Introduction
Methamphetamine is a psychostimulant with high abuse potential, which has seen a major resurgence across most demographic groups in the United States since 2010, with the annual use rate increasing to 59.7 per 1,000 adults, and overdose deaths involving stimulants increasing by over 300% from 2013 to 2019 (Jones et al., Reference Jones, Houry, Han, Baldwin, Vivolo-Kantor and Compton2022; Jones, Compton, & Mustaquim, Reference Jones, Compton and Mustaquim2020). Despite this, there is a dearth of literature examining the long-term neurobiological ramifications of methamphetamine use in humans. Given the widespread use of methamphetamine and its associated societal impact, understanding the effects of chronic methamphetamine use on brain function and illuminating the key factors that may trigger further use and/or reward-seeking behavior is critical and may provide new directions for treating methamphetamine addiction.
While our understanding of the effects of chronic methamphetamine use at the circuit-level is limited, the molecular effects have been extensively studied in model systems. Methamphetamine use triggers elevations in monoamine neurotransmitters such as dopamine, serotonin, and norepinephrine in the synapse (Jayanthi, Daiwile, & Cadet, Reference Jayanthi, Daiwile and Cadet2021; Kish, Reference Kish2008), leading to both acute and chronic cognitive effects. Acutely, methamphetamine use has been associated with behavioral disinhibition, increased energy, and short-term improvements in attention, but chronic methamphetamine use has been linked with impaired executive function and poor impulse control, leading to a collection of adverse mental health consequences (Hart et al., Reference Hart, Gunderson, Perez, Kirkpatrick, Thurmond, Comer and Foltin2008; Hassan, Wearne, Cornish, & Goodchild, Reference Hassan, Wearne, Cornish and Goodchild2016; Rusyniak, Reference Rusyniak2013). Further, methamphetamine use induces neuroinflammatory cascades that may ultimately lead to neurotoxicity and damage to nerve terminals (Moratalla et al., Reference Moratalla, Khairnar, Simola, Granado, García-Montes, Porceddu, Tizabi, Costa and Morelli2017; Sj, Reference Sj2008). While human neurobiological studies remain relatively scarce, a handful have demonstrated functional alterations in the frontal, striatal, and limbic regions (Dakhili et al., Reference Dakhili, Sangchooli, Jafakesh, Bidoki, Soleimani, Batouli, Kazemi, Faghiri, Oghabian and Ekhtiari2021; London, Kohno, Morales, & Ballard, Reference London, Kohno, Morales and Ballard2015; Zhou et al., Reference Zhou, Hu, Wang, Yang, Li, Ma, Wu, Chen, Yang, Hao, Wang, Li, Peng, Liu and Yang2023), accompanied by reports of mild cognitive impairments, particularly in domains involving attention, processing speed, and inhibitory control (Boutros et al., Reference Boutros, Gooding, Sundaresan, Burroughs and Johanson2006; Dean, Groman, Morales, & London, Reference Dean, Groman, Morales and London2013; Fitzpatrick, Rubenis, Lubman, & Verdejo-Garcia, Reference Fitzpatrick, Rubenis, Lubman and Verdejo-Garcia2020; Hart, Marvin, Silver, & Smith, Reference Hart, Marvin, Silver and Smith2012; Jan et al., Reference Jan, Lin, McLaren, Kirk, Kydd and Russell2014).
Regular methamphetamine use is thought to disrupt feedforward and feedback processes within the cortico-basal-ganglia-cortical loops, ultimately altering plasticity and leading to a strengthening of cortical responsiveness to drug-related cues through Pavlovian conditioning (Everitt & Robbins, Reference Everitt and Robbins2005; Gremel & Lovinger, Reference Gremel and Lovinger2017). Altered plasticity in corticostriatal circuitry may further disrupt inhibitory control and reward processing, promoting the eventual shift to habitual drug use (Everitt & Robbins, Reference Everitt and Robbins2013). To probe such processes, neuroimaging studies have used substance-related cue reactivity paradigms, which have consistently shown heightened neural responses to salient drug cues among those with substance use disorders, and this hyperactivation has been linked to an increased likelihood of relapse (Courtney et al., Reference Courtney, Schacht, Hutchison, Roche and Ray2016; Dakhili et al., Reference Dakhili, Sangchooli, Jafakesh, Bidoki, Soleimani, Batouli, Kazemi, Faghiri, Oghabian and Ekhtiari2021; Ekhtiari, Kuplicki, Aupperle, & Paulus, Reference Ekhtiari, Kuplicki, Aupperle and Paulus2021; Liu et al., Reference Liu, Wang, Zhang, Xu, Shao, Chen, Yang, Liu and Yuan2021). Notably, most of this work has focused on subcortical reward circuitry and its associations with behavioral measures, with far fewer studies examining cortical mechanisms of inhibitory control immediately following the onset of drug cues.
One established method for assessing cortical inhibitory circuits in humans is through sensory gating paradigms, which involve presenting two identical stimuli in rapid succession. Typically, the neural response to the second stimulus is sharply suppressed, reflecting the filtering out or inhibition of redundant information to optimize resource utilization (Grunwald et al., Reference Grunwald, Boutros, Pezer, von Oertzen, Fernández, Schaller and Elger2003; Hsiao, Cheng, Chen, & Lin, Reference Hsiao, Cheng, Chen and Lin2013). Sensory gating has been demonstrated across multiple sensory modalities (Gjini, Sundaresan, & Boutros, Reference Gjini, Sundaresan and Boutros2008; Hsiao et al., Reference Hsiao, Cheng, Chen and Lin2013; Jones et al., Reference Jones, Hills, Dick, Jones and Bright2016) and work in the somatosensory cortices has shown that local gamma oscillations are essential for effective gating (Proskovec, Spooner, Wiesman, & Wilson, Reference Proskovec, Spooner, Wiesman and Wilson2020; Spooner, Eastman, Wiesman, & Wilson, Reference Spooner, Eastman, Wiesman and Wilson2020; Spooner et al., Reference Spooner, Wiesman, Proskovec, Heinrichs-Graham and Wilson2019). Such oscillations are highly dependent on the integrity of local GABAergic circuitry, particularly parvalbumin-positive interneurons, which support sensory gating through reciprocal excitatory-inhibitory loops with pyramidal cells (Bartos, Vida, & Jonas, Reference Bartos, Vida and Jonas2007; Buzsáki & Wang, Reference Buzsáki and Wang2012; Fries, Reference Fries2005; Hagiwara et al., Reference Hagiwara, Okamoto, Shigeto, Ogata, Somehara, Matsushita, Kira and Tobimatsu2010; Uhlhaas & Singer, Reference Uhlhaas and Singer2012). Magnetic resonance spectroscopy (MRS) studies further demonstrate direct links between GABA levels and the strength and peak frequency of gamma oscillations (Balz et al., Reference Balz, Keil, Roa Romero, Mekle, Schubert, Aydin, Ittermann, Gallinat and Senkowski2016; Edden, Muthukumaraswamy, Freeman, & Singh, Reference Edden, Muthukumaraswamy, Freeman and Singh2009; Kujala et al., Reference Kujala, Jung, Bouvard, Lecaignard, Lothe, Bouet, Ciumas, Ryvlin and Jerbi2015; Muthukumaraswamy et al., Reference Muthukumaraswamy, Edden, Jones, Swettenham and Singh2009). Consistent with this, human neurophysiological studies show that gamma power and peak gamma frequency are tightly coupled (Wilson et al., Reference Wilson, McDermott, Mills, Coolidge and Heinrichs-Graham2018) and that aberrant gamma oscillations in the somatosensory cortices are associated with impaired inhibitory control (Arif, Wiesman, Christopher-Hayes, & Wilson, Reference Arif, Wiesman, Christopher-Hayes and Wilson2021; Casagrande et al., Reference Casagrande, Lew, Taylor, Schantell, O’Neill, May, Swindells and Wilson2021; Schantell et al., Reference Schantell, Lulli, McDonald, Horne, John, Coutant, Okelberry, Glesinger, Arif, O’Neill, Bares, May-Weeks and Wilson2025; Spooner et al., Reference Spooner, Wiesman, Mills, O’Neill, Robertson, Fox, Swindells and Wilson2018; Thoma et al., Reference Thoma, Hanlon, Huang, Miller, Moses, Weisend, Jones, Paulson, Irwin and Cañive2007; Wiesman et al., Reference Wiesman, Mundorf, Casagrande, Wolfson, Johnson, May, Murman and Wilson2021; Wood et al., Reference Wood, Cummings, Jung, Patterson, Okada, Guo, O’Neill, Dapretto, Bookheimer and Green2021). Together, these findings strongly support the use of sensory gating as a proxy for inhibitory processing and establish that effective gating depends on the integrity of gamma oscillations via local GABAergic circuitry. Finally, although initially conceptualized as a preattentive, bottom-up inhibitory process, growing evidence indicates that while attention may not modulate time domain (i.e. evoked) gating responses, the strength of at least some oscillatory gating responses is modulated by attention (Gjini et al., Reference Gjini, Burroughs and Boutros2011; Kho et al., Reference Kho, Verkes, Eling, Zwarts, Ellenbroek and van Luijtelaar2003; Wiesman & Wilson, Reference Wiesman and Wilson2020). Critically, oscillations within the gamma band have been linked to multiple cognitive functions, including inhibitory control (Arif et al., Reference Arif, Wiesman, Christopher-Hayes and Wilson2021; Cheng et al., Reference Cheng, Chan, Niddam, Tsai, Hsu and Liu2016; Proskovec et al., Reference Proskovec, Spooner, Wiesman and Wilson2020; Schantell et al., Reference Schantell, Lulli, McDonald, Horne, John, Coutant, Okelberry, Glesinger, Arif, O’Neill, Bares, May-Weeks and Wilson2025; Spooner et al., Reference Spooner, Wiesman, Proskovec, Heinrichs-Graham and Wilson2019; Spooner, Eastman, et al., Reference Spooner, Eastman, Wiesman and Wilson2020), and are known to be modulated by attention (Jensen et al., 2007; McCusker et al., Reference McCusker, Wiesman, Schantell, Eastman and Wilson2020; Wiesman et al., Reference Wiesman, Mundorf, Casagrande, Wolfson, Johnson, May, Murman and Wilson2021). However, whether attentional biases toward substance-related cues may affect gamma oscillatory responses and/or sensory gating in people who use methamphetamine, and possible relationships with higher order cognitive function, have not been examined.
In the current study, we leveraged a dynamic functional mapping approach based on magnetoencephalography (MEG) and a novel substance-cue paradigm to determine whether regular methamphetamine use is associated with impaired gating and the impact of substance-related cues. We used MEG mapping because it has excellent reliability (Lew et al., Reference Lew, Fitzgerald, Ott, Penhale and Wilson2021; McCusker, Lew, & Wilson, Reference McCusker, Lew and Wilson2021; Wilson, Heinrichs-Graham, & Becker, Reference Wilson, Heinrichs-Graham and Becker2014), provides good spatial precision, and has the millisecond temporal resolution necessary to quantify gamma activity during sensory gating. Our primary hypotheses were that the methamphetamine use group (MG) would exhibit weaker gating responses compared to controls, that this effect would be accentuated by the presence of substance-related cues, and that the well-known negative relationship between gating strength and peak gamma frequency would be aberrant in the MG. We also hypothesized that these gating indices would be predictive of attention and processing speed abilities across both groups and that the MG would exhibit elevated spontaneous gamma levels during the prestimulus baseline period, consistent with dysregulated cortical excitability.
Methods
Participants
Participants were recruited via advertisements and referrals. Thirty-seven participants who regularly use methamphetamine met criteria for the MG, and 37 demographically matched participants with no illicit substance use were selected for the control group. Participants in the MG had frequent methamphetamine use over the past year (i.e. >30 episodes) and regular use over the past 6 months. Exclusionary criteria for both groups included any medical illness affecting CNS function, history of head trauma, pregnancy, and standard exclusion criteria for MEG/MRI (e.g. ferromagnetic implants). Given the high prevalence of polysubstance use and psychiatric symptoms in people who use methamphetamine, individuals with anxiety/depression symptoms were not excluded from either group, and those who used illicit substances other than methamphetamine were not excluded from the MG, as long as methamphetamine was their primary illicit substance. Schizophrenia and other severe disorders involving psychotic symptoms were exclusionary. Participants were asked to abstain from using illicit substances affecting CNS function on the days of their two visits. All participants completed a urinalysis to assess recent substance use, as well as a breathalyzer and a NarcoCheck saliva drug test (Montluçon, France) to ensure they were not under the influence during their study visits. The study protocol was approved by the Boys Town National Research Hospital’s Institutional Review Board (IRB), and all participants provided written informed consent after receiving a detailed description of the study.
Substance use assessments and neuropsychological testing
All participants completed a substance use interview that included the NIDA-Modified ASSIST and Module E of the Structured Clinical Interview for DSM-5 Research Version. Participants also completed the Alcohol Use Disorders Identification Test (AUDIT), Cannabis Use Disorders Identification Test-Revised (CUDIT-R), Fagerström Test of Nicotine Dependence, the Beck Depression Inventory-II (BDI-II), and questionnaires about their regular drug use (e.g. administration route, frequency, and total amount), as well as whether their past use of alcohol, tobacco, and/or other substances had recently changed. In addition, participants completed a battery of neuropsychological tests. Raw scores per assessment were converted to z-scores, and domain-specific composite scores were computed by averaging the assessment scores comprising each domain, yielding six cognitive domains (i.e. learning, memory, attention, executive function, language, and processing speed). See the Supplementary Materials for details.
Task paradigm and MEG data acquisition
All recordings were conducted in a two-layer magnetically shielded room (Vacuumschmelze, Hanau, Germany). Participants were seated at 68° with their head positioned within the MEG sensor array as neuromagnetic responses were sampled continuously at 1 kHz with an acquisition bandwidth of 0.1–330 Hz using a MEGIN Neo MEG system (Helsinki, Finland) with 306 sensors, including 204 planar gradiometers and 102 magnetometers. During the scan, participants completed a novel substance cue paradigm in which either a neutral or a substance-related visual cue was presented during paired-pulse left median nerve stimulation. Each visual cue appeared 0.05 s before the first stimulation and remained on the screen for 1.2 s, with the images across the two conditions being matched on complexity, luminance, emotional valence, and the presence of people/faces in the images. The paired-pulse paradigm involved electrical stimulation of the left median nerve with the Digitimer DS7A constant-current stimulator system (Digitimer Ltd, Garden City, UK) at an amplitude 10% below the threshold for eliciting a thumb twitch, with an interstimulus interval of 0.5 s and an interpair interval that varied randomly between 3 and 3.3 s (Spooner, Eastman, et al., Reference Spooner, Eastman, Wiesman and Wilson2020; Spooner et al., Reference Spooner, Wiesman, Proskovec, Heinrichs-Graham and Wilson2019, Spooner et al., Reference Spooner, Taylor, L’Heureux, Schantell, Arif, May, Morsey, Wang, Ideker, Fox and Wilson2021; Figure 1). Participants completed a total of 180 pseudo-randomized trials that were equally split between the neutral and substance-cue conditions, along with eight oddball trials where participants responded with a button press based on the presence of a musical instrument. The total MEG recording time was approximately 10 minutes.
Substance cues somatosensory gating MEG paradigm. Participants performed a novel substance cue paradigm that consisted of 90 trials per condition (neutral/substance cue), plus eight oddball trials containing a musical instrument, during paired-pulse electrical stimulation of the left median nerve. Participants were instructed to fixate on the crosshair and respond with a button press each time they detected a musical instrument. In each trial, a fixation cross was presented, and this was followed by a visual image that appeared 0.05 s before the first electrical stimulation and had a duration of 1.2 s. The second electrical stimulation occurred 0.5 s after the first. The interpair interval between paired pulses was 3.15 ± 0.15 s. IPI: interpair interval, ISI: interstimulus interval.
Figure 1. Long description
The diagram is divided into a dashed box on the left containing the experimental timeline and a photo of a MEG scanner on the right.
Timeline sequence from left to right:
* Fixation: A black screen with a white crosshair labeled 1.95 plus or minus 0.15 s.
* Visual Cues: An arrow points to two stacked images. The top image is a Neutral cue showing dried leaves and debris. The bottom image is a substance cue showing drug paraphernalia. A large arrow above these images is labeled Duration 1.2 s.
* Electrical Stimulation: To the right of the cues are two illustrations of a human forearm and hand. Each has a green dot on the wrist and a blue lightning bolt symbol.
* Timing intervals:
- A bracket between the visual cue and the first pulse is labeled 0.05 s.
- A bracket between the two pulses is labeled I S I 0.5 s.
- A large bottom bracket spanning the entire trial is labeled I P I 3.15 plus or minus 0.15 s.
- Text at the bottom states Total task time approximately 10 minutes.
On the right, a white and grey MEGIN TRIUX Neo MEG scanner is shown, with a dashed line connecting the experimental task box to the head-positioning area of the machine.
MEG data processing and statistical analyses
Our MEG and MRI data acquisition, structural coregistration, preprocessing, and sensor- and source-level analysis methods closely followed the pipeline detailed in previous manuscripts (Arif et al., Reference Arif, Wiesman, Christopher-Hayes and Wilson2021; Spooner et al., Reference Spooner, Wiesman, Proskovec, Heinrichs-Graham and Wilson2019; Spooner, Eastman, et al., Reference Spooner, Eastman, Wiesman and Wilson2020; Wiesman & Wilson, Reference Wiesman and Wilson2020) and are fully described in the Supplemental Materials. Each MEG dataset was individually corrected for head motion and subjected to noise reduction using the signal space separation method with temporal extension (Taulu & Simola, Reference Taulu and Simola2006). Artifacts were removed from the data using signal space projection, which was accounted for during source reconstruction (Uusitalo & Ilmoniemi, Reference Uusitalo and Ilmoniemi1997). The continuous magnetic time series was filtered between 0.5 and 200 Hz and divided into epochs of 1.8 s duration, with 0.0 s defined as the onset of the first stimulation and the baseline defined as the −0.5 to −0.1 s time window. Note that the visual image onset was at −0.05 s, and thus was outside of the baseline period. The epochs that contained artifacts were rejected using a fixed threshold method that was based on the full distribution of amplitude and gradient values of each participant. On average, 160.88 (SD = 10.11) trials per participant were used for further analysis. A linear mixed-effects (LME) model revealed no significant differences in the number of trials by group, condition, or the group-by-condition interaction (all ps > .05). Thus, source-amplitude values were not biased by SNR differences due to differential trial counts. Epochs were transformed into the time-frequency domain using complex demodulation and averaged over trials to generate time-frequency plots of mean spectral density. The time-frequency windows used for imaging were determined through a rigorous statistical analysis of the sensor-level spectrograms across the entire array of gradiometers. First, two-tailed paired-samples t-tests against baseline were conducted on each pixel per spectrogram, and this was followed up with nonparametric permutation testing (Ernst, Reference Ernst2004; Maris & Oostenveld, Reference Maris and Oostenveld2007) to control for multiple comparisons.
The resulting neural responses were imaged at an isotropic resolution of 4.0 mm using the dynamic imaging of coherent sources (DICS) beamforming approach (Gross et al., Reference Gross, Kujala, Hämäläinen, Timmermann, Schnitzler and Salmelin2001), with units (i.e. pseudo-t) that reflect noise-normalized power differences per voxel. These whole-brain 3D maps were transformed into standardized space, spatially resampled, and then averaged across participants, conditions, and stimulations to enable the peak voxel to be identified. Voxel time series (i.e. virtual sensors) were then extracted per participant using this peak voxel and used to compute the relative (i.e. baseline-corrected) and absolute (i.e. not baseline-corrected) power time series of the gamma response envelope (30–75 Hz) per participant, which were used to assess neural responses to the stimulation and spontaneous cortical activity during the baseline, respectively. We also estimated the peak gamma frequency per neural response by identifying the frequency with the greatest power within each stimulation window. To examine sensory gating, we used the relative time series to compute the gating ratio (Stimulation 2/Stimulation 1) for each participant per task condition. Estimates of spontaneous power during the baseline were based on the absolute power time series averaged from −0.5 to −0.1 s. Note that we examined gating using a ratio instead of directly comparing the amplitudes of each response because this approach is more widely used in the gating literature (Arif et al., Reference Arif, Wiesman, Christopher-Hayes and Wilson2021; Hsiao et al., Reference Hsiao, Cheng, Chen and Lin2013; Schabrun, Christensen, Mrachacz-Kersting, & Graven-Nielsen, Reference Schabrun, Christensen, Mrachacz-Kersting and Graven-Nielsen2016; Spooner et al., Reference Spooner, Wiesman, Mills, O’Neill, Robertson, Fox, Swindells and Wilson2018; Wiesman & Wilson, Reference Wiesman and Wilson2020), and because this approach reduces the impact of individual differences in overall somatosensory response amplitude. Note that higher gating ratios reflect poorer gating.
An LME model was used to examine the effects of group, task condition, and their interaction on the gating ratio, controlling for age. Since the power of spontaneous activity is estimated during the baseline (i.e. prior to visual image onset), group differences were assessed using a linear regression model that controlled for age. Relationships between gating ratio, peak frequency, and cognitive performance were probed using general linear models in an exploratory framework. Post hoc tests were corrected for multiple comparisons using a false discovery rate of .05 (Benjamini & Hochberg, Reference Benjamini and Hochberg2000).
Results
Two participants in the MG were excluded due to incidental findings during their clinical interview or excessively noisy MEG data. Thus, the final sample included 35 MG participants and 37 controls with no illicit substance use. Of the 35 participants in the MG, 34 met criteria for methamphetamine use disorder. For assessing group demographic differences, Mann–Whitney U tests were employed for continuous variables to account for nonnormal distributions, and chi-square tests were used for categorical variables. The two groups were matched on age (U = 703, P = .532), sex (χ2 = .05, P = .824), handedness (χ2 = .199, P = .656), race (χ2 = 2.489, P = .288), ethnicity (χ2 = 1.751, P = .417) and AUDIT score (U = 672, P = .972), but differed on education level (U = 167.5, P < .001), CUDIT-R score (U = 1191, P < .001), nicotine dependence (U = 1097.5, P < .001), and depression symptoms (U = 1019.5, P < .001). The latter factors were included as nuisance covariates as appropriate. Demographic details per group are shown in Table 1.
Participant characteristics and demographics
Table 1. Long description
The table is organized into four columns: Category, MG (N = 35), Controls (N = 37), and P value.
Demographics:
* Age: MG 45.2 (8.93); Controls 44.6 (12.2); P = .532.
* Sex (M/F): MG 18/17; Controls 20/17; P = .824.
* Handedness (R/L): MG 30/5; Controls 33/4; P = .656.
* Ethnicity (H/NH/U): MG 4/31/0; Controls 2/34/1; P = .417.
* Race (White %): MG 27 (77.14%); Controls 33 (89.19%); P = .288.
* Education (years): MG 12.71 (1.20); Controls 16.22 (2.08); P < .001.
Clinical Scores (M/SD):
* AUDIT: MG 6.37 (10.11); Controls 2.83 (2.50); P = .972.
* CUDIT - R: MG 9.89 (7.24); Controls .25 (1.05); P < .001.
* BDI: MG 21.42 (13.37); Controls 4.92 (4.78); P < .001.
* Fagerström nicotine dependence: MG 3.40 (2.69); Controls .11 (.67); P < .001.
* Methamphetamine involvement score: MG 24.59 (7.84).
Past 3-month drug use (M G only):
* Cannabis: 19 participants.
* Cocaine: 5 participants.
* Sedatives: 5 participants.
* Methamphetamine: 35 participants.
Neuropsychological assessments (M/SD):
* Learning: MG -0.82 (.83); Controls 0.10 (.62); P < .006.
* Memory: MG -0.61 (.67); Controls 0.06 (.44); P < .007.
* Attention: MG -0.83 (.81); Controls 0.22 (.66); P < .001.
* Processing speed: MG -0.67 (.60); Controls 0.32 (.66); P < .001.
* Language: MG -0.41 (.56); Controls 0.23 (.66); P < .017.
* Executive function: MG -0.88 (1.12); Controls 0.13 (.63); P < .007.
p-values indicate significance of between-groups comparisons for demographic measures using Mann–Whitney U tests to account for the nonnormal distribution of continuous variables and chi-square tests for categorical variables. ANCOVAs were performed controlling for education to compare group differences in neuropsychological assessments. Note that the reported number of participants for past 3-month drug use overlaps between drug categories, reflecting that some participants used multiple substances over the past 3 months, with methamphetamine as their primary substance since this was part of the inclusion criteria.
a Data missing for one control.
b Data missing for two participants in the MG.
AUDIT, ‘Alcohol Use Disorders Identification Test’; CUDIT-R, ‘Cannabis Use Disorders Identification Test – Revised’; BDI, ‘Beck Depression Inventory’.
Somatosensory responses: MEG sensor-level and source-level analyses
We first identified the spectral and temporal extent of the transient responses to the paired-pulse stimulation. Significant increases in gamma power were detected from 30–75 Hz after each somatosensory stimulation in MEG sensors near the sensorimotor and parietal regions (30–75 Hz, 0–50 and 500–550 ms poststimulation; P < .001, corrected; Figure 2). These time-frequency windows were subjected to beamformer source reconstruction, which revealed identical locations in the right primary somatosensory cortices for the first and second stimulation. Next, we grand-averaged these beamformer images, and the resulting peak voxel was used to extract voxel time series data per condition and participant.
Neural responses to somatosensory stimulation of the left median nerve. (a): The grand-averaged time-frequency spectrogram from a representative gradiometer sensor near the right sensorimotor cortices (MEG 1342), with the x-axis denoting time (s) and the y-axis showing frequency (Hz). The two dotted white lines represent the onset of the paired-pulse stimulation (0 and 0.5 s), and the white boundary boxes denote the time-frequency windows where neural responses to the somatosensory stimulation were strongest in our two-stage cluster-based permutation analysis. The prestimulus baseline period used in our analysis of spontaneous activity is denoted by the white dotted boundary box. The color bar on the top of the figure displays the amplitude scale for the signal power data in the spectrogram, which is expressed as percent change from baseline (−0.5 to −0.1 s). (b): Averaged beamformer images (pseudo-t) across both visual image types and stimulation separately for each group. The color scale bars on the top denote the response amplitude (in pseudo-t) for each cluster. (c): Baseline-normalized (i.e. relative) power time series envelope for the 30–75 Hz band. Note: MG, ‘methamphetamine use group’.
Figure 2. Long description
The figure is divided into three panels labeled a, b, and c.
Panel a is a time-frequency spectrogram. The x-axis represents Time in seconds from minus 0.5 to 1.0. The y-axis represents Frequency in Hz from 0 to 100. A color bar at the top ranges from minus 40 percent in cyan to 40 percent in yellow. A white dashed box on the left marks the baseline period. Two vertical white dotted lines at 0 and 0.5 seconds mark Stim 1 and Stim 2. High-intensity yellow and orange vertical bands appear immediately after each stimulus within white boundary boxes in the 30 to 75 Hz gamma range.
Panel b shows two axial brain MRI slices labeled Controls and MG. A color scale at the top indicates amplitude in pseudo-t from 7.5 in red to 12 in yellow. Both brain images show a concentrated orange-yellow cluster in the right sensorimotor cortex.
Panel c is a line graph of relative power in percent units on the y-axis against time in milliseconds on the x-axis. Two sharp peaks occur at approximately 50 and 550 milliseconds. The blue line representing Controls and the maroon line representing MG follow nearly identical paths, with the Controls peak slightly higher than the MG peak at both stimulus intervals.
Methamphetamine-related alterations in gating and spontaneous activity
For each participant, we computed the gating ratio (Stimulation 2/Stimulation 1) and used a 2 x 2 LME model to assess differences by group (controls, MG) and task condition (neutral, substance cue), with age included as a covariate of no interest. We found a significant main effect of group (F = 11.47, P = .001) and a group-by-condition interaction (F = 6.58, P = .01; Figure 3a). Post hoc testing revealed that the MG exhibited significantly weaker gating (i.e. higher gating ratio) compared to controls during the substance cue (b = .336, SE = .083, t = 4.03, p FDR = .0007) and the neutral condition (b = .174, SE = .833, t = 2.08, p FDR = .048). Further, MG participants showed significantly weaker gating in the substance cue compared to the neutral condition (b = .131, SE = .046, t = 2.86, p FDR = .01), while controls showed no such difference (t = .70, p FDR = .488). To ensure group differences in factors other than methamphetamine use were not driving these effects, we computed the same LME model with depression symptoms, nicotine dependence, cannabis use disorder symptoms, and educational attainment as nuisance covariates, which indicated that the main effect of group was no longer significant (F = .014, P = .907), while the interaction effect became stronger (F = 7.46, P = .008).
Substance cues modulate somatosensory gating in participants who use methamphetamine. (a) A linear mixed effects (LME) model revealed a significant task condition-by-group interaction. Post hoc testing indicated that those in the methamphetamine use group (MG) exhibited significantly weaker gating (i.e. higher gating ratio: Stim2/Stim1) relative to controls in both the neutral and substance cue conditions, and that only the MG had weaker gating during the substance cue relative to the neutral condition. Note that the interaction remained significant when controlling for age, depression symptoms, nicotine dependence, cannabis use disorder symptoms, and educational attainment. The violin plots show the probability density of the gating ratio for each group and condition, with individual residual points controlling for age. Boxes and asterisks display the upper and lower quartiles and the mean, respectively. Error bars denote the 95% confidence intervals. (b) The gating ratio significantly predicted processing speed domain scores above and beyond the effect of group and age. Note: Lower gating ratios are indicative of stronger (better) gating, while higher processing speed domain scores are indicative of better processing speed performance. (c) A general linear model revealed a significant peak gamma frequency-by-group interaction predicting the gating ratio. Post hoc testing revealed a significant negative relationship between peak gamma frequency and gating responses in the control group, while such a relationship was not found in the MG. Note: Peak gamma frequency was averaged across Stim1 and Stim2 responses; MG, methamphetamine use group. * P < .05, *** P < .005.
Figure 3. Long description
Panel a is a violin plot with a vertical Y-axis labeled Gating Ratio (residuals) ranging from -1 to 1. A vertical double-headed arrow on the left indicates higher values are worse gating and lower values are better gating. Two groups, Controls and M G, are compared across neutral image (teal) and substance cue (pink) conditions. The MG shows higher gating ratios than Controls in both conditions. Significant differences are marked with asterisks: a triple asterisk for the group-by-condition interaction and single asterisks for post hoc comparisons. An inset brain map shows a localized activation area in the right hemisphere with an amplitude scale from 8 to 14 pseudo-t.
Panel b is a scatter plot with Gating Ratio (residuals) on the Y-axis and Processing Speed (residuals) on the X-axis. A horizontal double-headed arrow indicates that higher X-axis values represent better Processing Speed. Data points and regression lines for Controls (blue) and MG (red) show a negative correlation, where lower gating ratios predict better processing speed. A dashed line represents the trend across groups.
Panel c is a scatter plot with Gating Ratio (residuals) on the Y-axis and Peak Frequency (residuals) on the X-axis. The MG (red) shows a slight positive slope (b = .005), while the Controls (blue) show a significant negative slope (b = -.01 asterisk). A dashed horizontal line at zero represents the across-groups baseline.
To ensure that this difference in the gating ratio was not driven by group differences in response power to either stimulation, the same LME was performed for Stim1 and Stim2. These results showed that the response power did not significantly differ by group (Stim 1: F = 0.023, p = 0.879; Stim2: F = 1.247, P = 0.268), condition (Stim1: F = 3.628, P = 0.061; Stim2: F = 0.498, P = 0.483), nor group by condition interaction (Stim1: F = 0.336, P = 0.564; Stim2: F = 3.487, P = 0.066). Likewise, an LME on the amplitude difference (Stim1–Stim2) showed the same group main effect (F = 6.056, P = 0.016) captured by the ratio, as well as a conditional main effect (F = 4.686, P = 0.034) of weaker differences for substance cue trials, which was driven by the MG, although the interaction was not significant. These tests support our use of the gating ratio. Finally, to assess specificity to oscillatory activity, we repeated the LME model on the gating ratio using time-domain (i.e. evoked) ratios and found no significant effects (see Supplementary Material S1).
To evaluate whether cognitive performance was related to these gating indices, we performed two separate general linear models including group and gating indices as predictors of cognitive performance scores on the attention and processing speed domains, controlling for age. We focused on these domains given evidence linking them to sensory gating (see section ‘Introduction’). Our findings indicated that gating responses significantly predicted processing speed above and beyond the effect of group and age (F = 4.72, P = .033; Figure 3b), but not attention (F = 0.03, P = .854).
Next, to probe for differences in spontaneous (baseline) neural activity, we used a linear regression approach with spontaneous gamma power (30–75 Hz, −0.5 to −0.1 s; Figure 4a) as the dependent variable, controlling for age. We found significantly stronger gamma power in the MG compared to controls in the primary somatosensory cortices (b = 12.10, SE = 3.76, F = 10.35, P = .002; Figure 4b). These results remained significant after controlling for all other covariates noted above (b = 18.30, SE = 7.54, F = 5.88, P = .018).
Elevated spontaneous gamma power during the baseline and differential relationship between gating ratio and peak frequency in the methamphetamine use group (MG). (a) Absolute power voxel time series were extracted from the same peak voxel used in the gating analysis (i.e., right primary somatosensory cortices) to estimate the spontaneous gamma power during the baseline period. These data were averaged across both conditions in each group, with the yellow window highlighting the baseline period. Note that we averaged across the visual image task conditions because this time period preceded the onset of the image in each trial. (b): Linear regression revealed significantly stronger gamma power in the MG relative to the controls, controlling for age. Note that this finding remained significant when controlling for age, depression symptoms, nicotine dependence, cannabis use disorder symptoms, educational attainment, and the removal of one datapoint that was nearly an outlier in the MG. The violin plots show the probability density for spontaneous gamma power. Boxes and asterisks display the upper and lower quartiles and the mean, respectively. Error bars denote the 95% confidence intervals. *** P < .005 * P < .05.
Figure 4. Long description
Panel a is a line graph showing Spontaneous Power residuals on the y-axis from 20 to 100 and Time in ms on the x-axis from minus 500 to 1000. Two lines are plotted: a blue line for Controls and a red line for MG. The MG line is consistently higher than the Controls line. A yellow shaded region from minus 500 to 0 ms is marked with three asterisks. Two sharp peaks occur in both groups at approximately 50 ms and 550 ms. An inset shows a transverse brain slice with a yellow and orange heat map indicating activity in the right primary somatosensory cortex. Panel b is a violin plot comparing Spontaneous Power residuals for Controls in blue and MG in red. The y-axis ranges from 0 to 200. The MG violin shows a higher mean and a wider distribution with a long upward tail compared to the Controls. Three asterisks between the plots indicate a significant difference. Each violin contains a box plot showing quartiles and individual data points.
Altered relationship between peak frequency and gating response in MG
Given the established linkage between peak gamma frequency and GABAergic function, we conducted an exploratory analysis to test for the relationship between the peak gamma frequency and sensory gating ratio. The peak frequency of the responses to Stim1 and Stim2 was averaged to obtain a mean peak frequency per participant. A general linear model was then performed with the sensory gating ratio as the dependent variable, and group, peak frequency, and their interaction as independent variables, controlling for age. We found a significant group-by-peak frequency interaction (F = 7.00, P = .01), such that there was the expected negative association between gating ratio and peak gamma frequency in controls (b = −.01, SE = .004, pFDR = .04), but no such relationship in the MG (b = .005, SE = .004, pFDR = .196; Figure 3c).
Discussion
We used a novel paradigm and an advanced dynamic functional mapping approach to investigate whether substance-related visual cues alter inhibitory processing among participants who use methamphetamine relative to controls. Our key findings were that the MG exhibited weaker sensory gating than controls and that substance cues further attenuated such gating, while the type of visual cues had no such effect in controls. We also found that gating indices significantly predicted processing speed above and beyond the effects of age and group, and that the negative relationship generally observed between peak gamma frequency and sensory gating in normative studies (Cheng et al., Reference Cheng, Chan, Niddam, Tsai, Hsu and Liu2016; Cheng, Liu, Hsu, & Tseng, Reference Cheng, Liu, Hsu and Tseng2021) was absent in the MG. Finally, sharply elevated spontaneous gamma activity was observed within the same tissue as the altered gating responses in the MG relative to controls. Together, these results provide novel insight into how substance-related cues can modulate inhibitory processes in those who regularly use methamphetamine.
One of our most important findings was the aberrant gating in those who regularly use methamphetamine. Sensory gating has been widely studied and reflects the capacity of the CNS to filter redundant sensory input (Cromwell, Mears, Wan, & Boutros, Reference Cromwell, Mears, Wan and Boutros2008). Thus, the weaker gating response observed in the MG relative to controls, which was further attenuated in the presence of substance cues, supports the notion that regular methamphetamine use is associated with aberrant inhibitory processes that generally gate out excessive sensory information, and that such gating is sensitive to attentional biases in the presence of salient substance cues. While gating was historically considered a preattentive, bottom-up inhibitory process, recent evidence suggests that attention modulates the strength of gating under certain conditions (Golubic et al., Reference Golubic, Jurasic, Susac, Huonker, Gotz and Haueisen2019; Wiesman & Wilson, Reference Wiesman and Wilson2020), and our findings in the MG strongly support this. More broadly, these findings align with animal models showing that pre-exposure to drugs increases cue-reactivity, and with human studies showing decreased response inhibition to cues that trigger maladaptive behaviors (Ekhtiari et al., Reference Ekhtiari, Kuplicki, Aupperle and Paulus2021; Jafakesh et al., Reference Jafakesh, Sangchooli, Aarabi, Helfroush, Dakhili, Oghabian, Kazemi and Ekhtiari2022; Liu et al., Reference Liu, Wang, Zhang, Xu, Shao, Chen, Yang, Liu and Yuan2021). As discussed further below, our sensory gating findings were all within the high-frequency gamma range, which has been broadly linked with GABA circuitry (Bartos et al., Reference Bartos, Vida and Jonas2007; Buzsáki & Wang, Reference Buzsáki and Wang2012; Fries, Reference Fries2005; Hagiwara et al., Reference Hagiwara, Okamoto, Shigeto, Ogata, Somehara, Matsushita, Kira and Tobimatsu2010; Uhlhaas & Singer, Reference Uhlhaas and Singer2012 ). Thus, future studies could extend these findings by integrating MRS measures of GABA and sensory gating, which could confirm this linkage and suggest possible therapeutic approaches. In addition, future MRS investigations could probe whether the relationship between gating and processing speed observed across participants in the current study is mediated by GABAergic activity. Previous work has also implicated striatal dopamine as crucial for encoding the incentive motivational value of drug cues in a bottom-up, cue-driven manner, a system known to be dominant among substance users (Volkow et al., Reference Volkow, Wang, Fowler, Tomasi, Telang and Baler2010). This view corresponds with the molecular action of methamphetamine, where excessive phasic dopamine release enhances the attribution of incentive value to associated contexts, thereby strengthening the connection between the cue and compulsive drug-seeking behavior (Chen & Bruchas, Reference Chen and Bruchas2021; Eckart & Bunzeck, Reference Eckart and Bunzeck2013; Narita et al., Reference Narita, Aoki, Takagi, Yajima and Suzuki2003). Thus, there are likely both dopaminergic and GABAergic aberrations involved in our finding that substance cues alter sensory gating in those who regularly use methamphetamine.
We also observed an altered relationship between gating and peak gamma frequency in the MG. Specifically, controls exhibited the expected negative relationship between higher peak gamma frequency and lower gating ratio (i.e. better gating; Cheng et al., Reference Cheng, Chan, Niddam, Tsai, Hsu and Liu2016, Reference Cheng, Liu, Hsu and Tseng2021), while this relationship was disturbed in the MG. This methamphetamine-related perturbation in the coupling between peak gamma frequency and inhibitory functioning could reflect deficits in hierarchical control that regulate the perception of incoming sensory stimuli. Note that a similar pattern has been reported in individuals with panic disorder (Cheng et al., Reference Cheng, Liu, Hsu and Tseng2021). Additionally, repetitive stimulant use can alter receptor responsiveness, while simultaneously heightening behavioral reactivity to sensory input, which could further disrupt the relationship between peak gamma frequency and inhibitory functioning, as the circuitry governing inhibitory control may become compromised (Evans et al., Reference Evans, Pavese, Lawrence, Tai, Appel, Doder, Brooks, Lees and Piccini2006; Gainetdinov et al., Reference Gainetdinov, Premont, Bohn, Lefkowitz and Caron2004). While the etiology remains unclear, such decoupling may be a result of alterations in GABAergic neuronal transmission, as the mechanistic link between GABA and high-frequency gamma oscillations could point to local GABA dysfunction (Heba et al., Reference Heba, Puts, Kalisch, Glaubitz, Haag, Lenz, Dinse, Edden, Tegenthoff and Schmidt-Wilcke2016; Wood et al., Reference Wood, Cummings, Jung, Patterson, Okada, Guo, O’Neill, Dapretto, Bookheimer and Green2021). Previous studies have demonstrated an association between GABA concentrations and gamma oscillations in human motor and occipital cortices (Edden et al., Reference Edden, Muthukumaraswamy, Freeman and Singh2009; Heba et al., Reference Heba, Puts, Kalisch, Glaubitz, Haag, Lenz, Dinse, Edden, Tegenthoff and Schmidt-Wilcke2016; Muthukumaraswamy et al., Reference Muthukumaraswamy, Edden, Jones, Swettenham and Singh2009). Animal studies also show alterations in GABA and glutamate levels, and a downregulation of GABA receptor-mediated inhibition following methamphetamine exposure (Mark, Soghomonian, & Yamamoto, Reference Mark, Soghomonian and Yamamoto2004; Padgett et al., Reference Padgett, Lalive, Tan, Terunuma, Munoz, Pangalos, Martínez-Hernández, Watanabe, Moss, Luján, Lüscher and Slesinger2012; Zhang et al., Reference Zhang, Lee, Xiong, Chen, Davidson, Wetsel and Ellinwood2006). Those with methamphetamine use disorder have also been shown to exhibit altered GABA and GABA/glutamate ratios, though these findings were in the prefrontal cortices (Su et al., Reference Su, Chen, Zhong, Jiang, Du, Xiao, Xu, Song and Zhao2020). As the present investigation did not directly measure GABA levels, this interpretation should be considered speculative. Future MRS studies should directly investigate the link between GABA and gamma oscillations in people who regularly use methamphetamine to better characterize the functional relationship between these neurophysiological and neurochemical processes.
Beyond sensory gating, we found elevated spontaneous gamma activity in the contralateral somatosensory cortices in the MG relative to controls during the baseline period. This apparent hyperexcitability may represent a pathological marker reflecting the effect of regular methamphetamine use on GABAergic circuitry, which is consistent with prior work showing elevations in resting cortical excitability in the motor cortices among stimulant users (Flavel, White, & Todd, Reference Flavel, White and Todd2012). Moreover, this pattern of elevated spontaneous gamma has been observed in other populations, such as healthy aging individuals and among people with HIV (Casagrande et al., Reference Casagrande, Lew, Taylor, Schantell, O’Neill, May, Swindells and Wilson2021, Reference Casagrande, Wiesman, Schantell, Johnson, Wolfson, O’Neill, Johnson, May, Swindells, Murman and Wilson2022; Petro et al., Reference Petro, Rempe, Schantell, Ku, Srinivas, O’Neill, Kubat, Bares, May-Weeks and Wilson2024; Spooner et al., Reference Spooner, Wiesman, Mills, O’Neill, Robertson, Fox, Swindells and Wilson2018, Reference Spooner, Wiesman, Proskovec, Heinrichs-Graham and Wilson2019, Reference Spooner, Wiesman, O’Neill, Schantell, Fox, Swindells and Wilson2020; Wiesman et al., Reference Wiesman, O’Neill, Mills, Robertson, Fox, Swindells and Wilson2018; Wilson et al., Reference Wilson, Lew, Spooner, Rezich and Wiesman2019), with corresponding data suggesting GABA decreases in healthy aging (Salsano et al., Reference Salsano, Glesinger, John, Garrison, Horne, Ende, Rice, Coutant, Kress, Petts, Oeltzschner and Wilson2025). Intriguingly, evidence from animal and human neuroimaging studies suggests that regular methamphetamine use is associated with accelerated aging and neuronal damage such as neuroinflammation, gliosis (Miller et al., Reference Miller, Bu, Gopinath, Martinez and Khoshbouei2021; Shukla & Vincent, Reference Shukla and Vincent2020), and increased permeability of the blood–brain barrier (O’Shea, Urrutia, Green, & Colado, Reference O’Shea, Urrutia, Green and Colado2014; Ramirez et al., Reference Ramirez, Potula, Fan, Eidem, Papugani, Reichenbach, Dykstra, Weksler, Romero, Couraud and Persidsky2009). Thus, the elevated spontaneous gamma activity could be related to neuroinflammation, which may contribute to deficits in inhibitory control and broader cognitive impairment. Future studies should incorporate inflammation-based biomarkers to further elucidate the mechanisms driving hyperexcitability in the somatosensory cortices.
The current study is the first to examine how substance cues modulate inhibitory processes in people who regularly use methamphetamine. While this study possesses significant strengths and novelty, several limitations warrant consideration. First, participants in the two groups differed in terms of their educational attainment, nicotine dependence, cannabis use, and depression symptoms, which could potentially introduce confounding effects. While follow-up testing indicated that these factors did not significantly affect our primary findings, a more robust approach would be to match across all of these variables, and that should be a goal of future work. Second, our study included only a single time point using MEG to assess inhibitory function, and future work should include multiple time points to assess whether gating is affected by the recency of last methamphetamine use (e.g. 2 days ago versus a week ago). Third, as with all studies of methamphetamine use, those in our MG generally used other illicit substances as well (i.e. polysubstance). While our inclusion criteria required that methamphetamine was the primary drug of use, this remains a limitation of this study and the majority of work in this area. Lastly, while relatively high rates of methamphetamine use were required for inclusion in the study, there remained high variability between participants regarding total duration and frequency of use, dose, substance purity, and route of administration, which complicates interpretation of frequency and duration of use effects. Future studies should aim to minimize such variability in the MG.
In conclusion, we found that substance-related visual cues modulate sensory gating in people who regularly use methamphetamine but not controls who do not use illicit substances, suggesting that cue-reactivity may interfere with inhibitory processing through both dopaminergic pathways and local GABAergic circuitry. Our study builds upon findings from the mesolimbic dopaminergic system, suggesting that drug-related cues may also interfere with the oscillatory gamma activity underlying sensory gating processes. Finally, our novel finding of elevated spontaneous gamma activity provides a valuable therapeutic target, as studies using electrical stimulation methods have shown promise in modulating spontaneous activity levels in sensory cortices (Arif et al., Reference Arif, Spooner, Wiesman, Proskovec, Rezich, Heinrichs-Graham and Wilson2020; Heinrichs-Graham et al., Reference Heinrichs-Graham, McDermott, Mills, Coolidge and Wilson2017; McDermott et al., Reference McDermott, Wiesman, Mills, Spooner, Coolidge, Proskovec, Heinrichs-Graham and Wilson2019; Wiesman et al., Reference Wiesman, Mills, McDermott, Spooner, Coolidge and Wilson2018; Wilson et al., Reference Wilson, McDermott, Mills, Coolidge and Heinrichs-Graham2018), and such approaches could be used to mitigate the hyperexcitability found in the sensory cortices of people who regularly use methamphetamine.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S0033291726105686.
Data availability statement
The data used in this article will be made publicly available through the COINS framework at the completion of the study (https://coins.trendscenter.org/).
Author contribution
Conceptualization: P.J.H., M.S., and T.W.W.; Data collection: R.G., L.K.H., A.T.C., H.J.O., G.C.E., G.M.G., K.R.D., O.R.C., M.C.H., L.D.V., M.E.V., and J.A.J.; Formal analysis: P.J.H., Y.A., and T.W.W.; Funding acquisition: T.W.W. and M.S.; Methodology and software: P.J.H., Y.A., and T.W.W.; Resources: T.W.W. and M.S.; Reviewing and editing: all authors; Supervision: T.W.W. and Y.A.; Writing – original draft: P.J.H. and T.W.W. All authors approved the final version for publication and can certify that no other individuals not listed as authors have made substantial contributions to the paper.
Funding statement
This work was supported by the National Institutes of Health through grants R01-DA056223 (T.W.W.), R01-DA059542 (T.W.W.), R01-DA059877 (T.W.W.), S10-OD28751 (T.W.W.), P20-GM144641 (T.W.W.), F31-DA056296 (M.S.), and R36-DA059323 (M.S.). The funders had no role in study design, data collection, analysis, decision to publish, or preparation of the manuscript. The data presented in this manuscript have not been published or presented elsewhere.
Competing interests
The authors of this manuscript acknowledge no conflicts of interest, financial or otherwise.
Open access
