Functional MRI detection of blood-oxygenation-level-dependent responses in both white and gray matter of the spinal cord after optogenetic cortical stimulation

Abstract

We report the first demonstrations of neural activity detected using functional MRI within both the gray and white matters of the cervical spinal cord of nonhuman primates in response to stimulation by optogenetic excitation of brain cortex. The right secondary somatosensory cortices (S2) of two squirrel monkeys were transfected with Adeno-associated virus that introduced a light-sensitive cation channel (ChR2) into neurons. An optical fiber was inserted to selectively activate excitatory neurons in S2 targets using a blue laser. MRIs were acquired of the brain and cervical spine at 9.4 Tesla before, during, and after optical excitation. Robust blood-oxygenation-level-dependent (BOLD) signal changes (P < 0.05 post-FDR correction) were detected at the probe location in S2 region and in multiple connected brain regions as expected. At the same level of significance, we also detected robust and focal BOLD signal changes in spinal gray and white matters in response to the light stimulation of S2 cortex. For example, the gray matter of the ipsilateral ventral horn and left and right dorsal horns exhibited a classical BOLD signal hemodynamic response, with peak signal amplitudes of ∼0.5–0.9%. Several white matter tracts also demonstrated robust responses. For example, the right and left spinocerebellar, left corticospinal, and dorsal column tracts showed transient increases in BOLD signal that were comparable to those seen in gray matter, and with similar time courses. These effects may represent both antidromic and orthodromic excitation produced within the neurons that normally carry sensory information to S2.

Introduction

We report the first functional MRI (fMRI) detections of activations in the spinal cord (SC) produced by descending signals from the brain, the consequences of optogenetic excitation of somatosensory cortex. Conventional fMRI relies on the detection of blood-oxygenation-level-dependent (BOLD) signals evoked by changes in neural activity in response to an excitatory stimulation, and much has been learned over the past 30+ years about the functional organization of the brain by analyzing the effects of a variety of stimuli and tasks. However, there has been much less attention paid to using fMRI to assess neural networks within the SC, partly because of the greater technical challenges of imaging the spine. Nonetheless, several reports have successfully demonstrated neural activity in response to, e.g. tactile sensory or noxious heat stimuli within the central gray matter (GM) of the SC, and different tasks have been used to identify activation within the different horns (dorsal and ventral, left and right) in human subjects and animals (1). Moreover, distinct hubs of synchronized MRI signal fluctuations have been identified in the GM of the cord at rest (2), and correlations between these regions have led to the identification of intraspinal networks that are altered following injury to the SC (2). A few studies have also reported efforts to record fMRI responses to a stimulus within both the cord and the brain simultaneously, with the goal of identifying how these different parts of the nervous system interact in tasks in which both are engaged, but such studies require specialized radiofrequency coils and acquisitions to capture activity in both regions of interest (ROI). Here, we report an alternative strategy for measuring evoked brain and cord responses together using optogenetic excitation of cortical neurons which then induce BOLD effects in the SC after signals propagate “top down.” We injected the adeno-associated virus (AAV) Type 9 into the S2 region of two squirrel monkey brains to introduce an optically sensitive cation channel (3), and inserted an optical fiber to selectively activate excitatory neurons. We acquired fMRI images of the brain and cervical SC at 9.4 Tesla using gradient echo pulse sequences both during and after a series of optical stimuli. We successfully detected evoked BOLD signals in both central GM regions of the SC as well as selected white matter (WM) tracts engaged in information transfer between brain and spine.

Results

Robust BOLD signal changes (P < 0.05 after FDR correction) were detected at the probe location in left S2 cortex and in connected brain regions in response to optical stimuli (Fig. 1).

Fig. 1.

Experimental setup and fMRI BOLD activation evoked by light stimulation of S2 cortex. A) Schematic illustration of the experimental setup. Optical fibers were implanted on both sides of the S2 cortex, but only the left side was transfected with AAV. B) Stimulus cycle—averaged BOLD time courses (mean ± SD) from S2 ROIt during the blue light stimulation of 6 s duration at three intensities (1, 2, and 4 mW). n = 2 animals and a total of 34 runs. C) BOLD activation maps displayed on coronal images placed around the implanted optical fibers (shown as signal voids). Inserts show the zoomed-in view of the activation in S2 near the tip of the fiber and the ROI for time courses in (B).

The main areas activated are described in Table S3. Figure 1C shows two activation maps on two contiguous coronal images. The time courses of BOLD signals in S2 showed a classical time-locked hemodynamic response function after the start of 6-s optical stimulation (Fig. 1B). As a control, no reliable BOLD activation was detected in S2 on the contralateral right hemisphere, which is AAV naive, by the same blue light stimulation.

Of more novel significance, we also detected robust and focal BOLD signal changes at the same level of significance in SC GM and WM in response to the light stimulation of S2 cortex. Figure 2 shows the locations and time courses of the epoch-averaged % signal changes extracted from GM (left dorsal and ventral horns) which exhibited a classical BOLD signal hemodynamic response to 6-s laser stimulation.

Fig. 2.

fMRI BOLD signal detected in cervical SC in response to S2 light stimulation. A) Group level (n = 2 animals and n = 34 fMRI runs) light stimulation evoked activation maps (thresholded at t > 1.2, P < 0.05, FDR corrected) showing the gray-matter horns on slice 1–5. B) Group level activation in WM tracts. C) Schematic set up and plots of stimulus cycle averaged percentage BOLD signal change (red lines, mean ± SE) and solid two-gamma fitted curve shown on two representative regions within the GM and WM (ROI: CI-II: GM Slice2-LD and LV, WM ROI: CIII-IV; WM Slice1 CS-L: WM-Slice4-DC). Blue curve on each panel refers to the mean signal variation of voxels unresponsive to stimulus in the same ROI. Tables S1 and S2 report the % of each ROI significantly activated.

The peak signal amplitudes varied from ∼0.5–0.9%. Time courses extracted from some WM ROIs also demonstrated a robust response of similar form (Fig. 2). In particular, the left corticospinal and left and right and spinocerebellar tracts and dorsal column showed transient increases in BOLD signal that were comparable (∼0.5–0.8%) to those seen in GM and had similar time courses. Note that we also interrogated the signal changes in ROIs defined from a previous study in which 8 WM and 7 GM regions were identified using independent component analysis as showing synchronous BOLD signal fluctuations in a resting state. Not all voxels within each of those ROIs showed a strong response. The proportions of each ROI showing robust changes are summarized in Tables S1 and S2.

Discussion

The SC serves as the first processing and relay station of ascending peripheral inputs to brain as well as descending outputs. The functional integrity of the SC may be disrupted in several clinical disorders and after traumatic injury, so understanding the functional organization of normal SC is of high clinical significance. Current concepts of SC function posit that neural signals from peripheral stimuli are processed by the dorsal horn of GM and transmitted to the brain via specific WM tracts, while information from the cortex is transmitted down the WM of the SC and produce ventral horn GM responses. In previous work (2, 4), we identified the presence of synchronized hubs of resting state BOLD signals in a set of GM regions in each section of the SC, which together form intraspinal networks, as well as in a set of regions that co-locate with major WM tracts. These previous observations underscore the complexity of SC functional organization and indicate that the interactions among SC ascending and descending pathways are more extensive than previously recognized. Here, we show some of those same GM regions and WM tracts are selectively affected by top-down stimulation of the secondary somatosensory cortex (S2).

The S2 cortex lies adjacent to the primary area 3b and both are involved in processing somatosensory stimuli. By examination of connectivity and neural circuits, it has been proposed that S2 is a higher-order cortical area, processing distinct features of somatosensory stimuli, but more recent work (5) has also suggested both S1 (e.g. area 3b) and S2 may operate in parallel. Studies of the brain connections to S2 have revealed widespread engagement with other brain areas including the primary motor area (M1) and thalamus. Moreover, several studies have shown direct corticospinal connections from somatosensory cortices (S1 and S2) to the SCs of mice, rats, and monkeys (6–8) while direct cortical modulation of normal and pathological tactile sensory processing in the SC has also been reported (9). In line with these reports, laser stimulation of the left S2 cortex elicited brain-wide activation in cortical and subcortical regions (Fig. 1), including area 3b/1 of S1, M1 cortex, and MCC cortex. Some of these regions are known to have extensive anatomical connections to S2. For example, Onishi (9) showed that passive movements without motor commands activate not only the primary somatosensory cortex, but also M1 and bilateral S2 areas. The relay of effects to WM tracts accords with previous concepts of the roles of specific tracts. For example, Dum et al (10). found that the spinothalamic system reaches multiple cortical areas in the contralateral hemisphere including S2. Suter and Shepherd (11) showed interareal projections from S2 to M1 and M1 to S2 in mice excited pyramidal neurons across multiple layers of cortex, and that corticospinal projections from S2 and M1 converged in the cervical SC. The spinocerebellar tract is also part of the somatosensory nervous system that relays unconscious proprioceptive information from the body to the cerebellum and combines sensory and motor information. In line with the activation of ascending and descending WM tracts, both ventral and dorsal horns were activated in response to S2 stimulation, supporting the functional relevance of the activated GW and WM regions. The ability to induce BOLD signals within the SC by stimulation of the cortex potentially provides a new way to interrogate functional connections more precisely. The neuronal mechanisms underlying potential antidromically evoked BOLD signals within the SC in gray and WM warrant future investigation.

A notable result of these studies is the reliable detection of BOLD effects in WM. BOLD signals in WM have been largely ignored by others, and are often treated as nuisance regressors in fMRI analyses. BOLD signal increases in GM are usually interpreted as physiological responses to increased demands for nutrients that are required for increased activity, but the nature and driving force for such responses in WM are less clear. The ratio of glial cells to neurons is much higher in WM than in GM, so it is possible that BOLD effects in WM may partially reflect the energy demands of different cellular processes than in GM.

Materials and methods

All procedures involving animals were conducted following the National Institutes of Health and ARRIVE guidelines and were approved by the Institutional Animal Care and Use Committee of Vanderbilt University under protocol M1600079 (originally approved on 5/03/2016, and most recently reviewed and approved on 2025 April 25). The complete information is provided in the Supplementary material.

Supplementary Material

Supplementary material is available at PNAS Nexus online.

Funding

These studies were supported by grants R01NS092961, R01NS078860, and S10OD025085 provided by the National Institutes of Health.

Author Contributions

John C. Gore (Conceptualization, Funding acquisition, Methodology, Project administration, Resources, Supervision, Writing—original draft), Pai-Feng Yang (Formal analysis, Investigation, Methodology, Writing—review & editing), Arabinda Mishra (Data curation, Formal analysis, Methodology, Software, Writing—review & editing), Feng Wang (Data curation, Investigation, Methodology, Writing—review & editing), Zhangyan Yang (Formal analysis, Investigation, Software, Writing—review & editing), Anirban Sengupta (Formal analysis, Investigation, Software, Methodology, Writing—review & editing), and Li Min Chen (Conceptualization, Supervision, Funding acquisition, Investigation, Methodology, Project administration, Writing—review & editing)

Data Availability

All original data are available from the Openneuro (https://openneuro.org/datasets/ds006561) repository. The processing code and instructions may be found in GitHub at https://github.com/arimishra/repositoryPNAS.