In this study, targeted metabolomics was used to identify JIA-associated biomarkers in plasma. Subtype analysis revealed potential diagnostic metabolic biomarkers in plasma for sJIA. A strength of our approach was the analysis of four JIA subtypes. A previous untargeted liquid chromatography/mass spectrometry metabolomics study reported some differences in metabolites between JIA (n = 20) and controls (n = 20) but did not investigate JIA subtype-specific differences (19). Additionally, by using a targeted platform with clinically validated biomarkers, the biological context of our findings can be more readily interpreted than previous studies (35, 36).
Of the 71 metabolomic measures investigated, there were clear differences for a small number of measures in children with JIA relative to controls, suggesting that JIA is accompanied by specific metabolic alterations in plasma (higher GlycA and DHA, and lower acetate compared to non-JIA). Acetate was also negatively associated with GlycA, suggesting that the lowered acetate in JIA group may be caused by chronic inflammation. There are few previous JIA data to compare these findings to, as previous metabolomics studies have used mass-spectrometry metabolomics platforms with very few biomarkers overlapping with the targeted NMR platform we have used in this study. However, there have been adult studies investigating NMR metabolomic profiles associated with RA (37). In line with the findings of our study, GlycA levels were observed to be elevated in individuals with RA and correlated with the severity of RA (37). However, in contrast to our finding of higher DHA in JIA, RA was associated with lower DHA (37). Our finding that DHA is higher in the JIA group than controls is surprising as DHA has anti-inflammatory properties (38). One possible explanation for differences in DHA findings could be differences in medication used. The UK cohort considered medications such as non-steroid anti-inflammatory drugs (NSAIDs) and lipid-lowering drugs (37) whereas our study involved larger administrations such as steroids, NSAIDs and particularly, disease modifying anti rheumatoid drugs (DMARDs), which may have a stronger influence on lipid metabolism in JIA patients (39). The observed differences between two studies could be attributed, in part, to the age of the subjects. The concentration of fatty acids including DHA are generally higher in infants and children compared to adults, associated with the essential nature of DHA which is crucial for brain development (40, 41). The elevated levels in children may be related to intake of DHA-rich human milk and certain foods. This result also contrasts with a previous JIA study measuring serum fatty acids with mass spectrometry that reported DHA was lower in JIA (42). This study examined changes in serum DHA levels in response to dietary intake and reported lower DHA levels in patients with active and short-lasting disease activity, particularly the number of affected joints in pJIA subtype (42). On the other hand, our study directly measured DHA levels in plasma, without considering dietary intake, and primarily focused on the systemic inflammation associated with sJIA. Our approach of not accounting for dietary fluctuations may introduce biases into the findings.
While we did not see an association between mediation status and metabolomic biomarkers in the sJIA subgroup in this study (Fig. 2B, Supplementary Table S4), this may be due to insufficient power as there were only n = 5 non-medicated sJIA individuals in this study. In addition, we lacked detailed information such as treatment period. Therefore, further research is required to investigate the extent to which medication can influence lipid metabolism in JIA patients.
In subtype-specific analysis, metabolic variation was most evident for sJIA, which has the most severe inflammatory disease course among four subtypes (4, 8, 9, 43). While no metabolic biomarkers were significantly associated with oJIA and pJIA (RF- and RF+), sJIA was associated with 24 metabolic biomarkers. These results may reflect that sJIA is characterised by systemic inflammation affecting the circulation, in contrast to oJIA and pJIA, which show localised inflammation at the joints (44). As GlycA was strongly associated with sJIA, we explored if this chronic inflammation marker explained the other metabolic changes in sJIA. GlycA was positively associated with 4 sJIA-associated biomarkers and negatively associated with 2 sJIA-associated biomarkers, acetate, and histidine. This negative association is consistent with previous evidence that acetate and histidine have anti-inflammatory properties and regulate the production of pro-inflammatory cytokines and the activation of inflammatory cells (45, 46). Particularly, histidine-rich glycoprotein (HRG) has been identified as a biomarker in adult RA (47). However, acetate has not previously been linked to JIA nor RA. Another reason for the strongest evidence being observed in sJIA may be the smaller sample size of oJIA (n = 19), pJIA(RF-) (n = 20) and pJIA(RF+) (n = 14) subtypes compared to sJIA (n = 22). Of note, the pJIA(RF+) group in this cohort were all female and older on average than the other subtypes, in line with its known preponderance of girls (8–9 females : 1 male) and age range (generally 10 to 13 years of age) (48).
Overall, our findings indicate that only some of the JIA-associated metabolic biomarkers were also GlycA-associated, suggesting that there are metabolic signatures in sJIA that are independent of chronic inflammation. Further, as GlycA is a composite marker of several inflammatory processes in the body, such as elevated inflammatory cytokines and neutrophils, it is difficult to determine which of these cellular processes are contributing most to metabolome differences (27, 49). Future research is needed to characterise other soluble inflammatory mediators such as cytokines and inflammatory cells in JIA.
Strengths and limitations
This study investigated the association between the metabolome and JIA, as well as chronic inflammation, using a targeted NMR metabolomic approach. Our pediatric JIA cohort is large compared to previous metabolomic studies in JIA and included four distinct subtypes. Certain limitations merit consideration. The limited number of rarer subtypes, pJIA (RF+), comprising only 14 cases in the cohort, affected the statistical power for relevant subtype analyses. The lack of data for the medication history, including the duration of medication and symptom improvement after the treatment, means we were unable to account for these potential confounding factors. Furthermore, the exclusive representation of female patients in pJIA (RF+) subtype may impact the results, even after adjusting for sex in the linear regression model. Finally, our control cohort was younger on average than our JIA cohort, which may influence the results. However, age was included as a covariate in all statistical models, and our strongest signature in sJIA was not observed in the other JIA subtypes, all of which had similar mean age.