Copyright � 2009 by the Genetics Society of AmericaDOI: 10.1534/genetics.108.096487

Rapid High Resolution Single Nucleotide Polymorphism–ComparativeGenome Hybridization Mapping in Caenorhabditis elegans

Stephane Flibotte,* Mark L. Edgley,† Jason Maydan,† Jon Taylor,† Rick Zapf,† Robert Waterston‡

and Donald G. Moerman†,§,1

*Canada’s Michael Smith Genome Sciences Centre, British Columbia Cancer Agency, Vancouver, British Columbia V5Z 4S6, Canada,†Michael Smith Laboratories, University of British Columbia, Vancouver, British Columbia V6T 1Z4, Canada, ‡Department of

Genome Sciences, University of Washington, Seattle, Washington 98195-7730 and §Department of Zoology,University of British Columbia, Vancouver, British Columbia V6T 1Z4, Canada

Manuscript received September 19, 2008Accepted for publication October 24, 2008

ABSTRACT

We have developed a significantly improved and simplified method for high-resolution mapping ofphenotypic traits in Caenorhabditis elegans using a combination of single nucleotide polymorphisms (SNPs)and oligo array comparative genome hybridization (array CGH). We designed a custom oligonucleotidearray using a subset of confirmed SNPs between the canonical wild-type Bristol strain N2 and the Hawaiianisolate CB4856, populated with densely overlapping 50-mer probes corresponding to both N2 and CB4856SNP sequences. Using this method a mutation can be mapped to a resolution of�200 kb in a single geneticcross. Six mutations representing each of the C. elegans chromosomes were detected unambiguously and athigh resolution using genomic DNA from populations derived from as few as 100 homozygous mutantsegregants of mutant N2/CB4856 heterozygotes. Our method completely dispenses with the PCR,restriction digest, and gel analysis of standard SNP mapping and should be easy to extend to any organismwith interbreeding strains. This method will be particularly powerful when applied to difficult or hard-to-map low-penetrance phenotypes. It should also be possible to map polygenic traits using this method.

POLYMORPHISMS have long been useful in geneticmapping because they are abundant and usually

without phenotypic consequences. With genome se-quences now at hand, single nucleotide polymorphisms(SNPs) are readily detectable and new assay methodsare being developed. In fact, in humans, SNPs are nowat the heart of whole genome association mapping ofcommon variants (Carlson et al. 2004). SNPs have alsobeen used in Caenorhabditis elegans (Wicks et al. 2001;Swan et al. 2002; Davis et al. 2005) and other modelorganisms, (Van Eijk et al. 2004; Chen et al. 2008) butto date these usually require multiple handling stepsand are relatively complicated. Because the mapping ofnew mutants, particularly suppressors and enhancers,remains central to the study of these organisms, wesought to develop a more efficient method that wouldexploit advances in SNP detection on a large scale.

Our alternative strategy, called SNP–comparativegenome hybridization (CGH), exploits the high densityof SNPs between N2 and CB4856 for genetic mapping inC. elegans. It has been estimated that the wild-type strainsN2 and CB4856 exhibit a genetic variation of one SNPper 840 bp in the nuclear genome (Wicks et al. 2001).An evenly distributed set of SNPs at sufficient density

can therefore be used to map virtually any new muta-tion. Using chips designed by ourselves and manufac-tured by Roche NimbleGen we have shown that afteronly a single genetic cross and a single DNA preparationa mutation can be mapped to an �200–400-kb intervalon a chromosome. The SNP-CGH mapping protocoldescribed here is easier and more efficient than any SNPmapping protocol that we are aware of that is currentlyavailable for C. elegans.

MATERIALS AND METHODS

Strains used: The marker strains used for this study weredpy-5(e61) I (F27C1.8), dpy-10(e128) II (T14B4.7), dpy-17(e164)III (F54D8.1), unc-22(e66) IV (ZK617.1), dpy-11(e224) V(F46E10.9), and unc-2(ra612) X (T02C5.5). The Hawaiianstrain used for SNPs was CB4856.

Genetic crosses, DNA preparations, and hybridization tomicroarrays: For each chromosome marker, CB4856 maleswere crossed to marker homozygotes and 10 WT F1 progenywere singled on 60-mm petri plates containing NGM spreadwith Escherichia coli OP50. Ten F2 mutant homozygotes werepicked from each clone to corresponding single 60-mm platesfor a total of 100 animals on 10 plates for each marker line.Populations were allowed to grow to starvation, then collectedby washing with M9 buffer containing 0.01% Triton X-100,pelleted by centrifugation, and pooled by marker onto single150-mm petri plates containing rich NGM (standard recipewith 83 peptone) spread with E. coli X1666. Each such pooledpopulation was grown to starvation and collected by washing

1Corresponding author: Department of Zoology, 6270 University Blvd.,Vancouver, BC V6T 1Z4, Canada. E-mail: moerman@zoology.ubc.ca

Genetics 181: 33–37 ( January 2009)

and centrifugation, frozen at�80� in 2.5 volumes of worm lysisbuffer (50 mm KCl, 10 mm Tris HCl pH 8.3, 2.5 mm MgCl2,0.45% NP-40, 0.45% Tween-20, 300 mg/ml proteinase K), andincubated at 55� for 2–3 hr to produce crude worm lysates.High molecular weight DNA preparation, DNA fragmenta-tion, DNA labeling, sample hybridization, image acquisition,and determination of fluorescence were all performed aspreviously described (Maydan et al. 2007). Each step, fromDNA fragmentation to raw signal detection, was done byRoche NimbleGen.

Microarray manufacture and data normalization: Themicroarrays were manufactured by Roche NimbleGen, witholigonucleotides synthesized at random positions on thearrays. For all the experiments, normalization of fluorescenceintensity ratios was performed with a LOESS regression aspreviously described (Maydan et al. 2007).

Design of preliminary microarray: Two thousand sixhundred thirty-nine SNPs without nearby mutations in thestrain CB4856 were selected from WormBase data freezeWS170. Alternating between the plus and minus strandtemplates, up to 150 distinct 50-mer oligonucleotides weredesigned to densely tile a 200-base window centered on eachSNP. Probes including known repeats were excluded. Dye-flipexperiments were performed and the results averaged aftercorrecting for the change in sign of the log2 ratio.

Design of the mapping microarray and data analysis: Themapping microarray was designed to simultaneously probe4576 small mutations (3169 single nucleotide substitutions,567 insertions, 840 deletions) distributed across the CB4856genome and listed as ‘‘SNP’’ in WormBase data freeze versionWS180. For each mutation, 50-mer oligonucleotide probesaffected by the mutation were designed with the CB4856 andN2 sequences. Those probes were designed to follow eitherthe plus or minus strand template to keep the mutation as faraway as possible from the tethered end of the probe knowingthat with NimbleGen’s manufacturing process the oligonu-cleotides are synthesized from 39 to 59. Control 50-mer probesalternating between the plus and minus strands were designedto tile the immediate flanking regions 100 bases upstream anddownstream of each mutation. To avoid significant cross-hybridization, probes containing repeats annotated in Worm-Base have been excluded together with probes showingsignificant homology (.75% identity over the whole probeaccording to a MegaBLAST search) (Zhang et al. 2000) withother locations in the C. elegans genome. This resulted in atotal of 380,058 probes with 187,335 of those being controlprobes targeting flanking regions. The fit of the mappingsignal described in the text was performed with the cubicsmoothing spline function provided in the standard distribu-tion of the R statistical software.

The list of oligonucleotides used for mapping the geneticmarkers in these experiments is available from Roche Nim-bleGen as design name 071128_CE_HawaiianMap_DM_tiling.Note that we are currently refining the chip design and a newlist of oligonucleotides will be available in the near future. Weare also in the process of creating a user-friendly analysispackage. New chip designs as well as the software necessary toanalyze the data will be posted at http://www.zoology.ubc.ca/%7Edgmweb/research1_cgh.htm. If you cannot find what youneed on the Web site you can contact us directly (sflibotte@bcgsc.ca or moerman@zoology.ubc.ca).

RESULTS AND DISCUSSION

In a preliminary experiment, 2639 previously knownsingle nucleotide substitutions between the CB4856 and

N2 strains were characterized by array CGH using 50-mer oligonucleotide probes. The object of this experi-ment was to determine which SNPs could be detectedusing hybridization and would therefore serve as usefulmapping markers. Besides accruing a set of useful SNPswe also observed that the magnitude of the perturbationdue to a single base substitution on the hybridizationsignal is a function of the position of the substitutionwithin individual probes (Figure 1). As expected, theperturbation is smaller when the substitution is locatedclose to the surface of the slide as opposed to the end ofthe probe, which is freely floating in the solution. For 50-mer probes, the effect is maximal when the substitutionis near the seventh nucleotide from the unattached end,and moving the substitution to the center of the probereduces the magnitude of the effect by nearly a factor of2. In the manufacturing process the probes are synthe-sized onto the microarray from the 39 end to the 59 endwith the 39 end of the oligonucleotide being closer to thesurface of the glass slide. Consequently, when denselytiling a small region of a chromosomal sequence fromleft to right one can maximize the effect of a substitutionby first designing probes on the minus strand templateand then switching to the plus strand template when the

Figure 1.—Effect of the position of a single nucleotide sub-stitution within oligonucleotide probes in a comparative ge-nomic hybridization experiment. The figure shows theaverage comparative genomic hybridization profile for 2639single nucleotide substitutions between the CB4856 and theN2 strain of C. elegans. More specifically, the average of log2

(fluorescence intensity CB4856/fluorescence intensityN2) isshown as a function of the chromosomal coordinate of the leftend of each 50-mer probe relative to the coordinate of thesubstitution. Probes following the plus and minus strand tem-plates of the N2 Bristol sequence are represented by red andblue circles, respectively. Probes with relative positions , �50or .0 are in the immediate flanking region of a substitutionbut do not overlap known substitutions. Since the probes arealways synthesized from the 39 to the 59 end on the microarray,the probes following the plus and minus strand sequencesshow nearly identical hybridization patterns with a maximumperturbation when the substitution is located near the sev-enth nucleotide from the end away from the glass slide andfreely floating in solution.

34 S. Flibotte et al.

substitution is near the center of the probe; in otherwords, the substitution should be kept far from the sur-face of the slide to maximize its perturbation on thehybridization signal.

Armed with this critical piece of information, we nextdesigned a microarray to simultaneously probe SNPsand indels (insertions and deletions) distributed acrossthe CB4856 genome. Indels did not behave as expectedand so were excluded from further analysis. We tested3169 SNPs with mean spacing of 15 kb and medianspacing of 31 kb, avoiding both annotated repeats andprobes with significant homology elsewhere in thegenome. The design sought to incorporate 22 50-meroligonucleotide probes of each SNP for each strain.Elimination of repeats and remote homology reducedthe number of probes per SNP somewhat, but .85%had the full complement of 44 probes. Subsequentanalysis of the information content suggests that onlyhalf as many probes per SNP would suffice. We alsoincluded probes from flanking regions as controls in thearray design, but later found these were unnecessary forsuccessful mapping. The initial chip design used inthese experiments had �380,000 probes. By excludingindels, flanking probes and half the oligos covering aSNP, this number could be reduced to 76,000 to obtainthe results presented here.

We tested the method using mutants from near thecenter of each of the 6 chromosomes (except on the X,where a distal marker was used), where recombination islow and resolution inherently most limited. Mappingcrosses were performed as indicated in Figure 2 (seelegend to Figure 2 for details on the crosses and markersutilized). After homozygous mutant animals were rese-gregated from F1’s derived from the cross, their progenywere grown together as a mixed population on a singlelarge plate. Note that there is but a single cross and asingle DNA preparation in this procedure. The map-ping analysis procedure for data from the arrays involvesfirst normalizing the log2 ratios, then for each sub-stitution represented on the array to calculate themedian log2 ratio separately for the probes with N2and CB4856 sequences. The difference between thosetwo median values is then used as the signal to map thelocation of the mutation of interest. Near the site of themutation the alleles in the sample should be derivedfrom N2 and at unlinked positions the two alleles shouldbe equally represented because of recombination andassuming the absence of selection (but see Seidel et al.2008). Using cy3 to label the sample DNA and cy5 tolabel N2 control DNA, near the mutation all the probesshould show a ratio close to 1 (log2 ratio ¼ 0), but inunlinked, unselected regions the N2 probes shouldhave a ratio of ,1 (log2 ratio , 0) and the CB4856probes should have a ratio of .1 (log2 ratio . 0).

We used knowledge of the mutation site only for thefirst experiment [dpy-5(e61) on chromosome I] to de-velop a satisfactory analysis procedure, and data for the

other five experiments were analyzed blindly. Figure 3shows whole-genome views of this mapping signal for thesix experiments. Even though the mapping signal isfairly noisy, a simple visual inspection of the genomeview can determine on which chromosome each muta-tion is located (note that individual chromosomes arecolor coded); an automated procedure looking for thechromosome with the largest mean or median mappingsignal would of course find the proper chromosome. Ascan be seen in Figure 4, fitting a cubic smoothing splineto the mapping signal of the affected chromosomeallows a finer mapping of the mutation with surprisinglygood accuracy. Other smoothing techniques such as theLOESS local regression and Friedman’s SuperSmoothercould also be used. In the six cases, the mapping signalmaximum fell between 40 and 427 kb of the position ofthe mutation with an average of 182 kb and a median of171 kb.

Figure 2.—Schematic of genetic crosses and downstream pro-cessing required to perform SNP-CGH mapping. For this analysiswe chose six marker mutations of known physical location, onefor each chromosome. These were dpy-5(e61) I (F27C1.8), dpy-10(e128) II (T14B4.7), dpy-17(e164) III (F54D8.1), unc-22(e66)IV (ZK617.1), dpy-11(e224) V (F46E10.9), and unc-2(ra612) X(T02C5.5). Details of crosses, DNA preparations, microarray hy-bridization, and analysis are described in the materials and

methods (also see Maydan et al. 2007).

SNP–CGH Mapping 35

The procedural ease and mapping resolutionachieved from a single genetic cross and a single DNApreparation using this SNP-CGH mapping protocolshould increase the speed of mapping new mutationsafter forward genetic screens as well as enhancer andsuppressor screens. This protocol should also make it

much less difficult to map complex or subtle behavioralphenotypes as well as phenotypes with low penetrance,as only a handful of animals with the confirmedphenotype need be isolated and grown after the initialcross. This method should also prove extremely usefulfor experiments utilizing worms to study natural varia-tion in genetic pathways (in methods such as ‘‘geneticalgenomics’’; reviewed in Kammenga et al. 2008). This is ageneral method that can be applied to other C. elegansstrains as well as other nematode species (for example,C. briggsae Stein et al. 2003; Hillier et al. 2007). Therecent sequencing of the Pasadena C. elegans strainCB4858 (Hillier et al. 2008) was initially thought tooffer a possible alternative mapping strain, but CB4858has large blocks where there are few variants so it may notbe the best choice. Other wild-type strains may have to beinvestigated. The SNP platform we describe here can betransferred to any organism where a pair of interbreed-ing strains exists with significant variation at the nucle-otide level.

What is not yet clear is the limit to the resolution thatmay yet be achievable with this system. Even as it standsnow the method offers relatively high resolution forminimum effort. We see two routes to increasing map-ping resolution, increasing the number of recombinantsand increasing SNP density on the chip. For three of oursix hybridizations, DNA from 100 mutant segregantsprovided resolution to individual SNPs, so increasingthe number of segregants alone would not confer anincrease in mapping resolution for many mutants. Asdetailed earlier the density of substitutions representedon the array could easily be increased by a factor of 5 infuture microarray designs without increasing the totalnumber of probes on the array. This increase in the

Figure 3.—Whole genome views of themapping signal for six previously knownmutations. For each single nucleotide sub-stitution represented on the microarraythe difference between the median log2 ra-tio for probes with N2 and CB4856 sequen-ces is plotted by increasing chromosomenumber and location from left to right.The mapping signal for each chromosomeis shown with data points of different color(red for I, yellow for II, green for III, cyanfor IV, blue for V, and magenta for X). Theknown location of each mutation is (a)chromosome I coordinate 5,432,448 withclosest SNP index of 140; (b) chromosomeII coordinate 6,712,800 with closest SNPindex of 606; (c) chromosome III coordi-nate 5,107,908 with closest SNP index of1057; (d) chromosome IV coordinate12,011,000 with closest SNP index of1590; (e) chromosome V coordinate6,512,776 with closest SNP index of 2043;and finally (f) chromosome X coordinate2,717,348 with closest SNP index of 2629.

Figure 4.—Views of the mapping signal for a mutation lo-cated on chromosome I. The mapping signal is plotted withopen black circles for each substitution represented on themicroarray as a function of chromosome I coordinate. Theblue lines and blue squares represent a cubic smoothingspline fit to the mapping signal and the vertical red lines in-dicate the actual location of the mutation being mapped. Thewhole chromosome view shown in a and b focuses on a smallinterval around the mutation. The maximum of the fit can betaken as our best estimate of the location of the mutation withthe current mapping array. Since our prediction is just onedata point away from the real position of the mutation, weshould be able to improve the mapping resolution by increas-ing the number of substitutions represented on the array.

36 S. Flibotte et al.

number of SNPs coupled to analyzing more animalscould potentially improve the mapping resolutionsignificantly.

The conservation of many genetic pathways betweennematodes and humans makes the study of single-locusmutations in C. elegans an important source and inroadinto the details of both known and novel biologicalpathways. The development of this new techniqueshould quicken the pace of identifying novel mutationsaffecting these various genetic pathways.

We thank Aleksandra Rogula for providing Figure 2. Some nema-tode strains used in this work were provided by the CaenorhabditisGenetics Center, which is funded by the National Institutes of HealthNational Center for Research Resources. This work was supported bygrants from Genome Canada, Genome British Columbia, and theMichael Smith Research Foundation (to S.F. and D.G.M.).

LITERATURE CITED

Carlson, C. S., M. A. Eberle, M. J. Rieder, Q. Yi, L. Kruglyak et al.,2004 Selecting a maximally informative set of single-nucleotidepolymorphisms for association analyses using linkage disequilib-rium. Am. J. Hum. Genet. 74: 106–120.

Chen, D., A. Ahlford, F. Schnorrer, I. Kalchhauser, M. Fellner

et al., 2008 High-resolution, high-throughput SNP mapping inDrosophila melanogaster. Nat. Methods 5: 323–329.

Davis, M. W., M. Hammerlund, T. Harrach, P. Hullett, S. Olsen

et al., 2005 Rapid single nucleotide polymorphism mapping inC. elegans. BMC Genomics 6: 118.

Hillier, L. W., R. D. Miller, S. E. Baird, A. Chinwalla, L. A. Fulton

et al., 2007 Comparison of C. elegans and C. briggsae genomesequences reveals extensive conservation of chromosome organiza-tion and synteny. PLoS Biol. 5: e167.

Hillier, L.W., G. T. Marth, A. R. Quinlin, D. Dooling, G. Fewell

et al., 2008 Whole-genome sequencing and variant discovery inC. elegans. Nat. Methods 5: 183–188.

Kammenga, J. E., P. C. Phillips, M. De Bono and A. Doroszuk,2008 Beyond induced mutants: using worms to study naturalvariation in genetic pathways. Trends Genet. 24: 178–185.

Maydan, J. S., S. Flibotte, M. L. Edgley, J. Lau, R. R. Selzer et al.,2007 Efficient high-resolution deletion discovery in Caenorhab-ditis elegans by array comparative genomic hybridization. GenomeRes. 17: 337–347.

Seidel, H. S., M. V. Rockman and L. Kruglyak, 2008 Widespreadgenetic incompatibility in C. elegans maintained by balancing se-lection. Science 319: 589–594.

Stein, L. D., Z. Bao, D. Blasiar, T. Blumental, M. R. Brent et al.,2003 The genome sequence of Caenorhabditis briggsae: a plat-form for comparative genomics. PLoS Biol. 1: E45.

Swan, K. A., D. E. Curtis, K. B. McKusick, A. V. Voinov, F. A. Mapa

et al., 2002 High-throughput gene mapping in Caenorhabditiselegans. Genome Res. 12: 1100–1105.

van Eijk, M. J., J. L. Broekhof, H. J. van der Poel, R. C. Hogers, H.Schneiders et al., 2004 SNPWave: a flexible multiplexed SNPgenotyping technology. Nucleic Acids Res. 32: e47.

Wicks, S. R., R. T. Yeh, W. R. Gish, R. H. Waterston and R. H.Plasterk, 2001 Rapid gene mapping in Caenorhabditis elegansusing a high density polymorphism map. Nat. Genet. 28: 160–164.

Zhang, Z., S. Schwartz, L. Wagner and W. Miller, 2000 A greedyalgorithm for aligning DNA sequences. J. Comput. Biol. 7: 203–214.

Communicating editor: R. Scott Hawley

SNP–CGH Mapping 37