D-cycloserine modulates breathlessness related brain activity over pulmonary rehabilitation

Rationale: For people with Chronic Obstructive Pulmonary Disease (COPD), improvements in breathlessness from pulmonary rehabilitation are neither long lasting nor guaranteed. Previously, we showed that pulmonary rehabilitation induced brain activity changes akin to those seen in exposure based cognitive behavioural therapies (CBT) in other conditions. D-cycloserine is a partial NMDA-receptor agonist which has been shown to enhance CBT. Objectives: Here, we tested whether D-cycloserine would augment the effects of pulmonary rehabilitation on activity in brain areas that process breathlessness expectation. Methods: 72 participants with mild-to-moderate COPD were recruited to a double-blind experimental medicine study running parallel to a pulmonary rehabilitation course. Participants were randomised to 250mg D-cycloserine or matched placebo, administered 15-30 minutes prior to the first four sessions of pulmonary rehabilitation. Brain functional magnetic resonance imaging, self-report questionnaires and clinical measures of respiratory function were collected at three time points: before, during (2-3 weeks) and after pulmonary rehabilitation (6-8 weeks). Measurements: Primary and secondary outcome measures were difference in mean and voxel-wise brain activity across key brain regions of interest. An exploratory analysis determined the interaction with breathlessness-anxiety. Main results: No difference was observed in either primary or secondary outcome measures. However, in the exploratory analysis, D-cycloserine attenuated the relationship between brain activity and breathlessness-anxiety within prefrontal cortex, superior frontal gyrus and precuneus. Conclusions: The observed effects suggest that D-cycloserine augments pulmonary rehabilitation by dampening reactivity to breathlessness cues in brain areas associated with breathlessness expectation and anxiety. This work highlights the opportunity to test brain-active drugs in the context of augmenting behavioural interventions.


Introduction
Chronic breathlessness is a central symptom of chronic obstructive pulmonary disease (COPD). Currently, pulmonary rehabilitation offers the most effective treatment strategy for chronic breathlessness in COPD, where over a 6-8 week programme, participants learn to manage their symptoms by engaging in supervised physical exercise and educational sessions. However, around 30% of patients derive no clinical benefit [1], and for those that do, health-related benefits plateau within the first 6months following pulmonary rehabilitation, returning to pre-rehabilitation levels for the majority of patients after 12-18 months [2,3]. Thus, strategies to enhance the therapeutic process capable of increasing or prolonging treatment effects are of considerable clinical relevance.
A body of evidence has shown that improvements in breathlessness over pulmonary rehabilitation result from a reappraisal of the sensory experience [4][5][6][7], which arrests the downward spiral of fear, avoidance and physical deconditioning [6,8]. Clinical studies are supported by brain imaging evidence demonstrating changes in brain activity after pulmonary rehabilitation within areas including cingulate cortex, angular gyrus, insula and supramarginal gyrus which are linked to changes in ratings of breathlessness intensity and breathlessness anxiety [7]. The changes to activity within these networks, which are associated with attention and learned sensory and emotional expectations are similar to those observed during exposure-based cognitive behavioural therapies (CBT) for anxiety and panic disorder [9,10].
In the field of psychiatry, drug interventions have been shown to enhance exposurebased CBT, successfully reducing anxiety and promoting cognitive reappraisal in some studies [11][12][13][14]. These findings suggest that targeting similar brain processes with drug interventions in chronic breathlessness may be an important focus for improving the effectiveness of pulmonary rehabilitation. One candidate drug is D-cycloserine, a partial N-Methyl-D-Aspartate (NMDA) receptor agonist, which neuroscientific evidence suggests is associated with a reduced threat response in amygdala when paired with exposure based therapy [11]. The mechanism of effect is thought to occur at the glycine modulatory site of the NMDA receptor, where high affinity binding enhances synaptic plasticity, promoting emotional learning processes [13,15] and boosting therapeutic effects as a result [12,[16][17][18].
Based on clinical and neuroscientific evidence in psychiatry and the parallels observed in pulmonary rehabilitation, we hypothesised that D-cycloserine would augment the neural responses to learned breathlessness associations connected with pulmonary rehabilitation. Neuroimaging techniques offer advantages in terms of detecting subtle changes in brain activity, which may precede clinical effects [19]; thus, an imaging study was could determine the value of proceeding to larger definitive clinical trials of brain-targeted agents to enhance pulmonary rehabilitation. Our primary hypotheses focused on brain activity changes within five key regions of interest, identified in previous studies of breathlessness [7,[20][21][22]. The regions of the anterior insular cortex, posterior insular cortex, anterior cingulate cortex, amygdala and hippocampus have all been linked to body and symptom perception as well as emotional salience [20,23]. Secondary hypotheses examined the effect of D-cycloserine across a wider region of interest containing fifteen pre-defined brain areas. The fifteen brain areas encompassed regions associated with sensory and affective processing of breathlessness as well as body and symptom perception.
Since the inception of this trial, there has been an increasing number of null results from trials using D-cycloserine in combination with exposure-based CBT [24,25].
While this trend may in part be explained by technical differences in study design, a number of well-powered and well-controlled studies have also revealed a more nuanced picture of D-cycloserine action. Several studies have shown that D-cycloserine speeds therapeutic improvements rather than increasing them overall [14,18], while Hofmann et al suggest that "D-cycloserine not only makes "good exposures" better but also may make "bad exposures" worse" [26]. To account for this updated literature, we conducted two additional analyses. Firstly, we examined the temporal action of D-cycloserine. Secondly, we tested whether brain activity changes relating to D-cycloserine were linked to improvements in breathlessness related anxiety during pulmonary rehabilitation.

Methods and Materials
An overview of the methodology is presented here. Full details, including sample size calculations, non-completion and sensitivity analysis can be found within supplementary materials. The study and statistical analysis plan were pre-registered on clinicaltrials.gov (ID: NCT01985750) prior to unblinding.
Participants 91 participants (30 female, median age 70 years; range 46-85 years) with COPD were recruited immediately prior to their enrolment in a National Health Service-prescribed course of pulmonary rehabilitation (full demographic information including noncontinuation is shown in Supplementary material Table 1). From this population, 72 participants completed all study visits (18 female, median age 71 years (46-85 years)) (Table 1 and Figure 2). Written informed consent was obtained from all participants prior to the start of the study. Study approval was granted by South Central Oxford REC B (Ref: 118784, Ethics number: 12/SC/0713). Study inclusion criteria were: a diagnosis of COPD and admittance to pulmonary rehabilitation. Exclusion criteria were: inadequate understanding of verbal and written English, significant cardiac, psychiatric (including depression under tertiary care) or metabolic disease (including insulin-controlled diabetes), stroke, contraindications to either D-cycloserine (including alcoholism) or magnetic resonance imaging (MRI), epilepsy, claustrophobia, regular therapy with opioid analgesics or home oxygen therapy.

Screened
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 28, 2021.

Study drug
Participants were randomised in a double-blinded procedure to receive either 250mg oral D-cycloserine or a matched placebo, administered by the study nurse 30 minutes prior to the onset of their first four pulmonary rehabilitation sessions. Study participants, investigators and those performing the analysis were blinded to the treatment allocation. Both D-cycloserine and placebo were over-encapsulated to appear identical. Full study drug description, randomisation protocol and minimisation criteria can be found within the supplementary materials.

Study visit protocol
Following telephone screening, participants were invited to attend their first research session (baseline) prior to starting pulmonary rehabilitation. A second study visit took place following the fourth pulmonary rehabilitation session but before the sixth session.
Participants completed the remainder of their pulmonary rehabilitation course before attending a third study session ( Figure 2) that occurred as close to the termination of pulmonary rehabilitation as possible and always within two weeks.

Physiological measures:
Spirometry and two Modified Shuttle Walk Tests (MSWT) were collected using standard protocols [37,38]. Participant height and weight were recorded at each visit.
Oxygen saturations and heart rate were measured with pulse oximetry and were collected at rest and following the MSWT.

MRI measures
Image acquisition: Magnetic resonance imaging of the brain was carried out using a Siemens 3T MAGNETOM Trio. A T1-weighted (MPRAGE) structural scan (voxel size: 1 x 1 x 1 mm) was collected and used for registration purposes. A T2*-weighted, gradient echo planar image (EPI) scan sequence (voxel size: 3 x 3 x 3 mm), TR, 3000ms; TE 30ms was used to collect FMRI data.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 28, 2021. ; https://doi.org/10.1101/2021.06.24.21259306 doi: medRxiv preprint Word cue task: To probe the neural responses of breathlessness-related expectations we examined the activity of brain regions responding to breathlessness-related word cues [7,27,39]. Brain activity was correlated with corresponding visual analogue ratings of breathlessness and breathlessness-anxiety collected during scanning.
During the fMRI scanning, participants were presented with a word cue, e.g., "climbing stairs" in white text on a black background for 7 seconds. Participants were then asked, "how breathless would this make you feel" (wB) and "how anxious would this make you feel" (wA). To each question participants responded within a 7 second window using a button box and visual analogue scale (VAS). The response marker always initially appeared at the centre of the scale, with the anchors "Not at all" and "Very much" at either end.
Control task: A validated task of emotional faces was used as a control to separate generalized anxiety from breathlessness specific anxiety. Fearful or happy faces were presented on a black background was used to examine whether any differences in brain activity patterns between D-cycloserine and placebo groups was specific to breathlessness processing. Each face was shown for 500ms in blocks of 30 seconds.
A fixation cross was interspersed for 30 seconds between the blocks of faces.
Participants were instructed to respond via a button box to indicate facial gender.
Reaction time and accuracy were recorded throughout the task.

Outcomes
The primary outcome comparison was brain activity difference between D-cycloserine and placebo in five pre-specified brain regions. Secondary outcomes were a voxelwise difference in brain activity between D-cycloserine and placebo groups across one large region of interest which consisted of fifteen smaller brain regions. Exploratory . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) analyses examined the temporal action of D-cycloserine and updated the models used for primary and secondary outcomes to incorporate psychological variables.

Analysis
A summary of analyses is outlined here. Full details, including procedures for dealing with missing data and sensitivity analysis can be found within the supplementary materials. Our analysis for primary and secondary outcomes was pre-registered and made publicly available prior to unblinding (https://tinyurl.com/yxzkyp3d).
Data were pre-processed according to standard protocols before being entered into single subject general linear models. These models captured brain activity during the periods in which the breathlessness-related word cues were presented allowing us to examine expectation-related processes.

Group level analysis
For each patient, the following metrics were extracted from each of the five regions of interest: anterior insula cortex, posterior insula cortex, anterior cingulate cortex, amygdala and hippocampus, at visits one, two and three: 1. Mean brain activity in response to breathlessness word-cue presentation 2. Mean brain activity for control task of emotional faces . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 28, 2021. To test for a drug effect across each metric, the values from visit two were entered into independent linear mixed effects models where they were adjusted for age, gender and scores at visit one. To correct for multiple comparisons across regions, permutation testing (with Family Wise Error Rate (FWE) 5%) was carried out. This process was repeated separately for data collected at visit three. Models were programmed using the lme4 function and permuco package within R version 3.6.1 (2019-07-05).
To test for a drug effect across the larger region of interest (panel B of Supplementary   Figure 1), the following voxel-wise information was collected from within the region of interest at visits one, two and three: 1. Voxel-wise brain activity in response to breathlessness word-cues presentation 2. Voxel-wise brain activity in response the control task of emotional faces Each of the values from visit two were entered into independent general linear model (GLM), controlling for age, gender and scores at visit one. Permutation testing was performed with threshold free cluster enhancement (TFCE) (a non-parametric test) [40] using FSL's Randomise tool [41] at family wise error corrected p<0.05. The process was then repeated separately for data collected at visit three.
We repeated the primary and secondary analysis but with an additional term included in each model for the change in breathlessness related anxiety (wA). This allowed us to ask the question "Was there a difference in the relationship of brain activity and changes in breathlessness anxiety between the D-cycloserine and placebo groups?" To answer the question "do changes to brain activity patterns occur more quickly in the D-cycloserine group" The difference between time points of each metric listed . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 28, 2021. ; https://doi.org/10.1101/2021.06.24.21259306 doi: medRxiv preprint above under secondary analysis was estimated using a form of linear regression known as the Robust Sandwich Estimator technique [42]. This implements a 2-way repeated measures ANOVA, 1 intrasubject factor of time (2 time points), 1 between subject factor (2 groups), and their interaction. Further details of this technique can be found in the supplementary materials.

Results
Of the 91 participants recruited (Figure 1), 72 participants completed all three study visits. Reasons for drop-out or exclusion included illness, scanner error and issues with data quality. One further participant was excluded due to an error in task-data collection. 71 participants were therefore assessed for study objectives. Sensitivity analysis was performed and is reported within supplementary materials.
There was no significant overall effect of D-cycloserine on mean brain activity within the five key regions of interest of anterior cingulate, anterior insula cortex, amygdala, hippocampus or posterior insula cortex (family wise error rate corrected, p>0.05) at visit two or visit three (Table 2). is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 28, 2021. ; https://doi.org/10.1101/2021.06.24.21259306 doi: medRxiv preprint There was no significant overall effect of D-cycloserine across the broader mask of fifteen regions measured voxel-wise (family wise error rate correct, p>0.05) at visit two or visit three. No significant differences in the questionnaire measures or physiology scores were found between the D-cycloserine and placebo groups at any point during the study.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 28, 2021.   . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 28, 2021.    . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 28, 2021. ; https://doi.org/10.1101/2021.06.24.21259306 doi: medRxiv preprint compared to the placebo group, for a given improvement in breathlessness anxiety there was an attenuated neural response to breathlessness cues (z=2.3, p<0.05) ( Figure 4). This difference in brain activity was observed within the dorsolateral and medial prefrontal cortices, superior frontal gyrus, and precuneus. No significant relationship was observed between breathlessness related anxiety, drug allocation and brain activity at visit two. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 28, 2021. ;

Key Findings
The results from the exploratory analysis give important insights into potential mechanisms of action and novel therapeutic avenues. In the D-cycloserine group, for a given change in ratings of breathlessness anxiety, there was correspondingly less brain activity in response to breathlessness cues than the placebo group. This effect was observed across a network of emotional salience regions which included superior frontal gyrus, precuneus, dorsolateral prefrontal cortex and medial prefrontal cortex. Less corresponding increase in brain activity for D-cycloserine . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 28, 2021. ; https://doi.org/10.1101/2021.06.24.21259306 doi: medRxiv preprint improve exercise capacity over a course of pulmonary rehabilitation [43], while the long-lasting bronchodilator Tiotropium [44] also improved scores on the St Georges Respiratory Questionnaire which were sustained for three months following rehabilitation. While these studies focused on improving exercise capacity, the strong cognitive component to breathlessness suggests that the brain may be an important target that so far has been relatively ignored.
Work in social anxiety, post-traumatic stress disorder and panic disorder has shown that exposure-based CBT, which reduces anxiety and promotes reappraisal, may be enhanced by D-cycloserine [11][12][13][14]. Mechanisms of drug effect may occur via mediated activity within hippocampus, amygdala, dorsolateral prefrontal cortex and insula, regions which overlap with brain networks of internal bodily sensation (interoception) and reappraisal [11,17,18]. Our previous work, which examined changes in brain activity over a course of pulmonary rehabilitation, identified increased brain activity within posterior cingulate cortex, supramarginal gyrus and angular gyrus correlating with improvements to breathlessness anxiety [7]. It was hypothesised that these changes represented a shift in expectation-related brain activity towards that of healthy controls via somatosensory integration and attentional regulation. Based on these findings, we examined whether D-cycloserine would similarly augment the effect of pulmonary rehabilitation on neural responses to learned breathlessness associations.
Our primary and secondary outcomes (group mean differences in brain activity) were performed in predefined brain regions of interest. We tested for an overall group difference and observed no difference between D-cycloserine and placebo upon brain activity over the course of pulmonary rehabilitation.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 28, 2021. ; https://doi.org/10.1101/2021.06.24.21259306 doi: medRxiv preprint Our exploratory analyses addressed more recent literature which has found Dcycloserine acts to reinforce both positive and negative experiences during exposurebased CBT [26]. This may render brain activity differences unobservable in a simple contrast of group means. Following the completion of pulmonary rehabilitation, for a given improvement in breathlessness anxiety, D-cycloserine suppressed brain activity in response to breathlessness-related word cues within superior frontal gyrus, precuneus, dorsolateral prefrontal cortex and medial prefrontal cortex compared to placebo. This interaction can be considered as a difference in slopes of the relationship between breathlessness anxiety and brain activity.
The networks targeted by D-cycloserine are associated with attentional regulation. In a previous study these networks were also shown to change over pulmonary rehabilitation [7]. However, while our previous work found co-activation within angular and supramarginal gyrus, regions associated with somatosensory integration, the current study's co-activated networks are associated with emotional responsiveness. This finding demonstrated parallels between an earlier comparison of brain activity between patients with COPD and healthy controls, where breathlessness-related word cues elicited greater activity within medial prefrontal cortex than patients with COPD [27]. This difference was thought to reflect differences in emotional-cognitive aspects of breathlessness processing. Therefore, D-cycloserine, which we have shown here modulates the medial prefrontal cortex, may be driving activity towards that seen in healthy controls for whom the breathlessness-words hold less expectation-related significance. We speculate that our findings represent a down regulation of emotional responses to the breathlessness word cues as a result of D-cycloserine. This was particularly the case for people who had derived positive change (measured by breathlessness anxiety ratings) from pulmonary rehabilitation.
Our work has highlighted a potential brain-derived pathway of effect, for which other drugs may become more attractive candidates. As a stand-alone treatment for . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 28, 2021. ; https://doi.org/10.1101/2021.06.24.21259306 doi: medRxiv preprint treatment resistant depression, ketamine, which blocks pre-synaptic NMDA receptor signalling, increasing glutamate and thereby synaptic plasticity, has been linked with reductions in fear and anxiety [45], and rapid relief from symptoms [46]. While glucocorticoids such as cortisol, combined with exposure-based CBT, have shown promise in reduction of fear in phobias and post-traumatic stress disorder [12].
Collectively these candidate drugs boost synaptic plasticity, although via different neurochemical pathways, which may facilitate the re-setting of fearful associations within the brain. These could be used either during pulmonary rehabilitation, or as part of a precursor programme, helping to recruit harder to reach patients and support selfmanagement. Successful self-management has been highlighted as a key objective by the department of health [47] and often follows on from CBT-based programmes such as the talking therapies recommended by the British Lung Foundation

Future considerations and limitations
Our findings, which show that the action of D-cycloserine seems to relate to reductions in breathlessness anxiety, lend support to personalised approaches to treating breathlessness. We therefore suggest that future research should collect postrehabilitation session outcome measures. Paired with post treatment outcome assessment, self-report measures of session success could affect the decision to administer drugs such as D-cycloserine. Questions remain regarding D-cycloserine's optimum dosage, dose timing and number of administered sessions [24,48]. We selected a dose of 250mg for this study based on its previous successful usage [11,13] and practical considerations regarding drug availability. However, future trials may consider further investigating D-cycloserine (or alternative drug) dosage and dose timing in relation to this population of older adults who may demonstrate different responsivity to D-cycloserine given the well-established changes to NMDA receptor function as the brain ages [49]. Additionally, the action of D-cycloserine is known to be . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 28, 2021. ; https://doi.org/10.1101/2021.06.24.21259306 doi: medRxiv preprint curtailed near the therapeutic ceiling [50], and pulmonary rehabilitation is a highly effective treatment [2], this may leave insufficient scope for improvement in some individuals. Future work may consider pairing D-cycloserine (or alternative) with a weaker behavioural intervention. Finally, given D-cycloserines' supposed effect on learning, longer term follow-up would provide a clearer picture as to whether clinical scores were sustained for longer or whether relapse rates are affected.

Conclusions
We have shown evidence that D-cycloserine has the potential to augment the effects of pulmonary rehabilitation on the brain's breathlessness perception networks. This  . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 28, 2021. ;

33.
Bestall, J.C., et al. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 28, 2021. ; . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 28, 2021. 250mg of D-cycloserine. While 50mg remains the most common dosage for Dcycloserine, there is no evidence to suggest that 50mg is more effective than higher dosages [1]. Furthermore, work conducted in healthy volunteers suggests that hippocampal learning only occurred at 250mg dose and not at 50mg [2]. The efficacy of 250mg dosage on brain plasticity is supported by more recent work [3] which found changes to amygdala reactivity after a single administration of D-cycloserine.

Randomisation Procedure
Once the participant gave written consent to the trial and completed the MRI scan, a member of the team submitted a randomisation form, entering eligibility criteria and . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 28, 2021. which at the first stage of unblinding an independent researcher provided study researchers with a coded binarised system for analysis. Researchers remained blinded to group identity until analysis was completed.

Pulmonary Rehabilitation
Pulmonary rehabilitation courses were run by either Oxford Health NHS Foundation Trust, West Berkshire NHS Foundation Trust, or Milton Keynes University Hospitals

Sample Size
At the time of study inception (and to a large extent still to date), the literature regarding D-cycloserine's effects on functional brain activity is very limited. Therefore, in order to calculate the sample sizes required for this study we first took into account the described effects of d-cycloserine in clinical studies of augmentation for cognitive behavioural therapy for anxiety disorders, where effect sizes of up to 1.06 have been reported (although more commonly 0.4 to 0.7) [4][5][6][7]. The most relevant paper (on . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 28, 2021. ; https://doi.org/10.1101/2021.06.24.21259306 doi: medRxiv preprint treatment of snake phobia [8]) demonstrated that effects observed with neuroimaging were more sensitive than behavioural effects, therefore we believe that powering for a behavioural outcome measure (breathlessness-anxiety) provided a safe margin and was likely to be sufficiently conservative to detect our measures of interest. This was particularly the case as compared to the relatively blunt nature of behavioural data collection, functional neuroimaging carries considerably more specificity and statistical power. The study was not therefore specifically powered to investigate the clinical effects of D-cycloserine. In our previous study we observed an 11% (SD15%) improvement in breathlessness-related anxiety, measured with our FMRI word task (pre-treatment mean score 38%, post treatment mean score 27%, difference 11%, SD of difference 15%) [9]. Making a conservative assumption, we estimated that Dcycloserine augments this response with an effect size of 0.4. Assuming a similar coefficient of variation we anticipated an 18% (SD24%) improvement in breathlessness-anxiety (i.e. pre-treatment mean score 38%, post treatment mean score 20%, difference 18%, SD of difference 24%). Assuming α=0.05 and power 0.80, then we estimated a sample size of 36 in each group randomised 1:1. As this is a behavioural outcome, we expected this to have sufficient power to detect change in BOLD signalling.

Missing Data
The potential effect of missing brain imaging data was explored using a sensitivity analysis. Missing questionnaire and physiology data points were imputed using the Markov chain Monte Carlo method (multiple imputation technique) within the MICE package in R. A summary of missing data is provided below.

Behavioural Measures
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Questionnaire Measures
Dyspnoea-12 (D12) Questionnaire: This is a 12-item questionnaire designed to measure the severity of breathlessness and has been validated for use in patients with respiratory disease [10].

Centre for Epidemiologic Studies Depression Scale (CES-D):
Depressive symptoms are commonly observed in patients with respiratory disease. This brief questionnaire consists of 20 items investigates the symptoms of depression across a number of factors [11].

State-Trait Anxiety Inventory (STAIT-T):
This questionnaire assesses participant's general level of anxiety in particular scenarios via 20 questions asking "how anxious you generally feel" [12].
Fatigue Severity Scale: This 9-point questionnaire quantifies patient fatigue, which is well documented in its association with COPD [13].

St George's Respiratory Questionnaire (SGRQ):
There are 50 questions in this questionnaire, which has been developed and validated for use in COPD and asthma.
The questions measure the impact of overall health, daily life and well-being [14].

Medical Research Council (MRC) breathlessness scale:
The MRC scale quantifies perceived difficulty due to respiratory restrictions on a scale of 1 to 5 [15].

Mobility Inventory (MI):
This questionnaire collects data regarding the extent to which a participant avoids certain situations, either alone or accompanied (21-items in each category) [16].
Breathlessness Catastrophising Scaleadapted from the catastrophic thinking scale in asthma: This 13-point questionnaire was modified for this study by substituting the word "asthma" for "breathlessness" in order to measure catastrophic thinking [17] [18].
Breathlessness Vigilance Scaleadapted from the pain awareness and vigilance scale: This questionnaire was modified by substituting the word "breathlessness" for . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Physiological Measures
A trained respiratory nurse collected spirometry measures of FEV1 and FVC using Association for Respiratory Technology and Physiology standards [20]. Participants performed two modified incremental shuttle walk tests (MSWT) [21], and heart rate and oxygen saturations (SpO2) were measured immediately before the MSWT and subsequently every minute until 10 minutes post-exercise (or until participants returned to their baseline state) using a fingertip pulse oximeter (Go2; Nonin Medical Inc). Before and after the MWST participants also rated their breathlessness on a modified Borg scale [22]. In a MWST participants must walk between and around two cones, placed 10m apart in time to a set of auditory beeps played from a laptop. Initially the speed of beep repetition is slow, but the participant must increase their walking speed each minute in order to reach the cone before the next beep. Participants continue to walk (or run) until they are too breathless to continue, at which point the total distance walked is recorded.

MRI Acquisition
Prior to each MRI session participants were screened for standard MRI contraindications including metal in or about their person, epilepsy and claustrophobia.

Image acquisition:
Hardware: A Tim System (Siemens Healthcare GmbH) 12-channel head coil. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 28, 2021. Field map scans of the B0 field were obtained to aid the distortion correction of the functional scans: TR, 488ms; TE1, 5.19ms; TE2, 7.65ms; flip angle 60°; voxel size, 3.5 x 3.5 x 3.5 mm.

Word Task
This task was developed and published by Herigstad and colleagues in 2016 for use in the COPD population [23]. Word cues were developed in three key stages; firstly in collaboration with respiratory practitioners, academics and physiotherapists, a set of 30 word cues associated with breathlessness were created. Next, these cues were provided to patients with COPD alongside a VAS rating scale, allowing patients to rate how breathless and anxious the situations identified by the cues would make them feel. Following adjustments based on participant feedback, the word cues were then computerised and tested in a larger population of COPD patients [23]. Further validation was carried out in the fMRI environment and by for clinical sensitivity with comparisons between changes in key questionnaire measures and word-cue rating.
Before the first scan session, participants were given the opportunity to practice using the button box with a set of test words.
Symptom perception has been shown to be modulated by expectations [24][25][26]. Even in healthy volunteers, a cue paired with the expectation of a breathlessness-inducing situation can produce breathlessness and drive brain activity patterns in absence of afferent input [24]. When expectation incorrectly matches afferent inputs, maladaptive breathlessness can occur, and in populations with chronic symptoms of breathlessness the relationship between expectation and breathlessness-cues can become deeply reinforced. As a result, breathlessness-expectation is one target for . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 28, 2021. ; the CBT elements of pulmonary rehabilitations, and the most likely point of action for D-cycloserine.
Control tasks 1) A control condition, used as a baseline measure of activity in response to the presentation of a visual stimulus was presented 4 times over the course of the wordcue scan, consisting of a string of "XXXXXXXXXXXXXXX" with fixed length of 15 characters, and each time was presented for 7 seconds. No rating period followed these control blocks [9,18,23].
2) A validated task of emotional faces was used as a control to separate generalized anxiety from breathlessness specific anxiety. Emotional facial expressions are widely recognised to activate the same brain pathways as the behavioural emotion conveyed by the expression itself. Fearful facial expressions, for example, have been shown to correspond to activity within the amygdala, a region known to modulate fear processing [26]. Faces were drawn from a set first developed by Ekman and Friesen [27] and furthered by Young et al [28]. Photographs of 10 faces (5 male, 5 female) with fearful or happy expressions of 100% intensity were used. Each face was shown for 500ms in blocks of 30 seconds. A fixation cross was interspersed for 30 seconds between the blocks of faces. Participants were instructed to respond via a button box to indicate facial gender. Reaction time and accuracy were recorded throughout the task. The task contrasting fear and happy facial expressions has been extensively used in previously studies and has been found to activate the amygdala in both healthy volunteers and in depressed patients [29,30]. Neutral faces are not typically used as they can be interpreted as threatening or ambiguous in different settings [31].

Functional MRI Preprocessing
Data denoising was carried out as follows: Before the first level analysis, each . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 28, 2021. ; https://doi.org/10.1101/2021.06.24.21259306 doi: medRxiv preprint functional scan was decomposed into maximally independent components using FMRIB's MELODIC tool (Multivariate Exploratory Linear Optimised Decomposition into Independent Components). "Noise" components were identified by FIX (FMRIB's autoclassification tool, [32,33]) using the WhII.Standard.RData [34] trained classifier with aggressive clean up option. A Principal Component Analysis (PCA) was run on the FIX identified components to retrain 99% of the variance. Separately, the cardiac and respiratory related physiological signals (recorded via a pulse oximeter and respiratory bellows) were transformed into a series of regressors, (three cardiac and four respiratory harmonics) as well as an interaction term and a measure of respiratory volume per unit of time (RVT), using FSL's physiological noise modelling tool (PNM).
The signal associated with these waveforms (modelled using retrospective image correction (RETROICOR) [35,36]) was then used to form voxelwise noise regressors.
The confounds identified by FSL's FIX and PNM tools, along with sources of noise arising from motion, were then combined into a single model. This single noise model approach builds upon the technique outlined by [37]; and fully detailed by [38]. In these preceding works we employed a step-wise technique whereby physiological noise (identified by PNM) and FIX-identified noise were each removed from the data in separate steps prior to data entry into the lower level model. In the new cleanup pipeline, a single text file containing time-course information relating to FIX identified noise components along with white matter or CSF related noise was included as additional confound EV's within the lower level model, while the PNM-identified noise was entered into the model as a standard voxel-wise confound list. In this updated denoising pipeline, confounds identified above are added to model at the stage of firstlevel analysis and thus the functional dataset can be corrected for sources of noise arising from motion, scanner and cerebro-spinal fluid artefacts, cardiac, and respiratory noise in a single step, rather than three.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Functional MRI Analysis
MRI processing was performed using FEAT (FMRI Expert Analysis Tool within the FSL package). The data were corrected for movement using MCFLIRT (Motion correction using FMRIB's Linear Image Registration Tool [39]). Non-brain structures were removed using BET (Brain Extraction Tool [40]). Spatial smoothing was carried out using a full-width-half-maximum Gaussian kernel of 5mm, while high-pass temporal filtering (Gaussian-weighted least squares straight line fitting; 90 s) removed low frequency noise and slow-drift. Distortion correct of EPI data was carried out using a combination of FUGUE (FMRIB's Utility for Geometrically Unwarping EPI's [41,42] and BBR (Boundary Based Registration; part of the FMR Expert Analysis Tool, FEAT version 6.0 [43]). The data were corrected for physiological noise using FSL's FIX-PNM pipeline. Functional scans were registered in a two-step process to the MNI152 (1x1x1 mm) standard space brain template. Firstly, each subject's EPI was registered to their associated T1-weighted structural image using BBR (6 DOF) with nonlinear field map distortion correction [43]. In the second step the subject's structural image was registered to 1mm standard space via an affine transformation followed by nonlinear registration (using FNIRT: FMRIB's Non-linear Registration Tool [44]).

Region of interest extraction
The five bilateral regions of interest (ROI) were anterior insula cortex, posterior insula cortex, anterior cingulate cortex, amygdala and hippocampus. Seed voxels for each region of interest were identified as the peak voxel co-ordinates (Supplementary table   2) responding to breathlessness word-cues published by Herigstad et al 2017 [9] within the boundaries of each region of interest identified by standard atlas maps. The seed voxels were expanded to include the surrounding voxels within a 5 mm radius.
Left and right masks of bilateral regions of interests (anterior insula, posterior insula, anterior cingulate, amygdala and hippocampus) were added together to form one . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 28, 2021. ; https://doi.org/10.1101/2021.06.24.21259306 doi: medRxiv preprint mask for each region of interest. Following registration each mask was re-thresholded at 40% probability to avoid interpolation errors before being binarised.

Network mask
A second network mask region of interest was created from the 5 core regions of interest outlined above and an additional 11 regions defined by standard anatomical atlas maps (Harvard-Oxford Atlas and Destrieux' cortical atlas) (Supplementary Figure   1). A 40% probability threshold was applied to each region, before they were combined along with the original 5 regions into one network mask. This network mask was then registered to each individual before being re-thresholded at 40% probability to avoid interpolation errors and binarized. Combining the 16 regions into a single mask enabled us to appropriately correct for multiple comparisons.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 28, 2021. ;  CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Control task
At the individual subject level, a GLM was created with explanatory variables for the 30 second stimulus presentation periods of happy and fearful faces, along with the associated (de-meaned) reaction times. Two additional explanatory variables were created to model participant (de-meaned) accuracy in identifying whether the presented faces were male or female.  . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Sensitivity analysis essentially asks the question "if all of the participants who were enrolled but did not complete the study actually had, would the result of the primary analysis have changed?" To answer this, a number of different brain activity levels are simulated to account for reasonable extreme scenarios.
Using Equation 1, three quantiles of brain activity within each of the five key regions of interest in response to breathlessness-related word cues were calculated for = CC + 1 2 Y P 1 Y P -2 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 28, 2021.    . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 28, 2021. ; https://doi.org/10.1101/2021.06.24.21259306 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 28, 2021. ; https://doi.org/10.1101/2021.06.24.21259306 doi: medRxiv preprint

Sensitivity Analysis and Study completeness
Ten participants who were randomised to the D-cycloserine group and eleven participants who were randomised to the placebo group did not complete all three visits. Sensitivity analysis revealed that the inclusion of the 21 missing participants would not have altered the primary outcome.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 28, 2021. ; https://doi.org/10.1101/2021.06.24.21259306 doi: medRxiv preprint  is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 28, 2021. ; https://doi.org/10.1101/2021.06.24.21259306 doi: medRxiv preprint