Introductory
IMAGE Syndrome (Intrauterine growth restriction, Metaphyseal dysplasia, Adrenal hypoplasia
congenita, and Genital anomalies) is an undergrowth developmental disorder with life-threatening
consequences
1
. Identity-by-descent analysis in a family with IMAGE syndrome
2
identified a 17.2 megabase (Mb) locus on 11p15 that segregated in affected family
members. Targeted exon array capture of the disease locus, followed by high-throughput
genomic sequencing and validated by dideoxysequencing, identified missense mutations
in imprinted gene CDKN1C (P57KIP2) in two familial and four unrelated patients. Familial
analysis demonstrated an imprinted mode of inheritance where only maternal transmission
of the mutation resulted in IMAGE syndrome. CDKN1C inhibits cell-cycle progression
3
and targeted expression of IMAGE-associated CDKN1C mutations in Drosophila caused
severe eye growth defects compared to wild type CDKN1C, suggesting a gain-of-function
mechanism. All IMAGE-associated mutations clustered in the PCNA-binding domain of
CDKN1C and resulted in loss of PCNA binding, distinguishing them from CDKN1C mutations
that cause Beckwith-Wiedemann Syndrome, an overgrowth syndrome
4
.
Since the initial description of IMAGE syndrome (OMIM #300290)
1
, a number of isolated and familial cases have been reported
1,5–11
. In order to identify a causative gene for IMAGE syndrome, we performed a 250K Nsp
Affymetrix SNP Array on seven affected and one unaffected sibling from Family A, a
five-generation family from Argentina (Fig. 1a). Further analysis using a custom script
detected a 17.2 Mb identical-by-descent (IBD) region on chromosome 11, shared by seven
affected family members but not an unaffected sibling (Fig. 1b), with a LOD score
of 5.4. Despite the multisystem involvement of IMAGE syndrome, we did not identify
a contiguous gene deletion or duplication in the affected individuals (Supplementary
Fig. 1).
To determine the causative mutation, we performed targeted high-throughput genomic
sequencing of all the exons within a conservative IBD region. An Agilent 244K custom
CGH array was designed to capture all exons and splice sites within the region spanning
0–22.6 Mb on chromosome 11. In total, five custom bar-coded genomic DNA libraries
(Family A, V-1 and V-6; unrelated Patients 1, 2, and 3) were prepared, pooled, and
captured on a single custom array, and the DNA enriched for the IBD region was sequenced
on one lane of the Illumina Genome Analyzer II. Our targeted approach yielded an average
coverage of 32x. Patient 3 had a significantly lower coverage of 9x across the targeted
intervals and was not used in the initial bioinformatics analysis. In the remaining
four samples analyzed, ~85% of all targeted regions were covered at ≥10x.
The pedigree and IBD analysis led us to hypothesize that IMAGE syndrome was inherited
as a rare autosomal dominant disorder. Our bioinformatics analysis required that both
V-1 and V-6 share the same rare gene variant, and at least one of the non-related
IMAGE syndrome patients harbor a rare variant (defined as a variant not present in
dbSNP129) in the same gene. This approach identified a single gene, CDKN1C.
Upon further examination, we noted that CDKN1C was captured and sequenced at a much
lower rate compared to other targeted genes due to a high GC-content of ≥80%. To compound
this low gene coverage, we re-sequenced CDKN1C by dideoxysequencing (primers listed
in Supplementary Table 1) in all five individuals sequenced by high-throughput sequencing
and in an additional sporadic case (Patient 4). Affected individuals from Family A
carried a c.825T>G change resulting in a F276V missense mutation. The four unrelated
patients with IMAGE syndrome harbored four mutations in CDKN1C: F276S, R279P, D274N,
and K278E (Table 1). In total, we identified five rare heterozygous missense mutations
in CDKN1C that cluster within six amino acids of the PCNA-binding domain
12
(Fig. 2a). All variants localized to a highly conserved region
13
(Fig. 2b) and are predicted to be “damaging” to the structure and function of CDKN1C
by Polyphen analysis
14
.
CDKN1C is located on chromosome 11 and encodes a protein known to play a key role
in inhibiting cell cycle progression. In most tissues, the paternal allele is repressed
by distant imprinting control regions, such that expression is primarily from the
maternal allele
15,16
. Inheritance of IMAGE syndrome in Family A was only through maternal transmission
of the CDKN1C mutation (Fig. 1a). Sequencing for the c.825T>G mutation in 24 members
from Family A confirmed that only individuals who inherited the mutation on the maternal
allele are affected. A c.825T>G mutation inherited on the paternal allele was not
expressed, presumably because of epigenetic silencing of the mutated allele.
To confirm the pathogenicity of these mutations, we used an in vivo functional model
in which IMAGE-associated human CDKN1C mutants were expressed in Drosophila melanogaster
using the GAL4 UAS system
17
. Ubiquitous overexpression of wild type or mutant CDKN1C resulted in early larval
lethality. Targeted expression of the IMAGE-associated CDKN1C mutants resulted in
altered wing vein patterning and decreased wing size (see Supplementary Fig. 2 & 3).
Expression of wild type CDKN1C restricted to the eye did not have any effects on adult
eye size, while expression of IMAGE-associated CDKN1C mutants showed a moderate to
severe reduction in eye size (Fig. 3).
Overexpression of IMAGE-associated CDKN1C mutants in HEK293T cells did not interfere
with the ability of CDKN1C to inhibit the cell cycle in G0/G1 through binding of the
CDK-domain (Supplementary Fig. 4). These data suggest that IMAGE-mutations within
the PCNA-binding domain do not inhibit cell-cycle progression and likely act through
a different mechanism resulting in IMAGE syndrome.
CDKN1C is located within an imprinted cluster of genes that regulate prenatal and
postnatal growth and development. Genetic alterations in CDKN1C have been shown to
give rise to Beckwith-Wiedemann Syndrome (BWS; OMIM #130650), an overgrowth disorder
18,19
Here, we show that an undergrowth condition, IMAGE syndrome, is caused by domain-specific
mutations in the maternally inherited allele of CDKN1C.
Both BWS human patients and CDKN1C
−/− knockout mice exhibit adrenal hyperplasia
20,21
in contrast to the adrenal hypoplasia in IMAGE patients. We therefore verified that
CDKN1C mRNA and protein are expressed in the developing human adrenal gland (Fig.
4). Quantitative RT-PCR demonstrated that the expression of CDKN1C is greater in adrenal
tissue during early human development than in brain or muscle (Fig. 4a). Immunohistochemistry
showed strongest expression of CDKN1C within a subset of cells in the subcapsular
or developing definitive zone of the adrenal gland (Fig. 4b).
To determine if BWS and IMAGE mutations work through the same mechanism, we repeated
the above functional studies with BWS-specific mutants. In the cell cycle analysis,
transfection of the BWS-mutant CDKN1C-L42P resulted in a loss of cell cycle inhibition
at G0/G1. Ubiquitous expression of BWS-associated CDKN1C mutations in Drosophila melanogaster
were early larval lethal, however targeted expression had no effect on eye size, wing
size, or wing vein patterning (Fig. 3g,h, Supplementary Fig. 2e,2f,3e,3f). Thus, in
vitro and in vivo, BWS mutants have different effects relative to the IMAGE mutants
suggesting that domain-specific mutations have differential effects on cell cycle
progression and developmental processes.
BWS-associated CDKN1C mutations are either missense mutations localized to the cyclin-dependent
kinase binding domain or nonsense mutations, both of which result in protein loss-of-function,
over proliferation, and predisposition to cancer
22
due to loss of cell-cycle inhibition
3
. In contrast, we show that missense mutations localized to a highly conserved region
of the PCNA-binding domain of CDKN1C in IMAGE syndrome resulted in excess inhibition
of growth and differentiation—a gain of function (Fig. 5).
Since two of the five mutations fall into a putative nuclear localization signal,
we determined whether IMAGE mutants interfere with active nuclear transport of CDKN1C.
H295R and M1 cells transfected with GFP-CDKN1C fusion constructs (Supplementary Fig.
5), showed that none of the tested IMAGE mutants interfered with active transport
mechanisms or with binding affinity to α-importin.
Since IMAGE mutations cluster in a domain known to bind PCNA, we performed co-immunoprecipitation
experiments to test the effect of IMAGE mutations on PCNA binding. HEK293T cells were
transfected with flag-tagged CDKN1C constructs bearing the wild type or IMAGE alleles
(F276V and K278E). Endogenous PCNA was recovered from the wild type but not IMAGE
immunoprecipitates (Fig. 4c), suggesting that PCNA binding is disrupted in mutants.
Since one of the roles of PCNA is to facilitate ubiquitination of cell cycle proteins
23
, we investigated the role of IMAGE mutations in PCNA-dependent ubiquitination of
CDKN1C. HEK293T cells were co-transfected with flag-tagged wild type or IMAGE-mutant
F276V CDKN1C and with HA-tagged Ubiquitin (12 kDa) and subjected to co-immunoprecipitation.
CDKN1C migrates at ~50 kDa, and therefore we expect the mono-, di-, and poly-ubiquitinated
CDKN1C to migrate at ~62 kDa or higher depending on the number and the branching of
the ubiquitin moieties. Here we show that a band at 63 kDa, the approximate size of
a monoubiquitinated CDKN1C protein, is present in the wild type CDKN1C but absent
in the IMAGE mutant sample (Fig.4d).
Our data reveal a role for PCNA binding in a specific ubiquitination modification
of CDKN1C. Many cell cycle proteins are subject to PCNA-dependent ubiquitination
24
, which has pleiotropic effects. Monoubiquitination, as observed in our data, may
have a number of functional consequences, such as modulation of protein localization,
of protein interactions, and of proteosomal degradation
25,26,27
. The latter is less likely because it typically requires, at minimum, tetra-ubiquitination
28
, but cannot be ruled out without information on CDKN1C protein stability.
Next generation sequencing has emerged as a powerful tool in identifying rare Mendelian
disease genes, utilizing existing linkage analysis data to identify a disease gene
in an unbiased fashion. Our findings show that missense mutations in the PCNA-binding
domain have an inhibitory effect on growth, via loss of binding of PCNA to CDKN1C,
thereby altering ubiquitination of CDKN1C and presumably promoting its function. The
contrast between BWS and IMAGE mutations in CDKN1C highlights the dual and opposing
effects of specific CDKN1C mutations. Mutations within the PCNA-binding site of CDKN1C
blocked in vivo growth and differentiation and may illuminate novel mechanisms regulating
cell transformation
12
, tumor growth, and cell cycle progression.
METHODS
Study Subjects
All participants were patients diagnosed clinically with IMAGE syndrome. This study
was approved by the Institutional Review Board (IRB) at the University of California,
Los Angeles, the Hospital de Niños Ricardo Gutierrez in Argentina, or the Institute
of Child Health at University College London. All participants provided informed consent.
Phenotypes are summarized in Table 1.
Identity-by-descent analysis
Genomic DNA from 7 affected individuals (IV-10, V-1, V-2, V-5, V6, V-7, V-12) and
1 non-affected individual (V-13) from a large Argentine family (Family A) were genotyped
on the Affymetrix 250K Nsp1 SNP arrays, as per manufacturers’ protocol. Familial relationships
were confirmed through checking the sharing statistics in all pairs of samples and
ancestral identity-by-descent (IBD) analysis was performed using a custom script (B.
Merriman, available on request). The IBD analysis script was designed to search for
long continuous intervals compatible with a common extended haplotype among all the
affected individuals but not with the unaffected individuals.
29
A conservative error rate of 1% was used to allow the algorithm to tolerate possible
genotyping errors. A rare dominant model of inheritance was assumed with rare frequency
of this haplotype of 0.1% and penetrance of 100% under the assumption that the carriers
(parents of the affected individuals) were not showing the phenotype for a different
reason (i.e. imprinting) than low penetrance of the disease phenotype.
Capture and Sequencing of Genomic DNA
For the capture of the genes within the IBD interval, we used Agilent Custom 244K
comparative genomic hybridization Array. The 60-bp oligonucleotide probes were tiled
every 20-bp against all exonic regions on chromosome 11 between 2.45-20.15Mb (hg18,
March 2006, build 36.1) and every 30-bp in the flanking 5’ and 3’ regions spanning
0-2.45 Mb and 20.15-22.6 Mb. We included all gene models identified in RefSeq, Genbank,
CCDS and UniProt. Location and sequence of all probes are available upon request.
Genomic DNA libraries were created for patients V-1 an V-6 from Family A and 3 isolated
cases of IMAGE
1
(Patient1, 2, and 3) following manufacturer’s protocol (Illumina protocol Preparing
Samples for Sequencing Genomic DNA, p/n 11251892 Rev. A) except for the adaptor ligation
step where we used custom-made, internally validated bar-coded adaptors. After the
PCR amplification, five bar-coded libraries were pooled together and captured by hybridization
to custom-designed CGH array, as described in
30
. After the capture, the array was washed and the captured DNA was eluted, amplified
and diluted to 10nM final concentration based on the Qubit concentration measure and
Agilent Bioanalyzer. One flowcell lane of single-end sequencing was performed at the
UCLA Genomic Sequencing Center on the Illumina Genome Analyzer II for 76x cycles following
manufacturer’s protocol. The base-calling was performed by the real time analysis
software (Illumina).
Sequence Data Analysis
The Illumina output files (.qseq) were converted to fastq formats using BFAST
31,32
script ill2fastq.pl and then parsed into multiple fastq files, each for one unique
barcode. Only the reads with 100% matching barcode sequences were carried over to
the second set of fastq files. The sequence reads in each fastq file were aligned
to the human reference genome (hg18, March 2006, build 36.1) using Novoalign from
Novocraft Short Read Alignment Package. The output format was set to SAM and default
settings were used for all options. Using SAMtools, the SAM file of each sample was
converted to a BAM file, sorted and merged and potential PCR duplicates were removed
using Picard
33
. The variants, both single nucleotide variants (SNVs) and small INDELs (insertions
and deletions), within the captured coding exonic intervals were called using the
SAMtools pileup tool. For SNV calling, the last 5 bases were trimmed and only the
reads lacking INDELs were retained. For INDEL calling, only the reads that contained
one contiguous INDEL, not occurring on either end of the read, were retained
34
. The variants were further annotated using the SeqWare project and loaded into the
SeqWare QueryEngine database
35
. Variants from each sample with the following criteria were identified: (i) variant
base or INDEL observed at least twice and at ≥5% of the total coverage per base, (ii)
variant observed at least once on both forward and reverse strands, (iii) SNV quality
score ≥10. As IMAGE syndrome is a rare condition, only variants with coding consequences
not present in dbSNP129 were further analyzed.
CDKN1C Sequencing
All CDKN1C (RefSeq: NM_000076.2) mutations were PCRed using Phusion HF polymerase
(NEB) with 5% DMSO and 0.1M Betaine. PCR products were sent for dideoxysequencing
at Laragen, Inc. PCR primers used are listed in Supplementary Table 1. All mutation
locations were reported using CCDS7738.1 and P49918 as the normal transcript and protein
sequences, respectively.
Plasmid Constructs
The cDNA of CDKN1C cloned into pBluescript was purchased from ATCC(#99411)
36
. Mutations were generated using site-directed mutagenesis (Stratagene, primers in
Supplementary Table 1). 4 mutants were created, corresponding to BWS mutations (p.L42P
and c.826delT) and IMAGE syndrome mutations (p.F276V and p.K278E). Wild type and mutant
versions of CDKN1C were subcloned into pCDNA3.1(+) and pEGFP-C2 for mammalian cell
culture experiments, into pFLAG-pcDNA3.1 for the immunoprecipitation, and into pUAST
for the generation of transgenic flies.
Quantitative Reverse Transcription PCR
Human tissue from 7–8 weeks post-conception (wpc) was provided by the Medical Research
Council/Wellcome Trust-funded Human Developmental Biology Resource with Research Ethics
Committee approval and informed consent. RNA was extracted from adrenal, brain and
muscle samples using the TRIzol reagent and first-strand cDNA was generated using
SuperScriptR II reverse transcriptase (Invitrogen). Expression of CDKN1C transcript
was assessed by quantitative PCR using the StepOnePlus Real-time PCR System, TaqManR
Gene Expression Assays for human CDKN1C (Hs00175938_m1) and human GAPDH as endogenous
control (4333764T; all Applied Biosystems). Data were analyzed with StepOne software
v2.1 according to the 2-ΔΔCT method.
Immunofluorescence
Fourteen-micron sections of human fetal adrenal tissue (8 weeks post conception) were
fixed briefly in 4% PFA and blocked in 1% BSA before incubating overnight with antibody
to CDKN1C (Fisher AFMA121866) and CYP11A1 (Sigma HPA016436). Primary antibodies were
detected using Alexa647 goat anti-mouse (Invitrogen, A21235) and Alexa555 (Invitrogen,
A21429) goat anti-rabbit conjugates. Nuclei were counterstained with DAPI. IMAGEs
were collected on a Zeiss 710 confocal microscope (Carl Zeiss).
Immunoprecipitation
HEK293T cells were transfected with constructs encoding Flag-CDKN1C with Lipofectamine
2000 (Invitrogen). For ubiquitination assays, cells were co-transfected with pCI-neo-(HA)3-human
Ub
37
construct and treated 3 hours in 10μM MG-132 (Millipore) prior to cell lysis. Flag-CDKN1C
was immunoprecipitated from cell lysates using the ANTI-FLAG M2-Agarose Affinity Gel
(Sigma-Aldrich). Western blot was performed on immunoprecipitated samples and cell
lysates with primary antibody to HA (Covance MMS101R), Flag (Abcam, ab1162), PCNA
(Abcam ab29), and CDKN1C (Santa Cruz biotechnology, sc-1040). Secondary antibodies
used were goat-anti-mouse-HRP (Biorad, 1:20,000) and goat-anti-rabbit-HRP (Santa Cruz,
1:5,000).
Drosophila experiments
Five independent constructs, CDKN1CWT, CDKN1CL42P, CKDN1C826delT, CDKN1CF276V, and
CDKN1CK278E were injected into embryos and each construct generated multiple independent
transgenic lines. Overexpression was achieved by using the GAL4-UAS system
17
and the following drivers: Ubi-gal4 (ubiquitous expression), ey-gal4 (eye-specific
expression), MS1096-gal4 (wing-specific expression), and salPE-gal4 (wing pouch-specific
expression). MS1096-gal4 is expressed in the entire wing imaginal disc but at higher
levels in the dorsal compartment. This higher level of expression in the dorsal compartment
causes enhanced phenotypic effects in the dorsal versus the ventral side of the wing.
salPE-gal4 is specifically expressed in the pouch of the wing disc, which only gives
rise to the wing proper. All UAS-CDKN1C constructs were larval lethal when expressed
with Ubi-gal4, confirming their expression efficiency. At least two independent lines
were used for each experiment and all yielded similar results. All eye images were
taken on a Leica Z16 AP0 Camera. Wing IMAGEs were taken on Leica DFC 300 FX R2 Camera.
Flow Cytometry
Wild-type or mutant versions of CDKN1C were co-transfected with pCMV-GFP into serum
starved HEK293T cells using Lipofectamine 2000 (Invitrogen). After 24 hours, cells
were grown in media containing 5% serum for 48-hours. Cells were resuspended in a
hypotonic buffer with propidium iodide and GFP-positive events were analyzed on a
Becton Dickinson FACScan Analytic Flow Cytometer. All experiments were performed in
2 biological replicates. Statistical significance was assessed using a 2-proportion
z-test. Flow cytometry was performed in the UCLA Johansson Comprehensive Cancer Center
(JCCC) and Center for AIDS Research Flow Cytometry Core Facility.
Analysis of nuclear/cytoplasmic distribution of GFP-fusion proteins
H295R cells were transfected using Lipofectamine 2000. After incubation for 24 hours,
cells were fixed in 4% paraformaldehyde and IHC was performed using an anti-GFP antibody
(Invitrogen), a FITC-labeled secondary antibody (Jackson Lab), and a mounting medium
with DAPI (VectorLabs). Cells were imaged on an Olympus AX70 microscope.
M1 fibroblasts were transfected with 2ug/well of purified expression plasmid using
the X-tremeGENE HP DNA transfection reagent (Roche). 24-hours after transfection,
cells were fixed in 2% formaldehyde and analyzed using fluorescence microscopy. Fluorescent
and bright-field images of randomly selected fields containing transfected M1 cells
were saved electronically using a 'blind-code' file nomenclature. The distribution
of GFP-fusion proteins in each transfected cell was annotated by an experienced observer
(who was unaware of the blind code) as one of the following: (1) nuclear only, (2)
nuclear and cytoplasmic, and (3) cytoplasmic only.
Supplementary Material
1