Fig 1.
The first tarsal segment of the male Drosophila melanogaster foreleg carries multiple functionally and structurally distinct sensory organs.
(A) Anatomy of the Drosophila melanogaster foreleg. The first tarsal segment (ta1), the focal region of this study, is distal to the tibia. Two chordotonal organs (COs) are present outside of the tarsal segments (approximate positions shown in purple). One is situated in the proximal femur (FeCO) and the other in the distal tibia (tCO) [4,5]. (B) The ta1 of the D. melanogaster foreleg is enriched for a range of functionally and structurally diverse sensory organs. This region has the highest concentration of mechanosensory bristles of any part of the leg. Here, mechanosensory bristles are arranged in transverse rows on the ventral side, an arrangement thought to aid in grooming, and longitudinal rows on the anterior, dorsal, and posterior sides [9]. In males, the most distal transverse bristle row is transformed into the sex comb: The mechanosensory bristles, now “teeth,” are modified to be thicker, longer, blunter, and more heavily melanized, while the whole row is rotated 90° [11,12]. Males also show a sex-specific increase in the number of chemosensory taste bristles in ta1, bearing approximately 11 compared to the female’s approximately 7 [16]. Three campaniform sensilla are present in ta1, two on the dorsal distal end of ta1 and one on the proximal ventral side [Ta1GF and Ta1SF, respectively, using the nomenclature of [8]; no campaniform sensilla are present in the distal tibia, ta2, or proximal ta3. (C-E) Campaniform sensilla, mechanosensory bristles, and chemosensory bristles are all composed of modified versions of four core cell types: a socket (or “tormogen”), shaft/dome (or “trichogen”), sheath (or “thecogen”), and neuron [186]. The shaft and socket construct the external apparatus that provides the point of contact for mechanical or chemical stimuli and form a subcuticular lymph cavity that provides the ion source for the receptor current [142,187]. The sheath has glia-like properties, ensheathing the neuron and, as is thought, providing it with protection [187]. Ultimately, however, the contributions of these nonneuronal cells to sensory processing remain poorly characterized [26]. (C) Campaniform sensilla detect strain in the cuticle. They are singly innervated and capped with a dome, rather than a hair-like projection, which extends across the surface of the socket cell [25]. The dendrite tip attaches to the dome cuticle [187]. (D) Mechanosensory bristles detect deflection of the hair-like projection. They are innervated by a single neuron, the dendritic projections of which terminate at the base of the shaft. Specific to this bristle class, the most proximal epithelial cell to the developing sense organ is induced to become a bract cell [12,80]. Bract cells secrete a thick, pigmented, hair-like, cuticular protrusion. (E) The chemosensory taste bristles of the leg differ in their morphology from mechanosensory bristles, appearing less heavily melanized and more curved. They also house a pore at the terminus of the shaft and lack bracts. Each is innervated by a single mechanosensory neuron and 4 gustatory receptor neurons (GRNs) [16]. Figure created using Biorender.com.
Fig 2.
Clustering and iterative subsetting of integrated 24 h and 30 h APF scRNA-seq datasets identifies tarsal cell types with high resolution.
(A-J) UMAP plots showing cell clustering in the 24 h (A-E) and 30 h (F-J) datasets separately. Expression of a series of cluster markers is overlaid on the full 24 h (B-E) and 30 h (G-J) dataset UMAPs. At this resolution, each higher-level cluster in the 24 h dataset has a clear homolog in the 30 h dataset based on a selected subset of marker genes and vice versa. fne for neurons; ct for nonepithelial cells; NimC4 for hemocytes; aos for bracts; Sox100B and repo for different subtypes of glia and axon-associated cells; Su(H) for socket cells; sv for shafts and sheaths. (K) The central UMAP shows the clustering pattern observed in an integrated dataset containing just the epithelial joint and nonjoint cells from both 24 h (gold dots) and 30 h (green dots) samples. Joints are circled with a dashed light gray line, nonjoints with a dashed dark gray line. Reclustering the nonjoint and joint cells gave rise to the 2 flanking UMAP plots. See Fig 3 for details on how the annotations were determined. (L) The number of cells in the postfiltration, doublet-removed epithelial joint (yellow), epithelial nonjoint (gray), and nonepithelial cell (blue) datasets, plotted separately based on which sample (24 h APF or 30 h APF) the cells originated from. Numerical data with cell barcodes are listed in S1 Data. (M) UMAP showing the clustering pattern observed in an integrated dataset containing all nonepithelial cells from both the 24 h (gold dots) and 30 h (green dots) samples. Three major subsets of cells are grouped by colored shapes: neurons, nonsensory cells, and sensory support cells. (N) The number of cells in the postfiltration, doublet-removed sensory support (gold), neuron (navy), and nonsensory (pink) datasets, plotted separately based on which sample the cells originated from (24 h APF or 30 h APF). Numerical data with cell barcodes are listed in S2 Data. (O-Q) UMAPs showing the clustering pattern observed in integrated datasets containing all neurons (O), nonsensory cells (P), and sensory support cells (Q) from both the 24 h (gold dots) and 30 h (green dots) samples. See Figs 4–8 for details on how the annotations were determined. GRN, gustatory receptor neuron; MSNCB, mechanosensory neuron in chemosensory bristle. Data and code for generating the figure are available at https://www.osf.io/ba8tf.
Fig 3.
Single-cell sequencing recovers positional information in the leg epithelium.
(A, B) UMAP plots of the integrated epithelial joint and nonjoint dataset overlaid with the expression of (A) drm (red) and bab2 (blue), and (B) nub (red) and TfAP-2 (blue). The 2 major joint clusters are circled and can be distinguished from one another based on the expression of these 4 genes. bab2 is known to be restricted to the tarsus starting with the distal first tarsal segment [56,57], providing one means through which to distinguish between the tibia and intertarsal joints. (C, D) As in (A) and (B) but with a UMAP plot of just the joint cells. (E) Confocal images showing 24 h pupal legs. On the left is a leg from a TfAP-2-GFP male. Staining is concentrated at the joints, which are each marked with a white triangle. Staining appears stronger at the intertarsal joints compared to the tibia/ta1 joint, consistent with the expression pattern of TfAP-2 in the scRNA-seq data. On the right is a leg from a nub-GAL4 > UAS-GFP.nls male. Staining is concentrated in the distal tibia, proximal to the tibia/ta1 joint (marked with a white arrow). Note that some nonspecific staining from contaminating fat body is present in ta1 and ta2. (F) Joint UMAP with clusters identified through shared nearest neighbor clustering and annotated based on the data presented in (A-E). For details of the high mt. % cluster, see S2A Fig. (G) Dot plot of a selection of top marker genes for each of the joint clusters given in F (excluding the high mt. % cluster). Marker genes were identified by comparing each cluster to the remaining joint clusters. Dot size reflects the number of cells in the cluster in which a transcript for the marker gene was detected, while color represents the expression level. (H, I) Expression of a selection of top marker genes identified in the analysis presented in G overlaid on the joint UMAP plot. (H) sob (red) and Ser (blue). (I) fj (red) and Lim1 (blue). (J-M) UMAP plots of the nonjoint epithelial cells overlaid with markers of spatial identity. Both (J) and (K) show expression of dorsal (red: bi and dpp) and ventral (blue: H15 and wg) markers. (L) shows anterior (blue: ci) and posterior (red: hh) markers. (M) shows proximal (blue: bab2) and distal (red: rn) markers. As is clear from the expression patterns, separation based on spatial markers is apparent for each axis, although stronger for dorsal–ventral and anterior–posterior than proximal–distal. This is likely due to us recovering only a small fraction of the proximal–distal axis by focusing in on just a single tarsal segment. (N) UMAP plot of the nonjoint epithelial cells colored by cluster identity as determined through shared nearest neighbor clustering. Spatial axes are illustrated by arrows based on the expression data presented in (J-M). (O) A dot plot showing the expression of positional markers across each cluster given in (N). Clusters are assigned to regions based on the positional gene expression signature they display. (P) A dot plot of the top markers for each cluster given in (N, O). Marker genes were identified by comparing each cluster to the remaining nonjoint epithelial clusters. Data and code for generating the scRNA-seq elements of this figure are available at https://www.osf.io/ba8tf.
Fig 4.
The tarsus contains several types of nonsensory, nonepithelial cells.
(A) Annotated UMAP plot of nonsensory cells. Cell labels provided in purple indicate populations that are discussed in this figure. Those in black are discussed in Fig 5. (B) Dot plot of the expression of top differentially expressed genes identified through comparisons between each named cluster and all remaining clusters in (A). (C-F) The nonsensory UMAP shown in (A) overlaid with expression of key marker genes for each cluster. Note that dsx (D) and dally (F) are expressed in a distinct subset of bract cells, which likely corresponds to sex comb bracts. (G) 24 h APF male pupal upper tarsal segments showing staining from 1151-GAL4 > UAS-mCherry.nls (magenta) and the neuronal marker anti-Futsch (green). 1151-GAL4 marks tendons [84]. The arrangement of tendon cells is clearly distinct from the paired nerve fibers that run along the same axis. (H-J) 24 h APF first tarsal segment and distal tibia from an 1151-GAL4 > UAS-mCherry.nls (magenta) male counterstained with anti-Vvl (green). Costaining is clearer in the levator and depressor tendons at the distal tibia/ta1 joint (marked with an arrow) than in the long tendon, which extends along the proximal–distal axis of the tarsal segments. This may be due to the greater concentration of tendon cells in this region and difficulties distinguishing between anti-Vvl staining in mechanosensory bristle cells (see Fig 6Q–6S) and tendon cells. Data and code for generating the scRNA-seq elements of this figure are available at https://www.osf.io/ba8tf.
Fig 5.
Noncanonical expression patterns in leg glia and a new cell type associated with the neural lamella.
(A) Annotated UMAP plot of nonsensory cells. While the Oaz+ and svp+ glia cells express the canonical glia marker repo, the Sox100B+ cells do not. Cell labels provided in purple indicate populations that are discussed in this figure. Those in black are discussed in Fig 4. (B) Dot plot of the expression of repo and gcm, the canonical glia markers, along with top differentially expressed genes identified through comparisons between each of Sox100B+ glia, Oaz+/repo+ glia, and svp+/repo+ glia against all the clusters named in (A). Of these genes, sr is known to induce tendon cell fate, and to the best of our knowledge, no functions have previously been reported for sr in glia. Unc-5 and Fas2 are both required for glial migration (reviewed in [93]). Genes identified as differentially expressed through more targeted between-glia comparisons are given in S7A Fig. (C-G) UMAP plots of the subsetted Sox100B+ cells, Oaz+/repo+ glia, and svp+/repo+ glia from (A) overlaid with the expression of a series of top marker genes for each cluster. (H-M) The UMAP plot shown in (A) overlaid with the expression of a series of glia markers. moody and Gli are subperineural glia markers, nrv2 and Ntan1 are wrapping glia markers, apt is a surface glia marker (i.e., a marker of both perineural and subperineural glia), and gcm is the upstream determinant of glial identity [89,96]. (N-P) 24 h APF legs from repo-GAL4>UAS-mCherry.nls males counterstained with anti-Oaz, a marker of wrapping glia [95]. Oaz+ cells are denoted by an arrow in the left-hand image. In Oaz+ cells at 24 h APF, the repo-GAL4+ signal was often weak and in one of the 4 legs we imaged, the one shown here, we observed a single Oaz+ cell that appeared repo-GAL4−. This is the topmost of the Oaz+ cells to which an arrow is pointing. (Q, R) Brain, ventral nerve cord, and leg discs (Q) and a close-up of a leg disc (R) from repo-GAL4>UAS-mCherry.nls males counterstained with anti-Sox100B. Note how repo-GAL4+ cells can be seen migrating into the disc from the CNS, while Sox100B+ cells appear to originate within the disc itself. (S-U) 24 h APF male upper tarsal segments from repo-GAL4 > UAS-mCherry.nls counterstained with anti-Sox100B. Both show a similar, but nonoverlapping, distribution of stained cells. (V-X) 24 h APF male upper tarsal segments from Lim1-GAL4 > UAS-mCD8::GFP counterstained with anti-Futsch. Above the tibia/ta1 joint, Lim1-GAL4 was expressed in the epithelial cells of the distal tibia, as predicted by our epithelial joint analysis (Fig 3I). Below the joint, the staining surrounded and spanned the distance between the 2 central axon trunks into which the sensory neuron axons project. (Y-AA) 24 h APF male upper tarsal segments from repo-GAL4 > UAS-mCD8::GFP counterstained with anti-Futsch. Unlike the Lim1-GAL4, repo-GAL4 staining does not span the gap between the 2 axon trunks with which it is closely associated, and cell bodies are clearly seen branching away from the fibers. (AB-AG) The UMAP plot shown in (A) overlaid with the expression of a series of top markers identified in this study. Data and code for generating the scRNA-seq elements of this figure are available at https://www.osf.io/ba8tf.
Fig 6.
Identification of a combinatorial transcription factor code for leg sensory neurons.
(A) Annotated UMAP plot of neuronal cells from the integrated 24 h AFP and 30 h APF first tarsal segment dataset. GRN, gustatory receptor neuron; MSNCB, mechanosensory neuron in chemosensory bristle. See S9A–S9H Fig for details on the TkR86C+ mechanosensory neurons. (B) Dot plot showing the expression of a series of canonical neuronal markers (elav, nSyb, and para) and transcription factors across the major neuron class clusters labeled in the UMAP given in (A). Each cluster expresses a unique combination. The dotted lines separate the canonical neuronal markers and then the chemoreceptor from mechanoreceptor organs transcription factor markers. (C-F) UMAP plot of neuronal cells from the integrated 24 h AFP and 30 h APF first tarsal segment dataset overlaid with the expression of members of the transcription factor code depicted in (B). (C) Note how the MSNCB cluster branching off from the top of the mechanosensory neuron population is negative for both vvl and pros. (D) Ets65A is present in all non-GRN populations in the UMAP, while an effector of sex differentiation, dsx, is expressed in GRNs, sex comb neurons, and MSNCBs. (E) fru, the other effector of sex differentiation, is enriched in 2 GRN populations and sex comb neurons, while eyg is restricted to campaniform sensilla neurons. (F) CG42566 is the only nontranscription factor plotted. It is a top marker of MSNCBs and its expression in both MSNCBs and GRNs contributed to this cluster’s chemosensory bristle annotation. ham is enriched in mechanosensory neuron classes and 2 GRN populations. (G) Annotated UMAP plot of male neuronal cells subsetted from the Fly Cell Atlas single-nuclei RNA-seq leg dataset [38]. Note the presence of 3 clusters, annotated as “putative chordotonal,” which are absent from the pupal dataset—chordotonal organs are not present in the upper tarsal segments. No clear MSNCB or sex comb clusters could be resolved in this dataset. (H) As (B) but for the male neuronal cells subsetted from the Fly Cell Atlas single-nuclei RNA-seq leg dataset. Only those clusters present in the pupal single-cell data are shown. (I) UMAP plot of male neuronal cells subsetted from the Fly Cell Atlas single-nuclei RNA-seq leg dataset overlaid with expression of the mechanosensory neuron marker vvl (blue) and a top marker of the putative chordotonal organs, the predicted transcription factor CG9650 (red). (J) A subset of (I), showing only the putative chordotonal clusters overlaid with expression of 2 transcription factors, bab1 (blue) and erm (red). (K-V) Confocal images of 24 h APF male first tarsal segments. (K-M) Mechanosensory bristles from elav-GAL4 > UAS-mCherry.nls (magenta) stained with anti-Pros (green). Two elav-GAL4+ cells are present per mechanosensory bristle, one of which, the sheath, is Pros+. elav-GAL4 expression in the sheath is likely due to the legs being imaged soon after the division of the common pIIIb progenitor cell from which they derive (see also [113] and S9K–S9M Fig). MSN, mechanosensory neuron. (N-P) Two chemosensory bristles (circled) from elav-GAL4 > UAS-mCherry.nls (magenta) stained with anti-Pros (green). Note that each bristle includes 4 Pros+/elav-GAL4+ cells (the gustatory receptor neurons), 1 Pros+/elav-GAL4− cell (the chemosensory sheath cell), and 1 Pros−/elav-GAL4+ cell (the MSNCB, mechanosensory neuron in chemosensory bristle). (Q-S) Two chemosensory (CS) bristles and 1 mechanosensory (MS) bristle from elav-GAL4 > UAS-mCherry.nls (magenta) stained with anti-Vvl (green). Note that anti-Vvl staining is entirely absent from the CS bristle including, therefore, the mechanosensory neuron (MSNCB) that innervates it. Conversely, anti-Vvl staining is observed in all 4 constituent cells of a MS bristle. (T-V) The same stainings performed in (Q-S) but centered on the sex comb. Anti-Vvl staining is present in both the neuronal (elav-GAL4+) and nonneuronal cells of the sex comb. The “central bristle,” which develops from the same bristle row as the sex comb is labeled. (W-Y) Confocal images of 48 h APF male first tarsal segments showing the expression of fru-GAL4 (magenta) and anti-Futsch (green). fru-GAL4 expression is restricted to the sex comb and chemosensory (CS) neurons. The later 48 h time point was used as fru-GAL4 was undetectable up until 40 h and weak up until 48 h. (Z) Confocal image of the first tarsal segment from a 24 h male from eyg-GAL4 > UAS-GFP.S65T. Campaniform sensilla are marked with asterisks. The axonal projections can be seen as parallel lines running either side of the central autofluorescence. Note that some nonspecific fat body staining is also present in this image. (AA) Confocal image of a distal first tarsal segment campaniform sensillum from a 24 h male where eyg-GAL4 is driving the expression of UAS-mCherry.nls. The top and bottom image in this panel show the same sensillum but with different levels of saturation to variously highlight the domed structure (top) and the individual cells of the organ (bottom). (AB) As (Z) but showing an adult haltere. Note that the staining is restricted to the campaniform sensilla field on the pedicel (“Ped.”) and apparently absent from the field on the scabellum (“Sca.”). Data and code for generating the scRNA-seq elements of this figure are available at https://www.osf.io/ba8tf.
Fig 7.
Four gustatory receptor neuron (GRN) classes express a combinatorial transcription factor code and unique gene repertoires.
(A) Annotated UMAP of the pupal integrated neuron data. GRN, gustatory receptor neuron; MSNCB, mechanosensory neuron in chemosensory bristle. The number of cells in each GRN cluster is presented. The numbers are generally similar between each GRN population, with the exception of the fru+ male-sensing GRNs. This population was closely associated with the fru+ female-sensing GRNs, more so than were any other 2 GRN subtypes, and the interface between them in UMAP space contained several cells bearing intermediate characteristics. Consequently, the discrepancy in cell numbers between fru+ GRN populations may reflect classification errors due to transcriptomic similarities. (B-F) The UMAP shown in (A) overlaid with the expression of 5 transcription factors (pros, acj6, nvy, fkh, and fru) that are expressed in unique combinations in each of the 4 GRN clusters. (G-AA) Testing the GRN transcription factor code derived from the scRNA-seq data on 24 h APF male first tarsal segments. (G-I) anti-Nvy (green) and anti-Pros (magenta). Note that 1 Pros+ cell is partially obscuring the Pros+ sheath cell, which has a distinct, elongated morphology. (J-L) Anti-Pros (green) and anti-Fkh (magenta). (M-O) Anti-Pros (green) and fru-GAL4 > UAS-mCherry.nls (magenta). (P-R) Anti-Acj6 (green) and fru-GAL4 > UAS-mCherry.nls (magenta). (S-U) Anti-Acj6 (green) and anti-Fkh (magenta). (V-X) Anti-Nvy (green) and anti-Acj6 (magenta). (Y-AA) Anti-Nvy (green) and anti-Fkh (magenta). (AB) A schematic summarizing the expression patterns of each transcription factor across GRNs, along with a selection of other genes detected in each subtype. Gene names are colored pink, green, or blue when shared across multiple GRN subtypes. (AC-AH) Recovery of the same transcription factor code in the Fly Cell Atlas single-nuclei adult male leg neuron data. Note that in the adult data, nvy is barely detected. Correspondence between the nvy+ cluster in the pupal and adult data was supported by additional marker genes, such as foxo and Fer1 (see S11B, S11C, S11H, and S11I Fig). (AC) As in (A), the number of cells in each GRN population is presented. In this dataset, a subregion of what unsupervised clustering labeled as the fru+/acj6− population showed acj6 expression, suggestive of a classification error. This conclusion is further supported by the VGlut, ppk25, and ppk10 data below (see S10Y–S10AB Fig). We therefore manually labeled these as part of the fru+/acj6+ cluster. As in the pupal data, the interface between the 2 fru+ populations appeared particularly close. (AI) A dot plot summarizing the expression of a selection of top differentially expressed genes for each cluster that we identified in the Fly Cell Atlas single-nuclei adult male leg neuron data. Data and code for generating the scRNA-seq elements of this figure are available at https://www.osf.io/ba8tf.
Fig 8.
Distinct and shared modules of gene expression between sensory organ support cells.
(A) The 4 constituent cell types of external sensory organs, such as the mechanosensory bristle in this schematic, originate through asymmetric divisions of a sensory organ precursor (SOP) cell (reviewed in [138]). The SOP divides to produce a pIIa and pIIb daughter cell. pIIa further divides to generate a socket and shaft cell. In the notum, where it’s been studied, pIIb divides into a pIIIb cell and glial cell, the latter of which enters apoptosis soon after birth [188]. pIIIb further divides to produce the sheath and neuron. To the best of our knowledge, whether the pIIb glial division occurs in the leg remains untested. (B) Annotated UMAP plot of the bristle cells from the integrated 24 h AFP and 30 h APF first tarsal segment dataset. The campaniform support cluster included Su(H)+ cells, which suggests that it corresponds to socket cells, but it’s possible that it includes a mix of campaniform sensilla accessory cell types. (C-F) The UMAP shown in (B) overlaid with the expression of a series of marker genes, either previously published or demonstrated in this study, for different sensilla classes or accessory cell types. (G) A dot plot summarizing the expression patterns of a selection of genes identified as being differentially expressed in each of the clusters given in the UMAP shown in (B). Dotted lines separate the 3 major classes of sensory support cell. Of the socket markers, CG31676 is known to be expressed in a subset of olfactory projection neurons [133]; nw is a C-type lectin-like gene; stan, a cadherin that controls planar cell polarity [189]; and nrm, ed, and hbs are cell adhesion molecule genes [143–147]. Of the shaft markers, CG9095 encodes a C-type lectin-like gene; disco-r encodes a transcription factor; dUTPase encodes a nucleoside triphosphate; spdo encodes a transmembrane domain containing protein that regulates Notch signaling during asymmetric cell division [190–192]; and sha encodes a protein involved in the formation of bristle hairs [148]. Aside from pros and nompA, the top markers of the sheaths include the following: the transcription factors Glut4EF, pnt, and SoxN; jv, which encodes a protein involved in actin organization during bristle growth [149]; qua, which encodes an F-actin cross-linking protein [150]; and the midline glia marker wrapper, which encodes a protein involved in axon ensheathment [99–101]. (H-S) The UMAP shown in (B) overlaid with the expression of genes identified in this study as markers of sensory organ support cell subtypes. Data and code for generating the figure are available at https://www.osf.io/ba8tf.
Fig 9.
First and second order differences in transcription factor expression between sensory neuron classes.
GRN, gustatory receptor neuron; MSNCB, mechanosensory neuron in chemosensory bristle.
Table 1.
Reagents and resources used in this study.