A Potent HIV Protease Inhibitor, Darunavir, Does Not Inhibit ZMPSTE24 or Lead to an Accumulation of Farnesyl-prelamin A in Cells*

HIV protease inhibitors (HIV-PIs) are key components of highly active antiretroviral therapy, but they have been associated with adverse side effects, including partial lipodystrophy and metabolic syndrome. We recently demonstrated that a commonly used HIV-PI, lopinavir, inhibits ZMPSTE24, thereby blocking lamin A biogenesis and leading to an accumulation of prelamin A. ZMPSTE24 deficiency in humans causes an accumulation of prelamin A and leads to lipodystrophy and other disease phenotypes. Thus, an accumulation of prelamin A in the setting of HIV-PIs represents a plausible mechanism for some drug side effects. Here we show, with metabolic labeling studies, that lopinavir leads to the accumulation of the farnesylated form of prelamin A. We also tested whether a new and chemically distinct HIV-PI, darunavir, inhibits ZMPSTE24. We found that darunavir does not inhibit the biochemical activity of ZMPSTE24, nor does it lead to an accumulation of farnesyl-prelamin A in cells. This property of darunavir is potentially attractive. However, all HIV-PIs, including darunavir, are generally administered with ritonavir, an HIV-PI that is used to block the metabolism of other HIV-PIs. Ritonavir, like lopinavir, inhibits ZMPSTE24 and leads to an accumulation of prelamin A.

HIV protease inhibitors (HIV-PIs) 3 are designed to inhibit the HIV aspartyl protease, which is required for generating viral core proteins (1). HIV-PIs have become essential elements of modern antiretroviral regimens, but they have been associated with significant side effects, including partial lipodystrophy and metabolic syndrome (2)(3)(4). Similar disease phenotypes have been observed in association with missense mutations in LMNA (the gene for lamins A and C) (5,6) and with genetic defects associated with defective conversion of prelamin A to mature lamin A (7)(8)(9).
In 2003, Caron et al. (10) reported that a pair of HIV-PIs, indinavir and nelfinavir, appeared to lead to the accumulation of small amounts of prelamin A in a preadipocyte cell line. This finding was intriguing, but the biochemical mechanism was obscure. Potentially, this finding could have been due to the inhibition of any of four different enzymatic steps in prelamin A metabolism. The biogenesis of mature lamin A from prelamin A involves 1) farnesylation of a C-terminal cysteine by protein farnesyltransferase; 2) the removal of the last three amino acids of prelamin A (a redundant enzymatic activity of Ras converting enzyme 1 (RCE1) and ZMPSTE24); 3) the methylation of a newly exposed farnesylcysteine by isoprenylcysteine carboxyl methyltransferase (ICMT); and 4) the removal of the last 15 residues of the protein, including the farnesylcysteine methyl ester, by ZMPSTE24 (11). Steps 2-4 are utterly dependent on the first post-translational processing step, protein farnesylation.
Recently, our laboratories showed that several HIV-PIs, but notably lopinavir, lead to substantial prelamin A accumulation in cultured cells at therapeutically relevant concentrations, and we went on to identify the biochemical mechanism (12). Lopinavir blocked the enzymatic activity of ZMPSTE24, but it had no effect on protein farnesyltransferase.
In our initial studies (12), we noted that the electrophoretic migration of prelamin A differed in cells treated with lopinavir and cells treated with a farnesyltransferase inhibitor (FTI). We speculated that the difference was due to the fact that prelamin A was farnesylated in the setting of the HIV-PIs, but this issue was never settled. In the current study, one of our goals was to determine if the prelamin A that accumulates in the setting of HIV-PIs is farnesylated, because the farnesylation of prelamin A is highly relevant to its toxicity (11).
A second goal of the current studies was to determine if ZMPSTE24 inhibition is a universal property of HIV-PIs. Most of our early studies focused on lopinavir, a peptidomimetic drug and one of the most widely used and recommended HIV-PIs (13). In recent years, rational drug design strategies have yielded potent HIV-PIs that are structurally unrelated to lopinavir. One such compound, darunavir, exhibits very high affinity for the HIV protease and is potent in reducing viral loads (14). In the current study, we tested if darunavir inhibits ZMP-STE24 and leads to prelamin A accumulation in cells. We also examined another HIV-PI, ritonavir. Ritonavir is nearly always prescribed for HIV patients, not for its HIV-PI activity, but because it inhibits an enzyme that metabolizes protease inhibitors (15). Because it blocks the metabolism of HIV-PIs, ritonavir increases drug levels and improves drug efficacy (15).

EXPERIMENTAL PROCEDURES
Reagents-All antiretroviral drugs were obtained from the National Institutes of Health (NIH) AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, NIH. 4 The HIV-PIs were prepared as 20 mM stock solutions in Me 2 SO. An FTI, lonafarnib (Schering-Plough Kenilworth, NJ), was prepared as a 10 mM solution in Me 2 SO.
Assessing Protein Farnesylation-To assess protein farnesylation, cells were incubated for 36 h with an analogue of farnesol, 8-anilinogeraniol (AG, 30 M in Me 2 SO) (18). After entering cells, 8-anilinogeraniol is incorporated into anilinogeranyl diphosphate, which is used by protein farnesyltransferase as a substrate for the modification of CaaX proteins such as prelamin A. The 8-anilinogeraniol moiety of anilinogeranyl diphosphate is transferred onto cellular proteins in a protein farnesyltransferase-dependent manner, competitive with endogenous pools of farnesyl diphosphate. AG incorporation into endogenous cellular proteins can be detected readily by Western blotting with a mouse monoclonal antibody specific for AG (18,19), diluted 1:5000. AG is not incorporated into geranylgeranylated proteins by geranylgeranyltransferase I (18).
Cell Transfections-HeLa cells or immortalized mouse fibroblasts were grown to 80% confluency and transfected with prelamin A fusion constructs using the Fugene HD Transfection Reagent (Roche Applied Science) (12). Stably transfected HeLa cells expressing the GFP-prelamin A fusion were isolated after 10 days of selection in hygromycin (500 g/ml, Sigma). In some studies, stably transfected HeLa cells were grown for up to 16 days with HIV-PIs, and the cells were split 1:4 every 4 days.
Defining the Intracellular Localization of Prelamin A Fusion Proteins-Cells were plated on coverslips in 24-well plates and grown in the presence or absence of different HIV-PIs. After removing the cell culture medium, cells were washed with phosphate-buffered saline containing 1 mM CaCl 2 and 1 mM MgCl 2 . Cells were fixed with 4% paraformaldehyde in phosphate-buffered saline containing 1 mM CaCl 2 and 1 mM MgCl 2 for 10 min at room temperature and then incubated for 10 min with 2 g/ml 4Ј,6-diaminido-2-phenylindole, dihydrochloride (Invitrogen) to visualize nuclear DNA. Coverslips were mounted onto slides with the Prolong Gold Antifade reagent (Invitrogen). Images of cells were recorded sequentially on a Leica TCS-SP MP laser-scanning confocal microscope (Heidelberg, Germany) with a 100ϫ objective, and merged images were generated with Leica confocal software (version 2.5).

RESULTS
We showed previously that treatment of fibroblasts with lopinavir and atazanavir caused prelamin A to accumulate (12). In many experiments, prelamin A detected in extracts from HIV-PI-treated cells migrated more rapidly on SDS-polyacrylamide gels than prelamin A obtained after FTI treatment. The more rapid electrophoretic mobility of prelamin A in HIV-PItreated cells might be explained by a difference in protein farnesylation. To explore this issue, we metabolically labeled lopinavir-treated human fibroblasts with AG, an analogue of farnesol (18). This compound is taken up by cells, phosphorylated, and then used as a substrate by protein farnesyltransferase for modifying proteins (of note, AG is not a substrate for geranylgeranyltransferase I). The farnesyltransferase-dependent modification of proteins by AG can be detected by Western blotting with an AG-specific monoclonal antibody (18). We found that the prelamin A that accumulates in lopinavir-and atazanavir-treated cells was farnesylated, as judged by the incorporation of AG (Fig. 1A). This prelamin A protein was readily detectable with antibodies against lamin A/C or prelamin A (Fig. 1A). The intensity of the prelamin A band in lopinavir-treated cells was 28.5% of that of lamin C, as judged by infrared image quantification. Of note, darunavir did not lead to an accumulation of prelamin A (Fig. 1A).
In a parallel experiment, AG was incorporated into the prelamin A of ZMPSTE24-deficient fibroblasts isolated from a patient with restrictive dermopathy, and also into lopinavirtreated wild-type human fibroblasts (Fig. 1B). As expected, the incorporation of AG into the prelamin A of restrictive dermopathy fibroblasts could be eliminated with a highly specific FTI (Fig. 1B). The FTI, but not lopinavir, inhibited the farnesylation of HDJ-2, a farnesylated protein widely used as a biomarker for protein farnesylation (24).
The studies with human fibroblasts indicated that darunavir does not cause prelamin A accumulation (Fig. 1A). To explore this observation, we treated mouse wild-type fibroblasts and fibro-  blasts with half-normal levels of ZMPSTE24 (i.e. Zmpste24 ϩ/Ϫ fibroblasts) with darunavir as well as several other HIV-PIs (Fig. 2). Prelamin A accumulation was observed in wild-type fibroblasts treated either with lopinavir or the FTI, but not in cells treated with darunavir (Fig. 2). Small amounts of prelamin A accumulation were observed in atazanavir-treated wild-type fibroblasts. In Zmpste24 ϩ/Ϫ fibroblasts, treatment with lopinavir or atazanavir (20 M for 48 h) led to a significant accumulation of prelamin A (the prelamin A bands were 27 and 10.9%, respectively, compared with lamin C) (Fig. 2). Prelamin A was detectable in the darunavirtreated cells, but not above the levels observed in cells treated with vehicle (Me 2 SO) alone (Fig. 2).
Additional studies revealed that lopinavir and atazanavir, but not darunavir, interfered with the processing of a GFP-prelamin A fusion protein that contained the last 117 amino acids of prelamin A (Fig. 3A). In the absence of HIV-PIs, most of the fusion protein was fully processed, as judged by Western blots with antibodies specific for prelamin A or GFP. Lopinavir and atazanavir (20 M for 6 days) partially blocked the processing of the fusion protein, resulting in an accumulation of an unprocessed protein (the band intensities of the unprocessed protein were 32.4 and 20.6%, respectively, of the fully processed protein). Darunavir (20 M for 6 days) had no effect on the processing of the fusion protein (the amount of the unprocessed protein was 4.3% of the fully processed protein, compared with 7.2% for the Me 2 SO control) (Fig. 3B). Increasing the concentration of darunavir to 80 M had no effect on the processing of the GFP-prelamin A fusion protein (not shown).
Processing of the GFP-prelamin A fusion protein by ZMP-STE24 would cause the loss of its hydrophobic C terminus (i.e. the segment containing the farnesylcysteine methyl ester). We suspected that this proteolytic processing step would alter the intracellular localization of the fusion protein. Indeed, this was the case. In wild-type mouse fibroblasts, the GFP-prelamin A fusion protein was uniformly distributed within the cell (Fig.  4A). In Zmpste24 Ϫ/Ϫ fibroblasts, the fusion protein was excluded from the nucleoplasm (Fig. 4A). We hypothesized that a mutation designed to block ZMPSTE24 cleavage would also prevent the fusion protein from entering the nucleus. To test this hypothesis, we generated a DsRed-prelamin A fusion protein and changed the leucine at residue 647 to arginine. This missense mutation prevents the ZMPSTE24 cleavage reaction (25) and completely blocked the processing of the mutant (L647R) fusion protein (Fig. 4B). As expected, the mutant DsRed-prelamin A fusion protein was not present in the nucleoplasm of transfected HeLa cells (Fig. 4B). In contrast, a GFPprelamin A fusion without the missense mutation was uniformly distributed in the cell (Fig. 4B); the same uniform distribution was observed for a DsRed-prelamin A fusion protein without the missense mutation (not shown). When HeLa cells expressing the GFP-prelamin A fusion protein were treated with lopinavir (20 M), the fusion protein was excluded from the nucleoplasm (Fig. 4C), similar to when the GFP-prelamin A fusion protein was expressed in Zmpste24 Ϫ/Ϫ cells (Fig.  4A). When the transfected HeLa cells were treated with darunavir (20 M), there was no change in the intracellular localization of the fusion protein (Fig. 4C).
ZMPSTE24 enzymatic activity can be measured with an endoprotease-coupled methylation assay using a farnesylated a-factor substrate (21). In this assay, ZMPSTE24 cleaves the C terminus of a-factor, rendering it susceptible to methylation by Ste14p in the presence of S-adenosyl-L-[methyl-14 C]methionine. Methylation of the substrate can then be quantified by scintillation counting. Lopinavir strongly inhibited ZMP-STE24, whereas atazanavir inhibited ZMPSTE24 modestly (Fig.  5A). Neither darunavir nor its parent molecule amprenavir inhibited ZMPSTE24 (Fig. 5A). None of the drugs affected the enzymatic activities of RCE1 or ICMT (Fig. 5, B and C).
The fact that darunavir does not block ZMPSTE24 is a potentially attractive property. However, darunavir and most other HIV-PIs are generally administered with ritonavir, an HIV-PI that is given not for its antiretroviral effects but because of its ability to block the metabolism of HIV-PIs (15). Of note, ritonavir inhibits the enzymatic activity of ZMPSTE24 (Fig. 5A) and leads to a modest accumulation of prelamin A in human fibroblasts (Fig. 6A). Adding ritonavir (20 M) to HeLa cells expressing the GFP-prelamin A fusion protein inhibited the processing of the fusion protein (Fig. 6B). Also, in the presence of ritonavir, the GFP-prelamin A fusion was excluded from the nucleoplasm (Fig. 4C).

DISCUSSION
We previously showed that lopinavir treatment of mouse and human fibroblasts inhibits ZMPSTE24 and leads to an accumulation of a prelamin A species that has more rapid electrophoretic mobility than the prelamin A that accumulates in FTItreated cells. We speculated that the difference in electrophoretic mobility might be due to the fact that the prel- amin A in HIV-PI-treated cells was farnesylated. In the current study, we show, using metabolic labeling experiments with a chemical analogue of farnesol, that the prelamin A accumulating in HIV-PI-treated cells is indeed farnesylated. We also show that the prelamin A in ZMPSTE24-deficient human fibroblasts is farnesylated. Documenting the farnesylation status of the prelamin A in HIV-PI-treated cells is important, simply because substantial evidence indicates that farnesyl-prelamin A is toxic to cells (11,26,27). The accumulation of farnesyl-prelamin A in restrictive dermopathy is known to cause misshapen cell nuclei and severe progeroid syndromes in humans (9,28). Moreover, treating ZMPSTE24-deficient cells with an FTI reduces the frequency of misshapen nuclei (29), and treating Zmpste24 Ϫ/Ϫ mice with an FTI significantly reduces disease phenotypes (30).
We also found that a new and potent HIV-PI, darunavir, does not inhibit ZMPSTE24 activity, unlike lopinavir. Our data were very consistent: darunavir did not block the biochemical activity of ZMPSTE24; it did not lead to an accumulation of prelamin A in mouse and human fibroblasts; and it did not interfere with the processing of the wild-type GFP-prelamin A fusion protein. In our cell culture experiments, we generally used HIV-PI concentrations of 20 M because that level is near the upper limits of therapeutic levels in lopinavir-treated patients (31). Typical therapeutic levels for darunavir ranged 0. 2-18.6 M (32), similar to the 20 M concentration used in the current study. Even at 80 M concentrations, darunavir did not cause prelamin A accumulation in cells. In view of these considerations, it is unlikely that therapeutic levels of darunavir would result in prelamin A accumulation in human patients. In our studies, atazanavir was less potent than lopinavir in inhibiting ZMPSTE24 or causing an accumulation of prelamin A.
Although both HIV protease and ZMPSTE24 are clearly inhibited by lopinavir, ritonavir, and atazanavir, the molecular basis for this inhibition remains to be determined. The two enzymes display no sequence homology and have very different structures (the HIV protease is a soluble dimeric aspartyl protease, and ZMPSTE24 is an integral membrane zinc metalloproteinase). However, the overall structures of the transition states of the two proteases contain similarities that could explain how a single inhibitor could affect both enzymes (33). A detailed molecular understanding of this issue, however, will likely require a crystal structure of ZMPSTE24, with and without HIV-PIs. We hope that this information will ultimately become available, because the structure of another membrane zinc metalloproteinase, the site-2 protease, was recently solved (34).
With regard to the different HIV-PIs, it is noteworthy that they vary quite substantially in terms of their structures. Early

. Blocking the ZMPSTE24-mediated cleavage of the GFP-prelamin A fusion protein alters its localization within cells.
A, confocal fluorescence microscopy showing differences in the intracellular localization of the GFP-prelamin A fusion protein in transiently transfected wild-type and Zmpste24 Ϫ/Ϫ mouse fibroblasts. The fusion protein is excluded from the nucleus in Zmpste24 Ϫ/Ϫ cells. B, localization of the GFP-prelamin A fusion protein and a noncleavable DsRed-prelamin A fusion protein (containing the L647R substitution in prelamin A) in HeLa cells that were transfected with both constructs. The noncleavable fusion protein was excluded from the nucleoplasm. The Western blot study (lower left panel) shows that lopinavir (LPV) partially inhibits the processing of a DsRed-prelamin A fusion protein (WT). As expected, the noncleavable DsRed-prelamin A fusion protein (L647R) was not processed in cells and LPV had no effect on its electrophoretic migration. HIV-PIs were designed to mimic the binding of substrate to the catalytic site of the HIV protease. Thus, most of the early inhibitors were peptidomimetic compounds based on the scissile bond Phe-Pro; lead compounds were then further modified to improve bioavailability and other pharmacologic properties (35). Both lopinavir and ritonavir were derived from the same first-generation HIV-PI, saquinavir, a transition-state analogue. Two heterocyclic substitutions in ritonavir increased potency and bioavailability, and the addition of a cyclic urea moiety at position P2 in lopinavir (see structure in Table 1) improved potency against the HIV protease (35). Atazanavir was also derived from saquinavir, with further substitutions designed to improve efficacy against mutant HIV proteases. Darunavir was derived from a different first-generation HIV-PI, amprenavir (14). Amprenavir is also a transition-state analogue but with a sulfonamide group at P2Ј and a tetrahydrofuran group at P2. In darunavir, the addition of a bicyclic tetrahydrofuran group at P2 (see structure in Table 1) dramatically increases its interactions with the active site of the HIV protease (14). The binding affinity of darunavir for the HIV protease is far higher than that of lopinavir or other HIV-PIs (36). We suspect that the chemical moieties at positions P2 and P2Ј of darunavir, which were designed to maximize interactions with the active site of the HIV protease, could diminish its ability to bind to and inhibit ZMPSTE24 (Table 1).
Our current studies show that lopinavir, but not darunavir, causes prelamin A accumulation. This study was purely biochemical, and we did not attempt to investigate the pathological significance of farnesyl-prelamin A accumulation in HIV-PI-treated cells or HIV-PI-treated experimental animals. However, we would contend that it is at least plausible that an accumulation of farnesyl-prelamin A could cause adverse side effects. First, the farnesylated form of prelamin A is targeted to the nuclear rim, where it interferes with the structural integrity of the nuclear lamina and causes misshapen nuclei (26,27,29). Nonfarnesylated prelamin A is found in the nucleoplasm (29,37) and appears not to interfere with the nuclear lamina. Second, restrictive dermopathy, which is caused by a complete loss of ZMPSTE24, causes farnesyl-prelamin A accumulation and severe disease phenotypes. Finally, Caron and colleagues (38) found that the inhibition of ZMPSTE24 by HIV-PIs is more than a cell culture curiosity. They documented increased amounts of prelamin A in the adipose tissue of HIV patients treated with indinavir or nelfinavir. Those early HIV-PIs are no longer widely used, but we have documented that nelfinavir inhibits the activity of ZMPSTE24 and leads to an accumulation of prelamin A in fibroblasts (not shown).
In our studies, lopinavir and ritonavir but not darunavir clearly changed the localization of a GFP-prelamin A fusion protein in HeLa cells, preventing the fusion protein from entering the nucleus. An identical mislocalization of the GFP-prelamin A fusion protein was produced by interfering with the final prelamin A cleavage reaction, either by inactivating Zmp-ste24 or by introducing a missense mutation that renders the fusion protein resistant to cleavage by ZMPSTE24. The finding that the noncleavable, farnesylated prelamin A fusion was excluded from the nucleus was actually not surprising. More than 10 years ago, long before ZMPSTE24 or its role in prel-  , and amprenavir (APV) on the enzymatic activity of mouse ZMPSTE24. Shown are the results from a coupled endoproteolysis/ methylation assay that tested the ability of membranes from ⌬ste24⌬rce1 yeast overexpressing mouse ZMPSTE24 to cleave a yeast a-factor substrate, rendering it susceptible to methylation by yeast Ste14p. The results are presented as percentages of the specific activity measured in the presence of the vehicle (Me 2 SO (DMSO)). Each assay was repeated 5-7 times, except for APV (3 repeats) and LPV (2 repeats), with each point from each assay containing duplicate or triplicate measurements. B and C, all the drugs have a mild or no effect on the activities of RCE1 (B) and ICMT (C) at a concentration of 100 M. amin A processing had been identified, Hennekes and Nigg (39) constructed a noncleavable prelamin A fusion protein and found that it could not enter the nucleus. A similar phenomenon occurs with the farnesylated protein N-Ras. Choy and colleagues (40) showed that a "wild-type" GFP-N-Ras is excluded from the nucleus, whereas the introduction of a mutation preventing protein farnesylation resulted in a uniform distribution of the fusion protein within the cell. The finding that a GFPprelamin A fusion is excluded from the nucleoplasm should be interesting to the pharmaceutical industry, because the prelamin A fusion, in combination with fluorescence microscopy, provides a simple methodology for identifying HIV-PIs that block ZMPSTE24 and those that do not.
The lipodystrophy syndrome in HIV patients taking highly active antiretroviral therapy likely involves multiple drug components (2) and multiple biochemical mechanisms (41). Additional studies will undoubtedly be required to ascertain the pathological relevance of farnesyl-prelamin A in the lipodystrophy syndrome in humans. Atazanavir has been reported to be associated with fewer lipodystrophy/metabolic syndrome side effects (42). Atazanavir is clearly less potent than lopinavir in inhibiting ZMPSTE24 and causing prelamin A accumulation, but whether this biochemical property underlies the more favorable side effect profile is unknown. To the extent that prelamin A is important, one would imagine that darunavir would be far less likely to cause the lipodystrophy/metabolic syndrome, inasmuch as this drug does not inhibit ZMPSTE24 or lead to prelamin A accumulation. Darunavir has been in clinical use for only a short period of time, so its side effect profile, relative to the older HIV-PIs, is not completely established. Sorting out whether darunavir is associated with fewer side effects in human studies could prove to be difficult. Darunavir is nearly always given alongside ritonavir (14), an HIV-PI that impedes the metabolism of other HIV-PIs, thereby increasing their plasma levels and their potency in reducing viral loads (15). In our studies, ritonavir inhibited ZMPSTE24 and led to an accumulation of prelamin A in cultured cells.

TABLE 1
Comparison of the structure and activities of lopinavir and darunavir a 50% inhibitory activity against wild-type HIV protease determined by measuring p24 GAG production (43). b 50% effective concentration against wild-type HIV-1 viral activity (44,45). c affinity constant for the binding of the drug to wild-type HIV-1 protease, measured by surface plasmon resonance (36). d Trough and maximum drug concentrations in the plasma (31,32,46). e 50% inhibitory activity against mouse ZMPSTE24 (12). f Treatment leading to an accumulation of farnesyl-prelamin A in mammalian cells. The molecular structures were edited on PubChem Server Side Structure Editor v1. 21 (pubmed.ncbi.nlm.nih.gov/edit).