The present study revealed four new cytoarchitectonic areas at the caudal parahippocampal gyrus and the collateral sulcus. The comparison with functional imaging data suggests that they also differ in terms of their role in visual–spatial orientation as well as encoding associative memories. The cytoarchitectonic segregation of the caudal parahippocampal gyrus and the collateral sulcus may provide a microstructural correlate of the functional heterogeneity and complexity of this region that has been described in the literature. Moreover, stereotaxic maps were provided as probability maps, i.e., they captured intersubject variability at each position in space. The variability, however, was less high than in other associative areas of previous studies of our group, e.g., Broca’s region. Moreover, the relationships of the areas with respect to neighbouring sulci was rather stable.
The four areas differed in cytoarchitecture, which was quantified by cluster trees, measuring their similarities and dissimilarities as Euclidian distances. The cytoarchitecture of areas Ph2 and CoS1 was most similar as compared to all others: both revealed a light layer II without a clear-cut border to III, a low cell density in layer III and a thin and cell-poor layer IV. The two areas are located further rostrally than the other two areas. These similarities resulted in a small distance between the two areas in the cluster analysis suggesting also a functional similarity of these areas. Area Ph1 was characterized by a higher density of larger pyramidal cells than Ph2 and CoS1 that were especially found in layer IIIc, layer V as well as in layer VI. This difference in structure was also represented by a greater distance between Ph1 and Ph2/CoS1 in the cluster analysis. Area Ph3 differed from the other three by clear horizontal stripes resulted of a high cell density of layer II, IV and VI with lighter layers III and V in between. Ph3 showed the greatest distance to the other three areas meaning suggesting functional differences.
Comparison with previous architectonic maps
With respect to Brodmann’s cytoarchitectonic map (Brodmann
1909), the four new areas were localized within the caudal part of his area 28. Brodmann has described BA28 as an area with a high cell density, medially adjacent to the rhinal sulcus. He did not further subdivide this area, and did not show the cytoarchitecture within the sulcus. In contrast, we identified four areas in this region, including one area in the depth of the collateral sulcus.
Areas Ph1, Ph2, Ph3 and CoS1 seem to cover areas TH as well as partly PH of von Economo and Koskinas (
1925) (Fig.
1). In correspondence to the localisation of areas Ph1 and Ph3 of the present study, von Economo described the medially located parts of area PH in the posterior collateral sulcus and parahippocampal gyrus. Area PH was described as inhomogeneous depending on the neighbouring areas from which PH took on individual characteristics. Overall, area PH had clear and broad layers II and IV. While layer IV had a high density, V consisted of small cells. The boundary between the cortical ribbon and white matter was distinct. Matching Ph1, subtype PHT (temporal) was characterized by a layer III with occasional large cells in IIIc and a clear distinction of V and VI. Ph3 seems to correspond to subtype PHO (occipital) with occasional large cells in IIIc, while cells of layer IV were smaller than in PHT. The delimitation between layers V and VI was sharp. Similar to TH, area CoS1 was found in the depth of the collateral sulcus, while Ph2 was located at the medial bank of the collateral sulcus. The description of an indistinct layer II and a low cell density of layer III by von Economo and Koskinas was also consistent with the cytoarchitecture of Ph2 and CoS1. Similar to Ph2, layer V showed a higher density of larger pyramidal cells, and layer VI also had a high cell density. The border between the cortex and the white matter was distinct. In summary, TH was consistent with Ph2 and CoS1 in both localisation and cytoarchitecture.
It is difficult to go beyond this point and discuss in more detail correspondence between different maps, since an important added value of the probabilistic maps is the inclusion of the variability between brains. Any historical brain map, such as that from von Economo and Koskinas, based on a selection of sections of one or a few individual brains cannot cover this aspect, and it is not possible to compare them with sufficient accuracy in a common reference space.
Comparison with neuroimaging studies
Compared to the representation of the examined region in previous anatomical maps, the functional segregation of this region is much more heterogeneous and more fine-grained subdivisions have been proposed. Several studied demonstrated that the PHC participates in visual–spatial tasks including map reading, spatial orientation, navigation and spatial memory (Epstein and Kanwisher
1998; Aguirre et al.
1996; Janzen et al.
2007; Maguire et al.
1998; Aguirre and D'Esposito
1999; Mellet et al.
2000). The coordinates of area PPA as identified by Epsteins and colleagues (
1999) mainly overlapped with Ph2 in both hemispheres (Fig.
9). Aguirre et al. (
1996) coordinates for topographical learning were localized in area Ph2 on the right and area Ph3 on the left side (Fig.
9). Janzen et al. (
2007) reported coordinates of increased activity in the recognition of objects that had previously been placed at decision points of mazes and thus served as landmarks, which mainly matched area Ph2 in both hemispheres (Fig.
9). Another study described increased activity during exploration of an environment that contained salient objects as well as textures (Maguire et al.
1998), which matched the transition zone between Ph1, Ph2 and Ph3 of the right hemisphere (Fig.
9). Sommer et al. (
2005) showed activities elicited by encoding of object–location associations, while coordinates matched the localisation of the parahippocampal areas: while objects as retrieval cues elicited activity within the transition between Ph1 and Ph3 in the right hemisphere, activity in the left hemisphere was mainly localized within CoS1. Activity for location as retrieval cue was localized within the transition between area Ph1 and Ph3 on the right side and within area Ph2 on the left side. Common activity pattern for both cue types was found in the transition between Ph1 and Ph3 right as well as in the transition between Ph1 and Ph2 left. Activity elicited by remembering the location associated with a particular object revealed a coordinate that was located within the transition between CoS1 and Ph2 in the left hemisphere (Fig.
9).
Furthermore, the PHC is involved in formation of episodic memory (Davachi et al.
2003; Kirwan and Stark
2004; Tendolkar et al.
2008; Düzel et al.
2003; Henke et al.
1999; Hales et al.
2009; Yang et al.
2008). Henke et al. (
1999) provided coordinates of an area exhibiting great activity elicited by association of words that was localized at the height of CoS1 without matching it (Fig.
9). Kveraga reported coordinates for activity caused by objects with strong contextual associations that were located close to CoS1 (Kveraga et al.
2011) (Fig.
9). In addition, increased activity during associative encoding of pairs of images was localized within Ph2 (Hales et al.
2009) (Fig.
9). Furthermore, coordinates of increased activity for associative memory during encoding and retrieval of face–name associations were located at the height of CoS1 without matching it on the right side, within Ph2 on the left side and at the height of the right transition zone between Ph2 and CoS1 without matching them (Kirwan and Stark
2004) (Fig.
9). The coordinates of these studies are listed in Table
4.
Most recently, Rosenke et al. (
2021) published a map of early visual and category-selective regions in human ventral and lateral occipito-temporal cortex. A comparison of this map with our brain maps revealed a match of the area mFus-faces, which corresponded to the FFA-2, with CoS1. In addition, there was a match of area CoS-places, which corresponded to the PPA, with our areas Ph1, Ph2 and Ph3.
Thus, functional imaging data result in a detailed pattern of segregation pf the PHC: while the anterior part of the PHC seems to be activated by nonspatial associations, activity in the posterior part is elicited by stimuli associated with spatial contexts (Aminoff et al.
2007; Bar and Aminoff
2003; Baumann and Mattingley
2016). The present results provide further evidence for a subdivision into an anterior and a posterior part. In line with this, the discriminant analysis of our four new areas also resulted in a grouping of the rostrally localized areas CoS1 and Ph2 on the one hand, and the caudal located areas Ph1 and Ph3 on the other. Moreover, the comparison with functional imaging data showed activations elicited by associations within the rostral areas. However, activations elicited by visual–spatial information were found in the caudal areas. Interestingly, Weiner et al. provided data about the most probable location of the parahippocampal place area (Weiner et al.
2018), that partly covered area Ph1, Ph2 as well as Ph3 (Fig.
9). This leads to the hypothesis that the PPA can also be further subdivided. Appropriately, it was demonstrated by Baldassano, Beck, and Fei-Fei that the connections of the PPA also have an anterior–posterior subdivision: while the posterior part had stronger connections to visual associated regions, the anterior part was more strongly connected to the parietal and retrosplenial cortex (Baldassano et al.
2013).
Some stereotaxic coordinates of studies from the literature didn’t match exactly the new four areas, although functionally, there was an overlap with activation data. Such mismatch may have different reasons. For example, filtering of functional imaging data with a radius of 5 mm and more may lead to such divergence. Second, different reference spaces were used and terms such as “MNI space” were not always clearly defined. Though, depending on the position of the line between anterior and posterior commissure, the position of stereotactic coordinates may vary. This makes it difficult to assess the correlation of cytoarchitectonic areas with functional imaging data solely by comparing their stereotaxic coordinates. Anatomical precision can be added, for example, using the same workflows to align both anatomical and functional data into the same reference space, and/or using surface-based alignments, which seems to reduce intersubject variability, e.g., Fischl et al., Rosenke et al.. This remains to be a project of future research.
In conclusion, the present study identified four new cytoarchitectonic areas in the parahippocampal gyrus and the collateral sulcus, and provided probability maps in stereotaxic spaces. These maps represent a further step towards a higher coverage of the cortex in this functionally so important region. They can be used as an anatomical reference to relate data of other modalities and functional neuroimaging and connectivity studies to this region, and could be a tool to further sharpen our concepts on visuo-spatial processing and episodic memory.