AMP‐activated protein kinase regulates cancer cell growth and metabolism via nuclear and mitochondria events

Abstract Adenine monophosphate‐activated protein kinase (AMPK) is a fuel sensing enzyme that is activated in shortage of energy and inhibited in its surplus. Cancer is a metabolic disease characteristic of aerobic glycolysis, namely Warburg effect, and possesses heterogeneity featured by spatiotemporal hypoxia and normoxia, where AMPK is deeply implicated. The present study delineates the regulation of mitochondrial functions by AMPK in cancer cells. On the one hand, AMPKα subunit binds to mitochondria independently of β subunit and targeting AMPK to mitochondria facilitates oxidative phosphorylation and fatty acid oxidation, and inhibits glycolysis. As such, mitochondrial AMPK inhibits the growth of cancer cells and tumorigenesis. On the other hand, ablation of the β subunits completely abolishes AMPK activity and simultaneously leads to decreases in mitochondria DNA and protein contents. The effect of the β deletion is rescued by overexpression of the active mutant of bulky AMPKα1 subunit. In conjunction, the transcriptional factors PGC1α and Nrf‐1 are up‐regulated by LKB1/AMPK, an event that is abolished in the absence of the β subunits. Intriguingly, the stimulation of mitochondria biogenesis is not achieved by mitochondria‐targeted AMPK. Therefore, our study suggests that AMPK inhibits cancer cell growth and tumorigenesis via regulation of mitochondria‐mediated metabolism.


| INTRODUC TI ON
Cancer arises from mutations of a series of genes regulating cell growth including tumour promoters and tumour suppressors. To suffice the demands of rapid proliferation, transformed cells must generate sufficient energy to fuel the increased anabolic program.
Indeed, the energy generating mechanism in cancer cells is different from normal cells, which was first described by Warburg in early 1920s, also called Warburg effect. 1 Namely, cancer cells utilize glycolysis as the major energy source even under aerobic conditions, while normal cells mainly entail mitochondrial oxidative phosphorylation.
Although the generation of ATP through glycolysis is less efficient, the glycolytic rate in cancer cells can be up to 200 times higher than normal cells even in the presence of oxygen, assuring immediate availability of ATP for anabolic demands of cancer cells. 7 The adaptation of cancer cells to glycolysis appears to have several advantages. 2 First, it is beneficial to accelerated growth. The cancer cells are adapted to acidic microenvironment resulting from release of lactic acid, the end-product of glycolysis, which is beneficial to activation of matrix metalloproteinases, enzymes for cleavages of extracellular matrix and promotion of cell migration and invasion. 3 Second, glycolysis provides cancer cells with various metabolic precursors that are used for the synthesis of amino acids, nucleotides and lipids, as well as reducing power. 4 Third, many oncogenes and tumour suppressors, previously known to control cell growth, have now emerged as modulators of glycolysis. 5 For example, the TP53 tumour suppressor, the most frequently mutated gene in human cancer, supports oxygen-dependent energy production by promoting mitochondrial biogenesis and suppresses glycolysis. 6 Critical components in the glycolytic pathway can be directly or indirectly regulated at all levels by oncogenic or tumour suppressive proteins.
For instances, hypoxic inducing factor 1α (HIF1α), a master regulator for glycolysis, is activated under hypoxic condition and destructed on normoxia by proteosomal degradation. However, HIF1α is activated by many oncogenic proteins and thus contributes to aerobic glycolysis. 5 Adenine monophosphate-activated protein kinase (AMPK) is highly conserved in eukaryotes and acts as a fuel gauge that senses energy crisis. In mammalian cells, AMPK consists of three subunits, catalytic α (α1, α2) subunits and regulatory β (β1, β2) and γ (γ1, γ2, γ3) subunits. 8 When AMP level or AMP to ATP ratio is increased, AMPK activation is initiated by binding of AMP to the γ subunit, followed with phosphorylation of T172 in the activation loop of the α subunit by the liver kinase B1 (LKB1). 9,10 Likewise, pharmacological agents that mimic calorie restriction to increase cellular AMP can activate AMPK, which include

| Cell culture and transfection
The lung adenocarcinoma A549 cell line and A549-LKB1 cell line were established previously 26

| Preparation of virus expressing AMPK
cDNA encoding the active mutant of AMPK was engineered by tagging mitochondrial binding sequence from cytochrome C oxidase and a flag epitope at its aminoterminus. The chimeric cDNA was subcloned into a lentiviral expression vector under the control of Tet-off promoter, the lentivirus was prepared, and cells were infected as previously described. 26 Adenovirus expressing the active mutant of AMPK tagged only with the flag epitope was prepared as described previously. 37

| Mitochondria fractionation
Mitochondria fraction was prepared using a kit (MitoSciences) provided by Abcam according to the protocol provided.
To prepare mitochondria from rat, the liver was minced into small piece and about 0.6 mg was homogenized in 6 mL of buffer A at 4°C.
The following steps are the same as for the cultured cells.

| Western blot
Western blot was performed as we described previously. 37

| MTT assay
The assay was performed according to standard protocol. 37

| Wound healing
The assay was performed as described previously. 37

| Clonogenic assay
The assay was conducted according to the protocol described by Franken et al. 51 When the colonies were visible, the plates were washed with PBS, fixed with fixation solution (acetic acid/methanol, 1:7 v/v), stained with crystal violet (0.5%), and washed twice again.
Photographs were taken.

| Measurement of mitochondrial DNA to genomic DNA ratio
The assay was performed according to Rooney et al. 52

| Glycolysis and oxygen consumption assays
Both assays were performed with Seahorse Bioscience XF Analyzer.
Glycolysis rate was measured using Seahorse XF Glycolysis Stress Test Kit and oxygen consumption rate (OCR) using Seahorse XF Cell Mito Stress Test Kit. Animals were killed when tumour diameter reached 1.5 cm, which takes approximately 39 days after inoculation, and tumour removed.

| Tumour animal model
For allograft model, B16F10 cells (5 × 10 6 ) were injected subcutaneously into the back of C57 mice (Purchased from Nanjing University Biomedicine Institute, Nanjing, China) (6 each group) using similar method also described above except without matrigel.
One week after injection, animals were gavaged with 0.1 mL suspension of berberine in PBS (27 mg/mL, 135 mg/kg) every other day.
The tumour sizes were measured every 3 days.

| Immunofluorescence
Cells were fixed in 4% paraformaldehyde in PBS and then washed with PBS. For AMPK β subunit, samples were blocked with 10% normal goat serum (NGS) for 30 minutes, and incubated with rabbit monoclonal antibody for 2 hours. After washing three times with PBS, the samples were incubated with Cyanine-3-conjugated goat anti-rabbit antibody (Life Technologies) for 1 hour. For ATP synthase β subunit, the second antibody was FITC-conjugated goat antimouse IgG. All antibodies were diluted in PBS containing 2% NGS. After immunofluorescent staining, the cells were also counterstained with DAPI (0.5 μg/mL for 5 minutes) to detect the nuclei and the coverslips mounted in spectrometric grade glycerol and sealed with nail polish. Fluorescent images were taken under confocal microscope.

| Statistical analysis
Significance of differences amongst groups was determined by twotailed Student's t test. Statistical analysis of the mean differences between groups that have been split on two independent variables (eg volume × time) were examined with GraphPad InStat by using two-way ANOVA. P < 0.05 was set for significance.

| Localization of AMPK in mitochondria
To examine if AMPK is localized to mitochondria, we first performed biochemical fractionation of cell extracts from rat liver F I G U R E 1 Localization of AMPK in mitochondria. Mitochondria and cytosol fractions were prepared from rat liver (A) and A549 cells (B) and blotted with antibodies as indicated. C, HEK293 and A549 cells were immunofluorescently labelled with anti-AMPKβ subunit (red), ATP synthase β subunit (green), and DAPI. Images were taken under confocal microscope. Bar scale: 50 μm and A549 cells. As shown in Figure 1A and B, both AMPKα and β subunits were found in the mitochondrial fraction. Since biochemical mitochondrial fractionation might be contaminated by other cellular compartments, we then conducted immunofluorescent staining of AMPKβ and mitochondrial ATP synthase β subunit ( Figure 1C). In this experiment, we labelled both HEK293 cells and A549 cells with anti-β subunit antibody (red) and subsequently with anti-ATP synthase β subunit antibody (green  Figure 2B). Since this domain contains a binding site for β subunit, we ascertained if the association is mediated through β subunit. Thus, we employed Crispr/Cas9 gene editing technology to knock out both β1 and β2 subunits in A549 cells expressing LKB1 (A549-LKB1) ( Figure 2C). Using this cell line, we infected adenovirus expressing the active mutant of AMPK or GFP virus as a control, and then performed mitochondria fractionation to determine the distribution of AMPK. Figure 2D shows that both endogenous α and recombinant α1 subunits associated with mitochondria regardless that the β subunits were absent ( Figure 2C).

| Targeting AMPK to mitochondria inhibits tumorigenesis
Previous studies have shown that AMPK suppresses Warburg effect, so as to inhibit lymphomagenesis. 25 To test if binding of In an attempt to test if the effect of mitochondria-targeted AMPK α1 is due to sequestering factors for cell growth to mitochondria, we prepared mitochondria-targeted AMPK wild-type and kinase-dead mutant and expressed them in A549 cells. Our results did not suggest this was the mechanism, but rather the kinase activity is required (data not shown).
We conducted subcutaneous injection of these two cells into nude mice and observed the tumour growth curve. The results showed that the cells with the mitochondria-targeted AMPK mutant developed tumour slower and their average tumour weight was also less than the control cells (P < 0.05) ( Figure 3E-G). Altogether, our data demonstrate that binding of AMPK to mitochondria suppresses the growth of cancer cells, suggesting that it occurs via inhibition of the Warburg effect.

| The effect of AMPKβ subunits knockout on mitochondria biogenesis
A549 cells contained a loss-of-functional mutation of LKB1. We have found that the growth of A549-LKB1 was compromised and the cells failed to develop tumour in nude mice. 26 To assess if the inhibitory effect of LKB1 on the cancer cell growth is mediated by AMPK, we deleted β1 and β2 subunits on the context of A549-LKB1 ( Figures 2C   and 4A). Our results showed that cell viability and wound healing were increased after ablation of the β subunits, suggesting that the inhibitory effects of LKB1 on the growth and cell migration were mediated by AMPK ( Figure 4B and C).
We postulated that AMPK counteracted the Warburg effect, leading to the inhibition of the cancer cell growth. To test this, we compared the content of mitochondria DNA as opposed to genomic DNA. Our result revealed that LKB1 increased the ratio of mitochondria DNA, which was markedly diminished by ablation of the β subunits ( Figure 5A). Interestingly, when the ratio of mitochondria protein to total cellular protein was examined, a moderate but non-significant increase in the ratio was observed in the cells containing LKB1; however, a marked decrease in mitochondrial proteins Student's t test was used to test significance. *P < 0.05. E-G, Cells expressing RFP or active mito-AMPK were injected subcutaneously into back of nude mice (six for each cell lines, five developed from RFP cells). Tumour growth were monitored and volumes calculated (E). At the end of experiments, tumours were removed (F) and weighed (G). Statistical analysis for tumour volumes was performed with two-way ANOVA and t test for tumour weight. *P < 0.05 occurred in the absence of the β subunits, suggesting that the basal level of AMPK activity is required to maintain mitochondria mass ( Figure 5B). Furthermore, when the cells were infected with the active mutant of AMPK or GFP as a control, the ratio of mitochondrial to genomic DNA was enhanced to different degrees depending on whether the cells contained β subunits or LKB1( Figure 5C). However, the mitochondrial DNA was not increased when the active mutant of AMPK was targeted to mitochondria ( Figure 5D). F I G U R E 5 The effect of AMPK on mitochondria content. A, Ratio of mitochondrial DNA to genomic DNA was determined using qPCR from cell lines as indicated. B, Ratio of mitochondrial protein to total proteins was determined by protein assay. C, Ratio of mitochondria DNA to genomic DNA in the A549-LKB1 cells with or without β subunits knocked out after infection of adenovirus expressing GFP or active α1 subunit. D, Ratio of mitochondria DNA to genomic DNA in A549 cells expressing mitochondria-targeted active AMPK mutant under the control of Tet-off system (the cells were treated with doxycycline [2 μg/mL] for 2 d). Student's t test was used to test significance. *P < 0.05, **P < 0.01, ***P < 0.001 Our next study showed that nuclear transcription factors crucial for the mitochondria biogenesis were modulated by AMPK. Thus, mRNAs for PGC1α and Nrf-1 were up-regulated by the presence of LKB1 and/or AICAR ( Figure 6A and B). When AMPK β subunits were deleted, mRNAs for PGC1α and Nrf-1 were down-regulated ( Figure 6C and D). Under this context, the expression of the active mutant of AMPKα1 restored the levels of their mRNAs to some degrees.

| AMPK regulates energy metabolism
Using Seahorse Bioscience XF Analyzer, we carried out extracellular flow analysis to examine the effect of total AMPK and mitochondrial AMPK on mitochondrial respiration and the glycolytic rate. The mitochondrial respiratory rate is reflected by OCR, while extracellular acidification rate (ECAR) measures glycosylation rate. Our results showed that targeting the active mutant of AMPK to mitochondria led to a decrease in basal level of ECAR and an increase in OCR ( Figure 7A and B). This was consistent with findings from biochemical measurement of lactic acid in tissue culture medium (data not shown) and assays on cellular levels of malonyl CoA ( Figure 7C). The latter is a product of carboxylation of acetyl CoA catalysed by acetyl CoA carboxylase (ACC), which was the first substrate of AMPK identified to be phosphorylated and inhibited. Excessive malonyl CoA inhibits carnitine acyltransferase 1 in the mitochondrial membranes.
This enzyme transports long-chain fatty acyl CoA into mitochondria for β-oxidation. Thus, targeting AMPK to mitochondria decreased malonyl CoA, leading to increased fatty acid oxidation. We then examined OCR, ECAR and release of lactic acid in the A549-LKB1 cells with or without ablation of the β subunits. Deletion of the β subunits caused a decrease in ECAR, but no significant change in OCR ( Figure 7D and E). Biochemical assays also showed the decrease in lactic acid concentration (data not shown). To our surprise, the overall OCR levels in A549-LKB1 cells were lower than those in A549 cells. The underlying reason is not known, but it is possibly that this was caused by the fact that reduced proliferation of the cells expressing LKB1 diminished the demand for ATP.

| D ISCUSS I ON
The present study delineated the regulation of mitochondrial functions in cancer cells. On the one hand, we found that AMPKα subunit bound to mitochondria independently of β subunit and that targeting AMPK to mitochondria facilitated fatty acid oxidation and oxidative phosphorylation, and inhibited glycolysis. As a result, mitochondrial AMPK inhibited the growth of cancer cells and tumorigenesis. On the other hand, our study showed that knockout of the β subunits completely abolished AMPK activity, reduced mitochondria DNA and protein contents, which was rescued to some degrees by overexpression of the active mutant of bulky cellular AMPK, but not the mitochondrial counterpart. Consistently, our data revealed that the transcriptional factors PGC1α and Nrf-1 were regulated by LKB1/ AMPK and that the cells lacking the β subunits failed to up-regulate these two factors in the presence of active mutant of α1 subunit.
Finally, we found that lack of total AMPK activity diminished glycolysis, suggesting its requirement for basal energy metabolism.
Collectively, our results demonstrate that AMPK is required for regulation of mitochondrial mass through nuclear events and that AMPK regulates oxidative phosphorylation through direct binding to mitochondria. Therefore, we conclude that AMPK exerts integrated effects on mitochondrial function and inhibits tumorigenesis. shown to reduce muscle mitochondria contents. 47 In addition, AMPK activation increases fission rate and reduces fusion rate of mitochondria, processes associated with mitophagy and apoptosis. 48 In the present study, we deleted both β1 and β2 subunits and found that mitochondria mass was greatly diminished. Without the β subunits, the expression of α subunits was also reduced which is consistent with a previous report. 49 It would be plausible to predict that glycolysis would increase in the absence of AMPK activity, if AMPK favoured mitochondrial oxidative phosphorylation. However, to our surprise, our results revealed that cells lacking total AMPK activity exhibited a decrease in glycolysis without a significant compensation of oxidative phosphorylation, as compared to control cells. One explanation is that AMPK regulates extra-mitochondrial events in glucose catabolism. Indeed, AMPK was reported to increase flux through the glycolysis pathway by phosphorylating 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase 3 (PFKBP3), which via fructose-2,6 bisphosphate regulates the activity of PFK1, a ratelimiting enzyme in glycolysis. 50 It is possible that in the absence of AMPK activity, this step is suppressed, resulting in reduced glucose flux to both glycolysis and mitochondria oxidation.
In conclusion, our current work presents two novel pieces of findings. First, our study bifurcates the function of AMPK in regulation of mitochondria. While cytosolic AMPK regulates mitochondria biogenesis and glucose catabolism, mitochondria-associated AMPK stimulates oxidative phosphorylation and inhibits aerobic glycolysis. Second, we have shown that mitochondrial AMPK inhibits the growth of cancer cells both in vitro and in vivo. Our next task is to identify specific substrates of AMPK located in mitochondria and elucidate the mechanism underlying tumour suppressive function of mitochondrial AMPK.

CO N FLI C T O F I NTE R E S T
We hereby declare that all co-authors have no conflict of interest in this work, including affiliations, financial relationships, personal relationships or funding sources that could be perceived as influencing an author's objectivity regarding the manuscript content.

AUTH O R CO NTR I B UTI O N S
SSJ, YW, LYL, FLS, JRZ, HL, YY, YFL, ZZ and BZ performed the experiments and contributed to data analysis; SSJ, PJL and ZJL contributed to writing of the manuscript; SSJ, DQH and ZJL contributed to experimental design and interpretation of experimental data.
All authors gave final approval.