Osteopenia in Sparc (osteonectin) deficient mice: characterization of phenotypic determinants of femoral strength and changes in gene expression

Abstract

Sparc null mutants have been generated independently via targeted mutations in exons 4 and 6. Previous studies have identified low turnover osteopenia in the 129Sv/C57Bl/6 exon 4 knockout. Since both Sparc null mutations result in complete absence of Sparc protein, similar phenotypic outcomes are likely. However, genetic background (strain) and/or linkage disequilibrium effects can influence phenotype. Different inactivating mutations should be tested in various mouse strains; similar phenotypic outcomes can then confidently be assigned to the mutated gene. We have evaluated the bone phenotype in the 129Sv/EvSparc^tm1cam exon 6 knockout at 4 and 9 months, using physical measurement, mechanical strength tests and DXA scanning. We have also quantified bone marrow adiposity and circulating leptin levels to assess adipose tissue metabolism. 129Sv/EvSparc^tm1cam null mice show decreased bone mineral density, bone mineral content, and increased mechanical fragility of bone, in line with previous studies. Differences were also noted. Increased body weight and levels of bone marrow adiposity, but decreased circulating leptin concentrations were identified at 4, but not 9 months and 129Sv/EvSparc^tm1cam null mice also had shorter femurs. Molecular phenotyping was carried out using mouse HGMP NIA microarrays with cortical femur samples at various ages, using semi-quantitative RT PCR validation. We identified 429 genes highly expressed in normal bone. Six genes (Sparc, Zfp162, Bysl, E2F4, two ESTs) are differentially regulated in 129Sv/EvSparc^tm1cam cortical femur versus 129Sv/Ev controls. We confirm low turnover osteopenia as a feature of the Sparc null phenotype, identifying the usefulness of this mouse as a model for human osteoporosis.

INTRODUCTION

Total bone mass varies throughout life and is regulated via relative rates of bone resorption and deposition. Osteoblasts deposit bone (and are responsible for its developmental formation), while osteoclasts resorb bone, resulting in the constant remodelling of skeletal homeostasis. Osteoblasts are derived from mesenchymal cell populations (20),while osteoclasts are formed via fusion of cells of the monocyte/macrophage family (37); precursors of both are present in the adult bone marrow (27). Osteoporosis is a disorder of skeletal turnover; bone is resorbed faster than it is reformed as a result of imbalances in the activity, or relative numbers of, osteoblasts and osteoclasts (37). Risk factors include gender, hormonal status, race, nutritional status, inadequate resorption/retention of calcium via the digestive system and kidney, lack of weight bearing exercise and low peak bone mass in youth.

Human and animal linkage studies have identified multiple genetic loci that influence bone mass, (which is highly variable in humans and between different inbred mouse strains (1,4, 17,,24, 32)). Polymorphisms in the COL1A1, TGFβ1, SOST,VDR, ERα, LRP5,BMP2, and IL-6 genes, among others, have been implicated in increased osteoporosis risk (17,21,32). Moreover, genetically defined mouse mutants have been of particular utility in identifying underlying causes of bone thinning in humans. Gain of function p53 mutants, PASG hypomorphs, PolgA knock-in mutants and knockouts of BGN, SHIP, OPG, c-Abl, Irs1, klotho, Ku86, XPD and Sparc all show reductions in bone density (15,30,35,38).

Sparc (a.k.a osteonectin or BM40) is a calcium, hydroxyapatite and collagen binding protein, implicated in cell proliferation, tissue morphogenesis and repair and the modelling of extracellular matrix⁹. It is widely expressed in development and is present at high levels in bone, possibly linking organic and mineral phases of bone tissue (27). Two independent Sparc knockout lines have been generated. One, Sparc^tm1cam, was generated on both a purebred 129Sv/Ev and a mixed 129Sv/Ev/Mf1^bb background, with an inactivating mutation in exon 6 of the Sparc gene (19). The other knockout (not specifically named but referred to as the Sparc exon 4 knockout) was generated on a 129Sv/C57Bl/6 background with an inactivating mutation in exon 4 (2). Both Sparc null mutations have been shown to result in complete absence of Sparc protein via Western blot. Furthermore, testing of a wide variety of tissues and cell lines identified no splice variants of Sparc in mouse (29). Therefore, similar phenotypic outcomes would be likely. Initial analyses revealed subcortical posterior cataract in both lines (2,19); differences in the timing of cataract onset is most probably genetic background dependent (28). Progressive low turnover osteopenia has been identified in the 129Sv/C57Bl/6 Sparc exon 4 knockout (5,6,7,15,16). The numbers of osteoblasts and osteoclasts have been shown to be reduced, resulting in a decrease in the rate of bone formation which exceeds the reduction in bone resorption (15). Increased mineral content, crystallinity and collagen maturity were noted, consistent with decreased bone formation and remodelling (5). These Sparc^-/- mice attain a lower peak bone mass with progressive bone loss thereafter (15). Elevated serum leptin concentrations with advancing age, and excessive fat deposits have also been reported; marrow stroma from exon 4 Sparc -/- mice contained fewer osteoblasts and showed an increased tendency to form adipocytes (30,35,38). Interestingly, others have reported that Sparc seems to play a role in adipose tissue physiology (11), and Sparc expression is also altered in at least some forms of human obesity (23).

However, genetic background would be expected to influence this phenotype; C57Bl/6 mice have comparatively low bone density, while that of 129 strains is typically high (3,4). We have now studied the bone phenotype in purebred 129Sv/Ev Sparc^tm1cam null mice. Given data suggesting increased adiposity at the expense of bone formation, we have also assessed body weight, bone marrow adipocyte size and number and circulating leptin concentrations. Measurements of femoral size, strength, and mineral content have confirmed that loss of Sparc causes an osteopenic phenotype, regardless of the nature of the null mutation or the genetic background of the mice. Some differences have been noted between the two Sparc knockout lines, namely differences in femoral length and leptin metabolism not found in previous work with the exon 4 Sparc knockout animals.

We have also used our purebred mice with matched controls to carry out microarray analysis to assess knock-on changes in femoral gene expression in ageing Sparc null mice. Six genes, including Sparc, are downregulated in 129SvEv Sparc^tm1cam null mice; we have also used data from controls to identify 429 ESTs that are highly expressed in normal bone. These may be of importance to the aetiology of bone loss during ageing, and also provide a valuable resource for identifying candidate genes for association or linkage studies.

MATERIALS AND METHODS

Mouse nomenclature and strain choice

129SvEv Sparc^tm1cam null mice: 129SvEv refers to mice of inbred strain 129 and substrain SvEv. Sparc^tm1cam refers to “Sparc, targeted mutation 1, University of Cambridge”. This targeted mutation is in exon 6. (The exon 4 knockout has not been similarly named, and is referred to here as the exon 4 knockout). Mice used in this study were derived either from mutant homozygote lines or from 129Sv/Ev controls. Male mice were used in order to minimize the effect of hormonal variation on gene expression levels. For all measurements, experimental and control male animals were sacrificed at 17 and 40 week time points (4 and 9 months). Sample sizes of at least 6 controls and 6 knockouts were used for each assay. Our inbred mice have been inbred for generations prior to the introduction of the Sparc^tm1cam mutation, moreover, the ES cell lines used to generate this line were themselves of 129SvEv origin. There should therefore be no genetic variation between control and experimental lines other than the Sparc^tm1cam mutation. This should eliminate the possibility of error arising from linkage disequilibrium or from genetic drift between control and experimental lines derived from a mixed background. Elimination of such sources of error is particularly important when assessing results from microarray analyses.

Animal husbandry and tissue extraction

Mice were maintained and killed under Home Office licence in accordance with British law (comparable with U.S. Public Health Service Policy on Humane Care and Use of Laboratory Animals). Transgenic and control animals were maintained on RM3 diet (1.15% calcium, 0.82% phosphorus, 4088.65I.U./kg Vitamin D, Special Diet Services, Witham, Essex CM8 3AD, UK), ad libitum. Most animals (except for those used in leptin assays) were sacrificed via cervical dislocation. Whole femurs and tibias were dissected free of surrounding tissue. Bones intended for measurement and/or structural assay were wrapped in sterile PBS soaked gauze and frozen at -20°C prior to assessment. Tibiae to be sectioned prior to adipocyte counting were placed in 4% paraformaldehyde (PFA). Trunk blood samples for the determination of circulating leptin levels were obtained from halothane anaesthetised mice via decapitation.

Leptin assay

Blood samples were collected in heparinised tubes and separated plasma was stored at - 20°C prior to determination of plasma leptin concentrations via radioimmunoassay (Linco Research Inc, St. Charles, MO, USA). Intra-assay CV was 4.3%

Bone marrow adipocyte counting

Tibiae were fixed in 4% paraformaldehyde (PFA) for 2 days after dissection. The PFA was then replaced with a 10% EDTA solution (in 0.3 M NaOH). This solution was refreshed every second day for 3 weeks. Following decalcification, the final EDTA wash was replaced with 70% ethanol. Bones were then embedded in paraffin wax, sectioned, and stained with toluidine blue. Sections were visualised using a Leica DMLB microscope and photographed with a Leica DFC300FX camera. Image analysis was carried out using Scion Image. Longitudinal sections of mid-diaphyseal marrow were assessed for adipocytes following the method previously described by Gevers et al. 2002 (18). The total area counted was measured, along with adipocyte number, the area of each adipocyte, total adipocyte area and % of the field occupied by adipocytes. 3 sections were counted per animal; sections from 6 animals were assessed for each of the following 4 sample types; 4 month male 129SvEv, 4 month male 129SvEv Sparc^tm1cam, 9 month male 129SvEv, 9 month male 129SvEv Sparc^tm1cam. Statistical analysis of results was carried out using GraphPad Prism v. 2.0.

Femoral length, strength and morphology

Previously frozen femurs were thawed at room temperature, and the lengths measured with a hand held micrometer. Each bone was then loaded in three-point bending between 2.5 mm diameter rollers (6.5 mm apart), with the middle roller 3.25 mm from the outer rollers and positioned over the thinnest part of the femoral shaft, level with the distal end of the lateral ridge that runs along the proximal part of the femur toward the greater trochanter. The posterior aspect of the condyles rested on the side of the outer roller, and the bones were orientated such that they were loaded in a roughly posterior direction. Each bone was loaded at a crosshead speed of 2 mm/minute until failure, with load and displacement data recorded by the computer controlled testing machine (Lloyd LRX tensile testing machine with 100N load cell; Lloyd Instruments, Segensworth, Hants, UK). Mid-diaphyseal cortical medio-lateral (M-L) and anterior posterior (A-P) diameters, and lateral, medial, anterior, and posterior wall thicknesses were measured at the fracture site using a Pye travelling microscope. Using these measurements and simple beam theory, ultimate tensile stress (UTS) was calculated using:

$$\sigma =My∕I$$

where the bending moment, M, is one-half the applied load multiplied by the distance from the central support, y is one-half the outside depth, and the second moment of area, I, is given by:

$$I=\frac{\pi }{64}({b_{\mathrm{o}}{d}_{\mathrm{o}}}^3-{b_i{d}_i}^3)$$

where b and d are the breadth and depth of the cross-section, respectively, and the subscripts o and i indicate the outside and inside dimensions, respectively.

Femoral mineralisation

Bone mineral content (BMC) was measured by DXA using the Lunar Pixi small animal scanner. The accuracy of this technique in measuring calcium content was confirmed in preliminary studies, in which a highly significant correlation between femoral total BMC and ash weight (r = 0.86; p < 0.0001) was obtained (34). The CV for femoral bone mineral density (BMD) for five repeated scans of 30 femurs, with repositioning between scans, was 2.7%. Bones were thawed at room temperature for 30 minutes before measurement and aligned anterior-posteriorly relative to the scanning beam. Measurements of total BMC (g) and scanned bone area (BA; cm²) were made, and areal BMD (aBMD; g/cm²) was calculated as BMC/BA. In addition distal femur aBMD were determined by measuring BMC in a fixed area (0.03 cm²) of each region of interest (ROI).

Statistical analysis

All data are presented as mean ± SD. The differences between 2 experimental groups were compared by the unpaired Student t-test (*, ** and *** p<0.05, 0.01 and 0.001 respectively vs control of the same age; +, ++ and +++ p<0.05, 0.01 and 0.001 respectively vs same animal at different age).

RNA extraction for array analysis

Whole femurs were dissected free of surrounding tissue. Bone marrow was flushed via needle and syringe, using 0.5ml PBS per femur. Cellular components of bone marrow were isolated via centrifugation for 5 minutes at 2000rpm. Pellets were resuspended in TRIzol (Invitrogen, Paisley, UK) prior to RNA extraction. Flushed femurs were also retained. A 0.4 cm section of the femoral midshaft was obtained in order to minimize contamination by muscle and cartilage adhering to femoral epiphyses and standardise anatomical localization. Material from 2-6 femoral cortices was pooled in RNAlater (Ambion, Huntingdon, Cambs, UK). Where tissue from more than one animal was used, material from litter mates was pooled. Femoral midshaft samples were later transferred into TRIzol (Invitrogen, Paisley, UK) and immediately homogenized using a Yellowline DI18 basic electronic homogenizer (Yellowline, IKA, Staufen, Germany) prior to RNA extraction via the manufacturer’s protocol. RNA samples were quantitated using formaldehyde gel electrophoresis and spectrophotometry (Camspec, Sawston, Cambs., UK). Different pooled RNA samples were used for labelling array repetitions. 10ug of total RNA was labelled with either Cy3 or Cy5 dyes using the CyScribe labelling system (GE Healthcare, Chalfont St.Giles, Bucks, UK), according to the manufacturer’s protocol. 1ul of labelled cDNA was combined with 2ul 50% glycerol, run on a ‘John gel’ (a microscope slide sized, 1.5% agarose gel) and scanned using a GeneTac LS IV scanner (Genomic Solutions, Huntingdon, Cambs. UK) in order to assess successful incorporation of label. Control and experimental samples were then combined and prepared for hybridization.

Arrays: hybridization, image and data analysis

Microarray analysis was performed using the mouse NIA 15000 set from the Human Genome Mapping Project (HGMP), as described at the following website: (http://www.hgmp.mrc.ac.uk/Research/Microarray/HGMP-RC_Microarrays/array_description_files.jsp ).

The following comparisons were carried out:

1. 17 week 129SvEv Sparc^tm1cam null, male vs. 17 week 129Sv/Ev wild type, male, bone marrow

2. 17 week 129SvEv Sparc^tm1cam null, male vs. 17 week 129Sv/Ev wild type, male, femoral midshaft

3. 40 week 129SvEv Sparc^tm1cam null, male vs. 40 week 129Sv/Ev wild type, male, bone marrow

4. 40 week 129SvEv Sparc^tm1cam null, male vs. 40 week 129Sv/Ev wild type, male, femoral midshaft

Hybridization and scanning were carried out as previously described¹⁷. Arrays were repeated 5 times with fluor switching, in order to counteract any issues of dye bias that may have arisen from direct labelling. As the arrays contained duplicate spots, this allowed analysis of 10 spots per EST. Repetitions was derived from different RNA samples, in order to control for biological variation. Scanned images were stored and filtered, then analysed using the GeneTac Analyser spot finding software (Genomic Solutions, Huntingdon, Cambs. UK). Data were analysed as previously described (28), but, in brief, ESTs that were deemed to be significantly differentially regulated were changed in expression levels by at least 2 fold in 8 out of 10 replicates, were above background plus 2 standard deviations in at least one channel and showed a delta value of at least 0.5 using Significance Analysis of Microarrays software (http://www-stat.stanford.edu/~tibs/SAM ). These ESTs were subjected to bioinformatic analysis and were assessed via semi-quantitative RT-PCR in order to confirm array results. Data presented here are in full compliance with MIAME standards (8). Data are available from the GEO database, under series number GSE8381. We have also exceeded the requirements of MIAME in PCR testing every potentially differentially regulated gene. We also analysed the femur dataset from the control animals in order to identify genes highly expressed in normal bone. ESTs that were expressed above background + 2SD in all five array repetitions, in both 4 month and 9 month animals were selected. These were analysed using DAVID (http://niaid.abcc.ncifcrf.gov/). Gene details were later checked and amended where necessary, with reference to the NCBI website (http://www.ncbi.nlm.nih.gov ). Gene details are presented in Supplementary Table 1.

RT PCR array confirmations

We tested all genes arising from the four experiments using semi-quantitative RT-PCR. RNAs were quantified as described above. Even quantities of control and experimental RNA (usually 2-4μg) were treated with DNAfree (Ambion) according to the manufacturer’s protocol. Reverse transcription (RT) reactions were carried out using the Superscript II First Strand Synthesis System for RT PCR (Invitrogen), according to the manufacturer’s instructions. A “no RT” control corresponding to each sample was also produced; these were treated in exactly the same way as the samples except that Superscript II reverse transcriptase (Invitrogen) was not added. Standard primers were used for housekeeping genes; beta-actin, beta-microglobulin and GAPDH (see Table 1 for primer sequences). We used 3 housekeeping genes in conjunction in order to ensure that biased expression of one would not adversely affect results. Moreover, the 3 housekeeping genes used were present on the NIA arrays and were not significantly altered in expression. Primers for ESTs were designed as previously described (28) (see also Table 1). PCRs were initially carried out at the optimum Tm indicated, including No RT controls. Conditions were varied in order that only single bands appeared and the number of cycles for minimum visibility on a gel was always identified. We ensured that PCR bands obtained for housekeeping controls were even at minimum visibility cycles before testing other genes. In order to minimise false positive results, we used at least two sets of Sparc null and 2 sets of control cDNAs from different animals. Variation was observed in both sets in order for a gene to be confirmed as variably expressed. PCRs were carried out in 20μl volumes using 0.025μmol concentrations of each dNTP (100mM dNTP set, Invitrogen) and 1 unit Taq DNA polymerase (Sigma) per reaction. PCRs were usually carried out using PCR buffer (Sigma) containing 15mM MgCl₂, however, we varied magnesium concentrations where necessary in order to optimise results. Annealing temperature and the number of cycles were varied, up to a maximum of 50 cycles; finally, some primers were also redesigned. Final PCR conditions are indicated for confirmed genes in Figure 4. PCR changes were quantified from .tif files using Scion Image (Scion corporation), via the Scion Image protocol for analysing electrophoretic gels. PCR intensity changes (Figure 4) were calculated via the following equation:

$$\frac{\left(\text{Sparc average band intensity/wild type average band intensity}\right)}{\left(\text{Sparc average housekeeping gene intensity/wild type average housekeeping gene intensity}\right)}=\mathrm{X}$$

Changes <1 were converted to fold changes as follows: -1/X

Table 1. Primers used for amplification.

Genes detected after array analysis as significantly differentially regulated in 129SvEv Sparc^tm1cam null bone. All significant genes are listed here, including those NOT eventually confirmed by RT-PCR. The columns list the GenBank accession numbers of the EST originally printed on the array, the gene to which this EST is homologous (if any), and the forward and reverse primer sequences. Our standard housekeeping gene primer sequences are listed separately at the bottom. The final date of search is given beside uncharacterised ESTs

4 month bone
Accession	Gene	Forward primer	Reverse primer
BG082955	Pabpc4	GGTGGATGATGGGAATCTGAA	TGCACGGCCTACAAATATGG
BG075669	EST (26.2.7)	TTTATACAGCTATACAAGTT	GGTAAAGAATTTTGTCAGTC
BG072511	EST (26.2.7)	CGAGACAGGGTTTCTCTGTGTAGC	TGGGCGGTGGTGGCACACAC
BG069910	Canx	CAGGGTCCTTTACATTACAACTGC	TTTAAGAACTTGATCTGTGATTTCCTC
BG064802	Sparc	GCTGTGTTGGAAACGGAGTTG	CTTGCCATGTGGGTTCTGACT
9 month bone
Accession	Gene	Forward primer	Reverse primer
AW538347	E2F4	CAGGAGCTGGAGCCATTCTT	TCAGAGAGAGGGTGGTGCTG
BG065404	Esdn	GAGTGCAAGGAGAGCAGTGG	CGCTTAGTGGGCAGAGGGTA
BG065585	EST (26.2.7)	AACAGAGCTGAGTCTGTTGTG	TGGAGAGATAGATGGCTCAG
BG066371	EST (26.2.7)	CCGGCTCAACATCCTTTCTT	GGAACATCCCCAAACTCACA
BG066667	Mtap4	CCTTGCCTCTTCTGGAACTCA	TCCATGCTGCTAATCCTAGCC
BG066811	Ccdc96	TCCTGTCAAGGGCTGTCAAG	GTGGGCTTCTGGGTAAGGAA
BG077386	Zfp162	GACGCGCAGCATTACCAATA	CCCAGAGGTAGAGCCCACAG
BG079111	Wnt3a	TCTTAAATCCACGCCCACAG	AGGTTGGAGGGGCTCTGTCT
BG079188	Bysl	TGAGGGCTAGGCTTACAATCTTT	TGAACGTGATACAGCAAGGTGA
AW548270	Tpm1	ACTGGGCGAATTGCTTCTGT	GGACCACGCTCTCAACGATA
BG071230	Rbbp5F	AGCTGAGAGTGGACGGAAGG	TCCCTGTGTGAAGTGGCTTG
BG072628	Scmh1	AGGGAGAGGGGAAACTCCAT	TTGGAAGAGGGAGACCATCC
BG075868	Hebp1	GCCTTCACAAGCCAGACCTC	CGGGAGGGCATCACTGTCTA
BG087183	Hnrpr	CAAGTGAGGGCTTGAAAACG	GCCCAATAATGTCCAAGAGCA
BG064802	Sparc	GCTGTGTTGGAAACGGAGTTG	CTTGCCATGTGGGTTCTGACT
Housekeeping genes
Gene	primer name	Forward primer	Reverse primer
β-actin	Actin	CGTGGGCCGCCCTAGGCACCA	TTGGCCTTAGGGTTCAGGGGG
β-μ-globulin	BetaUG	GCTATCCAGAAAACCCCTCAA	CATGTCTCGATCCCAGTAGACGGT
Gapdh	Gapdh	ACCACAGTCCATGCCATCAC	TCCACCACCCTGTTGCTGTA

Figure 4.

RT-PCR confirmation of genes differentially regulated in 4 and 9 month 129SvEv Sparc^tm1cam null mice. Gene name, number of PCR cycles and annealing temperature used are given to the right. Lane labels identify the following: C = control, 129SvEv bone. Sp = 129SvEv Sparc^tm1cam null bone. NC and NSp indicate the corresponding No RT controls. Two sets of samples from different control and experimental animals were used. β-actin, βμglobulin and Gapdh are housekeeping controls. Residual Sparc expression is often detected by PCR in knockouts, but Western blots have confirmed the absence of functional protein (19).

RESULTS

Body weight and adiposity

The body weight of wild-type 129SvEv mice increased by 20% from 4 months to 9 months (p<0.001; Fig 1A) and this gain was accompanied by a 40% increase in circulating leptin (p<0.05; Fig 1E). At 4 months 129SvEv Sparc^tm1cam mice were significantly (20%) heavier than their wild type counterparts (p<0.001; Fig 1A), but this difference was not sustained, 129SvEv Sparc^tm1cam mice being similar to their wild type counterparts at 9 months (Fig 1A). Despite being heavier than their wild type counterparts at 4 months, circulating leptin was 40% lower in 129SvEv Sparc^tm1cam mice than in wild type animals (p<0.01; Fig 1E). Plasma leptin concentration increased at a much higher rate in 129SvEv Sparc^tm1cam mice than control animals between 4 and 9 months of age, however, such that leptin concentrations were not significantly different between the 2 groups at 9 months of age.

Figure 1.

Development of body weight, bone marrow adiposity, and circulating leptin concentrations in control (■) and 129SvEv Sparc^tm1cam null mice. Parameters shown are A) body weight, B) bone marrow adipocyte number, C) bone marrow adipocyte size, D) % area occupied by adipocytes, and E) plasma leptin concentrations. Values shown are mean ± SD. (n = 6 - 9; *, ** and *** p<0.05, 0.01 and 0.001 respectively vs control of the same age; +, ++ and +++ p<0.05, 0.01 and 0.001 respectively vs same animal at different age).

In the tibial marrow compartment total marrow adiposity did not change between 4 and 9 months in 129 SvEv mice (Fig 1D). At 4 months of age marrow adiposity was doubled in 129SvEv Sparc^tm1cam mice (p<0.05), but declined significantly by 9 months (p<0.05), becoming similar to that in their wild-type counterparts (Fig 1D). These differences in marrow adiposity were not dependent upon any changes in the number of adipocytes (Fig 1B), but were entirely attributable to the parallel differences in adipocyte size, adipocyte size being increased by 55% at 4 months (p<0.01) and declining by 25% at 9 months (p<0.05; Fig 1C).

Femoral length and morphology

In 129SvEv mice femoral length increased by 4% between 4 and 9 months (p<0.01; Fig 2A). Over the same period, there was no significant increase in femoral length in 129SvEv Sparc^tm1cam mice, femoral length becoming 4% shorter by 9 months (p<0.001; Fig 2A). Measurement of mid-diaphyseal femoral diameters revealed that in 129SvEv mice neither medio-lateral, nor anterior-posterior diameters changed over the same period (Fig 2 B & C). However, in 129SvEv Sparc^tm1cam mice, medio-lateral diameter increased by 10% over this age range (p<0.01; Fig 2B), whilst anterior-posterior diameter remained unaltered. At no point were either measures of diameter different between 129SvEv Sparc^tm1cam mice and their wild-type counterparts.

Figure 2.

Development of femoral geometry and mineral content in control (■) and 129SvEv Sparc^tm1cam null mice (●). Parameters shown are A) femoral length, B) mid-diaphyseal M-L diameter, C) mid-diaphyseal A-P diameter, D) aBMD, E) BMC. Values shown are mean ± SD. (n = 10 - 16 for A, B and C, and 6 - 13 for D and E; *, ** and *** p<0.05, 0.01 and 0.001 respectively vs control of the same age; +, ++ and +++ p<0.05, 0.01 and 0.001 respectively vs same animal at different age).

Femoral mineralisation

Total femoral mineral content (BMC) increased by 18% between 4 and 9 months in 129SvEv mice (p<0.05; Fig 2E), but when corrected for femoral area, areal BM density (aBMD) was not significantly increased (Fig 2D). Total femoral BMC in 129SvEv Sparc^tm1cam mice was reduced by 20% in 4 month old 129SvEv Sparc^tm1cam mice (p<0.01), but despite increasing significantly by 9 months (p<0.01), total BMC remained 14% lower than that in 129SvEv mice (p<0.05; Fig 2E). When corrected for femoral area, aBMD remained significantly reduced in 129SvEv Sparc^tm1cam mice (Fig 2D). At 4 months aBMD was 17% lower in 129SvEv Sparc^tm1cam mice (p<0.01), but despite a 13% increase (P<0.001), remained 11% lower than that in 129SvEv mice at 9 months (p<0.001; Fig 2D).

Femoral strength

Strength testing of mid-diaphyseal femori revealed that failure load did not increase significantly between 4 and 9 months in 129SvEv mice (Fig 3A). Despite a similar femoral strength at 4 months, failure load in 129SvEv Sparc^tm1cam mice diverged from that in their wild-type counterparts, becoming 20% lower at 9 months of age (P<0.01; Fig 3A). The absence of any significant age- or strain-related differences in the second moment of area revealed that there was no significant contribution of geometric variables to the recorded differences in femoral strength (Fig 3B). Although in 129SvEv Sparc^tm1cam mice the mean ultimate tensile stress at 9 months was only 78% of that in 4 month-old 129SvEv Sparc^tm1cam mice, none of the means were significantly different (Fig 3B).

Figure 3.

Development of femoral strength in control (■) and 129SvEv Sparc^tm1cam null mice (●). Parameters shown are A) failure load, B) second moment of area C) ultimate tensile stress. Values shown are mean ± SD. (n = 13 - 16; ** p<0.01 vs control of the same age).

Array results

As stated above, four different sets of array comparisons were carried out:-

1. 17 week 129SvEv Sparc^tm1cam null, male vs. 17 week 129Sv/Ev wild type, male, bone marrow

2. 17 week 129SvEv Sparc^tm1cam null, male vs. 17 week 129Sv/Ev wild type, male, femoral midshaft

3. 40 week 129SvEv Sparc^tm1cam null, male vs. 40 week 129Sv/Ev wild type, male, bone marrow

4. 40 week 129SvEv Sparc^tm1cam null, male vs. 40 week 129Sv/Ev wild type, male, femoral midshaft

Bone marrow

Analysis of bone marrow samples at both 4 and 9 months showed no statistically significant differences between 129SvEv Sparc^tm1cam null mice and 129SvEv controls. When we looked at Sparc expression in controls, it was often so low that values were beneath background +2SD, thereby eliminating Sparc from further analysis. If Sparc is not highly expressed in a majority of marrow cells, the effect of knocking it out in this tissue may be minimal. Another possibility is that, owing to high heterogeneity of cell type within the bone marrow, significant expression differences between cell subtypes may be difficult to detect.

Femoral midshaft

Analysis of femoral midshaft bone arrays showed that 5 genes (including Sparc) were differentially regulated at 4 months between 129SvEv Sparc^tm1cam null mice and 129SvEv controls, while 15 genes (also including Sparc) were differentially regulated at the 9 month timepoint (see Table 1). The presence of the Sparc gene acted as a good internal control and was reliably detected in each of the femur datasets as downregulated.

Via semi-quantitative RT-PCR, we were able to confirm the downregulation of Sparc at 4 and 9 months and of 5 other genes, zinc finger protein 162 (Zfp162), Bystin-like 1 (Bysl), transcription factor E2F4, and two uncharacterised ESTs, BG066811 and BG065585, at 9 months (Figure 4). Bioinformatic analysis of these genes is presented in Table 2. We also analysed the femur dataset from the control animals in order to identify genes highly expressed in normal bone. Details of 429 highly expressed ESTs are presented in Supplementary Table 1. Some of these represent the same gene, having taken redundancies into account, there are a total of 403 entries. 204 genes are mapped to a UniGene cluster; some of these are well characterised functionally. 16 map to the mouse mitochondrion. 183 are completely uncharacterised as of 23/02/2007. We have compared the composition of our dataset with other array studies previously presented. Microarray datasets involving biglycan deficient pre-osteoblasts, glucocorticoid treated osteoblasts and early osteoblast differentiation did not contain any of the genes we noted above (12,14,26). However, all of these studies involved cultured osteoblasts as opposed to live bone samples. Similarly, we have compared the chromosomal locations of our genes with a quantitative trait loci (QTL) mapping study using 129S1/SvImJ mice (the only mapping study we could find where strain 129 mice were used to map bone density) (24). None of the locations of our genes matched to QTLs identified; however, we have included mouse and human chromosomal locations in Table 2 for ease of comparison with other such studies.

Table 2.

Bioinformatic analysis of genes downregulated in Sparc null bone. Data from human homologs are included

Acc no.	Array	Identity	Regulation + fold change	Unigene	Mouse chromosome	human homolog, Unigene + chromosome	Comments	Hs. position + Reference interval
AW538347	9m femur	E2f4: E2F transcription factor 4, mRNA (cDNA clone MGC:38438 IMAGE:5346688)	down, -3.82	Mm.34554	8 (8 D3)	E2F4, Hs.108371 , 16q21-q22	Role in early adipogenic differentiation . Knockout mice show craniofacial and erythropoietic defects. OMIM 600659. Human Unigene cluster lists bone and bone marrow as tissue source
BG064802	4m + 9m femur	Sparc Secreted acidic cysteine rich glycoprotein (Osteonectin)	4mth, down, -9.04 9mth down, -11.15	Mm.35439	11	SPARC, Hs.111779, 5q31.3-q32	knocked out in Sparc mice, working control. Mouse Unigene cluster lists bone and bone marrow as tissue source	584.07 D5S470-D5S487 (152.8-157.6 cM)
BG065585	9m femur	EST	down, -4.83	no Unigene			uncharacterised
BG066811	9m femur	4921513E08Rik: RIKEN cDNA 4921513E08 gene (4921513E08Rik), mRNA	down, -5.23	Mm.42368	5 (5 B2)	Hypothetical protein FLJ90575, Hs.381181, 4p16.1	uncharacterised	50.48 D4S412-D4S1601 (3.7-28.2 cM)
BG077386	9m femur	Zinc finger protein 162, mRNA (cDNA clone MGC:7095 IMAGE:3157495) (Zfp162)	down, -4.99	Mm.256422	19 (19 B)	Splicing factor 1 (SF1), Hs.502829, 11q13	Highly expressed in macrophages. Contains the KH module, a sequence motif indicating a major role in regulating cellular RNA metabolism. Caslini et al. 1997, OMIM 601516. Mouse Unigene cluster lists bone and bone marrow as tissue source.	2809 D11S1357-D11S913 (62.5-70.9 cM)
BG079188	9m femur	Bystin-like, mRNA (cDNA clone MGC:27710 IMAGE:4925307) (Bysl)	down, -5.52	Mm.27291	17	Bystin-like (BYSL), Hs.106880, 6p21.1	Bystin is involved in implantation, also cell adhesion. OMIM 603871. Mouse Unigene cluster lists bone and bone marrow as tissue source.

Sp femur 9mth accession no.	Identity	Primers
AW538347	RIKEN cDNA 2010111M04 gene	AW538347F: CAGGAGCTGGAGCCATTCTT AW538347R: TCAGAGAGAGGGTGGTGCTG
BG065404	Esdn-pending Endothelial and smooth muscle cell-derived neuropilin-like molecule	EsdnF: GAGTGCAAGGAGAGCAGTGG EsdnR: CGCTTAGTGGGCAGAGGGTA
BG065585	EST	BG065585F: AACAGAGCTGAGTCTGTTGTG BG065585R: TGGAGAGATAGATGGCTCAG
BG066371	Weakly similar to RIKEN cDNA 5730493B19	BG066371F: CCGGCTCAACATCCTTTCTT BG066371R: GGAACATCCCCAAACTCACA
BG066667	ESTs	BG066667F: CCTTGCCTCTTCTGGAACTCA BG066667R: TCCATGCTGCTAATCCTAGCC
BG066811	RIKEN cDNA 4921513E08 gene	BG066811F: TCCTGTCAAGGGCTGTCAAG BG066811R: GTGGGCTTCTGGGTAAGGAA
BG077386	Zfp162 Zinc finger protein 162	Zfp162F: GACGCGCAGCATTACCAATA Zfp162R: CCCAGAGGTAGAGCCCACAG
BG079111	RIKEN cDNA 5730427N09 gene	BG079111F: TCTTAAATCCACGCCCACAG BG079111R: AGGTTGGAGGGGCTCTGTCT
BG079188	Bysl Bystin-like	ByslF: TGAGGGCTAGGCTTACAATCTTT ByslR: TGAACGTGATACAGCAAGGTGA
AW548270	Tpm1 tropomyosin 1, alpha	Tpm1F: ACTGGGCGAATTGCTTCTGT Tpm1R: GGACCACGCTCTCAACGATA
BG071230	Rbbp5 retinoblastoma binding protein 5	Rbbp5F: AGCTGAGAGTGGACGGAAGG Rbbp5R:TCCCTGTGTGAAGTGGCTTG
BG072628	Scmh1 sex comb on midleg homolog 1	Scmh1F: AGGGAGAGGGGAAACTCCAT Scmh1R: TTGGAAGAGGGAGACCATCC
BG075868	Hebp1 heme binding protein 1	Hebp1F: GCCTTCACAAGCCAGACCTC Hebp1R: CGGGAGGGCATCACTGTCTA
BG087183	RIKEN cDNA 2610003J05	BG087183F: CAAGTGAGGGCTTGAAAACG BG087183R: GCCCAATAATGTCCAAGAGCA
BG082955	Pabpc4;Poly A binding protein, cytoplasmic 4	Pabpc4F: GGTGGATGATGGGAATCTGAA Pabpc4R: TGCACGGCCTACAAATATGG
BG075669	2700023B17Rik;RIKEN cDNA 2700023B17 gene	BG075669F: TTTATACAGCTATACAAGTT BG075669R: GGTAAAGAATTTTGTCAGTC
BG072511	EST	BG072511F: CGAGACAGGGTTTCTCTGTGTAGC BG072511R: TGGGCGGTGGTGGCACACAC
BG069910	Canx;Calnexin	CanxF: CAGGGTCCTTTACATTACAACTGC CanxR: TTTAAGAACTTGATCTGTGATTTCCTC

Primers	Primer type	No bases	Price
AW538347F: CAGGAGCTGGAGCCATTCTT AW538347R: TCAGAGAGAGGGTGGTGCTG	3OD select	20 20	5.80 5.80
EsdnF: GAGTGCAAGGAGAGCAGTGG EsdnR: CGCTTAGTGGGCAGAGGGTA	3OD select	20  20	5.80 5.80
BG065585F: AACAGAGCTGAGTCTGTTGTG BG065585R: TGGAGAGATAGATGGCTCAG	3OD select	21  20	6.09 5.80
BG066371F: CCGGCTCAACATCCTTTCTT BG066371R: GGAACATCCCCAAACTCACA	3OD select	20  20	5.80 5.80
BG066667F: CCTTGCCTCTTCTGGAACTCA BG066667R: TCCATGCTGCTAATCCTAGCC	3OD select	21  21	6.09 6.09
BG066811F: TCCTGTCAAGGGCTGTCAAG BG066811R: GTGGGCTTCTGGGTAAGGAA	3OD select	20  20	5.80 5.80
Zfp162F: GACGCGCAGCATTACCAATA Zfp162R: CCCAGAGGTAGAGCCCACAG	3OD select	20  20	5.80 5.80
BG079111F: TCTTAAATCCACGCCCACAG BG079111R: AGGTTGGAGGGGCTCTGTCT	3OD select	20  20	5.80 5.80
ByslF: TGAGGGCTAGGCTTACAATCTTT ByslR: TGAACGTGATACAGCAAGGTGA	3OD select	22 22	6.38 6.38
Tpm1F: ACTGGGCGAATTGCTTCTGT Tpm1R: GGACCACGCTCTCAACGATA	3OD select	20  20	5.80 5.80
Rbbp5F: AGCTGAGAGTGGACGGAAGG Rbbp5R:TCCCTGTGTGAAGTGGCTTG	3OD select	20  20	5.80 5.80
Scmh1F: AGGGAGAGGGGAAACTCCAT Scmh1R: TTGGAAGAGGGAGACCATCC	3OD select	20 20	5.80 5.80
Hebp1F: GCCTTCACAAGCCAGACCTC Hebp1R: CGGGAGGGCATCACTGTCTA	3OD select	20  20	5.80 5.80
BG087183F: CAAGTGAGGGCTTGAAAACG BG087183R: GCCCAATAATGTCCAAGAGCA	3OD select	20  21	5.80 6.09
Pabpc4F: GGTGGATGATGGGAATCTGAA Pabpc4R: TGCACGGCCTACAAATATGG	3OD select	21 20	6.09 5.80
BG075669F: TTTATACAGCTATACAAGTT BG075669R: GGTAAAGAATTTTGTCAGTC	3OD select	20 20	5.80 5.80
BG072511F: CGAGACAGGGTTTCTCTGTGTAGC BG072511R: TGGGCGGTGGTGGCACACAC	3OD select	24  20	6.96 5.80
CanxF: CAGGGTCCTTTACATTACAACTGC CanxR: TTTAAGAACTTGATCTGTGATTTCCTC	3OD select	24 27	6.96 7.83
	total +	Total shipping	215.76 218.76

Acc.no. = GenBank accession number of the NIA EST, Array = array experiment where the differential regulation was detected, Identity = most homologous sequence match to the NIA EST (if any)

DISCUSSION

Sparc is an important structural constituent of bone; it binds hydroxyapatite and collagen, in addition to other functions including the regulation of cell proliferation and cell-matrix interactions, stimulation of angiogenesis and MMP production (15). We have now confirmed that the inactivation of Sparc results in osteopenia, regardless of mutation location or genetic background. This is important, as numerous phenotypes attributed solely to the lack of the ablated gene, have later been explained by genetic background differences between experimental and control animals or by linkage disequilibrium (13).

Our Sparc null mutation is present in a genetic background that normally specifies high bone density (3,4); as a result we have a milder phenotype than that previously reported (15) Differences in BMC and aBMD were noted at both 4 and 9 month time points, but mechanical strength only differed significantly at 9 months; this appears to be accounted for by a change in material properties rather than a change in the size of the bones. We also noted a small but significant reduction (3-4%) in femoral length in 129SvEv Sparc^tm1cam null mice. Length is inversely related to strength, but the potential of this variable is largely negated in our study via the use of a fixed span in the 3-point bending protocol. Previous reports had indicated no change in skeletal size (2,19). However, Sparc^tm1cam null mice were studied initially on a Mf1^bb/129Sv/Ev outbred mixed background. Preliminary observations using these mice and Mf1bb controls showed such high degrees of individual variability in most skeletal parameters (unpublished data), that we subsequently used inbred 129SvEv mice. Given the small difference observed, it is also likely that this would be undetectable even on a “fixed” mixed background derived from two inbred lines (such as that studied in the exon 4 knockout).

Previous reports note increased adiposity in exon 4 Sparc null mice (7), conversely, Sparc expression is increased in 3 different mouse models of obesity and altered in human obesity cases (36). These results point to a complex role for Sparc in adipose regulation. We therefore assessed body weight and adiposity in these animals. Adipocytes, osteoblasts and cartilage originate from the same lineage, mesenchymal stem cells (6,7,20); absence of Sparc may perturb mesenchymal differentiation at the expense of the osteoblast lineage. Skeletal size differences are minor and only appear in later life, so an effect on cartilage differentiation and skeletal patterning is unlikely. However, marrow stroma from 129SvC57BL/6 Sparc null mice contains fewer osteoblasts and shows an increased tendency to form adipocytes (16), suggesting preferential adipocytic differentiation at the expense of the osteoblast lineage. In order to address these issues, we have assessed the size and number of bone marrow adipocytes. These studies, involving staining and counting of adipocytes in fixed tibial sections, are advantageous in that this method rules out any effect of cell culture on cell numbers and may be more indicative of any perturbations in stem cell differentiation than analysis of peripheral fat depots. The numbers of marrow adipocytes do not vary significantly (although their volume does at one time point), indicating that the lack of Sparc has more effect on lipid accumulation than on differentiation or proliferation. However, adipocytes in marrow secrete small amounts of leptin. Larger adipocytes may secrete more, which may have a paracrine influence on bone formation without changing levels of circulating leptin.

Exon 4 129SvC57BL/6 Sparc null mice have been shown to have larger deposits of subcutaneous and epididymal fat and to display raised serum leptin levels between 6-8 months (7). In contrast, we failed to note a difference in circulating leptin concentration at 9 months, but noted a significant decrease in leptin concentrations at 17 weeks. We also noticed a substantial increase in body weight at 17 weeks (a time point not tested in the prior study (7), which had disappeared at 40 weeks (9 months). These differences are probably not diet related, as food fat contents were similar (4.3% fat, our study vs. 4% fat, previous study (7). Mirroring the raised 17 week body weight, we noted significant increases in the size of marrow adipocytes and the percentage of the marrow cavity occupied by adipocytes at the same timepoint. No increase in marrow adipocyte number was noted, differing from subcutaneous and epididymal fat in 129Sv/C57Bl/6 Sparc null mice, in which both adipocyte number and size were increased. Although peripheral and marrow fat deposits may be regulated in a differential manner, our results would concur with previous reports that absence of Sparc appears to cause abnormalities of adipose metabolism. Genetic aspects of adipose biology differ between inbred strains of mice and may lead to variations in adipose response to the lack of Sparc.

Since we have confirmed that 129SvEv Sparc^tm1cam null mice show an impaired skeletal phenotype, these mice are therefore an ideal model with which to carry out array studies in bone; experimental and control lines used in this study should be genetically identical apart from the Sparc null mutation, removing a common source of noise. Sparc was downregulated at 4 and 9 months; the presence of Sparc within the NIA set acts as an internal control and confirms array validity. Aside from Sparc, we have identified 5 downregulated genes in 9 month 129SvEv Sparc^tm1cam null mice. This mirrors previous findings in the lens, where we were unable to detect differential regulation of anything other than Sparc at 4 months, whereas a large number of genes were differentially regulated at 9 months (28). As we have found in bone, the majority of differentially regulated genes in early stage cataract were also downregulated (a result also noted by others investigating cataract via array) (28,33) .

These results imply that the absence of Sparc does not immediately result in the dysregulation of other genes. Repetition of this result in bone was somewhat surprising though, as, in the genetic backgrounds we used, Sparc^tm1cam lenses are asymptomatic at four months; overt cataract has yet to develop. In contrast, 129SvEv Sparc^tm1cam bone is already compromised at this stage. Possibly the initial cause of bone weakness (and cataract) in Sparc null mice is simply its absence, this may then be followed by progressive alterations in gene expression as the phenotype develops over time. Other than Sparc itself, we have not noted any genes that are differentially regulated in both lens and bone datasets. This implies that Sparc is not having a major effect on signalling pathways common to both lens and bone (tissues that remodel constantly throughout life). There are a number of genes that have a high homology to Sparc (hevin/Sc1, Spock 1-3 (also known as testicans 1-3), Smoc1 and Smoc2). Our arrays have not detected differential regulation of any of these in order to compensate for the absence of Sparc. It should be noted however, that the NIA set contains only 15,000 ESTs, in which there is a degree of redundancy (28). Therefore, this hypothesis has not been tested on an array dataset that is representative of the entire mouse genome.

The nature of the differentially regulated genes is interesting. The Unigene clusters representing 3 of these genes, E2F4, Bysl and Zfp162, quote bone and sometimes bone marrow as a tissue source, demonstrating their relevance to bone biology. E2F4 knockout mice have been generated; erythroid abnormalities during development and craniofacial abnormalities have been noted (22). Of further relevance is the fact that this gene regulates adipocyte differentiation. Zfp 162 and Bysl are less well investigated, however, high levels of expression in macrophages (which are bone marrow derived and which fuse in some cases to form osteoclasts) have been noted for Zfp 162 (10). The remaining two genes are functionally uncharacterized. Notably, despite the number of ESTs present on the array, we have not seen relative changes in biomarkers for bone metabolism, in either bone or in bone marrow, which contains both osteoblast and osteoblast precursors.

Using data from control animals, we have also identified 403 genes that are highly expressed in bone. Notably, 183 of these are completely uncharacterised. These data may be a useful resource in identifying novel candidate genes for skeletal disease, especially as a large number of linkage and association studies have delineated wide intervals in which important genes reside, but, owing to the number of genes within the critical regions, many have not yet been identified (21).

Despite the wealth of data from various genome and EST projects (23,25), biochemical pathways identified in bone are still incomplete, while large proportions of genes or ESTs identified by large scale sequencing projects are relatively functionally uncharacterized, implying that many vital regulators of skeletal homeostasis remain to be elucidated. The use of microarrays in combination with knockout mice explores phenotype in a complex world of development, ageing and environment, and can also be used to elucidate currently unidentified components of essential biological processes such as skeletal metabolism. Identification of genes dysregulated during abnormal bone thinning may lead us to novel routes of therapeutic intervention for osteoporosis. This topic is also important in that it will allow us to develop our knowledge of the biology and biochemical mode of action of Sparc, which we have confirmed as an potentially important therapeutic target for osteoporosis.

Supplementary Material

Supplementary Table 1

Supplementary Table 1: 429 ESTs were detected as highly expressed in bone (above). Some of these represent the same gene, having taken redundancies into account, 204 genes are mapped to a UniGene cluster. 16 are mitochondrial, while a further 183 are still completely uncharacterised. Of these, 144 are mapped to the mouse genome. Given the fact that the databases are continually updated, the final date of search for each uncharacterised gene is given in the entry.