Although Ubiquitylation encompasses two E1s and at least ten E2s, there is solely one of each enzyme reported to date for the downstream UFMylation cascade (21). Following UFM1 maturation, UBA5, which acts as the E1 and initiates the conjugation reaction, forms a thioester bond with UFM1 via adenylation in an ATP-dependent manner. In the second step, UFM1 is transferred from UBA5 to UFC1, the E2, using a thioester linkage as well (22).

While most components of the UFM1 pathway, as discussed later, are mainly localized at the cytosolic side of the ER membrane, both UBA5 and UFC1 exhibit a cytosolic and also nuclear distribution (9, 21). Earlier studies in fibroblasts revealed that UBA5 mutations impair the process of UFMylation, resulting in an abnormal ER structure. Experimental and clinical findings from animal models and humans further support UBA5 aberrance as a pathophysiological cause for many abnormalities during neurodevelopment (23, 24). In addition to UFM1, UBA5 was also reported to conjugate SUMO2, another ubiquitin-like protein, and activate SUMO2 in the nucleus, though details of this interaction remain to be elucidated (21). Meantime, biallelic UFC1 variants disrupting UFM1-UFC1 formation have been identified in children with severe early-onset encephalopathy, also arguing for an essential role for UFMylation in brain development (18). Moreover, cell adhesion molecules NCAM140 and KIRREL3 were found to interact with UFC1, implicating a potential function of UFM1 modification for cell surface proteins (25, 26).

In contrast to the more than 600 ligases utilized by Ubiquitylation reaction, the ligation of UFM1 to its target proteins is mediated by UFL1, again, the only E3 defined thus far for UFMylation (14). Surprisingly, UFL1 does not have any typical domains conserved in other E3s except for a highly conserved N-terminal region across species. For this reason, it was believed that UFL1 may be a scaffold-type E3 that brings UFM1 substrate and UFC1 in close proximity, and the E3 ligase activity of UFL1 requires the assistance of other proteins (27). Indeed, evidence is growing that UFL1 forms a tripartite complex with DDRGK1 and C53 proteins, which are regarded as substrate adapters and key regulators for UFMylation system (28, 29) ( Figure 1 ).

Intriguingly, interaction with UFM1 not only stabilizes DDRGK1 and C53, but also provokes their post-translational modifications by UFMylation. In fact, DDRGK1 (also known as C20orf116 and UFBP1) was the first characterized substrate of UFM1, and soon emerged as an E3 cofactor, whose UFMylation is not necessary for E3 complex assembly, but could alter the UFMylation of other substrates, such as ASC1, a nuclear receptor co-activator vital for breast cancer development (14, 28) ( Table 1 ). While DDRGK1 possesses a proteasome-COP9-initiation factor (PCI) domain that embraces the major lysine site for UFMylation and mediates the binding with UFL1, both DDRGK1 and UFL1 contain a transmembrane peptide and a nuclear localization signal (NLS) with the latter likely working merely when the transmembrane domain is deleted (34, 35). Due to the transmembrane helix at its N-terminus, DDRGK1 is predominantly tethered to the ER membrane and recruits UFL1 and C53 to form a heteromeric complex at the cytosolic surface of ER (30, 36, 37). Defect in this hydrophobic N-terminal region of DDRGK1 would lead UFL1 to localize in the cytoplasm, where UFL1 was found to be cleaved with unknown mechanism (30). In addition, numerous functions have been described for DDRGK1, including ER homeostasis, stress responses and cell differentiation (38–40). However, it is still undetermined whether all these functions involve UFMylation.

The second putative adaptor protein for UFL1 is C53 (also named CDK5RAP3 or LZAP). Initially classified as a tumor suppressor, C53 was observed to translocate from ER to nucleus upon UFL1 depletion to inhibit cyclin D1 and cell cycle progression in glioma cells (41). This phenomenon probably was resulted from the direct binding of C53 to RelA. RelA is the regulatory subunit of NF-κB in nucleus, essential for the transcriptional activity of NF-κB, which is crucial for cyclin D1 expression, tumor formation and metastasis (42). Conversely, knockdown of UFL1 has also been documented to elicit elevated NF-κB activity and cell invasion in Hela and U2OS cells. Surprisingly the efficacy of UFL1 ablation on NF-κB was stronger than C53 knockdown, indicating, besides C53, UFL1 may also modulate other components of the NF-κB pathway (30). More controversially, an article by Xi and coworkers depicted that DDRGK1 could positively regulate the degradation of IκBα, the inhibitor of NF-κB, and consequently facilitate NF-κB activity (40). In short, the disagreement in this scenario was suggestive that components of the UFM1 pathway may have function independent of UFMylation, and the precise relationship among UFL1, DDRGK1 and C53 is still a mystery.

In steady state, the ER undergoes continuous renovation to sustain proteostasis, which is achieved by two major mechanisms, ER-phagy and ERAD, to eliminate defective polypeptides. It is well-known that arrested protein products in the cytosol rendered by ribosome stalling are degraded by the ubiquitin-proteasome system. On the other side, removal of translation-faulty ER proteins requires lysosome (or vacuole in plants) associated ER-phagy, in that arrested proteins during co-translational protein translocation are situated inside the translocon with their N-termini inside the ER lumen (44, 45). Thus ER-phagy is usually defined as the lysosomal degradation of part of the ER network containing membranous compartments via partially or no intersecting with the selective autophagy machinery.

Recently a genome wide CRISPR screen identified UFL1 and DDRGK1, but not C53, as critical regulators for starvation-induced ER-phagy (33). They unveiled that nascent proteins trapped in the ribosome stimulates UFMylation of ER substrates such as RPN1 and RPL26 ( Table 1 ), thereby targeting ER sheets for degradation through ER-phagy, though the downstream played by UFMylated RPN1 and RPL26 during ER marking and autophagic engulfment remain to be investigated (33) ( Figure 1 ). On the contrary, C53 in plants and in human cell lines was justified to mediate ER stress-induced autophagic flux upon ribosome stalling after C53 was recruited to the ER by forming a functional ternary complex with UFL1 and DDRGK1. A working model was proposed simultaneously that UFM1 was transferred from C53 to RPL26 under proteotoxic stress so that C53 could act as an autophagy receptor to interact with ATG8 (Autophagy-related gene 8), another ubiquitin-like protein, and initiate autophagosome formation (46) ( Figure 1 ). By comparing the two studies, the involvement of C53 in ER-phagy seems dependent on the physiological context, since phenotyping experiments revealed that C53 mutant was asymptomatic during carbon or nitrogen starvation but was highly sensitive to ER-stress treatments (46, 47). In that case, an in-depth exploration then is urgently needed to address whether UFMylation of DDRGK1 and C53 is a collateral effect or is in turn required for the further UFMylation of other substrates.

Misfolded proteins in the secretory pathway ordinarily are monitored by molecular chaperons and eliminated via the ERAD system to ensure that only correctly folded proteins exit the ER (48, 49). When a protein fails to acquire its correct conformation, it is switched to the ERAD pathway and is immediately transported over the ER membrane towards the cytosol, where the ERAD substrate is ubiquitinated and degraded by the proteasome. Although significant progress has been made in dissecting the role of ERAD for ER homeostasis, little is known about the concrete link between UFM1 modification and ERAD. A study by Schuren, et al. aiming to isolate novel factors associated with ERAD reported that nearly all components in UFMylation affected the degradation of ER-resident HLA-I, a protein for cell immunity, in a different manner than ubiquitin (50). They uncovered that interference with the UFM1 conjugation machinery specifically inhibited the ER-to-cytosol dislocation of HLA-I and misdirected the ERAD-mediated protein sorting. Notwithstanding, they failed to detect any UFMylated substrate directly contributing to ERAD, in accordance with the ERAD phenotype in K562 and U2OS cells, which has been hinted as an indirect consequence of aberrant ER protein biogenesis impacted by erroneous UFMylation activities (33) ( Figure 1 ). On the other hand, it will be intriguing to examine whether the same side effect also occurs to hormone production and secretion in endocrine glands.

Various pathological conditions, such as nutrient deprivation, hypoxia, lipotoxicity, calcium depletion and viral infection, could perturb ER homeostasis (51–53). When the regulatory capacity of ER-phagy and ERAD is exhausted, misfolded proteins and their aggregates accumulate and trigger ER stress (49). To cope with this circumstance, cell has evolved a set of adaptive reactions, collectively called the unfolded protein response (UPR) that halts protein translation, enhances ER-folding capacity and elevates the activity of ER-phagy and ERAD. While temperate ER stress has the ability to protect cell and restore homeostasis, excessive or prolonged UPR would cause cell apoptosis and tissue damage (54).

There are three parallel branches of the UPR that are initiated by distinct ER stress sensors, including the kinase/RNase IRE1 (a synonym for ERN1, Endoplasmic reticulum to nucleus signaling 1), eIF2α kinases PERK and GCN2, and transcription factor ATF6 (55–59) ( Figure 1 ). IRE1 is the most conserved sensor coordinating the UPR from the early stage of stress response. IRE1 oligomerizes upon ER stress and allows for trans-autophosphorylation and activation of its endonuclease domain, which splices the mRNA of the transcription factor XBP1 to drive the expressions of genes encoding chaperones, foldases, ERAD components and lipid synthesis enzymes (60–65). When restoration of ER homeostasis fails, IRE1 also represses the adaptive cellular responses and activates apoptosis through JNK signaling (66–68). The second branch of the UPR is dominated by PERK, which phosphorylates eIF2α and attenuates the initiation of global protein translation (53). Paradoxically, phosphorylation of eIF2α could preferentially translate the transcription factor ATF4, which subsequently regulates several target genes, such as CHOP involved in cell death (69). Distinct from PERK that shares similar ER domains to IRE1, cytoplasmic GCN2 has also been showed to phosphorylate eIF2α cooperatively with PERK, but it is still questionable how it communicates stress signals from the ER (58). Additionally, PERK is known to regulate proliferation and differentiation of insulin-secreting β-cells of endocrine pancreas, as well as to promote adipocyte differentiation via intrinsic lipid kinase activity. Both, however, might be unrelated to the UPR pathway. The last transducer is ATF6 that is released from ER under stress boost. ATF6 controls gene expressions in nucleus to increase ER volume and reinforce the processing and degradative capacities of ER, serving overlapping functions with IRE1 and PERK (70).

So far, there has been a persuasive accumulation of data implying that UFMylation has an intricate relation to ER stress and UPR. For example, Zhang, et al. showed that UFMylation genes were transcriptionally up-regulated by disturbance of the ER homeostasis and inhibition of vesicle trafficking. They illustrated UFM1 promoter contained a putative binding site for the ER stress-responsive transcriptional activator XBP1 (71). UFM1 system thereafter was indicated to be indispensable for ER stress tolerance, exerting a protective role to prevent cells from death in the presence of stress (72). Yet in sharp contrast, skin fibroblasts from individuals with mutations in UBA5 were found to be aberrant in UFMylation and resistant to ER stress-induced toxicity with abnormally expanded ER structure (73). These conflicts in general may be because UFM1 modification differentially influences the ER network in different cell types, as in a subset of cultured cells, targeted disruption of genes encoding UFM1 machinery only weakly activates an ER stress response but would prompt UPR and ER stress-induced apoptosis in cells like U2OS, hematopoietic stem cells (HSC) and pancreatic beta cells.

Mechanistically, Liu, et al. claimed that DDRGK1 could regulate IRE1 protein stability through an UFMylation-dependent protein-protein interaction (74) ( Figure 1 ). They reported that depletion of DDRGK1 diminished IRE1-XBP1 signaling and activated the PERK-eIF2α-CHOP apoptotic pathway by targeting IRE1, probably owing to the similar induction requirements for the two UPR branches. As a result, absence of DDRGK1 in HSCs and cancer cells would lead to ER stress and finally disturb ER homeostasis. Nevertheless, neither the DDRGK1-mediated UFMylation nor stabilization of IRE1 can be verified by other scholars applying different cell lines and experimental settings (33, 75). Instead, higher expression and activation of IRE1-XBP1 was observed by Zhu et al. when DDRGK1 was deficient in plasma cell, an antibody secreting cell derived from B cell (75). Consistent with earlier literatures (36, 71), Zhu and his colleagues confirmed that the IRE1/XBP1 axis also upregulated the expression of DDRGK1 and UFM1 pathway genes in the differentiated B cell. They proved that DDRGK1 downstream of XBP1 was fundamental to the secretory function of plasma cell, and that lack of UFMylation retarded immunoglobulin production and impaired ER expansion, which likely fed back to the expression and activation of IRE1. In line with Liu’s speculation, however, they also suggested that DDRGK1 was required for the suppression of PERK during plasma cell development through an unknown mechanism but independent of DDRGK1 UFMylation.

Besides IRE1 and PERK, until now it is still not sure whether the UFM1 process interacts with GCN2 or ATF6 as well. Taken together, there definitely is a long way before the complicated interconnection between UFMylation and ER stress can be fully understood.

The pancreas is well-known for its dual function - the exocrine function engaged in digestive enzymes active in the small intestine, which are pivotal for nutrient digestion, and the endocrine effect, which is stimulated to control over the blood glucose level via pancreatic hormones (85). Actually UFM1 system appears to be involved in both exocrine and endocrine cells. In light of the theme of this review, merely endocrinal regulation in pancreatic beta cell will be discussed below.

Compared with other endocrine hormones that are ultimately guided by the pituitary gland, the pancreatic beta cell is unique in its capacity to synthesize, store and secrete insulin with precise rates in response to circulating sugar, covering the metabolic needs of the organism (86). The fine-tuning of insulin production and secretion is modulated at multiple levels, ranging from transcription, mRNA stability to translation and protein folding. To fulfill the task of insulin biosynthesis, the beta cell has evolved a rich ER whose pathologies are tightly associated with metabolic disturbance and obesity. PERK null mice, for instance, have increased beta cell apoptosis and early onset diabetes, whereas CHOP deletion mice have improved beta cell function and cell survival when insulin demand rises (87, 88). Reciprocally, ER stress is activated under circumstances of cellular nutrient overload, including lipotoxicity (89). Genetic or diet-induced models of obesity would act via PERK to enhance obesity-induced insulin resistance, in which IRE1 was also shown to lower insulin responsiveness (90).

Notably, in diabetic MKR mice expressing dominant-negative mutant IGF-IRs specifically in skeletal muscle, Lu et al. observed marked up-regulation of UFM1 and ERAD proteins for the first time in pancreatic islets using a quantitative proteomics approach combined with microarray. They hence identified UFM1 as a potential factor for the development of type 2 diabetes (81). Thereafter, Lemaire et al. testified that UFM1 and DDRGK1 were mainly expressed in beta cells in the mouse islets of Langerhans. Their expression in pancreas was increased upon feeding and ER stress but not on other stressors, such as cyclohexamide and H₂O₂ (36). Furthermore, evidence was provided that failing UFMylation by RNAi triggers ER stress-induced apoptosis of beta cells although neither UFM1 nor DDRGK1 was required for the glucose-mediated insulin release.

The ovary is the female gonad whose major function is the differentiation and release of mature oocytes for fertilization. It’s in charge of the synthesis and secretion of steroid hormones, including estrogen and progesterone, allowing the development of female secondary sexual characteristics and supporting pregnancy. In mammals, each follicle, the functional unit of ovary, is composed of an innermost germ cell (so called oocyte), surrounding granulosa cells (GCs), and outer layers of thecal cells (91). Controlled by endocrine and paracrine factors, the follicles develop with the proliferation of GCs through primary, secondary and antral phases, during which most follicles undergo atretic degeneration with a few of them eventually reaching the point of ovulation (39). After the secondary follicular stage, more protein synthesis and post-translational modifications are demanded in order to satisfy the accelerated GC proliferation and oocyte growth, which may create a deleterious situation for the ER (92–94). The follicular atresia thereby is initiated by GC apoptosis commonly thought to be evoked by the UPR pathway. By comparing the expression of UFL1 in mouse ovaries at different developmental stages, it was speculated that the increased expression of UFL1 during follicular development must dedicate to the maintenance of ER homeostasis in GCs, providing a stable internal environment for the maturation of follicles (95). Consistently, the expression levels of UFM1 and DDRGK1 were reported to be higher in GCs derived from antral atretic follicles than healthy ones in ovaries from ruminants (39).

In human, ovarian disease has gradually turned into hot topic due to the tendency of delaying the age of having children. Premature ovarian failure (POF) is a gynecological syndrome defined as an ovarian functional defect occurring before the age of 40 years. It is characterized by amenorrhea, hypogonadism and estrogen deficiency, severely reducing the fertility of women, especially in female cancer patients receiving chemotherapy drugs (96). A mouse model of POF was set up by intraperitoneal injection of cisplatin. In this model, cell apoptosis and follicular atresia were triggered by elevated ER stress in GCs associated with a significant decrease of UFL1 protein expression there (79). Interestingly, overexpression of UFL1 could alleviate ER stress and reduce cisplatin-induced damage to the ovary, in the end augmenting follicle number and restoring estrogen secreted by GCs to some extent. Similarly, UFM1 silencing was found to enhance ER stress but not oxidative stress-induced GC apoptosis during goat follicular development and atresia, suggesting a conserved role of UFMylation in the prevention of cell death related to the ER in ovaries (97).