|Home | About | Journals | Submit | Contact Us | Français|
The basal forebrain contains several interdigitating anatomical structures, including the diagonal band of Broca, the basal nucleus of Meynert, the ventral striatum, and also cell groups underneath the globus pallidus that bridge the centromedial amygdala to the bed nucleus of the stria terminalis. Among the cell populations, the magnocellular, cholinergic corticopetal projection neurons have received particular attention due to their loss in Alzheimer’s disease. In MRI images, the precise delineation of these structures is difficult due to limited spatial resolution and contrast. Here, using microscopic delineations in ten human postmortem brains, we present stereotaxic probabilistic maps of the basal forebrain areas containing the magnocellular cell groups. Cytoarchitectonic mapping was performed in silver stained histological serial sections. The positions and the extent of the magnocellular cell groups within the septum (Ch1-2), the horizontal limb of the diagonal band (Ch3), and in the sublenticular part of the basal forebrain (Ch4) were traced in high-resolution digitized histological sections, 3D reconstructed, and warped to the reference space of the MNI single subject brain. The superposition of the cytoarchitectonic maps in the MNI brain shows the intersubject variability of the various Ch compartments and their stereotaxic position relative to other brain structures. Both the right and left Ch4 regions showed significantly smaller volumes when age was considered as a covariate. Probabilistic maps of compartments of the basal forebrain magnocellular system are now available as an open source reference for correlation with fMRI, PET, and structural MRI data of the living human brain.
The basal forebrain comprises heterogeneous structures located close to the medial and ventral surfaces of the cerebral hemispheres. The most prominent feature of the primate basal forebrain is the presence of continuous collection of aggregated and non-aggregated, large, hyperchromic neurons, often referred to as the ‘magnocellular basal forebrain system’ (Hedreen et al., 1984). Compact cell clusters, especially those underneath the posterior limb of the anterior commissure and subpallidal (sublenticular) areas, are called the basal nucleus of Meynert. Large neurons in this area are continuous with similar cells in the medial septum and in the nuclei of the diagonal band and with scattered large cells in association with various fiber bundles, including the internal capsule and anterior commissure. A substantial proportion of the magnocellular neurons in primates represent cholinergic corticopetal projection neurons (Mesulam, 1983; Saper and Chelimsky, 1984; Mesulam and Geula, 1988) that have received particular attention due to their loss in Alzheimer’s and related disorders (Perry et al., 1984; Price et al., 1986). Cholinergic neurons in rodents are intermingled with GABAergic, and glutamatergic corticopetal neurons and various peptidergic interneurons (reviewed in Zaborszky and Duque, 2003; Hur and Zaborszky, 2005).
The term basal nucleus of Meynert* has been used in the past as a synonym with the substantia innominata (“Ungenante Mark-Substanze”, Reil, 1809), especially in the clinical literature. This latter term, however, lost its significance in the light of recent tracer and histochemical studies indicating that the main part of the basal forebrain that was previously called the substantia innominata belongs to nearby and better defined anatomical systems. The rostral, subcommissural part of the substantia innominata is mainly occupied by the ventral extensions of the striatum and globus pallidus, i.e. the ventral pallidum and the core/shell subdivisions of the nucleus accumbens (ventral striatum). More caudally, the sublenticular part of the substantia innominata is occupied by the ‘extended amygdala’, which refers to the subpallidal cell bridges extending from the centromedial amygdala to the bed nucleus of the stria terminalis (Heimer et al., 1985, 1991; Zaborszky et al., 1985; Sakamoto et al., 1999; Heimer et al., 1999; Heimer, 2000; Riedel et al., 2002; de Olmos, 2004; Heimer and van Hoesen, 2006).
Electrophysiological studies in rodents lend support to the notion that basal forebrain cholinergic and GABAergic neurons are important modulators of cortical activation (reviewed in Detari, 2000; Zaborszky and Duque, 2003; Lee et al., 2005). Furthermore, a substantial amount of experimental data in animals, including primates, suggest the involvement of the basal forebrain in attention, learning, memory, reward and cortical plasticity (Wilson and Rolls, 1990; Richardson and DeLong, 1991; Woytko et al., 1994; Chiba et al., 1995; Everitt and Robbins, 1997; Wang et al., 1997; Gaffan et al., 2002; Muir et al., 2004; Conner et al., 2005; Turchi et al., 2005; Sarter et al., 2006; McGaughy et al., 2006; Weinberger 2007). Similarly, imaging studies in humans - as detailed below - lend support for a role of the basal forebrain in a range of cognitive functions. Table 1 lists some of the imaging studies that reported coordinates with reference to the Talairach system (Talairach and Tournoux, 1988).
Paus et al. (1997) investigated the time course of changes in brain activity during a continuous performance auditory vigilance task. As the level of vigilance is shifted from high alert to drowsiness, regional cerebral blood flow (rCBF) decreased in the fronto-parietal network and in several subcortical structures, including the substantia innominata, medial thalamus, and ponto-mesencephalic tegmentum. These changes were interpreted by Paus et al., (1997) as coordinated functions of a separate cortical-subcortical arousal system and a cortical auditory attention network. In the study of Morris et al (1998), conditioning-related, frequency-specific modulation of tonotopic responses in the auditory cortex of young male volunteers was studied. The modulated regions of the auditory cortex co-varied positively with activity in the basal forebrain, amygdala, and orbitofrontal cortex. Furthermore, in an aversive conditioning paradigm, Morris et al. (1997) manipulated the salience of visual stimuli and found that augmented activation of the right pulvinar co-varied with a region of the basal forebrain.
It has been long known that basal forebrain lesions in humans, including those from trauma, tumor, or rupture of the anterior cerebral or anterior communicating arteries result in impaired memory functions (Damasio et al., 1985; Morris et al., 1992; Diamond et al., 1997; Abe et al., 1998). Since multiple anatomic regions were often damaged in these cases, it has been difficult to determine the exact area of the lesion that caused the cognitive symptoms. A recent fMRI study in young male volunteers suggested that activity in the nucleus of the diagonal band of Broca and the subcallosal area is related to episodic memory recall (Fuji et al., 2002). In another study, using delayed match to sample paradigm sensitive for working memory, activations were found in the dorsolateral prefrontal cortex and basal forebrain regions (Swartz et al., 1995).
Cocaine infusion into cocaine-dependent volunteers leads to significant fMRI signal changes in a basal forebrain region that is populated by cholinergic cells (Breiter and Rosen, 1999). However, the same region has been claimed as part of the ‘sublenticular extended amygdala’, a continuum between the centromedial amygdala and the bed nucleus of the stria terminalis (Heimer and Van Hoesen, 2006). It remains an open question whether or not activations related to the basal forebrain corticopetal system can be segregated from those cell groups that belong to the extended amygdala system.
Finally, consistent with the state-dependent changes of unit recordings in rodents across sleep stages (e.g., Lee et al., 2005), basal forebrain areas in humans were among those that showed significant changes in glucose metabolism (rCMRglu) or rCBF throughout the sleep-wake cycle (Braun et al., 1997; Maquet et al., 1997; Nofzinger et al., 2002).
Due to the complex anatomy of the basal forebrain, the contribution of a specific anatomical system to a particular cognitive function has not been well understood. The superimposition of postmortem anatomical and in vivo functional data into the same reference space allows correlations between cerebral microstructure and functional imaging data (Roland and Zilles, 1994; Amunts et al., 2002; Zilles et al., 2002; Amunts et al., 2004). A microstructural map of the basal forebrain region, however, is not available. Therefore, the goal of the present study was to provide cytoarchitectonic probabilistic 3D maps of basal forebrain structures that contain cholinergic projection neurons. Since cholinergic neurons are often aggregated in clusters and constitute most of the large neurons (> 20 μm long axis) in this brain area (Mesulam et al., 1983), the areas containing such magnocellular cell groups within the septum, the horizontal and vertical limbs of the diagonal band, and in the sublenticular basal forebrain can be easily delineated in histological sections stained with a modified Gallyas’s silver method for cell bodies (Merker, 1983). This method has been proven as an excellent marker for cytoarchitectonic studies (Uylings et al., 1999; Zilles et al., 2002). The cytoarchitectonic maps of ten human brains were warped to the reference space of the Montreal Neurological Institute (MNI) single subject brain (Collins et al., 1994; Holmes et al., 1998). The resulting probabilistic maps can now be correlated with functional neuroimaging data for a better understanding of the role of various basal forebrain systems in cognitive functions.
Brains from five males and five females with no record of neurological or psychiatric diseases were obtained at autopsy (Table 2). All subjects were included in the body donor program of the Institute of Anatomy at the University of Duesseldorf (Germany). The mean age of the subjects was: 64.9 years (range 37-75 years, Table 2). Seven brains were fixed in 4% formaldehyde for several months; three brains were fixed in Bodian-fixative. The brains were suspended at the basilar artery to avoid compression during fixation. T1-weighted MRI scans were obtained using a 1.5 T Siemens Magnetron SP scanner (3D FLASH pulse sequence; flip angle = 40°; TR = 40 ms; TE = 5 ms; voxel size = 1.17 mm [mediolateral] × 1 mm [rostrocaudal] × 1 mm [dorsoventral]) for documentation of brain size and shape before subsequent histological processing, which induces shrinkage and distortion. The brains were dehydrated in graded alcohols, embedded in paraffin, and sectioned coronally (20 μm whole brain sections). Images of the paraffin block-faces were obtained after each 15th section. Each 15th section was mounted on a gelatin-coated slide and stained for cell bodies with a modified silver method of Gallyas (Merker, 1983). This staining yields high contrast of cell bodies for cytoarchitectonic analysis and delineation of regions of interest. Each stained section was digitized using a flat bed scanner.
The outlines of various basal forebrain compartments were traced on 2D images of the sections (1.2 mm apart) with a final resolution of 7,000 × 6,000 pixels using in-house software. The outlines were processed as contour line for each histological section. For delineation of the magnocellular basal forebrain compartments we used a modified version of the Ch1-Ch4 nomenclature of Mesulam (Mesulam et al., 1983), as detailed in the Results section. Magnocellular neurons in the basal forebrain often constitute cell aggregates or clusters that facilitate their recognition. As a rule, borders of delineated areas were drawn where the number of large cells precipitously diminished toward surrounding regions (e.g. Fig. 5B).
The histological sections of the postmortem brains were reconstructed in 3D (“histological volume”) from the digitized histological sections using the MRI scans of the fixed brain as a shape reference with a resolution of 1 mm isotropic. A major difficulty of the reconstruction results from inevitable distortions of the histological sections, which occur during cutting and mounting on glass carriers: These distortions were modelled by the reconstruction algorithm, to arrange the histological sections correctly within the MR volume of the fixed brain (Schormann and Zilles 1998; Hoemke, 2006).
Each 3D reconstructed histological volume was spatially normalized to the single subject T1-weighted MNI reference brain (Collins et al., 1994; Holmes et al., 1998) using a combination of affine linear and non-linear, elastic transformations (Henn et al., 1997; Hoemke, 2006). The MNI reference space is frequently used in functional imaging studies, e.g. in the software environment of SPM (http://www.fil.ion.ucl.ac.uk/spm/). The origin of this reference space is not aligned to the intersection of the AC-PC line with the interhemispheric plane as in the Talairach stereotaxic space (Talairach and Tournoux 1988; T-T). To obtain a stereotaxic space with the origin of the anterior commissure at the level of the interhemispheric plane, the present data were shifted in Y and Z plane by 4 and 5 mm, respectively (“anatomical MNI space”; Amunts et al., 2005). Note that the coordinates are not in the T-T space, but in the anatomical MNI space.
Finally, based on the whole 3D brain reconstructions, volume and shape of the magnocellular cell groups were computed. We applied therefore the transformations, which have been calculated for the 3D-reconstruction of the whole histological sections, for the contour lines. Since the spatial resolution of the digitized histological sections (and the contour lines) was larger than that of the MR data set of the postmortem brain (reference for 3D-reconstruction), we decreased the resolution of the contours accordingly. In the next step, a stack of filled contour lines was generated, which was visualized by surface rendering techniques (marching cube approach). In this study, the software Amira® (Mercury Computer Systems®, Chelmsford, USA) was applied for this purpose.
Probability maps were calculated separately for each Ch compartment. Based on the sample of ten brains, they describe the relative frequency with which the same basal forebrain area (e.g., Ch1-2; Ch3, Ch4, or Ch4p) was represented in each voxel of the reference space as shown in Fig. 8 for Ch1-2, Fig. 9 for Ch3, Fig 10 for Ch4, and Fig. 11 for the posterior part of Ch4 compartment. Figure 12 shows the probability maps of the Ch space in the reference brain without subdividing into several compartments. The center of gravity for each Ch compartments was then calculated (Table 3). For a spatial comparison of the Ch cell groups with the surrounding BF structures, including the ventral pallidum, nucleus accumbens and the bed nucleus of the stria terminalis, all structures were combined in one data set by assigning labels to each voxel. A voxel was assigned to a structure that had the highest probability or that exceeded a threshold of 40% (Fig. 13). The generation of these maximum probability maps (MPM) is described elsewhere (Eickhoff et al., 2005, 2006). Histograms were calculated for quantifying the volume of overlap of the individual structures (Figure 14). Since the absolute volume of overlap is larger in large nuclei and smaller in small nuclei, we normalized the histograms by the sum of voxels to make the histograms comparable for all nuclei (see Supplementary Table 1. for the original data). Figure 16 shows the 3D rendering of an individual brain using the software Amira®.
The volumes of the individual Ch compartments (Table 4) were calculated using Cavalieri’s principle as previously described (Amunts et al., 2005). Volumes of Ch1-2, Ch3, Ch4 and Ch4p were analyzed statistically with respect to left-right and sex differences (ANOVA with repeated-measures design; within factors: area, and hemisphere; between factor: sex). A significance level of α=0.05 was set for all tests. Age was used as a covariate. Bonferroni correction was used to control for the alpha error.
Figure 1 shows the position of all magnocellular compartments as delineated in this study from a male brain #1696 at six coronal levels. Figures Figures22--77 illustrate the cytoarchitecture of the delineated areas as traced from high resolution images of the same sections shown in Figure 1.
Magnocellular neurons at the level of the area subcallosa and gyrus rectus stretching from just below the fused midline of the septum to the ventral edge of the brain and extend close to the medial border of the nucleus accumbens. This compartment largely corresponds to the medial septal nucleus and the nucleus of the diagonal band of Broca as defined by Andy and Stephan (1968). In the more dorsal part of this area, corresponding to the Ch1 group of Mesulam et al. (1983), the cell bodies and initial dendrites of the neurons tend to be oriented perpendicularly to the crossing fibers of the anterior commissure in coronal sections. The ventral part of this compartment (Ch2 group of Mesulam) merges laterally with the Ch3 cell groups. No distinction was made between Ch1 and Ch2 groups.
The Ch3 group of the present account corresponds to the horizontal nucleus of the diagonal band of Broca as defined by Hendreen et al., (1984), labeled by DB in Halliday et al., (1993) and Saper (2004) and labeled with DBB in Swaab (2003). The Ch3 compartment can be regarded as linking cell groups in the medial septum-vertical diagonal band nucleus (Ch1-2) with the more elaborate magnocellular cell groups in the sublenticular forebrain (Ch4). Figure 3A is taken at the level where the fibers of the anterior commissure reached the opposite side through the midline, to show the characteristic position of the Ch3 group. The cells in this compartment (Fig. 3B) tend to be oriented in all directions and are somewhat looser arranged than the more laterally located, often fusiform cells of the Ch4 cell groups (compare Fig. 3B with Fig. 4B).
Cell aggregates in the subpallidal region, largely correspond to the basal nucleus of Meynert as defined by Vogels et al., (1990) and De Lacalle et al., (1991). We did not differentiate subcomponents of the Ch4 compartment as suggested by Mesulam et al., (1983). The large neurons in the Ch4 compartment extend as far rostrally as cell aggregates underneath the nucleus accumbens (Fig. 3A). Large aggregates of darkly stained cell group within the ventrolateral edge of the bed nucleus of the stria terminalis at the border of the internal capsule/anterior commissure were also classified as Ch4 (Fig. 3A). Figure 4A, taken at the rostral level of the supraoptic nucleus where darkly stained cells of the Ch4 group are present in the area ventral to the internal part of the globus pallidus, stretching between the lateral hypothalamus and the temporal aspect of the anterior commissure, some large neurons are in close relation to the anterior commissure. Neurons often build cell clusters and their nuclei are oriented more or less parallel to the brain surface (Fig. 4B). Figure 5A is taken slightly more caudally, at the level of the anterior hypothalamic area. The delineated area contains scattered large neurons, in addition to cell clusters of various sizes. Figure 6 is from a more caudal level, showing small clusters of Ch4 neurons ventral to the globus pallidus and medial and dorsal to the temporal limb of the anterior commissure. Some of these cell groups are arranged around the ansa peduncularis.
We retained the term Ch4p for magnocellular cell groups in a well-defined area that begins caudal to the supraoptic nucleus at the level where the optic tract adjoins the internal capsule/cerebral peduncle and extends progressively laterally up to the most caudal level of the medial mammillary nucleus and the central nucleus of the amygdala. Figure 7B is an enlarged view of the medial, triangular cell aggregate of Ch4p from figure 7A, which is located underneath the internal component of the globus pallidus. At the level of the subthalamic nucleus, a few large, darkly stained cells invade the medial aspect of the putamen, the lateral part of the globus pallidus, and the area above the central amygdaloid nucleus (not shown). These small cell clusters also belong to Ch4p.
The probabilistic maps of Figure Figure88--1111 show the interindividual microstructural variability for each delineated Ch compartment using a color code: blue colour shows that only one brain is represented in an actual voxel, while red color indicates that all ten brains are overlapping. Those parts of the Ch 1-2 that have the lowest inter-subject variability (i.e., highest overlap of individual brain structures) are located rostrally and immediately adjoining the crossing fibers of the anterior commissure (y=3 to y=0). Approximately 28% of the voxels in the Ch1-2 region in the reference brain represent overlap of five or more brains (Fig. 8).
Figure 9 shows the probabilistic map of the Ch3 cell groups, which are located more posterior than the Ch1-2 compartment. It reveals a slightly higher intersubject variability than Ch1-2. Most of the voxels representing a high overlap of Ch3 cells are found between levels y=0 and y=-3. Approximately 24% of all voxels in Ch3 represent an overlap of five or more brains.
The probabilistic maps of the Ch4 region show a relatively large overlap at y=-6 (Fig. 10). However, a few voxels with high overlap stretch rostrally in front of the decussating fibers of the anterior commissure (between y=0 and y=-3). Approximately 28% of all voxels in the Ch4 region in the reference brain represent an overlap of five or more brains.
The Ch4p region shows the highest inter-subject variability among the delineated structures, accordingly only few red voxels can be seen in Figure 11. This compartment occupies the most caudal position among the magnocellular basal forebrain structures. On the average, 19% of all voxels in the Ch4p of the reference space overlap in five or more brains.
Figure 12 displays probabilistic maps at 6 coronal levels when all Ch groups are considered without subdividing into compartments. Since neighboring Ch compartments overlap at their borders, these maps show many more red voxels that originate from all ten brains than the previous probabilistic maps of the individual compartments. Figure 13 shows the maximum probability maps (MPM) of the various Ch regions. Each voxel was assigned to the cytoarchitectonic area that showed the greatest overlap (“winner take all”) among the ten examined post-mortem brains. As a result, MPMs show a non-overlapping representation of the Ch compartments in the anatomical MNI space.
Figure 14 shows the normalized distributions of overlapping voxels of the individual Ch regions in the anatomical MNI space. The comparison of the frequency distributions of these basal forebrain structures clearly demonstrates a similar degree of intersubject variability across Ch compartments (see for original data Supplementary Table 1.).
The mean volume of the whole Ch1-4 complex was 191 mm3 on the right side and 184 mm3 on the left side (Table 4). Among the compartments, the Ch4 group had the largest volume while the Ch3 constituted the smallest one. Interhemispheric and sex related differences in volume were not significant (p>0.05). There was, however, a significant age effect: both left and right Ch4 showed a negative correlation with age (α=0.05) using ANOVA with a repeated measures design and Bonferroni correction (Figure 15).
Figure 16 displays surface rendering of case #1696 to illustrate the spatial distribution of cell clusters in more detail. Although the full configuration can be appreciated only when rotating the model in a coordinate system with surrounding structures, it is obvious even from these panels, showing various aspects of the model, that the entire magnocellular system occupies a complicated space, extending from medially in the septum to caudo-laterally toward the amygdaloid body. Within this space, magnocellular neurons of the Ch4 compartment form several bands (clusters) of various size and shape that are in part parallel and in part connected with each other. The cell aggregates of the Ch3, and especially Ch1-2 have a simpler configuration. If other structures were added to this rendering it becomes obvious that the various cell aggregates of the Ch4 compartment penetrate the nucleus accumbens, ventral pallidum, and the bed nucleus of the stria terminalis to some extent.
In the present study, we manually delineated the magnocellular cell aggregates of the human basal forebrain in silver-stained sections in order to generate cytoarchitechtonic probabilistic maps in stereotaxic space. Most of the neurons of the extended amygdala are small or medium-sized and occupy a different position than the cholinergic neurons in the subcommissural/subpallidal territory in primates (Ghashghaei and Barbas, 2001). Since most of the cholinergic neurons are larger than cells in the surrounding areas, and 90% of the magnocellular perikarya in the Ch4 compartment are cholinergic (Mesulam et al., 1983), the space identified based upon the presence of large neurons in high-resolution histological sections, is likely represent the cholinergic space; i.e., the volume occupied by the cortically projecting cholinergic cell bodies. Mapping from ten brains resulted in probabilistic maps with high overlap, comparable to other subcortical nuclei (e.g., amygdala; Amunts, et al., 2005). The resulting probabilistic maps can be correlated with future functional neuroimaging data for a better understanding of the role of various basal forebrain systems in cognitive functions.
The widely dispersed, more or less continuous collection of aggregated and non-aggregated large neurons that extend obliquely from the septum rostrally to the level of the lateral geniculate body caudally is often called the ‘magnocellular basal nucleus’ (Saper and Chelinsky, 1984), or ‘magnocellular basal forebrain system’ (Hedreen et al., 1984). The rostral portion of these neurons are associated with the medial septum and the nucleus of the diagonal band of Broca, and primarily project to the hippocampus. On the other hand, neurons located more caudally in the subcommissural and sublenticular space, often referred to as the basal nucleus of Meynert, project topographically to various cortical areas (Jones et al., 1976; Mesulam et al., 1983; Pearson et al., 1983). The topographical organization of corticopetal cholinergic projection in humans has been confirmed in cases of Alzheimer’s disease that had relatively selective cell loss in various sectors of the basal forebrain. From these cases one can conclude that medial frontal and cingulate cortices are innervated from more medial cell groups, whereas the lateral neocortex, including the prefrontal, temporal, parietal and occipital cortices receive their projections from more lateral subdivisions of the magnocellular system (Arendt et al., 1985; Mesulam and Geula, 1988).
Mesulam et al. (1983) proposed a Ch nomenclature to differentiate cholinergic neurons according to their cortical target areas. According to this scheme the entire Ch4 space is subdivided into 6 compartments termed anteromedial (Ch4am), anterolateral (Ch4al), anterointermediate (Ch4ai), intermediodorsal (Ch4id), intermedioventral (Ch4ai) and posterior (Ch4p). However, the original subdivisions in standardized MRI space was only partly applied in the present study because we did not apply specific staining for cholinergic neurons. Moreover, in many instances it is not possible to consistently define boundaries between individual compartments, as noted already in 1988 in a symposium on nomenclature of the basal forebrain cholinergic system (Butcher and Semba, 1989). For example, the boundaries between Ch1 and Ch2 seem to be arbitrary and these two regions were taken together in our study, similar to the delineations used by Vogels et al. (1990). The Ch3 group, according to the original description of Mesulam et al. (1983), refers to a band of fusiform neurons close to the ventral surface of the brain in the subcommissural area (see for example, Figs. Figs.7A7A and 10A in Mesulam and Geula, 1988; Fig. 3A in Mufson et al., 1989). The cell group designated as Ch3 in our paper represents a transitory area between the septum-diagonal band complex and the more caudal cell groups of the magnocellular basal forebrain system and may correspond to be part of the anteromedial Ch4 group in some publications (Mesulam et al., 1983; Lehericy et al. 1993; Gilmor et al., 1999). We, and others, noted that a separation of the anterior Ch4 group into medial and lateral sectors by a vascular structure, as proposed by Mesulam and his coworkers, is ambiguous due to the variation of the vessels (Halliday et al., 1993; Swaab, 2003). Similarly, delineations of other Ch4 subdivisions are difficult to reproduce due to lack of specific landmarks, except for a well-defined posterior subgroup behind the supraoptic nucleus where the optic tract attaches to the internal capsule/cerebral peduncle. Our Ch4p probably corresponds to the nucleus subputaminalis of Ayala (Ayala, 1915; Simic et al.,, 1999). Hyperchromic, large neurons can often be found underneath the nucleus accumbens, and around the borders of the posterior limb of the anterior commissure along the putaminal border or more caudally within the ventral putamen itself. We incorporated these cell groups into the Ch4 group, since these cells seem to be connected to the bulk of Ch4 cell aggregates in 3D renderings.
The inhomogeneous, dense clusters of neurons in the magnocellular basal complex are interrupted by regions of low cellular density in humans (Holliday et al., 1993) as well as in rodents and monkeys (Zaborszky et al., 1999). The heterogeneous distribution of cholinergic cells and the highly variable shape of the cholinergic space across the rostrocaudal extent of the basal forebrain may have contributed to differing renderings of the magnocellular basal forebrain complex in primates, including humans (see Pearson et al., 1983; Mesulam et al., 1986; Ulfig, 1989). A publication of a preliminary computer-assisted 3D reconstruction of the magnocellular cells from gallocyanin-stained human sections by Heinsen and his colleagues (2006) showed a similar configuration as depicted in figure 16 of the present paper. Moreover, the magnocellular space in humans as shown in Figure 16 in this study is remarkably similar in its architecture to the cholinergic space in computerized 3D reconstructions of rodents (Zaborszky et al., 1999; Nadasdy and Zaborszky, 2000; Zaborszky et al., 2005).
Obstacles to a more thorough examination of basal forebrain structures in imaging studies of the living human brain include the low intensity and interdigitating nature of the different anatomical systems. In addition, variations in the size of the ventricles and the exact angle of brain orientation may affect the assessment of ventral areas in MRI. In an attempt to overcome these problems, Buchsbaum et al. (1998) took advantage of the thin-plate spline warping technique, developed by Bookstein (1991) using geometrical landmarking, including the anterior commissure, the internal capsule, and the fornix that are easily identifiable in MRI scans. However, because the positions of the magnocellular cell groups differ across brain sections, a 2D warping method seriously limits the conclusions that can be reached using this method for evaluating the magnocellular basal forebrain complex. Moreover, some structures like the ventral border of nucleus accumbens have indistinct margins that are difficult to trace on MRI images.
The method employed here overcomes these problems. The magnocellular cell groups are identified in high resolution histological sections with an appropriate staining. The 3D reconstruction of the histological sections allows for the transfer of delineated structural information into 3D MRI space. The spatial normalization of the postmortem brains to the common reference space, which consists of several steps, enables to study differences between the architecture of the individual brains and the reference brain. The affine linear transformations reduce global differences in scaling (size), rotation (orientation), translation (position), and shearing. The subsequent non-linear transformations were pixel/voxelwise transformations that account for local differences in brain shape. The modeling of the MR data sets of each brain as an elastic medium ensures that the non-linear transformations were smooth. Hence, the topology of the brain regions was preserved.
Variability in the probabilistic maps is, therefore, mainly an indicator for interindividual structural variability on a microscopic level rather than gross brain morphology. Normalizing the overlapping voxels to the total volume of a particular Ch compartments suggests that the intersubject variability of the four Ch regions is similar (Fig. 14). Intersubject variability can be expressed in various ways, for example, we can compare the percentage of voxels in the probability maps that represent more than five brains (see Supplementary Table 1.). Taken all of Ch compartments together (volume for one hemisphere in MNI space: 4,785 mm3), the Ch volume that originates from five or more brains is 37% of the total. This percentage value is compatible to that of the amygdala (44%; Amunts et al., 2005). It exceeds that for the previously studied neocortical area 44 and area 45 of Broca’s regions (11% and 18%, respectively; Amunts et al., 1999). Another example of a neocortical area is sensory area 3b, which is represented in the anatomical MNI space by a volume of 107,514 mm3, however; only 16 % of this space is devoted to overlapping voxels originating from five or more brains (Eickhoff et al., 2005). Due to relatively high overlap (i.e., low interindividual variability) of subcortical structures, as compared to some neocortical areas, probability maps should be efficient tools for the precise topographical analysis of structural changes in the basal forebrain in aging and Alzheimer’s disease.
Severe involvement of the magnocellular basal complex has been found in Alzheimer’s disease, where various studies reported cell loss up to 90% (Whitehouse et al., 1981; Vogels et al., 1990; Lehericy et al., 1993; Cullen and Halliday, 1998). Estimations of the neuronal loss of basal nucleus of Meynert during normal aging also vary greatly from 23% to 50% to no neuronal loss at all (Whitehouse et al., 1981; Chui et al., 1984; De Lacalle et al., 1991). In addition to Alzheimer’s disease, there are structural changes in the basal forebrain occasionally in Parkinson’s disease, Rett syndrome, progressive supranuclear palsy, Parkinson dementia complex of Guam, dementia pugilistica, Pick’s disease, Korsakoff’s syndrome, Down syndrome, Wernicke’s encephalopathy, and Cretuzfeldt-Jacob disease (reviewed in Swaab, 2003). However, much controversy remains whether or not the neuropathological changes are primary or secondary, what is the time course of changes, and whether or not the neuropathological changes occur in the entire extent of the cholinergic space or are restricted to its specific regions. Some of the unresolved issues are at least partly due to the fact that neuropathological examinations are restricted to postmortem cases from a single time point of advanced stages of the disease.
In vivo measures of morphological changes in the basal forebrain would be of great importance for a better understanding of how structural changes correlate to cognitive symptoms in Alzheimer’s disease. These measures are difficult to obtain since smaller subdivisions of the basal forebrain cannot be distinguished in vivo in routine MRI, but at a microscopic resolution. An alternative to an identification of microstructural nuclei in the individual brain, geometrical landmarks have often defined volumes of interest of the magnocellular basal forebrain cell groups. For example, age-related gray matter loss in the basal forebrain was reported in a voxel-based MRI study (Ishii et al., 2005). However, it is not clear from this study to what extent shrinkage of other brain areas contributed to the result. In two another studies, Sasaki et al. (1995) and Hanyu et al. (2002) measured the thickness of an area below the anterior commissure in coronal MR images in patients with Alzheimer’s disease and elderly controls. They found a significant correlation between age and thickness of the measured values in control subjects, in agreement with the age-dependent decrease of the volume of Ch4 in our study. However, methodological problems may have influenced the results of this study. For example, a signal loss in the globus pallidus, caused by its magnetic susceptibility effect with the T2-weighted spin-echo sequence (Schenker et al., 1993), may limit the dorsal delineation of the ROI. Furthermore, in linear measurements the angle of the section plane affects the thickness of the tissue. In another study, Teipel et al. (2005) defined an ROI about 20 times larger than the Ch4 area in our study that included the anterior commissure, large parts of the ventral globus pallidus and a small part part of the medioventral putamen in addition to Ch4 of present account. Teipel et al. (2005) found a signal decrease in proton density MRI in this ROI in patients with Alzheimer’s disease compared with healthy subjects. The degree of signal decline was related to the degree of grey matter atrophy in cortical areas. However, taking into consideration that many neighboring structures of the magnocellular space were included, the measurements based on such large ROIs may lead to erroneous conclusions. Future research is needed to validate these findings.
Probabilistic maps of the magnocellular cell groups in standardized stereotaxic space can provide the fine-grained parcellation necessary to localize neuropathological changes in a heterogeneous region such as the basal forebrain. The ability to delineate precisely microstructural details together with the improved sensitivity of advanced MRI, and with neuropsychological battery for memory disorders should allow the identification of patients with mild cognitive impairment, a patient population who will convert to Alzheimer’s disease, but in which early intervention may slow the disease progress. Probabilistic maps from neighboring basal forebrain areas, including the ventral pallidum, nucleus accumbens, and the bed nucleus of the stria terminalis, will soon be available (Voss, Amunts, Zilles and Zaborszky, in preparation). Using an anatomical toolbox integrated in the SPM program the superimposition of postmortem anatomical and in vivo functional data in the same space of a reference brain allows correlations to be made between microstructural details and functional imaging data (Eickhoff et al., 2005). The anatomical volume of interest can be defined by using the maximum probability maps or thresholding of individual probabilistic maps (e.g., >50%). Calculating the intersection between a volume of activation and anatomically defined volume, hypotheses can be tested and the structural substrates of functional activations can be inferred in a probabilistic context (Eickhoff et al., 2006). For example, it has been hypothesized, based upon anatomical data, that specific clusters of projection neurons in the basal forebrain together with specific prefrontal and posterior cortical associational regions constitute distributed elements of functional parallel circuits (Zaborszky, 2002). The close connectional relationship between prefrontal and basal forebrain regions in rodents and in primates are well known (Sesack et al., 1979; Ghasghhaei and Barbas, 2001; Golmayo et al., 2003; Spiga and Zaborszky, 2006), and imaging and neurophysiogical studies have found neural correlates of both top-down [task-driven] and bottom-up [stimulus-driven] attention processes in frontal and posterior parietal cortices (Buschman and Miller, 2007). However, it is not known whether or not the cholinergic system is activated as part of a general arousal system or if spatially specific cholinergic and other basal forebrain compartments would be correlated with activations in specific cortical targets. Such knowledge is critical for a better understanding of the directionality of information flow in attentional networks.
This project was supported by PHS/NINDS grant NS023945 to L.Z. The Human Brain Project/Neuroinformatics research was funded jointly by the National Institute of Mental Health, NINDS, the National Institute of Drug Abuse, the National Cancer Center (KZ and KA), the Helmholtz Gemeinschaft (KZ and KA), and by the Bundesministerium fur Bildung and Forschung, Brain Imaging Center West (KZ). The authors thank Dr. Susan Feldman for providing with constructive criticisms of a previous version of this article.
*Because of the lack of agreements on the definition of basal nucleus of Meynert or the substantia innominata, when describing the various compartments of the magnocellular neurons, we will use topographical terms and well known fiducial markers.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.