The mode of action of the Protein tyrosine phosphatase 1B inhibitor Ertiprotafib

Protein tyrosine phosphatase 1B (PTP1B) is a validated therapeutic target for the treatment of diabetes and obesity. Ertiprotafib is a PTP1B inhibitor that reached the clinical trial stage for the treatment of diabetes. Interestingly, Ertiprotafib reduces the melting temperature of PTP1B in differential scanning fluorimetry (DSF) assays, different from most drugs that increase the stability of their target upon binding. No molecular data on how Ertiprotafib functions has been published. Thus, to gain molecular insights into the mode of action of Ertiprotafib, we used biomolecular NMR spectroscopy to characterize the molecular details of the PTP1B:Ertiprotafib interaction. Our results show that Ertiprotafib induces aggregation of PTP1B in a concentration dependent manner. This shows that the insufficient clinical efficacy and adverse effects caused by Ertiprotafib is due to its tendency to cause aggregation of PTP1B.


Introduction
Protein tyrosine phosphorylation is a key post-translational modification that plays essential roles in cell growth, cell differentiation, cell-cycle regulation and the immune response [1,2]. The level of protein tyrosine phosphorylation is maintained by the concerted action of protein tyrosine kinases (PTKs;~90 members) and protein tyrosine phosphatases (PTPs;~107 members). Aberrant protein tyrosine phosphorylation due to imbalances in the activities of PTKs and PTPs has been implicated in numerous human diseases including cancer and diabetes [3,4]. Hence, targeting the activities of PTKs and PTPs have emerged as a promising avenue for the development of drugs to cure cancer and diabetes [5][6][7][8][9].
Protein tyrosine phosphatase 1B, the founding member of the PTP superfamily, is critical for the regulation of insulin signaling as it dephosphorylates both the insulin receptor (IR) and the insulin receptor substrate (IRS) [10][11][12]. Hence, PTP1B is an exceptionally well-studied target for the treatment of diabetes and obesity [13]. We and others have recently shown that PTP1B is a highly regulated enzyme that can be inhibited by either active site or allosteric inhibitors [14,15]. While its core catalytic functions are driven by rigid conformational changes, allostery in PTP1B is controlled by conformational and dynamic changes, especially in a PTP1B specific secondary structure element, helix α7 [16]. A large number of PTP1B inhibitors have been developed over last two decades [17][18][19]. Most of these inhibitors target the catalytic site and are mimics of the phosphotyrosine (pTYR) moiety. However, the catalytic site of PTP1B is both conserved and charged. Thus, PTP1B active site inhibitors are often not selective nor are they cell permeable. More recently, a number of allosteric inhibitors have been developed that target an allosteric binding pocket that lies at the intersection PTP1B helices α3, α6 and α7 [15,20]. Thus far, only two small-molecule PTP1B inhibitors, namely Ertiprotafib [21] and Trodusquemine (also known as MSI-1436) [22] have successfully reached clinical trials. We and others have recently shown that Trodusquemine inhibits PTP1B activity in a unique fashion, namely by binding to the intrinsically disordered C-terminus of PTP1B, demonstrating it functions in an allosteric manner [15].
Ertiprotafib is a monocarboxylic acid pTyr mimetic and thus it was assumed that it binds at the PTP1B active site (Fig 1a). Interestingly, Ertiprotafib inhibits both PTP1B and the dual peroxisome proliferator-activated receptor (PPAR) alpha and gamma agonist with hypoglycemic and anti-lipidemic activity [23]. Ertiprotafib was in phase II clinical studies for the treatment of type 2 diabetes in both the USA and Europe, but was discontinued due to unsatisfactory phase II clinical efficacy and dose-limiting adverse effects in some patients [21]. Although Ertiprotafib is no longer under investigation, it might provide a platform for the development of other PTP1B inhibitors based on its affinity, acceptable selectivity and oral availability. Strikingly, in differential scanning fluorimetry (DSF) assays, which are commonly used as primary drug screen, Ertiprotafib lowered the melting temperature of PTP1B [24,25]. This is different to most drugs that bind tightly to their targets, resulting in an increase in their melting temperatures. This suggests Ertiportafib may have an atypical mode-of-action for PTP1B inhibition. We set out to determine the molecular basis for the Ertiprotafib interaction with PTP1B in order to determine if further development of this class of inhibitor is warranted.
Here we use solution state bimolecular NMR spectroscopy to determine the mode of binding to and inhibition of PTP1B by Ertiprotafib. Unexpectedly, our data show that Ertiprotafib binds non-specifically to the catalytic domain of PTP1B, ultimately leading to PTP1B aggregation. Thus, Ertiprotafib inhibits the activity of PTP1B by inducing aggregation and not by either directly blocking the active site or inhibiting PTP1B allosterically. This mode-of-action also explains the reduction of the PTP1B melting temperature and highlights that destabilizers identified in DSF screens should be tested for their molecular modes-of-action prior to extensive efforts for compound optimization.
Cell pellets were resuspended in Lysis Buffer (50 mM Tris pH 8.0, 500 mM NaCl, 5 mM imidazole, 0.1% Triton X-100) containing EDTA-free protease inhibitor cocktail tablets (Roche) and lysed using a high-pressure homogenization (Avestin). The bacterial lysate was clarified by centrifugation at 45,000 xg and filtered through a 0.22 μm filter before loading onto a His-trap HP column (GE Healthcare). Bound proteins were washed with Buffer A (50 mM Tris pH 8.0, 500 mM NaCl, 5 mM imidazole) and eluted with increasing amounts of Buffer B (50 mM Tris pH 8.0, 500 mM NaCl, 500 mM imidazole) using a 5-500 mM imidazole gradient. Peak fractions were pooled and dialyzed overnight at 4˚C in Dialysis Buffer (50 mM Tris pH 8.0, 500 mM NaCl) with 5:1 volume ratio of TEV protease overnight. The next day, a subtraction His 6 -tag purification was performed to remove the TEV protease and the cleaved His 6 -tag. Cleaved PTP1B was concentrated and further purified using Size Exclusion Chromatography (SEC, Superdex 75 26/60 [GE Healthcare]) in Assay/DLS/NMR Buffer (10 mM HEPES pH 7.4, 150 mM NaCl, 0.5 mM TCEP). Purified PTP1B was either used immediately or flash frozen in liquid nitrogen for storage. Typical PTP1B protein yields were~45 mg/L in Luria Broth and~35 mg/L in 2 H/ 15 N M9 medium.

DLS measurements
Dynamic light scattering (DLS) data for PTP1B 1-301 and PTP1B 1-393 with increasing molar ratios of Ertiprotafib (0, 2, 5, 10 and 15) were acquired using a Wyatt DynaPro Nanostar. DLS size distribution histograms were calculated with DYNAMICS 6.12 (Wyatt Technology Corporation). All of the proteins were measured in triplicate at a concentration of 1 mg/ml in 10 mM HEPES pH 7.4, 150 mM NaCl, 0.5 mM TCEP. Each experiment reflects the average of ten measurements, each of which was acquired for 10 s at 25˚C.

pNPP activity assay
The activities of PTP1B 1-301 and PTP1B 1-393 were measured in assay buffer containing varying concentrations of p-nitrophenyl phosphate (pNPP; 0 to 6 mM final concentrations) with increasing concentrations of Ertiprotafib (0 μM, 0.16 μM, 0.4 μM, 0.8 μM and 1.2 μM). Enzymes were incubated with the substrate at 30˚C for 30 min. The reaction was stopped using 1 M NaOH and the absorbance was measured at 405 nM using a plate reader (BioTek). Ertiprotafib stock solution (25 mM) was prepared in 100% DMSO, resulting in 0.1-0.8% of DMSO in the reactions. Measured absorbance from blanks that contained substrate and inhibitors but no protein was subtracted from all measurements. The rate of dephosphorylation of pNPP was analyzed using the molar extinction coefficient for pNPP of 18000 M -1 cm -1 and an optical length of 0.3cm (96 well plates). K m and V max were determined by fitting to the Michaelis-Menten equation, y = V max � x/(K m +x); k cat was extracted using y = E t � k cat . The catalytic efficiency was obtained as k cat/ K m . Sigmaplot 13.0 was used for data analysis. The experiments were carried out in triplicate.

Molecular basis of PTP1B:Ertiprotafib interaction
Our inhibition data is inconsistent with previously published data that suggests Ertiprotafib is an active site inhibitor (IC 50 >20 μM) of PTP1B [23]. The main motivation for this mode of action was computational docking of Ertiprotafib to a 3D structure of the catalytic domain of PTP1B [29]. Thus, we used biomolecular NMR spectroscopy to experimentally determine the molecular mode of the Ertiprotafib:PTP1B interaction. Unexpectedly, a direct comparison of the 2D [ 1 H, 15 N] TROSY spectrum of free and Ertiprotafib-bound PTP1B 1-301 revealed only very small non-statistically meaningful chemical shift perturbations (CSPs; direct indicator of changes in chemical environment upon binding; Δδ av = 0.01 ± 0.01 ppm; S1a and S1b Fig; top  panel). This was surprising as other PTP1B inhibitors, such as TCS-401 or CPT157633, show large, statistically significant CSPs in residues that define the PTP1B active site [14,16]. Furthermore, these minor CSPs did not change in magnitude with increasing concentrations of Ertiprotafib, as would be expected for a specific interaction. Finally, multiple PTP1B peaks in the 2D [ 1 H, 15 N] TROSY spectrum showed line broadening upon addition of Ertiprotafib and this behavior became more and more apparent at higher PTP1B:Ertiprotafib molar ratios. In particular, at high ratios (1:10), nearly all the peaks of PTP1B 1-301 disappeared (Fig 2a). The simplest interpretation of this data is that PTP1B 1-301 aggregates upon the addition of increasing amounts of Ertiprotafib.

The PTP1B C-terminal tail does not contribute to the action of Ertiprotafib
We repeated these NMR experiments using PTP1B  . As for PTP1B 1-301 only minor CSPs were detected (S1b Fig, bottom panel and S1c Fig). Further, a small number of residues in the IDR region showed CSPs including L317, V334, I346, S352, A357, Y359, V374 and A385 (Fig  2b, top panel). As these are mostly hydrophobic residues, we speculated that these might be very weak non-specific interactions. Consistent with the data for PTP1B 1-301 , at high Ertiprotafib ratios (1:15), all peaks belonging to PTP1B 1-301 residues disappeared (Fig 2c). Residues of the PTP1B IDR C-terminus (residues 302-393) showed increased intensity upon binding to Ertiprotafib (Fig 2b; middle and bottom panel). This increase in intensity for residues in the IDR C-terminal tail shows that this disordered region is highly dynamic and behaves as an independent moiety from the PTP1B catalytic domain, resulting in faster tumbling and higher intensity peaks than those of the PTP1B catalytic domain. Together, these data show that Ertiprotafib induces aggregation of PTP1B 1-301 independent of the IDR C-terminus and, further, that the nonspecific interaction of Ertiprotafib with the C-terminal residues is not important for Ertiprotafib function. This contrasts with our study of PTP1B with Trodusquemine, where we did not see aggregation when we performed similar experiments [15].

Ertiprotafib induced aggregation of the PTP1B catalytic domain
To confirm that PTP1B 1-301 and PTP1B 1-393 aggregate upon Ertiprotafib binding, we used dynamic light scattering (DLS). DLS measurements report the radius of hydration (R H ) of 25.7 Å ± 0.6 Å and 35.7 Å ± 0.6 Å for PTP1B 1-301 and PTP1B 1-393 , respectively, consistent with previously reported data [15]. However, PTP1B 1-301 and PTP1B  showed exponential increases in their radii of gyration with increasing concentrations of Ertiprotafib (Fig 3a and 3b). These data not only confirm that PTP1B aggregates in the presence of Ertiprotafib but also that the  size of the aggregates increases with increasing concentrations of compound. This behavior was confirmed by measuring the melting temperature (T m ) of PTP1B in the presence of differing concentrations of Ertiprotafib (Ertiprotafib is known as a 'destabilizer' in DSF screens). Different to commonly used DSF measurements, our measurements relied on intrinsic tryptophan fluorescence and not a hydrophobic dye that might lead to aggregation. At low Ertiprotafib concentrations, the T m decreases, similar to reported DSF data. In contrast, at high Ertiprotafib concentrations, the T m increases (Fig 3c and 3d). This is consistent with the fact that Ertiprotafib decreases the stability of PTP1B at lower concentrations but becomes increasingly stable as a result of extensive aggregation.

Discussion
Ertiprotafib is the first PTP1B inhibitor that made it into clinical trials. In DSF measurements, experiments commonly used in high throughput drug screening assays to identify compounds that bind protein targets, Ertiprotafib was discovered to be a 'destabilizer', as it decreased PTP1B thermal stability (T m ). This is in contrast to most inhibitors/drugs which increase the T m of their protein targets. To define the molecular basis of these observations, we used biomolecular NMR spectroscopy. Our data showed unequivocally that Ertiprotafib induces oligomerization/aggregation of PTP1B. It also showed, mechanistically, how this aggregation is achieved. Namely, Ertiprotafib induces aggregation of the catalytic domain of PTP1B, with the regulatory, intrinsically disordered C-terminal tail playing no significant role in this process. This contrasts with another small molecule inhibitor of PTP1B, MSI-1436, which binds specifically to the disordered C-terminal tail of PTP1B and does not lead to PTP1B aggregation [15]. Thus, of the PTP1B inhibitors molecularly characterized, the aggregation of PTP1B is unique and specific to Ertiprotafib. During the last two decades, multiple studies have shown that at least a subset of compounds identified primarily via high-throughput screening also 'inhibit' by aggregating their target proteins. For example, small molecular inhibitors that target β-lactamase, malarial protease, the insulin receptor and farnesyl transferases, among others, have also been shown to aggregate in solution and it is the aggregation that indirectly results in enzyme inhibition [30,31]. Our data clearly show that the failure of Ertiprotafib in clinical trials is due to its ability to aggregate PTP1B and hence, its failure to manipulate insulin signaling in a specific, PTP1Bdependent manner. Together, these results suggest that compounds identified to lower the target thermal stability, as observed in the presence of low concentrations of Ertiprotafib, should be tested for their molecular mode of interaction prior to extensive efforts for compound optimization.