Association of cerebral blood flow with myelin content in cognitively unimpaired adults

Introduction

Production and maintenance of myelin integrity through oligodendrocytes is critical for saltatory conduction and normal axonal function. Indeed, accumulating evidence is establishing a close relationship between myelin degeneration and several neuropathologies, including multiple sclerosis and dementia.1 2 In animal studies, it has been shown that oligodendrocytes are vulnerable to blood flow deficits, and loss of these cells may occur rapidly in response to reductions in blood flow.3 Indeed, myelin maintenance through oligodendrocyte metabolism is an energy-intensive process, so that myelin homeostasis is particularly sensitive to hypoxia, hypoperfusion or ischaemia.4 In addition to substrate delivery, adequate cerebral blood flow (CBF) is crucial for removal of metabolic by-products and neurotoxins. The oligodendrocyte cells and myelin sheets are vulnerable to various insults, including iron accumulation as well as aggregations of tau and amyloid-beta proteins.5 Aside from potential neuronal damage, these insults can lead to loss of oligodendrocytes or impairment of their myelin synthetic capacity; this may result in deficits in myelin production and repair during turnover, or frank demyelination.

Growing evidence indicates that the breakdown of the myelin sheath may be an early phenomenon in neurodegeneration, including Alzheimer’s disease.1 6 7 Furthermore, there is evidence of a direct association between reduced brain perfusion and neurodegeneration.8 These findings suggest a positive association between hypoperfusion and white matter damage. Surprisingly, only a few magnetic resonance imaging (MRI)-based studies have examined the potential association between white matter integrity and CBF status. These studies have been limited to the context of leucoaraiosis, white matter lesions and Parkinson’s disease,8–10 all indicating that reduced CBF may induce brain tissue degeneration. However, association between blood supply and tissue integrity in unimpaired subjects has received very little attention. Yet evaluating the extent and patterns of this potential association in normal ageing is a critical step towards understanding the pathophysiological basis of the mechanisms and outcomes of neurological diseases.

Two previous pioneering MRI studies of the age-dependent relationship between CBF and white matter integrity have been conducted on healthy adults.11 12 Using diffusion tensor imaging (DTI) to assess white matter integrity and arterial spin labelling (ASL) to quantify CBF, Chen and colleagues11 found that cortical CBF is associated with white matter integrity, and a greater association of radial diffusivity, a DTI metric generally used to probe demyelination, with CBF, as compared with other DTI metrics; this was interpreted as an indicator of myelin degeneration. Furthermore, Giezendanner and colleagues12 found that subcortical CBF is associated with the integrity of different white matter tracts. All these compelling findings suggest that the blood supply to the brain may be an important determinant of white matter health in normal ageing. However, DTI outcomes, including fractional anisotropy and radial diffusivity, while sensitive to white matter microstructural changes, are not specific. Indeed, multiple factors aside from myelination can affect the DTI-derived eigenvalues from which fractional anisotropy, radial diffusivity and other DTI indices are derived; these include axonal degeneration, hydration, temperature, flow, macromolecular content and architectural features, including fibre fanning or crossing. The potential association of CBF with a more specific index of myelin content has not yet been undertaken.

Advanced MRI methods based on multicomponent relaxometry to assess myelin water fraction (MWF), a surrogate of myelin content, have led to much greater specificity in non-invasive MRI myelin mapping.2 13 14 We previously developed the Bayesian Monte Carlo multicomponent-driven equilibrium single pulse observation of T₁ and T₂ analysis (BMC-mcDESPOT) as an alternative approach to multicomponent relaxometry.15–17 This method provides rapid, accurate and precise whole-brain MWF maps,15–18 and has been used to provide quantitative evidence of myelin loss in mild cognitive impairment and dementia and to investigate myelination patterns in normative ageing.1 19 20

Here, we investigated the potential association between CBF and local myelin in a large cohort of well-characterised adults with no cognitive impairment (n=67), across the extended age range of 24–88 years. The inclusion of subjects across a wide age range ensures a large dynamic range for blood flow and myelin content measurements. Myelin content was measured using BMC-mcDESPOT-based MWF,16 17 a specific measure of myelin content, as well as longitudinal and transverse relaxation rates (R₁ and R₂), sensitive but non-specific myelin measures, to provide contact with previous studies.21–23 Indeed, R₁ and R₂ values depend on water mobility as well as macromolecular tissue composition, including local lipid and iron content; these are the main constituents of myelin. Therefore, changes in R₁ and R₂ are directly associated with microstructural changes, including changes in myelin content. CBF was measured using the nonlocal estimation of multispectral magnitudes (NESMA)-ASL analysis for accurate and precise CBF determination.24 Our goal is to characterise the regional associations between cortical or subcortical CBF and local myelination variations in critical white matter regions.

Materials and methods

Results

Figure 1 shows the MWF, R₁, R₂ and CBF maps derived from the brain of all participants within the youngest decade of our cohort, that is, between 24 and 29 years. This limited age range minimises the potential effect of age on derived values. Visual inspection indicates that, overall, participants with the smallest values for regional MWF, R₁ or R₂ exhibit the smallest regional CBF values, while participants with the highest regional MWF, R₁ or R₂ values exhibit the highest regional CBF values. These qualitative results indicate a potential association between CBF and myelin content variations, seen in all three of these MRI measures, but most directly through the MWF results.

Request permission

Figure 2 shows the results of the regressions of MWF with CBF, after adjusting for age and age², for the six indicated WM and GM regions. As seen, decreased CBF corresponds to decreased MWF in all ROIs examined, with the best-fit curves displaying regional variation; these results agree with those shown in figure 1. Statistical analysis indicates that this positive correlation between MWF and CBF was statistically significant (p_CBF<0.05) or close to significance (p_CBF<0.1) in most WM and GM brain regions (table 1), with the exception of the ROIs defined by the whole-brain WM and the cerebellum WM. In addition, the greatest positive slopes in CBF with MWF were found in the frontal and temporal WM lobes regions, while the smallest slope was found in the whole-brain WM region. Comparison of these slopes indicated statistically significantly different slopes between the temporal WM lobes and the whole-brain WM region (p<0.05; Z-test computed as the difference between the two slopes divided by the square root of the sum of the squared SE of the slopes33). Furthermore, as expected, the effect of age was significant for all brain regions evaluated (table 1). Similarly, the quadratic effect of age, age², was significant or close to significance in all brain regions (table 1).

Request permission

Figure 3 shows the results of the regressions of R₁ with CBF for the six indicated WM and GM regions. Visual inspection indicates that decreased CBF corresponds to decreased R₁ in most of the ROIs examined, with the best-fit curves displaying regional variation, in agreement with the results presented in figure 1. The statistical analysis indicates that this positive correlation between R₁ and CBF was statistically significant (p_CBF<0.05) or close to significance (p_CBF<0.1) in all WM and GM brain regions (table 1), except in the whole brain, parietal lobes and cerebellum WM and the frontal lobes GM. In addition, the greatest positive slopes in CBF with R₁ were found in the frontal and temporal WM lobes regions, while the smallest slopes were found in the parietal lobes GM and cerebellum GM regions. Comparison of each of the greatest and smallest slopes indicated statistically significantly different slopes between temporal lobe WM and the parietal lobes GM and cerebellum GM. Furthermore, the main age effect was significant (p_age<0.05) in only one brain structure, namely the parietal lobes WM, after FDR correction (table 1). In addition to this region, the effect of age was significant or close to significance in the whole brain, frontal lobes, parietal lobes, temporal lobes WM, and the parietal lobes and temporal lobes GM before FDR correction. The quadratic effect of age, age², was significant or close to significance in all brain regions except in the cerebellum WM (table 1).

Request permission

Figure 4 shows the results of the regressions of R₂ with CBF for the indicated six WM and GM regions. As seen, decreased CBF corresponds to decreased R₂ in most ROIs examined, with the best-fit curves displaying regional variation; these results are in overall agreement with figure 1. Statistical analysis indicates that this positive correlation between R₂ and CBF was statistically close to significance (p_CBF<0.1) in only three WM brain regions (table 1), namely the whole brain, the occipital lobes and the cerebellum GM. In addition, the greatest positive slope of CBF versus R₂ was found in the temporal lobes WM, while the smallest slope was found in the parietal lobes WM. Comparison of the greatest and smallest slopes indicated non-significant differences. Furthermore, the main age effect was significant in all brain structures (table 1). Similarly, the effect of age² was significant in all brain regions (table 1).

Request permission

Discussion

Using advanced MRI methodology for myelin content quantification and CBF mapping, we have provided what we believe is the first demonstration of associations between cortical and subcortical variations in blood supply and a direct measure of myelin content. These associations were observed in a cohort of healthy subjects and were significant in critical brain regions. Notwithstanding the fact that these results do not prove causality, they provide evidence that limitations in CBF may lead to lower myelin content, thus linking neurovascular physiology and myelin integrity throughout normative ageing.

Although not specific to myelin content, R₁ and R₂ are very sensitive to myelin.1 21 We found significant regional decreases in R₁ with CBF (figure 3, table 1). This further supports our hypothesis of potential associations between CBF and myelin content as measured using a direct and specific myelin content measure, MWF (figure 2, table 1). Indeed, while not specific to myelin content, R₁ is sensitive to lipid content,21 the main constituent of myelin. Of note, this association was less pronounced between CBF and R₂ (figure 4, table 1); this is expected since R₂ is sensitive to other tissue properties including hydration, macromolecular content, temperature and flow, while being less sensitive to lipids.21 Studies have indicated that the R₂ to R₁ ratio could serve as a sensitive metric of myelin content. Our analysis, detailed in the online supplementary material, indicates non-significant correlation between R₂/R₁ and CBF in all ROIs (online supplementary table 1). Interestingly, unlike MWF, R₁ and R₂, there was no significant correlation between R₂/R₁ and age² (online supplementary table 1). These results indicate that the R₂ to R₁ ratio may be less sensitive than MWF as a marker of myelin. In fact, we observed only low-to-modest regional correlations between R₂/ R₁ and MWF (online supplementary figure 1); this agrees with the recent observation of Uddin and colleagues.34

Our results also indicate an inverted U-shaped association of MWF with age in all cerebral ROIs investigated (table 1), consistent with our previous study of a larger cohort of participants with no cognitive impairment, and with other previous work.19 31 We attribute this association to myelination from youth through middle age, followed by demyelination in later years,23 31 in agreement with postmortem observations35 and with different investigations based on myelin-sensitive, but non-specific, quantitative MRI methods such as DTI or relaxometry.22 23 36

We found that the brain regions investigated exhibited different trends of CBF versus myelination. Indeed, autopsy studies of human brain microvasculature have revealed a complex brain vascular system exhibiting regional differences in blood supply.37 This complex vascular network offers some protection to certain cerebral structures, while leaving other areas vulnerable to chronic hypoperfusion, especially the deep white matter regions.37 Moreover, studies have shown that, overall, the occipital lobes exhibit delayed demyelination as compared with the other lobes.19 31 This pattern is consistent with the retrogenesis paradigm (last-in-first-out hypothesis) which suggests that, compared with anterior regions, the posterior brain regions are spared from neurodegeneration. Interestingly, the greatest slopes of CBF versus myelination were in the anterior lobes, indicating a more rapid MWF decrease with CBF decline compared with the more posterior regions investigated. However, larger cohorts and longitudinal follow-up are needed to draw definitive conclusions regarding regional differences in the association of CBF with myelination. Finally, given the emerging concept of a possibly causative association of demyelination with cognitive impairment,7 our work, although carried out in a normatively ageing cohort, further indicates the potential significance of decreased CBF in the development of cognitive impairment, including dementia.

Our calculated MWF mean values were overall in good agreement with those previously reported.19 31 However, we note that mcDESPOT, including BMC-mcDESPOT, provides somewhat higher MWF values as compared with multiple spin echo (MSE)-based methods.38 Indeed, various experimental and physiological factors, which are not considered in the MSE or mcDESPOT signal models, could be involved in this discrepancy. This includes, but is not limited to, the effects of magnetisation transfer between macromolecules and free water protons, longitudinal relaxation time resulting from use of short repetition times in MSE, exchange between water pools, J-coupling, off-resonance, spin locking, water diffusion within different compartments, and internal gradients. These represent major challenges in MRI studies of myelin content, and further technical developments are required to improve accuracy of myelin imaging. Indeed, comprehensive comparison analyses between BMC-mcDESPOT and MSE-decay curve MWF measurements are needed.18 38

Although our investigation examines a relatively large cohort and uses advanced MRI methodology, our work has limitations. Our data set is cross-sectional, with additional longitudinal studies being required to further establish the potential causality of the CBF and MWF associations observed here. Such work, motivated by the present results, is under way. Furthermore, we used identical ASL experimental parameters for all subjects, and acquired CBF images at a single postlabelling delay; this assumes negligible effects of spatial variation in arterial transit time (ATT), the time of the arterial bolus to transit from the labelling plane to the imaging volume. Although this is a reasonable assumption,28 ATT may vary spatially and may differ between subjects due to arterial blood velocity differences39; this could introduce a small bias in derived CBF values. A multiple postlabelling delay ASL technique may provide more accurate CBF determination.39 Moreover, certain physiological and experimental parameters could bias MWF determination as discussed above. Finally, a number of physiological parameters such as body mass, blood pressure, blood lipids, insulin resistance and inflammatory markers could affect myelin content as well, although a meaningful study of these variables would require a much larger cohort size than in the present study.