NeuroImage 60 (2012) 426–435

Contents lists available at SciVerse ScienceDirect

NeuroImage

j ourna l homepage: www.e lsev ie r .com/ locate /yn img

Taste laterality studied by means of umami and salt stimuli: An fMRI study

Emilia Iannilli a,⁎, P. Bano Singh a,b, Benno Schuster a, Johannes Gerber c, Thomas Hummel a

a Smell & Taste Clinic, Department of Otorhinolaryngology, University of Dresden Medical School (“Technische Universität Dresden”), Dresden, Germanyb Department of Oral Biology, Faculty of Dentistry, University of Oslo, Oslo, Norwayc Department of Neuroradiology, University of Dresden Medical School, Germany

Abbreviations: BOLD, blood oxygenation level depenic resonance; fO, frontal operculum; I, insula; MSG, mondium chloride; OFC, orbitofrontal cortex; PET, positroprefrontal cortex; ROI, region of interest; BA, Brodmann⁎ Corresponding author at: Smell & Taste Clinic, Depar

University of Dresden Medical School, Fetscherstrasse 7E-mail address: Emilia.Iannilli@uniklinikum-dresden

1053-8119/$ – see front matter © 2012 Elsevier Inc. Alldoi:10.1016/j.neuroimage.2011.12.088

a b s t r a c t

a r t i c l e i n f o

Article history:Received 14 October 2011Revised 21 December 2011Accepted 27 December 2011Available online 8 January 2012

Keywords:Functional imagingDynamic causal modelingGustatory afferentsGustatory cortexGustometerThalamus

Aim of the present study was to investigate laterality of the gustatory system in the human brain for the tastequalities elicited by MSG (monosodium glutamate) and NaCl (sodium chloride). A total of 23 subjects partic-ipated in a block-design functional magnetic resonance imaging (fMRI) study. Liquid stimuli were presentedat supra-threshold concentrations and delivered by means of a computer controlled gustometer. Left andright sides of the mouth were stimulated separately in order to correlate statistical parametrical maps toboth the site of the stimulus and the specific taste quality. Following the effects of the site of stimulationthrough primary and secondary gustatory cortices an effort was made to explore the laterality of the gusta-tory pathways. Our results showed for both tastants a predominance of ipsilateral connections at the thala-mus level. Insula left and right regions were both involved for both tastants. In these regions we found ahigh proportion of ipsilateral connection again for both the tastants. Considering orbitofrontal/prefrontal cor-tex, left-sided stimulation with NaCl or MSG produced left-sided activation of the orbito-frontal cortex clearlyindicating the presence of an ipsilateral path. Finally, the hypothesis of frontal operculum as primary gusta-tory cortex and lateral prefrontal cortex as secondary was also supported by the results from dynamic causalmodeling.

© 2012 Elsevier Inc. All rights reserved.

Introduction

The knowledge of the cerebral gustatory pathways in humans isbased mainly on anatomical dissections and in vivo electrophysiologyin animals (Rolls and Scott, 2003; Simon et al., 2006). The taste budslocated in the oral cavity and the pharynxmediate the transduction offive taste qualities, sour, sweet, bitter, salty and umami. It is a well-known fact that the primary gustatory afferents from the taste recep-tor cells project ipsilaterally to the nucleus of the solitary tract (Gotoet al., 1983; Jyoichi et al., 1985; Nakajima et al., 1983). However, thepathway from the secondary neurons to the gustatory cortex inhumans is not yet established (Kobayashi, 2006). In fact, some papersreported an ipsilateral projection from the nucleus of the solitary tractto the thalamic nuclei (Landis et al., 2006; Shikama et al., 1996;Uesaka et al., 1998), or described a contralateral connection throughthe thalamus (Fujikane et al., 1998; Lee et al., 1998; Onoda andIkeda, 1999), while Aglioti et al. (2000, 2001) suggested a bilateralpathway with ipsilateral predominance.

dent; fMRI, functional magnet-osodium glutamate; NaCl, so-n emission tomography; PFC,'s area.tment of Otorhinolaryngology,5, Dresden 01307, Germany..de (E. Iannilli).

rights reserved.

Regarding the cerebral processing of gustatory information it iswell established in primates that the primary taste cortex is locatedinside the anterior insula/frontal operculum (I/fO) (Haase et al.,2009; Rolls and Scott, 2003; Rolls et al., 1996; Scott et al., 1986;Yaxley et al., 1990) and the secondary cortical taste area is situatedin the caudolateral orbitofrontal cortex (OFC) (Rolls et al., 1990). Sim-ilar relations seem to be present in humans but the definition as wellas the positions of primary and secondary taste cortices is not yet def-initely clarified. Using positron emission tomography (PET) Small etal. (1997a) reported taste induced activations of the fO and bilateralOFC. Frey and Petrides (1999) showed a bilateral activation in the I/fO. Kinomura et al. (1994) among several regions, found activationsin the insula. By means of functional magnetic resonance imaging(fMRI) de Araujo et al. (2003a) found that stimuli such as sucroseand umami activated areas inside the I/fO and caudolateral OFC.Schoenfeld et al. (2004), found taste-activated brain areas close tothe ones mentioned above, with stable activation patterns during re-peated stimulation. According to Barry et al. (2001) electric tastestimulation of the tongue activated I/fO, which has also been shownin numerous other studies using natural stimuli (de Araujo et al.,2003b; Faurion et al., 1998; Haase et al., 2009; Small et al., 1997b).On the other hand magnetoencephalography data published byOgawa, Kobayakawa and their colleagues have argued that humanprimary taste cortex is in the parietal operculum (Kobayakawa etal., 1996, 1999; Ogawa et al., 2005) as well as a PET study made byFrey and Petrides (1999) the taste response was observed in multiple

427E. Iannilli et al. / NeuroImage 60 (2012) 426–435

regions of insula and overlying operculum but not in the anteriorInsula of fO. Further details about the debate on the definitions ofthe primary and secondary gustatory cortices are extensively exposedin Small (2010) but also in Veldhuizen et al. (2011).

Because of the open question of the lateralization of gustatory in-formation processing, we wanted to study the laterality of the gusta-tory pathway in relation to salty (NaCl) and umami (monosodiumglutamate: MSG) stimuli using an fMRI approach. While everybodyis familiar with salt, the taste of umami (Ikeda, 1909) is the taste ofproteins. Umami is present in palatable foods such as meat, fish, to-matoes, mushrooms, and dairy products and it has been shown tostimulate food intake in mammals (Prescott, 2004; Yamaguchi,1991). In the present study we tried to clarify whether there is later-alization in the processing of gustatory information dependent ontaste quality. Specifically, we applied salty and umami taste in liquidform to the left and right sides of the tongue/oral cavity, and werecorded the subsequent brain activity by means of fMRI. Based on aregion of interest (ROI)-analysis we followed brain activations alongthe ‘taste map’, as defined in Veldhuizen et al. (2011) in regionssuch as thalamus, I/fO, orbitofrontal cortex (OFC), prefrontal cortex(PFC) and limbic system. The results should guide us to depict a hypo-thetical connectivity among brain areas involved with the lateralizedtaste stimulus that we intended to asses using a formal analytic meth-od, Dynamic casual modeling (DCM) (Friston et al., 2003).

Materials and methods

Participants

Twenty-three healthy, right-handed volunteers participated inthis study (13 women; mean age±standard error=28.3±1.4 years). The study was approved by the Ethics Committee ofthe Technical University of Dresden Medical School. Written in-formed consent was obtained from all subjects prior to theexperiment.

Stimuli

Two tastants, NaCl and MSG, were used in diluted form. Stimuliwere presented at supra-threshold concentrations of 50 mM forboth NaCl and MSG, based on previous studies performed on humanumami taste determination (Singh et al., 2010) (de Araujo et al.,2003a; Lugaz et al., 2002; McCabe and Rolls, 2007). Stimulation wasrestricted to the lateral parts of the tongue and mouth, more than2 cm distant from the anterior tip of the tongue. Each side was stim-ulated separately.

Taste delivery system: the gustometer

The liquid stimuli were delivered by means of a computer con-trolled gustometer (Burghart GU002 — variant GM04; Burghart in-struments, Wedel, Germany). The gustometer was placed outsidethe scanner room; tubings for stimulation were funneled through adedicated opening into the scanner room. Stimuli were delivered ina pulse design (total pulse volume 1 ml, total pulse duration 3.3 s)at room temperature (24 °C) through two Teflon™ tubes placed in-side the subject's mouth. Apart from taste, stimulation was void ofany cues that would have made subjects aware of the stimulusonset. The small dimension (1.3 mm inner diameter, 1.5 mm outer di-ameter) made it possible for the tubes to be placed in a comfortablemanner inside the mouth. The tubes were positioned on the lateraledges of the tongue in order to separately stimulate the sensorycells in the taste buds located on each side of the tongue.

Olfactory and gustatory screening

The olfactory function of all subjects was investigated using thevalidated “Sniffin' Sticks” test (Hummel et al., 2007; Kobal et al.,1996). The subjects included in the study demonstrated a normalsense of smell. Similarly, the gustatory function of the subjects wasevaluated by standardized taste test kit, the “taste strips” (Landis etal., 2009; Mueller et al., 2003; Schuster et al., 2009). All subjects in-cluded in this study had test scores within the normal range.

Experimental procedure

The fMRI paradigm was built in a 50 s-block design (Fig. 1(b)).According to visually presented information on an MRI-room dedicat-ed screen, subjects were told to maintain the tastants in the mouth aslong as a gray field was visible. Then the subjects received the infor-mation “swallow” and after 5 s the information “rinse” was given to-gether with 1 ml of water (10.8 s of duration). As evidence that theliquid tastant was maintained to the side of the mouth we did a sim-ulation of the experiment using a food coloring (blue) diluted inwater, with a final solution having almost the same fluidity of theoriginal one. The tubings (1.3 mm inner diameter and 1.5 mm outerdiameter) were positioned in the tongue in the same way as in thereal experiment, the volunteer lied in supine position and we deliv-ered 1 ml of blue-solution for 30 s in his mouth. Right after the in-struction ‘swallow’ we took a picture of his mouth; it is visible inFig. 1(a).

The paradigm was such as AsrBsrBsrAsrBsrAsr, where A and B re-fers to the two different taste qualities, s refers to the instruction“swallow” and r to “rinse”. At the end of each session subjects wereinstructed to move the set of tubings from one side of the tongue tothe other side, without altering their head position in order to avoidany undesired movement artifacts. The order of presentation of Aand B conditions and the site of stimulus application (L=left andR=right) were pseudo-randomized across sessions and subjects.

The whole experiment included 4 sessions (two L and two R), in-cluding 6 repetitions for condition A and six for condition B, for anacquisition-time total of 20 min per subject.

Psychophysics rating and assessment

After every fMRI session subjects were asked to verbally rate theintensity of the stimuli (on a scale between 0, not perceived, and10, extremely intense). The rating scores were analyzed by meansof the mathematical functions available in Excel (Office 2010;Microsft, Redmond, WA, USA).

fMRI acquisition

To detect the BOLD (blood oxygenation level dependent) signal, a1.5 T scanner (SONATA-MR; Siemens, Erlangen, Germany) was used.For each subject the functional images were a total of 168 volumes/session and they were acquired by means of 27 axial-slice mosaic2D SE/EP sequence (TR=2500 ms/TE=45 ms/FA=90°/ma-trix=64×64/voxel size=3×3×3.75 mm3). Moreover, a structuralhigh resolution (0.78×0.78×1 mm3) reference image was added foreach volunteer dataset (3D IR/GR sequence; TR=2180 ms/TE=3.93 ms).

fMRI data analysis

The fMRI data analysis was performed by means of SPM5 (Statis-tical Parametric Mapping; Wellcome Department of Cognitive Neu-rology, London, UK) implemented in Matlab 7.5 R.2007b (MathWorks Inc., MA, USA). Spatial pre-processing included: slice timingto reduce the differences in slice acquisition times, realignment and

Fig. 1. (a) Simulation of the mouth stimulation with food coloring. The blue color localizes the liquid-bolus inside the mouth. (b) Time frame description of the 50 s-block designapplied in the fMRI sessions. The graph describes the subject's task highlighted in a marbled field. Subjects were told to watch a computer screen within the scanner. Gustatorystimuli were presented a few seconds after a gray field was projected on the screen. Subjects were instructed to expect and keep in mouth the tastant while the gray field was vis-ible. Then the word “swallow” guided subjects to swallow the tastant. Finally, a pulse of water (“w” in the graph) was delivered to the tongue, while the word “rinse” on the screeninstructed subjects to rinse the mouth and swallow the water. After that, a new block started. In the fMRI analysis only the scans inside the 30 s corresponding to the time whensubjects received the taste stimuli were included. The whole sequence included 4 sessions consisting of 6 block repetitions each, with a total duration of 20 min.

428 E. Iannilli et al. / NeuroImage 60 (2012) 426–435

unwarp to minimize movement effects and susceptibility artifacts,normalization in a stereotactic space and smoothing by means of an8∗8∗8 mm3 FWHM Gaussian kernel (Ashburner and Friston, 2003)in order to improve the signal-to-noise-ratio and reduce residual dif-ferences between subjects. Pre-processed functional data were mod-eled in single-subject first level analyses using the canonicalhemodynamic function and its derivative set available in SPM. Allscans acquired during mouth rinsing and swallowing were excluded.Statistical parametric maps for group inferences were produced by asecond-level random-effects analysis (Penny et al., 2003) using a2×2 factorial design (two stimulus conditions, NaCl, and MSG×twosites of stimulus application, L and R). The data from one subjectwas discarded due to movement artifacts, leaving a final group of 23subjects.

To assess the laterality of the gustatory pathway, based on the hy-pothesis described in the Introduction, we applied a ROI analysis ofthe fMRI imaging data choosing the following regions of interest:thalamus, insula, frontal operculum, orbitofrontal cortex, prefrontalcortex and some areas of the limbic system (including anterior cingu-late cortex, amygdala, and hippocampus). The ROIs were depictedusing the human atlas available in WFU PickAtlas v2.4 software(Mai et al., 2004; Maldjian and Burdette, 2004; Maldjian et al.,2003). Specifically the thalamus ROI comprehended left and right re-gions of thalamus with a total of voxels of 728, the insula regionswere based on 604 voxels on the left side of the brain and similarlyon the right, the frontal operculum comprehended 396 voxels ineach side separately, the orbitofrontal cortex in two areas, one wasconstituted only by Brodmann's area 47 (615 voxels for each side)

and a separate area that was constituted by Brodmann's areas 10and 11 (467 voxels); for the prefrontal cortex we used the union ofBA 44, 45, and 46 (596 voxels). Finally, for the limbic system we con-sidered amygdala and hippocampus left and right in one ROI of521voxels, and separately the anterior cingulate cortex (ACC) definedby the union of BA25 and BA32 (594 voxels).

Spm-fMRI assessmentThe statistical parametrical maps were assessed at a voxel height

threshold of pFWEb0.05, a cluster level of a minimum of 3 voxels,and exclusion masked with p-value in a range of 4E-4bpb2.5E-4(pFWE equivalent) dependent on the dimension of the ROI, by anspm-contrast with relative extent depending on the condition (5 vox-els minimum to 32 voxels maximum per region on interest). For ex-ample, to assess activations inside the I/fO as an effect of MSGapplied to the left side of the tongue (namely MSG_L), the t-contrast MSG, exclusion masked by the contrast MSG_R was used. Be-cause Guest et al. (2007) demonstrated that thermal stimulation eli-cited activation in the insular cortex, and Iannilli et al. (2008) foundsomatosensory processing to involve areas in insular cortex and thal-amus, this masking was applied in order to highlight only the areasactivated by the tastants, and to hide potential effects by other senso-ry inputs such as thermal, mechanical, visual or auditory responses.Moreover, this mask allowed to focus on brain activations related tothe specific quality of tastants, directly linked with the site of applica-tion in the mouth.

All coordinates reported in this manuscript are in MNI space. ThePick-Atlas software toolbox (Mai et al., 2004; Maldjian and Burdette,

Table 1ROI analysis showing the effect of tastants applied on the left or right side of thetongue.

p(k-cor) k p(FWE-cor) Z L R

x,y,z {mm}

ThalamusNaCl_L

0.034 3 0.016 3.11 −12 −12 6NaCl_R

0.051 2 0.062 (punc=0.001) 2.94 9 −15 6MSG_L

0.013 9 0.004 3.71 15 −18 00.022 4 0.024 3.13 −15 −18 6

MSG_R0.020 5 0.004 3.72 15 −21 0

Insula/fOperculumNaCl_L

0.092 3 0.018 3.88 −39 −12 15 fONaCl_R

0.078 4 0.002 4.52 −36 −12 15 I0.040 9 0.024 3.81 −45 9 −9 ** fO0.078 4 0.026 3.78 −33 6 6 I/fO0.044 9 0.047 3.63 39 12 30° fO

MSG_L0.029 6 0.050 3.73 −36 18 3°° fO

MSG_R0.008 26 0.001 4.59 45 15 33° fO0.044 9 0.002 4.50 48 0 0 fO0.035 11 0.017 3.93 42 12 −6 fO0.058 6 0.011 4.03 −42 9 3** fO0.067 5 0.013 3.97 −48 3 24 fO0.045 8 0.020 3.86 −45 0 −3** I/fO

OFC/PFCNaCl_L

0.027 26 0.017 4.19 −48 39 −12NaCl_R

No suprathreshold voxelsMSG_L

0.033 14 0.035 4 −48 36 −3MSG_R

No suprathreshold voxels

Maximum coordinate x,y,z in MNI-space. P value corrected at FWE level. Z: statisticalvalue. k: number of voxels inside the cluster. p(k-cor): p value at the cluster level. I:Insula; fO: frontal Operculum. **Brain area in the left fO shared by both tastants. °°Brain area in the right fO shared by both tastants.

429E. Iannilli et al. / NeuroImage 60 (2012) 426–435

2004; Maldjian et al., 2003) and Atlas of the Human Brain (Mai et al.,2004) were used to identify the brain areas.

Dynamic causal modelingTo estimate coupling among brain regions showing activity in re-

lation to lateralized stimulation we implemented a DCM, that is amathematical method to identify connectivity in functional brain net-work (Friston et al., 2003). It permits model comparisons among hy-potheses, using a Bayesian statistical approach (Bayesian modelselection or BMS) (Penny et al., 2004; Stephan et al., 2009).

To develop DCM the following approach was applied: (1). Brainregions resulting from fMRI group analysis were considered as brainareas potentially connected. (2). Using those areas various modelswere defined, specifying the input, the bilinear and the intrinsicterms. (3). The BOLD fMRI time series was extracted, for each regionin the single subject where the activation was statistically significant.(4). Eachmodel was estimated using the single subjects. (5). BMS wasdefined in order to find the best model, first in single subjects usingan FFX approach and then. (6). At a group-level using fixed and ran-dom effect methods (Stephan et al., 2009, 2010). Models were fittedusing the DCM10 software in SPM-8 (Friston et al., 2003).

The anatomical coordinates of the regions considered in themodels were based on the maxima of statistically significant activa-tion clusters resulting from the group analysis (see Spm-fMRIassessment and ROI in functional imaging sections). It included theleft VPM thalamic nucleus, left fO and left lPFC (see Fig. 6(a)). Foreach subject the BOLD fMRI time series was extracted in the volumeof interest (VOI) with 10 mm diameter centered in the previously de-scribed location, where a subject had no significant activation at thespecific coordinate the center of the VOI was adjusted to the nextsuprathreshold voxel. The threshold for the single subject was set atpb0.001. Among the 23 subjects 12 of them showed suprathresholdactivation in those areas. For each subject each model was fitted,and then for each subject the comparison among the models wasevaluated. The accuracy of the model is indicated by the free-energycriterion, F (Friston et al., 2007), that approximates what is calledthe negative log-evidence, so the best model is characterized by thegreatest log-evidence. That is why this parameter is used to comparemodels within and between subjects.

Results

Intensity ratings

The results of the intensity ratings of the two taste stimuli obtainedduring functional neuroimaging acquisition showed that subjects didnot perceive a significant difference in intensity (I) between the NaCland the MSG stimuli [(meanI±s.e.)NaCl=3.6±2.0; (meanI±s.e.)MSG=4.8±1.9; 2-sample t-testdf=44=1.58, p=0.12], indicat-ing that stimulus intensity waswell matched between the two differenttastants. Moreover, the effect between intensity and stimulus presenta-tion site was neither statistically significant for NaCl [(meanI±s.e.)NaCl-right=3.1±0.5; (meanI±s.e.)NaCl-left=3.8±0.5; tdf=44=1.25, p-left-right=0.22) nor for MSG ((meanI±s.e.)MSG-right=4.8±0.5; (meanI±s.e.)MSG-left=4.9±.4tdf=44=0.26, pleft-right=0.80].

ROI in functional imaging

ThalamusTo assess the activations related to the site of stimulus application

we used the t-contrast ‘tastant-A(B)’, exclusion masked by the t-contrast ‘tastantA(B)_L′ (R′ )’. Thus, to look in the thalamus at the ef-fect of NaCl applied on the left side of the tongue (NaCl_L) we usedthe contrast ‘salt’ exclusion masked by the t-contrast ‘NaCl_R’, and,similarly, for the effects of stimulation of the other side as well asfor the effect of the other tastant.

The results are summarized in Table 1 and Fig. 2. The conditionMSG_L yielded clusters both in the left and right side of the VPM-thalamus, whereas right-sided stimulation with MSG produced aright-sided VPM-thalamic activation. The condition NaCl_L generateda cluster of activation in the left side of the brain; in contrast, the con-dition NaCl_R produced no activation in the thalamus at the set statis-tical level, although at a lower p-value (punc=0.001, pFWE=0.06) anactivated voxel at the right side was found. Altogether this informa-tion seems to indicate a predominantly ipsilateral processing for thestimulus NaCl, as opposed to MSG.

I/fOThe results of the ROI analysis in the I/fO insula and frontal

operculum are summarized in Table 1 and Fig. 3. The contrastsused were ‘tastant-A(B)’, exclusion masked by the t-contrast‘tastant-A(B)_L′ (R′ )’, similar to the one illustrated in the previousparagraph. We found that the left-sided application of NaCl pro-duced only one cluster of activation in the left fO. When the samestimulus was applied on right side of the tongue, activations werelocalized in left insula/fO (3 clusters) and one in the right fO.

MSG stimuli presented on the right side of the tongue pro-duced activations in the right and left side of insula and fO. Nosuprathreshold clusters were found when MSG was applied onthe left side of the tongue. However, searching for a small volume

Fig. 2. Activations in the thalamic ROI-analysis (PFWE-corrb0.05, cluster level=3, t-level in color coded labeled) by the stimulus conditions umami on the left side of the tongue (MSG_L),umami on the right side of the tongue (MSG_R), salt on the left side of the tongue (NaCl_L) and salt on the right side of the tongue (NaCl_R). Apart from the statistical parametric map theROI is shownwith the crosshair position in the same voxel location. The conditionMSG_L activated the right and left VPM thalamic nucleus, the conditionMSG_R the left VPM tn, NaCl_Lthe left VPM t.n. and NaCl_R a region between the medio dorsal and VPM nucleus. The reported coordinates are in the MNI space, on the pictures R=right.

430 E. Iannilli et al. / NeuroImage 60 (2012) 426–435

correction driven by insula/fO coordinates found in the metaanalysis presented in the work of Veldhuizen et al. (2011), a clus-ter of 6 voxels was found to be significantly activated in the leftfO (Table 1).

OFC and PFCWith a contrast similar to the one described above, we found that

the effect of the tastants, when presented on the left side of the ton-gue, produced an activation in the left OFC for both MSG_L and

Fig. 3. Activations in the insular/frontal opercular ROI-analysis (PFWE-corrb0.05, cluster level=3, t-level in color coded labeled) by the stimulus conditions salt on the left side of thetongue (NaCl_L in green), salt on the right side of the tongue (NaCl_R in orange) and umami on the right side of the tongue (MSG_R in blue). The relative bargraph showing thecontrast estimated and the 90% of confidence interval at the maximum is reported in each condition, for each activation and it is color coded with the relative stimulus condition.For every slice the relative ROI is shown in a small window. The reported coordinates are in the MNI space, on the pictures R=right.

431E. Iannilli et al. / NeuroImage 60 (2012) 426–435

NaCl_L. The application of the liquid stimuli on the right side of thetongue did not produce significant activation in the ROI analyzed(Table 1, Fig. 4).

Limbic lobeIn the limbic lobe the effects of the lateralized stimulation

assessed by means of t-contrasts as described above were not statis-tically significant for either taste quality, NaCl and MSG.

Discussion

Laterality of MSG and NaCl

The primary goal of this study was to elucidate by means of fMRIthe laterality of the gustatory system when stimulated by umami

and salt. Our results are based on ROI analysis along the ‘taste map’,that, as recently identified by activation likelihood estimation analysis(ALE) (Veldhuizen et al., 2011), includes: thalamus, I/fO, OFC/PFC andsome areas in the limbic system. Spm-contrasts were defined in a waythat the results would stress the neuronal connection specifically re-lated to the taste quality presented to the left or right side of thetongue.

Results in the thalamic (Fig. 2) region showed bilateral activationfollowing MSG stimulation, with a predominance of ipsilateral pro-cessing. Similarly, left and right-sided stimulation of the oral cavitywith NaCl produced ipsilateral activation at the thalamus level.

Activations of the I/fO (Fig. 3) were more pronounced for bothtastants when they were presented to the right side of the tongue.Overlap among several areas was also observed: both tastants sharedan area of activation in the left side of the insular cortex (stressed in

Fig. 4. Activations in the OFC ROI-analysis (PFWE-corrb0.05, cluster level=3, t-level in color coded labeled) by the stimulus conditions salt on the left side of the tongue (NaCl_L inorange) and umami on the left side of the tongue (MSG_L in blue). The stimulus conditions NaCl_R and MSG_R did not produce any significant activation at the set statistical level.The relative bargraph showing the contrast estimated and the 90% of confidence interval at the maximum is reported for each activation and it is color coded in each condition, withthe relative stimulus condition. For every slice the relative ROI is shown in a small window. The reported coordinates are in MNI space, on the pictures R=right.

432 E. Iannilli et al. / NeuroImage 60 (2012) 426–435

Table 1 by two asterisks) as well as an area in the right frontal oper-culum (stressed in Table 1 by two open circles).

At the level of the OFC/PFC we found similar activations for bothstimuli in the left inferior frontal gyrus (Fig. 4). The two areas are al-most overlapping (Table 1). No supra-threshold voxels survived forright-sided stimulation in the OFC/PFC region.

Finally, the absence of activations inside the limbic lobe indicatedthat – at this level – the information related to the lateralized stimu-lation is lost.

Altogether our results show a predominance of ipsilateral proces-sing for both tastants. The left stimulus condition has effects on thethalamus, insula and left OFC, while the effects of right-sided stimula-tion were detectable in thalamus and insula. We want to stress thatthe effect of an ipsilateral link highlighted by this fMRI study doesnot exclude a more complex network of gustatory areas, which inthis case could be hidden by the specific contrast we used to followthe stronger effect of the lateralized stimulus.

Fig. 5. Gustatory pathway laterality drawn based on our fMRI-results for NaCl (salt) t

Gustatory pathway models

Information of how activated areas are connected is not directlydeducible from fMRI results. Specifically, in the gustatory system itis established, from clinical/anatomical studies, that afferents formthe tongue project to the ipsilateral nucleus of the solitary tract(Goto et al., 1983; Jyoichi et al., 1985; Nakajima et al., 1983); fromhere, studies based mostly on single case report, suggest that secondorder fibers project to the ventral posterior medial (VPM) thalamus,but the laterality of this projection is controversial. Moreover follow-ing the hypothesis presented in neuroimaging studies, the insula andfO seem to be seats of the primary gustatory cortex (Rolls and Scott,2003) even though it is not precisely defined the location (Small,2010; Veldhuizen et al., 2011) while the lPFC is the putative second-ary gustatory cortex ((Rolls, 2000; Rolls and Scott, 2003)).

Combining this previews knowledge with our results we can drawa model of the gustatory pathway for the simplest emerging network

aste and preview knowledge from neuroimaging, clinical and anatomical studies.

Fig. 6. Models description. (a) regional location and position in mm (NMI-space reference) of the three VOIs considered in the model: thalamus (th), frontal operculum (fO) andlateral prefrontal cortex (lPFC). The coordinates are listed for second level group analysis (group), the mean value positions across single subject activation and relative standarddeviation. (b) Definition of the three models. The red arrows indicate the exogenous input. The connections between the interested regions were specified as intrinsic connections.Model 1 (M1) considers direct connectivity (black arrows) between thalamus and respectively fO and lPFC. In Model 2 (M2) fO receives direct input from thalamus and this, in turn,give input to the lPFC. Model 3 (M3) considers lPFC connected both to thalamus and fO, and fO connected to thalamus. Our Bayesian model selection suggested a model rankingM2>M1>M3, indicating M2 as the best.

Fig. 7. Single subject fixed effects Bayesian model selection among DCMs for the threemodels, M1, M2 and M3, expressed relative to a null model. The graph shows the neg-ative free-energy approximation to the log-evidence for each subject. For 9 subjectsover 12 the highest log-evidence is related to M2 and for 3 subjects over 12 the highestlog-evidence is related to M1.

433E. Iannilli et al. / NeuroImage 60 (2012) 426–435

that is the neural correlate to the left-sided stimulation with NaCl.Fig. 5 illustrates it. Following the classical hypotheses, that fO/Insulaconstitutes the primary gustatory cortex and lPFC the secondary gus-tatory cortex, our analyses revealed the path indicated with (a) inFig. 5. We will refer to it as Model 1 (M1). Beside this first classicalmodel we cannot exclude another possible path, indicated with (b)in Fig. 5. It considers insula and lPFC, both receiving direct inputform the thalamus. This idea is supported by neuroimaging studieson humans about the OFC (Elliott et al., 2000), and by findings of pro-nounced connections between the most anterior section of the lateralsubdivision in the OFC/PFC, which is exactly the area that we found tobe involved, and both the mediodorsal thalamus and the granularfield of the insula (Fuster, 1997; Goldman-Rakic, 1987). We willrefer to this model as Model 2 (M2).

Finally for the sake of completeness, we consider a third model inwhich we hypothesize lPFC receiving input from both insula and thal-amus (Model 3, M3).

In order to study the connectivity among those emergent areas in-volved in lateralization of gustatory perception we implemented aDCM based on the three models described above (Fig. 6(b)). We de-cided to focus on the specific activations correlated with the NaCl-stimulus that gave us the most linear results on the left side of thebrain. In the model the exogenous input (Fig. 6(b) section, red ar-rows) was thought to influence the thalamus state, while the connec-tions between the involved regions were modeled as intrinsicconnection (Fig. 6(b)-section, black arrows).

Comparison of modelsBayesian model selection among the three models was performed

for each subject. The results are reported in Fig. 7. M1 was best in ninesubjects out of twelve subjects, whereas for three subjects the highestlog-evidence was for M2. More relevant are the results about the BMSfor the whole group, which were achieved using both the approachffx and rfx (Fig. 8). The fixed effect results show a maximum log-evidence for M2 equivalent to 2.62 that means a BF over 13, this byconvention (Raftery, 1995) indicate positive evidence for M2 at the

Fig. 8. Fixed effects (ffx) and Random (rfx) effects Bayesian model selection at group level estimated for the three models. Ffx-analysis is presented as log-evidences and modelposterior probability. Rfx-analysis is presented as model expected probability and model exceedance probability. It can be seen that by both approaches model M2 is preferredamong the three, and the two methods of model selection produce the same ranking that is M1 best than M2 best than M3.

434 E. Iannilli et al. / NeuroImage 60 (2012) 426–435

group level. Moreover the BF for M1 was over eight, that means thefollowing model ranking M2>M1>M3 emerged, which was alsosupported by the model posterior probability with 60% to the advan-tage of M2, and 36% for M1. The same ranking resulted from the rfxanalysis comparisons where it provided 40% expected probability infavor of M2 and the 38% in favor of M1, while the model exceedanceprobability was estimated as 43% for M1 and 40% for M2. Thus, thecurrent results provide positive evidence that lateralized taste stimu-lation activates the left VPM thalamic nucleus fibers of which directlyproject to the left fO, and from here, second order fibers are sent tothe left lPFC.

Conclusions

In conclusion the main finding of the present study is that activa-tion by both applied tastants, MSG and NaCl, indicated a predominantipsilateral link between the nucleus of the solitary tract and the ven-tral posterior medial nucleus of the thalamus. Moreover, in relation tothe stimulation of the left side of the oral cavity the ipsilateral pre-dominance remained evident also at the level of insular cortex andlOFC. Dynamical casual modeling implemented on the left brain net-work, involving thalamus, fO and lPFC, in the salt condition, indicatedas most likely connection the one that goes from thalamus to fO tolPFC. This supports the hypothesis of a primary gustatory cortex locat-ed in fO and a secondary g.c. in the lPFC.

Acknowledgment

The authors would like to thank the anonymous reviewers fortheir valuable comments and suggestions. This work was supportedpartly by grant from UNIFOR.

References

Aglioti, S., Tassinari, G., Corballis, M.C., Berlucchi, G., 2000. Incomplete gustatory later-alization as shown by analysis of taste discrimination after callosotomy. J. Cogn.Neurosci. 12, 238–245.

Aglioti, S.M., Tassinari, G., Fabri, M., Del Pesce, M., Quattrini, A., Manzoni, T., Berlucchi,G., 2001. Taste laterality in the split brain. Eur. J. Neurosci. 13, 195–200.

Ashburner, J., Friston, K.J., 2003. Spatial normalization using basis function. HumanBrain Function. Academic Press, Amsterdam.

Barry, M.A., Gatenby, J.C., Zeiger, J.D., Gore, J.C., 2001. Hemispheric dominance of corti-cal activity evoked by focal electrogustatory stimuli. Chem. Senses 26, 471–482.

de Araujo, I.E., Kringelbach, M.L., Rolls, E.T., Hobden, P., 2003a. Representation ofumami taste in the human brain. J. Neurophysiol. 90, 313–319.

de Araujo, I.E., Kringelbach, M.L., Rolls, E.T., McGlone, F., 2003b. Human cortical re-sponses to water in the mouth, and the effects of thirst. J. Neurophysiol. 90,1865–1876.

Elliott, R., Dolan, R.J., Frith, C.D., 2000. Dissociable functions in the medial and lateralorbitofrontal cortex: evidence from human neuroimaging studies. Cereb. Cortex10, 308–317.

Faurion, A., Cerf, B., Le Bihan, D., Pillias, A.M., 1998. fMRI study of taste cortical areas inhumans. Ann. N. Y. Acad. Sci. 855, 535–545.

Frey, S., Petrides, M., 1999. Re-examination of the human taste region: a positron emis-sion tomography study. Eur. J. Neurosci. 11, 2985–2988.

Friston, K.J., Harrison, L., Penny, W., 2003. Dynamic causal modelling. NeuroImage 19,1273–1302.

Friston, K., Mattout, J., Trujillo-Barreto, N., Ashburner, J., Penny, W., 2007. Variationalfree energy and the Laplace approximation. NeuroImage 34, 220–234.

Fujikane, M., Nakazawa, M., Ogasawara, M., Hirata, K., Tsudo, N., 1998. Unilateral gus-tatory disturbance by pontine infarction. Rinsho Shinkeigaku 38, 342–343.

Fuster, J.M., 1997. The Prefrontal Cortex. Anatomy, Physiology and Neuropsychology ofthe Frontal Lobe. Raven, New York.

Goldman-Rakic, P.S., 1987. Circuitry of primate prefrontal cortex and regulation of be-havior by representational memory. Handbook of physiology, vol. 5, pp. 372–417.

Goto, N., Yamamoto, T., Kaneko, M., Tomita, H., 1983. Primary pontine hemorrhage andgustatory disturbance: clinicoanatomic study. Stroke 14, 507–511.

Guest, S., Grabenhorst, F., Essick, G., Chen, Y., Young, M., McGlone, F., de Araujo, I., Rolls,E.T., 2007. Human cortical representation of oral temperature. Physiol. Behav. 92,975–984.

Haase, L., Cerf-Ducastel, B., Murphy, C., 2009. Cortical activation in response to puretaste stimuli during the physiological states of hunger and satiety. NeuroImage44, 1008–1021.

Hummel, T., Kobal, G., Gudziol, H., Mackay-Sim, A., 2007. Normative data for the “Snif-fin' Sticks” including tests of odor identification, odor discrimination, and olfactorythresholds: an upgrade based on a group of more than 3,000 subjects. Eur. Arch.Otorhinolaryngol. 264, 237–243.

Iannilli, E., Del Gratta, C., Gerber, J.C., Romani, G.L., Hummel, T., 2008. Trigeminal acti-vation using chemical, electrical, and mechanical stimuli. Pain 139, 376–388.

Ikeda, K., 1909. New seasonings. J. Tokyo Chem. Soc. 30, 820–836.Jyoichi, T., Honda, H., Noda, Y., Shimojo, S., Miyahara, T., 1985. A case of sensory distur-

bance of left 5th finger associated with gustatory disturbance. Rinsho Shinkeigaku25, 1066–1069.

Kinomura, S., Kawashima, R., Yamada, K., Ono, S., Itoh, M., Yoshioka, S., Yamaguchi, T.,Matsui, H., Miyazawa, H., Itoh, H., et al., 1994. Functional anatomy of taste percep-tion in the human brain studied with positron emission tomography. Brain Res.659, 263–266.

Kobal, G., Hummel, T., Sekinger, B., Barz, S., Roscher, S., Wolf, S., 1996. “Sniffin' sticks”:screening of olfactory performance. Rhinology 34, 222–226.

Kobayakawa, T., Endo, H., Ayabe-Kanemura, S., Kumagai, T., Yamaguchi, Y., Kikuchi,Y., Takeda, T., Saito, S., Ogawa, H., 1996. The primary gustatory area in humancerebral cortex studied by magnetoencephalography. Neurosci. Lett. 212,155–158.

Kobayakawa, T., Ogawa, H., Kaneda, H., Ayabe-Kanemura, S., Endo, H., Saito, S., 1999.Spatio-temporal analysis of cortical activity evoked by gustatory stimulation inhumans. Chem. Senses 24, 201–209.

Kobayashi, M., 2006. Functional organization of the human gustatory cortex. J. OralBiosci. 48, 244–260.

Landis, B.N., Leuchter, I., San Millan Ruiz, D., Lacroix, J.S., Landis, T., 2006. Transienthemiageusia in cerebrovascular lateral pontine lesions. J. Neurol. Neurosurg. Psy-chiatry 77, 680–683.

Landis, B.N., Welge-Luessen, A., Brämerson, A., Bende, M., Mueller, C.A., Nordin, S.,Hummel, T., 2009. “Taste Strips” — a rapid, lateralized, gustatory bedside identifi-cation test based on impregnated filter papers. J. Neurol. 256, 242–248.

Lee, B.C., Hwang, S.H., Rison, R., Chang, G.Y., 1998. Central pathway of taste: clinical andMRI study. Eur. Neurol. 39, 200–203.

Lugaz, O., Pillias, A.M., Faurion, A., 2002. A new specific ageusia: some humans cannottaste L-glutamate. Chem. Senses 27, 105–115.

435E. Iannilli et al. / NeuroImage 60 (2012) 426–435

Mai, J.K., Assheuer, J., Paxinos, G., 2004. Atlas of the Human Brain. Elsevier AcademicPress, U.S.A.

Maldjian, J.A.L.P., Burdette, J.H., 2004. Precentral gyrus discrepancy in electronic ver-sions of the Talairach atlas. NeuroImage 21, 450–455.

Maldjian, J., Laurienti, P., Burdette, J., Kraft, R., 2003. An automated method for neuro-anatomic and cytoarchitectonic atlas-based interrogation of fMRI data sets. Neuro-Image 19, 1233–1239.

McCabe, C., Rolls, E.T., 2007. Umami: a delicious flavor formed by convergence of tasteand olfactory pathways in the human brain. Eur. J. Neurosci. 25, 1855–1864.

Mueller, C., Kallert, S., Renner, B., Stiassny, K., Temmel, A.F.P., Hummel, T., Kobal, G.,2003. Quantitative assessment of gustatory function in a clinical context using im-pregnated “taste strips” made from filter paper. Rhinology 41, 2–6.

Nakajima, Y., Utsumi, H., Takahashi, H., 1983. Ipsilateral disturbance of taste due topontine haemorrhage. J. Neurol. 229, 133–136.

Ogawa, H., Wakita, M., Hasegawa, K., Kobayakawa, T., Sakai, N., Hirai, T., Yamashita, Y.,Saito, S., 2005. Functional MRI detection of activation in the primary gustatory cor-tices in humans. Chem. Senses 30 (7), 583–592.

Onoda, K., Ikeda, M., 1999. Gustatory disturbance due to cerebrovascular disorder. La-ryngoscope 109, 123–128.

Penny, W.D., Holmes, A.P., Friston, K.J., 2003. Random effects analysis. In: Frackowiak,R.S.J., Frith, C., Dolan, R., Friston, K.J., Price, C.J., Zeki, S., Ashburner, J., Penny, W.D.(Eds.), Human Brain Function. Press, A., New York, pp. 127–145.

Penny, W.D., Stephan, K.E., Mechelli, A., Friston, K.J., 2004. Comparing dynamic causalmodels. NeuroImage 22, 1157–1172.

Prescott, J., 2004. Effects of added glutamate on liking for novel food flavors. Appetite42, 143–150.

Raftery, A., 1995. Bayesian model selection in social research. In: Marsden, P. (Ed.), So-cial Methodology. MA, Cambridge, pp. 111–196.

Rolls, E.T., 2000. The orbitofrontal cortex and reward. Cereb. Cortex 10, 284–294.Rolls, E.T., Scott, T.R., 2003. Central taste anatomy and neurophysiology. In: Doty, R.L. (Ed.),

Handbook of Olfaction and Gustation. Marcel Dekker, New York, pp. 679–705.Rolls, E.T., Yaxley, S., Sienkiewicz, Z.J., 1990. Gustatory responses of single neurons in

the caudolateral orbitofrontal cortex of the macaque monkey. J. Neurophysiol.64, 1055–1066.

Rolls, E.T., Critchley, H.D., Wakeman, E.A., Mason, R., 1996. Responses of neurons in theprimate taste cortex to the glutamate ion and to inosine 5′-monophosphate. Phy-siol. Behav. 59, 991–1000.

Schoenfeld, M.A., Neuer, G., Tempelmann, C., Schussler, K., Noesselt, T., Hopf, J.M.,Heinze, H.J., 2004. Functional magnetic resonance tomography correlates of tasteperception in the human primary taste cortex. Neuroscience 127, 347–353.

Schuster, B., Iannilli, E., Gudziol, V., Landis, B.N., 2009. Gustatory testing for clinicians.B-ENT 5 (Suppl. 13), 109–113.

Scott, T.R., Taxley, S., Sienkiewicz, Z.J., Rolls, E.T., 1986. Gustatory responses in thefrontal opercular cortex of the alert cynomolgur monkey. J. Neurophysiol. 56,876–890.

Shikama, Y., Kato, T., Nagaoka, U., Hosoya, T., Katagiri, T., Yamaguchi, K., Sasaki, H.,1996. Localization of the gustatory pathway in the human midbrain. Neurosci.Lett. 218, 198–200.

Simon, S.A., de Araujo, I.E., Gutierrez, R., Nicolelis, M.A., 2006. The neural mecha-nisms of gustation: a distributed processing code. Nat. Rev. Neurosci. 7,890–901.

Singh, P.B., Schuster, B., Seo, H.S., 2010. Variation in umami taste perception in the Ger-man and Norwegian population. Eur. J. Clin. Nutr. 64, 1248–1250.

Small, D.M., 2010. Taste representation in the human insula. Brain Struct. Funct. 214,551–561.

Small, D.M., Jones-Gotman, M., Zatorre, R.J., Petrides, M., Evans, A.C., 1997a. Flavor pro-cessing: more than the sum of its parts. Neuroreport 8, 3913–3917.

Small, D.M., Jones-Gotman, M., Zatorre, R.J., Petrides, M., Evans, A.C., 1997b. A role forthe right anterior temporal lobe in taste quality recognition. J. Neurosci. 17,5136–5142.

Stephan, K.E., Penny, W.D., Daunizeau, J., Moran, R.J., Friston, K.J., 2009. Bayesian modelselection for group studies. NeuroImage 46, 1004–1017.

Stephan, K.E., Penny, W.D., Moran, R.J., den Ouden, H.E., Daunizeau, J., Friston, K.J.,2010. Ten simple rules for dynamic causal modeling. NeuroImage 49,3099–3109.

Uesaka, Y., Nose, H., Ida, M., Takagi, A., 1998. The pathway of gustatory fibers of thehuman ascends ipsilaterally in the pons. Neurology 50, 827–828.

Veldhuizen, M.G., Albrecht, J., Zelano, C., Boesveldt, S., Breslin, P., Lundstrom, J.N., 2011.Identification of human gustatory cortex by activation likelihood estimation. Hum.Brain Mapp. 32, 2256–2266.

Yamaguchi, S., 1991. Basic properties of umami and effects on humans. Physiol. Behav.49, 833–841.

Yaxley, S., Rolls, E.T., Sienkiewicz, Z.J., 1990. Gustatory responses of single neurons inthe insula of the macaque monkey. J. Neurophysiol. 63, 689–700.