Synthesis and evaluation of bifunctional PTP4A3 phosphatase inhibitors activating the ER stress pathway

Abstract

We developed JMS-053, a potent inhibitor of the dual specificity phosphatase PTP4A3 that is potentially suitable for cancer therapy. Due to the emerging role of the unfolded protein response (UPR) in cancer pathology, we sought to identify derivatives that combine PTP4A3 inhibition with induction of endoplasmatic reticulum (ER) stress, with the goal to generate more potent anticancer agents. We have now generated bifunctional analogs that link the JMS-053 pharmacophore to an adamantyl moiety and act in concert with the phosphatase inhibitor to induce ER stress and cell death. The most potent compound in this series, 7a, demonstrated a ca. 5-fold increase in cytotoxicity in a breast cancer cell line and strong activation of UPR and ER stress response genes in spite of a ca. 13-fold decrease in PTP4A3 inhibition. These results demonstrate that the combination of phosphatase inhibition with UPR/ER-stress upregulation potentiates efficacy.

Introduction

Protein tyrosine phosphatases (PTPs) are responsible for the removal of phosphate groups from biomolecules and, in combination with protein tyrosine kinases (PTKs), control phosphorylation homeostasis.¹ Aberrant phosphorylation can lead to an imbalance that is associated with many pathophysiological conditions, including metabolic diseases, inflammatory disorders, and cancer. Even though complementary in functionality, PTKs and PTPs have historically been perceived differently with respect to drug discovery. While many PTK inhibitors have achieved FDA approval, there are no clinically employed PTP inhibitors yet, contributing to the “undruggable” stigma for this class of enzymes.² At least for cancer, this can be attributed to a misconception that phosphatases act exclusively as tumor suppressors. Another issue is the difficulty in developing a small molecule inhibitor to bind in the shallow, positively charged active site typical of protein phosphatases.

Modulators of the unfolded protein response (UPR) have to-date faced a similar predicament as phosphatase inhibitors.³ While it is generally accepted that UPR signaling in response to endoplasmatic reticulum (ER) stress can dictate cell fate and is involved in numerous pathologies, including neurodegeneration, protein misfolding disorders, diabetes, ischemia, and cancer,^4,5 translational development of UPR and ER stress modulators has been lacking due to the dearth of selective small molecule drug candidates and viable pharmacologic strategies.^3,6

We have reported the synthesis of 7-iminothieno[3,2-c]pyridine-4,6 (5H,7H)-dione (1, JMS-053) through both batch and in-flow photo-oxygenation reactions.⁷ This novel compound was found to be a low-nanomolar inhibitor of the oncogenic, dual-specific phosphatase PTP4A3,⁸ as well as its family members PTP4A1 and PTP4A2.⁹ An initial SAR study on our lead compound focused on structural changes to the pendant aryl group, imide substitution, and conversion of the imine to a carbonyl group (Fig. 1, zone model). We determined that zone 1 tolerated para-alkoxy chains R¹ and zone 3 was able to accommodate small R² groups.^7b In contrast, switching the sulfur atom from Y1 to Y2 in zone 2 led to a 5-fold drop in PTP4A3 inhibitory activity. Collectively, the data suggest that 1 is a robust pan-inhibitor of PTP4A with many analogs retaining nanomolar potency.

Fig. 1.

From monofunctional PTP4A inhibitor chemotype to bifunctional ER stress inducer/phosphatase inhibitor.

A recent report suggested that PTP4A1 was an ER stress-inducible gene product.¹⁰ When a toxic ER insult is applied, PTP4A1 suppresses apoptosis, suggesting that the function of this phosphatase is connected to a mechanism that rescues cancer cells from death. Accordingly, we investigated if (1) PTP4A inhibitors were sufficient to induce ER stress, and (2) bifunctional molecules that combined a PTP4A inhibitor with an ER stress inducer showed enhanced potency in decreasing cancer cell viability. Therapeutic regimens are increasingly combining several active ingredients and bifunctional hybrid biomolecules such as the BiTEs (bispecific T-cell engagers, f. ex. blinatumomab), and some of the resulting compounds have been successfully translated into the clinic.¹¹ In contrast, traditional small molecule drug discovery efforts optimize drugs composed of a single pharmacophore.^12,13 By identifying and optimizing a dual-pathway agent at an early drug discovery stage that eliminates a compensatory mechanism for preventing cancer cell death, we hoped to generate a more effective drug lead for cancer therapy. Our SAR information on lead compound 1 suggested that attachment of an ER stress inducer (ERSI) at the para-alkoxy group in 2 was a feasible approach to generate an analog that retained potent PTP4A3 inhibitory activity (Fig. 1).

The highly lipophilic and chemically relatively inert adamantyl group has previously been found to induce acute ER stress resulting in transient UPR activation without direct induction of programmed cell death.¹⁴ This approach engages the vast network of molecular chaperones responsible for maintaining protein homeostasis.¹⁵ The adamantyl group has also been used to degrade endogenous oncogenic proteins, such as HER3¹⁶ and AR,¹⁷ by appending it onto known small molecule inhibitors of these targets.

Based on this precedence, we chose to employ the adamantyl moiety as the ERSI connected via a variable linker moiety “Z” to the core structure of JMS-053 (Fig. 1). The linker length and composition of 2 would likely affect protein binding, cell permeability and solubility characteristics,¹⁸ and we therefore allowed for some variation at this site, which would also provide useful control compounds and shed light on the SAR in the hybrid series.

Our initial target molecule 7 contained an aryl ether-based linker (Scheme 1). Its synthesis commenced with an amide bond formation from commercially available adamantyl acetic acid 3. This amide bond was found to be best constructed over a 2-step sequence where acid 3 was treated with neat SOCl₂ to generate the corresponding acid chloride followed by addition of 4-amino-1-butanol to give 4 in excellent yield.

Scheme 1.

a) SOCl₂, reflux; b) 4-amino-1-butanol, Et₃N, CH₂Cl₂, 0 °C to rt; c) 10, DIAD, PPh₃, THF, 0 °C to rt; d) 11, Pd(PPh₃)₄, K₂CO₃, dioxane/H₂O (5:1), reflux; e) H₂, Pd/C, dioxane/MeOH (2:1); f) air, 425 nm-LED, MeOH, in flow; g) NH₄OAc, MeOH, 60 °C.

A Mitsunobu reaction between 4 and phenol 10¹⁹ was used to form the aryl ether bond of 5. While in our previous SAR studies Suzuki coupling conditions (Pd(PPh₃)₄ and Na₂CO₃ in dioxane/H₂O (2:1)) at reflux temperature were successful for the segment condensation between the phenyl ring and the thiophene,^7b these conditions failed to effect coupling of 5 and 11. After some experimentation, we found that 5 could be successfully attached to thiophene bromide 11 by replacing Na₂CO₃ with K₂CO₃ and increasing the dioxane/H₂O ratio to 5:1. These conditions led to the formation of 6 in good yield. It is noteworthy that during the workup of the Suzuki reaction the crude material needs to be acidified in order to protonate the highly acidic nitro-pyridone moiety in 6, otherwise recovery of the product is low-yielding. Also, the Suzuki product 6 could be purified by a series of triturations, avoiding chromatography entirely.

The synthesis of the 4-amino-1-butanol linked conjugate 7 was completed using a 3-step sequence. Nitro compound 6 was first reduced to the corresponding amine by catalytic hydrogenation. The resulting enamine then underwent photooxygenation using our continuous-flow methodology.^7c During the aminothienopyridone photooxygenation reaction, a trione byproduct 7b is typically generated in varying amounts in addition to the major product, imine 7a, in the ratio depending on the substitutions on the phenyl ring. Though some imine/trione mixtures of JMS-053 analogs can be separated chromatographically, complete purification of the imine/trione mixture 7a,b was unsuccessful. Fortunately, an NH₄OAc imination reaction step converted the imine/trione mixture 7a,b exclusively into imine 7a. After filtering the crude photooxygenation material through a short plug of SiO2, the filtrate was stirred in a methanolic suspension of NH4OAc in a sealed tube at 60 °C. After re-filtration through a bed of SiO2, pure imine 7a could be isolated in moderate yield over 3 steps.

The corresponding ester analog (8) was prepared following a similar protocol but starting with a DCC-mediated esterification between 3 and 1,4-butanediol.²⁰ Notably, the synthesis of 8 did not require the NH4OAc imination step, as chromatographic separation of the crude imine/trione mixture after the in-flow photooxygenation provided pure imine 8. Additionally, a zone 3 N-alkylated analog (9) was synthesized in which the tethered adamantyl group was attached to the imide nitrogen of JMS-053.²⁰

In addition to the ether linkages in 7 and 8, we were also interested in conjugating JMS-053 with a more polar 2-hydroxyacetamide to the adamantane residue. This modification broadened the scope of our bifunctional analogs by: (1) allowing access to the wide variety of potential diamine linkers, and (2) avoiding the Mitsunobu reaction conditions which required multiple rounds of chromatography to purify the desired products from the excess of DIAD and its hydrazodicarboxylate byproduct. Therefore, phenol 10 was first alkylated with ethyl bromoacetate. While this reaction was high yielding, generating the carboxylic acid without hydrolyzing the boronic ester proved to be difficult. Instead, phenol 10 was alkylated with tert-butyl bromoacetate. The tert-butyl group was then cleaved by stirring in a solution of TFA in CH₂Cl₂, providing acid 12 in 71% yield over 2 steps.²⁰

Starting with N-Boc protected 1,3-diaminopropane, amide coupling with the acid chloride derived from 3 provided compound 13 in excellent yield (Scheme 2). The Boc group in 13 was removed with TFA, and the resulting primary amine was acylated with acid 12. Using our optimized Suzuki conditions, cross coupling of boronic ester 14 and thiophene bromide 11 cleanly generated intermediate 15 without the need for chromatography. Finally, nitro compound ₁₅ was transformed into conjugate ₁₆ by catalytic hydrogenation of the nitro group, in-flow photooxygenation of the resulting enamine, and imination of the imine/trione mixture with NH₄OAc. Following the same synthetic sequence, conjugate 17 was generated using 1,2-bis(2-aminoethoxy)ethane as the diamine starting material.

Scheme 2.

a) SOCl₂, reflux; b) N-Boc-1,3-diaminopropane, Et₃N, CH₂Cl₂, 0 °C to rt; c) TFA/CH₂Cl₂, 50 °C; d) 12, HATU, i-Pr₂NEt, THF, 0 °C to rt; e) 11, Pd (PPh₃)₄, K₂CO₃, dioxane/H₂O (5:1), reflux; f) H₂, Pd/C, dioxane/MeOH (2:1); g) air, 425 nm-LED, MeOH, in flow; h) NH₄OAc, MeOH, 60 °C.

We next examined the bifunctional PTP4A3 inhibitors in a biochemical enzyme assay (Table 1). While less potent than the parent monofunctional compound, JMS-053 (Entry 7, IC₅₀ = 84 nM) in these assays,^7b the conjugates retained different abilities to inhibit PTP4A3 enzymatic activity in vitro, with 2-hydroxyacetamide analogs 16 and 17 (Entries 5 and 6, IC₅₀ = 100–200 nM) showing substantially higher potency than the aliphatic ether linked analogs 7 and 8 (Entries 1–3, IC₅₀ = 1–7 μM). These significant differences in potency support our assumption that linker length and composition are critical features of bifunctional conjugates. Interestingly, conjugate 7b containing a trione motif instead of the imine functionality was ~ 6-fold less potent than 7a. This result confirms that the imino-thienopyridinedione motif remains important for the PTP4A3 inhibitory interaction in the bifunctional compounds. In agreement with the data for JMS-053 analogs with large substituents at the imide nitrogen in zone 3,^7b conjugate 9 also lost substantial inhibitory activity (Entry 4).

Table 1.

In vitro inhibition of PTP4A3 phosphatase activity.^a

Entry	Compound	IC50 (μM)	±SEM	N
1	7a (EJR-876-35)	1.16	0.51	6
2	7b (EJR-925-45)	6.59	1.99	3
3	8 (EJR-876-34)	4.98	0.31	3
4	9 (EJR-980-67)	1.92		1
5	16 (EJR-887-24)	0.206	0.011	3
6	17 (EJR-887-35)	0.107	0.016	3
7	1 (JMS-053)	0.084	0.041	6

^aRecombinant human PTP4A3 phosphatase was used and the enzymatic assay was performed as previously described.⁷. See SI for additional details.

Although Ser/Thr kinases and phosphatases are more commonly associated with the UPR,²¹ a mechanistic understanding of the links between PTPs and ER stress and cell death has also been emerging.¹⁰ For example, PTP1B regulates the UPR,²² and PTPN2 mediates ER stress response-dependent neuronal damage.²³ While the adamantyl moiety in bifunctional compounds should induce a basal level of ER stress, we were most interested in PTP4A inhibitors that potentiated this effect. Therefore, we evaluated the ability of the most active bifunctional inhibitors 7a, 16, and 17 to induce ER stress, as well as diminish cancer cell viability. Dithiothreitol (DTT), a potent UPR activator, served as a positive control in the ER stress measurements (Fig. 2). Notably, compound 7a-dependent induction of the canonical UPR reporter, BiP (HSPS5A),²⁴ along with a moderate accumulation of the spliced Xbp1 transcription factor that initiates BiP synthesis²⁵ in MDA-MB-231 breast cancer cells is consistent with known links between PTP inhibition and cancer cell death (Fig. 2).²⁶ In fact, MDA-MB-231 cells also express high levels of p53,²⁷ a key survival enzyme which in turn regulates PTP4A3 levels.²⁸ BiP induction, concomitant with a strongly enhanced expression of the pro-apoptotic gene, CHOP (DDIT3),²⁹ provides a mechanism to explain the prominent effect of 7a on MDA-MB-231 viability (Fig. 3 and Table 2, Entry 3; EC₅₀ ~ 8.6 μM). Interestingly, analog 7a was significantly more potent in these cell-based assays than 16 and 17, even though it was 6–10 times less potent in the biochemical PTP4A3 enzyme assay. The monofunctional adamantyl linker 4 served as a negative control in the cancer viability assay and indeed had negligible inhibitory effects on cell growth (Fig. 3 and Table 2, Entry 2). Similarly, the monofunctional parent PTP4A3 inhibitor 1 (JMS-053) was approximately equipotent to 17 in the MDA-MB-231 cell viability assay (Table 2, Entry 1, IC₅₀ ~ 42.7 μM), but 4–5 times less potent than 16 and 7a, in spite of its superior (2–13 times more potent, respectively) inhibition of PTP4A3. We also evaluated trione 7b in these assays. While the MDA-MB-231 growth inhibitory effects and UPR/ER stress pathway gene patterns were similar to compound 16, it was not as potent as imine 7a and produced a significantly lower accumulation of the spliced Xbp1 transcription factor gene and CHOP than 7a (Figs. 2 and 3, and Table 2, Entry 4). These effects might be due to its ca. 6-fold diminished ability to inhibit PTP4A3.

Fig. 2.

Induction of genes corresponding to UPR and ER stress pathway enzymes by JMS-053 (1), 7a (EJR-876-35), 7b (EJR-925-45), 16 (EJR-887-24), 17 (EJR-887-35) and DTT. UPR induction experiments were performed by qPCR. MDA-MB-231 cells were treated for 6 h with complete media containing 2 times the EC₅₀ calculated using the cell viability assay (see Table 2 and SI). A 2-h treatment with 2 mM DTT was used as a positive control for UPR induction. The corresponding fold increase of the indicated UPR markers relative to the DMSO control are plotted, +/−SEM. N = 3. sXbpl = spliced/activated Xbpl; HSPA5A = gene encoding BiP; DDIT3 = gene encoding CHOP.

Fig. 3.

Cancer cell viability assay with JMS-053 (1), adamantyl linker 4 (negative control), 7a (EJR-876-35), 7b (EJR-925-45), 16 (EJR-887-24), and 17 (EJR-887-35). A total of 3000 MDA-MB-231 cells/well were seeded and subjected to drug treatment for 72 h. The CellTiter Glo assay was used to measure cell viability and EC₅₀ values were generate using a sigmoidal nonlinear regression in PRISM software. Cell viability data represent the average of 2 independent experiments, +/−SEM.

Table 2.

Inhibitory effects of JMS-053 (1), adamantyl linker 4 (negative control), 7a (EJR-876-35), 7b (EJR-925-45), 16 (EJR-887-24), and 17 (EJR-887-35) on MDA-MB-231 cancer cell growth.^a

Entry	Compound	EC50 (μM)	±SD
1	1 (JMS-053)	42.7	3.0
2	4	>50	–
3	7a (EJR-876-35)	8.6	1.1
4	7b (EJR-925-45)	11.5	0.9
5	16 (EJR-887-24)	11.8	1.1
6	17 (EJR-887-35)	31.6	1.3

^aN = 2.

We also investigated a possible additive or synergistic effect between JMS-053 and adamantyl linker 4, but failed to detect any changes in viability upon addition of 20–60 μM 4 to MDA-MB-231 cells treated with JMS-053.²⁰ Previously, JMS-053 proved to be selective for PTP4A1-3 in a panel of 25 other phosphatases and 50 kinases.³⁰ Similarly, compound 7a demonstrated < 11% inhibition at 1 μM concentration for three closely related phosphatases, PTP1B, DUSP3, and CDC25B.²⁰ Overall, these results suggest that it is the combined action of the adamantyl group in covalent conjugation with the PTP4A3 inhibitory heterocycle that is responsible for the activation of ER stress responsive genes and the suppression of MDA-MB-231 breast cancer cell growth. The differences among the bifunctional analogs 7a, 16, and 17 further suggest that the linker moiety plays an important role in the biological properties of these hybrid molecules.

In conclusion, we have designed a set of bifunctional JMS-053 conjugates that inhibit the dual specificity phosphatase PTP4A3 enzymatically as well as activate the ER stress-induced UPR pathway. These conjugates utilize a highly hydrophobic adamantyl moiety linked to the PTP4A3 inhibitory scaffold at a site that was identified by a prior SAR campaign to be relatively tolerant to chemical modification. Both the site of attachment as well as the length and functionality of the linker moiety determined biological outcomes. While several of the newly synthesized conjugates show low nanomolar IC₅₀ values for inhibition of PTP4A3, the slightly less potent analog 7a was superior to derivatives with alternate linker moieties, i.e. 16 and 17, in inducing expression of the ER stress response genes HSPA5A, sXbpl, and DDIT3, resulting in the most potent growth inhibitory effect in MDA-MB-231 breast cancer cell viability assays. In contrast, the parent monofunctional PTP4A3 inhibitor, JMS-053 (1), did not appear to cause ER stress or activate the UPR and had the least effect on cell growth. In conclusion, these results highlight the promise of identifying dual-pathway agents at an early drug discovery stage that are able to modulate compensatory pathways used by cancer cells to bypass apoptosis.³¹