markup_maintext for pdf - Genes | Genomes | Genetics

G3: Genes|Genomes|Genetics Early Online, published on March 3, 2017 as doi:10.1534/g3.117.039883
1
AutomatedphenotypingindicatespupalsizeinDrosophilaisahighlyheritabletrait
2
withanapparentpolygenicbasis
3
4
R.GuyReevesandDiethardTautz
5
6
MaxPlanckInstituteforEvolutionaryBiology,Plön,Germany,
7
1
© The Author(s) 2013. Published by the Genetics Society of America.
8
SHORTRUNNINGTITLE
9
Pupalsizephenotypingandgenetics
10
11
KEYWORDS
12
Automatedphenotyping,Drosophilamelanogaster,pupallength,humanheight,complex
13
trait
14
15
Correspondingauthorscontactdetails
16
GuyReeves,DepartmentofEvolutionaryGenetics,MaxPlanckInstituteforEvolutionary
17
Biology,August-Thienemannstrasse2,24306Plön(Germany)
18
[email protected]
19
office+494522763376
20
lab+494522763296
21
22
2
23
24
ABSTRACT
Theintensefocusonstudyinghumanheighthasdonemorethananyothergeneticanalysis
25
toadvanceourunderstandingoftheheritabilityofhighlycomplexphenotypes.Herewe
26
describeindetailthepropertiesofapreviouslyunexploredtraitinDrosophila
27
melanogasterthatsharesmanysalientpropertieswithhumanheight.Thetotallengthof
28
thepupalcasevariesbetween2.8and3.9mmamongnaturalvariantsandwereportthatit
29
isamongthemostheritabletraitsreportedinthisspecies.Wehavedevelopedasimple
30
semi-automaticphenotypingsystemwithwhichasingleoperatorcanreliablyscore>5000
31
individualsinaday.Theprecisionoftheautomatedsystemis0.042mm(±0.030SD).All
32
phenotypedindividualsareavailabletobematedinsubsequentgenerationsoruniquely
33
archivedforfuturemolecularwork.Wereportbothbroadsenseandnarrowsense
34
heritabilityestimatesfortwobiologicallydistinctdatasets.Narrowsenseheritability(h2)
35
rangedfrom0.44-0.50andbroadsenseheritability(H2)rangedfrom0.58to0.61.We
36
presentresultsformappingthetraitin195recombinantinbredlines,whichsuggeststhat
37
therearenolociwith>10%effectsizeinthispanel.Weproposethatpupalsizegeneticsin
38
Drosophilacouldrepresentamodelcomplextraitamenabletodeepgeneticdissection
39
usingtheautomatedsystemdescribed.
40
3
41
INTRODUCTION
42
Elucidatingthegeneticarchitectureofaphenotypictraitfundamentallyrequiresthatit
43
istosomedegreeheritable.Phenotypictraitswithlowheritabilitydonotgenerallyrequire
44
additionalexplanationastowhyquantitativetraitlocicannotberobustlyidentified.
45
However,whiletraitswithhighheritabilityincreasethepowertoidentifycausativeloci,
46
thegeneticarchitectureoftraitsremainsakeyfactorinpredictingsuccessinidentifying
47
quantitativetraitloci.Forexample,ahighlyheritabletraitthatisdependentonalarge
48
numberofinteractingallelesorlocimaystillrequiresubstantialsamplesizesandcomplex
49
analysestoidentifycausativeloci(Visscheretal.2010;Bartonetal.2016).The
50
archetypicalexampleofthisishumanheightwhichbyatleasttwodistinctmeasuresof
51
heritabilityisamongthemostheritableofquantitativehumantraits,h2≈0.4-0.7using
52
regressionoftrios(Galton1886;Hanley2004;Visscheretal.2010)orh2≈0.7usingtwin
53
studies(Poldermanetal.2015).Howevertheidentificationofthelociinvolved,andany
54
interactionsamongthem,hasprovenfarfromtrivialandsuccessestodatehavebeen
55
dependentonutilizingverylargedatasets(upto250,000individuals(Woodetal.2014)).
56
Thecurrentunderstandingbeingthatofthe180-4,000lociimplicatedinimpactinghuman
57
heightoveritstypicalrangeallhaveadditiveeffectsizesofmuchlessthan1mm(Weedon
58
etal.2008;Visscheretal.2010;Woodetal.2014).Althoughstudiesofhumanheight
59
representavaluablemodelwithwhichtotesthowcomplexheritablephenotypescanbe
60
dissectedtheabilitytoexperimentallycontrolallelefrequenciesandenvironmental
61
conditionswouldbeavaluablecapacitythatisnotpossibleinhumanstudies.Whilethisis
62
apossibilityinmodelorganismstherearerelativelyfewmorphologicaltraitsthathave
4
63
beenrobustlydemonstratedtobebothhighlyheritableandamenabletoautomatedhigh
64
throughputphenotyping(thoughseebelow).Hereinwedescribeindetailanew
65
quantitativetrait–lengthofthepupalcase-inDrosophilamelanogasterthatplacesitinthe
66
15-20%mostheritablemorphologicaltraitsdescribedinthisspeciesandisamenableto
67
reliableandautomatedhigh-throughputphenotyping(RoffandMousseau1987).
68
Theutilityofautomatedphenotypingsystemsforawidevarietyoforganismshas
69
increasedgreatlyoverthelast10years.Whiletherehavebeenautomatedmorphological
70
phenotypingsystemsdevelopedforDrosophila,asyetonlyautomationanalysisofimages
71
ofwingshasbeenwidelyemployed(Houleetal.2003).However,acquiringthewing
72
imagesrequiresmanualmanipulationtopositionfliesindividuallyandsoisdifficultto
73
scaleup.Likewiseasystemtomeasureheartbeatfunctionrequiresthateachflybe
74
manipulatedintoposition.Analternativeapproachtophenotypinginaselection
75
experimentforgrossbodysizewasachievedusingaseriesofgraduatedsieves(Turneret
76
al.2011).Morerecentlyasophisticatedplatformcalledthe‘flycatwalk’wasdescribedthat
77
hasthecapacitytoreliablyphenotype700fliesadayforawiderangeofmorphological
78
traitssimultaneously(Medicietal.2015).Thisrequiresnousermanipulationofindividual
79
fliesandisnon-destructive.Whilethecostofconstructingtheequipmentisnotdetailed
80
(https://github.com/IMSB/FlyCatwalk/)itislikelythatitwouldrepresentasignificant
81
investmentoftime,expertiseandresourcesformostlabs.
82
Oursetupusesaninexpensivecamerainalightproofboxandtheopensourceimage
83
analysissoftwareCellprofilerwithwhichasingleusercanphenotype5,000pupaeinaday.
5
84
Pupaearephotographedinsituonflattenedsquaresoftransparentfilmthatlinedthe
85
entireverticalsurfaceofthevials.Wedemonstratehowthesystemcanbeusedto(1)
86
measurepanelsofrecombinantinbredlines(RILs)and(2)generatelargenumbersof
87
parentoffspringtriosthatareparticularlyusefulinexploringcomplextraitsusingartificial
88
selectiontechniques.Furthermore,theincreasedthroughputfacilitatesexploringthe
89
heritabilityoffamilymeansratherthansingleindividualmeasurementsthatareassociated
90
withincreasedmeasurementvariance.Hence,pupalsizecouldbecomeamodelphenotype
91
thatwillallowdeepdissectionofitsgeneticarchitecture,sincetheavailabilityofautomated
92
phenotypingwillallowtoscreenverylargemappingpanelsortodesignnewcomplex
93
mappingstrategies.
94
95
MATERIALSANDMETHODS
96
97
Flystocks
98
Theautomatedphenotypingsystemwasappliedtotwoindependentdatasets.
99
Thefirstdataset,referredtohereas‘8-way’isacollectionof195recombinantinbred
100
lines(RILs)thatarepartoftheDrosophilaSyntheticPopulationResource
101
(http://wfitch.bio.uci.edu/~dspr/).TheseRILsarealloriginallyderivedfromacross
102
between8globalstocksofDrosophilamelanogaster,theirgenerationisdescribedindetail
103
in(Kingetal.2012a;b).Narrowsenseheritabilityh2forthe8-waydatasetwasestimated
104
bymeasuringtheprogenyof67singlepairmatingsbetweenindividualsfrom11different
6
105
RILsthatspannedthefullrangeofphenotypemeasurements(IDsofcrossedstocksare
106
giveninFileS5,6ofthe67crossesareduplicatesalsoindicatedinFileS5).Broadsense
107
heritabilityestimatesH2ofthe8-waydatasetweregeneratedbyrepeatedmeasurementsof
108
RILs(IDsgiveninFileS5).
109
Theseconddatasettermed‘4-way’wasinitiatedbyacrossbetweentwoJapaneseand
110
twoAfricanstocks(SeetableS1).Narrowsenseheritabilityh2wasestimatedbymeasuring
111
theprogenyofsinglepairmatingsfromthe2ndto6thgenerationwherephenotyped
112
parentswererandomlyselectedfromdifferentvialstoformsubsequentgenerations.The
113
followingnumberofsinglepairsweremeasuredineachgeneration[G2,15pairs],[G3,81
114
pairs],[G4,78pairs],[G5,88pairs],[G6,154pairs].Noduplicatevialsweregeneratedfrom
115
thesamematingpair.Notethatunlikethe8-waydataset,levelsofheterozygosityinthe
116
parentsandtheiroffspringarelikelytobesimilar.BroadsenseheritabilityestimatesH2of
117
the4-waydatasetweregeneratedbymeasuring83RILsestablishedfromthe4-waycross
118
bygeneration37(seeFileS5forstockIDs).Detailsofallprogenitorstocksofthe4-wayand
119
8-waydatasetswithestimatesoftheirpupallengtharegiveninTableS1.
120
121
Imageacquisition
122
Fliesweremaintainedonstandardfooddispensedinto28.5mmdiameter,95mm
123
heightvials(GeneseeScientific).Oncethefoodvialshadfullycooled,10cmx10.5cm
124
squaresofoverheadprojectorfilmwereslidintoeachvialliningtheirentireverticalwall
125
(nobo,plainpapercopierfilm,33638237).Amoredetaileddescriptionoftheentire
126
procedureandequipmentsetupisprovidedasFileS1.Acustomprintedsemi-transparent
7
127
labelincludingauniquebarcodewasaffixedtotheoutsideofeachvial.Vialswere
128
incubatedat24°ina12hourlight12hourdarkincubator.Adultswhereremovedafter1-2
129
nightsinvials(sometimes3to4nights,iffertilityappearedtobeloworduetoholidays).
130
Generallybythe10thdayaftertheparentswereinitiallyintroduced,themajorityof
131
offspringinthevialswerepresentaspupaeattachedtothetransparentfilm,fewifany
132
larvaeremainedinthefood.Thefilmfromeachvialwasremovedandplacedintoa
133
purpose-madeplasticframe(thisframecanbe3DprintedusingafileprovidedasFileS2)
134
thatholdsthefilmflatforphotographing.Foodfromthelowerpartofthefilmwasscraped
135
awayandanylarvaeoranywhitepupariumstage(P1)removed.Theframewasthen
136
photographedusingbottomilluminationinalighttightbox.Batchesoftheresultingimages
137
werethenanalyzedusingtheprocedurebelow.
138
139
Automatedimageanalysis
140
ACellprofiler(v2.1.0)pipelinewasdevelopedtosimultaneouslyrecognizepupaeandto
141
measureavarietyofattributesincludinglength.Theoutputsofallmeasurementsare
142
writtento.xlsfilesthatcanthenbeviewedorimportedintoanydatabaseprogram.
143
Cellprofilerisfreeopenaccesssoftwareprovidingasuiteofflexibleimageanalysistools
144
(Lamprechtetal.2007).InbrieftheCellprofilerpipelinefirstidentifies‘primaryobjects’
145
distinctfromthebackgroundwithoutrestrictionontheirsize(module:identifyprimary
146
objects).Thenapplyingascalablemodelofpupalshapetoallobjectsthosethatare
147
composedofmultipletouchingpupaeareseparatedintodistinctpupae(module:Untangle
148
Worms,(Wählbyetal.2012)).Theresultingputativepupaearetheneachshrunkandthen
8
149
re-propagatedoutwardstomorepreciselyidentifytheedgesofeachpupabasedon
150
boundarychangesinpixelintensity(module:IdentifySecondaryobjects).Finally,pupae
151
arecrudelyfilteredonsizeattributesandtheproximityofneighboringobjectstoplace
152
theminoneofthethreeconfidenceclassesdescribedintheresults.Thedigitaloutlinesof
153
pupaeareoverlaidontoacroppedversionoftheoriginalimagetoallowuserstoeasily
154
visuallyassessquality.
155
Overlaidimagesandfilesofmeasurementswereimportedinbatchesintoadatabase
156
programFileMaker(v14,FileMakerInc.).Auniquebarcodestickeridentifyingthefilmin
157
theimagewasautomaticallyreadbythedatabase.Thisenabledtheimagefilestobe
158
automaticallyrenamedwiththebarcodeastheirnameforarchiving.Inadditiontopupal
159
measurementsa1€centcoin(16.25mmdiameter)presentinallimageswasmeasuredto
160
controlforcamerachangesandallowconversionofmeasurementinpixelstomm.Quality
161
filteringofpupaeandbasicanalyseswereperformedusingthedatabase.Anannotated
162
copyoftheCellprofilerpipelineisprovidedasFileS3.
163
164
Estimatesofheritability
165
BroadsenseheritabilityH2wasestimatedusingSPSSversion22withthe‘variance
166
components’function,withpupallengthasthedependentvariable,andRILnameasa
167
randomfactor.ThiswasdoneusingtheMinimumNormQuadraticUnbiasedmethod
168
(thoughANOVAproducedidenticalresultsto2decimalplaces).Themodelforthesingle
169
factorwasyij=μ+αj+eij.WhereyijisthejthobservationoftheithRIL,μistheoverallmean,
170
αjistherandomfactorandeistheassociatederror.
9
171
Narrowsenseheritabilityh2estimatesweredoneusing‘linearregression’function,with
172
pupallengthasthedependentvariableandparent-midpointsasanindependentvariable.
173
AllotherstatisticalanalysesandgraphswerealsoperformedusingSPSSorFilemaker.All
174
dataforindividualpupaareavailableasFileS4andforvialmeansasFileS5(fieldsusedin
175
eachgraphareindicatedintheassociated‘readme’files).
176
177
Humanheightdata
178
Estimatesofh2forhumanheightdatawasdoneusingthesamemethodsasforpupaewith
179
theoriginaldataofFrancisGaltonfor898individuals,transcribedfromhis1880s
180
laboratorynotebooks((Hanley2004),available
181
http://www.math.uah.edu/stat/data/Galton.html).The760individualsin123families
182
wherefourormoreindividualsweremeasuredwereusedforestimatesoftheh2
183
heritabilityofmeanfamilyheight(femaleswerenottransmutedintomales).
184
185
Genomescanforlocioflargeeffectimpactingpupallengthusing8-wayRILs
186
ScanswereperformedinR(version3.2.2)forthe195DSPRRILsmentionedabove(Kinget
187
al.2012a;b)usingthe‘DSPRscan’commandwithintheDSPRqtl-toolsprocedures
188
describedin(Kingetal.2012a;b).Theseutilize1SNPper10kbthroughoutthegenome
189
reference(withtheexceptionoftheYchromosomeandthemitochondrialgenome).
190
SignificanceLODscorethresholdswereestimatedusingthe‘DSPRperm’commandwith
10
191
1,000replicates.ThedatainputfileisavailableasFileS6.Thisscanwasundertakento
192
asseswhetheroneorfewlocioflargeeffectcontrolpupallengthi.e.thetraitisnot
193
complex.Notethatifthetraitiscontrolledbymanylociofsmallormoderateeffectsize,the
194
scanistoounderpoweredtorealisticallyidentifyindividualloci.Butthegoalofthescanis
195
toshowthatpupalsizeisnotarelativelysimpletraitcontrolledbyonlyafewmajoreffect
196
loci.
197
Dataavailability
198
AllstartingflystocksareavailablefromthesourcesdetailedinTableS1.RILsareavailable
199
fromthewebaddressgivenabove.FilesofrawdataareavailableasFilesS4,S5andS6.
200
201
RESULTS
202
Thelargenumbersofindividualsandvialsmeasuredmadepossiblethroughtheuseof
203
theautomatedmeasuringsystemprovidesrobustinsightsintohowphenotypicvariationis
204
partitionedforpupallength.Whilethemethodautomaticallyidentifiesandmeasures
205
pupaeformultipledifferentparameters(seeFileS1forafulllist),thelengthofthepupaeis
206
theprinciplefocusofthisstudy.Twodistinctsetsofbiologicalmaterialwereusedtodefine
207
thebiologicalpropertiesofpupallength.Thefirstdatasetisallderivedfroman8-waycross
208
(Kingetal.2012a;b).Thesecondisfroma4-waycross.Noneofthefoundingstocksare
209
sharedbetweenthe8-wayandthe4-waydatasets.BroadsenseheritabilityH2(reflecting
210
allpotentiallygeneticcontributionstotraitvariancee.g.,additive,epistatic,dominance,
211
maternalandpaternaleffects)wasestimatedthroughtheanalysisofrepeated
212
measurementsofrecombinantinbredlines(RILs).Narrowsenseh2heritability(reflecting
11
213
onlytheimpactofadditivegeneticeffects)wasmeasuredbyregressionofmid-parent
214
againsttheiroffspringmeasurements.Samplingpropertiesofthe8-wayand4-way
215
datasetsaredetailedinTable1andFigure1(seealsoFigureS1forsummaryofrepeat
216
measurementsofRILs).
217
Dataset Heritability Vial
Vialsmeasured=1184,RILs=195
8-way
H2
Individuals
n=83,402,
meanreplicatesperRIL=6.7(±2.4SD,
meanpupallength=3.2(±0.21SD)
min=4,max=15))
meanvialpupallength=3.2(±0.15SD)
rangeof95%ofpupallengths=2.8-3.5mm
meannumberofmeasuredpupaepervial=70
(±34SD)
crossesmeasured=67
n=3,113,
h2
meanvialpupallength=3.5(±0.11SD)
meanpupallength=3.4(±018SD)
rangeof95%ofpupallengths=3.1-3.5mm
meannumberofmeasuredpupaepervial=47
(±15SD)
Vialsmeasured=436,RILs=81
n=25,356,
4-way
H2
meanreplicatesperRIL=5.54(±0.78SD,
meanpupallength=3.6(±0.24SD)
min=3,max=6))
meanvialpupallength=3.6(±0.15SD)
rangeof95%ofpupallengths=3.4-4.0mm
meannumberofmeasuredpupaepervial=59
(±24SD)
crossesmeasured=363
n=22,487,
h2
meanvialpupallength=3.4(±0.14SD)
meanpupallength=3.4(±0.23SD)
rangeof95%ofpupallengths=3.1-3.7mm
meannumberofmeasuredpupaepervial=62
(±17SD)
Table1Summaryofsamplingof8-wayand4waydatasets.Onlyvialswhere≥15pupaeweremeasuredbythe
automatedsystemareconsidered.ForH2estimatesonlyRILswhere≥3replicatemeasurementswereavailableare
considered.
218
219
220
221
222
12
223
A
5,000.0
Frequency
4,000.0
3,000.0
2,000.0
1,000.0
0.0
2.0
2.5
3.5
4.0
4.5
pupal length (mm)
224
225
3.0
B
400.0
Frequency
300.0
Page 1
200.0
100.0
0.0
2.0
2.5
3.0
3.5
4.0
4.5
vial average pupal length (mm)
226
227
228
229
230
231
232
Figure1.Samplingdistributionsofpupallengthforindividualsandvials.A,distributionsforindividualpupal
lengths,4-waydataset(blue)n=47,843,8-waydataset(green)n=86,515.B,Correspondingdistributionsforvialmean
pupallengths,4-waydataset(blue)n=799vials,8-waydataset(green)n=1251vials.Onlyvialswhere≥15pupaewere
measuredbytheautomatedsystemareconsidered.AstockpossessingtheTb1mutationresultinginthewellknown
tubbypupalphenotype(Bloomingtonstock3644)wasmeasuredandfoundtohaveameanpupallengthof2.7mm,which
isatthelowerboundsofthesmallestwildtypeindividualsorvialsshownhere.
13
Page 1
233
234
Performanceofautomatedpupalphenotyping.
235
Theautomatedpupalrecognitionpipelineattemptstoidentifytheexternaloutlinesof
236
threeclassesofobjects.Highconfidencepupae:conformtoasetofexpectedpropertiesand
237
thathavenoneighboringpupaewithin20pixels(≈0.7mm,Cellprofilermodule:filter
238
objects).Mediumconfidencepupae:conformtothesamesetofpropertiesbutwithcloser
239
than20pixelneighboringobjects.Thisoftenoccursaslarvaeselectpupationsitestouching
240
eachother,resultinginamorechallengingtargetforimagerecognition.Lowqualityobjects:
241
unlikelytobepupae.ExamplesofthethreeclassesareshowninFigure2,(seeFileS1for
242
furtherdetails).
243
14
244
245
246
247
248
249
250
251
252
Ofthepupaeretainedforanalysis47%and53%werehighandmediumconfidence
253
respectively.Theprobabilityofpupaebeingmanuallyexcludedasaberrantlymeasured
254
was0.02forhighqualitypupaeand0.10formediumconfidencepupae(Figure2).Note
255
thatnotallimagesweremanuallyexamined,assometimesduetotherobustnessofthe
256
automatedestimatesgenerallyonlyvialswithatypicallyhighvarianceofthemeanwere
257
examined.
258
Theprecisionoftheautomatedsystemwasassessedbyre-measuringasubsetoffilmsafter
259
rotatingthemby180°andcomparing516duplicatedmeasurementsofthesamepupae
260
(Figure3).Thedifferencebetweenthetwomeasurementswasanaverageof0.043mm
261
(±0.030SD).Re-measuringpupaeafterdelaysof15or30hoursalsogeneratedsimilar
262
precisionestimates,indicatingthatoncepupaebecomebrownpuparium(P2)thereisno
263
detectablechangeinpupallength.Furthermoretheexuviaofeclosedindividualscanalso
264
bereliablymeasured(datanotshown).
265
266
Thecountofautomaticallymeasuredpupaeisareasonableproxyfordensitywithinvials.
267
Unsurprisinglynotallpupaewerecorrectlycapturedbytheautomatedsystemandbased
268
onasampleof148vialsanaverage20%±11.7SDpupaepervialweremissed,calledaslow
269
confidenceobjectsormanuallyexcluded(Figures4and2).Whilethereisanincreased
Figure2.Visualoutputofautomatedphenotypingsystem.Blueoutline=highconfidencepupae,redoutline=medium
confidencepupae,yellowoutline=lowqualityobjects.Asmallproportionofpupaeareentirelymissedbytheautomated
system(greenarrow).Othersareaberrantlymeasuredduetocloseproximitytootherpupaeresultingintruncationor
extensionoftheiroutline.Eventhosewithaahighconfidenceassignmentcanalsoappearmalformed(orangearrow).
Theyshouldbemanuallyexcludedfromanalysistoreducenoise(seeFileS1formoreexamplesofaberrantandcanonical
pupae).AllobjectsarenumberedinCellprofileroutputfilestopermitnumericalmeasurementstobeeasilyrelatedto
images.
15
varianceintheproportionofunmeasuredpupaeinhigherdensityvials(e.g.wherethe
271
automatedcountis>80),evenheretheautomatedcountprovidesareasonableproxyfor
272
densityofindividualsineachvial.Thisassumesthatfewifanylarvaeremainedonthefood
273
surfaceafterfilmtransfer,aswasgenerallythecase(dueinparttotheshortperiodparents
274
remainedinthevials).Inadditionthesmallnumberofpupaeremovedfromfilmsthatwere
275
obscuredbylarvalfoodrepresentsaconstantproportionacrossallvials(seeFileS1).
pupal length reverse orientation (mm)
270
4.5
4.0
3.5
3.0
2.5
2.0
276
277
278
279
280
281
2.5
3.0
3.5
4.0
4.5
pupal length normal orientation (mm)
Figure3.Precisionofautomatedpupallengthestimates.Comparisonofpupallengthmeasurementsoffilmsinthe
normalorientationcomparedtowherethesamefilmwasrotatedby180°Allpupaeisolatedfromothersbyupto15
pixelsinanydirectionwereused(n=516pupae,across10vials).Slope=0.94andR2=0.98.Anx=ylineisshownfor
reference.
Page 1
16
estimated vial density (automated
count photographed pupae)
250
200
150
100
50
0
0
50
100
150
200
250
true vial density (manual count of photographed
pupae)
282
283
284
285
286
287
288
Impactofvialdensityonpupallength.
289
OnecommonmajorenvironmentalcovariateofmanyDrosophilatraitsisdensityof
290
individualswithinthevialtheydevelopin.ConsequentlymanyDrosophilaresearchers
291
controlforthisinexperimentsbycollectinglargenumbersofzygotesandplacinga
292
controllednumberineachvial.Thisisafairlylaboriousprocesstoroutinelyperform,
Figure4.Comparisonofautomatedcountofpupaepervialversusmanualcounting
148vialsfromacrosstheentirerangeoftheautomateddensitiesobservedweremanuallyrecounted.Whileathigher
densitiestheaccuracyoftheautomatedcountdecreasesitremainsareasonableproxyforvialdensityacrosstheentire
observedrange.Slope=0.81andR2=0.92.Anx=ylineisshownforreference.
Page 1
293
whichisconsiderablycomplicatedwheresinglepaircrossesarerequired.
294
Heredensitywasonlyindirectlycontrolledthroughlimitingthenumberofparentsused
295
pervialandrestrictingthenumberofnightstheyremainedbeforebeingcleared(generally
296
2nightsforsinglepaircrosses,andonenightforsmallgroups(n=10-20)ofRIL
17
297
individuals,seeFileS1).Throughoutallexperimentstwostockswerecontinuallyre-
298
measuredtoactascontrols(stock335and329,seeTableS1).Thelargenumberofrepeat
299
measurementsofthesetwostocksacrossarangeofdensitiespermitstheexaminationof
300
anyrelationshipbetweendensityandpupallength(Figure5).Furthermore,inthesame
301
wayitisalsopossibletoexploretherelationshipbetweendensityandpupallengthusing
302
therepeated8-wayand4-wayRILmeasurements(seeFigureS2).AllRILsandcontrol
303
stocksexhibitauniformlynegativerelationshipbetweendensityandpupallengththeslope
304
variesfrom-0.0006to-0.0041.Thisobservedvariabilitymayreflecteithervariancein
305
estimatingslopesoralsothatRILsexhibitdifferentreactionnorms
306
.Ifthemodeslopeof-0.002(FigureS2)isusedtocorrectthemeanlengthofvials(or
307
individuals)tothatobservedatthemeanobservedvialdensitythiscanbeachievedwith
308
thefollowingequation.
309
Equation1
310
311
312
313
314
M=meanvialdensityacrosswholeexperiment
315
((D-M)S)+Q=Pupallengthcorrectedforvialdensity
316
317
Applyingthisformulatothe431vialsestablishedassinglepaircrosses>99%ofthem
318
wouldrequireacorrectionoflessthan±0.1mmandofthe1,620vialsestablishedfrom
319
smallgroupsofRILindividuals95%ofthemrequireacorrectionoflessthan±0.118mm
320
(Figure6).Giventhemodestnumberofvialssubjectedtoalargecorrectionand
S=slopeofregressionofdensityagainstmeanviallength(-0.002)
D=automatedestimateofdensityinvialtobecorrected
Q=individuallengthmeasurementorvialmeantobecorrected
18
uncertainlyaboutwhetherthereisatrulyuniversallinearrelationshipbetweendensity
322
andpupallengthnoneofthemeasurementspresentedinthemanuscripthavebeen
323
analyticallycorrectedfordensity.FurthermorewithrespectstorepeatedRIL
324
measurements,reducinganyconfoundingimpactofdensitycanbeachievedbyeither
325
experimentallyincreasingthenumberofreplicatemeasurementsclosertothemean
326
density,oranalyticallybyexcludingorweightingdownvialswithextremedensityvalues.
Pupa length vial average (mm)
321
4.5
4.0
3.5
3.0
2.5
20
40
60
80
100
120
Automated estimate of pupal density (count of
measured pupae in a vial)
327
328
329
330
331
332
0
Figure5.Relationshipbetweenvialdensityandpupallengthforthetwocontrolstocks.Repeatedmeasurementsof
thesamestocks(oneshortonelong)atdifferentdensitiespermitstheslopeandcorrelationoftherelationshiptobe
estimated.Whilethereisahighdegreeofcorrelationforthelongerstock-329(redcircles)R2=0.48p=>0.001the
correlationfortheshorterstock-335(blacktriangles)isminimalR2=0.07p=0.07
19
Page 1
333
A
B
100%
Cumulative percentage
Cumulative percentage
100%
80%
60%
40%
80%
60%
40%
20%
20%
0%
0%
0.00
0.02
0.04
0.06
0.08
0.10
0.12
Estimate of length correction due to vial density (mm)
0.0
0.1
0.2
0.3
0.4
Estimate of length correction due to vial density
(mm)
334
335
336
337
338
339
340
341
342
343
Varianceintheestimatesofvialmeans.
344
AscriptwaswritteninFilemakertoexploretheextenttowhichestimatesofmeanvial
Figure6.Magnitudeofpotentialcorrectionofpupallengthforvialdensity,usingequation1.
Themoreuniformvialdensitiesfromsinglepairmatings(A)resultedinallbutonevialbeingwithin50pupaeofthe
meanobserveddensityof65,correspondingtoamaximumcorrectionfordensityof0.1mm(n=431vials).Theincreased
varianceofthevialdensitiesresultingfromestablishingvialsfromsmallgroupsofRILindividuals(B)ledto
approximately5%ofvialsbeingcorrectedby>0.1mm(n=1620vials).Acalculatedmeandensity(M)of65wasusedfor
allvials.AslopevalueofS=-0.002wasused,ifasteeperslopeisusedthentherewouldbeacorrespondingincreaseinthe
magnitudeofthecorrections.Allvialwithadensityof<15wereexcluded.
Page 1
Page 1
345
pupallengthbasedonsub-samplesofindividualswithinavialdeviatefromthevialmeans
346
basedonallindividuals.Thisprovidesinsightintoatwhatpointvialswithlowdensities
347
generatemeanestimateswithunacceptablyhighvariance.Figure7indicatesthatwhile
348
selectingsingleindividualstoestimatevialmeansunsurprisinglyresultsinahighdegreeof
349
varianceofupto0.4mm.Whiletheestimatesareunbiasedthisislikelytoprove
350
unacceptablyhighgiventhatthetotalobservedrangeofpupallengthsis1.1mmand0.8-
351
0.9mmwithindatasets.However,forrandomsub-samplesizesequaltoadensityof15or
352
greaterthevarianceisgreatlyreducedtolessthan0.07mmfor95%ofvials(Figure7).On
20
thisbasisonlyvialswithmorethat15measuredpupaewereincludedforanalysisor
354
presentationthroughoutthisstudy.
355
Range of 95% of replicate differences
between the mean vial pupal length
estimate based on all individuals
compared to those based on subsubsamples of individuals (mm)
353
0.4
0.3
0.2
0.1
0.0
-0.1
-0.2
-0.3
-0.4
1
6
11 16 21 26 31 36 41 46 51
Size of sub-sample (count of individuals)
356
357
358
359
360
361
362
363
364
TheimpactofusingsmallernumbersofvialstoestimateRILmeanlength
365
InmeasuringpanelsofRILsitisusefulforpracticalreasonstominimizethenumberof
366
replicatevialsmeasuredforeachRILtoestimateanRILmeanwithanacceptabledegreeof
367
variance.Using8RILsandthetwocontrollinesallofwhichweremeasure12ormore
Page 1
Figure7.Deviationofmeanvialpupallengthofrandomsub-samplescomparedtothatbasedonallindividuals
withinavial.
Eachvialwasresampledselectingonlyasub-sampleofthepupallengthmeasurementswithinavialtocalculateasubsamplemeanlengththatwasthensubtractedfromthemeanlengthbasedonallsampledindividuals.Eachvialwasresampled100timespersub-samplesize.Whiskersrepresentrangeof95%ofsub-sampledeviationsandcirclesthemean
deviationof100replicates.Basedon729vialswithdensitiesrangingfrom51-60measuredpupae.
21
368
timesitispossibletoexploretherelationshipbetweenRILmeansandthenumberof
369
replicatevialmeasurementsused.Figure8indicatesthatsixreplicatedmeasurements
370
generallyresultin95%ofRILestimatesbeinglessthat0.1mmdifferentfromthatbasedon
371
largernumbersofreplicatevialmeasurements.Themeannumberofreplicatevial
372
measurementsperRILinthisstudywas6.4±2.1SD(5.8%theminimum4replicates,
373
18.0%=5replicates,61.2%=6replicatesand15%>6replicates).
374
22
375
376
A
Deviation of sub-sample pupal length
mean from RIL mean (mm)
0.30
0.25
329 (n=64)
335 (n=31)
11021 (n=12)
11210 (n=13)
11229 (n=12)
0.20
0.15
0.10
11236 (n=14)
11237 (n=15)
11257 (n=12)
11259 (n=12)
11265 (n=12)
0.05
0.00
-0.05
-0.10
-0.15
-0.20
-0.25
1
3
5
7
9
11 13 15 17 19 21 23 25 27 29 31 37 43 49 55 61 63
Size of sub-sample (count of individuals)
377
378
B
Deviation of sub-sample pupal length
mean from RIL mean (mm)
0.30
0.25
0.20
0.15
0.10
0.05
0.00
-0.05
-0.10
-0.15
-0.20
-0.25
1
3
5
7
9
11 13 15 17 19 21 23 25 27 29 31 37 43 49 55 61 63
Size of sub-sample (count of individuals)
379
380
Figure8.DeviationofmeanRILpupallengthutilizingrandomsub-samplesofvialscomparedtothatbasedonall
381
measuredvials.ForalleightRILandtwocontrolstockswith12ormorevialreplicatemeasurementstheRILmeanpupal
382
lengthwasrecalculatedusingarandomsub-sampleofvials.EachRILwasresampled100timespernumberofvialsub-
383
samplesizeused.Whiskersrepresentrangeofsub-samplemeans,circlesrepresenttheoverallmeans.(A)providesthe
384
rangeof95%ofsub-samplesand(B)50%ofsub-samples.TotalnumberofreplicatevialmeasurementsperRILwereas
23
Page 1
385
follows,[RILs11229,11257,11259,11265,11021=12replicates],11210=13replicates,11236=14replicates,11237=
386
15replicates,335=31replicates,and329=64replicates.Noteallpupaeinvialswereusedtocalculateeachviallength
387
mean,onlythenumberofvialmeansineachsub-samplewasvariedinestimatinglengthRILmeans.
388
h2(mid-parent
regression)
4-way
8-way
Human
Height
Offspringvialmeans
0.44±0.04SE
R2=0.31
Offspringindividual
0.42±0.08SE
R2=0.10
Offspringvialmeans
0.50±0.09SE
R2=0.33
Offspringindividual
0.54±0.02SE
R2=0.14
Offspringfamily(≥4)means
0.68±0.09SE
R2=0.31
Offspringindividual
0.68±0.06SE
R2=0.11
Vialmeans
0.58
Vialmeans
0.61
Notpossible
Vialmeans
0.61
Vialmeans
0.71
-
H2
ExpectedH2
assuming
additiveonly
model.
(2h2)/(1+h2)a
389
390
391
392
393
Estimatesofnarrowsenseheritabilityh2.
394
Estimatesoftheadditivegeneticimpactonthevarianceofthetraitcanbegainedfromthe
395
slopeofregressingtheparentalmidpoint((lengthoffather+lengthofmother)/2)against
396
thelengthoftheprogeny(Galton1886).Figure9indicatesthatdespitethe8-way(67
397
crosses)and4-waydatasets(363crosses)havingnooverlapinbiologicalmaterialand
398
havingquitedistinctsamplingpropertiestheybothresultinremarkablysimilarestimates
399
ofh2,whenrepresentingalloffspringofacrossasamean.Ifalloffspringarerepresented
400
individuallythentheestimateofheritabilityislargelyunchangedforbothdatasets(Table
401
2).Thisrobustnessintheestimatesissharedwithhumanheight(Table2and(Hanley
402
2004)).Likewise,regressionsthatuseonlythefatherormothermeasurements,ratherthan
a(Mackay2013)
Table2Summaryofheritabilityestimatesforpupallengthandhumanheight.DataforbothHumanandDrosophila
isuncorrectedforanysexualdimorphism.
24
theirmidpoint,confirmthatpaternalandmaternaleffectsareofequalmagnitudeforboth
404
pupallengthandhumanheight(seeFigureS3).
405
Offspring pupa length vial average
(mm)
403
4.5
4.0
3.5
3.0
2.5
2.5
3.0
3.5
4.0
4.5
Parental-midpoint (mm)
406
407
408
409
410
411
412
EstimatesofbroadsenseheritabilityH2
413
Estimatesofallpotentiallygenetic(H2)impactsonthevarianceofthetraitcanbeobtained
414
byestimatingtheproportionofthetotalvarianceinmeanviallengthmeasurements
415
relatedtoRILstock.Table2summarizestheresultswithboththe8-way(195RILs)and4-
416
way(81RILs)datasetsgeneratingverysimilarestimatesofH20.58and0.61respectively.
Figure9Estimatesofmeanviallengthnarrowsenseheritabilityh2for8-wayand4-waydatasets.
Despitenooverlapinthestocksusedtoinitiatebothdatasetsthereisaremarkablesimilarityintheslope(h2)andthe
degreeofcorrelation(R2)amongthem,seeTable2forallparameterestimates.Greenclosedcircles=8-waycrosses,blue
opencircles=4-waycrosses,
Page 1
25
417
ThesignificanceofANOVAtestsreflectsthestronggeneticsignal8-wayp=2.7x10-140(F=
418
10.86,MeanSquare=0.093,df=194,sumofsq=18.3)4-wayp=8.8x10-50((F=9.026,Mean
419
Square=0.087,df=80,sumofsq=6.9)
420
421
Wholegenomescanforlociimpactingpupallength.
422
Using195RILsfromthe8-waydatasetandtheDSPRtoolsagenomescanforgenomic
423
regionsassociatedwithmeanRILpupallengthwasconducted(un-weightedmeanof
424
replicatevialmeans).Thedensityofmarkersusedintheanalysisis1SNPper10kbacross
425
almosttheentirereferencegenome(excludingthemtDNAandYchromosome).NoSNPsof
426
significancewereidentifiedatα=0.05,thiswasalsothecaseifdensitycorrectedRILmeans
427
wereused(Figure10).
428
429
430
431
Figure10GenomescanofRILpupallengthin8-waydataset(195RILsmeansuncorrectedforvialdensity).No
peaksaresignificantatap=0.05thresholdof7.02LOD(estimatedbypermutation).
432
26
433
434
DISCUSSION
435
436
While(Visscheretal.2010)statedthat“Onecouldarguethatheightinhumansisthe
437
equivalentofbristlenumberinDrosophila,intermsofitsroleasamodelphenotype.”we
438
proposethatpupallengthisasgood,ifnotabetteranalogue,basedonitsbiological
439
propertiesdescribedindetailforthefirsttimehere.Thisisinadditiontothecapacityto
440
automatephenotypingusingareliablelowcostsystem.Heritabilityisestimatedusingtwo
441
distinctmeasures,broadsense(H2)andnarrowsense(h2)forbothoftwobiologically
442
independentdatasets.Thisallowstheconsistencyofheritabilityestimatestobeassessed
443
betweentwodatasets,whichlikelyspanmostoftherangeofthetraitwithinDrosophila
444
melanogaster.Furthermorethecontrastbetweenh2(additivegeneticeffectsonly)andH2
445
(additive+non-additiveeffects)permitsthemagnitudeofnon-additiveeffectstobe
446
estimated.Thepropertiesofpupallengthestablishedinthismannerarepotentiallysalient
447
toeffortstodevelopresourceefficientmeanstodissectgeneticallycomplextraitsand
448
similartothosereportedforhumanheight(Table3).Thekeyparameterbeingthehighand
449
consistentheritabilityestimates,whichinpartreflectthereliabilityofautomated
450
measurementsweretheestimatedprecisionofmeasurementsof0.04mmissmallrelative
451
totherangeofthetraitasaspecies(1.1mm)orwithindatasets(0.8-0.9mm).Thereliability
452
ofpupallengthmeasurementsalsolikelycontributestothenormalityofthedistributions
453
observedateveryhierarchicallevelinthedatasets(e.g.Figure1).
27
454
AdultHuman
height
Drosophilamelanogaster
pupallength
(basedonthisstudy)
Amongthemost
Intop80%ofmostheritable
heritablereportedin
traitsinD.melanogaster
Humansh2=0.4-0.7a
h2=0.4-0.5b
Manual
Automated
≈70cm
≈1.1mm
≈1.5cmc
0.042mm(±0.030SD)
≈0.02
≈0.04
Equal
Equal
≈180-4000d
Unknown
Notpossible
Easy
Relativeheritabilityoftrait
Measurementmethod
Typicalrangewithinspecies(ignoring
sexualdimorphism)
Measurementerror
Ratioofmeasurementerror:typical
rangeoftraitinspecies
Paternalandmaternalimpacton
offspringtraitvariance.
Numberoflociestimatedtoimpact
varianceoftrait
Experimentalmanipulationofallele
frequencies
455
456
457
458
Table3Salientpropertiesofpupallengthrelativetohumanheight.
a(Hanley2004;Visscheretal.2010)correctedforsexualdimorphism,bnotadjustedforsexualdimorphism,c(Vossetal.
1990),d(Aulchenkoetal.2009;Vinkhuyzenetal.2013;Woodetal.2014)
28
459
Theeaseofphenotypinglargenumbersofindividualsatadevelopmentalstagewhere
460
theyareobligatorilystationaryfor2-3dayshasenabledrobustinsightintotheheritability
461
ofpupallengthat24°.Theverysimilarestimates(Table2)ofheritabilityfortwo
462
independentdatasets,providesincreasedconfidencethatthistraitrepresentsanexcellent
463
prospectwithwhichtoattempttoidentifythegenesunderlyingthiscomplextrait.
464
Furthermore,thesimilaritybetweentheobservedbroadsenseheritabilityH2estimates
465
andthoseexpectedbasedonapurelyadditivemodelofinheritance(seelastrowinTable
466
2)indicatesthatepistatic,dominanceandpaternalormaternaleffectsarerelatively
467
modest(Mackay2013).Theobservedsimilarityinthetwoh2estimates(andtheassociated
468
correlationR2)betweenthe4-wayand8-waydatasets(Figure9)couldalsobearguedto
469
reflectamodestrolefordominance.Thisisbecausewhiletheparentsandoffspringofthe
470
4-waycrosseshavesimilarlevelsofheterozygosity,the8-wayparentsareinbredRILs
471
whiletheiroffspringareheterozygous.Undersomeformsofdominanceeffectsboththe
472
slopeandthecorrelationoftheregressioncouldbereducedinthelattercase,whichisnot
473
thecaseforpupallength.Withrespectstomaternalandpaternaleffectsthedatapresented
474
here(FigureS3)provideevidencethatanynon-additivepaternalormaternaleffectsare
475
small,asthesizeoftheeffectmothersandfathersexertontheiroffspringissymmetrical
476
(whichisalsoapropertyofhumanheight(Hanley2004)).
477
Ithasbeendetailedabovethattheimpactofdensitywithinvialsonpupallengthis
478
relativelymodest(mostlylessthan0.1mm,Figures5and6)andcanbeconveniently
479
controlledfor.Whileotherenvironmentalvariableswerenotsystematicallyexplored,
480
limiteddatacollectedforthetwocontrollinesindicatesthatbetween18°and24°pupal
29
481
lengthchangesby0.2-0.4mm(FigureS4),withpupallengthbeinggreateratcooler
482
temperatures.Thecorrelationbetweenadultbodyweightandpupallengthwasalsobriefly
483
examinedandfoundtobeR2=0.6forbothmalesandfemales(seeFigureS5).Any
484
correlationbetweenadultbodylengthandpupallengthwasnotexamined,thoughitis
485
potentiallynoteworthythatindividualspossessingtheTb1mutationresultingina
486
strikinglyshorttubbypupaarenotreadilydistinguishedfromwildtypeasadultsbasedon
487
theirbodylength.Giventhelargenumbersofgenesimpactinghumanheight(Table1)itis
488
unsurprisingthatitcanbecorrelatedwithdisparatetraitsofparticularinterest(e.g.cancer
489
risk,cardiovasculardiseaseandlongevity(Nelsonetal.2015;(NCD-RisC)2016)).Future
490
examinationoftraitscorrelatedwithpupallength(includingtheirreactionnorms)may
491
proveinformativewithrespectstotheirgeneticarchitectureandthatofpupallength.
492
Currentlytheautomatedsystemdoesnotattempttodistinguishmalepupaefrom
493
femalepupaebutitislikelythatitcouldbeextendedtogenerateprobabilisticsex
494
assignmentsbasedonthefactthatmalepupaeareonaverage≈8%smallerthanfemales
495
(datanotshown).Whenhumanheightdataiscorrectedforsexualdimorphism(alsoan8%
496
averagedifferencebetweensexes,(Galton1886;Hanley2004)),thereisacorresponding
497
increaseoftheheritabilityestimateby9%(seeTable1in(Hanley2004)).Thisimpliesthat
498
analyzingsexedpupallengthdatawillgenerateheritabilityestimatesthatexceedthe
499
valuesreportedhereforunsexedpupae(Table2).
500
Theobservationthatanaverage20%±11.7SDofpupaeinavialarenotmeasuredby
501
theautomatedsystem(Figure4)isgenerallynotanissueformostexperiments(evenfor
502
selectandre-sequenceapproaches),asitisoftenonlynecessarythatasubstantial
30
503
proportionofindividualsaremeasuredthatissufficienttoprovidearangeofparentsfor
504
thenextgenerationandtoreducethevarianceofmeanvialestimates(Figure7).
505
Furthermore,parametersprovidedinFigures7and8willprovidefutureresearchersthe
506
capacitytoreadilydesignmaximallyresourceefficientexperimentalstrategies.
507
Theobservationthatthegenomescanusing195RILs(8-waydataset)failedtoidentify
508
anysignificantlociimpactingpupallengthshouldbeconsideredinthelightthatwiththis
509
modestnumberofRILsthepowertodetectSNPsof10%or5%effectsizeisonly≈0.37and
510
≈0.08respectively(basedonsimulationsincorporatingtheseexactRILs,seefigure9(King
511
etal.2012b)).Consequently,theresultspresentedhereshouldbeviewedasonly
512
potentiallysufficienttoindicatethatfewifanylocioflargeeffectofsizearelikelytoexist
513
forthistraitamongtheRILsused.Itisofcourseconceivablethatphenotypingmoreofthe
514
>1,600RILscurrentlyavailablewouldidentifysignificantloci(Kingetal.2012a;Huanget
515
al.2014),andthereareseveralinFigure10whichareclosetothe0.05significance
516
threshold.Alternativelyanexpandingvarietyofotherapproachescouldbeappliede.g.
517
selectandre-sequence(Turneretal.2011;NuzhdinandTurner2013;Koflerand
518
Schlötterer2014)orcomparisonsbetweenparallelselectedlines(Chanetal.2012).While
519
thisneedstobefurtherexploredinthefuture,itisclearthattheautomatedphenotyping
520
systemwillbeakeytodesigningmappingstrategiesthatshouldallowtheidentificationof
521
manyloweffectsizegenesforthistrait.
522
Theapparatususedtophotographpupaecanberapidlyassembledforapproximately
523
500€(includingacamera)ormayalreadybepresentinmanylabsasgeldocsystems.
524
Furthermorethenecessarysoftwareisopensourceandfreeandrunsonanystandard
31
525
desktopPCorMAC.Thecapacityoftheautomatedsystemtoprovidealargenumberof
526
phenotypedindividualsprovidesincreasedpowertoexaminethegeneticarchitectureof
527
traitsbymostapproaches.Thisislikelytoprovekeytoidentifyinggenesoralleles
528
influencingmeanpupallengthanditsvariance(Woodetal.2014).Thisisinadditionto
529
potentiallyfacilitatingexploringthepoorlyunderstoodgeneticbasisofsexualdimorphism
530
(Rawliketal.2016))andlociwhichimpactthevarianceoftraits(ratherthantheircentral
531
values,(Ayrolesetal.2015)).Inadditionthecapacitytoselectfromlargenumbersof
532
individualsfromwhichtoestablishfuturegenerationshasthepotentialtoenhanceselect
533
andre-sequenceapproaches(Turneretal.2011;NuzhdinandTurner2013;Koflerand
534
Schlötterer2014).Furthermore,thelargenumbersofpowerfulapproachesbasedonsingle
535
pairmatingsdevelopedbyanimalandplantbreeders(GianolaandRosa2015)mayalso
536
becomeamenabletoexaminecomplextraitsinDrosophilathroughtheuseofthissimple
537
lowcostsystem.Aswithhumanheightoveritstypicalrangethegeneticarchitectureof
538
pupallengthisoflimitedpracticalinterest.However,forthelast130yearstheformer
539
continuestoprovidekeyinsightsintothemethodsthroughwhichcomplextraitscanbe
540
bestunderstood,thisisinpartduetotheeaseandreliabilityofhumanheight
541
measurement.Thehighheritabilityofpupallengthandthecapacitytoeasilyautomate
542
phenotypingcombinedwiththesmallandwell-describedgenomeofDrosophila
543
melanogastercouldmakepupallengthasimilarlyvaluablemodeltrait.
544
545
32
546
Bibliography
547
548
AshburnerM.,GolicK.,HawleyR.S.,2011Drosophila:ALaboratoryHandbook.ColdSpringHarborLaboratory
Press,ColdSpringHarbor,N.Y.
549
550
551
AulchenkoY.S.,StruchalinM.V,BelonogovaN.M.,AxenovichT.I.,WeedonM.N.,HofmanA.,UitterlindenA.G.,
KayserM.,OostraB.A.,DuijnC.M.Van,JanssensA.C.J.W.,BorodinP.M.,2009Predictinghuman
heightbyVictorianandgenomicmethods.17:1070–1075.
552
553
554
AyrolesJ.F.,BuchananS.M.,O’LearyC.,Skutt-KakariaK.,GrenierJ.K.,ClarkA.G.,HartlD.L.,BivortB.L.de,
2015Behavioralidiosyncrasyrevealsgeneticcontrolofphenotypicvariability.Proc.Natl.Acad.Sci.
U.S.A.112:6706–6711.
555
BarronA.B.,2000AnaesthetisingDrosophilaforbehaviouralstudies.J.InsectPhysiol.46:439–442.
556
BartonN.H.,EtheridgeA.M.,VéberA.,2016Theinfinitesimalmodel.bioRxiv:039768.
557
558
GaltonF.,1886RegressionTowardsMediocrityinHereditaryStature.TheJournaloftheAnthropological
InstituteofGreatBritainandIreland15:246–263.
559
560
GianolaD.,RosaG.J.M.,2015OneHundredYearsofStatisticalDevelopmentsinAnimalBreeding.Annual
ReviewofAnimalBiosciences3:19–56.
561
562
HanleyJ.A.,2004“Transmuting”WomenintoMen:Galton’sFamilyDataonHumanStature.TheAmerican
Statistician58:237–243.
563
564
HouleD.,MezeyJ.,GalpernP.,CarterA.,2003AutomatedmeasurementofDrosophilawings.BMC
EvolutionaryBiology3:25.
565
566
567
568
569
570
571
572
HuangW.,MassourasA.,InoueY.,PeifferJ.,RamiaM.,TaroneA.M.,TurlapatiL.,ZichnerT.,ZhuD.,LymanR.
F.,MagwireM.M.,BlankenburgK.,CarboneM.A.,ChangK.,EllisL.L.,FernandezS.,HanY.,Highnam
G.,HjelmenC.E.,JackJ.R.,JavaidM.,JayaseelanJ.,KalraD.,LeeS.,LewisL.,MunidasaM.,OngeriF.,
PatelS.,PeralesL.,PerezA.,PuL.,RollmannS.M.,RuthR.,SaadaN.,WarnerC.,WilliamsA.,WuY.-Q.,
YamamotoA.,ZhangY.,ZhuY.,AnholtR.R.H.,KorbelJ.O.,MittelmanD.,MuznyD.M.,GibbsR.A.,
BarbadillaA.,JohnstonJ.S.,StoneE.A.,RichardsS.,DeplanckeB.,MackayT.F.C.,2014Natural
variationingenomearchitectureamong205DrosophilamelanogasterGeneticReferencePanellines.
GenomeResearch24:1193–1208.
573
574
575
KingE.G.,MerkesC.M.,McNeilC.L.,HooferS.R.,SenS.,BromanK.W.,LongA.D.,MacdonaldS.J.,2012a
GeneticdissectionofamodelcomplextraitusingtheDrosophilaSyntheticPopulationResource.
GenomeResearch22:1558–1566.
576
577
KingE.G.,MacdonaldS.J.,LongA.D.,2012bPropertiesandPoweroftheDrosophilaSyntheticPopulation
ResourcefortheRoutineDissectionofComplexTraits.Genetics191:935–949.
578
579
KoflerR.,SchlöttererC.,2014Aguideforthedesignofevolveandresequencingstudies.Mol.Biol.Evol.31:
474–483.
33
580
581
LamprechtM.R.,SabatiniD.M.,CarpenterA.E.,2007CellProfiler:free,versatilesoftwareforautomated
biologicalimageanalysis.BioTechniques42:71–75.
582
583
MackayT.F.C.,2013Epistasisandquantitativetraits:usingmodelorganismstostudygene–gene
interactions.NatureReviewsGenetics15:22–33.
584
585
MediciV.,VoneschS.C.,FryS.N.,HafenE.,2015TheFlyCatwalk :AHigh-ThroughputFeature-BasedSorting
SystemforArtificialSelectioninDrosophila.5:317–327.
586
(NCD-RisC)N.R.F.C.,2016Acenturyoftrendsinadulthumanheight.eLife5:e13410.
587
588
589
590
591
592
593
594
595
NelsonC.P.,HambyS.E.,SaleheenD.,HopewellJ.C.,ZengL.,AssimesT.L.,KanoniS.,WillenborgC.,BurgessS.,
AmouyelP.,AnandS.,BlankenbergS.,BoehmB.O.,ClarkeR.J.,CollinsR.,DedoussisG.,FarrallM.,
FranksP.W.,GroopL.,HallA.S.,HamstenA.,HengstenbergC.,HovinghG.K.,IngelssonE.,Kathiresan
S.,KeeF.,KönigI.R.,KoonerJ.,LehtimäkiT.,MärzW.,McPhersonR.,MetspaluA.,NieminenM.S.,
O’DonnellC.J.,PalmerC.N.A.,PetersA.,PerolaM.,ReillyM.P.,RipattiS.,RobertsR.,SalomaaV.,Shah
S.H.,SchreiberS.,SiegbahnA.,ThorsteinsdottirU.,VeronesiG.,WarehamN.,WillerC.J.,ZallouaP.A.,
ErdmannJ.,DeloukasP.,WatkinsH.,SchunkertH.,DaneshJ.,ThompsonJ.R.,SamaniN.J.,2015
GeneticallyDeterminedHeightandCoronaryArteryDisease.NewEnglandJournalofMedicine372:
1608–1618.
596
597
NuzhdinS.V.,TurnerT.L.,2013Promisesandlimitationsofhitchhikingmapping.CurrentOpinioninGenetics
&Development23:694–699.
598
599
600
PoldermanT.J.C.,BenyaminB.,LeeuwC.A.de,SullivanP.F.,BochovenA.van,VisscherP.M.,PosthumaD.,
2015Meta-analysisoftheheritabilityofhumantraitsbasedonfiftyyearsoftwinstudies.Nature
Genetics47:702–709.
601
602
RawlikK.,Canela-XandriO.,TenesaA.,2016Evidenceforsex-specificgeneticarchitecturesacrossaspectrum
ofhumancomplextraits.GenomeBiology17.
603
604
RoffD.A.,MousseauT.A.,1987QuantitativegeneticsandfitnesslessonsfromDrosophila.Heredity58:103–
118.
605
606
607
TurnerT.L.,StewartA.D.,FieldsA.T.,RiceW.R.,TaroneA.M.,2011Population-BasedResequencingof
ExperimentallyEvolvedPopulationsRevealstheGeneticBasisofBodySizeVariationinDrosophila
melanogaster(GGibson,Ed.).PLoSGenetics7:e1001336.
608
609
610
VinkhuyzenA.aE.,WrayN.R.,YangJ.,GoddardM.E.,VisscherP.M.,2013Estimationandpartitionof
heritabilityinhumanpopulationsusingwhole-genomeanalysismethods.Annualreviewofgenetics
47:75–95.
611
612
VisscherP.M.,McEvoyB.,YangJ.,2010FromGaltontoGWAS:quantitativegeneticsofhumanheight.Genetics
Research92:371–379.
613
614
VossL.D.,BaileyB.J.,CummingK.,WilkinT.J.,BettsP.R.,1990Thereliabilityofheightmeasurement(the
WessexGrowthStudy).ArchivesofDiseaseinChildhood65:1340–1344.
615
616
617
WählbyC.,KamentskyL.,LiuZ.H.,Riklin-RavivT.,ConeryA.L.,O’RourkeE.J.,SokolnickiK.L.,VisvikisO.,
LjosaV.,IrazoquiJ.E.,GollandP.,RuvkunG.,AusubelF.M.,CarpenterA.E.,2012Animageanalysis
toolboxforhigh-throughputC.elegansassays.NatureMethods9:714–716.
34
618
619
620
621
622
623
624
WeedonM.N.,LangoH.,LindgrenC.M.,WallaceC.,EvansD.M.,ManginoM.,FreathyR.M.,PerryJ.R.B.,
StevensS.,HallA.S.,SamaniN.J.,ShieldsB.,ProkopenkoI.,FarrallM.,DominiczakA.,JohnsonT.,
BergmannS.,BeckmannJ.S.,VollenweiderP.,WaterworthD.M.,MooserV.,PalmerC.N.A.,MorrisA.
D.,OuwehandW.H.,ZhaoJ.H.,LiS.,LoosR.J.F.,BarrosoI.,DeloukasP.,SandhuM.S.,WheelerE.,
SoranzoN.,InouyeM.,WarehamN.J.,CaulfieldM.,MunroeP.B.,HattersleyA.T.,McCarthyM.I.,
FraylingT.M.,2008Genome-wideassociationanalysisidentifies20locithatinfluenceadultheight.
NatureGenetics40:575–583.
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
WoodA.R.,EskoT.,YangJ.,VedantamS.,PersT.H.,GustafssonS.,ChuA.Y.,EstradaK.,LuanJ.,KutalikZ.,
AminN.,BuchkovichM.L.,Croteau-ChonkaD.C.,DayF.R.,DuanY.,FallT.,FehrmannR.,FerreiraT.,
JacksonA.U.,KarjalainenJ.,LoK.S.,LockeA.E.,MägiR.,MihailovE.,PorcuE.,RandallJ.C.,ScheragA.,
VinkhuyzenA.A.E.,WestraH.-J.,WinklerT.W.,WorkalemahuT.,ZhaoJ.H.,AbsherD.,AlbrechtE.,
AndersonD.,BaronJ.,BeekmanM.,DemirkanA.,EhretG.B.,FeenstraB.,FeitosaM.F.,FischerK.,
FraserR.M.,GoelA.,GongJ.,JusticeA.E.,KanoniS.,KleberM.E.,KristianssonK.,LimU.,LotayV.,Lui
J.C.,ManginoM.,LeachI.M.,Medina-GomezC.,NallsM.A.,NyholtD.R.,PalmerC.D.,PaskoD.,
PechlivanisS.,ProkopenkoI.,RiedJ.S.,RipkeS.,ShunginD.,StancákováA.,StrawbridgeR.J.,SungY.J.,
TanakaT.,TeumerA.,TrompetS.,LaanS.W.vander,SettenJ.van,Vliet-OstaptchoukJ.V.Van,Wang
Z.,YengoL.,ZhangW.,AfzalU.,ÄrnlövJ.,ArscottG.M.,BandinelliS.,BarrettA.,BellisC.,BennettA.J.,
BerneC.,BlüherM.,BoltonJ.L.,BöttcherY.,BoydH.A.,BruinenbergM.,BuckleyB.M.,BuyskeS.,
CaspersenI.H.,ChinesP.S.,ClarkeR.,Claudi-BoehmS.,CooperM.,DawE.W.,JongP.A.De,DeelenJ.,
DelgadoG.,DennyJ.C.,Dhonukshe-RuttenR.,DimitriouM.,DoneyA.S.F.,DörrM.,EklundN.,EuryE.,
FolkersenL.,GarciaM.E.,GellerF.,GiedraitisV.,GoA.S.,GrallertH.,GrammerT.B.,GräßlerJ.,
GrönbergH.,GrootL.C.P.G.M.de,GrovesC.J.,HaesslerJ.,HallP.,HallerT.,HallmansG.,Hannemann
A.,HartmanC.A.,HassinenM.,HaywardC.,Heard-CostaN.L.,HelmerQ.,HemaniG.,HendersA.K.,
HillegeH.L.,HlatkyM.A.,HoffmannW.,HoffmannP.,HolmenO.,Houwing-DuistermaatJ.J.,IlligT.,
IsaacsA.,JamesA.L.,JeffJ.,JohansenB.,JohanssonÅ.,JolleyJ.,JuliusdottirT.,JunttilaJ.,KhoA.N.,
KinnunenL.,KloppN.,KocherT.,KratzerW.,LichtnerP.,LindL.,LindströmJ.,LobbensS.,Lorentzon
M.,LuY.,LyssenkoV.,MagnussonP.K.E.,MahajanA.,MaillardM.,McArdleW.L.,McKenzieC.A.,
McLachlanS.,McLarenP.J.,MenniC.,MergerS.,MilaniL.,MoayyeriA.,MondaK.L.,MorkenM.A.,
MüllerG.,Müller-NurasyidM.,MuskA.W.,NarisuN.,NauckM.,NolteI.M.,NöthenM.M.,OozageerL.,
PilzS.,RaynerN.W.,RenstromF.,RobertsonN.R.,RoseL.M.,RousselR.,SannaS.,ScharnaglH.,
ScholtensS.,SchumacherF.R.,SchunkertH.,ScottR.A.,SehmiJ.,SeufferleinT.,ShiJ.,SilventoinenK.,
SmitJ.H.,SmithA.V.,SmolonskaJ.,StantonA.V.,StirrupsK.,StottD.J.,StringhamH.M.,SundströmJ.,
SwertzM.A.,SyvänenA.-C.,TayoB.O.,ThorleifssonG.,TyrerJ.P.,DijkS.van,SchoorN.M.van,Velde
N.vander,HeemstD.van,OortF.V.A.van,VermeulenS.H.,VerweijN.,VonkJ.M.,WaiteL.L.,
WaldenbergerM.,WennauerR.,WilkensL.R.,WillenborgC.,WilsgaardT.,WojczynskiM.K.,WongA.,
WrightA.F.,ZhangQ.,ArveilerD.,BakkerS.J.L.,BeilbyJ.,BergmanR.N.,BergmannS.,BiffarR.,
BlangeroJ.,BoomsmaD.I.,BornsteinS.R.,BovetP.,BrambillaP.,BrownM.J.,CampbellH.,Caulfield
M.J.,ChakravartiA.,CollinsR.,CollinsF.S.,CrawfordD.C.,CupplesL.A.,DaneshJ.,FaireU.de,Ruijter
H.M.den,ErbelR.,ErdmannJ.,ErikssonJ.G.,FarrallM.,FerranniniE.,FerrièresJ.,FordI.,ForouhiN.
G.,ForresterT.,GansevoortR.T.,GejmanP.V.,GiegerC.,GolayA.,GottesmanO.,GudnasonV.,
GyllenstenU.,HaasD.W.,HallA.S.,HarrisT.B.,HattersleyA.T.,HeathA.C.,HengstenbergC.,HicksA.
A.,HindorffL.A.,HingoraniA.D.,HofmanA.,HovinghG.K.,HumphriesS.E.,HuntS.C.,HypponenE.,
JacobsK.B.,JarvelinM.-R.,JousilahtiP.,JulaA.M.,KaprioJ.,KasteleinJ.J.P.,KayserM.,KeeF.,
Keinanen-KiukaanniemiS.M.,KiemeneyL.A.,KoonerJ.S.,KooperbergC.,KoskinenS.,KovacsP.,
KrajaA.T.,KumariM.,KuusistoJ.,LakkaT.A.,LangenbergC.,MarchandL.Le,LehtimäkiT.,LupoliS.,
MaddenP.A.F.,MännistöS.,ManuntaP.,MaretteA.,MatiseT.C.,McKnightB.,MeitingerT.,MollF.L.,
MontgomeryG.W.,MorrisA.D.,MorrisA.P.,MurrayJ.C.,NelisM.,OhlssonC.,OldehinkelA.J.,OngK.
K.,OuwehandW.H.,PasterkampG.,PetersA.,PramstallerP.P.,PriceJ.F.,QiL.,RaitakariO.T.,
RankinenT.,RaoD.C.,RiceT.K.,RitchieM.,RudanI.,SalomaaV.,SamaniN.J.,SaramiesJ.,Sarzynski
M.A.,SchwarzP.E.H.,SebertS.,SeverP.,ShuldinerA.R.,SinisaloJ.,SteinthorsdottirV.,StolkR.P.,
TardifJ.-C.,TönjesA.,TremblayA.,TremoliE.,VirtamoJ.,VohlM.-C.,AmouyelP.,AsselbergsF.W.,
35
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
AssimesT.L.,BochudM.,BoehmB.O.,BoerwinkleE.,BottingerE.P.,BouchardC.,CauchiS.,Chambers
J.C.,ChanockS.J.,CooperR.S.,BakkerP.I.W.de,DedoussisG.,FerrucciL.,FranksP.W.,FroguelP.,
GroopL.C.,HaimanC.A.,HamstenA.,HayesM.G.,HuiJ.,HunterD.J.,HveemK.,JukemaJ.W.,Kaplan
R.C.,KivimakiM.,KuhD.,LaaksoM.,LiuY.,MartinN.G.,MärzW.,MelbyeM.,MoebusS.,MunroeP.B.,
NjølstadI.,OostraB.A.,PalmerC.N.A.,PedersenN.L.,PerolaM.,PérusseL.,PetersU.,PowellJ.E.,
PowerC.,QuertermousT.,RauramaaR.,ReinmaaE.,RidkerP.M.,RivadeneiraF.,RotterJ.I.,Saaristo
T.E.,SaleheenD.,SchlessingerD.,SlagboomP.E.,SniederH.,SpectorT.D.,StrauchK.,StumvollM.,
TuomilehtoJ.,UusitupaM.,HarstP.vander,VölzkeH.,WalkerM.,WarehamN.J.,WatkinsH.,
WichmannH.-E.,WilsonJ.F.,ZanenP.,DeloukasP.,HeidI.M.,LindgrenC.M.,MohlkeK.L.,SpeliotesE.
K.,ThorsteinsdottirU.,BarrosoI.,FoxC.S.,NorthK.E.,StrachanD.P.,BeckmannJ.S.,BerndtS.I.,
BoehnkeM.,BoreckiI.B.,McCarthyM.I.,MetspaluA.,StefanssonK.,UitterlindenA.G.,DuijnC.M.van,
FrankeL.,WillerC.J.,PriceA.L.,LettreG.,LoosR.J.F.,WeedonM.N.,IngelssonE.,O’ConnellJ.R.,
AbecasisG.R.,ChasmanD.I.,GoddardM.E.,VisscherP.M.,HirschhornJ.N.,FraylingT.M.,2014
Definingtheroleofcommonvariationinthegenomicandbiologicalarchitectureofadulthuman
height.NatureGenetics46:1173–1186.
685
Acknowledgements
686
Wewouldliketothankthefollowingindividuals.StuartMacdonaldforhishelpwiththe
687
DRSPstocksandtoolsandformaintainingthisvaluablecommonresource.Masayoshi
688
Watadaforhisinvaluableadviceandaccesstohispersonalflystocks.PredragKalajdzicfor
689
karyotypingthefounderlinesofthe4-waycross.OliverHirschforhelpwiththe
690
developmentofthesemi-transparentbarcodestickers.CarolinaWählbyforhelpwith
691
extendingtheuseofthe‘untangleworms’moduleofCellprofiler.AnitaMoellerforher
692
excellenttechnicalhelpandadvicewithconductingthisexperiment.
36