2.1. Assays

2.2. Probe Chemical Characterization

The chemical structure of the probe was verified by analysis of its 400 MHz ¹H NMR spectra (Figure 3) obtained on a Brüker 400 MHz instrument and supported by analysis of a ¹⁹F NMR spectrum obtain on the same instrument. The chemical structure was also corroborated by its LC/MS molecular ion (calc for M+1: 480.1, found 479.9, using a Thermo/Finnegan LCQ Duo system). Purity was measured at >98% (LC/MS analysis, confirmed by analytical HLPC analysis; HPLC purity data is shown in Figure 4). HPLC data was obtained using an Agilent 1200 analytical HPLC with an Agilent Eclipse XDB-C18 column, 4.6×150mm. The HPLC solvents used were acetonitrile and water with 0.1% formic acid added to each mobile phase as the pH modifier.

Figure 3

¹H NMR spectrum.

Figure 4

HPLC Spectrum of ML345 at various wavelengths; purity >98%.

The solubility of the probe in PBS at pH 7.4 was determined to be moderately low (0.4 μM), however, the solubility in simulated general assay buffer (DMEM + FBS) was substantially higher (5.1 μM). Its solubility is fully adequate to provide the high potency seen in cell-based assays. The measured solubility suggests that, when appropriately formulated, the solubility of ML345 is adequate for broad use as a biological probe. The FBS effect could indicate high protein binding, which we did not directly measure.

The probe has a half-life of >>48 hours in PBS at room temperature when tested at 10 μM (Figure 5).

Figure 5

48 hour stability study of ML345 in PBS.

Disappearance of the LC peak for the probe is altered slowly but significantly by the addition of excess (50 μM) glutathione, with ~25% erosion of the parent peak area after 3 hours and ~41% erosion after 6 hours, suggesting a half-life of ~7.5 hours with 10-fold excess glutathione present. This agrees with the expectation that, while ML345 can be made to react with free thiols, it has significant selectivity for IDE, and specifically for reacting with Cys819 of IDE. The probe is stable in DMSO solution at 23°C (no erosion of peak intensity after 7 days) and is also stable as a free base dry powder. It is also stable under assay conditions, as indicated by potency in cell-based assays that is independent of incubation time.

The following compounds have been submitted to the MLSMR (MLS numbers pending):

Table 2Probe ML345 and Analogs

View in own window

Designation	Structure	CID	SID	SR ID	Amount submitted (mg)	Date Submitted
Probe ML345		57390068	144241486	SR-03000002959-2	20	12/17/2012
Analog 1		57390067	144241485	SR-03000002958-2	15	12/17/2012
Analog 2		5295225	144241488	SR-03000003056-1	15	12/17/2012
Analog 3		4317404	144241489	SR-03000003057-1	15	12/17/2012
Analog 4		5023882	144241490	SR-03000003058-1	15	12/17/2012
Analog 5		1510390	144241491	SR-03000003059-1	15	12/17/2012

2.3. Probe Preparation

Overview. The probe ML345 was synthesized in a straightforward fashion in a convergent manner, in six steps overall, with the longest linear sequence being 4 steps, as summarized in Figure 6. The overall yield of the process (after preparative HLPC purification of the final product) is 26%. Analogs for SAR evaluation were prepared by similar methods.

Figure 6

Synthesis of probe ML345.

Step 1

A mixture of acid chloride 1 (2.36 g, 10 mmol) and triethylamine (5.06 g, 50.0 mmol) in CH₂Cl₂ (50 mL) was treated with dropwise with morpholine (2.61g, 30 mmol) at room temperature. After addition was complete the reaction was stirred at room temperature for 14 h, quenched with saturated NH₄Cl, and extracted with ethyl acetate. The combined organic extracts were washed with brine and dried over Na₂SO₄. The solvent was removed and the residue was purified by flash column chromatography (Hexanes: ethyl acetate = 1:1, R_f = 0.15) to afford 3.261 g (91%) of compound 2 as a yellow solid. Calc’d for C₁₄H₁₉N₃O₆S: 357.1; found [M+H]⁺: 358.0; ¹H NMR (400 MHz, CDCl₃) δ (ppm) 3.03–3.06 (m, 4H), 3.21–3.23 (m, 4H), 3.76–3.79 (m, 4H), 3.87–3.89 (m, 4H), 7.18 (d, J=8.8 Hz, 1H), 7.79 (dd, J=2.0, 8.8 Hz, 1H), 8.17 (d, J=2.0 Hz, 1H).

Step 2

To a solution of compound 2 (488 mg, 1.37 mmol) in THF (10 mL) and MeOH (10 mL) was added Pd/C (10%, 50 mg). The reaction mixture was then stirred under atmosphere of H₂ for 5 h, filtered and concentrated to afford 378 mg (85%) of compound 3 as white solid. Calc’d for C₁₄H₂₁N₃O₄S: 327.1; found [M+H]⁺: 328.1.

Step 3

A solution of Acid 4 (344 mg, 2.0 mmol) in MeOH (12 mL) was treated with NaOH (160 mg, 4.0 mmol). The reaction mixture was stirred at room temperature for 10 min. The solvent was removed to afford a solid which was dissolved in acetone (18 mL). Benzyl bromide (376 mg, 2.2 mmol) was added. The reaction was sonicated for 5 min and then stirred at 0 °C for 1 h. The precipitate was collected by vacuum filtration. The solid was dissolved in H₂O and then treated with 1 N HCl. The precipitate was collected by vacuum filtration and dried in air, affording 462 mg (88%) of compound 5 as white solid. ¹H NMR (400 MHz, acetone-d₆) δ (ppm) 4.25 (s, 2H), 7.27–7.36 (m, 4H), 7.45–7.47 (m, 2H), 7.54–7.57 (m, 1H), 7.70–7.73 (m, 1H).

Steps 4 and 5

A suspension of acid 5 (570 mg, 2.17 mmol) in CH₂Cl₂ (24 mL) was treated with (COCl)₂ (441 mg, 3.48 mmol) and a drop of DMF at 0 °C under atmosphere of N₂. The reaction was stirred at 0 °C for 30 min, and then room temperature for 3 h. The solvent was removed. The residue was dissolved in CH₂Cl₂ (24 mL). Compound 3 (781 mg, 2.39 mmol) was added and cooled to 0 °C. Diisopropylethylamine (841 mg, 6.51 mmol) was added dropwise. The reaction mixture was stirred at room temperature overnight, quenched with H₂O, extracted with CH₂Cl₂. The combined organic extracts were washed with brine, dried over Na₂SO₄. The solvent was removed and the residue was purified by flash column (hexanes: ethyl acetate = 1:1, R_f = 0.20) to afford 1.04 g (84%) of compound 6 as a yellow solid. Calc’d for C₂₈H₃₀FN₃O₅S₂: 571.1; found [M+H]⁺: 572.0;. ¹H NMR (400 MHz, CDCl₃) δ (ppm) 2.93–2.95 (m, 4H), 3.13–3.15 (m, 4H), 3.78–3.83 (m, 8H), 4.05 (s, 2H), 7.07–7.09 (m, 3H), 7.19–7.20 (m, 3H), 7.30–7.33 (m, 2H), 7.42–7.45 (m, 1H), 7.58–7.60 (m, 2H), 8.92 (br s, 1H); ¹⁹F NMR (400 MHz, CDCl₃) δ (ppm) −111.6.

Step 6

A mixture of [bis(trifluoroacetoxy)iodo]benzene (PIFA, 1.02 g, 2.36 mmol) and TFA (0.415 g, 3.64 mmol) in CH₂Cl₂ (18 mL) was cooled to 0 °C under atmosphere of N₂. A solution of compound 6 (1.04 g, 1.82 mmol) in CH₂Cl₂ (35 mL) was added dropwise. The reaction mixture was stirred at 0 °C for 30 min, room temperature 30 min, and then refluxed for 40 h. The crude product was concentrated and the residue was purified by flash column chromatography (hexanes : ethyl acetate = a gradient of 1:1 to 1:4) and then by preparative HPLC to afford 343 mg (40%) of ML345 as a white solid. Calc’d for M+H = C₂₁H₂₃FN₃O₅S₂: 480.1; found [M+H]⁺: 479.9; ¹H NMR (400 MHz, CDCl₃) δ (ppm) 2.98–3.01 (m, 8H), 3.67–3.71 (m, 8H), 7.11 (d, J=8.8 Hz, 1H), 7.36–7.41 (m, 1H), 7.48–7.52 (m, 1H), 7.63 (dd, J=2.4, 8.4 Hz, 1H), 7.70 (dd, J=2.4, 8.0 Hz, 1H), 7.84 (d, J=2.0 Hz, 1H); ¹⁹F NMR (400 MHz, CDCl₃) δ (ppm) −115.4.

3.1. Summary of Screening Results

An overall summary of the screening campaign in shown in Figure 7. Briefly, uHTS primary screening of ~325,000 compounds gave 1,316 primary IDE INH hits. Counterscreening for toxicity and for (artifact) fluorescence interference gave 139 hits of highest interest. 127 available compounds were profiled in dose-response format for cell-based IDE activity and a total of 43 non-toxic compounds were then considered for follow-up testing of independent, freshly prepared powder samples by the SRIMSC. All results are available in PubChem using the indicated AIDs.

Figure 7

Screening campaign overview.

The series of efforts used to select probe candidate scaffolds and ultimately to identify molecular probes is summarized in the general hit-to-probe optimization flow chart (Figure 8). Chemistry and cheminformatics-based analyses of the results of the μHTS campaign allow for reactive, non-selective, assay-promiscuous, and otherwise non-tractable hits to be removed from further consideration.

Figure 8

Probe development strategy, post-uHTS.

Following these analyses, the most tractable compounds are ordered through for confirmation screening of a second batch. Any non-repeating samples are ignored for further probe development, as are compounds lacking targeted activity in secondary assays. Confirmed hits then allow for selection of scaffolds for probe development, aided by computational analysis of active and inactive analogs from the uHTS effort, to gauge potential of scaffolds to have meaningful non-flat SAR. Iterative rounds of medicinal chemistry and SAR development, complemented by assessment of compound properties, aids in selection of a high-quality probe.

The most interesting IDE screening hits are shown in Table 3 below. The screening hit selection criteria include structural desirability, high potency in IDE Inhibition dose-response (cell-based EC₅₀ <10 μM), high maximal degree of inhibition (>75%), and lack of activity in the HEK toxicity and fluorescence artifact counterscreens (toxic and non-selective agents were excluded from probe development consideration).

Table 3

Hit summary, promiscuity analysis (reflects PubChem data available at time of uHTS).

Each of the compounds shown in Table 3 also displayed IDE activity when an independent batch was assayed (these “powder confirmation” EC₅₀ values were within 2-fold of the uHTS result). All new batches of compounds were also inactive in the HEK toxicity counterscreen and also in the fluorescence artifact counterscreens (EC₅₀ or IC₅₀ in each case >>10 μM).

Our initial focus was upon three scaffolds shown in the middle of Table 3:

• triazoles (CIDs 667188 and 664941),
• the thiazole amide CID 1483690, and
• the diacylpiperazine CID 17583130.

Analogs in each series are synthetically readily accessible. Further, literature analysis suggested that each series was tractable. We also preferred hits that were expected to be non-covalent inhibitors. Nine close structural analogs of the triazole CID667188 were prepared. However, seven of these analogs were found to be completely inactive vs. IDE, while the remaining two compounds were less potent than were the original hits. Even minor structural changes tended to eliminate all IDE activity. Poor water solubility was also a concern in the triazole series (e.g.; the cLogP for CID667188 is ~5.0). For these reasons, non-triazole probe development scaffolds were instead pursued. Seven structural analogs of the thiazole amide CID 1483690 were prepared and all were found to be completely inactive vs. IDE, prompting us to also discontinue efforts in this chemical series.

We then focused our attention upon the diacylpiperazine CID 17583130, initially encouraged by the simple tractable structure, IDE inhibitory activity of this compound, and by similar potency observed for several selected analogs that were purchased from a commercial vendor. However, when a sample of CID 17583130 was prepared internally (SID 124756577), this analytically pure sample was completely inactive vs. IDE (EC₅₀ >100 μM). HPLC traces showed that the internally-generated sample was essentially identical in composition, at least by UV absorption, to both purchased samples of the “same” compound (SID 49825176 and SID 117689475). 21 structural analogs of CID 17583130, prepared internally in high purity, were also found to be inactive, though several purchased analogs were at least weakly active. When the purchased (previously active!) sample of CID 17583130 was rigorously purified by preparative HPLC (given SID 134420263 after purification), all IDE inhibitory activity was lost.

The above series of observations indicates that a non-organic and non-UV-active impurity is responsible for the IDE inhibition observed for batches of diacylpiperazines tested as purchased, without purification, samples that nevertheless had passed LCMS purity criteria owing to the absence of secondary peaks in the UV spectrum. We feel that this disappointing series of false positive results is likely due to the high sensitivity of the zinc or thiol moieties in IDE toward trace impurities that were present in the commercial samples as obtained from the vendor. Efforts to identify the active impurity or impurities have been unsuccessful. Thus the diacylpiperazine series was also eliminated from further consideration as a probe class.

The two hits shown at the bottom of Table 3, CID 4869155 and CID 2574532, were also eliminated from consideration for follow-up due to undesirable properties. Specifically, the high molecular weight and high LogP of CID 4869155, as well as the lack of active analogs in the uHTS collection, were concerns. The rhodanine-like hit CID 2574532 was also considered undesirable for chemical tractability reasons.

We expected that the four compounds shown at the top of Table 3, in a structural class termed the benzoisothiazolidones, were acting as covalent reversible modifiers of a cysteine residue at the active site of IDE through the formation of a disulfide bond. This feature distinguishes the class from the other entries in Table 3, which were expected to be non-covalent inhibitors, but that had, in every instance, proved ill-suited for probe development. This proposed mechanism of action for inhibition by the benzoisothiazolidones is illustrated in Figure 9.

Figure 9

Benzoisothiazolones target a cysteine residue of IDE.

Notably, the most potent HTS hit, CID 2325815 (top entry in Table 3) has a morpholine group present at the R² position (see the highlighted R² group depicted in Figure 9). Were these benzoisothiazolidone inhibitors merely acting as non-selective and chemically reactive “traps” for a cysteine residue, without significant cooperative binding to the enzyme, such a bulky and electron-donating R² group would be expected to instead decrease the ability of the ligand in interact with IDE. Further, we found non-flat SAR found in this series, first as determined by cheminformatics analysis of uHTS results (percent IDE inhibition results for structural analogs of CID 2325815 that were present in the screening collection), and then verified by dose-response studies of several of these analogs (SAR data is shown later in section 3.4). These two lines of evidence suggested that there would be a high potential for finding a potent and selective IDE inhibitor probe in the benzoisothiazolidone scaffold.

When a freshly prepared second sample of CID 2325815 was obtained (SID 104144357) this analytically pure powder confirmation sample nearly duplicated the uHTS cell-based results, as shown in Figure 10, with EC₅₀ ~2 μM (green curve, comparable with EC₅₀ = 1.4 μM found for SID 24840227 in the uHTS campaign). Moreover, we saw promisingly high potency in an IDE biochemical assay (IC₅₀ ~600 nM, black curve). Further, no HEK toxicity was seen at any dose (red curve) and no artifact fluorescence interference was seen below 10 μM (blue curve). This favorable profile supported the selection of the benzoisothiazolidones as a probe series.

Figure 10

Profile of most potent uHTS hit, CID 2325815. Data is for SID 104144357. Black = cell-free IDE assay

The use of covalent modifiers as molecular probes, and even as drugs, has been an area of considerable interest and substantial ongoing debate[27]. The stable covalent complex can confer favorable properties associated with long target residence time. Selectivity for the target in question (in this instance IDE) over other members of the proteome having reactive cysteine residues is of higher concern, however. A second concern is possible immunogenic effects of a covalently modified protein target, though this concern is lessened when the covalent modification is reversible, as in the case of benzoisothiazolidones. Our initial probe development plan expressed a preference for development a non-covalent probe, for the reasons listed above. In this case, however, we were thwarted by repeated difficulty in finding non-covalent hits that were not assay artifacts.

Keeping the concerns over covalent modifiers in mind, while also factoring in the potency, counterscreen selectivity, and high maximal inhibition seen in the benzothiazolidone class (see Figure 10 above) we chose this series for probe optimization. We ultimately identified even more potent IDE inhibitors (shown later in section 3.4, SAR table) and found a probe with suitable target selectivity and potency, the declared probe ML345, a fluorinated analog of the screening hit CID 2325815 (Figure 11).

Figure 11

ML345 is a fluorinated analog of CID 2325815.

3.2. Dose-Response Curves for Probe

We found that other benzoisothiazolones that are related to the screening hit CID 2325815 maintain the same general profile shown above in Figure 10: highest potency in the biochemical assay, followed by the cell-based IDE assay, with minimal autofluoresence at inhibitory levels, and no inherent toxicity in the HEK counterscreen.

The IDE inhibition curves are shifted to the left when the benzoisothiazolone ring bears an appropriately-positioned electron withdrawing group, consistent with the covalent reactivity of the probe as discussed earlier and depicted in Figure 9. Figure 12 below shows the inhibition curve for ML347 in the biochemical (cell-free) assay, using DTT-free, wild-type, recombinant IDE. The IC₅₀ measured in the Leissring lab was 188 nM (SID 136920229). When a larger, equally pure probe batch was tested at a later date (SID 144241486), its IC₅₀ was measured at 1.1 μM. This slight discrepancy may be due to a variation in activity of different batches of IDE used in each assays (positive controls were also higher in the latter run). The IDE enzyme is quite sensitive to handling and purification methods, showing significant batch-to-batch variability. The lower-end values for the probe (~200 nM) are more typical and (we feel) more accurate, particularly considering the positive control was on outlier in the experiment giving the higher IC₅₀ value. Nevertheless, the probe is a potent IDE inhibitor in biochemical assays, is only active vs. native IDE and not a mutant form lacking Cys819, and is a significant advance in the field given the absence of small molecule prior art.

Figure 12

Probe dose-response curve, IDE inhibition (AID 624065, SID 136920229).

3.3. Scaffold/Moiety Chemical Liabilities

We have calculated various chemical descriptors for the probe using the Accelrys Pipeline Pilot software and by applying standard “rule-of-five” and other lead-likeness or drug-likeness criteria[28–30]. As shown in Figure 13 below, all calculated parameters meet these criteria. The probe has a moderately cLogP and modestly high polar surface area, with suitable molecular weight, H-bond donor/acceptor counts, and with no significant structural-based concerns from a stability, lead-likeness, drug-likeness, or general toxicity alert[31] perspective.

Figure 13

Lipiniski and other drug-likeness parameters for ML345.

The calculated partition coefficients shown in Figure 13 for ML345 (e.g., LogD_7.4 = 2.0) suggests the potential for significant aqueous solubility that could aid its utility as a probe. As was described earlier, the solubility of the probe in PBS was somewhat lower than anticipated, 0.4 μM. The solubility in simulated general assay buffer was more reasonable (5.1 μM) suggesting that formulation strategies, with solubilizing additives, should be considered by researchers using the probe. Also, as shown earlier (Figure 5), the probe is chemically stable in a PBS solution: neither a substantial erosion of parent LCMS peak intensity nor an emergence of new peaks was seen over a 48-hour study.

The potential for biological stability was assessed by incubation of the probe with hepatic microsome preparations (1 mg/mL) using known methods[26]. The probe is highly stable to liver microsomes, with a measured half-life of >120 minutes (human liver microsomes), 24 minutes (mouse microsomes), and 50 min (rat microsomes). High hepatic microsome stability is consistent with attributes of a compound intended to be used in the study of relevant in vivo effects.

Inhibition of cytochrome P450 enzymes (CYPs) is a marker for drug-drug interaction potential. An in vitro study used four relevant CYP isoforms (1A2, 2C9, 2D6, and 3A4). The appropriate positive controls, compounds known to inhibit individual CYP isoforms (furafylline, sulfaphenazole, quinidine, and ketoconazole), were used.[26] The analysis, shown in Table 4, determined that ML345 inhibits these CYP isoenzymes between 57–94% at 10 μM. This suggests that ML345 requires significant improvements with regard to CYP450 inhibition profile for safe use therapeutically; however, its profile should not hinder the utility of ML345 as a molecular probe and biological tool compound to study the importance of IDE, especially since it is a first-in-class small molecule inhibitor of IDE.

Table 4

CYP450 Inhibition of Probe ML345.

The selection of a covalently-reactive probe series, as discussed previously, demanded that we clearly show that the probe does not behave simply as a broadly reactive and non-selective cysteine trap. We have compiled several lines of evidence, including non-flat SAR (shown later SAR discussed later in Section 3.4) and profiling assays (discussed later in Section 3.6) to show that the probe has suitable target selectivity to be useful as a molecular probe.

The probe scaffold is termed a benzoisothiazolidone. We are not aware of any undesirable attributes of the scaffold, other than potential reactivity with cysteine-containing proteins (an issued addressed in Section 3.6, assay profiling) and the CYP450 inhibition we have observed, which is likely compound-specific rather than scaffold-specific. While in vivo studies are beyond the scope of the probe development effort, the electron-deficient aryl groups present in ML345 would tend to protect it from susceptibility to rapid oxidative metabolism. The sulfonamide group of ML345 is not normally prone to redox issues. The aniline moiety present is tri-substituted, sterically hindered, and electronically deactivated, thus this group is also unlikely to contribute to potential metabolic liabilities for the probe.

The relative ease of synthesis of the probe molecule and analogs by a convergent six step sequence (4 linear steps) in a reasonable yield (see Figure 6 in Section 2.3) is a strong asset for this scaffold and for this probe. Scale-up synthesis, if needed for extensive long-term studies, will be highly feasible.

An MLPCN probe with the benzoisothiazolidone scaffold has already been approved. This compound, ML089, is a potent, selective, and orally bioavailable phosphomannose isomerase inhibitor (IC₅₀ = 1.3 μM)[32]. This same compound (entry 3 in Table 5, shown later in Section 3.4) is also an IDE inhibitor, though it is ~20-fold less potent vs. IDE than is ML345. Very importantly, however, the reported highly favorable DMPK properties of ML089, including high oral bioavailability[32], support the assertion that analogs tailored for potent and selective IDE inhibition may also have similarly favorable DMPK properties.

Table 5

Abbreviated SAR summary in the probe scaffold showing cell-free (biochemical) results.

3.4. SAR Tables

After the team selected CID 2835215 as a probe development scaffold, we investigated structure-activity relationships (SAR) of the series that were apparent from uHTS results, since several related compounds were in the MLSMR library and had been screened. We also analyzed SAR of purchased analogs as well as related inhibitors synthesized internally. IDE inhibitory activity was measured in triplicate, using the cell-free IDE inhibition assay, and selected results are shown above in Table 5 (values for three other analogs present in the uHTS are shown in Table 3 and are not duplicated here). The screening hit and the eventual probe ML345 are shown at the top (entry 1) and bottom (entry 7) of Table 5, respectively, for comparison.

Briefly, entry 1 (the uHTS hit) has suitable activity for SAR development (The Scripps IC₅₀ value was ~600 nM, as shown earlier in Figure 10, with EC₅₀ in the cell-based assay = 1.4 μM; values shown in Table 5 are from the cell-free assay, measured in the Leissring lab).

Entry 2 shows that deletion of both the morpholine and morpholine sulfonamide groups present in the uHTS hit diminishes IDE activity more than 10-fold, suggesting that both of these groups play a role in IDE recognition.

Entries 3 and 4 show that electron withdrawing groups in either aryl ring enhances IDE inhibition (compare with entry 2).

Entry 5 shows that a pyrrolidine ring is less well-tolerated than is the morpholine ring present in the uHTS hit, again suggesting that morpholine group plays a important role in IDE recognition.

Entry 6 shows the very weak activity of a prototypical non-selective covalent “cysteine trap” N-ethyl maleimide.

Finally, the probe compound ML345 is shown in entry 7, which, like entry 3, bears a para-fluorine substituent to further augment IDE inhibitory activity (as suggested by entry 3). IDE inhibition is also aided by the presence of the morpholine ring and the sulfonamide group (as suggested by entries 3 and 4).

We draw another important conclusion from analysis of Table 5: the SAR in the benzoisothiazolidone series is not flat, i.e.; all compounds that are capable of interacting with a cysteine residue do not have roughly equal activity.

It is also noteworthy that the ortho morpholine group present in entries 1 and 7 enhances activity, since both the electron-donating effect of this group, as well as the steric hindrance it imparts, would deactivate the sulfur atom in the benzoisothiazolidone ring for nucleophilic attack. This again indicates that the morpholine ring present in the probe plays a positive role in IDE recognition that overrides its deleterious effects upon ligand reactivity.

3.5. Cellular Activity

The uHTS primary assay is cell-based, so all hits (and the probe) demonstrate cellular activity. As shown in Figure 10 for the uHTS hit, cell-based EC₅₀ is typically within 2- to 5-fold of biochemical IC₅₀.

3.6. Profiling Assays

In order to ascertain the selectivity of ML345 for IDE over other enzymes in the proteome with reactive cysteine residues, a proteome-wide activity-based protein profiling (ABPP) assay was conducted in collaboration with Ben Cravatt and Andrea Zuhl at TSRI-La Jolla. These assays were performed with the general cysteine-reactive probe chloroacetamide-rhodamine (CA-Rh)[33–35]. Briefly, the probe compound was incubated for 30 minutes at 37 °C in HEK293T proteomes then labeled for 30 minutes with 5 μM CA-Rh at 25 °C. Analysis by SDS-PAGE and in-gel fluorescence scanning gave the results shown below in Figure 14.

Figure 14

Proteome-wide activity-based protein profiling for cysteine reactivity of ML345.

When ML345 was added to either membrane (A) or soluble (B) proteomes at concentrations of 2.0 μM or below, it did not inhibit labeling of cysteine-reactive proteins as determined by the retention of fluorescence signal compared to a DMSO control lane. For both proteome fractions, ML345 was significantly reactive at 20 μM, as indicated by bands that do not appear in that lane using the highest concentration of test compound. Overall these results show that the probe is not a broadly reactive “cysteine trap” and that it likely has on the order of at least 10-fold selectivity for IDE in a whole cell environment.

Ideally the proteome-wide profiling assay would be complemented by profiling assays, especially including thio-reactive enzymes. Such assays much be thiol-free to have meaningful results, however, and a suitable thiol-free “ready to go” panel was not available from screening vendors. We hope to gather profiling data against related targets in the future, however.

4.1. Comparison to existing art and how the new probe is an improvement

Prior art compounds that act as inhibitors of IDE are known, but none are both potent (IDE IC₅₀ or EC₅₀ <1 μM) and truly selective for IDE (Figure 15). These compounds include the cyclic peptide bacitracin (which has very low potency), chloroquine, 4-chloromercuribenzenesulfonate (aka PCMBS; CID11136; a highly cytotoxic compound)[36], and related p-chloromercuribenzoic acid (PCMBA). Weak and non-selective thiol-alkylating agents such as N-ethyl maleimide (NEM) and 2-iodoactetamide are known, as are weak and non-selective general zinc chelators such as 1,10-phenanthroline and EDTA[37]. IDE can also be inhibited in a dose-dependent manner with excess insulin[37,38]. Long-chain fatty acids and related acyl-CoA thioesters, such as linoleoyl-CoA and palmitoyl-CoA, act as IDE inhibitors with in vitro IC₅₀ values ~10 μM and above[39]. ATP also inhibits IDE activity in vitro[37]. The protease inhibitor nelfinavir has been reported to inhibit insulin degradation mediated by partially purified IDE in cell-based assays; however, the IC₅₀ was found to be greater than 100 μM[39].

Figure 15

Prior art compounds. Exceptionally weak, non-selective, and non-drug-like prior art IDE inhibitors are not suitable starting points for small molecule probe development.

Recently, and subsequent to the initiation of this project, a peptide hydroxamic acid inhibitor of IDE that is highly potent (K_i for insulin ~23 nM) was developed by the assay submitter, Dr. Leissring[5]. However, this compound suffers from several limitations. First, due to its peptidic nature its utility in vivo is limited, which also limits its prospects for development into a more drug-like probe. Second, it is not a small molecule, with a molecular weight ~750 Daltons. Finally, the hydroxamic acid group present in this IDE inhibitor is a promiscuous zinc-binding moiety which will likely cross-react with other zinc-metalloproteases, thereby limiting its specificity.

From the above analysis, the potent and selective small molecule IDE inhibitor ML345 is well-differentiated from existing art and will be a significant advance in the field.

4.2. Mechanism of Action Studies

We made significant efforts to show precisely how benzoisothiazolidones, including the probe ML345, interact with IDE. Much of this work was performed using the uHTS hit, CID 2325815. This compound and its analogs similarly studied display activity in the cell-based IDE inhibition assays (AIDs 434962, 435028, 463220, and 588712) and also consistently have higher activity in related cell-free biochemical assays (AIDs 688711 and 624067). Typically a two- to ten-fold cell-based shift is seen, typical of the barrier to cell penetration seen for most effective on-target small molecules.

We were interested in testing the proposed covalent mechanism of action of this series of compounds (see Figure 9, in section 3.1). Toward this end, mutant forms of IDE that have alanine replacements for either Cys-819 or Cys-812 were prepared[25] and used in IDE inhibition assays. The benzoisothiazolidone screening hit CID 2325815, an inhibitor of the native IDE enzyme with an IC₅₀ = 1.4 μM (Table 5) does not inhibit mutant IDE lacking Cys-819 (IC₅₀ >100 μM), though this compound is an inhibitor of the mutant enzyme lacking Cys-812 but with Cys-819 intact (IC₅₀ ~3 μM). This shows that benzoisothiazolidones are not indiscriminately reactive, since they can distinguish between these two cysteine residues of IDE, preferentially reacting at Cys-819 to inhibit the enzyme.

The covalent mechanism of action for ML345 was further confirmed by using electrospray ionization-mass-spectrometry (ESI-MS) to show that ML345 bonds covalently to a thiol-containing cysteine analog present in large excesss (N-acetylcysteine), yielding a product of the predicted mass (625.2 amu).

Finally, the inhibition of IDE by ML345 was shown to be reversible by addition of reducing agents, such as ß-mercaptoethanol or dithiothreitol (data not shown).

These experiments, along with non-flat SAR (Table 5 in Section 3.4), and combined with the results of the activity-based protein profiling experiment (Figure 14 in section 3.6) show that potent benzoisothiazolidones are direct, covalent, and reversible inhibitors of wild-type IDE that act preferentially at Cys-819 in the internal cavity of IDE.

4.3. Planned Future Studies

Professors Leissring, Bannister, and collaborators intend to extend and expand upon this work through future grant proposals. Such efforts, if funded, will aim to develop a diverse set of potent, selective, IDE inhibitors, including zinc-targeting IDE inhibitors, more potent cysteine-targeting IDE inhibitors, and compounds in each class that are particularly suited for animal use, while also further studying mechanistic aspects of IDE inhibition by these compounds.

4.4. Acknowledgement

We would like to gratefully recognize the assistance of Dr. Andrea Zuhl and Professor Benjamin Cravatt in obtaining the activity-based protein profiling data, for helpful discussions, and for preparing the relevant figure ABBP data (Figure 14 in Section 3.6).