Nature of Pharmacophore Influences Active Site Specificity of Proteasome Inhibitors*

Proteasomes degrade most proteins in mammalian cells and are established targets of anti-cancer drugs. The majority of proteasome inhibitors are composed of short peptides with an electrophilic functionality (pharmacophore) at the C terminus. All eukaryotic proteasomes have three types of active sites as follows: chymotrypsin-like, trypsin-like, and caspase-like. It is widely believed that active site specificity of inhibitors is determined primarily by the peptide sequence and not the pharmacophore. Here, we report that active site specificity of inhibitors can also be tuned by the chemical nature of the pharmacophore. Specifically, replacement of the epoxyketone by vinyl sulfone moieties further improves the selectivity of β5-specific inhibitors NC-005, YU-101, and PR-171 (carfilzomib). This increase in specificity is likely the basis of the decreased cytotoxicity of vinyl sulfone-based inhibitors to HeLa cells as compared with that of epoxyketone-based inhibitors.

The chymotrypsin-like sites have long been considered the only suitable targets for anti-neoplastic agents and are the primary targets of all these agents. However, our recent work indicates that cytotoxicity of proteasome inhibitors correlates poorly with exclusive inhibition of the chymotrypsin-like sites and that co-inhibition of other sites is usually needed to achieve maximal cytotoxicity (12). In this regard, we have considered it of interest to determine whether inhibitors with increased specificity for ␤5 display decreased cytotoxicity.
Many structural classes of proteasome inhibitors are known (2,13). The majority of these are N-terminally capped short peptides (2-4 residues) with an electrophilic trap at the C terminus (e.g. aldehydes, boronates, epoxyketones, and vinyl sulfones). This electrophile reacts with the catalytic N-terminal threonines of the proteasome. The peptide portion binds in substrate-binding pockets and defines the active site specificity of inhibitors. It has long been assumed that the nature of the pharmacophore, while influencing reactivity of the compound, does not affect specificity, at least when it comes to proteasome active sites. However, we have recently discovered that changing pharmacophores without altering the peptide portion of the inhibitor can affect active site specificity (14). For example, in the process of development of active site probes, we have made the surprising observation that changing epoxyketone to vinyl sulfone in the ␤5-specific inhibitor NC-005 increases the ␤5 specificity of this agent (15). In the study presented here, we address the question of whether the same is true for other ␤5-specific (e.g. carfilzomib, YU-101) (3,16) and ␤5i-specific (e.g. PR-957) (17) epoxyketones and, if so, whether this increase in specificity leads to a decrease in cytotoxicity of these compounds.
Another indication that the pharmacophore may affect the specificity of inhibitors is a recent report by Marastoni et al. (18) that Hmb 5 -Val-Ser-Leu-vinyl ester (Hmb-VSL-ve) is a specific inhibitor of the trypsin-like (␤2) sites. Trypsin-like sites cut peptide bonds after basic residues, and inhibitors with leucine in the P1 position would not be expected to be specific for the trypsin-like sites (19), unless one assumes that the vinyl ester moiety contributes to ␤2-specific targeting. To determine whether the ␤2 specificity of this compound is determined by the vinyl ester pharmacophore or by its peptide fragment, we have swapped the pharmacophores and peptide fragments between this compound and the ␤5and ␤1-specific epoxyketone and vinyl sulfones we synthesized previously (12,20).
The combined arguments outlined above led to the design of several new peptide-based proteasome inhibitors, on which we report here. Our data reveal the following findings: 1) peptide-based vinyl esters have no inhibitory activity toward proteasomes; 2) replacement of epoxyketones by vinyl sulfones increases the specificity of inhibitors for the ␤5 sites (but not for the ␤5i sites); and 3) this increase in specificity decreases cytotoxicity of the compounds, confirming our previously reported observation that inhibition of other sites in conjunction with the chymotrypsin-like sites is a prerequisite for potential anti-tumor activity (12).
Purification of 26 S Proteasomes-For the purification of constitutive proteasomes, young rabbit muscles (200 g, Pel-Freeze Biologicals) were homogenized in a blender in 500 ml of buffer containing 50 mM Tris-HCl, pH 7.5, 1 mM DTT, 1 mM EDTA, 0.25 M sucrose, 5 mM MgCl 2 , and 2 mM ATP. The homogenate was centrifuged for 15 min at 10,000 ϫ g and then for 30 min at 40,000 ϫ g. The supernatant was filtered through a 5-micron filter, and proteasomes were batch-absorbed on 50 ml of DE52 DEAE-cellulose. After 30 min of stirring with the supernatant, the resin was washed on a glass filter with ϳ500 ml of buffer A (20 mM Tris-HCl, pH 7.5, 10% glycerol, 1 mM ATP, 5 mM MgCl 2 , 1 mM DTT, 0.5 mM EDTA) and then with 250 ml of 50 mM NaCl in buffer A. Proteasomes were eluted with 150 mM NaCl in the same buffer, and 40 -50-ml fractions were collected. All fractions were monitored for activity using Suc-LLVY-amc as a substrate. Active fractions were pooled (ϳ200 mg of total protein) and loaded on a 10-ml Source Q (GE Healthcare) column, which was eluted by a gradient of 0.15-0.35 M of NaCl in 120 ml of buffer A at a flow rate of 3 ml/min. Fractions containing proteasome activity (eluting approximately at 0.28 M NaCl) were combined to give ϳ40 mg of total protein, diluted 2-fold, and loaded on a 1.3-ml Uno Q column. 26 S proteasome was separated from 20 S proteasome by a gradient of 0.13-0.3 M NaCl in 30 ml of buffer A. 20 S proteasome-containing fractions were distinguished from the 26 S containing fractions based on SDS activation in the peptidase assays (21). The reason for two consecutive high resolution cation exchange chromatography steps is that Source Q column provided better separation from contaminating proteins than the Uno Q column but did not separate 20 S and 26 S proteasomes. Fractions containing 26 S proteasomes (1-2 mg of protein in a total volume of 1-2 ml/per tube) were loaded on a 32-ml 20 -40% glycerol gradient (in 20 mM HEPES, pH 7.5, 1 mM DTT, 0.5 mM EDTA, 5 mM MgCl 2 , 0.5 mM ATP). After 16 h of centrifugation at 130,000 ϫ g, gradients were fractionated and active fractions pooled, concentrated using Centriprep YM-50 devices, aliquoted, and stored at Ϫ80°C. Purification of immunoproteasomes was carried out from frozen rabbit spleen using a similar procedure, except that the amount of tissue was 10 g.
Inhibitor Assays-Purified 26 S proteasomes (ϳ10 ng/ml) were incubated with various concentrations of inhibitors at 37°C for 30 min in the assay buffer (50 mM Tris-HCl, pH 7.5, 1 mM ATP, 50 g/ml BSA, 2 mM EDTA, 40 mM KCl). Immediately after the end of this incubation, an aliquot of the inhibitor-treated proteasome was mixed with the 100 M substrate (Suc-LLVY-amc for the ␤5 or ␤5i sites, Ac-nLPnLDamc for ␤1/␤1i sites, and Ac-RLR-amc or Ac-RQR-amc for ␤2/␤2i sites), and fluorescence of released amc was measured continuously for 30 min at 37°C. (Substrate solutions did not contain inhibitors except when reversible inhibitor MG-132 was tested; in this case, MG-132 was added to the substrate at the same concentration as in the enzyme/inhibitor preincubation mixture.) The rate of reaction was determined from the slope of the reaction progress curves. Residual activity was calculated as the slope of reaction in inhibitor-treated sample divided by the slope of reaction in the control sample (i.e. proteasomes incubated under the same conditions but in the absence of inhibitor).
Extracts of HEK-293T cells (10 g of protein, prepared as described previously (15)) were incubated with inhibitors for 1 h at 37°C, then with 1 M MV-151 for an additional hour at 37°C, and then fractionated on 12.5% SDS-PAGE. Upon completion of electrophoresis, gels were scanned on a Typhoon imager (excitation laser, 532 nm; emission filter, 560 nm).
Tissue Culture Experiments-HeLa S3 cells were cultured in DMEM supplemented with 5% newborn calf serum and penicillin and streptomycin. Proteasome activity in inhibitortreated cells was measured with luminogenic substrates using Promega ProteasomeGlo TM cell-based assay (Promega) (22).
Inhibitors were washed out prior to measurements. See the supplemental material in Ref. 12 for details of the procedure. Cell viability measurements were performed using Alamar Blue mitochondrial dye conversion assay (12).
Preparation of Cytosol-depleted Extracts for Cathepsin Activity Measurements-Cells were harvested, washed with PBS, and permeabilized on ice with 0.05% digitonin in 4 -5 volumes of 50 mM Tris-HCl, pH 7.5, containing 250 mM sucrose, 5 mM MgCl 2 , 1 mM DTT, 1 mM ATP, and 0.5 mM EDTA. Cytosol was squeezed out by centrifugation for 15 min at 20,000 ϫ g at 4°C, and residual cell pellet was lysed with a buffer containing 50 mM BisTris-HCl, pH 5.5, 10% glycerol, 5 mM MgCl 2 , 1 mM EDTA, 2 mM DTT, and 0.5% CHAPS. These cytosol-depleted acidic extracts were used for measurement of cathepsin activity. Protein concentration in extracts was determined using Pierce 660 nM protein assay reagent.
Measurements of Cathepsin Activity-An aliquot of cytosoldepleted acidic extracts was added to 100 l of 40 M pancathepsin substrate Z-FR-amc in 100 M phosphate buffer, pH 6.0, 2 mM EDTA, 4 mM DTT (23). Increase of fluorescence of released amc was recorded continuously for 30 min, and the rate of reaction was calculated from the slopes of the linear reaction progress curves. Cleavage of this substrate was completely blocked in extracts of cells treated with 5 M E-64d.
All these compounds were tested for their ability to inhibit purified 26 S proteasomes from rabbit muscle and proteasomes in extracts of HEK-293T cells. After 30 min of incubation with 40 M vinyl esters compounds, none of these inhibited activity of purified proteasomes (Fig. 1B). Vinyl esters were then incubated with extracts of HEK-293T cells, and proteasome inhibition was evaluated based on ability to prevent subsequent modification of the catalytic subunits by the fluorescent activity-based probe MV-151 (24). Except for a weak inhibition of ␤5 site at 100 M of Hmb-VSL-ve (instead of expected inhibition of ␤2 site; Fig. 1C), no inhibition was observed. Thus, in contrast to what has been reported, peptide vinyl esters do not inhibit any proteasome active sites.
In contrast, epoxyketone and vinyl sulfone derivatives of Hmb-VSL-ve (Fig. 1D) inhibited proteasomes in both assays (Fig. 1, E-H) but were not ␤2-specific. The preferred target of these compounds was the ␤5 site. The vinyl sulfones (Fig. 1, G and H) were more ␤5-specific than the epoxyketones (Fig. 1F).
Comparison of ␤5-specific Vinyl Sulfones and Epoxyketones-The observation that Hmb-VSL-pvs and Hmb-VSL-mvs are more ␤5-specific than Hmb-VSL-ek is consistent with the earlier observation that vinyl sulfone derivatives of NC-005 (NAc-mYFL-ek) are more ␤5-specific than NC-005 itself (15). This effect was originally observed in HEK-293T lysates with the MV-151 activity-based probe (15). Here, we confirm this observation using purified 26 S proteasomes and fluorogenic substrates (Fig. 2, B-D). Although the vinyl sulfones are less potent inhibitors of the ␤5 sites than the epoxyketone, they do not inhibit ␤1 and ␤2 sites. In contrast, the epoxyketones markedly inhibited ␤2 sites and reduced activity of ␤1 sites partially (Fig. 2B).
To test the generality of these findings, we have synthesized methyl vinyl sulfone derivatives of two other ␤5-specific epoxyketones, YU-101 (16) and PR-171 (carfilzomib) (3). In both cases, vinyl sulfones were more ␤5-specific than epoxyketones (Fig. 2, E-H). YU-101-vs is the most ␤5-specific, as it did not inhibit ␤1 and ␤2 sites even at 100 M (Fig. 2F). It should be noted that among parental epoxyketones, YU-101 is also more ␤5-specific than PR-171 (compare Fig. 2, E and G). Thus, replacement of epoxyketone by vinyl sulfones increases selectivity of inhibitors to the ␤5 sites (at least in the context of leucine in the P1 position).
Comparison of MG-132 Derivatives with Different Pharmacophores-Proteasome inhibitors with different pharmacophores are widely used by the scientific community. Because blocking the ␤5 site alone is not sufficient to block the bulk of protein degradation (25), the question of how the chemical nature of the pharmacophore affects active site specificity of inhibitors is of great importance to the scientific community. For example, a scientist using MG-132 (Z-L3aldehyde(al)) or its vinyl sulfone analogue Z-L3-mvs may need to substitute for these an inhibitor that does not block lysosomal proteases, such as MG-262 (Z-L3-boronate) or Z-L3-ek. Information on the impact of this substitution on the active site specificity would be very useful. We have analyzed inhibition of purified 26 S proteasomes by MG-132 and its boronate (MG-262), methyl vinyl sulfone (Z-L3-mvs), and epoxyketone (Z-L3-ek) derivatives. Although the ␤5 site was the primary target of all four compounds, only vinyl sulfone (Fig. 3B) was truly ␤5-specific, achieving 95% inhibition of ␤5 sites before significant inhibition of ␤1 and ␤2 sites was observed. The epoxyketone (Fig. 3C) was slightly less specific; 85% inhibition of ␤5 sites was achieved before inhibition of ␤1 and ␤2 sites was observed. MG-262 (Fig. 3D) was ␤5-specific up to 70% inhibition, after which both ␤2 and ␤1 sites were rapidly inhibited. In MG-132 (aldehyde)-treated proteasomes (Fig.  3A), only 50% inhibition of ␤5 sites could be achieved before inhibition of ␤1 sites was observed; inhibition of ␤2 sites was observed at higher inhibitor concentrations. We conclude that the nature of the pharmacophore affects the secondary active site specificity of proteasome inhibitors.

Pharmacophore Effect on Specificity of Proteasome Inhibitors
Replacement of the pharmacophore produced results different from those observed in constitutive proteasomes. With the exception of YU-101-mvs (Table 1), which was more ␤5ispecific than YU-101, all other vinyl sulfones were less ␤5ispecific than their epoxyketone counterparts. Thus, vinyl sulfones do not improve the targeting of inhibitors to the chymotrypsin-like sites of immunoproteasomes.
Increasing ␤5 Site Specificity Decreases Inhibitor Cytotoxicity-In our previous study, we showed that cytotoxicity of proteasome inhibitors poorly correlates with the inhi-bition of ␤5 sites and that co-inhibition of ␤2 and/or ␤1 sites is observed under cytotoxic conditions (12). This result predicts that increasing ␤5 specificity would decrease cytotoxicity of inhibitors. We tested this prediction by comparing effects of NC-005 and homologous phenol vinyl sulfone NC-005-pvs on HeLa cells. This pair of inhibitors was chosen for comparison as they offered more distinct differences in specificity than YU-101-and PR-171-based pairs (Fig. 2). Between the two NC-005-derived vinyl sulfones, 4-hydroxyphenyl vinyl sulfone was chosen over methyl vinyl sulfone as a more potent inhibitor. HeLa cells were chosen over the myeloma cells used in our previous study (12) because they do not express immunoproteasomes, in which differences in active site specificity between vinyl sulfone and epoxyketone would be less dramatic due to the lack of pharmacophore effect on ␤5i targeting (Table 1). When HeLa cells were treated with these agents (Fig. 4 and supplemental Fig. S2), differences in potencies and specificities were the same as with purified proteasomes (Fig. 2), with the epoxyketone being an ϳ10-fold more potent inhibitor of ␤5 sites than the methyl vinyl sulfone. The most noticeable difference between purified proteasomes and proteasomes in HeLa cells was activation of ␤2 activity by vinyl sulfone in cells ( Fig. 4 and supplemental Fig. S2).
As in our previous work (12), we treated cells with inhibitors for 1 h and then removed the inhibitors and cultured  DECEMBER 17, 2010 • VOLUME 285 • NUMBER 51 JOURNAL OF BIOLOGICAL CHEMISTRY 40129 cells for an additional 48 h, at which point cell viability was measured with an Alamar Blue mitochondrial dye conversion assay. Immediately after the removal of the drug, inhibition of the proteasome was confirmed by measuring activity of ␤1, ␤2, and ␤5 sites with site-specific luminescent substrates. Recovery of activity was followed throughout the washout period with the same assay.

Pharmacophore Effect on Specificity of Proteasome Inhibitors
We observed different effects on cell viability from 1-h exposure with NC-005 and NC-005-pvs. Vinyl sulfone was not cytotoxic at concentrations as high as 80 M (Fig. 4A), at which ␤5 activity was inhibited by 90% and remained inhibited by Ͼ85% during the 24-h washout period (Fig. 4C). It should be noted that ␤2 activity was activated by 20 -40% by this treatment and stayed activated through the washout period (supplemental Fig. S3A). Contrary to the vinyl sulfone, 1-h exposure to the epoxyketone NC-005 induced cytotoxicity (Fig. 4B). However, induction of cytotoxicity coincided with the inhibition of the ␤1 and ␤2 activities but not with the inhibition of the ␤5 sites (Fig. 4D). Stronger cytotoxicity of NC-005 cannot be explained by the slower recovery of proteasome activity during the washout period as this recovery was in fact faster in NC-005-treated cells (Fig. 4D) than in NC-005-pvs-treated cells (Fig. 4C). Thus, a more specific targeting of proteasome inhibitors to the ␤5 site decreases their cytotoxic potential.
If stronger cytotoxicity of the epoxyketone is due to its ability to co-inhibit ␤1 and/or ␤2 sites, co-inhibiting ␤1 sites by ␤1-specific inhibitor NC-001 (12) in NC-005-pvs-treated cells should sensitize them to this agent. Indeed, adding NC-001 to the media during recovery of NC-005-pvs-treated cells led to a dramatic 70 -80% decrease in viability under conditions where ␤5 was almost completely inhibited (Fig. 4E). Contrary to this, the same ␤1-specific inhibitor did not cause significant sensitization of HeLa cells to the epoxyketone NC-005 (Fig. 4F). (We confirmed by activity measurements that NC-001 treatments inhibited activity of ␤1 sites by more than 90%

TABLE 1
Effect of inhibitors on the purified immunoproteasomes 26 S proteasomes, purified from rabbit spleens, were incubated with different concentrations of inhibitors for 30 min at 37°C followed by measurements by activities as described under "Experimental Procedures." Residual activity was plotted against concentration on semi-log plots (supplemental Fig. S1), which were used to determine IC 50 values. (To allow for easy comparison with data on Fig. 2 but did not change inhibition of ␤5 and ␤2 sites, see supplemental Fig. S3.) Thus, complete or nearly complete inhibition of ␤5 and either ␤1 or ␤2 sites is needed to decrease viability of HeLa cells to less than 10%.
We then asked what would be the effects on cells of specific inhibition of ␤5 sites under the conditions when recovery of proteasome activity is not possible. We treated HeLa cells with NC-005-pvs continuously. Under these conditions, inhibition reaches maximum within 6 h (supplemental Fig. S2) and does not recover (Table 2). We found that specific 70% inhibition of ␤5 sites at 0.3 M NC-005-pvs (40% activation of ␤2 activity was observed at this concentration) did not lead to cytotoxicity (Table 2). 95% inhibition of ␤5 sites at 1 M NC-005-pvs (with simultaneous inhibition of ␤1 sites by 20% and activation of ␤2 sites by 15%) led to a 65% decrease in viability. Thus, specific inhibition of ␤5 sites is cytotoxic to HeLa cells only when it is nearly total and is long lasting.
As NC-005-pvs concentrations increased, its specificity decreased leading to a slight increase in cytotoxicity, but even at the highest concentration used some cells remained viable. In a parallel experiment, we treated cells continuously with the epoxyketone NC-005 (Table 2), and the percentage of surviving cells appeared to be lower than with the vinyl sulfone treatment.
To determine how increased selective targeting of inhibitors to ␤5 sites affects residual survival rates, we performed clonogenic survival assay of cells treated with NC-005 and NC-005-pvs for 24 h (Table 3). We used concentrations of compounds that completely inhibited ␤5 sites but varied in the inhibition of ␤1 and ␤2 sites. Of cells treated with 3 M NC-005, which inhibited ␤1 and ␤2 sites by more than 80%, none survived. A smaller percentage of cells treated with 1 M NC-005, which inhibited ␤1 and ␤2 sites by more than 60%, survived than cells treated with 9 M of vinyl sulfone, which inhibited ␤1 sites by 46% and ␤2 sites by 20%. Thus, strong co-inhibition of ␤1 and ␤2 sites, as occurs in NC-005 treated cells, is needed to suppress residual survival of HeLa cells. . NC-001 completely inhibited ␤1 activity but did not inhibit ␤2 activity and did not alter recovery rate of ␤5 activity (supplemental Fig. S3). On all graphs, values are average Ϯ S.E. of 2 or 3 independent measurements (i.e. biologic replicates).
Vinyl sulfones can potentially inhibit cysteine proteases such as lysosomal cathepsins (27). To determine whether a 6-h treatment with NC-005-pvs leads to inhibition of these enzymes, we have measured cathepsin activity in extracts of inhibitor-treated cells (Table 4). We used the fluorogenic peptide substrate Z-FR-amc, which is cleaved by the majority of cathepsins (23). Cleavage of this substrate was inhibited in cells treated by EST (E-64d), a cell-permeable precursor of the class-specific inhibitor of cysteine proteases E-64 (Table 5). Indeed, NC-005-pvs but not NC-005 inhibited this activity in a concentration-dependent manner (Table 4). To determine whether inhibition of cathepsin by NC-005-pvs contributes to cytotoxicity of NC-005-pvs, we determined whether E-64d is cytotoxic to cells under similar treatment conditions as used in this experiment for NC-005-pvs. Because E64-d did not cause any reduction in cell viability after 48 h of treatment (Table 5), we conclude that inhibition of cathepsins is unlikely to contribute to the cytotoxicity of NC-005-pvs.

DISCUSSION
Vinyl Sulfones Are More Specific ␤5 Inhibitors than Epoxyketones-Although we noticed a few years ago that the nature of the electrophilic group may affect the active site specificity of proteasome inhibitors (14), we report here the first systematic comparison of vinyl sulfones and epoxyketones in specific targeting of inhibitors to the chymotrypsinlike sites of the proteasome. Our conclusion that vinyl sulfone inhibitors are more ␤5-specific than epoxyketone inhibitors is supported by the data on five series of compounds. 1) Hmb-VSL-mvs and Hmb-VSL-pvs are clearly more specific than Hmb-VSL-ek (Fig. 1). 2) Replacement of epoxyketone in NC-005 by either of the vinyl sulfone pharmacophores dramatically decreases its ability to co-inhibit ␤2 and ␤1 sites (Figs. 2 and 4) (15), even with prolonged treatment of cells (Table 2 and supplemental Fig. S2). 3) Conversion of the epoxyketone YU-101 into a vinyl sulfone abolishes inhibition of ␤1 and ␤2 sites (Fig. 2). 4) PR-171-mvs is a more specific ␤5 inhibitor than the parent epoxyketone PR-171. 5) Z-L3-mvs is more ␤5-specific than Z-L3-ek (Fig. 3). This conclusion does not extend to the immunoproteasomes, as vinyl sulfones do not improve selectivity of inhibitors to the ␤5i sites (Table 1 and supplemental Fig. S1).
Vinyl Esters Do Not Inhibit Proteasomes-As we clearly demonstrate the effect of a vinyl sulfone pharmacophore on the ␤5 specificity, we reject the previous claim that vinyl esters composed of the same peptide sequence are selective inhibitors of the ␤2 sites (18). In fact, we found that said peptidyl vinyl ester does not have any proteasome inhibitory activity at all, at least in our assays (Fig. 1). An explanation for the differences between our results and that of Marastoni et al. (18) might be that the inhibitory activity in the preparation of the vinyl ester used by Marastoni et al. belongs not to a major component but to a minute contaminant (or possibly a contaminating diastereomer) that was not separated by HPLC or detected by NMR. We prepared the vinyl ester via two synthetic routes (see supplemental material), including the reported route, and took care to purify the compound to homogeneity, and we are therefore confident that we have in fact prepared the compound claimed by Marastoni et al. (18) as a ␤2-specific inhibitor.
Increasing ␤5 Specificity Decreases Cytotoxicity of Inhibitors-The proteasome inhibitor bortezomib is being used clinically for the treatment of multiple myeloma, and second generation inhibitors are at different stages of development (3)(4)(5)(6). Development of all these compounds has been focused on inhibition of ␤5 sites. However, most of them co-target ␤1 and/or ␤2 sites, and it is not completely clear whether co-inhibiting these sites is important for their anti-neoplastic activity. Thus, an important issue for the development of next-generation compounds is whether targeting ␤5 sites is sufficient to achieve optimal anti-neoplastic activity. In our previous study, we have shown that, for the majority of multiple myeloma cell lines, cytotoxicity of NC-005 poorly correlates with ␤5 inhibition (12) and that adding a ␤1-specific inhibitor sensitizes them to NC-005. These data suggest that specific inhibition of ␤5 sites would not be sufficient to induce cytotoxicity in the majority of cell lines. Development of a more selective ␤5-specific inhibitor has allowed us to test this prediction in this study. Upon a 1-h pulse treatment of HeLa S3 cells, we found that as ␤5 specificity increases, cytotoxicity of inhibitors decreases dramatically (Fig. 4). Specific inhibition of ␤5 sites leads to the loss of viability only if inhibition exceeds 95% and is continuous ( Table 2). Even under these conditions, loss of viability is only partial (65%). Strong co-inhibition (80%) of other sites is needed to suppress residual viability (Table 3). These data are consistent with the observations of Parlati et al. (28), who found that specific inhibition of the chymotrypsin-like activity causes partial loss of viability of cell lines derived from hematologic malignancies. It should be noted that the conditions under which we observed that specific inhibition leads to cytotoxicity (e.g. 95% inhibition of ␤5 sites lasting 20 h, with only 20% inhibition of ␤1 sites and activation of ␤2 sites) could not be achieved with any other reported ␤5specific inhibitors as they all lose specificity when ␤5 inhibition is so strong.
A caveat in using vinyl sulfones is their potential for inhibition of cysteine proteases (e.g. cathepsins) (27). We have addressed this concern by measuring cathepsin inhibition (Table 4). Even though we found such an inhibition, these off-target effects of vinyl sulfones are unlikely to contribute to the cytotoxicity of the compounds because the class-specific inhibitor of thiol proteases E-64d was not cytotoxic to HeLa cells (Table 5).
In certain situations, conferring the ability to inhibit cathepsins to proteasome inhibitors may improve their therapeutic utility. Bortezomib was recently shown to have additive effects with a cathepsin S inhibitor in a mouse model of multiple sclerosis (29). Thus, peptide vinyl sulfones that target proteasome chymotrypsin-like activity and cathepsins may find therapeutic applications in the treatment of autoimmune disease.
In summary, this work clearly demonstrates the importance of pharmacophores in determining active site specificity of proteasome inhibitors and provides new tools for highly specific inhibition of proteasome ␤5 sites.