The short mRNA isoform of the immunoglobulin superfamily, member 1 gene encodes an intracellular glycoprotein

Mutations in the immunoglobulin superfamily, member 1 gene (IGSF1/Igsf1) cause an X-linked form of central hypothyroidism. The canonical form of IGSF1 is a transmembrane glycoprotein with 12 immunoglobulin (Ig) loops. The protein is co-translationally cleaved into two sub-domains. The carboxyl-terminal domain (CTD), which contains the last 7 Ig loops, is trafficked to the plasma membrane. Most pathogenic mutations in IGSF1 map to the portion of the gene encoding the CTD. IGSF1/Igsf1 encodes a variety of transcripts. A little studied, but abundant splice variant encodes a truncated form of the protein, predicted to contain the first 2 Ig loops of the full-length IGSF1. The protein (hereafter referred to as IGSF1 isoform 2 or IGSF1-2) is likely retained in most individuals with IGSF1 mutations. Here, we characterized basic biochemical properties of the protein as a foray into understanding its potential function. IGSF1-2, like the IGSF1-CTD, is a glycoprotein. In both mouse and rat, the protein is N-glycosylated at a single asparagine residue in the first Ig loop. Contrary to earlier predictions, neither the murine nor rat IGSF1-2 is secreted from heterologous or homologous cells. In addition, neither protein associates with the plasma membrane. Rather, IGSF1-2 appears to be retained in the endoplasmic reticulum. Whether the protein plays intracellular functions or is trafficked through the secretory pathway under certain physiologic or pathophysiologic conditions has yet to be determined.


Introduction
Loss-of-function mutations in the immunoglobulin superfamily, member 1 gene (IGSF1/Igsf1) cause central hypothyroidism in humans (OMIM #300888) and mice [1][2][3]. IGSF1 is abundantly expressed in the developing and adult pituitary gland [3][4][5]. According to observations in both humans and mice, IGSF1-deficiency is associated with impaired hypothalamic thyrotropin-releasing hormone (TRH) stimulation of thyroid-stimulating hormone (TSH) synthesis and/or secretion by thyrotrope cells of the anterior pituitary [2,3,6,7]. IGSF1's normal function in cells and how its absence leads to impaired TRH action are presently unknown. a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 The IGSF1/Igsf1 gene encodes several mRNA transcripts in a tissue-specific manner [4,[8][9][10]. The most thoroughly characterized transcript derives from 20 exons and encodes a large transmembrane glycoprotein of 12 C2-type immunoglobulin (Ig) loops. This protein is cotranslationally cleaved into N-and C-terminal domains (NTD and CTD) [11]. According to in vitro analyses, the CTD traffics to the plasma membrane, whereas the NTD is retained in the endoplasmic reticulum (ER). The CTD contains the last 7 of 12 Ig loops (from the full-length IGSF1), a transmembrane domain, and a short cytoplasmic tail, and is generally regarded as the functional part of the protein. This concept derives from at least two observations. First, most intragenic IGSF1 mutations map to the part of the gene encoding the CTD [1][2][3][12][13][14][15][16][17][18]. Second, the CTD can be produced from a mRNA isoform derived from an intronic promoter, at least in mouse [4,7,11]. Therefore, expression of the NTD is not required for expression or proper membrane trafficking of the CTD. Nonetheless, the 5 Ig loop-containing NTD is conserved across mammalian species, suggesting that it may have currently unappreciated functions.
During the initial characterization of Igsf1 mRNAs in rats, a truncated but abundant transcript was identified in both pituitary and testis [10]. It was cloned from human and murine pituitaries shortly thereafter [4,19]. This mRNA variant shares 5' sequence through the first 5 exons with the full-length IGSF1/Igsf1. However, it retains part of the 5 th intron and protein translation terminates within this sequence. The open-reading frame of this variant (hereafter IGSF1 isoform 2 or IGSF1-2) is predicted to encode the first 2 of the 12 Ig loops in the fulllength IGSF1. As the protein contains an N-terminal signal peptide but lacks a transmembrane domain, it was predicted to be secreted [10], but this has not been demonstrated experimentally. Indeed, to date, there are no published reports on characterization of IGSF1-2. Here, we investigated whether IGSF1-2 is a secreted protein.

Constructs
The rat IGSF1-2 expression vector was generated by PCR amplifying the coding sequence of Igsf1-2 from a rat testis cDNA library clone and ligating it into the KpnI and ApaI sites of pcDNA4/Myc-HIS-A (Invitrogen, Carlsbad, CA). All PCR primer sequences are provided in Table 1. The murine IGSF1-2 expression vector was generated through a three-step process. First, the coding sequence was PCR amplified from adult mouse pituitary cDNA and ligated into the HindIII and XhoI sites of pcDNA3.0. Two additional amplification reactions were performed starting with this plasmid to enable in-frame ligation into pcDNA4/Myc-HIS-A. Asn43Gln (N43Q) and N44Q mutations were introduced into the murine and rat IGSF1-2 expression vectors, respectively, using the QuikChange protocol (Agilent Genomics, Santa Clara, CA). The transthyretin (Ttr) coding sequence was PCR amplified from murine liver cDNA and ligated into the HindIII and XbaI sites of pcDNA4/Myc-HIS-A. The BMPR1A-Myc expression construct was previously described [20]. All constructs were confirmed by Sanger sequencing (Genome Québec, Montréal, Québec).

Protein purification
Twenty-four h post transfection, culture medium was collected and cell debris removed by centrifugation (10 min at 4000 rpm). HisPur Cobalt Resin (89964, ThermoFisher) was washed with PBS and then added to the medium. The resin-medium mixture was incubated overnight at 4˚C with rotation. Resin was allowed to settle by gravity, supernatant was removed, and the resin was washed with 10 mmol/L imidazole (I5513, Sigma Aldrich) in PBS. Bound proteins Table 1. Primers used for cloning and mutagenesis.

Plasmid
Sense Antisense were eluted with 250 mmol/L imidazole in PBS for 5 min at 95˚C. Four μg protein samples were resolved on 14% (v/v) Tris-glycine polyacrylamide gels and immunoblotted as described above.

Cell surface biotinylation
Transfected HEK293 cells were washed with PBS and incubated with EZ-link Sulfo-NHS-LC-Biotin (21335, Pierce, Nepean, Ontario) diluted to 0.5 mg/mL in PBS for 30 min at 4˚C. Cells were washed with PBS containing 100 mmol/L glycine and lysed in RIPA buffer as described above. Supernatant was incubated overnight at 4˚C with EZview Red Anti-c-Myc Affinity gel (E6654, Sigma Aldrich) and eluted with Laemmli buffer containing 2% (v/v) βmercaptoethanol. Blotting proceeded as described above; however, the membrane was blocked with TBS-T containing 5% (w/v) bovine serum albumin (BSA) and incubated with Vectastain ABC Elite as per manufacturer's instructions (PK-6100, Vector Laboratories, Burlingame, CA) before being exposed to film.

Results and discussion
To enable investigations of the IGSF1-2 protein isoform, we cloned the murine cDNA into an expression vector that added a Myc/His tag to the C-terminus. This was necessary as antibodies against the N-terminus of IGSF1 produced inconsistent results (data not shown). When expressed in CHO cells, murine IGSF1-2 migrated as a single protein species of~28 kDa ( Fig  1A, lane 2). The previously described IGSF1-NTD and-CTD are glycoproteins [11]. IGSF1-2 is similarly predicted to be glycosylated at an asparagine residue (Asn43) in the first of its 2 Ig loops. Consistent with this idea, removal of N-linked sugars with either PNGaseF or EndoH hastened migration of the protein on SDS-PAGE (Fig 1A; lanes 3 and 4). Moreover, mutation of Asn43 to Gln (N43Q) caused a similar increase in IGSF1-2's mobility (lane 5), which was not further altered by PNGaseF or EndoH (lanes 6 and 7). Collectively, these data indicate that murine IGSF1-2 is a glycoprotein, which is glycosylated at Asn43. Moreover, the equivalent effects of EndoH and PNGaseF suggest that the protein only acquires immature sugars and may therefore not transit from the ER to the Golgi in the secretory pathway.
To assess the generality of these results, we repeated the analyses with rat IGSF1-2. Both murine and rat IGSF1-2 migrated as single protein species when expressed in CHO cells ( Fig  1B, lanes 4 and 7). Again, treatment with EndoH or PNGaseF caused similar patterns of deglycosylation in both species (lanes 5, 6, 8, and 9). Comparable results were observed in a second cell line (HEK293; data not shown and Fig 2). Thus, IGSF1-2 is a glycoprotein containing only immature N-linked sugars in both mouse and rat. Interestingly, though the rat and murine IGSF1-2 proteins are of similar length (233 and 232 amino acids, respectively), the rat protein consistently migrated more rapidly on SDS-PAGE than the mouse (compare lanes 4 and 7 in Fig 1B). This likely resulted from differences in the cloning strategies for the expression constructs between the two species and differences in the lengths of their signal peptides (18 and 10 amino acids, respectively, according to SignalP-4.1).
Although IGSF1-2 was originally predicted to be secreted [10], the EndoH-sensitivity of the protein suggested that it might not exit the ER. We therefore asked whether IGSF1-2 could be detected in culture medium of transfected cells. HEK293 cells were transiently transfected with wild-type or glycosylation-deficient forms of murine (N43Q or NQ) or rat (N44Q or NQ) IGSF1-2 or the thyroid hormone binding protein transthyretin (TTR). The latter is a secreted protein [21] and was included as a positive control. As with IGSF1-2, the TTR expression vector included a C-terminal Myc/His tag. As demonstrated by immunoblotting of whole cell  lysates, all proteins were expressed at roughly equivalent levels (Fig 2, lanes 2-6, bottom  panel). In contrast, only TTR could be detected in culture medium, whether or not proteins were enriched via Ni-NTA chromatography (Fig 2, lane 6, middle and top panels). Equivalent results were observed in a second heterologous cell line (CHO; data not shown).
Although these results suggested that IGSF1-2 was not secreted, they did not definitively show that the protein failed to transit out of the ER. For example, it is possible that the protein might navigate the secretory pathway, but remain associated with the plasma membrane. Though IGSF1-2 lacks a transmembrane domain or motif for GPI-linkage, this did not rule out association through protein-protein or other biophysical interactions. We therefore used cell-surface biotinylation to assess whether IGSF1-2 associates with the plasma membrane. HEK293 cells were transfected with C-terminally Myc-tagged forms of IGSF1-2 (mouse and rat) or the human BMP type IA receptor (BMPR1A). The latter is a transmembrane glycoprotein and was used as a positive control. Roughly equivalent amounts of the proteins were immunoprecipitated with a Myc antibody (Fig 3A, lanes 2-4, bottom panel). BMPR1A runs as a doublet, with the higher molecular weight band representing the mature, plasma membrane form of the protein. As shown in the cell-surface biotinylation analysis, only this form of the protein was detected at the plasma membrane ( Fig 3A, lane 4, top panel). Neither murine nor rat IGSF1-2 was detected at the membrane (lanes 2 and 3, top panel). In complementary immunofluorescence analyses, IGSF1-2 protein could be detected in permeabilized, but not non-permeabilized cells (Fig 3B, compare top and bottom panels). Moreover, IGSF1-2 colocalized with GRP-78/BiP, a well-described luminal ER protein (Fig 3C). These results are consistent with the hypothesis that IGSF1-2 is not a secreted or plasma membrane-associated protein, but rather remains within the cell, most likely in the ER.
One caveat to the above results is that all the analyses were performed in heterologous cells (CHO or HEK293). It was therefore possible that mechanisms required for IGSF1-2 secretion were absent. Thus, we turned to a homologous cell system. IGSF1 is expressed in developing liver [5] and in hepatocellular carcinoma [22]. IGSF1-CTD is expressed in the liver cancer cell line HepG2 (Fig 4A). Unfortunately, we were unable to confirm expression of endogenous IGSF1-2 in HepG2 cells, as we lack antibodies against this isoform. Nonetheless, when we expressed murine or rat IGSF1-2 in HepG2 cells, we were again unable to detect the proteins in the culture medium, in contrast to TTR (Fig 4B, top panel). These data further support the interpretation that IGSF1-2 is not a secreted protein, at least not in the cell lines investigated here.
The IGSF1-2 protein is highly conserved in amino acid sequence across mammals. A BLASTP search of the murine IGSF1-2 sequence returned definitive alignments to IGSF1-2 orthologs (as opposed to full-length IGSF1) in 16 mammalian species. Sequence identity ranged from 77% (big brown bat) to 96% (rat), with 87% identity in human IGSF1-2. This level of conservation suggests a functional role for the protein. It is presently unclear, however, what this role might be in the pituitary or other tissues. It is possible that IGSF1-2 is secreted under conditions we were unable to recapitulate in cultured cells. As there is precedent for Ig superfamily members playing roles in the ER [23], we also cannot rule out intracellular functions. That said, we generated two Igsf1-deficient mouse models, which have similar phenotypes [3,4,7]. The first model [4] lacks multiple IGSF1 isoforms (including IGSF1-2), whereas IGSF1-2 is intact in the second model [7] (Brûlé and Turgeon, unpublished). This suggests that IGSF1-2 is not functional or that its function depends on the co-expression of other IGSF1 isoforms. medium were analyzed directly. In the top panel, proteins in the media were analyzed following Ni-NTA purification and enrichment. https://doi.org/10.1371/journal.pone.0180731.g002 In summary, the 2 Ig loop isoform of IGSF1 (IGSF1-2) is highly expressed in the pituitary gland and is conserved across mammalian species. Despite earlier predictions [10], IGSF1-2 does not appear to be secreted. Therefore, any functions of the protein may be intracellular, perhaps in the ER. Most pathogenic mutations in IGSF1 map to the portion of the gene encoding the CTD. As a result, IGSF1-2 should be intact in these individuals. In a few families, the entirety of the IGSF1 gene is deleted; however, their phenotypes are similar to those of individuals harboring missense, nonsense, or frame-shift mutations in the CTD [3,18]. Therefore, HEK293 cells were transfected with expression plasmids for wild-type murine (M.) or rat (R.) IGSF1-2, murine BMP type IA receptor (BMPR1A), or empty vector (pcDNA4). Note, IGSF1-2 proteins were expressed with Myc/His tags at their C-termini, whereas BMPR1A had the Myc tag alone. Cell surface proteins were labeled with biotin prior to collection of whole cell lysates. Lysates were immunoprecipitated (IP) with Myc-beads and then subjected to SDS-PAGE and transferred to nitrocellulose membranes. Total proteins were immunoblotted with anti-Myc (bottom), whereas biotinylated proteins were identified with streptavidin-HRP (top). B) HEK293 cells were cultured on coverslips and transiently transfected with wild-type murine IGSF1-2-Myc/His. Cells were then fixed and subjected to immunofluorescence with a Myc antibody (green) under non-permeabilizing (top) and permeabilizing conditions (bottom). C) Cells were transfected as in panel B, premeabilized, and processed for double-label immunofluorescence with the Myc antibody (green) and an antibody against GRP-78/BiP (red). The overlay is shown in yellow. In B and C, nuclei were stained with DAPI. Images were captured by confocal microscopy. Scale bar, 10 μm.