Bone 48 (2011) 496–506

Contents lists available at ScienceDirect

Bone

j ourna l homepage: www.e lsev ie r.com/ locate /bone

High dietary intake of retinol leads to bone marrow hypoxia and diaphysealendosteal mineralization in rats☆

Thomas Lind a,⁎, P. Monica Lind b, Annica Jacobson a, Lijuan Hu a, Anders Sundqvist a,1, Juha Risteli c,Africa Yebra-Rodriguez d, Alejandro Rodriguez-Navarro e, Göran Andersson f, Håkan Melhus a

a Department of Medical Sciences, Section of Clinical Pharmacology, University Hospital, S-75185, Uppsala, Swedenb Department of Medical Sciences, Section of Occupational and Environmental Medicine, University Hospital, S-75185, Uppsala, Swedenc Department of Clinical Chemistry, Oulu University Hospital, FI-90029, Oulu, Finlandd Department of Geology, IACT Associated Unit (CSIC-UGR), University of Jaen, Campus Las Lagunillas s/n, 23071 Jaen, Spaine Departamento de Mineralogía y Petrología, Facultad de Ciencias, Universidad de Granada, Avenida Fuentenueva s/n, 18071 Granada, Spainf Division of Pathology, Department of Laboratory Medicine, Karolinska Institutet, Karolinska University Hospital, S-14152, Huddinge, Sweden

☆ Funding: The Swedish Society of Medicine and tCouncil.⁎ Corresponding author. Department of Medical

Pharmacology, University Hospital, Ing 70 3tr Foa2 LabFax: +46 18 553611.

E-mail addresses: thomas.lind@medsci.uu.se (T. Lind(P.M. Lind), annica.jacobson@medsci.uu.se (A. Jacobson(L. Hu), anders.sundqvist@medsci.uu.se (A. Sundqvist), jayebra@ujaen.es (A. Yebra-Rodriguez), anava@ugr.es (Agoran.andersson@ki.se (G. Andersson), hakan.melhus@m

1 Current address: Ludwig Institute for Cancer ResearchCentre, Uppsala, Sweden.

8756-3282/$ – see front matter © 2010 Elsevier Inc. Aldoi:10.1016/j.bone.2010.10.169

a b s t r a c t

a r t i c l e i n f o

Article history:Received 26 March 2010Revised 27 September 2010Accepted 14 October 2010Available online 28 October 2010

Edited by: Bjorn Olsen

Keywords:NutritionOsteoporosisRetinolRodentBone mineralization

Vitamin A (retinol) is the only molecule known to induce spontaneous fractures in laboratory animals and wehave identified retinol as a risk factor for fracture in humans. Since subsequent observational studies inhumans and old animal data both show that high retinol intake appears to only have small effects on bonemineral density (BMD) we undertook a mechanistic study of how excess retinol reduces bone diameter whileleaving BMD essentially unaffected. We fed growing rats high doses of retinol for only 1 week. Bone analysisinvolved antibody-based methods, histology, pQCT, biomechanics and bone compartment-specific PCRtogether with Fourier Transform Infrared Spectroscopy of bone mineral. Excess dietary retinol inducedweakening of bones with little apparent effect on BMD. Periosteal osteoclasts increased but unexpectedlyendosteal osteoclasts disappeared and there was a reduction of osteoclastic serum markers. There was also alack of capillary erythrocytes, endothelial cells and serum retinol transport protein in the endosteal/marrowcompartment. A further indication of reduced endosteal/marrow blood flow was the increased expression ofhypoxia-associated genes. Also, in contrast to the inhibitory effects in vitro, the marrow of retinol-treated ratsshowed increased expression of osteogenic genes. Finally, we show that hypervitaminotic bones have a higherdegree of mineralization, which is in line with biomechanical data of preserved stiffness in spite of thinnerbones. Together these novel findings suggest that a rapid primary effect of excess retinol on bone tissue is theimpairment of endosteal/marrow blood flow leading to hypoxia and pathological endosteal mineralization.

he Swedish Medical Research

Sciences, Section of Clinical22, S-75185, Uppsala, Sweden.

), monica.lind@medsci.uu.se), lijuan.hu@medsci.uu.seuha.risteli@ppshp.fi (J. Risteli),. Rodriguez-Navarro),edsci.uu.se (H. Melhus).

, Uppsala University, Biomedical

l rights reserved.

© 2010 Elsevier Inc. All rights reserved.

Introduction

Vitamin A is the only molecule known to induce spontaneousfractures in laboratory animals and can only be derived from the diet asno animal species has the capability for de novo synthesis [1,2]. Theimportant regulatory role of vitamin A was recently highlighted whenthe nuclear receptors of its active metabolite, retinoic acid (RA), were

shown to have 5310 target genes, in breast cancer cells [3]. Vitamin Adeficiency is a public health problem in over 120 developing countriesand significantly increases the risk of severe illness and death fromcommon childhood infections [4]. In the industrializedpart of theworld,we have identified high serum retinol levels as a risk factor for fracture[5,6]. However subsequent observational studies relying on BMDmeasurements were unable to associate increased vitamin A intakewith reduced BMD. For example, of the nine published observationalstudies, investigating the association between circulating levels ofvitamin A and adverse skeletal effects, three relied on fracture as anendpoint and in twoof theseapositive associationwith fracture riskwasdemonstrated. More importantly, none of the studies relying on BMD asendpoint where able to demonstrate that serum retinol affected BMDnegatively, on the contrary, three of the studies noticed a positivecorrelation [7]. This may indicate that excess retinol induces fragilebones without a large effect on BMD. In line with this, severalhypervitaminosis A studies on laboratory animals were unable todemonstrate significant differences in ash weight between the thinnerhypervitaminotic bones and control bones [8–10]. Hypervitaminosis A

497T. Lind et al. / Bone 48 (2011) 496–506

has been studied since 1925 and consistently show induced loss ofappetite together with reduced weight gain, bone thinning, spontane-ous fractures of long bones together with general (including bone)tissue hemorrhage and increased periosteal resorption [11]. Recently,Kneissel and colleagues showed that excess retinoid treatment inducedbone thinning in both young and aged rodents and also suggested abone compartment-specific action of retinoids by means of measuringdecreased cortical BMD but increased trabecular BMD in animalsreceiving RA [12]. Furthermore, our recent finding in an experimentalstudy of subclinical hypervitaminosis A showed that although bendingstrength decreased, there was no reduction of BMD (cortical ortrabecular) [13] and as the ash weight was unchanged and since thehypervitaminotic bone volume was smaller than in control bones, theactual volumetric BMDwas increased [14]. As osteoporosis is defined asa reduction of BMD of 2.5 standard deviations or more, compared tohealthy young bone, it is important to further investigate themechanism behind this phenomenon.

Therefore the aim of this study was to identify the mechanismbehind how excess retinol induces fragile bones with minimal effectson BMD. We have used the rat as a model, since this species has beenconsistently used in hypervitaminosis A investigations. Importantly,as bone thinning and lack of reduced ash weight were noticed instudies using both high and subclinical doses of vitamin A, we have inthe present study mimicked earlier hypervitaminosis A protocolsusing high doses of retinol to induce distinct changes in a short time.Also important, it has previously been shown that high retinol doses,comparable to what is used in this report, given to rats for 2 weeksresult in plasma concentrations of about 3–4 μM, which is similar towhat is seen in hypervitaminosis A toxicity in humans [56,57]. Wehave applied modern methods, not used before in hypervitaminosis Astudies, such as specific antibody-based immunohistochemistry andELISA analysis, differential bone compartment-specific real-time PCRto assess changes in gene expression together with Fourier TransformInfrared (FTIR) Spectroscopy, to analyze the bonemineral content andcrystallinity [15].

Materials and methods

Animals and experimental design

Twenty-four male Sprague–Dawley rats, 5 weeks of age, wereobtained from Möllegaards Breeding Centre, Ltd. (Skensved,Denmark). They were acclimatized for 1 week and kept in groups ofthree animals and had free access to water and commercial pellet diet(Lactamin R36, Stockholm, Sweden). The rats were divided into twogroups, each with 12 animals. They were fed a standard dietcontaining 12 IU vitamin A/g pellet (“cont”), or a standard dietsupplemented with 1700 IU (“Ex. vitA”) vitamin A/g pellet. Thevitamin Awas added to the pellets in the form of retinyl palmitate andretinyl acetate. Food intake was continuously measured by weighing(grams of pellet per cage of three rats). Body weight was measured atthe beginning of the study (initial weight), at day three and before therats were killed (final weight). After 8 days, the rats were killed byexsanguinations from the abdominal aorta under Eqvitesin anesthesia(chloral hydrate 182 mg/kg, pentobarbital 41.7 g/kg). Blood was leftat room temperature for at least 30 min before centrifuging at 200×gfor 10 min to separate sera, aliquots were stored at −70 °C pendingbiochemical analyses. Animal experiments were conducted inaccordance with Swedish regulations.

Peripheral quantitative computed tomography (pQCT)

The left femora were scanned using pQCT (Stratec XCT ResearchSA+) with version 5.50 R software (Norland-Stratec MedizintechnikGmbH, Birkenfeld, Germany) using a voxel size of 0.07 mm. Theprecision and accuracy of the method have been verified previously

[13,16]. The femur diaphysis was scanned at the midshaft and thethreshold value was set to 710 mg cm−3, values above were definedas cortical bone. Trabecular variables were determined by metaphy-seal scans at a point located at 14% of the total bone length from theproximal tip. Values ranging from 280 to 400 mg cm−3 wereconsidered to be trabecular bone. The thresholds were set accordingto the manufacturer's recommendations.

Determination of serum markers

Commercially available kits were utilized for measurement ofserum markers as follows: osteocalcin, Rat-MID Osteocalcin (NordicBioscience Diagnostics, Herlev, Denmark); TRACP 5b, RatTRAP(Immunodiagnostic Systems, Boldon, UK); N-terminal propeptide oftype I collagen, PINP EIA (Immunodiagnostic Systems, Boldon, UK); C-terminal telopeptides of type I collagen (CTX-1), RatLaps (NordicBioscience Diagnostics, Herlev, Denmark); free soluble RANKL(Biomedica Medizinprodukte, Vienna, Austria), soluble Icam1 (R&DSystems) according to the manufacturer's instructions and rat ICTP aspreviously published [17]. Data represents 11 control and 12 excessretinol rats and are presented as mean±SD.

Immunohistochemistry

The bone (humeri) preparation and immunohistochemistry havepreviously been described in detail [18]. The bones from all animalswere sectioned in the same orientation in order to make comparablesections. Immunostaining for cathepsin K was achieved by the use ofpolyclonal antimouse cathepsin K antiserum at a dilution of 1:300 [19]and for preparation of a polyclonal antibody against osteopontin asynthetic peptide corresponding to amino acid 46–58(PDPSQKQNLLAPQ) to the rat osteopontin protein sequence wasconjugated to keyhole limpet hemocyanin (Innovagen AB, Lund,Sweden). A rabbit was immunized subcutaneously with 200 μg of thepeptide conjugate mixed with Freund's complete adjuvant at week 0and followed by booster injections of 100 μg peptide conjugate mixedwith Freund's incomplete adjuvant at week 2, 5 and 11. Serum wascollected 2 weeks after the last booster injection. The specificity of theantibody was verified by Western blot analysis and used inimmunohistochemistry at a dilution of 1:800. For protein localization,sections were incubated with goat antirat Icam1/CD54 (1:200, R&DSystems), rabbit antimouse Pecam1/CD31 (1:400, SantaCruz-1506),rabbit antirat Dmp1 (1:400, Takara Bio), rabbit antihuman vWF(1:800, DAKO), goat antimouse Mmp2 (1:400, R&D Systems), goatantihuman osteocalcin (1:800, SantaCruz-18319), and mouse anti-human Rbp4 (L48, 1:800) [20]. Visualization of the antibodies whereachieved by incubation with secondary biotinylated antibody at adilution of 1:200 in 10% serum and PBS followed by an avidin–biotin–peroxidase complex incubation using the Vectastain ABC-kit (VectorLaboratories) and the substrate diaminobenzidine tetrahydrochloride(DAB, DAKO).

Cell culture

Mouse stromal bone marrow derived cell line, ST2, was culturedessentially as described previously [21]. Briefly, cells were maintainedin α-minimal essential medium containing 10% fetal bovine serumand 100 units/ml penicillin/streptomycin. To induce osteoblastogen-esis, confluent cells were switched to media containing 25 μg/mlascorbic acid and 10 mM β-glycerophosphate, treated with ±400 nMretinoic acid (Sigma-Aldrich). RA was dissolved in 95% ethanol in adark room under the flow of nitrogen. The 0.5 mg/ml (1.66 mM) stocksolution was stored in −70 °C and shielded from light until use.Experiments were stopped and total RNA was extracted using TRIReagent® (Sigma-Aldrich).

Table 1Phenotypic characteristics from 1 week of excess ingestion of vitamin A.

Control Excess vitamin A p-value(n=12) (n=12) t-test

End weight (g) 124±2.1 105±1.6 b0.001Weight increase (g) 50.7±1.06 32.4±0.60 b0.001Femur length (mm) 25.5±0.58 24.8±0.72 b0.05Food intake (g) 102±4.16 89.1±28.1 b0.05

Values are mean±SE. n=number of individuals.

498 T. Lind et al. / Bone 48 (2011) 496–506

Diaphyseal bone tissue RNA isolation

Humeri were isolated, and all connective tissue, includingperiosteum, were completely removed as were also both epiphyses(including growth plates). For samples with cortical bone includingmarrow (CortB+M) the samples (n=3 for cont and n=4 for Ex. vitA)were snap-frozen in liquid nitrogen and quickly crushed into a finepowder using a mortar and pestle followed by total RNA extractionwith TRI Reagent® (Sigma-Aldrich). For cortical bone withoutmarrow (CortB) (n=4 for cont and n=3 for Ex. vitA), diaphysealbones were cut in pieces, vortexed in ice-cold PBS three times at 10 sto remove marrow cells followed by snap-freezing in liquid nitrogen,crushing into a fine powder and RNA extraction. Isolated RNA wasquantitated using spectrophotometry by measuring the absorbance at260 nm, and the 260/280 nm ratio was calculated, and RNA was keptonly when this ratio was 1.9–2.0 to ensure the absence of proteincontamination and limited RNA degradation. The integrity of sampleRNAs was confirmed by capillary electrophoresis separating 18S and28S ribosomal RNA on an Agilent Technologies 2100 Bioanalyzer(Agilent Technologies, Palo Alto, CA).

Quantitative RT-PCR

Four hundred nanograms of total RNA were transcribed to cDNAusing the TaqMan system (Applied Biosystems). Quantitative PCRwasperformed using inventoried TaqMan® Gene Expression Assays forCyp26a1 (Rn00590308), Stra6 (Rn01418200), Hif1α (Rn00577560),Twist1 (Rn00585470), Mmp2 (Rn01538174), tissue non-specificalkaline phosphatase (ALP or Alpl) (Rn00564931), osteocalcin(Rn00566386), cathepsin K (Rn00580723), Runx2 (Rn01512298),OPG (Rn00563499) and RANKL (Rn00589289) according to themanufacturer's protocol, on a TaqMan 7000 apparatus. Cyclingprotocol: 50 °C for 2 min followed by 95 °C for 10 min and then 40cycles of 95 °C 15 s followed by 60 °C for 1 min. Expression levels weredivided by β-actin (Rn00667869) levels for standardization. Eachsample was run as single points at each experiment and experimentswere performed at least twice.

Mechanical three-point bending analysis of femur diaphyses

Three-point bending test was performed on right femora using anelectromechanical material testing machine (Avalon technology Inc.,Rochester, MN, USA)with a span length of 15 mm and a loading speedof 0.48 mm/s, as described previously [13]. Bone strength wasmeasured at 50% of the total length of the bone. The load anddisplacement were sampled with a frequency of 50 Hz and storeddigitally. Displacement at failure (the total amount of movement ofthe load in mm from the moment of contact with the bone surfaceuntil the bone fractured) and the load applied (Newtons; N) at themoment when the bone fractured were recorded. The data accumu-lated during each test were used to construct a load–displacementcurve. The area under the curve defines the amount of energyabsorbed until failure (breakage) measured in Nmm. The maximumslope of the load–displacement curve is defined as the bone's stiffness(N/mm).

Bone chemical composition

Bone chemical composition was analyzed by Fourier transforminfrared spectrometry (FTIR). The entire left femora, after the pQCTanalysis, were cleaned with a scalpel, rinsed with Milli-Q water andpowdered using a cryogenic mill (CertiPrep 6750 Freezer/Mill, SPEX).For the FTIR analyses, 5 mg of bone powder was mixed with 90 mg ofFTIR-grade KBr and pressed under vacuum. Infrared spectral datawere collected on a Fourier transform infrared spectrometer (MagnaIR200, Nicolet) from 4000 to 400 cm−1 at 2 cm−1 resolution over 128

scans. The amounts of phosphate, carbonate, and organic matrix inbone were determined from the peak area of absorption bandsassociated with phosphate, carbonate, and amide groups in theinfrared spectra. Overlapping peaks under the above mentionedbands were resolved and their integrated areas measured using acurvefitting software (Peakfit v.4.11) and applying a secondderivativemethodology. Peak areas of isolated bandswere normalized to the areaof 3800–2800 cm−1 band region associated with OH groups and wererepresented by a capital “A” (e.g. A1660). On the other hand, theindividual sub-peak area under a band were normalized to theassociated band and were designated using “a” (e.g. a956, a994,a1020). The procedure is described in detail elsewhere [22]. Thismethodology allows for a detailed and quantitative analysis ofdifferent molecular constituents of bone mineral. In particular,different compositional variables were determined to quantify thepotential effects of hypervitaminosis on bone mineralization. Todescribe bone degree of mineralization and carbonate content in themineral from FTIR analyses: degree of mineralization of bone(mineral)was defined as the ratio between the peak area of phosphateand amide I bands: (mineral=(A1200_900)/A1660). Carbonate inbonemineral (minCO3)was defined as the ratio between the peak areafor 1405 cm−1 (carbonate type B substitution) to phosphate bandarea: (minCO3=A1405/(A900_1200)).

Statistical analyses

STATVIEW 5.0 software (SAS Institute, Inc., Cary, NC) was used forall statistical analyses. The datawere evaluated by t-test. In every case,pb0.05 was considered statistically significant.

Results

One week of excess retinol induces brittle bones

Growing (6-week old) male rats were fed either standard chow ora diet fortified with a high dose of vitamin A for 7 days. All animalssurvived and appeared healthy during the experimental period of8 days (which includes fasting for the last 24 h) although the classicalgross phenotypical characteristics of excess retinol consumption suchas reduced food intake (−13%) andweight gain (−36%)were noticed(Table 1). The length of femur was slightly affected by the large excessof retinol (−2.4%) (Table 1). pQCT measurements showed thatmetaphyseal trabecular and total bone mineral contents weredrastically reduced (34% and 32%, respectively), whereas therespective densities were less reduced (9% and 5%) (Table 2). Atmid-diaphysis, total cross-sectional areas were reduced by 15% andcortical mineral content was lowered by 15%, while the corticaldensity was only 3.2% lower than control bones. Importantly, a three-point bending test showed that bones from retinol-treated rats wereweaker as the energy absorption and load at failure was reduced by40% and 19%, respectively (Table 2).

Table 2Results of pQCT analysis and three-point bending test (Biomech.) of rat femurs.

Control Excess vitamin A p-value(n=12) (n=12) t-test

pQCT Distal metaphysisTotal BMC (mg/mm) 5.74±0.18 3.90±0.01 b0.001Total BMD (mg/cm³) 349±3.6 331±4.5 b0.01Trab BMC (mg/mm) 1.90±0.067 1.25±0.043 b0.001Trab BMD (mg/cm³) 192±3.88 174±3.77 b0.01Total CSA (mm²) 16.4±0.44 11.8±0.28 b0.001Trab CSA (mm²) 9.87±0.20 7.18±0.22 b0.001PERIC (mm) 14.4±0.19 12.2±0.15 b0.001

DiaphysisTotal BMC (mg/mm) 3.78±0.06 3.13±0.06 b0.001Total BMD (mg/cm³) 507.4±4.75 497.4±11.79 nsTotal CSA (mm²) 7.45±0.13 6.28±0.15 b0.001Cort BMC (mg/mm) 2.76±0.06 2.21±0.06 b0.001Cort BMD (mg/cm³) 1053±5.82 1019±8.13 b0.01Cort CSA (mm²) 2.62±0.05 2.16±0.05 b0.001Cort THKC (mm) 0.30±0.004 0.27±0.07 b0.01PERIC (mm) 9.67±0.087 8.88±0.105 b0.001ENDOC (mm) 7.78±0.083 7.18±0.13 b0.001Polar SSI (mm4) 4.01±0.141 2.71±0.096 b0.001Marrow cavity (mm²) 4.83±0.10 4.12±0.15 b0.01

Biomech. Displacement (mm) 2.08±0.4 1.65±0.29 b0.01Load (N) 33.5±3.3 27.1±4.54 b0.001Energy (N×mm) 47.8±9.7 28.7±5.40 b0.01Stiffness (N/mm) 16.8±4.3 17.1±5.26 ns

Values are mean±SE. n=number of individuals; BMC=bone mineral content;BMD=bone mineral density; trab=trabecular; CSA=cross-sectional area;PERIC=periosteal circumference, cort=cortical; THKC=thickness; ENDOC=endostealcircumference.

499T. Lind et al. / Bone 48 (2011) 496–506

Excess retinol induces low bone turnover

Serum markers for osteoclast activity (cathepsin K degradation offibrillar type I collagen) show a reduction by 36% (CTX-1) andosteoclast number was reduced 31% (TRACP 5b) (Fig. 1A). Thus, aresorption index (CTX-1/TRACP 5b), taking the osteoclast number andactivity into account, reveals that the mean resorptive activity of eachosteoclast is unchanged in hypervitaminotic bones (Fig. 1A). Further-more, serum soluble RANKL, a potential osteoclast activator, wasreduced (−12%) but levels of osteoprotegerin (OPG), a potent

ATRACP 5b

0

4

8

12

U/L

cont Ex. vitA

***

CTX-1

**

0

40

80

120

ng

/ml

cont Ex. vitA

B Osteocalcin

0

500

1000

1500

ng

/ml

cont Ex. vitA

*

PINP

0

100

200

300

ng

/ml

cont Ex. vitA

*

Fig. 1. Serum bone cell markers, measured with immuno-assays. (A) Osteoclast markers andshow results from rats on excess vitamin A (Ex. vitA, n=12). Not significant (ns), *pb0.05

inhibitor of osteoclast formation, were not significantly altered,resulting in only a small reduction of the sRANKL/sOPG ratio(Fig. 1A and not shown). Two osteoblastic markers, one for activity(collagen synthesis, PINP) and one for mature osteoblasts (osteocal-cin) were also significantly reduced, 28% and 9%, respectively(Fig. 1B). Matrix metalloproteinase (Mmp) activity, measured asICTP (cross-linked carboxyterminal telopeptide of type I collagen) andgelatin zymography, were not affected by excess retinol (not shown).

Excess retinol reverses diaphyseal osteoclast positioning andendothelial staining

In diaphyseal bones from control rats, osteoclasts were scatteredalong the endocortical surface and were not readily observed on theperiosteal surface except in the highly active metaphyseal area closeto the growth plate. In contrast, osteoclasts in retinol-treated animalswere abundant at the periosteal surface of the mid-diaphysis togetherwith an almost complete lack of endosteal osteoclasts (Fig. 2A).Counting osteoclasts along the periosteal and endosteal surfaces ofthe diaphysis confirmed this reversed pattern of osteoclast position-ing (Fig. 2B). Notably, there were no apparent osteoclast differences inthe trabecular compartment (Fig. 2C). Hough et al. previouslydescribed hypervitaminosis A rats with a pathognomonic appearanceof endosteal bone, characterized by “highly vascularized bonelacunae” [23]. These corresponding areas were highly cellular(Fig. 3A) but completely negative for the osteoclastic proteins,cathepsin K (Fig. 2A), TRACP and Mmp9 (not shown). Upon closeinspection, control sections showed numerous erythrocytes denselypacked in vessel-like structures. In contrast, sections from hypervi-taminotic animals showed either no or scattered erythrocytes outsidevessel structures, indicative of hemorrhage and/or a non-functionalblood supply (Figs. 3A and B). Counting of erythrocytes around theendosteal site revealed a significant reduction of microvesselassociated erythrocytes in pathognomonic areas of hypervitaminoticanimals compared to similar areas in controls (Fig. 3C). These bonelacunae were also negative for the endothelial markers intercellularadhesion molecule 1 (Icam1), platelet endothelial cell adhesionmolecule 1 (Pecam1/CD31) and von Willebrand factor (vWF)(Fig. 4A not shown). In fact, most of the marrow in retinol-treatedrat bones, except for the large central vessel, was negative for

sRANKL

0

5

10

pm

ol/L

cont Ex. vitA

*

Resorp. index: CTX-1/TRACP 5b

0

4

8

12

16

rati

o

cont Ex. vitA

ns

, (B) osteoblast markers. White bars represent control (cont, n=11) rats and black bars, **pb0.01 and ***pb0.001.

Fig. 2. Osteoclast positioning. (A) Cathepsin K staining of diaphyseal sections revealsthat osteoclasts (arrowheads) have changed position from endosteal in controls toperiosteal in hypervitaminosis A rats. Asterix (*) indicates periosteal cell layer.Bar=100 μm. Counting of cathepsin K positive osteoclasts (B) on the endosteal andperiosteal sites shows a clear redistribution of osteoclasts in bones from rats eatingexcessive amounts of vitamin A. (C) Counting of cathepsin K positive osteoclasts in thetrabecular area reveals no change in osteoclast number. n=3 per group.

500 T. Lind et al. / Bone 48 (2011) 496–506

endothelial Icam1 and showed very few intact blood vessel structures(Fig. 4A). Conversely the periosteal blood vessels appeared both largerin diameter and closer to the bone compared to vessels found incontrol periosteum (Fig. 4B). Again, the trabecular compartment wasunaffected by retinol treatment and endothelial vessel staining wasunaltered (Fig. 4C). As a further indication of endothelial dysfunctionserum levels of soluble Icam1, a marker for endothelial cell injury [24],were significantly increased (15%) in hypervitaminotic rats (Fig. 4D).These results show a compartment-specific and drastic effect ofexcess retinol on osteoclasts and the vascular system around corticalmid-diaphysis.

Excess dietary retinol induces bone marrow hypoxia

The lack of normal erythrocyte appearance together with the lackof endothelial staining in the diaphyseal endosteal/marrow compart-ment prompted us to investigate if retinol was delivered properly tothis site via its serum transport protein, Rpb4. Immunostaining ofbone sections for Rbp4 showed an obvious reduction in intraluminalvessel staining in the bone marrow close to the endosteal compart-ment in rats receiving excess retinol (Fig. 5A). Importantly, there wasstill a clear intraluminal vessel Rbp4 staining at periosteal sites(Fig. 5A). To test if the local differential delivery of vitamin A in

diaphyseal bone tissue was reflected in expression of its direct targetswe compared transcript levels from cortical bone (CortB) alone withintact diaphyseal bone including marrow (CortB+M). Notably, themetaphyses and growth plates are not included in these samples. Theamount of RNA extracted from CortB+M was approx. 4 times theamount of RNA in CortB and is here considered to mainly representmarrow cells. Reflecting the Rbp4 staining, real-time PCR analysisshowed the expected clear increase in RA targets (Cyp26a1 [53] andStra6 [54]) in CortB whereas there were no significant differences inexpression levels when marrow compartment was included (Fig. 5B).In line with reduced blood transport to the marrow we also foundincreased expression of hypoxia inducible factor 1α (Hif1α) and twoof its direct target genes (Twist1 [43] and Mmp2 [55]) in the marrowof retinol-treated rats (Fig. 5C). Here, there was no difference incathepsin K expression between control and retinol bones with orwithout marrow, thus osteoclast re-localization was not detected bygene expression analysis (not shown). Remarkably, the expression ofseveral osteoblast markers such as alkaline phosphatase, Runx2, OPGand RANKL were all strongly increased in the marrow of retinolsupplemented animals (Fig. 5D). Importantly and a further indicationof reduced delivery of retinol to the marrow, RA directly represses orhas little effect on the expression of the same osteoblastic and hypoxiagenes in a mouse bone marrow stromal cell line (Fig. 5E). Together,marrow hypoxia and lack of induction of RA targets further suggestreduced blood delivery to the endosteal/marrow compartment inhypervitaminotic animals.

Bones of hypervitaminotic animals have increased degreeof mineralization

Gene expression analysis suggests that the marrow cells inhypervitaminotic animals are osteogenic. Immunostaining of thenumerous cells in the marrow/endosteal pathognomonic compart-ment confirm themRNA expression data and show that these cells arepositive for osteoblastic proteins, including osteocalcin, osteopontinand Mmp2 (Fig. 6A). A further indication of increased endosteal/marrow osteoblast activity (as mineral producing cells) in retinol-treated rat bones is from the biomechanical data as a 3-point bendingtest show preserved stiffness although the bones are thinner(Table 2). This indicates that hypervitaminotic bones are harder(contain more mineral) which is in line with the significantlydecreased displacement, also noticed for the hypervitaminotic bones(Table 2). This, together with the measured reduced force needed forbreaking show that excess retinol induce brittle bones (Table 2). Todirectly test if we could measure changed mineral composition inhypervitaminotic bone tissue we analyzed its mineral content,composition and crystallinity using FTIR Spectroscopy. This revealedthat retinol bones indeed have a higher degree of mineralizationcompared to control bones. Although not statistically significant theamount of mature highly crystalline phosphate in hypervitaminoticbones were reduced (−5.1%), but contained more poorly crystallinephosphate mineral (+2.7%) compared to normal bone (Fig. 6B).Finally, as mineral trafficking in bone is believed to be controlled byosteocytes which are embedded inside mineralized bone and areorganized in an osteocyte lacunar canalicular system (OLCS), weanalyzed those structures. The OCLS can be visualized by Dmp1staining and it has been shown that when the speed of bonedeposition is high, the OLCS are irregular [25]. Significantly, Dmp1staining in the pathognomonic area of hypervitaminotic bones wasdiffuse and irregular, indicative of fast growing immature/non-remodeled endosteal bone (Fig. 6C). Importantly, at both theperiosteal site and mid-cortical the pattern of the OCLS in hypervi-taminotic bones was preserved indicating that these sites are notaffected although the staining intensity at the periosteal site wasstronger in hypervitaminotic bones (Fig. 6C). These results suggest a

Fig. 3. Hematoxylin staining of diaphyseal bone. (A) Lowmagnification (upper panel) of decalcified sections of humeri shows an apparent excessive vascularization at the endostealsite in excess vitamin A (Ex. vitA) bone (b=cortical bone, asterix=pathognomonic area). Higher magnification (lower panel) reveals clear streaks of erythrocytes (arrows) in bothlacunae and close to endosteal surface, in control (cont) sections. In excess vitamin A bone no erythrocytes are visible in the pathognomonic area or close to the endosteal surface.Bar=100 μm in upper panel and 25 μm in lower panel. (B) In controls, a detailed examination of bone sections consistently show intraluminal capillary erythrocytes (arrows) at theendosteal and intracortical sites but rarely visible in the periosteum (top panel). On the contrary, bones that have received excessive amounts of vitamin A appear to lack distinct,capillary associated erythrocytes and instead show either scattered cells in stromal tissue (arrowhead) adjacent to bone or no erythrocytes. Asterix=periosteum, arrows=normalvessel structures, arrowhead=enlarged periosteal vessel or hemorrhage. Bar=25 μm. (C) Number of endosteal microvessel associated erythrocytes (within 50 μm of the endosteallining cells) in hypervitaminotic pathognomonic areas and representative areas in control bone (3 animals/group and 2 sites/animal).

501T. Lind et al. / Bone 48 (2011) 496–506

local endosteal/marrow increase in mineralization in hypervitamin-osis A animals.

Discussion

We have previously shown experimentally that a subclinical doseof vitamin A for three months causes weakening of bones in rats,

without reduction or slight increase of BMD, depending on themethod of calculation [14]. Here we show that three times thissubclinical dose of vitamin A in food induces a measurable reductionin both diaphyseal cortical area and bone strength after only 8 days inyounger male rats. Serummarkers for osteoclast number, (TRACP 5b),differentiation, (RANKL) and activity, (CTX-1) were also found to bedecreased in vitamin A supplemented animals. In previous rat studies,

Fig. 4. Endothelial marker analysis of bone and serum. (A) Staining blood vessel cells (endothelial) with Icam1 in controls show thin circular staining close to endosteal bone (leftpart of picture, b=bone, asterix=central large vessel). In bones from rats on a high vitamin A diet there is a distinct reduction in staining throughout the marrow except for the largecentral vessel. Lower panel show a close-up view of the endosteal/marrow area. Bar=100 μm in upper panel and 25 μm in lower panel. (B) In controls, periosteal vessels are hardlyvisible using different endothelial immunological markers (left panel, arrows) whereas in hypervitaminotic animals endothelial staining of the periosteum shows presence ofengorged vessels (arrowheads). Bar=25 μm. (C) Icam1 staining of trabecular compartment. (D) Serum levels of soluble Icam1measured by ELISA (cont, n=11 and Ex. vitA, n=12).

502 T. Lind et al. / Bone 48 (2011) 496–506

hypervitaminosis A has been associated with increased markers forbone resorption, as determined by urinary secretion of hydroxypro-line [23] or serum tartrate resistant acid phosphatase (TRACP) activity[12]. The discrepancy with our results could be explained by severalexperimental differences such as time of sampling (day 8 here, day 5for Tracp activity and day 20 for hydroxyproline secretion) or the factthat earlier results rely on less specific biochemical analyses and thusdoes not specifically reflect osteoclast number/function [23] or theroute of administration i.e. oral and subcutaneous [12]. Subcutane-ously injected synthetic retinoids have been used together withthyroparathyroidectomy in rats as a model to study increased boneresorption to evaluate osteoclast inhibitors in vivo [26]. Today thestandard osteoporosis animal model is ovariectomy (OVX), whichmimics postmenopausal osteoporosis. OVX affects mostly trabecularbone and has less effect on cortical bone as reflected in reducedresistance to compression of vertebrae but preserved bendingstrength of long bones. It is important to stress the differencesbetween vitamin A-induced fragile bones and the OVX model. Inaddition, although the present study on hypervitaminosis A and boneweakening involves young male rats, similar results are noticed with

both young and aged female rats [12,13]. The dissimilarity observed inlong-bone bending strength between OVXed and hypervitaminosis Aanimals could at least in part be explained by the observation thatvitamin A affects bone perimeter considerably more than BMD,compared to the opposite in ovariectomized animals. Furthermore,OVX induces a persistent increase in serummarkers of both osteoclastand osteoblast activities [27,28], which is contrary to our findings ofexcess dietary vitamin A. Also, OVX treatment induces, whereasexcess vitamin A reduces weight gain. It is well known that growthand food intake influences the growth hormone/insulin-like growthfactor 1 (IGF-1)-axis which is a major regulator of postnatal growth,including bones. In our study hypervitaminotic rats gained less weightwhich could be a consequence of reduced serum levels of IGF-1 [29].However, the consequence of different serum IGF-1 levels on bonegrowth is complex and it appears that local tissue produced IGF-1 ismore important for maintaining skeletal integrity than the totalserum levels [30]. Also a recent study investigating the impact of shortterm calorie restriction on bone parameters showed that calorierestriction reduced serum IGF-1 and PINP without affecting CTX-1levels [31]. They also showed that calorie restriction decreases

Fig. 5. Bone compartment-specific vitamin A delivery and gene expression. (A) Immunohistochemical staining of bone sections for the serum retinol-binding protein, Rbp4, showsclear intralumenal vessel staining in endosteal and periosteal compartment in control rats (top panel, arrows). In contrast, although animals on excess vitamin A diet showintraluminal vessel staining at the periosteal site (right lower panel, arrows) there is complete lack of staining at the endosteal site (left lower panel, arrowheads). (Bar=25 μm).(B) TaqMan analysis of RA direct target genes in CortB and CortB+M (n=3 or 4 per group). (C) Hypoxia induced gene expression in CortB and CortB+M (n=3 or 4 per group).(D) CortB+M expression of osteoblastic genes (n=3 or 4 per group). (E) In vitro analysis of how RA affects osteoblastic and hypoxia gene expression in a murine bone marrowstromal cell line (n=3 per group).

503T. Lind et al. / Bone 48 (2011) 496–506

expression of marrow osteoblastic genes and decreases bone stiffnesswhich is in sharp contrast to our findings indicating that the effects wesee in our study are mainly vitamin A specific. Furthermore, manyhypervitaminosis A studies have investigated other serum factorsinfluencing bone remodeling such as: Ca, phosphate, vitamin D andPTH but there are generally very little changes in these parameters[9,10,14,23,32].

We did not notice any changes in the staining of markers forosteoclast or endothelial cells in metaphyseal trabecular bone

compartment which probably reflects that this area is better perfusedthan the diaphysis. Instead the indications of reduced diaphysealendosteal/marrow blood flow in hypervitaminotic animals are: 1)lack of visible capillary erythrocytes, 2) absence of endothelialstaining, 3) loss of luminal blood vessel labeling for the retinol-binding protein, 4) lack of induced expression of RA targets, and 5)increased expression of hypoxia induced genes. A further indication ofdisturbed blood distribution in hypervitaminotic animals was theincreased serum levels of a marker for endothelial injury, sIcam1 [24].

Fig. 6. Immunohistological appearance of osteogenic cells, bone mineral analysis andosteocyte network analysis. (A) Osteoblastic staining of control and pathognomonicendosteal area in hypervitaminosis A diaphyseal long bones (bar=25 μm). (B) FTIRanalysis of the bonemineral shows that hypervitaminotic bones have a higher degree ofmineralization and that the mineral shows a tendency towards being less crystalline(n=12 per group). (C) Staining of the osteocyte network with Dmp1 shows in controlsstrong staining (arrowheads) close to the marrow and weak staining (arrows) close tothe periosteum. In the hypervitaminosis A bone (lower panel) there is strong stainingclose to the periosteum and staining of the endosteal pathognomonic area is weak anddiffuse.

504 T. Lind et al. / Bone 48 (2011) 496–506

The endosteum is a niche with abundant osteoclasts in young bone,and with age there is an increase of periosteal osteoclasts. In line withearlier observations we find increased numbers of periostealosteoclasts in hypervitaminotic rat bones. However, the novelobservation in this study is that hypervitaminotic animals display aremarkable lack of endosteal osteoclasts. Since the normal life span ofosteoclasts in vivo are between 2–4 weeks, inhibition of formation isunlikely to be a major mechanism. Instead, since vitamin A has knownantiangiogenic effects, a reduced number of blood vessels penetratingthe cortical bone may explain how periosteal osteoclasts suddenlyappear at the same time as endosteal ones disappear. An associatedprolonged endosteal hypoxia could in turn induce osteoclastapoptosis [52]. It is also interesting to note the appearance ofperiosteal osteoclasts considering the potent capacity of RA to inhibit

osteoclast formation in vitro [33]. However this discrepancy couldinvolve special local osteoclast precursors which readily transforminto mature osteoclasts upon stimulation [34]. In line with this, organculture experiments show that RA induces bone degradation of bothcalvarial and long bones [35,36]. Although the endosteum inhypervitaminosis A animals has been described with a pathogno-monic appearance resembling highly vascularized bone lacunae wewere unable to label these structures with endothelial markers and aclose histological examination revealed a lack of erythrocytes in theseareas, indicating that these structures do not contain functional bloodvessels. In fact, hypervitaminotic bones showed a clear gradualreduction of labeling for endothelial markers beginning from the largecentral marrow vessel. A disturbed endothelial function is consistentwith regular findings of hemorrhage in hypervitaminosis A and Ingberand Folkman observed over two decades ago that retinoids are potentantiangiogenic factors [42]. Similar observations have also beenobserved in the clinic, in RA treated patients with acute promyelocyticleukemia (APL). Twenty-five percent of APL patients treated with RAdevelop an adverse complication called retinoic acid syndrome (RAS)which is manifested by fever, respiratory distress, pulmonary edema,episodic hypotension, and occasionally bone pain [37], bone marrowfailure [38] and bone marrow necrosis and fibrosis [39]. Importantly,post-mortem studies have suggested that the series of events leadingto RAS involves damaged microvasculature and moreover bonemarrow biopsies from APL patients before and after treatment withRA, show that RA suppresses angiogenesis [40,41]. These clinicalobservations are in line with our findings of disturbed endosteal/marrow blood flow. Also consistent with this, marrow fromhypervitaminotic animals was found to have increased expressionlevels of hypoxia induced genes. It is well acknowledged thatepithelial (or endothelial) cells transform to a mesenchymal type(EMT, a key profibrotic stimulus) during hypoxia. EMT involves anincreased expression of alpha smooth muscle actin (αSMA) andTwist1 expression [43]. In accordance with this we find in hypervi-taminotic bone marrow an exclusive increase of Twist1 expressionand an αSMA induction capacity of RA has been shown in previousangiogenesis experiments [44]. To get a feeling for the complexity ofhypoxia at the gene expression level it was shown in transcriptomeexperiments with colon carcinoma cells that 4005 genes were alteredby a 24 h hypoxic exposure [45]. In vivo experiments with geneticallymanipulated mice harboring osteoblastic overexpression of Hif1αdevelop extremely dense bones [46]. It was concluded that increasedbone formation was tightly connected to increased blood flow.However, it has also been shown that reduced or lack of blood flowin bones increases BMD or pathological calcification of the marrow,respectively [47,48]. Here we found increased labeling and expressionof osteoblastic genes in the marrow of hypervitaminotic animalsapparently lacking functional blood vessels, which resembles descrip-tions of pathological calcification occurring in ischemic bone marrow[49]. The mechanism behind this may involve precipitation ofhydroxylapatite emanating from ischemic tissue, as describedrecently when the appearance of abnormal calcification after livertransplants were investigated [50]. Here a direct measurement of thebone mineral content using FTIR showed clearly that hypervitamino-tic bones contain more phosphate mineral relative to organic matrixcontent compared to control bones. Although apparently contradic-tory as this increase is not reflected as an increase in cortical BMDfrom the pQCT data, this probably reflects the limitations of the pQCTmethod to accurately measure the density of thin cortical bone. It hasbeen shown that pQCT underestimates BMD values if cortical bonethickness decreases, as is the case with our hypervitaminosis A bones[51]. The FTIR analysis further indicated that the mineral was ofimmature nature, suggesting that it was not normal bone. Abnormalendosteal bone was also suggested by the diffuse and irregularappearance of the osteocyte lacunar canalicular system in thepathognomonic areas of hypervitaminotic bones. Additional evidence

505T. Lind et al. / Bone 48 (2011) 496–506

of increased mineralization of hypervitaminotic bones was providedby the biomechanical properties which changed after 1 week ofexcess vitamin A. Although hypervitaminotic bones were thinner,stiffness was unchanged and the displacement before failure wassignificantly less. Both these parameters indicate that hypervitami-notic bones were harder and contained more mineral compared tocontrol bones. Furthermore, bending strength is not only dependenton cortical area but also of bone width as is shown by the formula thatthe radius of long-bone diaphysis will decrease its resistance totorsion or bending loads by a factor raised to the fourth power. Thus, adecrease in failure load cannot readily be explained by a similardecrease in cortical area without taking the bone width into account.Also, in this study hypervitaminotic rats show decreased osteoblasticmarkers in serum, in spite of increased osteoblastic expression andmineralization in their bones. We believe this is a consequence ofreduced blood flow around the endosteal site, the very site where theincreasedmineralization takes place, thus resulting in local inability ofproper delivery of osteoblastic markers from these sites by secretionto the bloodstream. Collectively, our data indicate that excessiveingestion of vitamin A induces both bone thinning and increasedendosteal mineralization, leading to brittle bones.

In conclusion, we have identified that a relatively rapid primaryeffect of excess dietary retinol appears to be a reduction in the numberof blood vessels penetrating the cortical bone. This in turn leads tohypoxia in the endosteal/marrow compartment, reversal of osteoclastpositioning around diaphyseal bone and ultimately to pathologicalendosteal mineralization and brittle bones. Revealing the keyinitiating mechanism of how excess retinol impacts bone tissue isimportant for designing new observational studies and for under-standing/interpretation of acquired data. It is important to recognizenovel types of brittle bone conditions which differ from the “standard”postmenopausal osteoporosis condition and is not detected by BMDmeasurements. Notably, in spite of the rapid bone thinning, serummarkers for bone resorption were reduced in hypervitaminoticanimals and the resorption index (CTX-1/TRACP 5b) was unaltered.Together these results suggest that excess dietary vitamin A (andpossibly other potent antiangiogenic substances) reduces long-bonebending strength without a large effect on BMD or resorption index,possibly explaining the apparent inconsistencies in observationalstudies.

Acknowledgments

This work was supported by, The Swedish Society of Medicine(TL), The Swedish Medical Research Council (HM and GA) andMagnus Bergvalls Stiftelse (AS).We thankMaria Norgård for excellenttechnical assistance and Mark Birch for critical review of themanuscript.

References

[1] Binkley N, Krueger D. Hypervitaminosis A and bone. Nutr Rev 2000;58:138–44.[2] Blomhoff R, Blomhoff HK. Overview of retinoid metabolism and function. J

Neurobiol 2006;66:606–30.[3] Hua S, Kittler R, White KP. Genomic antagonism between retinoic acid and

estrogen signaling in breast cancer. Cell 2009;137:1259–71.[4] WHO. Global prevalence of vitamin A deficiency in populations at risk 1995–2005.

Geneva: World Health Organization; 2009.[5] Melhus H, Michaëlsson K, Kindmark A, Bergström R, Holmberg L, Mallmin H, et al.

Excessive dietary intake of vitamin A is associated with reduced bone mineraldensity and increased risk for hip fracture. Ann Intern Med 1998;129:770–8.

[6] Michaëlsson K, Lithell H, Vessby B, Melhus H. Serum retinol levels and the risk offracture. N Engl J Med 2003;348:287–94.

[7] Ribaya-Mercado JD, Blumberg JB. Vitamin A: is it a risk factor for osteoporosis andbone fracture? Nutr Rev 2007;65:425–38.

[8] Moore T, Wang YL. Hypervitaminosis A. Biochem J 1945;39:222–8.[9] Wolbach SB. Vitamin-A deficiency and excess in relation to skeletal growth. J Bone

Joint Surg 1947;29:171–92.[10] Clark I, Smith MR. Effects of hypervitaminosis A and D on skeletal metabolism. J

Biol Chem 1964;239:1266–71.

[11] Melhus H. Vitamin A and fracture risk. In: New SA, Bonjour J-P, editors. Nutritionalaspects of bone health, The Royal Society of Chemistry. Cambridge RSC publishing;2003. p. 368–402.

[12] Kneissel M, Studer A, Cortesi R, Susa M. Retinoid-induced bone thinning is causedby subperiosteal osteoclast activity in adult rodents. Bone 2005;36:202–14.

[13] Johansson S, Lind PM, Hakansson H, Oxlund H, Orberg J, Melhus H. Subclinicalhypervitaminosis A causes fragile bones in rats. Bone 2002;31:685–9.

[14] Lind PM, Johansson S, Rönn M, Melhus H. Subclinical hypervitaminosis A in rat:measurements of bone mineral density (BMD) do not reveal adverse skeletalchanges. Chem Biol Interact 2006;159:73–80.

[15] Boskey AL, Gadaleta S, Gundberg C, Doty SB, Ducy P, Karsenty G. Fourier transforminfrared microspectroscopic analysis of bones of osteocalcin-deficient miceprovides insight into the function of osteocalcin. Bone 1998;23:187–96.

[16] Lind PM, Larsson S, Johansson S, Melhus H, Wikström M, Lindhe O, et al. Bonetissue composition, dimensions and strength in female rats given an increaseddietary level of vitamin A or exposed to 3, 3′, 4, 4′, 5-pentachlorobiphenyl(PCB126) alone or in combination with vitamin C. Toxicology 2000;151:11–23.

[17] Risteli J, Elomaa I, Niemi S, Novamo A, Risteli L. Radioimmunoassay for thepyridinoline cross-linked carboxy-terminal telopeptide of type I collagen: a newserum marker of bone collagen degradation. Clin Chem 1993;39:635–40.

[18] Hollberg K, Hultenby K, Hayman A, Cox T, Andersson G. Osteoclasts from micedeficient in tartrate-resistant acid phosphatase have altered ruffled borders anddisturbed intracellular vesicular transport. Exp Cell Res 2002;279:227–38.

[19] Hollberg K, Marsell R, Norgård M, Larsson T, Jonsson KB, Andersson G. Osteoclastpolarization is not required for degradation of bone matrix in rachitic FGF23transgenic mice. Bone 2008;42:1111–21.

[20] Melhus H, Båvik CO, Rask L, Peterson PA, Eriksson U. Epitope mapping of amonoclonal antibody that blocks the binding of retinol-binding protein to itsreceptor. Biochem Biophys Res Commun 1995;210:105–12.

[21] Kang S, Bennett CN, Gerin I, Rapp LA, Hankenson KD, Macdougald OA. Wntsignaling stimulates osteoblastogenesis of mesenchymal precursors by suppres-sing CCAAT/enhancer-binding protein alpha and peroxisome proliferator-activat-ed receptor gamma. J Biol Chem 2007;282:14515–24.

[22] Rodriguez-Navarro AB, Romanek CS, Alvarez-Lloret P, Gaines KF. Effect of in ovoexposure to PCBs and Hg on Clapper Rail bone mineral chemistry from acontaminated saltmarsh in coastal Georgia. Environ Sci Technol 2006;40:4936–42.

[23] Hough S, Avioli LV, Muir H, Gelderblom D, Jenkins G, Kurasi H, et al. Effects ofhypervitaminosis A on the bone andmineral metabolism of the rat. Endocrinology1988;122:2933–9.

[24] Horstman LL, Jy W, Jimenez JJ, Ahn YS. Endothelial microparticles as markers ofendothelial dysfunction. Front Biosci 2004;9:1118–35.

[25] Hirose S, Li M, Kojima T, de Freitas PH, Ubaidus S, Oda K, et al. A histologicalassessment on the distribution of the osteocytic lacunar canalicular system usingsilver staining. J Bone Miner Metab 2007;25:374–82.

[26] Trechsel U, Stutzer A, Fleisch H. Hypercalcemia induced with an arotinoid inthyroparathyroidectomized rats. New model to study bone resorption in vivo. JClin Invest 1987;80:1679–86.

[27] Rissanen JP, Suominen MI, Peng Z, Halleen JM. Secreted tartrate-resistant acidphosphatase 5b is a marker of osteoclast number in human osteoclast cultures andthe rat ovariectomy model. Calcif Tissue Int 2008;82:108–15.

[28] Rissanen JP, Suominen MI, Peng Z, Morko J, Rasi S, Risteli J, et al. Short-termchanges in serum PINP predict long-term changes in trabecular bone in the ratovariectomy model. Calcif Tissue Int 2008;82:155–61.

[29] Singer CF, Hudelist G, Schreiber M, Kubista E. Pharmacological modulation of IGFserum concentrations as a therapeutic approach to control the growth ofmalignant breast tumors. Drugs Today (Barc) 2003;39:115–25.

[30] Yakar S, Rosen CJ, Bouxsein ML, Sun H, Mejia W, Kawashima Y, et al. Serumcomplexes of insulin-like growth factor-1 modulate skeletal integrity andcarbohydrate metabolism. Faseb J 2009;23:709–19.

[31] Bielohuby M, Matsuura M, Herbach N, Kienzle E, Slawik M, Hoeflich A, et al. Shortterm exposure to low-carbohydrate/high fat diets induces low bone mineraldensity and reduces bone formation in rats. J Bone Miner Res 2010;25:275–84.

[32] Nerurkar MK, SahasrabudheMB. Metabolism of calcium, phosphorus and nitrogenin hypervitaminosis A in young rats. Biochem J 1956;63:344–9.

[33] Conaway HH, Persson E, Halén M, Granholm S, Svensson O, Pettersson U, et al.Retinoids inhibit differentiation of hematopoietic osteoclast progenitors. Faseb J2009;23:3526–38.

[34] Mizoguchi T, Muto A, Udagawa N, Arai A, Yamashita T, Hosoya A, et al.Identification of cell cycle-arrested quiescent osteoclast precursors in vivo. J CellBiol 2009;184:541–54.

[35] Kindmark A, Melhus H, Ljunghall S, Ljunggren O. Inhibitory effects of 9-cis and all-trans retinoic acid on 1,25 (OH)2 vitamin D3-induced bone resorption. CalcifTissue Int 1995;57:242–4.

[36] Fell HB, Mellanby E. The effect of hypervitaminosis A on embryonic limb bonescultivated in vitro. J Physiol 1952;116:320–49.

[37] Larson RS, TallmanMS. Retinoic acid syndrome:manifestations, pathogenesis, andtreatment. Best Pract Res Clin Haematol 2003;16:453–61.

[38] Perrotta S, Nobili B, Rossi F, Criscuolo M, Iolascon A, Di Pinto D, et al. Infanthypervitaminosis A causes severe anemia and thrombocytopenia: evidence of aretinol-dependent bone marrow cell growth inhibition. Blood 2002;99:2017–22.

[39] Hatake K, Uwai M, Ohtsuki T, Tomizuka H, Izumi T, Yoshida M, et al. Rare butimportant adverse effects of all-trans retinoic acid in acute promyelocyticleukemia and their management. Int J Hematol 1997;66:13–9.

[40] Kini AR, Peterson LA, Tallman MS, Lingen MW. Angiogenesis in acute promye-locytic leukemia: induction by vascular endothelial growth factor and inhibitionby all-trans retinoic acid. Blood 2001;97:3919–24.

506 T. Lind et al. / Bone 48 (2011) 496–506

[41] Tallman MS, Andersen JW, Schiffer CA, Appelbaum FR, Feusner JH, Ogden A, et al.Clinical description of 44 patients with acute promyelocytic leukemia whodeveloped the retinoic acid syndrome. Blood 2000;95:90–5.

[42] Ingber D, Folkman J. Inhibition of angiogenesis through modulation of collagenmetabolism. Lab Invest 1988;59:44–51.

[43] Yang MH, Wu MZ, Chiou SH, Chen PM, Chang SY, Liu CJ, et al. Direct regulation ofTWIST by HIF-1alpha promotes metastasis. Nat Cell Biol 2008;10:295–305.

[44] Suzuki Y, Komi Y, Ashino H, Yamashita J, Inoue J, Yoshiki A, et al. Retinoic acidcontrols blood vessel formation by modulating endothelial and mural cellinteraction via suppression of Tie2 signaling in vascular progenitor cells. Blood2004;104:166–9.

[45] Lendahl U, Lee KL, Yang H, Poellinger L. Generating specificity and diversity in thetranscriptional response to hypoxia. Nat Rev Genet 2009;10:821–32.

[46] Wang Y, Wan C, Deng L, Liu X, Cao X, Gilbert SR, et al. The hypoxia-inducible factoralpha pathway couples angiogenesis to osteogenesis during skeletal development.J Clin Invest 2007;117:1616–26.

[47] Catto M. Ischaemia of bone. J Clin Pathol Suppl R Coll Pathol 1977;11:78–93.[48] Bloomfield SA, Hogan HA, DelpMD. Decreases in bone blood flow and bonematerial

properties in aging Fischer-344 rats. Clin Orthop Relat Res 2002;396:248–57.[49] Lehner CE, AdamsWM, Dubielzig RR, Palta M, Lanphier EH. Dysbaric osteonecrosis

in divers and caisson workers. An animal model. Clin Orthop Relat Res 1997;344:320–32.

[50] Kalantari F, Miao D, Emadali A, Tzimas GN, Goltzman D, Vali H, et al. Cellular andmolecular mechanisms of abnormal calcification following ischemia–reperfusioninjury in human liver transplantation. Mod Pathol 2007;20:357–66.

[51] Brodt MD, Pelz GB, Taniguchi J, Silva MJ. Accuracy of peripheral quantitativecomputed tomography (pQCT) for assessing area and density of mouse corticalbone. Calcif Tissue Int 2003;73:411–8.

[52] Knowles HJ, Athanasou NA. Acute hypoxia and osteoclast activity: a balancebetween enhanced resorption and increased apoptosis. J Pathol 2009;218:256–64.

[53] White JA, Beckett-Jones B, Guo YD, Dilworth FJ, Bonasoro J, Jones G, et al. cDNAcloning of human retinoic acid-metabolizing enzyme (hP450RAI) identifies anovel family of cytochromes P450. J Biol Chem 1997;272:18538–41.

[54] Taneja R, Bouillet P, Boylan JF, Gaub MP, Roy B, Gudas LJ, et al. Reexpression ofretinoic acid receptor (RAR) gamma or overexpression of RAR alpha or RAR beta inRAR gamma-null F9 cells reveals a partial functional redundancy between thethree RAR types. Proc Natl Acad Sci USA 1995;92:7854–8.

[55] Lee SH, Lee YJ, Han HJ. Effect of arachidonic acid on hypoxia-induced IL-6production in mouse ES cells: involvement of MAPKs, NF-kappaB, and HIF-1alpha.J Cell Physiol 2010;222:574–85.

[56] Mallia AK, Smith JE, Goodman DW. Metabolism of retinol-binding protein andvitamin A during hypervitaminosis A in the rat. J Lipid Res 1975;16:180–8.

[57] Smith FR, Goodman DS. Vitamin A transport in human vitamin A toxicity. N Engl JMed 1976;294:805–8.