Courtesy of Jurgen Berger. Used with permission.

1

Mushroom body endows organisms with a degree of free will or intelligent control over instinctive actions

- Dujardin, 1850

(Strausfeld et al., 1998) © Source Unknown. All rights reserved. This content isexcluded from our Creative Commons license. For moreinformation, see https://ocw.mit.edu/help/faq-fair-use/.

Courtesy of Cold Spring Harbor Laboratory Press. Used with permission.

2

Photos removed due to copyright restrictions. Please see the video.

3

Reprinted by permission from Macmillan Publishers Ltd: Nature.Source: Heisenberg, Martin. "Mushroom body memoir: From maps to models."Nature Reviews Neuroscience 4, no. 4 (2003): 266-275. ©2003.

4

AL

MB

20 µm

Courtesy of eLife. Used with permission.

5

poolingnoise

reductionnormalization

input

N~200/50 D ~ 35

PN

N ~ 1000/50D ~ 30

ORN

6

ORN

odor

odor

PN

firing rate (Hz)

firing rate (Hz)

D = 30

D = 35

with Sean Luo, Richard AxelOlsen, Bhandawat, Wilson 2010

Hallem & Carlson 2006

Courtesy of PNAS. Used with permission. 7

readout layer(s)

random, high-D hidden layer

poolingnoise

reductionnormalization

input

N ~ 34/21 D ~ 20

MBON

N ~ 2000/7 D ~ 1000

KC

N~200/50 D ~ 35

PN

N ~ 1000/50D ~ 30

ORN

8

Courtesy of Yoshi Aso. Used with permission.

9

Projection neurons

 1-8 cells/type

~ 50 types

~ 200 cells

Yoshi Aso, Daisuke Hattori

Courtesy of Yoshi Aso. Used with permission.10

Kenyon cells  

70-600 cells/type

~ 7 types

~ 2000 cells

Yoshi Aso, Daisuke Hattori

Courtesy of Yoshi Aso. Used with permission.

11

Yoshi Aso, Daisuke Hattori

Courtesy of Yoshi Aso. Used with permission.12

Antennal Lobe

Caron, Ruta, Abbott, Richard Axel, 2013

Reprinted by permission from Macmillan Publishers Ltd: Nature.Source: Caron, Sophie JC, Vanessa Ruta, L. F. Abbott, and Richard Axel. "Random convergence ofolfactory inputs in the Drosophila mushroom body." Nature 497, no. 7447 (2013): 113-117. © 2013.

13

glomerulus

yon

cell

enK

Reprinted by permission from Macmillan Publishers Ltd: Nature.Source: Caron, Sophie JC, Vanessa Ruta, L. F. Abbott, and Richard Axel. "Random convergence ofolfactory inputs in the Drosophila mushroom body." Nature 497, no. 7447 (2013): 113-117. © 2013.

14

reordered Kenyon cell

reordered Glomerulus

15

glomerulus

yon

cell

enK

Reprinted by permission from Macmillan Publishers Ltd: Nature.Source: Caron, Sophie JC, Vanessa Ruta, L. F. Abbott, and Richard Axel. "Random convergence ofolfactory inputs in the Drosophila mushroom body." Nature 497, no. 7447 (2013): 113-117. © 2013.

16

all tests consistent with random draws from this glomerular distribution

No

Further structure?

Supplementary Figure 1. | Distribution of glomerular connections onto KCs. a, The glomerular connections in the data sorted according to their observed frequencies display a non-uniform distribution. We observe that the frequency of connections to a given glomerulus roughly correlates with the number and size of its PN boutons in the MB calyx. b-b’’’, The VA2 glomerulus represents 1.3% of the total number of connections. Photo-labeling the VA2 glomerulus in four different flies expressing SPA-GFP under the control of the pan-neuronal promoter synaptobrevinGAL4 reveals each time a single PN forming boutons that are similar in number, size and overall location in the MB calyx. c, The DA3 glomerulus is absent in the data set and photo-labeling this glomerulus reveals one PN forming a single bouton in the MB calyx. d, The VL1 glomerulus, also absent in the data set, is innervated by 2 PNs extending small calycal boutons. e, The DA2 glomerulus represents 0.4% of the total number of connections and is innervated by 5 PNs, each of which extends only a single bouton into the MB calyx. f, The VA7l glomerulus represents 0.7% of the total number of connections and is innervated by a single PN also extending small boutons. g, The DP1m glomerulus accounts for 3.0% of the total number of connections and its associated PN forms larger and more numerous boutons than the PNs of less frequently represented glomeruli. h, Similar observations are made with the DL1 glomerulus (1 PN) representing 4.4% of the total number of connections. i, The most frequent DA1 glomerulus (5.1% of the total number of connections) is innervated by 6 PNs forming large boutons in the MB calyx. j, Similarly, the DC3 glomerulus (also forming 5.1% of the total number of connections) is innervated by 4 PNs also extending large boutons in the MB calyx. Scale bar = 10 mm.

Murthy, Fiete, Laurent 2008

Caron, Ruta, Abbott, Richard Axel 2013

Gruntman, Turner 2013

~7 connections per KC

17

0.10 1 2 3 4 5 6 7

1

10

100

1,000

10,000

00,0001

number of connections per KC

tical

pai

rsnu

mbe

r of i

den

Why 7 connections?

18

MB-Output neurons  

1-7 cells/type  

21 types

34 cells

Yoshi Aso, Daisuke HattoriCourtesy of Yoshi Aso. Used with permission.

19

γ lobe

2

3

4

20 µm

↵ lobe

� lobe

� lobe

�0 lobe

↵0 lobe

2 12 1

2

1

3 3

2

Tanaka, Tanimoto, Ito 2008

5 4 3 21

ped

© Wiley. All rights reserved. This content is excluded from our Creative Commonslicense. For more information, see https://ocw.mit.edu/help/faq-fair-use/.

20

Mushroom Body Extrinsic Neurons

Registration of 13 Gal4 lines

Yoshinori Aso, Daisuke Hattori, Yang Yu, Rebecca M Johnston, Nirmala A Iyer, Teri-TB Ngo, Heather Dionne, LF Abbott, Richard Axel, Hiromu Tanimoto, Gerald M Rubin, 2014

Courtesy of eLife. Used with permission. Source: Aso, Yoshinori, Divya Sitaraman, Toshiharu Ichinose, Karla R. Kaun, Katrin Vogt, Ghislain Belliart-Guérin, Pierre-Yves Plaçais et al. "Mushroom body output neurons encodevalence and guide memory-based action selection in Drosophila." Elife 3 (2014): e04580.

21

3

2

3

2

1

1

pedc

2

21

11

3 4 5

2

CS

KCs

calyx

pro

jection n

euro

ns

MBON transmitter

Acetylcholine

GABA

Glutamate

ourtesy of eLife. Used with permission. ource: Aso, Yoshinori, Divya Sitaraman, Toshiharu Ichinose, Karla R. Kaun, Katrin Vogt,

Ghislain Belliart-Guérin, Pierre-Yves Plaçais et al. "Mushroom body output neurons encodevalence and guide memory-based action selection in Drosophila." Elife 3 (2014): e04580.

22

↵1 ↵2 ↵3 �1 �2 ped �1 �2 �3 �4 �5 ↵01 ↵02 ↵03 �01 �02

cholinergic glutamatergic GABAergic output neurons

23

Dopaminergic neurons

Mushroom body lobe synaptic units

Neurons innervating α/β lobes

Output neurons

Yoshinori Aso, Daisuke Hattori, Yang Yu, Rebecca M Johnston, Nirmala A Iyer, Teri-TB Ngo, Heather Dionne, LF Abbott, Richard Axel, Hiromu Tanimoto, Gerald M Rubin, 2014

Courtesy of eLife. Used with permission. Source: Aso, Yoshinori, Divya Sitaraman, Toshiharu Ichinose, Karla R. Kaun, Katrin Vogt, Ghislain Belliart-Guérin, Pierre-Yves Plaçais et al. "Mushroom body output neurons encodevalence and guide memory-based action selection in Drosophila." Elife 3 (2014): e04580.

24

↵1 ↵2 ↵3 �1 �2 ped �1 �2 �3 �4 �5 ↵01 ↵02 ↵03 �01 �02

cholinergic glutamatergic GABAergic

PAM PPL1

output neurons

dopamine neurons

25

↵1 ↵2 ↵3 �1 �2 ped �1 �2 �3 �4 �5 ↵01 ↵02 ↵03 �01 �02

cholinergic glutamatergic GABAergic output neurons

dopamine neurons

26

punishment

KCs

calyx

PPL1 PAMreward

dopaminergic neurons

3

2

3

2

1

1

pedc

2

21

11

3 4 5

2

MBON transmitter

Acetylcholine

GABA

Glutamate

pro

jection n

euro

ns

Courtesy of eLife. Used with permission. Source: Aso, Yoshinori, Divya Sitaraman, Toshiharu Ichinose, Karla R. Kaun, Katrin Vogt, Ghislain Belliart-Guérin, Pierre-Yves Plaçais et al. "Mushroom body output neurons encode valence and guide memory-based action selection in Drosophila." Elife 3 (2014): e04580.

27

↵1 ↵2 ↵3 �1 �2 ped �1 �2 �3 �4 �5 ↵01 ↵02 ↵03 �01 �02

cholinergic glutamatergic GABAergic

28

↵1 ↵2 ↵3 �1 �2 ped �1 �2 �3 �4 �5 ↵01 ↵02 ↵03 �01 �02

29

↵1 ↵2 ↵3 �1 �2 ped �1 �2 �3 �4 �5 ↵01 ↵02 ↵03 �01 �02

30

↵1 ↵2 ↵3 �1 �2 ped �1 �2 �3 �4 �5 ↵01 ↵02 ↵03 �01 �02

31

↵1 ↵2 ↵3 �1 �2 ped �1 �2 �3 �4 �5 ↵01 ↵02 ↵03 �01 �02

↵/� (3) � (2) ↵0/�0 (2)

1

234

32

↵1 ↵2 ↵3 �1 �2 ped �1 �2 �3 �4 �5 ↵01 ↵02 ↵03 �01 �02

crepine, superior medial, intermediate and lateral protocerebrum, lateral horn

behaviorstate

learnedor

modulatedresponse

33

classical conditioning

o

electric shock

odor A dor B

Moshe Parnas, Andrew C. Lin, Wolf Huetteroth, Gero Miesenbock 201334

A

A

A

AUS

KC MBON

odor - CS

DAN

35

A

A

A

A

KC MBON

odor

DAN

36

Moshe Parnas, Andrew C. Lin, Wolf Huetteroth, Gero Miesenbock, 2013

intrinsic neurons of the MBs. Switching off the efferent synap-ses of KCs during testing occluded the effects of learning: thedecision bias of trained flies now followed the same distance-discrimination function as that of untrained flies (Figure 4B).Both parental control strains showed wild-type (WT) perfor-mance at the elevated temperature (Figure S4). Thus, prevent-ing the retrieval of memory in trained animals re-exposedtheir innate behavioral state. In contrast, blocking KC outputin untrained flies had no discernible behavioral consequence;the distance-discrimination functions of untrained animalswith intact and blocked MB output overlapped precisely (Fig-ure 4A). We conclude that flies use two parallel odor represen-tations in a state-dependent manner: they rely on the LH alonein the untrained state and engage the MB only after training.Failures of untrained flies to discriminate behaviorally betweenodors that are separated by small ePN distances, despitestrong and opposing preferences to each odor alone, mustreflect the coarse grain of odor representation in the LH anda lack of incentive to draw on the fine discrimination systemof the MB.

Inhibition by GABAergic PNs Enhances Innate OdorDiscriminationThe enhancer trap line Mz699-GAL4 (Lai et al., 2008; Okadaet al., 2009) labels 39.3 ± 0.5 GABA-positive PNs (mean ± SD,n = 4 hemispheres) located in a cluster at the ventral face ofthe antennal lobes (Figures 5A and 5D; Movies S1 and S2).Most of these GABAergic iPNs extend dendrites into multipleglomeruli (Lai et al., 2008; Tanaka et al., 2012) and project theiraxon via the mediolateral antennal lobe tract (mlALT, formerlythe medial antennocerebral tract or mACT) to the LH (Figures5A and 5C; Movie S3) (Lai et al., 2008; Tanaka et al., 2012). Incontrast, the vast majority of the !90 ePNs marked by GH146-GAL4 possess uniglomerular dendrites and project via themedial antennal lobe tract (mALT, formerly the inner antenno-cerebral tract or iACT) to both theMB and LH (Figure 5B) (Tanakaet al., 2012).Because iPN dendrites sample many glomerular channels,

odor-evoked iPN activity, like that of multiglomerular local neu-rons (Olsen et al., 2010), might scale with overall excitation inthe olfactory system. To test this idea, we expressed GCaMP3underMz699-GAL4 control and imaged the bundle of iPN axonsinnervating the LH as a proxy for iPN output. As expected, thetime integral of odor-evoked fluorescence changes correlatedwith two estimates of olfactory stimulus strength (Figures 5F,5G, and S5A): the sum of spike rates across the 24 characterizedORN classes (Figure S5A); and the number of active glomerularchannels, which was determined by thresholding ORN spikerates at 30 Hz (Figure 5G; see Figure S5B for a justification ofthreshold). The odor responses of iPNs were predicted moreaccurately by the number of active glomerular channels thanby the summed spike rates in these channels (Figures 5G andS5B). This result can be understood as a consequence ofshort-term depression at ORN synapses (Kazama and Wilson,2008), which clips excitation to iPNs when only a few ORNclasses are highly active but generates an effective drive whenmany ORN types fire at moderate rates.Interference with synaptic transmission from iPNs via the

expression of shits1 under Mz699-GAL4 control altered thebehavioral responses to odors in a subtle but characteristicway. Blocking iPN output preserved the sigmoid shape of thedistance-discrimination function but displaced the foot of thecurve to the right, compressing the range of distances that eli-cited a behavioral bias (Figures 6A and 6B; Table S5). Thus,iPN output facilitates the discrimination of closely related ePNactivity patterns. Inhibition had no general effect on the attrac-tiveness or repulsiveness of odors determined individuallyagainst air (Figures 6D and S2A; Table S2).However, the interpretation of this experiment is compli-

cated by the activity of the Mz699 enhancer element in agroup of 86 ± 1 neurons (mean ± SD, n = 4 hemispheres) inthe ventrolateral protocerebrum (vlpr) whose dendrites enterthe LH (Figures 5A and 5C; Movie S1). Because shits1 imposesa transmission block on all neurons in which it is expressed instoichiometric amounts (Kitamoto, 2001), we cannot ascribethe behavioral phenotype with confidence to a loss of iPNinhibition; impairment of vlpr neurons remains a viable alterna-tive. To eliminate this alternative, we manipulated the capacityto synthesize and package the transmitter GABA, which is

Figure 4. Innate versus Trained DiscriminationAbsolute decision bias scores of flies carrying mb247-LexA:lexAop-shits1

transgenes (mean ± SEM, n = 40–60 flies per data point) as functions

of Euclidean or cosine distances between ePN activity vectors. The distance-

discrimination functions obtained in Figure 2D are reproduced for reference.

(A) Innate discrimination at the restrictive temperature (32"C) when synaptic

output from KCs is blocked. Absolute decision bias scores as functions of

Euclidean (A1) or cosine (A2) distances between ePN activity vectors (mean ±

SEM, n = 30–60 flies per data point).

(B) Avoidance of the innately more aversive odor in a pair was reinforced during

a 1 min cycle of electric shock training at the permissive temperature (25"C).

After a 15 min rest interval, odor discrimination was analyzed at either the

permissive or the restrictive temperature when synaptic output from KCs is

intact (25"C, blue) or blocked (32"C, red). Absolute decision bias scores as

functions of Euclidean (B1) or cosine (B2) distances between ePN activity

vectors (mean ± SEM, n = 30–60 flies per data point).

See also Table S4 and Figure S4.

Neuron

Odor Discrimination from Population Activity

Neuron 79, 932–944, September 4, 2013 ª2013 The Authors 937

of PN activity of PN activity

KC output blocked

intrinsic neurons of the MBs. Switching off the efferent synap-ses of KCs during testing occluded the effects of learning: thedecision bias of trained flies now followed the same distance-discrimination function as that of untrained flies (Figure 4B).Both parental control strains showed wild-type (WT) perfor-mance at the elevated temperature (Figure S4). Thus, prevent-ing the retrieval of memory in trained animals re-exposedtheir innate behavioral state. In contrast, blocking KC outputin untrained flies had no discernible behavioral consequence;the distance-discrimination functions of untrained animalswith intact and blocked MB output overlapped precisely (Fig-ure 4A). We conclude that flies use two parallel odor represen-tations in a state-dependent manner: they rely on the LH alonein the untrained state and engage the MB only after training.Failures of untrained flies to discriminate behaviorally betweenodors that are separated by small ePN distances, despitestrong and opposing preferences to each odor alone, mustreflect the coarse grain of odor representation in the LH anda lack of incentive to draw on the fine discrimination systemof the MB.

Inhibition by GABAergic PNs Enhances Innate OdorDiscriminationThe enhancer trap line Mz699-GAL4 (Lai et al., 2008; Okadaet al., 2009) labels 39.3 ± 0.5 GABA-positive PNs (mean ± SD,n = 4 hemispheres) located in a cluster at the ventral face ofthe antennal lobes (Figures 5A and 5D; Movies S1 and S2).Most of these GABAergic iPNs extend dendrites into multipleglomeruli (Lai et al., 2008; Tanaka et al., 2012) and project theiraxon via the mediolateral antennal lobe tract (mlALT, formerlythe medial antennocerebral tract or mACT) to the LH (Figures5A and 5C; Movie S3) (Lai et al., 2008; Tanaka et al., 2012). Incontrast, the vast majority of the !90 ePNs marked by GH146-GAL4 possess uniglomerular dendrites and project via themedial antennal lobe tract (mALT, formerly the inner antenno-cerebral tract or iACT) to both theMB and LH (Figure 5B) (Tanakaet al., 2012).Because iPN dendrites sample many glomerular channels,

odor-evoked iPN activity, like that of multiglomerular local neu-rons (Olsen et al., 2010), might scale with overall excitation inthe olfactory system. To test this idea, we expressed GCaMP3underMz699-GAL4 control and imaged the bundle of iPN axonsinnervating the LH as a proxy for iPN output. As expected, thetime integral of odor-evoked fluorescence changes correlatedwith two estimates of olfactory stimulus strength (Figures 5F,5G, and S5A): the sum of spike rates across the 24 characterizedORN classes (Figure S5A); and the number of active glomerularchannels, which was determined by thresholding ORN spikerates at 30 Hz (Figure 5G; see Figure S5B for a justification ofthreshold). The odor responses of iPNs were predicted moreaccurately by the number of active glomerular channels thanby the summed spike rates in these channels (Figures 5G andS5B). This result can be understood as a consequence ofshort-term depression at ORN synapses (Kazama and Wilson,2008), which clips excitation to iPNs when only a few ORNclasses are highly active but generates an effective drive whenmany ORN types fire at moderate rates.Interference with synaptic transmission from iPNs via the

expression of shits1 under Mz699-GAL4 control altered thebehavioral responses to odors in a subtle but characteristicway. Blocking iPN output preserved the sigmoid shape of thedistance-discrimination function but displaced the foot of thecurve to the right, compressing the range of distances that eli-cited a behavioral bias (Figures 6A and 6B; Table S5). Thus,iPN output facilitates the discrimination of closely related ePNactivity patterns. Inhibition had no general effect on the attrac-tiveness or repulsiveness of odors determined individuallyagainst air (Figures 6D and S2A; Table S2).However, the interpretation of this experiment is compli-

cated by the activity of the Mz699 enhancer element in agroup of 86 ± 1 neurons (mean ± SD, n = 4 hemispheres) inthe ventrolateral protocerebrum (vlpr) whose dendrites enterthe LH (Figures 5A and 5C; Movie S1). Because shits1 imposesa transmission block on all neurons in which it is expressed instoichiometric amounts (Kitamoto, 2001), we cannot ascribethe behavioral phenotype with confidence to a loss of iPNinhibition; impairment of vlpr neurons remains a viable alterna-tive. To eliminate this alternative, we manipulated the capacityto synthesize and package the transmitter GABA, which is

Figure 4. Innate versus Trained DiscriminationAbsolute decision bias scores of flies carrying mb247-LexA:lexAop-shits1

transgenes (mean ± SEM, n = 40–60 flies per data point) as functions

of Euclidean or cosine distances between ePN activity vectors. The distance-

discrimination functions obtained in Figure 2D are reproduced for reference.

(A) Innate discrimination at the restrictive temperature (32"C) when synaptic

output from KCs is blocked. Absolute decision bias scores as functions of

Euclidean (A1) or cosine (A2) distances between ePN activity vectors (mean ±

SEM, n = 30–60 flies per data point).

(B) Avoidance of the innately more aversive odor in a pair was reinforced during

a 1 min cycle of electric shock training at the permissive temperature (25"C).

After a 15 min rest interval, odor discrimination was analyzed at either the

permissive or the restrictive temperature when synaptic output from KCs is

intact (25"C, blue) or blocked (32"C, red). Absolute decision bias scores as

functions of Euclidean (B1) or cosine (B2) distances between ePN activity

vectors (mean ± SEM, n = 30–60 flies per data point).

See also Table S4 and Figure S4.

Neuron

Odor Discrimination from Population Activity

Neuron 79, 932–944, September 4, 2013 ª2013 The Authors 937

Courtesy of Elsevier, Inc., http://www.sciencedirect.com. Used with permission.Source: Parnas, Moshe, Andrew C. Lin, Wolf Huetteroth, and Gero Miesenböck. "Odor discriminationin Drosophila: From neural population codes to behavior." Neuron 79, no. 5 (2013): 932-944.

37

classical conditioning

odor A odor B

electric shock

dopamine neuron activation

38

↵1 ↵2 ↵3 �1 �2 ped �1 �2 �3 �4 �5 ↵01 ↵02 ↵03 �01 �02

recognition memory

Daisuke Hattori, Richard Axel39

MBON-α'3 activity depresses upon repetitive odor stimulation

4-methylcyclohexanol (MCH)

Time (s)

Raw GCaMP Signal

PID Signal (odor)

40

MBON-α'3 depression is specific to repeated odor

-2 0 2 4

0

0.5

1

Pulses 1-3

-2 0 2 4

0

0.5

1

-2 0 2 4

0

0.5

1

Pulses 4-6 Pulses 7-9

MCH

-2 0 2 4

0

0.5

1

benzaldehyde (BEN)

ΔF/F0

-2 0 2 4

0

0.5

1

-2 0 2 4

0

0.5

1

-2 0 2 4

0

0.5

1

BEN

-2 0 2 4

0

0.5

1

3-octanol

1 s

MBON-α'3 depression is specific to repeated odor

-2 0 2 4

0

0.5

1

Pulses 1-3

-2 0 2 4

0

0.5

1

-2 0 2 4

0

0.5

1

Pulses 4-6 Pulses 7-9

MCH

-2 0 2 4

0

0.5

1

benzaldehyde (BEN)

ΔF/F0

-2 0 2 4

0

0.5

1

-2 0 2 4

0

0.5

1

-2 0 2 4

0

0.5

1

BEN

-2 0 2 4

0

0.5

1

3-octanol

1 s

41

MBON-α'3 depression is persistent

-2 0 2 4

0

0.5

1

-2 0 2 4

0

0.5

1

-2 0 2 4

0

0.5

1

-2 0 2 4

0

0.5

1

1 s

MCH (1-3) MCH (28-30) BEN (1-3) BEN (8-10)

-2 0 2 4

0

0.5

1

MCH (31-33)

~ 3 min

ΔF/F0

43

↵1 ↵2 ↵3 �1 �2 ped �1 �2 �3 �4 �5 ↵01 ↵02 ↵03 �01 �02

recognition memory

Daisuke Hattori, Richard Axel43

↵1 ↵2 ↵3 �1 �2 ped �1 �2 �3 �4 �5 ↵01 ↵02 ↵03 �01 �02

internal state

Raphael Cohn, Vanessa Ruta, 201544

Raphael Cohn, Vanessa Ruta, 2015

a2 a3 a4 a5

DANs > sytGCaMP

a

1s

a2 a3 a4 a5

b

e

Sucrose Feeding

0-0.7 42

Su

cro

se

0 21-0.2

Sh

ock

0 21-0.5

Fla

il

0.2

0.3

5HODWLYH�%DVDO�$FWLYLW\

a2 a3 a4 a5

d

Flail

Still

Correlation

&RHI¿FLHQW0 1-1

a2 a3 a4 a5

a2

a3

a4

a5

Flail

Flail

5 s

Basal DAN sytGCaMP Intensity

a2

a3

a4

a5

Flail

Still

Motion

No

rma

lize

d I

nte

nsity

0

1

0

1

1s

Electric Shock1s

0

1

0

1

a2 a3 a4 a5

a4�a5<DANs > P2X2

c

0 2-0.5

AT

P

ATP Injection1s

0

1

0

1

Still

0

1

a2 a3 a4 a5

Fla

il

0

1

0 1-0.9

Still

g

*

**

ns

MBONsDANs

KC

calyx

MBONs

DANs

KCcalyx

a2 a3 a4 a5

MBONsDANs

KC

calyx

MBONs

DANs

KCcalyx

a2 a3 a4 a5

f

Figure 4. DAN activity is coordinated by external stimuli and behavioral state. a, sytGCaMP was expressed in DANs of all γ-lobe compartments and fly was fed sugar (top). Representative heat map images and timecourse of DANs evoked by sugar feeding (middle). Triggered averages aligned to start of sucrose ingestion (bottom) (n=10 traces in 9 flies). Fluorescence in other lobes masked for clarity. b, DAN response to an electric shock ap-plied to the fly’s abdomen. Representative heat map images and timecourse of DANs evoked by electric shock (middle). Triggered averages aligned to electric shock (bottom) (n=21 traces in 11 flies). c, ATP injection activates γ4 and γ5 DANs expressing P2X2 under the R58E02 promoter. Rep-resentative heat map images and timecourse of DANs evoked by ATP injection (middle). Triggered averages aligned to ATP stimulation (bottom) (n=34 traces in 8 flies). d, Representative fluorescence traces of γ lobe DANs aligned to fly’s motion (top). Bouts of flailing are highly correlated with increased activity in γ2 and γ3 DANs and decreased activity in γ4 and γ5 DANs (bottom, n = 12 traces in 6 flies). e, Schematic and video frames representing the fly in flailing (orange) and still (pink) behavioral states. f, Representative heat map images of DAN activity in response to initiation and stopping of flailing (left). Triggered averages of DAN fluorescence in each compartment aligned to the start and stop of flailing bouts (right, n=14 traces in 6 flies). g, Integrated basal DAN activity in each compartment (n=6 flies).

DANs > sytGCaMP DANs > sytGCaMP

DANs > sytGCaMP

�2 �3 �4 �5

Courtesy of Elsevier, Inc., http://www.sciencedirect.com. Used with permission.Source: Cohn, Raphael, Ianessa Morantte, and Vanessa Ruta. "Coordinated and compartmentalizedneuromodulation shapes sensory processing in Drosophila." Cell 163, no. 7 (2015): 1742-1755.

45

Raphael Cohn, Vanessa Ruta, 2015

a2 a3 a4 a5

DANs > sytGCaMP

a2 a3 a4 a5

Sucrose Feeding

-0.7 0 2 4

Su

cro

se

0 21-0.2

Sh

ock

0 21-0.5

Fla

il

0.2

0.3

5HODWLYH�%DVDO�$FWLYLW\

a2 a3 a4 a5

d

Flail

Still

Correlation

&RHI¿FLHQW0 1-1

a2 a3 a4 a5

a2

a3

a4

a5

Flail

Flail

5 s

Basal DAN sytGCaMP Intensity

a2

a3

a4

a5

Flail

Still

Motion

Norm

alized Inte

nsity

0

1

0

1

1s

Electric Shock1s

0

1

0

1

a2 a3 a4 a5

a4�a5<DANs > P2X2

c

0 2-0.5

AT

P

ATP Injection1s

0

1

0

1

Still

0

1

a2 a3 a4 a5

Fla

il

0

1

0 1-0.9

Still

g

*

**

ns

MBONsDANs

KC

calyx

MBONs

DANs

KCcalyx

a2 a3 a4 a5

MBONsDANs

KC

calyx

MBONs

DANs

KCcalyx

a2 a3 a4 a5DANs > sytGCaMP

igure 4. DAN activity is coordinated by external stimuli and behavioral state. a, sytGCaMP was expressed in DANs of all γ-lobe compartments and fly was fed sugar (top). Representative heat map images and timecourse of DANs evoked by sugar feeding (middle). Triggered averages aligned to start of sucrose ingestion (bottom) (n=10 traces in 9 flies). Fluorescence in other lobes masked for clarity. b, DAN response to an electric shock ap-plied to the fly’s abdomen. Representative heat map images and timecourse of DANs evoked by electric shock (middle). Triggered averages aligned to electric shock (bottom) (n=21 traces in 11 flies). c, ATP injection activates γ4 and γ5 DANs expressing P2X2 under the R58E02 promoter. Rep-resentative heat map images and timecourse of DANs evoked by ATP injection (middle). Triggered averages aligned to ATP stimulation (bottom) (n=34 traces in 8 flies). d, Representative fluorescence traces of γ lobe DANs aligned to fly’s motion (top). Bouts of flailing are highly correlated with increased activity in γ2 and γ3 DANs and decreased activity in γ4 and γ5 DANs (bottom, n = 12 traces in 6 flies). e, Schematic and video frames representing the fly in flailing (orange) and still (pink) behavioral states. f, Representative heat map images of DAN activity in response to initiation and stopping of flailing (left). Triggered averages of DAN fluorescence in each compartment aligned to the start and stop of flailing bouts (right, n=14 traces in 6 flies). g, Integrated basal DAN activity in each compartment (n=6 flies).

F

�2 �3 �4 �5�2 �3 �4 �5Courtesy of Elsevier, Inc., http://www.sciencedirect.com. Used with permission.Source: Cohn, Raphael, Ianessa Morantte, and Vanessa Ruta. "Coordinated and compartmentalizedneuromodulation shapes sensory processing in Drosophila." Cell 163, no. 7 (2015): 1742-1755.

46

1 2 3

F/F0

1

2

a2 a4a3 a5

1

3

Time

Odor (In vivo) Stimulation (Ex vivo)1 2 3 4

a2 a4a3 a5

a b

d

�&K5�&K5

'$16�!����&K5

MBONsDANs

KC

calyx

MBONs

DANs

KCcalyx

a2 a3 a4 a5

a4 MBON

a4�a5<'$1V�!�3�;��&K5

c

F/F0

1

3

DANS > P2X2 a�<MBON > GC0 1 2 3

TPP

re-A

3F/

F0

1

0.5 s

TPP

ost-A

MBONsDANs

KC

calyx

MBONs

DANs

KCcalyx

a2 a3 a4 a5a2,a4 MBONs > GCaMP

MBONsDANs

KC

calyx

MBONs

DANs

KCcalyx

a2 a3 a4 a5a3,a5 MBONs > GCaMP

e

1 3 5

Figure 3. Dopaminergic modulation of KC-MBON synaptic transmission. a, Schematic showing pairs of MBONs expressing GCaMP6s used for functional imaging (left). Representative heat map of evoked fluorescence (top right) and timecourses (bottom right) of responses evoked by odor stimuli (black line) in vivo in pairs of MBONs. Timecourses of MBONs recorded simultaneously in the same trials are signifi-cantly different (n=8 for 2 vs. 4, n=11 for 3 vs. 5). b, Representative heat map of evoked fluorescence and timecourses of same MBON pairs in response to calyx stimulation in a brain explant (n=8 for each pair). c, Schematic of recording configuration in c-f. Whole-cell recordings of the γ4 MBON (green) were performed while the R58E02+ DANs (magenta) were stimulated through either Channelrhodopsin (d,f) or P2X2 (e). d, Comparison of whole cell voltage clamp recordings from γ4 MBONs in flies with (right) or without (left) ReaChR expressed in R58E02+ DANs. e, Representative whole-cell voltage clamp recording from γ4 MBON in fly expressing P2X2 under the 58E02 promoter, prior to and after ATP stimulation. f, Average γ4 MBON EPSC profiles and histogram of EPSC amplitudes with (red) and without (black) ReaChR expressed in R58E02+ DANs (n=5 ReaChR, n=6 control, p<.0005). GABA injection into the MB calyx suppresses spontaneous EPSCs measured in the γ4 MBON (bottom). g, Representative heat map fluorescence (left) and timecourses (right) of response in γ4 MBON expressing GCaMP evoked by calyx stimulation, before and after ATP injection to activate R58E02+ DANs expressing P2X2 (n=10, p<.0005, control animals data not shown n=6, p>0.2).

GABA

�&K5

�&K5

50 ms

2 pA

20 pA0 10

400

300

200

100

0

Eve

nts

f

10 pA1s

*ns ns ns

Pre-ATP 10 pADANS > P2X2

ATP Injection 0.5 s

Courtesy of Elsevier, Inc., http://www.sciencedirect.com. Used with permission.Source: Cohn, Raphael, Ianessa Morantte, and Vanessa Ruta. "Coordinated and compartmentalizedneuromodulation shapes sensory processing in Drosophila." Cell 163, no. 7 (2015): 1742-1755.

47

↵1 ↵2 ↵3 �1 �2 ped �1 �2 �3 �4 �5 ↵01 ↵02 ↵03 �01 �02

internal state affects routing

48

odor A odor B

hungry

odor A + “sweet”

Nobuhiro Yamagata, Toshiharu Ichinose, Yoshinori Aso, Pierre-Yves Plaçais, Anja B. Friedrich, Richard J. Sima, Thomas Preat, Gerald M. Rubin, Hiromu Tanimoto, 2014

internal state affects memory

49

↵1 ↵2 ↵3 �1 �2 ped �1 �2 �3 �4 �5 ↵01 ↵02 ↵03 �01 �02

50

later prioritize caloric contents. Similarly, ethanol exposure initiallyacts as an aversive reinforcer, but eventually turns into reward andinduces LTM (51). The sequential regulation of appetitive behav-ior by the same stimulus may be conserved across relevant appe-titive stimuli. As palatability is not always a faithful predictor of itsnutritional value, it may be a general design of reward systems tobalance short-term benefit and long-term fitness.

Materials and MethodsThe two GAL4 drivers, R48B04 and R15A04, were identified by screening ourconfocal image database for PAM cluster neurons (40). Appetitive condition-ing and reward substitution with thermoactivation were performed accord-ing to previously described protocols of differential conditioning (19, 29, 30).A group of flies received 1 min of odor exposure with sugar reward ortemperature-sensitive cation channel (dTrpA1)-mediated thermoactivationof GAL4-expressing neurons, followed by a 1-min presentation of another

odor in the absence of reward or temperature elevation. Most of the datadid not violate the assumption of the normal distribution and the homo-geneity of variance and were therefore subjected to parametric statistics.For data that significantly deviated from the normal distribution (Figs. 3Band 4I), nonparametric statistics were applied. Fluorescent immunohisto-chemistry was performed as described previously (19, 29, 30). All of theconfocal images in the main figures underwent landmark matching-basedimage registration using BrainAligner (59). For in vivo calcium imaging, fe-male flies expressing GCaMP3 (60) with R48B04-GAL4 and R15A04-GAL4were prepared essentially as described (19, 61). A droplet of 2 M sucrose, 3 Marabinose, or a mixture of 3 M arabinose and 2 M sorbitol solution in mineralwater was delivered on a plastic plate controlled by a micromanipulator (19).See SI Materials and Methods for more detailed experimental procedures.

ACKNOWLEDGMENTS. We thank T. Templier and A. Abe for technicalassistance and experiments that inspired this study; P. Garrity, H. Amrein,B. Pfeiffer, the Kyoto Drosophila Genetic Resource Center, and theBloomington Stock Center for fly stocks; K. Ito and the members of the

-0.2

0

0.2

0.4

0.6

0.8

-0.2 0 0.2 0.4 0.6 0.8

B

E

A

F

C

-4

0

log

P

-4

0

log

P

STM LTM

0

20

40

60

80

-0.2 0 0.2 0.4 0.6 0.80

20

40

60

80

-0.2 0 0.2 0.4 0.6 0.8

LTM (LI)

GAL

4 ex

pres

sion

(A.U

.)

STM (LI)

**

GAL

4 ex

pres

sion

(A.U

.)

STM (LI)

LTM

(LI)

D

-0.4-0.2

00.20.40.60.8

1

2min

24hrs

***

tneiciffeoc noitalerroC

-0.100.10.20.30.40.5

MB299B/+MB299B/UAS-ShiUAS-Shi/+

-0.100.10.20.30.40.5

MB299B/+MB299B/UAS-ShiUAS-Shi/+

-0.100.10.20.30.40.5

MB299B/+MB299B/UAS-ShiUAS-Shi/+

-0.100.10.20.30.40.5

MB299B/+MB299B/UAS-ShiUAS-Shi/+

-0.10

0.10.20.30.40.5

-0.10

0.10.20.30.40.5

-0.10

0.10.20.30.40.5

-0.10

0.10.20.30.40.5

IH

J K

ns

2M Sucrose

33°C

c t

2min

STM

(LI)

** 25

33°C c t

0 24hrs

2M Sucrose

LTM

(LI)

ns

3M Arabinose 3M Arabinose+2M Sorbitol

*

STM

(LI)

LTM

(LI)

G

33°C

c t

2min

25

33°C

c t

0 24hrs

stm-PAM

sugar

sweetness nutrition

α1

octopamine Gr43a

STM LTM

MB299Bβ2 α1

Fig. 4. PAM-α1 neurons are necessary and sufficient for appetitive LTM formation. (A) Correlation of STM and LTM scores. For each SplitGAL4 line, LTM scoreinduced by thermoactivation is plotted against STM score. Linear regression is overlaid. Significance of the correlation was tested by bootstrap, revealing noassociation between STM and LTM (n = 20). (B and C) GAL4 expression levels in each PAM cell type plotted against learning indices of appetitive STM (B) orLTM (C) of all 20 SplitGAL4 lines. Different colors and markers indicate different cell types. Linear regression is shown together for PAM-α1, which markedsignificant positive association with LTM (C). (D) Correlation coefficient of GAL4 expression level versus STM and LTM in each cell type. (E and F) Color map oflogarithm of P values calculated from correlation analyses of GAL4 expression and learning indices of STM (E ) or LTM (F). (G) Expression pattern of MB299B-GAL4 shown in a standardized brain. (Scale bar, 20 μm.) (H and I) Blockade of the PAM-α1 neurons by MB299B-GAL4 during memory acquisition with sucrosetested immediately (H) or 24 h later (I). n = 6–12. c, conditioning; t, training. Results with error bars are means ± SEM. **P < 0.01; ns, not significant. (J and K)Blockade of the PAM-α1 neurons has no significant effects during acquisition of arabinose memory tested immediately (J) but impairs formation of 24-hmemory induced by arabinose supplemented with sorbitol (K). (L) Model of a reward circuit in the fly brain. Reinforcement properties of sugar sweetness andnutrition are independently conveyed by octopamine or Gr43a neurons to distinct subclusters of the PAM cluster. The stm-PAM neurons collectively mediatereinforcement signal of STM. In contrast, the reinforcement signal of LTM is mainly conveyed by the PAM-α1 neurons.

Yamagata et al. PNAS Early Edition | 5 of 6

NEU

ROSC

IENCE

fructose

Nobuhiro Yamagata, Toshiharu Ichinose, Yoshinori Aso, Pierre-Yves Plaçais, Anja B. Friedrich, Richard J. Sima, Thomas Preat, Gerald M. Rubin, Hiromu Tanimoto, 2014

arabinose

>24 hrhours

X

Courtesy of Proceedings of the National Academy of Science. Used with permissionSource: Yamagata, Nobuhiro, Toshiharu Ichinose, Yoshinori Aso, Pierre-Yves Plaçais, Anja B. Friedrich,Richard J. Sima, Thomas Preat, Gerald M. Rubin, and Hiromu Tanimoto. "Distinct dopamine neuronsmediate reward signals for short-and long-term memories." Proceedings of the National Academy ofSciences 112, no. 2 (2015): 578-583.

51

later prioritize caloric contents. Similarly, ethanol exposure initiallyacts as an aversive reinforcer, but eventually turns into reward andinduces LTM (51). The sequential regulation of appetitive behav-ior by the same stimulus may be conserved across relevant appe-titive stimuli. As palatability is not always a faithful predictor of itsnutritional value, it may be a general design of reward systems tobalance short-term benefit and long-term fitness.

Materials and MethodsThe two GAL4 drivers, R48B04 and R15A04, were identified by screening ourconfocal image database for PAM cluster neurons (40). Appetitive condition-ing and reward substitution with thermoactivation were performed accord-ing to previously described protocols of differential conditioning (19, 29, 30).A group of flies received 1 min of odor exposure with sugar reward ortemperature-sensitive cation channel (dTrpA1)-mediated thermoactivationof GAL4-expressing neurons, followed by a 1-min presentation of another

odor in the absence of reward or temperature elevation. Most of the datadid not violate the assumption of the normal distribution and the homo-geneity of variance and were therefore subjected to parametric statistics.For data that significantly deviated from the normal distribution (Figs. 3Band 4I), nonparametric statistics were applied. Fluorescent immunohisto-chemistry was performed as described previously (19, 29, 30). All of theconfocal images in the main figures underwent landmark matching-basedimage registration using BrainAligner (59). For in vivo calcium imaging, fe-male flies expressing GCaMP3 (60) with R48B04-GAL4 and R15A04-GAL4were prepared essentially as described (19, 61). A droplet of 2 M sucrose, 3 Marabinose, or a mixture of 3 M arabinose and 2 M sorbitol solution in mineralwater was delivered on a plastic plate controlled by a micromanipulator (19).See SI Materials and Methods for more detailed experimental procedures.

ACKNOWLEDGMENTS. We thank T. Templier and A. Abe for technicalassistance and experiments that inspired this study; P. Garrity, H. Amrein,B. Pfeiffer, the Kyoto Drosophila Genetic Resource Center, and theBloomington Stock Center for fly stocks; K. Ito and the members of the

-0.2

0

0.2

0.4

0.6

0.8

-0.2 0 0.2 0.4 0.6 0.8

B

E

A

F

C

-4

0

log

P

-4

0

log

P

STM LTM

0

20

40

60

80

-0.2 0 0.2 0.4 0.6 0.80

20

40

60

80

-0.2 0 0.2 0.4 0.6 0.8

LTM (LI)

GAL

4 ex

pres

sion

(A.U

.)

STM (LI)

**

GAL

4 ex

pres

sion

(A.U

.)

STM (LI)

LTM

(LI)

D

-0.4-0.2

00.20.40.60.8

1

2min

24hrs

***

tneiciffeoc noitalerroC

-0.100.10.20.30.40.5

MB299B/+MB299B/UAS-ShiUAS-Shi/+

-0.100.10.20.30.40.5

MB299B/+MB299B/UAS-ShiUAS-Shi/+

-0.100.10.20.30.40.5

MB299B/+MB299B/UAS-ShiUAS-Shi/+

-0.100.10.20.30.40.5

MB299B/+MB299B/UAS-ShiUAS-Shi/+

-0.10

0.10.20.30.40.5

-0.10

0.10.20.30.40.5

-0.10

0.10.20.30.40.5

-0.10

0.10.20.30.40.5

IH

J K

ns

2M Sucrose

33°C

c t

2min

STM

(LI)

** 25

33°C c t

0 24hrs

2M Sucrose

LTM

(LI)

ns

3M Arabinose 3M Arabinose+2M Sorbitol

*

STM

(LI)

LTM

(LI)

G

L

33°C

c t

2min

25

33°C

c t

0 24hrs

stm-PAM

sugar

sweetness nutrition

octopamine Gr43a

α1

STM LTM

MB299Bβ2 α1

Fig. 4. PAM-α1 neurons are necessary and sufficient for appetitive LTM formation. (A) Correlation of STM and LTM scores. For each SplitGAL4 line, LTM scoreinduced by thermoactivation is plotted against STM score. Linear regression is overlaid. Significance of the correlation was tested by bootstrap, revealing noassociation between STM and LTM (n = 20). (B and C) GAL4 expression levels in each PAM cell type plotted against learning indices of appetitive STM (B) orLTM (C) of all 20 SplitGAL4 lines. Different colors and markers indicate different cell types. Linear regression is shown together for PAM-α1, which markedsignificant positive association with LTM (C). (D) Correlation coefficient of GAL4 expression level versus STM and LTM in each cell type. (E and F) Color map oflogarithm of P values calculated from correlation analyses of GAL4 expression and learning indices of STM (E ) or LTM (F). (G) Expression pattern of MB299B-GAL4 shown in a standardized brain. (Scale bar, 20 μm.) (H and I) Blockade of the PAM-α1 neurons by MB299B-GAL4 during memory acquisition with sucrosetested immediately (H) or 24 h later (I). n = 6–12. c, conditioning; t, training. Results with error bars are means ± SEM. **P < 0.01; ns, not significant. (J and K)Blockade of the PAM-α1 neurons has no significant effects during acquisition of arabinose memory tested immediately (J) but impairs formation of 24-hmemory induced by arabinose supplemented with sorbitol (K). (L) Model of a reward circuit in the fly brain. Reinforcement properties of sugar sweetness andnutrition are independently conveyed by octopamine or Gr43a neurons to distinct subclusters of the PAM cluster. The stm-PAM neurons collectively mediatereinforcement signal of STM. In contrast, the reinforcement signal of LTM is mainly conveyed by the PAM-α1 neurons.

Yamagata et al. PNAS Early Edition | 5 of 6

NEU

ROSC

IENCE

Nobuhiro Yamagata, Toshiharu Ichinose, Yoshinori Aso, Pierre-Yves Plaçais, Anja B. Friedrich, Richard J. Sima, Thomas Preat, Gerald M. Rubin, Hiromu Tanimoto, 2014

fructose

arabinose+

sorbitolarabinose

hours >24 hrCourtesy of Proceedings of the National Academy of Science. Used with permissionSource: Yamagata, Nobuhiro, Toshiharu Ichinose, Yoshinori Aso, Pierre-Yves Plaçais, Anja B. Friedrich,Richard J. Sima, Thomas Preat, Gerald M. Rubin, and Hiromu Tanimoto. "Distinct dopamine neuronsmediate reward signals for short-and long-term memories." Proceedings of the National Academy ofSciences 112, no. 2 (2015): 578-583.

52

internal state and gating

Michael J. Krashes, Shamik DasGupta, Andrew Vreede, Benjamin White, J. Douglas Armstrong, Scott Waddell, 2009

odor A + sweet

odor A odor B

hungry or fed

53

↵1 ↵2 ↵3 �1 �2 ped �1 �2 �3 �4 �5 ↵01 ↵02 ↵03 �01 �02

activated in fed statesilenced in hungry state

Michael J. Krashes, Shamik DasGupta, Andrew Vreede, Benjamin White, J. Douglas Armstrong, Scott Waddell, 2009

54

↵1 ↵2 ↵3 �1 �2 ped �1 �2 �3 �4 �5 ↵01 ↵02 ↵03 �01 �02

modulationby food & hunger

Laurence Lewis, K.P. Siju, Yoshinori Aso, Anja B. Friedrich, Alexander J.B. Bulteel, Gerald M. Rubin, and Ilona C. Grunwald Kadow, 2015

CO2 avoidance

55

highlystructured

perceptron-like readout

random, high-D representation

poolingnoise reductionnormalization

input

multiple forms of modulation and plasticity

random

highlystereotyped

56

MIT OpenCourseWarehttps://ocw.mit.edu

Resource: Brains, Minds and Machines Summer CourseTomaso Poggio and Gabriel Kreiman

The following may not correspond to a p articular course on MIT OpenCourseWare, but has beenprovided by the author as an individual learning resource.

For information about citing these materials or our Terms of Use, visit: https://ocw.mit.edu/terms.