Select HIV protease inhibitors alter bone and fat metabolism ex vivo.

Human immunodeficiency virus (HIV) therapies have been associated with alterations in fat metabolism and bone mineral density. This study examined the effects of HIV protease inhibitors (PIs) on bone resorption, bone formation, and adipocyte differentiation using ex vivo cultured osteoclasts, osteoblasts, and adipocytes, respectively. Osteoclast activity, measured using a rat neonatal calvaria assay, increased in the presence of nelfinavir (NFV; 47.2%, p = 0.001), indinavir (34.6%, p = 0.001), saquinavir (24.3%, p = 0.001), or ritonavir (18%, p < 0.01). In contrast, lopinavir (LPV) and amprenavir did not increase osteoclast activity. In human mesenchymal stem cells (hMSCs), the PIs LPV and NFV decreased osteoblast alkaline phosphatase enzyme activity and gene expression significantly (p < 0.05). LPV and NFV diminished calcium deposition and osteoprotegrin expression (p < 0.05), whereas the other PIs investigated did not. Adipogenesis of hMSCs was strongly inhibited by saquinavir and NFV (>50%, p < 0.001) and moderately inhibited by ritonavir and LPV (>40%, p < 0.01). Expression of diacylglycerol transferase, a marker of adipocyte differentiation, decreased in hMSCs treated with NFV. Amprenavir and indinavir did not affect adipogenesis or lipolysis. These results suggest that bone and fat formation in hMSCs of bone marrow may be coordinately down-regulated by some but not all PIs.

Highly active antiretroviral therapy (HAART) 1 is a therapeutic approach for HIV infection that involves combined treatment with three classes of anti-HIV drugs: protease inhibitors (PIs), non-nucleoside reverse transcriptase inhibitors, and nucleoside reverse transcriptase inhibitors (NRTIs; Ref. 1). Over the last few years a number of unusual adverse events have been observed when HAART is used as a long term therapy (2). Some common complications associated with HAART are: adverse events related to the use of NRTIs (e.g. neuropathy, myopathy, pancreatitis, and lactic acidosis; Ref. 3), metabolic alterations or lipodystrophy (fat redistribution, insulin resistance, and dyslipidemia; Ref. 2), and bone disorders (osteonecrosis and osteoporosis; Refs. 4 -7). Osteonecrosis has been documented in case reports of HIV patients; however, some of these reports predate HAART, and there is no firm evidence that osteonecrosis is associated with HAART (8,9). Many factors may influence bone and fat metabolism and could lead to bone dysfunction in HIV patients including the presence of viral infection, therapeutic drugs (PIs, NRTIs, non-nucleoside reverse transcriptase inhibitors, or combination therapies), cellular response to the virus/drug, and immune function in the affected individual.
Decreased bone formative and increased bone resorptive serum markers have been observed in subjects receiving HAART (4,7). However, there are conflicting reports on the cause of the bone disorders observed in these patients. Carr et al. (6) have reported a connection between low pre-therapy body weight, asymptomatic lactic acidemia, and osteopenia in HAART patients. Regardless of specific HIV drug treatment, Huang et al. (7) reported a lower bone mineral density (BMD) associated with an increase in abdominal visceral fat. Interestingly a longitudinal study report (10) demonstrates that there may be an increase or no change in BMD after treatment with certain PIs. Contrary to the above reports, Tebas et al. (5) found no link between increased lipodystrophy and lower BMD in their crosssectional studies. Instead they suggest a link in the development of osteopenia and osteoporosis to the PIs received by HAART patients, although the role of current or previous NRTI use on bone mineral density was not addressed (11).
These studies reveal that the effect of HAART on alterations in BMD remains unclear. This is probably due to the complexity of HAART, which can involve a treatment choice of up to 16 drugs in various combinations. Furthermore the response to HAART can be influenced by pre-therapy body weight, progression of the viral infection, and its effect on bone metabolism (12). In addition, individual patients may have genetic traits or be exposed to environmental factors that influence their response to HIV and to HAART and alter their risk for the development of osteoporosis.
Osteoblasts (OBs) and osteoclasts (OCs) are derived from different cell lineages and play important roles in bone metabolism. OBs are derived from stromal cells or mesenchymal stem cells (hMSCs) within the bone marrow. OCs are derived from hematopoietic cells and are distantly related to monocytes and macrophages. OBs are involved in active bone formation while OCs are involved in bone degradation and resorption. The functions of OBs and OCs balance one another during normal bone metabolism and any alterations in the function or formation of either of these cell types may result in the development of osteopenia or osteoporosis. The effect of individual HIV-PIs on isolated OB and OC cells remains unknown.
PIs have proven to be a very effective drug class for the control of HIV infection by inhibiting the aspartyl HIV protease and interfering with formation of mature viral particles. It is important to note that these drugs vary structurally, and * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  mechanistic studies are essential to determine interclass variability in the development of adverse events (such as osteoporosis). This study uses hMSCs and rat calvaria to study how PIs may influence bone and fat metabolism ex vivo. All commercially available PIs were individually studied in these HIVfree systems. To elucidate and differentiate the individual effect of each PI on bone, specific measurements of osteoclast activity/bone resorption (measured in rodent calvaria), bone formation, adipocyte formation, and fat degradation in the presence of physiological PI concentrations were examined.

MATERIALS AND METHODS
Dexamethasone, sodium ␤-glycerophosphate, Me 2 SO, ascorbic acid 2-phosphate, isobutylmethylxanthine, alkaline phosphatase (ALP) diagnostic kit no. 86, and calcium diagnostic kit no. 587 were purchased from Sigma. The PIs used in these studies, amprenavir (APV), indinavir (IDV), lopinavir (LPV), nelfinavir (NFV), ritonavir (RTV), and saquinavir (SQV), were obtained from GlaxoSmithKline Inc. chemical stores. Since PIs are insoluble at Ͼ40 M in aqueous solutions, Me 2 SO was used to make concentrated stock solutions as reported previously (13). We have observed (13) 2 that less than 0.1% Me 2 SO has no effect on cell growth and differentiation and hence is a satisfactory vehicle and control for these experiments. Human mesenchymal stem cells were purchased from BioWhittaker Inc. (Walkersville, MD). The measurements for osteoblastic and adipogenic parameters were conducted on differentiated cells that were no longer proliferating. Cell culture and lipid accumulation assays were done following published procedures (14,15). In summary, hMSCs (passages 3-5) were plated at a density of 10 4 cells/cm 2 and cultured in Dulbecco's minimum essential medium containing 10% fetal bovine serum purchased from BioWhittaker Inc. Osteogenic differentiation was induced using 0.1 M dexamethasone, 0.05 mM ascorbic acid 2-phosphate, and 10 mM ␤-glycerophosphate (osteogenic stimulation (OS) medium) within 24 h of plating. Simultaneously various concentrations of PIs were added to the cells, which were almost Ͼ90% confluent. Measurements of ALP activity and histochemical staining in hMSCs (BioWhittaker, Inc.) were performed as described previously (14). After 7-21 days ALP activity and histochemical staining were measured using the Sigma Fast para-nitrophenyl phosphate substrate and the ALP leukocyte staining kit, respectively, according to the manufacturer's specifications (Sigma). For calcium accumulation measurements, Sigma diagnostic kit no. 587 was used following the manufacturer's instructions. Cells were incubated in 0.6 N HCl for 24 h, and an aliquot was diluted 20ϫ into the calcium working reagent and shaken for 3 min followed by an OD measurement at 575 nm. Alternatively adipogenesis was induced by treating confluent cells for 14 -21 days with 0.01 mg/ml insulin, 1 M BRL49653, 1 M dexamethasone, 0.5 mM isobutylmethylxanthine, and 1 M LG100268 for a period of 3 days followed by the removal of dexamethasone and isobutylmethylxanthine for a period of 2 days. After 14 -21 days in culture, cellular lipid content was measured using the Sigma diagnostic glycerol-triglyceride assay (i.e. Trinder reagent 337, Sigma).
Neonatal rat calvaria were surgically removed from pregnant Wistar female rats and placed in basal medium as described by Vargas et al. (16). After a 24 h stabilization period, the calvaria were incubated in basal medium containing 10 M PI for a period of 48 h. After the incubation, the medium was collected, and an aliquot was analyzed for calcium release using Sigma diagnostic kit no. 588 following the manufacturer's instructions. All assays were run in triplicate, and results from each group were compared with untreated calvaria and parathyroid hormone (PTH)-treated calvaria.
Total RNA was isolated using Qiagen RNeasy kits (Qiagen Inc., Valencia, CA) and quantified with Ribogreen (Molecular Probes, Eugene, OR). Taqman probes and primers were designed to match Gen-Bank TM sequences for human diacylglycerol transferase, alkaline phosphatase, and lipoprotein lipase. Real time polymerase chain reactions (RT-PCR) were performed as described by Lenhard et al. (17). All samples were assayed in duplicate with three samples per group. Results from each group were averaged and compared with untreated cells to provide a p value using the Student's t test.

RESULTS AND DISCUSSION
OC activity was measured using a rat neonatal calvaria assay that monitors calcium release as a measure of bone resorption. The effect of a 10 M concentration of the six PIs (APV, IDV, LPV, NFV, RTV, and SQV) on OC activity was examined. Recombinant PTH was used as a positive control. OC activity was calculated and expressed as the percentage of calcium released. Fig. 1A shows that OC activity increased in the presence of NFV (47.2%, p Ͻ 0.01), IDV (34.6%, p Ͻ 0.01), SQV (24.3%, p Ͻ 0.01), and RTV (18%, p Ͻ 0.01). APV and LPV did not alter OC activity significantly (p Ͼ 0.05), indicating that these two drugs do not alter bone resorption in ex vivo experiments.
hMSCs isolated from marrow aspirates have the potential to differentiate into several mesenchymal tissues, including bone, cartilage, adipose, tendon, muscle, and marrow stroma (Ref. 15; i.e. adipocytic, chondrocytic, and osteogenic lineages). After exposure to OS medium, hMSCs deposit a calcium-enriched matrix after 15 days. This deposit is readily measured using a sensitive colorimetric calcium assay (18). In the absence of OS, hMSCs do not deposit detectable amounts of calcium during cell culture. hMSCs were exposed to 10 -20 M PIs in OS medium, and mineralization of the extracellular matrix was measured (Fig. 1B). Calcium accumulation was inhibited in hMSCs treated with 10 M NFV (37.6%, p ϭ 0.016) and LPV (20.89%, p ϭ 0.057; see For each treatment three different RNA samples were isolated, and each one was analyzed in duplicate during the real time PCR. Results from each group were averaged and compared with untreated cells to provide a p value using the Student's t test. Significant differences (p Ͻ 0.05) between control (Me 2 SO) and treated groups are denoted by *. In all experiments, the solvent used for the PIs, 0.1% Me 2 SO, was included as a control. DMSO, Me 2 SO.

HAART Alters Bone Metabolism 19248
receptor for OPGL. OPG sequesters OPGL, thereby preventing binding to receptor-activated NF-B, OC activity, and bone resorption. Decreases in OPG expression by an osteoblast will lead to an increase in osteoclast activity due to the increased availability and hence binding of OPGL to the osteoclast. Gene expression of OPG was measured in hMSCs induced to differentiate into OBs for 16 days in the presence of 5-10 M PIs (Fig.  1C). Treatment with 5 M NFV decreased OPG expression by ϳ50% (p ϭ 0.04) compared with Me 2 SO-treated control cells. LPV (10 M) also decreased OPG expression significantly (33%, p ϭ 0.03), whereas the other PIs did not (p Ͼ 0.05). These results indicate that some PIs (NFV and LPV) alter expression of OPG and bone formation/resorption pathways, while other PIs do not alter these pathways ex vivo. One hypothesis is that the decreased osteogenesis and OPG expression could lead to increased osteoclastogenesis and bone resorption potentially explaining an underlying mechanism associated with NFV treatment. However, these data also indicate that not all PIs activate the OPG/OPGL receptor pathway in osteoblasts, and there may be other unknown mechanisms by which these PIs (e.g. RTV, IDV, and SQV) stimulate osteoclast activity in the rat calvaria assay.
Select PIs Inhibit Osteogenesis in hMSCs-To further assess the effects of HIV PIs on osteogenic differentiation, hMSCs were cultured in the presence of various PIs under conditions permissive for osteogenesis (15). Osteogenic differentiation is associated with increased ALP activity, calcium accumulation, increased expression of osteogenic genes, and morphological change (spindle shape becomes cuboidal; Refs. 14 and 18). Exposure to OS medium for 8 days results in a significant increase in ALP activity, and ALP activity continues to increase linearly for the next 8 days (14,18). ALP activity was measured on Day 14 after hMSCs were treated with 10 M HIV PIs in OS medium (control cells were treated with Me 2 SO). ALP activity was significantly inhibited (p Ͻ 0.01) in the presence of NFV (Ͼ63%), SQV (53%), LPV (48.8%), and RTV (29.5%, p Ͻ 0.05, Fig. 2A). Other PIs did not alter ALP activity significantly (APV and IDV by 20%, p Ͼ 0.05), indicating that PIs have pharmacologically distinct effects on osteogenic differentiation ex vivo. ALP gene expression was examined in differentiated hMSCs on Day 16 after exposure to PIs in OS medium. ALP gene transcription decreased significantly in the presence of LPV, NFV, and RTV (p Ͻ 0.03); a smaller decrease in ALP transcription was observed in IDV-and APV-treated cells (p Ͼ 0.05) (Fig. 2B). Upon microscopic examination, there was a 10-50% reduction in the total number of cells in the NFV-(and to a certain degree LPV-) but not the other PItreated cells in addition to the decrease in the total number of cells stained positive for ALP enzyme activity as visualized by light microscopy (Fig. 2C). Likewise NFV has demonstrated cellular toxicity under some in vitro conditions as reported by Dowell et al. (20).
Adipogenic Differentiation of hMSCs in the Presence of PIs-A number of in vitro studies have examined the effects of PIs on murine (13,20) and human adipocyte differentiation (21). These studies indicate that different PIs have different effects on adipocyte function and differentiation. In this study, we utilized the inherent properties of a select population of bone marrow cells, the hMSCs, to study the effect of PIs on adipocyte formation and degradation.
The effects of PIs on hMSC adipocyte differentiation were examined in the presence of 10 -20 M PIs. Adipogenic differentiation was assessed on Day 11 by measuring total lipid accumulation (Fig. 3A). Total lipid accumulation was significantly reduced in the presence of 20 M SQV (59.2%, p Ͻ 0.001) and NFV (51.6%, p Ͻ 0.001) and moderately by LPV (48.8%, p Ͻ 0.01) and RTV (44%, p Ͻ 0.01). We previously demonstrated that select PIs stimulate lipolysis in murine adipocytes (13). Lipolysis was also measured in fully differentiated hMSC adipocytes exposed to PIs for 18 -24 h (data not shown) at multiple doses. Free fatty acids were released in a dose-dependent manner in cells treated with Ͼ20 M NFV. No other PIs increased lipolysis, indicating that different PIs alter fat metabolism within the bone marrow through different mechanisms.
Adipocyte differentiation involves changes in expression of certain genes essential for lipid metabolism. Therefore, RT-PCR was used to measure expression of diacylglycerol acyltransferase (DGAT) and lipoprotein lipase (LPL) in hMSCs differentiating into adipocytes that were exposed to PIs. DGAT expression decreased in the presence of 5 M NFV (Fig. 3B), whereas the other PIs did not have a significant effect. NFV caused a similar decrease in expression of LPL (data not shown). This is consistent with our previous observation that NFV decreases LPL expression in differentiating murine adipocytes (13). Taken together, these observations suggest that the loss of fat in subjects receiving HAART may, in part, be due to decreased expression of genes involved in lipogenesis and increased lipolysis.
In summary, this study demonstrates that select PIs alter OC, OB, and adipocyte activity and differentiation ex vivo. The observation that LPV, NFV, and SQV inhibit both osteogenesis and adipogenesis in hMSCs suggests that bone and fat metabolism may be coordinately down-regulated in the bone marrow potentially contributing to altered bone mineral density. Direct effects of PIs on OCs may contribute also to the altered bone mineral density associated with HAART. Not all PIs exhibited the same effects ex vivo, indicating that different PIs have distinct effects on bone and fat metabolism. For example, LPV decreased OPG expression, inhibited calcium accumulation by   20) and a modest increase in BMD with IDV-containing therapy (n ϭ 34). In a prospective study on subjects receiving an APV-containing regimen (n ϭ 14) there was an increase in the total body bone mineral content by 0.04 Ϯ 0.01 kg (p ϭ 0.02) over 48 weeks (22). Hence, these data support the idea that clinical studies are needed that discriminate between the effects of individual PIs and discern whether adverse events should be grouped together as a class effect (11). As the studies reported here are limited to ex vivo conditions, further studies in vivo are needed to deduce the mechanism(s) of PI action in human subjects. While our results suggest a possible mechanism by which some PIs alter bone and fat metabolism, the data do not account for many factors including active metabolites, pharmacokinetic parameters, environmental factors, and genetic predisposition, which may influence the development of osteoporosis in the clinic. Additionally these ex vivo treatments do not account for the effect of serum protein binding and drug-drug interaction on the activities of the PIs within the patient. There are reports of NRTIs (for review, see Refs. 2 and 3) and combination therapy of PIs ϩ NRTIs also influencing fat metabolism and mitochondrial toxicity indicating additional mechanistic studies using the NRTIs are needed.
The use of anti-HIV drugs should be evaluated based on their therapeutic benefits and potential adverse effects. Each PI used in anti-HIV treatment needs to be assessed for its specific effects with respect to lipodystrophy, osteoporosis, hyperlipidemia, and other conditions. A clinician might consider a patient's treatment history, risk factors, and quality of life before determining the best therapy for that patient. These results indicate that certain PIs may have a minimal effect on osteoblast and osteoclast activity and could aid in the development of safer anti-HIV drugs. For each treatment three different RNA samples were isolated, and each one was analyzed in duplicate during the real time PCR. Results from each group were averaged and compared with untreated cells to provide a p value using the Student's t test. Significant differences (p Ͻ 0.05) between control (Me 2 SO) and treated groups are denoted by *. In all experiments, the solvent used for the PIs, 0.1% Me 2 SO, was included as a control. DMSO, Me 2 SO.