Transglutaminase 2 accelerates neuroinflammation in amyotrophic lateral sclerosis through interaction with misfolded superoxide dismutase 1

Authors

  • Miki Oono,

    1. Molecular Neuroscience Research Center, Shiga University of Medical Science, Otsu, Shiga, Japan
    2. Department of Neurology, Kyoto University Graduate school of Medicine, Kyoto, Japan
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Ayako Okado-Matsumoto,

    1. System Glycobiology Research Group, RIKEN, Wako, Saitama, Japan
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Akemi Shodai,

    1. Molecular Neuroscience Research Center, Shiga University of Medical Science, Otsu, Shiga, Japan
    Search for more papers by this author
  • Akemi Ido,

    1. Molecular Neuroscience Research Center, Shiga University of Medical Science, Otsu, Shiga, Japan
    Search for more papers by this author
  • Yasuyuki Ohta,

    1. Department of Neurology, Okayama University Graduate School of Medicine, Dentistry and Pharmaceutical Sciences, Okayama, Japan
    Search for more papers by this author
  • Koji Abe,

    1. Department of Neurology, Okayama University Graduate School of Medicine, Dentistry and Pharmaceutical Sciences, Okayama, Japan
    Search for more papers by this author
  • Takashi Ayaki,

    1. Department of Neurology, Kyoto University Graduate school of Medicine, Kyoto, Japan
    Search for more papers by this author
  • Hidefumi Ito,

    1. Department of Neurology, Wakayama Medical University, Graduate School of Medicine, Wakayama, Japan
    Search for more papers by this author
  • Ryosuke Takahashi,

    1. Department of Neurology, Kyoto University Graduate school of Medicine, Kyoto, Japan
    Search for more papers by this author
  • Naoyuki Taniguchi,

    Corresponding author
    1. System Glycobiology Research Group, RIKEN, Wako, Saitama, Japan
    • Address correspondence and reprint requests to Makoto Urushitani, Molecular Neuroscience Research Center, Shiga University of Medical Science, Seta-Tsukinowa-cho, Otsu, Shiga 520-2192, Japan. E-mail: uru@belle-shiga-med.ac.jp or Naoyuki Taniguchi, System Glycobiology Research Group, RIKEN, 2-1 Hirosawa, Wako, Saitama 351-0198, Japan. E-mail: tani52@wd5.so-net.ne.jp

    Search for more papers by this author
  • Makoto Urushitani

    Corresponding author
    1. Molecular Neuroscience Research Center, Shiga University of Medical Science, Otsu, Shiga, Japan
    • Address correspondence and reprint requests to Makoto Urushitani, Molecular Neuroscience Research Center, Shiga University of Medical Science, Seta-Tsukinowa-cho, Otsu, Shiga 520-2192, Japan. E-mail: uru@belle-shiga-med.ac.jp or Naoyuki Taniguchi, System Glycobiology Research Group, RIKEN, 2-1 Hirosawa, Wako, Saitama 351-0198, Japan. E-mail: tani52@wd5.so-net.ne.jp

    Search for more papers by this author

Abstract

Although the aberrant assembly of mutant superoxide dismutase 1 (mSOD1) is implicated in the pathogenesis of familial amyotrophic lateral sclerosis (ALS), the molecular basis of superoxide dismutase 1 (SOD1) oligomerization remains undetermined. We investigated the roles of transglutaminase 2 (TG2), an endogenous cross-linker in mSOD1-linked ALS. TG2 interacted preferentially with mSOD1 and promoted its oligomerization in transfected cells. Purified TG2 directly oligomerized recombinant mutant SOD1 and the apo-form of the wild-type SOD1 proteins in a calcium-dependent manner, indicating that misfolded SOD1 is a substrate of TG2. Moreover, the non-cell-autonomous effect of extracellular TG2 on the neuroinflammation was suggested, since the TG2-mediated soluble SOD1 oligomers induced tumor necrosis factor-α, interleukin-1β, and nitric oxide in microglial BV2 cells. TG2 was up-regulated in the spinal cord of pre-symptomatic G93A SOD1 transgenic mice and in the hypoglossal nuclei of mice suffering nerve ligation. Furthermore, inhibition of spinal TG2 by cystamine significantly delayed the progression and reduced SOD1 oligomers and microglial activation. These results indicate a novel role of TG2 in SOD1 oligomer-mediated neuroinflammation, as well as in the involvement in the intracellular aggregation of misfolded SOD1 in ALS.

image

A new role of transglutaminase 2 (TG2) in misfolded SOD1-linked neuroinflammation has been clarified. TG2 recognized and oligomerized only misfolded forms of SOD1, which robustly activated microglia, and induce the expression of proinflammatory molecules such as iNOS, IL-1β, and TNF-α. The inhibition of spinal cord TG2 of mutant SOD1 transgenic mice successfully suppressed neuroinflammation and delayed the progression.

Abbreviations used
ALS

amyotrophic lateral sclerosis

IL-1β

interleukin-1β

iNOS

inducible nitric oxide synthase

SOD1

superoxide dismutase 1

TG2

transglutaminase 2

TNF-α

tumor necrosis factor-α

Amyotrophic lateral sclerosis (ALS) is characterized by selective loss of motor neurons with unknown etiology. Genetic mutations in superoxide dismutase 1 (SOD1) account for 2% of all ALS cases and are the most common cause of familial ALS. Diverse biological pathways of mutant SOD1 (mSOD1)-dependent motor neuron degeneration have been proposed, including mitochondrial dysfunction, proteasome impairment, endoplasmic reticulum stress, neuroinflammation, and excitotoxicity (Lambrechts et al. 2007; Ilieva et al. 2009), all of which are induced by misfolded mSOD1 species. Although the molecular pathologies shared by more than 150 types of mSOD1 proteins remain elusive, aberrantly assembled forms of disease-causing proteins have been implicated in diverse neurodegenerative disorders. Oligomeric forms of misfolded proteins may underlie several neurodegenerative disorders, including Alzheimer's disease (Benilova et al. 2012), Huntington's disease (HD) (Hoffner et al. 2007), and Parkinson's disease (Waxman and Giasson 2009). The presence of mSOD1 oligomers in ALS models has been documented in several previous reports (Urushitani et al. 2002; Furukawa et al. 2006; Liu et al. 2012). Cysteine residues reportedly mediated the formation of intermolecular disulfide bonds in SOD1 oligomers (Furukawa et al. 2006); however, a recent report suggested that this linkage may not be completely responsible for mSOD1 oligomer formation (Roberts et al. 2012). Moreover, the exact mechanism through which oligomeric SOD1 confers neurotoxicity has not been clarified.

In this study, we investigated the role of an endogenous cross-linker transglutaminase 2 (TG2) in mSOD1 oligomer formation. TG2 covalently binds glutamine and lysine residues in a calcium-dependent manner and is ubiquitously expressed in both intracellular and extracellular spaces of the central nervous system (Ruan and Johnson 2007). TG2 has been reported to oligomerize pathogenic proteins such as α-synuclein (Andringa et al. 2004), β-amyloid (Ho et al. 1994), and huntingtin (Htt) (Karpuj et al. 1999). In addition, targeting of TG2 has been shown to have therapeutic effects in animal models of HD (Karpuj et al. 2002) and TG2 was elevated in the brain and CSF of patients with ALS and HD (Fujita et al. 1998; Lesort et al. 1999). However, the role of the catalytic activity of TG2 in ALS pathogenesis remains unknown. We here show evidence that TG2 is involved in the aberrant assembly of misfolded SOD1 proteins, which contributes to neuroinflammation and disease progression in a mouse model of ALS.

Materials and methods

Plasmid construction and purification of recombinant proteins

Mammalian expression plasmids containing either wild-type (WT), Gly85Arg (G85R), or Gly93Ala (G93A) mutants of FLAG-tagged SOD1 (pcDNA3-hSOD1-FLAG), and Escherichia coli expression plasmids containing WT or a G93A mutant of human SOD1 (pGEX6p-1-hSOD1) were constructed as reported previously (Urushitani et al. 2002). The human TG2-coding region was subcloned into a pcDNA3 plasmid (Invitrogen, Carlsbad, CA, USA) containing a hemagglutinin (HA)-tag at the 3′ end to generate an HA-tagged protein by conventional PCR using pcDNA3.1-TG2 (Tucholski et al. 2001) as a template (a generous gift from Dr Gail V. W. Johnson at the University of Rochester, NY, USA). pCMV-Myc-α-synuclein plasmid was a gifted from Dr. Masafumi Ihara (Ihara et al. 2003). Recombinant human SOD1 proteins were purified as described previously (Urushitani et al. 2002). To generate the apo-form, hSOD1 proteins were demetallated by overnight incubation in 100 mM EDTA, and in acetate pH 3.8 for additional overnight. Finally, proteins were dialyzed against phosphate-buffered saline (PBS). Metallation was performed to generate holo-SOD1, by incubation in two equivalent parts of zinc chloride for 24 h, followed by further incubation with two-equimolar copper chloride for 24 h. Lipopolysaccharide was eliminated by intensive dialysis and with an endotoxin removal beads (Miltenyi Biotec Inc., Auburn, CA, USA). Synthetic WT SOD1 proteins generated from 2-mercaptoethanol-fused SOD1 were used for size-separation of the gel-filtrated oligomers (Fujiwara et al. 2007).

Antibodies and chemicals

The rabbit polyclonal anti-SOD1 antibody and anti-cyclooxygenase-2 were purchased from Enzo Life Sciences (Farmingdale, NY, USA). The rabbit polyclonal anti-TG2 antibody was obtained from Abcam (Cambridge, UK) and Cell Signaling (Danvers, MA, USA). The rabbit anti-glyceraldehyde-3-phosphate dehydrogenase (GAPDH) antibody was obtained from Santa Cruz Biotechnology (Santa Cruz, CA, USA). The mouse rabbit polyclonal anti-FLAG antibody was obtained from Abnova (Taipei, Taiwan). The rat monoclonal anti-HA antibody (3F10) was purchased from Roche (Basel, Switzerland). The rat monoclonal antibodies against Mac2 (TIB-166) and CD11b (M1/70) were obtained from the American Type Culture Collection (Manassas, VA, USA) and Abcam, respectively. The mouse monoclonal antibody against NeuN and glial fibrillary acidic protein (GFAP) were obtained from Millipore (Billerica) and Sigma-Aldrich (St. Louis, MO, USA), respectively. Mouse monoclonal recognizing misfolded human SOD1 proteins (C4F6 and D3H5) was obtained as described previously (Urushitani et al. 2007; Okamoto et al. 2011). Peroxidase-conjugated secondary antibodies were obtained from Jackson ImmunoResearch Laboratories (West Grove, PA, USA). Unless otherwise mentioned, all chemicals were purchased from Nacalai Tesque (Kyoto, Japan).

Cell culture, transfection, and immunoprecipitation

HEK293A cells and COS-7 were maintained in Dulbecco's modified Eagle's medium (Nacalai Tesque) containing 10% heat-inactivated fetal bovine serum. Murine microglial BV-2 cells were maintained at 37°C and 5% CO2 in Dulbecco's modified Eagle's medium/Ham's F-12 medium (Nacalai Tesque) supplemented with 10% fetal bovine serum. In immunoprecipitation studies, the FuGene HD transfection reagent (Roche) was used to co-transfect HEK293A cells with plasmids expressing FLAG-tagged WT, G85R, or G93A mutant human SOD1 and HA-tagged human TG2. Twenty-four hours after transfection, the cells were treated with the proteasome inhibitor lactacystin for a further 24 h to prevent the degradation of misfolded SOD1 species. Following this, the cells were harvested. Immunoprecipitation was performed as described in our previous report (Urushitani et al. 2006). In brief, the cells were washed twice with PBS and then dissolved in TNG-T buffer [50 mM Tris–HCl (pH 7.4), 150 mM NaCl, 10% glycerol, 1% Triton X100, and protease inhibitor cocktail] with brief sonication on ice. The cell lysates were centrifuged at 21,400 g for 20 min at 4°C. The supernatant was incubated with affinity gel coupled with anti-FLAG M2 (Sigma–Aldrich) or anti-HA (Roche) at 4°C overnight. The immunoabsorbed complex was washed five times with radio immunoprecipitation assay (RIPA) buffer [20 mM HEPES (pH 7.4), 150 mM NaCl, 2 mM EDTA, 1% Nonidet P-40, 1% sodium deoxycholate, and 0.1% sodium dodecyl sulfate (SDS)]. The precipitates were released from affinity gels by heating to 70°C for 20 min in 2% SDS sampling buffer containing 100 mM dithiothreitol (DTT) and then analyzed by immunoblotting.

Immunofluorescence of cultured cells and confocal laser microscope

At 48 h after transfection, the cells expressing enhanced green fluorescent protein (EGFP)-fused SOD1 and/or HA-tagged human TG2 in multichamber slides (Lab-tek; Thermo–Fischer Scientific, Waltham, MA, USA) were fixed in 4% paraformaldehyde for 20 min at 22°C. After blocking and membrane permeabilization with 4% normal goat serum and 0.2% of Triton X100 in PBS, the cells were incubated with rat monoclonal antibody against HA [1:500 in PBS containing 0.1% Triton X100 (PBS-T)] for 1 h at 22°C. After washing three times in PBS-T, the cells were reacted with CF-568-conjugated anti-rat IgG (1 : 500 in PBS-T) for 30 min at 22°C. 4′,6-diamidino-2-phenylindole (DAPI) was finally applied to counter-staining. An inverted confocal laser microscope (C1si; Nikon, Tokyo, Japan) and the analytical software (EZ-C1 viewer) bundled with the microscope were used to view the stained cells.

In vitro or in vivo oligomerization of recombinant SOD1 proteins by TG2

Native human SOD1 from erythrocytes (Sigma-Aldrich) and WT or G93A mSOD1 proteins purified from E coli were incubated with 10 nM guinea-pig TG2 (ORIENTAL YEAST CO., LTD, Tokyo, Japan) for 1 h at 37°C in a buffer containing 50 mM Tris–HCl (pH 7.5) and 0.1 mM DTT, with varying concentrations of CaCl2 (0–5 mM), with or without EDTA (20 mM) (Junn et al. 2003). The reaction was stopped by heating for 5 min at 95°C with SDS sample buffer and the products were analyzed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE), followed by immunoblotting using a mouse monoclonal antibody against SOD1 (D3H5). To analyze the TG2-mediated oligomerization of SOD1 or α-synuclein, HEK293A cells were co-transfected with plasmids for human TG2 and FLAG-tagged SOD1 or cMyc-tagged α-synuclein for 92 h, with or without calcium ionophore ionomycin, or TG2 inhibitor cystamine, for 24 h before the cell harvest. The cells were suspended in RIPA buffer eliminating SDS (RIPA-), and centrifuged at 17 400 g at 4°C for 30 min. The supernatant and the resultant pellets were resuspended in 2% SDS-sampling buffer, and were analyzed by western blotting, designated as detergent-soluble and -insoluble fractions, respectively.

SDS–PAGE, perfluoro-octanoic acid–PAGE, immunoblotting

For SDS–PAGE, cell lysates or mouse tissue lysates were mixed with sampling buffer containing 2% SDS and 100 mM DTT, and then incubated for 20 min at 70°C. Perfluoro-octanoic acid–PAGE was performed to assess the molecular weight of SOD1 oligomers containing non-covalent bonds because this technique preserves protein–protein interactions, according to the previous report (Ramjeesingh et al. 1999). The membranes were probed with primary antibodies in blocking buffer and subsequently with peroxidase-conjugated secondary antibodies in Tris-buffered saline with 0.1% Triton X-100 (TBS-T). The membranes were washed three times with TBS-T between each step. Enhanced chemiluminescence substrate (Pierce, Rockford, IL, USA) was used to visualize target proteins.

Treatment of murine microglial BV-2 cells with TG2-mediated SOD1 oligomers

TG2-catalyzed SOD1 oligomers in PBS were separated into five fractions on the basis of size by gel filtration chromatography. After in vitro cross-linking of recombinant SOD1 using guinea-pig TG2, the protein mixtures were passed through a filtration membrane with a 100-kDa cutoff. Superdex 200 HR 10/30 columns with a flow rate of 0.25 min/mL were used to fractionate the residual oligomers (380 μg in 25 mL of PBS) by size, and five 0.5-mL fractions larger than 440 kDa were used. Thyroglobulin (669 kDa), ferritin (440 kDa), human IgG (150 kDa), human transferrin (81 kDa), ovalbumin (43 kDa), myoglobin (17.6 kDa), and vitamin B12 (1.35 kDa) as protein standards (all purchased from GE Healthcare, Buckinghamshire, UK) were used to calibrate the column for estimation of the molecular weight. Equal amounts of SOD1 monomer, guinea-pig TG2, and size-fractionated SOD1 oligomers (3 μg/mL) were added to the BV-2 nutrient medium. Twenty-four hours after the treatment, the cells and medium were collected for real-time PCR and suspension array analysis, respectively. The Trizol Plus RNA Purification Kit and Superscript III reverse transcriptase (Invitrogen) were used to prepare RNA and cDNA, respectively.

Cell death assay of motor neuron NSC-34 cells

A NSC-34 murine motor neuron cell line was purchased from CELLutions (Toronto, ON, Canada). The cells were plated onto 96-well multi-well culture plates at the density of 2 × 104 cells/mL. The cells were exposed to 1-Hydroxy-2-oxo-3,3-bis(2-aminoethyl)-1-triazene (NOC-18, DOJINDO, Kumamoto, Japan), a long-acting nitric oxide (NO) donor, murine tumor necrosis factor-α (TNF-α) (PEPROTECH, Rocky Hill, NJ, USA), or both at the indicated concentrations for 48 h. In a staurosporine experiment, chemicals were applied for 24 h. The cell viability was assessed by simultaneous measurement of indicators for dead cells and live cells (Multitox Fluor Multiplex Cytotoxicity Assay; Promega, Madison, WI, USA) according to the manufacture's protocol. To minimize the bias derived from well-to-well differences in cell density, cell toxicity was expressed as bis-AAF-R110 normalized to glycyl-phenylalanyl-amino-fluorocoumerin (GF-AFC). To measure caspase 3/7 activity and the viability of NSC-34 exposed to the agents simultaneously, the cells were treated with luminogenic caspase substrates containing peptides Asp-Glu-Val-Asp (DEVD) for caspase 3/7 and GF-AFC, and luminescence and fluorescence were obtained by the multi-plate reader (ApoLive-Glo Multiplex Assay; Promega).

Animal breeding and tissue sampling

Transgenic mice harboring human G93A SOD1 (B6SJLTgN[SOD1-G93A]1Gur, SOD1G93A; Jackson Laboratory, Bar Harbor, ME, USA) were backcrossed with the C57BL/6 strain for more than 20 generations (G93AGurdl). The mice were maintained as hemizygotes by breeding transgenic males with C57BL/6 females. Genotyping of G93A SOD1 transgenic (mSOD1) or non-transgenic (WT) mice were identified by PCR using primers specific for human SOD1 (Takeuchi et al. 2010). The spinal cord was homogenized in a tissue lysis buffer (20 mM HEPES–KOH (pH 7.4), 120 mM NaCl, 2 mM EDTA, 10% glycerol, 1% Triton X-100, and protease inhibitor cocktail) or RIPA buffer and then incubated for 1 h at 4°C. After centrifugation at 21 400 g for 20 min at 4°C, a Bradford assay reagent was used according to the manufacturer's protocol (Bio-Rad, Hercules, CA, USA) to determine the protein concentration of the supernatant. All study protocols were performed respecting the dignity of animal lives and were approved by the animal experimental committee of the Shiga University of Medical Science (Document# 2009-12-2).

Immunohistochemistry of murine spinal cords

After deep anesthesia with intraperitoneal infusion of 10% chloral hydrate, the mice were transcardially perfused with 0.1 M phosphate buffer and 4% paraformaldehyde. A 10-μm-thick microtome (Yamato Kohki, Saitama, Japan) was used to prepare spinal cord frozen slices. Subsequent to the quenching in 3% hydrogen peroxide for 10 min at 22°C, the slices were incubated in the blocking buffer containing 5% normal goat serum, 1% bovine serum albumin, and 0.1% TritonX100 in PBS for 30 min and then with the primary antibodies in the antibody buffer containing 1% bovine serum albumin in Tris-buffered saline for 1 to 4 days at 4°C and washed in the wash buffer containing 0.1% PBS-T three times at 22°C. For optical immunohistochemistry, the slices were incubated with a biotinylated secondary antibody (1:2000) for 1 h at 22°C and then reacted with the avidin–biotin immunoperoxidase complex (Vectastain ABC Elite Kit; Vector, Burlingame, CA, USA) according to the manufacturer's protocol. Antibodies were visualized by chromogen diaminobenzidine (Nacalai Tesque). For immunofluorescence study, the specimens were stained with secondary antibody conjugated with CF488 or CF568 for 1 h (1 : 500; Biotum, Hayward, CA, USA), and were counter-stained with DAPI. The dilution of the primary antibodies used in the immunohistochemistry is as follows; anti-TG2 (1 : 500), anti-C11b (1 : 100), anti-NeuN (1 : 1000), anti-GFAP (1 : 1000), D3H5 (1 : 1000).

Preparation of cDNA from mouse hypoglossal nuclei and real-time PCR

WT or the mSOD1 mice at 20 weeks of age were subjected to tightening of the right hypoglossal nerve. After 24 h, the mice were perfused, and laser capture microdissection was used to collect the hypoglossal nuclei. The Picopure RNA Isolation Kit (Arcturus, Mountain View, CA, USA) was used to purify mRNA from approximately 1500 cells, a mixture of oligo dT and random primers (PrimeScript RT Reagent Kit; Takara, Kyoto, Japan) was used to generate cDNA by reverse transcription. Gene induction in mouse hypoglossal nuclei and BV-2 cells was analyzed by real-time PCR using SYBR green I and Lightcycler 4800 (Roche), as described previously (Takeuchi et al. 2010). Primer sequences are provided in the Supplementary Information. GAPDH was used as an internal standard. A suspension array system (Bio-Plex Pro Mouse Cytokine; Bio-Rad) was used to measure of TNF-α in the culture media according to the manufacturer's protocol, as described previously (Takeuchi et al. 2010).

Intrathecal infusion of a TG2-inhibitor in mSOD1 mice

An osmotic mini-pump (Alzet; Durect, Cupertino, CA, USA) was used to intrathecally administer cystamine, an inhibitor of TG2 to the mSOD1 mice, as described previously (Ohta et al. 2006). Cystamine chloride salt (200 μL per mouse; Nacalai Tesque) was loaded into the pump, which constantly delivered solution for 42 days once installed in the mice subcutaneously. After determination of the optimal concentration of cystamine using WT C57BL/6 mice (60 mg/mL), the mSOD1 mice were treated at 30 weeks of age (i.e., the timing of clinical onset) (n = 8 for cystamine; n = 11 for the saline control). The therapeutic effects were evaluated by weekly measures of grip power and body weight and by life span, in which the inability of mice to turn around from the supine position within 30 s was used as the endpoint (Takeuchi et al. 2010). To investigate the effect of cystamine for SOD1 oligomers or neuroinflammation, mice were killed and total spinal cords were resected for western blotting 30 days after the pump installment.

Statistical analyses

Student's t-test was used for the comparison between two groups, while one-way anova with a multiple comparison test of the Bonferroni's method was used to judge the significance from more than three groups. Differences in chronological profiles of averaged parameters (body weight and grip power) were estimated by two-way anova. The effect of the treatment on the onset delay or longevity was estimated by Kaplan–Meier log-rank test. Statistical analysis was performed using Graphpad Prism 5 (Graphpad, San Diego, CA, USA).

Results

TG2 recognizes familial ALS-linked mutant SOD1 proteins

To test our hypothesis that familial ALS-linked mutant forms of SOD1 are recognized by TG2 such as Htt or α-synuclein, HEK293A cells were co-transfected with mammalian expression plasmids containing FLAG-tagged human SOD1 (WT, G85R, or G93A) and HA-tagged human TG2. Western blotting of the immunoprecipitates with anti-FLAG or anti-HA monoclonal antibodies, using anti-HA or anti-SOD1 antibodies, respectively (Fig. 1ai or ii, respectively), clearly showed that the TG2 protein was preferentially pulled down with the mSOD1 proteins (G85R and G93A) (Fig. 1ai and ii). As reported previously, G85R mutant showed a different migration pattern from WT and G93A mutant (Urushitani et al. 2002). We also confirmed that endogenous TG2 was pulled-down with over-expressed mutant SOD1 in COS-7 cells (Fig. 1aiii). Moreover, confocal microscopic analysis showed that HA-tagged TG2 colocalized with EGFP-tagged human SOD1 (WT, G85R, or G93A) in the transfected HEK293A cells (Fig. 1b). Notably, occasional WT SOD1 aggregates merged with TG2 (Fig. 1biv–vi), implying that misfolded forms of SOD1 were recognized by TG2.

Figure 1.

Transglutaminase 2 (TG2) preferentially interacts with amyotrophic lateral sclerosis (ALS)-linked mutant superoxide dismutase 1 (SOD1) in cultured cells. (a) HEK293A cells were co-transfected with plasmids expressing FLAG-tagged wild-type (WT), Gly85Arg (G85R) mutant, or Gly93Ala (G93A) mutant human SOD1, and hemagglutinin (HA)-tagged-human TG2. Immunoprecipitates using affinity gels coupled with anti-FLAG (i) or anti-HA (ii) antibodies were analyzed by immunoblotting using antibodies targeting HA (to detect TG2) and SOD1 to investigate their interaction. Arrowheads indicate the transfected human SOD1-FLAG protein. *Endogenous mouse SOD1. (iii) COS-7 cells were transiently transfected with human FLAG-tagged SOD1 of WT or mutants (G85R and G93A). VC indicates vector control. (b) Confocal micrographs of HEK293A cells co-transfected with EGFP-fused SOD1 and HA-tagged TG2. Cells were treated with the proteasome inhibitor lactacystin (10 μM) to induce SOD1 aggregates. Cytosolic aggregates of mutant SOD1 have been frequently colocalized with TG2 (arrowheads). Note that some rare aggregates of WT SOD1 also contain TG2 (arrow). Scale bars; 20 μm.

Misfolded forms of SOD1 are susceptible to cross-linking by TG2 both in transfected cells and in vitro

Next, the transamidation activity of TG2 against mSOD1 was investigated in transfected cells. Western blot analysis of HEK293A cells, transiently co-transfected with human SOD1 (WT or G93A mutant) and human TG2 for 92 h revealed that TG2 generated high molecular species of mSOD1, but not of WT only in the detergent-insoluble fraction (Fig. 2ai and ii). The oligomer formation was enhanced by the ionomycin, a calcium ionophore, and was reduced by TG2 inhibitor cystamine. α-synuclein also formed the oligomers in the presence of TG2 solely the detergent-insoluble fraction (Figure S1). In addition, we analyzed a direct effect of TG2 on the cross-linking of the SOD1 using in vitro oligomerization assay (Junn et al. 2003). The incubation of recombinant apo-G93A SOD1 with purified TG2 readily formed oligomers in a calcium-dependent manner (Fig. 2bi). This effect required CaCl2 concentrations > 0.5 mM (Fig. 2bii). Moreover, TG2 clearly oligomerized the apo form of WT SOD1 and both the apo- and holo-forms of the G93A SOD1 mutant, but did not oligomerize native WT SOD1 (Fig. 2biii). These data indicate that TG2 preferentially recognized ALS-relevant forms of SOD1 proteins and directly cross-link them. Recently, it is reported that physiological concentrations of calcium ion promote fibrillar formation of apo-WT SOD1 by overnight incubation with agitation at 37°C (Leal et al. 2013). However, misfolded SOD1 required TG2 together with 2.5 mM CaCl2 to form oligomers by 1 h static incubation at the same temperature, otherwise SOD1 remained as a monomer only with CaCl2 (Fig. 2biii).

Figure 2.

Transglutaminase 2 (TG2) cross-links and forms oligomers of only misfolded forms of superoxide dismutase 1 (SOD1) in vitro and in transfected cells. (a) Over-expressed TG2 forms oligomers of the mutant SOD1, but not the wild-type (WT) in the detergent-insoluble fraction. HEK293A cells were transiently transfected with human-TG2 together with FLAG-tagged SOD1 (WT, G93A) in the presence or absence of 0.5 μg/mL of ionomycin, a calcium ionophore to allow the calcium entry into cells, or 0.4 mM cystamine. Chemicals were applied for 24 h before the cell harvest. Ninety-two hours after the transfection, the cell lysates were separated into detergent-soluble (i) and -insoluble (ii) fractions, both of which were analyzed by western blotting using antibodies against FLAG. The asterisk indicates SOD1 monomer. G93A SOD1, but not the WT was prominently oligomerized in the presence of TG2 (lanes 7, 8), which was reduced by cystamine (lane 9). (b) In vitro oligomerization assay showing the effect of TG2 on SOD1 proteins. Panel (i) shows calcium-dependent formation of apo-SOD1 G93A oligomers in the presence of TG2. 1 μg of recombinant G93A apo-SOD1 protein was reacted with 10 nM guinea-pig TG2 in the presence of 5 mM CaCl2 for 1 h at 37°C. SOD1 remained monomeric when calcium was chelated by 5 mM EDTA. Oligomerization reaction was stopped by incubation in 20 mM EDTA. Panel (ii) shows a titration of the calcium concentration; SOD1 was efficiently oligomerized at calcium concentrations greater than 0.5 mM. Panel (iii) shows that misfolded forms of SOD1 were preferentially cross-linked by TG2. The apo form of WT SOD1 was more prominently cross-linked than the native form, whereas G93A SOD1 was readily oligomerized. The vertical bar indicates high-molecular-weight SOD1 species. *Monomeric SOD1. **Dimeric SOD1.

Microglial cells are activated by extracellular SOD1 oligomers of misfolded forms

Our data revealed that > 0.5 mM CaCl2 was required for SOD1 oligomerization in vitro (Fig. 2aii). Moreover, the double transfection did not accelerate cell death in neuronal cell-lines (data not shown). Because, TG2 is abundantly distributed in extracellular spaces and in the cytosol (Ruan and Johnson 2007) and the transamidation activity of TG2 in the cerebrospinal fluid of ALS patients was reported previously (Fujita et al. 1998), we focused on the extracellular space as the site responsible for TG2-mediated SOD1 oligomerization in ALS pathogenesis. We and others previously documented that extracellular SOD1 proteins activated immortalized or primary microglial cells, and injured motor neurons (Urushitani et al. 2006; Zhao et al. 2010). On the other hand, a recent report by Roberts et al. (2013), showed that the aggregate form of mutant SOD1 was a much more potent neurotoxin than the monomeric form. Therefore, murine BV-2 cells were used to examine the effect of TG2-catalyzed WT or G93A mutant apo-SOD1 oligomers on the activation of microglia. First, mRNA levels of the pro-inflammatory molecules interleukin-1β (IL-1β), TNF-α, and inducible nitric oxide synthase (iNOS) were measured by real-time PCR. As shown in Fig. 3a, TG2-mediated apo-G93A SOD1 oligomers augmented the effect of monomeric G93A. Notably, the WT oligomer acquired proinflammatory properties, which indicated that TG2 transformed WT SOD1 into a pathogenic species via cross-linking. The relationship between the pathogenicity and molecular size of highly assembled species of disease-linked proteins remains unclear. Thus, we used BV-2 cells to examine the effect of oligomer size on microglial activation. PBS-soluble components of TG2-induced WT SOD1 oligomers were size fractionated by FPLC, and five serial fractions (F1–F5) > 440 kDa were used (Fig. 3b). Real-time PCR analysis showed that the WT SOD1 oligomer induced IL-1β and TNF-α expression in an oligomer size-dependent manner (Fig. 3ci and ii). On the other hand, iNOS expression was readily activated by the small WT SOD1 oligomer but declined in response to the larger fraction (F5) (Fig. 3ciii). Similarly, an increase in the TNF-α protein in the medium, which displayed a similar trend to the results of real-time PCR, was confirmed by suspension array analysis of media from BV-2 cells exposed to WT SOD1 oligomers (Fig. 3d).

Figure 3.

Transglutaminase 2 (TG2)-mediated superoxide dismutase 1 (SOD1) oligomers induce pro-inflammatory factors in microglial cells. (a) Real-time PCR analyses of pro-inflammatory factors in BV-2 microglial cells following exposure to wild-type (WT) SOD1 monomer or oligomer or to mutant SOD (mSOD)1 monomer or oligomer. Oligomer indicates synthetic SOD1 oligomer, generated through the cross-linking by purified TG2. Expression levels of IL-1β (i), tumor necrosis factor-α (TNF-α) (ii), and inducible nitric oxide synthase (iNOS) (iii) were significantly increased by exposure to WT or mSOD1 oligomers (3 μg/mL). In panel (i): *p < 0.05 versus WT SOD1 monomer; **p < 0.05 versus WT SOD1 monomer; #p < 0.05 versus G93A SOD1 monomer. In panel (ii): *p < 0.05 versus WT SOD1 monomer, #p < 0.05 versus G93A SOD1 monomer. In panel (iii): *p < 0.05 versus control; #p < 0.05 versus WT SOD1 monomer; **p < 0.05 versus WT SOD1 monomer; ##p < 0.05 versus G93A SOD1 monomer. Statistical analyses were performed by one-way anova with Bonferroni's test. The data represent the mean ± SEM for n = 3. (b) Gel filtration chromatography analysis of TG2-catalyzed WT SOD1 oligomers separated into five fractions on the basis of size (440–670 kDa), separated by perfluoro-octanoic acid–polyacrylamide gel electrophoresis (PFO–PAGE). (c) Real-time PCR analysis showing the effect of SOD1 oligomer size on induction of IL-1β (i), TNF-α (ii), and iNOS (iii). Apo-WT SOD1 oligomers induced IL-1β and TNF-α in a size-dependent manner. Oligomers larger than the F5 fraction abolished the pro-inflammatory effect. Expression of iNOS was robustly induced by the F1 oligomer. In all lanes, cells were treated with the same concentrations of proteins (3 μg/ml) *p < 0.05 (vs. WT SOD1 monomer) by one-way anova with Bonferroni's test. The data represent the mean ± SEM for n = 3. (b) Gel filtration chromatography analysis. (d) Suspension array assay analysis of TNF-α protein levels in the culture medium of BV-2 cells exposed to WT SOD1 oligomer. The SOD1 oligomers increased the amount of the TNF-α protein, which is consistent with real-time PCR data. *p < 0.05 (vs. WT SOD1 monomer) by one-way anova with Bonferroni's. Data represent the mean ± SEM for n = 3.

TNF-α and NO synergistically induced apoptotic cell death of motor neuron NSC-34 cells

No significant damage was found in NSC-34 cells exposed to the conditioned medium of SOD1-oligomer-treated BV-2 by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay (MMT assay, data not shown). We thus tested the direct effect of NO and TNF-α on NSC-34 focusing on apoptosis because the mechanism underlying neuroinflammatory motor neuron death is undetermined. NOC-18 is a NO donor with a half-life as long as 21 h. NSC-34 motor neuron cells were exposed to NOC-18 or murine TNF-α for 48 h at the indicated concentrations. The cell toxicity estimated by the simultaneous measurement of the fluorescent indicators bis-AAF-R110 and GF-AFC for extracellular protease from dead cells and intracellular protease inside cells, showed that 48 h treatment of NSC-43 cells with NOC-18 induced cell death in a dose-dependent manner (Fig. 4a), whereas TNF-α alone did not induce toxicity, even at doses as high as 80 ng/mL (Fig. 4b). As reported by He et al. (2002), the addition of sublethal doses of TNF-α augmented the toxicity of NOC-18 (Fig. 4c). Next, we measured caspase 3/7 activity and cell viability simultaneously, by adding a luminogenic substrate containing the DEVD sequence and GF-AFC. As shown in Fig. 4d, the caspase activity negatively correlated with the cell viability exposed to NOC-18 > 200 μM (Fig. 4c: viability ratio to the control, 0.598 and 0.519 for 200 and 400 μM NOC-18, respectively). However, TNF-α-challenged cells remained completely viable even when the caspase activity was significantly elevated (Fig. 4e: viability ratio, 1.129 for 80 ng/mL TNF-α). As expected, add-on treatment with TNF-α significantly elevated the caspase 3/7 activity of NSC-34 cells exposed to NOC-18 (Fig. 4f), which correlated to an increase in cell toxicity (Fig. 4c). Moreover, a pan-caspase inhibitor, z-VAD-FMK, provided a dose-dependent protection of NSC-34 cells exposed to NOC-18 with increasing doses of TNF-α (Fig. 4g). Interestingly, the effect of z-VAD-FMK was comparable among toxicity between NOC-18 alone and NOC-18 plus TNF-α, implying that apoptotic cascade is the final common pathway in neuroinflammation associated with NO and TNF-α. The susceptibility of NSC-34 cells to apoptosis stress by staurosporine and the protective effect of z-VAD were also confirmed (Fig. 4h).

Figure 4.

Synergistic effect of pro-inflammatory cytokines and nitric oxide to cause apoptotic cell death of motor neuron NSC-34 cells. (a–c) Cell toxicity analysis of motor neuron NSC-34 cells exposed to NOC-18 (a), mouse tumor necrosis factor-α (TNF-α) (b), and both (c). Forty-eight hours after the treatment, cells were incubated with test medium containing bis-AAF-R110 and GF-AFC and with extracellular and intracellular proteases, representing dead cells and live cells, respectively. Fluorescence of bis-AAF-R110 was normalized to that of GF-AFC to minimize bias from the well-to-well differences in the cell density. Data are expressed as a ratio to the control, representing the mean ± SEM for n = 3.*p < 0.05 versus control in a and b. p < 0.05 versus NOC-18 for 100 μM NOC-18 plus 20 or 40 ng/mL TNF-α. #p < 0.05 versus control for 200 μM NOC-18 plus TNF-α (20 and 40 ng/mL) in c. (d, e, f) The relationship between caspase 3/7 activity and the toxicity of NOC-18 and/or TNF-α. Forty-eight hours after the exposure to NOC-18 (d), mouse TNF-α (e), or both (f), cells were reacted with luminogenic caspase 3/7 substrates containing DEVD, to measure the caspase 3/7 activity. Following this, the cells were incubated with a live cell indicator, GF-AFC, to measure the cell viability in the same well. *p < 0.05 versus control for viability, and #p < 0.05 versus control for caspase activity by one-way anova with Bonferroni's test. In (f) sublethal doses (20, 40 ng/mL) of TNF-α was added on NOC18-challenged NSC-34 cells, and caspase activity in each well was normalized to that of GF-AFC for a quantitative comparison between different treatments. TNF-α add-on significantly increased caspases activity compared with NOC18 alone (#p < 0.05). Data are expressed as a ratio to the control, representing the mean ± SEM for n = 3. (g) Toxicity of NOC-18 alone or NOC-18 plus TNF-α in NSC-34 was ameliorated by a pan-caspase inhibitor, z-VAD-FMK, in a dose-dependent manner. Addition of TNF-α significantly augmented the toxicity of NOC-18 against NSC-34 cells (*p < 0.05). (h) z-VAD-FMK prevented staurosporine-induced apoptotic cell death of NSC-34 in a dose-dependent manner. *p < 0.05 versus control. #p < 0.05 versus toxic reagents [NOC-18, NOC-18 plus TNF-α in (g), or staurosporine in (h)] by one-way anova with Bonferroni's test. The data represent the mean ± SEM for n = 3.

TG2 is expressed in spinal motor neurons and glial cells of mutant SOD1 transgenic mice

TG2 is reportedly distributed through the brain (Maggio et al. 2001). We investigated the expression profiles of TG2 in vivo using spinal cords of the mSOD1 mice at the late symptomatic stage (40 weeks). Immunohistochemistry for diaminobenzidine staining revealed that TG2 was expressed in the cytosol of spinal motor neurons in the WT mice (Fig. 5ai and ii). Most of the remaining motor neurons of the mSOD1 mice displayed increased and aberrantly localized immunoreactivity for TG2 (Fig. 5aiii and iv). Since the surrounding cells also showed increased staining, we performed a double immunofluorescence study of G93A SOD1 mice to more precisely analyze cell types expressing TG2. Confocal microscopic analysis revealed that some TG2 stained cells were colocalized with NeuN and CD11b but not with GFAP, indicating that TG2 is expressed in neurons and active microglia (Fig. 5bi–ix). Our mice form SOD1 aggregates infrequently because of the low copy number of the transgene; however, in several motor neurons, aggregated SOD1 colocalized with TG2 (Fig. 5bx–xii). Immunoprecipitation analysis of spinal cord lysates of mSOD1 mice to investigate if mSOD1 and TG2 were co-precipitated, failed to show the compelling interaction (data not shown), probably because of the weak affinity of the two proteins and utility of used antibodies for the immunoprecipitation. Optical and confocal laser microscopic analyses showed considerable staining of vasculature, and it is also possible that Mac2-negative microglia expressed TG2.

Figure 5.

Immunohistochemical analysis of transglutaminase 2 (TG2) expression in amyotrophic lateral sclerosis (ALS) model mice. Spinal cord slices from the mutant SOD (mSOD)1 mice, and wild-type (WT) littermates at the age of 40 weeks were used to analyze the expression profiles of TG2 in ALS model mice. (a) Diaminobenzidine (DAB) staining for TG2 in the spinal cord sections of WT (i, ii) and, the mSOD1 mice (iii, iv). TG2 is chiefly expressed in motor neurons in the WT mice. In the mSOD1 mice, motor neurons show increased staining, and surrounding cells are immunoreactive to TG2. Photos ii and iv are magnified images of the demarcated areas in (i and iii), respectively. Scale bars, 50 μm in a,c; 25 μm in (ii, iv). (b) Confocal micrographs for the double-immunofluorescence study of TG2 and cell-type markers (i–iii, NeuN; iv–vi, CD11b; vii–ix, GFAP) and human SOD1 (x–xii). TG2 (red) colocalizing with cell-type markers and human SOD1 (green) is indicated by arrowheads. Nuclei were stained with DAPI (blue). Scale bars; 20 μm

TG2 is induced in the spinal cord and hypoglossal nucleus in in vivo ALS models

Because the potential linkage of TG2 was suggested in mSOD1 oligomerization in ALS, the expression profiles of TG2 in affected regions in ALS model mice were examined in spinal cord lysates from G93A SOD1 mice at the pre-symptomatic stage (20 weeks) and the disease onset stage (30 weeks). Immunoblot analysis using anti-TG2 antibody revealed that compared with WT controls, mildly but significantly higher amounts of the TG2 protein were present in the pre-symptomatic spinal cord of the mSOD1 mice (Fig. 6ai and ii). However, no significant difference was observed at the time of onset between the mSOD1 and WT mice (Fig. 6ai and iii). Notably, the TG2 level increased with age in the spinal cord of the WT mice. This result is consistent with that of a previous study, which documented the age-dependent increase in the expression and activity of TG2 in the spinal cord but not in the cerebrum or cerebellum of rats (Virgili et al. 2001). To obtain further insight into the mechanism underlying the induction of TG2 in the context of ALS pathogenesis, the induction of TG2 in the hypoglossal nucleus was examined after nerve ligation, because axonal flow impairment is one of the major pathogenic pathways of ALS (Julien 2001). Quantification of TG2 mRNA by real-time PCR showed that the axonal ligation induced significantly higher expression of TG2 in the WT mice (Fig. 6bi); TG2 expression in the mSOD1 mice subjected to nerve ligation displayed a similar trend; however, the difference was not statistically significant (Fig. 6bii). Collectively, these data suggest that TG2 is transiently up-regulated in mSOD1 mice, which may be mediated in part the axonal flow impairment.

Figure 6.

Transglutaminase 2 (TG2) is up-regulated in the spinal cord of pre-symptomatic G93A superoxide dismutase 1 (SOD1) Tg mice and in the hypoglossal nuclei of mice subjected to axonal ligation. (a) Panel (i) shows the result of immunoblot analyses of TG2, G93A SOD1, and GAPDH (loading control) in spinal cord lysates from G93A SOD1 and the wild-type (WT) mice at the pre-symptomatic (20 weeks) or early symptomatic (30 weeks) stage. Panels (ii) and (iii) show densitometric analyses of the immunoblotting results; TG2 was significantly higher in the mutant SOD (mSOD)1 mice than in the WT mice at the pre-symptomatic stage (ii), but not in the early symptomatic age (iii). *p < 0.0001 (vs. WT) by Student's t-test. The difference was not significant between the G93A SOD1 and the WT mice at the early symptomatic stage. The data represent the mean ± SEM for n = 3. (b) Analysis of TG2 expression by real-time PCR. Hypoglossal nuclei were collected from the WT mice or mice that had been exposed to hypoglossal axonal ligation (WT/G93A SOD1 Tg, 20 weeks). In the WT mice, TG2 was significantly increased following hypoglossal axonal ligation (i), whereas the similar trend was observed in G93A SOD1 mice, although not significant (ii). *p < 0.05 (vs. control) by Student's t-test. The data represent the mean expression ratios ± SEM for n = 3.

Inhibition of TG2 delayed disease progression and microglial activation in G93A SOD1 transgenic mice

Although TG2 is implicated in several neurodegenerative diseases such as HD, Parkinson's disease, and Alzheimer's disease, the extent to which the transamidation activity of TG2 contributes to the pathogenesis remains controversial. Thus, we examined the effect of the TG2 inhibitor cystamine on the phenotype of ALS model mice. Because, TG2 is a systemically distributed enzyme with diverse physiological effects, intrathecal infusion of cystamine using an osmotic mini-pump for 42 days starting at 30 weeks of age (285 μg/day) was employed. The chronological profiles of the average body weight or grip power expressed as a percentage relative to starting points were significantly different between cystamine and saline treatments during 10 weeks after the infusion (Fig. 7a, b; p < 0.05 by two-way anova). Kaplan–Meier analysis of the lifespan (Fig. 7c, blue lines) showed a trend for a therapeutic benefit of cystamine infusion (294 ± 11.2 days for cystamine; 272 ± 6.8 days for the saline control), although it did not reach statistical significance (p = 0.0932 for life span). Late progression, determined as a 10% steady decrease in the body weight (Fig. 7c, red lines), also showed a non-significant trend toward delayed progression (p = 0.1873 for time to late progression). However, the average individual duration from 10% grip decline to the endpoint was significantly increased by the cystamine infusion, which indicated a benefit in slowing the disease progression (Fig. 7d). Western blot analysis to assess the amount of the SOD1 oligomer (Fig. 7ei, vertical line) normalized to monomeric SOD1 (asterisk) revealed that intrathecal cystamine significantly inhibited TG2 on SOD1 oligomerization (Fig. 7eii). Moreover, the quantitative assessment of Mac2 and cyclooxygenase-2, chemical markers for active microglia (Fig. 7fi and Figure S2) and inflammation, demonstrated their significant decrease in cystamine-treated mice relative to that in control mice (Fig. 7fii and iii).

Figure 7.

Intrathecal infusion of the transglutaminase 2 (TG2) inhibitor cystamine suppresses disease progression, superoxide dismutase 1 (SOD1) oligomerization, and microglial activation in G93A SOD1 transgenic mice. (a, b) The effects of intrathecal infusion of cystamine or saline into mutant SOD (mSOD)1 mice for 42 days starting at 30 weeks of age on body weight (a) and grip power (b). The effect of cystamine treatment was analyzed by two-way anova. Data represent the mean ± SEM of the body weight or grip power as a percentage of the maximum value (n = 5–8 for cystamine treatment; n = 4–11 for saline). (c) Kaplan–Meier curves of the disease onset and late disease progression, as assessed by a 10% steady decrease in the grip power (gray) and body weight (red), respectively, and longevity (blue). Bar with four arrows indicates the duration of cystamine infusion. (d) The effect of intrathecal cystamine to delay the disease progression. Graph bars indicate the average of symptomatic days in each mouse from the time of 10% grip loss to the endpoint. cystamine, n = 8; saline, n = 11. The data represent the mean ± SEM. p < 0.05 by Student's t-test. (e) Immunoblot analysis of SOD1 oligomer in spinal cord lysates from mSOD1 mice with intrathecal cystamine or saline infusion (i). The amount of SOD1 oligomer was determined by obtaining the ratio of high molecular species of SOD1 (vertical bar) to monomeric human SOD1 (asterisk) through densitometry (ii). Data represent the mean ± SEM for n = 3. *p < 0.05 (vs. saline) by Student's t-test. (f) Immunoblot analysis of Mac2 and cyclooxygenase-2 (COX2) in spinal cord lysates from mSOD1 mice. Panel (i) shows that Mac2 is expressed in spinal cords of mSOD1 mice, but not in those of the wild-type (WT) mice (30 weeks of age). Asterisk indicates endogenous mouse SOD1; arrowhead indicates G93A human SOD1. Panel (ii) shows immunoblotting results, and panel (iii) shows the results of densitometric analysis of the data. Mac2 (left) or COX2 (right) expression was normalized to GAPDH expression. Data represent the mean ± SEM for n = 3. *p < 0.001 (vs. saline) by Student's t-test.

Discussion

In this study, we provided evidence that TG2 recognizes and oligomerizes familial ALS-linked mSOD1 proteins but not native WT. Notably, apo-WT SOD1 was also subjected to cross-linking by TG2, which indicated that misfolded forms of SOD1 are its substrates. Although we used tagged SOD1 to analyze the interaction between mSOD1 and TG2 in cell culture experiment, the effect of the tagging is unlikely, since the oligomerization of misfolded SOD1 with no tag was observed in vitro. To our knowledge, this is the first report documenting that the disease-relevant conformations are preferentially recognized by TG2. Moreover, the up-regulation of TG2 in the spinal cords of pre-symptomatic mSOD1 mice, and in the hypoglossal nuclei subjected to axonal ligation strongly suggests the involvement of TG2 in ALS pathogenesis.

The transamidation activity of TG2 in neurodegeneration is a matter of debate. Although TG2 had been an attractive candidate, which could mediate cross-linking of various proteins associated with neurodegenerative diseases, such as β-amyloid, tau, α-synuclein, and Htt (Jeitner et al. 2009), this hypothesis may not be universally applicable in several diseases, as challenged recently. For instance, Chun et al. (2001) demonstrated that TG2 did not interact with mutant Htt. Moreover, ablation of TG2 from HD model mice did not affect Htt inclusions (Mastroberardino et al. 2002). Furthermore, non-enzymatic transcriptional dysregulation of TG2 is linked to neurodegeneration in the mouse model of HD (McConoughey et al. 2010). On the other hand, TG2 has been shown to be involved in Parkinson's disease, in which both WT and mutant α-synuclein are cross-linked to form Lewy bodies (Junn et al. 2003; Schmid et al. 2009). Furthermore, it has also been reported that neurotoxicity caused by glutamate exposure or oxidative stress is mediated by the transamidation activity of TG2 in primary culture neurons (Basso et al. 2012). In addition, the novel roles of γ-glutamylamines as novel substrates, including free γ-glutamyl-ε-lysine, γ-glutamylamines and γ-glutamylamine cyclotransferase have been proposed recently (Jeitner et al. 2013). In the case of SOD1, it was clearly shown that misfolded SOD1 can be a substrate of TG2 for transamidation both in cultured cells and in vitro.

In our study, the long-term culture as long as 92 h was required for TG2 to generate SOD1 oligomers in transfected cells, which was augmented by calcium ionophore. Importantly, it has been reported that TG2 is enzymatically silent in cells because of the insufficient concentrations of calcium and the tight negative regulation by GTPs or nitric oxide (Beninati et al. 2013), although the transient and excessive elevation of intracellular calcium that occurs through glutamate receptors. It is also reported that the calcium efflux from endoplasmic reticulum is enough for the transamidation activity of TG2 (Jeitner et al. 2009). On the other hand, the extracellular space contains a higher concentration of calcium (as high as 1–2 mM), which suffices to activate TG2 (Jeitner et al. 2009). Although extracellular TG2 is also inactive in the presence of GTP, it is transiently activated upon tissue injury and inflammation (Siegel et al. 2008), which is in line with our view. The catalytic activity of TG2 in the CSF of ALS patients was also reported previously (Fujita et al. 1998). In light of growing evidence that both WT and mSOD1 are secreted and present in the CSF of ALS patients and mSOD1 mice (Urushitani et al. 2006; Zetterstrom et al. 2011; Liu et al. 2012), we proposed a new pathway in which the transaminase activity of extracellular TG2 may mediate SOD1 oligomer-based neuroinflammation. In our in vitro oligomerization assay, calcium concentrations greater than 0.5 mM, the extracellular concentrations, were required for SOD1 cross-linking by TG2 (Fig. 1bii). Extracellular TG2 is localized on the surface the plasma membrane, forming various complexes with transmembrane proteins such as integrin, fibronectin, and low-density lipoprotein receptor-related proteins (Belkin 2011), which may allow the interaction of TG2 with secreted misfolded SOD1 with high affinity to the lipid membrane because of their hydrophobic nature. Recently, a direct role of calcium ion has been documented in the aggregate formation of misfolded SOD1 (Leal et al. 2013). However, this reaction requires overnight agitation at 37°C, while our TG2 transamidation is completed within 1 h static incubation at 37°C. Nevertheless, synergistic roles of calcium and TG2 could be considered in the misfolding of extracellular SOD1 proteins.

We demonstrated that a soluble oligomer of misfolded SOD1 robustly activated BV2 microglia cells, and promoted the expression of putative neurotoxic pro-inflammatory molecules such as TNF-α, IL-1β, and iNOS (Fig. 3-7a–d). Although iNOS was readily induced by the smaller size of SOD1 oligomers than TNF-α and IL1β were, the underlying mechanism remains unclear. The different susceptibility of the induction might be simply applied to the experimental conditions, including primer settings, and type of cells. It would be also plausible that iNOS is a predisposing factor for the oligomer-mediated neuroinflammation. The toxicity of TNF-α and/or nitric oxide in NSC-34 cells was in part mediated by caspase activation. However, the different toxicities between NOC-18 and TNF-α, despite the comparable levels of caspases 3/7, suggest that TNF-α induces both neurotoxic and neuroprotective pathways, whereas perpetual NO stimulation predominately induces apoptotic motor neuron death. Collectively, our results suggest that the oligomerization propensity of misfolded SOD1 underlies the molecular basis of microglial activation by extracellular SOD1 mutants (Urushitani et al. 2006; Ezzi et al. 2007; Zhao et al. 2010). Recently, it was documented that aggregated forms of recombinant SOD1 proteins activated microglial cell line through the lipid raft (Roberts et al. 2013). Our results provide further evidence that soluble SOD1 oligomers readily activate microglia. Interestingly, our data showed that microglial activation increased in an oligomer size-dependent manner, although the effect was regressed at very large oligomer sizes. The results presented here are consistent with the recent consensus that soluble oligomeric species induce toxicity, whereas extremely high molecular aggregates may do no harm (Ross and Poirier 2005).

TG2 was up-regulated in the spinal cord from mSOD1 mice at the pre-symptomatic stage; however, the mechanism of its induction remains unclear. A previous study demonstrated the induction of TG2 under conditions of ischemia, trauma, and inflammation (Tolentino et al. 2002). In the present study, we demonstrated that TG2 transcription in the hypoglossal nucleus was elevated following nerve ligation (Fig. 6b). This induction was statistically significant in the WT mice, whereas a non-significant, but clearly similar trend was observed in the mSOD1 mice. Because impairment of axonal transport is an early event in ALS model mice (Williamson and Cleveland 1999), it is possible that TG2 is secondarily induced the axonal transport damage concomitantly at the pre-symptomatic stage. Fujita et al. (1995) reported that TG2 activity was markedly lower in spinal cord lysates from sporadic ALS patients than in those from non-ALS patients. The same group also reported that TG2 activity in the CSF of ALS patients was significantly higher than that in the CSF of control subjects at the initial stage of the disease and ultimately decreased at the late stage of ALS (Fujita et al. 1998). These reports support our notion that extracellular TG2 participates in the progression of the early stage of the disease.

Intrathecal infusion of the TG2 inhibitor cystamine hydrochloride into G93A SOD1 mice significantly suppressed phenotypic decline, SOD1 oligomers and chemical markers for active microglia and neuroinflammation. Cystamine hydrochloride has previously been shown to be very stable for long time use in the osmotic minipump (Santhanam et al. 2010). However, caution must be taken to interpret the effect of cystamine, which is multifunctional. For example, cystamine reportedly inhibits caspases and increases antioxidant levels (Lesort et al. 2003). Moreover, TG2 was shown to directly induce iNOS (Takano et al. 2010) or nuclear factor-kappa B (NFκB) (Lee et al. 2004), and cystamine treatment reportedly ameliorated the symptoms of the HD model mice lacking TG2 (Bailey and Johnson 2006). For more integrated interpretation, it is necessary to test the effect of more specific TG2 inhibitors. The use of membrane-permeable and -impermeable inhibitors may also help to specify the site of TG2 involved in the pathogenesis of ALS.

In conclusion, our results demonstrated that TG2 is a promising therapeutic target for mSOD1-linked ALS. Further investigation of the diverse pathways in which TG2 is involved will greatly contribute to our understanding of ALS.

Acknowledgements

We thank Dr Gail V. W. Johnson for kindly providing human TG2 plasmids. The technical help by the Central Research Laboratory at the Shiga University of Medical Science is gratefully appreciated. We thank Dr Atsushi Saito for technical help with laser capture microdissection. We thank Drs Yoko Okamoto and Masafumi Ihara for the generous gift of pCMV-Myc-α-synuclein plasmid, and experimental support of this work. This work was supported by a Grant-in Aid for Scientific Research on Innovative Areas from MEXT (23111002), a Grant for Research on Neurodegenerative Diseases from Japan Health and Labour Science, and an Intramural Research Grant for Neurological and Psychiatric Disorders (22-4) from the NCNP. The authors declare that there are no conflicts of interest.

Ancillary