servin B, Frantz L, Megens H, Milan D, Rohrer G, et al. A high densityrecombination map of the pig reveals a correlation between sex-specificrecombination and GC content. BMC Genomics. 2012;13:586–98.49. Beye M, Gattermeier I, Hasselmann M, Gempe T, Schioett M, Baines JF, et al.Exceptionally high levels of recombination across the honey bee genome.Genome Res. 2006;16:1339–44.50. Backström N, Forstmeier W, Schielzeth H, Mellenius H, Nam K, Bolund E, etal. The recombination landscape of the zebra finch Taeniopygia guttatagenome. Genome Res. 2010;20:485–95.51. Webster MT, Hurst LD. Direct and indirect consequences of meioticrecombination: implications for genome evolution. Trends Genet. 2012;28:101–9.52. Noor MA, Cunningham AL, Larkin JC. Consequences of recombination ratevariation on quantitative trait locus mapping studies: Simulations based onthe Drosophila melanogaster genome. Genetics. 2001;159:581–8.•  We accept pre-submission inquiries •  Our selector tool helps you to find the most relevant journal•  We provide round the clock customer support •  Convenient online submission•  Thorough peer review•  Inclusion in PubMed and all major indexing services •  Maximum visibility for your researchSubmit your manuscript your next manuscript to BioMed Central and we will help you at every step:Dukić et al. BMC Genetics  (2016) 17:137 Page 13 of 13


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items