Atherosclerosis 161 (2002) 85–94

Induction of calcification in rabbit aortas by high cholesterol diets:roles of calcifiable vesicles in dystrophic calcification

Howard H.T. Hsu a,*, Nancy C. Camacho b, Ossama Tawfik a, Francis Sun c

a Department of Pathology and Laboratory Medicine, Uni�ersity of Kansas Medical Center, Kansas City, KS 66160-7410, USAb Research Di�ision, Hospital for Special Surgery, Cornell-weill Medical Center, New York, NY, USA

c Laboratory Animal Resources, Uni�ersity of Kansas Medical Center, Kansas City, KS 66160-7410, USA

Received 5 February 2001; accepted 18 June 2001

Abstract

Atherosclerotic calcification may weaken the aorta wall and thereby lead to rupture of the vessel. The mechanism wherebyaortas undergo calcification remains unclear. Previous reports in this laboratory showed that, after 2 months of cholesterol-sup-plemental feeding, an increase in calcifiability of membrane vesicles isolated from rabbit aortas precedes substantial arterialcalcification. Further, the mineral was deposited by isolated calcifiable vesicles as an amorphous phase similar to minerals inhuman aortas at an early stage of atherosclerosis. In the current study, atherosclerotic calcification was induced by exposingrabbits to a 1% cholesterol-rich diet for 3 or 6 months. After 3 months of dietary interventions, atherosclerotic lesions were fullydeveloped. Fatty streaks were evident in areas proximal to the heart and became less frequent in the distal areas. However,calcification was not yet identifiable histologically or by using Fourier transform spectroscopy (FT-IR). After 6 months of highcholesterol treatment, aortas were partially calcified. Histochemical staining for mineral revealed that calcification appeared tooccur predominantly in the intimal areas immediately adjacent to the media. Fourier Transform Imaging analysis demonstratedthat the mineral deposited in atherosclerotic rabbit aortas was a hydroxyapatite-like phase. To determine whether aorta vesiclesplay a role in mineral formation in aortas, vesicles were isolated from calcified aortas and then their calcifiability was comparedto that in normal vesicles. Interestingly, during the course of vesicle isolation, we found that calcifiable vesicles with much highercalcifiability than normal vesicles could be readily isolated from atherosclerotic aortas simply by suspending minced tissues inPBS. The characteristics of the calcification process and the enzymatic contents of isolated vesicles were similar to those obtainedusing collagenase digestion. Correlatively, mineral deposited by calcifiable vesicles isolated from the calcified aortas was also ofhydroxyapatite-like phases. Altogether, these observations indicate that (1) aortic calcification is a later event during atherogene-sis, (2) calcifiable vesicles are loosely bound to the matrices of the lesions as the result of the disease process and (3) similaritiesin the mineral phases between those in aortas and by vesicles during atherogenesis further support the role of calcifiable vesiclesin dystrophic calcification. © 2002 Elsevier Science Ireland Ltd. All rights reserved.

Keywords: Atherosclerosis; Calcification; ATP; Calcifiable vesicles

www.elsevier.com/locate/atherosclerosis

1. Introduction

Mineralization which occurs in atherosclerotic aortascan decrease elasticity and lead to rupture of the arte-rial wall [1–3]. It is also the main cause of the failure ofbioprosthetic cardiac valves prepared from chemicallypreserved animal tissues [4,5]. Because of these devas-tating effects of calcification, considerable effort has

been made to identify the underlying mechanisms in-volved in this process. Evidence to date indicates acomplex process whereby mineral is deposited, involv-ing the interplay of several promoting and regulatoryfactors in the atherosclerotic vessel wall (see Ref. [6] forreview). These factors include matrix vesicles and mito-chondria [7,8], bone morphogenetic protein-2a [9], os-teopontin [10], osteocalcin [11], Ca–phospholipid–phosphate complex [12], collagen I [13] and cholesterol[14,15]. Interestingly, these putative factors also exist inskeletal tissues, suggesting common initiation and regu-latory mechanisms for both types of calcification [6].

* Corresponding author. Tel.: +1-913-588-5395; fax: +1-913-588-7073.

E-mail address: hhsu@kumc.com (H.H.T. Hsu).

0021-9150/02/$ - see front matter © 2002 Elsevier Science Ireland Ltd. All rights reserved.PII: S 0 0 2 1 -9150 (01 )00623 -2

H.H.T. Hsu et al. / Atherosclerosis 161 (2002) 85–9486

In spite of these studies, the precise role of theabove-mentioned factors in the mechanisms of dys-trophic calcification remains unclear [6]. Ultrastructuralevidence indicates that mineral deposits are associatedwith mitochondria and vesicles which are present in theextracellular matrix of both atheromatous regions andbioprosthetic valves [4,5,7,8]. Experimental atheroscle-rosis suggests that exocytosis or necrosis in the lesions islikely responsible for the release of these subcellularorganelles from intimal, medial, endothelial cells, ormacrophages [16,17]. Recently, we developed an in vitrotechnique to demonstrate that the vesicles isolated fromatherosclerotic human and rabbit aortas have highercalcifiability than non-atherosclerotic vesicles [18–20].Rabbit data indicate that the appearance of calcifiablevesicles precedes formation of substantial arterial calcifi-cation during atherogenesis [19]. In addition, severalvesicle-associated Pi-yielding enzymes related to calcifi-cation (including ATPase, NTP-pyrophosphohydrolase,and AMPase) are also more active than those fromnormal vesicles. These observations are consistent witha causal role of these vesicles in the development ofabnormal aortic calcification suggesting that the accu-mulation of these vesicles may reach a critical thresholdthat would then initiate significant calcification.

A recent ultrastructural study using humanatherosclerotic aortas showed the presence of amor-phous mineral as the major mineral in areas close tocalcification zones [21]. This observation suggests thatmineral may initially be deposited as an amorphous formand later as an apatite phase. We also reported that theamorphous mineral was deposited by rabbit calcifiablevesicles isolated from aortas at earlier stages ofatherosclerosis [19]. In human subjects, hydroxyapatitehas been shown to be a final form of mineral in advancedatherosclerosis. Correlatively, we also found a hydroxya-patite phase was deposited by human aorta vesicles atthis stage [18]. The purpose of this report was todemonstrate for the first time using Fourier transformmicrospectroscopy that chronic cholesterol feeding torabbits induced hydroxyapatite formation in aortas andby isolated vesicles. Secondly, we hereby show thatcalcifiable vesicles are loosely bound to atheromatouslesions and can be released from the lesions by mincingtissues in phosphate-buffered saline without the aid ofbacterial collagenase. These observations further sup-port the role of calcifiable vesicles in dystrophic calcifica-tion

2. Experimental procedures

2.1. High cholesterol diets

Four-month-old white New Zealand male rabbitswere fed ad libitum with a standard laboratory diet

supplemented with 1% of cholesterol (supplied by Har-lan Teklad) for 3–6 months. Four rabbits were assignedto each group. For the control group, rabbits were fedwith the standard laboratory diet.

2.2. Isolation of calcifiable �esicles from aortas

Segments of ascending thoracic aortas (about 10 cm)were collected and immediately submerged in phos-phate-buffered saline (PBS) at 4 °C. After the attachedfat and residual blood were removed, the tissues wereminced into fine pieces, suspended in 5 ml PBS, andimmediately centrifuged at 800 g to precipitate cells andcell debris. The cell-free supernatants were centrifuged at30,000 g for 10 min to spin down mitochondria andmicrosomes. The resultant supernatant fractions werecentrifuged at 300,000 g for 20 min. The pellets wereresuspended in 10 mM Tris-buffered saline-0.25 Msucrose, pH 7.6 (TBSS) and centrifuged. The resultantprecipitates were then resuspended in 100 �l of TBSS.Vesicles were also isolated from normal aortas ascontrols.

2.3. Calcification

For calcification study, the method of Hsu et al. wasused [19]. Unless otherwise stated, the freshly preparedstandard calcifying medium (2 ml) consisted of 50 mMTris, pH 7.65 (pH was adjusted at 37 °C), 85 mM NaCl,15 mM KCl, 1 mM MgCl2, 30 mM NaHCO3, 1.45 mMCaCl2, 2.3 mM Pi, and �1 mM ATP. For blankcontrols, vesicles were omitted. The reaction was ini-tiated by addition of an aliquot of calcifiable vesicles tocalcifying media and then incubated for 5 h at 37 °C ina water vapor saturated incubation chamber. The min-eral deposits were then precipitated at 250,000 g for 20min. The deposits were washed twice in water (pH 7.8)by means of resuspension and centrifugation. Thewashed precipitates were then subjected to Ca depositionassay, scanning microscopy, and FT-IR analysis asdescribed below.

2.4. 45Ca deposition

45Ca2+ (1×106 cpm) was used as tracer. The reactionwas initiated by addition of aliquots of calcifiable vesi-cles (final protein concentration: about 15 �g protein/ml)to 100-�l calcifying media. The mixtures were incubatedfor 5 h at 37 °C in the incubation chamber saturatedwith water vapor, which was used to minimize vaporiza-tion of the reaction mixture during incubation. At theend of incubation, the reaction mixtures was filteredthrough 0.1 �m pore-size Durapore membranes (Mil-lipore Inc.) attached to a Millipore vacuum trap device.The membrane filters were washed twice each with 1 mlof Tris-buffered saline, pH 7.6 (TBS) and then trans-

H.H.T. Hsu et al. / Atherosclerosis 161 (2002) 85–94 87

ferred to vials containing scintillation fluids for radioac-tivity counting. The nonspecific 45Ca2+ binding isdefined as the radioactivity nonspecifically bound to thefilters under the identical conditions in the absence ofcalcifiable vesicles (0.6�0.2% of the total radioactiv-ity). These nonspecific counts were then subtractedfrom the radioactivity in the presence of calcifiablevesicles with or without ATP under various experimen-tal conditions. Ca deposition is expressed as ‘the totalnanomole of Ca deposited/5 h by vesicles from a 1-cmaorta.’ Transmission electron microscopy and the AT-Pase activity measurement indicate that no significantamounts of vesicles and mineral passed through filtersalthough the size-ranges of mineral and vesicles can besmaller than the pore-size of the filters (not shown).

2.5. Alizarin red staining for mineral

For histological study, a small segment of aortas wasfixed with 2% glutaraldehyde in PBS overnight at 4 °C,blocked, 3 �m thin sectioned, dehydrated in 95% etha-nol, and then stained with 2% alizarin red. The sectionswere washed briefly in water and followed by dehydra-tion in acetone–xylene.

2.6. Scanning electron microscopy

Calcifiable vesicles in 100 �l TBS were centrifugedand then collected on Nuclepore 0.1 �m-filters using avacuum trap. The filters were then fixed, dehydrated,and then processed by a scanning electron microscopicprocedure. For examination of mineral deposited byisolated vesicles, the following steps were taken. After a5-h exposure of vesicles to calcifying media with orwithout 1 mM ATP, calcified vesicles and mineraldeposits were collected in Beckman polyallomer tubesby centrifugation at 270,000 g for 30 min. The resultingpellets were washed twice with TBS by centrifugationand then resuspended in 50-�l water (pH 7.6). Thesuspensions were then spotted directly onto the sampleholders and air-dried. The dried residues were examinedand photographed in a Hitachi scanning electronmicroscope.

2.7. Fourier transform-infrared spectroscopy (FT-IR)for the assessment of calcification

Since the conventional measurement of calcificationwith 45Ca uptake may measure either Ca2+ uptake ormineral formation per se, FT-IR was used to identifythe solid phase of mineral formed [18]. A 20 �l-aliquotof vesicle preparations (�6–10 �g protein/ml calcify-ing media) was exposed to calcifying medium (2 ml) inthe presence and absence of 1 mM ATP for 5 h andthen the reaction mixtures were centrifuged at 250,000 gfor 30 min. The vesicle precipitates were washed twice

with distilled and deionized water at pH 7.6 and sus-pended in the presence of 10 mg KBr to assure totalrecovery of the sample. The precipitate/KBr mixtureswere freeze-lyophilized. The lyophilized powders wereplaced onto micro-sample holders pre-filled with KBrand analyzed on a Mattson Satellite FT-IR spectrome-ter (Mattson Instruments, Madison, WI) equipped witha diffuse reflectance accessory. Typically, 64 scans wereacquired at 4 cm−1 resolution under N2 purge. WIN-

FIRST software was used for data analysis. Protein, lipidand phosphate species (from the mineral phase) absorbat infrared frequencies characteristic of that particularmolecule and thus can be used to identify the compo-nents present in a sample. To calculate the relativeamount of mineral deposited, spectra of non-mineral-ized vesicles were subtracted from the mineralized spec-tra to isolate the mineral phase. Integrated areas ofindividual absorbance bands are proportional to thequantity of the component present in the sample. Thus,the ratio of the integrated areas of the absorbance inthe phosphate region from 900 to 1200 cm−1 (mineral)to that of absorbance in the protein amide I region(1585–1725 cm−1) were used to obtain a ‘mineral/ma-trix’ ratio. The relative specific and total vesicle activi-ties were then calculated and defined as follows: Therelative specific activity is expressed as ‘integrated areaof absorbance or ratios/�g vesicle proteins/5 h’ calcu-lated by the following formula: Integral or ratios/mlreaction mixture divided by the concentration of cal-cifiable vesicle proteins (�g/ml). The total activity isthen calculated from the specific activity multiplied by�g vesicle protein per mg aortas.

To estimate the amount of mineral in whole tissues,aortas were minced and then incubated with theproteinase K-SDS mixture (10 mM Tris, pH 7.8, 100�g/ml proteinase K and 0.5% SDS) at 37 °C overnight.The mixture was then centrifuged at 300,000 g toprecipitate mineral. The pellets were washed once withdeionized water. The resulting pellets were thoroughlymixed with a small amount of KBr and freeze-dried.The lyophilized pellets were layered carefully onto KBr-filled steel container for FT-IR analysis. The mineral tomatrix ratios were used to estimate the total relativeamount of mineral in whole aortas.

2.8. Fourier transform-infrared imaging and microscopyfor determination of mineral phases

Intact rabbit aortas were cryosectioned at 6 �mthickness and transferred to a BaF2 window for analy-sis. A Bio-Rad (Cambridge, MA) UMA 300A FT-IRmicroscope with an FTS-60A step-scanning FT-IRspectrometer and a 64×64 MCT FPA detector(Stingray Imaging Spectrometer) was used to acquiredata at 8 cm−1 resolution under N2 purge. This allowsinformation on mineral, protein and lipid distribution

H.H.T. Hsu et al. / Atherosclerosis 161 (2002) 85–9488

to be obtained from a 400×400 �m2 region at 64×64individual points of 5–7 �m diameter, resulting in 4096individual spectra. Imaging data were analyzed withWINIR PRO software (Bio-Rad Inc). The areas under themineral peak centered at 1030 cm−1 and the lipid peakcentered at 1730 cm−1 were calculated for each spec-trum, and infrared images created based on these inte-grated areas.

Infrared microscopy was utilized to identify the min-eral phase in intact histological sections of aorta. ABio-Rad (Cambridge, MA) FTS-40 infrared spectrome-ter coupled to a Bio-Rad UMA 500 microscopeequipped with a mercury–cadmium– telluride (MCT)detector was used to acquire data at 4 cm−1 resolutionunder N2 purge. Spectra of regions 20×20 �m2 diame-ter were acquired from the tissue.

2.9. Ca and Pi contents in aortas

To extract total Ca and Pi, 800 g precipitates (about100 mg) containing cells, debris, and mineral obtainedfrom collagenase digest were extracted using a mi-crofuge pestle with 1 ml 10% acetic acid over night at50 °C. To estimate ionic species of Ca and Pi, aortaswere homogenized with deionized water which replacedacetic acid. The extracts were then centrifuged in amicrofuge for 2 min and the supernatant fractions werethen subject to Ca and Pi determinations. Ca concen-trations were determined by an Arsenazo III clinical kit(Sigma Chemical, Inc.) inorganic orthophosphate wasdetermined by the method of Martin and Doty [22].

2.10. Pi-yielding enzymes assay

ATPases, AMPase, NTP-pyrophosphohydrolase, andalkaline phosphatase were assayed by the method ofHsu et al. [18].

2.11. Protein and phosphate assays

Inorganic orthophosphate and protein concentra-tions were determined by the procedures of Martin andDoty [22] and Lowry et al. [23], respectively.

3. Results

3.1. Effect of high lipid diets on arterial calcification

A previous report [19] in this laboratory demon-strated that feeding rabbits high cholesterol supplemen-tal diet for 2 months induced typical atheroscleroticregions with abundant macrophage-like cells and foamcells in the intima. The lesions generally occur in theareas proximal to the heart and become less prominentin the distal areas (not shown). However, at this stage

of atherosclerosis, the aortas did not display histochem-ically detectable calcification. To induce histochemicallyidentifiable arterial calcification, rabbits were subjectedto longer exposure of high cholesterol feeding. A 3-month intervention with 1% cholesterol yielded thesame degree of atherosclerosis without calcification as

Fig. 1. Histochemical alizarin red staining of mineral in rabbit aortas:(A) normal aorta, no mineral is present; (B) atherosclerotic aorta.The calcified areas in (B) are stained in red.

Fig. 2. FT-IR images showing the distributions of components fromthe thin section of a rabbit aorta. Mineral images are based on 1030band, matrix imaged on 1650 band, and lipid imaged on 1730 band(see Section 2 for details).

H.H.T. Hsu et al. / Atherosclerosis 161 (2002) 85–94 89

those obtained with 2-month exposure. Six monthsafter cholesterol interventions severe atheroma withfatty streaks spread over the entire aortas. Conspicuouscalcification appeared in the proximal areas and gradu-ally decreased in intensity in the distal parts. Mineral-specific staining with Alizarin red indicated that calcifi-cation occurred mostly in the intimas adjacent to themedia (Fig. 1). The presence of mineral was furtherdemonstrated by Fourier transform-infrared imaging(Fig. 2). Fourier transform-infrared spectromicroscopicanalysis (FT-IR) in situ revealed the presence of acarbonated apatite phase similar to that occurs in bone(Fig. 3). The relative amounts of mineral deposited inaortas were estimated by the spectral ratios of mineralto matrix per gram of aorta residues. Typical FT-IRspectra for extracts from whole aortas were used forestimating the relative quantity of mineral in affectedaortas (Fig. 4). The ratios of mineral to matrix were2.2�1.5 and 103�75, respectively, for seven controlsand five calcified aortas, respectively. The differences inthe ratios are statistically significant between these twogroups (P�0.001). Thus, these spectral ratio data areconsistent with the histochemical observations that aor-tas became calcified as a result of prolonged high lipidfeeding.

The effects of high lipid diets on ionized and non-ionized Ca and Pi contents in the aortas were alsoexamined. The levels of total and ionized Ca and Piwere estimated from acid and water extracts of aortas,respectively. Mineral contents were estimated by sub-tracting ionic Ca and Pi levels in water extracts from

Fig. 4. FT-IR spectroscopic estimation of relative amounts of mineralin calcified aortas. The isolation of mineral is described in Section 2.The spectroscopic pattern of FT-IR of isolated mineral is character-ized by its high matrix to mineral ratio as compared to that prior toisolation of the mineral. The calculated ratios of spectra can be usedto estimate relative amounts of mineral from whole aortas: (A)atherosclerotic aorta; (B) synthetic hydroxyapatite; and (C) normalaorta.

Fig. 3. Infrared spectromicroscopy of aorta and bone in situ. Theaorta specimen was fixed in 10% formalin and followed by soaking inpolyvinyl alcohol for 3 days and sliced into 5 �m-thick sections. Thesections were placed on a BAF2 window for analysis by Fouriertransform microscopy spectroscopy. These observations indicate thataortas were mineralized after 6 months of chronic cholesterol inter-ventions.

those in acid extraction. As shown in Fig. 5, both ionicand non-ionized Ca and Pi contents in atheroscleroticaortas were elevated (P�0.01). The higher levels ofnon-ionized Ca and Pi likely representing the depositedmineral are therefore consistent with FT-IR and histo-chemical data.

3.2. Isolation of calcifiable �esicles from calcified aortaswithout the use of collagenase digestion

Collagenase digestion was previously used to isolatecalcifiable vesicles from human and rabbit atheroscle-rotic aortas [18–20]. However, a report by Kristiansenet al. [24] indicated that bacterial collagenase treatmentof skeletal muscle cells in vitro can lead to the forma-tion of membrane vesicles. In addition, the extent ofvesicle release from aortas was found to vary withdifferent commercial collagenase preparations probablydue to variations in collagenase activity (not shown).

H.H.T. Hsu et al. / Atherosclerosis 161 (2002) 85–9490

To minimize the artificial production of calcifiable vesi-cles from aortas by exogenous collagenase that mayobscure the results, an attempt was made to reduce andstandardize the concentration of collagenase. As a re-sult, 45Ca deposition experiments revealed that signifi-cant amounts of particles with calcifying activity werereleased from minced aortas after mixing the aortafragments with phosphate buffered saline (PBS) with-out the necessity of adding bacterial collagenase (Fig.6). As shown by scanning electron microscopy (Fig.7A), the vesicles were identifiable although the prepara-tions were less homogenous than those from collage-nase digestion in an earlier report (Fig. 7B and Ref.[20]). The lipid-like substances in the preparations arelikely released from abundant fatty streaks. An attemptto remove fat-like substances by sucrose density cen-trifugation proved to be difficult probably due to thepossibility of similar densities to those of vesicles. Inspite of the encountered heterogeneity of the prepara-tions, the specific activities of various Pi-yielding en-zymes and calcifying activity as described below weresimilar to those previously reported values using colla-genase digestion [19], indicative of the presence ofactive calcifiable vesicles. FT-IR analysis of mineraldeposited by calcifiable vesicles indicated the presenceof a mineral phase similar to poorly crystal hydroxyap-atite (Figs. 3 and 8). Scanning electron microscopyshowed the presence of large globular mineral deposits

Fig. 6. Calcifying activity in vesicles isolated from control andexperimental aortas. The activity is expressed as the total 45Cadeposition by vesicles per cm aortas (nmol 45Ca/5 h/cm). Values areexpressed as means�SE from four vesicle preparations. Total vesicleproteins are also calculated (�g/cm). Pared t-test is used to calculatestatistical significance. The activity in experimental aortas is signifi-cantly higher than that from control aortas (P�0.05).

similar to those previously reported (Fig. 7C). Thereleased vesicles from atherosclerotic aortas displayedsubstantially higher protein yields and calcifiability thanthose from the control aortas as assessed by 45CaCl2uptake (Fig. 6) and by the mineral to matrix ratios ofFT-IR spectra (Fig. 8). After a shorter period of highlipid dieting (2 months), substantial calcification wasnot detected by alizarin. However, at this stage, calcifia-bility of isolated vesicle (24.2�0.57 nmol 45Ca/cmaorta, n=4) was substantially higher than the activityof normal vesicles but was lower than those from6-month cholesterol fed rabbits (P�0.05).

Several Pi-yielding enzymes including ATPase, NTP-pyrophosphohydrolase, AMPase, alkaline phosphatase,and various phosphatases that may have a role invesicle calcification were investigated. Similar to thoseenzyme activities in calcifiable vesicles obtained fromcollagenase digestion [19], ATPase, AMPase, and NTP-pyrophosphohydrolase in vesicle prepared without col-lagenase digestion were also very active and found to be731�319, 304�11, and 338�64 nmol/mg protein(n=4 vesicle preparations), respectively. �-Glyc-erophosphatase and pyrophosphatase activities at pH7.6 were not detected within the limits of the assays.The total activities (nmol Pi per cm aortas) was signifi-cantly elevated in calcified aortas as compared to con-trol values (P�0.01 see Fig. 9). Similar to vesiclesprepared by collagenase, alkaline phosphatase activity(less than 2–5 nmol/mg protein) in these vesicles wasalso much lower than the above-mentioned enzymaticactivities.

To estimate the relative amounts of microsomes andmitochondria in calcifiable vesicle fractions, variousrespective marker enzymes for subcellular organelleswere assayed. As described above, plasma membrane-associated enzymes such as ATPase, ATP pyrophos-phohydrolase, and AMPase were active whereasmicrosomal and mitochondrial enzymes includingNADPH-cytochrome C reductase and succinate dehy-

Fig. 5. Determinations of Ca and Pi content in normal and calcifiedaortas. The extraction and estimation of the ionic and non-ionic Caand Pi from aortas are described in detail in Section 2: (A) Cacontent; (B) Pi content.

H.H.T. Hsu et al. / Atherosclerosis 161 (2002) 85–94 91

drogenase, respectively, were not detected within theassay sensitivity (data not shown).

4. Discussion

In the current study, we have demonstrated by FT-IR spectroscopy and imaging the presence of a hydrox-yapatite-like phase in rabbit aortas as the result ofprolonging cholesterol supplemental feeding (Fig. 3).Before histochemically identifiable calcification couldoccur, atheromas were markedly developed after 3months of high lipid feeding. The lesions then becamepartially calcified after 6 months of feeding high lipiddiets, suggesting that calcification gradually developedduring atherogenesis. Interestingly, both calcified areasand fatty streaks were more prominent at proximalregions than the distal part. The reason for these differ-ences is unclear. Since the narrowing proximal aortasare prone to the force of blood flow because of curva-ture, they may be readily damaged as the result of theimpact [25]. Consequently, the injury could lead tocalcification, a frequent occurrence of local damage.Whether the injury could lead to the release of calcifia-ble vesicles from vascular cells thereby causing thevessel walls to calcify remains to be established. Sincecalcification was most prominent at the atheromatousregions close to media, it would be of interest todetermine whether calcifiable vesicles are concentratedin the same areas. To date no such information isavailable although this observation may also suggestthat interactions between media and atheroma maytrigger calcification.

Despite the presence of mineral-associated vesicles inatherosclerotic aortas [7,8], the precise role of vesicles inthe aortic calcification process remains to be defined. Itis not known whether the mineral association was dueto vesicle calcifying activity or a secondary effect of

aortic calcification through other unknown mecha-nisms. To support the role of vesicles in dystrophiccalcification, it is necessary to demonstrate that vesiclesare able to initiate calcification. Recently, we developedan in vitro technique to demonstrate that the mem-brane vesicles isolated from atherosclerotic human andrabbit aortas can initiate calcification and are moreactive than non-atherosclerotic vesicles [18–20]. Obser-vations from cholesterol-induced atherosclerosis in rab-bits demonstrated that the appearance of calcifiablevesicles precedes formation of substantial calcification[19]. The mineral deposited by isolated vesicles appearsto be amorphous at this stage of atherosclerosis. Inter-estingly, an amorphous type of mineral has also beenreported to be present in the areas surrounding athero-matous lesions in atherosclerotic human aortas prior toadvanced calcification [21]. In contrast to early stagesof atherosclerosis, hydroxyapatite was deposited in hu-man [26,27] and rabbit aortas (Fig. 3) and by isolatedvesicles from both humans [18] and rabbits at laterstages (Fig. 7). Altogether, the strong correlation in theappearance and types of mineral in aortas and de-posited by isolated vesicles during atherogenesisstrongly supports the role of vesicles in arterialcalcification.

Whether or not amorphous mineral is a precursorform of hydroxyapatite in normal calcified tissues in-cluding bone and cartilage has been the subject ofintensive investigation and has remained contentious(for a review see Ref. [28]). Electron microscopic exam-ination revealed that a non-atherosclerotic region ofhuman atherosclerotic aortas contains mineral-associ-ated vesicles within dendritic cells [21]. Most of theminerals in this region appear to be amorphous al-though a few hydroxyapatite crystals are present. Theability of isolated membrane vesicles to deposit amor-phous mineral at early stages of atherosclerosis [19]raises an interesting question of whether it is a precur-

Fig. 7. Scanning microscopy of calcifiable vesicles. (A) Vesicles prepared without the use of bacterial collagenase (magnification: ×45,000). Thepreparation appears to be contaminated with some lipid-like substances, which coat most of 0.1-�m Nuclepore filter pores rendering the pores lessapparent. (B) Vesicles prepared using the bacterial collagenase procedure (magnification: 30,000× ). (C) Mineral deposits by isolated vesicles(magnification: 1000× ). The mineral deposits appear to be globular and much larger than vesicles.

H.H.T. Hsu et al. / Atherosclerosis 161 (2002) 85–9492

Fig. 8. Fourier transform-infrared spectroscopic analysis of mineral deposited by isolated membrane vesicles from calcified aortas. Normal vesicleswere also included as control. The integral intensities from mineral (900–1207) and matrix (1488–1725) are shown. The integral ratios of mineralto matrix for atherosclerotic and control vesicles are 1.09 and 0.48, respectively. The pattern of the mineral phase deposited by atheroscleroticvesicles is similar to apatite.

sor form of hydroxyapatite, which is a predominantmineral at advanced atherosclerosis in human [26,27]and rabbit aortas in this study. Alternatively, calcifiablevesicles may initially deposit amorphous mineral at anearly stage of atherosclerosis and become capable ofdirectly depositing hydroxyapatite at the later stage.The latter possibility is more likely, since vesicles iso-lated from human [18] and rabbit aortas (Fig. 3) withsevere atherosclerosis have been shown to deposit amineral phase similar to hydroxyapatite. This may alsosuggest that different populations of calcifiable vesiclescapable of directly depositing the two types of mineralsmay coexist in various stages of atherosclerosis.

The ability of tissue mincing in releasing calcifiablevesicles from aortas suggests that vesicles are looselyembedded in the lesions. The mechanisms wherebyvesicles are present in lesions remain unclear. Experi-mental and human atherosclerosis [16,17] suggests thatexocytosis or necrosis that occurs in the lesions result-ing from hyperlipidemia can cause the release of cellu-lar fragments and apoptotic bodies. Recently, Kockx etal. [8] provided the evidence that these apoptotic bodiesor vesicles indeed are enriched in calcium. Themarkedly elevated levels of collagenase and metallo-proteinases expression during atherogenesis [29,30] mayalso likely underlie the mechanisms of vesicle formationsince it has been shown in vitro that exogenous collage-nase can lead to the formation of membrane vesicles incultured muscle cells [24].

Like human and rabbit atherosclerotic vesicles iso-lated by collagenase digestion [18–20], the mincing-re-leased rabbit vesicles had little or no alkalinephosphatase activity (less than 0.01 units per mg

protein within the detection limit of the assay measuredat both 25 and 37 °C). In contrast, alkaline phos-phatase in fetal bovine cartilage matrix vesicles is about25 units/mg vesicle protein [31]. The lack of strongalkaline phosphatase activity is surprising because ofthe prevailing consensus for the role of this enzyme inskeletal calcification [32]. Since the activity was mea-sured using the whole aorta, it would be difficult to ruleout the possibility that a small population of cells withhigh alkaline phosphatase activity may escape the de-tection but may be responsible for dystrophic calcifica-tion. To identify these alkaline phosphatase-positivecells, a future study using PCR and immunocytochem-istry will be required. However, such approach will bedependent on availability of specific antibody againstaortic alkaline phosphatase.

Fig. 9. Various Pi-yielding enzyme activities in calcifiable vesiclesisolated from control and experimental aortas. The activities areexpressed as the total activity per cm aortas (mU (nmol Pi released)/min)/cm. Values are expressed as mean�SE from four vesicle prepa-rations. Pared t-test is used to calculate statistical significance. Allenzyme activities in experimental aortas are significant higher thanthose from control aortas.

H.H.T. Hsu et al. / Atherosclerosis 161 (2002) 85–94 93

Together with these observations, the enrichment ofplasma membrane enzymes such as ecto-NTP py-rophosphohydrolase, AMPase, and possibly ecto-Mg-ATPase supports the contention that the isolatedcalcifiable vesicle fractions are of plasmalemmal origin.These observations also are similar to those with vesi-cles isolated from human and rabbit atheroscleroticaortas by collagenase digestion [18–20]. An attempt tofurther purify calcifiable vesicles to homogeneity in sizeor shape by sucrose density centrifugation proved to bedifficult most likely due to in situ heterogeneity in size,shape, density with varying mineral content [9,10], andthe presence of low-density fat-like substances (Fig. 7).

In summary, this report demonstrates that aortasundergoes calcification as a result of cholesterol dietaryinterventions and that production of calcifiable vesiclesprecedes substantial calcification. Further, the data sug-gest that calcifiable vesicles are loosely bound to thematrices in the course of atheromatous formation. Al-together these observations provide further insights intothe role of calcifiable vesicles in atheroscleroticcalcification.

Acknowledgements

The technical assistance of L. Spevak (FT-IRM anal-ysis), M. Seller (electron microscopy), L. Gorman andD. Crow (histology), and E. Roach and D. Friesen(photos) is greatly appreciated. This work was in partsupported by the Heartland-affiliated American HeartAssociation.

References

[1] Rokita E, Cichocki T, Heck D, Jarczyk L, Strzalkowski A.Calcification of aortic wall in cholesterol-fed rabbits. Atheroscle-rosis 1993;7:183–93.

[2] Agatston AS, Janowitz WR, Hildner FJ, Zusmer NR, ViamonteM, Detrano R. Quantitation of coronary artery calcium usingultrafast computed tomography. J Am Coll Cardiol1990;15:827–32.

[3] Farb A, Burke P, Tang AL, Liang Y, Mannan P, Smialek J,Virmani R. Coronary plaque erosion without rupture into a lipidcore: a frequent cause of coronary thrombosis in sudden coro-nary death. Circulation 1996;93:1354–63.

[4] Schoen FJ, Kujovich JL, Levy RJ, Sutton SJ. Bioprostheticvalve failure. Cardiovasc Clin 1988;18:289–317.

[5] Mako WJ, Vesely I. In vivo and in vitro models of calcificationin porcine aortic valve cusps. J Heart Valve Dis 1997;6:316–23.

[6] Demer LL, Watson KE, Bostrom K. Mechanism of calcificationin atherosclerosis. Trends Cardiovasc Med 1994;4:45–9.

[7] Tanimura A, McGregor DH, Anderson HC. Matrix vesicles inatherosclerotic calcification. Proc Soc Exp Biol Med1983;172:173–7.

[8] Kockx MM, Muhring J, Bortier H, de Meyer GRY, Jacob W.Biotin-digoxigenin-conjugated nucleotides bind to matrix vesiclesin atherosclerotic plaques. Am J Pathol 1996;148:1771–7.

[9] Bostrom K, Watson KE, Horn S, Wortham C, Herman IM,Demer LL. Bone morphogenetic expression in humanatherosclerotic lesions. J Clin Invest 1993;91:1800–9.

[10] Giachelli CM, Bae N, Almeida M, Denhardt DT, Alpers CE,Shwartz SM. Osteopontin is elevated during neointima forma-tion in rat arteries and is a novel component of humanatherosclerotic plaques. J Clin Invest 1993;92:1686–96.

[11] Ginsberg B, van Haarlem LJM, Soute BAM, Ebberink RHM,Vermeer C. Characterization of a gla-containing protein fromcalcified human atherosclerotic plaques. Atherosclerosis1990;10:991–5.

[12] Boskey AL, Bulloiugh PG, Vigorita V, DiCarlo E. Calcium–phospholipid–phosphate complexes in human hydroxyapatite-containing pathologic deposits. Am J Pathol 1988;133:22–9.

[13] Katsuda S, Okada Y, Minamoto T, Oda Y, Matsui Y, Nakan-ishi I. Collagens in human atherosclerosis. Arterioscler Thromb1992;12:494–502.

[14] Watson E, Bostrom K, Ravindranath R, Lam T, Norton B,Demer I. TGF-b and 25-hydroxycholesterol stimulate osteoblast-like vascular cells. J Clin Invest 1994;93:2106–13.

[15] Hirsch D, Azoiury R, Sarig S, Kruth HS. Colocalization ofcholesterol and hydroxyapatite in human atherosclerotic lesions.Calcif Tissue Int 1993;52:94–8.

[16] Robinson KA, Apkarian RP. Ultrastructure of coronary arterialendothelium in a atherosclerotic swine suggest lipid retro-endo-cytosis. Scanning Microsc 1991;5:533–9.

[17] Hegyi L, Skepper JN, Cary NR, Mitchinson MJ. Foam cellapoptosis and the development of the lipid core of humanatherosclerosis. J Pathol 1996;180:423–4.

[18] Hsu HHT, Camacho NP. Isolation of calcifiable vesicles fromhuman atherosclerotic aortas. Atherosclerosis 1999;143:353–62.

[19] Hsu HHT, Sun F, Tawfik O, Camacho N. Isolation of calcifiablevesicles from high cholesterol-fed rabbit aortas. Atherosclerosis2000;153:337–48.

[20] Hsu HHT, Tawfik O, Sun F. Effects of lectins on calcification byvesicles isolated from aortas of cholesterol-fed rabbits. BiochimBiophys Acta 2000;1464:262–72.

[21] Babryshev YV, Lord RSA. Detection of vascular dendritic cellsaccumulating calcified deposits in their cytoplasm. Tissue & Cell1998;30:383–8.

[22] Martin JB, Doty DM. Determination of inorganic phosphate.Anal Chem 1949;21:965–7.

[23] Lowry OH, Rosebrough NJ, Farr AL, Randall RJ. Proteinmeasurement with the Folin phenol reagent. J Biol Chem1951;193:265–75.

[24] Kristiansen S, Youn J, Richter EA. Effect of vanadate onglucose transporter (GLUT4) intrinsic activity in skeletal muscleplasma membrane giant vesicles. Biochim Biophys Acta1996;12182:71–5.

[25] Haja L, Evans AI, Chen M, Hyduk SJ, Collins T, Cybulsky MI.The NF-kB signal transduction pathway in aortic endothelialcells is primed for activation in regions predisposed toatherosclerotic lesion formation. Proc Natl Acad Sci USA2000;97:9052–7.

[26] Manoharan R, Baraga JJ, Rava RP, Dasari RR, Fitzmaurice M,Feld MS. Biochemical analysis and mapping of atherosclerotichuman artery using FT-IR microspectroscopy. Atherosclerosis1993;103:181–93.

[27] Rogers KD. Initial Rietveld characterization of biological calcifi-cations. Powder Diffraction 1997;12:175–9.

[28] Boskey AL. Amorphous calcium phosphate: the contention ofbone. J Dent Res 1997;76:1433–6.

[29] Mach F, Schonbeck U, Bonnefoy JY, Pober JS, Libby P.Activation of monocyte/macrophage functions related to acuteatheroma complication by ligation of CD 40: induction ofcollagenase, stromelysin, and tissue factor. Circulation1997;96:396–9.

H.H.T. Hsu et al. / Atherosclerosis 161 (2002) 85–9494

[30] Li Z, Li L, Zielke HR, Cheng L, Xiao R, Crow MT, Stetler-Stevenson WG, Froehlich J, Lakatta EG. Increased expressionof 72-kd type IV collagenase (MMP-2) in human aorticatherosclerotic lesions. Am J Pathol 1996;148:121–8.

[31] Hsu HHT, Munoz PA, Barr J, Oppliger I, Morris DC,Vaananen HK, Tarkenton N, Anderson HC. Purification and

partial characterization of alkaline phosphatase of matrix vesi-cles from bovine epiphyseal cartilage. Purification by monoclonalantibody affinity chromatography. J Biol Chem 1985;260:1826–31.

[32] Robison R. The possible significance of hexosephosphoric estersin ossification. Biochem J 1923;17:286–93.