A Structural Basis for the Acute Effects of HIV Protease Inhibitors on GLUT4 Intrinsic Activity

Human immunodeficiency virus (HIV) protease inhibitors (PIs) act as reversible noncompetitive inhibitors of GLUT4 with binding affinities in the low micromolar range and are known to contribute to alterations in glucose homeostasis during treatment of HIV infection. As aspartyl protease inhibitors, these compounds all possess a core peptidomimetic structure together with flanking hydrophobic moieties. To determine the molecular basis for GLUT4 inhibition, a family of related oligopeptides containing structural elements found in PIs was screened for their ability to inhibit 2-deoxyglucose transport in primary rat adipocytes. The peptide oxybenzylcarbonyl-His-Phe-Phe-O-ethyl ester (zHFFe) was identified as a potent inhibitor of zero-trans glucose flux with a K(i) of 26 mum. Similar to PIs, transport inhibition by this peptide was acute, noncompetitive, and reversible. Within a Xenopus oocyte expression system, zHFFe acutely and reversibly inhibited GLUT4-mediated glucose uptake, whereas GLUT1 activity was unaffected at concentrations as high as 1 mm. The related photoactivatable peptide zHFF-p-benzoylphenylalanine-[(125)I]Tyr-O-ethyl ester selectively labeled GLUT4 in rat adipocytes and indinavir effectively protected against photolabeling. Furthermore, GLUT4 bound to a peptide affinity column containing the zHFF sequence and was eluted by indinavir. These data establish a structural basis for PI effects on GLUT4 activity and support the direct binding of PIs to the transport protein as the mechanism for acute inhibition of insulin-stimulated glucose uptake.


Introduction
The clinical use of HIV protease inhibitors (PIs) has led to dramatic improvements in HIV-related morbidity and mortality (1). However, it is now recognized that these drugs directly contribute to several metabolic changes that significantly increase the risk of developing diabetes mellitus and cardiovascular disease (2,3).
Insulin resistance appears early in the development of this dysmetabolic syndrome, and can occur in the absence of lipodystrophy or hyperlipidemia (4). While the complexity of drug regimens used during highly active antiretroviral therapy (HAART) has complicated efforts to characterize adverse effects that are the direct result of PI use, a definitive link to the development of insulin resistance has been established by both in vitro and in vivo studies (5)(6)(7).
Several mechanisms have been proposed to mediate PI-induced insulin resistance including changes in insulin signaling (8,9), SREBP processing (10), and adipokine secretion (11). Our previous studies establishing that the insulin responsive facilitative glucose transporter GLUT4 is acutely inhibited by PIs at pharmacologically relevant drug levels (12) have identified a direct molecular target for the cellular effects of these drugs.
Several observations support the hypothesis that GLUT4 inhibition is produced by direct, noncovalent binding of PIs to a unique structural domain within the transport molecule: Despite these data, without direct evidence that PIs bind to GLUT4, it remains possible that the effects of PIs on GLUT4 activity are indirect. For example, the drugs could interact with a unique regulatory molecule that either binds to GLUT4 or reversibly alters its structure such as by phosphorylation. Elucidation of the specific structural features of PIs that confer their ability to inhibit GLUT4 would not only facilitate efforts to define the molecular mechanism for this effect but could also provide a rationale way to design newer generations of PIs that retain their efficacy in treating HIV infection without producing insulin resistance.
We report here the identification of acute, potent, and isoform-selective peptide inhibitors of GLUT4 and provide evidence that this inhibition is due to direct binding of these compounds to the transporter protein.  GLUT4 content was assessed by Western blot analysis using polyclonal antibody directed toward the carboxy terminus of rat GLUT4.

Results.
Aromatic peptide effects on GLUT4 activity. All HIV protease inhibitors share a similar core structure that resembles the aromatic peptide backbone that acts as the native substrate for the HIV-1 protease ( Figure 1). We therefore hypothesized that synthetic peptides containing this basic structure would serve as effective inhibitors of glucose transport. To test this hypothesis we screened a family of aromatic di and tripeptides for their affects on insulin stimulated glucose uptake in primary rat adipocytes, which predominately express the insulin responsive facilitative glucose transporter GLUT4 (13).
With the addition of 200 µM of each peptide to primary rat adipocytes, four compounds were identified with significant inhibitory effects on zero-trans 2-deoxyglucose flux ( Figure 2). Like PIs, all four of these peptides contain a highly aromatic core peptide structure flanked by hydrophobic moieties. None of the peptides with charged amino or carboxy terminus affected transport activity. The most potent inhibitor of glucose transport identified in this screen was z-His-Phe-Phe-o-Et (zHHFe), which is also a known substrate for the aspartyl protease pepsin (14). and 205 ± 5 µM, respectively. This difference is comparable to that seen with indinavirmediated inhibition of glucose uptake in these two cell types (5,12). The intercept on the negative x-axis is indicative of non-competitive inhibition under the kinetic conditions used, which is also identical to the inhibition pattern observed for indinavir in primary rat and 3T3-L1 adipocytes (12). Due to the rapidity of the inhibitory effects of zHFFe (occurring within minutes after adding the drug) in adipocytes following insulin stimulation, it is unlikely that zHFFe produces its effect through alterations in insulin signaling or GLUT4 translocation. This is further supported by the inhibitory effect of this peptide in the Xenopus oocyte system, in which cell surface expression of the glucose transporter is independent of insulin signaling (15).

Kinetic characterization of the inhibition of GLUT4 by z-His
Isoform selectivity and reversibility of zHFFe effects on glucose transport-One of the distinguishing features of PI-mediated effects on glucose transport is their selectivity for inhibition of GLUT4. In the Xenopus oocyte system, GLUT1 is resistant to inhibition by indinavir at millimolar concentrations, whereas GLUT4 is highly sensitive with a K i of approximately 50 µM (12). To determine the isoform selectivity of zHFFe, oocytes injected with mRNA for GLUTs 1-4 were tested for their ability to acutely and reversibly inhibit zero-trans 2-deoxyglucose transport. As shown in figure 4A, only GLUT4 was inhibited by 250 µM peptide, whereas concentrations of zHFFe up to 1 mM had no effect on GLUT1 or GLUT3. A small (32%) but statistically insignificant (P=0.087) reduction in GLUT2 activity was observed with 1 mM zHFFe . This compares favorably with the behavior of PIs, which have been shown to inhibit GLUT2 at high concentrations in Xenopus oocytes (16). As shown in Figure 4B, acute inhibition of GLUT4 was fully by guest on March 20, 2020 http://www.jbc.org/ Downloaded from 11 reversible following removal of the peptide. Full restoration of transport activity was observed irrespective of the time in which the oocytes were exposed to drug (1-30 minutes, data not shown).
Photolabeling with a reactive peptide derivative. Having identified zHFFe as a peptidebased inhibitor of GLUT4 that shares inhibitory characteristics identical to those observed with the PI indinavir, a radiolabeled and photoreactive derivative of this peptide was synthesized and used to test the hypothesis that transport inhibition is due to direct binding of the drug to the transport protein. The peptide z-His-Phe-Phe-Bpa-Tyr-o-Et was found to effectively inhibit glucose transport in primary rat adipocytes with a binding affinity comparable to the parent zHFFe compound (Figure 5a). When the adipocytes were exposed to UV light to activate the Bpa residue, a concentration dependent and irreversible inhibition of transport activity by the photactivatable peptide was observed ( Figure 5B). As shown in Figure 6A, Together, these observations support the conclusion that the photolabel is bound to the same GLUT4 site as PIs.
Peptide affinity chromatography-The relative ease in modifying the core zHFF epitope that confers selective binding to GLUT4 provided a means to further investigate the relationship between aromatic peptide and PI-mediated inhibition of glucose transport.
The peptide z-His-Phe-Phe-Phe-Lys, which contains the zHFF epitope, was linked to an agarose resin to generate an affinity column which could bind GLUT4. As shown in Figure 7, when homogenates of primary rat adipocytes were applied to this resin, immunoreactive GLUT4 was found to bind to this column and could be selectively eluted with 500 µM indinavir. The native structure of GLUT4 appeared to be necessary for this interaction, since boiling of the adipocytes homogenate prior to loading the peptide column resulted in no immunodetectable GLUT4 binding to the column (data not shown). Since the peptidomimetic phenylalanine moiety is the only feature that is consistently shared between PIs and the newly identified peptide inhibitors, this structure is likely to be a primary determinant in conferring the ability to inhibit GLUT4 activity.
The shared participation of this structural feature in the binding of PIs to both the HIV 14 protease and GLUT4 may present a challenge in new drug design. Attempts to change this core structure to prevent GLUT4 inactivation may render the candidate drug ineffective in inactivating the HIV protease. The recent development of atazanavir, a PI which does not appear to alter glucose uptake in 3T3-L1 adipocytes (19), provides optimism that the problem of GLUT4 inactivation by PIs can be circumvented.
Reflecting the close correlation between the effects of PIs on GLUT4 activity in vitro and insulin sensitivity in vivo, atazanavir does not appear to contribute to changes in glucose homeostasis in HIV infected patients who have been treated with this drug (20).
Atazanavir differs from the other available PIs in several respects. While atazanavir contains a similar peptide-like core, with two phenylalanine-like structures arranged in a unique C-2 axis of symmetry, an additional pyridine ring is attached to one of the phenylalanine residues (Figure 1). While the location of the GLUT4 domain involved in PI binding remains to be identified, it is possible that this additional ring structure sterically prevents access of the drug to this region of the protein.
Atazanavir also differs from other PIs by lacking the same degree of hydrophobicity in the flanking structural domains. While the mechanism by which these hydrophobic moieties contribute to GLUT4 inhibition remains to be determined, all of the peptide inhibitors identified in this study contain an oxobenzylcarbonyl group at the amino terminus of the peptide. It is possible that this hydrophobicity facilitates entry of the drugs into the cell, thereby allowing access to a GLUT4 binding domain on the cytosolic surface of the transporter. The involvement of a cytosolic binding domain is suggested by our observation that the affinity labeled transporter could not be immunoprecipitated using antibody directed toward the carboxy terminus. This could by guest on March 20, 2020 http://www.jbc.org/ Downloaded from either be due to direct steric interference of the affinity label or an inhibitor-induced change in the protein conformation that leads to masking of this epitope. Since the epitope was recognized in the denatured protein on Western blots, the latter appears more likely. Unmasking of the carboxy terminus of GLUT4 in response to insulin treatment has previously been reported in both primary rat adipocytes (21) and skeletal muscle (22).
It has been postulated that this unmasking may contribute, together with GLUT4 translocation, to the activation of this transporter following insulin stimulation. The identification of peptide inhibitors thus raises the interesting possibility that an endogenous protein or peptide interacts at this same location to regulate GLUT4 intrinsic activity.
Although the current study supports the hypothesis that PIs directly bind to