The AMPK Pathway: Molecular Rejuvenation of Metabolism and Mitochondria

Abstract

Cells must constantly adapt their metabolism to the availability of nutrients and signals from their environment. Under conditions of limited nutrients, cells need to reprogram their metabolism to rely on internal stores of glucose and lipid metabolites. From the emergence of eukaryotes to the mitochondria as the central source of ATP and other metabolites required for cellular homeostasis, survival, and proliferation, cells had to evolve sensors to detect even modest changes in mitochondrial function in order to safeguard cellular integrity and prevent energetic catastrophe. Homologs of AMP-activated protein kinase (AMPK) are found in all eukaryotic species and serve as an ancient sensor of conditions of low cellular energy. Here we explore advances in how AMPK modulates core processes underpinning the mitochondrial life cycle and how it serves to restore mitochondrial health in parallel with other beneficial metabolic adaptations.

AMPK DOMAIN STRUCTURE AND ACTIVATION

AMPK Subunits and Structure

Across eukaryotes, AMP-activated protein kinase (AMPK) is an obligate heterotrimer composed of three subunits: the α catalytic subunit and the β and γ regulatory subunits (Figure 1a). Mammals express two α-isoforms (α1 and α2), two β-isoforms (β1 and β2), and three γ-isoforms (γ1, γ2, and γ3), with each isoform expressed by independent genes. Each subunit class interacts in a 1:1:1 heterotrimeric ratio, enabling the formation of 12 different AMPK complexes (Ross et al. 2016, Smiles et al. 2024). Although α1, β1, and γ1 isoforms are expressed ubiquitously across different tissues and cell types, AMPKα2 and AMPKβ2 are the dominant isoforms in skeletal and cardiac muscle in humans, while AMPKγ3 is restricted mainly to select subsets of skeletal muscle (Smiles et al. 2024). A detailed understanding of AMPK architecture has emerged from studies that resolved the crystal structure of AMPK under conditions examining the effects of nucleotides or synthetic ligands (Calabrese et al. 2014, Li et al. 2015, Ovens et al. 2022, Yan et al. 2019). Several structural features within each subunit enable the dynamic regulation of AMPK activity and response. AMPKα subunits contain the kinase domain at the N terminus followed by an autoinhibitory domain (α-AID) coupled to a linker region and a βγ subunit binding domain at the C terminus (see Figure 1b). AMPKβ subunits contain an N-myristoylation site and a conserved carbohydrate binding module (CBM) that enables glycogen interaction and stabilization of the AMPK complex (Hoffman et al. 2020); this is followed by an αγ binding domain at the C terminus. AMPKγ subunits provide the energy-sensing capacity of AMPK through four tandem cystathionine-β-synthase (CBS) repeats, which bind adenine nucleotides. Although there are four CBS domains, CBS3 appears to be the key regulatory site to which AMP or ADP binds and mediates activation (Steinberg & Hardie 2023).

Figure 1.

Structure of AMPK and sequence specificity of its substrates. (a) Schematic of each AMPK subunit. Mammals contain two catalytic (α) subunit genes, two β subunit genes, and three γ subunit genes. (b) Crystal structure of the AMPKα2β1γ1 trimeric complex. Major structural domains are indicated. The structure shows the activator A769662 bound to a pocket formed by the interface between the kinase domain and the CBM. Structure sourced and adapted from PDB ID: 4CFF using PyMOL software. (c) The optimal AMPK substrate motif as determined using spatially arrayed degenerate peptide libraries. Amino acids selectively enriched at various positions upstream and downstream of the phospho-acceptor site (O) are indicated, with the corresponding flanking sequences of nine reported AMPK substrates below. Abbreviations: AID, autoinhibitory domain; AMPK, AMP-activated protein kinase; CBM, carbohydrate binding module; CBS, cystathionine-β-synthase; CTD, C-terminal domain; NTD, N-terminal domain; PDB ID, Protein Data Bank identifier.

Activation of AMPK also requires phosphorylation at a conserved threonine residue within the activation loop of the α kinase domain, which in humans is Thr172 in α2. Displacement of ATP by AMP at the CBS3 domain causes a major conformational change in the heterotrimer that results in a three-part activation mechanism: (a) promotion of Thr172 phosphorylation, (b) prevention of Thr172 dephosphorylation, and (c) allosteric activation of complexes already phosphorylated at Thr172 (Steinberg & Hardie 2023). Given that phosphorylation of Thr172 is essential for the activation of AMPK, the discovery of the upstream kinase, LKB1 (STK11), was a pivotal breakthrough reported by three different teams using distinct approaches (Hawley et al. 2003, Shaw et al. 2004, Woods et al. 2003). These studies for the first time linked AMPK, which had been associated mainly to metabolic regulation, to LKB1, identified a few years earlier in human genetics studies as a tumor suppressor mutated in the inherited cancer disease Peutz–Jeghers syndrome (Hemminki et al. 1998, Jenne et al. 1998) and subsequently identified as the fourth most commonly sporadically mutated gene in human lung cancer (Shackelford & Shaw 2009). One of the early mechanistic observations about LKB1 was that its kinase activity was active only when in a heterotrimeric complex with the STE20-like pseudokinase STRAD and the scaffold protein MO25a (Baas et al. 2003, Boudeau et al. 2003). Phosphorylation of Thr172 in the α subunit is the principal event required for full activation of AMPK and can increase AMPK activity up to 100-fold in vitro (Gowans et al. 2013, Oakhill et al. 2011, Stein et al. 2000, Suter et al. 2006). In LKB1 null cells, AMPK activation by energy stress is defective but can be rescued upon LKB1 reconstitution. Subsequently, it was demonstrated that LKB1 is a master kinase that also phosphorylates 12 other kinases closely related to AMPK termed the AMPK-related kinases (Jaleel et al. 2005, Lizcano et al. 2004). Thus, LKB1 inactivation leads to loss of the kinase activity of 14 kinases with wide-ranging roles in cellular function, though only the originally defined AMPKα1 and AMPKα2 subunits are phosphorylated following energetic stress, as only they appear capable of binding the AMPKγ energy-sensing subunit.

AMPK can also be phosphorylated on Thr172 by another kinase, CAMKK2, in response to increased intracellular calcium, linking calcium signaling to AMPK regulation of energy metabolism (Hawley et al. 2005, Hurley et al. 2005, Woods et al. 2005). Sequence alignments reveal that CAMKK2 contains the highest degree of sequence similarity to LKB1 of any kinase in the mammalian genome, thus explaining their common shared substrate in AMPK. CAMKK2-mediated activation of AMPK occurs independently of LKB1 and nucleotide levels. This mode of activation occurs in response to hormones and stresses that acutely control intracellular calcium levels (Marcelo et al. 2016a,b). CAMKK2 phosphorylation of AMPK is involved in the regulation of food intake by hypothalamic neurons (Anderson et al. 2008) and other neuronal processes (Lee et al. 2022, Mairet-Coello et al. 2013) and in the activation of antigen and thrombin receptors in immune (Tamas et al. 2006) and endothelial (Stahmann et al. 2006) cells. Other cellular stresses that perturb calcium flux and trigger CAMKK2-dependent AMPK activation include amino acid starvation (Ghislat et al. 2012), hypoxia (Mungai et al. 2011, Salle-Lefort et al. 2016), and cell detachment from the matrix (Sundararaman et al. 2016). Moreover, CAMKK2 maintains basal AMPK activity in the absence of LKB1, explaining why a reduced but present basal AMPK activity can be detected in LKB1-deficient tumors (Eichner et al. 2019; Fogarty et al. 2010, 2016). CAMKK2 activation of AMPK plays important roles in prostate cancer, where CAMKK2 messenger RNA (mRNA) is directly upregulated by androgen receptor during tumor formation and progression (Lin et al. 2021; Penfold et al. 2018, 2023; Pulliam et al. 2022).

Regulation of AMPK

Exercise is one of the most well-documented triggers of mitochondrial biogenesis, leading to increases in the oxidative capacity of muscles ( Jornayvaz & Shulman 2010). Dating back to the 1950s, muscle activity in birds was associated with increased mitochondrial content in muscle fibers. In 1996, a landmark study by Winder & Hardie (1996) demonstrated that AMPK is phosphorylated and activated in skeletal muscle in rats after as little as 5 min of treadmill running. In humans, moderate-intensity cycling exercise for 90 min similarly led to increased AMPK activity and phosphorylation in muscle biopsies (Clark et al. 2004). Subsequent studies expanded on these findings, demonstrating that both acute exercise and exercise training increase AMPK activity in skeletal muscle, which was associated with increased glucose uptake and fatty acid oxidation, thus highlighting AMPK’s central role in mediating metabolic adaptations to exercise (Kjøbsted et al. 2018, Spaulding & Yan 2022). Given that exercise potently activates AMPK, this finding raised the question whether AMPK was involved in regulating aspects of exercise-induced mitochondrial biogenesis or whether other kinases or enzymes responsive to exercise may mediate these effects. Early studies demonstrated that transgenic mice bearing an Arg225Gln gain-of-function mutation in the AMPKγ3 subunit increased mitochondrial biogenesis in glycolytic skeletal muscle (Garcia-Roves et al. 2008), while mice bearing dominant negative AMPKα2 mutants failed to induce mitochondrial biogenesis under energy stress (Zong et al. 2002). Furthermore, mice bearing muscle-specific knockouts of AMPKβ1/AMPKβ2 subunits or the upstream LKB1 kinase showed reduced mitochondrial content in muscle and failed to increase mitochondrial biogenesis after exercise (O’Neill et al. 2011, Thomson et al. 2010). Similarly, increases in mitochondrial content after 6.5 weeks of exercise training were blunted in dominant negative AMPKα2 kinase dead transgenic mice (Brandauer et al. 2015). Notably, both mice bearing muscle-specific double knockout of AMPKβ1 and AMPKβ2 and mice with muscle-specific double knockout of AMPKα1 and AMPKα2 exhibited diminished mitochondrial biogenesis, lower mitochondrial content, and compromised mitochondrial function even under basal conditions (Lantier et al. 2014, O’Neill et al. 2011). AMPK also regulates mitochondrial mass in other tissues, including hepatocytes (Egan et al. 2011, Hasenour et al. 2014), adipocytes (Mottillo et al. 2016), and macrophages (Galic et al. 2011).

Any intervention that raises AMP/ADP levels relative to ATP triggers the activation of AMPK. Unsurprisingly, therefore, compounds that inhibit the mitochondrial electron transport chain (ETC) or lead to mitochondrial uncoupling, both of which reduce ATP production, activate AMPK (Herzig & Shaw 2018). Beyond drugs with known mitochondrial inhibitory activity, this is a common off-target effect of thousands of chemical compounds (Hoogstraten et al. 2023). One of the best-known pharmacological AMPK activators is metformin, a biguanide compound that is a derivative of a natural product (Witters 2001). Metformin, a frontline medication for type 2 diabetes that is prescribed worldwide, molecularly inhibits complex I in the mitochondrial respiratory chain, reducing mitochondrial respiration and ATP production (El-Mir et al. 2000, Owen et al. 2000) and thus indirectly activating AMPK (Zhou et al. 2001), as do any other complex I inhibitors such as rotenone (Toyama et al. 2016). A landmark paper in 2023 by Hirst and colleagues (Bridges et al. 2023) identified a specific pocket within the complex I structure where metformin binds, located primarily in the region where coenzyme Q (ubiquinone) interacts with the complex. With the increasing interest in targeting AMPK for metabolic disorders such as type 2 diabetes, obesity, and cardiovascular diseases, multiple companies sought to improve on metformin by identifying direct AMPK activators that may be more selective than indirect activators (Steinberg & Carling 2019). One of the first direct AMPK activators reported was A769662, developed by Abbott Laboratories in 2006 following a screen of 700,000 compounds (Cool et al. 2006). Although A769662 was initially reported to activate AMPK allosterically, its effects were additive with AMP, suggesting that this compound binds a site distinct from the nucleotide binding sites. By 2012, a review of patent databases reported 26 patents with 10 different classes of AMPK activators. Most of these compounds were closely related to A769662 and later studies confirmed that these drugs shared a common mechanism of action via the allosteric drug and metabolite (ADaM) site, a distinct pocket formed at the interface of the α subunit kinase domain and the β subunit CBM; a range of small-molecule AMPK activators bind at this site to activate AMPK independently of nucleotide levels (Calabrese et al. 2014, Ngoei et al. 2018). These include compounds like MK-8722 and PF-739, which activate AMPK with greater potency than A769662 and show favorable bioavailability properties in rodents (Cokorinos et al. 2017; Esquejo et al. 2018, 2022; Myers et al. 2017; Zhou et al. 2019). A study utilizing newly developed long-acting and short-acting pan-AMPK activators profiled gene expression changes in mice subjected to acute pharmacological treatment versus those undergoing exercise and showed striking concordance between exercise and even short-term AMPK activation (Muise et al. 2019). Additional studies are needed to best define minimal amounts and dosing intervals of direct AMPK activators for a myriad of human pathologies.

The ADaM site, to which many allosteric AMPK activators bind, may also play a direct role in fatty acid sensing. Until now, natural ligands for this site had remained elusive but long chain fatty acid–coenzyme A (LCFA-CoA) esters were identified as natural ligands for the ADaM site (Pinkosky et al. 2020). LCFAs typically are the primary cellular energy reserve in most eukaryotic organisms. Their oxidation in mitochondria is crucial for sustaining a favorable ratio of ATP to ADP. Like A769662, these LCFA-CoA esters allosterically activated β1-containing heterotrimers but not β2-containing complexes and an S108A mutation dampened activation of AMPKβ1 (Rahman et al. 2024). This binding appears to be necessary for AMPK-mediated acetyl-CoA carboxylase (ACC) phosphorylation and stimulation of whole-body fatty acid oxidation, in response to acute increases in fatty acid availability (Pinkosky et al. 2020). This fatty acid–sensing mechanism helps to explain how interventions such as endurance exercise, intermittent fasting, caloric restriction, or ketogenic diets elicit beneficial effects by activating AMPK without lowering cellular ATP or blood glucose. Moreover, AMPK activation by LCFA-CoA esters also appears to be critical for mitochondrial biogenesis and mitophagy (Desjardins et al. 2022). The question remains as to whether a natural metabolite exists that can activate β2-containing complexes, which are the predominant forms in skeletal and cardiac muscle.

AMPK CONTROL OF METABOLISM

The primary role of AMPK across evolution is to maintain energy homeostasis by adjusting metabolic pathways in response to changes in the AMP:ADP:ATP ratio (González et al. 2020). Under low ATP conditions, AMPK promotes cell survival and metabolic homeostasis by shifting cellular metabolism toward catabolic pathways through the breakdown of macromolecules to increase energy production while suppressing energy-consuming anabolic processes. Importantly, compared to other kinases reported to sense metabolic or mitochondrial stresses, AMPK is exquisitely sensitive, becoming rapidly and maximally activated after modest ATP loss from ETC inhibitors such as rotenone, unlike other kinases like PINK1 (Hung et al. 2021). Moreover, even under conditions of greater energetic stress, such as following carbonyl cyanide m-chlorophenyl hydrazone (CCCP) or a combination of antimycin and oligomycin, AMPK is maximally activated in under 5 min, whereas PINK1 and TBK1 only become activated after 30–60 min as revealed in timecourse studies (Hung et al. 2021).

The mechanisms through which AMPK molecularly controls metabolism have been greatly aided through pioneering work that revealed that AMPK contains a robust sequence specificity in its substrates, needing three or four specific amino acids in specific positions relative to the phospho-acceptor site (Gwinn et al. 2008, Hardie 2022) (see Figure 1c). Initial studies employing alanine-scanning and alignments of the first few AMPK substrates identified were enhanced with the advent of unbiased peptide library methods that allowed all theoretical combinations of amino acids at each position before and after a putative phospho-acceptor, revealing an optimal motif consistent with the initial well-validated ACC and HMGCR substrates and extending to predict many more. The sequence alone is not selective enough to indicate what may be an AMPK substrate; indeed, many of the roughly 50 members of the CAMK superfamily of kinases to which AMPK belongs phosphorylate similar sequences in vitro, including the DNA damage response kinases Chk1 and Chk2, CAMKII, PKD1, and nearly all 12 of the closely AMPK-related kinases (SIKs, MARKs, NUAKs, BRSKs/SADs) ( Johnson et al. 2023). An additional strategy that has worked well for many laboratories has been the combination of affinity purification of such low-abundance peptides from AMPK substrates and the use of cells from conditional genetically engineered AMPK knockout mice or, in the last 5 years, cells genetically disrupted in both AMPKα1 and AMPKα2 with the use of CRISPR/Cas9 technology (Chen et al. 2019, Ducommun et al. 2015, Hoffman et al. 2015, Nelson et al. 2019, Stein et al. 2019). From these advances has emerged a consensus across several laboratories about key targets of AMPK in the control of metabolism (see Figure 2). For more details on how AMPK controls glucose and lipid metabolism, see the Supplemental Text.

Figure 2.

AMPK signaling controls multiple specific components of cellular metabolism. AMPK is a central regulator of metabolism, ensuring localized metabolic stability within cellular compartments while coordinating overall cellular energy balance. Its activation and phosphorylation in specific regions of the cell are regulated by the upstream kinases LKB1 and CAMKK2, each responding to diverse stimuli within these compartments. AMPK modulates macronutrient metabolism by regulation of processes such as glucose uptake (via TXNIP/Glut1 and TBC1D1/Glut4), glycolysis (via PFKFB3), glycogen storage (via GS and GP), lipid and cholesterol synthesis (via SREBP1C, HMGCR, and ACC1/2), lipolysis (via ATGL and HSL), and fatty acid oxidation. Additionally, AMPK supports the maintenance of metabolically active organelles, including mitochondria (via MFF, FNIP1/FLCN, Parkin, MTFR1L, ARMC10, GBF1, PDZD8, and AKAP1) and the endolysosomal system (via FNIP1/FLCN, TFEB/TFE3, PIKfyve, ULK1, Vps34, Beclin1, and ATG9). AMPK also acts antagonistically toward key metabolic regulators, such as mTORC1, to suppress energy-consuming pathways, coordinating the balance between cellular anabolic and catabolic activities. Abbreviations: AMPK, AMP-activated protein kinase; DRP1; dynamin-related protein 1; ER, endoplasmic reticulum; ERRα, estrogen related receptor α; FLCN, folliculin; FNIP1, folliculin interacting protein 1; GP, glycogen phosphorylase; GS, glycogen synthase; MFF, mitochondrial fission factor; PKA, protein kinase A; TFEB, transcription factor EB; TFE3, transcription factor E3. Figure adapted from images created in BioRender; Malik N. 2025. https://BioRender.com/8j8ur2i.

AMPK Coordination Cycle of Mitochondrial Fission, Mitophagy, and Biogenesis

Mitochondria are highly dynamic organelles that form an interconnected network that is continuously remodeled upon physiological stress and in pathological scenarios (Mishra & Chan 2016). Mitochondria displaying an extensively interconnected and tubular network tend to exhibit high levels of oxidative phosphorylation, while small and fragmented mitochondria correlate with reduced oxidative phosphorylation capacity and diminished ATP synthesis (Wai & Langer 2016). Mitochondrial fission, fusion, mitophagy, and biogenesis enable remodeling of this mitochondrial network and govern its shape, size, quantity, and cellular distribution (Herzig & Shaw 2018). While mitochondrial fission divides a single mitochondrion into two or more distinct organelles, mitochondrial fusion allows amalgamation of the outer and inner membranes of mitochondria into a single organelle. Mitophagy requires fission to first separate damaged mitochondrial compartments, which are then sequestered and delivered to and degraded in lysosomes (Youle & Narendra 2011). The synthesis of new mitochondrial proteins is accomplished via transcriptional induction of hundreds of mitochondrial genes, which in turn then mix with existing mitochondrial networks, a process termed mitochondrial biogenesis. A major function of AMPK is the maintenance of mitochondrial health and AMPK governs at least three of these critical facets of mitochondrial equilibrium: biogenesis, fission, and mitophagy (Figure 3). Dysregulation of these processes plays critical roles in pathological processes, from neurodegeneration to cancer, and may mechanistically underlie many features of aging.

Figure 3.

AMPK regulates multiple aspects of the mitochondrial life cycle. Upon activation, AMPK rapidly phosphorylates MFF to mediate mitochondrial fission via Drp1. Mitochondrial fission facilitates mitophagy by targeting damaged mitochondria for degradation. AMPK plays diverse roles in the regulation of autophagy through various phosphorylation events, including the phosphorylation of Raptor to inhibit the anabolic mTOR complex. This inhibition relieves mTOR-mediated suppression of ULK1, thereby facilitating mitophagy. AMPK also directly phosphorylates ULK1 itself to control specific spatial and temporal aspects of macroautophagy. AMPK directly phosphorylates Parkin to promote PINK1-Parkin-dependent mitophagy of damaged mitochondria. Furthermore, AMPK phosphorylates FNIP1, initiating a cascade of events that begins with the inhibition of the GAP activity of the FLCN/FNIP1 complex. This enables TFEB nuclear localization and activity, leading to enhanced lysosomal biogenesis for the clearance of defective mitochondria. Simultaneously, AMPK promotes increased expression of PGC1α and ERRα mRNAs, which over several hours culminates in enhanced mitochondrial biogenesis and improved mitochondrial function. Abbreviations: AMPK, AMP-activated protein kinase; Drp1, dynamin-related protein 1; ERRα, estrogen related receptor α; FLCN, folliculin; FNIP1, folliculin interacting protein 1; GAP, GTPase-activating protein; MFF, mitochondrial fission factor; mRNA, messenger RNA; OPA1, optic atrophy type 1; PGC1α, peroxisome proliferator-activated receptor gamma coactivator 1-alpha; TFEB, transcription factor EB. Figure adapted from images created in BioRender; Malik N. 2025. https://BioRender.com/j53n727.

AMPK Controls Multiple Aspects of Mitochondrial Fission and Mitochondrial Morphology

In response to mitochondrial insults such as mitochondrial depolarization or inhibition of ATP synthesis, mitochondria undergo fragmentation to excise damaged mitochondrial regions and target them for degradation via mitophagy (Mishra & Chan 2016). Mitochondrial fission encompasses a complex series of steps, including initiation, constriction, division, and release. These events are driven by core mitochondrial fission machinery and involvement from several other organelles such as endoplasmic reticulum (ER) (Friedman et al. 2011).

AMPK phosphorylation of MFF controls fission and much more.

Across eukaryotes, mitochondrial membrane fusion is regulated by GTPase proteins: two mitofusin isoforms (Mfn1 and Mfn2) on the outer mitochondrial membrane (OMM) and optic atrophy type 1 (OPA1) on the inner mitochondrial membrane (IMM) (Chan 2020). Mitochondrial fission is also driven by another GTPase protein, dynamin-related protein 1 (DRP1), which resides in the cytoplasm but is recruited to bind specific receptors found on the OMM. The initiation step in fission involves recruitment of DRP1 to mitochondrial constriction sites on the OMM by mitochondrial fission factor (MFF), the long-sought receptor for DRP1 (Chan 2020). In addition, adaptor proteins MiD49 and MiD51 and fission protein Fis1 have also been proposed to be involved, though studies suggest that only MFF may serve as a receptor that directly binds to DRP1. Contact sites between ER and mitochondria known as mitochondria-associated membranes also serve as platforms for recruitment of Drp1 (Chan 2020). Upon recruitment, Drp1 assembles into spiral-shaped oligomers, which in coordination with ER–mitochondria contact sites wrap around the mitochondrion to form constriction rings at the fission sites, constricting the mitochondrial membrane and leading to narrowing of the mitochondrial structure at those points. Division occurs as the constriction ring tightens, creating a mechanical force on the OMM that leads to scission of the mitochondrial membrane, which requires the GTPase activity of Drp1, eventually dividing the mitochondrion into two or more smaller organelles. The newly formed daughter mitochondria are then released into the cytoplasm, where ER–mitochondria interactions continue to play a role in postfission events facilitating mitochondrial distribution and dynamics. A striking study in 2021 used live-cell structured illumination microscopy to capture mitochondrial dynamics, which revealed two functionally and mechanistically distinct types of fission (Kleele et al. 2021). Notably, division at the mitochondrial periphery enabled damaged material to be shed into smaller mitochondria destined for mitophagy, whereas division at the midzone led to the proliferation of mitochondria. Both types are mediated by DRP1, but ER- and actin-mediated preconstriction and MFF governed only midzone fission. Peripheral fission is preceded by lysosomal contact and is regulated by FIS1 (Kleele et al. 2021).

While mitochondrial depolarization was known to stimulate fission via inhibition of mitochondrial fusion machinery (Baker et al. 2014), it was not clear one decade ago whether modest ETC inhibitors also remodeled the mitochondrial network. In addition, nutrient starvation by different methods had been reported to promote mitochondrial elongation and fusion; thus, there were conceptual arguments for AMPK playing either profission or profusion roles. In this backdrop, multiple teams employing unbiased phosphoproteomics following pulldowns with the AMPK substrate motif identified MFF, an OMM protein and Drp1 receptor, as a candidate AMPK substrate (Ducommun et al. 2015, Toyama et al. 2016). Utilizing CRISPR/Cas9 to delete AMPK from U20S cells (which are excellent for imaging organelles), Toyama et al. (2016) reported that rotenone and antimycin A (inhibitors of ETC complex I and complex III, respectively) induce rapid fragmentation of the mitochondrial network, an effect that is surprisingly fully abolished by loss of AMPK. Moreover, treatment of cells with small-molecule activators of AMPK in the absence of mitochondrial damage was sufficient for induction of mitochondrial fission, confirming an acute profission function for AMPK. Mutation of the two conserved canonical AMPK sites in MFF, Ser155Ala and Ser173Ala, prevented DRP1 recruitment to mitochondria following rotenone and AICAR, demonstrating the critical role of AMPK phosphorylation of these sites in MFF for energy stress–induced recruitment of Drp1. Strikingly, substitution of these two serine sites in MFF to phosphomimetic residues resulted in chronically short mitochondria, in the absence of any stimuli (Toyama et al. 2016). The importance of these exact phosphorylation sites was confirmed in a different context, the fragmentation of mitochondria during mitosis, which is controlled by protein kinase D (PKD) phosphorylation of exact sites AMPK phosphorylates during energy stress (Pangou et al. 2021). As PKD is highly related to AMPK in its kinase domain, it is not surprising that they might share common substrates, with specific conditions dictating which kinase mediates the phosphorylation event.

Another widely studied form of energetic stress is induced by the diabetes drug metformin, which triggers AMPK-dependent MFF phosphorylation in hepatocytes, where mitochondrial fission improves mitochondrial respiration and restores the mitochondrial life cycle in vivo. Phosphorylation of MFF was correlated with the AMPK-dependent capacity of metformin to lower blood glucose levels in mice fed a high-fat diet, which is lost upon liver-specific knockout of AMPKα1/AMPKα2 (Wang et al. 2019). Another indication that regulation of MFF could be relevant for metabolic disease came from an unexpected finding from Hammerschmidt et al. (2019), interrogating whether specific lipids mediated the known effects of ectopic lipid deposition on the development of obesity and insulin resistance. The authors found that the C16:0 sphingolipid-synthesizing ceramide synthases, CerS5 and CerS6, affected distinct sphingolipid pools and that abrogation of CerS6 but not CerS5 protects from obesity and insulin resistance. Remarkably, proteomics revealed MFF as a protein that selectively bound CerS6-derived C16:0 sphingolipids. CerS6 and Mff deficiency protected from fatty acid–induced mitochondrial fragmentation in vitro, and the two proteins genetically interacted in vivo in obesity-induced mitochondrial fragmentation and development of insulin resistance (Hammerschmidt et al. 2019). A study performing unbiased phosphoproteomics of food perception without food consumption uncovered phosphorylation of Mff on a Ser131 site, two residues after the first AMPK site (equivalent to human MFF Ser157), as a hallmark phosphorylation event (Henschke et al. 2024). This site bears the Akt/S6K consensus motif and serves as an Akt1 substrate in vivo in the liver. Using elegant knock-in mice bearing nonphosphorylatable or phosphomimetic alleles of MFF at Ser131, the study also reveals that AKT-dependent phosphorylation of MFF and dynamic mitochondrial fragmentation are required for efficient insulin-induced suppression of hepatic glucose production, critical for suppressing insulin resistance and type 2 diabetes (Henschke et al. 2024). Notably, Akt phosphorylation of Ser131/Ser157 had the same effect on mitochondrial fission as AMPK or PKD phosphorylation of Ser129 and Ser173: They induce rapid fission.

Another context in which AMPK phosphorylation plays a critical role is innate immunity and the antiviral response (Hanada et al. 2020). Mitochondrial antiviral signaling (MAVS) is an OMM protein essential for the anti-RNA viral immune response, previously known to be inhibited by respiration inhibitors and sensitive to mitochondrial dynamics. MFF is essential for MAVS-mediated antiviral response, where it was shown to be required for the formation of MAVS clusters on the mitochondria, seemingly independent of its role in fission based on the behavior of MFF truncation mutants. AMPK phosphorylation of the aforementioned two MFF sites detailed in Toyama et al. (2016) is essential for immune tolerance during chronic infection through its ability to promote mitochondrial MAVS clustering (Hanada et al. 2020). Notably, MFF exhibits extensive alternative splicing events, some of which control phosphorylation of one of the AMPK sites (Toyama et al. 2016) and also regulation of MAVS and the antiviral response (Hanada et al. 2024).

Another AMPK substrate that may modulate the MFF–Drp1 interaction came from the discovery that AMPK directly phosphorylates the OMM protein AKAP1 (Hoffman et al. 2015), which anchors protein kinase A (PKA) near the mitochondria, where it can in turn directly phosphorylate DRP1 (Cribbs & Strack 2007, Gomes et al. 2011). Indeed, prior to the discovery of AMPK and Akt regulation of MFF, the major posttranslational event governing mitochondrial fission was reported to be PKA phosphorylation of DRP1.

AMPK phosphorylates several other proteins that contribute to mitochondrial dynamics.

In the last 5 years, several additional AMPK substrates that play roles in mitochondrial dynamics have been uncovered. In light of these discoveries, AMPK is firmly establishing itself as a pivotal orchestrator of mitochondrial fission. Although the exact mechanism was not elucidated, AMPK phosphorylation of ARMC10, which is located on the OMM, at Ser45 is important for promoting mitochondrial fission, as assessed by mitochondrial morphology, with shorter and fragmented mitochondrial staining in wild-type ARMC10 cells after AMPK activation compared to longer and fused mitochondria in ARMC10 knockout cells (Chen et al. 2019). The study also revealed ARMC10 interaction with core fission proteins, including Fis1 and Drp1, though many of the mechanistic details remain to be determined (Chen et al. 2019). The same study (Chen et al. 2019) and two earlier AMPK substrate screen studies (Hoffman et al. 2015, Schaffer et al. 2015) detected an uncharacterized protein, MTFR1L, as a potential AMPK substrate. Work has now validated MTFR1L as a bona fide AMPK substrate with Ser103 and Ser238 mapped as the phosphorylation sites. Upon cellular stress, AMPK phosphorylation of MTFR1L led to reduced levels of the IMM fusion protein Opa1, hence potentially indirectly tipping the balance toward fission (Tilokani et al. 2022). Expression of wild-type MTFR1L or phosphomimetic mutants (Ser103Asp/Ser238Asp) resulted in decreased OPA1 protein in MTFR1L knockout cells and mitochondrial fragmentation. On the other hand, in AMPK-deficient cells, an increase in OPA1 levels was detected, which could be reversed by expressing the phosphomimetic mutant MTFR1L but not the nonphospho-rylatable mutant (Ser103Ala/Ser238Ala). Loss of MTFR1L led to mitochondrial elongation and increased fusion events, which resulted in an interconnected mitochondrial network and increased mean mitochondrial area demonstrated by live-cell imaging and electron microscopy (Tilokani et al. 2022).

Another protein involved in mitochondrial dynamics, INF2, was suggested to be a novel AMPK substrate (Ding et al. 2024). INF2 nucleates actin polymerization at the ER, facilitating Drp1 recruitment and enhancing ER–mitochondria contacts (Fung et al. 2023, Korobova et al. 2013). In the recent study, the authors reported that AMPK phosphorylates INF2 on Ser1077, enhancing INF2 localization to the ER and promoting recruitment of Drp1 to the mitochondria (Ding et al. 2024). Importantly, however, the reported AMPK sites in INF2 do not match the AMPK consensus well nor do they score as well as a likely AMPK substrate using the latest whole kinome library motif technology ( Johnson et al. 2023). Additional studies are needed to validate this regulation.

Another modulator of mitochondrial networks is the Arf1 small GTPase (Ackema et al. 2014) and its activator GBF1, which is the direct target of the Golgi inhibitor compound Brefeldin A (Peyroche et al. 1999). AMPK phosphorylates GBF1 on Thr1337, disassociating GBF1 from the Golgi membrane and abolishing the action of GBF1 as an Arf1 GEF (Mao et al. 2013, Miyamoto et al. 2008). In addition, GBF1 interacts with Miro, the critical mitochondrial motility factor, and inhibition of GBF1 by Brefeldin A induced retrograded movement of mitochondria (Walch et al. 2018). Miro is itself targeted by PINK1 to induce recruitment of Parkin and its targeted destruction of Miro during mitophagy (Shlevkov et al. 2016). The connection between AMPK and GBF1 affords another route through which AMPK could coordinate mitochondrial motility, Golgi complex disassembly, and ER–mitochondria contacts depending on the energy state of the cells. Notably, AMPK activity was found localized near the Golgi complex in a study using a novel reporter (Miyamoto et al. 2015), and a pool of the upstream LKB1/STRAD/MO25 kinase complex localized to the Golgi complex based on interaction with the Golgi protein TBC1D23 (Tu et al. 2024). From a disease perspective, AMPK phosphorylation of GBF1 was proposed to be important in growth control of human oncocytoma tumors ( Joshi et al. 2015). Elegant work corroborated GBF1 Thr1337 as a direct AMPK phosphorylation site and, using CRISPR-generated Thr1227Ala GBF1 mutant knock-in cells, demonstrated that deletion of either AMPK or the single T–A mutant GBF1 mutation was sufficient to block the ability of AMPK activators to slow vesicular trafficking, suggesting that AMPK suppression of secretion may serve as an energetic restraint (Freemantle et al. 2024).

Finally, a study by Li et al. (2024b) also unbiasedly identified the ER protein PDZD8 as an AMPK substrate, which the authors showed was also conserved back to Caenorhabditis elegans, where its regulation by AMPK was critical for lifespan extension (Li et al. 2024a). The authors demonstrated that AMPK is required for promoting glutaminolysis at early times after glucose deprivation, and deletion of AMPK or PDZD8 blocked the ability of this metabolic adaptation, importantly without affecting the shift to fatty acid oxidation. They detected an interaction between PDZD8 and the rate-limiting glutaminase enzyme GLS1, which they proposed to be modulated by AMPK phosphorylation of PDZD8 (Li et al. 2024b). As GLS1 is localized to the inner mitochondrial matrix, whether PDZD8 could possibly be directly binding to GLS1 in vivo remains to be seen. Notably, however, PDZD8 is a well-established component of mitochondria-associated membranes, the aforementioned ER–mitochondria contacts. Using 3D serial scanning electron microscopy, Polleux and colleagues (Hirabayashi et al. 2017) demonstrated that the number and average surface of individual ER–mitochondria contacts were decreased by approximately 80% in PDZD8 knockout HeLa cells compared with control HeLa cells, suggesting that PDZD8 is required for tethering ER and mitochondria membranes in human cells. Indeed, in a study aimed at distinguishing mitochondrial fission initiated at the periphery from that at the midzone, PDZD8 was used as a marker of the ER contacts at the mitochondrial midzone, where all fission events required MFF (Kleele et al. 2021).

Taken collectively AMPK appears to be orchestrating a network of phosphorylation events to coordinate, promote, and fine-tune mitochondrial fission and dynamics, which is consistent with its role to sense and respond to perturbations to mitochondrial energy production. Mitochondrial fission of elongated mitochondrial networks into numerous smaller mitochondria following exposure to mitochondrial toxins is required for timing mitophagy and degradation of said damaged mitochondria (Burman et al. 2017), processes that AMPK also directly regulates, which we discuss in the next section.

AMPK Controls Macroautophagy and Mitophagy

AMPK plays a significant role in both general macroautophagy and mitophagy, the specialized form of autophagy that selectively degrades damaged mitochondria. Autophagy is an essential cellular quality control process, in which proteins, organelles, and macromolecules are delivered to the lysosomes for degradation. The process begins with the generation of an autophagosome and cargo recognition, which is followed by autophagosome maturation and its fusion with the lysosome (Dikic & Elazar 2018). Ultimately, autophagy serves two primary purposes. First, it facilitates degradation of cellular structures too large for other surveillance pathways, such as the ubiquitin–proteasome system; second, it enables cells to withstand starvation by recycling building blocks such as amino acids to sustain essential cellular functions (Dikic & Elazar 2018). Autophagy can also take a more targeted form, particularly when it comes to removal of mitochondria via autophagy, known as mitophagy. Mitophagy requires canonical autophagy machinery, in addition to specific markers signaling the need for mitochondria removal.

AMPK control of macroautophagy.

The autophagy machinery includes several multiprotein complexes, and AMPK regulates this machinery at multiple steps in the autophagy cascade. The turnover of aged and damaged molecules is a continuous process but the autophagy response to amino acid starvation specifically involves the mTORC1 pathway. Generally, nutrient starvation leads to mTORC1 inactivation, relieving the inhibitory mTORC1 phosphorylation of ULK1/ATG1, the only protein kinase in the canonical conserved autophagy pathway and one of four proteins (ULK1, ATG13, ATG101, and FIP200) that form the ULK complex, which mediates the initiation of autophagy. Once activated, ULK1 phosphorylates multiple components of the next step in the autophagy cascade, including the Beclin/Vps34/Vps15/ATG14L complex (Mercer et al. 2021), other downstream effectors such as ATG9 (Zhou et al. 2017), and the trafficking proteins Sec16A (Joo et al. 2016) and Sec23A (Gan et al. 2017), among others.

AMPK regulates autophagy in yeast and mammalian cells. A series of laboratories made discoveries in the early 2010s suggesting that AMPK regulates autophagy at a minimum via a two-pronged mechanism involving both direct AMPK phosphorylation of ULK1 on multiple phosphorylation sites and indirect suppression of the inhibitory mTORC1 phosphorylation of ULK1 on Ser757 via AMPK-dependent inhibition of mTORC1 (Alers et al. 2012). As these topics have been covered extensively elsewhere (e.g., Trefts & Shaw 2021), we do not go in depth here except to reexamine some of the early findings in light of data with a revised role for AMPK in the control of autophagy. The first direct link between AMPK and the core autophagy pathway was made when AMPK was shown to directly phosphorylate at least four residues on ULK1, Ser467, Ser555, Thr574, and Ser637, under low-energy conditions (Egan et al. 2011). Several contemporary studies reported additional sites in ULK1 as being regulated by AMPK (Kim et al. 2011, Loffler et al. 2011, Shang et al. 2011), and additional studies soon after reported that AMPK phosphorylation of Ser555 of ULK1 induced 14-3-3 binding to ULK1 (Bach et al. 2011, Mack et al. 2012). AMPK phosphorylation of ULK1 Ser555 has been widely used as a biomarker of AMPK activation in a wide variety of cell types and tissues ever since, but the precise impact of AMPK phosphorylation on ULK1 function was not clear from these early studies. At the time, there were no phospho-specific antibodies against downstream ULK1 substrates and none of the AMPK sites lie in canonical regulatory regions of the ULK1 kinase domain, so the biochemical consequences of AMPK phosphorylation on ULK1 kinase activity remained initially unclear despite both AMPK and ULK1 genetically playing pro-autophagy roles. However, genetically, ULK1 null cells, reexpressing AMPK nonphosphorylatable ULK1 mutants, accumulated defective mitochondria similar to those of ULK1 knockout cells, suggesting that the AMPK sites were somehow required for the ability of ULK1 to control mitochondrial turnover (Egan et al. 2011). Additional substrates for AMPK in the next step of the autophagy cascade, the Beclin1/Vsp34 complex, were reported (Kim et al. 2013). Following this finding, ULK1 phosphorylation of distinct sites in several of these same proteins in the Beclin1/Vps34 complex was decoded (Egan et al. 2015, Mercer et al. 2021, Russell et al. 2013). Other key proteins in autophagy are also dually targeted by AMPK and ULK1 on distinct sites, including ATG9 (Weerasekara et al. 2014, Zhou et al. 2017) and PIKfyve (Karabiyik et al. 2021, Liu et al. 2013, Yordanov et al. 2019).

Importantly, in the last few years, the original model that AMPK activation simply activates ULK1 while mTOR activity inhibits ULK1 has been brought into question. Early studies in the 2010s indicated that glucose starvation, which activated AMPK, was not increasing autophagy flux and was specifically preventing autophagosome maturation (Ramírez-Peinado et al. 2013), though glucose deprivation regulates many pathways besides AMPK. Subsequent studies of cell culture with direct AMPK activators similarly found that these activators or glucose deprivation suppresses autophagic flux, in a manner genetically requiring AMPK (Nwadike et al. 2018). In a technical tour de force using Halo-tagging of four endogenous autophagosome proteins, video microscopy, and the latest selective AMPK activators and inhibitors, Barnaba et al. (2024) provided compelling data indicating that AMPK activation after glucose starvation leads to the accumulation of mobile phagophores bound by WIPI2, which fail to tether to donor membranes, leading to a reduction in autophagosome maturation. A key unaddressed question detailed in this study is why AMPK activation simultaneously increases initial autophagy foci formation and inhibits autophagosome maturation, but the authors speculate that phagophores exist in a primed state when AMPK is active and can be rapidly tethered to membranes to expand and mature into autophagosomes, perhaps when the energetic state becomes more favorable. Aspects of these findings mesh well with other cellular studies reporting that AMPK inhibits the kinase activity of ULK1 toward its canonical substrates Atg13 and Atg14 (Park et al. 2023, Longo et al. 2024). A model in which AMPK binding to ULK1 and phosphorylation of Ser555 inducing 14-3-3 binding function to protect ULK1 from degradation was derived from extensive analysis of ULK1 phosphorylation site mutants and coimmunoprecipitation analysis of the AMPK/ULK1 complex (Park et al. 2023).

While the data in these studies indicate that AMPK is inhibiting ULK1 phosphorylation of Atg13 and Atg14 in these conditions, there are other published studies in which CCCP or the combination of antimycin and oligomycin activates AMPK and contemporaneously induces phosphorylation of reported ULK1 phosphorylation sites in Beclin Ser30 and ATG16L1 Ser275, all within the first 5 min of mitochondrial damage (Hung et al. 2021). It is possible that the rapid activation of ULK1 kinase activity toward Beclin and ATG16L1, parallel to the activation of AMPK kinase activity in the study by Hung et al. (2021), may reflect AMPK suppression of mTORC1 dominating the activation of some subcellular pools of ULK1, despite Ser555 phosphorylation on other pools. It is also possible that another kinase besides ULK1/2 is responsible for targeting Beclin Ser30 and ATG16L1 Ser275 in these cells after CCCP. Another possibility is that distinct subcellular pools of ULK1 complexes, AMPK complexes, and mTORC1 complexes are differentially responding to some of these stimuli and the combination of specific phosphorylation events on individual ULK1 molecules dictates whether ULK1 is near the lysosome, near a damaged mitochondrion, or out near ER involved in catalyzing initiation of omegasome formation at the start of macroautophagy. AMPK phosphorylation of both ULK1 and ATG9A triggers their binding to 14-3-3 (Bach et al. 2011, Longo et al. 2024, Mack et al. 2012, Park et al. 2023, Weerasekara et al. 2014), consistent with the concept holding them in a locked and inactive state until conditions improve (Park et al. 2023). Combined with the fact that roughly half of the core components of the autophagy cascade are directly phosphorylated on distinct sites by AMPK, ULK1, and mTORC1 and that all three of these protein kinase complexes phosphorylate one another, it becomes clear that additional studies are needed to decipher the precise timing and subcellular localization governing the tug-of-war consequences of each regulatory event on autophagy induction and resolution. That said, AMPK is not simply activating ULK1 kinase activity and macroautophagy; thus, additional studies are needed to define the relative contributions of and timing requirements for AMPK in macroautophagy in different tissues after different stimuli.

Upon examining the effects from tissue-specific deletion of AMPK in mice, a consistent observation emerges: Phosphorylation of ULK1 Ser555 is ablated and the mTOR site in ULK1 Ser757 is increased, coincident with increases in both p62 levels and defective mitochondria, as observed in AMPK knockout livers and hepatocytes (Egan et al. 2011), skeletal muscle (Bujak et al. 2015), and brown adipose tissue (Mottillo et al. 2016). Furthermore, studies of mice bearing a dominant negative AMPKα2 transgene demonstrated that AMPK was needed for exercise-induced colocalization of mitochondrial reporters with lysosomal reporters, as was ULK1, with both appearing to play a positive role in autophagy induction (Laker et al. 2017). A follow-up study uncovered a specific pool of AMPK localized to the OMM, and electroporating mouse muscle with a mitochondria-targeted AMPK-substrate-blocking peptide prevented the ability of endurance exercise to promote colocalization of the Mito reporter with lysosomes in vivo (Drake et al. 2021). A study using genetically encoded mitophagy reporters revealed that MK-8722 alone, in the absence of exercise, increased mito-lysosomes in skeletal muscle, though notably while decreasing mito-lysosomes in liver, illustrating the potential nuances and tissue-specific aspects to this regulation (Longo et al. 2024). Despite all these findings, it remained unclear from early studies whether the accumulation of altered mitochondria in AMPK or ULK1 null cells or tissues was due to their roles in regulating proteins in general macroautophagy or whether they also directly phosphorylated proteins dedicated to mitophagy itself.

AMPK control of mitophagy.

Currently, there are two well-established forms of mitophagy: ubiquitin-dependent mitophagy and ubiquitin-independent mitophagy (Zachari & Ktistakis 2020). The more deeply studied form is ubiquitin-dependent mitophagy, which relies on the mitochondria-localized serine/threonine kinase PINK1 and the E3 ubiquitin ligase Parkin, both of which are mutated in familial forms of early-onset Parkinson’s disease (Narendra & Youle 2024). What was known for a decade is that Parkin, which is normally cytoplasmic in unstressed cells, translocates to the mitochondria upon exposure to membrane depolarizers like CCCP or combinations of antimycin and oligomycin. Contemporaneously, PINK1 is stabilized on the OMM, where it phosphorylates Parkin Ser65 and ubiquitin Ser65 (Harper et al. 2018, Lazarou et al. 2015). Once activated by PINK1 at the mitochondria, Parkin ubiquitinates mitochondrial proteins, which signals the recruitment of autophagy machinery. Work has shed light on earlier steps in Parkin activation prior to its phosphorylation by PINK1, unexpectedly highlighting an additional point at which AMPK signaling is required (Hung et al. 2021) with the identification of a highly conserved phospho-site, Ser108, that lies within a 10–amino acid activation (ACT) domain of Parkin (Gladkova et al. 2018). Treatment of cells with both ETC inhibitors and mitochondrial depolarizing agents resulted in maximal AMPK activation and Parkin phosphorylation on Ser108 within 5 min in the cytoplasm but not mitochondria (Hung et al. 2021). However, association of Parkin with the mitochondria did not happen until considerably later, approximately 20–30 min after mitochondrial insults; Pink1 phosphorylation of Parkin at Ser65 and ubiquitin Ser65 also occurred only after this time, by which point, Parkin Ser108 phosphorylation was rapidly declining. In Hung et al. (2021), endogenous Parkin Ser108 was genetically reliant on AMPK in mouse embryonic fibroblasts and livers and on ULK1/2 in livers, though studies suggest that Parkin Ser108 may be a direct AMPK phosphorylation site in most cell types (Longo et al. 2024). The sequences surrounding Ser108 in the ACT domain of Parkin conform to both the optimal AMPK substrate consensus and the ULK1 optimal consensus, so perhaps it could be regulated by either in some circumstances, but both studies find it AMPK dependent in all circumstances (Longo et al. 2024). This phosphorylation event may be an early surveillance step to partially activate Parkin, causing its translocation to mitochondria to start surveying for mitochondria lacking membrane potential. By this model, only fragmented mitochondria with depolarized membranes would have stabilized Pink1 to fully activate Parkin to mark them for degradation (Trefts & Shaw 2021). A study by Ganley and colleagues (Longo et al. 2024) proposed a model in which AMPK may suppress NIX-dependent mitophagy of functional mitochondria while promoting Parkin-dependent mitophagy of damaged mitochondria. Additional studies are needed to assess whether all components of ubiquitin-independent mitophagy (e.g., via BNIP3, NIX, FUNDC1, BCL2L13, FKBP8) are suppressed by AMPK or whether this is specific to NIX/BNIP3 or to any mitophagy receptors whose LIR motifs are phosphorylated by ULK1 (Onishi et al. 2021).

AMPK was independently tied to phosphorylation of a distinct nonconsensus site in Parkin (Ser9), and AMPK-activated Parkin was proposed to inhibit necroptosis by preventing the formation of the RIPK1/RIPK3 complex by promoting polyubiquitination of RIPK3 (Lee et al. 2019). Other studies have also connected AMPK to RIPK1, first proposing that RIPK1 is a required scaffold for AMPK to phosphorylate TSC2 Ser1387 and inhibit mTORC1 (Najafov et al. 2021) and subsequently indicating that AMPK can control phosphorylation of RIPK1 Ser415, which suppresses its function, although this site also does not conform to the consensus AMPK substrate motif (Zhang et al. 2023). Nonetheless, in mouse tissues bearing a RIPK1 Ser415Ala knock-in mutant or bearing AMPKα1 deletion, RIPK1 was constitutively activated (Zhang et al. 2023). More work is needed to illuminate the interconnection between AMPK-mTOR and RIPK1-RIPK3 control of cell death and how PINK1-Parkin signaling interfaces with cell death decisions. Importantly, AMPK phosphorylation of Parkin Ser108 occurs following any stress that activates AMPK, including weak ETC inhibition and glucose deprivation, most of which do not activate Pink1 and the classic ubiquitin-dependent mitophagy cascade. Thus, more mysteries of Parkin function after energy stress but without PINK1 activation remain to be investigated.

AMPK Drives Transcriptional Induction of Lysosomal and Mitochondrial Biogenesis

Mitochondrial biogenesis is triggered by increasing energy demands to generate more ATP and involves the growth and division of preexisting mitochondria (Hock & Kralli 2009). The existing mitochondrial network is expanded through addition of new material, which requires increased production of mitochondrial proteins, as well as enhanced lipid production that supports enlargement of IMMs and OMMs. The mitochondrial genome (mtDNA) itself encodes 13 proteins, all of which are essential components of the oxidative phosphorylation system, but the vast majority of mitochondrial proteins (over 1,000) are encoded by the nuclear genome. Therefore, the signal to stimulate mitochondrial biogenesis must be relayed to the nucleus, enabling transcription factors to induce transcription of genes that encode mitochondrial proteins (Hock & Kralli 2009). Despite being one of the first mitochondrial processes shown to be regulated by AMPK (Bergeron et al. 2001, Reznick & Shulman 2006, Zong et al. 2002), the molecular mechanism by which AMPK controls the transcriptional induction of mitochondrial biogenesis remained unresolved for decades. Similarly, despite several studies exploring this process, a specific AMPK substrate responsible for mediating lysosomal biogenesis, which is essential to sustained autophagy and mitophagy, had also remained elusive. Folliculin interacting protein 1 (FNIP1), which forms a complex with folliculin (FLCN) (Baba et al. 2006), has been identified as a missing link between AMPK and both lysosomal and mitochondrial biogenesis (Malik et al. 2023). Lysosomal biogenesis is dominantly regulated by transcription factor EB (TFEB) and TFE3, closely related members of the microphthalmia/transcription factor E (MiT/TFE) family (Napolitano & Ballabio 2016). Similarly, though different transcription factors are involved, the master regulator of mitochondrial biogenesis across cell and tissue types is the transcriptional coactivator peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC1α) (Lin et al. 2005). Overexpression of PGC1α in muscle is sufficient to convert type IIb glycolytic fibers into mitochondria-rich type II and I fibers (Lin et al. 2002). While PGC1α is controlled by several posttranslational mechanisms, many studies noted linkages between AMPK and PGC1α. Overexpression of a constitutively active form of the AMPKγ3 subunit in the absence of treatments increased PPPARGC1A mRNA expression (Garcia-Roves et al. 2008). The upregulation of GLUT4, cytochrome c, and UCP2 mRNAs after metformin and AICAR treatments was ablated in PGC1α knockout muscle ( Jäger et al. 2007). Over the next 15 years, various mechanisms were reported as to how AMPK controls PGC1α, including two sites that were reported to be directly phosphorylated by AMPK in vitro (Thr177 and Ser538) ( Jäger et al. 2007), and how PGC1α acetylation is controlled via indirect regulation of SIRT1 activity (Canto et al. 2010), but no consensus model of how AMPK activation leads to mitochondrial biogenesis has been consistently reported across studies. We reported a step-by-step, temporal mechanism through which AMPK phosphorylation of the protein FNIP1 on at least five distinct sites alters the GTPase activating protein (GAP) activity of the FLCN/FNIP1 complex to tightly control the nuclear localization of TFEB, resulting in the induction of mRNAs for lysosomal genes and PGC1α (Malik et al. 2023).

AMPK and mTORC1 are well-established to serve antagonistic ying-yang functions, controlling catabolism and anabolism across all eukaryotes (González et al. 2020). mTORC1 is active when energy and nutrients, especially amino acids, are plentiful. Under nutrient-rich conditions, mTORC1 phosphorylates TFEB on three sites, two of which are conserved in TFE3 (Ser142 and Ser211), leading to their retention in the cytosol by 14-3-3 binding and exclusion from the nucleus (Puertollano et al. 2018). In contrast, AMPK is activated under the opposite circumstances. During energy stress and starvation of nutrients such as glucose and other carbon sources like glutamine, AMPK and mTOR converge on several rate-limiting effectors, through direct phosphorylation, exerting opposing effects. Previous work demonstrated that TFEB and TFE3 were stimulated by AMPK (Eichner et al. 2019, Young et al. 2016), though the precise mechanism of how AMPK regulated TFEB remained elusive. Although FLCN was identified in 2002 as a tumor suppressor mutated in Birt–Hogg–Dubé syndrome, which is a hamartoma syndrome clinically related to other syndromes caused by loss of the genes encoding TSC1/TSC2, LKB1, or PTEN, the precise function of FLCN remained obscure for over a decade after its discovery. Its binding partners FNIP1 and FNIP2 were identified in 2006 (Baba et al. 2006). Early attempts to understand their function led to coimmunoprecipitation approaches that identified AMPK and mTOR in association, but precisely what FNIP1/FLCN was controlling and whether AMPK or mTOR regulated FLCN/FNIP1 remained unknown for many additional years. One important function of this complex was identified in 2013, when two teams revealed that the FLCN/FNIP1/2 complex served as a GAP for the RagC and RagD GTPases (Petit et al. 2013, Tsun et al. 2013). The Rag GTPases function as obligate heterodimers, in which the active complex consists of GTP-bound RagA or RagB in complex with GDP-bound RagC or RagD. Activation of mTORC1 by intracellular amino acids occurs when amino acids stimulate GTP binding to RagA and RagB, promoting binding to Raptor and assembly of the activated mTORC1 complex. In the absence of amino acids, the Rag proteins take up an inactive conformation (GDP-bound RagA or RagB and GTP-bound RagC or RagD), inactivating and relocalizing mTORC1 to the cytosol (Liu & Sabatini 2020). Active Rag heterodimers also interact with and promote recruitment of TFEB to the lysosomes, leading to mTORC1-dependent phosphorylation and retention of TFEB in the cytosol. Depletion or inactivation of Rag proteins prevents recruitment of TFEB to lysosomes (Martina et al. 2012, Roczniak-Ferguson et al. 2012, Settembre et al. 2012). A catalytic arginine in FLCN is required for amino acid–dependent translocation of TFEB and TFE3, connecting control of FLCN/FNIP1 GAP activity to amino acid regulation of RagC (Lawrence et al. 2019). The GAP activity of the FLCN/FNIP1 complex allows accumulation of RagC or RagD in the GDP-bound, active state, promoting mTORC1 activation ( Jansen et al. 2022). Elegant cryo-electron microscopy and crystallography studies have now fully elucidated the structural underpinnings of these changes (Cui et al. 2023, Jansen et al. 2022, Lawrence et al. 2019).

In our work, we demonstrate that AMPK phosphorylation of FNIP1 at five serines (Ser220, Ser230, Ser232, Ser261, and Ser593) leads to TFEB dephosphorylation and TFEB nuclear translocation in cells expressing wild-type FNIP1 but not in cells expressing a mutant SA5 FNIP1 (where all five serines are mutated to alanine) (Malik et al. 2023). Mechanistically, AMPK phosphorylation of FNIP1 may be inhibiting the GAP activity of the FLCN/FNIP1 complex, even in the presence of amino acids, which would result in accumulation of RagC in its GTP loaded, inactive form. In SA5 FNIP1 mutant conditions, there would be accumulation of the active GDP–RagC conformation, resulting in constitutive TFEB recruitment to the lysosome and mTOR phosphorylation of TFEB. If this were true, overexpressing a GTP-locked mutant of RagC in SA5 FNIP1 mutant conditions should be able to reverse the strong TFEB phosphorylation consistently observed in these cells, which was indeed observed, while the opposite effect was seen with GDP-locked RagC mutants. These findings suggest that AMPK phosphorylation of FNIP1 blocks GAP activity of the FNIP1/FLCN complex, though more work is needed to directly assess GAP activity (Malik et al. 2023). Mechanistically, AMPK activation in wild-type FNIP1 cells but not SA5 FNIP1 cells resulted in RagC and TFEB dissociation from the lysosome and mTOR dissociation from TFEB. In contrast, in cells in which FNIP1 cannot be phosphorylated by AMPK, mTOR was continuously associated to TFEB, keeping it inactive and sequestered in the cytoplasm. Thus, AMPK phosphorylation of FNIP1 tightly controls TFEB nuclear translocation by suppressing formation of the Rag/mTOR/TFEB complex at the lysosome surface. AMPK phosphorylation of FNIP1 temporally controls the initial induction of lysosome mRNAs, subsequently inducing mitochondrial biogenesis via TFEB induction of PGC1α (Malik et al. 2023).

Within 2–4 h following mitochondrial damage or direct AMPK activators, TFEB nuclear translocation results in transcription of the CLEAR network lysosomal genes and a parallel increase in PGC1α mRNA in wild-type cells but not in AMPK knockout cells or FNIP1 SA5 cells (Malik et al. 2023). PGC1α was previously reported to be a canonical CLEAR gene, with TFEB binding motifs in its originally defined proximal promoter (Settembre et al. 2013), though this has not been extensively studied since. In HEK293 cells, the PPARGC1A gene product most strongly induced appeared to be a shorter 270–amino acid splice variant, N-terminal-PGC1α (NT-PGC1α). This variant contains the N-terminal transactivation domain and the two LXXLL-like (L2–L3) motifs that interact with nuclear receptor family members but lacks the C-terminal degron motif that triggers degradation of the full-length protein ( Jannig et al. 2022, Zhang et al. 2009). Thus, NT-PGC1α is a more stable form with a longer protein half-life while retaining full transactivation functionality. AMPK activation increased NT-PGC1α protein expression in wildtype cells but not in SA5 FNIP1 mutant cells, and depletion of PGC1α prevented induction of AMPK-/FNIP1-dependent mitochondrial genes in RNA-sequencing and quantitative PCR experiments. This finding supports the hypothesis that NT-PPARGC1A gene induction observed in the first wave of AMPK-dependent CLEAR network gene transcription may be in part responsible for the increased expression of mitochondrial genes seen in the second wave of gene expression several hours later (8–16 h). Consistent with this, AMPK activators increased mitochondrial DNA content and enhanced mitochondrial marker staining in imaging experiments in wild-type cells but not AMPK knockout or SA5 FNIP1 cells. Mitochondrial respiration was also reduced in AMPK knockout and SA5 FNIP1 cells compared to wild-type cells (Malik et al. 2023).

Given that PGC1α is a transcriptional coactivator, PGC1α coordinates with other transcription factors to induce mitochondrial biogenesis. Indeed, in addition to PGC1α, AMPK also enhances expression of the estrogen related receptor ERRα in KRAS-driven colon cancer (Fisher et al. 2015). Malik et al. (2023) noted that mRNAs of genes encoding the nuclear receptors ESRRA and ESRRG were increased from 2 to 16 h after AMPK activation and depletion of PPARGC1A suppressed this effect, consistent with previous reports that PGC1α stimulates expression of ERRα (Hock & Kralli 2009). Deletion of ERRα, which can bind to PGC1α via LXXLL-like binding sites, prevented the upregulation of many of the AMPK-/FNIP1-dependent mitochondrial genes. Additional papers further highlight the critical role of ERRα and ERRγ in regulating mitochondrial function and energy metabolism in skeletal muscle (Sopariwala et al. 2023, Wattez et al. 2023). Deletion of both ERRα and ERRγ in double knockout mice led to significant mitochondrial dysfunction, characterized by decreased expression of genes related to mitochondrial biogenesis and oxidative phosphorylation, resulting in reduced mitochondrial number and size, as well as impaired oxidative enzymatic capacity (Sopariwala et al. 2023). This dysfunction contributed to diminished aerobic capacity and exercise tolerance in double knockout mice. The AMPK–ERRα axis also plays a pivotal role in regulating mitochondrial function and promoting ischemic revascularization and muscle recovery in the context of diet-induced obesity (Sopariwala et al. 2022). AMPK enhanced expression of ERRα, which was crucial for transcriptional activation of angiogenic genes that enable muscle regeneration following ischemic injury. In diet-induced-obesity mice, endogenous ERRα levels were suppressed, which correlated with reduced AMPK activity and impaired mitochondrial function. Treatment with AMPK activators increased ERRα expression and its recruitment to mitochondrial gene promoters, demonstrating a direct link between AMPK and mitochondrial function via ERRα (Sopariwala et al. 2022). These findings suggest that targeting the AMPK–ERRα pathway could enhance mitochondrial function and improve ischemic recovery in diabetic peripheral artery disease.

Work by Xiao et al. (2024) reaffirms the central role of FNIP1 phosphorylation by AMPK at Ser220 in mitochondrial biogenesis and function following exercise-induced AMPK activation. Using muscle-specific mouse models, the authors demonstrated that mice expressing a non-phosphorylatable FNIP1 Ser220Ala mutant exhibited significantly reduced exercise performance, characterized by decreased running distance and increased blood lactate levels, indicating impaired metabolic adaptation due to compromised ETC complex formation and respiratory function in Ser220Ala transgenic muscle. In contrast, mice expressing a phosphomimetic FNIP1 (Ser220Asp) mutant displayed no such deficits. Moreover, Xiao et al. (2024) reported that FNIP1 phosphorylation at Ser220 promoted mitochondrial biogenesis and function, as evidenced by increased mitochondrial protein levels and improved metabolic responses during exercise. Overall, the observations by the authors further underscore FNIP1’s pivotal role in mitochondrial biogenesis and function; however, further research is needed to determine whether differences exist between mitochondrial damage and exercise-induced FNIP1 phosphorylation by AMPK. Notably, studies from a completely different aspect of cell biology have also identified FNIP1 as a Cys sensor in the cell, and mapped specific Cys residues sensitive to reactive oxygen species, somehow modulating mitochondrial health (Manford et al. 2020). One striking aspect of this work is that the Cys residues in FNIP1 lie within 10 amino acids of one of the AMPK sites, neither of which is conserved in FNIP2, so much additional work is needed to clarify the roles and regulation of these proteins that dominantly control lysosome and mitochondrial biogenesis.

Finally, several earlier studies reported hyperactivation of AMPK signaling in animal models of genetic disruption of FLCN (El-Houjeiri et al. 2019, Possik et al. 2014, Yan et al. 2014). Now observing the acute need for AMPK-mediated regulation of the FLCN/FNIP1 complex to maintain both lysosomal and mitochondrial homeostasis, one interpretation of the earlier results could be a feedback loop in which chronic TFEB activation in the FLCN-deficient state may trigger energetic deficiencies in response to which AMPK is biochemically activated. Future studies examining temporal and spatial control of AMPK, mTORC1, TFEB, PGC1α, and the FLCN/FNIP1 complex may shed further light on which components dominate in different physiological and pathological states.

CONCLUDING REMARKS AND FUTURE PERSPECTIVES

This review has centered largely on the direct substrates of AMPK that mediate its rewiring of metabolism under conditions of low energy, with a focus on the plethora of connections made between AMPK and mitochondrial biology over the last decade. Given the depth of molecular detail that has emerged regarding how AMPK controls mitochondrial biology, it will be important to define how critical AMPK is to the control of each of these processes in different tissues in vivo. The magnitude of the requirement for AMPK, compared to other kinases that play roles in different aspects of mitochondrial health, such as PINK1 (Harper et al. 2018, Singh & Muqit 2020), ULK1 (Lazarou et al. 2015, Vargas et al. 2019), TBK1 (Richter et al. 2016, Vargas et al. 2019), and HRI (Chakrabarty et al. 2024, Fessler et al. 2020, Guo et al. 2020), remains a critical area for future studies as well. Much of the historical focus has been on exercise and the role of AMPK in the therapeutic response to mitochondrial inhibitors such as metformin. However, considerable work is still needed to determine the requirement of newly reported AMPK targets in these processes. How critical AMPK is for autophagy and maintenance of mitochondrial integrity over time in different tissues, specifically in the context of mitochondrial diseases, muscular dystrophies, and neurodegenerative diseases, all of which share biochemical defects in bioenergetics and mitochondrial biology, remains an interesting and important question that still needs to be resolved. Finally, whether direct AMPK activators can mimic beneficial effects of exercise, metformin, or caloric restriction on the human health span remain critical topics in need of much further study.