SOX5 Controls the Sequential Generation
of Distinct Corticofugal Neuron Subtypes
Tina Lai,1,2Denis Jabaudon,1,2Bradley J. Molyneaux,1,3Eiman Azim,1,3Paola Arlotta,1,4Joao R.L. Menezes,1,5
and Jeffrey D. Macklis1,*
1MGH-HMS Center for Nervous System Repair, Departments of Neurosurgery and Neurology, Program in Neuroscience, Harvard Medical
School; Nayef Al-Rodhan Laboratories, Massachusetts General Hospital; and Department of Stem Cell and Regenerative Biology and
Harvard Stem Cell Institute, Harvard University, Boston, MA 02114, USA
2These authors contributed equally to this work (co-first authors).
3These authors contributed equally to this work (co-second authors).
4Present address: Center for Regenerative Medicine, Department of Neurosurgery, Massachusetts General Hospital,
Harvard Medical School, Boston, MA 02114, USA.
5Present address: Laborato ´rio de Neuroanatomia Celular, Departamento de Anatomia, Instituto de Cie ˆncias Biome ´dicas,
Programa em Cie ˆncias Morfolo ´gicas, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brasil.
The molecular mechanisms controlling the develop-
ment of distinct subtypes of neocortical projection
neurons, and CNS neuronal diversity more broadly,
are only now emerging. We report that the transcrip-
tion factor SOX5 controls the sequential generation
of distinct corticofugal neuron subtypes by prevent-
ing premature emergence of normally later-born
corticofugal neurons. SOX5 loss-of-function causes
striking overlap of the identities of the three principal
sequentially born corticofugalneuronsubtypes:sub-
plate neurons, corticothalamic neurons, and sub-
cerebral projection neurons. In Sox5?/?cortex,
subplate neurons aberrantly develop molecular hall-
marks and connectivity of subcerebral projection
ferentiated, while differentiation of subcerebral pro-
ysis reinforces the critical role of SOX5 in controlling
the sequential generation of corticofugal neurons—
causes re-emergence of neurons with corticofugal
features. These data indicate that SOX5 controls the
timing of critical fate decisions during corticofugal
neuron production and thus subtype-specific differ-
entiation and neocortical neuron diversity.
Projection neurons of the adult neocortex consist of two main
classes: corticofugal (CFu) neurons, which send their axons
away from the cortex, and intracortical projection neurons,
which send their axons to the ipsi- or contralateral cortex. CFu
neurons can be further divided into subplate (SP) neurons and
corticothalamic (CTh) neurons (together, subcortical projection
tion neurons (corticospinal and related corticobrainstem), which
project more caudally to the tectum, pons, and spinal cord.
These distinct subtypes of cortical projection neurons are born
in a tightly orchestrated sequence and reside at distinct laminar
locations in the neocortex (Gupta et al., 2002; Molyneaux et al.,
2007). Subcortical and subcerebral projection neurons are con-
fined to deep cortical layers (layers V, VI, SP), while intracortical
projection neurons are found mostly within superficial cortical
layers (layers II/III), as well as in layer V.
CFu neurons are exclusively generated during the first few
days of murine cortical development. Neurons born around em-
bryonic day (E)11.5 in the mouse form the deepest cortical layer,
the SP, and send pioneering subcortical projections toward the
thalamus, forming the first corticofugal tract (Allendoerfer and
Shatz, 1994; McConnell et al., 1989). The thalamus is also the
target of layer VI CTh neurons, which are born around E12.5.
Although SP neurons and CTh neurons share the same target,
these two neuronal subtypes have distinct functions and fates.
SP neurons die postnatally in most species but are thought to
play a critical developmental role in instructing corticothalamic
and thalamocortical connectivity (Allendoerfer and Shatz,
1994; Deng and Elberger, 2003). CTh neurons, on the other
hand, persist throughout life in layer VI and establish permanent
connections between the cortex and the sensory and motor
The next-born CFu neurons are the subcerebral projection
neurons, whose peak time of generation is E13.5 (Angevine
and Sidman, 1961); these migrate to layer V. A subset of
subcerebral projection neurons project to the spinal cord (corti-
cospinal motor neurons [CSMN]) and are of particular clinical
importance, since these are the brain neurons that specifically
degenerate in amyotrophic lateral sclerosis (ALS) and whose
damage in spinal cord injury contributes centrally to loss of
Intracortical projection neurons are born together with and
reside alongside subcerebral projection neurons. While CFu
neuron production ceases after E14.5, intracortical projection
neurons continue to be generated and populate more superficial
232 Neuron 57, 232–247, January 24, 2008 ª2008 Elsevier Inc.
cortical layers. The latter neurons are phylogenetically younger
and are thought to originate from progenitors distinct from those
giving rise to CFu neurons (Molyneaux et al., 2007; Zimmer et al.,
2004; Aboitiz and Montiel, 2003; Gupta et al., 2002; Tarabykin
et al., 2001; Desai and McConnell, 2000; McConnell and
Despite substantial advances in our understanding of the
molecular mechanisms regulating general aspects of precursor
specification during early telencephalic development (Guillemot
et al., 2006), only a few genes have been shown to play a
layer-specific role in the differentiation of neocortical neurons.
Examples include Tbr1, whose loss affects SP formation and
corticothalamic connectivity, and Otx1, which plays a role in
the refinement of subcerebral projections (Hevner et al., 2001,
2006; Weimann et al., 1999). Understanding of the mechanisms
underlying specification of individual subtypes of projection
neurons is even more limited.
By isolating pure populations of projection neurons, our labo-
ratory has recently identified what appear to be combinatorial
programs of gene expression that specifically control the devel-
opment of distinct subtypes of projection neurons: subcerebral
projection neurons (including CSMN), callosal projection neu-
rons (CPN), and striatal medium spiny neurons (Arlotta et al.,
2005, 2008; Molyneaux et al., 2005, 2007). The first two genes
belonging to the subcerebral neuron subtype-specific develop-
mental program reported so far are central to proper CSMN
development. Ctip2 is specifically expressed in subcerebral pro-
jection neurons in layer V and is critical for their axonal fascicula-
tion and pathfinding, and Fezf2 (previously called Fezl orZFP312)
is both necessary and sufficient for the birth and specification of
subcerebral projection neurons from cortical progenitors (Arlotta
et al., 2005; Molyneaux et al., 2005; Chen et al., 2005a, 2005b).
Here, we investigate the molecular mechanisms that control
the generation of distinct cortical projection neuron subtypes
by investigating the role of Sox5, a transcription factor belonging
to the SRY-box (Sox)-containing gene family. Sox5 is expressed
at high levels in developing CFu neurons and is excluded from
CPN of layer V and superficial layers. We find that SOX5 controls
the sequence and pace of generation of the different subtypes of
corticofugal neurons during corticogenesis, leading to a model
of progressive withdrawal of SOX5 repression over coordinately
regulated subtype-specific genes.
The function of Sox5 has been best characterized in chondro-
genesis, where it is required for the proper differentiation of
prechondrocytes into early chondroblasts by preventing prema-
ture differentiation (Smits et al., 2004, 2001; Lefebvre, 2002).
Members of the Sox family of transcription factors are highly
conserved and have precisely defined temporal and spatial pat-
terns of expression in many tissues during embryonic develop-
ment, including in the neuroectoderm (Hong and Saint-Jeannet,
2005; Wegner and Stolt, 2005; Bylund et al., 2003; Wegner,
1999). SOX5 itself is expressed in the chick neural crest and
controls the duration of segregation of this structure (Perez-
Alcala et al., 2004). More recently, SOX5 was shown to regulate
oligodendrocyte differentiation in the spinal cord by maintaining
precursor immaturity (Stolt et al., 2006). The specific expression
function of Sox genes in other systems, prompted us to investi-
gate whether this transcription factor plays a role in the develop-
ment of these neurons.
We find that SOX5 controls the differentiation of postmitotic
CFu neurons by repressing subcerebral projection neuron fate
during early corticogenesis. In Sox5-null mutants, the normal se-
quential generation of subtypes of CFu neurons is accelerated,
leading to the premature generation of subcerebrally projecting
neurons. Asa consequence, the delineation between thedistinct
CFu neuron subtypes is lost in Sox5?/?cortex, as earliest-born
istics, and projections of normally later-born subcerebral projec-
tion neurons. Similarly, subcerebral projection neurons prema-
turely differentiate before they reach their final positions in
ter and within deep cortical layers. Thus, the cellular, anatomic,
and functional diversity of the output of the neocortex are criti-
cally controlled by SOX5.
In strong support for SOX5’s central role in setting the proper
sequence of generation of CFu neurons, gain-of-function exper-
iments demonstrate that overexpression of SOX5 past the nor-
mal period of generation of CFu neurons is sufficient to disallow
callosal projection neuron differentiation and cause the re-emer-
gence of neurons with corticofugal projections.
lated, yet distinct, CFu neuron subtypes during development by
controlling the pace of generation of the sequentially born neu-
rons, preventing the premature and inappropriate emergence
of later-born CFu neuron features in early-born neurons.
SOX5 Is Specifically Expressed in Corticofugal Neurons
We previously found that Sox5 mRNA is expressed at high levels
in subcerebral projection neurons at embryonic and early post-
natal stages, while expression remains low throughout develop-
ment in intracortical projection neurons (Figure 1A and Arlotta
et al., 2005). We therefore sought to investigate whether this
transcription factor might function in the development of sub-
cerebral and other CFu neurons.
We first examined the temporal course and cell-type-specific
expression of SOX5 protein in the neocortex during embryonic
corticogenesis. We find that SOX5 is first expressed at about
E12.5 in the emerging cortical plate (CP). At E13.5, during CFu
neuron generation, SOX5 is expressed at high levels in the CP
and subplate (SP). SOX5 is expressed in postmitotic cells, but
not in the dorsal ventricular zone or subventricular zone where
neural progenitors are located (Figure 1B and data not shown).
By P0, neocortical SOX5 expression is confined to neurons of
deep cortical layers (SP, VI, and V), where CFu neurons reside
(Figures 1C, 1D, and 1G), and decreases rapidly after P3 (data
not shown). SOX5 is not expressed in astrocytes or interneurons
(Figures 1E and 1F). Outside of the neocortex, SOX5 is ex-
pressed at low levels in the subpallial ventricular wall, as well
as in the ventral olfactory nucleus, and in the dorsal thalamus
(Figure 1D). In all of these locations, levels of expression
decrease rapidly postnatally.
In the SP and layer VI, SOX5 is present in TBR1-positive CTh
neurons (Figures 1G, 1I, 1K–1K00, and 1L–M00). Within layer V,
SOX5 Controls Development of Corticofugal Neurons
Neuron 57, 232–247, January 24, 2008 ª2008 Elsevier Inc. 233
Figure 1. SOX5 Is Specifically Expressed in Corticofugal Neurons of the Subplate, Layer VI, and Layer V
(A) Normalized levels of Sox5 mRNA in CSMN and CPN determined by microarray experiments during development (from Arlotta et al., 2005).
(B–D) SOX5 immunocytochemistry (ICC). Coronal brain sections at E13.5 (B) and P0 (C) and a sagittal section at P0 (D) reveal SOX5 expression throughout deep
(E and F) SOX5 (red, solid arrows) is not expressed in neocortical astrocytes ([E]; S100b-positive cells, green, empty arrows) or interneurons ([F]; GABA-positive
cells, green, empty arrows).
(G) SOX5-expressing neurons are located in layers V, VI, and SP at P0.
(H) CTIP2 is expressed in layer V and at lower levels in layer VI and SP.
(I) TBR1 labels neurons in layer VI and SP.
(J–J00) Magnified views show that SOX5 and CTIP2 are coexpressed at high levels in layer V neurons.
(K–K00) Magnified views show that SOX5 and TBR1 are coexpressed in layer VI and SP.
(L–M00) Fluorogold (FG) injection into the thalamus at P0 retrogradely labels CTh neurons (green cells) in layer VI and SP neurons at P3. Boxed region in (L) is
magnified in (M)–(M00). FG-labeled CTh neurons (M) are SOX5 positive ([M0]; red, solid arrows). (M00) Merged.
SOX5 Controls Development of Corticofugal Neurons
234 Neuron 57, 232–247, January 24, 2008 ª2008 Elsevier Inc.
SOX5-positive neurons express CTIP2, a specific marker for
subcerebral projection neurons in this layer (Arlotta et al., 2005)
(Figures 1G, 1H, and 1J–J00). Each subtype of CFu neurons,
therefore, has a characteristic combinatorial expression of the
transcription factors SOX5, CTIP2, andTBR1: (1) SP neurons ex-
press an intermediate level of SOX5, a high level of TBR1, and
a low level of CTIP2; (2) CTh neurons in layer VI strongly express
SOX5 and TBR1 and little CTIP2; and (3) subcerebral projection
neurons in layer V express a high level of CTIP2, an intermediate
level of SOX5, and little TBR1 (Figures 1G–1I).
In order to confirm the specific expression of SOX5 in CFu
neurons, we retrogradely labeled neurons from their axonal pro-
jection sites. We find that CTh neurons and SP neurons express
SOX5 (retrograde labeling by FluoroGold [FG] injection in the
thalamus; Figures 1L–1M00). Strikingly, in layer V, SOX5 is ex-
pressed by subcerebral projection neurons (labeled by injection
in the cerebral peduncle) (Figures 1N–O00), but not by callosal
(intracortical) projection neurons (labeled by injection in the con-
tralateral hemisphere) (Figures 1P–1Q00). The temporal and sub-
type-specific patterns of SOX5 expression therefore suggest
that SOX5 might be involved in the postmitotic differentiation
of the three principal corticofugal neuron subtypes.
Loss of SOX5 Function Specifically Impairs Early
Differentiation of Corticofugal Neurons
We find that SOX5 is expressed across the distinct CFu neuron
plement of transcription factors. This suggests that these neuro-
nal subtypes might share a common lineage, within which SOX5
could act in a combinatorial manner to coordinately regulate fate
decisions. To test this hypothesis, we analyzed Sox5 null mutant
mice in which a lacZ/neo cassette is knocked into the Sox5
locus, yielding b-galactosidase (b-gal) expression in a SOX5-
specific pattern (Smits et al., 2001). b-gal and SOX5 strictly
colocalize in Sox5+/?mice, and b-gal and SOX5-expressing
CFu neurons are confined to cortical layers V, VI, and SP at P0
(Figures 2A, 2C, and 2E–E0). Using b-gal immunoreactivity to
trace the fate of CFu neurons in null mutants, which die at birth
(Smitsetal.,2001),wefindthatlossof SOX5dramatically affects
the laminar distribution of these neurons at P0 (Figures 2B and
2D). First, there is seemingly a failure of neurons of layers VI
and V to segregate from each other, with subcerebral projection
neurons nearly absent from layer V. Instead, b-gal-expressing
neurons are found in dense clusters within layer VI and form het-
erotopias in the subcortical white matter (Figures 2D and 2F,
open arrowheads). Second, despite the high density of b-gal-
expressing neurons in layer VI, the SP is not discernable as an
individual layer and is cell-sparse (Figure 2D and Figures S1A
and S1B available online). Instead, in the superficial layers of
Sox5?/?cortex, there is an aberrant population of heterotopic
b-gal-expressing neurons (Figure 2D, solid arrowheads). These
findings indicate that CFu neurons in developing Sox5?/?cortex
are unable to assume proper laminar position, a critical and very
early differentiation step of postmitotic neurons.
We next examined how each of the three sequentially born
subtypes of CFu neurons is specifically affected by loss of
SOX5. Nissl staining shows absence of a morphological SP in
Sox5?/?cortex (Figures S1A and S1B). Using the SP-specific
marker Ctgf (Heuer et al., 2003), we observed a lack of Ctgf-ex-
pressing neurons in the SP and abnormal Ctgf-expressing neu-
rons in superficial layers (Figures 2G and 2H). Although there
are fewer Ctgf-positive neurons in Sox5?/?cortex, the number
of cells labeled by BrdU at E11.5 (the peak of SP neuron birth)
is similar in WT and Sox5?/?cortex (data not shown). This
strongly suggests that the reduced number of Ctgf-positive neu-
rons reflects an abnormal differentiation of SP neurons rather
than an absolute decrease in the size of this cell population.
We observed a similar distribution of SP neurons using another
SP/layer V marker, Crim1 (Figures S1K–S1N) and Nurr-1 (a
deep layer VI/SP marker, data not shown; see Molyneaux
et al., 2007). The lack of a morphological SP in Sox5?/?cortex,
therefore, appears to result from the reallocation of SP neurons
to more superficial locations, suggesting that at least some of
the abnormal b-gal-expressing neurons within these layers are
We next examined the fate of CTh neurons in Sox5?/?cortex.
While TBR1-expressing neurons remain mostly confined to layer
VI, there is a subtle but distinct local dispersion of these neurons
and blurred laminar borders (Figures 2I and 2J). Additional anal-
ysis using TLE4, a marker for layer VI and V neurons (Yao et al.,
1998), further confirms an intralaminar redistribution of layer VI
neurons in Sox5?/?neocortex (Figures S1E–S1H).
Finally, we used the layer V subcerebral projection neuron
specific transcription factor Fezf2 (Molyneaux et al., 2005) to ex-
amine the fate of these neurons in Sox5?/?cortex (Figures 2K
and 2L). We find that Fezf2-positive neurons are almost
completely absent from layer V in Sox5?/?cortex, but instead
accumulate within layerVI and the white matter(Figure 2L).Anal-
ysis using another subcerebral projection neuron specific
transcription factor, CTIP2, confirms this redistribution of neu-
rons (Figures 2M–2Q). Additionally, CTIP2 labeling reveals a het-
erotopic neuronal population in superficial cortical layers, remi-
niscent of the abnormally located Ctgf-positive SP neurons
(Figure 2P, solid arrowheads). These results are further con-
firmed using a series of recently identified markers of subcere-
bral projection neurons (Arlotta et al., 2005; Molyneaux et al.,
2005), including Cry-mu, Crim1, and Diap3 (Figures 5D and 5E
and Figures S1K–S1N and data not shown), indicating that
these neurons are lacking from layer V. Instead, neurons with
molecular characteristics of subcerebral projection neurons are
found in the white matter and in the deepest and most superficial
layers of Sox5?/?cortex. Of note, lateral to motor cortex, the
neocortex is almost completely devoid of CTIP2-positive neu-
rons, resulting in a medialization of CTIP2-expressing neurons
(N–O00) FG injection into the cerebral peduncle at P0 labels subcerebral projection neurons in layer V (green cells) at P3. Boxed region in (N) magnified in (O)–(O00).
FG-labeled subcerebral projection neurons (O) are SOX5 positive ([O0]; red, solid arrows). (O00) Merged.
(P–Q00) FG injection into one hemisphere at P1 retrogradely labels contralateral CPN (green cells) in layers II/III and V at P4. Boxed region in (P) magnified
in (Q)–(Q00). FG-labeled layer V CPN ([Q]; solid arrow) are SOX5 negative (Q0). (Q00) Merged.
Scale bars: 0.5 mm (B–D), 50 mm (G–I, L, N, and P), 20 mm (J–J00, K–K00, M–M00, O–O00, and Q–Q00), 10 mm (E and F).
SOX5 Controls Development of Corticofugal Neurons
Neuron 57, 232–247, January 24, 2008 ª2008 Elsevier Inc. 235
in Sox5?/?cortex (Figures 2N and 2Q), and suggests that SOX5
functions in an areal-specific manner in combination with other
The laminar and areal redistributions of projection neurons in
the absence of SOX5 function are restricted to CFu neurons;
non-CFu and superficial neuron populations are unaffected. In
particular, expression of the layer IV marker RORb (Figures 2R
and 2S), the layer II/III marker Cux2 (Figures 2T and 2U), and
the marginal zone (MZ, layer I) marker reelin (Figures 2V and
2W) are all unchanged in Sox5 null mutants. Similarly, b-gal
and CTIP2 expression by non-neocortical neuron populations
in striatum, thalamus, and piriform cortex appear unchanged in
Sox5?/?brains (data not shown), reinforcing the specificity of
SOX5 function in CFu neuron development.
In summary, loss of SOX5 function highly selectively causes
sequentially differentiating classes of CFu neurons, indicating
a critical role for this transcription factor in the generation of
diversity of cortical output neurons.
Abnormal Differentiation of Corticofugal Neurons Leads
to Aberrant Subcerebral Projections in Sox5?/?Brain
We next examined whether the early abnormal differentiation
of CFu neurons would have consequences on their later
Figure 2. Loss of SOX5 Function Specifically Disrupts the Laminar Positioning of Corticofugal Neurons
(A and C) b-gal (visualized by ICC) is expressed from the Sox5 locus in Sox5+/?mice and serves as an indicator of neurons normally expressing SOX5.
(B and D) In Sox5?/?neocortex, b-gal-expressing cells are abnormally found in deep layer VI, in the white matter (empty arrowheads in [D]), and throughout the
developing cortical plate (solid arrowheads in [D]). Red dashed line in (A) and (B): boundary between layer V and VI.
(E–E0) Magnifications showing colocalized expression of b-gal (green) and SOX5 (red) in Sox5+/?neocortex.
(F) Magnified view from boxed area in (D).
(G and H) In situ hybridization for Ctgf, a specific marker for SP neurons, shows Ctgf-expressing cells (solid arrowheads) abnormally distributed throughout the
(I and J) ICC for TBR1, a SP and layer VI marker. (J) In Sox5?/?neocortex, TBR1-positive neurons are dispersed within layer VI, which is less well defined than in
(K and L) In situ hybridization reveals abnormal location of neurons expressing the subcerebral projection neuron marker Fezf2 (normally restricted to layer V [K])
into layer VI and the white matter (empty arrowheads) in Sox5?/?neocortex (L).
(M–Q) ICC for CTIP2, a subcerebral projection neuron marker in layer V. (O) Magnified view of WT cortex taken from the boxed area in (M) shows high CTIP2
expression in layer V. Magnified views of the medial (P) and lateral (Q) Sox5?/?cortex are shown from the boxed areas in (N). (P) CTIP2-positive neurons are
aberrantly found in the deep and superficial (solid arrowheads) layers of motor cortex, as well as in the white matter (empty arrowheads). (Q) Lateral to motor
cortex, few CTIP2-positive neurons are present. Red dashed line in (M) and (N): boundary between layer V and VI.
(R–W) Neuronalpopulations insuperficial cortical layers arenotaffected,as assessedby(Rand S)the layer IVmarker RORb,(T and U)the layer II/IIImarker Cux2,
and (V and W) the layer I marker reelin.
Scale bars: 0.5 mm (A, B, M, and N), 50 mm (C, D, G–L, and O–W), 20 mm (E–E0and F).
SOX5 Controls Development of Corticofugal Neurons
236 Neuron 57, 232–247, January 24, 2008 ª2008 Elsevier Inc.
developmental processes leading to establishment of proper
To investigate whether heterotopic CTIP2-positive neurons
in Sox5?/?cortex send axons to subcerebral targets, we retro-
gradely labeled neurons from the cerebral peduncle using ultra-
sound-guided in utero injections of FG into this subcerebral
target at E17.5 and examined brains at P0 (Figure 3A). We find
that, in contrast to WT mice, in which subcerebral projection
neurons are sharply confined to layer V (Figure 3B), subcere-
brally projecting neurons (all still CTIP2 positive) are abnormally
positioned throughout all laminae in the neocortex of Sox5?/?
mice (Figures 3C–3F00). This demonstrates that the abnormal
neurons in Sox5?/?neocortex, which heterotopically express
CTIP2 and other subcerebral projection neuron markers, have
a corresponding aberrant connectivity and send their axons
We directly visualized subcerebral projections by placing
small crystals of DiI in the motor cortex of WT and Sox5?/?
mice (Figures 3G–3H000). Though many efferent axons in Sox5
null mutants have a rostrocaudal course comparable to that of
WT mice and reach the pons, there is substantial caudal defas-
ciculation in null mutants (Figures 3H and 3H0). Quite notably,
there is an additional subcerebral projection tract in Sox5?/?
mice, which runs aberrantly within the external capsule and
reaches the cerebral peduncle and pons (Figures 3H0–3H000). In
contrast, the course of corticothalamic axons is seemingly
unaffected in Sox5?/?mice (Figure S2). These data suggest
that these two subpopulations of CFu neurons are differentially
affected by the absence of SOX5 function. Consistent with the
interpretation that SOX5 does not control the development of
layer V callosal neurons and superficial layer projection neurons,
callosal projections in Sox5?/?mice are largely unaffected,
although a small subset of aberrantly located CTIP2-positive
neurons extend anomalous callosal projections (Figure S3).
Taken together, these data indicate that SOX5 functions to
regulate early differentiation of CFu neurons, including proper
laminar positioning and subcerebral connectivity.
Sox5?/?Subplate Neurons Aberrantly Differentiate into
Normally Later-Born Subcerebral Projection Neurons
As a first step in assessing the origin of the abnormally located
subcerebrally projecting neurons in Sox5?/?cortex, we per-
formed migrational analysis via BrdU birthdating. We compared
neocortex at P0. BrdU was injected in utero at four stages of
cortical development: E12.5 (SP and CTh neuron production),
E13.5 (peak subcerebral neuron production), E14.5 (peak layer
tion). Proportionally more E12.5- and E13.5-labeled birthdated
cells in Sox5?/?cortex localize to superficial layers than in WT
cortex (Figures 4A, 4B, 4E, 4F, and 4I). This abnormal migration
is limited to neurons born early during corticogenesis: the lami-
nar distributions of E14.5- and E15.5-labeled birthdated cells
are similar in WT and Sox5?/?cortex (Figure 4C, 4D, 4G, 4H,
and 4I). These data are in keeping with the specific expression
of SOX5 in CFu neuron populations and reinforce that SOX5
functions selectively during early stages of corticogenesis, at
the time when the sequential birth of CFu neurons normally
The presence of early-born (BrdU-labeled at E12.5) and Ctgf-
positive neurons in superficial layers of Sox5?/?cortex suggests
that these cells are abnormally located and/or aberrantly differ-
entiated SP neurons. We thus specifically investigated the fate
of SP neurons in Sox5?/?cortex by BrdU labeling at their peak
time of birth, E11.5. Cells born at E11.5 are almost exclusively
confined to the SP in WT mice (Figure 5A). In striking contrast,
inSox5?/?mice,E11.5 BrdU-labeled cellsareinappropriately lo-
cated in superficial laminae (Figure 5B). Importantly, these E11.5
BrdU-labeled cells strongly express CTIP2 (Figures 5C–5C00),
a protein normally expressed at high level in subcerebral projec-
tion neurons only, and not significantly in WT SP neurons (Arlotta
et al., 2005). Together, these data indicate that SP neurons dif-
ferentiate abnormally inthe absence ofSOX5: theyadopt molec-
ular and projection characteristics of subcerebrally projecting
The abnormal location of early-born neurons in Sox5?/?cor-
tex, therefore, appears to reflect primarily a fate switch with ab-
normal differentiation and secondary migrational abnormalities.
Our analyses identify substantial further evidence for this
interpretation. Superficially located neurons in Sox5?/?cortex
express Cry-mu, a gene normally expressed by subcerebral
projection neurons, but not by SP neurons (Figures 5D and 5E;
Arlotta et al., 2005). Calretinin (calret)-positive neurons, which
are largely restricted to the SP during corticogenesis (Fonseca
et al., 1995; Schierle et al., 1997), are present in superficial layers
of Sox5?/?cortex, where they strongly and aberrantly express
CTIP2 (Figures 5F–5I00). In contrast to other SP neuron markers,
calret has the advantage of being present in the cytoplasm, al-
lowing analysis of cell morphology. While calret-positive SP neu-
rons in WT cortex typically have an elongated horizontal shape
(Figures 5F and 5H), the superficially positioned calret-CTIP2
double-positive neurons in Sox5?/?cortex are distinctly pyrami-
dal shaped (Figures 5G and 5I), resembling that of normal layer V
subcerebral projection neurons (Figure 1O). Of note, many neu-
rons also express calret in deep layers of Sox5?/?cortex, sug-
gesting that calret is not suppressed in later-born CFu neuron
Finally, we directly assessed the date of birth of the aberrantly
labeling, which confirmed that these neurons in Sox5?/?mice
are born earlier than WT CTIP2-positive subcerebral projection
neurons of layer V (Figure S4).
Taken together, these molecular, morphological, and connec-
tivity data indicate that SOX5 functions to regulate and coordi-
nate the timing of emergence of the sequentially generated
populations of CFu neurons during early corticogenesis.
Combined Loss of CTIP2 and SOX5 Rescues
the Preplate Splitting Defect Caused by Aberrant
Differentiation of Subplate Neurons in Sox5?/?Mice
We next investigated whether the abnormal differentiation of SP
neurons inSox5?/?mice disruptstheirnormal function.Normally
at E12.5, neurons that will form layer VI migrate into the preplate,
dividing this structure into marginal zone (MZ) and SP in a
process called ‘‘preplate splitting.’’ We examined whether the
SOX5 Controls Development of Corticofugal Neurons
Neuron 57, 232–247, January 24, 2008 ª2008 Elsevier Inc. 237
Figure 3. Loss of SOX5 Function Causes Aberrant Subcerebral Projections
(A) Sagittal schematic view of the mouse brain, showing FG injection site in the cerebral peduncle (CeP) for retrograde labeling of subcerebral projection neurons
in the neocortex (Cx).
(B) FG injection into the CeP of WT mice at E17.5 exclusively labels CTIP2-positive neurons in layer V.
(C–F00) In Sox5?/?littermates, retrogradely labeled neurons (green) are CTIP2 positive (red) but are abnormally located in superficial (D–D00and E–E00) and
deep (F–F00) cortical layers (arrows).
(G–H000) DiI anterograde labeling of the corticospinal tract (CST) in P0 WT (G–G000) and Sox5?/?(H–H000) mice. (G and H) Ventral views of whole-mount brains show-
ing theCST (arrowheads)and anaberrant fibertractinSox5?/?brain (arrow). (G0and H0)Sagittalcomposite viewsof theareaoutlined bytheredbox (inset in[G0]);
same brain as in (G) and (H), respectively. (G00and H00) Rostral coronal sections of brains in (G) and (H), at the positions indicated by the left dotted lines. (G000
and H000) High-magnification views of the area in the red box (inset in [G000]) at caudal locations, as indicated by the right dotted lines in (G) and (H), respectively. The
CST in Sox5?/?brains exhibits fasciculation defects and a substantially limited caudal extension, but its general course is preserved, and many axons reach
the pons. An additional component of the CST inappropriately courses through the Sox5?/?external capsule, extending ventrocaudally toward the cerebral
peduncle (arrows in [H]–[H000]).
Scale bars: 0.5 mm (G–G00and H–H00), 100 mm (G000and H000), 50 mm (B and C), 10 mm (D–F00). AC, anterior commissure; CeP, cerebral peduncle; Cx, cortex; IC,
internal capsule; Mid, midbrain; OB, olfactory bulb; Th, thalamus.
SOX5 Controls Development of Corticofugal Neurons
238 Neuron 57, 232–247, January 24, 2008 ª2008 Elsevier Inc.
aberrant differentiation of SP neurons in Sox5?/?cortex af-
fects their segregation away from the MZ and impairs preplate
We find that defects in SP neurons in Sox5?/?cortex result in
incomplete preplate splitting. Cajal Retzius cells of the MZ nor-
mally express reelin and calret, while SP neurons express calret
but not reelin (Meyer et al., 2000; data not shown). At E13.5,
when preplate splitting is normally complete (Figure 6A), calret-
positive reelin-negative neurons, i.e., SP neurons, in Sox5?/?
cortex are abnormally dispersed throughout the developing cor-
esis (Figure 6B). These findings indicate that preplate splitting is
disrupted in Sox5?/?mice, likely a consequence of abnormal
differentiation of SP neurons.
A preplate splitting defect might theoretically affect the devel-
opment and migration of later-born neurons and account, at
least in part, for the aberrant differentiation of CFu neurons in
Sox5?/?cortex. We tested this possibility by examining whether
re-expression of SOX5 in Sox5?/?cortical neurons, after abnor-
mal preplate splitting has already occurred, is sufficient to re-
store proper differentiation of subcerebral projection neurons.
We find that in utero retroviral transfer of the Sox5 gene in
Sox5?/?E12.5 embryos enabled CTIP2-positive (i.e., subcere-
sition, whereas the laminar location of CTIP2-positive neurons
infected with a control virus (expressing solely GFP), in the
same embryo, remained abnormal. Only 23% of CTIP2-positive
neurons infected with the control virus were located in layer
V (7/30) versus 66% of CTIP2-positive neurons infected with
the SOX5-expressing virus (37/56, n = 2 mice, p = 0.0002)
(Figure S5). Re-expression of SOX5 at a time when subcerebral
projection neurons are being born is thus sufficient to rescue the
laminar positioning of CTIP2-positive subcerebral projection
neurons in Sox5?/?mice, indicating that the abnormal differenti-
ation of these neurons is at least predominantly cell-autono-
In the absence of SOX5, CTIP2 is abnormally expressed at
high levels in mutant SP neurons, while it is minimally expressed
in WT SP neurons (Figures 5H–5H00and 5I–5I00). We therefore
hypothesized that SOX5 normally regulates the emergence of
CTIP2 expression (and potentially of other coordinately regu-
lated genes) in CFu neurons and that loss of SOX5 function
Figure 4. Early Corticogenesis Is Selectively Impaired in the Sox5?/?Neocortex
(A–D) Distributions of BrdU birthdate-labeled cells in P0 neocortex after BrdU injection at four time points during embryonic development in WT and Sox5?/?
mice. The cortical thickness was divided into ten equal bins from pia to white matter, and the distribution of BrdU-labeled cells across the bins was determined.
Dark-blue-outlined empty bars, WT; light-blue bars, Sox5?/?. Average of two (C and D) or three (A and B) experiments.
(A, B, E, F, and I) In Sox5?/?mice, E12.5-born and E13.5-born cells are located in excess in superficial cortical layers (bins 1–3) at P0.
(C, D, G, H, and I) Later-born neurons are not significantly affected.
(E–H) Representative photomicrographs of neocortical sections with ICC for BrdU. The strongest E12.5 and E15.5 BrdU birthdate-labeled cells are circled in red
in (E)–(F) and (G)–(H), respectively. The bold white line indicates the lower limit of the cortical plate (CP, bins 1–3); thin lines indicate bin boundaries.
(I) Graphical summary of the results from (A)–(D), indicating a significantly increased proportion of early-born cells in the developing CP of Sox5?/?mice (E12.5,
E13.5), while later-born cells (E14.5, E15.5) distribute normally. Empty circles represent values from individual experiments; connected circles indicate litter-
mates. Asterisked brackets indicate statistically significant differences between WT and Sox5?/?at E12.5 and E13.5 (p < 0.05).
Scale bars, 50 mm (E–H).
SOX5 Controls Development of Corticofugal Neurons
Neuron 57, 232–247, January 24, 2008 ª2008 Elsevier Inc. 239
would result in inappropriate and premature expression of
role of SOX5 as a transcriptional repressor. Given the critical role
of CTIP2 in normal differentiation of subcerebral projection neu-
rons (Arlotta et al., 2005), inappropriate expression of CTIP2
might be causative in the aberrant differentiation of SP neurons
Figure 5. Sox5?/?Subplate Neurons Abnormally Differentiate into Normally Later-Born Subcerebral Projection Neurons
(A and B) Distribution of E11.5 BrdU birthdate-labeled cells (green, circled in red) in P0 neocortex of WT and Sox5?/?mice.
(C–C00) Magnification of the boxed region in (B). Cells born at E11.5 are found throughout Sox5?/?cortex and express the subcerebral projection neuron marker
(D and E) In situ hybridization for the normally layer Vsubcerebral projection neuronmarkerCrystallin-mu (Cry-mu)shows abnormally located Cry-mu-expressing
neurons in the superficial (arrows) and deepest layers of Sox5?/?cortex.
(F–I00) ICC for CTIP2 (red) and the SP neuron marker calretinin (calret, green). (F and H–H00) In WT cortex, calret-positive neurons are found mostly in the SP and
have low or absent CTIP2 expression (empty arrows in [H]–[H00], magnified from boxed region in [F]). (G and I–I00) In Sox5?/?cortex, calret-positive neurons are
aberrantly located throughout the cortex, quite notably in the superficial cortical plate, and both have a pyramidal morphology and strongly express the subcere-
bral projection neuron marker CTIP2 (solid arrows in [I]–[I00]), magnified from boxed region in (G).
Scale bars: 50 mm (A, B, and D–G), 20 mm (C–C00, H–H00, and I–I00).
SOX5 Controls Development of Corticofugal Neurons
240 Neuron 57, 232–247, January 24, 2008 ª2008 Elsevier Inc.
into mixed-phenotype subcerebrally projecting neurons and
secondarily disrupt preplate splitting.
To test this hypothesis, we generated Sox5;Ctip2 double null
mutant mice. We reasoned that, if inappropriate expression of
CTIP2 causes abnormal differentiation of SP neurons, then the
preplate splitting defect should be less severe in Sox5;Ctip2
double null cortex than in Sox5 single null cortex. In agreement
with this hypothesis, we find that genetic ablation of Ctip2
partially rescues preplate splitting; twice as many SP neurons
are apposed to the MZ in Sox5?/?than in Sox5?/?;Ctip2?/?
mice (8.5% versus 4.0%, n = 2 for Sox5?/?, n = 3 for Sox5?/?;
Ctip2?/?; Figures 6C–6G). Interestingly, this rescue shows
a dose dependence with CTIP2 levels, since in Sox5?/?;Ctip2+/?
mice, the proportion of MZ-apposed SP neurons is intermediate
between Sox5?/?and Sox5?/?;Ctip2?/?mice (5.5%, p < 0.05;
Figure 6G). Taken together, these findings strongly support the
interpretation that, during early corticogenesis, SOX5 function
leads to CTIP2 repression in SP neurons and that absence of
this repression results in inappropriate adoption of mixed and
aberrant characteristics of normally later-born subcerebral
Gain-of-Function Analysis: SOX5 Misexpression
Disrupts Normal Differentiation of Callosal Projection
Neurons and Shifts Their Differentiation toward
Corticofugal Neuron Fate
Our loss-of-function results indicate that lack of SOX5 function
leads early-born CFu neurons to inappropriately and prema-
turely develop characteristics of later-born CFu neurons. We
next investigated whether overexpression of SOX5 past the nor-
mal period of generation of CFuneurons (here referred to as mis-
expression) might induce the converse in later-born, non-CFu
neuron populations, i.e., abnormal acquisition of characteristics
of an earlier-born neuron subtype. Specifically, we examined
whether misexpression of SOX5 in the developing cortex after
E13.5, past the normal period of generation of CFu neurons,
would cause differentiating CPN to develop characteristics of
(previously born) CFu neurons.
We misexpressed SOX5 at multiple developmental stages.
First, we misexpressed SOX5 from E12.5 onward via in utero
electroporation in WT mice using a plasmid construct in which
expression of the Sox5 coding sequence is under the control of
a constitutively active CMV/b actin promoter with an IRES-GFP
Figure 6. Loss of CTIP2 Function Partially Rescues the Preplate Splitting Defect in Sox5?/?Cortex
(A and B) Calret and CTIP2 ICC. (A) At E13.5, in control Sox5+/?cortex, calret-positive neurons of the MZ and SP (arrows) are separated by CTIP2-positive neu-
rons of the cortical plate (CP). (B) In contrast, in Sox5?/?littermate cortex, these two calret-positive neuron populations are not separated, indicating incomplete
preplate splitting; there is abnormal laminar positioning and intermingling of the calret-positive (arrows) and CTIP2-positive corticofugal neuron populations.
VZ, ventricular zone.
(C–G) ICC for calret in Sox5?/?;Ctip2+/+and Sox5?/?;Ctip2?/?cortex at P0. (E and F) Boxed regions from (C) and (D) at higher magnification. The dashed lines
in (C)–(F) demarcate the most superficial 10% of the cortex, in which calret-positive neurons are abnormally located in Sox5?/?mice. This abnormality is partially
rescued in Sox5?/?;Ctip2?/?mice: fewer calret-positive SP neurons remain apposed to the MZ in Sox5?/?;Ctip2?/?cortex (D and F) than in Sox5?/?;Ctip2+/+
cortex (C and E), while Sox5?/?;Ctip2+/?cortex has intermediate values, as quantified in (G). Empty circles represent values from individual mice, connected
circles indicate littermates.
Scale bars: 100 mm (C and D), 50 mm (A, B, E, and F).
SOX5 Controls Development of Corticofugal Neurons
Neuron 57, 232–247, January 24, 2008 ª2008 Elsevier Inc. 241
element for identification of electroporated cells (Molyneaux
et al., 2005). A control vector containing only GFP was used in
parallel control experiments.
Misexpression of SOX5 dramatically affects the differentiation
of CPN (Figure 7), which normally do not express this transcrip-
tion factor. When the control vector was used, axons of GFP-
positive CPN always crossed the midline (5/5 embryos) and
had typically reached lateral regions in the opposite hemisphere
by E18.5 (3/5 embryos; Figures 7C and 7C0). In striking contrast,
when the SOX5 expression vector was used, axons of SOX5-
GFP double-positive CPN never extended laterally into the
opposite hemisphere (0/5 embryos) and only occasionally even
reached the midline (2/5 embryos; Figures 7E and 7E0). CFu neu-
rons in layers V and VI were not significantly affected by overex-
pression of SOX5, as demonstrated by the similar numbers of
subcortical and subcerebral GFP-positive axons in the internal
Figure 7. SOX5 Misexpression Impairs Callosal Axonal Projections and Generates De Novo Corticofugal Axons
(A) Schematic view of a coronal brain section indicating regions imaged in (B)–(E). (C0) and (E0) are enlarged from boxed regions in (C) and (E), respectively.
(B–E0)Imagesfrom coronal brain sections after electroporationat E12.5, collected at E18.5. (Band D) The laminar locations of neurons labeledby electroporation
with the Sox5GFPconstruct at E12.5 (green cells in [D]) do not differ detectably from those electroporated with the ControlGFPconstruct (green cells in [B]). (B, C,
and C0) Callosal axons are unaffected (solid arrowheads in [B]) and project to the opposite hemisphere in control electroporations (solid arrowheads in [C0]). (D, E,
and E0) Callosal projections are dramatically impaired by SOX5 misexpression, not extending for more than a few hundred microns (empty arrowheads in [D]).
(F) Schematic view of a coronal brain section indicating regions imaged in (G)–(J).
(G–J0) Images from coronal brain sections after electroporation at E14.5, collected at E18.5. (G and I) Neurons labeled by electroporation with either the Con-
trolGFP(G)or Sox5GFP(I) construct at E14.5 are equivalently localized to the most superficial portion of the cortical plate. (J and J0) SOX5 misexpression generates
neuronsthatsendaxons throughthestriatum(Str)(arrowheads),whereassubcorticalprojections areneverpresentincontrolelectroporatedmice(H).Thedotted
lines in (H) and (J) outline the striatum, where CTIP2 is expressed (Arlotta et al., 2005, 2008). The subcortical fibers in (J) are shown in magnified view in (J0).
Scale bars: 200 mm (B–C0, D–E0, and G–J), 100 mm (J0).
SOX5 Controls Development of Corticofugal Neurons
242 Neuron 57, 232–247, January 24, 2008 ª2008 Elsevier Inc.
capsule and cerebral peduncle when either plasmid was used
(data not shown). Taken together, these data indicate that
misexpression of SOX5 specifically disrupts the proper differen-
tiation of CPN.
We investigated the possibility that this reduced callosal axon
extension might reflect aberrant development of CFu neuron
characteristics in callosal neurons when SOX5 is misexpressed.
misexpressing SOX5 might develop anomalous corticofugal
axons. Therefore, we examined whether corticofugal axons
might originate from normally callosal-projecting neurons in
Sox5-electroporated brains. In order to avoid the confounding
presence of normal corticofugal projections, we expressed
SOX5 from E14.5 on, a time after which CFu neurons are no
longer normally generated. We find that, whereas CPN never ex-
tended corticofugal axons in brains electroporated with the con-
trol vector (0/4 mice; Figure 7H), corticofugal axons were always
visible in the internal capsule of Sox5-electroporated mice (3/3
mice; Figures 7J and 7J0). Taken together, these results strongly
indicate that misexpression of SOX5 beyond the normal period
of generation of CFu neurons is sufficient to cause a partial
fate-switch of callosal neurons and the abnormal development
of neurons with corticofugal features at late stages of cortico-
Integrating the loss-of-function and gain-of-function data,
these experiments indicate a critical role for SOX5 in regulating
the temporal precision of development of sequentially born
populations of CFu neurons. SOX5 thus regulates the complex
emergence of neuronal diversity of CFu neurons. This regulation
defines and controls the proper organization, output circuit
formation, and function of the neocortex.
The identification and characterization of molecular controls
over the differentiation of newborn neocortical neurons into dis-
ing development. We show here that the transcription factor
SOX5 is specifically expressed in the three central and closely
related corticofugal neuron subtypes and that it controls the tim-
sity of neocortical output neurons.
By combining anatomical, molecular, and birthdating ap-
proaches in null mutant mice, we show that SOX5 is required
principal CFu neuron subtypes, during which SP neurons are
born first, CTh neurons second, and subcerebral projection neu-
rons last. Loss of SOX5 function specifically disrupts the sub-
type-specific differentiation programs of CFu neurons, causing
the premature and inappropriate acquisition of subcerebral pro-
jection neuron features by early-born SP neurons, leading to
a blurring of the identities of these normally precisely defined
Remarkably, while early-born neurons incorrectly develop
characteristics of later-born neurons in the absence of SOX5,
misexpression of this transcription factor after the normal period
of generation of CFu neurons has a reciprocal effect: the abnor-
mal development of neurons with features of (normally earlier-
born) CFuneurons atlate stages of corticogenesis. Weconclude
from these developmental expression, loss-of-function, and
gain-of-function studies that SOX5 is a critical regulator of CFu
neuron differentiation programs, acting to precisely control the
temporal sequence of differentiation and the identity of the
closely related, yet distinct, subtypes of neurons.
Role of SOX5 in the Delineation of the Sequentially
Born Corticofugal Neuron Subtypes
In contrast to WT mice, in which the sequentially born CFu neu-
ronal subtypes express characteristic levels and combinations
of genes and have a defining connectivity, several idiosyncratic
features that normally distinguish these distinct populations are
found across CFu neuron subtypes in Sox5?/?brain. This results
in an overlap, or ‘‘confusion,’’ in the molecular, laminar, and con-
nectivity identities of these neurons. Interestingly, even these
abnormal and partially ‘‘confused’’ neurons can still be distin-
guished somewhat by their relative expression of a core set of
subcerebral projection neuron (including CSMN) specific genes
and SP markers, and by their dates of birth (Figures 2 and 5). A
first subset of neurons in Sox5?/?cortex, located in the subcor-
tical white matter and deep layer VI, exclusively expresses
CSMN-specific genes (in particular Ctip2 and Fezf2) and not
SP neuron markers (Ctgf, calret) or TBR1. This strongly suggests
that these neurons are the ‘‘genuine’’ subcerebral projection
neurons that have failed to complete their migration into layer
V. A second subset of abnormal neurons, located mostly within
layer VI of Sox5?/?cortex, expresses a ‘‘confused’’ combination
ing mixed molecular features of all CFu neurons. Finally, abnor-
mal neurons located in the superficial layers of Sox5?/?cortex
are earliest born and express only a subset of CSMN-specific
genes (e.g., CTIP2, Cry-mu, but not Fezf2) while expressing
both SP markers calret and Ctgf; these are therefore likely to
be ‘‘genuine’’ SP neurons with aberrant differentiation. Taken to-
gether, these findings demonstrate that loss of Sox5 function
causes a striking lack of precision of CFu neuron differentiation
and dramatically reduces neuronal diversity, eliminating the ac-
curate molecular and connection features of CFu populations
that underlie the precision of cortical output function.
SOX5 is expressed in CFu neurons, but not in CPN of layers
II/III and V, suggesting that the sequentially born CFu neuron
subtypes share a common lineage distinct from that of intracort-
ical projection neurons. Loss of SOX5 function, however, does
not affect the three CFu neuron subtypes equally; this might
be informative regarding their relative lineage relationships. For
example, differentiating SP neurons are exquisitely sensitive to
the absence of SOX5 function and undergo a dramatically aber-
rant differentiation in its absence, while next-born CTh neurons
mostly remain within layer VI and establish relatively normal pro-
jections to the thalamus (Figure S2). The contrast between the
relative sparing of layer VI CTh neurons and the clearly abnormal
differentiation of SP neurons in Sox5?/?mice is somewhat
lecular similarities of these neuronal subtypes. The fact that sub-
cerebral projection neurons and SP neurons are both severely
SOX5 Controls Development of Corticofugal Neurons
Neuron 57, 232–247, January 24, 2008 ª2008 Elsevier Inc. 243
affected by loss of SOX5 function suggests that they might be
more closely related in lineage to each other than to layer VI
CTh neurons. In support of this hypothesis, we find that a signif-
icant proportion of CSMN-specific genes (9/18 examined) are
also expressed in SP neurons during cortical development, but
not in layer VI neurons. In contrast, callosal projection neuron
specific genesin layerVareonlyrarelyexpressedbySPneurons
(Arlotta et al., 2005; P.A. and B.J.M., unpublished data), demon-
strating that the correlated gene expression pattern of subcere-
bral projection neurons with SP neurons is not simply due to
subcerebral projection neurons’ layer V location. These data
regarding multiple molecular linkages between neuron subtypes
therefore suggest a previously unsuspected level of ontogenetic
connection between SP neurons and subcerebral projection
It is interesting to note that the period of corticogenesis during
which SP neurons are born progressively lengthens with mam-
malian evolution, resulting in a considerably expanded SP in pri-
mates (Kostovic and Rakic, 1990). The anatomical and temporal
expansion of the SP during evolution is thought to have played
a critical role in enabling afferent fibers to invade the cortex
from below, instead of from the cortical surface, as is the case
in sauropsids (reptiles and birds). This, in turn, is thought to
have allowed the cortex to expand from three to six layers in
an inside-out fashion (Marin-Padilla, 1992; Molnar et al., 2006;
Super and Uylings, 2001). It is interesting to speculate that
SOX5 function might have facilitated SP expansion by prevent-
ing the emergence of normally later-born neuron features in SP
neurons, enabling this critical neuronal population to be gener-
ated over a prolonged period of corticogenesis, simultaneously
with other subtypes of CFu neurons.
Role of SOX5 in the Differentiation
of Projection Neurons
Thesubcortical locationandgeneexpressionprofile ofthe‘‘gen-
uine’’ subcerebral projection neurons in Sox5?/?cortex are quite
similar to those of heterotopic CTIP2-positive subcerebrally pro-
jecting neurons generated after acute overexpression of the
transcription factor FEZF2, a gene that is critical for subcerebral
projection neuron specification (Molyneaux et al., 2005). In these
previous experiments, overexpression of FEZF2 induced a fate
switch of progenitors for superficial layer cortical neurons and
in secondary migratory arrest and white matter heterotopias
(Molyneaux et al., 2005). The similar presence of heterotopic
white matter subcerebral projection neurons in Sox5?/?mice
suggests thatpremature expressionof laterdifferentiation genes
like CTIP2, a critical control over CSMN axon outgrowth and fas-
ciculation (Arlotta et al., 2005), may lead to premature migratory
arrest. SOX5 therefore might act in migrating prospective sub-
cerebral projection neurons, to control the proper timing and
emergence of their early differentiation program. In support of
this interpretation, SOX5 has been reported to set the proper
pace and sequence of gene expression in chondroblasts during
cartilage development and in oligodendrocyte precursors during
spinal cord development, as shown by the premature differenti-
ation of these cell types in the absence of SOX5 function (Smits
et al., 2001; Stolt et al., 2006).
Consistent with an intrinsic role of SOX5 in regulating the
sequential emergence of distinct CFu neuron subtypes, misex-
pressionof SOX5inlate-born CPNissufficientto induceapartial
CFu neuron differentiation program and cause the acquisition of
corticofugal features. SOX5 misexpression from E12.5 onward
resulted in impaired callosal axonal extension, possibly reflect-
ing the partial adoption of a CFu neuron differentiation program
by the layer V CPN and the subsequent interference of normal
callosal projection neuron development.
The specificity of SOX50s effects in callosal neurons, which do
not normally express SOX5 (Figures 1P–1Q00), likely reflects the
lineage proximity of neocortical projection neurons. Despite
some distinctive and fundamental molecular and functional dif-
ferences, all cortical projection neurons share highly similar gen-
eral features and transcriptional profiles and are therefore likely
to include SOX5 ‘‘downstream’’ effector genes. As also recog-
nized in other systems, the cell-subtype-specific differentiation
of individual subpopulations of neocortical projection neurons
therefore likely relies on differences in the combinatorial interac-
etal., 2005, and unpublished observations) ratherthan on funda-
mental differences in the overall components of differentiation
Neocortical Efferent and Afferent Tracts
The aberrant differentiation and redistribution of CFu neurons in
Sox5?/?cortex have dramatic consequences on cortical con-
nectivity. In Sox5?/?cortex, CFu neurons that do not normally
send subcerebral projections extend their axons to subcerebral
targets, including via an exuberant ectopic tract, presumably
a consequence of the aberrant emergence of neurons with sub-
cerebral projection neuron features from the earliest stages of
corticogenesis. As demonstrated by retrograde labeling from
a subcerebral target, both nominal SP neurons and ‘‘genuine’’
subcerebral projection neurons send subcerebral projections
in Sox5?/?brains. The presence of layer VI subcerebrally projec-
ting CTIP2-negative neurons in lateral cortical areas suggests
that at least some CTh neurons also send subcerebral projec-
tions. Therefore, the lack of precision of neuronal subtype-spe-
cific expression of subcerebral projection neuron genes across
responding lack of precision and diversity in the connectivity of
these neurons, leading to an increased propensity to project
A surprising axonal pathfinding feature of Sox5?/?mice is the
seemingly preserved ability of thalamocortical neurons to find
their target despite the lack of a morphological SP (Figures
S1B and S2). The SP has been proposed to guide the develop-
ment of thalamocortical axons, which is perturbed to varying
degrees when preplate splitting is abnormal or when the SP is
injured (Allendoerfer and Shatz, 1994; Hevner et al., 2001;
cortical and corticothalamic interactions have been thought to
contribute to each other’s axonal pathfinding and connectivity
(Molnar and Blakemore, 1995). Thus, the globally preserved
anatomy of afferent and efferent thalamic pathways in Sox5?/?
mice suggests that, despite aberrant differentiation, at least
SOX5 Controls Development of Corticofugal Neurons
244 Neuron 57, 232–247, January 24, 2008 ª2008 Elsevier Inc.
a subset of SP neurons retains the ability to enable appropriate
thalamocortical and corticothalamic connectivity.
Proposed Mechanisms of SOX5 Action
The nominal SP neurons of Sox5?/?cortex are an extremely in-
teresting population, since they can shed light on the functional
consequences and mechanisms of the inappropriate initiation
rant differentiation of Sox5?/?SP neurons appears to dramati-
cally affect their normal function during preplate splitting, by
impeding their proper segregation away from the MZ. This im-
pediment results at least in part from increased CTIP2 expres-
sion in these earliest-born neurons, since genetic elimination of
CTIP2 function in Sox5;Ctip2 double null mice partially rescues
Consistent with the nominal role of SOX5 as a transcriptional
repressor, we propose that SOX5 normally inhibits the expres-
sion of coordinately regulated genes, including CTIP2, whose
levels normally progressively increase as the distinct subtypes
of CFu neurons are sequentially generated. Interestingly, loss
of SOX5 function has strikingly different consequences in medial
motor versus lateral nonmotor regions, even though this tran-
scription factor is normally expressed in both areas evenly, sug-
gesting that areal-specific interactions occur with other proteins
expressed along gradients. This inhibitory role of SOX5 likely re-
quires interactions with additional temporally regulated and cell-
type-specific gene products, since elimination of CTIP2 function
only partially corrects preplate splitting.
Repression of the inappropriate expression of CTIP2 (and
other genes)bySOX5 duringearly corticogenesis offers interest-
ing parallels with the ‘‘derepression model’’ proposed for spinal
cord development, according to which cell fate is progressively
specified by the sequential disinhibition of effector genes (Lee
and Pfaff, 2001). Similarly, other genes have been shown to act
by repressing differentiation of one cell subtype into another
stitutively repress Cajal-Retzius differentiation in deep layer neu-
rons (Hanashima et al., 2004). Another example is Notch1, which
has been shown to inhibit photoreceptor fate in the mammalian
retina (Jadhav et al., 2006).
In conclusion, our results indicate that SOX5 exerts critical
control over the sequential and precise differentiation of the dis-
tinct CFu neuron subtypes. The data indicate that SOX5 coordi-
nates the temporal generation of these sequentially born neuron
subtypes by repressing coordinately regulated genes and by
preventing inappropriate and premature emergence of subcere-
bral projection neuron features early in corticogenesis. Because
in the neocortex SOX5 is not detectably expressed in dividing
neural precursors, but is exclusively expressed in postmitotic
neurons, it is likely that it regulates the expression of genes
poised for activation in each of the CFu neuron populations. Fur-
ther elucidation of the combinatorial molecular programs by
which SOX5 functions to regulate fate specification should offer
important new insights into the evolutionary and functional rela-
tionships between distinct neuronal populations, the generation
of neuronal diversity, the precision of neuronal differentiation,
and potential plasticity in the mammalian neocortex.
Sox5+/?mice were the generous gift of V. Lefebvre (Smits et al., 2001) (Sox5
GenBank accession number 20678). Ctip2+/?mice were the generous gift of
R. Kominami and colleagues (Wakabayashi et al., 2003).The day of the vaginal
plug detection was designated as embryonic day 0.5 (E0.5). The day of birth
was designated postnatal day 0 (P0). All mouse studies were approved by
the Massachusetts General Hospital IACUC and were performed in accor-
dance with institutional and federal guidelines.
Immunocytochemistry and In Situ Hybridization
Brains were fixed and stained using standard methods (Fricker-Gates et al.,
2002). Primary antibodies and dilutions were used as follows: rat anti-CTIP2,
1:1000 (Abcam); rabbit anti-TBR1 antibody, 1:8000, gift of R. Hevner; rabbit
anti-RORb, 1:2000, gift of H. Stunnenberg (Gawlas and Stunnenberg, 2000);
rat anti-BrdU, 1:750 (Accurate); mouse anti-BrdU, 1:750 (Chemicon); rabbit
anti-GFP, 1:1000 (Molecular Probes); goat anti-SOX5, 1:250 (Santa Cruz
Biotech); rabbit anti-b-gal, 1:5000 (MP Biomedical); mouse anti-reelin, 1:500
(Chemicon); rabbit anti-FG, 1:500 (Chemicon); rabbit anti-calretinin, 1:1000
(Chemicon). Appropriate secondary antibodies were from the Molecular
Probes Alexa series.
Nonradioactive in situ hybridization was performed using reported methods
(Berger and Hediger, 2001). Riboprobes were generated as previously de-
scribed (Arlotta et al., 2005).
Timed pregnant females received a single intraperitoneal injection of BrdU
(100 mg/kg) at E11.5, E12.5, E13.5, E14.5, or E15.5. Pups were collected at
birth and processed for BrdU immunocytochemistry (ICC) (Magavi et al.,
2000). Quantification of first-generation BrdU-labeled cells and distribution
within cortical layers were analyzed using established methods (Molyneaux
et al., 2005). Four anatomically matched sections from each mouse were se-
lected, BrdU ICC and DAPI staining were performed, and fluorescence photo-
micrographs were obtained spanning motor cortex of both hemispheres. Two
orthreesetsofWTandSox5?/?littermates wereexamined foreachtimepoint.
The full cortical thickness was divided evenly into ten bins, and the distribution
of first-generation BrdU-positive cells (defined a priori as having strong and
homogeneous nuclear labeling) was determined for all bins. A priori criteria
were defined for blinded analysis by two independent investigators; interob-
server assessments for distribution of labeled cells were nearly indistinguish-
able. One investigator was not involved in the study, and both were blinded to
the genotype and time of BrdU injection. The unpaired t test was used for
Anterograde and Retrograde Labeling
Anterograde and retrograde DiI tracing in postfixed brains were performed as
previously described (Arlotta et al., 2005; O’Leary and Terashima, 1988). Sub-
cerebrally, subcortically, and callosally projecting neurons in the motor cortex
were retrogradely labeled via FG injections into the peduncle, thalamus, and
contralateral hemisphere, respectively, at E17.5, P0, or P1 under ultrasound
guidance (Vevo 660, VisualSonics; Arlotta et al., 2005). E17.5 pregnant mice
were deeply anesthetized with Avertin, and each embryo was injected through
the uterine wall (Arlotta et al., 2005). Injected mice were collected at P0, P3, or
P4 and processed for ICC.
For control experiments, a vector containing IRES-GFP under the control of
aconstitutively active CMV/bactin promoterwasused (controlGFP,agenerous
gift of C. Lois, MIT; Molyneaux et al., 2005). Sox5 was cloned into this vector to
create the construct Sox5GFPfor misexpression. 750 nl of purified DNA (0.5–
1.0 mg/ml) mixed with 0.005% Fast Green was injected in utero into the lateral
ventricle of CD1 embryos at E12.5 or E14.5 under ultrasound guidance and
electroporated into the neocortical ventricular zone essentially according to
Saito and Nakatsuji (2001) and Molyneaux et al. (2005). For E12.5 electropora-
tion, five controlGFP-electroporated pups from four different litters and five
Sox5GFP-electroporated pups from five different litters were analyzed. For
E14.5 electroporation, four controlGFP-electroporated pups from three
SOX5 Controls Development of Corticofugal Neurons
Neuron 57, 232–247, January 24, 2008 ª2008 Elsevier Inc. 245
different litters and three Sox5GFP-electroporated pups from two different lit- Download full-text
ters were analyzed.
The Supplemental Data for this article can be found online at http://www.
This work was partially supported by grants from the National Institutes of
Health (NS45523 and NS49553), the Harvard Stem Cell Institute, and the
Spastic Paraplegia Foundation to J.D.M., and by the Jane and Lee Seidman
Fund for CNSResearch,and the Emilyand Robert Pearlstein Fund for Nervous
System Repair. T.L. was partially supported by NIH Training Grant T32
NS22215. D.J. was partially supported bythe SwissNational Science Founda-
tion. We thank K. Bilmers, K. Quinn, and A. Palmer for technical assistance;
U. Berger for assistance with in situ hybridization; Drs. V. Lefebvre, R. Hevner,
H. Stunnenberg for generous sharing of mice and antibodies; and members of
the Macklis lab for helpful suggestions.
Received: July 5, 2007
Revised: November 30, 2007
Accepted: December 19, 2007
Published: January 23, 2008
Aboitiz, F., and Montiel, J. (2003). One hundred million years of interhemi-
spheric communication: the history of the corpus callosum. Braz. J. Med.
Biol. Res. 36, 409–420.
Allendoerfer, K.L., and Shatz, C.J. (1994). The subplate, a transient neocortical
structure: its role in the development of connections between thalamus and
cortex. Annu. Rev. Neurosci. 17, 185–218.
during histogenesis of cerebral cortex in the mouse. Nature 192, 766–768.
Arlotta, P., Molyneaux, B.J., Chen, J., Inoue, J., Kominami, R., and Macklis,
J.D. (2005). Neuronal subtype-specific genes that control corticospinal motor
neuron development in vivo. Neuron 45, 207–221.
Arlotta, P., Molyneaux, B.J., Jabaudon, D., Yoshida, Y., and Macklis, J.D.
(2008). CTIP2 controls the differentiation of medium spiny neurons and the
establishment of the cellular architecture of the striatum. J. Neurosci. 3,
Berger, U.V., and Hediger, M.A. (2001). Differential distribution of the gluta-
mate transporters GLT-1 and GLAST in tanycytes of the third ventricle.
J. Comp. Neurol. 433, 101–114.
Bylund, M., Andersson, E., Novitch, B.G., and Muhr, J. (2003). Vertebrate neu-
rogenesis is counteracted by Sox1-3 activity. Nat. Neurosci. 6, 1162–1168.
Chen, B., Schaevitz, L.R., and McConnell, S.K. (2005a). Fezl regulates the dif-
ferentiation and axon targeting of layer 5 subcortical projection neurons in
cerebral cortex. Proc. Natl. Acad. Sci. USA 102, 17184–17189.
Chen, J.G., Rasin, M.R., Kwan, K.Y., and Sestan, N. (2005b). Zfp312 is re-
quired for subcortical axonal projections and dendritic morphology of deep-
layer pyramidal neurons of the cerebral cortex. Proc. Natl. Acad. Sci. USA
Deng, J., and Elberger, A.J. (2003). Corticothalamic and thalamocortical path-
finding in the mouse: dependence on intermediate targets and guidance axis.
Anat. Embryol. (Berl.) 207, 177–192.
Desai, A.R., and McConnell, S.K. (2000). Progressive restriction in fate poten-
tial by neural progenitors during cerebral cortical development. Development
Fonseca,M., del Rio, J.A., Martinez, A.,Gomez, S.,and Soriano, E. (1995). De-
velopment of calretinin immunoreactivity in the neocortex of the rat. J. Comp.
Neurol. 361, 177–192.
Fricker-Gates, R.A., Shin, J.J., Tai, C.C., Catapano, L.A., and Macklis, J.D.
(2002).Late-stage immature neocorticalneuronsreconstructinterhemispheric
connections and form synaptic contacts with increased efficiency in adult
mouse cortex undergoing targeted neurodegeneration. J. Neurosci. 22,
Gawlas, K., and Stunnenberg, H.G. (2000). Differential binding and transcrip-
tional behaviour of two highly related orphan receptors, ROR alpha(4) and
ROR beta(1). Biochim. Biophys. Acta 1494, 236–241.
Guillemot, F., Molnar, Z., Tarabykin, V., and Stoykova, A. (2006). Molecular
mechanisms of cortical differentiation. Eur. J. Neurosci. 23, 857–868.
Gupta, A.,Tsai, L.H., and Wynshaw-Boris,A.(2002).Life is ajourney: agenetic
look at neocortical development. Natl. Rev. 3, 342–355.
Hanashima, C., Li, S.C., Shen, L., Lai, E., and Fishell, G. (2004). Foxg1 sup-
presses early cortical cell fate. Science 303, 56–59.
Heuer, H., Christ, S., Friedrichsen, S., Brauer, D., Winckler, M., Bauer, K., and
Raivich, G. (2003). Connective tissue growth factor: a novel marker of layer VII
neurons in the rat cerebral cortex. Neuroscience 119, 43–52.
Hevner, R.F., Shi, L., Justice, N., Hsueh, Y., Sheng, M., Smiga, S., Bulfone, A.,
Goffinet, A.M., Campagnoni, A.T., and Rubenstein, J.L. (2001). Tbr1 regulates
differentiation of the preplate and layer 6. Neuron 29, 353–366.
Hevner, R.F., Hodge, R.D., Daza, R.A., and Englund, C. (2006). Transcription
factors in glutamatergic neurogenesis: conserved programs in neocortex,
cerebellum, and adult hippocampus. Neurosci. Res. 55, 223–233.
Hong, C.S., and Saint-Jeannet, J.P. (2005). Sox proteins and neural crest
development. Semin. Cell Dev. Biol. 16, 694–703.
Jadhav, A.P., Mason, H.A., and Cepko, C.L. (2006). Notch 1 inhibits photore-
ceptor production in the developing mammalian retina. Development 133,
Kostovic, I., and Rakic, P. (1990). Developmental history of the transient sub-
plate zone in the visual and somatosensory cortex of the macaque monkey
and human brain. J. Comp. Neurol. 297, 441–470.
Lee, S.K., and Pfaff, S.L. (2001). Transcriptional networks regulating neuronal
identity in the developing spinal cord. Nat. Neurosci. Suppl. 4, 1183–1191.
Lefebvre, V. (2002). Toward understanding the functions of the two highly
related Sox5 and Sox6 genes. J. Bone Miner. Metab. 20, 121–130.
Magavi,S.S., Leavitt,B.R.,andMacklis, J.D.(2000).Inductionofneurogenesis
in the neocortex of adult mice. Nature 405, 951–955.
Marin-Padilla, M. (1992). Ontogenesis of the pyramidal cell of the mammalian
neocortex and developmental cytoarchitectonics: a unifying theory. J. Comp.
Neurol. 321, 223–240.
McConnell, S.K., and Kaznowski, C.E. (1991). Cell cycle dependence of lam-
inar determination in developing neocortex. Science 254, 282–285.
McConnell, S.K., Ghosh, A., and Shatz, C.J. (1989). Subplate neurons pioneer
the first axon pathway from the cerebral cortex. Science 245, 978–982.
McConnell, S.K., Ghosh, A., and Shatz, C.J. (1994). Subplate pioneers and the
formation of descending connections from cerebral cortex. J. Neurosci. 14,
Meyer, G., Schaaps, J.P., Moreau, L., and Goffinet, A.M. (2000). Embryonic
and early fetal development of the human neocortex. J. Neurosci. 20, 1858–
Molnar, Z., and Blakemore, C. (1995). How do thalamic axons find their way to
the cortex? Trends Neurosci. 18, 389–397.
Molnar, Z., Metin, C., Stoykova, A., Tarabykin, V., Price, D.J., Francis, F.,
Meyer, G., Dehay, C., and Kennedy, H. (2006). Comparative aspects of cere-
bral cortical development. Eur. J. Neurosci. 23, 921–934.
Molyneaux, B.J., Arlotta, P., Hirata, T., Hibi, M., and Macklis, J.D. (2005). Fezl
is required for the birth and specification of corticospinal motor neurons.
Neuron 47, 817–831.
Molyneaux, B.J., Arlotta, A., Menezes, J.R., and Macklis, J.D. (2007). Neuronal
subtype specification in the cerebral cortex. Nat. Rev. Neurosci. 8, 427–437.
SOX5 Controls Development of Corticofugal Neurons
246 Neuron 57, 232–247, January 24, 2008 ª2008 Elsevier Inc.