Bring it back, bring it back, don't take it away from me – the sorting receptor RER1

The endoplasmic reticulum (ER) is the largest membranous organelle in the eukaryotic cell and constitutes the first station of the secretory pathway. Almost one-third of the proteome comprises secreted soluble and membrane proteins that need to be co-translationally translocated into the ER (Ghaemmaghami et al., 2003). Putative cargo first undergoes extensive primary quality control mediated by the COPII machinery, to ensure that only correctly folded and post-translationally modified proteins are sorted for the ER (Barlowe and Helenius, 2016; Zanetti et al., 2012). Cargo destined for export to distal parts of the secretory pathway is sorted away from ER-resident proteins, whereas misfolded proteins are destined for degradation via the ER-associated degradation pathway (ERAD), involving the proteasome (Barlowe and Helenius, 2016). This mechanism, therefore, restricts the secretion of non-native protein conformations to downstream secretory environments. Imbalance in ER quality control results in accumulation of proteins that are incorrectly folded, which can lead to ER-stress. This imposes a threat to secretory proteostasis, the branch of the proteostasis network that – at equilibrium – keeps translation, folding and trafficking out of and into the ER. Against threats to secretory proteostasis, cells have developed a stress-response signaling pathway, the so-called unfolded protein response (UPR) (Plate and Wiseman, 2017).

Despite the efficient quality control and sorting mechanisms of the ER, proteins sometimes escape and are erroneously transported to the Golgi complex, where they are separated from ‘real’ cargo and brought back to the ER as part of a secondary ER quality control mechanism. This might be required, for instance, during the assembly of multi-subunit complexes, upon which disturbances in transcription, translation or folding of single subunits may lead to a temporary non-stoichiometric imbalance in individual subunit levels. Such an imbalance can increase the number of any surplus subunits escaping out of the ER and retrieval pathways will help to re-synchronize assembly processes. Retrieval of escaped proteins is often mediated by specific sorting receptors that bear specific ER-retrieval signals. ER-retrieval signals are also present on ER-resident proteins, as, for example, the lysine–aspartic acid–glutamic acid–leucine (KDEL)-sequence on chaperones, such as immunoglobulins and protein disulfide-isomerase (Munro and Pelham, 1987). Interestingly, retrieval signals are sometimes also found in proteins that are only supposed to reside temporarily in the ER. Such proteins are often part of the export or vesicle machinery and, after retrieval, are re-used in the next round of export. An example for this are the p24 family of cargo receptors (Nickel and Wieland, 1998; Pastor-Cantizano et al., 2016), the lectin ERGIC53 (LMAN1) (Hauri et al., 2000) and the KDEL-receptor (Cancino et al., 2013). In other cases, as mentioned above, short-term ER-residents are subunits of multimeric complexes, such as ion channels or cell-surface receptors (Schwappach, 2008). The ER-retrieval signals identified so far are short sequences, such as KKxx [with x denoting any amino acid (aa)] or KDEL (reviewed by Nilsson and Warren, 1994) located at the C-terminus or, within cytoplasmic loops, the arginine (R) comprising sequence RxR (reviewed by Michelsen et al., 2005). Sometimes, i.e. in the case of the γ-aminobutyric acid (GABA_B) receptor subunit 1, such cytoplasmic ER-retrieval signals are exposed in unassembled subunits but obscured upon complex assembly – providing a mechanism referred to as ‘hide and run’ (reviewed by Michelsen et al., 2005). Retrieval information within transmembrane domains (TMDs) is emerging as yet another mechanism of secondary quality control in the regulation of transport between ER and Golgi (see Fig. 1).

In this Review, our focus is on RER1, thus far the only known receptor for such TMD-based signals. We review our current knowledge regarding molecular functions of RER1 in yeast and mammalian cells, discuss its role in vivo, and end with a critical outlook, discussing open questions and how they can be addressed.

RER1 was first described in 1993 in the yeast Saccharomyces cerevisiae in a screen for mutants defective in the ER-export associated protein Sec12 (Nishikawa and Nakano, 1993). It was termed RER1 by the authors because they were expecting mutants to be “defective in the retention in the ER or in the return from the early Golgi” (Nishikawa and Nakano, 1993). The gene was cloned independently in two labs and found to encode RER1, a four-TMD protein of 188 aa (Boehm et al., 1994; Sato et al., 1995; see Fig. 2A). RER1 localizes dynamically to the ER-Golgi intermediate compartment (ERGIC) and the cis-Golgi (Sato et al., 1995), and rapidly cycles between ER and Golgi with the help of the coat protein complex COPI (Boehm et al., 1997; Sato et al., 2001). For retrograde Golgi-to-ER transport, RER1 binds directly to COPI; this involves both a KKxx-like motif and two tyrosine (Y) residues located in its C-terminal tail (Sato et al., 2001), see Fig. 2A. The first identified RER1-dependent cargo proteins in yeast included Sec12 (Boehm et al., 1997; Sato et al., 2001), the guanine-exchange factor (GEF) of the small GTPase Sar1 that is required for the formation of COPII budding vesicles from the ER (d’,Enfert et al., 1991). Sato and colleagues also demonstrated that the RER1-binding motif is localized in the TMD of yeast Sec12 (Sato et al., 1996). Other cargoes are Sec63 and Sec71 (Sato et al., 1997), respectively, the essential and non-essential components of the protein translocation machinery at the ER membranes, and the putative COPII regulator Sed4 (Sato et al., 1996). Although topologically very different membrane proteins, their interaction with RER1 all rely on the presence of spaced polar residues within the hydrophobic core of their respective TMDs (Sato et al., 2003). In Δrer1 mutant yeast cells, these membrane proteins become mislocalized to the trans-Golgi or the vacuole (Boehm et al., 1997; Sato et al., 1997, 2003). On RER1, a Y residue in the fourth TMD is required for its interaction with Sec12 but not Sec71, suggesting that different binding modes exist for different ligands (Sato et al., 2003). The role of RER1 in ER quality control is not limited to retrieving escaped ER-resident proteins. RER1 also controls the assembly of oligomeric membrane complexes by retrieving unassembled subunits, as shown for Fet3, a subunit of the yeast iron transporter (Sato et al., 2004). Another cargo of RER1 with a retrieval signal within its TMD is the yeast α1,2-mannosidase Mns1 (Massaad et al., 1999; Massaad and Herscovics, 2001). A list of all RER1 substrates in yeast can be found in Table 1; see also Jurisch-Yaksi and Annaert, 2013).

In 1997 the human RER1 was cloned (Füllekrug et al., 1997). With 196 aa residues, the human RER1 protein is slightly longer than but highly similar (65%) to yeast RER1. The human RER1 gene functionally complements a RER1 deletion in yeast, suggesting its localization and molecular function are highly conserved in mammalian cells (Füllekrug et al., 1997). Sequence comparison of RER1 from different species reveals three highly conserved domains (Fig. 2). One, stretching from glutamic acid (E) residue 106 to tryptophan (W) 122 in the cytoplasmic loop between TMD2 and TMD3, is identical to that in yeast, fruit fly, zebrafish and mammals. The second conserved domain, i.e. from phenylalanine (F) 141 to F158, comprises the small lumenal loop between TMD3 and TMD4, and most of TMD4. This domain is also highly conserved, with 14 aa residues out of 18 being identical and an additional three similar between yeast and human RER1. This stretch includes a conserved Y residue, which is the binding site for the RER1 cargo substrate Sec12 (Sato et al., 2003). Binding sites for the other cargoes have not yet been determined. The third highly conserved domain, isoleucine (I) 160 to F177, is located within the cytoplasmic C-terminal domain. Here 12 out of 18 aa residues are identical, and another six similar between yeast and human.

With this knowledge, the stage was set for the identification of mammalian cargoes of RER1.

RER1 is expressed in all cells and tissues (see the Human Protein Atlas). A knockout of Rer1 in mice is early embryonic lethal already at the blastocyst stage (Hara et al., 2018; Valkova et al., 2011), thereby demonstrating an essential role of RER1 in mammalian development. In this section, we discuss the known cargoes of RER1 in mammalian cells and what is known about the function of RER1 in vivo.

Ten years after its initial cloning, two cargoes of RER1 in human cells were reported independently, first nicastrin (NCSTN, hereafter referred to as NCT) (Spasic et al., 2007) and, shortly after, presenilin enhancer 2 (PSENEN, hereafter referred to as PEN2) (Kaether et al., 2007). Interestingly, both these cargoes assemble together with presenilin 1 or 2 (PSEN1 or PSEN2) and an anterior pharynx-defective 1 isoform (APH1), to form the tetrameric γ-secretase complex, an intramembrane protease targeting more than 90 substrates, including Notch receptors and the amyloid-β precursor protein (APP) (Haapasalo and Kovacs, 2011; Jurisch-Yaksi et al., 2013b). Here, γ-secretase mediates the final cleavage of APP, thereby liberating amyloid-β peptides of different lengths, with longer more aggregation-prone ones precipitating in senile plaques in the brains of Alzheimers disease (AD) patients (Wolfe, 2020). For this reason, γ-secretase has attracted a lot of attention and its structural composition, biogenesis and function has been discussed in many reviews (see, for example, De Strooper et al., 2012; Escamilla-Ayala et al., 2020; Kaether et al., 2006; Kanatsu and Tomita, 2016; Spasic and Annaert, 2008; Wolfe, 2019).

Molecularly, it has been demonstrated that RER1 binds to the TMD of NCT, crucially depending on three polar aa residues that cluster on one side of the membrane-spanning α-helix and are similarly spaced as those of known yeast RER1 substrates (Spasic et al., 2007). The same residues in the TMD of NCT are crucial for assembly of the γ-secretase complex (Capell et al., 2003), as they bind to the APH1 subunit and, as such, RER1 competes with APH1 for binding to the TMD (Spasic et al., 2007). RER1 has, therefore, been proposed to act as a negative regulator of γ-secretase assembly and, consequently, cellular activity (Spasic et al., 2007). Support for this model comes from a later study that analyzed amyloid-β production in cells that overexpress all four γ-secretase components (Park et al., 2012). There, all four γ-secretase subunits interact with overexpressed RER1 reducing the surface localization of γ-secretase; knockdown of RER1, however, increases its surface transport, as well as amyloid production (Park et al., 2012). Likewise, the ubiquitin ligase synoviolin degrades RER1, and synoviolin knockout results in a decrease in surface levels and activity of γ-secretase, leading to reduced levels of amyloid-β (Tanabe et al., 2012).

In a different study, RER1 was shown to bind to the first TMD of PEN2 (Kaether et al., 2007), which contains a single asparagine (N) residue that is essential for the binding to RER1. Furthermore, RER1 bound to unassembled PEN2 that had not (yet) assembled into the γ-secretase complex. Accordingly, knockdown of RER1 resulted in increased cell-surface accumulation of unassembled PEN2, suggesting that RER1 is part of a quality control mechanism that allows fully assembled complexes to leave the ER (Kaether et al., 2007).

Both – Spasic et al. as well as Kaether et al. – described for the first time a function for mammalian RER1 in retrieving unassembled complex subunits (Spasic et al., 2007; Kaether et al., 2007). We further discuss our understanding of the molecular function of RER1 below.

A third cargo that has been described for mammalian RER1 is the skeletal muscle nicotinic acetylcholine receptor (nAChR) and, more specifically, its α1-subunits (nAChRα1) (Valkova et al., 2011). The nAChR is a pentameric ligand-gated ion channel composed of the α1, β1, δ, and ε subunits in a 2:1:1:1 stoichiometry (Millar, 2003). The first TMD of human nAChRα1 contains an asparagine-based motif that is responsible for its ER-localization in its unassembled form (Wang et al., 2002), and is very similar to the RER1-binding motif within PEN2 (Kaether et al., 2007). In fact, RER1 specifically interacts with unassembled nAChRα1; however, it neither interacts with the fully assembled pentameric nAChR nor with any of the other subunits (Valkova et al., 2011). The in vitro knockdown of RER1 results in the escape of unassembled nAChRα1 from the ER and subsequent degradation of nAChRα1, presumably in lysosomes (Valkova et al., 2011). In vivo, small hairpin RNA (shRNA)-mediated knockdown of Rer1 in mouse skeletal muscle reduced the amount of fully assembled nAChR and, consequently, the size of neuromuscular junctions (Valkova et al., 2011). Moreover, this study showed for the first time that loss of RER1 is embryonic lethal in mammals (Valkova et al., 2011).

Brain-specific deletion of RER1 results in early postnatal death (Valkova et al., 2017). However, deletion of RER1 in only the Purkinje cells of the cerebellum results in viable fertile mice without any obvious phenotype early in life and, moreover, development of the cerebellum – as indicated by the number and arborization of Purkinje cells and the thickness of the molecular layer of the cerebellum – is normal (Valkova et al., 2017). Since expression of the Purkinje cell-specific Cre recombinase transgenic line results in deletion of RER1 in Purkinje cells during postnatal week 2, which was completed at postnatal day 16, a role for RER1 in neuronal development cannot be inferred from this model. However, older mice showed age-dependent motor deficits and loss of Purkinje cells (Valkova et al., 2017). Furthermore, electrophysiological and biochemical experiments demonstrated that Purkinje cells lost their ability to constantly generate action potentials owing to massive loss of the voltage-gated Na⁺-channel subunit Na_v1.6 (SCN8A) at the axon initial segment (Valkova et al., 2017). In a CNS-specific knockout in mice, loss of RER1 resulted in strong reduction in the levels of Na_v1.1 (SCN1A) and NA_v1.6 (Valkova et al., 2017). This suggests that, similarly to its role in AChR assembly, RER1 is also involved in the quality control of Na_v channel assembly, as its deletion results in the escape and degradation of unassembled or misfolded subunits. However, a direct binding of RER1 to Na_v channel subunits or a retrieval motif or binding region therein, have not been identified. Thus, the exact mechanism of RER1 function remains to be elucidated.

Recently, two publications described the role of RER1 in vivo, i.e. in murine neuronal stem cells (NSCs) from the ventricular zone of the mouse cortex (Hara et al., 2018; Kim et al., 2018). By using different technologies, apparent opposite conclusions were reached with regard to the function of RER1. Kim et al. showed that the protein tweety homolog 1 (Ttyh1) membrane protein maintains NSC stemness through binding of RER1 and destabilization of the latter. According to the authors, this results in increased γ-secretase activity and, consequently, increased Notch signaling, a key determinant for NSC stemness (Kim et al., 2018). In contrast, Hara and co-workers concluded that the loss of RER1 leads to a decrease of γ-secretase activity and, consequently, reduced Notch signaling and fewer NSCs (Hara et al., 2018). Whereas the former study used virally infected neurospheres in vitro as well as in-utero electroporation of viral shRNA and cDNA constructs (Kim et al., 2018), Hara and colleagues used a murine conditional forebrain-specific RER1 knockout (Hara et al., 2018). Whether the different conclusions are attributable to the different methodologies used – i.e. knockdown versus knockout – is a question that requires further analysis. In addition to the mouse models, the only other study of RER1 function in vivo was carried out in zebrafish (Jurisch-Yaksi et al., 2013a). Here, zebrafish RER1 was shown to be highly expressed in ciliated organs during development. Consequently, morpholino-mediated knockdown of zebrafish RER1 resulted in a prominent ciliopathy phenotype, including shortened primary cilia. This, in turn, resulted in impaired motile and sensory functions, reflected in defects regarding hearing, vision and left–right asymmetry. Furthermore, Hedgehog signaling, a cellular signaling pathway that needs intact cilia, was decreased (Jurisch-Yaksi et al., 2013a). Shortened cilia were also recapitulated upon RER1 knockdown in mammalian cells, such as porcine kidney and human retinal pigment epithelial cells, underscoring a conserved effect that could be attributed to a role of RER1 in negatively regulating γ-secretase activity, thereby, enhancing Notch signaling. It remains poorly understood how increased Notch signaling affects ciliogenesis. Notch downstream target genes, including zebrafish delta-like protein D (deltaD; officially known as Dld) or hairy-related 12 (Her12), may affect cellular differentiation processes and/or ciliogenesis program (Jurisch-Yaksi et al., 2013a; Lopes et al., 2010; Marcet et al., 2011; Tavares et al., 2017). In addition, interaction between Notch signaling and cilia master regulators has been reported (Boskovski et al., 2013; Tavares et al., 2017). As zebrafish RER1 is highly expressed in ciliated organs, disturbances in these regulations may explain some of the ciliopathy phenotype.

Taken together, although three studies clearly link RER1 to Notch signaling and the implications thereof for development, some data are conflicting and necessitate further studies for clarification.

Two additional substrates of mammalian RER1 have been described. One is rhodopsin, the pigment in photoreceptor cells (Hara et al., 2014; Yamasaki et al., 2014). Overexpression of either wild-type rhodopsin or of the rhodopsin G51R mutant – which carries a mutation within its first TMD, leading to accumulation of the protein within the ER and is found in patients diagnosed with retinitis pigmentosa – has been shown to be dependent on RER1 (Yamasaki et al., 2014). RER1 also interacts with the peripheral myelin protein 22 (PMP22), a protein mutated in inherited Charcot–Marie–Tooth (CMT) neuropathy (Hara et al., 2014). Similar to rhodopsin, ER localization of overexpressed wild-type PMP22 or of its neuropathy-causing L16P mutant, was shown to depend on RER1 (Hara et al., 2014). However, further work is needed to clarify the role of RER1 on either cargo, i.e. on endogenous rhodopsin and PMP22 within photoreceptor cells and Schwann cells, respectively.

On the basis of the above studies, we propose the following model for RER1 function (Fig. 3). RER1 retrieves unassembled subunits of heteromeric complexes from the cis-Golgi and returns them to the ER. Depending on the complex and its assembly location, RER1 can functionally contribute to maintain the stoichiometry of subunits within the ER and this may either increase or decrease the probability of final complex assembly. If the complex assembles in the ER, retrieval of escaped subunits from cis-Golgi increases their probability of assembly. However, if assembly preferentially occurs in the ERGIC or the cis-Golgi, retrieval of unassembled subunits to the ER decreases the probability of complex assembly. In the absence of RER1, unassembled subunits will no longer be retrieved but, instead, escape to later compartments of the secretory pathway. There, they are being recognized by secondary (Ellgaard and Helenius, 2003) or tertiary quality control mechanisms, such as the Golgi quality control (GQC) (reviewed by Sun and Brodsky, 2019), and are degraded as is, for example, the nAChRα1 (Valkova et al., 2011). If the complex assembles in the ER and in the absence of RER1, interaction between subunits is reduced, which, subsequently, results in reduced levels of fully assembled complexes at the plasma membrane. For complexes that assemble in the ERGIC or the cis-Golgi, absence of RER1 increases complex formation and, thus, results in increased numbers of fully assembled complexes at the plasma membrane.

Support for the latter scenario, in which loss of RER1 leads to more complexes at the plasma membrane, comes from studies of the γ-secretase subunit NCT (Spasic et al., 2007) and the RER1-binding protein Ttyh1 (Kim et al., 2018). In both cases, levels of the γ-secretase complex at the plasma membrane and, consequently, its activity, were increased upon reduction of RER1 levels.

The opposite scenario, whereby absence of RER1 leads to reduced complex levels at the plasma membrane, is also supported by findings regarding γ-secretase (Hara et al., 2018), as well as by data for AChR and voltage-gated Na⁺ channels (Valkova et al., 2011, 2017). In all these cases knockdown or knockout of RER1 results in decreased plasma membrane levels of the respective complex.

Our model for RER1 function thus provides a hypothetical framework for future studies to fully elucidate the impact of ER-retrieval of unassembled subunits on the physiological functions of the respective complex.

In this section we discuss some of the outstanding questions that, we feel, need to be addressed in future studies. They essentially relate to the molecular function of RER1, the range of cargoes that are retrieved by RER1 and, consequently, the overall physiological role of RER1, including in disease.

Thus far, there is no evidence that RER1 has a crucial role in the overall transport of secretory cargo. This is indicated by the two conditional knockouts described so far, as both show specific cellular phenotypes without any obvious impairment of general neuronal homeostasis – as expected from a global effect on the secretory pathway (Hara et al., 2018; Valkova et al., 2017). Rather, RER1 might target substrates more selectively in the light of maintaining their stoichiometry in the ER. Disturbing this balance might impact on processes that extend beyond the ER, depending on the nature of the substrate. An example is the role of Rer1 in multimeric complex assembly, as described above. An open question is also, whether mammalian Sec12 is a cargo of RER1. Sec12 has been implicated in the generation of autophagosomes (reviewed by Davis et al., 2017), and autophagy is increased in yeast and C. elegans RER1 mutants (Ghavidel et al., 2015). How absence of RER1 expression impacts autophagy in mammalian cells has not been studied, and the exact role of RER1 herein awaits further clarification.

RER1 probably targets a defined set of membrane proteins, some of which are known; however, probably many more await identification. The RER1 cargoes identified so far fall into two categories, ER-resident proteins that escaped the ER and need to be retrieved, and unassembled subunits of multimeric protein complexes that function in distal post-Golgi compartments. Other candidate cargoes that belong to either of these categories will, therefore, probably be identified. Determination of the entire cargo repertoire is needed to elucidate the physiological impact of defective RER1-mediated retrieval and quality control in the secretory pathway. At least the lethality observed in RER1 KO mice (Hara et al., 2018; Valkova et al., 2011) supports the notion that RER1 has an indispensable role in early development. At this point, however, no efforts have been reported to identify a comprehensive catalogue of RER1 cargoes.

From the results to date it appears that, in yeast, most – but not all – RER1 substrates possess a motif composed of spaced polar aa residues within one of their TMDs (Sato et al., 2003). This is also the case for mammalian NCT, which has similarly spaced polar aa residues that mediate RER1 binding (Spasic et al., 2007). By contrast, in PEN2 (Kaether et al., 2007) and nAChRα1 (Christina Valkova and C.K., unpublished observation), a single polar aa – asparagine (N) – mediates binding to RER1, whereas for other cargoes in mammalian cells the retrieval signal has not yet been identified. The apparent lack of a specific RER1-binding consensus motif aggravates database searches for potential candidates. However, with the availability of CRISPR-mediated gene targeting and conditional RER1 KO-mice (Hara et al., 2018; Valkova et al., 2017), new cargo proteins might be identified by comparing wild-type and knockout cells or tissues. In addition, interactomics studies that use classic co-immunoprecipitation or the more recently developed proximity labeling approaches combined with subsequent mass spectrometry analysis (Varnaite and MacNeill, 2016) may help to elucidate the full ‘RER1-ome’.

By contributing to ER residency or assembly of multimeric complexes, RER1 is – through its cargoes –indirectly relevant in disease. An example is its role in assembly of γ-secretase, an enzyme responsible for the production of amyloid-β peptides of different lengths, including longer aggregation-prone peptides. Also, interaction of RER1 with PMP22 and how RER1 differently affects ER retention of CMT disease-related mutant versus wild-type PMP22, potentially associates RER1 function to CMT. Interestingly, human RER1 is located on the short arm of chromosome 1 within the subtelomeric region that is deleted in the monosomy 1p36, which – with one in 5000 births – is the most commonly observed deletion syndrome (Jordan et al., 2015). Among other symptoms, patients suffer from intellectual disability, brain anomalies and seizures as well as characteristic craniofacial features (Jordan et al., 2015). Of note, one of the crucial regions regarding seizures was mapped to chromosome 1, nucleotides 2,045,453 - 2,622,423 (build GRCh38/hg19) and includes the RER1 locus (Zhu et al., 2013). Mice with a brain-specific deletion of Rer1 suffer from severe seizures before they die very early postnatally (Christina Valkova and C.K., unpublished observations). Although not studied yet, other clinical features observed in monosomy 1p36, such as hearing problems, heart abnormalities, cleft palate and others (Rosenfeld et al., 2010) might relate to ciliary defects, as observed upon rer1 silencing during zebrafish development (Jurisch-Yaksi et al., 2013a). In support for a crucial role of RER1 in 1p36 deletion syndrome, our search in the database for genome variation (http://dgv.tcag.ca/dgv/app/home) reveals no deletion in healthy individuals (Fig. S1A). This intolerance to gene variation is further confirmed in the clinical copy number variation (CNV) according to the DECIPHER database (Fig. S1B). Here, RER1 scores with a pLI≥0.9, indicating a high probability for RER1 being haploinsufficient, i.e. that a single copy RER1 is insufficient to maintain normal function. The pLI, i.e. the probability of being loss-of-function intolerant, is a score computed by the genome aggregation database (gnomAD), which indicates the probability that a gene is intolerant to a heterozygous loss-of-function (LoF) mutation.

Together, the potential contribution of RER1 haploinsufficiency in 1p36-related symptoms remains an interesting speculation but requires more scrutiny in studying existing and new in vivo models.

Another pertinent question with regard to the role of RER1 is its exact molecular function in the cell. RER1 has several highly conserved domains but none of them have been associated to a specific function. Therefore, to fully understand the molecular function of RER1, a more in-depth structure–function analysis of these domains needs to be performed. Ideally, a molecular structure of RER1 together with a given cargo would enable the identification of the functional protein–protein interfaces; however, given its very hydrophobic nature, this remains a challenge.

What do we know so far? Our current model of the molecular mechanism, in which RER1 is involved in the assembly of multi-subunit complexes is depicted in Fig. 4. According to this model, a subunit of a complex harbors a RER1-interacting motif (RIM) and a complex-interacting motif (CIM) that are adjacent or overlapping. RER1, which has a cargo-interacting motif (CaIM), competes with the partially assembled complex for this subunit. As long as RER1 is bound to the subunit, no interaction between the partially assembled complex and the subunit is possible because the CIM on the subunit is masked, i.e. concealed. By contrast, when the complex is fully assembled, the RIM is masked and RER1 can no longer bind its substrate. CIM, RIM and CaIM are all located within TMDs and, at least in the well-characterized cases, contain polar aa residues. Prerequisite for this mode of action is that CIM and RIM on the TMD of the RER1-binding subunit are very close to each other or overlapping. A motif could also serve as CIM and RIM at the same time. Indeed, at least in the case of PEN2, a conserved N residue in its first TMD (the CIM) binds to an N residue in the fourth TMD of presenilin (the CIM) within the fully assembled γ-secretase complex (Fassler et al., 2011). The same N residue of PEN2 is also involved in binding to RER1, suggesting that CIM and RIM are identical (Fassler et al., 2011, 2010). Likewise, in NCT, the RIM overlaps with the motif that interacts with γ-secretase subunit APH1A (the CIM) (Spasic et al., 2007). A similar masking of a RIM in a subunit has been suggested for the yeast Fe²⁺ transporter Fet3 (Sato et al., 2004). Whether this is a general feature of all RER1 cargoes remains to be shown. The masking, i.e. concealing, of retrieval signals as an elegant way to pass retrieval receptors, resembles the aforementioned ‘hide and run’ mechanism (reviewed by Michelsen et al., 2005). Unmasking and masking of specific motifs might, thus, be a general principle to facilitate the assembly of multimeric membrane protein complexes as part of the protein quality control beyond the primary control mechanisms of protein folding within the ER. As this also occurs in distal compartments, including the ERGIC and cis-Golgi, the ‘hide and run’ principle extends protein quality control beyond the ER (reviewed by Sun and Brodsky, 2019).

In this Review, we have taken a closer look at the small hydrophobic membrane protein RER1. It is, thus far, the only known retrieval receptor that recognizes its substrates through signals located within TMDs. RER1 is highly conserved in all eukaryotes, yet it has a specific set of substrates in different cell types and, possibly, different organisms. RER1 haploinsufficiency potentially contributing to 1p36-related symptoms remains an interesting speculation and should be further investigated in appropriate in vivo models. Future focus should go towards fully elucidating the physiological role of RER1 by identifying its full interactome and suite of substrates, as well as further structural analysis of its conserved domains.

We thank Christina Valkova and Koret Hirschberg for critical reading and constructive comments on the manuscript.