Elevated Aromatase Expression in Osteoblasts Leads to Increased Bone Mass Without Systemic Adverse Effects

JOURNAL OF BONE AND MINERAL RESEARCHVolume 24, Number 7, 2009Published online on February 16, 2009; doi: 10.1359/JBMR.090208� 2009 American Society for Bone and Mineral Research

Elevated Aromatase Expression in Osteoblasts Leads to IncreasedBone Mass Without Systemic Adverse Effects

Klara Sjogren,1 Marie Lagerquist,1 Sofia Moverare-Skrtic,1 Niklas Andersson,1 Sara H. Windahl,1

Charlotte Swanson,1 Subburaman Mohan,2 Matti Poutanen,3 and Claes Ohlsson1

ABSTRACT: The stimulatory effects of testosterone (T) on bone can either be through a direct activation ofthe androgen receptor (AR) or mediated through aromatization of T to estradiol (E2), followed by activationof estrogen receptors (ERs) in bone. Aromatase expression in osteoblasts and reproductive tissues is de-pendent on different promoters, which are differentially regulated. To study the effect of elevated localaromatization of T to E2 in bone, we developed a transgenic mouse model (Coll-1a1-Arom) that over-expresses the human aromatase gene under the control of the osteoblast specific rat type I a I procollagenpromoter. The Coll-1a1-Arom mice expressed human aromatase mRNA specifically in bone and had un-affected serum E2 and T levels. Male Coll-1a1-Arom mice had clearly increased total body BMD, trabecularBMD, cortical BMD, and cortical thickness associated with elevated osteoprotegerin mRNA levels andreduced number of osteoclasts (p < 0.01). Treatment of ovariectomized mice with T increased cortical andtrabecular thickness in the Coll-1a1-Arom mice (p < 0.001) but not in the wildtype mice. In conclusion,elevated aromatase expression specifically in osteoblasts results in stimulatory estrogenic effects in bonewithout increasing serum E2 levels. Because osteoblast-specific aromatase expression results in an increasedER to AR activation ratio in bone, we propose that activation of ERs results in a more pronounced increasein bone mass than what is seen after activation of the AR. Development of osteoblast-specific inducers ofaromatase expression might identify substances with stimulatory effects on bone without systemic adverseeffects.J Bone Miner Res 2009;24:1263–1270. Published online on February 16, 2009; doi: 10.1359/JBMR.090208

Key words: bone, hormone replacement therapy, estrogen, CYP19A1, sex hormones

Address correspondence to: Claes Ohlsson, MD, PhD, Division of Endocrinology, Department of Internal Medicine,Sahlgrenska University Hospital, SE-41345 Goteborg, Sweden, E-mail: [email protected]

INTRODUCTION

SEX STEROIDS ARE important for the maintenance of boththe female and the male skeleton.(1–3) However, the

relative contribution of androgens versus estrogens in theregulation of the skeleton is unclear. The effects of tes-tosterone (T) can be exerted either directly through theandrogen receptor (AR) or indirectly through aromatiza-tion to estrogens and further through estrogen receptor(ER)a and/or ERb.(2,3) All these receptors are expressedboth in growth plate cartilage and in bone.(4–7) ERa isthe major ER responsible for the trabecular bone-sparingeffect of estrogens both in males and females.(8–10) Fur-thermore, it has been shown that the in vivo effect ofER activation on bone is distinct from the effect of ARactivation.(9–11)

The important role of aromatase for the skeleton isclearly shown from clinical studies of men with mutation incytochrome P450, family 19, subfamily A, polypeptide1 (CYP19A1), the gene that encodes aromatase, and from

studies in aromatase-deficient mice, showing that defi-ciency of aromatase results in reduced bone mass.(12–15)

Some studies showed a reduced bone size and a reducedperiosteal apposition in aromatase-deficient men, suggest-ing that estrogen is required for a normal periosteal ap-position during sexual maturation.(16–18) Furthermore,male transgenic (TG) mice with a general overexpressionof human aromatase, using the human ubiquitin C pro-moter/human P450 aromatase fusion gene, have increasedbone mass.(19) However, as a consequence of elevated se-rum E2 levels in this mouse model, male mice showedseveral systemic pathological phenotypes, including infer-tility, rudimentary prostate/seminal vesicles, reduced se-rum T, gynecomastia, and adrenal gland hyperplasia.(20–22)

Thus, the role of aromatase expression specifically in os-teoblasts without affected circulating E2 levels could notbe determined.

In humans, the aromatase is expressed in several dif-ferent tissues including in bone/osteoblasts.(3) Therefore,local levels of E2 and T in osteoblasts are not only de-pendent on serum levels of E2 and T but also on the ar-omatase activity in osteoblasts. There are some indications

1Center for Bone Research, Institute of Medicine, University of Gothenburg, Sahlgrenska Academy, Goteborg, Sweden; 2Musculo-skeletal Diseases Center, Jerry L Pettis VA Medical Center, Loma Linda, California, USA; 3Institute of Biomedicine, Department ofPhysiology, University of Turku, and Turku Center for Disease Modeling, Turku, Finland.

The authors state that they have no conflicts of interest.

1263

that endogenous aromatase activity in bone might be moreimportant in humans than in mice. First, aromatase ex-pression in bone is relatively higher in humans than inmice. Second, mice have no expression of CYP 17, theenzyme responsible for sex steroid synthesis, in the adrenalgland, and the gonads are the exclusive site for sex steroidproduction in mice.(23) The situation is different in humansthat, in addition to gonadal-derived androgens, humanshave androgens secreted from the adrenal gland that canbe converted to E2 by aromatase in peripheral tissues suchas bone.(24) The human CYP19A1 gene is located in thechromosome 15q21.2 region and is comprised of a 30-kbcoding region and a 93-kb regulatory region. The unusuallylarge regulatory region contains 10 tissue-specific pro-moters (I.1, I.2, I.3, I.4, I.5, I.6, I.7, 2a, I.f, and PII) that arealternatively used in various cell types.(25) Each promoteris regulated by a distinct set of regulatory sequences inDNA and transcription factors that bind to these specificsequences. Aromatase expression in ovary, which is themajor source of E2 in fertile women, is mainly regulatedby the PII promoter and is strongly induced by follicle-stimulating hormone (FSH).(26) In contrast, promoters I.4and 1.6 have been described as the major promoters reg-ulating aromatase expression in bone/osteoblasts, and ithas been shown that vitamin D, dexamethasone, prosta-glandin E2 (PGE2), and dehydroepiandrosterone regulatearomatase expression in cultured osteoblasts.(25,27–32)

Moreover, in estrogen-dependent pathologic tissues suchas breast cancer and endometriosis, aromatase is upregu-lated through inappropriate activation of aberrant pro-moters.(25) Because the use of aromatase promoters inhumans differs between tissues, one can postulate that it ispossible to achieve a tissue-specific estrogenic effect (e.g.,on the bone) by substances that specifically activates bonespecific aromatase promoters. These substances wouldavoid the adverse effects in nonskeletal tissues. However,an osteoblast-specific induction of aromatase would resultnot only in elevated local estradiol (E2) but also reducedlocal T, and the effect of the resulting change in the balancebetween ER and AR activation in osteoblasts is unknown.

To study the importance of high local aromatization of Tto E2 and thereby increased ER to AR activation ratiospecifically in osteoblasts, we developed a TG mousemodel (Coll-1a1-Arom mice) that overexpresses the hu-man aromatase gene under the control of the osteoblastspecific rat type I a I procollagen promoter.

MATERIALS AND METHODS

Construction of the transgene

A 2.3-kb fragment of the rat type I a I procollagenpromoter was subcloned into the multiple cloning site ofthe pGL3-Basic Vector. This promoter fragment has ear-lier been shown to give efficient expression restricted toosteoblasts and osteocytes with no detectable expressionin chondrocytes.(33–35) A 2.2-kb fragment coding for full-length human aromatase cDNA was subcloned into themultiple cloning site of the pGL3-Basic Vector after thepromoter followed by a SV40 polyadenylation signal.

The 5.4-kb-long Coll 1a1-Aromatase fragment was re-leased from the vector backbone by digestion with NotIand HindIII enzymes.

Development of Coll-1a1-Arom transgenicfounder mice

TG mice were produced by a standard technique in aC57Bl6 strain. Positive founders for the transgene wereidentified by Southern blot analysis of DNA obtained fromtail biopsies. Briefly, genomic DNA (10 mg) was digestedwith XhoI and resolved by electrophoresis in 0.5% agarosegel. The DNA was blotted onto nylon membrane andprehybridized for 15 min at 658C in hybridization buffer(1.5 3 sodium chloride–sodium phosphate–EDTA buffer[SSPE], 10% polyethylene glycol [PEG] 6000, 7% wt/volSDS, 100 mg/ml salmon sperm DNA) and with [a-32P]dCTP(Amersham Pharmacia Biotech)-labeled Coll 1a1 aro-matase fragment overnight at 658C. The membranes werewashed in 13 SSC and 0.1% SDS three times at 658C for 15min and exposed to a PhosphorImager screen.

Establishment of a Coll-1a1-Arom transgenicmouse line

Coll-1a1-Arom mice showed normal fertility. To avoidexposure to the transgene during fetal life, all experimentswere carried out on mice born from crossing a male Coll-1a1-Arom mouse with a female C57Bl6 female mouse. WTlittermates were used as control groups. The mice werehoused in a standard animal facility under controlledtemperature (228C) and photoperiod (12 h of light, 12 h ofdark) and fed standard phytoestrogen-free pellet diet adlibitum. Animal care was in accordance with institutionalguidelines. All animal experiments had been approved bythe local Ethical Committees for Animal Research at theUniversity of Gothenburg.

PCR genotyping of transgenic mice

For routine genotyping of the Coll-1a1-Arom mice,PCR analyses were carried out using DNA extracted fromtail biopsies. Primer sequences are available on request.

Testosterone treatment of mice

Twelve-week-old female Coll-1a1-Arom and WT micewere either sham operated or ovariectomized (OVX). TheOVX mice were treated with vehicle or T (25 mg/d) for4 wk. The administration was done using slow releasepellets (Innovative Research of America, Sarasota, FL,USA). The dose corresponds to a replacement dose givinga physiological T level in orchidectomized male mice atsimilar weight (unpublished data).

Quantitative real-time PCR analysis

Total RNA was prepared from retroperitoneal fat, gas-trocnemius muscle, and femur (n = 4–10/group), usingTriZol Reagent (Life Technologies) according to themanufacturer’s instructions. The RT-PCR analysis wasperformed using the ABI Prism 7000 Sequence DetectionSystem (PE Applied Biosystems). The mRNA abundance

1264 SJOGREN ET AL.

of each gene was calculated using the ‘‘standard curvemethod’’ (User Bulletin 2; PE Applied Biosystems) andadjusted for the expression of 18S. Primer and probe se-quences are available on request.

DXA

Analyses of total body areal BMD (aBMD) and spineBMD were performed by DXA using the Lunar PIXImusMouse Densitometer (Wipro GE Healthcare).

pQCT

CT scans were performed with the pQCT XCT RE-SEARCH M (version 4.5B; Norland), operating at a reso-lution of 70 mm as described previously.(36) TrabecularBMD was determined ex vivo, with metaphyseal pQCTscans of the distal femur and the proximal tibia. Bonelengths were measured with a slide caliper. The scans werepositioned in the metaphysis at a distance from the growthplate corresponding to 3% of the total length of the bone,and the trabecular bone region was defined as the inner 45%of the total cross-sectional area. Cortical bone parameters(cortical BMD, cortical BMD, cortical thickness, periostealcircumference, and endosteal circumference) were analyzedin the mid-diaphyseal region of femur and tibia.(37)

mCT

mCT analyses were performed on the proximal tibia byusing Skyscan 1072 scanner (Skyscan, Aartselaar, Belgium)and imaged with an X-ray tube voltage of 50 kV and cur-rent 200 FA, with a 1-mm aluminium filter.(9) The scanningangular rotation was 1801 and the angular increment was0.451. The voxel size was 4.36 Fm isotropically. Datasetswere reconstructed using a modified Feldkamp algorithmand segmented into binary images using adaptive localthresholding.(38,39) Trabecular bone distal of the proximalgrowth plate was selected for analyses within a conformingvolume of interest (cortical bone excluded) starting at adistance of 200 Fm from the growth plate and extending afurther longitudinal distance of 1.3 mm in the proximaldirection. Trabecular thickness and separation were cal-culated by the sphere-fitting local thickness method.(40)

Histomorphometric analyses of osteoclast number

Tibias were fixed in Burckhardt’s fixative (24 h), dehy-drated, and embedded in methylmethacrylate (Technovit9100 New; Heraeus Kulzer, Hanau, Germany). Four-mi-crometer longitudinal sections of the proximal tibia werestained for TRACP activity and counterstained withMayer’s hematoxylin (Histolab, Goteborg, Sweden). Im-ages were captured using a Nikon Eclipse 80i light micro-scope connected to a Sony DXC-S500 video camera usingthe Osteomeasure software (OsteoMetrics, Decatur, GA,USA). Number of osteoclasts per bone surface is reportedaccording to the recommended American Society for Boneand Mineral Research nomenclature.(41)

Measurement of serum hormone levels

Commercially available radioimmunoassay (RIA) kitswere used to assess serum concentrations of T (MP

Biomedicals) and E2 (Siemens Medical Solutions Diag-nostics) in 3-mo-old mice.

RESULTS

Development of a transgenic mouse model withosteoblast specific aromatase overexpression

We developed a TG mouse model (Coll-1a1-Arom),which overexpresses the human aromatase gene under thecontrol of a 2.3-kb fragment of the rat type I a I procollagenpromoter (Fig. 1A). RT-PCR measurements showed thatthe Coll-1a1-Arom mice overexpressed aromatase specif-ically in bone (Fig. 1B). At 3 mo of age, Coll-1a1-Arommice showed normal body weight (males: WT, 26.9 g; Coll-1a1-Arom, 25.8 g; females: WT, 20.2 g; Coll-1a1-Arom,20.0 g), crown-rump length (males: WT, 53.7 mm; Coll-1a1-Arom, 52.9 mm; females: WT, 51.4 mm; Coll-1a1-Arom, 51.0 mm), and body mass index (males: WT, 9.3kg/m2; Coll-1a1-Arom, 9.2 kg/m2; females: WT, 7.7 kg/m2;Coll-1a1-Arom, 7.7 kg/m2) compared with WT mice. Se-rum IGF-I levels were unchanged in Coll-1a1-Arom mice(females: WT, 358 ± 18 ng/ml; Coll-1a1-Arom, 310 ± 21 ng/ml;males: WT, 356 ± 17 ng/ml; Coll-1a1-Arom, 318 ± 19 ng/ml)compared with WT mice. Both the male and female Coll-1a1-Arom mice had normal fertility (data not shown) andnormal serum E2 levels (females: 14.0 ± 2.3 versus 10.3 ±1.0 pg/ml; males: not detectable in WT or Coll-1a1-Arommice) and T levels (females 0.01 ± 0.001 versus 0.01 ± 0.001ng/ml; males: 0.12 ± 0.03 versus 0.08 ± 0.02 ng/ml) com-pared with WT mice (n = 8–12 in all groups). In addition,the weights of some organs known to be sensitive to sys-temic E2 treatment (uterus weight, thymus weight and fatmass) were normal in the Coll-1a1-Arom mice, indicatingthat neither circulating E2 levels nor E2 levels in non-skeletal tissues were affected in the Coll-1a1-Arom mice.

FIG. 1. Development of transgenic mice with osteoblast-specificoverexpression of human aromatase. (A) Development of a TGmouse model (Coll-1a1-Arom) that overexpresses the human ar-omatase gene under the control of a 2.3-kb fragment of the rat typeI a I procollagen promoter. This promoter will direct the ar-omatase expression to osteoblasts in bone. (B) Bone-specific ex-pression of human aromatase mRNA in 3-mo-old Coll-1a1-Arommice. Values are means ± SE. ND, not detectable.

AROMATASE EXPRESSION LEADS TO INCREASED BONE MASS 1265

In addition, the weight of the seminal vesicles, which are asensitive androgen-responsive tissue, was unaffected in theColl-1a1-Arom mice, supporting the notion that the ele-vated aromatase expression in osteoblasts did not result inreduction of circulating serum T levels.

Increased aBMD in adult male Coll-1a1-Arom mice

DXA analyses before sexual maturation (5 wk) showedthat there was no difference in the aBMD (total body andlumbar spine; Fig. 2) between males and females and thatColl-1a1-Arom mice did not have altered aBMD. However,after sexual maturation (3 mo old), male but not femaleColl-1a1-Arom mice had clearly increased aBMD in bothtotal body and lumbar spine compared with WT mice (Fig. 2).

Increased trabecular BMD and cortical BMC in adultmale Coll-1a1-Arom mice

The trabecular and cortical bone compartments wereanalyzed separately using pQCT in adult Coll-1a1-Arom

mice and WT mice. These analyses showed that the above-described elevated aBMD in adult male Coll-1a1-Arommice compared with WT mice was a consequence of in-creased trabecular BMD (+31%), cortical BMC (+16%),cortical BMD (+5.6%), and cortical thickness (+15%;Table 1). The increased cortical thickness was a result ofreduced endosteal circumference, whereas the periostealcircumference was unaffected in the male Coll-1a1-Arommice compared with the WT mice (Table 1). The effect ofelevated osteoblast specific aromatase expression on cor-tical thickness in male mice was confirmed by mCT analysesof the diaphyseal region of tibia, showing increased corticalthickness (+18.5 ± 4.7%, p < 0.01) and reduced endostealcircumference (29.8 ± 3.4%, p < 0.01) in male Coll-1a1-Arom mice compared with the WT mice (n = 9 inboth groups). Similar as seen using DXA, no major skel-etal phenotype except a minor increase in cortical thick-ness (+4.3%, p < 0.05) was evident for the adult fe-male Coll-1a1-Arom mice compared with the WT mice(Table 1).

At 3 mo of age, a more pronounced sex difference wasseen in Coll-1a1-Arom than in WT mice for lumbar spineaBMD (WT, 5.4 ± 3.6%; Coll-1a1-Arom, 21.9 ± 3.6%males over females; p < 0.01, Coll-1a1-Arom versus WT;Fig. 2A), total body aBMD (WT, 8.6 ± 1.9%; Coll-1a1-Arom, 15.9 ± 2.2% males over females; p < 0.05, Coll-1a1-Arom versus WT; Fig. 2B), and cortical thickness (WT, 7.6 ±2.1%; Coll-1a1-Arom, 18.2 ± 2.3% males over females;p < 0.01, Coll-1a1-Arom versus WT; Table 1).

Elevated OPG levels and reduced numberof osteoclasts in male Coll-1a1-Arom mice

To explore the mechanism behind the increased bonemass in the male Coll-1a1-Arom mice, the mRNA levels ofseveral genes known to be involved in bone metabolismwere analyzed in bone. Interestingly, male Coll-1a1-Arommice showed clearly increased OPG mRNA levels (+82 ±11%, p < 0.001) and unchanged RANKL mRNA levels(+15 ± 10%, not significant) compared with WT mice.There were no significant differences in TRACP5b, ca-thepsin K, osteocalcin, ERa, and ERb mRNA levels be-tween the Coll-1a1-Arom and WT mice.

Histomorphometric analyses of the proximal metaphy-seal region of tibia showed that the elevated OPG mRNAlevels in the male Coll-1a1-Arom mice were associatedwith reduced number of osteoclasts per bone surface in themale Coll-1a1-Arom mice compared with WT mice (258.9 ±4.7%, p < 0.01, n = 7–8).

Effect of OVX in Coll-1a1-Arom mice

To determine the role of increased local aromatization inthe absence of high endogenous ovary-derived circulatingE2, 12-wk-old Coll-1a1-Arom and WT mice were OVX.There was no difference in trabecular BMD in female go-nadal intact (sham operated) Coll-1a1-Arom and WT mice(Table 1). Four weeks after OVX, both Coll-1a1-Aromand WT mice had lost trabecular bone mass to the sameextent (tibia trabecular BMD—Coll-1a1-Arom: sham, 180 ±10 mg/cm3; OVX, 142 ± 9 mg/cm3; p � 0.01; WT: sham,

FIG. 2. Increased aBMD in adult male Coll-1a1-Arom miceLumbar spine (A) and total body (B) aBMD analyzed by DXA in5-wk-old and 3-mo-old Coll-1a1-Arom mice and WT mice (n = 8–12). Values are means ± SE. **p < 0.01 vs. WT.

1266 SJOGREN ET AL.

159 ± 10 mg/cm3; OVX, 124 ± 6 mg/cm3; p � 0.01). Therewas a small increase in cortical thickness in gonadal intactColl-1a1-Arom compared with WT (Table 1). The corticalthickness was clearly reduced after OVX in Coll-1a1-Arom (tibia cortical thickness: sham, 218 ± 4 mm; OVX,198 ± 4 mm; p � 0.01) but not in WT (tibia cortical thick-ness: sham, 209 ± 3 mm; OVX, 207 ± 3 mm) mice, resultingin cortical thickness that was not significantly dif-ferent between female Coll-1a1-Arom and WT mice afterOVX.

Enhanced skeletal T response in femaleColl-1a1-Arom mice

The minor skeletal phenotype in the female Coll-1a1-Arom mice is probably because they have low serum levelsof T, which is the substrate for aromatase. To determinewhether the female Coll-1a1-Arom mice showed a morepronounced skeletal T response than WT mice, exogenousT was given to OVX Coll-1a1-Arom and WT mice. Serumlevels of E2 were not detectable in OVX mice before andafter T treatment. DXA analyses showed that T treatmentresulted in a clearly more pronounced increase in totalbody aBMD in the Coll-1a1-Arom mice than in theWT mice (p < 0.01; Fig. 3). Detailed bone compartment–specific analyses of the diaphyseal region of femur, usingpQCT (Table 2), and of tibia, using mCT (Fig. 4), showedthat T treatment increased the cortical bone mass as aconsequence of increased cortical thickness and corticalBMD in the OVX Coll-1a1-Arom mice but not in theOVX WT mice compared with placebo-treated mice (Ta-ble 2; Figs. 4A and 5). The increased cortical thickness wasa result of reduced cortical endosteal circumference,whereas the periosteal circumference was unaffected (Ta-ble 2; Fig. 5). T increased the trabecular BMD, as measuredby pQCT (Table 2; Fig. 5), and the trabecular BV/TV andtrabecular number, as analyzed by mCT (Figs. 4B, 4C, and5), in both OVX Coll-1a1-Arom and in OVX WT mice, butthe effect of T treatment was significantly more pro-nounced (p < 0.01) in the OVX Coll-1a1-Arom mice thanin the OVX WT mice. Interestingly, T increased the tra-becular thickness in OVX Coll-1a1-Arom mice, whereas itactually reduced the thickness in OVX WT mice comparedwith placebo treatment (Fig. 4D).

DISCUSSION

Human aromatase is tissue specifically regulated, and inthis study, we developed a TG mouse model with osteo-blast specific aromatase overexpression, resulting in stim-ulatory estrogenic effects in bone without systemic adverseeffects. Because elevated aromatase expression in osteo-blasts results in increased E2 and reduced T levels in os-teoblasts and, as a consequence, an increased ER to ARactivation ratio in bone, we propose that activation of ERsresults in a more pronounced increase in bone mass thanwhat is seen after activation of the AR.

Ubiquitous overexpression of aromatase in male miceresults in elevated serum E2 levels associated with infer-tility, rudimentary prostate/seminal vesicles, gynecomastia,and adrenal gland hyperplasia.(20–22) In this study, we de-veloped a mouse model (Coll-1a1-Arom mice) with ele-vated osteoblast-specific expression of human aromatasewith unaffected serum E2 levels using the rat type I a Iprocollagen promoter that has previously been shown to be

TABLE 1. Trabecular and Cortical Bone Parameters in Coll-1a1-Arom Mice

Female Male

WT Coll-1a1-Arom WT Coll-1a1-Arom

Trabecular BMD (mg/cm3) 159 ± 10 180 ± 9 302 ± 12 397 ± 19*

Cortical BMC (mg/mm) 0.87 ± 0.02 0.90 ± 0.02 1.08 ± 0.03 1.25 ± 0.04†

Cortical BMD (mg/cm3) 1136 ± 7 1131 ± 6 1138 ± 6 1202 ± 6*

Cortical area (mm2) 0.76 ± 0.01 0.80 ± 0.02 0.95 ± 0.02 1.04 ± 0.03‡

Cortical thickness (mm) 209 ± 3 218 ± 4‡ 225 ± 4 258 ± 5*

Cortical periosteal circumference (mm) 4.31 ± 0.04 4.34 ± 0.04 4.93 ± 0.06 4.83 ± 0.07

Cortical endosteal circumference (mm) 3.00 ± 0.04 2.97 ± 0.05 3.52 ± 0.06 3.21 ± 0.05†

Trabecular BMD was analyzed in the proximal metaphyseal region of tibia, whereas cortical bone parameters were analyzed in the mid-diaphyseal region

of tibia using pQCT in 3-mo-old Coll-1a1-Arom mice and WT mice (n = 8–12). Values are expressed as means ± SE.* p < 0.001, † p < 0.01, and ‡ p < 0.05, vs. WT mice.

FIG. 3. More pronounced effect of testosterone on total bodyaBMD in OVX Coll-1a1-Arom mice than in WT mice. Total bodyaBMD analyzed by DXA in 3-mo-old OVX Coll-1a1-Arom miceand WT mice that were either vehicle treated (OVX) or testos-terone treated (OVX + T; n = 10–11). Values are means ± SE. *p <0.05 and ***p < 0.001 vs. OVX. ++p < 0.01, effect of testosterone inColl-1a1-Arom mice vs. in WT mice.

AROMATASE EXPRESSION LEADS TO INCREASED BONE MASS 1267

exclusively active in osteoblasts and osteocytes.(33–35) Coll-1a1-Arom mice showed a normal fertility and normal se-rum E2 and T levels, showing that there was no significantleakage of osteoblast-derived E2 into the circulation of theColl-1a1-Arom mice.

Adult male Coll-1a1-Arom mice had increased trabec-ular and cortical BMD and increased cortical thicknesscompared with the WT mice. This is a proof of conceptmodel providing evidence that induction of aromataseexpression specifically in osteoblasts results in bone-spe-cific stimulatory estrogenic effects without systemic ad-verse effects. E2 treatment in aromatase-deficient men atthe age of sexual maturation has been associated with anincreased periosteal bone expansion,(16) and ERa-deficientmale mice exhibit decreased periosteal bone growth, as-sociated with decreased serum IGF-I levels during sexual

maturation.(42) These finding indicated that E2 mightpromote periosteal bone expansion in males during sexualmaturation and that this might be caused by altered en-docrine serum IGF-I levels. Neither serum IGF-I levelsnor periosteal circumference was affected in male Coll-1a1-Arom mice, suggesting that the high local E2 levels inthe Coll-1a1-Arom mice mainly results in an endostealcontraction. However, the increased cortical endostealcontraction observed in adult male Coll-1a1-Arom mice isprobably an effect of local very high pharmacologicalE2 levels coming from osteoblast-mediated aromatizationof T. In line with this, an increased cortical endostealcontraction is also observed in OVX Coll-1a1-Arom micetreated with T. Thus, because the currently used transgenicmouse model probably results in very high local E2 levels,one can not directly interpret these data as a physiologicalcomparison of the role of estrogens versus androgens inbone.

As expected, the female Coll-1a1-Arom mice, withlow levels of endogenous serum T, and thereby, littlesubstrate for aromatase, had no major skeletal phenotype.OVX resulted in a similar loss of trabecular bone in Coll-1a1-Arom and WT mice. However, the cortical thicknesswas clearly reduced after OVX in Coll-1a1-Arom but notin WT mice. The reason behind this could be that the fe-male gonadal intact Coll-1a1-Arom mice have increasedcortical thickness as a result of aromatization of low levelsof endogenous ovary-derived androgens,(23) and after OVX,the substrate for osteoblast-expressed aromatase is lost,resulting in a cortical bone loss in the female Coll-1a1-

TABLE 2. Effect of Testosterone on Trabecular and Cortical BoneParameters in OVX Mice

WT Coll-1a1-Arom

Trabecular BMD (% increase) 43 ± 8* 90 ± 10†‡

Cortical BMC (% increase) 3.6 ± 1.5 16.7 ± 1.8†x

Cortical BMD (% increase) 1.6 ± 0.6 4.8 ± 0.8†‡

Cortical area (% increase) 2.3 ± 1.2 11.5 ± 1.5†x

Cortical thickness (% increase) 3.3 ± 1.2 13.4 ± 0.9†x

Cortical periosteal

circumference (% increase)

20.70 ± 0.8 20.1 ± 10.9

Cortical endosteal

circumference (% increase)

21.9 ± 1.3 24.4 ± 1.2{

Trabecular BMD was analyzed in the distal metaphyseal region of femur,

whereas cortical bone parameters were analyzed in the mid-diaphyseal re-

gion of femur using pQCT in 3-mo-old OVX Coll-1a1-Arom mice and

OVX WT mice that were either vehicle treated or testosterone (T) treated

(n = 10–11). Values are given as percent increase of T vs. vehicle and ex-

pressed as means ± SE.* p < 0.01, † p < 0.001, and { p < 0.05 for T vs. vehicle.‡ p < 0.01 and x p < 0.001 for effect of T in Coll-1a1-Arom mice vs. in WT

mice.

FIG. 4. More pronounced effect of testosterone on cortical andtrabecular bone parameters in OVX Coll-1a1-Arom mice than inWT mice. Cortical (A) and trabecular (B–D) bone parameterswere analyzed by mCT analyses of proximal tibia in 3-mo-old Coll-1a1-Arom mice and WT mice that were either vehicle treated(OVX) or testosterone treated (OVX + T; n = 10–11). Values aremeans ± SE. BV, bone volume; TV, total volume. *p < 0.05, **p <0.01, and ***p < 0.001 vs. OVX. ++p < 0.01 and +++p < 0.001, effectof testosterone in Coll-1a1-Arom mice vs. in WT mice.

FIG. 5. Summary of the effects of T in OVX WT mice and inOVX mice overexpressing aromatase in osteoblasts, resulting in anincreased ER to AR activation in bone (Coll-1a1-Arom). For allcortical and trabecular bone parameters, the stimulatory effect ofT was more pronounced in the OVX Coll-1a1-Arom mice than inOVX WT mice. Importantly, for all cortical bone parameters andfor trabecular thickness, elevated aromatase levels in osteoblasts,resulting in elevated E2 levels specifically in osteoblasts andthereby increased ER activation in osteoblasts, were required forT to increase these bone parameters. Based on these data, wepropose that activation of ERs results in a more pronounced in-crease in bone mass than what is seen after a direct T-mediatedactivation of AR in OVX mice. [, increased vs. vehicle-treatedOVX; /, unchanged vs. vehicle-treated OVX; Y, decreased vs.vehicle-treated OVX; [[, significantly more pronounced effect ofT in Coll-1a1-Arom mice than in WT mice.

1268 SJOGREN ET AL.

Arom. For all cortical and trabecular bone parameters,the stimulatory effect of exogenous T was morepronounced in the OVX Coll-1a1-Arom mice than in theOVX WT mice. Importantly, for all cortical bone parameterand for trabecular thickness, elevated aromatase levels inosteoblasts, resulting in elevated E2 levels in osteoblastand thereby increased ER activation in osteoblasts, wererequired for T to increase these bone parameters. Similarly,we have previously shown that E2, activating the ERsbut not dihydrotestosterone, which is a nonaromatizableandrogen exclusively activating the AR, increases corticalthickness, trabecular thickness, and cortical density in or-chidectomized mice.(9) Based on these data, we proposethat activation of ERs results in a more pronouncedincrease in bone mass than what is seen after a direct Tactivation of the AR in both female and male gonadecto-mized mice.

Much of the current research on estrogens aims to sep-arate beneficial estrogenic effects from harmful side ef-fects. The main approach has been to develop selectiveestrogen receptor modulators (SERMs), and some of theclinically available SERMs show at least partial tissue se-lectivity in ER activity, including block of estrogen actionin breast while still maintaining the beneficial effects ofestrogen in other tissues such as bone. As an alternativeapproach, we hypothesize that it is possible to target tissue-specific local E2 synthesis, resulting in tissue-specific es-trogenic effects without systemic adverse effects. Thepresent finding that Coll-1a1-Arom mice have increasedbone mass without systemic adverse effects, together withthe fact that human aromatase expression in osteoblastsand reproductive tissues is dependent on differentiallyregulated promoters,(25,27–32) suggests that development ofosteoblast-specific inducers of aromatase expression mightprovide means to stimulate bone mass without systemicadverse effects.

Bone resorption is dependent on RANKL, which is es-sential for osteoclast formation, activity, and survival. Thecatabolic effects of RANKL are prevented by OPG, whichbinds RANKL and thereby prevents activation of its re-ceptor RANK. Thus, osteoclast activity is likely to dependon the relative balance of RANKL and OPG.(43) Impor-tantly, in this study, the OPG levels were significantlyincreased, whereas the number of osteoclasts per bonesurface was reduced in the male Coll-1a1-Arom mice. Thissuggests that the increased bone mass in these mice was aresult of reduced RANK activation and thereby reducedbone resorption. However, not only osteoclasts but alsoosteoblasts express estrogen receptors and is affected byE2 treatment when studied in vitro.(2,3) Therefore, it ispossible that the probably very high local E2 levels in thetransgenic mice, besides affecting bone resorption, alsomight regulate osteoblast number, survival, and/or forma-tion, resulting in the observed endosteal contraction andtrabecular thickening. Furthermore, although our data ofincreased OPG mRNA levels in male Coll-1a1-Arom micesuggest that OPG might be involved in the estrogen action,one should interpret this finding with caution because thiswas not confirmed in osteoblast cultures and only at themRNA level.

In conclusion, elevated aromatase expression specificallyin osteoblasts results in stimulatory estrogenic effects inbone without systemic adverse effects. Because osteoblast-specific aromatase expression results in an increased ER toAR activation ratio in osteoblasts, we propose that acti-vation of ERs results in a more pronounced increase inbone mass than what is seen after activation of the AR.Furthermore, because human aromatase expression in os-teoblasts and reproductive tissues is dependent on differ-entially regulated promoters, development of osteoblastspecific inducers of aromatase expression might identifysubstances with stimulatory effects on bone without sys-temic adverse effects.

ACKNOWLEDGMENTS

This study was supported by the Swedish ResearchCouncil, the Swedish Foundation for Strategic Research,the ALF/LUA research grant in Gothenburg, the Lund-berg Foundation, the Torsten and Ragnar Soderberg’sFoundation, the Novo Nordisk Foundation, MagnusBergvall Foundation, Ake Wiberg Foundation, Tore Nil-son Foundation, and The Swedish Society for MedicalResearch.

REFERENCES

1. Orwoll ES 2003 Men, bone and estrogen: Unresolved issues.Osteoporos Int 14:93–98.

2. Riggs BL, Khosla S, Melton LJ III, 2002 Sex steroids and theconstruction and conservation of the adult skeleton. EndocrRev 23:279–302.

3. Vanderschueren D, Vandenput L, Boonen S, Lindberg MK,Bouillon R, Ohlsson C 2004 Androgens and bone. Endocr Rev25:389–425.

4. Arts J, Kuiper GG, Janssen JM, Gustafsson JA, Lowik CW,Pols HA, van Leeuwen JP 1997 Differential expression ofestrogen receptors alpha and beta mRNA during differentia-tion of human osteoblast SV-HFO cells. Endocrinology 138:5067–5070.

5. Nilsson LO, Boman A, Savendahl L, Grigelioniene G, OhlssonC, Ritzen EM, Wroblewski J 1999 Demonstration of estrogenreceptor-beta immunoreactivity in human growth plate carti-lage. J Clin Endocrinol Metab 84:370–373.

6. Onoe Y, Miyaura C, Ohta H, Nozawa S, Suda T 1997 Ex-pression of estrogen receptor beta in rat bone. Endocrinology138:4509–4512.

7. Vidal O, Kindblom LG, Ohlsson C 1999 Expression and lo-calization of estrogen receptor-beta in murine and humanbone. J Bone Miner Res 14:923–929.

8. Lindberg MK, Weihua Z, Andersson N, Moverare S, Gao H,Vidal O, Erlandsson M, Windahl S, Andersson G, Lubahn DB,Carlsten H, Dahlman-Wright K, Gustafsson JA, Ohlsson C2002 Estrogen receptor specificity for the effects of estrogen inovariectomized mice. J Endocrinol 174:167–178.

9. Moverare S, Venken K, Eriksson AL, Andersson N, Skrtic S,Wergedal J, Mohan S, Salmon P, Bouillon R, Gustafsson JA,Vanderschueren D, Ohlsson C 2003 Differential effects onbone of estrogen receptor alpha and androgen receptor acti-vation in orchidectomized adult male mice. Proc Natl Acad SciUSA 100:13573–13578.

10. Sims NA, Clement-Lacroix P, Minet D, Fraslon-Vanhulle C,Gaillard-Kelly M, Resche-Rigon M, Baron R 2003 A func-tional androgen receptor is not sufficient to allow estradiol toprotect bone after gonadectomy in estradiol receptor-deficientmice. J Clin Invest 111:1319–1327.

AROMATASE EXPRESSION LEADS TO INCREASED BONE MASS 1269

11. Tivesten A, Moverare-Skrtic S, Chagin A, Venken K, SalmonP, Vanderschueren D, Savendahl L, Holmang A, Ohlsson C2004 Additive protective effects of estrogen and androgentreatment on trabecular bone in ovariectomized rats. J BoneMiner Res 19:1833–1839.

12. Bilezikian JP, Morishima A, Bell J, Grumbach MM 1998 In-creased bone mass as a result of estrogen therapy in a man witharomatase deficiency. N Engl J Med 339:599–603.

13. Morishima A, Grumbach MM, Simpson ER, Fisher C, Qin K1995 Aromatase deficiency in male and female siblings causedby a novel mutation and the physiological role of estrogens.J Clin Endocrinol Metab 80:3689–3698.

14. Oz OK, Hajibeigi A, Howard K, Cummins CL, van Abel M,Bindels RJ, Word RA, Kuro-o M, Pak CY, Zerwekh JE 2007Aromatase deficiency causes altered expression of moleculescritical for calcium reabsorption in the kidneys of femalemice*. J Bone Miner Res 22:1893–1902.

15. Oz OK, Zerwekh JE, Fisher C, Graves K, Nanu L, Millsaps R,Simpson ER 2000 Bone has a sexually dimorphic response toaromatase deficiency. J Bone Miner Res 15:507–514.

16. Bouillon R, Bex M, Vanderschueren D, Boonen S 2004 Es-trogens are essential for male pubertal periosteal bone ex-pansion. J Clin Endocrinol Metab 89:6025–6029.

17. Rochira V, Zirilli L, Madeo B, Aranda C, Caffagni G, Fabre B,Montangero VE, Roldan EJ, Maffei L, Carani C 2007 Skeletaleffects of long-term estrogen and testosterone replacementtreatment in a man with congenital aromatase deficiency:Evidences of a priming effect of estrogen for sex steroids ac-tion on bone. Bone 40:1662–1668.

18. Vanderschueren D, Venken K, Ophoff J, Bouillon R, BoonenS 2006 Clinical Review: Sex steroids and the periosteum–reconsidering the roles of androgens and estrogens in perios-teal expansion. J Clin Endocrinol Metab 91:378–382.

19. Peng Z, Li X, Makela S, Vaananen HK, Poutanen M 2004Skeletal changes in transgenic male mice expressing humancytochrome p450 aromatase. J Bone Miner Res 19:1320–1328.

20. Li X, Nokkala E, Yan W, Streng T, Saarinen N, Warri A,Huhtaniemi I, Santti R, Makela S, Poutanen M 2001 Alteredstructure and function of reproductive organs in transgenicmale mice overexpressing human aromatase. Endocrinology142:2435–2442.

21. Li X, Strauss L, Kaatrasalo A, Mayerhofer A, Huhtaniemi I,Santti R, Makela S, Poutanen M 2006 Transgenic mice ex-pressing p450 aromatase as a model for male infertility asso-ciated with chronic inflammation in the testis. Endocrinology147:1271–1277.

22. Li X, Warri A, Makela S, Ahonen T, Streng T, Santti R,Poutanen M 2002 Mammary gland development in transgenicmale mice expressing human P450 aromatase. Endocrinology143:4074–4083.

23. Perkins LM, Payne AH 1988 Quantification of P450scc,P450(17) alpha, and iron sulfur protein reductase in Leydigcells and adrenals of inbred strains of mice. Endocrinology123:2675–2682.

24. Labrie F 1991 Intracrinology. Mol Cell Endocrinol 78:C113–C118.

25. Bulun SE, Sebastian S, Takayama K, Suzuki T, Sasano H,Shozu M 2003 The human CYP19 (aromatase P450) gene:Update on physiologic roles and genomic organization ofpromoters. J Steroid Biochem Mol Biol 86:219–224.

26. Parakh TN, Hernandez JA, Grammer JC, Weck J, Hunzicker-Dunn M, Zeleznik AJ, Nilson JH 2006 Follicle-stimulatinghormone/cAMP regulation of aromatase gene expression re-quires beta-catenin. Proc Natl Acad Sci USA 103:12435–12440.

27. Sasano H, Harada N 1998 Intratumoral aromatase in humanbreast, endometrial, and ovarian malignancies. Endocr Rev19:593–607.

28. Shozu M, Simpson ER 1998 Aromatase expression of humanosteoblast-like cells. Mol Cell Endocrinol 139:117–129.

29. Takayanagi R, Goto K, Suzuki S, Tanaka S, Shimoda S,Nawata H 2002 Dehydroepiandrosterone (DHEA) as apossible source for estrogen formation in bone cells: Correla-tion between bone mineral density and serum DHEA-sulfateconcentration in postmenopausal women, and the presenceof aromatase to be enhanced by 1,25-dihydroxyvitaminD3 in human osteoblasts. Mech Ageing Dev 123:1107–1114.

30. Wang L, Wang YD, Wang WJ, Li DJ 2009 Differential regu-lation of dehydroepiandrosterone and estrogen on bone anduterus in ovariectomized mice. Osteoporos Int 20:79–92.

31. Watanabe M, Noda M, Nakajin S 2007 Aromatase expressionin a human osteoblastic cell line increases in response toprostaglandin E(2) in a dexamethasone-dependent fashion.Steroids 72:686–692.

32. Watanabe M, Simpson ER, Pathirage N, Nakajin S, Clyne CD2004 Aromatase expression in the human fetal osteoblastic cellline SV-HFO. J Mol Endocrinol 32:533–545.

33. Dacquin R, Starbuck M, Schinke T, Karsenty G 2002 Mousealpha1(I)-collagen promoter is the best known promoter todrive efficient Cre recombinase expression in osteoblast. DevDyn 224:245–251.

34. Kalajzic I, Kalajzic Z, Kaliterna M, Gronowicz G, Clark SH,Lichtler AC, Rowe D 2002 Use of type I collagen green flu-orescent protein transgenes to identify subpopulations of cellsat different stages of the osteoblast lineage. J Bone Miner Res17:15–25.

35. Kalajzic Z, Liu P, Kalajzic I, Du Z, Braut A, Mina M, CanalisE, Rowe DW 2002 Directing the expression of a green fluo-rescent protein transgene in differentiated osteoblasts: Com-parison between rat type I collagen and rat osteocalcinpromoters. Bone 31:654–660.

36. Windahl SH, Vidal O, Andersson G, Gustafsson JA, OhlssonC 1999 Increased cortical bone mineral content but unchangedtrabecular bone mineral density in female ERbeta(2/2) mice.J Clin Invest 104:895–901.

37. Vidal O, Lindberg MK, Hollberg K, Baylink DJ, AnderssonG, Lubahn DB, Mohan S, Gustafsson JA, Ohlsson C 2000Estrogen receptor specificity in the regulation of skeletalgrowth and maturation in male mice. Proc Natl Acad Sci USA97:5474–5479.

38. Hildebrand T, Ruegsegger P 1997 Quantification of BoneMicroarchitecture with the Structure Model Index. ComputMethods Biomech Biomed Engin 1:15–23.

39. Lindberg MK, Svensson J, Venken K, Chavoshi T, AnderssonN, Moverare Skrtic S, Isaksson O, Vanderschueren D, CarlstenH, Ohlsson C 2006 Liver-derived IGF-I is permissivefor ovariectomy-induced trabecular bone loss. Bone 38:85–92.

40. Waarsing JH, Day JS, Weinans H 2004 An improved seg-mentation method for in vivo microCT imaging. J Bone MinerRes 19:1640–1650.

41. Parfitt AM, Drezner MK, Glorieux FH, Kanis JA, MallucheH, Meunier PJ, Ott SM, Recker RR 1987 Bone histomor-phometry: Standardization of nomenclature, symbols, andunits. Report of the ASBMR Histomorphometry Nomencla-ture Committee. J Bone Miner Res 2:595–610.

42. Vidal O, Lindberg MK, Hollberg K, Baylink DJ, AnderssonG, Lubahn DB, Mohan S, Gustafsson JA, Ohlsson C 2000Estrogen receptor specificity in the regulation of skeletalgrowth and maturation in male mice. Proc Natl Acad Sci USA97:5474–5479.

43. Kearns AE, Khosla S, Kostenuik PJ 2008 Receptor activatorof nuclear factor kappaB ligand and osteoprotegerin regula-tion of bone remodeling in health and disease. Endocr Rev29:155–192.

Received in original form November 5, 2008; revised formDecember 19, 2008; accepted February 11, 2009.

1270 SJOGREN ET AL.