Copyright 0 1995 by the Genetics Society of America

Major Quantitative Trait Loci Affecting Honey Bee Foraging Behavior

Greg J. Hunt,* Robert E. Page, Jr.,t M. Kim Fondrkt and Charles J. Dullumt

*Department of Entomology, Purdue University, West Lafayette, Indiana 47907-1 158 and tDepartment of Entomology, University of California, Davis, California 9561 6

Manuscript received March 25, 1995 Accepted for publication August 30, 1995

ABSTRACT We identified two genomic regions that affect the amount of pollen stored in honey bee colonies and

influence whether foragers will collect pollen or nectar. We selected for the amount of pollen stored in combs of honey bee colonies, a colony-level trait, and then used random amplified polymorphic DNA (RAPD) markers and interval mapping procedures with data from backcross colonies to identify two quantitative trait loci (plnl and pln2, LOD 3.1 and 2.3, respectively). Quantitative trait loci effects were confirmed in a separate cross by demonstrating the cosegregation of marker alleles with the foraging behavior of individual workers. Both plnl and pln2 had an effect on the amount of pollen carried by foragers returning to the colony, as inferred by the association between linked RAPD marker alleles, D8-.3f and 301-.55, and the individual pollen load weights of returning foragers. The alleles of the two marker loci were nonrandomly distributed with respect to foraging task. The two loci appeared to have different effects on foraging behavior. Individuals with alternative alleles for the marker linked to pln2 (but not p ln l ) differed with respect to the nectar sugar concentration of their nectar loads.

T HE number and effects of genes influencing natu- rally occurring behavioral traits is an issue of great

importance to studies of behavior and behavioral evolu- tion. Current evolutionary theory assumes that most adaptive traits are polygenic in inheritance, influenced by many genes, each having infinitesimally small addi- tive effects (BARTON and TURELLI 1989; ORR and COYNE 1992; LAI et al. 1994). This assumption of polygenic inheritance is particularly strong in quantitative and behavioral genetics. However, there have been few studies that were sufficient to determine the numbers and relative effects of genes on adaptive traits. These determinations can be made when mapping quantita- tive trait loci (QTL) involved in particular crosses. Most QTL mapping studies of crop plants have revealed a few major loci with relatively large effects and additional loci with smaller effects (reviewed by TANKSLEY 1993).

Single genes or loci with major effects have been identified for behavioral disorders of humans, mice, rats, and Drosophila (GILLIAM 1992; CARDON et al. 1994; CRABBE et al. 1994; HALL 1994; MCKLUSICK 1994; TAKA- HASHI et al. 1994). For example, QTL with major effects on behavior have been found for human behavioral traits such as dyslexia (CARDON et al. 1994) and Alzhei- mer’s disease (ST. GEORGE-HYSLOP et al. 1990). Behav- ioral genes have also been mapped in studies of Dro- sophila mutants (reviewed by KYRIACOU and HALL

1994). There are only a few mapped loci affecting be- havioral traits in nonlaboratory animal populations. Ex- amples include a sex-linked locus for mice with effects

Corresponding author: Greg J. Hunt, Entomology Hall, Purdue Uni- versity, West Lafayette, IN 47907-1 158. E-mail: greg-hunt@entm.purdue.edu

Genetics 141: 1537-1545 (December, 1995)

on aggression (CARLIER et al. 1990; VAN OORTMERSSEN and SLUV~ER 1994) and recently, the dominant locus, for, which controls a component of foraging behavior in Drosophila larvae (DE BELLE et al. 1989,1993). Mapping studies to find loci that affect animal behavior are im- peded in many of the commonly studied species by large environmental effects and by the difficulties of testing large progeny sets.

The honey bee, Apis mellijiia, offers advantages as an experimental organism for studying the behavioral genetics of foraging. Honey bees are social insects that live in large colonies, so that many individuals can be generated and tested at one time. The single reproduc- tive female, the queen, mates with -12-17 haploid males (reviewed by PAGE 1986) to produce as many subfamilies as possible in the colony, but artificial in- semination may be used to control matings (LAIDLAW 1977). Nonreproductive females (workers) from the same subfamily share an average of 75% of their genes by descent, because a haploid male (drone) transmits an identical genome to each of his worker progeny. Therefore, in colonies containing a queen mated to a single drone, the genetic similarity of workers within each colony makes it easier to assess the paternal contri- bution to worker behavior when comparing between colonies. Behavioral variation among subfamilies has been found for pollen foraging (CALDERONE and PAGE 1988, 1991; PAGE and ROBINSON 1989, 1991; OLDROYD et al. 1991), foraging distance (OLDROYD et al. 1993) and floral resource preference (OLDROYD et al. 1992), thus demonstrating heritable variation for these traits. Interactions between subfamilies (genotypes) and col- ony environments have also been observed (CALDE-

RONE and PAGE 1992).

1538 G. J. Hunt rt al.

A

nd Generation, .ow-Pollen Strain ,5-0// ’

2nd Generation, High-Pollen Strain

d

Fl Queen ?,

\ 3

segregating% haploid \ high-strain inbred

drones queens

38 colonies to test for segregation of QTLS inherited from fathers of the colonies

5th Generation 5th Generation High-Pollen Low-Pollen Strain Queen Strain Drone

High-Pollen Queen Strain Drone

???????? Many worker progeny segregating

for pollen-foraging QTLs

FIGURE I.-Backcross schemes for studying pollen-foraging quantitative trait loci. (A) Crossing scheme for detection and mapping of colony-level QTL for pollen-foraging (measured as area of pollen stored in combs) was based on the contribu- tion of QTL from segregating haploid drone fathers of each colony. The cross to produce the FI queen was made after the second generation of selection. Segregating haploid drones were each individually mated to a single virgin queen. The source of unfertilized eggs resulting in haploid drones is shown by dashed arrows. (B) Crossing scheme to confirm effects of QTL on individual foraging behavior based on seg- regation in worker progeny of the Fl queen in a single colony. The cross to produce the F1 queen was made after the fifth generation of selection.

Pollen and nectar foraging in the honey bee are com- plex behaviors expressed in a social context. Pollen is an important resource for honey bee colonies because it serves as the protein source for raising brood. The quantity of pollen stored in the comb is regulated by both positive and negative stimuli, such as the presence of brood and the size of the pollen stores (FREE 1967; FREE and WILLIAMS 1971; AL-TIKRITY et al. 1972; C M A - ZINE 1993). As new sources of pollen and nectar become available, foragers are recruited to specific resources through a system involving recruitment dances (vON FRISCH 1967; SEELEY 1985; MICHELSEN et al. 1992). Indi-

vidual bees respond to removal of colony pollen stores with both increased foraging effort and the recruitment of new foragers to pollen resources until pollen stores return to previous levels (FEWELI. and WINSTON 1992). In addition to environmental stimuli, the probability that an individual worker will forage for pollen us. nec- tar also depends on her genotype.

The pollen-hoarding trait (storing pollen in the nest) of the honey bee has relatively high heritability and can be selected as a colony-level phenotype (HELLMICH et al. 1985). Recently, high and low pollen-hoarding strains of honey bees were established by selecting for the amount of pollen stored in the combs of the nest (PAGE and FONDRK 1995). The effect of the colony-level selection was to differentially change the proportion of pollen foragers in the high and low strain colonies. The effect of two-way selection was observable at the level of amounts of stored pollen within the colony, as well as at the level of individual foraging behavior. There- fore, we crossed high and low pollen-hoarding lines to produce a mapping population of colonies for de- tecting QTL involved in pollen-hoarding behavior and to study the effects of the QTL on individual behavior. Here we report the map location of two such loci and their effects on both colony and individual phenotypes.

MATERIALS AND METHODS

Detection of pollen-hoarding QTL at the colony level: We conducted five generations of two-way selection to establish strains of bees that hoarded high and low quantities of pollen based on the amount of pollen stored in the combs of the hive (PAGE and FONDRK 1995). After two generations, a low- strain virgin queen was instrumentally inseminated with se- men of a male derived from a generation 2 queen of the high strain (LAIDLAW 1977). An F, virgin queen was produced and exposed to COP to induce egg laying, without mating; normal males are genetically haploid and derived directly from the unfertilized gametes of queens. A generation 2 high-strain virgin queen was instrumentally inseminated with the sperm of a male derived from her mother to produce inbred, high- strain virgin queens (see Figure IA). Males derived from un- fertilized eggs of one F1 queen were mated individually to inbred high-strain virgin queens. This backcross was made to the high line because of evidence that high pollen-hoarding behavior is a recessive trait (PAGE et al. 1995). Colonies con- sisting of backcross worker progeny were established from these queens. Thirty-eight of these colonies were evaluated for the amount of stored pollen by measuring the area of wax cells filled with pollen in all of the combs in each colony (PAGE and FONDRK 1995). Measurements were made in an almond orchard near Davis, California during the spring bloom in the first week of March, 1992.

Linkage analyses of RAPD markers for colony-level data: DNA was extracted from 95 haploid male progeny of the FI queen and was used to construct a saturated genomic map with RAPD markers (HUNT and PAGE 1995). RAPD markers (WILLIAMS et al. 1990) were produced by amplifying honey bee DNA in PCR using commercially available 10-base oligo- nucleotides as primers, as previously described (HUNT and PAGE 1995). Markers were resolved in gels containing 1 - 1.2% Synergel (Diversified Biotech, Newton Centre, MA), 0.7% agarose and 0.5X TBE. The QTL analyses compared the in-

QTLs Affecting Foraging Behavior 1539

A Group I1 I I I I

31.7 -

23.5 -

- 386-1.37

- 31 2-.74f - 438-.62 U15-.51f

- Dl 6-57 - 308-.27

D8-.3f, S9-

388-.76f

1.7 T20-.84

1.4 P9-.34h 331-.855

2.1 239-.96f

I I

I 1

(ii 4-36

I 1 I 1

LOD: 2.0 3.0

B Group X

U15-.8

44447 326.641 N6-.51f

536-1 09f

H7-.51

361-34 L12-.63f 456-.38 N11-33,

3 0 1 4 5 F8-1.04

P13-.67f

w5-1.m S15.16

SlC.87

LOD 2.0

I I I I I I I I I I I I I I I 3.0

FIGURE 2.-Major linkage groups containing the most likely locations of QTL plnl and pln2 as determined by Mapmaker and Mapmaker-QTL (LANDER et al. 1987; LANDER and BOTSTEIN 1989). Linkage group designations (I1 and X) are from HUNT and PAGE (1995). Numbers to the left of the bar are distances in cM between markers. Numbers to the right designate specific RAE’D markers. Numbers preceded by an alphabetic character represent primers produced by Operon Technologies, Alameda, California, while those without an alphabetic character were produced by University of British Columbia Biotechnology Labora- tory. Numbers following the dash represent the size of the RAPD marker fragment in kb. For instance, linkage group 11, marker D8-.3f, is a fragment amplified with Operon primer D8 and is 0.3 kb in length. A final letter designation of “f’ shows that this particular marker is a fragment length polymorphism while those followed by “h” are fragment length polymorphisms that form distinguishable heteroduplexes (HUNT and PAGE 1992). The line tracing to the right shows the LOD score for the likelihood that a QTL-exists at each location along the linkage group.

heritance of specific RAPD markers from each drone with pollen stores within the colony that the drone sired. Inter- val mapping with Mapmaker-QTL software (LANDER and BOTSTEIN 1989) was used to screen 364 markers belonging to 26 linkage groups that each contained at least three RAPD markers, spanning 3100 cM of the honey bee genome at an average spacing of -9 cM (HUNT and PAGE 1995).

Confirmation of QTL effects with individual behavior: A second FI cross was made between two individuals from the fifth generation of selection for pollen-hoarding behavior (Figure 1B). An FI queen was mated to a single drone from a high pollen line. Three thousand newly emerged adultwork- ers from the reulting colony were marked with paint. Marked workers were collected as they returned from foraging trips carrying nectar and/or pollen over a 3-day period during warm weather of August, 1993 (near Davis, CA). Returning foragers were collected, and nearly all of the marked individu- als that were currently engaged in foraging were removed by the end of the 3-day period. The bees were anesthetized with COS within 10 min of collection, and the weights of individual pollen and nectar loads were recorded. The nectar was with- drawn from each forager by squeezing the bee so that the nectar entered a capillary tube. The relative sugar concentra- tion of the nectar load was determined with a hand-held re- fractometer (GARY and LORENZEN 1976; ROBINSON and PAGE 1989). Pollen was removed from the corbiculae on legs of individual pollen foragers and weighed. The few individuals that were not carrying pollen or nectar were discarded. Three hundred and thirty-two foragers were frozen on dry ice, and DNA was extracted from each individual (HUNT and PACE

1995). The inheritance of RAPD markers linked to the two putative pollen-foraging QTL was then correlated with indi- vidual foraging behavior. We used data from one informative RAPD marker locus linked to each of the two QTL to infer the QTL genotype. Some error in scoring QTL genotypes is expected because we did not have informative flanking mark- ers. RAF’D markers from individual lanes that were considered too dificult to score were entered as missing data.

Statistical methods: Bees often specialize on either pollen or nectar, resulting in relatively large classes of individuals with zero values for load weights for each resource and hence a bimodal distribution for both nectar and pollen load weights. The assumptions of analysis of variance (ANOVA) were met when only the individuals with nonzero values were used in analyses, so individuals that did not carry pollen or nectar were eliminated from those analyses. Analysis of vari- ance was used for data sets with nonzero values for the three variables pollen load weight, nectar load weight and nectar concentration. In addition, nonparametric methods were used on complete data (with zero values) to correlate individ- ual foraging behavior with the inheritance of RAPD marker alleles at the two loci. Weights of pollen and nectar loads and nectar concentrations were first compared using a Kruskal- Wallis test for individuals with the four possible genotypes at these two loci (SOW and ROHLF 1981). When a statistically significant Kruskal-Wallis test was obtained, Mann-Whitney U tests were applied to data from individuals with alternative maternally inherited alleles at each locus. The effects of the two pollen-collecting QTL were also evaluated by separating individual foragers into task groups (collecting pollen only,

1540 G. J. Hunt et al.

nectar only, or both) and performing a Gtest for heterogene- ity (SOKAL. and ROHI.F 1981). This analysis emphasized forag- ing choice, independent of load size. Finally, the correlation of nectar concentration and nectar load sizes carried by re- turning foragers was evaluated by linear regression for alterna- tive alleles and for task groups.

RESULTS

Colony-level QTL mapping: The 38 haploid drone progeny of the first F, queen were analyzed for RAPD markers after they were individually backcrossed to sis- ter queens from the high-pollen line (Figure 1A). As a consequence of male haploidy, backcross workers within each of the 38 colonies inherited the same ge- nome from the drone father of that particular colony. The high-strain queen mothers of the colonies each shared an average of 287.5% of their genes by common descent because of male haploidy and inbreeding. QTL analyses for the total area of pollen stored in colonies showed that one marker interval, D8-3f to 388-.76f (Fig- ure 2A), had a LOD score (logarithm of odds ratio) for likelihood of containing a QTL of 3.1 and explained 38% of the total phenotypic variance. This putative QTL was designated plnl. Another interval, P13-.67f to S15-.16 (pln2), contained a putative QTL with a LOD of 2.3 that explained 33% of the total phenotypic variance (Figure 2B). In combination, the two loci explained 59% (LOD 5.3) of the phenotypic variance. Flanking marker data showed that the high-pollen plnl allele was inherited by the F, queen from her low-strain parent, the high pln2 allele had a high-strain origin. The low- strain origin ofplnl can be explained by lack of fixation for all QTL after just two generations of selection or by chance fixation of some QTLs among strains as a consequence of the initial establishment of the strains. It is also possible that this QTL has an overdominant effect, in which case it would appear that a high allele had been inherited from the low strain in the backcross.

Confirmation of the effects of plnl and pln2 on indi- vidual behavior: High-strain foragers individually are more likely to return from foraging trips with loads of pollen than are workers of low pollen-hoarding strains (CALDERONE and PACE 1988, 1991; PAGE et al. 1995). Therefore, to confirm the behavioral effect of the QTL, we backcrossed a second FI queen from generation 5 to a generation 5 high-strain drone (Figure 1B). In this cross workers differed for both putative QTL and their linked RAPD markers only as a consequence of meiotic recombination in the F1 queen. Data for pollen and nectar load sizes and nectar concentration of incoming foragers were not normally distributed because there were many zero values for each variable (Figure 3, A and B). In addition, nectar and pollen loads of incoming foragers were not independent variables because they were inversely correlated (9 = 0.513) (Figure 3C). This correlation was significant, even if we considered only those individuals that carried both pollen and nectar (9 = 0.174, p < 0.0001).

I ~~ ~

M rA A 3 50 - 2 2 40 - a F;

YI m 30 - 0

P E 20 - 2 10-

0 0 10 20 30 40 50 60 70

Nectar Load Weight (mg)

I I

$I$ 4 B z 2 40 a . F; +I 30' + I , 0

P ' 8 20'

ti 10- z "1 "

0 5 10 15 20 25 Pollen Load Weight (mg)

- I - v I 'G 40

j 20 I4 2 10 c u o a, k

? = 0.513 C

0 0 0 0

' 8': 8 Q 8 ~ ! !,eo! 0 ~ 0 ~ ~ ~ 0 0

0 Q 8 0

, 0 0 .le&! 0 0 o o io " - -

A ! ! I 0 5 10 15 20

Pollen Load Weight (mg)

FIGURE 3,"Distribution of load sizes collected from incom- ing foragers. (A) The distribution of nectar load sizes expelled from the honey crops of individual foragers and weighed in capillary tubes. (B) Distribution of pollen load sizes collected from incoming foragers. Pollen loads were removed from both legs of each forager and weighed. (C) The correlation of nectar load weights and pollen load weights (correlation is significant at p < 0.0001).

We determined the inheritance of alleles from two RAPD marker loci linked to plnl and pln2 (D8-.3f and 301-.55, respectively). RAPD-marker alleles at D8-.3f and 301-.55 did not have a detectable effect on viability or on the likelihood of becoming a forager, because the alleles at both loci segregated according to the ex- pected 1:l ratio ( p = 0.845 and 0.506; G = 0.038 and 0.443, respectively, 1 d.f.). The alleles at the two marker

QTIs Alkaing Foraging Ikhitvior I .7)4 I

0.6

0, a A 0 c 0.4

GI’VOIYPE: % 1)s - 14s

Y

G w 0 c 0 - 0 . 2 0 a c

.e L

El

Pollen 150th Sectar Foraging Task Group

c.88 INFERRED QTL GENOTYPE: IIICtj + H I G I I

Y ‘Q HIGH + LO\V 2 0.6 0 LOW + I.0W Q,

c3 u4 0 c 0.4 0

0 a c

.e Y L

El 0.2

Pollen Both Nectar Foraging Task Group

-

-

FW’RF. 4.--Proporrion o f h o n c y Iwr fo~xgc.rs ofc;trh gcwl- type performing tlifkrcnt rasks. ( A ) Thr grnolypir cm-gorirs show the maternal ronrrihurion of RAW markrr alldcs linkrtl 10 pollen-hoarding @TI, / h / and phr2 ( l L \ l W markers I)X-..?f antl 301-..5.5. resprctivdy). lnclivid~tal rvturning Ibragcrs w r c - separated into thrce task g r o ~ ~ p s 1);wrtl on wlwtlwr 111ry wrrc carrying only pollrn (Pollen), I m h pollrn m t l nectar (Both) or j u s t nrctar (Ncc~ar). Intlivitluals w r c f w ~ l t r r sc*p;~l;~trd according IO which R\PD marker allc-lr linkctl 1 0 p o l l r n forag- ing QTI. that they inhrritrtl from t h r F, q u w n . RWI) mitrkrr alleles are used t o infc-r QTI. allrles, h t somr error i n as- signing individuals t o specific @TI. gtrnotypt*s is cxpc*ctrtl due to recombination tx.twrcn thr singlr Ki\I’l) markrr linkrtl t o each @TI.. The designations HS ; t n d IS rcfcr IO r h r high- strain and low-srrain origins of the Krll’I) markrr ~tlldc-s. For the @TI. ph-l linked t o IIK.3f. t h r high-pollen ; t l l c l r w a s i n - herited from the lo~v-strain parrnt i n 1x)rh scts o f crosscs usrd in this srudv. Workers were nonrantlonlly tlisrril)trtcd Ix*twrcn genotypic and task group categoric*s (G = 16.48, p = 0 . 0 I I , tt = 320. 6 d.f.) and ;dso \vert* l)iawtl b y task g r o ~ ~ p a t rach of the two marker loci, I)K.Xand :301-.5.5 (G = 8.140 ; t n d 6.486. p = 0.017 and 0.039. rr = 3’25 ; t n d 527. rr.spcr~ivc*ly, 2 (1.1.). The bias \vas in the direction pr(dictetl by colony-lrvrl t lata. (B) The same task groups wcrc* romp;lrctl whcn individtlals were grouped into t h r tI1l-c-e possihlc QTI. grnoryprs a t I M ) I ~ I loci t h a t resrlltcd from this hackcross. For t w c - of~~r(.s~.~lt;\lioll, the inferred QTI. grnotypr is shown larhrr r h a n t h r strain of origin of rnarkrr allrlrs. Thc “ I ligh“ ;~llrlc o f p / n / ;~ctu;tlly had a l o w strain origin ( p h r /-IS = IHigh). Thcrr W;IS signilicant hrtrrogencity I)ct\vrcn grnotyps and foraging ritsks ( G = I.i. .i , p = 0.004, tr = 320. 4 (1.1.).

loci ;dso WTC inhcrircd i n workers i n d r l ~ c ~ ~ d ~ ~ ~ l t l y , bc- cilllsc’ there \\‘;IS 110 significant s k c ~ liolll 1 : 1 : 1 : 1 (/) >

gcnotypcs varictl significantly for wcighr and conrc*nt~~- tion of ncct;w and for pollcn load wight ( / I = 0 . 0 1 4 , 0 . 0 1 0 and 0.009. rcycctivcly, Kruskd-\Vallis tcs t ) . <;c- notypcs also \.;lrietl w i t h rcspcct t o t h c s proportion of t lw t o t a l l o a d weight (nectar plus pollcn) t h a t w s pollen (11 = 0.00’7. Kruskal-M’allis tcst). This may be t h c more appropriate tcs t hcc;~usc~ o f tllc lack o f inrlcpcntlcncc ol’ t h c t w o v;Iriablcs.

EfTccts on lo;ltl wcigllts were also obsc~vctl ;It c%;1ch ol’ the t w o marker loci. Inhcritancc o f alternative ;~llclcs for m;lrker DK.3f w a s significantly associated wit11 tllc weight of pollc*n and ncctar carricd ( p = 0 . 0 1 3 and 0.026. rcspectivcly. I ) = 32.5, Mann-M’hitncy CQcst). 11s i n t h e c;~sc o f the colony-lcvcl analyscs, the high-pollen QTI. allclc linkctl t o DK.3fthat rcwltctl i n larger pollen lo;~tls (anti smaller nectar loads) w a s inhcritcd from the low-strain p;Irc*nt. Inhcritancc of marker alleles of301- . x> also w a s associ;rtctl with pollct~ and nectar l o a d wights ( p = 0.012 and p = O.O!!S. respectively, 11 = 32’7, M;lnn-”hitncy /kcst). Xllclic substitutions ;It the t w o marker loci, I)K.3f;lnd 301-.5.5. explainctl I .7 and 1.1%. rcsspcctivc.ly, ofthc t o t a l viu-iancc for pollen loat1 wights r , f individual loragcrs i n this Ix~ckcross. \Zhcn an;dyzed i n combin;~tion. they cxplaitlctl 3 3 % o f this variance. This COIIIKIS~S with the colony-level ;unalyscs. i n which allc-lic suhstittltions a t these IWO m;~rker loci explained 9.5 and 183% 01’ the colony-lcvc-l plwnotypic \.ariancc for [he quantity of pollcn stored (and 29.8% whcn ana- l y ~ c d i n combin;\tion). Inheritance o f R-lPD m;lrker allelcs o f DK.3f and 301-..5.5 did not havc significant effects on nectar or pollen l o a d wights whew the zero values were cxclr~rlcrl from ;~n;~lyscs.

Thc correlation of msrkcr allclcs, linkctl t o / h I and / h 2 , with individual foragw bcllavior c m bc clcmon- stratctl by separating returning foragers into task groups t h a t reflect foraging choices: those t h a t carried only pollen loads, those t h a t had both pollen and nectar and those that had o n l y nectar ( 1 1 = 48, I22 and 16‘2, respectively). Each o f these three groups can t h t r n bc scparatcd into their QTL genotypes. a s inferred from the linked markers. Since ;dl of the workers inherited the sams high-strain (13s) m;lrkcr allele from the h a p loid drone father, workers were homozygous ;It ;I

marker loc~ts if they also inhcritcd the HS allcblc frOlT1

the FI qwen and were heterozygous at the markcr locus if they inherited the maternal low-strain ( I S ) dlelc. The maternal markcr-;lllcle contribution t o ~vorkcr progeny i n t h c threc foraging task groups is sho\vn in Figure 4A. Overall, the data showctl ;1 nonr;lndonl distri- brltion o f workers bctwccn RAPD marker genotypes and task-group categories ( G = I(i.481, j ) = 0.03’7. 11 = 320. 6 c1.f.) (Soka!. and Rol11.r 1981) a n d , the tlistrihttion among task groups also w a s hiasrd at each of t h e two marker loci, I)K.:3f and 301-..55 ( G = 8. 140 and 6.486.

0.05. (; = 3.583, I / = 3“O. 3 ~ 1 . f . ) . Th(* f ~ t ~ r l L ~ l l ’ l ~ - ~ ~ ~ i ~ ~ ~ k ( ~ ~ ~

”

1542 G. J. Hunt et al.

Marker DS-.3f, linked to pln-1

Low-strain allele 0 0 0

0 0

0 0 o o o

0 0

. . . . . . 10 2 0 3 0 4 0 50 60

a . . I . .

1 0 2 0 3 0 4 0 5 0 6 0

.6:] 2 .55

Pollen and nectar Nectar only 0 o o o o 0

C 3

$ = 1.5 x 1W5 .2

$ = .os1

5 1 0 1 5 2 0 25 30 35 4 0 10 20 3 0 4 0 5 0 6 0

Nectar Load Weight (mg)

FIGURE 5.-Regression of nectar load weights on relative sugar concentration for individual honey bee foragers that were carrying nectar. (A) Bees inheriting alternative alleles for RAPD marker, D8-.3f linked to plnl. The allele from the low-strain parent conferred the high-pollen phenotype for this QTL. (B) Bees inheriting alternative alleles for marker 301-.55 linked to pln2. Correlations for the two allelic classes of pln2 were significantly different ( p < 0.05). (C) Bees that were carrying pollen and nectar us. those that had only nectar loads.

p = 0.017 and 0.039, n = 325 and 327, respectively, 2 .55, suggesting additivity across the two loci. We there- d.f.) . These results demonstrate a bias toward specific fore pooled the latter two groups and tested for hetero- tasks in the direction predicted by the colony-level data. geneity of the two-locus genotypes: those that inherited

No difference in task group could be detected be- two maternal “high-pollen’’ QTL alleles, those that in- tween individuals that were heterozygous only at D8-.3f herited only one, and those that inherited no high- compared to those that were heterozygous only at 301- pollen alleles from the F1 queen (as inferred by marker

QTLs Affecting Foraging Behavior 1543

genotypes, Figure 4B). For ease of presentation, we show the inferred QTL alleles based on the phenotypic effect seen in the crosses used for this study. The “high” plnl allele actually was inherited from the low- strain parent but the high pln2 allele had a high-strain origin. The heterogeneity of combined genotypes with respect to task groups was apparent ( G = 15.5, p = 0.004, 4 d.f.).

Inheritance of alternative marker alleles for 301-.55 was associated with the nectar concentration in the crops of returning foragers that had nectar ( p = 0.0017, n = 302, one-way ANOVA). The high-strain RAPD marker allele was associated with lower sugar concentra- tion. In contrast, inheritance of alleles of marker D8- .3f had no significant effect on nectar concentration ( p = 0.672, n = 301). Another way to analyze the effect of pln2 on concentration is to look at the relationship between nectar concentration and nectar load size of returning foragers. The regression of nectar concentra- tion on nectar load size has a significantly positive slope for three of the four possible genotypes observed, indi- cating that bees inheriting these marker alleles tended to bring back larger loads if the nectar was sweet (Figure 5, A and B). However, for those individuals inheriting the maternal high-pollen QTL allele of pln2 (as inferred from the linked marker), there was no significant rela- tionship between load size and concentration. Similarly, individual nectar foragers that also carried pollen showed a lack of correlation between nectar load size and concentration (Figure 5C). The correlation coeffi- cients for individuals carrying the alternative alleles of 301-.55 (linked to pln2) were significantly different ( p < 0.05). This relationship demonstrates a differential response to sugar concentration between individual for- agers with alternative alleles for pln2, but not for plnl.

DISCUSSION

We have demonstrated that two loci have major ef- fects on the amount of stored pollen in colonies and detectable effects on individual foraging behavior. These two QTL explained roughly 59% of the total phenotypic variance for quantities of stored pollen in our backcross population of colonies. Although the esti- mates of the individual effects of these two QTL on colony pollen stores are in the upper range reported for any QTL from mapping studies (TANKSLEY 1993), our small sample size of 38 colonies greatly reduces the precision for estimating QTL effects and may be biased upward (BEAVIS 1994). In spite of the uncertainty con- cerning the magnitude of QTL effects, two lines of evi- dence indicate that both of these loci must have a major influence on pollen foraging behavior. The first evi- dence that these loci have a major effect is that we were able to identify them with only 38 colonies. Simulations have shown that backcross progeny sets with 100 indi- viduals have little power to detect a single QTL ex-

plaining 5% of the phenotypic variance (frequency of correct identification = 0.11) (VAN OOIJEN 1992). A similar result was obtained with 10 QTLs that each ex- plained 6.3% of the variance in a simulated data set of 100 F2 progeny (frequency = 0.33) (BEAVIS 1994). The second indication that plnl and pln2 have major effects is the fact that we were able to validate their behavioral effects in a separate cross and using a different pheno- type than the one used to map these loci.

The effect of the two QTL was to alter the probability that an individual would collect pollen us. nectar and, consequently, influenced the quantity of pollen col- lected and stored by the colony. For one of the two loci identified with effects on foraging behavior (plnl), the high-pollen allele was inherited from the low-pollen strain. QTL with allelic effects that are opposite to what would be expected based on parental strain phenotype are not uncommon (PATERSON et al. 1988; DOEBLEY and STEC 1991; DE VICENTE and TANKSLEY 1993). It is possible that alleles of plnl are overdominant in expres- sion (the heterozygote having a more extreme pheno- type than either homozygote), thus giving the appear- ance that the low strain had a high QTL allele because we backcrossed to the high-strain parent. We cannot distinguish between overdominance or complementary QTL fixation at this locus because we used backcrosses to one line only. However, other observations suggest an overdominant expression pattern. The quantity of stored honey (processed nectar) in colonies and the nectar collecting preferences of individual foragers both demonstrate overdominant (or heterotic) pat- terns of inheritance in our selected strains (PAGE et al. 1995; PAGE and FONDRK 1995). Further research is needed to determine the degree of dominance or over- dominance for plnl and pln2.

Results from this study indicate that plnl and pln2 have different effects on individual forager behavior. Only RAPD marker alleles linked to pln2 had a signifi- cant correlation with the relative sugar concentration of the nectar. Figure 5 shows that the correlation of these marker alleles with nectar sugar concentration was due to the relatively low and constant concentra- tions carried by individuals with the RAF’D marker linked to the high-pollen allele of pln2. The lack of a relationship between nectar concentration and load weights for individuals with the high-strain marker linked to pln2 is similar to that observed for pollen foragers that were carrying nectar, but contrasts with the significant relationship observed for individuals with either of the marker alleles linked to plnl. This study is inadequate to determine the specific manner in which these two loci are affecting behavior. However, the availability of markers that are linked to major genes affecting honey bee foraging behavior will enable stud- ies that identify specific genetic components of the be- havior and suggest mechanisms of gene action.

The traditional quantitative genetic models of evolu-

1544 G. J. Hunt et al.

tion assume that many genes with small additive effects are responsible for variation in adaptive behavioral traits (FISHER 1958; BARTON and TURELLI 1989; ORR and COYNE 1992; LAI et al. 1994). However, our data suggest that two loci have a major effect on colony pol- len stores and also influence individual foraging behav- ior. Individual foragers that inherited both RAPD marker alleles linked to the high-pollen QTL alleles were 1.7 times more likely to bring back pollen to the colony than to carry only nectar. Tightly linked markers that flank these two QTL could be used in studies to determine the specific individual effects of these two loci on worker foraging behavior and their contribu- tions to observed phenotypic variability within different honey bee populations.

We thank GENE ROBINSON, CIAUIIIA DRELLER, KEITH WAIIDINGTON and CHRISTIE WIILIAMS for useful comments. JOHAN VAN OOIJEN made valuable suggestions for the statistical analyses in the first cross. This research was funded by contracts from the California Depart- ment of Food and Agriculture and U.S. Department of Agriculture grant 93-37302-8880.

LITERATURE CITED

AI.-TIKRITY, W. S., A. W. BENTON, R. C. HII.I.MAN and W. W. CIARKE, JR., 1972 The relationship between the amount of unsealed brood in honeybee colonies and their pollen collection. J. Apic. Res. 11: 9-12.

BARTON, N. H., and M. TURELIJ, 1989 Evolutionary quantitative genetics: how little do we know? Annu. Rev. Genet. 23: 337- 370.

BEAVIS, W. D., 1994 The power and deceit of QTL experiments: lessons from comparative QTL studies, pp. 250-266 in Proceed- ings of the 49th Annual Corn and Smghum Industry Research Confer- ence. American Seed Trade Association, Chicago.

CALDERONE, N. W., and R. E. PAGE, JR., 1988 Genotypic variability in age polyethism and task specialization in the honey bee, Apis mell+ra, (Hymenoptera:Apidae). Behav. Ecol. Sociobiol. 22: 17- 25.

CALDERONE, N. W., and R. E. PAGE, JR., 1991 Evolutionary genetics of division of labor in colonies of the honey bee (Apis mellifea). Amer. Nat. 138: 69-92.

CALDERONE, N. W., and R. E. PAGE, JR., 1992 Effects of interactions among genotypically diverse nestmates on task specialization by foraging honey bees (Apis mrllifera). Behav. Ecol. Sociobiol. 30:

CAMAZINE, S., 1993 The regulation of pollen foraging by honey bees: how foragers assess the colony’s need for pollen. Behav. Ecol. Sociobiol. 3 2 265-272.

CARLION, L. R., S. D. SMITH, D. W. FULKER, W. J. KIMBERIJNG, B. F. PENNINGTON et al., 1994 Quantitative trait locus for reading disability on chromosome 6. Science 266: 276-279.

CARLIER, M., P. L. ROUBERTOUX, M. L. KO~TLER and H. DEGREILE, 1990 Y chromosome and aggression in strains of laboratory mice. Behav. Genet. 20: 137-156.

CRABBE, J. C., J. K. BEKN~P and K. J. BUCK, 1994 Genetic animal models of alcohol and drug abuse. Science 2 6 4 1715-1723.

DE BELLE, J. S., A. J. HILIKER and M. B. SOKOLOWSKI, 1989 Genetic localization of foraging (for) : a major gene for larval behavior in Drosophila melanogaster. Genetics 1 2 3 157-163.

DE BELLE, J. S., M, B. SOKOI.OWSKI and A. J. HILIXER, 1993 Genetic analysis of the foragmg microregion of Drosophila melanogaster. Genome 36: 94-101.

DE VICENTE, M. C., and S. D. TANKSLEY, 1993 QTL analysis of trans- gressive segregation in an interspecific tomato cross. Genetics

DOEBLEY, J., and A. STEC, 1991 Genetic analysis of the morphologi- cal differences between maize and teosinte. Genetics 129: 285- 295.

219-226.

134 585-596.

FEWEIL, J. H., and M. L. WINSTON, 1992 Colony state and regulation of pollen foraging in the honey bee, Apzs mellijera. L. Behav. Ecol. Sociobiol. 30: 387-393.

FISHER, R. A,, 1958 The Genetical Theory of Natural Selection. Dover Publications, Inc. New York.

FREE, J. B., 1967 Factors determining the collection of pollen by honeybee colonies. Anim. Behav. 1 5 134-144.

FREE, J. B., and I. H. WII.I.IAMS, 1971 The effect of giving pollen and pollen supplement to honeybee colonies on the amount of pollen collected. J. Apic. Res. 10: 87-90.

GARY, N. E., and K. LORENZEN, 1976 A method for collecting the honey-sac contents from honey bees. J. Apic. Res. 15: 73-79.

GILLIAM, T. C., 1992 Mapping psychiatric disease genes: impact of new molecular strategies. J. Psyciat. Res. 26: 309-326.

HALL, J. C., 1994 The mating o f a fly. Science 264: 1702-1714. HELLMICH, R. L., J. M. KULINCEWK and W. C. ROTHENBUHLER, 1985

Selection for high and low pollen-hoarding honey bees. J. Hered. 76: 155-158.

HUNT, G. J., and R. E. PAGE, JR., 1992 Patterns of inheritance of RAPD molecular markers in the honey bee reveal novel types of polymorphism. Theor. Appl. Genet. 85: 15-20.

HUNT, G. J., and R. E. PACE, JR., 1995 Linkage map of the honey bee, Apis mellijera, based on RAPD markers. Genetics 139: 1371- 1382.

KYRIAC:OU, C. P., and J. C. HALL, 1994 Genetic and molecular analy- sis of Drosophila behavior. Advan. Genet. 31: 139-185.

LAI, C., R. F. LYMAN, A. D. LONG, C. H. LANGLEY and T. F. MACKAY, 1994 Naturally occurring variation in bristle number and DNA polymorphisms at the scabrous locus of Drosophila melanogaster. Science 266: 1697-1702.

LAIDIAW, H. H. , 1977 Instrumental Insemination of Honqr Bee Queens. Dadant and Sons, Hamilton, IL.

LANDER, E. S., and D. BOTSTEIN, 1989 Mapping Mendelian factors underlying quantitative traits using RF’LP linkage maps. Genetics

MCKLUSICK, V. A,, 1994 Mendelian Inhedance in Man: A Catalog of Human Genes and Genetic Disorders. Johns Hopkins University Press, Baltimore.

MICHEISEN, A,, B. B. ANDERSEN, J. STORM, W. H. KIRCHNER and M. LINDAUER, 1992 How honeybees perceive communication dances, studied by means of a mechanical model. Behav. Ecol. Sociobiol. 30: 143-150.

OLDROYD, B. P., T. E. RINDERER and S. M. Buco, 1991 Honey bees dance with their super-sisters. Anim. Behav. 42: 121-129.

OLDROYD, B. P., T. E. RINDERER and S. M. BUCO, 1992 Intra-colonial foraging specialism by honey bees (Apis mellijima) (Hymenop tera:Apidae). Behav. Ecol. Sociobiol. 30: 291-295.

OLDROYD, B. P., T. E. RINDERER, S. M. Buco and L. D. BEAMAN, 1993 Genetic variance in honey bees for preferred foraging distance. Anim. Behav. 45: 323-332

ORR, H. A,, and COWE, J. A,, 1992 The genetics of adaptation: a reassesment. Am. Nat. 140: 725-742.

PACE, R. E., JR., and G. E. ROBINSON, 1991 The genetics of divi- sion of labour in honey bee colonies. Adv. Ins. Physiol. 23: 117-169.

PAGE, R. E., JR., and M. K. FONDRK, 1995 The effects of colony-level selection on the social organization of honey bee (Apis mellijiera L.) colonies: colony-level components of pollen hoarding. Behav. Ecol. Sociobiol. 36: 135-144.

PAGE, R. E., JR., K. D. WADDINGTON, G. J. HUNT and M. K. FONDRK, 1995 Genetic determinants of honey bee foraging behavior. Anim. Behav. (in press).

PATERSON, A. H., E. S. LANDER, J. D. HEWITT, S. PETERSON, S. E. LINCOLN et al., 1988 Resolution of quantitative traits into Men- delian factors by using a complete linkage map of restriction fragment length polymorphisms. Nature 335: 721 -726.

ROBINSON, G. E., and R. E. PAGE, JR., 1989 Genetic determination of nectar foraging, pollen foraging, and nest site scouting in honey bee colonies. Behav. Ecol. Sociobiol. 24: 317-323.

SOKAI., R. R., and ROHLF, F. J., 1981 Biometry. W. H. Freeman and Company, New York.

SEEILEY, T. D., 1985 Honeybee Ecology. Princeton University Press, Princeton, NJ.

ST. GEOR(;E-HBI.OP, P. H., J. L. HAINKS, L. A. FARRER, R. POI.INSKY, C. VAV BROECKHOVEN et al., 1990 Genetic linkage studies sug-

121: 185-199.

QTLs Affecting Foraging Behavior 1545

gest that Alzheimer’s disease is not a single homogenous disor- der. Nature 347: 194-197.

TAKAHASHI, J. S., L. H. PINTO and M. H. VITATERNA, 1994 Forward and reverse genetic approaches to behavior in the mouse. Sci- ence 264: 1724-1732.

TANKSLEY, S. D., 1993 Mapping polygenes. Annu. Rev. Genet. 27:

VAN OOIJEN, J., 1992 Accuracy of mapping quantitative trait loci in

VAN OORTMERSSEN, G. A,, and F. SLUITER, 1994 Studies on wild

205-233.

autogamous species. Theor. Appl. Genet. 8 4 : 803-811.

house mice. V. Agression in lines selected for attack latency and their Y-chromosomal congenics. Behav. Genet. 2 4 73-78.

VON FRISCH, K, 1967 The Dance Language and Orientation of Bees. Belknap Press of Harvard University Press, Cambridge, MA.

WILLIAMS, J. G. K, A. R. KUBELIK, K J. LIVAK, J. A. RAFALSKI and S. V. TINGEY, 1990 DNA polymorphisms amplified by arbitrary primers are useful as genetic markers. Nucleic Acids Res. 18: 6531-6535.

Communicating editor: A. G. CLARK