Highly conserved non-coding elements on either side
of SOX9 associated with Pierre Robin sequence
Sabina Benko1,14, Judy A Fantes2,14, Jeanne Amiel1,3, Dirk-Jan Kleinjan2, Sophie Thomas1, Jacqueline Ramsay2,
Negar Jamshidi4, Abdelkader Essafi2, Simon Heaney2, Christopher T Gordon4, David McBride2, Christelle Golzio1,
Malcolm Fisher2, Paul Perry2, Ve´ronique Abadie5,6, Carmen Ayuso7, Muriel Holder-Espinasse8, Nicky Kilpatrick4,
Melissa M Lees9, Arnaud Picard10,11, I Karen Temple12, Paul Thomas4, Marie-Paule Vazquez10,11,
Michel Vekemans1,3,5, Hugues Roest Crollius13, Nicholas D Hastie2, Arnold Munnich1,3,5, Heather C Etchevers1,
Anna Pelet1, Peter G Farlie4, David R FitzPatrick2,14 & Stanislas Lyonnet1,3,5,14
Pierre Robin sequence (PRS) is an important subgroup of cleft
palate. We report several lines of evidence for the existence
of a 17q24 locus underlying PRS, including linkage analysis
results, a clustering of translocation breakpoints 1.06–1.23 Mb
upstream of SOX9, and microdeletions both B1.5 Mb
centromeric and B1.5 Mb telomeric of SOX9. We have also
identified a heterozygous point mutation in an evolutionarily
conserved region of DNA with in vitro and in vivo features of
a developmental enhancer. This enhancer is centromeric to the
breakpoint cluster and maps within one of the microdeletion
regions. The mutation abrogates the in vitro enhancer function
and alters binding of the transcription factor MSX1 as
compared to the wild-type sequence. In the developing mouse
mandible, the 3-Mb region bounded by the microdeletions
shows a regionally specific chromatin decompaction in cells
expressing Sox9. Some cases of PRS may thus result from
developmental misexpression of SOX9 due to disruption of
very-long-range cis-regulatory elements.
Pierre Robin sequence (PRS, OMIM 261800)1, defined by
micrognathia, glossoptosis and a posterior U-shaped cleft palate,
is a complex anomaly resulting in life-threatening feeding
and breathing difficulties in 1/2,000–1/10,000 of neonates. PRS
represents an embryological sequence in which the primary
abnormality is in mandibular growth, with a retropositioned
tongue resulting in a physical obstruction of palatal shelf elevation
and/or fusion. The core triad of features suggests that PRS may be
considered a cranial neurocristopathy2.
We mapped an autosomal dominant and highly penetrant PRS
locus to chromosome 17q24.3–25.1 (the PRS1 locus) by genetic
linkage analysis in 12 affected individuals from a four-generation
PRS-affected family (F1) (Fig. 1a). The maximum lod score of 3.32
(recombination fraction y ¼ 0) for the PRS1 locus was obtained in
between polymorphic DNA markers D17S795 and D17S929. This
genetic interval of 5.4 Mb encompasses a gene desert of 2.46 Mb
(Fig. 1b). SOX9, KCNJ2, KCNJ16 and MAP2K6 were selected as
candidate genes on the basis of expression pattern, their involvement
in molecular pathways relevant to mandibular development, or the
phenotypes of available knockout mice. No gross genomic alterations
or coding-sequence mutations could be detected by direct sequencing
of those genes in six individuals with PRS from family F1.
Concomitantly, we identified three independent families with
autosomal dominant, isolated PRS segregating with different recipro-
cal translocations, each of them sharing one breakpoint at 17q24
(Fig. 1a). Family T1 is a father and daughter with PRS who carry
the translocation t(2;17)(q32;q24), which occurred de novo in the
father. The 2q and 17q breakpoints lie within the BAC clones CTD-
2053I13 and RP11-1003J3, respectively. The 17q breakpoint was
narrowed to 113–149 kb from the centromeric end of RP11-1003J3
(Supplementary Fig. 1a online). In family T2, the breakpoint-span-
ning clones were RP11-420O5 on 17q and RP11-496M2 on 5q
(ref. 3; Supplementary Fig. 1b). In family T3, PRS segregates with
t(2;17)(q24.1;q24.3)4. RP11-510H11 on 2q and RP11-1003J3 (as
above) on 17q spanned the breakpoints. Southern blotting and inverse
PCR localized the breakpoint to Chr17:66,400,448 (NCBI Build 36.1;
Supplementary Fig. 2 online). The 2q32, 5q15 and 2q24.1 breakpoints
Received 19 May 2008; accepted 12 January 2009; published online 22 February 2009; doi:10.1038/ng.329
1INSERM U-781, Hoˆpital Necker–Enfants Malades, Paris, France. 2Medical Research Council Human Genetics Unit (MRC HGU), Institute of Genetic and Molecular
Medicine, Edinburgh EH4 2XU, UK. 3Assistance Publique–Hoˆpitaux de Paris (AP-HP), De´partement de Ge´ne´tique, Hoˆpital Necker–Enfants Malades, Paris, France.
4Murdoch Children’s Research Institute, Royal Children’s Hospital, Parkville, Australia. 5Universite´ Paris Descartes, Faculte´ de Me´decine, Paris, France. 6AP-HP,
Service de Pe´diatrie, Hoˆpital Necker–Enfants Malades, Paris, France. 7Fundacio´n Jime´nez Dı´az, Gene´tica, Ciberer Madrid, Spain. 8CHRU de Lille, Hoˆpital Jeanne de
Flandre, Lille, France. 9North Thames Regional Genetics Service, Great Ormond Street Hospital, London, UK. 10AP-HP, Service de Chirurgie Maxillo-Faciale et
Chirurgie Plastique, Hoˆpital d’Enfants Armand Trousseau, Paris, France. 11Universite´ Pierre et Marie Curie–Paris 6, UFR de Me´decine Pierre et Marie Curie, Paris.
12Wessex Clinical Genetics Academic Group, Division of Human Genetics, University of Southampton, Southampton, UK. 13Department of Biology, E´cole Normale
Supe´rieure, CNRS UMR-8541, Paris, France. 14These authors contributed equally to this work. Correspondence should be addressed to S.L.
(stanislas.lyonnet@inserm.fr) or D.R.F. (david.fitzpatrick@hgu.mrc.ac.uk).
NATURE GENETICS VOLUME 41 [ NUMBER 3 [ MARCH 2009 35 9
L E T T E R S


© 2009 Nature America, Inc. All rights reserved.

in families T1, T2 and T3 did not disrupt any known genes. The 17q24
breakpoints cluster within 160 kb in the gene desert between the genes
KCNJ2 (Chr17:65,677,271–65,687,778) and SOX9 (Chr17:67,628,756–
67,634,155; Fig. 1b). Another PRS-associated translocation in the
same region has been reported5. Comparative genomics analysis
indicated that this region contains over 200 highly conserved non-
coding elements (HCNEs; Fig. 1c) with 475% identity over 350 bp
across humans, rhesus, dog and mouse6–8. The notable clustering of
these breakpoints led us to hypothesize that one or more HCNEs
centromeric to the T3 breakpoint had a critical regulatory function in
mandibular development. We constructed multiple stable reporter
transgenic lines of mice to test the potential enhancer function of an
HCNE immediately centromeric to the T3 breakpoint (9CE4Z,
Chr17:66373309–66376106; Fig. 2a–c). Embryos from these lines
showed reporter expression in the proximal mandibular mesenchyme,
which is compatible with the pathogenesis of PRS (Fig. 2a,b).
As a first step to determine whether specific regulatory mutations at
this chromosomal locus cause PRS, we designed a high-density-tiling-
path comparative genomic hybridization (CGH) array extending
1.94 Mb centromeric (5¢) and 1.76 Mb telomeric (3¢) to SOX9
(Chr17:65,689,756–69,390,437; Fig. 1b), as breakpoints down-
stream of SOX9 have been associated with campomelic dysplasia
featuring PRS9. We observed heterozygous deletions in 3 of 12
unrelated cases of PRS (Table 1, Fig. 1b): a centromeric 75-kb deletion
(Chr17:66,175,000–66,250,000; Fig. 1c, Supplementary Fig. 3a online)
segregating with PRS in family F1, and de novo deletions in two
individuals with sporadic PRS, respectively a centromeric deletion of
4319 kb (Chr17:65,730,750–66,049,600; Supplementary Fig. 3b) in
individual Sp4 and a de novo telomeric 36-kb deletion
(Chr17:69,153,000–69,189,000; Supplementary Fig. 3c) in individual
Sp2. Each of the three alterations comprised at least 1 and up to 427
HCNEs (Table 1).
Subsequent DNA sequencing analysis, in the remaining nine
individuals with PRS who did not carry deletions, of the ten
HCNEs located in the F1 deletion identified a heterozygous T-to-C
transition in PRS-affected family F2 (Fig. 1a)—absent in 440 control
chromosomes—within an HCNE that shows 94% identity between
human and mouse over a distance of 220 bp (HCNE-F2;
Chr17:66,187,898; Table 1, Fig. 1c). To determine whether HCNE-
F2 has tissue-specific enhancer properties, we performed in vivo
reporter assays. A transient transgenic assay using the wild-type
HCNE-F2 showed strong activity in the craniofacial region in
mouse embryos at 11.5 days post coitum (d.p.c.) (Fig. 2c,d). In
addition, transcription-factor binding-site predictions identified
MSX1, EN1 and ZNF628 (ZEC) as potential transcription factors
for which binding to the HCNE-F2 would be altered by the F2
mutation. Although ZNF628 and EN1 were excluded on the basis of
lack of expression within the human first branchial arch, MSX1
(OMIM 142983) is a transcriptional regulator expressed in the
human first pharyngeal arch10 (Supplementary Fig. 4q online) that
is required for the development of craniofacial skeletal elements,
including the palate and mandible, in mice11 and has been found to
be mutated in some families with orofacial clefting. Electrophoretic
mobility shift assay (EMSA) analysis, using in vitro–transcribed and –
translated MSX1, showed significantly greater MSX1 binding (by 33%,
P ¼ 0.031) to the mutant target sequence than to the wild-type
consensus sequence (Fig. 3a). A control EMSA experiment was
performed with the nearby POU2F1 (Supplementary Fig. 4q) binding
site, in which we observed no affinity difference in POU2F1 binding to
either the wild-type or the mutant probes (data not shown). Chro-
matin immunoprecipitation (ChIP) from mouse cell lines derived
from 11.5-d.p.c. embryonic mandibular and maxillary mesenchyme
(mdMEPA and mxMEPA) confirmed that endogenous Msx1 binds
wild-type HCNE-F2 (Fig. 3b). In addition, we performed in vitro
assays using three cell types (HEK293, SKNBE(2c) and mdMEPA)
with wild-type or mutant HCNE-F2 or minimal promoter reporter
constructs12. As compared to the minimal promoter, the wild-type
HCNE-F2 had an enhancer activity only in mdMEPA that was
abolished when the element was mutated (P ¼ 0.017, Fig. 3c).
These findings suggest that the family F2 mutation would abolish
enhancer activity of the HCNE-F2 in a tissue-specific manner.
Finally, further ChIP analysis of the HCNE-F2 using the mdMEPA
cell line showed strong binding of p300, CTCF and K4-methylated
histone H3 (K4Me) to this region (Fig. 4a), a pattern that suggests a
role in both chromatin remodeling and transcriptional activation and
is consistent with the hypothesized active mandibular mesenchymal
+/tc
t(2;17)(q32;q24) t(5;17)(q15;q24) t(2;17)(q24.1;q24.3)T1
F1 F2
T2 T3
+/tc
del/+
del/+
65,600,000 66,000,000 66,500,000 67,000,000 67,500,000 68,000,000 68,500,000 69,000,000
100,000 bp
← SDK2SOX9 →KCNJ2 →
KCNJ16 → ← SLC39A11 Sp2
F2
Sp4 F1 RP11-1003J3
65
,82
0,0
00
65
,85
0,0
00
65
,88
0,0
00
Chr17
HCNE-F2
Del-F1
100%
50%
50%
50%
100%
100%
Mouse
Rhesus
Dog
0.740
0.000
–0.600
–1.099 104206_02000 bp
65
,91
0,0
00
65
,94
0,0
00
65
,97
0,0
00
66
,00
0,0
00
66
,03
0,0
00
66
,06
0,0
00
66
,09
0,0
00
66
,12
0,0
00
66
,15
0,0
00
66
,18
0,0
00
66
,21
0,0
00
66
,24
0,0
00
66
,27
0,0
00
66
,30
0,0
00
66
,33
0,0
00
66
,36
0,0
00
66
,39
0,0
00
66
,42
0,0
00
66
,45
0,0
00
T3 T2
T1
+/+
+/+
+/+ +/m
+/m
m/+ m/+
+/+
+/+
del/+
+/+
+/+ +/++/tc +/tc
+/tc +/tc +/+ +/tc +/tc +/tc +/tc+/+ +/+ +/+ +/+
+/tc +/tc +/tc
a
b
c
Figure 1 Family trees and highly conserved noncoding element (HCNE)
rearrangements. (a) Segregation of full (black symbols; posterior cleft palate,
micrognathia, glossoptosis) and partial (gray symbols; micrognathia,
glossoptosis) PRS phenotypes with the mutant chromosomes in families T1–
T3 (tc: translocation t(N;17)(N;q24)), family F1 (del: deleted chromosome
17q24.3) and family F2 (m: mutated HCNE). All affected individuals are
heterozygous for the mutant allele. (b) Genomic organization of the SOX9
locus and its 5¢ and 3¢ flanking regions. Black arrows, T1, T2 and T3
translocation breakpoints; green box, human BAC clone RP11-1003J3
(within which translocation breakpoints T1 and T3 were located);
red boxes, deletions found in family F1 and individuals Sp2 and
Sp4 (as detected by CGH and confirmed by semiquantitative PCR;
Supplementary Figs. 4–6 online); red arrow, point mutation detected by
sequence analysis (not shown) in family F2. (c) CGH profile of the F1
deletion, presented with a conservation plot (ECR browser; conservation
throughout human, rhesus monkey, dog and mouse of fragments 4350 bp
at 75% identity indicated in red). Arrow indicates the HCNE harboring the
point mutation in family F2 (HCNE-F2).
3 60 VOLUME 41 [ NUMBER 3 [ MARCH 2009 NATURE GENETICS
L E T T E R S


© 2009 Nature America, Inc. All rights reserved.

enhancer function. The SP2 region also showed strong binding of
p300, CTCF and K4Me in mdMEPA cells but also showed weaker
binding to acetylated histone H3 (H3Ac) and K4-trimethylated
histone H3 that was not seen in HNCE-F2 (data not shown).
Such enhancer activity is often associated with changes in the
degree of chromatin condensation of the locus13. We reasoned that
an in vivo assay of chromatin alterations during development of the
mandible would help determine which gene(s) is being transcription-
ally regulated by the HCNEs in question. The closest flanking genes
are KCNJ2 and SOX9. We measured the three-dimensional (3D)
distance between pairs of BAC probes from the orthologous genomic
region in mouse embryos using interphase fluorescence in situ
hybridization (FISH; Table 2 and Supplementary Fig. 5 online).
These BACs contained Kcnj2 (RP23-408D5 Chr11:110,809,076–
111,002,979), the breakpoint–HCNE-F2 region (BP-F2; RP23-76P19
Chr11:111,402,479–111,629,783) and the telomeric de novo deletion
detected in individual Sp2 (delSp2; RP23-418P13 Chr11:113,809,869–
114,025,688) (Fig. 4b). Mouse embryos were examined at 13.5 d.p.c.,
just before palate fusion, a stage critical for PRS. The interphase
distance between Kcnj2 and the BP-F2 probe remained unchanged in
all tissues tested (Table 2). However, the distance between the BP-F2
probe and the delSp2 probe was significantly greater in Sox9-expres-
sing mandibular arch cells than in the periocular mesenchyme, which
does not express Sox9 (Table 2); a similar tissue distinction was seen
for the distance between BP-F2 and Sox9, although this was not
statistically significant. This chromatin decompaction implicates Sox9
as the best candidate target of the enhancer activity associated with the
PRS1 locus. Indeed, the pattern of enhancer activity in the transgenic
reporter mice is consistent with the pattern of endogenous SOX9
expression in human embryos (Supplementary Fig. 4a–p), which,
among other locations, is found in the nervous system and skeletal
structures, including Meckel’s and Reichert’s cartilages.
Collectively these data support the deregulation of tissue-specific
SOX9 expression following mutation or disruption of regulatory
HCNEs as a highly plausible pathogenic mechanism at the PRS1
locus. HCNEs located in the gene desert surrounding SOX9, such as
the Sox9Cre1 or E1–E7 regulatory elements14–16, have already been
reported as putative regulators of Sox9 tissue-specific expression.
Heterozygous loss-of-function coding-sequence mutations of the
SOX9 gene cause campomelic dysplasia (CD, MIM114290) a severe,
often lethal skeletal dysplasia associated with sex reversal. PRS is a
feature both of CD and of a milder variant of the condition (acampo-
melic campomelic dysplasia, ACD) caused by hypomorphic intragenic
mutations or ‘position effect’ translocation breakpoints9. Moreover,
Sox9 is an early, direct activator of Col2a1, Col11a1 and Col11a2 in
avian cranial neural crest (CNC) cells17 and mouse chondrocytes18,19.
Sox9CE4Z-096
9CE4Z
2,797 bp Hsp68
minimal promoter
Mx
Md
Mx
Md
Orthologous mouse region
100,000 bp
Orthologous mouse region
T3 T2
T1
F1
HCNE-F2
200 bp
HCNE-F2 HCNE-F2
9CE4ZKCNJ2 → SOX9 →
KCNJ16 →
LacZβ-globin
minimal promotor
Sp4 F2
Oto
Mx
Md
Oto
High
Medium
Low
LacZ
a
b
c
d
Table 1 SOX9 locus alterations in isolated Pierre Robin sequence (PRS)
Position relative to SOX9 Conserved element
Molecular event Genomic position 5¢ 3¢ HCNEs Human/mouse, 4200 bp, 490% Heredity
F1 75-kb deletion Chr17:66,175,000–66,250,000
1.38 Mb 10 5 +
F2 T4C mutation Chr17:66,187,898
1.44 Mb +
Sp4 4319-kb deletion Chr17:65,730,750–66,049,600
1.58 Mb Z27 Z5 De novo
T1 Translocation Chr17:66,431,112–66,467,874
1.16 Mb 75 29 +
T2 Translocation Chr17:66,518,875–66,602,885
1.03 Mb 98 39 +
T3 Translocation Chr17:66,400,448
1.23 Mb 66 25 +
Sp2 36-kb deletion Chr17:69,153,000–69,189,000 +1.56 Mb 1 1 De novo
Nucleotide positions at the SOX9 locus are numbered according to Human NCBI Build 36.1. Position of the genomic alteration reported is given relative to SOX9 start codon.
The number of highly conserved noncoding elements (HCNEs) involved in SOX9 genomic alterations in PRS are indicated, including the elements showing a higher conservation
(4200 bp in length, with a 90% identity between Homo sapiens and Mus musculus). Bold, regions included in deletions or mutated; italics, regions located in between 17q
translocation breakpoints and the 3¢ end of the copy number–polymorphic BAC RP11-300G13 (Chr17:65,725,854).
Figure 2 HCNEs at the PRS1 locus have tissue-specific enhancer activity.
(a) The 9CE4Z element, located just upstream of the translocation break-
point cluster, was cloned 5¢ to an Hsp68 minimal promoter–LacZ reporter
gene cassette and used to make transgenic lines of mice. At 9.5 d.p.c.,
these lines showed consistent LacZ expression in a subset of cells in the
mandibular mesenchyme and some extracraniofacial expression. (b) Left, 3D
digital reconstructions from an OPT scan of the embryo shown in a. Middle,
digital dissection of the craniofacial region viewed from below. Right, digital
section through the plane indicated in the whole-embryo image in a with
expression in the proximal mandible (Md) and otocyst (Oto). (c) Cartoon of
the genomic region, showing the location of 9CE4Z and HNCE-F2 (F2).
(d) Left, LacZ expression at 11.5 d.p.c. in a whole embryo containing three
copies of HCNE-F2 cloned 5¢ into a minimal promoter–LacZ reporter gene
cassette, showing a pattern consistent with a subset of the normal mouse
Sox9 and human SOX9 expression pattern (compare with Supplementary
Fig. 3). Right, digital section from an optical projection tomography (OPT)
scan of the same embryo using the section plane illustrated. This shows
expression in the proximal mandible (Md), the maxilla (Mx) and the
otocyst (Oto).
NATURE GENETICS VOLUME 41 [ NUMBER 3 [ MARCH 2009 36 1
L E T T E R S


© 2009 Nature America, Inc. All rights reserved.

Dominant negative mutations in these collagen genes cause Stickler
syndrome, the most common syndromic diagnosis associated with
PRS20. It is possible that a tissue-specific loss of SOX9 enhancer activity
could lead to a coordinated reduction in the transcription of all three
targets during development, cumulatively phenocopying mutations in
the individual genes. Furthermore, conditional inactivation of mouse
Sox9 in CNC cells results in a phenotype of reduced jaw and cleft palate
strongly reminiscent of PRS21.
Still, the expression of other genes in the vicinity may well be
affected by deletions or mutations at the PRS1 locus. The most
obvious alternative candidate is KCNJ2, dominant negative mutations
in which cause Andersen cardiodysrhythmic periodic paralysis syn-
drome22 (OMIM 170390). This condition is occasionally associated
with micrognathia and cleft palate, although not in the context of
PRS. The absence of any detectable alteration in chromatin compac-
tion within the Kcnj2–BP-F2 interval in the developing mouse
mandible would argue against KCNJ2 being a target of the enhancers
reported here. Conversely, we demonstrated chromatin decompaction
in areas normally expressing Sox9, and campomelic and acampomelic
dysplasias are due to SOX9 haploinsufficiency that frequently culmi-
nates in PRS features23. Collectively, these data support SOX9 as a
more likely enhancer target than KCNJ2.
The finding of very distant mutations on both sides of a gene
promoter suggests that cis-regulatory domains of developmental
genes may extend over megabases of DNA flanking its coding
sequences. Long-range regulatory mutations have been identified in
several diseases in humans and animals24. These can be broadly
classed into two groups: those that phenocopy intragenic null
mutations and those that result in a phenotype distinct from that
associated with loss or gain of function. In the first category, the best-
studied disease/gene combinations are aniridia/PAX6 (ref. 24) and
campomelic dysplasia/SOX9 (refs. 9,16). These mutations cause failure
of normal transcriptional activation of the affected allele. The
most clearly defined example of the second category is preaxial
mirror polydactyly/SHH25, in which the disease is due to ectopic
transcriptional activation. We propose that PRS/SOX9 mutations
represent a site- and stage-specific loss of transcription most closely
analogous to a mouse conditional knockout—in this case, of SOX9 in
human CNC cells21,26,27.
In conclusion, disruption of noncoding DNA sequences with site-
and stage-specific enhancer function surrounding master develop-
mental genes14,28 may be regarded as a general mutational mecha-
nism for some human congenital malformations. It is likely that
MSX1
+ : 3′
m : 3′
MSX1
5′
Biotin-DNA
Unlabeled DNA
+ m
+
+
–
+
+
+
+
+
++
+
+
+++
+
+
–
+
+
+
+
+
++
+
+
+++
Pro
mo
ter
wt
.HC
NE
-F2
m
.HC
NE
-F2
0.05
0.04
0.03
0.02
0.01
0
301 bp
ConF2
mdMEPAmxMEPA
Msx1
IgG
ConF2
257 bp
301 bp
257 bp
Msx1 binding
Control
mdMEPA qPCR
HCNE-F2
1.20
1.00
0.80
0.60
0.40
0.20
0.00
a
c
b
-
-
-
5′-
*
HCNE-F2mdMEPA cells
18.00
16.00
14.00
12.00
10.00
Fo
ld
e
nr
ic
hm
en
t o
ve
r
Ig
G
8.00
6.00
4.00
2.00
0.00
CTCF H3Ac Input K27Me3
ChIP
K4Me K4Me3 p300 Pol II
Sox9 promoter
a
HCNE-F2
KCNJ2 →
KCNJ16 → Sp4
RP23-408D5
(MMU)
RP23-76P19
(MMU)
610 kb (MMU) 1,120 kb (MMU)
2,401 kb (MMU)**
RP23-229L12
(MMU)
RP23-418P13
(MMU)
F1
Sox9 promoter
Sp2SOX9 → ← SLC39A11
← SDK2
b
Figure 4 Chromatin characteristics of the region around SOX9. (a) ChIP
using mdMEPA cell line combined with quantitative real-time PCR analysis
shows that HCNE-F2 strongly binds K4 methylated histone H3 (K4Me),
CTCF and p300, suggesting that this region has a role in transcriptional
activation and chromatin remodeling consistent with its functioning as a
mandibular mesenchymal enhancer. The Sox9 promoter region shows a
different pattern consistent with a regulated promoter region, with strong
binding RNA polymerase II (Pol II), acetylated histone H3 (H3Ac) and
trimethylated K4 histone H3 (K4Me3) and weaker binding to p300.
K27Me3, trimethyllysine 27 form of histone H3. Error bars are equivalent
to 95% confidence intervals (see Supplementary Methods). (b) Genomic
organization of the SOX9 locus. Family F1 and individuals Sp2 and Sp4
(all carrying deletions) are indicated as red boxes. Blue arrow, HCNE-F2; red
arrow, SOX9 promoter region. Orange boxes below represent orthologous
positions of mouse BAC probes around Sox9 relative to the human SOX9
locus. MMU, M. musculus. Distances between BAC probes in the mouse
genome are indicated in kbp.
Figure 3 HCNE-F2 has enhancer activity. (a) Nucleotide sequences of
normal and mutant transcription-factor binding probes derived from HCNE-
F2 and used in EMSA, with the MSX1 binding site underlined. The mutant
base is enlarged and boxed in red. MSX1 was incubated with biotin-labeled
probes (normal (+) and mutated (m), respectively) with or without an
unlabeled probe for the competition reactions (normal unlabeled probe
for + and mutated unlabeled probe for m). (b) Above, ChIP results showing
that endogenous Msx1 binds HCNE-F2 (301-bp amplicon; F2) but not a
257-bp control amplicon (Chr11:111450487–111450743; Con), in
mandibular (md) and maxillary (mx) mouse embryonic pharyngeal arch
(MEPA) cells. Anti-rabbit IgG antibody was used as a negative control.
Below, quantitative real-time PCR analysis of the same ChIP experiment
revealed a B2.5-fold greater binding of Msx1 to HCNE-F2 compared to the
control region. Errors bars represent the calculated fold change error (see
Supplementary Methods). (c) Enhancer function of HCNE-F2 in mdMEPA
cells. The cells were transfected with reporter constructs containing only the
Sox9 minimal promoter (Promoter) or with either the wild-type HCNE-F2
(wt.HCNE-F2) or the HCNE-F2 harboring the F2 mutation (m.HCNE-F2)
inserted 5¢ to the promoter. The wild-type HCNE-F2 showed enhancer
function, which was abolished when the element harbored the mutation
found in the PRS-affected F2 family (P ¼ 0.001).
3 62 VOLUME 41 [ NUMBER 3 [ MARCH 2009 NATURE GENETICS
L E T T E R S


© 2009 Nature America, Inc. All rights reserved.

high-resolution genome-wide analyses will identify mutations with
cis-regulatory effects for increasing numbers of diseases. Our study
suggests that these may have a much wider range of action than that
traditionally understood.
METHODS
Sample selection. Only individuals with isolated PRS were selected for study
(see Supplementary Note online for details of inclusion and exclusion criteria).
Controls were healthy, unrelated middle-aged individuals of European ancestry
with no known craniofacial defects. Informed consent was obtained from all
study participants. Studies were approved by the ethical institutional commit-
tees of the Comite´ Consultatif de Protection des Personnes dans la Recherche
Biome´dicale, the Hoˆpital Necker–Enfants Malades and the Royal Children’s
Hospital and by the UK national Multicentre Research Ethics Committee.
Genetic linkage. Genome-wide microsatellite genotyping was performed in the
PRS-affected family F1 by deCODE Genetics at an average marker density of
4 cM (1,000 markers). We computed two-point LOD score values using the
linkage package (MLINK program). PRS was assumed to behave as an auto-
somal dominant trait with incomplete penetrance (90%) and no gender bias.
Nucleotide variation screening. PCR products were directly sequenced on
both strands on an ABI PRISM 3130XL DNA sequencer (Perkin Elmer–Applied
Biosystems) using the Big Dye Terminator method according to the
manufacturer’s instructions. All primer sequences are listed in Supplementary
Table 1 online.
Metaphase FISH and breakpoint cloning. Metaphase chromosome analysis
using two-color FISH was performed as previously described13 (Supplemen-
tary Methods online). For Southern blots, PCR-amplified probes were
sequenced and verified probes radiolabeled with [a32P]dCTP using Rediprime
II Random Prime Labeling System (Amersham Biosciences) (Supplementary
Methods and Supplementary Table 1).
Production and analysis of reporter transgenic mice. To make the 9CE4Z
reporter construct, a 2.8-kb fragment was PCR amplified from human
BAC RP11-1003J3 using primers containing NotI and EagI restriction sites
(Supplementary Table 1). This fragment was subcloned into the NotI site of
the p610+ reporter construct containing an Hsp68 minimal promoter–LacZ
cassette. To generate the HCNE-F2(3xwt)bZ reporter construct, the wild-type
HCNE-F2 element was PCR amplified with three sets of attB site–containing
primers (Supplementary Table 1) and cloned into the three-way Gateway
(Invitrogen) entry vector, which was then recombined with a vector bearing a
LacZ reporter cassette containing the human b-globin minimal promoter. To
create transgenic animals, linearized 9CE4Z and HCNE-F2(3xwt)bZ constructs
were microinjected using standard procedures (Supplementary Methods).
Embryos were collected at relevant stages, fixed and analyzed for reporter
activity by Xgal staining (Supplementary Methods).
Optical projection tomography. Optical projection tomography (OPT) was
performed as previously described29 (Supplementary Methods).
Table 2 Localized tissue-specific chromatin decompaction at the Sox9 locus
BAC 408D5-76P19 76P19-229L12 76P19-418P13
Region Kcnj2–BP-F2 BP-F2–Sox9 BP-F2–delSP2
Genomic distance (kb) 610 1,120 2,401
Stage Tissue
Sox9
expression Nuclei
Mean 3D
separation
(nm)
95% CI,
± Nuclei
Mean 3D
separation
(nm)
95% CI,
± Nuclei
Mean 3D
separation
(nm)
95% CI,
±
9.5 d.p.c. Pharyngeal arch ++ 107 419 31 130 344 22 – – –
13.5 d.p.c. Meckel’s cartilage ++ 106 418 29 93 332 23 123 705** 59
13.5 d.p.c. Palatal shelf/maxillary mesenchyme + 51 425 43 70 322 31 118 590 42
13.5 d.p.c. Periocular mesenchyme
91 417 30 99 289 24 133 585 49
Measurements of the 3D distance between interphase FISH signals in sections of embryonic mouse tissues. The mouse BAC probes positions spanning the orthologous regions
around mouse Sox9 relative to human SOX9 locus are shown in Figure 1b. Optical sectioning of individual nuclei in sections of craniofacial tissue was used to determine the
distance between probe signals from interphase 3D FISH. ‘**’ indicates a significant transcription-dependent chromatin decondensation; the lower 95% CI for the BP-F2–delSP2
distance in the high-expressing region does not overlap with the upper 95% CI of either the medium- or low-expressing regions.
Fine-tiling CGH-array analysis. For the fine-tiling CGH (NimbleGen), a tiling
array was designed containing 385,000 probes of 50–75-mer length with a
median probe spacing of 5 bp covering the region Chr17:65,689,756–
69,390,437. Identified microrearrangements were tested by semiquantitative
PCR (Supplementary Table 1 and Supplementary Methods).
SAGE corroboration. Publicly available COGENE SAGE data10 were mined to
examine expression of SOX9, EN1, ZNF628, MSX1 and POU2F1; the ACTG
web tool was used for tag to gene mapping of the two SAGE libraries
(Supplementary Methods).
Electrophoretic mobility shift assays (EMSA). MSX1 and POU2F1 expression
vectors were constructed by insertion of the human MSX1 or POU2F1 cDNA
(Geneservice) into pcDNA3.0/zeo+ (Invitrogen) and used to synthesize MSX1
and POU2F1 protein following the TNT Coupled Reticulocyte Lysate Systems
protocol (Promega). EMSA was performed following the LightShift Chemi-
luminescent EMSA Kit (Pierce) protocol. The results were quantified using
ImageJ software. The applied statistical test was the nonparametric Wilcoxon
test (n ¼ 15, a ¼ 0.05) (Supplementary Methods).
Cell culture. Mandibular and maxillary processes were excised from 11.5 d.p.c.
embryos from CD1 females crossed with a male ‘Immortomouse’. The
mandibular and maxillary tissue was dissociated and plated in medium (1
DMEM/10% FCS/1% penicillin/streptomycin) containing 100 U ml–1 murine
g-interferon (Peprotech). Cells were cultured at 33 1C in an atmosphere
containing 5% CO2 (Supplementary Methods). The cell lines were designated
MEPA (mouse embryonic pharyngeal arch) with the prefix of mx (maxillary)
and md (mandibular). Adherent HEK293 and SKNBE(2c) cells were cultured
in 1 DMEM/10% FCS (FCS)/1% penicillin/streptomycin at 37 1C in an
atmosphere containing 5% CO2.
Chromatin immunoprecipitation (ChIP). The ChIP protocol was carried out
as previously described30. For the immunoprecipitation of chromatin-bound
proteins, the following antibodies were used: PolII, H3K4me3, H3K4me and
p300 from Abcam; CTCF and H3ac from Upstate. The MSX1 ChIP antibody
was from Santa Cruz Biotechnology (Supplementary Methods).
In vitro enhancer activity assays. To generate the pr.Sox9 reporter construct, the
Sox9 minimal promoter15 was amplified with PCR primers containing
BglII and HindIII restriction sites. This fragment was subcloned into the BglII/
HindIII restriction site upstream of luc+ of the pGL3-basic vector (Promega).
The HCNE-F2 wild-type and mutated elements were amplified using PCR
primers containing MluI and XhoI restriction sites, and were subcloned into the
MluI/XhoI restriction site of the pr.Sox9 reporter construct. This generated the
wtHCNE-F2-pr.Sox9 and m.HCNE-F2-pr.Sox9 reporter constructs, respectively.
Transfection experiments were performed on adherent HEK293 and
SKNBE(2c) cells cultured in the conditions described above, and in mdMEPA
cells cultures at 33 1C in an atmosphere containing 7% CO2. For the
transfection experiment, cells were grown to 80% confluency in 12-well plates.
Cells were transfected with 600 ng of plasmid DNA, 30 ng of pRL-CMV
(Promega) and 3 ml of FuGeneHD (Roche) according to the FuGeneHD
NATURE GENETICS VOLUME 41 [ NUMBER 3 [ MARCH 2009 36 3
L E T T E R S


© 2009 Nature America, Inc. All rights reserved.

transfection protocol. Cells were harvested and lysed 24–48 h after transfection.
Firefly and Renilla luciferase activities were measured (Dual-Luciferase Repor-
ter Assay System, Promega). The firefly luciferase activity of each construct was
normalized to the Renilla luciferase internal control, pRL-CMV. The applied
statistical test was the nonparametric Mann-Whitney test (n ¼ 18, a ¼ 0.05).
95% confidence intervals (CI 95%) given as the error bars in Figure 3c were
calculated using the following formula: CI 95% ¼ mean ± t0,25  s.e.m., where
t ¼ Student t-test value.
Sox9 immunohistochemistry and interphase FISH. Rabbit polyclonal anti-
Sox9 (Chemicon) was used as primary antibody and Alexa Fluor 488–
conjugated goat anti-rabbit F(Ab)2 fragments (Invitrogen) as secondary anti-
body. The slides were counterstained with 4¢,6-diamidino-2-phenylindole
(DAPI) upon mounting (Supplementary Methods).
Interphase FISH used mouse BAC probes obtained from BACPAC resources
of the Children’s Hospital Oakland Research Institute. BAC clones were labeled
with digoxygenin or with the direct fluorescent label Spectrum Orange-dUTP
(Vysis) (Supplementary Methods).
Image capture and analysis were performed using in-house scripts written for
IPLab Spectrum (Scanalytics). Inter-spot distances were calculated using the
custom software IP lab script (available from P. Perry, MRC HGU). 3D distance
means, s.d. and confidence intervals were calculated (Supplementary Methods).
In situ expression pattern study. SOX9 in situ expression was examined in
normal human embryos (CS13, CS15 and CS18) obtained from electively
terminated pregnancies in concordance with French legislation (94-654 and
08-400) and with oversight by a local ethics committee. Tissue fixation,
sectioning and in situ hybridization were carried out according to standard
protocols (Supplementary Methods and Supplementary Table 1).
Accession codes. RefSeq: SOX9, NM_000346; Sox9, NM_011448; KCNJ2,
NM_000891; KCNJ16, NM_170741, NM_018658, NM_170742; MAP2K6,
NM_002758; human chromosome 2, NC_000002.10; human chromosome 5,
NC_000005.8, human chromosome 17, NC_000017.9; rhesus chromosome 16,
NC_007873; mouse chromosome 11, NC_000077; dog chromosome 9,
NC_006591. GenBank: RP11-1003J3, AC005181.1; CTD-2053I13, AC098485.6;
RP11-420O5, AC007642.5; RP11-496M2, AC009126.2; RP11-510H11,
AC092662.2; MSX1 cDNA clone, BC021285; POU2F1 cDNA clone, BC052274.
OMIM: SOX9, *608160; campomelic dysplasia, #114290; Pierre Robin sequence,
261800; KCNJ2, *600681; Andersen syndrome, #170390; MSX1, *142983.
URLs. rVista 2.0, http://rvista.dcode.org/; ECR browser, http://ecrbrowser.dco-
de.org/; eShadow, http://eshadow.dcode.org/; SNP, http://www.ncbi.nlm.nih.
gov/projects/SNP/; Database of Genomic Variants, http://projects.tcag.ca/varia
tion/; DECIPHER, http://decipher.sanger.ac.uk/; Human Genome Structural
Variation Project, http://humanparalogy.gs.washington.edu/structuralvaria
tion/; COGENE, http://hg.wustl.edu/COGENE/SAGE/index.html; ACTG,
http://retina.med.harvard.edu/ACTG.
Note: Supplementary information is available on the Nature Genetics website.
ACKNOWLEDGMENTS
We are grateful to the affected individuals and their families who participated in
this study, to the Associations Franc¸aises du Syndrome de Robin, to the Centres
de Re´fe´rences Anomalies Cranio-Faciales Rares (AP-HP, Necker and Trousseau
hospitals), and to C. Ozilou, G. Staub and G. Gue´du-Molina for assistance.
We thank T. Attie´-Bitach, G. Couly and L. Legeai-Mallet (Necker) and V. van
Heyningen and R. Hill (MRC HGU) for useful discussion. This study was
underwritten by grants from the Agence Nationale de la Recherche (ERARE grant
CraniRare), EUROCRAN FP5, the Fondation pour la Recherche Me´dicale (FRM),
the MRC (UK) and the National Health and Medical Research Council
(Australia). S.T. was supported in part by grant NS039818 from the US National
Institutes of Health and S.B. by the FRM.
AUTHOR CONTRIBUTIONS
S.B., J.A.F. and A.P. performed molecular genetics studies. J.A.F., C.T.G. and
N.J. performed chromosomal studies. S.B. and J.R. performed the in vitro
enhancer activity experiments. S.B., S.T., C.G., M.V. and H.C.E. performed
human expression studies. J.R., S.B. and A.E. performed immunoprecipitation
experiments. D.-J.K. performed transgenic assays. J.A.F., S.H., P.P. and D.B.
performed the in vivo chromatin compaction studies. M.F. did the OPT image
analysis. S.B. and H.R.C. performed the comparative genomic analysis. J.A., V.A.,
C.A., M.H.-E., N.K., M.M.L., A.P., I.K.T., M.V., P.T., M.-P.V., D.R.F. and S.L.
recruited individuals and families affected with PRS. S.L., D.R.F., P.G.F. and J.A.
contributed to the concept, strategy, study design and project management. S.B.,
H.C.E., S.T., P.G.F., D.R.F. and S.L. contributed to the writing of the manuscript.
All authors discussed the results.
Published online at http://www.nature.com/naturegenetics/
Reprints and permissions information is available online at http://npg.nature.com/
reprintsandpermissions/
1. Robin, P. A fall of the base of the tongue considered as a new cause of nasopharyngeal
respiratory impairment: Pierre Robin sequence, a translation. 1923. Plast. Reconstr.
Surg. 93, 1301–1303 (1994).
2. Etchevers, H.C., Amiel, J. & Lyonnet, S. Molecular bases of human neurocristopathies.
Adv. Exp. Med. Biol. 589, 213–234 (2006).
3. Vintiner, G.M., Temple, I.K., Middleton-Price, H.R., Baraitser, M. & Malcolm, S.
Genetic and clinical heterogeneity of Stickler syndrome. Am. J. Med. Genet. 41,
44–48 (1991).
4. Jamshidi, N. et al. Isolated Robin sequence associated with a balanced t(2;17)
chromosomal translocation. J. Med. Genet. 41, e1 (2004).
5. Jakobsen, L.P. et al. Pierre Robin sequence may be caused by dysregulation of SOX9
and KCNJ2. J. Med. Genet. 44, 381–386 (2007).
6. Ahituv, N., Prabhakar, S., Poulin, F., Rubin, E.M. & Couronne, O. Mapping cis-
regulatory domains in the human genome using multi-species conservation of synteny.
Hum. Mol. Genet. 14, 3057–3063 (2005).
7. Dermitzakis, E.T., Reymond, A. & Antonarakis, S.E. Conserved non-genic sequences—
an unexpected feature of mammalian genomes. Nat. Rev. Genet. 6, 151–157 (2005).
8. Woolfe, A. et al. Highly conserved non-coding sequences are associated with vertebrate
development. PLoS Biol. 3, e7 (2005).
9. Velagaleti, G.V. et al. Position effects due to chromosome breakpoints that map
approximately 900 Kb upstream and approximately 1.3 Mb downstream of SOX9
in two patients with campomelic dysplasia. Am. J. Hum. Genet. 76, 652–662 (2005).
10. Cai, J. et al. Gene expression in pharyngeal arch 1 during human embryonic develop-
ment. Hum. Mol. Genet. 14, 903–912 (2005).
11. Satokata, I. & Maas, R. Msx1 deficient mice exhibit cleft palate and abnormalities of
craniofacial and tooth development. Nat. Genet. 6, 348–356 (1994).
12. Cirulli, E.T. & Goldstein, D.B. In vitro assays fail to predict in vivo effects of regulatory
polymorphisms. Hum. Mol. Genet. 16, 1931–1939 (2007).
13. Morey, C., Da Silva, N.R., Perry, P. & Bickmore, W.A. Nuclear reorganisation and
chromatin decondensation are conserved, but distinct, mechanisms linked to Hox gene
activation. Development 134, 909–919 (2007).
14. Bagheri-Fam, S., Ferraz, C., Demaille, J., Scherer, G. & Pfeifer, D. Comparative geno-
mics of the SOX9 region in human and Fugu rubripes: conservation of short regulatory
sequence elements within large intergenic regions. Genomics 78, 73–82 (2001).
15. Bagheri-Fam, S. et al. Long-range upstream and downstream enhancers control
distinct subsets of the complex spatiotemporal Sox9 expression pattern. Dev. Biol.
291, 382–397 (2006).
16. Bien-Willner, G.A., Stankiewicz, P. & Lupski, J.R. SOX9cre1, a cis-acting regulatory
element located 1.1 Mb upstream of SOX9, mediates its enhancement through the
SHH pathway. Hum. Mol. Genet. 16, 1143–1156 (2007).
17. Suzuki, T., Sakai, D., Osumi, N., Wada, H. & Wakamatsu, Y. Sox genes regulate type 2
collagen expression in avian neural crest cells. Dev. Growth Differ. 48, 477–486
(2006).
18. Akiyama, H. et al. Interactions between Sox9 and beta-catenin control chondrocyte
differentiation. Genes Dev. 18, 1072–1087 (2004).
19. Bi, W., Deng, J.M., Zhang, Z., Behringer, R.R. & de Crombrugghe, B. Sox9 is required
for cartilage formation. Nat. Genet. 22, 85–89 (1999).
20. Holder-Espinasse, M. et al. Pierre Robin sequence: a series of 117 consecutive cases.
J. Pediatr. 139, 588–590 (2001).
21. Mori-Akiyama, Y., Akiyama, H., Rowitch, D.H. & de Crombrugghe, B. Sox9 is required
for determination of the chondrogenic cell lineage in the cranial neural crest. Proc.
Natl. Acad. Sci. USA 100, 9360–9365 (2003).
22. Andelfinger, G. et al. KCNJ2 mutation results in Andersen syndrome with sex-specific
cardiac and skeletal muscle phenotypes. Am. J. Hum. Genet. 71, 663–668 (2002).
23. Wagner, T. et al. Autosomal sex reversal and campomelic dysplasia are caused by
mutations in and around the SRY-related gene SOX9. Cell 79, 1111–1120 (1994).
24. Kleinjan, D.A. & van Heyningen, V. Long-range control of gene expression: emerging
mechanisms and disruption in disease. Am. J. Hum. Genet. 76, 8–32 (2005).
25. Lettice, L.A. et al. A long-range Shh enhancer regulates expression in the developing
limb and fin and is associated with preaxial polydactyly. Hum. Mol. Genet. 12,
1725–1735 (2003).
26. Hong, C.S. & Saint-Jeannet, J.P. Sox proteins and neural crest development. Semin.
Cell Dev. Biol. 16, 694–703 (2005).
27. McCauley, D.W. & Bronner-Fraser, M. Importance of SoxE in neural crest development
and the evolution of the pharynx. Nature 441, 750–752 (2006).
28. Pheasant, M. & Mattick, J.S. Raising the estimate of functional human sequences.
Genome Res. 17, 1245–1253 (2007).
29. Sharpe, J. et al. Optical projection tomography as a tool for 3D microscopy and gene
expression studies. Science 296, 541–545 (2002).
30. Essafi, A. et al. Direct transcriptional regulation of Bim by FoxO3a mediates STI571-
induced apoptosis in Bcr-Abl-expressing cells. Oncogene 24, 2317–2329 (2005).
3 64 VOLUME 41 [ NUMBER 3 [ MARCH 2009 NATURE GENETICS
L E T T E R S


© 2009 Nature America, Inc. All rights reserved.