Identification of Split-GAL4 Drivers and Enhancers That Allow Regional Cell Type Manipulations of the Drosophila melanogaster Intestine

Abstract

The Drosophila adult midgut is a model epithelial tissue composed of a few major cell types with distinct regional identities. One of the limitations to its analysis is the lack of tools to manipulate gene expression based on these regional identities. To overcome this obstacle, we applied the intersectional split-GAL4 system to the adult midgut and report 653 driver combinations that label cells by region and cell type. We first identified 424 split-GAL4 drivers with midgut expression from ∼7300 drivers screened, and then evaluated the expression patterns of each of these 424 when paired with three reference drivers that report activity specifically in progenitor cells, enteroendocrine cells, or enterocytes. We also evaluated a subset of the drivers expressed in progenitor cells for expression in enteroblasts using another reference driver. We show that driver combinations can define novel cell populations by identifying a driver that marks a distinct subset of enteroendocrine cells expressing genes usually associated with progenitor cells. The regional cell type patterns associated with the entire set of driver combinations are documented in a freely available website, providing information for the design of thousands of additional driver combinations to experimentally manipulate small subsets of intestinal cells. In addition, we show that intestinal enhancers identified with the split-GAL4 system can confer equivalent expression patterns on other transgenic reporters. Altogether, the resource reported here will enable more precisely targeted gene expression for studying intestinal processes, epithelial cell functions, and diseases affecting self-renewing tissues.

The Drosophila midgut is a model system for understanding tissue homeostasis and adaptation in response to environmental challenges, including dietary changes, alterations in microbiota, and the ingestion of pathogens and harmful chemicals (Miguel-Aliaga et al. 2018). While physiologically and functionally orthologous to the mammalian intestine, the fly midgut epithelium is comparatively simple and composed of only four main cell types (Zwick et al. 2019): intestinal stem cells (ISCs), their undifferentiated enteroblast (EB) daughters, and two differentiated cell types, absorptive enterocytes (ECs) and hormone-producing enteroendocrine cells (EEs) (Micchelli and Perrimon 2006; Ohlstein and Spradling 2006). These four cell types display distinct regional identities along the length of the midgut, which has been divided into five main regions (R1–R5) that carry out the sequential processes of food breakdown, nutrient absorption, and waste elimination (Buchon et al. 2013; Marianes and Spradling 2013). In contrast to most tissues, which show little cell turnover during adult life, the midgut is continually remodeling through careful adjustments to stem cell division rates, division patterns, and the decisions of ISCs and EBs to differentiate into EEs or ECs [for review, see Li and Jasper (2016) and Miguel-Aliaga et al. (2018)]. These dynamic cell composition changes make the midgut exceptionally well suited for studying cellular processes common to all self-renewing tissues under normal and pathological states.

A current limitation to the use of the Drosophila midgut as a model system is the lack of tools to manipulate gene expression in a cell type and/or regional manner. The most common tool for experimentally controlling gene expression in Drosophila is the GAL4–UAS (upstream activating sequence) system (Brand and Perrimon 1993). Pioneering studies reported the intestinal expression profiles of dozens of GAL4 drivers and identified several that were expressed in specific regions (Buchon et al. 2013; Marianes and Spradling 2013), but many of these drivers have limited utility because their cell type expression was not characterized or they express in multiple cell types. In addition, few midgut enhancer sequences have been molecularly characterized, so it is difficult to design new transgenes targeting defined cells. Most known enhancers do not limit expression to a specific cell type (e.g., the mex1 enhancer, present in a widely used EC driver, also directs expression in stem cells and EBs) (Lucchetta and Ohlstein 2017), or they do not direct expression in all cells of a single type (e.g., EE enhancers from AstA, AstC, Dh31, and Tk and EC enhancers from Myo31DF, drive expression in only a subset of these cell types) (Beehler-Evans and Micchelli 2015; Chen et al. 2016). Better tools would enable investigators to target specific cells for genetic experimentation based on their type and their position within biochemically and physiologically specialized intestinal regions.

We have addressed this limitation by developing a strategy for characterizing the intestinal cell type expression profiles of split-GAL4 drivers and applying it to the large collection of split-GAL4 driver stocks maintained at the Bloomington Drosophila Stock Center (Tirian and Dickson 2017; Dionne et al. 2018). In the split-GAL4 system, the DNA-binding domain of GAL4 and a transcriptional activator are expressed from separate transgenes with different regulatory sequences (Luan et al. 2006). Combining the two transgenes with simple crosses activates UAS transgenes only in cells where both regulatory sequences are active (Figure 1A). While regulatory sequences typically direct transcription in widespread and complex patterns, this split-GAL4 intersectional scheme often defines very limited sets of cells or even single cells. The expression of these split-GAL4 drivers has been characterized most extensively in the adult brain, a complex, terminally differentiated tissue with little cell turnover in adults (Dionne et al. 2018; Wolff and Rubin 2018; Chen et al. 2019; Dolan et al. 2019; Sekiguchi et al. 2020). In this study, we extend the analysis of these drivers to the intestine. We first identify the subset of split-GAL4 drivers that are expressed in the midgut, and then characterize the cell type expression patterns of these drivers relative to a set of cell type-specific reference strains. We also show that enhancer sequences characterized by this approach can generate equivalent expression patterns when used in other transgenes. Altogether, this analysis identifies a large set of useful drivers and enhancer sequences that will enable more precise gene manipulations for studying intestinal processes, epithelial cell functions, and diseases affecting self-renewing tissues.

Figure 1.

PM split-GAL4 drivers label most cells in the Drosophila intestinal epithelium. (A) Identifying intestinally active enhancers using the split-GAL4 system. In this example, a midgut-only DBD reference driver restricts expression of UAS-GFP to a specific intestinal region when combined with a p65AD driver expressed more widely. (B and C) Larval and adult expression of UAS-6XGFP driven by PM split-GAL4DBD and p65AD drivers (CG10116-p65AD ∩ CG10116-GAL4DBD). (D and E) UAS-6XGFP expression driven by PM-AD and PM-DBD drivers in female (D) and male (E) intestines counterstained with DAPI (blue). Regions are labeled for reference. (F) UAS-GFP.nls expression driven by PM split-GAL4 drivers in the five regions (R1–5) of female intestines stained for GFP (anti-GFP in green), Ps (anti-HRP in red), EEs (anti-Pros in white), and nuclei (DAPI in blue). Individual channels of each panel are shown in Figure S2. (G) Graph showing percent of Ps, EEs, ECs, and all cells labeled by PM-AD and PM-DBD driving UAS-GFP.nls expression in each of the five intestinal regions of female intestines. No obvious differences were noted in male intestines. Ps were defined as cells showing anti-HRP staining, EEs as cells showing anti-Pros staining, and ECs as cells lacking both anti-HRP and anti-Pros staining. All cells showed DAPI staining. Cells were counted from representative fields of view for each region from 5 to 6 intestines. Graph shows means with SD. Bar, 500 μm (A–E), 25 μm (F). Complete genotypes are listed in Table S2. AD, activation domain; DBD, DNA-binding domain; EC, enterocyte; EE, enteroendocrine cell; P, progenitor cell; PM, pan-midgut; Pros, Prospero.

Materials and Methods

Drosophila strains and husbandry

All fly strains were cultured at 25° on standard Bloomington media (https://bdsc.indiana.edu/information/recipes/bloomfood.html) with ingredients in the following proportion: 15.9 g inactive dry yeast, 9.2 g soy flour, 67.1 g yellow cornmeal, 5.3 g agar, 71 ml light corn syrup, 4.4 ml propionic acid, and 1000 ml water. The full genotypes of all starting stocks used in this study are listed in Supplemental Material, Table S1, and the drivers and responders used in each figure are listed in Table S2. The PBac{UAS-DSCP-6XEGFP}VK00018 and PBac{UAS-DSCP-6XEGFP}attP2 strains were gifts from Steve Stowers (Montana State University).

Transgenes

Transgene construction was performed using NEBuilder HiFI DNA Assembly Master Mix (New England Biolabs, Beverly, MA: catalog number E2621), restriction enzyme-digested plasmids, and PCR products generated with primers listed in Table S3. Plasmids were sequence-verified, amplified, and sent to Rainbow Transgenic Flies (Camarillo, CA) for injection.

Pan-midgut split-GAL4 transgenes:

The Zip-GAL4DBD and p65AD-Zip open reading frames were assembled downstream of the 1-kb CG10116 enhancer fragment by combining the XhoI/XbaI-digested CG10116p-KD::PEST plasmid (Buddika et al. 2020b) with either a Zip-GAL4DBD-containing PCR fragment amplified with oligo pair 3855/3856 (see Table S3 for oligo sequences) from pBPZpGAL4DBDUw or a p65AD-Zip-containing PCR fragment amplified with oligo pair 3857/3858 from pBPp65ADZpUw [both plasmids (Addgene 26233 and 26234) were gifts from Gerald Rubin (Pfeiffer et al. 2010)]. Transgenesis of these plasmids yielded P{CG10116-GAL4.DBD}su(Hw)attP6 and P{CG10116-p65.AD}attP40.

Cell type split-GAL4 reference lines:

Cell type-specific split-GAL4 reference lines were generated by swapping the CG10116 enhancer fragment in the plasmids described above for either the 2.4-kb GMR57F07 enhancer fragment from Dh31 or the synthetic GBE fragment containing tandem arrays of Grainyhead- and Suppressor of Hairless-binding sites (Furriols and Bray 2001; Beehler-Evans and Micchelli 2015). The GMR57F07 enhancer fragment was amplified from genomic DNA using oligo pair 3955/3956, while the GBE sequence was amplified from the previously described 3Xgbe-smGFP::V5::nls plasmid (Buddika et al. 2020a) with oligo pair 3450/3451. Transgenesis of these plasmids yielded P{R57F07-p65.AD.A}attP40, P{R57F07-GAL4.DBD.A}attP2, and P{GBE-GAL4.DBD}attP2.

Enhancer-GAL4 and -smGFP.V5.nls transgenes:

Enhancers were PCR-amplified using oligonucleotide primers listed in Table S3. These enhancers were cloned via Gibson Assembly into 3Xgbe-excised versions of either the 3Xgbe-smGFP::V5::nls plasmid (Buddika et al. 2020a) or the same plasmid in which smGFP::V5::nls had been replaced with GAL4.

Dissections and immunostaining

For most analyses of UAS-6XGFP and UAS-Stinger expression, gastrointestinal tracts were dissected in 1× PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, and KH₂PO₄, pH 7.4), fixed in 4% paraformaldehyde (Electron Microscopy Sciences, catalog number 15714) in PBS for 45 min, washed multiple times in 1× PBT (1× PBS and 0.1% Triton X-100), including one wash with 5 μg/ml DAPI in PBT, and mounted in 1,4-Diazabicyclo[2.2.2]octane–containing mounting medium. For all antibody labeling except Delta (Dl) staining, samples were fixed as above, and then incubated at 4° overnight with primary antibodies, including mouse anti-Prospero (Pros) (MR1A, 1:100; Developmental Studies Hybridoma Bank), mouse anti-V5 (MCA1360GA, 1:250; Bio-Rad, Hercules, CA), and rabbit anti-GFP (A11122, 1:1000; Life Technologies). The following day, samples were washed in 1× PBT and incubated for 2–3 hr with secondary antibodies, including AlexaFluor 488- and 568-conjugated goat anti-rabbit, -mouse, -rat, and -chicken antibodies (1:1000; Life Technologies). AlexaFluor 647-conjugated goat-HRP antibodies (123-605-021; Jackson ImmunoResearch) were used in the secondary antibody solution whenever required. Finally, samples were washed multiple times in 1× PBT, including one wash with 5 μg/ml DAPI in PBT, and mounted in Vectashield mounting medium (Vector Laboratories, Burlingame, CA). An alternative staining protocol was used for mouse anti-Dl (C594.9B, 1:500; Developmental Studies Hybridoma Bank) staining as described in Buddika et al. (2020b) and these samples were mounted in ProLong Diamond mounting medium (P36970; Invitrogen, Carlsbad, CA). Intestines stained with anti-Dl antibody were also stained with anti-GFP antibody, since the methanol steps required in this protocol quenched GFP fluorescence.

Microscopy and image processing

Images of whole flies and dissected intestines during the screens were collected on a Zeiss (Carl Zeiss, Thornwood, NY) Axio Zoom microscope. Images of immunostained intestines were collected on a Leica SP8 Scanning Confocal microscope. Samples to be compared were collected under identical settings on the same day, image files were adjusted simultaneously using Adobe Photoshop, and figures were assembled using Adobe Illustrator.

Expression pattern analysis

Intestines from 4–8-day-old flies were analyzed and the expression patterns of at least five independent intestines were scored per genotype. For the primary screen, third-instar larvae and adults were examined for fluorescence. When fluorescence was observed in adults, males and females were dissected, and expression patterns were characterized by scoring the presence or absence of GFP expression in each of the five main intestinal regions (R1–R5). Variability among samples was noted. For the secondary screen, males and females were examined for drivers where a dimorphism was noted in the initial screen; only females were examined for the remainder. Expression patterns were characterized by semiquantitatively estimating the number of cells that displayed GFP expression in each of 11 intestinal subregions (R1A, R1B, R2A, R2B, R2C, R3, R4A, R4B, R4C, R5A, and R5B) (Buchon et al. 2013) on a scale from 0 to 3, where 0 indicated no cells, 1 indicated some cells (∼1–49%), 2 indicated most cells (∼50–99%), and 3 indicated all cells (∼100%). The variability in GFP intensity among cells of labeled intestines was also analyzed semiquantitatively: given a 0 when no midgut expression was present, a 1 when there was more than a twofold difference among ≥25% of cells, and a 2 when all labeled cells were roughly the same intensity. The variability of expression among intestines of the same genotype was scored on a scale from 0 to 3, where 0 indicated that all intestines had the same pattern, 1 indicated that the majority had the same pattern, 2 indicated that there was a pattern common to all samples but also some notable differences among them, and 3 indicated that all intestines displayed different patterns.

Data availability

Plasmids are available upon request. Extant stocks are available from the Bloomington Drosophila Stock Center or upon request. Figures S1 and S3 show the results of control experiments. Figure S5 shows representative images summarized in Table 2. Figures S2, S4, S6, and S7 show individual channels of micrographs presented in Figure 1, Figure 3, Figure 5, and Figure 7, respectively. Table S1 is the reagent table. Table S2 lists the genotypes of all strains shown in figures. Table S3 is a list of all oligonucleotide primers. Table S4 lists expression patterns of new GAL4 and GFP transgenes. File S1 describes driver characterizations in the primary screen. File S2 lists expression patterns characterized in the secondary screen. The results of our primary and secondary screens are also available online at https://bdsc.indiana.edu/stocks/gal4/midgut_splitgal4.html. File S3 lists drivers with clear regional expression boundaries. The authors affirm that all data necessary for confirming the conclusions of the article are present within the article, figures, and tables. Supplemental material available at figshare: https://doi.org/10.25386/genetics.13011518

Table 2. UAS-GFP expression in anterior, middle, and posterior regions when reference drivers were combined^a.

	PM-DBD	P-DBD	EE-DBD	EC-DBD
PM-AD	+ – +	+ + +	– – –	+ – +
P-AD	+ – +	+ – +	– – –	– – –
EE-AD	– – –	– – –	+ + +	– – –
EC-AD	+ – +	– – –	– – –	Lethal

AD, activation domain; DBD, DNA-binding domain; EC, enterocyte; EE, enteroendocrine cell; PM, pan-midgut; P, progenitor cell.

^aPresence (+) or absence (−) of GFP expression indicated for anterior (R1–R2), middle (R3), and posterior (R4–R5) regions.

Figure 3.

Reference split-GAL4 drivers label specific intestinal cell types. (A–D) GFP expression in intestinal regions R1–5 driven by (A) P drivers (VT004241-p65AD ∩ VT024642-GAL4DBD), (B) EE drivers (R57F07-p65AD ∩ R57F07-GAL4DBD), (C) an EC-AD driver (VT004958-p65AD ∩ CG10116-GAL4DBD), and (D) an EC-DBD driver (VT004958-GAL4DBD ∩ CG10116-p65AD). Anti-GFP antibody staining is shown in green with counterstaining for Ps (anti-HRP in red), EEs (anti-Pros in white), and all cell nuclei (DAPI in blue). Bar, 25 μm. (E–H) Histograms showing the percentage of (E) Ps, (F) EEs, and (G and H) ECs labeled with GFP by region. Cell identities were defined and cells counted as in Figure 1. Individual channels of all micrographs are shown in Figure S4. Complete genotypes are listed in Table S2. AD, activation domain; DBD, DNA-binding domain; EC, enterocyte; EE, enteroendocrine cell; P, progenitor cell; Pros, Prospero.

Figure 5.

The secondary screen identified split-GAL4 drivers that label specific progenitor and EE cell populations. (A) GFP expression in R1, R2, R4, and R5 driven by EB-DBD combined with PM-AD. Anti-GFP antibody staining is shown in green with counterstaining for ISCs (anti-Dl, red cytoplasmic staining), EEs (anti-Pros, red nuclear staining), progenitors (anti-HRP, white), and the DNA marker DAPI (blue). Individual channels of each panel are shown in Figure S6. (B) Quantification of the percent of EBs labeled by GFP in four intestinal regions. EBs are defined as cells showing anti-HRP staining but not anti-Dl staining. (C–E) Intestinal expression of UAS-6XGFP (green) when EB-DBD is combined with VT055990-p65AD, R78B06-p65AD, or R10F10-p65AD. (F) Intestinal expression of UAS-Stinger when R10F08-p65AD is combined with P-DBD (yellow arrowheads indicate locations of small subsets of labeled cells). (G) UAS-Stinger expression driven by R10F08-p65AD and PM-DBD in an intestine counterstained for EEs (anti-Pros, red), Ps (anti-HRP, white), and DAPI (blue). Yellow arrowheads indicate HRP+, Pros+, Stinger + cells. Bar, 25 μm (A and G), 500 μm (C–F). Complete genotypes are listed in Table S2. AD, activation domain; DBD, DNA-binding domain; EB, enteroblast; EE, enteroendocrine cell; ISC, intestinal stem cell; P, progenitor cell; PM, pan-midgut; Pros, Prospero.

Figure 7.

Enhancer fragments can confer similar expression patterns in different transgenic contexts. (A) UAS-Stinger expression (green) in adult female intestines when driven by six enhancer fragments in the split-GAL4 (left) or GAL4 (right) systems. Yellow arrowheads indicate minor expression differences. DAPI staining is in blue. (B) UAS-Stinger expression (green) driven by VT024642-GAL4 in intestinal regions R1–5 with counterstaining for Ps (anti-HRP in red), EEs (anti-Pros in white), and all cell nuclei (DAPI in blue). Individual channels of each panel are shown in Figure S7. (C) Quantification of P labeling by VT024642-GAL4 driving UAS-Stinger expression in R1–5. Bar, 500 μm (A), 25 μm (B). Complete genotypes are listed in Table S2. EE, enteroendocrine cell; P, progenitor cell; Pros, Prospero.

Results

Construction and verification of reference drivers for primary screen

To screen the split-GAL4 collection efficiently, we first designed reference stocks to restrict expression of individual split drivers to the midgut. These reference stocks contained an intestine-specific split-GAL4 transgene [encoding either the p65 activation domain (p65AD) or GAL4 DNA-binding domain (GAL4DBD)] and a clearly detectable UAS-reporter transgene. We cloned the p65AD or GAL4DBD open reading frame downstream of an enhancer from the CG10116 gene, whose activity in other transgenic contexts was limited to the gastrointestinal tract (Buddika et al. 2020b). To verify that these new CG10116-p65AD and CG10116-GAL4DBD drivers were expressed throughout the intestine as expected, we examined flies carrying both transgenes and either P{20XUAS-6XGFP}attP2 or P{20XUAS-DCSP-6XEGFP}attP2, which express a hexameric version of GFP under the control of 20 copies of UAS (Shearin et al. 2014; Williams et al. 2019). Because these insertions differ only in promoter sequences and were interchangeable in marking intestinal cells, we will refer to them both as UAS-6XGFP. Both larvae and adults displayed clear, strong intestinal fluorescence, which could be seen through the abdominal cuticle, and little fluorescence elsewhere (Figure 1, B and C). All midgut regions from both females and males were strongly fluorescent except for region 3 (R3), which had no apparent expression (Figure 1, D and E). Importantly, we did not see UAS-6XGFP expression when CG10116-p65AD or CG10116p-GAL4DBD were combined, respectively, with “empty” GAL4DBD or p65AD transgenes carrying no enhancer sequences (Figure S1). These results indicated that we could use these new drivers, which we refer to as PM-AD (pan-midgut-p65AD) and PM-DBD (PM-GAL4DBD) in the remainder of the text, to identify split-GAL4 drivers that have midgut expression, with the possible exception of drivers that express only in R3.

To evaluate the intestinal cell types expressing PM-AD and PM-DBD in detail, we combined these drivers with a nuclear GFP reporter (UAS-GFP.nls) that was easier to score on a cell-by-cell basis than UAS-6XGFP. We detected nuclearly localized GFP expression with an anti-GFP antibody and counterstained intestines with anti-HRP and anti-Pros antibodies, and DAPI, which label progenitor cells (Ps), EEs, and all cells, respectively (Shiga et al. 1996; Micchelli and Perrimon 2006; O’Brien et al. 2011; Miller et al. 2020). Ps include both ISCs and EBs, two cell types that have highly similar transcriptional profiles (Dutta et al. 2015; Hung et al. 2020). While not directly labeled, we classified ECs as cells with neither anti-HRP nor anti-Pros antibody staining. We found that most cells in the intestinal epithelium expressed GFP, ranging from 75% in R3 to 94% in R2 and R4 (Figure 1, F and G and Figure S2). This included almost all ECs (85–100%) and the majority of Ps (65–97%). However, EEs in most regions were not efficiently marked. While 68% were labeled in R2, only 5% were labeled in R5 and none were labeled in R3. These results indicated that the CG10116 enhancer was active in the majority of intestinal cells and that, consequently, PM-AD or PM-DBD could be paired with any other split-GAL4 driver to demonstrate intestinal expression with the possible exception of drivers expressed in only a subset of EEs.

Within the intestine, we saw no differences in the patterns of GFP fluorescence with either UAS-6XGFP construct, UAS-GFP.nls or UAS-Stinger, which expresses a nuclear GFP variant (Barolo et al. 2000) (Figure S3), although each construct showed a characteristic expression level (with UAS-GFP.nls expression low enough to require anti-GFP antibody staining for reliable detection). There was a single idiosyncratic exception: only UAS-GFP.nls was strongly detected in R3 in the presence of PM drivers. The general interchangeability of UAS-GFP responders in the intestine meant that we could choose the one with optimal expression for each experimental approach.

Primary screen identifies 424 split-GAL4 drivers with adult midgut expression

We used PM-AD or PM-DBD in combination with UAS-6XGFP to screen the Bloomington Drosophila Stock Center collection of split-GAL4 stocks, which carry driver transgenes generated at Janelia Research Campus and the Institute of Molecular Pathology in Vienna (Tirian and Dickson 2017; Dionne et al. 2018). From the 7304 stocks screened, we identified 590 total drivers, including 417 GAL4DBD and 173 p65AD drivers, giving abdominal fluorescence in third-instar larvae, young female adults, and/or young male adults (Table 1). A total of 424 of these included adult midgut expression. As a control for our detection method, we randomly chose 50 drivers that were scored as negative in whole animals and verified that dissected intestines lacked GFP fluorescence. However, we cannot rule out the possibility that our approach failed to detect drivers that sparsely or weakly labeled intestines. Among the 590 positive drivers, we noted several expression categories, including drivers that labeled subsets of cells based on sex, cell morphology, stage, and/or region (Figure 2, Table 1, and File S1). The 35 sex-dependent drivers included 6 expressed only in adult males, 7 expressed only in adult females, and 22 expressed in distinct patterns in each sex (Figure 2A). Cell morphology-specific drivers labeled subsets of cells of similar appearance throughout the adult midgut (Figure 2B) (these drivers were quantified and cell types were determined systematically in the subsequent secondary screens.) Stage-specific drivers were expressed in either larvae or adults (n = 358, Figure 2C). Finally, 142 region-dependent drivers selectively labeled patches of cells along the anterior–posterior (A–P) axis of the adult intestine, and ranged in expression from a small number of cells in a subregion to one-half of the gut (Figure 2D). An additional 71 drivers displayed variable expression patterns among adult intestines. We noted this variability (File S1) and included these drivers in subsequent analyses.

Table 1. Driver characterization in primary screen.

Characteristic	GAL4DBD	p65AD
Drivers screened	4,296	3,008
Intestinal expression	417	173
Adult	302	122
Sex-dependent	21	14
Stage-dependent	257	101
Larva only	115	51
Adult only	142	50
Regional	100	42
Variable expression	47	24

GAL4DBD, GAL4 DNA-binding domain; p65AD, p65 activation domain.

Figure 2.

The primary screen identified a variety of split-GAL4 expression patterns in the intestine. These micrographs show intestinal UAS-6XGFP expression (green) when split-GAL4 drivers were combined with pan-midgut drivers and stained with DAPI (blue). (A) Examples of sex-dependent patterns. R13G06-p65AD drives expression only in females and VT033050-GAL4DBD only in males. R83H01-GAL4DBD drives expression in different patterns in females and males. (B) Examples of cell morphology-specific patterns. VT063961-GAL4DBD and R48A01-GAL4DBD drive expression in similar-looking cells in distinctly different patterns. Insets are enlargements of indicated regions. (C) Examples of stage-dependent midgut patterns. VT014984-GAL4DBD drives expression only in adults and VT061206-GAL4DBD only in larvae. (D) Examples of region-dependent patterns including anterior (R94E02-p65AD), middle (VT019301-GAL4DBD), posterior (R85G08-GAL4DBD), or both anterior and posterior (R52A06-GAL4DBD) expression. Bar, 500 μm. The scale bar in D refers to all panels showing adult intestines. Complete genotypes are listed in Table S2.

Split-GAL4 reference strains label distinct intestinal cell types

To identify the cell types labeled by the 424 split-GAL4 drivers with adult expression, we assembled a collection of p65AD and GAL4DBD reference drivers for each of the three main midgut cell types: Ps, EEs, and ECs. For the Ps and ECs, we used drivers identified in the primary screen (VT004241-p65AD and VT024642-GAL4DBD for Ps; VT004958-p65AD and VT004958-GAL4DBD for ECs). For EEs, we generated new drivers containing the GMR57F07 enhancer fragment, which was previously shown to drive GAL4 expression in EEs (Beehler-Evans and Micchelli 2015). Because the extent and specificity of expression of these drivers was critical to their utility, we rigorously benchmarked their cell type expression using the UAS-GFP reporters and anti-HRP and anti-Pros staining described above (Figures 3 and Figure S4). VT004241-p65AD and VT024642-GAL4DBD combined with UAS-Stinger labeled 97–100% of Ps in R1, R2, R4, and R5, and 63% in R3 (Figure 3, A and E). Driver expression was specific to Ps, since only 4% of GFP-positive cells lacked anti-HRP staining. R57F07-p65AD and R57F07-GAL4DBD combined with UAS-6XGFP labeled 52–57% of EEs in R1, R2, R3, and R5, but only 3% in R4 (Figure 3, B and F). Although these drivers did not label all EEs, we adopted them as reference drivers since they labeled more EEs than any other driver tested (including a split-GAL4 version of the widely used P{GawB}pros^V1 driver, which labels most EEs, that we generated using the HACK method) (Balakireva et al. 1998; Lin and Potter 2016). Because the EC drivers, VT004958-p65AD and VT004958-GAL4DBD, caused lethality when combined, we evaluated their expression using the PM-AD and PM-DBD drivers with UAS-GFP.nls. Both EC drivers labeled 98–100% of ECs in all regions except in R3, where VT004958-p65AD labeled 93% and VT004958-GAL4DBD labeled 77% of ECs, with almost no other cells labeled (Figure 3, C, D, G, and H). Collectively, this characterization indicated that these six drivers showed the expected cell type expression. We refer to these drivers by their cell (P-, EE-, or EC-) and split type (-AD or -DBD) below, where we use them in combination with UAS-6XGFP.

To further validate the cell specificity of these reference drivers, we made pairwise combinations of the reference drivers (Figure S5 and Table 2). For the most part, the results were as expected: no GFP was detected when reference drivers expressed in distinct cell types were combined with each other, but GFP was detected when reference drivers labeling the same cell types were combined. We noted two exceptions. First, GFP was poorly expressed in R3 with the PM drivers, as noted above, as well as with the P and EC drivers. Second, we detected GFP expression in some large cells in R5 and occasionally R4 with the P and EE drivers, suggesting driver expression in a few ECs, which are the largest intestinal cells. Nevertheless, these results confirmed the specificity and effectiveness of the reference drivers in most regions.

Secondary screen defines cell type expression of 424 split-GAL4 drivers

To define the cell type expression patterns of the 424 split-GAL4 transgenes identified in the primary screen that drive in adults, we combined each with P, EE, and EC reference drivers and UAS-6XGFP. Expression patterns were analyzed in females for all drivers, and in males for drivers with sex-dependent expression in the primary screen. We semiquantitatively scored the intensity and variability of expression in the 11 subregions along the midgut A–P axis (R1A, R1B, R2A, R2B, R2C, R3, R4A, R4B, R4C, R5A, and R5B) (File S2). In females, we identified 173 drivers with P expression, 113 with EE expression, and 323 with EC expression (Table 3). While many of the EC-expressing drivers expressed only in ECs (152), fewer lines displayed P- or EE-specific expression (35 and 8, respectively). Many cell type-specific drivers labeled cells in only a specific subregion or in a subset of cells throughout the midgut (Figure 4, A–D). In contrast, 120 lines displayed expression in some combination of two cell types, and 58 lines displayed expression in all three cell types. By comparing driver expression profiles in the presence of all four reference drivers, we could deconstruct some complex patterns seen in our primary screen into combinations of simpler cell type-specific patterns (Figure 4, E–H), but the lack of simple “additivity” among other drivers reflects the complexity of the split-GAL4 system. We also identified drivers that labeled two or three cell types, each displaying a distinct regional pattern (Figure 4, E–G). A total of 81 drivers (19%) displayed some variability in expression among the intestines of a single sample, but, interestingly, no driver showed variable expression with all reference drivers, suggesting that the expression variability was cell type-dependent.

Table 3. Number of drivers with cell type expression from secondary screening in females.

Driver	GAL4DBD	p65AD
P only	27	8
EE only	5	3
EC only	114	38
P and EE only	6	1
P and EC only	52	21
EE and EC only	31	9
P, EE, and EC	30	28
No expression	34	11

EC, enterocyte; EE, enteroendocrine cell; P, progenitor cell.

Figure 4.

Cell type expression of representative split-GAL4 drivers. (A–H) Intestinal expression of UAS-6XGFP (green) when DBD drivers are combined with PM-, P-, EE-, and EC-AD drivers. (A) VT022223-GAL4DBD drives expression in a subset of Ps throughout the midgut, (B) VT046792-GAL4DBD in a subset of R5 Ps, (C) VT047163-GAL4DBD in a subset of R4 ECs, (D) VT063223-p65AD in most R5 ECs, (E) VT016117-GAL4DBD in Ps and EEs, (F) VT004977-GAL4DBD in anterior ECs and posterior Ps, (G) VT027938-GAL4DBD in ECs and EEs, and (H) R84D07-GAL4DBD in Ps, ECs, and EEs. Intestines were counterstained with DAPI (blue). Bar, 500 μm. Complete genotypes are listed in Table S2. AD, activation domain; DBD, DNA-binding domain; EC, enterocyte; EE, enteroendocrine cell; P, progenitor cell; PM, pan-midgut.

We also used the cell type-specific data to further examine the drivers with sex-dependent expression profiles identified with the PM reference drivers (File S2). Of the 22 drivers showing distinctly different expression patterns in males and females, differences in 12 of these drivers could be explained by EC expression alone, while dimorphisms in another three drivers could be explained by expression in ECs in combination with other cell types (EE, P, or both). With one other driver, we saw no expression with the P, EE, or EC reference drivers in either sex, and, with three drivers, we saw no cell type clearly responsible for the sex-dependent expression observed in our primary screen. Altogether, this analysis provides a comprehensive resource of split-GAL4-dependent expression profiles that can be used to target cells in a region- and cell type-specific manner. A database of expression profiles can be viewed at https://bdsc.indiana.edu/stocks/gal4/midgut_splitgal4.html.

Split-GAL4 drivers label subsets of Ps

To extend classification of the split-GAL4 drivers, we prepared an additional split-GAL4 reference driver that expresses GAL4DBD under the control of the synthetic GBE enhancer, known to direct expression in EBs, but not ISCs (Furriols and Bray 2001). To verify expression of this EB-DBD driver, we combined it with PM-AD and UAS-GFP.nls, and stained intestines with Fluor-tagged antibodies against GFP, HRP, and Pros as well as Dl protein, which is an established marker of ISCs (Figures 5A and Figure S6) (Micchelli and Perrimon 2006). We classified EBs as cells detected by anti-HRP antibodies, but not anti-Dl antibodies. We found that 82–99% of EBs expressed GFP in R1, R2, R4, and R5 (very few R3 cells were labeled), but only 6–13% of non-EB cells expressed GFP, indicating that EB-DBD effectively labels the EB subpopulation of Ps in most regions (Figure 5B). We then screened the 59 p65AD drivers that had shown P expression and found 44 with EB expression (File S2). Some of these drivers labeled EBs predominantly in one region, such as R78B06-p65AD and VT055900-p65AD, while some labeled EBs throughout the intestine, such as R10F10-p65AD (Figure 5, C–E). We also fortuitously identified two drivers (VT017650-p65AD and VT063223-p65AD) with EB expression that were not identified as expressing in Ps with P-AD or P-DBD. To complement this EB analysis, we attempted to generate EB-AD and ISC reference drivers, but they were inviable or not specific to the targeted cell type.

Identification of a split-GAL4 driver that specifically labels a novel subtype of EEs

In an attempt to identify subpopulations of ISC cells, we focused on the 15 p65AD drivers that showed GFP expression with P-DBD, but not EB-DBD. We observed that two drivers labeled a small population of small cells in the central midgut. One driver, R10F08-p65AD, restricted GFP expression to these cells when combined with P-DBD (Figure 5F). To determine the cell type identity of this cell population, we combined R10F08-p65AD with P-DBD and UAS-Stinger, and stained intestines with antibodies against HRP and Pros. Surprisingly, we found that the majority of cells that expressed UAS-Stinger were labeled by both anti-HRP, a P marker, and anti-Pros, an EE marker (Figure 5G). Although these two markers are mutually exclusive in most intestinal cells, Hung et al. (2020) recently identified a subclass of EEs with similar location and morphology that also express EE and P markers. We conclude that R10F08-p65AD in combination with P-DBD labels this novel EE subtype, demonstrating that split-GAL4 drivers can be used to identify novel cell populations with distinct molecular properties.

Split-GAL4 combinations label regional cell types and subregions

As described above, the expression of many drivers is limited to specific intestinal regions. We reexamined these regional drivers and identified 48 p65AD and 62 GAL4DBD drivers that have well-defined expression boundaries along the A–P axis (File S3). Figure 6A shows five EC drivers expressing in progressively more posterior locations within R1 and R2. It was clear from such overlapping patterns that we should be able to identify driver pairs defining smaller cell populations. Figure 6B shows that two EC drivers expressed in the posterior midgut (VT004417-p65AD and VT045598-GAL4DBD) could be combined to define a narrower band of cells with a particularly distinct anterior boundary in R4. Figures 6, C and D show other examples. Additional studies are needed to determine the functional significance of discrete cell populations defined by driver pairs, but our database of driver expression patterns (https://bdsc.indiana.edu/stocks/gal4/midgut_splitgal4.html) provides information for predicting which driver combinations might be experimentally useful for delineating new intersectional patterns.

Figure 6.

The expression of split-GAL4 drivers can display clear regional boundaries and, when combined, drivers can label small subsets of cells. (A) Intestinal expression of p65AD drivers in anterior regions demonstrated in combination with PM-DBD. Other tests showed these five drivers are expressed solely in ECs. Yellow arrowheads indicate regional boundaries. (B) VT004417-p65AD and VT045598-GAL4DBD are expressed broadly in the posterior intestine as shown in combinations with PM drivers, but they define a more restricted expression pattern when combined with each other. (C–D) Similarly, other driver pairs define discrete cell populations. UAS-6XGFP expression shown in green; DAPI staining in blue. Bar, 500 μm (A–D). Complete genotypes are listed in Table S2. AD, activation domain; DBD, DNA-binding domain; EC, enterocyte; PM, pan-midgut.

Identification of midgut enhancers

To address the lack of molecularly defined enhancers with adult midgut activity, we tested whether enhancer fragments from the split-GAL4 transgenes identified above could drive similar expression in other transgenic contexts. We cloned 10 enhancer fragments into a transformation vector encoding full-length GAL4 and saw 7 examples where the new GAL4 driver induced UAS-Stinger expression in a similar, although not always identical, pattern as a split-GAL4 driver with the same fragment combined with PM-AD or PM-DBD (Figure 7A and Table S4). Figure 7A shows that the VT043613 and GMR61H08 fragments drove almost identical expression in both systems, while GMR16G08, GMR86G08, and GMR28G08 drove expression in both systems with relatively minor regional differences. In addition, VT024642 drove expression in Ps throughout the midgut in both systems (shown in Figure 7A and quantified for the GAL4 version in Figure 7, B and C and Figure S7). While these examples showed enhancer fragments can drive similar patterns in different transgenic contexts, we saw three enhancer fragments that failed to replicate the split-GAL4 pattern in the GAL4 vector and 6 out of 11 fragments that failed when transferred to a smGFP.V5.nls vector (Table S4). Others have seen that expression patterns are not always recapitulated when enhancer fragments are transferred among vectors (Dionne et al. 2018; Chen et al. 2019). Such discrepancies may reflect regulatory interactions between transgene sequences or between transgene and adjacent genomic sequences. Regardless of the reasons for inconsistency, our analysis shows that enhancer sequences characterized using the split-GAL4 system can be used successfully to generate new transgenic tools for labeling and manipulating defined intestinal cells and highlights the value of our extensive documentation of split-GAL4 driver expression patterns.

Discussion

Here, we describe the expression of 424 split-GAL4 drivers in the Drosophila adult midgut that, altogether, represent 352 distinct fragments of regulatory sequence lying within 10 kb of 907 genes. By pairing these drivers with reference drivers specific to intestinal cell types, we describe 653 split-GAL4 combinations that identify a myriad of region-, cell type-, and sex-specific cell populations. Descriptions of these patterns are available both in File S2 and online at https://bdsc.indiana.edu/stocks/gal4/midgut_splitgal4.html, allowing the design of additional driver pairings to target particular intestinal cells, as well as the creation of new, cell-specific transgenes incorporating the enhancer sequences.

This new resource will provide tools for manipulating specific cell subtypes with more precision. We identified R10F08-p65AD as an example of a driver expressed in a novel cell population and we anticipate that split-GAL4 drivers will also identify subsets of cells based on physiological state, such as age, disease, or response to environmental conditions. Integrating split-GAL4 drivers into experiments with sophisticated molecular assays seems particularly promising. For example, one or more of the 83 region-specific, EC-specific split-GAL4 combinations described here might specifically label some of the regional EC subtypes recently identified using single-cell RNA sequencing (scRNA-seq) analysis (Hung et al. 2020). Indeed, we identified six enhancer fragments from drivers expressed in region-specific EC subpopulations that originated within 10 kb of genes expressed in single EC subtypes in the scRNA-seq analysis (Table 4), and several more fragments associated with genes expressed in a limited number of subtypes. While the relationships of driver expression, EC subtype, and specific genes need to be confirmed, these results show that split-GAL4 combinations have tremendous potential for experiments exploring the association of transcription factors and enhancers in specific intestinal cell populations.

Table 4. Drivers with enhancer fragments associated with EC subtype-specific genes.

Region-specific EC driver	EC subtype-specific gene	EC subtype
P{VT019063-GAL4.DBD}attP2	5-HT1A	pEC2
P{R61H08-GAL4.DBD}attP2	AstCC	LFC
P{VT063961-GAL4.DBD}attP2	CCKLR-17D3	LFC
P{VT020029-GAL4.DBD}attP2	hbn	pEC1
P{VT017655-GAL4.DBD}attP2	Sema2b	pEC3
P{R31E11-p65.AD}attP40	Ubx	aEC2

EC, enterocyte; LFC, large flat cell.

This work expands the number of defined enhancer sequences that label entire sets of intestinal cell types. Prior analyses had identified enhancer sequences from miranda and zfh2 that label Ps (Bardin et al. 2010; Rojas Villa et al. 2019), and here we identified two more enhancer-containing fragments labeling Ps: VT024642 from the region of Acp62F and VT004241 from the region of Akap200. All four of these chromosomal segments contain binding sites for the dorsal transcription factor (ModEncode Consortium et al. 2010), suggesting a regulatory logic that could be explored. We also identify VT004958 as a genomic fragment containing enhancer sequences active and specific to all ECs. However, in contrast to Ps and EC cells, we failed to identify regulatory sequences active in all EEs despite testing split-GAL4 and GAL4 drivers with six different genomic fragments associated with the EE-specific pros gene. Testing additional enhancer fragments and fragment combinations from pros may eventually identify a pan-EE element. Finally, we did not identify a regulatory fragment specific to either ISCs or EBs, consistent with recent sequencing results indicating that the transcriptomes of ISCs and EBs are highly similar (Dutta et al. 2015; Hung et al. 2020).

We also made observations that highlight the complexities of the split-GAL4 system. For example, 45 split-GAL4 drivers displayed expression with the PM reference drivers in the primary screen but did not promote expression with any of the cell type reference drivers in the secondary screen. In addition, we noted some drivers expressed in different patterns with the P, EE, and EC reference drivers than we expected from their expression patterns with a PM reference driver. Finally, we saw that split-GAL4 patterns can occasionally differ depending on the choice of UAS-responder. Altogether, these observations indicate that additional variables—such as genetic background, relative enhancer strength, and insertion site—may impact split-GAL4 transgene expression and should be considered when using the resources reported here.

We expect that this resource will have broad utility in experiments to label and/or genetically manipulate distinct subsets of intestinal cells. The split-GAL4 drivers used here have the disadvantage of not being compatible with popular gene expression systems employing GAL80 repression, because the drivers utilize the p65 activation domain, which, unlike the GAL4 activation domain, does not bind GAL80 (Ma and Ptashne 1987; Dionne et al. 2018). Nevertheless, as we have shown, enhancer fragments identified using the split-GAL4 system can be transferred to transgenic constructs compatible with GAL80 and to other gene expression systems for fine-tuned conditional or intersectional control. This resource can be expanded still by constructing additional split-GAL4 drivers, including ones that contain regulatory elements from genes known to be expressed in distinctive patterns in the intestine (Buchon et al. 2013; Marianes and Spradling 2013; Dutta et al. 2015; Hung et al. 2020). It should also prove valuable in future experiments combining the split-GAL4 system with other gene expression systems to target multiple cell types simultaneously. Such efforts will continue to enhance the capability of Drosophila experimental systems for studying conserved core biological processes.