Characterization of IVIG-resistant systemic inflammation in KD patients
At screening visit (prior to anakinra, Figure 1A) circulating levels of IL-6 (median fold change vs. HC: 48.25), CXCL10 (26.67), S100A12 (20.27) and IL-1Ra (15.18) were most prominently upregulated compared to healthy controls (Figure 1B). Multiple correlation analyses of blood biomarkers, cell counts, clinical routine inflammatory parameters as well as Z-score identifies numerous associations between inflammatory cytokines and immune cells, with two prominent clusters of strong positive correlation comprising IL-1b, IL-1RA, IL-6, S100A12, Galectin 3, CRP and particularly neutrophils as well as TNF, IFN-γ and MCP-2 (Figure 1C). Coronary artery Z-score at screening visit was predominantly associated with lympho- and thrombocyte counts as well as hemoglobin concentration rather than inflammatory mediators. Markers of liver function (Bilirubin, AST, ALP, ALT, GGT) revealed pronounced positive correlation with cellular adhesion molecules (sICAM-1, sVCAM-1) as well as IL-17A. In contrast to many other inflammatory parameters both IL-17A and bilirubin were positively associated with hemoglobin concentrations (Figure 1C).
Unsupervised hierarchical clustering according to correlation distance and Ward’s linkage of serum marker data obtained from samples at screening visit (prior to anakinra) highlights differing levels of systemic inflammatory activity at study entry (Figure 1D). Blood biomarkers in patients’ samples constituting cluster 1 were hardly elevated over those in healthy controls, whereas cluster 2 comprised patients with in part substantial elevation of serum markers (Figure 1D). From this as well as all subsequent analyses we excluded patient 15, who was retrospectively diagnosed with sJIA associated MAS (sJIA-MAS) and did already present with strongly elevated serum levels of both S100A12 and IL-18 as hallmarks of sJIA-MAS(18) at screening visit (Figure S1A). Of note, all study participants received only one IVIG-infusion prior to anakinra, except for patients 7 and 14 (two IVIG-infusions) and patient 15 (three IVIG-infusions).
Upon single marker analyses, we observed absolute levels of LRG1, MCSF as well as MCP3 to best separate patients in cluster 2 from cluster 1 and healthy controls (Figure 1E). At MFI level IL-1β and IFN-β were also significantly elevated among patients in cluster 2 but overall levels were too low to allow for robust conversion to absolute concentration (Figure S1B). While many other of the quantified blood biomarkers were elevated among individuals in cluster 2 over those in cluster 1 and healthy controls, those differences in expression remained far below significance level (Figure S1C). Furthermore, concentrations of CRP and hemoglobin as classical markers of inflammation indicated higher inflammatory activity among patients in cluster 2 (Figure 1F). This was not reflected by blood cell counts (Figure S1D).
As already obvious from correlation analyses of multiple laboratory and clinical markers at screening visit (Figure 1C), at expression level (MFI) we observed significant association between LRG1 and M-CSF (rs=0.57, p=0.038) or MCP3 (rs=0.60, p=0.025) and IFN-β (rs=0.76, p=0.003) as well as IL-1β by trend (rs=0.52, p=0.058). Expression of IL-1β – as the treatment target in this study – was further significantly associated with IL-1Ra (rs=0.76, p=0.002), M-CSF (rs=0.55, p=0.044), MCP3 (rs=0.80, p=0.001) and IFN-β levels (rs=0.71, p=0.006), as well as CRP (rs=0.59, p=0.028) and hemoglobin concentrations (rs=-0.59, p=0.029).
Association of anakinra treatment and dosage with blood biomarker levels
Most serum marker levels quantified in samples collected at screening visit, following 3 days of anakinra treatment (d3) and at day 14 (d14, end of anakinra treatment; Figure 2A) rapidly declined in course of anakinra treatment (Figure 2B-D). Reduction upon blockade of IL-1 signaling was most pronounced for IL-6, IL-10, CXCL10, sICAM-1, sVCAM-1 as well as S100A12, both at cumulative (Figure 2B, left panels) and individual patient follow-up level (Figure 2B, right panels). Compared to screening visit levels decline following three and 14 days of anakinra was most pronounced for CXCL10 (median Δ to d3=-14.64; median Δ to d14=-21.82), IL-6 (-5.25; -42.75), S100A12 (-0.59;-8.38), IL-10 (-1.48; -2.80) and IL-8 (-0.88; -2.0). Increase in IL-1Ra levels on days three and 14 compared to screening visit reflect anakinra treatment (median Δ to d3=68.79; median Δ to d14=46.82; Figure 2C, D and S2A).
Following three days of anakinra treatment, unsupervised hierarchical clustering separated the patient cohort into two main groups (Figure 2E), which overlapped strongly with the biomarker-induced clustering at the screening visit (Figure 1D) and reflected persistent systemic inflammatory activity among patients in cluster two over those in cluster one as well as patients 1 and 4 already associating with healthy controls. IL-1Ra levels quantified on study day three were not considered in this analysis, as – in contrast to measurements at screening visit - these levels would rather reflect anakinra than endogenous IL-1Ra concentrations. When analyzing peripheral IL-1Ra levels separately, we observed significantly higher levels of IL-1Ra in patients in cluster 2 than in cluster 1 (Figure 2F). Four out of five patients in cluster 2 versus three out of nine patients in cluster 1 (including patients 1 and 4) required anakinra dose escalation (Figure 2F). Patient 8 who only received a single but falsly too high dose of anakinra was excluded from this as well as subsequent analysis.
When further analyzing individual biomarker levels, we again observed absolute levels of LRG1 and MCP3 as well as MFI levels of IL-1β and IFN-β to best separate patients in cluster 2 from cluster 1 and healthy controls (Figure 2G, S3A). Other quantified blood biomarkers were likewise elevated among individuals in cluster 2 but expression differences compare to cluster 1 remained far below significance level (Figure S1B). Clinical routine amrkers of inflammation indicated increased inflammatory activity among patients in cluster 2 only be trend (Leukocyte/PMN/lymphocyte counts) or not at all (hemoglobin, CRP; Figure S3C).
Multiple correlation analyses of cluster differentiating serum markers (LRG1, MCP3, IL-1β, IFN-β), inflammatory parameters and cell counts indicated marked positive associations among blood biomarkers as well as leukocyte counts (Figure 2H).
Performing unsupervised hierarchical clustering analyses at the end of the anakinra treatment period (d14; Figure S4), biomarker-driven grouping of patients as observed at screening and d3 visit was almost lost.
Prediction of anakinra treatment regiment
Apart from unsupervised biomarker driven analyses of the study cohort we also stratified serum biomarker and inflammatory parameter data based on study outcome. We observed that over all samples, absolute serum concentrations of IL-6, LRG1 and S100A12 were significantly elevated among patients requiring anakinra dose escalation (Figure 3A, B). Similarly, CRP concentrations as well as blood PMN and leukocyte counts were higher among this subgroup of IVIG-resistant KD patients (Figure 3C, D). Furthermore, we wondered if serum biomarker and/or inflammatory parameter levels prior to IL-1Ra treatment might already facilitate a prediction on whether a patient may require future dose adjustment of anakinra. In these analyses of serum marker levels at screening visit we observed IL-6, LRG1 and S100A12 serum levels to indicate future dose escalation only by trend (Figure S5A, B), while CRP concentrations (AUC=0.85, p=0.028, 95% CI=0.64-1.0; cut-off >45.5 mg/L, 100% Sens, 66.7% Spec) as well as blood PMN (AUC=0.89, p=0.013, 95% CI=0.70-1.0; cut-off >9730 cells/mm3, 88.9% Sens., 83.3% Spec.) and leukocyte counts (AUC=0.89, p=0.014, 95%CI=0.72-1.0; cut-off >15335 cells/mm3, 87.5% Sens., 83.3% Spec.) in patients requiring future anakinra dose escalation were already significantly elevated at screening visit (Figure S5C, D).
Association of LRG1 expression with IL-1 signaling
Among all serum biomarker analyses in IVIG-resistant KD patients, LRG1 stood out most consistently as overexpressed in a subgroup of patients with enhanced inflammatory activity and largely associated with the need to increase respective anakinra dosage over the course of treatment. LRG1 belongs to the leucine-rich repeat (LRR) protein family and is expressed by many cell types including endothelial cells, but is mainly produced by hepatocytes and myeloid cells, particularly granulocytes (19). When analyzing association of collective LRG1 serum levels (all study visits) with circulating neutrophil (PMN), leukocyte, thrombocyte and lymphocyte counts as well as markers indicating hepatic stress (transferases), the only significant correlation we observed was between LRG1 and peripheral neutrophil levels (Figure 4A). Further, we performed a collective correlation analysis of LRG1 with all potentially proinflammatory markers assessed in our study. From these analyses, we excluded type 1 interferons and MCP-3, due to their generally low expression in our assays, as well as sVCAM-1, sICAM-1 and sFasL as molecules less involved in direct inflammatory signaling. IL-1Ra was excluded as quantified levels beyond screening visit rather reflect anakinra than endogenous expression. Among all included inflammation modulating cytokines, chemokines and DAMPs, this multiple correlation analysis indicated LRG1 to specifically associate with IL-1β concentrations throughout (Figure 4B).
In order to further investigate a link between IL-1β signaling and LRG1 expression we stimulated primary human coronary artery endothelial cells (HCAECs) with previously demonstrated high sensitivity to IL-1β signaling (8) healthy donor whole blood, and isolated neutrophils as among the main LRG1 producers (19) with recombinant human IL-1β. Here, we observed concentration dependent LRG1 release from particularly isolated human neutrophils, which were also responsive to recombinant human IL-1β on the level of other molecules (i.e. TNF, IL-6, IL-8, IL-1Ra, sICAM-1; Figure 4C, D). Compared to neutrophils, IL-1β stimulation induced only minor LRG1 release from HCAECs. Without stimulation, basal LRG1 levels in whole blood were already markedly elevated over those quantified in both PMN as well as HCAEC supernatants and in range with those previously observed in healthy control sera and could not be increased any further by stimulation with recombinant human IL-1β (Figure 4C, D).
Next, we aimed at validating our blood biomarker data in a different KD cohort and on gene rather than protein expression level. Therefore, we used already available blood transcriptional profile data of 146 KD patients in course of IVIG treatment (GSE63881), which had previously highlighted a prominent role of the IL-1β pathway in disease (17). In this retrospective analysis of z score on normalized data based on log10-transformation, we observed LRG1 expression not to differ between IVIG responders and non-responders (Figure 4E). In line with our data obtained from the KAWAKINRA study cohort, we further compared expression levels of LRG1 in acute and convalescent phase with those of molecules involved in IL-1 signaling (IL1R1, IL1B), S100A12 as neutrophil marker but also IL6 as top de-regulated in KAWAKINRA, as well as IL6RA and IL6RB. We found all significantly reduced upon convalescence (Figure 4F), except for IL6, which contrasts our serum marker data (Figures 1B, 2B). Further, we observed strongest correlations between LRG1 and IL1B (r=0.84, p=1.47⋅10−17), IL1R1 (r=0.77, p=6.44⋅10−13) and S100A12 (r=0.68, p=2.1⋅10−9; Figure 4G).
Finally, we compared LRG1 serum levels in IVIG-resistant KD patients with concentrations quantified in MAS as a hyperinflammatory condition, which can also complicate KD (5). Respective data were obtained from one patient enrolled in KAWAKINRA but retrospectively diagnosed as sJIA-MAS (patient 15) as well as patients previously reported in other context (15) (Table 1). Further, we included LRG1 levels in MIS-C in this comparison, as this condition is thought to share clinical features with KD (20, 21) (Table 1). Collectively, we observed elevated LRG1 levels to separate the investigated IVIG-resistant KD patients (particularly those with high inflammatory activity) from both sJIA-MAS as well as MIS-C (Figure 4H).