Coding exons function as tissue-specific enhancers
of nearby genes
Ramon Y. Birnbaum,1,2E. Josephine Clowney,3,4Orly Agamy,5Mee J. Kim,1,2
Jingjing Zhao,1,2,6Takayuki Yamanaka,1,2Zachary Pappalardo,1,2Shoa L. Clarke,7
Aaron M. Wenger,8Loan Nguyen,1,2Fiorella Gurrieri,9David B. Everman,10
Charles E. Schwartz,10,11Ohad S. Birk,5Gill Bejerano,8,12Stavros Lomvardas,3
and Nadav Ahituv1,2,13
1Department of Bioengineering and Therapeutic Sciences,2Institute for Human Genetics,3Department of Anatomy,4Program in
Biomedical Sciences, University of California, San Francisco, California 94143, USA;5The Morris Kahn Laboratory of Human Genetics,
NIBN, Ben-Gurion University, Beer-Sheva 84105, Israel;6Key Laboratory of Advanced Control and Optimization for Chemical
Processes of the Ministry of Education, East China University of Science and Technology, Shanghai 200237, China;7Department of
Genetics,8DepartmentofComputer Science,StanfordUniversity, Stanford,California94305-5329, USA;9IstitutodiGeneticaMedica,
Universita ` Cattolica S. Cuore, Rome 00168, Italy;10JC Self Research Institute, Greenwood Genetic Center, Greenwood, South Carolina
29646, USA;11DepartmentofGeneticsandBiochemistry, Clemson University,Clemson,South Carolina29634, USA;12Departmentof
Developmental Biology, Stanford University, Stanford, California 94305-5329, USA
Enhancers are essential gene regulatory elements whose alteration can lead to morphological differences between species,
developmental abnormalities, and human disease. Current strategies to identify enhancers focus primarily on noncoding se-
seq peaks overlap coding exons (after excluding for peaks that overlap with first exons). By using mouse and zebrafish enhancer
genes andthattheexonicsequenceis necessaryforenhanceractivity.Using ChIP,3C,andDNA FISH,wefurther show thatone
of these exonic limb enhancers, Dync1i1 exon 15, has active enhancer marks and physically interacts with Dlx5/ 6 promoter regions
900 kb away. In addition, its removal by chromosomal abnormalities in humans could cause split hand and foot malformation
1 (SHFM1),adisorderassociatedwith DLX5/ 6.These results demonstratethat DNA sequences can haveadual function,operating
mutations could be caused not only by protein alteration but also by disrupting the regulation of another gene.
[Supplemental material is available for this article.]
Precise temporal, spatial, and quantitative regulation of gene ex-
pression is essential for proper development. This tight transcrip-
tional regulation is mediated in part by DNA sequences called en-
hancers, which regulate gene promoters. By use of comparative
genomics or chromatin immunoprecipitation followed by next-
generation sequencing (ChIP-seq), candidate enhancer sequences
can now be identified in a relatively high-throughput manner
then be assayed for enhancer activity using various in vitro and in
vivo assays (Woolfe et al. 2005; Pennacchio et al. 2006; Heintzman
et al. 2009). However, the majority of these experiments remove
coding sequences from their analyses under the assumption that
they do not function as enhancers, due to their protein coding
Previous exonic enhancers (eExons) have been reported in
vertebrates (Neznanov et al. 1997; Lampe et al. 2008; Tumpel et al.
2008; Dong et al. 2010; Eichenlaub and Ettwiller 2011; Ritter et al.
2012). In addition, a recent study scanning for synonymous
constraint in protein coding regions (Lin et al. 2011) found an
overlap between two of these eExons (Lampe et al. 2008; Tumpel
et al. 2008) and synonymous constraint elements. Here, we ana-
lyzed 25 available ChIP-seq data sets of enhancer marks
(H3K4me1, H3K27ac, and EP300, also known as p300) for their
overlap with coding exons. Following this analysis, we wanted
to specifically determine whether eExons could regulate their
neighboring genes and not the gene they reside in. This was of
interest to us due to the phenotypic implications that coding
mutations could have on nearby genes. For this purpose, we ana-
lyzed a specific EP300 ChIP-seq data set from mouse embryonic
day (E) 11.5 limb tissue (Visel et al. 2009a), due to its ability to
identify active enhancers with high accuracy (88%) and tissue
specificity (80%) in vivo. At E11.5, mouse limb development
progresses along three axes: proximal-distal (P-D), anterior-poste-
limb bud create gradients and feedback loops that determine these
axes (Gilbert 2000; Nissim and Tabin 2004; Zeller et al. 2009), and
their alteration could lead to morphological differences. In this
study, we focused on identifying limb eExons involved in the de-
velopment along both the P-D and the A-P axes.
Article published online before print. Article, supplemental material, and pub-
lication date are at http://www.genome.org/cgi/doi/10.1101/gr.133546.111.
22:1059–1068 ? 2012, Published by Cold Spring Harbor Laboratory Press; ISSN 1088-9051/12; www.genome.org
The apical ectodermal ridge (AER) is the signaling center that
keeps the underlying mesenchyme in a proliferative state and al-
lows the limb to grow, thus governing the P-D axis. In the devel-
oping mouse limb bud, the distal-less homeobox 5 and 6 (Dlx5/6)
genes are expressed in the AER (Fig. 1D9,E9), and Dlx5 is also
expressed in the anterior mesenchyme (Fig. 1E9). Disruption of
both Dlx5 and Dlx6 in mice leads to a split hand and foot mal-
formation (SHFM) phenocopy (Robledo et al. 2002). In humans,
chromosomal aberrations in the DLX5/6 region, some of which do
not encompass the coding sequences of DLX5/6, cause SHFM1
(MIM 183600) and are associated with incomplete penetrance
with intellectual disability, craniofacial malformations and
deafness (Elliott and Evans 2006). Other than one family with
a DLX5 missense mutation (Shamseldin et al. 2012), no other cod-
ing mutation in either gene has been found in individuals with
SHFM1, suggesting that disruption of DLX5/6 gene regulatory ele-
ments could lead to a SHFM1 phenotype.
TheA-P axisis controlledbya signalingcentercalledthe zone
of polarizing activity, located in the posterior mesenchyme and
defined by the expression of Sonic Hedgehog (SHH). Twist1, is a
transcription factor that inhibits Shh expression in the anterior
limb bud by antagonizing HAND2, a Shh-positive regulator (Firulli
et al. 2005). Homozygous Twist1-null mice have limb bud devel-
opmental defects (Chen and Behringer 1995), and heterozygous
mice develop polydactyly (Bourgeois et al. 1998) and show ectopic
Shh expression (O’Rourke et al. 2002). In humans, mutations in
TWIST1 lead to various syndromes, the majority of which encom-
pass various forms of limb malformations (MIM*601622).
Here, by examining various enhancer-associated ChIP-seq data
sets, we characterized the general prevalence of peaks that overlap
exons. We then chose seven limb EP300 ChIP-seq exonic se-
quences and functionally tested them for enhancer activity using
sequences were shown to be functional limb enhancers in the
mouse. Further analysis of one of these enhancers, Dync1i1 eExon
15, using chromatin conformation capture (3C) and DNA fluo-
rescent in situ hybridization (FISH) showed that it physically in-
teracts with Dlx5/6 promoter regions in the developing mouse
limb. Mutation analysis of individuals with SHFM1 indicated that
chromosomal aberrations encompassing this enhancer could be
one of the causes of their SHFM1 phenotype. Combined, these
findings demonstrate that a DNA sequence can function both as a
coding exon in one tissue and as an enhancer in a different tissue
and suggest the need to be cautious when assigning a coding mu-
tation phenotype to protein function.
Exon overlap analyses of enhancer-associated ChIP-seq data sets
To determine the genome-wide prevalence of enhancer-associated
ChIP-seq peaks that overlap coding exons, we analyzed 25 available
ChIP-seq data sets of enhancer marks (H3K4me1, H3K27ac, and
EP300) from various human cell lines and mouse E11.5 tissues
(Supplemental Table 1; see Methods). Since these enhancer marks
could also identify potential promoters, we only looked for overlap
with coding exons after excluding for the first exon. In all the anal-
yses described in this study, coding exons are defined here as only
those exons that are not first exons. Analysis of the individual his-
peaks overlap coding exons, respectively (Supplemental Table 1). It is
worth noting that the average peak size in the H3K4me1 and
H3K27ac ChIP-seq data sets is 2441 and 3107 bp, and the average
size of peaks overlapping coding exons is 3476 and 4195 bp for
shows that it is ;280 bp (see Methods). Compared with the aver-
age peak size of the histone marks and those that overlap exons in
particular, it is quite possible that the functional entity of the peak
doesnot constitutetheexonandthat thepercentagesaboveare an
overestimate. Therefore, we analyzed six different EP300 ChIP-seq
data sets from various human cell lines
that have shorter peak sizes, the average
of which was 426 bp. In these data sets,
we found that on average, 4% of the
peaks overlapped with coding exons (Sup-
plemental Table 1). To get a better indi-
cation if these sequences could be func-
between exonic EP300 ChIP-seq peaks
and H3K4me1 and H3K27ac peaks. We
used ChIP-seq data from two different
cell lines, GM12878 and K562, where
all three enhancer marks were avail-
able. We found that 8% and 5% of the
ChIP-seq peaks that had all three en-
hancer marks overlapped coding exons
in GM12878 and K562 cells, respectively
(Supplemental Table 1). We next screened
coding sequences that had all three en-
hancer marks for their overlap with
a recently published study that scanned
the genome for synonymous constraint
in protein coding regions (Lin et al. 2011).
We found that 9%of coding exons withall
with synonymous constraint exons for
both GM12878 and K562 cell lines (Sup-
say. A schematic of the DYNC1I1-DLX5/6 (A) and HDAC9-TWIST1 (F) genomic regions. Black boxes and
orange ovals represent coding exons and positive eExons, respectively. The black arrows point to the
genes that are thought to be regulated by the eExons. (B–C9) Mouse enhancer assays of DYNC1I1 eExon
bud mesenchyme enhancer activity (red arrow). (D,E) Mouse whole-mount in situ hybridization of Dlx6
and Dlx5. (D9,E9) Dlx6 and Dlx5 limb expression pattern is similar to DYNC1I1 eExon 15 enhancer ac-
tivity. In addition, Dlx5 is also expressed in anterior limb bud as depicted by the red arrow (E9), similar to
DYNC1I1 eExon 17 enhancer activity (C9). (G–H9) Mouse enhancer assays of HDAC9 eExons 18 and 19.
(G,G9) HDAC9 eExon 18 shows anterior limb bud enhancer activity (red arrows), and (H,H9) HDAC9
eExon 19 shows posterior limb bud (red arrows) and branchial arch enhancer activity in E11.5 mice. (I)
Mousewhole-mount in situhybridization of Twist1 at E11.5. (I9)Twist1 limb expression patternis similar
to the HDAC9 eExon 18 anterior limb bud enhancer activity (G9)markedby red arrow and HDAC9 eExon
19 posterior limb bud enhancer activity (H9). For B, C, G, and H, the numbers in the bottom right corner
indicate the number of embryos showing this limb expression pattern/total LacZ stained embryos.
eExons within DYNC1I1 and HDAC9 characterized using a mouse transgenic enhancer as-
1060 Genome Research
Birnbaum et al.
peaks in the 25 enhancer-associated ChIP-seq data sets that we ana-
lyzed overlapped coding exons after removing first exons.
Analysis of a EP300 limb ChIP-seq data set and eExon
Given that several eExons were previously discovered to regulate
the gene they reside in (Neznanov et al. 1997; Lampe et al. 2008;
Tumpel et al. 2008; Ritter et al. 2012), we explicitly set out to
search for coding eExonsthat donotautoregulatebutrathercould
regulate their nearby genes. This was of interest to us due to the
phenotypic consequences that coding mutations could have on
their nearby genes. In order to do this, we needed a tissue-specific
ChIP-seq data sets that were shown to predict functional en-
hancers in three different mouse E11.5 tissues (forebrain, mid-
brain, limb) with high accuracy and tissue specificity (Visel et al.
2009a). In this data set, we observed that 4% of EP300 ChIP-seq
peaks from all three tissues overlap with coding exons after excluding
the first exon (Supplemental Table 1). These lower percentages could
be due to experimental differences such as cell line versus tissue. For
our functional assays, we next focused on the limb EP300 ChIP-seq
data set. We scanned this data set for exonic sequences that reside in
(see Methods). From the 252 limbEP300 ChIP-seqpeaks that overlaid
exons, 152 sequences overlapped coding exons and 134 were in
a gene that is not expressed in the limb (Supplemental Table 3).
Out of those 134 sequences, 90 had at least one limb expressed
gene up to 1 Mb away on either side of the gene. We chose seven
exons near important limb developmental genes (C14orf49 exon
16 [near DICER1], CDC14B exon 13 [near PTCH1], DYNC1I1 exon
15andexon17[near DLX5/6], HDAC9 exon18andexon19[near
TWIST1], STX18 exons 4–5 [near MSX1]) for subsequent mouse
enhancer assays (Supplemental Table 3).
Mouse enhancer assays
To test whether these exonic sequences function as enhancers, we
tested all seven sequences for their enhancer activity in mice. The
human sequences were cloned into the Hsp68-LacZ vector that
contains the heat shock protein 68 minimal promoter followed
by a LacZreporter gene (Kothary et al. 1988). Transgenicmice were
generated, and embryos were harvested at E11.5 and stained for
LacZ. We found that four out of the seven sequences showed limb
enhancer activity in mice. DYNC1I1 eExon 15 drove specific LacZ
expression in the limb mesenchyme and AER (Fig. 1B,B9; Supple-
mental Fig. 1A), and DYNC1I1 eExon 17 drove specific LacZ expres-
1C).HDAC9eExon18showedenhanceractivityinthe anterior limb
bud (Fig. 1G,G9; Supplemental Fig. 2A) and HDAC9 eExon 19 in the
posterior limb bud (Fig. 1H,H9; Supplemental Fig. 2B).
The exonic sequence is necessary for enhancer activity
The sequences tested in the mouse enhancer assays had some
intronic regions due to the ChIP-seq peak overlapping part of the
intron (Supplemental Table 3). We thus wanted to assess whether
the exonic sequence is necessary for enhancer activity. Since hu-
man limb and zebrafish fin development are considered highly
comparable on the molecular level (Hall 2007; Iovine 2007;
Mercader 2007) and since zebrafish enhancer assays are rapid and
cost-efficient, we carried out a deletion series analyses using this
assay. We first characterized whether our functional mouse limb
enhancers were positive for fin enhancer expression in zebrafish.
The four limb enhancers (Supplemental Table 3) were cloned from
human genomic DNA into a zebrafish enhancer assay vector, con-
taining anE1b minimal promoter followed by thegreen fluorescent
protein (GFP) reporter gene (Li et al. 2009). These vectors were
microinjected into one-cell-stage zebrafish embryos along with the
Tol2 transposase to facilitate genomic integration. GFP expression
was monitored at 48 and 72 h post-fertilization (hpf), both time
points when the pectoral fin can be observed. Two of our four
functional mouse limb enhancers, DYNC1I1 eExon 15 and HDAC9
eExon 19, were found to be functional fin enhancers in zebrafish
(Supplemental Table 4). At 72 hpf, DYNC1I1 eExon 15 drove GFP
expression in the pectoral fin, caudal fin, and somitic muscles (Sup-
plemental Fig. 1B), and HDAC9 eExon 19 exhibited enhancer ac-
In order to determine whether the actual exonic sequences are
necessary for enhancer activity, we used these two fin enhancers,
DYNC1I1 eExon 15 and HDAC9 eExon 19, for deletion series anal-
yses. DYNC1I1 eExon 15 was divided into three segments—59 in-
tron, exon, and 39 intron (Fig. 2A)—and HDAC9 eExon 19 was di-
vided into the following segments: 59 distal intron, 59 proximal
intron, and exon (Fig. 2C). We found that in both cases the exon
and 59 intron sequence adjacent to the exon had lower enhancer
activity by themselves, but when combined, their enhancer activity
respectively (Fig. 2B,D). These results demonstrate that the exonic
sequences are necessary but not sufficient for full enhancer activity.
Limb genes associated with eExons enhancer function
In order to identify the limb expressed genes that could be regu-
lated by our characterized eExons, we analyzed the RNA expres-
sion of nearby genes and carried out synteny block analysis (Ahituv
et al. 2005). Whole-mount in situ hybridization of neighboring
genes found that Dlx5/6 have similar limb expression patterns
to DYNC1I1 eExons 15 and17 (Fig. 1B–E9; Supplemental Figs. 1,3),
and Twist1 has a limb expression pattern that is similar to HDAC9
eExons 18 and 19 (Fig. 1G–-I9; Supplemental Figs. 2,3). We also
extractedRNA from E11.5 mouse limbs and adult mouse cortex and
heart and performed quantitative PCR (qPCR) to validate the tissue
specific expression of these genes. Dlx5, Dlx6, and Twist1 were
expressed in E11.5 limbs (Supplemental Fig. 3F,G). However,
Dync1i1 and Hdac9 were not detected in mouse E11.5 limbs but
expressed in the mouse adult cortex and heart, respectively (Sup-
plemental Fig. 3F,G). In addition, examination of the genomic
location of DYNC1I1-DLX5/6 and HDAC9-TWIST1 in various
vertebrate genomes shows that they remain adjacent to each
other from human tofish(SupplementalFig. 4). Basedonaprevious
analysis (Ahituv et al. 2005), the human–mouse–chicken DYNC1I1-
DLX5/6 block is 1.37 Mb in size and the HDAC9-TWIST1 is 2.52 Mb,
both above the 1.02 Mb average length (N50) of a human–mouse–
Dync1i1 eExon 15 is marked in the limb by an enhancer
eExon 15 for further functional analysis. We analyzed this eExon
Exonic enhancers of nearby genes
for histone modification signatures during limb development. We
15 for enhancer (H3K4me1, H3K27ac), promoter (H3K4me3), and
transcribed gene (H3K36me3) chromatin signatures (Hon et al.
2009). We found that in the mouse E11.5 limb, Dync1i1 eExon 15
is marked by H3K4me1 and H3K27ac (Fig. 3B,C) but not by
H3K4me3 or H3K36me3 (Fig. 3D,E). In contrast, Dync1i1 exon 6
was not marked by H3K4me1 or H3K27ac (Fig. 3B,C) in the limb,
and Dlx5/6 coding exons were marked by H3K36me3 (Fig. 3E).
Thus, the chromatin status correlates with the proposed limb en-
hancer activity of Dync1i1 eExon 15.
3C and DNA FISH show that Dync1i1 eExon 15 physically
interacts with the promoter regions of DLX5/ 6
To determine whether Dync1i1 eExon 15 physically interacts with
the Dlx5/6 promoter regions, we carried out 3C on mouse E11.5
heart and limb (AER enriched; see Methods) tissues. The mouse
heart tissue served as a negative control, as Dlx5/6 are not expressed
in the heart during that stage (Fig. 1D,E). We observed an increased
interaction frequency between Dync1i1 eExon 15 and the Dlx5/6
promoters in the limb tissue compared with the heart, indicating
a physical interaction between them in the limb (Fig. 4B). These
results suggest that Dync1i1 eExon 15 functions as an enhancer in
the AER through enhancer–promoter DNA looping.
To further analyze the chromosomal conformation around the
Dlx5/6 locus during limb development, we performed DNA FISH
using Dlx5/6 and Dync1i1 eExon 15 probes on mouse E11.5 limb
the physical distance between Dync1i1 eExon 15 and the Dlx5/6
coding region was calculated (Fig. 4C–J). Frequency distribution pat-
terns of the physical distance between Dync1i1 eExon 15 and Dlx5/6
AER, 35% of the Dync1i1 eExon 15 signals were in close proximity to
was greatly reduced (12%; P < 0.01, t-test) (Fig. 4L), and the overall
frequency of separated signals was higher compared to the AER
overlapping segments: 59intron, exon, 39intron. (Below)TheUCSC GenomeBrowser (http://genome.ucsc.edu)conservation trackshows thatonly the59
intron and exon are conserved between human and fish. (B) Zebrafish enhancer assay results for the different DYNC1I1 eExon 15 segments. While the 59
originally injected fragment of DYNC1I1 eExon 15. The 39 intron segment did not show enhancer activity. (C) HDAC9 eExon 19 was divided into three
overlapping segments: distal 59 intron, proximal 59 intron, and exon. (Below) The UCSC Genome Browser conservation track shows that the proximal 59
intron and exon are conserved between human and fish. (D) Zebrafish enhancer assay results for the different HDAC9 eExon 19 segments. While the
proximal 59 intron and exon show enhancer activity in the pectoral fin and branchial arches, only the combination of both gave comparable enhancer
expression to the previously injected 1098-bp HDAC9 eExon 19 sequence. The distal 59 intron segment did not show enhancer activity. Enhancer function
is plotted as percentage of GFP expression/total live embryos. Each of these segments was injected into at least 100 zebrafish embryos.
Segmental analysis of DYNC1I1 eExon 15 and HDAC9 eExon 19 enhancer function in zebrafish. (A) DYNC1I1 eExon 15 was divided into three
Birnbaum et al.
1062 Genome Research
(P < 0.01, t-test) (Fig. 4L), with a mean distance of 0.47 6 0.3 mm.
These results show that Dync1i1 eExon 15 is in close proximity with
Dlx5/6 promoter regions in the developing AER at E11.5, supporting
its proposed role as an enhancer during limb development.
Human chromosomal aberrations encompassing DYNC1I1
eExons 15 and 17 are associated with SHFM1
To test whether alterations of DYNC1I1 eExon 15 and 17 could
be associated with a limb phenotype, we analyzed available in-
dividuals and previously reported cases with SHFM1 (Fig. 5). We
translocation (Fig. 5). In addition, we mapped the inversion break-
points of a previously published SHFM1 family(Tackels-Horne etal.
2001) to be within chr 7: 96,219,611 and 109,486,136 (K6200
family) (Everman et al. 2005; Everman et al. 2006). In addition, we
referred to two recently reported SHFM1 cases: an individual with
SHFM1 who has a de novo pericentric inversion of chromosome 7:
46, XY, inv(7) (p22q21.3), with the breakpoint mapped to chr 7:
95.53–95.72 Mb (van Silfhout et al. 2009), and another individual
with a split foot phenotype who has an 880-kb microdeletion of
95.39–96.27Mb (Fig. 5; Kouwenhoven et al. 2010). It is worth not-
ing that an AER enhancer named BS1 was recently identified 300
kb centromeric to DLX5/6 (Kouwenhoven et al. 2010). However, at
least two individuals with SHFM1 that are described here have
chromosomal aberrations that do not include BS1 (Fig. 5), suggest-
ing that additional limb enhancers, such as DYNC1I1 eExon 15 and
17, could lead to SHFM1. All of the chromosomal abnormalities
described above overlap DYNC1I1 eExon 15 and 17 and suggest
that their removal could disrupt the transcriptional regulation of
Studies aimed at discovering gene regulatory elements usually
concentrate on noncoding DNA sequences as potential candidates
that protein coding sequences may have additional encrypted in-
formation in their sequence (Chamary et al. 2006; Itzkovitz and
Alon 2007; Lin et al. 2011). Here, by analyzing ChIP-seq data sets
for enhancer marks from various cell lines and tissues, we found
that on average 7% of peaks overlap with coding exons after ex-
cluding the first exon. With only ;1.6% of the human or mouse
genomes encoding for protein, eExons could be overrepresented
in these ChIP-seq enhancer-associated data sets. To test whether
exons are enriched in ChIP-seq enhancer data sets, we generated
a random data set from all mappable sequences that are used
for any whole-genome sequencing alignment from the UCSC
Genome Browser (http://moma.ki.au.dk/genome-mirror/cgi-bin/
tested how many peaks overlap exons compared to the ChIP-seq
data sets. We found a significantly higher percentage of peaks
overlapping with coding exons (after excluding the first exon) in
the EP300 ChIP-seq data sets of both GM12787 (P < 0.014; Fisher
addition, using a random sampling approach, we randomly sampled
1000 peaks (from all mappable sequences) having an equal distri-
bution to that of the two EP300 ChIP-seq data sets 1000 times and
10?16). Combined, these assays suggest an overrepresentation of
2 and promoter (pro.). (B) ChIP–qPCR analyses of H3K4me1, an enhancer histone mark. (C) ChIP-qPCR analyses of H3K27ac, an active enhancer histone
mark. (D) ChIP-qPCR analyses of H3K4me3, a promoter histone mark. (E) ChIP-qPCR analyses of H3K36me3, a transcribed gene histone mark. (X-axis)
Primer pairs; (y-axis) percentage of input recovery. (Error bars) SE from three technical replicates of a representative experiment.
Histone modification signatures of the Dync1i1 eExon 15 in the mouse E11.5 limb bud. (A) Schematic representation of the Dync1i1-Dlx5/6
Exonic enhancers of nearby genes
coding exons in ChIP-seq data sets. However, it is worth noting that
the technical variability of the ChIP-seq assay due to differences in
antibodies, cross-linking, pull down, sequencing depth, and others
along with sequence mappability are not taken into account in
Using a mouse transgenic enhancer assay for seven mouse
E11.5 limb EP300 ChIP-seq peaks, we show that four eExons are
functional limb enhancers and could regulate their neighboring
genes. The observed 57% (4/7) success rate does not imply that
;57% of exons overlapping enhancer-
associated ChIP-seq peaks are bona fide
enhancers and further functional assays
willbe neededin order todetermine this.
It is worth noting that Dlx5/6 and
Twist1 expression as detected by whole-
mount in situ hybridization is restricted
to the AER and the A-P domains of the
developing mouse limb, respectively.
However, the mouse limb enhancer ex-
pression pattern of DYNC1I1 eExon 15
extends into the limb bud mesenchyme
and the HDAC9 eExon 18 extends into
the posterior limb bud mesenchyme.
These expression pattern discrepancies
could be due to an inability of the whole-
mount in situ assay to detect low RNA
expression levels versus the more robust
staining seen through LacZ enhancer as-
says. One such example is the ability to
study the role and expression of the myo-
cyte enhancer factor 2C (MEF2C) gene in
the neural crest due to its enhancer func-
tion, which was previously confounded
and could not be observed through in
situ hybridization (Agarwal et al. 2011).
Alternatively, these limb enhancers could
potentially be regulating other limb-as-
sociated genes. However, our results for
DYNC1I1 eExon 15 demonstrate a physi-
cal interaction between this eExon and
the Dlx5/6 promoter regions, suggesting
that it regulates Dlx5/6 expression in the
limb. Another possibility for the discrep-
ancy in expression patterns could be as-
sociatedwith the ‘‘artificial’’natureofthe
transgenic enhancer assay, which could
lead to different results due to the use of
a minimal promoter instead of the pro-
moter of the regulated gene, the site of
transgene integration, the variation in
transgene copy number between trans-
Two of our characterized limb en-
of the cytoplasmic Dynein 1 motor pro-
tein complex that is not expressed in the
limb during development (Supplemen-
tal Fig. 3A,F; Crackower et al. 1999). Both
DYNC1I1 eExon 15 and 17 are present
in all three DYNC1I1 splice isoforms
NM_001135556.1), suggesting that alternative splicing does not
occur at these exons. DYNC1I1 eExon15 is also marked by an en-
hancerchromatin signature and physically interactswith the Dlx5/
6 promoter regions specifically in the limb. Both eExons encode
protein domains important for the cargo binding and specificity of
this Dynein (Kardon and Vale 2009). Cytoplasmic Dynein 1 is in-
volved in neuronal migration during brain development by inter-
acting with Dynactin and platelet-activating factor acetylhydrolase
1b (LIS1). Defects in neuronal migration can lead to brain malfor-
promoter regions in the mouse E11.5 limb. (A) Schematic of the Dync1i1-Dlx5/6 locus, showing the
relative location of the primers used for 3C and the BAC probes used for DNA-FISH. (B) Chromatin
looping events detected using 3C between Dync1i1 eExon 15 (orange oval) and promoters within the
Dync1i1-Dlx5/6 locus. The closest HindIII restriction sites (RS1 and RS2) of each promoter were used to
analyze the interaction frequencies to Dync1i1 eExon 15 (anchoring point). In the limb, the interaction
frequencies between Dync1i1eExon 15 and Dlx6 and Dlx5 promoter regions were significantly higher
compared to the heart negative control (more than 10- and 15-fold, respectively). No significant in-
teraction differences were found between Dync1i1eExon 15 and the Dync1i1 promoter, the closest
tested site to the anchoring point, or the two control regions (;900 kb away from the Dync1i1-Dlx5/6
locus) in limb versus heart tissues. (Error bars) SE of the average of three independent PCR reactions.
(C–L) DNA-FISH results with BAC probe RP23-430G21, which covers the Dync1i1eExon 15 region (red),
and BAC probe RP23-77O3, which covers the Dlx5/6 gene regions (green). (C) E11.5 limb section with
the dotted line highlighting the AER, as depicted by p63 staining in the nucleus. (D) BAC probes and
DAPI staining of E11.5 limbs. (Squares) Magnified regions in E and F that highlight the colocalization
of Dync1i1eExon 15 and Dlx5/6 signals. (G) E11.5 heart section shows p63 staining in the cytoplasm.
(H) BAC probes and DAPI staining of E11.5 heart. (Squares) Magnified regions in I and J that show
a separation of Dync1i1eExon 15 and Dlx5/6 signals. The white scale bars represent 5 mm length. (K,L)
Calculated frequencies for every 0.2 mm distance interval in mouse E11.5 AER (K) and heart (L) tis-
sues. (Black columns) Fraction of colocalized signals (0–0.2 mm). The number (n) of loci observed in
this experiment indicates a significant difference between the frequencies of the colocalized signals in
the AER and heart tissues (**P < 0.01; Student’s t-test).
3C and DNA-FISH show a physical interaction between Dync1i1 eExon 15 and Dlx5/6
Birnbaum et al.
mations such aslissencephaly, subcortical laminar heterotopias, and
pervasive developmental disorder-not otherwise specified (PDD-
NOS) (Kato and Dobyns 2003). Interestingly, an individual with
95.53–95.72 Mb) (Fig. 5), whose breakpoint has not been finely
characterized (van Silfhout et al. 2009). Further analysis would be
required in order to establish whether the PDD-NOS and SHFM
phenotypes in this individual could be due to the disruption of both
the DYNC1I1 gene and our characterized eExons.
Two other characterized limb enhancers, HDAC9 eExon 18
and 19, reside in the coding exons of HDAC9, a member of the
histone acetyltransferase class II family. HDAC9 eExon 19 has also
been shown to be an exonic remnant in zebrafish and speculated
to have a cis-regulatory function (Dong et al. 2010). Both HDAC9
eExons 18 and eExon 19 appear in 3/9 HDAC9 spice isoforms
(RefSeq: NM_178423.1, NM_058176.2, NM_178425.2). HDAC9
expression was shown to be more selective compared with that of
other HDAC family members (de Ruijter et al. 2003). Our RNA
analysis and whole-mount in situ hybridization results show that
Hdac9 is not expressed in the mouse limb at E11.5 (Supplemental
Fig. 3D,G). Hdac9-nullmice generatedby deletionof exons 4 and 5
are fertile and survive a normal life span but develop cardiac hy-
pertrophy with age and in response to pressure overload (Zhang
et al. 2002). Interestingly, despite Hdac9 not beingexpressed in the
limb at E11.5, Hdac9 homozygous knockout mice develop poly-
dactyly in their hindlimbs with partial penetrance (Morrison and
D’Mello 2008), similar to the polydactyly phenotype of Twist1
heterozygous knockout mice (Bourgeois et al. 1998). Although
the regulation of Twist1 by these and other potential Twist1 en-
hancers could be disrupted leading to the polydactyly phenotype.
The ability of eExons to enhance the expression of their nearby
genes, butnotthegenetheyreside in,couldsuggestthatmechanisms
such as those involved in epigenetic regulation and high-order chro-
matin organization might control their function in each tissue. To
can act as a protein coding sequence in one tissue but regulate the
expression of a nearby gene/s in another
tissue is novel. It raises the possibility
that mutations in a certain gene, even
synonymous ones, could potentially af-
fect the regulation of a nearby gene.
Therefore, careful analysis of the tissue-
would be required in order to determine
whether a phenotype is truly caused by
a mutation within its coding sequence.
Computational ChIP-seq data set
We identified exonic sequences in the
human hg18 and mouse mm9 genome
assemblies using the UCSC knownGene
track (http://genome.ucsc.edu). We down-
loaded all exonic sequences, including
59 UTR and 39 UTR, using the txStrat and
txEnd filter field. All exon sequence sizes
were divided by the number of exons to
calculate the average exon size. We down-
loaded coding exon sequences using the
cdsStart and cdsEnd filter field. The 22 ChIP-seq data sets of human
cell lines were obtained from Ernst et al. (2011), Myers et al. (2011),
and Rosenbloometal. (2012)and weredownloadedfrom theUCSC
Genome Browser, and the three EP300 ChIP-seq data sets of mouse
E11.5 tissues were obtained from Visel et al. (2009a) and down-
loaded from the Gene Expression Omnibus (http://www.ncbi.nlm.
nih.gov/geo) (Supplemental Table 1 includes links for all down-
loaded data). In order to unify our results, human sequences with
Browser LiftOver tool. A ChIP-seq peak was considered to overlap
an exon if at least 1 bp of exonic sequence overlapped. BED files of
all the ChIP-seq peaks that overlap exons in the various data sets
can be obtained at http://bts.ucsf.edu/ahituv/resources.html. First
exons for all splice isoforms of a gene were determined by the
exonStarts exonEnds field in the UCSC knownGene track. To
data from the Mouse Genome Informatics (MGI) gene expression
data query form (http://www.informatics.jax.org/javawi2/servlet/
or TS20 (E11.5–13.0).
Transgenic enhancer assays
By use of primers designed to amplify the EP300 ChIP-seq peaks
that overlap exons (Supplemental Table 5), we carried out poly-
merase chain reaction (PCR) on human genomic DNA (Qiagen).
Primers were designed to have up to 500 bp additional sequence
flanking the EP300 peak. Previous experiments have shown this to be
a reliable methodfor obtaining positive enhancer activity whenusing
evolutionary conserved regions (Pennacchio et al. 2006) and EP300
ChIP-seq peaks (Visel et al. 2009a). For the mouse enhancer assays,
containing the Hsp68 minimal promoter followed by the LacZ re-
mice were generated by the UCSF transgenic facility and by Cyagen
Biosciences using standard procedures (Nagy et al. 2002). Embryos
were harvested at E11.5 and stained for LacZ expression as previously
described (Pennacchio et al. 2006). For the zebrafish enhancer assays,
matic representation of the genomic positions of breakpoints from chromosomal rearrangements in
individuals with SHFM1 mapped to human genome assembly 18 (hg18) and compared to the location
of DYNC1I1 eExon 15 and 17. An 880-kb microdeletion in an individual with a split foot pheno-
type was found to be at 95.39–96.27 Mb (Kouwenhoven et al. 2010). In the GK family, the
the chromosomal inversion breakpoints mapped to chr 7: 96,219,611 and 109,486,136. The breakpoint
coordinates of a 7:46, XY, inv(7) (p22q21.3) with SHFM1 and pervasive developmental disorder-not
these chromosomal abnormalities overlap with DYNC1I1 eExon 15 and 17 (orange ovals). Two of these
chromosomal aberrations do not overlap with the BS1 AER enhancer (white oval). (Lightning bolts)
Translocation and inversion breakpoints; (diamonds) deletion.
Chromosomal abnormalities at chromosome 7q21-23 associated with SHFM1. A sche-
Exonic enhancers of nearby genes
the same human PCR products were cloned into the E1b-GFP-Tol2
enhancer assay vector containing an E1b minimal promoter followed
by GFP (Li et al. 2009). They were injected following standard
procedures (Nusslein-Volhard and Dahm2002; Westerfield 2007)
into at least 100 embryos per construct along with Tol2 mRNA
was observed and annotated at 48 and 72 hpf. An enhancer was
considered positive if 60% of the GFP expressing fish showed
a consistent expression pattern. All animal work was approved by
the UCSF Institutional Animal Care and Use Committee.
Whole-mount in situ hybridization
Mouse E11.5 embryos were fixed in 4% paraformaldehyde. Clones
Dlx5 (Depew et al. 1999), Dlx6 (OMM5895-99863403 Open Bio-
systems), Hdac9 (EMM1032-601163 and EMM1002-6974502, Open
Biosystems), and Twist1 (Chen and Behringer 1995) were used as
templates for digoxygenin-labeled probes. Mouse whole-mount in
(Hargrave et al. 2006).
RNA expression analysis
Mouse E11.5 limb and AER enriched tissues (limb buds where the
AER region was carefully dissected), and adult mouse heart and
cortex tissues were dissected. Total RNA was isolated using RNeasy
(Qiagen) according to the manufacturer’s protocol. qPCR was
performed using SsoFast EvaGreen Supermix (Biorad) and run on
the Eppendorf Mastercycler ep realplex 2 thermal cycler. Samples
were tested in duplicates. Specificity and absence of primer dimers
was controlled by denaturation curves. b-Actin (Actb) mRNA was
used for normalization. Primer sequences used for amplification
are listed in Supplemental Table 5.
ChIP followed by qPCR
ChIP following standard techniques (Nelson et al. 2006) was per-
formed on mouse E11.5 AER-enriched tissue. For each ChIP, 100–
500 mg of chromatin was used. For immunoprecipitation, we used
2 mg of H3K4me1 (ab8895, Abcam), H3K4me3 (ab8580, Abcam),
H3K27ac (ab4729, Abcam), and H3K36me3 (ab9050; Abcam) an-
tibodies. qPCR was carried out using SsoFast EvaGreen Supermix
(Biorad) and run on the Eppendorf Mastercycler ep realplex 2
thermal cycler. ChIP-qPCR signals were standardized to input
chromatin (percentage of input). Primer sequences used for am-
plification are listed in Supplemental Table 5.
3C was performed following standard procedures (Dostie and
Dekker 2007). Mouse E11.5 heart and AER enriched tissues were
dissected from 30 embryos, cross-linked with 1% formaldehyde,
and processed to get single cell preparations. Cells were lysed to
purify nuclei and digested with HindIII (1200 units) restriction
enzyme (New England Biolabs). Cross-linked fragments were li-
d at 4°C. The samples were reverse cross-linked, and purified DNA
was amplified by whole-genome amplification (WGA2, Sigma-
Aldrich). Product detection was done in triplicate by qPCR, as de-
scribed above for ChIP, and averaged for each primer pair (Sup-
plemental Table 5). Each data point was first corrected for PCR bias
by dividing the average of three PCR signals by the average signal
of an internal control template. Data from AER and heart were
normalized to a BAC library containing seven BACs obtained from
the CHORI BACPAC resource center covering the SHFM1 minimal
region (RP23-430G21, RP24-73K21, RP23-336P10, RP23-389M11,
DNA florescent in situ hybridization
DNA florescent in situ hybridization (FISH) was carried out as pre-
viously described (Lomvardas et al. 2006). BAC clones RP23-77O3
for Dlx5/6 and RP23-430G21 for Dync1i1 were obtained from the
CHORI BACPAC resource center. Probes were labeled with Digoxi-
genin-11-dUTP or Biotin-16-dUTP by Nick Translation (Roche).
Limb or hearttissues (E11.5) were embedded withoutfixation,and
10 mM cryosectionswere collected on Superfrost Plusslides (Fisher).
After drying, sections were fixed in 4% PFA for 5 min at 4°C. DNA
was fragmented by incubation with 0.1 M HCl for 5 min at room
temperature, and slides were treated with RNase A for 1 h at 37°C.
Slides were dried by an ethanol series, denatured in a solution of
75% formamide in 2xSSC for 5 min at 85°C, rinsed immediately in
ice-cold 2xSSC, and dried again by 4°C ethanol series. Pre-dena-
tured, CotI-annealed probes were applied overnight. The probe was
washed three times for 15 min in 55% formamide, 0.1% NP-40 in
2xSSC at 42°C. Probes were detected using Dylight 488 anti-digoxi-
genin and Dylight 549 anti-biotin (Jackson Immunoresearch).
Antibody washes were carried out in a solution of PBS containing
0.1% Triton-x-100 and 8% formamide at room temperature. All
images were obtained using confocal fluorescence microscopy
(Nikon C1 Spectral). FISH signals were recorded in three separate
RGB channels. The image stacks were reconstructed using the
Volocity program (PerkinElmer), and the shortest distance between
Subjects and chromosomal breakpoint mapping
The GK family consisted of a male who had ectrodactyly,
micrognathia, an elongated neck, and bilateral microtia with
neurosensory deafness and his female offspring who died before
birth and had ectrodactyly, micrognathia, and bilateral microtia.
balanced chromosomal translocation 46,XY,t(7;20)(q22;p13) that
was not found in GK’s healthy mother. By use of FISH, following
standard techniques (Trask 1991), with two BACs (RP11-94N7,
RP11-78B12), the breakpoint coordinates at chromosome 7 were
mapped to be between 96.2 and 96.47 Mb. The K6200 family had
autosomal dominant SHFM and variable sensorineural hearing loss
as previously reported (Tackels-Horne et al. 2001). Subsequent
studies of this family by pulse field gel electrophoresis and FISH
identified a chromosome inversion with breakpoints in the SHFM1
critical region (Everman et al. 2005; Everman et al. 2006). Southern
2002) were then used to identify the inversion breakpoints (D.B.
Everman, C.T. Morgan, M.E. Laughridge, T. Moss, S. Ladd, B.
DuPont, D. Toms, A. Dobson, K.D. Clarkson, F. Gurrieri, et al.,
unpubl.). The inversion in this family was balanced, with minimal
changes in the normal sequence at each breakpoint and segregated
with the SHFM/hearing loss phenotype.
We thankmembers of the Ahituv laboratory for helpfulcomments
on the manuscript. We also thank Juhee Jeong and John L.R.
Rubenstein for reagents. This research was supported by NICHD
grant no. R01HD059862. N.A. and G.B. are also supported by
NHGRI grant number R01HG005058, and N.A. is also supported
by NIGMS award number GM61390. M.J.K. was supported in part
by NIH Training Grant T32 GM007175 and the Amgen Research
Excellence in Bioengineering and Therapeutic Sciences Fellow-
Birnbaum et al.
ship. O.A. and O.S.B. were supported by the Morris Kahn family
foundation. D.B.E and C.E.S. were supported in part by a grant
from the South Carolina Department of Disabilities and Special
Needs, the Genetic Endowment of South Carolina, and a previous
grant (no.8510)from ShrinersHospitalsfor Children.Thecontent
is solely the responsibility of the authors and does not necessarily
represent the official views of the NIH, NICHD, NHGRI, NIDCR, or
E, Anderson JP, Wei ML, et al. 2011. The MADS box transcription factor
MEF2C regulates melanocyte development and is a direct
transcriptional target and partner of SOX10. Development 138: 2555–
Ahituv N,PrabhakarS, Poulin F,RubinEM, Couronne O.2005. Mappingcis-
regulatory domains in the human genome using multi-species
conservation of synteny. Hum Mol Genet 14: 3057–3063.
Bourgeois P, Bolcato-Bellemin AL, Danse JM, Bloch-Zupan A, Yoshiba K,
Stoetzel C, Perrin-Schmitt F. 1998. The variable expressivity and
incompletepenetranceof thetwist-nullheterozygous mouse phenotype
resemble those of human Saethre-Chotzen syndrome. Hum Mol Genet 7:
Chamary JV, Parmley JL, Hurst LD. 2006. Hearing silence: Non-neutral
evolution at synonymous sites in mammals. Nat Rev Genet 7: 98–108.
Chen ZF, Behringer RR. 1995. twist is required in head mesenchyme for
cranial neural tube morphogenesis. Genes Dev 9: 686–699.
Crackower MA, Sinasac DS, Xia J, Motoyama J, Prochazka M, Rommens JM,
Scherer SW, Tsui LC. 1999. Cloning and characterization of two
cytoplasmic dynein intermediate chain genes in mouse and human.
Genomics 55: 257–267.
Depew MJ, Liu JK, Long JE, Presley R, Meneses JJ, Pedersen RA, Rubenstein
JL. 1999. Dlx5 regulates regional development of the branchial arches
and sensory capsules. Development 126: 3831–3846.
de Ruijter AJ, van Gennip AH, Caron HN, Kemp S, van Kuilenburg AB. 2003.
Histone deacetylases (HDACs): Characterization of the classical HDAC
family. Biochem J 370: 737–749.
Dong X, Navratilova P, Fredman D, Drivenes O, Becker TS, Lenhard B. 2010.
Exonic remnants of whole-genome duplication reveal cis-regulatory
function of coding exons. Nucleic Acids Res 38: 1071–1085.
genomic elements using 5C technology. Nat Protoc 2: 988–1002.
Eichenlaub MP, Ettwiller L. 2011. De novo genesis of enhancers in
vertebrates. PLoS Biol 9: e1001188. doi: 10.1371/journal.pbio.1001188.
Elliott AM, Evans JA. 2006. Genotype-phenotype correlations in mapped
split hand foot malformation (SHFM) patients. Am J Med Genet A 140:
Ernst J, Kheradpour P, Mikkelsen TS, Shoresh N, Ward LD, Epstein CB,
Zhang X, Wang L, Issner R, Coyne M, et al. 2011. Mapping and
analysis of chromatin state dynamics in nine human cell types. Nature
Everman D, Morgan C, Clarkson K, Gurrieri F, McAuliffe F, Chitayat D,
Stevenson R, Schwartz C. 2005. Submicroscopic rearrangements
split-hand/foot malformation and sensorineural hearing loss. Proc
Greenwood Genet Cent 24: 137.
Everman D, Morgan C, Stevenson R, Schwartz C. 2006. Chromosome
rearrangements: An emerging theme in the causation of split-hand/foot
malformation. Proc Greenwood Genet Cent 25: 138–139.
Firulli BA, Krawchuk D, Centonze VE, Vargesson N, Virshup DM, Conway
SJ, Cserjesi P, Laufer E, Firulli AB. 2005. Altered Twist1 and Hand2
dimerization is associated with Saethre-Chotzen syndrome and limb
abnormalities. Nat Genet 37: 373–381.
Gilbert SF. 2000. Developmental Biology, 6th ed. Sinauer Associates,
Hall B.K. 2007. Fins into limbs. The University of Chicago Press, Chicago.
Hargrave M, Bowles J, Koopman P. 2006. In situ hybridization of whole-
mount embryos. Methods Mol Biol 326: 103–113.
Heintzman ND, Ren B. 2009. Finding distal regulatory elements in the
human genome. Curr Opin Genet Dev 19: 541–549.
Heintzman ND, Hon GC, Hawkins RD, Kheradpour P, Stark A, Harp LF, Ye Z,
Lee LK, Stuart RK, Ching CW, et al. 2009. Histone modifications at
human enhancers reflect global cell-type-specific gene expression.
Nature 459: 108–112.
Hon GC, Hawkins RD, Ren B. 2009. Predictive chromatin signatures in the
mammalian genome. Hum Mol Genet 18: R195–R201.
Iovine MK. 2007. Conserved mechanisms regulate outgrowth in zebrafish
fins. Nat Chem Biol 3: 613–618.
Itzkovitz S, Alon U. 2007. The genetic code is nearly optimal for allowing
additional information within protein-coding sequences. Genome Res
Kardon JR, Vale RD. 2009. Regulators of the cytoplasmic dynein motor. Nat
Rev Mol Cell Biol 10: 854–865.
Kato M, Dobyns WB. 2003. Lissencephaly and the molecular basis of
neuronal migration. Hum Mol Genet 12: R89–R96.
Kawakami K. 2005. Transposon tools and methods in zebrafish. Dev Dyn
Kothary R, Clapoff S, Brown A, Campbell R, Peterson A, Rossant J. 1988. A
transgene containing lacZ inserted into the dystonia locus is expressed
in neural tube. Nature 335: 435–437.
Kouwenhoven EN, van Heeringen SJ, Tena JJ, Oti M, Dutilh BE, Alonso ME,
de la Calle-Mustienes E, Smeenk L, Rinne T, Parsaulian L, et al. 2010.
Genome-wide profiling of p63 DNA-binding sites identifies an element
that regulates gene expression during limb development in the 7q21
SHFM1 locus. PLoS Genet 6: e1001065. doi: 10.1371/journal.
Lampe X, Samad OA, Guiguen A, Matis C, Remacle S, Picard JJ, Rijli FM,
Rezsohazy R. 2008. An ultraconserved Hox-Pbx responsive element
resides in the coding sequence of Hoxa2 and is active in rhombomere 4.
Nucleic Acids Res 36: 3214–3225.
Li Q, Ritter D, Yang N, Dong Z, Li H, Chuang JH, Guo S. 2009. A systematic
approach to identify functional motifs within vertebrate developmental
enhancers. Dev Biol 337: 484–495.
Lin MF, Kheradpour P, Washietl S, Parker BJ, Pedersen JS, Kellis M. 2011.
Locating protein-coding sequences under selection for additional,
overlapping functions in 29 mammalian genomes. Genome Res 21:
Lomvardas S, Barnea G, Pisapia DJ, Mendelsohn M, Kirkland J, Axel R. 2006.
Interchromosomal interactions and olfactory receptor choice. Cell 126:
Mercader N. 2007. Early steps of paired fin development in zebrafish
compared with tetrapod limb development. Dev Growth Differ 49: 421–
Morrison BE, D’Mello SR. 2008. Polydactyly in mice lacking HDAC9/HDRP.
Exp Biol Med (Maywood) 233: 980–988.
Myers RM, Stamatoyannopoulos J, Snyder M, Dunham I, Hardison RC,
Bernstein BE, Gingeras TR, Kent WJ, Birney E, Wold B, et al. 2011. A
user’s guide to the encyclopedia of DNA elements (ENCODE). PLoS Biol
9: e1001046. doi: 10.1371/journal.pbio.1001046.
Nagy A, Gertsenstein M, Vintersten K, Behringer R. 2002. Manipulating the
mouse embryo:A laboratory manual.Cold Spring Harbor Laboratory Press,
Cold Spring Harbor, NY.
Nelson JD, Denisenko O, Bomsztyk K. 2006. Protocol for the fast chromatin
immunoprecipitation (ChIP) method. Nat Protoc 1: 179–185.
Neznanov N, Umezawa A, Oshima RG. 1997. A regulatory element within
a coding exon modulates keratin 18 gene expression in transgenic mice.
J Biol Chem 272: 27549–27557.
Nissim S, Tabin C. 2004. Development of the limbs. In Inborn errors of
development (ed. C.E.R. Erickson and T. Wynshaw-Boris), pp. 148–167.
Oxford University Press, New York.
Nusslein-Volhard C, Dahm R. 2002. Zebrafish. Oxford University Press,
O’Rourke MP, Soo K, Behringer RR, Hui CC, Tam PP. 2002. Twist plays an
essential role in FGF and SHH signal transduction during mouse limb
development. Dev Biol 248: 143–156.
Pennacchio LA, Ahituv N, Moses AM, Prabhakar S, Nobrega MA, Shoukry M,
Minovitsky S, Dubchak I, Holt A, Lewis KD, et al. 2006. In vivo enhancer
Ritter DI, Dong Z, Guo S, Chuang JH. 2012. Transcriptional enhancers in
protein-coding exons of vertebrate developmental genes. PLoS One. doi:
Robledo RF, Rajan L, Li X, Lufkin T. 2002. The Dlx5 and Dlx6 homeobox
genes are essential for craniofacial, axial, and appendicular skeletal
development. Genes Dev 16: 1089–1101.
Rosenbloom KR, Dreszer TR, Long JC, Malladi VS, Sloan CA, Raney BJ, Cline
MS, Karolchik D, Barber GP, Clawson H, et al. 2012. ENCODE whole-
genome data in the UCSC Genome Browser: Update 2012. Nucleic Acids
Res 40: D912–D917.
a novel DLX5 mutation in a family with autosomal recessive split hand
and foot malformation. J Med Genet 49: 16–20.
Tackels-Horne D, Toburen A, Sangiorgi E, Gurrieri F, de Mollerat X, Fischetto
R, Causio F, Clarkson K, Stevenson RE, Schwartz CE. 2001. Split hand/
split foot malformation with hearing loss: First report of families linked
to the SHFM1 locus in 7q21. Clin Genet 59: 28–36.
Trask BJ. 1991. Fluorescence in situ hybridization: Applications in
cytogenetics and gene mapping. Trends Genet 7: 149–154.
Exonic enhancers of nearby genes
Tumpel S, Cambronero F, Sims C, Krumlauf R, Wiedemann LM. 2008. A Download full-text
regulatory module embedded in the coding region of Hoxa2 controls
expression in rhombomere 2. Proc Natl Acad Sci 105: 20077–20082.
van Silfhout AT, van den Akker PC, Dijkhuizen T, Verheij JB, Olderode-
Berends MJ, Kok K, Sikkema-Raddatz B, van Ravenswaaij-Arts CM. 2009.
Split hand/foot malformation due to chromosome 7q
aberrations(SHFM1): Additional support for functional
haploinsufficiency as the causative mechanism. Eur J Hum Genet 17:
Vervoort VS, Viljoen D, Smart R, Suthers G, DuPont BR, Abbott A, Schwartz
CE. 2002. Sorting nexin 3 (SNX3) is disrupted in a patient with
a translocation t(6;13)(q21;q12) and microcephaly, microphthalmia,
ectrodactyly, prognathism (MMEP) phenotype. J Med Genet 39: 893–
Visel A, Blow MJ, Li Z, Zhang T, Akiyama JA, Holt A, Plajzer-Frick I, Shoukry
M, Wright C, Chen F, et al. 2009a. ChIP-seq accurately predicts tissue-
specific activity of enhancers. Nature 457: 854–858.
Visel A, Rubin EM, Pennacchio LA. 2009b. Genomic views of distant-acting
enhancers. Nature 461: 199–205.
Westerfield M. 2007. The zebrafish book. University of Oregon, Eugene.
Woolfe A, Goodson M, Goode DK, Snell P, McEwen GK, Vavouri T, Smith SF,
North P, Callaway H, Kelly K, et al. 2005. Highly conserved non-coding
sequences are associated with vertebrate development. PLoS Biol 3: e7.
Zeller R, Lopez-Rios J, Zuniga A. 2009. Vertebrate limb bud development:
Moving towards integrative analysis of organogenesis. Nat Rev Genet 10:
histone deacetylases act as signal-responsive repressors of cardiac
hypertrophy. Cell 110: 479–488.
Received October 19, 2011; accepted in revised form March 19, 2012.
Birnbaum et al.
1068 Genome Research