Stroke-Like Episodes and Cerebellar Syndrome in Phosphomannomutase Deficiency (PMM2-CDG): Evidence for Hypoglycosylation-Driven Channelopathy

Stroke-like episodes (SLE) occur in phosphomannomutase deficiency (PMM2-CDG), and may complicate the course of channelopathies related to Familial Hemiplegic Migraine (FHM) caused by mutations in CACNA1A (encoding CaV2.1 channel). The underlying pathomechanisms are unknown. We analyze clinical variables to detect risk factors for SLE in a series of 43 PMM2-CDG patients. We explore the hypothesis of abnormal CaV2.1 function due to aberrant N-glycosylation as a potential novel pathomechanism of SLE and ataxia in PMM2-CDG by using whole-cell patch-clamp, N-glycosylation blockade and mutagenesis. Nine SLE were identified. Neuroimages showed no signs of stroke. Comparison of characteristics between SLE positive versus negative patients’ group showed no differences. Acute and chronic phenotypes of patients with PMM2-CDG or CACNA1A channelopathies show similarities. Hypoglycosylation of both CaV2.1 subunits (α1A and α2α) induced gain-of-function effects on channel gating that mirrored those reported for pathogenic CACNA1A mutations linked to FHM and ataxia. Unoccupied N-glycosylation site N283 at α1A contributes to a gain-of-function by lessening CaV2.1 inactivation. Hypoglycosylation of the α2δ subunit also participates in the gain-of-function effect by promoting voltage-dependent opening of the CaV2.1 channel. CaV2.1 hypoglycosylation may cause ataxia and SLEs in PMM2-CDG patients. Aberrant CaV2.1 N-glycosylation as a novel pathomechanism in PMM2-CDG opens new therapeutic possibilities.

Abstract: Stroke-like episodes (SLE) occur in phosphomannomutase deficiency (PMM2-CDG), and may complicate the course of channelopathies related to Familial Hemiplegic Migraine (FHM) caused by mutations in CACNA1A (encoding Ca V 2.1 channel). The underlying pathomechanisms are unknown. We analyze clinical variables to detect risk factors for SLE in a series of 43 PMM2-CDG patients. We explore the hypothesis of abnormal Ca V 2.1 function due to aberrant N-glycosylation as a potential novel pathomechanism of SLE and ataxia in PMM2-CDG by using whole-cell patch-clamp, N-glycosylation blockade and mutagenesis. Nine SLE were identified. Neuroimages showed no signs of stroke. Comparison of characteristics between SLE positive versus negative patients' group
Stroke-like episodes (SLE) are among the acute neurological complications that may occur in PMM2-CDG patients. SLE are typically triggered by infections and have been described in about 20-55% of patients with PMM2-CDG [5][6][7][8]. The term SLE was initially coined for MELAS (Mitochondrial myopathy, Encephalopathy, Lactic acidosis, and Strokelike episodes) syndrome to stress the non-ischemic origin of their events [9], but it may be used for different neurological diseases that are associated with focal deficits that mimic clinically, but not neuroradiologically, an ischemic injury. SLE in PMM2-CDG patients are characterized by confusional status, mono-or hemiparesis, and sometimes, epileptic seizures. The underlying pathomechanisms are not fully understood, and clinical guidelines helping their diagnosis, and prevention and treatment are missing, resulting in a very stressful situation for families and doctors.
Although coagulation abnormalities are very frequent in PMM2-CDG, no significant correlations have been found between the presence of thrombophilia and the occurrence of SLE [5][6][7][8]. Besides, the majority of patients with PMM2-CDG and SLE did not show vascular occlusions on magnetic resonance angiography, the affected brain lesions did not reveal restricted diffusion, or did not follow well-defined vascular territories on MRI (magnetic resonance imaging) [5][6][7][8][9][10][11].
SLE may also complicate the course of other neurological diseases such as channelopathies related to Familial Hemiplegic Migraine (FHM), a paroxysmal neurological disease caused by mutations in CACNA1A (encoding the neuronal pore-forming Ca V 2.1 channel α 1A subunit), ATP1A2 (encoding the Na + , K + -ATPase pump α 2 subunit) or SCN1A (encoding the neuronal Na V 1.1 channel α 1 subunit). As for PMM2-CDG, the pathogenesis of SLE in channelopathies is still not fully understood [12,13]. Interestingly, some clinical, neuroimaging and neurophysiological features of PMM2-CDG patients and CACNA1A mutated patients are similar [13], including not only SLE, but also ataxia, ocular motor disturbances, and cerebellar atrophy on MRI [14]. Ion channels are highly N-glycosylated and we are not aware of studies on the effect of an abnormal glycosylation of the α 1A channel subunit on Ca V 2.1 functional expression and gating. As other high-voltage activated (HVA) Ca 2+ channels, Ca V 2.1 also contains at least the regulatory α 2 δ and β subunits [15]. Impaired glycosylation of the auxiliary α 2 δ subunit reduces the functional expression of different HVA channels (including Ca V 2.1) in the plasma membrane [16][17][18], but its relevance (if any) on Ca V 2.1 channel gating is unknown. Moreover, increased or decreased Ca V 2.1-mediated Ca 2+ influx has been related to different clinical features. Thus, FHM relates to gain-of-function mutations on CACNA1A, while ataxia is associated with the alteration of a narrow window of Ca 2+ homeostasis in Purkinje cells due to either CACNA1A loss-or gain-of-function mutations [19].
We report on 9 SLE in a series of 43 PMM2-CDG patients and analyze different clinical, laboratory, and neuroimaging variables to detect potential SLE risk factors. We demonstrate the clinical similarities between the acute and chronic phenotypes of patients with PMM2-CDG or patients carrying CACNA1A mutations, establishing a novel link between both diseases. With those evident clinical similarities and knowing that gain-of-function CACNA1A mutations can lead to both paroxysmal neurological symptoms and cerebellar symptoms including cerebellar atrophy (such as progressive or congenital ataxia) [19], we explore the hypothesis of increased Ca V 2.1 activity due to deficient N-glycosylation as an underlying biological reason, offering a potential novel pathomechanism of SLE and cerebellar syndrome in PMM2-CDG.

Results
Forty-three children and young adults with PMM2-CDG were included. Seven patients had nine SLE (mean age 7.5 ± 4.8 years, range 3.3-15.0 years) representing 17.9% of the patients (SLE positive group). Using our SLE definition, two clinical episodes were excluded: one caused by a true cerebrovascular event as confirmed by the MRI, and one caused by an epileptic seizure occurring at the onset of an infection process of the brain.
Thirty-two PMM2-CDG patients (mean age 16.2 ± 9.6, range 4.2-38 years) did not have SLE and were used as PMM2-CDG control group (SLE negative group). Table 1 summarizes the age at the SLE, molecular studies, clinical presentation and follow-up, neurophysiological and neuroimaging findings of all patients.

Stroke-Like Episodes
In six SLE, a mild-to-moderate head trauma was identified as a potential trigger. SLE-related neurological symptoms started after a symptom-free interval ranging between 1 and 24 h. For three patients suffering from infections it was not possible to exactly quantify the symptom-free interval. Three out of nine episodes were characterized by focal neurological deficits and three patients presented with clinical seizures. All patients had abnormal consciousness (Supplementary Video 1).
Almost all patients developed hyperthermia at presentation leading to a laboratory work-up for a possible underlying infection or inflammatory process (Table 1). Acute phase proteins and complete blood counts were normal for all patients with SLE after head trauma.
In seven SLE a brain MRI was performed during the first 72 h. Neuroimaging showed the preexisting cerebellar atrophy with absence of new findings in all but Patient 1 (Table 1), who showed diffuse cortical swelling and FLAIR-hyperintense signal in the right hemisphere with no restricted diffusion, suggestive of vasogenic edema ( Figure 1). EEG revealed normal findings in six out of nine studies (Table 1). On the second day after SLE onset, the EEG of Patient 1 showed an asymmetric (right) slow background activity with moderately low voltage in temporal regions, and frontal intermittent rhythmic delta activity (FIRDA) in the right hemisphere ( Figure 1).
Antiepileptic drugs were used in six of nine SLE with two aims: to treat irritability using midazolam (two out of four cases), and to control or prevent epileptic seizures using levetiracetam, midazolam, valproic acid, phenytoin, and diazepam (Table 1). Neither anticoagulants nor antiaggregants were used in eight of nine episodes, as there was no evidence of a procoagulant situation. Patient 4 developed a deep venous thrombosis during her second SLE. Her coagulation tests revealed abnormal findings that required fibrinogen substitution.

SLE Versus SLE Negative Group
Statistical analysis of clinical, laboratory, and neuroimaging variables showed no significant differences between PMM2-CDG patients with and without SLE. Epidemiological and clinical data are shown in Table 2. Most of patients recovered completely within one week. Three patients, however, experienced a more delayed recovery (Table 1). After SLE onset, Patient 1 and Patient 6 developed episodic tonic upgaze that progressively decreased in frequency and disappeared three and five weeks after onset, respectively (Supplementary Video 2).

SLE Versus SLE Negative Group
Statistical analysis of clinical, laboratory, and neuroimaging variables showed no significant differences between PMM2-CDG patients with and without SLE. Epidemiological and clinical data are shown in Table 2.  Table 3 compares the acute and chronic clinical presentations that have been reported in patients with pathogenic variants in CACNA1A and PMM2, respectively. The similarities in acute and chronic neurological presentations suggest a possible common underlying pathomechanism for SLE. Table 3. Findings in patients carrying mutations in CACNA1A compared to findings in PMM2-CDG patients, including our cohort.
Infections, head trauma, angiography, alcohol ingestion [5]; Minutes to hours after the trigger.
Clonic convulsions/Low voltage pattern in affected areas, asymmetric slow background activity with moderately low frontal intermittent rhythmic delta activity (FIRDA) in the affected hemisphere [8].
Recovery 24 h to months. Complete recovery in the majority of the patients, but residual hemiplegia possible [14,22].
1h to several months. Complete recovery in the majority of the patients, but exceptionally residual motor symptoms may persist [5].

Deficient N-Glycosylation Alters Ca V 2.1 Functional Expression, Activation, and Inactivation
Whole-cell patch-clamp recordings obtained from HEK293 cells expressing Ca V 2.1 channels composed of α 1A , β 3 and α 2 δ 1 subunits, showed that tunicamycin treatment (which specifically blocks the first step of N-glycosylation by inhibiting N-acetylglucosamine transferase in the endoplasmic reticulum, Figure 2), reduced Ca V 2.1 Ca 2+ currents in a concentration-dependent manner. μg/mL, as indicated). The extent of α2δ glycosylation was identified by using antibody anti-α2 (1:500 dilution, Sigma D219, St. Louis, MO, USA) (top panels) and is shown as the difference in molecular weight between the glycosylated and unglycosylated forms. Molecular weight markers are indicated on the left. PNGase F was also added to protein extraction from tunicamycin-untreated cells in order to identify unglycosylated forms in vitro. Upper bands potentially corresponding to glycosylated α2δ progressively decrease as the concentration of tunicamycin increases, which in turn promotes the appearance of a lower band potentially corresponding to unglycosylated α2δ (lanes 2 to 6). In vitro PNGase F treatment (lane 1) shows two main bands potentially corresponding to unglycosylated α2δ (upper band) and unglycosylated α2 (lower band), as previously reported [26].Protein extraction from HEK293 cells transfected only with the vector plasmid was included as negative control (lane 7). For each experimental condition, the protein sample was probed with anti-tubulin (1:2000 dilution, Sigma T6074) (bottom panel) as loading control.
At lower concentrations, a smaller decrease of α2δ1 glycosylation by tunicamycin was found ( Figure 2) and the effect on the amplitude of CaV2.1 Ca 2+ currents was less important, with reductions in peak current density to −107.03 ± 25.1 pA/pF (n = 13) (p > 0.05) and to −26.7 ± 7.9 pA/pF (n = 16) (p < 0.001) induced by 0.2 and 0.6 μg/mL tunicamycin, respectively (Kruskal-Wallis test followed by Dunn post hoc test) ( Figure 4A,B). The CaV2.1 voltage-dependent activation curve was left-shifted and the potential for half-maximal channel activation (V1/2 activation, directly related to the energy necessary to open the channel) was significantly decreased in response to both 0.2 and 0.6 μg/mL tunicamycin treatments (by ~3.5 mV and ~5 mV, respectively) ( Figure 4C,D). Consistently, the maximum current amplitude was elicited by depolarizing pulses to +15 mV or +10 mV for control (vehicle) or tunicamycin treatments, respectively ( Figure 4B,C). The extent of α 2 δ glycosylation was identified by using antibody anti-α 2 (1:500 dilution, Sigma D219, St. Louis, MO, USA) (top panels) and is shown as the difference in molecular weight between the glycosylated and unglycosylated forms. Molecular weight markers are indicated on the left. PNGase F was also added to protein extraction from tunicamycin-untreated cells in order to identify unglycosylated forms in vitro. Upper bands potentially corresponding to glycosylated α 2 δ progressively decrease as the concentration of tunicamycin increases, which in turn promotes the appearance of a lower band potentially corresponding to unglycosylated α 2 δ (lanes 2 to 6). In vitro PNGase F treatment (lane 1) shows two main bands potentially corresponding to unglycosylated α 2 δ (upper band) and unglycosylated α 2 (lower band), as previously reported [26].Protein extraction from HEK293 cells transfected only with the vector plasmid was included as negative control (lane 7). For each experimental condition, the protein sample was probed with anti-tubulin (1:2000 dilution, Sigma T6074) (bottom panel) as loading control.
At lower concentrations, a smaller decrease of α 2 δ 1 glycosylation by tunicamycin was found ( Figure 2) and the effect on the amplitude of Ca V 2.1 Ca 2+ currents was less important, with reductions in peak current density to −107.03 ± 25.1 pA/pF (n = 13) (p > 0.05) and to −26.7 ± 7.9 pA/pF (n = 16) (p < 0.001) induced by 0.2 and 0.6 µg/mL tunicamycin, respectively (Kruskal-Wallis test followed by Dunn post hoc test) ( Figure 4A,B). The Ca V 2.1 voltage-dependent activation curve was left-shifted and the potential for half-maximal channel activation (V 1/2 activation, directly related to the energy necessary to open the channel) was significantly decreased in response to both 0.2 and 0.6 µg/mL tunicamycin treatments (by~3.5 mV and~5 mV, respectively) ( Figure 4C,D). Consistently, the maximum current amplitude was elicited by depolarizing pulses to +15 mV or +10 mV for control (vehicle) or tunicamycin treatments, respectively ( Figure 4B,C).     Tunicamycin (0.6 µg/mL) also impaired the inactivation of Ca V 2.1 Ca 2+ currents: compared to the control (vehicle) condition, the degree of inactivation at the end of 3s depolarizing pulses to +20 ( Figure 5A,B) or 0 mV ( Figure 5D,E) was lower (by~14% and~16%, respectively); and there was a tendency for slow Ca V 2.1 inactivation, which reached statistical significance for Ca 2+ currents elicited by depolarization to 0 mV ( Figure 5A,C,D,F).
Tunicamycin (0.6 μg/mL) also impaired the inactivation of CaV2.1 Ca 2+ currents: compared to the control (vehicle) condition, the degree of inactivation at the end of 3s depolarizing pulses to +20 ( Figure 5A,B) or 0 mV ( Figure 5D,E) was lower (by ~14% and ~16%, respectively); and there was a tendency for slow CaV2.1 inactivation, which reached statistical significance for Ca 2+ currents elicited by depolarization to 0 mV ( Figure 5A,C,D,F).

Mutation of α1A Potential Glycosylation Site Mimics Tunicamycin Effect on CaV2.1 Inactivation
To evaluate whether abnormal glycosylation of the α1A channel subunit may contribute to the effect produced by tunicamycin on CaV2.1 biophysical properties, we first searched for potential N-glycosylation sites in the sequence of the human α1A subunit (GenBank No. O00555) by using Uniprot Knowledgebase (www.uniprot.org, access on 30 November 2017). One single glycosylation site at residue N283, located in the extracellular P-loop region at domain I (DI) of α1A, was identified ( Figure 6A). This asparagine residue is highly conserved through evolution ( Figure 6B, top). In addition, potential glycosylation of a single N residue at P-loop-DI is a trait shared by all human high-voltage activated (HVA) Ca 2+ channels ( Figure 6B, bottom). Data are expressed as the mean ± SEM of the number of experiments shown in brackets. *** p < 0.001 (Kruskal-Wallis test followed by Dunn post hoc test) and ** p < 0.01 (Student's t-test) versus the control (vehicle) condition. (C,F) Average τ inactivation values of Ca 2+ currents through Ca V 2.1 channels expressed in HEK293 cells treated with DMSO (vehicle) or with tunicamycin (0.2 and 0.6 µg/mL), elicited by a 3 s depolarizing pulse to +20 mV or 0 mV, as indicated. Data are expressed as the mean ± SEM of the number of experiments shown in brackets. * p < 0.05 versus the control (vehicle) condition (Mann-Whitney U-test).

Mutation of α 1A Potential Glycosylation Site Mimics Tunicamycin Effect on Ca V 2.1 Inactivation
To evaluate whether abnormal glycosylation of the α 1A channel subunit may contribute to the effect produced by tunicamycin on Ca V 2.1 biophysical properties, we first searched for potential N-glycosylation sites in the sequence of the human α 1A subunit (GenBank No. O00555) by using Uniprot Knowledgebase (www.uniprot.org, access on 30 November 2017). One single glycosylation site at residue N283, located in the extracellular P-loop region at domain I (DI) of α 1A , was identified ( Figure 6A). This asparagine residue is highly conserved through evolution ( Figure 6B, top). In addition, potential glycosylation of a single N residue at P-loop-DI is a trait shared by all human high-voltage activated (HVA) Ca 2+ channels ( Figure 6B, bottom). Hence, we compared the biophysical properties of CaV2.1 channels composed by wild-type (WT) or N283Q glycosylation mutant α1A, β3 and α2δ1 subunits, heterologously expressed in HEK293 cells. Peak Ca 2+ current density for N283Q mutant CaV2.1 channels were ~72% lower than for WT channels ( Figure 7A,B). N283Q did not alter CaV2.1 voltage-dependent activation ( Figure 7C,D). Interestingly, tunicamycin (0.6 μg/mL) shifted V1/2 activation for N283Q CaV2.1 channels to less depolarized potentials (by ~4.5 mV) ( Figure 7E-H  Hence, we compared the biophysical properties of Ca V 2.1 channels composed by wild-type (WT) or N283Q glycosylation mutant α 1A , β 3 and α 2 δ 1 subunits, heterologously expressed in HEK293 cells. Peak Ca 2+ current density for N283Q mutant Ca V 2.1 channels were~72% lower than for WT channels ( Figure 7A,B). N283Q did not alter Ca V 2.1 voltage-dependent activation ( Figure 7C,D). Interestingly, tunicamycin (0.6 µg/mL) shifted V 1/2 activation for N283Q Ca V 2.1 channels to less depolarized potentials (by~4.5 mV) ( Figure 7E-H   illustrating the decrease in Ca 2+ current density through Ca V 2.1 channels containing the mutation at the α 1A glycosylation site N283, which exhibit voltage dependence of activation similar to that of WT channels. Dotted lines mark the zero current level. Average Ca 2+ current density-voltage relationships (B) and normalized I-V curves (C) for WT (open circles, n = 13) and N283Q (filled dark cyan triangles, n = 19) Ca V 2.1 channels. N283Q reduces maximal Ca 2+ current density (obtained by membrane depolarization to +15 mV) from −69.4 ± 12.9 pA/pF (n = 13) to −19.5 ± 3.2 pA/pF (n = 19) (p < 0.001, Mann-Whitney U-test). (D) N283Q mutation has no significant effect on the V 1/2 for Ca V 2.1 channel activation (estimated from normalized I-V curves shown in C as indicated in Materials and Methods, p = 0.798, Student's t-test). (E) Current traces elicited by 20 ms depolarizing pulses from −80 mV to the indicated voltages (inset) illustrating the shift of activation to lower depolarization induced by 0.6 µg/mL tunicamycin on N283Q mutant Ca V 2.1 channels containing β 3 and α 2 δ 1 subunits. The zero current level is indicated by dotted lines. Average Ca 2+ current density-voltage relationships (F) and normalized I-V curves (G) for N283Q mutant Ca V 2.1 channels expressed in DMSO-treated cells (filled cyan triangles, n = 10) and in cells treated with 0.6 µg/mL tunicamycin (filled green circles, n = 8). Peak Ca 2+ current density through N283Q Ca V 2.1 channels after vehicle (DMSO) and tunicamycin treatments were −23.3 ± 5.6 pA/pF (n = 10) and −11.6 ± 3.6 pA/pF (n = 8), respectively (p = 0.06, Student's t-test).

Patients
Patients with a molecular diagnosis of PMM2-CDG were recruited from the hospitals collaborating in the Spanish PMM2-CDG Network, and were followed from May 2014 to December 2016.
We used the following SLE definition: "Acute event consisting of sudden onset of a focal neurological deficit, irritability or decreased consciousness that may associate with seizures, headache or other transient symptoms, in the absence of another diagnosis explaining these symptoms". The definition does not include neurophysiological and neuroimaging findings. Patients with "true" ischemic stroke on MRI were excluded.
We collected epidemiological and molecular data, potential triggers of SLE, clinical findings that occurred shortly before the episode, symptoms during the episode and the recovery process, and laboratory, neurophysiological (EEG), and neuroimaging (computed tomography (CT) and MRI) findings. Laboratory studies included blood cell counts, liver enzymes, acute phase reactants Figure 8. N283Q glycosylation site mutation lessens Ca V 2.1 channel inactivation. Ca 2+ current traces (normalized to the corresponding peak amplitude) illustrating differential inactivation of WT (black traces) and N283Q (cyan traces) Ca V 2.1 channels, in response to a 3 s depolarizing pulse to +20 mV (A) or 0 mV (D). Dotted lines indicate the zero current level. (B,E) Average Ca 2+ current inactivation (in %) at the end of these 3s depolarizing pulses obtained from HEK293 cells expressing either WT or N283Q Ca V 2.1 channels. Data are expressed as the mean ± SEM of the number of experiments shown in brackets (*** p < 0.001 and ** p < 0.01 versus WT, Mann-Whitney U-test). (C,F) Average τ inactivation values of Ca 2+ currents through WT (open bars) and N283Q (cyan bars) Ca V 2.1 channels expressed in HEK293 cells, elicited by a 3 s depolarizing pulse to +20 mV or 0 mV, as indicated. Data are expressed as the mean ± SEM of the number of experiments shown in brackets (*** p < 0.001 and ** p < 0.01 versus WT, Mann-Whitney U-test).

Patients
Patients with a molecular diagnosis of PMM2-CDG were recruited from the hospitals collaborating in the Spanish PMM2-CDG Network, and were followed from May 2014 to December 2016.
We used the following SLE definition: "Acute event consisting of sudden onset of a focal neurological deficit, irritability or decreased consciousness that may associate with seizures, headache or other transient symptoms, in the absence of another diagnosis explaining these symptoms". The definition does not include neurophysiological and neuroimaging findings. Patients with "true" ischemic stroke on MRI were excluded.
We collected epidemiological and molecular data, potential triggers of SLE, clinical findings that occurred shortly before the episode, symptoms during the episode and the recovery process, and laboratory, neurophysiological (EEG), and neuroimaging (computed tomography (CT) and MRI) findings. Laboratory studies included blood cell counts, liver enzymes, acute phase reactants and coagulation parameters. Clinical severity of cerebellar syndrome was assessed through International Cooperative Ataxia Rating Scale (ICARS) [25].
To compare PMM2-CDG patients with SLE (SLE group) and without SLE (SLE negative group), we used epidemiological and molecular data, laboratory findings, personal history of epilepsy, personal or familial history of vascular events, and neuroimaging findings, using the midsagittal vermis relative diameter (MVRD) [25,27].
Biochemical studies were performed during the SLE. For the patients without SLE, the most abnormal liver function and coagulation laboratory values were included in the statistical analysis. For this comparison, four patients with early death in young childhood (before 2 years) were excluded, because of their severe systemic involvement.
Molecular studies had been carried out in all PMM2-CDG patients enrolled. Genetic analysis was performed in the "Centro de Diagnóstico de Enfermedades Moleculares" in Madrid. Total mRNA and genomic DNA were isolated from venous whole blood or patient-derived fibroblasts using a MagnaPure system following the manufacturer's protocol (Roche Applied Science, Indianapolis). Mutational analysis was performed by genomic DNA analysis both in patients' and parents' samples to assure that both changes are on different alleles and to rule out the presence of a large genomic rearrangement. In some cases, the effect on splicing was analyzed by cDNA profile analysis. The primers used for cDNA and genomic DNA amplifications were designed using the ENSEMBL database (http://www.ensembl.org/index.html, ENSG00000140650) and GenBank accession number NM_000303.2.

Literature Review
To compare the cerebellar syndrome and SLE in PMM2-CDG patients and the phenotype related to CACNA1A mutated patients, we searched PubMed for articles on PMM2, PMM2-CDG, CDG-Ia, congenital disorders of glycosylation, CACNA1A, familial hemiplegic migraine, episodic ataxia type 2 and spinocerebellar ataxia type 6 that have been published between 1 January 1980, and 31 May 2017, and used different combinations of these terms.

cDNA Constructs
cDNA of the human voltage-gated Ca 2+ (Ca V 2.1) channel α 1A wild-type (WT) subunit (originally cloned into a pCMV vector) was a gift from Professor J. Striessnig (University of Innsbruck, Austria). cDNAs of the rabbit α 2 δ 1 and rat β 3 regulatory subunits (subcloned into a pcDNA3 expression vector) were gifts from Dr. L. Birnbaumer (National Institutes of Health, Durham, NC, USA). Ca V 2.1 N283Q mutant channel was generated by site-directed mutagenesis of the human α 1A cDNA (GenScript Corporation, Piscatway, NJ, USA). All cDNA clones used in this study were sequenced in full to confirm their integrity.

Inhibition of N-Glycosylation in Live Cells and Western Blot
One day after transfection, cells were grown for another day in the presence of tunicamycin (0.2, 0.6 or 2 µg/mL) or vehicle (dimethyl sulfoxide, DMSO). At the end of the treatment, cells were washed with 1× PBS, returned to the incubator, and electrophysiological recordings were performed 4 h later at room temperature (22-24 • C). For Western Blot analysis of glycosylated fragments of heterologously expressed α 2 δ 1 subunit, after 4 h from recovery of the above mentioned tunicamycin or vehicle (DMSO) treatments, HEK293 cells were lysed on ice in lysis buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.5% v/v Nonidet P-40, 5 mM EDTA, 1 mM DTT, 10 mM β-glycerolphosphate, 0.1 mM Na 3 VO 4 , 1 µg/mL pepstatin, 2 µg/mL aprotinin, 0.1 mM phenylmethylsulfonyl fluoride) containing protease inhibitors (Complete Mini protease inhibitor cocktail, Roche). Lysates were vortexed for 15 min at 4 • C and centrifuged at 4 • C for 7 min at 13,500× g to pellet debris. Next, protein concentrations of cleared lysates were determined using Pierce BCA protein assay kit (Thermo Scientific, Madrid, Spain). Cell lysates (20 µg of proteins) were first incubated under denaturing conditions (0.5% SDS and 40 mM DTT) at 100 • C for 10 min, and then incubated in presence or absence of 500 units of peptide-N-glycosidase F (PNGase F, New England Biolab, Ipswich, MA, USA) during 1 h at 37 • C, according to the manufacturer's instructions. Incubated lysates were denatured to inactivate PNGase by incubation for 10 min at 80 • C, with 4× LDS Sample Buffer (Life Technologies, Carlsbad, CA, USA) and 10× sample reducing agent (Life Technologies). Samples were electrophoresed on an 8% SDS-polyacrylamide denaturating gel, transferred to a nitrocellulose membrane with iBlot (Invitrogen, Madrid, Spain), and probed with anti-rabbit Ca V α 2 antibody (1:500 dilution, Sigma D219) and anti-tubulin (1:2000 dilution, Sigma T6074) as a loading control, and mouse secondary antibody (GE Healthcare, Piscataway, NJ, USA). Signal was detected with the SuperSignal West Pico Chemiluminescent Substrate (Thermo Scientific). Blots were visualized with the ChemiDoc XRS documentation system (Bio-Rad, Hercules, CA, USA).

Electrophysiology
Ca 2+ currents (I Ca 2+ ) through Ca V 2.1 channels were measured using the whole-cell configuration of the patch-clamp technique as described in detail previously [28]. In brief, pipettes had a resistance Maximal inward Ca 2+ current (I Ca 2+ ) densities in response to 20 ms depolarizing pulses were measured from cells clamped at −80 mV, as detailed in previous report [28]. To evaluate the effect of tunicamycin and mutation N283Q on Ca V 2.1 voltage-dependent activation, normalized current-voltage (I-V) relationships were individually fitted with the modified Boltzmann equation, as previously reported [28]: where I is the peak current, G max is the maximal conductance of the cell, V is the membrane potential, V rev is the extrapolated reversal potential of I Ca 2+ , V 1/2 act is the voltage for half-maximal current activation, and k act is the slope factor of the Boltzmann term. Time constant for Ca V 2.1 inactivation (τ inactivation ) under the different experimental conditions was calculated after single exponential fits of the I Ca 2+ inactivation phase during a 3 s pulse from a holding potential of −80 mV to a test potential of +20 mV or 0 mV. The degree of I Ca 2+ inactivation (in %) at the end of these 3 s depolarizing pulses was also measured.

Ethics Statement
This study was approved by the Research and Ethics Committee of the "Sant Joan de Déu Hospital (SJDH)" (Internal code PIC-108-14) (Internal code PIC-108-14). Parents gave their written informed consent. Samples and data were obtained in accordance with the Helsinki Declaration of 1964, as revised in October 2013 (Fortaleza, Brazil).

Statistical Analysis
Statistical analysis was performed using Program R 3.2 (Vienna, Austria). Numerical variables are compared between groups by means of Mann-Whitney's U-test (non-parametric), and categorical variables by Fisher's exact test. For electrophysiological analysis, data are presented as the means ± S.E.M. and statistical tests included Kruskal-Wallis test followed by Dunn post hoc test, Student's t-test or Mann-Whitney U-test, as appropriate. Differences were considered significant if p < 0.05.

Discussion and Conclusions
In this multicenter cohort of 43 PMM2-CDG patients, we found a SLE incidence of 16.2% (7/39), slightly lower than previously reported (20-55%) [5][6][7]. Underdiagnosis due to the lack of defined clinical criteria for SLE, might explain such lower incidence. This situation also makes the differentiation from other more common complications of PMM2-CDG-such as epileptic seizures and true ischemic events-very challenging. To overcome this problem, we proposed and applied a SLE clinical definition.
Our results suggest two main trigger factors for SLE in PMM2-CDG: head trauma and viral infection. The association between cranial traumatisms and the occurrence of SLE in PMM2-CDG patients has been occasionally reported [5]. However, the role of head trauma and hyperthermia in triggering encephalopathic episodes is also described in other inborn errors of metabolism and child neurology conditions [29][30][31][32]. Here we report on six SLE occurring after mild accidental head trauma, a common situation in children with ataxia and hypotonia. Why only a minority of PMM2-CDG patients experiencing mild head trauma during their life develop SLE is currently unknown.
In our cohort, MRI during SLE revealed diffuse cortical edema in the parieto-occipital region of right hemisphere in Patient 1 (Figure 1), whereas no acute brain injuries were observed in the other eight SLE, as reported in many patients [8,20].
In PMM2-CDG different underlying pathomechanisms have been suggested for SLE, being hypoperfusion or ischemia the most discussed. However, they do not completely explain the nature and temporal course of the neuronal dysfunction [8,10]. In our PMM2-CDG cohort, no vascular injury was seen on brain MRI. Besides, no differences in coagulation factors were found between PMM2-CDG patients with and without SLE ( Table 2).
Three patients showed abnormal EEG findings, including asymmetrical background activity with FIRDA, a characteristic finding of severe encephalopathy, but nonspecific for SLE (Figure 1), supporting the need of EEG monitoring in SLE and the use of antiepileptic drugs to treat neuronal dysfunction [8].
Although rarely reported [20], our results show that hyperthermia without any sign of infection in blood or CSF is consistently present during SLE, presumably secondary to hypothalamic thermoregulatory dysregulation. Although contribution of hyperthermia to cerebral/cerebellar damage has not been documented in PMM2-CDG, aiming at normothermia appears as a reasonable goal in SLE, in line with general neuroprotective recommendations in stroke.
Analysis of epidemiological, clinical, laboratory, and neuroimaging findings did not reveal differences between PMM2-CDG patients with and without SLE. Regarding the genotype, all kind of pathogenic variants were distributed among both groups.
Similarities in clinical presentation encompass both the acute and chronic manifestations of patients with PMM2-CDG and CACNA1A channelopathies (Table 3). Those in the acute phenotype led to the not previously explored hypothesis of a similar underlying pathomechanism for FHM episodes (related to CACNA1A gain-of-function mutations) and SLE in PMM2-CDG. The characteristics of episodic encephalopathic crises after minor head trauma associated with seizures, hemiparesis, hyperthermia, altered consciousness and reversible hemispheric swelling on brain MRI were common features of both disorders [12,13]. Also, a chronic course including cerebellar ataxia, nystagmus, and episodes of tonic upgaze has been described in patients with Ca V 2.1 (CACNA1A) channelopathy [33], and in PMM2-CDG patients recovering from severe SLE as shown by this study (Video 2). Interestingly, in patients with CACNA1A-linked FHM, verapamil or acetazolamide may improve the symptoms and prevent recurrences [22][23][24]34]. This observation is of great importance because the similarity between the two diseases may be an argument to explore a potentially beneficial therapeutic use of these drugs in PMM2-CDG patients.
As a novel pathophysiological approach, we explored the hypothesis of abnormal Ca V 2.1 function due to aberrant N-glycosylation as the potential cause of SLE in PMM2-CDG. We studied the effect of hypoglycosylation of heterologously expressed Ca V 2.1 channels on their functional expression and gating. Our results suggest that α 2 δ 1 hypoglycosylation induced by strong inhibition of N-glycosylation with 2 µg/mL tunicamycin may mediate the reduction in Ca 2+ current density through Ca V 2.1 channels. This agree with previous reports showing that: (1) glycosylation of specific asparagine residues at α 2 δ increases cationic current density through distinct Ca V channels (including Ca V 2.1) [16][17][18], (2) mutation of several α 2 δ consensus N-glycosylation sites disrupt its cell surface expression and diminish protein stability [18], which may reduce the number of functional Ca V channels in the membrane by destabilizing the interaction between α 2 δ and pore-forming α 1 subunits [16][17][18]. At lower tunicamycin concentrations (0.2 and 0.6 µg/mL) the reduction of Ca V 2.1 Ca 2+ current density is less important and a double gain-of-function effect is observed: 1) the Ca V 2.1 voltage-dependent activation is favored, as indicated by the tunicamycin-induced shift of the current activation curve towards less depolarized potentials (by~3.5-5 mV); and 2) the Ca V 2.1 channel inactivation is impaired due to a lower degree of inactivation (that allows a persistent Ca 2+ influx at the end of long (seconds) depolarizing pulses) and the slowing of inactivation kinetic.
A gain-of-function in the voltage-dependence of Ca V 2.1 activation is the common feature described for all FHM-linked CACNA1A mutations that have been functionally analyzed, both in heterologous expression systems and in excitatory neurons from FHM Ca V 2.1 knock-in mice: they reduce the voltage threshold of channel activation by 5.6-20.9 mV [28,[35][36][37][38][39][40][41]. Besides, for some CACNA1A mutations, a relationship between impaired Ca V 2.1 inactivation and a greater severity in the clinical phenotype (including progressive cerebellar or congenital ataxia) has been described [41][42][43][44].
Despite the association with current density reduction, increase in Ca V 2.1 activity due to both favored voltage-dependent activation and impaired inactivation results in higher Ca 2+ influx into the cell in response to stimuli of physiological relevance in neurons, such as either single or trains of action potential [41]. An augment in intracellular Ca 2+ influx leads to the specific increase in cortical excitatory neurotransmission that promotes the generation and propagation of cortical spreading depression, which correlates with the neurological symptoms (auras) and triggers the headache phase that are typical for FHM [36][37][38][39][45][46][47][48]. In the cerebellum, enhanced Ca V 2.1 activity favors the generation of somatic action potentials and dendritic Ca 2+ spikes in Purkinje cells (PCs). This induces PCs hyperexcitability and, hence, mild constant cerebellar ataxia in a FHM knock-in mouse model [19,46].
Our functional analysis of the N283Q mutant Ca V 2.1 channel α 1A subunit suggests that N283 glycosylation is essential for proper Ca V 2.1 inactivation. To our knowledge, this is the first study on the importance of N-glycosylation at residues located at the P-loop-DI region of pore-forming α 1 subunits in the biophysical properties of HVA Ca V complexes formed by α 1 , β and α 2 δ. Only for low-voltage activated Ca V 3.2 channels it has been shown that α 1 subunit hypoglycosylation slowed inactivation kinetic [49,50].
Since α 1A mutation N283Q does not affect the voltage-dependence of Ca V 2.1 opening ( Figure 7A-D), the tunicamycin-induced gain-of-function effect on channel activation seems to be independent of aberrant α 1A subunit glycosylation, and is most probably due to α 2 δ hypoglycosylation. Supporting this observation, tunicamycin treatment of cells expressing the Ca V 2.1 channel formed by the N283Q mutant α 1A , β 3 and α 2 δ 1 subunits induces a reduction in the V 1/2 for channel activation of similar magnitude to that induced on the wild-type (WT) channel (Figures 4 and 7E-H). It is well known that α 2 δ subunits not only affect the surface expression of HVA Ca V channels but also their function. α 2 δ effect on Ca V activation most likely depends on the particular pore-forming α 1 subunit of the channel. α 2 δ 1 favored the opening of Ca V 1.2 (α 1C ) channels by shifting their activation 10 mV to more hyperpolarized voltages [51,52]. Accordingly, the presence of α 2 δ 1 glycosylation mutants impaired Ca V 1.2 activation by voltage [18]. On the contrary, α 2 δ 1 impaired the opening of Ca V 2.3 (α 1E ) channels by depolarizing their voltage-dependent activation (by 7-12 mV) [53]. The latter observation matches our data showing that tunicamycin-induced hypoglycosylation favors the activation of another Ca V 2 family member (Ca V 2.1) containing the regulatory α 2 δ subunit along with either WT or N283Q glycosylation mutant α 1A subunit (Figures 4 and 7E-H).
In summary, we propose a clinical definition for SLE to avoid misclassification of acute events in PMM2-CDG patients. The lack of well-defined diagnostic criteria and the high level of suspicion required for the diagnosis may result in SLE underdiagnosis. Our data suggest that mild cranial trauma and infections may trigger SLE in PMM2-CDG patients. Similarities in the clinical presentation of SLE in PMM2-CDG and CACNA1A-related encephalopathies suggest a possible common underlying pathomechanism. Our findings show that hypoglycosylation of different Ca V 2.1 subunits promotes gain-of-function effects that are similar to those induced by CACNA1A pathogenic variants linked to FHM and different forms of ataxia. Hence, our results support the hypothesis that aberrant Ca V 2.1 N-glycosylation may cause not only a cerebellar syndrome, but also SLE in PMM2-CDG patients. This is not to the detriment of the possible contribution of other hypoglycosylated proteins in the phenotypic expression of this disease, as we are aware that is difficult to extrapolate total cell hypoglycosylation to the effects on specific glycosylation sites at Ca V 2.1 channel subunits. Clarifying SLE pathogenesis in PMM2-CDG may prove paramount to develop prophylactic or therapeutic strategies for this acute and stressful complication.