Non-glycosylated Seipin to Cause a Motor Neuron Disease Induces ER stress and Apoptosis by Inactivating the ER Calcium Pump SERCA2b

A causal relationship between endoplasmic reticulum (ER) stress and the development of neurodegenerative diseases remains controversial. Here, we focused on Seipinopathy, a dominant motor neuron disease, based on the finding that its causal gene product, Seipin, is a hairpin-like transmembrane protein in the ER. Gain-of-function mutations of Seipin produce non-glycosylated Seipin (ngSeipin), which was previously shown to induce ER stress and apoptosis at both cell and mouse levels albeit with no clarified mechanism. We found that aggregation-prone ngSeipin dominantly inactivated SERCA2b, the major calcium pump in the ER, and decreased the calcium concentration in the ER, leading to ER stress and apoptosis. This inactivation required oligomerization of ngSeipin and direct interaction of the ngSeipin C-terminus with SERCA2b, and was observed in Seipin-deficient human colorectal carcinoma-derived cells (HCT116) expressing ngSeipin at a level comparable with that in neuroblastoma cells (SH-SY5T). Our results thus provide a new direction to controversy noted above.


INTRODUCTION
The BSCL1 and BSCL2 genes have been identified as causal genes of congenital generalized lipodystrophy (CGL) or Berardinelli-Seip congenital lipodystrophy syndrome (BSCL), a rare autosomal recessive disease characterized by insufficiency of adipose tissue from birth or early infancy and by severe insulin resistance. The BSCL1 gene encodes 1-acylglycerol-3-phosphate O-acyltransferase 2 (AGPAT2), which is present in the membrane of the endoplasmic reticulum (ER) and involved in phospholipid biosynthesis (Agarwal et al., 2002;Garg et al., 1999), whereas the BSCL2 gene encodes Seipin, a hairpin-like transmembrane protein in the ER with unknown function at that time (Magre et al., 2001). Loss of function mutations of the BSCL2 gene appear to produce more severe symptoms than those of the BSCL1 gene (Van Maldergem et al., 2002). Since the discovery that Seipin is involved in lipid droplet morphology in yeast (Szymanski et al., 2007), the role of Seipin in the biogenesis of lipid droplets has gained extensive attention (Bi et al., 2014;Cui et al., 2011;Sim et al., 2013;Sui et al., 2018;Tian et al., 2011;Wang et al., 2014;Wang et al., 2016;Yan et al., 2018).
To our interest, two missense mutations of the BSCL2 gene, namely N152S and S154L of Seipin, were found to dominantly cause distal hereditary motor neuropathy (dHMN) or distal muscular atrophy, which is characterized by almost exclusively degenerated motor nerve fibers, predominantly in the distal part of limbs (Windpassinger et al., 2004). Because N 152 , V 153 and S 154 of Seipin match the triplet code (Asn-X-Ser/Thr; X: any amino acid except Pro) for N-glycosylation, neither N152S nor S154L Seipin are glycosylated, leading to the proposal that the production of these aggregation-prone mutants results in neurodegeneration (Windpassinger et al., 2004).
Seipin was first identified as a protein of 398 aa (Magre et al., 2001), and was later found (Lundin et al., 2006) to have two splice variants, a short form of 398 aa and long form of 462 aa (see Fig. 1A), which are translated from three Seipin mRNA isoforms of 1.6 kb, 1.8 kb, and 2.2 kb. Both forms are translatable from 1.8 kb and 2.2 kb mRNA but the long form is more abundantly produced than the short form. In contrast, only the short form is translated from 1.6 kb mRNA (Lundin et al., 2006). Because 1.8 kb mRNA is predominantly expressed in human brain (Magre et al., 2001), it is considered that human brain expresses mainly the long form (Cartwright and Goodman, 2012).
The ER, where Seipin is located, is well known to be equipped with a quality control system for proteins. Productive folding of newly synthesized secretory and transmembrane proteins is assisted by ER-localized molecular chaperones and folding enzymes (ER chaperones hereafter). In contrast, proteins unable to gain their correct three-dimensional structures are dealt with by ER-associated degradation (ERAD), in which unfolded or misfolded proteins are recognized, delivered to the transmembrane complex termed the retrotranslocon, and retrotranslocated to the cytosol for ubiquitindependent proteasomal degradation.
Under a variety of physiological and pathological conditions, however, this quality control system misfunctions, resulting in the accumulation of unfolded or misfolded proteins in the ER. This ER stress is quite detrimental to the cell and may eventually cause cell death. In response, ER stress is immediately and adequately counteracted by a cellular homeostatic mechanism termed the unfolded protein response (UPR). In vertebrates, the UPR is triggered by three types of ubiquitously expressed ER stress sensor and transducers -PERK, ATF6 and IRE1 -which leads to general translational attenuation to decrease the burden on the ER; transcriptional upregulation of ER chaperons to increase productive folding capacity; and transcriptional upregulation of ERAD components to increase degradation capacity.
Daisuke Ito and colleagues showed for the first time that expression of nonglycosylated mutant Seipin by transfection in HeLa cells evokes ER stress, as evidenced by induction of the two major ER chaperones BiP and GRP94, the ERAD component Herp, and CHOP. HeLa cells expressing non-glycosylated mutant Seipin by transfection are subject to more extensive apoptosis (18%) than those expressing wild-type (WT) Seipin (6%) (Ito and Suzuki, 2007). Based on these findings, they proposed the designation of mutant Seipin-linked dominant motor neuron disease as Seipinopathy, which represents a novel ER stress-associated disease (Ito and Suzuki, 2009).
They further constructed a transgenic mice overexpressing human nonglycosylated mutant Seipin under the control of the neuron-specific murine Thy-1 promoter．They found that the levels of ER stress marker proteins BiP and PDI are elevated in brain of the transgenic mice, reproducing the symptomatic and pathological phenotypes observed in human patients with Seipinopathy (Yagi et al., 2011).
Here, we focused on the remaining and most critical question of how nonglycosylated mutant Seipin evokes ER stress.

Construction of Seipinopathy-causal mutant Seipin
We found that human HCT116 diploid cells derived from colorectal carcinoma (Roschke et al., 2002), which we use exclusively for gene knockout analysis, expressed only the short form of Seipin, designated Seipin S , whereas human neuroblastoma-derived SH- To produce the representative of Seipinopathy-causal mutants, we simultaneously mutated Asn 152 and Ser 154 of Seipin L to Ser and Leu, respectively (Fig.   1A). When expressed in HCT116 cells by transfection, N-terminally Myc-tagged WT Seipin L was sensitive to digestion with Endo H but N-terminally Myc-tagged mutant (N152S/S154L) Seipin L was not, as expected (Fig. 1B); WT Seipin L and the nonglycosylated mutant Seipin L are hereafter designated wtSeipin L and ngSeipin L , respectively. wtSeipin L expressed in HCT116 cells by transfection showed typical ER pattern, as expected, whereas ngSeipin L showed uneven distribution in the ER, suggesting that ngSeipin is prone to aggregation (Fig. 1C). It should be noted that as we carried out transfection in cell culture systems of various sizes, we express a transfection index as the amount of plasmid (ng) divided by the bottom area (cm 2 ) of the well/dish, i.e. 2.0 cm 2 7 for a 24-well plate, 9.6 cm 2 for a 6-well plate, and 11.8 cm 2 for a 3.5-cm dish, for easier comparison of results obtained from different experiments; accordingly, when 100 and 123 ng plasmid was transfected into cells in 6-well plates and 3.5 cm dishes, respectively, the transfection index was 10.4 ng/cm 2 .

Effect of ngSeipin expression on calcium concentration in the ER
We focused on SERCA2, the major calcium pump in the ER incorporating cytosolic calcium ion into the ER, based on the previous observation that Seipin physically associates with SERCA in fly as well as with SERCA2 in HEK293 cells (Bi et al., 2014).
It should be noted that the expression level of SERCA2 dominated that of SERCA1 and SERCA3 in both HCT116 and SH-SY5Y cells ( Fig. 1-S1B). It is also known that three splice variants exist for SERCA2 (Gélébart et al., 2003) and that SERCA2b is ubiquitously expressed, whereas SERCA2a and SERCA2c are expressed mainly in myocardium and skeletal muscle, in which the expression level of SERCA2b is low (Dally et al., 2006). Indeed, the expression level of SERCA2b dominated that of SERCA2a and SERCA2c in both HCT116 and SH-SY5Y cells ( Fig. 1-S1C). Interestingly, immunoprecipitation from HCT116 cells expressing wtSeipin L or ngSeipin L by transfection showed that ngSeipin L bound to SERCA2b more extensively than wtSeipin L (Fig. 1D).
Expression of ngSeipin L in HCT116 cells by transfection (10.4 ng/cm 2 ) markedly decreased [Ca 2+ ] in the ER compared with that of wtSeipin L (Fig. 1E). This effect of ngSeipin L on [Ca 2+ ] in the ER was also observed in SH-SY5Y cells ( Fig. 1-S1D). We noticed that small amounts (~20%) of wtSeipin S and ngSeipin S were produced from transfected plasmid [Fig. 1E(c); and see Fig. 3A(c) for clearer evidence]. Because they were not detected with anti-Myc antibody (data not shown), it is likely that they were translated from the second methionine M 65 (see Fig. 1A), given that the nucleotide sequences around M65 (ccgGccATGG) are more similar to the Kozak consensus sequence for translational initiation (gccRccATGG) than those around M1 (aggAagATGt).
Because Seipinopathy is an autosomal dominant disease, we next examined the effect of co-expression of wtSeipin L and ngSeipin L on [Ca 2+ ] in the ER. and 4 with bar 1]. Co-expression of wtSeipin L in a decreasing manner by transfection at 7.80, 5.20, and 2.60 ng/cm 2 and of ngSeipin L in an increasing manner by transfection at 2.60, 5.20, and 7.80 ng/cm 2 (transfection a total of 10.4 ng/cm 2 ) decreased [Ca 2+ ] in the ER [ Fig. 2B(b); compare bars 5, 6 and 7 with bar 1]. Furthermore, CDN1163 treatment did not increase [Ca 2+ ] in the ER significantly in HCT116 cells co-expressing wtSeipin L and ngSeipin L by transfection (5.20 ng/cm 2 each), in contrast to the case of untransfected HCT116 cells (Fig. 2-S2B). Thus, ngSeipin L dominantly inactivates SERCA2b and thereby dominantly decreases [Ca 2+ ] in the ER.

Construction and characterization of Seipin-knockout cells
To examine the effect of endogenous Seipin on SERCA2b, we knocked out (KO) the Seipin gene in HCT116 cells using CRISPR/Cas9-mediated cleavage of the Seipin locus at two sites ( Fig. 3-S1A). The deletion of almost the entire Seipin gene was confirmed by genomic PCR (Fig. 3-S1B), and the absence of Seipin mRNA and Seipin protein was In contrast, a higher expression of ngSeipin L in Seipin-KO cells by transfection at > 2.60 ng/cm 2 decreased [Ca 2+ ] in the ER in dose-dependent manner more robustly than that of wtSeipin L [ Fig. 3A(b); compare bars 6, 8 and 10 with bars 5, 7 and 9]. Higher expression of ngSeipin L by transfection at > 2.60 ng/cm 2 induced ER stress more extensively than that of wtSeipin L in Seipin-KO cells, as evidenced by increased levels of BiP mRNA (a target of the ATF6 pathway), XBP1(S) mRNA (a target of the IRE1 pathway) and CHOP mRNA (a target of the PERK pathway) ( Fig. 3B; compare bars 6, 8, and 10 with bars 5, 7 and 9). Higher expression of ngSeipin L by transfection at > 2.60 ng/cm 2 induced apoptosis more extensively than that of wtSeipin L in Seipin-KO cells, as shown by increased detection of Cleaved Caspase-3 by immunofluorescence ( Fig. 3C

Effect of oligomerization of ngSeipin on inactivation of SERCA2b
To elucidate the mechanism by which ngSeipin L inactivates SERCA2b, we examined the effect of oligomerization of Seipin, because human Seipin exists as a wheel-like undecamer ( Fig. 4A) (Yan et al., 2018). Interestingly, fly Seipin consisting of 370 aa is not glycosylated and exists as a wheel-shaped dodecamer (Sui et al., 2018). Because N-glycan wedges the interface of two protomers in the case of human Seipin (Fig. 4A, right panel), we hypothesized that non-glycosylated human Seipin expressed at a higher level becomes unable to maintain the undecamer structure, leading to aggregation.
Structural analysis revealed that the luminal region of each Seipin monomer consists of 8 b-strands, termed the b-sandwich domain, and 3 a-helices, and that an ER membrane-anchored core-ring is formed when the 3 a-helices of each of 11 monomers is gathered via multiple hydrophobic interactions between one protomer and its neighboring protomer, including those between L226 and L220 and between L233 and V227. The core ring is surrounded by 11 b-sandwich domains which are tightly associated via hydrogen bonds between one protomer and its neighboring protomer, including those between S181 and Y215, between Q239 and S217, and between H131 and R209, and via hydrophobic interactions between one protomer and its neighboring protomer, including those between Y134 and M189 ( To confirm the aggregation propensity of ngSeipin, we employed a proximity ligation assay (PLA), in which a PCR-mediated signal is produced when the distance between two proteins is less than 40 nm (Fig. 4D), and which was used to detect aggregates of a-synuclein, a causal protein of familial Parkinson's disease (Roberts et al., 2015). Results showed the production of a markedly strong signal when Myc-tagged ngSeipin L and Flag-tagged ngSeipin L were co-expressed by transfection each at > 1.30 ng/cm 2 (total at > 2.60 ng/cm 2 ), compared with co-expression of Myc-tagged wtSeipin L

Requirement of both the luminal and C-terminal regions of ngSeipin for inactivation of SERCA2b
To determine which region(s) of ngSeipin L is required for inactivation of SERCA2b, we constructed a series of deletion mutants in wtSeipin L and ngSeipin L , namely DN lacking the cytosolic N-terminal region, DLD lacking the luminal region, and DC lacking the cytosolic C-terminal region (Fig. 6A). We also constructed two swap mutants of wtSeipin L and ngSeipin L , in which the first and second transmembrane (TM) domains of Seipin L were replaced by the fourth and first TM domain of glucose 6-phosphatase, respectively ( Fig. 6A), in reference to the previous swapping experiments (Bi et al., 2014) and with further consideration of the topology of these TM domains ( Fig. 6-S1A). All constructs produced a band of the expected size in transfected WT cells ( Fig. 6B and Fig.   6-S1B, Input).
Immunoprecipitation using anti-SERCA2 antibody revealed that only Seipin L (DC) and ngSeipin L (DC) were hardly co-immunoprecipitated with SERCA2b ( Fig.   6B, lanes 10 and 11, and Fig. 6-S1B, lanes 23 and 24). We thus constructed a mutant which expresses only the C-terminal region of Seipin L and found that this Seipin(Cterm) was efficiently co-immunoprecipitated with SERCA2b (Fig. 6B, lane 12). These findings indicate that the cytosolic C-terminal region of Seipin L is necessary and sufficient for interaction with SERCA2b.
In the case of DN

Increase in the level of SERCA2b reverses the effect of ngSeipin
We finally examined whether the increase in the level of SERCA2b compensates for the decrease in [Ca 2+ ] in the ER caused by expression of ngSeipin L . Overexpression of SERCA2b (WT) but not inactive mutant SERCA2b (Q108H) (Miyauchi et al., 2006)

DISCUSSION
Seipin is conserved from yeast to humans, and yeast and fly orthologues are termed Sei1 (Fld1) and dSeipin, respectively. Interestingly, Sei1 and dSeipin do not have potential Nglycosylation sites, and dSeipin was shown to function as a dodecamer of non-  with Airyscan and Zen/Zen2.6 acquisition software (both from Carl Zeiss).

Genomic PCR
Non-homologous end joining in HCT116 cells was confirmed by genomic PCR using KOD-FX Neo (TOYOBO) and a pair of primers, all del Fw and all del Rv, and inside Fw and inside Rv. or IP3R3 were used as standards.

PLA
For PLA assay, cells grown on a glass-bottom dish were transiently transfected with plasmids to express Myc-tagged and Flag-tagged proteins simultaneously. After 24 h cells were fixed and permeabilized with methanol at -30°C for 6.5 min. PLA was performed using Duolink in situ Starter Set GREEN (Sigma-Aldrich) according to manufacturer's instructions using rabbit anti-Myc polyclonal antibody and mouse anti-Flag monoclonal antibody.

Determination of cell growth rate
Cells transfected with various plasmids were treated with trypsin, and equal amounts of detached cells were plated to 4 dishes each. Cell numbers were counted 0, 24, 48, and 72 h later.

Induction and imaging of lipid droplets
Cells grown on coverslips were transiently transfected with various plasmids and incubated 24 h later with medium containing 400 µM of fatty acid-free BSA (FUJIFILM)-          Fluorescence intensities were quantified and are expressed as in Fig. 1E(b) (n=3).