Network analysis of inflammation and symptoms in recent onset schizophrenia and the influence of minocycline during a clinical trial

Attempts to delineate an immune subtype of schizophrenia have not yet led to the clear identification of potential treatment targets. An unbiased informatic approach at the level of individual immune cytokines and symptoms may reveal organisational structures underlying heterogeneity in schizophrenia, and potential for future therapies. The aim was to determine the network and relative influence of pro- and anti-inflammatory cytokines on depressive, positive, and negative symptoms. We further aimed to determine the effect of exposure to minocycline or placebo for 6 months on cytokine-symptom network connectivity and structure. Network analysis was applied to baseline and 6-month data from the large multi-center BeneMin trial of minocycline (N = 207) in schizophrenia. Pro-inflammatory cytokines IL-6, TNF-α, and IFN-γ had the greatest influence in the inflammatory network and were associated with depressive symptoms and suspiciousness at baseline. At 6 months, the placebo group network connectivity was 57% stronger than the minocycline group, due to significantly greater influence of TNF-α, early wakening, and pathological guilt. IL-6 and its downstream impact on TNF-α, and IFN-γ, could offer novel targets for treatment if offered at the relevant phenotypic profile including those with depression. Future targeted experimental studies of immune-based therapies are now needed.


Sampled
Average correlation with original strengt Supplementary materials for network of inflammatory cytokines and symptoms at baseline Nonparametric bootstrapped confidence intervals.The Figure below presents the nonparametric bootstrapped confidence intervals of estimated edge-weights for the network comprising influential pro-inflammatory cytokines and symptoms whilst controlling for covariates.As can be seen, There were sufficiently narrow bootstrapped CIs around the estimated edge-weights.The IL-6-self-depreciation edge-weight was estimated to be above zero in 81% of the 1000 nonparametric bootstrapped procedures, IFN-γ-hopelessness in 64%, and TNF-α-suspiciousness in 63%. Figure Notes.The red line indicates the sample values and they grey area indicates the bootstrapped CIs.Each horizontal line represents one edge of the network, ordered from the edge with the highest edge-weight to the edge weight with the lowest edge-weight.In the case of equivalent sampled edge weights, the mean of the bootstrap samples was used in ordering the edges.
0.09 0.17 Supplementary materials for network of treatment allocation, influential pro-inflammatory cytokines and symptoms whilst controlling for covariates at 6-month follow-up Nonparametric bootstrapped confidence intervals.The Figure below presents the nonparametric bootstrapped confidence intervals of estimated edge-weights for the network comprising treatment allocation, influential pro-inflammatory cytokines, and symptoms whilst controlling for covariates at six-month follow-up.As can be seen, there were sufficiently narrow bootstrapped CIs around the estimated edge-weights.The treatment-TNF-α association was estimated to be above zero in 96% of the 1,000 nonparametric bootstrapped procedures, and treatment-early wakening in 60%.

Supplementary materials for network of influential pro-inflammatory cytokines and symptoms whilst controlling for covariates at 6-month follow-up in PLACEBO group
Nonparametric bootstrapped confidence intervals.The Figure below presents the nonparametric bootstrapped confidence intervals of estimated edge-weights for the network comprising influential pro-inflammatory cytokines and symptoms of the depression and suspiciousness community whilst controlling for covariates in the placebo group.As can be seen, there were wide bootstrapped CIs around the estimated edge-weights, suggesting that relative edge strength should be interpreted with caution.Nonetheless, the TNF-α-IL6 association was estimated to be above zero in 98% of the 1000 nonparametric bootstrapped procedures, TNF-α-IFN-γ in 97%, and IL6-IFN-γ in 35%.Supplementary materials for network of influential pro-inflammatory cytokines and symptoms whilst controlling for covariates at 6-month follow-up in MINOCYCLINE group Nonparametric bootstrapped confidence intervals.The Figure below presents the nonparametric bootstrapped confidence intervals of estimated edge-weights for the network comprising influential pro-inflammatory cytokines and symptoms of the depression and suspiciousness community whilst controlling for covariates in the minocycline group.As can be seen, there were wide bootstrapped CIs around the estimated edge-weights, suggesting that relative edge strength should be interpreted with caution.The TNF-α-IL6 association was estimated to be above zero in 49% of the 1000 nonparametric bootstrapped procedures, TNF-α-IFN-γ in 28%, IL6-IFN-γ in 41%, and IL6-observed depression in 56%.
bootstrapped procedures.The Figures below show the CS-coefficients for edge-weights for the network comprising influential pro-inflammatory cytokines and symptoms of the depression and suspiciousness community.The CS-co-efficient was 0.67.
bootstrapped procedures.The Figures below show the CS-coefficients for edge-weights for the network comprising treatment allocation, influential pro-inflammatory cytokines, and symptoms of the depression and suspiciousness community.The CS-co-efficient was 0 bootstrapped procedures.The Figures below show the CS-coefficients for edge-weights for the network comprising treatment allocation, influential pro-inflammatory cytokines, and symptoms of the depression and suspiciousness community in the placebo group.The CS-co-efficient was 0.36.
bootstrapped procedures.The Figures below show the CS-coefficients for edge-weights for the network comprising treatment allocation, influential pro-inflammatory cytokines, and symptoms of the depression and suspiciousness community in the placebo group.The CS-co-efficient was 0.43.