Principal neurons of the V glomerulus show distinct response properties for CO2.

(A–C) About 40 olfactory sensory neurons (ORNv) innervate the ipsilateral V glomerulus at the ventral AL surface (A). Single cell clones revealed the terminal branch pattern of ORNv axons (B). Each ORNv axon forms 3–5 non-overlapping terminal branches (C). (D–G) Morphology of central neurons of the V glomerulus, including 2 types of projection neurons (PNvuni and PNvbi, D), 2 classes of broad local interneurons (LNv1 and LNv2, E/E’) and 2 classes of sparse LNs (LN23 and LN19, F, G). Gal4 expression lines for sparse interneurons LN23 and LN19 (F’, G’). (H–N) CaLexA reporter expression in different neuron classes under ambient air (H–N) and elevated CO2 levels (H’–N’). CaLexA activity could be observed in ORNv, LNv1/2 and LN19, but not in glomerular PNs under ambient air conditions. (H’-N’) Following chronic CO2 stimulation, ORNv and LN23 show high activity, comparable to the GFP expression levels at ambient air. Based on the CaLexA-GFP intensity, previous activity was higher for PNvbi following stimulation as was the case for LNv1 and LNv2. PNvuni showed no activity as indicated by the GFP levels. (O) Quantification of GFP signal intensity in ORNv, LN23, PNvbi, PNvuni, LNv1, and LNv2. The experimental conditions were, that the flies were either kept in ambient air or in 5% CO2 for 3 consecutive days. For ORNs, LN23 and PNvbi we also exposed the flies to spermidine, which is a strong GR21a (ORNv) inhibitor. The effect was so strong for LN23, so that we also wanted to compare it to a complete bilateral ablation of both antennae. Further, since continuous exposure to CO2 did not produce any GFP signal for PNvuni, we tried a pulsed presentation of the CO2 stimulus, but that also showed no effect. The white bars at the lower right corners indicate the scale of 20 ym. This also applies to the following figures.

Circuit dynamics of the V glomerulus during CO2-induced habituation.

(A) T-maze setup to determine CO₂ valence identity of adult flies. (B) Performance index (PI) of naïve flies compared to flies, which have experienced either chronic CO2 exposure, removal of both antennae or a mutation of the CO2 receptor GR63a, each of it leading to a loss of innate aversion response towards CO2. A value of 100 for the PI represents a 100% attraction towards ambient air for a tested group of flies, while a negative value of –100 would indicate a 100% attraction towards 5% CO2. (C) Temporal pattern of CO2-induced behavioral habituation. It became evident, that even after 1 day of exposure, that aversion towards 5% CO2 already decreased drastically. (D, E) Structural changes of the V glomerulus volume (D) and axonal arborization of ORNv (E) following 5 days of chronic CO₂ stimulation. (F–Q) Synaptic connectivity among neurons of the V glomerulus analyzed by transgenic GRASP. Activity-dependent sybGRASP indicates enhanced ORNv→PNvbi connectivity (F’, H), but no significant changes for ORNv→LN23 (G’, H) during behavioral habituation. Activity-independent tGRASP expression demonstrates increased synaptic connectivity of ORNv axons with PNvbi and LN23, consistent with morphological changes observed in ORN terminals during habituation (I’, J’, K). ORN input is restricted to the glomerulus but not seen in the extraglomerular region (arrows, J).Differences in SybGRASP levels among central neurons of the V glomerulus PNvbi neurons showed higher GFP intensity after prolonged exposure to elevated CO2 levels (O, P compared to Q). The GFP levels for LN23 remained similar to the ambient air control (L-Q)

Neuronal polarity of CO2 glomerular circuit components.

(A-C) confocal sections illustrate syt::GFP and DenMark::Cherry expression in LN23 Quantification of GFP/cherry levels syt::GFP and DenMark::Cherry revealed distinct localization of presynaptic sites in PNvbi, PNvuni, and LN23. PNvbi displayed prominent presynaptic labeling throughout the glomerular neuropil, while PNvuni showed no detectable presynaptic signal, despite clear dendritic arborizations. LN23 exhibited scattered presynaptic puncta largely outside the glomerular boundaries. (D) PNvuni LN23 and PNvbi show different responses following habituation. While LN23 did not show increased syt::GFP or DenMark::cherry signal, both were increased for PNvbi while for PNvuni only DenMark::cherry was amplified. (E) Syb::GFP labeling further confirmed that presynaptic sites in PNvbi were restricted to the glomerular neuropil and increased after chronic CO₂ exposure, whereas LN23 signals remained scattered along the glomerular margin. (F-G) Higher-resolution optical sections confirm the presence of brp::GFP puncta in LN23 dendrites outside the V glomerulus (F, arrow) but confined to the V glomerulus in PNvbi. (H) Quantification of syt::GFP and DenMark::cherry. For LN23 we differentiated between intensity measurements in the center of the glomerulus and measurements near the glomerular boundary. (I) Hemibrain reconstructions show the enrichment of LN23 presynaptic sites (colored spheres) in extraglomerular domains compared to the intraglomerular distribution in PNvbi

Early LN23 development.

(A-E’) At the beginning of pupal development (0h APF), larval LN23 neurons remodel their ventral processes (arrow in A) and start extending into the developing adult AL (arrow in B). While a broad arborization can be observed in the ventral AL (arrows in C and D), a small process extends in the dorsal AL to target posterior to the V glomerulus (arrowheads in C and D). At 40h APF, the main LN23 dendrites are restricted to the glomerular boundaries (arrow in E). (F-G’’) In the reorganization from the larval to adult AL, LN23 processes (green) synchronize with uni-glomerular PN dendrites (red). Compared to wild-type (H), loss of Wnt5 (I-K) leads to a larger separation of ventral and dorsal LN23 branches (arrows in I and J), which later extend along axons towards higher brain centers (arrow in K).

Activity of LN23 controls CO2 aversion behavior.

(A) Cell type–specific silencing with UAS-shibirets revealed that neuronal activity of all three V-glomerulus relay neurons contributes to innate CO₂ avoidance. Silencing of PNvuni or PNvbi reduced avoidance behavior, with a stronger effect in PNvuni. In contrast, LN23 silencing abolished CO₂ aversion entirely. Control experiments performed at room temperature (rt; exp stands for experimental condition at restrictive temperatures), where shibirets should remain permissive, showed partial reduction of avoidance in PNvuni and LN23. Since experiments with PNvbi Gal4 (VT031497) showed inconclusive results, we used the alternative PNvbi (alt. PNvbi) Gal4 driver line (R53A05; BDSC#38859). (B) Cell type–specific expression of shibirets via CaLexA confirmed the essential role of LN23 in CO₂ behavior. LN23 CaLexA-shibirets flies showed reduced avoidance, and chronic CO₂ exposure, which should boost CaLexA activity, shifted behavior from avoidance to attraction. (C) Optogenetic activation using UAS-ReaChR demonstrated sufficiency of LN23 for driving aversive behavior. Light activation of ORNv (GR21a-Gal4) and LN23 (independent drivers LN23-Gal4 and alternative line 48408-Gal4) triggered robust avoidance, while activation of PNvbi or PNvuni failed to elicit significant changes. (D) CaLexA-driven ReaChR expression in LN23 also induced light-mediated aversion, confirming cell type specificity.

Two postsynaptic neurons of LN23 convey opposing odor valence.

(A) Connectomic analysis (neuPrint, Janelia Research Campus) revealed two major postsynaptic partners of LN23, PNm1 and PNm17. (B, C) Single-cell MCFO clones of sm_PNm1 and l2_PNm17 show axonal projections bypassing the canonical olfactory lateral horn (LH) and targeting adjacent regions including the posterior lateral protocerebrum (PLP), superior clamp (SCL), and superior lateral protocerebrum (SLP). (D) Optogenetic activation of PNm1 induced strong aversion, whereas PNm17 activation elicited attraction, demonstrating opposing behavioral valence. (E, F) Silencing experiments with shibirets confirmed that CO₂ aversion requires PNm1 activity but is unaffected by PNm17 silencing. (G–I) Hemibrain connectome data identified distinct downstream LH partners for the two CO₂ relay pathways: LHPD5c1 as the primary postsynaptic target of PNvbi, LHPV9b1 downstream of PNm1, and LHCENT4 downstream of PNvuni. The abbreviation “rt” stands for the room temperature control of shibirets while “exp” is the experimental condition at the restrictive temperature. The midline in the brain is indicated by the dashed line and “ml.”.

The “food” odor associated l2LN19.

(A) LN19 gets input from DM4, DC4/DP1 and VM7 while (B) the output is within the V glomerulus. (C) Table illustrating the post,– and presynaptic partners of LN19 as they can be found on https://neuprint.janelia.org/ (D) Optogenetic activation of LN19 using a specific driver line induced strong behavioral aversion, indicating that LN19 activity can bias olfactory responses towards avoidance. (E) In contrast, silencing LN19 via shibirets did not alter innate CO₂ avoidance, suggesting that while LN19 is sufficient to promote aversive responses, it is not essential for baseline CO₂-driven behavior. The abbreviation “rt” stands for the room temperature control of shibirets while “exp” is the experimental condition at the restrictive temperature.

Convergence of parallel channels for CO2 valence coding.

The activity of CO2 –responsive sensory neurons (ORNv, black) is relayed by 4 parallel PN channels to converge onto 2 central brain target neurons of opposite intrinsic valence identity. First, the glomerular PNvbi (red) and PNvuni (pink) target the ventral domain of LHPD5 (directly or via LNcent4, respectively) to trigger behavioral aversion (red). Second, the extraglomerular PNm1 (yellow) and PNm17 (turquoise), activated via the lateral LN23 relay (gray), converge onto LHPV9 (directly or via CRE108, respectively) which mediates attraction (green). While PNm17 supports LHPV9 activity, GABAergic PNm1 blocks LHPV9 thereby resulting in an aversive response. This core aversion domain (PNvbi/PNvuni /PNm1) can be modified either via the activation of the dorsal LHPD5 domain by fruit odor PNs (PNDM/VM/DC, green) or an increase of PNm17 by non-olfactory sensory neurons (turquoise). PNvbi is providing balancing activity between channels, for example through direct input onto LHPV9 or via receiving LN19-mediated inhibition by food PNs.

(A) Schematic of the canonical olfactory pathway (left) and the identified AL relay pathway (right). (B) CaLexA::GFP expression in vORN cell bodies on the 3rd antennal segment.

(A) CaLexA labeling of LN23 combined with anti-GABA staining revealed no colocalization, indicating that LN23 is not GABAergic.

Loss of the cell adhesion molecule Flamingo in ORNv.

In contrast to the (A) wild-type axonal convergence of ORNv within the glomerular boundaries, (B) fmi-mutant ORNv axons overshoot the ventral glomerular domain and extend into the posterior AL (arrow).