A randomized proof-of-mechanism trial of TNF antagonism for motivational deficits and related corticostriatal circuitry in depressed patients with high inflammation

Abstract

Chronic, low-grade inflammation has been associated with motivational deficits in patients with major depression (MD). In turn, impaired motivation has been linked to poor quality of life across psychiatric disorders. We thus determined effects of the anti-inflammatory drug infliximab-a potent tumor necrosis factor (TNF) antagonist-on behavioral and neural measures of motivation in 42 medically stable, unmedicated MD patients with a C-reactive protein >3 mg/L. All patients underwent a double-blind, placebo-controlled, single-dose, randomized clinical trial with infliximab (5 mg/kg) versus placebo. Behavioral performance on an effort-based decision-making task, self-report questionnaires, and neural responses during event-related functional magnetic resonance imaging were assessed at baseline and 2 weeks following infusion. We found that relative to placebo, patients receiving infliximab were more willing to expend effort for rewards. Moreover, increase in effortful choices was associated with reduced TNF signaling as indexed by decreased soluble TNF receptor type 2 (sTNFR2). Changes in effort-based decision-making and sTNFR2 were also associated with changes in task-related activity in a network of brain areas, including dorsomedial prefrontal cortex (dmPFC), ventral striatum, and putamen, as well as the functional connectivity between these regions. Changes in sTNFR2 also mediated the relationships between drug condition and behavioral and neuroimaging measures. Finally, changes in self-reported anhedonia symptoms and effort-discounting behavior were associated with greater responses of an independently validated whole-brain predictive model (aka "neural signature") sensitive to monetary rewards. Taken together, these data support the use of anti-inflammatory treatment to improve effort-based decision-making and associated brain circuitry in depressed patients with high inflammation.

Figure 1:

A. Schematic diagram of study design. All participants were screened for eligibility criteria, and eligible participants completed baseline symptom assessments and laboratory blood work. A baseline fMRI scanning session was followed by randomization and assessments at three days and at 14 days, when participants returned for a second fMRI session. Optional 24-hour and 7-day symptom and blood sample assessments were also collected (not shown). B. Schematic diagram of our previously developed EBDM task^, with descriptions of parametric modulator contrasts used at either the Cue 1 or Cue 2 epoch of each trial (see Online Methods for details). Abbreviations. H&P: history and physical; SCID: structured clinical diagnostic interview; SV: subjective value; SVPE: subjective value prediction error.

Figure 2:

Effects of infliximab on effort-based decision-making. A. Group × Time × Effort level interaction in EBDM task performance. Relative to the placebo group, patients receiving infliximab accepted more effortful options at the two highest effort levels at 14 days. B. Group × Time interaction on effort discounting. Relative to the placebo group, patients receiving infliximab showed a reduction in effort discounting (indicated by the parameter k) at 14 days. C. A reduction in sTNFR2 at 14 days was associated with increased willingness to accept effortful options at the highest effort level (100%). D. A reduction in sTNFR2 at 14 days was associated with a reduction in effort discounting (k). These associations mediated the relationship between drug condition and effort performance. E. Changes in the MDFI Reduced Motivation Subscale for each group, with lower scores indicating greater improvement. F. Changes in the IDS-SR anhedonia subscale for each group, with lower scores indicating greater improvement. Abbreviations: MDFI – multidimensional fatigue inventory; IDS-SR – Inventory for Depressive Symptomatology Self Report; sTNFR2 – soluble TNF receptor 2; EBDM – Effort based decision-making.

Figure 3:

Associations between dmPFC activation and change in inflammatory markers, effort discounting and subjective value encoding at 100% effort. A. Larger reductions in sTNFR2 at 14 days were associated with greater reductions in dmPFC subjective value encoding. Top: whole-brain contrast. Middle: shows scatter plot from dmPFC a priori ROI. Bottom: Change in mean dmPFC signal binned by change in sTNFR2 B. Larger reductions in dmPFC subjective value encoding at 14 days was associated with a reduction in effort-discounting. Top: whole-brain contrast. Middle: shows scatter plot from dmPFC a priori ROI. Bottom: Change in mean dmPFC signal binned by change in sTNFR2 C. Larger reductions in dmPFC subjective value encoding at 14 days was associated with a greater increase in willingness to accept effortful options at the highest effort level (100%). Top: whole-brain contrast. Middle: shows scatter plot from dmPFC a priori ROI. Bottom: Change in mean dmPFC signal binned by change sTNFR2 D. Conjunction of whole-brain maps for ΔsTNFR2, Δk, and Δ100% effort. E. Conjunction of whole-brain maps with our a priori ROI. F. Change in sTNFR2 in dmPFC subjective value encoding mediated the relationship between drug and dmPFC SVC encoding. Abbreviations: SVC – SVchosen; sTNFR2 – TNF soluble receptor 2; dmPFC; dorsomedial prefrontal cortex.

Figure 4:

Multivariate classification of drug condition from striatal activity and effects of infliximab on network connectivity. A. Receiver operating characteristic (ROC) curves for multivariate classification of drug condition based on regional striatal encoding of predicted subjective value at Cue 1. Estimated pattern of activity in ventral striatum shown in inlay. B.Scatter plots showing the association between change in sTNFR52 and change in dmPFC-putamen beta-series correlation (BSC) during Cue1 of the EBDM task when effort information was presented (top) or reward information was presented (bottom). C. Schematic diagrams of mediation models tested. Depiction of dmPFC and putamen ROIs shown in inlay. In both cases, change in sTNFR2 significantly mediated the association between drug condition and change in dmPFC-putamen BSC. NB: BSC analyses between dmPFC-ventral striatum and dmPFC-caudate were also tested (not shown). Abbreviations: AUC = area under the curve; BSC = beta-series correlation; VS = ventral striatum; dmPFC = dorsomedial prefrontal cortex.

Figure 5:

Results of neural signature analyses. A. Cosine similarity between the reward signature (blue) and a control signature for picture-induced negative emotions (“PINES”) (green) for low and high subjective value trials during the EBDM task in a sample of 47 healthy controls (see Supplemental Materials for details). Dark shaded area shows SEM and light shaded area shows standard deviation. B. Schematic illustration of cosine similarity analysis between change in whole-brain activity during Cue 2 of the EBDM task between 14-day and baseline scans and pre-trained “Reward Signature”. C. Association of cosine similarity with change in anhedonia (left) and change in 100% Effort choices (right).