Additional fiber orientations in the sagittal stratum—noise or anatomical fine structure?

The fiber architecture of the sagittal stratum was studied in postmortem brain sections of parietal and occipital lobes of an adult vervet monkey brain and a right human hemisphere. The vervet monkey brain (age: 2.5 years) was removed from the skull and then perfusion-fixed in accordance with the UCLA Chancellor’s Animal Research Committee (ARC #2011-135) and the Wake Forest Institutional Animal Care and Use Committee (IACUC #A11-219). The human brain hemisphere was obtained via the body donor program of the Department of Anatomy of the University of Rostock, from a female body donor (79 years, 26 h postmortem interval) without any known history of neurologic or psychiatric disease.

The human brain was fixed in 4% buffered formalin. The vervet monkey brain was perfusion fixed with 4% paraformaldehyde solution, after deep anaesthesia of the vervet with ketamine–pentobarbital. The brain was then cryoprotected with 20% glycerol before freezing the tissue blocks. Serial coronal sections through the parietal and occipital lobes, 60 µm thick each, were obtained using a large-scale cryostat microtome (Polycut CM 3500, Leica, Germany). Blockface images of the tissue blocks were obtained during sectioning using a CCD camera to have an undistorted reference image for image realignment after sectioning. This was supported by rigid registration of a fiducial marker system (ARTag), on which the frozen tissue block was mounted (Fiala 2005; Wagner and Schmalstieg 2007). Sections were then mounted on glass slides and left unstained for measurements by means of the 3D-PLI setup.

All vervet monkey and human brain sections were measured using the 3D-PLI setup for ultra-high-resolution analysis of the fiber architecture (Axer et al. 2011a,b). Taking advantage of the birefringent property of the myelin sheath, signals were obtained that allowed to compute fiber orientation maps (FOM), in which the arrangement of nerve fibers was studied within the sagittal stratum and surrounding white matter of the occipital and parietal lobes.

Whole coronal sections of the vervet monkey brain and selected regions of interest in the human brain were scanned using a polarizing microscope (LMP-1, Taorad, Germany), with an in-plane spatial resolution of 1.3 × 1.3 µm. The polarizing microscope was equipped with a rotatable linear as well as a circular polarizer, in between which the sections were placed on a specimen stage, and a light source underneath emitting green light through this setup. Rotation of the circular polarizer allowed for image acquisition at 18 equidistant rotation angles between 0° and 170°. The final image of each section was reconstructed from a series of partially overlapping single tiles (2.7 × 2.7 mm). To extract homogeneous intensity profiles for reconstruction of the local fiber orientations, raw images were preprocessed to correct for inhomogeneities due to noise from external sources, such as illumination, dust, or filters (Dammers et al. 2010). This allowed for reconstruction of different image types by fitting the individual rotational images to a three-parameter model based on a Fourier series solution to the Jones calculus (Axer et al. 2011b), of which the transmittance, inclination and fiber orientation maps were used for the present analysis. The transmittance maps result from light scattering and extinction in the tissue, thus providing information about the overall microstructure within the section, originating from cell bodies and fibers (see Fig. 2). The transmittance maps were normalized by dynamic range. The FOMs represent the finally reconstructed and color-coded fiber orientations per pixel of a section, whereby the colors coded for fiber direction within the cutting plane, and the saturation indicating the inclination of the fibers (see color wheels, e.g., Fig. 3A). FOMs were weighted by the normalized transmittance images. As transmittance maps and FOMs are both calculated from the same preprocessed image, they are automatically coregistered, allowing for immediate comparison of different data modalities.

The sagittal stratum and surrounding white matter were analyzed on consecutive coronal sections, supported by an additional region-of-interest (ROI) approach to zoom into more detail. From the original whole-brain FOMs, ROIs were selected for visualization of the courses of the fibers within the sagittal stratum as well as of those entering, leaving and passing through. For visualization purposes, the ROI images of the FOMs were additionally contrast-enhanced for optimal presentation of the fiber orientation of interest. This enabled delineation of fiber tracts within consecutive sections by a trained neuroanatomist (SC), taking advantage of the very high in-plane resolution of 3D-PLI. It needs to be stressed that 3D-PLI does not provide a definite answer to the question if fibers are entering or leaving, since it does not contain information about the direction of fibers. To solve ambiguities, all results were compared to tracer studies in monkeys as provided by the present “gold standard” from Schmahmann and Pandya (2006). This allowed to interpret results obtained using 3D-PLI in the human brain, as prerequisite for mutually comparing human 3D-PLI results with diffusion-based tractography results in-vivo.

The sagittal stratum was clearly identifiable on transmittance maps of consecutive coronal brain sections of the vervet monkey brain by its very dark, almost completely black appearance (Fig. 2A). From rostral to caudal, the SSt followed the occipital horn of the lateral ventricle. Its appearance changed from a straight band rostrally, running from medio-dorsal to latero-ventral and spanning the hemisphere almost from the bottom of the intraparietal sulcus (ips) to the bottom of the occipito-temporal sulcus (ots), to a C-shaped structure more caudally, which became thinner and consecutively less well delineable as it approached the occipital pole.

Zooming in revealed a clear sublayering of the SSt into a thicker external and a thinner internal layer (SSte and SSti; Fig. 2B). The SSte and SSti were identified throughout the extent of the SSt within each section as well as on all sections from rostral to caudal (except for the caudal-most sections, where the structure of the sagittal stratum dissolved). The two layers of the SSt were distinguishable by different transmittance properties: while the SSte appeared very dark (comparable to the appearance of the complete SSt in the full section views as in Fig. 2A), the SSti appeared lighter, although still considerably darker than the surrounding white matter, e.g., the inferior longitudinal fascicle (ILF) or the tapetum (Tp). In the transmittance maps, this reflected differences in fiber orientation: while the SSte was dominated by fibers coming out of the sectioning plane (thus running in rostro-caudal direction), the SSti contained more fibers running within the sectioning plane, intermingling with the rostro-caudal ones.

In addition, there were very small fiber bundles and single fibers visible in both the transmittance maps (Fig. 2B) and FOMs (Fig. 3A, B), entering or leaving the SSte or crossing through to join the SSti.

These could be subdivided into two groups: One group of fibers was entering or leaving the SSte almost perpendicularly from the laterally adjacent white matter. As these fibers approached the SSti, they changed direction to join the SSti, visible by the fading out of the fibers’ color as an indicator of steeper fiber inclination out of the sectioning plane (Fig. 3C, D; asterisks at arrow heads). The second group of fibers came out of the tapetum and revealed a straight course from the medially located tapetum through the SSti and SSte. These fibers intermingled with those from the first group, but to different degrees: while there was pronounced mixture of fiber directions in the dorsal-most part of the SSt in both SSte and SSti (Fig. 3E), the fibers of the two groups seemed to be arranged in parallel to each other, a fiber of one group running in between two fibers of the other group (Fig. 3C, D).

The origin of fibers entering the SSt is known from tracer studies in monkeys. We used this information for exemplary tracts (according to the experiments performed by Schmahmann and Pandya 2006) as basis for the analysis of the FOMs of the vervet monkey brain.

Focusing on a section, where fibers from parieto-occipital area DP approached the dorsal part of the SSt in tracer studies (Fig. 4A), accompanied by fibers running in medial and ventral direction, revealed a similar overall fiber direction in 3D-PLI: from dorso-lateral parieto-occipital cortex to the dorsal tip of the SSt (Fig. 4B). Zooming into the transition zone from the deep white matter to the SSt showed that this fiber bundle could be subdivided into three parts (Fig. 4C): a lateral one coursing into the occipito-temporal white matter (color change from blue to light green); a prominent middle part, where fibers kept their direction and intermingled with fibers perpendicular to the sectioning plane (visible by the dark triangular space in between the colored fiber bundles); and a medial bundle joining the fiber tract underneath the IPS, i.e., the dorsal occipital bundle (dOB).

At a comparable dorsal location, but in a more rostral section, the fibers from DP entered the SSte together with, but clearly separable from the fibers from medial superior parietal area PO according to tracer experiments (Fig. 5A). In 3D-PLI, these separate directions of the fiber bundles from both sides of the IPS were clearly visible (Fig. 5B): while fibers from the IPL were running from latero-dorsal to medio-ventral (blue color), the fiber bundle of the SPL was running perpendicular to it, from medio-dorsal to latero-ventral (green color). Zooming into the dorsal part of the SSt, fibers from different directions intermingled (Fig. 5C): fibers from the SPL entered the SSt at the medial side and changed direction upon entrance (change from green to yellow color and by the flashlight view with only fibers of relevant direction and inclination colored, Fig. 5E). Fibers from the IPL entered from the lateral side, keeping their latero-dorsal to medio-ventral course, as visible by the single fibers colored in blue within the dorsal SSt (particularly in the flashlight view of highlighted fibers in Fig. 5D). These two fiber directions additionally crossed through the tapetum fibers, entering the SSt from the medial side (yellow arrow).

Fibers within the SSt associated with area DP could, furthermore, be distinguished from fibers associated with more ventral areas, e.g., areas V4 and V4v. According to tracer experiments (Schmahmann and Pandya 2006), these fibers enter the SSt either at its ventral tip (area V4v) or with several separate small fiber bundles in its large middle part (area V4; Fig. 6A). Zooming into the respective regions revealed a similar pattern in 3D-PLI (Fig. 6B): fibers from the direction of area V4, i.e., from lateral temporo-occipital cortex bent around the STS, crossed through the ILF (red–orange coloring of the fibers vs. yellow coloring of the ILF) and entered the SSt in small parallel fiber bundles (Fig. 6C, fibers in red color within SSt, particularly visible in the flashlight view, Fig. 6D). These bundles could be clearly separated from those coming from the dorsal aspect of the section and thus from parieto-occipital cortex (Fig. 6E): the latter were not separated into different small bundles, but appeared as one large bundle of many single fibers. Furthermore, the direction differed, coming either from lateral (small fiber bundles colored in red) or from dorsal (fibers colored in light blue). Fibers at the ventral tip of the SSt, contrarily, showed a course oriented from latero-ventral to medio-dorsal (Fig. 6F, fibers colored in red–yellow). Within the underlying white matter, the course of these fibers before they enter the SSte could be determined, parallel to the adjacent U-fibers and crossing through the vertically oriented part of the ILF.

Having demonstrated similar evidence for the overall structure and courses of fibers entering and leaving the SSt in its course through the parieto-occipital white matter in previous tracer experiments and current 3D-PLI analyses in the same non-human primate lineage (macaques and vervets) led back to the question if with the new 3D-PLI technique, a comparable SSt structure might be detectable in humans. The overall appearance of the SSt with a clear delineation from the surrounding white matter and an internal and external layer is indeed comparable (Fig. 7A). Similarly, fibers entering/leaving the layers of the SSt from the tapetum or through the ILF are clearly visible in a high-resolution view (Fig. 7B), suggesting an at least similar or even more complex fiber architecture in humans as compared to non-human primates.

A bilaminar configuration was found, reflecting two separate rostro-caudal fiber bundles, namely, the optic radiation in the external and the cortico-subcortical projection tracts in the internal layer (Flechsig 1896; Pfeifer 1925; Sachs 1892; Schmahmann and Pandya 2006). Supplementing these studies, the current data showed a clear layering into an external and an internal segment throughout the whole extent of the sagittal stratum. As has been described previously, the layering of the sagittal stratum dissolves in caudal sections close to the occipital pole, as does the sagittal stratum as a whole (Vergani et al. 2014). Approaching the occipital pole region means that less cortical regions were present, until only primary visual area V1 is visible. This phenomenon was particularly pronounced in the vervet monkey brain as V1 covers major parts of mesial and lateral aspects of the occipital lobe. In the human brain, V1 (Brodmann area 17) is mainly confined to the mesial side and only partially encroaches on surrounding occipital lobe at the very caudal end (Amunts et al. 2000). This suggests that only the optic radiation and the external layer of the sagittal stratum are present when approaching the occipital pole. The optic radiation then fans out to reach all parts of V1 and adjacent areas V2 and V3v and V3d (Alvaréz et al. 2015; Wandell and Winawer 2011), resulting in a disappearance of the sagittal stratum, since the fibers are not as parallel as in rostral sections.

The internal layer of the sagittal stratum appears in rostral sections as more and more fibers from occipital, parietal and temporal areas enter the sagittal stratum. These fibers necessarily need to cross through the external layer of the sagittal stratum. This has been clearly shown in tracer experiments in monkeys, where the fibers from a particular brain region can be followed through the white matter as they enter and cross the external layer of the sagittal stratum and join the fiber bundle running within the internal layer on their way to the subcortical target zones (Schmahmann and Pandya 2006). These tracing results have been used as reference in the present study. Shown for different exemplary tracts, the pattern of fiber architecture as revealed by the tracing experiments could be clearly shown by means of 3D-PLI. Of note, not only the global connection could be demonstrated, e.g., that fibers from superior and inferior parietal or ventral occipital cortex enter the sagittal stratum at different dorsal or ventral levels. In addition, the exact distribution of fibers within the sagittal stratum could be reconstructed. This was particularly visible for the fibers originating in the region of area V4 and V4D as compared to those coming from parietal areas. According to tracer experiments (Schmahmann and Pandya 2006), the fibers from V4/V4D enter the sagittal stratum in its middle sector in several parallel interspaced fiber bundles with latero-medial orientation, while the fibers from parietal areas entered the sagittal stratum in its dorsal sector as a large bundle with dorso-ventral orientation. This differential pattern of several interspaced vs. one large fiber bundle could be clearly shown in the 3D-PLI sections of the present study. Thus, without using a tracing technique with injections, 3D-PLI is able to show the exact same connection pattern in the same non-human primate lineage. Providing the same results as the gold standard for studying structural connectivity, e.g., of interspaced separated fiber bundles in the middle of the SSt, which was also reported in the tracer experiments of Schmahmann and Pandya (2006), could be regarded as an external validation of the novel 3D-PLI technique. On the one hand, this facilitates systematic assessments of the complex white matter architecture and the different fiber bundles in one and the same brain (as compared to separate tracer injections in different hemispheres). On the other hand, it enables translating this approach to the human brain, where tracing using injections is ethically not feasible. 3D-PLI results obtained in the human brain could then be more easily understood as adequately revealing the fiber microarchitecture, which is a prerequisite for consecutive comparison and cross-validation with DWI data.

Thus, 3D-PLI could provide the bridge between tracing and diffusion imaging both within and across species (where comparable white-matter occipital fiber architecture can be detected, see, e.g., Kaneko et al. 2020), as the relevant intermediate step (Caspers and Axer 2019) and a valuable complement to comparisons between DWI and tracing experiments which also revealed comparable results (Schmahmann et al. 2007). At this, ex-vivo diffusion MRI serves as an additional relevant component when bridging between the scales (Roebroeck et al. 2019), as it could reach high resolutions on the mesoscale in the range of tens to hundreds of microns in different species (Aggarwal et al. 2015; Bech et al. 2020; Fritz et al. 2019; Liu et al. 2020; Ly et al. 2020; Wang et al. 2020).

It was also visible in the 3D-PLI data that these entering fibers indeed joined the internal layer of the sagittal stratum: while they could be clearly seen within the external layer, they faded out at the transition to the internal layer. This was indicative of fibers turning out of the sectioning plane into a perpendicular course, which in the case of our coronal brain sections pointed at a rostro-caudal trajectory. Following them continuously to their final subcortical target regions is challenging at the moment as fibers running out of the sectioning plane appear black in the 3D-PLI fiber orientation maps. This would be alleviated by 3D tractography in reconstructed stacks of 3D-PLI sections in the future.

We could additionally show that fibers crossed through both the external and internal segment of the sagittal stratum, originating in the medially adjacent tapetum. The tapetum contains the callosal fibers of the occipital and neighbouring parietal and temporal cortex. To reach their target zones within the cortex (Caspers et al. 2015; Clarke and Miklossy 1990), they have to cross through the sagittal stratum. We could show that they could be clearly distinguished from the fibers entering the sagittal stratum from the surrounding white matter, not only because of their major direction, but also by their complete course from leaving the tapetum, intermingling with and crossing through the different fibers of the sagittal stratum to leaving the sagittal stratum laterally.

Results obtained in this study rely on the analysis of one hemisphere of a vervet monkey brain and an exemplary comparison to a human brain section. Thus, future studies with a larger sample size are desirable to further verify the observations on the fiber architecture obtained here.

The packing density of fibers within the deep white matter of the human brain is even higher than in the monkey brain. This challenges current in-vivo diffusion imaging approaches, aiming at studying smaller bundles which cross the large fiber tracts (Dell’Acqua and Catani 2012; Jbabdi and Johansen-Berg 2011). This is particularly true within the occipital lobe, where several large fiber tracts run in parallel, mainly in rostro-caudal direction, such as the optic radiation (as part of the sagittal stratum), the inferior longitudinal fasciculus, the callosal fibers within the tapetum, or the inferior and superior fronto-occipital fasciculi (Catani et al. 2003; Cristina et al. 2014; Forkel et al. 2014; Takemura et al. 2020). Using additional MR contrasts, it was possible to identify another relatively large fiber tract perpendicular to them, running in dorso-ventral direction lateral to the sagittal stratum and crossing through the ILF, i.e., the vertical occipital fasciculus (Bugain et al. 2021; Takemura et al. 2016; Yeatman et al. 2014). These large bundles surrounding the sagittal stratum have also been identified in blunt dissections (De Benedictis et al. 2014).

When it comes to the fine structure beyond these large tracts, e.g., within the sagittal stratum, both dissection and diffusion imaging seemed to have reached their limits in terms of resolution: the bilaminar structure of an internal and external sagittal stratum could only partially be identified (Vergani et al. 2014).

Functional in-vivo mappings during neurosurgery suggest that several complex higher order functions (e.g., hemi-agnosia, spatial neglect, known to involve several occipital and parietal cortical regions) were hampered during direct electrical stimulation of the SSt, indicating involvement of different fiber systems intermingling within the SSt (Berro et al. 2021). The current results shed new light on this issue: the fibers enter, leave or cross through the sagittal stratum in very thin bundles or even as single fibers. This is far below the spatial and angular resolution of routine in-vivo diffusion imaging studies, even of high-angular resolution diffusion protocols, when it comes to identifying single fibers or disentangling crossings of such small bundles. Using advanced fiber orientation reconstruction algorithms, such as constrained spherical deconvolution (Tournier et al. 2007), might enable detection of some additional fiber orientations, but averaged across a considerably large volume of a few cubic millimeters.

The analysis of ultra-high-resolution images resulting from 3D-PLI showed that there are indeed small fiber bundles which cross through the inferior longitudinal or vertical occipital fasciculus on their way to or from the sagittal stratum. They could not only clearly be followed as they crossed through the external sagittal stratum to join the internal sagittal stratum, but also as they approached the sagittal stratum and crossed the large white matter tracts surrounding it.

Thus, the additional fiber directions as visible in the diffusion imaging data (Fig. 1C, D) might indeed hint at such anatomical fine structure of the sagittal stratum. Considering the reconstructed fiber orientations from the diffusion signal alone would prevent any distinction between a correct depiction of the underlying anatomy or false positives due to noise in the data (see, e.g., Leuze et al. 2021). Combination with an ultra-high-resolution technique such as 3D-PLI may provide detailed insights into the fiber architecture of the brain by decoding the local fiber configurations (Caspers and Axer 2019). This particularly holds true for regions with one major fiber direction dominating the local architecture, such as the sagittal stratum, which could lead to neglect of the minor directions. Based on the examples shown here for the human brain with a highly similar pattern of fiber architecture as found in the vervet monkey, future studies are needed to follow the individual fiber tracts which join the internal layer of the sagittal stratum to reach their subcortical target regions. This might be facilitated in the future by fiber tractography based on reconstructed stacks of 3D-PLI sections, based on aggregated 3D-PLI-derived fiber orientation distribution functions (Alimi et al. 2020; Axer et al. 2016; Reckfort et al. 2015; Schmitz et al. 2018), in combination with ultra-high-field ex-vivo DWI (Aggarwal et al. 2015; Fritz et al. 2019; Ly et al. 2020), as has been shown, e.g., for the trigeminal pathway in the human brainstem (Henssen et al. 2019). With ex-vivo DWI being inherently 3D when applied to a whole tissue volume, it could serve as relevant reference for the tractography data derived from 3D-PLI. With the high resolution of 3D-PLI of about one micrometer in-plane across species, these approaches are mutually complementary and could thus be used in combination in the future to understand the complex multi-level fiber architecture of the brain.

Similar to the knowledge gained from respective tracing experiments in monkeys, this high-resolution fiber tract analysis in humans is of particular importance to understand which parts of occipital and adjacent parietal and temporal cortices are connected with which subcortical structure, e.g., within the thalamus or among the pontine or mesencephalic nuclei. This will enable a better understanding of cortico-subcortical feedback loops for top-down control of incoming information from the subcortical relay stations as well as information flow to subcortical control centers.