1 Introduction

The western honey bee, Apis mellifera L. (Hymenoptera: Apidae), performs sophisticated social behaviors based on cognitive functions, such as foraging, waggle dance, and hygienic behaviors (Frisch 1967; Seeley 2009). Surprisingly, these behaviors are controlled by the tiny brain of honey bees, consisting of about a million neurons in size of less than 1 mm3 (Witthöft 1967). Moreover, the genome size of this eusocial animal is only a tenth that of human (Weinstock et al. 2006). Thus, the honey bee is a good model organism for understanding the relationship between behavior and cognition through a variety of approaches (Menzel 2012).

Honey bees inside or outside the hive can be exposed to a variety of xenobiotics that do not target honey bees but are still potentially toxic to bees (Johnson 2015). Among them, τ-fluvalinate (hereafter, fluvalinate), widely used in Varroa mites control, belongs to pyrethroid neurotoxins that interfere with the recovery of action potentials through binding to voltage-gated sodium channels (Gupta and Crissman 2013; Soderlund et al. 2002). Even though the voltage-gated sodium channel of the honey bee is as vulnerable to fluvalinate as that of Varroa (Gosselin-Badaroudine and Chahine 2017), the honey bee has a relatively high tolerance against fluvalinate due to rapid detoxification by cytochrome P450 monooxygenase (Johnson et al. 2006; Mao et al. 2011). Albeit fluvalinate can lead to death or several adverse effects on honey bees (Dai et al. 2017; Lim et al. 2020; Rangel and Tarpy 2015; Rinderer et al. 1999). Since efficient cognitive quantifying assays for honey bees in the laboratory conditions are well known (Giurfa and Sandoz 2012), cognitive impairments have been typically studied as a side effect of various xenobiotics (Belzunces et al. 2012; Hesselbach and Scheiner 2018; Williamson and Wright 2013). Previous research showed that the intake of fluvalinate affects gustation, learning, and memory (Frost et al. 2013). In addition, another study reported that fluvalinate and other miticides affected gene expression related to memory (Gashout et al. 2020). Furthermore, we recently demonstrated topical external contact with fluvalinate induces olfactory deficit in foragers (Lim et al. 2020). However, little is known about the mechanism of cognitive impairment caused by xenobiotics, including the relatively innocuous fluvalinate. The aim of the present study delves into investigating whether sub-lethal fluvalinate abdominal contact affects overall cognitive behavior as well as olfactory sensitivity (Lim et al. 2020). We hypothesize that the fluvalinate-induced negative effects on cognition may be an indirect side effect of detoxification, not directly acting as a neurotoxin. In addition, we examined the molecular mechanism underlying the physiological changes induced by xenobiotics. Our findings would be applied to neuroscience research for uncovering insufficient knowledge about cognition.

2 Materials and methods

2.1 Insects

The experiments were performed in April and May 2019 with honey bees (Apis mellifera) from three colonies raised in the Incheon National University Apiary, Republic of Korea (37°22′25″, 126°37′40″). Foragers, who exhibit the most consistent performance in the proboscis extension response (PER) assay and have the highest learning ability among the different tasks, were used for the test (Pham-Delegue et al. 1990; Ray and Ferneyhough 1999). Returning foragers with pollen were caught directly at the entrance of the hive.

2.2 Exposure to fluvalinate

The τ-fluvalinate (#46,294; Sigma–Aldrich, St Louis, MO) was diluted in acetone (#270,725; Sigma-Aldrich) at a concentration of 2 μg/μl, which was identified as a sub-lethal dose in our previous study (Lim et al. 2020). After bees were anesthetized on ice, prepared fluvalinate solution was topically on the ventral side of the abdomen. Then, they were kept in an incubator (darkness/30 ± 2 ˚C/RH 60 ± 5%) and fed sugar candy ad libitum. After 21 h, each bee was harnessed on the trimmed blue tip of a 1-ml pipette for behavioral experiments. Then, the harnessed bees were fed with 50% sucrose (#S0389; Sigma-Aldrich) and returned to the incubator for 3 h without feeding. As described in Figure 1a, the foreleg of harnessed bees was freed to allow self-grooming. For molecular experiments, the survivors were stored in a deep freezer (− 80 ˚C) immediately after 24 h of post-treatment. In entire experiments, the bees exposed to solvent (acetone) were used as a control group.

Figure 1.
figure 1

Comparison of sucrose responsiveness between control (white) and fluvalinate (blue) groups. a Schematic of harnessed bee on trimmed blue tip and its PER. b Sucrose concentration-PER ratio graph. PER rate increases with increasing sucrose concentration in both groups. c Box plot of gustatory response scores. Note that the markedly low scores in the fluvalinate group. Thick lines and whiskers represent median and min/max values, respectively. Statistical significance was determined by Mann–Whitney U test (***P < 0.001, ncontrol = 54, nfluvalinate = 61).

2.3 PER: quantifying gustatory responsiveness

To quantify the gustatory responsiveness, water and a series of sucrose concentrations (10−2.5, −2, −1.5, −1, −0.5, 0 M) were sequentially contacted with the antennae of each bee using a toothpick (Figure 1a). The inter-series interval was 3 min. For each series, individuals with PER were checked. The bees that showed a PER in the water trial were discarded. The sum of PER to the stimulation with six different sucrose concentrations constitutes the gustatory response score (GRS; 0 to 6), which is an excellent measure of a honey bee’s gustatory responsiveness (Scheiner et al. 2015).

2.4 PER: quantifying associative learning and memory

After 3 h of starvation, each harnessed bee was transferred to the arena to test associative olfactory learning and memory (Figure 2a). The arena consists of 24 rooms, each of which has a vacuum flow (0.05 L/min). We used 1 M linalool diluted in mineral oil as a conditioned stimulus (CS) and 1 M sucrose as the unconditioned stimulus (US) (#L2602 linalool, #330,760 mineral oil; Sigma–Aldrich). Twenty microliters of odorant was absorbed onto a filter paper, which was then inserted into a glass Pasteur pipette (#Z310727; Volac, 150 mm; Poulten & Graf Ltd, Barking, UK). The detailed procedure is as follows (Figure 2b).

Figure 2.
figure 2

Comparison of learning and memory between control and fluvalinate groups. a, b Overall schematic diagram of experiment method for proboscis extension response (PER). c Olfactory associative acquisition curve. The acquisition curve of bees treated with the fluvalinate differed significantly from the control (P < 0.0001, ncontrol = 100, nfluvalinate = 100). d Percentage of bees that passed the MTM = 1 h and LTM = 24-h memory test. The fluvalinate group showed a significant decrease in MTM = 1-h memory (Pearson chi-square test, ***P < 0.0001, ncontrol = 72, nfluvalinate = 41).

3 Pre-training

The tested bees were stimulated by constant airflow (0.04 L/min) with linalool for 6 s, which was automatically controlled by a stimulus controller (CS-55, Syntech, Hilversum, Netherlands). In this step, the bees showing spontaneous PER were discarded. After 15 min, the following steps were performed on individuals who passed pre-training.

4 Acquisition

For the six acquisition trials, the conditioned odorant was delivered for 6 s. Immediately after 3 s out of six, the sucrose solution was delivered using a toothpick to the antennae to derive a PER. Then, the bee was allowed to lick sucrose freely while the odorant was still applied. In each trial, the bee was recorded whether it showed PER during the first 3 s of odorant application. The inter-trial interval was 8 min. Each tested bee was given an acquisition score (0 to 5) as described in the previous paper (Scheiner et al. 1999). On the first trial, the bees that displayed PER without the US or did not respond to the US were discarded. After the last trial, the bees were placed in the incubator until the 1-h memory test.

5 Memory test

The memory test was conducted only on the bees that had learned during the acquisition phase. One hour after the acquisition step, the bees were re-transferred to the arena. Then, the bees were presented with the constant flow with CS for 6 s without the US. Of the three trials, individuals that performed at least two PERs were recorded. The inter-trial interval was 5 min. After a 1-h memory test, the bees were fed to satiation with 50% sucrose solution and placed in the incubator for a 24-h memory test. Twenty-four-hour memory test was performed the same way as the 1-h memory test.

5.1 Comparative analysis of gene expression patterns of neuropeptides in the bee brain

Quantitative real-time polymerase chain reaction (qRT-PCR) was performed to analyze gene expression patterns. For this, total RNA was isolated from the bee brains using a Qiagen RNeasy Mini Kit (#74,104; Qiagen, Valencia, CA, USA). Three brains were used in each replicate. Using 500 ng of total RNA, cDNA was synthesized with oligo-dT with Invitrogen Superscript III enzyme following the manufacturer’s instructions (#18,080,044; Invitrogen, Carlsbad, CA). qRT-PCR was performed on the AriaMx Real-Time PCR System (#G8830A; Agilent Technologies, Santa Clara, CA) using Brilliant III Ultra-Fast SYBR Green QPCR Master Mix (#600,882; Agilent Technologies). All primer sets were synthesized in Macrogen Inc. (Seoul, Korea). For all performed analyses, ribosomal protein 49 (RP49) was used as a housekeeping gene for normalization. Thermal cycling was set at 95 °C for 1 min, followed by 40 cycles of 95 °C for 5 s and 60–63 °C for 10 s. Three technical replicates were performed for each biological replicate, and relative mRNA expression levels were calculated using the 2−△△CT method as described by Livak and Schmittgen (2001).

5.2 Statistical analysis

To determine the effect of fluvalinate, the sucrose responsiveness and the olfactory acquisition curves were compared by generalized linear model (GLM) using binary (PER or not) logistic regression analysis. Mann–Whitney U tests were conducted to analyze the differences in gustatory response scores and acquisition scores between both groups. The number of bees showing PER in the pre-training step was tested using Pearson chi-square test (χ2). The number of individuals that passed the memory test was applied in the same statistical method. Differences in mRNA expression patterns were examined by Student’s t test. All statistical analyses were performed with SPSS® Statics 25 (IBM, Armonk, NY), and graphs were prepared in GraphPad® Prism 8 (GraphPad Software, Inc., La Jolla, CA). Final arrangements were carried out using CorelDRAW Graphics Suite 2020 (Corel Corporation, Ottawa, Canada).

6 Results

6.1 Sub-lethal effect on taste behavior

For forager bees exposed to fluvalinate and control solution, the PER rate increased with increasing sugar concentrations, demonstrating a preference for higher sucrose concentrations regardless of treatment. The comparison between the two groups, however, showed that abdominal contact with fluvalinate has a negative effect (GLM, Wald = 44.673, P < 0.0001; Figure 1b). Even at the highest concentration of sucrose, the response rate of the control group was 96.3% while that of fluvalinate group was 70.5%. Moreover, GRS was significantly reduced in foragers exposed to fluvalinate (Mann–Whitney U tests, U = 869, P < 0.0001, ncontrol = 54, nfluvalinate = 61; Figure 1c).

6.2 Sub-lethal effect on olfactory associative learning and memory

We investigated the effect of fluvalinate on the cognitive abilities of forager bees using classical conditioning of the PER. First, there was no significant difference between groups in the spontaneous response to odor (chi-square test, χ2 = 0.014, P = 0.9070, ncontrol = 129, nfluvalinate = 128; Figure S1). Acquisition tests of six trials demonstrated that fluvalinate abdominal contact reduced olfactory-associated learning ability (GLM, Wald = 78.228w, P < 0.0001, ncontrol = 100, nfluvalinate = 100; Fig. 2c). In the last trial, the acquisition rate of the fluvalinate group was about half of the control group (control = 72.0%, fluvalinate = 37.0%).

Then, we asked whether fluvalinate affected the memory retrieval of learned bees (Figure 2d). In 1-h memory test, the fluvalinate group showed significantly lower memory retrieval than the control (control = 86.1%, fluvalinate = 58.5%; chi-square test, χ2 = 10.923, P < 0.0001, ncontrol = 72, nfluvalinate = 41). In 24-h memory test, however, there is no significant difference between groups (control = 63.0%, fluvalinate = 50.0%; chi-square test, χ2 = 1.488, P = 0.223, ncontrol = 54, nfluvalinate = 36). The honey bee’s memory stages are largely divided into three stages: short-term memory (STM), mid-term memory (MTM), and long-term memory (LTM) (Reviewed in Menzel 2012). Based on this knowledge, our results showed that fluvalinate exposure to the abdomen of forager bees significantly affects MTM but not LTM.

6.3 Sub-lethal effect on gene expression of neuropeptides in the bee brain

Based on previous papers on several neuropeptides that might affect learning and memory (Hergarden et al. 2012; Knapek et al. 2013; Takeuchi et al. 2004), we considered that some neuropeptides might be involved in the mechanisms underlying the negative effects of fluvalinate topically applied to the abdomen of forager bees. Among the candidates, our results demonstrated that short neuropeptide F (sNPF) and tachykinin (TK) in the brain are not involved in the cognitive impairment induced by fluvalinate (t test, tsNPF = 1.869, PsNPF = 0.121; tTK = 0.716, PTK = 0.506; ncontrol = 6, nfluvalinte = 6; Figure S2). Then, we focused on the other candidates that are also associated with juvenile hormone (JH), which is known to affect cognition in the honey bee (Corona et al. 2007). qRT-PCR showed that fluvalinate significantly increases ILP1 and decreases ILP2 in forager’s brain (t test, tILP1 = 3.320, PILP1 = 0.021; tILP2 = 3.421, PILP2 = 0.019; ncontrol = 6, nfluvalinte = 6; Figure 3). In the case of allatostatic factors, only AstCCC was observed to increase significantly in fluvalinate treated bees (t test, tAstA = 1.201, PAstA = 0.283; tAstCC = 1.323, PAstCC = 0.243; tAstCCC = 5.171, PAstCCC = 0.004; ncontrol = 6, nfluvalinte = 6; Figure 4a). On the contrary, allatotropin (AT) was significantly reduced (tAT = 3.000, PAT = 0.030; ncontrol = 6, nfluvalinte = 6; Fig. 4b). C-type allatostatin receptor, the receptor of AstCC and CCC, also increased significantly in fluvalinate abdominal contact (t test, tAstC-R = 4.913, PAstC-R = 0.004; ncontrol = 6, nfluvalinte = 6; Figure S2). Thus, our qRT-PCR data demonstrated that fluvalinate abdominal contact affects the ILP-associated JH pathway in the honey bee brain.

Figure 3.
figure 3

The gene expression patterns of ILPs. Relative mRNA expression levels of ILPs in brain from control (white bar) and sub-lethal dose treated (grey bar) foragers. Each bar represents the mean ± SE. Data points represent values from biological replicates. Statistical significance was determined by student’s t test (*P < 0.05, ncontrol = 6, nfluvalinate = 6).

Figure 4.
figure 4

The gene expression patterns of allatostatic factors. Relative mRNA expression levels of a AstA, CC, CCC, and b AstC-R in brain from control (white bar) and sub-lethal dose treated (grey bar) foragers. Each bar represents the mean ± SE. Data points represent values from biological replicates. Statistical significance was determined by student’s t-test (n.s. = non-significant, *P < 0.05, ncontrol = 6, nfluvalinate = 6).

7 Discussion

Unperceived effects of fluvalinate on honey bee health have led beekeepers to abuse fluvalinate to control Varroa (Johnson et al. 2006). However, fluvalinate intake or injection was reported to have adverse effects on the various physiological activities of honey bees (Dai et al. 2017; Rangel and Tarpy 2015; Rinderer et al. 1999). Recently, we proved that fluvalinate is highly toxic enough to kill bees by abdominal contact alone (Lim et al. 2020). In the present study, we also identified that more bees in the fluvalinate group died between acquisition and memory testing than in the control group. It was also demonstrated that abdominal contact with fluvalinate induces a significant decrease in taste responsiveness. It was revealed that exposure to sub-lethal doses causes a reduction of olfactory sensitivity in foragers (Lim et al. 2020). In the current study, moreover, we demonstrated another cognitive impairment caused by fluvalinate abdominal contact. Along with the results of a previous study on intake experiments (Frost et al. 2013), these findings provide strong evidence that fluvalinate, which is known to be relatively safe for honey bees, may impair cognitive function. The study found negative effects of fluvalinate on honey bee learning, memory, responsiveness to sucrose, and survival, particularly at high oral doses. However, the high dose (1.25 μg) used in the study was lower than the dose used in our experiments (2 μg). Yet, little is known about molecular and neural mechanisms underlying the effects of fluvalinate on the cognitive function of honey bees.

7.1 Cognitive impairment indirectly induced by xenobiotics

Cognitive function is involved in various types of behavior and is essential to maintain the honey bee colony. In the insect brain, the main region of cognitive function, which plays the role of the hippocampus in mammals, is the mushroom body (MB) (Menzel 2012). MB might be affected by xenobiotics, as the hippocampus is affected (Nasuti et al. 2014). The cognitive function of honey bees can be impaired by the direct impact of xenobiotics on synaptic receptors, neuronal ion channels, and cellular signaling. Pyrethroids like fluvalinate can directly impair the olfactory-cognitive neurons of the bees by modifying the voltage-gated sodium channel (Kadala et al. 2014). In nature, however, due to the efficient detoxification mechanism of honey bees on pyrethroid, especially fluvalinate (Johnson et al. 2006), this xenobiotic may not be a direct factor in the cognitive impairment of honey bees. Interestingly, however, cognitive impairment was observed 1 h after acute contact with fluvalinate in our previous and current studies. Within this period, fluvalinate was sufficiently degraded by CYP450 systems in honey bees (Hillier et al. 2013). Interestingly, some metabolites converted from xenobiotics by detoxification are as toxic as their parent compounds (Guez et al. 2003; Nauen et al. 2001; Suchail et al. 2001). Fluvalinate is also transformed into metabolite (4′-hydroxylated subfragment) by CYP450s (Mao et al. 2011), which may cause severe toxic or indirect effects on neurophysiological impairment or mental disorders.

Another possibility is that there might be changes in energy demand from detoxification processes which may have an effect. The xenobiotic detoxification process, in general, sequentially involves the degradation of toxins (Phase I: functionalization), and these products are conjugated (Phase II: conjugation) and excreted from the cells (Phase III: excretion) (Reviewed in Berenbaum and Johnson 2015). Although the number of enzyme genes involved in the detoxification mechanism is relatively low (Claudianos et al. 2006), honey bees are no more susceptible to xenobiotics than other insects (Hardstone and Scott 2010). Similar to some insects that require a fitness cost for tolerance to insecticides, detoxification processes in honey bees increase energy demand (Kliot and Ghanim 2012). Furthermore, increased energy demand by detoxification upregulates energy metabolism which leads to increased production of reactive oxygen species (ROS) (Rand et al. 2015). ROS production, typically, is known to be a major factor in neurodegenerative diseases that accompany cognitive impairment, such as Parkinson’s and Alzheimer’s disease in mammals (Smith et al. 2002; Tabner et al. 2001). Surprisingly, ROS production not only affects the cognitive function of vertebrates (Liu et al. 2003) but can also cause cognitive impairment in honeybee and Drosophila models (Farooqui 2008; Haddadi et al. 2014). In the current study, we identified that fluvalinate significantly increases ILP1 and decreases ILP2; thus, fluvalinate abdominal contact affects the ILP-associated JH pathway in the honey bee brain (Shpigler et al. 2021). Thus, cognitive impairment may be caused by changes in energy metabolism induced by detoxification. Our study indicates that another pathway in neuropeptide expression has been downregulated by the detoxification processes.

7.2 Molecular mechanism underlying cognitive impairment

The cognitive ability of honey bees is highly dependent on behavioral development. Young honey bees have poor cognitive abilities, but they gradually improve with age and peak when they become foragers (Ray and Ferneyhough 1999). Juvenile hormone (JH), the key regulator of behavior development, might be involved in the mechanism of cognition by organizing the neuroanatomy of MB in the honey bee brain (Withers et al. 1995). Indeed, JH-injected bees showed significantly enhanced both appetitive and aversive olfactory learning and 1-h memory recall (MTM) (McQuillan et al. 2014).

JH of insects, generally, is synthesized in the corpora allata by regulation of neuropeptides. These regulatory peptides consist of three types of inhibitory Ast and one stimulatory AT, depending on the species (Weaver and Audsley 2009). In the A. mellifera genome, AT and one A-type (FGLamide) and two C-type (PISCF) AST (ASTCC and CCC) are present, but the B-type (W(X6)Wamide) AST is absent (Veenstra 2016; Veenstra et al. 2012). Although it is not clear which honeybee ASTs are JH synthesis inhibitors, previous reports imply that C-type AST (in particular AST-CCC) plays a role in inhibitory effects on JH synthesis in honey bees (Urlacher et al. 2016). In our study, qRT-PCR of Ast-CCC significantly increased, and it might also affect as an inhibitor of JH synthesis. Furthermore, the artificial increase of two C-type ASTs reduces the olfactory learning ability of foragers, but only AST-CC affects young bees (Urlacher et al. 2019). In the current study, we tested only forager bees, and qRT-PCR of Ast-CC was increased, but it was not significant. It might be due to the significant reduction of JH synthesis by C-type AST in the forager that was more affected by JH than young bees (Elekonich et al. 2001). Interestingly, C-type AST in Aedes aegypti acts as the JH synthesis inhibitor by blocking the transport of citrate from mitochondria to cytosol (Nouzova et al. 2015). Blocked citrate can be used in the mitochondrial TCA cycle rather than cytoplasmic JH synthesis to increase energy metabolism. In addition, JH synthesis relies on the balance of AST and AT, which is usually regulated by nutritional status (Noriega 2004). Therefore, our findings demonstrated that xenobiotic detoxification could increase energy demand and consequent reduction of JH (Shpigler et al. 2021). Upstream of allatostatic or -tropic mechanism may include ILPs that are highly related to nutrient status (Colombani et al. 2003; Corona et al. 2007). Although many lines of research on the interaction between JH and two types of ILP in honeybees have been conducted, its complex relationships with neural and molecular mechanisms underlying memory modulation are still unclear (Ament et al. 2008; Ihle et al. 2014; Nilsen et al. 2011). In our present study, the change in the ILPs genes induced by fluvalinate suggests a correlation between AST, AT, and JH, which may be downstream of ILPs. Thus, we propose a novel hypothetical model in which cognitive impairment can be induced by ILPs-JH interaction (Figure 5).

Figure 5.
figure 5

Hypothetical model of cognitive impairment caused by fluvalinate exposure. Fluvalinate causes the need for detoxification, thus increasing the energy demand. This in turn changes ILP expression in brain, which influences the expression of downstream Ast-CCC and AT. As a result, JH expression in corpora cardiaca (CC) and corpora allata (CA) is reduced, followed by cognitive impairment.

In summary, our findings suggest that cognitive impairment induced by abdominal contact with fluvalinate is an indirect side effect rather than direct damage to neurons associated with learning and memory. The initiation of this indirect mechanism is the change in energy demand by detoxification. Subsequently, neuropeptide functions in cellular and molecular cascades may lead to the regulation of JH synthesis, leading to physiological behavioral changes. Our research suggests a new approach to understanding the side effects caused by xenobiotics. More studies are needed to investigate further mechanisms underlying the interplay of neuropeptides and toxification processes in detail.