Connectomic reconstruction of a female
Drosophila
ventral nerve cord
In the format provided by the
authors and unedited
Nature | www.nature.com/nature
Supplementary information
https://doi.org/10.1038/s41586-024-07389-x
Supplementary Table 1.
FANC links to leg motor neurons
Nerve
Segment
Function
Muscle
MNs
DProN
thorax
swing
tergopleural promotor, pleural promotor
4
link
VProN
thorax
swing
sternal anterior rotator
2
link
ProAN
thorax
swing
sternal adductor
1
link
ProAN
thorax
stance
pleural remotor and abductor
2
link
ProAN
thorax
stance
sternal posterior rotator
4
link
Ventral
thorax
extend
tergotrochanter
4
link
Ventral
thorax
extend
sternotrochanter extensor
2
link
ProLN
coxa
extend
trochanter extensor
2
link
ProAN, VProN
coxa
flex
trochanter flexor
8
link
ProLN
coxa
flex
accessory trochanter flexor
3
link
ProLN
trochanter
reductor
femur reductor
6
link
ProLN
femur
extend
tibia extensor
2
link
ProLN
femur
flex
main tibia flexor
5
link
ProLN
femur
flex
accessory tibia flexor
10
link
ProLN
tibia
levate
tarsus levator
1
link
ProLN
tibia
depress
tarsus depressor
4
link
ProLN
tibia
depress
retro tarsus depressor
1
link
ProLN
femur
claw
ltm femur
2
link
ProLN
tibia
claw
ltm tibia
2
link
ProLN
claw
ltm uncertain
2
link
ProLN
femur and tibia
claw
ltm dipalpha
2
link
Supplementary Table 2.
FANC Links to wing motor neurons
Nerve
Muscle
Insertion
MNs
PDMN
dorsal longitudinal
thorax
5
link
ADMN
dorsoventral 1
thorax
3
link
mesoAN
dorsoventral 2
thorax
2
link
PDMN
dorsoventral 3
thorax
2
link
ADMN
i1
first axillary
1
link
ADMN
i2
first axillary
1
link
mesoAN
iii1
third axillary
1
link
mesoAN
iii3
third axillary
1
link
mesoAN
iii4
third axillary
1
link
ADMN
iv1
fourth axillary
1
link
ADMN
iv2
fourth axillary
1
link
ADMN
iv3
fourth axillary
1
link
mesoAN
iv4
fourth axillary
1
link
ADMN
b1
basalar apophysis
1
link
ADMN
b2
basalar apophysis
1
link
ADMN
b3
basalar apophysis
1
link
ADMN
tp1
meso
-
scutum
1
link
ADMN
tp2
meso
-
scutum
1
link
ADMN
tp & tp2
meso
-
scutum
1
link
mesoAN
PS1
pleural apophysis
1
link
mesoAN
PS2
pleural apophysis
1
link
PDMN
middle leg tergotrochanter
3
link
PDMN
peripherally synapsing interneuron
1
link
Supplementary Methods. Identification of leg motor neuron targets
1
Here we describe the effort to match leg motor neurons to their targets. We first describe the resources and general methods.
We
2
then give an overview of Appendix A., Figures A1-
17, the atlas of FANC left T1
MNs, including an explanation for how the figures
3
are laid out. Finally, specific evidence for each neuron or group of neurons is given in the legends for Figures A2-
17.
4
Anatomy datasets
5
Interpreting the connectome requires knowing which MNs control which muscles. We therefore sought to identify the peripheral
6
muscle targets of all MNs innervating the fly’s front (T1) leg. To do so, we used three imaging datasets that collectively span the
7
VNC and leg (main Figure 3A).
8
1)
The FANC dataset establishes the number of MNs in the T1 neuropils: 69 T1L, 70 in right
T1. It also shows which of four
9
nerves each MN axon exits the neuropil through: the Prothoracic Accessory Nerve (ProAN), the Dorsal Prothoracic Nerve
10
(DProN), the Prothoracic Leg Nerve (ProLN) and the Ventral Prothoracic Nerve (VProN).
11
2)
An X-
ray holographic nano-
tomographic (XNH) dataset of the fly’s front leg (Kuan et al., 2020)
(Kuan et al., 2020)
shows
12
each of the prothoracic nerves and their branches into the musculature, and shows both where sensory axons join and where
13
motor axons leave the nerve. For many MNs with large axons, we could even trace their axons to their target muscle fibers.
14
We also used the XNH dataset to determine leg muscle fiber origins, insertions and numbers, as well as how the tendons
15
move the leg joints (Figure A1). At the proximal end of the volume, the insertions of the thoracic muscles onto tendons and
16
apodemes that contact the coxa are visible. In most cases, the origins of the thoracic muscle fibers on the thoracic cuticle
17
are not visible. Distally, the dataset ends ~¾ the length of the tibia. Many of the tibia muscle fibers are visible except fo
r a
18
few notable fibers. The tibia
-tarsus joint is not visible.
19
3)
We screened a large collection of VNC neurons sparsely labeled with the multi
-color Flp
-out (MCFO) technique to identify
20
GAL4 driver lines labeling leg MNs
(Meissner et al., 2020)
. We imaged GFP expression of each genetic driver line in the
21
T1 leg to identify the muscle target of each MN axon. We then compared the dendritic morphology of the genetically-
22
labeled MNs to those reconstructed from FANC (Figure 3
C).
23
Past work showed that most leg MNs have clear matches on the left and right sides (Phelps et al., 2021)
, which we confirmed through
24
the identification exercise. Thus, for Figures A2
-17, we attempt to match only the FANC neurons in left T1
, but we use MCFO
25
clones in either left
T1 or
right T1
, and assume that they have a contralateral match.
26
Each anatomical tool has its drawbacks. For one, the EM volume does not show the muscle targets for each neuron. For another,
27
while it is possible to trace large neurons in the XHN volume, it is difficult to resolve thin axons, particularly within the
muscle
28
fibers, making it difficult to precisely count the number of neurons targeting each muscle, or the number of muscle fibers contacted
29
by every neuron. A third drawback, we found 196 images of single motor neurons in the Janelia MCFO collection, but the GAL4
30
driver line
s–from which the MCFO clones are generated
–are typically not sparse. In cases where a single MN was labeled by the
31
GAL4 line, we could make a direct one
-to-one match between dendrite morphology and muscle target. More often, several motor
32
neurons are labeled by a GAL4 line. However, along with evidence from the literature, the tools can together compensate for t
hese
33
and other drawbacks.
34
Drosophila leg motor neurons in the literature.
35
To confirm our findings in the anatomy datasets, we cross
-referenced the following studies that describe aspects of motor neuron
36
morphology and muscle innervation.
37
Baek and Mann
(Baek and Mann, 2009)
used the MARCM technique to label individual neuron clones with GFP, to image their
38
dendritic morphology in the VNC, and to image their axons in the leg. Based on their results, they determined which lineages
39
produced neurons targeting specific leg segments, and the birth order of neurons within different motor neuron lineages. The authors
40
generously shared their data to help confirm the MN identification in this study. Brierley et al. (Brierley et al., 2012)
used a similar
41
technique and made similar observations. Together, the studies complemented each other to label and image clones of many of t
he
42
neurons present in FANC.
43
Subsequent studies have probed the molecular mechanisms of motor neuron identity and muscle targeting. These papers include
44
additional context, as well as images of motor neuron morphology and their muscle targets, which we have found useful in
45
confirming our MN identification (Enriquez et al., 2015; Guan et al., 2022; Venkatasubramanian et al., 2019)
. Finally, in our own
46
previous work characterizing the electrophysiology, force generation and neural activity of several specific MNs, we filled t
ibia
47
flexor neurons with neurobiotin or biocytin, allowing us to definitively match dendritic morphology to axon morphology and
48
functional characteristics (Azevedo et al., 2020)
.
49
Drosophila leg musculature in the literature.
50
The names of
Drosophila
leg muscles differ across the literature. Miller
(Miller, 1950)
applied the nomenclature for locust leg
51
muscles
(Snodgrass, R.E., 1935)
to
Drosophila
, which has largely been adopted.
Here, we define and use synonyms for the
52
musculature to help clarify how a muscle actuates its joint (Table A1). In some cases, we have made novel observations from t
he
53
XNH volume on how a specific muscle actuates a joint, and we offer new names for the muscle as a result.
54
Soler et al.
(Soler et al., 2004)
used genetic techniques to label muscles and tendons, and established nomenclature for the leg
55
musculature, based on older work by Miller
(Miller, 1950)
. As we have argued previously (Azevedo et al., 2020)
, we believe Soler
56
et al. misidentified the accessory tibia flexors as a tibia reductor muscle, possibly a misreading of earlier work. They did not identify
57
the thoracic muscles, so we have relied on the work by Miller for their names and suspected function.
58
As a specific example, the tibia levator muscle (Snodgrass, 1935)
, a.k.a. “tilm” (Soler et al., 2004)
, refers to the muscle that extends
59
the tibia to “lift” it off the ground. We find the term “levator” unsatisfying for several reasons. First, the terms “levator
” and
60
“depressor” are not commonly applied to limbed vertebrates in modern literature, whereas “extensor” and “flexor” are common
61
terms. Second, the action that levator (or depressor) muscles have on a joint is not always the same: the trochanter levator
flexes
the
62
coxa
-trochanter/femur joint, whereas the tibia levator
extends
the femur
-tibia joint. Third, the well-
studied FETi and SETi MNs
63
extend the tibia, so it is simpler to refer to both the extensor muscle and to the extensor MNs. Thus, we call this muscle the tibia
64
extensor muscle in the main text and figures.
65
We retain the terms “levator” vs. “depressor” for the muscles that “lift” or “push down” the tarsus. When the fly is standing, the
66
tarsus bends back towards the tibia and the tibia
-tarsus joint flexes. The depressor muscle causes the tarsus to extend, to push the fly
67
off the substrate, while the levator muscle appears to flex the tibia
-tarsus joint further. We note that the terms levator vs depressor
68
imply the animal is standing upright with respect to gravity, and flies often hang from surfaces and walls.
69
Table A1.
Muscle nomenclature across the literature.
70
Updated muscle name
Action
Atlas Figure,
Appendix A
Snodgrass, 1935
Miller, 1950
Soler et al., 2004
Tergopleural promotor
Promote (move anteriorly) the
coxa
Figure A2
28
Pleural promotor
Promote (move anteriorly) the
coxa
Figure A2
30
Pleural remotor
and
abductor
Remote (move posteriorly) and
abduct (move laterally) the coxa
Figure A4
29
Sternal anterior rotator
Anterior movement of coxa
Figure A3
31
Sternal posterior
rotator
Posterior movement of coxa
Figure A4
32
Sternal adductor
Adduct (move medially) the coxa
Figure A3
33
Tergotrochanter
extensor
Extend the coxa
-trochanter joint
Figure A5
P
Extracoxal trochanteral
depressor
Sternotrochanter
extensor
Extend the coxa
-trochanter joint
Figure A6
Extracoxal
trochanteral
depressor
Trochanter extensor
Extend the coxa
-trochanter joint
Figure A6
Trochanter
depressor
Trochanter depressor
Trochanter
depressor (trlm)
Trochanter flexor
Flex the coxa
-trochanter joint
Figure A7
-8
Trochanter levator
Trochanter levator
Trochanter levator
(trlm)
Accessory trochanter
flexor
Flex
coxa
-trochanter joint
Figure A9
Trochanter reductor
(trrm)
Femur reductor
Unknown
Figure A10
Femur reductor
Femur reductor
Femur reductor
(ferm)
Femur
depressor
(fedm)
Tibia extensor
Extend the femur
-tibia joint
Figure A11
Tibia levator
Tibia levator
Tibia levator
(tilm)
Tibia flexor
Flex the femur
-tibia joint
Figure A12
Tibia levator
Tibia levator
Tibia depressor
(tidm)
Accessory tibia flexor
Flex the femur
-tibia joint
Figure A13
-14
Accessory tibia
levator
Accessory tibia levator
Tibia reductor (tirm)
Tarsus depressor
muscle
Extend the tibia
-tarsus joint. The
joint is flexed when the fly is
standing. Extension moves the
fly’s body away from the
substrate.
Figure A17
Tarsus depressor
Tarsus depressor
Tarsus depressor
muscle
(tadm)
Tarsus retro depressor
muscles
Muscle fibers originate on tibia
cuticle that is distal to their
insertion sites on the tarsus
depressor tendon.
Figure A17
Tarsus depressor
Tarsus depressor
Tarsus reductor
muscles 1 and 2.
(tarm 1 and 2)
Tarsus levator muscle
Flex the tibia
-tarsus joint.
The joint is flexed when
supporting the fly’s weight.
Flexion brings the fly closer to the
substrate.
Figure A17
Tarsus levator
Tarsus levator
Tarsus levator
muscle
(talm)
Long tendon muscle 2
Located in femur. Pull on the long
tendon.
Figure A15
-16
Long tendon muscle
2
Long tendon muscle 1
Located in the tibia. Pull on the
long tendon
Figure A15
-16
Long
tendon muscle
1
71
Matching motor neuron dendrite morphology across datasets
72
To match motor neurons, we relied on expert visual recognition of specific morphological features for each MN, rather than on
73
numerical algorithms like NBLAST (Costa et al., 2016)
. NBLAST was successfully used previously to match between left and right
74
T1 MNs, and to classify axon bundles (Phelps et al., 2021)
. We used the bundle identification together with the following
75
distinguishing characteristics to match motor neurons.
76
Prothoracic nerves.
The names and abbreviations of the peripheral nerves come from (Court et al., 2020)
. As reported in Phelps et
77
al (2021), the following number of motor neurons exit through the four prothoracic nerves in the FANC volume:
78
1)
Prothoracic accessory nerve (ProAN)
- 12 MNs. The ProAN follows the leg nerve but splits off just after leaving the
79
neuropil
.
80
2)
Dorsal prothoracic nerve (DProN) -
4 MNs. The DProN exits the neuropil laterally, more anteriorly and dorsally than the
81
other nerves.
82
3)
Prothoracic leg nerve (ProLN)
- 42 MNs.
83
4)
Ventral prothoracic nerve (VProN) -
11 MNs. The VProN exits the neuromere laterally, more anteriorly than the leg nerve,
84
but ventral
-posterior to the DProN.
85
We assumed that a similar number of motor neurons travel along each nerve in the XNH dataset. We could distinguish motor axon
s
86
from sensory axons in each nerve in the XNH dataset when they could be traced to their target muscles or source sensory organ,
87
respectively. Sensory neurons often have extremely thin axons, making them difficult to trace, but a limited number have larger
88
axons and are traceable.
89
Soma location.
Most MN somas are on the anterior cortex of T1 (Figure
4C). Six neurons have cell bodies on the posterior cortex
90
(Figures A4 and A7) and one additional neuron has a cell body on the dorsal cortex (Figure A6). For neurons with anterior som
as,
91
we did not assume that the specific location of the soma was a reliable indicator of identity; We and others have found that across
92
different flies, somas of identified motor neurons can be in different locations within the anterior cluster (Azevedo et al., 2020; Baek
93
& Mann 2009)
.
94
Neurite tracts and neurite bundles.
As shown previously, neurons exiting the same nerve could be further classified into specific
95
bundles based on close proximity of the primary neurites, i.e. the branch running between the soma and the axon (Phelps et al.,
96
2021)
. In some cases, 3D visualization of FANC MNs in neuroglancer revealed sub-
bundles. In many cases, these bundles were
97
associated with muscle targets, so once one MN was identified, we could estimate the number of neurons innervating the muscle
.
98
Dendrite morphological features.
Most MNs have distinctive, identifying projections within the VNC, which could be used to
99
match neurons in FANC to MCFO clones (Enriquez et al., 2015)
. We rely heavily on these distinguishing features to group motor
100
neurons together to estimate how many neurons share features, and how they differ morphologically, such as whether there is a
101
gradient in soma, primary neurite size, or number and extent of dendrites. Some of these features are subtle and can differ between
102
neurons that target the same muscle. Many of these features appear indistinguishable in 2D projections, but 3D visualization and
103
depth-
colored MCFO projections often reveal distinctions.
104
Axon pathways and targets.
Each MN axon innervates a stereotyped set of muscle fibers
(Venkatasubramanian et al., 2019)
.
105
Consequently, each axon also leaves the peripheral nerve to enter a muscle at a roughly stereotyped point. We can observe axo
ns
106
leaving the nerve in the XNH volume, allowing us to count the number of neurons we expect to innervate each leg segment. When
107
several axons innervate a given segment or muscle, the axons can have different thicknesses. Because thicker axons tend to come
108
from MNs with larger somas and larger diameter primary neurites (Azevedo et al., 2020)
, the gradient of axon thickness should
109
correlate with the number of EM
-reconstructed neurons and any gradient in their dendritic properties. Ideally, tracing the full axon
110
branching patterns in the XNH dataset would allow us to create an atlas of axon anatomy
that we could compare our light
-level leg
111
imaging with. Unfortunately
, many of the MN axons are too thin to be traced given the ~200
nm resolution of the XNH dataset, so
112
while the neurons we can fully trace give us valuable information, we were only able to reconstruct a subset of the complete
113
population.
114
By identifying these features across the datasets, we could estimate the numbers of neurons that sha
re features. We could then use
115
the GAL4 line expression to match axon targets to dendritic morphology, as well as rule out possible matches. Finally, we use
d
116
inference and process of elimination to buttress direct evidence.
117
Confocal imaging
118
Fly prothoracic (
front
) legs
were immersed in a 4% formaldehyde (PFA) PBS solution for 20 minutes, followed by three rinses in
119
PBS with 0.2% Triton X
-100 (PBT).
The legs were
then
incubated in a PBS solution containing 1:50 phalloidin
(Alexa
-phalloidin
-
120
647,
Fisher A222287)
and
the following reagents that
improve
tissue penetrance:
1% Triton X
-100, 0.5% DMSO, 0.05 mg/ml Escin
121
(Sigma
-Aldrich, E1378
), and 3% normal goat serum. Legs were incubated
for one week at 4 °C with occasional rocking. After
122
staining, legs were
rinse
d 3x
with PBS
-Tx, 1 rinse with PBS, and were mounted onto slides in Vectashiel
d.
To image MNs
123
innervating the coxal muscles in the thorax, the fly was fixed as above, then hemisected along the parasagital plane with a f
ine razor
124
blade in Tissue
-Tek O.C.T. compound (Sakura 4583) frozen for 10 seconds on dry ice. Hemisected thoraces were rinsed 3x in PBT
125
and stained as above.
126
Mounted legs or thoraxes were imaged on a Confocal Olympus FV1000.
At least one image stack of each segment of the leg was
127
acquired. If GFP
was expressed
in a motor neuron in a particular segment, two image stack
s of the segment were acquired. Images
128
are available
on Dryad at DOI https://doi.org/10.5061/dryad.k0p2ngfg0
.
Image stacks were processed in
FIJI
(Rueden
et al., 2017).
129
F
ly strains for genetically labeling motor neurons
130
We screened a large collection of VNC neurons sparsely labeled with the multi
-color Flp
-out (MCFO) technique to identify GAL4
131
driver or split
-GAL4 hemidriver lines labeling
leg MNs (Meissner et al., 2020). The confocal images in Figures A2
-A17
come from
132
a resulting
collection of 75 lines driving expression of
GFP expression
in the leg
(Table A2
). An additional 31 lines were
imaged
133
that showed no MN expression in the leg.
For “Gen1” Janelia GAL4 lines, the genotype was P{
<GMR
>- GAL4}attP2
/P{20XUAS
-
134
06XGFP}attP2
. For VDRC
Vienna Tile DBD hemidrivers, the genotype was TI{2A
-p65(AD)::Zip+}VGlut[2A
-p65AD]/
+;
135
P{y[+t7.7]w[+mC]=
<VT
>- GAL4.DBD}attP2
/P{20XUAS
-06XGFP}attP2
.
136
Blinding
137
Experimenters were not blinded to the genotype when acquiring images.
138
Randomization
139
Imaging experiments were not intentionally randomized, but lines were imaged as they were ordered and crossed,
without care for
140
any particular order.
141
Table A2.
Motor neurons expression driven by GAL4 lines and split
-GAL4 DBD hemidrivers. The 69 leg MNs in T1 in FANC are
142
indicated in the rows. The columns list specific GAL4 lines (Gen1 Janelia GMR lines, e.g. R10B11) and split
-GAL4 DBD
143
hemidrivers (VT lines). Numbers and gray scale indicate a
confidence
heuristic
that a specific MN is labeled by the GAL4 reagent.
144
145
146
147