Activation of AMPK by Bitter Melon Triterpenoids Involves CaMKKβ

We recently showed that bitter melon-derived triterpenoids (BMTs) activate AMPK and increase GLUT4 translocation to the plasma membrane in vitro, and improve glucose disposal in insulin resistant models in vivo. Here we interrogated the mechanism by which these novel compounds activate AMPK, a leading anti-diabetic drug target. BMTs did not activate AMPK directly in an allosteric manner as AMP or the Abbott compound (A-769662) does, nor did they activate AMPK by inhibiting cellular respiration like many commonly used anti-diabetic medications. BMTs increased AMPK activity in both L6 myotubes and LKB1-deficient HeLa cells by 20–35%. Incubation with the CaMKKβ inhibitor, STO-609, completely attenuated this effect suggesting a key role for CaMKKβ in this activation. Incubation of L6 myotubes with the calcium chelator EGTA-AM did not alter this activation suggesting that the BMT-dependent activation was Ca2+-independent. We therefore propose that CaMKKβ is a key upstream kinase for BMT-induced activation of AMPK.


Introduction
Type 2 diabetes (T2D) accounts for 90% of the world's escalating diabetes problem, with 439 million people expected to have diabetes by 2030 [1] at a predicted global healthcare cost of US $490 billion per annum [2]. However, current treatments for T2D, insulin resistance and obesity remain insufficient and often result in undesirable side-effects [3]. Traditional medicines have been an attractive source of novel anti-diabetic therapeutics. One such source is the bitter melon (BM, Momordica charantia) with extracts having been used to treat diabetes-like symptoms in humans for hundreds of years. Bitter melon extracts have been reported to have hypoglycaemic, hypolipidemic, antioxidant and anti-inflammatory effects and improve skeletal muscle insulin signalling in vivo [4][5][6][7]. Our recent studies have revealed that bitter melon-derived triterpenoids (BMTs) can increase fat oxidation and glucose tolerance in vivo and stimulate glucose transporter 4 (GLUT4) translocation via the activation of AMP-activated protein kinase (AMPK) in both muscle and fat cells [8]. Furthermore, we have recently demonstrated a sustained efficacy of a triterpenoid in eliminating hyperglycaemia in T2D mice [9]. These reports suggest that triterpenoids may be an attractive source of new anti-diabetic therapeutics via activating AMPK.
AMPK is a potential target for the treatment of T2D as activation of this kinase can promote fat oxidation to counteract insulin resistance [10][11][12]. AMPK activation is associated with the actions of anti-diabetic drugs including metformin and thiazolidi-nediones (TZDs, such as rosiglitazone and pioglitazone). However, there are limitations to these treatments. Metformin is not sufficient in the long term to control hyperglycemia and concerns have been raised about cardiac complications associated with TZDs [13,14]. One common feature of metformin and rosiglitazone is that both activate AMPK indirectly by inhibiting mitochondrial Complex I [15]. A similar effect has also been found with other AMPK-activating agents such as berberines [15,16] and resveratrol [17]. Conceivably, activation of AMPK without disturbing mitochondrial respiration may be a preferred mechanism to avoid some of the side-effects of these anti-diabetic agents [18,19].
AMPK is a heterotrimer consisting of a catalytic a subunit, a scaffolding b subunit [10][11][12]20] and regulatory c subunit, which is responsible for binding adenosine nucleotides, AMP, ADP and ATP [21,22]. AMPK is activated in response to changes in the adenylate charge [23], leading to an increase in AMP-or ADPassociated AMPK and phosphorylation by the major upstream kinases LKB1 [24] or CaMKKb [25,26]. LKB1-mediated activation of AMPK is dependent on binding to MO25a and STRADa, which regulates its subcellular localisation [27]. CaMKK, has two isoforms (a and b) with CaMKKb considered the major AMPK kinase due to its ability to directly bind to AMPK which then directs its kinase activity away from its other substrates and towards AMPK [28]. Phosphorylation inhibits the autonomous activity of CaMKKb which is relieved by Ca 2+ /CaM binding [29].
The purpose of this study was to identify the mechanism by which BMTs increase AMPK phosphorylation. Our results in the present study rule out direct allosteric activation, as well as indirect activation of AMPK through inhibition of mitochondrial respiration. We used LKB1-deficient HeLa cells together with the CaMKKb inhibitor and calcium chelators to confirm that BMTs activate AMPK through CaMKKb activation without altering intracellular calcium flux. These data support BMTs as being a novel class of AMPK activators and advocates the CaMKK-AMPK pathway as a potential target for novel anti-diabetic therapeutics.

Purification of Tetracyclic Triterpenoids from Bitter Melon
Purification of BMT-1 has been described previously [8]. Briefly, an initial ethanol extraction was obtained from freezedried bitter melon extracted with 80% aqueous ethanol, and then partitioned with dichloromethane and n-butanol, successively. The n-butanol soluble part was subjected to macroporous resin column chromatography eluting with ethanol/water mixture with various volume ratios to produce different subfractions. Based on preliminary tests in animals, one selected subfraction was then subjected to normal phase silica gel column chromatography with gradient elution (CHCl 3 : MeOH 100:7,100:35) to yield 8 fractions. BMT-1 was purified from fraction 6 by repeated column chromatography over MCI gel (MeOH in water 40,95%), silica gel (CHCl 3 : MeOH 8:1,8:3), and Sephadex LH-20 (MeOH). BMT-17 was also purified from this subfraction using repeated column chromatography over silica gel (CHCl 3 : MeOH 12:1,5:1), Sephadex LH-20 (CHCl 3 : MeOH 1:1), and finally preparative HPLC (CH 3 CN in water 45,70%, 90 min, 15 ml/min). The structure of BMT-17 was determined by NMR spectroscopic data and its purity (.98%) was determined by HPLC analysis. The structures of both compounds can be seen in Fig. 1A. BMT-1 is soluble in MeOH, pyridine and DMSO while BMT-17 is soluble in chloroform and DMSO. As such both compounds were solubilised in DMSO for this study.
Tissue Culture, Respiration and AMPK Assays L6 myoblasts (up to passage 17) were cultured in a-MEM supplemented with 10% heat-inactivated FBS and 1% antibiotic/ antimycotic. HeLa cells (ATCC) were cultured in DMEM (high glucose) supplemented with 10% heat-inactivated FBS, 16PSG (Gibco, Invitrogen). Both were kept at 37uC in a humidified 5% CO 2 incubator. For differentiation into myotubes, L6 cells were cultured in a-MEM supplemented with 2% heat-inactivated FBS and 1% antibiotic/antimycotic for 5-7 days before interventions. Before acute treatments, cells were moved into serum-free media for 18 hrs. Cells were then treated with either vehicle (0.1% DMSO), 2 mM AICAR, or BMTs, at concentrations indicated in the figures 1, 2, 3, 4, 5, for 30 min. As 1 mM ionomycin was toxic to the cells after 30 min, we incubated cells with ionomycin for 10 min. Cells were then quickly washed twice in ice-cold PBS before addition of lysis buffer (50 mM HEPES pH 7.4, 150 mM NaCl, 1% Triton X-100, 1 mM EDTA, 1 mM EGTA, 10% glycerol, 1 mM DTT) containing phosphatase inhibitors (40 mM NaF, 4 mM Na 4 P 2 O 7 , 2 mM Na 3 VO 4 ) and 16 Complete Protease Inhibitor Cocktail (Roche Applied Sciences, Indianapolis, IN, USA). Lysates were then snap-frozen in liquid nitrogen and stored at 280uC before use. Whole cell lysates were prepared by clearing cell debris by centrifugation before quantification of protein content by BCA (Pierce Technology, Rockford, IL, USA) or Bradford (Bio-Rad Laboratories, Hercules, CA, USA) protein assay.
For respiration experiments, L6 myotubes were cultured and differentiated in an XF24 culture plate (Seahorse Bioscience, Billerica, MA, USA). Differentiated L6 myotubes were then serum-starved 18 hrs in unbuffered a-MEM pH 7.4 (Invitrogen Cat No. 12000-022). Basal oxygen consumption rates (OCR) were measured in triplicate cycles of mix (3 min), wait (2 min) and measure (3 min). Repeated triplicate cycles were performed following each consecutive dose-dependent addition of vehicle, BMT-1, BMT-17 or berberine. For quality control, the OCRs were compared to those of untreated myotubes over time and total protein content of each well following the final measurement. To normalise well-well variation, the basal OCR of each well was set to 100% and the dose-dependent effects of each injection were compared to this.
For in vitro AMPK assays, AMPK complexes were first isolated by immunoprecipitation. Briefly, 50 mg of lysate was combined with 0.5 ml anti-pan-AMPKb antibody (to ensure all complexes were retained), 10 ml Protein A/G Sepharose (Amersham) and 700 ml of lysis buffer in the absence of protease inhibitor cocktail and incubated rotating at 4uC for 2 h. The beads were then washed in 700 ml lysis buffer, then in 700 ml HGE buffer (50 mM Hepes pH 7.4, 10% glycerol, 0.1 mM EDTA) and excess fluid was removed using a 30 G needle. The isolated AMPK was then incubated in assay buffer (50 mM HEPES pH 7.4, 5% glycerol, 5 mM MgCl 2 and 200 mM AMARA peptide) in the absence or presence of 200 mM AMP. Assays were initiated by addition of 32 P-ATP (200 cpm/pmole) at 30uC for 10-14 min. An aliquot of each sample was spotted onto P81 paper and stopped in 1% phosphoric acid. Incorporation of 32 P into the AMARA peptide was assessed by b-scintillation counter.

Western Blot Analysis
After protein quantification, whole cell lysates were reduced in Laemmli buffer containing 15 mM DTT at 65uC for 10 mins. Lysates (10 mg of total protein) were then subjected to SDS-PAGE on 9% gels and transferred onto Hybond PVDF membranes (GE Healthcare, Buckinghamshire, UK) overnight at a constant current (40 mA) in Tris/Glycine transfer buffer containing 10% methanol. Membranes were then blocked in 2% BSA in TBS containing 0.1% Tween 20 (TBST) for at least 1 hr before being incubated in rabbit polyclonal antibodies overnight (,16 hrs) at 4uC. Membranes were then washed in TBST and then incubated in HRP-conjugated donkey anti rabbit secondary antibody (Jackson Immuno Research Laboratories, West Grove, PA, USA) for 1-2 hrs at room temperature. Membranes were then washed in TBST and proteins were visualised by ECL (PerkinElmer, Waltham, USA) and exposed on X-ray film before scanning and densitometric analysis using ImageJ (NIH, Bethesda, MD, USA).

Statistical Analysis
Results are presented as means 6 SEM. When multiple doses were used, individual compounds were compared to vehicle control by one-way ANOVA using a Fisher's Least Significant Difference post-hoc test. When two separate treatments were considered, individual two-way ANOVA analyses were used to assess treatment effects and interactions between treatment effects, followed by post hoc tests to compare each group to the vehicle control group. An unpaired Student's t-test was used to assess statistical significance where one group was compared to vehicle control. For each statistical consideration, p#0.05 was taken to indicate a significant difference.

BMTs Increase AMPK Phosphorylation and Activity
Initially we confirmed that the lead candidate from the original study, BMT-1 [8], and a structurally similar but lower molecular weight compound, BMT-17 (Fig. 1A), resulted in AMPK phosphorylation in L6 myotubes. At the same time we assessed if this increase in AMPK phosphorylation led to phosphorylation of the key AMPK substrate ACC at Ser-79. Incubation of serum-starved L6 myotubes with 1 mM and 10 mM of these BMTs increased both AMPK Thr-172 and ACC Ser-79 phosphorylation, though not to the extent of the positive control berberine (Fig. 1B-D). To assess if these BMTs increased in vitro AMPK activity, we isolated AMPK complexes from these lysates using a pan-AMPKb antibody that recognizes the conserved C-terminus of both AMPKbsubunits essential for complex formation [30]. Both BMTs increased AMPK activity (15-50%) though, again, not to the extent of the positive control berberine (Fig. 1E). BMTs activated AMPK both in the absence and presence of the allosteric activator, AMP, to a similar degree. However, no interaction was evident with a two-way ANOVA (Fig. 1F) indicating that this activation was independent of and additive to AMPdependent activation.

BMTs do not Allosterically Activate AMPK
To assess if these BMTs could directly activate AMPK in vitro, in an allosteric manner similar to AMP or to the recently described compound A-769662 [31], we again isolated AMPK from assessment by in vitro AMPK assay; in the presence (both E and F), or absence (F only) of 200 mM AMP. Data are means 6 SEM for 3-6 independent experiments each containing 2-4 repeats. *p,0.05, **p,0.01 by one-way ANOVA to V; ## p,0.01 to V by Student's unpaired t-test. 1F was analysed by two-way ANOVA with h p,0.05 for compound effects, 1 p,0.05 for AMP effects. No interaction between treatment groups was apparent. doi:10.1371/journal.pone.0062309.g001 Figure 2. BMTs do not directly activate AMPK in isolated AMPK complexes. AMPK complexes were isolated from untreated 18 hr serumstarved L6 myotube lysates by immunoprecipitation using a pan-AMPKb antibody. The isolated AMPK was then incubated in vitro with either vehicle (V) alone, 10 mM BMT-17, BMT-1 or Abbott compound (A-769662) in the presence or absence of 200 mM AMP in AMPK assay buffer for 10 min at 30uC. The in vitro kinase assay was then initiated with the addition of 200 mM 32 P-Mg.ATP and assayed for a further 10 min at 30uC. Incorporation of the 32 P into the substrate peptide was then assessed by beta scintillation counting. Data are means 6 SEM for 5 separate experiments. **p,0.01 for compound effect, ## p,0.01 for AMP effects by two-way ANOVA. No interaction between treatment groups was apparent. doi:10.1371/journal.pone.0062309.g002 untreated serum-starved L6 myotubes. The isolated AMPK was then incubated in vitro with 10 mM BMT-1 or BMT-17 or Abbott compound A-769662 in the presence or absence of 200 mM AMP for 10 min at 30uC before initiation of the in vitro AMPK assay with the addition of 32 P.MgATP (Fig. 2). Unlike the allosteric activators AMP or A-769662, neither BMT-1 nor -17 could increase AMPK activity in vitro, nor could they enhance the AMPdependent activation of the complex. The reason why the Abbott compound (A-769662) only mildly increased AMPK activation may relate to the fact that L6 myotubes are derived from rat skeletal muscle which should predominantly express a2b2containing AMPK complexes, which are more AMP-and less Abbott compound-sensitive [32,33]. Furthermore, AMPK has been shown to be both nucleotide-and Abbott compoundinsensitive with increased basal activity [11,32], which may also contribute to the subtle increase in AMPK activity seen here. Together, these data suggest that BMTs do not activate AMPK in an allosteric manner similar to either AMP or the Abbott compound, nor do they enhance AMP-dependence of the AMPK complex.

BMTs do not Activate AMPK by Altering Cellular Respiration
As BMTs did not activate AMPK directly, indirect mechanisms for AMPK activation were investigated. We recently showed that metformin, rosiglitazone and another natural product, berberine, could activate AMPK by altering the cellular energy charge through inhibiting Complex I of the mitochondrial respiratory chain [15]. We therefore assessed the effect of BMTs on whole cell respiration in L6 myotubes using a Seahorse Bioscience XF24 Extracellular Flux Analyzer (USA). BMTs did not affect whole cell respiration, in contrast to berberine, which induced a dosedependent inhibition of respiration (Fig. 3). These data were consistent with respiration experiments performed on isolated mouse skeletal muscle mitochondria (data not shown). Together these results indicate that unlike many commonly used anti-diabetic therapeutic agents, BMTs do not activate AMPK by altering cellular respiration.

BMTs Activate AMPK through CaMKKband not through LKB1
As AMPK is phosphorylated and activated by AMPKKdependent phosphorylation, we assessed the involvement of the AMPKKs in the BMT-dependent AMPK activation. The predominant AMPKKs are LKB1 and CaMKKb. To evaluate the involvement of LKB1, we assessed the ability of these BMTs to increase AMPK phosphorylation and activity in LKB1-deficient HeLa cells [24]. Both BMT-17 and BMT-1 acutely phosphorylated and activated AMPK in HeLa cells, suggesting that LKB1 is not necessary for these BMT-dependent effects (Fig. 4A-C). To assess if CaMKKb itself was responsible for the increase in AMPK phosphorylation and activity, we pre-incubated the HeLa cells with 1 mM of the cell-permeable CaMKK inhibitor, STO-609, for 15 min before incubation with the BMTs. Pre-incubation with STO-609 completely blocked the BMT-dependent AMPK phosphorylation and activation confirming that CaMKKb activity was essential for the BMT-dependent AMPK activation. The partial inhibition of the ionomycin-dependent AMPK activation by STO-609 is consistent with previous reports under these conditions [26,34], while the minimal effects on AICAR-dependent activation can be attributed to STO-609 only attenuating one of the multiple actions of AMP on AMPK activation [35]. Together this suggests that activation of AMPK by BMTs involves CaMKKb.
As HeLa cells lack LKB1, we therefore investigated if inhibition of CaMKKb could attenuate the BMT-dependent increase in AMPK phosphorylation and activation in L6 myotubes, which contain LKB1. Accordingly, L6 myotubes were pre-incubated with or without 1 mM of the CaMKKb inhibitor STO-609 for 15 min before incubation with BMTs. As with the HeLa cells, BMTs increased AMPK phosphorylation and activation and this effect was significantly inhibited by STO-609 (Fig. 4D-F). Collectively these findings indicate that BMTs activate AMPK through CaMKKb and not through LKB1.

BMTs Activate CaMKKb in a Calcium-independent Manner
To assess if the BMTs increased CaMKKb activity per se, we investigated if BMTs could also increase phosphorylation of the well-established CaMKKb substrate, CaMKI, at Thr-177. Accordingly, L6 myotubes were incubated with increasing doses of BMT-17 for 30 min before analysis by Western blot (Fig. 5A  and B). Indeed, BMT-17 dose-dependently increased CaMKI Thr-177 phosphorylation along with AMPK and ACC phosphorylation, indicating that BMTs activate AMPK by activating CaMKKb.
As CaMKKb is regulated in part by Ca 2+ /CaM binding, we then assessed whether BMTs activate CaMKKb by increasing intracellular calcium levels. To do this L6 myotubes were preincubated with or without 150 mM of the cell-permeable calcium chelator, EGTA-AM, for 15 min before addition of BMT-17 and assessment of AMPK activity. Accordingly, incubation of L6 myotubes with BMT-17 significantly increased the AMPK activity. Pre-incubation with EGTA-AM to deplete the intracellular calcium levels did not alter the BMT-17-induced AMPK activation (Fig. 5C). Together these data suggest that BMTs activate AMPK through CaMKKb in a Ca 2+ -independent manner.

Discussion
Here we have interrogated the mechanism by which two lead BMTs (1 and 17) phosphorylate and activate AMPK. Unlike AMP and the Abbott compound A-769662, BMTs do not allosterically activate AMPK and, unlike berberine and metformin, BMTs do not alter cellular respiration. Use of the cell permeable CaMKKb inhibitor STO-609 on both LKB1-deficient HeLa cells and L6 myotubes has demonstrated the key role of CaMKKb in this event (Fig. 6). We therefore suggest that BMTs activate AMPK in a distinct manner from other anti-diabetic medications.  BMTs increased AMPK activity in the absence and presence of AMP in an additive but not synergistic way suggesting the activation of AMPK by BMTs involves a mechanism distinct from that of AMP. Furthermore, unlike A-769662 and the allosteric activator AMP, neither of the BMTs could activate AMPK in vitro suggesting that the BMTs do not activate AMPK in a direct manner. This finding alone is an attractive attribute in an antidiabetic therapy, as this would allow AMPK to remain responsive to other physiological stimuli such as an exercise bout.
Metformin, rosiglitazone, berberine and resveratrol all activate AMPK at least in part by inhibiting mitochondrial respiration [15,17]. This results in an increase in the cytosolic AMP:ATP ratio, leading to an indirect activation of AMPK. Conceivably, the inhibition of mitochondrial respiration may account for some of the undesirable side-effects associated with indirect AMPK activation including lactic acidosis as a result of the associated increase in glycolysis [18,19]. Conversely, BMTs do not activate AMPK by inhibiting mitochondrial or whole cell respiration and therefore may be more tolerated as anti-diabetic medications.
AMPK is also activated by phosphorylation by the major AMPKKs, LKB1 and CaMKKb. Using LKB1-deficient HeLa cells and LKB1proficient L6 myotubes together with the cell-permeable CaMKKb inhibitor STO-609, we provide evidence to indicate that BMTs activate AMPK through CaMKKb and not LKB1. As BMTs also increased phosphorylation of another CaMKKb substrate, CaMKI, this indicates that BMTs may activate CaMKKb itself and provides further evidence that BMTs are unlikely to activate AMPK through a direct interaction. Since binding to AMPK directs the CaMKKb activity away from its other substrates CaMKI and CaMKIV [36], it is therefore also unlikely that BMTs enhance the AMPK:CaMKKb interaction. As CaMKKb and its substrate CaMKI are both activated by Ca 2+ /CaM binding [37], we then used the cell-permeable calcium chelator EGTA-AM to examine if Ca 2+ /CaM binding was involved in the activation of AMPK by BMTs. The inability of EGTA-AM to block the BMT-dependent AMPK activation demonstrated that the BMT-dependent AMPK activation was independent of intracellular calcium levels. Conceptually, the BMTs could be binding directly to the Ca 2+ /CaM binding site and thereby enhancing CaMKKb activity. However, further experiments are needed to confirm this hypothesis. CaMKII, another Ca 2+ /CaM sensitive kinase can be activated by methionine oxidation or auto-phosphorylation which then maintains the protein in an active conformation [38]. CaMKKb too can be phosphorylated at its N-terminus by CDK5 and GSK3, which regulates its Ca 2+ /CaM-independent autonomous activity [39]. This phosphorylation event is inhibitory, however it can be overcome by Ca 2+ /CaM binding. Therefore, BMTs could be increasing CaMKKb activity by inhibiting phosphorylation by these proteins.
It is also unclear whether these BMTs can enter the cell to interact with CaMKKb. BMTs could, in fact, signal via surface receptors to increase CaMKKb activity. Indeed, bitter melon extracts have been shown to be estrogen receptor (ER) agonists [40], with activation increasing CaMKK and AMPK activities [41]. Accordingly, agonism of the membrane bound ER homolog GPR30 [42], a G-protein coupled receptor homolog, showed antidiabetic properties [43]. Alternatively, as BMTs are structurally similar to bile acids [44], they could agonise the bile acid receptor TGR5, another G-protein coupled receptor which has antidiabetic properties, though this has not been linked to CaMKKb or AMPK activities [45]. It remains to be clarified whether BMTs may signal through surface receptors.
Natural products may be metabolised by specialized enzymes of living organisms to produce pharmacologically active metabolites. Due to the limited availability of BMTs, we were unable to investigate this possibility. However, our previous study showed that a number of cucurbitane-type triterpenes exert similar biological activity in stimulating GLUT4 translocation in L6 myotubes via activation of AMPK [8]. This suggests that BMT-1 is likely to be the bioactive compound by itself rather than serving as a precursor of an active metabolite to produce the effects as observed in the present study.
In summary, our data indicate that BMTs activate AMPK in a manner distinct from other anti-diabetic small molecules (illustrated in Fig. 6). Our findings suggest that BMTs activate AMPK through CaMKKb activation, in a calcium-independent manner, and not through direct effects on AMPK, nor by inhibiting mitochondrial respiration. Although CaMKKb is expressed in tissues important to glucose and lipid metabolism, such as skeletal muscle, liver and adipose [46], the role of CaMKKb in the onset and treatment of insulin resistance and T2D is yet to be fully explored. By showing that BMTs increase AMPK activity through CaMKKb, we provide a proof-of-concept for triterpenoids, which are abundant in the plant kingdom [44], as an attractive source of novel AMPK activators with anti-diabetic efficacy mediated by the upstream kinase CaMKKb.