We generated GluN3B-deficienct mice by deletion of the 8-kb region encoding the entire GluN3B gene and found that it also resulted in partial deletion of membralin (Figure 1—figure supplement 1). Sequence analysis confirmed that the translation frame of the altered membralin transcript in null mice terminates 147 nucleotides (nts) beyond the end of partial exon XI, resulting in a slightly truncated membralin protein with 49 unrelated amino-acid (aa) residues replacing the wild-type (WT) 89 C-terminal aa residues. Thus, our targeting strategy generated a GluN3B/membralin C-terminal DKO mouse line.

DKO mice appeared normal at birth but unexpectedly died at postnatal day 5–6 (P5-6). The motor function of these mice was indistinguishable from WT or heterozygous littermates during the first three postnatal days, but motor strength was severely impaired, thereafter, preceding death. DKO mice were also significantly smaller than WT and heterozygous littermates (Figure 1A) but displayed normal gross anatomy and histology in the brain (Figure 1—figure supplement 2), visceral organs, and skeletal muscle.10.7554/eLife.06500.003Figure 1.

(A) DKO and WT littermate mice are shown at P5. The DKO mouse was unable to maintain correct posture. (B) Immunostaining with anti-Hb9 antibody revealed a dramatic decrease in the number of motor neurons in the lumbar spinal cord of DKO mice relative to WT littermate mice at P5 but not at P0. (C) The number of motor neurons in the lumbar spinal cord of KO and WT mice was not significantly different at P0 and P3. However, severe motor neuron loss was observed by P5 in DKO mice (n = 3) compared with WT mice (n = 3, *p < 0.05 by Student's t-test). (D) Glial fibrillary-associated protein (GFAP) and ionized calcium-binding adapter molecule 1 (IBA-1) immunostaining in the spinal cord of WT and DKO mice. GFAP-positive astrocytes and IBA-1-positive microglial cells surrounding motor neurons (counterstained with Giemsa) of DKO mice showed swelling cell bodies and processes, a typical sign of astrogliosis. (E,F) Electron microscopic analysis of spinal roots showed that DKO mice at age P5 had severe damage of motor neuron axons in the ventral root (E) but not of sensory neuron axons in the dorsal root (F). Bar, 2 µm. (G,H) Damaged axons in the ventral root showed either vacuoles or loss of axoplasm (G), as an early sign of axonal damage. At a later stage of axonal damage, dark disintegrated myelin sheaths and amorphous lipid were present (H). Bar, 1 µm. (I) Percentage of injured axons in the ventral and dorsal roots of DKO and WT mice (*p < 0.05 by Student's t-test, n = 3). Data are mean ± s.e.m.

DOI: http://dx.doi.org/10.7554/eLife.06500.003

Since both membralin and GluN3B are mutated in DKO mice, we needed to determine which gene was responsible for the motor neuron degeneration. Therefore, we tested whether transgenically expressed intact membralin or GluN3B could rescue DKO mice. We generated a transgenic mouse line carrying full-length membralin under the control of the murine prion promoter, followed by internal ribosome entry site (IRES) and enhanced green fluorescent protein (EGFP) cDNA sequences (Figure 2A). The membralin transgenic mice [Tg (membralin)] were fertile, normal in size, and did not display any gross physical or behavioral abnormalities compared with their littermate controls. When Tg (membralin) mice were crossbred with DKO mice, the DKO/Tg (membralin) mice were apparently normal, manifesting neither paresis nor premature death (Figure 2B). Thus, the membralin transgene rescued DKO mice, even though membralin mRNA levels remained somewhat lower than in WT mice (Figure 2C). GFP immunostaining in the ventral horn of the spinal cord showed that membralin/GFP transgene expression was predominantly neuronal, although we could not rule out relatively weak expression in glial cells (Figure 2—figure supplement 1). The DKO/Tg (membralin) mice showed no sign of motor defects or motor neuron loss, despite the absence of GluN3B expression (Figure 2D). In contrast, the GluN3B transgene did not rescue DKO mice (Figure 2—figure supplement 2), and GluN3B KO mice generated by deleting the first exon did not show any defect in motor neurons (Niemann et al., 2007). Taken together, these findings are consistent with the hypothesis that the loss of intact membralin in DKO mice causes motor neuron degeneration.10.7554/eLife.06500.008Figure 2.

(A) Generation of the transgenic constructs expressing membralin. The murine prion promoter was cloned with full-length mouse membralin cDNA followed by the IRES and the EGFP sequence. The entire transgenic sequence was isolated by enzymatic digestion and used to generate Tg (membralin) mice. (B) DKO/Tg (membralin) mice were viable and fertile. Left: genotyping by PCR showed that 3 littermates (#3, 4 and 5) from a breeding of DKO and Tg (membralin) mice were positive to DKO primers and negative to WT primers (same primers used as in Figure 1). Of the three DKO mice, the two receiving membralin transgene (#3 and 5, as positively identified by GFP primers) were rescued. Right: Weight of the 7 littermates was monitored after birth. Only DKO mouse #4, which was negative for the membralin transgene, showed weight loss after P3 and died at P5.5. Data shown here are from one representative litter (5 independent litters were analyzed with similar results). (C) Gene expression analysis by RT-PCR for mice #1, 3, 4 of the same litter as in B. Membralin transgene expression is seen in the brains of littermates #1 and 3, but not 4, whereas GluN3B is only expressed in #1. Membralin and GluN3B are not expressed in muscle samples. GAPDH is expressed in both brain and muscles of all mice and served as a positive control. (D) Rescued DKO/Tg (membralin) mice lived to adulthood without any sign of paresis.

DOI: http://dx.doi.org/10.7554/eLife.06500.008

Membralin is a novel protein with no conserved domains. RT-PCR analysis showed that membralin mRNA was expressed in multiple tissues, including brain and spinal cord, but not heart, skeletal muscle, or spleen (Figure 4A). Western blot analysis of spinal cord samples from WT and KO mice showed specific expression of membralin in the ER but not in the cytosol or mitochondria (Figure 4B). Since our anti-membralin antibody proved ineffective for immunostaining, we made a fusion protein by adding a myc tag to the C-terminus of membralin (membralin-myc). Immunocytochemistry demonstrated that membralin-myc co-localized with an ER marker PDI but not with markers for Golgi or mitochondria (Figure 4C). Membralin-myc was not expressed on the cell surface, since it was not detected in transfected cells without first permeabilizing the membrane (data not shown). Software for topology prediction from the Center for Biological Sequences (CBS, TMHMM) indicated that the membralin is most likely a transmembrane protein with 4–6 membrane-spanning regions, containing an ER-lumen domain between membrane-spanning regions 1 and 2. Additionally, the protein has multiple glycosylation sites and cytosolic regions at both the N-terminal ends and C-terminal ends (Figure 4—figure supplement 1A,B). These data suggest that membralin is a membrane protein most likely localized to the ER and/or outer nuclear envelope.10.7554/eLife.06500.012Figure 4.

(A) Semi-quantitative RT-PCR analysis showed the expression of membralin mRNA in brain (1), spinal cord (2), lung (4), liver (5), and kidney (6), but not in heart (3), spleen (7), and muscle (8) tissues of WT mice. (B) Western blot analysis showed the expression of membralin protein in the ER fraction, but not in the mitochondrial fraction (Mit) or the cytosolic fraction (Cyt), of the WT mouse brains. (C) Subcellular localization of Myc-tagged membralin in HEK 293 cells. Cells were transiently transfected with Myc-tagged membralin. At 48 hr after transfection, localization of membralin was detected by confocal microscopy (Zeiss 710). Organelles were labeled with specific markers (anti-PDI for ER, anti-mitochondria for mitochondria, and anti-GMP130 for Golgi). Region of interest shown at higher magnification in insets (bottom right corner). Merged pictures show membralin co-localized with the ER marker, PDI.

DOI: http://dx.doi.org/10.7554/eLife.06500.012

Increased ER stress is thought to be involved in some forms of motor neuron disease, including ALS (Atkin et al., 2006; Nagata et al., 2007; Nishitoh et al., 2008; Kanekura et al., 2009; Saxena et al., 2009). We, therefore, tested whether the absence of membralin impacted ER stress (Walter and Ron, 2011). Western blot analysis showed that the levels of GRP78 and ATF4, two molecules induced during ER stress via activation of the PERK signaling pathway, were consistently higher in the spinal cord of P3 membralin KO mice than WT mice (154 ± 35% and 412 ± 39% of WT, respectively; KO, n = 4; WT, n = 3; Figure 6A). In contrast, levels of spliced XBP-1, indicating activation of the IRE1 signaling pathway during ER stress, were not altered (Figure 6—figure supplement 1A). Furthermore, tunicamycin, an inducer of ER stress, produced a greater degree of cell death in MEFs from KO mice than from WT mice (Figure 6B). Tunicamycin exposure also increased CHOP and ATF4 levels earlier and to a greater extent in MEFs from KO mice; GRP78 levels were also higher at 24 hr post-exposure (Figure 6C). Basal levels of ATF4, but not CHOP, were higher in membralin KO mice than WT mice (Figure 6C), but XBP-1 splicing was not altered (Figure 6—figure supplement 1B). Collectively, these data suggest that loss of membralin increases basal ER stress and makes cells more vulnerable to additional ER stress-induced injury.10.7554/eLife.06500.016Figure 6.

(A) Upregulation of GRP78 and ATF4 in spinal cord of membralin KO mice. Left: immunoblot analysis shows higher levels of GRP78 and ATF4 in the spinal cord of P3 membralin KO mice compared to that of littermate WT mice. Right: increased expression level of GRP78 and ATF4 normalized to actin in membralin KO mice over WT mice. (B) Survival of MEFs after 24-hr exposure to tunicamycin or thapsigargin determined by a cytotoxicity assay. (C) The expression level of GRP78, CHOP, and ATF4 in MEFs from KO and WT mice is shown after exposure to tunicamycin at different time points. Levels of CHOP and ATF4 are elevated earlier and higher in MEFs from KO than that from WT mice. A representative immunoblot is shown; graphs include data from three experiments. For each panel, data are mean ± s.e.m.; n = 3 for each bar; *p < 0.05 by Student's t-test.

DOI: http://dx.doi.org/10.7554/eLife.06500.016

We used the 3.6 kb (Apa I/Spe I) and 3.2 kb (BamH I/Kpn I) DNA fragments flanking the target region as the 5′- and 3′-arms, respectively, for the ∼11.4 kb pGTN29/GluN3B/membralin targeting vector (see Figure 1—figure supplement 1A). Homologous recombination generated a ∼5 kb size difference between the original and recombinant genes, facilitating their analysis. During construction of the targeting vector, one SpeI site at the junction of the 5′-arm and the target region were eliminated, and another SpeI site was introduced at the junction of the target region and 3′-arm. These restriction site variations allowed us to designed a 3′ probe, a 436 bp fragment from Xho I digestion, that hybridized to the 14 kb and 6 kb fragments generated by SpeI digestion or 9.1 kb and 8.4 kb fragments generated by EcoR I digestion (see Figure 1—figure supplement 1A,B) from the WT and recombinant alleles, respectively.

All described procedures for animal were approved by the Institutional Animal Care and Use Committee of Sanford–Burnham Medical Research Institute and conducted in compliance with the Guide for the Care and Use of Laboratory Animals. Both sexes of mice were used for experiments and maintained in an institute facility accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC). The C1 ES cells were cultured in the presence of leukemia inhibitor factor on primary embryo fibroblast feeder cells and were transfected with the linearized pGTN29/GluN3B/membralin target vector DNA by electroporation at 400 V, 25 mF. Thirty-six hours after transfection, 0.48 mg/ml G418 was added to select for cells that have acquired the Neo^r gene by homologous recombination with the targeting vector. The ES cells with the correct recombination were confirmed by standard Southern blot analysis using the probe described above. GluN3B/membralin-targeted ES cell lines were used for blastocyst injection. Then, 6 to 10 embryos, including 2 uninjected blastocysts as carriers, were transferred into one uterine horn of each pseudopregnant foster mother (F1 of DBA X C57BL/J6j). Chimeric mice were identified by eye pigmentation. Further breeding was made to test for germ line transmission of the injected ES cells. Mouse tail DNA was used to identify homozygotes and heterozygotes by the same procedure used in testing ES cells. Once the DKO line was established, we also designed a PCR method to simplify genotyping using a pair of primers for the GluN3B sequence in WT alleles and a pair of primers for the Neo sequence in the DKO allele (see Supplementary file 1).

We have obtained the mouse full-length GluN3B cDNA clone from Dr Yuzaki as a gift (Matsuda et al., 2002). A full-length mouse membralin cDNA clone (ID# 3813678) was purchased from distributors of the Integrated Molecular Analysis of Genomes and their Expression (I.M.A.G.E) consortium (http://www.imageconsortium.org/). We used the mouse GluN3B promoter that drives gene expression mainly in motor neurons from late embryonic stages to adult (database from NINDS Gensat Bac Transgenic Project, http://www.gensat.org/index.html). In addition, we used the mouse prion promoter, which drives high transgene expression in mouse neurons (Baybutt and Manson, 1997; Loftus et al., 2002; Gispert et al., 2003), to express the transgene in a broader brain area. The selected full-length transgene was subcloned downstream of either promoter, followed by an IRES or a sequence coding the GFP. The insert containing promoter, transgene, and IRES/GFP sequences were isolated by restriction digest followed by gel purification. The purified DNA fragments were injected into fertilized eggs to generate transgenic mice (C57BL/J6j) in our institute's Transgenic Facility according to established protocol. We used PCR methods to genotype transgenic mice using a pair of primers (see Supplementary file 1).

Hemizygotes of transgenic mice were bred with DKO mice to obtain double heterozygotes [transgene^+/−/ (GluN3B/membralin)^+/−]. The double heterozygotes were further bred to obtain mice that express the transgene in the GluN3B/membralin null background [transgene^+/−/(GluN3B/membralin)^−/−], and tested for GFP and transgene expression using RT-PCR. The crossed transgene/DKO mouse line was subjected to histological tests for examining motor neuron integrity. The survival time of the [transgene^+/−/(GluN3B/membralin)^−/−] mice were monitored and compared with that of DKO mice to determine the ability of the transgene to rescue the DKO mice phenotype.

We generated membralin KO mice by using ES cells (gift of Sanger Institute) with a disrupted membralin gene by a gene trapping method (Brennan and Skarnes, 1999; Skarnes et al., 2004). Briefly, a trapping vector containing an RNA splicing acceptor sequence was inserted between exon 1 and 2 of the membralin gene in ES cells to disrupt normal RNA splicing. The positive ES cell clone was confirmed by sequencing and used for blastocyst injection to generate membralin KO mice (C57BL/J6j). RT-PCR experiments using primers from exon 1 and 2 further confirmed that homozygote mice carrying membralin/trapping vector alleles did not express membralin mRNAs with sequences beyond exon 2.

Previously established protocols were used for staining spinal cord sections or cultured cells (Xing et al., 2006), with the following antibodies: mouse anti-GFAP (Sigma-Aldrich, St. Louis, MO), rabbit anti-IBA-1 (Wako Chemicals, Inc., Richmond, VA), and anti-Hb9 (a gift from Dr Pfaff), followed by incubation with either a fluorophore-attached (Life Technologies, Grand Island, NY) or a biotinylated secondary antibody (Vector Laboratories, Burlingame, CA). Immunosignals were detected either directly under epifluorescence microscopy or after using a Vectastain Elite ABC kit (Vector Laboratories) with 3,3′-diaminobenzidine visualization (Roche Applied Science, Indianapolis, IN). The number of motor neurons in layers VIII and IX of the lumbar spinal cord was counted using stereological methods in WT, DKO, and membralin KO mice (Coggeshall and Lekan, 1996). Briefly, consecutive sections (12 μm in thickness) of the lumber enlargement (L1–L5) were collected for immunostaining with anti-Hb9 antibody to identify motor neurons. The number of motor neurons in each section was counted stereologically using adjacent sections for reference and look-up (physical disector). The total number of motor neurons was obtained for each mouse, and the mean value for 3 mice at each age group was calculated and normalized to that of WT mice. The general appearance of motor neurons was examined by conventional staining methods used in analysis of SOD1 mutant mice, including hematoxylin/eosin and cresyl violet staining.

HEK 293 cells were used for investigating the subcellular localization of membralin. A membralin-myc fusion protein was constructed by tagging the C-terminal of membralin with a Myc sequence using the pcDNA3.1 (−) Myc/his vector (Life Technologies). HEK 293 cells were plated onto coverslips, transfected with the membralin-Myc overnight, and fixed 48 hr after transfection with 4% PFA with 0.5% Triton X-100 in PBS. After blocking with 10% normal goat serum in PBS for 60 min, the cells were double stained by anti-Myc rabbit polyclonal antibody (Sigma) for membralin-Myc detection, followed by Alexa fluor-conjugated secondary antibodies (Life Technologies). Subcellular organelles were stained by either anti-PDI antibody (Enzo Life Science, Farmingdale, NY) for ER labeling, anti-GMP130 antibody (BD bioscience, Franklin Lakes, NJ) for Golgi labeling, or anti-mitochondria antibody (113-1, Abcam, Cambridge, MA) for mitochondrial labeling, followed by Alexa fluor-conjugated secondary antibodies (Life Technologies). DAPI was used to stain nuclei. Images were captured under confocal microscopy (Zeiss 710).

Dorsal and ventral roots of lumbar segments (L3–L5) of the spinal cord were dissected out from P5 WT and DKO mice. Samples were immediately placed in 4% paraformaldehyde plus 1% glutaraldehyde in 0.1 M phosphate buffer and simultaneously processed for electron microscopy as described previously (Liu et al., 1998). Briefly, samples were osmicated in 1% OsO₄ for 10–20 min followed by washing in 0.1 M phosphate buffer and then dehydrated in graded ethanol and 100% acetone. Each sample was oriented and placed in a Flat Embedding Mold (Ted Pella, Redding, CA) filled with Araldite (EMS, Fort Washington, PA). Embedded dorsal or ventral roots of different genotypes were dissected, and the cross sections of the roots were re-cut on an ultramicrotome (Ultracut, Leica) at 70–80 nm. Serial thin sections were collected on Formva-coated single slot nickel grids and stained with uranyl acetate and lead citrate. Ultrathin sections were examined in a Philips CM120 electron microscope at 80 KV. Digitized images were acquired by a high-resolution (2K × 2K) CCD camera (Gatan, Inc., Pleasanton, CA), processed using software provided by the manufacturer (DigitalMicrograph), and displayed with Photoshop CS (Adobe Systems, San Jose, CA).

Two DNA fragments were purified by enzymatic digestion of membralin cDNA using BamH I/Bgl II and Not I/BamH I, respectively, to generate two probes specific to regions of exon VII-X and the last exon (XI) of membralin. Northern blot analysis was performed using these two probes labeled with ³²P by random priming. 10 μg of total mouse RNA per lane were analyzed by electrophoresis on a 1.1% denaturing gel and subsequently transferred to a nylon membrane. The blot was hybridized at 42°C in a solution containing 50% formamide, 6x SSC, 5x Denhardt's reagent, and 0.5% SDS. The blots were dried and exposed to autoradiographic films for analysis.

cDNAs for mouse membralin and Erlin2 were purchased from distributors of the Integrated Molecular Analysis of Genomes and their Expression (I.M.A.G.E) consortium (http://www.imageconsortium.org/) and subcloned into a pcDNA3.1 vector for mammalian cell expression. C-terminal tagged membralin (membralin-myc or membralin-mCherry) and N-terminal or C-terminal tagged Erlin2 (Erlin2-EGFP or Erlin2-HA) were constructed by introducing corresponding tag fragments generated by the PCR method, and all constructs were confirmed by sequencing. Membralin-myc and Erlin2-HA were co-transfected in HEK 293 cells and total cell lysates collected in 48 hr after transfection for co-immunoprecipitation-immunoblot assay.

The interaction of Erlin2-EGFP and membralin-mCherry was validated by the acceptor photobleaching method for FRET detection (Karpova and McNally, 2006). Briefly, HEK 293 cells were transfected with Erlin2-EGFP and membralin-mCherry for 24 hr. A Zeiss 710 NLO microscope (Carl Zeiss Inc.) was used to record the fluorescence of EGFP and mCherry in transfected cells. Three pre-bleached and five post-bleached images were acquired. Averaged fluorescence intensities of the donor were calculated from the measurement of regions of interest for each experimental set before and after bleaching. The efficiency of FRET was calculated by E_fret = 1 – (I_a/I_b), where I_a and I_b represent the steady-state donor fluorescence in the presence and the absence of the acceptor, respectively. Seven Ala-linked EGFP-mCherry plasmids were transfected into HEK 293 cells and were used as a positive FRET control (FRET efficiency = ∼50%, data not shown). FRET efficiencies were reported as mean ± s.e.m.

We measured levels of GRP78, CHOP, and ATF4, components of three canonical branches of UPR, by Western blot analysis using the spinal cord tissues and MEFs from membralin KO mice and WT littermates. The antibodies against these proteins were purchased from commercial sources: GRP78 (H-129, 1:1000; Santa Cruz Biotechnology, Santa Cruz, CA), CHOP (2895S, 1:1000; Cell Signaling, Denvers, MA), and ATF4 (SC200, 1:500; Santa Cruz Biotechnology). Total RNA was isolated from the spinal cord of membralin mutant mice and WT littermate using TRIzol reagent (Life Technologies). The following set of primers was used to detect the expression of mouse Xbp-1 (GATCCTGACGAGGGTCCAAGA and ACAGGGTCCAACTTGTCCAG).

MEFs from membralin KO mice and WT littermates were isolated from day 12.5 embryos and cultured by conventional methods. In the toxicity assay, the percentage of cell death was measured by counting cells or using the CellTiter 96 Aqueous Non-Radioactive Cell Proliferation Assay kit (Promega, Madison, WI) after MEF cells were exposed to the ER stress inducer, tunicamycin, for 24 hr.

The sample size was 3–10 per genotype for animal histology and 3–6 for protein or transcript expression as well as for cell viability studies, as determined by Power Analyses of previous data. The experiments were not randomized. Although the initial investigator performing the experiments was not blinded, the samples and animal results were then examined by a group of the investigators masked to experimental identity in order to evaluate the results. All data points demonstrated a normal distribution and were all included in the analysis. Data are presented as mean ± s.e.m. and analyzed by a Student's t-test for pairwise comparisons. Statistical analyses were conducted using GraphPad Prism software (version 6). A p value <0.05 was considered statistically significant.