Genetic Variants Affect Distinct Metabolic Pathways in Pediatric Multisystem Inflammatory Syndrome and Severe COVID-19

Abstract

The coronavirus disease 2019 (COVID-19) pandemic has triggered a global health crisis, with over 700 million confirmed cases and at least 7 million deaths reported by early 2024. Children are less vulnerable to severe SARS-CoV-2 infection than adults and typically experience milder respiratory symptoms. However, a rare but significant complication, known as multisystem inflammatory syndrome in children (MIS-C), can develop weeks after infection, characterized by a spectrum of inflammatory symptoms. This study employed whole-exome sequencing and over-representation analysis to identify genetic variants of potential clinical significance related to MIS-C or severe COVID-19 in a group of children with acute respiratory distress syndrome (ARDS), all of whom were unvaccinated for COVID-19. We observed the enrichment of potentially pathogenic genetic variants in genes related to carbohydrate metabolism, particularly glycogen breakdown, in severe COVID-19 pediatric patients, and in genes related to cholesterol and lipoprotein metabolism in MIS-C patients. These findings offer insights into the genetic underpinnings of MIS-C and severe COVID-19, suggesting potential genes and biological pathways for further research.

Keywords: COVID‐19; MIS‐C (multisystem inflammatory syndrome in children); carbohydrate metabolism; cholesterol metabolism; genetic variants; whole‐exome sequencing.

Figure 1

Study design overview. (a) Whole‐exome sequencing was performed on 44 pediatric patients infected with SARS‐CoV‐2 – 15 diagnosed with severe COVID‐19 (sCOVID‐19) and 29 with Multisystem Inflammatory Syndrome in Children (MIS‐C) – yielding 766 filtered variants across 674 candidate genes. (b) Over‐representation analysis (ORA) using Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), and Reactome databases was conducted on sCOVID‐19‐specific, MIS‐C‐specific, and shared candidate genes, resulting in 115 enriched biological terms (q < 0.01 for GO/KEGG, q < 0.02 for Reactome). (c) From 382 ORA‐enriched genes, 172 were prioritized using logistic regression with an L2 penalty (coefficient ≥ |0.13 |). Hierarchical clustering based on Euclidean distance was then applied to these prioritized genes to group patient samples. (d) Of the 115 enriched terms identified, 23 were found to be most associated with sCOVID‐19 or MIS‐C after applying logistic regression with an L1 penalty, highlighting key pathways that distinguish the two conditions.

Figure 2

Hierarchical clustering of patient samples. Hierarchical clustering was conducted using the per‐gene variant counts for each of the 44 pediatric patients (sCOVID‐19 in purple, MIS‐C in green). Euclidean distances were computed and Ward′s method was applied, revealing two principal clusters: Cluster 1 (green) predominantly includes sCOVID‐19 samples, whereas Cluster 2 (orange) is mainly composed of MIS‐C cases. This clustering indicates distinct genetic variant profiles between the two clinical groups.

Figure 3

Terms prioritized by logistic regression and over‐representation analysis. (a) Logistic regression with an L1 penalty was applied to 115 enriched terms to identify those most closely associated with either severe COVID‐19 (sCOVID‐19) or MIS‐C. Negative regression coefficients indicate terms more likely linked to sCOVID‐19, whereas positive coefficients indicate those preferentially associated with MIS‐C. (b) Dot plot illustrating the key terms (coefficient > |0.5| in panel a) and closely related enriched terms from the over‐representation analysis (ORA). Terms were derived from GO, KEGG, and Reactome databases, each meeting the respective significance thresholds (q < 0.01 for GO/KEGG, q < 0.02 for Reactome). Color coding denotes broad functional categories, including cholesterol and lipoprotein metabolism, digestive system, motor protein activity, muscle cell development, cardiomyopathies, carbohydrate metabolism, and serine‐type peptidase activity.

Figure 4

Genes affected in sCOVID‐19‐enriched terms. Heatmaps display the presence of genetic variants in each patient sample (columns) across key genes (rows) within the sCOVID‐19 group (purple) and MIS‐C group (green). The intensity of red indicates the number of variants detected in each gene‐sample pair. Genes are organized into three main categories enriched in sCOVID‐19: (a) Carbohydrate metabolism (Starch and sucrose metabolism [KEGG], Glycogen breakdown [glycogenolysis] [Reactome], and Galactose metabolism [KEGG]). (b) Serine‐type peptidase activity (GO). (c) Cardiomyopathy (Hypertrophic cardiomyopathy [KEGG], Dilated cardiomyopathy [KEGG], and Arrhythmogenic right ventricular cardiomyopathy [KEGG]).

Figure 5

Genes affected in MIS‐C‐enriched terms. Heatmaps display the presence of genetic variants in each patient sample (columns) across key genes (rows) within the sCOVID‐19 group (purple) and MIS‐C group (green). The intensity of red indicates the number of variants detected in each gene‐sample pair. Genes are grouped into categories enriched in MIS‐C. (a) Cholesterol metabolism (Cholesterol transport [GO], Cholesterol metabolism [KEGG], Cholesterol homeostasis [GO]). (b) Lipoprotein metabolism (Plasma lipoprotein clearance [Reactome], Plasma lipoprotein assembly, remodeling, and clearance [Reactome], Lipoprotein particle receptor binding [GO], Lipoprotein particle binding [GO]). (c) Motor proteins (Motor proteins [KEGG], Cytoskeletal motor activity [GO]). (d) Digestive system process (GO). (e) Muscle cell development (GO).

Figure 6

Carbohydrate metabolism and glycogen breakdown pathways. Adapted from the KEGG Starch and Sucrose Metabolism pathway, this diagram highlights key enzymes involved in carbohydrate metabolism as identified through over‐representation analysis (ORA). Genes harboring variants in sCOVID‐19 are shown in purple, whereas those in MIS‐C appear in green. Light purple outlines indicate components of the KEGG glycogen degradation pathways (N00718 and N00720), underscoring the prominent role of glycogenolysis in the pathophysiology of severe COVID‐19 and MIS‐C.

Figure 7

Cholesterol metabolism pathway. Adapted from the KEGG Cholesterol Metabolism pathway, this diagram highlights genes within the Cholesterol Metabolism supercategory identified by over‐representation analysis (ORA). Genes harboring variants in sCOVID‐19 are shown in purple, while those in MIS‐C appear in green. The pathway illustrates the interplay of lipoproteins – HDL, LDL, VLDL, and chylomicrons – in cholesterol transport, uptake, and metabolism, emphasizing processes that may be disrupted by potentially pathogenic genetic variants.

Figure 8

Structural consequences of variants in carbohydrate metabolism genes. In silico analyses were conducted to model the effects of missense variants on protein conformation and function, focusing on key residues and potential disruptions in active sites or binding pockets: (a–c) GALT and NP_000146.2:p. Gln188Arg (rs75391579): (a) GALT structure. (b) Active site close‐up: Gln188 bonding with Trp190, G1P, and H2U. (c) Variant model: p. Gln188Arg disrupts Trp190 and ligand interactions, likely reducing catalytic activity. (d–f) PYGM and NP_005600.1:p. Ala365Val (rs116135678): (d) PYGM structure with AMP and glucose. (e) Close‐up of Ala365 (pink). (f) Variant model: p.Ala365Val causes steric clashes with Met351, Met360, and Leu354.

Figure 9

Structural consequences of variants in cholesterol metabolism genes. In silico analyses were conducted to model the effects of missense variants on protein conformation and function, focusing on key residues and potential disruptions in active sites or binding pockets. (a–g) CFTR chloride channel and NP_000483.3:p.[Arg1162Leu; Ile285Phe; Leu206Trp]: (a) CFTR with predicted membrane orientation. (b) Close‐up of Arg1162, which forms hydrogen bonds with Asp979 and Glu1046. (c) Variant model: p. Arg1162Leu (rs1800120) disrupts these hydrogen bonds. (d) Close‐up of Ile285. (e) Variant model: p. Ile285Phe (rs151073129) introduces steric bulk, potentially clashing with Thr262. (f) Close‐up of Leu206 near cholesterol (CLR). (g) Variant model: p. Leu206Trp (rs121908752) may perturb protein‐solvent and membrane interactions. All models depict extracellular surfaces in red and intracellular surfaces in blue. (h–j) ABCB11 and NP_003733.2:p.Arg517His: (h) Overall structure of the bile salt export pump ABCB11. (i) Detailed view of Arg517, which forms hydrogen bonds with Glu514 and Glu521 within the nucleotide‐binding region. (j) Variant model: p.Arg517His disrupts both stabilizing hydrogen bond networks. Electrostatic potential surfaces color scheme demonstrates the transitions from neutral to negative charge distribution upon aminoacidic substitution: electronegative regions in red, electropositive regions in blue, and neutral areas in white.