The KRÜPPEL-Like Transcription Factor DATILÓGRAFO Is Required in Specific Cholinergic Neurons for Sexual Receptivity in Drosophila Females

Female decision-making in Drosophila flies requires the expression of a transcription factor in a small number of cholinergic neurons in discrete brain regions.


Introduction
Animals are capable of a staggering array of complex behaviors and many of them rely on innate abilities to compare different scenarios and generate specific and appropriate responses. For instance, most animals can determine with ease whether the best option is to confront or retreat from a predator or opponent. Risk assessment and similar mutually exclusive behaviors are likely to rely on neural circuits that collect information, remove irrelevant and noisy information, and quickly determine a course of action.
Courtship rituals are ancient forms of communication that allow animals to identify and rank potential mates in the midst of a noisy and usually complex environment. Thus, it is not surprising that courtships usually deploy a series of displays that involve bright colors, unusual sounds, and rhythmicities. The recipients of these displays, which in many species are females, evaluate their quality and generate the mutually exclusive behaviors of accepting or rejecting courtship.
One of the most fascinating aspects of the ability to generate courtship and respond with a decision is the fact that both behaviors are largely genetically encoded; that is, animals are capable of executing them perfectly with minimal practice and no instruction every generation. Pioneering work has established clear associations between individual male courtship behaviors with specific genes and alleles in Drosophila [1], and even led to the mapping of foci in the central nervous system required to generate discrete behaviors [2][3][4][5]. However, little is known about how females interpret and integrate aspects of the male's displays and decide if and when to accept male courtship [6]. This is a longstanding question of significance not only to our understanding of the molecular mechanisms of reproductive behavior but also to any comprehensive understanding of how neural circuits generate mutually exclusive decisions.
In Drosophila, males show their interest in females by making wing displays, singing a courtship song, dispersing airborne and contact pheromones, and physically contacting them [7][8][9]. In response to these cues, receptive females slow their movement and allow the male to proceed, to finally posture themselves to allow the male to mount them for copulation. In contrast, a disinterested or unreceptive female will engage in a number of rejection behaviors, such as fleeing, kicking the male, extruding her ovipositor, and raising or curling her abdomen [7]. Early studies have shown that no single sensory modality alone determines acceptance or rejection in mature females. Instead, the likelihood of acceptance or rejection relies on different sensory modalities that individually contribute to the final behavioral output [10][11][12][13]. Genes and alleles that either enhance or inhibit female receptivity have been isolated [14][15][16][17]. Mutations in these genes provide a unique opportunity to determine the genetic contribution to cell organization and physiological responses required to generate female mate choice [18]. In addition to mutations, somatic mosaics have been employed to determine the regions of the brain underlying female behavior [18,19]. Nevertheless, critical information about the neural circuitry involved in female decisionmaking behavior and the genes that pattern these circuits is still sorely lacking. Here, we describe dati, a neural-specific transcription factor that is required for female courtship acceptance and locomotion, and use it to begin probing the nature of the circuit by which females integrate the signals they receive from courting males to reach the correct behavioral output.
Experimental crosses were raised at 25uC, unless otherwise specified. Both males and females used for mating experiments were collected as pupae and aged 3-6 d posteclosion before mating tests. All mating tests were performed at 22uC, between 1 and 4 pm EST.

Recording of Mating Behavior
Mating tests were performed in small arenas made by superimposing two sliding sheets of transparent polycarbonate containing 24 wells each (2.54 cm in diameter and 1.27 cm depth) [24]. Each well was divided in half by a thin removable sheet of plastic. Canton-S males and experimental females 3 to 6 d old were loaded into opposing sides of each chamber without anesthesia with a manual aspirator. Once all wells were loaded, the thin plastic sheet was removed and all pair matings began simultaneously. The chamber was lit from below by an Artograph LED LightPad (Artograph Inc.), and mating behavior was recorded using a Sanyo FH1-A (Sanyo Inc.) camcorder for 1 h. For each experimental group, we calculated the courtship acceptance rate, defined as the number of pairs that successfully copulated in the 1 h observation period divided by the number of pairs observed. The average Courtship Index (CI) was calculated for each experimental group. CI is defined as the fraction of time a male spent courting in a given observation period [25]. Male courtship for each pairing was observed for 10 min, starting at the onset of courtship. CIs for each pair mating in an experimental group were then aggregated into an average CI. Sample sizes are shown in the corresponding figure in results.

Quantification of Discrete Female Behaviors
Females 3 to 5 d old of experimental and control genotypes were pair-mated to Canton-S males in the mating chamber described above and video recorded for 1 h. For each pair mating, female behavior was analyzed for 10 min from the onset of courtship or until mating occurred. For this time period, every time a male initiated a step of courtship, the female reaction to courtship was recorded. The following six discrete rejection behaviors were quantified: fleeing, kicking, extruding ovipositor, jumping, flicking wings, and standing still. For each female, a Behavioral Index (BI) was obtained by calculating the frequency of each behavior displayed over the frequency of all behaviors, and these indices were averaged for each genotype.

Analysis of Locomotor Behavior
Locomotor behavior was analyzed using an adaptation of the negative geotaxis assay [26]. Five to eight flies of 3 to 5 d old were

Author Summary
Males of the fruit fly Drosophila melanogaster generate a series of courtship displays that convey visual, auditory, and olfactory information that females must decode in order to accept or reject mating. Despite the central role of female decision in sexual selection, relatively little is known about how genes and neural circuits generate this behavior. Here we show that the transcription factor datilografo (dati) is required to organize and maintain the neural circuitry required for acceptance in the central brain. Strikingly, dati is required in an excitatory circuit involving few neurons that express acetylcholine as their neurotransmitter and are located in the olfactory lobe, the first entry point for odor processing in the brain. In addition, dati is required in two other brain centers: a region where olfaction and presumably other senses are integrated and a novel region. Together these results show that a complex behavior can be generated by very few excitatory neurons, suggesting that the sharp cutoffs between acceptance and rejection may involve different thresholds of stimulation as postulated decades ago.
placed in a 15 mL Falcon tube without anesthesia and allowed to acclimate for 5 min. After this period of acclimation, the Falcon tube was inverted and rapped sharply against a fly transfer pad three times to knock flies to the bottom of the tube. The tube was then placed in front of the camcorder and flies were allowed to climb the walls. The heights reached by each of the flies after 5 s was assessed from the camcorder footage. Over 30 flies were analyzed for each experimental group. For each Falcon tube of flies, this assay was repeated for a total of five trials, spaced 30 s apart, and the heights of all flies from each trial averaged together.

Automated Image Analysis and Cell Counting
Automated cell counting was performed on confocal slices using Fiji software [28]. Briefly, a two-channel stack stained for dati (green) and Cha-Gal4 UAS-RFP.NLS (red) was converted to RGB, and the yellow overlap was segmented with white color using ''Threshold Color'' function. The blue channel containing the segmented nuclear overlaps was extracted and the noise removed by filtering the stack with the function ''Despeckle.'' Three-dimensional segmentation counts were generated by the plugin ''3D Object Counter'' [29]. Due to the large size of posterior brain stacks, they were stitched together using the plugin ''Pairwise Stitching'' before segmentation [30].

Clonal Analyses
Clonal analyses were performed using the FYT (FLP-recombinase recognition target site-yellow + -Translocation) system previously described [23]. After clone induction, third instar larvae containing GFP+ clones were handpicked, placed in a single vial, and allowed to develop up to adults 3 to 6 d old. Single females carrying clones were tested with single Canton-S male in the courtship arenas described above and video recorded for 1 h. After this time the number of couples that mated was recorded and the Courtship Indices determined. In the next day, the females that rejected males were retested with new Canton-S males for rejection, and only those that passed in the double rejection test were analyzed further [19]. Females that accepted and rejected males were referenced to specific wells and had their brains dissected. Each clone was located in a grid that divides the brain in 40 anterior and 40 posterior sectors. Because each brain may vary slightly in size or in the way it is mounted, the grid was manually stretched to find the best fit for each sample. In total, 491 clones from 83 brains were analyzed.

Olfactory Behavior Assays
In these experiments, we used a T-Maze [31] with 2 ml of benzaldehyde in one of the ends. Individual flies were loaded into the elevator of the apparatus and immediately lowered to the level containing the two ends with and without odor. After 10 s, the number of flies that moved away from the aversive odorant or towards it was recorded.

Mushroom Body and Antennal Lobe Image Analysis
Brains of different genotypes were dissected and stained for the mushroom body marker FasII and imaged as z-stacks at 1 mm intervals. Selected z-stacks containing the gamma lobe were manually segmented using the Fiji plugin ''Segmentation Editor.'' Measurement of c lobe morphological defects was done in Fiji. A similar procedure was done to segment the antennal lobe, except that the limits of the segmented structures were defined by the expression of GFP in the pattern of CHA.

Statistical Analysis
All statistical analyses were performed using MiniTab 16.1.0 (Minitab Inc.). For all comparisons of courtship acceptance rate between control and experimental groups, a 2-Proportion Test was performed, and Fisher's exact p test value was used for the determination of significance level between two groups, unless otherwise indicated. CI data were arcsine transformed prior to statistical analysis as previously described [32] and analyzed by one-way ANOVA. The difference in climbing ability in locomotor tests was analyzed by one-way ANOVA. Behavioral indices ( Figure S2) were analyzed by Mann-Whitney U Test. All other tests are two sample t tests unless otherwise noted.

Identification of dati on the Drosophila Fourth Chromosome
We previously generated a series of molecularly mapped terminal deletions on the fourth chromosome that define relatively small genomic intervals that can be used to map mutations [22,33]. These deletions were then used to map a collection of mutants available for this chromosome, to later test for locomotion and other behavioral abnormalities. One of them was l(4)102CDd 2 , an unmapped embryonic lethal mutation isolated nearly 50 years ago by Ben Hochman [20].
While mapping l(4)102CDd 2 we found that 5%-8% of the heterozygotes between this mutation and two deletions (Df(4)B6-2D and Df(4)B6-4A) escaped the lethality of l(4)102CDd 2 and exhibited a phenotype of uncoordinated movements, which becomes stronger with age (compare Movies S2 and S3). Due to the tapping of the forelegs of these genotypes, we named the mutation datilógrafo (dati) [34], which means typist in Portuguese. Subsequent analyses revealed that mutations in dati also render females completely unable to accept male courtship, as will be shown later. We located molecularly the mutation in l(4)102CDd 2 , which corresponds to a deletion that disrupts dpr7 and CG2052 plus eight other genes in between, and renamed it as deficiency on the fourth chromosome of Ben Hochman [Df(4)BH] ( Figure 1A and Text S1). Because two single fourth chromosome P-element insertions localized at the breakpoint of these deletions in the CG2052 gene (KG02689 and KG01667) exhibited the same phenotypes as homozygotes or heterozygotes for Df(4)BH, we focused our analyses on the insertion KG02689 (dati 1 ), the strongest of these two alleles. dati encodes a zinc finger transcription factor closely related to rotund (rn) and squeeze (sqz) with homologs in several species, including humans ( Figure 1B).
Consistent with its reported requirement in specifying late born neurons during embryogenesis [35], dati is specifically expressed in the central nervous system in embryos ( Figure S1A). In larval stages, dati is expressed in the brain and ventral nerve cord ( Figure  S1B) but not in other larval tissues (e.g., wing, leg, eye, and antennal discs; unpublished data). In adults, dati is broadly expressed in the brain ( Figure S1C).
dati Mutant Females Are Courted Normally But Fail to Accept Male Courtship dati 1 mutants usually stand still for long periods of time, but when courted by males, they can flee at considerable speed. In addition, when cornered by a courting male, they engage in a series of rejection behaviors that include kicking and curling their abdomen (Movies S1, S2, S3) [7,36]. To investigate how the behavior of dati 1 females departs from the wild type, we quantified six discrete behaviors normally displayed by wild-type females in response to male courtship (i.e., fleeing, kicking, extruding ovopositor, jumping, flicking wings, and standing still). dati females display all of the aforementioned behaviors but spend more time kicking and less time standing still than the wild type ( Figure S2).  (4)BH] spans a region of 10 genes, between dpr7 and dati. Indicated below Df(4)BH is the sequence of the breakpoint in dpr7 (gray) and dati (blue). (B) Neighbor-joining distance tree with Kimura twoparameter distances of dati sequences across multiple species. Branches of the tree with green termini represent orthologs of dati, whereas branches with red termini represent homologs. doi:10.1371/journal.pbio.1001964.g001 To further quantify the abnormal mating behavior of dati 1 mutant females, we compared their mating success with that of wildtype Canton-S, y w, and dati precise excision revertant females. From these data it becomes evident that the behavior of dati 1 is significantly different from the wild-type Canton-S, y w and the revertant dati F11.4 females, which exhibit normal acceptance rates ( Figure 2A). To test whether the deficit in matings was exclusively due to the female rejection, we assessed the sex appeal of dati 1 homozygous females using the CI ( Figure 2B) [25,37]. These experiments reveal that males respond to dati 1 females normally, with courtship indices indistinguishable between all four groups.  The rejection of dati mutants was tested over a longer time by measuring the frequency of females that produced progeny with a wild-type male in 6 d. The difference between the two groups is not significant (dati 1 6 d = 2/14 versus dati 1 1 h = 0/32, p = 0.08), but both are significantly different than Canton-S (dati 1 6 d versus Canton-S 6 d = 29/30, p,0.0001 and Figure 2A). This result is consistent with the fact that females that fail to accept males within 30 min are unlikely to mate afterwards [38].

dati Is Required In Neurons for Normal Acceptance and Locomotion
To determine in which tissues dati is required for normal courtship behavior and locomotion, we knocked down its expression using RNAi and UAS-dcr-2 to enhance the knockdown. The knockdown of dati with the ubiquitous Actin-Gal4 [23] at 25uC resulted in few adult individuals that died shortly after eclosion with extreme locomotor abnormalities (unpublished data).
To obtain a less severe phenotype more similar to dati 1 homozygotes, the UAS-dcr-2 construct was removed from the genotype and the flies were reared at 18uC. Under these conditions, females expressing the dati RNAi from Actin-Gal4 showed defects in acceptance and locomotion (Figure 2C,D; unpublished data). Similarly, the knockdown of dati with elav-Gal4 caused rejection and locomotor defects ( Figure 2C). elav is a bona fide postmitotic marker, except for a transient embryonic expression in glial cells and neuroblasts in thoracic and abdominal segments [39]. However, we show that the knockdown of dati in glial cells using repo-Gal4 produced no effect (Figure 2), indicating that the courtship behavioral phenotypes are not generated in these cells. In addition, we later provide evidence that the behavioral effects of dati knockdown with elav-Gal4 are not associated with neuroblasts of the embryonic ventral nerve cord.

The Removal of dati in Cholinergic Neurons Impairs Normal Female Acceptance But Not Locomotion
Because our previous results suggested that dati might be required in some capacity in neurons, we next asked whether a specific neuronal population could phenocopy the mating deficit observed. The fly brain employs several neurotransmitters including dopamine, acetylcholine, GABA, glutamate, serotonin, histamine, octopamine, and tyramine [40][41][42][43][44][45]. To begin an unbiased search for specific neuronal populations, we first knocked down the expression of dati by RNAi using four Gal4 drivers of genes involved in the synthesis of different neurotransmitters (Dopa decarboxylase, pale, Choline Acetyltransferase, and Glutamic acid decarboxylase 1) ( Figure 2E,F) to later test other neuronal types if necessary. Out of the four drivers tested, Choline Acetyltransferase Gal4 (Cha-Gal4) produced a strong and significant reduction in courtship acceptance ( Figure 2E,F). Thus, the inability of dati females to accept males affects a particular neuronal type.
Interestingly, the removal of dati in cholinergic neurons does not impair locomotion as can be observed from ''negative geotaxis'' escape response tests [26,46]. In these tests, dati 1 homozygous females normally achieve a much lower mean height 5 s after being knocked to the ground compared to wild-type Canton-S females ( Figure 2G). Revertants also have a significantly better climbing ability than dati homozygotes ( Figure 2G). However, their climbing ability was not completely restored to the levels of y w, indicating that although most of the climbing deficits can be ascribed to the mutation in dati, other genes in the genetic background contribute to the locomotor deficits observed. In contrast, the climbing abilities of dati RNAi knockdowns with the Cha-Gal4 driver were not different from wild-type Canton-S ( Figure 2G), indicating that the male rejection behavior of dati 1 mutants is separable from the locomotor deficits. The results above revealed that the acceptance deficits of the dati 1 mutant are generated in cholinergic neurons. Because the mushroom bodies in Drosophila express CHA and have been implicated in memory formation, learning, and olfactory processing, we initially tested whether this neuropile was abnormal in dati 1 mutants [47,48]. The alpha and beta lobes of dati 1 mutants appear indistinguishable from the wild-type mushroom bodies, but the gamma lobes are malformed with a generally withered appearance ( Figure 3A,B,D) and have significantly different curvature ( Figure 3E). To determine if the gamma lobe defects could be responsible for the behavioral rejection, we asked whether the knockdown of dati expression in CHA+ cells could recapitulate the morphological defect in the gamma lobe and behavioral phenotypes observed. These experiments revealed that although Cha-Gal4 UAS-dati-RNAi females reject males, the gamma lobe is not affected (Figure 3C,E). Together these experiments allowed us to conclude that although the loss of dati disrupts the gamma lobe neuropile, the focus of dati-mediated courtship acceptance lies elsewhere in the brain.

dati Is Expressed in a Large Set of Neurons But in a Small Subset of the Cholinergic Neurons
To narrow the region where dati is required for female acceptance, we asked whether DATI-and CHA-positive neurons corresponded to a smaller subset than CHA neurons. DATI is broadly expressed in a complex pattern that involves a few thousand neurons. Automated cell counts indicate that there are around 2,400 neurons of the central brain that express dati, which corresponds roughly to 6.6% of all neurons of the fly's central brain ( Figure S3) [49]. The overlap between CHA-and DATI-positive neurons is much smaller, comprising 345655.3 (mean 6 s.d) neurons of the anterior central brain (N = 5), and 1,0496134 neurons of the posterior central brain (N = 8). Based on these cell counts, DATI-and CHA-positive cells (i.e., cells that cause rejection with RNAi) correspond to a modest 4% of the total neurons in the central brain. Besides reducing the complexity of the neural circuit required for acceptance, these experiments revealed that dati is not required to determine cholinergic cell identity. Instead, dati appears to specify a subtype of neuronal identity that is presumably shared by neurons that express different neurotransmitters. To determine the brain regions that mediate acceptance, we performed a clonal analysis using a new genetic tool we developed that allows for the systematic and efficient generation of somatic clones of fourth chromosome mutants, named the FYT system ( Figure 4) [23]. In these experiments, we randomly removed dati in different positions in the brain, tested whether females accepted or rejected males, and located the position of each clone within a grid that divides the brain in 80 sectors ( Figure 5A-D). By compiling a collection of 491 clones in the brain of females that either produce acceptance or double rejection (i.e., rejection in 2 consecutive days), it becomes clear that some regions in the brain produce significant deficits in acceptance while others do not. In the anterior brain, a single statistically significant region was identified in anterior sector B2 (AntB2, p = 0.029, Figure 5A). In the posterior brain, two regions stood out as highly significant (PosA3, p = 0.004 and PosC4, p#0.001) ( Figure 5B).
The anterior region AntB2 encompasses the first focus identified for female acceptance behavior using gynandromorphs [19] and also a region populated by extensively characterized local neurons (LNs) that express Sex lethal [50,51]. The posterior region PosA3 is located in the posterior superior lateral protocerebrum (pslpr) immediately above the lateral horn. In contrast, PosC4 spans over the ventral part of the lateral horn, the edge of the posterior inferior lateral protocerebrum (pilpr), and posterior lateral protocerbrum (plpr) ( Figure 5D). Together, these results show that dati is required in discrete neurons along a known olfactory path [52,53], which involves second-order olfactory neurons and also third-order neurons located around  the lateral horn. Interestingly, the ventral lateral horn has been recently identified as the region that processes pheromones [54,55]. In contrast, PosA3 appears to be a novel focus implicated in female receptivity.

Rejection Foci Contain Few dati-Positive, Cholinergic Neurons
To narrow down the position of the neurons in each sector, we analyzed the neurons that express CHA and DATI within these regions. In the anterior brain, within the region AntB2, we can discern 13 dati Is Required to Generate a Subtype of Cholinergic Neurons Because we had observed that the removal of dati in olfactory neurons in the region AntB2 impairs female acceptance and dati is required in the specification of late born neurons [35], we expected that the mutant might fail to specify a neuronal subtype DATI CHA. To begin addressing this issue, we compared the GFP expression patterns of Cha-Gal4 in wild-type and dati 1 glomerulus DL3 (not visible from this angle), and the heart shape outlines the position of DA1 shown in (G). (E) Frontal view of a dati mutant brain, highlighting the antennal neuropile (dashed lines), the position of the glomerulus DL3 (asterisk), the glomerulus DA1 (also segmented in red to the left of DL3, but not visible from this angle in the 3D rendering), and the lateral neurons (square). Note that the antennal lobe neuropile in (E) is smaller than in (B), which indicates defects in innervation. Also note that the neurons in the square in (B) are smaller than those in (C). (F) Mutant antennal lobe in (E) viewed from the brain outwards. Asterisks indicate glomerulus DL3, red indicates glomerulus DA1, and arrows point to lateral neurons. Note that the size of DA1 in (F) is smaller than in (C) and that the lateral neurons are larger in (F) than in (C) (arrows). (G) Side view slightly from the top of another antennal lobe mutant for dati 1 . The DL3 (asterisk), DA1 (red), and lateral neurons (square) are indicated. Note that the neuropile where DA1 is indicated in (G) appears darker than other parts of the antennal lobe and also darker than the corresponding region in (D) (heart-shaped outline), indicating that this region is less dense. Also note that the density of lateral neurons in (G) (square) is smaller than in ( Figure 7A-G). These experiments revealed severe abnormalities in the cholinergic tracts of the antennal lobes ( Figure 7B-G). A closer examination reveals that the population of dorsal lateral neurons in the region AntB2 are either reduced or transformed to cholinergic neuronal types with a distinct morphology than those normally found in this region ( Figure 7C and F). These transformations within antennal lobe neurons affect several glomeruli, which include DA1, the target of the male pheromone cis-vaccenyl acetate (cVA) (Figure 7D-H) [56].
The experiments above revealed that the loss of dati disrupts olfactory glomeruli. To test whether these disruptions lead to olfactory deficits, we assayed the performance of dati mutant females in a T-Maze in which flies are tested for moving away or towards an aversive odor. In this test, only 3% of the Canton-S flies (1 out of 30) moved towards the aversive odor compared to 32% of the dati mutant females (11 out of 34), indicating that olfactory behavior is indeed impaired in dati mutants (wild type versus dati, p = 0.001).
In the lateral horn, the loss of dati leads to a reduction of approximately 10% of the lateral horn neuropile area ( Figure 8C and D; dati, N = 8; WT, N = 10) and the cholinergic projections from the antennal lobe towards the lateral horn are also affected ( Figure S4). Like in the antennal lobe, we note the presence of larger neurons in the lateral horn of dati mutants, which are not present in the wild type ( Figure 8A and B). Furthermore, there are more CHA-positive cells around the lateral horn, suggesting that in the absence of dati some neuronal precursors can proliferate to later assume a cholinergic fate or, alternatively, that in the absence of dati some cells assume a cholinergic fate ( Figure 8A-D). Together, these results show that dati is required in postmitotic neurons as well as in the precursors of these cells.

Nuclear Bar Coding Reveals That DATI CHA Neurons Mediate Short-and Long-Range Connections
From the previous experiments, we found evidence that dati specifies a subpopulation of cholinergic neurons that project into the antennal olfactory glomeruli. Olfactory neurons in the antennal lobe descend from few neuroblast lineages that generate remarkably different neurons within and across lineages [50], and it has been suggested that morphologically different neurons are dedicated to specific neurocomputations [57]. This heterogeneity has been traditionally investigated in great detail in clones of single or few neurons using Gal4 drivers that reveal discrete neuronal populations [58][59][60]. However, we are often confronted with the opposite problem, which is to estimate whether a selected neuronal population makes simple, complex, or both simple and complex connections when a discrete Gal4 driver for these neurons is not available. This distinction is important to determine whether dati intrinsically modifies cell shapes or other aspects of neuronal physiology [61]. To that end, we developed a simple system of nuclear bar coding that distinguishes different DATI CHA neurons by color. Nuclear Bar Coding (NBC) consists of labeling nuclei of neurons with small or large volumes with different colors by expressing a localized nuclear RFP (mCherry.NLS) and GFP-S65T (nuclear and cytoplasmic) under the control of a Gal4 driver (in this case Cha-Gal4). Cells expressing the two fluorescent proteins from the same promoter are expected to be produced and degraded at comparable rates and result in nuclei with an overlay of two colors ( Figure S5) [62,63]. Assuming that these two proteins are not subject to a different regulation, the overlay of two colors should vary depending on the cellular volume. In cells with long or more intricate processes, GFP-S65T should be expected to fill up the cellular processes and shift the overlay of the two signals in the cell bodies towards that of the localized nuclear fluorescence (i.e., red color from RFP). Evidence for this shift was obtained in comparisons between cells with short and long cell processes ( Figure 9A-C). Conversely, when both GFP and RFP are targeted to the nucleus, the shifts of nuclear bar coding are abolished ( Figure S5). If, to this simple bar coding, we add a third color that detects DATI-positive cells ( Figure 9D), then we can globally assess whether dati cholinergic neurons have simple or more complex projections. NBC allowed us to easily identify the descending neurons ( Figure 9A), as well as long projection neurons located immediately above the antennal lobe, known as anterior-dorsal projection neurons (adPNs; Figure 9C,D), and LNs imbedded in antennal lateral neurons ( Figure 9C-E). In addition, the NBC method reveals that the DATI CHA neurons within the region AntB2 make both short and long connections ( Figure 9C-E). Thus, we conclude that dati does not specify only one type of cell shape, like other transcription factors that specify particular neurons [61].

Discussion
dati Encodes a Conserved ZNF Transcription Factor Related to Rotund/Squeeze and ZNF384 Required for Female Decision Making and Locomotion Here we described DATI, a zinc finger transcription factor related to the Drosophila Rotund and Squeeze and the vertebrate ZNF384, one of the three genes known to be involved in acute lymphoblastic leukemia (ALL) [64,65]. A survey of the sequences related to dati suggests that it descends from a Krüppel/rotund prototype present in cnidarians (e.g., Nematostella, gb|ABAV01025004.1|). Later this prototype evolved to become the rotund-like found in nematodes (e.g., C. elegans, Lin29) and mollusks (e.g., M. galloprovincialis, gb|GAEN01018610.1|) and was inherited by both vertebrates and invertebrates. Due to its similarity with Lin29, dati was previously referred to as Dmel/Lin29. However, orthology tests show that the ortholog of the C. elegans Lin29 is rotund, not dati. The first true ortholog of dati is found in marine arthropods (e.g., Daphnia pulex, Dpdati, gb|ACJG01001740.1|), which appeared in the Cambrian some 540 Mya [66].
Like its vertebrate homolog, dati is expressed in the nervous system and required for stem cell development [35,[67][68][69]. During embryogenesis, dati is one of the last genes to be activated in a serial activation of transcription factors that determines the identity of specific neuronal lineages in the ventral nerve cord [35]. The present study shows that dati is later required to specify regions of the central brain required for appropriate female acceptance.
are lower than in short-range neurons. Compare with results of the control experiment shown in Figure S4. (C) Overall view of the central brain. Anterior-dorsal projection neurons appear as dark red (adPNs, dotted line indicated by arrow), consistent with their long projections. LNs appear as light orange/yellow (dotted line indicated by arrowhead). (D) Superimposition of the channel detecting signal for anti-DATI staining (blue) to the image in (C). Note that adPNs appear in purple, indicating that these cells with long projections also express DATI (dotted line indicated by arrowhead). (E) High magnification of AntB2 region (box in D). DATI is expressed in cells with both small and large volumes, as indicated by color bar legend shown on the right. Neuron cell types are indicated by arrows and arrowheads, as explained in color bar legend. 3D image rendering was done in Image J. doi:10.1371/journal.pbio.1001964.g009 dati mutant flies are moderately uncoordinated and almost invariably reject male courtship ( Figure S1C and Figure 2). This rejection is so intense and persistent that it does not seem to be due to the mere loss of single sensory modalities, which inhibit but do not abolish acceptance [36]. Because of this strong rejection, we expected that dati might impair either more than one path required to generate acceptance in the brain or an area in which sensory information converges. In addition, we also tested if the locomotor and decision-making defects were associated or separable.

Locomotor Defects Are Separable From the Inability to Make Decisions
The mapping of foci by clonal analyses revealed individuals with clones that exhibited rejection but not locomotor defects (unpublished data). Conversely, we also found individuals with locomotor defects that were perfectly capable of accepting courtship and mating properly (unpublished data). Further evidence that locomotion and female behavior are separable was obtained in the experiments in which dati was knocked down in neurons that express different neurotransmitters ( Figure 2). In this case, we found that none of the four drivers used (Ddc, Gad, ple, and Cha-Gal4) produced locomotor defects like those observed using either a ubiquitous driver or the neuronal driver elav-Gal4, but the removal of dati in CHA neurons resulted in strong female behavior deficits. Thus, we conclude that the locomotor defects and female acceptance map to different brain regions and distinct cells that express specific neurotransmitters.

DATI Adds an Additional Layer to the Identity of Cholinergic Neurons That Is Shared by Noncholinergic Neurons
Our results suggest dati has two roles in the nervous systemone developmental and another constitutive-both affecting female behavior. The over/underproliferation of cholinergic neurons in dati homozygotes suggests a requirement in neuronal precursors, which is consistent with the previous study that showed dati is transiently expressed in developing ganglion mother cells [35]. However, there is a requirement in neurons, as the courtship behavioral phenotype is recapitulated when dati is removed in postmitotic neurons. Further evidence for this requirement in adult neurons is the fact that dati is indeed expressed in neurons well into adulthood, and in fact, we identified a small group of neurons that only initiates expression of dati in adult neurons (unpublished data). Together these results suggest that dati may be required to maintain a neuronal identity. Because not all datipositive neurons are cholinergic, and vice versa, it is unlikely that its primary role would be to determine the expression of this neurotransmitter. The Nuclear Bar Coding analysis suggests that dati does not evidently define any specific cell morphology either. We speculate that dati specifies a type of neuronal identity that allows neurons to respond to neurotransmitters that other cholinergic neurons without dati cannot. In this scenario, it is easy to see that removing dati from mature neurons would deprive them from the appropriate receptor(s) needed to receive input from their synaptic partners, and consequently silence female receptivity. Future tests should resolve whether dati indeed regulates channels/receptors to generate courtship acceptance.
The Regions Where dati Is Required Agree with Previous Mapping and Suggest the Existence of a Core Circuit for Female Decision Making Different mutants and experimental approaches, including gynanders, spinster mosaics, mapping of cVA processing neurons, and the use of dati mosaics, here have identified some common and other distinct foci for female decision making. For instance, the first focus AntB2 that we identified maps to Sp11, the first brain region identified for female acceptance using mosaic gynandromorphs [19]. AntB2 also maps within the Spin-D site identified by mosaics of spinster [18], a gene also required for female behavior. In addition, the two other highly significant regions, PosC4 and PosA3, flank the lateral horn, and we note that the focus PosC4 co-maps with regions previously implicated in pheromonal processing in the female brain [52,54]. Notably, the lateral horn may have a larger role in sensory integration, as it receives projections from centers that process visual and mechanosensory information [52]. Thus, the picture that emerges from previous work and the present study suggests that female decision making in Drosophila is modulated by a core circuit involving the antennal lobe and the lateral horn. However, we note that there are regions with ratios of acceptance and rejection that intuitively may appear to be relevant but that failed to reach statistical significance. In particular, there are three regions in the anterior brain (AntB3, AntB4, and AntD3) and seven regions in the posterior brain (PosB3, PosB4, PosC1, PosC2, PosC3, PosD2, and PosD3). We believe that these regions are unlikely foci for female receptivity, as our sample had resolution to identify the great significance of a relatively small focus like PosA3. Also, a similar study that analyzed a larger sample of Spinster foci for female receptivity also found brain regions that did not reach statistical significance but had ratios that could be intuitively interpreted as almost significant like ours. Like us, these authors disregarded these data as significant [18].

dati's Requirement in Few Excitatory Neurons in Three Discrete Brain Foci Reveals a Simple, Yet Fundamental, Mechanism of Female Decision Making in Drosophila
Besides providing the locations where courtship acceptance decisions are generated in the brain and the type of neurotransmitter involved, our results also reveal a significant neural mechanism at play. The DATI-CHA neurons mapped in the antennal lobe correspond to a subset of extensively studied cholinergic population known as the excitatory dorsal lateral Projection Neurons (ePNs) and excitatory lateral neurons (eLNs) [70][71][72][73][74][75]. The central role of excitatory cholinergic neurons revealed by our study and the localization of a region where sensory information is integrated constitute a nearly perfect cellular and molecular representation of the ''Summation Hypothesis,'' elaborated by Manning and others several decades ago based on behavioral inference [38,76,77]. This hypothesis states that acceptance of courtship involves the convergence of multiple excitatory stimulations provided by different sensory modalities until the stimulation reaches a critical threshold point that generates acceptance [76]. Most importantly, the Summation Hypothesis predicts that the two opposite female responses (i.e., rejection or acceptance) are not the result of opposing neural activities (e.g., excitation and inhibition) but rather the result of two different levels of excitation. Until now, there was no molecular and cellular evidence in support of this prediction. In this regard, our results are in agreement with this prediction, as the absence or presence of DATI in an excitatory circuit generates either complete rejection or overwhelming acceptance, respectively.
Corroborating our results, recent findings show that pheromone processing is not subject to the inhibitory mechanisms that apply to the processing of other odors [78]. Taken altogether, our results suggest that few dozen excitatory neurons converging in as few as three brain foci make the core components to generate a mating decision in Drosophila. Given that dati-related genes are present in a wide variety of organisms, it is likely that their common ancestor had the same or a similar mechanism of female acceptance. Figure S1 Embryonic and larval expression of dati. (A) Antibody staining of a wild-type late stage embryo for DATI (green) and ELAV (magenta). Note the presence of DATI in neurons of each hemisegment of the ventral nerve cord. Image is from a single slice of a confocal image stack. (B) Antibody staining of a wild-type L3 larval brain for DATI (gray). (C) Adult female brain stained with anti-DATI (green). The images are maximum intensity projections of confocal stacks. (TIF) Figure S2 Quantification of discrete responses to male courtship displayed by dati mutant females versus wildtype females. Female response to male courtship was quantified for 10 min after initiation of courtship by wild-type males. Bars show BI of each discrete response type of control group (Canton-S females, dark bars) and experimental (dati homozygous females, light bars) (see Materials and Methods for details). dati females are capable of displaying the same array of rejection behaviors to courtship as wild-type females (i.e., fleeing, kicking, extruding ovipositor, jumping, and flicking wings). Compared to wild-type females, dati females spend more time kicking males. In contrast, dati females spend significantly less time standing still, which is considered an accepting behavior displayed by wild-type females after being courted for some time. The statistical significance of differences was evaluated by the Mann-Whitney U test (***p, 0.001; **p,0.01), and error bars represent 6SEM.  Figure S4 dati mutants exhibit defects in the trajectory of projection neurons. (A) A wild-type brain and (B) a dati 1 mutant brain viewed from the brain neuropile towards the rear surface of the brain. In both images, 3D-rendered images were superimposed to 3D segmentation of the major cholinergic tracts (magenta). The lateral horn (LH) is indicated by the dashed circle. (C and D) Isolated segmentations of the major cholinergic tracts of the brain shown in (A and B, respectively) viewed from the rear brain surface. Note the thickness and complexity of the cell projection coming from the antennal lobe in (C) (red bracket) and the thinner and ill-defined projections in dati 1 mutants (D) (red bracket). The coordinates P (posterior), A (anterior), D (dorsal), V (ventral), L (lateral), and C (central) are indicated. (TIF) Figure S5 Nuclear GFP and nuclear RFP have comparable rates of degradation. Fluorescent signals of nuclear GFP and nuclear RFP driven by the same ubiquitin promoter in the adult brain were captured and quantified. (A) A highmagnification confocal slice of the antennal lobe. The GFP and RFP pixel intensity values were collected along the white arrow. (B) Greyscale view of GFP.NLS expression from (A). (C) Greyscale view of RFP.NLS expression from (A). (D) Quantification of signals captured along the line shown in (A). Note that the levels of RFP and GFP are similar across the intensity peaks and valleys in contrast to when nuclear bar coding is performed (Figure 9). (TIF)

Supporting Information
Movie S1 Movie of a 2-d-old Canton-S female being courted by a Canton-S male. Note that although the female flees to some degree, she does so slowly and eventually stops to fully take in the males courtship display, and eventually allows him to mount. (MP4) Movie S2 Movie of a 2-d-old dati homozygous female being courted by a Canton-S male. Note that the female flees quickly from the male, and when she is caught up to by the male, she engages in rejection behaviors such as kicking. Movie S4 A 3D rendering of segmented images of the mushroom body of wild-type, Cha-Gal4 UAS-dati-RNAi and dati 1 females labeled with anti-FasII. Labels for each genotype are shown in the movie. 3D rotations showing a (red), b (blue), and c (yellow) lobes. Note that the c lobe has a hammer-like shape in wild-type and Cha-Gal4,UAS-dati-RNAi females, whereas in dati 1 mutants the c lobe has an accentuated curvature in the center. This morphological defect is quantified in Figure 3D. (MP4) Text S1 Determination of the molecular limits of l(4)102CD d2 . (TIF)