NeuroImage 95 (2014) 29–38

Contents lists available at ScienceDirect

NeuroImage journal homepage: www.elsevier.com/locate/ynimg

Complex relationship between BOLD-fMRI and electrophysiological signals in different olfactory bulb layers Bo Li a,c,1, Ling Gong a,c,1, Ruiqi Wu a,c, Anan Li a,⁎, Fuqiang Xu a,b,⁎⁎ a Key Laboratory of Magnetic Resonance in Biological Systems and State Key Laboratory of Magnetic Resonance and Atomic and Molecular Physics, Wuhan Institute of Physics and Mathematics, Chinese Academy of Sciences, Wuhan 430071, China b Wuhan National Laboratory for Optoelectronics, Wuhan 430074, China c University of the Chinese Academy of Sciences, Beijing 100049, China

a r t i c l e

i n f o

Article history: Accepted 17 March 2014 Available online 25 March 2014 Keywords: Functional MRI Olfactory bulb Blood oxygenation level dependent (BOLD) Local field potentials (LFP) Layer-dependent

a b s t r a c t Blood oxygenation level dependent functional magnetic resonance imaging (BOLD-fMRI), one of the most powerful technologies in neuroscience, measures neural activity indirectly. Therefore, systematic correlation of BOLD signals with other neural activity measurements is critical to understanding and then using the technology. Numerous studies have revealed that the BOLD signal is determined by many factors and is better correlated with local field potentials (LFP) than single/multiple unit firing. The relationship between BOLD and LFP signals under higher spatial resolution is complex and remains unclear. Here, changes of BOLD and LFP signals in the glomerular (GL), mitral cell (MCL), and granular cell layers (GCL) of the olfactory bulb were evoked by odor stimulation and sequentially acquired using high-resolution fMRI and electrode array. The experimental results revealed a rather complex relationship between BOLD and LFP signals. Both signal modalities were increased layer-dependently by odor stimulation, but the orders of signal intensity were significantly different: GL N MCL N GCL and GCL N GL N MCL for BOLD and LFP, respectively. During odor stimulation, the temporal features of LFPs were similar for a given band in different layers, but different for different frequency bands in a given layer. The BOLD and LFP signals in the low gamma frequency band correlated the best. This study provides new evidence for the consistency between structure and function in understanding the neurophysiological basis of BOLD signals, but also reminds that caution must be taken in interpreting of BOLD signals in regard to neural activity. © 2014 Elsevier Inc. All rights reserved.

Introduction Blood oxygenation level-dependent functional magnetic resonance imaging (BOLD-fMRI) is one of the most powerful technologies for neuroscience because of its capability to non-invasively study any brain region with high spatiotemporal resolution (Bailey et al., 2013; Bandettini, 2012; Logothetis, 2008). It has been widely used not only to map brain activation, but also to study the dynamics of neural networks (Logothetis, 2008; Logothetis and Wandell, 2004; Mishra et al., 2011; Sanganahalli et al., 2013). Comprehensive studies have revealed the

Abbreviations: BOLD, blood oxygenation level dependent; fMRI, functional magnetic resonance imaging; LFP, local field potentials; OB, olfactory bulb; GL, glomerular layer; MCL, mitral cell layer; GCL, granule cell layer; IAA, iso-amyl acetate; OCT, octanal. ⁎ Corresponding author at: Wuhan Institute of Physics and Mathematics, The Chinese Academy of Sciences/State Key Laboratory of Magnetic Resonance and Atomic and Molecular Physics, Wuhan 430071, China. Fax: +86 27 87199543. ⁎⁎ Corresponding author at: Wuhan Institute of Physics and Mathematics, The Chinese Academy of Sciences/State Key Laboratory of Magnetic Resonance and Atomic and Molecular Physics, Wuhan 430071, China. Fax: +86 27 87199543. E-mail addresses: [email protected] (A. Li), [email protected] (F. Xu). 1 Equal contribution.

http://dx.doi.org/10.1016/j.neuroimage.2014.03.052 1053-8119/© 2014 Elsevier Inc. All rights reserved.

relationship among neural activity, hemodynamics, and BOLD signals in many brain regions, and found that the BOLD signals are affected by multiple factors, including the local neural activity, metabolic capacity, the blood vessel system (blood flow and volume), and the neuroanatomy of the examined regions (Huttunen et al., 2008; Logothetis et al., 2001; Mishra et al., 2011; Mullinger et al., 2013; Shmuel et al., 2006). Therefore, the BOLD signal measures local changes in brain hemodynamics and metabolism, and should be interpreted with regard to the specific brain area where it is measured. Extracellular electrophysiological signals, especially single/multiple unit activity and local field potentials (LFP), which reflect the neural activities of single/a few neurons and cell assembly directly, are another type of the most commonly used methods in neuroscience (Buzsaki, 2010; Buzsaki et al., 2012; Panagiotaropoulos et al., 2012; Rasch et al., 2009). Compared with MRI, electrophysiological recordings have much higher sensitivity and temporal resolution. Instrumentally, they are simpler, less expensive and easier for maintenance. Functionally, they are capable of monitoring neuronal and neural activities. Thus they can be used to study the properties and functions of a specific neuron, circuit, and system under different conditions such as rest and stimulated, normal and diseased, developing and adult, learning and

30

B. Li et al. / NeuroImage 95 (2014) 29–38

learned, etc. (Buzsaki, 2010). Since they directly measure the electrical signals of neurons, it is generally accepted that the LFP and neuron firing from single/multiple units are gold standards for measuring the local neural activities (Buzsaki, 2010; Buzsaki et al., 2012). It is critical to correlate the indirect measurement of neural activity by BOLD signals with a gold standard of neural activity, such as by LFP, to justify the usage of fMRI and to help with interpretation of the fMRI data. Therefore, large amounts of studies have amassed evidence on the relationship between BOLD and LFP signals from different species and brain regions (Herman et al., 2013; Lippert et al., 2010; Logothetis et al., 2001; Magri et al., 2012; Maier et al., 2008; Mishra et al., 2011; Young et al., 2011). The results have demonstrated that, generally, the BOLD signal is more correlated with the LFP (Logothetis, 2008; Logothetis and Wandell, 2004; Logothetis et al., 2001), although, in some other regions, it also correlates well with neuronal firing (Hyder et al., 2002; Maandag et al., 2007). The olfactory bulb (OB), the first center in the olfactory system, has one of the most regular laminar structures in the brain (from the exterior to the interior): the olfactory nerve layer, the glomerular layer (GL), the external plexiform layer, the mitral cell layer (MCL), the internal plexiform layer, the granule cell layer (GCL), and the subependymal zone (Shipley and Ennis, 1996). Each layer has distinct molecular and cellular compositions, different physical and biological properties, and different functions (Shipley and Ennis, 1996). In addition, the OB is one of two regions containing a significant amount of newborn neurons in the adult brain (Alonso et al., 2012; Gheusi et al., 2000; Lazarini and Lledo, 2011). Therefore, the OB has served as a popular model for a variety of neuroscience topics (Shepherd and Charpak, 2008), including the correlation of vascular density, synaptic transmission, metabolism, and neurovascular coupling in optical imaging (Gurden et al., 2006; Petzold et al., 2008). BOLD-fMRI has been successfully applied to the OB in both rats and mice. High resolution MRIs can clearly identify the OB layers (Xu et al., 2000; Yang et al., 1998), map the activity patterns evoked by odorants and pheromones (Martin et al., 2007; Schafer et al., 2006; Xu et al., 2000, 2003; Yang et al., 1998), and reveal the dynamics of neural responses and the property of adaptation (Schafer et al., 2005; Xu et al., 2005). These previous fMRI studies were basically focused on the GL, while the other layers, which play important roles in olfactory information processing and transmission, rarely received attention. Further, the underlying neural basis of the BOLD signals and the correlation of the BOLD and LFP signals in the OB layers remains to be explored. More importantly, the unique laminar structures of the OB can provide a unique model to correlate BOLD and LFP signals under different brain states. In attempting to answer these questions, we sequentially obtained the BOLD and LFP signals evoked by different odorants in three major OB layers (GL, MCL and GCL) of the same rats. The results revealed that the relationships are distinct for BOLD and LFP signals in different OB layers, magnitudinally, temporally and spatially, leading to a rather complex relationship between the two signal modalities. These results can help us understand the neurophysiological basis and thus the interpretation of BOLD signals. Materials and methods Animal preparation and odorant delivery Odor stimulation experiments were performed on 18 adult SpragueDawley rats (250–300 g). These animals were pre-anesthetized with ~2.5% isoflurane for surgery, and the skin on top of the OB was removed to expose the skull. After surgery, the anesthetic was switched to urethane (i.p., 1.5 g/kg), which was used to maintain stable anesthesia in the odor stimulation experiments. The odorants, iso-amyl acetate and octanal (IAA and OCT, Sigma, MO, USA) were delivered to freely breathing animals. The odorants were dissolved in paraffin oil to make the concentration at 50% or 5% (H-IAA at 50%, L-IAA at 5% and OCT at only

5%). A stream of charcoal-filtered, warm air flowed over the oil, and then was diluted to 1/5 by an olfactometer. The stimulation was synchronously controlled with the data acquisition system by a solenoid valve, which was driven by a digital to analog converter. Air (off) or odorized air (on) was delivered to the nose at a constant rate of 1 l/min to eliminate the effect of the airflow. Odor stimulation lasted 32 s with an inter-stimulation interval N6 min to minimize habituation, and was repeated at least 4 times for each odor. Of the 18 rats used in the present study, 9 were used for BOLD-fMRI only (4 with all the odorants, 4 with L-IAA and OCT, 1 with H-IAA and L-IAA. Five were used only for LFP recording with all the odorants. Four were sequentially used in BOLD-fMRI and LFP recording with H-IAA. Therefore, 9, 9 and 8 (9, 5 and 5) rats were used for H-IAA, L-IAA and OCT stimulation in BOLDfMRI (LFP recording), respectively. BOLD-fMRI data acquisition The imaging experiments were performed on a horizontal-bore 7.0 T BioSpec (Bruker, Ettlingen, Germany). The procedures for animal preparation were similar to those previously reported (Xu et al., 2000, 2003, 2005). Briefly, the animal's head was placed in a head holder to minimize head movement, and the animal was laid on a water heating bed. The respiration rate was monitored by recording chest wall movements using a piezoelectric device. A circular transmitting and receiving surface coil (10 mm diameter) was centered on top of the OB between the eyes of the animal to maximize signal to noise ratio. Functional MRI data was mainly acquired using multi-segment gradient echo planar sequence (TR, 1000 ms; TE, 14 ms; flip angle, ~ 45°; FOV, 12.8 × 12.8 mm2, spatial resolution, 0.1 × 0.1 × 0.5 mm3, 10 slices), which decreased the inflow effects compared with the previous fast low angle shot sequence (Gao and Liu, 2012; Liu et al., 2008). The segment number was 8 in order to minimize distortions caused by the EPI method, leading to a temporal resolution 8 s per frame. The rapid acquisition with relaxation enhancement (RARE) T2-weighted anatomical images were obtained for the same slices (TR, 2500 ms; effective TE, 60 ms; RARE factor, 4; matrix, 128 × 128; spatial resolution, 0.1 × 0.1 × 0.5 mm3, 10 slices; average number, 25). In order to construct global OB's topological graphs, higher inter-plane resolution (0.2 × 0.2 × 0.25 mm3, 22 slices) was used. LFP recordings Each rat was placed prone on a stereotaxic holder on a vibration-free table inside a Faraday cage. The skull was exposed by an electric cranial drill and covered with mineral oil to prevent drying (~2 mm lateral to midline, ~8 mm anterior to Bregma). Then the linear array microelectrode (NeuroNexus, MI, USA), with 16-channel and 0.1 mm interchannel spacing, was inserted into the OB (with an angle perpendicular to the surface) using a stereotaxic micromanipulator (Stoelting, IL, USA). The recording site was selected according to the activation maps of both odorants in fMRI studies (http://senselab.med.yale.edu/ OdorMapDB/search.aspx), both odorants evoke strong responses in this area. A steel screw was fixed on the parietal lobe to record EEG signals. As the reference site for both LFP and EEG signals, the skin of the neck was hooked with a silver wire. Body temperature of the animal was maintained at 37 °C and the respiration rate was monitored by recording chest wall movements using a piezoelectric device. All the LFP, EEG and respiratory signals were amplified (× 2000, PGA32, BW, Germany) prior to digitization (μ-1401; CED, Cambridge, UK, sampling rate: 4000 Hz). Determination of the layers of interest Based on the anatomic images, features of LFP signals and electric lesion marks, the sites of GL, MCL and GCL were determined (Fig. 1). In the MRI, these layers can be defined in anatomical images: the GL, MCL

B. Li et al. / NeuroImage 95 (2014) 29–38

Fig. 1. Selection of the layers of interest in the OB. (A) The raw LFP signals are shown with the deepest site at the bottom. The gray box indicates transient sharp potentials, which reversed their direction in the MCL. The signals in the GL, MCL and GCL predicated by the depths and signal features are shown in blue, red and green, respectively. (B) The high frequency signals in the MCL after a high-pass filter (N300 Hz, right). The spikes are also a cue for the location of MCL (left). (C) The recording sites, marked by electrocoagulation, are confirmed by MRI. All three observations (site for potential reversing, spikes, and microlesion) were used together to identify the recording sites.

were the outer and inner dark bands, respectively, and the GCL was the outer band of the subependymal zone, which showed a super-intensity in the inner core of OB. In LFP recordings, high signal to noise ratio spikes (after filtering the 300–3000 frequency signals) were easily found in the MCL, very few and small spikes were found in other layers, which were consistent with previous studies (Doucette and Restrepo, 2008; Doucette et al., 2011). The GL and GCL were determined according to their relative distances from the surface and the MCL. Furthermore, the channel containing LFP signals that reversed their polarities was identified, based on previous studies, as the presumable MCL, and the GCL was identified from LFP signals showing the largest amplitude of the LFP oscillation and the strongest transient sharp potentials (Neville and Haberly, 2003; Rojas-Libano and Kay, 2008), as shown in Fig. 1A. In order to verify the consistency of the above two methods, electrocoagulation was performed at selected LFP recordings sites (GL, MCL and GCL) by using a tiny current pulse (current intensity, 0.1 mA; duration, 0.025 s; frequency, 20 Hz; number of pulses, 100). Generally, these marks last for about 10 min, and the fast low angle shot image (TR, 400 ms; TE, 5 ms; spatial resolution, 0.1 × 0.1 ×0.5 mm3; matrix, 128 × 128; FOV, 12.8 × 12.8 mm2; average number, 8; total duration, about 6 min) was acquired to observe the dots within the effective time, then the T2 weighted anatomical images described above were obtained to further verify the recording sites (Fig. 1C). Data processing All LFP and fMRI data were analyzed off-line with Matlab (MathWorks, Inc., USA). The statistical analyses were performed by SPSS (IBM, NY, USA). Student's t-test, ANOVA and a post hoc test (least significant difference, LSD) were used to compare the difference between two-group or multi-group responses. Pearson correlation analysis was used to compare the time courses of LFP and BOLD signals, similar to previously reported analysis (Bailey et al., 2013). Results in the text were expressed as mean ± s.e (standard error). BOLD-fMRI For odor stimulation experiments, a total of 160 s (20 frames) data were acquired for each trial, in which odor stimulation started at 65 s (9th frame), lasted 32 s (4 frames). Twelve frames were analyzed,

31

including the 4 frames before odor stimulation, which were used as baseline, 4 frames during stimulation as activation, and 4 frames after the end of stimulation as recovery. Each odor was repeated for 4–5 trials. All raw data were first realigned using SPM to eliminate head movement and image shift, and then Gaussian smoothing was performed for every slice to improve the signal to noise ratio (Gaussian low-pass filter, 3 × 3; sigma, 0.8). The differences between the baseline and activated periods were compared on pixel-by-pixel basis to obtain Student's t-values as an indicator for neural activity changes. The normalized dot product (NDP) was used to quantitatively evaluate the reproducibility in our experiment, similar to previously reported methods (Schafer et al., 2006). The three-dimensional t-values matrix was calculated for each trial and a globe OB's mask was created from the anatomical images to remove pixels out of the OB. Then, NDPs (L-IAA vs. L-IAA, OCT vs. OCT, L-IAA vs. OCT) were obtained trial-by-trial for each animal. The individual slice activation was presented, by overlapping the t-values that were calculated from all trials, onto the anatomical images. The global odor maps in the different layers (Fig. 3A) were constructed as previously described (Liu et al., 2007; Xu et al., 2003) from separate rats with a higher inter-plane resolution. The time courses of signal changes (ΔS/S) were generated by averaging all pixels in the selected layers for each frame. For the comparison of response intensity, we averaged the data from all trials. When signals across different layers were compared, we also averaged the data from different odorants (for example, Fig. 3D). Local field potential The data used for analysis were selected similarly to that for the imaging studies as described above. For the odor stimulation experiment, a total of 80 s data were used for analysis for each trial (odor stimulation started at 17 s, stimulation lasted 32 s), in which the first 16 s were used as baseline, 32 s during stimulation as activation, and 32 s after the end of stimulation as recovery. The LFP signals (1–90 Hz) were filtered from raw data using the Spike2 software, and binned with a width of 1 s (4000 sample points), and the power spectral density of 1 Hz resolution was calculated with a Fourier transform to build a time–frequency graph. The LFP signals contain several frequency bands, which presumably represent different functional properties of neurophysiology (Kay et al., 2009; Neville and Haberly, 2003). Three frequency bands of LFP, beta (13–34 Hz), low gamma (L-gamma, 35–64 Hz) and high-gamma (H-gamma, 65–90 Hz), were selected as in previous studies (Li et al., 2011, 2012), and the signals were collapsed to one dimension to generate time courses. For the comparison of signal intensity, we averaged the data from all trials. When signals across different layers were compared, we also averaged the data from different frequency bands for different odorants (for example, Fig. 4D). Correlation analysis of temporal response profiles from different layers or different frequency bands, or different modalities The response profiles of LFP (n = 9) and BOLD (n = 9) induced by the same odor (H-IAA) were used for the correlation analysis. The time courses (between 16 s before and 64 s after the onset of stimulation) of LFP (80 points) and BOLD (10 points) were normalized by subtracting the baseline and dividing by the difference between the baseline and maximum to eliminate the effects of the difference of response intensity. The Pearson correlation coefficient (R-value) was used to compare the similarity among the time courses animal by animal. In the cross-layers comparison, for each pair of rats (e.g., R01 and R02), two sets of R-values were obtained and averaged to a pair of R-values, one for the same layers (mean of Gl01 × GL02, MCL01 × MCL02 and GCL01 × GCL02, Fig. 5A, right-top triangle) and another for different layers (mean of GL01 × MCL02, MCL01 × GCL02, GCL01 × GL02, GL02 × MCL01, MCL02 × GCL01, GCL02 × GL01, Fig. 5A, left-bottom triangle). Then all averaged R-values were shown on a pseudo-color matrix. A similar process was used for the cross-layer comparison for

32

B. Li et al. / NeuroImage 95 (2014) 29–38

each frequency band of the LFP. Because of the similar temporal profile among the different layers, the time courses of three layers for BOLD and each frequency band were averaged for cross-band and cross-modality analyses. In the cross-band analysis, the averaged time courses (80 points) of each frequency band were used to calculate R-values for the same bands and different bands by a similar method aforementioned. In the cross-modality analysis, the temporal resolution of LFP signals was reduced to 8 s, same as the BOLD time courses (10 points), and the R-values between the BOLD time course and each frequency band (i.e., BOLD × Beta, BOLD × L-gamma and BOLD × H-gamma; Fig. 6D) were used to compare the relationship of functional BOLD and LFP signals. Results Odorant-evoked responses in the OB To reveal the properties of the BOLD signals in the OB and further, to correlate them with electrophysiological recordings, activity patterns across the OB were obtained first. Stimulation with odorants IAA and OCT yielded robust activation. The same odorant at different concentrations (L-IAA vs. H-IAA) elicited patterns with similar topography but different intensities, with the higher concentration being stronger (Fig. 2A). The BOLD signals were correlated with specific OB layers, and revealed by superimposing the functional images over the corresponding anatomical MRI images of the same slices. The signals were stronger in the GL, but also significant in the other layers, including the GCL and MCL. IAA vs. OCT generated quite different patterns: dorsal and lateral activation by IAA, and mainly lateral activation by OCT. Comparing the activated patterns induced by different odorants or same odorant with similar concentration (Fig. 2B), found that the same odorant evoked more similar patterns than different odorants. Compared with previous OB fMRI studies (Schafer et al., 2005), there is a little difference between medial region of the OB, probably because of individual differences revealed previously (Schafer et al., 2006). Generally, these properties of layer- and odorant-specific, high reproducibility and similar topography patterns for different concentrations of the same odorant were consistent with previous fMRI and intrinsic optical imaging studies (Lang et al., 2013; Xu et al., 2000, 2003). Layer-specific BOLD activation in the OB To examine and compare the BOLD signals in the deeper OB layers, the signals in each layer of each slice were extracted, unfolded, and reconstructed from contiguous coronal sections into the twodimensional odor map using home-made software (Liu et al., 2007),

as shown in Fig. 3A. Although the different map sizes made it difficult to statistically analyze the similarities and differences among these odor maps, it was obvious that these odor maps were quite similar. All pixels were selected and averaged from different layers to compare the BOLD signals in these layers. The numbers of averaged and the most active 30% of pixels in different layers are shown in Table 1. The most activated pixels were focused mainly in the GL. Among the three examined OB layers, the order of the signal intensity is GL N MCL N GCL (Figs. 3B–C), and all differences are significant (Fig. 3D). To assess whether this phenomenon was dependent on odor type or odor concentration, signals elicited by different odorants (IAA vs. OCT) and concentrations (L-IAA vs. H-IAA) were compared. It was found that the orders of the signal intensity among these three layers were the same, although the absolute intensity of the signals in a given layer was affected significantly by odor type and concentration (Figs. 3C–D).

LFP signals recorded simultaneously in the different OB layers To compare the layer specific responses between BOLD and underlying electrophysiological signals, the LFPs in different layers were recorded by linear 16-channel electrode arrays. To ensure the accuracies of the comparison and correlation of the BOLD and LFP signals, within or cross modality, the precise identification of the OB layers in the recording experiments were made as described in Materials and methods. The recording sites (GL, MCL and GCL) were determined by both the characteristic of LFP signals, the depth of the electrode and electric lesions, and then verified by MRI (Fig. 1, for details see Materials and methods). After the OB layers were assigned, the LFP signals in the different layers were analyzed. Both the signal distribution over different frequencies and the changes in temporal and magnitudinal domains contain rich neural information. Odor stimulation significantly increased the local neural activity, similar to the BOLD signals, as demonstrated by the enhanced power over the frequency range from 10 to 90 Hz (Fig. 4A). Analysis of the power spectrum revealed that in these three OB layers all neural activities were increased after odor stimulation, more concentrated in the L-gamma band, and reduced as stimulation was prolonged (Fig. 4B). Quantitative comparison of the strength of LFP responses from different layers revealed that the strongest responses were located at the GCL (Figs. 4C–D), which was different from the situation of BOLD signals where the strongest responses were in the GL (Fig. 3). This layer-specific response strength was independent of odor type, odor concentration and the frequency band of LFP (Figs. 4C–D).

Fig. 2. Activity maps induced by odorants in the OB. (A) The BOLD responses are visualized by overlapping the t-maps (1.5 b t b 6) to the corresponding anatomic images. Two coronal slices are shown from one rat, which had been repeatedly exposed to multiple odorants. H-IAA, left; L-IAA, central; and OCT, right. (B) Comparison of the activated patterns in different trials from different odorants or same odorants with similar concentrations. The activated patterns with similar signal to noise ratios and induced with similar concentrations were suitable to estimate the similarity of different odorants. The normalized dot product (NDP) of the activated pattern induced by L-IAA and OCT, respectively, were significantly higher than that between L-IAA and OCT (**p b 0.01). This indicated the different odorants evoked different activated patterns.

B. Li et al. / NeuroImage 95 (2014) 29–38

33

Fig. 3. Layer-dependent BOLD responses in the OB. (A) The global activity maps in the three layers elicited by IAA and OCT. (B) The averaged time courses of the BOLD response in the three layers with 32 s of repeated exposure (mean ± s.e; H-IAA, n = 9; L-IAA, n = 9; OCT, n = 8). (C) Statistical analyses of the BOLD responses. Line charts: there are significantly different responses among the three layers (ANOVA, p b 0.01); every layer-pair is significantly different (post hoc (LSD) tests, p b 0.01 for all comparisons). (D) Statistical analyses of the layernormalized BOLD responses to all odorants. The order of responses is GL N MCL N GCL (t-test, p b 0.01 for all pairs). Data processing: the three layers' responses were normalized by dividing the response of the MCL.

The characteristics and relationship of temporal response profiles of BOLD and LFP signals in the OB Despite different response intensities, the time courses of both BOLD and LFP in different layers were similar (Figs. 3B and 4B). R-values were calculated and used to assess the similarities of response profiles for BOLD and LFP signals in the different layers. Theoretically, a R N 0.7 is required for significant correlation (p b 0.05) of the BOLD time courses (10 points) within the different layers. Of the 324 R-values calculated from 9 animals, 100% (324/324) were significant. These R-values were classified, averaged and presented on pseudo-colored matrix (Fig. 5A, for details see Materials and methods). The R-values from the same and different layers were rather high and approximately equal, indicating high similarity in temporal domain among the different layers. This was confirmed by statistical analysis (t-test, p = 0.78, Fig. 5B, left first pair bars). Similar analyses were performed for the time courses of three frequency LFP bands (80 points, R N 0.3 is required for significant correlation p b 0.05). Over 99% (968/972) of the R-values were significant. The signals of each frequency band in different layers also had similar temporal features, indicated by the approximately equivalent R-values (Fig. 5B right three pairs of bars, p = 0.68, 0.49 and 0.75, for beta, L-gamma and H-gamma, respectively). The cross-band correlation analysis demonstrated that the signals of different frequency bands had different temporal features, suggested by the lower R-values for interband correlations (Fig. 5C, pixels in the left-lower triangle) than for the intra-band correlations (Fig. 5C, pixels in the right-upper triangle).

This was confirmed by statistical comparisons (t-test, p b 0.0001, Fig. 5D). Although the inter-band time courses were different because of the same changing direction in different frequency bands, 98% (319/324) of the R-values were significant (p b 0.05) in cross-band comparisons. The normalized time courses of the temporal differences among the different bands provide a detailed description (Fig. 6A, n = 9). For the three oscillations, the beta band was unique because the response decreased rapidly after an initially strong response, while decreases for L- and H-gamma were slower. To compare the time courses of LFP and BOLD signals, the temporal resolution of LFP signals was adjusted to be the same as the BOLD signals. Both types of signals reached the maxima at the first time point during odor stimulation (Fig. 6B), and the statistics showed no significant difference (ANOVA, p = 0.13, Fig. 6C). After onset of odor stimulation, all signals decreased with time, but the beta oscillation decreased significantly earlier after the onset of odor stimulation (Figs. 6B–C). With odor stimulation off, the Hgamma signals remained high for the next 8 s, and the BOLD signals fell under baseline, which was significantly different from any of the LFP bands (Figs. 6B–C). For the entire period of stimulation, the BOLD signal was better correlated with L-gamma oscillation during stimulation (Fig. 6D) and the observation was confirmed by quantitative analyses of all pairs (Fig. 6E, post hoc (LSD) test, p b 0.01 for paired comparison). The percentages of significant (p b 0.05) R-values were also calculated in a cross-modality comparison (BOLD × Beta, 72%; BOLD × L-gamma, 100%; BOLD × H-gamma, 86%). The order of the

Table 1 The numbers of total and the most active 30% pixels in different layers. Layer

GL MCL GCL

Total (n = 13)

954 ± 22 732 ± 15 543 ± 11

Most active 30% pixels (percentage of each layers) H-IAA (n = 9)

L-IAA (n = 9)

OCT (n = 8)

397 ± 11 (41.6 ± 1.0%) 184 ± 10 (25.0 ± 0.8%) 88 ± 6 (16.3 ± 1.1%)

383 ± 17 (39.8 ± 1.0%) 192 ± 8 (25.9 ± 0.9%) 101 ± 8 (18.5 ± 1.5%)

377 ± 18 (39.0 ± 0.8%) 194 ± 6 (26.5 ± 1.1%) 104 ± 6 (18.9 ± 1.0%)

In the second column, total numbers of pixels in the selected layers are shown. The values shown in the last three columns are the distribution of the most active 30% pixels, which were evoked by H-IAA, L-IAA and OCT respectively, in all selected pixels from three layers. This indicated that the three layers were all activated by the odorants, especially the response in GL which was the strongest.

34

B. Li et al. / NeuroImage 95 (2014) 29–38

Fig. 4. Odor induced LFP signals in three OB layers. (A) The LFP signals from a single trial with 32 s (thick black line) of IAA exposure in the GL (left), MCL (middle), and GCL (right). Row 1, the raw LFP; Row 2, the LFP through a band-pass filter from 13 to 90 Hz; and Row 3, enlarged view of 10 s of the filtered signals. (B) The normalized time-frequency graphs of LFP responses in the three OB layers, averaged from all rats. In all cases, the strongest response is observed in the GCL and the strongest frequency band is L-gamma. (C) Statistical analyses of the response to multiple odorants in different frequency bands. The responses in the GCL are the strongest (post hoc (LSD) tests, p b 0.01 for all bands), and the signals in the GL are stronger than MCL but not significant when they are compared using a single odorant (Post Hoc (LSD) tests, Beta, p = 0.53; L-gamma, p = 0.24; H-gamma, p = 0.44). (D) Statistical analyses of LFP responses to all odorants in the different layers. The order of responses is GCL N GL N MCL (t-test, p b 0.01 for all pairs). Data processing: the layers were normalized similar to the process for BOLD signals.

correlation between BOLD and LFP was L-gamma N H-gamma N Beta, which agreed with the statistical results (Fig. 6E).

signal correlated best with that of the LFP in the L-gamma frequency band.

Discussion

BOLD and LFP signal responses evoked by odor stimulation in different OB layers

The advantages of noninvasive measurements and high spatial and temporal resolutions have made BOLD-fMRI extensively used in neuroscience (Bandettini, 2012; DeSalvo et al., 2010; Logothetis, 2008; Sanganahalli et al., 2013). Since BOLD signal measures neural activity indirectly, correlation between BOLD signal and signals measuring neural activity directly has long been a fundamental topic for neuroimaging (Goense and Logothetis, 2008; Huttunen et al., 2008; Kahn et al., 2013; Kayser et al., 2009; Lippert et al., 2010; Logothetis et al., 2001; Shmuel et al., 2006; Sloan et al., 2010; Wang et al., 2012; Yen et al., 2011; Young et al., 2011). This study is the first attempt to correlate BOLD and LFP signals in the OB. The results revealed similar and distinct properties of the two types of signals in different OB layers. Odor stimulation evoked positive responses from both the BOLD and LFP in different frequency bands, while the strongest signals were located in the GL for BOLD and GCL for LFP, respectively. The temporal profile of the BOLD

The different layers of the OB, which are well defined according to the cellular architecture, play different roles in odor information coding, processing and transmitting. For example, the GL, receiving all olfactory information from the olfactory epithelium, contains large amount of synapses and a high density of capillaries. The MCL contains the bodies of mitral cells, which send the processed information to higher olfactory centers (Restrepo et al., 2009; Wilson and Mainen, 2006). The GCL contains the bodies of granular cells which are known to regulate the information process in the OB and the transmission to the olfactory cortexes via the dendro-dendritic networks in the external plexiform layer formed by the dendrites of granular cells and the lateral dendrites of mitral cells (Schoppa and Urban, 2003). The activity patterns in the GL have been extensively studied by a variety of techniques including 2-deoxyglucose (Johnson and Leon, 2007; Johnson et al., 1999, 2009), intrinsic optical imaging (Igarashi and Mori, 2005; Lang et al., 2013;

B. Li et al. / NeuroImage 95 (2014) 29–38

35

Fig. 5. The similarity of BOLD and LFP response profiles in the different layers. (A) Similarity of BOLD response profiles in the different layers. Almost all pixels, either in the left-bottom triangle for the correlation between different layers or in the right-top triangle for the correlation between the same layers, have high and similar R-values. This indicated that the response patterns in different layers are similar. The correlation coefficients from self-paired animals were ignored. (B) Statistical analyses showing the similarity of response profiles for both BOLD and different frequency LFP bands in the different layers (t-test, p = 0.78, 0.68, 0.49 and 0.75 for BOLD, Beta, L-gamma and H-gamma, respectively; n = 36). (C) Band-cross similarity comparison of LFP signals. Most pixels in the left-bottom triangle for the correlation between different bands have smaller R-values than these in the right-top triangle for the correlation between the same bands, indicating that the response patterns in different frequency bands are different. (D) Statistical analyses of the averaged R-values from the triangles for different (green) and same (red) frequency bands in (C) (t-test, p b 0.01, n = 36).

Mori et al., 2006; Takahashi et al., 2004) and fMRI (Xu et al., 2000, 2003), and has been proved to be the basis for olfactory behavior (Cleland et al., 2007; Mandairon et al., 2006). However, few studies have focused on the deep OB layer, due to a lack of proper techniques. By taking advantage of BOLD-fMRI, the activity patterns of the GL, MCL and GCL induced by odorants were obtained in this study, and found to be similar in topography and strongest in the GL (Fig. 3A), similar to previous studies with 2-deoxyglucose (Johnson et al., 1999) and simulation (Nawroth et al., 2007). As functional units for olfaction, columnar structures that radially extend across all layers of the OB have been revealed by recent viral tracing studies (Willhite et al., 2006). Therefore, the similar topography patterns in the different OB layers demonstrated that anatomical connectivity results in functional connectivity. The major factors affecting BOLD signals are blood supply structure, flow and volume, and glucose and oxygen usages (Logothetis, 2008). The capillary density and the levels of metabolism related enzymes in the OB are in the order of GL N MCL N GCL (Borowsky and Collins, 1989), as are all matching general neural activity and energy consumptions. Indeed, our results showed that the response intensities in these layers are in the same order (Figs. 3B–D). Other than demonstrating the response intensity, our data also revealed that in all OB layers the BOLD signals declined similarly with time (Figs. 5A–B). Adaptation is a well-known neuronal process and has been shown to be found in the OB by different studies (Chaudhury et al., 2010; Lecoq et al., 2009; Schafer et al., 2005). Therefore, our results provide further evidence that these layers are anatomically and functionally connected, and act coherently to external stimulations. Taking all the magnitudinal and

temporal features together, the results are consistent with the known neurophysiological basis for BOLD signals. The parallel experiments using electrical recording in the same animals revealed that the LFP signals evoked by odorants in all three frequency bands, beta, and L-gamma and H-gammas, in all three OB layers were in the order of GCL N GL N MCL (Figs. 4B–D). The particular spatial organization and dynamic evolution of current sinks and sources in a certain neuronal group is the generator of the LFP (Buzsaki et al., 2012). In the OB, the reciprocal dendro-dendritic networks between mitral and granular cells play a critical role in generating LFP oscillations (Kay et al., 2009; Rojas-Libano and Kay, 2008). However, the intensity of the LFP is mainly determined by the local neural activity, the parallelism of the fibers, the neuron density, and the geometrical arrangements of neurons (Buzsaki et al., 2012; Rasch et al., 2009). The GCL with the superior arrangements of fibers and the largest neuron density, conforming to the factors referred above, should generate the strongest LFP signals in the OB, which is in agreement with our observations. Not surprisingly, the current source density analysis has revealed that the strongest LFP locates in the GCL by electrical stimulation of the olfactory nerve (Neville and Haberly, 2003). Similar results have also been reported in other studies (Cenier et al., 2009; Gervais et al., 2007; Li et al., 2010, 2011). Temporal correlations between BOLD-fMRI and LFP signals Temporally, the LFP signals in the three frequency bands have different time courses, but are similar for a given band in different layers

36

B. Li et al. / NeuroImage 95 (2014) 29–38

Fig. 6. The response dynamics of LFP and BOLD and their correlation in the OB. (A) The normalized time courses of the LFP signals averaged from all tested animals show that the three bands have different temporal features. (B) Comparison of the normalized time courses of BOLD and different LFP bands. The points marked by black arrows represent four phases: odor onset, odor period, odor off and recovery. (C) Statistical analyses of the temporal similarity and difference between BOLD and LFP signals at the four phases in (B), post hoc (LSD) tests, **p b 0.01. (D) Similarity and difference between the time courses of BOLD with three different LFP bands. The color matrices of the R-values calculated from the BOLD response profile and one of the three frequency bands demonstrate that the R-values for the BOLD and L-gamma are the highest, as confirmed by statistical analyses in (E) post hoc (LSD) tests, p b 0.01 for all pairs (n = 81).

(Figs. 5C–D). This is logical because different bands of the LFP in the OB have different resources and thus should reflect different neural processes which have different temporal features, although their exact roles in olfaction are still barely understood (Kay et al., 2009; Neville and Haberly, 2003). In the three OB layers, the same intensity order and the highly similar time courses for each of the three LFP bands provided further evidence that the anatomically connected layers function coherently, as shown by the BOLD-fMRI results above. In the present study, BOLD and LFP signals were recorded across different OB layers under the same experimental condition, and are, therefore, directly comparable. Odor stimulation increased responses in all frequency bands of the LFP and BOLD signals, indicating the electrophysiological and BOLD signals both reflect changes of neural activities in all three examined layers. However, detailed analysis revealed significant differences between the two modalities, including (1) the layer with the strongest responses (Figs. 3D vs 4D); (2) the orders of response intensity in the three OB layers (Figs. 3C–D vs 4B–D); and (3) the temporal features (Fig. 6B). The differences in the intensity domain reflected that the BOLD and LFP signals are affected by different factors, as discussed above, but are masked when the spatial resolution is low. That is, if the OB layers are collapsed to one pixel (~ 1 × 1 mm2), the two types of signals would be well correlated. However, different frequency bands of the LFP showed different time courses (Figs. 6A–B) and for prolonged stimulation, the L-gamma band correlated the highest with the BOLD signal, and the beta band had the worst correlation (Figs. 6B, D–E). Previous studies on the origin of the LFP in different

frequency bands in the OB have suggested that the gamma oscillations reflect local neural activity in the bulb, while beta oscillations reflect global neural activity across the olfactory system (Kay et al., 2009; Neville and Haberly, 2003). Therefore, our present study indicates that the BOLD signals reflect more local than global neural activity, consistent with studies on the visual system (Logothetis et al., 2001; Magri et al., 2012; Maier et al., 2008; Viswanathan and Freeman, 2007). The use of fMRI technology in awake rats has been reported (Febo, 2011; Martin et al., 2013). Therefore, it would be interesting to test how the BOLD and LFP signals correlate in the OB under an awake state in future studies, especially with the simultaneous registration of both BOLD and LFP signals, as reported in other brain regions (Huttunen et al., 2008; Lippert et al., 2010; Logothetis et al., 2001). Interpretation of BOLD signals BOLD signal changes originate from changes of local neural activity, but are affected by changes in hemodynamics, metabolism, and neuroanatomy, and thus reflect neural activity only indirectly (Bailey et al., 2013; Logothetis, 2008; Logothetis and Wandell, 2004; Mishra et al., 2011). Making the situation more complicated is that different researchers might have different definitions of “neural activity”. It has been demonstrated by the simultaneous registration of BOLD and electrical recordings that BOLD is better correlated with LFP signals than single/multiple unit firing (Logothetis, 2008; Logothetis and Wandell, 2004; Logothetis et al., 2001). However, BOLD and LFP signals originate

B. Li et al. / NeuroImage 95 (2014) 29–38

from overlapping but different sources, and are affected by common as well as specific factors. Thus, these signals generally agree with each other at lower spatial resolution. Under high spatial resolution as in this study, the differences between them can be readily differentiated. Within a small region (1 mm2), the locations for the highest responses, as well as the intensity order across different layers were rather different. Therefore, one must be cautious with interpreting the BOLD signals when the change is related to the corresponding neural activity (such as the LFP), especially in studies under low spatial resolution. For a given small region (e.g., a glomerulus), the degree of BOLD changes should be able to be correlated to the neural activity there, with the larger the BOLD changes, the stronger the activation. This is especially true in comparing the maps of the same odor with different concentrations in this study (Figs. 3B–C and 4C). However, for different areas (like a pixel in the GL and another pixel in the MCL), there are different scenarios. For example, the same BOLD change does not mean the same neural activity change; the one with larger increase in BOLD signal might have smaller change in neural activity, even a decreased local neural activity (as in the case of different anesthesia levels, unpublished data). The discrepancy between BOLD and LFP signals revealed in the present study suggests that other signals measured by fMRI, i.e. cerebral blood volume (CBV) and blood flow (CBF), which in turn have been used to calculate changes in oxidative metabolism (CMRO2) with a calibrated fMRI, are probably better correlated with LFP signals than with BOLD. In a recent study, LFP, multiunit activity (spikes), BOLD, CBV, CBF and CMRO2 were compared across different layers of the somatosensory cortex with electrical forepaw stimulation (Herman et al., 2013). It revealed that although the BOLD and CBV signals were not correlated well with either LFP or multiunit activity, changes in CMRO2 and multiunit activity were strongly correlated across layers. Therefore, the correlation between LFP and other fMRI signals in the different layers of the OB is an open and interesting question for future studies. Conclusions In summary, using high-resolution fMRI and a micro-array electrode, the layer specific activation in the OB evoked by an odor were examined. Our study revealed that the magnitudinal and temporal relationships between the BOLD and LFP signals are rather complex. The two modalities of signals can be correlated: strongly and positively, weakly and positively, not significantly, and even strongly and negatively. These results provide new evidence for correlation of structure and function, reveal the relationship between BOLD and LFP signals, and help us to understand and interpret BOLD-fMRI and electrical recording data in future. Acknowledgments This work was supported by the National Natural Science Foundation of China (NSFC, 91132307/H09, 31171061/C090208 and 20921004), the Chinese Academy of Sciences (XDB02050005) and the Ministry of Science and Technology of China (2012BAI23B02) grants to F. Xu and the NSFC (31100799) grant to A. Li. References Alonso, M., Lepousez, G., Sebastien, W., Bardy, C., Gabellec, M.M., Torquet, N., Lledo, P.M., 2012. Activation of adult-born neurons facilitates learning and memory. Nat. Neurosci. 15, 897–904. Bailey, C.J., Sanganahalli, B.G., Herman, P., Blumenfeld, H., Gjedde, A., Hyder, F., 2013. Analysis of time and space invariance of BOLD responses in the rat visual system. Cereb. Cortex 23, 210–222. Bandettini, P.A., 2012. Twenty years of functional MRI: the science and the stories. Neuroimage 62, 575–588. Borowsky, I.W., Collins, R.C., 1989. Metabolic anatomy of brain: a comparison of regional capillary density, glucose metabolism, and enzyme activities. J. Comp. Neurol. 288, 401–413. Buzsaki, G., 2010. Neural syntax: cell assemblies, synapsembles, and readers. Neuron 68, 362–385.

37

Buzsaki, G., Anastassiou, C.A., Koch, C., 2012. The origin of extracellular fields and currents—EEG, ECoG, LFP and spikes. Nat. Rev. Neurosci. 13, 407–420. Cenier, T., David, F., Litaudon, P., Garcia, S., Amat, C., Buonviso, N., 2009. Respiration-gated formation of gamma and beta neural assemblies in the mammalian olfactory bulb. Eur. J. Neurosci. 29, 921–930. Chaudhury, D., Manella, L., Arellanos, A., Escanilla, O., Cleland, T.A., Linster, C., 2010. Olfactory bulb habituation to odor stimuli. Behav. Neurosci. 124, 490–499. Cleland, T.A., Johnson, B.A., Leon, M., Linster, C., 2007. Relational representation in the olfactory system. Proc. Natl. Acad. Sci. U. S. A. 104, 1953–1958. DeSalvo, M.N., Schridde, U., Mishra, A.M., Motelow, J.E., Purcaro, M.J., Danielson, N., Bai, X., Hyder, F., Blumenfeld, H., 2010. Focal BOLD fMRI changes in bicuculline-induced tonic–clonic seizures in the rat. Neuroimage 50, 902–909. Doucette, W., Restrepo, D., 2008. Profound context-dependent plasticity of mitral cell responses in olfactory bulb. PLoS Biol. 6, e258. Doucette, W., Gire, D.H., Whitesell, J., Carmean, V., Lucero, M.T., Restrepo, D., 2011. Associative cortex features in the first olfactory brain relay station. Neuron 69, 1176–1187. Febo, M., 2011. Technical and conceptual considerations for performing and interpreting functional MRI studies in awake rats. Front. Psychiatry 2, 43. Gao, J.H., Liu, H.L., 2012. Inflow effects on functional MRI. Neuroimage 62, 1035–1039. Gervais, R., Buonviso, N., Martin, C., Ravel, N., 2007. What do electrophysiological studies tell us about processing at the olfactory bulb level? J. Physiol. Paris 101, 40–45. Gheusi, G., Cremer, H., McLean, H., Chazal, G., Vincent, J.D., Lledo, P.M., 2000. Importance of newly generated neurons in the adult olfactory bulb for odor discrimination. Proc. Natl. Acad. Sci. U. S. A. 97, 1823–1828. Goense, J.B., Logothetis, N.K., 2008. Neurophysiology of the BOLD fMRI signal in awake monkeys. Curr. Biol. 18, 631–640. Gurden, H., Uchida, N., Mainen, Z.F., 2006. Sensory-evoked intrinsic optical signals in the olfactory bulb are coupled to glutamate release and uptake. Neuron 52, 335–345. Herman, P., Sanganahalli, B.G., Blumenfeld, H., Rothman, D.L., Hyder, F., 2013. Quantitative basis for neuroimaging of cortical laminae with calibrated functional MRI. Proc. Natl. Acad. Sci. U. S. A. 110, 15115–15120. Huttunen, J.K., Grohn, O., Penttonen, M., 2008. Coupling between simultaneously recorded BOLD response and neuronal activity in the rat somatosensory cortex. Neuroimage 39, 775–785. Hyder, F., Rothman, D.L., Shulman, R.G., 2002. Total neuroenergetics support localized brain activity: implications for the interpretation of fMRI. Proc. Natl. Acad. Sci. U. S. A. 99, 10771–10776. Igarashi, K.M., Mori, K., 2005. Spatial representation of hydrocarbon odorants in the ventrolateral zones of the rat olfactory bulb. J. Neurophysiol. 93, 1007–1019. Johnson, B.A., Leon, M., 2007. Chemotopic odorant coding in a mammalian olfactory system. J. Comp. Neurol. 503, 1–34. Johnson, B.A., Woo, C.C., Hingco, E.E., Pham, K.L., Leon, M., 1999. Multidimensional chemotopic responses to n-aliphatic acid odorants in the rat olfactory bulb. J. Comp. Neurol. 409, 529–548. Johnson, B.A., Xu, Z., Ali, S.S., Leon, M., 2009. Spatial representations of odorants in olfactory bulbs of rats and mice: similarities and differences in chemotopic organization. J. Comp. Neurol. 514, 658–673. Kahn, I., Knoblich, U., Desai, M., Bernstein, J., Graybiel, A.M., Boyden, E.S., Buckner, R.L., Moore, C.I., 2013. Optogenetic drive of neocortical pyramidal neurons generates fMRI signals that are correlated with spiking activity. Brain Res. 1511, 33–45. Kay, L.M., Beshel, J., Brea, J., Martin, C., Rojas-Libano, D., Kopell, N., 2009. Olfactory oscillations: the what, how and what for. Trends Neurosci. 32, 207–214. Kayser, C., Petkov, C.I., Logothetis, N.K., 2009. Multisensory interactions in primate auditory cortex: fMRI and electrophysiology. Hear. Res. 258, 80–88. Lang, J., Li, A., Luo, W., Wu, R., Li, P., Xu, F., 2013. Odor representation in the olfactory bulb under different brain states revealed by intrinsic optical signals imaging. Neuroscience 243, 54–63. Lazarini, F., Lledo, P.M., 2011. Is adult neurogenesis essential for olfaction? Trends Neurosci. 34, 20–30. Lecoq, J., Tiret, P., Charpak, S., 2009. Peripheral adaptation codes for high odor concentration in glomeruli. J. Neurosci. 29, 3067–3072. Li, A.A., Gong, L., Liu, Q., Li, X., Xu, F., 2010. State-dependent coherences between the olfactory bulbs for delta and theta oscillations. Neurosci. Lett. 480, 44–48. Li, A., Gong, L., Xu, F., 2011. Brain-state-independent neural representation of peripheral stimulation in rat olfactory bulb. Proc. Natl. Acad. Sci. U. S. A. 108, 5087–5092. Li, A., Zhang, L., Liu, M., Gong, L., Liu, Q., Xu, F., 2012. Effects of different anesthetics on oscillations in the rat olfactory bulb. J. Am. Assoc. Lab. Anim. Sci. 51, 458–463. Lippert, M.T., Steudel, T., Ohl, F., Logothetis, N.K., Kayser, C., 2010. Coupling of neural activity and fMRI-BOLD in the motion area MT. Magn. Reson. Imaging 28, 1087–1094. Liu, N., Xu, F., Miller, P.L., Shepherd, G.M., 2007. OdorMapComparer: an application for quantitative analyses and comparisons of fMRI brain odor maps. Neuroinformatics 5, 105–114. Liu, H.L., Wei, P.S., Wai, Y.Y., Kuan, W.C., Huang, C.M., Wu, C.W., Buckle, C., Wan, Y.L., Gao, J.H., 2008. Inflow effects on hemodynamic responses characterized by event-related fMRI using gradient-echo EPI sequences. Med. Phys. 35, 4300–4307. Logothetis, N.K., 2008. What we can do and what we cannot do with fMRI. Nature 453, 869–878. Logothetis, N.K., Wandell, B.A., 2004. Interpreting the BOLD signal. Annu. Rev. Physiol. 66, 735–769. Logothetis, N.K., Pauls, J., Augath, M., Trinath, T., Oeltermann, A., 2001. Neurophysiological investigation of the basis of the fMRI signal. Nature 412, 150–157. Maandag, N.J., Coman, D., Sanganahalli, B.G., Herman, P., Smith, A.J., Blumenfeld, H., Shulman, R.G., Hyder, F., 2007. Energetics of neuronal signaling and fMRI activity. Proc. Natl. Acad. Sci. U. S. A. 104, 20546–20551.

38

B. Li et al. / NeuroImage 95 (2014) 29–38

Magri, C., Schridde, U., Murayama, Y., Panzeri, S., Logothetis, N.K., 2012. The amplitude and timing of the BOLD signal reflects the relationship between local field potential power at different frequencies. J. Neurosci. 32, 1395–1407. Maier, A., Wilke, M., Aura, C., Zhu, C., Ye, F.Q., Leopold, D.A., 2008. Divergence of fMRI and neural signals in V1 during perceptual suppression in the awake monkey. Nat. Neurosci. 11, 1193–1200. Mandairon, N., Stack, C., Kiselycznyk, C., Linster, C., 2006. Broad activation of the olfactory bulb produces long-lasting changes in odor perception. Proc. Natl. Acad. Sci. U. S. A. 103, 13543–13548. Martin, C., Grenier, D., Thevenet, M., Vigouroux, M., Bertrand, B., Janier, M., Ravel, N., Litaudon, P., 2007. fMRI visualization of transient activations in the rat olfactory bulb using short odor stimulations. Neuroimage 36, 1288–1293. Martin, C.J., Kennerley, A.J., Berwick, J., Port, M., Mayhew, J.E., 2013. Functional MRI in conscious rats using a chronically implanted surface coil. J. Magn. Reson. Imaging 38 (3), 739–744. Mishra, A.M., Ellens, D.J., Schridde, U., Motelow, J.E., Purcaro, M.J., DeSalvo, M.N., Enev, M., Sanganahalli, B.G., Hyder, F., Blumenfeld, H., 2011. Where fMRI and electrophysiology agree to disagree: corticothalamic and striatal activity patterns in the WAG/Rij rat. J. Neurosci. 31, 15053–15064. Mori, K., Takahashi, Y.K., Igarashi, K.M., Yamaguchi, M., 2006. Maps of odorant molecular features in the Mammalian olfactory bulb. Physiol. Rev. 86, 409–433. Mullinger, K.J., Mayhew, S.D., Bagshaw, A.P., Bowtell, R., Francis, S.T., 2013. Poststimulus undershoots in cerebral blood flow and BOLD fMRI responses are modulated by poststimulus neuronal activity. Proc. Natl. Acad. Sci. U. S. A. 110 (33), 13636–13641. Nawroth, J.C., Greer, C.A., Chen, W.R., Laughlin, S.B., Shepherd, G.M., 2007. An energy budget for the olfactory glomerulus. J. Neurosci. 27, 9790–9800. Neville, K.R., Haberly, L.B., 2003. Beta and gamma oscillations in the olfactory system of the urethane-anesthetized rat. J. Neurophysiol. 90, 3921–3930. Panagiotaropoulos, T.I., Deco, G., Kapoor, V., Logothetis, N.K., 2012. Neuronal discharges and gamma oscillations explicitly reflect visual consciousness in the lateral prefrontal cortex. Neuron 74, 924–935. Petzold, G.C., Albeanu, D.F., Sato, T.F., Murthy, V.N., 2008. Coupling of neural activity to blood flow in olfactory glomeruli is mediated by astrocytic pathways. Neuron 58, 897–910. Rasch, M., Logothetis, N.K., Kreiman, G., 2009. From neurons to circuits: linear estimation of local field potentials. J. Neurosci. 29, 13785–13796. Restrepo, D., Doucette, W., Whitesell, J.D., McTavish, T.S., Salcedo, E., 2009. From the top down: flexible reading of a fragmented odor map. Trends Neurosci. 32, 525–531. Rojas-Libano, D., Kay, L.M., 2008. Olfactory system gamma oscillations: the physiological dissection of a cognitive neural system. Cogn. Neurodyn. 2, 179–194. Sanganahalli, B.G., Herman, P., Behar, K.L., Blumenfeld, H., Rothman, D.L., Hyder, F., 2013. Functional MRI and neural responses in a rat model of Alzheimer's disease. Neuroimage 79, 404–411. Schafer, J.R., Kida, I., Rothman, D.L., Hyder, F., Xu, F., 2005. Adaptation in the rodent olfactory bulb measured by fMRI. Magn. Reson. Med. 54, 443–448.

Schafer, J.R., Kida, I., Xu, F., Rothman, D.L., Hyder, F., 2006. Reproducibility of odor maps by fMRI in rodents. Neuroimage 31, 1238–1246. Schoppa, N.E., Urban, N.N., 2003. Dendritic processing within olfactory bulb circuits. Trends Neurosci. 26, 501–506. Shepherd, G.M., Charpak, S., 2008. The olfactory glomerulus: a model for neuro-gliovascular biology. Neuron 58, 827–829. Shipley, M.T., Ennis, M., 1996. Functional organization of olfactory system. J. Neurobiol. 30, 123–176. Shmuel, A., Augath, M., Oeltermann, A., Logothetis, N.K., 2006. Negative functional MRI response correlates with decreases in neuronal activity in monkey visual area V1. Nat. Neurosci. 9, 569–577. Sloan, H.L., Austin, V.C., Blamire, A.M., Schnupp, J.W., Lowe, A.S., Allers, K.A., Matthews, P.M., Sibson, N.R., 2010. Regional differences in neurovascular coupling in rat brain as determined by fMRI and electrophysiology. Neuroimage 53, 399–411. Takahashi, Y.K., Kurosaki, M., Hirono, S., Mori, K., 2004. Topographic representation of odorant molecular features in the rat olfactory bulb. J. Neurophysiol. 92, 2413–2427. Viswanathan, A., Freeman, R.D., 2007. Neurometabolic coupling in cerebral cortex reflects synaptic more than spiking activity. Nat. Neurosci. 10, 1308–1312. Wang, L., Saalmann, Y.B., Pinsk, M.A., Arcaro, M.J., Kastner, S., 2012. Electrophysiological low-frequency coherence and cross-frequency coupling contribute to BOLD connectivity. Neuron 76, 1010–1020. Willhite, D.C., Nguyen, K.T., Masurkar, A.V., Greer, C.A., Shepherd, G.M., Chen, W.R., 2006. Viral tracing identifies distributed columnar organization in the olfactory bulb. Proc. Natl. Acad. Sci. U. S. A. 103, 12592–12597. Wilson, R.I., Mainen, Z.F., 2006. Early events in olfactory processing. Annu. Rev. Neurosci. 29, 163–201. Xu, F., Kida, I., Hyder, F., Shulman, R.G., 2000. Assessment and discrimination of odor stimuli in rat olfactory bulb by dynamic functional MRI. Proc. Natl. Acad. Sci. U. S. A. 97, 10601–10606. Xu, F., Liu, N., Kida, I., Rothman, D.L., Hyder, F., Shepherd, G.M., 2003. Odor maps of aldehydes and esters revealed by functional MRI in the glomerular layer of the mouse olfactory bulb. Proc. Natl. Acad. Sci. U. S. A. 100, 11029–11034. Xu, F., Schaefer, M., Kida, I., Schafer, J., Liu, N., Rothman, D.L., Hyder, F., Restrepo, D., Shepherd, G.M., 2005. Simultaneous activation of mouse main and accessory olfactory bulbs by odors or pheromones. J. Comp. Neurol. 489, 491–500. Yang, X., Renken, R., Hyder, F., Siddeek, M., Greer, C.A., Shepherd, G.M., Shulman, R.G., 1998. Dynamic mapping at the laminar level of odor-elicited responses in rat olfactory bulb by functional MRI. Proc. Natl. Acad. Sci. U. S. A. 95, 7715–7720. Yen, C.C., Fukuda, M., Kim, S.G., 2011. BOLD responses to different temporal frequency stimuli in the lateral geniculate nucleus and visual cortex: insights into the neural basis of fMRI. Neuroimage 58, 82–90. Young, C.K., Brown, A.R., Robinson, J.H., Tuor, U.I., Dunn, J.F., Bland, B.H., Teskey, G.C., 2011. Functional MRI response and correlated electrophysiological changes during posterior hypothalamic nucleus deep brain stimulation. Neuroimage 56, 35–44.