1250 The Journal of Clinical Investigation http://www.jci.org Volume 115 Number 5 May 2005
A novel COL1A1 mutation in infantile cortical
hyperostosis (Caffey disease) expands the
spectrum of collagen-related disorders
Robert C. Gensure,1 Outi Mäkitie,2,3 Catherine Barclay,2 Catherine Chan,2
Steven R. DePalma,4 Murat Bastepe,1 Hilal Abuzahra,1 Richard Couper,5 Stefan Mundlos,6
David Sillence,7 Leena Ala Kokko,8 Jonathan G. Seidman,4 William G. Cole,2 and Harald Jüppner1
1Endocrine and Pediatric Endocrine Units, Departments of Medicine and Pediatrics, Massachusetts General Hospital and Harvard Medical School,
Boston, Massachusetts, USA. 2Division of Orthopaedics, The Hospital for Sick Children, Toronto, Ontario, Canada. 3Department of Pediatric Endocrinology,
Helsinki University Hospital, Helsinki, Finland. 4Department of Genetics, Harvard Medical School and Howard Hughes Medical Institute,
Boston, Massachusetts, USA. 5University of Adelaide, Department of Paediatrics, Women’s & Children’s Hospital, North Adelaide, South Australia, Australia.
6Max-Planck Institute for Molecular Genetics and Institute for Medical Genetics, Berlin, Germany. 7Departments of Paediatrics and Child Health,
The Children’s Hospital at Westmead Clinical School, Westmead, New South Wales, Australia. 8Collagen Research Unit,
Biocenter and Department of Medical Biochemistry and Molecular Biology, University of Oulu, Oulu, Finland.
Infantile cortical hyperostosis (Caffey disease) is characterized by spontaneous episodes of subperiosteal new
bone formation along 1 or more bones commencing within the first 5 months of life. A genome-wide screen
for genetic linkage in a large family with an autosomal dominant form of Caffey disease (ADC) revealed a locus
on chromosome 17q21 (LOD score, 6.78). Affected individuals and obligate carriers were heterozygous for a
missense mutation (3040C→T) in exon 41 of the gene encoding the α1(I) chain of type I collagen (COL1A1),
altering residue 836 (R836C) in the triple-helical domain of this chain. The same mutation was identified
in affected members of 2 unrelated, smaller families with ADC, but not in 2 prenatal cases and not in more
than 300 chromosomes from healthy individuals. Fibroblast cultures from an affected individual produced
abnormal disulfide-bonded dimeric α1(I) chains. Dermal collagen fibrils of the same individual were larger,
more variable in shape and size, and less densely packed than those in control samples. Individuals bearing
the mutation, whether they had experienced an episode of cortical hyperostosis or not, had joint hyperlaxity,
hyperextensible skin, and inguinal hernias resembling symptoms of a mild form of Ehlers-Danlos syndrome
type III. These findings extend the spectrum of COL1A1-related diseases to include a hyperostotic disorder.
Infantile cortical hyperostosis (Caffey disease; OMIM 114000) is
a genetic disorder characterized by an infantile episode of massive
subperiosteal new bone formation that typically involves the diaphy-
ses of the long bones, mandible, and clavicles (1). In addition to these
changes, which can appear quite prominently on x-ray, the involved
bones may also appear inflamed, with painful swelling and systemic
fever often accompanying the illness. Consistent with the observed
increased bone formation, laboratory findings include an elevated
level of alkaline phosphatase. Furthermore, there can be an eleva-
tion in white blood cell count and erythrocyte sedimentation rate,
which indicate an inflammatory response. The bone changes usually
begin before 5 months of age and resolve before 2 years of age (1, 2).
Recurrent episodes of cortical hyperostosis are uncommon, and
there are few reports concerning adverse sequelae of the hyperos-
totic lesions in affected individuals (3, 4). The inflammatory nature
of the symptoms has led to treatment trials with antiinflammatory
agents, and resolution of symptoms has been reported with the use
of indomethacin (5) and with glucocorticoid therapy (6).
Information on several unrelated pedigrees indicates that
Caffey disease can be inherited in an autosomal dominant man-
ner (7–9). Review of these pedigrees showed evidence of incom-
plete penetrance, as there were obligate carriers who lacked
episodes of cortical hyperostosis (7–9). However, the signs and
symptoms of Caffey disease can be subtle and are typically only
manifested at a preverbal age, which thus makes it plausible
that the diagnosis could have been missed in some of these
apparently unaffected carriers. A more severe and often lethal
prenatal form of Caffey disease appears to be inherited as an
autosomal recessive disorder (10).
In addition to the inherited forms of Caffey disease, sporadic
cases of infantile cortical hyperostosis have also been described
(reviewed in ref. 11). At least some of these nonfamilial cases can
be attributed to administration of prostaglandin E1 (PGE1) and
PGE2 for treatment of ductal-dependent cardiac lesions (12–14).
In fact, the incidence of this complication of prostaglandin
therapy is high, with radiographic changes occurring in 62% of
all infants receiving PGE1 infusion for more than 60 days (14).
Other conditions that can cause cortical bone lesions in children
less than 2 years of age, and may thus mimic infantile cortical
hyperostosis, include trauma (15), hypervitaminosis A (16), hyper-
phosphatemia (17), and infection (18). Sporadic cases of infantile
cortical hyperostosis not associated with PGE1 or PGE2 infusion
appear to be on the decline, but the reason(s) for the diminished
frequency of these bone abnormalities remain(s) uncertain (8).
Nonstandard abbreviations used: ADC, autosomal dominant form of Caffey
disease; COL1A1, gene encoding the α1(I) chain of type I collagen; CB6, cyanogen bro-
mide peptide 6; PGE1, prostaglandin E1.
Conflict of interest: The authors have declared that no conflict of interest exists.
Citation for this article: J. Clin. Invest. 115:1250–1257 (2005).
Related Commentary, page 1142
The Journal of Clinical Investigation http://www.jci.org Volume 115 Number 5 May 2005
The gene or genes responsible for the prenatal and postnatal
inherited forms of Caffey disease have not been reported. We there-
fore undertook genetic linkage studies using genomic DNA from
a large kindred with the autosomal dominant form of the condi-
tion (autosomal dominant Caffey disease [ADC]) (7, 8). Affected
individuals and obligate carriers in this kindred, as well as affected
individuals in 2 other kindreds, were shown to be heterozygous for
the same novel missense mutation affecting the gene that encodes
the α1(I) chain of type I collagen (COL1A1).
Genetic linkage analysis. Three unrelated families with at least 2
members affected by ADC (7, 8, 19) (Figure 1), 1 sporadic case,
and 2 prenatal cases were available for analysis (20, 21). DNA
from family 1 was used for a genome-wide scan, which revealed
linkage to chromosome 17q21 (Figure 2A). Fine mapping
reduced the size of the linked region to a 2.3-Mb interval delim-
ited by markers D17S1868 and D17S1877. The maximal LOD
score obtained within this region by 2-point analysis was 6.78
for marker D17S1795 (θ = 0), which is only slightly lower than
the maximal theoretical LOD score of 7.12. Nucleotide sequence
analysis of 13 candidate genes within the linked region (GNGT2,
NESH, PHB, LOC81558, ITGA3, LOC84687, XT2, PRO1855,
CHAD, EPN3, FLJ21347, OA48-18, and TOB1) did not reveal any
heterozygous mutations that segregated with ADC. However,
affected individuals and obligate carriers were heterozygous
for a missense mutation in exon 41 of COL1A1 (also referred
to as exon 42, corresponding to the exon structure of COL1A2,
which contains 1 additional exon; ref. 22) (Figure 3). This C→T
transition at nucleotide position 3040 of the cDNA (relative to
the translation initiation codon), which alters codon 1014 of
prepro-α1(I) mRNA, was predicted to produce an R836C amino
acid substitution within the triple-helical domain of the α1(I)
chain of collagen. The nucleotides were numbered relative to the
ATG start site of prepro-α1(I) cDNA, and amino acid residues
were numbered relative to the first glycine residue of the main
triple-helical domain of the α1(I) protein chain (22).
To confirm the mutation identified by sequence analysis, a
495-bp fragment comprising exon 41 was PCR amplified and
digested with HpyCH4IV (ACGT). The PCR product derived from
the mutant allele lacked 1 recognition site for HpyCH4IV, thus
generating an additional DNA fragment of 319 bp (Figure 2B).
The same COL1A1 mutation was also found in the 3 affected
members but not in the unaffected father of family 2 (Figure 2C).
Furthermore, the mutation was found in the clinically affected
identical twins but not in the clinically unaffected parents and
brother of family 3 (Figure 2D). The affected twins and the unaf-
fected sibling had inherited identical parental alleles in this
region. Consequently, the clinically affected twins were likely to
have acquired a de novo COL1A1 mutation.
Haplotype analysis of the affected members of families 1–3 did
not reveal any shared alleles, which supported the historical evi-
dence that the 3 families that were heterozygous for the R836C
substitution were not related to each other. The 3040C→T
transition was not detected in genomic DNA from more than 150
healthy controls (more than 300 chromosomes), in SNP databas-
es (dbSNP, http://www.ncbi.nlm.nih.gov/SNP; JSNP, http://snp.
ims.u-tokyo.ac.jp; Celera Discovery System, http://publication.
celera.com/humanpub/index.jsp; and Applied Biosystems
SNPbrowser Software, http://marketing.appliedbiosystems.
com/mk/get/snpb_landing), or in the Mutations in COL1A1
(22). The latter findings suggest that the 3040C→T transition
in exon 41 of COL1A1 and the resulting R836C amino acid sub-
stitution within the triple-helical domain of the α1(I) chain can
be a cause of ADC.
A possible sporadic case of Caffey disease with hyperostosis of a
rib did not have any demonstrable mutations in COL1A1 or in the
COL1A2 gene, which encodes the α2(I) chain of type I collagen.
Mutations of COL1A1 or COL1A2 were also not detected in 2 cases
of prenatal Caffey disease (20, 21).
Dermal collagen morphology and collagen biosynthesis. To further
investigate whether the Col1A1 mutation results in changes in
type I collagen, we obtained skin biopsies from an affected adult
female in family 1 (IV-2). This individual had an episode of infan-
tile cortical hyperostosis and from early childhood had clinical
signs of hyperextensible skin and generalized joint instability, as
well as a history of several fractures. As an adult, she had the
same clinical signs, as well as more severe voluntary subluxation
of the shoulders. Electron microscopy of her dermis showed
normal fibroblast morphology (data not shown). The collagen
fibrils were more variable in shape and size and were less densely
Radiographic features of infantile cortical hyperostosis. Radiographs
of 3 affected individuals of 2 unrelated kindreds showing subperiosteal
thickening of the femur, tibia, and fibula (left and upper middle panels:
tibia and fibula, respectively, of twin II-1 of family 3; upper right panel:
fibula of patient II-1 of family 2), and metacarpals of the left foot (lower
right panel: twin II-2 of family 3). Note that the periosteum, which is
normally anchored at the growth plates where it is continuous with
the perichondrium, was frequently elevated circumferentially from the
proximal to the distal growth plates (open arrows). The bone marrow
cavities were also narrowed (filled arrows).
1252 The Journal of Clinical Investigation http://www.jci.org Volume 115 Number 5 May 2005
packed (Figure 4A) than fibrils in age- and sex-matched control
dermal samples (Figure 4B). The fibrils were also significantly
larger and were more variable in size in the affected individu-
al (Caffey disease, 108 ± 15 nm; control, 91 ± 8 nm; n = 1,000;
P < 0.001). Granular material was visible in the matrix surround-
ing many of the collagen fibrils (Figure 4A).
One-dimensional gel electrophoresis showed that cultured der-
mal fibroblasts from the proband of family 1 produced a mix-
ture of apparently normal type I, III, and V collagens as well as
an abnormal disulfide-linked dimer (Figure 5A). As the dimer
was likely to include 2 α1(I) chains bearing the R836C substitu-
tion, it was designated β11′. Two-dimensional gel electrophoresis
showed that the abnormal β11′ dimers dissociated after reduc-
tion of disulfide bonds into α1(I) chains that were designated
as α1(I)′ chains (Figure 5B). Although β-aminopropionitrile,
an inhibitor of the formation of lysine-derived collagen cross-
linkages, was added to the cultures, a small amount of dimeric
collagen chains (β11 and β12) was formed. These lysine-derived
crosslinked dimers do not dissociate into α chains after reduc-
tion of disulfide bonds with dithiothreitol.
Genotype/phenotype correlations. Of the 24 members of family 1 in
our study with the R836C mutation, only 19 members had expe-
rienced an episode of cortical hyperostosis, while 5 were obligate
carriers. Thus, 79% of the individuals who were heterozygous for
the COL1A1 mutation had an episode of cortical hyperostosis,
and 21% of them did not. Detailed interviews of family members
did not reveal any precipitating events other than the consistent
report that the onset of cortical hyperostosis usually occurred
before 5 months of age. None of the children had been given pros-
taglandins, which can induce subperiosteal new bone formation in
neonates (23). The clinical course of the hyperostotic episode was
often protracted but resolved after about 2 years. By mid- to late
childhood, the hyperostotic long bones in the affected children
of families 2 and 3 had remodeled extensively. However, in each
affected bone, the central portion of the diaphysis, correspond-
ing to the complete diaphysis of infancy, showed mild external
and internal thickening of the cortex. In contrast, the diaphyseal
and metaphyseal bone above and below the central region, which
would have formed after the episode of hyperostosis, was radio-
graphically normal (data not shown). The epiphyses and growth
plates were also radiographically normal.
In family 1, individuals with the R836C substitution (affected
individuals and obligate carriers) had varying degrees of joint
hyperlaxity that was first observed in early childhood (24). The
hyperlaxity was apparent in large and small joints and allowed
some individuals bearing the Caffey allele to voluntarily sublux
Haplotype analyses of families with infantile cortical hyperostosis. (A) Family 1: Fine mapping of the locus for Caffey disease on chromosome 17q21.
Standard techniques were used to establish the genetic locus of the disease in a large Canadian family with ADC (7, 8). Black symbols, affected
individuals; white symbols with black dot in the center, obligate carriers; white symbols, unaffected individuals; gray symbols, deceased individuals.
The unaffected individual II-9, who has unaffected children and grandchildren (data not shown), was only included to deduce the haplotypes of I-1
and I-2 (indicated by italics); for LOD score calculations, his phenotype was entered as unknown. Marker D17S1795 provided a LOD score of 6.78
(maximal theoretical LOD score, 7.12). Markers D17S1868 and D17S1877 (indicated in white on a black background) define the centromeric and
telomeric boundary, respectively. The disease-associated haplotype is shown by black numbers on gray; markers consistent with a recombination
are shown by white numbers on black. Uninformative data and haplotypes not associated with the disease are shown by black numbers on white.
A C→T mutation at nucleotide 3040 of COL1A1 (see Figure 3) was identified by direct nucleotide sequence analysis only in affected members and
obligate carriers. (B) Haplotypes of portions of the family shown in A and PCR-based confirmation of the identified mutation (see Methods). (C)
Haplotypes and PCR analysis for family 2, comprising 2 affected brothers, their affected mother, and their healthy father. (D) Haplotypes and PCR
analysis for family 3, comprising identical twin sisters affected by Caffey disease and their healthy parents and brother (19).
The Journal of Clinical Investigation http://www.jci.org Volume 115 Number 5 May 2005
their shoulders and finger joints (Figure 6). Instability of the
joints, particularly the shoulder joints, worsened with age. These
individuals had abnormally soft and hyperextensible skin, which
was otherwise normal in appearance. They did not bruise easily.
Members of family 1 who did not bear the Caffey allele did not
have the skin or joint anomalies. The 3 members of family 2 with
the R836C mutation (see Figure 2C) also had hyperextensible skin,
joint hyperlaxity, and inguinal hernias, while the identical twins in
family 3 did not demonstrate such anomalies.
None of the affected individuals or obligate carriers in any of
the families had clinical signs of the major type I collagen disor-
der, osteogenesis imperfecta (25). In particular, they lacked gray-
blue sclerae, dentinogenesis imperfecta, premature hearing loss,
short stature, deformities, and scoliosis (25, 26). Skull radiographs
from the clinically affected identical twins did not contain worm-
ian bones, which can occur in patients with osteogenesis imper-
fecta (25, 26). Neither of the twins had radiographic evidence of
platyspondyly, which is also a common feature of osteogenesis
imperfecta. In family 1, 9 of the 18 individuals with Caffey disease,
for whom detailed fracture histories were available, had a total of
14 peripheral fractures (1–4 in each). Bone mineral density was
assessed in 1 affected adult with Caffey disease from family 1 and
the results were within normal limits for the lumbar spine and
femoral neck; spinal radiographs in this patient showed no evi-
dence of platyspondyly or compression fractures.
In family 1, we mapped ADC to an interval on chromosome 17q21.
All affected individuals and obligate carriers were heterozygous for
a 3040C→T transition in a CpG dinucleotide of exon 41 of COL1A1,
which encodes the α1(I) chain of type I collagen. The same 3040C→T
transition was also present in affected members of families 2 and
3. The recurrent nature of the mutation indicates that the involved
CpG dinucleotide is a mutational “hot spot” in COL1A1 (27–29).
The mutation is predicted to introduce an R836C substitution
into the triple-helical domain of α1(I) chains of type I collagen (22).
Results from collagen biosynthetic studies supported this propos-
al, as cultured dermal fibroblasts from an affected individual pro-
duced an abnormal disulfide-bonded α1(I) dimer. Ultrastructural
studies of the dermis revealed that in vivo, the mutation was associ-
ated with an abnormal collagen fibril architecture.
The absence of the nucleotide change in any of the databases of
normal DNA sequence variants as well as in normal DNA samples
supports our proposal that it is a cause of ADC. While the R836C
mutation therefore appears to be required for the development of
ADC, it remains uncertain how the identified amino acid change
can result in a spatially and temporally limited increase in cortical
bone formation. COL1A1 and COL1A2 mutations were not found
in a sporadic case of Caffey disease in which, according to radio-
graphy, changes were limited to a single rib, which suggests that
other genetic defects or environmental factors can lead to localized
hyperostosis. Mutations of the type I collagen genes were also not
detected in 2 cases of the prenatal form of the disease, which is
usually more severe and may be inherited in an autosomal reces-
sive manner (21). Consequently, it is likely that the prenatal form
of Caffey disease constitutes a separate disorder (11).
The R836C amino acid substitution changes the X posi-
tion of one of the 338 Gly-X-Y repeating amino acid triplets of
the triple-helical domain of the α1(I) chain of type I collagen.
Nucleotide sequence analysis of portions of exon 41 of COL1A1 and
adjacent intronic regions of an affected individual from family 1. The
individual was heterozygous for the transition, which is indicated by the
arrow, and the approximate location of both primers for PCR amplifica-
tion is indicated by arrowheads within the schematic, partial drawing
of COL1A1. Partial nucleotide sequence of wild-type and mutant (mut)
exon 41 (capital letters), as well as adjacent intronic sequence (lower-
case letters), along with the encoded amino acid sequence, are shown.
The 3040C→T transition (indicated by an asterisk), which alters the
first nucleotide of the second-to-last codon in exon 41, is predicted to
result in the substitution of an arginine (R) 836 to a cysteine (C) residue
(R836C). The approximate location of the R836C mutation within the
triple-helical region of the collagen fibril is schematically shown [bot-
tom: thin lines, α1(I) chains; thick line, α2(I) chain], as is the location of
the R134C mutation previously described in 2 unrelated patients with
Ehlers-Danlos syndrome type I (31).
Dermal ultrastructure from proband IV-2 of family 1. (A) Ultrastructure of
the extracellular matrix of the proband’s dermis (magnification, ×60,000).
The collagen fibrils are more variable in shape and size and are less
densely packed than in control samples. The fibrils are also larger
(Caffey disease, 108 ± 15 nm; control, 91 ± 8 nm; n = 1,000; P < 0.0001).
Granular material is visible in the matrix surrounding the collagen fibrils.
(B) Ultrastructure of control dermis. In contrast to the proband’s dermis,
the collagen fibrils in the control dermis are round, uniform in size, tightly
packed, and are not surrounded by granular material.
1254 The Journal of Clinical Investigation http://www.jci.org Volume 115 Number 5 May 2005
Cysteine residues are normally present in the amino and carboxyl
propeptides of type I procollagen chains, but they are removed
when type I procollagen is enzymatically processed to type I col-
lagen (30). While many G→C mutations have been described in
osteogenesis imperfecta, only 1 other R→C amino acid substitu-
tion has been reported in the α1(I) chain of type I collagen. Two
unrelated patients with classical Ehlers-Danlos syndrome type 1
were heterozygous for a 934C→T transition at a CpG dinucleo-
tide in exon 14 of COL1A1 (31). The resulting amino acid substi-
tution, R134C, also involved the X position of a Gly-X-Y triplet
within the main triple-helical domain of the α1(I) chain of type I
collagen (Figure 3). Ultrastructural studies of skin biopsies from
these patients revealed variability in diameter and irregularity of
dermal collagen fibrils, with granulofilamentous material visible
along the collagen fibrils (similar to what is shown in Figure 4A),
and collagen biosynthetic studies using fibroblast cultures bear-
ing the R134C substitution also showed disulfide-bonded α1(I)
dimers within the fibroblasts. Thus, the ultrastructural and
biosynthetic anomalies produced by the R134C and the R836C
substitutions were similar. However, in contrast to the R134C
substitution that was associated with a classical Ehlers-Danlos
syndrome type I phenotype (soft, velvety, and hyperextensible
skin; numerous atrophic paper scars of the skin; ecchymoses on
the legs; and generalized joint hyperlaxity), the present R836C
Collagen biosynthesis by cultured fibroblasts from proband IV-2 of fam-
ily 1. (A) One-dimensional gel electrophoresis of dermal and fibroblast
collagens: lane 1, pepsin-solubilized collagen from normal dermis (ND);
lane 2, proband fibroblast cell layer collagens; lane 3, proband medium
collagens; lane 4, reduced proband fibroblast cell layer collagens; lane 5,
reduced proband medium collagens; lane 6, control fibroblast cell layer
collagens; lane 7 , control medium collagens. All samples contained α1(I)
and α2(I) monomeric chains of type I collagen. Control dermis (lane 1)
contained α1(I) dimers (β11) and α1(I)/α2(I) dimers (β12) with lysine-
derived cross-linkages; these cross-linkages were partially blocked with
the addition of β-aminopropionitrile in all fibroblast cultures. The unre-
duced dermal and fibroblast culture samples contained disulfide-bond-
ed type III collagen trimers [α1(III)3]. There was an additional protein
band in the unreduced proband samples (arrowheads, lanes 2 and 3),
designated β11′, which was more abundant in the cell layer than in the
medium. This band migrated slightly slower than the dermal β11 dimer.
The abnormal band disappeared, along with the type III collagen trimer,
after reduction of disulfide bonds with DTT. (B) Two-dimensional gel
electrophoresis of the proband’s fibroblast cell layer collagens. Disulfide
bonds were unreduced in the first dimension and reduced with DTT in
the second dimension. The abnormal protein band in A, lane 2, was
dissociated by DTT into proteins, designated α1(I)′, which migrated in a
similar manner to control α1(I) chains.
Voluntary subluxation of various joints in ADC patients. Photographs of 4 individuals carrying the R836C mutation (III-13, IV-2, IV-3, and IV-4
of family 1): upper left, 2 individuals with subluxation of the shoulders; lower left, 2 views of patellar subluxation; right: views of hyperextension
and subluxation in the fingers and the elbow.
The Journal of Clinical Investigation http://www.jci.org Volume 115 Number 5 May 2005
substitution was associated with episodes of infantile cortical
hyperostosis, and 2 of the 3 Caffey families showed features of
Ehlers-Danlos syndrome type III (24). Interestingly, neither R→C
mutation was associated with osteogenesis imperfecta.
There are several possible mechanisms by which the R836C
mutation could result in the observed ADC phenotype. The
analysis of collagen fibrils in skin fibroblasts showed increased
disulfide crosslinking, either within or between mutant collagen
fibrils, which may be responsible for the observed alterations
in collagen architecture. Disulfide crosslinking may also occur
between the mutant collagens and other cystein-containing pro-
teins in the bone matrix. The Caffey mutation may furthermore
disrupt an important site of interaction with other peptides or
proteins. Arg836 is located within the carboxyterminal cyano-
gen bromide peptide 6 (CB6) of the α1(I) chain, which has been
shown to bind with high affinity to IL-2 and to the amyloid pro-
tein precursor (APP) (32–34). However, the CB6 peptide compris-
es 218 amino acid residues, i.e. approximately 21% of the α1(I)
chain (22), and the specific binding sites for IL-2 and APP within
the CB6 peptide are currently unknown. It also is plausible that
the R836C substitution within the α1(I) chain reduces the ther-
mal stability of the collagen triple helix. Based on studies with
a host-guest collagen peptide, Ac-(Gly-Pro-Hyp)3-Gly-X-Y-(Gly-
Pro-Hyp)4-Gly-Gly-CONH2, in which an arginine residue had
been replaced with a cysteine residue in the X position (30, 35),
the R836C mutation identified in Caffey patients would be pre-
dicted to decrease the melting temperature of the collagen tri-
ple helix by 4.5°C. This change could provide another possible
explanation for the observed alterations in collagen architec-
ture. Last, prostaglandins may be involved in the development
of hyperostosis, as short-term PGE1 or PGE2 administration for
the treatment of ductal-dependent cardiac lesions typically leads
to bone changes that resemble those present in ADC (12–14, 36).
The inflammatory signs that often accompany the hyperostot-
ic lesions in Caffey disease support this hypothesis. However,
examination of additional ADC families, as well as in vitro and
in vivo studies, are needed to help define the mechanism(s) that
lead to the disease phenotype.
While the hyperostotic lesions in ADC occur only in infancy,
the mutant α1(I) chain is presumably present in the bone matrix
throughout life. Based on our radiographic observations, detach-
ment of the periosteum from the underlying bone appears to be
a key pathological process in the onset of episodes of cortical
hyperostosis. However, there was no consistent reporting of
events in our families that may have precipitated this periosteal
detachment. The only recurrent finding was the onset of episodes
of cortical hyperostosis within the first 5 months of postnatal
life, which suggests that infants in this age group may have an
increased susceptibility to periosteal injury. Consistent with this
hypothesis, a radiographic study of babies who died of sudden
infant death syndrome between 1 and 4 months of age showed
that mild subperiosteal new bone formation, which was distinct
from the underlying diaphysis and metaphysis, had occurred
in the long bones of 35% of the cases (37). Furthermore, large
amounts of subperiosteal new bone characteristically accompany
healing fractures (38), osteomyelitis (39), and scurvy (40, 41) in
neonates and infants. In all of these conditions, the bone can
enlarge in diameter by several fold but remodels over the follow-
ing 2 years, similar to what occurs with the hyperostotic lesions
in Caffey disease. In contrast, subperiosteal elevation is more lim-
ited in older children with fractures and osteomyelitis (42). Thus,
intrinsic differences in periosteal bone formation in infants and
adults may explain the absence of hyperostotic lesions in adults
with the R836C mutation. Furthermore, differences in the clini-
cal threshold for periosteal detachment, influenced by the genet-
ic background and/or environmental exposures, may explain the
incomplete penetrance of the hyperostotic phenotype. Similar
explanations have been provided to account for the widely dif-
fering fracture rates among affected individuals in families with
classical osteogenesis imperfecta type I due to autosomal domi-
nant COL1A1 haploinsufficiency (26, 43, 44).
Our clinical evaluations of the affected individuals in the 3 fami-
lies have extended the reported phenotype of Caffey disease (1).
Individuals bearing the Caffey allele in 2 of the 3 families had fea-
tures that were similar to those observed in patients with Ehlers-
Danlos syndrome type III (24), and the absence of such features
in the third family may be attributed to the current young age of
the affected individuals. The relatively high number of fractures in
family 1 suggested that bone fragility may also be part of the phe-
notype. However, bone densitometry studies in an affected adult
patient (IV-2) were found to be normal, and similar fracture rates
can occur in the normal population (45). Nonetheless, bone health
of the ADC families requires further study, as radiographs taken
during the episodes of cortical hyperostosis show that abnormali-
ties in cortical and endosteal bone accompany the more obvious
formation of subperiosteal new bone.
In summary, ADC is associated with a mutation in COL1A1 and
thus belongs to the family of type I collagen–related conditions
that include osteogenesis imperfecta types I–IV (46); Ehlers-Dan-
los syndrome types I (31) and VII (47, 48) and a recessive cardiac
valvular form (49); idiopathic osteoporosis (50); arterial dissec-
tions (51); and dermatofibrosarcoma protuberans, which was
found to be associated with translocations between chromosomes
17 and 22 leading to the production of a fusion protein consisting
of α1(I) collagen and PDGF-β (52). The findings in the present
study extend the phenotypic spectrum of disorders associated with
genetic changes in the type I collagen genes to include a hyperos-
totic disorder, but it remains uncertain how the R836C mutation
results in a temporally and spatially limited defect in bone.
Families with ADC. A large, 3-generation Canadian ADC family (family 1)
was ascertained, and historical details concerning episodes of infantile cor-
tical hyperostosis were obtained from family members. Previous reports
on this family also included detailed accounts of all episodes of possible
infantile cortical hyperostosis (7, 8). These sources of information were
combined in order to validate disease assignments within the pedigree. For
the purposes of genetic linkage analysis, individuals in this family were
designated as affected if they had radiographically proven Caffey disease
(affected members) or if they had children diagnosed with Caffey disease
(obligate carriers). Genomic DNA from 24 such individuals was available
for study. Unrelated spouses were designated as unaffected.
Besides family 1, two additional small kindreds (families 2 and 3) were
ascertained, which had 3 and 2 affected members, respectively. The small
Canadian kindred (family 2) included the mother and her 2 children (see
Figure 2C), who were clinically and radiographically affected by Caffey dis-
ease (see Figure 1), while the small Australian family (family 3) consisted
of affected monozygotic twins, a healthy brother, and healthy parents (see
Figure 2D) (19). A single sporadic case had hyperostosis of a rib, which was
possibly due to ADC. Furthermore, 2 sporadic cases of the more severe
1256 The Journal of Clinical Investigation http://www.jci.org Volume 115 Number 5 May 2005
prenatal form of Caffey disease were studied (20, 21). Autosomal recessive
inheritance was postulated for the prenatal cases (21).
Detailed information was obtained from all families concerning possible
events that may have precipitated episodes of infantile cortical hyperosto-
sis. These enquiries were undertaken soon after the onset of the episodes
of cortical hyperostosis in the affected children in families 2 and 3 as well
as in some of the recently born affected members of the large family 1. The
long-term sequelae of the condition in clinically affected individuals and in
obligate carriers were determined. The long-term clinical features of the indi-
viduals in each family that did not bear the Caffey allele were also recorded.
Genetic screening and DNA analysis. In family 1, we performed a genome-wide
scan with microsatellite markers spaced at about 10-centimorgan intervals
(Webber set 9; Invitrogen Corp.) followed by fine mapping using commer-
cially available markers for the linked interval. We performed LOD score
calculations using the Mlink program (http://linkage.rockefeller.edu/)
assuming a disease frequency of 0.1%, equal allele frequencies, and
autosomal dominant inheritance with full penetrance. Candidate genes
in the linked region were identified and sequenced by the Massachusetts
General Hospital Sequencing Core using Applied Biosystems Taq DyeDe-
oxy Terminator cycle sequencing kits and methods.
Once the putative Caffey gene and its mutation were identified, we
screened genomic DNA from affected individuals and obligate carriers of
the various families, as well as the sporadic cases and controls, by restric-
tion mapping of PCR products. For these studies, we used the forward
primer AGCAGGGGAATATGGGTCAG and the reverse primer GGCCCT-
GAGAAAAACCATC to generate a 495-bp PCR product. The mutant PCR
product lacked 1 of the cleavage sites for the restriction endonuclease
HpyCH4IV, which thus allowed detection of the mutation by agarose gel
electrophoresis after enzymatic digestion of the amplified DNA (wild-
type, bands of 259, 60, and 176 bp; mutant, bands of 319 and 176 bp; note
that the 60-bp band is not shown in Figure 2B). In individuals who did
not bear the R836C mutation in COL1A1, mutations were sought in other
parts of COL1A1 as well as in COL1A2, as described previously (53).
Informed consent was obtained from all study participants prior to their
inclusion in these studies. These studies have been approved by the institu-
tional review boards at the respective institutions and were performed in
accordance with the guidelines of the Declaration of Helsinki.
Dermal histology and fibroblast cultures. Skin biopsies were obtained
from the proband (IV-2) of family 1. We processed 1 biopsy for elec-
tron microscopy using previously described methods (54). The same
methods were used to process the patient and control samples for elec-
tron microscopy. The second biopsy specimen was used to establish
fibroblast cultures, which were grown to confluency in the presence
of β-aminopropionitrile, which blocks the formation of lysine-derived
cross-linkages, and ascorbate, which is required for normal prolyl- and
lysyl-hydroxylation of collagen. The cultures were analyzed 5–7 days
after plating, as previously described (55, 56). Fibroblast cultures from
an age- and sex-matched individuals served as controls. Cell layer and
medium fractions were analyzed separately, and the procollagens were
converted to collagen by limited pepsin digestion. The collagens were
resolved by 1-dimensional and 2-dimensional SDS-PAGE and stained
with Coomassie G-250 (Invitrogen Corp.) (28). Where applicable, disul-
fide bonds were reduced with DTT.
We thank the affected individuals and their families for their
assistance. We also thank C.S. Houston for his assistance with the
radiology of Caffey disease. This work was supported by grants
from the Canadian Institutes of Health Research and the Cana-
dian Arthritis Network (to W.G. Cole); grants from the Founda-
tion for Pediatric Research and the Paivikki and Sakari Sohlberg
Foundation and a Research Fellowship from the European Society
for Paediatric Endocrinology, sponsored by Novo Nordisk A/S (to
O. Mäkitie); and NIH grants R01 46718-10 (to H. Jüppner) and
K08 HD41512 (to R.C. Gensure).
Received for publication July 19, 2004, and accepted in revised
form February 15, 2005.
Address correspondence to: Harald Jüppner, Endocrine and Pedi-
atric Endocrine Units, Departments of Medicine and Pediatrics,
Massachusetts General Hospital and Harvard Medical School,
Boston, Massachusetts, USA. Phone: (617) 726-3966; Fax: (617)
726-7543; E-mail: firstname.lastname@example.org.
1. Caffey, J. 1957. Infantile cortical hyperostosis; a
review of the clinical and radiographic features.
Proc. R. Soc. Med. 50:347–354.
2. Bernstein, R.M., and Zaleske, D.J. 1995. Familial
aspects of Caffey’s disease. Am. J. Orthop. 24:777–781.
3. Caffey, J. 1952. On some late skeletal changes in
chronic infantile cortical hyperostosis. Radiology.
4. Blank, E. 1975. Recurrent Caffey’s cortical hyperosto-
sis and persistent deformity. Pediatrics. 55:856–860.
5. Heyman, E., Laver, J., and Beer, S. 1982. Prostaglan-
din synthetase inhibitor in Caffey disease. J. Pediatr.
6. Bush, L., and Merrell, O. 1952. Infantile cortical
hyperostosis. Report of a case responding to treat-
ment with corticotropin. J. Pediatr. 40:330–333.
7. Gerrard, J.W., Holman, G.H., Gorman, A.A., and
Morrow, I.H. 1961. Familial infantile cortical
hyperostosis. J. Pediatr. 59:543–548.
8. Maclachlan, A.K., Gerrard, J.W., Houston, C.S., and
Ives, E.J. 1984. Familial infantile cortical hyperos-
tosis in a large Canadian family. Can. Med. Assoc. J.
9. Van Buskirk, F.W., Tampas, J.P., and Peterson, O.S.,
Jr. 1961. Infantile cortical hyperostosis; an inquiry
into its familial aspects. Am. J. Roentgenol. Radium
Ther. Nucl. Med. 85:613–632.
10. Barba, W.P., and Freriks, D.J. 1953. The familial
occurrence of infantile cortical hyperostosis in
utero. J. Pediatr. 42:141–150.
11. Saul, R.A., Lee, W.H., and Stevenson, R.E. 1982.
Caffey’s disease revisited. Further evidence for
autosomal dominant inheritance with incomplete
penetrance. Am. J. Dis. Child. 136:55–60.
12. Parker, S., and Griffiths, H. 1990. Prostaglan-
din-induced cortical hyperostosis. Orthopedics.
13. Kaufman, M., and El-Chaar, G. 1996. Bone and
tissue changes following prostaglandin therapy in
neonates. Ann. Pharmacother. 30:269–274, 277.
14. Woo, K., Emery, J., and Peabody, J. 1994. Cortical
hyperostosis: a complication of prolonged prosta-
glandin infusion in infants awaiting cardiac trans-
plantation. Pediatrics. 93:417–420.
15. Snedecor, S., Knapp, R., and Wilson, H. 1935. Trau-
matic ossifying periostitis of the newborn. Surg.
Gynecol. Obstet. 61:385–387.
16. Rothman, P., and Leon, E. 1948. Hypervitamin-
osis A. Report of two cases in infants. Radiology.
17. Talab, Y., and Mallouh, A. 1988. Hyperostosis with
hyperphosphatemia: a case report and review of
the literature. J. Pediatr. 114:1010–1013.
18. Caffey, J. 1946. Infantile cortical hyperostosis.
J. Pediatr. 29:541–559.
19. Couper, R.T., McPhee, A., and Morris, L. 2001.
Indomethacin treatment of infantile cortical perios-
tosis in twins. J. Paediatr. Child Health. 37:305–308.
20. Dahlstrom, J.E., et al. 2001. Lethal prenatal onset
infantile cortical hyperostosis (Caffey disease).
21. Schweiger, S., et al. 2003. Antenatal onset of corti-
cal hyperostosis (Caffey disease): case report and
review. Am. J. Med. Genet. 120A:547–552.
22. Dalgleish, R. 1997. The human type I collagen
mutation database. Nucleic Acids Res. 25:181–187.
23. Drvaric, D.M., et al. 1989. Prostaglandin-induced
hyperostosis. A case report. Clin. Orthop. 246:300–304.
24. Beighton, P., De Paepe, A., Steinmann, B., Tsipou-
ras, P., and Wenstrup, R.J. 1998. Ehlers-Danlos
syndromes: revised nosology, Villefranche, 1997.
Ehlers-Danlos National Foundation (USA) and
Ehlers-Danlos Support Group (UK). Am. J. Med.
25. Byers, P.H., and Steiner, R.D. 1992. Osteogenesis
imperfecta. Annu. Rev. Med. 43:269–282.
26. Cole, W.G. 2002. Advances in osteogenesis imper-
fecta. Clin. Orthop. 401:6–16.
27. Chan, D., Rogers, J.F., Bateman, J.F., and Cole, W.G.
1995. Recurrent substitutions of arginine 789 by
cysteine in pro-alpha 1 (II) collagen chains produce
spondyloepiphyseal dysplasia congenita. J. Rheuma-
tol. Suppl. 43:37–38.
28. Chan, D., Taylor, T.K., and Cole, W.G. 1993. Char-
acterization of an arginine 789 to cysteine substi-
tution in alpha 1 (II) collagen chains of a patient
with spondyloepiphyseal dysplasia. J. Biol. Chem.
The Journal of Clinical Investigation http://www.jci.org Volume 115 Number 5 May 2005
29. Tomso, D.J., and Bell, D.A. 2003. Sequence con-
text at human single nucleotide polymorphisms:
overrepresentation of CpG dinucleotide at poly-
morphic sites and suppression of variation in CpG
islands. J. Mol. Biol. 327:303–308.
30. Persikov, A.V., Ramshaw, J.A., Kirkpatrick, A., and
Brodsky, B. 2000. Amino acid propensities for the
collagen triple-helix. Biochemistry. 39:14960–14967.
31. Nuytinck, L., et al. 2000. Classical Ehlers-Danlos
syndrome caused by a mutation in type I collagen.
Am. J. Hum. Genet. 66:1398–1402.
32. Beher, D., Hesse, L., Masters, C.L., and Multhaup,
G. 1996. Regulation of amyloid protein precur-
sor (APP) binding to collagen and mapping of the
binding sites on APP and collagen type I. J. Biol.
33. Somasundaram, R., et al. 2000. Collagens serve as
an extracellular store of bioactive interleukin 2.
J. Biol. Chem. 275:38170–38175.
34. Di Lullo, G.A., Sweeney, S.M., Korkko, J., Ala-Kokko,
L., and San Antonio, J.D. 2002. Mapping the ligand-
binding sites and disease-associated mutations on
the most abundant protein in the human, type I
collagen. J. Biol. Chem. 277:4223–4231.
35. Persikov, A.V., Ramshaw, J.A., and Brodsky, B. 2000.
Collagen model peptides: sequence dependence of
triple-helix stability. Biopolymers. 55:436–450.
36. Velaphi, S., et al. 2004. Cortical hyperostosis in an
infant on prolonged prostaglandin infusion: case
report and literature review. J. Perinatol. 24:263–265.
37. Kwon, D.S., Spevak, M.R., Fletcher, K., and Klein-
man, P.K. 2002. Physiologic subperiosteal new
bone formation: prevalence, distribution, and
thickness in neonates and infants. AJR Am. J. Roent-
38. Morris, S., Cassidy, N., Stephens, M., McCormack,
D., and McManus, F. 2002. Birth-associated femo-
ral fractures: incidence and outcome. J. Pediatr.
39. Abernethy, L.J., Lee, Y.C., and Cole, W.G. 1993.
Ultrasound localization of subperiosteal abscesses
in children with late-acute osteomyelitis. J. Pediatr.
40. Front, D., Hardoff, R., Levy, J., and Benderly, A. 1978.
Bone scintigraphy in scurvy. J. Nucl. Med. 19:916–917.
41. Ratanachu-Ek, S., Sukswai, P., Jeerathanyasakun,
Y., and Wongtapradit, L. 2003. Scurvy in pediat-
ric patients: a review of 28 cases. J. Med. Assoc. Thai.
42. Cole, W.G., Dalziel, R.E., and Leitl, S. 1982. Treat-
ment of acute osteomyelitis in childhood. J. Bone
Joint Surg. Br. 64:218–223.
43. Willing, M.C., Deschenes, S.P., Slayton, R.L., and
Roberts, E.J. 1996. Premature chain termination
is a unifying mechanism for COL1A1 null alleles
in osteogenesis imperfecta type I cell strains. Am. J.
Hum. Genet. 59:799–809.
44. Willing, M.C., et al. 1994. Osteogenesis imper-
fecta type I: molecular heterogeneity for COL1A1
null alleles of type I collagen. Am. J. Hum. Genet.
45. Jones, I.E., Williams, S.M., Dow, N., and Goulding,
A. 2002. How many children remain fracture-free
during growth? a longitudinal study of children
and adolescents participating in the Dunedin
Multidisciplinary Health and Development Study.
Osteoporos. Int. 13:990–995.
46. Byers, P.H., and Cole, W.G. 2002. Osteogenesis
imperfecta. In Connective tissue and its heritable disor-
ders molecular, genetic and medical aspects. P.M. Royce
and B. Steinmann, editors. Wiley-Liss. New York,
New York, USA. 385–430.
47. Cole, W.G., Chan, D., Chambers, G.W., Walker, I.D.,
and Bateman, J.F. 1986. Deletion of 24 amino acids
from the pro-alpha 1(I) chain of type I procollagen
in a patient with the Ehlers-Danlos syndrome type
VII. J. Biol. Chem. 261:5496–5503.
48. Byers, P.H., et al. 1997. Ehlers-Danlos syndrome
type VIIA and VIIB result from splice-junction
mutations or genomic deletions that involve exon
6 in the COL1A1 and COL1A2 genes of type I col-
lagen. Am. J. Med. Genet. 72:94–105.
49. Schwarze, U., et al. 2004. Rare autosomal recessive
cardiac valvular form of Ehlers-Danlos syndrome
results from mutations in the COL1A2 gene that
activate the nonsense-mediated RNA decay path-
way. Am. J. Hum. Genet. 74:917–930.
50. Grant, S.F., et al. 1996. Reduced bone density and
osteoporosis associated with a polymorphic Sp1
binding site in the collagen type I alpha 1 gene. Nat.
51. Mayer, S.A., Rubin, B.S., Starman, B.J., and Byers,
P.H. 1996. Spontaneous multivessel cervical artery
dissection in a patient with a substitution of
alanine for glycine (G13A) in the alpha 1 (I) chain
of type I collagen. Neurology. 47:552–556.
52. Simon, M.P., et al. 1997. Deregulation of the plate-
let-derived growth factor B-chain gene via fusion
with collagen gene COL1A1 in dermatofibrosar-
coma protuberans and giant-cell fibroblastoma.
Nat. Genet. 15:95–98.
53. Korkko, J., Annunen, S., Pihlajamaa, T., Prockop,
D.J., and Ala-Kokko, L. 1998. Conformation sensitive
gel electrophoresis for simple and accurate detection
of mutations: comparison with denaturing gradient
gel electrophoresis and nucleotide sequencing. Proc.
Natl. Acad. Sci. U. S. A. 95:1681–1685.
54. Cole, W.G., Evans, R., and Sillence, D.O. 1987. The
clinical features of Ehlers-Danlos syndrome type
VII due to a deletion of 24 amino acids from the
pro alpha 1(I) chain of type I procollagen. J. Med.
55. Chan, D., Lamande, S.R., Cole, W.G., and Bate-
man, J.F. 1990. Regulation of procollagen syn-
thesis and processing during ascorbate-induced
extracellular matrix accumulation in vitro.
Biochem. J. 269:175–181.
56. Hata, R., and Senoo, H. 1989. L-ascorbic acid 2-
phosphate stimulates collagen accumulation, cell
proliferation, and formation of a three-dimension-
al tissuelike substance by skin fibroblasts. J. Cell.