A deletion in the gene encodingsphingomyelin phosphodiesterase3 (Smpd3) results in osteogenesisand dentinogenesis imperfecta inthe mouseIsabelle Aubin1, Carolyn P Adams2, Sibylle Opsahl3,Dominique Septier3, Colin E Bishop2,4, Nathalie Auge5,Robert Salvayre5, Anne Negre-Salvayre5, Michel Goldberg3,Jean-Louis Guenet1 & Christophe Poirier2

The mouse mutation fragilitas ossium (fro) leads to a syndromeof severe osteogenesis and dentinogenesis imperfecta with nodetectable collagen defect. Positional cloning of the locusidentified a deletion in the gene encoding neutralsphingomyelin phosphodiesterase 3 (Smpd3) that led tocomplete loss of enzymatic activity. Our knowledge of SMPD3function is consistent with the pathology observed in mutantmice and provides new insight into human pathologies.

Osteogenesis imperfecta is a heterogeneous group of skeletal disorderscharacterized by extreme phenotypic variation, ranging from multiplefractures in utero, sometimes leading to perinatal death, to nearlynormal stature with low fracture incidence. Osteogenesis imperfectasyndromes affect B1 in 20,000 human live births1 and are commonlyclassified into seven types. The first four types historically described(types I–IV)2 represent 90% of all cases and are caused by mutationsin the genes encoding type I procollagen (COL1A1 or COL1A2). Theother three types (V–VII), which were recently added to the classifica-tion3–5, have clinical and radiological features similar to collagenousosteogenesis imperfecta but lack mutations in either COL1A1 orCOL1A2. For these noncollagenous forms, the underlying genes arestill unknown.

Mouse models for collagenous osteogenesis imperfecta arose spon-taneously or have been generated with transgenic or knock-in tech-nologies. In 1981, the recessive mutation fragilitas ossium (fro)6 wasreported as the first animal model for a severe form of osteogenesisimperfecta that could not be attributed to a collagen defect. Here, wereport the results of the positional cloning of this mutation, which wefound to be a deletion in Smpd3, the gene encoding sphingomyelinphosphodiesterase 3 (nSMase2).

Morphological, radiological and biological studies of homozygousfro/fro mice have been published previously7,8. At birth, affected miceare smaller than normal with deformities and multiple fractures of ribs

a +/fro

Ab

Eo

O

P

fro/fro

Ab

EoO

P

b

G630049C14RikAW539964

1810019D21Rik

2010007L18Rik

Smpd3

6030452D12

AW413431

Nutf2

Cdh3

D8Mit12

BC022641

CtrlPsmb10LcatSlc12a4Dpep3LOC384901

Ddx282310016K04Rik

Nfatc3

9530027K23Rik

Lypla3AI643885

BC006705

Zfp90

Cdh1

Dpep2

MMU8

AI413596

105.2 M

105.3 M

105.4 M

105.5 M

105.6 M

105.7 M

105.8 M

105.9 M

106.0 M

106.1 M

Figure 1 Abnormal mineralization in mutant mice and genetic localization

of the Smpd3 locus. (a) von Kossa staining shows defective alveolar bone

(Ab) mineralization and delayed formation of dentin in the incisor of fro/fro

mice compared with heterozygotes (+/fro). Scale bar, 100 mm. Eo, enamel

organ, O, odontoblasts, P, pulp. (b) High resolution mapping of the fro locus

in a 980-kb interval of chromosome 8. Candidate genes for underlying fro

are shown in bold. The physical size of this interval seems much larger than

expected for a genetic distance r 0.12 cM (computed from the

recombination frequency at 95% confidence level).

Published online 17 July 2005; doi:10.1038/ng1603

1Unite de Genetique des Mammiferes, Institut Pasteur, 25 rue du Docteur Roux, 75724 Paris Cedex 15, France. 2Department of Obstetrics and Gynecology, BaylorCollege of Medicine, 6550 Fannin St., Houston, Texas 77030, USA. 3Faculte de Chirurgie Dentaire, Paris 5, EA 2496; 1, rue Maurice Arnoux, 92120 Montrouge,France. 4Department of Molecular and Human Genetics, Baylor College of Medicine, 6550 Fannin St., Houston, Texas 77030, USA. 5INSERM Unit U-466 andBiochemistry Department, IFR-31, CHU Rangueil, 31059 Toulouse cedex 9, France. Correspondence should be addressed to J.-L.G. (guenet@pasteur.fr).

NATURE GENETICS VOLUME 37 [ NUMBER 8 [ AUGUST 2005 803

BR I E F COMMUN ICAT I ONS©

2005

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

egen

etic

s

and long bones. Cartilage formation is normal, but the matrix ofdeveloping bones is severely undermineralized. Mortality is elevated inthe perinatal period (up to 30%), but the condition stabilizes inmutant mice that survive to weaning. Most fro/fro adults breed andhave normal behavior and lifespan. In mutant mice, blood calciumlevel is normal (2.9–3.1 mmol l�1), parathyroid hormone is elevatedand bone osteonectin is decreased by 30%. In addition to thepreviously reported skeletal anomalies, we also observed severetooth and alveolar bone abnormalities linked to impaired mineraliza-tion (Fig. 1a). This pathology is consistent with the fro mutationrepresenting a model of noncollagenous osteogenesis imperfecta.

To clone the fro locus, we genotyped DNA samples from 1,448 F2

mice (335 fro/fro and 1,113 heterozygous or wild-type) born from aninterspecific cross. We mapped the fro locus to a 980-kb criticalinterval on chromosome 8 containing 26 genes, of which 10 couldbe considered candidates for underlying fro on the basis of theirexpression patterns (GNF-Symatlas; Fig. 1b).

To assess these candidates, we compared their transcription profilesby PCR amplification of reverse-transcribed RNA samples from 2- to3-d-old fro/fro and wild-type littermates, using primer pairs designedfrom the 3¢ region of each gene. We could not obtain an amplifiedcDNA product from the RNAs of fro/fro mice with primer pairs fromexon 9 of the gene Smpd3, encoding nSMase2, but we amplified aproduct of the expected size from wild-type samples (Fig. 2a andSupplementary Table 1 online). With the same RNA samples andprimers encompassing most of the coding sequence of Smpd3, theamplified product from fro/fro samples was shorter than that fromcontrol samples (Fig. 2b and Supplementary Table 1 online). Thisfinding suggested that the gene in question might be deleted in fro/fromice. We confirmed this idea at the molecular level after sequencingfro/fro genomic DNA. The mutated allele has a deletion of 1,758 bpencompassing part of intron 8 and most of exon 9 (Fig. 2c andSupplementary Fig. 1 online), reshuffling the 3¢ end of the Smpd3coding frame. The deletion of the intron 8–exon 9 splice acceptor siteleads to the transcription of the undeleted portion of intron 8(Supplementary Fig. 1 online). The occurrence of a stop codon inthis intronic segment, upstream of the deletion breakpoint, probably

results in the substitution of the last 33 amino acids of the wild-typeprotein with 13 amino acids translated from the transcribed intronicsequence in question (Supplementary Fig. 2 online). These substituted33 amino acids include a histidine residue9 that is required for nSMase2catalytic activity. We hypothesized that the protein translated from themutated allele would be nonfunctional. Consistent with this hypothesis,the enzymatic activity of the mutant protein was abolished (Fig. 2d,e).

Neutral sphingomyelinases are ubiquitously expressed enzymeswith elevated expression in the brain and in the embryonic growthplate of bone. They cleave sphingomyelin into ceramide, which is, inturn, a substrate for ceramidase resulting in the production ofsphingosine. Finally, the sphingosine is modified by a specific kinaseand converted into sphingosine 1-phosphate (S1P).

S1P has a mitogenic activity on the bone-forming cells, osteo-blasts10. Sphingomyelinase, ceramide and S1P are inhibitory factors ofbone resorption11. Impaired function of SMPD3, one of the twoneutral sphingomyelinases, and a defective ceramide pathway couldaffect bone development and remodeling, leading to excessive boneresorption. Increased bone resorption is commonly associated withbone fragility and osteogenesis imperfecta.

Notably, a progressive disappearance or loss of sphingomyelin hasbeen observed during cartilage matrix vesicle-induced mineralization,presumably due to hydrolysis by sphingomyelinases12. Sphingomyelincould also contribute to initial mineralization through the formationof hydrophobic spaces where calcium may combine with phosphate13.The precise relationship between the impairment of the gene encodingnSMase2, its cascade of intracellular effects and the production ofdefective extracellular mineralized tissue is not yet fully understood,but the data presented here indicate that sphingomyelinases are deeplyinvolved in bone and dentin mineralization.

Experiments on living animals were approved by the FrenchMinistry of Agriculture-Veterinary Department. J.-L.G. is fully accre-dited (N1 75-299).

GenBank accession number. Smpd3, AY836550.

Note: Supplementary information is available on the Nature Genetics website.

c

∆

b d eMWfro/fro +/+

2 kb

3 kb4 kb

a BrainForelim

b

KidneyLive

rRibs

SkinMW Hindlim

b

BrainForelim

b

KidneyLive

rRibs

SkinHindlim

b

MW

600 bp

600 bp

fro/fro+/+

*

fro/fro

Control

0

1

2

Sph

ingo

mye

linas

e ac

tivity

(nm

ol/h

.mg

prot

ein)

TNF-stimulatedsphingomyelinase

0

100

200

300

5 6 7pH

Sph

ingo

mye

linas

e ac

tivity

(nm

ol/h

.mg

prot

ein)

Control

fro/fro

Figure 2 Analysis of Smpd3 structure, expression and function. (a) RT-PCR with primers designed from the sequence of the last exon of Smpd3 and RNAs

isolated from fro/fro and wild-type (+/+) mice (red arrows in c). No product could be amplified from RNAs prepared from mutant mice. (b) A shortened

messenger is transcribed from the Smpd3 mutated allele. RNA samples were prepared from 2- to 3-d-old mouse femurs. MW, molecular weight scale (1 kb

ladder, Invitrogene; primers are indicated as blue arrows in c). (c) Schematic representation of Smpd3 with the deletion in the fro allele (D). Primers used forthe various RT-PCR are indicated. (d) The activity of acid (SMPD1) and neutral magnesium dependent (SMPD3) sphingomyelinases of brain homogenates

was determined at pH 5.0 and pH 7.0, under conditions described elsewhere14. In fro/fro brain extracts, the activity of the neutral sphingomyelinase was

severely deficient (12 7 4% of the control), whereas that of the acid sphingomyelinase was in the normal range. These data are consistent with a major

structural defect of SMPD3, the residual activity probably resulting from the other neutral sphingomyelinase SMPD2. (e) In cultured fibroblasts, the neutral

sphingomyelinase was activated by TNFa in controls but not in fro /fro fibroblasts.

804 VOLUME 37 [ NUMBER 8 [ AUGUST 2005 NATURE GENETICS

BR I E F COMMUN ICAT I ONS©

2005

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

egen

etic

s

ACKNOWLEDGMENTSWe thank G. Langsley and K. McElreavey for critical reading of the manuscriptand C. Kress for cell culture support. This work was partly supported by a SeedGrant from the Osteogenesis Imperfecta Foundation to C.P. and by a Fellowshipfrom the CANAM to I.A. This paper is dedicated to the memory of RittaStanescu who, in collaboration with her husband Victor Stanescu, made manycareful analyses aimed at a better understanding of the physiopathology ofthis mutation.

COMPETING INTERESTS STATEMENTThe authors declare that they have no competing financial interests.

Received 12 January; accepted 27 May 2005

Published online at http://www.nature.com/naturegenetics/

1. Forlino, A. & Marini, J.C. Mol. Genet. Metab. 71, 225–232 (2000).2. Sillence, D.O., Senn, A. & Danks, D.M. J. Med. Genet. 16, 101–116 (1979).3. Glorieux, F.H. et al. J. Bone Miner. Res. 15, 1650–1658 (2000).4. Glorieux, F.H. et al. J. Bone Miner. Res. 17, 30–38 (2002).5. Ward, L.M. et al. Bone 31, 12–18 (2002).6. Guenet, J.-L., Stanescu, R., Maroteaux, P. & Stanescu, V. J. Hered. 72, 440–441

(1981).7. Muriel, M.P. et al. Bone 12, 241–248 (1991).8. Sillence, D.O., Ritchie, H.E., Dibbayawan, T., Eteson, D. & Brown, K. Am. J. Med.

Genet. 45, 276–283 (1993).9. Goni, F.M. & Alonso, A. FEBS Lett. 531, 38–46 (2002).10. Grey, A. et al. Calcif. Tissue Int. 74, 542–550 (2004).11. Takeda, H. et al. FEBS Lett. 422, 255–258 (1998).12. Wu, L.N., Genge, B.R., Kang, M.W., Arsenault, A.L. & Wuthier, R.E. J. Biol. Chem.

277, 5126–5133 (2002).13. Goldberg, M. & Boskey, A.L. Prog. Histochem. Cytochem. 31, 1–187 (1996).14. Auge, N. et al. J. Biol. Chem. 273, 12893–12900 (1998).

NATURE GENETICS VOLUME 37 [ NUMBER 8 [ AUGUST 2005 805

BR I E F COMMUN I CAT I ONS©

2005

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

egen

etic

s