Ultrasound localization microscopy and functional ultrasound reveal atypical features of the live trigeminal ganglion

The functional imaging of the neurovascular coupling within the trigeminal ganglion (TG) is highly challenging due to its small size and its deep localization. This study combined a methodological framework able to dive into the rat trigeminal nociceptive system by jointly providing first imaging of the trigeminal ganglion blood vasculature at microscopic resolution and the measurement of its neurovascular coupling in the rat TG evoked by corneal stimulations, a robust and clinically-relevant model. Using functional ultrasound imaging (fUS), we were able to image and quantify a strong hemodynamic response in the ipsilateral TG from anesthetized rats, evoked by mechanical or chemical stimulations of corneal nociceptive fibers to intact cornea, even though TG involves less than 300 sensory neurons. The in vivo quantitative imaging of the TG’s vasculature using ultrasound localization microscopy (ULM) combined with ex-vivo (DiI) staining reveals particular features of the vascularization of the area containing the sensory neurons, that is likely the origin of this strong vaso-trigeminal response and due to the nature of this structure at the interface between the peripheral and central nervous systems. This innovative imaging approach opens the path for future studies on the mechanisms underlying changes in trigeminal local blood flow and neurovascular coupling, key mechanisms and readouts for the understanding and treatment of debilitating trigeminal pain conditions.


INTRODUCTION
The trigeminal ganglion (TG) contains the cell body of the primary sensory neurons from the ophthalmic (V1), the maxillary (V2), and the mandibular (V3) nerves. These sensory neurons are highly specialized, as they detect and respond to a variety of chemical, mechanical and thermal stimuli applied on these branches.
Because the TG is relatively small and localized in Merkel's trigeminal cave in both human and rodents, only a limited number of studies were able to perform functional neuroimaging studies. While Bererra's and Borsook's teams published seminal works on the existence of a neurovascular coupling in the human TG (1)(2)(3)(4), only one preclinical contrast MRI study imaged macrophage infiltration in the mouse TG using ultrasmall superparamagnetic iron oxide nanoparticle contrast in a model of alkani burn cornea (5). But, to the best of our knowledge, dynamic functional imaging of the TG was never performed in rodents due to the difficulties to access the TG. Preclinical studies investigating the physiological activity within the TG of rodents are classically based on electrophysiological recordings of single and/or clusters of neurons (6,7), as well as immunohistochemical staining using indirect markers of neuronal activation (see for review (8)). Despite the cellular resolution of these surrogates, this approach lacks the ability to follow the dynamics of these neuronal changes. Recently, the visualization of trigeminal sensory neuron activities in response to orofacial stimuli was reported ex-vivo using either voltage sensitive dye approach in decerebrated animals (9) or calcium imaging in GCaMP6 mouse line (10,11). However, these experimental paradigms used were highly invasive requiring the decerebration of the animal, and therefore the disconnection between TG and the CNS.
Functional ultrasound imaging (fUS) imaging is a relatively new versatile neuroimaging modality that allows imaging and measurement of cerebral blood volume in both human (12,13) , non-human primates (14) and rodents (15)(16)(17)(18)(19) with excellent spatial (100 to 300 µm) and temporal resolutions (down to 20 ms). One of its biggest characteristics is its high sensitivity compared to fMRI (20). Indeed, during a task, the locally increased neuronal activity due to the neurovascular coupling leads to a hemodynamic response (21). In the past, fUS imaging showed to be sensitive enough to measure the cortical hemodynamic changes induced by sensory (18), olfactory (22), visual (23) stimuli in anesthetized animals, as well as auditory (24), motor tasks (14)(15)(16) in awake animals.
Interestingly functional ultrasound imaging can be coupled on the same device with another emerging modality, Ultrasound Localization Microscopy (ULM), able to provide the brain vascular anatomy and hemodynamic quantification up to microscopic resolution both in rodents (25) and humans (26). The corneal trigeminal system is particularly interesting as the cornea is the most densely innervated tissue in the body (8) whose nerve terminals are directly accessible for stimulation. Moreover, the cornea is exclusively innervated by unmyelinated C-and thinly myelinated A-delta fibers, including mechano-nociceptors that are triggered by noxious mechanical stimulation, polymodal nociceptors that are excited by mechanical, chemical, and thermal stimuli, and cold thermoreceptors that are activated by cooling (8,27,28).
Taking advantage of the high sensitivity of fUS imaging and the corneal nociceptive model, this study has several main objectives: first to image the TG in rat , second measure the velocity of blood flow in the TG using ULM and third to detect and measure functional activations in the ophthalmic division following corneal stimulations. We provide the first proof of concept of imaging the rat's TG, with a detection of local blood flow at a microscopic scale, and of the measurement of the hemodynamic responses evoked by the activation of corneal nociceptors in anesthetized animals. Our results bring forward an innovative approach to study the TG's neurovascular coupling, a key mechanism for deciphering the mechanisms of trigeminal sensitization and concomitant pain associated with trigeminal pathologies.

Localization / imaging of the rat trigeminal ganglia using ultrafast Doppler imaging
Taking into account that TG is a deep structure, we imaged much deeper under the brain as compared to previous studies in anaesthetized rodents (18,19). Despite the signal attenuation due to the depth of imaging (15 mm), two-dimension scans (2D) revealed two bilateral longitudinal structures detected between the antero-posterior coordinates Bregma -3.3 and Bregma -5.0 mm ( Figure 1A, B, F). In addition to two-dimension scans ( Figure 1A), three-dimension scans (3D, Figure 1B, Supplementary video 1) precisely localized the structure respectively to the location of the brain's vasculature.

Comparison
with the tractography atlas of the Waxholm atlas (https://scalablebrainatlas.incf.org/rat/PLCJB14, (29,30)) confirmed that these bilateral structures are located under the brain at the same anteroposterior coordinates and laterality than the TG they described using tractography of peripheral fibers ( Figure 1C, Considering that the doppler signal of the TG was visible at this depth despite the signal attenuation, suggests that TG is a richly vascularized structure. This was confirmed by comparison between the doppler signal in the TG and in the internal carotid ( Figure 1A).
We found that the blood volume measured in the TG (without correction for the attenuation) represents 44.0 % ± 1.4 % of that of the internal carotid.
Next, ex vivo staining of the TG vessels revealed a high density of DiI stained vessels which exhibited a tortuous morphology. Altogether, the ex vivo staining and in vivo fUS imaging data, support the idea that TG receive an important blood supply.

Corneal nociceptor stimulations induce functional hyperhaemia in the ipsilateral trigeminal ganglion
Our anatomical study confirmed a high density of blood vessels stained in vivo using DiI Finally, the activation of TRPV1 polymodal corneal nociceptors using the TRPV1 ligand Capsaicin (10 µM) induced a phasic and robust increased blood volume (18.9 % increase, p = 9.3 10 -3 ) in the ipsilateral TG as compared to the contralateral TG ( Figure 2B-F). The BV increase lasted during all the time of exposure to the drug (1 min), but was still observed afterwards ( Figure 2E).

Ultrasound Localization Microscopy (ULM) imaging reveals vascular features of the trigeminal blood flow
To assess novel information regarding the vascular characteristics of the blood supply within the TG (direction of the flow and its velocity) , we used ULM implemented on the same ultrafast ultrasound scanner. This method allowed the measurement of flow of microbubbles inside blood vessels, at a microscopic scale in anesthetized animals (25,32).
We were able to provide a thin definition of blood vessel and the measure and direction of the blood flow in 2D. Our data confirmed i) that the TG is highly vascularized, with a dense network of tortuous blood vessels. ii) The organization of vessels does not look like any vascularized cerebral structures in the sagittal plane imaged ( Figure

Imaging the hemodynamic responses in the ipsilateral TG induced by corneal stimulations: a valuable model to study the trigeminal neurovascular coupling.
This study aimed at providing a proof of concept for the hemodynamic responses in the TG following corneal nociceptive activation. We postulated that corneal stimulations and imaging of the TG in anesthetized animals constitutes a highly interesting model because corneal nociceptors (C and A delta corneal fibers) can be easily activated by mechanical stimulations (35) and by chemical stimulation of the TRPV1 receptor by capsaicin (28).
The ability to detect clear hemodynamic changes in response to corneal stimulations was a particular challenge in this study due to the small number of neurons innervating the cornea.
Despite these challenges, our study demonstrates significant real-time (400 ms) imaging of hemodynamic responses in TG to repeated sequences of C and A delta fibers mechanical activations. This ipsilateral activation, located in the V1 branch of the TG, is consistent with the known anatomy of the TG in rats (36) and somatotopy of corneal afferents within the TG (31). We went further in our investigations by evaluating the hemodynamic changes after activation of polymodal nociceptors by capsaicin, a TRPV1 agonist. We observed a strong phasic hemodynamic response in the ophthalmic division; the phasic aspect of this response is thought to be related both to difficulty to perform block design experiments, and also to the potent action of capsaicin known to strongly activate and sensitize the sensory afferents (1, 7).

The atypical vascularization of the TG
Until recently (37), the vascularization of the TG has been mostly studied and described in human subjects (38). Here we precisely localized the TG in rats using 2D or 3D fUS imaging and noted a high signal of blood volume, demonstrating that TG is richly vascularized. In toto staining of the vasculature (DiI experiments) and ULM in vivo, confirmed that rat TG is richly vascularized, with highly tortuous vessels, especially in the . Interestingly, a similar striking regional difference in the expression of tight junction proteins and the presence of functional blood-brain -barrier (BBB) in the NFRA of the dorsal root ganglion (DRG), but not the CBRA of the DRG has been reported (40). Indeed, blood vessels that vascularize the CBRA have large fenestrations, when compared to peripheral nerves, see for review (39,41). Some studies exploring the permeability of the BBB showed that injected macromolecules are kept in the blood vessels in the NFRA, while they leak out from the blood vessels in the CBRA (42). Therefore, there is an important dichotomy in the nature of the vascularization in the TG (and DRG) ganglia: while the NFRA has a minor vascularization and is well protected by the BBB, the CBRA on the other hand, has a dense vascularization, with a large lack of BBB. In addition to the high metabolic demand at the level of the cell bodies, these elements are likely the fundamental elements underlying the strong hemodynamic response to orofacial stimulations observed in rats in our study and human subjects.

Mechanisms of neurovascular coupling in the trigeminal ganglion
Much is still to be understood in the mechanism underlying the neurovascular coupling in the TG and its particular vascularization in health and disease. Trigeminal nociceptive nerve fibers are known to express vasoactive neuropeptides, including CGRP and substance P (43), that are released upon stimulation causing vasodilation that results in perivascular changes, CGRP is considered as the strongest vasodilating neuropeptide in human and also participate in sensitization of the trigemino-vascular system, peripheral sensitization and hyperalgesia (44). CGRP is expressed by 50% of the corneal neurons, 50% of these them also co-express TRPV1 (45). Therefore, the stimulation of corneal nociceptors both through mechanical stimulation and through TRPV1 activation by capsaicin application in our experimental design induced the release of CGRP. This release was both local in the TG, through a paracrine mechanism, as previously demonstrated (46) and central (in the trigeminal brainstem sensory complex, (44)).
Once released, CGRP increases local tone, thus further amplifying CGRP-mediated activation of cell population bearing CGRP receptor such as endothelial, glia and mast cells. Interestingly, a preclinical study showed previously, that while retrogradely traced neurons are localized in clusters in the TG (each branch V1, V2 and V3 being separated), capsaicin induces a spread of the tracer to all branch of the TG in hours following injection in the temporomandibular joint. The tracer was not restricted anymore to neuronal soma, but was detected in satellite cells (46). This phenomenon of cross-excitation within the entire TG induced by capsaicin is thought to be due paracrine neuro-glial communication that leads to communication via gap junctions (46). We propose that these mechanisms, hypothesized previously to be involved in orofacial sensitization and hypersensitivity (46) may also be involved in the increased neurovascular coupling imaged in the TG in our study and in previous clinical studies (4). In

Supplementary video 1: Video showing 3-dimension tomographic scan performed in a rat
(after removal of the skull to avoid attenuation). It shows the entire vasculature of the rat brain and TGs (below).

Supplementary video 2: Video showing, in real time in a typical experiment (using Iconeus
One imager and the software 'Neuroshop'), the increased blood volume in the trigeminal ganglion during the application of repeated corneal mechanical stimulations.

Acknowledgments
The authors wish to thank the CNRS, Inserm and ESPCI for their financial support. and and mounted on Superfrost slides (Thermofisher scientific, Waltham, Massachusetts, USA). Note: these animals were also used in a previous study on the vasculature of the spinal cord architecture (48).

CGRP immunohistochemistry in the trigeminal ganglion
After three washes in 0.1 M PBS, TG sections were incubated for 1 h in a blocking solution of 0.1 M PBS containing 3% normal donkey serum and 0.1% triton X-100, followed by incubation with primary antibody at 4°C for 24 h. The primary antibody used in this study was mouse anti-CGRP (Sigma-Aldrich: Lot #083M4785, 1:250). CGRP was revealed using Alexa Fluor 594-conjugated donkey anti-rabbit antibody (1:500; Invitrogen). Finally, the sections were cover slipped.

Microscopic analysis
Tissue sections were examined using a Zeiss M1 epifluorescence microscope (Axio ImagerM1; Carl Zeiss) equipped with a digital camera (C11440-42U30; Hamamatsu Photonics) and an image acquisition software (Zen; Carl Zeiss).

Surgical procedures and preparation for imaging
Under deep anesthesia (intraperitoneal (IP) bolus of Medetomidine (Domitor, 0,4 mg.kg --1 ) and ketamine (Imalgène, 40 mg.kg -1 )), the animal was placed on a stereotaxic frame and a craniotomy (removal of the skull) was performed as previously described (19) between Bregma and Lambda. Some ELMA cream (AstraZeneca, UK) was placed in the ear bars in order to prevent discharge from nociceptors at the level of the ears. This window allowed the scanning of a large part of the brain, under which the TG is located. During the surgical procedure and the imaging session, the animals' body temperature was kept at 37°C using a heating blanket and an intrarectal probe (Physitemp, USA), and the heart and respiratory frequencies were monitored continuously (MouseOxPlus, Ugo Basile, Italy). As previously described (19,48), 45 min after induction (when the craniotomy was finished), the anesthesia was maintained but reduced, using subcutaneous perfusion of Medetomidine (0.1 mg/kg/h) and ketamine (12.5 mg/kg/h) using a syringe pump. As previously observed (48), in order to obtain reproducible results, it was preferable to wait in order to reach stable physiological parameters and a reproducible level of anesthesia (respiratory frequency around 80-90 rpm). Each imaging session lasted from 3 to 4 hours.
Two milliliters of saline solution were gently dropped on the brain (the dura mater was kept intact), followed by echographic gel (Dexco Médical, France). The ultrasonic probe was then positioned just above the window using a 4-axis motorized system on which the ultrasound probe was fixed (19).

2D and 3D imaging scans of the brain and TG
In

Doppler Signal Analysis and Activation maps
Doppler data were analyzed using a generalized linear model approach (GLM) implemented in Matlab in order to obtain the Z-score and p-value maps (48). The activation maps show the Z-score of all significant pixels in the image (p-value < 0.05 (before Bonferroni correction).
We drew the ipsilateral region of interest (ROI) around the activation area thanks to the thresholded Z-score map and the contralateral ROI was drawn by symmetry. The two signals were averaged along the 2 spatial dimensions in order to obtain a single temporal signal. The signal was then expressed as a BV (Blood Volume) variation (in percent), or ∆BV, by subtracting the BV baseline (calculated for each acquisition by averaging all the temporal data within the TG where the stimulation pattern was strictly equal to 0) and by dividing by the BV baseline. Mean ∆BV values over time were computed at baseline, i.e.
during the periods without chemical or mechanical stimulation of the cornea, and during stimulation, i.e. during the periods with stimulation, and are denoted by ∆BVBL and ∆BVSTIM.

Statistical analysis of the evoked trigeminal hemodynamic responses
The statistical analysis was performed using Matlab Version 9.7.0.1261785 (R2019b).
The data were modelled using linear mixed models (LMM), which are suited to the case of non-independent, hierarchical data (∆BV values from several acquisitions, in several animals), and of factors of interest having fixed effects (ipsilateral vs contralateral TG, baseline vs stimulation) as well as random effect factors (acquisition, animal).
In the case of the corneal stimulation, our aim was to establish the significance of the

Ultrasound localization microscopy (ULM)
Three animals were used for this procedure. A catheter filled with saline was inserted in the rat jugular vein before the positioning of the animal on the stereotaxic frame. ULM was performed similarly to the methods described in (26), but using continuous injections of Sonovue (Braco, Italy) reconstructed in 5mL of saline at the rate of 3.5mL/h.
A total of 750 blocks composed of 400 compounded frames at a 1000Hz framerate (with angles at -5°, -2°, 0°, +2°, +5°, PRF=5000Hz, 12mm imaging depth) were acquired using the same system as above. The total acquisition lasted 300s. Beamformed data were filtered using the SVD spatio-temporal filter described in (67) and the N=10 first singular values were removed to extract microbubbles signals from the surrounding tissues.
Microbubbles were detected as the brightest local maxima in the images. Tracking of the maxima positions was performed using a classical particle tracking algorithm

Data availability statement:
Data supporting the findings associated with the figures of this study will be available after publication on a repository website. Figure 1 Anatomical localization and vascularization of the rat trigeminal ganglia (TG). A, F: fUS imaging through the whole brain depth reveals the vascularization and localization of the rat TG in a coronal plane (A or sagittal plane: E). B: Capture of the 3D tomographic scan detailed in Supplementary video 1, illustrating