Evolution of temperature preference in flies of the genus Drosophila

Abstract

The preference for a particular thermal range is a key determinant of the distribution of animal species. However, we know little on how temperature preference behaviour evolves during the colonization of new environments. Here we show that at least two distinct neurobiological mechanisms drive the evolution of temperature preference in flies of the genus Drosophila. Fly species from mild climates (D. melanogaster and D. persimilis) avoid both innocuous and noxious heat, and we show that the thermal activation threshold of the molecular heat receptor Gr28b.d precisely matches species-specific thresholds of behavioural heat avoidance. We find that desert-dwelling D. mojavensis are instead actively attracted to innocuous heat. Notably, heat attraction is also mediated by Gr28b.d (and by the antennal neurons that express it) and matches its threshold of heat activation. Rather, the switch in valence from heat aversion to attraction correlates with specific changes in thermosensory input to the lateral horn, the main target of central thermosensory pathways and a region of the fly brain implicated in the processing of innate valence^1-5. Together, our results demonstrate that, in Drosophila, the adaptation to different thermal niches involves changes in thermal preference behaviour, and that this can be accomplished using distinct neurobiological solutions, ranging from shifts in the activation threshold of peripheral thermosensory receptor proteins to a substantial change in the way temperature valence is processed in the brain.

Extended Data Figure 1:. Alignment of Gr28b.d protein sequences across Drosophila species.

(a) Multiple alignment of Gr28b.d protein sequences from 25 Drosophila species (names abbreviated to first three letters, e.g. D. mel = D. melanogaster), (b) and of the 4 species D. melanogaster, D. persimilis, D. mojavensis, and D. mettleri. (c) Location of amino acid differences between the indicated species (red=different). (histogram: bars = mean pairwise amino acid sequence identity across species, green = 100%, yellow = <100% and ≥30%, orange = <30%; highlighted residues below denote divergence from the consensus sequence and are colored by amino acid identity using the default Geneious color scheme).

Extended Data Figure 2:. The D. mojavensis Gr28b.d null mutant displays normal avoidance of noxious heat.

For each panel, the data plotted on the left is the preference index for wild-type D. mojavensis (WT), and the data plotted on the right is the preference index of D. mojavensis Gr28b.d^Δ1 mutants. In all cases, flies were given a choice between 25°C and a TT of 40° or 45°C, respectively. The transparent box on the left represents wild type D. mojavensis data re-drawn from Figure 1 to allow for direct comparison. Ø indicates the wild type and mutant preference distributions are not statistically different (two-tail t-test, p<0.05; p=[0.678,0.720] for WT vs. mutant for 40°C and 45°C, respectively; red lines=mean, grey boxes=one standard deviation, empty boxes=95% confidence interval for the mean, empty circles=datapoints; each datapoint is one group of 15 flies; N=11 and 7 datapoints /condition for 40°C and 45°C, respectively).

Extended Data Figure 3:. Labelling, reconstructing, and analyzing thermosensory projection neurons in Drosophila brains.

(a) TPN labeling strategy using electroporation of dextran-conjugated dyes, followed by imaging in Front view (on anteroposterior axis) and Top view (on dorsoventral axis) of the brain. (b) Following the acquisition of 2-photon z stacks, 2D outlines of brain regions are segmented to produce a 3D model of the CA and LH which is used for image registration (to align images to a representative brain for each species). Neurons are traced in the registered volumes using neuTube and visualized alongside the 3D CA and LH volumes. (c) Maximum projections and corresponding 3D neuronal reconstructions of five labelled TPN-V in D. melanogaster and D. mojavensis brains. (d) Neuron reconstructions are exported in SWC format, and intrinsic neuronal morphology is quantified with the CAJAL package. UMAP embedding of pairwise distances from CAJAL analysis alone shows neuronal morphologies sort out by species in the Lateral Horn (LH) but not as much for the Calyx (CA). (e) Center of Mass (COM) of each projection pattern is measured in FIJI/ImageJ using a z-projection for each z-stack cropped around either the CA or LH (see methods for details); XY plots represent average ± STD of 5 COM measurements for each species, the COM of the D. melanogaster EM reconstruction is shown for comparison. Next, a matrix of pairwise taxicab distances is computed between the COMs. (f) Pairwise distance matrices from CAJAL and COM analysis are normalized to their max values, summed, and used as input for hierarchical clustering analysis and UMAP embedding. Abbreviations: Thermosensory Receptor Neurons, TRNs; Thermosensory Projection Neurons, TPNs; Posterior Antennal Lobe, PAL; LH, Lateral Horn; CA, Calyx. Scale bars in a = 10 μm and in c = 25 μm.

Extended Data Figure 4:. Frequency of Thermosensory Projection Neuron cell types labelled in dye electroporation experiments across Drosophila species.

(a-d) Counts (tables) and frequencies (visualized as radial bar graphs) of Thermosensory Projection Neuron (TPN) cell types labeled in dye-filling experiments, arranged by species. Included are all experiments that successfully labelled a small number of TPNs (up to 3–4) allowing unambiguous identification of each cell type. We excluded experiments in which only local interneurons of the PAL were labeled, where too many TPNs were labeled, or where labeling of olfactory projections prevented us from classifying TPNs. (Note that individual percents for cell types add up to more than 100% due to double- and triple- labeling of TPNs in the same brains).

Extended Data Figure 5:. Thermosensory Lateral Horn Output Neuron II (TLHON-II) is the primary output of TPN-V and is critical for heat avoidance behavior in D. melanogaster.

(a) Connectomic diagram illustrating the pathways that relay temperature information from the hot thermosensory receptor neurons (TRNs) to the Thermosensory Lateral Horn Output Neuron II (TLHON-II). TPN-V is the major output of the aristal hot TRNs (comprising 52.67% of hot TRN output synapses to TPNs) and the major direct path of connectivity from hot TRNs to Thermosensory Lateral Horn Output Neuron II (TLHON-II), which is in turn the major target of TPN-V (black arrows represent alternative direct pathways of connectivity between hot TRNs and TLHON-II, Ns=number of synapses). (b) D. melanogaster EM reconstruction of TPN-V (red) and TLHON-II (light blue). (c) TPN-V::TLHON-I synapses (yellow dots) and TPN-V::TLHON-II synapses within the LH (blue dots) tile the TPN-V axonal arbor within the LH (synapse location was reconstructed from EM data, see methods for details). (d) A split-Gal4 driver exclusively expressed in TLHON-II (maximum projections of 2-photon z-stacks of the brain and VNC of TLHON-II-Gal4>GFP animal). Scale bars = 25 μm. (e) Genetic silencing using TLHON-II:Gal4>Kir2.1 impacts heat avoidance (*p < 0.05, 2-way ANOVA, p = [0.748, 1.95e-3, 4.15e-5, 4.91e-7] for test temperatures 25, 30, 35, and 40, respectively) (f) Optogenetic activation via TLHON-II:Gal4>CsChrimson is sufficient to induce avoidance of red light (*p < 0.05 indicates significant difference from zero in two-way one sample t-test, p = [0.440, 0.983, 2.54e-5] for Gal4/+, UAS/+, and Gal4>UAS, respectively). (in e-f red lines=mean, grey boxes=one standard deviation, empty boxes=95% confidence interval for the mean, circles=groups of 15 and 25 flies each for e and f, respectively; N = 8 and 9 groups/condition for e and f, respectively). Abbreviations: Lateral Horn, LH; Calyx, CA; Posterior Antennal Lobe, PAL; Mushroom Body, MB; Superior Medial Protocerebrum, SMP.

Extended Data Figure 6:. Labelling olfactory projections in the D. mojavensis brain reveals largely conserved Lateral Horn organization.

(a) Schematic of the Drosophila olfactory system. Olfactory Receptor Neurons (ORNs) send axons to the Antennal Lobe (AL) where they connect with specific second-order Olfactory Projection Neurons (OPNs) that relay information to higher brain centers such as the CA and LH. The ORN/OPN circuit of glomerulus DM1 is schematized next to an EM reconstruction of the D. melanogaster DM1 OPN. (b) Loading of dextran-conjugated dyes on the antennal nerve allows visualization of stochastic subsets of ORN axons that can be recognized due to the stereotype position of glomeruli (a D. mojavensis DM1 is shown) and used as a target for electroporation. (c) Innervation of the CA and LH by DM1-OPN is identical in D. mojavensis (in neuronal reconstructions and 2-photon z-stacks) as compared to D. melanogaster (EM reconstructions). Stacks from three individual D. mojavensis flies are shown to highlight stereotypy of innervation (right). (d) To label both olfactory and thermosensory projections in the same brain, a red dye was first used to load the aristal nerve while a green dye was later used to load the antennal nerve (see methods for details). This allows independent visualization of PAL and AL glomeruli as targets for 2-color dye-labelling experiments. (e) 2-photon imaging demonstrates that DM1-PN innervation (white, labelled with AlexaFluor^™ 594 dye) of the LH is essentially identical in D. mojavensis as compared to D. melanogaster (green, EM). In contrast, D. mojavensis TPN-V innervates only the posterior LH (red, labelled with AlexaFluor^™ 488 dye), while the D. melanogaster TPN-V counterpart innervates more broadly (red, EM). Abbreviations: Lateral Horn, LH; Calyx, CA; Antennal Lobe, AL; Posterior Antennal Lobe, PAL; Olfactory Receptor Neurons, ORNs; Olfactory Projection Neurons, OPNs. Scale bars = 25 μm.

Extended Data Figure 7:. Temperature preference behavior and TPN-V innervation are species-specific and consistent across populations.

(a-c) Preference indexes measured from additional strains of D. persimilis, D. melanogaster, and D. mojavensis, demonstrate consistent behavior within populations of the same species (compare to data shown in Fig. 1b–d, based on different strains -see methods). (a) A D. persimilis strain from Santa Rosa, California (strain 24) (N = 8,8,8,7,7,7 groups of 15 flies for test temperatures 15, 20, 25, 30, 35, and 40 °C, respectively). (b) A wild-derived D. melanogasterstrain from Egypt (EG16; N = 13,11,12,12 groups of 15 flies for test temperatures 15, 20, 25, 30, and 35 °C, respectively). (c) A D. mojavensis wrigleyi strain from Catalina Island (D. m. wrigleyi is a subspecies of D. mojavensis; N = 20,19,18,20 groups of 15 flies for 25, 30, 35, and 40 °C, respectively) (d-f) 2-photon z-stacks of dye-labelled TPN-V in each of the three strains shows consistent anatomy within species (compare to data shown in main Figs. 4 and 5, based on different strains of the same species -see methods). (g) A distance matrix of reconstructions (see methods) shows additional strains cluster by species (highlighted branches/datapoints represent additional strains generated from 2P images in d-f, whereas greyed out ones are re-plotted from Fig. 5). Abbreviations: Lateral Horn, LH; Calyx, CA. Scale bars = 25 μm. Asterisks: lateral Accessory Calyx (lACA). Arrowheads = ventral anterior LH, the dotted line is for reference. (In a-c: red lines=mean, filled boxes=one standard deviation, empty boxes=95% CI of the mean, empty circles=groups of flies, grey shading = approx. favorite thermal range, and Ø denotes preference indexes not significantly different from zero in two-tail one sample t-test; p < 0.05, p = [6.30e-3, 3.33e-4, 0.747, 2.47e-5, 7.67e-6, 1.55e-7] for a, p = [2.22e-9, 9.38e-6, 0.488, 1.34e-5, 1.76e-10] for b, and p = [0.888, 3.66e-8, 1.58e-7, 1.13e-5] for c).

Figure 1.. Drosophila species display different thermal preference related to conditions in native habitats.

(a) 2-choice thermal preference assay: groups of 15 flies are given a choice between 25°C and a variable test temperature in alternating spatial configurations, a preference index is calculated based on the time spent at each temperature. (b) Drosophila persimilis is restricted to the northern pacific coastal range of North America, where it is found in cool forest habitats. In 2-choice preference assays, it prefers ~15–20°C. (c) Drosophila melanogaster is a cosmopolitan human commensal mostly found in temperate regions and prefers ~25°C over lower or higher temperatures. (d) Drosophila mojavensis is endemic to the deserts of Southwestern United States and Mexico and prefers ~30–35°C. (in b-d red lines=mean, grey boxes=one standard deviation, empty boxes=95% confidence interval for the mean, empty circles=groups of 15 flies each; N=8,6 and 11 groups/condition, respectively; grey shading = approx. favorite thermal range). (e) Phylogenetic tree representing species relationships (see methods for details).

Figure 2.. The ion channel Gr28b.d displays species-specific thresholds of heat activation.

(a) Gr28b.d protein alignment from 25 Drosophila species (histogram= mean pairwise identity across species, green = 100%, red = <30%; orange blocks below= variable residues). (b) AlphaFold3 model of a Gr28b.d tetramer. (c) Color-coded conservation based on the alignment in a and (d) between the indicated species (red=different; see also Extended Data Figure 1). (e) In vitro recording schematic. (f) Representative current-clamp recordings. Unlike controls cells, cells expressing D. melanogaster Gr28b.d (DmelGr28b.d) produce an inward current in response to heating. (g) Representative current–voltage plots from control cells (grey trace) and DmelGr28b.d-expressing cells (black trace) at ~40°C (h,i) DmelGr28b.d threshold for heat activation. (h) A representative Arrhenius plot (see methods for details) reveals two distinct processes characterized by different Q10s, the threshold is recorded as the intersection of the two fit lines. (i) Q10 quantifications (N=13 cells). (j,k) Representative Arrhenius plot from a cell expressing DpersGr28b.d. (l,m) Representative Arrhenius plot from a cell expressing DmojGr28b.d. (n) Systematic differences in threshold reveal species-specificity (N=9, 13, 13, respectively). (o) The threshold recorded from Drosophila S2R+ cells expressing DmelGr28b.d (N=13) is the same as that recorded in HEK293T cells (compare to n); In h, k and m: red shading = increasing temperature. In i, n, and o: black lines and red lines=mean, outlined boxes=one standard deviation, empty boxes=95% CI of the mean, empty circles=datapoints, each datapoint corresponds to one cell threshold recording; *=p<0.05, n.s.=p>0.05, p=[2.9E-6,9.9E-10,1.39E-4, 0.986] for per vs. mel, per vs. moj, mel vs. moj, and mel HEK293T vs. mel S2R+, respectively; 1-way ANOVA.

Figure 3.. Cross-species transfer of Gr28b.d reveals a switch from heat avoidance to attraction in D. mojavensis.

(a,b) Gr28b genomic locus, mutations used, and their predicted effect. (a) DmelGr28b. (b) DmojGr28b. (c-e) RNAseq from D. mojavensis aristae demonstrates expression of the Gr28b.d variant. (c) Schematic. (d) Alignment of reads to the locus. (e) Quantification of expression; neuronal marker nSyb, Ir25a, and Ir93a are shown for comparison. (f-i) A Gr28b.d promoter fusion drives GFP expression in the 3 hot cells of the arista. (f,g) D. melanogaster. (h,i) D. mojavensis. (g, i) Confocal stack of transgenics (blue: cuticle, green: GFP; scale bars = 20 μm). (j) Control D. melanogaster avoid heat above 25°C (N=11 each), DmelGr28b^exc. mutants lose avoidance of 27.5°C and 30°C (empty arrowheads; N=8 each), expression of DmelGr28b.d rescues avoidance starting at 27.5°C (arrowheads; N=8 each), while expression of DmojGr28b.d rescues avoidance to 30°C (arrowhead) but not 27.5°C (empty arrowhead; N=10 each). (k) Wild type D. mojavensis are indifferent to 27.5°C but attracted to 30°C and 35°C (N=6 each), DmojGr28b.d^Δ1 mutants lose attraction to 30°C and 35°C (empty arrowheads; N=8 each), while expression of DmelGr28b.d in mutants rescues attraction to 30°C and 35°C (arrowheads) but also produces attraction to 27.5°C (empty arrowhead; N=7,7,7,6 for test temperatures from low to high). (in j and k: 2-choice, base temperature = 25°C, red lines=mean, filled boxes=one SD, empty boxes=95% CI of the mean, empty circles=groups,15 flies/group; a black box and/or Ø = not different from zero, two-tail one sample t-test, p<0.05; p-values for test temperatures from low to high in j are: p=[0.331,9.16e-3,2.21e-3,6.87e-7,1.11e-9] for control; p=[0.854,0.671,0.864,4.94e-4,1.08e-6] for mutant; p=[0.331,9.16e-3,2.21e-3,6.97e-7,1.11e-9] for DmelGr28b.d rescue; and p=[0.362,0.685,6.94e-3,6.49e-6,5.38e-10] for DmojGr28b.d rescue. In k are: p=[0.833,0.986,2.57e-3,8.56e-3] for control; p=[0.845,0.443,0.080,0.282] for mutant; and p=[0.701,7.15e-4,2.43e-4,8.14e-3] for DmelGr28b.d rescue).

Figure 4.. Differential morphology of central pathways that relay heat signals in the D. mojavensis brain.

(a) Thermosensory cell types. (b) PAL glomeruli. (c) Confocal stacks from D. melanogaster (left) and D. mojavensis (right) brains stained by the neuropil marker nc82 (magenta). Below, the axon terminals of hot-activated TRN of the antenna are independently labeled by transgenic expression of GFP (stained by anti-GFP, green; arrows); Scale bars = 50 μm. (d) EM reconstruction of TPN pathways in D. melanogaster; front view. (e) D. melanogaster EM reconstructions paired with representative 2-photon images of similar dye-labeled D. mojavensis TPNs. Cell types: TPN-Ia, TPNIb, TPN-VI, and TPN-V (arrowheads; empty arrowheads= additional pathways). Scale bars = 25 μm. Dye fills statistics are in Extended Data Figure 4. (f) Schematic of cell types involved and pie chart of the common TPN synaptic output of hot-activated TRNs. (g) EM reconstruction of D. melanogaster TPN-V showing innervation of the lACA (asterisk) and ventral LH (arrowhead; front view = AP axis). (h) Representative 2-photon z-stacks of dye-labeled TPN-V, in either (h’) D. melanogaster or (h”) D. mojavensis; front view. In D. mojavensis TPN-V more densely innervates the lACA, as compared to D. melanogaster (asterisks). In both cases, TPN-V veers to innervate the ventral aspect of the LH (arrowheads). (i) EM reconstruction of TPN-V terminals in the CA and LH; top view. Similarly oriented 2-photon stacks demonstrate a broader innervation within the anterior-ventral LH in (i’) D. melanogaster (arrowheads) compared to (i”) D. mojavensis (empty arrowheads). Scale bars = 25 μm. Dotted circles in h and i denote the outlines of the CA and LH, dotted line is for reference across panels. Abbreviations: TRNs, antennal temperature receptor neurons; PAL, Posterior Antennal Lobe; TPNs, 2^nd order thermosensory projections neurons; MB, Mushroom Body; LH, Lateral Horn; CA, Calyx; lACA, lateral Accessory Calyx.

Figure 5.. TPN-V innervation of the Lateral Horn matches behavior rather than following phylogeny.

(a) Phylogenetic tree. (b) Representative 2-photon z-stacks of dye-labeled TPN-V terminals in the CA and LH of D. persimilis; upper panel= front view (Z projection of a scan acquired on the AP axis), illustrating innervation of the CA (lACA, asterisks); lower panel= top view (Z projection of a scan acquired on the DV axis), illustrating innervation of the anterior-ventral LH (arrowheads). (c) D. mettleri avoids heat above 27.5°C in 2-choice assays. (d) The in vitro threshold for heat activation of DmetGr28b.d is ~30°C. (in c and d: red lines=mean, outlined boxes=one standard deviation, empty boxes=95% CI of the mean, empty circles=datapoints, in c N=8 groups of 15 flies, in d N=13 recordings). (e) Representative 2-photon z-stacks of TPN-V terminals in the CA and LH of D. mettleri (upper panel= front view, lower panel= top view, in b and e dotted circles= outlines of CA and LH; dotted line is for reference). (f) 3D reconstructions of TPN-V arborizations from D. melanogaster, D. persimilis, D. mojavensis, and D. mettleri, respectively (N=5 each, aligned within species, see methods for details). In top views, the LH innervation of the D. mojavensis TPN-V stands out as lacking any projections in the anterior-ventral LH (empty arrowhead). (g) A distance matrix for reconstructions in f (see methods for details) demonstrates distinct clustering of the D. mojavensis TPN-V arborizations within the LH. Cell-cell distances are represented both as a dendrogram (left) and UMAP (clustering P value=1.97e-5; SigClust test). In g every dendrogram line or UMAP dot represents a reconstruction in f. In a and f, red/green dots= heat avoidance/attraction. In g color represents species. LH: Lateral Horn; CA: Calyx; lACA: lateral Accessory Calyx. Scale bars in b,e and f = 25 μm.