Non-Canonical Odor Coding in the Mosquito

Female Aedes aegypti mosquitoes are a persistent human foe, transmitting arboviruses including dengue and yellow fever when they bite us to obtain a blood meal. Mosquitoes are intensely attracted to human-emitted body odor, heat, and carbon dioxide, which they detect using three different large multi-gene families encoding odor-gated ion channels. Genetic mutations that cause profound disruptions to the olfactory system have modest effects on human attraction, suggesting significant redundancy in odor coding. The canonical view is that olfactory sensory neurons each express a single chemosensory receptor that defines its ligand selectivity. Using immunostaining, RNA in situ hybridization, and single nucleus RNA sequencing, we discovered that Aedes aegypti uses an entirely different organizational principle, with many neurons co-expressing multiple chemosensory receptor genes. In vivo electrophysiology demonstrates that the broad ligand-sensitivity of mosquito olfactory neurons is due to this non-canonical co-expression. The redundancy afforded by an olfactory system in which many neurons co-express multiple receptors with different chemical sensitivity may greatly increase the robustness of the mosquito olfactory system and explain our longstanding inability to engineer new compounds that disrupt the detection of human body odor by mosquitoes.


SUMMARY
Female Aedes aegypti mosquitoes are a persistent human foe, transmitting arboviruses 27 including dengue and yellow fever when they bite us to obtain a blood meal. Mosquitoes are 28 intensely attracted to human-emitted body odor, heat, and carbon dioxide, which they detect 29 using three different large multi-gene families encoding odor-gated ion channels. Genetic 30 mutations that cause profound disruptions to the olfactory system have modest effects on 31 human attraction, suggesting significant redundancy in odor coding. The canonical view is that 32 olfactory sensory neurons each express a single chemosensory receptor that defines its ligand 33 selectivity. Using immunostaining, RNA in situ hybridization, and single nucleus RNA 34 sequencing, we discovered that Aedes aegypti uses an entirely different organizational 35 principle, with many neurons co-expressing multiple chemosensory receptor genes. In vivo 36 electrophysiology demonstrates that the broad ligand-sensitivity of mosquito olfactory neurons 37 is due to this non-canonical co-expression. The redundancy afforded by an olfactory system in 38 which many neurons co-express multiple receptors with different chemical sensitivity may 39 greatly increase the robustness of the mosquito olfactory system and explain our longstanding 40 inability to engineer new compounds that disrupt the detection of human body odor by 41 mosquitoes. 42 Increased global travel, a growing world population, and rising temperatures increase the 44 emergence and transmission of novel disease-causing pathogens spread by "vector" 45 organisms such as mosquitoes, ticks, sandflies, and fleas. Diseases spread by these 46 arthropods collectively account for more than 700,000 deaths every year (WHO, 2020). 47 Female Aedes aegypti mosquitoes spread arboviruses including dengue, Zika, yellow fever, 48 and chikungunya. Only female mosquitoes bite, and they do so because they require a blood-49 meal for reproduction (Allan et al., 1987). Aedes aegypti prefer to bite human hosts, which 50 contributes to their effectiveness as a disease vector (Brown et al., 2014;Gouck, 1972;51 McBride et al., 2014). To identify human hosts, mosquitoes rely heavily on chemosensory 52 cues, including carbon dioxide (CO2) emitted from breath, and human body odor, which is a 53 mixture of hundreds of different individual odorants including alcohols such as 1-octen-3-ol and 54 volatile amines such as ammonia ( human host cue that activates and attracts mosquitoes (Gillies, 1980). In Aedes aegypti, 72 Gustatory Receptor 3 (Gr3) encodes an essential subunit of the CO2 receptor, and Gr3 mutant 73 mosquitoes lose all sensitivity to CO2 (McMeniman et al., 2014).

75
Because mosquitoes specialize on humans and require blood to reproduce, the drive to find 76 humans is strong and innate. Indeed, even mosquitoes genetically engineered to eliminate 77 genes critical for peripheral detection of host sensory cues can find and bite people. Animals despite widespread efforts by humans to mask our odor with chemical repellents (Tawatsin et 87 al., 2006;Travis et al., 1949).
We have yet to identify long-lasting interventions to prevent this 88 deadly biting behavior, and it is not known how the mosquito olfactory system is seemingly 89 infallible in its ability to detect humans. 90 The cloning of the first odorant receptors in 1991 (Buck and Axel, 1991) led to the subsequent 91 discovery that each vertebrate olfactory sensory neuron expresses a single odorant receptor 92 that specifies its functional properties. With few exceptions, the well-studied olfactory system of 93 Mus musculus mice features olfactory sensory neurons that are thought to express a single 94 olfactory receptor (Bashkirova and Lomvardas, 2019; Chess et al., 1994). The same 95 organization was reported in Drosophila melanogaster flies (Clyne et al., 1999; Gao and 96 Chess, 1999;Vosshall et al., 1999), although recent work challenges this model (McLaughlin 97 et al., 2021; Task et al., 2021). In both species, decades of evidence has supported the model 98 that neurons expressing a given receptor project axons to dedicated olfactory glomeruli in the 99 first sensory processing center in the brain, the antennal lobe in insects (Couto et al., 2005;100 Fishilevich and Vosshall, 2005; Vosshall et al., 2000) and the olfactory bulb in vertebrates 101 (Mombaerts et al., 1996;Ressler et al., 1994;Vassar et al., 1994). This "one-receptor-to-one-102 neuron-to-one-glomerulus" organization is believed to be a widespread motif in insect olfactory 103 systems, and the convergence onto discrete glomeruli is hypothesized to permit the brain to 104 utilize combinatorial coding and parse which subpopulation of olfactory neurons is activated by 105 a given odorant (Bisch-Knaden et al., Semmelhack and Wang, 2009;Wang et al., 2003).

107
Consistent with this "one-receptor-to-one-neuron-to-one-glomerulus" organization in insects, 108 the number of expressed chemosensory receptors in the OR and IR gene families in many 109 insects roughly correlates to the number of olfactory glomeruli. This holds true in the honey 110 bee Apis mellifera (~180 receptors/~160 glomeruli) ( 2020). This leads to the question of how 120 the mosquito olfactory system is organized to accommodate so many receptors and whether 121 this deviation from rules established in other species explains their exquisite ability to locate 122 human hosts. 123 124 In this study, we developed a CRISPR-Cas9-based genetic knock-in strategy in Aedes aegypti 125 to generate genetically-modified mosquito strains that label molecularly distinct populations of 126 olfactory sensory neurons. We used these strains to understand how the mosquito olfactory 127 system is organized and discovered that OR-and IR-expressing olfactory sensory neurons 128 frequently innervated the same antennal lobe glomeruli. To ask if this was a feature of 129 individual olfactory neurons expressing multiple chemosensory receptors, we profiled receptor 130 expression in peripheral sensory organs using RNA in situ hybridization and by 131 immunostaining with antibodies that recognize endogenous OR and IR co-receptors. To 132 complement these studies, we carried out single nucleus RNA sequencing to profile gene 133 expression in the antennae and maxillary palps. Through these experiments, we found that the 134 olfactory system of Aedes aegypti does not have the expected "one-receptor-to-one-neuron-to-135 one-glomerulus" organization seen in other organisms. We frequently observed co-expression 136 of multiple chemosensory receptors from at least two of the three receptor gene superfamilies 137 within individual olfactory sensory neurons. We also saw expression of multiple receptors from 138 a single family within the same olfactory sensory neuron. To test if multiple receptors function 139 to detect different ligands within the same olfactory sensory neuron, we used in vivo 140 electrophysiology to examine odorant responses in the maxillary palp. We discovered a class 141 of neurons that expresses members of both the OR and IR gene family and that responds to 142 odorants that activate either OR or IR pathways. When we mutated either the OR or IR 143 pathway by deleting the major co-receptors, neurons retained responsivity to the odorant 144 sensed by the pathway that was still intact. Therefore, both ORs and IRs are required to detect 145 different classes of odorants in the same sensory neuron. This sensory organization, in which 146 multiple receptors responding to different chemosensory stimuli are co-expressed, suggests a 147 redundancy in the code for human odor. We speculate that this unconventional organization 148 underlies the robust, seemingly unbreakable properties of the Aedes aegypti olfactory system 149 in detecting human odor and driving human host-seeking in this olfactory specialist. 150 151

152
Mismatch in chemosensory receptor and olfactory glomerulus number suggests a novel 153 olfactory organization 154 In the mosquito, olfactory cues are sensed by olfactory sensory neurons in the antenna and 155 the maxillary palp, whose axons project to the ipsilateral antennal lobe of the brain (Distler and 156 Boeckh, 1997; Ignell et al., 2005) ( Figure 1A-D, Figure S1A-C). The antennal lobe, the insect 157 equivalent of the vertebrate olfactory bulb, is organized into discrete olfactory glomeruli in 158 which axons from peripheral olfactory sensory neurons terminate and synapse with local 159 interneurons and projection neurons that relay olfactory information to the higher brain 160 (Stocker, 1994 glomeruli (64.9 ± 0.9, mean±SEM), obtained by counting antennal lobe glomeruli in the left 164 hemisphere of 12 female Aedes aegypti brains stained to reveal synaptic neuropil ( Figure  165 1B,I,K, Figure S2-5). The glomerulus count ranged from 60-72 glomeruli per antennal lobe, 166 indicating a high level of variability in the organization of the antennal lobe. We generated 3-D 167 reconstructions of complete antennal lobes and saw considerable variability in the size and 168 shape of the glomeruli ( Figure S1). We were able to consistently identify certain landmark 169 glomeruli, most notably the three glomeruli that are innervated by the maxillary palp (Ignell et  The canonical "one-receptor-to-one-neuron-to-one glomerulus" organization posits that the 173 number of chemosensory receptors should roughly match the number of glomeruli in the 174 antennal lobe ( Figure 1D could function in the olfactory system ( Figure 1E). Reanalysis of previously published antennal 179 and maxillary palp RNA-sequencing data (Matthews et al., 2016) using multiple expression 180 thresholds demonstrates that even at the conservative threshold of 5 transcripts per million 181 (TPM), the mosquito olfactory system expresses 102 chemosensory receptors, and adjusting 182 the threshold to 2, 1, or 0.5 TPM increases the number of receptors plausibly expressed to 183 134, 156, and 178, respectively ( Figure 1F,G). Thus, there are many more chemosensory 184 receptors expressed in the olfactory system than available antennal lobe glomeruli, suggesting 185 that the organization of the Aedes aegypti olfactory system must differ from the canonical 186 scheme. We speculate that the mismatch can be resolved by expressing multiple receptors per 187 neuron or having multiple molecularly distinct neurons co-converge on a single glomerulus or 188 both ( Figure 1H).

190
To begin to distinguish between these two organizational principles, we generated a collection 191 of CRISPR-Cas9 gene-targeted strains that label subpopulations of olfactory neurons using 192 the Q-system, a binary expression system similar to Gal4/UAS (Brand and Perrimon, 1993) 193 that uses cell type-specific expression of the QF2 transcription factor to induce expression of 194 an effector from the QF2 binding QUAS enhancer (Potter et al., 2010;Riabinina et al., 2015;195 Riabinina et al., 2016). We introduced an in-frame insertion that replaced the stop codon of 196 each of the co-receptors Orco, Ir25a, Ir8a, and Ir76b, as well as the CO2 receptor subunit Gr3  197 with the transcription factor QF2 ( Figure 1I, Figure  These gene-sparing knock-in strains were designed to cause minimal disruption to the locus to 200 increase the likelihood that they would faithfully report expression of the endogenous gene. We 201 crossed these QF2 driver lines individually to a QUAS-CD8:GFP reporter to label neuronal 202 membranes and visualized axonal projection patterns in the antennal lobe.

204
Orco, Ir25a, Ir8a, and Ir76b co-receptor driver lines were expressed in olfactory sensory 205 neurons with distinct projection patterns in the antennal lobe ( Figure 1I-K). Unexpectedly, 206 neurons that expressed Ir25a projected to almost all glomeruli in the antennal lobe (89.9 ± 207 1.4%, mean±SEM, n=3) ( Figure 1I-K, Figure S3), and expression overlapped extensively with 208 glomeruli labeled by Orco ( Figure 1I-K, Figure S2, Figure S3). While these co-receptor driver 209 lines labeled glomeruli in the same regions from brain to brain, the interindividual expression 210 patterns were not identical, consistent with the variability in glomerular arrangement that we 211 have observed. Neurons that detect CO2 are located in the maxillary palp (Grant et al., 1995;212 Lu et al., Omer and Gillies, 1971) and we saw that Gr3-expressing neurons projected to 213 a large glomerulus in the posterior antennal lobe, Glomerulus 1 ( Figure 1I-K) which is also 214 innervated by Ir25a-expressing neurons. We also noted the presence of a second small 215 glomerulus that was often innervated by Gr3-expressing neurons in the antenna ( Figure S6B). 216 These initial findings point to the overlap of OR-, IR-, and GR-expressing neurons in the 217 antennal lobe of Aedes aegypti, consistent with recent observations in Drosophila 218 melanogaster (Task et al., 2021). 219 220 Co-expression of Orco and Ir25a in the mosquito olfactory system 221 The high degree of overlap between glomeruli labeled by Orco-and Ir25a-expressing olfactory 222 sensory neurons suggests that there is either widespread Orco and Ir25a co-expression within 223 individual sensory neurons or that Orco and Ir25a are expressed in different neurons whose 224 axons co-converge onto individual antennal lobe glomeruli or both ( Figure 1H). To determine if 225 Orco and Ir25a are co-expressed, we adapted the Split-QF2 system (Riabinina et al., 2019) for 226 use in the mosquito. This system "splits" the transcription factor QF2 into two components, the 227 DNA binding domain (QF2-DBD) and the activation domain (QF2-AD) each tagged with a 228 synthetic leucine zipper (Figure 2A,B). When both the QF2-DBD and QF2-AD are co-229 expressed in the same cell, the two domains associate via the leucine zipper, reconstitute a 230 functional QF2 protein, initiate transcription at the QUAS enhancer, and drive expression of a 231 reporter gene ( Figure 2C).

233
Using the same stop-codon replacement approach that we used to generate the QF2-lines, we 234 inserted the QF2-AD into the Ir25a locus to generate IR25a-QF2-AD and the QF2-DBD into 235 the Orco locus to generate Orco-QF2-DBD. When either IR25a-QF2-AD or Orco-QF2-DBD 236 was used to drive expression of dTomato (Shaner et al., 2004), we did not see fluorescence in 237 the female antenna, maxillary palp, or the antennal lobe ( Figure 2D-F, Figure S7). Therefore, 238 neither QF2-DBD nor QF2-AD alone can activate expression from the QUAS enhancer. 239 However, when Orco-QF2-DBD and IR25a-QF2-AD were crossed into the same animal, we 240 saw expression of dTomato in antennal and maxillary palp neurons of female mosquitoes, as 241 well as axonal projections in the antennal lobe ( Figure 2D-F, Figure S7, Figure S8). Nearly half 242 of the glomeruli in the antennal lobe were labelled with dTomato ( Figure 2G-I, Figure S7, 243 Figure S8). This points to widespread Orco and Ir25a co-expression within Aedes aegypti 244 olfactory sensory neurons, although we note that these findings do not rule out the possibility 245 that co-convergence may be present as well.

247
We then examined the olfactory system of male mosquitoes ( Figure 2J-U). We observed 248 extensive expression of dTomato throughout the antenna and in axons that terminate in the 249 antennal lobe when we drove expression with either the Orco-QF2 or IR25a-QF2 driver lines 250 ( Figure 2N,O,R,S). We again saw no expression of dTomato when driven by either the IR25a-251 QF2-AD or Orco-QF2-DBD control lines ( Figure 2L,M,P,Q). However, when Orco-QF2-DBD 252 and IR25a-QF2-AD were crossed into the same animal, we observed widespread dTomato 253 expression in the male antenna and antennal lobe, indicating co-expression of Ir25a and Orco 254 in male olfactory sensory neurons ( Figure 2T-U). It is not currently possible to compare male 255 and female glomerular position due to differences in antennal lobe volume and shape and lack 256 of glomerulus-specific driver lines, however, the general expression pattern was similar 257 between males and females, with innervation predominant in the anterior-medial antennal 258 lobes ( Figure 2F,G,U). 259 Co-expression of Orco and Ir25a in the mosquito taste system 260 Another source of olfactory information in Aedes aegypti may derive from olfactory neurons on 261 the proboscis, the mouthpart of the mosquito that engages in taste and food ingestion.

262
Drosophila melanogaster flies express IRs and GRs in the proboscis, but not ORs (Larsson et 263 al., 2004). In contrast, Orco neurons are widespread in both the proboscises of Anopheles 264 gambiae and Aedes aegypti mosquitoes, and RNA-sequencing data from Aedes aegypti has 265 shown there are many ligand-selective ORs expressed in this taste tissue (Matthews et  was extensive expression of both Orco and Ir25a alone in the proboscis ( Figure 3A) as well as 275 co-expression of Orco and Ir25a as defined by dTomato expression in the Split-QF2 animals 276 ( Figure 3C). Ir25a-expressing neurons send extensive projections to the subesophageal zone, 277 with axons terminating in the anterior and posterior regions of the subesophageal zone. There 278 is a small cluster of glomeruli in the central subesophageal zone that receives dense 279 innervation as well ( Figure 3B). Orco-expressing neurons do not project to the anterior region 280 and send sparse projections to the posterior subesophageal zone and subesophageal zone 281 glomeruli ( Figure 3B). Innervation by the neurons that co-express both Orco and Ir25a send 282 projections only to the posterior-ventral subesophageal zone, with the densest innervation in 283 the medial region and sparser lateral arborizations ( Figure 3D). 284 Since Ir25a complexes mediate not only detection of volatile odorants but also gustatory cues, 285 it is possible that sensory afferents in the Aedes aegypti proboscis are able to detect olfactory 286 as well as gustatory information within the same neurons. Alternatively, IRs and ORs in these 287 neurons may function as olfactory receptors that relay olfactory information to the taste center 288 of the mosquito brain.

290
Extensive co-expression of chemosensory co-receptors in the antenna 291 The observation that nearly half of antennal lobe glomeruli receive projections from neurons 292 co-expressing Ir25a and Orco suggested that there is extensive co-expression of IRs and ORs 293 throughout the antenna. To determine if neurons co-express Orco and Ir25a protein we 294 generated an antibody to Ir25a and conducted whole mount antennal immunostaining with this 295 antibody and a previously characterized Orco antibody (Basrur et al., 2020) to label 296 endogenous Orco and Ir25a proteins in wild-type mosquitoes ( Figure 4A-D). We observed 297 extensive co-expression of Orco and Ir25a ( Figure 4A-D), confirming the co-expression of 298 these distinct chemosensory genes seen using our QF2 and Split-QF2 driver lines. In addition 299 to neurons that contain both Orco and Ir25a protein, we also observed neurons that express 300 either Orco or Ir25a alone ( Figure 4A-D), indicating that OR cells, IR cells, and mixed OR+IR 301 cells exist. We validated the specificity of the Orco and Ir25a antibodies by performing whole-302 mount immunostaining on antennae from Orco and Ir25a mutants ( Figure 4E-H). To confirm 303 and extend these results, we performed RNA in situ hybridization on wild-type antennae with 304 probes designed to target endogenous Orco, Ir76b, and Ir25a RNAs ( Figure 4I-K). These 305 experiments replicated patterns of co-expression observed in immunostained antennae ( Figure  306 4A-D), with almost half of the Orco cells co-expressing Ir25a, and vice versa. In contrast we 307 saw that few Orco cells co-express Ir76b by RNA in situ hybridization. Both the RNA in situ 308 hybridization and immunostaining data indicated that widespread co-expression is not an 309 artifact of the QF2 and split-QF2 driver lines. 310 To gain additional resolution on the degree of overlap between Orco and the three major IR 311 family co-receptors, we carried out whole-mount antennal immunostaining with an antibody to 312 the endogenous Orco protein and to GFP expressed from each sensory neuron QF2 driver.

313
We confirmed extensive co-expression of Orco and Ir25a ( Figure 4L-N) and found that 314 substantially fewer cells co-express either Orco and Ir8a or Orco and Ir76b, even after 315 accounting for fewer total Ir76b and Ir8a cells ( Figure 4L-N). We also note that in addition to 316 widespread co-receptor co-expression, some mosquito olfactory neurons express just one co-317 receptor ( Figure 4O), highlighting the complexity in the rules that govern receptor co-318 expression in Aedes aegypti antennal olfactory sensory neurons. 319 320 Single nucleus RNA sequencing reveals that many antennal neurons co-express 321 multiple ligand-selective receptor subunits 322 Functional ORs and IRs are composed of a complex of co-receptor and ligand-selective 323 receptor subunits. Because there are hundreds of ligand-selective OR and IR genes, it was not 324 feasible to examine combinatorial co-expression of the full complement of receptors by RNA in 325 situ hybridization or immunostaining. Instead, we developed a method for single nucleus RNA 326 sequencing (snRNA-seq) in mosquito antennae based on previously described nucleus 327 extraction protocols from Drosophila melanogaster antennae McLaughlin et al., 328 2021). We isolated antennae from female mosquitoes, extracted nuclei, performed droplet 329 microfluidics to barcode reads from each cell, and performed droplet-based snRNA-seq using 330 the 10x Genomics platform. For clarity of presentation, we use "cell" as the unit of analysis to 331 refer to expression profiling of single nuclei. These experiments were carried out in two 332 batches, Batch 1 at Rockefeller University and Batch 2 at Baylor College of Medicine (see 333 Methods). We filtered for cells based on quality control parameters and combined data from 334 two batches, to capture a total of 14,161 cells ( Figure 5A, Figure S9A-G). Unsupervised 335 clustering was used to categorize cells into broad subtypes, which revealed cells that express 336 epithelial or glial markers ( Figure 5B, Figure S9H). This analysis also yielded 19 neuron 337 clusters based on expression of at least 3 of 4 neural markers (CadN, brp, syt1, and elav) in 338 50% or more of the cells within that cluster ( Figure S9I). A total of 6,645 cells were classified 339 as neurons. We then used unsupervised clustering on this population of neurons and identified 340 54 clusters of antennal neurons ( Figure 5C).

342
To examine the distribution of chemosensory receptors, we averaged expression among cells 343 within an entire cluster and saw cases where multiple receptors were co-expressed ( Figure  344 5D, Figure S10A). Among mean expression levels in the cluster, highly-expressed 345 chemosensory receptors generally belonged to only 1 cluster. Because highly-expressed 346 ligand-selective receptors displayed a strong relationship to individual clusters, we 347 hypothesized that more complex co-expression patterns could be obscured when looking at 348 cluster-level expression patterns. For instance, if a lower-expressed receptor subunit was co-349 expressed with several different highly expressed receptor subunits, cells exhibiting these 350 combinations might be distributed among several clusters and might not be apparent at this 351 level of analysis. We therefore investigated co-expression within individual cells.

353
We first looked at the most highly-expressed receptor subunit pairs in a chord plot and saw 354 several co-expression patterns that were not apparent in the cluster-based analyses ( Figure  355 5E). By replotting the expression of individual cells within clusters using heatmaps, we 356 observed many cases of cells co-clustering that expressed discrete combinations of 357 chemosensory receptors. This indicates that clusters are not a faithful representation of all 358 chemosensory receptor combinations with a given cell ( Figure S10C). This prompted us to 359 investigate other ways of categorizing cells besides clustering to look more broadly at receptor 360 co-expression patterns.

362
To analyze the co-expression partners of a given receptor, we filtered our population of 363 neurons for cells that express a receptor gene above a normalized expression threshold of 0.5 364 log(UMI of gene*10,000 / total UMI of cell +1) ( Figure S11A). We then visualized co-expression 365 using heatmaps. Because receptor expression has a bearing on clustering, we performed 366 unsupervised clustering on these cells as a sorting mechanism to group cells by similarity to 367 visualize patterns on heatmaps. Again, cells often grouped into clusters with clearly identifiable 368 receptor expression patterns, including some that contained multiple receptor subunits ( Figure  369 S11B). Simplified heatmaps of groups of cells with distinct receptor co-expression patterns are 370 illustrated in Figure 5F-I and S11C-F. This analysis revealed extensive co-expression that 371 points to a far greater variety of cell types than previously anticipated. For instance, Or82 372 marks at least 6 different cell types, some that appear to be Or82-specific and others that 373 express an additional one, two, four, or five different ligand-selective receptors. Several of 374 these OR-expressing cell types include one or more ligand-selective IR gene. Ir64a marks at 375 least 6 different cell types that each expresses one or more ligand-selective OR genes ( Figure  376 5F-I, S11C-F). These results reveal extensive and unexpected chemosensory receptor co-377 expression in the mosquito antenna.

379
Coordinated co-expression of chemosensory receptors in the maxillary palp 380 We next examined receptor co-expression in the maxillary palp, a smaller and simpler olfactory 381 organ than the antenna that detects important host cues including CO2 and 1-octen-3-ol, as 382 well as other host odorants (Grant et al., 1995;Lu et al., 2007;McMeniman et al., 2014;Omer 383 and Gillies, 1971). Each female Aedes aegypti maxillary palp contains approximately 35 384 capitate-peg sensilla that each house three chemosensory neurons (McIver, 1982). Based on 385 prior work that examined the morphology and function of the maxillary palp, the neurons within 386 each sensillum are termed the "A", "B", and "C" cells based on their size, from largest to 387 smallest respectively ( Figure 6A-C). The A cell responds to the important host cue CO2 and 388 houses the Gr3 CO2 receptor. The B and C cells both express Orco. The B cell is believed to 389 express Or8, which detects 1-octen-3-ol, while the C cell expresses Or49, which has a less 390 well-defined odorant response profile ( Figure 6A).

392
We hypothesized that these three cell types project to three glomeruli in the antennal lobe. To 393 delineate the organization of maxillary palp projections in the brain, we used our QF2 and 394 Split-QF2 driver lines to examine maxillary palp sensory innervation of antennal lobe glomeruli 395 ( Figure 6C-K). We discovered that Glomerulus 1, which is the largest glomerulus in the 396 antennal lobe (Shankar and McMeniman, 2020), received input from Gr3-expressing sensory 397 afferents. Glomerulus 1 was also innervated by Ir25a-expressing sensory neurons ( Figure 6F-398 K). Glomerulus 2 and Glomerulus 3 received input from Orco-, Ir25a-, and Ir76b-expressing 399 neurons ( Figure 6F-K). Co-expression of Orco and Ir25a in neurons that project to these two 400 glomeruli was confirmed using the Split-QF2 system. In Orco-QF2-DBD, IR25a-QF2-AD 401 animals, Glomerulus 2 and Glomerulus 3 were labeled, but Glomerulus 1 was not ( Figure 6E).

402
These findings suggest that the A, B, and C cells express multiple co-receptors, spanning IR-403 OR and IR-GR classes.

405
To form functional odorant-gated IR or OR complexes, olfactory sensory neurons must carried out multiplexed whole mount RNA in situ hybridization (Choi et al., 2018) in the 410 maxillary palp ( Figure 6L-P, Figure S12). The maxillary palp expresses many fewer 411 chemosensory receptor genes than the antenna, with 18 receptors detected at the 1 TPM 412 threshold in the maxillary palp compared to 138 in the antenna at the same threshold ( Figure  413 1F,G), simplifying the task of selecting genes for expression analysis. We performed RNA in 414 situ hybridization with probes for 10 of the 18 chemosensory co-receptors and ligand-selective 415 receptors that were present in maxillary palp RNA-seq at a threshold of TPM>1. This technique 416 visualized gene expression with sufficient sensitivity that even Or71 and Ir75g (present at 1.93 417 and 1.67 TPM, respectively) were readily detected ( Figure S12O,P).

419
We found no overlap in expression of Orco and Gr3 in the maxillary palp, but Ir25a was 420 expressed in all Orco and all Gr3 cells ( Figure 6L), consistent with our observation that Ir25a-421 expressing neurons project to all three antennal lobe glomeruli. Previous work in Anopheles 422 gambiae suggested that Orco-expressing neurons in the maxillary palp can be evenly divided 423 into two non-overlapping groups: an Or8 population and an Or49 population (Lu et al., 2007). It 424 is widely thought that the same is true in Aedes aegypti. We show definitively that Or8 and 425 Or49 are expressed in segregated populations of Orco-expressing neurons in Aedes aegypti 426 ( Figure 6M) and, when combined with the results of the previous experiment ( Figure 6L), that 427 these cells are also all Ir25a-positive. Additional RNA in situ hybridization experiments 428 revealed that Or8-and Or49-expressing cells also often express Ir76b, with a bias towards 429 expression in Or8-expressing cells ( Figure 6N, Figure S12I-O, Data File 1). Taken together 430 these data show that Orco-expressing olfactory sensory neurons co-express the co-receptor 431 Ir25a and either of the ligand-selective subunits Or49 or Or8, and often co-express the co-432 receptor Ir76b as well.

434
When we analyzed IR ligand-selective subunit expression, we found that Ir100a and Ir93a are 435 selectively expressed in a subset of Or49-expressing neurons ( Figure 6O,P, Figure S12L,M), 436 suggesting that these cells can form functional OR and IR complexes with their respective co-437 receptors in the same neuron and that co-expression of ORs and IRs may be transcriptionally 438 coordinated. Or71 and Or49 were found to be co-expressed, further supporting the idea that 439 multiple ligand-selective ORs can be expressed in an olfactory sensory neuron in Aedes 440 aegypti ( Figure S12O). We also discovered that the ligand-selective receptor Ir75g was 441 expressed in some but not all Gr3-expressing cells, which also express Ir25a ( Figure S12P). 442 Therefore, it is plausible that Gr3 neurons can functionally express both GRs and IRs.

444
Single nucleus RNA sequencing of maxillary palp reveals unanticipated neuronal 445 complexity 446 As mentioned above, the current view in the field is that the Aedes aegypti maxillary palp has a 447 simple organization in which all 35 capitate-peg sensilla are molecularly and functionally 448 identical, each containing the same A, B, and C cells ( Figure 6A). Our RNA in situ hybridization 449 results called this model into question. To examine receptor co-expression in the maxillary palp 450 in greater detail, we carried out snRNA-seq using similar tissue collection and analysis 451 pipelines used for the antenna ( Figure 7A, Figure S13, Figure S14A-C), yielding data from 452 2,298 cells. Using unsupervised clustering, we categorized these cells into epithelia, muscle, 453 glia, and neurons ( Figure 7B-C, Figure S13F). The neuron cluster comprised 630 cells that 454 were further subdivided into four classes that showed remarkable correspondence to cell types 455 previously described in the maxillary palp ( Figure 7C-D, Figure S13G-H). Cluster 4 consists of 456 putative mechanosensory neurons marked by expression of nompC and hamlet. Clusters 1, 2, 457 and 3 were enriched for Gr3, Or8, or Or49, and likely correspond to A, B, and C cells, 458 respectively ( Figure 7D,F,J,K, Figure S14D-G).

460
To investigate co-expression patterns of receptor genes within these clusters, we generated 461 chord plots and found that Or49 was co-expressed with Ir93a ( Figure 7E), confirming our RNA 462 in situ hybridization results. Two ligand-selective IR subunits, Ir41a and Ir161, were co-463 expressed in both the B cell and C cell. We used feature plots to further visualize the 464 expression of individual receptor subunits within the clusters of maxillary palp neurons. A 465 number of ligand-selective receptor genes were present in discrete clusters ( Figure 7F-P, 466 Figure S14H). Confirming our RNA in situ hybridization findings, both the Gr3 cluster and Orco 467 clusters also expressed the Ir25a co-receptor, and many Ir76b-expressing neurons were found 468 in both the B and C cell clusters ( Figure 7F-I, Figure S14D-E).

470
Consistent with RNA in situ hybridization data, Ir93a and Ir100a were expressed in the C cell 471 cluster ( Figure 7L-M). Ir161 was expressed in both B and C cells ( Figure 7O-P), and Or44, 472 Ir41a, and Ir41o were found in all three chemosensory clusters ( Figure 7N, Figure S14H).

473
We did detect low levels of Orco expression in the Gr3 cluster ( Figure 7G) in these snRNA-seq 474 experiments, an observation at odds with results from the three other methods used in the 475 paper and one that we do not currently understand but could be due to differences in the 476 sensitivity of detection methods. We also observed significant variability in expression of 477 ligand-selective and co-receptor subunits across cells ( Figure 7F-P), consistent with RNA in 478 situ hybridization data ( Figure 6L-P). Exploration of the effect of expression level on in vivo 479 receptor function in individual neurons will be an interesting direction of future study.

481
A summary of maxillary palp chemosensory receptor gene expression based on all the data in 482 this study is presented in Figure 7Q. This represents a significant departure from the current 483 view of this sensory system ( Figure 6A). Notably, these data suggest that many B and C cells 484 have all the necessary ligand-selective and co-receptor subunits to form both functional OR 485 and IR receptors.

487
Receptor co-expression expands the functional responses of olfactory neurons 488 We next asked whether this extensive chemosensory receptor co-expression would allow 489 maxillary palp neurons to respond to odorants detected by both ORs and IRs. We used single characteristic ligands in Aedes aegypti. We therefore focused our analysis on stimulus-evoked 500 activity of the A cell and B cell while the mosquito was exposed to CO2 or odorants likely to 501 stimulate either the OR or IR pathway ( Figure 8A-E). 502 503 To determine which family of receptors -GRs, ORs, or IRs -detects a given ligand we 504 recorded odorant responses in wild-type mosquitoes as well as mosquitoes with mutations in 505 Gr3, Orco, or Ir25a. Because the Gr3 and Orco receptor mutants were generated in a different Turner et al., 2011) and it has been proposed that Gr3 is a broadly-tuned receptor that 515 responds to many odorants. We examined the response to a recently identified CO2-neuron 516 activator, acetone (Ghaninia et al., 2019), which like CO2 also activated the A cell but not the B 517 cell ( Figure 8B). The response to acetone was abolished in the Gr3 mutant ( Figure 8B), which 518 suggests that the CO2 receptor can interact with non-CO2 ligands. 519 520 The host-emitted odorant 1-octen-3-ol has been shown to activate the B cell. Both Aedes 521 aegypti and Anopheles gambiae Or8-Orco, which are expressed in the B cell, respond to 1-522 octen-3-ol when expressed in heterologous cells (Bohbot and Dickens, 2009;Lu et al., 2007). 523 We found that firing of the B-cell, but not the A cell, increased in the presence of 1-octen-3-ol 524 ( Figure 8C). The B cell response to 1-octen-3-ol was abolished in the Orco mutant, but not in 525 Gr3 or Ir25a mutants ( Figure  We therefore examined the response of maxillary palp neurons to two volatile amines, hexyl 531 amine and triethyl amine. We found that both amines activated the B cell in wild type, Gr3 532 mutants, and Orco mutants ( Figure 8D-F). Average responses to hexyl amine and triethyl 533 amine were strongly reduced but not abolished in the Ir25a mutant. When we scrutinized the 534 raw data carefully, we noted that responses to both stimuli in Ir25a mutants fell into two clear 535 types of neurons. The majority of Ir25a mutant neurons did not respond to these stimuli at all, 536 but a few responded even more robustly than wild type (Data File 1). 537 538 To determine if there are two different functional types of B neurons, we generated a second 539 independent dataset using these stimuli to examine responses in an additional 17 wild-type 540 (+/+ LVP ) neurons and 23 additional Ir25a mutant neurons. The response to the water control 541 stimulus never exceeded 30 spikes/sec firing frequency in either genotype, and we used this 542 as a threshold to classify neurons as "responders" or "non-responders" (Figure 8H-I). We 543 found that all +/+ LVP neurons responded to triethyl amine, and 16 out of 17 +/+ LVP neurons 544 responded to hexyl amine ( Figure 8H-I). Responses to both stimuli were significantly higher 545 than the water control in wild-type (+/+ LVP ) neurons. In contrast, most neurons in the Ir25a 546 mutant did not respond to either triethyl amine or hexyl amine (78.3% n = 23) and neither 547 stimulus elicited significantly different responses from the water control when taking the entire 548 population of 23 recorded neurons into account. 549 550 We noted that 5 out of 23 neurons (21.7%) showed strong responses to both amines that 551 exceeded the corresponding response in+/+ LVP neurons ( Figure 8G-I). These neurons were 552 considered outliers by a ROUT analysis (Q=1%), consistent with the classification system that 553 we used to categorize neurons as responders or non-responders. Given our discovery of 554 multiple additional IRs and ORs co-expressed along with Or8 in the B cells ( Figure 7Q), we 555 speculate that there are at least two distinct types of B neurons, one that requires Ir25a to 556 respond to amines and the other that does not. We hypothesize that this second type of B 557 neuron expresses the Ir76b co-receptor, which could form functional amine receptors with one 558 or more of the ligand-selective IRs expressed in the B cell. Our findings that the B cell 559 responds to 1-octen-3-ol in an Orco-dependent manner and to triethyl amine and hexyl amine 560 in an Ir25a-dependent manner is consistent with the hypothesis that ORs and IRs are 561 functionally co-expressed in the same neurons and that co-expression enables these cells to 562 respond to ligands that activate both classes of receptors. 563

564
Combinatorial chemosensory receptor co-expression in Aedes aegypti 565 The mismatch between the number of receptors in the Aedes aegypti genome and the number 566 of glomeruli in the antennal lobe is resolved in part by co-expression of multiple odorant 567 receptors in individual olfactory sensory neurons. We found that co-expression is extremely 568 widespread, both between and within OR and IR receptor families, and that the number of 569 receptors expressed in a neuron can vary substantially. While some neurons express only an 570 individual co-receptor and ligand-selective receptor pair, others express "sets" of frequently co-571 expressed receptor subunits. We were surprised to find that a single receptor subunit could be 572 co-expressed with completely different combinations of receptor subunits. The biological 573 significance of this finding remains to be seen and the exact number of receptor groupings will 574 require extensive additional study. 575 576 We found that many commonly co-expressed IRs and ORs belong to mosquito gene family 577 expansions. The Ir41 clade of IRs was among the most common IRs to be co-expressed with 578 ORs. This clade is greatly expanded in Aedes aegypti relative to Drosophila melanogaster, 579 with 16 members compared to only 3 orthologous genes, respectively (Matthews et al., 2018). 580 The Aedes aegypti enhances the ability of these mosquitoes to detect amines present in human 586 odor, although this remains to be tested. Similarly, many of the commonly co-expressed ORs 587 are also members of gene expansions in Aedes aegypti as well as Anopheles gambiae but 588 have no direct orthologues in Drosophila melanogaster (Matthews et al., 2018). Because these 589 mosquito olfactory specialists maintain a very different diet than Drosophila melanogaster, it 590 will be fascinating to examine if the ligands for these receptors are enriched in human body 591 odor. 592 593 Among the OR subunits found to be co-expressed with other ORs, Or4 was notable. The possibility of neuronal co-convergence in antennal lobe glomeruli 608 The mismatch between the number of chemosensory receptors in the genome and the number 609 of glomeruli in the antennal lobe originally pointed to two simple models: co-expression or co-610 convergence. In this study we presented extensive evidence for widespread co-expression in 611 the Aedes aegypti olfactory system. However, our findings also point to the likelihood that co-612 convergence exists in this olfactory system as well. Previous work in Drosophila melanogaster 613 and Mus musculus has shown that the identity of an olfactory neuron is defined by the 614 chemosensory receptor it expresses. However, our snRNA-seq results identify a wide variety 615 of cell types that cannot be defined by expression of a single chemosensory receptor in a 616 given neuron. Rather Aedes aegypti chemosensory cell types are defined by the entire 617 complement of odorant receptors they express. For example, we identified 7 types of Ir64a 618 neurons and 4 types of Or4 neurons each co-expressed with many different types of ORs and 619 IRs. Given this revised view of olfactory sensory neuron types, the number of cell types in the 620 antenna far exceeds the number of glomeruli in the antennal lobe. The most likely solution to 621 this problem is co-convergence of neurons in the antennal lobe, although this remains to be 622 tested directly.

624
What rules govern co-convergence in the mosquito? Do neurons that express the same 625 dominant ligand-selective receptor subunit but also co-express different combinations of 626 ligand-selective receptors project to the same glomerulus or separate glomeruli? These 627 contrasting organizational principles would result in very different models of odor coding. We 628 demonstrate that Aedes aegypti mosquitoes express many receptors in some neurons but only 629 a single receptor in others. The presence of multiple receptors in a given neuron could serve 630 as a mechanism to integrate odorant information in the primary sensory neuron itself rather 631 than at the first synapse in the antennal lobe. Co-convergence of olfactory sensory neurons 632 onto the antennal lobe could allow for the early integration of olfactory cues while still retaining 633 discrete input channels that could be selectively modulated during changes in behavioral state, 634 such as the suppression of host-seeking after a blood meal. Exploring the organization of this 635 sensory system and the downstream circuitry will be essential, including the question how 636 projection neurons encode olfactory information given such extreme diversity in sensory 637 afferent types.

639
Coordinated co-expression between IR, OR, and GR ligand-selective receptors 640 We identified co-expression of co-receptors and ligand-selective receptors belonging to distinct 641 chemosensory families in single neurons in both the antenna and the maxillary palp. This co-642 expression poses a gene regulatory problem for an olfactory neuron. For ORs and IRs to form 643 functional chemosensory receptors, at least one co-receptor and one ligand-selective receptor 644 must be expressed in a cell. We have demonstrated that multiple ORs and IRs are expressed 645 in specific receptor "sets". Thus, the transcriptional landscape in Aedes aegypti olfactory 646 neurons is not only permissive to co-expression, but ensures certain receptors are expressed 647 with others.

649
How might this complex code of chemosensory receptor co-expression be regulated? In 650 vertebrates, an elaborate epigenetic silencing mechanism ensures that each olfactory neuron 651 stochastically expresses only a single allele of a single odorant receptor (Bashkirova and 652 Lomvardas, 2019). In contrast, Drosophila melanogaster is thought to use a more conventional 653 transcription factor code in which the specification of a neuron and the expression of its 654 chemosensory receptor is tightly regulated ( neurons has been reported (Karner et al., 2015). This differs from co-expressed receptors in 667 Aedes aegypti, which are often not closely associated genes within the genome and suggests 668 that other mechanisms of gene regulation must account for the co-expression we observe. A 669 recent study also identified neurons in Anopheles gambiae that co-express Orco and Ir76b, but Interestingly these neurons respond to a number of cues that regulate innate behaviors, such 689 as food odors and pheromones. Perhaps chemosensory receptor co-expression is more 690 conducive to sensory systems that drive innate rather than learned behaviors. 691 692 Maxillary palp chemosensory neurons go beyond a simple A, B, and C organization 693 The maxillary palp is a multi-modal sensory organ that responds to CO2 ( simple cellular organization. We demonstrate through multiplexed RNA in situ hybridization 703 and snRNA-seq that the receptor composition of these neurons is far more complex, and they 704 can be subdivided into many more than three cell types. Consistent with this idea, we found 705 that B cells can be separated into different types based on their physiological response to 706 volatile amines. This is revealed in Ir25a mutant animals, where the response to triethyl amine 707 and hexyl amine is abolished in most B cells, but a subset of neurons retains their responses 708 to this compound. We found that Ir76b is expressed in a subset of the This work was supported in part by grant # UL1 TR000043 from the National Center for 755 Advancing Translational Sciences (NCATS) National Institutes of Health (NIH) Clinical and 756 Translational Science Award (CTSA) program. Funding for this study was provided by Jane 757 Coffin The single sensillum recordings in Figure 8A-E were carried out by M.G. and those in Figure  786 8F-I were carried out by G.C.-V. R.I. supervised M.G. and G.C.-V. and analyzed all of the data 787 in Figure 8  The authors declare no competing interests. 792         receptor expression in selected clusters (see Figure S10A for full dot plot). Circle size 876 represents % of cells in each cluster that express a given gene above a normalized expression 877 threshold of 1 UMI of gene*10,000/total UMI of cell. Scale indicates mean expression within a 878 cluster. All circles representing a mean expression value greater than 20 have the same color. 879 Circles for clusters with below 35% of cells expressing the indicated chemoreceptor gene are 880 not included in plot (See Figure S10B).   receptors that are present in more than 10 cells. To be considered positively expressed within 916 a cell, gene must meet a normalized expression threshold of 1 log(UMI of gene*10,000/total 917 UMI of cell+1). (F-P) Feature plots illustrating normalized expression [log(UMI of gene*10,000 / 918 total UMI of cell +1)] of indicated genes visualized on tSNE plot (see Figure S14). (Q) 919 Summary of chemosensory receptor expression in the maxillary palp based on all 920 experimental data in this study (RNA FISH: fluorescent RNA in situ hybridization). 921  showing For routine strain maintenance, animals were primarily blood-fed on live mice and occasionally 956 on live human volunteers. Newly generated strains were blood-fed on human volunteers until 957 they were established. All experiments except those in Figure  To identify effective sgRNAs, 5 candidate sgRNAs per gene were first injected into separate 970 pools of 500 Liverpool embryos and CRISPR-Cas9-mediated cut rate was evaluated as 971 previously described (Kistler et al., 2015). Either a single sgRNA or 2 sgRNAs with the highest 972 cut rates were then chosen to be injected with donor plasmids to target chemosensory gene 973 loci using homology-directed repair. sgRNAs targeted the respective gene near the stop 974 codon, target sequence with protospacer adjacent motif (PAM) underlined: above the indicated threshold were plotted in Figure 1F,G and are available in Data File 1.

1070
Whole brain fixation and immunostaining 1071 Dissection of adult brains and immunostaining was done as previously described (Matthews et  1072 al., 2019). 6-14 day-old mosquitoes were anesthetized on wet ice. Heads were carefully 1073 removed from the body by pinching at the neck with sharp forceps. Heads were placed in a 1.5 1074 mL tube for fixation with 4% paraformaldehyde, 0.1 M Millonig's Phosphate Buffer (pH 7.4), 1075 0.25% Triton X-100, and nutated for 3 hr. Brains were then dissected out of the head capsule 1076 in ice-cold Ca +2 -, Mg +2 -free phosphate buffered saline (PBS, Lonza 17-517Q) and transferred 1077 to a 24-well plate. All subsequent steps were done on a low-speed orbital shaker. Brains were 1078 washed in PBS containing 0.25% Triton X-100 (PBT) at room temperature 6 times for 15 min. 1079 Brains were permeabilized with PBS, 4% Triton X-100, 2% normal goat serum (Jackson 1080 ImmunoResearch #005-000-121) for ~48 hr (2 nights) at 4°C. Brains were rinsed once and 1081 then washed with PBT at room temperature 6 times for 15 min. Primary antibodies were 1082 diluted in PBS, 0.25% Triton X-100, 2% normal goat serum for ~48 hr (2 nights) at 4°C. Brains 1083 were rinsed once then washed in PBT at room temperature 6 times for 15 min. Secondary 1084 antibodies were diluted in PBS, 0.25% Triton X-100, 2% normal goat serum for ~48 hr (2 1085 nights) at 4°C. Brains were rinsed once then washed in PBT at room temperature 6 times for 1086 15 min. Brains were equilibrated overnight in Vectashield (Vector Laboratories H-1000) and 1087 were mounted in Vectashield. The following primary antibodies were used: anti-Brp/nc82 1088 (mouse; 1:50, Developmental Studies Hybridoma Bank -see below) and/or anti-GFP (rabbit: 1089 1:10,000; Life Technologies A-11122). The secondary antibodies used in all experiments 1090 except Figure S1 and Figure S6 were anti-mouse-Cy5 (1:250; Life Technologies A-10524) and 1091 anti-rabbit-Alexa Fluor 488 (1:500; Life Technologies A-11034). In Figure S1, the secondary 1092 antibody was anti-mouse-Alexa Fluor 488 (1:500; Life Technologies A-11001) and in Figure  1093 S6 Hybridoma-SFM + 0.5% fetal bovine serum (production media) at 3 million viable cells / ml. 1108 The media compartment of the flask contained 350 ml of production media. The bioreactor was 1109 incubated at 37 o C with 7% CO2 for 3 days, at which time the cells and media containing nc82 1110 were harvested. 30 million viable cells from the harvest were re-inoculated back into the cell 1111 compartment in 30 ml fresh production media. The media in the media compartment was 1112 replaced the following day with 650 ml production media. Three days later, the media in the 1113 media compartment was replaced with 1,000 ml production media, with the next harvest 3 1114 days later (7 days after the previous harvest). Cells were harvested weekly and fed bi-weekly 1115 until the desired amount of monoclonal antibody was reached. After the first harvest, each one 1116 contained about 3 mg of monoclonal antibody nc82/ml production media. The harvests to be 1117 purified were pooled, centrifuged at 12,855 x g for 15 min. 6.5 mg / run were loaded onto a 1118 Cytiva (formerly GE Life Sciences) 1 ml HiTrap Protein G HP antibody purification column (Cat 1119 # 29048581) at 1 ml / min. The column was then washed with 0.02 M Sodium Phosphate (pH 1120 7.0) before the monoclonal antibody was eluted with 0.1 M Glycine-HCl (pH 2.7). One ml 1121 fractions were collected and immediately neutralized with 60 ml of 1.0 M Tris-HCl (pH 9.0). The 1122 harvest, flow through and fractions from the peak were run on an a 10% SDS-PAGE (Bio-Rad 1123 Cat # 345-0010) to confirm purity and determine which should be pooled. The pooled fractions 1124 of monoclonal antibody were dialyzed into PBS overnight using dialysis tubing (Spectrum™ 1125 132544) with a 50 kDa MWCO. Another 10% SDS-PAGE was run, and the concentration 1126 determined using the absorbance at 280 using an extinction coefficient of 1.43.

1128
Generation of the IR25a polyclonal antibody 1129 Rabbit polyclonal antibodies were raised against IR25a by Proteintech Group Inc. Antibodies 1130 were raised against a protein fusion of the 67 C-terminal amino acids of IR25a and glutathione 1131 S-transferase. cDNA corresponding to the C-terminal region was inserted into the expression 1132 vector PGEX-4T using primers TTTTGGATCCAAATACCGCAAGAACGTAAAG and 1133 TTTTCTCGAGTTAGAAACGAGATTTAAAGTTG and expressed in bacterial strain BL21. A 1134 purified 31 kDA fusion protein was used to immunize 2 rabbits. Serum was affinity purified to a 1135 final concentration of 450 µg/mL and tested by whole mount antenna immunostaining 1136 comparing +/+ LVP to IR25a BamHI/BamHI . Antibodies from one of the two rabbits were found to 1137 selectively label +/+ LVP antennae, and only this antibody was used in all further studies.

1139
Female Antennal lobe confocal imaging 1140 All brains were imaged using a Zeiss Inverted LSM 880 laser scanning confocal microscope 1141 with a 25x / 0.8 NA immersion-corrected objective unless otherwise noted. Glycerol was used 1142 as the immersion medium to most closely match the refractive index of the mounting medium 1143 Vectashield. Antennal lobes in Figure 1, Figure 2, Figure 6, Figure  oversampled relative to the objective resolution and no differences were noted between 1147 imaging conditions. The laser intensity and gain were adjusted along the Z-axis to account for 1148 a loss of intensity due to depth and care was taken to avoid saturation and ensure that the 1149 deepest glomeruli were visible for segmentation. We note that all confocal imaging was 1150 conducted in a manner that would maximize our ability to visualize the boundaries between 1151 glomeruli and to determine the presence or absence of a given fluorophore in each 1152 glomerulus, and was not intended as a quantitative measure of fluorescence intensity. 3xP3 1153 was used as a promoter to express fluorescent proteins as markers for the knock-ins and 1154 QUAS transgenes used in this study, and care was taken to distinguish expression derived 1155 from the 3XP3 promoter from the expression of the QF2 driver and QUAS effector lines under 1156 investigation. 3xP3 drives expression in the optic lobes, as well as some cells in the dorsal 1157 brain. Neither area overlaps with the antennal lobes. As reported previously (Matthews et al., 1158 2019), we saw no 3xP3-driven expression in the antennal lobes in the reporter lines alone 1159 (data not shown). Representative antennal lobe images presented in the figures were cropped 1160 to remove 3xP3-driven expression elsewhere in the brain.

1162
Male brain confocal imaging 1163 All male brains ( Figure 2M,O,Q,S,U) were imaged using a Zeiss Inverted LSM 880 laser 1164 scanning confocal microscope with a 25x / 0.8 NA immersion-corrected objective. Glycerol was 1165 used as the immersion medium to most closely match the refractive index of the mounting 1166 medium Vectashield. Brains were imaged at 1024 x 1024 pixel resolution in X and Y with 0.5 1167 µm Z-steps for a final voxel size of 0.2372 x 0.2372 x 0.5 µm 3 . The laser intensity and gain 1168 were adjusted along the Z-axis to account for a loss of intensity due to depth and care was 1169 taken to avoid saturation and ensure that the deepest regions of the brain were visible. 1170 Confocal images of the brain were processed in ImageJ/FIJI (NIH).

1175
Glycerol was used as the immersion medium to most closely match the refractive index of the 1176 mounting medium Vectashield. Brains were imaged at 1024 x 1024 pixel resolution in X and Y 1177 with 0.5 µm Z-steps for a final voxel size of 0.2076 x 0.2076 x 0.5 µm 3 . The laser intensity and 1178 gain were adjusted along the Z-axis to account for a loss of intensity due to depth and care 1179 was taken to avoid saturation and ensure that the deepest regions of the subesophageal zone 1180 were visible. Confocal images of the subesophageal zone were processed in ImageJ/FIJI 1181 (NIH). 1182

Antennal lobe glomerulus quantification 1183
Confocal images of the antennal lobes in Figure 1, Figure 2, Figure 6, Figure S2-S8 were 1184 processed in ImageJ/FIJI (NIH). The number of glomeruli was quantified as follows: a single 1185 region of interest (ROI) was manually drawn around each glomerulus at a section 1186 approximately central along the Z-axis. Every glomerulus was outlined and an ROI set was 1187 collected that contained the outlines of all glomeruli. Glomeruli were then separated into two 1188 groups, GFP-positive and GFP-negative glomeruli. A count of each was made to determine the 1189 number of glomeruli labeled by each line as well as the total number of glomeruli. The ROIs 1190 were flattened along the Z-axis to enable representation of the data in two dimensions in 1191 Figure 1, Figure 2, Figure S2-S5, Figure S7. The left antennal lobe in 3 brains was analyzed 1192 for each genotype in Figure 1 except for Gr3, for which the left antennal lobe was analyzed in 1 1193 brain, and both left and right antennal lobes were analyzed in an additional 4 brains in Figure  1194 S6. Although we were able to recognize general regions of the antennal lobe, the 1195 interindividual variability made it impossible to identify most glomeruli by shape alone. We for this discrepancy in coordinate axes (Shankar and McMeniman, 2020), and throughout this 1205 paper we use the same coordinate axes they have implemented. While Shankar and 1206 McMeniman renamed most antennal lobe regions and glomeruli, they chose not to rename the 1207 MD (Medio-Dorsal) cluster of glomeruli comprising MD1, MD2, and MD3 whose sensory input 1208 derives from the maxillary palp. We have observed in our study that the MD glomeruli are 1209 medial, but they are not notably dorsal, and therefore refer to them as Glomerulus 1, 1210 Glomerulus 2, and Glomerulus 3 in this paper for simplicity. While there is utility in naming 1211 glomeruli, we suspect that the Aedes aegypti mosquito antennal lobe atlas will be refined in the 1212 future with the advent of new genetic tools that will unambiguously allow the field to distinguish 1213 and name genetically identifiable glomeruli. We found that the size, shape, and number of 1214 antennal lobe glomeruli in Aedes aegypti was variable from animal to animal. It is possible that 1215 the boundaries between glomeruli are not easily distinguished by synaptic staining and that 1216 specific glomeruli will become identifiable once there are genetic tools available that label 1217 smaller populations of olfactory sensory neurons. The anatomical variability we see is 1218 consistent with both the original map that identified 50 glomeruli , which 1219 divided glomeruli into 3 classes based on their variability in location, as well as a recent study 1220 that looked specifically at the size and shape of glomeruli across animals (Shankar and 1221 McMeniman, 2020) and revised the original map to a count of ~80 glomeruli. Shankar and 1222 McMeniman named and numbered these glomeruli across animals, but they noted that they 1223 were only able to consistently identify 63 glomeruli. This is similar to the ~65 glomeruli we 1224 observed in our work. While there is not yet a clear consensus on the exact number of 1225 antennal lobe glomeruli in Aedes aegypti, the number of chemosensory receptors expressed in 1226 the antenna and maxillary palp is at least twice as large as any of the estimates of glomerulus 1227 number. The variability in antennal lobe structure appears at first to contrast with Drosophila 1228 melanogaster, where each glomerulus can be clearly identified and named. However, we note 1229 that the antennal lobe map in Drosophila melanogaster has been refined with the advent of 1230 new genetic techniques, starting with 35 glomeruli in the original atlas (Stocker et al., 1990), 1231 then modified to 40 glomeruli (Laissue et al., 1999), and further refined in numerous studies 1232 (Couto et al., 2005;Fishilevich and Vosshall, 2005;Tanaka et al., 2012) including a recent 1233 count of 54 (Grabe et al., 2015) and 58 (Task et al., 2021) glomeruli. We have refrained from 1234 naming glomeruli in Aedes aegypti at this time because we believe that a more stereotyped 1235 arrangement will emerge as new genetic lines are generated that allow cell-type-specific 1236 labelling. A recent study in the mosquito Anopheles gambiae using mosquitoes that label Orco-1237 expressing olfactory neurons also noted that the antennal lobe was variable between animals 1238 relative to Drosophila melanogaster (Riabinina et al., 2016). It is therefore possible that 1239 mosquito antennal lobes are more variable than Drosophilids ( vary between animals more than initially appreciated and appears to be determined by both 1244 genetic factors and activity in olfactory sensory neurons during the early life of the animal. In 1245 Drosophila melanogaster, glomerulus size is highly genetically determined and correlates 1246 strongly with the number of olfactory sensory neurons that innervates each glomerulus (Grabe 1247(Grabe et al., 2015. Whether the variability in glomerulus size in the mosquito is due to activity-1248 dependent changes in structure or other factors remains to be seen.

1250
Additional technical notes on expression and projection patterns of chemosensory 1251 receptor knock-in strains 1252 Orco-QF2>QUAS-mCD8:GFP: We noted that the intensity of GFP varies between glomeruli in 1253 this driver line, with some bright and others comparably dim. We speculate that this is due to a 1254 combination of the variability in Orco expression levels in individual neurons and variability in 1255 the density of innervation in individual glomeruli. A large region of the anterior ventral antennal 1256 lobe was previously referred to as the Johnston's organ center and was thought to comprise a 1257 single large glomerulus . In other insect species, Johnston's organ mediates 1258 detection of auditory cues. Consistent with a recent study (Shankar and McMeniman, 2020), 1259 we segmented this region into multiple glomeruli based on anatomical boundaries revealed 1260 with Brp immunofluorescence. Glomeruli in this region are innervated by Orco-expressing 1261 neurons, calling into doubt the original report that these glomeruli process auditory stimuli and 1262 suggesting instead that they serve an olfactory function. In support of this hypothesis, the 1263 analogous area of the Anopheles coluzzii antennal lobe has been shown to receive projections 1264 from Orco-expressing olfactory sensory neurons (Riabinina et al., 2016). We also observed 1265 GFP projections into the subesophageal zone in Orco-QF2>QUAS-mCD8:GFP animals, which 1266 appear to derive from expression in the proboscis, the primary taste organ in insects. This is 1267 consistent with similar expression in Anopheles coluzzii (Riabinina et al., 2016) and functional 1268 data in Anopheles gambiae showing that olfactory responses are detected in this gustatory 1269 organ (Kwon et al., 2006). 1270 1271 Ir25a-QF2>QUAS-mCD8:GFP: The intensity of GFP projections varies between glomeruli in 1272 this driver line, with some bright and other comparably dim, as noted for Orco-QF2. The 1273 brightest glomeruli are primarily medial and anterior. We see the dimmest innervation in the 1274 area previously described as Johnston's organ center as well as in the central antennal lobe.

1275
Labeling was also seen in other areas of the brain, most notably the subesophageal zone and 1276 anterior mechanosensory motor center.

1278
Ir8a-QF2>QUAS-mCD8:GFP: Depending on the brain being analyzed there were either 2 or 3 1279 medial glomeruli labelled in this line. In the cases where there were 3 medial glomeruli, this 1280 third medial glomerulus was innervated by a few large-diameter axons. These were larger and 1281 sparser than the smaller axons that densely innervated most other glomeruli in this line. We 1282 also note that there are 2-3 cell bodies that express GFP located in the cell body rind lateral to 1283 the antennal lobe (rALl). We are unable to definitively describe where these cells project 1284 without genetic reagents that selectively label these cells, but they appear to send bilateral 1285 processes that cross the midline within what appears to be the saddle to innervate the anterior 1286 mechanosensory motor center outside the antennal lobe. All naming is in accordance with the 1287 new insect brain nomenclature presented in Ito et al. (Ito et al., 2014). 1288 1289 Ir76b-QF2>QUAS-mCD8:GFP: In addition to projections to the antennal lobe, this line shows 1290 innervation of the subesophageal zone of the brain. 1291 1292 Gr3-QF2>QUAS-mCD8:GFP: All antennal lobes in this line show innervation of a single 1293 glomerulus (also referred to as "MD1" and here referred to as "Glomerulus 1"; (Ignell et al., 1294 2005; Shankar and McMeniman, 2020). In several brains, we saw a second small medial 1295 glomerulus that derives its innervation from the antenna and is in a small medial cluster of 1296 landmark glomeruli midway down the anterior-posterior axis closest to the center of the brain. 1297 Innervation appears to come from only a few axons. This low and variable reporter expression 1298 is consistent with the low level of expression of Gr3 in the antennal transcriptome (Matthews et 1299(Matthews et al., 2016. Because this line only shows innervation of these 1-2 glomeruli, we analyzed all 1300 glomeruli only in the single brain in Figure 1I, and additionally analyzed 8 more antennal lobes 1301 in 4 brains for the presence or absence of labelling in these two glomeruli. We analyzed both 1302 left and right antennal lobes from 4 brains and found that in 3 of the 4 brains there was a 1303 second glomerulus in one or both antennal lobes ( Figure S6). The presence of the second 1304 glomerulus was not specific within a single animal as we found all variations of presence and 1305 absence of this glomerulus across both antennal lobes in these 4 animals. In some Gr3-1306 QF2>QUAS-mCD8:GFP animals, we detected a small number of processes that extended 1307 beyond the antennal lobe and into the higher brain, although the exact termination site varied. 1308 We never saw CO2-evoked activity in the variable second glomerulus or these projections 1309 outside the antennal lobe. Images in Figure S6 were taken as described above with the 1310 following changes: Secondary antibodies used were anti-mouse-Alexa Fluor 594 (1:500; Life 1311 Technologies A-11005) and anti-Rabbit-Alexa Fluor 488 (1:500; Life Technologies A-11034). 1312 Images were taken using a Zeiss Inverted LSM 880 laser scanning confocal microscope with a 1313 Plan-Apochromat 40x/1.4 Oil DIC objective. Images were taken at 1024 x 1024 in XY to 1314 generate images with a final voxel size of 0.1384 x 0.1384 x 0.5 µm 3 . Images were scored as 1315 containing GFP in one or two glomeruli.

1317
Additional technical notes on expression and projection patterns of Split-QF2 strains 1318 All antennal lobe immunostaining in Figure 2, Figure 6, Figure S7, Figure S8 was carried out 1319 as described above with slight modifications to utilize the 15xQUAS-dTomato-T2A-GCaMP6s 1320 effector line. The same primary antibodies were used because of the structural similarity 1321 between GCaMP6s and GFP. Intrinsic dTomato was detected without antibody amplification, 1322 as it retained fluorescence after fixation and staining. Brp (Cy5), dTomato, and GCaMP6s 1323 (Alexa Fluor 488) were imaged as three separate confocal channels as described above.

1324
Glomeruli labelled by dTomato completely overlapped with those labelled by GCaMP6s 1325 immunofluorescence, so both channels were used during the quantification of positive and 1326 negative glomeruli. dTomato labeling was used to generate sample images. There was no 1327 staining in the antennal lobes of the individual split effector lines crossed to 15xQUAS-1328 dTomato-T2A-GCaMP6s (n=3 per genotype) (Figure 2, Figure S7).

1330
Antennal lobe anterograde dye fill 1331 For images in Figure S1, mosquitoes were anesthetized on wet ice until immobile and then 1332 transferred to a cold dissection dish. A single antenna or maxillary palp was loaded with 1333 Texas-red conjugated dextran (Molecular Probes D3328) diluted 10 mg in 100 μL external 1334 saline (103 mM NaCl, 3 mM KCl, 5 mM 2-[Tris(hydroxymethyl)methyl]-2-aminoethanesulfonic 1335 acid (TES), 1.5 mM CaCl2, 4 mM MgCl2, 26 mM NaHCO3, 1 mM NaH2PO4, 10 mM trehalose, 1336 10 mM glucose, pH 7.3, osmolality adjusted to 275 mOsm/kg). To load the dye a small drop 1337 (approximately 0.5-1 µL) of dye was placed onto the surface of the dish and the animal was 1338 moved such that the intended cut-site on a single antenna or maxillary palp was placed in the 1339 drop of dye. The antenna or maxillary palp was then removed with sharp forceps and a fine 1340 scalpel (F.S.T 10315-12) while it was submerged in the dye. Care was taken to remove the 1341 maxillary palp proximal to the fourth segment, to include all the capitate-peg sensilla, and to 1342 remove the antenna near the base but to leave the antennal pedicel completely intact. The 1343 animal remained immobile on ice with the antenna or maxillary palp submerged and the dye 1344 was loaded for 2-5 min. After this time the animal was placed in a small soup cup with access 1345 to 10% sucrose and returned to standard rearing conditions overnight to give the dye time to 1346 diffuse throughout the neurons and fill the length of the axon. The next morning dissection of 1347 adult brains and immunostaining was carried out as described above.

1349
Antennal lobe 3-D reconstructions 1350 In an attempt to develop a map of the Aedes aegypti antennal lobe, 3 brains from the +/+ LVP 1351 strain were immunolabeled with Brp to identify the boundaries between antennal lobe 1352 glomeruli. The left antennal lobe in each brain was independently reconstructed from confocal 1353 sections taken with a Plan-Apochromat 63x/1.40NA oil immersion objective, at 1024 x 1024 1354 pixel resolution in X and Y with 0.5 µm Z-steps for a final voxel size of either 0.1318 x 0.1318 x 1355 0.5 µm 3 using the software Imaris (Bitplane). Although the area previously termed Johnston's 1356 organ center was considered a single glomerulus in a previous study , we 1357 noted anatomical boundaries in this region, suggesting that it contains multiple glomeruli. This 1358 observation is consistent with recently published work (Shankar and McMeniman, 2020) and 1359 this area was segmented by an individual researcher to generate the final reconstructions. Two 1360 of these are shown in Figure S1. Each glomerulus was manually segmented into an individual 1361 surface using Surpass View. We were consistently able to identify the three glomeruli 1362 innervated by the maxillary palp, previously termed MD1, MD2 and MD3 ) 1363 which we refer to in this study as Glomerulus 1, Glomerulus 2, and Glomerulus 3 (Figure 1, 1364 Figure 6). The overall structure of the antennal lobe varied considerably from animal to animal 1365 and although we were able to identify certain regions and certain landmark glomeruli including 1366 those that are targeted by the maxillary palp, we were unable to assign an unambiguous 1367 identity to every glomerulus, as is possible in Drosophila melanogaster (Couto et al., 2005;1368 Fishilevich and Vosshall, 2005). This variability makes it essentially impossible to identify a 1369 given glomerulus between animals and we therefore have decided to avoid referring to 1370 glomeruli by previous naming schemes, including MD1, MD2, MD3. An authoritative atlas of 1371 the Aedes aegypti antennal lobe awaits genetic reagents that label subpopulations of sensory 1372 neurons that will permit the field to refer to glomeruli by their molecular identity.

1374
Antennal whole mount immunofluorescence 1375 Whole-mount immunostaining of adult antennae was performed as described ( Vanessa Ruta), and rabbit anti-Ir25a (1:50 dilution). Orco monoclonal antibody and Ir25a 1421 polyclonal antibody specificities were verified in Aedes aegypti by staining orco mutant and 1422 We performed two independent snRNA-seq experiments on the antenna to collect a large 1662 number of nuclei for our analysis. The two batches of antenna snRNA-seq data were merged 1663 and split using merge() and SplitObject() functions in Seurat. Split objects were normalized and 1664 selected for highly variable genes independently. To reduce the batch effects from two 1665 samples, we first selected genes for integrating two batches using the 1666 SelectIntegrationFeatures() function in Seurat ( Figure S9F). Two batches were then integrated 1667 using the FindIntegrationAnchors() and IntegrateData() functions. Batch-corrected samples 1668 were then analyzed following the procedures described in the previous section from scaling to 1669 clustering to identify cluster-specific genes.

1671
To classify cells as neurons, we first identified genes that are orthologous to the neuronal 1672 marker genes used in Drosophila melanogaster using pBLAST. Four mosquito genes, 1673 LOC5565901, LOC5570204, LOC5564848, and LOC5570381, are orthologous to the 1674 Drosophila melanogaster neural markers, syt1, elav, CadN, and brp, respectively. We saw that 1675 expression largely overlapped with the olfactory sensory neuron co-receptors Orco, Ir25a, 1676 Ir76b, Ir8a, and Gr3, consistent with the idea that these are neuronal markers. We defined 1677 neural clusters based on the expression of syt1, elav, CadN, and brp, and clusters expressing 1678 at least three neuronal markers in more than 50% of cells in the corresponding cluster were 1679 defined as neural clusters ( Figure S9I, Figure S13H-I). These neural clusters were then 1680 examined for ligand-selective receptor and co-receptor expression.

1682
Antenna heat map: 1683 The normalized expressions of genes in all nuclei were utilized to plot heatmaps using the 1684 ComplexHeatmap package in R. Epithelia-, glia-, and neuron-enriched genes in the Drosophila 1685 melanogaster antenna were considered as references of the corresponding marker genes in 1686 Aedes aegypti. 1687 1688 Antenna tSNE plot: To generate tSNE plots in Figure 5C of all antennal nuclei and antennal 1689 neurons, expression matrices were first log-normalized, selected for highly variable genes, and 1690 scaled. Scaled data were applied to the RunTSNE() with 1 to 50 dimensions. All antenna 1691 nuclei and antennal neurons were clustered using the Louvain algorithm with resolutions 0.5 1692 and 3 respectively ( Figure S9G, Figure 5C).

1694
Antenna dot plot: The dot plot of cluster-enriched chemosensory receptors in Figure 5D was 1695 based on the DotPlot() function in Seurat and customized using the ggplot2 package. The 1696 mean normalized expression and expression percentage of each chemosensory receptor were 1697 extracted by the DotPlot() function. Chemosensory receptors expressed in more than 35% of 1698 nuclei in the corresponding cluster with mean expression values (UMI of gene*10,000 / total 1699 UMI of cell +1) larger than 1 were considered cluster-dominant chemosensory receptors 1700 ( Figure S10A-B). The expression percentages of all dominant chemosensory receptors were 1701 scaled and clustered. For visualizing differences lower-expressed ligand-selective receptor 1702 subunits, circles representing a mean expression value greater than 20 have the same color. 1703 The expression percentage and mean expression of each chemosensory receptor were plotted 1704 using the geom_point() function in ggplot2. The hclust() function was used to cluster genes. 1705 1706 Antenna chord plot: The chord plot of co-expressed chemosensory receptor in Figure 5E  1707 was generated using the chorddiag package in R. Normalized expressions of the top 20 1708 expressed chemosensory receptor were examined for the co-expression in the antenna 1709 neuron population. Receptors that express more than 1 normalization value were considered 1710 as positively expressed. Each expressed chemosensory receptor was iteratively compared to 1711 the expression of the remaining 19 chemosensory receptor in the corresponding nuclei. If more Figure S1 Approximate number of antennal lobe glomeruli per brain hemisphere innervated by the 2157 indicated sensory structure, derived from quantification of the left antennal lobe in 12 brains 2158 presented in Figure 1I-J, Figure     confocal Z-stack of the left antennal lobe shown in Figure 1I with immunofluorescent labeling 2208 of GFP (green) and Brp (synaptic marker, magenta). A single plane is shown every 5 µm in Z 2209 to capture each glomerulus. Scale bar (C-D): 50 µm. Orientation: d=dorsal, m=medial.