Differential expression of both extracellular and intracellular proteins is involved in the lethal or nonlethal phenotypic variation of BrtlIV, a murine model for osteogenesis imperfecta

RESEARCH ARTICLE

Differential expression of both extracellular and

intracellular proteins is involved in the lethal or nonlethal

phenotypic variation of BrtlIV, a murine model for

osteogenesis imperfecta

Antonella Forlino1, Chiara Tani2, Antonio Rossi1, Anna Lupi1, Elena Campari1,Benedetta Gualeni1, Laura Bianchi2, Alessandro Armini2, Giuseppe Cetta1, Luca Bini2

and Joan C. Marini3

1 Department of Biochemistry “A. Castellani”, Section of Medicine and Pharmacy,University of Pavia, Pavia, Italy

2 Department of Molecular Biology, University of Siena, Siena, Italy3 Bone and Extracellular Matrix Branch, NICHD, NIH, Bethesda, MD, USA

This study used proteomic and transcriptomic techniques to understand the molecular basis ofthe phenotypic variability in the bone disorder osteogenesis imperfecta (OI). Calvarial bonemRNA expression was evaluated by microarray, real-time, and comparative RT-PCR and the boneproteome profile was analyzed by 2-DE, MS, and immunoblotting in the OI murine modelBrtlIV, which has either a moderate or a lethal OI outcome. Differential expression analysisshowed significant changes for eight proteins. The expression of the ER stress-related proteinGadd153 was increased in lethal mice, whereas expression of the chaperone aB crystallin wasincreased in nonlethal mice, suggesting that the intracellular machinery is involved in the mod-ulation of the OI phenotype. Furthermore, in lethal BrtlIV, the increased expression of the carti-laginous proteins Prelp, Bmp6, and Bmp7 and the lower expression of the bone matrix proteinsmatrilin 4, microfibril-associated glycoprotein 2, and thrombospondin 3 revealed that both adelay in skeletal development and an alteration in extracellular matrix composition influence OIoutcomes. Differentially expressed proteins identified in this model offer a starting point forelucidating the molecular basis of phenotypic variability, a characteristic common to manygenetic disorders. The first reference 2-DE map for murine calvarial tissue is also reported.

Received: August 3, 2006Revised: January 30, 2007

Accepted: February 21, 2007

Keywords:

2-DE / Allele specific real-time PCR / Microarray / Osteogenesis imperfecta / Phenotypicvariability

Proteomics 2007, 7, 1877–1891 1877

1 Introduction

The skeletal dysplasia osteogenesis imperfecta (OI), alsoknown as brittle bone disease, is an autosomal dominantnegative disorder caused by mutations in the COL1A1 orCOL1A2 genes, coding respectively for the a1 and a2 chainsof type I collagen [1, 2]. OI occurs in all ethnic and racialgroups with an equal incidence of about 1 in 15 000–20 000,

Correspondence: Dr. Antonella Forlino, Department of Biochem-istry “A. Castellani”, Section of Medicine and Pharmacy, Univer-sity of Pavia, Via Taramelli 3/B, 27100 Pavia, ItalyE-mail: [email protected]: 139-0382-423108

Abbreviations: DTE, dithioerythritol; ECM, extracellular matrix;OI, osteogenesis imperfecta; WT, wild type

DOI 10.1002/pmic.200600919

© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

1878 A. Forlino et al. Proteomics 2007, 7, 1877–1891

although the mildest forms are probably underestimated dueto the difficulty of diagnosis and underreporting. The mostcommon molecular defects are point mutations causing thesubstitution of one of the glycine residues present at the firstposition of each of the 338 Gly-X-Y triplets composing the achain of type I collagen, in which X is often proline and Yhydroxylproline. The glycine residue is necessary for thefolding and stability of the triple helical structure of type Icollagen; in fact, it is the only amino acid whose side chain, ahydrogen atom, is small enough to be accommodated in theinternal space of the triple helix [3–5]. The second mostcommon molecular defect in OI patients affects the splicesites, causing exon skipping or alternative transcripts. Dele-tions and insertions have also been reported as causativedefects of the disease [1].

The bone fragility and skeletal deformity of OI are asso-ciated with a broad range of clinical outcomes, from verymild to lethal. The relationship between phenotype andgenotype is only partially understood and the underlyingmechanisms are unknown [6, 7]. In a general sense, it isclear that structural mutations of collagen are associatedwith lethal and moderate forms of the disease (types II, III,and IV, according to Sillence’s classification) [8, 9]. Theseverity of the OI phenotype depends in part on the type andposition of the causative mutation. During helix formation,these mutations delay propagation and folding. In the extra-cellular matrix (ECM), they may compromise procollagenprocessing or interfere with collagen–collagen and/or col-lagen–ECM protein interactions. The quantitative defects, incontrast, are usually due to a null COL1A1 allele and result insynthesis of a reduced amount of structurally normal col-lagen and mild OI (type I, according to Sillence’s classifica-tion).

In addition to variation of OI severity with mutation typeand location, severity of the condition may vary greatlyamong family members or between different families withthe same molecular defect [7, 10–12]. Although this pheno-typic variability among individuals with the same mutationhas been attributed to mosaic conditions, difference ingenetic background, or stochastic events; the mechanismwas not investigated in detail [1, 6, 13, 14]. In addition, envi-ronmental factors, transcriptional and translational factorsin the cells as well as molecular and biochemical interactionswith modifier genes may play a role in this variability.

The study of phenotypic variability in human OI patientsis particularly difficult because it requires direct investigationof bone tissue in a significant number of individuals with thesame mutation, but different phenotypes. For these reasons,it is preferable to address this issue first in animal models. In1999, Forlino et al. [15] created a knock-in murine model forOI (BrtlIV) carrying a Gly349Cys substitution in one col1a1allele; this mutation was first identified in a child with themoderately severe form of OI (OI type IV) [16]. BrtlIV repro-duces the most common type of molecular defect, a glycinesubstitution, and the dominant inheritance most commonlypresent in human OI patients. Interestingly, it also shows the

phenotypic variability reported for human OI. Some BrtlIVmice have a moderately severe form of OI, with growth retar-dation, bone fragility, and deformity, while others die a fewhours after birth from respiratory stress. Furthermore, BrtlIVmice are an outbred strain, thus reproducing the hetero-geneity of genetic background found in OI patients [15, 17].

Hence, they are a valid tool to investigate the molecularbasis of clinical variability.

In a previous study, we reported that the mutant andnormal type I collagen molecules were incorporated equallyinto the matrix of lethal and nonlethal mice and formedmature covalent crosslinks with the same efficiency [18].

We examined the collagen in different tissues and, al-though we found a variable extent of overglycosilation ofmutant molecules, we did not detect consistent differencesbetween lethal and nonlethal animals. We also did not detectany changes in the thermal stability or rate of thermal dena-turation of mutant collagen. Furthermore, no changes incollagen–collagen recognition or interactions were found,except for disruption of quasicrystallin lateral packing ofmolecules in tendon from some mutant animals, which didnot correlate significantly with a lethal or nonlethal outcome[18]. We suggested that OI symptoms and phenotype varia-tions were probably related to abnormal interactions ofmutant collagen helices with other matrix molecules or toabnormal functions of bone cells.

In the work presented here, we continued our investiga-tion of the molecular basis of phenotypic variability usingcomplementary molecular and proteomic approaches.Microarray technology was employed to screen the expres-sion level of a significant number of ECM proteins, whosedirect detection at the protein level may be difficult using the2-DE proteomic approach due to the size and hydrophobicityof most of those proteins [19]. Since it is ultimate differencesin protein levels that act physiologically in cells, accuratemeasurement of protein expression levels remains a funda-mental goal in proteomics. In our research, we combined themicroarray approach with Western blotting and the use of2-DE, MS, and bioinformatics analysis to investigate thecomplete protein profile of bone tissue. Furthermore, since a2-DE protein map for murine neonatal calvarial bone is notavailable in the public databases, we created the first refer-ence protein profile for this tissue.

Our final goal was the identification of phenotype spe-cific markers for OI, which can be tested for their relevanceto human OI and potentially used for the development ofnew therapeutic approaches.

2 Materials and methods

2.1 Animals

The BrtlIV heterozygous mice and wild-type (WT) litter-mates were maintained in the designated pathogen-freefacility at NICHD, NIH, Bethesda, USA under the standard


Proteomics 2007, 7, 1877–1891 Animal Proteomics 1879

institutional animal care protocol. One-day-old mice wereused for the present studies. Tissue specimens were dis-sected and stored at 2807C until processed.

2.2 Type I collagen expression level and allele specific

real-time PCR quantitation

Calvarial bone specimens were obtained from BrtlIV micewith lethal (MTL) or nonlethal (MTA) OI type IV phenotypeand from WT littermates on day 1 of life (n = 3 for each type).The surrounding connective tissue was removed, and thecalvariae were cut in small pieces, immediately frozen inliquid nitrogen and stored at 2807C. Total RNA was isolatedusing TriReagent (Sigma, Milan, Italy) according to themanufacturer’s protocol and treated with DNAse (TurboDNA-free, Ambion, Cambridgeshire, UK). The concentra-tion of total RNA was determined by measuring the OD at260 nm.

The expression of the two genes col1a1 and col1a2, cod-ing, respectively, for the a1 and a2 chains of type I collagenwas evaluated by quantitative real time PCR (Q-RT-PCR).Briefly, 5 mg of total RNA for each sample was retrotrascribedby cDNA Archive kit (Applied Biosystems, Monza, Milan,Italy) following the manufacturer’s suggestions. The expres-sion level of col1a2 was evaluated using the Assay-on-Demand primers–probe set (Mm00801665-g1), whereas, forthe expression level of col1a1 an Assay-by-Design primers–probe set was created (Applied Biosystems). All ABI Assaysare designed to generate amplicons of 50–150 bp and arecarried out using identical cycling conditions: 507C, 2 min;957C, 10 min; 957C, 15 s; and 607C, 1 min for 40 cycles. Taq-Man Universal master mix (Applied Biosystems) was usedfor the assays. Serial dilution of WT cDNA was used todetermine expression values for both target genes and anendogenous housekeeping gene (Gapdh, Mm99999915_g1).Relative expression values for each gene were expressed as aratio of target gene expression level to Gapdh in the samespecimen and the ratio of the normalized values was calcu-lated between MTL and MTA mice 6 SEM. For allele specificidentification and relative quantitation of the col1a1 mutantand WT alleles, we designed two primers spanning an 88 bpamplicon P1 50-TGCTGGTCCCAAAGGTTCTC-30 (1614–1633 bp, GenBank NM_007742) and P2 50-GCCAGGA-CTGCCAGTGA-30 (1701–1685 bp, GenBank NM_007742)and two probes, respectively, specific for the WT and mutantallele, 50-Fam-TGGTGAAGCT(G/T)GTCG-30 (1635–1649 bp,GenBank NM_007742), containing 1 and 3 locked nucleicacid (LNA) nucleotides, respectively (Proligo, USA). TaqManUniversal master mix (Applied Biosystems) was used foramplification in 25 mL reactions. The amplification condi-tions were 507C, 2 min; 957C, 10 min; 957C, 15 s; and 607C,1 min for 45 cycles. The MX3000 (M-Medical, Cornaredo,Milan, Italy) was used; each sample was run in triplicate on96-well plates. The serial dilution method was used for rela-tive quantitation of the mutant and WT alleles and standardcurves were generated using ten-fold dilutions from 100 to

0.01 pg of two plasmids (PCRII, Invitrogen, San GiulianoMilanese, Milan, Italy), containing respectively, the 283 bpfragment (1529–1812 bp, GenBank NM_007742) of themutant or WT allele. No mutant allele signal was detectedabove the threshold in the reaction containing the WT probeand, conversely, no WT allele signal was detected above thethreshold in the reaction containing the mutant probe.Results were expressed as the ratio of WT versus mutant(WT/MT) allele expression values in MTL and MTA ani-mals.

Statistical analysis was performed by Student’s t-test anda p,0.05 was considered significant. The statistical calcula-tions were done with SigmaStat 3.5.

2.3 RNA preparation and microarray analysis

Calvarial bone specimens were dissected from BrtlIV micewith a lethal or an OI type IV phenotype and from WT litter-mates few hours after birth (n = 2 for each type). The sur-rounding connective tissue was removed; the bone was cutinto small pieces, immediately frozen in liquid nitrogen andstored at 2807C. Total RNA was isolated using TriReagentaccording to the manufacturer’s protocol. The concentrationof total RNA was determined spectrophotometrically and itsquality was verified using the Bioanalyzer 2001 (AgilentTechnologies, Santa Clara, CA, USA). Microarray analysiswas done by Memorec Biotec (Köln, Germany). Since theamount of total RNA in each sample was not sufficient fordirect labeling, the samples were amplified and the presenceof uniformly distributed fragment length in the aRNAs wasverified. To determine the expression profiles of lethal (MTL)and nonlethal (MTA) knock-in BrtIV mice, the aRNA (2 mg)of the respective samples was labeled by reverse transcriptionwith Cy5 and a pool of aRNA (2 mg) from both WT mice waslabeled with Cy3. To analyze differences between MTL andMTA knock-in mice directly, the aRNAs of these animals(2 mg) were labeled differentially with Cy5 or Cy3 and hybri-dized on an additional array.

Pairs of Cy5/Cy3 labeled samples were hybridized to-gether on a PIQOR™ cRNA Array (PIQORECMatrix cDNAArray, #130-091-973, Memorec, see Supporting InformationTable 1 for the complete genes’ list) containing 194 ECMgenes, each spotted in duplicate. After hybridization, thearrays were read by laser scanning. Image capture and sig-nal quantitation were done with the ScanArray 3000 (GSILuminomics) and the ImaGene software v4.1. The imageswere analyzed digitally. For data evaluation and the sub-sequent generation of expression profiles, signal and back-ground intensities were determined using ImaGene v4.1software (Bio Discovery). Background values were sub-tracted from signal intensities to obtain the net intensities.The Cy5/Cy3 ratio for single spots was computed, the meanratio of two corresponding spots was built and the relativeCV (%) was estimated from the two ratios with respect totheir mean.



2.4 Comparative RT-PCR and real-time PCR

Validation of microarray and mRNA expression levels ofsome differentially expressed proteins from the proteomicanalysis were performed on three independent samples foreach phenotype by comparative RT-PCR and/or real-timePCR. We confirmed that the two validation methods werecomparable by demonstrating that the expression level oftwo transcripts, Gadd153 and thrombospondin 3, was simi-lar in both assays.

Comparative RT-PCR was performed using the Intra-Spec Comparative RT-PCR Kit (Ambion) according to themanufacturer’s protocol. Briefly, 5 mg of total RNA of MTL,MTA, or WT mice were mixed with 1 mL of Tag10RT primerand, in a separate tube, the same amount of RNA from miceof a different phenotype was mixed with 1 mL of Tag50RTprimer. RNAs were denatured for 10 min at 707C and thenequilibrated at 487C for 10 min. In a separate tube, a reactionmixture containing 2 mL of 106buffer, 2 mL of RT dNTPsmix, and 0.5 mL of 32P-dATP was assembled. Then the reac-tion mixture was added to the RNA and RT primer mix andthe samples were incubated with reverse transcriptase at487C for 2 h. After removing the free nucleotides from thecDNA templates using ProbeQuant G-50 microcolumn(Amersham Biosciences, now GE Healthcare, Milan, Italy),the specific activity of each cDNA was measured by b coun-ter. Equal amounts of cDNAs were used in the followingPCR amplification reaction. Nested PCR was performed in a25 mL reaction mixture using EuroTaq polymerase (Euro-clone, Pero, Milan, Italy), the annealing temperature wasoptimized experimentally for each primer set and 40 cycleswere used. The gene specific 50-primers designed for nestedPCR are listed in Table 1. The 30-primers were the outer andinner primers included in the Ambion Kit. PCR products (5–10 mL) were analyzed on 6–10% polyacrylamide gels, stainedwith SYBR GREEN II (Sigma). Fluorescence signal wasdetected and analyzed by VersaDoc 3000 imaging system(BioRad, Segrate, Milan, Italy) equipped with Quantity Onesoftware (BioRad). Each reaction was run in triplicate.

For real-time experiments, 5 mg of total RNA from eachsample was reverse transcribed with the cDNA Archive Kit(Applied Biosystems). PCR primers and Taqman probeswere obtained from Assay-on-Demand (Applied Biosys-tems): agrin (Mm01264855_m1), bone morphogenic pro-tein 2 (Bmp2, Mm01340177_m1), bone morphogenic pro-tein 4 (Bmp4, Mm00432087_m1), bone morphogenic pro-tein 6 (Bmp6, Mm00432095_m1), growth arrest and DNAdamage inducible protein (Gadd153, Mm00492097_m1),matrilin 4 (Mm00522095_m1), prolargin precursor(Prelp, Mm00475328_m1), thrombospondin 3 (THBS3,Mm00449802_m1), aB crystallin (Mm00515566_m1). TheQ-RT-PCR was performed following standard protocol [20]using MX3000 apparatus (M-Medical, Cornaredo, Milan,Italy). Each sample was run in triplicate on 96-well plates.TaqMan Universal master mix (Applied Biosystems) wasused for the assays with the recommended cycling condi-tions described above. The serial dilution method of WTcDNA was adopted to determine expression values for eachgene and for an endogenous housekeeping gene (Gapdh,Mm99999915_g1). Relative expression values were expres-sed as a ratio of target gene expression level to Gapdh in thesame specimen. Data were expressed as mean 6 SEM. Thesignificance of differences was evaluated by Student’s t-test,with p,0.05 indicating significance. All the statistical cal-culations were performed with SigmaStat 3.5.

2.5 2-DE

Calvarial bone specimens were obtained from 1-day-old MTLand MTA BrtlIV mice and WT littermates (at least n = 3 foreach type). After the surrounding connective tissue wasremoved, the calvariae were cut into small pieces, immedi-ately frozen in liquid nitrogen, and kept at 2807C until pro-cessed. Each sample was solubilized in 350 mL of 8 M urea,4% CHAPS, 65 mM dithioerythritol (DTE) for 45 min atroom temperature with occasionally stirring. After centri-fugation at maximum speed for 10 min, the supernatant wasremoved for protein quantitation by RC-DC Protein

Table 1. Gene specific primers used for comparative RT-PCR

Gene GenBank Primer sequence Nucleotideposition

Bmp7A X56906 50-GATGCTTCACCATCTAAGTCTCT-30 1446–1468Bmp7B X56906 50-ACCTTGGCGAGGAGCCCACAGA-30 1477–1498Gadd153A NM_007837 50-GAAAGTGGCACAGCTAGCTGAA-30 458–479Gadd153B NM_007837 50-AGCGGCTCAAGCAGGAAATCGA-30 487–508PrelpA BC019775 50-ACTGATCCTGCAAACCTGAACTA-30 3183–3205PrelpB BC019775 50-ACCATCAGCATCCCATTATCCAA-30 3213–3235Magp2A AF180805 50-ACCATGCTTCGTGGCTTGTCCTA-30 1126–1148Magp2B AF180805 50-GGAAGGTTGTGGGTGAGGCAAGA-30 1176–1198THBS3A L04302 50-ATCTCCAGTACCGATGCAATGACA-30 2831–2854THBS3B L04302 50-AGTGCCTGAGGACTTTGAGCCATT-30 2856–2879



Assay (BioRad). Serum albumin was used for the standardcurve. For 2-D plasma electrophoresis, blood was collectedfrom three animals of each type on day 1 and plasma wasseparated by centrifugation in Terumo Capiject EDTA (Na21)Tubes (Cardinal Health 200, NC, USA). To 10 mL of plasmawe added 20 mL of 10% SDS, 2.3% DTE, then heated at 957Cfor 5 min, and cooled slowly at room temperature. Subse-quently, 470 mL of 8 M urea, 4% CHAPS, and 65 mM DTEwere added and proteins were quantitated by RC-DC ProteinAssay (BioRad).

2-DE was performed using immobiline/polyacrylamidesystem [21, 22]. Briefly, 60 mg (for analytical runs) or 0.5–1 mg (for preparative runs) of total proteins were used ineach electrophoresis. IEF was done on immobiline strips(18 cm), providing a nonlinear pH 3–10 gradient (GEHealthcare) in an Ettan IPGPhor protein apparatus. Hydra-tion was performed in the strip holder for 1 h without voltageand for 8 h at 30 V, and then 120 V was applied for 1 h and300 V for 30 min. The voltage was then linearly increased to3500 V in 3 h and kept at 7950 V until a total voltage of60 000–70 000 was reached. IPG strips were equilibrated intwo consecutive steps: (i) 12 min in 20 mg/mL of DTE and(ii) 5 min in 25 mg/mL of iodoacetamide, both dissolved inSDS equilibration buffer (6 M urea, 50 mM Tris, pH 6.8,30% glycerol, 2% SDS, and trace of bromophenol blue). Thesecond dimension was carried out on 9–16% polyacrylamidelinear gradient gels (18 cm620 cm6 1.5 mm) at 40 mA/gel,at 107C until the bromophenol blue front reached the bottomof the gels. IPG strips were sealed on the top of the gels with0.4% agarose. Each sample was electrophoresed at least onthree separated gels. Analytical gels were stained withammoniacal silver nitrate [23, 24], preparative gels with CBB(Sigma), according to the manufacturer’s suggestion. Gelimages were acquired by VersaDoc 3000 imaging system(BioRad).

2.6 Analysis of the images of 2-DE

2-D gel images were analyzed by PDQuest software (BioRad)to generate the reference protein profile for WT calvarialbone tissue and to compare the 2-D maps of the MTL, MTA,and WT samples. Spot detection and matching were per-formed automatically followed by manual matching. Thequality of the matches made by the computer was criticallyevaluated in each case, and necessary editions and correc-tions were done manually. The experimental molecular massand pI were obtained from molecular weight protein stand-ards (BioRad). Normalization of the spot intensities wasperformed against whole-gel densities. Only well-definedspots were considered. Only differences greater than two-fold and consistent among all the samples and sample repli-cates were taken into account. The significance of differenceswas evaluated by Student’s t-test with a p,0.05 taken asindicating significance. All statistical calculations were donewith SigmaStat 3.5. Normality and equal variance tests wereapplied before running the statistical analysis.

2.7 MALDI-TOF-MS protein identification

Protein identification by MALDI-TOF mass fingerprintingwas carried out as previously described [25]. Briefly, elec-trophoretic spots were manually excised and destainedwith 25 mM ammonium bicarbonate/50% ACN. Proteinswere trypsin-digested [26] overnight and aliquots of 0.5 mLwere applied on the target disk and air dried. Then 0.5 mLof matrix solution (CHCA saturated in 50% v/v ACN/0.5%v/v TFA) was added to the samples and again allowed todry. Spectra were obtained using the Ettan MALDI-TOFPro mass spectrometer (GE Healthcare). Mass fingerprint-ing database searching in the NCBIm, Swiss-Prot, andTrEMBL was carried out using ProFound (www.proteometrics.com), PeptIdent (http://www.expasy.org), and MAS-COT (www.matrixscience.com) softwares, available online.The searches were performed with a peptide mass toler-ance of 50 ppm, carbamilation modification of methionineresidues and allowed a single missed tryptic cleavage. Pro-found Software, which allowed for hydroxylisine modifica-tion was used for spot 32 identification. Peptide sequenc-ing by MS/MS analysis was performed to confirm thesequence of some peptides when the PMF was not clear,using a nanospary/LCQ DECA IT mass spectrometer(Thermo, West Palm Beach, FL, USA). Database searchingwas carried out using TurboSEQUEST (Thermo) andMASCOT MS/MS ion search software (www.matrixscience.com).

The geneontology tool (www.geneontology.org), Target P(www.cbs.dtu.dk/services/TargetP), DAVID (apps1.niaid.nih.gov/david/), and Expasy (www.expasy.org) were used todetermine subcellular localization and functional classifica-tion of the proteins.

2.8 Western blotting

Using independent samples, total proteins were extractedfrom calvarial bone of 1-day-old MTL and MTA BrtlIV miceand WT littermates (n = 3 for each type), as described above.For each sample, 20 mg of proteins were separated on 12%polyacrylamide gel. After electrophoresis, the proteins weretransferred onto a PVDF membrane (PVDF, GE Healthcare).Hybridization with primary antibodies (Santa Cruz, Heidel-berg, Germany) against Gadd153 (sc-7351), fibrinogen b(sc218027), aB crystallin (sc-22744), and b-actin (sc-8432),were performed at 47C overnight at the dilutions suggestedby the manufacturer’s. Incubation with specific secondaryantibody was performed for 1 h at room temperature. ECLPlus (GE Healthcare) was used for signal detection. Eachsample analysis was repeated in three independent gels.Films were acquired by EPSON Scanner and band inten-sities were quantified by Multi Gauge v2.2 software. Nor-malization was performed using b-actin as housekeepingprotein.



3 Results

3.1 Type I collagen expression and microarray

analysis

To investigate the molecular basis of the variable phenotypeassociated with the type I collagen mutation in the Brtlmouse, we first explored possible differences in total col-lagen transcription and mutant versus WT col1a1 alleleexpression in day 1 murine calvariae, using Q-RT-PCR. Nosignificant differences were detected (data not shown).

Since ECM proteins are often difficult to extract andsolubilize from whole tissues and to analyze by 2-DE, weevaluated their expression at the transcript level by micro-array using a specific PIQOR custom array (SupportingInformation Table 1), containing 194 ECM genes. We com-pared total calvarial RNA from Brtl mice with a lethal ormoderate phenotype, respectively, to WT littermates and toeach other. We accepted only differences that were greaterthan 1.5-fold. The results obtained by DNA microarray werevalidated by comparative RT-PCR and by real-time RT-PCR.The RNA preparations used for validation were independentfrom those utilized for the microarray, to increase the samplenumber and to reduce distortions due to variability among

the animals (Table 2). The data obtained from microarrayand the validation methods were not satisfactorily con-gruent for all genes, although the fold differences weresimilar when they were congruent. We confirmed three ofthe six transcripts identified on microarray as differentiallyexpressed in the two mutant phenotypes of Brtl; Gadd153,Bmp7, and Prelp had higher expression in lethal animals.For transcripts identified on microarray as differentiallyexpressed in mutant lethal versus WT mice, we validatedseven of the nine transcripts. Bmp6, Bmp7, Gadd153, andPrelp had higher expression in lethal animals while matri-lin 4, Magp2, and thrombospondin 3 had higher expressionin WT mice.

3.2 Bone specific expression of Gadd153

Gadd153 is a transcription factor which is activated by cel-lular stress, in particular by ER retention. It is also known tobe involved in apoptosis activation and/or growth arrest. Toinvestigate whether the increase in Gadd153 expression levelfound in calvarial bone in MTL mice was limited to bonetissue, we examined total RNA extracted from skin and lungof BrtlIV mice with lethal and nonlethal outcome and fromWT littermates. These tissues were chosen because most of

Table 2. Microarray and validation data

Gene name Gene symbol EntrezGene ID

Microarraya) Validation(mean 6 SEM)

p value

MTL/MTAAgrin 25592 2.21/22%; n.d. 1.20 6 0.17d) 0.54Bone morphogenic protein 2 Bmp2 12156 1.70/–%; n.d. n.d.Bone morphogenic protein 4 Bmp4 12159 1.68/–%; n.d. 1.28 6 0.085d) 0.85Bone morphogenic protein 7 Bmp7 12162 1.84/2%; n.d. 1.34 6 0.03c) 0.001Growth arrest and DNA damage

inducible protein 153GADD153/CHOPb) 29467 2.06/22%; 2.32/3% 1.72 6 0.11c)

3.25 6 0.43d),0.001; 0.003

Pro–Arg-rich and Leu-rich protein Prelpb) 116847 2.03/9%; n.d. 2.03 6 0.12c) ,0.001

MTL/WTAgrin 25592 1.55/2%; n.d. 1.11 6 0.22d) 0.97Bone morphogenic protein 6 Bmp6b) 25644 1.99/–%; n.d. 1.52 6 0.17d) 0.01Bone morphogenic protein 7 Bmp7b) 12162 2.54/12%; n.d. 3.25 6 0.65c) ,0.001Growth arrest and DNA damage

inducible protein 153GADD153/CHOPb) 29467 1.78/3%; 2.14/5% 2.99 6 0.45c)

2.95 6 0.54d),0.001; 0.004

Matrilin 4 MAT-4b) 17183 22.08/6%; 21.66/0% 21.92 6 0.23d) 0.05Microfibril-associated glycoprotein 2 Magp2b) 50530 23.33/6%; 21.56/8% 21.87 6 0.08c) 0.004Pro–Arg-rich and Leu-rich protein Prelpb) 116847 2.03/9%; n.d. 3.73 6 0.56c) 0.004Stathmin Lap18 16765 1.89/6%; n.d. 1.90 6 0.44d) 0.248Thrombospondin 3 THBS3b) 21827 22/6%; 21.72/1% 22.53 6 0.63c)

22.43 6 0.35d)0.047; 0.05

MTA/WTAgrin 25592 n.d.; 1.55/2% 1.07 6 0.14d) 0.515Thrombospondin 3 THBS3b) 21827 21.59/2%; 21.61/11% 21.62 6 0.23d) 0.161

a) The values represent the mean ratio Cy5/Cy3 of the two corresponding spots and the CV (%).b) Microarray data validated by c) comparative RT-PCR; and d) real-time RT-PCR. MTL = BrtlIV mice with lethal outcome; MTA = BrtlIV mice

with moderate severe outcome; and WT = control mice.



the investigations of OI have involved dermal fibroblasts,which synthesize abundant type I collagen, and becauselethal Brtl mice die from respiratory distress, as frequentlyoccurs in patients with severe OI. No difference was found inGadd153 expression between the genotypes in either skin(MTL/MTA: 0.95 6 0.03, p = 0.77; MTL/WT: 1.2 6 0.03,p = 0.06) or lung (MTL/MTA: 1.08 6 0.06, p = 0.164; MTL/WT:1.09 6 0.06, p = 0.72) using real-time RT-PCR. Impor-tantly, we confirmed that the difference in Gadd153 boneexpression between genotypes results in a difference inGadd153 protein levels by Western blot of extracts of an in-dependent set of MTL, MTA, and WT littermate using com-mercially available mAb. The transcription factor Gadd153was detectable only in mutant mice with lethal outcome,confirming the translational impact of the increase found atthe transcription level (Fig. 1A).

Figure 1. Western blot confirmation of microarray and 2-D pro-teomic data. Proteins (20 mg) extracted from bone were loaded onSDS-PAGE for each sample. Expression of Gadd153 (A), fibrino-gen b (B) and a crystallin B chain (C) were determined with spe-cific antibodies. Equal loading of the samples was evaluated byprobing the membranes with b-actin antibody after stripping.

3.3 2-DE map of calvarial bone tissue

Protein solubilized from calvarial bone tissue of day 1 WTmice were separated by 2-DE, using a broad pH 3–10 non-linear gradient. About 1300 spots were identified and theirpIs and molecular weight were characterized. The number ofspots detected in each set of three replicates was reproducible(CV 8–30%) as was spot intensity (CV 32 6 15%). So far, 164spots have been excised, subjected to in-gel digestion andanalyzed by PMF using MALDI TOF MS. In some cases,MS/MS was run to confirm the identification (Fig. 2, Table3). The identified spots represented 96 distinct proteins, pri-marily intracellular (82), of which 49 are cytosolic, 17 occurin organelles, 7 are nuclear, 7 are both cytosolic and nuclear,1 is both cytosolic and in an organelle, and 1 is cytosolic and

associated with the plasma membrane. The remainders weresecreted proteins present either in the blood (10) or incorpo-rated in the ECM (3). One protein was unknown (# 162). Thesmall number of ECM proteins in our 2-DE map of bonetissue may be the result of the known difficulty of solubiliz-ing collagen and/or matrix proteins from whole tissue, or theconsequence of their high molecular weight, which does notallow them to resolve on our 2-D SDS-PAGE. Moreover,some matrix proteins could be present in the unidentifiedand fainter spots.

Among the identified spots, we recognized proteinsinvolved in physiological processes (73%) and with structuralfunctions (15%), while function was unknown for 10% ofthem. With few exceptions, the positions of the identifiedprotein spots in the 2-DE correlated well with the theoreticalvalues of molecular weight and pI values of the proteins. Inthe few exceptions, the protein spots might be PTM prod-ucts, adducts, or degradation products of the identified pro-teins.

We also noted that 24 unique proteins have multiple iso-forms, as each of them is represented by more than one dis-tinct 2-DE gel spot, as shown in Table 3. These proteinsincluded both intracellular molecules and secreted proteins.For one of these proteins, serum albumin, all the isoformsare smaller than the theoretical molecular weight, whichindicates a profound fragmentation of the protein.

The a1 chain of type I collagen is present in full length (#32) and in the C-terminal form (# 155).

In some cases, the sensitivity of the MS allowed therecognition of more than one protein in apparently uniquespots. For instance, # 25 as protein disulfide isomerase A6precursor and a enolase, # 59 as phosphatidilethanolaminebinding protein and peroxiredoxin 2, # 74 as heat shock pro-tein cognate 71 kDa protein and a fetoprotein, and # 129 asproteasome subunit a type 2 and heterogeneous nuclearribonuclear protein H0. For spot # 162, identified as hypo-thetical protein LOC70984, our data provide the first evi-dence for its expression in bone tissue.

3.4 2-DE map profile comparison of lethal and

nonlethal BrtlIV mice and WT littermates

To compare the protein profile of calvarial bone in lethal andnonlethal BrtlIV mice and WT littermates, three independ-ent samples from each phenotype were analyzed at least intriplicate. In order to reduce genetic, epigenetic, and experi-mental variability in expression of some proteins, the gelswere compared to each other using PDQuest Software andsynthetic gels were created containing the spots consistentlypresent or absent in all gels from mice with each genotype/phenotype. While we did not observe any protein whosepresence or absence was limited to one phenotypic group, wefound greater than two-fold differences in the expressionlevel of five spots, corresponding to three proteins (Fig. 3),which were reproducible in all examined mice. Fibrinogen a(# 89, 107) and fibrinogen b (# 53, 103) have higher expres-



Figure 2. A representative 2-DE image of the calvarial bone protein extract. The gel was stained by ammoniacal silver nitrate and the imagewas acquired by VersaDoc 3000. One hundred and sixty-four spots that have been identified by MS are indexed and the complete list of theproteins is in Table 3.

sion in mutant than WT mice (Fig. 3A) and a crystallin Bchain (# 109) expression was increased in surviving Brtl mice(Fig. 3B and Table 4). Several other proteins showed less thantwo-fold differences between phenotypes or were not con-sistent among the samples and are therefore not reported(data not shown). Western blots, using independent tissuesamples and commercially available antibodies confirmed anincreased level for the fibrinogen b in lethal (6.35 6 3.1) andnonlethal (6.0 6 2.88) mice with respect to the WT and anincrease expression of the aB crystallin in moderately severeBrtlIV mice (2.25 6 0.77) with respect to lethal animals(Figs. 1B and C). Q-RT-PCR showed that the a crystallin Bchain expression level was higher in MTA, paralleling thedifference found at the protein level (1.67 6 0.2) (data notshown).

Since fibrinogen is a circulating blood protein, its sys-temic expression was evaluated directly in blood from BrtlIV

mutant and WT mice. Quantitative analysis of triplicate gelsfrom three different mice from each genotype was per-formed using PDQuest software. No statistical significantdifferences were found among the phenotypes in theexpression level of either fibrinogen a or b, suggesting thatthe increased level of these proteins in bone is tissue specific(Fig. 4).

4 Discussion

The occurrence of “variable expression”, that is, of variableclinical outcomes from the identical molecular defect in bothrelated and unrelated individuals, is a well-known character-istic of many human monogenic disorders [27, 28]. Its mo-lecular basis is generally unknown; association with envi-ronmental elements or genetic factors such as cellular



Table 3. Proteins identified in murine calvarial bone

Protein Acc.no.

EntrezGeneID

Mth/pIth Mob/pIob Numberof massvaluesmatched

Sequencecoverage(%)

MASCOTScore

MS/MS peptides SpotID

IntracellularCytosolActin, a skeletal muscle P02568 11459 42.0/5.23 19.0/4.84 9 24 71 20

53.0/5.16 11 37 138 48Actin, cytoplasmatic 1 (b-actin) P02570 11461 41.7/5.29 20.1/5.01 7 22 64 21

19.7/4.94 6 18 59 2245.1/5.15 11 41 194 50

Adenylate kinase isozyme 1 Q9R0Y5 11636 21.5/5.67 26.0/5.34 9 56 154 63a crystallin B chain P23927 12955 20.0/6.76 23.7/6.94 6 44 103 109Calmodulin P02593 12313 16.7/4.09 20.0/4.09 10 54 96 23Creatine kinase M chain P07310 12715 43.0/6.58 44.0/6.76 7 23 71 95

27.9/6.38 6 17 74 139Dihydropyrimidinase–related protein 2 O08553 12934 62.1/5.95 62.2/6.17 8 20 93 102Elongation factor 1-b O70251 55949 24.5/4.53 32.4/4.55 5 37 88 44Enolase a P17182 13806 47.0/6.36 35.8/6.05 6 23 94 161

43.8/6.24 5 19 46 AAVPSGASTGIYEAL-ELRYITPDQLADLYK

25

Enolase b P21550 13808 46.8/6.81 48.3/6.57 6 19 52 AAVPSGASTGIYEAL-ELRVNQIGSVTESIQACK

98

Eukaryotic translation initiation factor 4H Q9WUK2 22384 27.2/6.91 31.9/6.19 6 36 85 141Fascin Q61553 14086 54.2/6.21 53.9/6.39 7 15 78 121Fructose-bisphosphate aldolase A P05064 11674 39.2/8.4 41.8/8.04 11 53 162 92Gluthatione S-transferase P 1 P19157 14870 23.4/8.13 27.1/7.98 6 37 99 99Heat shock 27 kDa protein P14602 15507 23.0/6.12 28.7/6.10 5 30 84 135Heat shock cognate 71 kDa protein P63017 15481 70.8/5.37 69.0/5.34 20 43 341 74

32.8/4.74 9 23 113 15871.7/– 6 44 133 112

Isocitrate dehydrogenase (NADP)cytoplasmic

O88844 15926 46.6/6.48 46.0/6.72 12 36 184 97

Myosin light chain 1, skeletal muscleisoform

P05977 17901 20.4/4.98 25.3/4.97 9 56 136 58

26.1/4.98 9 56 138 60Peptidyl-prolyl cis–trans isomerase

A N-nonacetylatedP17742 268373 17.8/7.88 20.2/7.80 12 50 129 9

Peptidyl-prolyl cis–trans isomeraseA N-acetylated

P17742 268373 17.8/7.88 20.4/7.40 14 43 128 VNPTVFFDITADDEPLGR 10

Peptidyl-prolyl cis–trans isomerase C P30412 19038 22.8/6.96 22.9/6.60 6 25 108 127Peroxiredoxin 1 P35700 18477 22.2/8.26 26.3/8.11 10 63 165 18Peroxiredoxin 2 Q61171 21672 21.6/5.2 25.1/4.92 8 47 87 59Peroxiredoxin 4 O08807 53381 31.0/6.67 28.6/6.03 7 29 137 134Phosphatidylethanolamine-binding protein P70296 23980 20.7/5.19 25.1/4.92 8 72 94 59Phosphoglycerate kinase 1 P09411 18655 44.4/7.52 45.5/7.74 8 35 119 90

40.2/5.68 11 37 137 108Pyruvate kinase isozyme M2 P52480 18746 57.7/7.42 40.0/6.02 10 27 148 145Rho GDP-dissociation inhibitor 1 Q99PT1 192662 23.4/5.12 28.4/4.97 7 51 89 15240S ribosomal protein SA P14206 16785 32.6/4.74 44.9 /4.71 8 29 96 157SEC13-related protein Q9D1M0 110379 35.4/5.15 37.8/5.00 7 30 82 154Selenide water dikinase 1 Q8BH69 109079 42.9/5.65 44.1/5.69 7 25 57 147Septin-11 Q8C1B7 52398 49.5/6.26 51.8/6.46 5 15 59 122Stathmin P54227 16765 16.1/5.76 20.6/5.60 5 31 97 126Stress-induced phosphoprotein 1 Q60864 20867 52.5/6.4 61.5/6.42 15 34 203 88Superoxide dismutase (Cu-Zn) P08228 20655 15.8/6.03 19.8/6.12 7 43 154 12T-complex protein 1 b subunit P80314 12461 57.3/5.98 54.7/6.10 7 23 73 104T-complex protein 1 e subunit P80316 12465 59.6/5.72 59.4/5.71 9 23 118 106



Table 3. Continued

Protein Acc.no.

EntrezGeneID


Sequencecoverage(%)

MASCOTScore


Transgelin 2 Q9WVA4 21346 23.4/6.83 25.0/6.83 10 55 161 100Translationally controlled tumor protein

(p23)P14701 22070 19.4/4.76 26.1/4.68 5 29 78 57

Triosephosphate isomerase P17751 21991 26.6/7.09 28.7/7.16 8 39 127 128Tropomyosin a 1 chain P58771 22003 32.8/4.66 34.9/4.75 15 44 179 42Tropomyosin a 3 chain P21107 59069 32.8/4.68 32.8/4.74 13 37 178 46Tropomyosin a 4 Q6IRU2 326618 28.3/4.65 32.4/4.69 9 38 140 45Tropomyosin b chain P58774 22004 32.8/4.66 37.9/4.67 12 32 127 40

36.2/4.75 12 35 179 41Troponin C, skeletal muscle P20801 21925 17.9/4.07 23.3/– 5 40 65 150Tubulin a 1 chain P68369 22142 50.1/4.94 38.7/5.36 8 30 91 29Tubulin b 5 chain P99024 22154 49.6/4.78 36.1/5.32 13 36 155 28

53.8/4.85 16 52 219 3941.5/5.62 14 41 187 14839.4/5.63 9 24 111 14933.9/4.93 11 26 142 156

Ubiquitin C-terminal hydrolase-L1 Q9R0P9 22223 24.8/5.14 28.6/5.09 10 60 152 153Vimentin P20152 22352 53.5/5.06 46.6/4.70 23 53 335 35

48.3/4.81 25 55 304 3650.6/4.88 19 50 286 37

OrganelleAcetyl-coenzyme A acetyltransferase 1

precursorgi)21450129 110446 44.8/8.71 43.8/7.86 6 26 79 91

Aconitase2, mitochondrial Q99KI0 11429 85.4/8.08 82.0/7.30 10 14 212 SQFTITPGSEQIRNAVTQE-FGPVPDTAR

83

ATP synthase a chain, mitochondrialprecursor

Q03265 11946 59.7/9.22 26.5/6.85 6 17 88 130

25.8/6.71 10 22 122 131ATP synthase b chain, mitochondrial

precursorP56480 11947 56.3/5.19 51.5/4.93 18 51 253 38

ATP synthase D chain, mitochondrial Q9DCX2 71679 18.6/5.53 25.1/5.38 8 63 127 64Calreticulin precursor P14211 12317 47.9/4.33 62/– 11 41 148 33

48.5/– 6 25 83 70Clathrin light chain A O08585 12757 25.5/4.45 35.4/– 5 14 80 43Endoplasmatic reticulum protein ERp29

precursorP57759 67397 28.8/5.9 30.1/6.07 6 20 66 FDTQYPYGEK 137

30.1/5.70 6 25 90 144GRP 78 (BiP) P20029 14828 72.4/5.07 62/– 26 47 339 51Hsp60 P19226 15510 60.9/5.91 58.4/5.28 14 43 165 55Protein disulfide isomerase precursor (PDI) P09103 18453 57.1/4.79 56.7/4.71 20 49 247 34Protein disulfide isomerase A3 precursor

(PDI)P27773 14827 56.6/5.99 57.2/5.78 17 40 208 52

57.3/5.71 11 26 86 11832.2/5.87 6 15 68 14232.4/5.70 7 16 79 14333.0/6.05 7 17 78 163

Protein disulfide-isomerase A6 precursor Q922R8 71853 48.1/5 43.8/6.24 5 15 46 TGEAIVDAALSALR 25Reticulocalbin-1 precursor Q05186 19672 38.1/4.7 44.4/4.57 8 33 108 164Reticulocalbin 3 precursor Q8BH97 52377 38.0/4.74 46.9/4.81 9 34 128 68

47.6/4.56 7 29 65 69Stress-70 protein, mitochondrial precursor P38646 3313 73.7/5.87 71.1/5.38 15 29 139 72Ubiquinol-cytochrome-c reductase

complex core protein I, mitochondrialQ9CZ13 22273 52.7/5.75 49.5/5.34 16 42 222 160



Table 3. Continued

Protein Acc.no.

EntrezGeneID


Sequencecoverage(%)

MASCOTScore


NucleusChromobox protein homolog 1 P23198 12417 20.8/5.13 26.8/4.71 99 IIGATDSSGELXFLMKKV-

EEVLEEEEEEYVVEK151

Heterogeneous nuclear ribonucleoproteinsA2/B1

O88569 53379 35.9/8.67 35.9/– 11 43 189 94

Heterogeneous nuclear ribonucleoprotein A3 Q8BG05 229279 39.6/9.1 38.3/8.46 9 28 117 93Heterogeneous nuclear ribonucleoprotein H0 P70333 56258 49.2/5.89 27.6/6.71 6 19 64 129Laminin A P48678 16905 74.2/6.54 67.1/7.09 15 27 60 ITESEEVVSR 84

66.8/6.71 9 18 91 8566.8/6.55 12 23 122 8667.0/6.49 8 17 93 87

Poly (rC) binding protein 1 P60335 23983 37.4/6.66 41.6/6.59 7 36 79 96Putative RNA-binding protein 3 O89086 19652 16.6/6.84 20.2/6.59 5 37 85 124

Cytosol and organellePeroxiredoxin 6 O08709 11758 24.7/5.72 28.4/6.22 10 48 137 138

Cytosol and nucleusCofilin, nonmuscle isoform P18760 12631 18.4/8.26 21.0/8.01 9 60 146 8DJ-1 protein Q99LX0 57320 20.0/6.32 26.1/6.05 7 30 68 136KH-type splicing regulatory protein Q99PF5 171137 76.8 80.0/6.72 10 24 120 79

80.1/6.55 9 17 120 110Nucleoside diphosphate kinase A P15532 18102 17.2/6.84 19.6/6.54 5 42 98 125Proteasome subunit a type 2 P49722 19166 25.8/8.42 27.6/6.71 6 45 87 129Proteasome subunit a type 5 Q9Z2U1 26442 26.4/4.74 29.5/4.61 9 56 122 56Protein C14orf166 homolog Q9CQE8 68045 28.1/6.4 28.9/6.36 8 48 115 140

Cytosol and cellular membraneNucleoside diphosphate kinase B Q01768 18103 17.3/6.97 20.2/6.59 7 46 69 11

SecretedBloodApolipoprotein A-1 Q00623 11806 30.06/5.64 27.2/5.33 9 31 153 26Fetoprotein a precursor P02772 11576 67.3/5.65 71.3/5.28 24 52 299 65

71.7/5.27 25 57 315 6679.5/5.19 20 42 264 6765.8/5.46 15 36 177 7369.0/5.34 14 32 341 7459.7/5.81 9 23 93 10558.7/6.09 10 20 122 119

Fibrinogen a polypeptide Q99K47 14161 61.3/7.16 59.5/6.99 9 19 92 8959.4/7.12 12 25 143 GLIDEANQDFTNR 10756.5/6.67 10 22 107 GLIDEANQDFTNR 11456.6/6.54 8 17 91 115

Fibrinogen b polypeptide Q8K0E8 110135 54.7/6.68 56.0/5.98 17 45 206 5355.9/6.15 9 25 118 103

Hemoglobin a chain P01942 15122 14.9/8.08 17.8/8.26 6 44 80 771.7/– 6 44 133 112

Hemoglobin b-1 chain P02088 15129 15.7/7.26 17.8/7.23 10 66 127 118.5/7.78 10 74 139 5

Hemoglobin b-2 chain P02089 15130 17.7/7.97 17.8/7.65 11 63 154 2Serotransferrin Q921I1 22041 76.7/6.94 19.2/7.79 8 9 71 4

18.8/7.47 7 11 58 676.0/6.55 23 36 297 8075.4/6.49 20 36 292 8176.3/6.45 18 33 245 82



Table 3. Continued

Protein Acc.no.

EntrezGeneID


Sequencecoverage(%)

MASCOTScore


64.0/6.34 10 20 119 10176.9/6.38 7 14 94 11170.8/6.38 16 24 207 11363.9/6.11 14 22 161 11658.7/6.09 10 16 126 11756.2/6.34 7 10 56 120

Serum albumin P07724 11657 68.7/5.75 18.4/6.42 9 13 93 329.6/6.47 11 25 130 1429.6/6.25 10 23 133 1525.0/6.76 12 25 135 1624.4/6.77 12 21 145 1719.1/5.08 13 21 117 1947.46.14 19 36 247 2434.4/5.57 7 16 66 2737.5/5.49 14 29 151 3037.5/5.49 24 50 277 3143.7/5.19 17 35 152 4743.8/5.27 11 26 123 4954.9/5.37 21 40 293 5426.8/4.98 11 20 111 6127.6/4.98 8 15 67 6262.8/5.41 8 20 81 7128.6/5.71 6 34 101 13229.3/5.92 6 34 105 13333.9/5.43 8 19 109 14646.1/5.42 13 27 178 159

Transthyretin P07309 22139 15.7/5.77 18.5/6.02 7 60 139 13

ECM proteinsCollagen a 1 (I) precursor Q810J9 12842 138.0/5.65 142.2/6.89 35 31 261024a) 32

35.7/4.95 7 5 60 155Collagen a 2 (VI) precursor Q02788 12834 109.8/5.95 125.1/6.03 8 9 – NFVINVVNR 75

124.8/5.95 8 9 58 76124.9/5.87 – Probable 77125.3/5.79 6 7 43 78

Matrilin 1 P51942 17180 54.4/7.88 54.9/7.51 9 27 96 FINQIVDTLDVSDR 123

UnknownHypothetical protein LOC70984 gi)19526926 70984 34.9/5.86 35.5/5.94 9 33 79 162

Acc. no., accession number in the Swiss-Prot/TrEMBL or NCBInr databases; Mth, theoretical molecular mass (kDa) of the full-length protein;Mob, observed molecular mass (kDa); pIth, theoretical pI of the full-length protein; and pIob, observed pI. Mob and pIob were determined whenpossible using BioRad 2-D molecular markers. Spot IDs correspond to those given in Fig. 2.ProFound software was used for spot 32 identification since it allows for hydroxylisine modification.a) Expectation value.

mosaicism, differential expression of the mutant allele, ormodifier genes have been suggested. OI, also known as brit-tle bone disease, is a skeletal dysplasia with autosomaldominant inheritance. The consortium for OI mutationreports dozen of instances of independent occurrence of thesame substitution at a particular glycine residue leading todivergent clinical outcomes especially in the a1(I) chain oftype I collagen [7, 11, 12]. Understanding the molecular basis

of phenotypic variability is relevant to elucidating the rela-tionship between cells and matrix, which will illuminate OIpathophysiology and may identify target proteins for devel-oping a new therapy for this disorder.

In the present study, we investigate the expression ofbone proteins in WT and Brtl mice with a lethal or nonlethalphenotype by using the complementary approaches ofmicroarray and proteomics.



Figure 3. Representative 2-DE images that displayed differentialexpression of some bone proteins. The gels were stained byammoniacal silver nitrate and then acquired by VersaDoc 3000.PDQuest software was used for the quantitative analysis. (A) Fi-brinogen a and b were more highly expressed in Brtl mice withlethal (MTL) and nonlethal (MTA) phenotypes, than in WT litter-mates. (B) a crystallin B chain expression was increased in MTA.The numbers refer to spot ID as shown in Table 3 and Fig. 2.

Table 4. Differential protein expression levels detected by 2-DEanalysis

Protein Acc. no. Fold difference 6 SEM Spot ID p value

Fibrinogen a Q99K47 MTL/WT = 3.73 6 1.01 89 0.004MTA/WT = 2.51 6 0.53 0.004MTL/WT = 4.48 6 0.98 107 0.003MTA/WT = 3.50 6 1.07 0.001

Fibrinogen b Q8K0E8 MTL/WT = 2.55 6 0.47 53 ,0.001MTA/WT = 1.62 6 0.27 0.08MTL/WT = 2.39 6 0.47 103 0.007MTA/WT = 1.87 6 0.40 0.06

a crystallin Bchain

P23927 MTA/MTL = 2.76 6 0.47 109 0.001MTA/WT = 1.6 6 0.99 0.156MTL/WT = 0.7 6 0.27 0.286

While we were mindful of the complexity of whole bonetissue, which includes a variety of cell types and a complexECM, we used bone tissue from our previously validatedmurine model for OI to investigate phenotypic variability invivo. As recently pointed out by Pastorelli et al. [29], the use ofin vitro models can easily tell us which proteins can beexpressed by a cell-type under a particular status, but only theanalysis of the full tissue can delineate the actual proteinprofile in vivo, which takes into consideration the tight net-works among the different cell types.

Figure 4. Representative 2-DE images that displayed expressionof fibrinogen a (FNa) and b (FNb) in plasma from mutant and WTmice. The gels were stained by ammoniacal silver nitrate andthen acquired by VersaDoc 3000. PDQuest software was used forthe quantitative analysis.

We used calvarial bone, which also expresses the bonedysplasia of OI that has been well described in long bones[30, 31], since the lethal outcome occurs within hours of birthand only calvarial bone can be cleanly dissected from sur-rounding connective tissue at that age.

We excluded differential expression of the collagengenes, col1a1 and col1a2 and differential expression of themutant or WT col1a1 alleles as the cause of the phenotypicvariability of BrtlIV. These results were obtained using real-time PCR and allele-specific probes with LNA nucleotides,techniques which are sensitive to even small variations.

Intracellular retention of mutant collagen and intra-cellular ER stress are common features in OI [32–34]. Wehave previously demonstrated their presence in Brtl fibro-blasts and osteoblasts, in vitro and in vivo in skin [15] (Forlinoet al. submitted). Here we demonstrated that differences inhandling the mutant collagen by the intracellular cellmachinery in ECM proteins expression and a delay in skele-tal development contribute to the modulation of the OI out-come.

In lethal animals, activation of the ER stress-relatedprotein Gadd153 was detected. CHOP/Gadd153 (growtharrest and DNA damage inducible gene) is a member ofthe C/EBP family of transcription factors which regulatecellular growth, differentiation, and death. Gadd153 isincreased at both transcriptional and protein levels in thecalvaria of lethal mice, but its expression is unchanged inskin and lung, which also synthesize large amount of typeI collagen. These data suggest a bone-specific ER stressresponse. While higher expression of Gadd153 could havebeen associated with greater retention of mutant collagenin lethal mice, this is not supported by in vitro biochemicaldata [18]. The trigger for Gadd activation must be a morecomplex secondary response to mutant collagen, perhapsreflecting other changes in mutant osteoblast metabolism.

In nonlethal BrtlIV, we found indeed an overexpressionof the heat shock protein aB crystallin. The a crystallinscomprise 35% of soluble proteins in the ocular lens and hadchaperon-like function. a-Crystallin B chain has previouslybeen demonstrated in various tissues, including bone, and inosteoblast-like cells. Since it occurs in nonlethal BrtlIV, it isparticularly interesting that overexpression of a crystallin



had been demonstrated to have the opposite effect from thatof Gadd153 and protects cells from apoptosis, which is rele-vant to the regulation of osteoblast number [35]. Additionalevidence indicates the involvement of aB crystallin in theubiquitine proteasome system and in the aggresomal re-sponse to misfolded proteins [36, 37]. The increase of theheat shock protein a crystallin B chain both at transcriptionaland protein levels suggests the activation of an intracellularsystem in the bone cells of nonlethal BrtlIV, which is protec-tive against an apoptotic response to ER stress and proteinmisfolding.

ECM structure and composition are known to be abnor-mal in OI [38]. Our study demonstrated differential expres-sion of minor ECM components in mice with lethal or mod-erate severe outcome compared to the WT. In particular wedetected lower expression level of three ECM proteins:matrilin 4, Magp2 and thrombospondin 3. Matrilin 4 is amember of the matrilins family of noncollagenous ECMproteins. All matrilins are expressed in skeletal tissueincluding calvaria. Matrilin 4 is the most ubiquitous of thematrilins, with a broad tissue distribution. It has been sug-gested that matrilins function as a bridge between differentECM macromolecules [39]. Magp2 is an ECM protein with anarrow tissue distribution. This is the first report of its pres-ence in murine calvarial bone. Magp2 function is not wellunderstood, but it appears to play a role in cell–microfibrilinteractions [40]. Finally, thrombospondin 3 is a heparinbinding protein. Recently, Hankenson et al. [41] generated aknock-out murine model for this protein and demonstratedthat thrombospondin 3 has a role in the regulation of skeletalmaturation. In the lethal phenotype of Brtl, the down reg-ulation of these ECM protein expressions could represent acompromise of cellular viability, although the specific de-crease in these proteins is more supportive of the existence ofa specific feedback response to their expression from abnor-mal collagen matrix.

A third factor involved in the variable outcome in BrtlIVmice is a difference in skeletal development. The calvarialsamples collected for this study also included the calvarialsutures. The higher level of cartilage specific proteins, suchas Prelp and Bmps, identified in lethal animals, is consistentwith delay in bone maturation in mice with a lethal outcome.The specific increase in the expression of fibrinogen a and bfound in all BrtlIV compared to WT was also intriguing. Fi-brinogen is a secreted blood protein involved in the coagula-tion cascade; it is also present on osteoclast membranes infetal bone [42, 43]. The role of surface fibrinogen in osteo-clasts is unknown, but it may be necessary for aggregation orcellular adhesion of mononuclear osteoclast precursors priorto fusion and for formation of multinuclear osteoclasts. Fur-thermore, osteoclasts may also require fibrinogen to bind tothe bone surface. Our data on Brtl pups suggest an increasein osteoclast recruitment; this is in agreement with histo-morphometric observations in mature mice [7] and withobservations of increased bone remodelling in children withOI.

Very little proteomic data are available for whole bonetissue. Previously proteomic data on bone tissue focusedmainly on in vitro systems using bone cells [44–47]. Tworecent papers described a proteome analysis of bone from rat[48] and murine [29] femura and humeri, respectively. Inboth cases, 2-D gels and MS were used although no com-prehensive map was presented.

We performed the first investigation of the murine cal-varial bone tissue using proteomic techniques, and gener-ated the first 2-D map for such a tissue because it was notavailable in public databases. We were able to identify mainlyintracellular proteins (85%).

In conclusion, our study provides new insight into theunderstanding of OI phenotypic variability using two pow-erful and complementary approaches, microarray and prote-omic analysis. Factors related both to intracellular machineryand ECM appear to determine the response of bone cells tothe presence of mutant collagen and to influence the phe-notypic OI outcome. The in vivo function of such proteins aspotential modifier genes is under investigation in our labo-ratory.

This work was supported by the Osteogenesis ImperfectaFoundation (USA), Fondazione Cariplo (Milan, Italy), FondoAteneo per la Ricerca (FAR) and MIUR PRIN (grant. no.2006050235 to A. F.).

5 References

[1] Byers, P. H., Cole, W. G., in: B. Steinmann (Ed.), ConnectiveTissue and Heritable Disorders, Wiley-Liss, New York 2002,pp. 385–330.

[2] Marini, J. C., in: H. B. Jensen (Ed.), Nelson’s Textbook ofPediatrics, Saunders, Philadelphia, PA 2004, pp. 2336–2338.

[3] Rick, A., Crick, F. H. C., J. Mol. Biol. 1961, 3, 483–506.

[4] Fraser, R. D. B., MacRae, T. P., Suzuki, E., J. Mol. Biol. 1979,129, 463–481.

[5] Prockop, D. J., Kivirikko, K. I., Annu. Rev. Biochem. 1995, 64,403–434.

[6] Forlino, A., Marini, J. C., Mol. Genet. Metab. 2000, 71, 225–232.

[7] Marini, J. C., Forlino, A., Cabral, W. A., Barnes, A. M. et al.,Hum. Mutat. 2007, 28, 209–221.

[8] Sillence, D. O., Rimoin, D. L., Danks, D. M., Birth DefectsOrig. Artic. Ser. 1979, 15, 113–129.

[9] Sillence, D. O., Senn, A., Danks, D. M., J. Med. Genet. 1979,16, 101–116.

[10] Byers, P. H., Wallis, G. A., Willing, M. C., J. Med. Genet. 1991,28, 433–442.

[11] Dalgleish, R., Nucleic Acids Res. 1998, 26, 253–255.

[12] Dalgleish, R., Nucleic Acids Res. 1997, 25, 181–187.

[13] Cabral, W. A., Marini, J. C., Am. J. Hum. Genet. 2004, 74,752–760.



[14] Pereira, R., Halford, K., Sokolov, B. P., Khillan, J. S. et al., J.Clin. Invest. 1994, 93, 1765–1769.

[15] Forlino, A., Porter, F. D., Lee, E. J., Westphal, H. et al., J. Biol.Chem. 1999, 274, 37923–37931.

[16] Sarafova, A. P., Choi, H., Forlino, A., Gajko, A. et al., Hum.Mutat. 1998, 11, 395–403.

[17] Kozloff, K. M., Carden, A., Bergwitz, C., Forlino, A. et al., J.Bone Miner. Res. 2004, 19, 614–622.

[18] Kuznetsova, N. V., Forlino, A., Cabral, W. A., Marini, J. C. etal., Matrix Biol. 2004, 23, 101–112.

[19] Lammi, M. J., Hayrinen, J., Mahonen, A., Electrophoresis2006, 27, 2687–2701.

[20] Zhao, H., Granberg, F., Elfineh, L., Pettersson, U. et al., J.Virol. 2003, 77, 11006–11015.

[21] Gorg, A., Postel, W., Gunther, S., Weser, J. et al., Electro-phoresis 1988, 9, 37–46.

[22] Bjellqvist, B., Sanchez, J. C., Pasquali, C., Ravier, F. et al.,Electrophoresis 1993, 14, 1375–1378.

[23] Oakley, B. R., Kirsch, D. R., Morris, N. R., Anal. Biochem.1980, 105, 361–363.

[24] Hochstrasser, D. F., Patchornik, A., Merril, C. R., Anal. Bio-chem. 1988, 173, 412–423.

[25] James, P., Quadroni, M., Carafoli, E., Gonnet, G., Protein Sci.1994, 3, 1347–1350.

[26] Hellman, U., Wernstedt, C., Gonez, J., Heldin, C. H., Anal.Biochem. 1995, 224, 451–455.

[27] Lalouel, J. M., White, R., in: A. E. H. Emery, D. L. Rimoin, J. A.Sofaer, I. Black (Eds.), Principles and Practice of MedicalGenetics, Vol. 1, Churchill-Livingston, New York 1990, pp.149–164.

[28] Ott, J., in: R. Edition (Ed.), Analysis of Human Genetic Link-age, The Jhon Hopkins University Press, Baltimore, MD1991, pp. 146–164.

[29] Pastorelli, R., Carpi, D., Airoldi, L., Chiabrando, C. et al., Pro-teomics 2005, 5, 4936–4945.

[30] Lund, A. M., Muller, J., Skovby, F., Arch. Dis. Child 1999, 80,524–528.

[31] Kaplan, S. B., Kemp, S. S., Oh, K. S., Radiol. Clin. North. Am.1991, 29, 195–218.

[32] Lamande, S. R., Chessler, S. D., Golub, S. B., Byers, P. H. etal., J. Biol. Chem. 1995, 270, 8642–8649.

[33] Fitzgerald, J., Lamande, S. R., Bateman, J. F., J. Biol. Chem.1999, 274, 27392–27398.

[34] Delahunty, M., Bonifacino, J. S., Connect Tissue Res. 1995,31, 283–286.

[35] Mehlen, P., Schulze-Osthoff, K., Arrigo, A. P., J. Biol. Chem.1996, 271, 16510–16514.

[36] Boelens, W. C., Croes, Y., de Jong, W. W., Biochim. Biophys.Acta. 2001, 1544, 311–319.

[37] Chavez Zobel, A. T., Loranger, A., Marceau, N., Theriault, J.R. et al., Hum. Mol. Genet. 2003, 12, 1609–1620.

[38] Fedarko, N. S., Vetter, U., Robey, P. G., Connect Tissue Res.1995, 31, 269–273.

[39] Klatt, A. R., Paulsson, M., Wagener, R., Matrix Biol. 2002, 21,289–296.

[40] Gibson, M. A., Finnis, M. L., Kumaratilake, J. S., Cleary, E. G.,J. Histochem. Cytochem. 1998, 46, 871–886.

[41] Hankenson, K. D., Hormuzdi, S. G., Meganck, J. A., Born-stein, P., Mol. Cell. Biol. 2005, 25, 5599–5606.

[42] Athanasou, N. A., Quinn, J., Heryet, A., McGee, J. O., J. Cell.Sci. 1988, 89, 115–122.

[43] Quinn, J. M. W., Athanasou, N. A., McGee, J. O., His-tochemistry 1991, 96, 169–176.

[44] Hwang, R., Lee, E. J., Kim, M. H., Li, S. Z. et al., J. Biol. Chem.2004, 279, 21239–21247.

[45] Chang, E. J., Kwak, H. B., Kim, H., Park, J. C. et al., FEBS Lett.2004, 564, 166–170.

[46] Kubota, K., Wakabayashi, K., Matsuoka, T., Proteomics 2003,3, 616–626.

[47] Roman-Roman, S., Garcia, T., Jackson, A., Theilhaber, J. etal., Bone 2003, 32, 474–482.

[48] Fan, Y., Liu, J., Wang, S., Wang, H. et al., Life Sci. 2005, 76,2893–2901.


Documents

Differential expression of both extracellular and intracellular proteins is involved in the lethal or nonlethal phenotypic variation of BrtlIV, a murine model for osteogenesis imperfecta