Using a combination of CRISPR/Cas9, behavioural experiments and functional analysis to characterise taste receptors in honeybees

Honeybees (Apis mellifera) need their fine sense of taste to evaluate nectar and pollen sources. Gustatory receptors (Grs) translate taste signals into electrical responses. In vivo experiments primarily demonstrate collective responses of the whole Gr-set, but little is known about the individual impact of receptors. Here, we disentangle for the first time the contributions of three gustatory receptors (AmGr1-3) in sugar sensing of honeybees by combining CRISPR/Cas9 mediated genetic knock-out, electrophysiology and behaviour. AmGr1 responds to multiple sugars. Bees lacking this receptor have a reduced response to sucrose and glucose but not to fructose. AmGr2 acts as co-receptor of AmGr1 in a heterologous expression system, but honeybee knock-out mutants perform normally. Eliminating AmGr3 while preserving AmGr1 and AmGr2 abolishes the perception of fructose but not of sucrose. We thus dissociate the roles of AmGr1, AmGr2 and AmGr3 in honeybee taste perception.


Introduction 30
Honeybees are important pollinators of crops and wild plants world-wide [1]. They depend on floral 31 nectars as their main source of carbohydrates, which are converted into honey stores for provisioning 32 3 A2). Although glucose was present for 60 s, the glucose-induced currents appeared only transiently for 63 about 20 s (see also below).

64
To verify the impact of AmGr1 on overall sugar response, co-expression of only AmGr2 and AmGr3 in 65 oocytes was tested, simulating a honeybee mutant lacking functional AmGr1 receptor. Under this 66 scenario, only fructose-induced macroscopic currents could be observed ( Fig. 1 A3), indicating that 67 AmGr2 does not appear to act as a co-receptor of the fructose-specific receptor AmGr3 [12] [13] in 68 Xenopus oocytes.

69
Testing sugar responsiveness at the behavioural level using the established proboscis extension 70 response assay revealed that AmGr1 honeybee mutants were significantly less responsive to sucrose 71 than wild type bees when the sugar was applied to their antennae ( Fig. 1 B1). The mutants also 72 displayed a reduced responsiveness to glucose ( Fig. 1 B2) compared to wildtype bees. However, the 73 fructose responses of AmGr1 mutants did not differ from those of wild type bees ( Fig. 1 B3).

75
AmGr2 is a co-receptor for sucrose and glucose perception 76 Although expression of AmGr2 in oocytes could be confirmed using a N-terminal fused YFP as a 77 genetically encoded reporter protein (Fig. S2 A), AmGr2-expressing oocytes did not reveal any indicating the assembly of homomeric AmGr2 receptors of low electric activity (Fig. S2 B). Taken 90 together, the sugar-induced inward currents, along with the physical interaction between AmGr2 91 subunits proven with BiFC, suggest that AmGr2 is able to assemble to a functional homomeric channel 92 4 that builds up an ion pore, thus being able to perform ligand-gated channel activity with low conductance 93 in oocytes by itself.

94
Sucrose and glucose stimulation of oocytes co-expressing solely AmGr1 and AmGr3, lacking the 95 expression of AmGr2, led to a very similar current response pattern as that seen for the sole expression 96 of AmGr1, with additional fructose-induced currents upon fructose application (Fig. 2 A3). Honeybee 97 AmGr2-mutants did not differ from wild-type bees in their responses to sucrose, glucose or fructose 98 when tested in the behavioural assay ( Fig. 2 B1, B2 and B3). These findings indicate that AmGr2 on its 99 own does not provide sufficient ion channel performance on a comparable scale such as sugar receptors 100 AmGr1 nor AmGr3 but can unfold its function in honeybee sugar sensation acting as a co-receptor for 101 AmGr1.

103
AmGr3 is a specific fructose receptor

104
The sole expression of AmGr3 in oocytes only led to fructose-induced current responses (Fig. 3 A1).

105
Neither sucrose nor glucose acted as ligands of AmGr3 in our experiments, supporting earlier findings 106 [12] [13]. Oocytes co-expressing AmGr1 and AmGr2, simulating the honeybee AmGr3 knock-out 107 mutant, did not elicit any fructose-induced currents in the cells (Fig. 3 A3). Honeybee AmGr3 108 homozygous mutants displayed a significantly reduced response to fructose compared to wild-type 109 bees, what was found out and published previously [13]. This difference was not observed when tested 110 with sucrose in the PER paradigm. These findings show that AmGr3 is unequivocally a fructose-specific 111 receptor in the honeybee demonstrated by heterologous expression system and behaviour in vivo.

114
Xenopus oocytes expressing all three receptors showed robust current deflections in TEVC when 115 exposed to sucrose, glucose or fructose. Interestingly, glucose-induced inward currents were of

117
Following the decay, glucose-induced steady-state currents reached similar levels of maltose, trehalose 118 and melezitose (Fig. S1 F, bar diagram). This behaviour is also apparent when AmGr3 is absent (Fig. 3 119 A3. Fig. S1 C). However, when AmGr1 was expressed alone, the glucose current trace was stronger 120 and did not decay over time (Fig. 1 A1). In agreement with Jung et al. [14], AmGr2 itself did not show 121 macroscopic sugar-induced currents (Fig. 2 A1) which is well in line with honeybee AmGr2 mutants that 122 did not show significant differences in responses to sucrose, glucose or fructose compared to controls 123 5 ( Fig. 2 B1, B2 and B3). Moreover, our BiFC experiments indicate a direct physical interaction between 124 AmGr1 and AmGr2, strongly suggesting that heteromerization occurs (Fig. S2 B). Thus, AmGr2 seems 125 to act exclusively as co-receptor for the sucrose signal of AmGr1. AmGr2 additionally modulates the 126 strength [14] and the time-dependent characteristics of AmGr1-derived glucose signals.

127
For fructose induced currentscaused by AmGr3-expression as described aboveno obvious 128 modulations could be found when it was co-expressed with the other receptors ( Fig. 1 A3 or Fig. 2 A3).

129
Mutant bees lacking AmGr2 (but expressing AmGr3 and AmGr1) as well as AmGr1 homozygous 130 mutants (with AmGr3 and AmGr2 being present) did not show any significant difference in their fructose 131 response compared to wildtype bees ( Fig. 1 B3 and Fig. 2 B3). Thus, among the tested gustatory 132 receptors, AmGr3 is irreplaceable as the specific fructose receptor [13]. Moreover, its 133 electrophysiological properties are not modulated by any of the other sugar receptors.

136
Taste plays a critical role for honeybees when evaluating and choosing profitable food sources in terms 137 of concentration and type of sugar [28]. Bees rely on pollen as a source of protein and nectar or 138 honeydew as their main sources of carbohydrates. It has been shown that honeybees perceive e.g. the 139 sugars sucrose, glucose, fructose, maltose and melezitose, which are the main constituents found in 140 their food [5] [6] [28]. It is therefore astonishing that honeybees only have a reduced set of gustatory 141 receptors and, among them, only three are sugar receptors (AmGr1-3) [2] [3]. Bees show a broad 142 spectrum of food-related and taste-related behaviours, like division of labour [29] [25] or the ability to 143 memorize sugars and associate odours with them [30]. Differences in sugar receptor expression were 144 discovered as a possible regulatory mechanism of division of labour [3] [31].

145
To elucidate the complexity of sugar perception in honeybees, we carried out a combined study on

157
(sucrose, fructose, glucose, maltose and melezitose) or their blood sugar (trehalose). In contrast, we 158 could not detect any responses to the less relevant sugars such as arabinose, mannose or galactose 159 and none of them contains a glucose unit. Additionally, no signal was generated by raffinose application.

160
We therefore assume that the glucose unit in raffinose is embedded and difficult to access. The

164
The gustatory receptor AmGr2 acts as a co-receptor modulating the specificity of AmGr1 but not of

186
AmGr2 knock-out mutants did not show any significant changes in behaviour in response to the sugars 187 sucrose, glucose and fructose. Though the PER paradigm works very well to fully elucidate receptor 188 activity [13], it might be rather unspecific for the co-receptor AmGr2. For instance, for the fruit fly (which 189 inheres 68 gustatory receptors [2]), it was shown that the neurons are widely distributed throughout the 190 main gustatory organs, each expressing distinctive types and sets of Grs. These heterogeneous Gr 191 populations can be found in the same organ or sensilla [38], they might respond to different sugars [39] 192 and even interact with each other [40]. Assuming a comparable complexity in the honeybee as well as 193 further processing in the brain, the actual influence of the co-receptor AmGr2 in the behavioural outcome 194 might be blurred. This could also be the reason for our finding that an absolute loss of response could 195 not be observed in all mutants [36]. Fixation could also influence natural behaviour [41]. Earlier 196 experiments showed that freely moving bees or bees in cages prefer sucrose over all other sugars [4], 197 [42] comparable to fixed bees in more recent PER experiments [43]. In all TEVC experiments, however, 198 the sucrose signals measured were weaker or similar to those of glucose. The yet uncovered co-199 receptor function of AmGr2 or differences in general receptor expression might be factors modulating 200 the receptor signal and the actual behaviour, but these points need further investigation.

201
Our results clearly show that AmGr3 is a specific fructose receptor. Inward currents induced by this 202 sugar can no longer be measured when AmGr3 is missing in the cell system and the homozygous 203 honeybee mutants are significantly less responsive to fructose. The fructose signal does not seem to 204 be influenced by any other receptor (such as AmGr1 by AmGr2). The homolog of AmGr3 in Drosophila 205 appears to be a sensor for saturation of these insects, displayed by the fructose level, which seems to 206 react to food intake regardless of actual blood sugar [44] [45]. The specificity of AmGr3 might indicate a 207 comparable function in honeybee saturation with sugars.    transfected with the ligation mix and plated on agar plates as described previously [13]. E. coli could 263 grow over night at 37 °C, clones were picked with the blue-white selection and transferred to an 264 overnight culture (as described in detail in [13]). After pelleting, the isolation was performed according        Target-sites for the sgRNAs (single guide RNAs) were found in the first exons of the ORFs of the 300 respective genes (AmGr1-3, see suppl. Tab. S2). Target specific sequences were found via benchling 301 (https://benchling.com, San Francisco, USA) following the criteria mentioned elsewhere [13]. For each 302 11 target an overlapping phusion PCR (NEB, Ipswich, USA) was performed with a forward primer including 303 the T7 promoter and the certain crRNA sequence (specific for each sgRNA, see suppl. Tab. S2) and a 304 stable revers primer (tracrRNA sequence, [13]). PCR products were purified, checked and used as 305 template for the sgRNA synthesis. Purification of sgRNA was performed with the MEGAclear 306 Transcription Clean-Up Kit (Invitrogen, Carlsbad, USA) after the DNAse digestion. As described, the 307 quantified and pre-tested sgRNAs were aliquoted and frozen in their favorable concentration (best 308 hatching rates and mutation rates were determined in pre-tests, data is not shown). During the

319
AmGr2 or AmGr3 and water controls), so that the mutational event occurred in the single-cellular state, 320 leading to fully mutated embryos without mosaic patterns. Roth et al. recently showed that frequently 321 both alleles were mutated (homozygous mutants) and that the entire bee was affected (absence of 322 mosaicism) [20]. Mutations were controlled via next generation sequencing (NGS, after preselection via 323 fluorescence length analysis, FLA). Additionally, Yu and Omholt demonstrated in 1999 that division of 324 the honeybee zygote is completed after 120±6.9 minutes, confirming our choice of procedure [21].

326
Following the protocol of [20], eggs were processed and injected in a 35°C climate chamber, using the 327 same set-up, procedure and material described in our prior work [13] [22]. Each egg was injected with 328 400 pl volume (water or sgRNA/Cas9 as described above, also see suppl. Tab. S2). Eggs were 329 maintained in a sulfuric atmosphere until a few hours before hatching (1 ml of 16 % sulfuric acid per liter 330 of volume, separated from the rings by a grid) and monitored. Hatched larvae were removed carefully 331 with a modified Chinese grafting tool and transferred to prepared Nicot-wells (NICOTPLAST, Maisod, France) containing larval food. Larval rearing was carried out as described earlier [13] based on the 333 protocol of [22]. One wing was removed from the freshly hatched and fully dried bees. Individuals were 334 marked with coloured number plates (Opalith queen-marking plates) using super glue (UHU GmbH & 335 Co. KG, Germany). All marked bees of one replicate (treated with sgRNA/Cas9 and the water controls) 336 were placed into a cage with pollen and sugar water and maintained in an incubator at 35°C.

337
Testing responsiveness to sugars 338 One-week-old bees were tested for their proboscis extension response (PER) to increasing 339 concentrations of sucrose, glucose and fructose (in two replicates per gene) according to [13]. Prior to 340 the test, each bee was immobilized on ice, carefully mounted in brass tubes and fixed with adhesive 341 tape [23]. Subsequently, both antennae were stimulated with a droplet of a certain sugar water 342 concentration. All three sugars (two for AmGr3), alternatingly starting with sucrose, fructose or glucose, 343 were tested. After the water pre-test, an increasing concentration series (16 %, 20 %, 25 %, 32 %, 40 344 %, 50 % and 63 % (w/v)) of a certain sugar was tested. Contaminations at the antennae were 345 immediately removed and rinsed with water. An inter-trial interval of two minutes was used to prevent 346 intrinsic sensitization [24]. It has been shown that the order of sugar water concentration does not 347 influence responsiveness to sucrose [23]. For each sugar and each bee, the positive PER of the

352
Directly after the PER tests, bee's genomic DNA (gDNA) was isolated as described before [13]. The  Illumina HiSeq 2500 (2x250bp, Rapid Run). Bioinformatic analysis followed the strategy with a 358 handwritten Perl script as used previously by Değirmenci et al. [13]. De-multiplexing of the samples was 359 carried out via their identifier tags using HMMer v3.2.1 [26]. All reads (forward and reverse) were merged 360 and subsequently quality filtered (maxEE=1, minlen=100) using USEARCH v11 [27]. We identified and 361 counted variants of each sample making use of USEARCH for dereplication and counting, therefore, 362 13 taking into account both alleles of each individual. In this script, also MUSCLE was used for alignment 363 with the reference and afterwards indel positions were counted. For AmGr1, the alignment was split into 364 segments to cover only the relevant site to account for splice variants at other positions [27] before 365 counting indels. Wild-type alleles were classified as "wt" and in-frame indels with a multiple of 3 bps 366 were marked as "if", resulting in an intact open reading frame and exclusion from our investigation. The 367 label "ns" was used for "nonsense" mutation, since these mutations would result in reading frame shifts 368 (nonsense code), leading to non-functional proteins. We thus followed the proved genotyping approach 369 of Roth et al. [20] and only included animals with a proven homozygous mutant (ns/ns) or homozygous 370 wildtype (wt/wt) genotype.

381
Acknowledgments 382 We thank Markus Thamm for molecular lab expertise and Karin Möller for technical assistance. We

Ethics approval and consent to participate 400
No ethics approval or consent to participate was required for this study.

Competing interests 402
The authors declare that they have no competing interests.

453
AmGr2 mutants (ns/ns) did not show any significant differences in their responsiveness towards all three 454 sugars when compared to wild-type (wt/wt) bees, neither to sucrose (B1; Mann-Whitney-U, ns/ns vs.    The following supporting information is available for this article:  (Fig. 1-3). 26 27