Liver disease continues to be a leading cause of death worldwide, with metabolic liver
diseases such as nonalcoholic fatty liver disease (NAFLD), its progressive component
nonalcoholic steatohepatitis (NASH), and alcoholic liver disease (ALD) becoming increasingly
prevalent.[1]
[2] Experimental disease models and observational human data demonstrate that endoplasmic
reticulum (ER) stress is a feature of acute and chronic liver diseases.[3]
[4]
[5]
[6]
[7]
[8] The key signaling pathways activated by ER stress are termed the unfolded protein
response (UPR) due to their characterization under conditions of accumulated misfolded
or unfolded proteins in the ER lumen. Physiologically, the UPR is essential for maintaining
cellular homeostasis in both hepatocytes and hepatic stellate cells (HSCs) during
metabolism and protein secretion. However, UPR signaling also drives pathogenesis
of liver disease through its involvement in inflammatory responses, steatosis, hepatocyte
apoptosis, and fibrosis via HSC activation. We discuss the UPR signaling pathways,
causes of ER stress in liver disease, and discuss UPR signaling in hepatocytes and
HSCs. While ER stress plays a physiologic and pathophysiologic role in other liver
cell types such as Kupffer cells, recent literature provides critical insight into
the pathophysiologic mechanisms of the UPR in hepatocytes and HSCs in metabolic liver
disease and fibrosis, and is the focus of this review.[9]
[10]
[11]
[12]
UPR Signaling Pathways
Unfolded protein response signaling is driven through three major pathways ([Fig. 1]), mediated by the ER transmembrane proteins inositol requiring enzyme 1 α (IRE1α),
protein kinase RNA-like ER kinase (PERK), and activating transcription factor 6 α
(ATF6α).[13]
[14] During homeostatic conditions, the luminal side of each UPR sensor interacts with
ER resident chaperones, primarily immunoglobulin-binding protein/glucose regulatory
protein 78 (BiP/GRP78). When the ER experiences an abundance of unfolded or misfolded
proteins, the sensors activate. This mechanism involves recruitment of BiP away from
the luminal domains of the sensors,[15] and there is also evidence for direct binding of unfolded proteins to the luminal
domains of IRE1α and PERK.[16]
[17] Either mechanism allows for oligomerization and activation of the sensors and UPR
signaling.[15]
[18]
[19]
[20]
[21] At a high level, UPR signaling is initially a prosurvival mechanism; reducing the
protein folding load by blocking general protein translation, increasing chaperone
expression, and increasing degradative protein export from the ER. If UPR signaling
is insufficient to relieve ER stress, proapoptotic signaling ensues. Chemical induction
of ER stress, which mimics many aspects of canonical UPR signaling, is useful for
studying the effects and mechanisms of UPR signaling. Two chemicals used in studies
discussed here are tunicamycin, which blocks protein folding in the ER through inhibiting
N-linked glycosylation, and brefeldin A (BFA), which disrupts protein trafficking
from the ER to the Golgi leading to ER protein accumulation. We will discuss each
UPR signaling pathway briefly; they are reviewed elsewhere in detail.[22]
[23]
[24]
[25]
Fig. 1 UPR sensors and signaling pathways. ER stress is sensed by three ER transmembrane proteins, ATF6α, IRE1α, and PERK, which
are crucial for mediating the adaptive and apoptotic signaling of the UPR. ATF6α translocates
to the Golgi upon sensing ER stress, where it is cleaved and subsequently trafficked
to the nucleus where it upregulates chaperones and proteins involved in ERAD. IRE1α
oligomerizes and autophosphorylates in response to ER stress, and acts through several
mechanisms. The endonuclease domain is involved in RIDD, as well as transcription
through activating the transcription factor XBP1. IRE1α also promotes apoptosis through
activation of ASK and subsequent phosphorylation of JNK. PERK oligomerizes upon sensing
ER stress, and autophosphorylates. The canonical target of PERK kinase activity is
eIF2α, which acts through ATF4 to promote expression of chaperones, and proteins involved
in ERAD, amino acid metabolism, the oxidative stress response, and UPR-mediated apoptosis.
eIF2α also serves to attenuate nonessential mRNA translation. CHOP, a stress-induced
transcription factor, is upregulated downstream of ATF4 and mediates ER stress-induced
apoptosis. Gad34, another transcriptional target of this pathway, dephosphorylates
eIF2α, thus resuming translation which can lead to apoptosis by increasing oxidative
protein folding. ASK, apoptosis-signal-regulating kinase; CHOP, CCAAT-enhancer-binding
protein homologous protein; ER, endoplasmic reticulum; ERAD, endoplasmic reticulum-associated
degradation; JNK, c-Jun N-terminal kinase; PERK, protein kinase RNA-like endoplasmic
reticulum kinase; RIDD, regulated IRE1-dependent decay; UPR, unfolded protein response.
IRE1α: Signaling through IRE1α is the most conserved of the three UPR pathways. There are
two homologs of IRE1 in humans, IRE1α and β, though only IRE1α is expressed in the
liver.[26] IRE1α is a type 2 transmembrane protein and contains two cytosolic domains critical
for downstream signaling: a kinase and an endoribonuclease domain. Under conditions
of normal protein load, IRE1α associates with chaperones in the ER lumen, which prevents
IRE1α signaling. When unfolded or misfolded proteins exceed ER capacity, IRE1α oligomerizes
and autotransphosphorylates, leading to IRE1α downstream signaling.[15]
[27] Autotransphosphorylation of IRE1α via its kinase activity activates the endoribonuclease
domain, which primarily acts through unconventional splicing of the transcription
factor X-box-binding protein 1 (XBP1). IRE1α endoribonuclease activity excises a 26-base
sequence from total or unspliced XBP1 which is then spliced by the conserved RNA ligase
RtcB to generate a shorter messenger RNA (mRNA) with a frame shift facilitating translation
of a longer, transcriptionally active XBP1 protein.[28]
[29]
[30]
[31] Spliced XBP1 (sXBP1) translocates into the nucleus and activates transcription of
several chaperones and components of the secretory pathway in an attempt to alleviate
the protein load within the ER.[32] sXBP1 also induces transcription of machinery involved in ER-associated degradation
(ERAD),[33] which facilitates the export of misfolded proteins out of the ER and targets them
for degradation. Another function of the IRE1α endoribonuclease domain is to limit
translation of nonessential mRNA and some microRNA through a process called regulated
IRE1-dependent decay (RIDD).[34]
[35]
[36] Together, the XBP1-mediated pathways are associated with the proadaptive arm of
the UPR which attempts to restore cellular homeostasis; however, RIDD is associated
with both prosurvival and proapoptotic effects via promiscuous mRNA decay and degradation
of mRNAs that encode chaperones and microRNAs that repress caspase2, respectively.[34]
[36]
[37] The IRE1α kinase domain also signals through a separate pathway via recruiting tumor
necrosis factor receptor-associated factor 2 (TRAF2), an adaptor protein that activates
apoptosis-signal-regulating kinase 1 (ASK1).[38] ASK1 phosphorylates c-Jun N-terminal kinase (JNK), which promotes apoptosis through
activating the proapoptotic protein Bim while inhibiting antiapoptotic Bcl2 proteins.[39]
[40]
[41] In addition to ASK1 and JNK, both extracellular signal-regulated kinase (ERK) and
p38 kinase are implicated downstream of IRE1α kinase activity and could play a role
in prosurvival or proapoptotic UPR signaling.[39]
[42]
PERK: PERK is a type 1 transmembrane protein that is typically activated through recruitment
of chaperones away from PERK leading to oligomerization and activation of the cytosolic
kinase domain. There is also evidence for an activating role for direct binding of
unfolded proteins to the luminal domain of PERK.[17]
[43] The major substrate of PERK phosphorylation is eukaryotic initiation factor 2α (eIF2α),
though eIF2α is also a substrate for additional kinases (heme-regulated inhibitor,
general control nonderepressible-2, and protein kinase R). eIF2α phosphorylation results
in global translation attenuation while permitting selective translation of a subset
of mRNAs, the best studied of which is activating transcription factor 4 (ATF4).[44]
[45]
[46] ATF4 transcriptionally upregulates several UPR-target genes that mediate prosurvival
or proapoptotic signaling. A critical protein that mediates proapoptotic signaling
downstream of PERK and ATF4 is CCAAT enhancer-binding protein (C/EBP) homologous protein
(CHOP), which is implicated in the progression of liver disease.[47]
[48]
ATF6α: Upon ER stress, full length ATF6α (ATF6p90) interacts with unfolded or misfolded
proteins, which exposes two Golgi localization signal sequences and initiates ATF6α
translocation to the Golgi.[18]
[49]
[50]
[51] At the Golgi, ATF6p90 is cleaved to the active form of ATF6α, ATF6p50: a bZIP-domain
containing transcription factor. ATF6p50 translocates into the nucleus and promotes
transcription of several chaperones such as BiP. ATF6p50 also heterodimerizes with
sXBP1; this heterodimer upregulates components of the ERAD machinery.[52]
Endoplasmic Reticulum Stress in Hepatocytes
Physiologic ER functions: Hepatocytes are structurally rich in ER, with ER membrane comprising 50% of total
cell membrane, and have more abundant rough ER than smooth ER, except in the pericentral
hepatocytes where hepatocytes have an equal proportion of rough and smooth ER.[53]
[54] Hepatocytes are professional secretory cells, synthesizing and secreting the majority
of plasma proteins with the exception of immunoglobulins.[55]
[56] The majority of secreted proteins and proteins destined for the plasma membrane
are cotranslationally folded in the ER followed by quality control checks before trafficking
of correctly folded proteins to either the secretory pathway or the plasma membrane.
The ER is also the site for the biosynthesis of several classes of lipids including
cholesterol, ceramide, and phospholipids; their respective synthetic machineries enriched
in the smooth ER.[57] Very low density lipoprotein (VLDL) particles begin biogenesis in the rough ER where
newly translated apolipoprotein B is lipidated by microsomal triglyceride transfer
protein (MTTP)-mediated lipid transfer as it translocates across the ER membrane,
with subsequent addition of triglycerides and other lipids in the smooth ER and Golgi.[58]
[59] The smooth ER in hepatocytes also houses the lipophilic drugs and xenobiotic detoxification
machinery that includes the family of cytochrome P450 oxidases which are involved
in alcohol metabolism. Maintenance of calcium homeostasis is an integral ER function
in hepatocytes, as in other cell types. Thus, the ER is central to the physiologic
functions of hepatocytes.
Global disruption of ER function: In spite of abundant ER, hepatocytes are sensitive to the disruption of normal ER
function. This is highlighted by observations that liver-specific deletion of the
chaperone BiP/GRP78 in mice led to disorganization and dilation of the ER compartment,
an ER stress response, despite no other stimuli. Loss of BiP also led to hepatic steatosis,
hepatocyte apoptosis, and spontaneous liver injury, highlighting the importance of
the ER in physiologic hepatocyte processes.[60] These mice were further sensitized to myriad acute and chronic hepatic insults including
alcohol, high fat diet, and acetaminophen. Similarly, loss of Sec61α1 function, a
component of ER translocon, led to hepatic steatosis and hepatomegaly with the activation
of an ER stress response.[61] Data from pharmacologic ER stress induction in genetic knockout mouse models of
UPR components also supported hepatic steatosis as a common occurrence in livers with
sustained, excessive, or unresolved ER stress.[62] These data suggest that hepatic steatosis is a conserved response to unresolved
ER stress in the liver, and perhaps UPR pathways prevent hepatic steatosis under physiologic
conditions. Notably, several of these observations linking ER stress to the mechanisms
leading to hepatic steatosis employed pharmacologic ER stress, mostly tunicamycin.
This would be in contrast to high fat diet-induced activation of the ER stress transducers,
where steatosis occurs before activation of the ER stress response, which is discussed
in subsequent sections.
Unfolded protein response transducers regulate hepatic lipid homeostasis: Several recent studies have advanced our understanding of the physiologic hepatic
functions of the individual UPR transducers in regulating hepatic lipid synthesis,
export and oxidation, which are distinct from their well-defined roles in the canonical
UPR. XBP1 liver-specific knockout mice demonstrated profound hypocholesterolemia and
hypotriglyceridemia due to reduced de novo synthesis of fatty acids.[63] IRE1α hyperactivation occurred in the livers of XBP1 liver-specific knockout mice,
and silencing of IRE1α partially restored plasma lipid levels, suggesting a key role
for IRE1α-mediated RIDD in maintaining lipid homeostasis in the liver.[64] Together, both XBP1 and RIDD regulated hepatic lipid homeostasis: genes encoding
lipogenic enzymes such as Dgat2, Acacb, and Scd1 were suppressed in XBP1 liver-specific knockout mice and the Ces1 gene family and Angptl3 were identified as RIDD substrates.[64] Furthermore, deletion of XBP1 in genetically obese (ob/ob) mice lowered hepatic triglycerides and plasma cholesterol. Thus, in the absence
of XBP1, IRE1α-mediated RIDD predominates lipid regulation with profound reductions
in plasma and hepatic lipids; however, XBP1 deletion lowers hepatic lipids in ob/ob mice.
Additional observations demonstrate that IRE1α regulates other aspects of hepatic
lipid homeostasis such as suppression of lipogenic transcription factors and maintaining
VLDL production. IRE1α repressed several transcriptional regulators of hepatic steatosis
including C/EBPβ, C/EBPδ, and peroxisome proliferator-activated receptor γ (PPARγ).[65] IRE1α-XBP1-induced protein disulfide isomerase (PDI) expression maintained VLDL
assembly and secretion, as PDI is a component of the MTTP and essential for normal
MTTP activity.[59] Altogether these data suggest a dominant role for IRE1α in maintaining hepatic lipid
homeostasis. An intact IRE1α-XBP1 signaling axis maintains hepatic VLDL secretion
and prevents hepatic steatosis in acute ER stress by inhibiting lipogenic transcription
factors, and XBP1 regulates a subset of hepatic lipid synthesis genes. Yet, when IRE1α
is hyperactivated, RIDD-induced hypolipidemia predominates.
There is evidence that other UPR sensors are involved in mediating lipid homeostasis.
Cleaved ATF6αp50 regulated lipids through binding to and inhibiting activated (cleaved)
sterol-regulatory element-binding protein 2 (SREBP-2). ATF6α served this function
in part through recruitment of HDAC1 and subsequent decreased transcription of SREBP-2
target genes.[66] Thus, regulation of hepatic lipid homeostasis is a recently identified function
of the UPR transducers that merits further study. We propose that the lipid synthesis
function of the ER in hepatocytes imparts a hepatocyte-specific, noncanonical role
to the UPR sensors such that they regulate hepatic steatosis, as discussed above,
and maintain sensitivity to lipid perturbations such as increased palmitate or phosphatidylcholine
(PC) depletion (discussed in subsequent sections).
Nonalcoholic fatty liver disease: Obesity associated-nonalcoholic NAFLD is associated with activation of all three
UPR transducers in human liver samples, and genetic and dietary mouse models of NAFLD,
in both isolated steatosis or nonalcoholic fatty liver (NAFL) and NASH.[67]
[68]
[69]
[70] We have used the term NAFLD when discussing UPR signaling with respect to hepatic
steatosis and insulin resistance, as this is conserved between NAFL and NASH, and
the term NASH when discussing inflammation and fibrosis. In recent years, observations
in genetic knockout mouse models have led to significant advances in understanding
the contribution of the UPR transducers toward NAFLD. Early observations implicated
UPR transducers in insulin resistance. Activation of JNK in livers of high fat-fed
and ob/ob mice led to inhibitory phosphorylation of the insulin receptor substrate-1 (IRS-1)
and impaired insulin signaling secondary to ER stress.[68] Newer studies have implicated a regulatory role for IRE1α-XBP1 signaling in hepatic
steatosis ([Fig. 2]), and also in liver injury and inflammation as discussed below.
Fig. 2 IRE1α activation states in NAFLD. In this model we propose two alternative activation states of IRE1α: (A) unconstrained IRE1α s-nitrosylation as observed in ob/ob mice, which may increase over time. Initial activation of XBP1 (in pink) and RIDD (in green) may favor adaptive UPR. Over time a loss of endoribonuclease activity (in purple)
would occur with increased s-nitrosylation (in blue) leading to a loss of adaptive signaling. (B) In the absence of s-nitrosylation IRE1α activation increases over time and shifts
from adaptive to maladaptive. However, whether this switch is a function of time;
regulated by the activation of alternative signaling pathways such as PERK; a cell-
and tissue-specific activation of IRE1α; or specific to accumulated toxic lipids,
remain to be determined. NAFLD, nonalcoholic fatty liver disease; PERK, protein kinase
RNA-like endoplasmic reticulum kinase; RIDD, regulated IRE1-dependent decay; UPR,
unfolded protein response.
IRE1α-XBP1: In ob/ob mice and high fat diet-fed mice, s-nitrosylation of IRE1α was reported.[69] This resulted in a progressive decline in IRE1α endoribonuclease activity with reduced
XBP1 splicing; however, kinase function was not compromised. Furthermore, reconstitution
of hepatic IRE1α expression with s-nitrosylation-resistant IRE1α restored XBP1 levels,
which is associated with improved glucose homeostasis in ob/ob mice. Interestingly, XBP1 splicing and other UPR target genes were significantly
upregulated in 7- and 12-week-old ob/ob mouse livers, before a reduction was noted in 16-week-old mouse livers; however,
no other analyses of hepatic steatosis, inflammation, and injury were reported. Separately,
Wang et al reported an increase in hepatic triglyceride content due to lack of IRE1α-mediated
degradation of precursor forms of microRNAs, miR-200 and miR-34 families, resulting
from impaired endoribonuclease activity of s-nitrosylated IRE1α,[70] suggesting that functional IRE1α mitigates hepatic steatosis by RIDD-mediated decay
of target microRNA. They also observed increased hepatic inflammation and fibrosis
in 20 week high fat-fed hepatocyte-specific IRE1α knockout mice, suggesting that IRE1α
protects from NASH.
Liver-specific XBP1 overexpression in dietary or genetic obesity models had an antisteatotic
effect due to reduced fatty acid synthesis rates resultant of Srebp1c inhibition, reduced expression of fatty acid synthase (FasN) and stearoyl-CoA desaturase 1 (Scd1), and increased macrolipophagy.[71] These observations were contrary to earlier reports that XBP1 liver-specific knockout
mice manifest hypocholesterolemia and hypotriglyceridemia and low hepatic lipid synthesis
rates.[63] Later work clarified that the hypolipidemia in XBP1 liver-specific knockout mice
is mostly due to IRE1α hyperactivation and ensuant RIDD.[64] On the other hand, when challenged with a diet known to recapitulate human NASH
in mice,[72] hepatocyte-specific XBP1 knockout mice exhibited greater liver injury, apoptotic
and inflammatory signaling, hepatocyte apoptosis, and fibrosis; however, reduced hepatic
steatosis was observed which may be attributable to RIDD.[73] IRE1α was hyperactivated in the hepatocyte-specific XBP1 knockout mice, linking
a deleterious outcome with IRE1α hyperactivation. A similar phenotype was observed
in the absence of Bax inhibitor 1 (BI-1), an ER membrane protein which negatively
regulates IRE1α activity. Nine month high fat diet-fed BI-1 knockout mice (BI-1−/−
) developed exaggerated hepatocyte cell death, liver injury, inflammation, and liver
fibrosis.[74] This was associated with overactivation of IRE1α endonuclease activity as measured
by sXBP1 protein levels and the nod-like receptor protein 3 (NLRP3) inflammasome.
Inhibition of IRE1α endonuclease activity with the pharmacologic inhibitor STF-083010
was associated with a reduction in NLRP3 inflammasome and proinflammatory markers
in BI-1−/−
mice. Thus, BI-1 constrained IRE1α endonuclease activity, and in its absence, hyperactivation
of IRE1α endonuclease activity with resultant increased sXBP1 expression was associated
with increased hepatocyte apoptosis, worse liver injury, and inflammation.
How do we reconcile these observations to explain the important yet variable role
of the IRE1α-XBP1 axis in NAFLD pathogenesis? We propose that IRE1α s-nitrosylation
may facilitate the switch between NAFL and NASH, and that BI-1 facilitates IRE1α s-nitrosylation
([Fig. 2]). However, we propose this cautiously as these models are based mostly on knockout
mouse studies and require further validation. Given the well-established paradigm
that inflammation and fibrosis in NASH occur secondary to hepatocyte apoptosis and
injury, the question whether the increase in hepatocyte apoptosis occurs in high fat-fed
hepatocyte-specific IRE1α knockout mice needs to be addressed. Kinetic analysis will
be needed to determine if the UPR shifts from adaptive to maladaptive due to sustained
ER stress, leading to hepatocyte apoptosis and proinflammatory signaling ([Fig. 2]).
Hepatic steatosis is promoted by free fatty acids derived from enhanced adipose tissue
lipolysis in an insulin-resistant state.[75] This is driven by macrophage-mediated adipose tissue inflammation. In keeping with
this, IRE1α signaling in immune cells or adipose tissue may play an understudied role
in NAFLD. Macrophage IRE1α promoted proinflammatory polarization of adipose tissue
macrophages, such that when IRE1α was deleted from myeloid cells, mice were resistant
to diet-induced weight gain, insulin, hyperlipidemia, and hepatic steatosis.[76] Thus, the tissue-specific role of IRE1α in hepatic macrophages merits examination.
IRE1α s-nitrosylation and BI-1 function in adipose tissue may also influence the development
of NAFLD, providing another scenario for experimental testing. In summary, negative
regulation of steatosis by XBP1 and RIDD appears to be beneficial in reducing hepatic
steatosis, whereas inactivation of IRE1α endoribonuclease activity promotes hepatic
steatosis and hyperactivation of IRE1α promotes NASH. There remain multiple unanswered
mechanistic questions, as discussed above, on cell autonomous functions and the tissue-specific
role of IRE1α in NAFLD pathogenesis.
PERK: PERK signaling is also implicated in NAFLD. One of the earliest observations in
this regard demonstrated that enforced expression of GADD34, a downstream effector
of PERK, improved hepatic steatosis and insulin tolerance in high fat-fed mice due
to attenuated expression of the lipogenic transcriptional regulators PPARγ and C/EBPα
and β.[77] ATF4 deficiency protected against high carbohydrate diet (HCD)-induced hepatic steatosis
by downregulating HCD-induced SCD1 expression.[78] When challenged with a high fructose diet, ATF4 knockout mice were protected from
hepatic steatosis and hypertriglyceridemia due to reduced induction of PPARγ and other
lipogenic genes.[79] Mechanistically, CHOP-dependent hepatocyte lipotoxicity is an important mediator
of NASH pathogenesis and links ER stress to NASH ([Fig. 3]). In isolated hepatocytes, saturated free fatty acid-induced apoptosis was CHOP-dependent
by increasing the expression of the death receptor 5 (DR5) and proapoptotic Bcl-2
family protein p53 upregulated modulator of apoptosis (PUMA).[80]
[81] The palmitate-repressed microRNA miR-615–3p repressed CHOP expression in hepatocytes
such that augmentation of miR-615–3p led to reduced CHOP expression and protected
hepatocytes from palmitate-induced apoptosis.[82] CHOP knockout mice (CHOP−/−
), however, developed greater liver injury, inflammation, and hepatocyte apoptosis
when fed a high-fat diet.[83]
[84] This was reportedly due to impaired macrophage apoptosis in CHOP−/−
mice which increased proinflammatory signaling. In contrast, it was reported that
CHOP−/−
mice are either protected from a methionine- and choline-deficient diet-induced steatohepatitis,[85] or demonstrated a similar degree of liver injury.[86] Further, hepatocyte-specific CHOP deletion did not protect or worsen liver injury
and ER stress in high fat-fed MUP-uPA mice. These mice express the urokinase-type
plasminogen activator under the hepatocyte-specific major urinary protein promoter
(MUP-uPA); uPA accumulation in the ER activates the ER stress response.[87]
[88] Thus, the role of CHOP remains incompletely understood in NASH pathogenesis. Complicating
these studies is the varied mechanism of dietary injury, which may account for some
of the observed differences in the role of CHOP in steatohepatitis. Additionally,
tissue-specific roles of CHOP, especially in myeloid cells, may explain some of the
disparate observations.
Fig. 3 Lipotoxicity and ER stress in NASH. The saturated free fatty acid palmitate, and other toxic lipids such as lysophosphatidylcholine
(LPC) are known to activate the three UPR sensors. IRE1α, via phosphorylation and
activation of the stress kinase c-Jun N-terminal kinase (JNK) targets the insulin
receptor substrate 1 (IRS1) for inhibitory phosphorylation leading to insulin resistance.
IRE1α-dependent RNA degradation (RIDD) negatively regulates hepatic steatosis via
degradation of lipogenic mRNAs and regulatory microRNA. IRE1α s-nitrosylation inhibits
its endoribonuclease activity with a resultant decrease in XBP1 splicing and an increase
in microRNAs that are degraded via RIDD. The PERK pathway is implicated in the regulation
of hepatic steatosis; CHOP mediates palmitate-induced hepatocyte apoptosis, though
its in vivo role in NASH is less clear. ATF6α is activated in NASH; however, its contribution
is not well defined. Lipid perturbation-induced UPR activation activates a subset
of genes distinct from misfolded protein-induced UPR; however, the specific ATF4 and
XBP1 induced lipotoxic target genes have not been defined. CHOP, CCAAT-enhancer-binding
protein homologous protein; ER, endoplasmic reticulum; NASH, nonalcoholic steatohepatitis;
PERK, protein kinase RNA-like endoplasmic reticulum kinase; UPR, unfolded protein
response.
ATF6α and sarco/endoplasmic reticulum calcium ATPase (SERCA): ATF6α signaling is likely important in NASH pathophysiology; however, not much is
known. High fat-fed ATF6α−/− mice demonstrated greater hepatic steatosis with associated increased XBP1 splicing,[89] suggesting a role for ATF6α in promoting adaptation to dietary obesity-induced fatty
liver. Additionally, disrupted phospholipid composition, leading to inhibition of
the SERCA activity, was reported in livers of obese mice.[90] A limiting factor in these studies is the lack of validated reagents to test the
ATF6α pathway, highlighting a major limitation in fully understanding ATF6α in NASH
and other liver diseases.
Endoplasmic reticulum stress and inflammasome activation: NLRP3, an inflammasome component, is associated with NASH progression.[91] ER stress is also associated with activation of the NLRP3 inflammasome by lipopolysaccharide
and hepatocyte cell death,[92] with a correlation in liver biopsies from NASH subjects between NLRP3 expression,
ER stress markers, and liver injury. ER stress-induced inflammasome activation likely
occurred independently of classical UPR signaling pathways.[93] Thus the UPR transducers and ER stress play myriad roles in NAFLD pathogenesis.
It is possible that IRE1α hyperactivation, CHOP-dependent macrophage apoptosis, and
ER stress-induced inflammasome activation are determinants of NASH; these concepts
require further experimental validation.
Mechanism of lipid-induced ER stress: Lastly, though we have focused mostly on in vivo observations, there have been considerable
advancements in the mechanisms by which ER stress pathways are activated in hepatocytes
([Fig. 3]). NASH is a lipotoxic disease, where the following lipids have been identified as
being elevated and injurious to hepatocytes: the saturated free fatty acid palmitate,
the phospholipid lysophosphatidylcholine (LPC), sphingolipids including C16:0 ceramide,
and free cholesterol.[94] Palmitate-induced increases in membrane saturation activated IRE1α and PERK in a
mechanism that relied on the transmembrane domain of each protein.[95] This and additional studies demonstrated activation of all three UPR sensors by
palmitate via induction of downstream signaling molecules including sXBP1, eIF2α phosphorylation,
increased abundance of ATF4 and CHOP, and transcriptional targets of ATF6α.[80]
[81]
[82] CHOP was of particular interest, as it was implicated in hepatocyte apoptosis via
upregulation of DR5, which induces apoptosis upon engagement of its ligand tumor necrosis
factor-related apoptosis-inducing ligand (TRAIL), and the proapoptotic protein PUMA.[80]
[81] CHOP also suppress the antiapoptotic protein Bcl-2,[96] though this has limited relevance to hepatocytes, which do not express Bcl-2.[97]
In contrast to palmitate, monounsaturated fatty acids such as oleate and palmitoleate
are known not to be toxic, but rather mitigate the palmitate toxicity by promoting
more efficient esterification into neutral triglyceride.[98]
[99] In keeping with this, palmitoleate attenuated palmitate-induced XBP1 splicing.[100] However, prolonged (16 hours) or high concentration (800 and 1600 μM) exposure to
oleate activated the UPR sensors in a rat hepatocyte cell line.[101] This may be due to conversion of exogenously loaded oleate to palmitate, as has
been previously reported;[102] though, this is unclear as this study did not measure palmitate levels in oleate
loaded cells. At equimolar concentrations, the magnitude of BiP induction and eIF2α
phosphorylation was greater in palmitate-treated cells in comparison to oleate-treated
cells,[103] perhaps due to only partial conversion of exogenously loaded oleate into palmitate.
Ceramide is also implicated in NASH. Ceramide accumulation from de novo synthesis
occurs in palmitate-treated cells, and in rat hepatoma cells this was variably reported
as decreasing or not impacting palmitate-induced ER stress.[103]
[104] However, release of extracellular vesicles from palmitate-treated cells depended
on the de novo synthesis of ceramide in an IRE1α-XBP1 dependent fashion.[105] Another possible mechanism for palmitate-induced UPR signaling may be post-translational
s-palmitoylation, which has been described for the ER chaperone calnexin but not for
any of the three UPR sensors.[106]
[107]
Another lipotoxic lipid implicated in NASH is LPC, which recapitulates many of the
features of palmitate-induced ER stress and lipotoxicity when applied exogenously.[108]
[109] Endogenous LPC is synthesized from several pathways: (1) hydrolysis of the fatty
acid at sn-2 position by the action of phospholipase A2 (PLA2) on PC, (2) lecithin
cholesterol acetyl transferase (LCAT) activity in high-density lipoprotein and low-density
lipoprotein, and (3) endothelial lipase and hepatic lipase.[110] Circulating LPC levels in plasma are high under normal conditions and its hepatocyte
lipotoxicity was dependent on hepatocyte-PLA2, implying intracellular generation of
LPC from PC.[108]
[111] Unlike palmitate, LPC generation from PC may lead to plasma membrane PC depletion
with subsequent activation of an ER stress response. This mechanism has been demonstrated
in budding yeast and Caenorhabditis elegans but not yet in mammalian hepatocytes.[112]
[113] Interestingly, PC depletion activated a subset of genes unique from those activated
following tunicamycin treatment, suggesting that lipid perturbation-induced ER stress
differs from chemically induced ER stress.
In summary, lipid-induced perturbations promote ER stress, the best described of which
are palmitate-induced increased membrane saturation and PC depletion. It remains to
be seen whether mammalian lipid perturbation-induced ER stress differs from unfolded
protein-induced ER stress, but this would provide an interesting avenue of study and
increase our understanding of pathological mechanisms of NAFLD and NASH.
Alchoholic liver disease: Alcohol metabolism is a major driver of ER stress in hepatocytes in ALD. Alcohol
is primarily metabolized into acetaldehyde in the cytosol by alcohol dehydrogenase,
followed by conversion to acetate in the mitochondria. During excessive ethanol consumption,
the smooth ER localized p450 cytochrome CYP2E1 participates in the ethanol-to-acetaldehyde
metabolism; however, this process generates reactive oxygen species (ROS) and activates
ROS-induced signaling pathways. Recent studies showed that ethanol-induced hyperhomocysteinemia
due to reduced methionine synthase activity led to ER stress due to interference with
protein folding.[114]
[115] The administration of betaine, a methyl donor used for the conversion of homocysteine
to methionine, to mice receiving intragastric ethanol to induce alcoholic steatohepatitis
(AH) prevented an increase in homocysteine, while reducing ER stress.[116] ER chaperones including GRP78 and glucose regulated protein 94 (GRP94) were upregulated
in an early and sustained manner in mice fed a chronic ethanol diet.[116] This was associated with increases in SREBP-1 and HMG CoA reductase expression,
possibly contributing to ethanol-induced steatosis. CHOP was also induced by ethanol
feeding; furthermore, ethanol-fed CHOP −/− mice were protected from hepatocyte apoptosis,
though they were equally susceptible as wild-type mice to ethanol-induced hyperhomocysteinemia,
liver steatosis, and injury.[115] Acid sphingomyelinase (ASMase) was upregulated in human liver biopsies in AH, and
ASMase knockout mice were resistant to alcohol-induced activation of ER stress response
in spite of comparable increases in homocysteine levels. This suggested that homocysteine-independent
effects of alcohol also contributed to the activation of the ER stress response in
AH.[117] Mallory–Denk bodies (MDBs), protein aggregates of ubiquitinated keratin 8 and keratin
18, were a hallmark feature of AH. Failure of protein quality control along with activation
of an ER stress response was associated with the formation of MDBs in human liver
biopsy samples.[118] Overall, though ER stress is associated with many features of ALD, the mechanistic
pathways and contribution of each UPR sensor and their downstream effectors in pathogenesis
of AH are less defined than NAFLD.
Unfolded Protein Response Signaling in HSC Activation and Fibrogenesis
Hepatic stellate cells are the key fibrogenic cells in the liver. Activation of HSCs
in response to liver injury leads to production and secretion of extracellular matrix
(ECM) proteins.[119] Enhanced protein secretion is associated with ER stress and UPR signaling in numerous
secretory cells, indicating that UPR signaling may be crucial for processing and secretion
of ECM proteins.[120] Indeed, UPR signaling was observed in HSCs in response to liver injury in vivo (i.e.,
alcohol and carbon tetrachloride (CCl4)), and in response to fibrogenic stimuli in vitro (i.e., transforming growth factor
β [TGFβ], oxidative stress, and stiffness).[8]
[121]
[122]
[123]
[124] This suggests that UPR induction results from increased ECM production; however,
UPR-mediated ER expansion prior to increased protein production was observed in other
cell types.[125] Conversely, chemical induction of UPR signaling through tunicamycin and BFA promoted
expression of fibrotic genes such as procollagen I, and these effects were reversed
by inhibiting one or all three UPR sensors.[8]
[121]
[122]
[126] Similar to the paradox in hepatocytes that ER stress can induce cellular injury
but cellular injury can cause ER stress, ER stress may be a driver and a result of
HSC activation. Adding another layer of complexity to UPR signaling in HSCs, inducing
the UPR in HSCs could serve as an antifibrotic strategy through activating proapoptotic
signaling pathways. Here we will discuss recent studies looking at each of the three
UPR sensors and their role in HSC activation, and how they mediate adaptive or apoptotic
responses ([Fig. 4]).
Fig. 4 UPR signaling pathways during HSC activation and potential strategies for antifibrotic
targeting. (A) HSC activation through signals such as inflammation, TGFβ, or oxidative stress leads
to ER stress and induction of the UPR. It is yet unclear how activation signals lead
to UPR induction, whether through upregulation of ECM proteins or a separate mechanism
independent of increased gene transcription. UPR signaling downstream of PERK, ATF6α,
and IRE1α further promotes HSC activation and fibrogenesis through increased SMAD2/3
phosphorylation, ER expansion, and enhanced protein secretion, as well as additional,
unexplored mechanisms. (B) Targeting the UPR provides potential strategies for either limiting HSC activation
or promoting apoptosis of activated HSCs, both favorable for fibrosis resolution.
General UPR inhibition through chemical chaperones or BiP overexpression has already
shown to be antifibrotic in animal models, and could be pursued further. In addition,
preferential targeting of activated HSCs for apoptosis, whether through general activation
of UPR signaling (etoposide, caffeine, quercetin, or azithromycin), increased CHOP
expression, ASK1-JNK signaling, or disrupted protein export from the ER (inhibition
of TANGO1), could prove to be a useful strategy for antifibrotic therapies. ASK1,
ASK, apoptosis-signal-regulating kinase 1; BiP, immunoglobulin-binding protein; CHOP,
CCAAT-enhancer-binding protein homologous protein; ECM, extracellular matrix; ER,
endoplasmic reticulum; HSC, hepatic stellate cell; JNK, c-Jun N-terminal kinase; PERK,
protein kinase RNA-like endoplasmic reticulum kinase; UPR, unfolded protein response.
IRE1α: IRE1α signaling plays an important role in HSC activation. Inhibition of IRE1α through
4μ8C, a noncompetitive inhibitor that blocks IRE1α kinase and endoribonuclease signaling,
effectively blocked both TGFβ- and BFA-induced activation of HSCs in vitro, and in
an alcohol/CCl4 liver fibrosis model in vivo.[122]
[127]
[128] Despite this clear relationship, the mechanisms by which IRE1α mediates HSC activation
and the regulation of IRE1α signaling during HSC activation are less understood. Both
the endonuclease and the kinase domains of IRE1α are implicated in HSC activation
and will be reviewed in detail.
IRE1α endonuclease activity and HSC activation: A critical role for IRE1α signaling in HSC activation was initially reported upon
observations of increased sXBP1 in response to ethanol or oxidative stress.[121] Furthermore, TGFβ-promoted XBP1 splicing in HSCs, while 4μ8C-mediated IRE1α inhibition
reduced TGFβ-induced sXBP1 and expression of BiP, α smooth muscle actin (αSMA), collagen
Iα1, and connective tissue growth factor (CTGF).[121]
[122]
[124] Several pieces of evidence suggested that XBP1 acts to facilitate HSC activation
through increasing protein secretion. First, sXBP1 correlated with TGFβ-induced ER
dilation during HSC activation, a phenotype associated with increased ER secretory
capacity.[122] Second, XBP1 was critical for expression of transport and Golgi organization protein
1 (TANGO1), a critical component of the collagen I ER export machinery.[124] Furthermore, XBP1 overexpression was sufficient to drive HSC activation and increased
expression of several genes involved in protein secretion.[123] Interestingly, signaling pathways activated by XBP1 overexpression did not include
TGFβ-responsive signaling, as evidenced by unchanged expression of SMAD2, a predominant
mediator of TGFβ signaling, or other downstream components.[123] This suggested that XBP1 contributes to HSC activation through facilitating cargo
secretion by expanding ER capacity and upregulating protein secretion machinery as
opposed to enhanced TGFβ signaling. Another less explored role for the IRE1α endonuclease
domain in HSC activation may be through cleavage of miRNAs. IRE1α-mediated cleavage,
and subsequent degradation of miR-150, led to increased αSMA expression, a known miR-150
target.[122] Additional miRNA targets of RIDD may mediate HSC activation and fibrogenesis, and
remain to be explored.
IRE1α kinase activity and HSC activation: IRE1α kinase activity also influences HSC activation through one of its downstream
targets, p38. p38 is a mitogen-activated protein kinase (MAPK) involved in regulating
cell proliferation and autophagy, but also promotes apoptosis downstream of IRE1α.[42]
[129] In HSCs, UPR induction of p38 activity regulated TGFβ signaling. Tunicamycin induced
p38 phosphorylation downstream of IRE1α, and p38 inhibition decreased expression of
collagen Iα1 and αSMA.[121] Later studies showed that IRE1α activation in response to BFA led to SMAD2/3 phosphorylation
and procollagen Iα1 expression in a p38-dependent manner, suggesting crosstalk between
TGFβ and IRE1α kinase activity.[126] Finally, little is understood in regards to UPR regulation of the other downstream
effectors of the IRE1α kinase domain (ASK1 and JNK) during HSC activation, but the
importance of JNK in HSC activation suggests that UPR signaling could further regulate
HSC activation and fibrogenesis through modulation of these kinases.[130]
PERK: Mouse models of liver disease, such as CCl4 injection or administration of a high-fat diet, identified a positive correlation
between increased PERK phosphorylation and elevated expression of HSC activation markers.
The mechanisms by which PERK signaling promotes HSC activation have not received much
attention, but recent evidence suggested that PERK augments TGFβ signaling to maintain
HSC activation in the presence of chronic liver injury. A noncanonical PERK substrate
found to play a role in liver fibrosis is heterogeneous nuclear riboprotein A1 (HNRNPA1).
Phosphorylation of HNRNPA1 by PERK was necessary for tunicamycin to upregulate proteins
involved in TGFβ signaling, including SMAD2.[8] HNRNPA1 processed miR-18, which in turn induced the degradation of SMAD2 mRNA. Upon
phosphorylation by PERK, HNRNPA1 was targeted for degradation, leading to reduced
miR-18 processing and increased SMAD2 expression. Thus, PERK promoted SMAD2/3-mediated
expression of profibrotic genes. This study also identified HNRNPA1 as a potential
drug target as lentiviral delivery of HNRNPA1 limited CCl4-induced fibrosis.[8] While a role for PERK in HSC activation is clear, no direct role for the primary
target of PERK, eIF2α, has been identified, and merits further investigation.
ATF6α: Little is known regarding ATF6α signaling in HSC activation. ATF6α inhibition reduced
HSC activation in response to BFA; however, no mechanistic studies are reported thus
far.[126] ATF6α likely plays a role in HSC activation, as ATF6α was implicated in cardiac
fibrosis, where ATF6α activation promoted production and folding of ECM proteins.[131]
[132] Furthermore, UPR signaling in cardiomyocytes, the primary fibrogenic cells in the
heart, was implicated in prolonged TGFβ signaling, and may play a wider role in cardiac
fibrogenesis. More studies are required to understand the role of ATF6α signaling
during HSC activation.
Potential therapies: Therapeutic targeting of ER stress in HSCs is attractive as an antifibrotic therapy,
through either inhibition of ER stress and HSC activation, or induction of UPR-mediated
apoptosis. Fibrosis was limited in murine models through inhibition of ER stress by
4µ8C or 4-phenylbuterate (4-PBA), a small molecule that functionally reduces ER stress
via a mechanism that remains poorly understood. Another study utilized a viral delivery
system where BiP expression was driven by an αSMA promoter, intended to specifically
overexpress BiP in activated HSCs.[8] This treatment effectively reduced CCl4- or tunicamycin-driven fibrogenesis. While not yet feasible for human application,
this work shows that targeting UPR signaling in HSCs may be just as important as targeting
UPR signaling in hepatocytes.
Another approach to targeting ER stress as an antifibrotic therapy is driving UPR-mediated
HSC apoptosis. This approach is more difficult, since any drug would need to specifically
target activated HSCs for apoptosis, with little effect on hepatocytes. Studies using
the anticancer drug etoposide showed increased apoptosis of LX-2 cells through a UPR
signaling mechanism, with reduced effects on the hepatocyte cell lines LO-2 and QSG-7701.[133] Compounds such as caffeine and quercetin also promoted HSC apoptosis through UPR
signaling.[134]
[135] Additionally, a promising treatment for idiopathic pulmonary fibrosis (IPF) may
translate to liver fibrosis. The drug azithromycin significantly increased survival
of patients with acute exacerbation of IPF (AE-IPR), and acted in vitro to limit differentiation
of lung fibroblasts in part through increased UPR signaling.[136]
[137] HSC apoptosis was also induced through forced accumulation of secretory proteins
in the ER. RNA interference (RNAi)-mediated knockdown of TANGO1 led to procollagen
I retention within the ER and increased cell death in vitro, whereas HepG2 cells were
unaffected by TANGO1 knockdown.[138] The drawback of targeting HSCs directly lies in the lack of effective and selective
delivery mechanisms. HSC-specific targeting by adenoviral delivery or targeted liposomes
have been explored, but with only moderate success. This is a major hurdle which needs
to be addressed if antifibrotic therapies are to specifically target activated HSCs
for apoptosis.