This post is adapted from this X thread by @ZsoltLenkei about the main results of my PhD thesis published as a preprint on BioRxiv server.
I will explain briefly in the following paragraphes the main points of Mariani et al. 2024. This study was designed to ellicit the fingerprint of opioids on Functional Connectivity (FC). We leveraged the possibility of Functional Ultrasound Imaging (fUSI) to image transcranially awake behaving mice and investigate for the first time the influence of opioids on FC in absence of anaesthesic bias. Using mice we could demonstrate the causal involvement of the \(\mu\)-Opioid Receptor (MOP) in the observed FC dysconnectivity pattern.
Fig1. Top: imaging setup for awake transcranial fUSI on behaving mice in the Neurotar system head fixed system. Middle imaging slice of interest (left: Ultrasound Localisation Microscopy, right: overlay of Power Doppler with the Allen template). Bottom: experimental design, the drug is injected after a \(20min\) of baseline, we compute differences of FC with matching saline injections, significant difference in seed-based maps show the influence of morphine on the FC.
We use fUSI to image, through the skull, coronal brain slices of head fixed and behaving mice. Using the Neurotar system, the animal is the clamped to reduce artefacts and facilitate precise positioning of the probe. The animal can explore the Mobile Home Cage freely as the cage is floating on some air cushion. After \(20min\) of baseline we inject either drugs (opioids: Morphine, Fentanyl, Methadone, Buprenorphine or the MOP antagonist Naloxone) or saline for controls. The animal is followed up for one hour to observe the effect of the drug on its brain haemodynamics. We focused on FC shift illustrated here (Fig1.) as Seed Based Maps (SBM) with voxels whose correlation with the seed (in red) is significantly altered.
Fig2. Top: Correlation matrix, low triangle average correlation after the injection of morphine, top triangle displays the difference only for the edges with a significant change as compared to baseline. Middle: significant edges displayed as a network. Bottom: seed based maps showing correlation voxel wise with 3 different regions of interest.
Looking at static FC (zero lag correlation), we observe after injection of morphine the apparition of two main anticorrelated super-synchronous networks: thalamo-cortical vs. thalamo-hippocampal. This shift highlighted in the correlation matrix (Fig2.) is driven by the hippocampus which reduces cortical synchronization while enhancing correlation with the thalamus, the hippocampus and the hypothalamus. These network are better illustrated comparing the SBMs after saline injection or morphine.
Fig3. Evolution of the connectome under different conditions, from left to right network computed over 10min windows for two phases of baseline and 6 phases post injection. Top: increasing doses of morphine show a time and dose dependent effect. Middle: the effect is absent in MOP-KO mice, two weeks induction of tolerance by continuous perfusion of morphine reduces the magnitude of change and accelrate its onset. Bottom: similar dysconnectivity profile are observed with different opiates in a dose dependence manner.
Morphine induced FC change is time- and dose-dependent, i.e. higher the dose, stronger and faster the effect. The pattern is absent in MOP KO mice, we show in the article inhibition of this pattern after pre-treatment with Naloxone. We also invesigated tolerance, by implantation of an osmotic micropump which infused the mice with opiates (either morphine or fentanyl) for multiple weeks. Tolerance induction reduces the effect intensity but also accelerates the FC shift. Finally we observed the same pattern in other opiates such as Fentanyl and Methadone.
Fig4. Top: Power spectrum shows a redistribution of power towards higher half of FC band with a relatively fast oscillations peak. Middle: looking at cross corelation we observe the spatio-temporal evolution of correlation structure by introducing delays with the seed signal in SBA. Bottom, looking at ROIs pair cross-correlation, the fast oscillatory phenomenon seems to drive dynamic reconfiguration of FC with different effects depending on baseline profile.
To understand the apparition of thalamo-cortical anticorrelation we looked at dynamical properties of the brain signals. In addition to a changes of correlation pattern, power spectra show that morphine induces an oscillatory phenomenon in the higher half of the FC bandwidth. This oscillation is apparent in voxel-wise cross-correlation maps.
Fig5. Cross correlation structure with a seed in the ventral Thalamus. Left: after saline injection, Middle: after morphine injection, Right: difference in voxels whose changes after morphine injection are significant as compared to saline.
To illustrate this spatio-temporal reorganisation, we introduced a delay in the seed signal before computing the correlation (Fig4. and Fig5.). This way we can reconstruct the correlation structure with regions “in the past” or “in the future” of the seed signal. The anticorrelation appears as a phase shift in slow oscillations. The peak in the spectrum emerges as the main driver of changes in the dynamics of this correlation structure.
Fig6. Top: influence of various opiates on awake mice connectome. Bottom: analgesia profile (time and dose dependence) for the same drugs measured with the hot plate test.
We observed major similarities in the FC patterns induced by various opiates (Morphine, Fentanyl, Methadone, Buprenorphine). This pattern, that we call the opioid fingerprint, matches the time- and dose- dependence of analgesia induced by each drug, as measured with the hot plate test. The relationship matches increasing doses of the same compounds, but also explains the weak pattern of Buprenorphine, the least potent analgesic of the four.
Fig7. Top: superposition of regional CBV, FC index, mobility, MOP phosphorylation and analgesia evolution with time after injection of 4 different doses of morphine. Bottom: cross correlation shows two different dynamics, while hyperlocomotion correlates with perfusion, FC correlates with analgesia and MOP phosphorylation.
We compared the evolution of FC, Cerebral Blood Volume (CBV), analgesia, MOP activation and mobility after morphine injection. This analysis suggests a sequential action of the drug. First, a peak of mobility co-occurs alongside with a CBV burst. It is followed by a slower shift of FC that correlates with analgesia and MOR phosphorylation. The correlation highlights differences in the two clusters timings. We further validated this observation with Principal Component Analysis PCA which also identified these two dynamics.
Fig8. Summary illustration.
To summarise this study. We investigated the influence of opioid injection on brain dynamics as measured with fUSI. We identified:
- The emergence of 2 anticorrelated super-synchronous networks (cortex VS hippocampo-thalamo-hypothalamic)
- The presence of a CBV and hypermobility burst which precedes a FC shift
- The FC pattern change conversely correlates with MOP activation and analgesic dynamics
- FC pattern change could be linked to a dynamic phase and frequency shift in correlation structure
Beyond demonstrating the ability of fUSI to investigate the action of important drugs on brain function, we expect that this work may also help to bridge the explanatory gap between neuron-level mechanisms and large-scale neural dynamics that underlie behavior. This study, by showing for the first time in awake and behaving animals the effect of opioids on FC, reaches a milestone in neuropharmacology at the system level, without interactions from potential anaesthetic agents. The recent advances in fUSI technology Bertolo et al. 2023, will leverage our optimised imaging protocol for whole brain fingerprinting. We expect this platform to significantly improve high throughput drug screening and the development of new therapeutic strategies.
I am grateful to Zsolt Lenkei for supervising me during this project, Andrea Kliewer for leading this project to its end and acquiring all these data, Samuel Diebolt for his critical role in analysis implementation. I would also like to thank our main collaborators: Mickael Tanter and Thomas Deffieux for their invaluable expertise and support on fUSI, Stefan Schultz on the pharamacology side, Renata Santos for supporting us all times and Laurianne Beynac who helped with the machine and acquisitions.