Autophagy is the process by which lysosomes degrade organelles and cellular components. Three different forms of autophagy coexist in most cells (Figure 1): Microautophagy, chaperone-mediated autophagy and macroautophagy [1].

Microautophagy, a still poorly characterized type of autophagy in mammalian cells, consists in the direct engulfment of cytoplasmic components in intralysosomal tubules and vesicles, which are produced by various modifications of the lysosomal membranes. Chaperone-mediated autophagy is a selective lysosomal pathway for the degradation of cytosolic proteins with KFERQ-like sequences, in which both cytosolic and lysosomal chaperones facilitate the transport of these proteins for degradation into the lysosomal lumen through a multimer of the lysosomal protein LAMP2A spanning the lysosomal membrane. Macroautophagy is the most important and better known form of autophagy. Initially, cytoplasmic components are surrounded by a double membrane, with still an unclear origin, forming a vacuole, which is called an autophagosome. Autophagosomes fuse with endosomes and lysosomes to form autolysosomes with an acidic pH and where the degradation of the sequestered material by lysosomal hydrolases occurs. While autophagy has been usually considered a nonselective and an in bulk degradation process, there are also selective autophagic pathways. For example, the selective degradation of mitochondria, peroxisomes or ribosomes by autophagy is called, respectively, mitophagy, pexophagy or ribophagy. In this paper, unless otherwise stated, the term “autophagy” will only refer to macroautophagy, and we will not review here the two other forms of autophagy.

Autophagy is regulated by a high number of extracellular stimuli, which are transduced through a large variety of signalling pathways with the kinase mTORC1 occupying a central position in most of them [2,3]. In mammalian cells, the main regulators of autophagy are hormonal and nutritional factors. Among those, the regulation by amino acids and insulin, their signalling pathways and the underlying mechanisms on the autophagic machinery have been investigated in most detail (see e.g. [4]). In contrast, the information on other molecules that also regulate autophagy is much scarce.

Autophagy, like protein degradation [5], in spite of being a catabolic process requires an input of energy, as first established by Plomp et al. [6]. ATP (adenosine triphosphate) is required for, at least, two autophagic steps: (i) sequestration of the cytosolic material in autophagosomes; and (ii) maintenance of the activity of the proton pump (vacuolar-type ATPase or V-ATPase) at the lysosomal membrane to produce the lysosomal acidification required in the autophagic flux. Since most ATP in the cell derives from glucose metabolism, this metabolite should be important for the normal functioning of autophagy. However, glucose is also a nutrient, and it is well known that nutrient depletion induces autophagy to provide the cell with energy and building blocks. Therefore, whether glucose affects positively or negatively autophagy is a matter of discussion, as we will see below. In this review, we aim to present an overview of what is specifically known on the signalling pathways that regulate autophagy in response to glucose or glucose-deprivation. Due to space limitations, we will mention only those papers dealing, more or less directly, with this question, while other specific stimuli regulating autophagy, other signalling pathways and the autophagy machinery will not be reviewed here. We will mainly concentrate on mammalian cells, but first we will give in the next section a brief and general account on this subject in yeast.

Most of our current knowledge on the molecular machinery of autophagy derives from studies in yeast that, initially, were mainly carried out in the laboratories of Yoshinori Ohsumi in Japan, Michael Thumm in Germany and Dan Klionsky in USA. In the late 90s, two different forms of autophagy were recognized in yeast: micro- and macroautophagy [7]. The first was described in yeast as a process by which cell components are surrounded by cytoplasmic invaginations or finger-like protrusions of the vacuole, which eventually form intravacuolar vesicles that, together with their content, are degraded by the vacuolar hydrolytic enzymes. In macroautophagy, cytoplasmic components are surrounded by a membrane from a compartment called the preautophagosomal structure (PAS) that closes to form a double-membrane autophagosome [8]. Then, the outer membrane of the autophagosome fuses with the vacuolar membrane and the single membrane autophagic body that has entered into the vacuole is degraded by vacuolar hydrolases [9]. The identification of the ATG (autophagy-related) genes provided insights into the molecular basis of autophagy. Many of the codified proteins in yeast are conserved in mammals, but there are also mammalian specific ATG proteins. Now, the role of some, but not all, of the more than 30 identified Atg proteins is known and has been discussed elsewhere [10,11,12]. Most of them participate in the initiation of autophagy, but some others are essential for other autophagic steps.

In yeast, autophagy is mainly affected by nitrogen starvation, but glucose starvation plays also an important role, as we will see below in a few examples. The growth of Saccharomyces cerevisiae in a medium without glucose induces the synthesis of gluconeogenic enzymes (fructose-1,6-bis-phosphatase, malate dehydrogenase, phosphoenolpyruvate carboxykinase) and also of the galactose and maltose transporters. When these cells are subsequently incubated in a medium containing glucose, the transcription of these genes is suppressed and the induced proteins are degraded [13] in the general process of catabolite repression. In the specific case of fructose-1,6-bisphosphatase, it has been proposed that glucose induces its sequestration within double membrane vesicles, which eventually fuse with the vacuole by a process that requires protein synthesis [14].

Something similar occurs with the methylotrophic yeasts Pichia pastoris and Hansenula polymorpha when they are incubated in a medium containing methanol [15,16]. This induces the synthesis of enzymes that facilitate methanol uptake, such as alcohol oxidase in peroxisomes and formate dehydrogenase in the cytosol, as well as an overproduction of peroxisomes. When these yeasts are next cultured in a medium with glucose, formate dehydrogenase and the peroxisomes produced in excess are selectively degraded [15]. In the specific transfer from methanol to glucose, peroxisomes are degraded either by macroautophagy (macropexophagy, which sequesters single peroxisomes within pexophagosomes that fuse with the vacuole to degrade their content) in H. polymorpha [16] or by microautophagy (micropexophagy, which sequesters and degrades a cluster of peroxisomes within vacuolar membranes and a micropexophagy specific apparatus) in P. pastoris [15]. Pexophagy requires some specific proteins, such as for example Vac8 on the vacuolar membrane, the peroxisome receptor Atg30 or the peroxins Pex3 and Pex14 on the membrane of peroxisomes [17,18,19]. Other proteins involved in the activation of the selective autophagy triggered by glucose have been identified in yeast. One of those, Ras2, is at least involved in the degradation of fructose-1,6-bisphosphatase, because its mutants block the degradation of this protein [20], and other proteins, such as Gpr1 and Gpa2, are required in pexophagy for glucose sensing [21].

The data mentioned above indicate that some types of specific autophagy are induced by glucose. However, other experiments indicate that glucose could also produce in yeast the opposite effect, namely an inhibition of autophagy, at least in the case of the nonselective autophagy. For example, this can occur in S. cerevisiae via inhibition by glucose of the expression of Snf1p, the closest yeast homologue of AMPK (AMP-activated protein kinase). Both Snf1p and the cyclin-dependent kinase Pho85p have been described to control autophagy in opposite directions, activating and inhibiting it, respectively, via the autophagic proteins Atg1 and Atg13 [22]. Another example is provided by S. cerevisiae maintained in a medium with plenty of nutrients. Under these conditions, autophagy is inhibited via PKA (cAMP-dependent protein kinase A) and also, and independently, via TorC1 [23,24]. Both kinases phosphorylate Atg13 at different sites and inhibit, in a different way, the Atg1 complex [24], a key site that integrates different autophagic signals, as indicated also by the above reported effects of Snf1p and Pho85p [22]. In this regard, it is relevant that under glucose deprivation the regulatory and catalytic subunits form the PKA inactive complex [25], because this supports the inhibitory role of glucose on autophagy. Sch9, the closest yeast homologue of mammalian S6K1 and a substrate of TorC1, is also involved in nutrient sensing as an inhibitor of autophagy and in cooperation with PKA [26], although it is apparently less important.

In summary, while it is known that depletion of nitrogen mainly inactivates TorC1 and induces autophagy, the signalling pathways by which glucose-deprivation produces the same effect are less known. It is possible that in yeast TorC1 mainly responds to nitrogen levels while the response to carbon levels occurs mainly via the Ras/PKA pathway [27] and also, but to a lesser extent, via Snf1, TorC1 and Sch9, as described before.

In addition to the sequestration step, glucose in yeast can also affect positively autophagy at later steps, for example regulating the assembly of the V-ATPase, which, as mentioned before, consumes ATP. The V-ATPase is a supramolecular complex responsible for the acidification of organelles in eukaryotic cells [28] and has two main domains, V₁, responsible for ATP hydrolysis, and V_o, responsible for proton translocation. In yeast, glucose is important to modify the activity of the V-ATPase independently of protein synthesis [29]. Thus, in the absence of glucose, only 15–20% of the subunits of the V₁ domain remain attached to V₀, while with glucose this figure rises to 50–70% [30]. This reduces ATP consumption when the availability of glucose is limited, because the separation of the V₁ and V₀ portions produces the loss of their respective functions. To explain how glucose increases the V₁–V_o coupling, an activation of the proton pump Pma1 at the plasma membrane that produces an increase in cytosolic pH [31] and a regulation by Ras/cAMP/PKA [32] have been postulated.

In summary, glucose in yeast can both induce and inhibit autophagy. For example, glucose induces a selective autophagy during an adaptive phase to specifically remove enzymes or organelles previously overproduced, and increases also the assembly and the activity of the V-ATPase and, thus, of the vacuolar acid hydrolases. However, glucose in a nutrient-rich medium can also suppress nonspecific autophagy to prevent an unnecessary autodigestion of the cell, and this mainly occurs via Ras/cAMP/PKA but also via TorC1 and Sch9 and by inactivation of Snf1p.

In mammalian cells, as in yeast, autophagy is a process highly regulated by nutrients. Over the last years, the effects of amino acids on autophagy have been extensively studied. However, the role of glucose is less known, despite its implication in diseases such as diabetes or different types of cancer. In mammals low and high blood glucose levels induce the production and release of, respectively, glucagon by alpha cells and insulin by beta cells of pancreas. It is also well known that, for example in liver, glucagon activates autophagy while insulin inhibits it. However, the amount of glucose in the environment may also produce an autophagic response unrelated to these hormones. This response is mainly due to the role of glucose as an important determinant of the cellular energy status, but also to other glucose effects, such as oxidative stress and ROS accumulation. In the following sections we will not review the glucose effects on autophagy mediated by hormones, because they have been extensively described elsewhere, and we will deal exclusively with those other glucose effects.

As in yeast, in mammalian cells the availability of glucose promotes the assembly, via PI3K (phosphatidylinositol 3-kinase), of the V-ATPase [33] favouring the acidification of lysosomes required for the optimal activity of the lysosomal hydrolases. Although further details on this signalling pathway are unknown, AKT and p70S6K have been discarded as regulators [33]. Since lysosomal degradation is a late step in autophagy, these data could suggest a positive role for glucose in autophagy. However, only few works have postulated a positive effect of glucose on autophagy. For example, Ravikumar et al. [34] showed in cultured cells that glucose induces autophagy by inhibition of the main negative regulator of autophagy, mTOR. In addition, trehalose, a disaccharide composed of two molecules of glucose, activates autophagy, as demonstrated by an increased degradation of aggregates of huntingtin and α-synuclein mutants [35,36].

In contrast to these data, most investigators consider that in mammalian cells glucose inhibits autophagy [37,38,39,40]. When energy depletion occurs in cells incubated in a glucose-free medium, autophagy is activated as a mechanism to restore ATP from cellular components [41]. This drop in energy results in an increase in the AMP/ATP ratio, which will be sensed by AMPK to induce autophagy [42] by different mechanisms that will be later described. However, other authors conclude that ATP depletion is not necessary to induce autophagy by glucose deprivation. For example, endoplasmic reticulum (ER) stress caused by 2-deoxy-D-glucose (a non-metabolizable glucose analogue, 2-DG) induces autophagy [43]. Although 2-DG treatment reduces intracellular ATP content and activates AMPK, it is the restoration of the ER homeostasis (by mannose) what causes a decrease in LC3-II levels without affecting ATP or AMPK [43]. In addition, autophagy can be also stimulated by an increase in reactive oxygen species (ROS), caused both by glucose deprivation or a high glucose concentration, which can produce ER stress [44,45,46].

In the following sections, we will discuss different signalling pathways that regulate autophagy by glucose. In spite of the apparent importance of PKA in the signalling pathways by which glucose regulates autophagy in yeast, its relevance in mammalian cells remains elusive and, therefore, will not be mentioned here. Some of the implicated proteins, such as AMPK, p53 or p27^KIP1, induce autophagy when ATP decreases in cells incubated without glucose. Others, such as p38, ERK (extracellular signal-regulated protein kinase) or JNK (c-Jun amino-terminal kinase), respond to ROS generated by an excessively low or high availability of glucose, activating autophagy. Finally, we will also mention the possible role of IKK (IκB kinase) and FOXO (forkhead class O) proteins in the induction of autophagy by glucose, which occurs without an apparent relation with ATP or ROS. All this reflects the existence in mammalian cells of a complicated network that interconnects different signalling pathways in the regulation of autophagy by glucose.

In addition to its obvious importance for a better understanding of the physiological roles of autophagy, the issue of the regulation of autophagy by the available glucose levels has also important pathological implications. We will present here only two examples in relationship with two important pathologies: cancer and diabetes.

The induction of autophagy under conditions of glucose deprivation is very important for the maintenance of cell homeostasis. When the availability of glucose is a limiting factor and thus depletion of energy occurs, as is the case in most cancers, an increased autophagy allows to obtain energy from the superfluous cell constituents. Cancer cells survive and even grow in poor nutrient microenvironments thanks to an induction of autophagy via AMPK, p53 or JNK [124]. In the case of K-RAS transformed cells, JNK induces autophagy and when these cells are incubated under low glucose conditions their growth is not affected [104]. In addition, human hepatoma tissues with low glucose uptake and high K-RAS expression show increased LC3-II levels [104]. Thus, induction of autophagy in environments with low glucose availability seems to be essential to maintain the survival of tumour cells.

We have also discussed above that glucose deprivation, as well as hyperglycemia, induces autophagy by signalling pathways mediated by ROS production. This is of great importance in the pathology of diabetes, where the production of ROS by hyperglycemia has been shown to be an important factor influencing its clinical complications [125]. The hyperglycemic environment in which the cells survive leads to ROS accumulation, mitochondrial dysfunction and ER stress that causes an incorrect folding of the proteins, which as a consequence tend to form aggregates [126]. Therefore, in pancreatic β-cells an increase in autophagy takes place in order to degrade these aggregates that cannot be degraded by proteasomes. In this sense, autophagy plays a protective role [127], favouring the structure maintenance, mass and function of β-cells [128]. For example, it has been reported that incubation for 72 h of pancreatic cells in the presence of 30 mM glucose induces autophagy as a survival mechanism and that inhibition of this autophagy with siRNAs targeting ATG7 produces an increased activity of CASPASE 3 that compromises the survival of β-cells [58]. Along the same line, a deficiency in autophagy during hyperglycemia will produce an increase in oxidative stress, causing the formation of atherosclerotic plaques [129]. Although autophagy in hyperglycemia is mainly protective, it has been also shown that an excess can also have negative consequences by diminishing mitochondrial mass, as it occurs in non-obese diabetic G-K rats where autophagy inhibition can lead to a reduction in oxidative stress, allowing the recovery of the mitochondrial function [105]. The ER stress caused under hyperglycemic conditons, and also under high fat diet or obesity, causes a sustained activation of JNK1, which has been implicated in the development of obesity and insulin resistance [130,131]. In fact, treatment of mice with JNK inhibitors reduces hyperglycemia and improves insulin sensitivity [132,133]. A direct relationship among diabetes, JNK1 activation and autophagy has been shown in G-K rats [105].

Diabetic patients are at significantly higher risk of ischemia/reperfusion injury [134]. During ischemia, there is a constraint of blood supply, causing a deficiency in the oxygen and glucose provided to the cells for its cellular metabolism. After ischemia, reperfusion must occur to allow the survival of the tissue. In cardiomyocytes, glucose deprivation during ischemia produces an increase in autophagy dependent on AMPK, RHEB/mTORC1 and SIRT1/FOXO3 [48,135]. Presence of oxygen during this step is of main importance for autophagy to take place, as its absence impairs autophagy [136]. This increased autophagy confers a protective role, leading to survival of the cells under these conditions [48,135]. The same has been also shown in vivo during ischemia/reperfusion [137,138]. Treatment with an mTOR inhibitor to induce autophagy after myocardial infarction decreases the infarct size [137]. In the same line, during reperfusion, autolysosome maturation is inhibited [136]. This impairment of late steps in autophagy, associated with increased ROS, causes cardiomyocyte death [138]. However, it has been also shown that an excessive autophagy later during reperfusion can cause extensive cell death [48].

Although some of the signalling pathways that orchestrate in yeast the regulation of autophagy by glucose have been conserved in mammalian cells, a more complex signalling network appears to be involved in these cells. Here, AMPK, which is activated by a deficit in energy, is the most important kinase of these signalling pathways. This kinase exerts its effect mainly, but not exclusively, through the mTOR kinase, the main nutrient sensor in the cells. However, there are other proteins that are also implicated in the regulation of autophagy downstream of AMPK in the response to glucose deprivation, e.g.,p53 and p27^KIP1, also activated by ATP depletion. In addition, ROS also influence autophagy in response to glucose deprivation or hyperglycemia, by signalling pathways that implicate other kinases such as p38MAPK, JNK or ERK1/2. Finally, glucose-induced autophagy can be also regulated by IKK/NF-κB or FOXO, independently of ROS or the ATP levels and AMPK function.

The importance of the regulation of autophagy by glucose is highlighted by human diseases in which glucose levels are affected. For example, autophagy serves to preserve cell integrity in cancer cells growing in glucose-limiting environments and in pancreatic β-cells under diabetes. Therefore, understanding the mechanisms of autophagy regulation by glucose is of major interest in order to better understand these and other pathologies.