Hox Repertoires for Motor Neuron
Diversity and Connectivity Gated
by a Single Accessory Factor, FoxP1
Jeremy S. Dasen,1,3,* Alessandro De Camilli,1Bin Wang,2Philip W. Tucker,2and Thomas M. Jessell3,*
1Smilow Neuroscience Program, Department of Physiology and Neuroscience, New York University School of Medicine, New York,
NY 10016, USA
2Department of Molecular Genetics and Microbiology, The University of Texas at Austin, Austin TX 78712, USA
3Howard Hughes Medical Institute, Kavli Institute for Brain Science, Departments of Neuroscience, and Biochemistry and Molecular
Biophysics, Columbia University, New York, NY 10032, USA
*Correspondence: email@example.com (J.S.D.), firstname.lastname@example.org (T.M.J.)
The precision with which motor neurons innervate
target muscles depends on a regulatory network of
Hox transcription factors that translates neuronal
identity into patterns of connectivity. We show that
a single transcription factor, FoxP1, coordinates mo-
tor neuron subtype identity and connectivity through
its activity as a Hox accessory factor. FoxP1 is ex-
pressed in Hox-sensitive motor columns and acts
as a dose-dependent determinant of columnar fate.
Inactivation of Foxp1 abolishes the output of the
system to an ancestral state. The loss of FoxP1 also
changes the pattern of motor neuron connectivity,
and in the limb motor axons appear to select their
trajectories and muscle targets at random. Our find-
ings show that FoxP1 is a crucial determinant of
motor neuron diversification and connectivity, and
clarify how this Hox regulatory network controls the
formation of a topographic neural map.
The versatility of motor behaviors relies on the ability to activate,
on demand, a select few of the many hundred skeletal muscle
groups. Motor neurons (MNs) lie at the core of this action plan.
Each muscle is innervated by a dedicated set of MNs, and
many of the inputs to the spinal cord are designed to activate
MNs in temporal patterns that meet the mechanical require-
ments of motor performance. How the neurons that implement
these motor programs are assembled into functional circuits
remains poorly understood.
During development spinal MNs segregate into discrete col-
umns, each innervating a different peripheral domain (Figure 1A).
Median motor column (MMC) neurons innervate axial muscles,
hypaxial motor column (HMC) neurons innervate body wall
muscles, preganglionic motor column (PGC) neurons innervate
sympathetic ganglia, and lateral motor columns (LMC) innervate
limb musculature (Fetcho, 1992; Gutman et al., 1993; Land-
messer, 2001). Within these columns, MNs exhibit finer-grained
positional identities that are also matched with the location of
their targets (Laskowski and Sanes, 1987; McHanwell and Bis-
coe, 1981; Gutman et al., 1993). The developmental logic that
underlies nerve-muscle connectivity is best understood for
neurons of the LMC, which acquire divisional and pool identities
that determine their axonal trajectory and muscle target within
the limb (Jessell, 2000).
The construction of this topographic motor map is directed by
innervation (Figures 1B and 1C). Graded Hedgehog and Wnt4/5
signals control the dorsoventral expression of homeodomain
(HD) transcription factors that specify MMC and HMC fates
(D. Agalliu and T.M.J., unpublished data; Briscoe and Ericson,
2001). In contrast, PGC and LMC neurons are specified by a
rostrocaudal FGF signaling gradient that establishes regional
ify brachial LMC neurons, Hox9 paralogs specify PGC neurons,
and Hox10 proteins specify lumbar LMC neurons (Dasen et al.,
2003; Shah et al., 2004; Wu et al., 2008). Within the LMC,
changing the profile of Hox expression in specific pools results in
corresponding changes in the pattern of muscle innervation (Da-
sen et al., 2005). This core Hox network appears to direct motor
innervation patterns by activating a diverse array of downstream
transcriptionfactors andcellsurfacereceptors(De Marco Garcia
and Jessell, 2008; Livet et al., 2002; Kania and Jessell, 2003).
motor columns induced by the dorsoventral and rostrocaudal
signaling pathways emerged at different stages of evolution, in
parallel with the elaboration of peripheral target structures. Early
aquatic vertebrates with simple locomotor behaviors that are
driven by axial and hypaxial muscles appear to possess neurons
of MMC and HMC character but lack PGC and LMC neurons
(Fetcho, 1992; Kusakabe and Kuratani, 2005). The appearance
304 Cell 134, 304–316, July 25, 2008 ª2008 Elsevier Inc.
of LMC and PGC neurons is linked to the formation of paired
appendages (lateral fins and limbs) and a sympathetic nervous
system - structures that emerged later in vertebrate evolution
(Fetcho, 1992; Freitas et al., 2006; Funakoshi and Nakano,
2007). The Hox-dependent program of MN diversification may
set of peripheral target tissues, and more elaborate motor
The workings of the motor neuron Hox regulatory network
remain obscure. One puzzle is that Hox proteins are expressed
by all spinal MNs, yet PGC and LMC neurons alone depend on
Hox function (Dasen et al., 2003), implying a constraint on Hox
activity within certain motor columns. It is also unclear whether
Hox programs of columnar specification involve common down-
stream effectors. Consequently, there has not been an effective
formation of topographic motor maps. We reasoned that insight
intotheseissuesmight emergefromanexamination ofthespinal
motor system under conditions in which the entire Hox-depen-
dent program of MN subtype differentiation has been inacti-
vated, while leaving intact the Hox-independent program. Such
a perturbation might consign spinal MN differentiation to an
ancestral vertebrate state, yet encase these primitive neurons
in a body that is hundreds of millions of years more advanced.
Figure 1. Expression of FoxP1 in Hox-
Dependent Motor Columns
(A) Organization of Hox proteins, motor columns
paxial motor column (after Fetcho, 1992, formerly
lateral MMC; Gutman et al., 1993); PGC: pregan-
glionic motor column; LMC: lateral motor column.
Symp: sympathetic chain ganglion neurons.
(B) 21 Hox proteins assign spinal MN identity.
(C) Hox interactions specifying MN identity (Dasen
et al., 2005).
(D–F) FoxP1 in e12.5 spinal cord. Inset: magnified
(G) FoxP1 levels in motor columns and dorsal
spinal cord (dSC). Similar Foxp1 mRNA levels are
detected (Figures S1K–S1P).
(H) Profile of columnar subtypes in thoracic (Th)
(I–L) FoxP1 and markers of Th motor columns.
(M) Markers of MN columnar subtypes at brachial
(Br) and lumbar (Lu) levels.
(N and O) Expression of FoxP1 in RALDH2+LMC
neurons but not Lhx3+MMC neurons or V2
neurons. Dotted line in L separates MNs and inter-
(P) FoxP1 and Hox expression in MN columns.
Analysis of patterns of motor innervation
in such atavistic chimeras could reveal
how Hox regulatory networks enable
MNs to innervate their targets with spec-
ificity. Achieving this condition through
direct genetic perturbation of Hox genes
is unrealistic given the involvement of
so many. We therefore searched for
other ways to disable the Hox-dependent program of MN
Hox proteins typically rely on transcriptional cofactors that re-
fine and constrain their activities (Mann and Affolter, 1998). Two
Drosophila cofactors, Extradenticle [Exd] and Homothorax [Hth]
(Meis and Pbx/Prep proteins in vertebrates), have pervasive
roles as regulators of Hox activity (Mann and Affolter, 1998;
Moens and Selleri, 2006). In the spinal cord Meis and Pbx/Prep
proteins have broad patterns of expression (Dasen et al.,
2005), and thus they are unlikely candidates as cell-type specific
regulators of MN Hox activity. Recent studies in Drosophila have
identified a distinct group of Hox ‘accessory’ factors, notably the
HD protein Engrailed and the forkhead proteins Slp1/2 which
work together with Exd and Hth to gate Hox activities (Gebelein
et al., 2004). These factors have more restricted domains of ex-
pression and activity: Engrailed regulates Hox activity in poste-
rior compartment cells whereas Slp1/2 regulate Hox activity in
anterior cells (Gebelein et al., 2004). Their vertebrate counter-
parts, the Engrailed and Fox proteins, are expressed by subsets
of spinal neurons (Jessell, 2000; Tamura et al., 2003), but their
function and potential roles as regulators of Hox activity have
not been defined.
We have explored the role of accessory factors as regulators
of Hox-dependent programs of spinal MN differentiation,
Cell 134, 304–316, July 25, 2008 ª2008 Elsevier Inc. 305
scription factor with functions in cardiac and hematopoetic de-
velopment (Wang et al., 2004; Hu et al., 2006), controls the out-
put of the entire Hox-dependent program of spinal MN diversity
FoxP1 Is Restricted to Hox-Sensitive Motor Columns
To explore whether Fox and Hox proteins work together in the
specification of MN subtype identity we defined Fox genes ex-
pressed by MNs. Analysis of forty-one Fox genes in embryonic
mouse spinal cord revealed that three members of the FoxP
sub-family, FoxP1, FoxP2, and FoxP4, are expressed by ventral
neurons. FoxP2 is expressed by interneurons but not MNs,
FoxP4 is expressed transiently by a subset of MNs, whereas
FoxP1 is expressed by MNs at brachial, thoracic and lumbar
levels of the spinal cord (Figures 1D–1F and Figures S1A–S1G
available online). We have focused on the role of FoxP1 as a
potential regulator of Hox activity during MN differentiation.
To determine whether FoxP1 is restricted to columnar classes
of MNs we compared its expression with that of Hox proteins
and other markers of columnar subtype, from e11.5 to e14.5.
This analysis revealed that FoxP1 is expressed selectively by
LMC and PGC neurons, and that its onset of expression occurs
tiation (Figures 1H–1O and S2A–S2D). In brachial and lumbar
spinalcord,FoxP1isdetected athighlevels inLMCneurons, de-
fined by retinaldehyde dehydrogenase-2 (RALDH2) expression
(Figure 1N) Brachial FoxP1+LMC neurons coexpress Hoxa6
and Hoxc6, whereas lumbar FoxP1+LMC neurons express
Hoxd10 (Figures S3A, E; data not shown). In thoracic spinal
cord, the domain of Hoxa9 and Hoxc9, FoxP1 is detected at
low levels by PGC neurons, defined by phospho-Smad1/5/8
(pSmad) and neuronal nitric oxide synthase (nNOS) expression
(Figures 1I, 1J, and S3C). In contrast, FoxP1 is excluded from
MMC neurons, defined by expression of Hb9, Isl1/2 and Lhx3,
as well as from HMC neurons, defined by Hb9 and Isl1/2 expres-
sion in the absence of Lhx3 (Figures 1K, 1O, and S3H). A similar
columnar profile of FoxP1 expression is detected in developing
chick spinal cord (Figures S1H–S1J and S4A–S1I). Thus FoxP1
marks Hox-dependent motor columns.
We also examined, more quantitatively, the level of FoxP1 ex-
pression in MNs. We found that the level of FoxP1 in the nuclei of
LMC neurons is ?6-fold greater than that in PGC neurons
(Figure 1G; p < 0.001). The difference in FoxP1 expression level
is evident at the onset of MN differentiation (?e9.5) and persists
until at least e14.5. Thus, LMC neurons constitute a FoxP1high,
and PGC neurons a FoxP1low, population (Figure 1P).
Hox and Homeodomain Activities Determine
the Columnar Profile of FoxP1
To determine whether FoxP1 expression in MNs depends on
Hox activity we used chick electroporation to express engrailed
entiation, yet block the emergence of PGC and LMC identities
(Dasen et al., 2003). We examined the impact of EnR-Hoxc6
on FoxP1 expression in brachial MNs, and of EnR-Hoxc9 on
expression of EnR-Hoxc6 failed to express FoxP1 (Figures 2A
and 2C), or RALDH2 (Figures S5A–S5D). The HD profile of this
novel set of brachial MNs (Isl1/2+, Hb9+, Lhx3off) matches that
sion of EnR-Hoxc9 lacked FoxP1 (Figures 2B,C) as well as PGC
markers (Dasen et al., 2003), and the number of neurons with an
HMC profile was increased (Figures S5E-S5H). Thus Hox activity
is required for the expression of FoxP1 in MNs, and the loss of
PGC and LMC identity after Hox blockade is accompanied by
the appearance of MNs with an HMC profile.
Is the difference in FoxP1 level in PGC and LMC neurons de-
termined by Hox paralog expression? To assess this, we altered
the profile of Hox proteins along the rostrocaudal axis of chick
spinal cord and examined whether the interconversion of colum-
nar identities is accompanied by changes in FoxP1 expression
level. Thoracic misexpression of Hoxc6 or Hoxd10, which re-
presses Hox9 proteins and elicits a switch from PGC and HMC
to LMC columnar fates (Dasen et al., 2003; Figures S5I and
S5L; data not shown), induced FoxP1 expression at ?5-6 fold
greater levels than that in non-electroporated MNs, close to
the level normally detected in LMC neurons (Figures 2D-I). Con-
versely, brachial expression of Hoxc9, which represses Hox6
proteins and generates PGC neurons at the expense of LMC
neurons (Dasen et al., 2003), reduced the level of FoxP1 ?5-
fold, to values approaching those of PGC neurons (Figures 2J–
2L). Ectopic expression of Hox6/9/10 proteins did not induce
FoxP1 in Lhx3+MMC MNs (Figures S5J and S5L, and data not
shown). These findings provide evidence that the difference in
We next examined why FoxP1 expression is excluded from
MMC and HMC neurons, despite expression of Hox proteins
(see Figure S3). Lhx3 has been implicated in the assignment of
MMC fate (Sharma et al., 2000), and Hb9 in the suppression of
PGC fate (William et al., 2003). We found that ectopic Lhx3 pre-
vented FoxP1 expression in brachial and thoracic MNs and that
ectopic Hb9 repressed FoxP1 expression in prospective PGC
neurons (Figures 2M–2R and S5M–S5P). These findings show
is achieved through the convergence of rostrocaudal and dorso-
ventral transcriptional programs: the inductive activity of Hox
paralog proteins in concert with a restrictive influence of HD
proteins that assign MMC and HMC fates (Figure 2S).
FoxP1 Level Determines Motor Neuron Columnar
Does the level of FoxP1 expression in MNs determine PGC and
LMC identities? We first asked whether ectopic expression of
FoxP1 in thoracic MNs converts prospective HMC neurons to
a PGC fate, and whether an increase in FoxP1 level converts
HMCandPGC neurons to aLMCfate. WeusedHb9and CAGGs
promoters to generate Foxp1 transgenic mouse and chick em-
bryos in which the number of thoracic MNs expressing FoxP1
is increased from ?25% to ?70% (Figures S7A and S7R).
Founder analysis of e12.5 Hb9::Foxp1iresGFP transgenic mice
revealed that the level of FoxP1 in individual thoracic MN nuclei
was ?3-fold greater than that detected in wild-type PGC
306 Cell 134, 304–316, July 25, 2008 ª2008 Elsevier Inc.
neurons (Figures S6A–S6C). Inchick, electroporation of a diluted
CAGGs::Foxp1 (Foxp1[low]) plasmid directed FoxP1 expression
at low levels, similar to that of PGC neurons, whereas Hb9::
Foxp1iresGFP and high concentrations of CAGGs::Foxp1
(Foxp1[high]) resulted in expression of FoxP1 at levels ?7-fold
Expression of these constructs did not impair general features of
(Figures 3N, S7C–S7F, and S7S).
Analysis of columnar fates in the thoracic spinal cord of
Hb9::Foxp1iresGFP mouse embryos revealed a doubling in the
number of pSmad+, and nNOS+PGC neurons, compared to
controls (Figures 3A–3D and 3K). In addition, a few MNs ex-
pressed the LMC marker RALDH2 (Figures 3E, 3F, and S7G–
S7L). Conversely, the number of neurons with an HMC profile
(Isl1/2+, Hb9+, Lhx3off) was markedly reduced (Figures 3K and
S7O–S7R). In mouse, ?40% of HMC neurons express the ETS
protein Er81 (Figure 1L) (Cohen et al., 2005), and this subset
was similarly reduced in Hb9::Foxp1iresGFP embryos (Figures
3G, 3H, and 3K). In contrast, the number of Lhx3+MMC neurons
was unchanged (Figures 3I–3K), and the motor nerve projecting
to axial muscles was maintained (data not shown). We next
compared the impact of FoxP1 dosage on thoracic MN differen-
tiation in chick embryos. Expression of CAGGs::Foxp1[low]eli-
cited a 50% increase in the number of PGC neurons (Figures
3L andS6F–S6H). Incontrast,
CAGGs::Foxp1[high]markedly reduced the number of pSmad+
PGC as well as Hb9+, Isl1/2+, Lhx3offHMC neurons (Figure 3M;
data not shown) and induced many more RALDH2+LMC
neurons (Figure 3O). These findings provide evidence that incre-
mental changes in the level of FoxP1 in thoracic MNs result in
a step-wise ‘HMC to PGC to LMC’ interconversion of columnar
fate (Figure 3R), under conditions in which Hox paralog profiles
Figure 2. Induction and Restriction of FoxP1 Expression in Motor Neurons
Regulation of FoxP1 expression in chick MNs analyzed by electroporation of genes at stages 14–17 and analysis at stages 26–30.
(B) Thoracic (Th) EnR-Hoxc9 expression abolishes FoxP1 in MNs.
(C) Impact of expression of EnR-Hox fusions on FoxP1 levels. FoxP1 levels in the nuclei of GFP+MNs (labeled EnRc6 and EnRc9) and non-electroporated neu-
rons (LMC and PGC).
(D–F) Expression of Hoxc6 in Th MNs elevates nuclear FoxP1 level.
(G–I) Th Hoxd10 expression elevates nuclear FoxP1 level.
(J–L) Br Hoxc9 expression reduces FoxP1 level. C, F, I, L show nuclear FoxP1 levels from R 50 MNs, from electroporated and control sides.
(M and N) Br Lhx3 blocks FoxP1 and RALDH2.
(O and P) Th Lhx3 blocks FoxP1 and pSmad. Lhx3 does not change Br Hoxc6 or Th Hoxc9 (Figures S5M and S5N).
(Q and R) Th Hb9 blocks FoxP1 and pSmad. Expression of Hb9 at limb levels does not block FoxP1 or LMC fate (Figures S5O–S5P).
(S) Interactions of Hox, Hb9, and Lhx3 control MN FoxP1.
Cell 134, 304–316, July 25, 2008 ª2008 Elsevier Inc. 307
Is FoxP1 simply a Hox intermediary, or an accessory factor
that functions together with ongoing Hox activity? To resolve
this issue we sought to eliminate Hox activity while maintaining
FoxP1 expression. We expressed a EnR-Hoxc6 construct which
blocks Hox6 activator functions at brachial levels (Dasen et al.,
2003), and in addition expressed CAGGs::Foxp1[high], to com-
pensate for the loss of endogenous FoxP1 caused by repression
of Hox6 activity. If FoxP1 functions as a Hox intermediary, its
target, RALDH2, should be expressed robustly despite the loss
of Hox activity. Conversely if FoxP1 requires ongoing Hox
activity, the expression of RALDH2 should be abrogated despite
high level FoxP1 expression. RALDH2 expression was drasti-
cally reduced in brachial MNs, despite evident high level
FoxP1 expression (Figure 3P,Q). This finding reveals an ongoing
requirement for Hox activity during the FoxP1-dependent
assignment of LMC identity, and indicates that FoxP1 and Hox
proteins act in a convergent manner to specify MN columnar
Switches in Columnar Fate and Connectivity
in Foxp1 Mutants
To test the requirement for FoxP1 activity in the assignment of
MN columnar fates we analyzed Hox expression profiles, MN
subtype identity and connectivity in Foxp1 mutant mice.
In e11.5 Foxp1 mutants the expression of Hb9, Isl1/2, and the
cholinergic marker VAChT, was similar in Foxp1 mutant and
wild-type embryos (Figures S8A-H). Thus, FoxP1 is not required
for the emergence of generic MN characteristics. Moreover, the
profile of Hox4 to Hox10 paralog expression was unchanged in
Foxp1 mutants (Figures S8I-V). Thus Hox protein expression
regulates, but is not itself regulated by, FoxP1.
columnar differentiation. At thoracic levels we detected a > 90%
reduction in PGC neurons, assessed by the absence of dorsal
Isl1+MNs, as well as by the loss of pSmad and nNOS expression
(Figures 4A–4F and 4K; data not shown). At brachial and lumbar
levels we detected a > 90% reduction in RALDH2 expression
(Figures 4L–4O). The persistence of a few neurons of PGC and
LMC character is likely to result from a compensatory activity
of FoxP4, which is expressed transiently by MNs and shares
FoxP1’s inductive activity (Figures S1Q–S1S). RALDH2-depen-
dent retinoid synthesis by LMC neurons provides a feed-back
signal that promotes the proliferation of MN progenitors (Jessell,
2000) and the loss of this signal may account for the ?30%
decrease in the number of MNs detected at limb levels of
e13.5 Foxp1 mutants (data not shown).
What becomes of prospective PGC and LMC neurons in
Foxp1 mutants? At thoracic and limb levels of Foxp1 mutants
Figure 3. FoxP1 Induces PGC and LMC Identity
(A–K) Thoracic MN differentiation in e12.5-e13.5 Hb9::Foxp1iresGFP embryos.
(A–D) Expression of FoxP1 increases the number of nNOS+and pSmad+neurons.
(E and F) FoxP1 induces RALDH2 in a few Th MNs, close to Lhx3+MNs (see also Figure S7).
(G and H) Ectopic FoxP1 blocks Er81.
(I and J) Expression of FoxP1 has no effect on Lhx3+neurons.
(K) MN columnar identity after Th FoxP1 expression. Similar results obtained in 3 founder embryos. Mean ± SEM/ventral quadrant/15 mm section. Experimental
and control embryos differ at p < 0.01.
(L–O)FoxP1 expression inchickThspinalcord.(L)MorepSmad+neuronsafterexpressionofFoxP1 atlowlevels. (M)LossofpSmad+neuronsafterexpression of
FoxP1 at high levels. (N) FoxP1 does not induce Hoxc6. (O) FoxP1highexpression induces RALDH2.
(P and Q) Lack of RALDH2 after expression of FoxP1 under CAGGs control in the presence of EnR-Hoxc6.
(R) MN columnar fate after Th elevation of FoxP1 expression in chick [c] and mouse [m].
308 Cell 134, 304–316, July 25, 2008 ª2008 Elsevier Inc.
most Lhx3offMNs coexpressed Isl1/2 and Hb9 - the profile of
HMC neurons (Figures 4G, 4H, 4K, 4P, and 4Q). We also de-
tected a 2-fold increase in Er81+HMC neurons at thoracic levels
of Foxp1 mutants (Figure 4I- K). In contrast, the number of MMC
(Hb9+, Isl1/2+, Lhx3+) neurons was unchanged (Figure 4I–4K,
data not shown). Thus in Foxp1 mutants the spinal cord loses
quence is transformed along its length into a twinned columnar
system that comprises MMC and HMC neurons (Figure 6S).
These findings provide genetic evidence that FoxP1 activity nor-
mally diverts MNs from an HMC-like ground state toward PGC
and LMC fates.
We next examined how the appearance of a continuous HMC
column influences MN axonal projections. We introduced an
Hb9::GFP transgene (Arber et al., 1999) into Foxp1 mutant and
heterozygote backgrounds, and analyzed the trajectory of
GFP-labeled motor axons from e11.5 to e14.5. At thoracic levels
of Foxp1 mutants we observed a severe reduction in the projec-
tion of the axons of PGC neurons to sympathetic chain ganglia
(Figures 5A–5C). We also injected rhodamine-dextran (RhD)
into the intercostal nerves that supply body wall muscles in
Foxp1?/?and wild-type embryos and monitored the transcrip-
tional profile of retrogradely-labeled MNs at e12.5. Virtually all
RhD-labeled MNs in Foxp1 mutants exhibited an HMC-like
profile (Isl1/2+, Hb9+, Lhx3off; Er81+; Figures 5G–5L). Thus tho-
racic MNs deprived of FoxP1 fail to pursue a PGC-like trajectory
and instead project their axons distally toward body wall mus-
cles, the normal target of HMC neurons. Lhx3+MMC neurons
were not labeled after intercostal tracer injections in Foxp1 mu-
tants (Figures 5G and 5H), indicating that their axons continue
to project to axial muscles.
At brachial and lumbar levels of Foxp1 mutants we detected
a GFP-labeled motor nerve branch that reached axial muscles,
suggesting that a MMC-like axonal trajectory is also preserved
at limb levels of the spinal cord (Figures 5D–5F). We also de-
tected a distally-directed motor nerve branch that reached the
base of the limbs (Figures 5D–5F), and MNs retrogradely labeled
after tracer injection into the limbs exhibited an HMC transcrip-
tional profile (data not shown). Thus the HMC-like neurons
generated at limb levels of the spinal cord of Foxp1 mutants
embark on a distal trajectory, but by virtue of their aberrant
position of origin encounter limbs rather than body wall
The Fate of Motor Axons in the Limbs of Foxp1 Mutants
How do the axons of these misplaced HMC neurons behave on
entering the limb? The acquisition of LMC divisional and pool
Figure 4. FoxP1 Is Required for the Differentiation of PGC and LMC Neurons
(A and B) Motor columns at Th levels in wild-type and Foxp1 mutants.
(C–F) In Foxp1 mutants, Isl1+MNs are not detected in a dorsal position, and little nNOS or pSmad is detected.
(G and H) Increase in the number of Hb9+/Isl1/2+MNs.
(I and J) Increase in Er81+MNs in Foxp1 mutants. The number of Lhx3+neurons is unchanged.
(K) Quantification of MN columnar subtype markers in Th spinal cord of wild-type and Foxp1 mutant embryos/15 mm section, mean ± SEM. Values for PGC and
HMC differ in experimental and control embryos (p < 0.01).
(L and M) Motor columns at Br levels in e12.5 wild-type and Foxp1 mutant embryos.
(N and O) Loss of RALDH2 in Foxp1 mutants. Residual expression likely reflects a redundant role of FoxP4 (Figures S1Q-S1S).
(P and Q) At e13.5, medial (m) and lateral (l) divisions of the lumbar LMC are defined by segregated expression of Isl1 and Hb9. In Foxp1 mutants, Lhx3offMNs
coexpress Hb9 and Isl1.
Cell 134, 304–316, July 25, 2008 ª2008 Elsevier Inc. 309
of motor axons within the limbs (Figure 7A). Medial and lateral
divisional identity dictates the selection of ventral and dorsal ax-
onal trajectories (Kania and Jessell, 2003). And diverse MN pool
identities direct the formation of muscle nerve branches as well
as the pattern of axonal arborization within target muscles (Livet
et al., 2002; De Marco Garcia and Jessell, 2008). To determine
the contribution of the FoxP1/Hox program to these patterns of
connectivity we compared the extent of LMC divisional and
pool differentiation with the trajectory of motor axons within
the limbs of Foxp1 mutants.
We found that FoxP1 is needed for the expression of divisional
transcription factors and surface receptors that normally deter-
mine the dorsoventral trajectory of LMC axons in the limb. In
wild-type embryos, medial LMC neurons express Isl1 and lateral
LMC neurons Hb9 and Lhx1 (Figures 4P and 6A) (Kania and
Jessell, 2003). The HMC-like MNs generated at limb levels of
Foxp1 mutant embryos did not segregate Isl1 from Hb9, lacked
Lhx1 expression (Figures 4Q and 6B) and failed to express
EphA4, a guidance receptor that directs the axons of lateral
LMC neurons into the dorsal limb (Figures 6C and 6D) (Kania
and Jessell, 2003). FoxP1 also controls the expression of tran-
and connectivity. Three Hox-regulated transcription factors nor-
mally expressed in selected LMC motor pools, Nk6.1, Pea3 and
Er81, were absent from HMC-like MNs in Foxp1 mutants, (Fig-
ures 6G–6L) (De Marco Garcia and Jessell, 2008). These MNs
also lacked expression of Cad-8 and Cad-20, type II cadherins
ligand implicated in motor axon growth and guidance (Figures
6M–6R) (Livet et al., 2002; Price et al., 2002). In contrast, expres-
sion of Npn-1, a Sema receptor expressed by many classes of
spinal MNs (Huber et al., 2005) was maintained by HMC-like
neurons at limb levels of Foxp1 mutants (Figures 6E and 6F).
Thus, FoxP1 activity controls LMC divisional and pool character
Analysis of the projections of motor axons in the limb of e11.5
Foxp1?/?; Hb9::GFP mice revealed an apparently normal bifur-
cation of the main motor nerve trunk into dorsal and ventral
divisional branches (Figures 7B and 7C). And between e11.5
and e14.5, the overt pattern of muscle nerves was similar to
that observed in control Hb9::GFP embryos (Figures 7D–7G
and S9A–S9F). Muscle nerves branches tended to be thinner in
Foxp1 mutants, probably a consequence of the reduction in
MN number at limb levels. In addition, a few nerve branches to
specific muscles were missing (Figure S9). Nevertheless, most
limb muscles were innervated by motor axons in Foxp1 mutants
(Figures 7H and 7I; data not shown).
focused on the pattern of innervation of muscles supplied by
motor pools that express Hox-dependent transcription factors.
Nerve branches supplying the tibialis anterior (Ta) and gracilis
posterior (Gp) muscles derive from Nkx6.1+motor pools, and
Figure 5. Altered Motor Axon Projections in Foxp1 Mutants
(A–F) Motor axon projections in wild-type and Foxp1 mutant embryos.
(A) Projections of Th MNs. Location of RhD injection site for panels (G)–(L) shown.
(B and C) In Foxp1 mutants, axonal projections to sympathetic chain ganglia (SCG) are dramatically reduced at e13.5 (arrows). GFP-labeled motor axons (green),
Isl1/2+DRG and SCG neurons (red). Axonal projections to axial muscles are preserved.
limb may reflect the depletion of Lhx3offBr MNs. Projections to axial muscles are preserved.
(G–L) Labeled MNs after RhD injection into intercostal nerves. Er81+, Isl1/2+, Lhx3offMNs are labeled in wild-type and Foxp1 mutants.
310 Cell 134, 304–316, July 25, 2008 ª2008 Elsevier Inc.
in Nkx6.1 mutants the axons that normally project to these mus-
cles are re-routed to different targets (De Marco Garcia and
Jessell, 2008). In Foxp1 mutants we found that the Ta and Gp
muscles were innervated by motor axons despite the loss of
Nkx6.1 (Figures 7F,G, S9A,B). This paradoxical finding appears
to have its basis in the erosion of Ta and Gp pool identities in
Foxp1 mutants, whereas these pools retain aspects of their
molecular character in Nkx6.1 mutants (De Marco Garcia and
Jessell, 2008, see Discussion). In contrast, motor pool programs
that specify the pattern of axon terminal arborization within
target muscles were severely affected by the loss of FoxP1.
MNs that project to the cutaneus maximus (CM) forelimb muscle
normally establish an expansive intramuscular arbor through
a developmental program that depends on the pool-restricted
Pool Identity of LMC Neurons
(A and B) Loss of expression of Lhx1 by Br MNs in
e13.5 Foxp1 mutants. LMC confines marked with
dotted line. Similar results obtained at Lu levels.
(C and D) Loss of EphA4 expression from Br MNs
in Foxp1 mutants.
(E and F) Persistence of Npn-1 in Br MNs in Foxp1
(G–R) Loss of pool markers in Lhx3offMNs in
Foxp1 mutants at Br and Lu levels.
(S) Summary indicating how FoxP1 controls the
formation of PGC and LMC columns and LMC
(T) FoxP1 gates the output of Hox networks that
assign MN columnar and pool identities.
ysis of Foxp1
revealed a failure of motor axon arboriza-
tion within the CM muscle (Figures 7J-M),
a phenotype similar to that observed in
Pea3 mutants (Livet et al., 2002). Thus
this later, muscle-type specific, connec-
tivity program is drastically impaired in
?/?; Hb9::GFP embryos
Erosion of Motor Neuron-Muscle
Topography in Foxp1 Mutants
Does the overtly normal pattern of motor
nerves observed in the limbs of Foxp1
mutants conceal a disruption in the topo-
graphic link between MN position and
muscle nerve trajectory? To address this
possibility, we injected HRP and RhD
branches of e13.5 wild-type and Foxp1
mutant embryos, and assessed the posi-
tion of labeled neurons within the cohort
of HMC-like neurons present at limb
levels of the spinal cord. In wild-type
embryos, axons that enter the ventral
divisional branch derive from medially-
positioned LMC neurons whereas axons
that enter the dorsal divisional branch derive from laterally-posi-
tioned LMC neurons (Figure 7N). In Foxp1 mutants there was no
discernable relationship between the mediolateral position of
RhD- or HRP-labeled MNs within the HMC-like cohort and the
selection of ventral or dorsal divisional branches (Figures 7N,
S9G-I; data not shown). This finding, together with the erosion
of LMC divisional identity revealed by loss of Lhx1 and EphA4
expression, provides strong evidence that the inactivation of
FoxP1 scrambles the selection of dorsal and ventral axonal
trajectories in the developing limb.
We also examined the topographic relationship between MN
position and axonal projections into individual muscle nerves.
We labeled the forelimb ulnar and hindlimb obturator muscle
nerves by focal RhD injection, and assessed the position of
Cell 134, 304–316, July 25, 2008 ª2008 Elsevier Inc. 311
labeled MNs (Figures 7O and S9J). In wild-type embryos, RhD-
labeled MNs that send axons into the ulnar and obturator nerve
branches were clustered into coherent pools within the medial
LMC (Figures 7O and S9J). But in Foxp1 mutants RhD-labeled
neurons were scattered throughout the HMC-like column (Fig-
ures 7O and S9J). Thus the loss of FoxP1 degrades the topogra-
phy of motor axonal projections into individual muscle nerves,
suggesting that MNs deprived of FoxP1 project axons into
divisional and muscle nerve paths in a haphazard manner.
A network of twenty-one Hox genes endows spinal MNs with
identities that define the specificity of target innervation.
FoxP1, at one fell swoop, controls the output of the Hox reper-
toire that directs MN identity and connectivity. Short-circuiting
this network by inactivation of FoxP1 has revealed insight into
the origins of MN diversity and key aspects of the logic through
which Hox genes impose a topographic motor map.
Figure 7. Limb Innervation in Foxp1 Mutants
(A) Development of limb innervation.
(B and C) Projection of motor axons along dorsal (d) and ventral (v) divisional branches in e11.5 wild-type and Foxp1 mutants.
(D and E) Forelimb innervation (dorsal view) in wild-type and Foxp1 mutants at e14.5. Motor nerve branches are similar in wild-type and Foxp1 mutants. bbr,
biceps brachialis; ecr, extensor carpi radials longus and brevis; ed, extensor digitorum; fcu, flexor carpi ulnaris; fdp, flexor digitorum profundis. Motor axons
visualized by GFP in Hb9::GFP and Foxp1?/?; Hb9::GFP embryos.
(F and G) Motor axons in the tibialis (Ta) nerve in wild-type and Foxp1 mutant embryos at e14.5.
(H and I) Near-normal pattern of motor innervation of distal forelimb muscles (My32: myosin light chain, red) in e13.5 Foxp1 mutant embryos.
(J–M) A vestigial motor nerve is oriented toward the CM muscle (arrow in K), but the arborization of motor axons is drastically reduced in e13.5 Foxp1 mutants.
at forelimb levels (Figure S9H).
(O) RhD-labeled, Isl1/2+MNs after tracer injection into the obturator nerve branch of wild-type and Foxp1 mutant embryos. Similar results obtained after ulnar
nerve injection (Figure S9J).
312 Cell 134, 304–316, July 25, 2008 ª2008 Elsevier Inc.
FoxP1 and the Origins of Motor Neuron Diversity
Spinal motor systems exhibit remarkable variation in organiza-
tion and complexity (Fetcho, 1992). Our findings point to an es-
sential role for the FoxP1/Hox program in expanding the range
of MN columnar subtypes so as to ensure effective innervation
of peripheral motor targets. In particular they suggest that the
formation of HMC neurons from an ancestral MMC-like group
was a crucial step in MN diversification; providing a malleable
population of MNs that serve as the substrate for the FoxP1/
Hox program of columnar and pool specification (Figure 6S).
The ancestral state of spinal MNs is marked by expression of
Lhx3, in the context of Isl1/2 and Hb9 (Landgraf and Thor, 2006).
This transcriptional profile defines early-born ‘primary’ MNs in
zebrafish and amphibian embryos (Appel et al., 1995; Borodin-
sky et al., 2004), presumed counterparts of the MNs of jawless
vertebrates (Fetcho, 1992), as well as the MMC neurons of birds
file of this ancestral set of MNs may reflect common patterns of
connectivity—the innervation of segmentally arrayed muscles
involved in undulatory locomotor behaviors. The diversification
of columnar subtypes from this ancestral group requires relief
from the confining influence of Lhx3 (Sharma et al., 2000; William
et al., 2003). This evasive step may have involved a decrease in
the strength of the Wnt signaling component of the dorsoventral
inductive pathway, since reducing Wnt4/5 activity in mice pro-
motes the generation of HMC neurons at the expense of MMC
neurons (D. Agalliu and T.M.J., unpublished data). Thus the spi-
nal motor system induced by the dorsoventral signaling pathway
comprises MMC and HMC neurons, arrayed in coextensive col-
umns (Figure 6S).
How did this twinned columnar plan undergo further diversifi-
cation to generate segmentally restricted PGC and LMC neu-
of the influence of Lhx3, became sensitive to the rostrocaudal
patterning activities of Hox proteins, which diverted them away
from their constitutive HMC ground state and toward PGC or
LMC fates. FoxP1, through its dose-dependent inductive activ-
ity, is a key participant in this Hox program of columnar specifi-
cation. Under the influence of Hox9 activity, newly-generated
Lhx3offthoracic MNs acquire the capacity for low-level FoxP1
expression, and appear to resolve their bi-potential, HMC or
PGC, fate through mutual repressive interactions between
FoxP1 and Hb9 (Figure S7). In the limb-level realms of Hox6/10
activity, FoxP1 can be induced to high levels without repression
by Hb9, which is likely to account for the progression of all
Lhx3offneurons to a LMC fate.
How the FoxP1/Hox network was recruited to the task of MN
columnar diversification remains unclear. The absence of PGC
and LMC neurons from early vertebrates could have its basis
in evolutionary changes in cis-regulatory elements that control
Foxp1 expression (Shubin et al., 1997; Prud’homme et al.,
2007), such that Hox-sensitive elements responsible for expres-
sion in spinal MNs were configured only at the time of formation
of the sympathetic nervous system and paired appendages. Al-
ternatively, the rostrocaudal pattern of expression of Hox genes
in the spinal cord of jawless vertebrates may differ from that in
birds and mammals (Force et al., 2002; Takio et al., 2007), and
thus may fail to produce a productive Hox code capable of acti-
vating FoxP1 expression. An evolutionary change in the rostro-
caudal profile of Hox genes in the lateral plate mesoderm has
been proposed to regulate the formation and positioning of fins
and limbs (Cohn et al., 1997; Cohn and Tickle, 1999). Thus, the
coordinate reshaping of mesodermal and neural Hox expression
patterns during vertebrate evolution provides a plausible basis
for linking the formation and diversification of LMC neurons to
the appearance and elaboration of paired appendages (Tabin,
1995; Shubin et al., 1997).
FoxP1 as a Hox Accessory Factor
Thepervasive activitiesof insectand vertebrateHoxproteins are
constrained by other transcription factors that enhance the
specificity of Hox-DNA interactions and thus determine the
selection of Hox target genes (Mann and Affolter, 1998; Moens
and Selleri, 2006). Our findings show that FoxP1 is an indispen-
sible Hox accessory factor during the programming of spinal MN
diversity and connectivity. FoxP1 appears to act jointly with Hox
proteins ratherthanasalinear intermediary inthe pathwayofco-
lumnar specification, since forced expression of FoxP1 is unable
to induce LMC character under conditions in which Hox activity
is repressed. The idea that FoxP1 and Hox proteins interact to
specify MN columnar identity is also supported by analysis of
mice lacking Hox10 gene function, where the erosion of lumbar
LMC character is accompanied by the appearance of neurons
with an HMC-like transcriptional profile (Wu et al., 2008), a
phenotype similar to that of Foxp1 mutants.
Two distinct sets of Hox regulatory interactions specify the
columnar and pool character of LMC neurons. Our data suggest
that FoxP1 gates the output of both Hox networks (Figure 6T).
But the burden of specificity during the transcriptional program-
ming of MN columnar and pool identity alternates between Hox
proteins and FoxP1. Initially, distinctions in Hox6, 9, and 10
paralog identity set the level of MN FoxP1 expression. Once in-
duced, however, FoxP1 is the primary determinant of columnar
specificity, since its dose-dependent assignment of PGC and
LMC fate is independent of Hox6/9/10 paralog status. But for
the program of motor pool allocation, the onus of specificity in
target gene activation reverts to Hox proteins, since the Hox
code that determines pool identity (Dasen et al., 2005) acts in
the context of uniformly high-level FoxP1 expression.
The contributions of FoxP1 to the Hox programming of spinal
MN differentiation have parallels with those of the Drosophila
Slp1/2 forkhead proteins, which serve as accessory factors for
Ubx and Abd-A during embryonic patterning (Gebelein et al.,
2004). The Drosophila Hox protein scr directly activates expres-
sion of the forkhead gene (Joshi et al., 2007), raising the possibil-
feed-forward loop’ that can introduce delays in target gene
activation (Mangan et al., 2003), and may represent an intrinsic
mechanism for matching the time of expression of guidance
receptors to the arrival of axons at relevant cues, on the path
to their targets.
FoxP1 is critical for the output of the entire Hox program of
spinal MN subtype specification, but not all MN Hox functions
depend on FoxP1. Expression of FoxP1 in PGC and LMC neu-
rons is induced by Hox proteins, and thus this early phase of
Cell 134, 304–316, July 25, 2008 ª2008 Elsevier Inc. 313
Hox activity does not depend on FoxP1. Moreover, mutual re-
pressive interactions between Hox6/10 and Hox9 paralogs are
evident in MMC and HMC neurons (Dasen et al., 2003), even
though they lack FoxP1 expression. As a general rule, regulatory
interactions between Hox genes occur independently of FoxP1,
whereas Hox activation of downstream effectors relies on
FoxP1. The well-documented activities of Hox genes in hind-
brain MN specification (Trainor and Krumlauf, 2000) must also
by hindbrain MNs (Figure S1). But Meis and Pbx/Prep cofactors
are expressed in the hindbrain, and Pbx proteins regulate cranial
MN identity (Moens and Selleri, 2006). Thus, the activities of
FoxP1 as an arbiter of Hox output during spinal MN differentia-
tion may be exerted in a broader context of Meis and Pbx/Prep
activities. Meis and Pbx/Prep cofactors may impart a first level
of Hox specificity (Joshi et al., 2007), with FoxP1 providing
additional filters on target gene activation.
FoxP1 and Strategies of Motor Axon Targeting
How does FoxP1/Hox network activity control the decisions
that loss of FoxP1 blocks the output of the entire MN Hox reper-
toire provided an opportunity to assess, wholesale, the contribu-
tion of this network to target connectivity.
The switch from PGC to HMC fate in Foxp1 mutants is accom-
panied by a redirection of motor axons from sympathetic ganglia
to body wall targets. In contrast, HMC-like MNs generated at
brachial and lumbar levels project to limb rather than body wall
muscles, following a proximal path that resembles that of LMC
axons. Is this the expected behavior of HMC neurons, or an indi-
cation that cryptic elements of LMC character have been pre-
served? Evidence that this is the predicted trajectory of HMC
neurons unlucky enough to find themselves aligned with limbs
comes from the finding that ectopic limbs induced adjacent to
thoracic spinal cord are innervated selectively by axons of
published data). Thus, HMC and LMC neurons are alike in their
initial pursuit of a distal trajectory that takes them to body wall
or limb targets.
In the limb, LMC axons establish stereotypic projections to
individual muscles (Landmesser, 2001). Transcription factors,
recognition molecules and guidance receptors contribute to the
establishment of this topographic motor map. The loss of
FoxP1 abolishes the expression of all molecular features of
LMC divisional and pool identity yet motor nerve branching and
ble explanation for this dichotomy is that elimination of FoxP1
simply scrambles MN cell body position, leaving intact the pro-
grams that direct axonal trajectory. The loss of expression of
genes that determine the trajectory of LMC axons (Lhx1,
EphA4,andNkx6.1)aswellasthose thatcontrol motorpool sort-
ing (Pea3, Cad8, Cad20) argues against this. Our findings fit best
with the view that the overall pattern of motor nerve branching
hibitory domains within the limb mesenchyme, with the FoxP1/
Hox program providing LMC neurons with identities that enable
axons to respond to local cues that promote the selection of
just one of many available axonal conduits (Figure S9K). In this
view, the trajectory of MNs deprived of FoxP1 activity will still
be constrained by the existence of preordained paths, but
without intrinsic molecular programming, individual axons will
tants may underlie the preservation of muscle nerve branches
that are missing in mice deprived of downstream pool transcrip-
terminants in Nkx6.1 mutants accounts for the re-direction of
motor axons away from their intended muscles, to alternate tar-
gets (De Marco Garcia and Jessell, 2008). But in Foxp1 mutants,
be available. This view also helps to explain why motor axon ar-
borization within the CM muscle is similarly stunted in Foxp1 and
Pea3 mutants. We surmise that the random selection of axonal
paths ensures, by chance, that some motor axons will project
toward the CM muscle, but on arrival they will be unable to acti-
vate the Pea3-dependent program that promotes the expansive
arborization of motor axons (Livet et al., 2002).
Motor axon arborization is but one of several late steps in mo-
programs that are also controlled by motor pool transcription
factors (Livet et al., 2002; Vrieseling and Arber, 2006). The
dependence of these pool determinants on the FoxP1/Hox net-
work therefore predicts that the loss of FoxP1 will perturb the
specificity of inputs to spinal MNs that innervate an individual
muscle target. And since Hox and FoxP proteins are coex-
pressed by certain interneuron classes, their convergent activi-
ties in interneurons may also contribute to the assembly of local
microcircuits that coordinate motor output.
Screening for Fox Gene Expression in the Spinal Cord
PCR and in situ hybridization-based screens were performed to identify Fox
genes expressed in motor neurons. We initially screened against forty one
Fox genes by RT-PCR (One-Step, Invitrogen), using four or more unique com-
binations of oligonucleotides directed against each gene. RT-PCR analysis
was performed on e12.5-e14.5 mRNA isolated from tissue containing the
spinal cord and surrounding mesoderm. Twenty-one Fox genes were ampli-
fied by RT-PCR from this initial screen and were subsequently cloned into
a pcrII-TOPO vector (Invitrogen). Template plasmid DNAs were used to gener-
ate cRNA probes for in situ hybridization on e12.5-13.5 spinal cord. Of the
twenty-one probes tested, seven (Foxa2, Foxb1, Foxd3, Foxn4, Foxp1,
Foxp2, and Foxp4) were found to be expressed by neuronal populations within
the spinal cord.
Quantification of FoxP1 Levels
Nuclear FoxP1 levels were determined on cryostat sections at sub-saturating
identical laser and gain configurations. All analyses were performed on
sections on the same slide. Nuclear pixel intensity was determined using
Photoshop. Mean pixel intensities for > 50 MN nuclei are shown.
Plasmid Constructs for Transgenic Animals
Expression constructs for Lhx3, Hb9, Hox, and Engrailed-repressor (EnR) Hox
derivatives were subcloned into a CMV/b?actin promoter driver vector
(CAGGs) as described previously (Dasen et al., 2003, 2005; William et al.,
2003). The murine Foxp1 gene was cloned into an expression vector contain-
ing the regulatory elements of the Hb9 gene (Arber et al., 1999; Wichterle et al.,
2002). The Foxp1 cDNA was followed by an internal ribosomal entry site (ires)
314 Cell 134, 304–316, July 25, 2008 ª2008 Elsevier Inc.
to allow expression of enhanced green fluorescent protein (GFP). For genera-
tion of Hb9::FoxP1iresGFP transgenic mice the expression cassette was
excised from the plasmid prior to pronuclear injection.
In Ovo Electroporation
Electroporation was performed on stage 12–16 chick embryos (Dasen et al.,
2003). Levels of FoxP1 approximating those in PGC neurons were achieved
by diluting the CAGGs::Foxp1 plasmid to 10–15ng/ml using pBKS plasmid
DNA. Levels of FoxP1 approximating those in LMC neurons were achieved us-
ing Hb9::Foxp1iresGFP at 3 mg/ml or CAGGs::Foxp1 at 100 ng/ml. Results for
each experiment are representative of > 8 embryos with MN electroporation
efficiency > 60%.
The Foxp1 mutant strain is as described in Wang et al., (2004), the Hb9::GFP
line in Arber et al., (1999). The Hb9::Foxp1iresGFP line was generated as
described (Arber et al., 1999).
In Situ Hybridization Histochemistry and Immunohistochemistry
In situ hybridization and immunohistochemistry were performed on 15–20 mm
cryostat sections as described (Dasen et al., 2005). Whole-mount antibody
staining was performed as described (De Marco Garcia and Jessell, 2008)
and GFP-labeled motor axons were visualized in projections of confocal
other proteins were generated as described (Dasen et al., 2005). Antibodies
against additional proteins were generated in guinea pigs and rabbits using
the following peptide sequences FoxP1: ENSIPLYTTASMGNPTC, RALDH2:
CERAKRRIVGSPFDPTTE. Additional antibodies were obtained and used as
follows:rabbit anti-nNOS 1:5000 (Immunostar), goatanti-Hoxc6 1:2000 (Santa
Cruz), goat anti-Hoxd10 1:500 (Santa Cruz), goat anti-FoxP2 1:8000 (AbCam),
rabbit anti-FoxP4 1:8000 (kindly provided by Edward Morrisey), goat anti-HRP
1:2000 (Jackson Immunoresearch).
Retrograde Labeling of Motor Neurons
Retrograde labeling of motor neurons was performed as described (Dasen
et al., 2005). 3000 MW lysine-fixable dextran-tetramethylrhodamine (RhD,
Molecular Probes) or horseradish peroxidase (HRP, Roche) was injected into
severed nerves or individual muscles of e12.5-e13.5 embryos. Prior to injec-
tions, embryos were decapitated, eviscerated, and dissected to remove the
epidermis dorsal to the spinal cord. To aid in the identification of nerves, we
used GFP fluorescence from Hb9::GFP transgenic mouse embryos, visualized
using a MVX10 wide-field fluorescent macroscope (Olympus). Nerves were
severed using Oban Bioscissors and HRP or RhD was injected onto the cut
terminal. Embryos were incubated for 3–5 hr in oxygenated F12/DMEM
(50:50) solution at 32–34?C and subsequently fixed in 4% paraformaldehyde.
Supplemental Data include nine figures and can be found with this article
online at http://www.cell.com/cgi/content/full/134/2/304/DC1/.
for mousehusbandry, S. Morton for antibody generation, N. Permaul for assis-
tance, K. MacArthur and I. Schieren for help with the manuscript. N. De Marco
and G. Surmeli identified muscle nerves, R. Mann advised on Hox biology, J.
Fetcho and M. Cohn on motor system and limb evolution. E. Laufer showed
that pSmad is a PGC marker, C. Henderson that Er81 marks HMC neurons.
J. Brown kindly let us cite her unpublished data. S. Arber, R. Axel, G. Fishell,
C. Henderson, R. Mann and S. Poliak provided comments on the manuscript.
J.D. is supported by a Burroughs Welcome Fund CABS and a Whitehead Fel-
lowship, P.W.T. by a National Institutes of Health CA031534 and a Marie Betz-
ner Morrow Centennial Endowment, and T.M.J. by NINDS 5R01033245, The
Wellcome Trust, Project ALS and the NYSCIRP. T.M.J. is a Howard Hughes
Medical Institute investigator.
Received: March 12, 2008
Revised: May 6, 2008
Accepted: June 12, 2008
Published: July 24, 2008
Appel, B., Korzh, V., Glasgow, E., Thor, S., Edlund, T., Dawid, I.B., and Eisen,
J.S. (1995). Motoneuron fate specification revealed by patterned LIM homeo-
box gene expression in embryonic zebrafish. Development 121, 4117–4125.
Arber, S., Han, B., Mendelsohn, M., Smith, M., Jessell, T.M., and Sockana-
than, S. (1999). Requirement for the homeobox gene Hb9 in the consolidation
of MN identity. Neuron 23, 659–674.
Borodinsky, L.N., Root, C.M., Cronin, J.A., Sann, S.B., Gu, X., and Spitzer,
N.C. (2004). Activity-dependent homeostatic specification of transmitter
expression in embryonic neurons. Nature 429, 523–530.
Briscoe, J., and Ericson, J. (2001). Specification of neuronal fates in the ventral
neural tube. Curr. Opin. Neurobiol. 11, 43–49.
Cohen, S., Funkelstein, L., Livet, J., Rougon, G., Henderson, C.E., Castellani,
V., and Mann, F. (2005). A semaphorin code defines subpopulations of spinal
MNs during mouse development. Eur. J. Neurosci. 21, 1767–1776.
Cohn, M.J.,Patel, K.,Krumlauf,R.,Wilkinson, D.G.,Clarke,J.D.,and Tickle,C.
(1997). Hox9 genes and vertebrate limb specification. Nature 387, 97–101.
Cohn, M.J., and Tickle, C. (1999). Developmental basis of limblessness and
axial patterning in snakes. Nature 399, 474–479.
Dasen, J.S., Liu, J.P., and Jessell, T.M. (2003). MN columnar fate imposed by
sequential phases of Hox-c activity. Nature 425, 926–933.
Dasen, J.S., Tice, B.C., Brenner-Morton, S., and Jessell, T.M. (2005). A Hox
regulatory network establishes MN pool identity and target-muscle connectiv-
ity. Cell 123, 477–491.
De Marco Garcia, N.V., and Jessell, T.M. (2008). Early MN pool identity and
muscle nerve trajectory defined by postmitotic restrictions in Nkx6.1 activity.
Neuron 57, 217–231.
Fetcho, J.R. (1992). The spinal motor system in early vertebrates and some of
its evolutionary changes. Brain Behav. Evol. 40, 82–97.
Force, A., Amores, A., and Postlethwait, J.H. (2002). Hox cluster organization
in the jawless vertebrate Petromyzon marinus. J. Exp. Zool. 294, 30–46.
Freitas, R., Zhang, G., and Cohn, M.J. (2006). Evidence that mechanisms of fin
development evolved in the midline of early vertebrates. Nature 442, 1033–
Funakoshi, K., and Nakano, M. (2007). The sympathetic nervous system of
anamniotes. Brain Behav. Evol. 69, 105–113.
Gebelein, B., McKay, D.J., and Mann, R.S. (2004). Direct integration of Hox
and segmentation gene inputs during Drosophila development. Nature 431,
Gutman, C.R., Ajmera, M.K., and Hollyday, M. (1993). Organization of motor
pools supplying axial muscles in the chicken. Brain Res. 609, 129–136.
Hu, H., Wang, B., Borde, M., Nardone, J., Maika, S., Allred, L., Tucker, P.W.,
and Rao, A. (2006). Foxp1 is an essential transcriptional regulator of B cell
development. Nat. Immunol. 7, 819–826.
Huber, A.B., Kania, A., Tran, T.S., Gu, C., De Marco Garcia, N., Lieberam, I.,
Johnson, D., Jessell, T.M.,Ginty, D.D., and Kolodkin,A.L.(2005). Distinct roles
for secreted semaphorin signaling in spinal motor axon guidance. Neuron 48,
Jessell, T.M. (2000). Neuronal specification inthespinalcord: inductive signals
and transcriptional codes. Nat. Rev. Genet. 1, 20–29.
Joshi, R., Passner, J.M., Rohs, R., Jain, R., Sosinsky, A., Crickmore, M.A.,
Jacob, V., Aggarwal, A.K., Honig, B., and Mann, R.S. (2007). Functional spec-
ificity of a Hox protein mediated by the recognition of minor groove structure.
Cell 131, 530–543.
Kania, A., and Jessell, T.M. (2003). Topographic motor projections in the limb
imposed by LIM homeodomain protein regulation of ephrin-A:EphA interac-
tions. Neuron 38, 581–596.
Cell 134, 304–316, July 25, 2008 ª2008 Elsevier Inc. 315
Kusakabe,R.,andKuratani,S.(2005).Evolutionanddevelopmental patterning Download full-text
of the vertebrate skeletal muscles: perspectives from the lamprey. Dev. Dyn.
Landgraf, M., and Thor, S. (2006). Development of Drosophila motoneurons:
specification and morphology. Semin. Cell Dev. Biol. 17, 3–11.
Landmesser, L.T. (2001). The acquisition of motoneuron subtype identity and
motor circuit formation. Int. J. Dev. Neurosci. 19, 175–182.
Laskowski, M.B., and Sanes, J.R. (1987). Topographic mapping of motor
pools onto skeletal muscles. J. Neurosci. 7, 252–260.
Livet, J., Sigrist, M., Stroebel, S., De Paola, V., Price, S.R., Henderson, C.E.,
Jessell, T.M.,and Arber, S.(2002). ETS gene Pea3 controlsthe central position
and terminal arborization of specific MN pools. Neuron 35, 877–892.
Mangan, S., Zaslaver, A., and Alon, U. (2003). The coherent feedforward loop
serves as a sign-sensitive delay element in transcriptionnetworks. J.Mol. Biol.
Mann, R.S., and Affolter, M. (1998). Hox proteins meet more partners. Curr.
Opin. Genet. Dev. 8, 423–429.
McHanwell, S., and Biscoe, T.J. (1981). The localization of motoneurons
supplying the hindlimb muscles of the mouse. Phil. Trans. Roy. Soc. Lond.
Moens, C.B., and Selleri, L. (2006). Hox cofactors in vertebrate development.
Dev. Biol. 291, 193–206.
Price, S.R., De Marco Garcia, N.V., Ranscht, B., and Jessell, T.M. (2002).
Regulation of MN pool sorting by differential expression of type II cadherins.
Cell 109, 205–216.
Prud’homme, B., Gompel, N., and Carroll, S.B. (2007). Emerging principles of
regulatory evolution. Proc. Natl. Acad. USA 104 (Suppl 1), 8605–8612.
Shah, V., Drill, E., and Lance-Jones, C. (2004). Ectopic expression of Hoxd10
in thoracic spinal segments induces motoneurons with a lumbosacral molec-
ular profile and axon projections to the limb. Dev. Dyn. 231, 43–56.
Sharma, K., Leonard, A.E., Lettieri, K., and Pfaff, S.L. (2000). Genetic and
epigenetic mechanisms contribute to MN pathfinding. Nature 406, 515–519.
Shubin, N.,Tabin,C.,and Carroll, S. (1997). Fossils, genes and the evolution of
animal limbs. Nature 388, 639–648.
Tabin, C. (1995). The initiation of the limb bud: growth factors, Hox genes, and
retinoids. Cell 80, 671–674.
Takio, Y., Kuraku, S., Murakami, Y., Pasqualetti, M., Rijli, F.M., Narita, Y., Kur-
atani, S., and Kusakabe, R. (2007). Hox gene expression patterns in Lethen-
teron japonicum embryos–insights into the evolution of the vertebrate Hox
code. Dev. Biol. 308, 606–620.
Tamura, S., Morikawa, Y., Iwanishi, H., Hisaoka, T., and Senba, E. (2003). Ex-
pression pattern of the winged-helix/forkhead transcription factor Foxp1 in the
developing central nervous system. Gene Expr. Patterns 3, 193–197.
Trainor, P.A., and Krumlauf, R. (2000). Patterning the cranial neural crest: hind-
brain segmentation and Hox gene plasticity. Nat. Rev. Genet. 1, 116–124.
Turney, B.W., Rowan-Hull, A.M., and Brown, J.M. (2003). The innervation of
FGF-induced additional limbs in the chick embryo. J. Anat. 202, 83–92.
Vrieseling, E., and Arber, S. (2006). Target-induced transcriptional control of
dendritic patterning and connectivity in MNs by the ETS gene Pea3. Cell
Tucker, P.W. (2004). Foxp1 regulates cardiac outflow tract, endocardial cush-
ion morphogenesis and myocyte proliferation and maturation. Development
Wichterle, H., Lieberam, I., Porter, J.A., and Jessell, T.M. (2002). Directed
differentiation of embryonic stem cells into motor neurons. Cell 110, 385–397.
William, C.M., Tanabe, Y., and Jessell, T.M. (2003). Regulation of MN subtype
identity by repressor activity of Mnx class homeodomain proteins. Develop-
ment 130, 1523–1536.
regulate mouse columnar, divisional and motor pool identity of lumbar moto-
neurons. Development 135, 171–182.
316 Cell 134, 304–316, July 25, 2008 ª2008 Elsevier Inc.