N-Glycan Processing Deficiency Promotes Spontaneous Inflammatory Demyelination and Neurodegeneration*

Multiple sclerosis (MS) is characterized by inflammatory demyelination of axons and neurodegeneration, the latter inadequately modeled in experimental autoimmune encephalomyelitis (EAE). Susceptibility of inbred mouse strains to EAE is in part determined by major histocompatibility complex haplotype; however, other molecular mechanisms remain elusive. Galectins bind GlcNAc-branched N-glycans attached to surface glycoproteins, forming a molecular lattice that restricts lateral movement and endocytosis of glycoproteins. GlcNAc branching negatively regulates T cell activity and autoimmunity, and when absent in neurons, induces apoptosis in vivo in young adult mice. We find that EAE susceptible mouse strains PL/J, SJL, and NOD have reduced GlcNAc branching. PL/J mice display the lowest levels, partial deficiencies in N-acetylglucosaminyltransferase I, II, and V (i.e. Mgat1, -2, and -5), T cell hyperactivity and spontaneous late onset inflammatory demyelination and neurodegeneration; phenotypes markedly enhanced by Mgat5+/- and Mgat5-/- backgrounds in a gene dose-dependent manner. Spontaneous disease is transferable and characterized by progressive paralysis, tremor, dystonia, neuronophagia, and axonal damage in both demyelinated lesions and normal white matter, phenocopying progressive MS. Our data identify hypomorphic Golgi processing as an inherited trait that determines susceptibility to EAE, provides a unique spontaneous model of MS, and suggests GlcNAc-branching deficiency may promote T cell-mediated demyelination and neurodegeneration in MS.

Relapsing remitting multiple sclerosis is characterized by inflammatory destruction of the myelin sheath surrounding axons in the central nervous system (CNS), 3 producing relaps-ing and remitting attacks of neurological dysfunction (1). This is commonly followed by a secondary progressive neurodegenerative phase distinguished by axonal damage and neuronal loss (1). Primary progressive MS is similar to secondary progressive disease but lacks the initial relapsing-remitting phase. However, recent investigations have demonstrated that gray matter involvement and axonal damage in otherwise normal appearing white matter are present at the onset of relapsing remitting multiple sclerosis (2,3). This indicates neurodegeneration is an early and prominent feature of disease and questions the interpretation that MS is only a T cell-mediated demyelinating disease. Experimental autoimmune encephalomyelitis (EAE) is a useful model of T cell-dependent inflammatory demyelination, but fails to properly address the neurodegenerative phenotype of MS.
MS is characterized by adult onset and partially familial relationships, indicating complex interactions between environmental and genetic factors in disease pathogenesis (4). Whole genome screens have identified a number of candidate loci associated with MS (5) and EAE (6, 7), but non-MHC genes that strongly promote disease have yet to be described. This is despite long standing observations that T cell dysfunction is critical to development of EAE (8 -10) and the identification of multiple genes that alter EAE severity in susceptible mouse strains when deficient or overexpressed (11). Myelin-specific TCR transgenic mice develop spontaneous CNS autoimmune demyelinating disease in susceptible strains (9,10,13,14), however, spontaneous disease secondary to physiologically relevant gene dysfunction has not been reported.
Deficiency of the N-glycan processing gene Mgat5 in 129/Sv mice, an EAE-resistant strain, results in spontaneous kidney autoimmunity after 1 year of age, enhanced delayed type hypersensitivity, and increased susceptibility to myelin basic protein (MBP)-induced EAE (15). Mgat5 is near the end of a linear pathway of Golgi processing enzymes required for GlcNAcbranching in N-glycans, structures on glycoproteins that serve as ligands for the galectin family of N-acetyllactosamine binding lectins. Multivalent binding between N-glycans attached to surface glycoproteins and galectins forms a molecular lattice that restricts lateral movement of glycoproteins and their loss to endocytosis (15,16). Galectin binding to N-glycans increases proportionally for mono-, bi-, tri-, and tetra-antennary Glc-NAc-branched N-glycans, the products of N-acetylglucosami-nyltransferase I, II, IV, and V (i.e. Mgat1, -2, -4, and -5), respectively. The number of N-glycans, an encoded feature of protein sequence, varies widely between glycoproteins and is a critical determinant of galectin binding. This allows for differential regulation of surface glycoproteins by the galectin lattice (17,18).
In resting T cells, the Mgat5 mutation reduces T cell receptor (TCR) binding to galectin-3, thereby enhancing agonist-induced TCR clustering and signaling. This lowers activation thresholds and promotes subsequent T H 1 differentiation (15,19). After several rounds of cell division, T cells undergo growth arrest following translocation of CTLA-4 to the cell surface (20). CTLA-4 recycles rapidly from the surface to endosomes and retention of CTLA-4 at the cell surface is promoted by GlcNAc-branched N-glycan-mediated inhibition of endocytosis (17,21). In this manner, N-glycan GlcNAc-branching negatively regulates T cell growth early by increasing TCR activation thresholds and late by promoting growth arrest via CTLA-4.
The Golgi enzymes Mgat1, -2, -4, and -5 sequentially display decreasing protein levels and affinities for their shared sugarnucleotide donor UDP-GlcNAc (22). As such, tri-and tetraantennary GlcNAc-branched N-glycans produced by Mgat4 and -5 are subsaturating on mature glycoproteins and sensitive to changes in both enzyme expression and UDP-GlcNAc availability (17,23,24). Metabolic supplementation to UDP-Glc-NAc biosynthesis in T cells (e.g. GlcNAc, uridine, and glucose) enhances GlcNAc-branching, increases thresholds for T cell activation and T H 1 differentiation of naïve T cells, and promotes CTLA-4 surface retention in activated T cells. Moreover, GlcNAc supplementation of T cells inhibits adoptive transfer EAE in PL/J mice and as an oral supplement, reduces spontaneous diabetes in non-obese diabetic (NOD) mice (21). These data demonstrate that metabolism, via UDP-GlcNAc biosynthesis, conditionally regulates T cell-mediated autoimmunity by altering GlcNAc branching in N-glycans.
Here we report that among inbred mouse strains, N-glycan GlcNAc-branching in T cells is highly variable and inversely correlates with EAE susceptibility. PL/J mice display the lowest levels, partial deficiency of Mgat1, -2, and -5 enzyme activity, TCR hypersensitivity, and mild spontaneous inflammatory demyelination and neurodegeneration after 1 year of age. Spontaneous disease was markedly enhanced by Mgat5 ϩ/Ϫ and Mgat5 Ϫ/Ϫ backgrounds in a gene dose-dependent manner, demonstrating interactions between inherited and experimentally induced defects in N-glycan processing. PL/J mice with spontaneous disease displayed features of chronic MS, including progressive paralysis, tremor, focal dystonic posturing, paroxysmal dystonia, neuronophagia, and axonal damage in demyelinated lesions and normal white matter (1,25). Our results indicate that naturally arising hypomorphisms in multiple N-glycan GlcNAc-branching enzymes regulate EAE susceptibility among inbred strains of mice Moreover, Mgat5 Ϫ/Ϫ PL/J mice provide a unique spontaneous model of MS that arises from physiologically relevant gene dysfunction and displays the two critical phenotypes observed in MS, namely inflammatory demyelination and neurodegeneration.
MALDI-TOF Mass Spectroscopy-This was done as a service at the Glycotechnology Core Facility, a resource of the Glycobiology Research and Training Center at the University of California, San Diego. CD3 ϩ T cells were lysed with 1% SDS in 100 mM Tris, pH 7.4, dialyzed to remove SDS, digested with trypsin to generate glycopeptides, and treated with PNGase F to release the N-glycans. To focus on early processing, N-glycans were desialylated by mild acid digestion prior to permethylation and MALDI-TOF mass spectroscopy. Monoisotopic peaks in the spectra were identified using GlycanMass and GlycoMod online software, with specific targeting of the N-glycan intermediates occurring during Golgi processing.
Spontaneous Demyelinating Disease and Adoptive Transfer of Disease-PL/J mice at two facilities were assessed for clinical demyelinating disease and dystonia (Table 1 and supplemental  Table S1). The first cohort was at backcross 4 from 129/Sv (Table 1) and was housed at the Samuel Lunenfeld Research Institute vivarium, a colony infected with mouse hepatitis virus, EDIM, minute virus, mouse parvovirus, GDVII, pinworm, and fur mites. These mice were initially assessed by blindly examining all Mgat5 Ϫ/Ϫ (n ϭ 43), Mgat5 ϩ/Ϫ (n ϭ 22), and Mgat5 ϩ/ϩ (n ϭ 15) PL/J mice in the colony over 6 months of age. Only mice over 1 year of age were found to have weakness and this smaller cohort (n ϭ 21, 13, and 10, respectively) was scored for clinical severity every 1-2 weeks over the next ϳ4 months. Weakness was slowly progressive without recovery in all affected mice, an observation confirmed by daily assessment of a smaller cohort (n ϭ 12) of clinically affected mice over a 4-week period. At sacrifice, mice were perfused with paraformaldehyde via cardiac perfusion and harvested brain and spinal cord were embedded in paraffin, sectioned, and stained with Hematoxylin & Eosin or Luxol Fast Blue. The second cohort (supplemental Table 1) at backcross 6 were re-derived from the Samuel Lunenfeld Research Institute mice by embryo transfer and housed at the University of California, Irvine, vivarium that is pathogen-free except for some rooms with mouse parvovirus. Disease was observed in both pathogen-free and mouse parvovirus-containing rooms. For adoptive transfer of demyelinating disease, 1 ϫ 10 7 splenocytes harvested from spontaneously diseased Mgat5 ϩ/Ϫ and Mgat5 Ϫ/Ϫ donor mice were transferred by intraperitoneal injections into naïve wild-type PL/J recipient mice following 48 h of in vitro activation with anti-CD3/anti-CD28.
Electromyography and Nerve Conduction Studies-Mice were anesthetized with Avertin. Temperature was maintained at 35-37°C using infrared heat lamps. Monopolar needle electrodes (Ambu Inc., Glen Burnie, MD) were used for stimulation and recording motor nerve potentials. The active and indifferent recording electrodes were placed in medial gastrocnemius and ipsilateral footpad, respectively. The active and reference stimulating electrodes were placed percutaneously in the popliteal or sciatic notch and ipsilateral thoraco-lumbar paraspinal muscle, respectively. A pre-gelled strip electrode at the tail acted as a ground. Responses from supramaximal electrical stimulation (pulse width 0.05 ms) were analyzed with additional stimulations done to record late responses (F waves and H reflexes). H reflexes were identified when successive late responses had identical morphology and onset latency; F waves were identified when successive late responses had variable onset latency and morphology. For needle EMG recording, the recording monopolar needle electrode was inserted into one or more hindlimb muscles: quadriceps, hamstrings, lumbar paraspinals, gastrocnemius, and tibialis anterior. The presence of spontaneous muscle activity (i.e. fasiculations, fibrillations, or myokymia) was assessed in at least three regions of the muscle. All recordings were made on a Sierra LT portable machine (Cadwell Laboratories, Kennewick, WA) and analyzed using the proprietary software supplied by the manufacturer.

EAE Susceptible Mouse Strains Are Hypomorphic for N-Glycan GlcNAc Branching-Reducing
GlcNAc-branching by ϳ20 -25%, either by loss of a single Mgat5 allele or by partial inhibition of Golgi processing with the mannosidase II inhibitor SW, is sufficient to enhance TCR signaling and T cell proliferation (Ref. 21 and Fig. 1A and supplemental Fig. S1). Therefore, we explored whether susceptibility of inbred mouse strains to EAE correlates with levels of GlcNAc-branching. For this purpose we stained the cell surface with L-PHA (P. vulgaris leukoagglutinin), a plant lectin that specifically binds ␤1,6Glc-NAc-branched N-glycans produced by Mgat5 and serves as a marker of GlcNAc-branching (15,21). CD4 ϩ and CD8 ϩ T cells from the EAE susceptible strains PL/J, SJL, and NOD, which also develops spontaneous autoimmune diabetes, expressed ϳ30 -40% less ␤1,6GlcNAc-branched N-glycans than the three EAE-resistant strains 129/Sv, Balb/c, and B10.S (Fig. 1, A-C). Remarkably, CD4 ϩ T cells from wild-type PL/J mice express ϳ25% less ␤1,6GlcNAc-branched N-glycans than Mgat5 heterozygous 129/Sv cells (Fig. 1A). This indicates that the PL/J strain harbors genetic hypomorphisms that reduce GlcNAc-branching to a significantly greater degree than loss of an Mgat5 allele. The C57BL/6 strain is less sensitive than the SJL strain to EAE, as evidenced by differential requirement for CD28 co-stimulation to induce disease (26). C57BL/6 T cells display intermediate levels of ␤1,6GlcNAc-branched N-glycans relative to T cells from the EAE-sensitive and -resistant strains (Fig. 1, B and C). Co-staining CD4 ϩ T cells with L-PHA and the naïve T cell marker CD62L, the effector/regulatory T cell marker CD25, and/or the regulatory T cell marker Foxp3 demonstrated the same relative differences in ␤1,6GlcNAcbranched N-glycans, with PL/J ϽC57BL/6 Ͻ 129/Sv (Fig. 1D). However, in B220 ϩ B cells and F4/80 ϩ CD11b ϩ macrophages, ␤1,6GlcNAc-branched N-glycan levels were similar among the tested strains (Fig. 1, B and C). Therefore, susceptibility to EAE correlated inversely with ␤1,6GlcNAc-branched N-glycan expression in T cells in rank order PL/J Ͼ SJL, NOD Ͼ C57BL6 Ͼ Balb/c, 129/Sv, B10.S. NOVEMBER 16, 2007 • VOLUME 282 • NUMBER 46

JOURNAL OF BIOLOGICAL CHEMISTRY 33727
Next we explored potential mechanisms for reduced ␤1,6GlcNAc-branching in PL/J T cells. The final step in the linear pathway to ␤1,6GlcNAc-branched N-glycan biosynthe-sis is mediated by Mgat5. Mgat5 enzyme activity but not mRNA levels are reduced ϳ50% in splenocytes and T cells from PL/J and C57BL/6 mice relative to 129/Sv mice (Fig. 2, A and B). However, PL/J T cells display a greater reduction in ␤1,6Glc-NAc-branched N-glycans than C57BL/6 T cells, suggesting additional N-glycan processing defects proximal to Mgat5 are present in PL/J cells. Indeed, MALDI-TOF mass spectroscopy indicates that relative to 129/Sv and C57BL/6 T cells, PL/J T cells have reduced bi-and triantennary GlcNAc-branched N-glycans (E ions) and accumulate pathway intermediates upstream of Mgat2 (C and D ions) ( Fig. 2C and supplemental Fig. S2). Moreover, Mgat2 and Mgat1 enzymatic activities, but not mRNA transcript levels, differ significantly among the three strains, with PL/J Ͻ C57BL/6 Ͻ 129/Sv and PL/J, C57BL/6 Ͻ 129/Sv, respectively (Fig. 2, A and B). These data indicate that partial deficiencies at the post-transcriptional level in Mgat1, -2, and -5 combine to reduce GlcNAc-branching in PL/J Ͼ C57BL/6 Ͼ 129/Sv T cells and confirm that PL/J mice are naturally hypomorphic for GlcNAc-branched N-glycans. However, defects in other N-glycan processing enzymes may also contribute to the phenotype.

Deficiency of Multiple Golgi Glc-NAc Transferases Induces T Cell
Hyperactivity in PL/J Mice-Mgat5 ϩ/Ϫ 129/Sv T cells hyperproliferate relative to Mgat5 ϩ/ϩ 129/Sv T cells (supplemental Fig.  S1); indicating the greater reduction in GlcNAc-branching inherent to wild-type PL/J T cells should result in T cell hyperactivity. Indeed, wildtype PL/J T cells were more sensitive to TCR agonist than wild-type 129/Sv T cells as indicated by phosphorylation levels of lck at activating Tyr 394 and LAT, induction of the activation marker CD69, and proliferation as measured by CFSE dilution (Fig. 3, A-D). Consistent with intrinsically higher ␤1,6Glc-NAc-branching in the 129/Sv strain, the relative increase in lck-pY 394 , pLAT 394 , and CD69 is significantly greater in Mgat5 Ϫ/Ϫ versus Mgat5 ϩ/ϩ 129/Sv T cells than Mgat5 Ϫ/Ϫ versus Mgat5 ϩ/ϩ PL/J T cells (Fig. 3, A and B). However, in absolute terms, TCR signaling and CD69 expression were greater in PL/J Mgat5 Ϫ/Ϫ than 129/Sv Mgat5 Ϫ/Ϫ T cells (Fig. 3, A and B). Galectins bind N-glycans in proportion to GlcNAc-branching (17,18) and these data suggest that deficiency of bi-and triantennary Glc-NAc-branched N-glycans also contribute to PL/J T cell hypersensitivity. Indeed, co-incubation of wild-type 129/Sv cells with SW, which blocks N-glycan GlcNAcbranching beyond monoantennary, equalized L-PHA staining, CD69 expression, and proliferation to that of untreated wild-type PL/J T cells (Fig. 3, C and D, supplemental Figs. S3 and S4). In contrast, SW had nominal affects on wild-type PL/J T cells, consistent with pre-existing deficiency in bi-and triantennary structures. SW also had minimal affects on Mgat5 Ϫ/Ϫ PL/J T cells but enhanced CD69 expression in Mgat5 Ϫ/Ϫ 129/Sv T cells to that of Mgat5 Ϫ/Ϫ PL/J T cells co-incubated with or without SW (Fig. 3E). Together, these data indicate that GlcNAc-branching deficiency in wild-type PL/J T cells induces TCR hyperactivity relative to 129/Sv, with reductions in bi-, tri-, and tetra-antennary GlcNAc-branched N-glycans all contributing to the phenotype.
The above data suggest that genetic rescue of GlcNAcbranching deficiency in PL/J T cells will require enhancing the expression of at least three genes (i.e. Mgat1, -2, and -5). In contrast, metabolically supplementing the hexosamine pathway with GlcNAc increases bi-, tri-, and tetra-antennary Glc-NAc-branched N-glycans by increasing UDP-GlcNAc supply  NOVEMBER 16, 2007 • VOLUME 282 • NUMBER 46

Spontaneous Mouse Model of Multiple Sclerosis
to Mgat1, -2, and -5 (17,21,23), providing a simple experimental approach for rescue. Indeed, supplementing PL/J T cells with GlcNAc rescues N-glycan GlcNAc-branching and inhibits TCR signaling, CD69 expression, T H 1 differentiation, CTLA-4 endocytosis, and proliferation (21). These inhibitory phenotypes are all reversed by co-incubation with SW, confirming that GlcNAc supplementation acts by increasing GlcNAcbranching in N-glycans. Taken together, these data demonstrate that hypomorphic production of GlcNAc-branched N-glycans in PL/J mice is causal in T cell hypersensitivity.
Defective N-Glycan Processing in PL/J Mice Promotes Spontaneous Inflammatory Demyelination and Neurodegeneration-Mgat5 Ϫ/Ϫ 129/Sv mice develop spontaneous kidney autoimmunity after 1 year of age and are more susceptible to EAE (15). Neuron-specific depletion of the Mgat1 gene eliminates galectin ligands in N-glycans and results in neuronal apoptosis in young adult mice (27). High levels of galectin-1, which provokes T cell apoptosis by remodeling the galectin lattice (28,29), induces axonal and neuronal cell body degeneration (30). In addition to N-glycan processing deficiency, the PL/J strain possess the H-2 MHC class II haplotype that reduces negative selection of MBP 1-11 reactive T cells in the thymus (31,32). We reasoned that loss of central tolerance to MBP1-11 combined with reduced N-glycan GlcNAc-branching may induce spontaneous CNS inflammatory demyelination and neurodegeneration in PL/J mice. Indeed, clinical observation of Mgat5 ϩ/ϩ , Mgat5 ϩ/Ϫ , and Mgat5 Ϫ/Ϫ PL/J mice at backcross 4 and 6 from 129/Sv as well as non-congenic wild-type PL/J mice from Jackson Laboratories revealed signs of tail and/or hindlimb weakness after 1 year of age ( Table 1, supplemental Table 1, supplemental Video 1A, and data not shown). Importantly, incidence, severity, and mortality were inversely correlated with Mgat5 gene dose (Table 1 and supplemental Table 1), demonstrating genetic interactions of inherited and experimentally induced deficiencies in GlcNAc-branched N-glycans. The course was chronic and slowly progressive without relapses or recovery (supplemental Fig. 5F), a clinical picture typical of progressive MS (1). Mice frequently displayed involuntary movements such as tremor, focal dystonic posturing, and paroxysmal dystonia (Fig. 4A, supplemental Video 1, B-D, and Table 1), movement disorders that occur in MS (25) but rarely reported in EAE.
Pathological studies revealed submeningeal perivascular lymphocyte cuffing (Fig. 4B) and multifocal demyelination of the brainstem (Fig. 4, C and D), spinal cord (supplemental Fig.  S5, A and B) and spinal roots (Fig. 4, G and H, supplemental Fig.  S5, C and E). CNS pathology was similar to chronic MS plaques and characterized by mononuclear cells admixed with myelin debris centered around blood vessels, gliosis, axonal swelling (spheroids), and axonal degeneration. Phagocytosis of neurons in the gray matter (neuronophagia, Fig. 4E), a presumptive marker of neuronal apoptosis, and axonal pathology in otherwise normal appearing CNS white matter (Fig. 4F) were also frequently observed; phenotypes consistent with previous data demonstrating a direct affect of GlcNAc-branching on neurodegeneration (27). Gray matter disease and axonal damage in normal appearing white matter are early and diffuse features of  Fig. S5H). CNS and/or peripheral nervous system pathology was present and qualitatively similar in all mice with clinical weakness, and frequently co-existed in the same individual. Additional organ screening in 4 clinically affected mice suggested that the only autoimmune disease present was inflammatory demyelination. Peripheral nervous system demyelination was seen with similar frequency in all three Mgat5 genotypes (Fig. 4I). In contrast, CNS disease was ϳ2and 3-fold more frequent in Mgat5 ϩ/Ϫ and Mgat5 Ϫ/Ϫ PL/J mice than wild-type mice, respectively.
Anti-CD3 antibody stimulated splenocytes from Mgat5 Ϫ/Ϫ mice with moderate to severe, but not mild demyelinating pathology, frequently transferred disease to naïve wild-type recipients ( Table 2), confirming that spontaneous disease was in part immune mediated. Metabolic supplementation (i.e. Glc-NAc) enhances GlcNAc-branching in encephalitogenic wildtype PL/J T cells in vitro and inhibits their ability to induce EAE     when the cells are transferred to naïve recipients (21). This confirms that reduced GlcNAc-branching in PL/J T cells promotes T cell hypersensitivity and demyelinating disease in vivo.
Environmental pathogens have been implicated in demyelinating disease in MBP-TCR transgenic mice (9). However, we observed similar frequency of disease when mice were housed in vivariums containing a multitude of pathogens (Table 1), a single pathogen or no pathogens (supplemental Table 1), suggesting genetic rather than infectious factors dominate in disease pathogenesis.
Mgat5 Ϫ/Ϫ PL/J mice Ͻ1 year and diseased PL/J mice of all three genotypes Ͼ1 year displayed increased frequency of TIM-3 ϩ T H 1 cells (Fig. 5A), indicating negative regulation of T H 1 differentiation by GlcNAc-branching in vivo (19). Paradoxically, the Mgat5 deficiency, increasing age, and demyelinating disease are associated with increased numbers of CD25 ϩ CD4 ϩ and CD25 ϩ Foxp3 ϩ CD4 ϩ T cells (Fig. 5B, supplemental Fig. S6). This result is consistent with hyperproliferation of Mgat5 Ϫ/Ϫ regulatory T cells observed in vitro 4 and previous observations of age-associated increases in functional CD25 ϩ CD4 ϩ regulatory T cells in humans (33). The increased frequency of PL/J CD25 ϩ Foxp3 ϩ CD4 ϩ regulatory T cells in vivo may not only be a direct effect of GlcNAc-branching deficiency on proliferation, but also reflect homeostatic negative regulation of autoreactive T H 1 effector cells. Taken together these data indicate enhanced T H 1 effector responses associated with the GlcNAc-branching deficiency combined with H-2 induced loss of central tolerance to MBP 1-11 to promote spontaneous inflammatory demyelination in PL/J mice.

DISCUSSION
GlcNAc-branching in N-glycans attached to surface glycoproteins promotes multivalent galectin binding, which negatively regulates TCR signal strength and T H 1 differentiation of naïve T cells, CTLA-4 endocytosis in activated T cells, and susceptibility to autoimmunity (15,19). Here we demonstrate that variability in T cell-specific expression of GlcNAc-branched N-glycans is an inherited trait in inbred mice that regulates T cell function and susceptibility to autoimmune demyelinating disease. EAE susceptible strains PL/J, NOD, and SJL are hypomorphic for GlcNAc-branched N-glycans in T cells but not B cells and macrophages; a phenotype that induces T cell hyperactivity and promotes EAE susceptibility in PL/J mice. A mild inflammatory demyelinating disease similar to progressive MS develops spontaneously in PL/J mice after 1 year of age, a phenotype markedly enhanced by gene dose-dependant loss of Mgat5. This demonstrates inherited and induced defects in GlcNAc-branching interact at the genetic level to promote disease. However, our data does not exclude other molecular mechanisms in the development of spontaneous disease.
Loss of ␤1,6GlcNAc-branched N-glycans via genetic deletion of Mgat5 enhances demyelinating disease in both 129/Sv H-2 b (15) and PL/J H-2 mice. Mgat5 Ϫ/Ϫ 129/Sv mice develop late onset spontaneous kidney autoimmunity, but unlike the PL/J strain, Mgat5 ϩ/ϩ and Mgat5 ϩ/Ϫ 129/Sv mice do not develop spontaneous autoimmune disease; a result consistent with the higher levels of GlcNAc-branched N-glycans in the 129/Sv strain (15). These data suggest GlcNAc-branching regulates autoimmune thresholds and EAE irrespective of the MHC haplotype, whereas strain-dependent genetic factors such as MHC determine the targeted tissue in spontaneous disease.
Autoimmunity is a complex trait, where genetic susceptibility is distributed across multiple genes and modulated by the environment via unclear mechanisms. Genetic and environmental factors are often presumed to encompass various pathways that interact by unknown molecular mechanisms. Remarkably, we find that reduced GlcNAc-branching in PL/J T cells results from partial deficiencies of multiple N-glycan processing enzymes, including Mgat1, -2, and -5, but can be conditionally regulated by metabolite input to the biosynthesis of UDP-GlcNAc (21). This indicates a unique genetic model for autoimmunity, whereby multiple weakly penetrant genetic factors combine in the Golgi N-glycan GlcNAc-branching pathway to produce a highly penetrant phenotypic change to promote disease. This model raises the possibility that co-inheritance of genetic variants in multiple N-glycan pathway genes may combine to reduce GlcNAc branching and promote MS in humans.
MS is frequently a two-stage disease characterized by inflammatory destruction of the myelin sheath with mean onset ϳ29 years of age followed ϳ10 years later by a secondary progressive neurodegenerative phase distinguished by axonal damage and neuronal loss (1,34  progressive MS: spontaneous onset in mid-life, movement disorders such as tremor and dystonia, and a slow progressive decline in neurological function in association with neuronal loss and axonal damage (1,25). As such, Mgat5-deficient PL/J mice provide a unique model to study both the inflammatory and neurodegenerative components of MS. Neuron-specific loss of Mgat1, which eliminates galectin ligands in N-glycans, results in apoptosis of adult neurons in vivo (27). This suggests that defects in GlcNAc branching inherent to the PL/J strain, coupled with induced deficiency in Mgat5, may promote neurodegeneration in Mgat5 Ϫ/Ϫ PL/J mice independent from effects on T cell function. Additional investigation is required to confirm this hypothesis. Moreover, it will be important to determine the relative roles of T cell-mediated demyelination and neurodegeneration. Supplementing the hexosamine pathway with GlcNAc increases GlcNAc-branching in multiple cell types (17,21,23), raising the possibility that in addition to suppressing inflammatory demyelination (21), this therapeutic approach may also directly limit neurodegeneration in MS.
NOD and SJL mice also display reduced ␤1,6GlcNAcbranching in T cells, albeit less severe than the PL/J strain. Oral GlcNAc supplementation increases N-glycan GlcNAc branching in vivo and suppresses spontaneous autoimmune diabetes in NOD mice (21), indicating GlcNAc branching deficiency also promotes spontaneous autoimmunity in the NOD strain. Although NOD mice harbor the H-2 g7 MHC haplotype, a critical promoter of diabetes, they also develop other spontaneous autoimmune diseases at lower frequency (e.g. sialitis, autoimmune thyroiditis) (35) and when given pertussis toxin, autoimmune diabetes is suppressed and spontaneous CNS demyelinating disease develops (36). Similarly, B7-2 deficiency or interleukin-2 blockade in NOD mice induces spontaneous peripheral nerve autoimmune demyelination (37,38). Schwann cells, the cellular constituent of peripheral myelin, surround pancreatic islets and have been proposed to be targeted early in the development of insulitis in NOD mice (39). SJL mice are highly sensitive to EAE, in part because of high precursor frequency of proteolipid protein 139 -151 reactive T cells (40); however, spontaneous autoimmunity has not been reported in this strain. The level of N-glycan GlcNAc branching appears similar in NOD and SJL T cells, suggesting hypomorphic expression of GlcNAc-branched N-glycans in these mice promote T cell-mediated demyelinating disease, but other genetic factors are required for induction of spontaneous disease.
Modification of ␤1 integrins with ␤1,6GlcNAc-branched N-glycans reduces cell adhesion to fibronectin and increases cell motility (16,24,41,42), phenotypes that may contribute to T cell dysfunction and disease. Antibodies to ␣ 4 ␤ 1 integrin inhibit EAE and MS by limiting T cell recruitment to the CNS via interactions with vascular cell adhesion molecule-1 expressed on activated endothelium (43,44). ␤1,6GlcNAc branching reduces T cell adhesion to vascular cell adhesion molecule in vitro, 5 suggesting ␤1,6GlcNAc-branching may limit T cell recruitment to the CNS in vivo. Macrophage motility and phagocytosis are impaired by Mgat5 deficiency (16), a phenotype that may inhibit clearance of apoptotic cells and promote autoimmunity (12). However, macrophages from wild-type PL/J, C57BL/6, and 129/Sv mice display similar levels of ␤1,6GlcNAc-branched N-glycans, suggesting defective N-glycan processing does not significantly alter macrophage function to promote autoimmunity in wild-type PL/J mice. Defects in substratum and/or cell-cell adhesion may also occur in neurons and directly contribute to the neurodegeneration observed in Mgat5 Ϫ/Ϫ PL/J mice. With greater insight into the relative importance of GlcNAc branching in these various pathways, we will achieve a more complete molecular model of spontaneous demyelinating disease and neurodegeneration.