Reproductive Conflicts and Signal Evolution in Social Wasps and Bees

A Dissertation Presented to the Faculty of the Graduate School of Cornell University In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

by Kevin Joseph Loope August 2015

REPRODUCTIVE CONFLICTS AND SIGNAL EVOLUTION IN SOCIAL WASPS AND BEES

Kevin Joseph Loope, Ph. D. Cornell University 2015

Chapter 1 (published in Naturwissenshaften) is a side project on honeybee behavior. We showed that honeybee colonies that are headed by queens who are artificially inseminated with the sperm

of a single drone have egg-eating policing behavior, just like colonies with naturally mated,

highly polyandrous queens. Chapter 2 is an ESS-style model of worker reproduction in

honeybee colonies, suggesting that workers may invest in selfish reproduction if they sense the

queen may be about to die. Chapter 3 (published in BMC Evolutionary Biology) addresses the

evolution of multiple mating in the Vespine wasps. First, I used microsatellite markers to

describe how many patrilines are present in colonies of five wasp species, four of which are in

the enigmatic Vespula rufa species group, and the last a facultative social parasite of another

species of yellowjacket. I also performed a comparative analysis of paternity number and

paternity skew across 21 species of yellowjacket wasps and hornets (Vespidae: Vespinae).

Species with larger colonies have higher average paternity frequencies and lower average

paternity skew, with interesting implications for the evolution of polyandry in this group.

Chapters 4 and 5 focus on the adaptive significance of matricide in Dolichovespula arenaria, an aerially nesting yellowjacket wasp. I describe matricide for the first time in this species, and use experiments and genetic analyses to show that in natural and lab colonies, queens that are killed

are typically those who have mated few times, or who use sperm in a strongly biased way, resulting in high worker relatedness. Queens who have mated multiply and use sperm evenly are

rarely killed, supporting the hypothesis that workers kill queens as a result of conflict over the

production of males. Experiments suggested that queens laying only male eggs do not trigger

matricide, nor does an abrupt drop in queen fecundity, contrary to theoretical predictions.

Chapter 6 examines the evolution of cuticular hydrocarbon diversity across the polistine wasps,

and provides evidence that the diversity of recognition compounds correlates with social

organization, suggesting that these compounds have evolved in response to their function in

recognition behavior.

BIOGRAPHICAL SKETCH

Kevin was born and raised in Lincoln, Nebraska in a family who loves the natural world.

Frequent trips to the American West and abroad almost certainly bent him toward biology and graduate school. He attended the Lincoln Public Schools, with two years at “Zoo School”, the science focus program housed at the Lincoln Children’s Zoo. He left Nebraska for the University of Wisconsin, Madison, graduating in 2007 with degrees in Mathematics and Zoology. Most influential was the time spent in Costa Rica in 2006: a semester in Monteverde with CIEE, followed by a summer near Cañas, collecting data for a senior thesis project on nest construction in a social wasp. Here and afterward he was inspired by his advisor, Bob Jeanne, to take a stab at a career studying the social behavior of insects, particularly social wasps. But the reason he went to Costa Rica in the first place was the rave review from Elizabeth Hunter, his future spouse, who he met in Madison in 2005. After graduating, they spent a whirlwind year traveling, working field jobs in the west, and volunteering in Costa Rica, before moving to Ithaca, NY where he started grad school in 2008After several early field seasons working with tropical wasps in Costa Rica, he switched to projects in Ithaca with honey bees and local yellowjacket wasps. Along the way he spent considerable time in Syracuse, where Elizabeth lived during her

Masters, and Athens, GA, where she is working on a PhD. Next, they move to Riverside, CA, for a postdoc with Erin Wilson Rankin.

ACKNOWLEDGMENTS This dissertation was the result of seven years of support, discussion, collaboration and encouragement from many, many mentors, collaborators, friends and family members. First, I am indebted to my committee who gave me the freedom to explore the questions I wanted with the organisms I chose. Tom Seeley was a supportive advisor who let me figure out what I wanted to study, and made sure I had what I needed to make fieldwork a success at Liddell and elsewhere. Without his support and patience, my various projects may have been cut short before the exciting results turned up. It was also wonderful to spend a summer learning honey bees with Tom. Heather Mattila was also very generous and fun to work with that summer, hosting me at Wellseley and providing advice and encouragement later on. Kern Reeve shares my fascination with conflict within cooperative groups, and was at the center of the group of students and postdocs working on this topic when I first got to Cornell. His modeling advice, excitement, and typically brilliant suggestions about additional analyses or overlooked connections to previous work made meetings valuable and fun. Although Paul Sherman was only around for half of my time at Cornell, he was very encouraging of my work on matricide and pushed me to test my hypotheses more directly. Cole Gilbert was a phenomenal outside committee member. He helped me develop surgical methods, was always happy to chat about my work and about science in general, and he gave great comments on drafts of this manuscript. I must also express great gratitude to several non-committee faculty. First, I thank Rob Raguso, for training me in GC-MS methods and for helping me with so much of my methodology. Rob was generous with his lab space, his knowledge and his time. I regret that the CHC project didn’t end up being a collaboration with him, but I always very much appreciated our conversations, and his voluminous advice. Conversations with Kerry Shaw in my last year at Cornell really opened my eyes to the wider field of evolutionary biology and the different approaches to tackling evolutionary questions. Her friendship and mentorship were an unforeseen but extremely valuable part of my time at Cornell. Finally, Patrizia D’Ettorre, at U. Paris 13, generously trained me in CHC analysis methods, hosted me in Paris, and has been a

vii

great collaborator. For my comparative analysis in Chapter 6, Rick Hoebeke, Jason Dombroskie, Jim Carpenter, Jim Hunt and especially Joan Strassmann were extremely generous with their time and specimens. The vast majority of my training and support came from other graduate students and postdocs. Martin von Arx generously trained me to do EAG recordings. Julian Kapoor taught me microsatellite data collection and analysis, and most of my results would not exist were it not for his generosity. Julian has been a phenomenal sounding board for ideas and an inspirational field biologist, engineer, programmer, chef and friend. Julie Miller, Jessie Barker, Caitlin Stern all share my passion for the evolution of cooperation and conflict, and provided the core group I could always rely on for feedback and interesting discussion. They are also great friends. Michael Smith is a logistical genius, a great sounding board and collaborator, and a creepy but good friend. David Peck provided great conversations about parasites, social insects, experimental design and many other things. Paul Shamble made me understand how amazing mimicry is, and how curious naturalists should think. Becky Cramer is/was a great role model for productivity, curiosity and work/life balance. Barrett Klein provided great advice and was a wonderful role model and friend during our shared summer at Liddell. Gil Menda has been a great friend and fellow bee and wasp enthusiast. Shinichi Asao, Patty Jones, Susan Whitehead, Bonnie Waring, Kellie Kuhn, Bernal Matarrita and Danilo Brenes made La Selva fun, despite my experimental woes. Ben Freeman, Nick Mason, and Jake Berv helped me understand phylogenetics and comparative analyses. Shane Peace is an invaluable friend who helped me stay sane. Matt Lewis provided excellent rants. My parents have always been encouraging, supportive and proud of me. I’m grateful to my brother, Garrison, for his help and his patience putting up with my tropical flailing during a season in Costa Rica. Finally, and most of all, I thank Elizabeth Hunter, for helping me appreciate my successes, get over my failures, learn statistics, think about ecology, and live a happier and more balanced life.

viii

TABLE OF CONTENTS Biographical Sketch ...... vi

Acknowledgements ...... vii

Chapter 1. No facultative worker policing in the honey bee (Apis Mellifera L.) ...... 1 References ...... 11 ! ! Chapter 2. Queen loss and worker reproduction in honey bees and other social insects ...... 14 References ...... 44 Appendix 2.1. Selfish effort as the product of the probability of selfish investment and degree of selfish investment ...... 49! Appendix 2.2. Male reproductive value ...... 50 ! ! Chapter 3. Colony size is linked to paternity frequency and paternity skew in yellowjacket wasps and hornets ...... 55 References ...... 80 Appendix 3.1. Supplementary figures and tables ...... 87 !

! Chapter 4. Queen killing is linked to high worker-worker relatedness in a social wasp ...... 94 References ...... 110 Appendix 4.1. Supplementary figures and tables...... 113! Appendix 4.2. A description of observed matricides ...... 116! ! ! Chapter 5. Matricide, queen sex allocation, and queen fecundity in a yellowjacket wasp ...... 119 References ...... 145 Appendix 5.1. Details of observation box setup and surgical methods ...... 149! !

! Chapter 6. Social organization predicts cuticular hydrocarbon diversity in paper wasps ...... 153 References ...... 171 Appendix 6.1. Supplementary figures and tables ...... 177 ! !

LIST OF FIGURES

Figure 1.1. Mean number of queen-laid and worker-laid eggs remaining 24 hours after transfer into colonies with SDI, MDI and naturally mated queens...... 8 Figure 2.1. Possible outcomes during a focal worker’s lifetime...... 20 Figure 2.2. Worker selfishness increases with the probability of queen loss, and with queen mating frequency ...... 26 Figure 2.3. Worker selfish strategy when workers can replace the queen but cannot reproduce in queenright colonies (Model 2 when S = 0)...... 30 Figure 2.4. Evolutionarily stable worker selfish investment as a function of p, the probability of failing to re-queen, for several colony sizes and strengths of policing (Model 2 with S>0)...... 32 Figure 3.1. Aerial and subterranean yellowjacket nests...... 58 Figure 3.2. Predicted effects of colony size and nest site on paternity traits...... 60 Figure 3.3. Phylogeny and trait data used in comparative analyses...... 74 Figure 3.4. Intracolony relatedness, effective paternity and colony size in vespine wasps...... 75 Figure S3.1. Phylogeny used in analyses shown in Table S3.3...... 93 Figure 4.1. Queen and worker reproduction in Dolichovespula arenaria...... 96 Figure 4.2. Matricide, paternity and inbreeding in D. arenaria ...... 98 Figure S4.1. Paternity skew, queen loss and inbreeding ...... 114 Figure S4.2. Nest boxes used for matricide observation...... 115 Figure 5.1. Investigating matricide in Dolichovespula arenaria...... 125 Figure 5.2. Colony comb allocation and the timing of queen death...... 132 Figure 5.3. Stored sperm counts for D. arenaria queens ...... 135 Figure S5.1. A queen about to have her ovaries removed ...... 152 Figure 6.1. Cuticular hydrocarbon evolution in paper wasps...... 158 Figure S6.1. CHC diversity of 34 paper wasp species across social types and latitude...... 177 Figure S6.2. Alkane relative abundance and average chain length, as a function of social type and latitude...... 178 Figure S6.3. Extraction of CHCs from a pinned specimen of Ropalidia marginata...... 179 Figure S6.4. GC-MS traces of two Vespula maculifrons (Vespidae: Vespinae) worker CHC profiles ...... 180 Figure S6.5. Constraint tree used to generate trees ...... 181 Figure S6.6. Consensus Tree (50%) from BEAST output ...... 182

LIST OF TABLES Table 2.1. A summary of model variables and parameters ...... 22 Table 2.2. Life-for-life relatedness values for a focal worker ...... 24 Table 2.3. Probability of queen loss during mating flights for different populations of honey bees ...... 39 Table 3.1. Summary data from paternity analysis of five Vespula species...... 66 Table 3.2. PGLS models of the effect of colony size and nest site on paternity traits ...... 73 Table S3.1. The number of alleles and expected heterozygosity at each locus...... 87 Table S3.2. Descriptive data for colonies of Vespula acadica, V. atropilosa, V. consobrina, V. vidua, V. flavopilosa ...... 88 Table S3.3. Comparative analyses of colony size and nest site on effective mating frequency across 21 species of Vespine wasps using alternative phylogeny ...... 90 Table S3.4. Comparative analyses of maximum colony size and nest site on effective mating frequency ...... 91 Table S3.5. Genbank accession numbers for sequences used in phylogenetic anaylsis...... 92 Table S4.1. Detailed colony data ...... 113 Table 5.1. Genotypes reveal sex of the queen-laid brood at the time of matricide ...... 133 Table 5.2. Outcome of surgery experiments ...... 137 Table 6.1. Models of CHC diversity...... 159 Table S6.1. Models of CHC diversity (expanded) ...... 183 Table S6.2. Robustness analysis of preferred models with ! set to upper 95% CI value ...... 184 Table S6.3. Models of CHC diversity excluding 4 species from the literature ...... 185 Table S6.4. Diversity models using 0.1% rmean relative abundance threshold ...... 186 Table S6.5. Relative abundance and chain length models ...... 187 Table S6.6: Sample summary ...... 188 Table S6.7. Accession numbers for sequences used to create trees ...... 189

xi CHAPTER 1

NO FACULTATIVE WORKER POLICING IN THE HONEY BEE (APIS MELLIFERA L.)

Kevin J. Loope, Thomas D. Seeley, Heather R. Mattila

Published in Naturwissenschaften

Abstract

Kin selection theory predicts that in colonies of social Hymenoptera with multiply mated queens,

workers should mutually inhibit (“police”) worker reproduction, but that in colonies with singly

mated queens workers should favor rearing workers’ sons instead of queens’ sons. In line with

these predictions, Mattila et al. (2012) documented increased ovary development among workers

in colonies of honey bees with singly mated queens, suggesting that workers can detect and

respond adaptively to queen mating frequency, and raising the possibility that they facultative

police. In a follow-up experiment, we test and reject the hypothesis that workers in single-

patriline colonies prefer worker-derived males and are able to reproduce directly; we show that

their eggs are policed as strongly as those of workers in colonies with multiply mated queens.

Evidently, workers do not respond facultatively to a kin structure that favors relaxed policing and

increased direct reproduction. These workers may instead be responding to a poor queen or

preparing for possible queen loss.

Introduction

In many social insect species, workers favor rearing eggs that are laid by queens and eat

eggs that are produced by other workers ( Ratnieks and Visscher 1989; Ratnieks et al. 2006;

Wenseleers and Ratnieks 2006). Such mutual inhibition of worker reproduction is called

policing and is thought to have contributed to the evolution of complex insect societies by

1 reducing conflict over who produces males (Ratnieks et al. 2006; Wenseleers and Ratnieks 2006) and aligning workers’ interests in rearing their queens’ offspring (Bourke 1999; Wenseleers et al.

2004; Ratnieks and Helantera 2009). Because workers are more related to their own sons than to the sons of other colony members, kin selection theory predicts that workers may attempt to reproduce directly, even at the expense of colony productivity (Hamilton 1972). However, mutual eating of worker-laid eggs is also predicted to be advantageous for workers if they are more related to their queen’s sons than to the sons of other workers (Starr 1984; Woyciechowski and Lomnicki 1987; Ratnieks 1988). For species with one queen per colony, worker-worker policing is favored when queens have an effective mating frequency (me) greater than two because multiple mating dilutes worker relatedness to nephews but not to brothers. Conversely, worker policing is not favored when me is less than two, because workers are more related to nephews than to brothers. An alternative to this relatedness hypothesis for worker egg eating is the colony productivity hypothesis, which suggests that workers eat eggs to increase total colony output regardless of kin structure, by avoiding either costly laziness by egg-laying workers

(Ratnieks 1988) or costly investment in low viability brood produced by workers (Pirk et al.

2004; Nonacs 2006).

Empirical support for the relatedness hypothesis for worker policing comes largely from interspecific comparisons showing that the incidence of policing increases and percentage of males produced by workers decreases with decreasing worker-worker relatedness (Wenseleers and Ratnieks 2006). However, kin selection theory also predicts between-colony differences in policing for species in which some queens mate multiply and others mate singly (and for species with variable queen numbers; Hammond et al. 2003). The evolution of facultative policing based on relatedness differences requires: 1) natural between-colony variation in intracolony

2 relatedness that spans the predicted policing threshold, 2) information detectable to workers that

indicates intracolony relatedness, and 3) absence of other benefits of policing (such as avoiding

reduced colony productivity) that would favor it regardless of relatedness benefits. If these

requirements are met, then the relatedness hypothesis predicts that policing will be reduced or

absent in colonies possessing a queen with me < 2.

Although worker policing was first described in honey bees (Ratnieks and Visscher

1989), they are not an obvious choice for studying facultative policing in relation to queen mating frequency, primarily because honey bee queens nearly always mate multiply (Tarpy and

Nielsen 2002), and thus likely do not meet the aforementioned first requirement. As for the second and third requirements, it is unclear whether kinship information is available to and detectable by workers (Visscher 1986; Arnold et al. 1996) or whether there are productivity benefits of policing in this species (see Pirk et al. 2003 and Discussion). However, Mattila et al.

(2012) recently found that honey bee workers develop their ovaries more and work less in colonies headed by single-drone inseminated (SDI) queens versus colonies headed by multiple- drone inseminated (MDI) queens. This discovery raises the possibility that workers in colonies with singly mated queens can detect and can respond adaptively to low queen mating frequency.

This in turn suggests the intriguing hypothesis that honey bee workers in SDI colonies have reduced policing and, as a result, facultatively activate their ovaries to produce sons (Mattila et al. 2012; Van Zweden et al. 2012). We tested this explanation for worker ovary activation in

SDI colonies by comparing policing rates among colonies headed by SDI queens, MDI queens, and naturally mated queens. If honey bee workers do not relax policing when queens are singly mated, then this would suggest that one or more of the requirements for the evolution of facultative policing are not met for this species. It would also suggest that worker ovaries are

3 activated in colonies headed by singly mated queens for another reason that is independent of colony kin structure, such as the presence of a weak queen whose death could lead to reproductive opportunities for ovary-activated workers (Visscher 1989; Mattila et al. 2012).

Methods

Our test of policing involved a standard assay (Ratnieks and Visscher 1989) where worker-laid and queen-laid eggs from source colonies are transferred into test colonies and egg survival is measured over 24 hours.

Source Colonies for Eggs

To acquire worker-laid and queen-laid eggs, we divided queenright colonies in Ithaca,

NY to form pairs of queenright and queenless colonies. We split each colony by removing its queen, 2-4 frames of brood and adult bees, and several frames of food from the original hive and installing them in a new hive. The queenless portion remained in the original hive with most of the brood and adult bees. All colonies were checked weekly for eggs. Queen cells were removed from the queenless colonies to ensure that they remained queenless. We split five colonies on June 4, 2010 and four more on June 18. On July 11, we selected as our source colonies the 3 pairs of colonies that had strong egg laying by both queens (queenright colonies) and workers (queenless colonies).

Test Colonies for Policing Assay

To compare egg eating in colonies with singly mated (SDI) and multiply mated (MDI) queens, we obtained 9 queens of each type from a queen breeder (Glenn Apiaries, Fallbrook,

4 CA) who performs instrumental inseminations. All the test-colony queens were full sisters of

Apis mellifera carnica and drones were of mixed ancestry. The semen for the instrumental inseminations was collected from drones chosen randomly from a pool of 1,000 drones that came from 20 unrelated colonies. SDI queens received semen from one drone, while MDI queens received semen from 15 drones (mixed by stirring with a glass rod before insemination). SDI and

MDI queens received the same volume of semen (1 !L per queen). We introduced the 18 queens into test colonies that were maintained in Wellesley, MA. These introductions were performed on 18 May, 8 weeks prior to the policing tests, which ensured that each test colony’s workers were daughters of the colony’s queen at the time of the assay. To check for possible effects of working with instrumentally inseminated queens, we also performed tests of policing with 9 colonies headed by naturally (multiply) mated queens.

Policing Assay

On July 12, the 3 pairs of egg-source colonies were moved from Ithaca, NY to Wellesley,

MA and the policing assay was performed from July 14 to July 19. Twenty-four hours prior to egg transfer, frames of empty drone comb (comb built of drone cells) were inserted into source colonies (queenright and queenless) from a single pair, as well as into a trio of test colonies (SDI,

MDI, and naturally mated). In the test colonies, each frame of comb was placed inside a cage with walls made of queen excluder screen, enabling workers to clean the comb’s cells but preventing the queen from laying eggs in them (these frames would receive eggs from the source colonies shortly). In the source colonies, workers (in queenless colonies) and the queen (in queenright colonies) were free to lay eggs on the drone comb frames for the 24-hour period, ensuring that all eggs were less than 1 day old. The next day, these frames were removed from

5 all hives so that eggs could be transplanted out of frames from source colonies into frames from test colonies. Each test colony was given 30 eggs from both source colonies in a pair (queenright and queenless); the eggs were deposited in two adjacent rows of cells using modified forceps

(Taber 1961). When frames were not being handled, they were placed in an incubator (34ºC); combs containing eggs were covered with damp paper towel to prevent desiccation. Frames with rows of eggs were then placed back inside the queen excluder cages inside the test colonies.

These cages ensured any missing eggs were removed by workers and not the queen, who was also prevented from laying eggs in the focal cells. The number of eggs remaining in each row was checked after 24 hours.

We assayed 9 trios of test colonies (sets with SDI, MDI and naturally mated queens; 27 colonies in total). Source colony pairs A, B and C provided the eggs for four, three, and two of the test trios, respectively. In the second trial using eggs from colony pair C, all queen-laid and worker-laid eggs were removed by workers, possibly due to egg desiccation during transplanting.

This trial was removed from the analysis.

Statistical Analysis

To determine whether mating frequency affected the propensity of workers to favor queen-laid over worker-laid eggs, we analyzed the effect of queen type (SDI, MDI, naturally mated) and egg type (queen-laid, worker-laid) on the number of transferred eggs that survived the 24-hour assay using a two-way ANOVA, with a random effect of source colony pair. This test was performed in R 2.9.2 (R Development Core Team 2012). Because our data violated the assumption of homogeneity of variances for ANOVA, we performed an alternative test that did not; this alternative test yielded similar results (see Appendix 1.1).

6 Results

Mating frequency of queens did not affect survival of queen-laid or worker-laid eggs

after 24 hours in colonies (Figure 1; two-way ANOVA; effect of queen type: F2,40 = 0.38, p >

0.5; queen type "egg type interaction, F2,40 = 0.41, p > 0.5). Worker policing was strong across

all colonies, as many more queen-laid eggs survived than worker-laid eggs (Figure 1; two-way

ANOVA; effect of egg type: F1,40 = 106.52, p < 0.0001). Of the 810 worker-laid eggs that were transplanted into test colonies, only two eggs remained after 24 hours (0.08 ± 0.06 eggs per colony, SEM), whereas 340 of the 810 queen-laid eggs remained after 24 hours (12.6 ± 1.2 eggs per colony).

Discussion

This study was done to determine whether facultative worker policing based on relatedness can explain increased development of workers’ ovaries in honey bee colonies with

SDI queens (Mattila et al. 2012). Our results show unambiguously that workers in the colonies that we studied did not facultatively police as a function of their colony’s kin structure. Workers in colonies with singly inseminated queens removed worker eggs at a high rate that was similar to colonies with multiply inseminated and naturally mated queens.

The absence of facultative policing in honey bees is consistent with what has been found in the other two social insect species that have been tested for this phenomenon. In the wasp

Dolichovespula saxonica, Bonckaert et al. (2011) found no evidence for facultative policing based on mating frequency, and concluded that earlier hints of facultative policing in this species

(Foster and Ratnieks 2000) were likely due to a small sample size and a confound of relatedness with colony developmental stage. Similarly, in the ant Leptothorax acervorum, Hammond et al.

(2003) found that variation in intracolony relatedness due to differences in queen number did not

#" #! " -./0123..34536615.37/808065/293.5:;5<.1 ! $% &% $% &% $% &% '() *() +,-

Figure 1.1. Mean (± SEM) number of queen-laid (QL) and worker-laid (WL) eggs remaining 24 hours after transfer into colonies with SDI, MDI and naturally mated queens. For each colony, 30 eggs of each type were transferred (n = 8 colonies for each type of queen).

8 predict attempts by workers to reproduce. It appears, therefore, that social Hymenoptera

typically do not meet the requirements that are necessary for the evolution of facultative

policing, and that predictions regarding levels of policing based on relatedness may be upheld

only for interspecific comparisons based on average relatedness structure.

Why don’t honey bees exhibit facultative policing? Observations of non-policing

colonies suggest natural genetic variation for policing exists and could be a target for selection

(Beekman et al. 2002). The evolution of a facultative response to mating frequency would

require natural variation in me spanning the policing threshold, but such variation is evidently not

present in A. mellifera. Honey bee queens nearly always mate multiply. Of the 113 colonies

summarized by Tarpy and Nielson (2002), only 8 had me < 2, and estimates of me may be artificially low for these colonies because a only small number of workers were genotyped.

Even if variation in mating frequency was sufficient for facultative policing, its evolution would still require a mechanism for workers to assess mating frequency (or patriline diversity).

Theoretical arguments suggest that this information may not be evolutionary stable (e.g. Ratnieks

1991), though sufficient discriminatory information is present in some species (Sundström et al.

1996; Boomsma et al. 2003; van Zweden et al. 2010; Nehring et al. 2011), and may be present in the honey bee (Arnold et al. 1996).

Alternatively, it is possible that honey bees can detect and respond to mating frequency, but that selection maintains policing even in colonies with singly mated queens because colony- level productivity benefits discourage workers from reproducing (Ratnieks 1988, Pirk et al.

2004). This is an attractive but difficult-to-test hypothesis that is often proposed to explain the presence of worker policing in species in which colony kin structure does not predict that it should exist (e.g. Hammond and Keller 2004) and it has been invoked to explain why A.

9 mellifera capensis workers police worker-laid eggs in the absence of relatedness benefits (Pirk et al. 2003).

Regardless of the selective basis for policing in honey bees, our results refute the hypothesis that workers activate their ovaries in queenright colonies with singly mated queens because there is reduced worker policing in these colonies and thus increased opportunities for direct reproduction by workers in these colonies. This finding points to an alternative hypothesis to explain why more workers undergo ovary development in colonies headed by singly inseminated queens: workers assess a singly mated queen as one who is weak and likely to fail

(Mattila et al. 2012). The weak-queen hypothesis proposes that queen reliability, not colony kin structure, influences worker ovarian development. Workers may prime themselves for direct reproduction as the possibility of being able to reproduce directly in a queenless colony increases, even at a cost to the still queenright colony (Visscher 1989; Mattila et al. 2012).

10 REFERENCES

Arnold G, Quenet B, Cornuet JM, Masson C, Deschepper B, Estoup A, Gasqui P (1996) Kin recognition in honeybees. Nature 379: 498

Beekman M, Good G, Allsopp MH, Radloff SE, Pirk CWW, Ratnieks FLW (2002) A non- policing honeybee colony (Apis mellifera capensis). Naturwissenschaften 89:479-482

Bonckaert W, Van Zweden JS, d'Ettorre P, Billen J, Wenseleers T (2011) Colony stage and not facultative policing explains pattern of worker reproduction in the Saxon wasp. Mol Ecol 20:3455–3468

Boomsma JJ, Nielsen J, Sundström L, Oldham NJ, Tentschert J, Petersen HC, Morgan ED (2003) Informational constraints on optimal sex allocation in ants. Proc Natl Acad Sci USA 100:8799–8804

Bourke AFG (1999) Colony size, social complexity and reproductive conflict in social insects. J Evol Biol 12:245–257

Foster KR, Ratnieks FLW (2000) Facultative worker policing in a wasp. Nature 407:692–693

Hamilton WD (1972) Altruism and related phenomena, mainly in social insects. Annu Rev Ecol Syst 3:193–232

Hammond RL, Bruford M, Bourke AFG (2003) Male parentage does not vary with colony kin structure in a multiple-queen ant. J Evol Biol 16:446–455

Hammond RL, Keller L (2004) Conflict over male parentage in social insects. PLoS Biol 2:E248

Mattila HR, Smith ML, Reeve HK (2012) Promiscuous honey bee queens increase colony productivity by suppressing worker selfishness. Curr Biol 22:2027–2031

Nehring V, Evison SEF, Santorelli LA, d’Ettorre P, Hughes WOH (2011) Kin-informative recognition cues in ants. Proc R Soc Lond B 278:1942–1948

Nonacs P (2006) Nepotism and brood reliability in the suppression of worker reproduction in the eusocial Hymenoptera. Biol Lett 2:577–579

Pirk CWW, Neumann P, Ratnieks FLW (2003) Cape honeybees, Apis mellifera capensis, police worker-laid eggs despite the absence of relatedness benefits. Behav Ecol 14:347-352

Pirk CWW, Neumann P, Hepburn HR, Moritz RFA, Tautz J (2004) Egg viability and worker policing in honey bees. Proc Natl Acad Sci USA 101:8649–8651

R Core Team (2012) R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing. Vienna, Austria

Ratnieks FLW (1988) Reproductive harmony via mutual policing by workers in eusocial

11 Hymenoptera. Am Nat 132:217–236

Ratnieks FLW (1991) The evolution of genetic odor-cue diversity in social Hymenoptera. Am Nat 137:202–226

Ratnieks FLW, Foster KR, Wenseleers T (2006) Conflict resolution in insect societies. Annu Rev Entomol 51:581–608

Ratnieks FLW, Helantera H (2009) The evolution of extreme altruism and inequality in insect societies. Philos Trans R Soc B 364:3169–3179

Ratnieks FLW, Visscher PK (1989) Worker policing in the honeybee. Nature 342:796–797

Starr CK (1984) Sperm competition, kinship, and sociality in the aculeate Hymenoptera. In: Smith R (ed) Sperm competition and the evolution of animal mating systems. Academic Press, Orlando, FL, pp 428–459

Sundström L, Chapuisat M, Keller L (1996) Conditional manipulation of sex ratios by ant workers: a test of kin selection theory. Science 274:993–995

Taber S (1961) Forceps design for transferring honey bee eggs. J Econ Entomol 54:247–250

Tarpy DR, Nielsen DI (2002) Sampling error, effective paternity, and estimating the genetic structure of honey bee colonies (Hymenoptera: Apidae). Ann Entomol Soc Am 95:513–528

Van Zweden JS, Brask JB, Christensen JH, Boomsma JJ, Linksvayer TA, d’Ettorre P (2010). Blending of heritable recognition cues among ant nestmates creates distinct colony gestalt odours but prevents within-colony nepotism. J Evol Biol 23:1498–1508

Van Zweden JS, Cardoen D, Wenseleers T (2012) Social Evolution: When Promiscuity Breeds Cooperation. Curr Biol 22:R922–R924

Visscher PK (1986) Kinship discrimination in queen rearing by honey bee colonies (Apis mellifera). Behav Ecol Sociobiol 18:453-460

Visscher PK (1989) A quantitative study of worker reproduction in honey bee colonies. Behav Ecol Sociobiol 25:247–254

Wenseleers T, Hart A, Ratnieks FLW (2004) When resistance is useless: Policing and the evolution of reproductive acquiescence in insect societies. Am Nat 164:E154–E167

Wenseleers T, Ratnieks FLW (2006) Comparative analysis of worker reproduction and policing in eusocial hymenoptera supports relatedness theory. Am Nat 168:E163–E179

Woyciechowski M, Lomnicki A (1987) Multiple mating of queens and the sterility of workers among eusocial hymenoptera. J Theor Biol 128:317–327

12 Appendix 1.1. Methods for Alternative Statistical Approach

With the goal of conducting a statistical analysis that did not violate the assumption of homogeneity of variances for ANOVA, we calculated the difference in the number of surviving queen-laid and worker-laid eggs for each test colony to determine whether mating frequency affected the propensity of workers to police. A high positive difference would represent a strong bias in favor of queen-laid eggs, whereas a difference close to zero would represent no bias in favor of queen-laid eggs, and thus weak or absent worker policing. We compared this measure of policing among the treatment groups (SDI, MDI and naturally mated) using a one-way

ANOVA, blocking by source colony pair. Because there was no effect of source colony pair or treatment group on this difference (see below), we pooled colonies from all queen treatments and determined using a paired t-test whether worker policing was present in our colonies by asking whether there were on average more queen-laid than worker-laid eggs remaining after 24 hours.

Statistical tests were performed in R (R Development Core Team 2012).

Results for Alternative Statistical Approach

Mating frequency did not affect the degree of worker policing in colonies, measured as the difference in the number of queen-laid and worker-laid eggs that survived the 24-hour policing assay (Figure 1; one-way ANOVA, F2,21 = 0.401, p = 0.675; no significant block effect of source colony: p = 0.957). Worker policing was strong across all colonies, as many more queen-laid eggs survived than worker-laid eggs (Figure 1; paired t-test, df = 23, t = 10.7, p < 0.001).

13 CHAPTER 2

QUEEN LOSS AND WORKER REPRODUCTION IN HONEY BEES AND OTHER SOCIAL

INSECTS

Kevin J. Loope and H. Kern Reeve

Abstract

Explaining the adaptive basis of worker reproduction in social insect groups is an essential step in understanding the evolutionary transition from simple to complex societies. Recent empirical work suggests that under certain circumstances, worker honey bees facultatively develop their ovaries in queenright colonies. Mattila et al. (2012) present evidence that singly-mated queen colonies exhibit a higher degree of worker ovary activation than do multiply-mated colonies, while Woyciechowski and Kuszewska (2012) show that workers finish development primed for reproduction during a narrow window of queenlessness in the natural swarming process. Both of these phenomena can potentially be explained by (i) a change in colony kin structure, or (ii) a perceived risk of queen loss. Indeed, both selective pressures can operate in concert, making it difficult to verbally assess their relative importance in determining optimal worker selfishness.

Here, we present a game theoretic model investigating the interacting roles of relatedness, policing and the probability of queen loss in determining worker investment in personal reproduction. Our results suggest that the risk of queen loss and the probability of failing to replace a lost queen are overlooked factors that can maintain costly conflict over reproduction in

“super-organisms” like honey bees, despite policing and other mechanisms promoting cooperation.

14 Introduction

Eusocial insects present some of the most extreme examples of reproductive altruism in the animal kingdom (Wilson 1971). Within the eusocial Hymenoptera, species vary greatly in the degree to which reproduction within colonies is monopolized by one or a few individuals. Some societies consist of a single queen, who dominates reproduction in the colony, and workers who lack ovaries and thus are incapable of laying eggs (Holldobler and Wilson 1990). At the other end of the spectrum are species with colonies of entirely totipotent individuals, all capable of mating and producing offspring (Peeters and Crewe 1985; Strassmann et al. 2002). However, most species lie somewhere in the middle, typically with a morphologically distinct queen or queens, and workers who cannot mate but can compete with queens over the production of a colony’s males due to their haplodiploid genetic system. Within and between these species, there is great variation in the extent to which workers attempt to reproduce (Bourke 1988;

Hammond and Keller 2004; Wenseleers and Ratnieks 2006a). Understanding the selective factors that favor worker sterility is important because it appears to be associated with the transition from simple to complex societies, one of the major hierarchical transitions in evolution

(Bourke 1999; Bourke 2011). Although the evolutionary basis for worker reproduction has been a major focus of study, much variation remains to be explained (Hammond and Keller 2004;

Wenseleers and Ratnieks 2006a; Wenseleers et al. 2013).

Worker reproductive self-restraint can result from acquiescence due to coercion by other colony members (Ratnieks 1988; Ratnieks and Helantera 2009; Bourke 2011), as well as high relatedness to reproductive individuals within the colony. Coercive policing reduces the benefits of attempted reproduction to a reproductive individual relative to the cost to the colony of attempted reproduction (Wenseleers et al. 2004b). The evolution of mutual worker policing, and

15 thus of reduced worker reproduction, is favored by colony kin structures in which workers are more related to the sons of the queen or queens than to the sons of other workers (Ratnieks 1988;

Wenseleers et al. 2004b; Wenseleers and Ratnieks 2006a), as well as by sex-ratio and ergonomic benefits to workers (Foster and Ratnieks 2001a; Ohtsuki and Tsuji 2009; Wenseleers et al. 2013).

Kin structure also influences how much workers value the offspring of queens and other workers, and therefore the indirect benefits and costs of competing with the queen and other workers for reproduction. In this way, kin structure should also directly influence self-restraint beyond the effect of policing. Comparative analyses suggest that interspecific patterns of relatedness, policing, and worker reproduction are consistent with theoretical predictions regarding when policing should evolve, and policing is clearly a central factor explaining much of the interspecific variation in worker reproduction in insect societies (Wenseleers and Ratnieks

2006a; Wenseleers and Ratnieks 2006b).

However, variation in kin structure and policing does not explain intraspecific variation in worker reproductive investment (Hammond et al. 2003; Bonckaert et al. 2011; Loope et al.

2013), and there remain large differences in worker reproductive investment between species that cannot be explained solely by the presence or absence of worker policing (Bourke 1988;

Toth et al. 2004; Hammond and Keller 2004; Wenseleers and Ratnieks 2006a). Furthermore, it appears that workers of many species exhibit partial ovarian development (Smith et al. 2013), despite little opportunity for reproduction in queenright (queen-possessing) colonies due to policing. Motivated by these observations, and the recent discoveries of facultative ovarian development in honey bee (Apis mellifera L.) workers associated with experimentally manipulated queen mating frequency (Mattila et al. 2012) and swarming (Woyciechowski and

16 Kuszewska 2012), we develop a model exploring the hypothesis that the risk of queen loss may explain why workers of honey bees and other social insects sometimes invest in partial or complete ovarian development, even in the face of strong policing. If workers sense that queen loss is increasingly likely, then they may benefit from increasing investment in personal reproductive ability to prepare for the possibility of directly reproducing in a queenless colony.

Once the queen does die, workers that are more reproductively mature would be expected to outcompete those that are less reproductively mature in the ensuing scramble competition to lay male eggs. Even if the colony successfully replaces a queen, workers may also benefit from increased selfish investment at this point because they are less related to a replacement queen than to their mother (Woyciechowski and Kuszewska 2012). Both of these factors may interact in the evolution of worker selfishness, depending on the species.

Because of the ties to recent empirical work, we focus our analysis on honey bees (Apis mellifera and congeners), though the model applies to a variety of social insect species. Queen loss may be particularly important in the evolutionary ecology of honey bees, as colonies frequently undergo risky queen replacements (supersedures) and reproductive events, wherein the colony must rear a new queen to replace an old one that has departed with a swarm. Queens typically live fewer than 4 years, and workers will replace queens that are injured, diseased, or have reduced fecundity or depleted sperm stores leading to male-laying (Winston 1987). Most colonies in cavities of natural size will swarm at least once a year (Seeley 1978; Winston 1980).

If a colony fails to rear a replacement queen in either scenario, it runs the risk of becoming permanently queenless, leading to a period of worker reproduction prior to colony death

(Winston 1987).

17 In this paper we build a proof-of-concept model (Servedio et al. 2014) to evaluate the verbal hypothesis that queen loss can promote selfishness in sterile workers under a variety of conditions. First, using an ESS analysis on worker inclusive fitness, we consider how the possibility of queen loss within a focal worker’s lifetime changes her optimal investment in selfish effort under a simple scenario with no queen replacement. We then investigate two more realistic complications; the addition of queen replacement and of imperfect worker policing in queenright colonies. This provides a deeper exploration of the verbal hypothesis that the possibility of queen loss may explain worker ovarian development (Visscher 1989; Oldroyd et al.

2001; Mattila et al. 2012; Smith et al. 2013; Friend and Bourke 2014), and conversely, that queen longevity, and thus a stable kin structure, promotes altruism by workers in the social

Hymenoptera (Bourke 1999).

The Models

Assumptions

We assume that workers in a colony of size n are the offspring of a single queen who has an effective mating frequency of m. Hereafter we will use relatedness values appropriate for haplodiploids (i.e., social Hymenoptera). We use x to represent the fixed, lifetime selfish investment in personal reproductive ability by an individual worker. This can be thought of as the product of the probability of investing in personal reproduction and the degree of investment in personal reproduction (ranging from 0 to 1; see Appendix 2.1). Thus, a higher x corresponds either a higher probability of becoming reproductive, a greater investment in reproductive effort if reproductive, or both. We assume a tradeoff between investment in personal reproduction and

18 total colony output (see below). Importantly, we also assume that workers can accurately assess the probability of queen loss in their lifetime.

We will examine two models (Figure 2.1). The first represents a simple colony life history in which queens cannot be replaced if they die, and workers cannot reproduce in queenright colonies (perfect policing). In this version of the model, a positive value for x does not imply that workers are actively laying eggs; these workers can be investing in personal reproductive capacity below the level required for active, immediate reproduction. If the queen dies, workers compete in a reproductive tug-of-war over who produces males before the entire colony dies. A critical assumption of this model is that workers who invest selfishly in personal reproductive ability prior to actual queen loss are better competitors for reproduction after queen loss, in proportion to their selfish investment. We consider workers who invest in selfish effort during the queenright phase to be priming themselves for possible queen loss, even though in this scenario, these workers get no personal reproduction if the queen does not die in their lifetime.

The second, more complex model builds on the first to better fit the life history of honey bees and other perennial species that have the ability to rear a replacement queen in the event of queen loss. In this scenario, queen loss is followed by queen replacement, and colonies become queenless only if this replacement fails. We use queen loss to refer to when a queen dies, or when a queen’s fecundity drops below an acceptable level to workers, who then attempt to replace her. We consider old queen departure with a reproductive swarm (as happens in honey bees) functionally equivalent to queen loss for the remaining workers who do not leave the natal colony; a honey bee colony that has just cast a swarm must rear a replacement queen, and thus runs the risk of becoming queenless (Winston 1987). In this model we also allow worker direct reproduction in queenright colonies, with the strength of policing determining the fraction of

!"#$%&' !"#$%&( !"#$% !$%

!"#$% !$% !"#&% !&%

!"##$% !"##$*#&&% !"##$% .,*,$-% !"##$*#&&% 1.2/"3$ &"'()(#& +,*,$- &"'()(#& '#/0"##$& +,*,$- 4*"5657%728 !"#$% !$% !"#$% !"#$%!"#&% !$%!&% )"*+$*,& 1#& *$-*"#./$0 2, 1#& )3%456 )3%456

Figure 2.1. Possible outcomes during a focal worker’s lifetime. Workers only reproduce in the event of queen loss (Model 1), queen loss without replacement (Model 2), or in queenright colonies with imperfect policing (i.e., when S > 0).

20 worker offspring that survive to adulthood. Now, worker selfish reproductive investment (i.e., a positive value for x) represents direct attempts at reproduction in the queenright colony, and is still linearly related to the ability to compete for reproduction in the event of queen loss. All plotted examples assume a sex investment ratio (a colony’s fractional investment in males) at the optimum for workers (Pamilo 1991). However, for nearly all scenarios analyzed, queen and worker optima are both very close to 0.5 (data not shown).

In order to use an inclusive fitness approach (Hamilton 1964) with pedigree relatedness values, we analyze the invasion of a rare mutant continuous strategy, and determine the evolutionarily stable strategy (ESS; Maynard-Smith 1982), assuming weak selection and complete dispersal (no inbreeding or kin competition). All variables and parameter definitions are summarized in Table 2.1.

Model 1: No queen replacement and perfect policing

Let x be the selfish effort of a focal mutant worker, in a population of workers with selfish effort x’. In Model 1, we assume that this selfish investment does not lead to worker reproduction when the colony remains queenright, and will only benefit a worker if the colony becomes queenless, when workers compete over male production. Such selfish effort may include physiological adjustments such as partial or complete ovary activation as well as behavioral investment in low-cost activities that increase potential fecundity or even survival.

We assume for simplicity a linear colony cost function that represents the total colony output given costly worker selfishness: !!

! ! ! ! ! ! ! ! ! ! ! !!

Table 2.1. A summary of model variables and parameters Variable or Definition Parameter Proportion of males in the population that are produced by colonies that have re- A queened, used in calculating male reproductive value in Model 2 (Table 2.2; Appendix 2.2) Proportion of males in the population that are produced by queenless colonies, B used in calculating male reproductive value (Table 2.2; Appendix 2.2) f Probability that the queen dies in the focal worker’s lifetime G Colony productivity, a declining function of x and x’ m Queen effective mating frequency n Number of workers in a colony p Probability that the colony fails to produce a new queen given initial queen loss Proportion of males in the population that are produced by workers in queenright colonies, used in calculating male reproductive value (Table 2.2; Appendix 2.2) Relative fecundity of the queen, compared to a worker’s maximum reproductive q output Regression relatedness between two daughters of a queen mated m times (Table r sib 2.2) s Fraction of a queenright colony’s investment allocated to males Fraction of worker selfish investment in personal reproduction that escapes S policing in queenright colonies (Model 2) VM Reproductive value of males, assuming the reproductive value of females is 1

W Direct fitness of the focal worker playing strategy x. W’ Direct fitness of a non-focal worker playing strategy x’ WQM Male component of queen direct fitness WQF Female component of queen direct fitness Fraction of energy a mutant focal worker devotes to personal reproduction, i.e. the x worker’s selfish investment strategy x’ The selfish strategy of all the other workers in the population x* The ESS worker selfish strategy

22 With probability f, the queen dies during the lifetime of a focal worker. If the queen dies,

workers that have already invested in their own reproductive ability compete with each other in

tug-of-war competition (sensu Reeve et al. 1998) over reproduction. A focal worker receives a

fraction of the resources in proportion to her selfish effort, relative to the total worker selfish

effort in the colony, and no direct fitness if the queen does not die. If the queen does not die, we

assume that workers investing in selfishness do not reproduce, and all colony resources go to

producing the queen’s offspring. Thus, worker direct fitness can be written:

! ! ! ! ! ! ! ! ! ! !!

Other workers in the same colony each obtain:

!! !! ! ! ! ! ! ! ! ! !!

For simplicity, we assume that selfish investment by workers is equally detrimental to the colony’s production of each sex, and that worker’s strategy does not affect the sex ratio. If this is the case, a queenright colony will invest in males and in females, where s is the

colony’s sex investment ratio, expressed !"as proportion invested! ! ! in! males. Thus, a queen’s male

and female production can be written:

!" ! ! ! ! ! !"

!" ! ! ! ! ! ! ! ! !

To determine the effect of a focal worker’s behavior on its inclusive fitness, we require

the kin values (life-for-life relatedness values) of the various types of offspring (Table 2.2). The

Table 2.2. Life-for-life relatedness values for a focal worker. Kin values are calculated by multiplying regression relatedness and the sex-specific reproductive value. For Model 1, reproductive value of males is , where B is the proportion produced in queenless ! colonies; reproductive value is 1 for females (Pamilo 1991, Appendix 2.2). For Model 2, see !! ! ! !!! Appendix 2.2 for VM calculation. Regression relatedness between female siblings, rsib, is defined as , where m is the queen’s effective mating frequency. ! ! ! !!! ! ! ! ! ! Relatedness of focal worker to: Own sons VM Other worker’s sons rsib*VM !! Queen’s sons (!)*VM !!" Queen’s daughters rsib !!" New Q’s sons rsib*VM !!" New Q’s daughters (!)*rsib !!!" New W’s sons (!)* rsib*VM !!!"

!!!"

24 inclusive fitness increment that results from a slight increase in selfish effort x can be written

(Wenseleers et al. 2004b; Taylor et al. 2007):

!" !" !" ! !"! !" !! !" !! !" ! ! ! ! ! ! ! ! ! ! !" !" !" !" At the equilibrium (ESS; (Maynard-Smith 1982)) value of selfish effort, the inclusive fitness increment is equal to zero. We can solve for the ESS strategy by setting the above expression equal to zero, substituting for x and x’, and solving for , and checking that this solution is a ! ! maximum with a second derivative! test. !

! ! !! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !

In this solution, is the population’s proportion of males produced by queenless colonies, used

in computing male! reproductive value (Appendix 2.2).

This ESS solution for worker selfishness increases with f, the probability of queen death,

and m, the mating frequency of the queen, since the partial derivatives ! and ! are always !! !! !" !" positive given our definitions of m, f, s, n, and (analyses not shown). Assuming sex

investment at the worker optimum, moderate colony! size and low proportion of colonies with

failing queens, a moderate probability of queen death can favor much more selfish investment

than if workers are certain of queen survival (when f = 0; Figure 2.2). Importantly, this

investment in selfish effort occurs in queenright colonies without the possibility of direct

reproduction in the queenright stage.

Figure 2.2. Worker selfishness increases with the probability of queen loss, and with queen mating frequency. This assumes no worker reproduction in queenright colonies and no possibility of queen replacement (Model 1). Workers invest in selfishness in queenright colonies due to a non-zero probability of the colony entering a worker-reproduction state should the queen die in the worker’s lifetime. The depicted solution values are for n = 1000, B = 0.1, and s at the worker optimum . !" ! !!! !!!!! !!

26 Colony size has virtually no effect on selfishness. Varying the population fraction of

males produced by queenless colonies , and the sex investment ratio (s), likewise shift the

solution only slightly (data not shown),! !though! the overall effect of f and m remain the same.

Model 2: Re-queening and imperfect policing

The result of Model 1 makes clear that the possibility of queen loss can favor worker

investment in personal reproduction, but the assumptions of this model are not as widely

applicable as they could be. In order to investigate this hypothesis in honey bees and other

species, we extend Model 1 in two ways. First, we allow for another possible outcome following

queen loss. We now assume that once a queen has died, or left the colony with a reproductive

swarm, it is possible for workers to rear a new replacement queen, who will be a daughter of the

old queen. There is a chance that the re-queening attempt will fail, which occurs with probability

p; this could occur in honey bees, for example, if there are not enough female brood to rear a

replacement, or if a replacement queen dies on her mating flight. The three possible outcomes

for a focal worker are: (1) The old queen survives, with probability ; (2) The old queen is replaced by a new queen, with probability ; (3) The old queen! ! is! lost, is not replaced,

and workers compete to produce males, with! !probability! ! (Figure 2.1).

The second modification is to allow some worker!" reproduction in queenright colonies due

to imperfect policing. To more accurately represent policing, we allow a fraction S of workers’

investment in selfishness to escape policing and be invested in the production of male offspring,

even if the colony remains queenright. In this version of the model, selfish investment x

represents investment in actual reproduction in the queenright colony, and is still proportional to

future reproductive ability should the colony become queenless. We use an expression based on

27 that in Wenseleers et al. (2004a) to describe the effect of policing on a worker’s share of the colony’s male reproduction: a focal worker with selfish effort x, in a colony with policing determined by S, and a queen with a relative fecundity q times the maximum output of a worker, obtains a proportion of the colony’s male production equal to:

!" ! !" ! ! ! ! ! ! ! !

Other workers in the colony each obtain a similar share proportional to their selfish investment x’, while the queen obtains a share proportional to her fecundity relative to workers:

! ! !" ! ! ! ! ! ! ! !

We can write the direct fitness of the focal worker in each of the three possible outcomes. Here, we assume workers only compete for the fraction (s) of queenright colony resources invested in males. If the old queen survives, the worker has direct fitness:

!" !" ! !" ! ! ! ! ! ! ! !

If the queen is replaced, the worker obtains this same amount of direct fitness because policing continues. Finally, if the queen dies, the worker obtains a fraction resulting from a tug-of-war with the other workers, similar to Model 1:

! ! ! ! ! ! ! !!

28 From these expressions and the relative probabilities of the three possible outcomes (Figure 2.1),

we can write the total direct fitness of a focal worker, and write analogous expressions for the

other workers, the original queen and the replacement queen. Using the appropriate relatedness

values (see Appendix 2.2 for calculation of reproductive value using methods of (Alpedrinha et

al. 2013)), we can then determine the inclusive fitness effect of a small change in x (Wenseleers

et al. 2004a; Taylor et al. 2007). We apply the same approach as in Model 1, and solve for x*

under a range of parameter values. The ESS is an extremely complicated function, so we take

two simplifying approaches to analyzing the solution. First, we consider the case where S = 0,

representing the scenario when workers do not reproduce in queenright colonies, as in Model 1,

but are capable of rearing a replacement queen in the event of queen death. We then examine the

solution when S > 0 and, instead of determining its behavior analytically, we consider a range of

biologically realistic parameter values and examine the numerical solutions graphically.

Requeening with perfect policing (S = 0)

When we assume, as in Model 1, that workers cannot reproduce in queenright colonies (S

= 0), the solution resembles the solution to Model 1, but with a strong effect of p, the probability

of failure to re-queen (Figure 2.3; when p =1, the solution is identical to Model 1). When p = 0,

i.e. when re-queening is certain, selfishness is reduced to zero because workers never get a

chance to reproduce (Figure 2.3). However, even a small increase in p above zero can result in

moderate selfish investment by workers, if the probability of queen loss and queen mating

frequency are both high (Figure 2.3). Increasing the probability of failure to re-queen always

increases selfishness (that is, ! is always positive), and, as in Model 1, increasing f and m !! !" always increases selfishness (analyses not shown).

Figure 2.3. Worker selfish strategy when workers can replace the queen but cannot reproduce in queenright colonies (Model 2 when S = 0). The ESS worker selfish strategy x* increases with both the probability of re-queening failure p and the probability of queen loss f when S=0, A=0.1, B=0.1, m=15 and n = 1000, and s a the worker optimum. This result is virtually insensitive to variation in A, B, s and n, though lower mating frequency (m) favors lower selfishness. Requeening with worker reproduction in queenright colonies (S > 0)

30 To investigate whether the drop in relatedness to a new queen selects for increased worker investment in selfishness, we considered the scenario in which workers can produce male offspring in queenright colonies. Selfishness increases with increased risk of re-queening failure

(p) under all analyzed conditions (Figure 2.4), though the degree to which the probability of failure to re-queen affects selfishness depends strongly on the probability of queen loss (f), and on both colony size (n) and policing efficiency (S). We graphically analyze three scenarios: (i) a colony with a healthy queen that will not swarm soon (low f, red line in Figure 2.4), (ii), a colony with a weak queen, who may be replaced soon (green line in Figure 2.4, moderate f), and (iii) a colony preparing to swarm, or a colony with a dying queen (blue line in Figure 2.4, high f). The increasing probability of queen loss across these scenarios is always associated with increasing selfishness, regardless of other parameters. If policing is sufficiently strong to permit only very little worker reproduction in the absence of queen loss (purple dashed line near zero; upper left 6 panels in Figure 2.4), workers’ optimal selfishness is zero regardless of f unless p > 0. This suggests that if worker acquiescence (sensu Wenseleers et al. 2004a) is enforced by policing under normal queenright conditions, then the drop in relatedness that occurs during queen replacement is not sufficient to select for workers to attempt to reproduce even under weak queen or swarming conditions. Such reproductive effort would be wasted due to efficient policing after re-queening, unless there is a significant chance that re-queening will fail (p > 0; see discussion). We also find that increasing mating frequency (m) s elects for much greater selfishness than single mating, even under strong policing.

#$ 1.0 !"#"!" !"#""%"! !"#""%! !"#""%$

%& : ("#")*+ $&'& 0.5 !"" ("#")*, ("#")*)-

1.0

$&'& 0.5 !("""

1.0

0.5 $&'& )**&+,-./-&0/12304&356/078/57&9 !"("""

0.5 1.0 0.5 1.0 0.5 1.0 0.5 1.0 ;-,<=<3137>&,2&2=3135?&7,&-/#@A//5&9':

Figure 2.4. Evolutionarily stable worker selfish investment as a function of p, the probability of failing to re-queen, for several colony sizes and strengths of policing (Model 2 with S>0). For all plots, m = 15, q = 25, = 0.1, s is at worker optimum (~ 0.5; data not shown). The purple dashed line represents a normal colony with a low probability of queen loss in a new worker’s lifetime (f = 0.01).!! ! The!! !green dotted line represents a colony with a weak queen, with a moderate probability of queen loss within a worker’s lifetime (f = 0.5). The blue solid line represents a colony with a failing queen or a colony preparing to swarm, with a high probability of queen loss within a worker’s lifetime (f = 0.9). Plots in columns share a strength of policing parameter (S) value listed above, and rows share a colony size parameter (n) value listed at right.

32 Discussion

Our goal has been to determine how the threat of queen loss might influence worker reproductive development in honey bees and other social insects. The results suggest that the possibility of queen loss favors worker selfish investment in personal reproduction, despite a cost to the colony. A general prediction emerges that variation in the probability of queen loss (f) and successful queen replacement (p) are likely important factors in explaining variation in worker ovarian development among species, colonies, and colony life stages. These results are consistent with and might help to explain the observed widespread variation in partial ovary development by worker honey bees and other social insects, even if this investment is at a level too low for immediate egg laying (Smith et al. 2013). Model 1, and Model 2 with S = 0 predict selfish effort such as ovarian development that cannot be due to successful worker reproduction in the queenright colony. If we are correct in assuming that worker investment in ovarian tissue in the queenright colony corresponds to a greater ability to compete for resources in the queenless state, then these workers may be seen as hedging their bets against queen loss, while incurring a cost to the queenright colony. This key assumption is eminently testable, and such a test is a logical next step in determining the selective basis for worker partial ovarian development.

Importantly, this predicted selfishness may be lurking below the surface even in queenright colonies without actively reproducing workers, incurring colony-level costs without conspicuous evidence of conflict. Workers may be investing selfishly just to the extent that they are not detectable as reproductive to aggressively-policing workers, resulting in varying but significant partial ovarian development (Smith et al. 2013). We suggest that workers in some

33 advanced eusocial species may be modulating their reproductive behavior in response to the likelihood of an opportunity for direct reproduction, with a trade-off between indirect and direct benefits reminiscent of the dominance and inheritance hierarchies of Polistes wasps and other primitively eusocial insects (Cant and Field 2001; Liebig et al. 2005; Field et al. 2006). While the queen and other workers clearly benefit from worker reproduction after queen loss, the conflict we model here with a “tug-of-war” structure represents energy wasted in competition for reproduction that could be spent investing in further colony productivity. The queen would

“prefer” that only as few workers as is necessary to produce eggs would reproduce, which is likely to be rather few, given the substantial egg laying abilities of many species’ workers. That such wasteful selfish competition arises in queenless honeybee colonies is evidenced by the large fraction of workers that invest in reproduction (Wenseleers and Ratnieks 2006b), and by the supernumerary eggs in queenless colonies that reflect worker laying. Our model suggests such competition may extend into the period prior to queen loss.

Honey bees

Two recent studies document variation in worker selfishness in honey bee colonies that is consistent with our model’s predictions. In colonies headed by Single-Drone Inseminated (SDI) queens, Mattila et al. (2012) demonstrate that workers increase their ovarian development at a cost to colony productivity. The hypothesis that workers in these colonies are reproducing in the presence of a queen due to a facultative relaxation of policing has been ruled out (Loope et al.

2013), suggesting that some other attribute of SDI queens promotes worker ovarian development. If SDI queens have higher intrinsic mortality, and thus higher f, our model predicts increased ovarian development in workers. Alternatively, it could be that SDI queens

34 have the same intrinsic lifespan, but a reduced oviposition rate q. Lowering q also increases worker selfish investment under some parameter values in our model (analysis not shown; also see Wenseleers et al. 2004b), but may have the more important effect of forcing workers to attempt to replace her, and thus increasing f relative to a colony with a normal queen. In either case, the extensive partial ovarian development in SDI colonies suggests that workers are not currently reproducing, but may sense an opportunity for personal reproduction in the future. The evidence suggesting SDI queens are more likely to fail is only anecdotal (HR Mattila, personal communication), though this would be easy to test. In support of the hypothesis that workers in

SDI colonies are preparing for queen loss, SDI colonies produced more worker-derived males following experimental queen removal than did multiply-mated colonies (Bratkowski et al.

2012). A fruitful follow-up to these studies would be to examine worker ovarian development in colonies before and after a failing queen is replaced, to see whether partial ovarian development increases as it does in SDI colonies.

In a second recent study of honey bee worker selfishness, Woyciechowski and

Kuszewska (2012) found that during the period immediately following swarming, honey bee workers increase investment in personal reproduction, by both activating their ovaries and by eclosing with a greater number of ovarioles per ovary, thus permanently investing in reproductive capacity. This represents much greater worker reproductive investment than that observed by Mattila et al. (2012) in SDI colonies, with many workers developing their ovaries much more fully. This is predicted by our model, as a colony with a queen who has just departed with a swarm has an f (probability of queen loss) of 1, substantially higher than that for workers in a colony with a weak queen; these workers will experience the gamble of queen replacement

35 with complete certainty. Further support for the queen loss hypothesis would come from

evidence that workers increase ovarian development leading up to a swarming event. Some data

suggest this (Kropacova and Haslbachova 1970), though the result is far from conclusive. If

workers do increase ovarian development to prepare for possible queen loss, our model would

also predict that those workers elect to stay with the new queen, rather than depart with the

swarm, as the risk of queen loss is only experienced by the remaining fraction.

For honey bees, the risk of queen loss may select for worker selfishness for at least two

reasons (Woyciechowski and Kuszewska 2012). First, the colony may become queenless, in

which case workers benefit from being primed for reproduction, as outlined above. Second,

workers are less related to a replacement queen than to the old queen, and they should thus be

willing to waste more colony resources in competing with that queen for reproduction within the

queenright colony. The results of Model 2 suggest that it is the former rather than the latter that

actually selects for increased selfishness during swarming. For moderate colony sizes and

realistic policing levels, we can compare the predictions for when p = 0 (when re-queening is

certain) and when p is slightly positive (representing a small chance that re-queening fails). The

p = 0 scenario represents the predicted selfishness due only to the drop in relatedness to the new

queen’s offspring (compared to the old queen’s offspring) that occurs during re-queening, while

the p > 0 scenario includes the effects of both relatedness and of possible queenlessness on worker selfishness. For realistic values of policing and colony size, worker selfishness is zero when p = 0, suggesting that the relatedness drop alone does not select for worker selfishness.

This is because strong worker policing still selects for reproductive acquiescence (Wenseleers et

al. 2004b; Wenseleers et al. 2004a), despite the change in relatedness to the colony’s queen.

36 However, even a small probability of re-queening failure increases selfishness substantially for

colonies with a moderate or high chance of queen loss (Figure 2.4).

The conclusion that worker selfishness during queen turnover is caused by possible

failure to re-queen, rather than lower relatedness between workers and the new queen, depends

on continued policing throughout and following swarming. It is possible that policing is reduced

during re-queening, which would, along with the drop in relatedness, promote worker

reproduction. Purely on relatedness grounds, daughters of the old queen should no longer prefer

the new queen’s male eggs over workers’ male eggs, as their relatedness to both types of eggs is

equal (Table 2.2). However, this seems unlikely to have selected for relaxed policing; in this

scenario, any colony productivity benefit to policing should lead to maintained policing to

discourage worker reproduction (Ratnieks 1988). Empirical evidence also suggests facultative

policing is unlikely: worker policing in honey bees is not facultative in response to mating

frequency, although experimental colonies did not represent a natural level of variation (Loope et

al. 2013). More importantly, worker policing persisted or increased for several weeks following

experimental queen removal (Miller and Ratnieks 2001), suggesting that during queen

replacement, worker reproduction would be fruitless due to strong policing. Examining policing

throughout a queen replacement event would determine whether workers have a window of

opportunity for direct reproduction.

Our hypothesis rests on the assumption that failure to re-queen actually happens in

nature. What are estimates of p for natural colonies? For honey bees, we can approximate p by examining the fraction of virgin queens lost on their mating flights (Table 2.3). Estimates range

37 from 0.048 to 0.53, though some of these studies used colonies in unnatural densities or

locations, which could artificially increase the risk of queen loss because queens may mistakenly

return to the wrong hive (Ratnieks 1990). On the other hand, the lowest estimates come from

colonies kept outside of the honey bee’s native range, possibly masking a more substantial risk to

queens from native predators. For example, beekeepers in the Middle East lose many queens on

mating flights to migratory avian predators (bee-eaters; Merops spp.) during certain times of the

year, and a colony rearing an emergency queen during a bee-eater migration may face a high

probability of becoming permanently queenless (Yakobson and Rosenthal 1990; Ali and Taha

2012). In Japanese populations of Apis cerana, workers rarely rear a replacement queen following experimental queen removal (Sakagami and Akahira 1958), and thus have a much

higher effective p than A. mellifera. As predicted, workers exhibit greater ovarian development,

although this species has only moderately effective policing in queenright colonies, with S

between 0.05 and 0.29 and some workers successfully reproducing in queenright colonies

(Oldroyd et al. 2001; Holmes et al. 2014). This ovary development may help workers to begin

laying soon after queen loss (Oldroyd et al. 2001; Holmes et al. 2014). Oldroyd et al. further

suggested that this species may have a high risk of queen loss due to frequent swarming and

absconding, and therefore a high f value which would further select for worker selfishness.

Anarchy and reproductive revolutions in Apis

Extensive genetic analyses have revealed that some honeybee genetic lines produce

workers that lay eggs and are immune to policing (Oldroyd et al. 1994; Oldroyd and Osborne

1999; Barron et al. 2001; Chaline et al. 2002; Holmes et al. 2013). These “anarchistic” workers

lay eggs that are not policed, and can produce sons in queenright colonies. They are often

Table 2.3. Probability of queen loss during mating flights for different populations of honey bees. Asterisk marks the queen mating flight mortality before / during a bee-eater (Merops spp) migratory period. Species Location Estimate N Reference

A. mellifera New York, 0.049 52 Ratnieks 1990 lingustica USA

California, A. mellifera 0.062 32 Tarpy and Page 2000 USA

A. cerana Kashmir, India 0.14 310 Shah and Shah 1980

A. mellifera Island of capensis and A. m. Neuwerk, 0.35 20 Kraus et al. 2004 carnica Germany

A. mellifera 0.20 / Saudi Arabia 30 / 30 Ali and Taha 2012 jementica 0.53*

39 detected as rare patrilines within colonies of “normal” (non-anarchistic) workers (Oldroyd et al.

1994; Holmes et al. 2013), though anarchy may manifest through queen lines as well (Chaline et al. 2002). These workers substantially complicate an analysis of worker reproduction, given that anarchists have different optimal selfish efforts, and the costs and benefits of selfish investment for both types of workers will depend on the frequency and strategy of each type within a colony.

Evolutionary models like ours require simplifying assumptions, and, as it currently stands, our model does not accommodate the complexity of worker anarchy. However, anarchistic colonies appear to be (surprisingly) rare in European and North American populations (Visscher 1989;

Chaline et al. 2002), though a recent study suggests that this strategy may be more common in

Australia than was previously thought (Holmes et al. 2013). A similar phenomenon may also occur in A. cerana (Holmes et al. 2014), where workers may have recently undergone a

“revolution,” temporarily escaping policing (Wenseleers et al. 2004b; Nanork et al. 2007).

Though ignoring these very interesting phenomena, we feel it is still useful to examine worker reproduction in the absence of anarchy, as this may apply to many populations of Apis bees and other social insects.

Other social insects

The prediction that greater rates of queen loss select for worker selfish investment should apply to annual species, such as temperate social wasps (Vespinae and Polistinae) and bumblebees (Bombus spp). However, for such species we must consider another layer of complexity: the possibility of queen loss due to matricide (Bourke 1994; Foster and Ratnieks

2001b; Strassmann et al. 2003). Matricidal behavior includes its own tradeoffs and its coevolution with worker selfish behavior is beyond the scope of this paper. However, to other

40 workers in a colony, a matricidal event is effectively the same as queen loss due to other mortality factors, and may even be more predictable. Thus, our model still predicts that in species with matricide, worker selfish investment precedes queen killing, although such species are typically singly mated, thus the effect is likely to be small. Interestingly, vespine wasp species with a higher rate of queen loss appear to have more worker ovary development in queenright colonies (Foster and Ratnieks 2001b), and several of these species have been observed to kill their queen (Bourke 1994).

A more straightforward test of our hypothesis avoiding this complication would be to experimentally reduce queen life expectancy in a species without matricide (e.g. vulgaris-group

Vespula wasps), and look for increased worker ovary development. Vulgaris-group Vespula species resemble honey bees in effective queen mating frequencies typically exceed two (Loope et al. 2014), and worker reproduction in queenright colonies is near zero due to policing (Foster and Ratnieks 2001c; Bonckaert et al. 2008), though queenless colonies can produce many worker-derived males (Spradbery 1973). Our hypothesis predicts a strong effect of a failing queen on worker selfishness in these species, consistent with the observed presence of variation among colonies in worker ovarian development (Ross 1985).

Worker ants are often capable of producing male offspring in queenless colonies (Bourke

1988). A recent study in the facultatively polygynous ant Leptothorax acervorum nicely tested our central hypothesis that workers forecast the probability of becoming queenless and prepare for queen loss before it occurs (Friend and Bourke 2014). Worker reproduction began sooner in de-queened monogynous colonies than in de-queened polygynous colonies, suggesting workers are preparing for reproduction in colonies with only a single remaining queen. Behavioral analyses also strengthened the case that such “future reproductive” workers act in ways (e.g.

41 increased aggression and feeding from larvae, decreased brood care) that enhance their future reproductive success, but that may be costly to the colony, parallel to observations of laziness by honeybee workers with partially or completely active ovaries (Mattila et al. 2012). Although our

model does not capture the complex kin structure of polygynous colonies, the results of this

study clearly lend support to our hypothesis. In other ant species, a substantial amount of

colony fitness may come after the queen’s death (Franks et al. 1990; Evans 1996), which may select for maintained worker reproductive abilities (Bourke 1988). Further experimental evidence for the hypothesis would come manipulations of queen health in monogynyous, polyandrous species. Lasius niger is a prime candidate, especially given that this species has an identified queen pheromone that responds quantitatively to experimental immune challenges to

the queen (Holman et al. 2010).

Queen signals of fecundity and viability

If the selfish investments of workers track the risk of queen loss as predicted by our

model, queens would be selected to provide signals of their health and vigor, i.e., signals of

queen fecundity, viability, or both. Our model suggests that such signals will be especially well

developed in species or contexts associated with frequent queen loss. To our knowledge, there is

no direct evidence for chemical communication of queen viability, though the queen pheromone

of honey bees and other species could provide such information. Our current understanding of

queen pheromones is mostly limited to presence/absence effects on worker behavior (Holman et

al. 2010; e.g. Van Oystaeyen et al. 2014), though suggestively, the amount of queen pheromone

produced by immune-challenged Lasius niger queens declines within 24 hours (Holman et al.

2010), indicating that it could signal queen viability. In honeybees, workers respond more

42 strongly to the queen pheromone blends of young queens with greater ovary development

(Kocher et al. 2009), consistent with the idea that queen pheromones track fecundity, which may

be correlated with viability. Whether workers use queen pheromones to assess queen expected

viability, and respond with ovarian development and other direct fitness enhancing behavior,

remains to be seen.

Conclusion

Our model shows that insect colonies can fall short of being perfect super-organisms even when workers have zero reproduction in the presence of the queen and the queen is alive

(and there is no other conflict such as sex-ratio conflict). Colonies of honey bees and other social insects regularly experience queen loss, and preparation for queen loss can lead to significant within-colony reproductive conflict, which could explain observed variation in worker selfish investment.

43 REFERENCES

Ali MAA-SM, Taha E-KA (2012) Bee-Eating Birds (Coraciiformes: Meropidae) Reduce Virgin Honey Bee Queen Survival during Mating Flights and Foraging Activity of Honey Bees (Apis mellifera L.). International Journal of Scientific & Engineering Research 3:1–8.

Alpedrinha J, West SA, Gardner A (2013) Haplodiploidy and the evolution of eusociality: worker reproduction. Am Nat 182:421–438. doi: 10.1086/671994

Barron A, Oldroyd BP, Ratnieks FLW (2001) Worker reproduction in honey-bees (Apis) and the anarchic syndrome: a review. Behav Ecol Sociobiol 50:199–208.

Bonckaert W, Van Zweden JS, D'Ettorre P, et al (2011) Colony stage and not facultative policing explains pattern of worker reproduction in the Saxon wasp. Mol Ecol 20:3455–3468.

Bonckaert W, Vuerinckx K, Billen J, et al (2008) Worker policing in the German wasp Vespula germanica. Behav Ecol 19:272–278.

Bourke AFG (1988) Worker reproduction in the higher eusocial Hymenoptera. Q Rev Biol 63:291–311.

Bourke AFG (1999) Colony size, social complexity and reproductive conflict in social insects. J Evol Biol 12:245–257.

Bourke AFG (2011) Principles of Social Evolution. Oxford Univ Press, Oxford

Bourke AFG (1994) Worker matricide in social bees and wasps. J Theor Biol 167:283–292.

Bratkowski J, Pirk CWW, Neumann P, Wilde J (2012) Genotypic diversity in queenless honey bee colonies reduces fitness. J Apic Res 51:336–341.

Cant MA, Field J (2001) Helping effort and future fitness in cooperative animal societies. Proc R Soc Lond B 268:1959–1964.

Chaline N, Ratnieks FLW, Burke T (2002) Anarchy in the UK: Detailed genetic analysis of worker reproduction in a naturally occurring British anarchistic honeybee, Apis mellifera, colony using DNA microsatellites. Mol Ecol 11:1795–1803.

Evans JD (1996) Queen longevity, queen adoption, and posthumous indirect fitness in the facultatively polygynous ant Myrmica tahoensis. Behav Ecol Sociobiol 39:275–284.

Field J, Cronin A, Bridge C (2006) Future fitness and helping in social queues. Nature 441:214– 217.

Foster KR, Ratnieks FLW (2001a) The effect of sex-allocation biasing on the evolution of worker policing in hymenopteran societies. Am Nat 158:615–623.

44 Foster KR, Ratnieks FLW (2001b) Paternity, reproduction and conflict in vespine wasps: a model system for testing kin selection predictions. Behav Ecol Sociobiol 50:1–8.

Foster KR, Ratnieks FLW (2001c) Convergent evolution of worker policing by egg eating in the honeybee and common wasp. Proc R Soc Lond B 268:169–174.

Franks NR, Ireland B, Bourke AFG (1990) Conflicts, social economics and life history strategies in ants. Behav Ecol Sociobiol 27:175–181.

Friend LA, Bourke AFG (2014) Workers respond to unequal likelihood of future reproductive opportunities in an ant. Anim Behav 97:165–176.

Hamilton WD (1964) The genetical evolution of social behaviour I. J Theor Biol 7:1–16.

Hammond RL, Bruford M, Bourke AFG (2003) Male parentage does not vary with colony kin structure in a multiple-queen ant. J Evol Biol 16:446–455.

Hammond RL, Keller L (2004) Conflict over male parentage in social insects. PLoS Biol 2:E248.

Holldobler B, Wilson EO (1990) The Ants. Harvard University Press, Cambridge, MA

Holman L, Jørgensen C, Nielsen J, D'Ettorre P (2010) Identification of an ant queen pheromone regulating worker sterility. Proc R Soc Lond B 277:3793.

Holmes MJ, Oldroyd BP, Duncan M, et al (2013) Cheaters sometimes prosper: targeted worker reproduction in honeybee ( Apis mellifera) colonies during swarming. Mol Ecol 22:4298– 4306.

Holmes MJ, Tan K, Wang Z, et al (2014) Why acquiesce? Worker reproductive parasitism in the Eastern honeybee ( Apis cerana). J Evol Biol 27:939–949.

Kocher SD, Richard FJ, Tarpy DR, Grozinger CM (2009) Queen reproductive state modulates pheromone production and queen-worker interactions in honeybees. Behav Ecol 20:1007– 1014.

Kropacova S, Haslbachova H (1970) The Development of Ovaries in Worker Honeybees in Queenright Colonies Examined Before and After Swarming. J Apic Res 9:65–70.

Liebig J, Monnin T, Turillazzi S (2005) Direct assessment of queen quality and lack of worker suppression in a paper wasp. Proc R Soc Lond B 272:1339–1344.

Loope KJ, Chien C, Juhl M (2014) Colony size is linked to paternity frequency and paternity skew in yellowjacket wasps and hornets. BMC Evol Biol 14:277.

Loope KJ, Seeley TD, Mattila HR (2013) No facultative worker policing in the honey bee (Apis mellifera L.). Naturwissenschaften 100:473–477.

Mattila HR, Reeve HK, Smith ML (2012) Promiscuous honey bee queens increase colony

45 productivity by suppressing worker selfishness. Curr Biol 22:2027–2031.

Maynard-Smith J (1982) Evolution and the Theory of Games. Cambridge Univ Press

Miller D, Ratnieks FLW (2001) The timing of worker reproduction and breakdown of policing behaviour in queenless honey bee (Apis mellifera L.) societies. Insectes Soc 48:178–184.

Nanork P, Wongsiri S, Oldroyd BP (2007) Preservation and loss of the honey bee (Apis) egg- marking signal across evolutionary time. Behav Ecol Sociobiol 61:1509–1514.

Ohtsuki H, Tsuji K (2009) Adaptive reproduction schedule as a cause of worker policing in social hymenoptera: a dynamic game analysis. Am Nat 173:747–758.

Oldroyd BP, Halling L, Good G, et al (2001) Worker policing and worker reproduction in Apis cerana. Behav Ecol Sociobiol 50:371–377.

Oldroyd BP, Osborne KE (1999) The evolution of worker sterility in honeybees: the genetic basis of failure of worker policing. Proc R Soc Lond B 266:1335–1339.

Oldroyd BP, Smolenski AJ, Cornuet JM, Crozler RH (1994) Anarchy in the beehive. Nature 371:749.

Pamilo P (1991) Evolution of colony characteristics in social insects. I. Sex allocation. Am Nat 137:83–107.

Peeters C, Crewe R (1985) Worker reproduction in the ponerine ant Ophthalmopone berthoudi: an alternative form of eusocial organization. Behav Ecol Sociobiol 18:29–37.

Ratnieks FL (1990) The evolution of polyandry by queens in social Hymenoptera: the significance of the timing of removal of diploid males. Behav Ecol Sociobiol 26:343–348.

Ratnieks FLW (1988) Reproductive harmony via mutual policing by workers in eusocial Hymenoptera. Am Nat 132:217–236.

Ratnieks FLW, Helantera H (2009) The evolution of extreme altruism and inequality in insect societies. Phil Trans R Soc B 364:3169–3179.

Reeve HK, Emlen S, Keller L (1998) Reproductive sharing in animal societies: reproductive incentives or incomplete control by dominant breeders? Behav Ecol 9:267–278.

Ross KG (1985) Aspects of worker reproduction in four social wasp species (Insecta: Hymenoptera: Vespidae). J Zool 205:411–424.

Sakagami SF, Akahira Y (1958) Comparison of ovarian size and number of ovarioles between the workers of Japanese and European Honeybees (Studies on the Japanese honeybee, Apis indica cerana Fabricius. I). Jpn J Entomol 26:103–109.

Seeley TD (1978) Life history strategy of the honey bee, Apis mellifera. Oecologia 32:109–118.

46 Servedio MR, Brandvain Y, Dhole S, et al (2014) Not Just a Theory—The Utility of Mathematical Models in Evolutionary Biology. PLoS Biol 12:e1002017.

Smith ML, Mattila HR, Reeve HK (2013) Partial ovary development is widespread in honey bees and comparable to other eusocial bees and wasps. Communicative & Integrative Biology 6:e25004.

Spradbery J (1973) Wasps. Sidgwick and Jackson, London

Strassmann JE, Nguyen J, Arevalo E, et al (2003) Worker interests and male production in Polistes gallicus, a Mediterranean social wasp. J Evol Biol 16:254–259.

Strassmann JE, Sullender BW, Queller DC (2002) Caste totipotency and conflict in a large- colony social insect. Proc R Soc Lond B 269:263–270.

Taylor PD, Wild G, Gardner A (2007) Direct fitness or inclusive fitness: how shall we model kin selection? J Evol Biol 20:301–309.

Toth E, Queller DC, Dollin A, Strassmann JE (2004) Conflict over male parentage in stingless bees. Insectes Soc 51:1–11.

Van Oystaeyen A, Oliveira RC, Holman L, et al (2014) Conserved class of queen pheromones stops social insect workers from reproducing. Science 343:287–290.

Visscher PK (1989) A quantitative study of worker reproduction in honey bee colonies. Behav Ecol Sociobiol 25:247–254.

Wenseleers T, Hart A, Ratnieks FLW (2004a) When resistance is useless: Policing and the evolution of reproductive acquiescence in insect societies. Am Nat 164:E154–E167.

Wenseleers T, Helantera H, Hart A, Ratnieks FLW (2004b) Worker reproduction and policing in insect societies: an ESS analysis. J Evol Biol 17:1035–1047.

Wenseleers T, Helanterä H, Alves DA, et al (2013) Towards greater realism in inclusive fitness models: the case of worker reproduction in insect societies. Biol Lett 9:20130334–20130334.

Wenseleers T, Ratnieks FLW (2006a) Comparative analysis of worker reproduction and policing in eusocial hymenoptera supports relatedness theory. Am Nat 168:E163–E179.

Wenseleers T, Ratnieks FLW (2006b) Enforced altruism in insect societies. Nature 444:50–50.

Wilson EO (1971) The Insect Societies. Belknap Press, Cambridge, MA

Winston ML (1980) Swarming, afterswarming and reproductive rate of unmanaged honeybee colonies. Insectes Soc 27:391–398.

Winston ML (1987) The Biology of the Honey Bee. Harvard University Press, Cambridge, MA

Woyciechowski M, Kuszewska K (2012) Swarming Generates Rebel Workers in Honeybees.

47 Curr Biol 22:707–711.

Yakobson BA, Rosenthal C (1990) The status of bee pests in Israel. Proceedings of the International Symposium on Recent Research on Bee Pathology, Sept 5-7, 1990, Gent, Belgium 213–214.

48 Appendix 2.1. Selfish effort as the product of the probability of selfish investment and degree of

selfish investment

Here we demonstrate, in a simplified scenario, that the selfish strategy x can sensibly be defined

as the product of the probability that a focal worker invests in selfishness (e.g. activates her

ovaries) and the degree to which they invest, given that they do (e.g. the extent to which they

develop their ovaries). This allows us to relax the assumption that reproductive workers invest

maximally in reproduction.

Let a* be the probability that an individual activates her ovaries, and let b* be the ESS for the

personal investment in personal reproduction of an ovary-activated individual (at the expense of

group productivity). For a group of n individuals, a mutant that activates her ovaries with

probability a will have a direct fitness of:

! !" ! ! ! ! ! ! ! ! !" ! !! ! !!! ! !" ! !! ! !!! !

Assuming relatedness of r between group members, solving for the ESS by setting the first

derivative of the mutant’s inclusive fitness equal to zero yields the solution constraint:

! ! !! ! !!!! ! !! ! ! ! !

Clearly the same analysis on b will yield the same constraint, given the symmetry of a and b.

Any values of a and b that together satisfy this constraint will be evolutionarily stable. Thus, we

49 can simply define x as the product of a and b, allowing consideration of colonies comprised of

some individuals that have intermediate investments in personal reproduction. This finding is

dependent on the linear costs of selfish investment, and will not hold generally if this assumption

is broken.

Appendix 2.2. Male reproductive value

Model 1: Worker reproduction in queenless colonies

The reproductive value of males in the population is a function of , the proportion of the

population’s males produced by queenless colonies (Pamilo 1991). Under!! the assumption that

no workers reproduce in queenright colonies, if VF is set to 1,

! !! ! For both Model 1 and Model 2, the proportion! ! ! of males produced in queenless colonies is positive, and likely to be influenced by many variables not in our model, including the longevity of queens, and the productivity and longevity of queenless colonies relative to queenright colonies. We set for our numerical analyses above, though changing this value does not qualitatively alter !the! predictions!!! (analyses not shown). We expect to be low, given the relatively low frequency of queenless colonies encountered in nature.!

Model 2: Daughters can produce queens and males

We use the approach of Alpedrinha et al. (2013) to determine the reproductive value of males and females. Reproductive value represents the contribution of each class (in our case, each sex) to the gene pool after many generations. First, we must define generations. We

50 consider 'old' queens to be females of a given generation, that produce males and new queens of the next generation. They can also produce workers (which can produce males) and replacement queens (daughters that can produce new queens and males), both of which we call

'intergenerational' females.

For simplicity, we assume that replacement queens mate with males of the same generation as the old queens, and do not survive to the next generation. Given this setup, when considering the probabilities that genes in the next generation are inherited from females or males, we must also consider new queens and reproductive workers, as these add new pathways for males genes to reach the next generation (for example, without workers or new queens, male genes would never be transmitted to new males) (Pamilo 1991; Alpedrinha et al. 2013).

To calculate reproductive value for each sex, we first determine the transition probabilities Pij that represent the proportion of genes in individuals of sex i in a new generation that derive from individuals of sex j in the previous generation.

We have three types of colonies, those with old queens that survive, those that re-queen, and those that become queenless and have only worker reproduction. We will assume that of all of the males produced in a generation, a fraction B of them are produced in queenless colonies and a fraction A of them are produced in requeened colonies.

Then, according to Alpedrinha et al. 2013, we can calculate Pmf as:

!!" ! ! ! !! !! !!!!!!!

51 Where is the fraction of males produced in a given generation that are sons of an original

! (non-replacement)! queen, defining as the fraction of males from queenright colonies that are produced by workers. !

! We also define as the fraction! of !males! !!! !that!! !are! !sons!!!!! of! ! intergenerational!!! females (workers

! and replacement! queens). This is all of the males not produced by old queens:

! ! To calculate the proportion of genes in males! ! that!! !derive! !! from males of the previous generation,

we simply write:

!! !" ! ! ! !! ! !

Similarly, we can write (Alpedrinha et al. 2013):

!!!

! ! !!! ! ! ! !! !!! !!!" where is the fraction of next generation queens! that !are daughters of old queens, while is

! !" the fract!ion produced by replacement queens. !

We will assume the proportion of queens produced by replacement queens is similar to the

proportion of males produced by colonies with replacement queens, and thus:

!! ! ! ! !! !! ! !! ! !!

! !!" ! !! ! !!

52 We here, and later, assume that the sex investment ratio in queen replacement colonies is equal

to that in original queen colonies.

Following Alpedrinha et al. 2013, we have

!" !! ! ! ! !! ! !

Finally, to calculate the sex-specific class reproductive values, we calculate the left eigenvector of the transition probability matrix (Frank 1998; Alpedrinha et al. 2013):

!!!! !!"! !" !! ! ! ! !

From this solution, after setting VF = 1, we obtain a reproductive value for males:

! ! !!! ! !! !! ! !!! ! !!!! ! ! ! ! ! ! ! ! ! !

In Model 2, the proportion of worker-derived males in queenright colonies depends on the outcome of the competition among workers and the queen (Wenseleers et al.!! 2004b)! :

!"!!!! ! ! ! !" ! ! !

53 This is a function of x’, and thus must be substituted into the inclusive fitness increment

! expression! before solving for the ESS value of x*. This is only true in Model 2 when S > 0

(analysis not shown).

54 CHAPTER 3

COLONY SIZE IS LINKED TO PATERNITY FREQUENCY AND PATERNITY SKEW IN

YELLOWJACKET WASPS AND HORNETS

Kevin J. Loope, Chun Chien, Michael Juhl

Published in BMC Evolutionary Biology

Abstract

The puzzle of the selective benefits of multiple mating and multiple paternity in social insects has been a major focus of research in evolutionary biology. We examine paternity in a clade of social insects, the vespine wasps (the yellowjackets and hornets), which contains species with high multiple paternity as well as species with single paternity. This group is particularly useful for comparative analyses given the wide interspecific variation in paternity traits despite similar sociobiology and ecology of the species in the genera Vespula, Dolichovespula and Vespa. We describe the paternity of 5 species of yellowjackets (Vespula spp.) and we perform a phylogenetically controlled comparative analysis of relatedness, paternity frequency, paternity skew, colony size, and nest site across 22 vespine taxa. We found moderate multiple paternity in four small-colony Vespula rufa-group species (effective paternity 1.5 – 2.1), and higher multiple paternity in the large-colony Vespula flavopilosa (effective paternity ~3.1). Our comparative analysis shows that colony size, but not nest site, predicts average intracolony relatedness.

Underlying this pattern, we found that greater colony size is associated with both higher paternity frequency and reduced paternity skew. Our results support hypotheses focusing on the enhancement of genetic diversity in species with large colonies, and run counter to the hypothesis that multiple paternity is adaptively maintained due to sperm limitation associated with large colonies. We confirm the patterns observed in taxonomically widespread analyses by

55 comparing closely related species of wasps with similar ecology, behavior and social organization. The vespine wasps may be a useful group for experimental investigation of the benefits of multiple paternity in the future.

Background

The mating frequency of social insect queens is a central factor shaping the evolution of social behavior within colonies. Polyandry, and the genetic diversity created by multiple paternity, is the foundation of many evolutionary conflicts within insect societies (Ratnieks et al. 2006). It also has important consequences for sexual selection, sperm competition, and the evolution of male reproductive strategies (Boomsma et al. 2005; Boomsma 2007). But the evolution of multiple mating and multiple paternity itself is an evolutionary puzzle. It has arisen several times in the social Hymenoptera, including in a handful of ant genera, the honey bees (Apis spp.), and some vespine wasps (e.g. Vespula spp.) (Hughes et al. 2008). Multiple paternity presents automatic costs of increased exposure to sexually transmitted disease, greater predation risk while mating, and greater potential conflict among colony members due to lower relatedness

(Crozier and Fjerdingstad 2001). So, given these costs, what are the benefits that underlie the adaptive maintenance of multiple paternity?

Many hypotheses have been proposed to explain the fitness benefits of multiple mating and multiple paternity in the eusocial Hymenoptera (reviewed in (Crozier and Page 1985;

Boomsma and Ratnieks 1996; Crozier and Fjerdingstad 2001; Strassmann 2001)). The most successful suggest a colony-level benefit derived from the greater genetic diversity of colony members created when a queen uses sperm from multiple males. The pathogens and parasites hypothesis proposes that multiple paternity results in a colony with diverse genetic defenses

56 against coevolving natural enemies, thus reducing intracolony disease transmission and increasing colony survival (Hamilton 1987; Sherman et al. 1988; Schmid-Hempel 1998; Brown and Schmid-Hempel 2003). Alternatively, the division of labor hypothesis suggests that colonies with greater genetic diversity have greater genetically determined behavioral diversity or broader task performance thresholds, which allows for a more efficient division of labor or the exploitation of rare genetic specialists (Crozier and Page 1985; Robinson 1992; Fuchs and

Moritz 1998; Oldroyd and Fewell 2007; Jeanson et al. 2007). A third popular hypothesis not based on genetic diversity is that queens mate with multiple males to acquire a sufficient number of sperm (Cole 1983). In this paper we explore the evolution of multiple paternity, and the predictions of these hypotheses, by comparing species of vespine wasps, i.e., the yellowjackets and hornets (Figure 3.1). The vespine wasps share many features of their ecology and sociobiology but vary dramatically in paternity (Matsuura and Yamane 1990; Greene 1991;

Foster and Ratnieks 2001a), making them a useful group for comparative studies of the evolution of multiple paternity and its consequences.

The pathogens hypothesis, the division of labor hypothesis and the sperm limitation hypothesis all predict that colony size (i.e. the number of workers in a colony) will be positively associated with paternity frequency (i.e. the number of fathers represented in the workers of a given colony; Figure 3.2) (Cole 1983; Schmid-Hempel 1998; Bourke 1999). A larger workforce inevitably results in increased traffic by returning foragers. If each foraging trip represents a possible entry into the nest of an externally encountered pathogen, then larger colonies will acquire forager-borne diseases and parasites sooner and more often (Schmid-Hempel 1998).

This higher parasite pressure would cause larger colonies to benefit more from disease resistance conferred by multiple paternity. Species with large colonies also tend to have greater division of

Figure 3.1. Aerial and subterranean yellowjacket nests. A. Young colony of Dolichovespula arenaria, a typical aerial-nesting yellowjacket. B. Excavated, subterranean nest of an anesthetized colony of Vespula flavopilosa.

58 labor and specialization, and thus will benefit more from multiple paternity if the genetic diversity it brings enhances the division of labor (Bourke 1999; Jeanson et al. 2007). Both of these hypotheses also predict more even sperm use (reduced paternity skew) for species with greater colony size because reducing skew leads to reduced intracolony genetic similarity (Figure

3.2; (Jaffé et al. 2012)). Finally, queens who create large colonies may require more sperm than is provided by her first mate, and would thus be selected to mate with more males, increasing paternity frequency (Cole 1983). In the only study of male sperm quantity in vespine wasps, adult males were found to contain over 2 million sperm, approximately 100 times the average number found in the sperm storage organs of spring queens (Stein and Fell 1996), suggesting sperm may not be limiting. However, it remains theoretically possible that males benefit from incompletely inseminating queens due to sexual conflict (Wedell et al. 2002). Unlike the genetic-diversity based hypotheses, the sperm limitation hypothesis predicts that paternity skew should increase with colony size, because queens in large colonies should be selected to use all available sperm and males likely range widely in the amount of sperm they provide (Figure 3.2;

(Jaffé et al. 2012).

Comparative analyses have shown that across taxonomically broad sets of species, there is a positive relationship between paternity frequency and colony size (Cole 1983; Schmid-

Hempel 1998; Jaffé et al. 2012), as well as a negative relationship between paternity skew and paternity frequency, in ants, bees, and wasps (Jaffé et al. 2012). As the selective forces that drive the evolution of polyandry may differ across the social insects (Crozier and Fjerdingstad

2001), our aim in this work is to examine these predictions in a clade possessing a wide range of paternities, with species that differ little in social complexity and other traits that might confound the explanation of multiple paternity.

Colony size s + & + p,d,s Nest site - p,d (% enclosed) paternity paternity

skew frequency

p p

(B) - + - (k)

intra-colony genetic similarity (r) p: pathogens hypothesis d: division of labor E ects predicted by... hypothesis s: sperm limitation hypothesis

Figure 3.2. Predicted effects of colony size and nest site on paternity traits. Red arrows indicate the predicted positive (+) or negative (-) effects of increasing colony size and frequency of cavity nesting on paternity skew and paternity frequency. The predictions of the pathogens hypothesis and division of labor hypothesis stem from the effects of paternity traits on intracolony genetic similarity (black arrows).

60 For the vespine wasps, we extend the logic of the pathogens hypothesis to propose another factor–nest site–that may be involved in the evolution of multiple mating and multiple paternity. Vespine wasps all construct similar nests of multiple combs surrounded by insulating layers of paper envelope (Figure 3.1). Some species build nests hidden in cavities, typically excavated rodent burrows, rotten logs or tree holes. Other species construct exposed, aerial nests in shrubs or suspended from tree branches. It seems possible that these nest sites, and their microenvironments, expose colonies to different types and quantities of pathogens, as has been suggested for canopy- and soil-dwelling ants (Walker and Hughes 2011). Cavity-nesting and ground-nesting species may experience greater exposure to fungal and other microbial pathogens due to the increased proximity to damp soil and rotting wood, which could in turn favor multiple paternity. A casual examination of the vespine species included in the most extensive comparative analysis of colony size and paternity (Jaffé et al. 2012) suggests an association between colony size, nest site, and paternity. The large-colony, vulgaris-group Vespula species are subterranean nesting and have high paternity, while the small-colony Dolichovespula species are aerial nesting and have a low paternity. Because nest site and colony size are tightly associated in this data set, it is unclear whether it is colony size or nest site (or both) that distinguishes high-paternity species from low-paternity species.

Here we assume that colony size and nest site are determined by ecological factors

(Bourke 1999), such as climate, prey type and availability, nest site availability and predator type and abundance. These hypotheses predict that selection then modifies mating and paternity traits to reflect the ecologically determined colony size and nest site. In short, these hypotheses predict that colony size and nest site cause changes in mating behavior and paternity, rather than vice versa (Figure 3.2).

61 In this study we describe the paternity of five species of North American Vespula wasps.

Four of these species are members of the enigmatic Vespula rufa group, which have small colonies and subterranean nests (Akre et al. 1980; Archer 2012). These features make this clade attractive for testing the link between nest site and paternity as these species break the correlation between colony size and nest site found in the species that have been previously studied. The fifth species, Vespula flavopilosa, has small colonies compared to other species in the Vespula vulgaris group (MacDonald et al. 1980), and may thus provide an interesting intermediate position in the comparison of colony size and mating frequency. It is also a facultative social parasite (MacDonald et al. 1980), a feature sometimes associated with reduced paternity (Sumner

et al. 2004; Hoffman et al. 2008). We then perform a phylogenetically controlled comparative

analysis, including both colony size and nest site, to examine the species-level traits associated

with the evolution of intracolony genetic similarity (relatedness), paternity frequency and

paternity skew in the Vespinae.

Methods

Colony collections

We collected workers from active, mature colonies located by responding to pest control

calls (V. consobrina, V. atropilosa and V. acadica in Thurston, Co., WA), by nest searching (V.

flavopilosa; Tompkins Co., NY), or by “wasp-lining” foragers back to their nests (V. vidua;

Tompkins Co., NY). All collections adhered to state and federal regulations. Most colonies

would have been destroyed as pests regardless of our collection, and none are species known to

be endangered or threatened. Samples were collected between 2008-2013 and stored frozen (-20

deg C). We collected entire colonies of V. consobrina, V. vidua and V. flavopilosa to obtain

62 colony size data; a sample of worker V. acadica and V. atropilosa were collected from the nest

entrance with a vacuum. All V. consobrina and some V. vidua colonies were collected during

the day using a battery-powered vacuum. The collector waited at least 30 minutes to collect

returning foragers, and for V. vidua, we returned hours later to collect the last remaining foragers

and escapees. All colonies of V. flavopilosa and some colonies of V. vidua were anesthetized

overnight with CO2 and excavated in the morning. For colonies of V. consobrina, V. vidua and

V. flavopilosa, we counted all adult workers and determined the presence or absence of the mother queen.

Genetic analysis

We extracted DNA from approximately twenty workers or gynes per colony, as well as the mother queen when present, by placing a single antenna or leg in 100 µL of 10% Chelex solution (Chelex 100, 100-200 mesh, Bio-Rad), then incubating for 20 minutes at 95°C. We then refrigerated or froze the supernatant before PCR. Variable loci were selected based on preliminary screening of published loci (Thorén et al. 1995; Hasegawa and Takahashi 2002;

Daly et al. 2002). We used dye-labeled primers (Applied Biosystems) in combination with a 3- primer labeling method (Schuelke 2000) to perform multiplex PCR with 4-6 primers, depending on the species (Table S3.1). Each 10 µL PCR reaction included 1ul extracted DNA, 5 µL Qiagen master mix (Qiagen Type-It Microsatellite Kit, Qiagen Inc.), 0.2 µL of each reverse primer, 0.2

µL (dye-labeled) or 0.1 µL (3-primer labeled) of each forward primer, 0.15 µL FAM-labeled 3- primer tag for each 3-primer-labeled primer pair, and water to total 10 µL. PCR reaction conditions were 95°C for 15 minutes, 35 cycles of 95°C for 30 seconds, 50°C for 90 seconds,

72°C for 60 seconds, followed by 60°C for 30 minutes. Fragment analysis was performed on an

63 ABI-3730xl sequencer using 0.5 µL PCR product combined with 15 µL HiDi Formamide and

0.15 µL LIZ 500 internal size standard (Applied Biosystems). Allele sizes were called using

GeneMarker (SoftGenetics LLC) and checked twice by eye.

Estimating paternity

We used Colony2 v2.0.4.1 (Jones and Wang 2010) to find the maximum likelihood

configuration of paternity assignments for all genotyped workers. Workers that failed to amplify

at more than two loci were excluded from the analysis. In the few cases where Colony2 assigned

a worker to a matriline from a different colony, the anomalous worker genotype was checked

against the genotype of the queen from that colony. If she shared an allele with the queen at all

loci, a maternal sibship constraint was entered into Colony2 containing all workers in that colony that did not differ from the queen for both alleles at any locus. In all cases, the subsequent run of the likelihood analysis assigned the worker to an additional patriline from that colony. In the few cases where anomalous workers were inconsistent with being a daughter of the queen, these workers were removed from the dataset and the analysis was run again. From the paternity assignments, we calculated each colony’s observed paternity frequency (k), an uncorrected estimate of effective paternity: ke = where p is the proportion of offspring in the ! ! !!! ! sample fathered by male i (Starr 1984;! Boomsma! and Ratnieks 1996; Nielsen et al. 2003), and a corrected estimate of effective paternity (ke3), which adjusts for sample size (Nielsen et al. 2003).

Mean intracolony relatedness was calculated from effective paternity using the formula . ! ! ! ! !!!! We calculated paternity skew using the B index (Nonacs 2000). Table S3.1 reports allelic

diversity and expected heterozygosity calculated from allele frequencies determined by Colony2.

64 Non-detection and non-sampling error

Two types of error—non-detection and non-sampling of fathers—could lead to an inaccurate estimation of effective paternity (Boomsma and Ratnieks 1996). Non-detection error, when two males have the same multilocus genotype, was estimated using the formula found in ref (Jaffé 2014). For all five species studied, the non-detection error was <0.0025 (Table 1), suggesting that such errors did not bias our estimate of effective paternity. However, this calculation assumes maternal and paternal genotypes are known (Boomsma and Ratnieks 1996;

Foster et al. 1999). When examining the maternal genotype assignments following maximum likelihood analysis in Colony2 for colonies without a genotyped queen, queen genotypes were occasionally ambiguous (i.e., assigned a genotype with probability <0.9). There were 7 ambiguous single-locus genotypes for queens of V. acadica, 3 for V. vidua, and 2 for V. atropilosa. This never occurred at more than one locus per queen, so a simple and conservative estimate of the upper bound of male non-detection error is to remove the single most variable of the problematic loci from the non-detection calculation for each of these species. Even with this adjustment, error rates were low enough to be confident that paternity estimates were not biased due to male non-detection error (Table 1).

Non-sampling error occurs when a father is not detected because the sampled daughters do not include his offspring. We used the formula presented in (Cornuet and Aries 1980;

Oldroyd et al. 1997) to estimate the expected number of patrilines per colony, given the observed number of patrilines and the number of individuals sampled, assuming no skew among patrilines.

The species-level average expected paternity frequencies were virtually identical to the observed values (V. acadica: 2.00, V. atropilosa: 2.40, V. vidua: 3.02, V. consobrina: 2.85, V. flavopilosa:

3.84; compare to observed values in Table 1). Our estimates of effective paternity and thus

Table 3.1. Summary data from paternity analysis of five Vespula species. Colony data used to generate these summary values are presented in Table S3.2. Site: WA collection occurred in Thurston, Co., Washington. NY collection occurred in Tompkins Co., New York. nc: number of colonies analyzed. nw: arithmetic mean (range) of the number of female offspring genotyped per colony. k: arithmetic mean (95% CI) of the number of male mates detected. ke: harmonic mean (95% CI) of the uncorrected estimate of effective paternity. ke3: harmonic mean (95% CI) of effective paternity corrected for sample size (Nielsen et al. 2003). B: arithmetic mean paternity skew using the B index (Nonacs 2000). # W: arithmetic mean (SD) number of workers collected in mature colonies. NDE: male non-detection error (see text). a. estimate for male non- detection error assuming all parental genotypes are known. b. upper estimate for male non- detection error, accounting for uncertain parental genotypes. Site Year nc nw k ke ke3 B # W NDE V. 2008- 19.9 2.00 1.47 1.51 0.044 - 4.7e-4a WA 10 acadica 2013 (19, 20) (1.42, 2.58) (1.17, 1.98) (1.18, 2.08) (1.3e-3)b V. 2008- 20.1 2.40 1.67 1.72 0.10 - 2.5e-3a WA 10 atropilosa 2013 (19, 23) (1.80, 3.00) (1.37, 2.14) (1.39, 2.26) (1.3e-2)b V. 2008- 20.4 2.83 1.82 1.89 0.069 98 8.0e-6 WA 12 consobrina 2013 (18, 24) (2.00, 3.66) (1.43, 2.50) (1.46, 2.68) (52.0) V. 2012- 19.8 3.00 2.01 2.12 0.056 172 1.7e-5a NY 10 vidua 2013 (18, 21) (2.12, 3.88) (1.62, 2.67) (1.67, 2.90) (103.3) (4.1e-5)b V. 2012- 19.6 3.80 2.80 3.07 0.033 899 1.4e-3 NY 10 flavopilosa 2013 (18, 20) (3.10, 4.50) (2.34, 3.49) (2.50, 3.97) (718.2)

66 relatedness are already corrected for non-sampling, since the calculation of ke3 accounts for non- sampling error (Nielsen et al. 2003). The values of ke3 were typically only slightly larger than ke

(Table S3.2), further suggesting that our sample sizes were sufficient to describe paternity in these species.

Comparative analysis

To analyze how mean colony size parameters and nest site preference influence paternity frequency, paternity skew, and resulting intracolony genetic similarity (relatedness), we compared these traits across 22 vespine taxa, including 2 subspecies. Because species are not independent due to shared ancestry, we performed our analyses using methods accounting for phylogeny (Harvey and Pagel 1991; Symonds and Blomberg).

Data – We searched the literature for all vespine species with reports of paternity based on genetic data. We recorded the arithmetic mean number of patrilines per colony and used harmonic mean effective paternity to calculate intracolony genetic similarity I. Where possible, we acquired colony-level paternity distribution data from authors to calculate the species mean B index of paternity skew for colonies with multiple paternity (Nonacs 2000). We then searched for data on average and maximum mature colony size (worker number). For species with multiple reports of colony size or effective mating frequency, we chose the study with the largest sample size, attempting to use reports of the two variables from the same population whenever possible, and (with the exception of Vespa velutina) only reports from colonies studied within the species’ native ranges. Our goal was to describe the average peak colony size of each species.

Because colony sizes vary over the season, whenever possible we used averages of colonies that

67 had adult reproductives present but had not entered decline. We similarly found data on the maximum number of workers recorded in an annual colony of each species, for colonies within the species’ native range. Finally, we also found estimates of the nest-site preferences of each species, recording the fraction of colonies with enclosed or subterranean nests (as opposed to aerial, exposed nests).

Phylogeny – We generated a phylogenetic hypothesis for our dataset using phyloGenerator

(Pearse and Purvis 2013) based largely on molecular data reported in two recent studies of vespine phylogeny (Perrard et al. 2013; Lopez-Osorio et al. 2014) (Genbank accession numbers available in Table S3.5). Sequences were aligned with MAFFT v6.847b (Katoh 2002), and the phylogeny was estimated with RaxML 7.3.0 (Stamatakis 2006) using a GTRGAMMA model and a single ML run with 1000 integrated bootstraps to determine support. We ultrametricized the resulting tree using the chronopl function in the R package ape with # = 0 to approximate non-parametric rate smoothing (Sanderson 2002). Three species with trait data (Vespula rufa, V. atropilosa, and D. norwegica) were omitted from the analysis because no sequence data were available. Although our resulting tree closely matched the topology for Vespula and

Dolichovespula of Lopez-Osorio et al. (Lopez-Osorio et al. 2014), the additional data used by

Perrard et al. (Perrard et al. 2013) led to a topology for Vespa that differed from ours. To ensure that this difference did not influence the outcome, we performed the same comparative analysis on a tree with a Vespa topology consistent with Perrard et al. (Perrard et al. 2013) and all branch lengths set equal to 1.0 (Garland et al. 1992) (Table S3, Figure S3.1).

68 Statistics – Because species data are not independent due to shared history (Harvey and Pagel

1991), we analyzed comparative data with phylogenetic generalized least squares (PGLS) models (Grafen 1989) implemented in R version 3.0.3 using the caper package (Orme et al.

2013; Team). To improve homoscedasticity we log-transformed colony size and inverse- transformed the skew index B. In each model we determined how well colony size and nest site predict intracolony genetic similarity I, paternity frequency (k) or paternity skew (B), while accounting for phylogenetic history. We also performed a similar analysis using the maximum observed colony size rather than mean colony size for each species (Table S3.4). In PGLS models, branch lengths of a phylogeny are used to correct for the non-independence of regression residuals, the basic assumption of ordinary least squares models that is violated by species data. The transformation parameter # scales the branch lengths of the phylogeny and thus the degree to which the phylogeny affects the regression’s residuals. When # = 0, the phylogeny is scaled to a star phylogeny and the analysis is equivalent to an ordinary least squares model. When #=1, PGLS is equivalent to Felsenstein’s independent contrasts (Felsenstein

1985), the most conservative analysis allowing the greatest effect of phylogeny. Using maximum likelihood methods, # can be optimized to best fit the residuals of the regression model (Revell 2010). We ran each model twice, once with the maximum likelihood value of # and, because the likelihood surfaces were often shallow, once with # set to the upper 95% confidence interval value for the maximum likelihood estimate, for the most conservative analysis (i.e., allowing the greatest effect of phylogeny). To confirm that residuals were distributed normally, residual density plots were checked by eye.

69 Results and Discussion

The first aim of this study was to describe paternity in five species of Vespula wasps.

The second aim was to determine if colony size and nest site predict intracolony genetic

similarity (relatedness) across species, and if this pattern arises through effects of colony size and

nest site on paternity frequency, paternity skew, or both. These patterns are used to address the

question of why multiple mating and multiple paternity evolve.

Paternity

All four species of the Vespula rufa group exhibited moderate multiple paternity, with

most queens mating more than once, and some using the sperm from as many as 6 males (Table

S3.2). The estimated mean effective paternity values for species in the Vespula rufa group were

moderate, varying between 1.5 and 2.1 (Table 1), and similar to V. rufa, which has a mean

effective paternity of 1.5 (Wenseleers et al. 2005). Consequently, these species may provide

interesting fodder for investigating the evolution of worker reproduction: when effective

paternity is near 2.0, relatedness is not predicted to determine whether workers favor worker- or

queen-derived males, as their relatedness to these two types of males is equal (Ratnieks 1988).

Thus, other factors such as costs of worker reproduction or the effectiveness of queen policing

may usefully explain differences between these species. Behavioral observations suggest that

such differences exist: some species appear to have frequent worker reproduction (e.g., V. rufa

(Wenseleers et al. 2005) and V. consobrina (Akre et al. 1982)), while others may have little or no

worker reproduction in queenright colonies (V. acadica (Reed and Akre 1983a), V. atropilosa

(Landolt et al. 1977), and V. vidua, Chien and Loope, unpublished data).

The high average effective paternity of Vespula flavopilosa (~3.1) is similar to other members of the Vespula vulgaris group, though lower than other large-colony species in eastern

70 North America (Hoffman et al. 2008). This suggests no reversion to low paternity due to facultative social parasitism, as was also found in the similar social parasite Vespula 71quamosal

(Hoffman et al. 2008). This makes sense given that queens of both of these species eventually produce colonies with many workers, and thus likely benefit from greater paternity for the same reasons as other large-colony Vespula. It remains to be seen whether the rufa-group social parasite species (e.g., V. infernalis and V. austriaca), which lack a worker caste but are probably derived from a moderately polyandrous ancestor, have reverted to monandry as predicted by the hypothesis of benefits of a genetically diverse workforce (Sumner et al. 2004; Thurin and Aron

2011).

The effects of colony size and nest site on paternity frequency and paternity skew

Our hypotheses for the evolution of multiple paternity all assume that ecology is largely responsible for colony size and nest site variation (Bourke 1999), and that paternity traits evolve in response to these two traits. To test whether colony size and nest site predict paternity traits, we conducted a phylogenetically controlled comparative analysis across 22 taxa (Figure 3.3).

Our results suggest that (1) intracolony genetic diversity increases with colony size across species, and (2) large colony size is associated with both increased paternity frequency and reduced paternity skew (Table 2). We also confirm that paternity frequency and paternity skew explain much of the variation in intracolony genetic diversity, given that they, by definition, should together determine average intracolony relatedness.

Our PGLS models simultaneously estimate regression coefficients and optimize the error structure of the residuals using the # transformation (Grafen 1989; Pagel 1999; Revell 2010), the recommended procedure for analyzing comparative data that may have phylogenetic signal.

71 When # is zero, the phylogenetic signal is estimated to be zero and the analysis is equivalent to ordinary least-squares regression, whereas when # is set to 1, the model incorporates the maximum amount of phylogenetic covariance and is equivalent to independent contrasts analysis

(Felsenstein 1985). The maximum likelihood estimates of # from our models are low to moderate, between zero and 0.35, for models including both colony size and nest site (Table 2), though the uncertainty in the # estimates suggests a possibly large phylogenetic signal.

Regardless, the most conservative analyses (models with # set to the upper 95% confidence interval) yield similar significant effects of colony size on relatedness, paternity frequency, and paternity skew (Table 2). The alternative analyses using a different phylogeny (Table S3.3) or maximum size as a proxy for colony size (Table S3.4) give similar results. Unsurprisingly, actual variation in paternity skew and paternity frequency both significantly predict intracolony relatedness (Table 2). Overall, these results are consistent with the hypothesis that colony size influences intracolony relatedness through changes in paternity frequency and paternity skew.

These findings for the Vespinae are in concord with previous analyses of more phylogenetically diverse sets of species (Schmid-Hempel 1998; Jaffé et al. 2012).

We predicted that the nesting habit influences species’ paternity traits, as subterranean and cavity nesting may expose colonies to more pathogens. The higher paternities of the rufa- group Vespula species, compared with Dolichovespula species of similar colony size, suggest a role of nest site, as does the fact that most high paternity species (with the notable exception of

Vespa velutina) are ground-nesting (Figure 3.4). Furthermore, when including colony size and nest site in a model predicting relatedness, both factors are significant when using the maximum likelihood value for # in the PGLS model (Table 2). However, given the shallow likelihood surface for #, the more conservative analysis with # set to the upper 95% confidence interval

Table 3.2. PGLS models of the effect of colony size and nest site on paternity traits across 22 Vespine taxa Model estimat Response factors "a t p rb e Relatedness log (size) + -0.14 -3.89 <0.001 0.67(0.3, 0.86) 10 0.24(na,0.83) (r) nest site -0.11 -2.27 0.036 0.46(0.01, 0.76) (0.31, 0.86) Paternity log10(size) + 0.86 3.94 <0.001 0.67 frequency 0(na,0.82) nest site 0.96 1.53 0.14 0.33(-0.15, 0.68) "=ML (k) Paternity log (size) + 12.08 3.91 <0.01 0.72(0.31, 0.91) 10 0.35(na,na) skew (B-1) nest site 3.13 0.63 0.53 0.17(-0.40, 0.64) Relatedness k + -0.04 -6.03 <0.001 0.85(0.58, 0.95) 1(na,na) (r) B-1 -0.004 -2.85 0.014 0.61(0.11, 0.86) Relatedness log (size) + -0.13 -3.07 0.007 0.58(0.16, 0.82) 10 0.83 (r) nest site -0.08 -1.30 0.211 0.29(-0.20, 0.66) (0.25, 0.85) Paternity log10(size) + 0.92 3.57 0.002 0.63 "= frequency 0.82 upper nest site 0.57 0.66 0.51 0.15(-0.33,0.57) (k) 95% Paternity log (size)c + 8.81 2.13 0.052 c 0.49(-0.05, 0.81) CI 10 1 skew (B-1) nest site 1.45 0.24 0.81 0.06(-0.49, 0.58) Relatedness k + -0.04 -6.03 <0.001 0.85(0.58, 0.95) 1 (r) B-1 -0.004 -2.85 0.014 0.61(0.11, 0.86) a. values in the #=ML model show the maximum likelihood estimate of # and the 95% confidence interval. “na” means the estimate is outside of the bounds (0, 1). b. effect size calculated from t-values and sample size using compute.es package in R (Del Re). Parenthetical values are 95% confidence intervals. c. effect of colony size on paternity skew is significant if nest site is omitted as a cofactor (t = 2.23, p = 0.042

Polistes dominula (outgroup) Colony size Nest site r K B 100 Dolichovespula saxonica 122 0.33 0.59 1.96 0.079 Dolichovespula arenaria 378 0.30 0.71 1.35 ? 99 100 Dolichovespula sylvestris 134 0.20 0.68 1.36 0.143 100 Dolichovespula media 74 0 0.71 1.18 0.109 Dolichovespula maculata 181 0 0.75 1.00 - 96 Vespula maculifrons 1847 1 0.37 5.64 0.027 70 Vespula avopilosa 97 899 1 0.41 3.80 0.034 82 Vespula vulgaris 100 2098 1 0.51 2.29 0.040 Vespula germanica 1511 1 0.42 3.50 0.026 100 Vespula pensylvanica 681 1 0.45 4.33 0.074 Vespula squamosa 1409 1 0.36 7.25 0.043 100 59 Vespula acadica 139 1 0.58 2.00 0.044 100 Vespula consobrina 98 1 0.51 2.83 0.069 Vespula vidua 172 1 0.49 3.00 0.056 Vespa analis 53 0 0.74 1.05 0.160 74 Vespa crabro 182 1 0.72 1.29 0.189 100 26 Vespa a nis 585 0 0.67 1.50 0.083 36 Vespa mandarinia 136 1 0.74 1.10 0.220 36 Vespa ducalis 18 1 0.75 1.00 - 100 Vespa s. simillima 236 0.46 0.71 1.53 ? V. rufa group 100 Vespa s. xanthoptera 510 0.46 0.46 ? ? V. vulgaris group Vespa velutina 1000 0 ? 4.70 ?

Figure 3.3. Phylogeny and trait data used in comparative analyses. The rate-smoothed phylogeny includes support values from 1000 integrated bootstraps in a single ML run using RaxML 7.3.0. Intracolony relatedness I was calculated from ke or ke3. Paternity frequency (k) is the mean number of observed patrilines present in each colony. Colony size estimates are the arithmetic mean number of adult workers collected from mature colonies. Nest site values are the fraction of observed nests that are in enclosed sites (ie subterranean, tree and man-made structural cavities), as opposed to exposed, aerial nests. The B skew index was averaged across all multiple-paternity colonies. Data come from references (Spradbery 1973; MacDonald et al. 1974; Greene et al. 1976; Edwards 1980; MacDonald and Matthews 1981; Makino 1982; Reed and Akre 1983b; Keyel 1983; MacDonald and Matthews 1984; Matsuura 1984; Matsuura and Yamane 1990; Archer 1993; Archer 1997; Makino and Yamane 1997; Foster et al. 1999; Archer 2000; Foster et al. 2000; Foster and Ratnieks 2001b; Foster et al. 2001; Archer 2002; Takahashi et al. 2002; Takahashi et al. 2004; Wenseleers et al. 2005; Takahashi 2006; Takahashi et al. 2007; Archer 2008; Bonckaert et al. 2008; Hoffman et al. 2008; Martin et al. 2009; Archer 2011; Bonckaert et al. 2011; Archer 2012; Arca 2012; Hanna et al. 2013; Rome et al. 2015).

Figure 3.4. Intracolony relatedness, effective paternity and colony size in vespine wasps. Each point represents mean trait values for a species (data shown in Figure 3.3). This figure includes the 21 species with relatedness and colony size data used in the analyses shown in Table 2, as well as three species (Vespula atropilosa, V. rufa and Dolichovespula norwegica) that were omitted due to lack of phylogenetic information (plotted as diamonds). Model results describing statistical relationships are reported in Table 2. Notably, Vespa velutina is not depicted due to lack of data on relatedness (only paternity frequency was reported), though this species has large mean colony size, high paternity frequency, and is an aerial nester.

75 suggests that colony size, but not nest site, drives the evolution of low intracolony relatedness.

Nest site is also not a significant predictor of paternity frequency or paternity skew when included as a cofactor with colony size (Table 2). Nest site may be significant in the first model predicting relatedness (which is, because # = 0, equivalent to an ordinary least squares model) because this trait is phylogenetically correlated with colony size and highly conserved in the genera Vespula and Dolichovespula. Although nest site is more variable in Vespa, it does not appear to be correlated with paternity in this group (analysis not shown). These mixed results provide little support for the hypothesis that pathogen-laden nest sites select for higher paternity, and the inclusion of additional species providing more phylogenetic contrasts (if they exist) would be useful.

The evolution of multiple paternity in the Vespinae

We have considered the factors associated with the evolution of multiple paternity in a group of social wasps sharing similar annual life histories, colony founding strategies, natural enemies, temperate and subtropical distributions, foraging behaviors, and food sources

(Spradbery 1973; Edwards 1980; Matsuura and Yamane 1990; Greene 1991). Our results confirm an important role of large colony size in the evolution of high paternity in this group.

This pattern is based partly on the correlated independent origins of large colony size and high paternity in the vulgaris and squamosa groups (or a single such origin and then correlated reduction of colony size and paternity in the rufa group; Figure 3.3). It is also supported by transitions to moderate colony size and moderate multiple mating in the hornets, by Vespa simillima, Vespa velutina and Vespa affinis, though these groups are in need of more study.

Further evidence for a link between colony size and paternity would come from confirming the

76 difference reported between subspecies of V. simillima (Martin et al. 2009), and a detailed

description of paternity and colony size for V. velutina in its native range.

What does the strong association with colony size tell us about the selective factors

leading to multiple mating and multiple paternity in these species? The pathogens hypothesis is

consistent with the observed effect of colony size, but remains to be directly tested in vespine

wasps. The predictions that pathogens enter colonies via foragers, that larger colonies within

species have more pathogens, and that species with larger colonies have more pathogens, are all

testable. Numerous parasites and pathogens of social wasps have been identified (Spradbery

1973; Edwards 1980; Schmid-Hempel 1998) but studies of their relative occurrence and

association with colony traits are lacking. The strongest evidence for this hypothesis come from

experimental manipulation of mating frequency in bees (Baer & Schmid-Hempel, 1999; Seeley

& Tarpy, 2007). Such an experiment would provide a powerful test of this hypothesis in vespine

wasps.

The colony size prediction of the division of labor hypothesis rests on the existence of a

difference between small- and large-colony species in behavioral organization. This could be the

case if large-colony species possess more morphological or behavioral castes, improved division

of labor based on genetically determined task thresholds, or a larger behavioral repertoire

(Oldroyd & Fewell, 2007). Such a pattern has been well documented in the polistine wasps, with

greater specialization and evidence supporting a genetic basis underlying specialization, in large- colony species (Jeanne, 1991; O'Donnell, 1996; 1998), as well as differences in task partitioning within species according to colony size (Jeanne, 1986). However, there is little evidence for such a difference in the Vespinae (Jeanne, 2003). Large-colony Vespula workers typically lack long-term specialization (Potter, 1964; Hurd et al., 2007), and there is no evidence that they

77 possess more complex, coordinated behaviors such as recruitment signals or task partitioning of nest construction (Jeanne, 2003). On the other hand, studies of the division of labor and specialization in these wasps are few, and overlooked specialization or complexity in large- colony species, consistent with the division of labor hypothesis, may yet be discovered.

A third popular hypothesis, that queens require multiple mates to obtain enough sperm to last a lifetime (Cole, 1983; Kraus et al., 2004), also predicts an association between multiple paternity and colony size. However, this hypothesis does not predict that queens of large- colony, highly polyandrous species use males’ sperm more evenly than their small-colony, slightly polyandrous counterparts (Table 2; (Jaffé et al., 2012)). Queens that are sperm limited should use all available sperm, and thus paternity skew should reflect the (presumably high) variation in male sperm availability. Therefore, our data suggest that selection for increasing genetic diversity, rather than selection to increase sperm quantity, may explain high paternity in the vespine wasps.

Additional hypothesized benefits of high paternity, such as reducing sex ratio conflict

(Crozier & Page, 1985; Boomsma & Ratnieks, 1996) or obtaining rare, critical patrilines (Fuchs

& Moritz, 1998; Mattila & Seeley, 2010; 2011), are not obviously linked to colony size, and thus seem less likely to explain multiple paternity in the vespine wasps. However, this group may provide useful subjects for further tests of these and other hypotheses, once more is known about their natural enemies, division of labor, mating biology, and sex investment. It will also be valuable to explore an alternative explanation for a link between colony size and paternity not considered here: it may be that increased paternity reduces intra-colony conflicts, increasing productivity and resulting in larger colonies (Bourke, 1999; Ratnieks et al., 2006; Mattila et al.,

2012).

Conclusion

Our results show a strong association between colony size, paternity frequency and paternity skew in the vespine wasps, consistent with earlier, taxonomically broad, analyses. The observed patterns are consistent with hypotheses for the benefits of multiple paternity based on intracolony genetic diversity, but do not support the sperm limitation hypothesis. Clearly, further, more direct, tests of the pathogens hypothesis and the division of labor hypothesis are needed. Comparing closely related species with dramatically different paternity traits but otherwise similar natural history will help to reveal the details of when and why multiple paternity evolves.

79 REFERENCES

Akre RD, Greene A, MacDonald JF, et al (1980) The Yellowjackets of America North of Mexico. United States Department of Agriculture Washington, DC

Akre RD, Reed HC, Landolt PJ (1982) Nesting Biology and Behavior of the Blackjacket, Vespula consobrina (Hymenoptera:Vespidae). J Kansas Entomol Soc 55:375–405.

Arca M (2012) Caractérisation génétique et étude comportementale d’une espèce envahissante en France: Vespa velutina Lepeletier (Hymenoptera, Vespidae). 1–212.

Archer ME (2012) Vespine wasps of the world: behavior, ecology & taxonomy of the Vespinae. Siri Scientific Press, Manchester, UK

Archer ME (2000) The life history and a numerical account of colonies of the social wasp, Dolichovespula norwegica (F.)(Hym., Vespinae) in England. Entomologist's Monthly Magazine (United Kingdom) 136:1–14.

Archer ME (2002) A numerical account of the development of colonies of the social wasp, Dolichovespula sylvestris (Scopoli) (Hym. Vespinae) in England and overseas. Entomologist's Monthly Magazine (United Kingdom) 138:209.

Archer ME (1997) A numerical account of successful colonies of the social wasp, Vespula rufa (L.)(Hym., Vespinae). Entomologist's Monthly Magazine (United Kingdom) 133:205–215.

Archer ME (1993) The life history and colonial characteristics of the hornet, Vespa crabro L. (Hym., Vespinae). Entomologist's Monthly Magazine (United Kingdom) 129:151–163.

Archer ME (2008) Taxonomy, distribution and nesting biology of species of the genus Paravespula or the Vespula vulgaris species group (Hymenoptera, Vespidae). Entomologist's Monthly Magazine (United Kingdom) 144:5–29.

Archer ME (2011) An adventure into the history of a nest of the superwasp Dolichovespula media. Naturalist 136:235.

Bonckaert W, Van Zweden JS, D'Ettorre P, et al (2011) Colony stage and not facultative policing explains pattern of worker reproduction in the Saxon wasp. Mol Ecol 20:3455–3468.: 10.1111/j.1365-294X.2011.05200.x

Bonckaert W, Vuerinckx K, Billen J, et al (2008) Worker policing in the German wasp Vespula germanica. Behav Ecol 19:272–278.

Boomsma JJ (2007) Kin Selection versus Sexual Selection: Why the Ends Do Not Meet. Curr Biol 17:R673–R683.

Boomsma JJ, Baer B, Heinze J (2005) The evolution of male traits in social insects. Annu Rev Entomol 50:395–420.

80 Boomsma JJ, Ratnieks FLW (1996) Paternity in Eusocial Hymenoptera. Phil Trans R Soc B 351:947–975.

Bourke AFG (1999) Colony size, social complexity and reproductive conflict in social insects. J Evol Biol 12:245–257.

Brown M, Schmid-Hempel P (2003) The evolution of female multiple mating in social Hymenoptera. Evolution 57:2067–2081.

Cole BJ (1983) Multiple mating and the evolution of social behavior in the Hymenoptera. Behav Ecol Sociobiol 12:191–201.

Cornuet J-M, Aries F (1980) Number of sex alleles in a sample of honeybee colonies. Apidologie 11:87–93.

Crozier RH, Fjerdingstad E (2001) Polyandry in social Hymenoptera - disunity in diversity? Ann Zool Fenn 38:267–285.

Crozier RH, Page RE (1985) On being the right size: male contributions and multiple mating in social Hymenoptera. Behav Ecol Sociobiol 18:105–115.

Daly D, Archer ME, Watts PC, et al (2002) Polymorphic microsatellite loci for eusocial wasps (Hymenoptera: Vespidae). Mol Ecol Notes 2:273–275.

Del Re AC compute.es: Compute Effect Sizes. In: R package version 0.2-2. http://cran.r- project.org/web/packages/compute.e. Accessed 28 Jun 2014

Edwards R (1980) Social wasps: their biology and control. Rentokil, London

Felsenstein J (1985) Phylogenies and the comparative method. Am Nat 125:1–15.

Foster KR, Ratnieks FLW (2001a) Paternity, reproduction and conflict in vespine wasps: a model system for testing kin selection predictions. Behav Ecol Sociobiol 50:1–8.

Foster KR, Ratnieks FLW (2001b) Convergent evolution of worker policing by egg eating in the honeybee and common wasp. Proc R Soc Lond B 268:169–174.

Foster KR, Ratnieks FLW, Gyllenstrand N, Thoren P (2001) Colony kin structure and male production in Dolichovespula wasps. Mol Ecol 10:1003–1010.

Foster KR, Ratnieks FLW, Raybould A (2000) Do hornets have zombie workers? Mol Ecol 9:735–742.

Foster KR, Seppa P, Ratnieks FLW, Thoren P (1999) Low paternity in the hornet Vespa crabro indicates that multiple mating by queens is derived in vespine wasps. Behav Ecol Sociobiol 46:252–257.

Fuchs S, Moritz R (1998) Evolution of extreme polyandry in the honeybee Apis mellifera L. Behav Ecol Sociobiol 45:269–275.

81 Garland T, Harvey PH, Ives AR (1992) Procedures for the analysis of comparative data using phylogenetically independent contrasts. Syst Biol 41:18–32.

Grafen A (1989) The phylogenetic regression. Phil Trans R Soc B 326:119–157.

Greene A (1991) Dolichovespula and Vespula. In: Ross K, Matthews RW (eds) The Social Biology of Wasps. Cornell University Press, Ithaca, NY, pp 263–305

Greene A, Akre RD, Landolt PJ (1976) The aerial yellowjacket, Dolichovespula arenaria (Fab): nesting biology, reproductive production, and behavior (Hymenoptera: Vespidae). Melanderia 26:1–34.

Hamilton WD (1987) Kinship, recognition, disease and intelligence: constraints of social evolution. In: Ito Y, Brown JL, Kikkawa J (eds) Animal societies: theory and facts. Japanese Scientific Society, Tokyo, pp 81–102

Hanna C, Cook ED, Thompson AR, et al (2013) Colony social structure in native and invasive populations of the social wasp Vespula pensylvanica. Biol Invasions 16:283–294.

Harvey PH, Pagel MD (1991) The comparative method in evolutionary biology. Oxford University Press, Oxford

Hasegawa E, Takahashi J (2002) Microsatellite loci for genetic research in the hornet Vespa mandarinia and related species. Mol Ecol Notes 2:306–308.

Hoffman EA, Kovacs JL, Goodisman MAD (2008) Genetic structure and breeding system in a social wasp and its social parasite. BMC Evol Biol 8:239.

Hughes WOH, Oldroyd BP, Beekman M, Ratnieks FLW (2008) Ancestral monogamy shows kin selection is key to the evolution of eusociality. Science 320:1213–1216.

Jaffé R (2014) An Updated Guide to the Study of Polyandry in Social Insects. Sociobiology 61:1–8.

Jaffé R, Garcia-Gonzalez F, Boer den SPA, et al (2012) Patterns of paternity skew among polyandrous social insects: what can they tell us about the potential for sexual selection? Evolution 66:3778–3788.

Jeanson R, Fewell JH, Gorelick R, Bertram SM (2007) Emergence of increased division of labor as a function of group size. Behav Ecol Sociobiol 62:289–298.

Jones OR, Wang J (2010) COLONY: a program for parentage and sibship inference from multilocus genotype data. Mol Ecol Resour 10:551–555.

Katoh K (2002) MAFFT: a novel method for rapid multiple sequence alignment based on fast Fourier transform. Nucleic Acids Res 30:3059–3066.

Keyel R (1983) Some aspects of niche relationships among yellowjackets (Hymenoptera:

82 Vespidae) of the northeastern United States. 1–180.

Landolt PJ, Akre RD, Greene A (1977) Effects of colony division on Vespula atropilosa (Sladen)(Hymenoptera: Vespidae)[Insects]. J Kansas Entomol Soc 50:135–147.

Lopez-Osorio F, Pickett KM, Carpenter JM, et al (2014) Phylogenetic relationships of yellowjackets inferred from nine loci (Hymenoptera: Vespidae, Vespinae, Vespula and Dolichovespula). Mol Phylogenet Evol 73:190–201.

MacDonald JF, Akre R, Hill W (1974) Comparative biology and behavior of Vespula atropilosa and V. pensylvanica (Hymenoptera: Vespidae). Melanderia 18:1–66.

MacDonald JF, Matthews RW (1981) Nesting biology of the Eastern yellowjacket, Vespula maculifrons (Hymenoptera: Vespidae). J Kansas Entomol Soc 54:433–457.

MacDonald JF, Matthews RW (1984) Nesting biology of the southern yellowjacket, Vespula squamosa (Hymenoptera: Vespidae): social parasitism and independent founding. J Kansas Entomol Soc 57:134–151.

MacDonald JF, Matthews RW, Jacobson R (1980) Nesting biology of the yellowjacket, Vespula flavopilosa (Hymenoptera: Vespidae). J Kansas Entomol Soc 53:448–458.

Makino S (1982) Nest structure, colony composition and productivity of Dolichovespula media media and D. saxonica nipponica in Japan (Hymenoptera, Vespidae). Kontyu 50:212–224.

Makino S, Yamane S (1997) Nest Contents and Colonial Adult Productivity in a Common Hornet, Vespa simillima simillima SMITH, in Northern Japan (Hymenoptera, Vespidae). Jpn J Entomol 65:47–54.

Martin SJ, Takahashi J, Katada S (2009) Queen condition, mating frequency, queen loss, and levels of worker reproduction in the hornets Vespa affinis and V. simillima. Ecol Entomol 34:43–49.

Matsuura M (1984) Comparative biology of the five Japanese species of the genus Vespa (Hymenoptera, Vespidae). The Bulletin of the Faculty of Agriculture, Mie University 69:1– 131.

Matsuura M, Yamane S (1990) Biology of the Vespine Wasps. Springer-Verlag, Berlin

Nielsen R, Tarpy DR, Reeve HK (2003) Estimating effective paternity number in social insects and the effective number of alleles in a population. Mol Ecol 12:3157–3164.

Nonacs P (2000) Measuring and using skew in the study of social behavior and evolution. Am Nat 156:577–589.

Oldroyd BP, Clifton MJ, Wongsiri S, et al (1997) Polyandry in the genus Apis, particularly Apis andreniformis. Behav Ecol Sociobiol 40:17–26.

83 Oldroyd BP, Fewell JH (2007) Genetic diversity promotes homeostasis in insect colonies. TREE 22:408–413.

Orme D, Freckleton R, Thomas G, et al (2013) caper: Comparative Analyses of Phylogenetics and Evolution in R. R package version 0.5.2 2:

Pagel M (1999) Inferring the historical patterns of biological evolution. Nature 401:877–884.

Pearse WD, Purvis A (2013) phyloGenerator: an automated phylogeny generation tool for ecologists. Methods Ecol Evol 4:692–698.

Perrard A, Pickett K, Villemant C, et al (2013) Phylogeny of hornets: a total evidence approach (Hymenoptera, Vespidae, Vespinae, Vespa). J Hymenoptera Res 32:1–15.

Ratnieks FLW (1988) Reproductive harmony via mutual policing by workers in eusocial Hymenoptera. Am Nat 132:217–236.

Ratnieks FLW, Foster KR, Wenseleers T (2006) Conflict resolution in insect societies. Annu Rev Entomol 51:581–608.

Reed HC, Akre RD (1983a) Comparative colony behavior of the forest yellowjacket, Vespula acadica (Sladen)(Hymenoptera: Vespidae). J Kansas Entomol Soc 56:581–606.

Reed HC, Akre RD (1983b) Nesting biology of a forest yellowjacket Vespula acadica (Sladen)(Hymenoptera: Vespidae), in the Pacific Northwest. Ann Entomol Soc Am 76:582– 590.

Revell LJ (2010) Phylogenetic signal and linear regression on species data. Methods Ecol Evol 1:319–329.

Robinson GE (1992) Regulation of Division of Labor in Insect Societies. Annu Rev Entomol 37:637–665.

Rome Q, Muller FJ, Touret-Alby A, et al. (2015) Caste differentiation and seasonal changes in Vespa velutina (Hym.: Vespidae) colonies in its introduced range. J Appl Ecol

Sanderson MJ (2002) Estimating absolute rates of molecular evolution and divergence times: a penalized likelihood approach. Mol Biol Evol 19:101–109.

Schmid-Hempel P (1998) Parasites in Social Insects. Princeton University Press, Princeton, NJ

Schuelke M (2000) An economic method for the fluorescent labeling of PCR fragments. Nat Biotechnol 18:233–234.

Sherman PW, Seeley TD, Reeve HK (1988) Parasites, Pathogens, and Polyandry in Social Hymenoptera. Am Nat 133:602–610.

Spradbery JP (1973) Wasps. University of Washington Press, Seattle

84 Stamatakis A (2006) RAxML-VI-HPC: maximum likelihood-based phylogenetic analyses with thousands of taxa and mixed models. Bioinformatics 22:2688–2690.

Stein KJ, Fell R (1996) Sperm use dynamics of the baldfaced hornet (Hymenoptera: Vespidae). Environ Entomol 25:1365–1370.

Strassmann JE (2001) The rarity of multiple mating by females in the social Hymenoptera. Insectes Soc 48:1–13.

Sumner S, Hughes WOH, Pedersen JS, Boomsma JJ (2004) Ant parasite queens revert to mating singly. Nature 428:35–36.

Symonds, M. R. E., & Blomberg, S. P. (2014). A Primer on Phylogenetic Generalised Least Squares. In L. Z. Garamsegi, Modern phylogenetic comparative methods and their application in evolutionary biology (pp. 105–130). Heidelberg: Springer-Verlag.

Takahashi J (2006) Evolutional biology in the hornet – cooperation and conflict within the colony. The Nature and Insects 41:9–14.

Takahashi J, Akimoto S, Hasegawa E, Nakamura J (2002) Queen mating frequencies and genetic relatedness between workers in the hornet Vespa ducalis (Hymenoptera: Vespidae). Appl Entomol Zool 37:481–486.

Takahashi J, Akimoto S, Martin SJ, et al (2004) Mating structure and male production in the giant hornet Vespa mandarinia (Hymenoptera: Vespidae). Appl Entomol Zool 39:343–349.

Takahashi J, Inomata Y, Martin SJ (2007) Mating structure and male production in Vespa analis and Vespa simillima (Hymenoptera: Vespidae). Entomol Sci 10:223–229.

R Core Development Team. R: A language and environment for statistical computing.

Thorén PA, Paxton RJ, Estoup A (1995) Unusually high frequency of (CT)n and (GT)n microsatellite loci in a yellowjacket wasp, Vespula rufa (L.) (Hymenoptera: Vespidae). Insect Mol Biol 4:141–148.

Thurin N, Aron S (2011) No reversion to single mating in a socially parasitic ant. J Evol Biol 24:1128–1134.

Walker TN, Hughes WO (2011) Arboreality and the evolution of disease resistance in ants. Ecol Entomol 36:588–595.

Wedell N, Gage MJ, Parker GA (2002) Sperm competition, male prudence and sperm-limited females. TREE 17:313–320.

85 Wenseleers T, Badcock N, Erven K, et al (2005) A test of worker policing theory in an advanced eusocial wasp, Vespula rufa. Evolution 59:1306–1314.

86 Appendix 3.1. Supplementary figures and tables

Table S3.1. The number of alleles (above) and expected heterozygositya (below) at each locus. List List List Rufa Rufa Rufa Rufa Rufa VMA VMA

2004 2009 2019 05 13 15 18 19 3 6 5 9 13 12 5 V. acadica 0.66 0.83 0.87 0.86 0.52 V. 5 10 12 11 atropilosa 0.68 0.81 0.74 0.84 V. 8 7 14 18 14 20 consobrina 0.83 0.66 0.89 0.88 0.86 0.93 3 8 20 15 12 18 V. vidua 0.5 0.79 0.90 0.89 0.85 0.91 V. 5 5 5 9 17 flavopilosa 0.55 0.67 0.69 0.77 0.87 a. Expected heterozygosity was calculated from adjusted allele frequencies determined by Colony2 after assigning parentage and accounting for family structure.

87 Table S3.2. Descriptive data for colonies of Vespula acadica (VAC), V. atropilosa (VAT), V. consobrina (VC), V. vidua (VV), V. flavopilosa (VF). nw is number of workers successfully genotyped, k is number of male mates detected, ke is an estimate of effective paternity, ke3 is a corrected estimate of effective paternity (see Methods). B is paternity skew, # W is number of workers collected, and Q indicates presence of queen at collection. a. queen not present in sample but genotyped eggs/young larvae were diploid, suggesting colony likely queenright. b. queen collected but not genotyped. c. queen collected but desiccation suggests she had died before collection. Colony nw k ke ke3 B # W Q? VAC1 19 1 1.00 1.00 - - - VAC2 20 1 1.00 1.00 - - - VAC3 20 2 2.00 2.10 -0.025 - - VAC4 20 3 2.47 2.65 0.0383 - - VAC5 20 3 1.94 2.03 0.1483 - - VAC6 20 3 2.06 2.17 0.1183 - - VAC7 20 1 1.00 1.00 - - - VAC8 20 1 1.00 1.00 - - - VAC9 20 3 2.67 2.89 0.0083 - - VAC10 20 2 2.00 2.10 -0.025 - - VAT1 20 2 2.00 2.10 -0.025 - - VAT2 20 4 2.60 2.81 0.0975 - - VAT3 20 3 2.38 2.55 0.0533 - - VAT4 19 1 1.00 1.00 - - - VAT5 23 2 1.91 1.98 0.002 - - VAT6 20 2 1.72 1.79 0.055 - - VAT7 20 2 1.10 1.11 0.38 - - VAT8 20 4 2.47 2.65 0.117 - - VAT9 20 2 1.34 1.36 0.22 - - VAT10 19 2 1.87 1.95 0.008 - - VC1 20 4 2.60 2.81 0.097 91 Ya VC2 20 2 1.98 2.08 -0.02 99 Y VC3 18 1 1.00 1.00 - 185 Y VC4 20 2 1.98 2.08 -0.02 49 Y VC5 19 4 3.42 3.89 -0.01 79 N VC6 20 4 3.23 3.60 0.022 - Y VC7 20 1 1.00 1.00 - 39 Ya VC8 20 2 1.92 2.01 -0.005 98 Y VC9 20 6 3.64 4.15 0.067 140 Y VC10 20 3 1.65 1.71 0.238 166 Ya VC11 24 3 2.80 3.01 -0.003 34 Ya VC12 23 2 1.19 1.20 0.319 91 Y VV1 18 4 2.66 2.91 0.085 128 Y VV2 21 4 2.64 2.85 0.093 193 Y VV3 20 5 2.47 2.65 0.165 316 Y VV4 20 3 2.06 2.17 0.118 336 Y

88 VV5 20 2 1.98 2.08 -0.02 153 Y b VV6 20 2 1.92 2.01 -0.005 69 Y VV7 20 1 1.00 1.00 - 111 Y VV8 20 5 3.85 4.44 0.02 72 Y VV9 20 2 2.00 2.10 -0.025 - - VV10 19 2 1.86 1.94 0.071 - - VF1 20 3 2.41 2.58 0.048 - - VF2 20 4 2.67 2.89 0.087 1839 Y a VF3 20 4 3.17 3.54 0.027 1755 Y VF4 20 3 1.94 2.03 0.148 214 Y VF5 20 3 2.41 2.58 0.048 1580 Y VF6 20 4 4.00 4.66 0.037 183 Y VF7 20 4 3.51 3.98 0.002 910 Y c VF8 19 5 3.97 4.65 0.01 130 Y VF9 19 2 1.87 1.95 0.008 280 Y VF10 18 6 4.76 5.92 0.003 1199 Y

Table S3.3. Comparative analyses of colony size and nest site on effective mating frequency across 21 species of Vespine wasps using alternative phylogeny based on topology of Lopez- Osorio et al. (2014) and Perrard et al. (2014). Model Response factors λa estimate t p rb (0.4, 0.89) Relatedness log10(size) + -0.16 -4.69 <0.001 0.73 0(na,0.88) (r) nest site -0.12 -2.60 0.018 0.51(0.07, 0.79) (0.36, 0.88) Paternity log10(size) + 0.90 4.28 <0.001 0.70 0(na,0.63) frequency (k) nest site 1.01 1.65 0.12 0.35(-0.12, 0.70) λ=ML Paternity log (size) + 12.47 3.07 <0.01 0.63(0.15, 0.87) 10 0.42(na,na) skew (B-1) nest site 0.94 0.20 0.84 0.05(-0.49, 0.57) Relatedness k + -0.05 -6.41 <0.001 0.86(0.61, 0.96) 0(na,na) (r) B-1 -0.006 -4.35 <0.001 0.76(0.38, 0.92) (0.10, 0.80) Relatedness log10(size) + -0.12 -2.76 0.013 0.53 0.88 (r) nest site -0.07 -1.20 0.25 0.27(-0.22, 0.64) Paternity log (size) + 0.89 3.60 0.002 0.64(0.25, 0.85) λ= 10 0.63 frequency (k) nest site 0.73 0.94 0.36 0.21(-0.27, 0.61) upper Paternity log (size) + 10.32 2.27 0.040 0.52(-0.02, 0.82) 10 1 95% CI skew (B-1) nest site -1.27 -0.26 0.79 0.07(-0.53, 0.66) Relatedness k + -0.04 -6.48 0.000 0.87(0.62, 0.96) 1 (r) B-1 -0.003 -2.21 0.045 0.51(-0.03, 0.82) a. values in the λ=ML model show the maximum likelihood estimate for lambda and the 95% confidence interval. “na” means the estimate is outside of the bounds (0, 1). b. Effect size was calculated from t-values and sample size using compute.es package in R. Parenthetical values are 95% confidence intervals.

Table S3.4. Comparative analyses of maximum colony size and nest site on effective mating frequency across 22 vespine taxa Model Response factors λa estimate t p rb (0.26, 0.85) Relatedness log10(size) + -0.11 -2.49 0.023 0.64 0.36(na,0.93) (r) nest site -0.06 -0.90 0.11 0.37(-0.11, 0.70) (0.30, 0.86) Paternity log10(size) + 0.94 3.90 0.001 0.67 0(na,0.91) frequency (k) nest site 0.79 1.22 0.24 0.27(-0.21, 0.65) λ=ML Paternity log (size) + 11.49 2.78 0.015 0.60(0.09, 0.86) 10 0.26(na,na) skew (B-1) nest site 3.26 0.62 0.55 0.16(-0.41, 0.64) Relatedness k + -0.038 -6.03 <0.001 0.85(0.58, 0.95) 1(na,na) (r) B-1 -0.004 -2.85 0.014 0.61(0.11, 0.86) Relatedness log (size) + -0.12 -2.76 0.013 0.50(0.05, 0.78) 10 0.93 (r) nest site -0.07 -1.20 0.25 0.27(-0.22, 0.64) Paternity log (size) + 0.99 3.71 0.002 0.65(0.27, 0.85) λ= 10 0.91 frequency (k) nest site 0.20 0.22 0.82 0.05(-0.44, 0.54) upper (-0.14, 0.78) Paternity log10(size) + 7.12 1.76 0.10 0.43 1 95% CI skew (B-1) nest site 0.81 0.13 0.90 0.03(-0.51, 0.56) Relatedness k + -0.038 -6.03 <0.001 0.85(0.58, 0.95) 1 (r) B-1 -0.004 -2.85 0.014 0.61(0.11, 0.86) a. values in the λ=ML model show the maximum likelihood estimate for lambda and the 95% confidence interval. “na” means the estimate is outside of the bounds (0, 1). b. Effect size was calculated from t-values and sample size using compute.es package in R. Parenthetical values are 95% confidence intervals.

Table S3.5. Genbank accession numbers for sequences used in phylogenetic anaylsis. Gene

Species 12S 16S 28S COI H3 EF1a COII CytB Apol2 win KJ147180.1 KJ147207.1 KF981699.1 KJ147236.1 KF981673.1 KJ147264.1 KJ147292.1 KF955646.1 Polistes dominula Dolichovespula KJ147202.1 KF981694.1 KJ147231.1 KF981668.1 KJ147259.1 KJ147287.1 KF981643.1 KF955641.1 maculata Dolichovespula KJ147175.1 KJ147228.1 KF981693.1 KJ147230.1 KF981667.1 KJ147258.1 KJ147286.1 KF981642.1 KF955640.1 arenaria Dolichovespula KJ147179.1 KJ147206.1 KF981698.1 KJ147235.1 KF981672.1 KJ147263.1 KJ147291.1 KF981647.1 KF955645.1 sylvestris Dolichovespula KJ147176.1 KJ147203.1 KF981695.1 KJ147232.1 KF981669.1 KJ147260.1 KJ147288.1 KF981644.1 KF955642.1 media Dolichovespula KJ147178.1 KJ147205.1 KF981697.1 KJ147234.1 KF981671.1 KJ147262.1 KJ147290.1 KF981646.1 KF955644.1 saxonica GU207849.1 Vespula vulgaris Vespula KJ147197.1 KJ147224.1 KF981715.1 KJ147254.1 KF981689.1 KJ147282.1 KJ147310.1 KF981662.1 KF955662.1 pensylvanica KJ147194.1 KJ147221.1 KF981712.1 KJ147251.1 KF981686.1 KJ147279.1 KJ147307.1 KF981659.1 KF955659.1 Vespula germanica Vespula KJ147196.1 KJ147223.1 KF981714.1 KJ147253.1 KF981688.1 KJ147281.1 KJ147309.1 KF981661.1 KF955661.1 maculifrons KJ147193.1 KJ147220.1 KF981711.1 KJ147250.1 KF981685.1 KJ147278.1 KJ147306.1 KF981658.1 KF955658.1 Vespula flavopilosa KJ147198.1 KJ147225.1 KF981716.1 KJ147255.1 KF981690.1 KJ147283.1 KJ147311.1 KF981663.1 KF955663.1 Vespula squamosa KJ147199.1 KJ147226.1 KF981717.1 KJ147256.1 KF981691.1 KJ147284.1 KJ147312.1 KF981664.1 KF955664.1 Vespula vidua KJ147191.1 KJ147218.1 KF981709.1 KJ147248.1 KF981684.1 KJ147276.1 KJ147304.1 KF981657.1 KF955657.1 Vespula consobrina KJ147189.1 KJ147216.1 KF981708.1 KJ147246.1 KF981682.1 KJ147274.1 KJ147302.1 KF981655.1 KF955655.1 Vespula acadica KJ147188.1 KJ147215.1 KF981706.1 KJ147244.1 KF981680.1 KJ147272.1 KJ147300.1 KF981653.1 KF955653.1 Vespa crabro KJ147186.1 KJ147213.1 KJ147242.1 KF981678.1 KJ147270.1 KJ147298.1 KF955651.1 Vespa affinis KF933053.1 KF933064.1 KF933076.1 KF933084.1 KF933092.1 Vespa ducalis Vespa simillima KF933049.1 KF933077.1 KF933080.1 KF933090.1 xanthoptera KF933055.1 KF933062.1 KF933068.1 KF933085.1 Vespa mandarinia AB585940.1 AB585948.1 AB585957.1 Vespa analis Vespa simillima HM180937.1 simillima Vespa velutina KF933050.1 KF933058.1 KF933073.1 KF933081.1

Figure S3.1. Phylogeny used in analyses shown in Table S3.3. Dolichovespula arenaria Dolichovespula saxonica Dolichovespula sylvestris Dolichovespula media Dolichovespula maculata Vespula consobrina Vespula acadica Vespula vidua Vespula squamosa Vespula penyslvanica Vespula germanica Vespula maculifrons Vespula avopilosa Vespula vulgaris Vespa analis Vespa velutina Vespa simillima xanthoptera Vespa simillima simillima Vespa crabro Vespa anis Vespa mandarinia Vespa ducalis

93 CHAPTER 4

QUEEN KILLING IS LINKED TO HIGH WORKER-WORKER RELATEDNESS IN A

SOCIAL WASP

Kevin J. Loope

Summary

Social insect colonies are pinnacles of evolved altruism, but can also be battlegrounds of conflict among relatives (Hamilton 1972; Ratnieks et al. 2006). In many species, a colony’s workers compete with the queen and each other over the production of males. Interspecific comparisons demonstrate the importance of within-colony relatedness in determining the outcome of this conflict (Foster and Ratnieks 2001a; Wenseleers and Ratnieks 2006), but facultative responses to within-colony relatedness are rare (Foster and Ratnieks 2000; Hammond et al. 2003; Bonckaert et al. 2011b). Here, I report facultative matricide (worker killing of a colony’s queen) in the social wasp, Dolichovespula arenaria. Matricide is strongly associated with high worker-worker relatedness, as predicted by theory, because closely-related workers value nephews more than brothers (Bourke 1994). This pattern is the result of variation in both paternity frequency and the paternity skew of colonies with multiple patrilines, implicating worker-worker relatedness rather than a direct effect of multiple mating on queen survival. Furthermore, occasional inbreeding can explain why some multiple-patriline colonies exhibit high paternity skew associated with matricide. In general, these results support the hypothesis that workers can facultatively respond to intracolony relatedness determined by queen mating behavior, and demonstrate a novel benefit of polyandry in annual social insects. Facultative matricide shows dramatically how workers are evolutionary actors with interests that can diverge from the queen’s, rather than ‘extrasomatic projections of her personal genome’ (Nowak et al. 2010).

94 Results and Discussion

In many species of social wasps and bees, colonies are founded in the spring by a single queen, produce several generations of female workers, rear new queens and males, and finally senesce in late summer or autumn. Workers do not mate but, due to haplodiploidy, can lay unfertilized male eggs (Figure 4.1). In colonies with the queen present, worker reproduction is inhibited through egg eating by the queen and other workers (Foster and Ratnieks 2001b;

Wenseleers et al. 2005; Zanette et al. 2012). It has been proposed that workers of some species kill their mother queen in order to evade this reproductive suppression, and matricide has been anecdotally observed or inferred in numerous species of wasps and bees (Bourke 1994; Foster and Ratnieks 2001a; Strassmann et al. 2003). Queen killing is potentially costly: the irreplaceable queen is the only colony member capable of producing the fertilized eggs that become new workers and queens (Trivers and Hare 1976). However, matricide during the reproductive phase, at the end of an annual colony’s life, could benefit aspiring reproductive workers because it both stops the queen from eating worker-laid eggs and attacking ovipositing workers, and removes a competing source of male eggs (Bourke 1994; Foster and Ratnieks

2001a). In large colonies, a single potentially matricidal worker is unlikely to dominate reproduction after the queen is dead. Thus, the gain from queen killing would largely come from replacing the queen’s sons with other workers’ sons (Bourke 1994). Workers are more related to their fellow workers’ sons than to the queen’s sons if the queen has fewer than two effective mates, when average worker-worker relatedness is greater than 0.5 (Starr 1984; Bourke 1994).

Therefore, matricide should be most common in colonies with effective paternity less than 2.0, when colony resources are redirected mostly to the sons of full siblings. Interspecific

! "

Figure 4.1. Queen and worker reproduction in Dolichovespula arenaria. a. A queen lays an egg while workers tend brood (photo by Barrett Klein). b. A paint-marked worker lays a male egg the day after her sister, another reproductive worker, stung the queen to death (see Appendix 4.2 for description).

96 comparisons support this prediction (Bourke 1994; Foster and Ratnieks 2001a), though it

remains untested within a partially polyandrous species.

To determine the role of worker-worker relatedness in the evolution of matricide, I

genotyped workers from 21 colonies of Dolichovespula arenaria, a yellowjacket wasp,

predicting that matricidal colonies would have higher worker-worker relatedness than colonies

that retained their queens. Video observations on two of these colonies revealed matricide

directly (see Appendix 4.2 for a description of observed matricides). The remaining 19 were

wild-collected, mature colonies: ten were collected with a queen present, and nine were

queenless with inferred matricide. For these latter colonies, estimates based on queen loss rates

in similar species without matricide suggest that ~89% of these queenless, mature D. arenaria

colonies experienced matricide (see Experimental Procedures; the natural frequency of such

colonies is 42%). Paternity analysis of workers shows that worker-worker relatedness is strongly

associated with whether or not a queen is killed (Figure 4.2a). Queen mating frequency explains

part of this association: queens from 6 of 7 single-paternity colonies did not survive, and 9 of 10

colonies with surviving queens had multiple patrilines. In addition, among the colonies with

multiple patrilines, paternity skew is significantly greater in queenless colonies (B index (Nonacs

2000); queenless: 0.24±0.069 mean±SEM; queenright: 0.014±0.020; unequal variances t-test:

n1,2=5,9; t=3.10; df=4.69; p=0.029; Figure S4.1). This higher skew results in marginally higher

relatedness in queenless multiple-patriline colonies compared to queenright multiple-patriline

colonies (queenless: 0.61±0.047; queenright: 0.50±0.0041; unequal variances t-test: n1,2 = 5,9; t =

2.33; df = 4.06; p = 0.079). Thus, workers also appear to distinguish among multiply-mated queens, suggesting that they are directly assessing the level of worker-worker relatedness

" !

!"%# @ : :"! /(<<(,7.=(/367(1).7>/ 2

!"$# 367(1).7>/+861( / !"# /

!"## (?

0-12(1/3(4.51((/1(*67(4)(++ ;

!"9# !"! &'(()1.587 &'(()*(++ B!"; !"!/ !"; ,-*-).(+ ,-*-).(+ A6*(B&'(()/1(*67(4)(++

Figure 4.2. Matricide, paternity and inbreeding in D. arenaria. A. Worker-worker relatedness within colonies that are queenless due to matricide (red points: observed matricide, black points: inferred matricide) is higher than within colonies that remain queenright (all queenless colonies: 0.69±0.030 mean±SEM, n=11; queenright: 0.52±0.025, n=10; unequal variances t-test on ranks: t=3.40, df=18.95, p=0.003; observed matricide only vs queenright: t=4.04, df=5.86, p=0.007). This difference is predicted by kin selection theory7, because when worker-worker relatedness is greater than 0.5 (dotted line), workers value other workers’ sons more than the queen’s sons. The only monandrous queenright colony (asterisk) had just entered the reproductive stage when it was collected, unlike the nine other queenright colonies, suggesting this queen may yet have been killed (see Chapter 5). For detailed paternity data, see Table S4.1. B. Males with a low paternity share are more related to their mates in queenless colonies. Points connected by lines represent males mated to the same queen. Male-queen relatedness, estimated using inferred parental genotypes, is an index of inbreeding. Inbreeding avoidance or depression could therefore explain high paternity skew in these colonies, which is associated with matricide. This pattern is not present in queenright colonies (see Figure S4.1b).

98 (influenced by paternity skew) rather than some cue indicating the number of times a queen has

mated.

Why would a multiply-mated queen skew the paternity in her colony, and thereby

increase worker-worker relatedness, if workers kill queens that do so? Partial inbreeding (queens

mating to both a relative and a non-relative) could result in the observed paternity skew

measured in adult workers, if inbreeding depression causes immature offspring mortality, or if

queens bias sperm use to avoid using sperm from relatives (Bretman et al. 2004). Given that

some yellowjacket wasps readily copulate with siblings (Kovacs et al. 2008), I looked for an

effect of inbreeding on paternity share in multiple-patriline, matricide colonies. Consistent with inbreeding-driven paternity skew, the minority father is significantly more related to the queen than is the majority father (paired t-test; mean difference=0.26; n=5, t=3.68, df=4, p=0.021;

Figure 4.2b), and these minority patriline fathers from queenless colonies are significantly more related to their mates than are the remaining males across all colonies (unequal variances t-test: n1,2=5,32; t=2.77, df=8.14, p=0.024). However, there is no such difference between majority

and minority fathers in queenright colonies (paired t-test; df = 9, p>0.05; Figure S4.1b), and

overall, male-queen relatedness was not different from zero (relatedness: -0.03±0.07, mean ±

95% CI, jackknifing over loci), consistent with a lack of inbreeding in low-skew, queenright

colonies. These patterns suggest that occasional inbreeding may cause the high paternity skew

associated with the killing of some multiply-mated queens.

The main result, that workers facultatively kill queens based on worker-worker

relatedness determined by colony paternity, is difficult to explain with alternative hypotheses.

Multiple mating typically decreases female lifespan in insects (Arnqvist and Nilsson 2000),

including annual social Hymenoptera (Baer and Schmid-Hempel 2001), and a direct benefit of

99 re-mating (e.g. a nuptial gift or nutrient transfer via seminal fluid) cannot explain the association

between paternity skew and polyandrous queen death. Multiple mating increases female egg

production rate in some insects (Arnqvist and Nilsson 2000), and this could, under some

circumstances, disfavor matricide behavior (Bourke 1994), but an experiment that reduced queen fecundity did not induce matricide (see Chapter 5). Instead, the hypothesis most consistent with the results reported here is that workers kill their queen when they are more related to nephews than brothers, as predicted by kin selection theory (Bourke 1994).

Most examples of facultative responses to intracolony relatedness come from studies of

sex allocation (Meunier et al. 2008), though only a few involve paternity variation in species

with single-queen societies (Sundström 1994; Sundström et al. 1996). To my knowledge, the

only other example of a facultative worker behavioral response to natural variation in colony

paternity is facultative policing of worker-laid eggs in Dolichovespula saxonica (Foster and

Ratnieks 2000), a wasp very similar to D. arenaria, though this finding was not replicated in a

different population (Bonckaert et al. 2011b). Why would facultative responses to relatedness

determined by queen mating behavior not be widespread? Information available to workers on a

colony’s paternity can be limited (Boomsma et al. 2003), given that kin-informative recognition

cues are likely evolutionarily unstable (Rousset and Roze 2007). Responses to intracolony

relatedness may arise and exist until the loss of recognition cues make them maladaptive,

producing a patchwork of populations, some of which have sufficient information to assess and

respond to relatedness variation. Although the mechanism by which workers detect relatedness

is unknown, facultative matricide in D. arenaria reiterates the importance of kin structure in

social evolution and suggests an additional conflict-driven benefit of polyandry (Crozier and

Fjerdingstad 2001; Mattila et al. 2012) for queens of the annual social insects.

100 Experimental Procedures

Study species

Dolichovespula arenaria is a common North American aerial-nesting vespine wasp. The nesting biology of this species is similar to most annual yellowjacket wasps (Vespula spp. and

Dolichovespula spp.) (Greene et al. 1976). Colonies are founded in the spring by a single queen.

After the first workers emerge, the colony grows rapidly, constructing 1-3 worker combs, one below the next, before the initiation of the reproductive stage in mid summer, when workers build larger cells to rear new queens and males. Colonies in my study population in Tompkins

Co., NY typically construct 3-6 combs in total, and the largest colonies have a peak worker population of over 800 (KJL unpublished data), sometimes producing hundreds of new queens or males (Table S4.1). Colonies senesce in late summer, and only new queens survive the winter after mating in the autumn.

Colony collections

D. arenaria colonies were collected on private property in Tompkins Co., NY from June-August,

2010-2013. This species is not threatened or protected, and I received verbal consent from all private land owners and from the groundskeepers of Cornell University. Colonies were collected at night, or during the day with a vacuum over the course of 30 minutes to ensure capture of nearly all adults. Care was taken to avoid losing the queen when collecting the colony. During colony disturbance, queens hide at the top of the nest and will not fly (KJL, unpublished observations). Thus collection with a vacuum does not risk queen evacuation.

Observations of matricide

101 I video-recorded three matricides over the course of four summers (2010-2013). Continuous video monitoring with webcams (Logitech C600) connected to PCs (software: Eyeline v1.18,

NCH Software) witnessed the first matricide in an observation colony on July 10, 2010. This colony was one of five that were established in horizontal style boxes (Akre et al. 1973) (Figure

S4.2a) that constrain new comb construction but allow a view of most colony activity. The matricide colony (number 20_10) was transplanted into the box on July 5, 2010 after daytime collection. One hundred and eleven workers were uniquely marked with paint pens (Sharpie) and the queen with a honey bee queen tag ("Opalithplättchen”). Combs were secured side by side with hot glue, the cold-anesthetized workers and queen returned to the box, with cotton balls soaked in honey and water. The box was positioned near a window where a tube connected the colony to the outdoors, and the tunnel was opened after dark. The colony was filmed from below under continuous lighting.

In subsequent years, colonies were established using a similar protocol, except colonies were transplanted when young, typically with fewer than 20 workers, and workers were not marked. Observation boxes were vertical with two glass walls (Figure S4.2b), and either 1.5 or

2.5 inches wide, depending on the size of the colony when collected. This type of box allows colonies to build additional combs, and most successful transplants eventually constructed reproductive combs, some rearing hundreds of reproductives (KJL unpublished data). Workers were unmarked, queens were marked with paint pens or a thin copper wire wrapped around her waist. These colonies were continuously filmed under infrared lighting, with two cameras per box, one on each side. Matricide was observed in one of five colonies in 2011 and one of eight unmanipulated colonies in 2012 (the queens of an additional 3 observation colonies in 2012 and

10 colonies in 2013 were surgically manipulated in an experimental attempt to induce matricide;

102 Chapter 5). For descriptions of matricide events, see Appendix 4.2. In two colonies, queen death during the reproductive stage was not observed on video due to technical problems, and in one colony the queen died because she slid to the bottom of the box while walking on the glass walls of the lower portion of the nest (shelves were installed in subsequent boxes to prevent this from happening).

Inferring matricide

Matricide is challenging to observe directly, but a careful consideration of yellowjacket wasp biology allows inference of matricide from queenless colonies collected during particular developmental stages. D. arenaria colonies can become queenless for several reasons. First, queens may die while foraging or during usurpation attempts prior to the stage when the first workers begin foraging and the queen no longer leaves the nest (Greene 1991). These queenless young colonies are common: in our population, queenless colonies made up 14 of 43 colonies with fewer than 20 post-emergence workers (this excludes an additional 20 colonies that were parasitized by the inquiline D. arctica). Such colonies can survive and produce males, though because they can rear no more workers, they remain small and do not expand the nest beyond the second or third comb (KJL, personal observation). By collecting colonies with at least four combs, we can avoid collecting queenless colonies with queen loss of this type. Given that matricide should occur after the worker production phase (Bourke 1994), and given that all three of our observed matricides occurred as workers were constructing the fourth comb, we are likely not overlooking many matricide colonies by excluding those with fewer than four combs.

Queens also senesce at the end of the season, when the colony is in decline and no more brood will be reared (Edwards 1980). By collecting active colonies with healthy eggs and

103 larvae, we can avoid collecting colonies that are queenless due to late-season senescence. The queenless and queenright colonies used to infer matricide did not differ in their collection dates, indicating that the queenless colonies used were not merely senescing queenright colonies (see

Colony sampling for genotyping section).

Reproductive-stage queens can die for reasons other than matricide, including disease or intrinsic weakness(Edwards 1980). However, this is likely a rather rare source of queen mortality. Mature-colony queenlessness rates for five Vespula yellowjacket species range from 0

– 11% (Foster and Ratnieks 2001a; Loope et al. 2014), with a pooled queenlessness rate of 5/106

=4 .7%; these species are very similar to Dolichovespula but probably do not have matricide

(Bourke 1994; Foster and Ratnieks 2001a; Loope et al. 2014). In contrast, Dolichovespula species have very high rates of queenlessness in mature colonies (20-86% for six species (Foster and Ratnieks 2001a)), and workers of at least three of these species have been observed stinging the queen to death (D. maculata (Akre and Myhre 1992); D. arenaria, this paper; D. sylvestris,

T. Wenseleers, personal communication). If we assume that intrinsic, non-matricide queen mortality is similar in Vespula and Dolichovespula, then the great majority of queen deaths in the mature, pre-senescence stage are likely due to matricide, and colonies collected in this stage can be used to study this phenomenon. In my collections in 2010, 2012 and 2013, 42% (13/31) of pre-decline, wild-collected colonies with at least 4 combs were queenless. Assuming a 4.7% non-matricide mortality in this stage (based on the Vespula data) implies that 89% of queenless colonies experienced matricide. This imperfect inference necessarily adds noise to the analysis of matricide, but patterns emerging from these colonies should still reflect matricide behavior, particularly if they agree with the findings from observed matricide colonies.

104

Colony sampling for genotyping

Colonies were selected for inclusion in the study based on the year they were collected (I only saved colonies from 2012 and 2013). I do not have genetic material from one of the three observed matricide colonies (20_10). In addition to the two additional observed matricide colonies, I analyzed all mature, pre-senescence colonies with at least 4 combs collected in 2013 and six of nine colonies collected in 2012 (three queenright colonies were randomly omitted). No other colonies were genotyped, and all genotyping and paternity calculations were performed blind to whether the colony was queenless or queenright, with the exception of the two observed matricide colonies that were genotyped. Queenless and queenright colonies did not differ in the

Julian date of collection, further suggesting that queenlessness is not due to colonies entering decline (queenless: 209.3 ± 14.8 SD; queenright: 208.2 ± 7.7; unequal variances t = 0.21, df =

11.75, p = 0.84).

Genotyping

I used standard Chelex DNA extractions, multiplex PCR and fragment length analysis to obtain the microsatellite genotypes of workers from 21 colonies at the loci Rufa05, Rufa13, Rufa15

(Thorén et al. 1995) and List2004 (Daly et al. 2002); these loci are known to be variable in D. arenaria (Freiburger et al. 2004). Protocols were identical to those reported in a previous study on wasp paternity (Loope et al. 2014). Briefly, DNA was extracted from a single antenna by boiling in 100 µL of a 10% Chelex solution. Each 10 µL PCR reaction included 1 µL extracted

DNA, 5 µL Qiagen master mix (Qiagen Type-It Microsatellite Kit, Qiagen Inc.), 0.2 µL of each dye-labeled forward and reverse primer, and water to total 10 µL. PCR reaction conditions were

105 95°C for 15 minutes, 35 cycles of 95°C for 30 seconds, 50°C for 90 seconds, 72°C for 60 seconds, 60°C for 30 minutes. Fragments were analyzed on an ABI-3730 sequencer, and alleles were scored and checked twice by eye using GeneMarker (SoftGenetics LLC). I analyzed approximately 16 adult workers per colony for 19 collected wild colonies, and all adult workers

(37 and 88) from two observation colonies collected immediately after observing matricide

(Table S4.1).

Suitability of microsatellite data for paternity analysis

To make sure loci were suitable for paternity analysis, I created 16 samples of 21 workers, one drawn randomly from each study colony. These 16 data sets were analyzed using Genepop 4.3

(Rousset 2008) to check for linkage, departures from Hardy-Weinberg equilibrium, and null alleles. Linkage analyses indicated loci were in linkage equilibrium (all p>0.05). Genotype data did not depart from Hardy-Weinberg equilibrium: only two probability tests yielded a p value <0.05, and with $ = 0.05 and a total of 16"4 = 64 tests, we expect three tests to be significant just due to chance. Null allele estimation suggested possible rare null alleles in

List2004 (0, 0.032; median, mean null frequency across 16 sampled datasets) and Rufa15

(0.032, 0.030; median, mean null frequency). However, subsequent analyses ensured that our paternity assignments were not influenced by rare null alleles.

To check for the influence of possible rare null alleles at List2004 and Rufa15, I first inspected all colony genotypes by eye. Unusual List04 genotype patterns in one colony (80_13) indicated a possible maternal null allele or sample contamination and were removed and

Colony2 analyses re-run. Genotypes at the remaining three loci in this colony all were consistent in assigning paternity for all 16 workers, further suggesting erroneous List2004 genotypes in this

106 colony. To further check for null paternal alleles in other colonies, I verified that no two

patrilines were inferred as distinct based on a single locus. I also re-ran paternity

assignment analysis omitting all genotypes at the List2004 locus or the Rufa15 locus. The

resulting paternity assignment was identical to the result of the four-locus analysis for all

colonies containing workers homozygous at the locus omitted. Because parental null alleles

generate erroneous paternity assignments due to false homozygosity in offspring (Dakin and

Avise 2004), these checks ensure that paternity assignments were not affected by null

alleles. The estimates of rare null alleles in Rufa15 may instead be due to occasional inbreeding,

which increases observed homozygosity, biasing null allele estimates (Chybicki and Burczyk

2009).

Paternity analysis

Colony2 v2.0.5.8 (Jones and Wang 2010) was used to determine the maximum likelihood

paternity assignment for 427 workers from 21 colonies. I input genotypes for four loci, as well

as a maternal sibship exclusion for all workers not collected from the same colony. Female

polygamy and male monogamy were specified, using a single “long” Full-Likelihood run. Allele

frequencies were updated during the run, and the program used a weak sibship scaling prior of

10 sibs per patriline and 16 sibs per matriline. I ran analyses both with and without inbreeding, and with genotyping error rates of 0.05 and 0.001. All runs gave the same patriline assignments to all workers, and zero estimated genotyping errors for offspring (after removing a possible null allele in one colony; see above).

The Colony2 results suggested that six workers in five colonies (three queenright, two queenless) were from a second matriline. These six individuals were removed from the dataset

107 prior to relatedness calculations. Such individuals are commonly detected in Dolichovespula

colonies (Foster et al. 2001; Bonckaert et al. 2011b) and result either from queen usurpation or drift. I genotyped all adult workers in both observed matricide colonies with genotypic data, and neither contained a second matriline, ruling out the possibility that queens are killed by unrelated workers.

Estimation of D. arenaria population paternity statistics

An estimate of population average effective paternity must account for the fact that the colonies genotyped in this study were not a random sample of colonies: I analyzed a greater proportion of queenless colonies than are found in nature. The natural frequency of mature (% 4 combs), pre- decline, queenless colonies is 42% in my population, from a collection of 31 colonies. Because paternity is correlated with presence of the queen, to estimate the mean paternity frequency in the population, I performed the following correction. First, I included all 10 queenright colonies sampled. Then seven queenless, wild-collected colonies were randomly selected from the nine used in paternity analysis. The result is a set of 17 wild-collected colonies, 41% of which are queenless, matching the population frequency.

Relatedness

I calculated worker-worker pedigree relatedness from estimates of effective paternity (Pamilo

1993) after correcting for sample size using the formula for ke3 (Nielsen et al. 2003). To estimate

inbreeding, I used Relatedness 5.0.8 (Queller and Goodnight 1989) to calculate pairwise

relatedness values of all fathers to their mates, using inferred parental genotypes from Colony2

and setting inferred single-locus genotypes with probability <0.8 as “missing.” The relatedness

108 of a male to his mate is equal to the offspring inbreeding coefficient in haplodiploid species

(Liautard and Sundström 2005). Although individual pairwise relatedness estimates are noisy for analyses with few loci, the estimator used by the Relatedness software is unbiased, and thus still useful for comparisons of inbreeding within a study (Liautard and Sundström 2005).

Statistics

I used the unequal variances (Welch’s) t-test (Ruxton 2006) in the base package of R version

3.0.3. For comparisons with very skewed distributions, values were ranked (averaging ties) before applying the unequal variances t-test (Ruxton 2006).

109 REFERENCES

Akre RD, Hill WB, MacDonald JF (1973) Artificial housing for yellowjacket colonies. J Econ Entomol 66:803–805.

Akre RD, Myhre EA (1992) Nesting biology and behavior of the Baldfaced Hornet, Dolichovespula maculata (L.) (Hymenoptera: Vespidae) in the Pacific Northwest. Melanderia 48:1–33.

Arnqvist G, Nilsson T (2000) The evolution of polyandry: multiple mating and female fitness in insects. Anim Behav 60:145–164.

Baer BC, Schmid-Hempel P (2001) Unexpected consequences of polyandry for parasitism and fitness in the bumblebee, Bombus terrestris. Evolution 55:1639–1643.

Bonckaert W, Tofilski A, Nascimento FS, et al (2011a) Co-occurrence of three types of egg policing in the Norwegian wasp Dolichovespula norwegica. Behav Ecol Sociobiol 65:633– 640.

Bonckaert W, Van Zweden JS, D'Ettorre P, et al (2011b) Colony stage and not facultative policing explains pattern of worker reproduction in the Saxon wasp. Mol Ecol 20:3455–3468.

Boomsma JJ, Nielsen J, Sundström L, et al (2003) Informational constraints on optimal sex allocation in ants. Proc Natl Acad Sci USA 100:8799–8804.

Bourke AFG (1994) Worker matricide in social bees and wasps. J Theor Biol 167:283–292.

Bretman A, Wedell N, Tregenza T (2004) Molecular evidence of post-copulatory inbreeding avoidance in the field cricket Gryllus bimaculatus. Proc R Soc Lond B 271:159–164.

Chybicki IJ, Burczyk J (2009) Simultaneous estimation of null alleles and inbreeding coefficients. J Hered 100:106–113.

Crozier RH, Fjerdingstad E (2001) Polyandry in social Hymenoptera - disunity in diversity? Ann Zool Fenn 38:267–285.

Dakin EE, Avise JC (2004) Microsatellite null alleles in parentage analysis. Heredity 93:504– 509.

Daly D, Archer ME, Watts PC, et al (2002) Polymorphic microsatellite loci for eusocial wasps (Hymenoptera: Vespidae). Mol Ecol Notes 2:273–275.

Edwards R (1980) Social wasps: their biology and control. Rentokil, London

Foster KR, Ratnieks FLW (2001a) Paternity, reproduction and conflict in vespine wasps: a model system for testing kin selection predictions. Behav Ecol Sociobiol 50:1–8.

Foster KR, Ratnieks FLW (2000) Facultative worker policing in a wasp. Nature 407:692–693.

110 Foster KR, Ratnieks FLW (2001b) Convergent evolution of worker policing by egg eating in the honeybee and common wasp. Proc R Soc Lond B 268:169–174.

Foster KR, Ratnieks FLW, Gyllenstrand N, Thoren P (2001) Colony kin structure and male production in Dolichovespula wasps. Mol Ecol 10:1003–1010.

Freiburger B, Breed M, Metcalf JL (2004) Mating frequency, within!colony relatedness and male production in a yellow jacket wasp, Dolichovespula arenaria. Mol Ecol 13:3703–3707.

Greene A (1991) Dolichovespula and Vespula. In: Ross K, Matthews RW (eds) The Social Biology of Wasps. Cornell University Press, Ithaca, NY, pp 263–305

Greene A, Akre RD, Landolt PJ (1976) The aerial yellowjacket, Dolichovespula arenaria (Fab): nesting biology, reproductive production, and behavior (Hymenoptera: Vespidae). Melanderia 26:1–34.

Hamilton WD (1972) Altruism and related phenomena, mainly in social insects. Annu Rev Ecol Syst 3:193–232.

Hammond RL, Bruford M, Bourke AFG (2003) Male parentage does not vary with colony kin structure in a multiple-queen ant. J Evol Biol 16:446–455.

Jones OR, Wang J (2010) COLONY: a program for parentage and sibship inference from multilocus genotype data. Mol Ecol Resour 10:551–555.

Kovacs JL, Hoffman EA, Goodisman MAD (2008) Mating success in the polyandrous social wasp Vespula maculifrons. Ethology 114:340–350.

Liautard C, Sundström L (2005) Estimation of individual level of inbreeding using relatedness measures in haplodiploids. Insectes Soc 52:323–326.

Loope KJ, Chien C, Juhl M (2014) Colony size is linked to paternity frequency and paternity skew in yellowjacket wasps and hornets. BMC Evol Biol 14:277.

Mattila HR, Reeve HK, Smith ML (2012) Promiscuous honey bee queens increase colony productivity by suppressing worker selfishness. Curr Biol 22:2027–2031.

Meunier J, West SA, Chapuisat M (2008) Split sex ratios in the social Hymenoptera: a meta- analysis. Behav Ecol 19:382–390.

Nielsen R, Tarpy DR, Reeve HK (2003) Estimating effective paternity number in social insects and the effective number of alleles in a population. Mol Ecol 12:3157–3164.

Nonacs P (2000) Measuring and using skew in the study of social behavior and evolution. Am Nat 156:577–589.

Nowak MA, Tarnita CE, Wilson EO (2010) The evolution of eusociality. Nature 466:1057–1062.

111 Pamilo P (1993) Polyandry and allele frequency differences between the sexes in the ant Formica aquilonia. Heredity 70:472–480.

Queller DC, Goodnight KF (1989) Estimating relatedness using genetic markers. Evolution 43:258–275.

Ratnieks FLW, Foster KR, Wenseleers T (2006) Conflict resolution in insect societies. Annu Rev Entomol 51:581–608.

Rousset F (2008) genepop’007: a complete re-implementation of the genepop software for Windows and Linux. Mol Ecol Resour 8:103–106.

Rousset F, Roze D (2007) Constraints on the origin and maintenance of genetic kin recognition. Evolution 61:2320–2330.

Ruxton GD (2006) The unequal variance t-test is an underused alternative to Student's t-test and the Mann–Whitney U test. Behav Ecol 17:688–690.

Strassmann JE, Nguyen J, Arevalo E, et al (2003) Worker interests and male production in Polistes gallicus, a Mediterranean social wasp. J Evol Biol 16:254–259.

Sundström L (1994) Sex ratio bias, relatedness asymmetry and queen mating frequency in ants. Nature 367:266–268.

Sundström L, Chapuisat M, Keller L (1996) Conditional manipulation of sex ratios by ant workers: a test of kin selection theory. Science 274:993–995.

Thorén PA, Paxton RJ, Estoup A (1995) Unusually high frequency of (CT)n and (GT)n microsatellite loci in a yellowjacket wasp, Vespula rufa (L.) (Hymenoptera: Vespidae). Insect Mol Biol 4:141–148.

Trivers R, Hare H (1976) Haplodiploidy and the evolution of the social insects. Science 191:249–263.

Wenseleers T, Ratnieks FLW (2006) Comparative analysis of worker reproduction and policing in eusocial hymenoptera supports relatedness theory. Am Nat 168:E163–E179.

Wenseleers T, Tofilski A, Ratnieks FLW (2005) Queen and worker policing in the tree wasp Dolichovespula sylvestris. Behav Ecol Sociobiol 58:80–86.

Zanette LRS, Miller SDL, Faria CMA, et al (2012) Reproductive conflict in bumblebees and the evolution of worker policing. Evolution 66:3765–3777.

112 Appendix 4.1. Supplemental figures and tables

Table S4.1. Detailed colony data. This table presents paternity data for the 21 colonies included in the study. Details are explained in notes below. nd Collection Queen 1 2 patrilines 4 5 6 7 8 9 10 11 Colony n 2 3 B k k k r W G M date present? matriline ratio e e3 7/15/11 09_11 observed matricide 88 0 - - 1 1.00 1.00 0.75 88 0 0

7/17/12 25_12 observed matricide 37 0 36:1 0.43 2 1.06 1.06 0.72 37 0 0

7/5/12 47_12 N 16 0 - - 1 1.00 1.00 0.75 289 0 0

7/21/12 68_12 N 15 0 - - 1 1.00 1.00 0.75 210 0 0

7/16/13 65_13 N 16 0 - - 1 1.00 1.00 0.75 424 0 0

8/4/13 83_13 N 16 0 - - 1 1.00 1.00 0.75 242 4 0

8/5/13 85_13 N 16 0 - - 1 1.00 1.00 0.75 169 1 22

8/25/13 88_13 N 15 0 14:1 0.34 2 1.14 1.15 0.69 97 0 79

8/6/13 86_13 N 16 1 13:2 0.24 2 1.30 1.32 0.63 156 4 181

7/29/13 81_13 N 16 0 12:4 0.09 2 1.60 1.66 0.55 300 11 31

7/18/13 69_13 N 16 2 8:5:1 0.08 3 2.18 2.36 0.46 99 0 38

7/29/13 82_13 Y 16 1 - - 1 1.00 1.00 0.75 208 0 0

7/27/12 70_12 Y 16 0 11:5 0.04 2 1.75 1.83 0.52 219 63 27

8/4/13 84_13 Y 16 0 11:4:1 0.16 3 1.86 1.95 0.51 702 0 343

8/13/13 87_13 Y 16 1 10:5 0.02 2 1.80 1.89 0.51 318 186 11

7/15/12 60_12 Y 16 0 10:6 0 2 1.88 1.98 0.50 260 21 0

7/24/12 69_12 Y 16 1 9:6 -0.01 2 1.92 2.04 0.50 200 0 0

7/21/12 67_12 Y 16 0 9:7 -0.02 2 1.97 2.09 0.49 322 0 0

7/24/13 75_13 Y 16 0 9:7 -0.02 2 1.97 2.09 0.49 287 0 0

7/24/13 76_13 Y 16 0 9:7 -0.02 2 1.97 2.09 0.49 389 27 0

7/25/13 80_13 Y 16 0 9:7 -0.02 2 1.97 2.09 0.49 401 42 0 Notes: 1the number of genotyped workers used in paternity analysis 2the number of workers in the colony sample identified as members of a second matriline 3the ratio of workers assigned to the patrilines in the majority matriline 4Nonac’s B, an index of skew (Nonacs 2000) 5The number of detected queen mates (paternity frequency). The population mean is 1.88, with 71% of colonies having multiple patrilines, from 17 colonies after correcting for the types of colonies sampled (see Experimental Procedures) 6The uncorrected estimate of effective paternity (Starr 1984) 7Estimate of effective paternity correcting for sample size (Nielsen et al. 2003). The population harmonic mean is 1.49 from 17 colonies after correcting for the types of colonies sampled (see Experimental Procedures) 8 Worker-worker relatedness calculated from effective paternity (ke3) 9The number of adult workers collected 10The number of adult gynes (new queens) collected 11The number of adult males collected

113

"#( 2/:./?/<*.+>0123,4,@A $#"" 1)B"#"( "#' )B"#$ )C"#$ "#D( "#&

"#% "#("

"#$ )*+,-./+012;*-, )*+,-./+0123,41567 "#%( "

!"#$ "#"" 89,,.-/:;+ 89,,.>,22 "#% "#" "#% "#' <=>=./,2 <=>=./,2 ! " E*>,!89,,.1-,>*+,@.,22

Figure S4.1. Paternity skew, queen loss and inbreeding. A. Paternity skew (Nonacs 2000) is higher in multiple-patriline queenless colonies than queenright colonies (see statistical comparison in main text). The probability of obtaining the observed skew values due to chance if the true paternity were evenly divided among males (p) was less than 0.1 for all five queenless colonies but only one of nine queenright colonies. B. Points represent males (n = 19) mated to queens from multiple-patriline queenright colonies (n=9; each color represents the mates of a single queen). Unlike in queenless colonies (Figure 4.2b), paternity share is not linked to inbreeding in males mated to queens from queenright colonies.

114

Figure S4.2. Nest boxes used for matricide observation. A. This style of box was used for observation colonies in 2010. Combs were glued side by side. The entrance tunnel connecting the box to the outdoors is visible at top. The photo is taken upward from below through the glass bottom of the box; the cells open downward. This is colony 20_10 several days after matricide. B. In 2011 and 2012, the vertical observation box was used. It measures 1.5 inches wide and allows colonies to add combs and expand the nest. This colony was installed when it had only 10 post-emergence workers and only a few cells present on comb 2, the subsequent nest expansion occurred inside the nest box. Visible in the photo are two worker combs and two reproductive combs. The cameras and infrared lights are visible, and the tunnel connecting the nestbox to the outdoors is at lower right.

115 Appendix 4.2. A description of observed matricides

I video-recorded three matricides in observation colonies, one each year in 2010, 2011 and 2012.

The timing of matricide in colony development

All three matricides occurred as the colony was initiating or expanding the fourth comb, and in all three colonies the first three combs consisted entirely of worker cells. As no colonies in this population have been observed with more than three combs of worker cells, the timing of matricide coincides with the start of the reproductive phase. However, in only two colonies

(20_10 and 09_11) were the colonies vigorous enough to continue into the reproductive phase.

Matricide in colony 20_10 occurred on July 10, 2010, five days after it was installed in an observation box. The colony was transplanted with 111 workers and the queen, and three combs of worker brood and one tiny comb with eggs of an undetermined sex. This colony was not collected immediately after matricide, but was instead allowed to continue to grow. This colony produced two new combs of male cells, though it reared out no adult males from these cells.

This style of observation box limits colony expansion, perhaps due to unnaturally heavy investment in envelope reconstruction, and thus this failure could have been a result of the transplant.

Colony 09_11 was in a stage remarkably similar to 20_10 when matricide occurred on July 15,

2011, with workers constructing a fourth comb (the first reproductive comb). Genotyping the 1st instar larvae and eggs from the edge of this comb indicated they were all diploid; therefore, the queen was laying female brood (i.e. at the start of a queen production phase) when she was killed

116 (Chapter 5).

At the time of matricide on July 17, 2012, Colony 25_12 had already entered decline, typified by many empty cells, few, discolored larvae and little foraging. This colony had only 37 workers when collected immediately after matricide, and would have reared no more brood to adulthood if it had not been collected. However, numerous cells contained fresh eggs and the queen’s ovaries were still active with numerous mature oocytes in the ovarioles and approximately

27,000 sperm in the spermatheca (Chapter 5), suggesting the early decline was not a result of a failing queen.

The queen killing act

In all three colonies in which matricide was observed, a single worker mounted the queen and began stinging her. In two cases, she was stung between the segments in a membrane on the dorsal side; once, in 09_11, she was likely stung in the thorax, though the video angle prohibited a clear view. Once the queen was stung, other workers became clearly agitated and, in colonies

09_11 and 25_12, were attracted to the struggling queen and worker, joining in stinging and biting the queen. In 20_10, workers rushed toward the lamp that illuminated the nest, suggesting that they were responding to alarm pheromone which, during daylight, likely stimulates departure from the nest. This did not occur in the 2011 and 2012 colonies, as they were illuminated with infrared light not visible to wasps, and the events occurred at night. This immediate response to the matricide event could be explained simply as a general response to alarm pheromone release; alternatively, as has been suggested based on similar observations in D. maculata, it could be behavior that is specific to the context of matricide (Akre and Myhre 1992).

117 The queen in all cases seemed ineffectual at defending herself. In 20_10 she appeared to be

stung before she could respond to the attack; in the other two colonies she attempted to sting her

attacker, but only the queen was seen to be killed in the struggle. However, given the observed

attraction and stinging by other workers, a matricidal worker may be at risk of death in the

subsequent frenzy.

Behavior prior to matricide

In colony 20_10, most workers were individually marked. In 20 hours of observation

immediately prior to matricide, three workers were observed laying eggs (the queen was also

observed laying several eggs, as well as eating a worker-laid egg). One of these reproductive

individuals, “White-Pink”, was the queen killer, consistent with the theoretical prediction that

matricidal workers should be reproductive, since they are most related to the resulting worker- laid male brood (Bourke 1994). Given that only three of approximately 100 adult workers were reproductive, the probability that the matricidal worker was a reproductive worker due to chance is quite low (p ~ 0.03; Fisher’s Exact Test). The matricidal worker was not seen in 5 hours of observation on the 5th day after matricide, suggesting there is high turnover in reproductive workers, as has also been observed in D. norwegica (Bonckaert et al. 2011a), and further suggesting that in this species, the benefit of matricide to the matricidal worker likely comes from replacing brothers with nephews, rather than her sons.

118 CHAPTER 5

MATRICIDE, QUEEN SEX ALLOCATION, AND QUEEN FECUNDITY IN A

YELLOWJACKET WASP

Kevin J. Loope

Abstract

In many colonies of social insects, the workers compete with each other and with the queen over the production of the colony’s males. This intra-colony conflict even results in matricide—the killing of the colony’s irreplaceable queen by a daughter worker—in some species of social bees and wasps with annual colonies. In colonies with low effective paternity and high worker-worker relatedness, workers value worker-laid males more than queen-laid males, and thus may benefit from queen killing. Workers gain by eliminating the queen because she is a competing source of male eggs and actively inhibits worker reproduction through policing. However, matricide may be costly to workers if it reduces the production of valuable new queens and workers. Here, I test two theoretical predictions about the timing of matricide in a wasp, Dolichovespula arenaria, recently shown to have facultative matricide based on intra-colony relatedness. First, using analyses of collected, mature colonies and a surgical manipulation preventing queens from laying female eggs, I show that workers do not preferentially kill queens who are only producing male eggs. Instead, workers sometimes kill queens laying valuable females, suggesting a high cost of matricide. Second, observations of colonies containing queens with experimentally- induced low fecundity suggest that declining queen fecundity does not trigger matricide.

Overall, although matricide is common and typically occurs only in low paternity colonies, it seems that workers pay substantial costs in this expression of conflict over male parentage.

119 Introduction

Evolutionary conflict occurs when individuals differ in their optimal outcomes for the

same event (Hamilton 1972; Trivers and Hare 1976; Ratnieks and Reeve 1992). How this

conflict plays out will depend primarily on two things: the different optima of the individuals in

conflict, and the relative power the individuals have to influence the outcome of conflict given

the biological details of the situation (Ratnieks and Reeve 1992; Beekman and Ratnieks 2003).

Reproductive conflict has been studied intensively in the eusocial Hymenoptera, particularly in

the contexts of sex allocation and male parentage (reviewed in (Queller and Strassmann 1998;

Bourke 2005; Ratnieks et al. 2006; West 2009). Conflict over male parentage in single-queen colonies arises because each female (queens and workers) prefers that the colony rears her sons instead of the sons of others. This is because the regression relatedness between a mother and her son is 1, while workers are related to the queen’s sons by 1/2, and to the average other workers’ sons by 1/4 - 3/4, depending on the relative frequency of full- and half-sibling workers.

In single-queen colonies, worker-worker relatedness is determined by the colony’s effective paternity, a measure that accounts for the number of fathers and their relative shares of paternity among workers (Boomsma and Ratnieks 1996; Jaffé 2014). When the colony has an effective paternity less than 2.0, worker-worker relatedness is relatively high, and workers prefer the sons of other workers to the sons of the queen (Starr 1984; Woyciechowski and Lomnicki 1987;

Ratnieks 1988). Workers can influence the outcome of this conflict by eating eggs of other workers or of the queen (e.g., (Ratnieks and Visscher 1989; D'Ettorre et al. 2004; Wenseleers et al. 2005; Zanette et al. 2012), by aggressing or dominating other workers (e.g., (Liebig et al.

1999; Wenseleers et al. 2005; Stroeymeyt et al. 2007), and by killing reproductive competitors, both workers (KJL unpublished observations) and the queen (Bourke 1994).

120 Matricide, the killing of the mother queen by workers, is predicted only in eusocial insect species with annual colonies, given the high cost of killing a queen if she is likely to live for many years (Bourke 1994). In most species with annual colonies, workers cannot produce a replacement queen, making the transition to queenlessness irreversible. Workers have been observed killing their queen in at least six species of vespine wasp, (Chapter 4; Bourke 1994; T.

Wenseleers, personal communication), and at least three species of bumble bee (summarized in

Bourke 1994). Here, I test theoretical predictions regarding the evolution of matricide using a common North American yellowjacket wasp, Dolichovespula arenaria (Vespidae: Vespinae).

A kin-selection model of matricide makes several predictions about when matricide should be most favorable to workers (Bourke 1994). First, when an individual worker is unlikely to dominate male production after the queen is dead, matricide should be favored only in colonies with high worker-worker relatedness, because there is a relatedness benefit to the worker of replacing brothers with nephews when producing male reproductives. This prediction is upheld in my study population of D. arenaria; most mature, queenright colonies have effective paternities near 2.0 while most queenless colonies (in which matricide was observed or inferred) have effective paternities near 1.0 (Chapter 4). In the present study, I test two additional predictions of Bourke’s model. The first is that workers should preferentially kill queens who have irreversibly switched to laying male eggs (Trivers and Hare 1976; van der

Blom 1986; Bourke 1994). This prediction stems from the fact that workers value the laying of female eggs by the queen: workers are highly related to new queens, and they benefit from increased colony productivity resulting from the addition of new workers. If the queen dies while producing females, workers thus suffer a inclusive fitness cost. But if the queen is only producing male eggs, then queen-killing involves no cost of lost female production; the colony

121 simply switches from producing brothers to producing sons and nephews, which are more valuable if the colony’s effective paternity is low. The second prediction I test is that workers should preferentially kill queens with low fecundity. This prediction is based on the assumption that such queens constrain colony productivity by their low rate of egg laying, and that the colony will produce more total offspring once she is dead and worker oviposition is uninhibited

(Bourke 1994).

The best support for these predictions to date come from Bombus terrestris, a bumblebee in which worker-queen conflict has been well-studied. Several authors report matricide in colonies whose queens have switched to laying exclusively haploid male eggs, and matricidal workers are reproductively active (van Honk et al. 1981; van Doorn and Heringa 1986), supporting predictions of the kin-selected matricide hypothesis (Bourke 1994). However, unlike

B. terrestris, most vespine wasps have a relatively long period of producing reproductives, and workers are often produced throughout this reproductive stage (Greene 1984). In most species, male and female reproductives are produced simultaneously, suggesting that a matricidal worker in a typical colony would incur a substantial cost by prematurely ending female production. For these reasons, some have argued against the likelihood of kin-selected matricide in vespine wasps (Martin et al. 2009), though this logic does not explain the observations of matricide, and it does not consider the possibility of facultative matricide in response to colony-specific conditions.

There is suggestive evidence for facultative matricide based on the sex of the queen-laid brood in yellowjackets. Montagner ((Montagner 1966) found, after radiolabelling queens to determine whether queens or workers produce males, that all three queens that produced only male eggs after reintroduction to the colony were killed by workers, while all seven queens that

122 laid only female eggs were not killed. This appears to be facultative matricide in response to queen primary sex ratio, but the methods employed make conclusive interpretation difficult. The male-laying queens may have been more damaged by the radiation than the others, and then died or were killed due to their poor condition. Additionally, Montagner provides no description of the queen killing event, or how this was observed. Also, he performed these experiments using colonies of several species, two of which (Vespula germanica and Vespula vulgaris) are in a genus with no other reports of matricide (reviewed in Bourke 1994). Thus, although these findings are suggestive, further investigation of the link between the sex of the queen-laid brood and matricide is clearly needed.

To test the brood sex and queen fecundity predictions for matricide, I described comb allocation patterns in collected queenless and queenright colonies to determine the context of matricide, and then I used genetic assessments of observed- and inferred-matricide colonies to determine the sex of the last queen-laid brood. I also performed a manipulative experiment similar to Montagner’s: some queens were forced to lay only male eggs by surgically removing their spermathecae, and others had surgically reduced fecundity. Colonies were then observed to document matricide. The results suggest that neither the laying of male eggs by queens nor the onset of low-fecundity in queens is a trigger for matricide in Dolichovespula arenaria.

Methods

Overview

This study used two types of colonies: collected, mature colonies, and observation colonies. The former were collected and killed, whereas the latter were left alive and transplanted into observation boxes. I located these colonies in Tompkins Co., NY using fliers,

123 advertisements on craigslist.org, and visual searching. All colonies were collected with verbal consent from landowners on private property.

From the collected colonies, I measured the surface area of different cell types within each nest. These data revealed how queenright and queenless, inferred-matricide colonies allocated resources among worker, queen and male cells. I also collected genetic data from male pupae from five inferred-matricide colonies to determine what sex the queen was producing immediately prior to her death.

From the observation colonies, I collected genetic data on the last queen-laid brood in one colony in which I observed matricide, the developmental stage (in terms of comb built) of three colonies at the time of the observed matricide, and the number of sperm in one observed- matricide queen. Finally, I surgically induced male-laying in some queens, and surgically reduced other queens’ fecundity. These colonies were monitored to see if these manipulations induced matricide.

Study species

Dolichovespula arenaria is an aerial nesting yellowjacket abundant throughout North

America (Greene et al. 1976; Figure 5.1a; Akre et al. 1980). In my study population in central

New York State, solitary Dolichovespula arenaria queens found new colonies in May and June.

The first workers emerge in mid June, and colonies grow rapidly, building 1-3 combs of worker- producing small cells before building 1-3 combs of larger, reproductive cells for rearing new queens and males (Greene et al. 1976; this study). Colonies can contain over 800 workers (KJL unpublished data), though the average peak colony size is around 300 (Chapter 4). Typically, colonies are in decline by mid August, and only the new queens survive the winter after mating

124

Figure 5.1. Investigating matricide in Dolichovespula arenaria. a. A very large, mature colony of D. arenaria. b. The third comb from colony 65_12 illustrates the transition from female production to male production. The 14 central, large queen pupae (marked “Q”) are surrounded by the shorter first cohort of males (marked “M”). c. A queen ready for spermatheca removal surgery, immediately before the incision is made in the exposed intersegmental membrane. CO2 flowing through the holding tube lightly anesthetizes her. d. A nest box permitting longitudinal observations of colony development. Wooden clips hold infrared LEDs and a webcam is visible on each side. The tube at lower left connects the box to the outdoors, allowing foragers to come and go freely.

125 in late summer and autumn. Worker paternity estimates suggest that queens mate 1-3 times, and that the harmonic mean effective paternity within colonies is 1.49, with 70% of queens (12 of 17) mating more than once (Chapter 4). Notably, this is substantially more polyandry than was found in previous estimates of paternity in D. arenaria populations from the western United

States (Foster and Ratnieks 2001; Freiburger et al. 2004).

Inferring matricide from collected colonies

I collected mature colonies in 2010, 2012 and 2013, and used a subset to infer matricide

(for a full justification, see Chapter 4). The subset included colonies with at least 4 combs but were not senescing, with many empty cells and shriveled brood. These colonies were are all in the reproductive stage, and did not include colonies that were queenless and producing males due to queen death in the pre-emergence period when queens forage away from the colony, since such colonies typically do not grow beyond the addition of a third comb (KJL, personal observation). Given the high rate of queenlessness in these mature colonies (13/31 = 42% in this population; Chapter 4), and the low rates in mature colonies of very similar wasps that lack matricide (4.7% queenlessness rate in Vespula spp.; (Foster and Ratnieks 2001; Loope et al.

2014), I inferred that ~89% of these mature queenless colonies were likely queenless due to matricide. These colonies were compared to colonies in the same stage that retained their queens to understand the factors leading to, and the consequences of, matricide.

Genetic analyses of collected colonies

I used standard Chelex extractions, multiplex PCR and fragment length analysis to obtain the microsatellite genotypes of workers from 21 colonies at the loci Rufa05, Rufa13, Rufa15

126 (Thorén et al. 1995) and List2004 (Daly et al. 2002). Protocols were identical to those reported in a previous study on wasp paternity (Loope et al. 2014). The genotypes of queens and their mates were known from a companion study of worker paternity in these colonies (Chapter 4).

I genotyped 18-24 male pupae from each of the five collected colonies that still contained pupae spanning the transition from queen cells to male cells (Figure 5.1b). The first males laid after the switch to male production can indicate what the queen was laying prior to her death. If the queen was laying females when she died, then the first males laid should all be the sons of workers. In this scenario, the transition from queen production to male production is the result of queen death. Alternatively, the transition from queen production to male production could be the result of the queen switching to laying male eggs. If this is the case, then we expect most or all of the first males laid to be sons of the queen. I inferred parentage of these first males by examining how many of them shared an allele with the queen at all four loci. Given that workers have one queen allele at each locus, there is a chance that a worker-laid male shares alleles with the queen at all four loci. The probability that this happens is (1/2)k where k is the number of informative loci (loci at which the male mate of the queen does not share an allele with the queen). A binomial test can then determine the probability of finding the observed number of all-queen-allele males under the null hypothesis that all males are the sons of workers. If this probability is low, we can infer that the queen laid some male eggs before dying.

In the one thriving observed matricide colony for which I have genetic material (09_11),

I genotyped 24 first-instar larvae and eggs from the edge of the incipient fourth comb (the first reproductive comb) to determine what the queen was laying when she was killed. These brood were certainly almost all queen offspring, since the colony was collected 12 hours after the queen

127 was killed. Determining whether these brood were male or female revealed what the queen was laying when she died.

Establishing observation colonies

Colonies were transplanted to observation boxes when young, typically with fewer than

30 workers and 2-3 combs. These observation boxes (Figure 5.1d) were established in outbuildings at the Liddell Field Station in Ithaca, NY, and the colonies in them were allowed to forage freely outdoors. For details of observation box setup, see Appendix 5.1.

Sperm counting

To determine whether the observed-matricide queen in 2012 had run out of sperm, I counted the stored sperm in her spermatheca, as well as those of five spring-caught queens, and four mid-season, reproductive-stage queens. I prepared a sperm-counting solution that was 1 part 0.05% Triton X-100 (Sigma Aldrich) diluted in 19 parts Ringer’s solution (Kronauer and

Boomsma 2007). Otherwise, I used the method of (Stein and Fell 1996). Each spermatheca was dissected in distilled water, transferred with forceps to an Eppendorf tube containing 20 µL of sperm-counting solution, then crushed thoroughly with a plastic pestle. The pestle was rinsed into the tube using four 20 µL aliquots of solution, resulting in a final volume of 100 µL of sperm in suspension. After mixing by thorough pipetting, the two sides of a Bright-Line

Haemocytometer were immediately filled with sperm in dilution. I allowed the cells to settle before counting the number of sperm in four 100 nL squares on each side, for a total of 8 counts. The mean of these counts was multiplied by 1,000 to estimate the number of sperm in the total 100 µL volume.

128

Surgical manipulation of queens

The goal of surgery was to induce matricide by creating queens that 1) only lay male eggs

(spermathecectomy), or 2) have reduced fecundity (ovariectomy).

Spermathecectomy – Because the wasps studied have haplodiploid sex determination, sperm is

required to produce female offspring, but not male offspring. If the stored sperm are removed,

only males can be produced (Koeniger 1970). I created queens who cannot lay male eggs by

surgically removing the spermatheca. Briefly, queens were extracted from colonies constructing

their third or fourth comb (i.e. early in the reproductive stage), and anesthetized with CO2. Their spermathecae were surgically removed (Figure 5.1c) and they were returned to the nest within a few minutes of surgery. Sham surgery queens experienced the same procedure, but the spermatheca was only touched with forceps instead of being pulled out. For a detailed description of surgical methods, see Appendix 5.1.

Ovariectomy – To experimentally reduce queen fecundity, I surgically removed part or all of the oocytes from the ovaries of live queens (for a similar manipulation, see (Strambi 1965). This procedure mirrored those described above for spermathecectomies. Briefly, queens were anesthetized with CO2, then held in place with pins, dorsal side up, in a wax dish while CO2

flowed into the dish. Ovaries were pulled out through a slit in the dorsal intersegmental

membrane. For a detailed description of surgical methods, see Appendix 5.1.

Three of four queens survived the spermatheca removal surgery in 2012. Five of eight

queens survived in 2013. The rest died within three days of surgery (video observations

129 confirmed that these deaths were not due to matricide, and were almost certainly caused by damage during surgery). Five of six sham surgery queens survived the operation in 2013.

Surgery-related mortality in ovariectomy queens was higher: only two of six queens survived for more than two days after the operation.

Monitoring colonies for matricide

After manipulation, colonies were checked daily for queen presence. During these daily checks, I opportunistically noted queen oviposition behavior. Continuous video recording ensured that I could determine the cause of queen death once it was discovered in daily checks.

Recordings were made using two Logitech C600 webcams per colony (one on each side). Video from each camera was recorded to hard drives using Eyeline 1.18 (NCH Software). I removed the infrared filters on each camera, and illuminated the colonies with infrared LEDs (Figure

5.1d). Spermathecectomy colonies were killed and examined approximately 30 days after surgery in 2012, and 21 days after surgery in 2013. The egg stage is typically 4-5 days (Edwards

1980), and at this stage in colony development, pupae are often capped within 11-18 days of laying (KJL, unpublished data). Thus, virtually all queen-derived eggs and larvae at the time of collection would be male.

Both surviving ovariectomy queens died by falling to the bottom of the box prior to the

21st day after surgery, when colonies were killed and examined.

Results

Comb area and sex allocation

130 The positions of cell types in all colonies suggest that cells were constructed in the order:

worker, queen, male, as previously observed in this species (Greene et al. 1976). As also noted

by Greene et al. (1976), many colonies specialized on the production of one sex of reproductives

(Figure 5.2). Eight of ten queenright colonies produced only queen-destined reproductive cells,

while two produced mostly male cells (Figure 5.2). Queenless colonies were typically male

specialists: they produced a greater area of male cells than queen cells (mean ± SD queen cell

2 2 area: 33.3 ± 28.3 cm ; male cell area: 170.1 ± 97.2 cm ; paired t-test: t = 4.18, df = 8, p = 0.003), and invested an average of 20.6% ± 26.6SD of reproductive comb area in queen cells.

The three observed matricide colonies had each constructed three combs of worker cells at the time of matricide, and two were starting construction of the first reproductive comb (see

Chapter 4 for further description). The area of worker cells in these nests was not different from collected queenless colonies with inferred matricide (observed matricide: 117.9 ± 289.0 cm2,

inferred matricide: 125.5 ± 42.7 cm2, Welch’s t = 0.34, df = 5.25, p= 0.76) but was significantly smaller than the area of worker cells constructed in queenright mature colonies (queenright:

178.5 ± 50.4 cm2; Welch’s t = 2.62, df = 6.15, p = 0.039). Similarly, the area of worker cells constructed in collected queenless colonies was less than that in queenright colonies (t = 2.49, df

= 16.95, p = 0.024). However, queenless and queenright collected colonies did not significantly differ in the mean total comb area (queenless: 328.9 ± 117.3 cm2; queenright: 463.5 ± 221.5 cm2;

Welch’s t = 1.68, df = 13.96, p = 0.12).

131

4$+#*+&0*..1& @==& 3)**"&0*..1& 2-.*&0*..1& !"#"$%"&'()**"&$+&,-.*/&0*..1& /

7 ?==&

>==& ! 5$,6&-+*-&'0, * 7==& !

=& C& 0$.$"IJ& 7ABC7&=DBC 7=BC=&?DBCE&@EBCE&?@BC7&@CBCE&@?BCE&@ABCE&>FBC7&@@BCE&?ABCE&?DBC7&@7BCE&FABCE&F?BCE&F=BC7&?FBC7&?=BC7&@=BCE&@FBCE&@>BCE& G61*+H*9 5$..*08*9 5$..*08*9 ,-8+:0:9* ()**".*11 ()**"+:;<8

Figure 5.2. Colony comb allocation and the timing of queen death. The relative area of combs with different cell types reflects how colonies invest in the production of new males and queens. Colonies often rear males in worker cells as well, making this index an approximate and queen- biased representation of overall sex allocation. Arrows indicate the inferred matricide timing in colony development, in terms of cell construction (Table 5.1). Green/light-grey arrows indicate queens known to be laying females at time of death (determined by genotyping brood; see Methods). Blue/dark-grey arrows indicate queens who had switched to male production before dying (the question mark indicates that the timing of matricide within the male phase is unknown). Black arrows indicate observed matricides for which the sex of the queen’s last brood was unknown. Colony 82_13, marked with an asterisk, was the only queenright colony with single paternity (Chapter 4). It was also the only queenright colony to have barely initiated the reproductive stage, suggesting it may yet have experienced matricide, were it not collected.

132

Table 5.1. Genotypes reveal sex of the queen-laid brood at the time of matricide # brood # # informative # observed males c Implication: Colony p genotyped males locib with all queen alleles queen last laid… 65_13 20 20 3 20 0.00 males 81_13 24 24 3 24 0.00 males 47_12 19 19 3 5 0.079 females?d 68_12 20 20 2 3 0.91 females 83_13 18 18 3 3 0.39 females 09_11a 24 0 - - - females amatricide was observed directly in this colony binformative loci are those at which the male mated to the queen has an allele different from both queen alleles, allowing detection of worker-laid brood at that locus c The probability of observing at least this many males with all queen alleles, assuming all workers are workers’ sons dThis result is ambiguous; p is not low enough to conclusively reject the hypothesis that males are all workers’ sons, though it is possible that a few of the five males with all queen alleles are queens’ sons.

133 Genetic analyses of collected colonies

Genotyping the brood of six colonies (five collected queenless, one observed matricide) shed light on the sex of the queenlaid brood when the queen died (Table 5.1). In two collected colonies, all males sampled from the first cohort of males possessed alleles in common with the queen at all loci. This demonstrates that the queen was alive when the switch to male production occurred, since only ~3 such males were expected if the workers had produced all of these males.

In three other collected colonies, most males possessed at least one allele not shared with the queen, indicating they were the sons of workers. In two of the colonies, the number of all-queen- allele males was close to or less than the number predicted if workers produced all of these males

(colonies 83_13 and 68_12), suggesting that the queen died before male production began. In one colony, 47_12, 5 males had all-queen alleles, while only 2.38 were expected if workers laid all of these males; the probability of observing 5 or more such males if they were all worker’s sons is 0.079 (Table 5.1). The result is therefore ambiguous: it is possible that the queen laid a few male eggs before dying, or alternatively, the workers may have produced all of the sampled males and the excessive number of queen alleles may be due to chance.

In the observed matricide colony 09_11, all genotyped young brood from the fourth comb were diploid, indicating that the colony had started a queen-production phase and the queen was killed while laying females (Table 5.1).

Spermatheca contents in observed matricide queen 25_12

The killed queen from colony 25_12 had ~26,600 sperm in her spermatheca, within the range of other queens collected in 2012 (Figure 5.3).

134

'#!"

'!!" ? >

&!"

%!"

/6,-47-* $!" (.84)9)*- 12--0 :3-4(")0",3-4(.8;-9."<='! #!"

!" ,34)05" ()*+,-.,/0 12--0, "12--0,

Figure 5.3. Stored sperm counts for D. arenaria queens. Spring queens (those collected on the wing in May and early June) and mid-season queens (those collected inside reproductive-stage colonies in July) had thousands of stored sperm in their spermathecae. The queen from the observed matricide colony 25_12, marked with a star, had approximately 26,600 stored sperm.

135 Surgery experiments

All eight surviving spermathecectomy queens were observed laying eggs within a few days of surgery. These observations included seeing the actual egg after oviposition, typically in the shallow cells at the edge of a lower comb. However, two of the eight queens were subsequently observed pumping their abdomens in an unusual manner, and partially everting the sting chamber. This eversion was also observed during attempted ovipositions in which these two queens failed to lay an egg. Thus, while these queens did lay some eggs, their egg-laying capacity was reduced after surgery. This behavior was not seen in any other queens.

The two surviving ovariectomy queens were also observed ovipositing. In these two surgeries, only the distal portion of the ovaries, including the germaria, was removed, leaving the larger oocytes in the posterior section of the abdomen intact. Thus, these queens had short-term oviposition capacity though it was not known how many viable oocytes remained after surgery.

Both queens died, one after 10 days and the other after 12 days, both associated with falls from the nest. These queens sometimes walked jerkily over the combs, as was seen in other surgery queens before they died as a result of the surgery. They were observed to fall from combs regularly, and eventually fell from the combs to the bottom of the nestbox where they were unable to return to the nest and died. They were carried out of the nestbox by workers, as observed on video.

Crucially, matricide was not observed in any of these colonies, even though the eight spermathecectomy and five sham surgery queens all survived for at least 21 days after surgery

(Table 5.2).

136

Table 5.2. Outcome of surgery experiments queens observed unobserved treatment survived matricides deaths spermathecectomy 8a 0 0 sham-surgery 5 0 0 ovariectomy 2b 0 0 unmanipulatedc 15d 3 2 aTwo of these queens had difficulty laying eggs. All survived for >21 days. bThese two queens died after 10 and 12 days, but their deaths were not due to matricide cThese observation colonies are reported in another manuscript, Chapter 4 dOne of these queens died because she slid to the bottom of the observation box while resting on the glass wall.

137 Discussion

The goals of this study were 1) to see if male-laying by queens triggers matricide; 2) to see if low queen fecundity triggers matricide.

Queen sex investment and matricide in collected colonies

The data do not support the prediction that workers preferentially kill queens who have switched to male production. This prediction stems from the fact that workers value females highly, so the cost of killing a queen is much higher if she is still producing females than if she is not. The queen was laying queen-destined females when she died in at least three of the six queen deaths for which there are data (Table 5.1). Additionally, two of ten queenright colonies were male specialists (Figure 5.2; see also (Greene et al. 1976), suggesting a male-laying queen does not trigger matricide. The latter result can be explained since these colonies had effective paternities near 2.0 (Chapter 4), giving little or no relatedness advantage to replacing brothers with nephews. But why would workers kill queens producing new queens in low-effective- paternity colonies, when they are highly related to those new queens? The strategy of delaying matricide until the queen switches to male production may not be stable if queen behavior coevolves with worker behavior. If matricide were triggered by the switch to male laying, individual queens could benefit from delaying the switch beyond the point where it is favorable to workers, making killing of female-laying queens beneficial earliest in single-paternity colonies, resulting in the observed facultative matricide based on paternity but not sex allocation. This would be compounded by a relatively low reproductive value of new queens if non-matricide colonies then continue to specialize on queen production, as appears to occur (Figure 5.2).

Furthermore, female reproductive value is already reduced, given that workers in queenless

138 colonies likely produce a large fraction of males in the population (Pamilo 1991). This admittedly convoluted hypothesis could explain the absence of facultative matricide in response to queen sex allocation, but clearly suggests that a formal coevolutionary model of queen and worker strategies would be helpful in clarifying the predictions regarding the timing of matricide.

Regardless of why workers ignore the sex of the queen’s offspring when committing matricide, the outcome appears to be split sex ratios in the direction opposite the classic pattern

(Boomsma and Grafen 1990; Boomsma and Grafen 1991; Sundström 1994; Sundström et al.

1996): high-effective-paternity colonies produce mostly queens, and low-effective-paternity, queenless colonies produce mostly males (Figure 5.2; Chapter 4). It is important to note that my estimates of sex investment are approximate: they are based on measurements of comb area instead of energetic investment, they do not include males that are regularly reared in worker cells (Bonckaert et al. 2011), and they do not reflect the fact that the reproductive stage is long enough that a point sample does not accurately capture total sex investment, particularly for colonies collected early in the summer. This being said, the investment patterns in reproductive cells are strongly bimodal, and these patterns undoubtedly correlate with overall sex allocation.

A similar pattern of higher female investment by lower-relatedness/higher-effective-paternity colonies occurs in another yellowjacket, Vespula maculifrons (Goodisman et al. 2007; Johnson et al. 2009), though the pattern is among queenright colonies. This species is always highly polyandrous (average effective paternity is ~5) and thus is not predicted to have matricide, and the few queenless mature colonies sampled had high effective paternity (Johnson et al. 2009).

Thus, it appears that different selection pressures are probably responsible for the effects of effective paternity on these two species.

139 Experimental manipulation of queen-laid sex ratio

The results of the spermathecectomy experiment also suggest that a queen laying only male eggs does not trigger matricide in this species. This is in some ways a better test of the hypothesis than the correlational data from collected colonies, because the experiment guarantees queens are not laying female eggs in the upper, worker-cell combs; this likely occurs in male- specializing queenright colonies in the reproductive stage (Greene et al. 1976; Greene 1984).

However, this experimental advantage is mitigated by the fact that there may be unknown factors in observation colonies that reduce the incidence of matricide: I only observed 3 matricides in 18 unmanipulated colonies (Chapter 4), though two additional queen deaths occurred but were not observed on video. This is a substantially lower fraction (17-28%) than the 42% of mature, queenless colonies observed in nature (Chapter 4). Regardless, the results are consistent with the data from collected colonies, suggesting that workers do not monitor and respond to the sex of the queen-laid brood.

The spermathecectomy experiment did not replicate the results reported by Montagner

(1966). This is somewhat surprising, given that D. arenaria is known to have matricide (unlike two of the three species used in his study), and because paternity patterns in D. arenaria support the relatedness prediction of kin-selected matricide (Chapter 4). There are several possible interpretations. It is possible that there are species differences in this behavior, though D. arenaria seems like a species most likely to exhibit responses to male-laying queens: this species is unusual for yellowjackets in that it has naturally occurring queens that exhibit irreversible male-production phases (Greene et al. 1976; Archer 2006). Alternatively, it’s possible that the male-laying queens Montagner reported as killed by the workers instead died from the irradiation treatment that caused the male egg laying, and were mistakenly observed being dismembered

140 and removed from the colony. It is difficult to know exactly what occurred in Montagner’s

experiments, as his brief report provides no details on how these queens were killed or how this

killing was observed.

Finally, the observation that a killed queen in colony 25_12 had ample sperm in her

spermatheca (Figure 5.3) suggests that queen killing (and the early decline of this colony; see

Chapter 4) is not triggered by queens switching to male production because of sperm depletion, a

plausible hypothesis for why some queens make this switch. Although sperm use patterns in D. maculata suggest that some queens may be sperm-limited (Stein and Fell 1996), there is no evidence that the male production phases of D. arenaria result from queens running out of sperm, as queenright male specialist colonies had queens with abundant stored sperm (Greene et al. 1976).

Queen fecundity and matricide

The results of the surgical manipulations do not suggest that workers preferentially kill low-fecundity queens (Bourke 1994). Four queens with compromised egg-laying abilities (two ovariectomy queens and two spermathecectomy queens that were observed to have difficulty laying eggs) survived for 10-21 days in reproductive-stage colonies, and none died because of matricide. For colonies with a short life cycle, this represents a substantial fraction of the time when reproductive-destined eggs are produced. That workers tolerated these queens suggests that low queen fecundity in the colony’s reproductive stage is not sufficient to trigger matricide.

The model predicting that queen fecundity should influence matricide assumes that colonies are egg-limited by a declining queen, and thus would benefit from killing a queen with reduced fecundity (Bourke 1994). The benefit of killing a low-fecundity queen who completely

141 inhibits worker reproduction is certainly greater than if her fecundity were higher. However, in

Dolichovespula wasps, workers obtain some direct reproduction even if the queen is present

(Foster et al. 2001; Bonckaert et al. 2011). This suggests that one of the benefits of killing a queen comes from removing a more fecund competitor in production competition (Cant 2012) for male eggs. If this is the case, then declining egg production by the queen means that the benefit of eliminating her is reduced; as her fecundity declines, a greater fraction of the eggs laid in the colony will already be workers’, even without killing the queen. Consistent with this effect, a greater fraction of worker-laid males are produced in queenright colonies with queens possessing a weaker fertility signal (Van Zweden et al. 2014). Furthermore, if low-fecundity queens have weaker policing behavior, then the benefit of matricide is further reduced, as workers may reproduce even more freely without having to kill the queen (van Honk et al. 1981).

These queens may be favored to cede some reproduction to workers if it prevents matricide; such a response is akin to the peace incentives predicted for multi-queen associations in reproductive skew models (Reeve and Ratnieks 1993). For these reasons, the relationship between queen fecundity and the benefits and costs of matricide are likely not as straightforward as the necessarily simplified scenario discussed in Bourke’s model. This could explain why our reduced-fecundity queens were not killed, though given that there were only four queens with experimentally reduced fecundity, this result is tentative.

Collected queenless (inferred matricide) colonies produced fewer worker cells than their queenright counterparts, which could be the result of low queen fecundity. However, nearly all of the collected queenless colonies switched from worker cell to queen cell construction, and workers likely control when this switch occurs. Thus, it may be that these colonies entered the reproductive stage earlier due to different worker behavior in response to intra-colony

142 relatedness (Chapter 4), rather than grew slowly due to low-fecundity queens. This would imply

that workers that kill their queen are paying an ergonomic cost, counter to the predictions of

Bourke’s model (Bourke 1994). Given these possible ergonomic costs, and the costs of killing

queens laying queen-destined eggs, the trade-offs facing matricidal workers in this species

requires further theoretical and empirical attention. Although workers appear to respond

adaptively to relatedness variation among colonies (Chapter 4), these additional aspects of queen

killing are at odds with the current theoretical framework for matricide in annual species (Bourke

1994).

Comparisons to other species

Several aspects of the reproductive biology of D. arenaria are unusual for vespine wasps,

but are remarkably similar to the bumblebee Bombus terrestris, in which conflict over male production has been studied extensively (Duchateau and Velthuis 1988; Bloch 1999; Bloch and

Hefetz 1999; Bourke and Ratnieks 2001; Duchateau et al. 2004; Zanette et al. 2012). First, D.

arenaria queens appear to have a critical switch point, after which they produce only males, at

least in reproductive cells (Greene et al. 1976; this study). To my knowledge, this does not occur

in other Dolichovespula, with the possible exception of D. norwegica (Akre and Myhre 1992;

Akre and Myhre 1994; Archer 2006); instead, other species appear to simultaneously produce

both males and queens in large reproductive cells. In two species of Vespa, a genus with reports

of matricide (Bourke 1994), colonies produce males before females, though their production

overlaps extensively (Martin et al. 2009). In contrast, the abrupt switch by D. arenaria queens

from diploid to haploid egg-laying mirrors the switch point that occurs in the annual colonies of

Bombus terrestris (Duchateau et al. 2004). Furthermore, like B. terrestris (Bourke and Ratnieks

143 2001; Duchateau et al. 2004), D. arenaria exhibits split sex ratios resulting from an early or late switch to male production in newly constructed combs (even when considering only queenright colonies; Figure 5.2; Greene et al. 1976).

Despite these similarities, worker-queen conflict unfolds differently in these two species.

Unlike in B. terrestris (Bloch 1999; Bourke and Ratnieks 2001), D. arenaria workers begin to lay eggs prior to the switch point (KJL, personal observation; also, see description of matricide colony 20_10 in Chapter 4). Furthermore, although B. terrestris workers have been observed to harass and eventually kill the queen after the switch point (van Doorn and Heringa 1986; van der

Blom 1986; Bourke 1994), D. arenaria workers sometimes kill queens who are still laying queen-destined eggs (Table 5.1). Worker reproduction in B. terrestris does not interfere with queen production, suggesting low costs of conflict for selfish workers (Lopez-Vaamonde et al.

2003), while matricide by workers in D. arenaria sometimes stops the production of new queens

(Figure 5.2; Table 5.1). Thus, despite many similarities in colony development, the conditions resulting in matricide are not particularly convergent in these two species, beyond the fact that matricide occurs in the reproductive stage, and predominantly in colonies with low effective paternity (Schmid-Hempel and Schmid-Hempel 2000). A detailed, quantitative study of individual worker inclusive fitness in this species would be useful to understand how the benefits of matricide outweigh the likely sizeable costs workers experience from killing their queen- producing mothers.

144 REFERENCES

Akre RD, Greene A, MacDonald JF, et al (1980) The Yellowjackets of America North of Mexico. United States Department of Agriculture Washington, DC

Akre RD, Myhre EA (1994) Nesting biology of Dolichovespula norvegicoides (Hymenoptera, Vespidae). Ent News 105:39–46.

Akre RD, Myhre EA (1992) Nesting biology and behavior of the Baldfaced Hornet, Dolichovespula maculata (L.) (Hymenoptera: Vespidae) in the Pacific Northwest. Melanderia 48:1–33.

Archer ME (2006) Taxonomy, distribution and nesting biology of species of the genus Dolichovespula (Hymenoptera, Vespidae). Entomol Sci 9:281–293.

Beekman M, Ratnieks FLW (2003) Power over reproduction in social Hymenoptera. Phil Trans Roy Soc Lond B 358:1741–1753.

Bloch G (1999) Regulation of queen-worker conflict in bumble-bee (Bombus terrestris) colonies. Proc R Soc Lond B 266:2465–2469.

Bloch G, Hefetz A (1999) Regulation of reproduction by dominant workers in bumblebee (Bombus terrestris) queenright colonies. Behav Ecol Sociobiol 45:125–135.

Bonckaert W, Van Zweden JS, D'Ettorre P, et al (2011) Colony stage and not facultative policing explains pattern of worker reproduction in the Saxon wasp. Mol Ecol 20:3455–3468.

Boomsma JJ, Grafen A (1990) Intraspecific variation in ant sex ratios and the Trivers-Hare hypothesis. Evolution 44:1026–1034.

Boomsma JJ, Grafen A (1991) Colony-level sex ratio selection in the eusocial Hymenoptera. J Evol Biol 4:383–407.

Boomsma JJ, Ratnieks FLW (1996) Paternity in Eusocial Hymenoptera. Phil Trans R Soc B 351:947–975.

Bourke AFG (2005) Genetics, relatedness and social behaviour in insect societies. In: Fellowes M, Holloway G, Rolff J (eds) Insect Evolutionary Ecology. CABI, Cambridge, MA, pp 1–30

Bourke AFG (1994) Worker matricide in social bees and wasps. J Theor Biol 167:283–292.

Bourke AFG, Ratnieks FLW (2001) Kin-selected conflict in the bumble-bee Bombus terrestris (Hymenoptera: Apidae). Proc R Soc Lond B 268:347–355.

Cant MA (2012) Suppression of social conflict and evolutionary transitions to cooperation. Am Nat 179:293–301.

145 D'Ettorre P, Heinze J, Ratnieks FLW (2004) Worker policing by egg eating in the ponerine ant Pachycondyla inversa. Proc R Soc Lond B 271:1427–1434.

Daly D, Archer ME, Watts PC, et al (2002) Polymorphic microsatellite loci for eusocial wasps (Hymenoptera: Vespidae). Mol Ecol Notes 2:273–275.

Duchateau MJ, Velthuis H (1988) Development and reproductive strategies in Bombus terrestris colonies. Behaviour 107:186–207.

Duchateau MJ, Velthuis HH, Boomsma JJ (2004) Sex ratio variation in the bumblebee Bombus terrestris. Behav Ecol 15:71–82.

Foster KR, Ratnieks FLW (2001) Paternity, reproduction and conflict in vespine wasps: a model system for testing kin selection predictions. Behav Ecol Sociobiol 50:1–8.

Foster KR, Ratnieks FLW, Gyllenstrand N, Thoren P (2001) Colony kin structure and male production in Dolichovespula wasps. Mol Ecol 10:1003–1010.

Freiburger B, Breed M, Metcalf JL (2004) Mating frequency, within!colony relatedness and male production in a yellow jacket wasp, Dolichovespula arenaria. Mol Ecol 13:3703–3707.

Goodisman MAD, Kovacs JL, Hoffman EA (2007) The significance of multiple mating in the social wasp Vespula maculifrons. Evolution 61:2260–2267.

Greene A (1984) Production schedules of vespine wasps: an empirical test of the bang-bang optimization model. J Kansas Entomol Soc 57:545–568.

Greene A, Akre RD, Landolt PJ (1976) The aerial yellowjacket, Dolichovespula arenaria (Fab): nesting biology, reproductive production, and behavior (Hymenoptera: Vespidae). Melanderia 26:1–34.

Hamilton WD (1972) Altruism and related phenomena, mainly in social insects. Annu Rev Ecol Syst 3:193–232.

Jaffé R (2014) An Updated Guide to the Study of Polyandry in Social Insects. Sociobiology 61:1–8.

Johnson E, Cunningham T, Marriner SM, et al (2009) Resource allocation in a social wasp: effects of breeding system and life cycle on reproductive decisions. Mol Ecol 18:2908–2920.

Koeniger G (1970) Bedeutung der tracheenhülle und der anhangsdrüse der spermatheka für die befruchtungsfähigkeit der spermatozoen in der bienenkönigen (Apis mellifica L.). Apidologie 1:55–71.

Kronauer DJC, Boomsma JJ (2007) Do army ant queens re-mate later in life? Insectes Soc 54:20–28.

Liebig J, Peeters C, Holldobler B (1999) Worker policing limits the number of reproductives in a

146 ponerine ant. Proc R Soc Lond B 266:1865–1870.

Loope KJ, Chien C, Juhl M (2014) Colony size is linked to paternity frequency and paternity skew in yellowjacket wasps and hornets. BMC Evol Biol 14:277.

Lopez-Vaamonde C, Koning JW, Jordan WC, Bourke AFG (2003) No evidence that reproductive bumblebee workers reduce the production of new queens. Anim Behav 66:577– 584.

Martin SJ, Takahashi J, Katada S (2009) Queen condition, mating frequency, queen loss, and levels of worker reproduction in the hornets Vespa affinis and V. simillima. Ecol Entomol 34:43–49.

Montagner H (1966) Sur l’origine des mâles dans les sociétés de Guêpes du genre Vespa. C R Acad Sc Paris 263:785–787.

Pamilo P (1991) Evolution of colony characteristics in social insects. I. Sex allocation. Am Nat 137:83–107.

Queller DC, Strassmann JE (1998) Kin selection and social insects. BioScience 48:165–175.

Ratnieks FLW (1988) Reproductive harmony via mutual policing by workers in eusocial Hymenoptera. Am Nat 132:217–236.

Ratnieks FLW, Foster KR, Wenseleers T (2006) Conflict resolution in insect societies. Annu Rev Entomol 51:581–608.

Ratnieks FLW, Reeve HK (1992) Conflict in single-queen hymenopteran societies: the structure of conflict and processes that reduce conflict in advanced eusocial species. J Theor Biol 158:33–65.

Ratnieks FLW, Visscher PK (1989) Worker policing in the honeybee. Nature 342:796–797.

Reeve HK, Ratnieks F (1993) Queen-queen conflicts in polygynous societies: mutual tolerance and reproductive skew. In: Keller L (ed) Queen number and sociality in insects. Oxford University Press, Oxford. pp 45–85

Schmid-Hempel R, Schmid-Hempel P (2000) Female mating frequencies in Bombus spp. from Central Europe. Insectes Soc 47:36–41.

Schneiderman HA (1967) Insect Surgery. In: Wilt FH, Wessells NK (eds) Methods in developmental biology. T.Y. Crowell Co, pp 753–766

Stein KJ, Fell R (1996) Sperm use dynamics of the baldfaced hornet (Hymenoptera: Vespidae).

147 Environ Entomol 25:1365–1370.

Strambi A (1965) Essai d'analyse de la dynamique ovarienne chez les guêpes polistes (P. gallicus L. et P. nimpha Christ) en activité reproductrice. Insectes Soc 1:1–18.

Stroeymeyt N, Brunner E, Heinze J (2007) “Selfish worker policing” controls reproduction in a Temnothorax ant. Behav Ecol Sociobiol 61:1449–1457.

Sundström L (1994) Sex ratio bias, relatedness asymmetry and queen mating frequency in ants. Nature 367:266–268.

Sundström L, Chapuisat M, Keller L (1996) Conditional manipulation of sex ratios by ant workers: a test of kin selection theory. Science 274:993–995.

Thorén PA, Paxton RJ, Estoup A (1995) Unusually high frequency of (CT)n and (GT)n microsatellite loci in a yellowjacket wasp, Vespula rufa (L.) (Hymenoptera: Vespidae). Insect Mol Biol 4:141–148.

Trivers R, Hare H (1976) Haplodiploidy and the evolution of the social insects. Science 191:249–263. van der Blom J (1986) Reproductive dominance within colonies of Bombus terrestris (L.). Behaviour 97:37–49. van Doorn A, Heringa J (1986) The ontogeny of a dominance hierarchy in colonies of the Bumblebee Bombus terrestris (Hymenoptera, Apidae). Insectes Soc 33:3–25. van Honk CGJ, Roseler PF, Velthuis HHW, Hoogeveen JC (1981) Factors influencing the egg laying of workers in a captive Bombus terrestris colony. Behav Ecol Sociobiol 9:9–14.

Van Zweden JS, Bonckaert W, Wenseleers T, D'Ettorre P (2014) Queen signaling in social wasps. Evolution 68:976–986.

Wenseleers T, Tofilski A, Ratnieks FLW (2005) Queen and worker policing in the tree wasp Dolichovespula sylvestris. Behav Ecol Sociobiol 58:80–86.

West S (2009) Sex Allocation. Princeton University Press, Princeton

Woyciechowski M, Lomnicki A (1987) Multiple mating of queens and the sterility of workers among eusocial Hymenoptera. J Theor Biol 128:317–327.

Zanette LRS, Miller SDL, Faria CMA, et al (2012) Reproductive conflict in bumblebees and the evolution of worker policing. Evolution 66:3765–3777.

148 Appendix 5.1. Details of observation box setup and surgical methods

Observation box setup

Workers were collected with a hand-held, battery-powered insect vacuum (BioQuip Inc.).

Collection took place during the day, and I waited for at least 30 minutes to collect nearly all returning foragers (initial observations suggest that 95% of foragers that return within 1 hour do so within the first 30 minutes of collection). The nest was then cut into a zip-close plastic bag, and the plastic chamber containing workers was placed on ice.

In the laboratory, the nest envelope was carefully removed by making an incision along one side. The combs typically consisted of a first comb of worker cells and the start of a second comb with no pupal caps; only occasionally did I use colonies that had begun construction of a third or fourth comb. The upper comb was glued into the top of a wooden frame observation box that was either 1.5 or 2 inches wide, depending on the size of the nest when collected (see Figure

5.1d). The envelope was trimmed on opposite sides to permit view of the comb, and then glued in place around the comb.

Workers and the queen were placed in a refrigerator until anesthetized. Queens were marked with a thin copper wire bent in a loop between abdomen and thorax, or with a large paint mark (Sharpie paint pen) on the thorax. This made identifying the queen easier in large colonies with post-emergence gynes. All individuals were placed inside the nest while still anesthetized.

Cotton balls soaked in honey-water were placed in the bottom of the box, and then the glass walls were attached with duct tape. The colony was kept in the dark until after nightfall, when the box was installed in one of several sheds at the Liddell Field Station, and connected via a 1- inch-diameter Tygon tube to the outdoors. Foraging resumed the next morning.

149 Spermathecetomy

Prior to surgery, the queen was extracted from the nest either after collecting workers

with a vacuum and net, or in 2013 after briefly anesthetizing the entire colony with CO2 (only enough to prevent workers from flying when the box was opened, typically no longer than 1 minute). The queen was placed abdomen-first into a modified 1500 µL Eppendorf tube (Figure

5.1c). This tube held the queen in place while CO2 was gently flowing from a Tygon tube connected to the mouth of the Eppendorf tube. The tube containing the queen was positioned using a movable jointed arm, and she was oriented with her ventral side upward, with the sting pointing toward me. Forceps on a micromanipulator grasped the ultimate sternite while a hook fashioned from the tip of an insect pin held the penultimate sternite. By retracting the ultimate sternite, the membrane between these two segments was exposed (Figure 5.1). I then made an incision in the membrane with a shard of sterilized razor blade, and immediately placed a drop of sterile Ringer’s solution into the aperture. Then, using two sets of fine forceps, the common oviduct was brought into view. Care was taken not to tear muscles that connect the two segments and run across this region, and the ventral nerve cord was also seen and avoided. Once the common oviduct was in view, it was a simple procedure to pull it to one side with one pair of forceps, and to grasp the spermatheca with the other pair. The spermatheca was then removed without rupturing it. Before closing the wound, a few crystals of phenylthiourea, penicillin G and streptomycin sulfate (ratio 2:1:1) were placed in the abdominal cavity (Schneiderman 1967).

The wound was not sealed, but the overlapping segments closed the wound effectively. The entire process usually took around 3-4 minutes. This method was practiced on dozens of spring queens of various species before implementation on observation colony queens.

150 Ovariectomy

The queen was extracted from the colony in the same manner as for spermathecectomies.

She was anesthetized with CO2 and held in place with a forked pin positioned at the junction between thorax and abdomen (Figure S5.1). The junction between the two tergites was held open using a forceps and pin hook mounted on micromanipulators. I made a small incision in the intersegmental membrane with a sterilized shard of razor blade and immediately added a drop of sterile Ringer’s solution to the opening. The anterior-most section of the ovaries was located, typically plastered to the foregut. The ovarioles were grasped between forceps and carefully drawn from the abdominal cavity, minimizing as much as possible the tearing of Malphigian tubules tangled around the ovaries. In most cases, the ovaries were not completely removed, as the ovarioles tore before the largest (most posterior) oocytes were pulled out. However, in all cases the germarium and the majority of small distal oocytes were removed, leaving at most a few dozen large oocytes in the ovaries. After pulling out the ovaries, the abdominal cavity was filled with Ringer’s solution and closed by allowing the segments to overlap. The entire procedure typically took 3-4 minutes.

This method was practiced on ~20 spring queens, and most survived for several days in wire cages with regular sugar-water feeding before being killed. However, the surgery was much more difficult on mid-season queens used in the experiment due to the much greater volume occupied by the ovaries in the abdomen.

151

Figure S5.1. A queen about to have her ovaries removed. This queen, anesthetized with CO2, had her ovaries removed via the visible aperture between tergite 2 and tergite 3.

152 CHAPTER 6

SOCIAL ORGANIZATION PREDICTS CUTICULAR HYDROCARBON DIVERSITY IN

PAPER WASPS

Kevin J. Loope and Patrizia D’Ettorre

Abstract

Cuticular hydrocarbons (CHCs) are a central modality for recognition in insect social interactions (Blomquist and Bagneres 2010). In the societies of bees, wasps, ants and termites, the complex, non-volatile chemical profiles on the exoskeleton can encode sex, caste, dominance, fertility, nest membership, task group, kin group and individual identity (1). Species vary greatly in the diversity of compounds found in their profiles (Martin and Drijfhout 2009a).

Although several authors have proposed that such differences could reflect variation in social organization (Boomsma et al. 2003; Wyatt 2010) there are few studies explaining variation across social insect species’ CHC profiles (Van Wilgenburg et al. 2011; Menzel and Schmitt

2012). Here we show that small-colony, independent-founding paper wasp species have more diverse CHC profiles than do their large-colony, swarm-founding relatives. These two types of social organization differ primarily in the number of methyl-branched alkanes, a compound class used in nestmate recognition (Dani et al. 2001; Guerrieri et al. 2009). In contrast, social organization poorly predicts a species’ diversity of linear alkanes, a class of compounds that is probably less useful for recognition (Dani et al. 2001; Chaline et al. 2005). These results provide rare support for the hypothesis that these important recognition compounds evolve in response to selection on their communication function, and may thus be true signals, rather than cues.

153 Introduction

Cuticular hydrocarbons are a central modality for communication in insects, and the CHCs of ants, bees, wasps and termites have become an important system for understanding the mechanisms and ontogeny of animal recognition (Howard and Blomquist 2005; Le Conte and

Hefetz 2007; Blomquist and Bagneres 2010). Many studies describe or manipulate hydrocarbon variation within a species to reveal how cuticular compounds facilitate recognition. These studies show that often, particular communicative functions are encoded in the relative abundance of particular subsets of hydrocarbons: for example, in some species, methyl-alkanes

(Dani et al. 2001), dimethyl-alkanes (Guerrieri et al. 2009), or alkenes (Martin et al. 2008) are used for nestmate recognition, while linear and methyl-alkanes often signal fertility (Liebig et al.

2000; Smith et al. 2009; Van Oystaeyen et al. 2014). Interestingly, numerous behavioral studies suggest that linear alkanes are not used as nestmate recognition cues (Dani et al. 2001; Dani et al.

2005; Martin et al. 2008; Brandt et al. 2009; Guerrieri et al. 2009), perhaps because their featureless chemical structures makes distinguishing them difficult (Chaline et al. 2005; Van

Wilgenburg et al. 2012).

This approach has been successful in revealing how CHC profiles facilitate communication, but it leaves open the question of how and why the profiles themselves evolve

(Van Wilgenburg et al. 2011; Menzel and Schmitt 2012). CHC profiles could evolve neutrally, or, because of their important waterproofing function (Gibbs 1998), CHCs could be entirely determined by insects’ water balance needs. Were this the case, communication based on CHCs would be cue-based, entirely dependent on non-communication functions to maintain cue diversity and reliability. However, given the central role of CHCs in communication, there are several hypotheses suggesting how social evolution may shape CHC profiles. Profile complexity

154 could expand in response to the greater complexity of communication in more social species

(Wyatt 2010), particularly if the multiple signals encoded within the CHC profile interfere with

one another (Moore and Liebig 2010; Oi et al. 2015). Alternatively, the costs and benefits of

recognition accuracy may alter profiles. Profile diversity could decline if accurate kin- informative cues permit costly nepotism (Boomsma et al. 2003). Benefits of more accurate nestmate recognition may favor rare kin-informative cues (Ratnieks 1991), which could result in the expansion of profile diversity by selecting for rare novel compounds. Similarly, benefits of accurate recognition of any type could lead to the addition of compounds to a profile, effectively adding channels to a multi-component signal (Tibbetts and Dale 2007; Ay et al. 2007; Wilson et al. 2013).

We tested the hypothesis that social organization drives cuticular hydrocarbon evolution by comparing compound diversity in the profiles of paper wasp species. The paper wasps

(Vespidae: Polistinae) are a diverse subfamily of 26 genera and 958 species (Pickett and

Carpenter 2010), and they divide into two main types of social organization: the independent- founding (IF) and the swarm-founding (SF) species (Jeanne 1980; Jeanne 2003). IF species typically have small colonies (<50 individuals) started by one or several aspiring queens. These groups can include both relatives and non-relatives (Queller et al. 2000), and are typically organized by dominance and reproductive hierarchies (Pardi 1948; Jeanne 1980). In contrast, swarm-founding species typically have large colonies of hundreds to tens of thousands of individuals. New colonies result from fission, with a large group of queens and workers departing to start a nest together. Colonies typically contain tens to hundreds of queens (Jeanne

1991). This derived social system has probably evolved at least four times (Henshaw et al. 2001;

Saito-Morooka 2014), allowing a comparison between the CHC profiles of these taxa and their

155 independent-founding relatives. Given that CHC profiles may also respond to climate (Chapman

et al. 1995; Frentiu and Chenoweth 2010; Van Wilgenburg et al. 2011), likely because of their

essential role in waterproofing the insect cuticle (Gibbs and Rajpurohit 2010), we also included

latitude as a proxy for climate.

Results

CHC diversity

Phylogenetic generalized least squares analyses on 34 species show that social type

strongly influences total CHC diversity (Figure 6.1; Table 6.1; Table S6.1). The slope estimates

from the best supported model suggest that swarm-founding species have substantially fewer

CHC compounds in their profiles, and that compound number increases with latitude across

species. The results were virtually the same for the diversity of non-linear compounds (all

compounds except linear alkanes), with the best model on average explaining 49% of the

variance (adjusted R2 = 0.49). Methyl-branched alkanes account for most of this difference

(Figure S6.1); means for methylalkanes all differ substantially between IF and SF species (Figure

S6.1; mean±SEM monomethyls: IF = 31.86 ± 2.66, SF = 20.77 ± 5.76; dimethyls: IF = 19.71 ±

2.17, SF = 7.08 ± 1.96; trimethyls: IF = 2.33 ± 0.82, SF = 0.08 ± 0.02). In contrast to the results for these “non-linear” CHCs, linear alkane diversity was not well explained by our predictors, with the intercept-only model included in the preferred model set and the best model (latitude alone) explaining only 7% of the variance (Table 6.1). The mean number of linear alkanes in the profiles of IF and SF species was virtually identical (IF = 6.1 ± 0.52; SF = 6.31 ± 1.75; Figure

6.1e).

156 Our main result, that social type strongly influences non-linear and total (but not linear)

CHC diversity, is robust to changes in several aspects our analysis. These findings are robust to

phylogenetic uncertainty, given the similarity of results across trees (Table 6.1;Table S6.1).

They are also robust to uncertainty in the estimate of Pagel’s # (Revell 2010): even when # in the

preferred model (Social + Latitude) is set to the upper 95% confidence interval value for each

tree, the mean confidence interval for the effect of social type does not overlap zero and the

estimate is only slightly lower (-18.9 vs -21.9 for total CHC diversity and -18.0 vs -21.1 for non-

linear CHC diversity; see Table S6.2). Whether or not we included the four species from

literature data had little qualitative effect on the results for total and non-linear diversity: when

only analyzing our data, slope estimates in the preferred models were slightly lower for social

type, though mean confidence intervals did not include zero (Table S6.3). However, for linear

alkane diversity, excluding the four species from the literature resulted in Social + Latitude

becoming the preferred model, with a moderate negative effect of both predictors. Finally, the

effect of social type on non-linear and total diversity was still strong when using 0.1% average

relative abundance as the inclusion threshold for GC-MS peaks instead of 0.5% (Table S6.4).

The relative abundance of linear alkanes was weakly predicted by social type and

latitude: the set of preferred models included the intercept-only model, and the mean slope

confidence intervals overlapped with zero in all models (Table S6.5; Figure S6.2). Average

chain length was similarly poorly explained by our predictors (Table S6.5; Figure S6.2). Social

type and latitude are therefore best able to explain CHC diversity, but not aspects of the profile

that are more likely to respond to abiotic factors. This suggests that diversity differences are not a consequence of selection for these other aspects of the profile that might indirectly result in a change in compound diversity.

157 2$03+, (*'3, +2@+03% ')5,(&+,6(0% ! ?" " !" $

1(2$%*3%

9053/30530*:;(-05$08

./$/(0$0$ %

#$%&'(&)**+,-%

4(/+2$5$+

1(2)6$($53%

1+,+/(2)6$+

732(0(8+%*3, <=+,>:;(-05$08

63,E(;E0(0:2$03+,EFGF% D->63,E(;E2$03+,EFGF%

Figure 6.1. Cuticular hydrocarbon evolution in paper wasps. a. One of 100 phylogenetic trees used in a comparative analysis of CHC diversity. Key clades are labeled with grey bands. Black bars depict the number of linear alkanes in each species’ profile, while green (IF) and orange (SF) bars represent the diversity of ‘non-linear’ compounds (all CHCs except linear alkanes). b. A colony of Mischocyttarus mexicanus, a typical independent-founding (IF) species. c. A colony of Polybia occidentalis, a typical swarm-founding (SF) species. d. Species’ diversity of non- linear CHCs, by social type (swarm-founding or independent-founding) and latitude. e. Species’ diversity of linear alkane diversity by social type and latitude.

158

Table 6.1. Models of CHC diversity. Response Predictor(s) #AICc Weight Slope Adj. R2 Social + -21.86 0 0.73 0.48 Latitude* 0.55 Total number of Social 2.33 0.23 -28.16 0.40 compounds Latitude 6.52 0.03 0.73 0.16

1 9.63 0.01 ------

Social + -21.1 0 0.81 0.49 Latitude* 0.64 Number of non- linear Social 3.63 0.13 -28.54 0.39 compounds Latitude 5.76 0.05 0.82 0.2

1 10.57 0 ------Latitude* 0 0.48 -0.07 0.07 1 1.32 0.25 ------Number of Social + -0.76 linear alkanes 1.79 0.19 0.06 Latitude -0.08

Social 3.5 0.08 0.26 -0.03 Notes: ‘Non-linear’ refers to the number of those compounds that are not linear alkanes. PGLS model statistics are mean values from analyses on 100 trees. For ranges and confidence intervals, see Table S6.1. Lines in bold are preferred models (mean &AICc <2). Slope of “Social” is the difference between SF and IF species. Latitude units for slope estimates are in degrees. * These models had the minimum &AICc in the model set for each of the 100 trees.

159 Discussion Our results suggest that social organization strongly affects CHC profile diversity, and that latitude may also influence CHC diversity. Previous studies have reported latitudinal clines in

CHC profile properties (Frentiu and Chenoweth 2010), and these patterns have been attributed to differences in climate. However, only a few comparative studies have attempted to explain interspecific variation in social insect CHCs (Van Wilgenburg et al. 2011; Menzel and Schmitt

2012), and ours is the first to suggest that intraspecific social organization may drive CHC profile evolution. In support of the hypothesis that selection for recognition function has created this difference, the compound classes that differ are those with more complex structures such as the methyl-branched alkanes; these compounds are known to be important for recognition in social insects, including independent-founding paper wasps. The diversity of linear alkanes, thought to be less useful for recognition, is not well predicted by social type. This suggests a selective maintenance of those compounds most useful for accurate recognition in independent founding, but not swarm-founding, taxa.

Our results imply that compound diversity has been maintained in independent founding species but lost in the derived swarm-founders. What costs would select against maintaining high CHC diversity in all species? In addition to the costs of producing and perceiving additional compound types, the compounds with methyl groups or double bonds that are useful for recognition are less effective at waterproofing than are linear alkanes (Gibbs and Rajpurohit

2010). Thus, it may be the trade-off with efficient desiccation resistance that reduces CHC diversity in species where selection for diversity is not strong.

Theory suggests that additional compounds can enhance recognition accuracy (Breed and

Buchwald 2008; Wilson et al. 2013), but how many compounds are necessary for accurate recognition? Independent-founding species have an average of 58 non-linear CHCs, which at

160 face value is much higher than the 8-16 recognition compounds predicted by Breed and

Buchwald (Breed and Buchwald 2008) to be necessary for accurate nestmate recognition.

However, there is evidence that the production of subsets of compounds are physiologically linked and thus highly correlated (Martin and Drijfhout 2009b), and that many compounds are likely generalized into sets that are perceptually indistinguishable (Chaline et al. 2005; Bos et al.

2012; Van Wilgenburg et al. 2012). This means that even if 58 compounds are present, there are many fewer independent channels of information. Furthermore, these profiles encode multiple messages beyond nestmate identity, including information about dominance (Dapporto et al.

2010) and relatedness (Leadbeater et al. 2014), which further reduces the number of available cues. Together, this suggests the difference between SF non-linear diversity (mean = 30) and IF non-linear diversity (mean = 58) likely represents a meaningful difference in the number of channels available for communication.

There are several aspects of paper wasp biology that could explain the more diverse CHC profiles in independent-founding wasp species, if such profiles facilitate more accurate recognition. First, in independent founding species, the founding stage involves assessment of relatedness(Reeve et al. 2000; Leadbeater et al. 2014), dominance (Dapporto et al. 2010), and perhaps fecundity (Liebig et al. 2005) among a handful of cooperating rivals for queenship.

These complex and dynamic social interactions have selected for sophisticated recognition systems (Tibbetts and James Dale 2004; Dapporto et al. 2010), sometimes including individual recognition(Tibbetts 2002; Tibbetts 2004; D'Ettorre and Heinze 2005). This founding stage does not exist in swarm-founding species, where colonies are always large, and one-on-one interactions and individual complexity are replaced by distributed interactions and colony-level complexity (Jeanne 2003). Second, interactions with potential usurpers, both inter- and intra-

161 specific, are more common in the independent founding species. These species often nest in dense aggregations (Strassmann 1991), making encounters with non-nestmates frequent.

Takeover of nests by non-nestmates is common in many species (Strassmann 1991; Tindo et al.

1997; Ito and Itioka 2008), in part because young nests often fail, creating a large pool of potential usurpers(Klahn 1988; Makino 1989; Dani and Cervo 1992). Furthermore, there are several known obligate and facultative interspecific social parasites of independent-founding species (Cervo 2006), and likely more that remain to be described (Buck et al. 2012).

Coevolution with such parasites could select for more accurate recognition and thus greater CHC diversity: this hypothesis likely explains CHC diversification in paper wasp (Lorenzi et al. 2014) and ant (Martin et al. 2011) populations with specialized social parasites. On the other hand, inter- and intra-specific social parasitism is, to our knowledge, unreported for swarm-founding wasps. Finally, the cost of recognition error in the small societies of independent-founding species is likely much higher. An erroneously accepted usurper in an IF colony with a single dominant reproductive could result in workers rearing a nest full of unrelated offspring.

However, in the large, polygynous colonies of the swarm-founding species, an acceptance error would, at worst, result in the addition of a single unrelated reproductive to the tens or hundreds of related queens in the nest, making the cost per error relatively low.

Whether the hypotheses we outline here, or others, explain the effect of social organization on CHCs, cannot be determined by our correlative study. Understanding the selective regimes responsible for this difference will require further behavioral experiments, especially in the understudied swarm-founding species. Further comparative analyses, for example comparing CHC diversity between social and non-social insects (Wyatt 2010), or among the simple and complex societies of other eusocial groups, may shed more light on the

162 links between social evolution and signal evolution in the social insects. Understanding the evolution of cue diversity from the perception side will also be valuable. Greater vomeronasal receptor protein diversity in mammal species with greater major urinary protein diversity suggests a co-evolutionary expansion of both recognition signals and signal receptor proteins

(Chamero et al. 2007). Recent comparative genomic work in bees suggests an expansion of an odorant-binding protein gene family with the origin of eusociality (Kapheim et al. 2015), though the role of these receptors in communication is unknown. Comparative studies at the perceptual end, both behavioral and genomic, will be important in further testing the hypothesis that CHC diversity evolves to facilitate recognition.

We have not yet discussed the role of multi-modal recognition, but other modalities are clearly important in wasp social interactions. Visual signals are known to be used in recognizing individuals (Tibbetts 2002) and assessing dominance (Tibbetts and James Dale 2004; Tannure-

Nascimento et al. 2008) in independent-founding Polistines, and for recognition of non- nestmates in stenogastrine wasps with similar social organization (Baracchi et al. 2013; Baracchi et al. 2015). If visual cues replace chemical cues in recognition behavior (Baracchi et al. 2015), we might predict lower CHC diversity in species that use visual as well as chemical cues.

Alternatively, the same selective forces that expand CHC diversity could similarly select for the use of cues in additional modalities to further enhance recognition. Of the ten Polistes species in our data set that were analyzed for variable facial cues in previous studies (Tibbetts 2004;

Tibbetts and James Dale 2004), the total compound diversity is higher in the four species with highly variable faces than the six species with invariant faces (t = 2.49, df = 6.31, p = 0.047).

Given that this analysis ignores phylogeny and relies on few species with poorly understood visual signaling systems, this preliminary pattern is merely suggestive, and serves to motivate

163 further investigation of the evolutionary link between visual and chemical signaling in wasp recognition systems.

Methods

Phylogeny

To perform our comparative analyses, we required a ultrametric phylogenetic tree for our 34 taxa.

No such tree exists, so we generated some from published sequence data. First, we obtained sequence data for 122 taxa in the Polistinae for the genes 12S, 16S, 28S and COI, as well as the outgroup species Vespula maculifrons (accession numbers in Table S6). Most sequences came from a recent, simultaneous analysis of wasp phylogeny using sequences and phenotypic data by

Pickett and Carpenter (Pickett and Carpenter 2010). Because this analysis included few species in the genera Belonogaster, Ropalida, Polybioides and Parapolybia, we searched Genbank for additional sequences from all species in these genera. We also sought sequences for additional species in the Epiponine genera present in our comparative dataset, as well as all Polistes and

Mischocyttarus species for which we had trait data but lacked sequences from Pickett and

Carpenter (Pickett and Carpenter 2010).

We prepared sequences using phyloGenerator (Pearse and Purvis 2013), aligning with

MUSCLE 3.8.31 (Edgar 2004), and trimming with trimAl 1.2 (Capella-Gutiérrez et al. 2009). A posterior distribution of trees was generated with BEAST 1.8.1 (Drummond et al. 2012) using default GTR-Gamma model settings and 4,000,000 generations, saving trees every 1000 steps.

Because we sought a set of trees consistent with the current consensus while also reflecting the ambiguous placement of basal genera, we edited the BEAST XML file to force monophyly in key clades present in Pickett and Carpenter (Pickett and Carpenter 2010) (Figure S6.5). As

164 desired, the resulting posterior tree topologies varied mostly in the placement of Belonogaster,

Ropalidia, Polybioides and Parapolybia taxa, and the topologies within the Epiponini, Polistes and Mischocyttarus were similar to that of Pickett and Carpenter (Pickett and Carpenter 2010).

We visually confirmed convergence, discarded the first 1500 of trees as burn-in, and randomly sampled 100 trees from the post-convergence distribution for use in comparative analyses

(available in supplementary data). For visualization, we also generated a 0.5 threshold clade frequency consensus tree from the post-burn-in distribution using sumTrees 3.3.1 within

DendroPy (Sukumaran and Holder 2010) (Figure S6.6).

These 100 trees were edited for use in comparative analyses with the R packages ape

(Paradis et al. 2004) and phytools (Revell 2011). Two species for which we had CHC data

(Metapolybia mesoamerica and Agelaia cajennensis) lacked sequence data but were the only members of their genera in our dataset. These were added to each tree in place of a species in each genus (because the remaining species in these genera would later be removed from the tree, branch length would be the same regardless of the replaced species). For the five species with congeners in the dataset but no sequence data for tree inference (Ropalidia montana, Ropalidia marginata, Polybioides raphigastra, Parapolybia nodosa, and Polybia micans), we inserted them in to the trees within genera. For Polybioides raphigastra and Parapolybia nodosa, with a single congener in each tree, we inserted each as a sister to its congener with branch lengths set to the average intrageneric branch length (across all genera) for that tree. The two Ropalidia species were inserted randomly within the genus using the phytools function add.species.to.genus(). Polybia micans is a member of the subgenus Trichinothorax (Richards

1978) and was similarly added to that clade (which was always monophyletic, given our constraint tree; see Pickett and Carpenter 2010). Finally, we pruned all but the 34 species in our

165 dataset, yielding 100 ultrametric trees for use in comparative analyses.

CHC data

We analyzed CHCs from the extracted profiles of three individuals from each of 30 species of

polistines. These individuals were collected by us, or obtained from private or museum

collections (Table S6.1). All three individuals for each species were collected from different

colonies, or, in the case of some museum specimens, from different dates or sites within a

population. Wasps collected by KJL were from private property in eastern North America or

were collected at La Selva Biological Station in Costa Rica (see Table S6.5). The Costa Rican

specimens (Agelaia cajennensis and Metapolybia mesoamerica) were collected under permit

number 018-2010-SINAC. These CHCs were extracted by immersion in hexane for 10 minutes

(see Figure S6.4 for museum specimen extraction). Preliminary analyses on profiles extracted

from frozen wasps from 2013 and museum specimens from the same location showed that CHC

profiles are highly stable, at least qualitatively, as has been found by others (Figure S6.3, S6.4;

Page et al. 1990; Martin et al. 2009). This sampling was not biased between social types, and

thus are not responsible for patterns that we observed.

Samples were extracted for 10 minutes in 1-10ml hexane, depending on the size of the specimen (each species had the same extraction volume for all samples). Extracts were partially evaporated, transferred to small vials, and then allowed to dry in a fume hood. These samples were shipped at room temperature from New York to Paris, where samples were re-eluted with hexane the internal standard octacosane. Two microliters of sample were injected in splitless mode on an Agilent 7890A GC coupled to a 5975C MSD. The GC was equipped with a HP-

5MS column (30m x 250!m x 0.25!m). Inlet temperature was 280 °C. Oven temperature was

166 held at 70 °C for 1 min, then ramped to 200 °C at 35 °C/min, followed by a 4 °C/min ramp to

320 °C, with a hold for 20 minutes at 320°.

We identified CHC compounds using diagnostic ions and retention times (Carlson et al.

1998). We did not include the occasional alcohols, esters, and other polar compounds we encountered (these were typically in trace abundance or varied greatly in presence/abundance between individuals, suggesting they resulted from glandular or internal contamination). To construct a species’ CHC profile, we included compounds from all peaks that were present in at least two of the three individuals analyzed. We averaged peak relative abundance across the three samples, then removed all peaks with less than a trace abundance set at 0.5% of the total peak area. This removes trace compounds unlikely to be involved in communication via relative abundance, and more importantly avoids variation in CHC total abundance from influencing the species’ profile. In other words, we were reliably able to identify compounds in peaks greater than this size, regardless of total quantity of CHCs of an individual. This prevents bias due to size of the individual or quantity of CHCs extracted. We also repeated the analyses using a cutoff of 0.1%.

We searched the literature for polistine species with quantitative, complete CHC profile descriptions determined with methods similar to ours by consulting recent reviews of wasp

CHCs (Dani 2006; Bruschini et al. 2010) and searching Google Scholar for “Polistinae CHC” with publication dates from 2010 to present. We omitted four species that are interspecific social parasites, as these species’ profiles may be under substantial selection resulting from their parasitic behavior (Polistes nimphus, P. atrimandibularis, P. sulcifer and P. semenowi (Bagnères and Cristina Lorenzi 2010)). We also omitted one species whose CHC profile is likely influenced by symbioses with other species (Parachartergus aztecus; Espelie and Hermann

167 1988). This left four species not already in our dataset: Polistes biglumis (Lorenzi et al. 1997),

Ropalidia opifex (Dapporto et al. 2006), Synoeca septentrionalis (Kelstrup et al. 2014b), and

Polybia micans (Kelstrup et al. 2014a).

For Ropalidia opifex (Dapporto et al. 2006), a swarm-founding species, we conservatively assumed that peaks reported as internally-branched monomethylalkanes include all possible methyl position isomers for that chain length. Specifically, “i.b.-meC29” was assumed to include 7, 9, 11, 13, and 15-meC29, and “i.b.-meC31” was assumed to include 7, 9, 11, 13, and

15 -meC31. This is a reasonable assumption, given that when an internally branched methylalkane peak is present, all such compounds were regularly found for the other paper wasp species in our dataset. Because this species is swarm-founding, including all possible positional isomers is conservative, as the pattern from our analyses is that swarm-founders have fewer compounds than do independent founders.

Latitude data

Because latitudinal clines are sometimes reported for cuticular hydrocarbon profile attributes, likely due to climate effects (eg (Frentiu and Chenoweth 2010)), we used Google Earth to gather latitude data for the collection location for specimens used in this analysis. Each species’ latitude was analyzed as the degrees of distance from the equator (ie all scores are positive, regardless of hemisphere).

Comparative methods

We analyzed species data with Phylogenetic Generalized Least Squares (PGLS) models (Grafen

1989; Freckleton et al. 2002; Revell 2010) using the R package ‘caper’ (Orme et al. 2013).

168 These models simultaneously fit a parameter # that scales the branch lengths of a phylogenic covariance matrix, determining the role of phylogeny in explaining trait covariance (Freckleton et al. 2002; Revell 2010). These models effectively determine the degree to which phylogenetic correction is necessary for a dataset; when # is zero, PGLS reduces to an ordinary least squares regression. When # is maximized at one, the analysis is equivalent to an independent contrasts analysis. Importantly, the appropriate # for a model should be estimated from the residuals of the covariance of predictor and response trait values, not the phylogenetic signature in the individual traits themselves (Revell 2010).

We used corrected Akaike Information Criteria (AICc) to select among models with different combinations of predictors (Burnham et al. 2011). Specifically, four models were compared with different combinations of the predictors Social and Latitude. We calculated model weights using the package qpcR (Ritz and Spiess 2008). Weights represent the relative probability of each model, given the data (Burnham et al. 2011).

We first investigated the effects of these parameters on the diversity of CHCs in species’ profiles. We ran analyses on the total number of different CHC compounds present (‘Total’), the number of linear alkanes present (‘Linear’), and the number of CHCs other than linear alkanes

(‘Non-linear’; note that for each species, Total = Linear + Non-linear).

To further test the hypothesis that the waterproofing properties of different compound types shape patterns in CHC profiles, we looked at how our predictors affected the relative abundance of linear alkanes, which are thought to be the most effective class of CHCs for waterproofing (Gibbs and Rajpurohit 2010). To calculate the relative abundance of a compound type, we summed the average relative abundance of peaks with a single type (linear alkane, monomethyl alkane, n-alkene, etc), and split the relative abundance of peaks containing

169 compounds of different classes evenly among the classes present in that peak. We predicted that

the relative abundance of linear alkanes would decrease with increasing absolute latitude, given

their superior waterproofing properties. Finally, although chain length is thought to be less

important than compound class in determining water retention properties of the cuticle (Gibbs

and Rajpurohit 2010), we tested the hypothesis that chain length increases in warmer climates (ie

lower latitudes). Models of alkane relative abundance failed to converge on a maximum

likelihood value for lambda when run on 8 trees from the first 100 trees selected randomly. We

inspected these trees but could not find any obvious commonalities among them. These trees

were removed from the set and additional randomly selected trees were included to make 100

trees. These same 100 trees were then used for analyses of all responses.

Assumptions of statistical models

To check for possible collinearity among predictors, we calculated variance inflation factors

(VIFs) according to Mundry (Mundry 2014). The VIF calculation gives an acceptably low value

of 1.35 for latitude and social. We checked the normality of the residuals visually with Q-Q

plots, and we checked the homogeneity of residual variance with plots of the residuals vs fitted

values. Given the large number of models resulting from our use of 100 phylogenies, we

checked all models for three random phylogenies. A few models had Q-Q plots suggesting that the residuals deviated from normality, but with a sample size of 34, this is often the case even for similar data with a normal distribution (KJL unpublished simulations). All models had residuals that were acceptably homoscedastic, which is the more important assumption for these tests

(Mundry 2014).

170 REFERENCES

Ay N, Flack J, Krakauer DC (2007) Robustness and complexity co-constructed in multimodal signalling networks. 362:441–447.

Bagnères A-G, Cristina Lorenzi M (2010) Chemical deception/mimicry using cuticular hydrocarbons. In: Blomquist GJ, Bagnères A-G (eds) Insect Hydrocarbons: Biology, Biochemistry, and Chemical Ecology. Cambridge University Press, pp 282–324

Baracchi D, Petrocelli I, Chittka L, et al (2015) Speed and accuracy in nest-mate recognition: a hover wasp prioritizes face recognition over colony odour cues to minimize intrusion by outsiders. Proc Biol Sci 282:20142750–20142750.

Baracchi D, Petrocelli I, Cusseau G, et al (2013) Facial markings in the hover wasps: quality signals and familiar recognition cues in two species of Stenogastrinae. Anim Behav 85:203– 212.

Blomquist GJ, Bagneres AG (eds) (2010) Insect Hydrocarbons: Biology, Biochemistry, and Chemical Ecology. Cambridge University Press

Boomsma JJ, Nielsen J, Sundström L, et al (2003) Informational constraints on optimal sex allocation in ants. Proc Natl Acad Sci USA 100:8799–8804.

Bos N, Dreier S, Jørgensen CG, et al (2012) Learning and perceptual similarity among cuticular hydrocarbons in ants. Journal of Insect Physiology 58:138–146.

Brandt M, van Wilgenburg E, Sulc R, et al (2009) The scent of supercolonies: the discovery, synthesis and behavioural verification of ant colony recognition cues. BMC Biol 7:71.

Breed MD, Buchwald R (2008) Cue diversity and social recognition. In: Gadau J, Fewell JH (eds) Organization of insect societies. Harvard University Press, pp 173–194

Bruschini C, Cervo R, Turillazzi S (2010) Pheromones in social wasps. Vitam Horm 83:447–492.

Buck M, Cobb TP, Stahlhut JK, Hanner RH (2012) Unravelling cryptic species diversity in eastern Nearctic paper wasps, Polistes (Fuscopolistes), using male genitalia, morphometrics and DNA barcoding, with descriptions of two new species (Hymenoptera: Vespidae). Zootaxa 3502:1–48.

Burnham KP, Anderson DR, Huyvaert KP (2011) AIC model selection and multimodel inference in behavioral ecology: some background, observations, and comparisons. Behav Ecol Sociobiol 65:23–35.

Capella-Gutiérrez S, Silla-Martínez JM, Gabaldón T (2009) trimAl: a tool for automated alignment trimming in large-scale phylogenetic analyses. Bioinformatics 25:1972–1973.

Carlson DA, Bernier UR, Sutton BD (1998) Elution patterns from capillary GC for methyl-

171 branched alkanes. J Chem Ecol 24:1845–1865.

Cervo R (2006) Polistes wasps and their social parasites: an overview. Ann Zool Fenn 43:531– 549.

Chaline N, Sandoz J-C, Martin SJ, et al (2005) Learning and discrimination of individual cuticular hydrocarbons by honeybees (Apis mellifera). Chemical Senses 30:327–335.

Chamero P, Marton TF, Logan DW, et al (2007) Identification of protein pheromones that promote aggressive behaviour. Nature 450:899–902.

Chapman RF, Espelie KE, Sword GA (1995) Use of cuticular lipids in grasshopper taxonomy: a study of variation in Schistocerca shoshone (Thomas). Biochem Sys Ecol 23:383–398.

D'Ettorre P, Heinze J (2005) Individual recognition in ant queens. Curr Biol 15:2170–2174.

Dani F (2006) Cuticular lipids as semiochemicals in paper wasps and other social insects. Ann Zool Fenn 43:500–514.

Dani F, Jones G, Destri S, et al (2001) Deciphering the recognition signature within the cuticular chemical profile of paper wasps. Anim Behav 62:165–171.

Dani FR, Cervo R (1992) Reproductive strategies following nest loss in Polistes gallicus (L.) (Hymenoptera Vespidae). Ethol Ecol Evol 4:49–53.

Dani FR, Jones GR, Corsi S, et al (2005) Nestmate recognition cues in the honey bee: differential importance of cuticular alkanes and alkenes. Chemical Senses 30:477–489.

Dapporto L, Bruschini C, Cervo R, et al (2010) Hydrocarbon rank signatures correlate with differential oophagy and dominance behaviour in Polistes dominulus foundresses. Journal of Experimental Biology 213:453.

Dapporto L, Fondelli L, Turillazzi S (2006) Nestmate recognition and identification of cuticular hydrocarbons composition in the swarm founding paper wasp Ropalidia opifex. Biochem Sys Ecol 34:617–625.

Drummond AJ, Suchard MA, Xie D, Rambaut A (2012) Bayesian phylogenetics with BEAUti and the BEAST 1.7. Mol Biol Evol 29:1969–1973.

Edgar RC (2004) MUSCLE: multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Res 32:1792–1797.

Espelie KE, Hermann HR (1988) Congruent cuticular hydrocarbons: Biochemical convergence of a social wasp, an ant and a host plant. Biochem Sys Ecol 16:505–508.

Freckleton RP, Harvey PH, Pagel M (2002) Phylogenetic analysis and comparative data: a test and review of evidence. Am Nat 160:712–726.

Frentiu FD, Chenoweth SF (2010) Clines in cuticular hydrocarbons in two Drosophila species

172 with independent population histories. Evolution 64:1784–1794. d

Gibbs AG (1998) Water-proofing properties of cuticular lipids. Amer Zool 38:471–482.

Gibbs AG, Rajpurohit S (2010) Cuticular lipids and water balance. In: Blomquist GJ, Bagnères A-G (eds) Insect Hydrocarbons: Biology, Biochemistry, and Chemical Ecology. Cambridge University Press, pp 100–120

Grafen A (1989) The phylogenetic regression. Phil Trans R Soc B 326:119–157.

Guerrieri FJ, Nehring V, Jorgensen C, et al (2009) Ants recognize foes and not friends. Proc R Soc Lond B 276:2461–2468.

Henshaw MT, Strassmann JE, Queller DC (2001) Swarm-founding in the polistine wasps: the importance of finding many microsatellite loci in studies of adaptation. Mol Ecol 10:185– 191.

Howard RW, Blomquist GJ (2005) Ecological, behavioral, and biochemical aspects of insect hydrocarbons. Annu Rev Entomol 50:371–393.

Ito Y, Itioka T (2008) Demography of the Okinawan eusocial wasp Ropalidia fasciata (Hymenoptera: Vespidae) II. Effects of foundress group size on survival rates of colonies and foundresses, and production of progeny. Entomol Sci 11:17–30.

Jeanne RL (2003) Social complexity in the Hymenoptera, with special attention to the wasps. In: Kikuchi T, Azuma N, Higashi S (eds) Genes, Behaviors and Evolution of Social Insects. Hokkaido University Press, Sapporo, pp 81–130

Jeanne RL (1980) Evolution of social behavior in the Vespidae. Annu Rev Entomol 25:371–396.

Jeanne RL (1991) The Swarm-founding Polistinae. In: Ross KG, Matthews RW (eds) The Social Biology of Wasps. Cornell University Press, Ithaca, NY, pp 191–231

Kapheim KM, Pan H, Li C, et al (2015) Genomic signatures of evolutionary transitions from solitary to group living. Science 348:1139–1143.

Kelstrup HC, Hartfelder K, Nascimento FS, Riddiford LM (2014a) Reproductive status, endocrine physiology and chemical signaling in the Neotropical, swarm-founding eusocial wasp, Polybia micans. Journal of Experimental Biology 217:2399–2410.

Kelstrup HC, Hartfelder K, Nascimento FS, Riddiford LM (2014b) The role of juvenile hormone in dominance behavior, reproduction and cuticular pheromone signaling in the caste-flexible epiponine wasp, Synoeca surinama. Frontiers in Zoology 11:78.

Klahn J (1988) Intraspecific comb usurpation in the social wasp Polistes fuscatus. Behav Ecol Sociobiol 23:1–8.

Le Conte Y, Hefetz A (2007) Primer pheromones in social hymenoptera. Annu Rev Entomol

173 53:523.

Leadbeater E, Dapporto L, Turillazzi S, Field J (2014) Available kin recognition cues may explain why wasp behavior reflects relatedness to nest mates. Behav Ecol 25:344–351.

Liebig J, Monnin T, Turillazzi S (2005) Direct assessment of queen quality and lack of worker suppression in a paper wasp. Proc R Soc Lond B 272:1339–1344.

Liebig J, Peeters C, Oldham N, et al (2000) Are variations in cuticular hydrocarbons of queens and workers a reliable signal of fertility in the ant Harpegnathos saltator? Proc Natl Acad Sci USA 97:4124–4131.

Lorenzi MC, Azzani L, Bagneres AG (2014) Evolutionary consequences of deception: Complexity and informational content of colony signature are favored by social parasitism. Current Zoology 60:137–148.

Lorenzi MC, Bagneres AG, Clément JL, Turillazzi S (1997) Polistes biglumis bimaculatus epicuticular hydrocarbons and nestmate recognition (Hymenoptera, Vespidae). Insectes Soc 44:123–138.

Makino S (1989) Usurpation and nest rebuilding in Polistes riparius: Two ways to reproduce after the loss of the original nest (Hymenoptera: Vespidae). Insectes Soc 36:116–128.

Martin SJ, Drijfhout F (2009a) A Review of Ant Cuticular Hydrocarbons. J Chem Ecol 35:1151– 1161.

Martin SJ, Drijfhout F (2009b) How reliable is the analysis of complex cuticular hydrocarbon profiles by multivariate statistical methods? J Chem Ecol 35:375–382.

Martin SJ, Helantera H, Drijfhout FP (2011) Is parasite pressure a driver of chemical cue diversity in ants? Proc R Soc Lond B 278:496–503.

Martin SJ, Vitikainen E, Helantera H, Drijfhout F (2008) Chemical basis of nest-mate discrimination in the ant Formica exsecta. Proc R Soc Lond B 275:1271.

Martin SJ, Zhong W, Drijfhout FP (2009) Long!term stability of hornet cuticular hydrocarbons facilitates chemotaxonomy using museum specimens. Biol J Linnean Soc 96:732–737.

Menzel F, Schmitt T (2012) Tolerance requires the right smell: first evidence for interspecific selection on chemical recognition cues. Evolution 66:896–904.

Moore D, Liebig J (2010) Mixed messages: fertility signaling interferes with nestmate recognition in the monogynous ant Camponotus floridanus. Behav Ecol Sociobiol 64:1011– 1018.

Mundry R (2014) Statistical Issues and Assumptions of Phylogenetic Generalized Least Squares. In: Garamszegi LZ (ed) Modern phylogenetic comparative methods and their application in evolutionary biology. Springer, Heidelberg, pp 131–153

174 Oi CA, Van Zweden JS, Oliveira RC, et al (2015) The origin and evolution of social insect queen pheromones: Novel hypotheses and outstanding problems. Bioessays 37:808–821.

Orme D, Freckleton R, Thomas G, et al (2013) caper: Comparative Analyses of Phylogenetics and Evolution in R. R package version 0.5.2 2:

Page M, Nelson LJ, Haverty MI, Blomquist GJ (1990) Cuticular hydrocarbons of eight species of north american cone beetles, Conophthorus Hopkins. J Chem Ecol 16:1173–1198.

Paradis E, Claude J, Strimmer K (2004) APE: Analyses of Phylogenetics and Evolution in R language. Bioinformatics 20:289–290.

Pardi L (1948) Dominance order in Polistes wasps. Physiological Zoology 21:1–13.

Pearse WD, Purvis A (2013) phyloGenerator: an automated phylogeny generation tool for ecologists. Methods Ecol Evol 4:692–698.

Pickett K, Carpenter JM (2010) Simultaneous analysis and the origin of eusociality in the Vespidae (Insecta: Hymenoptera). Arthropod Systematics & Phylogeny 68:3–33.

Queller DC, Zacchi F, Cervo R, et al (2000) Unrelated helpers in a social insect. Nature 405:784–787.

Ratnieks FLW (1991) The evolution of genetic odor-cue diversity in social Hymenoptera. Am Nat 137:202–226.

Reeve HK, Starks PT, Peters JM, Nonacs P (2000) Genetic support for the evolutionary theory of reproductive transactions in social wasps. Proc R Soc Lond B 267:75–79.

Revell LJ (2010) Phylogenetic signal and linear regression on species data. Methods Ecol Evol 1:319–329.

Revell LJ (2011) phytools: an R package for phylogenetic comparative biology (and other things). Methods Ecol Evol 3:217–223.

Richards OW (1978) The social wasps of the Americas excluding the Vespinae. 1–296.

Ritz C, Spiess AN (2008) qpcR: an R package for sigmoidal model selection in quantitative real- time polymerase chain reaction analysis. Bioinformatics 24:1549–1551.

Saito-Morooka F (2014) Double origin of swarm-founding in the genus Ropalidia? Congress of the International Union for the Study of Social Insects IUSSI

Smith AA, Holldobler B, Liebig J (2009) Cuticular hydrocarbons reliably identify cheaters and allow enforcement of altruism in a social insect. Curr Biol 19:78–81.

Strassmann JE (1991) Costs and benefits of colony aggregation in the social wasp, Polistes annularis. Behav Ecol 2:204–209.

175 Sukumaran J, Holder MT (2010) DendroPy: a Python library for phylogenetic computing. Bioinformatics 26:1569–1571.

Tannure-Nascimento IC, Nascimento FS, Zucchi R (2008) The look of royalty: visual and odour signals of reproductive status in a paper wasp. Proc R Soc Lond B 275:2555–2561.

Tibbetts EA (2004) Complex social behaviour can select for variability in visual features: a case study in Polistes wasps. Proc R Soc Lond B 271:1955–1960.

Tibbetts EA (2002) Visual signals of individual identity in the wasp Polistes fuscatus. Proc R Soc Lond B 269:1423–1428.

Tibbetts EA, Dale J (2007) Individual recognition: it is good to be different. TREE 22:529–537.

Tibbetts EA, James Dale (2004) A socially enforced signal of quality in a paper wasp. Nature 432:218.

Tindo M, D'Agostino P, Francescato E, et al (1997) Associative colony foundation in the tropical wasp Belonogaster juncea juncea (Vespidae, Polistinae). Insectes Soc 44:365–377.

Van Oystaeyen A, Oliveira RC, Holman L, et al (2014) Conserved class of queen pheromones stops social insect workers from reproducing. Science 343:287–290.

Van Wilgenburg E, Felden A, Choe D-H, et al (2012) Learning and discrimination of cuticular hydrocarbons in a social insect. Biol Lett 8:17–20.

Van Wilgenburg E, Symonds MRE, Elgar MA (2011) Evolution of cuticular hydrocarbon diversity in ants. J Evol Biol 24:1188–1198.

Wilson AJ, Dean M, Higham JP (2013) A game theoretic approach to multimodal communication. Behav Ecol Sociobiol 67:1399–1415.

Wyatt TD (2010) Pheromones and signature mixtures: defining species-wide signals and variable cues for identity in both invertebrates and vertebrates. J Comp Physiol A 196:685–700.

176 Appendix 6.1. Supplementary figures and tables

monomethylalkanes dimethylalkanes 40 30 20 20 10 0 0 umber of compounds umber of compounds n n IF SF IF SF

trimethylalkanes QïDONHQHV 8 # !" 6

8 $%&'&()* 4 4 2 0 0 umber of compounds umber of compounds n n IF SF IF SF Figure S6.1. CHC profile properties of 34 paper wasp species across social types and latitude. Diversity measures are the number of compounds of various types present in the profile of each species. Statistical tests (Table 6.1) were performed on the total number of compounds and the total number of non-linear compounds, which include these four compound classes, as well as the uncommon methylalkenes (not shown).

177

n-alkane average chain length relative abundance

# !" 34

80 $%&'&()* 30 40 carbons % of total CHC 0 26

IF SF IF SF

Figure S6.2. Alkane relative abundance and average chain length, as a function of social type and latitude. Relative abundance of linear alkanes is the fraction of the entire CHC total ion chromatogram area that was found in linear alkane peaks. Chain lengths were weighted by the relative abundance of each compound.

178

Figure S6.3. Extraction of CHCs from a pinned specimen of Ropalidia marginata. The specimen was lowered into hexane using a micromanipulator. The vial is partially filled with glass beads, reducing the volume of hexane required to bring liquid level up to a depth allowing immersion of specimen and pin.

179

1947 (pinned from museum collection)

2013 (frozen)

15.00 16.00 17.00 18.00 19.00 20.00 21.00 22.00 23.00 24.00 25.00 26.00 27.00 28.00 29.00 30.00 31.00 32.00 33.00 Retention time Figure S6.4. GC-MS traces of two Vespula maculifrons (Vespidae: Vespinae) worker CHC profiles. These specimens were both collected in Ithaca, NY. The specimen for the upper GC- MS trace was collected in 1947 and stored on a pin in the Cornell University Insect Collection under typical museum conditions before extraction by immersion in hexane in 2013. The 2013 sample was collected and frozen immediately, and stored at -20 until extraction by immersion in hexane in 2013.

180

Mischocyttarus pallidipectus Mischocyttarus

Mischocyttarus lecointeilecointei

Mischocyttarus phthisicus Mischocyttarus

Mischocyttarus carinulatusMischocyttarus weyrauchi Mischocyttarus lemoulti Mischocyttarusalfkenii MischocyttarusspKMP Mischocyttarus drewsenigigas Mischocyttarus melanarius Mischocyttarus immarginatus

Apoica pallida Apoica pallens Apoica strigata Apoica thoracica

Agelaia Apoicamultipicta arborea Polybia [Trichinothorax] sericea

Mischocyttarus flavitarsis Mischocyttarus mexicanus Mischocyttarus mexicanuscubicola Mischocyttarus mastigophorus Agelaia pallipes Mischocyttarus paraguayensis Mischocyttarus cearensis Polybia [Trichinothorax] affinis Mischocyttarus nrcollarellus Mischocyttarus injucundus Agelaia spKMP Mischocyttarus bertonii Polybia [Trichinothorax] flavitincta Mischocyttarus tolensis Mischocyttarus latior Mischocyttarus deceptus Polybia [Trichinothorax] emaciata Mischocyttarus punctatus

Metapolybia aztecoides Belonogaster petiolata Metapolybia cingulata Belonogaster somereni Belonogaster junceajuncea Belonogaster nrjuncea Belonogaster junceacolonialis Polybia jurinei Parapolybia varia Ropalidia opifex Polybia raui Ropalidia latebalteata Polybia occidentalisnigratella Polybioides melainusRopalidia spBR Ropalidia fasciata Ropalidia socialista Polybia striata Ropalidia spEA Ropalidia plebeiana Polybia scrobalis Ropalidia romandicabeti Ropalidia spKMP Polybia fastidiosuscula Vespula maculifrons Polistes gallicus Polybia occidentalisoccidentalis Polistes dominula Polistes nimpha Polybia ruficepsruficeps Polistes biglumis Polybia belemensis Polistes marginalis Polistes stigmabernardii Polybia velutina Polistes sagittarius Polistes japonicus Protopolybia emortualis Polistes snelleni Protopolybia sedula Polistes cinerascens Polistes testaceicolor Polistes bicolor Protopolybia scutellaris Polistes occipitalis Protopolybia spKMP Polistes geminatusgeminatus Polistes actaeon Protonectarina sylveirae Polistes pacificus Polistes cavapyta Protopolybia exiguaexigua Polistes lanio Polistes comanchusnavajoe Epipona niger Polistes annularis Brachygastra augusti Polistes satan Polistes goeldii Polistes apicalis Polistes biguttatus Brachygastra mellifica Polistes buyssoni Polistes simillimus Polistes melanotus Polistes canadensis Polistescrinituscrinitus Polistes crinitusamericanus Polistes Polistes instabilis Polistes Brachygastra lecheguana Polistes majormajor PolistesPolistes perplexus apachus

PolistesPolistes bellicosus carolina

Polistes metricus Polistes aurifer ParachartergusParachartergus colobopterus fraternus Polistes fuscatus

Polistes rothneyi

Polistes jokahamae Polistes exclamans

Polistes poeyihaitiensis Polistes tenebricosus

Polistes erythrocephalus Polistes dorsaliscalifornicus

Figure S6.5. Constraint tree used to generate trees. This topology (based on the more comprehensive analysis of Pickett and Carpenter 2010) was used to force monophyly for particular clades in BEAST runs.

181

Mischocyttarus_lecointeilecointei

Mischocyttarus_carinulatus Mischocyttarus_punctatus Mischocyttarus_lemoulti

Mischocyttarus_injucundus

Mischocyttarus_tolensis Mischocyttarus_immarginatus

Mischocyttarus_deceptusMischocyttarus_bertonii Mischocyttarus_latior

Parachartergus_colobopterus Parachartergus_fraternus

Metapolybia_cingulata Metapolybia_aztecoides Synoeca_septentrionalis

Mischocyttarus_weyrauchi Mischocyttarus_nrcollarellus Mischocyttarus_mastigophorus Mischocyttarus_cearensis Mischocyttarus_paraguayensis Mischocyttarus_flavitarsis Mischocyttarus_pallidipectus Polybia_flavitinctaPolybia_sericea Mischocyttarus_mexicanus Mischocyttarus_mexicanuscubicola Mischocyttarus_alfkenii Mischocyttarus_phthisicus Polybia_affinis Polybia_emaciata Mischocyttarus_spKMP Mischocyttarus_drewsenigigas Mischocyttarus_melanarius Epipona_niger Parapolybia_varia Polybia_scrobalis Belonogaster_junceacolonialis Belonogaster_nrjuncea Belonogaster_petiolata Polybia_occidentalisnigratellaPolybia_raui Polybia_striata Belonogaster_junceajuncea Belonogaster_somereni Polybioides_melainus Polybia_ruficepsruficepsPolybia_jurinei Ropalidia_opifex Polybia_occidentalisoccidentalis Ropalidia_socialista Ropalidia_spEA Polybia_fastidiosuscula Ropalidia_latebalteata Ropalidia_plebeiana Polybia_belemensis Ropalidia_romandicabeti Ropalidia_spKMP Polybia_velutina Ropalidia_fasciata Brachygastra_mellifica Ropalidia_spBR Brachygastra_lecheguana Vespula_maculifrons Polistes_nimpha Brachygastra_augusti Polistes_dominula Protonectarina_sylveirae Polistes_gallicus Protopolybia_spKMP Polistes_biglumis Polistes_marginalis Protopolybia_scutellaris Polistes_erythrocephalus Polistes_stigmabernardii Protopolybia_exiguaexigua Polistes_sagittarius Protopolybia_emortualis Polistes_japonicus Protopolybia_sedula Polistes_snelleni Agelaia_spKMP Polistes_tenebricosus Agelaia_pallipes Polistes_actaeon Polistes_pacificus Agelaia_multipicta Polistes_cinerascens Apoica_thoracica Polistes_testaceicolor Apoica_strigata Polistes_bicolor Polistes_geminatusgeminatus Apoica_pallens Polistes_occipitalis Apoica_pallida Polistes_biguttatus Polistes_simillimus Polistes_canadensis Apoica_arborea Polistes_goeldii Polistes_annularis Polistes_comanchusnavajoe Polistes_cavapyta Polistes_rothneyi Polistes_lanio Polistes_apicalis Polistes_buyssoni Polistes_melanotus

Polistes_crinitusamericanus Polistes_satan

Polistes_jokahamae Polistes_fuscatusPolistes_aurifer

Polistes_metricus Polistes_carolina

Polistes_perplexus Polistes_bellicosusPolistes_apachus

Polistes_instabilis

Polistes_majormajor Polistes_exclamans

Polistes_poeyihaitiensis

Polistes_crinituscrinitus Polistes_dorsaliscalifornicus Figure S6.6. Consensus Tree (50%) from BEAST output. 123 species were used to create the trees that were then pruned. Several species lacking sequence data were inserted randomly within their genera (see methods section).

182

0.03) - (range) 0.03, (0.48,0.5) - (0.13,0.19) (0.48,0.48) (0.28,0.42) (0.07,0.07) (0.06,0.06) ------( (0.28,0.42) (0.18,0.23) 0.4 0.4 0.2 0.2 0.49 0.49 0.16 0.16 adjR2 adjR2 0.07 0.07 0.48 0.48 0.39 0.39 0.06 0.06 0.03 0.03 -

(range)

" (0.9,1) (0.87,1) (0.63,1) (0.37,1) (0.4,0.97) (0.79,0.96) (0.66,0.92) (0.78,0.99) (0.49,0.77) (0.27,0.46) (0.27,0.45) (0.47,0.79) 0.96 0.96 0.97 0.97 0.83 0.83 0.64 0.64 0.6 0.6 0.64 0.64 0.81 0.81 0.86 0.86 0.35 0.34 0.85 0.85 0.62 0.62 U95%CI

(range)

(0,0.32) (0,0.31) (0,0) (0,0) (0,0) (0,0) (0,0) (0.64,0.8) " (0.32,0.57) (0.59,0.84) (0.32,0.58) (0,0.03) 0 0 0 0 0 0 0 0 0 0 0.05 0.05 0.04 0.04 ML 0.71 0.71 0.42 0.42 0.73 0.73 0.42 0.42

0.5) 0.55) 0.53) 0.55) 0.32) 0.13) 0.34) -

------0.51, 0.66, 0.53, 0.66, 0.32, 0.13, 0.34, - - - (0.45,0.5) - - - - ( (0.4,0.4) (0.39,0.46) (0.36,0.36) (0.05,0.05) ------( ( ( ( ( ( (range) r r 0.4 0.4 Effect size Effect 0.48 0.48 0.51 0.51 0.43 0.43 0.36 0.36 0.05 0.32 0.32 0.53 0.53 0.64 0.13 0.34 0.64 0.64 ------

9.53) 16.59) 16.71) - 8.66) - - -

0.13,0) 0.16,0) - - 2.73,1.21) 34.19, ( ( 1.51,2.03) (mean CI) 39.73, 40.38, - - 33.54, (0.2,1.25) (0.3,1.34) - - - (

( (0.05,1.06) - (0.13,1.14) ------( ( ( ( 0.07 0.07 0.08 0.73 0.73 0.82 0.82 - - 0.76 0.76 0.55 0.55 0.64 0.26 0.26 21.1 21.1 - Slope 21.86 21.86 - 28.54 28.54 28.16 28.16 - - -

(range) (0,0.02) (0.03,0.1) (0.02,0.08) (0.22,0.29) (0.66,0.75) (0.13,0.16) (0.76,0.84) (0.25,0.25) (0.48,0.48) (0.08,0.08) (0.19,0.19) (0,0.01) 0 0 0.01 0.05 0.05 0.03 0.03 0.23 0.19 0.73 0.73 0.81 0.48 0.13 0.13 0.25 0.08 weight

0 0 0 0 0 0 0 0 0 100 100 100 % best

CHC diversity

(range)

(6.99,11) (0,0) (0,0) (0,0) (3.1,3.67) (8.92,12.06) (3.5,3.5) (4.21,7.63) (1.67,2.38) (3.99,6.79) (1.32,1.32) (1.79,1.79) 0 0 0 0 3.5 3.5 AICc 9.63 9.63 3.63 3.63 1.79 1.79 5.76 5.76 1.32 2.33 2.33 6.52 6.52 # 10.57 10.57

. Models of

AICc <2).

& 1 1 1 Social Social Social

Latitude Social + Social Latitude + Social + Social Latitude Latitude Latitude Latitude

predictor

CHCs CHCs linear CHCs

- non of Number linear of Number of number Total Table S6.1 Notes:model PGLS statistics mean are andvalues range from analyses on 100 trees. Lines in bold preferred are ( models

183 Table S6.2. Robustness analyses for " estimate: preferred models with " set to upper 95% CI value response predictor Slope (mean CI) Effect size r (range) " (range) adjR2 (range)

Social + -18.89 (-34.61,-3.18) -0.39 (-0.43,-0.32) Total number of CHCs 0.6 (0.47,0.79) 0.23 (0.16,0.26) Latitude 0.42 (-0.1,0.94) 0.27 (0.22,0.33)

Social + (-33.94,-2.07) (-0.43,-0.31) -18.00 -0.37 (0.49,0.77) (0.2,0.28) Number of non-linear CHCs 0.62 0.25 Latitude 0.53 (0,1.05) 0.33 (0.27,0.39)

1 ------0.35 (0.27,0.46) ---

Latitude -0.11 (-0.19,-0.03) -0.44 (-0.57,-0.38) 0.64 (0.4,0.97) 0.17 (0.12,0.3) Number of linear alkanes Social + -0.98 (-3.51,1.56) -0.13 (-0.46,-0.07) 0.64 (0.37,1) 0.16 (0.1,0.51) Latitude -0.12 (-0.2,-0.04) -0.46 (-0.68,-0.39) Notes: PGLS model statistics are mean and range values from analyses on 100 trees.

184

(range) (0.24,0.3) (0.4,0.4) (NA,NA) (NA,NA) (NA,NA) (0.12,0.24) (0.24,0.31) (0.37,0.37) (0.19,0.29) (0.07,0.09) (0.01,0.01) (0.21,0.21) 0.4 0.4 NA NA NA 0.28 0.28 0.16 0.16 0.23 0.23 0.07 0.01 0.29 0.29 0.37 0.21 adjR2 adjR2

(range)

" (0.88,1) (0.83,1) (0.68,0.9) (0.66,0.85) (0.46,0.71) (0.64,0.87) (0.46,0.69) (0.25,0.42) (0.54,0.95) (0.23,0.37) (0.37,0.82) (0.71,0.9) 0.93 0.93 0.95 0.95 0.8 0.8 0.78 0.78 0.74 0.74 0.57 0.57 0.57 0.57 0.76 0.76 0.34 0.79 0.31 U95%CI

(range)

(0,0.44) (0,0.15) (0,0.38) (0,0.12) (0,0) (0,0) (0,0) (0,0) (0,0) " (0.53,0.68) (0.49,0.69) (0,0.04) 0 0 0 0 0 0 0 0 0 0 0.03 0.03 0.22 0.22 0.18 0.18 0.04

ML 0.59 0.59 0.59 0.59

(range) 0.49) 0.43) 0.49) 0.31) 0.46) - - - - - 0.4) 0.2) 0.4) - - - 0.4, 0.2, 0.4, 0.55, 0.43, 0.54, 0.32, 0.46, ------(NA,NA) (NA,NA) ( ( ( (0.36,0.49) (0.32,0.32) (0.45,0.54) (0.39,0.39) --- ( ( ( ( ( 0.4 0.4 0.4 0.4 0.2 0.2 NA NA - - - 0.32 0.32 0.42 0.42 0.49 0.49 0.39 0.39 0.46 0.46 0.43 0.43 0.52 0.31 0.54 0.54 - - - - - Effect size r

4.26) 9.15) 2.78) 10.06) - - - 0.4) - - 0.03) -

0.13,0) - 3.68, 2.59,0.69) 29.36, 32.75, 26.76, ( 0.01,1.04) - (mean CI) 33.88, - - - - (0.3,1.29) 0.17, - ( - (

( (NA,NA) ( ( (NA,NA) (0.11,1.11) --- - ( ( (0.17,1.22) ( values from analyses on 100 trees. Lines in bold are preferred models models preferred bold are in Lines 100 trees. on analyses from values

0.06 0.06 NA NA 0.7 0.7 0.1 0.1 0.79 0.79 - 2.04 2.04 0.95 0.95 0.61 0.61 0.51 0.51 - - - Slope 14.77 14.77 16.81 16.81 20.95 20.95 21.97 21.97 - - - -

(range) (0.5,0.6) (0.02,0.06) (0.07,0.15) (0.01,0.03) (0.13,0.24) (0.12,0.14) (0.61,0.72) (0.08,0.08) (0.13,0.13) (0.05,0.05) (0.74,0.74) (0.28,0.32) 0.58 0.58 0.3 0.3 0.09 0.09 0.03 0.03 0.69 0.69 0.74 0.02 0.02 0.16 0.14 0.08 0.13 0.05 weight

0 0 0 0 0 0 0 0 0 100 100 100 % best

(range)

(0,0) (0,0) (0,0) (3.1,3.29) (5.5,5.5) (2.39,4.34) (6.35,8.59) (1.87,3.39) (4.54,4.54) (3.46,3.47) (4.29,7.38) (1.09,1.32) 0 0 0 0 5.5 5.5 AICc 1.3 1.3 6.2 6.2 3.25 3.25 7.66 7.66 3.01 4.54 3.47 3.85 3.85 #

. Models of CHC diversity (excluding 4 species the from literature)

1 1 1 Social Social Social

Social + Social Latitude + Social Latitude + Social Latitude Latitude Latitude Latitude

predictor

CHCs CHCs linear CHCs

AICc <

- non of Number linear of Number of number Total & Table S6.3 range and mean are statistics model PGLS Notes: (

185

(range) (NA,NA) (NA,NA) (NA,NA) (0.03,0.08) (0.25,0.34) (0.31,0.43) (0.05,0.11) (0.21,0.31) (0.31,0.42) (0.04,0.04) (0.02,0.02) (0.19,0.19) NA NA NA 0.37 0.37 0.08 0.08 0.26 0.04 0.02 0.37 0.37 0.19 adjR2 adjR2 0.06 0.06 0.29

(range)

" (0.77,0.9) (0.7,0.83) (0.5,0.92) (0.83,0.95) (0.62,0.76) (0.77,0.91) (0.72,0.88) (0.21,0.31) (0.34,0.67) (0.64,0.8) (0.84,0.97) (0.23,0.37) 0.7 0.7 0.9 0.9 0.3 0.75 0.75 0.72 0.72 0.83 0.83 0.88 0.88 0.45 0.45 0.84 0.84 0.79 0.26 0.68 0.68 U95%CI

(range)

(0.6,0.7) (0,0) (0,0) (0,0) (0,0) " (0.43,0.56) (0.07,0.28) (0.59,0.73) (0.42,0.56) (0.22,0.43) (0.07,0.27) (0.2,0.37) 0 0 0 0 0 0 0.3 0.3 0.65 0.65 ML 0.49 0.49 0.16 0.15 0.15 0.67 0.67 0.49 0.33

(range) 0.49) 0.48) 0.46) 0.26) 0.23) 0.43) 0.45) 0.52) ------0.6, 0.54, 0.57, 0.51, 0.26, 0.23, 0.43, 0.45, ------(0.2,0.27) ( (NA,NA) (NA,NA) (NA,NA) (0.23,0.33) (0.28,0.37) (0.25,0.31) ( ( ( ( ( ( ( NA NA NA 0.24 0.24 0.56 0.56 0.28 0.28 0.33 0.33 0.29 0.29 0.43 0.43 0.53 0.53 0.26 0.23 0.52 0.52 0.49 0.45 0.45 ------Effect size r

16.08) 17.79) 20.39) - - 13.26) 0.03) - - - 0.7) - 0.13,0.02) 3.03,0.57) 63.7, 4.5, 0.19, 0.11,1.68) 0.26,1.47) 0.15,1.59) 60.84, 62.24, (mean CI) ------58.14, ( (

(NA,NA) ( (NA,NA) (NA,NA) ( ( ( ( ( ( ( - (0.02,1.78) ( 2.6 2.6 NA NA NA 0.9 0.9 - 0.06 0.06 1.23 0.11 0.11 0.61 0.61 0.72 0.72 0.79 0.79 Slope - - - 35.7 35.7 42.04 42.04 40.02 40.02 38.46 38.46 - - - -

(range) (0.5,0.64) (0.4,0.56) (0.01,0.05) (0.01,0.03) (0.33,0.45) (0.01,0.06) (0.02,0.06) (0.39,0.53) (0.08,0.08) (0.08,0.08) (0.06,0.06) (0.79,0.79) 0.57 0.57 0.47 0.47 0.02 0.02 0.39 0.02 0.02 0.79 0.79 0.03 0.03 0.04 0.08 0.08 0.06 0.46 0.46 weight

0 0 0 0 0 0 0 0 47 53 100 100 % best

(range) AICc <2).

(0,0.59) (0,0.74) (0,0) (0,0) (5.5,8.19) (5.12,8.14) (0.21,1.28) (4.17,7.16) (3.96,6.46) (4.69,4.69) (4.66,4.66) (5.09,5.09) & 0 0 0 0 AICc 0.13 0.13 0.09 0.09 6.77 6.77 0.75 0.75 5.77 5.77 5.05 4.69 4.66 5.09 6.69 6.69 #

. modelsDiversity using a mean 0.1% relative peak abundance threshold

1 1 1 Social Social Social

Social + Social Latitude + Social Latitude + Social Latitude Latitude Latitude Latitude

predictor

CHCs CHCs linear CHCs

- non of Number linear of Number of number Total preferred models ( Table S6.4 bold are in Lines 100 trees. on analyses from values range and mean are model statistics PGLS Notes:

186

(range) (0,0.9) (0,0.09) 0.02,0.27) 0.01,0.14) 0.02,0.18) (0.02,0.1) ------( ( ( 0.03 0.03 0.05 0.05 0.07 0.07 adjR2 adjR2 0.09 0.09 0.06 0.06 0.09

(range)

" (0.7,1) (0.68,1) (0.52,1) (0.57,1) (0.87,1) (0.86,1) (0.82,1) (0.83,1) 0.95 0.95 0.95 0.95 0.95 0.95 0.93 0.93 0.91 0.98 0.99 0.98 U95%CI

(0,1) (0,1) (range)

(0.24,1) (0.21,1) (0.48,1) (0.42,1) (0.52,1) " (0.35,0.68) 0.37 0.37 0.57 0.7 0.7 0.64 0.64 0.74 0.74 0.62 ML 0.53 0.53

(range) 0.02) 0.12) 0.23) - - - 0.3,0.04) 0.41, - 0.54, 0.36, - ( - - ( (0.16,0.41) (0.07,0.35) (0.19,0.95) (0.08,0.32) ------( ( 0.24 0.24 0.17 0.17 0.34 0.34 0.25 0.24 0.14 0.29 0.29 0.31 0.31 - - - -

Effect size r

CI) 5.83,33.7) 1.06,0.08) 0.89,0.32) 2.64,0.09) 2.54,0.44) 0.01,33.79) - 0.01,0.08) 0.03,0.07) (mean - - - - - ( - -

( ( ( ( --- ( --- ( ( 1.27 1.27 0.49 0.49 0.29 1.05 0.03 0.03 0.02 Slope - - - - 13.93 13.93 16.89 16.89

(range) (0.1,0.2) (0.08,0.5) (0.24,0.5) (0.12,0.62) (0.08,0.32) (0.15,0.39) (0.11,0.5) (0.13,0.45) 0.2 0.2 0.15 0.15 0.2 0.2 0.38 0.38 0.24 0.24 0.17 0.17 0.39 0.39 0.26 weight

0 0 6 8 1 91 32 62 2). % best

(range) AICc <

(0,3.38) (0,4.01) (0,2.93) (0,2.29) (0,2.68) (0,1.19) (0.46,3.59) (1.37,2.68) & AICc 1.66 1.66 1.57 0.51 0.86 1.33 0.04 1.99 1.99 1.86 #

. Relative abundance and chain length models chain and abundance Relative .

1 1 Social Social

Social + Social + Social Latitude Latitude Latitude Latitude

predictor

alkane linear

length chain Average is that CHC of % preferred models ( Table S6.5 bold are in Lines 100 trees. on analyses from values range and mean are model statistics PGLS Notes:

187 Table S6.6: Sample summary Species Social Sample type Source Location Agelaia_cajennensis SF Frozen personal collection 10°25'40.14" N 84°00'15.98" W Belonogaster_juncea IF Pinned CUIC 3°21'50.12" N 16°22'40.80" E Belonogaster_petiolata IF Pinned CUIC 25°42'05.00" S 27°45'46.87" E Brachygastra_mellifica SF Frozen J. Strassmann 26°05'49.30" N 97°57'42.35" W Metapolybia_mesoamerica SF Frozen personal collection 10°25'40.14" N 84°00'15.98" W Mischocyttarus_immarginatus IF Pinned CUIC 10°48'50.28" N 85°36'31.90" W Mischocyttarus_mexicanus IF Frozen personal collection 30°40'56.82" N 83°13'24.30" W Mischocyttarus_pallidipectus IF Pinned UGCA 9°56'45.12" N 84°05'41.87" W Mischocyttarus_phthisicus IF Pinned CUIC 18°00'01.03" N 66°35'40.06" W Parachartergus_colobopterus SF Frozen J. Strassmann 10°10'24.72" N 67°37'18.26" W Parapolybia_nodosa IF Pinned UGCA 24°00'16.52" N 121°05'26.04" E Parapolybia_varia IF Pinned UGCA 24°00'16.52" N 121°05'26.04" E Polistes_annularis IF Frozen personal collection 32°16'42.93" N 81°04'45.06" W Polistes_biglumis IF literature Lorenzi et al. 1997 44°55'51.29" N 6°43'27.68" E Polistes_canadensis IF Pinned UGCA 10°48'50.28" N 85°36'31.90" W Polistes_dominula IF Frozen personal collection 42°25'49.67" N 76°30'11.13" W Polistes_erythrocephalus IF Pinned CUIC 9°56'45.12" N 84°05'41.87" W Polistes_exclamans IF Frozen personal collection 31°19'35.13" N 81°27'02.90" W Polistes_fuscatus IF Frozen personal collection 42°25'49.67" N 76°30'11.13" W Polistes_instabilis IF Pinned UGCA 10°48'50.28" N 85°36'31.90" W Polistes_major IF Frozen personal collection 31°19'35.13" N 81°27'02.90" W Polistes_metricus IF Frozen personal collection 32°16'42.93" N 81°04'45.06" W Polistes_rothneyi IF Pinned CUIC 22°01'13.91" N 121°34'01.15" E Polybia_micans SF literature Kelstrup et al. 2014a 10°58'20.67" S 37°11'05.77" W Polybia_occidentalis SF Frozen J. Strassmann 10°10'01.57" N 68°07'40.00" W Polybioides_melainus SF Pinned CUIC 1°28'31.84" S 29°18'22.26" E Polybioides_raphigastra SF Pinned AMNH & UGCA 1°43'03.42" N 103°54'23.67" E Protopolybia_exigua SF Frozen J. Strassmann 10°08'56.23" N 68°09'25.62" W Ropalidia_fasciata IF Pinned UGCA 9°24'26.41" N 118°33'19.81" E Ropalidia_marginata IF Pinned J. Hunt 12°57'53.72" N 77°35'32.53" E Ropalidia_montana SF Pinned AMNH & CUIC 11°21'09.30" N 76°46'20.11" E Ropalidia_opifex SF literature Dapporto et al. 2006 3°21'02.45" N 101°47'30.55" E Ropalidia_romandi SF Pinned AMNH 16°56'21.34" S 145°42'17.22" E Synoeca_septentrionalis SF literature Kelstrup et al. 2014b 10°58'20.67" S 37°11'05.77" W SF: Swarm-founding AMNH: American Museum of Natural History IF: Independent-founding UGCA: University of Georgia Collection of Arthropods CUIC: Cornell University Insect Collection

188 Table S6.7. Accession numbers for sequences used to create trees Species 12S 16S 28S COI Agelaia_multipicta - - - AY382247.1 Agelaia_pallipes GU596571.1 GU596686.1 GU596702.1 GU596948.1 Agelaia_spKMP - - - AY918915.1 Apoica_arborea GU596562.1 - - GU596893.1 Apoica_pallens - GU596614.1 GU596707.1 GU596895.1 Apoica_pallida - GU596615.1 GU596708.1 GU596894.1 Apoica_strigata GU596572.1 GU596616.1 GU596709.1 GU596897.1 Apoica_thoracica GU596573.1 GU596617.1 GU596710.1 GU596896.1 Belonogaster_junceacolonialis - - GU596711.1 GU596899.1 Belonogaster_junceajuncea - GU596618.1 GU596712.1 GU596848.1 Belonogaster_nrjuncea - - EF190742.1 - Belonogaster_petiolata - AF066910.1 AF066897.1 - Belonogaster_somereni - AF066937.1 AF066927.1 - Brachygastra_augusti - - - AY382253.1 Brachygastra_lecheguana - GU596619.1 GU596713.1 GU596900.1 Brachygastra_mellifica - - - AY382254.1 Epipona_niger - GU596621.1 - GU596905.1 Metapolybia_aztecoides GU596589.1 GU596622.1 GU596732.1 GU596903.1 Metapolybia_cingulata GU596575.1 GU596623.1 GU596733.1 GU596904.1 Mischocyttarus_alfkenii - HQ163804.1 - HQ163838.1 Mischocyttarus_bertonii - GU596625.1 GU596734.1 GU596906.1 Mischocyttarus_carinulatus GU596514.1 GU596626.1 GU596735.1 GU596852.1 Mischocyttarus_cearensis GU596515.1 GU596627.1 GU596736.1 GU596853.1 Mischocyttarus_deceptus GU596513.1 GU596624.1 GU596737.1 GU596851.1 Mischocyttarus_drewsenigigas GU596516.1 GU596628.1 GU596738.1 GU596907.1 Mischocyttarus_flavitarsis GU596563.1 GU596629.1 GU596739.1 GU596908.1 Mischocyttarus_immarginatus - - - AY382233.1 Mischocyttarus_injucundus - HQ163802.1 - - Mischocyttarus_latior GU596517.1 GU596630.1 GU596740.1 GU596909.1 Mischocyttarus_lecointeilecointei GU596518.1 GU596631.1 GU596741.1 GU596910.1 Mischocyttarus_lemoulti GU596519.1 GU596632.1 GU596742.1 GU596911.1 Mischocyttarus_mastigophorus GU596576.1 GU596633.1 - GU596912.1 Mischocyttarus_melanarius GU596590.1 GU596634.1 GU596743.1 GU596913.1 Mischocyttarus_mexicanus - - - HQ163837.1 Mischocyttarus_mexicanuscubicola GU596591.1 GU596635.1 GU596744.1 GU596914.1 Mischocyttarus_nrcollarellus - HQ163801.1 - HQ163835.1 Mischocyttarus_pallidipectus - - GU596745.1 GU596915.1 Mischocyttarus_paraguayensis - GU596636.1 GU596746.1 GU596916.1 Mischocyttarus_phthisicus - - - AY382237.1 Mischocyttarus_punctatus - GU596637.1 GU596747.1 GU596917.1 Mischocyttarus_spKMP - - - AY918910.1 Mischocyttarus_tolensis - - GU596748.1 GU596918.1 Mischocyttarus_weyrauchi GU596520.1 GU596695.1 GU596749.1 GU596919.1 Parachartergus_colobopterus - - - AY382249.1 Parachartergus_fraternus - - EF190762.1 AY382250.1 Parapolybia_varia - - - AY382240.1 Polistes_actaeon GU596577.1 GU596638.1 GU596753.1 GU596947.1 Polistes_annularis GU596524.1 GU596639.1 GU596754.1 GU596854.1 Polistes_apachus - - GU596755.1 GU596855.1 Polistes_apicalis GU596525.1 GU596640.1 GU596756.1 GU596856.1 Polistes_aurifer GU596578.1 GU596641.1 GU596757.1 GU596857.1 Polistes_bellicosus GU596569.1 GU596642.1 GU596758.1 GU596858.1 Polistes_bicolor GU596526.1 GU596643.1 GU596759.1 GU596920.1 Polistes_biglumis - - - GU596859.1 Polistes_biguttatus - - GU596760.1 GU596860.1 Polistes_buyssoni GU596527.1 GU596644.1 GU596761.1 GU596861.1 Polistes_canadensis - - GU596762.1 GU596862.1 Polistes_carolina GU596528.1 GU596645.1 GU596763.1 GU596864.1 Polistes_cavapyta GU596529.1 GU596646.1 GU596764.1 GU596865.1 Polistes_cinerascens GU596530.1 GU596647.1 GU596765.1 GU596866.1 Polistes_comanchusnavajoe GU596586.1 GU596648.1 GU596766.1 GU596867.1

189 Polistes_crinitusamericanus - GU596649.1 GU596767.1 GU596868.1 Polistes_crinituscrinitus GU596531.1 GU596650.1 GU596768.1 GU596869.1 Polistes_dominula GU596549.1 GU596651.1 GU596769.1 GU596870.1 Polistes_dorsaliscalifornicus GU596532.1 GU596652.1 GU596770.1 GU596871.1 Polistes_erythrocephalus - - - GU596872.1 Polistes_exclamans GU596533.1 GU596693.1 GU596771.1 GU596873.1 Polistes_fuscatus GU596534.1 GU596653.1 GU596772.1 GU596874.1 Polistes_gallicus GU596535.1 GU596654.1 GU596773.1 GU596875.1 Polistes_geminatusgeminatus GU596536.1 GU596655.1 GU596774.1 GU596876.1 Polistes_goeldii GU596564.1 GU596656.1 GU596775.1 GU596925.1 Polistes_instabilis - - - EF136441.1 Polistes_japonicus GU596566.1 GU596657.1 GU596776.1 GU596877.1 Polistes_jokahamae - - GU596777.1 GU596878.1 Polistes_lanio GU596537.1 GU596658.1 GU596778.1 GU596879.1 Polistes_majormajor - GU596659.1 GU596779.1 GU596880.1 Polistes_marginalis GU596538.1 GU596660.1 GU596780.1 GU596881.1 Polistes_melanotus GU596539.1 GU596661.1 GU596781.1 GU596921.1 Polistes_metricus GU596540.1 GU596662.1 GU596782.1 GU596882.1 Polistes_nimpha GU596541.1 GU596663.1 GU596783.1 GU596883.1 Polistes_occipitalis GU596542.1 GU596664.1 GU596784.1 GU596884.1 Polistes_pacificus GU596543.1 GU596665.1 GU596785.1 GU596885.1 Polistes_perplexus GU596544.1 GU596666.1 GU596786.1 GU596886.1 Polistes_poeyihaitiensis GU596592.1 GU596667.1 GU596787.1 GU596887.1 Polistes_rothneyi - AB284540.1 - AB969807.1 Polistes_sagittarius GU596545.1 GU596668.1 GU596788.1 GU596922.1 Polistes_satan - - KP255853.1 EF136455.1 Polistes_simillimus GU596579.1 GU596669.1 GU596789.1 GU596888.1 Polistes_snelleni GU596580.1 GU596670.1 GU596790.1 GU596889.1 Polistes_stigmabernardii GU596546.1 - GU596791.1 GU596890.1 Polistes_tenebricosus GU596547.1 GU596671.1 GU596792.1 GU596891.1 Polistes_testaceicolor GU596548.1 GU596672.1 GU596793.1 GU596892.1 Polybia_affinis GU596581.1 - AF142528.1 GU596923.1 Polybia_belemensis - GU596673.1 GU596794.1 GU596924.1 Polybia_emaciata GU596582.1 GU596674.1 GU596795.1 AY382257.1 Polybia_fastidiosuscula - GU596675.1 - GU596926.1 Polybia_flavitincta - GU596676.1 GU596796.1 GU596927.1 Polybia_jurinei - - GU596797.1 GU596928.1 Polybia_occidentalisnigratella GU596587.1 GU596677.1 GU596798.1 GU596930.1 Polybia_occidentalisoccidentalis - GU596678.1 GU596799.1 GU596931.1 Polybia_raui - GU596698.1 GU596800.1 GU596932.1 Polybia_ruficepsruficeps - - - GU596934.1 Polybia_scrobalis - - - AY382261.1 Polybia_sericea - GU596679.1 GU596801.1 GU596935.1 Polybia_striata GU596594.1 GU596680.1 - GU596936.1 Polybia_velutina - - GU596802.1 - Polybioides_melainus GU596583.1 GU596681.1 - GU596937.1 Protonectarina_sylveirae GU596584.1 GU596682.1 GU596803.1 GU596938.1 Protopolybia_emortualis - GU596683.1 GU596804.1 GU596939.1 Protopolybia_exiguaexigua GU596565.1 GU596684.1 - GU596940.1 Protopolybia_scutellaris GU596593.1 GU596685.1 GU596805.1 GU596941.1 Protopolybia_sedula - - - GU596942.1 Protopolybia_spKMP - - - AY918914.1 Ropalidia_fasciata - - - AB969808.1 Ropalidia_latebalteata - - - AY382241.1 Ropalidia_opifex - - EF190765.1 - Ropalidia_plebeiana - - JF510016.1 JF510007.1 Ropalidia_romandicabeti - - AF146654.1 AF146677.1 Ropalidia_socialista - - - AY382242.1 Ropalidia_spBR - - - KM054517.1 Ropalidia_spEA - - - AY382244.1 Ropalidia_spKMP - - - AY918909.1 Synoeca_septentrionalis - - - AY382263.1 Vespula_maculifrons KJ147196.1 AY206804.2 EF190738.1 GU207859.1

190