Notch signalling mediates reproductive constraint in the adult worker honeybee

The hallmark of eusociality is the reproductive division of labour, in which one female caste reproduces, while reproduction is constrained in the subordinate caste. In adult worker honeybees (Apis mellifera) reproductive constraint is conditional: in the absence of the queen and brood, adult worker honeybees activate their ovaries and lay haploid male eggs. Here, we demonstrate that chemical inhibition of Notch signalling can overcome the repressive effect of queen pheromone and promote ovary activity in adult worker honeybees. We show that Notch signalling acts on the earliest stages of oogenesis and that the removal of the queen corresponds with a loss of Notch protein in the germarium. We conclude that the ancient and pleiotropic Notch signalling pathway has been co-opted into constraining reproduction in worker honeybees and we provide the first molecular mechanism directly linking ovary activity in adult worker bees with the presence of the queen.

Previous research has observed pycnotic nucleic in the germarium in worker bees and it is thought that the control of reproduction in worker bees occurs in this region of the ovariole 1

. (B)
Queen bees have large active ovaries that consist of up to 200 ovarioles. In contrast, queen-right worker bees have small-undeveloped ovaries with 2-10 ovarioles, these ovarioles are thin and generally show no signs of cell differentiation. In the absence of a queen and her brood a proportion of worker bees undergo ovary activation. This process can be separated into four discrete stages or scores based on a modified Hess scale 2 as detailed in the text. Arrowheads indicate signs of cell differentiation and asterisks indicate mature oocytes. Scale bars indicate 1 mm.

Supplementary Figure 2. Expression of genes of the E(spl)-C.
(A) bHLH2 RNA is detected in the anterior germarium of queen-right worker bees, but not in the germarium of queen-less or queen bees. In queen-less worker and queen ovaries expression of bHLH2 is restricted to the nurse cells and RNA is detected in the developing oocyte. RNA for bHLH2 is also detected in the follicle cells overlying the developing oocyte. (B) Her RNA is detected in the anterior germarium of queen-right worker bees, but not in the germarium of queen-less or queen bees. In queen-less worker and queen ovaries expression of bHLH2 is restricted to the nurse cells and RNA is detected in the developing oocyte. (C) Bearded RNA is detected in the anterior germarium of queen-right worker ovaries, but not in the anterior germarium of queen-less worker or queen bees. Bearded RNA is, however, detected in the posterior germarium of queen bees with RNA detected in the presumptive oocytes as they enlarge and differentiate within the cystocyte cluster. (D) bHLH2 RNA is weakly detected in the germarium of queen-right worker bees and in the nurse cells of queen-less worker bees and queens. bHLH2 RNA is only weakly detected in the developing oocytes. Following in situ hybridisation ovaries were counter-stained with DAPI (right panels). Scale bars indicate 100 µm. (A) Sequence conservation between Drosophila and Apis mellifera for the fragment of Notch protein used to generate the C17.9C6 antibody 3 . (B) The C17.9C6 antibody detects proteins of the expected sizes in the Drosophila lysate and honeybee lysates. The most abundant immunoreactive species is consistent in size with the full-length Notch receptor (288 kDa in Drosophila, and 266-71 kDa in honeybee). The band that we detect in the honeybee ovary lysate is smaller in size than that seen in Drosophila, as consistent with bioinformatic predictions of the size of the full-length Notch receptor. We also detect two species of around 120 kDa in Drosophila that are consistent in size with the cleaved Notch receptor. Two similar sized species are detected in the honeybee lysate. Blots were stripped and incubated with the tubulin antibody (E7) to allow comparisons of protein loading between lanes.

Supplementary Figure 4. Notch signalling differentiates follicle cell populations in the queen ovariole.
The Notch receptor is detectable on nurse cell membranes and follicle cell membranes (A).
However, as early as stage 3 of oogenesis we see increased fluorescence associated with the follicle cells located at the anterior and posterior of the egg chamber (B, indicated by arrowheads). Closer examination of these anterior and posterior follicle cells indicates that the Notch receptor is present in punctate dots within the cytoplasm of these cells as well as the cell membrane (C). These Notch receptor-rich dots do not colocalize with nuclei (D), nor do they colocalize with Rab11, a marker for recycling endosomes (E, and higher magnification in F).
This raises the possibility that the Notch receptor is sequestered in these anterior and posterior follicle cells rendering them refractory to Notch signalling. Consistent with this hypothesis, in situ hybridisation for bearded, a gene known to be responsive to Notch signalling, detects RNA in the main body follicle cells but not the anterior and posterior follicle cells (arrowheads in G).
This is most obvious at the anterior of the oocyte where bearded RNA is not detected in the five to ten follicle cells either side of the connection with the attached nurse cells (H). Examining these specimens by confocal reflectance microscopy confirms that the anterior follicle cells that are in the same plane as the germinal lumen do not express bearded (I) RNA for bearded is enriched at the anterior periphery of the follicle cells proximal to the nucleus (J). This indicates that the main body follicle cells are receiving a Notch signal, possibly via Delta signalling from the germline. In contrast, the cells at the anterior and posterior of the egg chamber to not express bearded, consistent with these cells being refractory to Notch signalling from the germline.
Following immunohistochemistry ovaries were counter-stained with DAPI and phalloidin to visualize nuclei and cortical-actin. Scale bars indicate 50 µm. RNA peaks in ovaries that are scored as a 2 (as yolk becomes deposited into the oocytes) with neuralized RNA detected at 5-fold higher levels than queen-right controls. Neuralized RNA then declines to queen-like levels in laying-workers (score = 3). (F) neuralized RNA is detected weakly throughout the ovariole in queen-right worker bees. In queen-less worker ovarioles and queen ovarioles no expression is detected in the terminal filament or germarium. Neuralized RNA is detected in the presumptive oocytes as they are specified and RNA accumulates strongly in these oocytes as they enlarge and begin to separate from the associated nurse cells. As the oocyte separates from the nurse cell bundle neuralized RNA is only detected in the posterior nurse cells. Late in oogenesis neuralized RNA is enriched on the dorsal surface of the oocyte and surrounds the oocyte nucleus. Following in situ hybridisation ovaries were counter-stained with DAPI (right panels). Scale bars indicate 100 µm. All RT-qPCR data is the mean of transcript levels (Log 10 ) in five biological samples for each condition. Boxplot whiskers indicate minimum and maximum, the box is defined by 25 th percentile, median and 75 th percentile.
Differences in target gene expression were determined by ANOVA with a Tukey's post-hoc test, statistical differences (P < 0.05) are denoted by different letters.

Supplementary Figure 6. Food intake and Kaplan-Meier survival curves for DAPT treated bees.
In the absence of QMP, supplementation of the diet with DAPT had no significant effect on food intake (A, grey = control (ethanol), black = DAPT treated (1mM)) or survival (B) as compared with the ethanol solvent control. Similarly, in the presence of QMP supplementation of the diet with DAPT had no significant effect on food intake (C, grey = control (ethanol), white = QMP control (QMP + ethanol), black = QMP and DAPT treated (1mM) ) or survival (D) as compared with the ethanol solvent control. Dotted lines on Kaplan-meier survival curves are 95% confidence intervals. Statistical significance for food intake was measured using t-tests with a Holm-Sidak correction for multiple testing and for survival was measured using a log rank test 4 .

Supplementary Tables
Supplementary Table 1: Gene expression as a predictor of physiological status of the ovary. Red indicates the most reliable predictor in this study, whilst bold font indicates genes that are equivalent or better predictors than a previously published gene, Anarchy 6 .

Target Gene % accurate prediction 'active' versus 'inactive'
Notch 40% Her 87% bearded 50%  PCR efficiencies for each of the possible reference genes were determined using a 10 fold dilution series of ovary cDNA. The PCR efficiencies ranged from 95% to 105% with R 2 values > 0.98. Analysis of the melting profiles showed no evidence of primer dimer, or amplification from genomic DNA (data not shown) and specificity of the amplification was confirmed by direct sequencing of the PCR product.

Supplementary
Raw expression data was analyzed using GeNormPlus. All eight of the candidate reference genes had high expression stability (M-values < 0.4, Supplementary Fig. 7A).
GeNorm indicated that for the experimental conditions tested (encompassing the full range of ovary activation illustrated in Supplementary Fig. 1B) that the geometric mean of two reference genes was optimal for normalization of the qRT-PCR data ( Supplementary Fig. 7B). Am-Rpn2 (26S proteasome non-ATPase regulatory subunit 1) and Am-mRPL44 (Mitochondrial ribosomal protein L44) were chosen for use in this study has they have the highest stability values (lowest M-values, Supplementary Fig. 7B).

Genes of the E(spl)-C are expressed in overlapping domains of the honeybee ovary.
The four genes of the E(spl)-C (bHLH1, Her, bearded, and bHLH1) are orthologous to the Drosophila E(spl)-C genes 7 which have well-defined roles in neurogenesis, are expressed in a Notch responsive manner and depend on Su(H) for initiation and maintenance of gene expression [8][9][10][11] . Expression of these genes during embryonic development is consistent with these genes also being regulated by Notch signalling in honeybees 7 .
Expression of these genes in the honeybee ovary varies with reproductive status (Fig. 2).
RNA for all four of these genes is detected in the germarium of queen-right but not queen-less or queen ovarioles (Fig. 2).
In the upper germarium, as the presumptive oocyte is specified from the cystocyte cluster, bearded RNA is detected in the presumptive oocytes ( Supplementary Fig. 2C). Bearded RNA accumulates in the presumptive oocyte before it is detected in the associated nurse cells.
As oogenesis proceeds bearded RNA becomes detectable in the nurse cells and continues to accumulate in the developing oocyte until it matures.
Genes of the E(spl)-C genes are, however, expressed more extensively, particularly in the queen ovary and queen-less worker ovary suggesting that Notch signalling also has other roles oogenesis. RNA for all four of these genes is detected in the nurse cell clusters, and three of these RNAs (bHLH2, bearded and to a lesser degree, Her) are maternally provided to the developing oocyte (Supplementary Fig. 2A,B,C). In contrast bHLH1 RNA is detected relatively weakly in the oocytes later in oogenesis (from stage 5 onwards) ( Supplementary Fig. 2D).
Maternal provision of these RNAs to the developing oocyte may suggest a possible role for these genes in regulating oocyte maturation or early developmental processes.

Notch signalling differentiates follicle cell populations in the honeybee queen ovary. In
Drosophila Notch signalling has multiple roles in oogenesis; in addition to the requirement for It has previously been reported that Notch signalling may play a similar role in regulating follicle cell migration and specification in the honeybee 16 . In the queen honeybee, Delta RNA is detected in the presumptive oocytes soon after they are specified, and is readily detectable by stage 1 of oogenesis (Fig. 3). Delta RNA continues to accumulate in the oocyte throughout oogenesis persisting until the oocyte is mature 16 . Using immunohistochemistry for the Notch intracellular domain we were able to show that the Notch receptor is present on all of the nurse cell membranes and on the follicle cell membranes ( Supplementary Fig. 4A,B). By mid-late oogenesis (stage 3 to stage 6) intracellular accumulation of the Notch receptor is observed in the anterior and posterior follicle cells (arrows in Supplementary Fig. 4B). These intracellular foci of Notch protein do not co-localize with follicle cell nuclei or F-actin ( Supplementary Fig. 4C,D).
To determine whether the Notch intracellular domain is colocalized with recycling endosomes we used immunohistochemistry with the Notch intracellular domain and Drosophila Rab11 as a marker of recycling endosomes. The Notch intracellular domain does not colocalize with recycling endosomes (Supplementary Fig. 4E,F), raising the possibility that the Notch intracellular domain may be sequestered intracellularly associated with degradative endosomes.
This sequestration may ensure that these anterior and posterior follicle cells are refractory to Notch signalling, and likely acts to differentiate these follicle cell populations. Consistent with this hypothesis, RNA for the E(spl)-C genes is detected at the apical surface of all of the main body follicle cells ( Supplementary Fig. 4G, H), suggesting that these cells are receiving the Notch signal from either Delta or Serrate present in the oocyte. However, RNA for the E(spl)-C genes is not detected in the anterior and posterior follicle cells ( Supplementary Fig. 4G,H, and confocal reflectance in I, J), confirming that these cells are not receiving a Notch signal.
While Notch signalling has a role in differentiating follicle cell populations in Drosophila and the honeybee, the details of the way that Notch signalling achieves this differs. In Drosophila, Delta signalling from the germ-line stimulates Notch receptor activation in the polar follicle cells, whereas it is the anterior and posterior follicle cells that are refractory to Notch signalling in the honeybee. These data imply that over evolutionary time the regulation of this pathway has altered but the specification of sub-populations of follicle cells is still dependent on Notch signalling from the germline.
Expression of the gene encoding the Notch receptor. Differences observed in Notch signalling between queen-less and queen-right worker ovaries (Fig. 2) could be due to differential expression of the Notch mRNA and protein. RT-qPCR demonstrates that mRNA levels for the Notch receptor are dynamic during the ovary activation process with a modest increase (1.7 fold) in Notch mRNA levels seen prior to yolk being deposited into the oocyte (score = 2). Notch RNA levels decline as queen-less worker ovaries become fully mature and produce oocytes ( Supplementary Fig. 5A). This dynamic expression, where queen-right workers have relatively low levels of Notch mRNA does not explain the high levels of Notch signalling we observe in queen-right worker ovaries (Fig. 2). Deltex RNA is subtly but consistently down-regulated in queen-less worker ovaries as compared to queen-right workers ( Supplementary Fig. 5B). In situ hybridisation revealed very similar expression patterns between queen-right, queen-less and queen worker ovaries ( Supplementary   Fig. 5C). Deltex RNA is expressed by nurse cells and is deposited into the developing oocyte, but is not detected in the germarium or terminal filament. This is consistent with Deltex having a role late in oogenesis, but not affecting Notch signalling in the germarium. RT-qPCR shows that expression of Am-neuralized is induced in queen-less worker bees very early in the ovary activation process, even before the ovaries are morphologically distinguishable from queen-right worker ovaries (3 fold induction in queen-less worker ovaries (score = 0) compared to queen-right workers). Expression peaks at levels 5-fold higher than queen-right workers in queen-less workers score = 2 (just as yolk begins to be deposited into the oocyte) ( Supplementary Fig. 5E). In situ hybridisation reveals that neuralized RNA is detected very weakly in the germarium of queen-less and queen bees but accumulates strongly in the newly specified oocytes (Supplementary Fig. 5F). Mid-way through oogenesis neuralized RNA becomes enriched at the dorsal surface of the oocyte and this enrichment persists as the oocyte matures ( Supplementary Fig. 5F). This expression pattern is similar to that seen for tailless in honeybees which has a role in patterning early honeybee embryos 24  Genes involved in Notch signalling also meet the testable-criteria outlined by Thompson 27 for genes underlying altruism; the expression of many of these genes associated with Notch signalling are responsive to the social environment (including bHLH2, Her, bearded, Delta, Serrate, Numb, Notch, Deltex and Neuralized). With the exception of Notch and Numb these genes are also differentially expressed in the ovaries of adult worker bees and queen bees;

Expression of
this reflects a theoretical constraint of altruistic genes linked to QMP exposure as genes must only be responsive to QMP in worker ovaries to prevent queens sterilizing themselves with their own pheromone 6 .
In addition to meeting the testable criteria for genes underlying altruism 27 we demonstrate, using functional manipulation, that inhibition of Notch signalling can overcome the repressive effect of QMP on ovary activity and that Notch signalling acts on the earliest stages of oogenesis, in the germarium, the region of the ovary that has been shown to differ morphologically between queen-right and queen-less worker bees 1 . We therefore conclude that Notch signalling is the proximate mechanism maintaining reproductive sterility in the worker honeybee and this sterility is crucial for conflict resolution over production of males 28 and social harmony 29,30 .