Patient characteristics

Koivuniemi et al. [24] have previously described the patient population comprising of 58 female RA patients. Their clinical characteristics are shown in supplementary table 1. Of the 58 patients, 30 had ERA and were treatment-naïve, whereas 28 patients had CRA with inadequate response to conventional synthetic disease-modifying anti-rheumatic drugs, and their medication was modified at the baseline. The follow-up visit took place 12 months later. Between the two visits, DAS28-CRP decreased significantly in both groups [24]. Based on active disease at the baseline, the analyses concern both ERA and CRA patients unless otherwise indicated. Missing data is presented in supplementary table 2.

Serum LPS bioactivity at baseline associates with inflammatory parameters and body mass index (BMI)

First, we explored for any correlations at baseline between serum-induced TLR4 signaling in the reporter cells (LPS bioactivity), inflammatory parameters, and disease activity. No correlation was found between LPS bioactivity and baseline disease activity as measured by DAS28-CRP (r=+0.13 [95% confidence interval (CI) −0.15 to +0.39], p=0.35) or the number of swollen (r=−0.02 [95% CI −0.29 to +0.25], p=0.88) or tender joints (r=−0.05 [95% CI −0.23 to +0.34], p=0.70). LPS bioactivity correlated with the inflammatory parameters erythrocyte sedimentation rate (ESR, r=+0.28 [95% CI +0.01 to 0.52], p=0.037), SAA (r=+0.35 [95% CI +0.09 to 0.57], p=0.008), YKL40 (r=+0.27 [95% CI +0.002 to 0.51], p=0.042), and E-selectin (r=+0.33 [95% CI +0.07 to 0.56], p=0.013), but not with hsCRP, resistin, visfatin, or IL-6 (Supplementary table 3a). Baseline LPS bioactivity correlated also with BMI (r=+0.42 [95% CI +0.17 to 0.62], p=0.002), the amount of adipose tissue (r=+0.47 [95% CI +0.21 to 0.67], p<0.001), blood pressure (systolic r=+0.40 [95% CI +0.14 to 0.61], p=0.002 and diastolic, r=+0.37 [95% CI +0.14 to 0.61], p=0.006), and advancing age (r=+0.45 [95% CI +0.21 to 0.64], p<0.001) (Supplementary table 3a), but not with the presence of ACPAs (mean±SD [standard deviation] 0.24±0.12 vs mean±SD 0.19±0.08 EU/ml, p=0.22).

As previous studies have utilized LBP, CD14, and CD163 as surrogate markers for LPS levels, we analyzed their correlations with LPS bioactivity. LPS bioactivity at baseline correlated significantly with the CD163 (r=+0.44 [95% CI +0.19 to 0.63], p<0.001), but not with LBP (r=+0.22 [95% CI −0.06 to +0.46], p=0.12) or CD14 (r=+0.06 [95% CI −0.19 to +0.35], p=0.53). Baseline LBP correlated with DAS28-CRP (r=+0.39 [95% CI +0.13 to 0.60], p=0.004), ESR (r=+0.47 [95% CI +0.22 to 0.66], p<0.001), and hsCRP (r=+0.47 [95% CI +0.22 to 0.66], p<0.001) (Supplementary table 3b). CD163 correlated also with the markers of inflammation and with disease activity (Supplementary table 3c). Furthermore, baseline LBP and CD163 correlated with each other (r=+0.50 [95% CI +0.26 to 0.68], p<0.001). CD14 correlated with LBP, CD163, and ESR, but not with disease activity (Supplementary table 3d). The total serum LPS concentrations as measured by the EndoLISA assay did not correlate with disease activity, inflammatory parameters, and had a significant correlation at the baseline with CD14 and CD163 but not LPS bioactivity or LBP (Supplementary table 3e).

Serum LPS bioactivity at the follow-up visit associates with inflammatory parameters and disease activity

At the follow-up visit after 12 months, the correlations between LPS bioactivity and disease activity and hsCRP became statistically significant (for DAS28-CRP r=+0.48 [95% CI +0.24 to 0.67] and p<0.001, for hsCRP r=+0.41 [95% CI +0.15 to 0.61] and p=0.004, and for ESR r=+0.29 [95% CI +0.02 to 0.52] and p=0.030) (Supplementary table 3a). LPS bioactivity correlated highly significantly also with the number of tender (r=+0.40 [95% CI +0.14 to 0.60], p=0.003) and swollen joints (r=+0.40 [95% CI +0.12 to 0.59], p=0.003). Furthermore, LPS bioactivity correlated significantly with patient-related outcomes (PROM) (Supplementary table 3a). LBP levels at 12 months also correlated with inflammatory parameters hsCRP (r=+0.54 [95% CI +0.31 to 0.71], p<0.001) and ESR (r=+0.45 [95% CI +0.20 to 0.64], p=0.001), as well as with DAS28-CRP (r=+0.37 [95% CI +0.11 to 0.58], p=0.005) and the number of swollen joints (r=+0.40 [95% CI +0.14 to 0.61], p=0.003), but not with tender joints (r=+0.22 [95% CI −0.06 to +0.46], p=0.11) (Supplementary table 3b). CD163 levels correlated with DAS28-CRP (r=+0.37 [95% CI +0.11 to 0.59], p=0.005) and ESR (r=+0.37 [95% CI +0.11 to 0.59], p=0.005), but not with CRP or number of swollen joints (Supplementary table 3c). CD14 levels correlated also with DAS28-CRP (r=+0.27 [95% CI +0.004 to 0.51], p=0.041) (Supplementary table 3d). In contrast, total serum LPS concentrations did not correlate significantly with most of the parameters of disease activity, inflammation, or LPS activity (Supplementary table 3e). Patients who failed to achieve ACR/EULAR remission after 12 months had higher levels of LPS bioactivity (mean±SD 0.22±0.09 vs 0.15±0.07 EU/ml, p<0.001), LBP (mean±SD 6985±2860 vs 5162±1715 ng/ml, p=0.008), and CD163 (mean±SD 1527±630 vs 1135±404 ng/ml, p=0.008) (Fig. 1a–c). Figure 2 and supplementary tables 3a-e present the correlations between all measured biomarkers and clinical characteristics.

Fig. 1

Remission (ACR/EULAR 2011) of rheumatoid arthritis is associated with lower levels of serum LPS bioactivity (A), LBP (B), and CD163 (C) concentrations at the 12-month follow-up visit. Notched boxplots represent interquartile ranges and 95% confidence intervals of the medians. *p≤0.05, **p≤0.01, ***p≤0.001

Fig. 2

A correlation plot of serum LPS bioactivity and the concentrations of LBP, CD163, CD14, and LPS (EndoLISA) with each other, disease activity, metabolic factors, and inflammatory biomarkers. Colors represent Spearman correlation coefficients. *p≤0.05, **p≤0.01, ***p≤0.001

Serum LPS bioactivity and LBP levels at baseline correlate with disease activity at 12 months and the likelihood of achieving remission

As expected, disease activity decreased between baseline and 12-month follow-up visits (DAS28-CRP; mean±SD 3.54±1.09 vs 2.28±0.95, p<0.001). LPS bioactivity in the entire patient population, however, remained unchanged (Supplementary table 4). As the associations between LPS bioactivity and disease activity parameters became significant at 12 months, we explored the possibility that continuously elevated circulating LPS levels could relate to higher disease activity and also to less favorable treatment response. Indeed, LPS bioactivity measured at the baseline correlated significantly with disease activity at 12 months (DAS28-CRP, r=+0.29 [95% CI +0.02 to 0.52], p=0.031) as well as with ESR (r=+0.28 [95% CI +0.003 to 0.51], p=0.042), CRP (r=+0.30 [95% CI +0.03 to 0.53], p=0.025) and, in patients with early RA, with ACR/EULAR remission (mean±SD 0.24±0.09 EU/ml in ERA patients without remission vs 0.15±0.07 EU/ml, p=0.009; for all patients mean±SD 0.21±0.09 vs 0.17±0.07 EU/ml, p=0.065). In line with this, the level of LBP at baseline correlated significantly with disease activity (r=+0.34 [95% CI +0.07 to 0.56], p=0.012), hsCRP (r=+0.49 [95% CI +0.24 to 0.69], p<0.001) and swollen joints (r=+0.37 [95% CI +0.10 to 0.59], p=0.007) in the entire cohort at 12 months. Thus, high baseline levels of LPS bioactivity and LBP were both predictive of poor treatment response. Of the 20 patients with LPS bioactivity above median on both visits, only 3 (15%) reached ACR/EULAR remission at 12 months, whereas of the remaining 33 patients with LPS bioactivity below median, 18 (55%) had reached remission (p=0.004). Finally, LPS bioactivity at baseline correlated significantly with the levels of surrogate markers LBP (r=+0.31 [95% CI +0.03 to 0.54], p=0.025) and CD163 (r=+0.45 [95% CI +0.19 to 0.65], p<0.001) but not CD14 (r=+0.21 [95 % CI −0.08 to +0.46], p=0.14) at 12 months. Together these data suggest that levels of LPS bioactivity and the surrogate markers of LPS are relatively constant in RA patients, and the higher the levels are the less likely the patients are to reach remission.

We also attempted to study the effect of anti-rheumatic treatment on LPS bioactivity, but the small number of patients did not allow any detailed analysis. The use of Mtx was associated with lower LPS bioactivity and LBP level at 12 months in patients with CRA (mean±SD 0.26±0.11 vs 0.19±0.07 EU/ml, p=0.053 and 7811±2225 vs 5585±2077 ng/ml, p=0.016, respectively), suggesting that Mtx may in part modify disease activity by reducing systemic LPS bioactivity. Accordingly, in a principal component analysis of LPS bioactivity, LBP level, and inflammatory biomarkers in CRA patients of those on Mtx appeared to cluster apart from those who were not (Fig. 3).

Fig. 3

Principal component analysis of patients with chronic rheumatoid arthritis at the 12-month follow-up visit including LPS bioactivity and the concentrations of LBP, SAA, hsCRP, E-Selectin, YKL-40, and IL-6 in sera demonstrates clustering according to methotrexate use

Neutralization of LPS abrogates the ability of serum to activate TLR4

Human serum contains various factors that potentially activate the TLR4 or NF-κB signaling, such as the acute phase proteins SAA, high mobility group box (HMGB) 1, and several cytokines. To explore the proportion of TLR4 activity contributed by serum LPS, we added polymyxin B to all serum samples to specifically neutralize LPS [27]. Polymyxin B abolished most of the TLR4 activity present in the sera of RA patients. No significant correlation existed between the residual TLR4 activity and parameters of inflammation, including SAA, or disease activity, except for the number of swollen joints at follow-up visit. In contrast, the proportion of TLR4 activity that was neutralized by polymyxin B correlated with the inflammatory parameters and disease activity in a similar manner to serum total TLR4 activity, suggesting that most of the TLR4 activity in the sera of RA patients is induced by LPS (Supplementary table 5a-b).

LPS bioactivity, metabolic syndrome, and cardiovascular risk factors in RA patients

Metabolic endotoxemia is strongly associated with the metabolic syndrome, and obesity is associated with a less favorable prognosis of RA [28]. Patients with RA have an increased risk of cardiovascular diseases [29], and therefore, we explored the association of LPS bioactivity with cardiovascular risk factors, although several risk factors such as diabetes had been excluded from the patient cohort. LPS bioactivity was associated with higher blood pressure both at baseline (r=+0.40 [95% CI +0.14 to 0.61], p=0.002 for systolic and r=+0.37 [95% CI +0.11 to 0.59], p=0.005 for diastolic blood pressure) and at 12 months (r=+0.26 [95% CI −0.01 to 0.50], p=0.055 for systolic and r=+0.36 [95% CI +0.10 to 0.58], p=0.007 for diastolic blood pressure). As expected, LPS bioactivity associated with increased BMI both at baseline (r=+0.42 [95% CI +0.17 to 0.62], p=0.002) and at the 12-month visit (r=+0.39 [95% CI +0.14 to 0.60], p=0.003) (Fig. 2). The proportion of adipose tissue correlated significantly with LPS bioactivity (at baseline r=+0.47 [95% CI +0.21 to 0.67], p=0.001; at 12 months r=+0.36 [95% CI +0.09 to 0.58], p=0.007) as did also CD163 levels (BMI: r=+0.26 [95% CI −0.02 to 0.50], p=0.061 at baseline, r=+0.45 [95% CI +0.21 to 0.64], p<0.001 at 12 months; adipose tissue: r=+0.35 [95% CI +0.07 to 0.58], p=0.014 at baseline, r=+0.43 [95% CI +0.17 to 0.63], p=0.001 at 12 months) (Fig. 2). High BMI at baseline also associated with a decreased likelihood of achieving ACR/EULAR remission at 12 months (mean±SD 25.8±4.2 vs 22.7±3.3, p=0.003). HbA1c levels correlated with LBP at the follow-up visit (r=+0.37 [95% CI +0.11 to 0.58], p=0.005). HbA1c also correlated with DAS28-CRP at 12 months (r=+0.31 [+0.05 to 0.53], p=0.019), consistent with the possibility that hyperglycemia can disturb the intestinal barrier function [30].