A cognitive neural circuit biotype of depression showing functional and behavioral improvement after transcranial magnetic stimulation in the B-SMART-fMRI trial

Abstract

We previously identified a cognitive biotype of depression characterized by treatment resistance, impaired cognitive control behavioral performance and dysfunction in the cognitive control circuit, comprising the dorsolateral prefrontal cortex (dLPFC) and dorsal anterior cingulate cortex (dACC). Therapeutic transcranial magnetic stimulation (TMS) to the left dLPFC is a promising option for individuals whose depression does not respond to pharmacotherapy. Here, 43 veterans with treatment-resistant depression were assessed before TMS, after early TMS and post-TMS using functional magnetic resonance imaging during a Go-NoGo paradigm, behavioral cognitive control tests and symptom questionnaires. Stratifying veterans at baseline based on task-evoked dLPFC-dACC connectivity, we demonstrate that TMS-related improvement in cognitive control circuit connectivity and behavioral performance is specific to individuals with reduced connectivity at baseline (cognitive biotype +), whereas individuals with intact connectivity at baseline (cognitive biotype -) did not demonstrate significant changes. Our findings show that dLPFC-dACC connectivity during cognitive control is both a promising diagnostic biomarker for a cognitive biotype of depression and a response biomarker for cognitive improvement after TMS applied to the dLPFC.

Fig. 1 |. Stratification and validation of the cognitive biotype groups defined by functional connectivity within the cognitive control circuit.

a, Task fMRI data were collected during a Go–NoGo task that requires response inhibition and is designed to engage the cognitive control circuit. ISI, interstimulus interval. b, From the scan we quantified connectivity to define regions of the cognitive control circuit, the left dLPFC and dACC, for the contrast of NoGo>Go (top). Left dLPFC–dACC connectivity was quantified for individual participants and connectivity scores were expressed in s.d. units relative to a healthy reference dataset. Cognitive behavioral performance was assessed using the same Go–NoGo task conducted outside of the scanner (bottom). Cognitive performance was quantified as the composite of commission and omission errors, and also expressed in s.d. units relative to a healthy reference dataset. c, We then stratified the sample into two biotypes based on their pretreatment baseline left dLPFC–dACC connectivity. The cognitive biotype (n = 26) was defined as participants with hypoconnectivity (<0) relative to a healthy norm, which we termed cognitive biotype + and those not in the cognitive biotype (n = 17) were defined as participants with intact connectivity (≥0), which we termed cognitive biotype − (top). d, Stratification on the basis of baseline cognitive control circuit connectivity was validated by biotype differences in baseline cognitive performance. Veterans in the cognitive biotype + group (n = 26) were distinguished from cognitive biotype − (n = 17) by significantly poorer accuracy for the composite measure of commission and omission errors (cognitive biotype + versus cognitive biotype −; t = −2.127, P = 0.041, 95% CI = [−2.003; −0.042]). The mean is shown as bars, the s.e.m. are shown as whiskers and individual data points are shown as dots.

Fig. 2 |. TMS restores impaired cognitive control circuit connectivity and cognitive control behavioral performance in a session-dependent manner.

a–d, Changes in cognitive control circuit connectivity (a,b) and behavioral performance (c,d) were analyzed using a linear mixed-effects model to test for the effects of biotype, session and their interaction while adjusting for time (in days) from baseline on the scan date and repeated measure effects. Data are mean and s.e.m. a, TMS produced an early improvement in cognitive control circuit connectivity, which was dependent on the pre-TMS stratification (biotype by session interaction, F = 7.485, d.f. = 2, P = 0.001). This early improvement was specific to the veterans in the cognitive biotype and was present both early after TMS and at the end of treatment (early versus baseline, t = 4.784, P < 0.001, d = 1.114, 95% CI = [0.773; 1.876]; post treatment versus baseline, t = 3.347, P = 0.001, d = 0.675, 95% CI = [0.661; 2.585]). b, Data for individual veterans in each of the two cognitive biotypes, indicated by individual data points and connected by faint lines. c, Cognitive control behavioral performance of both biotypes, mirroring the findings in a. We observed early TMS-induced improvement in cognitive control performance that was specific to the cognitive biotype + group (biotype by session interaction, F = 4.178, d.f. = 2, P = 0.021) and this improvement was present both early after TMS and at the end of treatment (early versus baseline, t = 3.862, P < 0.001, d = 1.075, 95% CI = [0.513; 1.622]; post treatment versus baseline, t = 3.177, P = 0.002, d = 0.767, 95% CI = [0.764; 3.343]). d, Go–NoGo performance values for individual veterans in both biotypes, indicated as data points and connected by faint lines. All statistical tests were two-sided and not adjusted for multiple comparisons.

Fig. 3 |. Restoration of cognitive control circuit connectivity by TMS correlates with improvement of cognitive control performance.

Changes in cognitive control circuit connectivity across sessions correlated positively with changes in cognitive control performance (r = 0.299, two-sided P = 0.004).

Fig. 4 |. Causal mediation model linking biotype, changes from baseline in cognitive control circuit connectivity and changes from baseline in cognitive control performance following TMS.

To determine whether the changes in cognitive control performance were driven by those in functional connectivity, we built a causal mediation model. In this model the independent variable was biotype (factor: cognitive biotype − and cognitive biotype +), the mediator variable was change in left dLPFC–dACC NoGo>Go connectivity from baseline and the dependent variable was change in cognitive control performance from the baseline. Average causal mediation effects and average direct effects were considered significant when P < 0.05. All statistical tests were two-sided and not adjusted for multiple comparisons. Δ, Change.

Fig. 5 |. TMS reduces depression severity.

a,b, Depression severity was measured pretreatment, post early treatment, and post treatment using the quick inventory of depression symptoms (QIDS) total score. Changes in depression severity were analyzed using a linear mixed-effects model to test for the effects of biotype, session and their interaction while adjusting for time from baseline on the scan date and repeated measure effects. The thick lines indicate the mean of depression severity and the error bars the s.e.m. a, TMS reduced depression severity of individuals in both cognitive biotypes (main effect of session, F = 4.366, d.f. = 2, P = 0.016). Depression severity significantly decreased from baseline to early post treatment (early versus baseline, t = −2.758, d.f. = 75, P = 0.007, 95% CI = [−3.24; −0.523]) and remained lower post treatment (post treatment versus baseline t = −2.337, d.f. = 103, P = 0.021, 95% CI = [−6.230; −0.510]). b, Data for individual veterans in each of the two cognitive biotypes, indicated by individual data points and connected by faint lines. All statistical tests were two-sided and not adjusted for multiple comparisons.