Plasma Protein Biomarkers Distinguish Multisystem Inflammatory Syndrome in Children From Other Pediatric Infectious and Inflammatory Diseases

Background: Multisystem inflammatory syndrome in children (MIS-C) is a rare but serious hyperinflammatory complication following infection with severe acute respiratory syndrome coronavirus 2. The mechanisms underpinning the pathophysiology of MIS-C are poorly understood. Moreover, clinically distinguishing MIS-C from other childhood infectious and inflammatory conditions, such as Kawasaki disease or severe bacterial and viral infections, is challenging due to overlapping clinical and laboratory features. We aimed to determine a set of plasma protein biomarkers that could discriminate MIS-C from those other diseases. Methods: Seven candidate protein biomarkers for MIS-C were selected based on literature and from whole blood RNA sequencing data from patients with MIS-C and other diseases. Plasma concentrations of ARG1, CCL20, CD163, CORIN, CXCL9, PCSK9 and ADAMTS2 were quantified in MIS-C (n = 22), Kawasaki disease (n = 23), definite bacterial (n = 28) and viral (n = 27) disease and healthy controls (n = 8). Logistic regression models were used to determine the discriminatory ability of individual proteins and protein combinations to identify MIS-C and association with severity of illness. Results: Plasma levels of CD163, CXCL9 and PCSK9 were significantly elevated in MIS-C with a combined area under the receiver operating characteristic curve of 85.7% (95% confidence interval: 76.6%–94.8%) for discriminating MIS-C from other childhood diseases. Lower ARG1 and CORIN plasma levels were significantly associated with severe MIS-C cases requiring inotropes, pediatric intensive care unit admission or with shock. Conclusion: Our findings demonstrate the feasibility of a host protein biomarker signature for MIS-C and may provide new insight into its pathophysiology.

][4] MIS-C leads to critical illness in ~70% of affected children and, as of October 2023, the Centers for Disease Control and Prevention have reported 9577 cases meeting the full clinical definitions of this novel disorder. 57][8] Common symptoms of MIS-C include persistent fever, oral mucosal inflammation, conjunctivitis, skin rash, elevated inflammatory markers, gastrointestinal (GI) involvement, cardiac manifestations, multisystem organ dysfunction and shock. 9D is an acute systemic vascular disease of unknown etiology affecting predominately the coronary arteries in infants and children. 10,11The standard treatment for KD patients is high-dose intravenous immunoglobulin (IVIG), which can quickly alleviate symptoms and reduce the incidence of coronary artery aneurysms. 12 there is currently no diagnostic test for KD, the diagnosis heavily relies on clinical signs and symptoms, which include high levels of markers of inflammation and mucosal changes.Approximately 40% of patients with MIS-C will meet the diagnostic criteria for KD and, due to the overlapping clinical and laboratory features, IVIG was recommended as a first-line treatment for MIS-C. 5,13owever, recent evidence has shown that treatment of MIS-C with corticosteroids has the same outcome as IVIG treatment, which is important because corticosteroids are cheap and readily available in low-and middle-income countries. 14,15A diagnostic test to distinguish between these 2 diseases would, therefore, greatly benefit treatment and patient outcome.
Host-derived protein biomarkers have been successfully used to distinguish bacterial from viral infection in children. 1617][18][19][20] From a proteomic perspective, it has also been suggested that there are clinical similarities between MIS-C and secondary causes of hemophagocytic lymphohistiocytosis such as macrophage activation syndrome (MAS) and thrombotic microangiopathy, in addition to its clinical similarities with KD and TSS. 19These studies have singled out potential individual biomarkers of MIS-C, such as cysteine-cysteine motif chemokine ligand 20 (CCL20), C-X-C motif chemokine ligands 9 and 10 (CXCL9, CXCL10), interferon-gamma (IFN-γ), interleukins 6, 7, 8 and 10 (IL-6, IL-7, IL-8 and IL-10), phospholipase A2 group IIA, tumor necrosis factor alpha (TNF-α), nuclear factor kappa B and zonulin.However, the results differ between studies, which may be due to the different comparator groups that were used.
In this study, we aimed to identify a set of proteins that could distinguish MIS-C from other disease groups, including KD, definite bacterial infections (DB) and definite viral infections (DV) by quantifying plasma protein levels of candidate MIS-C biomarkers.Such a protein signature could be used as the basis for a diagnostic test.We also aimed to associate significant changes in individual protein levels with clinical presentation, such association could be further investigated to predict disease severity or inform clinical management.

Ethics Statement
Informed consent was taken from all patients recruited to the study.Recruitment at all locations took place after ethical permissions were in place for that area.In the DIAMONDS study (Diagnosis and Management of Febrile Illness using RNA Personalized Molecular Signature Diagnosis, https://www.diamonds2020.eu) and the PERFORM study (Personalized Risk assessment in Febrile illness to Optimize Real-life Management, https://www.diamonds2020.eu/our-research-history/perform/),the consortium agreed on a finalized protocol and supporting documents that were translated into local languages.Each participating country took responsibility for gaining ethical approval in their region.For DIA-MONDS, the lead site received ethical approval for United Kingdom centers from the London-Dulwich research ethics committee (20/HRA/1714).For the PERFORM study, the lead site received ethical approval for United Kingdom centers from the London-Central research ethics committee (16/LO/1684).The Genetic Determinants of Kawasaki Disease for Susceptibility and Outcome study recruited in the United Kingdom only, and ethical approval was granted by the London-Fulham research ethics committee (13/ LO/0026).During the conduct of the study, clinical data and samples were identified only by anonymized study numbers.

Study Population
This study included 108 pediatric patients (≤16 years old) with clinical and laboratory-confirmed MIS-C (n = 22), KD (n = 23), DB (n = 28), DV (n = 27) and healthy controls (HC, n = 8).A breakdown of the DB and DV can be found in Figure, Supplemental Digital Content 1, http://links.lww.com/INF/F417.The World Health Organization (WHO) defines MIS-C as children 0-19 years of age with fever ≥3 days and elevated inflammatory markers [including C-reactive protein (CRP), procalcitonin and erythrocyte sedimentation rate] who have no other obvious microbial cause of inflammation, including bacterial sepsis, staphylococcal or streptococcal shock syndromes and evidence of SARS-CoV-2 infection or probable exposure. 21They must also exhibit 2 of the most common features of MIS-C including GI symptoms, rash or bilateral conjunctivitis, hypotension, shock or cardiac complications.All MIS-C patients met the WHO criteria, which was the recommended criteria at the time.Plasma samples from all of the other disease groups (KD, DB, DV) were collected before the COVID-19 pandemic (October 2013-January 2019).Plasma samples from KD patients were collected before receiving IVIG treatment in 21 of 23 cases used in this study.3][24] In brief, the DB group included patients in whom an appropriate bacterial pathogen was isolated from a normally sterile site and a diagnosis of DV was conditional upon identification of a virus compatible with the clinical syndrome with no evidence of bacterial infection and CRP ≤60 mg/L.No deaths occurred in this study population.Samples were taken at hospital presentation or admission for all disease groups.

Protein Selection
Proteins were selected either through candidate gene biomarkers identified by RNA sequencing (RNA-Seq) or through existing literature at the time of this study design.
Whole blood transcriptome profiling was performed using RNA-Seq on samples obtained from patients with MIS-C, KD, DB, DV and HC.Following preprocessing and normalization, which is described in detail in Jackson et al, 25 differential expression analysis was performed to identify the genes that were significantly differentially expressed (SDE) between MIS-C and KD, DB and DV.Differential expression analysis was performed using DESeq2 with models including age, sex at birth and RNA-Seq batch, as 2 experimental runs were used for sequencing. 26The following comparisons were made: MIS-C versus KD + DB + DV; MIS-C versus KD; MIS-C versus DB; MIS-C versus DV.P values were adjusted for multiple testing using the Benjamini-Hochberg (BH) adjustment and genes with adjusted P values <0.05 were considered SDE. 27enes of interest were selected based on the criteria that they were significantly upregulated in MIS-C cases as compared to one or more of the other disease groups.Due to the high number of SDE genes, we selected genes with BH-adjusted P values <10 −5 and a log-fold change >2 and those that encoded for extracellularly secreted proteins for which antibodies were commercially available were considered.The final proteins chosen based on the RNA-Seq data were ADAMTS2, ARG1, CD163, CORIN and PCSK9.
Two additional protein targets, CCL20 and CXCL9, were selected based on existing MIS-C literature at the time of this study because of their reported roles in the GI involvement and hyperinflammatory state observed early in the course of MIS-C disease, respectively. 6,17,28

Enzyme-linked Immunosorbent Assay (ELISA)
Plasma concentrations of ADAMTS2 were quantified using a commercial enzyme-linked immunosorbent assay kit, human ADAMTS2 (Abbexa, abx519366, Cambridge) according to the manufacturer's protocol.Samples were randomized, diluted 1:10 in sample diluent buffer and run in duplicate.Absorbance was measured at 450 nm on a SpectraMax microplate reader (Molecular Devices) and a standard curve was created using the SoftMax Pro software (version 5.0, San Jose, CA).Samples that fell below the lower limit of quantification were reanalyzed and values that were still below the lower limit of quantification were extrapolated as half of the lowest standard value.

Statistical Analysis
All statistical analyses were performed using R (version 4.1.1). 29Individual protein results were analyzed using the nonparametric Kruskal-Wallis test, followed by the Mann-Whitney U test to evaluate the differences between each cohort.Principal component (PC) analysis was used to visualize the data.Kruskal-Wallis P values were corrected using the Bonferroni procedure, as were Mann-Whitney U P values.Associations between each of the protein markers and clinical variables were tested, using the Mann-Whitney U test for binary categorical variables [pediatric intensive care unit (PICU) admission: yes or no; oxygen requirement: yes or no; noninvasive ventilation requirement: yes or no; inotrope requirement: yes or no; cardiac involvement: yes or no; GI involvement: yes or no; rash: yes or no; conjunctivitis: yes or no] and linear regression models for continuous variables [body mass index, days of symptoms when sample was taken, CRP levels, white blood cell (WBC) levels, neutrophil levels, lymphocyte levels, monocyte levels and platelet levels].All P values were corrected for multiple testing using the Bonferroni procedure (corrected per clinical/laboratory variable).Clinical and laboratory variables included admission to the PICU, oxygen or ventilation requirement (invasive or noninvasive), inotrope administration, duration of symptoms at the time samples were taken, neutrophils, lymphocytes, platelets, CRP, WBC counts and presentation of symptoms including cardiac and GI involvement, rash, shock and conjunctivitis.
The distinguishing ability of each of the proteins that were significantly differentially abundant between MIS-C and other disease groups was evaluated using the area under the receiver operating characteristic curve (AUC).The performance of the proteins when combined into a multivariate model was also evaluated, with the proteins combined using the simple disease risk score first described in Herberg et al, 22 and calculated as follows:

Clinical Features
One hundred eight samples were included in this study, with the patient demographic and clinical characteristics summarized in Table 1.All 22 MIS-C patients met the WHO case definitions for inclusion. 21A higher proportion of cases were male (68.2%, n = 15) and the median age of the patients was higher in MIS-C than in the other disease comparator groups.Over half of MIS-C cases (n = 12, 54.5%) required PICU admission and many cases (n = 18, 81.8%) presented with GI symptoms.

Candidate Protein Biomarkers
Seven protein biomarkers for distinguishing MIS-C from other pediatric infectious and inflammatory diseases were selected through unbiased analyses of host transcriptomic profiles obtained from children with MIS-C 25 as well as from the literature.These included ADAMTS2, ARG1, CD163, CORIN and PCSK9, which were selected from RNA-Seq, and CCL20 and CXCL9, which were selected from the literature (see

CD163, CXCL9 and PCSK9 Are Significantly Differentially Abundant in MIS-C Versus Other Diseases and Healthy Controls
Pediatric plasma samples were analyzed from MIS-C (n = 22), KD (n = 23), DB (n = 28), DV (n = 27) and HC (n = 8).The protein concentrations were compared between MIS-C and other diseases (KD, DB, ad DV combined) and HC to evaluate how well the proteins would be able to discriminate MIS-C from non-MIS-C groups (eg, MIS-C vs. KD, DB, DV, HC).The protein concentrations were then compared between MIS-C against each diagnostic group individually.
For CORIN, we observed a significant Bonferroni-adjusted P value of 0.02 for MIS-C versus DV; however, none of the other pairwise comparisons between MIS-C and the other phenotype groups were significantly different.Although CORIN was significant using the Kruskal-Wallis test, we chose to exclude this protein from our final signature as this difference was only present for MIS-C versus DV.ADAMTS2, ARG1 and CCL20 showed no significant difference (adjusted P value >0.05) between MIS-C and both HC and other diseases, nor between MIS-C and each disease group individually.
The performance of smaller combinations of proteins was also evaluated to determine whether all 3 markers (PCSK9, CD163 and CXCL9) were required in the model (Table, Supplemental Digital Content 5, http://links.lww.com/INF/F421).The combination of all 3 proteins was optimal for distinguishing MIS-C from all groups combined (DB, DV, KD), and also for MIS-C versus DB.For MIS-C versus DV and MIS-C versus KD, CD163 + PCSK9 demonstrated the best performance.Despite this improved performance for these comparisons, the 3-protein signature was taken forward for subsequent analyses given the importance in distinguishing MIS-C from severe bacterial infections and the implications for treatment administration.
CRP is a widely measured biomarker, often used for identifying bacterial infections. 30Given the widespread availability of CRP measurement, we evaluated whether adding CRP to the 3-protein model improved its diagnostic performance (Table, Supplemental Digital Content 5, http://links.lww.com/INF/F421).The 3-protein signature + CRP improved the performance for MIS-C versus KD by 9%; however, a 16% reduction in performance was observed for MIS-C versus DB.CRP could not be accurately assessed for the classification of MIS-C versus DV, as CRP was used in the initial clinical assignment of DV patients (DV patients were required to have CRP <60 mg/L).

Lower ARG1 and CORIN Plasma Levels Are Associated With Severity in MIS-C
Associations between plasma protein concentrations and clinical variables were tested (Table 2).Lower levels of ARG1 were observed in MIS-C patients admitted to PICU (Fig. 3A; Table 2; Bonferroni-adjusted P value: 0.014) and requiring inotrope administration (Fig. 3B; Table 2; Bonferroni-adjusted P value: 0.027), suggesting that reduced ARG1 concentration is associated with severity of MIS-C.Lower levels of CORIN were observed among patients who experienced shock (Fig. 3C; Table 2; Bonferroni-adjusted P value: 4.8 × 10 −3 ).Where data were available, associations were also tested in DB, DV and KD for ARG1 and inotrope requirement and PICU admission, and for CORIN levels and shock.No significant associations were observed.When evaluating whether the level of clinical

DISCUSSION
Weeks after SARS-CoV-2 infection, some children developed a hyperinflammatory condition termed MIS-C.The clinical presentation and laboratory findings in MIS-C are like other inflammatory and infectious diseases such as severe bacterial or viral infection and KD, which makes the appropriate diagnosis and treatment of MIS-C challenging.
Here, we evaluated 7 candidate host protein biomarkers in a cohort of 108 children with MIS-C, KD, DB, DV and HC.All MIS-C cases met the WHO case definitions and over half were admitted to PICU.These patients were compared to KD, DB and DV cases recruited before the COVID-19 pandemic.Five of the candidate biomarkers were selected from RNA-Seq and 2 from the literature because of their reported roles in the GI involvement and hyperinflammatory state observed early in the course of MIS-C disease, respectively. 6,17,28he results showed that 3 proteins (PCSK9, CD163 and CXCL9) were significantly higher in MIS-C cases when compared with the other comparator groups.More importantly, the performance of PCSK9, CD163 and CXCL9 when combined into a 3-protein signature could distinguish MIS-C patients from our other disease controls with an AUC of 85.7% (95% CI: 76.6%-94.8%).When we included CRP levels into the 3-protein model, we observed an increase in performance for distinguishing MIS-C from KD of 9%.Given the overlap in clinical features between MIS-C and KD, especially in incomplete KD cases, these findings could be clinically useful due to the importance of timely administration of IVIG to patients with suspected KD to prevent coronary artery aneurysm formation. 31Similarly, we observed an increase in distinguishing MIS-C from DV when including CRP levels into the 3-protein model; however, this finding needs to be further validated in an independent cohort comprised of patients with viral infections that have been phenotyped independent of CRP, as CRP levels were used in the clinical phenotyping of the DV cases in this study (ie, <60 mg/L).
The proteins included in the 3-protein signature may have biologic functions that can provide insight into the pathogenesis of MIS-C.CD163 is a transmembrane macrophage-specific hemoglobin-haptoglobin scavenger receptor that is elevated in MAS and other vasculitic conditions.The biomarker form of this protein is the soluble CD163 (sCD163) found in plasma and produced from increased sCD163 shedding mediated by TNF-α. 32Increased abundance of sCD163 and marked elevation of macrophage activity have been recognized as markers in several inflammatory diseases. 33Mostafa and colleagues 34 reported increased levels of sCD163 in children with SAR-CoV-2 infection and MIS-C, compared with HC, which likely reflects the exaggerated proinflammatory host response and suggests a potential therapeutic role for sCD163 antagonists.CD163 levels have also been shown to strongly correlate with the marked neutrophilia observed in MIS-C, suggesting that this is a result of macrophage activation and supporting the observation that the highest levels of this protein are found in the most severe cases of MIS-C. 20,35CXCL9 is 1 of 3 chemokines that selectively bind to the C-X-C motif chemokine receptor 3 (CXCR3) and its expression is primarily driven by IFNγ. 36CXCL9 is commonly expressed by peripheral blood mononuclear cells, but more specifically by macrophages, and is best known for its role in the inflammatory response by mediating immune cell migration and activation. 37CXCL9 and its mediator, IFN-γ, have been reported to be more abundant in MIS-C cases than in children with mild and severe COVID, confirming the hyperinflammatory state observed in MIS-C. 17,20,28Our results support these findings and validate CD163 and CXCL9 as viable biomarkers for the diagnosis of MIS-C when compared to KD or other common bacterial and viral childhood infections.
Higher levels of PCSK9 were observed in MIS-C cases when compared to all other disease comparator groups.9][40] Increased PCSK9 expression has been linked to lower levels of LDL receptors and consequently higher LDL levels.Cholesterol is essential for SARS-CoV-2 infection as the virus binds to cell surface angiotensin converting enzyme 2 fused to cholesterol, causing a signaling cascade and allowing the lipid-enveloped RNA to enter host cells. 41,42In systemic inflammatory conditions, disruption of the vascular endothelium can lead to dysregulation of lipid transport and metabolism and an increase in circulating LDL and PCSK9.This explains why higher cholesterol levels are associated with higher susceptibility to SARS-CoV-2. 42While LDL data was not available from the patients in this study, future studies to explore the effect of MIS-C on LDL levels could provide insight into the pathophysiology of MIS-C and further validate our findings.PCSK9 also exhibits a proinflammatory effect by promoting TNF-α expression while suppressing the anti-inflammatory markers ARG1 and IL-10. 43The beneficial effect of using PCSK9 inhibitors in adults with severe COVID-19 has been demonstrated [44][45][46] ; however, elevated levels of this protein in MIS-C cases have not, to our knowledge, been previously reported.
ARG1 was not found to be significantly different between our MIS-C cases and other comparator groups; however, lower levels of this protein were associated with cases of MIS-C requiring PICU admission and inotropes.ARG1 is the final enzyme involved in the urea cycle, hydrolyzing arginine to urea, and as such, is highly expressed in the liver. 47,480][51] ARG1 dysregulation has www.pidj.com| 451 Protein Biomarkers for Pediatric MIS-C become synonymous with various pathologic processes involving cardiovascular, immunologic, neurodegenerative and tumorigenic disorders. 52Elevated levels of ARG1 have been reported in COVID-19 cases with high viral loads 51,53 so it is possible that the high levels of PCSK9 observed in MIS-C cases are suppressing ARG1.
Lower levels of CORIN were also associated with severe MIS-C cases with shock.CORIN converts pro-atrial natriuretic peptide (pro-ANP) to biologically active ANP, a cardiac hormone that regulates blood volume and pressure by reducing plasma volume by renal excretion of salt and water, vasodilation and increased vascular permeability. 54Higher pro-ANP levels, which could result from lower levels of CORIN, are associated with poor survival prognosis in children and adults with severe sepsis and septic shock. 55,56The lower levels of CORIN we observed could therefore result in higher levels of pro-ANP and lower levels of ANP, which could contribute to the hypertension, inflammation and cardiac manifestations associated with severity and shock in MIS-C.Future studies to measure the ratio of pro-ANP to ANP levels in MIS-C could provide insight into the disease mechanisms and help in validating our findings.While the best performing diagnostic biomarkers (PCSK9, CD163 and CXCL9) were significantly elevated in MIS-C patients, they were not found to be associated with severity of MIS-C.This could reflect the wide spectrum of mild to severe cases in our cohort.
We did not observe significant differences in the levels of ADAMTS2 or CCL20 in the MIS-C cases when compared to our other disease group.ADAMTS2, a biomarker we selected from RNA data, is a metalloprotease that processes procollagen to provide strength and support to many body tissues.Subsequent to our screening of RNA-Seq data, 2 other groups have reported increased gene expression of ADAMTS2 in COVID-19 and MIS-C cases; however, these groups were compared to HC rather than other disease groups. 57,58Similarly, Gruber et al 6 reported an increased abundance of CCL20 in patients with MIS-C.Our inability to replicate their results could be due to the difference in the control groups Associations between protein levels and binary categorical variables including PICU admission, oxygen requirement, noninvasive ventilation requirement, inotrope requirement, cardiac involvement, GI involvement, rash and conjunctivitis were tested using the Mann-Whitney U test.The association between protein levels and continuous variables including body mass index, days of symptoms, CRP, WBC, neutrophil, lymphocytes, monocytes and platelet levels were tested using linear regression models.Bonferroni-adjusted P values <0.05 are in bold.P values are adjusted for each clinical/laboratory variable.Although our study provides useful information on potential protein biomarkers for MIS-C, there are several limitations that should be taken into consideration.First, our sample size for MIS-C (n = 22) was restricted by the number of children that had been recruited at the time of this study that met all WHO criteria.Second, these patients were recruited during the period of April 2020-August 2021 and, due to the variants that were circulating during this period, would not include the Omicron variant that was first detected in November 2021.Further investigation of the biomarkers we present in this study is required in a larger independent cohort that includes other relevant DB and DV groups with overlapping clinical features of MIS-C including TSS, MAS and hemophagocytic lymphohistiocytosis.Furthermore, our MIS-C cases were older than our other disease comparator groups and the performance of our signature should be tested in younger children with MIS-C.A limitation of all diagnostic studies comparing MIS-C and KD is the lack of a diagnostic test for either disease; however, we included KD patients that were recruited prepandemic to avoid errors in phenotyping.None of the KD patients in this study required inotropes and were, therefore, less severe than the MIS-C, DB and DV groups.The severity associations that we observed for ARG1 and CORIN require further validation in more severe KD cases or KD shock to confirm if this finding is unique to MIS-C.
Finally, this study reports the first protein-based diagnostic signature to discriminate MIS-C from KD and other common bacterial and viral infections.Our results show that the increase in specific proteins involved in macrophage activation, endothelial dysfunction and lipid metabolism can discriminate MIS-C from other inflammatory diseases in children.We further explored the associations between plasma levels and clinical outcomes to find both ARG1 and CORIN levels are lower in severe MIS-C.Once validated, our results could form the foundation for the development of a point-of-care diagnostic test that could assist pediatricians to diagnose and determine the best course of treatment for MIS-C.

©
2024 The Author(s).Published by Wolters Kluwer Health, Inc Yeoh et al TABLE 1. Demographic and Clinical Characteristics of the Patients With MIS-C, DB, DV and KD used for the Enzyme-linked Immunosorbent Assay and Meso Scale Discovery Validation Assay

FIGURE 3 .
FIGURE 3. Levels of ARG1 and CORIN among patients stratified according to whether they were admitted to PICU (A), required inotropes (B) or presented with shock (C).
Table, Supplemental Digital Content 2, http://links.lww.com/INF/F418).Genes were selected based on BH-adjusted P values <10 −5 , a logfold change >2 and if they encoded for extracellularly secreted proteins.All proteins were measured in a cohort of children with MIS-C, KD, DB or DV and HC.When visualized using PC analysis, the disease group was clearly captured by PC 1, with a correla-

TABLE 2 .
Bonferroni-adjusted P Values Showing the Association Between Clinical Variables and Each of the Proteins Measured in MIS-C Patients