True and apparent optogenetic BOLD fMRI signals

Authors


Abstract

Purpose

Optogenetic fMRI (ofMRI) is a novel tool in neurophysiology and neuroimaging. The method is prone to light-induced artifacts, two of which were investigated in this study.

Methods

ofMRI was performed in rats using two excitatory opsins (ChR2 and C1V1TT) virally transduced in somatosensory cortex or thalamus. Heat-induced apparent BOLD activation at the site of the optical fiber and stimulation light–induced activation of the visual pathways were investigated, and control experiments for these two artifacts were established.

Results

Specific optogenetic BOLD activation was observed with both opsins, accompanied by BOLD in the visual pathways. Unspecific heat-induced BOLD was ruled out by a control experiment employing low-level constant illumination in addition to pulsed optogenetic stimulation. Activation of the visual pathways was confirmed to be physiological by direct visual stimulation of the eyes and was suppressed by additional low-level constant light to the eyes. Light inside the brain was identified as one source of the BOLD signal observed in the visual pathways.

Conclusion

ofMRI is a method of tremendous potential, but unspecific activations in fMRI not caused by the activation of opsins must be avoided or recognized as such. The control experiments presented here allow for validating the specificity of optogenetic stimulation. Magn Reson Med 77:126–136, 2017. © 2016 Wiley Periodicals, Inc.

INTRODUCTION

During the last decade, optogenetics has revolutionized experimental research on brain networks in animal models [1, 2]. In this context, optogenetics affords to activate or inhibit genetically defined cells in the central nervous system (CNS), expressing light-sensitive ion channels or light-driven pumps (opsins), by light of a defined wavelength. Illumination can be achieved through an optic fiber that can be implanted into the CNS of a living subject. Expression of opsins is achieved either by using transgenic animals or by local transduction via viral gene transfer, driven by a cell-specific promoter. Thus, optogenetics renders a localized activation or inhibition of a genetically defined population of cells in the CNS possible, and thereby affords a novel level of control of neuronal networks in the brain. However, to assess the effect of optogenetic manipulation to the network, it needs to be combined with a method able to detect the activation of brain networks. BOLD fMRI (blood oxygenation level–dependent functional MRI) represents such a method, allowing for the brain-wide readout of neuronal activity by the mechanisms of neurovascular coupling. The feasibility of combining optogenetic control with BOLD fMRI and the aptitude of BOLD fMRI to monitor manipulations of brain networks has been demonstrated in the seminal papers by Lee et al [3] and Desai et al [4]. However, the mechanism leading to BOLD signals upon optogenetic stimulation has been questioned [5].

The combination of optogenetics and BOLD fMRI (ofMRI) has been explored in rodents, both anesthetized and awake, as well as in primates. Individual aspects of the optogenetically evoked BOLD response such as spatial limits [6], temporal characteristics [3, 7], stimulus strength dependence [7, 8], stimulus frequency dependence [7] and linearity upon repeated stimulations [9] have been assessed in rodents. Furthermore, sensory inputs in the barrel fields have been mimicked by optogenetics, under brain-wide fMRI recordings [10]. In primates, the interference of optogenetic stimulation and visual pathways, affecting saccade latencies, has been demonstrated [11]. Different aspects of how optogenetic stimulation activates networks in the brain such as frequency dependence and similarity to sensory stimulation have been addressed by ofMRI in a number of studies [3, 4, 12-16].

From a technical point of view, ofMRI is particularly challenging, as commonly used EPI sequences are prone to susceptibility artifacts at the fiber implantation site. Furthermore, artifacts may arise from the illumination of the brain. For optogenetic modulation, a number of excitatory and inhibitory opsins are available with characteristic activation wavelength and channel kinetics (eg, time to peak, time for channel closure τoff). A widely used excitatory opsin suitable for ofMRI is Channelrhodopsin-2 H134R (ChR2), which is activated by blue light. We additionally selected the red-shifted chimeric excitatory opsin C1V1TT [17], which has slower off kinetics (τoff of 50 ms versus 10 ms for ChR2) [18, 19], and is activated by green light. Typically, light intensities exceeding 1 mW mm−2 are required to stimulate opsins. However, light applied through an optic fiber is strongly absorbed and scattered in the brain. High light intensities in the fiber are therefore required to illuminate a sufficiently large volume above this threshold [20]. Such intensities may lead to heating effects at the tip of the fiber, which may give rise to an apparent BOLD signal, as previously characterized in detail [8]. To exclude apparent BOLD, careful determination of the threshold for heat-induced apparent BOLD is required for each experimental paradigm. Simply reducing light intensities may result in loss of optogenetic activation, as the volume with above (optogenetic) threshold activation may become too small. To overcome this problem, we present a novel control experiment that allows to quickly but unambiguously distinguish optogenetically evoked BOLD from apparent BOLD.

We further address another illumination artifact that may confound the analysis of optogenetically evoked BOLD signal. We investigated the activation of the visual network by the stimulation light applied to the brain via an implanted optic fiber, and show that this activation is independent of opsin expression. Also for this confounder, we present a control experiment to distinguish between optogenetic stimulation and unspecific activation of the visual pathways.

METHODS

All experiments were carried out according to the German Tierschutzgesetz and were approved by the Landesamt für Natur, Umwelt und Verbraucherschutz of Nordrhein-Westfalen, Germany (A787.54.04.2010.A274). We performed experiments on 30 female fisher rats with a body weight between 160 and 180 g.

Animal Preparation

Stereotactic injections in 18 animals were conducted under deep isoflurane anesthesia (Forene, Abbott, Wiesbaden, Germany). Viral solutions were delivered to the brain by a glass micropipette connected to a 10-mL syringe by gentle manual pressure.

We used two common opsins, either ChR2 or the red-shifted chimeric opsin C1V1TT [17], which were virally transduced in the cortex and thalamus of the rat brain. Animals were pretreated with the analgesic Metacam (1 mg/kg s.c.). For transduction of ChR2, two replication-deficient adeno-associated virus (rAAV) preparations were mixed at a ratio of one volume rAAV-CAG-Cre and four volumes rAAV-EF1A-DIO-hChR2(H134R)-mCherry (nine animals). For transduction of C1V1TT rAAV/CamK, IIa-C1V1(E122T/E112T)-TS-eYFP was used (nine animals). A total of 0.5 µl of the viral solution was slowly injected via a small craniotomy into the primary somatosensory cortex, front limb region (S1FL), at AP 0 mm, ML + 3.5 mm, DV 1.2 mm with an angle of 55 ° (vertical from medial (five animals with ChR2, three animals with C1V1TT) according to the stereotactic coordinates of the Paxinos Watson rat brain atlas [21], into the posterior thalamic nucleus (POm) at AP 3.3 mm, ML + 1 mm, DV 5 mm with an angle of 78 ° (vertical from medial) (four animals with ChR2, four animals with C1V1TT), or into VPM at AP 3.3 mm, ML + 1 mm, DV 6 mm with an angle of 75 ° (vertical from medial) (two animals with C1V1TT). After the injection, the pipette was held in place for 2 min before slowly retracting it. The scalp incision was closed with a suture, and postinjection analgesics were given to aid recovery. Optical stimulation and recordings were carried out after a minimum of 14 days after viral construct injection.

Before the imaging experiment, animal preparation of either virus-transduced (n = 18) or naïve animals (n = 10) was performed under surgical depth of isoflurane anesthesia. The optical fiber delivering the stimulation light was inserted perpendicular to the brain surface via a small craniotomy into S1FL at AP 0 mm, ML + 4 mm, DV 0.3 mm (1 mm lateral to the previous craniotomy into the transduced area), or into POm/VPM at AP 3.3 mm, ML + 2 mm, DV 4.6 mm/5.1 mm and glued to the skull with UV glue.

fMRI Procedures

MRI was performed on a 9.4 T small animal imaging system with a 0.7-T/m gradient system (Biospec 94/20, Bruker Biospin GmbH, Ettlingen, Germany) equipped with a RF surface coil with fiber lead-through, placed on the animal's head, and a custom-built optical setup consisting of a blue laser at 488 nm wavelength (Saphhire, Coherent, Dieburg, Germany) with intensity control by an acousto-optic modulator (AOM) for ChR2 activation, and a green laser at 552 nm with internal fast intensity modulation (OBIS, Coherent, Dieburg, Germany) for C1V1TT activation (Fig. 1a). The entire experimental setup was controlled by a custom-written LabView software (National Instruments), providing trigger pulses for MRI acquisition and operating the lasers.

Figure 1.

Experimental setup for ofMRI experiments. The customized surface coil has a fiber lead-through. The optical part of the setup consists of a blue laser with intensity control by an AOM, a green laser with direct intensity modulation, and coupling of both laser beams into optical fibers. Validation of ChR2 expression in cortex by confocal fluorescence microscopy: opsin expressing area (b) and detail (c). Validation of C1V1TT expression by confocal fluorescence microscopy: opsin expressing area (d) and detail (e).

Experiments were performed under medetomidine sedation. After placement of the animal in the scanner, sedation was induced by a subcutaneous bolus injection of 0.04 mg/kg medetomidine (Dormitor 1 mg/mL, Pfizer, Orion Pharma, Espoo, Finland), followed by continuous subcutaneous infusion of 0.05 mg/kg/h medetomidine. Isoflurane inhalation was stepwise reduced while monitoring the respiration rate and discontinued within the first 5 min after bolus injection. As described previously [22], we observed a drop of heart rate, respiration rate, and mean arterial blood pressure directly after starting the medetomidine infusion. fMRI experiments were performed at least 30 min after the change of anesthesia. The animal cradle was heated by a thermostat, and the respiration rate and body core temperature were routinely monitored during the experiments.

For the anatomical images, a T2-weighted two-dimensional RARE sequence was used with TR/TE 2000/12.7 ms, RARE factor 8, 256 matrix, 110 × 100 µm² spatial resolution, slice thickness 1.2 mm, 6–16 contiguous slices. For BOLD fMRI measurements, math formula-weighted images were acquired with a single-shot gradient echo echo planar imaging (EPI) sequence with TR = 1 s and TE = 18 ms. The spatial resolution of the MR images was 350 × 325 µm², slice thickness 1.2 mm, 6–9 contiguous slices.

For stimulation, a block paradigm was used with 10 s of stimulation followed by 20 s baseline for 10 min. Stimulation consisted of light pulses of 10-ms duration at 9 Hz with a light intensity of 80 mW mm−2 at the tip of the fiber, if not stated otherwise.

After the experiments, rats were transcardially perfused with 4% paraformaldehyde (PFA) under deep isoflurane anesthesia. Brains were excised, fixed overnight (4% PFA), and transferred to 30% sucrose solution. Axial sections (20 µm) were prepared using a cryotome (Leica CM1850, Wetzlar, Germany). Opsin expression was validated by confocal microscopy (SP 8, Leica, Mannheim, Germany) with 10x and 40x objectives for both opsins (Figs. 1b–1e) [23].

fMRI Analysis

fMRI data from individual animals were analyzed with a custom-written analysis script in ImageJ. The initial five scans were discarded from each imaging series to remove signal variations at the beginning of data acquisition. The remaining images were spatially filtered with a Gaussian kernel of 0.5 mm. A voxelwise t-test was calculated to compare data during stimulation (with a time shift of 2 s to account for the delay of the hemodynamic response) versus the rest, to identify significant signal changes following the stimulation. The statistical significance level for the activation maps was set at P < 0.001. The resulting maps showing % BOLD signal change in significant voxels were masked to include only brain and superimposed to EPI images. BOLD time courses were obtained by summing the image intensities over a region of interest (ROI) for all 20 stimulations of one experiment using a custom-written MATLAB script (The MathWorks, Natick, Massachusetts, USA). The ROI was chosen individually in each animal, including all activated pixels in the respective brain region. For thalamic stimulation, the ROI in the thalamus was defined below the fiber tip. For measurements with illumination of the blindfold, ROIs were copied from the intrabrain illumination experiments from the same animal. Averaged BOLD time courses were calculated for the individual groups and are reported as mean ± standard error of the mean (SEM).

ofMRI Experiments

BOLD fMRI experiments with optogenetic stimulation of ChR2 or C1V1TT were performed on all 18 opsin-expressing animals. Activation maps were calculated to identify areas of correlated activity. Averaged time courses were obtained in the cortex upon cortical (n=7) and thalamic (n=5) optogenetic stimulation, and at the respective fiber position in the thalamus (n=5).

fMRI Control Experiments to Rule Out Heat-Induced Apparent BOLD

Two different experiments were performed to address the issue of heat-induced apparent BOLD signal upon stimulation with blue light. Four naïve animals were used to exclude that heat-induced apparent BOLD occurs under our standard conditions for ofMRI. The intensity of the stimulation pulses and their duration and frequency determine the energy deposited in the tissue, which is proportional to the mean ( = time averaged) light intensity. We varied the mean intensities and applied 0.9, 2, 9, 22, and 92 mW mm−2 in three naïve animals, and 22, 46, 69, and 92 mW mm−2 in one naïve animal.

In six of the ChR2-expressing animals, a control experiment to “switch off” optogenetic activation was performed. A permanent constant low-level illumination (blue light, 11 mW mm−2, during both stimulation-on and stimulation-off periods) was switched on in addition to the optogenetic block stimulation (blue light, 10-ms pulses at 9 Hz and 80 mW/mm−2 for 10 s, mean intensity of 7 mW mm−2). In two additional naïve animals, this control experiment was validated using constant illumination (blue light, 11 mW mm−2) in addition to heat stimulation (100-ms pulses at 9 Hz, 102 mW mm−2 for 10 s).

fMRI for Stimulation of the Visual Pathways

A control experiment with additional visual stimulation (direct illumination of the eye) was performed in one C1V1TT-expressing and five naïve animals. In all animals, light was delivered to either S1FL or thalamus by an implanted optic fiber, with light intensities of 70 to 95 mW mm−2 at the fiber tip. For additional visual stimulation, a second fiber was placed approximately 2 cm in front of one eye. Either pulsed or continuous light for illumination of the eyes was applied with approximately 0.3–3.0 µW mm−2.

Stimulation of the visual pathways was analyzed by grouping animals according to light color (green and blue) and fiber position (cortex and thalamus), independent of opsin expression, and calculating the averaged BOLD time courses: green light in cortex (n=6), blue light in cortex (n=9), green light in thalamus (n=6), and blue light in thalamus (n=2), respectively.

To rule out stimulation by stray light, the control experiment was repeated with blindfolded rats (two with blue light on the blindfold and green light in the brain, one with green light on the blindfold and blue light in the brain).

RESULTS

Implementation of ofMRI

For ofMRI an optical setup allowing for control of two lasers, sensory stimulation, and MR acquisition was established. A blue laser at 488-nm wavelength with submillisecond intensity control by an acousto-optic modulator (AOM) for ChR2 activation, and a green laser at 552 nm with internal fast intensity modulation for C1V1TT activation was used (Fig. 1a). Stimulation light was delivered via a 200-µm optical fiber passing a lead-through of a custom-made surface coil.

Animals were injected with AAVs encoding for either ChR2 or C1V1TT (Figs. 1b–1e) in cortex (S1FL) or thalamus (VPM/POm) and subjected to optogenetic stimulation at the virus injection site. Confocal microscopy revealed strong expression of ChR2 (Fig. 1b) and C1V1TT (Fig. 1d) in layers II/III and V of rat somatosensory cortex and thalamus. The area of expression constitutes a sphere of approximately 500-µm diameter. At higher magnification, membrane-bound homogenous expression, excluding the nucleus, of both constructs could be demonstrated (Figs. 1c and 1e).

Slight MR signal intensity disturbances at the site of fiber implantation were observed regularly in anatomical T2-weighted and corresponding EPI MR images (Figs. 2a and 2b). For cortical stimulation (10-ms pulses at 9 Hz, 80 mW mm−2, mean intensity of 7 mW mm−2), a robust activation at the site of stimulation was detected (Fig. 2c). Optogenetic stimulation of POm or VPM led to activation of cortical projection targets being identified as primary or secondary somatosensory cortex and/or motorcortex (Figs. 2d and 2e). A typical BOLD time course with a signal increase 3 s after the start of the stimulation, reaching a maximum after 5–6 s, was observed. Averaged time courses of cortical BOLD signal upon optogenetic stimulation in the cortex or thalamus are shown in Figures 3a and 3b. BOLD amplitude was approximately 1% for optogenetic stimulation in the cortex (Fig. 3a). For thalamic stimulation, BOLD amplitude in the activated cortical area was approximately 0.7% (Fig. 3b). No robust BOLD signal was observed in the thalamus, only small signal changes upon thalamic stimulation (Fig. 3c).

Figure 2.

Representative BOLD activation maps upon optogenetic stimulation. The implanted fiber is discernible in both anatomical RARE (a) and functional EPI MR images (b), indicated by red arrows. BOLD activation upon cortical stimulation (c) and upon thalamic stimulation (d, e). Local optogenetic BOLD response at the fiber tip upon thalamic stimulation was not detected or very weak (d), but robust responses were detected at cortical projection targets (e).

Figure 3.

Averaged BOLD time courses upon optogenetic stimulation. BOLD response in cortex upon cortical optogenetic stimulation (n=7) (a), and BOLD response at cortical projection target where activation was detected upon thalamic optogenetic stimulation (n=5) (b). At the site of thalamic stimulation, no significant signal change was detected (same animals as in b) (c). Gray areas indicate SEM. In the cortex, ROIs were selected individually in each animal, including the activated pixels (a and b). In the thalamus, the ROI was defined to include the illuminated area under the fiber tip (c).

Heat-Induced Apparent BOLD

Because high light intensities are required for optogenetic stimulation, heating effects at the fiber tip are a potential concern, possibly leading to an unspecific, apparent BOLD response, independent of whether opsins are locally expressed. Occurrence of apparent BOLD is dependent on the energy deposited in the tissue, which is determined by the intensity of the stimulation pulses and their duration and frequency, as has been studied in detail recently [8]. In four naïve animals we performed light stimulation at different mean light intensities. No significant signal changes resulting from heat deposition were detected at mean intensities of up to 22 mW mm−2 (Figs. 4a–4c), an intensity more than three times the value used for effective optogenetic stimulation (7 mW mm−2). To confirm that apparent BOLD can be evoked by light application, we used a mean intensity of 92 mW mm−2 (100 ms pulses at 9 Hz, mean intensity of 102 mW mm−2), leading to an apparent BOLD response (Fig. 4d).

Figure 4.

Representative BOLD activation maps obtained upon stimulation with different mean light intensities. Light stimulation of naïve rat brain did not lead to heat-induced apparent BOLD activation up to a mean light intensity of 22 mW mm−2 (a–c). At a mean light intensity of 92 mW mm−2, a negative activation in the slice of fiber implantation and a positive activation in the adjacent slices was detected (d).

However, under different experimental conditions, higher light intensities may be required for optogenetic activation, and heat-induced apparent BOLD may not be easily distinguished from true BOLD. We therefore propose a control experiment to test for specificity of the BOLD response in ChR2-expressing animals, “switching off” the BOLD response while applying higher mean light intensities.

The response of opsins is determined by their specific intrinsic channel kinetics. For ChR2 and C1V1TT [1, 24], light activation leads to short, high-amplitude, peak currents followed by low-amplitude, long-lasting, steady-state currents. Short peak currents inactivate fast and require approximately 10 s to deinactivate [24]. Constant illumination results in continuous steady-state currents, preventing the recovery of peak current of pulsed excitation. To achieve this state, we used permanent constant illumination of 11 mW mm−2 in addition to the pulsed illumination (10 ms pulses at 9 Hz, mean intensity of 7 mW mm−2) (Fig. 5b) in six of the ChR2-expressing animals. No BOLD response was detected with constant additional illumination, but BOLD signal was restored when no constant illumination was applied in subsequent experiments (Fig. 5a). This observation rules out that the originally observed BOLD response was heat-induced, which would have been expected to occur at the higher mean intensity of 18 mW mm−2. To validate the specificity of this control experiment, it was repeated in two naïve animals under stimulation conditions for which a heat-induced apparent BOLD response was observed. Pulsed illumination at a mean intensity of 92 mW mm-2 caused apparent BOLD signal (Fig. 5c), which did not vanish when continuous illumination was added (Fig. 5d).

Figure 5.

Control experiment for specificity of ChR2 activation. BOLD activation map upon optogenetic stimulation in cortex (a) (pulse duration 10 ms, 9 Hz, pulse train duration 10 s). BOLD response was not observed in the same animal stimulated with identical light pulses, but with additional continuous illumination (b). In contrast, heat-induced activation (c) (pulse duration 100 ms, 9 Hz, 102 mW mm−2, pulse train duration 10 s) did not vanish upon additional continuous illumination of 11 mW mm−2 (d).

Activation of the Visual Pathways

In experiments with optogenetic stimulation in opsin-expressing animals, BOLD signal was observed regularly in subcortical regions (Fig. 6a) caudal to the optogenetically evoked BOLD in the somatosensory cortex (see Figure 2b). Based on anatomical landmarks, the activated regions are part of the visual pathways, namely the dorsal lateral geniculate nucleus (DLG), optic nerve layer of the superior colliculus, and the nucleus of the optic tract. We evaluated whether this activation was caused directly by the stimulation light without participation of the opsins. Visual stimulation was performed by illuminating the eyes of an opsin-expressing animal with pulsed light in the same block paradigm as the optogenetic fMRI experiments. Similar activation of the visual pathways was observed (Fig. 6b). As a control experiment, optogenetic stimulation was performed while one of the rat's eyes was continuously illuminated. In the activation map, the response of the contralateral visual pathway was strongly attenuated (Fig. 6c).

Figure 6.

Control experiment to investigate the activation of visual pathways. Optical stimulation of opsin-expressing rats, shown here for an animal expressing C1V1TT, regularly resulted in activation of subcortical regions in addition to the somatosensory cortex (a), namely the dorsal lateral geniculate nucleus (DLG), optic nerve layer of the superior colliculus, and the nucleus of the optic tract. When illuminating the eyes of this animal with pulsed light in the same block paradigm, similar activation of the visual pathways was observed (b). In the case of optogenetic stimulation, while one of the rat's eyes was continuously illuminated, the response of the visual pathway in the contralateral hemisphere of the brain was not present any more (c, left side in the image). When performing similar experiments with naïve animals, illumination of the brain with green (d) or blue (g) light resulted accordingly in activation of the visual pathways. Visual stimulation of one eye resulted in a similar activation pattern (e, h). Continuous illumination of one eye in addition to pulsed illumination of the brain abolished the BOLD response at the contralateral side (f, i).

To validate that this effect on the visual pathways was independent of the optogenetic stimulation, we repeated the same experiment in naïve animals: In one experiment, pulsed blue light was applied to S1FL, and green light to the left eye for visual stimulation. In a second experiment, intrabrain illumination (S1FL) was performed by green light and visual stimulation by blue light. Both experiments resulted in similar activation patterns. Pulsed light applied to the brain led to activation of the visual pathways (Figs. 6d and 6g). In contrast to the animals transduced with an opsin, no local activation at the fiber location in S1FL was detected. Visual stimulation (in block design) of one eye led to the activation of the visual pathways, primarily in the contralateral hemisphere (Figs. 6e and 6h). For visual stimulation, low intensities in the order of 1 µW mm−2 were sufficient. Continuous visual stimulation of one eye during pulsed brain illumination abolished the activation in the contralateral hemisphere (Figs. 6f and 6i). Group-averaged BOLD time courses for the different experimental conditions (fiber in S1FL or thalamus; blue light or green light) showed that BOLD responses with typical temporal profiles were observed for all groups (Figs. 7a–7d). Strong signal variations did not allow for analyzing intergroup differences.

Figure 7.

BOLD maps and time courses upon visual stimulation and intrabrain illumination. Mean BOLD time courses of the activation of the visual pathways are shown as a solid black line upon light delivery through the optic fiber implanted into the cortex ((a) green light, n=6; (b) blue light, n=9) and into thalamus ((c) green light, n=6; (d) blue light, n=2). In (a–d), ROIs were selected individually in each animal, including activated pixels in the respective brain region. Gray areas indicate SEM. To identify intrabrain illumination as a source of activation of visual pathways, the eyes and snout of the animals were covered. Intrabrain illumination (fiber in cortex) resulted in BOLD signal in the visual pathways for both green (e) and blue (f) light. Direct illumination of the covered eyes and snout showed no activation of the visual pathways for both green (g) and blue light (h). Averaged BOLD time courses of visual pathways from single experiments upon intrabrain illumination (green in i, blue in j). No response was detected for blindfolded animals upon direct illumination of the eyes with blue light (blue in i) and green light (green in j). The same ROIs were used for intrabrain illumination and illumination of the blindfold. For comparison, time courses from different animals with illumination of uncovered eyes are shown (black in i and j). For reasons of clarity, no SEM is shown.

Light from different sources may contribute to activation of the visual pathways (Fig. 8a). The optical fiber may emit light if bending or contact with other parts inside the narrow magnet bore is not carefully avoided. In addition, light from the fiber-brain interface or the massive diffuse illumination of the whole brain (Figs. 8b and 8c) may lead to activation of the visual pathways. We repeated the experiment with rats with covered eyes and snout. Direct illumination of the cover in front of the eye did not result in activation of the visual pathways (Figs. 7g–7j). However, light applied to the brain resulted in BOLD signal in the visual pathways (Figs. 7e, 7f, 7i, 7j), indicating that intrabrain illumination was partly responsible for the observed activation of the visual pathways.

Figure 8.

Diffuse brain illumination provides one possible source of activation of the visual pathways. Schematic illustration of potential sources of light that may contribute to stimulation of the visual pathways (a): [1] Due to fiber bending, stray light coupling back into the fiber or contamination of the dismantled fiber, light may leak from the fiber in the magnet bore; [2] light shining through the exposed skull or directly from the fiber brain interface; [3] diffusely scattered light inside the brain. Illumination of excised rat brain with blue (b) and green (c) light through an optic fiber lead to illumination of the whole brain.

DISCUSSION

The combination of optogenetic manipulation of brain networks and a read-out by BOLD fMRI may provide insights into neuronal network function in unprecedented detail. However, the complex surgical procedures, method integration, fMRI acquisition, and signal analysis are prone to various artifacts and render this multimodal approach challenging. In this study, we performed ofMRI in rats, resulting in robust activations either locally in S1FL when stimulating the cortex or at the thalamocortical projection targets in S1, S2, and/or motorcortex when stimulating thalamus. BOLD time courses followed a typical linear delayed hemodynamic response as reported by Khan et al [9]. However, optogenetic fMRI relies on BOLD as an indirect readout of neuronal activation. It is therefore of particular importance to avoid, or at least reliably recognize, unspecific BOLD signal originating from other sources than opsin activation. In this study, we investigated two possible mechanisms of unwanted activation in ofMRI experiments that are not caused by opsin activation: an apparent BOLD response caused by heating of the tissue [8], and activation of the visual pathways by light applied to the brain. For both mechanisms, we presented control experiments to test whether the detected fMRI signal changes are specifically caused by optogenetic stimulation.

We observed robust BOLD responses upon optogenetic stimulation at light intensities far below the threshold for heat-induced apparent BOLD, which was above 22 mW mm−2 under our experimental conditions. These values agree with a recent study by Christie et al [8], who systematically characterized temperature effects upon optical stimulation, and confirm that our observed BOLD was the result of optogenetic activation. We therefore conclude that our optogenetic stimulation (10-ms pulses at 9 Hz, mean intensity of 7 mW mm−2) in the ChR2-expressing animals was below the threshold for heat-induced apparent BOLD, and the observed BOLD signal was the result of optogenetic activation. To distinguish between an apparent heat-induced and an optogenetic BOLD response, we implemented a straightforward control experiment that may be used as a standard control experiment in ofMRI. As all current class I opsins (ie, light-gated channels and pumps) exhibit a peak and steady state kinetics, the only prerequisites are that the laser intensity must be controlled precisely and fast enough between low and high intensity. By simply applying constant illumination in addition to pulsed stimulation, it is therefore possible to rule out false positive fMRI activation. Although we showed this control experiment only for ChR2, it may in principle be applied to other class I opsins as well.

We have shown that application of light to the brain can lead to the activation of the visual system. Such activations may be particularly problematic when networks in subcortical regions of the brain are studied [13, 15], and may mask activations in projection targets of optogenetically activated regions. Regardless of the presence and the type of opsins, we regularly observed activation of the visual pathways in optogenetic stimulation experiments of the somatosensory network. Visual stimulation by light applied directly to the eyes led to similar activation patterns. These results correspond to a previous fMRI study by van Camp et al [25], in which strong bilateral activation of the visual pathways was observed upon visual stimulation with white light, and a weak activation of the visual cortex was reported. In our experiments, the activation of the visual cortex was only sporadically observed, as our imaging slices were positioned rostral to optimally detect activations in the primary and secondary sensory cortex, and also in thalamic regions.

Constant illumination of one eye abolished the BOLD response in the contralateral part of the visual pathway, hypothetically by rendering this eye less sensitive to the light pulses applied to the brain. From this observation, it can be concluded that indeed the same pathways are affected by both direct illumination of the eyes and indirect intrabrain illumination via the optic fiber. We showed that activation of the visual pathways also occurs when the animals are blindfolded. However, it is not possible to dissect individual contributions of different sources of light during a normal optogenetic experiment. Yet, diffuse light inside the brain contributes to the activation of the visual pathways. This is critical when the identification of brain network components is the matter of investigation. The occurrence and extent of BOLD signal in participating structures may be altered by unintentionally applying multiple stimulations, optogenetic and visual, at once, independent of the opsin used. A straightforward strategy to prevent the detection of BOLD response in the visual pathways is to uncouple the optogenetic stimulation paradigm from stimulation of the visual pathways by applying constant light to the eyes. Therefore, we suggest that ofMRI be performed routinely with background light inside the scanner.

In summary, we conclude that in ofMRI in particular, light-induced artifacts have to be considered to exploit the full potential of this novel method. Because optogenetics has evolved as a valuable tool in neurophysiology and neuroimaging, and was recently shown to reliably reproduce sensory stimulation in rodents [7, 23], ofMRI may become more widely used. The control experiments introduced by our study may add to the reliability of ofMRI studies, as they are generally applicable and allow for the validation of the specificity of optogenetic stimulation.

ACKNOWLEDGMENTS

This work was supported by the Bavarian State Ministry of Sciences, Research and the Arts (“ForNeuroCell”); the DFG (SFB 1080); the Focus Program Translational Neurosciences (ftn), the Interdisciplinary Center for Clinical Research Münster (Fa3/1603), and the Excellence Cluster Cells in Motion (DFG EXEC 1003). We thank Daniel Kalthoff for providing ImageJ macros for fMRI analysis, and Michaela Moisch and Kornelia Parusel for excellent technical assistance. We thank Karl Deisseroth, Stanford University, for providing optogentic constructs via Addgene / Penn Vector Core.

Ancillary