{"paper_id":"39d1ee42-ca50-43bf-b0d1-e549876c558f","body_text":"1 \nTranscriptional Interference Gates Monogenic Odorant Receptor Expression in Ants \n \nGiacomo L. Glotzer*1, P. Daniel H. Pastor1, Daniel J. C. Kronauer*1,2,3  \n1. Laboratory of Social Evolution and Behavior, The Rockefeller University, New York, \nNY 10065, USA \n2. Howard Hughes Medical Institute, New York, NY 10065, USA \n3. Lead contact  \n \n*Correspondence \ngglotzer@rockefeller.edu (G.L.G.); dkronauer@rockefeller.edu (D.J.C.K.) \n \n \nSUMMARY  \nCommunication is crucial to social life, and in ants, it is mediated primarily through olfaction. \nAnts have more odorant receptor (OR) genes than any other group of insects, generated \nthrough tandem duplications that produce large genomic arrays of related genes. However, \nhow olfactory sensory neurons produce a single functional OR from these arrays remains \nunclear. In ants, only mRNA from one OR in an array is exported into the cytoplasm, while \nupstream genes are silent and transcripts from downstream genes remain nuclear. Here, we \nshow that non-canonical readthrough transcription in the downstream direction generates \nnon-translated transcripts. We also find that OR promoters are bidirectional, producing \nantisense long non-coding RNAs that appear to suppress the expression of upstream genes. \nFinally, we present evidence that this regulatory architecture is conserved across ants and \nbees, suggesting that this mechanism for functionally monogenic OR expression is \nwidespread in insects with expanded OR repertoires. \n \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 2 \nINTRODUCTION  \nAlmost all organisms rely on chemosensation to sample and interpret cues from the \nenvironment. The ability to perceive different compounds is dictated by a species’ \nchemoreceptor repertoire. Odorant receptors (ORs), which mediate the sense of smell, are a \nmajor class of chemoreceptors. While vertebrate ORs are GPCRs that signal via secondary \nmessengers, insect ORs function as ligand-gated cation channels and form heterotetramers \nof a conserved coreceptor, Orco, and one of many tuning ORs.1  \nEusocial insects, like ants, some wasps, some bees, and termites, rely heavily on \nchemosensation, as they use pheromone communication to coordinate their sophisticated \nsocial and collective behaviors. In ants, this is reflected by expanded OR repertoires.2–4 \nWhile the fruit fly Drosophila melanogaster has roughly 60 ORs, ants possess up to 687 \nORs.5 When Orco function is disrupted, ants become essentially asocial, illustrating the \nimportance of ORs in mediating ant sociality.6,7  \nA longstanding question in sensory biology is how olfactory sensory neurons (OSNs) select \nwhich receptor from their repertoire to express.8–14 In fruit flies and mice, each OSN \ngenerally expresses a single OR, and the axons of OSNs expressing the same receptor \nconverge on shared regions, known as glomeruli, in the primary olfactory centers of the \nbrain.15–19 Fly OSNs rely on a combinatorial code of transcription factors to dictate OR \nexpression and axon guidance,20,21 whereas mouse OSNs exhibit semi-random OR \nchoice.10,22,23 \nOur understanding of how OSNs choose which OR to express in insects other than \nDrosophila remains limited. Importantly, many insects differ from Drosophila in the genomic \norganization of ORs, suggesting that they might use distinct mechanisms for receptor \nchoice. While in Drosophila almost all ORs occur as isolated singletons,24 many ORs in other \ninsects are arranged in tandem arrays, clusters of genes aligned on the same DNA strand \nthat are hotspots for gene duplication.25–34 Ant tandem arrays can be particularly large and \nnumerous. The clonal raider ant, Ooceraea biroi, for example, which has 503 ORs, has over \n40 tandem arrays of 2-82 ORs distributed across 13 of its 14 chromosomes.27  \nWithin insect tandem arrays, coexpression of multiple ORs is common.35–37 Specifically, ant \nOSNs express a primary “chosen” OR alongside additional ORs located downstream in the \nsame tandem array.35 Importantly, only RNA from the chosen OR is exported into the \ncytoplasm where it can be translated into protein, while RNA from downstream ORs remains \nsequestered in the nucleus.35 Furthermore, single nucleus multiomic sequencing (RNA-seq \nand ATAC-seq) of honeybee (Apis mellifera) OSNs revealed that a single active promoter \nwithin a tandem array can drive the coexpression of multiple downstream ORs.37 However, \nbeyond these initial insights, the gene regulatory mechanisms that dictate functionally \nmonogenic receptor expression at insect OR tandem arrays remain unknown.  \nHere, we show that bidirectional transcription from a single promoter enhances selectivity in \nreceptor expression. Non-canonical transcriptional readthrough ensures that downstream \ngenes do not produce transcripts with protein-coding potential. ORs located upstream of the \nchosen OR are transcriptionally inhibited by an antisense long non-coding RNA (lncRNA) \nthat originates from the same promoter as the chosen OR. This mechanism can explain \ntranscriptional patterns at the edge of tandem arrays and in rare cases where ORs within a \ntandem array have been inverted. Collectively, we interpret this extensive transcriptional \nactivity as a regulatory filter that ensures monogenic OR protein production in ants and other \nhymenopteran insects. We propose that this mechanism may facilitate the integration of \nnewly duplicated genes into the olfactory system and therefore contribute to the rapid \nexpansion of OR repertoires. \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 3 \nRESULTS  \nOR Expression Affects the Transcription of Downstream Non-OR Genes  \nIn principle, the mechanism that drives the expression and nuclear sequestration of ORs \ndownstream of the chosen OR could be a function of the local transcriptional landscape or \nencoded in the OR genes themselves. To test these possibilities, we examined the \nexpression of non-OR genes that are either adjacent to, or nested within, OR tandem arrays. \nWe first quantified the expression of genes directly flanking each tandem array using \npublished snRNA-seq data.35 Surprisingly, many of these flanking genes showed enrichment \nin the cluster of cells that expressed ORs in the corresponding tandem array. For non-OR \ngenes downstream of tandem arrays, this enrichment was limited to genes located on the \nsame DNA strand as the array (Figure 1A), whereas genes on the opposite strand exhibited \nno enrichment (Figure 1B). Even non-OR genes downstream of and in line with singleton \nORs were specifically upregulated in cells that had chosen these ORs (Figure S1A and \nS1B), indicating that this phenomenon is not limited to tandem arrays. This expression \npattern suggests that the transcriptional activity that drives downstream OR expression \nextends to non-OR genes. \nWe also found that non-OR genes nested within tandem arrays undergo OR-mediated \ninduction. Chymotrypsin (LOC105276652), a serine protease involved in digestion38,39 that \nhas no known function in insect OSNs, is nested in tandem array T19 between OR U40 and \nthe pseudogene U41. While chymotrypsin is not expressed in most OSNs, it is upregulated \nin the cluster of OSNs that express T19 ORs (Figure 1C). We conducted RNA-FISH staining \nfor chymotrypsin and U34 (Figure 1D), an OR located 29 kbp upstream (Figure 1E). To \nquantify the RNA-FISH data for this and subsequent experiments, we trained a custom \nCellpose 3.0 model40 to segment nuclei (Video S1) and cytoplasm (Video S2), and used \nconsistent thresholds across all experiments to classify cells with nuclear and cytoplasmic \ntranscripts (Methods). Across all OSNs, we found that cells expressing U34 generally also \nexpressed chymotrypsin, whereas cells expressing chymotrypsin did not necessarily express \nU34 (Figure 1F). Because transcription can initiate anywhere along the array, we suspect \nthat these chymotrypsin+/U34- cells had chosen an OR between U34 and U40. Cells that \nhad chosen U34 had strong nuclear U34 and moderate nuclear chymotrypsin signal (Figure \n1G). In contrast, the cytoplasmic intensity in these cells was high for U34 and very low for \nchymotrypsin, consistent with chymotrypsin RNA being sequestered in the nucleus (Figure \n1H). We confirmed that 95% of these cells had chymotrypsin transcripts in the nucleus, but \nonly 16% had transcripts in the cytoplasm, a level consistent with nonspecific background \nfluorescence (Figure 1I). Thus, while T19-expressing OSNs produce chymotrypsin \ntranscripts, these transcripts likely do not yield functional protein. As a positive control, we \nconfirmed that our signal segmentation pipeline detects cytoplasmic Orco transcripts in cells \nthat express U34 (Figure 1J).  \nFinally, we examined the expression of two non-OR genes that border tandem arrays, \nidentified in our analysis of flanking genes (Figure 1A). LOC105282603 and LOC105286072 \nare both uncharacaterized non-OR genes located directly downstream of T45 and T51, \nrespectively (Figure S1C and S1D). Both genes are highly enriched in the cluster of cells \nexpressing from their respective neighboring tandem array (Figure S1E). We found \nLOC105282603 coexpressed in 87% of cells that had chosen an upstream OR, with \ncytoplasmic localization in 6% of cases (Figure S1F). LOC105286072 was coexpressed in \n100% of cells that had chosen an upstream OR, 52% of which also had cytoplasmic \ntranscripts (Figure S1G and S1H). This demonstrates that non-OR genes located \ndownstream and on the same strand as a tandem array are coexpressed in cells expressing \nan OR in that array, and transcripts from these non-OR genes can also undergo nuclear \nsequestration. The mechanism that drives downstream expression at ant OR tandem arrays \nthus extends beyond the boundaries of tandem arrays and operates independently of the \nprotein products encoded by these genes. \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 4 \nAt least two alternative models could explain these results. First, RNA polymerase II \n(RNAPII) might produce a continuous polycistronic transcript spanning the chosen OR and \ndownstream genes. However, previous work using long-read mRNA sequencing and RNA-\nFISH suggests that each gene produces individual transcripts.35 Furthermore, mRNA from \ndownstream genes is abundantly represented in 10X 3’ snRNA-seq data, which relies on \noligo(dT) primers, implying that RNA from each downstream gene is also polyadenylated. An \nalternative possibility is that RNAPII terminates normally, but proximal genes produce \ntranscripts due to leaky regulation. This scenario still does not explain why transcripts from \ndownstream genes remain in the nucleus. The most parsimonious explanation involves a \nnovel form of transcriptional readthrough, where a single RNAPII produces transcripts from \nseveral tandemly arrayed genes. Transcripts from each gene are cleaved and \npolyadenylated, but we suspect only transcripts from the first gene benefit from the addition \nof a 5’ cap, a key signal for nuclear export that is added immediately after transcription \ninitiation.41–43  \nDownstream Genes are Expressed as a Product of Transcriptional Readthrough \nTo test this hypothesis, we investigated whether RNAPII fails to terminate when transcribing \nOR genes, in which case we should be able to detect RNA from intergenic regions. We used \nan existing dataset of rRNA-depleted RNA-sequencing from whole O. biroi pupae,44 which \nincludes nascent and non-polyadenylated transcripts. To compare the sequencing coverage \nof intergenic regions in tandem arrays with that of a comparable sample of paired non-OR \ngenes, we sampled pairs of genes in the genome that were located on the same DNA strand \nwith intergenic distances similar to those of typical OR gene pairs (Figure S2A; Methods). \nThe mean coverage of OR exons was lower than that of non-ORs (Figure S2B), which is \nexpected because ORs are expressed in a small fraction of cells in the whole pupa. We \ntherefore normalized to the mean coverage of the upstream gene’s exons. Intriguingly, we \nfound that OR intergenic regions had significantly higher relative coverage than intergenic \nregions between non-ORs (Figure 2A), consistent with the hypothesis that RNAPII fails to \nterminate transcription upon reaching the polyadenylation signal sequence (PAS) of OR \ngenes.  \nTo validate that intergenic regions are transcribed with their respective coding regions, we \ndesigned FISH probes against T79 that target either the exons of all its genes or its \nintergenic regions (Figure 2B). T79 exhibits the characteristic staircase-like pattern of OR \nexpression, where ORs downstream of the chosen OR are coexpressed (Figure S2C). The \nchoice of T79 as an example array was based on long-read RNA sequencing data that \nsuggested well-defined intergenic regions (Figure S2D). Antennae (n=5) contained 300 ±10 \n(range: 282-308) T79-expressing cells on average, with 99% of these cells labelled by both \nsets of probes (Figure 2C). In segmented nuclei, the T79 exon signal highly correlated with \nthe intergenic signal (R=0.81, p<0.001) and the exon signal colocalized with the brightest \nintergenic signal (Figure 2D), suggesting that they label the same set of molecules.  \nBecause downstream genes are also polyadenylated, we hypothesized that cleavage occurs \nco-transcriptionally but is insufficient to terminate transcription. We stained 9E213, 9E214 \nand the 3.3 kbp intergenic region (Figure 2E). In cells expressing 9E213 as the chosen OR, \nboth the intergenic region and 9E214 were highly coexpressed, and their transcripts \noverlapped spatially and remained in the nucleus (Figure 2F and 2G). This indicates that \ncleavage occurs immediately after the 3’ end of 9E213 and suggests that the intergenic \nregion and 9E214 are part of the same mRNA molecule. The 9E214 and intergenic \ntranscripts colocalized strictly to a single region of the nucleus, where 9E213 signal was \nmaximal, presumably the site of active transcription (Figure 2G). As expected, cells that \nexpressed 9E214 as the chosen OR did not express 9E213 (Figure 2G and S2E).  \nFor genes containing introns, splicing is a critical prerequisite for the nuclear export of \ntranscripts.45–47 Therefore, we sought to determine whether transcripts from downstream \nORs undergo proper splicing. We separately stained for 9E118 exons and introns, as well as \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 5 \n9E129, an OR located 65 kbp upstream (Figure 2H). We segmented the RNA-FISH signal \nas before and quantified the fraction of the nucleus occupied by 9E118 exon probes, which \nlabel nascent and mature transcripts, versus the region labeled by both 9E118 exon and \nintron probes, which label unspliced transcripts. In cells with 9E129 as the chosen OR, the \nexons occupied a larger area than the exon-intron overlap, suggesting that some degree of \nsplicing does occur (Figure 2I). We confirmed visually that exonic and intronic signals do not \nperfectly overlap (Figure 2J). This indicates that transcripts from downstream ORs are \nspliced, and that another feature must be responsible for confining them to the nucleus. \nCollectively, our results support the hypothesis that downstream OR expression is a product \nof a novel form of transcriptional readthrough in which transcripts from each gene are \ncleaved to produce monocistronic, spliced mRNAs. We surmise that, as in A. mellifera, a \nsingle O. biroi promoter can drive the expression of multiple downstream ORs.37 However, \nwe argue against a polycistronic mode of transcription, and given that RNA-FISH of \ncoexpressed ORs in the honeybee produces non-overlapping signal, we suspect the \nmechanism we describe here also occurs in bees. While the function of this non-canonical \nreadthrough remains unknown, we suspect that it serves as a form of transcriptional \ninterference, inhibiting the production of capped transcripts from downstream genes. \nAlthough this mechanism may explain why downstream ORs do not generate protein, an \nadditional process must account for the silencing of ORs located upstream of the active \npromoter. \nAnt OSNs Produce Antisense RNA from OR Tandem Arrays  \nReturning to our previous analysis of flanking non-OR genes, we next looked at the genes \nlocated directly upstream of each tandem array. Intriguingly, we found the exact opposite \npattern compared to downstream genes: Genes that were upstream and on the same strand \nexhibited minimal enrichment (Figure 3A), while genes upstream and on the opposite strand \nwere highly upregulated in their respective clusters (Figure 3B). The same enrichment of \nantisense non-OR genes was present upstream of singleton ORs (Figure S3A and S3B). \nThis pattern raises the question of whether these upstream antisense non-OR genes \nproduce protein in this context.  \nWe therefore investigated the expression of one of these genes using RNA-FISH. Krüppel \nhomolog 1 (Kr-h1; LOC105275104), a transcription factor and effector of juvenile hormone \nsignaling,48 is directly upstream of and antisense to T45 (Figure S4A), and its expression is \nspecifically enriched in T45-expressing cells (Figure S4B). Kr-h1 was coexpressed in 74% of \ncells that had chosen a T45 OR (Figure S4C). Surprisingly, only 5% of these cells exhibited \nKr-h1 transcripts in the cytoplasm (Figure S4C and S4D), a pattern associated with \ndownstream genes. \nAntisense transcripts, a class of lncRNA transcribed from the opposite DNA strand of a \nprotein-coding gene, have emerged as powerful regulators of gene expression in \neukaryotes.49–54 Two recent studies have highlighted the abundance of lncRNAs in insect \nOSNs,55,56 making them promising candidates for OR regulation.57 While annotated lncRNAs \noccur throughout the O. biroi genome, we noticed that OR loci are heavily enriched for \nantisense lncRNAs, with 27% of OR genes overlapping with annotated antisense lncRNAs, \ncompared to only 6% of non-OR genes (Figure S4E). Accordingly, the number of antisense \nlncRNAs nested within tandem arrays scales with the number of ORs in each tandem array \n(Figure S4F). Because lncRNA annotations are often incomplete, we wondered whether \nlncRNAs could in fact be even more ubiquitous in OR tandem arrays. \nTo enhance our coverage of lncRNAs, which may lack a polyA tail,58 we revisited our dataset \nof rRNA-depleted RNA-seq and compared its coverage with that of a control dataset that \nrelied on oligo(dT) primers.44 Compared to polyA-selected data, the rRNA-depleted dataset \nrevealed a pronounced enrichment of reads mapping to the non-coding strand of OR tandem \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 6 \narrays (Figure S4G). An example tandem array illustrating this phenomenon is shown in \nFigure 3C.  \nTo investigate whether the presence of antisense transcripts in the rRNA-depleted data was \nspecific to OR genes, we returned to the same pairs of ORs and non-ORs and quantified the \nrelative coverage of reads that align to the non-coding strand. Antisense transcripts are \ntypically expressed at much lower abundance than sense transcripts,59 and indeed we found \nthat non-OR gene pairs exhibited a ~100-fold reduction in opposite-strand read coverage \n(Figure 3D). However, sense and antisense transcripts were approximately equally abundant \nat OR loci across exons, introns, and intergenic regions (Figure 3D). This indicates that non-\ncoding antisense transcripts are pervasive across OR tandem arrays, and that many of \nthese transcripts are missing in current genome annotations, possibly due to a lack of \npolyadenylation. In fact, all regions antisense to OR genes that currently lack an antisense \nlncRNA annotation have similar relative coverage to existing annotated lncRNAs nested in \ntandem arrays (Figure S4H), suggesting that polyadenylation is not necessary for the \nproduction of these lncRNAs.  \nAntisense lncRNAs Emerge from Bidirectional OR Promoters \nThe high abundance of annotated and unannotated antisense transcripts at OR tandem \narrays provokes the question of how their expression is regulated. To better understand the \ninitiation sites of sense and antisense transcription, we analyzed capped small-RNA \nsequencing (csRNA-seq) data from whole adult ants. csRNA-seq captures transcriptional \nstart sites (TSSs) at single nucleotide resolution and can better detect transient or unstable \nRNA than conventional RNA-seq.60,61 We found twin sense and antisense peaks upstream of \ndozens of ORs, including those with (Figure 4A) and without (Figure 4B) annotated lncRNAs. \nImportantly, the antisense peaks tend to be farther from the OR than the sense ones, \npreventing RNAPII collision (Figure 4C). These upstream peaks are prominent irrespective \nof whether an OR is in a tandem array or not (Figure 4D). Studies have shown that many \npromoters can exhibit bidirectional initiation even if elongation and maturation of transcripts \nis predominantly unidirectional.62 We therefore analyzed the same non-OR pairs from the \nrRNA-depleted RNA-seq analysis and found that, on average, 34% of these genes had \nbidirectional csRNA-seq reads (Figure 4E). Peaks were less common at ORs than non-ORs \n(Figure 4E), likely due to the low relative expression of ORs in whole adult tissue. This \nshows that bidirectional initiation is not limited to ORs, but that ORs stand out in their ability \nto promote antisense elongation and transcript production.  \nAlthough we suspect that antisense transcripts are ubiquitous within ant OR tandem arrays, \nour snRNA-seq analysis only includes reads that map to the 74 annotated lncRNAs nested \nwithin tandem arrays. For each cell expressing these lncRNAs, we plotted the mean \nexpression level of nearby ORs against the genomic distance of these ORs from the \nlncRNA. An example is shown in Figures 4F and 4G, and aggregate data are shown in \nFigures 4H and 4I. This analysis revealed that ORs immediately upstream of lncRNAs are \nexpressed at significantly higher levels than ORs immediately downstream of the lncRNA \n(Figure 4H and 4I), suggesting that these lncRNAs, along with transcripts from the upstream \nOR, originate from a single active promoter region that produces bidirectional transcriptional \nactivity. \nBecause the vast majority of lncRNAs in the O. biroi genome currently appear to be \nunannotated, we aimed to verify their existence and to further investigate their coexpression \nwith upstream ORs. First, we probed for a putative lncRNA in T19 by targeting a 12 kbp \nregion antisense to U31-U33 (Figure 4J). In cells expressing the upstream U21 as the \nchosen OR, signal from the probed region was absent, whereas U34 RNA was present but \nrestricted to the nucleus (Figure 4K). In contrast, cells with the downstream U34 as the \nchosen OR lacked U21 transcripts but coexpressed the probed region (Figure 4L). \nTranscripts from the probed region also remained in the nucleus (Figure 4L), a feature \ncommon to lncRNAs.51 Some cells also expressed U34 alone or the probed region alone \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 7 \n(Figure 4M), which are likely cells with chosen ORs upstream and downstream of U34, \nrespectively.  \nWe then studied T70, a minimal tandem array consisting of only two ORs, Q1 and R2 \n(Figure S5A). Because of its high similarity to R3, a singleton on a different chromosome, we \nwere unable to design probes unique to R2. Of the cells that had chosen Q1, 98% were \nlabeled by R2/3 probes, indicating the presence of R2 transcripts (Figure S5B). In contrast, \nof the cells that had chosen R2/3, 71% coexpressed a putative antisense lncRNA antisense \nto Q1 (Figure S5C). The subset of cells that did not coexpress the putative lncRNA were \nlikely R3-expressing cells (Figure S5D).  \nTo ascertain whether this pattern of lncRNA expression is general to ant ORs, we looked for \nlncRNAs upstream of singleton ORs and confirmed coexpression where lncRNAs are \nannotated (Figure S4E and S4F) and unannotated (Figure S4G-I). Together, these results \nsuggest that every OR gene in the ant genome, including singleton ORs, has a bidirectional \npromoter region that produces an antisense lncRNA in cells where the respective OR is \nexpressed as the chosen OR.  \nLastly, we investigated the length of these lncRNAs. For comparison, we knew that \ntranscription in the sense direction can extend >100 kbp downstream (Figure S5J). Although \nthe annotated lncRNAs can appear short (Figure 4F), we hypothesized that many mapped \nreads may represent transcription initiation much further upstream, as our quantification of \ncoexpression suggests (Figure 4H). Using RNA-FISH, we found that 85% of cells produced \na lncRNA >30 kbp in length (Figure S5K and S5L), and 61% produced a lncRNA >100 kbp in \nlength (Figure S5M and S5N). Our results suggest that bidirectional transcription from a \nsingle OR promoter region can result in polymerase activity >100 kbp in both directions.  \nAntisense lncRNA Expression Negatively Correlates with the Expression of Upstream \nORs  \nWhile these experiments verify the existence of ubiquitous antisense lncRNAs at OR tandem \narrays, their function remains unclear. We hypothesized that lncRNAs inhibit the \ntranscription of ORs upstream of the active promoter, which may be prone to “off-target” \ntranscription due to their spatial proximity. We revisited our snRNA-seq data and identified \nOSNs expressing a lncRNA upstream of their respective chosen OR. For each OSN, we \ncorrelated the expression level of the lncRNA with ORs flanking the chosen OR. For \nexample, when we examined the cells with 9E300 as the chosen OR, we found that the \nexpression of an upstream antisense lncRNA (LOC113562279) was negatively correlated \nwith the expression of the upstream OR and positively correlated with the expression of the \ndownstream OR (Figure 5A and 5B).  \nAggregating these data across cells that express a chosen OR ≤2.5 kbp from a lncRNA, we \nconfirmed that the expression levels of the lncRNA and upstream OR are negatively \ncorrelated (Figure 5C). At these small distances, OSNs exhibit a switch-like behavior, with \n85% of cells showing either upstream OR expression and no lncRNA expression, or vice \nversa (Figure 5C). This correlation weakens in cells where the window is shifted to >2.5 kbp \nand ≤5 kbp (Figure 5D) and subsides entirely once the window increases to >5 kbp and ≤10 \nkbp (Figure 5E). For each of these distances, the correlation of lncRNA expression with \ndownstream OR expression was positive, confirming that the activity of the bidirectional \npromoter region has opposite effects on upstream vs. downstream gene expression (Figure \n5F-H). \nFinally, for each unique lncRNA within a window of 5 kbp from the chosen OR, we examined \nthe correlation of its expression level with that of upstream, chosen, and downstream ORs. \nAveraging across all unique lncRNAs confirmed that lncRNA expression is associated with a \ndecrease in upstream OR expression and an increase in the expression of both the chosen \nOR and any downstream ORs (Figure 5I). This suggests that, as the promoter increases its \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 8 \ntranscriptional activity, the resulting increase in lncRNA expression serves to shut down the \nproduction of upstream transcripts.  \nThese results help clarify why some cells in our heatmaps of tandem array expression show \nnon-zero expression of ORs upstream of the chosen OR. For example, if we isolate all cells \nthat express 9E121 as the chosen OR, we find that the heatmap becomes markedly cleaner \nwhen restricting the sample to cells with detectable coexpression of the upstream antisense \nlncRNA (Figure 5J) because these cells have greatly reduced expression of upstream ORs \n(Figure 5K). \nTo examine whether this pattern of lncRNA expression generalizes to ants and other insects, \nwe analyzed published snRNA-seq data from the antennae of the Indian jumping ant \nHarpegnathos saltator63 (Figure S6A) and the honeybee Apis mellifera37 (Figure S6B). In \nboth species, we identified tandem arrays that exhibit the staircase-like pattern of \ncoexpression described in O. biroi, where the chosen OR is coexpressed with downstream \nORs (Figures S6C and S6D). Publicly available genome annotations from H. saltator64 and \nA. mellifera65 contained twelve and eleven antisense lncRNAs nested within OR tandem \narrays, respectively (Figures S6E and S6F). First, we plotted the coexpression of ORs \nneighboring lncRNAs and found that, as in O. biroi, lncRNA expression was associated with \nthe expression of ORs immediately upstream (Figures S6G-J). We then analyzed the \ncorrelation of lncRNA expression with the expression of upstream, chosen and downstream \nORs. As in O. biroi, lncRNA expression was associated with a decrease in upstream OR \nexpression and an increase in downstream OR expression (Figures S6K and S6L). Taken \ntogether, our results suggest a comprehensive model of how transcriptional activity \ngenerates OR selectivity in hymenopterans and possibly other insects (Figure 6).  \nBidirectional Transcription Ensures Monogenic Expression in the Case of Inversions  \nIf correct, our model should also explain patterns of gene expression in the rare cases of OR \ninversions. While most tandem arrays are composed of genes oriented head-to-tail, we \nidentified a few exceptions in which one or multiple genes were inverted relative to the rest \nof the tandem array. These genes provide a unique opportunity to test our model, particularly \nin cases where the antisense lncRNA spans the coding sequence of an upstream OR.  \nFirst, we examined T51, a tandem array of 16 genes in which 11 genes in the middle of the \narray are flipped in orientation (Figure 7A). 9E89, the last OR in the array, is coexpressed \nwith all other ORs, which is expected because OSNs with a chosen OR in the same \norientation should express it as a downstream OR, while the inverted genes 9E92-9E102 \nshould express it as part of an antisense lncRNA. We co-stained 9E89 and 9E99 (Figure 7B) \nand found 9E89 reliably coexpressed in OSNs with 9E99 as the chosen OR, with nuclear \ntranscript localization (Figure 7C and 7D). Similarly, OSNs with 9E89 as the chosen OR \nshowed nuclear localization of 9E99 transcripts (Figure 7C and 7E). \nFinally, we examined OR 9E198, which is located in tandem array T35 and flipped relative to \nall other ORs in that tandem array (Figure S7A). 9E198 is coexpressed with all ORs in the \ntandem array other than 9E200 and 9E201, the two ORs downstream of 9E198 (Figure \nS7A). We co-stained 9E196, 9E197 and 9E198 (Figure S7B) and observed that cells that \nhad chosen 9E196 coexpressed the upstream gene 9E198 but not 9E197 (Figure S7C). \nSimilarly, cells that had chosen 9E197 coexpressed both the downstream gene 9E196 and \nthe upstream gene 9E198 (Figure S7D). In both cases, the non-chosen OR transcripts \nremained nuclear (Figure S7C and S7D).  \nThese examples of inverted ORs show that ant OSNs reliably prevent ORs neighboring the \nchosen OR from producing protein products, regardless of the relative orientation of these \ngenes. The mechanism of tandem gene duplication typically results in gene copies oriented \nin the same direction on the same strand.33 However, the mechanisms described here \nrobustly ensure monogenic expression even when inversions do occur.  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 9 \nDISCUSSION \nThe common tandem arrangement of chemosensory genes in insects25–34 and \nvertebrates66,67 presents a unique regulatory challenge. The accumulation of transcription \nfactors at any single gene could lead to off-target transcription of nearby genes, and if these \ngenes also produce protein, the tuning of the respective OSN may be affected. Here, we \ndescribe a novel solution to this problem that uses a bidirectional transcriptional interference \nmechanism. Increasing transcriptional activity on both sides of an active promoter in the form \nof downstream readthrough and upstream antisense lncRNAs enhances OR selectivity at \nthe protein level. \nThis study builds on previous work in honeybees37 and mosquitoes36 showing that single \npromoters can drive expression of 2-6 OR genes arranged in tandem. In O. biroi, we extend \nthis idea to tandem arrays containing dozens of genes, and we further propose that a novel \nform of transcriptional readthrough explains the data we observe. In our model, RNAPII \ncontinues transcribing after each cleavage event, producing independent polyadenylated \ntranscripts of which only the first is exported out of the nucleus. We suspect that this \ntranscriptional activity serves as a protective barrier that prevents the activation of \ndownstream promoters. Our model differs from the mechanism governing a tandem array of \nthree ionotropic receptor genes in Drosophila, where RNAPII produces a long polycistronic \nmRNA that spans all three receptors and lacks the first exons of the downstream genes, \nprohibiting their translation.68 While limited transcriptional readthrough (<5 kbp) typically \noccurs in healthy human cells,69 it appears that in ant OSNs, readthrough can extend \nsignificantly further from the active promoter (>100 kbp) to drive expression of non-translated \nOR transcripts. \nThe model we propose relies on an unusual failure of transcriptional termination that \npreserves proper cleavage. Canonically, once the nascent transcript is cleaved, RNAPII \ncontinues transcribing the downstream intergenic sequence until the exonuclease XRN2 \ndegrades the leftover RNA fragment and dislodges RNAPII.70–72 Surprisingly, we find that \neach OR mRNA is cleaved and polyadenylated, yet RNAPII continues transcribing into the \nnext OR gene, indicating a local barrier to termination. This unusual behavior is \nunprecedented at the scale we observe, although an example where a single RNAPII can \nproduce two polyadenylated transcripts has been documented in vertebrates. Here, small \nnucleolar RNAs (snoRNAs) embedded immediately downstream of a PAS enable continued \nRNAPII transcription after cleavage.73 The snoRNAs recruit ribonucleoproteins co-\ntranscriptionally, yielding a ribonucleoprotein cap that shields the freshly cleaved 5’ end from \nXRN2 and allows RNAPII to traverse multiple genes from a single promoter.73 An analogous \nprotection mechanism could underlie OR-specific readthrough in ants, obviating the need to \nglobally modulate XRN2 activity while ensuring that transcripts from downstream genes \nremain sequestered.  \nOR genes in the ant produce two qualitatively distinct mRNA species. The transcripts of \nchosen ORs are plentiful and exported into the cytoplasm, while transcripts produced via \nnon-canonical readthrough are low in abundance and typically colocalize in the nucleus with \nthe brightest signal from the chosen OR. We demonstrate that these downstream transcripts \nare spliced normally, and hypothesize that the distinguishing feature of chosen OR \ntranscripts is the presence of a 5’ cap, a 7-methylguanosine moiety attached to the first \ntranscribed nucleotide.41–43 In the absence of this cap, downstream transcripts would be \nvulnerable to post-transcriptional degradation. Although downstream genes may be \ntranscribed at similar rates to the chosen OR, increased degradation could account for their \nlow steady-state levels.  \nWe further propose that ant ORs possess bidirectional promoter regions that generate \nantisense lncRNAs. Although bidirectional gene pairs are common in insects,74 promoters in \nDrosophila are predominantly unidirectional.75 In the human genome, most promoters are \nstrongly directional when assessing mature transcripts, but many appear bidirectional when \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 10 \nmeasuring nascent transcripts.62 We present evidence that bidirectional initiation is common \nin the ant genome, but antisense elongation is specific to OR promoters.  \nNon-coding transcripts have emerged as important mediators of transcriptional interference \nand gene silencing,76,77 and the abundance of lncRNAs in the Drosophila antenna56 has led \nto speculation on the involvement of lncRNAs in insect OSNs.57 In some cases, lncRNAs \nthemselves are critical for silencing,78 whereas in others, the act of transcription alone is \nsufficient to repress gene expression.79 Transcriptional interference is thought to occur \nthrough several mechanisms, including promoter occlusion and RNAPII collision with \ntranscription factors.79 RNAPII can engage in transcriptional interference irrespective of \nwhether it transcribes in the same68,76,80–83 or opposite84,85 orientation as the gene it is \nsilencing. However, all studies to date on lncRNA-associated interference in tandemly \narrayed genes have focused on gene pairs or small clusters,68,76,86 rather than large tandem \narrays containing dozens of genes. Our analysis suggests that sense and antisense \ntranscripts from a bidirectional promoter region can silence dozens of genes located up to \n>100 kbp upstream or downstream. We suspect that neither the sense nor the antisense \ntranscripts serve a direct functional role beyond transcriptional interference, preventing \ninitiation from nearby promoters.  \nSocial insects rely extensively on chemical communication, and the rapid evolutionary \nturnover of ORs, particularly in ants, is believed to underpin between-species differences in \nbehaviors such as non-nestmate discrimination and prey recognition, as well as various \ncollective activities.6,87–91 Here, we present evidence suggesting that bidirectional \ntranscription from OR promoter regions ensures OR selectivity in dense, gene-rich tandem \narrays. Remarkably, this mechanism is active even at singleton OR genes and OR genes \nthat are inverted within their respective tandem array, showing that it emanates from the \npromoter region itself. This opens the possibility that when ant ORs are duplicated together \nwith this conserved regulatory region, they immediately give rise to a novel type of OSN that \nexclusively produces the respective receptor. This transcription-based mechanism might \nthus be crucial to maintaining monogenic OR selectivity in a clade characterized by frequent \ngene duplication events.  \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 11 \nRESOURCE AVAILABILITY  \nLead Contact  \nFurther information and requests for resources and reagents should be directed to and will \nbe fulfilled by the lead contact, Daniel J. C. Kronauer (dkronauer@rockefeller.edu).  \n \nMaterials Availability  \nAll materials other than ants are commercially available, and ants can be provided upon \nrequest in accordance with federal regulations.  \n \nData and Code Availability  \nConfocal RNA-FISH images are publicly available via the Brain Image Library. Links and \naccession numbers are provided in the Key Resources Table. All code used for image \nsegmentation, alignment, analysis, quantification, and figure generation can be found on \nGitHub (https://github.com/Social-Evolution-and-Behavior/Glotzer-Kronauer-2025).  \n \nACKNOWLEDGEMENTS  \nWe thank Bayley McDonald and Sascha Duttke for acquiring and sharing the csRNA-seq \ndata. Stephany Valdés-Rodríguez, Alek Rahman and Alejandra Hurtado-Giraldo maintained \nstock colonies of ants used for experiments. We thank Kip Lacy for his advice on quantifying \nthe rRNA-sequencing data, Anindita Brahma for her input on the manuscript, Shixin Liu, Bob \nDarnell and Vanessa Ruta for valuable discussions, and other members of the Kronauer lab \nfor their feedback. This is Clonal Raider Ant Project paper number 38.  \n \nFunding Sources  \nThis work was supported by the National Institute of Neurological Disorders and Stroke of \nthe National Institutes of Health under award number R01NS123899 to D.J.C.K. The content \nis solely the responsibility of the authors and does not necessarily represent the official \nviews of the National Institutes of Health. This work was also supported by the Howard \nHughes Medical Institute, where D.J.C.K. is an investigator.  \n \nAUTHOR CONTRIBUTIONS  \nG.L.G. analyzed the sequencing data, designed RNA-FISH probes, conducted RNA-FISH \nexperiments, and wrote code used for segmentation, analysis, and figure generation. \nP.D.H.P. provided input on the planning of experiments, design of RNA-FISH probes, and \ndata analysis. G.L.G. and D.J.C.K. designed the experiments and wrote the paper, and all \nauthors provided feedback on the manuscript. D.J.C.K. supervised the project. \n \nDECLARATION OF INTERESTS  \nThe authors declare no competing interests.  \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 12 \nMAIN FIGURES   \n \nFigure 1. Coexpression of Genes Downstream of Chosen ORs Extends to Non-OR \nGenes  \n(A-B) Expression of non-OR genes downstream of OR tandem arrays on the same strand \n(A) or opposite strand (B) as the ORs. Column n corresponds to the downstream gene of the \ntandem array in row n. Cells are assigned a tandem array based on chosen OR expression, \nand a tandem array’s strand reflects the orientation of most ORs. Dot size corresponds to \npercentage of cells in each group that express a gene at a detectable level (>0) and dot \ncolor reflects the log-normalized expression level. \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 13 \n(C) UMAP of antennal neurons colored by cluster (left), mean expression of T19 ORs \n(middle), and expression of chymotrypsin (right). \n(D) Slice from a confocal z-stack of an ant antennal club stained for U34 (cyan), \nchymotrypsin (yellow) and DAPI (grey). \n(E) Schematic of a subset of T19 highlighting the position of U34 (cyan) and the nested non-\nOR gene chymotrypsin (yellow).  \n(F) Normalized nuclear signal for chymotrypsin vs. U34 in segmented OSN nuclei from n=5 \nantennae. Images with colored borders show the cells labeled with the corresponding colors, \nand each channel is shown individually to the right of each image. Blue: cell with \ncytoplasmic U34 and nuclear chymotrypsin. Orange: cell with cytoplasmic chymotrypsin \nonly. Scale bars: 5 µm.  \n(G-H) Normalized nuclear (G) and cytoplasmic (H) signal for U34 and chymotrypsin in 319 \ncells across n=5 antennae that express U34 as the chosen OR. Each boxplot shows the \nmedian and quartiles; the whiskers extend to 1.5 times the interquartile range.  \n(I-J) Proportion of OSNs with U34 as the chosen OR per antenna (n=5) exhibiting U34 and \nchymotrypsin signal in the nucleus and cytoplasm.  \n(J) Proportion of OSNs with U34 as the chosen OR per antenna (n=5) exhibiting U34 and \nOrco signal in the nucleus and cytoplasm.  \n \nError bars: 95% CI centered on the mean.  \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 14 \n \nFigure 2. Intergenic Regions are Transcribed Along with Downstream Genes \n(A) Relative coverage of intergenic regions for 211 non-OR (magenta) and 211 OR (yellow) \ngene pairs, normalized to the mean coverage of upstream exons. Each boxplot shows the \nmedian and quartiles; the whiskers extend to 1.5 times the interquartile range. The y-axis \nhas a linear region from 0-0.001 and logarithmic scaling (base 10) outside this range. P-\nvalue from Wilcoxon rank-sum test. Dotted line at y=1. \n(B) Schematic of T79 with exon probe-binding regions in magenta and intergenic probe-\nbinding regions in cyan.  \n(C) Number of cells per antenna (n=5) with exonic and intergenic signal versus exon-only \nsignal. P-value from two-sided t-test.  \n(D) Normalized nuclear signal for T79 intergenics vs. T79 exons in segmented OSN nuclei \nfrom n=5 antennae. Images with colored borders show the cells labeled with the \ncorresponding colors, and each channel is shown individually to the right of each image. \nBoth images show example cells with both exonic and intergenic signal.  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 15 \n(E) Schematic of a subset of tandem array T37 with probe-binding regions that target 9E213 \n(cyan), 9E214 (yellow) and the intergenic probed region (magenta).  \n(F) Proportion of OSNs with 9E213 as the chosen OR per antenna (n=7) exhibiting 9E213, \nintergenic PR and 9E214 signal in the nucleus and cytoplasm.  \n(G) Normalized nuclear signal for 9E214 vs. 9E213 in segmented OSN nuclei from n=7 \nantennae. Blue: cell with cytoplasmic 9E213, nuclear 9E214 and nuclear IPR. Orange: cell \nwith cytoplasmic 9E214 only. Green: cell with nuclear 9E213, 9E214 and IPR.  \n(H) Schematic of a subset of tandem array T45 with probe-binding regions that target 9E118 \nexons (magenta), 9E118 introns (yellow) and 9E129 exons (cyan).  \n(I) Proportion of nucleus occupied by 9E118 exon signal alone and the overlap of exon and \nintron signal in cells with 9E129 as the chosen OR.  \n(J) Example cell with cytoplasmic 9E129 and nuclear 9E118. Only some of the 9E118 exon \nsignal overlaps with the 9E118 intron signal. \nError bars: 95% CI centered on the mean. Scale bars: 5µm. \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 16 \n \nFigure 3. Tandem Arrays are Hotspots of Antisense lncRNAs \n(A-B) Expression of non-OR genes upstream of OR tandem arrays on the same strand (A) \nor opposite strand (B) as the ORs. Column n corresponds to the upstream gene of the \ntandem array in row n. Cells are assigned a tandem array based on chosen OR expression, \nand a tandem array’s strand reflects the orientation of most ORs. Dot size corresponds to \npercentage of cells in each group that express a gene at a detectable level (>0) and dot \ncolor reflects the log-normalized expression level.  \n(C) Strand-specific mRNA sequencing coverage (reads per nucleotide) across T79, using \npolyA-enriched RNA-seq (top) and rRNA-depleted RNA-seq (bottom) from whole pupae. \nCoding strand coverage (green) is mirrored across the x-axis for visualization. Gray boxes \nindicate annotated OR gene boundaries.  \n(D) Relative antisense coverage of exons, introns, and intergenic regions for 211 non-OR \n(magenta) and 211 OR (yellow) gene pairs, normalized to the mean coverage of upstream \nexons on the sense strand. Each boxplot shows the median and quartiles; the whiskers \nextend to 1.5 times the interquartile range. The y-axis has a linear region from 0-0.001 and \nlogarithmic scaling (base 10) outside this range. P-values from Wilcoxon rank-sum tests with \nBenjamini–Hochberg FDR correction. Dotted line at y=1. \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 17 \n \nFigure 4. Antisense lncRNAs are Expressed from Bidirectional OR Promoters \n(A-B) Bidirectional peaks from csRNA-seq upstream of L35 (A), an OR in T3 with an \nannotated lncRNA (LOC113562688), and 9E244 (B), an OR in T13. The strand orientation of \neach gene is indicated next to its name. Reads aligning to the sense and antisense \ndirections are displayed in magenta and green, respectively. Gene models include exons \n(dark blue) and introns (grey), with coding sequences (CDS) in a thicker dark blue. \n(C) Histogram of the distance from each csRNA-seq read to the upstream OR CDS for reads \non the same (yellow) and opposite (cyan) strand as the OR. Only reads within 1 kbp \nupstream of an OR’s first CDS are shown.  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 18 \n(D) Proportion of ORs in tandem arrays (magenta) and singletons (grey) with csRNA-seq \nreads within 1 kbp upstream of the first CDS. The orientation of reads is given relative to the \nrespective OR. \n(E) Proportion of 211 non-OR and 211 OR gene pairs with csRNA-seq reads within 1 kbp \nupstream of the TSS for non-ORs and first CDS for ORs. The sample is restricted to gene \npairs arranged in tandem without antisense lncRNA annotations. \n(F) Schematic of T80, containing four ORs and a nested lncRNA (magenta), where the ORs \nare colored by their position relative to the lncRNA, either upstream (yellow) or downstream \n(cyan).  \n(G) Mean log-normalized coexpression of each OR in T80 across nuclei with detectable \nexpression (>0) of lncRNA LOC113562279 vs. TSS-TSS distance from LOC113562279. \n(H) Mean log-normalized coexpression of upstream (yellow) and downstream (cyan) ORs vs. \nthe TSS-TSS distance from lncRNAs. Data include 70 unique lncRNAs embedded within \ntandem arrays and each dot represents a cell in which the corresponding lncRNA is \ndetected.  \n(I) Boxplot of log-normalized OR coexpression within 100 kbp upstream (yellow) or \ndownstream (cyan) of antisense lncRNAs. Each boxplot shows the median and quartiles; the \nwhiskers extend to 1.5 times the interquartile range. P-value from Wilcoxon rank-sum test. \n(J) Schematic of a subset of T19, highlighting ORs U21 (cyan), U34 (yellow), and the probed \nregion (PR) targeting a putative antisense lncRNA (magenta). \n(K-L) Proportion of OSNs with U21 (K) or U34 (L) as the chosen OR per antenna (n=4) \nexhibiting U21, PR and U34 signal in the nucleus and cytoplasm. Error bars: 95% CI \ncentered on the mean. \n(M) Normalized nuclear signal for U34 vs. PR in segmented OSN nuclei from n=4 antennae. \nImages with colored borders reflect the cells labeled with the corresponding colors and each \nchannel is shown individually to the right of each image. Blue: cell with cytoplasmic U34 and \nnuclear PR. Green: only nuclear U34. Orange: only nuclear PR. Scale bars: 5 µm. \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 19 \n \nFigure 5. Antisense lncRNAs Prohibit Expression of Upstream ORs  \n(A) Schematic of a subset of T17 showing three ORs, 9E299 (green), 9E300 (yellow) and \n9E301 (cyan) and the antisense lncRNA LOC113562279 (magenta).  \n(B) Log-normalized expression of 9E301 versus LOC113562279 (left) and E299 \nvs. LOC113562279 (right), in cells where 9E300 is the chosen OR.  \n(C-H) Scatterplots of log-normalized lncRNA versus upstream OR (C-E) or downstream OR \n(F-H) expression, split by the genomic distance from the chosen OR TSS to the 3’ end of the \nnearest upstream lncRNA: 0-2.5 kbp (C, F), 2.5-5 kbp (D, G), and 5-10 kbp (E, H).  \nPearson correlation coefficients and p-values are indicated above each plot (B-H).  \n(I) Mean correlation of each lncRNA and either upstream ORs (cyan), chosen ORs (yellow), \nor downstream ORs (green). The 3’ end of each lncRNA is within ≤5 kb of the chosen OR \nTSS. P-values from one-sample t-tests against zero.  \n(J) Heatmaps of log-normalized expression of all ORs in T45 for cells with 9E121 (*) as the \nchosen OR. Split by absent (top) or detectable (bottom) expression of the antisense lncRNA \nLOC109611203 located 2 kbp upstream of 9E121. Arrows indicate strand orientation. \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 20 \n(K) Mean expression of upstream genes 9E122-129 for all cells in (J) split by absent or \ndetectable expression of the antisense lncRNA LOC109611203. P-value from Wilcoxon \nrank-sum test.  \nError bars: 95% CI centered on the mean.  \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 21 \n \nFigure 6. Non-coding Transcriptional Activity Enhances OR Selectivity at Tandem \nArrays \nSchematic model of transcriptional regulation at a hypothetical OR tandem array. A single \nbidirectional promoter initiates RNAPII activity in both directions. In the coding direction (left \nto right), RNAPII transcribes the chosen OR (yellow), and the transcript is capped at the 5′ \nend. Upon encountering the polyadenylation signal, the nascent transcript is cleaved and \npolyadenylated. RNAPII continues transcription past the 3′ end of the chosen OR into the \nintergenic region and beyond, producing downstream OR transcripts (blue) that are also \npolyadenylated but remain in the nucleus. We suspect that the intergenic region is retained \nin the 5’ UTR of each downstream gene. In the non-coding direction (right to left), RNAPII \ntranscribes an antisense lncRNA (purple) that inhibits transcription of upstream ORs (green). \nMost of these lncRNAs are not polyadenylated. Only the capped transcripts from the chosen \nOR are exported into the cytoplasm and translated into functional protein.  \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 22 \n \nFigure 7. Non-coding Transcription Explains Expression Patterns at OR Gene \nInversions  \n(A) Heatmap of log-normalized expression of all ORs in T51 across cells with a chosen OR \nin T51. Cells (rows) are sorted by the genomic position of their chosen OR. Arrows indicate \nstrand orientation.  \n(B) Schematic of T51, highlighting ORs 9E89 (magenta) and 9E99 (cyan).  \n(C) Normalized nuclear signal for 9E99 and 9E89 in segmented OSN nuclei from n=5 \nantennae. Images with colored borders reflect the cells labeled with the corresponding colors \nand each channel is shown individually to the right of each image. Blue: cell with cytoplasmic \n9E99 and nuclear 9E89. Orange: cell with cytoplasmic 9E89 and nuclear 9E99. Scale bars: \n5 µm. \n(D-E) Proportion of OSNs with 9E99 (D) or 9E89 (E) as the chosen OR per antenna (n=5) \nwith 9E89 and 9E99 signal in the nucleus and cytoplasm. Error bars: 95% CI centered on \nthe mean. \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 23 \nSUPPLEMENTAL FIGURES  \n \nFigure S1. Additional Characterization of Non-OR genes, Related to Figure 1 \n(A-B) Expression of non-OR genes downstream of singleton ORs on the same strand (A) or \nopposite strand (B) as the focal OR. Column n corresponds to the downstream gene of the \nsingleton OR in row n. Each group is composed of cells that express the OR as the chosen \nOR. Dot size corresponds to percentage of cells in each group that express a gene at a \ndetectable level (>0) and dot color reflects the log-normalized expression level. \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 24 \n(C) Schematic of a subset of T45 highlighting 9E116 (cyan) and LOC105282603 (yellow). \n9E116 is located 81 kbp upstream of LOC105282603.  \n(D) Schematic of a subset of T51 highlighting 9E99 (cyan) and LOC105286072 (yellow). \n9E99 is located 51 kbp upstream of LOC105286072.  \n(E) UMAPs of antennal neurons colored by cluster (left), mean expression of T45 ORs \n(second from left), expression of LOC105282603 (middle), mean expression of T51 ORs \n(second from right), expression of LOC105286072 (right). \n(F) Proportion of OSNs with 9E116 as the chosen OR per antenna (n=6) exhibiting 9E116 \nand LOC105282603 signal in the nucleus and cytoplasm.  \n(G) Proportion of OSNs with 9E99 as the chosen OR per antenna (n=4) exhibiting 9E99 and \nLOC105286072 signal in the nucleus and cytoplasm.  \nError bars: 95% CI centered on the mean (F, G).  \n(H) Normalized nuclear signal for LOC105286072 vs. 9E99 in segmented OSN nuclei from \nn=4 antennae. The image with a blue border shows an example cell with nuclear 9E99 and \ncytoplasmic LOC105286072 that is labeled in blue in the plot. Each channel is shown \nindividually to the right. Scale bar: 5 µm.  \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 25 \n \nFigure S2. Additional Analysis of Intergenic Regions, Related to Figure 2  \n(A) Histogram of intergenic distances for 211 non-OR gene pairs (magenta) and 211 OR \ngene pairs (yellow). P-value from Wilcoxon rank-sum test. \n(B) Mean rRNA-depleted RNA-seq coverage across exons of 211 non-OR (magenta) and \n211 OR (yellow) gene pairs. P-value from Wilcoxon rank-sum test. \n(C) Heatmap of log-normalized expression of all ORs in T79 across cells with a chosen OR \nin T79. Cells (rows) are sorted by the genomic position of their chosen OR. Arrows indicate \nstrand orientation. \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 26 \n(D) Alignment of long-read mRNA sequencing to the T79 locus. \n(E) Proportion of OSNs with 9E214 as the chosen OR per antenna (n=5) exhibiting 9E213, \nintergenic PR and 9E214 signal in the nucleus and cytoplasm. Error bars: 95% CI centered \non the mean. \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 27 \n \nFigure S3. Characterization of Non-ORs Upstream of Singleton ORs, Related to Figure \n3 \n(A-B) Expression of non-OR genes upstream of singleton ORs on the same strand (A) or \nopposite strand (B) as the focal OR. Column n corresponds to the upstream gene of the \nsingleton OR in row n. Each group is composed of cells that express the OR as the chosen \nOR. Dot size corresponds to percentage of cells in each group that express a gene at a \ndetectable level (>0) and dot color reflects the log-normalized expression level. \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 28 \n \nFigure S4. RNA from Upstream Antisense Non-OR Genes is Sequestered, Related to \nFigure 3  \n(A) Schematic of T45 highlighting 9E116 (cyan) and Kr-h1 (magenta). 9E116 is located 106 \nkbp upstream of Kr-h1.  \n(B) UMAPs of antennal neurons colored by cluster (left), mean expression of T45 ORs \n(middle), and expression of Kr-h1 (right).  \n(C) Proportion of OSNs with 9E1116 as the chosen OR per antenna (n=5) exhibiting 9E116 \nand Kr-h1 signal in the nucleus and cytoplasm. Error bars: 95% CI centered on the mean.  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 29 \n(D) Normalized nuclear signal for Kr-h1 vs. 9E116 in segmented OSN nuclei from n=5 \nantennae. The image with blue borders shows an example cell with cytoplasmic 9E116 and \nnuclear Kr-h1 that is labeled in blue in the plot. Each channel is shown individually to the \nright. Scale bar: 5 µm.  \n(E) Proportion of OR and non-OR genes that overlap with annotated antisense lncRNAs.  \n(F) Number of antisense lncRNAs nested within each OR tandem array (magenta), and the \nnumber of OR genes per array (grey).  \n(G) Ratio of non-coding to coding strand coverage for tandem arrays with ≥2 OR genes, \nusing rRNA-depleted RNA-seq (magenta) and polyA-enriched RNA-seq (yellow).  \n(H) Relative coverage of OR antisense regions (magenta) and annotated antisense lncRNAs \nnested in tandem arrays (yellow). Antisense coverage is normalized to the coding-strand \ncoverage.  \n(G, H) Each boxplot shows the median and quartiles; the whiskers extend to 1.5 times the \ninterquartile range. P-value from Wilcoxon rank-sum test. Dotted line at y=1.  \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 30 \n \nFigure S5. Additional Staining of lncRNAs, Related to Figure 4  \n(A) Schematic of T70, highlighting Q1 (cyan), R2 (yellow), and the probed region (PR) \ntargeting a putative antisense lncRNA (magenta).  \n(B-C) Proportion of OSNs with Q1 (B) or R2/3 (C) as the chosen OR per antenna (n=6) \nexhibiting Q1, R2/3 and PR signal in the nucleus.  \n(D) Normalized nuclear signal for PR vs. R2/3 in segmented OSN nuclei from n=6 antennae. \nImages with colored borders reflect the cells labeled with the corresponding colors and each \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 31 \nchannel is shown individually to the right of each image. Blue: cell with only cytoplasmic \nR2/3. Orange: cell with cytoplasmic R2/3 and nuclear PR.  \n(E) Schematic of the singleton OR G1 (yellow) and the antisense \nlncRNA LOC113562161 (magenta). G1 is 48 kbp away from the nearest other OR.  \n(F) Proportion of OSNs with G1 as the chosen OR per antenna (n=4) exhibiting G1 and \nLOC113562161 signal in the nucleus.  \n(G) Schematic of 9E88 and a probed region (PR) targeting a putative antisense lncRNA \n(magenta).  \n(H) Proportion of OSNs with 9E88 as the chosen OR per antenna (n=6) exhibiting 9E88 and \nPR signal in the nucleus.  \n(I) Normalized nuclear signal for 9E88 vs. PR in segmented OSN nuclei from n=6 antennae. \nThe blue dot indicates an example cell with cytoplasmic 9E88 and nuclear PR. The cell is \nshown in the blue-ordered image, and each channel is shown individually to the right. \n(J) Mean and standard deviation of OR expression vs. genomic distance from the chosen \nOR TSS using snRNA-seq data.  \n(K) Schematic of a subset of T3, highlighting L16 (yellow) and the probed region (PR) \ntargeting a putative antisense lncRNA (magenta) 30 kbp upstream.  \n(L) Proportion of OSNs with L16 as the chosen OR per antenna (n=4) exhibiting L16 and PR \nsignal in the nucleus.  \n(M) Schematic of a subset of T19, highlighting U54 (yellow) and the probed region (PR) \ntargeting a putative antisense lncRNA (magenta) 103 kbp upstream.  \n(N) Proportion of OSNs with U54 as the chosen OR per antenna (n=4) exhibiting U54 and \nPR signal in the nucleus.  \nError bars: 95% CI centered on the mean. Scale bars: 5 µm.  \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 32 \n \nFigure S6. Bidirectional Promoter Activity in Other Ants and Bees, Related to Figure 5  \n(A-B) Photographs of Harpegnathos saltator (A) and Apis mellifera (B) workers (images by \nAlex Wild). \n(C-D) Representative tandem arrays from H. saltator (C; 30 ORs) and A. mellifera (D; 53 \nORs). Heatmaps of log-normalized expression of all ORs in each tandem array across cells \nwith a chosen OR in the corresponding tandem array. Cells (rows) are sorted by the \ngenomic position of their chosen OR. Arrows indicate strand orientation. \n(E-F) Number of antisense-annotated lncRNAs (magenta) nested within each tandem array \nand the corresponding number of ORs per array (grey) in H. saltator (E) and A. mellifera (F). \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 33 \n(G-H) Mean log-normalized coexpression of upstream (yellow) and downstream (cyan) ORs \nvs. the TSS-TSS distance from lncRNAs, using antennal snRNA-seq data from H. \nsaltator (G) and A. mellifera (H). Each dot represents a cell in which the corresponding \nlncRNA is detected. \n(I-J) Boxplots of log-normalized OR coexpression within 100 kbp upstream (yellow) or \ndownstream (cyan) of nested antisense lncRNAs, using antennal snRNA-seq data from H. \nsaltator (I) and A. mellifera (J). Each boxplot shows the median and quartiles; the whiskers \nextend to 1.5 times the interquartile range. P-values from Wilcoxon rank-sum tests.  \n(K-L) Pearson correlation coefficients for each unique lncRNA and either upstream ORs \n(cyan), chosen ORs (yellow), or downstream ORs (green), using antennal snRNA-seq data \nfrom H. saltator (K) and A. mellifera (L). Each lncRNA has a 3’ end within ≤5 kb of the \nchosen OR TSS. P-values from one-sample t-tests against zero. Error bars: 95% CI \ncentered on the mean. \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 34 \n \nFigure S7. Additional Staining of Inverted OR Genes, Related to Figure 7.  \n(A) Heatmap of log-normalized expression of all ORs in T35 across cells with a chosen OR \nin T35. Arrows indicate strand orientation.  \n(B) Schematic of a subset of T35, highlighting ORs 9E196 (yellow), 9E197 (cyan), \nand 9E198 (magenta). \n(C–D) Proportion of OSNs with 9E196 (C) or 9E197 (D) as the chosen OR per antenna \n(n=4) exhibiting 9E196, 9E197 and 9E198 signal in the nucleus and cytoplasm. Error bars: \n95% CI centered on the mean. \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 35 \nMETHODS  \nKEY RESOURCES TABLE  \nREAGENT or RESOURCE SOURCE IDENTIFIER \nChemicals, peptides, and recombinant proteins \n10% Triton-X 100 Sigma-Aldrich Cat# 93443 \nPBS 1X Corning Cat# MT21040CV \nTween-20 Sigma-Aldrich Cat# P9416 \n16% Paraformaldehyde Electron Microscopy \nScience \nCat# 15710 \nAccuGENE 20X SSC Buffer  Thermo Fisher \nScientific \nCat# 51205 \nSlowFade Glass Thermo Fisher \nScientific \nCat# S36917 \nCritical commercial assays \nHCRv3.0 RNA FISH amplifiers & buffers Molecular \nInstruments  \n \nDeposited data \nConfocal RNA-FISH images This paper https://api.brainima\ngelibrary.org/web/v\niew?bildid=ace-\noak-dig \nO. biroi whole ant csRNA-seq data  This paper zenodo.org/record\ns/15866305 \nO. biroi reference genome v5.4  McKenzie and \nKronauer3 \nPRJNA420369 \nO. biroi reference transcriptome with curated \nRefSeq and GenBank annotations  \nBrahma et al.35 zenodo.org/record\ns/10079884 \nP14 O. biroi single-nucleus RNA-seq reads & \nPacBio Iso-Seq reads \nBrahma et al.35 PRJNA1010363 \nO. biroi whole pupae rRNA-depleted & polyA-\nenriched RNA-seq reads  \nLacy et al.44  PRJNA1075055 \nH. saltator single-nucleus RNA-seq & OR \nannotations \nSieriebriennikov et \nal.63 \nPRJNA987670 \nH. saltator reference genome v8.6 Shields et al.64  PRJNA445978 \nA. mellifera single-nucleus RNA-seq reads Zhang et al.37 PRJNA1041765 \nA. mellifera reference genome v3.1 Wallberg et al.65 PRJNA471592 \nExperimental models: Organisms/strains \nO. biroi clonal line B wild type Kronauer Lab  N/A \nOligonucleotides \nProbes for L16, 9E116, 9E88, 9E129, and Orco,  Molecular \nInstruments \nLOT RTE207 \n(L16), RTE496 \n(9E116), PRM633 \n(9E88), RTE497 \n(9E129), RTC284 \n(Orco) \nCustom RNA-FISH Probes IDT Oligo oPools  Table S1 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 36 \nSoftware and algorithms \nScanpy  Wolf et al.92 https://scanpy.read\nthedocs.io/en/stabl\ne/ \npygenomeviz Yuki Shimoyama https://github.com/\nmoshi4/pyGenome\nViz \nBLAST+ Camacho et al.93 https://blast.ncbi.nl\nm.nih.gov/Blast.cgi \nseaborn Waskom94 https://seaborn.pyd\nata.org \nBiopython Cock et al.95 https://biopython.or\ng \nCellpose 3 Stringer and \nPachitariu40 \nhttps://cellpose.rea\ndthedocs.io/en/late\nst/ \nnapari Chiu et al.96 https://napari.org/s\ntable/ \nnapari-czifile2 Jonas Windhager https://www.napari\n-\nhub.org/plugins/na\npari-czifile2 \nScikit-image  Walt et al.97 https://scikit-\nimage.org \nFIJI  Schindelin et al.98 https://fiji.sc \nCell Counter FIJI Plugin  Curtis Rueden, Mark \nHiner, Tim Wheeler  \nhttps://imagej.net/p\nlugins/cell-counter \nHisat2 Kim et al.99 http://daehwankiml\nab.github.io/hisat2/ \nHOMER Duttke et al.60 http://homer.ucsd.\nedu/homer/ngs/cs\nRNAseq/ \nsamtools Li et al.100 http://www.htslib.or\ng \nScanorama  Hie et al.101 https://scanpy.read\nthedocs.io/en/stabl\ne/generated/scanp\ny.external.pp.scan\norama_integrate.ht\nml \nHarmony  Korsunsky et al.102 https://scanpy.read\nthedocs.io/en/stabl\ne/generated/scanp\ny.external.pp.harm\nony_integrate.html \nIntegrative Genomics Viewer (IGV)  Robinson et al.103 https://igv.org \nOther \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 37 \nAdditional scripts This study https://github.com/\nSocial-Evolution-\nand-\nBehavior/Glotzer-\nKronauer-2025 \n \n \nEXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS  \nAnt Husbandry and Maintenance  \nOoceraea biroi ants were obtained from large stock colonies maintained at 25 °C in \nTupperware containers (32x14 cm) with a plaster of Paris floor. During the brood care phase, \ncolonies were fed three times per week with frozen Solenopsis invicta (fire ant) brood and \ncleaned and watered weekly as needed. Pupae from clonal line B were collected on the day \nof pupation (P0) and housed with adult workers in small Petri dishes also lined with plaster of \nParis. Pupae were aged to 14 days (P14), at which point antennal clubs were dissected.  \n \nMETHOD DETAILS  \nrRNA-depleted and polyA-enriched RNA-seq Data \nTo assess RNA sequencing coverage of OR and non-OR gene pairs, we used existing \ndatasets of rRNA-depleted and polyA-enriched RNA-sequencing (PRJNA1075055) from \nwhole O. biroi pupae.44 We retrieved the curated gene annotations that include RefSeq \n(GCF_003672135.1) and GenBank (GCA_003672135.1) annotations in GTF format from \nZenodo (https://zenodo.org/records/10079884).  \nFor our analysis of gene pairs, we first filtered for genes with a coding sequence length \nbetween 100 bp and 10 kbp. Candidate gene pairs were selected based on an intergenic \ndistance ranging from 50 bp to 10 kbp and were required to be located on the same strand. \nPairs were excluded if either gene or the intergenic region overlapped with any other \nannotated gene on either strand. This filtering yielded 211 OR gene pairs, and we randomly \nsampled an equal number of non-OR gene pairs for comparison.  \nStranded base-pair read coverage was calculated directly from raw BAM files \nusing Samtools.100 For each upstream gene in each pair, we calculated the mean coverage \nacross all annotated exons and introns. For intergenic regions, mean coverage was \ncalculated across the entire intergenic span. To account for expression differences, relative \ncoverage was computed by normalizing to the mean exon coverage of the upstream gene in \neach pair.  \ncsRNA-seq Data  \nTo assess the exact location of transcriptional start sites (TSSs), we analyzed an \nunpublished dataset of capped short RNA-sequencing (csRNA-seq) from whole adult O. \nbiroi ants. The data was acquired by Bayley McDonald, a graduate student in the lab of Dr. \nSascha Duttke at Washington State University, following an established protocol.60,61 Bulk \nextracted RNA was size selected and enriched for 5’-capped RNAs. After amplification, \nstrand-specific paired-end libraries were depleted of rRNA and sequenced on an Illumina \nNextSeq 2000. Sequencing reads were trimmed using HOMER60 and aligned to the O. biroi \nreference genome v5.4 using Hisat2.99 A multimapping threshold of 10 was used to ensure \nunique alignment. The resulting bedGraph files were loaded into Python for subsequent \nanalysis.  \nMany csRNA-seq peaks were within the annotated UTR of an OR gene, indicating that some \nof our gene models have incorrect TSSs. Thus, we assigned csRNA-seq reads to OR genes \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 38 \nby searching the 1 kbp upstream of the first CDS. For the analysis of non-ORs, the same \n211 gene pairs were used as in the rRNA-depleted RNA-seq analysis. This sampling \nensured that the downstream gene TSS was not in close proximity to an existing antisense \ngene annotation. Here, we searched the 1 kbp upstream of the first CDS for ORs and the 1 \nkbp upstream of the annotated TSS for non-ORs.  \nLong-read RNA-seq Data  \nTo produce the image of long-read sequencing coverage of T79 in Figure S2D, we loaded \nthe BAM files of aligned long-read Isoseq reads from P14 antennae (PRJNA1010363)35 into \nIntegrative Genomics Viewer103 and exported an image of the locus containing T79.  \nOoceraea biroi snRNA-seq Data \nWe used a published snRNA-seq dataset from wild-type P14 O. biroi antennae.35 We \nspecifically isolated only the neurons by thresholding for expression of previously-identified \nneuron markers.35 Each nucleus had at least two of the five markers (LOC105284916, \nLOC105280759, LOC105285306, LOC105276401, LOC105275115) with a UMI ≥2.  \nHarpegnathos saltator snRNA-seq Data \nWe analyzed three published snRNA-seq datasets from wild-type adult H. \nsaltator antennae63. Raw gene expression matrices were processed using Scanpy,92 and \ndatasets were integrated with batch correction using Scanorama.101 OR and lncRNA genes \nwere identified using the H. saltator reference genome v8.6.64 OR genes were assigned to \ntandem arrays using cluster definitions from the supplementary materials of Sieriebriennikov \net al.63 \nApis mellifera snRNA-seq Data \nWe analyzed three published snRNA-seq datasets from the antennae of wild-type forager, \nnurse, and newly emerged A. mellifera.37 We integrated the datasets using Harmony.102 OR \ngene annotations were obtained from the supplementary materials of Zhang et al.37 Tandem \narrays were defined as clusters of non-overlapping ORs separated by ≤10 kb. lncRNAs were \nannotated using the A. mellifera reference genome v3.1.65 \nPreprocessing of snRNA-seq Data \nEach dataset was imported into Scanpy92 and gene expression values were log-transformed \nand normalized to a target sum of 10,000 counts per cell. \nQuantification of lncRNAs using snRNA-seq Data  \nTo identify nested lncRNAs, we searched for transcripts overlapping in genomic coordinates \nwith OR tandem arrays. For each nested lncRNA, we annotated the nearest upstream and \ndownstream OR genes within the same array, relative to the lncRNA TSS. We calculated the \ncoexpression of each OR gene in all cells expressing the corresponding lncRNA (>0 counts). \nGenomic distance between each lncRNA and OR gene was defined as the distance \nbetween their TSSs. \nTo assess expression correlation, we identified cells in which the chosen OR (the highest-\nexpressing OR per cell) was located within 5 kbp upstream of an antisense lncRNA. In these \ncases, distance was calculated between the OR TSS and the lncRNA 3’ end.  \nSample Preparation for RNA-FISH  \nLine B ants were aged to 14 days post-pupation (P14). Ants were washed for 30 seconds in \nice-cold 95% ethanol followed by 1xPBS. Antennae were dissected in PBS using \nmicrodissection scissors and transferred to 1.5 mL microcentrifuge tubes containing 1 mL of \n4% paraformaldehyde (PFA) in 1xPBS with 0.5% Triton X-100. The samples were incubated \nin the 4% PFA solution for one hour at room temperature (RT) on a rocker. Antennae were \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 39 \nthen sonicated on a cooling block using a Q700 Sonicator (QSonica, Newtown, CT) fitted \nwith a 1.6mm microtip. Sonication was performed at amplitude 15 with the following \nparameters: 20 cycles of 2 seconds on and 20 seconds off. Following sonication, samples \nwere returned to RT and fixed for an additional hour on a rocker. After fixation, the antennae \nwere washed in 0.1% PBS-Tween (PBST) for 5 minutes and dehydrated using an ice-cold \nmethanol/PBST gradient (25%, 50%, 75%, 100%) for 10 minutes at each step. Samples \nwere then bleached in 3% hydrogen peroxide (H₂O₂) in methanol for 1 hour at 4 °C under \nbright light. Antennae were returned to 100% methanol and stored at -20 °C until staining. \nPrior to staining, samples were rehydrated stepwise using the reverse methanol/PBST \ngradient (75%, 50%, 25%, 0%). \nRNA-FISH Probes  \nFor the genes L16, 9E116, and 9E88, we used probes designed by Molecular Instruments \n(Los Angeles, CA). For all other genes, tandem arrays, intergenic regions, and putative \nlncRNAs, we designed custom RNA-FISH probes compatible with the Molecular Instruments \namplification system. To ensure probe specificity, we used command-line BLASTN93 against \nthe O. biroi transcriptome and excluded any target regions that shared consensus \nsequences with other genes. For intergenic and putative lncRNA probes, we further \nexcluded regions that were not unique to each intergenic and putative lncRNA. We designed \n30 probe pairs against each target, unless the unique regions were not long enough, in \nwhich case we ordered the maximum number of probes that would fit. All probe design code \nis available on GitHub (https://github.com/Social-Evolution-and-Behavior/Glotzer-Kronauer-\n2025), and a complete list of custom probe sequences is provided in Table S1. \nOligonucleotide pools were synthesized by IDT as oPools (50 pmol per oligo) and \nreconstituted in 50 µL of nuclease-free water to generate a 1 µM stock solution.  \nRNA-FISH Staining \nWe followed the HCR v3.0 FISH protocol for chicken embryos as described by Choi et al.,104 \nwith slight modifications. Hybridization was carried out using 4 µL of 1 µM probe stock in \n300 µL of probe hybridization buffer at 37°C for 16 hours. Amplification was performed for 16 \nhours in the dark using 6 µL of each hairpin in 300 µL of amplification buffer, with 1 µL DAPI \nadded. Samples were mounted at room temperature in SlowFade Glass mounting medium \n(Thermo Fisher, Cat# S36917).  \nConfocal Microscopy  \nAntennae were imaged on a Zeiss LSM 900 confocal microscope using 405 nm, 488 nm, \n561 nm, and 633 nm lasers. We used a Zeiss LD LCI Plan-Apochromat 40X / 1.2NA multi-\nimmersion objective lens immersed in glycerol to acquire z-stack images at 0.8x zoom with a \nz-step of 1 µm, capturing slices at 2048x2048 pixel resolution. Laser power was calibrated \nindividually for each stain to avoid underexposure or saturation. For samples stained in \nparallel, identical laser settings were used to enable accurate and consistent quantification of \nRNA-FISH signal. \nCell Segmentation of OSN Nuclei \nConfocal images were loaded using the napari-czifile2 plugin (www.napari-\nhub.org/plugins/napari-czifile2). We trained a custom 2D nuclei segmentation model for \nOSNs by fine-tuning the default Cellpose 3 nuclei model.40 We hand-labeled 24 \nrepresentative z-slices extracted from six images to use as ground truth, with 16 slices used \nfor training and 8 for testing. We specifically only labeled OSNs and did not label support \ncells, which are elongated, or IR-expressing cells, which have larger nuclei. The model was \ntrained for 500 epochs with a learning rate of 0.005 and weight decay of 10-4, using an \nestimated nuclear diameter of 3 µm. We monitored training and validation loss curves and \nconfirmed convergence of the test loss to approximately 0.2. The model was further \nvalidated on held-out images to ensure accurate segmentation. Following validation, the \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 40 \ntrained model was applied to all images, and the resulting nuclear regions of interest (ROIs) \nwere saved for downstream analysis. \nQuantification of Nuclear and Cytoplasmic Intensity \nEach image and its corresponding nuclear ROIs were processed to quantify signal \nintensities. To correct for depth-dependent signal attenuation, we normalized the intensity of \neach non-DAPI channel across z-slices using the mean DAPI intensity as a reference. \nSpecifically, we scaled each slice using a scalar defined as the ratio of the maximum DAPI \nintensity across slices to the DAPI intensity of the current slice. \nBackground subtraction was performed on each z-slice using a Gaussian filter with a sigma \nof 100 and FISH signal was segmented by applying the Triangle threshold from scikit-\nimage.97 The signal mask was further improved by removing small objects (<12 pixels in \narea). For each nuclear ROI, we recorded the mean signal intensity across channels, along \nwith centroid coordinates and area. \nTo quantify cytoplasmic signal, each nuclear ROI was dilated by 3 pixels (~270 nm), \nreflecting the estimated cytoplasmic thickness in OSNs. The cytoplasmic region was defined \nusing a binary subtraction of the original nucleus from the dilated mask, preserving the \ncytoplasmic periphery. Overlapping regions, either between cytoplasmic masks or between \ncytoplasmic and neighboring nuclear masks, were excluded to avoid signal contamination in \ndense regions. For each cytoplasmic ROI, we quantified the mean signal intensity in all \nchannels. \nNormalization of Nuclear and Cytoplasmic Signal  \nTo account for inter-image variability and differences in imaging conditions, nuclear and \ncytoplasmic mean signal intensities were normalized independently for each image and each \nchannel using robust quantile scaling. For each distribution, the lower (0.001) and upper \n(0.999) quantiles of the nuclear signal defined the normalization range, effectively reducing \nthe impact of outliers. The same nuclear quantile range was used to normalize the \ncytoplasmic signal.  \nLabeling of Chosen OR-Expressing Cells \nFor all RNA-FISH experiments, we designated cells as expressing an OR as the chosen OR \nif it expressed an OR with a normalized nuclear signal >0.75 and a normalized cytoplasmic \nsignal >0.2. Additionally, we applied standard size selection, ensuring that these cells had a \nnuclear area between 400-900 pixels (3.2-7.2 µm2) and cytoplasmic area >100 pixels (>0.8 \nµm2). We further selected for circular nuclei by ensuring that the ROIs had an eccentricity \n<0.8. The number of segmented cells in each antenna may be slightly different than the \ncounts provided in this paper because the z-step we used (1 µm) is small enough that the \nsame cell may be counted twice in two different z-planes. However, this does not \nsystematically affect our results, as we are interested in the proportion of cells that \ncoexpress two genes, rather than their raw counts. \nQuantification of Nuclear and Cytoplasmic Colocalization  \nA cell was classified as exhibiting nuclear transcript localization if its normalized nuclear \nsignal was >0.1. Cells were further classified as exhibiting cytoplasmic transcript localization \nif they also had a normalized cytoplasmic signal >0.2. These thresholds were uniformly \napplied across all channels, images, and experimental replicates to ensure consistency.  \nQuantification of Subnuclear RNA-FISH Expression Domains \nRNA-FISH signal was segmented using the same method applied for signal intensity \nmeasurements. To quantify whether probes label the same RNA molecules, we measured \nthe area of each signal domain and the area composed of the overlap between probes in \ndifferent channels.  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 41 \n \nQUANTIFICATION AND STATISTICAL ANALYSES \nAnalyses were performed in Python using standard scientific libraries including NumPy, \nPandas, Seaborn and Matplotlib. Code used for analysis, quantification and plotting is \navailable on GitHub (https://github.com/Social-Evolution-and-Behavior/Glotzer-Kronauer-\n2025). Statistical tests and p-values are noted in the figure legends.  \n \nSUPPLEMENTAL VIDEO AND EXCEL TABLE TITLES AND LEGENDS  \nSupplemental Video 1. Nuclear Segmentation of OSNs  \nExample of nuclei segmentation applied to each confocal z-stack for use in RNA-FISH \nanalysis. Scale bar: 50 µm.  \nSupplemental Video 2. Cytoplasmic Segmentation of OSNs  \nExample of cytoplasmic segmentation applied to each confocal z-stack for use in RNA-FISH \nanalysis. Scale bar: 50 µm.  \nSupplemental Table 1. Custom RNA-FISH Probe Sequences  \nExcel file containing a sheet with the custom probe sequences used for RNA-FISH for each \nprobe set.  \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 42 \nREFERENCES \n1. Benton, R., Sachse, S., Michnick, S.W., and Vosshall, L.B. (2006). Atypical Membrane \nTopology and Heteromeric Function of Drosophila Odorant Receptors In Vivo. PLoS \nBiol. 4, e20. https://doi.org/10.1371/journal.pbio.0040020. \n2. Slone, J.D., Pask, G.M., Ferguson, S.T., Millar, J.G., Berger, S.L., Reinberg, D., Liebig, J., \nRay, A., and Zwiebel, L.J. (2017). Functional characterization of odorant receptors in \nthe ponerine ant, Harpegnathos saltator. Proc. Natl. Acad. Sci. 114, 8586–8591. \nhttps://doi.org/10.1073/pnas.1704647114. \n3. McKenzie, S.K., and Kronauer, D.J.C. (2018). The genomic architecture and molecular \nevolution of ant odorant receptors. Genome Res. 28, 1757–1765. \nhttps://doi.org/10.1101/gr.237123.118. \n4. Zhou, X., Rokas, A., Berger, S.L., Liebig, J., Ray, A., and Zwiebel, L.J. (2015). \nChemoreceptor Evolution in Hymenoptera and Its Implications for the Evolution of \nEusociality. Genome Biol. Evol. 7, 2407–2416. https://doi.org/10.1093/gbe/evv149. \n5. Vizueta, J., Xiong, Z., Ding, G., Larsen, R.S., Ran, H., Gao, Q., Stiller, J., Dai, W., Jiang, \nW., Zhao, J., et al. (2025). Adaptive radiation and social evolution of the ants. Cell 0. \nhttps://doi.org/10.1016/j.cell.2025.05.030. \n6. Trible, W., Olivos-Cisneros, L., McKenzie, S.K., Saragosti, J., Chang, N.-C., Matthews, \nB.J., Oxley, P.R., and Kronauer, D.J.C. (2017). orco Mutagenesis Causes Loss of \nAntennal Lobe Glomeruli and Impaired Social Behavior in Ants. Cell 170, 727-735.e10. \nhttps://doi.org/10.1016/j.cell.2017.07.001. \n7. Yan, H., Opachaloemphan, C., Mancini, G., Yang, H., Gallitto, M., Mlejnek, J., Leibholz, \nA., Haight, K., Ghaninia, M., Huo, L., et al. (2017). An Engineered orco Mutation \nProduces Aberrant Social Behavior and Defective Neural Development in Ants. Cell \n170, 736-747.e9. https://doi.org/10.1016/j.cell.2017.06.051. \n8. Bozza, T., Feinstein, P., Zheng, C., and Mombaerts, P. (2002). Odorant receptor \nexpression defines functional units in the mouse olfactory system. J. Neurosci. 22, \n3033–3043. https://doi.org/10.1523/JNEUROSCI.22-08-03033.2002. \n9. Shykind, B.M., Rohani, S.C., O’Donnell, S., Nemes, A., Mendelsohn, M., Sun, Y., Axel, \nR., and Barnea, G. (2004). Gene Switching and the Stability of Odorant Receptor Gene \nChoice. Cell 117, 801–815. https://doi.org/10.1016/j.cell.2004.05.015. \n10. Lomvardas, S., Barnea, G., Pisapia, D.J., Mendelsohn, M., Kirkland, J., and Axel, R. \n(2006). Interchromosomal Interactions and Olfactory Receptor Choice. Cell 126, 403–\n413. https://doi.org/10.1016/j.cell.2006.06.035. \n11. Serizawa, S., Ishii, T., Nakatani, H., Tsuboi, A., Nagawa, F., Asano, M., Sudo, K., \nSakagami, J., Sakano, H., Ijiri, T., et al. (2000). Mutually exclusive expression of \nodorant receptor transgenes. Nat. Neurosci. 3. https://doi.org/10.1038/76641. \n12. Vosshall, L.B., Amrein, H., Morozov, P.S., Rzhetsky, A., and Axel, R. (1999). A Spatial \nMap of Olfactory Receptor Expression in the Drosophila Antenna. Cell 96, 725–736. \nhttps://doi.org/10.1016/S0092-8674(00)80582-6. \n13. Ray, A., Van Der Goes Van Naters, W., Shiraiwa, T., and Carlson, J.R. (2007). \nMechanisms of Odor Receptor Gene Choice in Drosophila. Neuron 53, 353–369. \nhttps://doi.org/10.1016/j.neuron.2006.12.010. \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 43 \n14. Buck, L., and Axel, R. (1991). A Novel Multigene Family May Encode Odorant \nReceptors: A Molecular Basis for Odor Recognition. Cell 65, 175–187. \nhttps://doi.org/10.1016/0092-8674(91)90418-x. \n15. Mombaerts, P., Wang, F., Dulac, C., Chao, S.K., Nemes, A., Mendelsohn, M., \nEdmondson, J., and Axel, R. (1996). Visualizing an Olfactory Sensory Map. Cell 87, \n675–686. https://doi.org/10.1016/s0092-8674(00)81387-2. \n16. Vosshall, L.B., Wong, A.M., and Axel, R. (2000). An Olfactory Sensory Map in the Fly \nBrain. Cell 102, 147–159. https://doi.org/10.1016/S0092-8674(00)00021-0. \n17. Gao, Q., Yuan, B., and Chess, A. (2000). Convergent projections of Drosophila \nolfactory neurons to specific glomeruli in the antennal lobe. Nat. Neurosci. 3, 780–785. \nhttps://doi.org/10.1038/77680. \n18. Couto, A., Alenius, M., and Dickson, B.J. (2005). Molecular, Anatomical, and Functional \nOrganization of the Drosophila Olfactory System. Curr. Biol. 15, 1535–1547. \nhttps://doi.org/10.1016/j.cub.2005.07.034. \n19. Fishilevich, E., and Vosshall, L.B. (2005). Genetic and Functional Subdivision of the \nDrosophila Antennal Lobe. Curr. Biol. 15, 1548–1553. \nhttps://doi.org/10.1016/j.cub.2005.07.066. \n20. Jafari, S., Alkhori, L., Schleiffer, A., Brochtrup, A., Hummel, T., and Alenius, M. (2012). \nCombinatorial Activation and Repression by Seven Transcription Factors Specify \nDrosophila Odorant Receptor Expression. PLoS Biol. 10, e1001280. \nhttps://doi.org/10.1371/journal.pbio.1001280. \n21. Lyu, C., Li, Z., Xu, C., Wong, K.K.L., Luginbuhl, D.J., McLaughlin, C.N., Xie, Q., Li, T., \nLi, H., and Luo, L. (2025). Dimensionality reduction simplifies synaptic partner matching \nin an olfactory circuit. Science 388, 538–544. https://doi.org/10.1126/science.ads7633. \n22. Pourmorady, A.D., Bashkirova, E.V., Chiariello, A.M., Belagzhal, H., Kodra, A., Duffié, \nR., Kahiapo, J., Monahan, K., Pulupa, J., Schieren, I., et al. (2024). RNA-mediated \nsymmetry breaking enables singular olfactory receptor choice. Nature 625, 181–188. \nhttps://doi.org/10.1038/s41586-023-06845-4. \n23. Markenscoff-Papadimitriou, E., Allen, W.E., Colquitt, B.M., Goh, T., Murphy, K.K., \nMonahan, K., Mosley, C.P., Ahituv, N., and Lomvardas, S. (2014). Enhancer Interaction \nNetworks as a Means for Singular Olfactory Receptor Expression. Cell 159, 543–557. \nhttps://doi.org/10.1016/j.cell.2014.09.033. \n24. Robertson, H.M., Warr, C.G., and Carlson, J.R. (2003). Molecular evolution of the \ninsect chemoreceptor gene superfamily in Drosophila melanogaster. Proc. Natl. Acad. \nSci. 100, 14537–14542. https://doi.org/10.1073/pnas.2335847100. \n25. Dittmann, M.A., Buczkowski, G., Scharf, M., and Harpur, B.A. (2024). Comparative \ntranscriptomics and phylostratigraphy of Argentine ant odorant receptors. PLOS ONE \n19, e0307604. https://doi.org/10.1371/journal.pone.0307604. \n26. Zhou, X., Slone, J.D., Rokas, A., Berger, S.L., Liebig, J., Ray, A., Reinberg, D., and \nZwiebel, L.J. (2012). Phylogenetic and Transcriptomic Analysis of Chemosensory \nReceptors in a Pair of Divergent Ant Species Reveals Sex-Specific Signatures of Odor \nCoding. PLoS Genet. 8, e1002930. https://doi.org/10.1371/journal.pgen.1002930. \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 44 \n27. McKenzie, S.K., Fetter-Pruneda, I., Ruta, V., and Kronauer, D.J.C. (2016). \nTranscriptomics and neuroanatomy of the clonal raider ant implicate an expanded \nclade of odorant receptors in chemical communication. Proc. Natl. Acad. Sci. 113, \n14091–14096. https://doi.org/10.1073/pnas.1610800113. \n28. Robertson, H.M., and Wanner, K.W. (2006). The chemoreceptor superfamily in the \nhoney bee, Apis mellifera: Expansion of the odorant, but not gustatory, receptor family. \nGenome Res. 16, 1395–1403. https://doi.org/10.1101/gr.5057506. \n29. Legan, A.W., Jernigan, C.M., Miller, S.E., Fuchs, M.F., and Sheehan, M.J. (2021). \nExpansion and Accelerated Evolution of 9-Exon Odorant Receptors in Polistes Paper \nWasps. Mol. Biol. Evol. 38, 3832–3846. https://doi.org/10.1093/molbev/msab023. \n30. Obiero, G.F., Pauli, T., Geuverink, E., Veenendaal, R., Niehuis, O., and Große-Wilde, \nE. (2021). Chemoreceptor Diversity in Apoid Wasps and Its Reduction during the \nEvolution of the Pollen-Collecting Lifestyle of Bees (Hymenoptera: Apoidea). Genome \nBiol. Evol. 13, evaa269. https://doi.org/10.1093/gbe/evaa269. \n31. Bohbot, J., Pitts, R.J., Kwon, H. -W., Rützler, M., Robertson, H.M., and Zwiebel, L.J. \n(2007). Molecular characterization of the Aedes aegypti odorant receptor gene family. \nInsect Mol. Biol. 16, 525–537. https://doi.org/10.1111/j.1365-2583.2007.00748.x. \n32. Nei, M., Niimura, Y., and Nozawa, M. (2008). The evolution of animal chemosensory \nreceptor gene repertoires: roles of chance and necessity. Nat. Rev. Genet. 9, 951–963. \nhttps://doi.org/10.1038/nrg2480. \n33. Ramdya, P., and Benton, R. (2010). Evolving olfactory systems on the fly. Trends \nGenet. 26, 307–316. https://doi.org/10.1016/j.tig.2010.04.004. \n34. Engsontia, P., Sanderson, A.P., Cobb, M., Walden, K.K.O., Robertson, H.M., and \nBrown, S. (2008). The red flour beetle’s large nose: An expanded odorant receptor \ngene family in Tribolium castaneum. Insect Biochem. Mol. Biol. 38, 387–397. \nhttps://doi.org/10.1016/j.ibmb.2007.10.005. \n35. Brahma, A., Frank, D.D., Pastor, P.D.H., Piekarski, P.K., Wang, W., Luo, J.-D., Carroll, \nT.S., and Kronauer, D.J.C. (2023). Transcriptional and post-transcriptional control of \nodorant receptor choice in ants. Curr. Biol. 33, 5456-5466.e5. \nhttps://doi.org/10.1016/j.cub.2023.11.025. \n36. Karner, T., Kellner, I., Schultze, A., Breer, H., and Krieger, J. (2015). Co-expression of \nsix tightly clustered odorant receptor genes in the antenna of the malaria mosquito \nAnopheles gambiae. Front. Ecol. Evol. 3. https://doi.org/10.3389/fevo.2015.00026. \n37. Zhang, W., Nie, Y., Xu, T., Li, Y., Xu, Y., Chen, X., Shi, P., Liu, F., Zhao, H., Ma, Q., et \nal. (2025). Evolutionary process underlying receptor gene expansion and cellular \ndivergence of olfactory sensory neurons in honeybees. Mol. Biol. Evol., msaf080. \nhttps://doi.org/10.1093/molbev/msaf080. \n38. Dunse, K.M., Kaas, Q., Guarino, R.F., Barton, P.A., Craik, D.J., and Anderson, M.A. \n(2010). Molecular basis for the resistance of an insect chymotrypsin to a potato type II \nproteinase inhibitor. Proc. Natl. Acad. Sci. 107, 15016–15021. \nhttps://doi.org/10.1073/pnas.1009327107. \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 45 \n39. Hemmati, S.A. (2020). Structural, functional, and phylogenetic studies of chymotrypsin \nenzyme genes in insects: a bioinformatics approach. Plant Prot. Sci. J. Agric. 43, 43–\n60. https://doi.org/10.22055/ppr.2020.16417. \n40. Stringer, C., and Pachitariu, M. (2025). Cellpose3: one-click image restoration for \nimproved cellular segmentation. Nat. Methods 22, 592–599. \nhttps://doi.org/10.1038/s41592-025-02595-5. \n41. Katahira, J. (2015). Nuclear Export of Messenger RNA. Genes 6, 163–184. \nhttps://doi.org/10.3390/genes6020163. \n42. Kachaev, Z.M., Lebedeva, L.A., Kozlov, E.N., and Shidlovskii, Y.V. (2020). Interplay of \nmRNA capping and transcription machineries. Biosci. Rep. 40, BSR20192825. \nhttps://doi.org/10.1042/BSR20192825. \n43. Gonatopoulos-Pournatzis, T., and Cowling, V.H. (2014). Cap-binding complex (CBC). \nBiochem. J. 457, 231–242. https://doi.org/10.1042/BJ20131214. \n44. Lacy, K.D., Lee, J., Rozen-Gagnon, K., Wang, W., Carroll, T.S., and Kronauer, D.J.C. \n(2025). Heterozygosity at a conserved candidate sex determination locus is associated \nwith female development in the clonal raider ant (Ooceraea biroi). eLife 14. \nhttps://doi.org/10.7554/eLife.106913.1. \n45. Zhou, Z., Luo, M., Straesser, K., Katahira, J., Hurt, E., and Reed, R. (2000). The protein \nAly links pre-messenger-RNA splicing to nuclear export in metazoans. Nature 407, \n401–405. https://doi.org/10.1038/35030160. \n46. Khan, M., Hou, S., Chen, M., and Lei, H. (2023). Mechanisms of RNA export and \nnuclear retention. WIREs RNA 14, e1755. https://doi.org/10.1002/wrna.1755. \n47. Luo, M., and Reed, R. (1999). Splicing is required for rapid and efficient mRNA export \nin metazoans. Proc. Natl. Acad. Sci. 96, 14937–14942. \nhttps://doi.org/10.1073/pnas.96.26.14937. \n48. Belles, X. (2020). Krüppel homolog 1 and E93: The doorkeeper and the key to insect \nmetamorphosis. Arch. Insect Biochem. Physiol. 103, e21609. \nhttps://doi.org/10.1002/arch.21609. \n49. Sebastian-delaCruz, M., Gonzalez-Moro, I., Olazagoitia-Garmendia, A., Castellanos-\nRubio, A., and Santin, I. (2021). The Role of lncRNAs in Gene Expression Regulation \nthrough mRNA Stabilization. Non-Coding RNA 7, 3. \nhttps://doi.org/10.3390/ncrna7010003. \n50. Deniz, E., and Erman, B. (2017). Long noncoding RNA (lincRNA), a new paradigm in \ngene expression control. Funct. Integr. Genomics 17, 135–143. \nhttps://doi.org/10.1007/s10142-016-0524-x. \n51. Sun, Q., Hao, Q., and Prasanth, K.V. (2018). Nuclear Long Noncoding RNAs: Key \nRegulators of Gene Expression. Trends Genet. TIG 34, 142–157. \nhttps://doi.org/10.1016/j.tig.2017.11.005. \n52. Pelechano, V., and Steinmetz, L.M. (2013). Gene regulation by antisense transcription. \nNat. Rev. Genet. 14, 880–893. https://doi.org/10.1038/nrg3594. \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 46 \n53. Mattick, J.S., Amaral, P.P., Carninci, P., Carpenter, S., Chang, H.Y., Chen, L.-L., Chen, \nR., Dean, C., Dinger, M.E., Fitzgerald, K.A., et al. (2023). Long non-coding RNAs: \ndefinitions, functions, challenges and recommendations. Nat. Rev. Mol. Cell Biol. 24, \n430–447. https://doi.org/10.1038/s41580-022-00566-8. \n54. Statello, L., Guo, C.-J., Chen, L.-L., and Huarte, M. (2021). Gene regulation by long \nnon-coding RNAs and its biological functions. Nat. Rev. Mol. Cell Biol. 22, 96–118. \nhttps://doi.org/10.1038/s41580-020-00315-9. \n55. Dikmen, F., Dabak, T., Özgişi, B.D., Özenirler, Ç., Kuralay, S.C., Çay, S.B., Çınar, Y.U., \nObut, O., Balcı, M.A., Akbaba, P., et al. (2024). Transcriptome-wide analysis uncovers \nregulatory elements of the antennal transcriptome repertoire of bumblebee at different \nlife stages. Insect Mol. Biol. 33, 571–588. https://doi.org/10.1111/imb.12914. \n56. Talross, G.J.S., and Carlson, J.R. (2023). The rich non-coding RNA landscape of the \nDrosophila antenna. Cell Rep. 42. https://doi.org/10.1016/j.celrep.2023.112482. \n57. Talross, G.J.S., and Carlson, J.R. (2025). New dimensions in the molecular genetics of \ninsect chemoreception. Trends Genet. https://doi.org/10.1016/j.tig.2025.04.003. \n58. Nojima, T., and Proudfoot, N.J. (2022). Mechanisms of lncRNA biogenesis as revealed \nby nascent transcriptomics. Nat. Rev. Mol. Cell Biol. 23, 389–406. \nhttps://doi.org/10.1038/s41580-021-00447-6. \n59. Ozsolak, F., Kapranov, P., Foissac, S., Kim, S.W., Fishilevich, E., Monaghan, A.P., \nJohn, B., and Milos, P.M. (2010). Comprehensive Polyadenylation Site Maps in Yeast \nand Human Reveal Pervasive Alternative Polyadenylation. Cell 143, 1018–1029. \nhttps://doi.org/10.1016/j.cell.2010.11.020. \n60. Duttke, S.H., Chang, M.W., Heinz, S., and Benner, C. (2019). Identification and \ndynamic quantification of regulatory elements using total RNA. Genome Res. \nhttps://doi.org/10.1101/gr.253492.119. \n61. McDonald, B.R., Picard, C.L., Brabb, I.M., Savenkova, M.I., Schmitz, R.J., Jacobsen, \nS.E., and Duttke, S.H. (2024). Enhancers associated with unstable RNAs are rare in \nplants. Nat. Plants 10, 1246–1257. https://doi.org/10.1038/s41477-024-01741-9. \n62. Dudnyk, K., Cai, D., Shi, C., Xu, J., and Zhou, J. (2024). Sequence basis of \ntranscription initiation in the human genome. Science 384, eadj0116. \nhttps://doi.org/10.1126/science.adj0116. \n63. Sieriebriennikov, B., Sieber, K.R., Kolumba, O., Mlejnek, J., Jafari, S., and Yan, H. \n(2024). Orco-dependent survival of odorant receptor neurons in ants. Sci. Adv. 10, \neadk9000. https://doi.org/10.1126/sciadv.adk9000. \n64. Shields, E.J., Sheng, L., Weiner, A.K., Garcia, B.A., and Bonasio, R. (2018). High-\nQuality Genome Assemblies Reveal Long Non-coding RNAs Expressed in Ant Brains. \nCell Rep. 23, 3078–3090. https://doi.org/10.1016/j.celrep.2018.05.014. \n65. Wallberg, A., Bunikis, I., Pettersson, O.V., Mosbech, M.-B., Childers, A.K., Evans, J.D., \nMikheyev, A.S., Robertson, H.M., Robinson, G.E., and Webster, M.T. (2019). A hybrid \nde novo genome assembly of the honeybee, Apis mellifera, with chromosome-length \nscaffolds. BMC Genomics 20, 275. https://doi.org/10.1186/s12864-019-5642-0. \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 47 \n66. Niimura, Y., and Nei, M. (2003). Evolution of olfactory receptor genes in the human \ngenome. Proc. Natl. Acad. Sci. 100, 12235–12240. \nhttps://doi.org/10.1073/pnas.1635157100. \n67. Pan, D., and Zhang, L. (2008). Tandemly Arrayed Genes in Vertebrate Genomes. Int. \nJ. Genomics 2008, 545269. https://doi.org/10.1155/2008/545269. \n68. Mika, K. (2021). Olfactory receptor–dependent receptor repression in Drosophila. Sci. \nAdv. 7. https://doi.org/10.1126/sciadv.abe3745. \n69. Caldas, P., Luz, M., Baseggio, S., Andrade, R., Sobral, D., and Grosso, A.R. (2024). \nTranscription readthrough is prevalent in healthy human tissues and associated with \ninherent genomic features. Commun. Biol. 7, 100. https://doi.org/10.1038/s42003-024-\n05779-5. \n70. Eaton, J.D., and West, S. (2020). Termination of Transcription by RNA Polymerase II: \nBOOM! Trends Genet. 36, 664–675. https://doi.org/10.1016/j.tig.2020.05.008. \n71. Eaton, J.D., and West, S. (2018). An end in sight? Xrn2 and transcriptional termination \nby RNA polymerase II. Transcription 9, 321–326. \nhttps://doi.org/10.1080/21541264.2018.1498708. \n72. Cortazar, M.A., Erickson, B., Fong, N., Pradhan, S.J., Ntini, E., and Bentley, D.L. \n(2022). Xrn2 substrate mapping identifies torpedo loading sites and extensive \npremature termination of RNA pol II transcription. Genes Dev. 36, 1062–1078. \nhttps://doi.org/10.1101/gad.350004.122. \n73. Wu, H., Yin, Q.-F., Luo, Z., Yao, R.-W., Zheng, C.-C., Zhang, J., Xiang, J.-F., Yang, L., \nand Chen, L.-L. (2016). Unusual Processing Generates SPA LncRNAs that Sequester \nMultiple RNA Binding Proteins. Mol. Cell 64, 534–548. \nhttps://doi.org/10.1016/j.molcel.2016.10.007. \n74. Behura, S.K., and Severson, D.W. (2015). Bidirectional promoters of insects: genome-\nwide comparison, evolutionary implication and influence on gene expression. J. Mol. \nBiol. 427, 521–536. https://doi.org/10.1016/j.jmb.2014.11.008. \n75. Core, L.J., Waterfall, J.J., Gilchrist, D.A., Fargo, D.C., Kwak, H., Adelman, K., and Lis, \nJ.T. (2012). Defining the Status of RNA Polymerase at Promoters. Cell Rep. 2, 1025–\n1035. https://doi.org/10.1016/j.celrep.2012.08.034. \n76. Shuman, S. (2020). Transcriptional interference at tandem lncRNA and protein-coding \ngenes: an emerging theme in regulation of cellular nutrient homeostasis. Nucleic Acids \nRes. 48, 8243–8254. https://doi.org/10.1093/nar/gkaa630. \n77. Shearwin, K.E., Callen, B.P., and Egan, J.B. (2005). Transcriptional interference – a \ncrash course. Trends Genet. 21, 339–345. https://doi.org/10.1016/j.tig.2005.04.009. \n78. Nesterova, T.B., Wei, G., Coker, H., Pintacuda, G., Bowness, J.S., Zhang, T., Almeida, \nM., Bloechl, B., Moindrot, B., Carter, E.J., et al. (2019). Systematic allelic analysis \ndefines the interplay of key pathways in X chromosome inactivation. Nat. Commun. 10, \n3129. https://doi.org/10.1038/s41467-019-11171-3. \n79. Palmer, A.C., Egan, J.B., and Shearwin, K.E. (2011). Transcriptional interference by \nRNA polymerase pausing and dislodgement of transcription factors. Transcription 2, 9–\n14. https://doi.org/10.4161/trns.2.1.13511. \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 48 \n80. Gummalla, M., Maeda, R.K., Alvarez, J.J.C., Gyurkovics, H., Singari, S., Edwards, K.A., \nKarch, F., and Bender, W. (2012). abd-A Regulation by the iab-8 Noncoding RNA. \nPLOS Genet. 8, e1002720. https://doi.org/10.1371/journal.pgen.1002720. \n81. Corbin, V., and Maniatis, T. (1989). Role of transcriptional interference in the \nDrosophila melanogaster Adh promoter switch. Nature 337, 279–282. \nhttps://doi.org/10.1038/337279a0. \n82. Petruk, S., Sedkov, Y., Riley, K.M., Hodgson, J., Schweisguth, F., Hirose, S., Jaynes, \nJ.B., Brock, H.W., and Mazo, A. (2006). Transcription of bxd noncoding RNAs \npromoted by trithorax represses Ubx in cis by transcriptional interference. Cell 127, \n1209–1221. https://doi.org/10.1016/j.cell.2006.10.039. \n83. du Mee, D.J.M., Ivanov, M., Parker, J.P., Buratowski, S., and Marquardt, S. (2018). \nEfficient termination of nuclear lncRNA transcription promotes mitochondrial genome \nmaintenance. eLife 7, e31989. https://doi.org/10.7554/elife.31989. \n84. Latos, P.A., Pauler, F.M., Koerner, M.V., Şenergin, H.B., Hudson, Q.J., Stocsits, R.R., \nAllhoff, W., Stricker, S.H., Klement, R.M., Warczok, K.E., et al. (2012). Airn \nTranscriptional Overlap, But Not Its lncRNA Products, Induces Imprinted Igf2r \nSilencing. Science 338, 1469–1472. https://doi.org/10.1126/science.1228110. \n85. Nevers, A., Doyen, A., Malabat, C., Néron, B., Kergrohen, T., Jacquier, A., and Badis, \nG. (2018). Antisense transcriptional interference mediates condition-specific gene \nrepression in budding yeast. Nucleic Acids Res. 46, 6009–6025. \nhttps://doi.org/10.1093/nar/gky342. \n86. Palmer, A.C., Ahlgren-Berg, A., Egan, J.B., Dodd, I.B., and Shearwin, K.E. (2009). \nPotent Transcriptional Interference by Pausing of RNA Polymerases over a \nDownstream Promoter. Mol. Cell 34, 545–555. \nhttps://doi.org/10.1016/j.molcel.2009.04.018. \n87. Ferguson, S.T., Bakis, I., Edwards, N.D., and Zwiebel, L.J. (2023). Olfactory sensitivity \ndifferentiates morphologically distinct worker castes in Camponotus floridanus. BMC \nBiol. 21, 3. https://doi.org/10.1186/s12915-022-01505-x. \n88. d’Ettorre, P., Deisig, N., and Sandoz, J.-C. (2017). Decoding ants’ olfactory system \nsheds light on the evolution of social communication. Proc. Natl. Acad. Sci. 114, 8911–\n8913. https://doi.org/10.1073/pnas.1711075114. \n89. Pask, G.M., Slone, J.D., Millar, J.G., Das, P., Moreira, J.A., Zhou, X., Bello, J., Berger, \nS.L., Bonasio, R., Desplan, C., et al. (2017). Specialized odorant receptors in social \ninsects that detect cuticular hydrocarbon cues and candidate pheromones. Nat. \nCommun. 8, 297. https://doi.org/10.1038/s41467-017-00099-1. \n90. McKenzie, S.K., Winston, M.E., Grewe, F., Vargas Asensio, G., Rodríguez-Hernández, \nN., Rubin, B.E.R., Murillo-Cruz, C., Von Beeren, C., Moreau, C.S., Suen, G., et al. \n(2021). The genomic basis of army ant chemosensory adaptations. Mol. Ecol. 30, \n6627–6641. https://doi.org/10.1111/mec.16198. \n91. Ferguson, S.T., Park, K.Y., Ruff, A.A., Bakis, I., and Zwiebel, L.J. (2020). Odor coding \nof nestmate recognition in the eusocial ant Camponotus floridanus. J. Exp. Biol. 223, \njeb215400. https://doi.org/10.1242/jeb.215400. \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint \n\n 49 \n92. Wolf, F.A., Angerer, P., and Theis, F.J. (2018). SCANPY: large-scale single-cell gene \nexpression data analysis. Genome Biol. 19, 15. https://doi.org/10.1186/s13059-017-\n1382-0. \n93. Camacho, C., Coulouris, G., Avagyan, V., Ma, N., Papadopoulos, J., Bealer, K., and \nMadden, T.L. (2009). BLAST+: architecture and applications. BMC Bioinformatics 10, \n421. https://doi.org/10.1186/1471-2105-10-421. \n94. Waskom, M.L. (2021). seaborn: statistical data visualization. J. Open Source Softw. 6, \n3021. https://doi.org/10.21105/joss.03021. \n95. Cock, P.J.A., Antao, T., Chang, J.T., Chapman, B.A., Cox, C.J., Dalke, A., Friedberg, I., \nHamelryck, T., Kauff, F., Wilczynski, B., et al. (2009). Biopython: freely available \nPython tools for computational molecular biology and bioinformatics. Bioinformatics 25, \n1422–1423. https://doi.org/10.1093/bioinformatics/btp163. \n96. Chiu, C.-L., Clack, N., and the napari community (2022). napari: a Python Multi-\nDimensional Image Viewer Platform for the Research Community. Microsc. Microanal. \n28, 1576–1577. https://doi.org/10.1017/S1431927622006328. \n97. Walt, S. van der, Schönberger, J.L., Nunez-Iglesias, J., Boulogne, F., Warner, J.D., \nYager, N., Gouillart, E., and Yu, T. (2014). scikit-image: image processing in Python. \nPeerJ 2, e453. https://doi.org/10.7717/peerj.453. \n98. Schindelin, J., Arganda-Carreras, I., Frise, E., Kaynig, V., Longair, M., Pietzsch, T., \nPreibisch, S., Rueden, C., Saalfeld, S., Schmid, B., et al. (2012). Fiji: an open-source \nplatform for biological-image analysis. Nat. Methods 9, 676–682. \nhttps://doi.org/10.1038/nmeth.2019. \n99. Kim, D., Paggi, J.M., Park, C., Bennett, C., and Salzberg, S.L. (2019). Graph-based \ngenome alignment and genotyping with HISAT2 and HISAT-genotype. Nat. Biotechnol. \n37, 907–915. https://doi.org/10.1038/s41587-019-0201-4. \n100. Li, H., Handsaker, B., Wysoker, A., Fennell, T., Ruan, J., Homer, N., Marth, G., \nAbecasis, G., Durbin, R., and 1000 Genome Project Data Processing Subgroup (2009). \nThe Sequence Alignment/Map format and SAMtools. Bioinformatics 25, 2078–2079. \nhttps://doi.org/10.1093/bioinformatics/btp352. \n101. Hie, B., Bryson, B., and Berger, B. (2019). Efficient integration of heterogeneous single-\ncell transcriptomes using Scanorama. Nat. Biotechnol. 37, 685–691. \nhttps://doi.org/10.1038/s41587-019-0113-3. \n102. Korsunsky, I., Millard, N., Fan, J., Slowikowski, K., Zhang, F., Wei, K., Baglaenko, Y., \nBrenner, M., Loh, P., and Raychaudhuri, S. (2019). Fast, sensitive and accurate \nintegration of single-cell data with Harmony. Nat. Methods 16, 1289–1296. \nhttps://doi.org/10.1038/s41592-019-0619-0. \n103. Robinson, J.T., Thorvaldsdóttir, H., Winckler, W., Guttman, M., Lander, E.S., Getz, G., \nand Mesirov, J.P. (2011). Integrative genomics viewer. Nat. Biotechnol. 29, 24–26. \nhttps://doi.org/10.1038/nbt.1754. \n104. Choi, H.M.T., Schwarzkopf, M., Fornace, M.E., Acharya, A., Artavanis, G., Stegmaier, \nJ., Cunha, A., and Pierce, N.A. (2018). Third-generation in situ hybridization chain \nreaction: multiplexed, quantitative, sensitive, versatile, robust. Dev. Camb. Engl. 145, \ndev165753. https://doi.org/10.1242/dev.165753. \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted August 21, 2025. ; https://doi.org/10.1101/2025.08.21.671318doi: bioRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}