Recurrent connectivity supports carbon dioxide sensitivity in Aedes aegypti mosquitoes

Abstract

The mosquito Aedes aegypti's human host-seeking behavior depends on the integration of multiple sensory cues. One of these cues, carbon dioxide (CO₂), gates odorant and heat pathways and activates host-seeking behavior. The neuronal circuits underlying processing of CO₂ information remain unclear. We used automated serial-section transmission electron microscopy (EM) to image and reconstruct the circuitry of the glomeruli that are innervated by the Ae. aegypti maxillary palp, including the glomerulus that responds to CO₂. Notably, CO₂-sensitive olfactory sensory neurons (OSNs) make high levels of recurrent synaptic connections with one another, while making a low density of feedforward synapses. At some of these contacts between CO₂ OSNs, we observe ribbon-like presynaptic structures, which may further enhance recurrent signaling. We compared both feedforward and recurrent connectivity with all olfactory glomeruli in Drosophila melanogaster, and we found more recurrent connections between the Ae. aegypti CO₂-responsive OSNs than in any D. melanogaster glomeruli. We developed a computational circuit model that demonstrates recurrent synapses are necessary for robust CO₂ detection under normal physiological conditions. Together, elevated levels of recurrent connectivity and ribbon-like structures may amplify sensory information detected by CO₂-sensitive OSNs to support mosquito activation and sensitization by CO₂, even in the presence of high levels of other odorants in the environment. We propose that this circuit organization supports the salience of CO₂ as a mosquito host cue.

Figure 1:. Aedes aegypti mosquito chemosensory driven behaviors and EM reconstruction of maxillary palp glomeruli.

(A) Schematic of CO₂ activation and sensitization in Ae. aegypti host-seeking. (left) Quiescent adult female mosquitoes become activated upon CO₂ detection. (right) Sensitization to additional host cues (eg. odor, heat, etc.) drives host-seeking behavior. (B) Diagrams (top) of an adult female Ae. aegypti head and (bottom) a basiconic sensillum on maxillary palp (gray) which contain dendritic processes of olfactory sensory neurons (OSNs), including CO₂-sensitive Gr3-expressing OSNs. OSN axons project centrally to the antennal lobes in the brain. (C) Volumetric rendering of the mosquito brain neuropils including the antennal lobes (light blue) (Heinze et al., 2021; Matthews et al., 2019). Scale bar 100 μm. (D) Schematic of an antennal lobe glomerulus and the circuit components analyzed here. Multiple OSNs selective to a chemosensory cue (eg. odor or CO₂) converge on a glomerulus where they make synapses onto a cognate projection neuron (PN). OSNs can also make reciprocal synapses onto other OSNs. (E) Coronal 40 nm-thick section through maxillary palp Glomerulus 1, acquired and aligned using large-scale, tape-based transmission electron microscopy (EM) (Phelps et al., 2021). Scale bar 25 μm. Magenta inset expanded in (F) (left) Example OSN (yellow) to OSN (blue) synapse (black arrowhead). (right) Polyadic synapse (black arrowhead) in a Glomerulus 1 OSN (yellow). Red: postsynaptic multiglomerular cell. Blue: postsynaptic Glomerulus 1 uniglomerular projection neuron (uPN). Scale bar 1 μm. (G) Sagittal view of reconstructed maxillary palp nerve OSNs: Glomerulus 1 (pink), 2 (blue), and 3 (green). Scale bar 25 μm. a, anterior; d, dorsal; l, lateral; m, medial; p, posterior; v, ventral. (H) Cable length of OSNs (Kruskal-Wallis test (two-tailed), p = 6.20 x 10⁻⁵, pairwise comparisons using Dunn’s post-hoc test with Bonferroni correction: n.s., p > 0.05; * p ≤ 0.01, ** p ≤ 0.001). (I) Coronal view of individual OSNs that innervate Glomerulus 1, 2, or 3 (gray surface meshes). Scale bar 25 μm. (J) Schematic with reciprocal OSN-to-OSN connections highlighted. (K) Axograms of representative OSNs from each glomerulus and their reciprocal output synapses to other OSNs. Scale bar 25 μm. (L) Total number of outgoing OSN-to-OSN synapses contained within Glomerulus 1, 2, and 3 for ten fully reconstructed OSNs (Kruskal-Wallis test (two-tailed), p = 1.28 x 10⁻⁵, pairwise comparisons using Dunn’s post-hoc test with Bonferroni correction: n.s., p > 0.05, * p ≤ 0.05; ** p ≤ 0.001). (M) Synapse density of outgoing OSN-to-OSN synapses per micrometer of cable overlap (length of postsynaptic OSN axon within 2 μm of presynaptic OSN axon, see Methods) (Kruskal-Wallis test (two-tailed), p = 0.0003, pairwise comparisons using Dunn’s post-hoc test with Bonferroni correction: n.s., p > 0.05; * p ≤ 0.05; ** p ≤ 0.001). (N) Reciprocal OSN-to-OSN synapses as a fraction of overall OSN output synapses (Kruskal-Wallis test (two-tailed), p = 0.0002, pairwise comparisons using Dunn’s post-hoc test with Bonferroni correction: n.s., p > 0.05, * p ≤ 0.01, ** p ≤ 0.001). (G) Schematic of reciprocal connectivity with arrow thickness reflecting synapse number.

Figure 2:. Feedforward connectivity of maxillary palp glomeruli.

(A) Schematic with feedforward OSN-to-PN connections highlighted. (B) Dendrograms (flattened 2D representations of the 3D dendritic arbor preserving distance information) of uPNs and their incoming feedforward synapses from OSNs. * denotes branch leading to soma, and arrow denotes branch leading to the inner antennocerebral tract (iACT). (top) uPN cable length, glomerular volume, and cable density. Scale bar 100 μm. (C) Coronal view of uPN reconstructions. Scale bar 25 μm. (D) Axograms (flattened 2D representations of the 3D axonal arbor preserving distance information) of representative OSNs with their feedforward synapses to uPNs shown as colored dots. Scale bar 25 μm. (E) Number of OSN-to-uPN synapses from ten fully reconstructed OSNs within Glomerulus 1, 2, and 3 (Kruskal-Wallis test (two-tailed), p = 0.0004, pairwise comparisons using Dunn’s post-hoc test with Bonferroni correction: n.s, p > 0.05; * p ≤ 0.05; ** p ≤ 0.001). (F) Density of OSN-to-uPN synapses per μm of cable overlap (length of uPN dendrite within 2 μm of OSN axon, see Methods) (Kruskal-Wallis test (two-tailed), p = 3.34 x 10⁻⁵, pairwise comparisons using Dunn’s post-hoc test with Bonferroni correction: n.s., p > 0.05; * p ≤ 0.01; ** p ≤ 0.001). (G) Feedforward OSN to uPN synapses as a fraction of overall OSN output synapses (Kruskal-Wallis test (two-tailed), p = 2.49 x 10⁻⁶, pairwise comparisons using Dunn’s post-hoc test with Bonferroni correction: * p ≤ 0.05, ** p ≤ 0.001). (H) Total number of outgoing OSN synapses to all cell types (Kruskal-Wallis test (two-tailed), p = 4.5 x 10⁻⁵, pairwise comparisons using Dunn’s post-hoc test with Bonferroni correction: n.s., p > 0.05, * p ≤ 0.01; ** p ≤ 0.001).

Figure 3:. Feedforward connectivity of Ae. aegypti Glomerulus 1 compared to D. melanogaster antennal lobe glomeruli.

(A) Schematics of cells and connectivity compared between Ae. aegypti and D. melanogaster. (left) The Ae aegypti CO₂-sensitive Glomerulus 1 with 45 unilateral OSNs and the 1 uPN that projects ipsilaterally through the iACT to higher brain regions. Dashed lines indicate areas beyond the main reconstruction. (middle) The D. melanogaster CO₂-sensitive glomerulus V with 41 unilateral OSNs and 3 uPNs: one unilateral projecting uPN and two bilateral projecting uPNs, with one bilateral uPN on each side of the brain. (right) Representation of D. melanogaster glomeruli that contain a single uPN. These contain between 5 and 60 bilateral OSNs, which each innervate the corresponding glomerulus on both sides of the brain and have 1 uPN that projects ipsilaterally to higher brain regions. (B) Number of feedforward OSN-to-uPN synapses for each glomerulus (Kruskal-Wallis test (two-tailed), p = 3.93 x 10⁻⁵⁶). Filled data are non-significantly different from Glomerulus 1 using Dunn’s post-hoc test with Bonferroni correction: p ≤ 0.01). (C) Density of feedforward OSN-to-uPN synapses per μm of cable overlap within each glomerulus (Kruskal-Wallis test (two-tailed), p = 1.39 x 10⁻⁴⁵. Filled data N.S. are non-significantly different using Dunn’s post-hoc test with Bonferroni correction: p ≤ 0.01). (D) Feedforward OSN-to-uPN synapses as a fraction of overall OSN output synapses (Kruskal-Wallis test (two-tailed), p = 4.87 x 10⁻³⁸. Filled data are non-significantly different using Dunn’s post-hoc test with Bonferroni correction: p ≤ 0.01). (E) OSN axonal cable length within each glomerulus. (Kruskal-Wallis test (two-tailed), p = 1.03 x 10⁻⁶². Filled data are non-significantly different using Dunn’s post-hoc test with Bonferroni correction: p ≤ 0.01).

Figure 4:. Ae. aegypti Glomerulus 1 has elevated reciprocal OSN-to-OSN connectivity.

(A) Schematic with reciprocal OSN-to-OSN connections highlighted. (B) Total number of outgoing OSN-to-OSN synapses contained within each glomerulus (Kruskal-Wallis test (two-tailed), p = 1.63 x 10⁻⁹⁸. Ten fully reconstructed OSNs are included for Ae. aegypti CO₂ sensitive Glomerulus 1. Filled data are non-significantly different using Dunn’s post-hoc test with Bonferroni correction: p ≤ 0.01). (C) Density of reciprocal OSN-to-OSN synapses per μm of cable overlap within each glomerulus (Kruskal-Wallis test (two-tailed), p = 1.39 x 10⁻⁷². Filled data are non-significantly different using Dunn’s post-hoc test with Bonferroni correction: p ≤ 0.01) (D) Reciprocal OSN-to-OSN synapses as a fraction of overall OSN output synapses (Kruskal-Wallis test (two-tailed), p = 1.13 x 10⁻⁷². Filled data are non-signficantly different using Dunn’s post-hoc test with Bonferroni correction: p ≤ 0.01). (E) Schematic summarizing difference between CO₂ selective OSN connectivity between Ae. aegypti and D. melanogaster.

Figure 5:. Ribbon-like synapses among CO₂ OSNs.

(A) Schematic of a canonical T-bar synapse and (B) a putative ribbon-like synapse in Ae. aegypti Glomerulus 1. Scale bars 500 nm. (C) Axogram of an example Glomerulus 1 OSN with reciprocal output T-bar (blue) and putative ribbon-like (red) synapses represented by colored dots. Arrow points to a putative ribbon-like synapse from this OSN. Scale bar 25 μm. (D) EM micrographs of canonical T-bar (blue arrows) synapses. Scale bars 500 nm. (E) EM micrographs of putative ribbon-like (red arrows) synapses. Scale bars 500 nm. (F) Higher magnification (0.5 nm/pixel) EM micrographs of putative ribbon-like synapses with vesicles docked to the presynaptic membrane (arrowheads). Scale bars 200 nm. (G) Histogram of putative ribbon-like and T-bar OSN-to-OSN synapse branch order (Mann-Whitney U test, p = 5.93 x 10⁻¹⁷). Inset: illustration of synapse branch order. (H) Histogram of distances (cable length) from the first branch point to putative ribbon-like and T-bar OSN-to-OSN synapses (Mann-Whitney U test, p = 6.78 x 10⁻⁹). Inset: illustration of quantified distance (geodesic distance or cable length) of synapse from the first branch point.

Figure 6:. Connectomically-informed network model reveals recurrent connectivity specifically enhances stimulus detection.

(A) Model architecture. CO₂ is processed by Glomerulus 1, which includes recurrent connections. Co-occurring odorants (‘background’) activate glomeruli which inhibit other glomeruli through inhibitory local neurons (LNs). (B) OSN to PN response curves for OSNs with and without recurrent connections, for two different background odor strengths (average and strong, see Methods). (C) OSN to PN response curve for differing values of g for an average background odor strength. Horizontal gray shaded region indicates OSN response rates for which CO₂ should be detected (35 sp/s), and vertical gray shaded region indicates the PN response strength that would correspond to CO₂ detection (0.4R_max). (D) Minimum value of g required to ensure CO₂ detection for varying background odor strengths. Vertical gray shaded region indicates an average level of background odor.