Differential risk of tuberculosis reactivation among anti-TNF therapies is due to drug binding kinetics and permeability

Abstract

Increased rates of tuberculosis (TB) reactivation have been reported in humans treated with TNF-α (TNF)-neutralizing drugs, and higher rates are observed with anti-TNF Abs (e.g., infliximab) as compared with TNF receptor fusion protein (etanercept). Mechanisms driving differential reactivation rates and differences in drug action are not known. We use a computational model of a TB granuloma formation that includes TNF/TNF receptor dynamics to elucidate these mechanisms. Our analyses yield three important insights. First, drug binding to membrane-bound TNF critically impairs granuloma function. Second, a higher risk of reactivation induced from Ab-type treatments is primarily due to differences in TNF/drug binding kinetics and permeability. Apoptotic and cytolytic activities of Abs and pharmacokinetic fluctuations in blood concentration of drug are not essential to inducing TB reactivation. Third, we predict specific host factors that, if augmented, would improve granuloma function during anti-TNF therapy. Our findings have implications for the development of safer anti-TNF drugs to treat inflammatory diseases.

Figure 1

Multi-scale model of the immune response to Mtb infection in the lung and TNF neutralization. Details are presented in Methods. (A) Selected cell-level ABM rules based on known immunological activities and interactions (M_i: infected macrophage, M_ci: chronically infected macrophage, M_a: activated macrophage, T_γ: pro-inflammatory IFN-γ producing T cell, T_c: cytotoxic T cell, T_reg: regulatory T cell). (B) Binding interactions and reactions controlling TNF/TNFR dynamics at the single-cell level. (C) Drug transport from a vascular source to the grid. Vascular permeability coefficient (k_c) determines the level of drug penetration from blood into lung tissue (relationship between C_p and C_source) as described in Methods. (D) Addition of TNF neutralizing drugs with either constant or varying blood concentrations (C_p), 100 days after Mtb infection. (E) Hypothetical classes of TNF inhibitors defined in this study based on TNF binding characteristics: Class 1 binds sTNF, but not mTNF, at a binding ratio of 1:1; Class 2 binds both sTNF and mTNF at a binding ratio of 1:1; Class 3 binds both sTNF and mTNF at a TNF/drug binding ratio of 1:3. Numbers represent reactions as listed in Table 1. (F) The effect of drug-induced cell death in TNF-expressing cells.

Figure 2

Effect of TNF/drug binding characteristics on bacterial levels within a granuloma 100 days after anti-TNF treatment. (A)–(C) Effect of variations of TNF/drug binding (k_{on_TNF/Drug}) and unbinding rate constants (k_{off_TNF/Drug}) on bacterial levels in a granuloma after treatment with TNF inhibitors of Class 1, 2 and 3, respectively. The black, red and green stars locate TNF inhibitors with TNF binding kinetics similar to etanercept, infliximab and adalimumab, respectively. (D), (E) Effect of variation of TNF/drug binding rate constant (k_{on_TNF/Drug}) on bacterial levels in a granuloma at large and small unbinding rate constants (large: k_{off_TNF/Drug} = 2×10⁻³ s⁻¹, small: k_{off_TNF/Drug} = 6.3×10⁻⁵ s⁻¹), respectively. (F) Effect of variation of TNF/drug binding rate constant (k_{on_TNF/Drug}) on bacterial levels in a granuloma at a constant drug affinity for TNF (K_{d_Drug} = k_{off_TNF/Drug} / k_{on_TNF/Drug} = 2×10⁻⁹ M). Simulations are run with drug blood concentrations of C_p = 1.25×10⁻⁸ M and vascular permeability coefficient of k_c = 1.1×10⁻⁷ cm/s representing an approximately 50% drug permeability in tissue. Simulation results are averaged over 5 runs. Error bars represent standard deviations.

Figure 3

Comparison of effects of etanercept, infliximab and adalimumab on bacterial numbers and granuloma snapshots at different blood concentrations and vascular permeability coefficients (k_c). (A) Effect of permeability coefficient variations on bacterial numbers within a granuloma for infliximab, etanercept, and adalimumab. Results are shown for drug-specific blood concentrations, corresponding to doses administered in RA patients (see Table II). Vascular permeability coefficients of 10⁻⁹–10⁻⁶ cm/s correspond to approximately 1%–90% drug permeability levels from blood into tissue. Simulation results are averaged over 10 runs. Error bars represent standard deviations. (B) Granuloma snapshot for a scenario of containment in the absence of TNF inhibitor. (C), (D) Granuloma snapshots 200 days post-infection for 100 day etanercept treatment for k_c = 1.1×10⁻⁸ cm/s and k_c = 1.1×10⁻⁷ cm/s, respectively. (E), (F) Granuloma snapshots 200 days post-infection for 100 day infliximab treatment for k_c = 1.1×10⁻⁸ cm/s and k_c = 1.1×10⁻⁷ cm/s, respectively. Cell types and status are shown by different color squares, as indicated in the bottom left corner of the figure (M_r: resting macrophage, M_i: infected macrophage, M_ci: chronically infected macrophage, M_a: activated macrophage, B_e: extracellular bacteria, T_γ: pro-inflammatory IFN-γ producing T cell, T_c: cytotoxic T cell, T_reg: regulatory T cell). Caseation and vascular sources are also indicated.

Figure 4

Effect of infliximab-induced cell death as a result of binding to mTNF on a granuloma at 100 days after anti-TNF treatment. (A) Bacterial levels within a granuloma controlling infection in the absence of infliximab and in the presence of infliximab with low and high vascular permeabilities (low: k_c = 1.1×10⁻⁸ cm/s, high: k_c = 1.1×10⁻⁷ cm/s) with or without apoptotic and cytolytic activities and at different concentration thresholds for drug-induced cell death (τ_death-Drug). (B)–(D) Levels of TNF and drug-induced cell death for T cells, activated macrophages (M_a) and infected and chronically infected macrophages (M_i and M_ci), respectively. Cell death numbers does not include death events induced by factors other than TNF and drug. Infliximab’s ability to induce apoptosis and cytolysis significantly contributes, at low and high drug permeabilities, to death of T cells, and only at high permeabilities to death of activated and infected macrophages. At low drug permeabilities, there is no statistically significant difference between activated and infected macrophage death with or without apoptotic and cytolytic activities of the drug. Simulation results are averaged over 10 runs. Error bars represent standard deviations.

Figure 5

Effect of pharmacokinetic (PK) fluctuations in the blood concentration of infliximab and variation of tissue half-life of infliximab on free drug concentration and bacterial levels within a granuloma. (A) The mono-exponential PK model with a first order elimination for blood concentration of infliximab in RA patients at a 3 mg/kg dose level as presented by (39), compared with an estimated steady state concentration. The PK model represents a loading period with infliximab infusions at weeks 0, 2 and 6, and then infusions every 8 weeks. (B) Dynamics of the average free infliximab concentration within a granuloma following anti-TNF treatment for different values of permeability coefficient (small k_c = 1.1×10⁻⁸ cm/s, large k_c = 1.1×10⁻⁷ cm/s). (C) Dynamics of bacteria numbers within a granuloma following anti-TNF treatment. (D) Bacterial levels within a granuloma in the absence of infliximab (containment baseline) and in the presence of infliximab at low and high vascular permeabilities (small k_c = 1.1×10⁻⁸ cm/s, large k_c = 1.1×10⁻⁷ cm/s) and different tissue half-lives (half-life of 4 days: k_{deg_Drug} = 2×10⁻⁶ s⁻¹, half-life of 8 days: k_{deg_Drug} = 1×10⁻⁶ s⁻¹, half-life of 12 days: k_{deg_Drug} = 5.35×10⁻⁵ s⁻¹) 300 days post-infection. Anti-TNF treatments are initiated at day 100 post-infection. Simulation results are averaged over 10 runs. Error bars represent standard deviations.

Figure 6

Sensitivity analysis results for the effect of cellular/tissue scale and TNF-associated molecular scale parameters on model outcomes in the presence of TNF-neutralizing drugs: etanercept and infliximab. Important cellular/tissue scale parameters are identified to be: chemokine degradation rate constant δ_chem, probability of T cell moving onto a macrophage-containing location T_moveM, TNF/chemokine concentration threshold for T_γ recruitment τ_recTgam, probability of T cell recruitment T_recr, and intracellular Mtb growth rate α_Bi. Important TNF-associated parameters include: sTNF degradation rate constant δ_TNF, mTNF synthesis rate for macrophages k_synthMac, mTNF synthesis rate for T cells k_synthTcell, TACE activity rate constant for macrophages k_TACEMac, equilibrium dissociation constant of sTNF/TNFR1 K_d1, apoptosis rate constant k_apop, rate constant for TNF-induced NF-κB activation in macrophages k_NFκB, and cell surface sTNF/TNFR1 threshold for TNF-induced NF-κB activation τ_NFκB. +/− signs show positive/negative correlations. Color intensities show the significance of correlations based on p-values. Significant correlation coefficient values are shown in Supplementary Tables 1, 2. White squares show non-significant correlations.