A schizophrenia risk locus alters brain metal transport and plasma glycosylation

A common missense variant (rs13107325 (C->T), A391T) in SLC39A8, a gene encoding a transporter of divalent cations including manganese (Mn), is convincingly associated with schizophrenia and has pleiotropic effects on several additional brain-related phenotypes. Homozygous loss-of-function mutations in SLC39A8 result in undetectable serum Mn and a Congenital Disorder of Glycosylation (CDG) due to the exquisite sensitivity of glycosyltransferases to Mn concentration. Here, we identified Mn-related changes in human carriers of the SLC39A8 missense allele. Analysis of structural brain MRI scans from the UK Biobank showed a dose-dependent change in the ratio of T2w to T1w signal in several brain regions, presumably from altered transport of paramagnetic cations including Mn. We confirmed a specific reduction of serum Mn and showed through comprehensive profiling reduced complexity and branching of the plasma protein N-glycome in both heterozygous and homozygous minor allele carriers. N-glycome profiling of two individuals with SLC39A8-CDG showed similar but more severe alterations in branching that improved with Mn supplementation, suggesting that hypofunction of the common missense variant exists on a spectrum with potential for reversibility. Characterizing the functional impact of this variant may enhance our understanding of schizophrenia pathogenesis and identify novel therapeutic targets and biomarkers of disease.

SLC39A8 (aka ZIP8) is a transmembrane protein that cotransports divalent cations with bicarbonate; though it is capable of transporting Zn, Fe, Cu, Co, and Cd in cells, multiple studies suggest the primary physiologic role in humans is the transport of Mn (12)(13)(14). Manganese is an essential trace element for human health and affects neuronal function and development of dopaminergic neurons, although excess Mn is associated with disease (15)(16)(17). In the brain, Mn is at highest concentrations in the striatum, and Mn toxicity, also known as manganism, is characterized by a Parkinsonian phenotype resulting from dysfunction of the nigrostriatal pathway (16). Manganese is a cofactor for many enzymes, including superoxide dismutase, glutamine synthetase, pyruvate carboxylase, and arginase, as well as many glycosyltransferases such as β(1,4)-galactosyltransferase (18). A large number of Golgi glycosyltransferases, generally those containing a DxD metal-binding motif, uniquely require Mn as a cofactor (19,20). Glycosylation is a highly regulated, step-wise process of covalently attaching branched sugar polymers to proteins and lipids, and is known to affect nearly all biological pathways (21). Glycosylation plays a critical MEALER SLC39A8 MANUSCRIPT 5 role in human disease, with over 100 known Mendelian conditions, termed congenital disorders of glycosylation (CDGs), associated with mutations in glycotransferases and related genes (22). On proteins, glycans are most commonly attached to asparagine (N-linked) or serine/threonine (Olinked) residues. Disorders of N-glycosylation are divided into two distinct groups (type I and type II CDG) based on the observed transferrin glycosylation pattern, with type I indicating a defect in glycan assembly in the ER and type II suggestive of impaired modification of the side chain in the Golgi apparatus (23,24).
Two case reports in 2015 demonstrated the importance of SLC39A8 in human disease, as individuals harboring rare homozygous mutations in SLC39A8 displayed a severe type II congenital disorder of glycosylation (SLC39A8-CDG) and near total absence of blood Mn (25,26). Other markers related to the Mn-dependent enzymes pyruvate carboxylase and glutamine synthetase were normal, suggesting a unique vulnerability of glycosylation enzymes to Mn concentration in SLC39A8-CDG. Importantly, a marker of impaired glycosylation and some clinical phenotypes such as seizure activity improved following supplementation with glycosylation precursors (galactose and uridine) or Mn (25,27). A recent study by Rader and colleagues showed that SLC39A8 regulated Mn homeostasis through uptake from bile in inducible-and liver-specific-knockout mice (28). Serum protein N-glycans analyzed by MALDI-TOF in these mice was suggestive of impaired N-glycosylation. Analysis of plasma N-glycans in human rs13107325 homozygous minor allele carriers (n=12) showed a slight but significant increase in one N-glycan precursor (monosialo-monogalacto-biantennary glycan, abbreviated A2G1S1), though a complete N-glycan analysis was not reported (28).
The primary objective of our study was to measure Mn-related phenotypes with relevance to schizophrenia risk and brain function in human carriers of the SLC39A8 missense variant. We identified a dose-dependent association of the schizophrenia risk allele with changes in the T2w/T1w ratio in several brain regions. Measurement of 23 serum trace elements confirmed the specific reduction of only serum Mn observed in other studies on this variant, and plasma protein MEALER SLC39A8 MANUSCRIPT 6 N-glycosylation was altered in both heterozygous and homozygous carriers characterized by decreased branching and complexity of N-glycans. Analysis of SLC39A8-CDG plasma identified similar but more severe changes in protein N-glycosylation that were improved with Mn therapy, suggesting therapeutic intervention may be feasible in those carrying the SLC39A8 risk allele. 7
Previous studies have linked the rs13107325 minor allele (T) with brain MRI changes attributed to regional volumetric differences (8,29). Given the paramagnetic properties of Mn and its known effect on MRI relaxation time (30)(31)(32), we hypothesized that the signal change resulted from changes in local concentrations of Mn and related ions. Decreased Mn would result in longer relaxation times of both T1-and T2-weighted images (T1w, T2w); however, longer T1 leads to lower signal intensity on T1w images, while longer T2 leads to increased signal on T2w images.
We predicted that the ratio between the signal intensity of T2w and T1w images (T2w/T1w) would be a more sensitive parameter than either alone, with decreased Mn concentration in A391T carriers increasing in the T2w/T1w ratio. As UK Biobank images do not include T1 or T2 mapping as part of their protocol, comparing the ratio avoids problems with inter-subject normalization of T1w and T2w signal intensities due, for instance, to different coil loading factors and body size.
Brain MRI data were downloaded from the UK Biobank for 48 participants with homozygous minor allele genotype at rs13107325 (TT) along with 48 heterozygous minor (CT) and 48 homozygous major (CC) individuals matched on age, gender, smoking status, living area, BMI, and Townsend Deprivation Index as a proxy for socioeconomic status (Supp. Table 1). Due to problems with either the T1 or T2 images in some subjects, the final number of individuals with both scans included in the analysis were 47, 45, and 44 subjects for CC, CT, and TT genotypes, respectively. T2w/T1w ratios were compared between minor allele carriers (CT, TT) and controls (CC) using a t-test corrected for a false discovery rate of 5% on a pixel by pixel basis. In TT carriers, increased T2w/T1w ratios were observed in lateral putaminal (LPut) areas and diffusely through white matter (Fig. 1a). Contrary to our prediction, a decrease in the T2w/T1w ratio was observed in the globus pallidus interna (GPi) and substantia nigra (SN) of TT carriers, two areas of the brain with links to Mn toxicity and high levels of the divalent metal ion transporter DMT-1 (SLC11A2) (33). CT carriers display changes in similar regions and in the same direction as TT were significantly different between CC vs TT genotypes, though the effect sizes were much larger for T2w signal in the GPi and SN. Additional analyses using MRI data to classify by genotype showed excellent separation of the CC and TT genotypes using only the T2w/T1w ratios of GPi, SN, and LPut (Supp. Fig. 1a, 1b). We note that use of raw T1w or T2w signal is not a useful comparator given the differences in BMI between the CC and TT genotypes. Regression analysis of all demographic data and T2w/T1w ratios found a significant correlation only between the LPut and BMI only in TT carriers (Supp. Fig. 1c).
Though our sample size is smaller than prior studies (8,29), a preliminary analysis using FreeSurfer software (http://surfer.nmr.mgh.harvard.edu/) showed no detectable difference in volume between genotypes in any brain region, suggesting that MRI signal changes in A391T carriers likely result from differences in the concentrations of paramagnetic ions such as Mn (data not shown).  Secondary sex-stratified analysis of genotype-related changes in serum Mn concentration showed a similar pattern but fell short of statistical significance in males (CT and TT) and TT females, likely due to the reduced sample size and the small effect size of the variant on Mn concentrations (5) (Supp. Fig. 3). There was no overall significant difference in serum Mn by sex, and linear regression showed no correlation between serum Mn concentration and age or BMI (Supp. Fig. 4).

Analysis of plasma protein N-glycosylation showed reduced branching in CT and TT carriers
After confirming decreased serum Mn in A391T carriers from the Partners Biobank, we sought measure the plasma protein N-glycome based on rs13107325 genotype in the same individuals.
N-glycans of plasma proteins were measured from 5 µL samples following peptide:N-glycosidase F (PNGaseF) cleavage and permethylation, and analyzed using MALDI-TOF MS based on standard protocols (34,35). Plasma was studied in lieu of serum given the abundance of literature on human plasma glycosylation (36), and because our pilot analysis of serum and plasma from the same donors found no substantial differences (data not shown). Fifty-seven individual N-glycans were quantified after normalization for percent abundance within each sample. The overall N-glycome pattern, as illustrated by the 20 most abundant plasma N-glycans, was consistent with previous reports and similar between genotypes ( Fig. 3a) (36). Several individual glycans differed significantly based on genotype, and the direction of change was the same for the majority of individual N-glycans in CT and TT carriers (Supp. Table 3). A heat map of percent change (relative to CC) showed that larger glycans (m/z > 2851) consistently trend towards decreased abundance in CT and TT carriers (Fig. 3b).
To determine whether specific enzymatic machinery is uniquely affected by genotype and Mn levels, glycans sharing structural similarities (such as branching, fucosylation, sialylation, etc.) were analyzed together, providing a more comprehensive evaluation than a single N-glycan. For example, bisection of N-glycans is performed by a single enzyme, MGAT3; analysis of all bisected N-glycans would be a more accurate readout of MGAT3 function than a single bisected N-glycan. Classification of each glycan by category is included in supplementary material (Supp. Table 4).
Branching of N-glycans, or antennarity, is defined as the number of N-acetylglucosamine (GlcNAc) linkages to core mannose (Man) residues, and is a proxy for the complexity of Nglycans (37). We observed reduced branching in CT and TT carriers compared to CC ( Fig. 4; Table 1). In both CT and TT genotypes, there is a statistically significant increase in bi-antennary N-glycans (CC 88.0% vs CT 90.2%, p =0.0017; CC vs TT 90.1%, p =0.0082) and decrease in tri-  Table 5).
No significant change was observed in hybrid, bisecting, or core-fucosylated N-glycans (Supp. Table 6). Fucosylation of antenna, shown to be primarily tri-and tetra-antennary N-glycans (as opposed to core-fucosylation primarily on mono and bi-antennary structures) (36), was reduced in both CT and TT carriers relative to CC, though only significantly in the TT group (CC 2.09% vs CT 01.56%, p =0.146; CC vs TT 1.43%, p =0.037). Analysis based on terminal monosaccharides showed no significant changes across any category, though there was a trend towards less sialylated species in CT and TT carriers (Supp. Table 6). Analysis stratified on the number of each residue showed a similar pattern of reduced abundance of larger and more complex structures in CT and TT carriers relative to CC. Finally, there was no change in the overall representation of each individual monosaccharide in the total protein N-glycan pool between genotypes, suggesting that differences in branching are not due to altered availability of the enzymatic substrate (UDP-Gal, UDP-GlcNAc, etc.) (Supp. Table 6).

SLC39A8-CDG patients have increased precursor N-glycans and decreased complex Nglycans that are reversed following Mn supplementation
After completing glycome analysis of a common variant with a small effect, we sought to determine if more intolerant mutations in SLC39A8 result in a similar but larger effect. Advances in next-generation DNA sequencing has resulted in an expansion of the known congenital disorders of glycosylation to more than one-hundred (22). A pair of case studies in 2015 (25,26) and a recent report in 2017 (38) have identified multiple individuals with congenital disorders of glycosylation resulting from intolerant mutations in SLC39A8 inherited in a recessive manner. The clinical phenotypes are dramatic and overlap in intellectual disability, seizures, brain structural abnormalities, low to undetectable Mn, and impaired transferrin glycosylation. Transferrin Nglycosylation is a common screening test for CDGs, though recent efforts have focused on performing mass spectrometry (MS) based methods given the more complete and sensitive nature of the test (34).
We performed plasma protein N-glycan profiling of two individuals with CDGs caused by homozygous SLC39A8 mutations before and after Mn supplementation. A full characterization of the clinical presentations as well as Mn supplementation protocol is described elsewhere (27). In brief: Subject A is an 8-month-old female with a more severe phenotype, harboring two mutations in highly conserved sites of SLC39A8 (Gly38Arg, Ile340Asn), treated for ~1 year with Mn-sulfate after a cross-titration from galactose supplementation; Subject B is a 19-year-old female with a milder phenotype found to have 3 mutations in SLC39A8 (Val33Met/Ser335Thr, Gly204Cys) treated with Mn-sulfate for ~1 year. The full plasma protein N-glycan profile and spectra for each individual pre-and post-Mn therapy is included in the supplementary material (Supp. Fig. 6, Supp. Table 7). Each sample was analyzed twice and produced similar results. We highlight specific glycans and groups of glycans which: 1) changed in the same direction in both subject A and subject B; 2) were similar to changes observed in other CDGs using MALDI-TOF; 3) changed with Mn treatment; and 4) showed parallel changes relative to those observed in A391T carriers.
Similar to prior studies on CDGs, identification of a single or few individuals with a particular CDG limits direct and statistical comparison to meaningful matched control populations due to small sample size, age differences, and clinical variability. We include values for CC carriers using as comparators for our experimental protocols. As described in our methods, only serum from subject A pre-Mn supplementation was available and analyzed in lieu of plasma.
Relative abundance of A2G1S1, a monosialo-monogalacto bi-antennary N-glycan with permethylated m/z of 2227, is consistently elevated in plasma/serum across multiple CDGs (34), and was the only N-glycan reported as significantly elevated by Rader and colleagues in A391T homozygotes (28 (Fig. 5). Analysis based on terminal monosaccharides showed more variable differences between the subjects without any clear trends in both subjects (Supp. Table 8). Analysis stratified by residue shows a similar pattern of increased abundance of more complex structures following Mn treatment, and the overall representation of each monosaccharide in the total N-glycan pool remained similar before and after Mn treatment aside from a 50% increase in fucose in subject A (Supp. Table 8). In summary, the MALDI-TOF N-glycan profiles of two individuals with SLC39A8-CDG showed reduced complexity of N-glycans, which was increased after one year of Mn supplementation.

Discussion:
In our study, we found several Mn-related changes in human carriers of a missense variant in We hypothesized that lower plasma Mn levels would be associated with lower brain Mn levels in rs13107325 minor allele carriers, leading to increased T1 and T2 relaxation times with higher T2w signal and lower T1w signal. We identified increased T2w/T1w ratios in putamen (primarily lateral putamen) and white matter tracts consistent with this hypothesis. In contrast, the GPi and SN showed decreased T2w/T1w ratios in both CT and TT carriers, suggestive of increased paramagnetic metal ion deposition. Mn transport in the brain is complex, regulated by numerous transporters and pathways with significant overlap between iron (Fe) and Mn homeostasis (15,16,33,39,40). Both increased and decreased Fe lead to increased Mn transport into the brain (41,42). We suspect that low levels of Mn lead to increased uptake of Fe in some regions, particularly those with high affinity for divalent cations such as the GPi and SN. The GPi has high levels of DMT-1 expression and the highest rate of Mn deposition in welders exposed to Mn (32).
Studies in primates have demonstrated that the pallidum is uniquely susceptible to Mn accumulation (43,44), and manganese toxicity results in a Parkinsonian phenotype (16,17).
Interestingly, A391T is protective against Parkinson disorder, presumably from reduced uptake of Mn in to the pallidum or SN (10). Two recent studies associated A391T with brain MRI signal changes attributed to volumetric differences including increased gray matter in the caudate, putamen and cerebellum (8,29). MRI signal is affected by any factor influencing the magnetic field including Fe and Mn concentrations (45). Voxel-based morphometry (VBM) detects intensity changes between brain regions and is often interpreted as changes in grey matter density and volume. In contrast to VBM techniques, our preliminary studies measured volume using FreeSurfer and found no volumetric difference in any brain region based on rs13107325 genotype. We conclude that the direction of the T2w/T1w signal change in A391T carriers resulted from regional changes in Mn, Fe, or both.
Our study confirms that the A391T missense mutation selectively lowers serum Mn levels in both heterozygous and homozygous carriers. No differences were detected in any of the other trace elements transported through SLC39A8 (Zn, Co, and Cu), though effects on Cd could not be assessed as levels were below our method of detection limit. Fe was not measured in our study, though GWAS on Fe levels and Fe-related traits have not identified any associations with the A391T missense variant, and two recent studies show no difference in Fe concentration in A391T carriers, suggesting that serum Fe is not affected (9,46) Multiple studies demonstrate the importance of SLC39A8 to health and disease. Hypomorphic mice expressing 10-15% of basal SLC39A8 show severe growth stunting, dysmorphogenesis, and anemia (47,48). SLC39A8 is required during cardiac development and regulates zinc transport in endothelium (49). Recent studies also highlight a role for the SLC39A8 common variant in regulation of the gut microbiome and metal homeostasis in Crohn's disease (4,50,51).
Two case series from 2015 identified individuals with severe mutations in SLC39A8 presenting with a constellation of severe symptoms including intellectual disability, developmental delay, cerebellar atrophy, growth abnormalities, and seizures (25,26). A more recent report identified a pair of siblings presenting with a Leigh-like syndrome of intellectual disability, dystonia, seizures, cortical atrophy and basal ganglia T2 hyperintensities (38). Importantly, these studies demonstrate that the glycosylation deficits and some of the clinical phenotypes can be improved by oral supplementation of galactose and uridine to yield UDP-Gal, the key enzymatic substrate for β(1,4)-galactosyltransferase, as well as the attendant obligatory cofactor for this activity, Mn (25,27,38).
Human plasma protein N-glycosylation is an extensive area of research, with detailed descriptions of the abundance of each glycan, identification of proteins harboring each glycan, and how these glycans affect protein function (36). The plasma N-glycome is increasingly explored as a potential biomarker in a variety of settings including depression (52-54), pregnancy (55), IBD (56), Down syndrome (57), inflammation and metabolic health (58), and post-surgical changes (59). Given the repeated association of SLC39A8 with glycosylation defects in prior studies (25)(26)(27)(28)38), and the exquisite sensitivity of certain glycosyltransferases for Mn as an irreplaceable co-factor, we focused our study on glycosylation changes in A391T carriers.
Plasma protein N-glycome changes were similar in individuals carrying one or two copies of the hypo-functioning allele, suggesting a dominant effect. The most notable findings in individuals carrying the missense mutation were reduced branching and decreased complexity of large Nglycans. Rader and colleagues describe a small but significant increase in the abundance of the monosialo-monogalacto-biantennary precursor glycan A2G1S1 (m/z 2227) in a group of A391T homozygotes carriers (TT) but did not report the abundance of other plasma N-glycans (28).
A2G1S1 is commonly increased in CDGs and suggestive of decreased β(1,4)galactosyltransferase activity (60). We observed a ~20% increased abundance of A2G1S1 in both CT and TT carriers, though it did not reach statistical significance. We hypothesize that reduced Mn availability in mutation carriers results in a modest but broad reduction of glycosyltransferase activity in glycosyltransferases containing DxD domains including β(1,4)galactosyltransferase and the numerous N-acetylglucosaminyltransferases (MGATs) that control N-glycan branching. In addition, the activity of some glycosyltransferases lacking a classic DxD domain are also affected by variations in Mn concentration including sialyltransferases and may be affected (61).
We report the first plasma N-glycome analysis using MALDI-TOF MS of two individuals with severe SLC39A8 mutations causing CDGs before and after Mn supplementation (25,27). In plasma from these patients, we observed an elevation of the precursor A2G1S1 (m/z 2227) and a reduction of larger, more complex glycans including A3G3S3 and A3FG3S3 (m/z 3603 and 3777, respectively), similar to what has been reported in other type-II CDGs (34). The abundance of these three individual glycans normalized following Mn supplementation, and grouped analysis showed a dramatic increase in branched N-glycans and reduced high-mannose precursors following treatment. These changes parallel the reduced complexity of N-glycans in A391T carriers and suggest that such changes could be targeted with Mn supplementation. When and how to safely and effectively administer such a treatment remains to be determined. However, tracking of the branching of N-glycans could be a useful biomarker of treatment response and dose titration in CDGs and conditions associated with the A391T variant such as schizophrenia.
Glycosyltransferases associated with common, complex diseases tend to have specific tissuespecific expression profiles, isoenzymes with redundant activity, and function across multiple pathways, whereas glycosyltransferases associated with CDGs tend to have diffuse expression, lack redundant isoenzymes, and function within specific glycosylation pathways (62). A minority of glycosyltransferases are associated with both common diseases and CDGs, similar to what is observed with SLC39A8. GWAS have identified loci influencing the glycosylation of specific plasma proteins such as IgG as well as the total glycome (63)(64)(65). These studies do not report an association of rs13107325 with plasma and IgG N-glycosylation patterns and primarily identify SNPs near glycosyltransferase genes. This may result from differences in methodologies employed between studies (MALDI-TOF vs UPLC (65) and LC-ESI-MS (64)), minor allele frequency between cohorts, power/sample size, and effect size; rs13107325 may not be a major regulator of total plasma N-glycosylation relative to all genetic variation in glycosylation-related genes. Large, branched N-glycans are not generally found on IgG (36), thus it is less likely changes associated with rs13107325 would be identified in such studies. Though MALDI-TOF MS is a standard tool in the study of glycosylation disorders, the semi-quantitative of this assay only allows determination of abundance changes within each sample after normalization. In addition to quantitative assays, future studies of SLC39A8-A391T should include analysis of protein O-glycans and glycolipids, as well as disease-relevant systems including primary tissue, cell lines, and murine models to determine how this variant elicits such pleotropic effects.
In summary, we have demonstrated that a common missense variant in SLC39A8 is associated with multiple Mn-related phenotypes in both heterozygous and homozygous carriers, including changes in brain MRI T2w/T1w signals, reduced serum Mn levels, and decreased branching and complexity of plasma N-glycans. In addition, we identify parallel changes in the plasma Nglycome of SLC39A8-CDG patients that are reversed following Mn supplementation. Although the effect size of the common variant in SLC39A8 on glycosylation is relatively modest, translation of validated genetic variants to functional biologic pathways can provide critical insight for a disease.

Methods:
MRI data (T1 and T2 images) were downloaded from the UK Biobank 2018 release of ~15,000 participants' imaging data (https://www.ukbiobank.ac.uk). The UK Biobank imaging protocol has been described previously (8). Equal numbers (n = 48) of individuals with each of the three rs13107325 genotypes and brain imaging data were identified after matching for age, sex, smoking status, living area, body mass index (BMI) and Townsend Deprivation Index as a proxy for socioeconomic status. Ratio images were created by dividing the T2-weighted images (T2w) by the T1-weighted (T1w) images. Ratio images in TT and CT carriers were compared to CC subjects on a pixel-by-pixel basis using a t-test corrected for a false discovery rate of 5% using standard tools in the AFNI image analysis program (https://afni.nimh.nih.gov). Regions of interest (ROI), including the substantia nigra (SN), globus pallidus interna (GPi) and lateral putamen (LPut), were drawn after averaging all the images from the CC cohort, and then propagated to each individual ratio image. Values determined over the whole ROI were then compared using ANOVA analyses as described in Statistical Analysis, below.  Table 1. Plasma and serum were provided by Drs. Park and Marquardt from two individuals with a CDG associated with mutations in SLC39A8, as described previously (25,27). Of note, the plasma sample provided for Subject A pre-Mn supplementation did not produce an interpretable N-glycan profile despite two repeat analyses as the signal intensity was too low, presumably due to a problem with storage or transfer of the sample. An available serum sample of Subject A pre-Mn supplementation was analyzed in lieu of plasma and produced a reliable and replicable N-glycan profile. Given the limited quantity of samples available from these raredisease cases, and our observation that serum and plasma have similar N-glycan profiles as described above, the serum sample from Subject A was included in our report. Purification of plasma protein N-glycans was performed using standard protocols consistent with prior studies on CDGs (34) and are available through The National Center for Functional Glycomics website. (www.ncfg.hms.harvard.edu). In brief, 5 µL of plasma was lyophilized and resuspended in 20 µL 1X Rapid PNGaseF buffer (NEB #P0710S) and incubated for 15 minutes at 70°C to denature proteins. After cooling to room temperature, 1 µL of Rapid PNGaseF (NEB #P0710S) was added and incubated at 50°C for 1 hour to cleave N-glycans from proteins.

Inductively Coupled Plasma-Mass Spectrometry (ICP-MS) Trace Element Analysis
PNGaseF treated samples were resuspended in 100 µl of 5% acetic acid and added to a C18 Sep-Pak (50 mg) column (Waters, #WAT054955) preconditioned with one column volume each of methanol, 5% acetic acid, 1-propanol, and 5% acetic acid. The reaction tube was washed with another 100 µL of 5% acetic acid and added to the C-18 column, followed by 1 mL of 5% acetic acid, and the entire flow-through was collected in a microcentrifuge tube (~1.2 mL). Samples were placed in a speed vacuum for two hours to reduce the volume to ~300 µL, covered in parafilm and lyophilized overnight.
N-glycan permethylation was performed using a fresh slurry of NaOH/DMSO daily. Seven pellets of NaOH (Sigma-Aldrich, #S8045) were dissolved in four glass pipettes volumes (~3 ml) of DMSO (Sigma-Aldrich, #D8418) and ground using a clean/dry mortar and pestle. 200 µL of the NaOH/DMSO slurry was added to the lyophilized N-glycans in addition to 100µL iodomethane (Sigma-Aldrich, #289566) and placed in on a vortex shaker for 20 minutes at room temperature with a microtube cap to prevent the lid from opening due to increased gas pressure. After the mixture became white, semi-solid and chalky, 200 µL ddH2O was added to stop the reaction and dissolve the sample. 200 µL chloroform and an additional 400 µL ddH2O were added for chloroform extraction and vortexed followed by brief centrifugation. The aqueous phase was discarded, and the chloroform fraction was washed three additional times with 800 µL ddH2O.
Chloroform was then evaporated by 20 minutes in a speed vacuum. Permethylated N-glycans were resuspended in 200 µL of 50% methanol and added to a C18 Sep-Pak (50 mg) column preconditioned with one column volume each of methanol, ddH2O, acetonitrile, and ddH2O. The reaction tube was washed with 1mL 10% acetonitrile and added to the column, followed by an additional 2 ml wash of 10% acetonitrile. Columns were placed in a 15 mL glass tube, and permethylated N-glycans were eluted with 3 mL 50% acetonitrile. The eluted fraction was placed in a speed vacuum for one hour to remove the acetonitrile, covered in parafilm and lyophilized overnight.
MALDI-TOF analysis of purified glycans was performed on permethylated N-glycans resuspended in 25 µL of 75% methanol and spotted in a 1:1 ratio with DHB matrix on a metal 384 spot. Spectra from the samples were obtained in a Bruker MALDI-TOF instrument using FlexControl Software in the Reflection Positive mode with a mass/charge (m/z) range of 1,500-5,000 kD. Twenty independent captures (representing 1,000 shots each) were obtained from each sample and averaged to create the final spectra file and exported in .msd format for analysis.
N-Glycan analysis. 57 N-glycans of known structure corresponding to the correct isotopic mass were annotated in each spectra using mMass software (72). The relative abundance of each Nglycan was calculated as the signal intensity for each peak divided by the signal intensity for all 57 measured N-glycans within a spectrum. N-glycans were grouped into different categories based on shared components such as monosaccharide composition, antennarity, or class based on deductive reasoning and prior MS/MS data where available(36) (Supp. Table 4). Absolute change and relative change compared either to CC genotype or pre-Mn supplementation is shown. Heat maps are scaled from dark blue ® white ® bright red representing (-5.0 ® 0 ® +5.0) for absolute abundance change and (-50.0% ® 0 ® +50.0%) for relative change. The contribution of each monosaccharide was determined by taking the percentage of each monosaccharide in a N-glycan multiplied by the abundance of the glycan, and then summated for the five monosaccharides present in human plasma N-glycans.
Statistical Analysis. Brain MRI data of T2w/T1w ratios from TT and CT carriers were compared to CC subjects on a pixel by pixel basis using a t-test corrected for a false discovery rate of 5% using AFNI Image Analysis Tools (https://afni.nimh.nih.gov) and StatistiXL Version 2 Software.
Regions of interest were compared as ratios of T2w/T1w signal intensities using a one-way ANOVA followed by a post-hoc comparison using a Dunnett's test, with the CC as a control (or TT for comparisons between TT and CT). Metal data was analyzed using GraphPad Prism Version 7 and included an initial ANOVA analysis (degrees of freedom, DF = 2) followed by individual unpaired t-tests assuming unequal variance between each genotype (CC vs CT, CC vs TT, CT vs TT), followed by linear regression between Mn concentration, age, and BMI.
Glycosylation data was analyzed using Microsoft Excel Version 16.27. The abundance of individual glycans and glycan classes were compared between genotypes using unpaired t-tests assuming unequal variance between genotypes, with significance thresholds applied at p *<0.05, **<0.01, and ***<0.001.