A GAL80 Collection To Inhibit GAL4 Transgenes in Drosophila Olfactory Sensory Neurons

Fruit flies recognize hundreds of ecologically relevant odors and respond appropriately to them. The complexity, redundancy and interconnectedness of the olfactory machinery complicate efforts to pinpoint the functional contributions of any component neuron or receptor to behavior. Some contributions can only be elucidated in flies that carry multiple mutations and transgenes, but the production of such flies is currently labor-intensive and time-consuming. Here, we describe a set of transgenic flies that express the Saccharomyces cerevisiae GAL80 in specific olfactory sensory neurons (OrX-GAL80s). The GAL80s effectively and specifically subtract the activities of GAL4-driven transgenes that impart anatomical and physiological phenotypes. OrX-GAL80s can allow researchers to efficiently activate only one or a few types of functional neurons in an otherwise nonfunctional olfactory background. Such experiments will improve our understanding of the mechanistic connections between odorant inputs and behavioral outputs at the resolution of only a few functional neurons.

For example, all OSNs expressing Or22a will send axons to the DM2 glomerulus in the antennal lobe while all OSNs expressing Or82a will send axons to the VA6 glomerulus ( Figure 1). The stereotyped organization of OSNs and their projections is known as the olfactory sensory map (Vosshall, Wong, and Axel 2000;Couto, Alenius, and Dickson 2005;Stocker et al. 1990). The regularity of this map is a key feature that makes Drosophila olfaction such a useful model, as any aberration to the typical pattern will be apparent. The apparent simplicity of the map (Figure 1), however, obscures mechanistic complexities that are yet to be discovered, in part because necessary tools remain unavailable.
Drosophila geneticists have traditionally relied on genetic mutations or deletions to understand how complex biological systems normally work. Most alleles are recessive, so homozygotes must be bred over multiple generations. Achieving homozygosity of a mutation while also adding transgenes to the system often requires the creation of recombinant chromosomes produced after multiple generations of crossing and PCR screening. Classical genetic strategies thus limit the number and complexity of combinatorial genotypes that one can achieve. More challenging experimental questions demand more facile and versatile genetic tools.
The GAL4/UAS gene regulation system has become a defacto standard in studies of Drosophila. GAL4 is a yeast transcription activator that binds to the Upstream Activating Sequence (UAS) and induces expression of downstream genes (Giniger, Varnum, and Ptashne 1985). By driving GAL4 expression from an OR promoter, specific expression of a UAS-transgene can be obtained for any OSN subtype. An OrX-GAL4 line exists for almost every OR. This collection of GAL4 lines is a powerful toolbox since different UAS-transgenes can be introduced into a line via conventional mating. For example, human a-synuclein has been expressed in OSNs to model human Parkinson's disease (A. Y. Chen et al. 2014). Alternatively, protein expression levels can be knocked down using any specified UAS-RNAi transgene.
A variety of existing compatible effectors can be used study different aspects of neuronal communication. The UAS-Kir2.1 effector is used as an example in experiments described below. This inward rectifier potassium channel electrically inactivates the neurons that express it (Hodge 2009;Baines et al. 2001;Johns et al. 1999). Similarly, shibire ts or tetanus toxin can be used to silence synaptic communications (van der Bliek and Meyerowitz 1991;kitamoto 2002;Kitamoto 2001;M. S. Chen et al. 1991;Sweeney et al. 1995;Baines et al. 1999), reaper/grim/ hid genes can be used to physically kill neurons by inducing their own apoptotic pathways (Song and Steller 1999;Abrams 1999), or ricin toxin can be expressed ectopically to kill neurons. Conversely, neurons can be selectively activated with trp1a or a variety of other channelrhodopsin transgenes (Boyden 2011;Pulver et al. 2009).
If GAL4 is a standard on-switch for nearly any desired transgene, GAL80 is the logical off-switch. GAL80 binds the GAL4 transcriptional activation domain, thereby preventing recruitment of RNA polymerase (Ma and Ptashne 1987). GAL80 crosses are much more convenient than classical breeding approaches ( Figure 2). In order to have a single functional OSN in an otherwise silent olfactory system, the traditional method uses an Orco null mutation (Larsson et al. 2004). In this genetic setup, Orco mutant flies are mostly anosmic, but function is restored to one OSN subset with Or-GAL4, UAS-Orco transgenes (Olsen, Bhandawat, and Wilson 2007;DasGupta and Waddell 2008;Hoare, McCrohan, and Cobb 2008;Hoare et al. 2011;Benton et al. 2006; (Figure 2a). An Orco-GAL4, UAS-effector, Or-GAL80 method can be used instead (Figure 2b). Kir2.1 is used as an example of an effector (Hodge 2009;Baines et al. 2001;Johns et al. 1999). Classical breeding strategies (Figure 2a) may look less complicated on paper than GAL80 crosses (Figure 2b) but are actually more time-consuming and limited. The Orco mutation must be homozygous. Since most Drosophila transgenes are embedded into the same two chromosomes (2 or 3) recombination and PCR screening may be required to achieve this homozygosity.
Neurons seldom operate autonomously, but rather groups of neurons coordinate within a circuit to provide an organism with perception and behavior. An investigation of the behavioral impact provided by a limited number of different functional neuronal types would require additional genes. The elaboration of genotypes (Figure 2) to restore pairs or groups of functional OSNs in a nonfunctional background normally requires generations of crosses (followed by PCR screens for desired recombinants). A GAL80 strategy can shorten this process by achieving similar results in only one or two generations with no necessary recombinant creation. Furthermore, a GAL80 strategy takes advantage of the interchangeable variety of existing UAS-transgene lines.
For example, Or22a and Or85b are co-expressed in the same sensillum. It is understood that the activity of one neuron in the sensillum affects the activity of another (Su et al. 2012). (Gao, Clandinin, and Luo 2015) showed that Or22a was sufficient to restore behavioral responses to E2-Hexenal. But, using a similar apparatus to theirs, we were unable to show any odorant restoration to Isoamyl-Acetate (IAA), another odorant for which Or22a shows a strong response. What is the minimum necessary circuit for restoring IAA perception?
We used our GAL80 system to restore both Or85b and Or22a in an otherwise null system. The two neurons together did not restore behavioral responses (data not shown). An investigator could go further, adding back more OR's that may contribute to IAA sensation, such as Or10a, which also responds strongly to IAA (Hallem and Carlson 2006).
Using the traditional method, restoration of Or22a and Or85b in an Orco null background would require the creation of at least one recombinant and at least 3 generations of crosses ( Figure 2C). And what if the researcher wanted to add a third functional neuron? Or instead of looking a behavior, wanted to look at neuroanatomy and needed to express a reporter gene? The required genotype would be prohibitive. Our GAL80 method required no recombination and only 2 generations. This flexible system could be easily adapted to add more functional neurons, an ectopic receptor, a reporter gene, or any number of transgenes. (Though admittedly, adding more GAL80 transgenes will be easier once the GAL80 construct is targeted to additional landing sites)." Here we describe a new collection of OrX-GAL80 lines designed to complement existing OrX-GAL4 lines, and demonstrate their potential utility for neuroanatomical studies of the Drosophila olfaction model.

Fly Stocks
Flies were reared on standard cornmeal/molasses food and kept at 25C with a 16 hr on/8hours off light cycle. All lines were obtained from the Indiana University Bloomington Stock Center and the Janelia Research Campus. Any recombinants made were validated with PCR.
Or22a-mcd8::GFP #52620 Gr21a-mcd8::GFP #52619 Orco 2 #23130 UAS-Orco #23145 UAS-mcd8::GFP #5130 UAS-mcd8::GFP #5137 UAS-Kir2.1 Janelia stock #3015545 UAS-Kir2.1 Janelia stock #3015298 UAS-Kir2.1::eGFP Janelia stock #BS00312 pJFRC19-13xLexAop2-IVS-myr::GFP-p10 (attP8) Janelia stock #1171146 pJFRC59-13xLexAop2-IVS-myr::GFP-p10 (attP40) Janelia stock #3015445 Figure 2 Advantages of using a GAL80 approach over a null mutation. a) Current method with available reagents. In order to examine a single type of Olfactory Sensory Neuron (OSN) without interference from other OSNs, one can use an Orco null mutant. Without Orco, ORs cannot reach the cell membrane or function properly. Orco mutants are mostly anosmic (unable to smell.) A single OR can then be restored using two transgenes, OrX-GAL4 and UAS-Orco. Or22a-GAL4 is shown here as an example. This fly may require the making and validating of one or more recombinant chromosomes, since the Orco mutation must be homozygous. In more complicated systems, e.g., restoring more than one OSN, multiple recombinants would need to be made and validated at a cost of several months of crossing. b) Using a GAL80. GAL80 is a potent GAL4 inhibitor. All olfactory neurons could be silenced using any number of transgenes in an Orco-GAL4, UAS-effector (such as UAS-Kir2.1) genotype. A single OSN subtype can then be restored using an OrX-GAL80 (such as Or22a-GAL80). This system requires no recombinant creation, and is amenable to the use of various effectors or additional transgenes without requiring recombinant construction. c) Example of a Complex Genotype. If a researcher wants to restore two neurons in an otherwise silent background, using the Orco null approach from 2A would require at least 3 generations and one recombinant creation. Using a GAL80 approach from 2B, one can create the same fly in only 2 generations with no recombination necessary. Or22a (teal) and Or85b (peach) share sensillum Ab3 and are used as an example. Receptor appearance, orientation, and heterodimerization is based on previous designs by Neuhaus et al. (2005), Benton et al. (2006), Smart et al. (2008), and (Smart et al. 2008;Neuhaus et al. 2005;Benton et al. 2006;Benton 2009).

GAL80 Creation
Primers were designed to capture the entire promoters described by (Couto, Alenius, and Dickson 2005) (see Table S1). Promoters were amplified from genomic DNA using Q5 High Fidelity PCR (NEB #M0491S) and added to entry vectors using the pENTR/D-TOPO system (Invitrogen 2012b). Recombination with the pBP-GAL80Uw-6 (Addgene #26236) destination vector was done using the LR Clonase II system (Invitrogen 2012a). To ensure no mutations, no gaps, and correct orientation, the complete promoters were sequenced in the destination vector using the sequencing primers shown in Table S2. PhiC31 site-directed transgenesis was performed by Genetivision Inc. All GAL80 transgenes were inserted at the attP2 site, because this landing site is known to produce very strong expression (Pfeiffer et al. 2010). Since the widely-used GAL4 collection was not created using the PhiC targeted-landing site system, it is not expected that expression at this site will interfere with the GAL4 genes. Furthermore, most Or-GAL4s are available on multiple chromosomes and therefore a researcher also has a choice of which chromosome to try in a crossing scheme. A single Or-LexA line was also created using the Or22apromoter entry vector and pBPnlsLexA::p65Uw (Addgene #26230).  (Wagh et al. 2006). Larval brains were collected from third instar larvae and fixed in 4% PFA. Primary antibodies: mouse a-neuroglian (Developmental Studies Hybridoma Bank. #BP104) at 1:50 and rabbit a-GFP at 1:500. Brains were washed 4x, 15min per wash while nutating with 2mL PBT. 200mL secondary antibodies in PBT were then added for 4 hr at room temperature and then 3 overnights at 4C while nutating upright. Secondary antibodies: AF568 goat a-mouse (Life Technologies #A11031) at 1:400 and AF488 goat a-rabbit (ThermoFischer #A11034) at 1:800. Tubes were protected from light at all times after secondary antibodies had been added. Brains were washed again 4x, 15min per wash while nutating with 2mL PBT. Then washed with 1xPBS and mounted using Vectashield mounting media (Vector Labs #H-1000). Confocal images were taken with Leica800 microscope.

Single Sensillum Recordings
SSRs were performed as described in Lin et al. (2015) (Lin and Potter 2015). GFP labeled ab1 and ab3 sensilla were identified using a Zeiss AxioExaminer D1 compound microscope with eGFP filter cube (FL Filter Set 38 HE GFP shift free). A glass recording electrode filled with ringers solution (7.5g of NaCl+0.35g of KCl+0.279g of CaCl 2 -2H 2 O in 1L of H 2 O) was inserted into the base of the sensillum. To test ab1 (Gr21a) response, CO 2 was delivered through a tube ending with a Pasteur pipette that was inserted for 1 sec into a hole in a plastic pipette directed at the antenna. This plastic pipette (Denville Scientific Inc, 10ml pipette) carried a purified continuous air stream (8.3 ml/s) that used a stimulus controller (Syntech) at the time of CO 2 delivery to correct for the increased air flow. To test ab3 (Or22a) response, 20 ml of E2-Hexenal or Isoamyl acetate (diluted to 1% in mineral oil) was pipetted on a piece of filter paper (1X2 cm) in a Pasteur pipette. The Pasteur pipette was then inserted into the hole of the plastic pipette that carried continuous air stream to the antenna. For odorant delivery, the stimulus controller (Syntech) was used to divert a 1 s pulse of charcoal-filtered air (5 ml/s) into the Pasteur pipette containing the odorant. Signals were acquired and analyzed using AUTOSPIKE software (USB-IDAC System; Syntech). Spikes were counted in a 500 ms window from 500 ms after CO 2 delivery and multiplied by 2 to calculate spikes/second. Then, the spikes in 1000ms before CO 2 delivery were subtracted to calculate the increase in spike rate in response to CO 2 (Dspikes/second). For each genotype, 6 flies (4-8 days old) were tested, with 1-3 sensilla tested in each fly.

Design of GAL80 Constructs
The following criteria were used to choose OR promoters for the collection. i) The ORs should be relevant to current research as shown by the number of studies that used it. ii) The ORs should represent a variety of expression patterns (larval or adult, antennae or maxillary palps, sensillary class etc.). iii) Finally, the ORs should reflect a variety of different odorant response profiles. The promoter regions were defined based largely on the work of Couto et al., (2005).
Equimolar expression of GAL4 and GAL80 is not always sufficient to effectively eliminate GAL4 activity so the pBP-GAL80uW-6 vector was used. This vector contains a modified GAL80 sequence, designed to increase the stability and expression of its gene product (Pfeiffer et al. 2010). A few OrX-GAL80s were already made with this vector and used effectively. (Gao, Clandinin, and Luo 2015) pioneered the technique by creating a limited number of OrX-GAL80s. This work is a logical extension and makes many additional OrX-GAL80s available for general use.
Testing GAL80 Efficacy and Specificity GAL80 lines were created for the following odorant receptor promoters: Or7a, Or9a, Or10a, Or13a, Or19a, Or22a, Or22b, Or33c, Or35a, Or42a, Or42b, Or43b, Or47a, Or56a, Or59b, Or59c, Or67a, Or67d, Or71a, Or82a, Orco, Or85a, Or85b, Or85c, and Gr21a. To examine GAL4 subtraction in vivo, OrX-GAL80 flies were crossed to flies with the genotype OrX-GAL4, UAS-mcd8::GFP. OSNs expressing the same OR can be identified from their specific glomerulus in the antennal lobe (Figure 1). OrX-GAL4, UAS-mcd8::GFP flies show robust expression of the membrane-bound GFP reporter gene in their respective glomeruli. However, when OrX-GAL80 is added to the genotype, GFP expression is entirely absent, indicating a robust antagonism of GAL4 activity (Figure 3). The efficacy of Or9a-GAL80 and Or22b-GAL80 could not be determined because their Or-GAL4, UAS-mcd8:: GFP controls did not show robust or reliable GFP signaling in the first place. The created Or7a-GAL80 line was not effective at subtracting GFP signal. Though these lines are not included in Figure 3, they will still be available in the Bloomington Stock Center. Several of the GAL80 lines also have expression in larvae. GAL4 subtraction was examined in larval brains using the UAS-mcd8::GFP reporter gene. In larvae, GAL80 reduced but did not eliminate GAL4 activity ( Figure S1a). Suppression may not be as strong in larvae because the GAL4 and GAL80 genes were under the control of the same promoter. It could require extra time for GAL80 to accumulate to a level that would efficiently inhibit GAL4 activity.
The OrX-GAL80 lines were checked to ensure they would not have aberrant expression in untargeted OSN subtypes. The pBP-GAL80uW-6 vector contains a Drosophila Synthetic Core Promoter (DSCP). DSCP is an effective means of using enhancer elements to drive strong expression (Pfeiffer et al. 2008), but it could also cause the GAL80s to have nonspecific or leaky expression. Therefore, a version of pBP-GAL80uW-6 was cloned with the DSCP removed. However, when the DSCP was absent, GAL80 expression was insufficient to subtract GAL4 activity ( Figure S1b). A few lines were tested to see if DSCP causes nonspecific GAL80 expression. For these lines, an OrY-GAL80 did not impede GAL4 activity of an OrX-GAL4 neuron ( Figure S1c). Due to the uneven expression in an Orco-GAL4, UAS-GFP line, it could not be determined if each OrX-GAL80 subtracts GAL4 from only one glomerulus in an otherwise fully-labeled brain, but results shown in Figure S1c give reasonable confidence that the GAL80s do not have widespread nonspecific expression. It can also be noted that the GAL80 subtraction does not interfere with reporter gene expression in a genetic system that does not use GAL4. When Or22a-GAL80 is used in conjunction with Or22a-mcd8::GFP, containing no GAL4/UAS intermediary, the GFP is still expressed ( Figure S1d). These images, showing subtraction of reporter gene expression, confirm that GAL4 activity is suppressed anatomically by the GAL80 lines.
To make sure the system also worked for neurons expressing an OR protein (in additional to a GR), SSR was also done with ab3 sensilla. Ab3 houses Or22a-expressing neurons. This receptor is known to be activated by a diverse set of odorants, including Isoamyl acetate and E2hexenal (Hallem and Carlson 2006;Hallem, Ho, and Carlson 2004;deBruyne:2001bs W. D. Jones et al. 2007Kwon et al. 2007;de Bruyne, Foster, and Carlson 2001). A transgene was necessary to visualize the neurons without interfering with the GAL4/UAS/GAL80 system, but Or22a-mcd8::GFP was insufficiently bright to identify sensilla for SSR. A new Or22a-LexA transgenic animal was therefore created using the same promoter that was used to create the Or22a-GAL80 gene (this line is also available through Bloomington). When crossed to a LexAop2-mcd8::GFP line, the ab3 sensilla showed bright fluorescence. When Or22a-LexA, LexAop2-mcd8::GFP flies were exposed to Isoamyl acetate or to E2-hexenal, their sensillar neurons showed robust responses (mean Dspikes/s = 67.86 and 27.71, N = 7 and 7 sensilla, respectively). Unlike the Gr21a neurons, Kir2.1 expression in the Or22a neurons effectively eliminated both spontaneous and odor-evoked activity. (mean Dspikes/s = 0, N = 7 sensilla, P , 0.001 for both odorants). Both activities could be restored with the addition of the Or22a-GAL80 gene (Isoamyl acetate: mean Dspikes/s =74.29, N = 7 sensilla, P , 0.001; E2-Hexenal: mean Dspikes/s =25, N = 7 sensilla, P , 0.001). Neurons showed some low-level responses to mineral oil alone, the solvent used for the odorants (Figure 4b). Only Figure 3 OR-GAL80 reagents eliminate GAL4 activity. All antennal lobes are stained with anti-nc82 (a general neuropil marker, gray) and anti-GFP (green). The orientation of each image is dorsal-up, ventral-down, lateral-right, medial-left. Scale bars indicate 20mm. Each of the brains shown has the genotype OrX-GAL4, UAS-GFP. The specific receptor promoter is given above each column. The top row in each set shows GFP expression in these lines without GAL80. Notice how each neuron's target in the antennal lobe glomeruli is expressing GFP. Each bottom row shows the brains containing an additional Or-GAL80 gene. Note how GAL80 effectively inhibits GAL4 activity, as seen by the elimination of GFP expression. The images are representative of the 5-20 brains examined per genotype. GAL4 inactivation was 100% penetrant in one day old female flies.

DISCUSSION
The collection of GAL80 lines subtracts GAL4 activity efficiently and specifically in OSNs. In anatomical studies, reporter gene expression from the GAL4/UAS system is suppressed. The tested neuronal subtypes, which were silenced with Kir2.1, have normal firing capacity restored when GAL4 is antagonized using the GAL80 lines.
In behavioral assays, using a GAL80 transgene will be more flexible than mutant lines and less cumbersome than crafting the required recombinants as the complexity of the genotype increases. Though in some special circumstances, olfactory sensory neurons have been shown to produce behaviors autonomously, this is not a widely applicable principle, and further investigation upon this principle requires better tools. For example,  used larvae in their study to restore aversion with a single functional OSN subtype, but the larval olfactory system may be fundamentally different from adults in this respect., (Bhandawat et al. 2010) also showed that single glomerular activity is sufficient to invoke a behavioral response, but that study was done using an intact and fully functional olfactory background, so some neuronal cooperation may still have occurred. (DasGupta and Waddell 2008) provided evidence that a single functional OSN subtype is sufficient to learn odor discrimination, and Gao et al. (2015) gave convincing evidence of aversive restoration in adults with only one functional OSN.
However, the extent to which the restoration of single-OSN behavior depends on the odorant and receptor used is still unknown. Only a small subset of receptors have been tested. The current models of odor coding by the olfactory system predict that a coordinated effort of many OSNs is usually required to produce a behavioral output. Paired neurons in sensilla can affect the firing dynamics of their neighbors in the periphery (Dobritsa et al. 2003;Su et al. 2012;Kazama and Wilson 2009), and downstream neurons such as interneurons and projection neurons may rely on synchronized input from multiple OSN types (Chou et al. 2010;Wilson 2011;Yaksi and Wilson 2010;Hong and Wilson 2015;Kazama, Yaksi, and Wilson 2011;Olsen, Bhandawat, and Wilson 2007;Ng et al. 2002;Acebes et al. 2011). GAL80 tools open more possibilities to combinatorially activate subsets of neurons. The hope is that additional researchers will use the reagents and validate them in their own assays.
Researchers encounter a significant technical obstacle to the understanding of olfactory function if they need to create genotypes with small groups of interacting neurons in isolation. The tools presented here facilitate the activation or deactivation of combinations of particular neurons, thereby overcoming this obstacle. The lines are available to order through Bloomington Stock Center. Figure 4 Olfactory neuron responses toward odors in Single Sensillum Recordings (SSR). In box plots on the left, each circle shows response in an individual sensillum, and filled squares indicate the means. On the right of each plot, example SSR traces are shown for each genotype. ( Ã indicates 0.01 . P . 0.005, ÃÃ indicates P , 0.001) a) Ab1C SSR responses. Ab1C neurons are visualized using the Gr21a-GFP gene. Top: Sensilla respond strongly to CO 2 , and adding Kir2.1 reduces response to CO 2. Response is restored when Gr21a-GAL80 is added. Bottom: Air was used as a control for CO 2 experiments. Air does not cause an odor-evoked neuronal response, and adding the Kir2.1 or GAL80 genes does not affect the spontaneous signaling responses. b) Ab3 SSR responses. Ab3 neurons are visualized using the Or22a-LexA and LexAop2-GFP genes. Sensilla respond to Isoamyl Acetate and to E2-Hexenal. Adding Kir2.1 eliminates both odor-evoked and spontaneous activity in these neurons . Spontaneous and odor-evoked activity is restored when Or22a-GAL80 is added. Odorants were diluted in mineral oil and neurons from the GAL80 restorative genotype did show low-level responses to mineral oil alone (bottom).

ACKNOWLEDGMENTS
Funding for this project comes from the National Science Foundation (MCB-1413062). The authors report no conflicts of interest. We thank Heather Dionne for cloning advice, Yoshi Aso for the use of his behavioral setup and Janelia FlyLight for consulting on imaging and immunostaining.