A Circuit Logic for Sexually Shared and Dimorphic Aggressive Behaviors in Drosophila

A Circuit Logic for Sexually Shared and Dimorphic Aggressive

Behaviors in

Drosophila

Hui Chiu

1,*

Eric D. Hoopfer

Maeve L. Coughlan

David J. Anderson

1,2,5,*

Division of Biology and Biological Engineering 156-29, Tianqiao and Chrissy Chen Institute for

Neuroscience

Howard Hughes Medical Institute, California Institute of Technology, Pasadena, CA USA 91125

Carleton College, 1 N. College St., Northfield, MN USA 55057

Mount Holyoke College, 50 College St., South Hadley, MA USA 01075

Lead Contact:

SUMMARY

Aggression involves both sexually monomorphic and dimorphic actions. How the brain

implements these two types of actions is poorly understood. We have identified three cell types

that regulate aggression in

Drosophila

: one type is sexually shared, and the other two are sex-

specific. Shared CAP neurons mediate aggressive approach in both sexes, whereas functionally

downstream dimorphic but homologous cell types, called MAP in males and fpC1 in females,

control dimorphic attack. These symmetric circuits underlie the divergence of male and female

aggressive behaviors, from their monomorphic appetitive/motivational to their dimorphic

consummatory phases. The strength of the monomorphic

→

dimorphic functional connection is

increased by social isolation in both sexes, suggesting that it may be a locus for isolation-

dependent enhancement of aggression. Together, these findings reveal a circuit logic for the neural

control of behaviors that include both sexually monomorphic and dimorphic actions, which may

generalize to other organisms.

Graphical Abstract

Authors for correspondence: hchiu@caltech.edu; wuwei@caltech.edu.

AUTHOR CONTRIBUTIONS

Conceptualization, D.J.A. and H.C.; Investigation, H.C. and M.L.C.; Writing - Original Draft, H.C. and D.J.A.; Writing - Review &

Editing, H.C. and D.J.A.; Resources, H.C. and E.D.H.; Funding Acquisition, D.J.A.

Publisher's Disclaimer:

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our

customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of

the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered

which could affect the content, and all legal disclaimers that apply to the journal pertain.

DECLARATION OF INTERESTS

The authors declare no competing interests.

HHS Public Access

Author manuscript

Cell

. Author manuscript; available in PMC 2022 January 21.

Published in final edited form as:

Cell

. 2021 January 21; 184(2): 507–520.e16. doi:10.1016/j.cell.2020.11.048.

Author Manuscript

In Brief

Chiu et al. uncover how a sexually dimorphic behavior is wired in the brains of male and female

flies. Looking for neuronal components controlling aggressive behaviors, they find a single type of

neuron that controls aggressive approach in both sexes connected hierarchically to different

neurons in males vs. females which control dimorphic attack responses (lunge vs. headbutt,

respectively). They then show that social isolation, known to augment aggressive behaviors in both

sexes, strengthens the connections between the monomorphic and dimorphic neurons.

INTRODUCTION

Males and females of a given species often show sex-specific differences in behavior

(reviewed in

Dulac and Kimchi, 2007

;

Manoli et al., 2013

). These dimorphic behaviors can

be roughly divided into two categories: “pure” dimorphic behaviors, in which males and

females exhibit non-overlapping motor patterns to achieve a similar goal; and “mixed”

monomorphic-dimorphic behaviors, in which certain actions are common to both sexes,

while other actions are dimorphic. Extensive research in multiple species has shown that

“pure” sexually dimorphic behaviors (e.g., mating) are controlled by sexually dimorphic

brain circuits (reviewed in

Asahina, 2018

;

Yamamoto and Koganezawa, 2013

;

Yang and

Shah, 2014

). However, much less is known about the configuration of circuits that control

and coordinate “mixed” behaviors.

Chiu et al.

Page 2

Cell

. Author manuscript; available in PMC 2022 January 21.

Author Manuscript

Aggressive behavior in

Drosophila

exhibits a mixed pattern: a monomorphic appetitive and a

dimorphic consummatory phase (

Chan and Kravitz, 2007

;

Chen et al., 2002

;

Craig, 1917

;

Hoyer et al., 2008

;

Lorenz, 1950

;

Nilsen et al., 2004

;

Vrontou et al., 2006

). Both sexes

approach the opponent during the appetitive phase and then initiate the consummatory

attack: males lunge and tussle while females headbutt (

Chan and Kravitz, 2007

;

Nilsen et al.,

2004

;

Vrontou et al., 2006

). A great deal has been learned about neural circuit nodes that

control aggression in

Drosophila

males (reviewed in

Hoopfer, 2016

;

Kravitz and Fernández,

2015

). Less work has been done on female aggression (

Deutsch et al., 2020

;

Palavicino-

maggio et al., 2019

;

Schretter et al., 2020

) and even less on the control of the monomorphic

vs. dimorphic aspects of this behavior.

Two extreme models could explain the control of this mixed mono-/dimorphic behavior by

the brain. In one model, all phases of male and female aggression are controlled by sex-

specific circuit nodes. In support of this, most identified neurons controlling male and

female aggression in flies are sex-specific, for example, Tk

FruM

, aSP2 and P1a cells in

males, as well as aIPg and pC1d neurons in females (

Asahina et al., 2014

;

Deutsch et al.,

2020

;

Hoopfer et al., 2015

;

Palavicino-maggio et al., 2019

;

Schretter et al., 2020

;

Watanabe

et al., 2017

). Furthermore, manipulation of sex-determination genes like

fruitless

transformer

switches the pattern of sex-specific fighting (

Chan and Kravitz, 2007

;

Vrontou

et al., 2006

). In an alternative model, the mono- and dimorphic phases of aggression could

be controlled by mono- and dimorphic circuit nodes, respectively. The fact that male and

female aggression can be induced by common external or internal triggers (

Lim et al., 2014

;

Ueda and Kidokoro, 2002

;

Wang et al., 2008

) is consistent with such a model. However,

sexually shared neurons that control monomorphic features of aggression have not yet been

identified.

Here we identify a pair of sexually shared neurons whose activation increases aggressive

approach towards same-sex targets in both males and females. We also identify homologous,

male- and female-specific interneurons that promote male- or female-specific attack

behavior, respectively. We show that the dimorphic neurons are functionally downstream of

the common neurons in both sexes. Moreover, we demonstrate that the functional

connectivity of this circuit motif is strengthened in isolated males and females, which are

more aggressive than group-housed flies (

Ueda and Kidokoro, 2002

;

Wang et al., 2008

Together, these data suggest a circuit logic for the neural control of a “mixed”

monomorphic-dimorphic social behavior, and identify a potential locus for experience-

dependent modulation of this behavior.

RESULTS

Fly aggression consists of sexually monomorphic and dimorphic behaviors

We first characterized aggression in wild-type flies of both sexes and developed automated

classifiers for quantifying the appetitive (approach) and the consummatory (attack) phases of

aggression (Figure 1;

Craig, 1917

;

Tinbergen, 1951

). An approach bout occurs similarly in

both sexes, and typically involves three motor elements: orientation from a distance,

advance, and contact (Figures 1A and S1A; Video S1). In contrast, attack is sexually

dimorphic: in males it includes lunging, in which the behaving fly raises its upper body with

Chiu et al.

Page 3

Cell

. Author manuscript; available in PMC 2022 January 21.

Author Manuscript

its front legs up, and slams down onto its target (

Chen et al., 2002

;

Hoyer et al., 2008

); in

females it includes headbutting which comprises a quick horizontal thrust in the direction of

the target fly, resulting in contact (Figure 1A; Video S2;

Nilsen et al., 2004

). These

monomorphic (approach) and dimorphic (lunge/headbutt) aggressive behaviors could be

reliably detected by our behavioral classifiers (Figure S1D; Table S2).

Aggression can be promoted by social isolation or by food in both males and females (

Lim

et al., 2014

;

Ueda and Kidokoro, 2002

;

Wang et al., 2008

). However, direct, quantitative

comparisons of these effects on the two sexes have not been performed. We therefore

investigated how males and females respond to food or social isolation (SI), during or prior

to aggressive encounters, respectively. Both SI and food increased aggression in both sexes,

albeit to different extents (Figures 1B and 1C). However, the group-housed females showed

a higher level of baseline aggression than their male counterparts (Figures 1Bi and 1Ci, GH).

Thus, male and female aggressiveness can be increased by similar environmental influences,

despite their dimorphic attack behavior. These influences may act in parallel on sexually

dimorphic circuits (Figure 1D, Model 1), or via a module common to both sexes (Figure 1D,

Model 2). To distinguish between these models, we next investigated the relationship

between neural circuits controlling aggression in males versus females.

Identification of sexually shared and dimorphic aggression-promoting neurons

A distinguishing feature of Model 2 is an aggression-promoting node common to both sexes

(Figure 1D, Model 2, ‘C’). We therefore searched for such neurons, by rescreening in

females Gal4 drivers identified previously in a large-scale screen for aggression-promoting

neurons in males (

Hoopfer et al., 2015

). This screen yielded a promising candidate, line

R60G08-Gal4 (R60G08 neurons, henceforth), optogenetic activation of which strongly

promoted both approach, and dimorphic attack, in both sexes (Figure S2A).

R60G08-Gal4 drives expression in roughly 80 neurons in males and 64 neurons in females

(Figure S2B). This sex difference could reflect quantitative or qualitative differences in

aggression neurons. We first confirmed that aggression is promoted by the R60G08 neurons

in the brain but not in the ventral nerve cord of both sexes (Figure S2C). To narrow down the

subset of neurons that controls aggression, we used an “enhancer bashing” strategy to further

fractionate the R60G08 population (Figure 2A) (

Hobert and Kratsios, 2019

;

Luo et al.,

2008

). To this end, we divided the 1.5-kb R60G08

cis-

regulatory module (CRM) sequence

(

Jenett et al., 2012

;

Pfeiffer et al., 2008

) into five 0.5-kb partially overlapping fragments and

generated corresponding Gal4 drivers inserted into the same genomic locus (Enhancer

bashing (Eb) 1–5 Gal4s).

This approach yielded new Gal4 drivers with sparse labeling of R60G08 neuron subsets in

the two sexes (Figures 2B and S3A). Among these, optogenetic activation of Eb5-Gal4

neurons (Eb5 neurons, henceforth) triggered robust male and female aggression (Figures

2C), suggesting that these cells may account for R60G08 neuron-induced aggression. To test

this hypothesis, we activated R60G08 neurons in the parental Gal4 driver line, while

“subtracting” Eb5 neurons using Gal80 (R60G08+/Eb5−). Such activation yielded little or

no detectable increase in male or female aggression (Figure S3B). These data indicate that

Eb5 neurons are required for R60G08 neuron-induced aggression.

Chiu et al.

Page 4

Cell

. Author manuscript; available in PMC 2022 January 21.

Author Manuscript

Eb5-Gal4 labeled a pair of neurons in each male hemi-brain and a single neuron in each

female hemi-brain (Figure 2B). A subset of these neurons expressed

dsx-Gal4

, but none

expressed FruM (Figure S3C). To further subdivide the two Eb5 neurons identified in males,

we visualized their morphology using photo-activatable GFP (PA-GFP) (

Datta et al., 2008

;

Ruta et al., 2010

). This revealed two morphologically distinct cell types. One of the cell

types resembled the Eb5 neuron labeled in females, whereas the other was dissimilar and

appeared only in males (Figures 2Di–ii and 2Ei). We then generated intersectional drivers

that separately labeled these two classes of Eb5 neurons in males, using R22F05 CRM

(Figures 2Diii-v, 2Eiii, S3D, S3E, and S4Ai;

Hoopfer et al., 2015

). We refer to these cells as

CAP (

Common

Aggression-

Promoting neurons) and MAP neurons (

Male-specific

Aggression-

Promoting neurons), respectively. Both the CAP and MAP drivers labeled

neurons exclusively in the brain (Figure S4B).

CAP neurons promote the approach phase of aggression in both sexes

We first investigated the function of the shared CAP neurons in males and females. We

optogenetically activated CAP neurons and examined the effect on interactions between

pairs of freely moving group-housed (GH) flies of both sexes. When the CAP neurons were

activated, both males and females initiated significantly more approach bouts during the

stimulation period (Figure 3Ai). We next asked whether CAP neurons evoke generic

approach towards any object, and whether approach was biased towards same- or opposite-

sex flies. To this end, we developed an approach preference assay, in which we provided

pairs of different “target objects” in the chamber, and asked whether the GH tester fly

preferentially approached one of the targets during CAP activation. Initially, we gave each

tester fly a choice between a same-sex dead fly or a fly-sized object (magnet; Figure 3Bi).

Both sexes of tester flies exhibited a higher percentage of approach bouts directed toward the

fly target (Figure 3Bi). This result suggests that stimulation of CAP neurons preferentially

promotes approach to conspecifics, relative to an inanimate object.

We then investigated whether the approach promoted by CAP neurons was biased towards

same- or opposite-sex conspecifics.

Drosophila

males normally do not attack females (even

if aggression-promoting neurons are activated (Figure S5A)), and female flies of other

species most frequently attack females, in competition for oviposition sites (

Fernández et al.,

2010

;

Shelly, 1999

). Consistent with this, during CAP stimulation male testers preferred

male over female targets, whereas female testers showed the opposite tendency (Figure 3Bii;

Video S3). The tester’s target preference was not affected by the duration of its interaction

with the targets prior to CAP stimulation (Figure S5B). More importantly, if both targets

were female, no difference in target preference was observed between CAP-stimulated male

vs. female testers (Figure 3Biii). Furthermore, wing extension, a male courtship behavior

(

Bennet-Clark and Ewing, 1967

;

von Philipsborn et al., 2011

) that was induced naturally by

the presence of a female (alive or dead), was not promoted by CAP activation (Figure 3Aiii).

These data therefore indicate that CAP neurons promote approach towards sex-appropriate

targets of aggressive behavior.

Lastly, we examined how aggression was affected in the two sexes when the CAP neurons

were silenced. To do this, we enhanced natural aggressiveness using social isolation, and

Chiu et al.

Page 5

Cell

. Author manuscript; available in PMC 2022 January 21.

Author Manuscript

silenced the CAP neurons using the inwardly-rectifying potassium channel Kir2.1 (

Baines et

al., 2001

). Silencing the CAP neurons greatly decreased overall aggressiveness in both

sexes, as indicated by a significant reduction in the number of both approach and attack

bouts (Figure 3C). The locomotor activity of CAP>Kir2.1 flies was comparable to that of

CAP>GFP controls (Figure S5C), suggesting that the effect to inhibit aggression is not due

to a general reduction in vigor. Conditional silencing with the optogenetic inhibitor, GtACR

(

Mohammad et al., 2017

) or the temperature-dependent inhibitory effector, Shibire

, was

not feasible due to adverse effects of green light or the non-permissive temperature on

aggression in control flies. Together, these data indicate that although activation of CAP

neurons only promotes approach, the activity of these cells is required for both the approach

and attack phases of male and female aggression. This further supports the idea that CAP

neurons promote aggressive rather than generic approach.

CAP neurons exhibit sex differences in aggression-promoting thresholds

As part of our optogenetic protocol, we performed a stimulation titration experiment, in

which we activated the CAP neurons in males and females with five different LED

intensities titrated from 0.1 to 0.62 μW/mm

(Figure 4A). Unexpectedly, this experiment

revealed sex differences in the threshold for CAP stimulation-induced aggression. In GH

females, approach was elicited at low photostimulation intensities (0.1 μW/mm

), while

headbutting was elicited at higher intensities (0.43 μW/mm

; Figures 4Ai

➀

and 4Aiii

③

By contrast, in GH males lunging behavior was not evoked, even at the highest light

stimulation intensity tested (Figure 4Aiv). Furthermore, in females the threshold intensity

for eliciting approach (0.1 μW/mm

; Figure 4Ai

➀

) was substantially (4-fold) lower than in

males (0.43 μW/mm

; Figure 4Aii

③

We asked what neural mechanism(s) might underlie the sex differences in the effects of CAP

stimulation. We hypothesized that in females, CAP neurons might be intrinsically more

excitable than in males. To test this idea, we performed an

in vivo

all-optical stimulation and

imaging experiment to quantify the responses of CAP neurons to optogenetic activation in

each sex (Figure 4Bi). When activated using photostimulation at the same frequency and

intensity, female CAP neurons showed a slightly higher GCaMP fluorescence increase than

did male CAP neurons (Figure 4Bi;

=0.034, but non-significant after the Bonferroni

correction for multiple comparisons). Importantly, the fluorescence intensities of both

Chrimson::tdTomato and GCaMP in the CAP neurons were statistically indistinguishable

between sexes (Figure 4Bii). While this sex difference in the excitability is consistent with

the observed difference in the threshold for CAP-evoked aggression, other mechanisms are

likely involved.

Together, our titration experiments indicated that in GH females, CAP neurons can evoke

approach and attack (headbutting), in a scalable manner. In contrast, CAP neurons only

evoked approach behavior in GH males. One explanation for this difference is that in males,

CAP neurons do not connect with neurons that control attack. Alternatively, they may make

such a connection, but one that is weaker than in females. To distinguish these alternatives,

we next sought to identify attack-promoting neurons in males and to determine their

functional interaction with CAP neurons.

Chiu et al.

Page 6

Cell

. Author manuscript; available in PMC 2022 January 21.

Author Manuscript

MAP neurons control the attack phase of aggression in males and are functionally

downstream of CAP neurons

First, we asked whether MAP neurons might control the attack phase of aggression in males.

Indeed, in GH males, optogenetic activation of MAP neurons strongly promoted lunging,

while activation of CAP neurons was insufficient to do so (Figure 5A). Furthermore, in

contrast to CAP neurons, activating MAP neurons did not increase the number of approach

bouts initiated during photostimulation, relative to baseline or to genetic controls (Figures

3Aii and S4Aii). Consequently, MAP-induced lunging typically occurred during

serendipitous close encounters between flies (Video S1). Consistent with these observations,

the frequency of lunge bouts, but not that of approach bouts, was significantly diminished

when MAP neurons were silenced using Kir2.1 (Figures 5B and S5C). To further confirm

these results, we activated both CAP and MAP neurons simultaneously using Chrimson,

while specifically inhibiting MAP neurons with Kir2.1 (Figure 5C). Both approach and

lunging behaviors were strongly increased when CAP and MAP neurons were co-activated

(Figure 2Ci–ii). The number of lunge bouts, and the fraction of approaches leading to

lunges, were suppressed by MAP inhibition, but not the number of approaches initiated

(Figure 5C, cf. Figure 3C). Taken together, these results suggest that the sexually shared

CAP neurons are required for the initiation of approach in both males and females, while the

male-specific MAP neurons are required for male-specific attack (lunging) (Figure 5E); both

populations may contribute to the transition from approach to attack.

Anatomical analysis indicated that CAP and MAP neurites lie in close proximity (Figure

2Bi–ii), suggesting that MAP neurons might be a downstream target of CAP neurons in

males. To test this hypothesis, we performed

in vivo

calcium imaging in MAP neurons while

optogenetically stimulating CAP neurons. Indeed, GCaMP fluorescence signals in MAP

neurons were significantly elevated during CAP photostimulation, suggesting that the former

lie functionally downstream of the latter (Figure 6Aii). However, these experiments do not

distinguish whether this functional connection is direct (monosynaptic) or indirect.

fpC1 neurons represent a functional homolog of MAP neurons in females

The foregoing results indicated that in GH males, CAP activation promotes approach, and

MAP activation in turn promotes attack. Paradoxically, in GH females, strong CAP

activation could evoke both approach and attack (Figures 4Aiii and 5D), but no MAP

neurons were labeled by the Eb5 or MAP drivers. One explanation for this paradox is that

CAP neurons in females directly control both approach and attack. Alternatively, females

may contain a sex-specific homolog of MAP neurons, which was not labeled by our Gal4

drivers, and which controls headbutting (attack). We therefore searched for MAP-like

neurons in females, using anatomical methods.

The morphology of the MAP cells resembles that of pC1 neurons, a cluster of cells labeled

by the intersection of NP2631-Gal4 and

dsx

-Flp (Figure 2E). Thermogenetic activation of

the pC1

NP2631;dsxFlp

cluster was reported to promote aggression in both males and females

(

Ishii et al., 2020

;

Koganezawa et al., 2016

). NBlast (

Costa et al., 2016

) searches performed

using MAP neurite traces as a query returned several pC1 clusters among the top hits

(Figure S4C). However, MAP neurons were not labeled by NP2631-Gal4 (Figure S4D),

Chiu et al.

Page 7

Cell

. Author manuscript; available in PMC 2022 January 21.

Author Manuscript

suggesting they might be a distinct pC1 subtype. The numbers of pC1 neurons differ in the

two sexes (53–65 and 5–10 neurons per hemisphere in males and females, respectively;

reviewed in

Asahina, 2018

), and these neurons can be further divided into morphologically

distinct subtypes (

Costa et al., 2016

;

Wang et al., 2020

). This raised the possibility that a

functional homolog of MAP neurons might exist among pC1 neurons in females.

Recently, a group of female-specific neurons intersectionally labeled by R26E01-Gal4 and

dsx

-Flp were shown to trigger intense female aggression when thermogenetically activated

(

Palavicino-maggio et al., 2019

). These cells were identified as a female-specific subtype of

pC1 neurons. We therefore labeled similar cells using a different intersectional strategy, in

which R26E01-AD and dsx-DBD hemi-drivers were used to generate a split-Gal4 (Figure

6Bi). Cells labeled by this split-Gal4 driver in females exhibited a main projection pattern

similar to that of MAP neurons, but additionally extended a short, lateral branch and a long,

ventral branch (cf. Figure 6Ai vs. 6Bi). No neurons were labeled in the male brain (Figure

S4Ei). The overall morphology of these cells resembles that of pC1d neurons, a subtype of

pC1 neurons that recently has been shown to promote female aggression (

Deutsch et al.,

2020

;

Schretter et al., 2020

;

Wang et al., 2020

). We therefore refer to these neurons as

“fpC1” cells. Unfortunately, it was not possible to establish definitively correspondence

between fpC1 and pC1d neurons, due to driver incompatibility. Because of their anatomic

similarity to MAP neurons in males, these data raised the possibility that fpC1 neurons

might be female analogs or homologs of MAP neurons.

Because previous studies of R26E01-Gal4; dsx-FLP neurons used thermogenetic activation,

and did not include analysis of approach behavior (

Palavicino-maggio et al., 2019

), we first

activated these neurons optogenetically (Figure 6C). Such activation promoted female attack

(headbutting; Figure 6Cii), but not approach behavior (Figure 6Ci). We confirmed that this

phenotype is due to activation of fpC1 neurons in the brain, but not of those in the ventral

nerve cord (Figure S4Eii). Finally, we silenced fpC1 neurons using Kir2.1. This

manipulation strongly suppressed headbutting, but did not affect approach (Figures 6D and

S5C). Thus, the behavioral phenotype of activating and silencing fpC1 neurons in females

was analogous to that of activating and silencing MAP neurons in males (Figures 3Aii, 5Aii,

and 5B).

Since MAP neurons receive excitatory input from CAP neurons in males, we hypothesized

that fpC1 neurons may likewise receive excitatory input from CAP neurons in females.

Indeed, fpC1 neurons responded strongly to CAP activation (Figure 6Bii), suggesting that

they are indeed direct or indirect synaptic targets of the latter cells. Next, we performed a

behavioral epistasis experiment, in which CAP neurons were strongly activated while

silencing fpC1 neurons with Kir2.1 (Figure 6E). In this compound genotype, CAP-induced

female aggression was suppressed. Together, these data suggest that fpC1 neurons are

functionally as well as physiologically downstream of CAP cells in females. If so, it would

suggest an analogous circuit motif controlling aggression in the two sexes, in which the

sexually shared CAP neurons target MAP neurons in males, and fpC1 neurons in females

(Figure 6F).

Chiu et al.

Page 8

Cell

. Author manuscript; available in PMC 2022 January 21.

Author Manuscript

We observed that CAP stimulation caused a relatively greater increase in ΔF/F in fpC1 than

in MAP neurons (Figures 6Aii vs. 6Bii, Stim). This difference was not due to sex differences

in Chrimson expression in CAP neurons (Figure 4Bii), or in baseline GCaMP expression in

MAP vs. fpC1 neurons (Figure S4F). One possibility is that the CAP

→

fpC1 functional

connection in females is stronger than the CAP

→

MAP connection in males. This would be

consistent with our observation that strong CAP activation in GH females evoked both

approach and attack, while in GH males it evoked only approach. Alternatively, fpC1

neurons may be intrinsically more excitable than MAP neurons.

Social isolation enhances aggressiveness by strengthening circuit connectivity

The effect of social isolation to increase aggression has been observed in many species

(

Chiara et al., 2019

;

Toth et al., 2008

;

Ueda and Kidokoro, 2002

;

Wang et al., 2008

;

Zelikowsky et al., 2018

). We thus wondered how social isolation can boost aggressiveness in

both males and females, despite their differences in aggression circuitry and behavior. Social

isolation may act to increase the excitability of CAP neuron in both sexes, the excitability of

both MAP and fpC1 neurons, or the strength of the connection between CAP

→

MAP/fpC1

neurons. To address this, we investigated the excitability and functional connectivity of these

neurons in group- (GH) vs single-housed (SH) flies.

We first compared the excitability of each of the three neuron types in GH vs. SH flies, by

co-expressing Chrimson and GCaMP in each cell, and optogenetically activating and

imaging them. In both sexes, the response amplitude of each of these neurons to direct

activation was comparable in SH versus GH flies (Figure S6). We next imaged MAP or fpC1

neurons while optogenetically stimulating CAP neurons. In both sexes, the response of

MAP/fpC1 neurons to CAP activation was significantly greater in SH than in GH flies

(Figure 7A). These results suggest that the CAP

→

MAP functional connection in males, and

the CAP

→

fpC1 functional connection in females, may be strengthened by social isolation.

To examine the behavioral consequences of this enhanced functional connectivity, we

optogenetically activated CAP neurons in SH flies of both sexes (Figures 7B and 7C), at the

lowest photostimulation intensity used for GH flies in our titration experiment (Figure 4A).

SH flies exhibited increased approach behavior in both sexes, in comparison to GH flies

(Figures 7Bi and 7Ci, GH

→

SH, ‘ChR’). More importantly, in both sexes, weak activation of

CAP neurons in SH flies triggered both approach and sex-specific attack behaviors (Figures

7Bii and 7Cii, GH

→

SH, ‘ChR’), a phenotype not observed in GH flies of either sex at this

stimulation intensity (Figures 7Bii and 7Cii, GH

→

GH, ‘ChR’). Thus, in SH males, weak

activation of CAP neurons promoted both approach and lunging, a response not observed in

GH males even at the highest photostimulation intensity tested (Figure 4Aiv). Taken

together, our

in vivo

imaging analyses and behavioral experiments support the conclusion

that social isolation elevates aggressiveness in both sexes, at least in part, by strengthening

the functional connectivity between CAP neurons and their downstream MAP/fpC1 targets.

DISCUSSION

How behaviors that exhibit sexually monomorphic, as well as dimorphic, action components

are implemented in the brain is poorly understood. Here we have identified three cell types

Chiu et al.

Page 9

Cell

. Author manuscript; available in PMC 2022 January 21.

Author Manuscript

that regulate aggression in

Drosophila

: one type is sexually shared, and the other two are

sex-specific. The shared cell type, called CAP neurons, mediates aggressive approach in

both sexes, and in turn activates the dimorphic cell types, called MAP in males and fpC1 in

females, which control dimorphic attack (Figure 7D). These mirrored circuit motifs

therefore underlie the divergence of male and female aggressive behaviors, from their

monomorphic appetitive/motivational to their dimorphic consummatory phases. This circuit

logic may generalize to other behaviors and organisms.

Dissociable neural control of aggressive approach vs attack

The results presented here suggest that in both sexes, the approach vs. attack phases of

aggression are controlled by different neuron types. Importantly, the approach behavior

controlled by CAP neurons is directed preferentially towards same-sex conspecifics, and is

required for natural attack, arguing that it expresses aggressive motivation, and not simply

generic social investigation. However, this conclusion presents a seeming paradox: if

approach behavior is required for attack during natural aggression, how can experimental

MAP stimulation cause attack without promoting approach behavior as well?

Close examination of fighting patterns during MAP stimulation suggests an explanation for

this paradox. Tester males lunged during MAP stimulation only when the target fly was in

close proximity. Such proximity resulted from both directed approaches, and through

serendipitous close encounters, such as when a climbing target fly fell off the chamber wall

next to the tester (Video S1). The relatively small arenas (16mm in diameter) used in our

experiments increased the frequency of the latter type of events. Indeed, the majority (75%)

of lunge bouts promoted by MAP stimulation occurred following such serendipitous close

encounters. Moreover, once an initial MAP-evoked lunge occurred, tester males could

perform lunges continuously towards the target as long as it remained in proximity (Figure

S5D). In this way, MAP stimulation can promote lunging without also promoting approach.

We find that approach and attack are triggered at progressively higher stimulation intensities.

Other studies have shown that a single pair of descending neurons controls sequential male

courtship actions in a ramp-to-threshold manner (

McKellar et al., 2019

). Our observations

are consistent with such a mechanism operating to control the transition from approach to

attack, but further studies will be required to validate this hypothesis (Figure 7E). Similarly,

in mammals, optogenetic stimulation of estrogen receptor 1-positive (Esr1

) neurons in the

ventrolateral subdivision of the ventromedial hypothalamus (VMHvl) has been shown to

promote sniffing/mounting and attack at low vs. high thresholds, respectively (

Lee et al.,

2014

). However calcium imaging has revealed substantial overlap between attack- vs. sniff-

tuned VMHvl

Esr1+

neurons (

Remedios et al., 2017

). In contrast, distinct populations of

neurons in the lateral hypothalamus (LH) have been shown to be active during the appetitive

vs. consummatory phases of feeding behavior (

Jennings et al., 2015

). Whether these distinct

populations are functionally interconnected, and whether they control their respective

behaviors at different thresholds, is not yet clear.

Chiu et al.

Page 10

Cell

. Author manuscript; available in PMC 2022 January 21.

Author Manuscript

Physiological sexual dimorphisms in anatomically similar circuits

Despite the overall similarity in circuit logic between males and females, the light-intensity

threshold for optogenetically induced CAP-mediated approach is lower in females. Our data

indicate that this difference may reflect, at least in part, a higher intrinsic excitability of CAP

cells in GH females than in males. This difference may also explain why wild-type GH

females are more aggressive than wild-type GH males (Figures 1Bi and 1Ci, GH), although

further studies are required to confirm this.

In addition, the intensity threshold for attack stimulated by optogenetic activation of CAP

cells is lower in females than in males, particularly in GH flies where stimulation of CAP

neurons in males does not promote attack at all (Figure 4A). In SH flies, however, attack as

well as approach could be evoked by CAP stimulation in both sexes (Figures 7B and 7C).

One explanation for this sex difference is that in GH males, the threshold for MAP activation

by CAP cells is higher than the threshold for fpC1 activation in females. In SH males,

however, because the excitability of the CAP

→

MAP connection is enhanced, this threshold

can be reached and therefore lunging can be evoked by CAP stimulation (Figures 7Ai and

7Bii). The synaptic basis for this physiological dimorphism remains to be investigated.

Dimorphic but homologous pC1 neuron subtypes may control dimorphic aspects of

aggressive behavior

The symmetry between the CAP

→

MAP circuit in males and the CAP

→

fpC1 circuit in

females raises the question of whether MAP and fpC1 neurons are sex-specific, analogous

cell types. pC1 neurons, a large (~50-cell) cluster of

doublesex

-expressing cells labeled by

NP2631-Gal4 (

Koganezawa et al., 2016

), promote both male and female aggression (

Ishii et

al., 2020

;

Koganezawa et al., 2016

). More recent studies have suggested that a sex-specific

subpopulation of pC1 neurons, called pC1d, controls aggression in females (

Deutsch et al.,

2020

;

Palavicino-Maggio et al., 2019

;

Schretter et al., 2020

;

Wang et al., 2020

). These cells

resemble fpC1 cells (Figure 6Bi); although establishing correspondence between these pC1

subtypes is currently challenging, due to Gal4 driver incompatibility and the multiplicity of

morphologically similar cell types within a cluster.

Our data suggest that MAP neurons may represent a male-specific subclass of pC1 neurons.

Although MAP neurons are not labeled by NP2631-Gal4 (Figure S4D), the cell body

location and the morphology of these cells appear similar to that of some pC1 neurons

(Figure 2Eiv), an observation supported by NBlast analysis (Figure S4C). MAP and fpC1

neurons share their main projection patterns, but the latter have two extra branches

projecting laterally and ventrally (Figures 6Ai and 6Bi). This suggests that these two cell

types are indeed functional homologues. This point should be further clarified once an EM-

level connectome of the male fly brain is available.

Relationship to other aggression-promoting neurons

While our data provide an overall logic for the functional organization of male vs. female

aggression circuits, they leave out the details of its implementation. For example, it is not yet

clear whether CAP

→

MAP/fpC1 connections are direct or indirect, and exclusively feed-

forward or also recurrent. Our initial examination of the female EM connectome identified

Chiu et al.

Page 11

Cell

. Author manuscript; available in PMC 2022 January 21.

Author Manuscript

several candidate CAP-like neurons, some of which make direct monosynaptic connections

with fpC1-like pC1d neurons. pC1d neurons in turn connect to aIPg neurons, which also

promote female aggression; however there is extensive recurrence between pC1d and aIPg

cells (

Deutsch et al., 2020

Schretter et al., 2020

Although a male EM connectome is not yet available, several other groups of male-specific

neurons, such as the FruM

Tachykinin(Tk)-expressing neurons, have been previously

shown to play important roles in male aggression (

Asahina et al., 2014

;

Hoopfer, 2016

;

Kravitz and Fernández, 2015

). Whether and how these male-specific neurons interact with

CAP and MAP neurons to regulate male aggression remains largely unexplored. Initial

epistasis experiments suggest that Tk neurons act downstream of the Eb5 (CAP+MAP)

neurons to regulate aggression (Figure S7). In the future, it will be interesting to clarify the

implementation of connectivity between CAP, MAP, and Tk neurons once a male

connectome becomes available. Nevertheless, the functional pathways defined by CAP and

MAP/fpC1 neurons establish a conceptual framework for understanding how monomorphic

and dimorphic aggressive behaviors are regulated in the two sexes.

Dimorphic aggressive behavior and internal states

A fundamental question that emerges from the sexual dimorphism in fly aggressive behavior

(

Nilsen et al., 2004

;

Vrontou et al., 2006

) is whether the central motive state of

aggressiveness is encoded similarly or differently in the two sexes. Our data suggest that

CAP neurons might be candidates for controlling an internal state of aggressiveness that is

common to both sexes. Two lines of evidence support this idea. Firstly, CAP stimulation

promotes the appetitive phase of aggression (Figure 3B). In other systems, the appetitive

phase of goal-directed behaviors has been shown to reflect an internal state of motivation or

drive (

Gentry et al., 2019

;

Jennings et al., 2015

;

Salamone and Correa, 2012

). Secondly, in

SH flies CAP neurons are able to evoke not only approach, but also attack, in both sexes.

The ability of CAP neurons to control different phases of aggression, in a scalable manner

(Figure 7E), is reminiscent of the function of VMHvl

Esr1+

neurons in mice (

Falkner et al.,

2014

;

Falkner et al., 2016

;

Lee et al., 2014

;

Remedios et al., 2017

;

Yang et al., 2013

); these

cells have been suggested to mediate an aggressive internal state (

Anderson, 2016

;

Anderson

and Adolphs, 2015

;

Hashikawa et al., 2016

Finally, our findings raise the question of whether sexually dimorphic attack neurons,

analogous to MAN and fpC1, are present in other species. In rodents, males and females

attack different body parts of a same-sex intruder (

Blanchard et al., 1975

;

Sgoifo et al.,

1992

). This topographic difference in biting may reflect sexual dimorphisms in other aspects

of rodent aggression that have not yet been fully characterized. Whether this behavioral

dimorphism reflects underlying sex differences in the neural control of attack is also unclear.

In mice, Esr1/progesterone receptor-expressing (Esr1

/PR

neurons) glutamatergic neurons

in VMHvl have been shown to control both male and maternal aggression (

Hashikawa et al.,

2017

;

Lee et al., 2014

;

Yang et al., 2013

). However, recent single-cell RNA sequencing data

have revealed at least 7 distinct transcriptomic cell types among VMHvl

Esr1+/PR+

neurons,

some of which are male- or female-specific (

Kim et al., 2019

). Several of these cells types

are specifically activated during aggression, including one that is male-specific. Which

Chiu et al.

Page 12

Cell

. Author manuscript; available in PMC 2022 January 21.

Author Manuscript

transcriptomic cell type(s) is activated during maternal aggression is not yet clear. Thus,

whether the same or different cell types regulate male vs. female aggression in vertebrates,

as well as how different phases of aggressive behavior are regulated in each sex, remains to

be determined.

STAR METHODS

RESOURCE AVAILABILITY

Lead Contact—

Requests for resources and reagents should be addressed directly to the

Lead Contact, David J. Anderson (wuwei@caltech.edu).

Materials Availability—

The plasmids and the transgenic flies generated in this study are

available upon request. They will also be deposited to Addgene and the Bloomington stock

center, respectively. The behavioral classifiers are available to download:

https://

github.com/H-Chiu/Classifiers

Data and Code Availability—

Source data and analysis codes supporting this study are

available upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Fly strains—

Canton-S (from the lab of Dr. Martin Heisenberg) was used as the wild type.

Please see Table S1 for the full genotypes of flies used in each figure and see the Key

Resource Table for the source of these flies. Briefly, R60G08-Gal4(attp2),

pBDPGal4U(attp2), BDP-p65AD(attp40), BDP-Gal4DBD(attp2), R22F05-Gal4DBD(attp2),

20xUAS-IVSSyn21-GFP(attp2), 10xUAS-eGFP::Kir2.1(attp2), 10xUAS-IVS-Syn21-

GFP(attp2), 20xUAS-FRT-myrTopHat2-FRT-Chrimson::tdT3.1(vk5), 20xUAS-IVS-Syn21-

Chrimson::tdT3.1 (su(Hw)attp5 and vk5), 13xLexAop2-CsChrimson::tdT3.1(vk5),

13xLexAop2-Chrimson::tdT3.1(su(Hw)attp5), 13xLexAop2-eGFP::Kir2.1(attp40),

10xUAS-nls::tdTomato(vk22), 10xUAS-nls::GFP(vk40) were kindly shared by the Gerald

Rubin laboratory (HHMI Janelia Research Campus) and Barret Pfeiffer.

dsx

-Gal4,

dsx-

Flp,

dsx

-DBD were kindly shared by the Stephen Goodwin laboratory (University of Oxford,

UK). NP2631; dsx-Flp was kindly shared by the Daisuke Yamamoto laboratory (Tohoku

University, JP). R26E01-Gal4(attp2)(RRID:BDSC_60510), R26E01-p65AD(attp40)

(RRID:BDSC_75740), 20xUAS-IVS-CsChrimson-mVenus(attp2) (RRID:BDSC_55136),

20xUAS-IVS-jGCaMP7b(vk5) (RRID:BDSC_80907), 13xLexAop-IVS-jGCaMP7b(vk5)

(RRID: BDSC_80915) were obtained from the Bloomington Stock Center (Indiana

University).

Rearing conditions—

Stocks and crosses were reared at 25°C and 50% humidity and

maintained on a 12hr:12hr light:dark cycle. To keep the fly density consistent across

experiments, each cross was set up with 10–12 virgin females and 5–6 males, and was

flipped every two days. Experimental flies were collected mostly as virgins on the same day

of eclosion and reared in isolation (single-housed (SH) condition; one fly per vial), or in

groups (group-housed (GH) condition; ~20 single-sex flies per vial). For optogenetic

experiments, flies were transferred to vials containing retinal food (regular fly food mixed

Chiu et al.

Page 13

Cell

. Author manuscript; available in PMC 2022 January 21.

Author Manuscript

with all-trans-retinal to a final concentration of 0.2 mM) after collection and reared in the

dark for 5–6 days. Flies were flipped to fresh retinal food vials one day before a behavioral

test.

Construction of transgenic animals—

Several strains were generated for this study:

Eb2-Eb5-Gal4s, Eb5-p65AD, Eb5-Gal80, R22F05-Gal80, R22F05-LexADBD, and Eb5-

iLexA (improved LexA). Enhancer-bashing fragments, Eb2-Eb5, were PCR-amplified from

R60G08 sequence (

Jenett et al., 2012

;

Pfeiffer et al., 2008

) and cloned into pBPGUw

(Addgene #17575) via Gateway and the LR reaction. A detailed description of the cloning

strategy can be found in

Pfeiffer et al., 2008

. Sequences of the primers used for PCR

amplification are listed in the key resources table. The Eb5 fragment was subcloned into

pBPp65ADZpUw (Addgene #26234) and pBPGAL80Uw-6 (Addgene# 26236) to generate

Eb5-p65AD and Eb5-Gal80, respectively. To make the construct R22F05-Gal80, the

R22F05 fragment was first amplified from the genomic DNA of the reference strain (

Adams

et al., 2000

) using the primers listed in the ‘Janelia_info’ excel sheet (

Jenett et al., 2012

;

downloaded from Bloomington Stock Center webpage:

https://bdsc.indiana.edu/stocks/gal4/

gal4_janelia.html

) and then cloned into pBPGAL80Uw-6 via Gateway and the LR reaction.

To improve the strength of LexA, we fused the activating domain of Gal4, GAD, to the C-

terminus of nlsLexA::p65 (Addgene #26230) to make pBPnlsLexA::p65::GADUw. The

GAD sequence was amplified from the pBPGUw. It has been shown that the expression of a

targeted gene can be enhanced by adding additional activating domains to its transcription

activator (

Chavez et al., 2016

). The Eb5 fragment was then cloned into

pBPnlsLexA::p65::GADUw via Gateway and LR reaction to create Eb5-iLexA. All

constructs were verified by sequencing. To make the construct R22F05-LexADBD, we first

generated the pBPZpnlsLexADBD construct. A nlsLexADBD fragment was PCR-amplified

from pattB-nsyb-MKII::nlsLexADBDo (Addgene #64725) and cloned into pBPZpGal4DBD

(Addgene #26233) plasmid using KpnI and HindIII sites. The R22F05 enhancer fragment

was then cloned into pBPZpnlsLexADBD via Gateway and LR reaction.

METHODS DETAILS

Immunohistochemistry and brain registration—

Brains or ventral nerve cords of 6–8

day old flies were dissected in cold PBS and fixed in 4% paraformaldehyde for 1 hour at

4°C. After fixation, samples were washed twice with 0.05% PBST (PBS containing 0.05%

Triton X-100) for 15 minutes at room temperature (RT), and incubated in 2% PBST (PBS

containing 2% Triton X-100) at RT for 30 minutes. Samples were then blocked in 5%

normal goat serum (NGS) solution for 2 hours or overnight at 4°C. After blocking, samples

were incubated with the primary antibody solution for two days at 4°C. The dilution ratios

for the primary antibodies used in this study were 1:1000 for anti-GFP or anti-RFP and 1:20

for nc82 or anti-NCad. To obtain even labeling of the neuropils, which is critical for

subsequent brain registration, samples were transferred to freshly diluted primary antibody

solution on the second day of incubation. Afterwards, the samples were washed with 0.05%

PBST for 30 minutes at 4°C three times and then incubated with the secondary antibody

solution for two days at 4°C. All secondary anti-bodies used in this study were diluted

1:1000. Finally, the samples were washed three times with 0.05% PBST for 30 minutes at

4°C and incubated overnight in the mounting media Vectashield (Vectorlabs, Inc.) at 4°C.

Chiu et al.

Page 14

Cell

. Author manuscript; available in PMC 2022 January 21.

Author Manuscript

Image stacks were obtained using confocal microscopy (Olympus Fluoview FV1000 or

FV3000). Representative images illustrating the expression patterns of each driver were

chosen from among 6–10 dissected samples. Brain images were first registered to

JFRCtemplate2010 (

Jenett et al., 2012

) using the CMTK registration GUI (Jefferies et al.,

2007;

Masse et al., 2012

) and z-projected with maximum intensity under Fiji. To create

overlays of two registered brain images, images were pseudocolored and layered in

Photoshop.

Behavioral assays—

Details for each of the five different assays performed in this paper

are listed below. In general, all experiments were performed in a room maintained at 25°C

and 50% humidity. The behavioral chamber is a 12mm-high 16mm-diameter acrylic cylinder

with a clear top and floor. The wall and the lid of the chambers were coated with Insect-A-

Slip and silicon fluid, respectively. The floor was covered with freshly prepared apple juice

agar (2.5% (w/v) sucrose and 2.25% (w/v) agarose in apple juice) and illuminated with an

850nm backlight (SOBL-200×150–850, SmartVision Lights, Muskegon, MI). Flies were

tested on the 6

or 7

day after eclosion. Flies were introduced into the chambers by gentle

mouth pipetting and allowed to settle for at least two minutes before the tests began.

Behaviors were recorded at 30fps from the top using a Point Grey Flea3 camera with a long

pass IR filter (780 nm, Midwest Optical Systems).

Group-housed (GH) vs. single-housed (SH) assay.:

Canton-S males and females were

collected on the day of eclosion and reared in the single-housed (one per vial; SH) or group-

housed (twenty single-sex flies per vial; GH) condition. In each behavior chamber, we

paired same-sex flies from different vials to avoid the influence of prior life history, e.g. a

GH male from the first vial 1 was paired with a GH male from the second vial. Interactions

were recorded for 10 minutes.

Food vs. no food assay.:

Canton-S males and females were collected on the day of eclosion

and reared in group-housed condition until the test day (the 6

day after eclosion). Same-sex

flies were paired in the chambers with or without a freshly prepared banana chunk (~2mm

)

in the center as the food resource, where indicated.

Optogenetic stimulation assay.:

Experimental flies were group-housed and raised on

retinal food in the dark until the test day. Flies were transferred to fresh vials one day before

the experiment. A detailed description of the photostimulation setup can be found in

Inagaki

et al., 2014

. Briefly, a high-powered 655nm LED was mounted ~8cm above the chamber at a

24° angle to provide photostimulation at various intensities and frequencies. With the

exception of the stimulation titration experiments shown in Figure 4, 10 Hz and 5Hz pulsed

light with a maximum intensity = 0.62 μW/mm

were used to activate neurons in males and

females, respectively. The stimulation protocol included a 30s baseline activity recording,

two 30s stimulation blocks separated by a 30s interstimulation interval, and a 30s post-

stimulation period. Behaviors observed during experiments were present as the average from

two stimulation blocks. For the stimulation titration experiment shown in Figure 4A, the

same set of experimental flies were tested with 5 different intensities of photostimulation

(0.1–0.62 μW/mm

) using the same stimulation paradigm. The order of the intensity was

Chiu et al.

Page 15

Cell

. Author manuscript; available in PMC 2022 January 21.

Author Manuscript

randomly applied and each experiment was followed by a 5min interval to allow the

experimental flies to recover to their baseline activity. Data were collected from two

independent sets of experimental flies.

Preference assay (Figure 3B).:

The chamber used was a 12mm-high cylinder with a 16mm

diameter. The two targets were mounted at the opposite sides of the floor, ~10mm apart

using a UV glue tool kit (Bondic). For Figure 3Bi, the left and the right targets were a fly

(same-sex as the tester) and a fly-size magnet (K&J Magnetics, Pipersville, PA; Cylinder

1/16”x1/16”, D11-N52), respectively. The fly targets were 5–6-day old Canton-S virgin

males or females frozen at −80°C for 20 minutes immediately before the experiment. For

Figure 3Bii, the male and the female fly targets were mounted at the left and the right sides

of the chamber, respectively. For Figure 3Biii, the left and right targets were both females.

Testers were introduced into the chamber from the top and were allowed to freely explore

the arena and the targets for two minutes before the standard optogenetic stimulation (30s

pre-stimulation period followed by 30s 655nm photostimulation) was applied. The

preference index was calculated as follows,

preference index =

n

left

−

n

right

n

left

n

right

where

left

is the number of approach bouts toward the left target and

right

is the number of

approach bouts toward the right target.

Loss-of-function assay.:

The inwardly-rectifying potassium channel Kir2.1 was expressed

in CAP (Figure 3C), MAP (Figures 5B–C), fpC1 (Figures 6D–E), Tk (Figure S7A), and Eb5

(Figure S7B) neurons to test how silencing these neurons affects naturally- or artificially-

induced aggression. Experimental flies were single-housed for six days under a normal 12hr-

light: 12hr-dark cycle to naturally enhance aggressiveness (Figures 3C, 5B, and 6D). For

behavioral epistasis experiments where fighting was optogenetically induced experimental

flies were group-housed and reared in dark (Figures 5C, 6E, and S7).

Optogenetic activation in isolated flies experiment (Figures 7B–C).:

GH>GH flies and

GH>SH testers were collected on the same day. Both fly groups were reared in the GH

condition for the first three days. On the fourth day, the control (GH>GH) flies were

transferred to a new vial and maintained in groups whereas the testers (GH>SH) were

isolated individually; both groups were maintained for an additional 3 days. The optogenetic

stimulation experiments were done on the morning of the 7

day. Here we purposely

shortened the length of social isolation from six days (our standard procedure;

Wang et al.,

2008

) to three days, in order to avoid a ceiling effect on aggression.

Functional imaging—

Six-day old experimental flies were briefly anesthetized on ice and

head-fixed on a customized holder with the UV glue in their normal standing posture. The

top of the fly head was immersed in fly saline (103mM NaCl, 3mM KCl, 5mM N-

Tris(hydroxymethyl)methyl-2-aminoethane-sulfonic acid, 8mM trehalose, 10mM glucose,

26mM NaHCO

, 1mM NaH2PO

, 4mM MgCl2, 1.5mM CaCl2; pH7.25; 270–275mOsm)

Chiu et al.

Page 16

Cell

. Author manuscript; available in PMC 2022 January 21.

Author Manuscript

(

Hong and Wilson, 2015

). A piece of cuticle (~350 μm by 350 μm) was removed from the

posterior side of the head capsule to create an imaging window. After surgery, the

experimental fly was placed under a 0.8 numerical aperture (NA) 40x objective

(LUMPLFLN40XW, Olympus) and habituated for at least 5 minutes. The optical setup for

two-photon imaging with optogenetic activation was as described in

Inagaki et al., 2014

Briefly, imaging was performed using a custom-modified Ultima two-photon laser scanning

microscope (Bruker). 920nm ultrafast light pulses for exciting GCaMP was provided by a

Chameleon Ultra II Ti:Sapphire laser (Coherent) and the GCaMP signals were detected by

photomultiplier-tubes (Hamamatsu). Images were acquired at 256×256 pixel resolution and

2 frames per second. Chrimson activation was provided by a fiber-coupled 660nm LED

(M660F1, Thorlabs), powered by a 1-channel LED driver with pulse modulation (DC2100,

Thorlabs) and delivered through an optical fiber (200μm in diameter, 0.39NA, M75L01,

Thorlabs) placed above the imaging window at a ~45° angle and ~500μm in distance. The

stimulation paradigm included a 20s baseline, 5s photo-stimulation at the designated

frequency (5–50Hz) and intensity (0–34.6 μW), and a 35s post-stimulation period.

Labeling neurons with photoactivatable GFP—

To trace the morphologies of

different Eb5 cell classes, photoactivatable-GFP (PA-GFP;

Datta et al., 2008

;

Ruta et al.,

2010

) was expressed by Eb5-Gal4 (attp2) in both sexes. The experimental preparation was

similar to functional imaging experiments except that the photoactivation was provided by

710 nm ultrafast light pulses using the two-photon microscope. The photoactivation was first

localized to the cell body of the targeted neuron and, after the diffusion of the activated PA-

GFP, applied subsequently to the terminus of the labeled segments to extend the labeling.

Photoactivation cycles were separated by 10min intervals to allow the spread of activated

PA-GFP and the entire experiment typically lasted 1.5–2 hours for each cell. Image stacks

were taken at 1024×1024 pixel resolution. For the purpose of clarity, the labeled neuron was

traced with Simple Neurite Tracer plugins using Fiji to mask the background fluorescence.

The morphology of each cell class was confirmed with at least two biological replicates.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistics and quantification—

Statistical analysis was performed using MATLAB. All

behavioral data were compared using nonparametric tests. The n number for each

experiment is indicated in the figures and listed in Table S3, along with the statistical

method used for each comparison, the

-value, and the test scores. Briefly, Mann-Whitney

U-tests and Wilcoxon signed-rank tests were used for between and within-group

comparisons, respectively. The cutoff for significance was set as an

<0.05. The central

mark of each boxplot indicates the median and the bottom and top are the 25

and the 75

percentiles, respectively.

Fly tracking and behavior classification—

Behavioral videos were tracked with

Caltech FlyTracker to determine the positions, the velocities, and the postures of the

experimental flies in 30 fps recordings (

Eyjolfsdottir et al., 2014

). These tracking data were

used by behavioral classifiers developed with the Janelia Automatic Animal Behavior

Annotator (JAABA;

Kabra et al., 2013

) to determine the behavioral bouts. Naturally

occurring fighting between single-housed male-male or female-female wild-type pairs were

Chiu et al.

Page 17

Cell

. Author manuscript; available in PMC 2022 January 21.

Author Manuscript