The role of ER chaperones in protein folding and quality control

Abstract

Molecular chaperones assist the folding of nascent chains in the cell. Chaperones also aid in quality control decisions as persistent chaperone binding can help to sort terminal misfolded proteins for degradation. There are two major molecular chaperone families in the endoplasmic reticulum (ER) that assist proteins in reaching their native structure and evaluating the fidelity of the maturation process. The ER Hsp70 chaperone, BiP, supports adenine nucleotide-regulated binding to non-native proteins that possess exposed hydrophobic regions. In contrast, the carbohydrate-dependent chaperone system involving the membrane protein calnexin and its soluble paralogue calreticulin recognize a specific glycoform of an exposed hydrophilic protein modification for which the composition is controlled by a series of glycosidases and transferases. Here, we compare and contrast the properties, mechanisms of action and functions of these different chaperones systems that work in parallel, as well as together, to assist a large variety of substrates that traverse the eukaryotic secretory pathway.

INTRODUCTION

Anfinsen’s seminal studies demonstrated that the primary sequence of the protein determines the three-dimensional structure of the folded protein and that small proteins can fold in isolation in the test tube (Anfinsen 1973). The folding of more complex proteins; however, frequently requires assistance from cellular folding factors. By transiently interacting with the maturing nascent chains molecular chaperones increase the folding efficiency in the suboptimal folding environment of the cell. Prolonged interactions with chaperones can also play a central role in quality control decisions that supports the targeting of irreparable misfolded proteins for degradation.

Hsp70 is a well conserved classical family of chaperones that is ubiquitously expressed throughout the cell and performs diverse roles (Balchin, Hayer-Hartl, and Hartl 2020). In the endoplasmic reticulum (ER) where proteins targeted to the eukaryotic secretory pathway mature the Hsp70 family member is BiP (Binding-immunoglobulin Protein or Grp78) (Pobre, Poet, and Hendershot 2019). In addition to assisting in protein folding, BiP aids in protein translocation, as well as the preparation and targeting of terminal misfolded for degradation.

Proteins that traverse the secretory pathway are frequently modified by N-linked glycans, and a chaperone system has evolved to utilize these glycans. The carbohydrate-dependent chaperone system binds to the N-glycan modification based on their dynamic composition to aid in folding and quality control of the cargo they are attached to (Hebert et al. 2014). Altogether, the Hsp70 and carbohydrate-dependent chaperone networks of the ER assist a broad range of clients that pass through the secretory pathway to help maintain cellular homeostasis and these chaperone networks are the focus of this chapter.

BIP BINDING CYCLE

BiP is a member of the classical Hsp70 molecular chaperone family. BiP substrate binding is regulated by adenine nucleotides and its associated co-factors. Because of its diverse roles in maintaining ER homeostasis, it has been referred to as a master regulator of the ER (Pobre, Poet, and Hendershot 2019). As a luminal chaperone of the ER, it assists in the translocation, folding and quality control of nascent chains as they emerge into the ER. Beyond its roles that involve interacting directly with secretory pathway client proteins, BiP also plays important roles in controlling the activity of ER resident signaling molecules that transcriptionally regulate the composition of the ER depending upon protein homeostasis demands (Hetz, Zhang, and Kaufman 2020). Altogether, BiP is a critical protein homeostasis factor of the ER.

BiP organization and structure

BiP is composed of two discrete functional domains that communicate with each other in part through a flexible linker. The N-terminus of BiP forms a nucleotide binding domain (NBD; Figure 1, green domain) followed by a C-terminal substrate binding domain (SBD; Figure 1, purple domain) (Chang et al. 2008). The NBD is characterized by its ability to bind adenine nucleotides, its ATPase activity and its capacity to control the conformation or state of the SBD.

Figure 1. An overview of the BiP substrate binding cycle.

In the ATP-bound state, BiP is in the open conformation (1), with both the SBD (purple) and NBD (green) domains docked together. Substrate may be bound directly or transferred to BiP via a co-factor. Here, ERdj6 brings substrate (red) bound in the TPR domain (orange) to pass to BiP (2), then stimulates the ATPase activity of BiP via the J-domain (grey) (PDB: 2Y4K). Upon stimulation, BiP hydrolyzes ATP to ADP and the SBD clamps onto the substrate (3), undocking from the NBD for more dynamic movement (PDB: 5E85). Nucleotide exchange is facilitated by a NEF (Sil1) (4), exchanging ADP for ATP, which causes BiP to unclamp from the substrate and occupy the ATP- bound open conformation once more (5) (PDB: 3QML). BiP sequestering can be achieved through modification, specifically AMPylation via FICD. FICD AMPylates BiP as a monomer and de-AMPylates BiP as a dimer (5) (AMP shown on the structure in red) (PDB: 6I7K, SO4P, 6I7G, respectively).

The NBD of BiP is a large globular domain organized into two large lobes with a deep cleft containing the nucleotide binding pocket between them (Pobre, Poet, and Hendershot 2019). The nucleotide binding pocket contains the adenine nucleotide sensing residue T^37, which interacts with the γ-phosphate of the nucleotide, distinguishing ADP and ATP (Wei, Gaut, and Hendershot 1995). Hsp70 proteins occupy two main conformations, an open or docked state (ATP bound), or a closed or undocked state (ADP bound). Transition between the two states is regulated allosterically by the nucleotide bound state of the NBD. The ATPase activity of BiP can be stimulated via interaction with J-domain containing proteins through their defining conserved H-P-D or J-motif (Pobre, Poet, and Hendershot 2019). The invariant D of the J-motif interacts with the NBD and leads to a destabilization of the domain-domain interface of the NBD, resulting in ATP hydrolysis.

The SBD is comprised of both the substrate binding pocket, located between two loops, as well as the extended alpha-helical lid region to cap or trap the substrate. The substrate binding pocket is located between the two regions of the SBD: SBDα and SBDβ (Yang et al. 2015). The SBDα region is largely alpha-helical and comprises the lid of the Hsp70, while the SBDβ consists of eight beta strands (Figure 1, N-terminal purple domain) (Yang et al. 2015; Chang et al. 2008). In the ATP bound state, the extended alpha-helical SBDα domain remains open, which allows for more promiscuous interactions with substrates, facilitating interactions with a high on- and off-rate. Hydrolysis of the bound ATP leaves the NBD in the ADP-bound state, which in turn causes an allosteric conformational shift in the SBD; the SBDα closes over the SBDβ as it undocks from the NBD. The two domains now only remain loosely connected by the extended flexible linker region, allowing for the dynamic independent movement of both domains in contrast to the docked ATP bound state (Yang et al. 2015). The substrate binding pocket resides between the SBDβ and the helical SBDα and the conformational shift in both domains leads to the trapping of the substrate ensuring prolonged interaction with the Hsp70 (Yang et al. 2015; Chang et al. 2008).

BiP binding specificity

Chaperones from the classical heat shock families bind exposed hydrophobic domains on substrates to drive proper folding and retard aggregation (Balchin, Hayer-Hartl, and Hartl 2020). Using an affinity panning assay, BiP was found to preferentially bind hydrophobic domains containing alternating aromatic and hydrophobic amino acids (Blond-Elguindi et al. 1993; Rüdiger et al. 1997). These stretches of 7–9 amino acid tend to be buried in the mature protein fold and are indicative of an immature, unfolded protein or exposed non-native domains. BiP binding sites have been statistically predicted to be located every ~36 amino acids (Rüdiger et al. 1997).

BiP binds additional sites dispersed throughout the protein rather than just rare sites with higher aggregation potential, indicating a strong pro-folding function (Behnke et al. 2016). These binding motifs within nascent polypeptides often times are seven amino acids long, however BiP can also bind larger epitopes, sometimes comprising tertiary structural components (Behnke et al. 2016). These larger structures can interact directly with the lid domain of the SBD of BiP suggestive of a larger pool of substrates.

Substrate delivery to BiP

BiP relies upon its cofactors to facilitate its ATPase cycle to transition it to its high affinity substrate binding state. The ER-localized DNAJ cofactors (ERdjs) facilitate the hydrolysis of ATP through the stimulation of BiP via an interaction with their J-domains; as such, all ERdj family members are marked by both ER localization and a luminal J-domain (Pobre, Poet, and Hendershot 2019). ERdj proteins can be broadly understood to interact with BiP primarily via their J-domain, but the majority harbor additional roles. The ERdjs can be split into two general families: (i) translocation regulators, or (ii) substrate binders along with their J-domain functionality. The ERdj proteins which work to facilitate translocation are ERdj1 and ERdj2, while the other ERdjs (with the exception of ERdj7 for which exists no known function) bind directly to substrates for quality control. Both subfamilies of ERdj proteins facilitate the ATPase and substrate binding cycle, allowing for BiP to efficiently function.

The only mammalian homologue of E. coli DNAJ localized to the ER is ERdj3 as it contains all the domains found in bacterial DNAJ. It is also highly conserved with other soluble DNAJ homologues such as Ydj1 from Saccharomyces cerevisiae. ERdj3 has been found to form tetramers instead of the more common dimeric assemblies associated with the other Hsp40 co-chaperones (Chen et al. 2017). These tetramers appear to be created via the dimerization of dimers. By binding directly to an unfolded protein substrate, ERdj3 retards aggregation of the substrate until it is transferred to BiP (Shen and Hendershot 2005). The stimulation of the ATPase activity of BiP is necessary for ERdj3 to release its bound substrate (Jin et al. 2008). ERdj3 interacts with the lid of the SBD to facilitate passage of the substrate to BiP (Marcinowski et al. 2011).

ERdj4 is a soluble ERdj that contains an N-terminal J-domain followed by a G-F rich flexible linker region and a C-terminal SBD. The expression level of ERdj4 is low under normal conditions, but is upregulated upon ER stress (Shen and Hendershot 2005). It is hypothesized that ERdj4 functions primarily in a pro-ERAD process with emphasis on binding substrates to facilitate their selection, preparation or entrance to the degradation process (Pobre, Poet, and Hendershot 2019). ERdj4 also appears to help recruit BiP to IRE1 thus functioning to quench the activated dimeric state of Ire1 to return it to an inactive monomeric state (Amin-Wetzel et al. 2017). Thus, ERdj4 may be an integral part of the unfolded protein response (UPR) transduction signaling pathway, a function further supported by the relative lack of expression under normal growth conditions.

ERdj5 and ERdj6 also directly bind substrate. ERdj5 is a unique ER localized J-protein as it is also part of the protein disulfide isomerase (PDI) family of co-chaperones. It contains an ER luminal J-domain as well as six thioredoxin domains (Cunnea et al. 2003). The crystal structure of ERdj5 indicates that the J-domain and the four C-terminal thioredoxin domains are oriented in the same plane (Hagiwara et al. 2011). These domains may work in concert by allowing ERdj5 to act on substrates bound by BiP by acting as a reductant to reduce non-native substrates to prepare them for dislocation from the ER for proteasomal clearance in the cytosol (Ushioda et al. 2008). ERdj6, also known as DNAJC3/P58^IPK, is a soluble ERdj protein that contains an N-terminal ER targeting sequence followed by nine tetratricopeptide repeat domains (TPRs) and a C-terminal J-domain (Figure 1 orange, grey domains respectively) (Pobre, Poet, and Hendershot 2019; Rutkowski et al. 2007). Under normal growth conditions, ERdj6 resides in the ER and binds to both unfolded protein substrates and BiP (Rutkowski et al. 2007; Petrova et al. 2008). In contrast, under stress, ERdj6 translocation into the ER is less efficient and consequently ERdj6 can accumulate in the cytosol (Rutkowski et al. 2007). The TPR domains in ERdj6 are organized into three subdomains of three TPRs each, where subdomain 2 binds to PKR and subdomain 1 is the primary substrate binding region (Melville et al. 1999; Svärd et al. 2011). The concerted effort between the TPR subdomain 1 and J-domain facilitates the exchange of bound unfolded substrate to BiP (Svärd et al. 2011). ERdj proteins play a critical role in the regulation of BiP, contributing to both the delivery of new substrate as well as facilitating the ATPase cycle of BiP.

Substrate release from BiP

Release from the BiP binding cycle relies on the exchange of ADP for ATP. This supports the raising of the SBD alpha-helical lid and freeing of the trapped substrate. The co-chaperones that assist this process are called nucleotide exchange factors (NEFs). There are two ER resident NEFs that act as cofactors for BiP: BiP-associated protein (Bap), and glucose-regulated protein 170 (Grp170). Both BiP cofactors were originally found in yeast. Bap is the human homologue to Saccharomyces cerevisiae Sil1p (Behnke, Feige, and Hendershot 2015). Similarly, luminal Hsp 70 (Lhs1p) is the homologue of human GRP170 (Craven, Egerton, and Stirling 1996).

Bap shares tissue specific expression profiles and strongly co-localizes with BiP (Chung, Shen, and Hendershot 2002). Consequently, Bap was found to preferentially interact with the ADP-bound state of BiP and facilitate the exchange of ADP for ATP (Rosam et al. 2018). The interaction of the Bap C-terminal domain is sufficient for its weak NEF activity, however full length Bap is necessary for its complete nucleotide exchange function (Rosam et al. 2018). The C-terminus of Sil1 interacts with BiP, resulting in the opening its nucleotide binding pocket to facilitate nucleotide exchange (Yan, Li, and Sha 2011). Disease-associated variants of Bap include missense mutations in the N- and C-termini impairing the interaction with BiP underscoring its role in ER homeostasis (Yan, Li, and Sha 2011).

Grp170 stimulates the exchange of ADP for ATP on BiP. It shares a similar structure to that of the Hsp70 family as it consists of an N-terminal NBD and a C-terminal SBD (Andréasson et al. 2010). The NBD domain of Grp170 interacts directly with the NBD of BiP, opening the nucleotide binding pocket to facilitate nucleotide exchange (Andréasson et al. 2010). The SBD of Grp170 has a C-terminal extension as well as a small unstructured loop in the ®-rich region (Andréasson et al. 2010). Both of these additional domains modulate the binding of Grp170, though the exact purpose of substrate binding by Grp170 remains elusive (Andréasson et al. 2010; Behnke and Hendershot 2014). Recent work has proposed that Grp170 acts as an aggregation deterrent, binding to less common, more aggregation prone epitopes rich in aromatic residues as compared to that of BiP (Behnke et al. 2016).

Terminal release from the BiP binding cycle occurs when the protein has either achieved its native mature fold and no longer presents interaction motifs, or the protein is deemed terminally misfolded. For irreparable substrates, exit from the BiP binding cycle is facilitated by the delivery of the misfolded protein to downstream ER-associated degradation (ERAD) machinery (Ushioda, Hoseki, and Nagata 2013; Pobre, Poet, and Hendershot 2019). When unstressed, the ER has two pathways for ERAD entrance, one though persistent BiP association of non-glycosylated proteins, and the other through the calnexin/calreticulin pathway for glycosylated proteins (Lamriben et al. 2016). However, the two distinct pathways become interchangeable when the cell is under stress, leading to BiP facilitated binding and subsequent degradation of glycosylated proteins as well (Ushioda, Hoseki, and Nagata 2013). ERAD machinery retro-translocates the BiP-associated cargo out of the ER where it is degraded by the ubiquitin-proteasome system. This release can be facilitated by J-domain proteins including ERdj5 contributing to their diverse roles in regulating the functions of BiP in the ER lumen.

BiP modification

Under non-stressed growth conditions, BiP is post-translationally modified to sequester an inactive store of the protein for immediate deployment and activation upon signaling. The AMPylation of BiP by FICD (also known as HYPE) maintains an inactivated pool of BiP under normal conditions (Preissler, Rohland, et al. 2017; Perera et al. 2019). AMPylation on T⁵¹⁸ inactivates BiP by biasing it towards a conformation that mimics the ATP bound state (Preissler, Rohland, et al. 2017). When client population is low, BiP can also be modulated by forming a multimeric state, which further contributes to creating an inactive pool of BiP. If the client load drastically increases, FICD can also facilitate the de-AMPylation of BiP and rapidly revert the chaperone to its functional state on a timescale faster than transcriptional regulation (Preissler et al. 2015; Perera et al. 2019; Preissler, Rato, et al. 2017). An oligomeric switch for FICD has been implicated as the important regulatory mechanism for promotion of FICD AMPylation or de-AMPylation activity, as dimeric FICD functions to de-AMPylate BiP, whereas monomeric FICD acts to AMPylate BiP (Perera et al. 2019). FICD-mediated AMPylation is not limited to BiP as a global proteomics analysis identified 25 novel AMPylated substrates, suggesting that this regulatory method may be employed more widely (Broncel et al. 2016).

BiP is the canonical Hsp70 chaperone of the ER that directly contacts non-native substrates through protein-protein interactions. These substrate protein sequences are often aggregation-prone as BiP acts to bind exposed hydrophobic sequences. This can allow the rest of the protein to fold correctly or if terminally misfolded, the substrate may be delivered to downstream ERAD machinery for eventual turnover.

CALNEXIN AND CALRETICULIN BINDING CYCLE

The majority of ER-targeted proteins are modified with at least one N-linked glycan, a carbohydrate group which plays multiple roles in ER protein maturation, quality control and trafficking. N-linked glycans are bulky and hydrophilic, significantly altering the interaction of a modified substrate with the surrounding environment, promoting solubility, productive folding, altering protein function, and acting as a reporter of glycoprotein folded status (Hebert et al. 2014). The role of the N-linked glycan in acting as a reporter of glycoprotein folding and age is key to glycoprotein quality control and proteostasis in the ER. N-linked glycans are dynamically modified during the folding and maturation pathway of the glycoprotein, and the glycoform promotes interaction with carbohydrate-binding proteins in the ER, altering the folding, trafficking, or degradation pathway of a glycoprotein (Adams, Oster, and Hebert 2019). At the center of glycoprotein quality control are the lectin chaperones calnexin and calreticulin and their associated proteins that act to promote proper glycoprotein folding through pro-folding interactions and the retention of non-native glycoproteins for repair or degradation targeting.

Calnexin/calreticulin organization and structure

Calnexin and calreticulin share about 39% identity and play similar, though not identical, roles in glycoprotein quality control. Both are comprised of a signal sequence for ER targeting, an N-terminal lectin domain, a P-rich domain (P-domain), and a C-terminal domain. Calnexin possesses a transmembrane domain while calreticulin is soluble ER luminal protein.

The N-terminal cleavable signal sequence of both calnexin and calreticulin directs delivery to the ER and the type-I topology for calnexin (Ou et al. 1995). The lectin domains of both proteins adopt a globular fold composed of a beta-sandwich with two anti-parallel beta sheets, which comprise a single carbohydrate binding site. Crystal structures of both lectin domains demonstrate a binding site for calcium, potentially promoting stability and lectin activity (Schrag et al. 2001; Kozlov et al. 2010). The luminal domains of calnexin and calreticulin also possess an ATP binding site which, upon phosphorylation, may lead to destabilization and decreased lectin activity (Ou et al. 1995; Wijeyesakere et al. 2015).

Extending from the lectin domain is the P-domain. The P-domain is comprised of sequential Pro-rich motifs, with calnexin containing four repeats of motif 1 (17 amino acids) followed by four repeats of motif 2 (14 amino acids), and calreticulin containing three repeats each of motif 1 and 2. Between the two sets of motifs is a hairpin turn, leading to the P-domain adopting an extended, hook-like fold (Ellgaard et al. 2001; Schrag et al. 2001). Multiple folding factors interact with calnexin and calreticulin in the region of the hairpin turn, which will be discussed in more detail later.

The C-terminal region is the most dissimilar region between calnexin and calreticulin. Calnexin contains a C-terminal region composed of a transmembrane domain and a cytoplasmic tail, which contains sites for phosphorylation and palmitoylation, as well as a di-R sequence that promotes ER retention (Rajagopalan, Xu, and Brenner 1994). Calreticulin has a C-terminal tail containing acidic residues, which provide low affinity calcium binding sites, therefore supporting the role of calreticulin as a more robust ER calcium buffer as compared to calnexin (Lamriben et al. 2016). ER retention of the soluble protein calreticulin is supported by its N-terminal K-D-E-L motif.

Calnexin/calreticulin binding specificity

The lectin domains of calnexin and calreticulin support binding to glycoproteins that possess monoglucosylated glycans presented on substrates (Hebert et al. 2014) (Figure 2). Crystallographic evidence has demonstrated that the lectin domain of calreticulin interacts with the A-branch tetrasaccharide Glcα1–3Manα1–2Manα1–2Man (Kozlov et al. 2010), while isothermal calorimetry (ITC) studies have shown that calreticulin has low micromolar affinity for the A-branch tetrasaccharide Glc₁Man₃ with decreasing affinity for Glc₁Man₂ and Glc₁Man₁ (Kapoor et al. 2003). The B- and C-branch mannoses do not appear to be required for interaction with calnexin and calreticulin, as removal of these residues does not abrogate interaction with calnexin. However, the B/C branch point mannose was found to be required for glycan interaction with calnexin and calreticulin, despite this mannose not being present in ITC experiments demonstrating calnexin and calreticulin lectin activity (Spiro et al. 1996; Vassilakos et al. 1998). More work is therefore required to determine the precise mannose composition required for interaction with calnexin and calreticulin, though it is clear a monoglucosylated glycan is necessary.

Figure 2. The calnexin and calreticulin cycle.

Nascent glycoproteins possess N-linked glycans with a Glc₃ Man₉ GlcNAc₂ structure. The terminal glucoses are sequentially processed by α-glucosidase I (PDB 4J5T) and α-glucosidase II (PDB 5F0E; Catalytic GRH31 domain, green). In the monoglucosylated state, the substrate is engaged by calnexin (CNX) or calreticulin (CRT) (PDB 1JHN; lectin domain, yellow; P-domain, teal). Upon release from CNX/CRT, the terminal glucose may then be trimmed by α-glucosidase II. Folding to a native state promotes release from the cycle, while non-native substrates are reglucosylated by UGGT1/2 (PDB 5MZO; Folding sensor domain, blue; catalytic domain, red). Reglucosylation promotes re-engagement by CNX/CRT.

In addition to lectin-based interactions, calnexin and calreticulin display protein-protein based interactions as well, though to a lesser degree. Cell-based experiments have found that calnexin can interact with some substrates lacking monoglucosylated glycans (Danilcyk and Williams 2001). Disruption of protein-protein interaction sites mapped to the lectin domain led to a small decrease in chaperone function, while impairment of lectin activity produced a severe impact on this chaperone function demonstrating that the carbohydrate binding site is the predominant site responsible for substrate binding (Lum et al. 2016).

Delivery to the calnexin/calreticulin cycle

Calnexin and calreticulin can engage substrates both co-translationally as well as post-translationally in mammalian cells. Upon translocation into the ER through the Sec61 channel, glycoproteins are modified with preassembled N-linked glycans as Glc₃Man₉GlcNAc₂ transferred en bloc from a dolichol phosphate in the ER membrane to the Asn site in the acceptor sequence N-X-S/T/C (where X is not a P) via the oligosaccharyltransferase (OST) complex. In mammals, there are two forms of the multimeric OST complex: one that contains the catalytic domains STT3A that transfers glycans co-translationally assisted by its proximity to the Sec61 translocon, and a second that contains STT3B that transfers N-glycans post-translationally (Cherepanova et al. 2019). Interestingly, Saccharomyces cerevisiae only possess the STT3B containing post-translational modification complex (Shrimal et al. 2019). The mode of glycan addition significantly impacts the temporal nature of glycan addition and initial lectin chaperone binding.

After glycosylation, the terminal glucose is rapidly trimmed off the glycan by glucosidase I, generating a glycan with two glucoses. Glucosidase I is a type I membrane protein requiring this initial trimming event to take place proximal to the membrane. The next step in the calnexin/calreticulin cycle is the trimming of the second glucose by glucosidase II, generating a monoglucosylated glycan. This glycoform can then be recognized by calnexin or calreticulin. Glucosidase II is a heterodimeric complex comprised of an α-domain that contains the catalytic domain and is highly expressed, and a β-domain that contains the ER retention sequence. Glucosidase IIβ also contributes to substrate recognition as it contains a mannose 6-phosphate receptor homologous (MRH) domain that supports binding to terminal mannose residues. Therefore, it helps to recruit the glucosidase II dimer to glycan rich regions for rapid trimming and generation of the monoglucosylated substrate (Deprez, Gautschi, and Helenius 2005).

As glycosylation can occur co-translationally in mammals, co-translational interaction with the calnexin/calreticulin cycle is therefore possible (Daniels et al. 2003). This however is dependent on the location of the glycan on the substrate, as C-terminal glycosylation would likely not support co-translational interaction with calnexin and calreticulin because translation would be terminated shortly after glycosylation occurs. As calnexin and calreticulin has been demonstrated to interact co-translationally with substrates, any delay in glycoprotein entry into this cycle due to the requirement of glucose trimming by glucosidases I/II is brief. Since glucosidases I/II do not appear to actively select substrates based on folding status, this initial entry step into the calnexin/calreticulin cycle appears to be promiscuous.

Calnexin/calreticulin substrate release and rebinding

As calnexin and calreticulin require a monoglucosylated glycan for substrate recognition, glucose trimming of the final glucose by glucosidase II represents a release of the substrate from the calnexin/calreticulin cycle. It is notable that unlike BiP that has different cofactors that support delivery and release, glucosidase II acts to both deliver and release substrates from the calnexin/calreticulin cycle (Table I).

Table I.

Comparison of BiP and carbohydrate-binding chaperone networks.

	BiP	calnexin/calreticulin
Binding specificity	extended hydrophobic domains	Glc1ManxGlcNAc2
Binding regulation
chaperone	ATP/ADP
substrate		carbohydrate composition
Delivery
chaperone	ERdj1-7
substrate	ERdj3-6	glucosidase II
rebinding		UGGT1/2
Release
chaperone	Grp170 Bap/Sil1
substrate		glucosidase II
Additional associated factors	EDEM1 (ERdj5)	PDIA3/PDIA9/cyclophilin B Sep15 (UGGT1/2)
Modifications	AMPylation	phosphorylation (calnexin) palmitoylation (calnexin)

After trimming of the final glucose, the glycoprotein has two general fates, depending on the folded status of the protein. If the protein has reached its native fold and oligomeric state, it can exit the ER to traffic to its proper destination (Hebert et al. 2014). Alternatively, non-native glycoproteins can be recognized by the folding sensor uridine diphosphate-glucose: glycoprotein glucosyltransferase I (UGGT1) and reglucosylated, placing the glycan back in a monoglucosylated state. Reglucosylation supports glycoprotein rebinding to calnexin and calreticulin, promoting further rounds of intervention and retention of non-native glycoproteins.

UGGT1 is suggested to preferentially recognize near-native substrates over terminally misfolded substrates, though how this fine-tuned selectivity is achieved is poorly understood (Caramelo et al. 2004; Adams et al. 2019). UGGT1 is the central gatekeeper of the calnexin/calreticulin cycle, recognizing non-native glycoproteins that should be retained in the ER as these may be non-functional or aggregation prone, and therefore potentially deleterious to cellular health. Therefore, the initial delivery (glucosidases I and II) and re-delivery (UGGT1) to the calnexin/calreticulin cycle are roles played by separate factors, with glycoprotein folding status playing the determinant role in persistent lectin chaperone binding.

UGGT1 has a paralogue, UGGT2, with 55% sequence identity, which has been shown to be active in vitro, though no cellular substrates have been identified (Takeda et al. 2014). Both UGGT1 and UGGT2 are soluble, ER-resident proteins with similar domain organization, consisting of an N-terminal folding sensor domain and a C-terminal glucosyltransferase domain. Recent crystal structures of Thermomyces dupontii and Chaetomium thermophilum UGGT1 have demonstrated that the N-terminal folding sensor is composed of four thioredoxin domains that lack catalytically active C-X-X-C motifs (Satoh et al. 2017; Roversi et al. 2017). These thioredoxin domains likely play a role in substrate recognition. The thioredoxin domains are orientated into a curved and flexible structure with a large central cavity possessing hydrophobic patches, suggesting a binding region for non-native substrates. The folding-sensor domain is more dissimilar between UGGT1 and UGGT2, displaying 49% identity, as compared to 84% identity between the C-terminal glucosyltransferase domains. This suggests that substrates may not completely overlap, but UGGT1 is expected to carry the bulk of the quality control load as it is expressed at a 25-fold higher level in HeLa cells than UGGT2 (Itzhak et al. 2016). Further studies are needed to delineate substrate specificities of UGGT1 and 2, and their differential roles in glycoprotein folding and quality control.

Terminal release of substrates from calnexin/calreticulin binding cycle

Persistent calnexin and calreticulin interaction may occur as long as non-native substrates are recognized and reglucosylated by UGGT. However, to clear the ER of terminally misfolded glycoproteins, there are a number of possible mechanisms in place for extracting irreparable glycoproteins from the carbohydrate-dependent chaperone cycle. These mechanisms use two general concepts: (1) the recognition or pass-off to downstream machinery for ERAD targeting; and (2) the alteration of the substrate N-glycan so it is no longer a suitable structure for reglucosylation. These mechanisms are not mutually exclusive as they have a number of overlapping components.

One early model for extraction is based on observations that misfolded substrates can interact with ER degradation-enhancing alpha-mannosidase-like protein 1 (EDEM1) (Oda et al. 2003; Molinari et al. 2003). This model is supported by the observation that a misfolded disease associated variant of alpha-1-antitrypsin was passed directly from calnexin to EDEM1, possibly assisted by the association of the transmembrane domains of calnexin and EDEM1 (Oda et al. 2003). Studies using a misfolded variant of BACE (BACE457) found that EDEM1 expression accelerated its deglucosylation and turnover (Molinari et al. 2003). These studies demonstrated a role for EDEM1 in extracting aberrant glycoproteins from the lectin chaperone binding cycle for subsequent targeting for dislocation to the cytosol and eventual proteasomal degradation.

The second general mechanism for removal of substrates from the calnexin/calreticulin binding cycle involves rendering the substrate unsuitable for reglucosylation. Human cells contain seven possible exo-mannosidases (MAN1a1/1a2/1b1/1c1 and EDEM1–3) and a single endomannosidase (MANEA) that might be responsible for trimming substrate glycans to make them poor substrates for UGGT. Removal of the terminal mannose on the A-branch could diminish reglucosylation, and ultimately reentry into the cycle, by creating a poorer substrate for UGGT (Avezov et al. 2008). Trimming of mannoses from the B- and C-branches also appears to diminish recognition by the UGGTs (Sousa, Ferrero-Garcia, and Parodi 1992). Alternatively, the trimming of α1,2-linked mannose residues exposes α1,6-linked mannose residues on the C-branch, which have been shown to recognized by downstream ERAD receptors Os-9 and XTP3B to support accelerated degradation of aberrant clients (Roth and Zuber 2017). Future studies will be needed to delineate the roles of the mannosidases in extraction of terminally misfolded proteins from the calnexin/calreticulin cycle. As some of these mannosidases are localized to ER subcompartments, vesicles or the Golgi, the recycling or trafficking of substrate or mannosidases might play an important role in substrate targeting for degradation.

Calnexin/calreticulin-associated factors

Calnexin and calreticulin associate with several co-chaperones in order to promote productive substrate folding. Interaction with co-chaperones is mediated via the P-domain (Kozlov et al. 2017). The P-domains of calnexin and calreticulin are structurally similar, though calnexin possesses four Pro-rich motifs while calreticulin possesses three, shortening its reach. The P-domain of calnexin and calreticulin functions as a binding site for three known co-chaperones: the protein disulfide isomerase (PDI) family members PDIA3 (ERp57) and PDIA9 (ERp29), and the peptidyl cis trans protein isomerase (PPI) family member cyclophilin B (CypB). Therefore, these lectin chaperones act as a platform to recruit foldases or protein folding catalysis to non-native nascent chains (Kozlov and Gehring 2020).

PDIA3 is an ER resident member of the protein disulfide isomerase family, which promotes native disulfide bond formation of substrates. It consists of four thioredoxin domains, a, b, a’, and b’, with a and a’ possessing catalytically active C-X-X-C motifs (Frickel et al. 2004). The PDIA3 binding site on calnexin is located on the hair-pin turn of the P-domain (Frickel et al. 2002), interacting with a positively charged patch on the b’ domain of PDIA3 (Kozlov et al. 2006). A similar binding orientation has been shown by cryo-EM in the context of the peptide loading complex, a multi-protein complex including calreticulin and PDIA3 in which MHC class I is loaded with peptides for cell surface antigen presentation (Blees et al. 2017). Notably, PDIA3 has been shown to display increased oxidoreductase activity when associated with calnexin or calreticulin (Zapun et al. 1998). This increased activity can likely be attributed to the increased local concentration of PDIA3 and substrates when both are engaged by calnexin or calreticulin, highlighting the role of calnexin and calreticulin as both chaperones and scaffolds wherein substrates can be placed in close proximity to multiple chaperone co-factors.

PDIA9, though a member of the PDI family, does not possess a full catalytically active thioredoxin site and functions as a dimer, with the thioredoxin-like domain mediating dimerization (Liepinsh et al. 2001). Its active site contains a single C residue that is required for its ability to assist trafficking of substrates, such as insulin, but is not required for client interactions or disulfide formation or isomerization (Viviano et al. 2020). Calnexin and calreticulin interact with PDIA9 at the tip of the P-domain, in a similar region to the PDIA3 interaction domain, via a hydrophobic pocket in the C-terminal D-domain of PDIA9 (Kozlov et al. 2017).

CypB is a predominantly ER resident member of the PPI family, which has been suggested to have a pro-folding function especially important for collagen folding (Terajima et al. 2017). CypB also binds to the tip of the P-domain of calnexin and calreticulin (Kozlov et al. 2017). It is therefore notable that all known associated factors of calnexin and calreticulin bind to a similar region of the P-domain, which acts as a mechanism to allow a glycoprotein substrate to be engaged by multiple folding factors in an efficient manner.

In addition to the co-chaperones of the calnexin/calreticulin cycle, UGGT1/2 engages the co-factor Selenoprotein F (Sep15) within the ER. Sep15 is a 15-kDa protein containing a selenocysteine (Sec, U) encoded by a UGA codon (Kasaikina et al. 2011). The mRNA for all known selenoproteins contain a unique stem-loop structure, known as a Sec insertion sequence (SECIS), that prevents the termination of translation and instead inserts a Sec at the UGA position (Kryukov et al. 2003). Immediately C-terminal to the cleaved signal sequence is a C-rich domain with six well conserved C. Furthermore, there is a highly conserved C-G-U-K motif that is the predicted redox-active site for Sep15 (Korotkov et al. 2001). Interestingly, Sep15 does not contain an ER retention signal at its C-terminus, suggesting it is retained by some other means. Immunoprecipitation of overexpressed mouse Sep15, in the presence of overexpressed UGGT1/2, has shown that Sep15 co-precipitates and forms a 1:1 complex with both UGGT1 and UGGT2. Mutagenesis of the N-terminal C-rich domain revealed that all six C are essential for its interaction with UGGT1/2. Notably, mutating U to C does not ablate interactions between Sep15 and UGGT1 (Labunskyy et al. 2005). Of note is that all of Sep15 is found in complex with UGGT1/2 and does not rely on intermolecular disulfide bonds to maintain the interaction (Korotkov et al. 2001). It is therefore likely that, Sep15 is retained in the ER through its interaction with UGGT1 and UGGT2. Sep15 has been shown to increase the ability of UGGT1 and UGGT2 to reglucosylate a synthetic target in vitro (Takeda et al. 2014). Further investigations are needed to understand how Sep15 contributes to the calnexin/calreticulin cycle in a cellular context.

Calnexin/calreticulin modification

Three major modifications of calnexin have been described that target its cytosolically localized C-terminal tail: palmitoylation, phosphorylation, and SUMOylation. The sites of palmitoylation have been identified as two C residues in the C-terminal tail of calnexin. Its palmitoylation has been suggested to either promote interactions between calnexin and the ER translocon complex or to enrich calnexin at ER mitochondrial-associated-membrane sites (Lynes et al. 2012; Lakkaraju et al. 2012). Phosphorylation of calnexin has been shown to occur at three sites in the C-terminal tail, mediated primarily by two cytosolic kinases, CK2 and ERK1 MAPK (Chevet et al. 2010). Phosphorylation at S⁵⁶³ recruits calnexin to ribosome engaged translocons as well as promotes prolonged interaction with the model secretory pathway protein alpha 1-antitrypsin under misfolding conditions (Cameron et al. 2009). Phosphorylation of S^534/544 appears to promote the redistribution of peripheral ER calnexin to a perinuclear localization, potentially through disrupted interaction with the cytosolic protein PACS-2 (Myhill et al. 2008). UBC9-dependent SUMOylation of calnexin at L⁵⁰⁶ in the cytosolic C-terminal tail has been demonstrated to support interaction with protein tyrosine phosphatase 1B (PTP1B) (Lee et al. 2015). While PTP1B is primarily targeted to the ER via a C-terminal transmembrane region, the SUMO-dependent interaction between calnexin and PTP1B promotes the localization of PTP1B to the cytosolic face of the ER membrane while potentially altering the pool of substrates which PTP1B can engage. Modifications of the calnexin C-terminal tail therefore act to alter the localization of calnexin and binding partners and consequently may modify the functional role of calnexin, but the mechanism of these processes requires further characterization.

CONCLUDING REMARKS

The BiP and calnexin/calreticulin chaperone pathways aid in the folding and quality control of a wide range of cargo that traverse the eukaryotic secretory pathway using diverse mechanisms for the selection and the regulation of binding to cargo. The Hsp70 ER family member BiP binds to exposed hydrophobic motifs that are normally sequestered in the core of a properly folded protein. BiP binding is regulated by its adenine nucleotide binding state. In contrast, calnexin and calreticulin binding is dictated by covalent changes to the substrate itself or by controlling the N-glycan composition of cargo protein. These distinct pathways provide a broad range of assistance for the thousands of different client proteins of the secretory pathway.

These two chaperone systems use opposing methods for substrate recognition and binding. BiP, like other Hsp70 family members, binds to short stretches of exposed hydrophobic residues that should be hidden in the core of the folded proteins. In converse, the lectin chaperones bind to large hydrophilic modifications on the client that generally remain exposed to the aqueous environment on the natively folded protein. How can two such dissimilar mechanisms for binding be used to select proteins that require assistance for folding?

The exposure of hydrophobic residues is a ubiquitous hallmark of a misfolded protein and it is directly utilized by the Hsp70 system regardless of the location in the cell to mediate substrate recognition and binding. While binding to calnexin and calreticulin is through a monoglucosylated protein, there are two different mechanisms to support binding. Binding commences after the recently transferred N-glycan is rapidly trimmed sequentially by glucosidases I and II to generate the monoglucosylated side chain. This initial binding event, therefore, is controlled by the timing of glycan addition by the OST and its subsequent trimming shortly thereafter. As glycans can be added co-translationally in mammals, this allows for early intervention of the lectin chaperones in the folding process. If the trimming of the second and third glucose that are both released by glucosidase II is rapid, it is likely that lectin chaperone binding to this glycan would be bypassed. Rebinding or additional rounds of lectin chaperone binding are mediated by UGGT1, and likely UGGT2. These proteins use their folding sensor domain comprised of four thioredoxin domains to recognize non-native proteins likely through exposed hydrophobic segments. Therefore, while the mechanisms of binding differ greatly for both chaperone systems, both systems share that persistent chaperone binding is at least in part based on exposed hydrophobic motifs. BiP scans for these motifs directly to mediate binding, while UGGT does the scanning for the lectin chaperones and then modifies the glycan signal accordingly to direct chaperone binding.

The ER is a large organelle that can spread throughout the cytosol of many cells and perform diverse functions. The seven J-proteins of the ER help to direct BiP to various location in the ER, as well as to a variety of functions. For example, ERdj2 (yeast Sec63) directs BiP to the translocon that resides in the ER membrane to assist in protein translocation and position it for early engagement and assistance for nascent chains as they emerge into the ER lumen (Pobre, Poet, and Hendershot 2019). ERdj5 helps with preparation and potential targeting of misfolded proteins for ERAD. In contrast, the lectin chaperones diversify their localization by having two paralogues, one membrane associated and the other soluble bypassing the need to recruit a chaperone to the membrane. Post-translational modification of the cytoplasmic C-terminal tail of calnexin appears to be involved in its localization. While BiP is recruited and controlled by its J-domain proteins to dictate its function, the lectin chaperones in some respects act as a platform to recruit diverse folding factors to the non-native substrate for modification of C residues (PDIA3 and PDIA9) or P orientation (CypB) (Kozlov and Gehring 2020).

Both chaperone systems appear to act as holdases or folding factors that protect and delay the folding of the region of the protein they are bound to thereby controlling the folding trajectory of the nascent chain (Daniels et al. 2003; Wang et al. 2008; Mayer and Gierasch 2019). This results in a higher yield of properly folded protein or more efficient folding by which the rate of folding can either be accelerated or delayed. In addition, persistent binding of both systems to non-native clients supports ER retention and subsequent targeting for degradation. The binding mechanisms of BiP and calnexin/calreticulin vary; however, their functions in protein folding and quality control have a number of important overlapping roles.

While the lectin chaperone system works largely on glycosylated client proteins, there has been some evidence for it interacting with non-glycosylated targets though its biological significance is poorly understood. Glycoprotein folding and quality control can be helped by both BiP and calnexin/calreticulin. The timing for initial chaperone binding is dependent upon the location of the glycans on the protein as calnexin/calreticulin binding appears to be dominant over BiP binding. It was found that prior binding to BiP occurred if there was no N-glycan in the first 50 amino acids for co-translational translocating nascent chains (Molinari and Helenius 2000). However, once a glycosylation site emerged into the ER lumen, if it was modified and rapidly trimmed to support lectin chaperone binding that this binding would dominate over BiP binding. The analysis of the primary sequence of a protein using algorithms that search for BiP binding sites and N-glycosylation sites can provide important insights into the timing of chaperones interactions with the nascent chain; however, empirical demonstration is required as substrate features do not always successfully predict which quality control systems will be utilized (Adams et al. 2019).

The ER serves as the site for the maturation of a large array of proteins. While small proteins likely fold efficiently unassisted by chaperones, larger proteins that fold and frequently assemble into multiprotein complexes in the ER are expected to be accommodated by region-specific engagement by BiP and calnexin/calreticulin along with their associated factors. These complex proteins that are reliant upon chaperones for their proper maturation and sorting are commonly the source of diseases rooted in misfolding underscoring the importance in understanding how chaperones direct the folding and trafficking of proteins in the early secretory pathway.

SUMMARY POINTS.

1. The binding of the ER Hsp70 family member BiP to non-native proteins is regulated by adenine nucleotides and its associated cofactors (J-proteins for delivery and NEFs for release).

2. The lectin chaperones bind to proteins possessing monoglucosylated N-linked glycans, therefore binding is regulated by glucosidases and transferases that dictate the glycan composition.

3. Both chaperone systems bind to non-native substrates; however, BiP binds to exposed hydrophobic motifs that should be sequestered by a properly folded protein and the lectin chaperones bind to exposed bulky hydrophilic modifications.

4. The classical and lectin chaperones prevent aggregation thereby increasing the properly folded protein yield.

5. Persistent chaperone binding supports ER retention and subsequent delivery of terminally misfolded proteins for degradation.

FUTURE ISSUES.

1. The identification of BiP and calnexin/calreticulin co-translational and post-translational binding sites using proteomic and ribosomal profiling approaches.

2. The determination of cellular UGGT1 and UGGT2 substrates and modification sites.

3. To understand the effects of the modifications of the C-terminal tail of calnexin on its cellular function.

4. To investigate the mechanism of extraction from the calnexin/calreticulin cycle by the variety of exo- and endo-mannosidases.

5. Are substrates engaged by the BiP and lectin chaperone cycles simultaneously or in a dedicated manner?

6. What diseases states or therapeutic approaches can be addressed through modifications of the ER chaperone cycles?

ABBREVIATIONS

ER

Endoplasmic Reticulum

BiP

Binding-immunoglobulin Protein

ERdj

ER-localized DNAJ

SBD

Substrate Binding Domain

NBD

Nucleotide Binding Domain

TPR

Tetratrico Peptide Repeat

AMP

adenosine 5’-monophosphate

ATP

Adenosine Triphosphate

CNX

Calnexin

CRT

Calreticulin

UGGT1(2)

UDP-glucose:glycoprotein glucosyltransferase I (II)

PDIA3

Protein Disulfide-Isomerase A3

PDIA9

Protein Disulfide-Isomerase A9

CypB

Cyclophilin B

Sep15

15 kDa selenoprotein