Article

Unique homeobox codes delineate all the

neuron classes of C. elegans

https://doi.org/10.1038/s41586-020-2618-9 Molly B. Reilly1, Cyril Cros1, Erdem Varol2, Eviatar Yemini1 & Oliver Hobert1 ✉

Received: 5 February 2020

Accepted: 4 June 2020 It is not known at present whether neuronal cell-type diversity—defined by

Published online: xx xx xxxx cell-type-specific anatomical, biophysical, functional and molecular signatures—can
be reduced to relatively simple molecular descriptors of neuronal identity1. Here we
Check for updates
show, through examination of the expression of all of the conserved homeodomain
proteins encoded by the Caenorhabditis elegans genome2, that the complete set of
118 neuron classes of C. elegans can be described individually by unique combinations
of the expression of homeodomain proteins, thereby providing—to our knowledge—
the simplest currently known descriptor of neuronal diversity. Computational and
genetic loss-of-function analyses corroborate the notion that homeodomain proteins
not only provide unique descriptors of neuron type, but also have a critical role in
specifying neuronal identity. We speculate that the pervasive use of homeobox genes
in defining unique neuronal identities reflects the evolutionary history of neuronal
cell-type specification.

The classification of neurons into distinct types is an important step
towards understanding the logic of the evolution, development and
function of the nervous system1. Traditionally, the classification of
neuron types has relied on anatomical features, and later expanded
to include electrophysiological features and eventually molecular
markers1. The emergence of high-throughput transcriptome profiling,
including single-cell sequencing, has deepened our appreciation of
the complexity of neuronal cell types among many different animal
species, from very simple (for example, cnidarian) to very complex
(mammals)3–6. Ongoing molecular classifications of neuron type raise
a number of questions, including whether there is a minimal descriptor
for neuronal identity—that is, whether specific subsets of molecular
features exist that are sufficient to capture the full complexity of all neu-
ronal cell types, or whether unique cellular identities can be described
only by their combined expression of many different types of gene.
Additionally, from a developmental standpoint, many questions remain
about how the molecular signatures that characterize individual neuron
types are genetically specified during differentiation.
Homeodomain transcription factors, which are encoded by
homeobox genes7, have emerged as possible answers to these
questions. Loss-of-function studies in a number of organisms have
demonstrated the importance of these transcription factors in the
specification of neuronal cell types. For example, in C. elegans, the
first neuronal-specification genes to be positionally cloned after unbi-
ased mutant screens were homeobox genes (mec-3, unc-4, unc-30 and
unc-86)8–11. Subsequent mutant analysis revealed that many additional
homeobox genes control neuronal identity in the nematode12. Home-
obox genes have also surfaced as specifiers of neuronal identity in other
organisms7,13–17, and recent single-cell profiling of many regions of the
mouse central nervous system has shown that homeobox genes are
the gene family that best distinguishes neuron classes of the central
nervous system4. A similar discriminatory power for homeobox gene
expression—particularly, the combinatorial expression of distinct
homeobox genes—has been revealed through the bulk sequencing of
179 distinct, genetically and anatomically identified cell populations
in mouse18. Transcriptome analysis in the visual system and the ventral
nerve cord of Drosophila has also revealed that homeobox genes display
a more-discriminatory expression profile than that of genes that encode
other types of transcription factors19,20. However, owing to the complex-
ity of the mouse and fly nervous system, and the resulting incomplete
coverage of all neuronal cell types, these previous studies have not
been able to test the possibility that the expression of homeobox genes
might uniquely identify every cell type in the entire nervous system.
We test this possibility here in the context of the nervous system of the
C. elegans model system. Fine-grained anatomical analysis of the adult
hermaphrodite worm has classified its 302 neurons  into 118 anatomi-
cally distinct types and several additional subtypes21,22. We set out to
systematically address how much of this diversity in neuronal cell type
can be explained by homeobox  gene expression and function.
C. elegans homeobox genes
The C. elegans genome encodes 102 homeobox genes (Methods), less
than half of the number of homeobox genes present in mammalian
genomes2,23,24. As in other animal genomes, C. elegans homeodomain
proteins do not constitute the largest family of transcription factors,
accounting for only about 10% of all genes that encode transcription
factors25,26. Of the 102 C. elegans homeobox genes, 70 have homologues
in other invertebrate and vertebrate genomes, 18 are conserved only
in nematodes and 14 are not conserved in any other known species of
Caenorhabditis2 (Fig. 1a). Caenorhabditis elegans contains representa-
tives of most subclasses of mammalian homeobox genes, characterized
by specific sequence features within the homeodomain (for example,
paired-type homeodomain) or by the presence of additional domains

1
Department of Biological Sciences, Howard Hughes Medical Institute, Columbia University, New York, NY, USA. 2Department of Statistics, Columbia University, New York, NY, USA.
✉e-mail: or38@columbia.edu

Nature | www.nature.com | 1
Article
a Homeodomain proteins in C. elegans No. Genes in C. elegans No. conserved
ANTP HD 9 ceh-2, ceh-5, ceh-7, ceh-12, ceh-23, ceh-43, ceh-62, ceh-63, pha-2 5

ANTP: HOXL HD 8 ceh-13, lin-39, mab-5, egl-5, nob-1, php-3, pal-1, vab-7 8
ANTP: NKL HD 14 ceh-1, ceh-9, ceh-19, ceh-22, ceh-24, ceh-27, ceh-28, ceh-30, 11
ceh-31, tab-1, mls-2, cog-1, vab-15, ceh-51
ANTP: En EH EH HD EH 1 ceh-16 1
ceh-8, ceh-10, ceh-17, ceh-36, ceh-37, ceh-45, ceh-53, ceh-54,
PRD-like HD 14 alr-1, ttx-1, unc-4, unc-30, unc-42, dsc-1 13
PRD PRD HD 3 pax-3, vab-3, eyg-1 2
POU POU HD 3 ceh-6, ceh-18, unc-86 3
HNF HNF HD 1 hmbx-1 1
Onecut Cut HD 6 ceh-21, ceh-38, ceh-39, ceh-41, ceh-48, ceh-49 6
Cut

CUX CASP Cut Cut Cut HD 1 ceh-44 1

CMP Compass HD HD 1 dve-1 0

LIM LIM HD 7 ceh-14, lim-4, lim-6, lim-7, mec-3, lin-11, ttx-3 7

ZF ZF HD ZF ZF ZF 1 zag-1 1
ZF
ZF ZF ZF ZF ZF ZF ZF ZF ZF ZF HD HD ZF ZF HD ZF ZF ZF 1 zfh-2 1

PROS HD Prospero 1 pros-1 1

4 ceh-33, ceh-34, ceh-32, unc-39 4
SIX/SO SIX/SO HD
HD 1 irx-1 1
IRO
TALE

PBC PBC HD 3 ceh-20, ceh-40, ceh-60 3

MEIS MEIS HD 1 unc-62 1
Conserved in Caenorhabditiae 1-2 HD 6 ceh-58, ceh-75, ceh-79, ceh-87, ceh-88, nsy-7 0
HD HD HD HD HD HD 1 ceh-92 0
Divergent

HD HD HD HD 1 ceh-93 0
De novo C. elegans 1–12 HDs 13 ceh-57, ceh-74, ceh-76, ceh-81, ceh-82, ceh-83, ceh-84, ceh-86, 0
ceh-89, ceh-90, ceh-99, ceh-100, duxl-1
HD HD HD HD 1 ceh-91 0

b pha-2fos::gfp ceh-8crispr::gfp ceh-93fos::gfp c tab-1crispr::gfp ttx-1crispr::gfp ceh-57fos::gfp ceh-16 ::gfp

fos

ADL ASI
RIA AIN M2
I4 ADF
AVD RIP ASG
AVE AIZ
AIB AFD RIB
ASJ RIG
RIF
Head Head Head Head Head Head Head RIG
RIF

d vab-15fos::gfp hmbx-1fos::gfp ceh-34fos::gfp

PVD
M5 M1
PVC ALM MI
PVQ I4 I6 M5
URX FLP I3 M4
PLM MC M3 M2
I2 I5
AVG NSM
PVT
Head Tail Head Ant. midbody Post. midbody Tail Head

irx-1fos::gfp Ex[vab-3fos::gfp] ceh-54fos::gfp

CEP RIV
I6 OLQ ASG
M5 PVQ AVD PLM
ASG SAA
M4 M2 I2 M3 BAG AFD
AWA
BAG AS AWC
AVF DD CEP DB VB VD ASE
RIH PVT SMBs SAA
SABV
Head Tail Head DB Tail Head

pros-1fos::gfp ceh-6fos::gfp ceh-14fos::gfp

AVH DVC
URYD ALA PVR
ASG RMDV PVN
I3 DVA
I5 PHB PVW
AWA PVQ PVQ
I2 RMDD AFD ADA PHC
AUA AVF
URYV
AVF PDA PHA PVC
Head Tail Head Tail Head AIM Tail PVT

Fig. 1 | The homeobox gene family in C. elegans, and representative neuron classes (b), 3 or 4 neuron classes (c) or 5 to 18 neuron classes (d). Ant,
expression patterns. a, Cartoon representations of homeodomain proteins anterior; post, posterior. Neurons were identified by overlap with the
and their associated domains by subfamily. Numbers and names of homeobox NeuroPAL landmark strain, outlined and labelled in yellow. Head structures,
genes in C. elegans, and the number of conserved homeobox genes in humans, including the pharynx, are outlined in white for visualization.
are based on ref. 2. HD, homeodomain. Yellow ‘HD’ symbol indicates Autofluorescence common to gut tissue is outlined with a white dashed line.
nematode-specific HOCHOB domain, a derivative of the homeodomain2. Ten worms were analysed for each reporter strain. Scale bars, 10 μm. All other
b–d, Representative images showing homeobox genes expressed in 1 or 2 expression patterns are shown in Extended Data Figs. 1–8.

(for example, the POU or LIM domain)2 (Fig. 1a). As in other animal our fosmid and/or endogenous reporter alleles reveal many novel
genomes, only a small fraction of all C. elegans homeobox genes are sites of expression of previously reported homeodomain proteins, in
of the Antennapedia-like HOX cluster23,24. addition to providing expression patterns of many dozen previously
uncharacterized homeodomain proteins (Supplementary Tables 1, 2).
It is important to emphasize that our analysis delineates protein
Analysis of homeodomain protein expression expression, thereby  capturing post-transcriptional regulatory
The expression patterns of a number of C. elegans homeobox genes events that are not revealed through mRNA-based transcriptomic
have previously been reported, but often without individual neuron approaches.
resolution and almost entirely with reporter reagents that do not We built an expression atlas of 101 of the 102 homeodomain proteins,
capture the full complement of regulatory sequences2,12,27 (Supple- including all of the 70 homeodomain proteins that are conserved out-
mentary Table 1). To comprehensively analyse the expression pattern side the nematode phylum, plus all of the 18 nematode-specific homeo-
of homeodomain proteins throughout the entire nervous system, we domain proteins, and 13 of the 14 C.-elegans-specific homeodomain
used fosmid-based reporter transgenes that contain the full inter- proteins (that is, no homologues in the genomes of other Caenorhab-
genic genomic context of the respective homeobox genes and/or we ditis species2). This atlas incorporates 97 homeodomain expression
engineered gfp (encoding green fluorescent protein, GFP) into home- patterns that we established ourselves using fosmid reporters and/or
obox gene loci using CRISPR–Cas9 genome engineering. As expected, CRISPR–Cas9-engineered reporter alleles, complemented with the

2 | Nature | www.nature.com
 a 120

60
Number of neuron classes

40

20

ceh-100
hmbx-1

unc-86

unc-30
unc-42

unc-39

unc-62
ceh-13

ceh-19
ceh-22
ceh-24
ceh-28
ceh-30
ceh-31
ceh-51

ceh-12
ceh-16
ceh-23
ceh-43
ceh-62
ceh-63
ceh-21
ceh-38
ceh-39
ceh-41
ceh-44
ceh-48
ceh-49

ceh-14

ceh-18

ceh-10
ceh-36
ceh-45
ceh-53
ceh-54

ceh-32
ceh-33
ceh-34
ceh-20
ceh-40
ceh-60

ceh-58
ceh-74
ceh-75
ceh-76
ceh-79
ceh-81
ceh-82
ceh-83
ceh-86
ceh-88
ceh-89
ceh-90
ceh-91
ceh-92
ceh-93
ceh-99
ceh-27

ceh-17
ceh-37

ceh-57

ceh-87
vab-15

pros-1
mab-5

mec-3

duxl-1
php-3
nob-1

pha-2
cog-1

pax-3

unc-4
ceh-1
ceh-9

ceh-5

ceh-2

ceh-6

ceh-8
ceh-7

dve-1

dsc-1
eyg-1
vab-3

zag-1
vab-7
lin-39

mls-2

lin-11

nsy-7
tab-1
pal-1

zfh-2
egl-5

lim-4
lim-6
lim-7

ttx-3

ttx-1
alr-1

irx-1
ANTP: HOXL ANTP: NKL ANTP Cut LIM POU PRD PRDL SIX TALE ZF Divergent

Number of homeobox genes



94.
(>*

94/
94+
(>)

5:4
:4+
:4)
7=>
(>(

94,
7=4

94-
)+<

:+8
(89

(=4

)(.

7/*

789

*(5

/:5
<9)
(34

638

734
(:.

7/)

<9(

+=*

7+)
(:/

7=8
<9?
<9@
7/(

(<(

+=)
(+(

7+(
7=*
*,7

7=5

:()
(=.
(+,

(:2

7+,

7=+

7=9
(=/

:((
(+-
(-+
(35

735
(+3

(:,

(=+

3<(
+=(

7=7
(=)

(=2
633

(3(

7=;
(:1

(=,

904
(=(

(=-
-37

(04

(=3
(=1

90.
90*

90/
90+
(05

90)

909
(0)

907
90:

:0)

4*
90(
(:0

(0@

90=
(0(

:0(
(0A

90-

4
4
4
4
4
++
+)
03
03

+(

=*
=+
(:

=)
=(

40
0
0
0
0
0
0
Sensory neurons Interneurons Motor neurons Pharynx

b
Jaccard distance

0.6
0.5
0.4
0.3
0.2
0.1
0
FLP
PVD
PVM
ALM
AVM

BDU
SDQ
AQR
RMG
ADE
PDE
AIZ
RIC
RIF
HSN
PDB
AIB

RIG
SAB
DD
VD

AS
DB
VB
VC
RMH
SMD
RID

SMB
AWB

SIB

RIR
RIB
AVH
RMD
CEP
OLQ
IL1
DVB

AVL
ADL
ASJ
PHB

URB

AIN

RIH
AVB
I1
MI
ASI
M2
M4
M5
I6
PVT
RIS
RIP
RMF
I4

AIY
CAN
BAG
OLL
URY
AIM
PVC

PQR
AVF
I5
PVQ

AVE
AVD
AVG
PVP
RIV

URX
IL2

I3
M3
NSM
ALN
PLN

MC
ASK

AVK

ADF
ASH
ASG
PHC
PVW
PVR
PLM
DVC
PVN
AVJ
RIM
M1
RME
ASE
AWC
AFD
I2
ADA

ALA

DA
VA

SAA

SIA

LUA

PDA

URA

AIA

AUA

RIA

PHA

AVA

DVA

AWA
Touch neurons VNC motor neurons Head motor

Fig. 2 | Summary of expression patterns of homeodomain proteins across
the nervous system, and similarity of neuron classes on the basis of
homeodomain codes. a, Top, number of neuron classes in which each
homeobox gene is expressed. Left, dot plot distribution. Each dot represents a
homeobox gene, and the value associated with the dot represents the number
of neuron classes in which this homeobox gene is expressed. Right, histogram
showing the number of neuron classes in which each homeobox gene is
expressed, organized and coloured by homeobox gene subfamily and shared
protein domains. The dashed line at 5 neuron classes denotes the median
number of neuron classes in which each homeobox gene is expressed. Bottom,
number of homeobox genes expressed in each neuron class. Left, dot plot
distribution. Each dot represents a neuron, and the associated value
represents the number of homeobox genes expressed in this neuron. Right,
histogram displaying the number of homeobox genes expressed in each
neuron class (excluding the pan-neuronal homeobox genes), ordered by the
neuron type (sensory, motor, interneuron and pharyngeal). The dashed line at
7 homeobox genes denotes the median number of genes per neuron class. b,
Dendrogram ordering neuron classes on the basis of the similarity of their
homeobox gene code. Some examples of functionally related neuron groups
are shaded. VNC, ventral nerve cord.

patterns of 4 previously fully characterized homeodomain patterns that
were also generated either using fosmid or CRISPR–Cas9-engineered
reporter alleles (Supplementary Table 1–3). We comprehensively
analysed the expression pattern of all these homeodomain proteins
at single-neuron resolution throughout all 302 neurons, using the
multicolour-landmark identification transgene NeuroPAL 28. We
focused our expression analysis on mature neurons in the nervous
system of late larval stage or young adult-stage worms, because con-
tinuous expression throughout the life of postmitotic neurons is usually
associated with transcription factors that specify and subsequently
maintain terminal neuron identity12,29.
Notably, we find that 80 of the 101 homeodomain proteins we exam-
ined are expressed in the mature nervous system (Fig. 1b–d, Extended
Data Figs. 1–7, Supplementary Tables 2, 3). Twelve are expressed in
all neurons and many major tissue types; two Cut-type homeobox
genes (ceh-44 and ceh-48), as well as the nematode-specific ceh-58
gene, are exclusively expressed in all neurons, but no other major
tissue types (Fig. 1, Extended Data Figs. 3, 7). At the other extreme,
seven homeodomain proteins are expressed exclusively in one neuron
class (Fig. 1, Extended Data Figs. 1, 2, 5, 7). More than two thirds of the
neuron-specific homeodomain proteins are expressed in less than 10%
of all neuron classes (Fig. 2a). Neurons that express the same homeodo-
main protein are not usually related by lineage or by neurotransmitter
identity (Extended Data Fig. 8). With the exception of pan-neuronally
expressed homeodomain proteins, no two homeodomain proteins
are expressed in the exact same combination of neuron classes (Sup-
plementary Table 2). The two homeodomain proteins with the closest
similarity in expression are UNC-62 (the orthologue of vertebrate MEIS

Nature | www.nature.com | 3
Article
Clustering neuron classes by similarity in homeobox gene expression

RMG

AWC
RMH

NSM
RMD
AWB
SMD

PVW
SMB
PVM

RME
AVM

SDQ

RMF

PQR
BDU

HSN

PHC
AQR

PLM
OLQ

URB

CAN
BAG
ALM

PHB

ASG
PDB

URA

URY

PVQ

DVC
PHA

AVG

ASH
DVB
PDA

URX

AUA

PVC

PVN
PVD

ADA

SAB

AVH
CEP
PDE

ASK

AVD

DVA

PVR
ADE

SAA

AVB
PLN

AVK

PVP
LUA

ALN

AFD
AVA

ADF

ASE
ADL

OLL

AVE
PVT
ALA

AVF
AVL
ASJ

RIM
AVJ
AIM
FLP

RIG
RIC

RIH
RID

RIR
RIB

AIN
SIB
AIB

RIS
RIP

RIA
MC
SIA

ASI
RIV

AIA

AIY
AIZ
RIF

M3

M2
M4
M5

M1
DD

DB

IL1

IL2
DA

VC
VD

VB
AS
VA

MI
I3

I1

I6

I4

I5

I2
pha-2
ceh-12
ceh-28
ceh-93
Clustering homeobox genes by similarity in neuron class expression

ceh-8
mls-2
ceh-10
ceh-7
ceh-2
unc-39
ttx-3
unc-30
dve-1
ceh-90
ceh-57
ceh-53
unc-4
ceh-6
ceh-9
ceh-19
lim-6
ceh-16
ceh-1
pros-1
ceh-37
ceh-17
lim-4
ceh-24
ceh-45
ceh-34
ceh-43
ceh-31
ceh-54
ceh-23
dsc-1
ceh-79
ceh-36
ceh-63
vab-7
ceh-14
php-3
nob-1
vab-15
irx-1
ttx-1
ceh-27
alr-1
lim-7
ceh-32
unc-86
mec-3
hmbx-1
egl-5
cog-1
vab-3
unc-62
ceh-20
lin-39
mab-5
ceh-13
tab-1
unc-42
lin-11
zfh-2
nsy-7
ceh-18
zag-1
ceh-89

Fig. 3 | Homeodomain expression atlas for entire nervous system of between neuron classes defined by the Jaccard index, as in Fig. 2b.
C. elegans. Neuron classes are coloured by neuron type (blue, sensory; pink, Homeodomain proteins are coloured by subfamily, and ordered by similarity of
motor; yellow, interneuron; and grey, pharyngeal) and ordered by similarity neuron-class expression and sparsity.

proteins), which is expressed in 33 neuron classes, and CEH-20 (orthol-
ogous to vertebrate PBX proteins), which is expressed in 32 classes
(31 of which are same as the classes that express UNC-62)—consistent
with the mutual dependency of function of MEIS and PBX proteins
in other organisms30. Tandem duplicated homeobox genes retain
overlaps in their expression, but in most cases one of the duplicates
shows an expression pattern that is much more restricted than the other
(Supplementary Table 2).
The expression pattern of members of subclasses of homeodomain
proteins (for example, POU, LIM and PRD) do not share obvious fea-
tures: for example, there is no enrichment of specific homeodomain
subclasses in sensory neurons versus interneurons or motor neurons, or
in neurons of a specific neurotransmitter identity. The only exceptions
are the above-mentioned Cut-type homeodomain proteins, which are
either ubiquitously or pan-neuronally expressed. The cellular specific-
ity of homeodomain protein within the nervous system appears to
correlate with the extent of conservation. Of the 70 conserved homeo-
domain proteins, 56 (80%) are expressed in specific subsets of neurons,
whereas only 10 out of the 18 (56%) nematode-specific proteins and
only 3 of the tested 13 (27%) C. elegans-specific homeobox genes are
selectively expressed in the nervous system (Extended Data Fig. 7, Sup-
plementary Table 2). Some of the highly unusual C. elegans-specific
homeodomain proteins2—such as CEH-100, which contains an unpar-
alleled number of twelve homeodomains—are expressed in all cells
and tissues, whereas the very unusual HOCHOB-type homeodomain
protein CEH-91 displays no expression in the mature nervous system
(Extended Data Fig. 7). The greater specificity of expression in individ-
ual neuronal cell types of conserved homeodomain proteins suggests
that neuron-type-specific expression may be an ancestral feature of
homeodomain protein expression.
Recently reported single-cell transcriptome sequencing recovered
mRNA profiles for 42 of the 118 neuron classes31,32. Although these
datasets recover homeobox gene transcripts in all of the 42 identi-
fied neuron classes, they uncover only little more than half (55%) of
the expression profiles that we recovered via our protein expression
analysis (Methods), which is probably a testament to the incom-
plete depth of single-cell RNA sequencing profiles (Supplementary
Table 4). Vice versa, there are cases in which a homeobox gene tran-
script can be detected in cells in which we observe no expression of
the corresponding protein (Supplementary Table 4), possibly owing
to post-transcriptional regulatory events. Together, the comparison
of our protein dataset with single-cell transcriptome data illustrates
the limitations of the depth of currently available single-cell transcrip-
tome datasets and the expected discordances between transcript and
protein expression.
Homeodomain combinations define neuron types
The most notable feature of the homeodomain protein expression atlas
becomes apparent when one considers the patterns of co-expression
of homeodomain proteins in distinct neuron classes: each neuron
class expresses its own entirely unique combination of homeodomain
proteins. Excluding the pan-neuronally expressed homeobox genes,
the combinatorial code consists of four homeodomain proteins on
average (Fig. 2a). Neuron-type-specific homeodomain codes are gen-
erated by the 70 phylogenetically conserved homeobox genes alone

4 | Nature | www.nature.com
 P = 1 × 10–5 P = 4 × 10–3
a WT ceh-32 (ok343) P = 9 × 10–27 P = 1 × 10–7 b WT ceh-31 (tm239) **** ***
**** ****
100 100

PVR neurons (%)

eat-4prom8::mCherry
ON

RIA neurons (%)

DIM
75 75
PVR
50 PVR 50

RIA RIA 25
25
RIA RIA 0
0 WT ceh-31 WT ceh-31
WT ceh-32 WT ceh-32 eat-4fos::YFP::H2B n = 40 n = 33 n = 10 n = 11
n = 50 n = 50 n = 10 n = 20
eat-4fosmid flp-10
eat-4prom8 glr-3
WT ceh-8 (gk116531) c P=2× 10–9
****
P = 5 × 10–9
****
P = 1 × 10–18 P = 5 × 10–9 P = 4 × 10–7
eat-4fos::mCherry ::H2B

100
**** **** ****

PVN neurons (%)

100 WT 2 PVN

RIA neurons (%)

75 PVN
PVN 75 1 PVN

50
PVN
PVN 50

25 10µm 25
RIA 0 ceh-9 0
RIA WT ceh-8 WT ceh-8 WT ceh-8 (tm2747) WT ceh-9 WT ceh-9
n = 29 n = 38 n = 15 n = 18 n = 10 n = 17
PVN
PVN PVN
PVN n = 26 n = 40 n = 26 n = 40
eat-4fosmid glr-3 dop-2 unc-17::TagRFP cho-1::YFP

Fig. 4 | Previously uncharacterized homeobox genes act as regulators of
neuronal identity. a–c, Each panel compares wild-type (WT) worms to
null mutants, with about 30 independent worms per condition for
neurotransmitter reporters and about 15 worms for other reporters. Graphs
show P values from Fisher’s exact test; ***P values between 10 −2 and 10 −3,
****P values below 10 −3. Characteristic images were chosen. In a, RIA neuron
identity is lost in ceh-8 and ceh-32 mutant worms, as assessed with multiple
markers. Left, eat-4 expression is lost from RIA neurons in a ceh-8- or
ceh-32-mutant background. Right, quantification of eat-4 loss in RIA neurons
for both ceh-8 and ceh-32 mutant worms, as well as glr-3 and dop-2 reporters
(see Extended Data Fig. 11 for reporter image).b, Glutamatergic identity of the
PVR neuron is lost in ceh-31-mutant worms. Left, eat-4 expression is lost in
ceh-31-mutant background. Right, quantification of eat-4 loss in PVR neurons,
as well as a flp-10 reporter (see Extended Data Fig. 11 for reporter image). c, PVN
neuron fate change in ceh-9-mutant worms. Left, PVN neuron expresses only
unc-17 in a wild-type background. In a ceh-9-mutant background, cho-1 fosmid
expression is ectopically activated and unc-17 expression is lost, indicating a
change in cell fate. Right, quantification of unc-17 and cho-1 expression in
wild-type and mutant worms. Scale bars, 5 μm (a, b), 10 μm (c).

(Extended Data Fig. 9a). Not all of the 70 conserved homeobox genes throughout the C. elegans nervous system and visualizing the sparsity
are required to generate neuron- class-specific codes. We calculated of this matrix (Fig. 3).
that the expression patterns of a minimal set of 24 conserved homeo- There are 118 anatomically defined neuron classes but 155 distinct
domain proteins uniquely identify all 118 neuron class (Extended Data combinatorial homeobox codes, which demonstrates that the home-
Fig. 9b). obox codes reveal additional neuronal subclass identities (Extended
We visualized the complete set of homeodomain codes using their Data Fig. 10a–c). For example, the six radially symmetric RMD neurons
Jaccard distance to construct a dendrogram, grouping neurons on the (composed of a dorsal and a ventral left–right symmetric neuron pair,
basis of the similarity of their unique homeodomain protein codes and a lateral left–right symmetric pair) are uniquely defined by the
(Fig. 2b, Methods). When comparing this clustering to the relatedness of combination of ceh-89, nsy-7, unc-42, zfh-2 and zag-1, but the dorsal and
neuron classes on the basis of other anatomical or functional criteria, a ventral neuron pair is further distinguished by additional expression
number of expected and unexpected relationships were revealed. Broad of ceh-32 and ceh-6 and the lateral pair by the additional expression
classes of functionally related neurons (such as ventral-nerve-cord of cog-1. The subclassification of the dorsal–ventral and the lateral
motor neurons, head motor neurons or touch-receptor neurons) clus- RMD pair is paralleled by synaptic connectivity differences21. Simi-
tered together on the basis of the similarity of their homeodomain larly, the inner labial neuron class IL1—composed of six class members
protein codes (Fig. 2b). Notably, neurons that share similar codes and (a dorsal, lateral and ventral pair)—can be subdivided into subclasses by
fall into related classes are not obviously related by lineage. However, differential homeodomain expression patterns (all three neuron pairs
functionally and anatomically related neuron classes can also display co-express ceh-43, ceh-32 and ceh-18, but only the dorsal and ventral
very different homeodomain protein codes, as seen—for example—in pairs express zfh-2). This subclassification also mirrors the distinct
the case of the two interconnected, anatomically similar and function- synaptic connectivity patterns of the dorsal and ventral IL1 pairs versus
ally related phasmid sensory-neuron classes PHA and PHB (Fig. 2b, Sup- the lateral IL1 pair21.
plementary Table 3). Conversely, several neuron classes that display no Yet another example of homeodomain codes subdividing neuron
obvious functional or anatomical similarity clustered together on the classes is evident in ventral-nerve-cord motor neurons that are aligned
basis of their homeodomain protein codes. For example, the amphid along the anterior–posterior axis (Extended Data Fig. 10c). Distinct
olfactory neuron AWB displays a code related to that of several head homeobox codes uniquely identify all known motor neuron classes
motor neurons. (that is, DA, VA, AS and so on), but the expression of HOX cluster pro-
We also clustered homeodomain proteins on the basis of similarity teins further subdivides the identity of individual members of specific
of their expression patterns. This dendrogram visualizes substantial motor neuron classes (for example, DA1 versus DA2)—not only towards
differences in the expression patterns of individual homeodomain the tail of the ventral nerve cord (as previously reported33,34), but also in
proteins (with a few notable exceptions, such as the MEIS and PBX mid- and anterior domains of the ventral nerve cord. Moreover, every
similarities noted above) (Fig. 3). We used both of our dendrograms single post-embryonically generated motor neuron class expresses
(that is, clustering homeodomain proteins on the basis of similarity a diverse set of additional, non-HOX homeodomain proteins in a
of expression patterns, as well as clustering of neuron classes on the subclass-specific manner, including VAB-3 (the C. elegans orthologue
basis of similarity of homeodomain expression) to order the axes of of PAX6), VAB-7 (EVX1 and EVX2) or COG-1 (NKX6) (Extended Data
our homeodomain expression matrix (Fig. 3). By grouping the most Fig. 10c). Lastly, our homeobox data also revealed left–right asym-
similar codes in proximity to each other, this illustrates the uniqueness metries in the functionally lateralized ASE neuron pair35, which we find
of each homeodomain code per neuron class, providing the most suc- express the homeobox genes alr-1 and ceh-23 in the left but not right
cinct summary of the expression patterns of homeodomain protein ASE neuron (Extended Data Figs. 1, 5).

Nature | www.nature.com | 5
Article
Homeodomain profiles predict neuron identity
We next set out to determine the extent to which the unique homeo-
domain expression code can account for the known molecular signa-
tures of all C. elegans neurons. To this end, we used a Wormbase-curated
list of 1,126 published reporter transgenes generated by the C. elegans
research community over the past few decades22. This reporter atlas
describes regulatory states for every single neuron type, with a sizable
average of 42 reporters expressed per neuron type22 (Supplementary
Table 5). We used a simple multivariate linear regression to ask how well
our homeodomain protein expression atlas (the independent variables)
fit the remaining genes observed in neurons (the dependent results).
We found that we could explain 74% of the reporter atlas expression
at single-neuron resolution, using our sparse set of homeobox pro-
tein expression. This a significantly better fit than our control dataset
(P < 0.0001), a randomly shuffled set of homeodomain protein expres-
sion data. To further illustrate the fit of our multivariate linear regression,
we used this regression to predict reporter expression in each neuron
class and correlated this prediction to the known reporter expression in
these neuron classes (Extended Data Fig. 11a, Supplementary Table 5).
Several classes of neuron have expression that is completely predicted by
homeodomain protein expression (exhibiting a correlation coefficient
of 1) and all of the remaining neuron classes show moderate-to-strong
positive correlations (exhibiting coefficients between 0.5 and 0.95).
Functional relevance of homeobox genes
Experimental validation of the importance of the homeobox code had
already been demonstrated by previous genetic loss-of-function analy-
sis, which had shown that 40 of the 80 neuronally expressed C. elegans
homeodomain proteins indeed have a role in the specification of neuronal
identity8–12 (Supplementary Table 2). We extended this functional analysis
by examining homeobox genes that were not previously implicated in
the specification of neuronal identity, and examining neurons for which
no identity-promoting factor had previously been reported. We found
that ceh-8, the C. elegans orthologue of the vertebrate RAX homeobox
gene, and the SIX/SO-type homeobox gene ceh-32—both of which were
uncharacterized in the context of the specification of neuronal identity—
define a unique homeodomain expression code for the RIA interneurons
(Extended Data Figs. 2, 5). In worms that carry a nonsense allele of either
ceh-8 or ceh-32, the RIA interneurons do not acquire a number of distinct
features of RIA identity (Fig. 4a, Extended Data Fig. 11b–e).
We further examined whether any of our newly identified expression
patterns of homeobox genes can distinguish previously defined, but
non-discriminatory homeobox codes. The unc-86 (an orthologue of
Brn3) POU and ceh-14 (an orthologue of LHX3) LIM homeobox genes were
previously found to specify the identity of distinct classes of neuron—
among them, the AIM and PVR neurons36,37. We discovered that the BarH
homologue ceh-31 is expressed in PVR—but not AIM— neurons, and that
in ceh-31-mutant worms, the glutamatergic as well as peptidergic identity
of PVR neurons is affected (Fig. 4b, Extended Data Fig. 11c). Similarly,
we discovered that the NK-like homeobox gene ceh-9 is required for the
specification of the neurotransmitter mechanistic identity of the PVN
neuron (Fig. 4c), a neuron that was previously found to be specified by
a combination of ceh-14 and unc-3, both of which also specify the PVC
neuron38. The ceh-9 homeobox gene therefore distinguishes ceh-14- and
unc-3-dependent PVN from ceh-14- and unc-3-dependent PVC identity.
Taken together, 74 of the 118 neuron classes of C. elegans have so far been
found to require at least one (if not multiple) homeobox transcription fac-
tors for the correct specification of their identity (Extended Data Fig. 11f).
Conclusions
We have shown here that the expression patterns of a single transcrip-
tion factor family fully describe the diversity of all neuronal cell types
6 | Nature | www.nature.com
throughout an entire nervous system. Several transcriptome datasets
from Drosophila and vertebrate nervous systems have also explicitly
noted that homeobox genes are the gene family that distinguishes
neuron types most effectively4,18–20. For example, bulk sequencing of
large collections of distinct, labelled cell types throughout the mouse
central nervous system also revealed that individual homeobox gene
combinations distinguish almost all unique populations of neuronal
cells18. However, to our knowledge, our analysis is the first to assign
unique homeodomain protein codes to a whole nervous system in its
entirety and with single-cell resolution. Transcriptome efforts from
more-complex nervous systems will need to be scaled up substantially
to assess the depth and breadth of combinatorial homeobox codes. As
transcriptome datasets do not capture post-transcriptional regulatory
events, ideally such transcriptome data need to be complemented by
protein expression data, as we have shown here.
Future analysis will reveal whether other families of transcription
factors display unique combinatorial expression patterns throughout
the nervous system. It is already clear that non-homeodomain types
of transcription factors also have critical roles in neuronal identity
specification (for example, ref. 12) but such non-homeodomain tran-
scription factors often cooperate with homeodomain transcription
factors in the control of neuron identity in C. elegans33,38–40. Inspired
by Dobzhansky’s dictum that ‘nothing in biology makes sense except
in the light of evolution’41, we speculate that the potential preponder-
ance of homeobox genes in the specification of neuronal identity may
hint at the possibility that homeodomain proteins were recruited into
the specification of neuronal identity very early in the evolution of the
nervous system. It is possible that a homeodomain transcription factor
was used to specify the signal properties of an ancestral ‘ur-neuron’
(the evolutionarily earliest, most-primitive form of a neuron). Different
neuronal cell types could have come into existence through homeobox
gene duplication, and an ensuing diversification of expression and
target specificity of the duplicated homeodomain proteins. Homeobox
expression codes may therefore provide a window in the evolutionary
history of neuronal cell types.
Online content
Any methods, additional references, Nature Research reporting sum-
maries, source data, extended data, supplementary information,
acknowledgements, peer review information; details of author con-
tributions and competing interests; and statements of data and code
availability are available at https://doi.org/10.1038/s41586-020-2618-9.
1. Zeng, H. & Sanes, J. R. Neuronal cell-type classification: challenges, opportunities and
the path forward. Nat. Rev. Neurosci. 18, 530–546 (2017).
2. Hench, J. et al. The homeobox genes of Caenorhabditis elegans and insights into their
spatio-temporal expression dynamics during embryogenesis. PLoS ONE 10, e0126947
(2015).
3. Sebe-Pedros, A. et al. Cnidarian cell type diversity and regulation revealed by
whole-organism single-cell RNA-seq. Cell 173, 1520–1534 (2018).
4. Zeisel, A. et al. Molecular architecture of the mouse nervous system. Cell 174, 999–1014
(2018).
5. Tasic, B. et al. Shared and distinct transcriptomic cell types across neocortical areas.
Nature 563, 72–78 (2018).
6. Hodge, R. D. et al. Conserved cell types with divergent features in human versus mouse
cortex. Nature 573, 61–68 (2019).
7. Gehring, W. J. Master Control Genes in Development and Evolution: The Homeobox Story
(Yale Univ. Press, 1998).
8. Way, J. C. & Chalfie, M. mec-3, a homeobox-containing gene that specifies differentiation
of the touch receptor neurons in C. elegans. Cell 54, 5–16 (1988).
9. Finney, M., Ruvkun, G. & Horvitz, H. R. The C. elegans cell lineage and differentiation gene
unc-86 encodes a protein with a homeodomain and extended similarity to transcription
factors. Cell 55, 757–769 (1988).
10. White, J. G., Southgate, E. & Thomson, J. N. Mutations in the Caenorhabditis elegans unc-
4 gene alter the synaptic input to ventral cord motor neurons. Nature 355, 838–841
(1992).
11. Jin, Y., Hoskins, R. & Horvitz, H. R. Control of type-D GABAergic neuron differentiation by
C. elegans UNC-30 homeodomain protein. Nature 372, 780–783 (1994).
12. Hobert, O. A map of terminal regulators of neuronal identity in Caenorhabditis elegans.
Wiley Interdiscip. Rev. Dev. Biol. 5, 474–498 (2016).
13. Tsuchida, T. et al. Topographic organization of embryonic motor neurons defined by
expression of LIM homeobox genes. Cell 79, 957–970 (1994).
14. Lindtner, S. et al. Genomic resolution of DLX-orchestrated transcriptional circuits driving
development of forebrain GABAergic neurons. Cell Rep. 28, 2048–2063 (2019).
15. Stettler, O. & Moya, K. L. Distinct roles of homeoproteins in brain topographic mapping
and in neural circuit formation. Semin. Cell Dev. Biol. 35, 165–172 (2014).
16. Tahayato, A. et al. Otd/Crx, a dual regulator for the specification of ommatidia subtypes in
the Drosophila retina. Dev. Cell 5, 391–402 (2003).
17. Blochlinger, K., Bodmer, R., Jack, J., Jan, L. Y. & Jan, Y. N. Primary structure and expression
of a product from cut, a locus involved in specifying sensory organ identity in Drosophila.
Nature 333, 629–635 (1988).
18. Sugino, K. et al. Mapping the transcriptional diversity of genetically and anatomically
defined cell populations in the mouse brain. eLife 8, e38619 (2019).
19. Davis, F. P. et al. A genetic, genomic, and computational resource for exploring neural
circuit function. eLife 9, e50901 (2020).
20. Allen, A. M. et al. A single-cell transcriptomic atlas of the adult Drosophila ventral nerve
cord. eLife 9, e54074 (2020).
21. White, J. G., Southgate, E., Thomson, J. N. & Brenner, S. The structure of the nervous system
of the nematode Caenorhabditis elegans. Phil. Trans. R. Soc. Lond. B 314, 1–340 (1986).
22. Hobert, O., Glenwinkel, L. & White, J. Revisiting neuronal cell type classification in
Caenorhabditis elegans. Curr. Biol. 26, R1197–R1203 (2016).
23. Bürglin, T. R. & Affolter, M. Homeodomain proteins: an update. Chromosoma 125, 497–521
(2016).
24. Bürglin, T. R., Finney, M., Coulson, A. & Ruvkun, G. Caenorhabditis elegans has scores of
homoeobox-containing genes. Nature 341, 239–243 (1989).
25. Lambert, S. A. et al. The human transcription factors. Cell 172, 650–665 (2018).
26. Fuxman Bass, J. I. et al. A gene-centered C. elegans protein–DNA interaction network
provides a framework for functional predictions. Mol. Syst. Biol. 12, 884 (2016).
27. Murray, J. I. et al. Automated analysis of embryonic gene expression with cellular
resolution in C. elegans. Nat. Methods 5, 703–709 (2008).
28. Yemini, E. et al. NeuroPAL: a neuronal polychromatic atlas of landmarks for whole-brain
imaging in C. elegans. Preprint at https://www.biorxiv.org/content/10.1101/676312v1 (2019).
29. Hobert, O. Terminal selectors of neuronal identity. Curr. Top. Dev. Biol. 116, 455–475
(2016).
30. Merabet, S. & Mann, R. S. To be specific or not: the critical relationship between Hox and
TALE proteins. Trends Genet. 32, 334–347 (2016).
31. Packer, J. S. et al. A lineage-resolved molecular atlas of C. elegans embryogenesis at
single-cell resolution. Science 365, eaax1971 (2019).
32. Cao, J. et al. Comprehensive single-cell transcriptional profiling of a multicellular
organism. Science 357, 661–667 (2017).
33. Kratsios, P. et al. An intersectional gene regulatory strategy defines subclass diversity of
C. elegans motor neurons. eLife 6, e25751 (2017).
34. Schneider, J. et al. UNC-4 antagonizes Wnt signaling to regulate synaptic choice in the C.
elegans motor circuit. Development 139, 2234–2245 (2012).
35. Hobert, O. Development of left/right asymmetry in the Caenorhabditis elegans nervous
system: from zygote to postmitotic neuron. Genesis 52, 528–543 (2014).
36. Serrano-Saiz, E. et al. Modular control of glutamatergic neuronal identity in C. elegans by
distinct homeodomain proteins. Cell 155, 659–673 (2013).
37. Serrano-Saiz, E., Oren-Suissa, M., Bayer, E. A. & Hobert, O. Sexually dimorphic
differentiation of a C. elegans hub neuron is cell autonomously controlled by a conserved
transcription factor. Curr. Biol. 27, 199–209 (2017).
38. Pereira, L. et al. A cellular and regulatory map of the cholinergic nervous system of C.
elegans. eLife 4, e12432 (2015).
39. Lloret-Fernández, C. et al. A transcription factor collective defines the HSN serotonergic
neuron regulatory landscape. eLife 7, e32785 (2018).
40. Doitsidou, M. et al. A combinatorial regulatory signature controls terminal
differentiation of the dopaminergic nervous system in C. elegans. Genes Dev. 27,
1391–1405 (2013).
41. Dobzhansky, T. Biology, molecular and organismic. Am. Zool. 4, 443–452 (1964).
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
© The Author(s), under exclusive licence to Springer Nature Limited 2020
Nature | www.nature.com | 7
Article
Methods
No statistical methods were used to predetermine sample size. The
experiments were not randomized and investigators were not blinded
to allocation during experiments and outcome assessment.
Homeobox gene list
Previous sequence analysis identified 103 C. elegans homeobox genes2.
A more recent evaluation of sequences revealed that one gene (ceh-85)
is a pseudogene (www.wormbase.org), which brings the total number
of homeobox genes considered here down to 102.
Generation of expression reagents
Previously reported expression patterns of homeodomain proteins
relied, in a few cases, on antibody staining, but the patterns of expres-
sion of these proteins in the nervous system were either incompletely
or not completely correctly identified (for example, VAB-7 and UNC-30,
which are revised in this Article), owing to a lack of molecular land-
marks for proper cellular identification. With only three exceptions
(ttx-3, unc-86 and unc-42, all of which used both fosmid and/or endog-
enous reporter alleles generated by CRISPR–Cas9), all other previously
reported expression patterns of homeobox genes were determined
using reporter transgenes that did not contain the entire gene locus,
which—as we show here—results in substantial underestimations of
expression patterns (summarized in Supplementary Tables 1, 2).
Here we examined the expression patterns of 20 homeodomain pro-
teins by tagging the respective endogenous locus with gfp via CRISPR–
Cas9 genome engineering. To this end, gfp was inserted at the 3′ end of
the gene, immediately before the stop codon. For vab-7, lin-11, ceh-37
and zfh-2, these reporter alleles were generated using the self-excising
cassette method for CRISPR–Cas9 genome engineering42. ceh-44 and
ceh-49 reporter alleles were provided by E. Leyva Díaz, and were gener-
ated as previously described43. CRISPR–Cas9-engineered strains with
the strain name PHX were created by Sunybiotech. Sixty homeodomain
proteins were examined using available chromosomally integrated
fosmid reporters lines generated by ModEncode (not previously exam-
ined for neuron-type-specific expression in the nervous system)44,
and an additional six homeodomain proteins were examined using
fosmid reporters (made by the ModEncode project44) that we injected
ourselves. All fosmid reporters included 3′ tagged protein fusions as
well. Injections were done into OH15430 (otis669;pha-1(e2123)) worms
at 10 ng/μl with 3 ng/μl pha-1(+) and 100 ng/μl OP50 genomic DNA to
create independent lines. A list of all reporter strains is provided below.
As expected from the usual compactness of C. elegans gene loci and
the size of fosmid reporters (about 40 kb of genomic sequence, usually
containing several genes up- and/or downstream of the gene of inter-
est), so far we have not found a single instance in which the fosmid
reporters do not fully recapitulate expression patterns observed with a
reporter allele generated by CRISPR–Cas9 genome engineering. Such
comparisons have been explicitly made with the transcription factors
unc-42 (E. Berghoff, pers. comm.), ttx-3 (V. Bertrand, pers. comm.),
lin-39 (ref. 45), unc-3(ref. 46) and che-1 (ref. 47).
Strain list for expression analysis
All newly generated strains used in this study are publicly available from the
Caenorhabditis Genetics Center. The strains for the respective homeobox
genes are as follows: alr-1: OP200; wgIs200 [alr-1::TY1::EGFP::3xFLAG +
unc-119(+)]; ceh-1: OP571; wgIs571 [ceh-1::TY1::EGFP::3xFLAG + unc-119(+)];
ceh-12: OH16368; otEx7486[ceh-12::TY1::EGFP::3xFLAG + unc-119(+)
+ pha-1(+)]; ceh-13: OH16366; otEx7484[ceh-13::TY1::EGFP::3xFLAG +
unc-119(+) + pha-1(+)]; ceh-14: OP73; wgIs73 [ceh-14::TY1::EGFP::3xFLAG
+ unc-119(+)]; ceh-16: OP82; wgIs82 [ceh-16::TY1::EGFP::3xFLAG + unc-
119(+)]; ceh-17: OH16369; otEx7487[ceh-17::TY1::EGFP::3xFLAG + unc-
119(+) + pha-1(+)]; ceh-18: OP533; wgIs533 [ceh-18::TY1::EGFP::3xFLAG
+ unc-119(+)]; ceh-19: OP739; wgIs739 [ceh-19::TY1::EGFP::3xFLAG +
unc-119(+)]; ceh-2: OP323; wgIs323 [ceh-2::TY1::EGFP::3xFLAG + unc-
119(+)]; ceh-20: RW12211; ceh-20(st12211[ceh-20::TY1::EGFP::3xFLAG]);
ceh-21, ceh-39, ceh-41: OP759; wgIs759 [ceh-41::TY1::EGFP::3xFLAG
+ unc-119(+)]; ceh-22: OP389; wgIs389 [ceh-22::TY1::EGFP::3xFLAG
+ unc-119(+)]; ceh-23: PHX1849; ceh-23(syb1849[ceh-23::GFP)]; ceh-
24: PHX1608; ceh-24(syb1608[ceh-24::GFP)]; ceh-27: OP135; wgIs135
[ceh-27::TY1::EGFP::3xFLAG + unc-119(+)]; ceh-28: OH16367; otEx748
5[ceh-28::TY1::EGFP::3xFLAG + unc-119(+) + pha-1(+)]; ceh-30: OP120;
wgIs120 [ceh-30::TY1::EGFP::3xFLAG + unc-119(+)]; ceh-31: OP370;
wgIs379 [ceh-31::TY1::EGFP::3xFLAG + unc-119(+)]; ceh-32: OH16477;
ceh-32(ot1040[ceh-32::GFP::LoxP::3x FLAG]) V; ceh-33: OP575; wgIs575
[ceh-33::TY1::EGFP::3xFLAG + unc-119(+)]; ceh-34: OP524; wgIs524
[ceh-34::TY1::EGFP::3xFLAG + unc-119(+)]; ceh-36: OP620; wgIs620
[ceh-36::TY1::EGFP::3xFLAG + unc-119(+)]; ceh-37: OH16345; ceh-37(ot10
23[ceh-37::GFP::FLAG]); ceh-38: OP241; wgIs241 [ceh-38::TY1::EGFP::3xFLAG
+ unc-119(+)]; ceh-40: OP232; wgIs232 [ceh-40::TY1::EGFP::3xFLAG + unc-
119(+)]; ceh-43: OH10447; otIs339 [ceh-43::gfp; ttx-3::dsred; rol-6]; ceh-
44: OH16219; ceh-44(ot1015[ceh-44::gfp]); ceh-45: OH16370; otEx748
8[ceh-45::TY1::EGFP::3xFLAG + unc-119(+) + pha-1(+)]; ceh-48: OP631;
wgIs631 [ceh-48::TY1::EGFP::3xFLAG + unc-119(+)]; ceh-49: OH16224;
ceh-49(ot1016[ceh-49::gfp]); ceh-5: PHX1592; ceh-5(syb1592[ceh-5::GFP)];
ceh-51: PHX1551; ceh-51(syb1551[ceh-51::GFP)]; ceh-53: OP444; wgIs
444 [ceh-53::TY1::EGFP::3xFLAG + unc-119(+)]; ceh-54: OP456; wgIs456
[ceh-54::TY1::EGFP::3xFLAG + unc-119(+)]; ceh-57: OP706; wgIs706 [ceh-57::
TY1::EGFP::3xFLAG + unc-119(+)]; ceh-58: PHX2015; ceh-58(syb2015[ceh-58::
GFP)]; ceh-6: RW10871; wgIs87[ceh-6::TY1::EGFP::3xFLAG + unc-119(+)]; ceh-
60: DLS395; ceh-60(rhd395 [HA-mCherry::ceh-60]); ceh-62: OP416;
wgIs416 [ceh-62::TY1::EGFP::3xFLAG + unc-119(+)]; ceh-63: OP742; wgIs741
[ceh-63::TY1::EGFP::3xFLAG + unc-119(+)]; ceh-7: OP168; wgIs681[ceh-7::
TY1::EGFP::3xFLAG + unc-119(+)]; ceh-74: OP680; wgIs680 [ceh-74::TY1::
EGFP::3xFLAG + unc-119(+)]; ceh-75: PHX1884; ceh-75(syb1884[ceh-75::GFP)];
ceh-76: OH16487; ceh-76(ot1042[ceh-76::GFP]) ceh-79: OP553; wgIs553 [ceh-
79::TY1::EGFP::3xFLAG + unc-119(+)]; ceh-8: PHX1656; ceh-8(syb1656
[ceh-8::GFP)]; ceh-81: OH16479; otEx7569 [ceh-81:TY1::EGFP::3xFLAG
+ unc-119(+)]; ceh-82: OP212; wgIs212 [ceh-82::TY1::EGFP::3xFLAG + unc-
119(+)]; ceh-83: OP727; wgIs727 [ceh-83::TY1::EGFP::3xFLAG
+ unc-119(+)]; ceh-86: PHX2517; ceh-86(syb2517[ceh-86::GFP)];
ceh-87: PHX1955; ceh-87(syb1995[ceh-87::GFP)]; ceh-88: OP593;
wgIs593 [ceh-88::TY1::EGFP::3xFLAG + unc-119(+)]; ceh-89: OH16505;
ceh-89(ot1050[ceh-89::GFP]); ceh-9: OP690; wgIs690 [ceh-9::TY1::
EGFP::3xFLAG + unc-119(+)]; ceh-90: OP210; wgIs210 [ceh-90::TY1::
EGFP::3xFLAG + unc-119(+)]; ceh-91: OH16480; otEx7570 [ceh-91:TY1::
EGFP::3xFLAG + unc-119(+)]; ceh-92: PHX1610; ceh-92(syb1610
[ceh-92::GFP)]; ceh-93: OP554; wgIs554 [ceh-93::TY1::EGFP::3xFLAG
+ unc-119(+)]; ceh-99: OH16481; otEx7571 [ceh-99:TY1::EGFP::3xFLAG
+ unc-119(+)]; ceh-100: OH16488; ceh-100(ot1043[ceh-100::GFP]);
cog-1: OP541; wgIs541 [cog-1::TY1::EGFP::3xFLAG + unc-119(+)]; dsc-1:
OP522; wgIs522[dsc-1::TY1::EGFP::3xFLAG + unc-119(+)]; duxl-1: OP470;
wgIs470 [duxl-1::TY1::EGFP::3xFLAG + unc-119(+)]; dve-1: OP398; wgIs398
[dve-1::TY1::EGFP::3xFLAG + unc-119(+)]; egl-5: OP54; wgIs54 [egl-5::
TY1::EGFP::3xFLAG + unc-119(+)]; eyg-1: OP441; wgIs441 [eyg-1::
TY1::EGFP::3xFLAG + unc-119(+)]; hmbx-1: OP655; wgIs655 [hmbx-1::TY1::
EGFP::3xFLAG + unc-119(+)]; irx-1: OP536; wgIs536 [irx-1::TY1::EGFP::3xFLAG
+ unc-119(+)]; lim-4: OP681; wgIs681 [lim-4::TY1::EGFP::3xFLAG + unc-119(+)];
lim-6: OP387; wgIs387 [lim-6::TY1::EGFP::3xFLAG + unc-119(+)]; lim-7: OP15;
wgIs15[lim-7::TY1::EGFP::3xFLAG + unc-119(+)]; lin-11: OH15910; lin-11(ot95
8[lin-11::GFP::FLAG]); lin-39: OP18; wgIs18 [lin-39::TY1::EGFP::3xFLAG + unc-
119(+)]; mab-5: OP27; wgIs27 [mab-5::TY1::EGFP::3xFLAG + unc-119(+)]; mec-
3: OP55; wgIs55 [mec-3::TY1::EGFP::3xFLAG + unc-119(+)]; mls-2: OP645;
wgIs654 [mls-2::TY1::EGFP::3xFLAG + unc-119(+)]; nob-1: JIM271;
stIs10286 [nob-1::GFP::unc-54 3′UTR + rol-6(su1006)]; nsy-7: OH16371;
otEx7489[nsy-7:TY1::EGFP::3xFLAG + unc-119(+) + pha-1(+)]; pal-1:
OP380; wgIs380 [pal-1::TY1::EGFP::3xFLAG + unc-119(+)]; pax-3: OP190;
wgIs190 [pax-3::TY1::EGFP::3xFLAG + unc-119(+)]; pha-2: OP687;
wgIs687 [pha-2::TY1::EGFP::3xFLAG + unc-119(+)]; php-3: PHX1549;
php-3(syb1548[php-3::GFP]); pros-1: OP500; wgIs500 [ceh-26::TY1::
EGFP::3xFLAG + unc-119(+)]; tab-1: PHX1587; tab-1(syb1587[tab-1::GFP)];
ttx-1: PHX1679; ttx-1(syb1679[ttx-1::GFP)]; unc-30: OP395; wgIs395 [unc-30::
TY1::EGFP::3xFLAG + unc-119(+)]; unc-39: OP186; wgIs186 [unc-39::
TY1::EGFP::3xFLAG + unc-119(+)]; unc-4: PHX1658; unc-4(syb1658[unc-4::
GFP)]; unc-62: SD1871; wgIs600 [unc-62::TY1::EGFP::3xFLAG + unc-119(+)];
vab-15: OP730; wgIs730 [vab-15::TY1::EGFP::3xFLAG + unc-119(+)]; vab-3:
FQ1092; wzEx302[vab-3::GFP + Pflp-17::DsRed]; vab-7: OH15912; vab-7(o
t959[vab-7::GFP::FLAG]); zag-1: OP83; wgIs83 [zag-1::TY1::EGFP::3xFLAG
+ unc-119(+)]; and zfh-2: OH16346; zfh-2(ot1024[zfh-2::GFP::FLAG]). The
unc-42 reporter lines will be described elsewhere (E. Berghoff and O.H.,
manuscript in preparation).
Microscopy
Worms were anaesthetized using 100 mM of sodium azide (NaN3) and
mounted on 5% agarose pads on glass slides. Images were acquired
using confocal laser scanning microscopes (Zeiss LSM800 and LSM880)
and processed using ImageJ software48. For expression of reporters,
representative maximum intensity projections are shown for the GFP
channel as greyscale, and gamma and histogram were adjusted for vis-
ibility. For mutant functional analysis, representative maximum inten-
sity projections are shown as an inverted greyscale. NeuroPAL images
provided in the Supplementary Information are pseudocoloured in
accordance with ref. 28. All reporter reagents and mutants were imaged
at 40× using fosmid or CRISPR reagents, unless otherwise noted.
Examination of expression reagents and neuron identification
Some obviously panneuronal or ubiquitous genes were determined
to be expressed in all neurons by crossing the reporter strain with
otIs314, a rab-3 fosmid driving TagRFP. For all of the remaining genes,
colocalization with the NeuroPAL landmark strain (otIs669 or otIs696)
was used to determine the identity of all neuronal expression28. For
CRISPR–Cas9-generated strains and integrated fosmid strains, the
reporter strain was crossed with the NeuroPAL landmark strain. To
analyse fosmid expression with available DNA but no integrated strain,
fosmid DNA was injected into the NeuroPAL landmark strain OH15430
(otIs669;pha-1(e2123)) as a rescuing array with pha-1(+) DNA. Three
extrachromosomal lines were created and analysed for each extrachro-
mosomal fosmid strain to determine the expression of that gene. Gener-
ally, the expression of a given reporter gene was stable over all worms
scored. In the few cases in which we observed variable expression of
fosmid reporter genes (for example, ceh-8 and ceh-24), we generated
reporter alleles by CRISPR–Cas9, resulting in more stable expression. In
terms of expression level, for every gene expressed in multiple neurons,
we noticed different levels of expression in different neuron classes
(Extended Data Figs. 1–8). Expression (even if very dim) was counted
as present if seen across multiple worms. This is because even the dim
expression of a homeodomain transcription factor has been shown
have functional phenotypes. For example, ceh-14 is bright in all neuron
types in which it is expressed except AFD and I2, but has previously been
shown to control the specification of the AFD neurons36,49.
We also noticed many cases of additional expression of
well-characterized homeobox genes, the expression of which had previ-
ously been studied with suboptimal reporter reagents. In some cases,
the new sites of expression are relatively dim, whereas in other cases
they are strong. Two such examples are a fosmid reporter of the LIM
homeobox mec-3, which is brightly expressed in previously identified
touch neurons50 and is less bright in posterior VA neurons (which we
describe here). By contrast, a CRISPR–Cas9-engineered reporter allele of
the unc-4 locus is—within the context of the ventral nerve cord—equally
bright in the previously identified VA and DA motor neuron classes51, as
it is in the AS motor neurons which we identify here as unc-4-expressing.
Although we did not notice obvious differences in expression pat-
terns between late larval stages and adult, a number of genes are clearly
expressed in additional cells in the embryo.
Clustering using the Jaccard index
To assess the similarities among neuron classes by homeobox genes,
we used the Jaccard index. This index is used to measure similarity
between finite sample sets by calculating the intersection of those sets
divided by their union. For our data, we calculated the number of shared
homeobox genes between each neuron class in a pairwise manner, and
then divided them by the number of shared and unshared homeobox
genes in those pairs. To cluster this data, we created a distance matrix
for the degree of dissimilarity between each neuron class based on
their homeobox gene codes, calculated as 1 − Jaccard similarity index.
With this distance matrix, we clustered our data using the hierarchical
clustering tool hclust (available in R), an open source software environ-
ment for statistical computing.
We did this same analysis for the degree of similarity among home-
obox genes by their expression in shared neuron classes. In this calcula-
tion, the number of shared neuron classes between each homeobox
gene was counted in a pairwise manner, and then divided by the number
of shared and unshared neuron classes in which those genes expressed.
We again created a distance matrix (1 − Jaccard index), clustered the
data using hclust.
Minimal code of homeobox genes
Given a set of redundant codes of homeodomain coexpression for each
neuron, we aim to reduce this codebook to one in which there are no
redundancies and each cell is represented by a unique barcode. The
problem of codebook reduction is cast as a multidimensional knapsack
problem52 with binary weight constraints. The goal of this problem
is to find the minimum set of homeodomain codes such that no two
neurons share the same barcode. The global optimum solution is then
found through a branch-and-bound scheme53 that yields the minimum
subset of bits that can be conserved from the genetic codebook and
that ensures uniqueness of cell barcodes.
Correlation of homeobox and reporter expression
We used a Wormbase-curated list of 1,126 published reporter transgenes
available, with new homeobox gene expression data added in Supplemen-
tary Table 5. To test for correlations between reporter transgene expres-
sion in specific neurons and homeobox gene expression, we removed all
homeobox gene expression profiles from the Wormbase-curated list. We
then performed a simple linear regression using the lm function in R: we
fitted lm(G ~ TF), in which G is the reporter expression by neuron class
matrix and TF is the homeobox expression by neuron class 1. To assess
the goodness of our fit, we also shuffled the homeobox expression matrix
1,000 times. This gave us an R2 value of 0.74 for our actual homeobox
expression dataset, which compared favourably to the 0.41 achieved
with the control shuffled homeobox expression dataset. We then set
out to verify how good this correlation was across individual neuron
classes, as the number of available reporters they express is variable.
The fitted values from the above regression predict an expected reporter
expression for each neuron class, on the basis of their homeobox gene
expression. For each neuron class, we extracted these fitted values and
compared them to the actual transgene expression profiles reported
using the correlation function in R (cor) using the standard Pearson
method. These correlation values are shown in Extended Data Fig.11a.
Mutant analysis scoring and statistics
Reporter expression was scored as an all-or-nothing phenotype per
neuron, with expression in 0, 1 or 2 neurons. Scoring data was processed
in R and converted as contingency tables of the number of expressing
neurons by genotype. Statistical analysis was then done using Fisher’s
exact test (under a two-sided null hypothesis), using Holm’s method to
correct for multiple comparisons. The resulting adjusted P values are
all below 0.001. No statistical methods were used to determine sample
size before the experiment. On basis of the common standard in the
Article
field, we aimed for about 30 worms per genotype for neurotransmitter
reporters and about 15 worms for other markers.
ceh-32(ok343) mutant worms arrested at L1 were maintained with an
otEx7146 ceh-32 fosmid rescue construct. Worms were only counted as
ceh-32 mutants when the myo-2::mCherry coinjection marker of this array
was not visible at all. The ceh-32 mutants arrested at L1 were scored against
their wild-type counterpart strain at L1, rather than with the rescued worm
of the same strain. Owing to the disorganization of their head ganglions,
glutamatergic identity in RIA neurons was instead scored using a short inte-
grated eat-4 promoter fragment (otIs521) with a restricted expression pat-
tern in only a subset of glutamatergic neurons36. Scoring was done under
a Zeiss stereo dissecting scope at high magnification, and representative
images from confocal microscopy are shown at 63×. One or two very dim
cells were seen in less than 15% of the ceh-32-mutant worms under confocal
microscopy, but these cells made no axonal projection and their cell body
did not match the shape of RIA neurons. Reported P values would still be
significant if they were conservatively counted as eat-4(+) RIA neurons.
For the mutant analysis, the following strains were used: OH13094
otIs354[cho-1fos::YFP]; otIs518 [eat-4fos::mCherry]; OH15958
otIs354[cho-1fos::YFP]; otIs518 [eat-4fos::mCherry]; ceh-8(gk116531);
IK705(njIs10[glr-3p::GFP]; OH15970 njIs10[glr-3p::GFP]; ceh-8(gk116531);
OH4793 otIs173 [F25B3.3::DsRed2 + ttx-3pB::GFP]; otEx980 [dop-2::GFP
+ pha-1(+)]; OH16478 otIs173 [F25B3.3::DsRed2 + ttx-3pB::GFP]; otEx980
[dop-2::GFP + pha-1(+)]; ceh-8(gk116531); OH16253 otIs354[cho-1fos::YFP];
ot907(unc-17::mKate2 CRISPR); OH16251 otIs354[cho-1fos::YFP];
ot907(unc-17::mKate2 CRISPR), ceh-9(tm2747); OH16256 otIs580 [cho-1
fos::mCherry + eat-4fos::YFP]; OH16201 otIs580 [cho-1fos::mCherry + eat-
4fos::YFP] ceh-31(tm239); OH16204 otIs92[ flp-10p::GFP]; OH16203
otIs92[flp-10p::GFP]; ceh-31(tm239); OH12525 otIs521[eat-4prom8::tagRFP;
ttx-3::gfp]; OH16314 otIs521[eat-4prom8::tagRFP; ttx-3::gfp],
otIs388[eat-4fos::YFP], ceh-32(ok343) otEx7146[ceh-32 fosmid rescue
WRM0637dA10 + myo-2 RFP]); IK705 njIs10[glr-3p::GFP]; and OH16476
ceh-32(ok343) V; njIs10[glr-3p::GFP]; otEx7146[ceh-32 fosmid rescue
WRM0637dA10 + myo-2 RFP].
Comparison of homeobox expression with single-cell
RNA-sequencing data
To analyse the congruence between available single-cell RNA sequencing
(scRNA-seq) data31,32 and our reported homeodomain expression, we used
the provided bootstrap median data (averaging resampled RNA levels
1,000 times) from refs. 31,32, and applied no cut off (that is, any transcripts
per million value >0 counted as real expression). We then directly com-
pared the binary expression profiles of the homeobox gene mRNA in iso-
lated neuron classes with our reported homeodomain protein expression
(coloured in keys in the figures). We found that the scRNA-seq expression
data from the 42 identified L2 neuron classes recapitulated only 38% of
our homeodomain protein expression. We calculated this percentage by
taking the agreed expression (blue) and dividing it by the agreed expres-
sion plus the expression seen only in the homeodomain protein analysis
(blue + red). We then asked whether scRNA-seq was able to detect mRNA
of our homeodomain proteins at earlier embryonic time points. To this
end, we added the scRNA-seq embryo data available for these 42 neuron
classes, and found that this increased the coverage to 55%. This percentage
was calculated as above with the agreed expression divided by the agreed
plus the expression seen only in the homeodomain protein analysis.
Reporting summary
Further information on research design is available in the Nature
Research Reporting Summary linked to this paper.
Data availability
All newly generated data, including the expression pattern of every
homeobox gene, are available in Supplementary Tables 1, 2. Addition-
ally, whole-worm confocal images of all homeobox genes analysed are
available in Extended Data Figs. 1–8. Newly generated reporter strains
made during this study are available from the Caenorhabditis Genet-
ics Center. The most up-to-date version of the community-curated
transgene expression resource used is available in Supplementary
Table 5.
Code availability
The R code used to generate the Jaccard distance matrix for the cluster-
ing of homeobox genes and neuron classes is available on the GitHub
of the O.H. laboratory, at https://github.com/hobertlab/Reilly_2020/
tree/master/Jaccard_Distance. The MATLAB code used to create the
minimal codebook of homeobox genes is available at https://github.
com/hobertlab/Reilly_2020/tree/master/Minimal_Codebook.
42. Dickinson, D. J., Pani, A. M., Heppert, J. K., Higgins, C. D. & Goldstein, B. Streamlined
genome engineering with a self-excising drug selection cassette. Genetics 200, 1035–
1049 (2015).
43. Dokshin, G. A., Ghanta, K. S., Piscopo, K. M. & Mello, C. C. Robust genome editing with
short single-stranded and long, partially single-stranded DNA donors in Caenorhabditis
elegans. Genetics 210, 781–787 (2018).
44. Sarov, M. et al. A genome-scale resource for in vivo tag-based protein function
exploration in C. elegans. Cell 150, 855–866 (2012).
45. Feng, W. et al. A terminal selector prevents a Hox transcriptional switch to safeguard
motor neuron identity throughout life. eLife 9, e50065 (2020).
46. Patel, T. & Hobert, O. Coordinated control of terminal differentiation and restriction of
cellular plasticity. eLife 6, e24100 (2017).
47. Leyva-Díaz, E. & Hobert, O. Transcription factor autoregulation is required for acquisition
and maintenance of neuronal identity. Development 146, dev177378 (2019).
48. Schindelin, J. et al. Fiji: an open-source platform for biological-image analysis. Nat.
Methods 9, 676–682 (2012).
49. Cassata, G. et al. The LIM homeobox gene ceh-14 confers thermosensory function to the
AFD neurons in Caenorhabditis elegans. Neuron 25, 587–597 (2000).
50. Way, J. C. & Chalfie, M. The mec-3 gene of Caenorhabditis elegans requires its own
product for maintained expression and is expressed in three neuronal cell types. Genes
Dev. 3 (12A), 1823–1833 (1989).
51. Miller, D. M., III & Niemeyer, C. J. Expression of the unc-4 homeoprotein in
Caenorhabditis elegans motor neurons specifies presynaptic input. Development 121,
2877–2886 (1995).
52. Kellerer, H., Pferschy, U. & Pisinger, D. Knapsack Problems (Springer, 2004).
53. Schrijver, A. Theory of Linear and Integer Programming (John Wiley & Sons, 1998).
54. Mukaka, M. M. Statistics corner: a guide to appropriate use of correlation coefficient in
medical research. Malawi Med. J. 24, 69–71 (2012).
Acknowledgements We thank Q. Chen for expert technical assistance in generating
transgenic lines; E. Berghoff for communicating the unpublished expression of unc-42;
E. Leyva-Díaz for CRISPR-tagging ceh-44 and ceh-48; R. Dowen for sending a ceh-60 reporter
allele; V. Bertrand for communicating unpublished results on ceh-10 and ttx-3 reporter alleles;
L. Glenwinkel for updating and sharing the community-curated gene expression resource;
J. Booth for creation of select extrachromosomal fosmid reporter strains; Y. Ramadan for help
with mutant analysis; and T. Bürglin for comments on the manuscript. This work was funded by
a predoctoral fellowship to M.B.R. (F31 NS105398), by NIH R21 NS106843, and by the Howard
Hughes Medical Institute. Some strains were provided by the CGC, which is funded by NIH
Office of Research Infrastructure Programs (P40 OD010440).
Author contributions M.B.R. and O.H. designed the experiments and wrote the manuscript,
M.B.R. generated constructs and conducted the expression pattern analysis, C.C. conducted
genetic loss-of-function experiments and contributed to writing the paper, E.V. and E.Y.
conducted the bioinformatic analysis and contributed to writing the paper.
Competing interests The authors declare no competing interests.
Additional information
Supplementary information is available for this paper at https://doi.org/10.1038/s41586-020-
2618-9.
Correspondence and requests for materials should be addressed to O.H.
Peer review information Peer reviewer reports are available.
Reprints and permissions information is available at http://www.nature.com/reprints.
 ANTP-HOXL homeodomain expression patterns in the L4 nervous system
HEAD Anterior MIDBODY Posterior MIDBODY TAIL

RMG
AS1 VD
DA1 VA2 DD DA VD VA VD DA VD
VB3 VD VA VB V
VA DD VD VA DD VD VA VA
DD1 AS2 VD3 DA2 VA2 DD2 VD DA VB AS DDVD VA VD DA
VA
DA VD
VD2

Anterior MIDBODY Posterior MIDBODY TAIL

SDQ

AVM PDE

V
PVD
AS V
AQR VB
VB DB DD
VA AS
DB2 DD VD DB VD
AS DD VD VD VA DD VD VA
AIYs AS2 DB3 DA VA VB VC DD AS VB VC AS DB VD DA
DB AS VD VA VB DA VC

HEAD Anterior MIDBODY Posterior MIDBODY TAIL

PQR
AS VD
VB1 VB2 AS1 VD2 VA2 VB VC VA VC DA
AVL AS2 VA DA VB VC
DD AS VD VA VB DA AS VD DB VA VB VA VB DADB AS VD VA AS VD DD
VB3 VD3 AS VC
DB DD AS VD
VA DA VAVB VA
VC VB VD

HEAD Anterior MIDBODY Posterior MIDBODY TAIL

AWA
DVB LUA PVC PLM
VA
DA
PHB
DD VD
HSN PDB PDA
V

HEAD MIDBODY TAIL

DVA
PVN
PVW
PVN
V PLN PHC

HEAD Anterior MIDBODY Posterior MIDBODY TAIL

DVC
PVR PVN
DVA
PVW
PLN PHC
V

HEAD Anterior MIDBODY Posterior MIDBODY TAIL

ADA
DB PVWs
VC DB VC PHCs
DB2 DB1
DB
V

ANTP-HOXL expression patterns Outside of L4 Nervous system

EMBRYO

ANTP homeodomain expression patterns in the L4 nervous system

HEAD MIDBODY TAIL

AIZ

RIGs
RIFs
V
HEAD MIDBODY TAIL

HEAD MIDBODY TAIL

HEAD MIDBODY TAIL

I3
M3s
NSM
V

HEAD MIDBODY TAIL

I3

HEAD MIDBODY TAIL

VB2 VB3 VB11

VB1 VB4 VB VB VB VB
VB
V

HEAD MIDBODY TAIL

PDE
CEPD SDQ
PVQ
URB IL1D SDQ
AIZ BDU CAN
IL1 IL1V ASJ
ADE
CEPVs
V
MIDBODY TAIL

ADL ASG
CAN
ADF AWA
BAG AFD PHB
AIY PHA
AWC
V

HEAD MIDBODY TAIL

URYD
DVC

V PVWs
URYV

HEAD MIDBODY TAIL

I4

ANTP expression patterns outside of L4 Nervous system

HEAD MIDBODY TAIL

early L1 EMBRYO

Extended Data Fig. 1 | ANTP and HOXL homeodomain protein expression worm. Heads, midbodies and tails of each reporter are shown and labelled
patterns. Neuronal cell identifications with NeuroPAL are provided for one accordingly. The outer body of the worm and pharynx are outlined in white. All
member of each class and shown as the GFP reporter alone, the NeuroPAL images are of L4 or young adult worms unless otherwise noted as an expression
landmark alone and a merged image of the GFP reporter with the NeuroPAL outside of the L4 nervous system. Ten worms were analysed for each reporter
landmark. Neurons are circled in yellow and the identities of those neurons by strain and characteristic images were chosen.
NeuroPAL identifier are immediately beside them. V indicates the vulva of the
Article
ANTP-NKL homeodomain expression patterns in the L4 nervous system
MIDBODY TAIL

RIV
AWA
V

HEAD MIDBODY TAIL

PVN
RIV

SAAV AWB
ADF V
ASJ
SAAD

HEAD MIDBODY TAIL

ADF
MC
PHA

V
HEAD MIDBODY TAIL

RMED

SIBD SIBV V
SMBV
RMEV SIAD
SMBD
SIAV

HEAD MIDBODY TAIL

RMED
I6
RIP RME
RIM
RMEV RMFs
V

HEAD MIDBODY TAIL

M4

HEAD MIDBODY TAIL

URAD
URB AVB PVR

URAV V

HEAD MIDBODY TAIL

ADL

ASER DVB
AIA AIZ
ASJ PHB
RMD VA12
RMF SM AVL
Bs
PDA
V
HEAD MIDBODY TAIL
ASK
AVJ
V
AIM

HEAD MIDBODY TAIL

AIN
AVD
AIB

HEAD MIDBODY TAIL

M5 PVC
PVQ
PVT
AVG

Ubiquitous ANTP-NKL expression patterns

HEAD MIDBODY TAIL

ANTP-NKL expression patterns outside of L4 Nervous system

HEAD Anterior MIDBODY Posterior MIDBODY TAIL

EMBRYO

Extended Data Fig. 2 | NKL homeodomain protein expression patterns.
Neuronal cell identifications with NeuroPAL are provided for one member of
each class and shown as the GFP reporter alone, the NeuroPAL landmark alone
and a merged image of the GFP reporter with the NeuroPAL landmark. Neurons
are circled in yellow and the identities of those neurons by NeuroPAL identifer
are immediately beside them. V indicates the vulva of the worm. Heads,
midbodies and tails of each reporter are shown and labelled accordingly. The
outer body of the worm and pharynx are outlined in white. All images are of L4
or young adult worms unless otherwise noted as an expression outside of the
L4 nervous system. Ten worms were analysed for each reporter strain and
characteristic images were chosen.
 CUT homeodomain expression patterns in the L4 nervous system
Panneuronal CUT Expression Patterns
Anterior MIDBODY Posterior MIDBODY TAIL

HEAD Anterior MIDBODY Posterior MIDBODY TAIL

HEAD Anterior MIDBODY Posterior MIDBODY TAIL

HEAD Anterior MIDBODY Posterior MIDBODY TAIL

Ubiquitous CUT expression Patterns

HEAD Anterior MIDBODY Posterior MIDBODY TAIL

HEAD Anterior MIDBODY Posterior MIDBODY TAIL

CUT expression Patterns - compass subfamily

HEAD MIDBODY TAIL

CAN

DD1
AVG VD3 VD13
VD4 V VD12
VD1&2 DD2 VD11 PVPs
VD VD VD DD VD

CUT expression Patterns outside of L4 nervous system

GERMLINE EMBRYO

LIM homeodomain expression patterns in the L4 nervous system

HEAD MIDBODY TAIL

RIA M1
URYD
RME AVB
AFD PVC
BAG AIB PVN
URYV AIM PHC
V

HEAD MIDBODY TAIL

HEAD MIDBODY TAIL

HEAD MIDBODY TAIL

BDU
ALA DVC
PVR
AFD PHB PVW
PVT PVQ
ADA PVNs
I2 AIM V PHA PVC PHC

HEAD MIDBODY TAIL

SMDV RIV
RID AWB
SAAV SIBs
SM RMH
B
RM
F SMB SIAs V

HEAD MIDBODY TAIL

RMEs DVB
ASEL

AVL RIS
V PVT
RIGs

HEAD MIDBODY TAIL

ADL ASG
AVH
ADF AVD DVA
RIM
AVA ASH RIC PVQ
AVE AVFs RIFs
AVG PVPs
VC
VC V

HEAD Anterior MIDBODY Posterior MIDBODY TAIL

PLMs
AVM PVD PVM
VA
FLP
VA VA
V VA
VA

Extended Data Fig. 3 | CUT and LIM homeodomain expression patterns. midbodies and tails of each reporter are shown and labelled accordingly. The
Neuronal cell identifications with NeuroPAL are provided for one member of outer body of the worm and pharynx are outlined in white. All images are of L4
each class and shown as the GFP reporter alone, the NeuroPAL landmark alone or young adult worms unless otherwise noted as an expression outside of the
and a merged image of the GFP reporter with the NeuroPAL landmark. Neurons L4 nervous system. Ten worms were analysed for each reporter strain and
are circled in yellow and the identities of those neurons by NeuroPAL identifier characteristic images were chosen.
are immediately beside them. V indicates the vulva of the worm. Heads,
Article
POU homeodomain expresssion patterns in the L4 nervous system
HEAD Anterior MIDBODY Posterior MIDBODY TAIL

MI AVM DVA
IL1 OLQ RID AVH DVB
MC AV PVD SDQ ALN
SAA D
AVA AIN
URY BDU PVQ PVR
AVB PVM PVC
AIB
RIC HSN VA PVP VA DAs
IL1Vs OLQ SAA
AUA
FLP
DB AS
VD PLN
s RIH ADA AS
SMB RIF SAB AVG DDVA AS DA DD VA VB AS DD VD DD VD PDB PDA PLM
s SIA VD VA VB AS DB DB VA VB DD VA VB VD VD
s
VB V VA VB AS DA VD DB
AS
DA
VA VB
DB VB SAB VD DA DA DD AS VD DA AS VD DB AS

HEAD Anterior MIDBODY Posterior MIDBODY TAIL

V
HEAD Anterior MIDBODY Posterior MIDBODY TAIL

HEAD MIDBODY TAIL

AVH

RMDV AUA
I5
PDA
RMDD AVFs

PRD homeodomain expression patterns in the L4 nervous system

HEAD MIDBODY TAIL

OLQ CEPD
RIV ALM
AVD PLMs
BAG SAAV
SAAD
CEPV RIF DB VB AS
SMBs VB DB VB DB VA
VB VB V
VB VB

HEAD MIDBODY TAIL

HEAD MIDBODY TAIL

PRD expression patterns outside of L4 nervous system

MIDBODY TAIL

HEAD MIDBODY TAIL

SIX homeodomain expression patterns in L4 nervous system

HEAD Anterior MIDBODY Posterior MIDBODY TAIL

ED
URYD RM RID
IL1D SMDV
OLL RME RIA
I2 IL1
IL1V RMDV RIB
RMHs
URYV RM
EV RM V
DD SMDD
V

HEAD MIDBODY TAIL

M1 I6
M5
I4
MI M4 M2s
I3 M3
MC I5
NSM
I2
V

HEAD MIDBODY TAIL

IL2
AIA

SIX expression patterns outside of L4 nervous system

HEAD MIDBODY TAIL

Extended Data Fig. 4 | POU, PRD and SIX homeodomain protein expression worm. Heads, midbodies and tails of each reporter are shown and labelled
patterns. Neuronal identifications with NeuroPAL are provided for one accordingly. The outer body of the worm and pharynx are outlined in white. All
member of each class and shown as the GFP reporter alone, the NeuroPAL images are of L4 or young adult worms unless otherwise noted as an expression
landmark alone, and a merged image of the GFP reporter with the NeuroPAL outside of the L4 nervous system. Ten worms were analysed for each reporter
landmark. Neurons are circled in yellow and the identities of those neurons by strain and characteristic images were chosen.
NeuroPAL identifier are immediately beside them. V indicates the vulva of the
 PRD-like homeodomain expression patterns in the L4 nervous system
Anterior MIDBODY Posterior MIDBODY TAIL

CEPD M1
AVM
URX
OLL ASEL PDE
I6 ALM PLM
MI RME
AVA AFD AIY
I2 BAG VD VD VD VD VD
CEPV VD VD VD
V VD
VD VD

HEAD Anterior MIDBODY Posterior MIDBODY TAIL

HEAD Anterior MIDBODY Posterior MIDBODY TAIL

HEAD Anterior MIDBODY Posterior MIDBODY TAIL

ALA

SIAD SIAV
V

HEAD MIDBODY TAIL

ASI

AWA ASE
AWC

V
HEAD MIDBODY TAIL

ASG
AWB
AWA
BAG
V

HEAD MIDBODY TAIL

MI
I1

HEAD MIDBODY TAIL

I5

HEAD Anterior MIDBODY Posterior MIDBODY TAIL

ASG

BAG AFD AWA

M3 ASE
I2 ASH
AWC
AUA
V

HEAD MIDBODY TAIL

RIA

AVE

HEAD MIDBODY TAIL

AFD ASE

AWC

HEAD MIDBODY TAIL

RIP M2

AFD RIB

V
HEAD MIDBODY TAIL

ASG

AVJ
DD
DD VD VD
VD VD
VD VD VD DD VD PVPs
VD DD V
DD

HEAD MIDBODY TAIL

I5
DA
SABV AVF VA DA AS VA DA
VA VD VA AS DA VA AS
AVF SABD VA VD AS VA AS
DA DA VD
V AS
VA

Extended Data Fig. 5 | PRD-like homeodomain protein expression worm. Heads, midbodies and tails of each reporter are shown and labelled
patterns. Neuronal cell identifications with NeuroPAL are provided for one accordingly. The outer body of the worm and pharynx are outlined in white. All
member of each class and shown as the GFP reporter alone, the NeuroPAL images are of L4 or young adult worms unless otherwise noted as an expression
landmark alone and a merged image of the GFP reporter with the NeuroPAL outside of the L4 nervous system. Ten worms were analysed for each reporter
landmark. Neurons are circled in yellow and the identities of those neurons by strain and characteristic images were chosen.
NeuroPAL identifier are immediately beside them. V indicates the vulva of the
Article
HNF and PROS homeodomain expression patterns in the L4 nervous system
MIDBODY TAIL

PVD
URX ALM
PLM
FLP

HEAD MIDBODY TAIL

HEAD MIDBODY TAIL

HEAD MIDBODY TAIL

URYD ASG
AWA DVA
I3 PVQ
URYV
I2

TALE homeodomain expression patterns in the L4 nervous system

HEAD MIDBODY TAIL

ASG
M4 M2s
I5
RIH DD PVQ
RIS AVFs
AVG PVT
DD DD
SABV DD V

HEAD MIDBODY TAIL

HEAD MIDBODY TAIL

HEAD Anterior MIDBODY Posterior MIDBODY TAIL

ALM
ALA I4
AIB
AIZ FLP ADE ADA PDE PQR
PHA
RMG VC PVD PVM DD PDB
VB DB VA VB SDQ AS
RIF RIG VC VD VD
V VC
SABV RIF DA VD AS DD
DA VA VB VC VA VBDA DD AS
VA VC AS VA DA
RIG DB
SAB AS
VA DB AS VD DA VA VB DD VD VB
VB
DB VD

HEAD Anterior MIDBODY Posterior MIDBODY TAIL

ALM
I4 DVA
ALA PVD PDE
ADA PVM PHA
AIBs AIZ ADE RMG VDDA
FLP AS DB VC SDQ VA VB AS
VA VB VD DA VA VB AS VC V DA VC AS DD VD VA VB VA DA PDB
RIC
VB VB RIF VD AS VA DB DB VC DB DA VD VA AS VD
VD AS VD VA VB
RIG DA
SABVs DB DB DA

TALE expression patterns outside of L4 nervous system

HEAD MIDBODY TAIL

HEAD MIDBODY TAIL

ZF homeodomain expression patterns in the L4 nervous system

HEAD MIDBODY TAIL

RID ALA
IL1 OLQ
CEP SMD DVA
URY SAA AWB LUA PVN
AFD PVM
OLQ RMD ASG
URY AV PVD
ASIBs BDU VA PDB PHB PQR
NSM ASH FLP ADE PDE
AIB
CE 1
RIR Ps

PHA
IL

SIBAA Ds

RIM VA VB DA DB AS VA
S M

SMB AS AS DA PDA
R

Bs
SM H

VB DA DB RIG AS AS
AS DB AS V DA VA VB
RM

VB VA VB DA DB AS VA VB VA VB DB AS DA
VA DB VA VB AS
SAB SAB VB

HEAD MIDBODY TAIL

HEAD MIDBODY TAIL

HEAD MIDBODY TAIL

ALA AVM
RME D RIV M1
SM AVH AIN DVA
AV
RMD AIBAWA J M2 DVB
DVC
PVW
M4 AVD RMG LUA
RIP A RMD RIB I5 ADA PVN
RME AV E PVQ
V SIB SIA RIC AVK AVK PDA PVC
RME A RMD AIZ AS VD PVP
PVP
PLM
AIM RIFs RIG RIG VD VD
RIH AVG PVT
AW B

AVF VD VC
C

VC
SI

DD VD
SIA

AVFs VD DD AS VC AS
SABs AS DD VC DD VD VD VD
AS AS V

Extended Data Fig. 6 | HNF, PROS, TALE and ZF homeodomain vulva of the worm. Heads, midbodies and tails of each reporter are shown and
protein expression patterns. Neuronal cell identifications with NeuroPAL are labelled accordingly. The outer body of the worm and pharynx are outlined in
provided for one member of each class and shown as the GFP reporter alone, white. All images are of L4 or young adult worms unless otherwise noted as an
the NeuroPAL landmark alone and a merged image of the GFP reporter with the expression outside of the L4 nervous system. Ten worms were analysed for
NeuroPAL landmark. Neurons are circled in yellow and the identities of those each reporter strain and characteristic images were chosen.
neurons by NeuroPAL identifier are immediately beside them. V indicates the
 Divergent homeodomain expression patterns in the L4 nervous system
Anterior MIDBODY Posterior MIDBODY TAIL

MI ASI M1
PHB
I2 AWA
AFD ASE
AWC

HEAD Anterior MIDBODY Posterior MIDBODY TAIL

HEAD Anterior MIDBODY Posterior MIDBODY TAIL

HEAD MIDBODY TAIL

ASI
ADF
ASG

ASJ
V

HEAD MIDBODY TAIL

ASI

HEAD MIDBODY TAIL

ADL

HEAD MIDBODY TAIL

URY OLQ CEP CAN

M4 RMD ADF AVH DVA PVC
ASH

OLL AVA AWA RIB VD DA DA PVP DA

URY RMD
A VA VB DD AS AS LUA
OLQ P RMD VE ASJ AIM AVG VA DB DB VA PHB
CE IR VD VA VB AS VD DA DB VB AS DD VC V VA VB DB VD VD PHA
R IH AWC VA RIG VA VB DA VD VC DA AS VD DA PVP
R DB
VBs SABs DD RIG

Panneuronal Divergent expression patterns

HEAD MIDBODY TAIL

HEAD Anterior MIDBODY Posterior MIDBODY TAIL

Ubiquitous Divergent expression patterns

HEAD MIDBODY TAIL

HEAD Anterior MIDBODY Posterior MIDBODY TAIL

V
HEAD MIDBODY TAIL

Divergent expression patterns outside of L4 Nervous system

EMBRYOs
EMBRYOs EMBRYOs

EMBRYOs EMBRYOs L1 EMBRYOs

EMBRYOs

HEAD Anterior MIDBODY Posterior MIDBODY TAIL

HEAD TAIL EMBRYO

Extended Data Fig. 7 | Divergent homeodomain protein expression worm. Heads, midbodies and tails of each reporter are shown and labelled
patterns. Neuronal cell identifications with NeuroPAL are provided for one accordingly. The outer body of the worm and pharynx are outlined in white. All
member of each class and shown as the GFP reporter alone, the NeuroPAL images are of L4 or young adult worms unless otherwise noted as an expression
landmark alone and a merged picture of the GFP reporter with the NeuroPAL outside of the L4 nervous system. Ten worms were analysed for each reporter
landmark. Neurons are circled in yellow and the identities of those neurons by strain and characteristic images were chosen.
NeuroPAL identifier are immediately beside them. V indicates the vulva of the
Article

Extended Data Fig. 8 | Homeobox gene expression ordered by lineage and of expression by the Jaccard index (as in Fig. 3). Neurons are further coloured by
neurotransmitter identity. Representation of homeobox gene expression their neurotransmitter identity: acetylcholine, red; glutamate, yellow; GABA,
pattern with neurons ordered by their lineage and genes ordered by similarity blue; amine, green; unknown, grey.
Extended Data Fig. 9 | Features of the homeobox gene code. a, The
70 conserved homeobox genes alone are sufficient to codify all neuron classes.
Neuron classes are coloured by neuron type (sensory, blue; motor, pink;
interneuron, yellow; and pharyngeal, grey) and ordered by similarity between
neuron classes defined by the Jaccard index as in Fig. 2b. Homeobox genes are
coloured by subfamily, and ordered by similarity of neuron class expression
and sparsity. b, Theoretical minimal code of conserved homeobox genes
required to distinguish every neuron class (determined mathematically,
as described in Methods). Colours are as in a.
Article
Extended Data Fig. 10 | Homeobox codes define neuronal subclasses.
a, Homeobox codes subdivide neuron classes in the head. Images of head
neuron classes are provided from http://www.WormAtlas.org. Homeobox
genes expressed in each neuron are listed. Yellow colour indicates the
homeobox gene is expressed in all members of that neuron class. Orange and
green colours indicate that the homeobox gene is expressed in only a subset of
neuron class members. b, Tabular representation of homeobox gene
expression subdividing members of the ventral nerve cord motor neurons. Red
and blue boxes correspond to cholinergic and GABAergic neurotransmitter
identity, respectively. Green boxes indicate the gene is a member of the HOX
subfamily; grey boxes indicate that a gene is expressed in some neurons of the
VNC; black boxes indicate that a gene is expressed in all neurons of the VNC.
Extended Data Fig. 11 | See next page for caption.
Article
Extended Data Fig. 11 | Homeobox genes affecting neuronal marker gene
expression. a, Correlation between predicted gene expression patterns and
actual expression patterns in each neuron class. Actual expression patterns for
each neuron class were determined by a community-curated list of 1,132
fluorescent reporter expression patterns in all 118 neuron classes
(Supplementary Table 5). Predicted gene expression was found by multivariate
linear regression of known reporter expression patterns using homeobox gene
expression atlas. Correlation coefficient was calculated by Pearson, in which a
coefficient of 0.5–0.7 is moderate, 0.7–0.9 is a strong and 0.9–1.0 is a very
strong correlation, consistent with ref. 54. b–e, Additional characteristic images
for the quantification shown in Fig. 4a–c. We analysed about 15 independent
worms per genotype, acquired over at least two separate imaging sessions
with an equal mix of both genotypes. Characteristic images were chosen.
b, Expression of dop-2, a dopamine receptor, is lost in RIA in the ceh-8-mutant
worm. c, Expression of flp-10, a neuropeptide, is lost or becomes dimmer in
PVR in the ceh-31-mutant worm. d, e, Expression of glr-3, a glutamate receptor,
is lost in RIA in both the ceh-8- (c) and the ceh-32- (d) mutant worms. f, Summary
of effects of the loss of homeobox gene on neuronal identity throughout entire
nervous system of C. elegans, based on previous studies (specified in
Supplementary Table 2) and the mutant analysis conducted here. Dark grey
boxes indicate gene activator function; black boxes indicate repressor
function.
 nature research | reporting summary
Corresponding author(s): Oliver Hobert
Last updated by author(s): Feb 25, 2020

Reporting Summary
Nature Research wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency
in reporting. For further information on Nature Research policies, see Authors & Referees and the Editorial Policy Checklist.

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.
n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement
A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly
The statistical test(s) used AND whether they are one- or two-sided
Only common tests should be described solely by name; describe more complex techniques in the Methods section.

A description of all covariates tested

A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons
A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient)
AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals)

For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted
Give P values as exact values whenever suitable.

For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings
For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes
Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated
Our web collection on statistics for biologists contains articles on many of the points above.

Software and code

Policy information about availability of computer code
Data collection Zeiss Black software was used to collect multicolor confocal images.

Data analysis An open source image processing software, FIJI version 2.0.0-rc.69, was used to process confocal images. An open source software
environment for statistical computing, R version 1.1.383, was used for all statistical tests including clustering, correlation, and t-tests.
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors/reviewers.
We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Research guidelines for submitting code & software for further information.

Data
Policy information about availability of data
All manuscripts must include a data availability statement. This statement should provide the following information, where applicable:
- Accession codes, unique identifiers, or web links for publicly available datasets
- A list of figures that have associated raw data
- A description of any restrictions on data availability

All newly generated data from this study is available in Supplementary Tables S2 and S3, and corresponding confocal images showing expression of every gene are
October 2018

available in Supplemental Figures S1-S8. All new reporter strains generated for this study will be available for order from the Caenorhabditis Genetics Center. The
most updated community-curated transgene expression resource used for comparison is also available in Supplemental Table S5.

1
 nature research | reporting summary
Field-specific reporting
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.
Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Life sciences study design

All studies must disclose on these points even when the disclosure is negative.
Sample size For expression pattern analysis between 10 and 15 transgenic animals were analyzed per gene of interest. No statistical test was used to
determine this sample size, but the high quality of the endogenous reporters yielded little to no variability across independent worms. For
functional analysis, between 10 and 30 mutant animals were analyzed in various reporter backgrounds. No statistical test was used to
determine this sample size, but the effect of mutant genes were so penetrant that it was easily scoreable and consistent with the current (n)
in the field.

Data exclusions No data was excluded.

Replication No steps were taken because expression pattern analysis was completed in relatively invariable reporter strains. And functional data was
acquired over multiple days and imaging sessions.

Randomization For functional analysis, groups were allocated based on genotype, confirmed either by PCR and double checked or based on fluorescent
marker expression for rescued strains.

Blinding Study was not blinded because mutant phenotypes were easily observable.

Reporting for specific materials, systems and methods

We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material,
system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response.

Materials & experimental systems Methods

n/a Involved in the study n/a Involved in the study
Antibodies ChIP-seq
Eukaryotic cell lines Flow cytometry
Palaeontology MRI-based neuroimaging
Animals and other organisms
Human research participants
Clinical data

Animals and other organisms

Policy information about studies involving animals; ARRIVE guidelines recommended for reporting animal research
Laboratory animals Caenorhabditis elegans - N2 Bristol strain. L4 to young adult stage hermaphrodites were imaged. Males and hermaphrodites
were used to generate new strains via crossing.

Wild animals Not applicable

Field-collected samples Not applicable

Ethics oversight No ethical oversight was required due to nature of organism.

Note that full information on the approval of the study protocol must also be provided in the manuscript.
October 2018