Comparative analysis of behavioral and transcriptional variation underlying CO₂ sensory neuron function and development in Drosophila

ABSTRACT

Carbon dioxide is an important environmental cue for many insects, regulating many behaviors including some that have direct human impacts. To further improve our understanding of how this system varies among closely related insect species, we examined both the behavioral response to CO₂ as well as the transcriptional profile of key developmental regulators of CO₂ sensory neurons in the olfactory system across the Drosophila genus. We found that CO₂ generally evokes repulsive behavior across most of the Drosophilids we examined, but this behavior has been lost or reduced in several lineages. Comparisons of transcriptional profiles from the developing and adult antennae for subset these species suggest that behavioral differences in some species may be due to differences in the expression of the CO₂ co-receptor Gr63a. Furthermore, these differences in Gr63a expression are correlated with changes in the expression of a few genes known to be involved in the development of the CO₂ circuit, namely dac, an important regulator of sensilla fate for sensilla that house CO₂ ORNs, and mip120, a member of the MMB/dREAM epigenetic regulatory complex that regulates CO₂ receptor expression. In contrast, most of the other known structural, molecular, and developmental components of the peripheral Drosophila CO₂ olfactory system seem to be well-conserved across all examined lineages. These findings suggest that certain components of CO₂ sensory ORN development may be more evolutionarily labile, and may contribute to differences in CO₂-evoked behavioral responses across species.

Introduction

Odor-guided behaviors are important for the survival of many different species and can vary dramatically, even among closely related taxa. During evolution, different sets of ecological constraints can create divergent behavioral responses to certain chemicals among closely-related species by reconfiguring the structure and function of the olfactory system underlying these behaviors.^1,2 However, how evolution modifies the genetic and developmental mechanisms underlying these behavioral differences remains poorly understood.³

Carbon dioxide (CO₂) is an important environmental cue for many insects, regulating many behaviors including some that have direct human impacts. For mosquitoes and other blood-feeding insects, CO₂ is a highly attractive cue and an important component of host-seeking behavior.⁴ On the other hand, CO₂ is a repulsive cue that elicits a robust aversive response for the ubiquitous laboratory model species D. melanogaster.⁵ Previous studies have suggested that these behavioral differences are associated with changes in the components of the CO₂ receptors, as well as changes in the organization of the CO₂ circuits.^6-8

One advantage of studying the insect olfactory system is that the structure and development of the system is fairly well-understood, especially in D. melanogaster. Studies on D. melanogaster have shown that fruit flies detect volatile chemicals via olfactory receptors expressed on olfactory sensory neurons (ORNs) in 2 olfactory appendages, the antenna and the maxillary palps. These appendages are covered with fine sensory hairs called sensilla that house a cluster of between 1 and 4 ORNs. Sensilla are categorized both by morphology and by the combinations of ORN classes that they house.^9,10 Additionally, each ORN class connects to a unique position on the antennal lobe of the fruit fly brain, called a glomerulus, where it forms synapses with higher order neurons.^1,11

In the antennae, each ORN cluster arises through asymmetric divisions of a single multipotent precursor cell. The identity of each precursor cell determines the number and the identity of the ORNs in each sensillum and is defined over 3 stages during the development of the antennae. First, the combinatorial expression of several cross-regulatory pre-patterning transcription factors patterns the antennal disk along the proximodistal and anteroposterior axis, and generates zones with unique differentiation potentials.^12,13 Next, selection of sensory organ precursors occurs within each zone through the expression of the proneural genes atonal and amos.^14,15 Finally, once a multipotent precursor is selected, it undergoes asymmetric cell divisions, regulated by Notch signaling and terminal selector transcription factors, to generate the ORN clusters for each sensilla subtype.^16-18 Transcription factors that regulate ORN expression in the maxillary palps have also been identified.¹⁹ Changes in the transcriptional profiles of the regulatory modules change these developmental programs, which restructures the olfactory circuits ultimately leading to changes in odor-guided behaviors.²⁰

In D. melanogaster, the CO₂ receptor is a heterodimer of 2 proteins: Gr21a and Gr63a.⁵ This receptor is expressed in the ab1 sensilla subtype from the large basiconic class of sensilla. The CO₂ ORNs that house this receptor connect to the ventrally-located V-glomerulus in the antennal lobe to mediate repulsive behavior.⁶ Developmentally, several of the transcription factors that define the ab1 sensilla have been identified.¹² Additionally, the expression of Gr21a and Gr63a in Drosophila species has been shown to be regulated in part by the MyB-MuvB/dREAM complex through the histone methyltransferase su(Var)3–9.²¹Previous studies have found that several proteins of the complex, such as MyB and mip130, are direct regulators of Gr63a expression, while others like mip120 and E2f2 are negative regulators. Interestingly, in D. melanogaster the microRNA miR-279 and the transcription factor prospero are required to inhibit the generation of CO₂ ORNs in the maxillary palps. Mutations in these genes generate ectopic addition of CO₂ ORNs to the maxillary palps in addition to the population in the antennae.⁶ More recently, species-specific differences in CO₂ behaviors were reported for a few Drosophila species. D. virilis and D. suzukii show decreased repulsion from CO₂ compared with D. melanogaster, yet still showed physiologic antennal responses to CO₂, suggesting that receptor function is not the sole regulator of such behavioral differences.²² Together, these studies suggest that transcriptional changes in the genetic and developmental programs that affect olfactory receptor expression, ORN development and connectivity all can affect olfactory coding, and might drive behavioral differences toward CO₂ between different Drosophila species.

In this study, we examined the behavioral response toward CO₂ for an expanded number of Drosophila species and asked how the molecular and developmental mechanisms underlying CO₂ olfactory receptor neurons have evolved across the genus using transcriptional profiles from the developing peripheral olfactory system of 6 of the species. We also examined the transcription profiles of genes known to be involved in the function and development of the CO₂ olfactory circuit to identify evolutionarily labile components as well as potential mechanisms that may be responsible for behavioral differences. We found that, as anticipated, CO₂ generally evokes repulsive behavior across most of the Drosophila species we examined. However, this behavior has been lost or reduced in a subset of species. Comparisons of transcriptional profiles from the developing and adult antennae for several these species suggest that these behavioral differences may be due to differences in the expression of the CO₂ co-receptor Gr63a. Furthermore, we found variations in the developmental expression of a few transcription factors known to be involved in the development of the CO₂ circuit, namely dac and mip120, which were previously shown to regulate Gr63a expression. In contrast, most of the other known molecular and developmental components of the peripheral CO₂ sensing ORNs show very stable transcriptional levels across all examined Drosophila lineages. These findings suggest that certain components of the CO₂ system may be more developmentally and evolutionarily labile, and may contribute to differences in CO₂-evoked behavioral responses across species.

Results

Behavioral response to carbon dioxide across the Drosophila genus

We investigated variation in carbon dioxide response across the genus. Previous studies have shown reduced repulsion toward CO₂ compared with D. melanogaster for a few Drosophila species, namely D. virilis and D. suzukii,²² but the ancestral state and the extent of behavioral differences across the genus remains unknown. To investigate those questions, we measured the CO₂ response from 15 different species across the genus (Fig. 1A). The species we examined were primarily from the melanogaster (D. melanogaster, D. simulans, D. sechellia, D. erecta, and D. ananassae) and nasuta (D. nasuta, D sulfurigaster, D. kohkoa, D. pallidifrons, and D. niveifrons) subgroups, along with several additional taxa in between (D. virilis, D. lummei, D. mojavensis, D. willistoni and Zaprionus indianus). The approximate phylogenetic relationships of these taxa are shown in Figure 1A.^23-26 The behavioral response of each species toward CO₂ was measured using a T-maze assay, which measures the instantaneous preference or avoidance response of walking flies toward an odor vs. a control. The results of our T-maze assays, represented as a preference index (see Methods) suggest that the majority of the Drosophila species we examined had a robust aversive response toward CO₂, including all members of the melanogaster subgroup and members of the virilis-repleta radiation (Fig. 1B). However, this aversive response is either greatly reduced or not present in many species that we examined, particularly most members of the nasuta subgroup, D. virilis, and D. wilistoni (Fig. 1B). Using ANOVA followed by post-hoc Tukey's HSD, we found a significant effect for species, which we were able to group into 2 categories (avoidance vs. indifference) based on whether their behavioral responses to CO₂ were significantly different (P < 0.05) from D. melanogaster or from D. melanogaster Gr63a mutant flies (Fig. 1B). Comparing these 2 categories to the Drosophila phylogenetic tree suggested that the robust aversive response toward CO₂ was the ancestral state for Drosophilids (fewest required derived characters), and this behavior has been independently lost in a few lineages, such has the ancestors of D. virilis and D. wilistoni, and the ancestors of the nasuta subgroup. Interestingly, if this hypothesis is true, it is possible that a member of the nasuta subgroup, D. kohkoa, independently re-acquired aversion toward CO₂. Overall, our results suggest that the behavioral response to CO₂ among Drosophilids is fairly diverse and evolutionarily labile.

Figure 1.

Behavioral response of different Drosophila species toward carbon dioxide. (A) Phylogenetic tree showing the phylogenetic relationships of all the Drosophila species examined in this study. Light blue colors indicate species that we found to have reduced aversion toward CO₂ relative to the other species. (B) Preference indices for different Drosophila species in response to a T-maze assay where flies are given a choice between regular air and 5% CO₂. The value given for each species is the average value over 3–10 trials of 5 runs (see Methods). Error bars indicate standard error of measure (SEM). Red lines indicate significant differences between each species and wild-type D. melanogaster (top line), and between each species and D. melanogaster Gr63a mutants (bottom line), shown as the results of an ANOVA followed by Tukey's post-hoc test for significance (NS = not significant, *** = P < 0.001).

Differential expression of CO₂ receptor components across the Drosophila genus

Previous studies have shown that physiologic antennal responses to CO₂ are conserved even in Drosophila species that have a reduced avoidance response toward CO₂.²² This suggests that behavioral differences for CO₂ are likely due to differences in the amount or pattern of receptor expression or to changes in neural circuitry, rather than differences in receptor function. To look for differences in CO₂ receptor expression, we used a publicly available RNASeq data set (see Methods) which contains the antennal transcriptome profiles of 6 Drosophila species (D. melanogaster, D. sechellia, D. simulans, D. erecta, D. ananassae, and D. virilis) from 4 different time points during development (3rd instar larva, 8 h after puparium formation (APF), 40 h APF, and adults). We first looked for differences in gene expression in the adult stage for Gr21a and Gr63a, the 2 components of the CO₂ receptor in D. melanogaster. Of the 6 species included in this data set, only D. virilis has reduced aversive response toward CO₂ (Fig. 1B), and this correlates with reduced expression in D. virilis for Gr63a (Fig. 2A), but not Gr21a (Fig. 2B) in the adult antennae, using normalized transcript counts. The reduced expression for Gr63a in D. virilis compared with D. melanogaster was confirmed using quantitative RT-PCR on adult antennae, which showed a greater reduction in expression for Gr63a in D. virilis relative to D. melanogaster compared with Gr21a (Fig. 2C), a result that is comparable to RNASeq data that has been normalized (Fig. 2D).

Figure 2.

Differential expression of CO₂ receptors across Drosophila. (A, B) DESeq2-normalized transcript counts for Gr63a and Gr21a, the 2 CO2 co-receptors, in the adult antennae across 6 Drosophila species as determined by RNASeq analysis. Error bars indicate SEM. Significantly different comparisons, as determined by DESeq2 p-adjusted values using a stringent criteria (P < 1⁻¹⁰), are indicated with an asterisk (NS = not significant). (C) Relative expression of Gr21a and Gr63a in the adult antenna for D. virilis relative to D. melanogaster as measured with real-time quantitative RT-PCR. Values shown indicate the mean over 3 runs normalized against Act5C (set to +18), and error bars indicate SEM. (D) To provide a direct RNASeq comparison to (C), this plot shows the relative expression of Gr21a and Gr63a in the adult antenna for D. virilis relative to D. melanogaster as measured with RNASeq, log₂ normalized against Act5C (set to +18). Error bars indicate SEM. (E) Preference indices for D. melanogaster wild-type and Gr63a heterozygous and homozygous mutants in response to a T-maze assay where flies are given a choice between regular air and 5% CO₂. The value given for each species is the average value over 6–10 trials of 5 runs (see Methods). Error bars indicate SEM. Significant differences shown are the results of an ANOVA followed by Tukey's HSD (NS = not significant, *** = P < 0.001).

Given the reduced Gr63a expression in D. virilis, which has reduced repulsion toward CO₂, we also tested the behavioral response of heterozygotic Gr63a mutant D. melanogaster which have only one functional copy of the Gr63a gene (and thus presumably reduced expression) toward CO₂ (Fig. 2E). However, the behavioral response of the Gr63a heterozygotes toward CO₂ was not significantly different compared with wild-type D. melanogaster, suggesting that a reduction in Gr63a transcription alone is not sufficient to cause behavioral differences, and that additional processes such as developmental or higher-order circuit changes must also be involved.

Conservation of genes regulating the development of CO₂ receptor expression across Drosophila

The availability of the RNASeq data set with the transcriptome profiles for adult and developing antennae across multiple Drosophila species also allowed us to investigate the broader question on how conserved the molecular and development programs of the CO₂ olfactory system are across species. To answer this question, we compiled RNASeq data for all of the genes known to be involved in the development of CO₂ ORNs, including factors that regulate the development of the ab1 sensillum and factors that regulate CO₂ receptor expression.^{6,12,17,21,27,28} We calculated the absolute log fold change (ALFC) in gene expression for each gene across each pairwise interspecies comparison. The ALFC value has been used in previous studies as a simple proxy for the variability of a gene across evolutionary time scales,²⁹ with higher ALFC values indicating that a gene has greater transcriptional variability, and lower values indicating that gene expression is conserved. We first plotted our results as a boxplot of median and quartile ALFC values for each gene (Fig. 3A), which showed that the genes regulating the development of CO₂ ORNs can be roughly divided into 2 categories, with some genes having similar transcription levels across species and others having more variable levels of transcription. However, it is important to note that the genes with higher variability are not unusually variable; the variability they present is consistent with the average across the transcriptome (Fig. 3A). Instead, the difference between the 2 groups seems to be due to the group with conserved gene expression being remarkably well conserved — that is, far below the transcriptome average. Interestingly, most of the genes belonging to the MMB/dREAM complex have highly conserved expression, while the members of mir-279 pathway and the transcription factors that specify ab1 identity have greater transcriptional variability across species. In addition, we also represented our results as a heatmap showing each individual species comparison (pairwise ALFC), along with a clustering analysis of the ALFC values for each gene using hierarchical clustering (Fig. 3B). The clustering analysis also supported the distribution of genes regulating the development of CO₂ ORNs into 2 groups with good bootstrap support, with one group in the upper part of the figure having high transcriptional variability across species, and the other group in the lower part of the figure being more transcriptionally conserved. Again, most of the genes belonging to the MMB/dREAM complex appear to have highly conserved levels of gene expression. In contrast, the members of the mir-279 pathway and the transcription factors that specify ab1 identity, along with Gr21a and Gr63a, have relatively higher transcriptional variability across species, comparable to the variability of the housekeeping genes Act5C and Gapdh2 (in other words, having average variability). Furthermore, the heatmap indicated that the transcriptional variability in the developmental genes was not restricted to any one species, nor did it seem to be primarily related to phylogenetic distance between species as large differences were observed even among closely related species. Overall, our results suggest that some components of the regulatory mechanisms underlying CO₂ receptor expression have very similar transcription levels across species while other components have transcription that is more evolutionarily labile.

Figure 3.

Transcriptional variability of developmental regulators of CO₂ receptors across 6 Drosophila species. (A) Boxplot and (B) heatmap plotting the absolute log fold change (ALFC) in the expression of known developmental regulators of CO₂ receptors across all pairwise interspecies comparisons and across 4 developmental time points (3rd instar larvae, 8 h APF, 40 h APF, adult). Color coding of genes is as follows: cyan – MMB/dREAM complex; magenta – miR-279 pathway; yellow – ORN/sensilla identity. The boxplot shows combined values for each transcription factor across development (n = 60), with the mean ALFC value for the antennal transcriptome shown as a red line for comparison. ALFC values for Orco, Gr63a and Gr21a during the adult stage are also included (n = 15). Statistical comparisons show results for ANOVA + Tukey's HSD (. = P < 0.1, * = P < 0.05, ** = P < 0.01, *** = P < 0.001, NS = not significant). In the heatmap, ALFC values shown are the mean ALFC value across all 4 developmental time points for each specific pairwise comparison, except for Orco, Gr63a and Gr21a which show values only from adults. Numbers on nodes indicate bootstrap support for the clustering analysis. (C) Plots of interspecies pairwise log fold change for Gr63a against interspecies pairwise log fold change for genes known to be involved in Gr63a expression, from RNASeq data (n = 15). Solid lines indicate linear regression. See also Figure S1. (D) DESeq2 and log-normalized RNASeq transcript counts for mip120 and dac across 6 species (see Methods) and 4 developmental time points (3L – 3rd instar larvae, p8 – 8 h APF, p40 – 40 h APF, AA – adult antennae). Error bars omitted for visual clarity.

Next, we hypothesized that the changes in Gr63a expression may result from changes in the developmental programs underlying receptor expression and/or sensilla identity. To investigate this, we used the same RNASeq data set to compare changes in adult Gr63a expression across 6 Drosophila species with changes in the transcription of genes known to be involved in regulating Gr63a expression or ab1 sensilla identity during development (such as pre-patterning transcription factors and the MMB/dREAM complex) to identify developmental genes that may potentially underlie behavioral differences toward CO₂. Regression analysis identified mip120 and dac as the 2 developmental genes with changes across the 6 species that most strongly correlate with changes in Gr63a expression (Fig. 3C, see also Fig. S1 for a combined figure with other developmental genes). Mip120 is part of the MMB/dREAM complex, an epigenetic regulator known to repress Gr63a expression,^21,30 while dac is a pre-patterning transcription factor known to be involved in the specification of large basiconic sensilla.²⁸ RNASeq data shows that mip120 has increased transcription in D. virilis compared other species, while dac has increased transcription in species with higher Gr63a expression such as D. ananassae and D. sechellia (Fig. 3D). Our data are consistent with prior work showing that mip120 and Gr63a expression are inversely correlated.²¹ Overall, our results suggest that the indifference of D. virilis toward CO₂ compared with members of the melanogaster subgroup may be due to a reduction in Gr63a expression, which in turn may be due to transcriptional changes in certain developmental regulatory genes.

Changes in miR-279 expression in D. virilis maxillary palps

Previously, miR-279 and its transcriptional regulator prospero were shown to repress the generation of ectopic CO₂ ORNs in D. melanogaster maxillary palps.^6,27 We next asked if there were differences in the miR-279 regulatory pathway that may be involved in behavioral differences between species. The RNASeq data set is from antennal samples and therefore does not include microRNA expression, which means we cannot directly determine differences in antennal miR-279 expression in other Drosophila species. Instead, we decided to compare the maxillary palp expression of the D. melanogaster (a species that is repelled by CO₂) miR-279 promoter to the D. virilis (a species that has reduced aversive response to CO₂) miR-279 promoter by generating transgenic reporters of miR-279 expression in D. melanogaster. In a previous study, we showed that a 3 kb miR-279 genomic construct that contains only 2 kb of upstream sequences and 1 kb downstream of the transcriptional start site was 100% efficient at rescuing the miR-279 mutant phenotype in the olfactory system.⁶ This 2 kb upstream sequence is also capable of expressing miR-279 in the maxillary palps.⁶ Additionally, a bioinformatics interrogation of miR-279 promoter in different Drosophilids found that Prospero binding sites upstream of mir-279 critical for its expression in melanogaster maxillary palp precursors,²⁷ are modified in D. virilis (Fig. 4A). These data suggest that the regulatory elements that are required for the developmental and tissue-specific expression of miR-279 are harbored within this 2 kb sequence, including the Prospero binding sites. From the sequence analysis, we also observed that the same Prospero binding sequence is changed from “nnnTAAGACAnnn” in D. melanogaster to “nnnTGCGCGAnnn” in D. virilis. We generated GAL4 reporter transgenes driven by the 3 kb D. virilis miR-279 promoter, and compared UAS-CD8GFP expression in the developing maxillary palps at 50 h after puparium formation to D. melanogaster (Fig. 4B). Interestingly, we found that D. melanogaster flies containing D. virilis mir-279-gal4 UAS-cd8GFP had less GFP expression in the developing maxillary palps compared with D. melanogaster flies containing the D. melanogaster equivalent, suggesting that mir-279 expression in D. virilis may be reduced (Fig. 4B-C). However, a comparison of Gr21a and Gr63a expression in the maxillary palps of D. melanogaster and D. virilis using quantitative RT-PCR (Fig. 4D), along with fluorescent RNA in situ hybridization experiments for Gr21a in D. virilis (data not shown), showed no indication of any ectopic expression of the CO₂ receptor in the D. virilis maxillary palps. As such, we believe that it is unlikely that the mir-279 regulatory pathway ultimately contributes to any reduction in the aversive behavioral response toward CO₂, at least for D. virilis.

Figure 4.

Conservation of the mir-279 regulatory system in D. virilis. (A) Comparisons of a prospero binding site in the promoter region of miR-279 across multiple Drosophila species. (B) Confocal images of maxillary palps from transgenic D. melanogaster flies with the D. virilis mir279-Gal4 UAS-cd8GFP promoter-fusion construct at 50 h APF during development (left), and equivalent D. melanogaster flies with a D. melanogaster mir279-Gal4 UAS-cd8GFP promoter-fusion construct (right). (C) Quantification of (B) as the difference in the florescent intensity of GFP expression by miR-279 promoters from D. mel. and D. vir. Asterisks indicate significant difference as determined by an unpaired t-test (n = 6, P < 0.001). (D) Relative expression of Gr21a, Gr63a, Or42a (positive control for maxillary palps) and Or47b (positive control for antennae) in maxillary palps (MP) and antennae (Ant.) for D. virilis and D. melanogaster as measured with real-time quantitative RT-PCR. Values shown indicate the mean over 3 runs normalized against Act5C (set to +18), and error bars indicate SEM.

Conservation of CO₂-sensing antennal structures and the target glomerulus in the antennal lobe

Finally, we asked if there were any differences in the CO₂-sensing antennal structures or the CO₂ circuitry in other Drosophila species and how conserved these are across the genus. To investigate this, we used different imaging methods to determine the existence and spatial location of Gr21a-expressing ORNs, large basiconic sensilla, and the V-glomerulus in a subset of the species that we examined for CO₂ behavioral response. First, using fluorescent whole-mount RNA in situ hybridization, we found Gr21a-expressing ORNs in all the species we examined (D. melanogaster, D. sechellia, D. simulans, D. virilis), which was unsurprising given our previous findings but also revealed that the spatial location of the CO₂ receptor in the proximal-medial region of the antennae is conserved (Fig. 5A). Quantification of this data indicated that the number of Gr21a-expressing cells in the antennae is conserved across the 4 species as well (Fig. 5B). Similarly, scanning electron micrographs of the antennae from D. melanogaster, D. sechellia, D. erecta, and D. virilis indicated that the presence and spatial distribution of large basiconic sensilla in the proximal-medial region of the antennae is also conserved across multiple Drosophila species (Fig. 5C). In addition, antibody stainings of the antennal lobe for the nc82 protein, a neuronal marker, in D. melanogaster, D. sechellia, D. willistoni, and D. virilis indicated that the V-glomerulus is present and distinct even in species with reduced aversion toward CO₂ (Fig. 5D). Interestingly, 3D reconstruction of the V-glomerulus for D. melanogaster and D. virilis indicated that the D. virilis V-glomerulus has a smaller volume relative to the rest of the antennal lobe compared with D. melanogaster (Fig. 5E). Given that the number of Gr21a ORNs appear to be comparable in both species (Fig. 5C), it is possible that the smaller volume reflects a reduced number of interneurons or projection neurons (PNs) connecting to the D. virilis V-glomerulus compared with D. melanogaster. Overall, our results suggest that the presence and spatial location of CO₂-sensing antennal structures and circuits are mostly conserved across the Drosophila genus, but higher order neuronal connections may potentially be different in species with reduced repulsion toward CO₂.

Figure 5.

Conservation of CO₂-sensing antennal structures and the V-glomerulus. (A) RNA fluorescent in situ hybridization (FISH) of the adult antennae from multiple Drosophila species with a Gr21a antisense probe. (B) Quantification of RNA FISH results for Gr21a expression (n = 6–12). Error bars indicate SEM. Significance shown as determined by ANOVA + Tukey's HSD (NS = not significant). (C) Scanning electron micrographs of the adult antennae from multiple Drosophila species. Blue dots indicate the location of basiconic sensilla, orange dots indicate the location of trichoid sensilla. (D) Images of the antennal lobe from multiple Drosophila species, stained with an anti-nc82 antibody, a neuronal marker. The location of the V-glomerulus is highlighted in white. (E) Comparison of the volume of the V-glomerulus in D. melanogaster and D. virilis. To normalize for body size, values are shown as the ratio between the volume of the V-glomerulus and the volume of the entire antennal lobe. Significance indicated as determined by an unpaired t-test (n = 10, P < 0.001).

Discussion

Here we investigated the relationship between changes in carbon dioxide response, which is critical for regulating many behaviors, and changes in the abundance and expression patterns of the genes giving rise the to CO₂ sensory system. We examined the behavioral response of various Drosophila species toward CO₂, and, for a subset of those species, we examined the transcription profiles of developmental regulators of CO₂-sensing ORNs, as well as the conservation of CO₂ sensory structures. We found that some Drosophila lineages strongly avoided CO₂ while others were indifferent, and our examination of transcription profiles suggests that these differences are accompanied by molecular and transcriptional changes in the expression of the Gr63a co-receptor and some of its developmental and transcriptional regulators.

Our results parallel a previous study examining the CO₂ response in Drosophilids²² that also found a reduced aversive response for D. virilis. Further, we showed that reduced aversion toward CO₂ is common among Drosophila species, and that the behavioral response toward CO₂ independently diverged in a few different Drosophila lineages such as the virilis-repleta radiation and the nasuta subgroup. The reason(s) for this behavioral divergence remains unclear, but likely relate to differences in each species ecological niche, as previously suggested.²² For example, in species that are repulsed by CO₂, CO₂ may repel certain species from ripening fruit,²² or function as an alarm pheromone⁸; in contrast, species that are indifferent toward CO₂ may not need those cues or use different odorants for those functions. Hypothetically, the loss of aversion toward CO₂ may allow such species to exploit new food sources that emit CO₂. For example, D. virilis is known to breed in slime fluxes on shedding trees³¹ which is a known bacterial infection. With such ecological pressures, CO₂ produced by the fermentation of bacteria might have transformed CO₂ sensing circuit function to have a positive attractive valence, which could lead to the weakening of the repulsive behaviors driven by the CO₂ sensing ORNs over time. Admittedly, our study tested behavioral response only at a specific concentration (5%); thus, it is very possible that the response of the species we examined may differ in non-laboratory conditions with different CO₂ concentrations. Previous studies have also found differences in the behavioral response toward CO₂ for D. melanogaster when the fly is in tethered flight as opposed to walking³²; our study only examined the latter. Additionally, our study only examines the olfactory system, and does not examine gustatory responses toward dissolved CO₂ (in D. melanogaster, this attractive response is mediated by CO₂ sensing neurons in the labial palps³³), which also may or may not be conserved between species. Nonetheless, we believe that our results are sufficient to suggest potentially interesting behavioral differences for an ecologically important odorant that warrants further investigation.

Our comparison of the gene expression profiles for CO₂-related genes across several Drosophila species found reduced Gr63a expression in the CO₂-indifferent D. virilis along with correlated changes in some Gr63a regulators during development, specifically mip120, a component of the MMB/dREAM complex,²¹ and dac, an important developmental transcription factor.²⁸ Whether or not these differences in gene expression are the actual causative mechanism behind changes in behavior will require additional study, but our findings identify candidate genes for future experiments. In addition, as our comparisons of gene expression profiles included only 6 Drosophila species, with only one of those having reduced aversion toward CO₂, perhaps a more thorough analysis of the entire Drosophila genus would give us a more complete picture to identify the influence of developmental regulators on Gr63a expression and the resulting behavioral outcomes.

Similarly, our finding that certain components of the regulatory mechanisms underlying CO₂ receptor expression, specifically the members of the MMB/dREAM complex, are more transcriptionally conserved than others may be useful as a starting point for future experiments intending to manipulate the CO₂ behavioral response in insects. Among other reasons, it is possible that these components are parts of systems that more sensitive to perturbation, have more important functions, or have greater pleiotropy.³⁴ This explanation is consistent with our finding that the most conserved components are members of the MMB/dREAM complex that are key epigenetic modifiers with critical roles during development.²¹

Besides differences in gene expression, another factor that could contribute to behavioral differences between species is changes in CO₂-sensing neural circuits. Previous studies have suggested that CO₂ olfactory transduction to the antennal lobe is conserved across Drosophilids,²² which is consistent with our findings regarding the general conservation of CO₂ sensing morphological structures. However, it is quite possible that that there are differences in neuronal connections in higher order circuits, such as projection neurons and local interneurons in the antennal lobe³⁵ that lead to a reduced CO₂ response in some species, as suggested by our limited comparisons of the V-glomerulus. There may also be higher order circuits in the mushroom body and lateral protocerebrum that could determine species-specific responses to CO₂. Further studies will be required to determine if these differences are present and play meaningful roles in the divergent behavioral responses of Drosophilids toward CO₂.

Materials and methods

Fly strains, rearing, and collections

The complete set of species genotypes used in our study is listed in Table S1. For our analyses of gene expression, the 6 Drosophila species genotypes examined were D. ananassae (14024–0371.00, Drosophila Species Stock Center (DSSC), University of California, San Diego), D. erecta (14021–0224.00, DSSC), D. sechellia (14021–0248.01, DSSC), D. simulans (14021–0251.165, DSSC), D. virilis (15010–1051.00, DSSC) and the w118 strain of D. melanogaster. All flies used (for both behavioral experiments and RNASeq) were reared on cornmeal medium at room temperature.

T-maze CO₂ behavioral Assay

All experimental runs were conducted in the dark with the use of far red light. A vial of 30–40 flies (∼7–10 d old) containing both males and females were loaded into transfer tubes and were then loaded into the T-maze elevator. Flies were left in the resting phase position for about 1 min. After the resting phase, 5% CO₂ gas (which has previously been shown to elicit robust repulsive behavior^5,6) was added to the CO₂ arm of the T-maze, and the flies were immediately moved to the choice point and allowed to decide between the CO₂ and air arms for 30 s. After 30 s, the flies in each arm were counted. Each vial of 30 flies counted as 1 trial, and each trial consisted of 5 runs through the T-maze protocol, with the final counts for each trial reported as the sum total of fly choices for each arm over the 5 runs. These counts were then used to determine the preference index value for each trial, calculated as the difference in the number of flies in the CO₂ arm vs. the air arm divided by the sum of flies in both arm.

RNASeq data analysis

Our RNASeq analysis used a publicly available data set of antennal transcriptomes from 6 Drosophila species. This data set includes the antennal transcriptomes from 4 developmental stages: 3^rd instar larvae, 8 h after puparium formation (APF), 40 h APF, and adults. The gender ratio for all adult samples was 1:1. The raw sequence data and metadata are available for download from the NCBI Gene Expression Omnibus (GEO) database using the accession numbers GSE85239 and GSE75986, and the normalized count tables are also available in the GSE85239 series. Differential expression of genes between species was analyzed using the DESeq2 suite in R. In our analyses, the conservation of a gene was determined by calculating the mean absolute log fold change for all pairwise species comparisons. Absolute log fold change (ALFC) values were calculated by comparing DESeq-normalized transcript counts for each species to every other species in a pairwise fashion, and taking the absolute of the log fold change in transcript counts to provide a non-directional estimate of the difference in gene expression between each pairwise species comparison for every gene.

Heatmaps and cluster analysis

Heatmaps were created in R using the heatmap.2 function from the gplot package (v3.0.1). Cluster analysis was done with the default clustering options from the heatmap.2 function that uses hierarchical clustering with the distance and cluster methods set as “euclidean” and “complete,” respectively. Bootstrap analysis of clusters was done using the pvclust package in R (v2.0–0) for 10,000 iterations.

Genetics

To compare the function of the D. melanogaster and D. virilis mir-279 promoters, we created transgenic D. melanogaster flies containing promoter-Gal4 fusion constructs for the mir-279 promoter region from D. virilis. To create the D. virilis mir-279 promoter-Gal4 construct, we cloned a 3 kb region upstream of the orthologous D. virilis mir-279 gene from genomic D. virilis DNA into a pCasperAUG-GAL4-X vector^11,36 using the custom primers 5′-CGCCACATTTCTACTCAGTTTC and 5′-AGTACGCATATTCGATCCACTC. From there, injection of the construct into D. melanogaster w¹¹¹⁸ eggs and screening for transgenic flies followed standard protocol.³⁷ Transgenic flies containing mir-279 gal4 constructs were crossed to flies containing UAS-cd8GFP¹² before imaging.

Real-time RT-PCR

Antennae or antennal discs from approximately 100 flies or 50 larvae, respectively, were dissected and analyzed for each species. RNA was extracted with an RNeasy kit (Qiagen), treated with on-column DNase digestion (Qiagen), and then reverse transcribed into cDNA using the SuperScript First-Strand Synthesis System for RT-PCR (Invitrogen). qPCR was performed with the FastStart Universal SYBR Green Master Mix (Roche) using standard protocol. Expression for each gene was analyzed in triplicate. RNA concentration was standardized to 15 ng/µl before reverse transcription, and cDNA was diluted 1:32 before use. Ct values were normalized to each species' Actin 5C (Act5c) expression. Primers used in the reactions are listed in Table S2.

Immunohistochemistry

Samples were fixed with 4% paraformaldehyde, washed with phosphate buffer with 0.2% Triton X-100, and stained as described previously.¹² Primary and secondary antibodies were used in the following dilutions: rabbit anti-GFP 1:1000 (Invitrogen), mouse anti-nc82 1:20 (Developmental Studies Hybridoma Bank), Alexa 488 goat anti-rabbit 1:1000, goat anti-mouse-Cy3 1:100. Confocal images were taken by an Olympus Fluoview FV1000. 3D reconstruction and calculations of glomerular or antennal lobe volumes were done using Imaris software (v8.2).

Flourescent whole-mount RNA in situ hybridization

Digoxigenin RNA probes were made using a Roche DIG RNA labeling kit (Indianapolis, IN, USA). Primers used can be found in Table S3. Drosophila heads were dissected into cold fixative (4% paraformaldehyde, 0.05% Tween 20 in 1X PBS) and fixed for one hour. Heads were then washed 3 × 10 min in PBST (1X PBS, 0.1% Tween 20). The third antennal segment was dissected and fixed for an additional 30 min. Tissue samples were washed 5 × 5 min in PTX (1X PBS, 1% Triton X) and incubated in hybridization (Hyb) buffer (50% formamide, 5X SSC, 0.05 mg ml−1 heparin, 0.1% Tween 20) for 2 h at 55°C, before hybridization overnight at 55°C with DIG-labeled RNA probe. Tissue was then washed 5 × 20 min in Hyb buffer at 55°C, with the last wash proceeding overnight. Samples were then washed for 20 min in Hyb buffer at 55°C, followed by 5 × 5 min washes in PBST at room temperature before incubation for 3 h in 1: 500 anti-DIG-AP in PBST and 1X BSA. Next, samples were washed 5 × 5 min in PBST and incubated in FastRed solution (Roche, Indianapolis, IN, USA) for 30 min. Samples were washed for a final 5 × 5 min in PBST and stored overnight in mounting solution before imaging.

Scanning electron microscopy

Fly heads were fixed (4% paraformaldehyde, 0.05% Tween 20 in 1X PBS) immediately after dissection for 20 min, then rinsed in PBST (1X PBS, 0.1% Tween 20) twice for 15 min each. Samples were then dehydrated using an ethanol series at 30%, 50%, 70%, 90 and 100% ethanol, twice for 10 min at each dilution. Samples were allowed to dry for 1 h, then sputter-coated with Au using a Denton Vacuum Desk IV sputter unit. SEM images were taken with a FEI XL30 SEM-FEG (Duke Shared Materials Instrumentation Facility) immediately after.

Abbreviations

APF

After puparium formation

ALFC

Absolute log fold change

CO₂

Carbon dioxide

Gr

Gustatory receptor

ORN

Olfactory receptor neuron

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Funding details

This study was supported by grants from the Whitehall Foundation and Duke University, as well as from the NSF to PCV and CDJ (DEB-1457690).

Author contributions

Conceptualization, JWP PCV; Methodology, JWP CDJ PCV; Investigation, JWP JM HY CL PR SO AM SC; Writing – Original Draft, JWP PCV; Writing – Review & Editing, JWP PCV; Funding Acquisition, CDJ PCV; Supervision, CDJ PCV.