Effects of Intravenous (IV) Citalopram Hydrochloride During Transcranial Magnetic Stimulation in Major Depressive Disorder (MDD)

Repetitive Transcranial Magnetic Stimulation (rTMS) is an increasingly common treatment for Major Depressive Disorder (MDD). rTMS applied to dorsolateral prefrontal cortex (DLPFC), the most common treatment target, appears to change neuronal excitability in this region. rTMS also changes function of brain circuits connected to DLPFC. This application proposes an innovative approach to elucidating the mechanism of action (MOA) underlying these circuit changes, using MDD as a significant translational model. One form of rTMS, theta burst stimulation (TBS), has particularly strong effects on cortical excitability: intermittent pulsing over left DLPFC (iTBS) increases excitability, while continuous pulsing over right DLPFC (cTBS) reduces excitability. TBS applied to DLPFC alters functional connectivity with the anterior cingulate, medial frontal, and orbitofrontal cortices (ACC, MFC, and OFC); however, the MOA underlying these changes in connectivity is incompletely understood. Pilot data will be obtained for a larger grant application that will test the hypothesis that changes in local excitability underlie the changes in functional connectivity and the therapeutic efficacy of TBS for MDD. TBS modulation of cortical excitability may be modulated in turn by the serotonergic (5HT) neurotransmitter system, which is also a key target of classical antidepressant medications. The investigators will use the 5HT transport inhibitor citalopram (CIT), a widely-used clinical antidepressant agent, to investigate the serotonergic modulation of functional connectivity and neurophysiologic measures of excitability. Intravenous citalopram hydrochloride has been available for prescription to patients with treatment-refractory major depression and anxiety disorders in most of continental Europe for more than 30 years. In the U.S., it has been used extensively as an investigational drug to study human neurochemistry and in clinical trials for depressive disorders. An IND for the University of Pittsburgh has utilized this compound safely as a research tool for more than 10 years (Smith et al., 2009). This double-blinded study will recruit 30 subjects diagnosed with Major Depressive Disorder (MDD). Subjects will be randomized to acute (single-dose) citalopram (CIT) 40 mg iv treatment or placebo, counterbalanced and combined with one of two forms of unblinded Transcranial Magnetic Stimulation, i.e., either intermittent Theta Burst Stimulation or continuous Theta Burst Stimulation. Assessments for this study include brain scans, qEEG recordings, and cognitive and mood scales. One citalopram/placebo infusion and 10 TBS treatments will be administered over the course of approximately two weeks. Thirty subjects ages 21-55 with a DSM-V diagnosis of MDD will be enrolled in a two-week treatment study. After a baseline diagnostic assessment, all subjects will undergo pretreatment assessment of brain functional connectivity (qEEG and fMRI) and structural connectivity (DTI). They then will be randomized 1:1 to treatment with an intravenous dose of CIT or placebo (PBO), followed by 1:1 randomized assignment to two weeks (10 sessions) of treatment with either iTBS to left or cTBS to right DLPFC (four treatment conditions). The end of Week 2 will constitute the primary endpoint. High-density qEEG will be recorded throughout the initial CIT-TBS treatment session using TMS-compatible qEEG. These recordings will be used to assess changes in excitability with CIT/PBO and iTBS/cTBS treatment (i.e., TMS evoked Local Mean Field Power, or LMFP). The investigators will determine whether changes in cortical excitability in DLPFC are modulated by 5HT neurotransmission. Next, changes in neuroplasticity in limbic circuits involving DLPFC will be examined, from pretreatment baseline to after the first treatment, and from baseline to the primary endpoint, using functional connectivity measures. qEEG will be used to measure changes in signal synchronization and information flow (i.e., lagged coherence, Granger Causality), as well as location and spread of current sources (LORETA source localization). fMRI will be used to examine resting state network function (BOLD signal). The investigators will test whether changes in excitability measured in the first treatment are correlated with changes in connectivity with the first treatment, as well as over the full course of treatment. Correlation would suggest that excitability may exert a metaplastic effect on functional connectivity in limbic circuits involving DLPFC. Finally, the investigators will examine symptom change at the primary endpoint, and determine whether excitability and connectivity changes correlate with symptom improvement. Integrated examination of changes in excitability, immediate and longer-term connectivity, and symptom improvement, will elucidate the MOA of TBS and could lead to strategies to enhance treatment outcomes. Clinical diagnosis and assessment. Thirty subjects with a DSM-V diagnosis of MDD established using the Mini International Neuropsychiatric Interview (MINI, version 7.0) (http://www.medical-outcomes.com/index/mini7fororganizations) will be enrolled, with all subjects having depressive symptoms of at least moderate severity as indicated by a 17-item Hamilton Depression Rating Scale \[Ham-D17\] score \> 17 (23 below) will be enrolled. Subjects will undergo clinical assessment using methods similar to those employed in our previous treatment studies using TMS (Leuchter et al., 2015). Subjects will have failed to enter remission with at least two prior antidepressant medications (Sackeim et al., 1990; Vasavada et al., 2016) and must have been free of any medications known to significantly affect brain function for at least ten days prior to enrollment (except fluoxetine, which will require a five-week washout) (Vasavada et al., 2016). Subjects will be excluded if they meet DSM-V criteria for any other current primary Axis I mood, anxiety, or psychotic disorder, depression secondary to a general medical condition, or substance-induced illness. Subjects also will be excluded if they have current suicidal intent or plan, a history of substance abuse or dependence within the past six months (except nicotine and caffeine), Bipolar Disorder or psychotic disorder (lifetime), eating disorder (current or within the past year), Obsessive Compulsive Disorder (lifetime), Post-Traumatic Stress Disorder (PTSD, current or within the past year), medical or neurologic illness that would contraindicate administration of study interventions or complicate interpretation of study results, or have an implanted medical device or metal in their body that would contraindicate an MRI or TMS treatment (Leuchter et al., 2015). Subjects also will be excluded if they have had prior history with IV citalopram. Women who are currently pregnant, not using a medically acceptable means of birth control, or are breastfeeding will be excluded. A urine drug screen will be performed, and subjects with a positive screen for illicit substances will be excluded. Mood symptoms will be examined after the first, second, and 10th treatment sessions using the Clinician Global Impressions-Severity of Illness (CGI-S) and Improvement (CGI-I) (Cohen et al., 2014), and the subject-rated Inventory of Depressive Symptomatology-Self Report (IDS-SR) (Connolly et al., 2012), with symptoms of suicidal ideation assessed using the Columbia- Suicide Severity Rating Scale (C-SSRS) (Cornwell et al., 2012). Assessments at time of entry and exit from the study will include symptom ratings with the Ham-D17 and quality of life and functional status ratings with the Quality of Life Inventory (QOLI) (Dandash et al., 2015). Treatment response at the primary endpoint is defined as a 50% or greater improvement from baseline on the Ham-D17 and remission as a final Ham-D17 score \< 7. Treatment procedures Citalopram (CIT) and placebo (PBO) infusion. CIT and saline PBO will be administered intravenously using established clinical procedures. A single dose of citalopram 40 mg in 60 cc normal saline, or saline PBO, will be delivered intravenously under double-blind conditions via pump over a 40-minute period. Blood will be drawn at the conclusion of infusion to obtain a plasma sample for CIT levels. Subjects will be fasting after midnight or for a minimum of 8 hours prior to undergoing infusion. Vital sign monitoring will include blood pressure, pulse oximetry, and respiratory rate recording every 3 minutes and a continuous cardiac rhythm strip. A change of greater than 25% increase in heart rate or blood pressure from baseline or absolute increase in heart rate above 120 bpm or systolic blood pressure ≥ 180 mm Hg or diastolic blood pressure ≥ 105 mm Hg sustained from more than 2 minutes will prompt the immediate discontinuation of double-blind infusion and evaluation for intervention. Similarly, a drop of O2 saturation via pulse oximetry to less than 92% will prompt evaluation of subject, initiate use of nasal cannula O2 at 2 liters/min or rate necessary to return O2 saturation to baseline at room air or greater than 95%. A physician investigator will administer the CIT or PBO infusion, perform subject assessment and direct nursing staff. Mental status monitoring will also occur during double-blind infusion to assess for any untoward behavioral or psychological effects. Subjects will be instructed not to drive for up to 24 hours after infusion and will need to be driven to and from UCLA for the procedure. Theta burst transcranial magnetic stimulation (TBS). Stimulation will be performed using a Magstim Rapid2 biphasic stimulator with a figure-8 coil (14 cm width) and 2 T peak field strength. Stimulation percentages are expressed as a proportion of this individual unit's Maximum Stimulator Output (MSO). This unit can generate the theta-burst stimulation patterns at intensities of 45% MSO or below, well within range of most participant's individual motor threshold (MT). To determine MT, the coil will be held mediolaterally over the region of the left motor cortex with the handle pointing backwards and 45° from sagittal midline. This technique induces current roughly perpendicular to the central sulcus. The right first dorsal interosseus muscle (FDI) will be monitored with surface electromyography (5000 Hz). TMS pulses will be delivered in a grid at suprathreshold intensities in order to identify the location which produces the largest, most consistent motor evoked potential (MEP) recorded from the FDI. Intensities at this hotspot will then be lowered 1% with each stimulation. The lowest intensity stimulation that produces peak-to-peak MEP amplitudes \>100 μV on at least 5 of 10 trials under conditions of gentle activation of the FDI is defined as the active motor threshold (AMT). TBS intensity is then set at 120% AMT. TBS consists of three TMS pulses given at 50 Hz, with this triplet repeated at a frequency of 5 Hz (every 200 ms) (Huang et al., 2005). 1800 pulses of cTBS will be delivered to the right DLPFC, and an equal number to left DLPFC following the iTBS paradigm of a 2 s train of repeated every 10 s. This number of pulses has been shown to have antidepressant efficacy after two weeks of treatment (Li et al., 2014). Left and right DLPFC will be targeted using the F3 and F4 EEG electrode locations, respectively. This method for magnet placement bears a close relationship to placing the magnet over radiographically defined Broadmann areas (BAs) in DLPFC (i.e., BA46) (Ahdab et al., 2010; Fitzgerald et al., 2009). This approach, widely used in clinical practice, will allow us to easily relate the findings of this project to clinical use. The investigators recognize that this probabilistic method for defining DLPFC may result in some variability of magnet placement relative to underlying neuroanatomy. Given interindividual variability in sulcal and gyral anatomy, however, it may not be possible to reliably identify a specific BA in all subjects, and there is no "gold-standard" method for standardizing neuroanatomic targeting across subjects. The investigators will have digitized surface electrode locations from all subjects, and these locations will be fused with the structural MRI images obtained from each subject at baseline and after the 10th iTBS treatment. This will allow us to utilize neuroanatomic coil placement data as a post-hoc covariate in data analyses (see MRI-EEG image integration below). Neurophysiologic recording and analyses. EEG recording. Data will be recorded using the "eego mylab" TMS-compatible EEG system at a sampling rate of 1000 Hz (Advanced Neuro Technology \[ANT\]; Enschede, Netherlands). Electrodes will be applied using the 64-electrode "WaveGuard" system with sintered Ag/AgCl electrodes mounted in an elastic cap and positioned according to the Extended 10-20 System. The material and shape of the electrodes prevents current loops is designed for minimal DC shifts and optimal stability of the incoming signal during TMS. The cap utilizes active shielding of each lead to limit electric noise. Data are recorded using full-band EEG DC amplifiers that return to physiologic baseline signal level within 10 ms after the end of the TMS pulse. Filters will not be applied during data acquisition, and recording will be performed using a common average reference with impedance kept below 5 kΩ. EOG will be recorded by placing two electrodes above and below the left eye. EEG will be processed off-line in BrainVisionAnalyzer2 (BVA2) (BrainProducts GmbH; Gilching, Germany) with a digital band-pass filter (Butterworth zero-phase shift, 0.5-70 Hz, 12 dB/oct; plus 60 Hz notch) before segmenting into 2 s epochs (100 ms pre-stimulus period), detrending, and baseline correction. Data initially will be processed using semi-automated artifact rejection methods previously described including ±100 μV peak-to-peak voltage step gradient or persistent low activity (Leuchter et al., 2012), followed by visual inspection by two independent technicians to eliminate data contaminated by muscle, head, or eye movement artifacts. Adaptive Mixture ICA (AMICA) also will be used to separate out non-brain source processes including eye blinks and saccades, scalp muscle, electrocardiographic artifact, and line noise, increasing signal-to-noise (artifact) ratio and increasing reliability of high frequency (beta and gamma) frequency analyses as well as brain source localization (Bigdely-Shamlo et al., 2013; Delorme et al., 2011). Excitability measures. Excitability will be measured using the Local Mean Field Power (LMFP) and Global Mean Field Power (GMFP) methods. Oscillatory voltage amplitude is the principal measurement that directly reflects cortical excitability, and therefore changes in EEG field power are used as the principal indicator of changes in excitability. LMFP can be used as a measure of excitability at the site of TMS stimulation in any region of cortex; GMFP is a measure of global excitability that has been used to study a number of non-invasive neuromodulation treatments (Casarotto et al., 2013; Chung et al., 2015; Huber et al., 2008; Pellicciari et al., 2013; Romero Lauro et al., 2014). After administration of a local stimulus, a focal change in excitability may come to elicit a global change; as a result, GMFP can be used to interpret LMFP and determine whether a local change in excitability remains focal, or becomes part of a global change in excitability. For the excitability determinations, single TMS pulses will delivered to left or right DLPFC (F3 or F4 electrode location, coinciding with the treatment site) 10 minutes before and after treatment sessions. Given the number of TBS pulses delivered in the protocol, changes in excitability can be expected to remain stable for at least one hour following TBS treatment (Huang et al., 2005). Pulses will be administered at a frequency of 0.25 Hz so that they are sufficiently infrequent so as not affect excitability. During excitability determination, subjects will wear earplugs and a masking noise will be played to reproduce the TMS "click" in time-varying frequency components in order to suppress auditory evoked potentials. TEPs will be computed by averaging valid artifact-free single trials, filtering between 2 and 40 Hz, and performing baseline correction before and after the TMS pulse. LMFP will be calculated from the amplitudes of the TEPs averaged from F3 or F4 and their respective four surrounding electrodes, while GMFP will be calculated from all 64 electrodes. GMFP and LMFP will be calculated in 30 ms time windows following the TMS pulse ranging from the end of the pulse to 400 ms (i.e., 0-30 ms, 30-60 ms, 60-90 ms, etc.) so that early and later changes in excitability can be detected. Early time points (\< 90 ms) will be used to assess LMFP, and later time points will be used to assess GMFP and spread altered excitability to other cortical regions. Regional measures of neurophysiologic activity. Current source density (CSD) in gray matter structures will be calculated using the eLORETA (exact Low Resolution Brain Electromagnetic TomogrAphy) method (http://www.uzh.ch/keyinst/loreta.htm) (Lehmann et al., 2014). eLORETA computes local neurophysiologic activity (Current Source Density, or CSD) as a linear weighted sum of scalp electrical potentials. CSD measurements using eLORETA have been validated as cortical sources of neurophysiologic activity, and obviate the the ambiguity of source localization and the reference-dependence that are inherent in scalp EEG measurements (Lehmann et al., 2014; Pascual-Marqui et al., 2011). The method identifies the smoothest possible distribution of sources in a three-shell spherical head model, consisting of CSD at each of 6239 cortical voxels (plus hippocampus and amygdala) (5 mm spatial resolution) in Montreal Neurological Institute (MNI) space with electrode coordinates assigned according to cross-registration between spherical and realistic head geometry (Towle et al., 1993). Reported Brodmann areas utilize MNI space corrected to coincide with Talairach space (Brett et al., 2002). Regions of interest (ROIs) will be created for subregions of the ACC, OFC, hippocampus, insula, and other cortical ROIs in the limbic system based upon our prior work and the broader literature (Arns et al., 2015; Korb et al., 2008; Korb et al., 2009; Korb et al., 2011; Whitton et al., 2016). Neurophysiologic connectivity measures. The investigators will use one primary neurophysiologic measure and several complementary techniques as secondary (exploratory) measures in order to examine baseline and changes in neurophysiologic connectivity. Our primary connectivity measure based upon CSD data will be examined using "lagged" coherence (omitting zero phase angle) based on intracortical source modeling techniques (Lehmann et al., 2014; Pasi et al., 1989). The investigators will identify cortical ROIs throughout the PFC, hippocampus, and amygdala based upon BAs and subcortical probabilistic atlases, and will apply the eLORETA option "all nearest voxels" to assigned voxels to the ROIs. The investigators will use the ROIs identified by eLORETA as showing high activity to seed the connectivity analyses; The investigators anticipate that the primary ROIs of interest will be those with strong connections to DLPFC in the limbic system, including OFC, MFC, ACC, hippocampus, and insula. Lagged coherence represents one of the two main components of neurophysiologic connectivity: instantaneous and lagged. The lagged component is mediated by physiological time delays, with the instantaneous contribution to the connectivity is eliminated. This method selectively retains connectivity due to physiologic processes (for any non-zero, measurable time delay) that is not confounded by low resolution and volume conduction effects (Lehmann et al., 2014). Secondary measures also will examine connectivity and cross-frequency coupling in the electrode space using the Source Information Flow Toolbox (SIFT) (Delorme et al., 2011). This method uses time-varying (adaptive) multivariate autoregressive modeling to detect and measure fluctuations in effective connectivity between sources of neural activity. this method is complementary to the lagged coherence method in using advanced multivariate measures of directed information flow (e.g., Granger Causality Modeling and Directed Transfer Function modeling). SIFT functions will be used to examine changes in effective connectivity among the different experimental conditions and between treatment groups, using bootstrap/resampling techniques to correct for multiple comparisons. 'Network Projection', which is an extension of Measure Projection (Bigdely-Shamlo et al., 2013) will be performed using pairwise connectivity to measure input, which is computed by blurring the dipole locations and loading them with connectivity measures. The brain space will then be segmented into 88 pre-defined anatomical ROIs provided by automated anatomical labeling (AAL) obtained from WFU Pick Atlas (which is a SPM plugin). At this point, all the subjects have 88x88 matrix of connectivity values, therefore every subject has 7744 (88x88=7744) possible combinations of directed causal flows. Diagnostic condition differences will be computed and bootstrap statistics will be performed for each condition to control for type 1 error. MRI scanning and analyses. Structural and functional MRIs including DTI will be acquired using a head-only Siemens Prisma-FIT 3 Tesla scanner at the UCLA Ahmanson-Lovelace Brain Mapping Center. This scanner is the latest Siemens 3T product modeled after the custom made Skyra system originally developed for the Human Connectome Project (HCP). The investigators will utilize the optimized HCP protocols (http://www.humanconnectome.org/) to ensure that the connectivity data are collected and processed with state-of-the-art methods that will be interpretable in the context of other ongoing NIH research. Total scanning time will about one hour. Structural MRI (sMRI). A whole brain structural scan will be obtained for registration with EEG data and evaluating morphology (MPRAGE, TR=2300 ms, TE=3 ms, FOV=256 mm, 208 slices, .8 x .8 x .8 mm voxels). The standardized HCP sMRI protocols include both 3D T1- and T2-weighted (T1w and T2w) high resolution scans. Analyzed together, these sequences improve the automated extraction of brain features, though both T1 and T2 scans may also be examined separately. Diffusion tensor MRI (DTI). The acquisition of DTI data will be used to measure the structural connectivity of white matter. DTI will be acquired with two diffusion-weighted gradients of b=0 and 1000 s/mm2 in 98 orthogonal directions (TE = 89.2 ms, TR = 3222 ms, FOV=210mm, 92 slices, 1.5 x 1.5 x 1.5 mm voxels). The HCP also uses HARDI (High Angular Resolution Diffusion Imaging) that provides greater precision for mapping the direction of fibers in regions where fibers are crossing and allows for probabilistic tractography to estimate fiber trajectories from ROIs defined with other imaging modalities (sMRI, fMRI, or EEG). These same acquisition protocols will be used for this project. Resting-state fMRI (fMRI). fMRI will be used to characterize resting state network function in a manner complementary to EEG. The investigators will use an echo-planar sequence (72 axial slices per volume, 104×90 matrix \[2.0×2.0×2.0mm3\], FOV=208mm, TE = 33.1ms, TR = 720ms, FA = 52°, 420 volumes) to gauge resting-state BOLD signals. The HCP fMRI multiband EPI sequence offers excellent spatial and temporal resolution and contain anterior-posterior and posterior-anterior phase encoding to correct for anatomic distortions. MRI image processing. The investigators already have installed the HCP pipeline developed by the Washington University group and have processed HCP-compliant data collected at UCLA using this pipeline, achieving high quality tissue segmentation on the T1w MRI, white matter tractography using HARDI, and resting brain networks based on fMRI. Each pipeline is described separately below. Structural MRI. Three HCP pipelines are applied to sMRI (T1w \& T2w) data. The PreFreeSurfer pipeline registers T1w and T2w images, performs bias-field correction, and normalizes these images to MNI space. The FreeSurfer pipeline recreates FreeSurfer's reconall pipeline, which segments tissue, reconstructs cortical surfaces, and aligns images to standard volume and surface atlases. DTI. The Diffusion Preprocessing pipeline performs standard preprocessing steps using FSL, including equalizing intensity of b0s across runs, removing EPI and eddy-current distortions, correcting for motion and gradient nonlinearities, and registration to T1w images. The investigators perform postprocessing using BrainSuite and its BrainSuite Diffusion Pipeline (BDP) (http://brainsuite.org), which enables us to compute deterministic tractography using the ODF data to generate whole-brain diffusion tracts. The investigators will then compute the average fractional anisotropy (FA), mean diffusivity (MD), radial diffusivity (RD), and number of tracts along the white matter pathways connecting the anatomical ROIs (specific circuits associated with depression, including tracts originating from the subgenual ACC and DLPFC). Resting state fMRI. Preprocessing protocols for functional data include two pipelines. The first, fMRIVolume, corrects images for spatial distortions and motion, registers to T1w images, corrects for subject motion, among other standard steps. The fMRISurface pipeline aligns the output of fMRIVolume to cortical surfaces, including mapping onto a "grayordinate" system to facilitate comparison and analyses across subjects. The investigators perform resting state fMRI postprocessing using a novel method that is sensitive to spatial differences between subjects. For graph theoretic analysis of brain networks, The investigators will use independent components computed by ICA as well as structural ROIs based on the Destrieux atlas as nodes, and the connectional relationships between each pair of components will be evaluated by partial correlation analysis. The brain networks will be characterized by network parameters computed from their adjacency matrixes to quantitatively describe topology of networks, including 1) node degree, degree distribution and assortativity; 2) clustering coefficient and motifs; 3) path length and efficiency; 4) connection density or cost; 5) hubs, centrality and robustness; and 5) modularity. MRI-EEG image integration. In order to identify the location of the F3 electrode scalp site at which TBS is being delivered into the MRI space, as well as fully integrate the EEG and MRI datasets, The investigators will employ the Advanced Neuro Technology Xensor 3D electrode digitizer system (ANT Neuro; Enschede, Netherlands). This system uses a digitizing wand with reflective markers and an infrared camera to identify the 3D locations of all 64 EEG electrodes and three major fiducials (nasion, left and right preauricular points) for each participant in less than 10 minutes. The 3D digitized locations then are coregistered to the Montreal Neurological Institute (MNI-Colin27) template brain (Montreal Neurological Institute, Montreal, Canada. URL www.bic.mni.mcgill.ca/brainweb). Electrode locations can be used in the Visor 2.0 neuronavigation system (ANT Neuro; Enschede, Netherlands) to visualize the stimulation target and associated scalp site that was actually used during TBS in each individual. After loading each subject's raw MRI images into the Visor software, the relationship of the F3 stimulation site to the DLPFC target, as well as any individual electrode to underlying neuroanatomy, can be identified using the MNI coordinates.