Small molecule branched-chain ketoacid dehydrogenase kinase (BDK) inhibitors with opposing effects on BDK protein levels

Abstract

Branched chain amino acid (BCAA) catabolic impairments have been implicated in several diseases. Branched chain ketoacid dehydrogenase (BCKDH) controls the rate limiting step in BCAA degradation, the activity of which is inhibited by BCKDH kinase (BDK)-mediated phosphorylation. Screening efforts to discover BDK inhibitors led to identification of thiophene PF-07208254, which improved cardiometabolic endpoints in mice. Structure-activity relationship studies led to identification of a thiazole series of BDK inhibitors; however, these inhibitors did not improve metabolism in mice upon chronic administration. While the thiophenes demonstrated sustained branched chain ketoacid (BCKA) lowering and reduced BDK protein levels, the thiazoles increased BCKAs and BDK protein levels. Thiazoles increased BDK proximity to BCKDH-E2, whereas thiophenes reduced BDK proximity to BCKDH-E2, which may promote BDK degradation. Thus, we describe two BDK inhibitor series that possess differing attributes regarding BDK degradation or stabilization and provide a mechanistic understanding of the desirable features of an effective BDK inhibitor.

Fig. 1. PF-07208254 improves cardiac function similar to BT2 and reduces pBCKDH in mice.

A–J Mice were subjected to TAC surgery. A Study design. Mice were switched to chow containing PF-07208254 or BT2 48 h prior to surgery. Four weeks after surgery, echocardiography was performed. Five to six weeks after surgery, the animals were euthanized. B, C Echocardiography measurements B. % Fractional shortening (%FS) (**p = 0.007, *p = 0.015), C % Ejection fraction (%EF). D–E Tissue weights were measured at euthanasia. D Heart weight normalized to tibia length (*p = 0.018). E Lung weight normalized to tibia length. (N = 14–26 animals/group for B–E; statistics performed by one-way ANOVA with Tukey post-hoc test for B–D and pairwise Wilcoxon test was performed for E). F–H Heart tissue was immunoblotted. F Representative Western blot images of pBckdh, Bckdh, BDK and Gapdh. G-H Densitometry from F (N = 3–6 animals/group; statistics performed by pairwise Wilcoxon test. #; p < 0.0001). I, J Plasma BCAA and BCKA levels were quantified at day 38. I Leucine. J Ketoleucine (N = 10–15 animals/group; statistics performed by one-way ANOVA with Tukey post-hoc test. ^#p = 0.0001). Data represent the mean ± SEM. Source data are provided as a Source Data file.

Fig. 2. PF-07208254 improves metabolism similar to BT2 and sustainably reduces BCAA and BCKA in mice.

Mice were fed HFD for 10 weeks, at which time animals were randomized into groups, and treated daily with vehicle, PF-07208254 or BT2. A Study design. B-H Oral glucose tolerance tests (oGTTs) were performed (N = 12 animals/group; a longitudinal mixed effects model with a random intercept and an AR(1) covariance structure was fit for each mouse for the glucose AUCs over the course of the study). B Day 2 oGTT. C Day 2 oGTT AUC. D Week 2 oGTT. E Week 2 oGTT AUC (**p = 0.009, ***p < 0.0001, **p = 0.009). F–H Plasma and livers were isolated 1 h post final compound dose, and steatosis was evaluated by histology (N = 12 animals/group). F Hepatic triglycerides (statistics performed with pairwise Wilcoxon test, ***p = 0.003). G Hepatic steatosis was graded by a veterinary pathologist on a scale of 0-4. (Statistics was performed using a Kruskal Wallis test followed by a Dunn test to compare within groups, ***p = 0.001). H Plasma insulin levels (N = 11–12 animals/group; statistics performed by one-way ANOVA). I, J A PK/PD assessment was performed after 7 weeks of treatment in which mice are dosed with compound, and timed bleeds were performed to measure drug levels, BCAA, and BCKA levels (N = 6–12 animals/group; statistics performed with one-way ANOVA with Tukey HSD test). I Leucine (^#p < 0.0001), J Ketoleucine (^#p < 0.0001). Data represent the mean ± SEM. Source data are provided as a Source Data file.

Fig. 3. Thiazoles bind to BDK in distinct mode adjacent to thiophenes.

Allosteric binding site from X-ray crystal structures of BDK bound to inhibitors. The regulatory domain of BDK is shown in yellow ribbons, and the kinase domain in aqua. Side chains of contacting residues are shown, and hydrogen bonds in dash. A S3 (stick model, C atoms colored in green) bound to hBDK. Superimposed is the BT2 X-ray crystal structure (PDB ID: 4E00) with BT2 C atoms colored in pink and side chains of residues that adopt different rotamers in the BT2 complex in gray. B PF-07247685 (stick model, C atoms colored in green) bound to hBDK. Superimposed are BT2 (C atoms colored in pink) and S3 (C atoms colored in salmon).

Fig. 4. Thiazole compounds do not improve metabolic endpoints after day 2, and BCAA/BCKA rebound occurs over time.

A–C Mice were fed HFD for 10 weeks, at which time animals were randomized into groups, and treated daily with vehicle or PF-07238025 for 9 days (N = 9–10 animals/group; statistics were performed using one-way ANOVA with Tukey HSD test, *p = 0.02). A Study design. B, C An oGTT was performed at day 2. B oGTT. C oGTT AUC. D–N Mice were fed HFD for 10 weeks, at which time animals were randomized into groups, and treated daily with vehicle or PF-07238025 for 8 weeks. D Study design. E, F An oGTT was performed at week 2 (N = 9–10 animals/group). E oGTT, F oGTT AUC. G, H Plasma and livers were isolated 1 h post final compound dose. G Hepatic triglycerides (N = 9–10 animals/group). H Plasma insulin levels (N = 8–10 animals/group). I–N PK/PD assessments were performed on day 1 (I, J) (N = 4 animals/group; statistics were performed by Welch’s t-test), after 7 days (K, L) (N = 4 animals/group; statistics were performed by Welch’s t-test), or 6 weeks (M, N) (N = 4–10 animals/group; statistics were performed using one-way ANOVA with Tukey HSD test) of treatment, in which mice were dosed with compound, and timed bleeds were performed to measure drug levels, BCAA, and BCKA levels. I Day 1 Leucine (***p = 0.0049), (J) Day 1 Ketoleucine (#p = 1.64 × 10⁻⁶, ***p = 0.003). K Day 7 Leucine, (L) Day 7 Ketoleucine (***p = 0.001). M Week 6 Leucine (*p = 0.019), (N) week 6 Ketoleucine (****p < 0.001, ***p = 0.001). Data represent the mean ± SEM. Source data are provided as a Source Data file.

Fig. 5. Washout PK/PD demonstrates different pharmacological profiles of thiazole and thiophene BDK inhibitors.

Mice were fed HFD for 10 weeks, at which time animals were randomized into groups, and treated BID or QD as indicated with vehicle, BT2, PF-07208254, PF-07247685, or PF-07238025 at maximal inhibitory doses. A Study design. Mice were bled immediately prior to dosing compound, 1 h post compound dose, 4 h post compound dose, 7 h post compound dose (immediately prior to 2nd daily compound dose if BID dosed), and 24 h after first compound dose on day 3 and day 17 (N = 8–9 animals/group; a longitudinal mixed effects model with an unstructured covariance was used for statistics for C–D, F–G). B After the final compound dose on day 19, animals were bled at 24 h, 36 h, 45 h, and 68 h on days 20–21 (N = 5 animals/group; a longitudinal mixed effects model with a random intercept for each mouse and an AR(1) covariance structure was fit for statistics for E, H). C Leucine was measured on day 3, (D) day 17 (*p = 0.043, **p = 0.01, ***p = 0.001, ^#p < 0.0001, +within group, p = 0.049 from day 3), and (E) upon compound washout (*p = 0.047, 0.013). F Ketoleucine was measured on day 3, (G) day 17 (^#p < 0.0001, ++ within group, p < 0.001 from day 3), and (H) upon compound washout (^#p < 0.0001, ***p = 0.001). Dashed lines in panels E, H represent the 0–7 h time points from D and G for visualization purposes. Data represent the mean ± SEM. Source data are provided as a Source Data file.

Fig. 6. Differing effects of thiazoles and thiophenes on pBCKDH and BDK protein levels in mice.

A, B HFD-fed mice were treated with PF-07208254 or BT2 for 8 weeks as described in Fig. 2A. Hearts were isolated and subjected to Western blot for pBckdh, Bckdh, Bdk, and Gapdh. A Representative Western blot images, (B) Densitometric analyses of pBckdh/Bckdh (left panel) and Bdk/Gapdh (right panel) (N = 12 animals/group, ^#p < 0.0001). C, D HFD-fed mice were treated with PF-07238025 for 8 weeks as described in Fig. 4D. Hearts were isolated and subjected to Western blot for pBckdh, Bckdh, Bdk, and Gapdh. C Representative Western blot images, D Densitometric analyses of pBckdh/Bckdh (left panel) and Bdk/Gapdh (right panel) (N = 4–5 animals/group, ^#p < 0.0001, *p = 0.03, ***p < 0.001). E, F HFD-fed mice were treated with PF-07247685 for 18 days as described in Supplementary Fig. 6A. E Representative Western blot images, F Densitometric analyses of pBckdh/Bckdh (left panel) and Bdk/Gapdh (right panel). (N = 3 animals/group; ^#p < 0.0001, *p = 0.011, **p = 0.008). A one-way ANOVA with Tukey HSD test was performed for B, D, and F n.s. non-specific band. Data represent the mean ± SEM. Source data are provided as a Source Data file.

Fig. 7. Differing effects of thiazoles and thiophenes on pBCKDH and BDK protein levels in human cells.

A–D Hek293 cells were treated in culture with compounds for 48 h. A, B Cells were treated with PF-07208254 or BT2 in a dose response. A Representative Western blot images. B Densitometric analyses of pBCKDH/BCKDH (left panel) and BDK/GAPDH (right panel) (pBCKDH *p = 0.036, 0.049, BDK *p = 0.014, 0.033 ***p = 0.004, 0.005). C, D Cells were treated with PF-07238025 or PF-07247685 in a dose response. C Representative Western blot images. D Densitometric analyses of pBCKDH/BCKDH (left panel, ***p = 0.002, ^#p < 0.0001) and BDK/GAPDH (right panel). n.s. non-specific band. (N = 4–8 independent experiments; one-way ANOVA with Tukey HSD test was performed for statistical analyses). Data represent the mean ± SEM. Source data are provided as a Source Data file.

Fig. 8. Thiazoles enhance binding of BDK to BCKDH complex while thiophenes reduce binding and promote degradation.

A BDK was immunoprecipitated from hearts of DIO mice that were treated with vehicle, 100 mg/kg BT2 QD or 10 mg/kg PF-07247685 BID for 18 days, and proteomics was performed on the immunoprecipitated proteins. Volcano plot demonstrating the most significantly enriched proteins bound to BDK after BT2 treatment (to the left) versus PF-07247585 treatment (to the right). B Representative Western blot of BDK after IgG or BDK immunoprecipitation from mouse hearts (N = 3 animals/group). C An AlphaLISA assay was performed with BDK and BCKDHE2. Fluorescence aligned with in vitro potency of BT2, PF-07208254, PF-07238025 and PF-0724785 (n = 5–26 independent experiments). Source data are provided as a Source Data file.

Fig. 9. Molecular dynamics simulation shows differing effects on fluctuations of BDK residues.

A, B Root-mean-square fluctuations (RMSFs) based on MD show differences in BDK loop 1 and lipoyl binding pocket residue dynamics when bound to destabilizers or stabilizers. A RMSFs of BDK Cα atoms over all residues. B RMSFs of loop 1 and lipoyl-binding pocket residues. C Model of BDK inhibitor modulation of BDK-BCKDHE2 binding and BDK levels.