European Journal of Neuroscience, Vol. 27, pp. 1534–1546, 2008

doi:10.1111/j.1460-9568.2008.06117.x

Covariance-based subdivision of the human striatum using T1-weighted MRI Michael X Cohen,1,2 Michael V. Lombardo3 and Robert S. Blumenfeld2 1

Department of Epileptology and Center for Life and Brain, University of Bonn, Germany Department of Psychology, University of California, Davis, USA 3 Autism Research Centre, Department of Psychiatry, University of Cambridge, England 2

Keywords: anatomy, basal ganglia, connectivity, limbic system, striatum, voxel-based morphometry

Abstract The striatum plays a key role in many cognitive and emotional processes, and displays an intricate pattern of connectivity with cortical and subcortical structures. Invasive tracing work in rats and non-human primates demonstrates that the striatum can be segregated into subregions based on similar clustering of input and output fibers. In contrast, the human striatum is typically segregated according to local anatomical landmarks without considering natural boundaries formed by functional ⁄ anatomical networks. Here, we used non-invasive magnetic resonance (MR) imaging in young, healthy adults to define subregions of the human striatum based on volume correlations with other subcortical and cortical structures. We present three methods to delineate anatomical volumetric correlations based on gray matter content estimated from T1-weighted MR images. We observed both consistencies with and divergences from invasive tracing work in animals, suggesting that magnetic resonance imaging (MRI)-based covariance likely does not correspond to direct anatomical connections, although it might index other forms of connectivity or tissue type similarity. These novel approaches may be useful in understanding connectivity of other regions, and changes in connectivity in patient or ageing populations.

Introduction The striatum plays a central role in reward-guided learning, habit formation and action selection (Rolls, 1994; Hollerman et al., 2000; Everitt & Robbins, 2005; Cardinal, 2006). These functions are supported by complex and dynamic interactions with a wide range of cortical and subcortical structures (Robbins et al., 1989; Morgane et al., 2005; Frank & Claus, 2006; Haber et al., 2006). Dysregulation between the striatum and other regions might contribute to clinical ⁄ psychiatric disorders, including substance abuse, pathological gambling, schizophrenia, Parkinson’s disease and depression (Grace, 1991; Heinz et al., 1994; Jentsch et al., 2000; Ikemoto, 2004; Volkow et al., 2004; Grant et al., 2006; Hyman et al., 2006). The striatum comprises several subregions that, in rats and nonhuman primates, can be roughly differentiated on the basis of their cytoarchitecture, patterns of connectivity and functional properties (Fig. 1), although precise boundaries are difficult to define (Haber, 2003; Voorn et al., 2004). Thus, many researchers who study nonhuman animals typically delineate subregions of the striatum based on patterns of connectivity with other cortical and subcortical structures, instead of by local anatomical markers (Haber & McFarland, 1999; Fudge et al., 2002; Haber, 2003; Voorn et al., 2004; Haber et al., 2006). In contrast, in studies on the human striatum, researchers delineate subregions based on local anatomical features, such as the appearance of ventricles or crossing of fiber pathways in magnetic resonance (MR) images. Although this approach has the advantage of being straightforward to implement and standardizable across research Correspondence: Dr M. X Cohen, 1Department of Epileptology and Center for Life and Brain, as above. E-mail: [email protected] Received 21 June 2007, revised 26 January 2008, accepted 31 January 2008

groups and study populations, it is potentially limiting because it ignores the functional ⁄ anatomical networks of the striatum, which are not likely to be bound by local anatomical landmarks. Here, we explored anatomical volumetric connectivity between striatum subcompartments and the rest of the brain using T1-weighted MR images in young, healthy adult humans. Following preprocessing of T1 images in preparation for voxel-based morphometry (VBM), the value in each voxel reflects local gray matter volume. It has previously been speculated that regions in which gray matter volume correlate are functionally or anatomically linked (Momenan et al., 2004; Mechelli et al., 2005a; Lerch et al., 2006; He et al., 2007). However, this has not been demonstrated, and it remains unknown the extent to which gray matter covariance reflects direct anatomical connectivity. Here, we present three approaches that use covariance in local gray matter volume to delineate shared and distinct networks of volumetric intercorrelations among striatal subregions and other structures in the brain. Because the anatomical pathways of the striatum are well known, this region is amenable to examining the extent to which gray matter covariance obtained via T1-weighted magnetic resonance imaging (MRI) maps onto the invasive tracing work in primates. These approaches help bridge a gap between research on humans and on animals, and also suggest a reconsideration of the anatomical organization of the human striatum that is more in line with that used by researchers who study the animal striatum.

Materials and methods Subjects A total of 103 right-handed volunteers with no history of neurological or psychiatric disorders were recruited across eight separate functional

ª The Authors (2008). Journal Compilation ª Federation of European Neuroscience Societies and Blackwell Publishing Ltd

Anatomical segregation of human striatum 1535 the spatially normalized gray matter by its Jacobian determinant, computed during spatial normalization. For normalization, we used the defaults in SPM5 (16 non-linear iterations and a regularization parameter of 1.0) with the exception of the output image voxel size, which we changed to 1 mm isotropic. Finally, the resulting gray matter probability images were smoothed with an 8 mm FWHM Gaussian kernel. All images were visually inspected to ensure that our preprocessing was successful and that each image was acceptable for analysis (e.g. in the correct orientation, not distorted).

VBM statistical analyses

Fig. 1. Delineations of the human and non-human primate striatum. (a) Projection field inputs from prefrontal cortical regions to the striatum in non-human primates. Adapted, with permission, from Haber et al. (2006), fig. 3J. (b) Standard delineation of a human striatum based on anatomical landmarks. Cd, caudate nucleus; dACC, dorsal anterior cingulate cortex; DLPFC, dorsolateral prefrontal cortex; ic, internal capsule; NAcc, nucleus accumbens; OFC, orbitofrontal cortex; Pu, putamen; vmPFC, ventromedial prefrontal cortex.

imaging studies that have been published elsewhere (Ranganath et al., 2004, 2005; Brozinsky et al., 2005; Cohen & Ranganath, 2005; Cohen et al., 2005b; Blumenfeld & Ranganath, 2006). There were 44 males and 59 females (18–30 years old). We did not have access to the precise ages of all of our subjects, but all were within this range. It is unlikely that our results could have been systematically influenced by age within the limited range of our sample. All subjects in these studies were self-reported free of psychiatric illnesses. Subjects in all experiments gave informed consent, according to the Declaration of Helsinki.

Data acquisition and preprocessing A structural MRI image was acquired from each subject using a 1.5 Tesla GE Signa scanner at the UC Davis Imaging Research Center. Scanning parameters were as follows: sequence: 3D IR-SPGR; TR: 9 ms; TE: 2 ms; flip angle: 15; inversion pulse (IRprep): 500 ms; matrix size: 256 · 124 · 256; field of view: 220 mm; slices: 124 in coronal orientation; voxel size: 0.86 · 1.5 · 0.86 mm. Structural images were preprocessed for VBM analyses with Statistical Parametric Mapping software (SPM5) running under Matlab 7.1 (MathWorks, Natick, MA, USA). VBM is a semiautomated technique used to examine differences in gray matter density or volume at each voxel in a whole-brain analysis (Ashburner & Friston, 2000; Mechelli et al., 2005b). The first preprocessing step was segmentation of the T1-weighted MR structural images into gray matter using a specialized VBM toolbox within SPM5 (VBM5; available at http://dbm.neuro.uni-jena.de/vbm/vbm5-for-spm5/). This segmentation procedure incorporates an intensity non-uniformity correction (Ashburner & Friston, 2000), gray matter segmentation and spatial normalization of the segmented gray matter to the Montreal Neurological Institute (MNI) template. An additional hidden random Markov field algorithm was used to remove noise from the segmentation procedure. After segmentation, the resolution of the data was 1 mm3 isotropic voxels. Next, we ‘modulated’ each segmented gray matter probability map to ensure that the intensity value in each voxel conserves the total amount of gray matter within that voxel before and after spatial normalization (Good et al., 2001). Modulation involves multiplying the intensity values at each voxel in

VBM is a standard tool for examining changes in gray matter across populations or within individuals across time. Here, we used a multiple regression approach (Ashburner & Friston, 2000) to delineate anatomical circuits related to striatal subregions by correlating the average gray matter volume values in three striatal regions of interest (ROIs) ) the caudate nucleus, putamen and nucleus accumbens ) with gray matter volume in every other voxel in the brain. Thus, this analysis identifies voxels in which gray matter volume correlates significantly with gray matter volume in the striatal subregion. These subregions were obtained from the Automatic Anatomical Labeling (AAL) atlas (Tzourio-Mazoyer et al., 2002), which is commonly used to identify brain structures. Gray matter volume in our three striatal seed ROIs was obtained in the following manner. A mask for each striatal seed ROI was extracted from the AAL atlas. This mask was then applied over our smoothed modulated gray matter images and an average value was extracted from the voxels within the seed ROI. Extracting the average gray matter value in each ROI, rather than counting the number of voxels within each mask that exceeds a certain threshold, might provide better sensitivity because: (1) gray matter probability is meaningfully related to the tissue content in each voxel; and (2) using a threshold is inherently arbitrary and thus different thresholds may produce slightly different results. We averaged seed ROIs across hemispheres because the mean volume of a seed in one hemisphere correlated highly with the volume of the same seed in the other hemisphere (interhemispheric correlation coefficients for nucleus accumbens, caudate nucleus and putamen: 0.88, 0.89 and 0.85 after correcting for total-brain volume). We performed three separate general linear models (one for each ROI, as previously done; Mechelli et al., 2005a), each with the following independent variables: (1) mean seed ROI volume values; (2) gender; and (3) total brain volume. Gender and total gray matter volume were included as covariates of no interest. In order to calculate total gray matter, the sum of all voxels in the gray matter segmented images (which were obtained using standard and published methods within SPM5; Ashburner & Friston, 2005) was taken (Alema´n-Go´mez et al., 2006; http://www.thomaskoenig.ch/Lester/ibaspm.htm). Thus, variability due to these variables is partialed out when examining specific contrasts. Statistical inferences were made based on a threshold of P < 0.01 false discovery rate (FDR)-corrected (Genovese et al., 2002) and a cluster threshold of at least 50 contiguous voxels (i.e. 50 mm3). In the figures, all displayed voxels are statistically significant at this level. Different views in different figures were selected to maximize the amount of information present for each analysis.

Correlation-based morphometry analysis In VBM and in other analyses designed to assess covariance in gray matter volume, regions of the brain are identified a priori, using

ª The Authors (2008). Journal Compilation ª Federation of European Neuroscience Societies and Blackwell Publishing Ltd European Journal of Neuroscience, 27, 1534–1546

1536 M. X Cohen et al. standardized maps or local landmarks. However, as discussed in the Introduction, anatomists who study the rat and primate striatum have suggested that a better anatomical characterization of the striatum might arise from defining subregions according to their similar patterns of inputs and outputs. In our next two sets of analyses, we present methods to reconstruct the human striatum, with no a priori anatomical restraints, according to patterns of gray matter covariance. Our first approach is called correlation-based morphometry, and involves characterizing each voxel in the striatum according to the region outside the striatum that voxel most strongly correlated with. This allowed us to spatially reconstruct the striatum into compartments that are most strongly correlated with target structures outside the striatum. To do this, we treated each voxel in the striatum as spatially independent, correlated its volume with the volume of selected target regions across subjects (see below), and labeled each voxel according to the target region with which it correlated most highly. This analysis is similar to analyses used in diffusion tensor imaging to segregate brain regions according to their patterns of connectivity with other regions of the brain (Behrens et al., 2003). We calculated the partial correlation coefficient (partialing out any correlations with total brain volume), across subjects, between each voxel in the entire striatum (comprising the nucleus accumbens, caudate and putamen AAL maps) and the average volume values from selected target regions outside the striatum. In other words, we generated a 4D matrix of correlation coefficients over space, in which the value at each voxel in the ith element in the 4th dimension (i.e. rx,y,z,i) is the correlation coefficient between volume values at that voxel and the average value from target region i. We used the following targets (and corresponding numerical descriptors in the left hemisphere of the AAL map for reference): medial orbitofrontal cortex (26); lateral orbitofrontal cortex (14); medial prefrontal cortex (24); dorsolateral prefrontal cortex (8); rostral anterior cingulate gyrus (32); mid-cingulate gyrus (34); posterior cingulate gyrus (68); amygdala (42); hippocampus (38); and V1 (48). We chose these targets based on prior tracing work demonstrating anatomical connectivity between these regions and the striatum. V1 was included as a control region, because although a V1– striatum connection exists in the rat, no known direct connections exist between early visual cortex and the striatum in non-human primates or cats (Lopez-Figueroa et al., 1995). It is not known whether such connections exist in the human. Note that these target regions are defined by the AAL maps, and are thus independent of any results from the VBM analyses. The result of this correlation analysis is 20 partial correlation coefficients for each striatal voxel (10 seed ROIs and two hemispheres). Finally, we assigned each target region with a color value (e.g. right medial orbitofrontal cortex has number 1, left medial orbitofrontal cortex has number 2), and labeled each voxel in the striatum with the color value of the target region that voxel is most highly correlated with. In other words, we collapsed across the 4th dimension by assigning each voxel to the target region with the highest correlation coefficient. This analysis was done once for target regions in the left hemisphere and then again for target regions in the right hemisphere. Note that the null hypothesis ) that no spatial patterns of correlations exist between striatal voxels and target regions ) would result in a map of random, non-contiguous, ‘checkerboard-looking’ colored voxels. Factor-based morphometry analysis The correlation-based morphometry method can demonstrate that there is spatial organization in the striatum that can be uncovered by

examining patterns of correlations with volume of target brain regions. In our next analysis approach, we wished to confirm and extend these findings using an alternative method. Factor-based morphometry involves conducting a factor analysis ) a data reduction technique that extracts principle axes of variance out of multidimensional data ) to cluster spatially both voxels in the striatum and target regions outside the striatum according to how strongly they intercorrelate. Whereas the correlation-based morphometry can demonstrate that there is an inherent spatial organization to the human striatum, factor-based morphometry can reveal more intricate patterns of organization that relate and reduce variance from multiple regions. Thus, factor-based morphometry also has the advantage of clustering multiple regions that show high intercorrelations. We entered all partial correlations between each striatal voxel and each target ROI (listed above) into a factor analysis. In the dataset, each row corresponds to a voxel in the striatum and each of 20 columns corresponds to the partial correlation (across subjects) between gray matter in that voxel and gray matter in the target region (e.g. hippocampus). The factor analysis uses covariance among the variables to find a minimum number of factors that explains a maximal amount of variance. To do this, we used a principal components extraction method and varimax rotation with Kaiser normalization (standard settings for factor analysis) in SPSS 12.0. Resulting factors represent orthogonal axes of covariation in the data, and each factor has an associated eigenvalue, which is related to the amount of variance explained. Typically, factors are selected for further investigation if they have an eigenvalue greater than one. Our analysis identified four main factors with eigenvalues greater than one. From these factors, four factor loading scores are calculated (i.e. how strongly each voxel loads onto each factor). The right hemisphere factor analyses produced a 5th factor with an eigenvalue of 1.1 that contained exclusively V1. This factor was not further considered. Next, each striatal voxel and each target region was color-coded according to the factor on which it loaded most strongly. This analysis was done separately for each hemisphere. As with the correlation-based morphometry, the null hypothesis ) that there is no inherent organization in the striatum related to the volume of target regions ) would produce a spatial map devoid of contiguous clusters.

Results VBM: regions of convergence Each ROI was highly related to itself and surrounding striatal structures in both hemispheres. In addition, we found that the volume of all striatal ROIs correlated with the volume of overlapping voxels in the amygdala, medial orbitofrontal cortex and the substantia nigra (Figs 2–4). The volume of the nucleus accumbens and the caudate both covaried with the volume of overlapping voxels in the anterior hippocampus and thalamus. For the caudate and putamen seed regions, volumetric relationships were observed in overlapping voxels in the right insula. All regions with significant covariation from this and other VBM analyses are reported in Table 1. The spatial overlap of volume correlations in the substantia nigra is particularly interesting because the substantia nigra provides dopaminergic input to the entire striatum (Groves et al., 1995; Williams & Goldman-Rakic, 1998). In Fig. 5 we show a zoomed-in view through the midbrain to demonstrate this overlap. To confirm that this spatial overlap in correlations was due to covariance with midbrain volume and not to smoothing artifacts (i.e. if the results were driven by gray matter outside the substantia nigra but appeared to be in the substantia

ª The Authors (2008). Journal Compilation ª Federation of European Neuroscience Societies and Blackwell Publishing Ltd European Journal of Neuroscience, 27, 1534–1546

Anatomical segregation of human striatum 1537

Fig. 2. Results from the voxel-based morphometry (VBM) analysis using the nucleus accumbens as a seed region (P < 0.01 FDR with a 50 contiguous voxel cluster threshold). Scatterplots on the right depict the relationship between nucleus accumbens, and lateral and medial orbitofrontal cortex (OFC) volumes across subjects. Female subjects are represented by pink X and male subjects are represented by violet dots. Values represent the fraction of total-brain volume from all activated voxels within the medial or lateral orbitofrontal masks from the AAL atlas (see Materials and methods for numerical identification value associated with these masks). T-statistics and MNI coordinates from these regions are listed in Table 1. LOFC, lateral OFC; MCC, mid-cingulate cortex; MOFC, medial OFC; NAcc, nucleus accumbens; PCC, posterior cingulate cortex; SN, substantia nigra.

nigra due to smoothing), we conducted an additional VBM analysis using the substantia nigra as the seed ROI. Here, we observed significant correlations between substantia nigra volume and volume in the ventral striatum, which was most pronounced in ventromedial regions, but also extended dorsally into the ventral anterior caudate. Correlations were also observed in the amygdala, extending into the anterior hippocampus (Fig. 6).

primary motor and somatosensory cortices (Fig. 3). Additionally, regions in the left anterior cingulate and bilateral middle cingulate cortex covaried with caudate volume. Finally, we observed a cluster in the ventral medulla, below the pons. The volume of the putamen seed region was uniquely related to the volume of the right parahippocampal gyrus and bilateral pons. No other striatal seed region displayed significant covariance to these two brain structures (Fig. 4).

VBM: regions of divergence In the previous section, we described brain regions for which the striatal ROIs had spatially overlapping volume correlations. We also found regions of the brain in which gray matter volume correlated with only one of the striatal ROIs. The nucleus accumbens seed volume showed a unique volumetric relationship with several areas of prefrontal cortex, including right lateral orbitofrontal cortex, left medial prefrontal cortex, left ventrolateral prefrontal cortex and right dorsolateral prefrontal cortex (Fig. 2). Additionally, the correlation with ventromedial prefrontal cortex volume (area 25) was more spatially extensive than was the caudate-ventromedial correlation. The left middle and posterior cingulate gyrus was also related to the volume of the nucleus accumbens seed region. In the temporal lobe, the nucleus accumbens seed volume was related to that of the left anterior middle temporal gyrus and fusiform gyrus as well as the right anterior temporal pole. The volume in the caudate seed region showed a pattern of volumetric covariance that was more dorsal and lateral than the results of the nucleus accumbens seed. Specifically, we observed clusters in bilateral dorsolateral and left ventrolateral prefrontal cortex, and

Correlation-based morphometry We used correlation-based morphometry to spatially reconstruct the striatum into compartments according to target structures outside the striatum with which the gray matter volume of each voxel was most strongly correlated. This procedure identified large clusters of voxels within the striatum that had similar patterns of gray matter volume correlation with target regions. These clusters had a ventromedial to dorsolateral gradient in the striatum with a roughly parallel gradient outside the striatum. That is, the volume of striatal voxels situated ventromedially correlated with the volume of the medial orbitofrontal cortex; the volume of striatal voxels situated ventrolaterally correlated with the volume of lateral orbitofrontal cortex; the volume of striatal voxels situated dorsomedially correlated with dorsomedial prefrontal cortex; and striatal voxels situated dorsolaterally correlated with dorsolateral prefrontal cortex (Fig. 7). Additionally, ventral striatal voxels roughly along the midline between medial and lateral correlated with amygdala volume, and more medial and slightly posterior were voxels that correlated with hippocampus volume. Although there were some differences between right- and left-

ª The Authors (2008). Journal Compilation ª Federation of European Neuroscience Societies and Blackwell Publishing Ltd European Journal of Neuroscience, 27, 1534–1546

1538 M. X Cohen et al.

Fig. 3. Results from the VBM analysis using the caudate nucleus as a seed region. Conventions are as in Fig. 1. ACC, anterior cingulate cortex; DLPFC, dorsolateral prefrontal cortex; MCC, mid-cingulate cortex; MPFC, medial prefrontal cortex; RDLPFC, right dorsolateral prefrontal cortex; SN, substantia nigra; VLPFC, ventrolateral prefrontal cortex.

Fig. 4. Results from the VBM analysis using the putamen as the seed region. Conventions are as in Fig. 1. SN, substantia nigra.

hemisphere target structures, the overall patterns of correlations were similar regardless of striatal or target hemisphere. Finally, we found that a region in posterior lateral putamen correlated with volume in V1

(i.e. primary visual cortex). Although this result was not anticipated, it arose consistently in this region in both the right and left hemisphere analyses.

ª The Authors (2008). Journal Compilation ª Federation of European Neuroscience Societies and Blackwell Publishing Ltd European Journal of Neuroscience, 27, 1534–1546

Anatomical segregation of human striatum 1539 Table 1. Results from VBM analyses Seed regions

Hemisphere

Regions with shared variance Nucleus accumbens Nucleus accumbens B Caudate B Putamen B Substantia nigra B Medial OFC B Amygdala B Hippocampus B Thalamus B Caudate Caudate B Nucleus accumbens B Putamen B Substantia nigra B Medial OFC B Amygdala B Hippocampus B Thalamus B Insula R Putamen Putamen B Nucleus accumbens B Caudate B Substantia nigra L Medial OFC L Amygdala B Parahippocampal R Insula R Regions with distinct variance Nucleus accumbens Lateral OFC R MPFC L VLPFC L DLPFC R Middle cingulate L Posterior cingulate L Middle temporal L gyrus Fusiform gyrus L Temporal pole R Caudate DLPFC L DLPFC R DLPFC R VLPFC L MPFC R Anterior cingulate L Middle cingulate L Middle cingulate R Precentral gyrus L Precentral gyrus R Precentral gyrus L Medulla B Putamen Pons B

BA

t-value

x

y

z

25 – 48 – 11 34 28 –

40.82

12

8

)11

– 25 48 – 11 34 28 – 48

18.71

15

16

9

48 25 – – 11 34 36 48

22.26

27

7

)2

11 ⁄ 47 10 45 9 23 23 20

4.71 4.27 4.3 3.91 3.7 3.74 3.97

)36 16 47 )28 5 3 38

47 55 38 31 )19 )34 6

)10 2 )2 34 36 32 )28

37 38

3.84 3.77

28 )38

)36 5

)16 )28

46 11 10 47 9 24 – 23 6 6 6 –

4.32 4.25 4.08 4.03 4.71 4.24 4.1 3.87 4.36 4.14 3.79 4.33

32 )28 )22 36 )7 5 7 )3 32 )40 35 4

57 60 52 56 58 30 )20 )14 )16 )5 )4 )27

14 1 13 )2 34 18 47 44 56 49 45 )48

–

4.46

)10

)13

)32

Because striatal clusters were spatially extensive, we report the voxels with the peak significance values and list, underneath them, other regions included in those clusters. B, bilateral. x, y, z coordinates are in MNI space. DLPFC, dorsolateral prefrontal cortex; MPFC, medial prefrontal cortex; OFC, orbitofrontal cortex; VLPFC, ventrolateral prefrontal cortex.

Factor-based morphometry We applied a factor analysis to each striatum separately (including bilateral target regions), and found four main factors for each hemisphere (see Materials and methods). In each hemisphere, we found a similar ventromedial to dorsolateral topographical segregation

(Fig. 8). We label these factors descriptively as ventral, middle, dorsomedial and dorsolateral. In Table 2, we list the target regions that comprised each factor. In general, the results were similar to findings observed in animal tracing studies, although divergences were also observed. The ventral segment loaded onto a factor containing the hippocampus, amygdala and orbitofrontal cortex in the right hemisphere, and medial prefrontal and anterior cingulate cortices in the left hemisphere. The middle segment, likely comprising the putamen and the ventral part of the caudate, loaded onto a factor containing the anterior cingulate and lateral prefrontal cortex. The dorsomedial segment contained dorsomedial caudate regions, loaded onto a factor containing the medial frontal cortex in the right hemisphere, and the hippocampus and amygdala in the left hemisphere. Finally, the dorsolateral striatal segment loaded onto a factor containing orbitofrontal and dorsolateral prefrontal cortex in the left striatum, and the posterior cingulate in the right striatum. Note that although the loading of the target regions differed between left and right hemispheres, the overall spatial organization within the striatum was similar between the two hemispheres.

VBM using factors as seed regions Results from the previous factor-based morphometry analysis suggest that the striatum can be empirically segregated, based on patterns of correlations with other cortical and subcortical structures. In a followup analysis, we conducted a VBM analysis, similar to the one reported earlier, except that here we used the striatal factors as seed regions instead of a priori selected regions from the AAL map. Here again we averaged across hemisphere due to high interhemispheric correlation between volumes in each factor across subjects (0.87, 0.82, 0.78 and 0.79, for the ventral, middle, dorsomedial and dorsolateral segments, respectively, correcting for total-brain volume). When comparing the results from the ventromedial and ventrolateral factors VBM to the AAL-defined accumbens seed region VBM, we found that the explained variance displayed a similar spatial extent, but was more restrictive, especially in the ventromedial prefrontal cortex and hippocampus (Fig. 9). In contrast, when comparing the dorsomedial and dorsolateral factors VBM with the AAL-defined caudate seed VBM, we observed more regions in which volume correlated with the factor-defined regions than with the AAL map-defined regions (Fig. 9). This was especially the case for the dorsomedial factor, in which the volume of this seed region predicted the volume in a more spatially extensive region in dorsomedial frontal cortex than did the caudate seed region. Results from these analyses are presented in Table 3.

Discussion The striatum must work in conjunction with other regions including frontal cortex and medial temporal lobes to support goal-directed behavior. These interactions are subserved by somewhat distinct but partly overlapping functional ⁄ anatomical loops that support different aspects of cognition and emotion (Haber, 2003). Here we used noninvasive imaging in humans to demonstrate that distributed, partly distinct and partly overlapping, anatomical networks can be linked to subregions of the striatum. Some aspects of our findings were consistent with fronto-striatal anatomical ⁄ functional networks, although there were also several discrepancies with pathways identified through invasive tracing methodologies. Because frontostriatal anatomical connectivity is unlikely to differ greatly between humans and other primates, and because our results were statistically

ª The Authors (2008). Journal Compilation ª Federation of European Neuroscience Societies and Blackwell Publishing Ltd European Journal of Neuroscience, 27, 1534–1546

1540 M. X Cohen et al.

Fig. 5. Overlap of volume relationships in the substantia nigra (SN). Rows display the volume correlations from VBM analyses for the nucleus accumbens (NAcc), caudate and putamen. Orange-colored voxels depict regions of significant covariance with the striatum. Green-colored voxels show the substantia nigra. Overlap between the subregions of the striatum and the substantia nigra is greatest for the nucleus accumbens, followed by the caudate, and the least amount of overlap is for covariation with the putamen.

robust over many subjects, these divergences suggest that gray matter covariance does not entirely map onto white matter fiber pathways.

Convergences with known white matter connections Our factor-based morphometry revealed four major subdivisions, roughly ventral, middle, dorsomedial and dorsolateral. The ventral subdivision likely contains the nucleus accumbens, ventral putamen and extended nucleus of the amygdala, similar to what Haber and colleagues have recently termed the ‘reward-related striatum’ (Haber et al., 2006). Invasive tracing studies in animals demonstrates that these regions receive convergent input from the hippocampus, medial and ventromedial prefrontal cortex, and midbrain dopamine regions (Mogenson et al., 1980; Groenewegen et al., 1999; Haber et al., 2006). Functional imaging and rat lesion studies confirm that the ventral striatum is critically involved in utilizing reward information to guide decision-making (Albertin et al., 2000; Schoenbaum & Setlow, 2003; Cardinal, 2006; Atallah et al., 2007; Heekeren et al., 2007; Schonberg et al., 2007). Another major subdivision was the dorsomedial striatum, which likely comprises the caudate nucleus and part of the putamen. This division was correlated with volume in dorsal prefrontal regions such as dorsolateral and dorsomedial frontal cortex, as well as the posterior cingulate. The dorsal striatum has been shown to support cognitive, prefrontal-mediated processes such as goal-directed behavior and working memory (Saint-Cyr et al., 1988; Rolls, 1994; Collins et al., 2000; Frank et al., 2001). The dorsolateral segment likely comprised the lateral putamen, the internal and external segments of the globus pallidus, and part of the lateral caudate. In the VBM analysis, the putamen as defined from the AAL correlated strongly with the insula, consistent with tracing studies in animals (McGeorge & Faull, 1989). In the correlation- and

factor-based morphometry approaches, this dorsolateral segment correlated highly with lateral prefrontal and mid-cingulate cortices. Thus, the correlation- and factor-based morphometry analyses revealed stronger and more extensive networks of anatomically connected regions than did the standard VBM analysis using the AAL atlas. In the VBM analysis, the putamen seed region exhibited relatively little covariance with brain regions outside the parahippocampal gyrus and pons. This was a striking divergence between our findings and white matter fiber tracing work, because tracing studies in animals (Inase et al., 1996; Brown et al., 1998) and diffusion-tensor imaging results in humans (Lehericy et al., 2004) have demonstrated white matter pathways between the putamen and sensory-motor regions.

Divergences with known white matter connections Several aspects of our results showed striking differences with what would be expected given the white matter anatomical connectivity of the striatum. In the VBM analyses, for example, volume in the ventral region identified from the factor analysis and the nucleus accumbens identified by the AAL map correlated with volume in the dorsolateral prefrontal cortex and lateral orbitofrontal cortex, although these are not pathways identified in invasive tracing work. Further, although the gray matter correlations with dorsal prefrontal regions are consistent with known fiber pathways (McGeorge & Faull, 1989; Leh et al., 2007), the volume of this section correlated with the volume in the amygdala, temporal pole and middle temporal gyrus. Also, caudate (identified from the AAL map) volume correlated with volume of the amygdala. These findings would not be expected based on invasive white matter tracing work. Further, the thalamus projects to all striatal subregions (Deschenes et al., 1995; Gimenez-Amaya et al., 1995; Haber & McFarland, 2001; McFarland & Haber, 2001; Parsons et al.,

ª The Authors (2008). Journal Compilation ª Federation of European Neuroscience Societies and Blackwell Publishing Ltd European Journal of Neuroscience, 27, 1534–1546

Anatomical segregation of human striatum 1541

Fig. 6. The substantia nigra seed region was used in a subsequent VBM analysis to confirm the volume relationship between this region and the striatum. (Also see Fig. 4 for location of substantia nigra seed region.)

2007), and the substantia nigra provides dopamine input into the entire striatum. However, we observed volume correlations between the thalamus and the nucleus accumbens and caudate, but not putamen, and substantia nigra volume correlated with volume significantly only in ventromedial striatal regions. Another unexpected finding was the gray matter volume correlations with V1. We originally included this region as a control, because no known direct connections exist between early visual cortex and the striatum in non-human primates or cats, although they are present in rats (Lopez-Figueroa et al., 1995). It is possible that V1–striatum pathways exist in the human, or that this was a spurious correlation, for example if both the dorsolateral striatum and V1 correlate with a third region. Nonetheless, this pattern of correlation was observed in somewhat analogous voxels in both hemispheres, suggesting it is a robust effect. Even more curious was that V1 loaded negatively onto the left ventral striatum factor and right dorsomedial factor. In the right hemisphere, the loading scores of V1 were relatively weak compared with loading scores in frontal ⁄ temporal regions (see Table 2, which shows that maximum factor loadings with V1 in the right hemisphere were about 0.3, compared with frontal ⁄ temporal factor loadings, which were about 0.6–0.9). Future studies, for example using diffusion-based imaging, could more closely examine this potential connection. We also found several hemispheric asymmetries, particularly in the factor-based morphometry analysis. For example, in the factor-based morphometry analyses, ipsilateral dorsolateral prefrontal cortex

volume correlated with the left dorsolateral but not ventral striatal section, whereas the reverse pattern was observed for the contralateral dorsolateral prefrontal cortex (Table 2). In another example, medial orbitofrontal cortex volume loaded on the right but not left ventral striatum segment, whereas in the left ventral striatum, medial orbitofrontal cortex volume loaded most highly on the dorsolateral segment. However, within each hemisphere, the patterns of factor loadings for ipsilateral and contralateral hemispheres were relatively similar. These asymmetries are curious in light of a lack of major white matter connectivity differences identified in invasive tracing work. Some previous work using MRI in humans has also demonstrated asymmetries in the striatum that relate to depression (Lacerda et al., 2003), schizophrenia (Glenthoj et al., 2007) and dopamine D2 receptor binding (Larisch et al., 1998). However, these previous findings relate to differences in regional volume and receptor binding; to our knowledge the major anatomical pathways do not show strong differences between hemispheres. Thus, from these previous studies and our own results, it appears that there are indeed hemispheric volumetric asymmetries in the human striatum, although more work is needed to understand the functional implications of these findings. These results further suggest that gray matter covariance is driven at least in part by factors other than direct anatomical connectivity. Based on these divergences, an important implication of our findings is that gray matter covariance cannot be taken as a proxy measure for anatomical connectivity. It does appear that gray matter covariance is related to anatomical connectivity to some extent.

ª The Authors (2008). Journal Compilation ª Federation of European Neuroscience Societies and Blackwell Publishing Ltd European Journal of Neuroscience, 27, 1534–1546

1542 M. X Cohen et al.

Fig. 7. The correlation-based morphometry analysis suggested a topographic organization of the striatum along a ventromedial to dorsolateral gradient that mirrored the topographic layout of subcortical and cortical target regions. The color of each voxel reflects the target region with which that voxel correlates most highly.

Several aspects of our findings were consistent with known anatomical pathways, previous studies have found gray matter covariance among putatively anatomically connected regions (Momenan et al., 2004; Mechelli et al., 2005a; Lerch et al., 2006; He et al., 2007), and several studies have related gray matter volume to personality measures, saccade efficiency, intelligence, language acquisition and various cognitive skills (Maguire et al., 2000; Knutson et al., 2001; Gaser & Schlaug, 2003; Draganski et al., 2004, 2006; Haier et al., 2004; Mechelli et al., 2004; Kaasinen et al., 2005; Omura et al., 2005). However, gray matter covariance must be driven by other factors as well, possibly related to different aspects of interregional functional or anatomical associations. Gray matter estimates obtained via

morphometry preprocessing can arise from a variety of tissue types, such as neuron cell bodies, glial cells, synapse density, water content and the presence of lipids; it is unknown which of these tissues types contribute to what extent to the observed signal (Mechelli et al., 2005b). It is possible that our approach is more amenable to detecting functional networks that may not have direct white matter connections. For example, functional connectivity can be observed among cortical, and between subcortical and cortical, regions that are only indirectly anatomically connected (e.g. Cohen et al., 2005a, 2008). In these studies, functional connectivity was not compared with gray matter covariance, but this could be examined in future studies. Finally, it is possible that positive volumetric correlations reflect

ª The Authors (2008). Journal Compilation ª Federation of European Neuroscience Societies and Blackwell Publishing Ltd European Journal of Neuroscience, 27, 1534–1546

Anatomical segregation of human striatum 1543 Table 2. Results from the factor-based morphometry analyses

Left striatum Ipsilateral DLPFC MPFC Medial OFC Lateral OFC ACC MCC PCC Amygdala Hippocampus V1 Contralateral DLPFC MPFC Medial OFC Lateral OFC ACC MCC PCC Amygdala Hippocampus V1 Fig. 8. The factor-based morphometry analysis suggested four factors that explained 87% and 84% of the total variance in the volume correlations between each striatal voxel and each target region for the left and right hemispheres, respectively. Each voxel in the striatum, and each extra-striatal target structure is colored according to the factor on which it loaded most strongly. The overall topographic organization of the striatum is similar to the correlation-based morphometry analysis (Fig. 6), and to previous anatomical delineations of the striatum in non-human primates and rats. Displayed target regions are ipsilateral. Numbers indicate the MNI coordinates of the cross-hairs and slices.

neither anatomical connectivity nor functional connectivity, but instead other factors such as tissue type.

Measuring striatal subregions In studies examining the volume of striatal structures using in vivo imaging in humans, researchers typically measure volume of the structures using predefined anatomical boundaries. In contrast, researchers who study the rat and primate striatum typically delineate subregions of the striatum according to functional ⁄ anatomical circuits (Heimer & Wilson, 1975; Mogenson et al., 1980; Groves, 1983; Groenewegen et al., 1999; Haber & McFarland, 1999; Haber, 2003; Voorn et al., 2004). This approach is also taken because striatal structures have similar types of neurons, making it difficult to draw precise boundaries based on cytoarchitecture (Voorn et al., 2004). We attempted to help bridge this gap between animal and human striatum delineations by grouping striatal voxels according to common patterns of gray matter volume correlations with other regions of the brain. A strength of our approaches is that each striatal voxel is considered independent, which means that any spatial organization that arose was not influenced by a priori hypotheses about possible striatal subregions, but rather naturally emerged as a result of covariance in gray matter volume. The fact that all of our subjects were young, healthy university community members demonstrates that such structural covariance is a robust and inherent feature of brain connectivity, and not an effect of, for example, disproportionate changes in structural anatomy observed in ageing or certain patient populations. Such delineations by anatomical intercorrelations might yield more fruitful results compared with delineations based on

Right striatum Ipsilateral DLPFC MPFC Medial OFC Lateral OFC ACC MCC PCC Amygdala Hippocampus V1 Contralateral DLPFC MPFC Medial OFC Lateral OFC ACC MCC PCC Amygdala Hippocampus V1

Ventral

Middle

Dorsomedial

Dorsolateral

0.1689 0.9095* )0.0182 )0.2521 0.6867* 0.1091 )0.3516 0.0954 )0.0712 )0.7814*

0.3886 )0.1901 0.3582 0.11 0.6254 0.8768* 0.7538* )0.2223 0.3084 )0.1688

)0.3621 )0.0995 0.2715 0.4376 0.0568 )0.3294 0.3444 0.8143* 0.8785* 0.1816

0.669* )0.2292 0.836* 0.8028* 0.2813 0.2208 0.1363 0.4189 0.1524 )0.0472

0.5858* 0.906* )0.0937 )0.1166 0.6717* 0.1194 )0.3656 )0.3338 )0.2008 )0.7528*

0.1967 )0.1293 0.3151 )0.1926 0.5999 0.8963* 0.7523* )0.2818 0.1842 0.2641

)0.4081 0.0061 0.2239 0.2585 0.2355 )0.0508 0.3489 0.7214* 0.9298* 0.3895

0.5847 )0.2669 0.8591* 0.8497* 0.0699 0.1432 0.0457 0.4042 0.0666 )0.0058

0.2338 )0.0504 0.7249* 0.8037* 0.1348 0.1067 )0.0504 0.7853* 0.8019* 0.059

0.8258* 0.1335 0.5645 0.4255 0.3618 0.7069* 0.0406 )0.4396 )0.1789 )0.1525

0.3207 0.9198* 0.0104 )0.0413 0.8027* 0.081 )0.0485 )0.1889 )0.1397 )0.1833

)0.1715 )0.2321 0.1821 )0.1165 0.3278 0.461 0.9229* )0.1492 0.4441 0.3005*

0.1327 )0.1362 0.7917* 0.8583* 0.1851 )0.1455 0.1062 0.8921* 0.7454* )0.0229

0.8898* 0.0821 0.4585 0.3154 0.487 0.7022* 0.1956 )0.0877 0.117 )0.25

0.2988 0.84* 0.1546 )0.1886 0.5792* 0.0024 )0.1756 0.2681 0.1479 )0.3324*

0.0072 )0.3941 0.1682 )0.094 0.2313 0.4661 0.9282* )0.1784 0.4948 )0.1024

Table of factor loading scores for each target region, organized into ipsilateral and contralateral targets. *Factor with highest loading score. Regions are listed in the same order for each section to facilitate comparison. ACC, anterior cingulate cortex; DLPFC, dorsolateral prefrontal cortex; MCC, medial cingulate cortex; OFC, lateral orbitofrontal cortex; PCC, posterior cingulate cortex.

anatomical landmarks, such as the crossing of a certain fiber path, when seeking to understand the relation between striatal structures and ageing or psychiatric conditions. We do not suggest, however, that using local anatomical landmarks to delineate subregions is incorrect or misleading, nor do we suggest that delineations based on a priori maps such as the AAL atlas are incorrect. Additionally, our results do not imply that these circuits are independent of each other; indeed, examination of Table 2 reveals that many brain regions loaded onto more than one single factor. Rather, we suggest that a fuller understanding of brain anatomy and its relation to disease or ageing might come from consideration of how brain regions covary with other regions instead of treating each structure as independent of a larger functional-anatomical circuit.

ª The Authors (2008). Journal Compilation ª Federation of European Neuroscience Societies and Blackwell Publishing Ltd European Journal of Neuroscience, 27, 1534–1546

1544 M. X Cohen et al. Table 3. Results from VBM using factors as seed regions Factor

Hemisphere

BA

t-value

x

y

Dorsolateral Putamen Putamen Globus pallidus Middle cingulate Precentral gyrus Precentral gyrus DLPFC Lateral OFC

R L R L L R R L

– – – 23 6 6 46 47

17.1 16.76 8.7 5.07 4.83 4.7 4.44 4.01

)20 21 )14 2 33 )40 )21 31

4 6 3 )20 )5 )5 45 48

12 12 0 34 41 49 14 )8

L R R L R L R L R R R R R R L R L L L R – R L L L R

– 25 48 48 – 34 9 10 9 ⁄ 32 11 – 46 45 6 8 6 9 8 6 10 – – 5 46 37 38

22.01 20.43 6.34 7.25 6.04 6.01 7.15 6.53 6.41 5.77 4.83 4.67 3.68 4.62 4.53 4.41 4.31 3.77 4.3 4.27 4.07 4.02 4.01 4 3.95 3.94

13 )11 )37 42 )6 18 )6 22 )4 )8 )5 )46 )54 )21 7 )25 22 30 4 )1 2 )57 5 48 66 )27

15 18 6 2 )3 0 56 63 36 30 )29 52 41 9 26 )7 39 25 )9 64 )37 22 )42 48 )56 11

13 8 6 )2 )22 )16 35 21 41 )27 49 4 6 65 61 60 50 56 68 )2 )8 )4 77 7 )7 )22

R L L L

11 6 32 48

3.92 3.68 3.61 3.6

)4 15 8 28

67 )17 40 )3

)16 73 11 9

Lateral Putamen Putamen Putamen Lateral OFC Lateral OFC DLPFC Calcarine sulcus

L R L L L R L

48 48 48 11 47 46 18

21.38 17.39 9.76 4.44 4.21 4.21 3.96

26 )25 31 29 32 )38 17

8 6 )9 38 49 47 )74

)3 )3 )4 )16 )10 9 13

Ventral Putamen Putamen Caudate DLPFC Precentral gyrus Thalamus DLPFC

L R R R R R R

48 48 – 46 6 – 46

18.39 16.52 13.73 4.9 4.25 3.94 3.94

22 )23 )18 )39 )40 )19 )22

11 9 16 48 )4 )23 49

)3 )1 3 10 49 18 13

Dorsomedial Caudate Caudate Insula Insula Amygdala Amygdala MPFC DLPFC MPFC Medial OFC Middle cingulate DLPFC DLPFC Dorsal premotor MPFC SMA DLPFC DLPFC SMA MPFC Cerebellum VLPFC Superior parietal DLPFC Middle temporal Anterior temporal pole Medial OFC Precentral gyrus Anterior cingulate Putamen

Fig. 9. Each striatal factor in the factor-based morphometry analysis (see Fig. 8) was subsequently used in a VBM analysis to identify voxels in which volume correlated with the volumes of the factors. Here we show a multiple overlay, with the VBM results from the nucleus accumbens and ventral and lateral factors (top), and from the caudate nucleus and dorsomedial and dorsolateral factors (bottom). Maps are thresholded at P < 0.01 FDR with a 50 contiguous voxel cluster. Numbers indicate the MNI coordinates of the crosshairs and slices.

z

See legends for Tables 1 and 2 for definitions of acronyms. SMA, supplemental motor area.

The approaches we present here elucidate the intrinsic connectional properties of the human striatum. These methods might be useful in identifying dysfunctional brain networks in individuals suffering from, for example, depression, autism or schizophrenia, for which evidence suggests anatomical deviance from matched controls. These methods might also be useful in studying ageing adults, in whom decreases in structural size tend to be disproportionate in specific fronto-temporalstriatal circuits (Raz et al., 1995, 2003). Finally, these methods might

be useful in functional imaging studies, to test whether regions that coactivate might be anatomically linked as well. Indeed, examining gray matter volume correlations among functionally coactivated regions might help shed light onto the factors that contribute to gray matter covariance (e.g. Momenan et al., 2004). To this end, we note that reliable patterns of results were obtained with fewer subjects than were included in our sample.

ª The Authors (2008). Journal Compilation ª Federation of European Neuroscience Societies and Blackwell Publishing Ltd European Journal of Neuroscience, 27, 1534–1546

Anatomical segregation of human striatum 1545

Acknowledgements We thank Charan Ranganath and Craig Brozinsky for providing additional MRI scans, the Dynamic Memory Laboratory at UC Davis, the Autism Research Centre at the University of Cambridge, Nicole David for helpful comments and discussion, and two reviewers for helpful comments and suggestions.

Abbreviations AAL, Automatic Anatomical Labeling; MNI, Montreal Neurological Institute; MR, magnetic resonance; MRI, magnetic resonance imaging; ROI, region of interest; VBM, voxel-based morphometry.

References Albertin, S.V., Mulder, A.B., Tabuchi, E., Zugaro, M.B. & Wiener, S.I. (2000) Lesions of the medial shell of the nucleus accumbens impair rats in finding larger rewards, but spare reward-seeking behavior. Behav. Brain Res., 117, 173–183. Alema´n-Go´mez, Y., Melie-Garcı´a, L. & Valde´s-Hernandez, P. (2006) IBASPM: toolbox for automatic parcellation of brain structures. Presented at the 12th Annual Meeting of the Organization for Human Brain Mapping, oHBM Florence, Italy. Available on CD-Rom in Neuroimage, Vol. 27, no. 1. Ashburner, J. & Friston, K.J. (2000) Voxel-based morphometry – the methods. Neuroimage, 11, 805–821. Ashburner, J. & Friston, K.J. (2005) Unified segmentation. Neuroimage, 26, 839–851. Atallah, H.E., Lopez-Paniagua, D., Rudy, J.W. & O’Reilly, R.C. (2007) Separate neural substrates for skill learning and performance in the ventral and dorsal striatum. Nat. Neurosci., 10, 126–131. Behrens, T.E., Johansen-Berg, H., Woolrich, M.W., Smith, S.M., WheelerKingshott, C.A., Boulby, P.A., Barker, G.J., Sillery, E.L., Sheehan, K., Ciccarelli, O., Thompson, A.J., Brady, J.M. & Matthews, P.M. (2003) Noninvasive mapping of connections between human thalamus and cortex using diffusion imaging. Nat. Neurosci., 6, 750–757. Blumenfeld, R.S. & Ranganath, C. (2006) Dorsolateral prefrontal cortex promotes long-term memory formation through its role in working memory organization. J. Neurosci., 26, 916–925. Brown, L.L., Smith, D.M. & Goldbloom, L.M. (1998) Organizing principles of cortical integration in the rat neostriatum: corticostriate map of the body surface is an ordered lattice of curved laminae and radial points. J. Comp. Neurol., 392, 468–488. Brozinsky, C.J., Yonelinas, A.P., Kroll, N.E. & Ranganath, C. (2005) Lagsensitive repetition suppression effects in the anterior parahippocampal gyrus. Hippocampus, 15, 557–561. Cardinal, R.N. (2006) Neural systems implicated in delayed and probabilistic reinforcement. Neural Netw., 19, 1277–1301. Cohen, M.X, Elger, C.E. & Weber, B. (2008) Amygdala tractography predicts functional connectivity and learning during feedback-guided decisionmaking. Neuroimage, 39, 1396–1407. Cohen, M.X, Heller, A.S. & Ranganath, C. (2005a) Functional connectivity with anterior cingulate and orbitofrontal cortices during decision-making. Brain Res. Cogn. Brain Res., 23, 61–70. Cohen, M.X & Ranganath, C. (2005) Behavioral and neural predictors of upcoming decisions. Cogn. Affect. Behav. Neurosci., 5, 117–126. Cohen, M.X, Young, J., Baek, J.M., Kessler, C. & Ranganath, C. (2005b) Individual differences in extraversion and dopamine genetics predict neural reward responses. Brain Res. Cogn. Brain Res., 25, 851–861. Collins, P., Wilkinson, L.S., Everitt, B.J., Robbins, T.W. & Roberts, A.C. (2000) The effect of dopamine depletion from the caudate nucleus of the common marmoset (Callithrix jacchus) on tests of prefrontal cognitive function. Behav. Neurosci., 114, 3–17. Deschenes, M., Bourassa, J. & Parent, A. (1995) Two different types of thalamic fibers innervate the rat striatum. Brain Res., 701, 288–292. Draganski, B., Gaser, C., Busch, V., Schuierer, G., Bogdahn, U. & May, A. (2004) Neuroplasticity: changes in grey matter induced by training. Nature, 427, 311–312. Draganski, B., Gaser, C., Kempermann, G., Kuhn, H.G., Winkler, J., Buchel, C. & May, A. (2006) Temporal and spatial dynamics of brain structure changes during extensive learning. J. Neurosci., 26, 6314–6317. Everitt, B.J. & Robbins, T.W. (2005) Neural systems of reinforcement for drug addiction: from actions to habits to compulsion. Nat. Neurosci., 8, 1481– 1489.

Frank, M.J. & Claus, E.D. (2006) Anatomy of a decision: striato–orbitofrontal interactions in reinforcement learning, decision making, and reversal. Psychol. Rev., 113, 300–326. Frank, M.J., Loughry, B. & O’Reilly, R.C. (2001) Interactions between frontal cortex and basal ganglia in working memory: a computational model. Cogn. Affect. Behav. Neurosci., 1, 137–160. Fudge, J.L., Kunishio, K., Walsh, P., Richard, C. & Haber, S.N. (2002) Amygdaloid projections to ventromedial striatal subterritories in the primate. Neuroscience, 110, 257–275. Gaser, C. & Schlaug, G. (2003) Brain structures differ between musicians and non-musicians. J. Neurosci., 23, 9240–9245. Genovese, C.R., Lazar, N.A. & Nichols, T. (2002) Thresholding of statistical maps in functional neuroimaging using the false discovery rate. Neuroimage, 15, 870–878. Gimenez-Amaya, J.M., McFarland, N.R., de las Heras, S. & Haber, S.N. (1995) Organization of thalamic projections to the ventral striatum in the primate. J. Comp. Neurol., 354, 127–149. Glenthoj, A., Glenthoj, B.Y., Mackeprang, T., Pagsberg, A.K., Hemmingsen, R.P., Jernigan, T.L. & Baare, W.F. (2007) Basal ganglia volumes in drugnaive first-episode schizophrenia patients before and after short-term treatment with either a typical or an atypical antipsychotic drug. Psychiatry Res., 154, 199–208. Good, C.D., Johnsrude, I.S., Ashburner, J., Henson, R.N., Friston, K.J. & Frackowiak, R.S. (2001) A voxel-based morphometric study of ageing in 465 normal adult human brains. Neuroimage, 14, 21–36. Grace, A.A. (1991) Phasic versus tonic dopamine release and the modulation of dopamine system responsivity: a hypothesis for the etiology of schizophrenia. Neuroscience, 41, 1–24. Grant, J.E., Williams, K.A. & Kim, S.W. (2006) Update on pathological gambling. Curr. Psychiatry Rep., 8, 53–58. Groenewegen, H.J., Wright, C.I., Beijer, A.V. & Voorn, P. (1999) Convergence and segregation of ventral striatal inputs and outputs. Ann. N Y Acad. Sci., 877, 49–63. Groves, P.M. (1983) A theory of the functional organization of the neostriatum and the neostriatal control of voluntary movement. Brain Res., 286, 109– 132. Groves, P.M., Garcia-Munoz, M., Linder, J.C., Manley, M.S., Martone, M.E. & Young, S.J. (1995) Elements of the intrinsic organization and information processing in the neostriatum. In Houk, J.C., Davis, J.L. & Beiser, D.G. (Eds), Models of Information Processing and the Basal Ganglia. MIT Press, Cambridge, MA, USA, pp. 51–96. Haber, S.N. (2003) The primate basal ganglia: parallel and integrative networks. J. Chem. Neuroanat., 26, 317–330. Haber, S.N., Kim, K.S., Mailly, P. & Calzavara, R. (2006) Reward-related cortical inputs define a large striatal region in primates that interface with associative cortical connections, providing a substrate for incentive-based learning. J. Neurosci., 26, 8368–8376. Haber, S.N. & McFarland, N.R. (1999) The concept of the ventral striatum in nonhuman primates. Ann. N Y Acad. Sci., 877, 33–48. Haber, S. & McFarland, N.R. (2001) The place of the thalamus in frontal cortical-basal ganglia circuits. Neuroscientist, 7, 315–324. Haier, R.J., Jung, R.E., Yeo, R.A., Head, K. & Alkire, M.T. (2004) Structural brain variation and general intelligence. Neuroimage, 23, 425–433. He, Y., Chen, Z.J. & Evans, A.C. (2007) Small-world anatomical networks in the human brain revealed by cortical thickness from MRI. Cereb. Cortex, 17, 2407–2419. Heekeren, H.R., Wartenburger, I., Marschner, A., Mell, T., Villringer, A. & Reischies, F.M. (2007) Role of ventral striatum in reward-based decision making. Neuroreport, 18, 951–955. Heimer, L. & Wilson, R.D. (1975) The subcortical projections of the allocortex: similarities in the neural associations of the hippocampus, the piriform cortex, and the neocortex. In Santini, M. (Ed.), Golgi Centennial Symposium: Perspectives in Neurobiology. Raven Press, New York, pp. 177–193. Heinz, A., Schmidt, L.G. & Reischies, F.M. (1994) Anhedonia in schizophrenic, depressed, or alcohol-dependent patients – neurobiological correlates. Pharmacopsychiatry, 27 (Suppl. 1), 7–10. Hollerman, J.R., Tremblay, L. & Schultz, W. (2000) Involvement of basal ganglia and orbitofrontal cortex in goal-directed behavior. Prog. Brain Res., 126, 193–215. Hyman, S.E., Malenka, R.C. & Nestler, E.J. (2006) Neural mechanisms of addiction: the role of reward-related learning and memory. Annu. Rev. Neurosci., 29, 565–598. Ikemoto, K. (2004) Significance of human striatal d-neurons: implications in neuropsychiatric functions. Prog. Neuropsychopharmacol. Biol. Psychiatry, 28, 429–434.

ª The Authors (2008). Journal Compilation ª Federation of European Neuroscience Societies and Blackwell Publishing Ltd European Journal of Neuroscience, 27, 1534–1546

1546 M. X Cohen et al. Inase, M., Sakai, S.T. & Tanji, J. (1996) Overlapping corticostriatal projections from the supplementary motor area and the primary motor cortex in the macaque monkey: an anterograde double labeling study. J. Comp. Neurol., 373, 283–296. Jentsch, J.D., Roth, R.H. & Taylor, J.R. (2000) Role for dopamine in the behavioral functions of the prefrontal corticostriatal system: implications for mental disorders and psychotropic drug action. Prog. Brain Res., 126, 433– 453. Kaasinen, V., Maguire, R.P., Kurki, T., Bruck, A. & Rinne, J.O. (2005) Mapping brain structure and personality in late adulthood. Neuroimage, 24, 315–322. Knutson, B., Momenan, R., Rawlings, R.R., Fong, G.W. & Hommer, D. (2001) Negative association of neuroticism with brain volume ratio in healthy humans. Biol. Psychiatry, 50, 685–690. Lacerda, A.L., Nicoletti, M.A., Brambilla, P., Sassi, R.B., Mallinger, A.G., Frank, E., Kupfer, D.J., Keshavan, M.S. & Soares, J.C. (2003) Anatomical MRI study of basal ganglia in major depressive disorder. Psychiatry Res., 124, 129–140. Larisch, R., Meyer, W., Klimke, A., Kehren, F., Vosberg, H. & Muller-Gartner, H.W. (1998) Left-right asymmetry of striatal dopamine D2 receptors. Nucl. Med. Commun., 19, 781–787. Leh, S.E., Ptito, A., Chakravarty, M.M. & Strafella, A.P. (2007) Fronto-striatal connections in the human brain: a probabilistic diffusion tractography study. Neurosci. Lett., 419, 113–118. Lehericy, S., Ducros, M., Krainik, A., Francois, C., Van de Moortele, P.F., Ugurbil, K. & Kim, D.S. (2004) 3-D diffusion tensor axonal tracking shows distinct SMA and pre-SMA projections to the human striatum. Cereb. Cortex, 14, 1302–1309. Lerch, J.P., Worsley, K., Shaw, W.P., Greenstein, D.K., Lenroot, R.K., Giedd, J. & Evans, A.C. (2006) Mapping anatomical correlations across cerebral cortex (MACACC) using cortical thickness from MRI. Neuroimage, 31, 993–1003. Lopez-Figueroa, M.O., Ramirez-Gonzalez, J.A. & Divac, I. (1995) Projections from the visual areas to the neostriatum in rats. A re-examination. Acta Neurobiol. Exp. (Wars.), 55, 165–175. Maguire, E.A., Gadian, D.G., Johnsrude, I.S., Good, C.D., Ashburner, J., Frackowiak, R.S. & Frith, C.D. (2000) Navigation-related structural change in the hippocampi of taxi drivers. Proc. Natl Acad. Sci. USA, 97, 4398– 4403. McFarland, N.R. & Haber, S.N. (2001) Organization of thalamostriatal terminals from the ventral motor nuclei in the macaque. J. Comp. Neurol., 429, 321–336. McGeorge, A.J. & Faull, R.L. (1989) The organization of the projection from the cerebral cortex to the striatum in the rat. Neuroscience, 29, 503–537. Mechelli, A., Crinion, J.T., Noppeney, U., O’Doherty, J., Ashburner, J., Frackowiak, R.S. & Price, C.J. (2004) Neurolinguistics: structural plasticity in the bilingual brain. Nature, 431, 757. Mechelli, A., Friston, K.J., Frackowiak, R.S. & Price, C.J. (2005a) Structural covariance in the human cortex. J. Neurosci., 25, 8303–8310. Mechelli, A., Price, C.J., Friston, K. & Ashburner, J. (2005b) Voxel-based morphometry of the human brain: methods and applications. Curr. Med. Imaging Rev., 1, 1–9.

Mogenson, G.J., Jones, D.L. & Yim, C.Y. (1980) From motivation to action: functional interface between the limbic system and the motor system. Prog. Neurobiol., 14, 69–97. Momenan, R., Rawlings, R., Fong, G., Knutson, B. & Hommer, D. (2004) Voxel-based homogeneity probability maps of gray matter in groups: assessing the reliability of functional effects. Neuroimage, 21, 965–972. Morgane, P.J., Galler, J.R. & Mokler, D.J. (2005) A review of systems and networks of the limbic forebrain ⁄ limbic midbrain. Prog. Neurobiol., 75, 143–160. Omura, K., Todd Constable, R. & Canli, T. (2005) Amygdala gray matter concentration is associated with extraversion and neuroticism. Neuroreport, 16, 1905–1908. Parsons, M.P., Li, S. & Kirouac, G.J. (2007) Functional and anatomical connection between the paraventricular nucleus of the thalamus and dopamine fibers of the nucleus accumbens. J. Comp. Neurol., 500, 1050–1063. Ranganath, C., Cohen, M.X & Brozinsky, C.J. (2005) Working memory maintenance contributes to long-term memory formation: neural and behavioral evidence. J. Cogn. Neurosci., 17, 994–1010. Ranganath, C., Cohen, M.X, Dam, C. & D’Esposito, M. (2004) Inferior temporal, prefrontal, and hippocampal contributions to visual working memory maintenance and associative memory retrieval. J. Neurosci., 24, 3917–3925. Raz, N., Rodrigue, K.M., Kennedy, K.M., Head, D., Gunning-Dixon, F. & Acker, J.D. (2003) Differential aging of the human striatum: longitudinal evidence. AJNR Am. J. Neuroradiol., 24, 1849–1856. Raz, N., Torres, I.J. & Acker, J.D. (1995) Age, gender, and hemispheric differences in human striatum: a quantitative review and new data from in vivo MRI morphometry. Neurobiol. Learn. Mem., 63, 133–142. Robbins, T.W., Cador, M., Taylor, J.R. & Everitt, B.J. (1989) Limbic–striatal interactions in reward-related processes. Neurosci. Biobehav. Rev., 13, 155– 162. Rolls, E.T. (1994) Neurophysiology and cognitive functions of the striatum. Rev. Neurol. (Paris), 150, 648–660. Saint-Cyr, J.A., Taylor, A.E. & Lang, A.E. (1988) Procedural learning and neostriatal dysfunction in man. Brain, 111, 941–959. Schoenbaum, G. & Setlow, B. (2003) Lesions of nucleus accumbens disrupt learning about aversive outcomes. J. Neurosci., 23, 9833–9841. Schonberg, T., Daw, N.D., Joel, D. & O’Doherty, J.P. (2007) Reinforcement learning signals in the human striatum distinguish learners from nonlearners during reward-based decision making. J. Neurosci., 27, 12860–12867. Tzourio-Mazoyer, N., Landeau, B., Papathanassiou, D., Crivello, F., Etard, O., Delcroix, N., Mazoyer, B. & Joliot, M. (2002) Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. Neuroimage, 15, 273–289. Volkow, N.D., Fowler, J.S., Wang, G.J. & Swanson, J.M. (2004) Dopamine in drug abuse and addiction: results from imaging studies and treatment implications. Mol. Psychiatry, 9, 557–569. Voorn, P., Vanderschuren, L.J., Groenewegen, H.J., Robbins, T.W. & Pennartz, C.M. (2004) Putting a spin on the dorsal-ventral divide of the striatum. Trends Neurosci., 27, 468–474. Williams, S.M. & Goldman-Rakic, P.S. (1998) Widespread origin of the primate mesofrontal dopamine system. Cereb. Cortex, 8, 321–345.

ª The Authors (2008). Journal Compilation ª Federation of European Neuroscience Societies and Blackwell Publishing Ltd European Journal of Neuroscience, 27, 1534–1546