During UPR, transmembrane proteins in the ER act as sensors to detect disturbances in protein handling (proteostasis) caused by misfolded/mutant proteins, or by alterations in signaling networks or in the ionic balance such as calcium in the cell. Three major signaling pathways characterize the UPR; the inositol-requiring enzyme 1α (IRE1α), the protein kinase RNA-like endoplasmic reticulum kinase (PERK) and the activating transcription factor 6 (ATF6; Figure 1). These proteins have distinct but overlapping functions in UPR and are structurally composed of specific subdomains. IRE1α has a cytosolic kinase domain that and phosphorylated IRE1α associates with molecules such the c-Jun N-aminoterminal kinase (JNK) involved in cell death regulation (Hetz et al., 2011; Walter and Ron, 2011). IRE1α has also an endoribonuclease activity that cleaves out a 26 bp intron from the pre-mRNA of X-box binding protein-1 (XBP-1) producing an active transcription factor that upregulates expression of genes involved in protein folding (like the ER chaperon GRP78/BiP), in ER-associated protein degradation (ERAD) pathway, and in lipid synthesis. The increased protein folding capacity and degradation of misfolded proteins induced by XBP-1 contribute to restored protein homeostasis and increased cell survival following stress. XBP-1 has also other functions in the cell that are now being explored in more details (see below). In addition, activated IRE1α can cleave certain RNA molecules in a process called IRE1α-dependent decay of mRNAs (RIDD) though the functional significance of this is not fully understood.

Recent studies on the crystal structure of yeast IRE1α have depicted a likely mechanism by which unfolded proteins and peptides can directly bind to and interact with the luminal domain of the molecule. IRE1α then self-associates to produce oligomers that are observed as large foci in the ER using microscopy (Gardner and Walter, 2011). The binding of unfolded peptides to IRE1 has so far been shown mainly using the yeast protein but the mammalian IRE1 shows a rather homologous structure, suggesting similar mode of regulation. This mode of IRE1α regulation by unfolded peptides is at variance with previous thoughts indicating that the binding of BiP is crucial in keeping IRE1α signaling at bay in the ER. However, it is likely that BiP plays a role in fine-tuning of the UPR and overexpression of BiP in different cells reduces ER stress and increases cell survival.

Protein kinase RNA-like endoplasmic reticulum kinase (PERK) is a protein kinase that phosphorylates downstream targets such as the eukaryotic translation initiation factor-2α (eIF2α). This leads to a reduction in general protein synthesis with a decrease in the load of misfolded/mutant proteins in the ER. However, the translational block is not complete and a subset of mRNAs including that encoding the transcription factor ATF4 are still being translated leading to increased expression of among others the growth arrest and DNA damage-inducible 34 (GADD34), the transcription factor C/EBP homologous protein (CHOP) as well as pro-apoptotic BCL-2 family proteins. GADD34, also called PPP1R15A, is a phosphatase for p-IF2α with the ability to restore protein synthesis. The degree of eIF2α phosphorylation is dynamically regulated in the cell and this can be interfered with using specific compounds to control ER stress in various diseases (see below).

The timing of induction of the different UPR pathways and their mutual interactions are important issues and ultimately determine the cell response to stress. Thus, an early and transient increase in IRE1α mediates cell protection, whilst an early-onset increase in PERK-ATF4 causes cell degeneration (Walter et al., 2015). An example of contrasting effects of UPR-controlled signals is also shown by the death receptor 5 (DR5) that is increased by the PERK-CHOP pathway but whose mRNA is cleaved by the endonuclease of IRE1α (Lu et al., 2014). DR5 is important for tumor cell killing and UPR signaling may accordingly influence cancer growth.

In contrast to IRE1α and PERK, ATF6 is first being transported from the ER to the Golgi compartment where it then becomes activated by intramembranous proteolysis mediated by site-1 and site-2 proteases causing the release of the transcriptionally active form of ATF6 (Walter and Ron, 2011). This mode of regulation resembles that of the sterol regulatory element-binding proteins (SREBPs) that are transcription factors for genes involved in lipid and cholesterol metabolism. However, like IRE1α, ATF6 upregulates gene expression of ER chaperons, ERAD components, as well as of XBP-1 (Tabas and Ron, 2011; Sano and Reed, 2013). There are two closely related ATF6 genes, ATF6α and ATF6β and double gene deleted mice are embryonically lethal. ATF6 plays a role in development as studied in the medaka fish (Oryzias latipes). ER stress is part of the normal development of the brain and of notochord in medaka, caused by the preponderate synthesis of extracellular matrix (EM) proteins required for organogenesis (Ishikawa et al., 2011). The lack of ATP6 decreased the levels of BiP and other chaperones reducing the folding and transport of EM proteins such as pro-collagen that interferes with early development (Ishikawa et al., 2013). Several proteins are involved in pro-collagen trafficking including the large protein particle (TRAPP) multi-subunit protein complex, and involves large vesicles different from ordinary coat protein complex II (COPII) ones, mediating ER-Golgi transport (Venditti et al., 2012).

Members of the astrocyte specifically induced substance (OASIS) protein family are transcription factors that are latent in the ER and activated in a process resembling that of ATF6 (Kondo et al., 2012). These proteins are expressed in a time and cell-specific manner important for their functions. The box B-binding factor-2 human homolog on chromosome 7 (BBF2H7) is present in developing chondrocytes and is activated by ER stress (Izumi et al., 2012). In the nucleus BBF2H7 increases gene expression of proteins involved in the secretion of EM proteins such as collagen. It is considered that BBF2H and ATF6 may act in parallel to enhance cell survival and ensure proper synthesis of EM components during development and after injury.

The nuclear factor erythroid-derived 2-related factor 1 (NRF1) is a transcription factor residing latent in the ER membrane involved in the regulation of expression of antioxidants and ERAD components. NRF1 shares many of its activities with the related factor NRF2, but NRF1 can specifically increase proteins associated with the proteasomes to promote protein degradation. NRF1 can be activated by different factors and events including ER stress (Zhang and Xiang, 2016). Recent studies showed that NRF1 is induced downstream of the target of rapamycin complex-1 (mTORC1) protein that is usually associated with the stimulation of protein synthesis and anabolic processes in the cell. These results indicate that mTORC1 signaling through NRF1 also increases proteasomes enhancing protein turnover, which is equally important for protein homeostasis. The physiological relevance of NRF1 and NRF2 in different cells and human diseases warrants further studies.

Wolframin (WFS1) is a transmembrane ER resident protein encoded by the WFS1 gene that in a mutant form causes Wolfram syndrome 1 (WS1), an autosomal recessive degenerative disease characterized by insulin-dependent diabetes mellitus, optic atrophy, deafness and various neurological symptoms (Hofmann et al., 2003). WFS1 is a component of the UPR and participate in the IRE1 and PERK signaling to maintain ER homeostasis (Fonseca et al., 2005). The promoter region of WFS1 gene has ER stress response element (ERSE) like sequence and WFS1 expression is increased during ER stress by XBP1 and ATF6 (Kakiuchi et al., 2006; Odisho et al., 2015) WFS1 stabilizes the yeast E3 ubiquitin ligase HRD1 (SYVN1 in humans) involved in ERAD and enhances the proteasome-mediated degradation of ATF6α thereby reducing ER stress signaling (Fonseca et al., 2010). Lack of functional WFS1 protein makes cells more vulnerable to different environmental insults (Kõks et al., 2013), and pancreatic β-cells degenerate in WFS1 gene deleted mice subsequent to ER-stress (Ishihara et al., 2004).

UPR has been shown to play a dual role in neurodegeneration as the lack of XBP1 increased survival in a ALS mouse model (Hetz et al., 2009). This was unexpected in view of the protective functions associated with IRE1α /XBP1 signaling. Further data showed that XBP1 deficiency increased autophagy through the forkhead box 01 (FOXO1) transcription factor improving the degradation of mutant SOD1 protein (Hetz et al., 2009). In dopaminergic neurons degenerating in PD, XBP1 deletion produce different outcomes depending on the age of the animals and reflecting compensatory changes in UPR-regulated factors. Thus, the lack of XBP1 during development induced a mild ER stress with protection of dopamine neurons against neurotoxicity of 6-hydroxydopamine, whilst silencing XBP1 in adulthood led to chronic ER stress and the loss of dopamine neurons (Valdés et al., 2014). In transgenic mice expressing mutant alfasunyclein as a model for PD, the administration of salubrinal inhibiting the PPP1R15A/GADD34 phosphatase inducing phospho-eIF2α levels was found to be neuroprotective suggesting an involvement of the PERK pathway in dopamine neuron degeneration (Colla et al., 2012).

In an acute neuronal injury, local delivery of XBP1 was beneficial increasing cell survival and the locomotion of animals after a spinal cord lesion (Valenzuela et al., 2012). These findings indicate that manipulation of IRE1α /XBP1 signaling is a possible target for development of therapeutics in different neurological diseases. XBP1 can influence also other functions in neurons than ER stress including learning and memory formation (Martínez et al., 2016). The effect of XBP1 was associated with changes in the expression of brain-derived neurotrophic factor (BDNF) that influences neuronal plasticity and long-term potentiation. Learning and memory are also regulated by the phosphorylation state of eIF2α influenced by PERK as well as other kinases in the neurons (Placzek et al., 2016). The role of UPR in memory-related processed in brain remain an interesting avenue to explore.

In ALS, spinal cord motor neurons and corticospinal motor neurons (CSMNs) in brain cortex undergo cell degeneration by mechanisms only partly understood. Transgenic mice expressing fluorescent labeled ubiquitin C-terminal hydrolase-L1 (UCHL1) in the CSMNs showed that these neurons are particularly sensitive to ER stress and to the lack of UCHL1 the ALS-associated gene, Alsin (Jara et al., 2015; Gautam et al., 2016). Cell stress and autophagy contributed to cell degeneration in the CSMNs pointing to novel targets for intervention. Recent data have shown that ALS and frontotemporal dementia share a major genetic risk factor related to the presence of a hexanucleotide repeat expansion in the gene, C9orf72 (Renton et al., 2011). Studies of CSMNs may give novels insights into the molecular mechanisms underlying the neuronal dysfunctions in these diseases.

One target for intervention in ALS may be PPP1R15A/GADD34 increasing the levels of phospho-eIF2α. Along this line it was shown that the compound guanabenz was able to prolong eIF2α phosphorylation and improve proteostasis in stressed human cells (Tsaytler et al., 2011). However, guanabenz as an adrenergic agonist causes also hypotension and therefore cannot be used in humans. Subsequent screening led to the identification of another molecule, Sephin1 (selective inhibitor of a holophosphatase) that can inhibit the stress-induced phosphatase pathway and restore protein homeostasis (Das et al., 2015). In vivo sephin1 was neuroprotective in the mutant SOD1-ALS mouse model, and improved behavior with no obvious adverse effects (Das et al., 2015). Sephin1 also prevented demyelination in Charcot-Marie-Tooth disease model caused by mutant myelin protein zero. Future studies will show whether sephin1 or similar compounds may potentially be useful for treatment of ALS and other protein misfolding disorders in humans.

Studies in models of HD have shown that the lack of XBP1 is neuroprotective by reducing the aggregation of mutant Huntingtin protein (Htt) and increasing cell survival (Vidal et al., 2012). Like in ALS, XBP1 deficiency increased autophagy and upregulated FOXO1 expression in HD (Vidal et al., 2012). As shown recently, the ubiquitin-specific protease 14 (Usp14) plays a role in degradation of mutant Htt by binding IRE1α and by regulating the proteasome activity, adding a further degree of complexity to HD pathophysiology (Hyrskyluoto et al., 2014). The aggregation of mutant Htt may takes place in different neuronal compartments, such as in the nucleus and the cytosol. Recently it was shown that Htt-aggregates are also present at the nuclear membrane interfering with the normal nucleocytoplasmic transport of proteins and RNA (Woerner et al., 2016). Interestingly, expression of TAR DNA binding protein-43 (TDP-43) that is linked to pathophysiology of ALS also affected this transport (Woerner et al., 2016). The significance of nucleocytoplasmic alterations in HD and in other protein misfolding diseases remains to be studied further.

AD is associated with accumulation of extracellular plaques consisting of amyloid-β (Aβ) peptides and with intracellular neurofibrillary tangles of hyperphosphorylated tau. The mechanisms for cell toxicity in AD are complex and related to changes in synaptic transmission, an altered calcium homeostasis, increased ER stress and a chronic state of neuroinflammation (Mattsson et al., 2016). Studies made in cell culture and in animal models of AD have shown that the IRE1/XBP1 branch of UPR influences amyloid-β toxicity and is either protective as in Drosophila or deleterious as in C. elegans (Cornejo and Hetz, 2013). Moreover, activation of JNK3 by UPR was found to increase production of amyloid-β that aggravated ER stress (Yoon et al., 2012). Inhibition of PERK reducing p-eIF2α levels reversed memory impairments in an AD mouse model (Ma et al., 2013). Together these studies indicate the involvement of specific ER signaling in AD that affects specific neurons and neuronal circuits in various brain regions in a time-dependent manner.

In AD, great efforts have been devoted to study the occurrence and toxicity of different forms of Aβ peptides (Tipping et al., 2015). Amyloid plaques contain also non-protein factors including glycosaminoglycan (GAG) polysaccharides that can increase the rate of Aβ polymerization (Stewart et al., 2016). Solid-state nuclear magnetic resonance spectroscopy has helped to identify the interaction site of Aβ with the polysaccharide heparin as a proxy for GAG (Madine et al., 2012). This information is useful for further structural studies and for design of compounds that may interfere with Aβ fibril formation. Notably, it has been shown that the structure of Aβ fibril in AD brains differ from that in vitro and may even be patient-specific for different clinical variants of AD (Lu et al., 2013). More studies into β-amyloid fibrils are therefore warranted both regarding the neuropathology and the possibility for development of novel treatments in AD.

Prion diseases are rare neurodegenerative disorders in humans (such as Creutzfeldt-Jakob disease) and in animals (Bovine Spongiform Encephalopathy) caused by the infectious prion protein (PrP; Colby and Prusiner, 2011). Mutant and misfolded PrP aggregates in brain tissue as amyloid fibrils leading to spongiform changes and loss of brain cells over time. Mounting evidence has shown that the spreading of the disease (propagation) occurs from cell to cell during PrP replication by misfolded PrP causing an abnormal folding of the cellular PrP (Colby and Prusiner, 2011). The self-templating ability of PrP is important for cell toxicity causing an exponential increase in amyloid fibril formation that ultimately causes cell stress and death. In PrP specific prion-like domains are present that influence protein aggregation. Similar domains are also found in other proteins such as in mutant TDP-43 that causes cytoplasmic inclusions in ALS (Smethurst et al., 2016). Prion diseases are important as models to understand the mechanisms of protein aggregation and spreading of infectious agents in neurodegenerative disorders. Recent studies indicate also that a prion-like mechanism can contribute to the cell spreading of α-synuclein in PD (Goedert et al., 2017).

In prion-infected mice, UPR and the PERK pathway was specifically activated with a sustained increase in phospho-eIF2α reduced the synthesis and levels of synaptic proteins causing neuronal death (Moreno et al., 2012). In contrast, inhibition of PERK or an overexpression of the GADD34 phosphatase decreased the p-eIF2α level and improved synaptic functions (Moreno et al., 2013). This lends credence to the view that the modulation of the PERK/p-eIF2α pathway may constitute a drug target for the development of novel therapeutics in prion diseases.

Recent studies using gene expression profiling have revealed changes in prion-infected mouse brain tissue including those in specific microRNAs (Majer and Booth, 2014), and in neuronal-specific genes that influence synaptic and brain functions (Boese et al., 2016). These findings may help to identify important preclinical stages and biomarkers for early diagnoses of prion diseases.

Increased level of the excitatory amino acid glutamate is part of the pathophysiology of brain ischemia and stroke leading to the degeneration and loss of neurons (excitotoxic damage) in the brain. Glutamate acts via different types of glutamate receptors and thereby influences cell signaling cascades and elevates intracellular calcium in neurons. Increased ER stress is linked to brain ischemia (Su and Li, 2016) and to excitotoxic injury as studied by the use of the glutamate receptor agonist, kainic acid (Sokka et al., 2007; Putkonen et al., 2011). Salubrinal that is a specific inhibitor of the eIF2α phosphatase, PP1/GADD34 (Boyce et al., 2005) and was found to protect neurons against excitotoxicity in rat hippocampus and after cerebral ischemia induced by 15 min of brain vessel occlusion in gerbils (Sokka et al., 2007; Anuncibay-Soto et al., 2016).

Epilepsy is characterized by an increased electrical activity in the brain causing repetitive seizures and neuronal dysfunctions. The severity and recurrence of seizures vary between patients and in most cases can be treated with drugs. In chronic temporal lobe epilepsy in humans it was recently shown that ER stress and particularly the IRE1α pathway are becoming activated contributing to brain damage (Liu et al., 2013). This suggests that alleviation of ER stress may be beneficial in reducing neuronal degeneration in epilepsy.

Studies of the role of ER stress responses in psychiatric and mental disorders are scarce. Bipolar disorder is characterized mood changes with alternating phases of depression and increase activity (mania). The pathophysiology of BD is complex and the disease is associated with other mental as well as metabolic disorders (Kim et al., 2017). Recent data using patient-derived blood or cell lines s showed that the ER stress response to tunicamycin or thapsigargin is compromised in BD compared with controls (So et al., 2007), suggesting a reduced resistance of cells toward stress in BD. Lithium is a mood stabilizer that is used in the treatment of BD. It was shown that lithium regulates gene network for ER adaptation in BD cells (Breen et al., 2016). Together this suggests that ER stress and associated pathways may prove beneficial targets to consider in treatment of BD (Bengesser et al., 2016) and possibly also in other mood disorders.

Regarding potential ER targets in neuropsychiatric diseases the sigma-1 receptor localized in ER subdomains interacting with mitochondria and is a target for several neuroactive drugs (Hashimoto, 2015). Activation of sigma-1 receptors can afford neuroprotection and modulates ER stress pathways in the brain (Hyrskyluoto et al., 2013; Omi et al., 2014). The precise roles of sigma-1 receptor and other ER resident molecules in neuropsychiatric disease warrant further studies.

Mounting evidence has shown a link between UPR and the development of diabetes with an altered insulin secretion, reduced β-cell viability and increased insulin resistance (IR) in peripheral tissues (Cardozo and Cnop, 2008; Salvadó et al., 2015). Lipotoxicity is a key trigger for β-cell dysfunction and saturated fatty acids are known to induce the PERK and IRE1α pathways (Sommerweiss et al., 2013). Increased insulin synthesis during IR leads to accumulation of misfolded insulin that further activates the UPR (Han et al., 2013). UPR causes accumulation of the human islet amyloid polypeptide, which occurs in 90% of T2D patients (Cadavez et al., 2014). Chaperons ameliorating the ER stress response can increase insulin secretion, pointing to novel therapeutic possibilities in β-cell rescue (Cadavez et al., 2014). The IRE1α/XBP1 branch of UPR plays a critical role in regulation of glucose and lipid metabolism in various diseases (Sha et al., 2011; Jiang et al., 2015). Activated JNK cascade contribute to obesity and insulin resistance by phosphorylation of the insulin receptor substrate, IRS-1 that attenuates insulin signaling (Ozcan et al., 2004).

In atherosclerosis, macrophages infiltrate the vessel wall and accumulate lipids and lipotoxic substances that sustain the local inflammation. In addition saturated fatty acids such as palmitate can active the IRE1α/XBP1 axis in these cells. The use of the chemical chaperone 4-phenyl butyric acid (4-PBA) was found to mitigate ER stress and reduce the amount of atherosclerotic lesions in vivo (Erbay et al., 2009). 4-PBA acts by decreasing the level of the fatty acid-binding protein-4 (aP2) that is crucial for inducing lipotoxic stress in macrophages. Lipid chaperones, acting as fatty acid sensors in these cells, may provide novel targets to consider for treating dyslipidemias (Erbay et al., 2009). In addition, the bioactive lipid palmitoleate also protected against macrophage ER stress and prevented atherosclerosis (Çimen et al., 2016). It remains to be studied whether supplementation with PAO can be of value in treatments of lipid and other metabolic disorders in humans. Recently, the use of the specific IRE1α inhibitor, STF-083010 in mice was shown to reduce atherosclerotic plaque size, lipid-induced inflammation and cytokine production in vivo (Tufanli et al., 2017), suggesting that alleviation of ER stress may provide therapeutic benefits also in cardiovascular diseases.

ER stress is also involved in hearth diseases such as myocardial ishemic/infarction and cardiomyopathies. Historically it was shown that heart failure is characterized by morphological changes in the ER, as a hallmark for the ER overload and ongoing stress (Maron et al., 1975). In cellular models, pressure overload induces ER stress and apoptosis in cardiac myocytes (Okada et al., 2004). These finding were confirmed in preclinical experiments with mice. Increased GRP78 expression has been shown in patients with heart failure (Minamino et al., 2010). Myocardial infarction also induces ER stress involving increased oxidative stress reactions. Recent data show a role for the ATF6 pathway in counteracting oxidative stress and counteracting ischemia and reperfusion injury in cardiomyocytes and in mouse hearts in vivo (Jin et al., 2017). Activation of XBP1 contributes to tissue adaptation and hypertrophy following heart injury by regulating the expression of the pro-angiogenic factor, vascular endothelial growth factor-A (VEGF-A) in vivo (Duan et al., 2016). The regulation of VEGF-A by the IRE1α/XBP1 pathway is probably of general significance in other ischemic disorders as well. Recently, it was reported that sirtuin-1 (SIRT1) counteracts ER stress and reduces the PERK/eIF2α pathway in the injured heart via deacetylation of eIF2α (Prola et al., 2017). This opens up a novel pathway for cardioprotection and for mitigation of ER stress via SIRT1 that itself can be activated by chemical compounds.

In the clinics the use of anticancer drugs can also impinge on the heart to cause dysfunctions in protein quality control events. Cardiotoxicity of such drugs are often linked to the induction of ER stress but also to changes in autophagy and the ubiquitin proteasome system; for a detailed discussion on this see the recent review (Fu et al., 2017).

Liver is the central organ in the regulation of metabolism and it ties together different metabolic challenges (lipids, toxins) with the responses in other organs. In many metabolic diseases ER stress and UPR are activated in hepatocytes that eventually can lead to liver damage. UPR in hepatocytes is important to balance fluctuations in blood glucose levels and under diabetic conditions sustained UPR leads to ER stress in the liver (Ji and Kaplowitz, 2006). ER stress itself can induce enzymes involved in gluconeogenesis modulating glucose cycling in the liver (Wang et al., 2006). Altered metabolism in form of diets containing high-carb and high saturated fats can induced accumulation of lipids (steatosis) in the liver characterized by increased ER stress (Sozen and Ozer, 2017). The induction of ER stress and steatosis precedes obesity and changes in insulin action and a sustained ER stress can worsen the metabolic dysregulation through activation of lipidogenic genes. Over-expression of ER chaperone, ORP150 was found to improve insulin resistance and ameliorate glucose tolerance in diabetic animals (Ji and Kaplowitz, 2006). Steatosis and chronic inflammation are also characteristics of alcoholic liver injury. Activation of signaling molecule such as p38 MAPK by inflammatory cytokines can further aggravate lipotoxicity and ER stress and lead to ER/Golgi remodeling (Koeberle et al., 2015). Activation of SREBP signaling by ER stress is responsible for the accumulation of lipids also in alcoholic liver models (Esfandiari et al., 2005). Targeting and relieving ER stress can reverse liver damage and protect also other tissues from harmful effects.

As ER stress has obvious role in the survival of cancer cells, IRE1α, and PERK are attractive candidates for drug development (Maly and Papa, 2014; Jiang et al., 2015). Several small molecules have been identified, that are able to inhibit IRE1α. These compounds are shown to have some anti-myeloma activity, but there are serious concerns about the selectivity of these compounds. Mores specific inhibitors, IRE1α kinase inhibiting RNase attenuators (KIRAs) that inhibit IRE1α RNAse, have been developed (Wang et al., 2012). KIRAs are able to inhibit XBP1 RNA cleavage, mRNA decay in ER and IRE1α activity and they are shown to protect against ER stress induced damages (Ghosh et al., 2014).

Potent and selective inhibitors of PERK's kinase domain have also been developed, and one such compound, GSK2656157 revealed promising anti-tumor effects in preclinical models of pancreatic adenocarcinoma and multiple myeloma (Atkins et al., 2013). Main issue with all these inhibitors are their side effects as for examples PERK inhibitors cause pancreatic β-cell loss and diabetes (Oakes, 2017). Therefore, while ER stress is necessary for the cancer progression and spread, the utility of ER stress inhibitors as anti-cancer drugs in the clinics will require careful consideration and critical evaluation of benefits and drawbacks. One possibility for treatment would be to combine UPR inhibitors with those targeting the proteasome as shown recently in case of models of hepatocellular carcinoma (Vandewynckel et al., 2016).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.