Introduction

Honey bees (genus Apis) are highly eusocial insects that have benefited humans for centuries1. Beekeeping, the culture of both Apis mellifera and Apis cerana, has been used widely to produce bee products2,3,4 and provide crop pollination5. Honey bees also help protect biodiversity6 and benefit human health7,8.

The biology of A. mellifera has been well studied9,10. Compared to A. mellifera, the biology of A. cerana in Asia is less studied11 and many questions remain. There are many differences between the two sister species of honey bees. A. cerana has smaller body sizes than A. mellifera, but is better at collecting sporadic nectar in mountainous or forested areas12. A. cerana also has natural resistance against Varroa destructor13, better ability to cope with extreme climate14, and the ability to fight against giant hornets such as Vespa mandarinia15. A. cerana does not collect propolis16 and also fan in their nest entrances in the opposite direction, toward the hive17 instead of drawing air out by A. mellifera. A. mellifera has the advantage of higher honey production, which is the main reason for its spread throughout the world including its introduction to China18. However, the largest difference, and one that impacts honey yield drastically, is that A. cerana has much smaller colonies19. This is partly because they swarm more often20, partly because their queen egg-laying capability is much lower, compared to that of A. mellifera.

The difference in queen egg-laying capacity between the two species is about 4-5 times. A. mellifera queens can lay 1500 to 2000 eggs per day21,22,23,24, while A. cerana queens lay only 420 to 600 eggs per day25,26,27. There are several possible mechanisms to cause this huge difference. One is that A. cerana queens have fewer ovarioles in their ovaries, with only 80-85% of the number in A. mellifera queens. However, the difference in ovariole count (across both ovaries) alone cannot fully explain the disparity in egg-laying capacity, because the ratio of ovarioles in queens of A. mellifera to A. cerana is only about 1.5 to 2.0: A. mellifera queens show an average of 300 ovarioles28 or a maximum of 40029, while A. cerana queens average of 20030. The possibility that A. mellifera queens are fed much more often and/or with more or better food compared to A. cerana queens is not investigated. While such data do exist for A. mellifera queens31,32, comparative data are missing in A. cerana. But even if the feeding rates are different, it still would require each ovariole to be more efficient, such that egg production per ovariole would be at least 2.5 times higher in A. mellifera than that in A. cerana.

We hypothesize that A. mellifera queen ovarioles are able to produce mature eggs faster than those of A. cerana. This would require differences in physiology, which can be tracked via differences in gene expression. In this study, we used ATAC-seq, RNA-seq, spatial transcriptomics, and cross-species sequence comparisons to compare the gene expression differences in the ovaries of virgin queens and laying queens between the two honey bee species. We identified agxt2l as a potential gene: its higher expression in trophocyte cells enhances glycerophospholipid metabolism, thereby accounting for the higher queen egg-laying rate in Apis mellifera compared to A. cerana.

Results

Difference in queen egg-laying power between A. mellifera and A. cerana

The number of eggs laid daily by A. mellifera queens was significantly higher than those in A. cerana (N = 10, Z = − 3.82, P < 0.01) (Fig. 1c). Body weight (N = 10, t = 4.32, df = 11.39, P < 0.01) (Fig. 1d) and the number of ovarioles (N = 10, t = 16.14, df = 18, P < 0.01) (Fig. 1e) in A. mellifera queens were significantly higher than those in A. cerana. The number of eggs laid per milligram of body weight per day (N = 10, t = 12.11, df = 18, P < 0.01) (Fig. 1f) and the number of eggs laid per ovariole per day in A. mellifera were significantly higher than those in A. cerana (N = 10, t = 20.80, df = 18, P < 0.01) (Fig. 1g). Details of statistical analyses are in Supplementary Data 1.

Fig. 1: Different egg-laying performance of queens in A. mellifera and A. cerana.
Fig. 1: Different egg-laying performance of queens in A. mellifera and A. cerana.
Full size image

a The newly laying queen in A. mellifera colony. b The newly laying queen in A. cerana colony. c The eggs laid daily of the newly laying queens of A. mellifera and A. cerana. d The body weight of the newly laying queens of A. mellifera and A. cerana. e The number of ovarioles in one ovary of A. mellifera and A. cerana. f The number of eggs laid per day per milligram of body weight. g The amount of eggs laid per ovariole per day in A. mellifera and A. cerana.

Sequence data quality control

In this study, ATAC-seq and RNA-Seq analyses were conducted for the virgin queens (A. mellifera, AmV; A. cerana, AcV) and the newly laying queens (about two months old) (A. mellifera, AmQ; A. cerana, AcQ). After quality control, each sample obtained an average of 6.52 Gb of clean data (43.48 million clean sequence reads) (Supplementary Data 2). The Q30 values of each sample in the two omics were greater than 90% (Supplementary Data 2). For spatial transcriptome, 2046 spot amounts were detected in the ovaries of A. mellifera queen (AmQ). For A. cerana queens (AcQ), 1625 spots were detected in the ovaries (Supplementary Data 2). These results showed that sequence quality was acceptable and the bioreplication was reliable.

DEGs between ovaries of A. mellifera and A. cerana queens

DEGs were identified based on two sets of reference genomes (the reference genomes of A. mellifera and A. cerana). Pearson correlation coefficients were > 0.95 for all A. mellifera bioreplicates using the A. mellifera reference genome (Supplementary Fig. 1a), and > 0.91 for all A. cerana bioreplicates using the A. cerana reference genome (Supplementary Fig. 1b).

Because the reference genomes were different between A. mellifera and A. cerana, the expression level of the same gene may be different, even if they were the same, if compared using different reference genomes. To avoid this problem, genes showing the same expression trend in a comparison using two reference genomes were selected. In the comparison AmV vs AcV, there were 8481 and 10,315 DEGs using the A. mellifera and the A. cerana reference genome, respectively (Fig. 2a, d). Among these genes, 5976 genes were homologs (Fig. 2b). Among these homologs, 1154 DEGs showed the same expression trend under the dual species genomes and most of them were up-regulated in the virgin queens (A. mellifera, Fig. 2e). In the comparison AmQ vs AcQ, there were 7341 and 9468 DEGs based on these two reference genomes respectively (Fig. 2a, d). Among these genes, 4821 genes were homologs (Fig. 2c). Among these homologs, 1543 DEGs showed the same expression trend in both bee species and most of them were up-regulated in the newly laying queens (A. cerana, Fig. 2f).

Fig. 2: Transcriptome data analysis results.
Fig. 2: Transcriptome data analysis results.
Full size image

a The numbers of DEGs and the up-down-regulation relationship in the two comparisons (AmV vs AcV, AmQ vs AcQ) using the A. mellifera reference genome. b The Venn diagram of overlapped DEGs between AmV vs AcV (using the A. mellifera reference genome, AmV-m vs AcV-m) and AmV vs AcV (using the A. cerana reference genome, AmV-c vs AcV-c). c The Venn diagram of overlapped DEGs between AmQ-m vs AcQ-m and AmQ-c vs AcQ-c. d The numbers of DEGs and the up-down-regulation relationship in the same two comparisons using the A. cerana reference genome. e The DEGs in AmV vs AcV that showed the same expression trend using either A. mellifera or A. cerana as a reference genome. f The DEGs in AmQ vs AcQ that showed the same expression trend using either A. mellifera or A. cerana as a reference genome.

AmV vs AcV common expression trend DEGs were enriched in Fatty acid degradation, mTOR, Glycerolipid metabolism and other pathways (Supplementary Fig. 2a). AmQ vs AcQ common expression trend DEGs were enriched in Wnt, mTOR, Noth, Glycerophospholipid metabolic pathway, and others (Supplementary Fig. 2b).

Differential open chromatin regions in queen ovaries between the two bee species

To further reveal the distribution of peaks in the genomes, 78,740 and 42,324 peaks in AmV-m (A. mellifera virgin queens using the A. mellifera reference genome) and AmQ-m (A. mellifera laying queens using the A. mellifera reference genome) respectively were annotated to the genome of A. mellifera (Fig. 3a, b), and there were 22,426 and 17,826 differential peaks in the comparisons AmV-m vs AcV-m and AmQ-m vs AcQ-m (Fig. 3c). While 95,834 and 55,140 peaks in AcV-c (A. cerana virgin queens using the A. cerana reference genome) and AcQ-c (A. cerana laying queens using the A. cerana reference genome) respectively were annotated to the genome of A. cerana (Fig. 3d, e). The peaks were mainly distributed in the promoter region and intron region, which accounted for more than 54% of the total peaks. There were 19,288 and 15,383 differential peaks in the comparisons AmV-c vs AcV-c and AmQ-c vs AcQ-c (Fig. 3f). There were more up-regulated peaks in A. mellifera (AmV, AmQ) under both reference genomes (Fig. 3c, f). The number of differential peaks distributed in intron and promoter regions were predominant (Fig. 3g, h).

Fig. 3: ATAC-seq data analysis results.
Fig. 3: ATAC-seq data analysis results.
Full size image

a, b The distribution of peaks in AmV and AmQ in the genome of A. mellifera. c The numbers of differential peaks and the up-down-regulation relationship in the two comparisons (AmV vs AcV, AmQ vs AcQ) using the A. mellifera reference genome. d, e The distribution of peaks in AcV and AcQ in the genome of A. cerana. f The number of differential peaks and the up-down-regulation relationship in the same two comparisons using the A. cerana reference genome. g The distribution of differential peaks in the three comparisons AmV-m vs AcV-m, and AmQ-m vs AcQ-m. h The distribution of differential peaks in the three comparisons AmV-c vs AcV-c and AmQ-c vs AcQ-c.

Combined analysis of ATAC-Seq and RNA-Seq

When comparing the differential peak-related genes with the DEGs, in AmV vs AcV, there were 6136 overlapped genes between ATAC-Seq and RNA-Seq, when using the A. mellifera reference genome. While there were 6604 overlapped genes between ATAC-Seq and RNA-Seq, when using the A. cerana reference genome. Between these two sets of DEGs, 4411 genes overlapped (Fig. 4a). Analyzing these overlapped genes with the consistent trend genes in Fig. 2e, it was found that 861 differential genes were shared (Fig. 4c). Of them, 88 genes were up-regulated and 84 genes were down-regulated in AmV in both ATAC-seq and RNA-seq under the A. mellifera reference genome (Fig. 4e), and the numbers were 131 and 85 under the A. cerana reference genome (Fig. 4g). Under the two reference genomes, there were 40 consistently up-regulated genes and 16 consistently down-regulated genes in AmV in both ATAC-seq and RNA-seq (Fig. 4i). AmV vs AcV common trend DEGs were enriched in the fatty acid metabolism pathway and other pathways (Fig. 4j).

Fig. 4: Analysis the result of ATAC-Seq and RNA-Seq.
Fig. 4: Analysis the result of ATAC-Seq and RNA-Seq.
Full size image

a, b Venn diagrams of DEGs and differential peak related genes in comparisons AmV vs AcV and AmQ vs AcQ using either A. mellifera or A. cerana as a reference genome. c, d Analyzing the common genes (a, b) with the consistent trend genes in Figs. 2e, f. e, f Using the A. mellifera reference genome, the RNA-seq and the ATAC-seq common DEGs of all comparison combinations in c, and d were statistically displayed. g, h Using the A. cerana reference genome, the RNA-seq and the ATAC-seq common DEGs of all comparison combinations in c and d were statistically displayed. i The panels e–h differential genes with the same trend in either reference genomes (the reference genome of A. mellifera or A. cerana). j, k The Kegg pathways of common trend DEGs.

In AmQ vs AcQ, there were 5016 and 5599 overlapped genes between ATAC-Seq and RNA-Seq, when using the A. mellifera and A. cerana reference genomes, respectively. Of these, 3263 DEGs overlapped (Fig. 4b). Analyzing these overlapped genes with the consistent trend genes in Fig. 2f, it was found that 1037 DEGs were shared (Fig. 4d). Of them, 100 genes were up-regulated and 172 genes were down-regulated under A. mellifera reference genome (Fig. 4f), and the numbers were 131 up-regulated genes and 127 down-regulated genes in AmQ in both ATAC-seq and RNA-seq under the A. cerana reference genome (Fig. 4h). Under the two reference genomes, there were 48 consistently up-regulated and 72 consistently down-regulated genes in AmQ in both ATAC-seq and RNA-seq (Fig. 4i). AmQ vs AcQ common trend DEGs were enriched on Wnt, mTOR and Glycerophospholipid metabolic pathway and other pathways (Fig. 4k).

Agxt2l, spr and nedd8 were essential for ovary development

To identify candidate genes that led to differences in queen egg-laying between A. mellifera and A. cerana, from the 48 consistently up-regulated and 72 consistently down-regulated genes in AmQ vs AcQ in both ATAC-seq and RNA-seq using the dual species genomes, we chose 49 genes that contained a differential peak in their promoter region (Fig. 5a) as candidates, and found three genes, agxt2l (LOC408817), spr (LOC725051) and nedd8 (LOC550745), were essential for ovary development. After RNAi of these three genes, their expression levels (Mann-Whitney U, P < 0.05), the number of ovarioles (Mann-Whitney U, P < 0.05) and body weight in the RNAi group (Mann–Whitney U, P < 0.05 and T-test, P < 0.05) (Fig. 5i–s) were all significantly reduced compared with the control group, except that the body weight showed no significant changes after RNAi of nedd8 (T-test, t = −1.78, df = 38, P > 0.05) (Fig. 5t) (Supplementary Data 3).

Fig. 5: Screening and validation of candidate genes.
Fig. 5: Screening and validation of candidate genes.
Full size image

a The 49 genes that contain a differential peak in their promoter region. b–e Bees ovaries in the control group and RNAi agxt2l, spr and nedd8. f–h agxt2l, spr and nedd8 differed between AmQ and AcQ in all two omics assays. i–k Expression of agxt2l, spr and nedd8 in queen ovaries of both species. l–n Expression of agxt2l, spr and nedd8 in ovaries in RNAi injected queens and control. o–t Body weight and number of ovarioles in A. mellifera injected with siRNA of agxt2l, spr, nedd8 and the control. u The glycerophospholipid metabolic pathway. Genes marked in red represent genes that have differences in gene expression be in AmQ vs AcQ. (we determined one side ovary because the other side was used for qRT-PCR. But we presented the total number here by timing it by 2 to be consistent with Fig. 1f).

Agxt2l was localized in nutrient cells

Spatial orientation indicated that A. mellifera queen ovaries had more cells (nutrient, germ) than that of A. cerana queens (Fig. 6a). agxt2l was localized in nutrient cells (Fig. 6b) and expressed more in A. mellifera queen ovaries than in that of the A. cerana queens (Fig. 6c), details in supplementary Fig. 3. The results of DNA template preparation are in Supplementary Fig. 4.

Fig. 6: Spatial orientation of agxt2l expression in ovary.
Fig. 6: Spatial orientation of agxt2l expression in ovary.
Full size image

a Cell type and proportion in ovary of A. mellifera queen and A. cerana queen. b Spatial orientation of agxt2l expression in ovary. c Results of in situ hybridization of agxt2l.

Brc-z1 regulates the transcription of agxt2l gene

To explore the molecular mechanisms leading to the differential expression of the agxt2l gene between A. mellifera and A. cerana, we analyzed the genomic sequences corresponding to the differential peaks located in the promoter region of agxt2l in A. mellifera and A. cerana, and predicted two transcription factors whose binding sites differed between A. mellifera and A. cerana (Supplementary Fig. 5). To verify these three candidate transcription factors regulating the transcription of agxt2l, RNAi was performed in both honey bee species. It was found that after brc-z1 (LOC552255) RNAi, expression level of brc-z1 (N = 21, Z = − 4.49, P < 0.05) (Fig. 7a) and agxt2l (N = 21, Z = − 2.10, P < 0.05) (Fig. 7b), ovariole number (Mann–Whitney U, Z = −4.98, P < 0.05) (Fig. 7c) and body weight (T-test, t = −2.69, df  = 38, P < 0.05) (Fig. 7d) of A. mellifera queens in the RNAi group, were significantly lower than those in the control group (Supplementary Data 4). While the other genes and hsf (list the genes here) had no effect on the transcription of agxt2l.

Fig. 7: Motif analysis, transcription factor prediction and change of phospholipid content after RNAi of agxt2l and brc-z1.
Fig. 7: Motif analysis, transcription factor prediction and change of phospholipid content after RNAi of agxt2l and brc-z1.
Full size image

a, b Expression of brc-z1 and agxt2l injected with siRNA of brc-z1. c, d Body weight and number of ovarioles in A. mellifera injected with siRNA of brc-z1. e Change of phospholipid content after RNAi of agxt2l and brc-z1.

Agxt2l regulates phospholipid content in ovaries

Phospholipid content in larval ovaries of agxt2l and brc-z1 RNAi groups were significantly lower than that of the control group (Kruskal–Wallis, H = 48.45, P < 0.001, Padj < 0.05), and the phospholipid content in agxt2l RNAi group was significantly higher than that in the brc-z1 RNAi group (Kruskal–Wallis, H = 48.45, P < 0.001, Padj < 0.05) (Fig. 7e) (Supplementary Data 5).

Agxt2l regulates egg-laying performance of queens

The difference in egg-laying number before and after treatment was significantly greater in agxt2l RNAi group compared to the control group (T-test, t = 3.24, df = 16, P < 0.05) (Fig. 8) (Supplementary Data 7).

Fig. 8: Agxt2l regulates egg-laying performance of queens.
Fig. 8: Agxt2l regulates egg-laying performance of queens.
Full size image

The X-axis shows the experimental (injected agxt2l-siRNA solution) and control (injected negative control siRNA solution) groups, while the Y-axis shows the difference obtained by subtracting the pre-interference egg-laying quantity from the post-interference quantity.

Discussion

The principal findings of this study are that we revealed the mechanisms underlying the higher egg-laying capability of A. mellifera over A. cerana. We identified agxt2l as a potential gene: its higher expression in trophocyte cells induces increased glycerophospholipid metabolism, accounting for the higher queen egg-laying rate in Apis mellifera compared to A. cerana. And agxt2l was in turn regulated by brc-z1 (LOC552255) (Figs. 5u,  7,  8,  9, and Supplementary Fig. 5). In addition, brc-z1 promotes follicular opening. We determined that most of the enrichment pathways in egg-laying queens were related to oviposition or oocyte formation. Finally, we also verified that the egg-laying rate of A. mellifera queen is significantly higher than that of A. cerana queen, under the same environmental conditions (colony strength, season, and nectar resources).

Fig. 9
Fig. 9
Full size image

Increased glycerophospholipid metabolism due to higher expression of agxt2l in trophocyte cell explains the higher queen egg laying rate in Apis mellifera than A. cerana.

The egg-laying capability of A. mellifera queen and A. cerana queen have been studied many times33,34,35,36 but no study controlled variables which could cause differences besides inherent capability. To rule out contribution due to other factors (such as seasons33,35, population sizes34, regions36, etc.), we tested the egg-laying performance of the two species under the same conditions. We found that the number of eggs laid per milligram of body weight (Fig. 1f) and eggs per ovarioles (Fig. 1g) in A. mellifera queen was significantly higher than A. cerana queen, we hypothesize that the higher egg-laying rate in A. mellifera queen is mainly due to higher efficiency of each ovariole or higher maturation rate of oocytes in each ovariole.

The differences in egg-laying rate might be due to differences in glycerophospholipid metabolic rates. In AmV vs AcV, the common trend DEGs were enriched in the fatty acid metabolism pathway (Fig. 5j). This pathway has been shown to play an important role in the development and reproduction of insects37,38,39,40,41. In AmQ vs AcQ, the common trend DEGs were enriched in Wnt, mTOR signaling pathway and Glycerophospholipid metabolic pathway (Fig. 5k). Wnt signaling pathway has been found to be involved in the development of reproductive system, such as follicular development, ovulation, and luteinization42. mTOR and glycerolipid metabolism pathways indirectly promote each other. mTOR signaling has been shown to promote reproduction and lifespan in C. elegans43,44, female D. melanogaster45 and female mice46. The first step of the Glycerophospholipid metabolic pathway is mediated by mTOR signal transduction47, which suggests that Glycerophospholipid metabolic pathway may be related to egg-laying difference between the two species. In addition, in honey bees, approximately 30–40% of the dry weight of an egg is consisted of lipids48. Glycerophospholipids are important lipid components in the queen ovaries and play crucial roles in the membrane structure and cell functions49, which suggests that glycerophospholipid metabolic rate is the reason for the difference in egg-laying rate.

Agxt2l, spr and nedd8 promotes ovary development. We found that agxt21, spr and nedd8 were expressed higher in A. mellifera queen than in A. cerana, and they enhanced queen ovary development (Fig. 5c–e). Previous studies indicated that spr enhances egg-laying and female fecundity in Bactrocera dorsalis (Diptera: Tephritidae)50, D. melanogaster51 and Spodoptera litura52. Zebrafish nedd8 facilitates ovarian development and the maintenance of female secondary sexual characteristics via suppression of androgen receptor activity53. This suggests that in honey bees, agxt2l, spr and nedd8 gene may be the candidate genes involved in differential egg-laying capacity between the two sister species.

The rate of glycerophospholipid metabolic is promoted by agxt2l, which itself is enhanced by brc-z1. agxt2l is located in glycerophospholipid metabolic pathway and belongs to a subfamily of proteins called “Class II transaminases”, and is a modulator of phospholipid metabolism54. Liang et al. found that phospholipids promote ovarian development in shrimps55. In addition, in pregnant rabbits56, pigs57 and sheep58, phospholipids in the ovaries reach high levels during ovulation and the levels are maintained throughout pregnancy. These results indicate that phospholipids are related not only to ovarian development but also to embryo formation. We found that brc-z1 bound to the agxt2l promoter region (in both A. mellifera and A. cerana.) and positively regulated agxt2l. The broad-complex (br-c) is an early ecdysone hormone response gene that encodes four zinc finger protein subtypes (Z1, Z2, Z3, and Z4). In Blattella germanica, these four zinc finger isomers all play an important role in embryogenesis59. When BR-C is mis-expressed, the follicular cells in the front of the oocyte do not migrate forward properly60. In D. melanogaster, brc-z1 is involved in the formation of egg and ovum61,62,63 and is specifically expressed in nurse cells, somatic follicle cells64, and egg chamber59 which indicate that brc-z1 not only regulates agxt2l but also affects ovary development.

Increased glycerophospholipid metabolism due to higher expression of agxt2l in trophocyte cell explains the higher egg-laying rate in A. mellifera queens. We found that agxt2l is highly expressed in trophoblast cells (Fig. 6b) and the proportion of trophoblast cells in A. mellifera queens is higher than that in A. cerana. In bees, trophoblast cells contain a large number of coarse endoplasmic reticulum and Golgi cytoplasm, mitochondria, lipids, carbohydrates and proteins65,66,67,68. The presence of these organelles indicates that trophoblast cells are capable of synthesizing, absorbing, metabolizing, and storing substances, especially those derived from hemolymph69. After interfering with agxt2l, glycerophospholipid metabolism of trophoblast cells was limited, and lipid synthesis/ transport was decreased (Fig. 9), thus perhaps resulting a delay (or reduction) in egg formation. Based on the present results, we provide evidence that the rate of lipid synthesis controlled by agxt2l is the main reason why the A. mellifera queens have higher egg-laying capability than the A. cerana queens.

Furthermore, evolutionary pressure disparities likely underlie the variation in reproductive phenotypes. Apis mellifera ligustica is native to the Apennine Peninsula in the central Mediterranean, and Apis cerana cerana is mainly distributed in the subtropical humid area of the southern Longnan River Valley, the warm temperate humid area of the northern Longnan, the temperate semi-humid area of the southern Longzhong, and the temperate semi-arid area of the northern Longzhong, east of the Wushaoling Mountains in China. The monsoon climate in China, compared to Mediterranean climates, is characterized by high diversity, more acute changes, and a more scattered nectar flow. There are also large differences among different regions in China. All of these factors favor the phenotype for more flexible reproduction to achieve better resource management and faster responses to changing environments (including swarming and absconding), which increases the chance of survival for the colony under a changeable climate.

Materials and methods

Queen egg-laying capacity

A. mellifera ligustica colonies (Jiangxi Agricultural University, 28.46°N, 115.49°E, Nanchang, China) and A. cerana cerana colonies (Yifeng County, 28.26°N, 114.39°E, Jiangxi, China) (N = 10 each species) with newly laying queens (approximately 2 months old) were randomly selected for this experiment. The distance between the two places is 144 km. The queen of each colony was restricted to laying eggs on a comb for 24 h. The number of eggs on the comb was counted by taking digital photos (Canon EOD 70D) and using Image J (v1.54 d). This egg-laying experiment was repeated twice. We then fixed the queens on wax dish (N = 10 for each species) and dissected the ovaries. Ovary sections were prepared according to the method reported by Yi et al.70. Then, the number of ovarioles from both ovaries was counted under a dissection microscope (40x Nikon DS. Fi3). The differences in egg number and ovariole number between the two species were analyzed using Mann–Whitney Test and T-test in SPSS Statistics version 25.0.

Insects and sample collection

A. mellifera ligustica colonies and A. cerana cerana colonies (N = 14 colonies each species) in Jiangxi Agricultural University (28.46°N, 115.49°E), Nanchang, China, were used in the rest of this study. For A. mellifera, the queen of each colony was confined on a comb to lay eggs for six hours (8:00 am-2:00 pm), and then the queen was separated from the experimental comb using a queen excluder. After egg hatching, the larvae were transferred to queen cells. On the tenth day after hatching, twelve queenless colonies were prepared as nucleus colonies. In the afternoon of the 11th day, some queen cells were placed in the queenless colonies to obtain laying queens, and some were placed in queen cages to obtain virgin queen samples. Seven ovary samples of the newly emerged virgin queen were collected. After the queens had mated, seven ovary samples of the newly laying queens (approximately two months old) were collected. The sampling methods of A. cerana were the same as A. mellifera.

A total of 28 pairs of ovaries were collected from four types of queens, these were virgin queens and newly laying queens of both bee species (N = 7 for each type).

RNA extraction, transcriptome sequencing, and data analysis

Total RNAs were extracted from the ovary samples using Trizol Reagent (Ambion/Invitrogen, USA). RNA integrity and total amount were detected by Agilent 2100 Bioanalyzer (Agilent Technologies, CA, USA). All RNA samples that passed quality control were used to construct the library71. The library preparations were sequenced on an Illumina Novaseq platform (performed by Novogene Co., Ltd. (Beijing, China) and 150 bp paired-end reads were generated. From the sequencing data, reads containing just adapters and undetermined bases as well as low-quality reads (Qphred≤20 reads with more than 50% of the total read length) were filtered out.

The cleaned reads were aligned to the A. mellifera genome version Amel_HAv3.1(ftp.ncbi.nlm.nih.gov/genomes/all/GCF/003/254/395/GCF_003254395.2_Amel_HAv3.1/) or to the A. cerana reference genome (ftp://ftp.ncbi.nlm.nih.gov/ assembly/ GCA_020141525.1) using HISAT2 software72. The expression level was calculated by Fragments Per Kilobase of transcript per Million (FPKM) base pairs, with standardized annotation of the length of the transcript and the total number of reads aligned with the transcriptome. Analysis of differential expression was performed using the DESeq2 R package (1.32.0). The p-value was corrected for multiple comparisons with a false discovery rate (FDR) < 0.0573,74,75. Genes with padj≤0.05 were defined as differentially expressed genes.

After that, all DEGs were mapped to the Gene ontology (GO) database for enrichment analysis of functional significance. Finally, pathway enrichment analysis was performed in the Kyoto Encyclopedia of Genes and Genomes (KEGG) database76.

Nuclear extraction, ATAC-seq, and data analyses

The ATAC-seq was performed according to Buenrostro et al.77,78,79 and performed by Novogene Co., Ltd. (Beijing, China). Cell nuclei were extracted from the ovary samples and the nuclei particles were re-suspended in the Tn5 transpotase reaction mixture. The transposition reaction was incubated at 37 °C for 30 min. After transposition of Novogene, equimolar adapter1 and adapter2 were added to amplify the library by PCR. Then, the PCR products were purified by AMPure beads, and the quality of the libraries was evaluated by Qubit. The libraries were sequenced on an Illumina Novaseq platform.

Index of the reference genome was built using BWA (v0.7.12) and clean reads were aligned to the reference genome using BWA mem, by default parameters. Peaks with q-value threshold of 0.05 were used for all data sets, peak-related genes were identified and annotate by ChIPseeker80. The peaks with fold change more than 2(|log2fc | >1) were considered as differential peaks.

Genome homology alignment and data analyses

The homologs between A. mellifera and A. cerana were identified using Blast software (2.2.26) with e value < 1e − 10. In the process of RNA-seq and ATAC-seq analyses, the homologous genes with the same trend were selected, and the differential genes and differential peaks showing difference when using the two sets of genomes as references were eliminated to avoid differences caused by different reference genomes. The analytical technique route was shown in Supplementary Fig. 6.

Larval RNAi

The queen of a colony was confined on a comb to lay eggs for six hours, and then the queen was separated from the experimental comb by a queen excluder. After hatching of the eggs, the larvae were transferred to 24-cell tissue culture plates containing 150 μL artificial food, and then incubated at 35 °C and 85 relative humidity (RH). Each larva was fed with 150 μL artificial food two times daily. The vitro larval rearing method followed Schmehl et al.81. The queen food formula consisted of 53% fresh bee royal jelly, 6% glucose, 6% fructose, 1% yeast extract, and 34% distilled water82. When the larvae reached 2.5-day-old they were randomly divided into two group; one group was injected with about 1 μL RNAi solution (2 μg/μL), another group was injected with 1 μL negative control siRNA. Injection was performed by an Eppendorf FemtoJet 4i with an injecting needle at 1 μm in size, injection pressure at 500 hpa, and Output pressure (PC) at 15 hpa. Survival after injection was between 70–80%. For RNAi of agxt2l, spr, and nedd8, 24 larvae’s ovaries were randomly collected 48 h later after injection to determine RNAi efficiency using qRT-PCR. The remaining larvae were further reared until they became adult bees (the queens were virgins at the time of sampling). The body weight and ovarian tubes of the newly emerged bees were measured. When we injected siRNA of brc-z1 to test the relationship between brc-z1 and agxt2l, 33 larvae’s ovaries were randomly collected 48 h after injection to determine RNAi efficiency using qRT-PCR. The siRNA sequences are shown in Supplementary Data 6.

RNAi of egg-laying queen

A. mellifera ligustica colonies (N = 10) in Jiangxi Agricultural University (28.46°N, 115.49°E), Nanchang, China, were used in this study. All colonies had similar population sizes, and each contained a newly laying queens (approximately 2 months old). Queens with similar 24-hour egg-laying capacities (measured over three consecutive days) were selected (N = 6) and randomly divided into two groups (n = 3 per group). One group (n = 3) was injected with about 3 μL agxt2l-siRNA solution (1 μg/μL), another group (n = 3) was injected with 3 μL negative control siRNA (1 μg/μL). RNAi in egg-laying queens was performed following the method reported by Han et al.83. Prior to injection, queens were transferred to the laboratory and anesthetized with CO₂. Injections were performed dorsally between the second and third abdominal segments using an Eppendorf FemtoJet 4i with injecting needle at 3 μm in size, with an injection pressure at 70 hPa and output pressure (PC) at 15 hPa. Injected queens were given time to recover and placed back into their original colonies. 12 h post-injection, the egg-laying capacity of each queen was quantified. The impact of the treatment and control groups on queen fecundity was assessed by comparing the difference in egg numbers before and after RNAi injection.

Quantitative real-time PCR (qRT-PCR)

Total RNA was extracted from ovary samples using Trizol Reagent (Ambion/Invitrogen, USA) and RNA integrity was detected by agarose gel electrophoresis. RNA concentration and purity (OD260/280 value) were measured by an ultraviolet spectrophotometer. The RNA samples (1 μg/μL) were reverse transcribed into cDNA using the PrimeScript™ RT reagent Kit (Takara, Japan). The gapdh and β-actin gene were used as internal reference. The primer sequences are shown in Supplementary Data 6. qRT-PCR was performed according to Pan et al. 84. Vandesompele’s method was used to analyze qRT-PCR data85.

Detection of insect phospholipid by ELISA

Larval ovaries were used in this experiment to detect the phospholipid content using the Shanghai Meilian Insect Phospholipid Enzyme-Linked Immunosorbent Assay Kit. The larval ovary samples were weighed at first and then were homogenized (homogenate proportion of 10%, that is, 1 milligram tissue plus 9 ml of PBS). Then, 78 larvae ovary samples (control group=21, agxt2l RNAi group=30 and brc-z1 RNAi group=27) were centrifuged for about 20 min (14,000 RPM) to collect the supernatant. Subsequently, the absorbance (OD value) of each supernatant was measured by BioTek Synergy_HTX Multimode Reader (BioTek Synergy HTX, Bio-TEK) at 450 nm wavelength, and the sample concentration was calculated.

Spatial transcriptomics

One ovary of egg-laying queen of A. mellifera or A. cerana was used as a sample. After the sample was embedded with OTC, tissue permeation was performed. The optimal permeation time (30 min) was selected, and RT Master Mix was added into the permeated tissue section for amplification. Finally, fragment screening was performed to obtain cDNA library. Transcriptome sequencing was performed by LC-Bio Technology Co., Ltd (Hangzhou, China). The original data underwent data filtering, comparison and quantification to obtain the gene expression matrix of spots using the official analysis software of 10x Genomics (https://support.10xgenomics.com/spatial-gene-expression/software/overview/welcome). Seurat86 was used for difference analysis and visual analysis.

in situ hybridization

For in situ detection of the agxt2l, sense and antisense probes were synthesized using agxt2l specific primers with a T7 promoter sequence at the 5′ -ends (Agxt2l-Fow1: 5′ -TAATACGACTCACTATAGGGCGATGTCTGTTGGTGGGCAAATTCTAC-3′; Agxt2l-Rev1: 5′ -TAATACGACTCACTATAGGGCGACCATAGGTTTACCTACAGTGACTATATCC-3′) in combination with primers lacking the T7 sequence (Agxt2l-Fow2: 5′-TGTCTGTTGGTGGGCAAATTCTAC-3′; Agxt2l-Rev2: 5′-CCATAGGTTTACCTACAGTGACTATATCC-3′). These primers generated a product of 223 bp. The antisense and sense products were produced by the primer combination Agxt2l-Fow2+Agxt2l-Rev1 and Agxt2l-Fow1+ Agxt2l-Rev2, respectively. Amplification conditions were: 94 °C for 2 min, 10 cycles of 94 °C for 15 s, 55 °C for 30 s, 72 °C for 45 s and 20 cycles of 94 °C for 15 s, 55 °C for 30 s, 72 °C for 45 s (+5 s per cycle), final extension step at 72 °C for 7 min, 4 °C, ∞.

Aliquots of the amplification products were checked on agarose gels and quantified spectrophotometrically. RNA probes were generated by in vitro transcription from the T7 promoter using the DIG RNA Labeling Kit (SP6/T7) (Roche Applied Science). Fixation of dissected ovary and the subsequent hybridization and detection reactions were performed following the protocol optimized by Osborne and Dearden87 for in situ hybridization studies on honey bees.

Statistics and reproducibility

Data statistical analyzes were performed using SPSS Statistics software version 25.0 (IBM Corp., Armonk, NY) or GraphPad Prism version 8 for Windows (GraphPad Software, San Diego, CA, USA). The Mann–Whitney U-test, the T-test, and the Kruskal–Wallis in SPSS Statistics version 25.0 were used to evaluate queen egg-laying capacity (Figs. 1c–g,  7c, d), changes of phospholipid content after RNAi of agxt2l and brc-z1 (Fig. 7e), egg-laying performance before and after RNAi treatment (Fig. 8), and qPCR results throughout the text (Figs. 5i–t,  7a, b).