(C) PLOS One
This story was originally published by PLOS One and is unaltered.
. . . . . . . . . .
Multisite imaging of neural activity using a genetically encoded calcium sensor in the honey bee [1]
['Julie Carcaud', 'Evolution', 'Genomes', 'Behavior', 'Ecology', 'Université Paris-Saclay', 'Cnrs', 'Ird', 'Gif-Sur-Yvette', 'Marianne Otte']
Date: 2023-02
Understanding of the neural bases for complex behaviors in Hymenoptera insect species has been limited by a lack of tools that allow measuring neuronal activity simultaneously in different brain regions. Here, we developed the first pan-neuronal genetic driver in a Hymenopteran model organism, the honey bee, and expressed the calcium indicator GCaMP6f under the control of the honey bee synapsin promoter. We show that GCaMP6f is widely expressed in the honey bee brain, allowing to record neural activity from multiple brain regions. To assess the power of this tool, we focused on the olfactory system, recording simultaneous responses from the antennal lobe, and from the more poorly investigated lateral horn (LH) and mushroom body (MB) calyces. Neural responses to 16 distinct odorants demonstrate that odorant quality (chemical structure) and quantity are faithfully encoded in the honey bee antennal lobe. In contrast, odor coding in the LH departs from this simple physico-chemical coding, supporting the role of this structure in coding the biological value of odorants. We further demonstrate robust neural responses to several bee pheromone odorants, key drivers of social behavior, in the LH. Combined, these brain recordings represent the first use of a neurogenetic tool for recording large-scale neural activity in a eusocial insect and will be of utility in assessing the neural underpinnings of olfactory and other sensory modalities and of social behaviors and cognitive abilities.
Funding: We thank the French Research National Agency (ANR to JCS, Project ANR-17-CE20-0003 Bee-O-Choc; ANR to JC, Project ANR-22-CE37-0029 Pherobrain), the CNRS (JCS) and the University Paris-Saclay (JC and JCS), the DFG (Deutsche Forschungsgemeinschaft to MB) and CIMeC, University of Trento (AH) for funding. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
To this aim, we developed the first pan-neuronal genetic driver in a Hymenopteran model organism, the honey bee, and expressed the calcium indicator GCaMP6f under the control of the honey bee synapsin promoter. We characterized its expression pattern in the honey bee brain and evaluated its potential as a functional tool by recording neural activity upon olfactory stimulation. We show that GCaMP6f expression allows to record olfactory responses from multiple brain regions, after simply opening the brain capsule. By using a controlled panel of well-characterized odorants, we show that the recorded signals reveal robust odor coding rules. This new pan-neuronal genetic driver also permits to record neural activity from poorly recorded regions of the brain such as the lateral horn (LH) and the mushroom body (MB) calyces and allowed to find that olfactory chemical features are less represented in the LH than in the antennal lobe. This study opens new possibilities for neuroethological research in the honey bee to study the neural basis of advanced social behaviors and cognitive skills.
Despite the difficulty, the recent use of the genome editing tool CRISPR/Cas9 in the honey bee allowed to knock out specific genes, such as the sex-determining dsx gene [ 35 ], the olfactory co-receptor gene orco [ 36 ], the gustatory receptor AmGr3 gene [ 37 ], or the Amyellow-y gene [ 38 ]. Even if these studies evaluated the effect of specific mutations at the neuronal [ 35 ] or the behavioral levels [ 37 ], versatile genetic tools allowing to investigate the neural basis of a wide range of higher-order social behaviors and learning in honey bees are still missing. The advent of genetically encoded neural activity sensors, in particular calcium sensors, has represented a major breakthrough in Drosophila or mouse research [ 39 – 41 ], but such a critical tool for circuit dissection is still lacking in Hymenoptera.
For now, honey bee neuroscience remains limited to the use of conventional neuroanatomy [ 16 ], pharmacology [ 17 ], electrophysiology [ 18 ], and imaging tools [ 13 , 19 , 20 ]. However, the development of neurogenetic tools during the 20th century has provided unprecedented progress in our understanding of the neural basis of behavior in other model species. Among others, this approach allowed to decipher the brain circuits underlying aggressive behavior [ 21 ], courtship [ 22 ], or memory [ 23 ] in a limited set of model species, both vertebrates such as mouse or zebrafish and invertebrates. In insects, the fruit fly Drosophila melanogaster has been for many years the leading model for investigating the neural basis of behavior, from gene expression to neural circuits [ 24 – 26 ]. Recently, neurogenetic approaches have been developed in other Diptera such as mosquitoes [ 27 , 28 ] and in some Lepidoptera [ 29 , 30 ] to understand specific behaviors such as human–host seeking [ 31 , 32 ] or insect–plant interactions [ 33 ], mainly focusing on their olfactory capabilities. However, genetic methods are particularly difficult to apply in eusocial insects, since genetic transformation rates are low, endogenous promoters for neuronal expressions are unknown and the genetically manipulated, reproductive individuals (the queens) have to be maintained in larger colonies with workers in containments [ 34 ].
Previous research found that this sophisticated social behavior is associated with higher cognitive abilities in these insects, and in this context, honey bees are a mainstream model for studying higher-order insect cognition such as navigation [ 7 ], rule learning [ 8 ], social learning [ 9 ], dance communication [ 10 ], and also olfactory perception and learning [ 11 – 13 ]. However, whether the evolutionary rise of these sophisticated social behaviors is associated with the formation of specific neuronal pathways and structures has not been sufficiently resolved [ 14 , 15 ]. Progress in understanding these sophisticated behaviors and how social cues are processed and integrated with higher-order cognitive abilities have in part been limited due to the lack of tools allowing to measure neuronal activity simultaneously in different regions of the honey bee brain.
Sociality is classified as one of the major transitions in evolution, and animals often form social groups because the benefits of grouping (either direct or indirect) outweigh the costs of breeding independently [ 1 , 2 ]. The most advanced level of sociality is found in eusocial insect societies [ 3 , 4 ], and the insect order Hymenoptera (including ants, bees, and wasps) presents the largest number of eusocial species. Among them, honey bees are a classical model for the study of eusocial behavior, as they live in colonies composed of up to 60,000 individuals, consisting of 3 adult castes (queen, worker, and male). Within the worker caste, honey bees show a clear division of labor with a specialization of roles [ 5 ]. The success of honey bee colonies lies in the capacity of all members of the society to behave in a well-organized and context-dependent manner, a social behavior mediated in part by olfactory cues such as pheromones used for communication within the colony [ 6 ].
( A ) GCaMP6f calcium signals in the LH evoked by a panel of pheromonal compounds, both volatile (IPA and ocimene in black) or not volatile (queen pheromone compounds in red and brood pheromone compounds in blue) and to the controls (isopropanol and air). Different odorants induce different activity patterns. Relative fluorescence changes (ΔF/F [%]) are presented in a false-color code (see colorbar), from dark blue (minimal response) to red (maximal response). ( B ) Amplitude of calcium responses (ΔF/F [%]) to the 12 tested odorants and to the controls (air and isopropanol). Volatile odorants and queen pheromone compounds induce significant activity in comparison to the controls (n = 7 honey bees, * p < 0.05). The data underlying the graphs shown in the figure can be found in S4 Data . LH, lateral horn.
A condition for using GCaMP bees for unraveling the neural basis of social communication is the ability to record robust responses to pheromonal odorants. We focused again on the LH and presented a panel of honey bee pheromonal odorants, including both volatile (IPA, from the alarm pheromone and ocimene, a volatile brood pheromone) and non-volatile compounds (queen and brood pheromone compounds). Calcium signals were visible for all compounds in the LH ( S4 Data and Fig 6A ), although the response intensity was significantly higher from the controls ( Fig 6B , n = 7 honey bees) for volatile compounds (t tests, t > 4.35, p < 0.005) and for queen pheromone compounds (Friedman ANOVA, odor effect F 11 = 30.23, p = 0.0008; t tests or Wilcoxon tests, t > 2.48, p < 0.05 except for 9-ODA with t = 1.8 and p = 0.11) but not for brood pheromone compounds (t-tests t < 2.09, p = NS).
( A ) Calcium signals in the calyces of the MBs in response to the presentations of 2-heptanone, heptanal, octanal, and the air control. ( B ) Example time courses (left, taken from the black square shown in A in 2-heptanone) or average time courses (right, n = 6 honey bees) of odor-evoked responses (ΔF/F [%]) recorded in the lip of the MB to 2-heptanone (in red), octanal (in blue), and to the air control (in gray). Calcium signals are weaker here than in other structures, with a smaller first component and a larger second component. ( C ) Amplitude of calcium responses (ΔF/F [%]) to the 5 tested odorants and to the air control. All odorants induce significant activity in comparison to the air control (n = 6 honey bees, * p < 0.05, except 2-heptanone with p = 0.07). The data underlying the graphs shown in the figure can be found in S3 and S7 Data files. MB, mushroom body.
Lastly, we turned to the second higher-order center, the MBs ( Fig 5 and S3 Data ), a multisensory integration center in the honey bee brain. Calcium signals upon odorant presentation were observed at the level of the calyx lip. Calcium signals systematically showed a biphasic time course (Figs 5B and S8 Fig and S7 Data ) like that recorded in the AL and in the LH, but in all cases with a small first component and a much stronger second component. In this structure, the 5 presented odorants induced significant activity in the MB calyx in comparison to the air control ( Fig 5C ; n = 6 honey bees, RM-ANOVA, odor effect F 5,25 = 7.99, p = 0.0033, post hoc Dunnett tests p < 0.05, except 2-heptanone with p = 0.07). However, the signal-to-noise ratio was considered too low for further analysis of inter-odor relationships.
We then compared similarity relationships (Euclidian distances among odorants, S2 Data ) measured in the LH and in the AL and found a weak but still significant correlation (R 2 = 0.07, Mantel test p = 0.023). This suggests a transformation of odor similarity relationships between the 2 structures. A cluster analysis performed on Euclidian distances in the LH confirmed this finding ( Fig 4D ). The analysis roughly segregated alcohols (-OH moiety) on the one side and aldehydes/ketones (= O moiety) on the other. However, different rules apply here, as the 2 longest chain molecules nonanal and 2-nonanone clustered with the former group, while the shortest chain secondary alcohols, 2-hexanol and 2 heptanol clustered with the latter group. Thus, the study of odor coding in the LH using GCaMP bees suggests a less clear dependence on chemical features than in the AL.
( A ) Calcium signals in the LH evoked by the same panel of 16 odorants. Different odorants induce different activity patterns in the LH. ( B ) Example time courses (top, taken from the black square shown in A in C7 ketones) and average time courses (bottom, n = 8 honey bees) of odor-evoked responses (ΔF/F [%]) recorded in the LH (in the black square shown in A) to 2-heptanone (in red), to the air control (in gray) and to octanal (in blue, only shown for individual time course). The calcium signals also show a biphasic response, with a fluorescence increase upon odor presentation (blue bar) followed by a long undershoot. ( C ) Amplitude of calcium responses (ΔF/F [%]) to the 16 aliphatic odorants and to the air control. All odorants induce a significant activity in comparison to the air control (n = 8 honey bees, * p < 0.04). ( D ) Cluster analysis showing similarity relationships among odorants (Ward’s classification method). The data underlying the graphs shown in the figure can be found in S2 and S6 Data files. LH, lateral horn.
We then studied olfactory information processing in higher-order brain centers, first focusing on the LH. All the aliphatic odorants used in this panel induced calcium signals in the LH ( Fig 4A ). Odor-evoked signals systematically followed a biphasic time course, like that recorded in the AL (Figs 4B and S7 Fig and S6 Data ). In the LH, all but one of the presented odorants induced a response that was significantly higher than to the air control ( S2 Data and Fig 4C , n = 8 honey bees, RM-ANOVA, odor effect F 16,112 = 8.091, p = 0.0014, post hoc Dunnett tests, p < 0.04 except p = 0.09 for 2-hexanol). As in the AL, response amplitudes correlated with odorants’ vapor pressures (R 2 = 0.49, F 1,14 = 13.87, p = 0.0023). However, this correlation appeared to be weaker in the LH with a near-significant difference between the 2 correlation coefficients (R 2 = 0.84 in the AL versus R 2 = 0.49 in the LH, Fisher test, z = 1.76, p = 0.078). In other words, response intensity in the LH was not a simple product of the number of volatile molecules in the odor puff, possibly revealing a yet undocumented gain control mechanism in the LH.
Finally, we asked if odorant response maps recorded with the GCaMP bees relate to bees’ actual perception of the odorants. We thus correlated the Euclidian distances measured between all odor pairs in this study with distances measured in an appetitive conditioning experiment ( S8 Data ) [ 49 ]. We found that behavioral distances can be predicted using signals recorded with genetically encoded GCaMP6f ( S6 Fig ; R 2 = 0.34, Mantel test p = 1.0 × 10 −4 ). Thus, odorants evoking similar activity patterns in the AL are treated as similar by honey bees in their behavior, a further confirmation that GCaMP6f expression allows to record meaningful neural signals.
We then evaluated odor coding in the GCaMP bees by comparing similarity relationships among odorants, by calculating Euclidian distances between response maps obtained for the different odor pairs ( S1 Data ). First, we confirmed that within each bee, each odorant evokes a specific activity pattern, as distances were lower for different presentations of the same odorant than for the presentation of different odorants ( Fig 3E , n = 11 honey bees, paired t test t = 6.15, p = 1.08 × 10 −4 ). This result fits with previous calcium imaging data both at AL input and output [ 47 , 48 ]. We then performed a hierarchical cluster analysis using Euclidian distances between odorants. We found a first segregation between odorants with long chain lengths and odorants with short chain lengths ( Fig 3F , C6-C7 versus C8-C9). Within the shorter chain length group, a further separation between functional groups appeared based on the oxygen moiety (C = O for aldehydes and ketones versus C-OH for alcohols). Within the longer chain length group, aldehydes were separated from the other odorants. This analysis shows a clear coding depending on the chemical structure of the odorants and with remarkably similar results as in previous studies using non-transgenic approaches [ 19 , 46 ]. Accordingly, inter-odorant distances found in the GCaMP bees were highly significantly correlated with those measured previously in OSNs using calcium dye bath application (R 2 = 0.53, Mantel test p = 1.0 × 10 −4 ) or measured in PNs using dye injection (R 2 = 0.44, Mantel test p = 1.0 × 10 −4 ). The correlation coefficients were not significantly different (R 2 = 0.53 versus R 2 = 0.44, Fisher test, z = 0.98, p = 0.33, NS), suggesting the possibility that both OSN and PN contribute to the signals recorded in the AL using genetically encoded GCaMP6f.
( A ) GCaMP6f calcium signals in the AL evoked by a panel of 16 odorants varying systematically according to their carbon chain length (C6–C9) and their functional group (primary and secondary alcohols, aldehydes, and ketones). Different odorants induce different glomerular activity patterns. The map shows the whole amplitude of the response including both positive and negative components (see text for calculation). ( B ) Example time courses (top, taken from the black square shown in A in C7 ketones) and average time courses (bottom, n = 11 honey bees) of odor-evoked responses (ΔF/F [%]) recorded in the AL to 2-heptanone (in red), to the air control (in gray) and to octanal (in blue, only shown for individual time course). ( C ) Amplitude of calcium responses (ΔF/F [%]) to the 16 aliphatic odorants and to the air control. All odorants induce significant activity in comparison to the air control (n = 11 honey bees, ** p < 0.0097). The intensity of the odor-induced response was obtained by averaging 3 consecutive frames at the end of the odor presentation (frames 19–21) and subtracting the average of 3 frames during the second, negative component of the signal (frames 49–51). ( D ) Amplitude of calcium responses (ΔF/F [%]) as a function of odorant vapor pressure (in log units). The linear regression shows a significant correlation (R 2 = 0.84, *** p = 6 × 10 −7 ). ( E ) Different presentations of the same odorants (2-octanone and heptanal) show similar glomerular patterns in the AL (left). Dissimilarity measures (Euclidian distance, right) between representations of the same or of different odorants. Activity maps are more similar (shorter distances) when the same odorant is presented (*** p = 1.1 × 10 −4 ), showing, as expected, a clear odor coding in this structure. ( F ) Cluster analysis showing similarity relationships among odorants (Ward’s classification method). Functional groups are shown in different colors: primary alcohols in blue, secondary alcohols in green, aldehydes in black, and ketones in red. The analysis shows a first separation between odorants with short and long carbon chain lengths. Odorants with a short carbon chain are then subdivided into alcohols (primary and secondary, C-OH functional group) and ketones/aldehydes (C = O functional group). The data underlying the graphs shown in the figure can be found in S1 and S5 Data files. AL, antennal lobe.
We then asked if the signals recorded using this genetically encoded calcium sensor represent neural signals that are meaningful in terms of sensory coding, focusing first on the primary olfactory center, the AL, since olfactory coding was best studied in this structure. To this aim, we presented 16 aliphatic odorants used in previous imaging studies either after bath-application of a cell-permeant calcium-sensitive dye (Calcium Green 2-AM, OSN recordings) [ 46 ] or after insertion of a migrating calcium-sensitive dye within a neural tract (Fura-2 dextran, PN recordings) [ 19 ]. The response patterns to the 16 aliphatic odors differed systematically according to their functional group (primary or secondary alcohols, aldehydes, or ketones) or their carbon chain length (from 6 to 9 carbons) ( Fig 3A ). As observed with other calcium reporters, the presentation of each odorant induced a biphasic signal in a different set of AL glomeruli while presentation of the air control or no stimulation did not induce such response (Figs 3B and S4A Fig and S9 Data files). Moreover, WT bees did not show any signal in response to odor presentation ( S4B Fig ). The averaged AL response amplitudes were significantly different from the air control for all 16 odors ( Fig 3C and S1 Data ; n = 11 honey bees, RM-ANOVA, odor effect F 16,160 = 12.49, p = 8 x 10 -21 , post hoc Dunnett tests p < 0.0097). Signal amplitudes differed according to the odor (odor effect, p = 8 × 10 −21 ) and were related to the quantity of molecules in the headspace, since odor-evoked responses were highly correlated with the vapor pressure of the odorants ( Fig 3D , n = 11 honey bees, R 2 = 0.84, Fisher test F 1,14 = 73.72, p = 6 × 10 −7 ). Such a correlation was also found in previous studies recording at AL input (OSNs) [ 46 ] or output (PNs) [ 19 ]. This strong dependence explains why odor-evoked intensities recorded using genetically encoded GCaMP6f in this study are highly correlated with odor-evoked intensities recorded in OSNs ( S5A Fig , R 2 = 0.84, F 1,14 = 74.23, p = 6 × 10 −7 ) or in PNs ( S5B Fig , R 2 = 0.78, F 1,14 = 49.73, p = 6 × 10 −6 ).
( A ) Hymenopteran olfactory pathway (adapted from [ 46 ]). Odorant molecules are detected by OSNs on the antenna, which send olfactory information to the AL. Then, PNs convey information to higher-order centers, the MBs and the LH. Lo: lobula, Me: medulla, vL: vertical lobe, hL: horizontal lobe. ( B ) Calcium signals in the whole brain evoked by 3 different odorants (2-heptanone, octanal, heptanal) and the air control. Relative fluorescence changes (ΔF/F [%]) are presented in a false-color code, from dark blue (minimal response) to red (maximal response). ( C ) Time course of odor-evoked responses (ΔF/F [%], taken from the black squares shown in B) for 1 individual to the presentation of 2-heptanone (in pink), octanal (in blue), and the air control (in gray), simultaneously recorded in the AL (left), the LH (middle), and the MB (right). The calcium signals show a biphasic time course, with a fluorescence increase upon odorant presentation (blue bars) followed by a long-lasting fluorescence undershoot, in all brain regions. AL, antennal lobe; LH, lateral horn; MB, mushroom body; OSN, olfactory sensory neuron; PN, projection neuron.
GCaMP6f expression in neurons was then used to record neural activity in the honey bee brain ( Fig 2 ). As a proof of concept for the use of these bees in neuroethological research, we decided to focus here on olfactory information processing, as olfaction is the best-understood sensory modality in honey bees. Right after placing the bee in a recording chamber and opening its head capsule, we first presented a few standard odorants while recording the whole brain surface accessible under the microscope objective. Odorant presentations triggered a clear GCaMP fluorescence change from several brain regions ( Fig 2A ), including the primary olfactory centers (the ALs) and both higher-order centers (the LHs and the MBs), known for their role in olfactory processing. The use of GCaMP-expressing bees allowed us to record neural activity simultaneously in all these structures ( Fig 2B ), while the presentation of an unscented air control did not induce strong activity. Response amplitudes were stronger in the AL ( Fig 2C left) than in the LH ( Fig 2C middle) or in the MB ( Fig 2C right). In all regions, the recorded calcium signals showed a biphasic response, with a fluorescence increase upon odor presentation followed by a long undershoot. Such biphasic signals are reminiscent of calcium signals previously recorded using bath-applied calcium-sensitive dyes [ 45 , 46 ]. This experiment represents the first recording of neural activity using neurogenetic tools outside of Diptera.
We next asked whether the cloned synapsin promoter used for GCaMP6f expression drives the same expression pattern as the one of the synapsin gene. To answer this question, we performed a complementary immunostaining targeting the synapsin protein ( Fig 1B right, anti-SYNORF1 in red). We found co-localized staining of GCaMP6f and synapsin within AL glomeruli both in the cortex and the core ( Fig 1C right), demonstrating co-expression within AL neurons. This experiment also shows differences between GCaMP6f and synapsin stainings. The synapsin immunostaining allows to clearly visualize presynaptic zones, for instance, in AL glomeruli or microglomeruli in the MB calyces, but is not present in the somata unlike GCaMP6f ( Fig 1D , white arrows). This result indicates that even if GCaMP6f is expressed in cells expressing synapsin (in theory all neurons), these 2 proteins appear to be partially differently localized within neurons.
Within neurons of the transgenic bees expressing GCaMP6f, we found evidence of staining in somata (Figs 1B (left) and S2A and S2B (white arrows)), neuronal processes (see for instance, neural tracts around the α-lobe; Fig 1B left, white arrows) as well as dendrites and terminal projections. At the level of the AL (Figs 1C and S2 ), lateral and medial clusters of projection neurons (PNs) and local interneurons somata show strong GCaMP6f immunostaining. In addition, glomeruli seem homogeneously stained: staining in the cortex suggests expression in olfactory sensory neurons (OSNs) terminal projections, whereas staining in the glomerulus core suggests expression in PNs dendrites and possibly local neurons. Finally, the clearest expression of GCaMP6f was found in somata at the level of the calyces of the MB (Figs 1D and S2B ). In these structures, different neuron types are stained, with somata of class I Kenyon cells in the cup-shaped calyces and class II Kenyon cell somata lying outside the calyces (see white arrow in Figs 1D and S2B ) [ 43 , 44 ]. These staining patterns were similar in the different bees we used for the immunostaining ( S2A and S2B Fig ). We found very broad expression in some groups of neurons like the somata of antennal lobe (AL) neurons (example shown in S3A Fig for a somata cluster of local/projection neurons in caudo-lateral position) and at the same time heterogeneous staining at the level of the Kenyon cells within MB calyces. We also found expression of GCaMP6f in neurons outside of the central brain, for example, in the photoreceptors of the ocelli ( S3B Fig ).
We first studied the pattern expression of GCaMP6f in the brain using immunostaining against GFP (Figs 1B–1D and S2 ), the green fluorescent protein that the GCaMP6f sensor contains. We found clear and widespread staining throughout the brain, with strong staining in the antennal lobes (ALs), the optic lobes (OLs), and the MBs, suggesting that the GCaMP6f protein is ubiquitously expressed in all major brain structures. In comparison, wild-type (WT) bees that do not express the GCaMP6f protein show no staining throughout the brain ( S2C Fig ).
( A ) Scheme of the honey bee synapsin promoter GCaMP6f expression cassette. CDS: coding sequence; UTR: untranslated region; ATG: translation start. ( B ) GCaMP6f expression in the honey bee brain revealed by anti-GFP immunostaining (left, in green). GCaMP6f is widely expressed in the bee brain, including in somata and neural tracts (white arrows). For comparison with synapsin expression, an anti-SYNORF1 immunostaining (right, in red) is superimposed on the anti-GFP signal in green. Scale bar = 200 μm. ( C ) GCaMP6f expression (anti-GFP in green) and synapsin expression (anti-SYNORF1 in red) in the antennal lobe. Remarkable and strong expression is observed in the somata of projection neurons and local neurons (near the AL, white arrows). Scale bar = 50 μm. (D) GCaMP6f expression (anti-GFP in green) and synapsin expression (anti-SYNORF1 in red) in the mushroom bodies with strong expression in some somata of Kenyon cells (in the cup of the calyces, see white arrows). Scale bar = 50 μm. AL: antennal lobe, MB: mushroom body, OL: optic lobe, Lo: lobula, Me: medulla, vL: vertical lobe, lc: lateral cluster of antennal lobe neuron somata, mc: medial cluster of antennal lobe neuron somata, Li: lip, BR: basal ring, Co: collar.
In order to generate a bee with a possibly pan-neuronally expressed calcium sensor, we introduced a synapsin (syn) promoter GCaMP6f expression cassette (syn-GCaMP6f) into the genome using the PiggyBac transposon system [ 34 ]. To generate this syn-GCaMP6f expression cassette, we cloned the 1 kb promoter region together with the entire 5′ untranslated region (5′ UTR) of the synapsin gene from the honey bee ( S1 Fig ), which we then fused with the coding sequence of the GCaMP6f sensor protein ( Fig 1A ). We obtained transgenic syn-GCaMP6f queens using previously published procedures and a hyperactive transposase [ 34 , 42 ]. We instrumentally inseminated the queens and reared from only 1 F0 queen the offspring second-generation queens (F1). Only 1% of those F1 queens were carrying the transgene, which we identified from transgene amplifications. The second-generation syn-GCaMP6f queens produced 50% syn-GCaMP6f worker offspring bees, which we used in the following neuroanatomy and imaging experiments.
Discussion
We developed a line of honey bees with pan-neuronal expression of a genetically encoded calcium sensor allowing to record neural activity from honey bee neurons. To our knowledge, this represents the first neurogenetic tool for recording neural activity in a non-dipteran insect and eusocial insect. Neurogenetic tools, especially the use of genetically encoded calcium indicators (GECIs), have been widely developed in both vertebrates [41,50] and invertebrates [51] and supported major progress in our understanding of the neural basis of behavior. In insects, such possibility was limited until now to a few Dipteran species (fruit flies and mosquitoes [27]) and was lacking in other insects with different lifestyles. The present development in a social Hymenoptera represents a major progress for the neuroethology of social behavior and also a significant achievement in a species like the honey bee in which neurogenetic tools are difficult to establish due to its peculiar reproductive biology and the lack of knowledge of appropriate promoter sequences driving the expressions.
To develop our pan-neuronal calcium indicator, we expressed GCaMP6f under the control of the promoter of the gene coding for synapsin, a vesicle-associated presynaptic protein found throughout the honey bee brain [52]. This strategy was successful, as honey bees displayed broad labeling with GCaMP6f in the major brain structures of the nervous system at sufficient amounts for performing calcium imaging. Our neuroanatomical data confirm significant expression throughout the bee brain (optic lobes, antennal lobe, mushroom bodies, and all other regions of the protocerebrum), but we noticed substantial cell-to-cell variation in some regions, for instance, in the cup of the MB calyx, which contains the somata of type I Kenyon cells (see Fig 1C): There, a strong contrast appeared among neurons, suggesting differences in the level of GCaMP expression among Kenyon cells or even some Kenyon cells lacking GCaMP expression. The GCaMP staining pattern did not correspond to any of the Kenyon cell subtypes previously described based on their differing gene expression profiles [44]. In contrast to the synapsin protein, GCaMP6f is located in the entire neuron, since we detected staining in the somata (Fig 1C), the neurites, and the synaptic regions. This widespread expression could represent a great advantage for performing functional recordings in the bee brain.
To assess the power of this new tool for studying brain activity, we focused on the olfactory system, which is the best-known sensory system in this animal. We compared the results obtained using the GCaMP6f bees with previous results acquired using conventional dye injections or bath application [19,46]. For the first time, we could observe the simultaneous activation of the 3 main brain structures involved in insect olfaction, the antennal lobe, the LH, and the MB calyx.
While odor stimuli elicited responses well confined to these structures, also pure air stimuli produced neuronal activation but distributed over further central brain regions. These could be responses of various mechanosensory neurons, which in principle should be largely suppressed by the olfactometer keeping a constant airflow. However, small fluctuations during the switching phases are inevitable. Besides neuropils that process mechanosensory information from the Johnston’s organ, such as the dorsal lobe, the posterior protocerebral lobe, and the subesophageal zone [53], also MB [54] and LH [55] are known to integrate mechanosensory information and recently mechanosensitivity has also been reported in olfactory neurons [56,57].
The possibility to record from different brain structures at the same time is a first and important step towards the long-term goal to understand brain circuits responsible for bees’ sophisticated social behaviors and cognitive abilities that may unravel yet undescribed sensory and/or behavior-related pathways.
To follow this, we first validated our tool using individual structures and we confirmed its validity for measuring relevant neural activity with regard to bees’ behavior. Recordings obtained using the GCaMP6f-expressing honey bees revealed the same olfactory coding rules in the AL as in previous studies [19,58]. Thus, odor-induced activity followed the same intensity rule (depending on the odorant vapor pressure—Fig 3D), and inter-odorant neural distance measures were organized according to odorants’ chemical features, chain length, and functional group. Accordingly, calcium imaging measured with the GCaMP-expressing bees allowed to predict how similarly honey bees perceive these odorants, since inter-odor relationships among GCaMP response maps correlated with inter-odor behavioral distances previously measured in an associative conditioning experiment [49]. This result, even if limited to a single sensory modality, is a strong confirmation of the biological pertinence of the signals measured with this new neurogenetic tool.
In order to demonstrate that the activity can be followed to higher-order brain structures, we studied the role of the LH in encoding the chemical features of our odorants. The LH is a poorly studied structure in Hymenoptera, which could only be recorded previously using invasive dye injection [59]. Importantly, while previous recordings were measured only from a limited subpopulation of PNs in this structure (the l-ALT), the GCaMP6f-expressing honey bees allowed recording from the whole structure. The recordings in response to the 16 aliphatic odorants showed that odor coding in this higher-order center is less clearly influenced by odorant quantity (i.e., lower response amplitude differences among odorants) or by the chemical structure of the odorants than in the AL. These first results suggest the existence of an additional gain control mechanism in this structure, fitting with current discussions on its more diverse role in Drosophila [60]. This includes a possible coding of hedonic odorant valence rather than chemical features [55,61], a segregation of food odors from pheromones [62], a central involvement in courtship behavior [63], or a multisensory integration of olfactory, visual, mechanosensory, and gustatory information [55,61]. The tool presented here will allow for comparable studies also in the honey bee.
Calcium signals were also recorded in the mushroom body calyx, and odorants clearly activated the lip, the olfactory input region of the calyx, as observed previously with a bath-applied dye [64]. The signal amplitude was generally low, which reproduces the known sparseness of the MB odor code [65]. Together with the random arrangement of PN terminals and KC dendritic arbors in the lip, this prevented the observation of odor-specific response maps at this point. However, advanced 3D microscopy techniques [66] will benefit from this neurogenetic marker to unravel coding mechanisms also in this neuropil, as experiments in Drosophila promise [67].
This new tool will provide a range of advantages over previous solutions. We expect that, thanks to the genetically encoded calcium sensor, it will facilitate imaging from other members of the colony, like queens and drones, in contrast to what has been done until now (an exception is [68]). Furthermore, it will also allow to extend the study of sensory processing to other sensory modalities than the most commonly studied olfaction and vision. For instance, the study of mechanosensory information processing notably used during the waggle dance remains limited [57,69] and should be facilitated using GCaMP-expressing honey bees. Lastly, transgenic bees could be instrumental for studying neural signals in response to multisensory inputs, for instance, in the MBs, the higher-order center known for its role in multimodal sensory integration [70].
The future of insect neuroethology beyond the use of model organisms like Drosophila will depend on the further development of neurogenetic tools in species like honey bees, with their rich behavioral repertoire involving intricate social behaviors as well as elaborated cognitive skills. First, the use of the CRISPR-Cas9 strategy, already developed in the honey bee [35], possibly allows target-specific insertion with an integration rate higher than the rate achieved with the PiggyBac transposon system. It should be noted that once a transgenic queen is produced, the line can be kept for longer periods of time by regularly raising new queens from the eggs laid by the initial queen. This procedure involves relatively standard beekeeping practices. The future neurogenetic tools will also have to improve the amplitudes of the calcium responses, with a better signal-to-noise ratio, a necessary condition to study sensory coding in higher-order centers. A potential strategy would be to amplify the GCaMP expression using the UAS-Gal4 or the Q binary expression systems commonly used in fruit flies and mosquitoes [71]. Lastly, a better resolution will have to be achieved using cell-type-specific gene expression to visualize (i.e., using fluorescent proteins such as GFP) or monitor activity (using GECIs as here) from specific neuronal populations. The same precision shall then be exploited to silence or activate specific neuron populations, allowing a temporal control of neural function. We hope that the new line of GCaMP-expressing bees will represent the first step in this direction.
[END]
---
[1] Url:
https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3001984
Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.
via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/
