Discovery of the First Lactate Dehydrogenase Proteolysis Targeting Chimera Degrader for the Treatment of Pancreatic Cancer

Design of Pyrazole-Based VHL-Recruiting LDH PROTAC Degraders.

The pyrazole-based LDH inhibitors, compounds 5–7, exhibited high potency in inhibiting both LDHA and LDHB and were effective in reducing lactate production in cancer cells. Compounds 6 and 7 also effectively inhibit LDH in vivo.²² Therefore, we chose pyrazole-based LDH inhibitors, 5–7, as LDH binders for the development of LDH PROTAC degraders. We carried out structure-based design to attach a linker to LDH binders. The crystal structure of LDHA in complex with compound 6 (Figure 2A)²² revealed that the thiophene group of 6 is solvent-exposed (indicated by the black dotted circle in Figure 2A) and is a suitable moiety for the linker attachment. We therefore designed precursor 8 by replacing the thiophene moiety with a four-carbon chain spacer bearing a terminal amine group to serve as a handle for linker installation (Figure 2B). Because the thiazole carboxylic acid moiety and the 4-benzyl sulfonamide group of 6 (and 5) form critical hydrogen bond interactions with LDHA (the magenta dotted lines in Figure 2A), we retained these moieties in precursor 8. To simplify chemical synthesis, we removed the fluorine atom ortho to the sulfonamide group based on inhibitor 5. We chose to hijack the E3 ligase VHL, instead of cereblon (CRBN), mainly to avoid the potential complication caused by degradation of CRBN neosubstrates. Given the fact that linkers play a crucial role in the development of PROTACs, we designed a number of putative VHL-recruiting LDH PROTAC degraders (13–32) featuring various alkylene and polyethylene glycol (PEG) linkers (Figure 3), which were connected to VHL-1, a classic VHL ligand.³⁰

Synthesis.

The synthetic route to obtain precursor 8 is outlined in Scheme 1. Cesium carbonate-mediated alkylation of known 1,3-diketone 9 with commercially available 4-(bromomethyl)-benzenesulfonamide in dimethyl sulfoxide (DMSO) at room temperature generated compound 10. Microwave heating of 10 with pyrrolidine and tosylic acid, followed by addition of ethyl 2-hydrazinylthiazole-4-carboxylate hydrobromide with continued microwave heating, yielded the desired pyrazole 11. Sonogashira coupling reaction between compound 11 and commercially available 12 afforded precursor 8. Compounds 13–32 were prepared by amide condensation reactions between linker-attached VHL ligands and precursor 8 and subsequent hydrolysis of the ethyl ester group under basic conditions (Scheme 1). The linker-attached VHL ligands were synthesized according to the previously reported procedures.³¹

Effects of Compounds 13–32 on Degrading LDHA.

We next evaluated the effect of compounds 13–32 on reducing the LDHA protein level by western blot (WB) analysis in PANC1 cells, a pancreatic cancer cell line (Figure 4). Inhibitor 5 was used as a control in this study, and PANC1 cells were treated with the test compound at 1 and 10 μM concentrations for 48 h. We chose to screen the effect of these compounds on the LDHA protein level because LDHA has been shown to be overexpressed in pancreatic cancers.^32–34 As expected, the parent inhibitor, compound 5, did not degrade LDHA. Compounds 13–15 bearing the shortest carbon linkers (C2–C4) did not induce obvious LDHA degradation (Figure 4A). On the other hand, compounds 16 to 23 (C5–C12) with increasing linker length induced significant LDHA degradation at 10 μM. In particular, compounds 19–23 induced substantial LDHA degradation even at 1 μM. Interestingly, among the putative LDH degraders featuring PEG linkers, only compound 24, which contains the shortest PEG linker, effectively induced LDHA degradation at both 1 and 10 μM (Figure 4B). Based on these initial screening results, we selected degraders 19–22 with relatively long carbon linkers and degrader 24 with a PEG linker for further testing the concentration-dependent LDHA degradation induced by these compounds to identify the best LDHA degrader among them. We did not select compound 23 because 23 has a longer linker than 22 but was not as effective as 22 (Figure 4A). From this concentration–response study, we identified compound 22 as the most effective LDHA degrader, which induced significant LDHA degradation even at 0.1 μM (Figure 5). Therefore, compound 22 was selected for further characterization.

Compound 22 Potently Induces LDH Degradation in a Time- and UPS-dependent Manner.

We further investigated the potency of degrader 22 in reducing the LDHA protein level in PANC1 cells and found that 22 potently degraded LDHA with a DC₅₀ value of 38 ± 0.5 nM and a D_max value of 93 ± 2.6% (Figure 6A,B). No “hook effect” was observed for compound 22 at concentrations up to 10 μM. Since LDH inhibitor 5 potently inhibited both LDHA and LDHB, we next assessed the effect of compound 22 on reducing the LDHB protein level in PANC1 cells. As expected, compound 22 also potently degraded LDHB with a DC₅₀ value of 74 ± 6.9 nM and a D_max value of 86 ± 7.7% (Figure 6A,B). Next, we assessed the kinetics of the LDHA and LDHB protein degradation induced by 22 by treating PANC1 cells with 1 μM of 22 for various durations. As shown in Figure 6C, LDHA and LDHB degradation was observed as early as 24 h and reached the maximum level at 48 h. Together, these results indicate that compound 22 degrades LDHA and LDHB in a concentration- and time-dependent manner. The compound treatment time of 48 h to achieve the maximum degradation presents a limitation since potential secondary cellular events may also be occurring.

To facilitate further assessment and characterization of compound 22, we designed two close analogues of compound 22, compounds 33 and 34, as negative control compounds for degrader 22 (Figure 7). Compound 33 was designed to prevent its binding to the VHL E3 ligase by benzylating the key hydroxy group of VHL-1, which is essential for binding VHL,³⁵ while keeping the rest of the molecule the same as 22 (Figure 7A, top). On the other hand, compound 34 was designed to reduce binding to LDHA by replacing the key carboxylic acid group with the dimethyl amide group while keeping the linker and VHL binding moiety the same as 22 (Figure 7A, bottom). The synthesis of 33 and 34 is outlined in Scheme S1. Briefly, amide coupling reaction between known compound 35 (VHL-1-Bn)³⁵ and dicarboxylic acid 36 afforded carboxylic acid 37. Another amide coupling reaction and subsequent deprotection of the ester group provided the desired compound 33. Compound 34 was obtained by amide coupling of 22 and dimethylamine. Indeed, compounds 33 and 34 did not degrade LDHA as expected (Figure 7B).

To further examine whether the degradation induced by 22 is mediated through the UPS, we conducted a set of rescue studies. PANC1 cells were pretreated with the VHL ligand Ac-VHL-Me (20 μM),^36,37 proteasome inhibitor MG132 (1 μM), neddylation inhibitor MLN4924 (1 μM), or parent LDH inhibitor 5 (1 μM) for 2 h, followed by treatment with compound 22 for 24 h. As shown in Figure 8, the degradation of LDHA by compound 22 was at least partially rescued by pretreatment with the VHL ligand, proteasome inhibitor, neddylation inhibitor, or parent LDH binder. Together, these results indicate that the LDHA degradation induced by 22 is dependent on the UPS and requires its engagement with both LDH and VHL E3 ligase.

Unbiased Global Proteomic Studies Confirm That Compound 22 Degrades LDHA and LDHB.

We next performed global proteomic profiling studies using a label-free quantitative mass spectrometry (MS) approach in PANC1 cells (Figure 9). The samples for MS analysis were prepared by treating PANC1 cells with compound 22 (3 μM), compound 33 (3 μM), or DMSO for 24 h. WB analysis confirmed that the LDHA protein level was significantly reduced by degrader 22 but not by the negative control 33 in these samples (Figure S1). As shown in Figure 9A–C, LDHA and LDHB are among a small number of proteins, whose levels were reduced by compound 22, compared to DMSO or compound 33. Interestingly, the proteomics analysis also showed that of the 52 down- and up-regulated proteins by compound 22 compared to DMSO and compound 33, 39 are the proteins involved in cellular metabolic process (Figure 9D and Table S1). These results suggest that compound 22 is relatively selective for LDHA and LDHB and in turn for cellular metabolic process. A potential limitation of this proteomic study is that the relatively long treatment time (24 h) might induce additional secondary cellular effects.

Compound 22 Effectively Degrades LDHA and Inhibits the Proliferation in Multiple Pancreatic Cancer Cell Lines.

The two major molecular subtypes consistently found in pancreatic ductal adenocarcinoma (PDAC) are the classical epithelial (E) and quasi-mesenchymal (QM) subtypes, and many data sets indicate the poor prognosis of the QM subtype compared to the E subtype.³⁸ Therefore, we chose three different pancreatic cell lines, MiaPaca2, BxPC3, and PANC1 cells, to further characterize LDH degrader 22. MiaPaca2 and BxPC3 cells belong to the QM and E subtypes, respectively, while PANC1 cells cannot be unambiguously classified due to the fact that PANC1 cells have the character of both mesenchymal cells and epithelial-type cells.³⁹ We first evaluated the anti-proliferative activities of 22 in PANC1 cells using a WST-8 assay, with parent inhibitor 5 and two negative compounds 33 and 34 as controls (Figure 10A, left). Compound 22 effectively inhibited the cell proliferation in PANC1 cells with a GI₅₀ value of 16.1 ± 5.7 μM. It was more potent than parent inhibitor 5, which inhibited the cell growth only at high concentrations (GI₅₀ > 30 μM). As expected, two negative controls, 33 and 34, which did not degrade LDHA (Figure 7B), did not display any cell growth inhibition. We also confirmed that compound 22, but not inhibitor 5, effectively degraded LDHA in PANC1 cells (Figure 10A, right). Notably, the potency of compound 22 in cell growth inhibition (GI₅₀) is much weaker than its LDHA/LDHB degradation potency (DC₅₀) in PANC1 cells. However, it is not uncommon to observe more than 100-fold difference between the DC₅₀ and GI₅₀ values in aggressive cancers such as pancreatic cancer.³⁷ We next compared the anti-proliferative effect of compound 22 to that of inhibitor 5 in MiaPaca2 cells. While compound 5 inhibited the cell growth in MiaPaca2 cells only at high concentrations (GI₅₀ > 30 μM), compound 22 inhibited the cell proliferation with a GI₅₀ value of 12.2 ± 3.7 μM (Figure 10B, left). Similar to PANC1 cells, the two negative controls, 33 and 34, did not show any effect on cell proliferation in MiaPaca2 cells. Using WB analysis, we determined that compound 22 induced LDHA degradation also in MiaPaca2 cells treated for 24 h, while compound 5 did not induce LDHA degradation (Figure 10B, right). Notably, the LDHA degradation effect induced by compound 22 was less profound in MiaPaca2 cells compared to PANC1 cells. In BxPC3 cells, compound 22, but not inhibitor 5, suppressed the cell proliferation and effectively degraded LDHA (Figure 10C). The two negative controls, 33 and 34, were also largely ineffective in suppressing the cell proliferation. Taken together, these results indicate that compound 22 can degrade LDH and inhibit the cell growth in multiple pancreatic cancer cell lines and suggest that the cell growth inhibition effect of compound 22 could be due to its LDH degradation activity.

Compound 22 is Bioavailable in Mice.

Last, we assessed in vivo mouse PK properties of compound 22. Compound concentrations in plasma from Swiss albino mice were determined following a single intraperitoneal (IP) administration of compound 22 at 5 mg/kg. We were pleased to find that a single IP injection at this relatively low dose resulted in good exposure levels over 24 h in plasma (Figure 11). The peak plasma concentration (1800 nM) was observed at 8 h. In addition, compound 22 was well tolerated by the treated mice, and no clinical signs were observed in the PK study. Taken together, these results suggest that compound 22 could be a valuable chemical tool for investigating the effects of LDH degradation in vivo.

Chemistry General Procedures.

All commercially available chemical reagents were used directly in syntheses without further purification. Microwave-heated reactions were performed with a Discover SP microwave system with an Explorer 12 Hybrid Autosampler by CEM (Buckingham, UK). A Teledyne ISCO CombiFlash Rf⁺ instrument and HP C18 RediSep Rf reverse phase columns equipped with UV detector were used to conduct flash chromatography. All final compounds for biological evaluation were purified with preparative high-performance liquid chromatography (HPLC) on an Agilent Prep 1200 series with the UV detector set to 254 or 220 nm with a flow rate of 40 mL/min at room temperature. Crude samples were injected into a Phenomenex Luna 750 × 30 mm, 5 μ C18 column, with the gradient program set to 10% of methanol or acetonitrile (B) in H₂O containing 0.1% TFA (A) progressing from 10 to 100% of methanol or acetonitrile (B). The purity of all compounds for biological testing is >95% assessed by high-performance liquid chromatography–high-resolution mass spectroscopy (HPLC-HRMS). All HPLC spectra were obtained by using an Agilent 1200 series system with DAD detector and a 2.1 mm × 150 mm Zorbax 300SB-C18 5 μ column for chromatography. Samples (0.5 μL) were injected onto a C18 column at room temperature with the flow rate of 0.4 mL/min. Chromatography was performed with the solvent as follows: water containing 0.1% formic acid was designated as solvent A, while acetonitrile containing 0.1% formic acid was designated as solvent B. The linear gradient was set such that 1% B was used from 0 to 1 min, 1 to 99% B was used from 1 to 4 min, and 99% B was used from 4 to 8 min. High-resolution mass spectra data were acquired in positive ion mode using an Agilent G1969A API-TOF with an electrospray ionization (ESI) source. All compounds were also characterized using either a Bruker (Billerica, MA) DRX nuclear magnetic resonance (NMR) spectrometer (400 and 600 MHz, ¹H NMR, 151 MHz ¹³C NMR). Chemical shifts for all compounds are reported in units of parts per million (ppm, δ) relative to residual solvent peaks. ¹H NMR data are reported in the following format: chemical shift, multiplicity (s = singlet, d = doublet, t = triplet, q = quartet, m = multiplet), coupling constant, and integration. VHL-based carboxylic acid linkers were prepared following the reported procedures.³¹

Cell Culture.

All cell lines were purchased from ATCC and authenticated. PANC1 and MiaPaca2 were cultured in Dulbecco’s modified Eagle’s medium, while BxPC3 was cultured in RPMI-1640, supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin. All cells were cultured at 37 °C and 5% CO₂.

Western Blotting Analysis.

WBs were performed as described previously.⁴⁰ Briefly, cells were lysed on ice for 30 min with the lysis buffer (50 mM Tris pH 7.4, 1% IGEPAL CA-630, 150 mM NaCl, 1 mM ethylenediaminetetraacetic acid (EDTA), and 1 mM AESBF), supplemented with protease and phosphatase inhibitor cocktail (A32961, Thermo Fisher Scientific). The samples were centrifuged at 12,000g for 10 min at 4 °C to obtain the supernatant as a cell lysate. Protein concentrations were determined using a Pierce rapid gold bicinchoninic acid (BCA) protein assay kit. The primary antibodies used were LDHA (Cat# 2021S, Cell Signaling Technology) and Vinculin (Cat#13901S, Cell Signaling Technology). Fluorescence-labeled IRDye 800CW goat anti-rabbit IgG (926–32211) were purchased from LI-COR, and an OdysseyCLx imaging system (LI-COR) was used to obtain protein signals. The WBs were analyzed by Image Studio Lite software (LI-COR). DC₅₀ values were obtained with GraphPad Prism 8 from the data of three independent experiments.

Cell Proliferation Assays.

Cells were seeded (5 × 10³ cell) in a 96-well microplate in triplicates (Corning 720089). After 24 h, cells were treated with 0.1% DMSO or indicated serial dilution of compounds for an additional 3 days. Cell viability was tested using WST-8 reagent (CK04, Dojindo). Briefly, 20 μL of CCK-8 was added to each well, and the plates were kept in an incubator at 37 °C for 3 h in the dark. Absorbance signals for WST-8 were read at 450 nm with 650 nm as a reference performed with Infinite F PLEX plate reader (TECAN, Morrisville, NC). GI₅₀ values were analyzed using GraphPad Prism 8 from the data of three independent experiments.

Global Proteomic Profiling Sample Preparation.

Mass Spectrometry Analysis.

Peptides were cleaned up using C18 stage tips, and the concentration was determined (Peptide assay, 23275, Thermo Fisher Scientific). The clean peptides were dissolved in 0.1% formic acid and analyzed on a Q Exactive HF-X coupled with an Easy nanoLC 1200 (Thermo Fisher Scientific, San Jose, CA). Then, 0.5 g of peptides were loaded onto an Acclaim PepMap RSLC C18 Column (150 mm 75 m ID, C18, 2 m, Thermo Fisher). Analytical separation of all peptides was achieved with a 130 min gradient. A linear gradient of 5–30% buffer B over 110 min was executed at a 300 nL/min flow rate, followed by a ramp to 100% B in 5 min, and 15 min wash with 100% B, where buffer A was aqueous 0.1% formic acid and buffer B was 80% acetonitrile and 0.1% formic acid.

Raw Proteomics Data Processing and Analysis.

Mass spectra processing and peptide identification were performed on the Andromeda search engine in MaxQuant software (Version 1.6.10.43) against a human UniProt database (UP000005640). All searches were conducted with a defined modification of cysteine carbamidomethylation, with methionine oxidation and protein amino-terminal acetylation as dynamic modifications. Peptides were confidently identified using a target-decoy approach with a peptide false discovery rate (FDR) of 1% and a protein FDR of 1%. A minimum peptide length of 7 amino acids was required, maximally two missed cleavages were allowed, the initial mass deviation for precursor ions was up to 7 ppm, and the maximum allowed mass deviation for fragment ions was 0.5 Da. Data processing and statistical analysis were performed on Perseus (Version 1.6.10.50). Protein quantitation was performed on technical replicate runs, and two-sample t-test statistics was used to report statistically significant expression fold changes.

Mouse Pharmacokinetic Study.

Six male Swiss Albino mice were administered intraperitoneally with solution formulation of compound 22 at a 5 mg/kg dose. The following formulation vehicle was used in the study: 5% v/v NMP, 45% v/v propylene glycol, and 50% v/v PEG 400. Blood samples (approximately 60 μL) were collected from the six test mice at 0.5, 2, 8, 12, 16, and 24 h. Plasma was harvested by centrifugation of blood and stored at −70 ± 10 °C until analysis. Plasma samples were quantified by fit-for-purpose LC–MS/MS method. Compound concentrations in plasma at each time point are average values from three test mice. Error bars represent ±SEM. Experiments involving mice were performed according to the Institutional Animal Care and Use Committee (IACUC)-approved protocol.