INTRODUCTION
Cells need to react promptly to changes in nutrients to co-ordinate intracellular bio-processes such as energy homeostasis, cell growth, and proliferation. In mammalian cells, the two protein kinases — mTOR (mechanistic target of rapamycin) and AMPK (AMP-activated protein kinase) are central regulators of cell metabolism and their activities are modulated by various nutrients, respectively
1. mTOR interacts with Raptor, mLST8, and other proteins to form mTOR complex 1 (mTORC1) to promote protein translation, lipid synthesis, glucose metabolism, and so on
2, and the activation process of mTORC1 by amino acids has been successfully elucidated over the recent decade. Briefly, amino acids bind with their specific sensors such as Sestrin2 and CASTOR
3,4, thereby initiating the recruitment of mTORC1 to the lysosome. More than a dozen of proteins (and protein complexes), especially small GTPases (RagA/B/C/D, Arf1, Rheb)
5-7, guanine nucleotide exchange factors (GEF) (Ragulator, ATP6AP1)
8,9, GTPase activating proteins (GAP) (GATOR1, folliculin, TSC1/2)
5,10,11, and their interaction partners (GATOR2)
10, play essential roles in that process to activate mTORC1. In contrast, AMPK is activated under energy stress or nutrient starvation conditions, and it phosphorylates numerous substrates to suppress anabolism and to cause breakdown of macromolecules, thereby affecting many bio-processes such as glucose and lipid metabolism, mitochondrial biogenesis, cell death, and so on
12. It is well established that AMPK could be activated via directly sensing the changes of the AMP/ADP ratio under the energy stress condition, which is known as the canonical nucleotide-dependent mechanism
13,14. However, the activation of AMPK under the glucose deprivation condition does not involve major changes in AMP/ADP ratio
15, but through phosphorylation of a conserved threonine residue (known as the Thr172 site) at its activation loop. Although CAMKK2 may also phosphorylate AMPK in response to enhanced Ca
2+ concentration triggered by upstream signals
16-18, LKB1 complex is regarded as the major kinase to phosphorylate Thr172 of AMPK under the nutrient deprivation condition
15,19-22. Therefore, glucose signals to AMPK in a phosphorylation-dependent manner, which is different from the amino acid-sensing process involving small GTPase-dependent activation of mTORC1.
Compared with the well-annotated amino acid-sensing/mTORC1 activation process, how the LKB1-AMPK axis senses nutrients has been a long-standing question. Recently, multiple mechanisms have been found to connect the regulation of LKB1-AMPK axis to various metabolites. For example, it was reported that ribulose-5-phosphate (Ru5P), a product of the pentose phosphate pathway (PPP), could break the integrity of LKB1 kinase complex and inhibit the activation of AMPK
23. Two free amino acids, Asp and Asn, were reported to modulate LKB1 complex-mediated AMPK activation
24, although the detailed mechanisms remain unclear. Importantly, recent studies revealed that LKB1 is recruited to the lysosome by AXIN under the glucose starvation condition to phosphorylate AMPK, while fructose-1,6-bisphosphate (FBP), a metabolite in the glycolysis pathway, could bind to aldolase A (ALDOA) to impair the recruitment of the LKB1 complex to the lysosome membrane and inhibit AMPK activation
15,19. These results implicate a clear but indirect link between glucose metabolism and the LKB1-AMPK pathway. However, since only a fraction of LKB1 and AMPK are recruited to the lysosome, there may be other mechanisms for the LKB1-AMPK axis to sense the glucose starvation condition.
Glycolysis is a central metabolic pathway composed of a series of enzymatic reactions that break down glucose to produce ATP and essential intermediates for biosynthesis. Within this pathway, ALDOA catalyzes a key step by cleaving the six-carbon FBP into two triose phosphates: dihydroxyacetone phosphate (DHAP) and glyceraldehyde 3-phosphate (G3P). Downstream of ALDOA, phosphoglycerate kinase 1 (PGK1) converts 1,3-bisphosphoglycerate (1,3-BPG) to 3-phosphoglycerate (3PG), generating one molecule of ATP. Although PGK1 has been reported to exert diverse functions as both a metabolic enzyme
25 and a protein kinase
26,27, its role in the context of signaling transduction and nutrient sensing remains largely unexplored.
Herein, we found that depletion of PGK1 caused activation of the LKB1-AMPK pathway in an AXIN-independent manner, and the glycolytic metabolites downstream of PGK1, 3PG, 2-phosphoglycerate (2PG), and pyruvate (PYR) reversed the activation of AMPK caused by PGK1 knockdown. With molecular dynamics (MD) simulation and in vitro assays, we further found that 3PG, 2PG, PYR could dynamically bind to and inhibit the kinase activity of the LKB1 complex to phosphorylate AMPK, and identified the key amino acid residues of LKB1 mediating the metabolite binding. Following these clues, we generated mice expressing LKB1 mutants that are deficient in metabolite interaction, and observed an increase in AMPK activation and resistance phenotypes to a high-fat diet (HFD) in them. Moreover, by generating PGK1 mutant mice and a potent PGK1 inhibitor, we demonstrate that PGK1 activity is crucial for modulating AMPK activity. Thus, we reveal a distinct lysosome-independent glucose-sensing mechanism, characterize LKB1 as a direct metabolite sensor, and further propose PGK1 as a potential drug target to modulate the LKB1-AMPK pathway.
RESULTS
Glycolytic metabolites downstream of PGK1 suppress AMPK activation
PGK1 is an enzyme catalyzing the conversion of 1,3-BPG to 3PG in the glycolysis pathway (Supplementary Fig. S1a). Through phosphoproteomic analysis, we first observed that PGK1 depletion caused alterations of multiple signaling pathways, among which the activation of AMPK (PRKAA1/2) was the most significant one (Fig. 1a; Supplementary Fig. S1b–e). Then we further validated the activation of AMPK signaling triggered by PGK1 knockdown via immunoblot (Fig. 1b) and found that PGK1 depletion caused AMPK activation in a high-glucose condition, but not in the glucose starvation condition (Fig. 1c). ALDOA has been identified as an FBP receptor that signals to AMPK, and the accumulation of FBP will suppress LKB1-dependent AMPK activation by disrupting the v-ATPase–Ragulator–AXIN–LKB1 complex, while depletion of ALDOA elevates AMPK signaling since the FBP-sensing mechanism is disrupted
15. Thus we wonder whether the AMPK activation caused by PGK1 knockdown is due to a reduction of cellular FBP. Surprisingly, although knockdown of ALDOA and PGK1 both activated AMPK (Fig. 1d; Supplementary Fig. S1j), quantitative analysis of the glycolytic metabolites showed that PGK1 knockdown significantly increased the amount of FBP, but greatly reduced the metabolites downstream of PGK1, such as 3PG, 2PG, PYR, and lactate (Fig. 1e; Supplementary Fig. S1f, k). And the ratios of AMP/ATP and ADP/ATP were not significantly affected by the depletion of ALDOA or PGK1 (Fig. 1f; Supplementary Fig. S1g–i, l–o). Therefore, the activation of AMPK caused by PGK1 depletion was not because of changes in FBP or the ratios of AMP/ATP and ADP/ATP.
Then we speculated that the decrease in the metabolites downstream of PGK1, but not FBP, may be the reason for the activation of AMPK. Interestingly, the addition of 3PG, 2PG, and PYR in the presence of streptolysin O (SLO) markedly suppressed the AMPK pathway in glucose-starved cells pre-treated with various shRNAs, while FBP had minimal effect in ALDOA-depleted cells (Fig. 1g) and SLO treatment itself did not alter the AMPK pathway (Supplementary Fig. S2a–f). So the effects of 3PG, 2PG, and PYR are possibly different from those of FBP. To further dissect the difference between the observed PGK1/AMPK regulation and the established ALDOA-FBP/AXIN-LKB1-AMPK pathway, we generated
AXIN knockout SNU-739 cells with the CRISPR/Cas9 gene editing method. Consistent with a previous report
28, the lysosome localization of LKB1 under the glucose starvation condition was largely abolished and the AMPK phosphorylation was greatly decreased in
AXIN-null cells (Fig. 1h, i). Moreover, the addition of FBP hardly had any effect on suppressing AMPK in glucose-starved
AXIN-null cells (Fig. 1i). In contrast, 3PG treatment or deletion of PGK1 by siRNA could still affect AMPK activation status in both parental control cells and
AXIN-null cells (Fig. 1i–k). On the other hand, depletion of GAPDH, the metabolic enzyme between ALDOA and PGK1 in the glycolytic pathway, did not affect the effects of metabolite treatment (Supplementary Fig. S2g). Furthermore, lithocholic acid (LCA) binds to TUB-like protein 3 (TULP3), activating the TULP3-sirtuin-v-ATPase-AMPK pathway
29. In TULP3-knockdown 293T cells, either 3PG treatment or PGK1 depletion via siRNA still modulated AMPK activation (Supplementary Fig. S2h, i). Thus these results suggest that glycolytic metabolites downstream of PGK1, such as 3PG, may modulate AMPK activation through mechanisms independent of FBP-ALDOA /AXIN-LKB1 axis and TULP3.
Glycolytic metabolites impair the activity of the LKB1 complex
It is reported that some metabolites such as Ru5P and free Asp/Asn may regulate the integrity or the kinase activity of the LKB1 complex (LKB1/MO25/STRAD) to modulate LKB1-dependent AMPK activation, which could be examined by
in vitro kinase assay
23,24. We purified the LKB1 complex from 293F cells and purchased a commercial LKB1 complex to perform
in vitro kinase assays with recombinant AMPKα1 as the substrate in the presence of various glycolytic metabolites. As detected by immunoblot and LKBtide kinase assay, 3PG, 2PG, and PYR dramatically inhibited the LKB1 complex-mediated AMPK phosphorylation (Fig. 2a, b; Supplementary Fig. S3a–e). And such an effect was dose-dependent since increasing concentrations of metabolites (3PG, 2PG, PYR) caused more significant inhibition of AMPK phosphorylation (Fig. 2b; Supplementary Fig. S3c). Since the three metabolites all inhibit the enzymatic activity of the LKB1 complex to phosphorylate AMPK, we continued to measure the inhibition constant (
Ki) of 3PG, 2PG, and PYR. As indicated (Fig. 2c–h), 3PG, 2PG, and PYR all efficiently suppressed the enzymatic activity of the LKB1 complex with
Ki values of 14.70, 22.92, and 11.50 (μM), respectively. Notably, metabolite treatment had minimal effect on CAMKK2-mediated AMPK phosphorylation
in vitro (Fig. 2i; Supplementary Fig. S3f–j) and decreased AMPK phosphorylation in cells with CAMKK2 depletion (Supplementary Fig. S3k).
LKB1 also phosphorylates other substrates in addition to AMPK
30-34 to modulate multiple biological processes. To determine whether the glycolytic metabolites generally inhibit the kinase activity of the LKB1 complex, we generated recombinant GST-SIK1 (1–303) protein and recombinant GST-fusion peptides of the reported LKB1-phosphorylation motifs of NUAK2, BRSK2, and MARK3 to perform an
in vitro kinase assay, and found that the phosphorylation of these LKB1 substrate motifs was all reduced by the metabolites
in vitro (Supplementary Fig. S4a–f). Moreover, depletion of PGK1 increased while the metabolite treatment decreased the phosphorylation of endogenous SIK kinases in a manner consistent with the changes in p-AMPK signal (Supplementary Fig. S4g, h).
Based on these results, we reason that the coexistence of cellular metabolites (3PG, 2PG, and PYR) may superimpose their effects in impairing LKB1 complex-mediated AMPK phosphorylation. To examine whether such an effect exists under physiological conditions, we collected skeletal muscle and liver tissues from mice under three conditions (regular-fed, overnight-fasted and fasted-refed for 2 h; 5 mice for each group), and measured the physiological concentrations of FBP, 3PG, 2PG, and PYR, as well as adenine nucleotides (ATP, ADP, and AMP), by LC-MS. Importantly, we found that the concentrations of these metabolites varied much among the three conditions in liver tissues, which was correlated with glycogen and blood glucose levels but inversely correlated with p-AMPK levels. In contrast, the amount of FBP did not change much in the skeletal muscle tissues (Fig. 2j, k; Supplementary Fig. S5a–d). On the other hand, adenine nucleotide levels and their relative ratios in the skeletal muscle tissues fluctuated more dramatically than those in liver tissues (Supplementary Fig. S5e–h). Notably, the total concentrations of the three metabolites (3PG, 2PG, and PYR) in the livers of regularly fed and fasted mice were about 170–180 μM and 70–80 μM, respectively (Fig. 2j). Then we performed in vitro kinase assays with the purified LKB1 complex and recombinant AMPKα1 in the presence of physiological concentrations of these metabolites. Importantly, the metabolites at concentrations corresponding to those in fed mouse liver tissues were more effective in suppressing LKB1 complex-mediated AMPK phosphorylation (Fig. 2l). These results suggest that glycolytic metabolites have an inhibitory effect against the LKB1 complex under physiological conditions.
Glycolytic metabolites dynamically interact with LKB1 and cause its allosteric changes
To elucidate the inhibitory effect of the metabolites on the LKB1 complex, we utilized the reported crystal structure of the LKB1 complex
35 to perform extensive MD simulations of LKB1 solvated with water and the metabolites to dissect the possible structural basis. To comprehensively sample binding possibilities, we initiated simulations from four distinct starting positions for the metabolite (3PG as an example), which were selected based on the electrostatic potential distribution across the LKB1 surface. A key finding from these simulations was that all four initial conformations dynamically converged into a single, large positively charged binding region (Supplementary Fig. S6a, b). This demonstrates the existence of a dominant and preferred binding pattern for 3PG on LKB1. Moreover, the analysis revealed a mechanism involving two correlated sub-sites within this converged binding region. 3PG may briefly interact with the exterior lysine 83 and arginine 87 (K83/R87) residues of LKB1 kinase, which facilitates its interaction with the relatively internal pocket formed by lysine residues 62, 64, 81, and 124 (K62/K64/K81/K124). The latter interaction renders a shift of the activation loop of the LKB1 kinase, which may eventually result in a reduction in enzymatic activity of the LKB1 complex (Fig. 3a, b; Supplementary Fig. S6c). MD simulation analysis showed that similar binding mechanisms also applied to 2PG and PYR, although the interaction with 3PG caused the most dramatic changes in LKB1 (Supplementary Fig. S6d–f). Interestingly, all the key residues mentioned above are highly conserved across evolution, indicating that they may be important for LKB1 regulation and function (Supplementary Fig. S7a). To validate the interaction between the metabolites and LKB1, we performed hydrogen–deuterium exchange mass spectrometry (HDX-MS) on LKB1 in the absence or presence of 3PG/2PG/PYR. The results showed that all three metabolites reduced the hydrogen-deuterium exchange of the peptides located in the predicted binding region, indicating the existence of the interaction (Supplementary Fig. S7b–e). Interestingly, the effect of 3PG is more dramatic than those of 2PG and PYR, which was consistent with the prediction of MD simulation analysis. Notably, the metabolites affected the HDX of the glycine-rich loop (56-GEGSYG-61) of LKB1 (responsible for ATP binding
36) (Supplementary Fig. S7f–h), and consistently, the interaction between the LKB1 complex and ATP was impaired by the metabolites (Fig. 3c).
MD simulations suggested that the population times of the three metabolites in the allosteric binding pocket all decreased for LKB1-K83AR87A (impairing the exterior interaction site) and LKB1-K64EK81E (disrupting the four-lysine pocket) mutants (Fig. 3b; Supplementary Fig. S6e, f). To validate the predicted interaction model, we generated constructs encoding two double mutants, and purified LKB1 complex with them to perform in vitro kinase assays. Importantly, in contrast to the WT control, metabolite treatment had only a minimal effect on blocking AMPK phosphorylation by kinase complexes containing LKB1-K83AR87A and LKB1-K64EK81E (Fig. 3d, f, h). Moreover, we also examined the interaction between purified LKB1 complex and the metabolites (3PG, 2PG, PYR, and FBP) by MicroScale Thermophoresis (MST) method. Interestingly, we observed relatively strong binding affinities of 3PG, 2PG, PYR to LKB1 complex with dissociation constants (Kd) at about 15–25 μM range (Fig. 3e, g, i), while no obvious affinity was found between 3PG/2PG/PYR and CAMKK2 kinase, which may be due to the differences between LKB1 and CAMKK2 at several arginine/lysine residues (Supplementary Fig. S8a–d). Notably, FBP did not interact with the LKB1 complex in the MST assay either (Supplementary Fig. S8e), in keeping with the results of the in vitro kinase assay. Consistently, the interaction between complexes containing the LKB1 mutants (K83AR87A and K64EK81E)and the three metabolites was significantly decreased by up to 10–20-folds compared with WT LKB1 kinase complex (Fig. 3e, g, i). Notably, mutation of the four key residues (K83AR87A and K64EK81E) did not affect the interaction between LKB1 and AXIN, and metabolite (3PG, 2PG, and PYR) treatment did not affect the integrity of LKB1 complex (Supplementary Fig. S8f, g). And the LKB1 mutants (K83AR87A and K64EK81E) both formed kinase complexes that had kinase activity very similar to WT LKB1 complex (Supplementary Fig. S8h–m). Taken together, these results suggest that LKB1 is a direct nutrient sensor, and glycolytic metabolites directly interact with LKB1 complex to allosterically inhibit its kinase activity.
Metabolite-dependent LKB1 regulation plays a role in body metabolism control
To evaluate the metabolite-sensing effect of LKB1, we stably expressed LKB1-K83AR87A and K64EK81E mutants as well as WT LKB1 in
Lkb1-null mouse tumor cells
37, and examined their responses to metabolite treatment. As indicated, the phosphorylation of AMPK and acetyl-CoA carboxylase (ACC) was reduced by metabolite treatment in cells expressing WT LKB1, but not LKB1-K83AR87A and LKB1-K64EK81E mutants (Fig. 4a). On the other hand, depletion of endogenous PGK1 increased the p-AMPK and p-ACC levels in cells expressing WT LKB1, but not in parental
Lkb1-null cells or cells expressing LKB1-K83AR87A and LKB1-K64EK81E mutants (Fig. 4b). Interestingly, the LKB1-K83AR87A and K64EK81E mutants could still be recruited to the lysosome under glucose starvation condition (Supplementary Fig. S9a). Consistently, the p-AMPK and p-ACC signals in cells expressing LKB1-K83AR87A and K64EK81E mutants were resistant to 3PG treatment but could still be inhibited by FBP (Supplementary Fig. S9b), suggesting that the metabolite-sensing function of LKB1 is independent of the reported ALDOA-FBP/AXIN-LKB–AMPK pathway. Importantly, expression of LKB1 mutants and 3PG treatment also affected the phosphorylation of SIK3, another known LKB1 substrate (Supplementary Fig. S9c). These results support a key role for the LKB1–metabolite interaction in regulating LKB1 activity and AMPK pathway.
To further investigate the physiological role of LKB1 metabolite-sensing function, we then generated mice expressing LKB1-K83AR87A mutant (Lkb1K83AR87A/+) and LKB1-K64EK81E mutant (Lkb1K64EK81E/+) by a gene editing method (Supplementary Fig. S9d). From our breeding crosses, we obtained 318 heterozygous and 148 WT offspring for the Lkb1-K83AR87A line, and 93 heterozygous and 54 WT offspring for the Lkb1-K64EK81E line. The complete absence of homozygous mutants from both lines indicates that homozygosity for these mutations may be embryonically lethal. Consequently, all subsequent experiments were performed using heterozygous animals. Interestingly, p-AMPK and p-ACC in liver tissues from WT mice were much lower under the regular-fed condition compared with the overnight-fasted condition, while the same group of phosphorylation signals largely remained in the liver tissues from Lkb1K83AR87A/+ and Lkb1K64EK81E/+ mice under the regular-fed condition as detected by immunofluorescence methods (Supplementary Fig. S9e, f). It is well known that the AMPK pathway plays vital roles in maintaining metabolic homeostasis, so we continued to challenge the Lkb1K83AR87A/+ and Lkb1K64EK81E/+ mice with an HFD and examined their phenotypes. Importantly, we found that the gain of body weight of both mutant mice was significantly less than that of WT mice under the HFD condition, which may be due to a much lower composition of fat tissue (Fig. 4c–f).
Furthermore, in the oral glucose tolerance test (OGTT) and insulin tolerance test (ITT) assays, the blood glucose levels of both mutant mice were only slightly enhanced after HFD treatment, in sharp contrast to the WT mice (Fig. 4g, h; Supplementary Fig. S10a, b). Moreover, the mutant mice had fewer liver and serum triglycerides than their WT counterparts (Fig. 4i, j; Supplementary Fig. S10c, d). Notably, both mutant mice showed increased oxygen consumption, heat and CO2 generation, and respiratory exchange ratio (RER), but not food intake compared with WT mice, (Fig. 4k, l; Supplementary Fig. S10e–l). We also collected metabolic tissues such as liver, skeletal muscle, and brown adipose tissue (BAT), and determined the activation status of AMPK pathway. As indicated, the p-AMPK and p-ACC levels were much higher, while the p-S6K levels were lower in tissues from both mutant mice than those in WT mice, and the mutant mice had less lipid composition in the liver after being challenged with HFD (Fig. 4m–p; Supplementary Fig. S11a, b). Importantly, LKB1 was expressed at similar levels and exhibited comparable in vitro kinase activity in the absence of metabolites (Supplementary Fig. S11c–e). These results suggest that disrupting the glycolytic metabolite-mediated LKB1 suppression could greatly alleviate the metabolic phenotypes caused by excessive nutrients.
Manipulation of PGK1 activity modulates body metabolism
Since all three metabolites (3PG, 2PG, PYR) that suppress LKB1 kinase activity are downstream of PGK1 in the glycolysis metabolism, we reason that the alteration of PGK1 enzymatic activity may cause metabolic phenotypes via the metabolite-sensing mechanism of LKB1. Previously, we found an elevation of PGK1 expression in hepatocellular carcinoma and identified K323 acetylation as an important modification that promotes PGK1 activity
25, and the acetylation-mimetic PGK1-K323Q mutant exhibited elevated enzymatic activity. Then we generated mice expressing PGK1-K323Q mutant by editing the
Pgk1 allele on the X chromosome and used male mice (namely
Pgk1K323Q/Y) for experiments (Supplementary Fig. S12a). Importantly, the liver tissues from
Pgk1K323Q/Y mice had lower phospho-AMPK signals and higher fat composition, compared with WT mice (Fig. 5a, b). Consistently, there were more glycolytic metabolites downstream of PGK1 (3PG, 2PG, PYR) in the liver tissues from
Pgk1K323Q/Y mice compared with WT mice (Fig. 5c; Supplementary Fig. S12b), implicating that the increased PGK1 activity may generate more metabolites to suppress the LKB1 complex and the activity of AMPK. Interestingly, the body weight of
Pgk1K323Q/Y mice increased faster than their WT counterparts, and higher glucose levels were observed in both OGTT and ITT assays (Fig. 5d, e; Supplementary Fig. S12c). Consistent with the insulin-resistant phenotypes,
Pgk1K323Q/Y mice showed lower oxygen consumption, and reduced heat and CO
2 generation (Supplementary Fig. S12d–f). These results suggest that the enzymatic activity of PGK1 may be a possible switch to regulate the LKB1-AMPK pathway and metabolic phenotypes.
To further investigate the potential of PGK1 as the possible target to modulate LKB1-AMPK pathway and regulate body metabolism, we developed a potent PGK1 enzymatic inhibitor MCB-330 (marked as PGK-I) based on the structure of Terazosin, a reported binding molecule and putative activator of PGK1
38 (Fig. 5f; Supplementary Fig. S13a, b). MCB-330 strongly inhibits PGK1 enzymatic activity
in vitro and it has a potent binding affinity to the WT PGK1, but not the PGK1-S256AV342AF343A mutant that was designed based on the predicted interaction model (Fig. 5g, h; Supplementary Fig. S14a–c). Importantly, in keeping with a key role of PGK1 activity in modulating the LKB1-AMPK axis, MCB-330 treatment dramatically enhanced the activation of AMPK pathway in a dose-dependent manner in cells expressing the WT PGK1, but not in cells expressing the PGK1-S256AV342AF343A mutant (Fig. 5i; Supplementary Fig. S14d). Notably, MCB-330 treatment also increased p-AMPK signal in cells depleted of TULP3, the reported binding protein of lithocholic acid to activate lysosome-dependent LKB1-AMPK pathway (Supplementary Fig. S14e). Then we further treated cells with MCB-330 and measured the changes in the glycolytic metabolites. As indicated, chemical treatment strongly increased the upstream metabolites, such as FBP, but reduced the metabolites downstream of PGK1 (3PG, 2PG, PYR, and lactate) without significantly affecting the AMP/ATP and ADP/ATP ratios (Fig. 5j, k; Supplementary Fig. S14f–o), suggesting an effective inhibition of PGK1 enzymatic activity. Next, we continued to investigate whether targeting PGK1 with MCB-330 would have any effect in mice. Oral administration of MCB-330 (10 mg/kg) dramatically enhanced the phosphorylation of AMPK and ACC in the tissues from WT mice, but had much less effect on tissues from
Lkb1K83AR87A/+ mice (Fig. 5l; Supplementary Fig. S15a), supporting the existence of PGK1-metabolite-LKB1 axis. To evaluate the potential of targeting PGK1 to treat metabolic abnormalities, we used MCB-330 to treat ob/ob mice with metformin as a positive control. Strikingly, continuous treatment of MCB-330 (14 days, 10 mg/kg/day) was as effective as metformin (14 days, 100 mg/kg/day) in reducing the levels of blood glucose, insulin, and triglycerides (Fig. 5m; Supplementary Fig. S15b), and in enhancing insulin sensitivity (Fig. 5n, OGTT assay). And such effects were well correlated with elevated AMPK phosphorylation and reduced glycolytic metabolites post MCB-330 treatment (Fig. 5o; Supplementary Fig. S15c–e). Additionally, we examined the tissue distribution (Supplementary Tables S1–7) and pharmacokinetic parameters (Supplementary Tables S7, S8) of MCB-330 in blood, liver, pancreas, and skeletal muscle following oral administration in mice. These results suggest that the inhibition of PGK1 activity is a promising means to activate AMPK pathway and treat metabolic diseases, via alleviating metabolite-mediated LKB1 suppression.
DISCUSSION
LKB1 has been shown to be recruited to the lysosome membrane via its interaction with AXIN, which binds v-ATPase and Ragulator in the glucose deprivation condition
19,28. The LKB1 complex then phosphorylates and activates AMPK regardless of the minimal changes in the AMP/ATP and ADP/ATP ratios
15. So, this mechanism is regarded as a nutrient-sensing pathway to activate AMPK in parallel with the canonical AMP-dependent energy-stress-sensing pathway. The accumulation of FBP impairs such mechanism through its interaction with ALDOA and promotes the ALDOA–v-ATPase interaction to release the AXIN–LKB1 complex from the lysosome and to diminish LKB1-dependent AMPK activation
15. Thus ALDOA serves as the nutrient sensor of the regulatory circuit while AXIN mediates the lysosomal membrane localization of LKB1. In the current study, we found that AXIN knockout indeed disrupted the localization of LKB1 to the lysosome, and reduced the level of p-AMPK, and abolished the effect of FBP treatment on p-AMPK. However, p-AMPK still responds to the addition of 3PG and PGK1 knockdown in the absence of AXIN and TULP3, suggesting alternative glucose-sensing mechanisms may exist. Following these clues, we further found that glycolytic metabolites (3PG, 2PG, PYR) directly bind LKB1 complex and impair its kinase activity to phosphorylate AMPK. Moreover, the LKB1 mutants (K83AR87A and K64EK81E) with impaired metabolite interaction still bind to AXIN and are recruited to the lysosome in glucose-starvation condition. In a previous study
39, 3PG was reported to increase LKB1-mediated AMPK phosphorylation and activity
in vitro, which is opposite to our results. However, the 3PG concentrations (0.5–10 mM) used in that study were much higher (10–100-folds) than the total physiological concentration of 3PG and 2PG we have measured in mouse liver and muscle tissues (30–100 μM). Therefore, we think that the discrepancy is due to different experimental conditions. Importantly, no homozygous mutant mice were obtained from either the LKB1-K83AR87A or LKB1-K64EK81E breeding lines. This suggests that metabolite-dependent inhibition of LKB1 may be essential for early development, and its complete disruption leads to embryonic lethality. These results suggest that LKB1 could act as a direct nutrient sensor connecting glucose metabolism to AMPK activation independent of the lysosome, revealing a distinct glucose-sensing mechanism complementary to the established lysosome glucose-sensing pathway.
Previously, Ru5P was found to impair LKB1 complex-mediated AMPK activation by disrupting the integrity of the LKB1–STRAD–MO25 complex
23, although the exact mechanism remained elusive. By structural simulation, we found that the binding of 3PG/2PG/PYR may cause allosteric changes in LKB1 kinase, resulting in a shift of its activation loop in the kinase domain to impair the enzymatic activity of LKB1 complex but not affect the complex integrity. In the interaction process, the exterior residues K83 and R87 of LKB1 may serve as the “guide” to initiate the recruitment of metabolites, and facilitate their binding to the relative inside pocket formed by other positively charged residues (K62/K64/K81/K124), which causes the conformation change in LKB1. The interaction mechanism may be affected by post-translational modifications on the key lysine residues under certain physiological conditions. For example, K64 of LKB1 was reported to be acetylated
40. Therefore, the interaction-inhibitory model we have identified may provide new regulatory potential for LKB1-AMPK pathway.
MD simulation analysis showed that similar binding mechanisms also applied to 2PG and PYR, although the interaction with 3PG caused the most dramatic changes in LKB1, which was supported by similar binding affinities (
Kd, ~15–25 μM) to the LKB1 complex and inhibitory effects in
in vitro kinase assays and culture cell-based assays. Furthermore, since 3PG, 2PG and PYR are predicted to interact with LKB1 at the same region with similar affinities, their effect on blocking LKB1 kinase activity may add together under physiological conditions. Interestingly, we found that the sum of 3PG/2PG/PYR in liver tissues from fasted mice was about 70–80 μM, while that from regular-fed mice was about 170–180 μM. And the 170–180 μM more significantly suppressed LKB1-mediated AMPK phosphorylation
in vitro than 70–80 μM of mixed metabolites (3PG/2PG/PYR), which was consistent with the difference in AMPK phosphorylation status in the mouse liver tissues from fed/fasted conditions (Fig. 2j). Thus we reason that the total amount of the three metabolites (instead of every single one) may be important in determining the activation status of AMPK. Moreover, different from FBP, the upstream six-carbon glycolytic metabolite, 3PG, 2PG, and PYR are three-carbon metabolites positioned at the interface between glycolysis and gluconeogenesis pathways that are regulated by AMPK
41,42. Accumulation of these gluconeogenic precursors may suppress AMPK activity via LKB1, thus promoting gluconeogenesis. Dynamic fluctuations in their levels modulate LKB1-AMPK signaling, which in turn regulates metabolic flux through both pathways, establishing a feedback mechanism critical for glucose homeostasis under physiological conditions. Moreover, 3PG/2PG serve as precursors for serine biosynthesis
43, while PYR participates in the TCA cycle and alanine synthesis. Therefore, the variation in other metabolic pathways may also affect the LKB1-AMPK signaling via the changes in 3PG/2PG/PYR to maintain the flexibility and sensitivity of cellular metabolic homeostasis. It is also possible that the LKB1 complex may interact with and be affected by other metabolites to create distinct sensing mechanisms for nutrients including glucose, which warrants more efforts to explore.
As a critical enzyme in glycolysis, PGK1 catalyzes the conversion of 1,3-bisphosphoglycerate to 3PG and generates ATP, serving as one of the only two ATP-producing enzymes in glycolysis. Elevated expression and activity of PGK1 have been observed in multiple cancers and are regarded as an important factor to drive the enhanced glycolysis of tumor cells, which is known as “Warburg effect”
44. PGK1 was also reported to function as a protein kinase and to regulate autophagy, mitochondrial function, and apoptosis to promote cancer progression
45. On the other hand, genetic mutation of PGK1 was also linked to hemolytic anemia and Parkinson’s disease
46,47. So PGK1 is an important physiological regulator. We first noticed that depletion of PGK1 activated the AMPK pathway, and we further discovered that glycolytic metabolites downstream of PGK1 (3PG/2PG/PYR) directly interacted with LKB1 and suppressed its enzymatic activity. Thus PGK1 may have a unique function in regulating LKB1-AMPK pathway by generating downstream metabolites. Following this direction, we generated
Pgk1K323Q/Y mice with increased PGK1 activity and found that indeed the levels of downstream metabolites increased and AMPK pathway was suppressed in the liver tissues of
Pgk1K323Q/Y mice. And
Pgk1K323Q/Y mice had higher fat composition and gained more body weight than the WT mice. On the other hand, by developing a potent PGK1 inhibitor MCB-330, we found that the inhibition of PGK1 reduced downstream metabolites and enhanced the p-AMPK signal in both cultured cells and mouse liver tissues. Importantly, MCB-330 treatment at a low dose (10 mg/kg/day) effectively dropped the levels of blood glucose, serum insulin, and serum triglycerides, in a manner very similar to metformin (100 mg/kg/day). These results suggest that PGK1 activity may be a potential target to modulate LKB1-AMPK pathway.
Taken together, our current study demonstrates a lysosome-independent glucose-sensing pathway via the direct interaction between LKB1 and glycolytic metabolites, and also implies a new means of elevating AMPK activation by targeting PGK1 (Fig. 6). These findings may contribute to the understanding of the complex nutrient-sensing network as well as shed light on intervening in metabolic abnormalities by developing specific inhibitors.
Although MD simulations predicted the binding modes and interaction regions of three glycolytic metabolites (3PG, 2PG, and PYR) with the LKB1 complex, and HDX experiments further validated these potential binding sites, direct structural evidence of their binding conformations remains elusive. Notably, to avoid aggregation during large-scale purification, we used truncated forms of LKB1 and STRAD. Consequently, the kinetic parameters we measured may differ from those of complexes comprising full-length proteins. Moreover, while metabolites reduced the interaction between the LKB1 complex and ATP, the functional significance of this reduction for LKB1 kinase activity remains unclear. We attempted to crystallize the LKB1 complex in the presence of these metabolites but were unsuccessful, likely due to their relatively weak binding affinity (Kd in the tens of micromolar range) and the dynamic binding-unbinding equilibrium observed in MD simulations — both of which complicate crystallization. MD simulations identified a conserved binding pocket in LKB1 formed by residues K62, K64, K81, K83, K84, R87, K122, and K124, which accommodates these metabolites. In addition to 3PG, 2PG, and PYR, we assessed other glycolytic intermediates (e.g., lactate and citrate) using MST and in vitro kinase assays; however, none of them exhibited detectable binding to the LKB1 complex. Due to experimental limitations, we chose to focus on glucose metabolism and could not systematically screen metabolites from other metabolic pathways that might interact with LKB1 in a similar manner.
Using an in vitro kinase assay and a commercially available phospho-SIK antibody, we found that metabolites also inhibited the LKB1-dependent T-loop phosphorylation of AMPK-related kinases. To maintain focus on the LKB1-AMPK axis and due to limited resources, our analysis was restricted to measuring the phosphorylation and total protein levels of SIKs. We did not investigate their expression patterns or activity regulation in greater depth. We developed a PGK1 inhibitor as the proof-of-concept tool to validate the proposed metabolite-dependent LKB1 regulation in vivo. And the PGK1 inhibitor alleviated some metabolic phenotypes of ob/ob mice in a manner similar to that of metformin. However, these results are still preliminary to support the use of the PGK1 inhibitor for clinical purposes. More efforts are required to fully determine the possible risks and side-effects of PGK1 inhibitors, since many organs and bioprocesses highly depend on glycolysis metabolism.
In summary, our integrative approach combining MD simulations and biochemical assays provides strong evidence that LKB1 can directly sense glycolytic metabolites (3PG, 2PG, and PYR). Nevertheless, further structural and functional studies are required to elucidate the precise mechanism by which these metabolites regulate LKB1 activity and determine whether additional metabolites possess similar regulatory potential.
MATERIALS AND METHODS
Antibodies and reagents
Rabbit anti-phospho-AMPKα-T172 (cat. #2535, 1:2,000 for immunoblot (IB), 1:200 for immunofluorescence (IF)), anti-AMPKα (cat. #2532, 1:2,000 for IB), anti-phospho-ACC-Ser79 (cat. #3661, 1:1,000 for IB), anti-ACC (cat. #3662, 1:1,000 for IB), anti-ALDOA (cat. #8060, 1:2,000 for IB), anti-AXIN1 (cat. #2074, 1:1,000 for IB), anti-LKB1 (cat. #3050, 1:2,000 for IB), anti-MO25α (cat. #2716, 1:1,000 for IB), anti-UCP1 (cat. #72298, 1:2,000 for IB), anti-p70 S6 Kinase (cat. #2078), anti-phospho-p70 S6 Kinase (cat. #9205), anti-SIK2 (cat. #6919), anti-SIK3 (cat. #39477), Mouse anti-LAMP1 (cat. #15665, 1:100 for IF) anti-phospho-Threonine (cat. #9386, 1:2,000 for IB) were purchased from Cell Signaling Technology. Rabbit anti-phospho-CAMK4 (cat. #ab195000) and anti-phospho-SIK (cat. #ab199474) were purchased from Abcam. Mouse anti-Vinculin (V4505) and rabbit anti-FLAG (F7425) were purchased from Sigma-Aldrich. Mouse anti-PGK1 (sc-48342), anti-STRADα (sc-515635) were purchased from Santa Cruz Biotechnology. Mouse anti-CAMK4 (cat. #68452-1-Ig), anti-LKB1 (cat. #68016-1-Ig), anti-GST (cat. #66001-2-Ig), Rabbit anti-CAMKK2 (cat. #11549-1-AP), anti-TULP3 (cat. #13637-1-AP), the HRP-conjugated goat anti-mouse IgG (SA00001-1, 1:5,000 for IB) and HRP-conjugated goat anti-rabbit IgG (SA00001-2, 1:5,000 for IB) antibodies were purchased from Proteintech. Rabbit anti-Calmodulin was purchased from ABclonal. Insulin (cat. #P3376), DAPI (cat. #C1002) were purchased from Beyotime Biotechnology. 3PG (cat. #P8877), 2PG (cat. #19710), G3P (cat. #G5251), sodium pyruvate (cat. #P5280) were purchased from Sigma. Glucose (cat. #63005518) was purchased from Sinopharm Chemical Reagent. FBP (cat. #sc-214805) was purchased from Santa Cruz Biotechnology. PEP (cat. #A602553) and sodium lactate (cat. #A604046) were purchased from Sangon Biotech. Streptolysin O (SLO) (cat. #S4470) was purchased from Solarbio. Protease inhibitor cocktail (B14001) and Phosphatase inhibitor cocktail (B15001) were purchased from Selleck. LKB1 kinase complex protein (cat. #14-596) was purchased from Millipore. CAMKK2 protein (cat. #C18-10G) was purchased from Sino Biological. Improved Oil Red O staining kit (cat. #C0158) was purchased from Beyotime Biotechnology. Mouse Ultrasensitive Insulin ELISA (cat. #80-INSMSU-E01) was purchased from ALPCO. Phospho-AMPKα (Thr172) Sandwich ELISA Kit (cat. #7959) was purchased from Cell Signaling Technology. Micro Triglyceride Assay Kit (cat. #KTB2200), Micro Free Fatty Acid Assay Kit (cat. #KTB2230) and Micro Total Cholesterol Assay Kit (cat. #KTB2220) were purchased from Abbkine. Glycogen Assay Kit (cat. #MAK465) was purchased from Sigma-Aldrich.
Cell culture
Cell lines used in this study were purchased from Cell Bank, Chinese Academy of Sciences or kept by Prof. Daming Gao’s lab. HEK293T cells were maintained in DMEM with high glucose (25 mM) (Gibco), and SNU739 cells were maintained in RPMI-1640 (Gibco) (glucose 11.1 mM) both supplemented with 10% fetal bovine serum (FBS) and 1% streptomycin/penicillin at 37 °C under 5% CO2. All the cell lines were tested to confirm no mycoplasma contamination.
Animal models
All mice (C57BL/6J background) were maintained in a specific-pathogen-free (SPF) facility, and all related experiments were performed following the ethical guidelines and protocols approved by the Institutional Animal Care and Use Committee of the Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences. The procedures used followed the recommendations from the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC). Lkb1 mutant mice were generated by Shanghai Decode Biotechnology through CRISPR/Cas9-mediated genome editing. Pgk1-K323Q mutant mice were generated by Dr. Jinsong Li Lab (CAS Center for Excellence in Molecular Cell Science, Shanghai, China), referred to as Pgk1K323Q/Y mice in the manuscript (Pgk1 is located on the X chromosome, and “Y” represents the Y chromosome). OGTT, ITT, and metabolic cage studies were conducted using 18-week-old male heterozygous Lkb1 K83AR87A or Lkb1 K64EK81E and WT mice after 10 weeks (treatment started at 8-week-old) of HFD (60% fat calories) housed under specific pathogen-free conditions. All mice were randomly stratified into experimental groups. For the fast-refeed experiment, mice were fasted for 12 h (7:00 p.m. to 7:00 a.m.), a period synchronized with the dark cycle, during which food was withheld. In the refed group, food was restored for 2 h (7:00 a.m. to 9:00 a.m.) prior to tissue harvest and analysis.
Plasmids
The expression plasmids were constructed in the lentiviral vector pLEX-MCS-CMV-puro. The shRNAs were cloned into pLKO.1-puro vector. CRISPR-sgRNA targeting AXIN1 was used to knock out the gene in SNU-739 cells and was constructed in pWST-lenti. The siRNA, shRNA and sgRNA oligonucleotide sequences are listed as follows:
Luciferase siRNA (5′-TTCCTGGAACAATTGCTTTTA-3′),
Human PGK1 siRNA-1 (5′-AGGAAGAAGGGAAGGGAAATT-3′),
human PGK1 siRNA-2 (5′-ACAAACAACCAGAGGAUUATT-3′),
human PGK1 siRNA-3 (5′-ACAGAAGGCUGGUGGGUUUTT-3′),
human PGK1 shRNA (5’-CCGGGCCAAGATTGTCAAAGACCTACTCGAGTAGGTCTTTGACAATCTTGGCTTTTTG-3’),
human ALDOA shRNA (5’-CCGGCCGAGAACACCGAGGAGAACTCGAGTTCTCCTCGGTGTTCTCGGTTTTTG-3’),
human AXIN1 sgRNA-1: (5′-TCCAGTAGACGGTACAGCGA-3’),
human AXIN1 sgRNA-2: (5′-GCTGCTTACGGATCCTGTAT-3’),
human AXIN1 sgRNA-3: (5′-GGAAGCACGTACCCAAGTCA-3’),
human TULP3 siRNA-1: (5′-GAGAAGAGGCAAAGGAAAATT-3’),
human TULP3 siRNA-2: (5′-AUUUAAAGGUCCUAGGAAATT-3’),
human TULP3 siRNA-3: (5′-UGAGGAGACUGAUGGAAUATT-3’),
Human GAPDH siRNA: (5′-AACCAUGUAGUUGAGGUCATT-3’).
Transfection and lentivirus infection
Plasmids were transiently transfected in HEK293T with polyethylenimine. SiRNAs were synthesized from Biotend Company (Shanghai, China) and transfected with Lipofectamine 3000 at 50 nM final concentration according to the manufacturer’s instructions. Lentivirus overexpressing or knocking down target genes was packaged in HEK293T cells with polyethylenimine transfection. Lentivirus was incubated with target cells for 24 h with 10 mg/mL polybrene, and cells should recover for 24 h before selection. Infected cells were selected in 1 mg/mL puromycin until the uninfected control cells were dead.
Lkb1–/– mouse tumor cells were generated as described previously
37.
Immunoblot and immunoprecipitation
Cells were lysed in EBC buffer (50 mM Tris-HCl, pH 8.0, 120 mM NaCl, 0.5% NP-40) with protease inhibitors and phosphatase inhibitors (Selleck Chemicals). 30 μg total protein was separated by SDS-PAGE gel and blotted with the indicated primary antibodies. For immunoprecipitation, proper cell lysate was incubated with anti-Flag M2 agarose beads for 2 h, and then the pellets were washed 4 times with NETN buffer (20 mM Tris-HCl, pH 8.0, 100 mM NaCl, 1 mM EDTA, 0.5% NP-40) and resolved by SDS-PAGE and immunoblotted with the indicated antibodies.
Cell permeabilization assay
As described previously
48, cells were washed once with SLO permeabilization buffer (137 mM NaCl, 3 mM KCl, 2 mM MgCl
2, 0.1 mg/mL bovine serum albumin, 100 nM CaCl
2 and 3 mM EGTA, pH 7.2), and fresh SLO-permeabilization buffer containing the indicated metabolites (200 μM) and SLO (200 ng/mL) was added and then incubated at 37 °C for 15 min. Cells were harvested and determined by immunoblotting.
IF analysis
Cells grown on glass coverslips were treated under the indicated conditions, followed by fixation with 4% paraformaldehyde in PBS for 15 min and permeabilization with 0.1% Triton X-100 in PBS for 10 min. Then, cells were blocked for 1 h with blocking buffer (3% BSA in PBS) and incubated with primary antibodies in blocking buffer overnight. After washing three times with PBS (5 min for each wash), the coverslips were incubated with Alexa-488-conjugated goat anti-mouse secondary antibody and Alexa-594-conjugated goat anti-rabbit secondary antibody (Thermo Fisher Scientific) for 1 h. Cells were washed three times with PBS and then stained with 4,6-diamidino-2-phenylindole (DAPI) for 10 min. After being rinsed twice with PBS, coverslips were mounted onto slides using fluorescence mounting medium (Agilent, DAKO). Images were obtained by the Leica TCS SP8 fluorescence microscope.
For tissue section staining, after OCT removal through three 5-min PBS washes, tissue sections were permeabilized using 0.3% (v/v) Triton X-100 supplemented with 5% (w/v) bovine serum albumin (BSA) in PBS for 30 min at room temperature. After three PBS washes (5 min each), sections were incubated with primary antibodies diluted in antibody dilution buffer (PBS containing 3% BSA and 0.3% Triton X-100) overnight at 4 °C. Then the sections were rinsed three times with PBS, followed by incubation with Alexa-Fluor 488-conjugated anti-rabbit secondary antibody in antibody dilution buffer during 1-h light-protected incubation at room temperature. Following final PBS washes, nucleus staining was achieved by DAPI for 10 min, and slides were mounted and visualized on an Axioscan 7 (Zeiss).
Phosphoproteomic analysis
Sample Preparation for MS Analysis
After knockdown of PGK1 using siRNA or negative control (NC), cultured HEK293T cells were washed twice with ice-cold PBS and then lysed in SDS lysis buffer (4% SDS, 100 mM DTT, 100 mM Tris, pH 8.5). The lysates were sonicated at 20% amplitude for 1 min (5 s on and 5 s off) using an Ultrasonic Homogenizer (JY92-IIDN, NingBo Scientz Biotechnology), and subsequently denatured and reduced at 95 °C for 5 min. The lysates were centrifuged at 12,000× g for 10 min, and the supernatants were collected to determine the protein concentrations using a tryptophan fluorescence emission assay.
Protein digestion was performed using the SP4 method as previously described
49. Silica beads/glass spheres (Sigma, cat. #440345) were washed and suspended to a final concentration of 50 mg/mL in Milli-Q water, then added to 1 mg protein lysates at a 10:1 bead-to-protein ratio. Four volumes of 100% ACN were added, and mixed for 5 s at 500 rpm. After centrifugation at 16,000×
g for 5 min, the supernatants were aspirated and carefully washed three times with 80% ethanol. Protein aggregates were digested using a 1:50 trypsin-to-protein ratio in 100 mM ammonium bicarbonate for 18 h at 37 °C at 1,000 rpm. After digestion, peptides were acidified with 0.1% TFA and desalted using C18 columns.
The High-Select Fe-NTA kit (Thermo Fisher Scientific) was used for phosphopeptide enrichment following the manufacturer’s instructions. In brief, desalted peptides were reconstituted in 200 μL of 80% ACN/0.1% TFA and incubated with 50 μL Fe3+-NTA agarose beads for 20 min at room temperature. The mixture was then transferred to a filter tip, and the unbound peptides were collected from the clarified peptide flow-through by centrifugation. After washing three times with 200 μL 80% ACN/0.1% TFA and three times with 200 μL H2O, the bound phosphopeptides were eluted twice with 200 μL of 50% ACN/5% NH3·H2O and dried using a Speed-Vac. All centrifugation steps were performed at 50× g at room temperature.
LC-MS/MS Analysis
LC-MS/MS was performed on an Orbitrap Q-Exactive HFX mass spectrometer (Thermo Fisher Scientific) coupled with an EASY nLC 1200 liquid chromatograph (Thermo Fisher Scientific). Enriched phosphopeptides were separated using a home-made micro-tip C18 column using a 90 min gradient as follows: 2–5% B in 1 min; 5–28% B in 70 min; 28–38% B in 10 min; 38–100% B in 2 min; 100% B in 7 min. Data-independent acquisition was performed using Xcalibur software in profile spectrum data type. The MS1 full scan was set at a resolution of 120,000 at m/z 200, AGC target 3e6, and maximum IT 100 ms by orbitrap mass analyzer (350–1650 m/z), followed by 32 DIA isolation windows with variable width. The MS2 scans were generated by HCD fragmentation at a resolution of 30,000 at m/z 200, AGC target 5e5 and maximum IT 45 ms. The normalized collision energy (NCE) was set at NCE 27%.
Data Processing and Bioinformatic analysis
The “DirectDIA” pipeline in Spectronaut v18 software was utilized to analyze the phosphoproteomic data using default parameters. All DIA runs were searched directly against the Swiss-Prot protein database (20,588 entries). Trypsin/P was set as the enzyme, with a peptide length of 7–52 amino acids. Carbamidomethyl (C) was set as a fixed modification, and acetyl (protein N-term), oxidation (M), and phospho(S/T/Y) were set as variable modifications. The PTM localization score > 0.75 was used as the criterion for highly reliable Class-I phosphosites.
The phosphosites matrix was median-normalized and log2-transformed. Phosphosites with > 50% missing values in each group were removed from the dataset. Statistical and bioinformatic analysis were conducted using R. The Benjamini-Hochberg adjusted
P value < 0.05 (Student’s
t-test) and a fold-change > 1.5 were used to identify differentially expressed phosphosites. KEGG enrichment analysis based on differentially expressed phosphosites (collapsed into gene-level) was performed using “clusterProfiler” R package
50. Kinase-Substrate Enrichment Analysis (KSEA) algorithm was employed to estimate kinase activities using the “KSEAapp” (
casecpb.shinyapps.io/ksea/)
51.
Phosphorylation detection of peptides in vitro kinase assay
An in vitro kinase reaction was performed by incubating the purified LKB1 complex with 200 μM LKB1tide in kinase buffer (50 mM Tris, pH 7.5, 10 mM MgCl2, 1 mM DTT, 100 μM ATP) at 30 °C for 30 min with constant rotation. The reaction was terminated by ultrafiltration centrifugation to remove the LKB1 complex proteins.
Peptides for each sample were desalted with SDB-RPS StageTips for LC-MS/MS analysis. The DDA analysis was performed on a hybrid TIMS quadrupole TOF mass spectrometer (Bruker timsTOF Pro) platform connected to an online nanoflow U3000 HPLC system (Thermo Fisher Scientific, USA). Raw data were processed using Maxquant software (version 2.4.2) (3) against the peptide database, with protein, PSM and site FDR at 1%. Phospho (STY) and Oxidation (M) were set as variable modifications.
Metabolite sample preparation and analysis
Sample preparation for metabolite analysis, including ATP, ADP, and AMP, was performed as follows. For cell samples, each sample required 2 × 106 cells collected from a 10-cm dish. The metabolites were extracted using an extraction solution of acetonitrile/methanol/water (2:2:1, v/v/v), pre-cooled to −80 °C. Cells were incubated with extraction solution in the plates for 1 h at −80 °C. Scrape the entirety of the contents of the plates, transfer to EP tubes, then vortex for 1 min. Samples were centrifuged at 12,000× g for 15 min at 4 °C to precipitate the insoluble material. The supernatant was transferred to EP tubes and evaporated to dryness at 4 °C using a vacuum concentrator. The samples were normalized to the determined protein concentration. For the tissue sample, 20 mg liver was homogenized in 200 μL ddH2O at 4 °C. The samples were normalized by protein concentration determined. Then add 800 μL acetonitrile/methanol (1:1, v/v) into samples, vortex and sonicate to extract the metabolites. Sample was centrifuged at 12,000× g for 15 min at 4 °C and the supernatant was evaporated and dried at 4 °C using a vacuum concentrator.
Extracts were dried in a vacuum concentrator at 4 °C and subjected to LC-MS analysis. The data were acquired using a UHPLC system (Vanquish, ThermoFisher Scientific) coupled to an orbitrap mass spectrometer (Exploris 480, Thermo Scientific). A Waters BEH amide column (particle size, 1.7 μm; 100 mm (length) × 2.1 mm (i.d.)) and a BEH Z-HILIC column (particle size, 1.7 μm; 100 mm (length) × 2.1 mm (i.d.)) were used for LC separation. For analysis using BEH Z-HILIC column, mobile phase A was 15 mM ammonium bicarbonate in water (pH = 9.0), while mobile phase B was 15 mM ammonium bicarbonate in 90% acetonitrile and 10% water (pH = 9.0) for both ESI positive and negative ionization modes. The linear gradient eluted from 90% B to 65% B (0–5.0 min, 0.5 mL/min), 65% B (5.0–6.0 min, 0.5 mL/min), 65% B to 90% B (6.0–6.5 min, 0.5 mL/min), 90% B (6.5–9.5 min, 1.0 mL/min), 90% B (9.5–9.51 min, 1.0 mL/min). The sample injection volume was 2 μL. Mobile phases, gradient elution for BEH amide column and ESI source parameters followed the previous publication
52. Metabolite annotation was performed using MetDNA (
metdna.zhulab.cn)
52-54.
Metabolite concentrations were calculated using standard curves generated from reference compounds. Cellular volume was determined from the ratio of total packed cell volume to cell count. Liver and muscle tissue volumes were derived using established tissue density values and measured tissue mass.
H&E staining
Liver tissues excised from mice were fixed in 4% (v/v) paraformaldehyde for 1 h at 4 °C, then were soaked in 30% sucrose solution (w/v in water) for 24 h. The dehydrated tissues were frozen at –80 °C. The tissues were embedded in optimal cutting temperature compound (OCT) on a freezing microtome (Leica), and sectioned at a thickness of 8 μm and then dried onto adhesion microscope slides at room temperature.
For H&E staining, the sections were soaked in PBS to remove OCT, then stained in hematoxylin solution for 5 min, then washed in distilled water for 5 min, differentiated in 1% hydrochloric acid (in ethanol) for 10 s, washed in distilled water for 1 min, and immersed in 0.2% (v/v in water) ammonium hydroxide solution for 30 s, washed in distilled water for 1 min, and stained in eosin Y solution for 5 s. The stained sections were dehydrated in 75% ethanol for 5 min, twice in 95% ethanol for 5 min each, and twice in anhydrous ethanol for 5 min each. The stained sections were mounted with Rhamsan gum and visualized on an Axioscan 7 (Zeiss).
Protein expression and purification
The pET-28a plasmids expressing the full-length human PGK1, AMPKα1 or Calmodulin were transfected into E.coli BL21 (DE3) Competent Cells and then the plasmids were expressed with induction by 0.1 mM Isopropyl β-D-1-thiogalactopyranoside (IPTG) for 20 h. The cells were harvested and resuspended in lysis buffer containing 30 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM DTT, and protease inhibitor cocktail (Selleck Chemicals), then lysed by sonication on ice and centrifuged at 12,000× g for 40 min. The supernatant was mixed with Ni-NTA (QIAGEN) and rotated for 1 h at 4 °C. Then the Ni-NTA was washed by wash buffer containing 30 mM Tris-HCl, pH 7.5, 150 mM NaCl, 30 mM imidazole. The protein was eluted with elute buffer containing 30 mM Tris-HCl, pH 7.5, 150 mM NaCl, and 300 mM imidazole. Purified proteins were used immediately or stored at −80 °C.
The LKB1–STRADα–MO25α complex was co-produced in 293F cells. LKB1 (residues 43–346) was cloned in pMLink plasmids tagged with Flag, and MO25α (residues 1–341) and STRADα (residues 59–431) were cloned in pMLink plasmids tagged with His. The plasmids were transfected into 293F cells with Polyethylenimine (Polysciences) together. After 72 h, cells were harvested and lysed by sonication in lysis buffer with 30 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1 mM DTT, and protease inhibitor cocktail (Selleck Chemicals), and then centrifuged at 12,000× g for 40 min. The supernatant was mixed with anti-Flag Beads and rotated for 3 h at 4 °C. Then the anti-Flag Beads were washed by washing buffer containing 30 mM Tris-HCl, pH 7.5, 150 mM NaCl. Add 3C protease into anti-Flag Beads with buffer containing 30 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, and 1 mM DTT to remove the Flag tag. After on-beads 3C digestion, the LKB1–STRADα–MO25α complex was eluted from anti-Flag Beads, further purified by gel filtration chromatography on Superdex 200 Increase 10/300GL column (GE AKTA PURE 25M). The LKB1 kinase complexes — comprising both the WT protein and its mutants — were purified in a single batch at a relatively large scale (0.7–1.0 mg for each complex) and used throughout this study. The purified complex was aliquoted (5 μg/μL, 20 μL per tube) and stored at –80 °C prior to use.
ITC
For ITC assay, Human recombinant PGK1 proteins and MCB-330 were diluted into the same buffer of 30 mM Tris-HCl, pH 7.5 and 150 mM NaCl. MCB-330 at 0.5 mM in the syringe were titrated into PGK1 proteins at 0.05 mM in the sample cell within a MicroCal PEAQ-ITC system (Malvern Panalytical, UK) at 20 °C. The titration curves were processed using the Origin 7.0 software program (Origin Lab) according to the ‘‘one binding site’’ fitting model. The Kd, enthalpy changes (DH), and the fitting errors were derived from the ITC plot.
MST
For MST measurements, the protein was dialyzed into PBS Buffer, then labeled with Protein Labeling Kit RED-NHS 2nd Generation (NanoTemper, cat. #MO-L011). Metabolite or chemical was gradient diluted into 16 concentrations. The labeled protein was added into metabolite or chemical within 1:1. The mixture was loaded into a capillary (NanoTemper, cat. #MO-K022) at room temperature. MST measurements were performed on MicroScale Thermophoresis instrument (NanoTemper, Monolith NT.115) with 20% Excitation Power and 20% MST Power. The data were analyzed by Affinity Analysis software.
In vitro LKB1 kinase activity
LKB1 complex was purified from 293F cells or purchased from Millipore (cat. #14-596), and recombinant His-AMPKα1 was purified from E.coli BL21 (DE3) Competent Cells. 100 ng LKB1 complex was incubated with the indicated metabolites at room temperature for 15 min, then 1 μg recombinant His-AMPKα1 protein was added to perform an in vitro kinase reaction in kinase buffer (50 mM Tris, pH 7.5, 10 mM MgCl2, 1 mM DTT, 100 μM ATP) at 30 °C for 30 min on a rotator. The products were detected by immunoblot with indicated antibodies or were detected by phospho-AMPKα (Thr172) Sandwich ELISA Kit (Cell Signaling Technology, cat. #7959).
In vitro CAMKK2 kinase activity
The CAMKK2 kinase assay protocol was implemented based on the approach described by Chen, Min
et al.
18,55 with minor modifications to accommodate our research context. Briefly, 300 ng commercial CAMKK2 (cat. #C18-10G, Sino Biological) was incubated with indicated metabolites at room temperature for 15 min, then the recombinant His-AMPKα protein was added to perform an
in vitro kinase reaction in the kinase buffer (50 mM Tris, pH 7.5, 10 mM MgCl
2, 1 mM CaCl
2, 1 mM DTT, 200 μM ATP, 200 μM AMP, 200 ng calmodulin) at 30 °C for 30 min on a rotator. The products were detected by immunoblot with indicated antibodies.
Glycogen quantification
Briefly homogenize 20 mg of tissue in 1 mL of 25 mM citric acid, pH 4.2, 2.5 g/L NaF on ice. Centrifuge at 14,000× g for 5 min at room temperature to remove debris. The glycogen levels in the supernatant were measured with Glycogen Assay Kit (Sigma MAK465).
Immunoprecipitation of endogenous LKB1 protein
For immunoprecipitation, tissue lysates were first precleared by incubation with Protein A/G beads at 4 °C for 1 h to remove nonspecific binding. After centrifugation, the supernatant was collected and incubated overnight at 4 °C with either an LKB1 antibody (1:200) or control IgG. Protein A/G beads were then added to the mixture and incubated at 4 °C for 1 h. The beads were pelleted by centrifugation, and the supernatant was discarded. The beads were washed three times with PBS and subsequently used for in vitro kinase assays.
MD simulation analysis
The conserved systems were constructed according to PDB files (PDB: 2wtk). The missing residues are reconstructed using the CHARMM-GUI input generator
56. The systems were then solvated by TIP3 water molecules with minimal margin of 15 Å from any protein atom to any edge of water box. Sodium and chloride ions were added to neutralize the system to a total concentration of ~150 mM by VMD software
57. The resulting solvated systems were energy-minimized for 50,000 steepest descent steps, followed by an additional 50,000 conjugate gradient steps, where all atoms could move. In the heating stage, each system was gradually heated to 50 K and then to 250 K. In the production stage, all simulations were performed using the NPT ensemble at 300 K, with a timestep of 2 fs. The particle mesh Ewald (PME) method was used to calculate the electrostatic interaction, and the van der Waals interactions were calculated using a cutoff of 8 Å. All MD simulations were performed using the amber20 software
58 and last 1,000 ns. MD trajectories were saved by every 0.1 ns for analysis.
RMSD and RMSF calculation: RMSD and RMSF for the backbone of each structure were calculated by VMD.
HDX-MS
The HDX-MS experiment was performed using a Tris-HCI buffer at pH 7.4, with two parallel buffers prepared using either water or deuterium oxide as the solvent. Sample preparation was automated and efficiently carried out using the PAL3 System (LEAP Technologies).
For continuous labeling, 4 μL of the LKB1/STRAD/MO25 complex (12 μM), either in the presence or absence of test compounds (3PG, 2PG, or PYR), was mixed with 30 μL of D2O buffer. The mixture was incubated at 10 °C for 10, 60, and 600 s to allow hydrogen–deuterium exchange under controlled conditions. The deuterium-exchanged samples were quenched with 30 µL of ice-cold quench buffer (4 M guanidine hydrochloride, 200 mM TCEP, 100 mM citric acid, pH 2.3). Fifty microliters of quenched samples were thawed and immediately injected onto an immobilized pepsin column (2.1 × 30 mm; Thermo Fisher Scientific, Waltham, MA, USA) at a flow rate of 50 µL/min with 0.1% formic acid in H2O at 4 °C. Peptide fragments were collected on an Acclaim PepMap300 C18 column (5 µm, 1.0 mm × 15 mm; Thermo Fisher Scientific) for desalting with 0.1% formic acid in H2O, followed by isolation via liquid chromatography using an ACQUITY UPLC Peptide CSH C18 column (130 Å, 1.7 µm, 1 mm × 50 mm; Waters, Milford, MA, USA). Chromatographic separation was performed at a flow rate of 45 µL/min with an acetonitrile gradient starting at 5% and increasing to 80% over 10 min. To minimize deuterium back-exchange, the system, including the trapping and UPLC columns, was maintained at 0.5 °C, and all buffers were adjusted to pH 2.5.
Mass spectral analyses were carried out using an Orbitrap Fusion Tribrid™ Mass Spectrometer (Thermo Fisher Scientific, USA) equipped with a heated electrospray ionization (HESI) source in positive ion mode. Tandem MS (MS/MS) data were processed using BioPharma Finder 2.0 software (Thermo Fisher Scientific, USA) for peptide identification, and HDX-MS data were analyzed with HDExaminer 3.0 software (Sierra Analytics, Modesto, CA, USA). Per-residue deuterium uptake differences were calculated for each time point, manually validated, and exported based on overlapping peptides. Statistical significance for differential HDX data was assessed using an unpaired t-test for each time point. All data were obtained from at least three independent experiments.
Oral glucose tolerance test and insulin tolerance test
Mice were placed in a clean cage and fasted overnight (14–16 h) for OGTT or 4 h for ITT. For OGTT, mice were given glucose (1.5 mg/g body weight) by oral gavage. For ITT, mice received an intraperitoneal injection of insulin (0.75 U insulin/kg body weight). Tail blood glucose levels were measured after the challenge using the FreeStyle Optium Neo blood glucose monitoring system (Abbott).
Measurement of insulin, triglycerides, cholesterol, and free fatty acids
Mouse liver tissue was homogenized in EBC buffer using OMNI International homogenizer, lysed, and centrifuged at 12,000× g for 20 min at 4 °C to collect the supernatant. Blood samples were collected through orbital blood collection, clotted for 15–30 min, and centrifuged at 2,000× g for 20 min at 4 °C, after which the serum supernatant was collected. Insulin levels were measured using a Mouse Ultrasensitive Insulin ELISA kit (ALPCO) following the manufacturer’s instructions. triglycerides (TG), free fatty acids (FFA), and total cholesterol (TC) were measured using the CheKine™ Micro Triglyceride Assay Kit (Abbkine), CheKine™ Micro Free Fatty Acid Assay Kit (Abbkine), and CheKine™ Micro Total Cholesterol Assay Kit (Abbkine), respectively.
Metabolic cage studies
Metabolic cage studies were performed by the Oxymax/CLAMS system (Coulumbus Instruments). The mice were maintained on a 12-h dark-light cycle for two days at room temperature. Metabolic parameters including oxygen consumption (VO2), CO2 generation (VCO2), RER, and heat were monitored and measured continuously using the Oxymax/CLAMS system. During the whole process, all experimental mice and their littermate controls were single-housed in the metabolic chambers and kept on normal chow or HFD as indicated, and water ad libitum.
In vitro measurement of PGK1 enzyme activity
PGK1 activity was measured at 37 °C by a continuous assay coupled with GAPDH. The reaction system contained 10 mM KH2PO4 (pH = 7.0), 20 mM MgSO4, 200 mM Glycine, 10 mM DL-Glyceraldehyde 3-Phosphate (G3P), 2 mM β-NAD, 2 mM ADP, 10 μg/mL GAPDH and 20 ng/mL recombinant human PGK1. The PGK1 inhibitor was added to the substrate buffer before the reaction. The production of ATP was determined by ATP Assay Kit (Beyotime) after 30 min of reaction.
Structure-based design of MCB-330
Compound MCB-19-330 was designed through a structure-based drug design approach followed by further optimization of the pharmacokinetic properties. Terazosin was identified as a potential PGK1 activator
38. In the crystal structure of mPgk1 in complex with the substrate 3PG and the PGK activator terazosin, the oxygen atom on the furan ring forms a hydrogen network with PGK1-S255 and T254 through a water molecule. We decided to retain this interaction by moving the piperazinyl nitrogen atom out of the ring to form an acyclic amide moiety to reduce the potential hERG inhibition. Meanwhile, the C1 amino group on the central quinazoline core was replaced by a morpholine moiety to increase the aqueous solubility and chemical stability of the primary amine. Meanwhile, by shedding light on a reported PGK1 inhibitor from Abbot company featuring a prop-2-yn-1-ol sidechain, we incorporated this component as the C7-substituent of our quinazoline skeleton to bring in additional H-bonding and π−π stacking, thus leading to the final compound MCB-330. Structure-based design of MCB-330 is shown in Supplementary Materials (Supplementary Fig. S13a). This compound shows a high inhibitory potency against PGK1 with an IC
50 value of 50.8 nM with a marginal inhibition on hERG (IC
50, 9.08 μM), a marker of potential cardiac toxicity.
Synthetic procedure of MCB-330
A mixture of commercially available methyl 2-amino-5-chloro-4-iodobenzoate (3.12 g, 10.0 mmol) and urea (6.06, 100 mmol) in a sealed tube was heated at 200 °C for 5 h. The reaction mixture was cooled to room temperature, and a large amount of solid was precipitated after water addition. After filtration and drying, compound 1 (3.06 g) was obtained in 95% yield.
Compound 1 (3.06 g, 9.5 mmol) was dissolved in POCl3 (8.8 mL, 94.6 mmol), and then DIPEA (6.3 mL, 36.1 mmol) was added to the resulting solution. The mixture was stirred at 110 °C for 4 h. The solution was concentrated in vacuo and the resulting mixture was poured into the ice-water mixture. The resulting mixture was extracted with EtOAc and the organic layer was dried over Na2SO4 and concentrated in vacuo. The residue was purified by column chromatography to give the target product 2 (2.79 g) as a white solid in 82% yield. 1H NMR (400 MHz, d6-DMSO) δ 8.73 (s, 1H), 8.40 (s, 1H).
To a solution of 2 (539 mg, 1.5 mmol) in THF (10 mL) was added morpholine (0.26 mL), and the resulting mixture was stirred at room temperature for 1 h. After completion of the reaction, a large amount of solid was precipitated after water addition to give compound 3 (575 mg) after filtration and drying in 92% yield. 1H NMR (400 MHz, CDCl3) δ 8.39 (s, 1H), 7.88 (s, 1H), 3.88 (d, J = 5.0 Hz, 8H).
To a solution of 3 (575 mg, 1.4 mmol) and tert-butyl methyl(piperidin-4-yl) carbamate (450 mg, 2.1 mmol) in 1,4-dioxane (8 mL) was added DIPEA (0.70 mL, 4.2 mmol). The resulting solution was stirred at 100 °C for 3 h. After reaction completion, the solution was concentrated in vacuo and the resulting mixture was extracted with EtOAc. The organic layer was dried over Na2SO4 and concentrated in vacuo. The residue was purified by column chromatography to give 4 (528 mg) as a white solid in 64% yield. 1H NMR (400 MHz, CDCl3) δ 8.09 (s, 1H), 7.68 (s, 1H), 4.97 (d, J = 13.4 Hz, 2H), 4.28 (s, 1H), 3.91 – 3.83 (m, 4H), 3.66 – 3.58 (m, 4H), 2.89 (t, J = 12.1 Hz, 2H), 2.71 (s, 3H), 1.79 – 1.59 (m, 4H), 1.48 (s, 9H).
To a solution of 4 (528 mg, 0.89 mmol), PdCl2(PPh3)2 (63 mg, 0.089 mmol), and CuI (34 mg, 0.178 mmol) in THF (6 mL) was added prop-2-yn-1-yl acetate (0.18 mL, 1.78 mmol) and DIPEA (0.60 mL, 3.56 mmol). The resulting solution was stirred at 80 °C under N2 atmosphere for 5 h. After reaction completion, the solution was concentrated in vacuo and the resulting mixture was extracted with EtOAc. The organic layer was dried over Na2SO4 and concentrated in vacuo. The residue was purified by column chromatography to give 5 (405 mg) in 80% yield. 1H NMR (400 MHz, CDCl3) δ 7.64 (d, J = 3.3 Hz, 2H), 4.98 (d, J = 10.8 Hz, 4H), 4.30 (br, 1H), 3.93 – 3.82 (m, 4H), 3.67 – 3.56 (m, 4H), 2.89 (t, J = 12.3 Hz, 2H), 2.70 (s, 3H), 2.14 (s, 3H), 1.73 (d, J = 11.3 Hz, 2H), 1.62 (d, J = 7.7 Hz, 2H), 1.48 (s, 9H).
To a solution of 5 (56 mg, 0.1 mmol) in CH2Cl2 (1 mL) was added TFA (74 μL, 1 mmol) at 0 °C under N2 atmosphere, and the resulting mixture was stirred at room temperature for 1.5 h. After reaction completion, the solution was concentrated in vacuo to give a brown residue, which was further dissolved in CH2Cl2. To the resulting solution was added (R)-(+)-tetrahydro-2-furoic acid (20 mg, 0.15 mmol), TBTU (64 mg, 0.2 mmol), DIPEA (83 μL, 0.5 mmol). The resulting mixture was stirred at room temperature for 3 h. After reaction completion, the solution was concentrated in vacuo and the resulting mixture was extracted with CH2Cl2. The organic layer was dried over the saturated Na2SO4 and concentrated in vacuo. The residue was purified by column chromatography to give 6 (28 mg) in 50% yield. 1H NMR (400 MHz, CDCl3) δ 7.64 (s, 2H), 5.08 – 4.92 (m, 4H), 4.72 (t, J = 11.6 Hz, 0.7H), 4.15 (d, J = 9.5 Hz, 0.3H), 4.08 – 4.03 (m, 2H), 3.88 (s, 4H), 3.61 (d, J = 3.5 Hz, 4H), 3.56 – 3.46 (m, 1H), 3.12 – 3.00 (m, 1H), 2.98 – 2.85 (m, 4H), 2.78 (s, 1H), 2.53 – 2.48 (m, 1H), 2.02 – 1.71 (m, 6H), 1.54 (d, J = 10.7 Hz, 2H).
To a solution of 6 (28 mg, 0.05 mmol) in MeOH (1 mL) was added 4N HCl 1,4-dioxane solution (1 mL) at 0 °C under N2 atmosphere, and the resulting mixture was stirred at room temperature for 3 h. After reaction completion, the solution was concentrated in vacuo to give a colorless residue. A large amount of solid was precipitated when ethanol was added into the residue. After filtration and drying, MCB-330 (22 mg) was obtained in 80% yield. MS (ESI, [M + H] +) m/z 514.3. HRMS (ESI) calcd for C26H33ClN5O4, 514.2216; found, 514.2225. 1H NMR (400 MHz, d6-DMSO) δ 12.82 (s, 1H), 8.23 (s, 1H), 8.07 (s, 1H), 4.86 – 4.76 (m, 3H), 4.67 – 4.51 (m, 1.5H), 4.42 (s, 2H), 4.21 (br, 0.5H), 4.01 (br, 4H), 3.82 – 3.70 (m, 6H), 3.21 (br, 2H), 2.84 (s, 2H), 2.66 (s, 1H), 2.10 – 1.96 (m, 2H), 1.91 – 1.68 (m, 5H), 1.62 (d, J = 10.1 Hz, 1H). 13C NMR (126 MHz, d6-DMSO) δ 170.80, 170.54, 160.50, 150.07, 137.06, 128.27, 127.37, 126.70, 110.24, 99.18, 79.29, 75.25 and 75.04 (1C), 68.23 and 68.17 (1C), 65.77 (2C), 52.86 and 50.24 (1C), 49.46 (2C), 49.15 (2C), 44.92, 29.12 and 27.18 (1C), 29.15 and 29.03 (1C), 28.20 and 27.11 (1C), 27.75, 25.43 and 25.19 (1C).
The detailed synthetic procedures are shown in Supplementary Materials (Supplementary Fig. S13b).
Quantification and statistical analysis
All experiments were performed using more than three mice or three independent repeated experiments. Unless otherwise indicated, data in the figures are presented as the mean ± SD. Statistical significance was determined by Student’s t-test, one-way ANOVA, two-way ANOVA, Pearson correlation, log-rank test, Fisher’s exact test, or χ2 test. The 0.05 level of confidence was accepted for statistical significance for all statistical tests. Different P values, i.e., P < 0.05, P < 0.01, P < 0.001 and P < 0.0001 are indicated as *, **, *** or ****, respectively. The Graphical Abstract was generated with BioRender.
DATA AVAILABILITY
HDX data have been deposited at PRIDE database (Project accession: PXD069361, log-in token: mNdzHRJ7ATSg). This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.
The Author(s) 2026. Published by Higher Education Press. This is an Open Access article distributed under the terms of the CC BY license (https://creativecommons.org/licenses/by/4.0/).
