Migraine monoclonal antibodies against CGRP change brain activity depending on ligand or receptor target – an fMRI study

Article
Figures and data
Abstract
Editor's evaluation
Introduction
Materials and methods
Results
Discussion
Data availability
References
Decision letter
Author response
Article and author information
Metrics

Abstract

Background:

Monoclonal antibodies (mAbs) against calcitonin gene-related peptides (CGRP) are novel treatments for migraine prevention. Based on a previous functional imaging study which investigated the CGRP receptor mAb (erenumab), we hypothesized that (i) the CGRP ligand mAb galcanezumab would alter central trigeminal pain processing; (ii) responders to galcanezumab treatment would show specific hypothalamic modulation in contrast to non-responders; and (iii) the ligand and the receptor antibody differ in brain responses.

Methods:

Using an established trigeminal nociceptive functional magnetic imaging paradigm, 26 migraine patients were subsequently scanned twice: before and 2–3 weeks after administration of galcanezumab.

Results:

We found that galcanezumab decreases hypothalamic activation in all patients and that the reduction was stronger in responders than in non-responders. Contrasting erenumab and galcanezumab showed that both antibodies activate a distinct network. We also found that pre-treatment activity of the spinal trigeminal nucleus (STN) and coupling between the STN and the hypothalamus covariates with the response to galcanezumab.

Conclusions:

These data suggest that despite relative impermeability of the blood-brain barrier for CGRP mAb, mAb treatment induces certain and highly specific brain effects which may be part of the mechanism of their efficacy in migraine treatment.

Funding:

This work was supported by the German Ministry of Education and Research (BMBF) of ERA-Net Neuron under the project code BIOMIGA (01EW2002 to AM) and by the German Research Foundation (SFB936-178316478-A5 to AM). The funding sources did not influence study conduction in any way.

Clinical trial number:

The basic science study was preregistered in the Open Science Framework (https://osf.io/m2rc6).

Editor's evaluation

Antibodies to CGRP or to the CGRP receptor are now approved for the management of migraine. Interestingly, although antibodies are presumed not to penetrate the blood-brain barrier, consistent with findings previously shown for antibodies directed against the receptor, the authors now report that galcanezumab, which targets the CGRP peptide, decreases hypothalamic activation, more so in responders. They suggest that these CNS actions may underlie part of the efficacy of these drugs in the management of migraine.

https://doi.org/10.7554/eLife.77146.sa0

Introduction

Since the introduction of the new generation of ‘migraine-specific drugs’, the monoclonal antibodies (mAbs) against the calcitonin gene-related peptide (CGRP-L-mAb) and its receptor (CGRP-R-mAb), their exact mode of action has been intensely studied. Whereas a central site of action is, due to limited permeability of the blood-brain barrier for large molecules, considered to be of minor relevance (Johnson et al., 2019), it is predominantly thought that CGRP mAb act peripherally by modulating the interaction between C-type and Aδ-type sensory neurons in the trigeminal ganglion and trigeminal nerve fibers (Edvinsson et al., 2018; Edvinsson et al., 2019), and that CGRP mAb may block CGRP responses in the dura mater (Edvinsson et al., 2018; Melo-Carrillo et al., 2017; Edvinsson, 2017).

In our recent study in migraine patients, we demonstrated that administration of the CGRP mAb erenumab results in a diminished activation of the thalamus, the insular cortex, the periaqueductal gray, and the secondary somatosensory cortex. Only responders showed less activation in the hypothalamus whereas non-responders did not, which suggests that CGRP mAb have direct or indirect central effects (Ziegeler et al., 2020).

The mode of action may however differ between different CGRP mAbs (Ziegeler and May, 2020). We focused on this question and studied 26 migraine patients before and after administration of the CGRP mAb galcanezumab using the same event-related functional magnetic resonance imaging (fMRI) and trigeminal nociceptive paradigm that we have used in the erenumab study (Stankewitz et al., 2010; Schulte et al., 2016). Given the results of the erenumab study, we hypothesized that (i) galcanezumab would alter central trigeminal pain processing in migraine patients, (ii) responders to galcanezumab treatment would show specific hypothalamic modulation in contrast to non-responders. We preregistered the study and the hypothesis (https://osf.io/m2rc6).

Materials and methods

Preregistration

Request a detailed protocol

This study was preregistered on May 12, 2020 (title ‘Galcanezumab in the migraine brain – fMRI’) on the Open Science Framework (https://osf.io/m2rc6).

Patient consent

Request a detailed protocol

The study was approved by the local ethics committee in Hamburg, Germany (PV 5964) and was conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained before initiation of the first study session. Participants could discontinue the study at any time.

Patients

Participants were recruited by headache specialists of the headache and facial outpatient clinic of the University Medical Center Hamburg-Eppendorf, Germany.

All study participants fulfilled the diagnosis of migraine (chronic and episodic) according to ICHD-3 criteria (No authors listed, 2018) and kept a headache diary. Drug-naive participants to any CGRP-antibody treatment were eligible when a therapy with galcanezumab 240 mg (loading dose) was planned following national treatment guidelines (Diener et al., 2020). In case of additional preventative treatments, the dose of this treatment must have been unchanged for at least 3 months prior to study participation and was not allowed to change during the study. Comorbid primary or secondary headache disorders (including medication overuse headache) were excluded and none of the participants suffered from severe comorbid psychiatric, neurological, or other somatic conditions. Two patients reported a mild restless legs syndrome.

Experimental paradigm

Request a detailed protocol

Patients were asked to attend two fMRI scanning sessions before and after the first administration of galcanezumab. Both visits took place approximately 3 weeks apart (pharmacokinetic drug peak blood level) and followed the same protocol. Following the first scan, the loading dose of galcanezumab 240 mg was administered subcutaneously by the patient under the prior instruction of a headache specialist.

During the experiment, which has been described in detail previously (Stankewitz et al., 2010; Schulte et al., 2016), participants received 15 repetitive painful trigeminal stimulations by administering gaseous ammonia (mixed in a stream of air) into the left nostril and 15 puffs of air, 15 trials of rose scent, and 15 repetitive visual stimuli as control conditions. All stimuli were presented in a pseudo-randomized order, with painful trigeminal stimuli not presented in immediate succession. After each trial, participants were asked to rate the intensity as well as the unpleasantness on a numeric rating scale from 0 to 100, each of which was tested with a paired t-test if the criteria of the Gaussian distribution were fulfilled or otherwise with a Wilcoxon signed-rank test. Standardized headache calendars were collected at visit 1, visit 2, and in a follow-up interview by telephone and email after a total of 3 months of therapy (Figure 2A).

MRI acquisition

Request a detailed protocol

All images were recorded at a Siemens PRISMA 3T MR system (Siemens, Erlangen, Germany) using a 64-channel head coil. During the experimental protocol, 1095 functional images were acquired for each subject and for each session using an echoplanar imaging sequence optimized for blood-oxygenation level-dependent (BOLD) brainstem imaging (Schulte et al., 2016): voxel size 1.3 × 1.3 × 2.5 mm³, 38 axial slices (no gap), repetition time 2.64 s, echo time 28 ms, flip angle 80°, GRAPPA acceleration mode, field of view readout 216 mm, phase partial Fourier 7/8, two saturation pulses were added anterior and posterior to the target volume, which covered the whole volume from the corpus callosum to the foramen magnum (MNI z-range 25–72). Simultaneously, we recorded pulse and breathing (Expression, Philipps, Best, The Netherlands) to correct for cardiovascular artifacts.

Functional imaging was followed by field mapping MRI sequence (repetition time 0,8 s, echo time 1: 5.51 ms, echo time 2: 7.97 ms, flip angle 40°, field of view readout 215 mm) and a high-resolution magnetization-prepared rapid gradient echo sequence image (voxel size 1 mm³, repetition time 2.3 s, echo time 2.98 ms, flip angle 9°, field of view 256 mm², 240 axial slices gap 50%).

Preprocessing

Request a detailed protocol

In general, the preprocessing followed the established basis for functional imaging preprocessing steps consisting of: denoising functional images (spatially adaptive non-local mean filter) (Manjón et al., 2010), realignment and unwarping by the aforementioned field maps, slice time correction, co-registration to the anatomical images, and transformation into Montreal Neurological Institute (MNI) space as implemented in SPM12 (Wellcome Trust Center for Neuroimaging, London, UK). Functional images were then smoothed using a 4 mm³ full width at half maximum Gaussian kernel.

Statistical analysis

Request a detailed protocol

A general linear model (GLM) analysis was used within each participant providing β-estimates which were used for group statistical analysis. These β-values were calculated for each voxel and signify the condition-specific neuronal activity. Therefore, we were using a hemodynamic response function to model all four stimulus conditions (ammonia, rose, air puff, visual) and three confound conditions (key press/assessment, attention task, anticipation phase) by convolving their onsets and durations and applying them as regressors in the GLM. For further correction of movements that were not intercepted by the realignment processing, we included the six movement regressors provided in the realign and unwarp step mentioned earlier. For physiological noise correction, we included an additional 18–20 regressors extracted from each participants’ breath and pulse signals using the approach described by Deckers et al., 2006. For the main effect, the results were corrected for multiple comparisons (family-wise error corrected, p<0.05), for the sub-analysis with a strong a priori hypothesis (see preregistration and Ziegeler et al., 2020), we calculated paired and independent one-sided t-test as implemented in the SPM toolbox and used an uncorrected statistical threshold of p<0.001.

Arterial spin labeling

Request a detailed protocol

As galcanezumab is potentially a vasoactive drug, arterial spin labeling (ASL) was also recorded to exclude that any BOLD changes were due to general changes in cerebral blood flow (CBF). To cover the blood flow of the entire brain, we performed separate measurements of the brainstem and the cerebrum. The ASL sequence used pulsed ASL recorded with 91 repetitions in 17 slices with a TR of 2.6 s (TE 12 ms, 90° flip angle, bolus duration 1800 ms, inversion time 700 ms, PICORE Q2T perfusion mode, voxel size 2 × 2 × 5 mm³). The relative CBF maps calculated by the scanner software were co-registered to the anatomical image, warped into MNI space using the transformation calculated on the anatomical image, and smoothed using a 12 mm isotropic Gaussian kernel again using SPM12 (Wellcome Trust Center for Neuroimaging, London, UK). Results were corrected for multiple comparisons (family-wise error corrected, p<0.05, T>6.1, df = 25).

Effect of trigeminal stimulation

Request a detailed protocol

First, both visits were pooled to observe the effect of the trigeminal nociceptive ammonia stimulation (main effect). A statistical threshold of p<0.05 (FWE corrected for multiple comparisons) was used for a one-sided, independent t-test.

Effect of galcanezumab on trigeminal stimulation

Request a detailed protocol

To estimate significant differences before vs. after administration of galcanezumab, a one-sided, paired t-tests in each voxel was performed using SPM12, 2014 for the contrast of pain > control at visit 1 vs. pain > control at visit 2 (further classified as [ammonia-air puffs]_{visit 1} > [ammonia-air puffs]_{visit 2} and [ammonia-air puffs]_{visit 1} < [ammonia-air puffs]_{visit 2}). A statistical threshold of p<0.001 (uncorrected) was used, since we had a strong a priori hypothesis, see preregistration and Ziegeler et al., 2020.

Differences between galcanezumab (CGRP-L-mAb) and erenumab (CGRP-R-mAb)

Request a detailed protocol

As this study followed a previously conducted study on the CGRP receptor mAb erenumab (Ziegeler et al., 2020), we also compared the two effects caused by these medications. To ensure direct comparability, we also included a subgroup consisting of patients with the same migraine phase in the analyses and compared the respective contrasts ([ammonia-air puffs]_{visit 1} vs. [ammonia-air puffs]_{visit 2}) in a one-sided, two-sample t-test. For this analysis, the existing erenumab data sets were reassessed and included in above-mentioned analysis routine. A statistical threshold of p<0.001 (uncorrected) was used, since we had a strong a priori hypothesis, see preregistration and Ziegeler et al., 2020.

Therapy prediction analysis

Request a detailed protocol

Following the protocol of the previous study (Ziegeler et al., 2020), we defined and preregistered ‘being a responder’ by showing a 30% reduction in headache frequency after 3 treatment months. Since a maximal response was not to be expected at 3 weeks, we used the accepted time span of 3 months to calculate responders (Goadsby et al., 2019). However, we asked all patients at the time of the second scan whether they felt the medication to be effective and the number of patients who answered this subjective question positively, corresponded with the 30% response at 3 months (two-tailed Pearson’s correlation, r=0.778, p<0.001). To detect possible predictors of treatment outcome in the processing of trigemino-nociceptive stimuli, we used the visit 1 contrast [ammonia-air puffs]_{visit 1} and co-varied the model with the individual response after 3 months (in % reduction of monthly headache days [MHD]) in all patients (n=26) in an one-sided, independent t-test as implemented into the SPM toolbox. An additional nuisance covariate of migraine phase was used to eliminate possible variance caused by the presence of headache. A statistical threshold of p<0.001 (uncorrected) was used for hypothesis generation.

Alterations in functional connectivity

Request a detailed protocol

Galcanezumab-induced functional connectivity changes were estimated for the whole group (n=26) by psychophysiological interaction (PPI) analysis (Friston et al., 1997), using the region of interest (i.e. spinal trigeminal nucleus [STN]) resulting from the previous analysis as a starting point and contrasting ammonia (nociceptive input) and air puffs (control condition) between the two visits as before. A statistical threshold of p<0.001 (uncorrected) was used in the one-sided, independent t-test, since we had a strong a priori hypothesis, see preregistration and Schulte et al., 2016; Schulte and May, 2016.

Results

Patient characteristics

Between May 2020 and January 2021, a total of 29 patients were enrolled in the study. One patient administered 120 rather than 240 mg loading dose and technical issues in two cases led to exclusion. Twenty-six patients (19 without aura and 7 with aura) (No authors listed, 2018) were included. Of these, 11 patients fulfilled the formal criteria for chronic migraine and 15 patients for episodic migraine, based on headache diaries. A total of 13 patients were taking medications other than hormonal contraception or dietary supplements for other indications: Eight patients took (because of migraine) antidepressant medication (selective serotonin reuptake inhibitor, serotonin and norepinephrine reuptake inhibitor, or tricyclic antidepressant), two patients took antihypertensive medication (metoprolol, bisoprolol), and another two anticonvulsive drugs (topiramate); two patients were taking L-dopa due to restless legs syndrome. Each of the concomitant medications was used as a covariate in the imaging analysis but had no effect on the results. Patient characteristics are shown in Table 1 and the study flowchart in Figure 1.

Figure 1

Download asset Open asset

Study flowchart.

On the left side the galcanezumab cohort, on the right side the erenumab cohort.

Table 1

Patient characteristics.

Patient characteristics	Galcanezumab	Erenumab
Number	26	27
Female, % (n)	96 (25)	81 (22)
Age, mean ± SD (range), in years	37.81±12.11 (21–60)	39.1±12.2 (22–60)
Disease duration, mean ± SD (range), in years	19.77±12.43 (2–50)	21.7±11.2 (6–43)
Headache frequency, mean ± SD (range), days/month EM: mean ± SD (range), days/month CM: mean ± SD (range), days/month	16.71±8.92 (4–30) 10.28±2.94 (4–14) 25.48±6.32 (16–30)	20.3±8.3 (8–30) 12.08±2.07 (8-14) 26.8±4.71(20–30)
Migraine with and without aura, n	7	9
Migraine without aura, n	19	18
Chronic migraine (ICHD-3), % (n)	42 (11)	56 (15)
Episodic migraine (ICHD-3), % (n)	58 (15)	44 (12)
Same headache state – episodic, % (n)	40 (6)	29 (5)
– chronic, % (n)	60 (9)	71 (12)

Abbreviation: ICHD-3=International Classification of Headache Disorders, 3rd edition, EM: episodic migraine, CM: chronic migraine.

Headache state group/covariation

Because headache state can have a major impact on the results (Schulte et al., 2017), we followed the same protocol as for the erenumab study (Ziegeler et al., 2020) and defined a subgroup with the same headache condition during both sessions (n=15), that is, headache-free or headache on both visit days to control for possible changes due to headache phase. Additionally, the headache state was also modulated as a covariate of nuisance (noted below in each respective analysis).

Fourteen patients had headache on visit 1, whereas only 11 patients had headache on visit 2, resulting in a total of 15 patients (episodic n=6, chronic n=9) who had equal headache (n=7) or no headache (n=8) on both dates (for erenumab study: 17 patients [episodic n=5, chronic n=12] with headache [n=11] and no headache [n=6] on both dates). In patients with headache, the intensity of headache did not differ between both visits.

Responder group

Ten out of 26 (38%) of the included patients were responders (50% decrease in MHD), while 18 out of 26 (69%) showed a decrease of 30% in MHD and were defined as responders since this group consisted (following the European prescribing instructions) of difficult-to-treat migraine patients and also according to previous studies (Ziegeler et al., 2020; Overeem et al., 2022) and the preregistration.

Behavior

The behavioral data between visit 1 and visit 2 for intensity and unpleasantness ratings of the nociceptive stimulation was not significantly different (p_*Bonferroni = 0.094 with T = 2.4, df = 25, average difference 6.62/100, 95% confidence interval [CI] –0.90 to 11.58 for intensity; p_*Bonferroni = 0.156 with T = 2.18, df = 25, average difference 5.2/100, 95%, CI –0.28 to 10.12 for unpleasantness ratings) tested by a two-sided, paired t-test (corrected for multiple comparisons by Bonferroni correction) with given normal distribution. The same was true for the other conditions: control (air puff), olfactory input (rose scent), and visual stimulation.

Arterial spin labeling

ASL was performed in a two-step whole-brain protocol and showed no changes in relative CBF (corrected for multiple comparisons [family-wise error corrected, p<0.05, T>6.1, df = 25]).

Effect of trigeminal stimulation

The (pooled) main effect of trigeminal nociceptive stimulation revealed cortical and subcortical structures mediating central pain/salience processing, like ipsilateral STN, contralateral thalamus, insula, and cerebellum with a statistical threshold of p<0.05, corrected for multiple comparisons (FWE).

Effect of galcanezumab on trigeminal stimulation

Galcanezumab treatment (n=26; [ammonia-air puffs]_visit1>[ammonia-air puffs]_visit2) was associated with a decrease in activation in the right hypothalamus, right cerebellum, and cerebellar vermis (Table 2A), with a statistical threshold of p<0.001 (uncorrected) and a minimum cluster extension of 25 voxels, following previous studies with previously generated and preregistered hypotheses. The inverse contrast ([ammonia-air puffs] _visit1<[ammonia-air puffs]_visit2) showed no significant effects.

Table 2

Details of the statistical results of the functional magnetic resonance imaging (fMRI) analyses comparing before and after treatment as well as responders and non-responders.

Anatomical region	Cluster size(voxels), n	T value	MNI coordinates(x,y,z)
*(A) Visit 1>visit 2, all participants (n=26, contrast* *[ammonia-air puffs]_visit1 > [ammonia-air puffs]_visit2,threshold: p<0.001 [uncorrected], T>3.45, minimum cluster extent 25 voxels, df = 25)*
R hypothalamus	49	4.99	8, –14, –7
R cerebellum	48	5.0	20, –40, –25
R cerebell. vermis	52	4.09	3, –67, –34
(B) Responder (n=8)>non-responder (n=7) with same headache state on both days *(contrast [ammonia-air puffs]_visit1 > [ammonia-air puffs]_visit2, threshold p<0.001 [uncorrected], T>3.85, minimum cluster extent 10 voxels, df = 14)*
R inf. parietal lobule	91	9.17	58, –44, 21
L precentral gyrus	26	5.99	−48, –11, 11
L parahippocampal gyrus	12	5.06	−18, –11, –19
L superior temporal gyrus	13	4.76	–36, 15,–26
L inf. parietal lobule	17	4.7	46, –42, 22
R lentiform nucleus	16	4.67	24, –2, 4
R parahippocampal gyrus	27	4.63	27, –28, –12
L insula	26	4.56	−45, –27, 19
L hypothalamus	13	4.47	−6, –17, –6

(A) Main findings of trigeminal pain processing alterations after administration of the CGRP-ligand monoclonal antibody galcanezumab. The contrast shows more neuronal activity on visit 1 than on visit 2, therefore decrease driven by galcanezumab and was tested with a one-way paired t-test. (B) Subgroup analysis of patients having the same migraine/headache state on both visits (ictal and ictal or interictal and interictal). Main findings specific for being responder (responder showing a higher decrease after galcanezumab than non-responder) tested with a one-way independent sample t-test. Note that left-right activation patterns in near proximity twin structures such as the hypothalamus are not necessarily side-locked (a left activation excludes a right activation) but may be due to statistical thresholding.

Comparing responders vs. non-responders in a subgroup consisting only of patients within the same migraine phases (visit 1 and visit 2 either with or without headache, n=15; eight responders), we found a responder-specific decrease in BOLD activity in central areas such as the inferior parietal lobule/operculum, the left insula, the parahippocampal gyrus, as well as the hypothalamus (Figure 2B, Table 2B). Non-responders only showed a specific decrease in the cerebellar tonsils (p<0.001, T=4.81, k_E = 66 voxels, −38, –62, –49 [x, y, z/mm]). Using the 50% responders showed the same results but at lower exploratory threshold levels since only 10 out of 26 of the included patients were 50% responders.

Figure 2

Download asset Open asset

Experimental design and responder-specific main findings.

(A) Flowchart of the experimental design. At both visits, objective rating of pain intensity and unpleasantness on a numeric rating scale (NRS), session-specific questionnaire: current headache (yes/no), strength of headache (NRS), headache frequency (per week), date of last headache. Three months of total therapy duration, patients were contacted via phone and/or email to acquire their headache diaries. (B) Responder-specific decreased neuronal activity during nociception at visit 2 compared to visit 1 in subgroup analysis of migraine-phase matched patients (n=15). Data is shown at statistical threshold of p<0.001 and a minimum cluster size of 10 voxels, masked with gray and white matter study template where a higher t-value means a stronger decrease of activity after treatment for responders in comparison to non-responders.

Galcanezumab (CGRP-L-mAb) vs. erenumab (CGRP-R-mAb)

Exclusion and inclusion criteria between this study and the previous erenumab study (Ziegeler et al., 2020) were the same for reasons of comparability. Patient characteristics (sex, age, migraine type) and behavioral ratings (pain severity and unpleasantness of stimuli) were also comparable and not significantly different. It is therefore feasible to directly compare both studies in the same analysis and we focused on the migraine-phase equal subgroup analysis (n=15 galcanezumab vs. n=17 erenumab) to eliminate possible influences of different migraine phases. Erenumab specifically decreased activation in the operculum, insula, thalamus, and cerebellum (Figure 3, Table 3A), whereas galcanezumab led to a specific decrease in activity in the hypothalamic area, left thalamus, as well as in pontine region (Figure 3, Table 3B). The inverse contrast showed no results.

Figure 3

Download asset Open asset

Galcanezumab vs erenumab.

Erenumab (red)/galcanezumab (blue)-specific decreased neuronal activity during nociception at visit 2 compared to visit 1 in subgroup analysis of migraine-phase matched patients (n=15 vs. 17). Data is shown at statistical threshold of p<0.001 and a minimum cluster size of 10 voxels.

Table 3

Details of the statistical results of the functional magnetic resonance imaging (fMRI) analyses comparing calcitonin gene-related peptides (CGRP) receptor monoclonal antibody erenumab and CGRP ligand monoclonal antibody galcanezumab.

Anatomical region	Cluster size(voxels), n	T value	MNI coordinates(x,y,z)
(A) Erenumab >Galcanezumab (Visit 1>visit 2, migraine-phase equal subgroup [n=15 vs. n=17], contrast [ammonia-air puffs]_visit1 > [ammonia-air puffs]_visit2,threshold: p<0.001 [uncorrected], T>3.39, minimum cluster extent 10 voxels, df = 30)
L operculum	416	5.61	−60, –5, 8
L cerebellum	122	5.59	−16, –65, –22
R cerebellum	207	5.23	42, –51, –32
R supramarginal gyrus	37	5.11	64, –45, 23
R thalamus	67	5.0	6, –15, 11
R hippocampus	74	4.88	21, –32, –10
L thalamus	53	4.71	−9, –7, 11
L temporal pole/insula	202	4.5	–54, 17, –1
L superior temporal gyrus	55	4.42	−55, –33, 16
Locus coeruleus	30	4.41	4, –37, –9
R insula	82	4.37	47, 14, –7
R operculum	19	4.15	62, –2, 7
R middle temporal gyrus	21	4.07	55, –6, –22
R lingual gyrus	11	3.9	16, –49, 0

(B) Galcanezumab >Erenumab (Visit 1>visit 2, migraine-phase equal subgroup [n=15 vs. n=17], contrast [ammonia-air puffs]_visit1 > [ammonia-air puffs]_visit2,threshold: p<0.001 [uncorrected], T>3.39, minimum cluster extent 10 voxels, df = 30)
Pons	40	4.09	−1, –18, –38
R substantia nigra	13	4.01	9, –14, –10
L thalamus	13	3.93	−18, –15, 0
R hypothalamus	12	3.89	4, –19, 0

Main findings of trigeminal pain processing alterations driven by the administration of the CGRP receptor monoclonal antibody erenumab in contrast to the CGRP ligand monoclonal antibody galcanezumab in the migraine-phase equal subgroup. (A) Erenumab-specific decrease in neuronal activity/BOLD across the two visits. (B) Galcanezumab-specific decrease in neuronal activity/BOLD across the two visits. Both results stem from one-sided independent t-tests.

Therapy prediction analysis

Contrasting the activation maps on visit 1 [n=26; ammonia-air puffs]_visit1 with the response (% reduction MHD) 3 months later showed that the STN highly co-varied with the response (Figure 4A, Table 4A) at a statistical threshold of p<0.001. Given that the STN has a central role in the ascending pathway for processing trigemino-nociceptive stimuli, we extracted the parameter estimates from the F-contrast in the peak voxel of the STN and correlated it with the response to galcanezumab (% reduction of MHD). The Pearson’s correlation showed a significant positive relationship between visit 1 STN activation and the response of each participant (two-tailed Pearson’s correlation, r=0.63, p<0.001, Figure 4B).

Figure 4

Download asset Open asset

Prediction/functional connectivity changes after galcanezumab administration.

(A) Two-tailed Pearson’s correlation of visit 1 spinal trigeminal nucleus (STN) neuronal activity (before administration of galcanezumab) and response (in % change of MHD). Responder (>30% reduction) are marked in blue; non-responders (<30% reduction) are colored red. STN activation is shown in the image above the legend – visit 1 trigeminal-nociceptive activation (ammonia-air puffs) covaries with response (in percent reduction of monthly headache days [MHD]). Data is shown at statistical threshold of p<0.001 and a minimum cluster size of 10 voxels. (B) Results of the psychophysiological interaction (PPI) analysis, showing the decreased connectivity between STN (seed region [A]) and hypothalamus. Data is shown at statistical threshold of p<0.001.

Table 4

Statistical results of the functional magnetic resonance imaging (fMRI) analyses regarding the coherence with the reduction of headache days after treatment and changes in functional connectivity.

Anatomical region	Cluster size(voxels), n	T value	MNI coordinates(x,y,z)
(A) Covariation of monthly headache day reduction (in % reduction) in the contrast (ammonia-air puffs) _visit1, all participants (n=26) threshold: p<0.001 [uncorrected], T>3.48, minimum cluster extent 25 voxels, df = 25
R middle temporal gyrus	110	4.9	49, –8, –14
STN	83	4.65	4, –42, –53
(B) PPI analysis of STN (region of interest 4, −42,–53), all participants (n=26) in the contrast (ammonia-air puffs) _visit1> [ammonia-air puffs]_visit2, threshold: p<0.001 [uncorrected], T>3.45, minimum cluster extent 25 voxels, df = 25
L superior temporal gyrus	62	5.57	−52, –47, 18
R hyopthalamus	52	4.24	10, 5, –12

(A) Main findings of the covariation of the response to galcanezumab by the activation pattern of neuronal activity at visit 1 (before administration of galcanezumab). This result was gained by introducing the covariate of reduction in headache days into a one-sided, one-sample t-test. (B) Results from the psychophysiological interaction analysis of the above-mentioned STN activation.

Alterations in functional connectivity

In the PPI connectivity analysis (n=26), we used the STN cluster found in the therapy prediction analyses and found a decrease in connectivity of the STN on visit 2 with the hypothalamus and also in the superior temporal gyrus (Figure 4C, Table 4B). Connectivity increases with the STN were found to areas of the cerebellum, middle temporal gyrus, and insula.

Discussion

Our functional imaging data suggest that galcanezumab therapy is associated with a decrease of hypothalamic activation in all patients following galcanezumab therapy, which is even more decreased in responders. This is in line with our previous study in erenumab patients (Ziegeler et al., 2020), where responders also showed decreased hypothalamic activity in comparison to non-responders. Unlike non-responders to galcanezumab in this study, we did not find decreased hypothalamic activation in non-responders to erenumab. Since this could be due to threshold effects (i.e. erenumab also reduced hypothalamic activity but at a lower statistical threshold), we compared both studies directly in the same analysis. This comparison showed that the hypothalamic decrease following galcanezumab is indeed specific to galcanezumab as it was not present in erenumab patients. Of note, the other comparison in this medication-specific analysis showed that the general effects seen after erenumab medication (Ziegeler et al., 2020), that is, a decrease in activation in the thalamus, right middle temporal gyrus, right lingual gyrus, and left operculum are specifically only found in the erenumab-treated patients and not after galcanezumab therapy. It is important to state that these findings are related to the response to a trigeminal-nociceptive stimulus, and not some general difference in brain state such as resting state changes for example.

Our findings suggest that the antibodies to the ligand and the receptor exhibit a similar modulatory change in the hypothalamic area if patients respond to treatment, whereas the general change due to administration of the medication is different and specific for each antibody type. Since the perception of nociceptive input elicited by our paradigm was not significantly different between the two sessions, neither in responders nor in non-responders or within the whole cohort, an external influence on the central processing of these stimuli can be excluded.

The hypothalamus plays a central role in migraine pathophysiology (Alstadhaug, 2009 Stankewitz et al., 2010; Overeem et al., 2022). Studies in the rhesus monkey confirmed the expression of CGRP receptors in these areas (Eftekhari et al., 2016). However, a direct effect of monoclonal CGRP antibodies in the central nervous system in these areas is unlikely, because the blood-brain barrier is relatively impermeable to substances of this large molecular mass (Johnson et al., 2019; Noseda et al., 2020). However, the hypothalamus, and indeed the entire central nervous system, is subject to cyclic changes in CGRP concentrations in migraine patients which has been shown in cubital (Greco et al., 2020) and also plasma cranial venous blood analyses (Goadsby et al., 1990). Since small amounts of CGRP can cross the blood-brain barrier (Johnson et al., 2019), one could argue that central nervous (hypothalamic) receptors are subsequently stimulated by increased cranial CGRP levels during a migraine attack. Systemic administration of galcanezumab binds free circulating CGRP (Kielbasa and Helton, 2019) and this modulates the ability to cross the blood-brain barrier from the periphery to the cerebrospinal fluid compartment. Assuming intra- and extrathecal CGRP concentrations pursue an equilibrium or equal concentration gradients, such a decrease in functional CGRP due to antibodies could subsequently lead to reduced CGRP-mediated activation of hypothalamic and other brain areas that express CGRP receptors. This would account for the reduction of hypothalamic activity in all subjects as well as in the responder vs. non-responder comparison. Of note, studies showing an elevated CGRP level in migraine attacks (Goadsby et al., 1990) have to be seen in the light of studies showing that administration of CGRP may trigger migraine-like attacks (Hansen et al., 2010) but is in itself not painful (Asghar et al., 2016), suggesting that the elevation of CGRP levels in migraine attacks may be a consequence and not the primary trigger of the attack. That being said, erenumab elicits other modes of action (e.g. intracellular) than just blocking or inactivating the receptor (Bhakta et al., 2021). This could explain why we see different central responses to erenumab and galcanezumab, and it could also explain why some non-responders to CGRP-Ab (erenumab or galcanezumab) may indeed have a better response by switching to the other antibody class (Ziegeler and May, 2020).

Given that certain areas of the central nervous system like the area postrema or the hypothalamus are not completely protected by the blood-brain barrier (Eftekhari et al., 2016), one could also argue that the antibodies could still reach the hypothalamus even with intact blood-brain barrier in a very small but effective amount. This could be different for the antibody type and thus could explain why we see different central responses to erenumab and galcanezumab, and it could also explain why some non-responders to CGRP-Ab (erenumab or galcanezumab) may indeed have a better response by switching to the other antibody class (Ziegeler and May, 2020).

We also found that the STN, which plays a central role in the ascending trigemino-nociceptive pathway, co-varies in its activity already at visit 1 with the response, that is, the efficacy to galcanezumab. In other words, the higher the STN activity to nociceptive stimuli before administration of the drug, the more likely that galcanezumab is effective. One could argue that a sensitization of the STN might be the missing link why some patients benefit from anti-CGRP therapy, others do not. The STN exhibits a particularly high density of CGRP due to termination of afferents from the trigeminal ganglion (Eftekhari et al., 2016; Eftekhari and Edvinsson, 2011). Recent findings on molecular brainstem and trigeminal ganglion mechanisms suggest that presynaptic CGRP-evoked intracellular mechanisms may lead to a possibly long-term modulation of signaling cascades, which in their consequence could lead to a facilitation and strengthening of (trigeminal) nociceptive transmission (Messlinger, 2018). However, we did not investigate sensitization levels or allodynia in our patients and explanations for this correlation between STN activity and treatment response stays speculative at the moment. It is interesting, however, that we found that galcanezumab also decreased the well-known functional connection between the STN and the hypothalamus (Schulte et al., 2016; Schulte and May, 2016; Schulte et al., 2017). Our data suggest that, in addition to peripheral mechanisms, the combination of decreased hypothalamic activity and a simultaneous decrease of the hypothalamic-brainstem connectivity is needed for CGRP antibodies to be effective. The question of whether this is the cause or consequence of blocking the CGRP ligand cannot be answered with our study.

Some limitations of this study have to be discussed: The number of patients (n=26) in this longitudinal design is difficult to achieve but allows only small subgroups (such as responder/non-responder). We note that the groups were big enough to find meaningful differences between groups and moreover, these findings confirmed our preregistered a priori hypothesis. One could ask the question why we defined treatment response of 30%, rather than 50%. We followed the IHS (International Headache Society) efficacy parameter ‘attack frequency’ although most of our patients reported a marked effect of attack severity rather than frequency reduction. In clinical practice, this means that many patients who show significant improvement in quality of life and reduction in acute medication consumption, and are therefore responders, may still have nominally a high migraine attack frequency. Therefore, because of the methodology used, the observations of this study in terms of responders may be limited to patients who are frequency responders. One could argue that only in frequency responders the hypothalamus is effectively targeted using mAbs and not in severity responders. Unfortunately, given the small group sizes of intensity vs. frequency responders, a subgroup analysis in this regard was not possible, which would have been necessary to answer this question directly.

We also note that, following national guidelines (Diener et al., 2020), patients receiving mAbs must have demonstrated an unsuccessful treatment effect of a least four different other preventatives and are so-called severe treatment non-responders. Both facts prompted us to focus on a 30% frequency reduction from baseline, which is sufficient to demonstrate robust differences between groups.

Data availability

All de-identified contrast imaging data and variables, involving the documentation of the processing procedure are available: Basedau, Hauke et al. (2022), Migraine monoclonal antibodies against CGRP change brain activity depending on ligand or receptor target: a functional magnetic resonance imaging study, Dryad, Dataset, https://doi.org/10.5061/dryad.w3r2280t2.

The following data sets were generated

1. Basedau H
2. Sturm L
3. Mehnert J
4. Peng K
5. Schellong M
6. May A
(2022) Dryad Digital Repository
Migraine monoclonal antibodies against CGRP change brain activity depending on ligand or receptor target: a functional magnetic resonance imaging study.
https://doi.org/10.5061/dryad.w3r2280t2

References

1. Alstadhaug KB
(2009) Migraine and the hypothalamus
Cephalalgia : An International Journal of Headache 29:809–817.
https://doi.org/10.1111/j.1468-2982.2008.01814.x
- PubMed
- Google Scholar
(2016) Calcitonin Gene-Related Peptide Modulates Heat Nociception in the Human Brain - An fMRI Study in Healthy Volunteers
PLOS ONE 11:e0150334.
https://doi.org/10.1371/journal.pone.0150334
- PubMed
- Google Scholar
1. Bhakta M
2. Vuong T
3. Taura T
4. Wilson DS
5. Stratton JR
6. Mackenzie KD
(2021) Migraine therapeutics differentially modulate the CGRP pathway
Cephalalgia 41:499–514.
https://doi.org/10.1177/0333102420983282
- PubMed
- Google Scholar
1. Deckers RHR
2. van Gelderen P
3. Ries M
4. Barret O
5. Duyn JH
6. Ikonomidou VN
7. Fukunaga M
8. Glover GH
9. de Zwart JA
(2006) An adaptive filter for suppression of cardiac and respiratory noise in MRI time series data
NeuroImage 33:1072–1081.
https://doi.org/10.1016/j.neuroimage.2006.08.006
- PubMed
- Google Scholar
1. Diener HC
2. Förderreuther S
3. Gaul C
4. Giese F
5. Hamann T
6. Holle-Lee D
7. Jürgens TP
8. Kamm K
9. Kraya T
10. Lampl C
11. May A
12. Reuter U
13. Scheffler A
14. Tfelt-Hansen P
(2020) Prevention of migraine with monoclonal antibodies against CGRP or the CGRP receptor: Addition to the S1 guideline: Therapy of migraine attacks and prevention of migraine: Recommendations of the Germany Society of Neurology and the German Migraine and Headache Society
Neurological Research and Practice 2:11.
https://doi.org/10.1186/s42466-020-00057-1
- PubMed
- Google Scholar
1. Edvinsson L
(2017) The Trigeminovascular Pathway: Role of CGRP and CGRP Receptors in Migraine
Headache 57 Suppl 2:47–55.
https://doi.org/10.1111/head.13081
- PubMed
- Google Scholar
(2018) CGRP as the target of new migraine therapies - successful translation from bench to clinic
Nature Reviews. Neurology 14:338–350.
https://doi.org/10.1038/s41582-018-0003-1
- PubMed
- Google Scholar
(2019) C-fibers may modulate adjacent Aδ-fibers through axon-axon CGRP signaling at nodes of Ranvier in the trigeminal system
The Journal of Headache and Pain 20:105.
https://doi.org/10.1186/s10194-019-1055-3
- PubMed
- Google Scholar
1. Eftekhari S
2. Edvinsson L
(2011) Calcitonin gene-related peptide (CGRP) and its receptor components in human and rat spinal trigeminal nucleus and spinal cord at C1-level
BMC Neuroscience 12:112.
https://doi.org/10.1186/1471-2202-12-112
- PubMed
- Google Scholar
1. Eftekhari S
2. Gaspar RC
3. Roberts R
4. Chen T-B
5. Zeng Z
6. Villarreal S
7. Edvinsson L
8. Salvatore CA
(2016) Localization of CGRP receptor components and receptor binding sites in rhesus monkey brainstem: A detailed study using in situ hybridization, immunofluorescence, and autoradiography
The Journal of Comparative Neurology 524:90–118.
https://doi.org/10.1002/cne.23828
- PubMed
- Google Scholar
1. Friston KJ
2. Buechel C
3. Fink GR
4. Morris J
5. Rolls E
6. Dolan RJ
(1997) Psychophysiological and modulatory interactions in neuroimaging
NeuroImage 6:218–229.
https://doi.org/10.1006/nimg.1997.0291
- PubMed
- Google Scholar
(1990) Vasoactive peptide release in the extracerebral circulation of humans during migraine headache
Annals of Neurology 28:183–187.
https://doi.org/10.1002/ana.410280213
- PubMed
- Google Scholar
1. Goadsby PJ
2. Dodick DW
3. Martinez JM
4. Ferguson MB
5. Oakes TM
6. Zhang Q
7. Skljarevski V
8. Aurora SK
(2019) Onset of efficacy and duration of response of galcanezumab for the prevention of episodic migraine: a post-hoc analysis
Journal of Neurology, Neurosurgery, and Psychiatry 90:939–944.
https://doi.org/10.1136/jnnp-2018-320242
- PubMed
- Google Scholar
1. Greco R
2. De Icco R
3. Demartini C
4. Zanaboni AM
5. Tumelero E
6. Sances G
7. Allena M
8. Tassorelli C
(2020) Plasma levels of CGRP and expression of specific microRNAs in blood cells of episodic and chronic migraine subjects: towards the identification of a panel of peripheral biomarkers of migraine?
The Journal of Headache and Pain 21:122.
https://doi.org/10.1186/s10194-020-01189-0
- PubMed
- Google Scholar
1. Hansen JM
2. Hauge AW
3. Olesen J
4. Ashina M
(2010) Calcitonin gene-related peptide triggers migraine-like attacks in patients with migraine with aura
Cephalalgia 30:1179–1186.
https://doi.org/10.1177/0333102410368444
- PubMed
- Google Scholar
(2019) Peripheral and central nervous system distribution of the CGRP neutralizing antibody [¹²⁵I] galcanezumab in male rats
Cephalalgia : An International Journal of Headache 39:1241–1248.
https://doi.org/10.1177/0333102419844711
- PubMed
- Google Scholar
1. Kielbasa W
2. Helton DL
(2019) A new era for migraine: Pharmacokinetic and pharmacodynamic insights into monoclonal antibodies with A focus on galcanezumab, an anti-CGRP antibody
Cephalalgia 39:1284–1297.
https://doi.org/10.1177/0333102419840780
- PubMed
- Google Scholar
(2010) Adaptive non-local means denoising of MR images with spatially varying noise levels
Journal of Magnetic Resonance Imaging 31:192–203.
https://doi.org/10.1002/jmri.22003
- PubMed
- Google Scholar
(2017) Fremanezumab-A Humanized Monoclonal Anti-CGRP Antibody-Inhibits Thinly Myelinated (Aδ) But Not Unmyelinated (C) Meningeal Nociceptors
The Journal of Neuroscience 37:10587–10596.
https://doi.org/10.1523/JNEUROSCI.2211-17.2017
- PubMed
- Google Scholar
1. Messlinger K
(2018) The big CGRP flood - sources, sinks and signalling sites in the trigeminovascular system
The Journal of Headache and Pain 19:22.
https://doi.org/10.1186/s10194-018-0848-0
- PubMed
- Google Scholar
1. No authors listed
(2018) Headache Classification Committee of the International Headache Society (IHS) The International Classification of Headache Disorders, 3rd edition
Cephalalgia: An International Journal of Headache 38:1–211.
https://doi.org/10.1177/0333102417738202
- PubMed
- Google Scholar
1. Noseda R
2. Schain AJ
3. Melo-Carrillo A
4. Tien J
5. Stratton J
6. Mai F
7. Strassman AM
8. Burstein R
(2020) Fluorescently-labeled fremanezumab is distributed to sensory and autonomic ganglia and the dura but not to the brain of rats with uncompromised blood brain barrier
Cephalalgia : An International Journal of Headache 40:229–240.
https://doi.org/10.1177/0333102419896760
- PubMed
- Google Scholar
1. Overeem LH
2. Peikert A
3. Hofacker MD
4. Kamm K
5. Ruscheweyh R
6. Gendolla A
7. Raffaelli B
8. Reuter U
9. Neeb L
(2022) Effect of antibody switch in non-responders to A CGRP receptor antibody treatment in migraine: A multi-center retrospective cohort study
Cephalalgia : An International Journal of Headache 42:291–301.
https://doi.org/10.1177/03331024211048765
- PubMed
- Google Scholar
1. Schulte LH
2. May A
(2016) The migraine generator revisited: continuous scanning of the migraine cycle over 30 days and three spontaneous attacks
Brain 139:1987–1993.
https://doi.org/10.1093/brain/aww097
- PubMed
- Google Scholar
(2016) Physiological brainstem mechanisms of trigeminal nociception: An fMRI study at 3T
NeuroImage 124:518–525.
https://doi.org/10.1016/j.neuroimage.2015.09.023
- PubMed
- Google Scholar
(2017) Hypothalamus as a mediator of chronic migraine
Neurology 88:2011–2016.
https://doi.org/10.1212/WNL.0000000000003963
- PubMed
- Google Scholar
Website
1. SPM12
(2014) Statistical Parametric Mapping
Accessed April 1, 2019.

https://www.fil.ion.ucl.ac.uk/spm/software/spm12/
1. Stankewitz A
2. Voit HL
3. Bingel U
4. Peschke C
5. May A
(2010) A new trigemino-nociceptive stimulation model for event-related fMRI
Cephalalgia : An International Journal of Headache 30:475–485.
https://doi.org/10.1111/j.1468-2982.2009.01968.x
- PubMed
- Google Scholar
1. Ziegeler C
2. May A
(2020) Non-Responders to Treatment With Antibodies to the CGRP-Receptor May Profit From a Switch of Antibody Class
Headache 60:469–470.
https://doi.org/10.1111/head.13729
- PubMed
- Google Scholar
1. Ziegeler C
2. Mehnert J
3. Asmussen K
4. May A
(2020) Central effects of erenumab in migraine patients: An event-related functional imaging study
Neurology 95:e2794–e2802.
https://doi.org/10.1212/WNL.0000000000010740
- PubMed
- Google Scholar

Decision letter

Allan Basbaum

Reviewing Editor; University of California San Francisco, United States
Jeannie Chin

Senior Editor; Baylor College of Medicine, United States
Andrew Charles

Reviewer; University of California, Los Angeles, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Decision letter after peer review:

Thank you for submitting your article "Migraine monoclonal antibodies against CGRP change brain activity depending on ligand or receptor target: a functional magnetic resonance imaging study" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Jeannie Chin as the Senior Editor. The following individual involved in the review of your submission has agreed to reveal their identity: Andrew Charles (Reviewer #3).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

1) The Reviewers have made several suggestions that should be addressed in revising your manuscript. Also it would be very helpful to elaborate on the interesting distinction between the effect of galcanezumab and erenumab.

Reviewer #1 (Recommendations for the authors):

The paper reports a well conducted, carefully analyzed data set with important implications for understanding migraine therapeutics. My comments are modest and advisory.

A. The Introduction states CGRP mABs are impermeable with regard to the blood-brain barrier. An article is cited that, consistent with what is known, shows their permeability is limited not zero (1). The authors could consider recasting the sentence in line with the citation.

B. Not all responders to galcanezumab will do so in the first two weeks (2); would this have blunted the responder/non-responder comparison at the two week point? Could it be argued that a longer study time might be considered going forward?

C. The authors cite six papers for the Method; perhaps one or two would be sufficient?

D. Given the highly personalized nature of pain, it could be argued that VAS scales by virtue of the underlying biological variability between patients, are poor candidates for being treated as more than within subject ordinal measures? It could be argued that the Wilcoxon test is preferable as has been mentioned by the authors.

E. Given response rates to galcanezumab vary by frequency of migraine days, as evidenced by headline responses in episodic (3, 4) and chronic migraine (5), did migraine day frequency effect outcomes? Tables one might usefully offer the frequency data for both the episodic and chronic migraine groups.

F. Regarding concomitant preventives, it is a practical limitation of studies that stopping preventives can be problematic. If the patients taking concomitant preventives with established efficacy in migraine are excluded, and not all those listed un characteristics would fulfill that criterion, does this alter the outcome?

G. It would be generally accepted that a 30% reduction in migraine days in episodic migraine is a modest effect; if a 50% reduction had been used as the binary response outcome for all participants, would the result be different? The authors very sensibly point out the severity change argument in the limitations section. It could be argued that applying the 50% rule might sharpen the differences.

References

1. Johnson KW, Morin SM, Wroblewski VJ, Johnson MP. Peripheral and central nervous system distribution of the CGRP neutralizing antibody [(125)I] galcanezumab in male rats. Cephalalgia. 2019;39:1241-8.

2. Goadsby PJ, Dodick DW, Martinez JM, Ferguson MB, Oakes TM, Zhang Q, et al., Onset of Efficacy and Duration of Response of Galcanezumab for the Prevention of Episodic Migraine: A Post-Hoc Analysis. Journal of Neurology, Neurosurgery and Psychiatry. 2019;90:939-44.

3. Skljarevski V, Matharu M, Millen BA, Ossipov MH, Kim BK, Yang JY. Efficacy and safety of galcanezumab for the prevention of episodic migraine: Results of the EVOLVE-2 Phase 3 randomized controlled clinical trial. Cephalalgia. 2018;38:1442-54.

4. Stauffer VL, Dodick DW, Zhang Q, Carter JN, Ailani J, Conley RR. Evaluation of Galcanezumab for the Prevention of Episodic Migraine: The EVOLVE-1 Randomized Clinical Trial. JAMA Neurol. 2018;75:1080-8.

5. Detke HC, Goadsby PJ, Wang S, Friedman DI, Selzler K, Aurora SK. Galcanezumab in chronic migraine. The randomized, double-blind, placebo-controlled REGAIN study. Neurology (Minneap). 2018;91:e2211-e21.

Reviewer #2 (Recommendations for the authors):

In this study, Basedau et al., seek to further understand the mechanism of action of CGRP monoclonal antibodies for migraine. These drugs likely do not gain substantial access to the CNS given their size, yet they show efficacy for a neurological disorder that has numerous aspects mediated by the CNS. Thus, the exact mechanism by which they treat migraine is not clear. This group performed a similar study recently using a distinct CGRP antibody that binds to the receptor (erenumab) and the current study extends this work using a CGRP antibody that binds to the peptide (galcanezumab). Using fMRI in 26 migraine patients, the authors performed scans to assess brain activity following nasal administration of ammonia, rose odor, or visual checkerboard stimulation both prior to and three weeks following galcanezumab treatment. Unpleasantness ratings were collected and headache diaries were compiled for each patient, including efficacy of the drug for migraine attacks during the study. Approximately 70% of the patients reported efficacy of galcanezumab for migraine attacks and were defined as responders to the drug. Despite efficacy for headache in the majority of subjects, there were no differences in intensity or unpleasantness ratings for any of the stimuli following drug treatment compared to prior. In contrast, galcanezumab treatment decreased trigeminal stimuli-mediated activation of a number of CNS regions, including the hypothalamus, in all subjects. The regions where stimulated activation were reduced by galcanezumab showed distinct patterns when broken down by responder vs non-responder groups with respect to drug efficacy for migraine attacks. Additionally, when the new data with galcanezumab were compared with prior data generated using erenumab, distinct drug-induced changes in activation patterns were observed. Finally, there was a highly co-variant relationship between the effect of galcanezumab on trigeminal stimulation mediated activation patterns and headache efficacy for responses within the spinal trigeminal nucleus (STN) and functional connectivity of the STN. This latter finding was the basis of a conclusion by the authors that response patterns in the STN could be predictive and/or required for galcanezumab efficacy for migraine. In general, the studies add to the understanding of how CGRP monoclonal antibodies influence CNS responses to painful stimuli.

The manuscript is well written and the conclusions are supported by the data. There were some questions and comments identified related to the study.

1. The authors suggest in the discussion that galcanezumab binding to peripheral CGRP could subsequently influence the circulating levels of CGRP and as a result, modulate its ability to cross the blood brain barrier and modulate hypothalamic function. While this is true for galcanezumab, this is unlikely to be true for erenumab since it binds the receptor. Is it possible that these types of differences are responsible for the distinct patterns of effects observed when comparing the two antibodies? The authors discuss other possible explanations in the paper (starting at line 344) but this.

2. Given that effects of gancanezumab were observed in the hypothalamus, and the hypothalamus has areas not protected by the blood brain barrier, could the observed effects on this region be due to direct access of drug to this CNS region? It might be helpful to discuss the specific regions of the hypothalamus not protected by the BBB and whether these could contribute to the observed effects. This was included in the prior erenumab paper but would be helpful to discuss here.

3. The prior erenumab study included 81% females while the current study is 96% female. While not a major difference, do the authors expect this has any influence on the findings of the two studies?

4. It might be helpful to include more detail on the information collected in the headache calendars. This may help readers assess the level of attention subjects paid to the qualities of their migraines across the 3 months. Since this study uses drug efficacy as a method to segregate patients responses to trigeminal stimulation, and efficacy was entirely subjective on the part of the subjects, it seems helpful to have more detail on what was asked of the patients between study visits.

The current work is quite similar to the prior study by these same authors using erenumab. It would help increase the impact for eLife if the authors could provide more information (or data) that clarifies what the major advance is in this work over the prior work.

Reviewer #3 (Recommendations for the authors):

It is now well established that CGRP plays a fundamental role in migraine, and therapies targeting CGRP can be highly effective in reducing frequency and severity of migraine. However, multiple aspects of the specific mechanisms of CGRP in migraine remain poorly understood. Ongoing questions include: What triggers the release of CGRP that is involved in migraine? Where does migraine-related CGRP release occur, and where does CGRP binding mediate migraine-related effects? Are the therapeutic effects of CGRP-targeted treatments mediated primarily by the peripheral or central nervous system, or both?

The current study aims to shed light on the potential therapeutic mechanisms of CGRP-targeted treatments by using fMRI to examine changes in brain activity and connectivity before and after treatment with galcanezumab. These results are compared with a previous similar study of erenumab. The results clearly show central effects of the antibodies, notably in the hypothalamus, and a comparison of the two treatments suggests that there are distinctive patterns of reduction in brain activity resulting from each of the two therapies. The results also show that activation of the spinal trigeminal nucleus by noxious stimulation is correlated with the response to therapy. The results are an important description of the brain effects CGRP-targeted antibodies. Although it is widely presumed that because of their size these antibodies must be working peripherally, the studies do not rule out the possibility that they could be acting in brain regions (like the hypothalamus) that are outside of the blood brain barrier. Although the present study does not definitively answer the question regarding the site(s) of action of the CGRP-targeted therapies, it provides key information about central effects that will enable future studies to address this question. In addition, it identifies patterns of brain activity that could at some point become "biomarkers" of treatment response.

The imaging studies are of high quality, and the experimental design importantly enables within-subject comparison before and after treatment.

There are, however, some sources of variability that require consideration in the interpretation of the results, and some interesting findings that warrant further discussion.

1. The study includes both episodic and chronic migraine patients who are on a variety of different medications. In order to reduce potential migraine phase-related variability, a subgroup of subjects who had either headache or non headache on scan days both pre and post treatment. As it turned out , 14/15 of the phase-consistent patients were scanned on a headache day pre- and post-treatment with galcanezumab, while only 1 was scanned on non-headache days. The fact that nearly all patients were scanned on headache days vs. non-headache days could have important effects on the results. It is also unclear how many of the phase-consistent patients had chronic migraine vs. episodic migraine, which could be another important contributing variable.

2. The responder rates for both the current galcanezumab study and the previously reported erenumab study that is used for comparison are significantly less than those reported in the clinical trials of these therapies. This suggests that the subjects examined in these studies may be different from the general clinical trial population.

3. The authors report that there was no difference in headache severity on the day of the scan pre- and post- treatment, despite reductions in brain activation in different regions, particularly in those who responded to treatment with reduced attack severity. This suggests that the reduction in brain activation is not correlated with attack severity, in contrast to the correlation with attack frequency.

4. The unpleasantness perception in response to trigeminal nociceptive stimulation apparently did not change after treatment despite reductions in the brain responses seen with fMRI. Here again there is a dissociation between the acute clinical response and the changes in brain activity.

5. Regarding the comparison of the erenumab data vs. the galcanezumab data, there is some concern regarding the comparison of these two groups. Although the demographics of the groups are generally similar, it is not clear how many of the phase-consistent scans in the erenumab study were performed on headache days vs. non-headache days, and how many were in chronic vs. episodic migraine patients. These details would be important to know in considering whether the comparison between the two groups is feasible.

6. Since the subjects were primarily women, another potential source of variability is the menstrual cycle. Is there any indication that this could be an issue regarding the interpretation of the data?

7. Regarding the conclusions, the authors should acknowledge the possibility that the antibodies could be acting directly in the hypothalamus, parts of which may be outside of the blood brain barrier. Some brain areas including the hypothalamus have been shown to express CGRP receptors. (Eftekhari S, Gaspar RC, Roberts R, et al., Localization of CGRP receptor components and receptor binding sites in rhesus monkey brainstem: A detailed study using in situ hybridization, immunofluorescence, and autoradiography. The Journal of comparative neurology. 2016;524:90-118).

8. Please indicate the number of subjects compared in the phase-consistent studies of galcanezumab and erenumab who had episodic vs. chronic migraine.

9. For the erenumab study that is compared, please indicate the number of scans in the phase-consistent studies that were performed on headache dys vs. non-headache days.

10. Please comment on the lack of correlation between the clinical measures of headache severity or unpleasantness of the noxious stimulation and the imaging findings on MRI in terms of the reduction in brain activity in response to galcanezumab.

11. A brief discussion of the possibility that galcanezumab (and erenumab) could be exerting effects via brain regions like the hypothalamus that may be outside the blood brain barrier is warranted.

12. p. 17, line 339 -- I think the authors mean "express" instead of "expose".

13. It would be helpful to have a greater description in the figure legends for the results tables or elsewhere regarding what each of the values mean and how they were calculated. For example, I am unclear as to how the T values were calculated.

https://doi.org/10.7554/eLife.77146.sa1

Author response

Reviewer #1 (Recommendations for the authors):

The paper reports a well conducted, carefully analyzed data set with important implications for understanding migraine therapeutics. My comments are modest and advisory.

A. The Introduction states CGRP mABs are impermeable with regard to the blood-brain barrier. An article is cited that, consistent with what is known, shows their permeability is limited not zero (1). The authors could consider recasting the sentence in line with the citation.

Thank you for the suggestion; we have changed the statement accordingly (p. 4 line 5f.).

B. Not all responders to galcanezumab will do so in the first two weeks (2); would this have blunted the responder/non-responder comparison at the two week point? Could it be argued that a longer study time might be considered going forward?

Good point. For this reason we took the subjective assessment of the participants after 3 weeks as well as the headache calendar after a total of 3 months of therapy into account. This interval is also specified in the German guideline on CGRP therapy for migraine (p. 6 line 16f. and p. 10 line 7). Interestingly, the outcome of the 2 time points was highly correlated.

C. The authors cite six papers for the Method; perhaps one or two would be sufficient?

We have reduced the number of citations in the methods section

D. Given the highly personalized nature of pain, it could be argued that VAS scales by virtue of the underlying biological variability between patients, are poor candidates for being treated as more than within subject ordinal measures? It could be argued that the Wilcoxon test is preferable as has been mentioned by the authors.

We totally agree. We changed the wording accordingly. Indeed we used a NRS (anchored at 0 – 100). NRS should be treated as a continuous variable and we therefore used paired t-tests.

E. Given response rates to galcanezumab vary by frequency of migraine days, as evidenced by headline responses in episodic (3, 4) and chronic migraine (5), did migraine day frequency effect outcomes? Tables one might usefully offer the frequency data for both the episodic and chronic migraine groups.

We conducted sub-analyses of the functional imaging data as well as the analyses mentioned in the manuscript with the diagnosis (episodic vs. chronic) as a covariate. No significant influence by migraine diagnosis could be found. The frequencies for episodic vs. chronic are now given in the manuscript (p.11 line 11 ) as well as in table 1.

F. Regarding concomitant preventives, it is a practical limitation of studies that stopping preventives can be problematic. If the patients taking concomitant preventives with established efficacy in migraine are excluded, and not all those listed un characteristics would fulfill that criterion, does this alter the outcome?

Concomitant preventives may well influence the results, especially when there is any change (including dosage) between T0 and T1. For this reason, concomitant preventives are only allowed where the kind and the dosage of any medication have remained unchanged starting 3 months before and throughout the entire study period.

G. It would be generally accepted that a 30% reduction in migraine days in episodic migraine is a modest effect; if a 50% reduction had been used as the binary response outcome for all participants, would the result be different? The authors very sensibly point out the severity change argument in the limitations section. It could be argued that applying the 50% rule might sharpen the differences.

We absolutely agree with the reviewer on this point. We therefore also carried out this analysis with 50% responders although not many patients qualified, and found the same result at a less stringent (explorative) significance level (p. 14 line 15f.) because of lower patient numbers.

References

1. Johnson KW, Morin SM, Wroblewski VJ, Johnson MP. Peripheral and central nervous system distribution of the CGRP neutralizing antibody [(125)I] galcanezumab in male rats. Cephalalgia. 2019;39:1241-8.

2. Goadsby PJ, Dodick DW, Martinez JM, Ferguson MB, Oakes TM, Zhang Q, et al., Onset of Efficacy and Duration of Response of Galcanezumab for the Prevention of Episodic Migraine: A Post-Hoc Analysis. Journal of Neurology, Neurosurgery and Psychiatry. 2019;90:939-44.

3. Skljarevski V, Matharu M, Millen BA, Ossipov MH, Kim BK, Yang JY. Efficacy and safety of galcanezumab for the prevention of episodic migraine: Results of the EVOLVE-2 Phase 3 randomized controlled clinical trial. Cephalalgia. 2018;38:1442-54.

4. Stauffer VL, Dodick DW, Zhang Q, Carter JN, Ailani J, Conley RR. Evaluation of Galcanezumab for the Prevention of Episodic Migraine: The EVOLVE-1 Randomized Clinical Trial. JAMA Neurol. 2018;75:1080-8.

5. Detke HC, Goadsby PJ, Wang S, Friedman DI, Selzler K, Aurora SK. Galcanezumab in chronic migraine. The randomized, double-blind, placebo-controlled REGAIN study. Neurology (Minneap). 2018;91:e2211-e21.

Thank you, all references cited, where appropriate

Reviewer #2 (Recommendations for the authors):

The manuscript is well written and the conclusions are supported by the data. There were some questions and comments identified related to the study.

1. The authors suggest in the discussion that galcanezumab binding to peripheral CGRP could subsequently influence the circulating levels of CGRP and as a result, modulate its ability to cross the blood brain barrier and modulate hypothalamic function. While this is true for galcanezumab, this is unlikely to be true for erenumab since it binds the receptor. Is it possible that these types of differences are responsible for the distinct patterns of effects observed when comparing the two antibodies? The authors discuss other possible explanations in the paper (starting at line 344) but this.

The reviewer is exactly right. We have discussed this issue. Relevant discussion is highlighted in in the text (see page 17).

2. Given that effects of gancanezumab were observed in the hypothalamus, and the hypothalamus has areas not protected by the blood brain barrier, could the observed effects on this region be due to direct access of drug to this CNS region? It might be helpful to discuss the specific regions of the hypothalamus not protected by the BBB and whether these could contribute to the observed effects. This was included in the prior erenumab paper but would be helpful to discuss here.

Thank you for this comment. We have now included this line of thought into the discussion (page 17)

3. The prior erenumab study included 81% females while the current study is 96% female. While not a major difference, do the authors expect this has any influence on the findings of the two studies?

Thank you for this comment. We agree that sex hormones play a great role. The difference between the 2 studies (3 males) is however too small to explain the difference in imaging results. Nevertheless, following your suggestion we also carried out a sub-analysis in this regard with only the female study participants and found the same result.

4. It might be helpful to include more detail on the information collected in the headache calendars. This may help readers assess the level of attention subjects paid to the qualities of their migraines across the 3 months. Since this study uses drug efficacy as a method to segregate patients responses to trigeminal stimulation, and efficacy was entirely subjective on the part of the subjects, it seems helpful to have more detail on what was asked of the patients between study visits.

The classification of the responder group and non-responder group was made after an interval of 3 months with the help of a standardised headache calendar (p. 6 line 16- https://www.dmkg.de/files/Kopfschmerzkalender_PDF/Kopfschmerzkalender_ENGLISCH_18.3.2021.pdf). We took the subjective assessment of the participants after 3 weeks as well as the headache calendar after a total of 3 months of therapy into account. This interval is also specified in the German guideline on CGRP therapy for migraine (p. 6 line 17 and p. 10 line 7). Interestingly, the outcome of the 2 time points were highly correlated. (p. 10 line 7).

The current work is quite similar to the prior study by these same authors using erenumab. It would help increase the impact for eLife if the authors could provide more information (or data) that clarifies what the major advance is in this work over the prior work.

The main advance of the current work compared to our previous study using erenumab lies in the mechanism of how the medication works. Different from erenumab, which targets the CGRP receptor, galcanezumab directly targets the peptide itself. For this reason, we additionally analysed the raw data from both studies (i.e. did not just take analysed data from a historical patient group), and such direct comparisons are therefore in a strong position to point out differences and similarities between the ligand and the receptor antibody on central pain modulation in migraine patients. We additionally and for the first time show that the STN, which plays a central role in the ascending trigemino-nociceptive pathway, co-varies in its activity already at visit 1 with the response, i.e. the efficacy to galcanezumab. These data are new and demonstrate, using functional imaging, a predictor for clinical efficacy of galcanezumab.

Reviewer #3 (Recommendations for the authors):

The imaging studies are of high quality, and the experimental design importantly enables within-subject comparison before and after treatment.

There are, however, some sources of variability that require consideration in the interpretation of the results, and some interesting findings that warrant further discussion.

1. The study includes both episodic and chronic migraine patients who are on a variety of different medications. In order to reduce potential migraine phase-related variability, a subgroup of subjects who had either headache or non headache on scan days both pre and post treatment. As it turned out , 14/15 of the phase-consistent patients were scanned on a headache day pre- and post-treatment with galcanezumab, while only 1 was scanned on non-headache days. The fact that nearly all patients were scanned on headache days vs. non-headache days could have important effects on the results. It is also unclear how many of the phase-consistent patients had chronic migraine vs. episodic migraine, which could be another important contributing variable.

We apologise for the misleading wording in the manuscript. The number of patients with headache on day 1 and day 2 referred to the total cohort. Overall, the ratio of no headache on both days (n=8) and headache on both days (n=7) is balanced, especially as we have evaluated the imaging data intraindividually and longitudinally. We have clarified this in the manuscript (see p. 12 line 4f)

2. The responder rates for both the current galcanezumab study and the previously reported erenumab study that is used for comparison are significantly less than those reported in the clinical trials of these therapies. This suggests that the subjects examined in these studies may be different from the general clinical trial population.

True. The participants fulfilled the official criteria for CGRP therapy according to the German guideline. This guideline requires that all oral prophylactics have already been tried unsuccessfully, and, in case of chronic migraine patients, have also failed botulinum toxin treatment. Thus, the patients included here formally correspond to a patient group that is difficult to treat (p. 13 line 4). We would argue that the principles of mechanism how the respective antibodies exert their effect (in this case centrally) are however the same in treatment naïve and non-responding patients. We agree that this needs to be investigated in future studies to give a valid answer.

3. The authors report that there was no difference in headache severity on the day of the scan pre- and post- treatment, despite reductions in brain activation in different regions, particularly in those who responded to treatment with reduced attack severity. This suggests that the reduction in brain activation is not correlated with attack severity, in contrast to the correlation with attack frequency.

That is an interesting thought. We used the (quantifiable) reduction in the number of monthly headache days to evaluate the efficacy of migraine prophylaxis. The subgroup analysis (same headache state between T0 and T1) that this reviewer mentions was indeed controlled for headache severity, and no difference between headache severity on that special day was found. This allowed making the assumption, that our results are frequency dependant.

We agree that over the whole time span of 3 months the mean severity (and other subjective measures) may well be reduced nevertheless. We cannot more than agree that our findings therefore may be restricted to patients who indeed are frequency responders. Since in our experience many patients report that indeed the severity is reduced more than frequency, one could argue that only in frequency responders the hypothalamus is effectively targeted using mABs and not in severity responders. Unfortunately, we had not enough patients who kept the frequency and only reported a severity decrease, what would have been necessary to directly answer that question. (we have added this in the discussion please see page 19)

4. The unpleasantness perception in response to trigeminal nociceptive stimulation apparently did not change after treatment despite reductions in the brain responses seen with fMRI. Here again there is a dissociation between the acute clinical response and the changes in brain activity.

Our remark regarding unpleasantness perception is explicitly for the response to (experimental) trigeminal nociceptive stimulation. Since CGRP monoclonal antibody are not pain killers, we would not expect a difference in perceiving this experimental stimulus and consequently controlled for this.

5. Regarding the comparison of the erenumab data vs. the galcanezumab data, there is some concern regarding the comparison of these two groups. Although the demographics of the groups are generally similar, it is not clear how many of the phase-consistent scans in the erenumab study were performed on headache days vs. non-headache days, and how many were in chronic vs. episodic migraine patients. These details would be important to know in considering whether the comparison between the two groups is feasible.

6. Since the subjects were primarily women, another potential source of variability is the menstrual cycle. Is there any indication that this could be an issue regarding the interpretation of the data?

We agree that the menstrual cycle could have an impact in general. However, In our previous study, we scanned seven patients daily over 30 days using the same experimental set-up and did not find changes in imaging results in regards to the menstrual cycle. Therefore, we did not control for the menstrual cycle in the current study[1].

1. Schulte, L. H., Mehnert, J. and May, A. Longitudinal Neuroimaging over 30 Days: Temporal Characteristics of Migraine. Annals of Neurology 87, 646–651 (2020).

7. Regarding the conclusions, the authors should acknowledge the possibility that the antibodies could be acting directly in the hypothalamus, parts of which may be outside of the blood brain barrier. Some brain areas including the hypothalamus have been shown to express CGRP receptors. (Eftekhari S, Gaspar RC, Roberts R, et al., Localization of CGRP receptor components and receptor binding sites in rhesus monkey brainstem: A detailed study using in situ hybridization, immunofluorescence, and autoradiography. The Journal of comparative neurology. 2016;524:90-118).

Thank you for this remark, we have now discussed this fact and cited the reference (page 17 line 19ff)

8. Please indicate the number of subjects compared in the phase-consistent studies of galcanezumab and erenumab who had episodic vs. chronic migraine.

Done (see table 1)

9. For the erenumab study that is compared, please indicate the number of scans in the phase-consistent studies that were performed on headache dys vs. non-headache days.

Done (see p. 12 line 5ff)

10. Please comment on the lack of correlation between the clinical measures of headache severity or unpleasantness of the noxious stimulation and the imaging findings on MRI in terms of the reduction in brain activity in response to galcanezumab.

11. A brief discussion of the possibility that galcanezumab (and erenumab) could be exerting effects via brain regions like the hypothalamus that may be outside the blood brain barrier is warranted.

Thank you for this remark, we have now discussed this fact and cited the reference (page 17 line 19ff)

12. p. 17, line 339 -- I think the authors mean "express" instead of "expose".

We corrected this.

13. It would be helpful to have a greater description in the figure legends for the results tables or elsewhere regarding what each of the values mean and how they were calculated. For example, I am unclear as to how the T values were calculated.

We explicitly state an extended description of the individual statistical results by exemplifying the exact comparison, the number of participants (n), the degrees of freedom, the p-values and t-values for all t-tests. We calculated T-test with the standard t-test family implemented in the SPM toolbox (Wellcome Trust Center for Neuroimaging, London, UK) SPM12 – Statistical Parametric Mapping. https://www.fil.ion.ucl.ac.uk/spm/software/spm12/ (accessed 4 Jan 2019).

We now also include a short description of the used test in the main text table, as well as in the Table land Figure legends.

https://doi.org/10.7554/eLife.77146.sa2

Article and author information

Author details

Hauke Basedau

Department of Systems Neuroscience, University Medical Center Hamburg-Eppendorf, Hamburg, Germany

Contribution
Formal analysis, Investigation, Writing – original draft

Competing interests
HB received consulting fees from Novartis and Teva. The author has no other competing interests to declare

"This ORCID iD identifies the author of this article:" 0000-0002-4605-0172
Lisa-Marie Sturm

Department of Systems Neuroscience, University Medical Center Hamburg-Eppendorf, Hamburg, Germany

Contribution
Investigation, Writing - review and editing

Competing interests
No competing interests declared
Jan Mehnert

Department of Systems Neuroscience, University Medical Center Hamburg-Eppendorf, Hamburg, Germany

Contribution
Formal analysis, Writing - review and editing

Competing interests
No competing interests declared

"This ORCID iD identifies the author of this article:" 0000-0001-7088-740X
Kuan-Po Peng

Department of Systems Neuroscience, University Medical Center Hamburg-Eppendorf, Hamburg, Germany

Contribution
Investigation, Writing - review and editing

Competing interests
No competing interests declared
Marlene Schellong

Department of Systems Neuroscience, University Medical Center Hamburg-Eppendorf, Hamburg, Germany

Contribution
Investigation, Writing - review and editing

Competing interests
No competing interests declared

"This ORCID iD identifies the author of this article:" 0000-0002-6499-2024
Arne May

Department of Systems Neuroscience, University Medical Center Hamburg-Eppendorf, Hamburg, Germany

Contribution
Conceptualization, Funding acquisition, Supervision, Validation, Writing – original draft, Writing - review and editing

For correspondence
a.may@uke.de

Competing interests
No competing interests declared

"This ORCID iD identifies the author of this article:" 0000-0002-3499-1506

Funding

Bundesministerium für Bildung und Forschung (BIOMIGA (01EW2002 to AM))

Arne May

Deutsche Forschungsgemeinschaft (SFB936- 178316478 - A5)

Arne May

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

Ethics

Human subjects: The study was approved by the local ethics committee in Hamburg, Germany (PV 5964) and was conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained before initiation of the first study session.

Senior Editor

Jeannie Chin, Baylor College of Medicine, United States

Reviewing Editor

Allan Basbaum, University of California San Francisco, United States

Reviewer

Andrew Charles, University of California, Los Angeles, United States

Publication history

Received: January 17, 2022
Accepted: April 30, 2022
Version of Record published: May 23, 2022 (version 1)

Copyright

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.