Autophagy inhibitors

Abstract

Autophagy is a lysosome-dependent mechanism of intracellular degradation. The cellular and molecular mechanisms underlying this process are highly complex and involve multiple proteins, including the kinases ULK1 and Vps34. The main function of autophagy is the maintenance of cell survival when modifications occur in the cellular environment. During the past decade, extensive studies have greatly improved our knowledge and autophagy has exploded as a research field. This process is now widely implicated in pathophysiological processes such as cancer, metabolic, and neurodegenerative disorders, making it an attractive target for drug discovery. In this review, we will summarize the different types of inhibitors that affect the autophagy machinery and provide some potential therapeutic perspectives.

Introduction

The maintenance of cellular homeostasis is continuously challenged by abrupt or long-lasting changes in the environment. In order to adapt, cells have to modify their anabolic and catabolic processes. Autophagy and the proteasome constitute the two universal intracellular catabolic pathways. The term autophagy, derived from the Greek word for “self-eating”, was coined by Christian de Duve to account for observations of vesicles containing components of the cytoplasm in various degrees of disintegration [1]. This was after his discovery of the lysosome in 1955, which won him the Nobel Prize in 1974 [2]. In contrast to the proteasome which allows the direct degradation of intracytoplasmic proteins, the process of autophagy enables cells to degrade proteins, protein complexes and organelles via a lysosome-dependent mechanism. Present in all eukaryotic cells, three different types of autophagy can be distinguished: macroautophagy, microautophagy and chaperone-mediated autophagy, that differ in their mechanisms and functions [3]. Macroautophagy (hereafter referred to as autophagy) is the most characterized form and will be the subject of this review. Readers interested by microautophagy and chaperone-mediated autophagy should consult other reviews [4, 5].

Autophagy machinery

Autophagy is characterized by the formation of double-membrane vesicles called autophagosomes within the cytoplasm, which engulf bulk cytosol or organelles. Autophagosomes then fuse with the lysosomal system converting them into autolysosomes. The content of the autophagosomes is therefore degraded by lysosomal hydrolases and the metabolites generated as a result of autophagy are reused either as sources of energy or building blocks for the synthesis of new macromolecules (Fig. 1). The origin of the autophagosomal membrane that forms the phagophore is still not well understood, but the endoplasmic reticulum seems to be one of the main components of nascent autophagosomes [6, 7].

Fig. 1.

Schematic representation of the signalling pathways that control autophagy machinery. Autophagy is regulated by various growth and nutrient signalling. Growth factors activate the class I PI3K/Akt signalling pathway and mTOR complex 1 (mTORC1). In this condition, mTORC1 negatively regulated the two key kinase complexes, Vps34 and ULK1, leading to a prevention of autophagy. Under growth factors starvation, the activity of mTORC1 is suppressed, releasing the inhibition of Vps34 and ULK1 complexes. mTORC1 can also sense the cellular concentration of amino acids. High amino acids concentration positively affects mTORC1. When a decrease in amino acids occurs, mTORC1 is inhibited. This leads to Vps34 and ULK1 complexes activation. The level of ATP is detected by the kinase AMPK that negatively regulate mTORC1 and positively affect ULK1. In case of a decrease in ATP, AMPK is activated and this results in mTORC1 inhibition and activation of ULK1 complex allowing the induction of the autophagy machinery. The first step of the autophagy machinery is the formation of a double membrane named phagophore. This membrane then expands and forms the autophagosome. This vesicle sequesters complex proteins and organelles and is characterized by the presence of LC3 (green circles) attached at their membranes. Autophagosomes finally fuse with lysosomes and their content is degraded by the lysosomal hydrolases. The resulting metabolites are transported into the cytoplasm and used for the synthesis of new macromolecules or as a source of energy

The machinery of autophagy is tightly regulated by complex signaling pathways and core autophagy proteins. The identification of the autophagy-related (ATG) genes, using genetic screens for autophagy-defective mutants in yeast, was historically a major achievement for the field [8]. More than 36 ATG genes have been identified in yeast, and most of them are evolutionarily conserved. The hierarchy and the recruitment of ATG proteins have been reviewed in mammals [9, 10]. Proteins involved in the autophagy machinery, including ATG proteins, can be categorized into six functional groups that cooperate to allow the formation of autophagosomes: (1) UNC-51-like kinase 1 (ULK1) kinase complex; (2) PIK3C3, the class III phosphatidylinositol 3-kinase (PtdIns3K), also known and hereafter named vacuolar protein sorting 34 (Vps34), kinase complex; (3) FYVE or PX domain-containing proteins that bind to PtdIns 3-phosphate (PtdIns3P); (4) the ATG5-12 ubiquitin-like conjugation system; (5) the microtubule-associated protein 1-light chain 3 (LC3) phosphatidylethanolamine conjugation system; and (6) ATG9a, the only transmembrane protein among the ATG proteins [7, 11, 12].

Autophagy signaling pathways

The signaling pathways upstream of the autophagy machinery are complex and involve multiple mediators. The best characterized regulator of autophagy is the mammalian target of rapamycin (mTOR), also known to control cellular homeostasis. Indeed, as a serine/threonine kinase, mTOR can sense multiple stimuli including the presence of growth factors, nutrient and energy levels and stress in order to maintain cellular metabolism and growth. At the molecular level, mTOR integrates the signaling of many upstream pathways such as for example the phosphoinositide 3-kinase (PI3K)-Akt and AMP-activated kinase (AMPK) pathways [13]. Of note, the complex 1 of mTOR (mTORC1) acts as a major checkpoint in signaling pathways regulating autophagy (Fig. 1), whereas mTORC2 is not a direct autophagy regulator [14]. Activated mTORC1 downregulates autophagy by preventing the activity of ULK1 and Vps34 complexes [15–18]. During cell starvation, i.e. when the amino acids or growth factors concentration in the cell environment drops, mTORC1 is inhibited. This releases its break on the two downstream ULK1 and Vps34 complexes and as a consequence allows the nucleation of autophagosomes (Fig. 1).

ULK1 is a serine/threonine protein kinase that exists within the cells as protein complexes with Atg13, FIP200 (FAK family kinase-interacting protein of 200 kDa) and Atg101. This complex is regulated by the energy sensor AMPK (Fig. 1). One of the substrates of the ULK1 complex is the Vps34 complex [19–24].

Vps34 is a lipid kinase that converts phosphatidylinositol (PtdIns) to phosphatidylinositol 3-phosphate (PtdIns3P) [25–27]. Of note the pool of PtdIns3P is tightly regulated by phosphatases such as the two myotubularins, MTMR3 and MTMR14 (also called Jumpy) and the dual-domain protein tyrosine phosphatase σ (PTPσ) [28, 29]. The formation of PtdIns3P by the Vps34 complex recruits downstream proteins containing either a FYVE (Fab1, YOTB, Vac1 and EEA1) zinc-finger domain or a PX (Phox homology) domain such as double FYVE-containing protein 1 (DFCP1) and WD-repeat domain phosphoinositide-interacting protein (WIPI). Subsequently, an expansion phase of the phagophore occurs with the recruitment of the two ubiquitin-like conjugation system, the ATG12-ATG5 complex and the LC3 complex, which ends with the formation of an autophagosome [7, 11].

Autophagy function in physiology

Autophagy occurs under many physiological conditions. Notably it has been demonstrated that autophagy is critical during mammalian development and also during the period of nutrient deprivation immediately after birth [30, 31]. At the cellular level, autophagy is crucial in removing unfolded proteins and damaged or superfluous organelles such as mitochondria, peroxisomes, ribosomes, endoplasmic reticulum, or endosomes, and therefore acts as a “quality control” process [32]. Under starvation and stress, autophagy is dramatically induced to recycle cellular components and to supply the building blocks so that cellular homeostasis is maintained. Autophagy is thus viewed as a cell survival process in response to stress.

Autophagy and diseases

Dysregulated autophagy is implicated in numerous pathophysiological processes such as neurodegenerative diseases, infectious or metabolic diseases and cancer [33–37]. In the case of cancer, most of the compounds in the clinic cause cellular stress and activate autophagy after drug treatment in vitro [38]. Even if it was originally proposed that autophagic cell death is part of the mechanism of anticancer drugs, the emerging picture is that the increased autophagy seen during anticancer drug treatment could be a survival response of the dying cells, rather than a cause of cell death. One illustration of this concept was highlighted using a chemical screening study with 59 compounds that induced both autophagy and cell death. It was found that one of the ATG proteins, ATG7, when knockdown prevented autophagy, but not cell death. This study led the authors to conclude that “cell death is rarely, if ever, executed by autophagy in human cells” [39, 40].

Autophagy inhibitors

Given its potential roles in many diseases, various preclinical studies have been undertaken to develop therapeutic agents targeting autophagy. However, the field had to face major challenges because autophagy is a dynamic process that is difficult to measure and quantify. For example, the accumulation of autophagosomes does not necessarily demonstrate an increase in autophagy by itself, but could simply reflect that autophagy is blocked at a late stage [41]. The recommendations on relevant assays for investigating autophagy are published elsewhere [42]. One of the widely used readout for measuring autophagy is LC3. Most researchers use a chimeric protein of LC3 conjugated to a fluorescent protein such as Green Fluorescent Protein (GFP). This probe, when transfected into cell lines, is a useful tool to measure autophagy and to perform cell-based screens.

Different compounds are described in the literature as potential inhibitors of the autophagy process. However, most of them are poorly selective, limiting their application. This review will be divided into two parts and will focus on the best characterized autophagy inhibitors (Table 1). The first part will cover the proximal inhibitors, targeting druggable proteins or pathways involved in the initial steps of the core autophagy machinery. The second part will present the late stage inhibitors of autophagy which target the functions of lysosomes.

Table 1.

List of the compounds discussed in the article

Target	Inhibitor	References
PI3K	3-methyladenine	[45–48]
	Wortmannin	[48, 52–54]
	LY294002	[47, 54, 58]
	PT210	[60]
	GSK-2126458	[63]
Vps34	Spautin-1	[79]
	SAR405	[82]
	Compound 31	[84]
	VPS34-IN1	[85]
	PIK-III	[86]
ULK	Compound 6	[98]
	MRT68921	[100]
	SBI-0206965	[24]
Proteases	Pepstatin A	[101, 103]
	E64d	[101, 103]
V-ATPase	Bafilomycin A1	[105–107]
Lysosomes	Clomipramine	[108]
	Lucanthone	[109]
	Chloroquine	[110–112, 118–124]
	Hydroxychlorquine	[110–112, 118–124]
	Lys05	[127]
	ARN5187	[128]
	Compound 30	[129]

Pan-PI3K inhibitors

Phosphoinositide 3-kinases (PI3Ks) are divided into class I, class II and class III, and encompass eight isoforms. The class III isoform, corresponding to Vps34, activates the autophagy machinery, while the class I PI3K triggers mTOR signaling pathway and inhibits autophagy. The contribution of class II PI3K activity on autophagy is unclear [25, 27, 43, 44]. A group of pan-PI3K inhibitors, including 3-methyladenine (3-MA), Wortmannin and LY294002 have been used as autophagy inhibitors. The following section will present the knowledge on these three compounds as well as two additional novel pan-PI3K inhibitors, PT210 and GSK-2126458 (Fig. 2).

Fig. 2.

Selected small molecule inhibitors of autophagy. Autophagy inhibitors can be classified as early stage inhibitors (in blue) and late stage inhibitors (in yellow) of the autophagy machinery. The early stage inhibitors encompass the Pan-PI3K, Vps34 and ULK inhibitors. The second category corresponds to compounds that affect the function of lysosomes. All of these compounds are at the preclinical phase except for GSK-2126458, which is being evaluated in clinical trials. Clomipramine, chloroquine and hydroxychloroquine have been FDA-approved. Compounds with a asterisk correspond to putative autophagy inhibitors, as no report clearly demonstrate their inhibitory effect on autophagy

3-Methyladenine

3-MA was first discovered via the screening of purine-related substances using isolated hepatocytes from starved rats [45], and is the most widely used autophagy inhibitor [46]. One of the targets for 3-MA is Vps34 [47], but this compound also inhibits the activity of class I PI3K, and, as a consequence, exerts a dual effect on autophagy. It was described that 3-MA not only suppresses starvation-induced autophagy, but it can also promote autophagic flux when administrated under nutrient-rich conditions with a prolonged period of treatment [48]. This pan-PI3K inhibitor has a low potency, requiring its use at ~10 mM concentrations to prevent autophagy in vitro on cellular assays. At such high concentrations, it is known that 3-MA can also affect other kinases, including class I PI3K as mentioned earlier, p38MAPK or c-Jun kinase. With such off-target activities, it is recognized that this compound affects many cellular processes such as glycogen metabolism, lysosomal acidification, endocytosis, and mitochondrial permeability transition [49]. Another caveat of this compound is its poor solubility. To improve the solubility and also the potency of 3-MA, a library of 29 derivatives was synthetized [50]. While the authors described three novel 3-MA derivatives with improved solubility, the potency of these compounds on autophagy readout remains limited with IC50s between 18 and 670 µM [50].

Wortmannin

Wortmannin, a mold metabolite initially described as having anti-inflammatory activities [51], is a highly potent pan-PI3K inhibitor with IC50s from 10 to 50 nM for PI3K classes I, II and III [52, 53]. Wortmannin seems to exert persistent effects on Vps34 while it inhibits transiently the class I PI3K regardless of the nutrient conditions [48, 54]. It is known that Wortmannin achieves its inhibitory effect via covalent irreversible binding [52, 55]. However, it also has inhibitory effects on phylogenetically related kinases such as mTOR, DNA-dependent protein kinase (DNA-PK) or ataxia telangiectasia mutated (ATM) protein kinase [56, 57].

LY294002

LY294002 is the first synthetic inhibitor of PI3K and it inhibits autophagy on cellular assays [47, 54, 58]. This compound has limited potency (IC50s at the µM level) for the class I PI3K, Vps34, and affects mTOR and DNA-PK [59].

Although the three inhibitors described above block autophagy, they have limited potency (excepted for Wortmannin) and have off-target effects on different related lipid and protein kinases (Fig. 3). Moreover, these compounds are not fully characterized for their pharmacological properties such as solubility, stability, ADME or pharmacokinetic (PK) parameters. Therefore great caution needs to be exercised when analyzing data with these inhibitors in preclinical studies and on in vivo experiments.

Fig. 3.

In vitro profiling of pan-PI3K and Vps34 inhibitors. The activity of the compounds in biochemical assays and recombinant enzymes are indicated. Results of the compounds on the inhibition of Vps34 target dependent cellular assay (GFP-FYVE transfected cells) and of autophagy cellular assay (GFP-LC3 transfected cells) are also mentioned in this table. The data correspond to IC50s (nM) determination and are from [55, 60, 79, 82, 84–86]. Compounds with a asterisk correspond to putative autophagy inhibitors, as no report clearly demonstrate their inhibitory effect on autophagy

Novel pan-PI3K inhibitors

PT210

Using the structure of Drosophila Vps34 in complex with pan-PI3K inhibitors, the group of RL. Williams performed structure-based design to identify compounds selective for Vps34 versus the isoform γ of the class I PI3K [60]. In particular, starting from PIK-93, another pan-PI3K inhibitor [61], the authors were able to select PT210 which shows a tenfold in vitro selectivity ratio between Vps34 (IC50 = 450 nM) and the isoform γ of the class I PI3K (IC50 = 4.4 µM) (Fig. 3). The effect of PT210 on autophagy is yet to be demonstrated.

GSK-2126458

The PI3K/Akt signaling pathway has been extensively investigated during the last 20 years, and oncogene activation or tumor suppressor gene inactivation has been described in this pathway. As a consequence, numerous novel pan-PI3K inhibitors have been developed by pharmaceutical industries as targeted therapies for cancer. These molecules demonstrate improved drug-like properties that led to many clinical trials [62]. Among the pan-PI3K compounds in clinical studies, GSK-2126458 [63], developed as an inhibitor for the class I PI3K and mTOR, is also highly active on Vps34 using biochemical assays and recombinant proteins (Fig. 3). To the best of our knowledge, no report was available on the regulation of autophagy by GSK-2126458. With such a pan-PI3K/mTOR/Vps34 profile, GSK-2126458 could, in theory, both activate and inhibit autophagy. While it will be challenging to demonstrate the inhibition of autophagy after treatment with GSK-2126458, the results from clinical trials could confirm the therapeutic advantage of this inhibitor.

Vps34 inhibitors

Overview of Vps34

Vps34, the most ancient paralog of the PI3Ks, was originally identified in S. cerevisiae as a protein involved in endosomal sorting of proteins towards the vacuole, the yeast equivalent of the mammalian lysosome [64]. It is now recognized that Vps34 is involved in a broad range of vesicular trafficking pathways [26, 27]. This includes not only autophagy [65, 66], but also the endosomal trafficking of receptors, for example the epidermal growth factor receptor (EGFR), platelet-derived growth factor receptor (PDGFR) or Transferrin receptor [67, 68]. The molecular explanation behind these multiple pathways comes from the fact that Vps34 is active within complexes with several accessory proteins [25, 69]. At least two distinct complexes are described, both encompass Vps34, Vps15, Beclin-1 plus either Atg14L/Barkor or UVRAG. Atg14L-containing Vps34 complex is localized in autophagosomal structures and participates in the generation of autophagosomes, while the UVRAG-containing Vps34 complex is found at the endosomes [25, 69]. Another dimension for the functions associated to Vps34 come from the multitude of PtdIns3P-binding proteins, including phosphatases, kinases and adaptor proteins that are present within the cells [26].

Target credentialing of Vps34 in the autophagy process

During the last 50 years, different in vitro studies have validated Vps34 as a key target for autophagy using antisense oligonucleotides or siRNA [47, 70, 71]. However, the recent results from Vps34 knockout mice have generated controversy about the role of Vps34 on the activation of autophagy in vivo. While two initial studies have shown that conditional knockout of Vps34 in sensory neurons and in T lymphocytes does not elicit an autophagy defect [72, 73], more recent results indicated that the deletion of Vps34 prevents the formation of autophagosomes [68, 74, 75]. In fact when looking at the constructs of the Vps34 knockout mice models, it appears that two different strategies were used, with the deletion of exon 17/18 and the targeting of exon 4, respectively. Whether this difference can explain the discrepancy observed on autophagy remains to be demonstrated.

Considering the key role of Vps34 in the autophagy machinery, many groups were interested in developing compounds targeting this lipid kinase. The following section will first present the characterization of an indirect Vps34 inhibitor, Spautin-1 and will then discuss the two recently described catalytic inhibitors of Vps34 (Fig. 2).

Spautin-1

4-[[3,4-(methylenedioxy)benzyl]amino]-6-chloroquinazoline (MBCQ) was originally characterized as a potent and selective inhibitor of phosphodiesterase type 5 (PDE5) [76, 77]. Using a cell-based screen with the GFP-LC3 cell line, this compound was able to inhibit autophagy [78, 79]. Derivatives of MBCQ were synthetized to improve the effect on autophagy and to decrease the off-target effect on PDE5 leading to the identification of the quinazoline derivative Spautin-1 (specific and potent inhibitor of autophagy 1) [79]. This compound does not affect the catalytic activity of Vps34, but promotes the degradation of Vps34 complexes. The exact mechanism of action of Spautin-1 was identified and it was demonstrated that it inhibits the ubiquitin-specific peptidases, USP10 and USP13 [79]. These two enzymes deubiquitinate the protein Beclin-1. Under glucose-free conditions, treatment with Spautin-1 indirectly promotes the ubiquitination of Beclin-1 and leads to its proteasomal degradation. The consequence of this is the destabilization and degradation of Vps34 complexes and an inhibition of autophagy. The IC50 of Spautin-1 in GFP-LC3 cellular assay is 740 nM. Moreover, this compound promotes the death of cancer cells under nutrient deprivation, when autophagy is activated, but not in nutrient-rich conditions. Spautin-1 also appears to be essentially non cytotoxic to normal cells. It could be interesting to explore whether Spautin-1 could also affect the endosomal trafficking of receptors. Preclinical studies using Spautin-1 showed a synergy with Imatinib, indicating a potential therapeutic application of this compound in chronic myeloid leukemia [80].

Catalytic inhibitors of Vps34

A major breakthrough in the field has been made with the very recent characterization of two series, the pyrimidinone and the bisaminopyrimidine, originated from compound libraries at Sanofi and Novartis, respectively.

Pyrimidinones

With the aim to screen for autophagy inhibitors in a physiologically relevant environment, HeLa cells stably expressing GFP-LC3 were employed in a cell image-based assay with a subset of our proprietary compound collection, consisting of ~250,000 compounds. The screen identified compounds with as yet unidentified targets, and chemical series active against autophagy-relevant biological enzymes, ULK and Vps34 in biochemical assays using recombinant enzymes [81]. The initial hit compounds displayed an activity on Vps34 in the nM range, but they showed a lack of selectivity regarding the class I PI3K. A chemical optimization program was initiated and allowed the identification of SAR405, a highly potent and selective inhibitor of Vps34. SAR405 has a binding equilibrium constant K _D of 1.5 nM. In order to evaluate its activity on a dedicated Vps34 cellular assay, HeLa cells were transfected with a fluorescent probe, GFP-FYVE, that binds to PtdIns3P, the reaction product of Vps34. SAR405 completely inhibited the signal and an IC50 activity was measured at 27 nM. As compared to the pan-PI3K inhibitors, SAR405 presented an exquisitely selective profile with regards to class I, class II PI3K, mTOR and other lipid kinases (Fig. 3). Moreover, no activity was found when profiled on a panel of more than 200 protein kinases present in a cellular extract at 1 µM [82, 83]. Using SAR405 we showed that the catalytic activity of Vps34 is necessary for maintaining the size of late endosomes/lysosomes and the function of lysosomes during vesicle trafficking, thus confirming previous data with knockdown of Vps34 [67, 68]. We then demonstrated that SAR405 is able to inhibit autophagy induced by two different stimuli, starvation and treatment with a catalytic mTOR inhibitor AZD8055 [82, 83]. These results suggest that SAR405 might prevent the catalytic activity of the two Vps34 complexes, Atg14L- and UVRAG-containing Vps34 complexes. Finally we showed that adding the Vps34 inhibitor in combination with the FDA-approved mTOR inhibitor everolimus synergized to trigger antiproliferative activity in renal tumor cell lines, indicating a potential therapeutic application in renal cell carcinoma.

As SAR405 demonstrated an excellent Vps34-specific inhibitory profile, we went on to find compounds that were chemically optimized for in vivo applications. Compound 31 was then characterized as a Vps34 inhibitor for oral route with excellent in vitro ADME and PK properties in rats and mice. When tested in vivo, compound 31 induced a sustained target inhibition after a single administration at 100 mg/kg. The in vivo effective plasma concentration to achieve 50 % inhibition was estimated at 0.37 µM [84]. To our knowledge, this is the first report of an optimized Vps34 inhibitor that can be used to investigate the biology of this target, including autophagy.

Bisaminopyrimidines

The bisaminopyrimidine family is exemplified by VPS34-IN1 and PIK-III, with an activity of 25 and 18 nM, respectively, using biochemical assays and Vps34 recombinant enzyme [85, 86]. These two compounds display a highly selectivity profile when tested on a panel of protein kinases and lipid kinases at 1 µM or using dose–response experiments (Fig. 3). The group of D. Alessi demonstrated that Vps34 regulates the activity of serum- and glucocorticoid-regulated kinase-3 (SGK3), a PX-bearing domain protein, indicating that SGK3 phosphorylation can be used as a biomarker of Vps34 activity in vitro [85]. Unfortunately, the drug-like properties of VPS34-IN1 and PIK-III, including their in vitro ADME properties and in vivo target engagement, were not disclosed.

While no data concerning autophagy was provided with VPS34-IN1 [85], the effect of the PIK-III compound was well described using LC3 fluorescent readout and western blot [86]. It was shown that PIK-III inhibits the degradation of autophagy substrate receptors such as p62, NBR1 and NDP52 when autophagy is induced by the catalytic mTOR inhibitor AZD8055. In addition PIK-III can inhibit mitophagy, a subtype of autophagy that triggers the selective degradation of mitochondria. Using SILAC proteomics, the group of L. Murphy identified a novel molecule namely NCOA4 (nuclear receptor co-activator 4) that is increased following treatment with PIK-III. NCOA4 interacts with ferritin and is required for its degradation into autophagolysosomes after starvation. This study and the discovery of PIK-III allowed them to identify a new function for autophagy, the regulation of iron levels [86].

Collectively, the three catalytic inhibitors mentioned above illustrate the dual role of Vps34 in autophagy and in endosomal pathway. Rather than being considered as autophagy-specific inhibitors, these agents are powerful tools to dissect the role of Vps34 in these two processes.

ULK inhibitors

ULK1, present in higher eukaryotes, corresponds to Atg1 kinase in yeast. This kinase has four orthologues in humans: ULK2, ULK3, ULK4 and STK36 [87]. ULK1 function as the most upstream protein kinase in the autophagy machinery [9, 16].

Target credentialing of ULK in the autophagy process

Multiple evidences have validated ULK as a target for autophagy modulation [88–92]. In particular, in a siRNA screen with a library that targeted 753 human kinases, ULK1 was found as a key protein involved in autophagy [93]. Of note, a possible redundancy between ULK1 and ULK2 was described in mouse embryonic fibroblasts (MEFs) [94]. The ULK knockout mouse indicated that this kinase plays a role in mitophagy during erythroid differentiation [95], but also in axon growth, endocytosis and in lung development [96, 97].

Catalytic inhibitors of ULK

Because of its role in autophagy and its druggable properties, ULK has attracted attention for drug discovery. Three groups very recently disclosed the structures of potential ULK inhibitors (Fig. 2). Of note our group also identified hit compounds on ULK2 after the phenotypic screen on GFP-LC3 HeLa cells [81].

Compound 6

The initial aim of the group of KM. Shokat was to solve the structure of human ULK1. As they met some issues with the apo protein, they decided to look for a compound that could bind to ULK1 and increase its stability in order to facilitate its crystallization. A screening of 764 molecules against ULK1 was done using a ³²P-ATP radioactive assay with myelin basic protein (MBP) as a generic substrate [98]. Using the enzymatic assay, they identified compound 6 as their best hit. This inhibitor displayed a high affinity with an IC50 on ULK1 of 8 nM. While the effect of this compound on autophagy was not depicted, the authors showed that it is not specific toward ULK1 when profiled in a small panel of protein kinases. Indeed compound 6 inhibited the activity of 37 kinases by more than 50 % at a concentration of 250 nM and is therefore not yet optimized for in vitro testing on cellular assays.

MRT68921

Starting from the TANK-binding kinase 1 (TBK1) inhibitor, MRT67307 (IC50 = 19 nM) [99], the group of IG. Ganley investigated its effect on kinases with similar catalytic domain [100]. They found that MRT67307 potently inhibits ULK1 and ULK2 (IC50 of 45 and 38 nM, respectively) using a similar biochemical assay as the one used by KM. Shokat. When looking at a closely related series of analogues generated during the original TBK1 screen, the authors selected MRT68921 as a more potent inhibitor for ULK. This compound displays a tenfold increased affinity as compared to MRT67307 on ULK1 and ULK2 (IC50 of 2.9 and 1.1 nM, respectively) [100]. However, when they investigated its potential off-target effect on a panel of 80 protein kinases, they found that 44 kinases were inhibited (<50 % activity remaining) by treatment of MRT68921 at 1 µM. Among them, TBK1 and AMPK, two kinases known to play a role in the autophagy process, were found. Using knockout MEF cells, it was demonstrated that MRT68921 was still capable of inhibiting autophagy in the absence of TBK1 and AMPK pathways. To rule out the fact that MRT68921 could inhibit autophagy through other kinases, the authors generated a drug-resistant ULK1 mutant. Using biochemical studies they selected the mutation of the gatekeeper residue of ULK1 methionine 92 to threonine (M92T) that retains kinase activity in the presence of MRT68921. It was shown that ULK1/2 double knockout MEF cells reconstituted with M92T ULK1 protein restored autophagy in the presence of MRT68921, while this compound was still able to inhibit autophagy in MEF cells reconstituted with wild type ULK1. This result demonstrates that MRT68921 inhibits autophagy specifically through ULK1.

SBI-0206965

The group of RJ. Shaw performed a small screen with analogues of a focal adhesion kinase (FAK) inhibitor [24]. SBI-0206965 was selected as the best ULK inhibitor (IC50 = 108 and 711 nM for ULK1 and ULK2, respectively). Having identified that the phosphorylation of Vps34 Ser249 is one of the reaction catalyzed by ULK1, they confirmed the activity of the compound on ULK1 using this biomarker in overexpressing Vps34 HEK293T cells. When profiled in a panel of 456 protein kinases at a concentration of 1 µM, this compound inhibited the activity of 33 kinases by more than 50 % [24]. Finally they investigated the effect of SBI-0206965 on the GFP-LC3 cell line. It was shown that this compound could prevent autophagy induced by the ATP-competitive mTOR kinase inhibitor AZD8055 [24]. Similar results were obtained with ULK1 siRNA suggesting that this target is essential under these experimental conditions [24]. One caveat of this recent study is that SBI-0206965 was used at 5 or 10 µM concentration on cellular assays. At such high concentrations and given the weak specificity of the compound, caution needs to be taken when interpreting the in vitro results with SBI-0206965.

As a conclusion, the three catalytic inhibitors of ULK1 mentioned above correspond to the very early stages of drug discovery. Although they demonstrate potency in biochemical assays (IC50s from 3 to 100 nM), their poor selectivity limits their use as chemical tools in vitro. Chemical optimization is needed for further deciphering the role of ULK in cellular processes, including autophagy, and also in human pathophysiologies using this class of inhibitors.

Late inhibitors

Another way to tackle autophagy is to act on the later stages of the autophagy process, the degradation of autophagosome content by the lysosomes (Fig. 1). In the following section we will briefly present the different category of drugs that were described as late inhibitor of autophagy in vitro. A main focus will be done on the lysosomotropic agents and their applications in cancer (Figs. 2, 4).

Fig. 4.

Selected phase II and III clinical trials of chloroquine (CQ) and hydroxychloroquine (HCQ) in cancer treatment. These information’s were queried from ClinicalTrial.gov website (http://clinicaltrials.gov/)

Acid protease inhibitors

Lysosomal proteases are involved in the degradation of autophagolysosomes and were reviewed earlier [101]. Among the lysosomal hydrolases and proteases, cathepsins have a major role. Pepstatin A and E64d inhibit cathepsins D, E and B, H, L respectively [102]. These two drugs by inhibiting lysosomal proteases can effectively suppress autophagy [103] and are commonly used to monitor the autophagy flux [104].

Vacuolar-type H⁺-ATPase inhibitors

Vacuolar-type H⁺-ATPases (V-ATPases) are found within the membranes of many organelles including lysosomes, endosomes, and secretory vesicles. V-ATPases couple proton transport with ATP hydrolysis and maintain the acidic pH microenvironment of the vacuole [105]. Bafilomycin A1 blocks lysosomal proton transport and lead to the inhibition of lysosomal hydrolases, which are only activated at low pH. As a consequence, Bafilomycin A1 inhibits autophagy flux [105–107].

Other inhibitors of autophagic degradation

Clomipramine is a FDA-approved drug used for the treatment of psychiatric disorders. Its active metabolite, desmethylclomipramine have been shown to interfere with autophagic flux by blocking autophagosome-lysosome fusion [108]. Lucanthone, an anti-schistosome agent, inhibits autophagy by disrupting lysosomal membrane permeabilization leading to markedly increased cathepsin D cytosolic levels [109]. Importantly, lucanthone was found more effective at reducing the viability on a panel of 7 breast cancer cells than chloroquine.

Lysosomotropic agents

Chloroquine (CQ) and Hydroxychloroquine (HCQ)

CQ and HCQ function primarily by inhibiting lysosomal acidification. At neutral pH, both drugs remain uncharged and freely diffuse across the plasma membrane. In acidic environments, CQ and HCQ become protonated and trapped in lysosomes, causing an increase in the lysosomal pH and therefore inhibiting the activity of the degradative enzymes [110–112]. Consequently, cells treated with such agents are unable to undergo lysosomal digestion and exhibit vesicular organelle accumulation in the cytoplasm consistent with blocked autophagy.

For the past 60 years, CQ and HCQ have been prescribed for malaria [110–112] and autoimmune disorders such as rheumatoid arthritis [113] and lupus erythematosus [114]. These compounds were also evaluated for HIV infection [115]. CQ and HCQ are oral drugs well tolerated in human. The only major side effect of CQ is retinal toxicity [116]. HCQ was later developed in response to this problem. Structurally, HCQ differs from CQ by the addition of a single hydroxyl group to the beta carbon of a tertiary amino ethyl located at the terminus of the side chain attached to the quinolone base. This addition limits the ability of the drug to cross the blood-retinal barrier and decreases the potential for retinal toxicity [117].

Many preclinical studies demonstrated that effective autophagy inhibition can be achieved with CQ and HCQ, and that they can have therapeutic application in cancer [118–121]. The potency of CQ as an anticancer agent seems to be very different among tumor cell lines [123]. It is still not clear if the cytotoxic effects of these agents in cancer cells are specifically from autophagy inhibition and not from another effect on lysosomal. However, cancer cells exhibiting increased autophagy flux are sensitive to, and could die, in response to treatment with lysosomotropic agents [124].

Because these drugs are well tolerated, numerous clinical trials for cancer patients were undertaken with the primary aim to measure efficacy and safety. Most of the trials are investigating HCQ in various tumors in combination with standard of care treatments, such as allosteric mTOR inhibitors (Temsirolimus, Everolimus), DNA damaging agents (Cyclophosphanmide, Oxaliplatin, Capecitabine), Tyrosine kinase inhibitors (Erlotinib, Imatinib), proteasome inhibitors (Velcade), microtubule inhibitors (Docetaxel, Paclitaxel) and radiation (Fig. 4). In patients suffering from glioblastoma multiforme, a phase III clinical trial combining conventional treatment, radiation and carmustine, with chloroquine demonstrated a median survival of 24 months in the chloroquine arm compared to 11 months in the placebo arm. However these results were not statistically significant due to the limited number of patients enrolled and in addition autophagy was not examined in this study [124].

Few of the current clinical trials for cancer patients included endpoints on autophagy biomarkers. HCQ dose-dependent autophagy inhibition was observed using electron microscopy assay, as a pharmacodynamic readout, on blood mononuclear cells [125, 126]. However, it is unclear today whether safely tolerated doses of HCQ or CQ can effectively and consistently inhibit autophagy in human tumors.

Lys05

Even if HCQ displays a very large tissue distribution resulting in high intratumoral concentration, it is anticipated that CQ or HCQ might not be sufficiently potent to demonstrate clinical efficacy in cancer at tolerable doses. To increase the activity of CQ, derivatives are being developed. A bivalent aminoquinoline analog of HCQ, Lys05, was identified [127]. This compound is water soluble and accumulates more readily within the lysosome. Lys05 causes robust lysosomal pH elevation, blocks autophagy and induces cytotoxicity in cancer cell lines with a ten-fold greater potency than HCQ. When tested in vivo, this compound induced an increase of autophagic vesicles, using electron microscopy and LC3 biomarker, in human tumor xenografts after 2 days of dosing at 76 mg/kg (138 nmol/g) i.p, as compared to equimolar dose of HCQ (60 mg/kg) or PBS. Administrated as a single agent, Lys05 showed a significant antitumor activity on two different tumor models without causing toxicity at effective doses in mice [127]. Of note, antitumor activity was not observed with equimolar dose of HCQ. While no extensive in vivo PK data was provided it this report, the authors measured a better accumulation of Lys05 within tumors at effective dose (138 nmol/g, i.p.) as compared to equimolar dose of HCQ, using mass spectrometry. These preclinical data with Lys05 are promising and investigations on alternative route of administration, such as intravenous or oral, would be of great interest for a clinical perspective.

ARN5187

Starting from an in silico screen of lysosomotropic small molecule interacting with the circadian nuclear receptor REV-ERBβ, ARN5187 was recently identified [128]. This lysosomotropic compound blocked autophagolysosome final maturation and inhibited autophagy. Although ARN5187 displayed a similar potency to equimolar concentrations of CQ on inhibiting autophagy readout, it was more cytotoxic in cancer cell lines. This was recently explained by structure–activity relationship (SAR) of the chemical series of ARN5187 and also using knockdown of REV-ERBβ in CQ-treated cancer cell lines [129]. The improved cytotoxic effect of ARN5187 as compared to CQ, is due to its dual inhibitory activity on REV-ERB and autophagy. The group of B. Grimaldi finally selected a lead compound, the compound 30, with 5–50 superior in vitro cytotoxic activity on 6 tumor cell lines, as compared to equimolar concentrations of CQ. Testing this dual inhibitor in vivo for PK and efficacy studies will demonstrate the benefit of this novel pharmacological approach.

Conclusions

The housekeeping function of autophagy has attracted major attention over the last decade as it has a broad variety of roles in health and diseases. While the learnings in the autophagy machinery markedly improved, the researchers still had to deal with poorly selective tools to manipulate it. During the last 2 years, significant progress was made with the discovery of catalytic inhibitors of the two main kinases involved in the autophagy process, Vps34 and ULK1. We must keep in mind that these enzymes do not exclusively belong to autophagy pathway, and chemical optimization is needed in the case of ULK inhibitors. Nonetheless, we anticipate that these new compounds would advantageously replace the pan-PI3K inhibitors, such as 3-methyladenine, to interrogate the fundamental questions in the field of autophagy.

Whether inhibiting autophagy could have some benefits for patients is still an open question. Targeting this process could be of great interest for the treatment of cancer, as autophagy has been associated with tumor initiation, progression and resistance to treatment [37]. New information could come from the current clinical trials with chloroquine analogs, that affect lysosomal functions. However, many challenges remain to be addressed because autophagy is a dynamic process and a clear and suitable pharmacodynamic readout is not yet available for patients. The identification of the responder patient population and the best combination will also be critical to increase the probability of success of such therapy. Given the “addiction to autophagy” of tumor associated K-RAS mutation, such as pancreas cancer or non-small cell lung cancer [123, 130, 131], and the possibility of combining with mTOR inhibitors in renal cell carcinoma [82, 132], the development of new autophagy inhibitors is an attractive field of investigations.

Abbreviations

ADME

Absorption, distribution, metabolism, excretion

AMPK

AMP-activated protein kinase

ATG

Autophagy-related genes or proteins

Beclin-1

Coiled-coil, myosin-like BCL2-ineracting protein

CQ

Chloroquine

DNA-PK

DNA-dependent protein kinase

FYVE

Fab1, YOTB, Vac1 and EEA1

GFP

Green fluorescent protein

HCQ

Hydroxychloroquine

LC3

Microtubule-associated protein light chain 3

mTOR

Mammalian target of rapamycine

mTORC1

Mammalian target of rapamycine complex 1

MEF

Mouse embryonic fibroblasts

PK

Pharmacokinetic

PI3K

Phosphoinositide 3-kinase

PIK3C3

Phosphatidylinositol 3-kinase, catalytic subunit type 3

PtdIns3K

Phosphatidylinositol 3-kinase

PtdIns3P

Phosphatidylinositol 3-phosphate

ULK

Unc-51-like kinase

ULK1

Unc-51-like kinase 1

Vps34

Vacuolar protein sorting 34

3-MA

3-methyladenine

Compliance with ethical standards

Conflict of interest

BP is a Sanofi employee and also a shareholder of Sanofi stock.