Advertisement

DLPFC stimulation alters working memory related activations and performance: An interleaved TMS-fMRI study

Open AccessPublished:May 26, 2022DOI:https://doi.org/10.1016/j.brs.2022.05.014

      Highlights

      • TMS increased high cognitive load related DMN deactivations.
      • FPN node activations were amplified in high versus low cognitive load TMS.
      • DMN node activations were attenuated in high versus low cognitive load TMS.
      • TMS delivered during high load enhanced cognitive processing.

      Abstract

      Background

      Findings from correlative neuroimaging studies link increased frontoparietal network (FPN) activation and default mode network (DMN) deactivation to enhanced high cognitive demand processing. To causally investigate FPN-DMN contributions to high cognitive demand processing, the current interleaved TMS-fMRI study simultaneously manipulated and indexed neural activity while tracking cognitive performance during high and low cognitive load conditions.

      Methods

      Twenty participants completed an n-back task consisting of four conditions (0-back, 0-backTMS, 2-back, 2-backTMS) while undergoing interleaved TMS-fMRI. During TMS concurrent with n-back blocks, TMS single pulses were delivered to the left DLPFC at 100% motor-threshold every 2.4s.

      Results

      TMS delivered during high cognitive load strengthened cognitive processing. FPN node activations and DMN node deactivations were increased in the high versus low cognitive load TMS condition. Contrary to our hypothesis, TMS did not increase high load related activation in FPN nodes. However, as hypothesized, increased DMN node deactivations emerged as a function of TMS during high load (right angular gyrus) and from interactions between cognitive load and TMS (right middle temporal gyrus). Load and TMS combined to dampen activation within the DMN at trend level (p = .058). Deactivation in a dorsomedial DMN node was associated with TMS driven improvements in high load cognitive processing.

      Conclusions

      Exogenous perturbation of the DLPFC via single pulse TMS amplified DMN node deactivations and enhanced high cognitive demand processing. Neurobehavioral findings linking these effects hint at a promising, albeit preliminary, cognitive control substrate requiring replication in higher-powered studies that use control stimulation.

      1. Introduction

      Cognitive control refers to the regulation of cognitive and emotional resources in the service of flexible, goal-directed behavior [
      • Braver T.S.
      The variable nature of cognitive control: a dual mechanisms framework.
      ]. Abnormal processing in cognitive control circuits and corresponding behavioral deficits have been reliably detected in several neuropsychiatric disorders, implicating cognitive control as a promising transdiagnostic treatment target [
      • McTeague L.M.
      • Goodkind M.S.
      • Etkin A.
      Transdiagnostic impairment of cognitive control in mental illness.
      ,
      • McTeague L.M.
      • Huemer J.
      • Carreon D.M.
      • Jiang Y.
      • Eickhoff S.B.
      • Etkin A.
      Identification of common neural circuit disruptions in cognitive control across psychiatric disorders.
      ].
      The dorsolateral prefrontal cortex (DLPFC), a key cognitive control locus [
      • Niendam T.A.
      • Laird A.R.
      • Ray K.L.
      • Dean Y.M.
      • Glahn D.C.
      • Carter C.S.
      Meta-analytic evidence for a superordinate cognitive control network subserving diverse executive functions.
      ], has been targeted in the vast majority of therapeutic repetitive transcranial magnetic stimulation (rTMS) studies [
      • McLean A.L.
      Publication trends in transcranial magnetic stimulation: a 30-year panorama.
      ] and is the FDA approved site for rTMS for depression. This large and ever growing evidence base demonstrates the capacity of rTMS targeted to left dlPFC to remediate multiple neuropsychiatric symptoms [
      • Cao X.
      • Deng C.
      • Su X.
      • Guo Y.
      Response and remission rates following high-frequency vs. low-frequency repetitive transcranial magnetic stimulation (rTMS) over right DLPFC for treating major depressive disorder (MDD): a meta-analysis of randomized, double-blind trials.
      ,
      • Liao X.
      • Li G.
      • Wang A.
      • Liu T.
      • Feng S.
      • Guo Z.
      • et al.
      Repetitive transcranial magnetic stimulation as an alternative therapy for cognitive impairment in Alzheimer's disease: a meta-analysis.
      ,
      • Nguyen T.D.
      • Hieronymus F.
      • Lorentzen R.
      • McGirr A.
      • Østergaard S.D.
      The efficacy of repetitive transcranial magnetic stimulation (rTMS) for bipolar depression: a systematic review and meta-analysis.
      ,
      • Vicario C.
      • Salehinejad M.A.
      • Felmingham K.
      • Martino G.
      • Nitsche M.
      A systematic review on the therapeutic effectiveness of non-invasive brain stimulation for the treatment of anxiety disorders.
      ], potentially by upregulating cognitive control [
      • Demirtas-Tatlidede A.
      • Vahabzadeh-Hagh A.M.
      • Pascual-Leone A.
      Can noninvasive brain stimulation enhance cognition in neuropsychiatric disorders?.
      ,
      • Lantrip C.
      • Gunning F.M.
      • Flashman L.
      • Roth R.M.
      • Holtzheimer P.E.
      Effects of transcranial magnetic stimulation on the cognitive control of emotion: potential antidepressant mechanisms.
      ,
      • Yang L.L.
      • Zhao D.
      • Kong L.L.
      • Sun Y.Q.
      • Wang Z.Y.
      • Gao Y.Y.
      • et al.
      High-frequency repetitive transcranial magnetic stimulation (rTMS) improves neurocognitive function in bipolar disorder.
      ].
      Accumulating evidence suggests that pairing stimulation with cognitive tasks/interventions that putatively engage prefrontal circuits may augment treatment response (for review, see: 13). However, despite the promising results emerging from this developing literature, the neurophysiological mechanisms undergirding this potential synergistic effect remain unclear. One compelling possibility is that TMS and cognitive tasks/interventions converge on an overlapping neural substrate, with TMS amplifying cognitive task/intervention driven effects. This possibility is supported by evidence demonstrating that both high cognitive demand and canonical ‘excitatory’ DLPFC TMS protocols (depolarizing single-pulse or high-frequency rTMS) modulate activation within and coordination between the frontoparietal network (FPN) and default mode (DMN) networks [
      • Chen A.C.
      • Oathes D.J.
      • Chang C.
      • Bradley T.
      • Zhou Z.W.
      • Williams L.M.
      • et al.
      Causal interactions between fronto-parietal central executive and default-mode networks in humans.
      ,
      • Denkova E.
      • Nomi J.S.
      • Uddin L.Q.
      • Jha A.P.
      Dynamic brain network configurations during rest and an attention task with frequent occurrence of mind wandering.
      ,
      • Philip N.S.
      • Barredo J.
      • van’t Wout-Frank M.
      • Tyrka A.R.
      • Price L.H.
      • Carpenter L.L.
      Network mechanisms of clinical response to transcranial magnetic stimulation in posttraumatic stress disorder and major depressive disorder.
      ,
      • Shang Y.
      • Chang D.
      • Zhang J.
      • Peng W.
      • Song D.
      • Gao X.
      • et al.
      Theta-burst transcranial magnetic stimulation induced functional connectivity changes between dorsolateral prefrontal cortex and default-mode-network.
      ]. Specifically, high cognitive demand and ‘excitatory’ DLPFC TMS protocols have generally been shown to increase FPN activation/decrease FPN intracortical inhibition [
      • Wang H.
      • He W.
      • Wu J.
      • Zhang J.
      • Jin Z.
      • Li L.
      A coordinate-based meta-analysis of the n-back working memory paradigm using activation likelihood estimation.
      ,
      • Eshel N.
      • Keller C.J.
      • Wu W.
      • Jiang J.
      • Mills-Finnerty C.
      • Huemer J.
      • et al.
      Global connectivity and local excitability changes underlie antidepressant effects of repetitive transcranial magnetic stimulation.
      ], increase DMN deactivation/weaken DMN connectivity [
      • Gu H.
      • Hu Y.
      • Chen X.
      • He Y.
      • Yang Y.
      Regional excitation-inhibition balance predicts default-mode network deactivation via functional connectivity.
      ,
      • Cui H.
      • Ren R.
      • Lin G.
      • Zou Y.
      • Jiang L.
      • Wei Z.
      • et al.
      Repetitive transcranial magnetic stimulation induced hypoconnectivity within the default mode network yields cognitive improvements in amnestic mild cognitive impairment: a randomized controlled study.
      ], and weaken connectivity between the FPN and DMN [
      • Chen A.C.
      • Oathes D.J.
      • Chang C.
      • Bradley T.
      • Zhou Z.W.
      • Williams L.M.
      • et al.
      Causal interactions between fronto-parietal central executive and default-mode networks in humans.
      ,
      • Cai W.
      • Ryali S.
      • Pasumarthy R.
      • Talasila V.
      • Menon V.
      Dynamic causal brain circuits during working memory and their functional controllability.
      ,
      • Liston C.
      • Chen A.C.
      • Zebley B.D.
      • Drysdale A.T.
      • Gordon R.
      • Leuchter B.
      • et al.
      Default mode network mechanisms of transcranial magnetic stimulation in depression.
      ]. FPN/DMN activation and connectivity changes associated with high cognitive load and TMS have been linked to working memory performance [
      • Keller J.B.
      • Hedden T.
      • Thompson T.W.
      • Anteraper S.A.
      • Gabrieli J.D.
      • Whitfield-Gabrieli S.
      Resting-state anticorrelations between medial and lateral prefrontal cortex: association with working memory, aging, and individual differences.
      ,
      • Anticevic A.
      • Repovs G.
      • Shulman G.L.
      • Barch D.M.
      When less is more: TPJ and default network deactivation during encoding predicts working memory performance.
      ] and clinical TMS response [
      • Philip N.S.
      • Barredo J.
      • van’t Wout-Frank M.
      • Tyrka A.R.
      • Price L.H.
      • Carpenter L.L.
      Network mechanisms of clinical response to transcranial magnetic stimulation in posttraumatic stress disorder and major depressive disorder.
      ,
      • Liston C.
      • Chen A.C.
      • Zebley B.D.
      • Drysdale A.T.
      • Gordon R.
      • Leuchter B.
      • et al.
      Default mode network mechanisms of transcranial magnetic stimulation in depression.
      ,
      • Fox M.D.
      • Buckner R.L.
      • White M.P.
      • Greicius M.D.
      • Pascual-Leone A.
      Efficacy of transcranial magnetic stimulation targets for depression is related to intrinsic functional connectivity with the subgenual cingulate.
      ,
      • Hopman H.
      • Chan S.
      • Chu W.C.W.
      • Lu H.
      • Tse C.Y.
      • Chau S.W.H.
      • et al.
      Personalized prediction of transcranial magnetic stimulation clinical response in patients with treatment-refractory depression using neuroimaging biomarkers and machine learning.
      ,
      • Weigand A.
      • Horn A.
      • Caballero R.
      • Cooke D.
      • Stern A.P.
      • Taylor S.F.
      • et al.
      Prospective validation that subgenual connectivity predicts antidepressant efficacy of transcranial magnetic stimulation sites.
      ], highlighting the functional significance of this substrate and suggesting a potential convergent pathway by which rTMS may alter cognition and mood [
      • Lantrip C.
      • Gunning F.M.
      • Flashman L.
      • Roth R.M.
      • Holtzheimer P.E.
      Effects of transcranial magnetic stimulation on the cognitive control of emotion: potential antidepressant mechanisms.
      ].
      To date, relationships between FPN and DMN activations, connectivity, and cognitive control have primarily been assessed via correlative neuroimaging methods. Although pre-post rTMS neuroimaging findings lend additional support to this work, pre-post designs cannot capture immediate and dynamic stimulation induced changes in neural activity/connectivity. Tools that can simultaneously manipulate and measure neural activity/connectivity are essential for shedding causal light on activation within and interactions between the FPN and DMN and their functional significance. Interleaved TMS-fMRI is a well validated technique capable of delivering exogenous probes to discrete cortical targets and indexing the effects of these probes on local circuits and distributed networks. Applying TMS-fMRI during systematic cognitive control paradigms could causally elucidate how FPN and DMN activation and connectivity support cognitive control.
      Given the centrality of cognitive control deficits to neuropsychiatric illness and correlative evidence that FPN and DMN activation and connectivity shape cognitive control, the present investigation applied interleaved TMS-fMRI during a working memory task to examine how causally probing the FPN with single pulse TMS alters concurrent working memory related activations and performance. Consistent with previous work, we hypothesized that [
      • Braver T.S.
      The variable nature of cognitive control: a dual mechanisms framework.
      ]: TMS would enhance high cognitive load related FPN activations and DMN deactivations and reduce connectivity between the FPN and DMN; and that [
      • McTeague L.M.
      • Goodkind M.S.
      • Etkin A.
      Transdiagnostic impairment of cognitive control in mental illness.
      ] FPN activation, DMN deactivation, and increased anti-correlations between the FPN and DMN would be linked to TMS related improvements in high cognitive load performance.

      2. Methods and materials

      2.1 Subjects

      Twenty healthy, right-handed volunteers between the ages of 22 and 38 were included in the current study. Demographic information is presented in Supplementary Table 1. Handedness was assessed with the Annett Handedness Scale [
      • Annett M.
      A classification of hand preference by association analysis.
      ]. A modified structured clinical interview (SCID) [
      • First M.B.
      • Gibbon M.
      The structured clinical interview for DSM-IV Axis I disorders (SCID-I) and the structured clinical interview for DSM-IV Axis II disorders (SCID-II).
      ] was performed to screen for psychopathology. To meet inclusion criteria, subjects could not be on any psychotropic medications and had to have a negative urine drug screen. The Hamilton Depression Rating Scale (HRSD17) [
      • Hamilton M.
      A rating scale for depression.
      ] and Inventory of Depressive Symptomatology Self Report (IDS-SR) [
      • Rush A.J.
      • Giles D.E.
      • Schlesser M.A.
      • Fulton C.L.
      • Weissenburger J.
      • Burns C.
      The inventory for depressive symptomatology (IDS): preliminary findings.
      ] were used for screening; to be included, subjects could not have scores exceeding 8 and 14 on the scales, respectively. All subjects provided written informed consent in accordance with the guidelines of the Institute of Psychiatry and the Internal Review Board at the Medical University of South Carolina (MUSC).

      2.2 N-back verbal working memory paradigm

      We used a block design with each discrete block presented pseudo-randomly (see Fig. 2). In each block, 12 upper and lower-case letter stimuli were visually presented at an inter-stimulus interval of 1.5 s for a total of 18 s, preceded by 6 s of instructions. There were 4 conditions, two delivered in a conventional fMRI n-back task format: ‘0-back’, ‘2-back’; and two similar n-back blocks with simultaneous single pulses of TMS: ‘0-back + TMS’, and ‘2-back + TMS’. In the low cognitive load condition (0-back, ‘0-back + TMS’), subjects were asked to press a button with their right index finger when the letter “X” appeared. In the high cognitive load condition (2-back, ‘2-back + TMS’), subjects were asked to press a button with their right index finger if the presented letter was the same as the letter presented 2-trials previously. All conditions were matched for the number of target letters (3 per block, 9 total) presented. The conditions were presented in the following fixed order: 0, 0 + TMS, 2, 2 + TMS, 0 + TMS, 0, 2 + TMS, 2, 0, 2, 0 + TMS, 2 + TMS, with three repetitions of each level of difficulty, for a total duration of 4 min 48 s. Subject performance during scanning was recorded including accuracy and response time (RT) to target letters. Immediately before scanning, all subjects received task training with a unique set of stimuli to ensure that they fully understood task requirements. Stimuli were presented using E-Prime version 2.0 (Psychology Software Tools, Inc, Pittsburgh, PA) and Integrated Functional Imaging System (IFIS; MRI Devices Inc.) to record behavioral data.

      2.3 Behavioral analyses

      Data from 4 subjects were excluded from behavioral analyses due to malfunctions with the data-collection software. Additionally, to ensure maximum continuity between neurobiological and behavioral analyses, we excluded behavioral data from a participant who was excluded from fMRI analyses due to excessive motion (final N = 15). Following tests confirming normality (Shapiro-Wilk's test) and homoscedasticity (Levene's test), we used repeated measures ANOVA to examine whether working memory performance (accuracy and response time) differed as a function of cognitive load, TMS, and/or their interaction. Significant ANOVA results were followed up using Bonferroni corrected, two-tailed t-tests.
      To account for both missed targets and incorrect responses to non-target items, accuracy % was calculated as the number of correct target ‘hits’ minus the number of incorrect non-target responses divided by the total number of target items. Per condition mean values of correct and incorrect responses as well as missed targets are included in Table 1. Because miss and incorrect response data were non-normally distributed, we used Bonferroni corrected Wilcoxon Signed Rank Tests to assess for differences between conditions. Statistical thresholds were set at p < .05. Significant ANOVA results were followed up using Bonferroni corrected, two-tailed t-tests.
      Table 1Behavioral results.
      Subjects (N = 15)Mean # correct responses (Std)Accuracy % (Std)Mean # incorrect responsesMean # missesRT (Std)
      0-back8.93 (.26)99.26 (2.87)0.07 (.26)455.12 (54.82)
      0-backTMS8.6 (.91)90.37 (13.84).47 (.52).4 (.91)518.41 (91.02)
      2-back6.6 (.99)63.7 (12.92).87 (.83)2.4 (.99)542.11 (119.29)
      2-backTMS8.4 (.74)88.89 (11.11).4 (.51).6 (.74)521.12 (103.81)
      2-back accuracy rates were significantly reduced relative to all other conditions (2-back vs. 0-back, adjusted p < .001; 2-back vs. 0-backTMS, adjusted p = < .001; 2-back vs. 2-backTMS, adjusted p < .001). 2-back RT was prolonged compared to 0-backRT (adjusted p = .075) at trend level.

      2.4 fMRI data collection

      Single-shot, gradient echo echoplanar imaging was used to acquire 120 T2∗-weighted image volumes on a neuro-optimized 3T Siemens scanner (Siemens AG, Erlangen, Bavaria, Germany) with a 12 channel head coil. Volumes were collected with the following parameters: TR = 2400 ms which included a 150 ms delay, whereby the first 2250 ms were dedicated to image acquisition after which there was a 150 ms delay until image acquisition commenced again. The TMS pulse was delivered in the middle of this delay so as not to disrupt image acquisition. TE = 35 ms, flip angle 90°, in-plane resolution = 3 × 3x3 mm3 voxels; FOV = 256 × 256, bandwidth = 2874 Hz). There were 32 slices.

      2.5 TMS protocol

      Single pulse TMS was used to interrogate functional networks. During the TMS block, subjects received 7 individual pulses separated by 2.4 s, equal to the TR interval of the fMRI protocol. As previously stated, the pulse was delivered in the middle of the 150 ms delay period between subsequent image acquisitions to avoid corrupting neuroimaging data. With 6 TMS blocks, subjects received a total of 42 pulses. Motor threshold was found by using the PEST [
      • Borckardt J.J.
      • Nahas Z.
      • Koola J.
      • George M.S.
      Estimating resting motor thresholds in transcranial magnetic stimulation research and practice: a computer simulation evaluation of best methods.
      ] method applied to left motor cortex and monitoring for right thumb twitch. We used 100% resting MT for the protocol. A higher output was not used due to increased perceived discomfort of TMS pulses within the MRI. We used the same coil and MagStim200 to measure MT and to deliver the pulses inside the MRI. Stimulation was applied to the left dorsolateral prefrontal cortex, which was estimated using the position of F3 according to the 10–20 EEG system of measurement [
      • Beam W.
      • Borckardt J.J.
      • Reeves S.T.
      • George M.S.
      An efficient and accurate new method for locating the F3 position for prefrontal TMS applications.
      ]. Subjects wore swim caps in order to more accurately mark the measurements and TMS target. A translucent template was centered on the target and aligned with the TMS coil to ensure accurate stimulation of the intended target. The coil was held in place by a specially designed non-ferromagnetic apparatus capable of full 3-dimensional placement and fixation.

      2.5.1 Imaging data pre-processing

      Pre-processing was carried out using AFNI and SPM. MR signal spikes were removed via AFNI 3dDespike using the default settings. Motion correction was performed whereby volumes were aligned to the volume with the fewest outliers. Data were then transformed to MNI space using SPM's template EPI, smoothed using an 8 mm FWHM gaussian kernel and scaled so that each voxel had a mean of 100.

      2.6 Statistical analysis

      The data were analyzed using a general linear model (GLM). Five regressors were used in the GLM, corresponding to blocks for instruction presentation, 0-back, 0-back with TMS, 2-back, and 2-back with TMS. Each condition was modeled as a box-car convolved with a gamma function. Data from subjects who moved more than 2 mm were inspected; three subjects for whom there was visual evidence of spin history artifacts were excluded from the analysis. Higher level neuroimaging analyses were thus carried out on 17 of 20 subjects.
      The beta weights computed for each voxel at the individual subject level analysis served as the dependent variable for the group-level analysis. The following contrasts were examined using AFNI 3dttest++: [2-back – 0-back], [2-backTMS – 2-back], [0-backTMS – 0-back], [2-backTMS – 0-backTMS], and a contrast assessing the interaction between stimulation and cognitive load conditions: [(2-backTMS – 0-backTMS) – (2-back – 0back)], with the statistical threshold set at p < .001. A non-parametric permutation test was implemented within 3dttest++ that estimated the probability of obtaining clusters of a given size that consisted entirely of false positives; cluster sizes that occurred 5% of the time or less were considered to be the minimum significant cluster size (i.e., p < .05).

      3. Results

      3.1 Behavioral results

      As predicted, conditions differed in terms of accuracy (See Table 1, Fig. 1a, Supplementary Fig. 2a). Accuracy % varied as a function of load [F(1,14) = 36.842, p < .001)], TMS condition [F(1,14) = 13.391, p = .003], and their interaction [(F(1,14) = 48.248, p < .001)]. Moreover, Bonferonni corrected Wilcoxon Rank Sum Test results revealed that conditions differed by number of missed targets and incorrect responses. Significantly more missed responses were logged in the 2-back condition compared to all other conditions (0-back vs. 2-back, p = .0074; 0-backTMS vs. 2-back, p = .0032; 2-backTMS vs. 2-back, p = .0081). Moreover, more incorrect responses were made in the 2-back condition compared to the 0-back condition (p = .021), and in the 0-back condition compared to the 0-backTMS condition at trend level (p = .064).
      Fig. 1a
      Fig. 1aLoad and TMS condition interacted significantly to predict accuracy [(F(1,14) = 48.248, p < .001)]. Error bars = 95% confidence interval.
      Post-hoc, Bonferroni corrected two-tailed t-tests (results depicted in Fig. 1a, Fig. 1ba and b and Supplementary Figs. 1a and b) revealed that [
      • Braver T.S.
      The variable nature of cognitive control: a dual mechanisms framework.
      ]: 2-back % accuracy rates were significantly reduced relative to all other conditions (2-back vs. 0-back, adjusted p < .001; 2-back vs. 0-backTMS, adjusted p < .001; 2-back vs. 2-backTMS, adjusted p < .001) and 2-backTMS was greater than 0-back at trend level (adjusted p = .078). Load emerged as a significant predictor of RT [F(1,14) = 7.883, p = .014)]. However, in contrast to accuracy, TMS condition and Load∗TMS were not significant predictors of RT. Regarding individual condition RT comparisons, aside from a trend level increase in 2back compared to 0-back (p = .075), no significant findings emerged for RT.
      Fig. 1b
      Fig. 1bLoad predicted RT [(F(1,14) = 7.883, p = .014)]. TMS condition [(F(1,14) = .607, p = .449)] and the interaction between Load and TMS condition [(F(1,14) = 2.756, p = .119)] did not significantly predict RT. Error bars = 95% confidence interval.

      3.2 fMRI results

      3.2.1 2-Back – 0-back

      Consistent with previous meta-analyses [
      • Owen A.M.
      • McMillan K.M.
      • Laird A.R.
      • Bullmore E.
      N-back working memory paradigm: a meta-analysis of normative functional neuroimaging studies.
      ], the 2-back versus 0-back contrast (Table 2 and Fig. 3) revealed heightened activation in bilateral portions of the dorsolateral prefrontal cortex (DLPFC, [Brodmann Areas (BA) 9, 10, [
      • Beynel L.
      • Davis S.
      • Crowell C.
      • Hilbig S.
      • Lim W.
      • Nguyen D.
      • et al.
      Online repetitive transcranial magnetic stimulation during working memory in younger and older adults: a randomized within-subject comparison.
      ,
      • Bakulin I.
      • Zabirova A.
      • Lagoda D.
      • Poydasheva A.
      • Cherkasova A.
      • Pavlov N.
      • et al.
      Combining HF rTMS over the left DLPFC with concurrent cognitive activity for the offline modulation of working memory in healthy volunteers: a proof-of-concept study.
      ]), dorsomedial prefrontal cortex (DMPFC, [BA6, 8, 9]), dorsal anterior cingulate cortex (dACC, [BA 32]), anterior insula, premotor cortex [BA 6], inferior parietal lobule [BA40], superior parietal lobule [BA7], and precuneus. Deactivations emerged in bilateral portions of the anterior ventromedial prefrontal cortex (vmPFC), [BA10]) and posterior cingulate cortex [BA23, 29, 30, 31].
      Table 2Neurobiological results for 2-back vs. 0-back.
      Peak Voxel
      Anatomical RegionsVoxelsL/RMNI Coordinatest value
      xyz
      Cluster 1455
      PrecuneusR6−67651.21
      L−9−73590.93
      Superior parietal lobuleR13−74560.85
      L−30−69560.98
      Inferior parietal lobuleL−37−66550.96
      Cluster 2369
      Inferior frontal gyrusR5516321.01
      Middle frontal gyrusR4547250.93
      Cluster 3282
      Inferior frontal gyrusL−5026361.02
      Middle frontal gyrusL−4927361.01
      Cluster 4280
      Inferior parietal lobuleR42−60561.02
      Angular gyrusR41−61561.02
      Supramarginal gyrusR44−42500.8
      Superior occipital gyrusR28−76490.92
      Middle occipital gyrusR32−73440.76
      Cluster 5272
      Superior medial gyrusR124470.86
      L−124470.86
      R119490.86
      Supplemental motor areaL−119490.86
      Cluster 6159
      Posterior cingulate cortexR1−4921−0.95
      L−1−5021−0.95
      PrecuneusR1−5120−0.93
      L−1−5321−0.93
      Cluster 7140
      Superior frontal gyrusR16320−0.72
      L−66820−0.93
      Mid orbital gyrusR1612−0.78
      L−8662−0.97
      Cluster 8124
      Superior frontal gyrusL−240740.90
      Precental gyrusL−30−4680.90
      Cluster 983
      Anterior insulaL−34202−0.60
      Cluster 1055
      InsulaR3722−10.70
      Cluster 1144
      Superior frontal gyrusR302711.01
      Cluster 1232
      Superior frontal gyrusL−156529−0.92
      Cluster 1331
      Middle frontal gyrusL−3959110.82
      Cluster 1430
      Superior frontal gyrusL−95347−0.69
      Cluster 1527
      Superior frontal gyrusR3056−40.42
      Cluster 1614
      PrecuneusL−27−85380.57
      Brain areas showing significant activation differences during 2-back vs 0-back. Abbreviations: MNI = Montreal Neurological Institute.
      Fig. 2
      Fig. 2The paradigm consisted of four conditions presented pseudo-randomly in a block-design: ‘0-back’, ‘2-back’; and two similar n-back blocks with simultaneous single pulses of TMS: ‘0-back + TMS’, and ‘2-back + TMS’. During each block, 12 upper and lower-case letter stimuli were visually presented at an inter-stimulus interval of 1.5 s for a total of 18 s, preceded by 6 s of instructions. All conditions were matched for the number of target letters presented. The conditions were presented in the following fixed order: 0, 0 + TMS, 2, 2 + TMS, 0 + TMS, 0, 2 + TMS, 2, 0, 2, 0 + TMS, 2 + TMS, with three repetitions of each level of difficulty, for a total duration of 4 min 48 s. During TMS blocks, 7 pulses were delivered in the middle of the 150 ms delay period between images, every 2.4 s, equal to the TR interval of the fMRI protocol. Abbreviations: TR = repetition time; TMS = transcranial magnetic stimulation.
      Fig. 3
      Fig. 3Neural activations/deactivations for the following contrasts: (A) 2-back vs 0-back; (B) 2-backTMS vs 0-backTMS; (C) 0-backTMS vs 0-back; and (D) i. 2-backTMS vs. 2-back; ii. Interaction contrast. Results are displayed at p < .001 (cluster size ≥10) on the MNI 27 T1 template.

      3.2.2 0-backTMS – 0-back

      The 0-backTMS versus 0-back contrast (Table 3 and Fig. 3) revealed increased activation in bilateral portions of the auditory processing superior temporal gyrus [BA21, 22, 42] – likely driven by stimulation related TMS pulse noise – and middle temporal cortex [BA22]. Additional activations emerged in postcentral gyrus and posterior insula (putatively corresponding to somatic sensations of stimulation), and portions of the left calcarine and right precentral gyrus [BA6]. We did not detect increased activation in the left DLPFC stimulation location. Inspection of functional images revealed MR signal drop-out in the targeted region (see Fig. S1) in several subjects, which may account for this unexpected finding.
      Table 3Neurobiological results for 0-backTMS vs. 0-back.
      Peak Voxel
      Anatomical RegionsVoxelsL/RMNI Coordinatest value
      xyz
      Cluster 1229
      Superior temporal gyrusR66−19140.93
      Middle temporal gyrusR65−2720.56
      Rolandic operculumR65−15180.72
      Heschls gyrusR58−12110.58
      InsulaR44−1580.36
      Cluster 270
      Superior temporal gyrusL−66−25170.74
      Postcentral gyrusL−65−22180.73
      Supramarginal gyrusL−59−30230.44
      Heschls gyrusL−51−18120.56
      Cluster 362
      Superior temporal gyrusL−63−1−10.98
      Middle temporal gyrusL−64−7−30.77
      Rolandic operculumL−59−380.65
      Cluster 453
      Middle temporal gyrusL−66−40110.93
      Superior temporal gyrusL−65−34140.87
      Cluster 551
      Calcarine gyrusL−9−79140.67
      CuneusL−10−79230.46
      Lingual gyrusL−6−70140.60
      Cluster 641
      Temporal poleR638−10.95
      Inferior frontal gyrusR62980.58
      Superior temporal gyrusR58−120.46
      Rolandic operculumR64790.58
      Cluster 724
      Middle frontal gyrusR48−1591.08
      Cluster 823
      Inferior temporal gyrusR60−61−10.71
      Cluster 918
      Inferior frontal gyrusR5711380.74
      Brain areas showing significant activation differences during 0-backTMS vs 0-back. Abbreviations: MNI = Montreal Neurological Institute.

      3.2.3 2-backTMS – 2-back

      The 2-backTMS versus 2-back contrast (Table 4 and Fig. 3) revealed deactivations in the right angular gyrus and middle occipital gyrus. Activation increases in the right superior temporal gyrus (STG) indicative of stimulation induced auditory processing were also detected. Similar to the previous contrast, we did not detect activation in the left DLPFC target.
      Table 4Neurobiological results for 2-backTMS vs. 2-back.
      Peak Voxel
      Anatomical RegionsVoxelsL/RMNI Coordinates
      xyz
      Cluster 113
      Angular gyrusR45−7335−0.55
      Cluster 212
      Temporal poleR60520.49
      Superior temporal gyrusR54−120.49
      Brain areas showing significant activation differences during 2-backTMS vs 2-back. Abbreviations: MNI = Montreal Neurological Institute.

      3.2.4 2-backTMS – 0backTMS

      The 2-backTMS versus 0-backTMS contrast (Table 5 and Fig. 3) revealed increased activation in bilateral portions of the DLPFC [BA9 right side; BA9, 46: left side, posterior to stimulation site], inferior parietal lobule [BA40), and portions of the left precentral gyrus and superior parietal lobule [BA7]. Deactivations emerged in bilateral portions of the anterior ventromedial prefrontal cortex [BA10], posterior cingulate cortex [BA18, 23, 29, 30, 31], precuneus, calcarine, and portions of the left superior medial frontal gyrus [BA8, 10], angular gyrus [BA39], superior temporal gyrus [BA 22], and middle temporal gyrus [BA39].
      Table 5Neurobiological results for 2-backTMS vs. 0-backTMS.
      Peak Voxel
      Anatomical RegionsVoxelsL/RMNI Coordinatest value
      xyz
      Cluster 1549
      Posterior cingulateR3−4914−1.14
      L−1−5015−1.10
      PrecuneusR−1−5321−1.05
      L5−5313−0.89
      CuneusR4−7632−0.55
      L−1−7630−0.60
      CalcarineR1−6317−0.80
      L−1−6215−0.78
      Lingual gyrusR4−6311−0.59
      L−7−618−0.75
      Cluster 2440
      Superior medial gyrusR16426−1.20
      L−9688−1.10
      Superior frontal gyrusR176629−0.78
      L−156529−1.2
      Cluster 3268
      Superior parietal lobuleL−28−69591.20
      Inferior parietal lobuleL−33−66550.89
      Middle occipital gyrusL−25−66430.56
      Angular gyrusL−31−63400.50
      Cluster 4100
      Inferior parietal lobuleR39−43440.74
      Cluster 571
      Inferior frontal gyrusL−5411380.74
      Precentral gyrusL−531470.72
      Cluster 650
      Superior temporal gyrusR59−169−0.45
      Middle temporal gyrusR57−13−10−0.50
      Cluster 747
      Angular gyrusL−44−7539−0.78
      Middle temporal gyrusL−53−7023−0.73
      Cluster 842
      Superior temporal gyrusR−60−113−0.67
      Middle temporal gyrusR−66−132−0.75
      Cluster 941
      HippocampusR42−13−13−0.64
      PutamenR36−16−3−0.34
      InsulaR40−13−4−0.35
      Superior temporal gyrusR46−11−10−0.47
      Cluster 1041
      Inferior frontal gyrusR4529200.51
      Cluster 1140
      Temporal poleR425−22−0.54
      AmygdalaR31−7−20−0.46
      HippocampusR30−9−18−0.51
      Parahippocampal gyrusR28−16−19−0.54
      Cluster 1233
      Inferior frontal gyrusR3323−70.55
      Cluster 1326
      Inferior frontal gyrusL−4532−22−0.8
      Cluster 1418
      Middle frontal gyrusR33−1620.58
      Cluster 1518
      Superior frontal gyrusR278620.5
      Cluster 1616
      Middle frontal gyrusL−272680.67
      Cluster 1715
      Middle temporal gyrusL−60−13−16−0.68
      Cluster 1812
      Middle cingulate gyrusR6−2547−0.5
      Cluster 1912
      Paracentral lobule0−2877−0.7
      Brain areas showing significant activation differences during 2-backTMS vs 0-backTMS. Abbreviations: MNI = Montreal Neurological Institute.

      3.2.5 Interaction contrast

      A contrast comparing effects of cognitive load on neural activity as a function of stimulation versus non-stimulation [(2-backTMS – 0-backTMS) – (2back – 0back)] (Table 6 and Fig. 3) revealed that stimulation induced greater deactivations in discrete portions of the middle and superior temporal gyri [BA21] during high versus low cognitive load.
      Table 6Neurobiological results for interaction contrast.
      Peak Voxel
      Anatomical RegionsVoxelsL/RMNI Coordinates
      xYz
      Cluster 140
      Middle temporal gyrusR60−25−7−0.27
      Superior temporal gyrusR59−240−0.17
      Results for contrast comparing effects of cognitive load on neural activity as a function of stimulation versus non-stimulation. Both areas showed increased deactivation in the high versus low load TMS conditions compared to the corresponding no TMS control conditions. Abbreviations: MNI = Montreal Neurological Institute.
      Abbreviations: MNI: Montreal Neurological Institute.

      3.2.6 DMN/FPN analyses

      To better assess the effect of TMS, load, and their interaction on the DMN and FPN, we plotted beta-values of significant voxels within canonical DMN and FPN networks (see Fig. 3a and b). First, we generated DMN and FPN structural masks based on Talairach coordinates using AFNI's draw plug-in. The DMN included bilateral medial PFC, middle temporal gyri, hippocampus, posterior cingulate cortices, angular gyri, and precunei; the FPN included bilateral F3 and inferior parietal lobules. Next, to isolate significant voxels within these masks, we multiplied each structural mask by a functional mask containing clusters that significantly separated from a mean condition baseline (see Fig. 4).
      Fig. 4
      Fig. 4FPN and DMN beta-values per condition.
      Consistent with voxel-wise results, FPN and DMN activations were amplified and attenuated as a function of high cognitive load, respectively (FPN p < .001, DMN p < .001). Moreover, consistent with a dampening effect of TMS on DMN in high load conditions (i.e., AG reduction in 2-backTMS vs. 2-back; high load by TMS MTG reduction), load and TMS combined at trend level (p = .058) to dampen DMN activation. In contrast, high load and TMS did not combine to strengthen FPN activation, possibly suggestive of a high-load related ceiling effect.
      To assess whether connectivity between the FPN and DMN shifted as a function of cognitive load, TMS, or their combination, we extracted the Pearson's correlation coefficient of BOLD signal changes between ROIs in FPN (F3, right F3) and DMN (AG, MTG) for all conditions. Following z-transformation of these correlation coefficients, we performed t-tests that assessed whether connectivity between FPN and DMN ROIs and the rest of the brain changed as a function of condition. No significant results emerged. Next, we examined whether connectivity between specific FPN and DMN ROIs changed as a function of condition. As in the whole-brain analyses, no significant results emerged. Combined neurobehavioral results.
      To shed light on the neural underpinnings of TMS induced working memory enhancement, we ran a post-hoc AFNI 3dttest++ (2-backTMS – 2-back) with working memory accuracy (% 2-back correct subtracted from % 2-back TMS correct) entered as a covariate. Deactivation in a dorsomedial DMN cluster (p < .04) and a cluster adjacent to the dACC (p < .01), and increased activation in an aspect of the right orbitofrontal cortex (p < .01), were linked to increased working memory performance during high cognitive load (See Fig. 5).

      4. Discussion

      To our knowledge, the present investigation is the first TMS-fMRI study to examine how DLPFC TMS applied during various levels of cognitive load modulates activity in FPN and DMN regions and working memory performance.

      4.1 Neurobiological effects: TMS pulses increase high-cognitive load related DMN deactivations but do not alter FPN activation

      In line with meta-analytic findings [
      • Owen A.M.
      • McMillan K.M.
      • Laird A.R.
      • Bullmore E.
      N-back working memory paradigm: a meta-analysis of normative functional neuroimaging studies.
      ,
      • Rottschy C.
      • Langner R.
      • Dogan I.
      • Reetz K.
      • Laird A.R.
      • Schulz J.B.
      • et al.
      Modelling neural correlates of working memory: a coordinate-based meta-analysis.
      ], high cognitive load increased activation in FPN (bilateral DLPFC and IPL) and cingulo-opercular regions (DMPFC/dACC and insula), and decreased activation in DMN regions relative to low cognitive load. As hypothesized, the high cognitive load TMS condition showed increased activation in FPN nodes (right DLPFC and bilateral IPL) and decreased activation in regions of the anti-correlated DMN (vmPFC, right MTG, left AG) compared to the low cognitive load TMS condition.
      Unexpectedly, voxel-wise and FPN ROI interaction analyses did not reveal stimulation by load effects in the FPN. These results run counter to our prediction that DLPFC TMS would amplify high cognitive load related activations in the engaged FPN. One interpretation of this unexpected result is that the stimulated FPN was ‘saturated’ during high cognitive load, resulting in a ceiling effect. This possibility is consistent with motor and visual cortex state-dependency findings showing that engaging circuits prior to excitatory stimulation can dampen stimulation effects or even yield inhibitory effects (for review, see 33). Alternatively, signal loss under the coil (see Fig. S1) may have obscured TMS driven FPN activation increases. A third possibility consistent with previous findings is that depolarizing TMS applied to the DLPFC induces more robust effects on the anti-correlated DMN than on the FPN itself [
      • Chen A.C.
      • Oathes D.J.
      • Chang C.
      • Bradley T.
      • Zhou Z.W.
      • Williams L.M.
      • et al.
      Causal interactions between fronto-parietal central executive and default-mode networks in humans.
      ,
      • Liston C.
      • Chen A.C.
      • Zebley B.D.
      • Drysdale A.T.
      • Gordon R.
      • Leuchter B.
      • et al.
      Default mode network mechanisms of transcranial magnetic stimulation in depression.
      ]. Regardless of its cause, the lack of TMS induced increases in high cognitive load FPN activation may account for the unexpected absence of TMS driven changes in high load FPN-DMN connectivity.

      4.2 As predicted, TMS increased high cognitive load related deactivation in DMN nodes

      Compared to the no-TMS high cognitive load condition, the TMS high cognitive load condition featured increased deactivation in the right angular gyrus, a key DMN structure. Moreover, a contrast that assessed interactions between stimulation condition and cognitive load revealed stronger right MTG deactivation in the high versus low load TMS conditions than in the corresponding no-TMS conditions. Finally, a DMN-wide ROI containing canonical DMN nodes showed increased deactivation as a function of stimulation by load at trend level. Collectively, these results align with previous evidence demonstrating that DLPFC TMS modulates DMN activity [
      • Chen A.C.
      • Oathes D.J.
      • Chang C.
      • Bradley T.
      • Zhou Z.W.
      • Williams L.M.
      • et al.
      Causal interactions between fronto-parietal central executive and default-mode networks in humans.
      ,
      • Liston C.
      • Chen A.C.
      • Zebley B.D.
      • Drysdale A.T.
      • Gordon R.
      • Leuchter B.
      • et al.
      Default mode network mechanisms of transcranial magnetic stimulation in depression.
      ] and provide the first causal demonstration that single-pulse DLPFC TMS pulses increase high cognitive load related deactivations in DMN nodes.
      TMS increased low-cognitive load related activations in the left calcarine, a region implicated in visual processing [
      • Klein I.
      • Paradis A.L.
      • Poline J.B.
      • Kosslyn S.M.
      • Le Bihan D.
      Transient activity in the human calcarine cortex during visual-mental imagery: an event-related fMRI study.
      ,
      • Morland A.B.
      • Lê S.
      • Carroll E.
      • Hoffmann M.B.
      • Pambakian A.
      The role of spared calcarine cortex and lateral occipital cortex in the responses of human hemianopes to visual motion.
      ]. The DLPFC shows heightened dynamic functional connectivity with visual cortices during visual attention [
      • Wiesman A.I.
      • Heinrichs-Graham E.
      • Proskovec A.L.
      • McDermott T.J.
      • Wilson T.W.
      Oscillations during observations: dynamic oscillatory networks serving visuospatial attention.
      ]. During the low cognitive load condition, subjects were tasked with attending to stimuli in the center of the screen which putatively engaged visual attention. Increased activation in the calcarine during low cognitive load TMS may thus reflect increased activation of a functionally active visual processing/attentional circuit. High load TMS did not increase activation in this circuit; indeed, high load TMS was associated with reduced activations in other visual processing regions (e.g., cuneus) compared to low load TMS. Moreover, activations in auditory cortex (STG) putatively driven by TMS clicking noises were more widespread in the low versus high load TMS condition. Reduced activation in sensory processing regions during high cognitive load may reflect attentional gating of task irrelevant sensory signals driven by high cognitive load demands [
      • Sörqvist P.
      • Dahlström Ö.
      • Karlsson T.
      • Rönnberg J.
      Concentration: the neural underpinnings of how cognitive load shields against distraction.
      ]. These results indicate that despite sharing the same sensory characteristics, high and low load TMS conditions induced distinct activations in primary sensory areas.
      As in the high cognitive load TMS condition, TMS delivered during low cognitive load did not increase left DLPFC activation. As mentioned above, this result may be due to MR signal drop-out at the target location (see Fig. S1). Alternatively, recent TMS-fMRI studies targeting the DLPFC have also not detected a BOLD response in the targeted DLPFC location [
      • Rafiei F.
      • Safrin M.
      • Wokke M.
      • Lau H.
      • Rahnev D.
      TMS alters multivoxel patterns in the absence of overall activity changes.
      ,
      • Jackson J.B.
      • Feredoes E.
      • Rich A.N.
      • Lindner M.
      • Woolgar A.
      Concurrent neuroimaging and neurostimulation reveals a causal role for dlPFC in coding of task-relevant information.
      ], suggesting that stimulation of prefrontal cortical tissue may evoke different neurophysiological responses than stimulation of other cortical areas (e.g., motor cortex) that show intensity-dependent BOLD responses [
      • Bohning D.
      • Shastri A.
      • McConnell K.
      • Nahas Z.
      • Lorberbaum J.
      • Roberts D.
      • et al.
      A combined TMS/fMRI study of intensity-dependent TMS over motor cortex.
      ,
      • Nahas Z.
      • Lomarev M.
      • Roberts D.R.
      • Shastri A.
      • Lorberbaum J.P.
      • Teneback C.
      • et al.
      Unilateral left prefrontal transcranial magnetic stimulation (TMS) produces intensity-dependent bilateral effects as measured by interleaved BOLD fMRI.
      ].

      4.3 Behavioral effects: TMS delivered during high cognitive load enhances cognitive processing

      As hypothesized, TMS delivered during high cognitive load enhanced cognitive processing. The facilitatory effects of stimulation delivered during high cognitive load are consistent with results from a recent study finding a preferential effect of TMS on high working memory load conditions (45, but see 46). The neurobiological mechanisms underlying this effect remain unclear. However, DMN deactivations during high cognitive load in the present study are consistent with results from a recent study showing that TMS delivered to an FPN target decreased DMN activation during incongruent Stroop trials [
      • Novakova L.
      • Gajdos M.
      • Rektorova I.
      Theta-burst transcranial magnetic stimulation induced cognitive task-related decrease in activity of default mode network: an exploratory study.
      ]. Our neurobehavioral findings linking deactivation in a dorsomedial aspect of the DMN to enhanced high load cognitive processing provide causal, albeit preliminary given the lack of a control arm, support for the notion that DMN deactivation facilitates cognitive processing during high demand.
      The DMN supports episodic recollection and prospection [
      • Raichle M.E.
      The brain's default mode network.
      ,
      • Xu X.
      • Yuan H.
      • Lei X.
      Activation and connectivity within the default mode network contribute independently to future-oriented thought.
      ], key internal mentation processes that may interfere with externally guided attention and other on-line cognitive processes. Increased DMN deactivation may therefore support working memory by reducing interference between on-line cognitive processes and non-task related internal mentation.

      4.4 Clinical implications

      As discussed, cognitive control deficits have been detected in several disorders [
      • Cao X.
      • Deng C.
      • Su X.
      • Guo Y.
      Response and remission rates following high-frequency vs. low-frequency repetitive transcranial magnetic stimulation (rTMS) over right DLPFC for treating major depressive disorder (MDD): a meta-analysis of randomized, double-blind trials.
      ,
      • Liao X.
      • Li G.
      • Wang A.
      • Liu T.
      • Feng S.
      • Guo Z.
      • et al.
      Repetitive transcranial magnetic stimulation as an alternative therapy for cognitive impairment in Alzheimer's disease: a meta-analysis.
      ,
      • Nguyen T.D.
      • Hieronymus F.
      • Lorentzen R.
      • McGirr A.
      • Østergaard S.D.
      The efficacy of repetitive transcranial magnetic stimulation (rTMS) for bipolar depression: a systematic review and meta-analysis.
      ,
      • Vicario C.
      • Salehinejad M.A.
      • Felmingham K.
      • Martino G.
      • Nitsche M.
      A systematic review on the therapeutic effectiveness of non-invasive brain stimulation for the treatment of anxiety disorders.
      ], and bolstering cognitive control is a core mechanism by which clinical DLPFC rTMS is proposed to ameliorate a diverse array of neuropsychiatric syndromes [
      • Demirtas-Tatlidede A.
      • Vahabzadeh-Hagh A.M.
      • Pascual-Leone A.
      Can noninvasive brain stimulation enhance cognition in neuropsychiatric disorders?.
      ,
      • Lantrip C.
      • Gunning F.M.
      • Flashman L.
      • Roth R.M.
      • Holtzheimer P.E.
      Effects of transcranial magnetic stimulation on the cognitive control of emotion: potential antidepressant mechanisms.
      ,
      • Yang L.L.
      • Zhao D.
      • Kong L.L.
      • Sun Y.Q.
      • Wang Z.Y.
      • Gao Y.Y.
      • et al.
      High-frequency repetitive transcranial magnetic stimulation (rTMS) improves neurocognitive function in bipolar disorder.
      ].
      Although stand-alone rTMS has been shown to augment cognitive control sub-functions, effects are modest [
      • Brunoni A.R.
      • Vanderhasselt M.A.
      Working memory improvement with non-invasive brain stimulation of the dorsolateral prefrontal cortex: a systematic review and meta-analysis.
      ]. As mentioned, results from a growing rTMS literature suggest that rTMS applied to prefrontal circuits engaged by cognitive tasks/interventions may enhance treatment effects [
      • Sathappan A.V.
      • Luber B.M.
      • Lisanby S.H.
      The dynamic duo: combining noninvasive brain stimulation with cognitive interventions.
      ]. The present findings shed light on the basic neurobehavioral mechanisms potentially undergirding these clinical effects by demonstrating that TMS increases DMN deactivation and augments cognitive performance during high cognitive load. If replicated in clinical samples, these results could spur clinical trials comparing the neurobehavioral effects of rTMS delivered during cognitively demanding tasks/interventions versus at rest.
      Contrary to the present findings, Bakulin et al., 2020 did not find a cognitive enhancing effect of TMS delivered during high cognitive load. Instead, effects were detected for TMS delivered at rest, which both enhanced and disrupted storage and maintenance phases of spatial working memory, respectively. Differences in our respective study designs may partially account for these discrepancies. For example, Bakulin et al., 2020 applied high-frequency rTMS and assessed working memory offline, while we applied single-pulse TMS and assessed working memory online. Regardless of its origin, the contrasting nature of these results necessitates additional multimodal investigations that concurrently measure neurobiological and cognitive effects of TMS delivered during different levels of cognitive load. Such designs may shed light on the optimal point at which to stimulate to maximally enhance cognitive control. Coupled with advances in closed-loop technology, this line of work could eventually yield temporally precise clinical rTMS protocols that sync stimulation timing with underlying neural activity [
      • Zrenner C.
      • Belardinelli P.
      • Müller-Dahlhaus F.
      • Ziemann U.
      Closed-loop neuroscience and non-invasive brain stimulation: a tale of two loops.
      ].

      5. Limitations and conclusions

      The present results should be interpreted in light of several limitations. First, although our sample size was similar to or exceeded several of our group's previous interleaved TMS-fMRI studies [
      • Nahas Z.
      • Lomarev M.
      • Roberts D.R.
      • Shastri A.
      • Lorberbaum J.P.
      • Teneback C.
      • et al.
      Unilateral left prefrontal transcranial magnetic stimulation (TMS) produces intensity-dependent bilateral effects as measured by interleaved BOLD fMRI.
      ,
      • Li X.
      • Tenebäck C.C.
      • Nahas Z.
      • Kozel F.A.
      • Large C.
      • Cohn J.
      • et al.
      Interleaved transcranial magnetic stimulation/functional MRI confirms that lamotrigine inhibits cortical excitability in healthy young men.
      ,
      • Webler R.D.
      • Hamady C.
      • Molnar C.
      • Johnson K.
      • Bonilha L.
      • Anderson B.S.
      • et al.
      Decreased interhemispheric connectivity and increased cortical excitability in unmedicated schizophrenia: a prefrontal interleaved TMS fMRI study.
      ], larger multimodal studies are warranted to replicate the relationship between DMN deactivation and cognitive effects of stimulation detected in the present study. Second, consistent with the nascent stage of this research area, we utilized a simple block-design with a relatively small number of target trials and did not deliver stimulation that was time-locked to specific stages of cognitive processing. Future event-related designs featuring more target trials are therefore necessary to replicate the present preliminary findings and delineate how TMS delivered during different phases of cognitive processing shapes working memory related neural activations and performance. Third, the present study did not utilize a control stimulation condition (i.e., sham TMS or an active control target). High and low cognitive load TMS conditions were designed to serve as their own respective sensory controls. However, as described above, low cognitive load TMS featured increased activation in primary sensory cortices compared to high cognitive load TMS, indicating that the sensory effects of TMS differed as a function of cognitive load. Therefore, future studies that employ a sham or active control are warranted to better isolate the neural and sensory effects of high vs. low cognitive load TMS. Finally, although the Beam method used in the present study has been shown to be a reliable method for targeting the DLPFC [
      • Trapp N.T.
      • Bruss J.
      • Johnson M.K.
      • Uitermarkt B.D.
      • Garrett L.
      • Heinzerling A.
      • et al.
      Reliability of targeting methods in TMS for depression: Beam F3 vs. 5.5 cm.
      ], selecting individualized targets on the basis of functional connectivity may further increase reliability and precision [
      • Cash R.F.
      • Cocchi L.
      • Lv J.
      • Wu Y.
      • Fitzgerald P.B.
      • Zalesky A.
      Personalized connectivity-guided DLPFC-TMS for depression: advancing computational feasibility, precision and reproducibility.
      ]. Future studies that leverage functional connectivity based targeting may induce more precise effects on functional networks.
      To conclude, our findings shed novel light on the effect of DLPFC TMS on working memory and its neural substrates, whereby TMS increased deactivation in DMN nodes and strengthened cognitive processing during high cognitive load.

      Financial disclosures

      All authors report no conflicts of interest related to this publication.

      Acknowledgements

      The current work was conducted within and supported by the Brain Stimulation laboratory at MUSC . The authors would like to thank Dr. Colleen Hanlon for her input regarding neuroimaging pre-processing and analyses.

      Appendix A. Supplementary data

      The following are the Supplementary data to this article:

      References

        • Braver T.S.
        The variable nature of cognitive control: a dual mechanisms framework.
        Trends Cognit Sci. 2012; 16: 106-113
        • McTeague L.M.
        • Goodkind M.S.
        • Etkin A.
        Transdiagnostic impairment of cognitive control in mental illness.
        J Psychiatr Res. 2016; 83: 37-46
        • McTeague L.M.
        • Huemer J.
        • Carreon D.M.
        • Jiang Y.
        • Eickhoff S.B.
        • Etkin A.
        Identification of common neural circuit disruptions in cognitive control across psychiatric disorders.
        Am J Psychiatr. 2017; 174: 676-685
        • Niendam T.A.
        • Laird A.R.
        • Ray K.L.
        • Dean Y.M.
        • Glahn D.C.
        • Carter C.S.
        Meta-analytic evidence for a superordinate cognitive control network subserving diverse executive functions.
        Cognit Affect Behav Neurosci. 2012; 12: 241-268
        • McLean A.L.
        Publication trends in transcranial magnetic stimulation: a 30-year panorama.
        Brain Stimul. 2019; 12: 619-627
        • Cao X.
        • Deng C.
        • Su X.
        • Guo Y.
        Response and remission rates following high-frequency vs. low-frequency repetitive transcranial magnetic stimulation (rTMS) over right DLPFC for treating major depressive disorder (MDD): a meta-analysis of randomized, double-blind trials.
        Front Psychiatr. 2018; 9: 413
        • Liao X.
        • Li G.
        • Wang A.
        • Liu T.
        • Feng S.
        • Guo Z.
        • et al.
        Repetitive transcranial magnetic stimulation as an alternative therapy for cognitive impairment in Alzheimer's disease: a meta-analysis.
        J Alzheimers Dis. 2015; 48: 463-472
        • Nguyen T.D.
        • Hieronymus F.
        • Lorentzen R.
        • McGirr A.
        • Østergaard S.D.
        The efficacy of repetitive transcranial magnetic stimulation (rTMS) for bipolar depression: a systematic review and meta-analysis.
        J Affect Disord. 2020;
        • Vicario C.
        • Salehinejad M.A.
        • Felmingham K.
        • Martino G.
        • Nitsche M.
        A systematic review on the therapeutic effectiveness of non-invasive brain stimulation for the treatment of anxiety disorders.
        Neurosci Biobehav Rev. 2019; 96: 219-231
        • Demirtas-Tatlidede A.
        • Vahabzadeh-Hagh A.M.
        • Pascual-Leone A.
        Can noninvasive brain stimulation enhance cognition in neuropsychiatric disorders?.
        Neuropharmacology. 2013; 64: 566-578
        • Lantrip C.
        • Gunning F.M.
        • Flashman L.
        • Roth R.M.
        • Holtzheimer P.E.
        Effects of transcranial magnetic stimulation on the cognitive control of emotion: potential antidepressant mechanisms.
        J ECT. 2017; 33: 73-80
        • Yang L.L.
        • Zhao D.
        • Kong L.L.
        • Sun Y.Q.
        • Wang Z.Y.
        • Gao Y.Y.
        • et al.
        High-frequency repetitive transcranial magnetic stimulation (rTMS) improves neurocognitive function in bipolar disorder.
        J Affect Disord. 2019; 246: 851-856
        • Sathappan A.V.
        • Luber B.M.
        • Lisanby S.H.
        The dynamic duo: combining noninvasive brain stimulation with cognitive interventions.
        Prog Neuro-Psychopharmacol Biol Psychiatry. 2019; 89: 347-360
        • Chen A.C.
        • Oathes D.J.
        • Chang C.
        • Bradley T.
        • Zhou Z.W.
        • Williams L.M.
        • et al.
        Causal interactions between fronto-parietal central executive and default-mode networks in humans.
        Proc Natl Acad Sci. 2013; 110 (9): 19944
        • Denkova E.
        • Nomi J.S.
        • Uddin L.Q.
        • Jha A.P.
        Dynamic brain network configurations during rest and an attention task with frequent occurrence of mind wandering.
        Hum Brain Mapp. 2019; 40: 4564-4576
        • Philip N.S.
        • Barredo J.
        • van’t Wout-Frank M.
        • Tyrka A.R.
        • Price L.H.
        • Carpenter L.L.
        Network mechanisms of clinical response to transcranial magnetic stimulation in posttraumatic stress disorder and major depressive disorder.
        Biol Psychiatr. 2018; 83: 263-272
        • Shang Y.
        • Chang D.
        • Zhang J.
        • Peng W.
        • Song D.
        • Gao X.
        • et al.
        Theta-burst transcranial magnetic stimulation induced functional connectivity changes between dorsolateral prefrontal cortex and default-mode-network.
        Brain Imaging Behav. 2019; : 1-9
        • Wang H.
        • He W.
        • Wu J.
        • Zhang J.
        • Jin Z.
        • Li L.
        A coordinate-based meta-analysis of the n-back working memory paradigm using activation likelihood estimation.
        Brain Cognit. 2019; 132: 1-12
        • Eshel N.
        • Keller C.J.
        • Wu W.
        • Jiang J.
        • Mills-Finnerty C.
        • Huemer J.
        • et al.
        Global connectivity and local excitability changes underlie antidepressant effects of repetitive transcranial magnetic stimulation.
        Neuropsychopharmacology. 2020; 45: 1018-1025
        • Gu H.
        • Hu Y.
        • Chen X.
        • He Y.
        • Yang Y.
        Regional excitation-inhibition balance predicts default-mode network deactivation via functional connectivity.
        Neuroimage. 2019; 185: 388-397
        • Cui H.
        • Ren R.
        • Lin G.
        • Zou Y.
        • Jiang L.
        • Wei Z.
        • et al.
        Repetitive transcranial magnetic stimulation induced hypoconnectivity within the default mode network yields cognitive improvements in amnestic mild cognitive impairment: a randomized controlled study.
        J Alzheimers Dis. 2019; 69: 1137-1151
        • Cai W.
        • Ryali S.
        • Pasumarthy R.
        • Talasila V.
        • Menon V.
        Dynamic causal brain circuits during working memory and their functional controllability.
        Nat Commun. 2021; 12: 1-16
        • Liston C.
        • Chen A.C.
        • Zebley B.D.
        • Drysdale A.T.
        • Gordon R.
        • Leuchter B.
        • et al.
        Default mode network mechanisms of transcranial magnetic stimulation in depression.
        Biol Psychiatr. 2014; 76: 517-526
        • Keller J.B.
        • Hedden T.
        • Thompson T.W.
        • Anteraper S.A.
        • Gabrieli J.D.
        • Whitfield-Gabrieli S.
        Resting-state anticorrelations between medial and lateral prefrontal cortex: association with working memory, aging, and individual differences.
        Cortex. 2015; 64: 271-280
        • Anticevic A.
        • Repovs G.
        • Shulman G.L.
        • Barch D.M.
        When less is more: TPJ and default network deactivation during encoding predicts working memory performance.
        Neuroimage. 2010; 49: 2638-2648
        • Fox M.D.
        • Buckner R.L.
        • White M.P.
        • Greicius M.D.
        • Pascual-Leone A.
        Efficacy of transcranial magnetic stimulation targets for depression is related to intrinsic functional connectivity with the subgenual cingulate.
        Biol Psychiatr. 2012; 72: 595-603
        • Hopman H.
        • Chan S.
        • Chu W.C.W.
        • Lu H.
        • Tse C.Y.
        • Chau S.W.H.
        • et al.
        Personalized prediction of transcranial magnetic stimulation clinical response in patients with treatment-refractory depression using neuroimaging biomarkers and machine learning.
        J Affect Disord. 2021; 290: 261-271
        • Weigand A.
        • Horn A.
        • Caballero R.
        • Cooke D.
        • Stern A.P.
        • Taylor S.F.
        • et al.
        Prospective validation that subgenual connectivity predicts antidepressant efficacy of transcranial magnetic stimulation sites.
        Biol Psychiatr. 2018; 84: 28-37
        • Annett M.
        A classification of hand preference by association analysis.
        Br J Psychol. 1970; 61: 303-321
        • First M.B.
        • Gibbon M.
        The structured clinical interview for DSM-IV Axis I disorders (SCID-I) and the structured clinical interview for DSM-IV Axis II disorders (SCID-II).
        2004
        • Hamilton M.
        A rating scale for depression.
        J Neurol Neurosurg Psychiatry. 1960; 23: 56
        • Rush A.J.
        • Giles D.E.
        • Schlesser M.A.
        • Fulton C.L.
        • Weissenburger J.
        • Burns C.
        The inventory for depressive symptomatology (IDS): preliminary findings.
        Psychiatr Res. 1986; 18: 65-87
        • Borckardt J.J.
        • Nahas Z.
        • Koola J.
        • George M.S.
        Estimating resting motor thresholds in transcranial magnetic stimulation research and practice: a computer simulation evaluation of best methods.
        J ECT. 2006; 22: 169-175
        • Beam W.
        • Borckardt J.J.
        • Reeves S.T.
        • George M.S.
        An efficient and accurate new method for locating the F3 position for prefrontal TMS applications.
        Brain Stimul. 2009; 2: 50-54
        • Owen A.M.
        • McMillan K.M.
        • Laird A.R.
        • Bullmore E.
        N-back working memory paradigm: a meta-analysis of normative functional neuroimaging studies.
        Hum Brain Mapp. 2005; 25: 46-59
        • Rottschy C.
        • Langner R.
        • Dogan I.
        • Reetz K.
        • Laird A.R.
        • Schulz J.B.
        • et al.
        Modelling neural correlates of working memory: a coordinate-based meta-analysis.
        Neuroimage. 2012; 60: 830-846
        • Klein I.
        • Paradis A.L.
        • Poline J.B.
        • Kosslyn S.M.
        • Le Bihan D.
        Transient activity in the human calcarine cortex during visual-mental imagery: an event-related fMRI study.
        J Cognit Neurosci. 2000; 12: 15-23
        • Morland A.B.
        • Lê S.
        • Carroll E.
        • Hoffmann M.B.
        • Pambakian A.
        The role of spared calcarine cortex and lateral occipital cortex in the responses of human hemianopes to visual motion.
        J Cognit Neurosci. 2004; 16: 204-218
        • Wiesman A.I.
        • Heinrichs-Graham E.
        • Proskovec A.L.
        • McDermott T.J.
        • Wilson T.W.
        Oscillations during observations: dynamic oscillatory networks serving visuospatial attention.
        Hum Brain Mapp. 2017; 38: 5128-5140
        • Sörqvist P.
        • Dahlström Ö.
        • Karlsson T.
        • Rönnberg J.
        Concentration: the neural underpinnings of how cognitive load shields against distraction.
        Front Hum Neurosci. 2016; 10: 221
        • Rafiei F.
        • Safrin M.
        • Wokke M.
        • Lau H.
        • Rahnev D.
        TMS alters multivoxel patterns in the absence of overall activity changes.
        bioRxiv, 2020
        • Jackson J.B.
        • Feredoes E.
        • Rich A.N.
        • Lindner M.
        • Woolgar A.
        Concurrent neuroimaging and neurostimulation reveals a causal role for dlPFC in coding of task-relevant information.
        Commun Biol. 2021; 4: 1-16
        • Bohning D.
        • Shastri A.
        • McConnell K.
        • Nahas Z.
        • Lorberbaum J.
        • Roberts D.
        • et al.
        A combined TMS/fMRI study of intensity-dependent TMS over motor cortex.
        Biol Psychiatr. 1999; 45: 385-394
        • Nahas Z.
        • Lomarev M.
        • Roberts D.R.
        • Shastri A.
        • Lorberbaum J.P.
        • Teneback C.
        • et al.
        Unilateral left prefrontal transcranial magnetic stimulation (TMS) produces intensity-dependent bilateral effects as measured by interleaved BOLD fMRI.
        Biol Psychiatr. 2001; 50: 712-720
        • Beynel L.
        • Davis S.
        • Crowell C.
        • Hilbig S.
        • Lim W.
        • Nguyen D.
        • et al.
        Online repetitive transcranial magnetic stimulation during working memory in younger and older adults: a randomized within-subject comparison.
        PLoS One. 2019; 14e0213707
        • Bakulin I.
        • Zabirova A.
        • Lagoda D.
        • Poydasheva A.
        • Cherkasova A.
        • Pavlov N.
        • et al.
        Combining HF rTMS over the left DLPFC with concurrent cognitive activity for the offline modulation of working memory in healthy volunteers: a proof-of-concept study.
        Brain Sci. 2020; 10: 83
        • Novakova L.
        • Gajdos M.
        • Rektorova I.
        Theta-burst transcranial magnetic stimulation induced cognitive task-related decrease in activity of default mode network: an exploratory study.
        Brain Stimul Basic Transl Clin Res Neuromodulation. 2020; 13: 597-599
        • Raichle M.E.
        The brain's default mode network.
        Annu Rev Neurosci. 2015; 38: 433-447
        • Xu X.
        • Yuan H.
        • Lei X.
        Activation and connectivity within the default mode network contribute independently to future-oriented thought.
        Sci Rep. 2016; 6: 1-10
        • Brunoni A.R.
        • Vanderhasselt M.A.
        Working memory improvement with non-invasive brain stimulation of the dorsolateral prefrontal cortex: a systematic review and meta-analysis.
        Brain Cognit. 2014; 86: 1-9
        • Zrenner C.
        • Belardinelli P.
        • Müller-Dahlhaus F.
        • Ziemann U.
        Closed-loop neuroscience and non-invasive brain stimulation: a tale of two loops.
        Front Cell Neurosci. 2016; 10: 92
        • Li X.
        • Tenebäck C.C.
        • Nahas Z.
        • Kozel F.A.
        • Large C.
        • Cohn J.
        • et al.
        Interleaved transcranial magnetic stimulation/functional MRI confirms that lamotrigine inhibits cortical excitability in healthy young men.
        Neuropsychopharmacology. 2004; 29: 1395-1407
        • Webler R.D.
        • Hamady C.
        • Molnar C.
        • Johnson K.
        • Bonilha L.
        • Anderson B.S.
        • et al.
        Decreased interhemispheric connectivity and increased cortical excitability in unmedicated schizophrenia: a prefrontal interleaved TMS fMRI study.
        Brain Stimul. 2020; 13: 1467-1475
        • Trapp N.T.
        • Bruss J.
        • Johnson M.K.
        • Uitermarkt B.D.
        • Garrett L.
        • Heinzerling A.
        • et al.
        Reliability of targeting methods in TMS for depression: Beam F3 vs. 5.5 cm.
        Brain Stimul. 2020; 13: 578-581
        • Cash R.F.
        • Cocchi L.
        • Lv J.
        • Wu Y.
        • Fitzgerald P.B.
        • Zalesky A.
        Personalized connectivity-guided DLPFC-TMS for depression: advancing computational feasibility, precision and reproducibility.
        Hum Brain Mapp. 2021;