An Epigenetic Analysis of the Robustness of the Honeybee (Apis Mellifera) Queen Developmental Pathway

Molecular Ecology (2017) 26, 1598–1607 doi: 10.1111/mec.13990

Making a queen: an epigenetic analysis of the robustness of the honeybee (Apis mellifera) queen developmental pathway

XU JIANG HE,* LIN BIN ZHOU,* QI ZHONG PAN,* ANDREW B. BARRON,† WEI YU YAN* and ZHI JIANG ZENG* *Honeybee Research Institute, Jiangxi Agricultural University, Nanchang, Jiangxi 330045, China, †Department of Biological Sciences, Macquarie University, North Ryde, NSW 2109, Australia

Abstract Specialized castes are considered a key reason for the evolutionary and ecological success of the social insect lifestyle. The most essential caste distinction is between the fertile queen and the sterile workers. Honeybee (Apis mellifera) workers and queens are not genetically distinct, rather these different phenotypes are the result of epige- netically regulated divergent developmental pathways. This is an important phe- nomenon in understanding the evolution of social insect societies. Here, we studied the genomic regulation of the worker and queen developmental pathways, and the robustness of the pathways by transplanting eggs or young larvae to queen cells. Queens could be successfully reared from worker larvae transplanted up to 3 days age, but queens reared from older worker larvae had decreased queen body size and weight compared with queens from transplanted eggs. Gene expression analysis showed that queens raised from worker larvae differed from queens raised from eggs in the expression of genes involved in the immune system, caste differentiation, body development and longevity. DNA methylation levels were also higher in 3-day-old queen larvae raised from worker larvae compared with that raised from transplanted eggs identifying a possible mechanism stabilizing the two developmental paths. We propose that environmental (nutrition and space) changes induced by the commercial rearing practice result in a suboptimal queen phenotype via epigenetic processes, which may potentially contribute to the evolution of queen–worker dimorphism. This also has potentially contributed to the global increase in honeybee colony failure rates.

Keywords: DNA methylation, epigenetic analysis, gene expression, honeybee, immunity, queen Received 7 March 2016; revision received 12 December 2016; accepted 19 December 2016

(Queller & Strassmann 1998; Linksvayer & Wade 2005; Introduction Foster et al. 2006). Oster and Wilson have particularly The evolution of cooperation, cooperative living and emphasized the importance of caste in the evolution of animal societies has been an enduring subject of fasci- social insect societies (Oster & Wilson 1978). Different nation for evolutionary biologists (Wilson 1975). Key castes within the society specialize on different func- insights into the processes of social evolution have tions. This specialization promotes efﬁciencies, which come from studies of the advanced social insects (Eilson provides a key selective advantage to social living. 1971; Andersson 1984; Robinson 1999). These have Oster & Wilson (1978) argue castes are one key reason shaped our understanding of the genetic and ecological for the ecological success of the social insect lifestyle. factors that can promote the evolution of sociality Queens and workers are the deﬁning caste distinction for the social insects. Queens have multiple morphological and behavioural specializations for extreme fecun- Correspondence: Zhi Jiang Zeng, Fax: +86 791 83828176; dity, whereas workers show a similar degree of E-mail: [email protected]

© 2016 John Wiley & Sons Ltd EPIGENETIC CHANGES IN HONEYBEE QUEEN REARING 1599 specialization for social roles supporting the queens’ and space) changes induced by the commercial rearing reproduction, and in many social insects workers are practice may potentially affect queen development via sterile. This is the case for honeybees (Apis mellifera). A epigenetic processes. Here, we explored the conse- typical colony contains a single reproductive queen sup- quence of age of transplant from worker cells to queen ported by up to 50 000 sterile workers (Winston 1991). cells on DNA methylation, gene expression and queen Studies of the bee have shown how the distinction morphology. We found that the domestic rearing prac- between queens and workers is not genetic: rather these tice altered queen morphology and induced epigenetic two phenotypes are the outcome of different develop- changes in developing queens, which supports our mental pathways (Nijhout 2003; Linksvayer et al. 2011). hypothesis. Both queens and workers develop from fertilized eggs, but differences in nutrition and the amount of Materials and methods food given to young larvae trigger different epigeneti- cally regulated developmental pathways (Kucharski Three European honeybee colonies (Apis mellifera) each et al. 2008; Maleszka 2014; Maleszka et al. 2014). with a single drone inseminated queen (SDI) were used Kucharski et al. (2008) reported that nutritional differ- throughout this study. These colonies were maintained ences between queen and worker at their larval stage at the Honeybee Research Institute, Jiangxi Agricultural control their development via DNA methylation. Shi University, Nanchang, China (28.46 uN, 115.49 uE), et al. (2011) showed that the amount of space in which according to the standard beekeeping techniques. a larva can develop alters the DNA methylation level of the larval genome and contributes to the process of Queen-rearing methods caste differentiation. Changes in gene regulation caused by these epigenetic mechanisms then establish diver- Queens were restricted for 6 h to a plastic honeybee gent developmental paths (Simola et al. 2013), involving frame developed by Pan et al. (2013) for laying. The particularly genes involved in signal transduction, frame is designed such that the plastic base of this gland development and carbohydrate metabolism frame with eggs or larvae can be transferred to plastic (Woodard et al. 2011). queen cells directly (Pan et al. 2013). Eggs that queen Since the 19th century in commercial beekeeping, it laid in worker cells were transplanted into queen cells has been a standard practice to raise queens by trans- for rearing new queens when eggs were less than 6 h planting eggs or young larvae into artificial queen cells, old (QWE). For the other experimental groups, day 1, which triggers workers to raise a queen (Doolittle 1888; day 2 and day 3 worker larvae were transplanted into Buchler€ et al. 2013). Within the commercial queen-rear- queen cells for rearing QWL1, QWL2 and QWL3, ing practice, there is variation in the age at which eggs respectively. Queen cells with worker eggs or larvae or worker larvae are transplanted to queen cells to be were returned into their natal colonies (the SDI colo- raised as queens. It is not clear how well the honeybees’ nies) for queen rearing. developmental processes are able to tolerate this kind For the morphological measurements, new emerging of intervention. Woyke (1971) reported that rearing queens were collected and their weight measured using queens from young worker larvae resulted in decreased an analytical balance (FA3204B; Shanghai Precision Sci- body size, a smaller spermatheca and fewer ovarioles. entific Instrument Co., Ltd.). Their thorax width and Rangel et al. (2012) reported that colonies from queens length were measured with a zoom stereo microscope reared from older worker larvae had significantly lower system (Panasonic Co., Ltd.) according to the manufac- production of worker comb, drone comb and stored turer’s instructions. food compared with colonies from queens reared from For epigenetic analysis, we sampled 3-day-old larvae young worker larvae. In fact, concern over the long- from QWE, QWL1 and QWL2, respectively, from their term consequences of commercial queen rearing for bee queen cell. The fourth group QWL3 sampled 3-day-old stocks is not new. In 1923, Rudolf Steiner predicted that worker larvae directly from worker cells. Each sample honeybees would become extinct within 100 years as a group collected three larvae and there were three bio- consequence of commercial queen rearing progressively logical replicates, each from different colonies, for each weakening bee stocks (Thomas 1998). In the current group. We weighed each larva from these four treat- environment of increased honeybee colony failure rates, ment groups with an analytical balance. All samples mass deaths of colonies and declining honeybee stocks, were immediately flash-frozen in liquid nitrogen. The there is a great deal of concern as to whether a decline DNA and RNA from each sample were both extracted in queen bee quality might be a factor in these prob- for further DNA methylation and RNA sequencing lems (van Engelsdorp et al. 2010; Delaney et al. 2011). analysis. DNA and RNA were extracted from the same Therefore, we hypothesize that environmental (nutrition samples.

< RNA-Seq analysis KEGG protein database by BLAST (E-value 1e-5) and used KOBAS 2.0 software to test the statistical enrichment Total RNA was extracted from larvae according to the of differential expression genes in KEGG pathways (Xie standard protocol for the TRIzol reagent (Life technolo- et al. 2011). gies, California, USA). RNA integrity and concentration were checked using an Agilent 2100 Bioanalyzer (Agi- DNA Methylation analysis by bisuphite sequencing lent Technologies, Inc., Santa Clara, CA, USA). mRNA was isolated from total RNA using a NEB- The DNA of each larval sample was extracted using the Next Poly(A) mRNA Magnetic Isolation Module (NEB, Universal Genomic DNA Extraction Kit (DV811A; E7490). A cDNA library was constructed following the TaKaRa). DNA concentration was measured and manufacturer’s instructions for the NEBNext Ultra RNA adjusted to the same level. Genomic DNA was sheared Library Prep Kit (NEB, E7530) and the NEBNext Multi- with Covaris ultrasonicator (Life Technology). The frag- plex Oligos (NEB, E7500) from Illumina. In brief, mented DNA was purified using AMPure XP beads enriched mRNA was fragmented into approximately and end-repaired. A single ‘A’ nucleotide was added to 200 nt RNA inserts, which were used as templates to the 30 ends of the blunt fragments followed by ligation synthesize the cDNA. End-repair/dA-tail and adaptor to methylated adapter with T overhang. 200- to 300-bp ligation were then performed on the double-stranded insert size targets were purified by 2% agarose gel elec- cDNA. Suitable fragments were isolated by Agencourt trophoresis. Bisulphite conversion was conducted using AMPure XP beads (Beckman Coulter, Inc.) and enriched a ZYMO EZ DNA Methylation-GoldTM Kit (ZYMO, by PCR amplification. Finally, the constructed cDNA Irvine, CA, USA). The final libraries were generated by libraries were sequenced on a flow cell using an Illu- PCR amplification. Bisulphite libraries were analysed mina HiSeqTM 2500 sequencing platform. by an Agilent2100 Bioanalyzer (Agilent Technology) Low-quality reads, such as adaptor-only reads or and quantified by QPCR (Agilent QPCR NGS Library reads with >5% unknown nucleotides, were filtered Quantification Kit). The construction of bisulphite from subsequent analyses. Reads with a sequencing er- libraries and paired-end sequencing using Illumina ror rate less than 1% (Q20 > 98%) were retained. These HiSeqTM 2500 (Illumina, San Diego, CA, USA) were remaining clean reads were mapped to the honeybee performed at Beijing Biomarker Technology Co., Ltd (A. mellifera) official genes (OGSv3.2) using TOPHAT2 (Beijing, China). (Kim et al. 2013) software. The aligned records from the After filtering adaptor sequences and PCR-duplicated aligners in BAM/SAM format were further examined to reads, genomic fragments from bisulphite libraries were remove potential duplicate molecules. Gene expression mapped against the honeybee genome (A. mellifera. levels were estimated using FPKM values (fragments Amel 4.5) using BOWTIE 2 software (Langmead & Salz- per kilobase of exon per million fragments mapped) by berg 2012). The bismark methylation extractor (Krueger the CUFFLINKS software (Trapnell et al. 2010). & Andrews 2011) was used to predict all methylation DESeq2 and Q-value statistical methods were used to sites. Only uniquely mapped reads were retained. The evaluate differential gene expression among the four ratio of C to CT was used to indicate methylation level. experimental treatments (Love et al. 2014). The false dis- Three methods for DNA methylation level analysis covery rate (FDR) control method was used to deter- were used: fraction of methylated cytosines, mean mine the appropriate threshold of P-values in multiple methylation level and weighted methylation level tests comparing gene expression differences by read (Schultz et al. 2012). The results are presented in Fig. S8 counts. Only genes with an absolute value of log2 ratio (Supporting information). ≥1 and FDR significance score <0.05 were used for subsequent analysis. However, gene expression levels of Data analysis each gene in all samples were presented using their ratio of FPKM values. Morphological analysis of queens and 3-day-old queen lar- Sequences differentially expressed between sample vae. All data from morphological experiments of each groups were identified by comparison against various group were analysed by ANOVA using STATVIEW 5.01 fol- protein databases by BLASTX, including the National Cen- lowed by a Fisher’s PLSD test (SAS Institute, Cary, NC, ter for Biotechnology Information (NCBI) nonredundant USA). protein (Nr) database, SWISS-PROT database with a cut- off E-value of 10 5. Furthermore, genes were searched Correlation analysis between expression and methylation and against the NCBI nonredundant nucleotide sequence (Nt) map construction of genes and chromosome. Methylated 5 database using BLASTN by a cut-off E-value of 10 . regions were deemed significantly differentially methy- Differentially expressed genes (DEGs) were mapped to lated across QWE, QWL1, QWL2 and QWL3 with a

© 2016 John Wiley & Sons Ltd EPIGENETIC CHANGES IN HONEYBEE QUEEN REARING 1601 false discovery rate (FDR) <0.05 and log2 fold change ≥1.5 in sequence counts using the BSmooth method in R package 3.1.1 (Hansen et al. 2012). Signiﬁcantly different methylated regions (DMRs) of each gene were mapped to the 16 honeybee chromosomes regions (A. mellifera. Amel 4.5) using integrative genomics viewer (IGV, http://www.broadinstitute.org/igv/).

Analysis of RNA-Seq quality and DNA methylation sequencing. In RNA-Seq, four libraries were generated from our experimental groups, and summaries of RNA sequencing analyses are shown in Table S1 (Supporting information). In each library, more than 98% clean reads were unique reads of which more than 89% reads were paired reads. Very few clean reads (<1.4%) were multiple mapped reads. Each library had a sufficient cover- age of the expected number of distinct genes (stabilized at 3M reads, Fig. S1, Supporting information). The Pear- son correlation coefficient among three biological replicates of each experimental group was all ≥0.80 (Table S2, Supporting information), a conventionally accepted threshold for valid replicates (Tarazona et al. 2011), indicating that there was acceptable sequencing quality and repeatability among the biological replicates of each group. The majority of methylation sites of all samples (77.55%) were the CG type, which was considerably more Fig. 1 (A) Mean (+SE) weight of new born queens and 3-day- than other two types (CHH: 20.5% and CHG: 1.95% old queen larvae, (B) Mean (+SE) thorax width and length of respectively; Fig. S2, Supporting information). new born queens, from QWE (open), QWL1 (grey), QWL2 (black) and QWL3 (diagonal stripes). Different letters above each bar indicate significant differences (P < 0.05, ANOVA test Results followed with Fisher’s PLSD test). As the age of transplant of the worker larvae increased, the size and mass of the emergent adult queen decreased CYP450 6a14-like and CYP450 305a1 were upregulated (Fig. 1). QWE had the highest thorax length (4.90 0.24 in QWL2 compared with QWE, whereas seven of ten mm, mean SE), thorax width (4.78 0.21 mm, body development-related genes were downregulated mean SE) and weight (267.21 2.49 mg, mean SE), in QWL2; In QWL3 and QWE comparison, 13 of 23 whereas the QWL3 had lowest, with 4.60 0.13 mm, immunity-related DEGs were upregulated in QWL3 4.45 0.26 mm and 226.00 2.82 mg, respectively. All while 29 of 41 body development-related genes were morphological indices were differed significantly across downregulated in QWL3, respectively. Interestingly, the the four treatments (P < 0.05, ANOVA test followed with hormone biosynthesis genes [vitellogenin precursor (Vg), Fisher’s PLSD test). juvenile hormone esterase precursor (JH), juvenile hormone RNA-Seq analyses comparing gene expression esterase-like (JH-like) and ecdysteroid-regulated 16 kDa pro- between QWE and the three larvae-transplanted groups tein-like] were upregulated in QWL3 compared with (QWL1, QWL2 and QWL3) showed that the number of QWE, whereas the major royal jelly protein 1 (MRJP1) differentially expressed genes increased as the age of was downregulated in QWL3 (Table S3, Supporting the transplanted worker larva increased (Fig. 2 and information). Similarly, the results of GO enrichment Fig. S3, Supporting information). In all comparisons, the analysis showed that the number of categories of DEGs differentially expressed genes contained a high propor- enriched between the QWE group and the other experi- tion of genes involved in immunity, body development, mental groups increased with the increasing age of the metabolism, reproductive ability and longevity (Fig. 2 grafted larva: from 27 categories (QWL1 vs. QWE) to 33 and Table S3, Supporting information). In particular, (QWL2 vs. QWE) and 44 (QWL3 vs. QWE), respectively one of cytochrome P450 family gene (CYP450 6a14-like) (Fig. S4–S6, Supporting information). Categories of DEG was significantly upregulated in QWL1 compared included growth, development process, reproductive QWE; Six of ten immunity-related DEGs such as process and immune system process. The results of

COG enrichment analysis showed a similar pattern in (Table S5, Supporting information). A correlation analy- that the number of categories between QWE and QWLs sis of gene expression and DNA methylation developed increased with their grafting age (Fig. S7, Supporting by (Lou et al. 2014) (Fig. S9, Supporting information) information). suggested a very weak correlation between DNA methy- Furthermore, queens from older grafted worker larvae lation and gene expression in honeybees for all compar- had a higher global DNA methylation level than QWE isons (Pearson correlation values were 0.0018, 0.0016, (Fig. S8 and Table S4, Supporting information). The 0.004 and 0.0034 in QWE, QWL1, QWL2 and QWL3, QWE vs. QWL3 (totally 146 DMRs) comparison had the respectively, and all P-values were >0.05). We also com- greatest number of differentially methylated regions pared the expression of DGEs involved in immune sys- (DMRs) than QWE vs. QWL2 (108) and QWE vs. QWL1 tem and hormone biosynthesis and their DNA (99) comparisons (Fig. 4). Mapping these DMRs to gene methylation among the four treatment groups. The dif- regions identiﬁed 2, 6 and 23 differentially methylated ference in expression of these genes increased as the dif- genes in the QWE vs. QWL1, QWE vs. QWL2 and QWE ference in age of the transplanted larvae increased vs. QWL3 comparisons, respectively, and no gene was (Fig. 3) (greater when comparing QWE vs. QWL3 than overlapped among these three comparisons. These genes when comparing QWE vs. QWL1), but the DNA methy- were different from the DEGs identiﬁed by RNA-Seq. lation of very few of these genes showed a clear nega- Most of them were involved in substance metabolism tive correlation with expression (Fig. 3 and Table S6, Supporting information).

Discussion Honeybee workers and queens are two very different phenotypes that come about as a result of divergent developmental pathways. Environmental differences in growing space and nutrition cause the divergence, and the two pathways are organized by epigenetic processes (Haydak 1970; Shi et al. 2011; Foret et al. 2012). Here, we explored the mechanisms and stability of the queen developmental path by transplanting worker larvae of different ages into queen cells causing a redirection of the worker developmental path onto the queen developmental path. Our results speak to both the plasticity and limita- tions of honey bee phenotypic plasticity. While it is known that worker larvae that are more than 3.5 days old when transplanted fail to develop into queens (Weaver 1966), here we found that transplanting even 3-day-old worker larvae to queen cells resulted in functional adult queens, but with reduced size and weight. Moreover, our results showed that the number of differentially expressed genes in QWLs compared with QWE (Fig. 2) increased with the age of the transplanted larvae. Many of these genes were involved in immunity, body development, metabolism, reproductive ability and longevity (Fig. 2 and Table S3, Supporting information). These are all functions that differentiate queens from workers and are critical to the quality and longev- Fig. 2 Signiﬁcantly differentially expressed genes in three com- ity of the queen. We also showed that queens raised parisons. Genes were identiﬁed as differentially expressed if from transplanted older worker larvae had a higher glo- < both the FDR 0.05, and the absolute value of the log2 fold bal genomic DNA methylation level when compared to change was ≥1. Genes involved in immunity (yellow), body QWE (Table S4, Supporting information) identifying a development (green), reproduction or longevity (purple) and other functions (open bars) are shown. Top bars represent possible epigenetic mechanism for these differences. In upregulated genes in QWE compared with other groups. summary, we could conclude that the queen develop- Lower bars are downregulated genes. mental pathway is quite robust. The queen phenotype

Fig. 3 Expression and DNA methylation of 38 immunity- and hormone-related genes among QWE, QWL1, QWL2 and QWL3. These genes were identiﬁed by their functions in immunity or hormone biosynthesis and were at least signiﬁ- cantly differentially expressed between one comparison of QWE and QWLs. The ratio of gene expression in QWLs against QWE was used for presenting the expression level of each gene. Green indicates downregulation in QWLs compared with QWE, red indicates upregula- tion, and black indicates no difference. Left side is the gene ID and gene function, middle is the ratio of gene expression of each gene, and right is the ratio of DNA methylation for each gene. More detailed information of these 38 refers to Table S6 (Supporting information).

could be attained even by 3-day-old worker larvae (the downregulated in QWL3 compared with QWE. Hence, larval developmental period is just 5 days long) and this identifies a candidate pathway for why transplant therefore developmental trajectory is certainly not fixed age alters adult queen ovariole number and spermath- until relatively late in larval development. However, eca size (Woyke 1971). The gene MRJP-1 (downregu- interfering with the normal developmental process by lated in QWL3 compared with QWE), JH (upregulated switching larvae from a worker to a queen path clearly in QWL3) and Vg (upregulated in QWL3) were also dif- had consequences for the resulting adults that were ferentially expressed between QWE and QWL3. These detectable at morphological and genetic levels. genes are also involved in the regulation of honeybee Our study identified many genetic and epigenetic caste differentiation and longevity (Amdam & Omholt changes related to the age at which worker larvae were 2002; Kamakura 2011), perhaps indicating why queens transplanted to queen cells to be raised as queens from late-stage larval grafts are undersized. The GO (Figs 2 and 4 and Table S3 and S4, Supporting informa- enrichment results also confirmed this result that DEGs tion). Of note, the differentially expressed genes between QWE and QWLs were enriched in growth, included insulin-like peptide A chain (ILP-A), a gene reproductive and development processes (Fig. S4–S6, involved in the mTOR pathway. The mTOR pathway is Supporting information). involved in queen ovary development, and caste differ- Previous studies have demonstrated that DNA entiation (Patel et al. 2007; de Azevedo & Hartfelder methylation is widespread in social Hymenoptera 2008; Mutti et al. 2011). The ILP-A was significantly (Kronforst et al. 2008; Kucharski et al. 2008; Lyko &

Fig. 4 Distribution of signiﬁcantly differentially methylated regions (DMRs) from three comparisons for the 16 honeybee chromosomes. DMRs were deemed sig- niﬁcantly differentially methylated across QWE, QWL1, QWL2 and QWL3 with a false discovery rate (FDR) <0.05 and log2 fold change ≥1.5 in sequence counts. The DMRs of QWE/QWL1, QWE/QWL2 and QWE/QWL3 comparisons are presented from outer to inner, respectively. Red plots are upregulated DMRs in QWE compared with other three groups, whereas green plots are downregulated ones. Chromosome name and scale are indicated on the outer rim.

Maleszka 2011; Foret et al. 2012). While the level of differentiation (Wang et al. 2006; Gabor Miklos & Mal- DNA methylation in honeybees is quite low compared eszka 2011; Shi et al. 2013; Maleszka 2014). Our results with mammals, differential methylation of the genome lend further credence to the view that the functions of has a key role in establishing the divergent worker and DNA methylation of sites in gene body regions in queen developmental pathways (Wang et al. 2006; hymenoptera have more complex functions than simply Gabor Miklos & Maleszka 2011; Shi et al. 2013; Mal- inhibiting expression. eszka 2014). We found that increasing the age at which In commercial apiculture, rearing queens from trans- worker larvae were transplanted to queen cells resulted planted worker larvae is a standard commercial prac- in an increasing number of differentially methylated tice, and the age of the worker larvae used can be regions of the genome. We propose this reﬂects the epi- anything up to and including 3 days old. In practice, genetic processes underlying the reorientation of a there is a preference to use older worker larvae for worker-destined developmental pathway to a queen transplant as these are more hardy and easier to handle developmental pathway. and give a higher success rate. But these larvae will There was not a strong negative correlation between have been fed worker jelly (brood food) rather than the gene expression and DNA methylation in this study queen jelly in their early life. Worker jelly is a very dif- (Fig. S9, Supporting information); however, in honey- ferent diet to queen jelly. It differs in sugar content bees most methylation sites are located in gene body (Asencot & Lensky 1977), amino acid (Brouwers 1984), regions rather than the upstream and downstream regu- vitamin (Brouwers et al. 1987), juvenile hormone (Asen- latory regions (Fig. S8, Supporting information). Foret cot & Lensky 1984) and major royal jelly protein content et al. (2012) showed that honeybee DNA methylation is (Kamakura 2011). Therefore, the commercial queen-rear- correlated with gene alternative splicing. Table S7, Sup- ing practice alters the nutritional environment of the porting information showed that the alternative splicing queen larvae, likely resulting in development and epi- also exists in differentially methylated genes. It is still genetic changes. This is consistent with the previous unclear what role these DNA methylation have for studies that nutritional differences control the caste dif- gene expression, although DNA methylation plays an ferentiation of queen and workers by DNA methylation important role in the regulation of honeybee caste (Kucharski et al. 2008; Shi et al. 2011).

There have been long-standing concerns about the References consequences of this queen-rearing method for queen Amdam GV, Omholt SW (2002) The regulatory anatomy of quality and colony productivity (Woyke 1971; Thomas honeybee lifespan. Journal of Theoretical Biology, 216, 209–228. 1998; Rangel et al. 2012). Our results have clearly shown Andersson M (1984) The evolution of eusociality. Annual that queens raised from older worker larvae are smal- Review of Ecology and Systematics, 15, 165–189. ler, and it is also telling that we found a difference in Asencot M, Lensky Y (1977) The effect of sugar crystals in expression between several genes involved in immune stored royal jelly and juvenile hormone on the differentiation function between QWE and QWLs (Fig. 2, Table S3, of female honey bee larvae (Apis mellifera L.) to queens. In: – and Fig. S4–S6, Supporting information for GO enrich- Proc. 8th Int. Congr. pp. 3 5. IUSSI, Wageningen. Asencot M, Lensky Y (1984) Juvenile hormone induction of ment results). These included the Cytochrome P450 fam- ‘queenliness’ on female honey bee (Apis mellifera L.) larvae ily (CYP450s, 11 of 15 were upregulated in QWL3 reared on worker jelly and on stored royal jelly. Comparative compared with QWE). CYP450s may contribute to both Biochemistry and Physiology B-Biochemistry & Molecular Biology, disease and insecticide resistance in queen honeybees 1, 109–117. (Claudianos et al. 2006; Boncristiani et al. 2012). We pro- de Azevedo SV, Hartfelder K (2008) The insulin signaling path- – pose that queens reared from older larvae may suffer way in honey bee (Apis mellifera) caste development differ- reduced immune function and insecticide resistance. ential expression of insulin-like peptides and insulin receptors in queen and worker larvae. Journal of Insect Physi- Evolutionary theories on the development of eusocial- ology, 54, 1064–1071. ity demonstrated that maternal care (nutrition and Bird A (2002) DNA methylation patterns and epigenetic mem- developmental space) dramatically contribute to the ory. Genes & Development, 16,6–21. evolution of queen–worker dimorphism in honeybees Boncristiani H, Underwood R, Schwarz R, Evans JD, Pettis J, (Linksvayer et al. 2011; Leimar et al. 2012). Our study van Engelsdorp D (2012) Direct effect of acaricides on patho- clearly showed that the domestic rearing practice artifi- gen loads and gene expression levels in honey bees Apis mel- – cially transformed the nutrition and developmental lifera. Journal of Insect Physiology, 58, 613 620. Brouwers E (1984) Glucose/fructose ratio in the food of honey- space of queen larvae and resulted in a partial inter- bee larvae during caste differentiation. Journal of Apicultural caste between queen and workers. In Leimar’s model Research, 2,94–101. (2012), honeybee caste dimorphism is produced by Brouwers EVM, Ebert R, Beetsma J (1987) Behavioural and maternal care rather than a switch-controlled polyphen- physiological aspects of nurse bees in relation to the compo- ism, and our results are consistent with this model. We sition of larval food during caste differentiation in the – also demonstrated that the domestic rearing practice honeybee. Journal of Apicultural Research, 1,11 23. € altered natural honeybee maternal behaviour, inducing Buchler R, Andonov S, Bienefeld K et al. (2013) Standard methods for rearing and selection of Apis mellifera queens. Journal various epigenetic changes. This is particularly interest- of Apicultural Research, 52,1–30. ing as epigenetic changes such as DNA methylation can Claudianos C, Ranson H, Johnson RM et al. (2006) A deficit of introduce environmental effects into the following detoxification enzymes: pesticide sensitivity and environ- generations (Bird 2002; Klironomos et al. 2013) and mental response in the honeybee. Insect Molecular Biology, 15, potentially influence evolutionary processes (Dickins & 615–636. Rahman 2012). If it is possible for epigenetic markers of Delaney DA, Keller JJ, Caren JR, Tarpy DR (2011) The physical, the genome to persist through gamete formation and insemination, and reproductive quality of honey bee queens (Apis mellifera L.). Apidologie, 42,1–13. operate transgenerationally in honeybees, as occurs in Dickins TE, Rahman Q (2012) The extended evolutionary some mammals (Klironomos et al. 2013), then the evolu- synthesis and the role of soft inheritance in evolution. Pro- tion of queen–worker dimorphism might be influenced ceedings of the Royal Society B-Biological Sciences, 279, 2913– by these mechanisms. Moreover, we propose that the 2921. often used commercial queen-rearing practice results in Doolittle GM (1888) Scientific Queen-rearing. American Bee Jour- queens of lower quality. As a proximal remedy, rearing nal, Hamilton, IL, USA. queens from eggs or very young larvae may yield a better Eilson E (1971) The Insect Societies. Harvard University Press, Cambridge, Massachusetts. outcome for queen performance and colony function. van Engelsdorp D, Hayes J Jr, Underwood RM, Pettis JS (2010) A survey of honey bee colony losses in the United States, fall – Acknowledgements 2008 to spring 2009. Journal of Apicultural Research, 49,7 14. Foret S, Kucharski R, Pellegrini M et al. (2012) DNA methyla- We thank Dr. Ying Wang for suggesting on experimental tion dynamics, metabolic fluxes, gene splicing, and alterna- design, Mr. Xue Chuan Zhang and Li Sha Huang for helping tive phenotypes in honey bees. Proceeding of the National on epigenetic data analysis. This work was supported by the Academy of Science of the United Statesof America, 109, 4968– National Natural Science Foundation of China (No. 31572469 4973. and No. 31460641) and the Earmarked Fund for China Agricul- Foster KR, Wenseleers T, Ratnieks FL (2006) Kin selection is ture Research System (CARS-45-KXJ12). the key to altruism. Trends in Ecology & Evolution, 21,57–60.

Gabor Miklos GL, Maleszka R (2011) Epigenomic communica- Nijhout HF (2003) Evelopment and evolution of adaptive poly- tion systems in humans and honey bees: from molecules to phenisms. Evolution and development, 5,9–18. behavior. Hormones and Behavior, 59, 399–406. Oster G, Wilson E (1978) Caste and Ecology in the Social Insects. Hansen KD, Langmead B, Irizarry RA (2012) BSmooth: from Princeton University Press, Princeton, New Jersey. whole genome bisulfite sequencing reads to differentially Pan QZ, Wu XB, Guan C, Zeng ZJ (2013) A new method of methylated regions. Genome Biology, 13, R83. queen rearing without grafting larvae. American Bee Journal, Haydak MH (1970) Honey bee nutrition. Annual Review of Ento- 153, 1279–1280. mology, 15, 143–156. Patel A, Fondrk MK, Kaftanoglu O et al. (2007) The making of Kamakura M (2011) Royalactin induces queen differentiation a queen: TOR pathway is a key player in diphenic caste in honeybees. Nature, 473, 478–483. development. PLoS One, 2, e509. Kim D, Pertea G, Trapnell C, Pimentel H, Kelley R, Salzberg Queller D, Strassmann J (1998) Kin Selection and Social SL (2013) TopHat2: accurate alignment of transcriptomes in Insects Social insects provide the most surprising predic- the presence of insertions, deletions and gene fusions. Gen- tions and satisfying tests of kin selection. BioScience, 3, ome Biology, 14, R36. 165–175. Klironomos FD, Berg J, Collins S (2013) How epigenetic muta- Rangel J, Keller JJ, Tarpy DR (2012) The effects of honey bee tions can affect genetic evolution: model and mechanism. (Apis mellifera L.) queen reproductive potential on colony BioEssays, 35, 571–578. growth. Insectes Sociaux, 60,65–73. Kronforst MR, Gilley DC, Strassmann JE, Queller DC (2008) Robinson GE (1999) Integrative animal behaviour and socioge- DNA methylation is widespread across social Hymenoptera. nomics. Trends in Ecology & Evolution, 14, 202–205. Current Biology, 18, R287–R288. Schultz MD, Schmitz RJ, Ecker JR (2012) ‘Leveling’ the playing Krueger F, Andrews SR (2011) Bismark: a flexible aligner and field for analyses of single-base resolution DNA methylomes. methylation caller for Bisulfite-Seq applications. Bioinformat- Trends in Genetics, 28, 583–585. ics, 27, 1571–1572. Shi YY, Huang ZY, Zeng ZJ, Wang ZL, Wu XB, Yan WY (2011) Kucharski R, Maleszka J, Foret S, Maleszka R (2008) Nutri- Diet and cell size both affect queen-worker differentiation tional control of reproductive status in honeybees via DNA through DNA methylation in honey bees (Apis mellifera, Api- methylation. Science, 319, 1827–1830. dae). PLoS One, 6, e18808. Langmead B, Salzberg SL (2012) Fast gapped-read alignment Shi YY, Yan WY, Huang ZY, Wang ZL, Wu XB, Zeng ZJ (2013) with Bowtie 2. Nature Methods, 9, 357–359. Genomewide analysis indicates that queen larvae have lower Leimar O, Hartfelder K, Laubichler MD, Page RJ (2012) methylation levels in the honey bee (Apis mellifera). Naturwis- Development and evolution of caste dimorphism in honey- senschaften, 100, 193–197. bees – a modeling approach. Ecology and Evolution, 2, Simola DF, Wissler L, Donahue G et al. (2013) Social insect gen- 3098–3109. omes exhibit dramatic evolution in gene composition and Linksvayer TA, Wade MJ (2005) The evolutionary origin and regulation while preserving regulatory features linked to elaboration of sociality in the aculeate Hymenoptera: mater- sociality. Genome Research, 23, 1235–1247. nal effects, sib-social effects, and heterochrony. Quarterly Tarazona S, Garcıa-Alcalde F, Dopazo J, Ferrer A, Conesa A Review of Biology, 80, 317–336. (2011) Differential expression in RNA-seq: a matter of depth. Linksvayer TA, Kaftanoglu O, Akyol E, Blatch S, Amdam GV, Genome research, 21, 2213–2223. Page RE Jr (2011) Larval and nurse worker control of Thomas B (1998) Bees-lecturers by Rudolf Steiner, pp. 222. developmental plasticity and the evolution of honey bee Anthroposophic Press, Great Barrington, MA, USA. queen–worker dimorphism. Journal of evolutionary biology, 9, Trapnell C, Williams BA, Pertea G et al. (2010) Transcript 1939–1948. assembly and quantification by RNA-Seq reveals unanno- Lou S, Lee HM, Qin H et al. (2014) Whole-genome bisulfite tated transcripts and isoform switching during cell differenti- sequencing of multiple individuals reveals complementary ation. Nature Biotechnology, 28, 511–515. roles of promoter and gene body methylation in transcrip- Wang Y, Jorda M, Jones PL et al. (2006) Functional CpG methy- tional regulation. Genome Biology, 15, 408. lation system in a social insect. Science, 314, 645–647. Love M, Anders S, Huber W (2014) Differential analysis of Weaver N (1966) Physiology of caste differentiation. Annual count data-the DESeq2 package. Genome Biology, 15, 550. review of entomology, 11,79–102. Lyko F, Maleszka R (2011) Insects as innovative models for Wilson E (1975) Sociobiology: The New Synthesis. Harvard functional studies of DNA methylation. Trends in Genetics, University Press, Cambridge, Massachusetts. 27, 127–131. Winston M (1991) The Biology of the Honey Bee. Harvard Univer- Maleszka R (2014) Epigenetic integration of environmental and sity Press, Cambridge, MA, USA. genomic signals in honey bees: the critical interplay of Woodard SH, Fischman BJ, Venkat A et al. (2011) Genes nutritional, brain and reproductive networks. Epigenetics, 3, involved in convergent evolution of eusociality in bees. Pro- 188–192. ceeding of the National Academy of Science of the United Statesof Maleszka R, Mason PH, Barron AB (2014) Epigenomics and the America, 108, 7472–7477. concept of degeneracy in biological systems. Briefings in Woyke J (1971) Correlations between the age at which honeybee Functional Genomics, 13, 191–202. brood was grafted, characteristics of the resultant queens, and Mutti NS, Dolezal AG, Wolschin F, Mutti JS, Gill KS, Amdam results of insemination. Journal of Apicultural Research, 10,45–55. GV (2011) IRS and TOR nutrient-signaling pathways act via Xie C, Mao X, Huang J et al. (2011) KOBAS 2.0: a web server juvenile hormone to influence honey bee caste fate. Journal of for annotation and identification of enriched pathways and Experimental Biology, 214, 3977–3984. diseases. Nucleic Acids Research, 39, W316–W322.

Z.J.Z. designed the experiments; X.J.H., L.B.Z. and Fig. S4 Gene ontology classification of DEGs between QWE Q.Z.P. performed the experiments; X.J.H., A.B.B. and and QWL1. W.Y.Y. analysed the data; and X.J.H., A.B.B. and Z.J.Z. Fig. S5 Gene ontology classification of DEGs between QWE written the paper. We have declared that no conflict of and QWL2. interests exist. Fig. S6 Gene ontology classification of DEGs between QWE and QWL3.

Fig. S7 COG enrichment analysis of DEGs between QWE and Data accessibility QWLs. The raw data for queen morphology are available from Fig. S8 DNA methylation level of protein coding genes in four groups: QWE (blue), QWL1 (green), QWL2 (lavender) and the Dryad Digital Repository: https://doi.org/10.5061/ QWL3 (red). dryad.bg4t9. The raw Illumina sequencing data are acces- sible through NCBI’s database: RNA-Seq and DNA meth Fig. S9 Correlation analysis of gene expression and DNA ylation data of QWE: NCBI Bioproject: PRJNA308280/ methylation. SAMN04390202. RNA-Seq and DNA methylation data of Table S1 Summary of DGE profiles and their mapping to the QWL1: NCBI Bioproject: PRJNA308280/SAMN04390203. reference genes. RNA-Seq and DNA methylation data of QWL2: NCBI Bioproject: PRJNA308280/SAMN04390236. RNA-Seq Table S2 Pearson correlation coefficient among three biological replicates of each group. and DNA methylation data of QWL3: NCBI Bioproject: PRJNA308280/SAMN04390201. Table S3 Significantly differentially expressed genes in three comparisons.

Table S4 Summary of DNA methylation sites of each sample. Supporting information Table S5 Differentially methylated genes between QWLs Additional supporting information may be found in the online ver- (QWL1, QWL2 and QWL3) and QWE. sion of this article. Table S6 Gene expression and DNA methylation of 38 selected Fig. S1 The saturation curve of RNA-Seq in each sample. genes in four treatments.

Fig. S2 Average distribution of three methylation types in Table S7 The Alternative splicing sites in differentially methy- 3-day-old queen larvae. lated genes between QWLs and QWE.

Fig. S3 Volcano plots of gene expression in three comparisons.