DOI: 10.1126/science.1247761
, 1256 (2014);343Science
et al.Bela S. Desai
Neurons
Gene Is Essential for Mechanical Sensing in ProprioceptivestumThe
This copy is for your personal, non-commercial use only.
clicking here.colleagues, clients, or customers by
, you can order high-quality copies for yourIf you wish to distribute this article to others
here.following the guidelines
can be obtained byPermission to republish or repurpose articles or portions of articles
):March 16, 2014www.sciencemag.org (this information is current as of
The following resources related to this article are available online at
http://www.sciencemag.org/content/343/6176/1256.full.html
version of this article at:
including high-resolution figures, can be found in the onlineUpdated information and services,
http://www.sciencemag.org/content/suppl/2014/03/12/343.6176.1256.DC1.html
can be found at:Supporting Online Material
http://www.sciencemag.org/content/343/6176/1256.full.html#ref-list-1
, 4 of which can be accessed free:cites 12 articlesThis article
http://www.sciencemag.org/cgi/collection/genetics
Genetics
subject collections:This article appears in the following
registered trademark of AAAS.
is aScience2014 by the American Association for the Advancement of Science; all rights reserved. The title
CopyrightAmerican Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005.
(print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by theScience
onMarch16,2014www.sciencemag.orgDownloadedfromonMarch16,2014www.sciencemag.orgDownloadedfromonMarch16,2014www.sciencemag.orgDownloadedfromonMarch16,2014www.sciencemag.orgDownloadedfromonMarch16,2014www.sciencemag.orgDownloadedfrom
Together with the lineage analysis, these observations
strongly suggest that the limb bud initiates
earlier than stage 17 to 18, through EMTand independent
of proliferation rate changes. When the
limb bud mesenchyme is generated, it induces a
source of fibroblast growth factor (Fgf) activity in
the overlying ectoderm, the apical ectoderm ridge,
which serves to maintain the limb’s high level of
proliferation (3). This does not occur in the trunk
region, which we hypothesize is the reason for
its relative decrease in mitotic activity as mesenchyme
is generated. We suggest that the difference
in proliferation of trunk versus limb bud
mesenchyme is a result of limb bud initiation, as
opposed to being a cause of it.
To verify that EMT of the somatopleure represents
a necessary step of limb initiation, we
blocked EMT in the presumptive limb region by
RhoA overexpression, which abrogates EMT by
inducing strong interaction of cells with extracellular
matrix components (4). We electroporated
RhoAwith GFPwithin the epithelial somatopleure.
In control embryos electroporated with GFP alone,
limb buds formed normally (Fig. 2A), the basement
membrane of laminin broke down, and cells
underwent EMT (Fig. 2, B to D). In contrast,
coelectroporation of RhoA with GFP completely
abrogated limb formation (Fig. 2, E and I).
Sections revealed that RhoA electroporated cells
were stuck in the epithelial somatopleure and
failed to undergo proper EMT (Fig. 2, F to H).
These cells were attached to aggregates of laminin
(Fig. 2, K and L) and did not exhibit enrichment in
vimentin staining (fig. S5, G to L). We found that
RhoA-overexpressing cells overexpressed F-actin,
maintained adherens junction as revealed by
N-cadherin and b-catenin stainings (fig. S5, A
to L), and failed to relocalize aPKC from their
apical cortex to the cytoplasm (fig. S5, A to F) as
observed in control GFP-electroporated cells
(fig. S4, N, O, R, and S). RhoA overexpression
did not lead to dramatic reduction of proliferation
(fig. S6, A to C) or increase in apoptosis in
the electroporated cells (fig. S6, D to G), confirming
that RhoA acts to prevent epithelialto-mesenchymal
cell state change. Thus as a
consequence of this failure to undergo EMT, the
electroporated cells did not participate in the initiation
and formation of the limb primordium.
To understand how EMT is regulated in this
context, we explored the possibility that Snail1,
a transcription factor upstream of EMT in other
contexts (5), plays a similar role here, but (perhaps
because of redundancy) we failed to find
such a connection. We therefore turned to factors
known to be involved in establishing the formation
of a limb bud. Ectopic Fgf protein, applied
up to stage 17, is sufficient to induce the formation
of an entire additional limb from trunk tissue
(6). To test whether Fgf10 promotes limb formation
by inducing EMT, we coelectroporated Fgf10
and GFP into the trunk somatopleure of stage-13
to -14 chick embryos. As expected, 36 hours after
electroporation Fgf10 induced swelling of the
trunk, indicating ectopic limb initiation (Fig. 3A,
red arrowheads). Sectioning through the electroporated
region showed that most of the GFPpositive
cells had left the epithelium and had
acquired a mesenchymal phenotype (81%, Fig. 3,
B to D), showing that indeed EMT had taken
place. The trunk is only competent to form an
ectopic limb up to stage 16 to 17 (7, 8). This was
previously interpreted as the time at which the
trunk mesenchyme becomes determined and is
no longer capable of being redirected to a limb
fate. However, our data show that this is precisely
the time at which the trunk mesenchyme is first
generated. Thus, we would reinterpret those results
as indicating that ectopic Fgf activity can
induce limb bud formation from epithelial trunk
somatopleure cells but not from mesenchymal
cells of the same rostrocaudal level.
Targeted mutation of Fgf10 and Tbx5 in mice
have demonstrated that these genes are necessary
to initiate limb bud formation (9–12). Transverse
sections of embryonic day 9.5 Fgf10−/−
and Tbx5−/−
embryos confirmed that neither Tbx5 nor Fgf10
mutant embryos exhibit swellings characteristic
of limb bud initiation (Fig. 3, E, H, and K). Both
FGF10 and Tbx5 mutant embryos showed the
presence of mesenchyme in the forelimb region;
however, the proportion of mesenchymal cells
compared to the proportion of epithelial cells was
significantly lower than that of a wild-type (WT)
sibling embryo (Fig. 3N), with a stronger phenotype
observed in Tbx5−/−
embryos (only 12% of
cells were epithelial in WTembryos, whereas 21%
and 51% were epithelial in Fgf10- and Tbx5deficient
embryos, respectively). In Tbx5 mutant
embryos, the epithelium appeared separated from
the mesenchyme (Fig. 3K, asterisks). aPKC and
b-catenin staining revealed hyperplasia of the
somatopleure epithelium, in support of failure of
these cells to undergo EMT (Fig. 3, K and L).
Last, in Fgf10−/−
as well as in Tbx5−/−
embryos,
the basement membrane of laminin did not properly
break down and appeared overstabilized, as
opposed to WTembryos (Fig. 3, F, G, I, J, L, and
M). Taken together, these data show that Tbx5
and Fgf10 act on the somatopleure epithelium to
regulate, at least partially, the early induction of
EMT in the limb fields, the process that is at the
heart of limb bud initiation.
References and Notes
1. R. L. Searls, M. Y. Janners, Dev. Biol. 24, 198–213 (1971).
2. A. Mauger, J. Embryol. Exp. Morphol. 28, 313–341
(1972).
3. H. Ohuchi et al., Development 124, 2235–2244 (1997).
4. Y. Nakaya, E. W. Sukowati, Y. Wu, G. Sheng, Nat. Cell
Biol. 10, 765–775 (2008).
5. J. P. Thiery, H. Acloque, R. Y. J. Huang, M. A. Nieto, Cell
139, 871–890 (2009).
6. M. J. Cohn, J. C. Izpisúa-Belmonte, H. Abud, J. K. Heath,
C. Tickle, Cell 80, 739–746 (1995).
7. H. Ohuchi et al., Biochem. Biophys. Res. Commun. 209,
809–816 (1995).
8. A. Vogel, C. Rodriguez, J. C. Izpisúa-Belmonte,
Development 122, 1737–1750 (1996).
9. P. Agarwal et al., Development 130, 623–633 (2003).
10. C. Rallis et al., Development 130, 2741–2751 (2003).
11. H. Min et al., Genes Dev. 12, 3156–3161 (1998).
12. K. Sekine et al., Nat. Genet. 21, 138–141 (1999).
Acknowledgments: We thank F. Constantini for help
rederiving the Fgf10 mutant mouse line, B. Bruneau for kindly
providing Tbx5 mutant embryos, and J. deMelo and P. Tschopp
for help with mouse work. J.G. was a fellow of the Human
Frontier Science Program. This work was supported by NIH
grant R01-HD045499 to C.J.T.
Supplementary Materials
www.sciencemag.org/content/343/6176/1253/suppl/DC1
Materials and Methods
Supplementary Text
Figs. S1 to S6
Reference (13)
7 November 2013; accepted 7 February 2014
10.1126/science.1248228
The stum Gene Is Essential
for Mechanical Sensing
in Proprioceptive Neurons
Bela S. Desai, Abhishek Chadha, Boaz Cook*
Animal locomotion depends on proprioceptive feedback, which is generated by mechanosensory
neurons. We performed a genetic screen for impaired walking in Drosophila and isolated a
gene, stumble (stum). The Stum protein has orthologs in animals ranging from nematodes to
mammals and is predicted to contain two transmembrane domains. Expression of the mouse
orthologs of stum in mutant flies rescued their phenotype, which demonstrates functional
conservation. Dendrites of stum-expressing neurons in legs were stretched by both flexion and
extension of corresponding joints. Joint angles that induced dendritic stretching also elicited
elevation of cellular Ca2+
levels—not seen in stum mutants. Thus, we have identified an
evolutionarily conserved gene, stum, which is required for transduction of mechanical stimuli in
a specific subpopulation of Drosophila proprioceptive neurons that sense joint angles.
A
nimal locomotion is achieved by coordination
of motor activity according to
proprioceptive mechanosensory inputs.
In Drosophila, mechanosensation is mediated
either by ciliated or multidendritic receptor neurons.
Multidendritic neurons can respond to direct
14 MARCH 2014 VOL 343 SCIENCE www.sciencemag.org1256
REPORTS
application of mechanical force to their membranes
(1, 2). It is less clear, however, how multidendritic
mechanosensory neurons can be tuned
to one mechanical modality, such as joint angle,
and disregard other mechanical stimuli that may
originate from external impacts or changes in
the shape of muscles during contraction. In order
to identify genes involved in proprioceptive sensation,
we screened for uncoordination in a collection
of ethyl methanesulfonate–mutagenized
Drosophila lines (3). Lines that exhibited walking
impairments were selected, and the phenotype
severity was quantified by measuring climbing
speed. We identified three lines, 204, 922,
and 4487, that showed lack of coordination in
homozygous flies and did not complement one
another (Fig. 1A), which suggested that they
represent alleles of the same gene. Two of the
lines, 204 and 4487, showed severe uncoordination,
and the phenotype in 922 was mild (Fig.
1A). Deficiency mapping pointed to a gene,
CG30263 (Fig. 1B), predicted to encode a large
protein with 1870 or 1959 amino acids (depending
on the splice variant). Because the mutants
had a walking impairment phenotype, we named
this gene stumble (stum). The stum ortholog in humans
is C1orf95, and it was categorized as a member
of the SPEC3 family (UNIPROT, INTERPRO),
with unknown function. We sequenced the stum
gene in the three mutant lines and found that
stum204
, stum4487
, and stum922
had stop codons at
amino acid positions 171, 202, and 1081, respectively
(Fig. 1C).
We generated transgenic flies that express
stum cDNA in neurons and found that the stum
Department of Molecular and Cellular Neuroscience,
The Scripps Research Institute, La Jolla, CA 92037, USA.
*Corresponding author. E-mail: bcook@scripps.edu
Fig. 2. Expression of stum in legs. (A) Brightfield image of a mesothoracic (middle) leg (anterior
view, D, dorsal; V, ventral; L, lateral; M, medial). (B) GFP fluorescence showing stum-positive neurons.
(i), (ii), and (iii) are high-magnification views of the corresponding regions in (B). Arrowhead, cell
body; short arrow, dendritic tip; and long arrow, side branch.
Fig. 1. The stum gene underlies walking uncoordination. (A) Climbing speed of
Drosophila with homozygous stum alleles and transheterozygote lines. Flies having
three alleles were significantly slower than controls (P < 0.01). Complementation tests
showed that stum922
heterozygotes were not different from stum922
homozygotes, yet they
are significantly different from homozygous stum204
and stum4487
(P < 0.05). (B) Schematic
representation of the stum gene in the 57F8 region of chromosome 2R. The dashed
line designated stum-Gal4 represents the regulatory genomic region of the gene. The
dashed line that encompasses the coding region (green bars) illustrates the genomic
arrangement of splice variant RB. (C) Schematic representation of the Stum protein in
Drosophila and mouse. Red bars designate mutations in the three stum alleles. Green
bars show the position of predicted transmembrane domains. aa, amino acids. (D) Expression
of stum rescues the climbing phenotype of the homozygous stum mutant. For both
fly and mouse rescues, the flies that had both the driver (elav-Gal4) and the cDNA were
significantly faster than the controls having only one of the transgenes (P < 0.01). For
(A) and (D), significance was determined using a Mann-Whitney nonparametric test and
was adjusted by Bonferroni correction for multiple comparisons. Error bars represent
SEM, n > 8 for each data point in (A) and (D).
www.sciencemag.org SCIENCE VOL 343 14 MARCH 2014 1257
REPORTS
phenotype was partially rescued (Fig. 1D), which
indicated that the mutations in stum were underlying
the uncoordination phenotype. We noted that
the mouse ortholog of stum (National Center for
Biotechnology Information, the U.S. National Insti-
tutesofHealth,referencesequence:NP_001074696.1),
which is only 141 amino acids long and shares 33%
sequence identity with Drosophila stum, was also
able to substantially rescue the uncoordination
phenotype (Fig. 1D). Therefore, the function of
stum appears to be conserved between distant
animal species. Moreover, the rescue of the phenotype
with such a short form of stum suggests that
the C-terminal region of the fly protein constitutes
the functional core.
Proprioceptive defects in adult flies have
been attributed to malfunction of type I (ciliated)
mechanoreceptor neurons (4–6). To test whether
such defects also underlie the phenotype in the
stum mutant, we performed electrophysiological
recordings of mechanical responses from
the ciliated mechanoreceptor neuron of the anterior
notopleural bristle (6). Type I neurons of
stum mutants were indistinguishable from controls
(fig. S1). Therefore, unlike known proprioception
mutants, the phenotype in stum mutants
does not arise from a general defect in type I
mechanoreceptor neurons. To identify which cells
give rise to the phenotype, we used the genomic
regulatory region of stum to drive a CD8 fused
to green fluorescent protein (CD8GFP) reporter
[stum-Gal4 driving the upstream activation sequence
(UAS)–CD8GFP]. We found that stum
expression in the legs was localized to three
labeled neurons: one at the femur-tibia joint, the
second at the tibia-tarsus joint, and the third
spanning the second tarsal segment (Fig. 2). The
cell bodies of these stum-expressing neurons were
located near the distal end of each leg segment,
and their dendrites terminated at the corresponding
joints (Fig. 2).
To study whether there are stum-expressing
cells within the ventral nerve cord (VNC), we
examined it in flies that express CD8GFP using
stum-Gal4. We found that the only fluorescent
signal in the VNC originated from the axons of
the leg neurons (fig. S2). The axons terminated
within the neuropil that corresponds to each particular
leg, branching into a bowl shape (fig.
S2). This pattern is typical of neurons that take
part in proprioception, such as the hair plate
neurons (7–9). Therefore, the stum-expressing
cells in the Drosophila body have characteristics
of proprioceptive neurons that sense a property
of specific joints.
We performed confocal imaging of the dendritic
region of stum-expressing neurons and
found that, close to its tip, the dendrite branches
toward the lateral aspect of the joint, and this side
branch terminates at a short distance from the
cuticle (Fig. 2). The tips of dendrite were not associated
with a cuticular structure or a scolopale.
These structural features are typical features of
type II (multidendritic) neurons (10) but are incompatible
with type I mechanoreceptor neurons
that terminate with a ciliary structure. Thus,
stum uncoordination mutations affect type II
mechanoreceptor neurons. Furthermore, the entire
dendritic terminal of these neurons was located
in a region that is devoid of musculature
(fig. S3), which suggests that they do not sense the
mechanical properties of muscles. Taken together,
these data suggest that stum-expressing neurons
sense a mechanical property of the joint.
To test whether stum-positive neurons encode
joint angles, we performed high-resolution
imaging of the tibia-tarsus joint area at different
angles. At each angle of the joint, we measured
the total length of the sensory dendrite and its
side branch. We found that the total dendrite
length had a minimum typically at 130° to 170°,
and it increased when the joint was shifted to
either more obtuse or more acute angles (Fig. 3).
These morphological changes indicate that these
neurons are mechanically affected by the position
of the joint. Because the tip of the side
branch is stationary, it is likely that the change
in total length results from the coupling of the
tip of the main dendrite to the motion of the
distal joint segment. The position of the dendrite
and the susceptibility of its morphology to
joint angle suggest that the role of stum-positive
neurons is to sense and encode the angle of
the joint.
We measured Ca2+
fluorescence while forcing
the tibia-tarsus joint to different angles and found
that the Ca2+
fluorescence in stum-expressing
neurons correlated with the angle of the joint
(Fig. 4). As in the morphological changes, the responses
correlated with joint angle in a U-shaped
manner, where both acute and obtuse angles induced
increasing Ca2+
elevations. These results
indicate that the stum-positive neurons encode
proprioceptive information about the angle of
joints. It is noteworthy that a similar U-shaped
encoding of joint angles was also described in
receptor neurons of mammalian joints (11), which
suggests that sensing the deviations from a neutral
joint range is universally critical for motor
function.
The walking impairment in the stum mutant
fly suggests that the gene is necessary for generating
the proper proprioceptive responses in
stum-expressing neurons. To test whether the
stum mutations affect coupling between joint
Fig. 3. Dendrite stretching of stumexpressing
neurons. (A) Fluorescence
of stum-driven CD8GFP at the
tibia-tarsus joint. Dotted lines connect
the medial and lateral ends of the
tibia. Note that, at both flexion (60°)
and extension (220°), the tip of the
dendrite (arrow) is distal to its position
at the neutral angle (150°). (B
and C) Summary of dendritic length
changes at different joint angles. The
genotype of control flies was w;stumGal4;UAS-CD8GFP
and the genotype
of tested stum mutants was w;stum204,
stum-Gal4;UAS-CD8GFP. The length
change ratio (Dl/l) was determined
by dividing the dendritic length at a
given angle by the length at the baseline
(120° to 140°). Error bars represent
standard errors, n < 6 for each
data point. Data were transformed
by using a square root transformation
and tested for normality using the
D’Agostino and Pearson omnibus normality
test. *P < 0.05, **P < 0.01,
***P < 0.001, one-way analysis of
variance (ANOVA) with post hoc
Bonferroni multiple comparison test.
14 MARCH 2014 VOL 343 SCIENCE www.sciencemag.org1258
REPORTS
angles and dendritic stretching, we quantified
the stretching in mutant flies that have stumexpressing
neurons labeled with CD8GFP. We
found that, in the stum mutant, the dendritic
stretching in response to joint angles was comparable
to that of control flies (Fig. 3). Thus,
stum is not required for the mechanical coupling
between joint angles and stretching of the sensory
dendrites.
Although the dendritic stretching was not
significantly affected in the stum mutant, we
found that the Ca2+
responses to both acute and
obtuse joint angles were abolished in the mutant
(Fig. 4). Therefore, we conclude that stum is
essential for transducing dendrite stretching into
cellular responses.
We examined the morphology of stumexpressing
neurons and found that, although
axons (fig. S2) and cell bodies (fig. S4) of stum
mutants were indistinguishable from controls,
the sensory dendrite in mutants exhibited abnormalities
(fig. S4). Most notably, in some of
the stum-expressing neurons the tip of the dendrite
was overgrown and extended into the distal
segment of the joint (fig. S4). Thus, the absence of
stum leads to a morphological defect, possibly
because of the lack of mechanical responsiveness.
The occasional morphological differences
may account for the slight difference in the
stretching profile between control and mutant
dendrites (Fig. 3, B and C). The fraction of
neurons demonstrating the abnormal morphology
in the mutant increased from the day of
eclosion to the following day (fig. S4C). As the
addition of morphological changes takes place
after eclosion, it is possible that stum-dependent
activity is essential for late shape determination
that can take place in adult multidendritic
neurons (12).
We generated transgenic flies that express a
GFP fused with the N terminus of Stum under
UAS regulation (UAS-GFP-Stum). We found
that the Stum fusion protein was specifically localized
to the distal part of the sensory dendrite,
although it did not accumulate substantially in
any other part of the cell (fig. S5). The fluorescent
signal started at the region of bifurcation
and extended to both distal tips of the dendrite
(fig. S5). This specific localization suggests that
Stum functions in the part of the dendrite that
senses stretching.
Taken together, stum expression in mechanosensory
neurons, Stum localization to the sensory
dendrite, and the abolition of responses to
stretching in the stum mutant suggest that stum
has an essential role in mediating mechanical
sensing in receptor neurons. Because the Stum
protein in most species is very small and because
Drosophila stum is expressed in limited populations
of receptor neurons, we propose that stum
is not the mechanically activated channel. Rather,
stum may serve as an accessory module that is
essential for the proper localization or function
of the transduction channels.
The stretch-receptor neurons that express stum
present an elegant engineering solution for generating
specificity to the modality of mechanical
stimulus. The distal part of their dendrite bifurcates
into two branches whose tips are anchored
to parts of the joint that shift their relative positions.
Sensing the stretching only between the two
dendritic tips may tune the nerve responses to
joint motions and filter out the effect of irrelevant
mechanical impacts. This specificity enables the
sensory neuron to relay reliable proprioceptive
information to the central nervous system.
References and Notes
1. S. E. Kim, B. Coste, A. Chadha, B. Cook, A. Patapoutian,
Nature 483, 209–212 (2012).
2. Z. Yan et al., Nature 493, 221–225 (2013).
3. E. J. Koundakjian, D. M. Cowan, R. W. Hardy, A. H. Becker,
Genetics 167, 203–206 (2004).
4. T. Avidor-Reiss et al., Cell 117, 527–539 (2004).
5. M. Kernan, D. Cowan, C. Zuker, Neuron 12, 1195–1206
(1994).
6. R. G. Walker, A. T. Willingham, C. S. Zuker, Science 287,
2229–2234 (2000).
7. D. J. Merritt, R. K. Murphey, J. Comp. Neurol. 322,
16–34 (1992).
8. R. K. Murphey, D. Possidente, G. Pollack, D. J. Merritt,
J. Comp. Neurol. 290, 185–200 (1989).
9. R. K. Murphey, D. R. Possidente, P. Vandervorst, A. Ghysen,
J. Neurosci. 9, 3209–3217 (1989).
10. R. Bodmer, Y. N. Jan, Roux’s Arch. Dev. Biol. 196, 69–77
(1987).
11. D. Burke, S. C. Gandevia, G. Macefield, J. Physiol. 402,
347–361 (1988).
12. K. Shimono et al., Neural Dev. 4, 37 (2009).
Acknowledgments: We thank K. Spencer for technical
assistance. The stum mutants were identified by screening
a collection of F3 Drosophila lines in the laboratory of
C. S. Zuker. We thank L. Stowers and T. Avidor-Reiss for
kindly providing reagents. Data for this paper may be found
in the supplementary materials.
Supplementary Materials
www.sciencemag.org/content/343/6176/1256/suppl/DC1
Materials and Methods
Figs. S1 to S5
References
28 October 2013; accepted 6 February 2014
10.1126/science.1247761
Fig. 4. Responses of stum-expressing neurons. (A) Fluorescence of stumdriven
calcium indicator GCaMP at the tibia-tarsus joint. Two joint angles
are presented (the region of the dashed box in the schematic drawing below),
and the contour of the neuron is highlighted with a dashed yellow line. Arrows
show the side branch of the dendrite, and the arrowheads indicate the location
of the cell body. Fluorescence is represented by an arbitrary scale that increases
from green to red. Note that at 220° the dendrite is stretched distally,
which results in an increase of fluorescent area over the dendrite and elevated
fluorescence in the cell body (red signal). (B) A summary of fluorescence levels
of the cell body at different joint angles. The genotype of control flies was w;
stum-Gal4;UAS-GCaMP3, and the genotype of tested stum mutants was w;stum204
,stum-Gal4;UAS-GCaMP3. The DF/F was determined by dividing the
fluorescence value at a given angle by the fluorescence at the preceding baseline (120° to 150°). Error bars represent standard errors, n > 6 for each data
point. Baseline and response fluorescence levels were compared pairwise for each data point. *P < 0.05, **P < 0.01, Wilcoxon signed-rank test.
www.sciencemag.org SCIENCE VOL 343 14 MARCH 2014 1259
REPORTS