This was a cross-sectional study of depressed, treatment-seeking adolescents. All participants underwent diagnostic and clinical assessments and TMS neurophysiology measures. A subset of eligible participants underwent MRI/MRS. The TMS measures were obtained during a single session, and MRI/MRS was subsequently performed on the same day. No treatment was provided as part of the study protocol, but following completion, participants were referred for additional care if clinically appropriate. Prior to enrollment, the study protocol and all procedures were approved by the Mayo Clinic institutional review board. Participants under the age of 18 years granted written assent, and parents or legal guardians gave written informed consent. Participants who were 18 years of age provided written consent.

The sample was comprised of 24 participants, aged 11–18 years, with clinically significant depressive symptoms. All participants were right-handed as determined by the Edinburgh Inventory (Oldfield, 1971). Additionally, all participants were fluent in the English language, and at least one parent or guardian also was fluent in English. Exclusion criteria included a primary psychiatric diagnosis other than a depressive disorder; history of unprovoked seizures, febrile seizures, seizure disorder or family history of epilepsy; significant positives on the TMS Adult Safety Screen (Keel et al., 2001); implanted metal in the head or other contraindications to MRI/MRS (of note, adolescents were eligible for participation in the study if they had metallic orthodontic hardware, although they were excluded from the MRI/MRS portion of the protocol); prior brain surgery or risk for increased intracranial pressure; pregnancy or suspected pregnancy; or other unstable medical conditions. Additionally, participants at risk of imminent self-harm or suicide as determined by a study psychiatrist (CPL, JLVV, PEC) were not eligible for participation.

Participants’ depressive symptoms were assessed with two instruments. The Children’s Depression Rating Scale—Revised (CDRS-R; Poznanski et al., 1984) is a 17-item, clinician-rated scale that incorporates both parent- and adolescent-reported data into composite scores for specific depressive symptoms. The CDRS-R clinician total score is reported. The Quick Inventory of Depressive Symptomatology—Adolescent (17 Item)—Self-Report (QIDS-A₁₇-SR) is a self-report version of the clinician-rated QIDS-A₁₇-C (Bernstein et al., 2010), a 17-item assessment of depressive symptoms. The QIDS- A₁₇-SR total score is reported.

All participants underwent TMS to assess cortical inhibition and excitability. Single- and paired-pulse stimulation paradigms were applied to the left primary motor cortex, and measures were obtained via surface electromyography (EMG) of the contralateral (right) abductor pollicis brevis (APB) muscle during TMS stimulation. Participants and research personnel wore earplugs throughout the procedure.

Two Magstim 200 stimulators with a BiStim module (Magstim Co. Ltd., Whitland, Wales, UK) and a figure-of-eight electromagnetic coil (70 mm diameter for each coil loop) were used. The coil was positioned tangentially on the surface of the scalp, superficial to the left primary motor cortex. In order to locate the optimal coil position, the position was adjusted in 1-cm increments while stimulating with single magnetic pulses and observing for the maximal contraction of the contralateral APB while the muscle was at rest. After determining the optimal position of the coil for APB stimulation, pulse intensity was increased gradually until the MEP on EMG reached at least 50 microvolts in 5 of 10 trials (Figure 2A); the intensity at which this occurred was defined as the RMT (Rossini et al., 2015). After establishing the RMT, other single- and paired-pulse measures were acquired.

Paired-pulse measures involved the application of two paired magnetic pulses in rapid succession to the left primary motor cortex, again while the APB was at rest. For both SICI and ICF paradigms, a subthreshold conditioning stimulus (80% of RMT) was followed by a suprathreshold test stimulus (calibrated to result in an MEP with a peak-to-peak amplitude of 1 mV on EMG). The interstimulus intervals (ISIs) between conditioning and test stimuli were 2 ms and 4 ms for the SICI paradigm (Figure 2B) and were 10 ms, 15 ms, and 20 ms for the ICF paradigm (Figure 2C). For both SICI and ICF, the amplitude of the conditioned MEP occurring after the test stimulus is expressed as a ratio to the mean unconditioned MEP amplitude. Paired-pulse measures were obtained in a randomized and counterbalanced fashion, with a total of twelve trials at each ISI. MEP amplitude ratios were averaged for each ISI.

To measure the CSP duration, participants maintained tonic, voluntary contraction of the right APB (at 20% of maximum contraction strength, measured by hand-held dynamometer), during which single magnetic pulses at 140% of the RMT were applied to the left primary motor cortex. The CSP was recorded via EMG of the right APB following each pulse (Figure 3). CSP duration was averaged over the 10 trials performed.

At the conclusion of the TMS session, the position of the center of the TMS coil was marked with a vitamin E capsule, affixed to the scalp with adhesive, as a point of reference for the primary motor cortex during the ¹H-MRS scan.

To assess relationships between TMS measures and glutamatergic metabolite concentrations, correlation coefficients were calculated for each TMS measure and [Glx] in each voxel, with an α level of 0.05 (two-tailed) for significance. Distributions of TMS measures of cortical excitability and inhibition and ¹H-MRS-measured Glx concentrations in each voxel were assessed for normality using the Shapiro-Wilk test. [Glx] in both voxels and all TMS neurophysiologic measures demonstrated normal distributions by the Shapiro-Wilk test, with the exception of the MEP amplitudes in the 2-ms SICI paradigm (W = 0.8507, p = 0.0227) and the 20-ms ICF paradigm (W = 0.7281, p = 0.0007). However, visual inspection of histograms and Q-Q plots revealed substantial skewness in many of the distributions. In light of the small sample size and low power to reject a null hypothesis of normality, we elected to report both the parametric Pearson product-moment correlation coefficient (r) and the non-parametric Spearman rank-order correlation coefficient (ρ) for all correlations between TMS measures and [Glx]. As this was an exploratory study aimed at detecting signals of relationships of interest for hypothesis generation, uncorrected p-values are reported and interpreted. However, for the interested reader, adjustment of p-values for multiple comparisons using the false discovery rate (FDR) method (Benjamini and Hochberg, 1995) was performed, and these are reported as well. JMP 10.0.0 software (SAS Institute, Inc., Cary, NC, USA) was utilized for all statistical analyses.

Twenty-four adolescents (16 female, 8 male) with a mean age of 15.2 ± 1.8 years enrolled in the study. Of these, 14 participants (11 female, 3 male; mean age 14.7 ± 1.9 years) completed ¹H-MRS procedures. Participant demographics, including mean depression severity scores on the CDRS-R and QIDS-A₁₇-SR, are reported in Table 1. Fifteen participants were taking psychotropic medications at the time of the study; this included 14 taking selective serotonin reuptake inhibitors, two taking psychostimulants (which were held on days of TMS and ¹H-MRS testing), one taking an atypical antipsychotic agent, and two taking sedative-hypnotics. For a full list of participants’ psychotropic medications and doses at the time of the study, see Supplementary Table S1.

Mean values and standard deviations for the TMS measures (RMT, CSP, SICI-2, SICI-4, ICF-10, ICF-15, and ICF-20) are reported in Table 2. In the left primary motor cortex, mean [Glx] obtained via ¹H-MRS was 11.35 ± 1.42 institutional units. In the medial ACC, mean [Glx] was 18.45 ± 5.03. Other neural metabolite concentrations obtained via ¹H-MRS are reported in Supplementary Materials (Table S2).

The question of whether properties of TMS-induced MEPs originating in the primary motor cortex correspond to neurotransmitter levels in the same region of cortex is intuitive, and indeed these relationships have been the focus of two prior studies in healthy adults and one study in adults with mild brain injury. Stagg et al. (2011) investigated the relationships of GABA and glutamate concentrations in the motor cortex to various TMS measures of excitability and inhibition. The authors found that the MEP I-O curve, considered to be a global measure of excitability, was positively correlated with both motor cortex glutamate and motor cortex GABA concentrations; however, the MEP I-O—glutamate relationship did not survive correction for GABA concentration. In another study, Tremblay et al. (2013) noted a significant positive association between the CSP duration and motor cortex [Glx], a relationship that remained significant after controlling for GABA concentration. No significant correlations between ¹H-MRS-measured [GABA] and CSP, SICI, or LICI were found. In another study, Tremblay et al. (2014) found a correlation between LICI and motor cortex [GABA] in young adult athletes with histories of concussion, a relationship that was absent in healthy control comparators, suggesting potential alterations in GABA-mediated inhibitory physiology due to brain pathology.

In the present study, we found a strong correlation between the RMT and motor cortex [Glx]. Interestingly, the relationship was positive (i.e., higher levels of [Glx] corresponded to higher thresholds for evoked potentials), which is somewhat counterintuitive and also differs from the direction of the relationship between RMT and ACC glutamate concentration (which did not quite meet the threshold for significance). Numerous factors could account for this. The physiology of the RMT is complex; both cortico-cortical and corticospinal neurons are involved (Sandbrink, 2008), and both voltage-gated sodium channels and glutamatergic receptors have been demonstrated to mediate the threshold level (Ziemann et al., 2015). The RMT may best be described as a global index of cortical and spinal excitability that has many inputs, including glutamate-mediated neurotransmission. Thus, it perhaps would be overly simplistic to assume that an increase in glutamate necessarily leads to a decrease in RMT, as autoregulatory feedback loops and other mechanisms may mediate the effect of local glutamate concentration on the RMT. Furthermore, concentrations of glutamate or glutamate + glutamine in a region of cortex may not be particularly accurate indices of the pool of glutamate available for excitatory functions. Vesicular glutamate is not well detected by ¹H-MRS, and the concentrations measured by the current spectroscopic techniques may better represent the total neuronal and glial stores of glutamate and glutamine available for both synaptic and metabolic functions (Yüksel and Öngür, 2010; Maddock and Buonocore, 2012). Magnetic fields greater than the 3 T employed for ¹H-MRS in this study are necessary to separate signals for glutamine and glutamate reliably. Additionally, there is growing recognition of the role of neuroactive chemicals in the extracellular matrix on intercellular communication in the brain (Marcoli et al., 2015). Nevertheless, it is notable that glutamatergic metabolites in the primary motor cortex demonstrated associations with the RMT, and further research is necessary to characterize the mechanisms by which glutamate mediates this aspect of excitatory physiology.

The few prior studies that examined relationships between TMS measures of cortical excitability and inhibition and cortical glutamatergic concentrations selected ¹H-MRS voxels only in the motor cortex. There is robust evidence from prior ¹H-MRS studies that glutamate metabolism in the ACC is disrupted in youth with MDD (Mirza et al., 2004; Rosenberg et al., 2004, 2005; Kondo et al., 2011), but it is not clear what effect this has on physiologic functions mediated by the motor cortex. Consequently, we investigated the relationships between ACC glutamate concentrations and TMS measures of cortical excitability and inhibition.

ACC [Glx] did not demonstrate significant correlations with inhibitory measures (CSP, SICI). However, [Glx] did demonstrate a strong positive correlation with MEP amplitude in the 10-ms ICF paradigm (in nonparametric correlations) and in the 20-ms paradigm (parametric), with higher concentrations of glutamate in the ACC corresponding to greater facilitation of the MEP. One should be cautious in assuming that increases in excitatory neurotransmitters in one area of cortex necessarily lead to increased excitatory physiology in another region. Nonetheless, these findings suggest that increased glutamate (or increased glutamate + glutamine stores) in the ACC may have downstream effects on the excitatory physiology mediated by the motor cortex, which is plausible in light of the dense connectivity between these two cortical regions (Paus, 2001). Replication, particularly in healthy control adolescents, would help to establish such a relationship. It is unclear why the nonsignificant ICF-ACC glutamate correlation was negative at the 15-ms ISI, but this could be related to subtle differences in the physiology of facilitatory responses at different ISIs, something that has been observed with GABA-mediated SICI paradigms (Fisher et al., 2002; Ilić et al., 2002; Hanajima et al., 2003; Stagg et al., 2011).

In contrast to these ICF findings, the correlation between ACC glutamate and the RMT, which did not quite reach significance, was negative. This differs from the statistically significant, positive correlation between the RMT and [Glx] in the primary motor cortex. The negative relationship observed between glutamate concentrations in the ACC and the RMT could be related to glutamate-mediated regulatory circuits between the ACC and the motor cortex. As noted above, the physiology of the RMT involves many elements, notably the voltage-gated sodium channels, and it is possible that within the motor cortex these dominate the excitability responsible for motor threshold, with glutamate playing a lesser role. Future studies investigating the relationship between ACC glutamate and TMS measures of cortical excitability in healthy adolescent populations may help to understand the excitatory physiology of ACC-motor cortex connections further.

The current study has several notable limitations. First, the sample size was small, particularly the subset that completed both TMS and ¹H-MRS procedures. Although similar in size to the few previous TMS-MRS studies, this limited our statistical approaches. Correlations that were significant based on raw p-values did not survive correction for multiple comparisons, which is unsurprising in light of the number of correlations performed and the small number of participants. Thus, the findings should be interpreted with caution. Larger studies with greater power are warranted, particularly as this could permit controlling for potential confounds. Additionally, the sample did not include healthy controls. In the absence of data from healthy youth, it is not possible to determine to what extent depression may impact these findings, and future studies should endeavor to compare healthy participants and participants with psychopathology. The participants in this study were not medication-naïve, unlike many of the previous studies utilizing ¹H-MRS or TMS methods. Although concurrent psychotropic medications or history of pharmacologic interventions do pose potential confounds, the sample of our study is potentially more representative of populations seen in typical clinical practice. An additional potential confound is the impact of the menstrual cycle on both TMS and ¹H-MRS data, particularly as the sample was comprised predominantly of female adolescents. Prior TMS research has shown cortical inhibition to increase during the luteal phase in healthy women (Smith et al., 1999, 2003) but to decrease in women with premenstrual dysphoria (Smith et al., 2003), and ¹H-MRS studies have shown alterations in cortical GABA levels across the menstrual cycle (Epperson et al., 2002, 2005; Harada et al., 2011). While menstrual phase and sex hormone levels were not assessed in the female adolescent participants in the present study, future investigations should consider the value of additional measures to assess and control for sex- and menstrual-related effects.

Previous research has established that significant changes occur in excitatory and inhibitory circuitry systems during the childhood and adolescent years. In early life, expression of various subunits that compose GABA_A receptors has been found to change over time (Duncan et al., 2010), while GABA_A receptor density changes at differing rates in various brain regions into adulthood (Chugani et al., 2001). GABA receptors also show paradoxically excitatory activity in early childhood, with a gradual shift toward inhibition with age (Rakhade and Jensen, 2009). Glutamatergic neurotransmission also undergoes developmental shifts, including alterations in expression of glutamatergic receptor types, as well as modifications in receptor subunit composition affecting ion permeability (Silverstein and Jensen, 2007). In the TMS literature, one of the most consistent age-related findings is that the RMT diminishes with increasing age, both in neurologically- and psychiatrically-healthy individuals (Nezu et al., 1997; Heinen et al., 1998; Moll et al., 1999; Garvey et al., 2003; Mall et al., 2004; Gilbert et al., 2011) as well as those with conditions such as attention-deficit/hyperactivity disorder (Gilbert et al., 2011), Tourette syndrome (Moll et al., 2006), and MDD (Croarkin et al., 2014). The impact of age on inhibitory TMS measures is less clear, with some studies finding CSP duration (Heinen et al., 1998; Moll et al., 1999), SICI (Mall et al., 2004), and LICI (Croarkin et al., 2014) to increase with age, while others have found no significant relationship with age for CSP (Garvey et al., 2003; Gilbert et al., 2011; Croarkin et al., 2014) or SICI (Moll et al., 1999; Gilbert et al., 2004, 2011; Croarkin et al., 2014). The statistical methods of the present study do not permit controlling for age; however, future studies utilizing neurophysiologic TMS in children and adolescents could consider regression analyses with age as a covariate. Further research is warranted to characterize more conclusively the developmental trajectories of TMS-measured cortical excitability and inhibition in the context of MDD.

In the processing of ¹H-MRS data, there are inherent challenges in separating signals for Glu, Gln and GABA. Our ¹H-MRS procedures did not permit identification of a distinct GABA signal, in contrast to other TMS-MRS studies (Stagg et al., 2011; Tremblay et al., 2013, 2014). However, there is significant interaction between GABAergic and glutamatergic systems, both at the level of their neurochemical pathways and in the excitatory-inhibitory balance achieved by circuits of GABAergic and glutamatergic neurons. Additionally, the interpretation of glutamatergic metabolite concentration data is difficult, as ¹H-MRS-measured concentrations reflect the combined neuronal and glial stores of glutamate and glutamine, and also are believed to represent primarily cytoplasmic (as opposed to mitochondrial or vesicular) concentrations (Maddock and Buonocore, 2012). While total quantities of glutamatergic metabolites would be expected to influence glutamate-mediated (and GABA-mediated) processes, neuronal vesicular glutamate concentrations may be more immediately related to the neurotransmission underlying the excitatory and inhibitory phenomena measured by TMS. Moreover, the spectroscopic data were obtained at a single point in time. This is relevant in light of evidence from human ¹H-MRS studies in which glutamatergic metabolites have been demonstrated to fluctuate in response to sensory stimuli (Mullins et al., 2005; Mangia et al., 2007; Gussew et al., 2010) and physical activity (Maddock et al., 2011), suggesting that glutamate concentrations are state-dependent and dynamic (Maddock and Buonocore, 2012). Although the ¹H-MRS data in the present study were obtained on the same day as TMS testing, they were not concurrent with the excitatory and inhibitory processes induced by TMS.