Cortical microstructure and hemispheric specialization – a diffusion-imaging analysis in younger and older adults

Hemispheric lateralization in structure and function

Based on comparative histology and DTI studies, we hypothesized to detect correlates of the structure-function relationship in the fundamental concept of hemispheric lateralization. Cortical FA was assessed as a surrogate parameter for cortical tissue complexity (34-36). Cortical AD was hypothesized to reflect the density of neurons, as these oblong structures facilitate diffusion along tensors’ main axis. Cortical RD was assumed to primarily correspond to structures facilitating diffusion oriented in parallel to the cortical surface, such as dendrites and interneurons. We hypothesized that hemispheric lateralization in these diffusion parameters would mirror hemispheric specialization, known from histology studies, particularly for the language or motor systems. We identified five regions exhibiting pronounced hemispheric specialization.

The first two regions were the planum temporale and the inferior frontal gyrus comprising Wernicke’s and Broca’s areas, core regions of the human language network. In the planum temporale we observed lower left-hemispheric FA, which could be induced by two different contributors. First, the accompanying reduction in left-hemispheric AD stands in line with existing histology findings: Greater intercolumnar gaps between neurons with less oblong neurons compared to the right-hemispheric homologue (43-45) can lead to smaller AD values and FA values on the left side. Second, denser and more widespread dendritic arbors in the left planum temporale (46-48), extending to a certain degree parallel to the cortical surface, might suggest a plausible link to the greater RD values observed in younger adults. Earlier studies proposed RD to be inversely linked with neuron density due to the influence of myelination (49). Broca’s area (assessed via the triangular inferior frontal cortex) has been characterized in a review by Keller et al. as poor in asymmetry on the macrostructural level and showing microstructural lateralization “more often than not” (10). While previous microscopy studies have reported multiple parameters to be lateralized in planum temporale, only a lower grey-level index (GLI) was observed for Broca’s area (50, 51), suggesting lower fraction of perikaryal structures, which could be based on greater amount of dendrites. One previous imaging study observed cortical microstructural lateralization for Wernicke’s area, but not for Broca’s area (19). In the present analysis, we found greater FA values in the left Broca’s area driven by data from younger adults. The link between cortical FA and the GLI remains speculative as the latter reflects the fraction of Nissl-stained ribosomal structures, and not the number of neurons per se as glia cells would also contribute to GLI. Such cell populations are less elongated, not as uniformly arranged as neurons in the cortex and highly different in cytomorphology. In contrast to the planum temporale, due to absence of positive AD or RD findings it remains elusive if the difference in FA is driven by differences in complexity or in density of axonal structures.

As a third region, greater AD and RD were measured in the left, dominant hemispheric precentral gyrus which comprises the primary motor cortex. It was striking not to observe a lateralization of FA in the precentral area. However, given the nature of FA as a ratio, even a more complex tissue may still show unchanged FA, if the amount of diffusivity along neuronal columns remains relatively equal compared to the diffusion in the plane perpendicular to neuronal column’s axis. This observation compels to evaluating cortical diffusion imaging in the light of diffusivity measures of AD and RD to allow more precise statements on the microstructure. For instance, greater AD in the left and dominant motor cortex indicates greater diffusivity perpendicular to cortical surface and could therefore corresponds to the neuronal density, which was shown to be greater in the left primary motor cortex (52). In turn, a greater left RD (diffusivity parallel to the cortical surface), significant only in older adults, could point towards a greater amount of dendritic or inter-neuronal structures.

Fourth, lateralization could be found beyond the eloquent areas of language and motor networks. Lower FA, AD, and greater RD in the left-hemispheric inferior parietal cortex and the temporo-parieto-occipital junction could correspond to specialization of the parietal cortex in higher-order integrative functions that was observed in imaging and stimulation studies (8, 53), while ischemic lesions affecting right inferior parietal cortices have been linked with neglect and apraxia (54, 55). Strikingly, one would expect greater AD and RD, in particular when considering the specialization, in the right hemisphere as these could represent greater neuronal count and dendritic arborization or interneuronal density in the specialized area. However, as the visualized data for the angular gyrus from both subgroups suggest (see Supplementary Figure 3), the lateralization of AD and RD becomes apparent in late adulthood. Concurrently, the absolute level of AD is greater and the absolute level of RD is lower in the older group, which could correspond to a greater axonal density and lower dendritic axonal complexity, induced by reduction of non–axonal structures along a constant amount of neurons. This is in line with earlier descriptions in review literature (56).

Fifth, the observed lateralization of the visual system in younger adults was a striking finding. Even though there are single reports on interhemispheric differences in activity of the calcarine cortex (9), the literature on this topic is sparse compared to data on asymmetry in motor and language function. Future studies will have to confirm this result.

Taken together, the number, the topography, and the side specificity in lateralization of cortical DTI measures encourage viewing these as a promising surrogate for functional hemispheric specialization. For instance, the RD findings appear to resemble histological findings on dendritic arborization in the planum temporale (43, 47) while AD lateralization corresponds with findings on neuronal/axonal density in the planum temporale (43, 57), the fusiform cortex (58) and the inferior parietal lobe (59). This agreement of lateralized cortical DTI metrics and classical histology supports the view that DTI is able to capture cortical microstructural traits which remain hidden for CT based morphometric analyses in the motor or language network. The absence of a positive finding for CT compared to the results of Kong et al. could rely on the larger sample size (n > 17000) which indicates that surrogates for microstructure could be not only more specific but also more sensitive (15).

Influence of age on microstructural hemispheric specialization

Our second hypothesis was that cortical DTI metrics are able to detect age-dependent changes in hemispheric lateralization. Multiple functional MRI and EEG studies have shown alterations in hemispheric activation with age, particularly in prefrontal cortices which has been contextualized with theoretical models such as the HAROLD model (4, 60). Reduction in hemispheric asymmetry of functional activity in older adults (i.e. the HAROLD model) has been argued to reflect compensatory mechanisms to overcome a decline in cortical integrative function (9). In this study, the inferior frontal sulcal and the inferior parietal cortex, two heteromodal integrative areas, displayed an age-dependent increase in lateralization due to a stronger right-hemispheric lateralization in absolute measures (Figures 1 and 2 – fourth column, Supplementary Figure 3). Such dynamic of lateralization towards the opposite hemisphere is highly suggestive for a reorganization that might be linked to compensatory activity. From a clinical perspective, the distribution of significant findings in the inferior parietal cortex would be in line with the assumption that cortical aging and degenerative disorders primarily affect higher cognitive and integrational processes due to subtle reorganization rather than the function of primary areas (61). Even though the second hypothesis appears to be confirmed considering the finding of lateralized RD in prefrontal microstructure, the results’ distribution (Figures 1 and 2 - fourth column), is highly suggestive that age-dependent organization is more strongly pronounced in posterior brain regions. Besides the HAROLD model, aging-related processes have been proposed to show a posterior to anterior shift, contextualized by the PASA model (62). The findings in the TPO cluster in this study seem to support the assumption of this shift, which has already been discussed by Rathi et al. in analyses on DTI measures on a gross, lobar level (37).

Of note, the focus on structure in this study allows only speculative judgements on the underlying mechanisms. The morphology of the aging cortex has been characterized by distinct and focal changes in dendritic branching rather than by extensive neuronal loss (56), which would be primarily reflected by a reduction of RD and to lesser extent also in FA. Such reorganization would unlikely cause reorganization measured by AD. However, an altered diffusion along the neuronal columns may be also explained by altered afferent fibers resulting from white matter lesions leading to a disconnection syndromes due to Wallerian degeneration (16, 17). An earlier study of our group showed that better performing older adults show a greater white matter integrity, operationalized by regional FA, compared to low performing peers (39). Again, cortical DTI measures, grasping such nuances based on single-shell imaging, appear to be superior to CT based measurements (15).

Implications

Cortical DTI with apparent advantages over CT bears a great potential for future clinical research questions. Current implications primarily focus on diagnostics. The observed relationship between AD and neuronal density as well as between RD and dendritic arborization may encourage researchers to further promote the usage of cortical DTI to address potential cortical microstructural alterations in neurological diseases (63). Furthermore, a better characterization of temporal and spatial aspects of neuroplasticity could reveal targets for stimulation techniques, secondarily improving their efficacy and efficiency.

Methodical limitations

There are several important limitations to note: First, five different studies were re-analyzed in this work. We controlled our models for study category, despite the acquisition on the same scanner. The inconsistent T1-data resolution in two studies is unlikely to confound the results as it had to be downgraded to the resolution of the diffusion data prior to DTI measure estimation. Second, the study does not allow a lifespan analysis due to the cross-sectional character, indicating the necessity to verify our observations in a longitudinal dataset. Third, beside absent information on education and socioeconomic background the behavioral data of some studies lacked standardized handedness testing such as the Edinburgh Handedness Inventory. Instead, the participants had to declare their handedness, which could potentially lead to inclusion of participants trained to right-handedness. Fourth, the diffusion image resolution was 2mm, this might be rather lower when aiming to assess cortical microstructure. However, given the strong differences observed in a conservative analysis and standing in line with existing histology reports, we consider the results as robust. Nevertheless, cortical DTI analyses should be further improved using high-resolution protocols in the future. Fifth, due to existing but still sparse literature it is challenging to compare the absolute DTI measures results to other studies. Even though we could reproduce insular cortex values known from previous studies (64), future work must validate our results. Differences in scanners, resolution of MRI datasets and different approaches of preprocessing might influence cortical DTI metrics. However, we assume that measuring lateralization, in which every participant serves as his or her own control, sufficiently accounts for this limitation.

Conclusion

Cortical DTI measures unambiguously grasped microstructural hemispheric lateralization in core areas known for asymmetry in function such as the motor and language networks, with FA showing different degree of variation between areas of higher and lower integrative function. This work further demonstrated age-related reorganization processes in integrative heteromodal, secondary cortices, which corresponds to findings from studies on cerebral activity. From a methodical perspective the present findings indicate, that (1) cortical DTI grasps not only myelinated structures within white matter but also the complexity from non-myelinated tissue in the cortical gray matter and (2) the diffusivity measures AD and RD may reflect specific histoarchitectural traits, such as neuronal density and dendritic arborization, being therefore compulsory for interpretation of FA in the cerebral cortex.

Participants Data

Imaging and clinical data were screened from five individual datasets contributing to studies which have previously been published by our laboratory (38-42). Datasets of healthy right-handed participants within two age-ranges (18-30 and 60-87 years) with sufficient data quality were included in this secondary analysis. In total, 91 of initially 123 datasets were analyzed (number of included datasets from each study - Study A (38): 14 of 23 [older adults], Study B (39): 39 of 50 [19 younger and 20 older adults], Study C (40): 18 of 20 [younger adults], Study D (41): 8 of 13 [6 younger and 2 older adults], Study E (42): 12 of 17 [older adults]). The exclusions were based on age beyond the range of interest and data quality. In sum, age groups consisted of 43 younger (mean 24.4 years, standard deviation [SD] 2.5, range 18-30, 26 female) and 48 older participants (mean 71.7 years, SD 6.0, range 60-87, 22 female). Brain imaging for all participants had been conducted on the same MRI scanner. The original studies were conducted following the Declaration of Helsinki and approved by the local ethics committee of the Medical Association of Hamburg. All participants gave written informed consent.

Brain Imaging and Processing

All participants were scanned on a 3 Tesla Siemens Skyra Scanner (Siemens Healthineers, Erlangen, Germany) with a 32-channel head coil. T1-weighted imaging was based on a magnetization-prepared, rapid acquisition gradient-echo sequence (MPRAGE) (repetition time (TR) = 2.500ms, echo time (TE) = 2.12ms, 256 slices with a field of view (FOV) = 240×192mm, interslice distance of 0.94mm and an in-plane resolution range between 0.94x0.94mm and 0.83×0.83mm). DTI datasets were consistent in imaging parameters, with 75 slices in 64 non-collinear gradient directions with a b-value of 1500s/mm² as well as one b₀ image (FOV = 208 × 256 mm², flip angle = 90°, TE = 82ms, TR = 10000ms and voxel size of 2×2×2mm). Each individual dataset underwent manual quality check excluding datasets with motion artifacts, for instance ringing, ghosting or tissue cuts, leading to insufficient reconstruction of grey and white matter boundary, and datasets with cerebral tissue outside the FOV.

T1-weighted images were segmented into 74 cortical anatomic regions per hemisphere from the Destrieux atlas (65) using the FreeSurfer based recon-all tool (version 6.0.1) (66) yielding CT values and individual labels for identification of regions in DTI data. The data was post-processed and registered to the individual pre-processed DTI using Advanced Normalization Tools (ANTs 2.3.4) (67) while adjusting the resolution to 2x2x2mm. DTI datasets were corrected for eddy currents using MRtrix3 (3.0.2) tools based on the FSL (6.0.1) correction method (68, 69). Distortion correction was applied by registering the b₀-weighted diffusion image to the T1-weighted image using ANTs, MRTrix3 and FSL. For free water (FW) correction of the cortical DTI signals, a custom written MATLAB script was used (ran on Matlab R2020a, The Mathworks, Natick, MA, USA). In brief, a bi-tensor model with a fixed diffusivity for FW and a second representing water diffusion in presence of tissue membranes (33) was fitted and mean FA, AD and RD values were derived from FW-corrected tensor eigenvalues for each cortical area.

Statistical Analysis

Statistics were performed using R software version 4.0.2 (70). Linear mixed-effects models with repeated measures were fitted to describe the extent of label-wise hemispheric lateralization in cortical microstructure. Specifically, in separate models for FA, AD, RD, and CT in all participants (N=91), mean absolute values were treated as the dependent variable (DV), SIDE (left, right) was treated as the independent variable of interest. Target effects were adjusted in a multivariate approach for the nuisance variables AGE_GROUP, AGE_GROUP × SIDE interaction, STUDY, SEX, as well as CT in models for diffusion measures. The necessity to adjust for CT is explained in the supplementary text 1, see Supporting Information. The amount of hemispheric lateralization is based on the coefficient SIDE (mean difference between hemispheres), its statistical significance (P_SIDE) is fully corrected for multiple comparisons using the false discovery rate (FDR) method for 74 Destrieux atlas regions per DTI parameter. The age-dependent change of lateralization is based on the coefficient AGE_GROUP × SIDE, its statistical significance (P_{AGE_GROUP × SIDE}) was corrected in the same manner as the SIDE effect. To specifically address age effects, separate models were also fitted for the younger (N=43) and older (N=48) participants separately, with AGE as covariate. Statistical significance was assumed at corrected P_FDR<0.05. Brain illustrations were generated using the fsbrain package in R.