Multiple combinations of RDL subunits diversify the repertoire of GABA receptors in the honey bee parasite Varroa destructor

In insects, γ-aminobutyric acid (GABA) is the major inhibitory neurotransmitter, and GABA-gated ion channels are the target of different classes of insecticides, including fipronil. We report here the cloning of six subunits (four RDL, one LCCH3, and one GRD) that constitute the repertoire of the GABA-gated ion channel family of the Varroa mite (Varroa destructor), a honey bee ectoparasite. We also isolated a truncated GRD subunit with a premature stop codon. We found that when expressed in Xenopus laevis oocytes, three of the four RDL subunits (VdesRDL1, VdesRDL2, and VdesRDL3) formed functional, homomultimeric anionic receptors, whereas GRD and LCCH3 produced heteromultimeric cationic receptors. These receptors displayed specific sensitivities toward GABA and fipronil, and VdesRDL1 was the most resistant to the insecticide. We identified specific residues in the VdesRDL1 pore-lining region that explain its high resistance to fipronil. VdesRDL4 did not form a functional receptor when expressed alone, but it assembled with VdesRDL1 to form a heteromultimeric receptor with properties distinct from those of the VdesRDL1 homomultimeric receptor. Moreover, VdesRDL1 physically interacted with VdesRDL3, generating a heteromultimeric receptor combining properties of both subunits. On the other hand, we did not detect any functional interaction between VdesLCCH3 and the VdesRDL subunits, an observation that differed from what was previously reported for Drosophila melanogaster. In conclusion, this study provides insights relevant to improve our understanding of the precise role of GABAergic signaling in insects and new tools for the development of Varroa mite–specific insecticidal agents that do not harm honey bees.

Varroa destructor is the most devastating pest for the Western honey bee Apis mellifera (1). The lifespan of infested honey bee colonies is significantly shorter unless they are treated with acaricides. Moreover, several honey bee viruses, including deformed wing virus or acute bee paralysis virus, are transmitted by Varroa mites (2). The combined effects of the parasite and the viral diseases play an important role in honey bee colony health, contributing to morphological deformities (small body, deformed wings) and immune system weakening. These infestations probably contribute to colony collapse disorder, a syndrome leading to the large-scale loss of managed bees and recorded in Europe and North America since 2006 (3). Colony collapse disorder is especially alarming because A. mellifera is commonly used for active crop pollination, and its role has been evaluated at over $200 billion worldwide. Synthetic agents, such as pyrethroids (fluvalinate) or organophosphates (coumaphos), are used to limit Varroa mite infestation. However, these treatments may have adverse effects on bee health (4). To minimize pesticide residues in bee colonies, beekeepers also use essential oil (thymol) and organic acids (formic acid, oxalic acid). Thymol, a constituent of thyme essential oil, is a positive allosteric modulator of GABA-gated receptors in humans and also in fruit flies (5). Moreover, insect GABA receptors are one of the major targets of insecticides including phenylpyrazoles (fipronil), cyclodienes (dieldrin), and metadiamides (broflanilide) (6).
GABA is the major inhibitory neurotransmitter in the vertebrate and invertebrate nervous systems. In insects, GABA is important for locomotion control, olfactory learning, and regulation of sleep and aggression (7)(8)(9)(10). GABA receptors are members of the cysteine-loop ligand-gated ion channel superfamily that also includes nicotinic acetylcholine receptors, which permeate cations, and glutamate-gated channels, which permeate anions (11,12). The receptors of this family are composed of five homologous subunits arranged around a central ion channel. Each subunit possesses an extracellular N-terminal domain that encompasses the ligand-binding domain (LBD), 2 and four transmembrane segments (TM1-4), with a variable intracellular loop between TM3 and TM4. The TM2 segment lines the channel pore (13). In insects, three genes encoding GABA receptors have been identified: RDL (resistance to dieldrin), LCCH3 (ligand-gated chloride channel homolog 3), and GRD (GABA and glycine receptor-like subunit from Drosophila). Mutations in the TM2 of the RDL subunits have been detected in insecticide-resistant strains of many arthropod species (14). In heterologous expression systems, RDL subunits assemble into homopentameric, chloride-selective, and thymol-sensitive channels (5,(15)(16)(17)(18). Less is known about LCCH3 and GRD. Only D. melanogaster subunits have been expressed in Xenopus laevis oocytes where they assemble into heteropentameric, cation-selective channels (19). Although a fragment encoding the TM2 of a RDL subunit was previously isolated in V. destructor (20), nothing is known about GABA-gated ion channels in this parasite, which is a major target of acaricides.
Here, we describe the molecular cloning of four RDL, one LCCH3, and two GRD subunits from V. destructor. We found that when expressed in X. laevis oocytes, the resulting GABAgated channels displayed specific pharmacological and biophysical features. Moreover, RDL subunits could assemble in heteromultimeric receptors. The properties of the three homomultimeric RDL receptors expressed in X. laevis oocytes VdesRDL1, VdesRDL2, and VdesRDL3 were clearly distinct from those previously reported for the honey bee RDL ortholog (21). This suggests structural specificities that could be exploited in a pharmacological screen to identify insecticidal agents that will be more efficient on Varroa receptors than on their ortholog in honey bee and therefore could potentially be used to control Varroa mites without harming honey bees.

Isolation of GABA-gated ion channel subunits from V. destructor
We obtained the cDNAs of the V. destructor RDL (VdesRDL), GRD (VdesGRD), and LCCH3 (VdesLCCH3) subunits by rapid amplification of cDNA ends and PCR using gene specific oligonucleotides. For VdesRDL2, however, we could initially isolate only an incomplete ORF (GenBank TM accession number KY748051). We then took advantage of the large number of new sequences obtained from the transcriptomic analysis in Varroa jacobsoni (assembly vjacob_1.0 GCF_002532875) and V. destructor (Assembly Vdes3.0 GCF_002443255) deposited in the GenBank TM during this work to obtain a full-length ORF also for VdesRDL2. The cDNAs for the RDL subunits included ORFs of 1758, 1986, 1776, and 2193 bp encoding proteins of 585, 661, 591, and 730 amino acids for VdesRDL1 (KY748050), VdesRDL2 (MH423074), VdesRDL3 (KY748052), and VdesRDL4 (KY748053), respectively (Fig. 1). The predicted molecular masses were approximately 65 kDa for VdesRDL1 and VdesRDL3 and 72 and 80 kDa for VdesRDL2 and VdesRDL4, respectively. VdesRDL1 predicted sequences found in GenBank TM (XP_02265189 to XP_022651192) were 10 residues longer than the deduced amino acid sequence of our clone, but we found this small insertion in the TM3-TM4 loop in another clone not studied here. The predicted sequences for VdesRDL2 found in GenBank TM differed at the N terminus and the TM3-TM4 loop (XP_022658105 to XP_022658113). The deduced amino acid sequence of VdesRDL2 was similar to the sequence XP_022658108 at the TM3-TM4 loop and to the sequence XP_022658109 at the N terminus. The deduced amino acid sequence of VdesRDL3 was similar to the sequences XP_022644956 to XP_02264958. In the three other predicted sequences found in GenBank TM for VdesRDL3, the TM4 (XP_022644960 and XP_02264961), and a conserved segment into the LBD and the TM1 (XP_02264959) were missing. Like for VdesRDL2, the amino acid sequences predicted for VdesRDL4 differed mainly at the TM3-TM4 loop and at the N terminus (XP_022665348 to XP_022665350 and XP_ 022665352 to XP_022665356). Although most of the deduced amino acid sequence of VdesRDL4 was similar to XP_ 022665348, the N terminus was absent in any of the predicted sequences available in GenBank TM . The four VdesRDL subunits shared ϳ50% identity. As expected, the LBD and the TM segments were well conserved between the VdesRDL subunits and the D. melanogaster RDL (DmelRDL) subunit (Fig. 1). Notably, all the residues involved in GABA binding in the pentameric receptor (Tyr 109 and Arg 111 in loop D, Ser 176 at the end of loop E, Glu 204 and Phe 206 in loop B, and Thr 251 and Tyr 254 in loop C, according to DmelRDL sequence numbering) (22,23) were conserved in the four VdesRDL subunits. However, VdesRDL1 sequence differed from the other sequences at residues critical for GABA receptor inhibition by the insecticides dieldrin and fipronil (14): a serine instead of alanine at 2Ј, a methionine instead of threonine at 6Ј in TM2 (when the charged residue at the cytoplasmic end is numbered 0Ј), and an alanine instead of a threonine at the cytoplasmic end of the TM3 (Fig. 1). Two of these differences (Met at 6Ј and Ala 306 ) were also found in VdesRDL4. Therefore, specific pharmacology could be developed for multimeric receptors that include one of these subunits. The cDNAs for the VdesGRD (KY748054) and VdesLCCH3 (KY748055) subunits included ORFs of 1662 and 1980 bp encoding proteins of 553 and 659 amino acids, respectively (Fig. 1). The predicted molecular masses were 63 kDa for VdesGRD and 75 kDa for VdesLCCH3. We also cloned a truncated GRD variant that we named Vdes⌬GRD, but we did not investigate its features. It is noteworthy that three predicted sequences found in GenBank TM (XP_022664708 to XP_022664710) are similarly truncated. The deduced amino acid sequence of VdesGRD differed from that of the two predicted proteins found in GenBank TM (XP_022653054 and XP_022653055) that lack a fragment of the LBD. VdesLCCH3 deduced amino acid sequence was similar to that of XP_022672719. Two other predicted proteins differed from VdesLCCH3 at the TM3-TM4 loop and by an insertion of 17 residues between Loop E and Loop B (XP_022672700 and XP_022672710). VdesGRD and VdesLCCH3 sequences shared 33 and 50% identity with the D. melanogaster subunits, respectively. The residues identified in the RDL sequences as involved in GABA binding were not well conserved in VdesGRD or VdesLCCH3. However, because of the absence of structural data and mutagenic studies, GABA-interacting residues have been not precisely identify in these subunits (22).

Functional expression of V. destructor GABA-gated channels
To evaluate the properties of the cloned V. destructor GABAgated channels, we performed two-electrode voltage-clamp recordings in X. laevis oocytes after injection of the cRNAs encoding each individual VdesRDL subunit or of a mixture of cRNAs encoding the VdesGRD and VdesLCCH3 subunits because these subunits do not form functional receptors when GABA-gated ion channels in Varroa destructor expressed separately. When maintained at Ϫ60 mV, perfusion of GABA induced an inward current only in oocytes that expressed VdesRDL1, VdesRDL2, VdesRDL3, or VdesGRD/ LCCH3, but not in oocytes that expressed VdesRDL4 (Fig. 2). This indicates that VdesRDL1, 2, and 3 subunits can form functional homomultimeric receptors. We then obtained concen-tration-response curves by challenging oocytes with GABA concentrations ranging from 0.1 to 250 M. Each VdesRDL subunit displayed a different GABA EC 50 with the smallest value for VdesRDL2 and the highest for VdesRDL1 (Table 1). In similar experiments, the GABA EC 50 value of the A. mellifera RDL subunit (25.2 Ϯ 3.9 M, n ϭ 5) was similar to value previ-

Inhibition of V. destructor GABA-gated channels by fipronil
The presence of a 2ЈS and 6ЈM in the TM2 of the VdesRDL1 amino acid sequence suggested that this subunit may display a different sensitivity toward insecticides than the other VdesRDL subunits. Fipronil potently inhibited the GABAevoked currents recorded in oocytes that expressed VdesRDL2, VdesRDL3, or VdesGDR/LCCH3 but not VdesRDL1 (Fig. 4). The VdesRDL1 IC 50 value was ϳ50 times higher than that of the other RDL subunits (Table 1). In our recording conditions, fipronil IC 50 for A. mellifera RDL (0.066 Ϯ 0.014 M, n ϭ 4) was similar to the value previously reported (21). To assess whether with increasing GABA concentrations in X. laevis oocytes after the injection of the cRNAs encoding VdesRDL1, VdesRDL2, VdesRDL3, or VdesRDL4. The duration of GABA perfusion was adjusted for each oocyte to reach the maximal current amplitude. Note the absence of current in oocytes injected with the cRNA encoding VdesRDL4. Bottom panel, GABA concentration-response curves obtained in X. laevis oocytes that express VdesRDL1, VdesRDL2, or VdesRDL3. B, same as in A but in X. laevis oocytes that express VdesGRD/LCCH3. The data are the means Ϯ S.E. of n ϭ 9 -23 oocytes. the specific differences in the TM2 sequence were responsible for VdesRDL1 resistance to fipronil, we determined the fipronil IC 50 for VdesRDL1 that harbored the mutations S2ЈA and M6ЈT (VdesRDL1mut). The significantly smaller IC 50 for VdesRDL1mut compared with VdesRDL1 (Table 1) demonstrated the importance of the 2Ј and 6Ј residues for RDL sensitivity toward fipronil. Our sequence analysis also indicated the presence of a glycine and an alanine at 2Ј in VdesGRD and VdesLCCH3, respectively. Although the mutation A2ЈG in RDL subunit has been frequently identified in strains resistant to fipronil (14), our results indicate that its presence in Vdes-GRD does not abolish VdesGRD/LCCH3 sensitivity to fipronil.

VdesRDL4 can form heteromultimeric receptors with VdesRDL1 in X. laevis oocytes
Western blotting analysis of lysates from X. laevis oocytes showed that oocytes injected with cRNA encoding VdesRDL4 tagged with GFP expressed a protein at the expected molecular

GABA-gated ion channels in Varroa destructor
mass (Fig. 5A). Because VdesRDL4 could not form functional homopentameric receptors, we tried to co-express this subunit with VdesRDL3 or VdesRDL1 ( Fig. 5 and Table 1). We did not observe any significant differences between the GABA EC 50 and fipronil IC 50 values obtained with VdesRDL3/RDL4 or with VdesRDL3 alone (Table 1). However, these results do not exclude that these two subunits might form heteromultimeric receptors in X. laevis oocytes. Indeed, the sensitivity toward GABA or fipronil of the heteromultimeric receptor might be similar to that of VdesRDL3 alone. This possibility was suggested by the finding that co-expression of VdesRDL4 and VdesRDL1 increased the sensitivity to fipronil compared with VdesRDL1 expressed alone ( Fig. 5C and Table 1). Conversely, the GABA EC 50 value obtained with VdesRDL1/RDL4 was not significantly different from that of VdesRDL1 alone. To determine whether the higher sensitivity toward fipronil of the heteromultimer VdesRDL1/RDL4 was due to the presence of VdesRDL4, we introduced the mutation A2ЈS in this subunit (VdesRDL4mut). The fipronil IC 50 value of VdesRDL1/ RDL4mut was not different from that of VdesRDL1 alone ( Fig.  5C and Table 1). As expected, the fipronil IC 50 value obtained when the VdesRDL1mut was co-expressed with VdesRDL4 was comparable with that of VdesRDL1mut alone ( Fig. 5C and Table 1). This demonstrates that VdesRDL1 can assemble in functional receptors with VdesRDL4 and that these heteromultimeric receptors display different features compared with the VdesRDL1 homopentameric receptors.

VdesRDL1 assembles with VdesRDL3 to form heteromultimeric receptors in X. laevis oocytes
To test whether VdesRDL1 and VdesRDL3 could assemble in heteromultimeric receptors, we co-injected equimolar concentration of RNAs encoding VdesRDL1 and VdesRDL3 in X. laevis oocytes. Indeed the expression of both receptors was similar with current amplitude values ranging from Ϫ635 to Ϫ8235 nA (mean Ϫ5087 Ϯ 679 nA, n ϭ 13) for VdesRDL1 and from Ϫ1026 to Ϫ5802 nA (mean Ϫ2882 Ϯ 335 nA, n ϭ 21) for VdesRDL3, 1-3 days after injection. The GABA EC 50 value obtained from oocytes that expressed VdesRDL1/RDL3 (89.2-121.2 M) was not significantly different from the value obtained for VdesRDL1 alone (Fig. 6A and Table 1). Conversely, the fipronil IC 50 value for VdesRDL1/RLD3 (1.4 -14.2 M) was close to the value obtained for VdesRDL3 alone (Fig. 6B and Table 1). These results are compatible with the expression of a homogenous population of heteromultimeric receptors rather than two populations of homomultimers. They also suggests that in X. laevis oocytes, VdesRDL1 and VdesRDL3 preferentially assemble to form heteromultimers, with properties different from those of the single receptors. Co-immunoprecipitation experiments using VdesRDL1 and VdesRDL3 tagged with GFP and mCherry, respectively, confirmed the direct interaction between VdesRDL1 and VdesRDL3 (Fig. 6C). This strongly supports that when the two RDL subunits are expressed together, they preferentially form heteromultimeric receptors.

VdesLCCH3 does not assemble in functional heteromultimers with VdesRDL subunits
Despite the fact that VdesRDL and VdesGRD/LCCH3 permeate different ions (anions/Cl Ϫ ions and cations/Na ϩ ions, respectively), a functional interaction between RDL and LCCH3 has been suggested in flies (25) and honey bees (8). Because the VdesGRD/LCCH3 heteromultimer was highly sensitive to fipronil inhibition (Table 1), VdesRDL1/LCCH3 heteromultimers should be fipronil sensitive on the basis of our results with VdesRDL1/RDL4. First, introduction of the mutation A2ЈS in VdesLCCH3 (VdesLCCH3mut) confirmed that this subunit was, at least in part, responsible for the VdesGRD/ LCCH3 sensitivity to fipronil. Indeed, the fipronil IC 50 value of the VdesGRD/LCCH3mut heteromultimer was significantly higher than that of VdesGRD/LCCH3 (Fig. 7C and Table 1). Then we co-injected into X. laevis oocytes the cRNA encoding VdesRDL1 with a 5-fold excess of the cRNA encoding VdesLCCH3 to promote the formation of heteromultimeric receptors. However, we did not observe any significant difference between the GABA EC 50 and fipronil IC 50 values for VdesRDL1/LCCH3 and for VdesRDL1 alone ( Fig. 7 and Table  1). On the basis of the results obtained after co-expression of VdesRDL1 with VdesRDL4 or VdesRDL3, this finding seems to exclude the formation of a functional heteromultimeric receptor between VdesRDL1 and VdesLCCH3. We next assessed whether VdesRDL4, which cannot form functional homomultimeric receptors, could assemble with VdesLCCH3 to form functional heteromultimeric receptors. We could not record any GABA-evoked current in oocytes co-injected with these subunits, suggesting that VdesRDL4 and VdesLCCH3 did not assemble or that the heteromultimeric receptors were not functional (Fig. 7A). In flies, the co-expression of LCCH3 decreased the picrotoxin sensitivity of RDL (25). Because VdesRDL1 (like VdesRDL4) harbors a methionine at 6Ј in TM2 that confers picrotoxin resistance (20), we tested whether VdLCCH3 could form functional heteromultimers with VdesRDL3 by evaluating the picrotoxin sensitivity (Fig. 7B). The picrotoxin IC 50 value determined in X. laevis oocytes that expressed VdesRDL3/ LCCH3 (33 Ϯ 2 nM, n ϭ 7) was not significantly different from

GABA-gated ion channels in Varroa destructor
that of VdesRDL3 alone (37 Ϯ 3 nM, n ϭ 7). These results seem to exclude a functional interaction between VdesLCCH3 and any of the VdesRDL subunits.

Discussion
In insects, there are four genes encoding potential GABAgated channels: RDL, LCCH3, GRD, and CG8916. Since their identification in D. melanogaster (15), many RDL orthologs have been isolated from various arthropod species. However, only few of them, for example in the tobacco budworm Heliothis virescens (Lepidoptera) (16), the American dog tick Dermacentor variabilis (Acari) (17), the planthopper Sogatella furcifera (Hemiptera) (18), or the honey bee A. mellifera (Hymenoptera) (21), have been functionally characterized in heterologous expression systems. Moreover, functional expression of the GRD/LCCH3 heteromultimer has only been done with the D. melanogaster subunits (19), although these subunits have been isolated in other arthropod species, such as the silkworm Bombyx mori (Lepidoptera) (26) or the planthopper Laodelphax striatellus (Hemiptera) (27). On the other hand, functional expression of the CG8916 subunit alone, as homomultimer, or with LCCH3 has never been obtained (19).

GABA-gated ion channels in Varroa destructor
In this study, we report the complete cDNA sequences of four RLD, one LCCH3, and one GRD subunits from V. destructor. This is the largest family of GABA-gated ion channel subunits cloned to date in insects. We also isolated a cDNA encoding an additional GRD subunit in which a large part of the TM3-TM4 loop and the TM4 segment are lacking. We identified a splice variant only for VdesRDL1, but future experiments will undoubtedly isolate other variants for the GABA receptors in V. destructor, like in other arthropod species. In many Lepidoptera, several genes encode RDL subunits. This is the case for the diamondback moth Plutella xylostella (28) and the rice stem borer Chilo suppressalis (29), in which two RDL subunits have been isolated, or the silkworm B. mori, in which three RDL subunits have been cloned (26). In Arachnoidea also several genes encode RDL subunits, for instance in V. destructor (four, this study) and in the two-spotted spider mite Tetranychus urticae (30), in which three RDL subunits have been isolated. This characteristic should be confirmed thanks to the availability of the genomic sequence of other Arachnoidea (31). Conversely, there is only a single gene encoding RDL subunit in Diptera (e.g. fruit fly) (12), in Hymenoptera (e.g. honey bee) (11), and in Hemiptera, (e.g. the brown planthopper Nilaparvata lugens) (32).
It has been hypothesized that the existence of multiple RDL subunits could result from gene duplication to enhance the tolerance to naturally occurring (e.g. picrotoxin-and picrotoxinlike molecules of plant origin) and synthetic insecticides (e.g. dieldrin and fipronil) (26,33). Differences in the amino acid sequence at 2Ј in TM2 support this hypothesis. Indeed, the three RDL subunits isolated in B. mori have alanine, serine, or glutamine residues at this position (26), and the two RDL subunits of C. suppressalis harbor an alanine or serine residue (29). In V. destructor, the VdesRDL1 subunit has a serine residue, and the other three RDL subunits have alanines. Mutations at this specific position in the inner mouth of the channel pore have been observed in many insecticidal-resistant arthropod strains (14,34). In a homology model, fipronil docks best in the channel pore, interacting with 2ЈA, 3ЈL, and 6ЈT (33). C. suppressalis RDL1 and RDL2 subunits have different residues at 2Ј (alanine and serine) in an otherwise conserved TM2, but they display a similar sensitivity toward fipronil (29). We did not evaluate the role of the specific residues found in VdesRDL1 in fipronil sensitivity, but it seems that the combination of 2ЈS, 6ЈM, and Ala 355 , at the end of TM3, is involved in the sensitivity to this insecticide (34).
Multiple RDL subunits with specific pharmacological or biophysical properties could also be advantageous for GABA signaling. The two RDL subunits isolated from C. suppressalis have a specific affinity for GABA (the GABA EC 50 is 38.3 M for CsRDL1 and 21.6 M for CsRDL2) (29). The situation is even more complex with the Varroa mite RDL subunits because the GABA sensitivity of VdesRDL2 was 50 times higher than that of VdesRDL1, and the GABA affinity of VdesRDL3 was intermediate. We also found that in X. laevis oocytes, the VdesRDL subunits can assemble in heteromultimeric receptors with specific properties. Therefore, in insects, multiple RLD subunits might provide a large toolbox to finely tune GABA signaling. Studies on the physiological roles of GABA receptors in insects are scarce, and it is difficult to attribute these roles to a specific GABA-gated receptor (35). This could be improved by the analysis of the expression pattern of the different GABA receptor subunits in the different tissues and during development. Such analysis has been partially done in Leodelphax striatellus (36) and in C. suppressalis where CsRDL1 and CsRDL2 show similar expression pattern but different expression levels (29). This suggests that in C. suppressalis, the two RDL subunits could assemble in heteromultimers. Additional experiments are needed to confirm the existence of such heteromultimeric receptors in vivo.
In mammals, GABA can also play the role of an excitatory neurotransmitter, and this effect results from the inversion of the chloride gradient that causes cellular depolarization (37). Excitatory GABA receptors have been described in Caenorhabditis elegans in which they play important roles in muscular contraction (38). Conversely, GABA-mediated excitatory neurotransmission has never been observed in insects (39,40), although it has been clearly demonstrated that the D. melanogaster GRD and LCCH3 subunits assemble in cationic ligandgated channels in X. laevis oocytes (19). Similarly, the Vdes-GRD/LCCH3 heteromultimer shows a cationic selectivity suggesting that this feature could be shared by other arthropod species in which these subunits are found. However, the existence of such heteromultimeric receptors in vivo remains an open question. Indeed, overlapping expression of these two subunits in the same cells has never been detected (8,27). It has been suggested that the LCCH3 subunit may assemble with RDL in A. mellifera antennal lobe (8), but in D. melanogaster LCCH3 and RDL exhibit a different pattern of expression (41). Although co-expression of LCCH3 with RDL in Sf21 cells generates an unusual receptor (25), we did not find any evidence for such an association in X. laevis oocytes for the V. destructor LCCH3 and RDL subunits. More experiments are needed to confirm the existence of cationic ligand-gated ion channel in vivo and to determine how subunits with opposite selectivity might assemble to form functional receptors.
In conclusion, we cloned eight GABA-gated ion channel subunits from the honey bee mite V. destructor and demonstrated the expression of three homomultimeric RDL receptors and one heteromultimeric GRD/LCCH3 receptor in X. laevis oocytes. These receptors have distinct biophysical and pharmacological features suggesting that they could specifically regulate GABA signaling in vivo. We also found that in X. laevis oocytes, the VdesRDL subunits assemble in heteromultimeric receptors with specific features. The existence of these heteromultimers in vivo, and their consequence for pest control should be investigated.

Molecular biology
Total RNA was isolated from whole V. destructor specimens or from honey bee brain, and first strand cDNAs were obtained as previously described (42). The 5Ј and 3Ј end regions of the different cDNAs were obtained by rapid amplification of cDNA ends-PCR using the GeneRacer kit (Thermo Fisher) and primers designed based on the sequences deposited in BeeBase, with

GABA-gated ion channels in Varroa destructor
the exception of VdesRDL2 for which we designed specific primers to obtain the 5Ј-end based on the predicted sequence obtained from transcriptomic data deposited in GenBank TM . The cDNA of amRDL (KJ485710) was cloned in the pBS-SK vector (Agilent). For the V. destructor subunits, full-length cDNAs covering the entire ORF were cloned in the pcDNA3.1(ϩ) vector, with the alfalfa mosaic virus sequence immediately before the start codon and the 3Ј-UTR sequence of the X. laevis ␤-globin gene immediately after the stop codon. The cDNAs encoding VdesRDL1 and VdesRDL4 were cloned in-frame with the GFP in pEGFP-N1 vector (Clontech) to obtain RDL1-GFP and RDL4-GFP, and the cDNA of VdesRDL3 was cloned in-frame with mCherry in pmCherry-N1 (Clontech) to obtain RDL3-Cherry. The mutants VdesRDL1mut (S306A and M310T), VdesRDL4mut (A257S), and VdesLCCH3mut (A444S) were obtained with the GeneEditor TM in vitro site-directed mutagenesis system (Promega), following the manufacturer's instructions. Mutations and cDNA integrity were checked by sequencing the entire ORF. For X. laevis oocyte injection, cRNAs were obtained from linearized plasmids using the mMessage mMachine transcription kit (Thermo Fisher) following the manufacturer's instructions. The cRNA concentration was adjusted to 1 g/l. For co-injection of subunits, we prepared a mixture of cRNA at a weight/weight ratio of 1:1, unless otherwise indicated.

Electrophysiology and data analysis
Expressed currents were recorded at room temperature using the two-electrode voltage clamp method 1-3 days after injection. Electrodes were pulled from borosilicate glass and filled with 3 M KCl. Oocytes were clamped at Ϫ60 mV, and GABA-gated currents were recorded with a Geneclamp 500 amplifier (Molecular Devices) and digitized with a Digidata 1200 converter (Molecular Devices) using Clampex software (Molecular Devices). The external solution, NDherg, (96 mM NaCl, 3 mM KCl, 0.5 mM CaCl 2 , 1 mM MgCl 2 , 5 mM Hepes, pH 7.4, with NaOH) was continuously perfused in the recording chamber at the rate of 1 ml/min. GABA (stock solution 500 mM in H 2 O), fipronil (stock solution 200 mM in DMSO), and picrotoxin (stock solution 10 mM in DMSO) were diluted in NDherg solution.
GABA concentration response curves were generated by challenging oocytes with increasing concentrations of GABA. Peak current amplitudes were plotted against GABA concentrations. The curves were fitted with a logistic function using Origin 6.0 (Microcal Software) to obtain the extrapolated maximal response that was used to normalize GABA response curves. Fipronil and picrotoxin inhibition curves were generated by inhibiting the response to a GABA concentration between its EC 30 and EC 50 for the tested receptor. Inhibition curves were normalized to the response induced by GABA in the absence of inhibitor. The concentration of GABA required to obtain 50% of the maximum response (EC 50 ), the concentration of inhibitor required to inhibit 50% of the GABA response (IC 50 ), and the Hill coefficient (n Hill ) were determined with a logistic function using Origin 6.0. The data are presented as the means Ϯ S.E. of n individual oocytes. The statistical significance of the difference between data were determined using the nonpaired Student's t test. The reversal potential of the different receptors was measured in a solution made of 100 mM NaCl, 2 mM MgCl 2 , 5 mM Hepes, pH 7.4, with NaOH with a 400-mslong ramp of voltage from Ϫ80 to ϩ80 mV. For ion exchange experiments, 50, 75, 87.5, and 100% NaCl was replaced by sodium acetate for chloride exchange and by TEACl for sodium exchange. The differences between the reversal potential measured in the different solutions and that measured in control solution (E rev shift) were plotted against the chloride or sodium concentration.