Clavulanic Acid Decreases Cocaine Cue Reactivity in Addiction-Related Brain Areas, a Randomized fMRI Pilot Study

Abstract

Background

Preclinical studies show that clavulanic acid (CLAV) inhibits cocaine self-administration. This study investigates the effect of CLAV on regions of brain activation in response to cocaine cues during functional magnetic resonance imaging (fMRI) in participants with cocaine use disorder (CUD).

Methods

A double-masked, placebo-controlled clinical trial with thirteen individuals with severe CUD who were randomized to treatment with CLAV (N = 10, 9 completers) 500 mg/day or matched placebo (PBO) (N = 3) for 3 days. fMRI was used to assess brain reactivity to 18 alternating six-second video clips of cocaine or neutral scenes. In this paradigm, participants were exposed to three different stimulus conditions: NEUTRAL, WATCH (passive watching), and DOWN (actively inhibiting craving while watching).

Results

Participants who received CLAV demonstrated a significant reduction in brain activity in the anterior cingulate gyrus (p = 0.009) and the caudate (p = 0.018) in response to DOWN cocaine cues. There was a trend toward lessened cue reactivity in other regions implicated in CUD.

Conclusion

CLAV reduced the response of the brain regions associated with motivation and emotional response during the DOWN condition compared to PBO, suggesting CLAV may strengthen voluntary efforts to avoid cocaine use. This pilot data supports the use of CLAV for CUD. (Trial registered in ClinicalTrials.gov NCT04411914).

Cocaine use disorder (CUD) is a devastating disease with no FDA-approved medications for treatment. In preclinical models, cocaine-associated drug cues trigger glutamate release¹ and clavulanic acid (CLAV), a potent GLT-1 activator, inhibited cocaine self-administration.² Our laboratory is performing human efficacy studies of CLAV, a component of the FDA-approved antibiotic Augmentin®, in CUD patients. Reactivity to drug cues using functional magnetic resonance imaging (fMRI) is a powerful method to understand brain function and treatment response in CUD subjects.³ We hypothesized that CLAV would decrease responses to cocaine cues during a fMRI craving-inhibition task in addiction-associated brain regions.

This Phase Ib, double-masked, placebo-controlled clinical trial was conducted in accordance with ethical standards after Temple IRB approval. This inpatient study was done at Temple University Hospital Health System (TUHS), from 2020–21, and registered in ClinicalTrials.gov NCT04411914. Study drug was manufactured at the Investigational Drug Service of the University of Pennsylvania with each capsule containing 250 mg CLAV or matched placebo (PBO).

Written informed consent was obtained from participants. Inclusion criteria: a DSM-V diagnosis of moderate to severe CUD and minimum of 7 days of abstinence based on Timeline Follow-Back assessment confirmed by urine drug screen. Exclusion criteria: a contraindication to MRI scan, significant medical or psychiatric illness, history of head injury, seizure or other brain disorder and allergy to penicillin or other beta-lactam drugs. Individuals taking medications that affect glutamate, glutamine, or GABA levels, and pregnant/nursing women were excluded.

Double-masked randomization to CLAV or PBO (3:1 ratio) was used with stratification based on sex and days since last cocaine use (<= or >30 days). Participants received either oral CLAV 500 mg/day or matched PBO each morning for 3 days. Structural MRI was followed by the fMRI craving-inhibition task.

During fMRI, a 6-second (s) craving-inhibition video cue task was used (a variation of published paradigms from the Childress laboratory).⁴ The task used quasi-random, 6s videos of cocaine or neutral scenes. Each cue was presented with instructions (2s) to either “Watch like a movie”, or to “Reduce your response.” Three stimulus conditions, NEUTRAL (watching neutral video), WATCH (watching cocaine video), and DOWN (reduce response to cocaine video) were recorded. A sequence of 18 cues lasted 5 minutes. As past research comparing the first task half to the full task showed influence by “carry-over” from the first to the second half^5,6 this analysis focuses on the first half (9 cues) yielding NEUTRAL1, WATCH1 and DOWN1 (Figure 1).

Figure 1.

Cocaine Cue Paradigm During fMRI

(A) Cues were presented in a quasi-random fashion as depicted here (s, seconds). Subjects were given instruction: “Watch” (like a movie) or “Down” (Inhibit their craving). Neutral cues preceded the cocaine cues. (B) A series of images were presented (left to right) to the participant during fMRI each for the duration noted. The interstimulus interval (4 s) was a grey screen with a small cross mark. Total duration of the task was 340 s including first and second halves. Subjects responded to the rating scale using either a toggle switch or verbally (‘1’ (no craving) to ‘9’ (extreme craving)).

A Siemens 3T scanner was used to acquire T2* weighted BOLD single shot gradient echo planar (GRE) images with the following parameters: time repetition (TR) = 2000 milliseconds (ms), time echo (TE) = 29 ms, flip angle = 70°, Field of view (FOV) = 208 mm, matrix size = 75 slices, and a voxel size of 2 × 2 × 2 mm with a slice thickness of 2 mm. Structural images were 3D SPGR (spoiled gradient recalled acquisition) acquired with the following parameters: TR = 2300 ms, TE = 2.32 ms, flip angle = 8°, FOV = 240 mm, 60 slices, resulting in voxel dimensions of 1 × 1 × 1 mm thick.

Preprocessing and first-level analyses of data was done using FEAT (FMRI Expert Analysis Tool).⁷ Preprocessing included motion correction for head movement, B0 unwarping for magnetic field inhomogeneities, slice timing correction for differences in slice acquisition times, and high-pass filtering to remove low-frequency noise. Spatial smoothing was performed using a 5 mm Full Width at Half Maximum (FWHM) Gaussian kernel to improve the signal-to-noise ratio. ICA-AROMA (Independent Component Analysis-Automatic Removal Of Motion Artifacts) removed motion-related artifacts. Preprocessed images were registered to the Montreal Neurological Institute (MNI) standard brain using linear and nonlinear transformations.

After preprocessing, first-level contrasts for “WATCH1-NEUTRAL1” and “DOWN1-NEUTRAL1” were used to generate averaged statistical Z-score maps for the two groups (CLAV, PBO). The threshold of statistical images (z-maps) was 2 < T < 5 for MNI template display. Regions of interest (ROIs) were defined using Harvard Oxford atlas (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/Atlases) (Figure 2).

Figure 2.

Regions of Interest Defined in CUD Participants

Nucleus Accumbens (NAcc), Amygdala, Anterior Cingulate Gyrus (CNG), Posterior Cingulate Gyrus (CNG), Caudate, Lateral Orbital Frontal Cortex (LOFC); Ventromedial Prefrontal Cortex (VMPFC), Insula, Putamen and Middle Frontal Gyrus-Dorsolateral Prefrontal Cortex (MFG DLPFC).

Data were expressed as counts/percentages for categorical variables and medians (ranges or interquartile ranges-IQR) for continuous variables. Wilcoxon exact test was used to calculate p-values for comparisons of the two treatments on brain activation intensity. Regular/Mixed-effects regression was used to model outcomes of interest without/with repeated measures over time, with interaction terms between treatment and a covariate of interest such as Arm*Time being included, one covariate at a time. Statistical tests were two-tailed. P-values less than 0.05 were considered statistically significant. SAS version 9.4 (SAS Institute Inc., Cary, NC) was used for the data analyses.

Thirteen participants were randomized to CLAV (N = 10) or PBO (N = 3). Twelve (8 men, 4 women) completed the study. One male developed hepatitis on CLAV and discontinued; his hepatitis resolved within 2 weeks. Participants’ median age was 55 years (range 28–61) with a median of 27 years (range 1.5–34.5) of cocaine use and a median of 17.5 days of abstinence (range 7–280). Participants identified as Black, Non-Hispanic (75%) or White, Non-Hispanic.

Compared to PBO participants, CLAV participants demonstrated significantly decreased activity during DOWN1-NEUTRAL1 in both the anterior cingulate gyrus (ACC) and the caudate (data expressed as Beta (β) coefficient (median activation (IQR)) (see Figure 3). The ACC showed CLAV β coefficient = 0.17 (0.15, 0.20), PBO β coefficient = 0.58 (0.57, 1.07); p = 0.009, and the caudate: CLAV β coefficient = 0.03 (0.01, 0.20), PBO β coefficient 0.53 (0.24, 0.57); p = 0.018, supporting CLAV associated tempering of brain responsivity to cocaine cues with conscious inhibition of craving response. There was also significantly reduced caudate activity with CLAV treatment when the participant simply watched the video cues (WATCH1-NEUTRAL1): CLAV β coefficient = 0.1 (0.0, 0.1), PBO β coefficient = 0.4 (0.3, 1.7); p = 0.036.

Figure 3.

CLAV Significantly Decreased Activation in ROIs During Voluntary Inhibition

CLAV participants voluntarily inhibited their response to cocaine cues during fMRI. Decreased activation with CLAV seen in several addiction associated ROI. (CLAV: clavulanic acid 500 mg/d for 3 days; PBO: similarly treated with matched PBO). **p < 0.05: Anterior cingulate gyrus p = 0.009 and Caudate p = 0.018; *p <= 0.10: Nucleus accumbens p = 0.094, Insula p = 0.064 and MFG DLPFC p = 0.096.

These areas are critical addiction associated regions. The ACC has both limbic and prefrontal cortical connections and plays a critical role in emotional regulation, executive control, reward processing and behavioral regulation. The caudate, part of the dorsal striatum, has connections to the orbitofrontal cortex thought to be necessary for drug motivation.⁸ Dampening ACC and dorsal striatum activity with CLAV may decrease motivation for cocaine addictive behaviors and strengthen the ability to consciously resist drug cues.

There was also a trend towards lessened reactivity to cocaine cues with CLAV seen with the DOWN1-NEUTRAL1 condition in other brain areas associated with addiction (Figure 3): the insula (CLAV β coefficient = 0.16 (0.03, 0.26), PBO β coefficient = 0.56 (0.52, 0.87); p = 0.064), nucleus accumbens (NAcc) (CLAV β coefficient = 0.15 (0.00, 0.33), PBO β coefficient = 1.33 (0.42, 1.60); p = 0.094), and middle frontal gyrus-dorsolateral prefrontal cortex (MFG-DLPFC) (CLAV β coefficient = 0.11 (0.10, 0.19), PBO β coefficient = 0.22 (0.18, 0.39); p = 0.10). These areas are critical for the interoception of pleasurable signals from drug use, the evaluation of drug reward and initiation of reward related behavior and craving.^9,10 While the p-values are only suggestive, the study has a small sample size and the findings deserve confirmation.

This is the first reported human fMRI study of the modulating effects of clavulanic acid, a GLT-1 activator, on brain response to drug cues in a CUD population. We postulated that modulating glutamate circuits with CLAV would decrease activity in brain regions critical to maintaining cocaine use and alter the processing of cocaine cues in individuals with CUD.

Our findings suggest that CLAV attenuates limbic system activity in response to cocaine cues in participants with CUD. One hypothesis is that CLAV interferes with the learned response and habituated connections between the reward centers and executive control networks conditioned in cocaine addiction.

Despite the limitation of small sample size of this pilot study, the current findings argue for a larger study to explore CLAV’s therapeutic potential for CUD, and possibly for other substance use disorders wherein cue-triggered glutamate release may lead to relapse.