Extensive hypoglycosylation of serum N-glycoproteins in SRD5A3 deficiency

Abstract

Polyprenal reductase is an enzyme encoded by the SRD5A3 gene, which is involved in the synthesis of dolichol from polyprenol. Dolichol serves as a carrier for glycan precursors or monosaccharides in N-linked glycosylation. Pathogenic variants in SRD5A3 can result in a congenital disorder of glycosylation (CDG), SRD5A3-CDG, which is inherited in an autosomal recessive manner. Most plasma proteins are glycosylated and changes in the glycosylation of several glycoproteins are associated with pathological consequences. Despite the critical role of SRD5A3 in glycosylation, the impact of its deficiency on the glycosylation of serum proteins remains largely unexplored. In this study, we used tandem mass tag-based multiplexed quantitative approach to analyze serum N-glycoproteomics and proteomics in SRD5A3-CDG patients and controls. We quantified 2,200 serum N-glycopeptides from 359 N-glycosites from 204 serum proteins. Extensive hypoglycosylation of serum proteins was observed in patients, with 245 of 291 altered glycopeptides decreased in SRD5A3-CDG. Altered glycopeptides included those derived from haptoglobin, plasma serine protease inhibitor, alpha-1-B glycoprotein, alpha-2-macroglobulin and ceruloplasmin. Some of these proteins have previously been reported to be associated with liver dysfunction, anemia and coagulopathy, which could underlie similar clinical features observed in SRD5A3-CDG patients. Overall, our study provides novel insights into alterations in the glycosylation status of specific serum proteins in SRD5A3-CDG. Some of these alterations could be further pursued to develop glycopeptide-based biomarkers as current diagnosis of SRD5A3-CDG by screening assays remains challenging. In addition, knowledge of altered glycoproteins could enhance our understanding of the disease spectrum and potentially unveil additional therapeutic avenues.

Take-home message

N-glycoproteomics analysis reveals site-specific alterations in N-glycosylation of circulating proteins in SRD5A3-CDG patient serum samples.

Introduction

Glycosylation is the most abundant post-translational modification present on proteins. It is a multistep process in which monosaccharides are covalently attached to the asparagine, serine, threonine, or tryptophan residues of proteins (1, 2). In addition to facilitating structural stability and folding of proteins, glycosylation is also implicated in a multitude of biological functions such as cell adhesion, cell signaling, protein secretion, transport, immune response etc. (3–8). The process of glycosylation involves over 200 genes and pathogenic variants in many of these genes can cause genetic disorders called congenital disorders of glycosylation (CDG) (1, 9, 10).

CDG are a group of rare genetic disorders with ~200 CDG identified thus far (11). The inheritance pattern is mainly autosomal recessive, but autosomal dominant and X-linked inheritance are also known to occur (12–15). SRD5A3-CDG is an autosomal recessive CDG caused by biallelic pathogenic variants in the SRD5A3 gene, which codes for the enzyme polyprenal reductase (SRD5A3). This enzyme is involved in the conversion of polyprenol to dolichol occurring in the membrane of endoplasmic reticulum (ER) (16). Dolichol acts as a carrier for N-glycan precursors as well as a donor of monosaccharides and its deficiency affects protein N-linked glycosylation (Figure 1A) (17, 18). The first confirmed case of SRD5A3-CDG was reported in 2010 (19, 20) and nearly 50 patients have been reported in the literature so far (19–31). Severe visual impairment and cerebellar ataxia are the primary characteristics of this disorder. Other symptoms include developmental delay, hepatopathy, anemia, recurrent infection and ichthyosiform dermatitis. (19–23, 25–27). Current management of the disease is focused on symptomatic management.

Figure 1. Investigation of changes in serum proteome and glycoproteome in SRD5A3-CDG.

(A) Overview of protein glycosylation and role of steroid 5α-reductase type 3 (SRD5A3) in glycosylation. (B) Experimental schematic for TMT labeling and relative quantitation of peptides and glycopeptides from serum of SRD5A3-CDG patients and controls.

Suspected cases of SRD5A3-CDG, which usually present with early onset of disease symptoms, are screened by serum carbohydrate-deficient transferrin analysis (19). In serum, the most abundant transferrin isoform normally is known to have two bi-antennary N-glycans at Asn-432 and Asn-630. This is called the di-oligo fraction of transferrin. Increased levels of transferrin isoforms with one (mono-oligo) or both the sites (a-oligo) without any glycan is observed to be associated with CDG affecting the early steps of the N-glycosylation pathway. Screening of CDG is carried out by monitoring the ratios of mono-oligo/di-oligo and/or a-oligo/di-oligo transferrin isoforms. These ratios are typically elevated in SRD5A3-CDG, although cases with normal transferrin patterns have also been reported (21, 24, 25, 32, 33). Definitive diagnosis of the disease is performed by molecular genetic testing although the presence of novel variants can still complicate interpretation.

Although SRD5A3 plays a vital role in glycosylation, the effect of its deficiency on abundance and glycosylation of the serum proteome has not been evaluated. In this study, we used high-resolution mass spectrometry to analyze the proteomics and glycoproteomics profiles of serum proteins from individuals with SRD5A3-CDG carrying different biallelic pathogenic variants in SRD5A3. Our detailed analysis revealed an overall decrease in the glycosylation of serum proteins. This study offers a molecular understanding of the alterations in the serum glycoproteome caused by SRD5A3 deficiency. It also provides insights for future research into the biological implications and dysregulation of glycosylation in SRD5A3-CDG.

Methods

Samples

Serum samples used in this study were deidentified residual samples from SRD5A3-CDG patients and age- and sex-matched volunteer donors, used as controls (approved by Mayo Clinic IRB: 21-012890). To evaluate the impact of SRD5A3 deficiency on serum glycoproteins, we examined serum samples from five patients clinically diagnosed with SRD5A3-CDG. Among these, one patient had a homozygous frameshift pathogenic variant leading to a premature termination, c.66delG (p.Thr23ProfsTer6). Two patients, who were siblings, shared the pathogenic variant, c.617G>T (p.Gly206Val) on one allele and c.645_670del26 (p.His216ArgfsTer7) on the other. Patient 4 had a heterozygous pathogenic variant with a pathogenic splice variant 562+3delG and c.869T>C (p.Leu290Pro) (identified as patient 4). The fifth patient had a homozygous pathogenic variant c.497G>A (p.Gly166Glu, corresponding to patient 5), classified as a variant of uncertain significance (VUS). Table 1 shows the demographic and genetic information of the patients included in the study along with their main clinical features.

Table 1. SRD5A3-CDG patients and their genetic variants.

Patient	Age	Sex	Genetic variants in SRD5A3, c.		Clinical features
			Allele 1	Allele 2
Patient 1	2 Months	F	66delG	66delG	Motor developmental delay, hypotonia, feeding difficulties, gastroesophageal reflux, failure to thrive, nystagmus, decreased reaction to light and eczema
Patient 2	14 Years	F	617G>T	645_670del26	Cataracts, nystagmus, optic atrophy and right aborted coloboma, hypotonia, global developmental delay, intellectual disability, autism spectrum disorder, psoriasis, ichthyosis, palmar and plantar keratosis, tremor, constipation, history of early breast development, cerebellar hypoplasia and ataxia
Patient 3	20 Years	M	617G>T	645_670del26	Gastroesophageal reflux disease, optic nerve hypoplasia, nystagmus, myopia, and poor vision, oculomotor apraxia, global developmental delay and intellectual disability, hypotonia, intention tremor, kyphosis and ataxia
Patient 4	19 Years	F	869T>C	562+3delG	Developmental delay and speech delay, intellectual disability, autism spectrum disorders, cerebellar hypoplasia, ataxia and decreased visual acuity due to bilateral optic nerve hypoplasia
Patient 5	4 Years	M	497G>A	497G>A	Global developmental delay, absent speech, visual impairment with bilateral optic nerve hypoplasia, hypotonia, ataxic gait and dysmorphic features

Sample preparation for glycoproteomics and total proteomics

Total protein from serum samples (controls and SRD5A3-CDG patients) was estimated using bicinchoninic acid protein estimation assay. Equal amount of protein from each sample was subjected to reduction and alkylation using 5 mM dithiothreitol and 20 mM iodoacetamide, respectively. Protein was digested overnight at 37°C using trypsin at a ratio of 1:20 (trypsin: protein). Subsequently, peptides were desalted using C₁₈ tips. The dried peptides were reconstituted in 50 mM triethyl ammonium bicarbonate and labeled with tandem mass tag (TMT) reagents using a 16-plex TMT pro kit. The labeled peptides were pooled and split into two aliquots for glycopeptide enrichment using size-exclusion chromatography and fractionation using basic reversed phase liquid chromatography followed by analysis by mass spectrometry.

Size-exclusion chromatography for glycopeptide enrichment

An aliquot of TMT-labeled total peptides was subjected to size exclusion-chromatography for glycopeptide enrichment as described earlier (34–36). Briefly, dried peptides were reconstituted in 0.1% formic acid followed by loading and separation on a Superdex peptide 10/300 column (GE Healthcare). Isocratic flow of 0.1% formic acid was maintained at a flow rate of 200 μl per minute. Size-exclusion chromatography was performed for a total run time of 130 minutes with fractionation and collection of the eluate into total 48 fractions. Twenty-four early fractions (based on UV profiles) were collected and pooled into 12 fractions.

Basic pH reversed-phase liquid chromatography

The other aliquot of TMT-labeled dried peptides (500 µg) was reconstituted in 5 mM ammonium formate (pH 10) and fractionated by basic pH reversed-phase liquid chromatography into 96 fractions on a C₁₈ column (3.5 µm, 4.6 mm × 250 mm). The total run time was 120 min and 5 mM ammonium formate (pH 10) in water was used as solvent A and 5 mM ammonium formate (pH 10) in 80% acetonitrile was used as solvent B. The gradient used for separation is as follows: 2% of solvent B was used for first two min and changed to 2% to 12% from 5 min to 8 min. Solvent B was increased to 55% from 8 min to 100 min, further increased to 70% over the next 10 minutes. The column was then washed with 80% of solvent B for 3 minutes and equilibrated with 5% of solvent B for 4 min. A total of 96 fractions were collected and concatenated to 12 fractions. These fractions were dried, reconstituted in 0.1% formic acid and analyzed on an Orbitrap Eclipse Tribrid mass spectrometer as described below.

Liquid chromatography – tandem mass spectrometry

Glycopeptides were analyzed by LC-MS/MS using previously described methods with modifications (34–36). Glycopeptide separation was performed using an Ultimate 3000 (Thermo Fisher Scientific Inc.) liquid chromatography system followed by analysis on an Orbitrap Eclipse mass spectrometer (Thermo Fisher Scientific Inc.). Glycopeptides were first trapped on a trap column (100 mm × 2 cm, Acclaim PepMap100 Nano-Trap, Thermo Fisher Scientific Inc.) at a flow rate of 5 µl/min and separated on an EASY-Spray analytical column (75 μm x 50 cm, PepMap RSCL C₁₈, Thermo Fisher Scientific Inc.) packed with 2 μm C₁₈ particles with temperature maintained at 50°C. Separation was performed using 0.1% formic acid in water as solvent A and 0.1% formic acid in acetonitrile as solvent B. Flow rate was maintained at 300 nl/min for a gradient time of 150 minutes as follows: column equilibration was done for 4 minutes at 3% of solvent B, a gradient of 3 to 35% of solvent B from 4 to 130 minutes, then 35 to 80% of solvent B from 130 to 135 minutes followed by 80% of solvent B for 10 minutes and finally equilibration for 5 minutes at 3% of solvent B.

Acquisition was done in data-dependent mode with the following parameters. Precursor ions were detected in the Orbitrap at a resolution of 120,000 with a scan range of 375 m/z to 2000 m/z with normalized automatic gain control target of 100% and maximum ion injection time of 50 ms. Precursor ions were isolated at a window of 0.7 m/z with a charge state ranging from +2 to +7 for MS/MS events. Fragmentation was done using normalized stepped higher-energy collisional dissociation at 15, 25 and 40%. Normalized AGC was set to 400% with a maximum ion injection time of 200 ms. Data acquisition was performed in centroid mode with option of lock mass (441.1200025 m/z) “on”. Total proteomics data was analyzed on the same instrument using the same parameters but with the normalized collision energy fixed at 35% for MS/MS fragmentation.

Database searching and data analysis

The proteomics data was searched in Proteome Discoverer 3.0, using the SEQUEST search engine, against the UniProt human reviewed protein sequences (Release 2021_02). Reporter ion quantification node was used for relative quantification of proteins between cases and controls. Enzyme specificity was set to trypsin with 2 allowed missed cleavages and carbamidomethylation of cysteine and TMTpro 16-plex modification of the peptide N-terminal set as fixed modifications and protein N-terminal acetylation, oxidation of methionine and TMTpro 16-plex modification of lysine set as variable modifications. Precursor mass tolerance was set to 10 ppm and fragment mass tolerance was set to 0.02 Da. The glycoproteomics data was searched using the publicly available pGlyco (v 3.0) software (37). Data was searched against UniProt human reviewed protein sequences and in-built N-glycan human database. For identification of potential N-X-V glycosylation sites, all the N-X-V sequences were manually converted to J-X-V, in the protein database, for consideration as potential modification site by pGlyco. The conversion of all N-X-T/S/C motifs present to J-X-T/S/C is one of the processing steps of pGlyco during the search. Enzyme specificity and modification parameters for glycoproteomics were used as described for total proteomics search. Precursor and fragment mass tolerances were set to 10 ppm and 20 ppm respectively and results were filtered for glycopeptides with 1% false discovery rate (FDR). Reporter ion intensity values were obtained by searching the data in Proteome Discoverer 3.0 using reporter ion quantifier node, as described for total proteomics, and were matched on a scan-to-scan basis with the glycopeptide IDs obtained from pGlyco.

Statistical analysis

Fold-change values for glycopeptides and proteins were calculated as average TMT intensity values in SRD5A3-CDG patients over controls. Unpaired two-tailed t-test was used to calculate the p-value. FDR adjusted p-value (q-value) was calculated by Benjamini–Hochberg procedure using the package rstatix in R (38). A q-value less than 0.05 was considered significant. MetaboAnalyst (v5.0) was used to generate PCA plots and heatmaps (39). Chord diagram and dot plots were generated in R using the packages circlize and ggplot2, respectively (40, 41).

Results

In this study, we performed multiplexed relative quantitation of serum proteins and glycopeptides in five individuals with SRD5A3-CDG relative to seven controls. Clinical and demographic characteristics of the included patients are provided in Table 1. All patients shared the classic cerebello-ocular syndrome presentation with central nervous system symptoms including developmental delay or intellectual disability, signs of ataxia, and visual impairment (20, 42). Three patients also had autism. In addition, abnormal optic nerve development and function were also shared features in three cases. Specific multi-organ features were only found in one patient with severe ichthyosis and endocrine abnormalities (Patient 2).

Proteins from serum samples were digested and labeled with tandem mass tag (TMT) reagents. Glycopeptides from these TMT-labeled peptides were enriched using size-exclusion chromatography and analyzed by mass spectrometry. In parallel experiments, the TMT-labeled total serum protein digest of the samples was also fractionated by basic reversed-phase liquid chromatography for relative quantitation by mass spectrometry. Figure 1B illustrates the experimental approach for proteomics and glycoproteomics analysis of these samples.

We detected 2,200 unique N-glycopeptides. These glycopeptides had 238 distinct N-glycan compositions that were located on 359 glycosylation sites derived from 204 glycoproteins. Notably, >93% of the identified glycopeptides were characterized by complex/hybrid types, with 42% being sialylated, 12% fucosylated, and 29% displaying both sialylation and fucosylation (Figure 2A). Only 5% of the glycopeptides were linked to high mannose glycans, and 2% were associated with paucimannose glycans.

Figure 2. Glycoproteomic alterations in SRD5A3-CDG.

(A) A donut chart representing the percentage of glycopeptides identified from serum samples, based on the categories of glycan compositions. (B) Principal component analysis plot for TMT-based glycoproteomics data. (C) Volcano plot showing global glycosylation changes in serum of SRD5A3-CDG (n=5) in comparison with controls (n=7); where the detected glycopeptide backbones contain multiple known glycosylation sites, the non-glycosylated sites are marked in blue; putative structures are shown using Symbol Nomenclature for Glycans (SNFG) and represent glycan composition inferred from mass spectrometry data (68). (D) Chord diagram showing the site-specific alterations in the glycosylation of serum proteins. Each connecting chord representing a glycopeptide from a protein glycosylation site shown on one side (represented by gene symbol and site of glycosylation in protein), and the composition of the glycan on the other side. The color of the chord represents the log₂-transformed fold-change (average of SRD5A3-CDG, n=5/average of controls, n=7). Glycopeptides with q values <0.05 are represented. H=hexose, N=N-acetylglucosamine, A=N-acetylneuraminic acid, F=fucose. (E) Heatmap showing the 50 most significantly altered (q<0.05) glycopeptides in SRD5A3-CDG; glycopeptides are represented by non-italicized gene symbol followed by the amino acid site of glycosylation and glycan composition; Hex=hexose, HexNAc=N-acetylglucosamine, NeuAc=N-acetylneuraminic acid, Fuc=fucose.

Among the 359 glycosylation sites identified, 276 were occupied by only complex/hybrid glycans, 22 by only high-mannose or paucimannose glycans, and 61 by both complex/hybrid and high-mannose/paucimannose glycans as seen in different glycopeptides. For instance, Asn²⁰⁷ in haptoglobin was found to be occupied by both paucimannose and complex/hybrid glycans while Asn²¹¹ exhibited only complex/hybrid glycans. Another example is complement factor C2 in which Asn¹¹² was associated with complex/hybrid glycans, Asn⁴⁶⁷ with high-mannose glycans and Asn⁶²¹ with both complex/hybrid and high-mannose glycans.

Regarding the distribution of glycopeptides among proteins, the highest number was observed in immunoglobulins, collectively contributing to ~17% of the total identified serum N-glycopeptides. The second-largest glycopeptide repertoire was observed in the case of haptoglobin, which accounted for 6.5% of the total identified glycopeptides with a total of 142 N-glycopeptides. This was followed by ceruloplasmin, contributing to 3.5% of the total identified glycopeptides. Supplementary Figure S1 illustrates the glycopeptide contributions of all identified serum proteins in our dataset.

SRD5A3-deficient cases exhibit altered glycosylation profiles

For relative quantitation of serum glycopeptides, fold-changes were calculated by using average TMT intensity values of patients over controls. A q-value <0.05 was considered significant. The serum glycopeptide profiles of SRD5A3-CDG cases and controls were separated into two distinct clusters by principal component analysis (Figure 2B). A list of all identified glycopeptides in this experiment is provided in Supplementary Table 1. Out of the total identified glycopeptides, 291 glycopeptides were observed to be significantly altered in SRD5A3-CDG with q<0.05. These include glycopeptides from 69 serum glycoproteins representing 105 unique glycosites. Among these dysregulated glycopeptides, 46 were observed to be increased and 245 were decreased (Figure 2C and 2D). The expression pattern of the top 25 up- or downregulated glycopeptides in individual samples is represented in the heatmap shown in Figure 2E.

A major decrease was observed in glycopeptides occupied by sialylated complex/hybrid glycans, comprising 174 glycopeptides derived from 87 unique N-glycosylation sites of 62 glycoproteins. Among these, 51 changing glycopeptides had both sialylated and fucosylated glycans, while only 15 were occupied with fucosylated glycans. Additionally, only 17 changing glycopeptides were occupied with high mannose/paucimannose.

A few serum glycopeptides were increased in SRD5A3-CDG

Amongst the 291 dysregulated glycopeptides, 37 were found to be significantly increased with ≥2 fold-change and a q-value <0.05. These 37 glycopeptides represented 7 unique sites from 6 different glycoproteins. The glycopeptide bearing three mannose residues (Hex₃HexNAc₂) on Asn⁸⁶⁹ of alpha-2-macroglobulin was one of the increased glycopeptides with 29-fold increase (q-value 0.02) in SRD5A3-CDG (Figure 3A). We also observed glycopeptide at Asn²²⁶ of complement C4 glycoprotein. This glycopeptide was occupied with a high-mannose glycan with the Hex₅HexNAc₂ composition and was increased four times in SRD5A3-CDG compared to the controls (q-value 0.005) (Figure 3B). In a recent study, we have reported this glycopeptide as a biomarker for PMM2-CDG which is another Type-I CDG (43).

Figure 3. Site-specific increase in the abundance of glycopeptides in SRD5A3-CDG.

Dot plots representing levels of significantly upregulated glycopeptides; putative structures are shown using SNFG nomenclature and represent glycan composition inferred from mass spectrometry data (68); *=q<0.05, **=q<0.01, ***=q<0.001. Glycopeptides identified from (A) Asn⁸⁶⁹ of alpha-2-macroglobin, (B) Asn²²⁶ of complement C4 and (C) Asn⁸⁵ of complement C3.

Four different glycoforms of the glycopeptide at Asn⁸⁵ of glycoprotein complement C3 were detected with upregulation in SRD5A3-CDG. Two of these glycoforms were occupied with sialic acid-containing complex glycans with the following glycan compositions: Hex₄HexNAc₃NeuAc₁ (fold-change 4.7; q-value 0.04) and Hex₅HexNAc₄NeuAc₂ (fold-change 3.3; q-value 0.01). The other two were high mannose glycoforms with the composition Hex₄HexNAc₂ (fold-change 2.6; q-value 0.03) and Hex₃HexNAc₂ (fold-change 2.5; q-value 0.01). Abundance level of these glycopeptides from complement C3 is represented in figure 3C.

Hypoglycosylation in SRD5A3-CDG

Interestingly, among the 37 increased glycopeptides, 29 were haptoglobin-derived glycopeptides representing two known N-glycosylation sites of haptoglobin i.e., Asn²⁰⁷ or Asn²¹¹. These two N-glycosylation sites of haptoglobin are located within the same tryptic peptide with the sequence NLFLN²⁰⁷HSEN²¹¹ATAK. We observed that glycopeptides in which only one of these sites was glycosylated were increased in SRD5A3-CDG while glycopeptides in which both sites were glycosylated were decreased. This suggests site-specific hypoglycosylation in haptoglobin. Figure 4 represents the relative abundance of these glycopeptides in patient and control groups including those when only Asn²⁰⁷ is glycosylated (Figure 4A), when only Asn²¹¹ is glycosylated (Figure 4B) and when both Asn²⁰⁷ and Asn²¹¹ were glycosylated (Figure 4C). The glycopeptide occupied with Hex₄HexNAc₃NeuAc₁ at Asn²¹¹ of haptoglobin showed the greatest increase in SRD5A3-CDG. It was increased 43-fold in SRD5A3-CDG (q-value 0.02) (Figure 4B). We also observed glycopeptides from another known N-glycosylation site of haptoglobin - Asn²⁴¹. These glycopeptides were observed to be decreased in SRD5A3-CDG (Figure 4D). The glycoform with glycan Hex₆HexNAc₄NeuAc₁ at Asn²⁴¹ of haptoglobin was among the ten most decreased glycopeptides (fold-change 0.24, q-value 0.03) (Figure 4D). This site-specific hypoglycosylation pattern of haptoglobin has also been observed in serum samples of another type-I CDG that we have analyzed, i.e., PMM2-CDG (43).

Figure 4. Site-specific alterations in glycosylation of haptoglobin in SRD5A3-CDG.

Dot plots representing levels of significantly upregulated glycopeptides; putative structures are shown using SNFG nomenclature and represent glycan composition inferred from mass spectrometry data (68); *=q<0.05, **=q<0.01, ***=q<0.001. Glycopeptides identified from haptoglobin with two N-linked glycosylation sites, Asn²⁰⁷ and Asn²¹¹ (A) when only Asn²⁰⁷ was glycosylated, (B) when only Asn²¹¹ was glycosylated, (C) when both Asn²⁰⁷ and Asn²¹¹ were glycosylated and (D) glycopeptides identified for Asn²⁴¹ of haptoglobin.

A total of 153 glycopeptides from 47 serum proteins were decreased in SRD5A3-CDG with a ≤0.5 fold-change and q-value <0.05. Interestingly, the glycopeptides with the most hypoglycosylation were derived from albumin which has long been considered as a non-glycosylated protein. In a recent study we have used multi-pronged approach to confirm the glycosylation of albumin at these sites (44). In the current study, we detected eight glycopeptides from albumin representing two glycosylation sites at Asn⁶⁸ and Asn¹²³. In our data, we observed a glycopeptide from Asn⁶⁸ with the glycan Hex₅HexNAc₄NeuAc₂ as most decreased (fold-change 0.17, q-value 0.003) (Figure 5A). Another glycopeptide from albumin at site Asn¹²³ was also amongst the ten most decreased glycopeptides in our data. This was occupied by the sialylated complex glycan Hex₅HexNAc₄NeuAc₂ (fold-change 0.23, q-value 0.001) (Figure 5B). These findings are in keeping with our findings in other CDG (45).

Figure 5. Site-specific decrease in the abundance of glycopeptides in SRD5A3-CDG.

Dot plots representing levels of significantly downregulated glycopeptides; putative structures are shown using SNFG nomenclature and represent glycan composition inferred from mass spectrometry data (68); *=q<0.05, **=q<0.01, ***=q<0.001. Glycopeptides identified from (A) albumin at Asn⁶⁸ and (B) Asn¹²³. Glycopeptides identified from (C) Asn⁶³ of alpha-1B glycoprotein, (D) Asn³⁵⁸ of ceruloplasmin, (E) Asn²²⁴ of antithrombin-III, (F) Asn²³⁸ of plasma protease C1 inhibitor, (G) Asn¹⁸⁸ of heparin cofactor 2 (H) Asn²³⁸ of kallistatin, (I) Asn¹⁷⁶ of corticosteroid-binding globulin and (J) Asn¹⁰⁶ of alpha-1-antichymotrypsin.

Another glycopeptide showing a decrease is from the protein alpha-1B- glycoprotein at Asn⁶³ occupied with the sialylated biantennary glycan Hex₅HexNAc₄NeuAc₂ (fold-change 0.20, q-value 0.006) (Figure 5C). Supplementary Figure S2 shows the dot plots representing the abundance levels of the other detected glycopeptides from alpha-1B-glycoprotein. Many glycopeptides from Asn³⁵⁸ of ceruloplasmin were also found to be decreased. These include glycopeptides with glycans Hex₄HexNAc₃NeuAc₁ (fold-change 0.32, q-value 0.001), Hex₆HexNAc₄NeuAc₂ (fold-change 0.32, q-value 0.02), Hex₅HexNAc₄NeuAc₂ (fold-change 0.33, q-value 0.002), Hex₅HexNAc₄NeuAc₁ (fold-change 0.34, q-value 0.002) and Hex₆HexNAc₄NeuAc₁Fuc₁ (fold-change 0.37, q-value 0.002). Figure 5D and Supplementary Figure S2 represents the abundance levels of these glycopeptides.

Several glycopeptides from the family of serine protease inhibitors were also found to be decreased. This includes glycopeptides from antithrombin-III, plasma protease C1 inhibitor, heparin cofactor 2, kallistatin, corticosteroid-binding globulin and alpha-1-antichymotrypsin,. Several glycopeptides were observed from three different sites of protein antithrombin-III. Five glycopeptides were observed from Asn²²⁴, three from Asn¹⁸⁷ and two from Asn¹²⁸, all of which were occupied with different complex/hybrid glycans. Amongst these, glycopeptide at Asn²²⁴ with the glycan Hex₄HexNAc₃ showed the most decrease (fold-change 0.34, q-value 0.01) (Figure 5E), followed by glycan Hex₅HexNAc₄NeuAc₂ at Asn¹⁸⁷ (fold-change 0.37, q-value 0.03) (Supplementary Figure S2). Two glycopeptides from plasma protease C1 inhibitor at Asn²³⁸ were observed to be decreased. These were occupied with Hex₆HexNAc₄NeuAc₁ (fold-change 0.35, q-value 0.02) (Figure 5F) and Hex₆HexNAc₅NeuAc₃ (fold-change 0.5, q-value 0.02) (Supplementary Figure S2). Heparin cofactor 2 had one glycopeptide showing downregulation (fold-change 0.5, q-value 0.02) at Asn¹⁸⁸ occupied with Hex₆HexNAc₅NeuAc₃ (Figure 5G). Two glycopeptides from kallistatin at Asn²³⁸ and corticosteroid-binding globulin at Asn¹⁷⁶ were observed occupied with glycan Hex₅HexNAc₄NeuAc₂ with a fold-change of 0.41 (q-value 0.01) and 0.47 (q-value 0.01) respectively (Figure 5H & 5I). Asn¹⁰⁶ and Asn¹²⁷ from alpha-1-antichymotrypsin were observed to be decreased (fold-change 0.37 and fold-change 0.5) and were occupied with Hex₆HexNAc₅NeuAc₃ and Hex₆HexNAc₅NeuAc₂ respectively (Figure 5J and Supplementary Figure S2). Overall, serum glycoproteins in SRD5A3-CDG cases showed significant hypoglycosylation as compared to the controls.

SRD5A3-CDG patients do not exhibit significant alterations in serum protein abundance levels

In the multiplexed proteomics experiment, we identified 686 proteins with 7,260 peptides. Interestingly, no statistically significant changes (q-value <0.05) were observed in protein levels in SRD5A3-CDG patients compared to the control group (Supplementary Figure S3A). Supplementary Figure S3B illustrates the principal component analysis (PCA) of the protein-level abundance data in these samples. On the PCA plot, the two groups (cases and controls) were observed to be mixed without any clear separation suggesting that the serum proteomics profiles are not significantly affected in SRD5A3-CDG. A list of all identified proteins in this experiment is provided in Supplementary Table 2.

Discussion

SRD5A3-CDG, a disorder of dolichol metabolism, affects multiple glycosylation pathways. Considering the key role of SRD5A3 in glycosylation and formation of dolichol being an early step in the multistep N-linked glycosylation pathway, we wished to investigate the effects of deficiency of SRD5A3 on global N-glycosylation of proteins in the serum. Thus, we performed TMT-labeled multiplexed relative quantitation of serum proteins and glycopeptides to analyze the changes in SRD5A3-CDG using previously established methods for protein and glycopeptide analysis with enrichment and deep profiling using LC-MS/MS (34–36). We observed that serum proteins in SRD5A3-CDG exhibit significant hypoglycosylation in a site-specific manner. Notably, we observed that 84% of the altered glycopeptides showed a decrease in SRD5A3-CDG. As the molecular defect is in the early steps of N-linked glycosylation, i.e., synthesis and transfer of a nascent N-glycan to a polypeptide chain (Figure 1A), this disorder is classified as a type I CDG (46). In this group of disorders, the overall glycosylation of proteins is expected to be decreased with an increase in unoccupied N-glycosylation sites (46). The observed overall decrease in N-linked glycopeptides in our data is in keeping with this expectation (Figure 2C).

The glycopeptides showing decrease in SRD5A3-CDG belong to many abundant serum proteins including haptoglobin, ceruloplasmin, antithrombin-III, plasma protease C1 inhibitor, alpha-1-antichymotrypsin, kallistatin and heparin cofactor 2. Changes in the glycoforms of some of these proteins have been reported earlier in case of CDG (47, 48). The role of antithrombin-III and sheparin cofactor 2 in the regulation of coagulation are well known (49). While coagulopathy is not a primary symptom of SRD5A3-CDG, it has been observed in a few patients (21, 22, 27). Ceruloplasmin is a glycoprotein involved in copper transport and iron metabolism (50, 51). It has been reported that low ceruloplasmin levels could lead to iron deficiency despite the high iron storage in the liver (52, 53) which could cause anemia. Anemia has been observed in patients of SRD5A3-CDG (54). Additionally, abnormal glycosylation of haptoglobin has been linked to liver diseases (55, 56) and is another clinical feature observed in SRD5A3-CDG patients (21, 57). Therefore, changes in the glycoforms of these proteins may shed light on associated systemic and symptomatic manifestations in SRD5A3-CDG. Ichthyosis is a clinical symptom which is mostly associated to abnormal lipid metabolism, and characteristically present in disorders of dolichol synthesis including SRD5A3-CDG (21). Abnormal dolichol metabolites have been reported earlier in SRD5A3-CDG (42, 58, 59).

Interestingly, the glycopeptides showing the greatest decrease were from albumin which has recently been described to be glycosylated (44). This is also in agreement with decreased levels of albumin-derived glycopeptides in PMM2-CDG and MPI-CDG (45). Another glycopeptide with limited reports is from the protein alpha-1B-glycoprotein at Asn⁶³. Alpha-1B-glycoprotein is a glycoprotein with four known N-linked glycosylation sites. One potential explanation for the scarcity of reports on the glycosylation of these sites could be that they are part of a non-canonical N-linked motif (N-X-C/V), which is known to exhibit a lower rate of glycosylation compared to the canonical motif (N-X-S/T) (30).

A subset of glycopeptides was increased even though most other glycopeptides were decreased in SRD5A3-CDG. Notably, 80% of the upregulated glycopeptides comprised those from haptoglobin, specifically involving the two closely spaced N-glycosylation sites (Asn²⁰⁷ and Asn²¹¹) within a single tryptic glycopeptide. Haptoglobin, an acute phase protein with four documented N-linked glycosylation sites, with each site reported to exhibit a glycosylation occupancy of >90% (3). Alterations in glycosylation levels of haptoglobin have been previously reported in pathological conditions, including cancer and CDG (60–62). In our dataset, in SRD5A3-CDG, there is an increase in abundance of glycopeptides which contain two N-glycosylation sites (Asn²⁰⁷ and Asn²¹¹) where one of the two sites remained non-glycosylated, indicating a decrease in N-glycosylation site occupancy. These findings corroborate at the glycopeptide-level what is known about hypoglycosylation at the protein-level in SRD5A3-CDG (42). This observation also aligns with findings in other type I CDG such as PMM2-CDG, where the early steps of the N-glycosylation pathway are affected (43). We wonder if this is indicative of skipping of closely spaced glycosylation sites by the N-glycosylation machinery in SRD5A3-CDG (63, 64). A limitation of mass spectrometry-based proteomics and glycoproteomics is the limited detection of lower-abundance plasma proteins due to the wide dynamic range of plasma proteins (65, 66). While the endocrine features of precocious puberty noted in patient 2 in our study could be due to glycosylation abnormalities of endocrine proteins (67), their low abundance in serum may have precluded their detection and correlation between their glycosylation status and endocrine features of Patient 2.

In conclusion, we noted a marked global hypoglycosylation of serum proteins in individuals with SRD5A3-CDG. To the best of our knowledge, this is the first report detailing the global N-linked glycosylation changes in serum proteins in SRD5A3-CDG patients. As there is no definitive treatment available for this disorder, therapeutic options currently remain constrained to symptom management. While various therapies, such as gene therapy or drug repurposing, are currently under investigation, such studies in this direction could enhance our understanding of the disease spectrum and potentially unveil additional therapeutic avenues. Further, the glycopeptide alterations that we describe in this study could be pursued towards biomarker development. Specifically, as previously noted, diagnosis of this CDG remains challenging due to false negatives observed in current screening tests. Glycopeptide-based biomarkers, such as those described here, could aid in more definitive and early diagnosis, thereby shortening patients’ diagnostic odyssey.

Data availability

The mass spectrometry proteomics data have been deposited with the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD053534.