Surviving protein quality control catastrophes – from cells to organisms

Cells are well equipped to survey and maintain the health of their proteomes; they employ chaperones that bind to non-native polypeptides to prevent aggregation and to facilitate the folding of proteins, as well as degradation systems – the ubiquitin-proteasome system (UPS) and autophagy – to degrade abnormal or damaged proteins. These diverse components of the protein quality control systems act in a concerted manner to prevent the accumulation of damaged or misfolded proteins and/or to promote their elimination. Because it is vital that cells avoid damage to proteins, cells not only keep an abundant supply of chaperones and protein degradation machineries but they have also evolved the ability to increase the abundance of the diverse components of the cellular defense systems against misfolded proteins when the need arises. These protein quality control systems have overlapping functions and are evolutionarily conserved, indicating that the evolutionary pressure to maintain protein homeostasis justified the cost of these elaborate pathways. The doubling of human life expectancy in the last 200 years (Finch, 2010) has resulted in a rise in the number of cases of age-related diseases. These diseases might have arisen as we reached the limits of our evolutionarily optimized protein quality control systems, because we have not had the time required to evolve effective defense mechanisms against age-related diseases. Therefore, identifying strategies to push the limits of protein quality control systems could reveal pathways to prevent age-related diseases. In this Commentary, we will give an overview of the cellular protein quality control systems, discuss how their failure might lead to disease and how we can exploit these natural cellular defense systems against misfolded proteins to survive experimental and pathological protein quality control catastrophes.

Life is harsh, and misfolded proteins represent a threat to proper cell and organism function. To survive this threat, organisms had to acquire resilience to misfolded proteins and have evolved an ability to fight harsh challenges (Fig. 1). An overview of the different cellular defense systems against misfolded proteins is presented below.

Studying the cellular responses to the harsh stresses that cause protein damage has revealed the mechanisms by which cells have acquired their resilience to such damages. Exposing cells to heat shock provides an ideal experimental system to identify protein chaperones because heat denatures existing proteins, leading to an overwhelming increase in the exposure of hydrophobic sequences. As a defense mechanism against the threat of protein aggregation, cells induce the expression of chaperones in order to neutralize the denatured proteins. Chaperones are a group of diverse proteins that interact with non-native intermediates to prevent their aggregation and facilitate their folding (Bukau et al., 2006; Kim et al., 2013). The expression levels of many chaperones upon heat shock increase to satisfy the increased demand that is generated by the heat-mediated denaturation of proteins (Lindquist and Craig, 1988; Morimoto, 2012), and many chaperones were therefore initially described as heat-shock proteins (HSPs). The master regulator of the heat-shock response is HSF1, a transcription factor that is normally in an inactive conformation and bound to molecular chaperones (Morimoto, 2012). When folding is compromised, HSF1 is activated and forms trimers that bind to the heat-shock element in the promoter region of heat-shock-responsive genes (Fig. 2), which includes those encoding chaperones, to activate their transcription (Morimoto, 2012).

Similar to the response in the cytosol, conditions that impair the folding of proteins in the endoplasmic reticulum (ER) create a stress (ER stress) to which cells adapt by mounting a response known as the unfolded protein response (UPR), aimed at restoring protein homeostasis in the ER (Wiseman et al., 2010) (see Fig. 2). Components of the mammalian UPR have been discovered by studying how cells react to ER stress, a condition elicited experimentally by treating cells with drugs that block ER function, such as tunicamycin and thapsigargin. Tunicamycin blocks N-linked glycosylation and thereby prevents the folding of a large fraction of ER proteins, whereas thapsigargin, a non-competitive inhibitor of the ER Ca²⁺ ATPases (ATP2A1, ATP2A2 and ATP2A3), perturbs ER function. The UPR did not evolve to protect cells from tunicamycin or thapsigargin. However, because these two drugs are potent inducers of the UPR, they have been instrumental in discovering the mammalian UPR. One component of the UPR, inositol-responsive enzyme 1 (IRE1; encoded by ERN1), is an ER-resident transmembrane protein that is conserved from yeast to mammals and senses stress in the ER to increase the transcription of genes encoding ER-resident chaperones (Mori, 2000). In addition to IRE1, metazoans have two additional ER-stress transducers – protein kinase RNA-like endoplasmic reticulum kinase (PERK; encoded by EIF2AK3) and activating transcription factor 6 (ATF6) (Mori, 2000). IRE1 and PERK sense protein misfolding in the ER lumen and convey this signal to their effector domain that is located on the other side of the ER membrane through oligomerization (Bertolotti et al., 2000). The three mammalian UPR sensors IRE1, ATF6 and PERK are kept inactive in unstressed cells in a complex with the ER chaperone binding immunoglobulin protein (BiP or GRP78; encoded by HSPA5), which dissociates upon stress, leading to activation of the sensors (Bertolotti et al., 2000). A popular interpretation given to these findings is the idea that misfolded proteins compete and sequester BiP away from IRE1 and PERK (Liu et al., 2000; Okamura et al., 2000). However, this interpretation disagrees with the abundance of the ER-stress signaling proteins relative to that of BiP (there might be 1000 more BiP than IRE1 and PERK molecules) and with the observed stability of the BiP–IRE1 and BiP–PERK complexes (Bertolotti et al., 2000). So how does it work? This long-standing question has taken fifteen years to answer and required the reconstitution of the BiP-mediated UPR-sensing mechanism in vitro with purified components (Carrara et al., 2015). It has been shown that BiP binds to the luminal domain of IRE1 and PERK through its ATPase domain, leaving the substrate-binding domain of BiP free to sample and bind to misfolded proteins (Carrara et al., 2015), in agreement with earlier findings (Todd-Corlett et al., 2007). The binding of unfolded proteins to BiP then triggers a conformational change that is transduced from the substrate-binding site of BiP to the ATPase domain, resulting in the dissociation of BiP, and activation of IRE1 and PERK (Carrara et al., 2015). Thus, UPR sensors have exploited the substrate-binding site of BiP, which has a high affinity for unfolded proteins, to survey and to respond to perturbations of folding in the ER. Like mammalian IRE1, the luminal domain of yeast IRE1 also binds to BiP, which dissociates upon stress (Okamura et al., 2000), but the importance of BiP dissociation in activating the yeast UPR has been debated (Kimata et al., 2004, 2007; Pincus et al., 2010). In addition, the luminal domain of yeast IRE1, unlike that of mammalian IRE1 (Oikawa et al., 2009), directly binds to unfolded proteins (Kimata et al., 2007; Gardner and Walter, 2011). Once activated, the three mammalian ER-stress transducers act in a coordinated manner to restore ER homeostasis, not only by increasing the expression of genes encoding chaperones – i.e. the ancestral branch of the UPR – but also by reducing global protein synthesis, which constitutes a rapid adaptive response to stress (Fig. 2). This is mediated by PERK, which, in response to the folding perturbation in the ER or simply when the folding supply does not match the demand, phosphorylates the α subunit of eukaryotic translation initiation factor 2 (eIF2α) on residue Ser51 to reduce translation initiation (Wiseman et al., 2010; Cao and Kaufman, 2012). This represents a first line of defense against the threat of protein misfolding. Reducing translation, in turn, favors protein folding by increasing the availability of chaperones (Tsaytler et al., 2011), probably because the chaperones that normally assist in the synthesis of new proteins become available when protein synthesis rates are decreased. Likewise, an immediate response to heat stress is also to reduce global protein synthesis (Richter et al., 2010); however, surprisingly, the molecular basis for the reduced protein synthesis in response to heat shock still remains unclear.

Phosphorylation of eIF2α and the resulting reduction in protein synthesis is a first line of defense that is essential for survival in response to many forms of stress. This signaling downstream of eIF2α is therefore referred to as the integrated stress response (ISR) (Harding et al., 2003). Viral infection and heme-deficiency signal to the eIF2α kinases PKR (EIF2AK2) and HRI (EIF2AK1), respectively, to result in the attenuation of protein synthesis (Ron and Harding, 2007). The shortage of amino acids is a stress that is sensed by the eIF2α kinase GCN2 (EIF2AK4), and activation of this kinase also results in the attenuation of protein synthesis (Sonenberg and Hinnebusch, 2009). When eIF2α is phosphorylated, global protein synthesis is attenuated, thereby sparing amino acids, which is a rapid and adaptive response conserved from yeast to mammals (Sonenberg and Hinnebusch, 2009). However, a few selective transcripts are preferentially translated under these conditions. One encodes the transcription factor ATF4, which controls the expression of genes that are involved in amino acid import and biosynthesis (Sonenberg and Hinnebusch, 2009). The fact that many forms of stress converge on the ISR, resulting in increased expression of genes that are involved in amino acid metabolism, indicates that perturbation of amino acid homeostasis might be a problem that is common to many forms of stresses (Fig. 2). This idea was highlighted with the finding that proteasome inhibition causes a shortage in amino acids, a stress to which cells react to by inducing the ISR and autophagy, in an attempt to restore amino acid homeostasis (Suraweera et al., 2012). Downstream of ATF4 is the transcription factor CHOP (encoded by DDIT3), which, in turn, induces transcription of PPP1R15A. PPP1R15A is one of two regulatory subunits of eIF2α phosphatases in mammals. PPP1R15A recruits one of the serine/threonine protein phosphatase 1 catalytic subunits (PP1c) in stressed cells to dephosphorylate eIF2α, thereby terminating stress signaling (Novoa et al., 2001).

Similar to the ER, mitochondria also adapt their supply of chaperones and proteases by mounting a mitochondrial UPR when needed (Pellegrino et al., 2013). The mitochondrial UPR comprises a mitochondria-to-nucleus signaling pathway that senses perturbation of homeostasis in the mitochondria, and the UPR responds by adjusting the expression of mitochondrial chaperones and proteases in the nucleus, a pathway that has been best characterized in Caenorhabditis elegans (Pellegrino et al., 2013).

In parallel to the heat-shock response and the UPR, cells also have the ability to increase the abundance of the two cellular degradation pathways – the proteasome system and autophagy – to match any arising need.

When the demand for protein degradation exceeds the proteolytic capacity, cells increase the expression levels of proteasome subunits in a concerted manner (Hanna and Finley, 2007). In yeast, this is controlled by Rpn4, a transcription factor that regulates the expression levels of proteasome subunits through a homeostatic negative-feedback loop (Xie and Varshavsky, 2001). Rpn4 is an unstable protein, which is normally rapidly degraded but accumulates when the proteasome is overwhelmed (Hanna and Finley, 2007). In mammals, the transcription factor erythroid-derived 2-related factor 1 (Nrf1; also known as NFE2L1) adjusts the expression of proteasome subunits to meet the cellular needs (Radhakrishnan et al., 2010).

Autophagy is also a tightly regulated process and was initially discovered as a response to starvation (Nakatogawa et al., 2009). When the supply of nutrients is sufficient, non-selective autophagy is repressed. This repression is under the control of a serine/threonine protein kinase, mammalian target of rapamycin (mTOR) (Zoncu et al., 2011). However, mTOR is much more than the molecular switch for autophagy. This kinase is in fact a central regulator of cellular metabolism and senses nutrients, growth factors and energy levels, and adjusts metabolic processes depending on the conditions. When the supply of nutrients is sufficient, mTOR promotes anabolic processes, including protein synthesis as well as ribosome biogenesis, while repressing autophagy. Under conditions that are unfavorable for cell growth – during stress or in the presence of the drug rapamycin – mTOR is inhibited (Zoncu et al., 2011). As a result, autophagy is induced and, concomitantly, translation and ribosome biogenesis are repressed.

When proteasome degradation is compromised, cells adapt by inducing autophagy, and cross-talk events between the two cellular degradation pathways have been identified (Korolchuk et al., 2009). Both pathways contribute not only to protein homeostasis but also to amino acid homeostasis. Under conditions of amino acid starvation, both autophagy and proteasomal degradation are required to maintain adequate amino acid levels in order to sustain protein synthesis (Onodera and Ohsumi, 2005; Vabulas and Hartl, 2005). In the absence of starvation, under normal conditions, proteasomal degradation contributes an important fraction of the intracellular pool of amino acids (Suraweera et al., 2012). Proteasome inhibition results in a lethal amino acid shortage, and this is the signal that leads to autophagy induction as an adaptive response that aims to restore homeostatic levels of amino acids (Suraweera et al., 2012).

Many components of the cellular defense systems have been identified by studying how cells respond to stress. This has conveyed, erroneously, the notion that the cellular defense systems against misfolded proteins are only required under stress. But this is not the case. The ensemble of sophisticated pathways (Fig. 2) that ensure resilience to protein damage are evolutionarily conserved and are needed at all times to maintain protein homeostasis. This is because the damages that occur during harsh stress also exist under non-stress conditions, albeit to a lower magnitude. As discussed above, heat stress denatures proteins and leads to an increased abundance of unfolded proteins; this condition also normally occurs during the synthesis of proteins when nascent chains are unfolded. Cells have acquired their resilience and gained their strength by surmounting difficulties (harsh stresses). As a result, cells normally thrive at maintaining protein homeostasis when milder stresses are continually encountered.

According to this model, cell fitness is not defined by an absence of errors, problems or failure but by an ability to cope and deal with errors, problems or failure. Consequently, cell fitness is lost when cells can no longer cope with failures. This is an important problem, at the origin of a broad range of diseases, which will be discussed below.

A broad range of age-related human diseases, including common and devastating neurodegenerative diseases, are caused by the deposition of misfolded proteins. These diseases include Alzheimer's disease, Parkinson's disease, Huntington's disease, amyotrophic lateral sclerosis (ALS) and prion diseases. Although clinically diverse, these diseases share a common pathological hallmark – they are caused by the progressive dysfunction and death of specific nerve cells in selective regions of the brain or the peripheral nervous system, which are the result of the accumulation of specific proteins of aberrant conformations (Soto, 2003). The major component of each disease-characteristic deposit has been identified, in most cases over two decades ago, and this has revealed that there are no common features amongst the different disease-causing proteins; they have different primary sequences and their native folds are also distinct (Chiti and Dobson, 2009). Some are tightly folded, such as superoxide dismutase 1 (SOD1) – the faulty proteins in some familial forms of ALS (Valentine et al., 2005) – whereas others are natively unstructured, such as α-synuclein – the major component of Lewy bodies in Parkinson's disease (Goedert et al., 2013). However, although extremely diverse, the disease-causing proteins share one common feature – they are usually soluble but have a propensity to misfold and aggregate.

It is now clear that neurodegenerative diseases are caused by the gain of toxicity that is associated with the misfolding propensity of a disease-causing protein; however, the pathogenic cascades that lead from the misfolding of a protein to neurodegeneration remain largely unknown, despite years of extensive research efforts. This represents a complex organismal problem, which, in fact, results from a cell biology problem that arises when cells become unable to withstand the pressure of misfolded proteins. Thus, the question becomes why proteins, which are normally soluble, eventually misfold and aggregate late in life. As discussed above, cells normally strive to ensure that proteins are correctly folded and have evolved powerful and sophisticated mechanisms to maintain protein homeostasis (proteostasis) (Balch et al., 2008). Protein quality control systems are normally very efficient at maintaining protein homeostasis over several decades of life. However, the fact that protein aggregates build up later in life suggests that the cellular defense systems against misfolded proteins gradually fail with age, resulting in the accumulation of misfolded proteins with devastating consequences for cells and organisms (Morimoto and Cuervo, 2014; Vilchez et al., 2014). If cells and organisms are able to cope with potentially harmful proteins for decades, perhaps identifying strategies that boost the cellular defense systems against misfolded proteins could have some value for the development of therapeutics against the diverse misfolding diseases.

In the past 20–30 years, many components of protein quality control systems have been identified, and we have learned, often in great detail, how protein quality control operates in cells. The challenge that remains is to use this knowledge to identify strategies to correct the broad range of diseases that arise when protein quality control is overwhelmed. Attempts to manipulate protein quality control pathways with small molecules have been made on several levels. For instance, the search for inducers of HSF1 is an actively pursued line of research because HSF1 is a master regulator of the expression of chaperones – key components of the cellular defense systems against misfolded proteins. Activating HSF1 mimics stress by inducing a heat-shock response, thereby increasing chaperone expression. Because chaperones neutralize misfolded proteins, induction of HSF1 could be an approach to combat protein misfolding diseases. Geldanamycin is an inhibitor of HSP90 proteins and induces the heat-shock response by activating HSF1 (Neckers and Workman, 2012). More recently, other inducers of HSF1 have been described (Calamini et al., 2011).

Furthermore, given that kinases are popular drug targets, the UPR kinases IRE1 and PERK have also been targeted by small molecule inhibitors. Because stress responses are essential for cell survival, inhibitors of stress signaling pathways are predicted to be deleterious, a property that could be advantageous for the discovery of cancer drugs. This possible effect on cell survival has motivated the development of inhibitors of PERK (Axten et al., 2012) and IRE1 (Cross et al., 2012; Ghosh et al., 2014), and of the integrated stress-response inhibitor ISRIB, which inhibits the signaling downstream of eIf2α phosphorylation (Sidrauski et al., 2013). Recent and comprehensive reviews on the pharmacological manipulation of stress responses are available elsewhere (Hetz et al., 2013; Maly and Papa, 2014). The next section will only focus on strategies to rescue cells from protein quality control catastrophes.

Despite our good understanding of many protein quality control pathways, one of the problems impeding the development of targeted therapies lies in the fact that it is currently impossible to predict which manipulations of the protein quality control pathways will be tolerable, detrimental or beneficial. For example, PERK inhibitors as well as ISRIB were developed to kill cancerous and stressed cells, respectively (Atkins et al., 2013; Sidrauski et al., 2013). In addition to their predicted cytotoxic properties, PERK inhibitors and ISRIB have been found to protect mice from prion diseases (Halliday et al., 2015; Moreno et al., 2013). These results could not be anticipated from 15 years of intensive explorations of the pathway. Thus, pharmacological manipulations of stress-response pathways bring surprises.

Because it is impossible to predict the outcomes of pharmacological intervention of stress-response pathways, we have focused on identifying unbiased approaches that can be used to rescue cells from protein quality control catastrophes with the view that if we manage to ‘cure’ cells from such profound defects, this might also be feasible in an organism (Fig. 3). Such strategies are appealing because they hold the promise to ameliorate the broad range of diseases that arise when protein quality control systems are overwhelmed.

In searching for an experimental system in which to identify such strategies, we have chosen to mimic experimentally the protein quality control failures that might arise during aging or in diseases by blocking key components of protein quality control systems (Fig. 3). Such perturbations are brutal and usually fatal to cells, and thereby provide robust and unbiased experimental conditions that can be used to search for strategies to help cells survive under such catastrophic circumstances. Recently, we have identified a number of approaches that have not only revealed interesting pathways that might have therapeutic benefits but that have also helped to identify new components of protein quality control systems.

The proteasome is essential for the degradation of many cellular proteins. Consequently, dysfunction of the ubiquitin-proteasome system (UPS) is associated with a broad range of diseases, including cancer and neurodegeneration (Tanaka and Matsuda, 2014). We use inhibition of the proteasome as an experimental model of global perturbation of protein quality control that could be relevant to human diseases. Aiming to identify strategies to help cells survive proteasome degradation failure, we found that the deleterious consequences of proteasome inhibition can be rescued through amino acid supplementation in yeast, mammalian cells and Drosophila (Suraweera et al., 2012). This demonstrates that cells can tolerate large amounts of undesired proteins but not the amino acid scarcity that results from proteasome inhibition (Suraweera et al., 2012). Thus, proteins that accumulate in cells upon proteasome inhibition appear to be deleterious largely because they sequester a pool of amino acids that would normally be recycled (Suraweera et al., 2012). These findings reveal that the proteasome is not just a waste disposer but is actually vital as an amino-acid-recycling machine (Fig. 2). Encouraged by these findings, which reveal a previously unappreciated aspect of proteasome degradation, we continued to search for strategies that rescue cells when the proteasome fails. Using an unbiased screen in yeast, we asked: can we identify genes that, when overexpressed, rescue the viability of the yeast cells that are defective for proteasome degradation? The screen was very stringent and led to the identification of (only) three potent suppressors of the proteasome defects of a Rpt6 thermosensitive mutant, two of which had been identified previously (Rpt6 and Rpn14) and one encoding a previously uncharacterized protein that we named Adc17 (encoded by TMA17). Because this newly identified suppressor was very potent, we could elucidate its function and found that Adc17 is a proteasome-assembly chaperone whose expression is increased upon proteasome stress (Hanssum et al., 2014).

As mentioned above, when the demand for protein degradation exceeds the proteolytic capacity, cells increase the expression levels of proteasome subunits in a concerted manner (Hanna and Finley, 2007). Although increasing the levels of proteasome subunits is necessary to increase proteasome abundance, this is not sufficient because the levels of functional proteasomes depend not only on the expression of proteasome subunits but also on their precise and correct assembly. The proteasome comprises 33 subunits, and its assembly is an extremely complex and challenging process in the crowded cellular environment (Beckwith et al., 2013; Murata et al., 2009; Tomko and Hochstrasser, 2013), but it was unknown how stressed cells overcome the challenging task of assembling proteasomes. We found that Adc17 promotes an early and rate-limiting step in the proteasome assembly process, and that its expression is increased upon proteasome stress to help cells assemble more proteasomes when the need arises. As a result, Adc17 is vital for cells so that they can cope with increased demands for proteasome-mediated degradation (Hanssum et al., 2014). However, how metazoans cope with overwhelming demands for proteasome assembly is currently unclear.

The identification of Adc17 through an unbiased suppressor screen has revealed a new component of the protein quality control system and provided new insights into the mechanisms by which cells maintain proteasome homeostasis. Although this discovery is still in its infancy, identifying such strategies that can boost proteasomal degradation might ultimately be of therapeutic value in order to ameliorate the large number of pathologies associated with proteasome deficiencies. Using the same concept, pharmacological inhibition of USP14 has been proposed as an approach to boost proteasome degradation and reduce the accumulation of unrelated aggregation-prone proteins in cells (Lee et al., 2010). It will be interesting to see whether inhibitors of USP14 can prevent the accumulation of aggregation-prone proteins in mouse models of neurodegenerative diseases.

Failure to maintain protein homeostasis in the ER is associated with a broad range of diseases (Hetz and Mollereau, 2014). Therefore, identifying strategies to help cells survive the deleterious treatment of tunicamycin could help to identify therapeutic pathways. We discovered that the small molecule guanabenz rescues HeLa cells from the otherwise lethal accumulation of misfolded proteins in the ER upon treatment with tunicamycin (Tsaytler et al., 2011). Deciphering the mechanisms underlying the cytoprotective effects of guanabenz shed light on an interesting aspect of cell biology. Guanabenz protects cells by prolonging translation attenuation, the first line of defense against the accumulation of misfolded proteins (Tsaytler et al., 2011). This occurs because guanabenz selectively binds to and inhibits the regulatory subunit of the stress-induced eIF2α phosphatase that comprises PPP1R15A and PP1c, whereas guanabenz does not interact with nor inhibit the constitutive eIF2α phosphatase PPP1R15B–PP1c, thereby avoiding persistent eIF2α phosphorylation (Tsaytler et al., 2011). This is very important because inhibition of the two eIF2α phosphatases leads to persistent eIF2α phosphorylation and results in persistent inhibition of protein synthesis, which is lethal (Harding et al., 2009). As a result, guanabenz increases the availability of chaperones to misfolded proteins and, consequently, rescues tunicamycin-stressed HeLa cells from collapse of proteostasis (Tsaytler and Bertolotti, 2012).

The approach of fine-tuning translation to overcome protein misfolding defects by inhibiting PPP1R15A is very attractive as it is straightforward, potent and selective, as well as potentially applicable to a broad range of conditions that are caused by the accumulation of misfolded proteins. Indeed, the misfolding of proteins in the ER is associated with many human diseases (Kim and Arvan, 1998). However, although guanabenz is a selective inhibitor of PPP1R15A in non-neuronal cells, this selectivity does not apply in an organismal context because guanabenz is a centrally active hypotensive drug with a nanomolar affinity for the α2-adrenergic receptor (Holmes et al., 1983). Guanabenz, initially, was marketed as a drug for the treatment of hypertension, but its use was associated with side effects that are a direct consequence of its activity as a α2-adrenergic agonist, such as hypotension, respiratory depression, bradycardia, drowsiness, lethargy and even coma (Hall, 1985).

Although guanabenz cannot be used to selectively inhibit PPP1R15A in vivo, we found that its adrenergic activity is a separate function from that of the inhibition of PPP1R15A (Tsaytler et al., 2011), indicating that it might be possible to generate guanabenz derivatives in which the adrenergic activity is ablated. To that end, we searched for selective PPP1R15A inhibitors and synthetized a range of guanabenz derivatives, and identified Sephin1 (a selective inhibitor of a holophosphatase), which, like guanabenz, rescues HeLa cells from cytotoxic tunicamycin treatment (Das et al., 2015). Sephin1 selectively binds to and inhibits PPP1R15A to safely prolong the benefit of translation attenuation following stress. Importantly, translation attenuation is prolonged only transiently because Sephin1 selectively inhibits the stress-induced eIF2α phosphatase PPP1R15A–PP1c, but not the constitutive PPP1R15B–PP1c enzyme, thereby avoiding the deleterious effects that result from inhibiting both eIF2α phosphatases (Harding et al., 2009). The cytoprotective activity of Sephin1 on stressed cells is abolished in cells that lack PPP1R15A, further confirming that the beneficial effect of Sephin1 on stressed cells is mediated entirely through inhibition of PPP1R15A (Das et al., 2015). Although it is closely related to guanabenz, Sephin1 has no α2-adrenergic activity. Like guanabenz, Sephin1 concentrates in the brain and nervous system, but does not cause any of the side effects that are associated with guanabenz because it is devoid of α2-adrenergic activity (Das et al., 2015).

Guanabenz also has another activity – it reduces the levels of prions in yeast and in mammals (Tribouillard-Tanvier et al., 2008a). Sephin1 is inactive in yeast; indeed, yeast lack both PPP1R15A and α2-adrenergic receptors, so the activity of guanabenz in yeast is not mediated by α2-adrenergic receptors or PPP1R15A, and different mechanisms for the activity in yeast have been proposed (Tribouillard-Tanvier et al., 2008b).

Having established that Sephin1 is a specific and selective PPP1R15A inhibitor that could be suitable for in vivo studies, we have tested whether Sephin1 can prevent protein-misfolding diseases in mice. For these proof-of-principle studies, we selected two protein-misfolding diseases for which there is robust genetic evidence that abnormal signaling through PPP1R15A is involved in the disease mechanism – Charcot-Marie-Tooth 1B (CMT-1B) (D'Antonio et al., 2013; Pennuto et al., 2008) and ALS that is due to mutation in the superoxide dismutase SOD1 (Wang et al., 2014). We found that oral administration of Sephin1 almost completely prevents the motor, histological and molecular defects of these two otherwise unrelated protein-misfolding diseases in mice (Das et al., 2015).

In humans as well as in mice, mutations in myelin protein zero (MPZ^mutant), a transmembrane protein that is produced by Schwann cells in the peripheral nervous system, causes the misfolding and ER-retention of the affected protein (Wrabetz et al., 2006). This results in the demyelinating neuropathy CMT-1B. At the molecular and cellular level, the pathology at the origin of CMT-1B is analogous to the perturbation created experimentally by treating cells with tunicamycin – a massive accumulation of misfolded proteins in the ER. CMT-1B is an extremely rare disease, but it is clear that it is a disorder resulting from malfunction of the ER and is therefore a useful model.

In the case of SOD1-ALS, the link between mutant SOD1 and ER stress is indirect, yet robust. SOD1 is a cytosolic protein, and mutant SOD1 proteins have been found to bind to Derlin-1 on the cytosolic side of the ER membrane, blocking the degradation of ER proteins, which then accumulate in the ER, thereby causing ER stress (Nishitoh et al., 2008). Supporting the notion that mutant SOD1 causes ER stress, deletion of one allele of PERK accelerates the disease progression in an ALS-SOD1 mouse model, whereas non-transgenic mice lacking one allele of PERK are normal (Wang et al., 2011). These findings reveal a strong link between the ISR and SOD1-ALS, and are consistent with the established function of PERK in helping cells to survive protein misfolding stress. Moreover, genetic inactivation of PPP1R15A markedly ameliorates the disease (Wang et al., 2014), and we have recapitulated these finding in a different ALS-SOD1 model through selective pharmacological inhibition of PPP1R15A (Das et al., 2015).

Translation begins with the recognition of the initiation codon AUG by the ternary complex comprising the GTP-bound form of eukaryotic initiation factor 2 (eIF2; comprising three subunits, α, β and γ), the initiator methionyl-tRNA (Met-tRNAi) and the small (40S) ribosomal subunit to form the 43S pre-initiation complex (PIC) (Hinnebusch and Lorsch, 2012). GTP in the ternary complex is hydrolyzed and GDP must be exchanged for GTP before another round of translation initiation can occur. Phosphorylation of eIF2α renders the interaction between the guanine nucleotide exchange factor eIF2B and eIF2 unproductive (Hinnebusch and Lorsch, 2012). Therefore a minute increase in eIF2α phosphorylation has a profound impact on translation rates (Kaufman, 1999).

Reducing translation rates by prolonging eIF2α phosphorylation increases chaperone availability in the ER (Tsaytler et al., 2011), and this in turn is likely to increase protein folding both in the ER and in the cytosol. Indeed, we found that Sephin1 completely prevented SOD1 aggregation and slightly increased the levels of the soluble protein. SOD1 mutants are known to have folding defects (Bruns and Kopito, 2007), and slowing down translation might improve their folding and prevent their aggregation (Das et al., 2015).

Importantly, use of Sephin1 is not only beneficial but also safe in mice (Das et al., 2015), which could be anticipated for two reasons. First, PPP1R15A-knockout mice are largely normal (Kojima et al., 2003), therefore selective inhibition of PPP1R15A function is not expected to have adverse effects. Second, if Sephin1 only inhibits PPP1R15A, then it is predicted to be safe, as we have found (Das et al., 2015).

In addition to demonstrating that selective inhibition of PPP1R15A can safely prevent diverse protein misfolding diseases in mice, our work also provides the proof-of-principle that regulatory subunits of phosphatases can be selectively and safely inhibited in vivo.

Moving from cells in culture to mice took time and effort. We now need to apply the same rigor to exploit this knowledge for the development of safe therapeutics. Although there is still much to do, our work provides proof-of-principle that approaches aimed at rescuing cells from protein quality control failure are relevant to diseases in mammals. The physiological relevance of studying the cellular responses of fibroblasts or HeLa cells to tunicamycin might have been unclear for many years, but our recent work underlines the importance of studying fundamental cellular processes for the discovery of effective and safe therapeutic approaches.

The recent investigations discussed here that have aimed to identify strategies to help cells survive when different components of protein quality control systems are faulty have newly revealed aspects and components of fundamental cell biology processes. It is fascinating to realize that our knowledge of cell biology is still only partial and that some important components of vital cellular pathways still remain to be discovered. This research area is not only interesting in its own right but also because we can reasonably anticipate that, ultimately, our knowledge will serve to develop rational therapeutics against a devastating group of diseases that affect an increasing number of individuals in aging populations.

We are grateful to members of the Bertolotti laboratory for discussions, in particular Indrajit Das, Anna Sigurdardottir and Marta Carrara for comments on this manuscript, as well as Lea Sistonen and Rick Morimoto for discussion on the heat-shock response.