Quantitative age-dependent differences in human brainstem myelination assessed using high-resolution magnetic resonance mapping

Abstract

Previous in-vivo magnetic resonance imaging (MRI)-based studies of age-related differences in the human brainstem have focused on volumetric morphometry. These investigations have provided pivotal insights into regional brainstem atrophy but have not addressed microstructural age-differences. However, growing evidence indicates the sensitivity of quantitative MRI to microstructural tissue changes in the brain. These studies have largely focused on the cerebrum, with very few MR investigations addressing age-dependent differences in the brainstem, in spite of its central role in the regulation of vital functions. Several studies indicate early brainstem alterations in a myriad of neurodegenerative diseases and dementias. The paucity of MR-focused investigations is likely due in part to the challenges imposed by the small structural scale of the brainstem itself as well as of substructures within, requiring accurate high-spatial resolution imaging studies. In this work, we applied our recently developed approach to high-resolution myelin water fraction (MWF) mapping, a proxy for myelin content, to investigate myelin differences with normal aging within the brainstem. In this cross-sectional investigation, we studied a large cohort (n = 125) of cognitively unimpaired participants spanning a wide age range (21-94 years) and found a decrease in myelination with age in most brainstem regions studied, while various regions exhibited a quadratic association between myelin and age. We believe that this study is the first investigation of MWF differences with normative aging in the adult brainstem. Further, our results provide reference MWF values.

1. INTRODUCTION

The human brain undergoes structural and macromolecular changes during normal aging which are often associated with neurocognitive and functional decline (1). As an important example, histopathological and magnetic resonance imaging (MRI) studies reveal marked degradation of the brain’s white matter, including myelin loss, with aging (2-14). However, while this myelination trajectory has been well-documented in the cerebrum using various quantitative MRI techniques (7-10, 14-21), little work has been conducted to investigate age-differences in the myelin content of brainstem structures. In addition to functioning as a relay and integrative brain center, the brainstem plays an important role in motor function and pain sensation, alertness, and regulation of cardiac, respiratory, and vasomotor functions. Multiple studies suggest significant brainstem involvement in the early development of Alzheimer’s and Parkinson’s diseases with accompanying neuropsychiatric symptoms including disturbances in mood, emotion, appetite, and sleep, as well as confusion, agitation, and depression (22-26). Therefore, characterizing age-differences in brainstem microstructure could provide important insights into the functional, emotional, and motor alterations that accompany normative aging and neurodegeneration.

The small scale of brainstem structures renders MR-based studies particularly challenging. In fact, the average brainstem volume is about 34 cm³; this in contrast to the cerebrum, with an average volume > 800 cm³ (27). The average diameter of cross-sectional area of the major brainstem subdivisions is only ~18 mm for the midbrain, ~30 mm for the pons, and ~14 mm for the medulla. Therefore, it is evident that imaging at high-spatial resolution is critical to minimize partial volume effects between gray and white matter as well as between different substructures. In recent groundbreaking work (28), Lambert and colleagues reported a linear relationship between various conventional quantitative MRI measures, including apparent transverse and longitudinal relaxation rates (R₂*, R₁), and magnetization transfer saturation (MTS, (29)), and age in the brainstem (28). It was further found that certain gray matter brainstem substructures exhibited decreased MTS and relaxation rates or increased proton density in old subjects. These results were interpreted as indicators of axonal loss, demyelination, or increased iron accumulation in these specific gray matter regions. These conventional quantitative MRI measures are sensitive to myelin content, with the work of Lambert and colleagues therefore providing critical insights into microstructural differences with aging and pathology. Importantly, these metrics also show correlation with histologic measures of myelin (30, 31). Nevertheless, interpretation can be problematic due to their sensitivity to other tissue properties as well (32-36). The development of advanced MR methods based on multicomponent relaxometry to map myelin water fraction (MWF) (37-39), a surrogate of myelin content, have resulted in MR studies with much greater specificity to myelin, although, similar to DTI, relaxation times, and MT, MWF determination is ultimately based on underlying modeling assumptions.

Unfortunately, conventional approaches to mapping MWF require lengthy and clinically unacceptable acquisition times for whole-brain coverage and are also very sensitive to noise and instable due to the small signal from the myelin water component, making them impractical for high-resolution imaging (40-45). To address these limitations, 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 (46-48). This method provides whole brain high-resolution MWF maps with higher accuracy and precision than conventional methods (47, 48) and within a clinically acceptable imaging time, and has been used to provide quantitative evidence of myelin loss in mild cognitive impairment and dementia (49). In the present work, building on Lambert’s et al. study (28), our main goal is to investigate age-differences in myelin in the brainstem using this stabilized approach to myelin quantification. Thus, we applied BMC-mcDESPOT to define differences in myelin content in the brainstem with normative aging on a large cohort of cognitively unimpaired participants (n = 125) spanning an extended and well-sampled age range between 21 and 94 years. Our main goals are to characterize the regional association of myelination with age in specific brainstem regions, to provide reference values for MWF in the adult brainstem, and to develop further insights into regional brainstem maturation and aging in a life-span sample of healthy adults. We believe this to be the first in-vivo study to investigate age-differences in MWF of human brainstem.

2. MATERIAL & METHODS

2.1. Subjects

Cognitively unimpaired subjects (57 women, 52.6 ± 20.3 years; 68 men, 56.5 ± 22.9 years) were recruited from the Baltimore Longitudinal Study of Aging (BLSA), an ongoing study of normative aging in adults (50, 51), and the Genetic and Epigenetic Signatures of Translational Aging Laboratory Testing (GESTALT) study, an ongoing epidemiological, observational and longitudinal study of adults. The BLSA is a longitudinal cohort study funded and conducted by the National Institute on Aging (NIA) Intramural Research Program (IRP). Established in 1958, the BLSA enrolls community-dwelling adults with no major chronic conditions or functional impairments. The GESTALT study is also a study of healthy volunteers, initiated in 2015, and funded and conducted by the NIA IRP. The goal of this study is to evaluate multiple biomarkers related to aging. Note that these studies do not differ in their population characteristics, so that combining subjects from them poses no difficulty. We note that the inclusion and exclusion criteria for these two studies are essentially identical. The cohort for our current study consisted of 125 subjects spanning the age range between 21 and 94 years (54.7 ± 21.7 years) after removal of six imaging datasets with technically limited scans, caused by, for example, excessive motion. Figure 1 provides a detailed distribution of the number of participants per sex and age-decade. Participants underwent a Mini Mental State Examination (MMSE) with a maximum score of 30 (ours sample of men = 28.5 ± 1.5; women = 29 ± 1.4). Age (p > 0.1) and MMSE (p > 05) were not statistically different between men and women. Experimental procedures were performed in compliance with our local Institutional Review Board, and all subjects provided written informed consent.

Figure 1. Number of participants per age-decade and sex within the study cohort.

2.2. Magnetic resonance imaging

All experiments were performed on a 3T whole body Philips MRI system (Achieva, Best, The Netherlands) using the internal quadrature body coil for transmission and an eight-channel phased-array head coil for reception. For each participant, a whole brain MWF map was derived using the BMC-mcDESPOT method (46-49). This protocol consists of ten 3D spoiled gradient recalled echo (SPGR) images acquired with flip angles of [2 4 6 8 10 12 14 16 18 20]°, echo time of 1.37 ms and repetition time of 5 ms, and ten 3D balanced steady state free precession (bSSFP) images acquired with flip angles of [2 7 11 16 24 32 40 60]°, echo time of 2.8 ms and repetition time of 5.8 ms; the TE value is slightly lower than TR/2 to ensure a complete spin refocusing (52). The bSSFP images were acquired with radiofrequency excitation pulse phase increments of 0° or 180° to account for off-resonance effects (47, 48, 53). All SPGR and bSSFP images were acquired with a voxel size of 1.6 mm × 1.6 mm × 1.6 mm. Further, we used the double-angle method (DAM) to correct for radiofrequency excitation inhomogeneity (54). The DAM protocol consists of two fast spin-echo images acquired with flip angles of 45° and 90°, echo time of 102 ms, repetition time of 3000 ms, and acquisition voxel size of 2.6 mm × 2.6 mm × 4 mm. All images were obtained with a field of view of 240 mm × 208 mm × 150 mm and reconstructed on the scanner to a voxel size of 1 mm × 1 mm × 1 mm. We emphasize that all MRI studies and ancillary measurements were performed with the same MRI system, running the same pulse sequences, at the same facility, and directed by the same investigators for both BLSA and GESTALT participants.

2.3. Image processing

2.3.1. Myelin water fraction mapping and image registration

For each subject, the scalp, ventricles, and other nonparenchymal regions within the acquired images were eliminated with the FMRIB Library Software (FSL) using the SPGR images averaged over all flip angles as the input image (55). A whole-brain MWF map was then generated for the extracted parenchymal regions using the BMC-mcDESPOT analysis (46-48). Briefly, BMC-mcDESPOT assumes a two-component non-exchanging system consisting of short and long T₁ and T₂ components. The short component corresponds to the signal of water trapped within the myelin sheets while the long component corresponds to intra/extra-cellular water. Analysis was performed explicitly accounting for nonzero TE as incorporated into our TE-corrected-mcDESPOT signal model (46). BMC-mcDESPOT permits determination of MWF in each voxel through marginalization over nuisance parameters, which in this case are the relaxation times. High-dimensional numerical integration is needed to perform this marginalization; this computational difficulty was addressed using the Monte Carlo sampling. MWF maps were visually inspected for motion artifacts, and six data sets with evident artifacts were omitted from further analysis. Finally, the averaged SPGR image was nonlinearly registered to the MNI space (MNI152_T1_1mm) using the FSL registration tools FLIRT and FNIRT (55-57). The computed transformation matrix was then applied to the MWF map using the applywarp tool function from the registration processing toolbox in FSL (55).

2.3.2. Region of interest segmentation

Fourteen brainstem structures were chosen as regions of interest (ROIs) from the Johns Hopkins University (JHU) ICGM-DTI 81 atlas (58, 59) and the Talairach structural atlas provided in FSL in order to cover all the ROIs for this investigation (Figure 2). Six white matter ROIs were derived from the JHU atlas; these were within the superior cerebellar peduncle, middle cerebellar peduncle, inferior cerebellar peduncle, corticospinal tract, lemniscus tract, and pontine tract. Four additional white matter ROIs were derived from the Talairach atlas corresponding to the midbrain, pons, medulla, and whole brainstem white matter, while three gray matter ROIs were derived from the same atlas corresponding to the substantia nigra, red nucleus, and subthalamic nucleus. All ROIs were eroded to reduce partial volume effects and imperfect image registration using a kernel box of 2 voxels × 2 voxels × 2 voxels with the FSL tool fslmaths. The ROIs were further superimposed on the Harvard-Oxford brainstem atlas in FSL and manually corrected as indicated to avoid anatomic overlap with nearby brain regions. For each ROI, the mean MWF value was calculated for each subject from the normalized space, as well as the mean and standard deviation (SD) MWF values averaged over participants for each age-interval described by Fig. 1. We note that the MWF map calculation, image registration, and ROI segmentation were performed blinded to any information pertaining to the participants’ age, sex and cognitive status.

Figure 2: Visualization of the white matter and deep gray matter ROIs in the brainstem.

1) Subthalamic nucleus, 2) Cerebral peduncle, 3) Substantia nigra, 4) Red nucleus, 5) Middle cerebellar peduncle, 6) Corticospinal tract, 7) Pontine tract, 8) Lemniscus tract, 9) Superior cerebellar peduncle, 10) Inferior cerebellar peduncle, 11) Midbrain, 12) Pons, and 13) Medulla. The brain images were obtained from the standard MNI atlas provided in FSL.

2.4. Statistical analysis

For each ROI, a linear regression model was evaluated with MWF as the dependent variable and with sex and age as the independent variables. The initial model incorporated an interaction between sex and age which was removed if not significant, with the resulting parsimonious model then evaluated without this nonsignificant interaction term. Further, it has been shown that the myelin association with aging generally follows an inverted U-shape in the cerebrum and cerebellum (7, 21). Therefore, for each ROI we also evaluated a model that incorporates a quadratic age term, that is, age². Here again, MWF was the dependent variable, with sex, age, and age² as the independent variables, and interaction terms were assessed as above. Of the several available methods, we simply used the F-test as implemented in MATLAB; incorporation (or not) of the interaction terms or the quadratic age term was based on the significance of the F statistic. In all cases, the threshold for statistical significance was taken as p < 0.05 after correction for multiple ROI comparisons using the false discovery rate (FDR) method (60, 61).

3. RESULTS

Figure 3 illustrates the differences in MWF across the adult lifespan represented as maps averaged over all experimental subjects within the indicated 10-year intervals, for four representative slices covering the main anatomical subdivisions of the brainstem. These MWF maps exhibit substantial tissue contrast between different brainstem substructures. Visual inspection shows that different regions exhibit different trends of MWF as a function of age with the superior brainstem regions, especially the midbrain, exhibiting greater MWF values in comparison to the inferior brainstem regions such as the medulla. All of these results indicate the sensitivity of BMC-mcDESPOT for detecting microstructural differences in myelin content in various brainstem regions as well as at different ages.

Figure 3: Myelin water fraction (MWF) represented as averaged participant maps calculated over 10-year intervals.

Results are shown for four representative slices covering the main brainstem structures. Slices 1 and 2 are representative axial views of the midbrain while slices 3 and 4 are representative axial views of the pons and medulla, respectively. The red bars on the anatomical images indicate the location of the slices. Visual inspection of MWF maps shows an increase in MWF values from early adulthood through middle age, that is, from 20-29 years through 40-49 years, followed by a more rapid decrease, in several brainstem substructures.

Figure 4 shows the MWF derived values from all participants as a function of their age for 14 white matter and gray matter regions. We note that all ROIs included a large number of voxels (Table 1). The plots indicate a decrease in MWF from young adulthood through old age. The main effect of age was significant in all brain regions evaluated except for the middle cerebellar peduncle (p > 0.1 before FDR correction) (Table 1). In addition, the most rapid declines in MWF with age were found in the midbrain, red nucleus, and subthalamic nucleus regions, which exhibited the greatest negative slopes, while the slowest declines in MWF with age were found in the middle cerebellar peduncle region. The slope of MWF versus age for each of the rapidly declining regions was statistically significantly different from that of the slowest declining region (p < 0.01; Z-test computed as the difference between the two slopes divided by the standard error of the difference between the slopes (62)). Furthermore, the highest MWF values were found in the cerebral peduncle, midbrain, and the red nucleus, while the lowest MWF values were found in the subthalamic nucleus, the lemniscus tract, the substantia nigra, and the medulla (Table 2), with the regions of highest MWF values significantly different from the regions of lowest MWF (p < 0.01).

Figure 4: Plots of myelin water fraction (MWF) ROI values as a function of age (N = 125) for white and deep grey matter substructures within the brainstem.

For each region, the coefficient of determination, R², and the significance of the linear regression model, p, are reported. All regions investigated showed linearly decreasing trends of MWF with age while exhibiting variation in slope.

Table 1.

Significance of sex and age terms in the linear regressions. All p-values presented are obtained after FDR correction.

	Sex		Age
	p	F	p	F
Superior cerebellar peduncle	0.48	0.9	0.006	9.4
Middle cerebellar peduncle	0.21	6.0	0.12	2.4
Inferior cerebellar peduncle	0.48	0.6	0.002	12.6
Cerebral peduncle	0.35	3.2	0.043	4.3
Corticospinal tract	0.28	4.3	0.009	7.5
Pontine tract	0.38	1.5	0.016	6.4
Lemniscus tract	0.48	0.8	0.008	8.3
Whole white matter	0.47	1.0	0.007	8.8
Midbrain white matter	0.37	1.7	0.0001	19.9
Pons white matter	0.36	2.0	0.006	9.4
Medulla white matter	0.73	0.1	0.02	5.6
Red nucleus	0.36	2.2	0.0001	19.9
Subthalamic nucleus	0.35	2.5	0.0001	20.0
Substantia nigra	0.46	0.6	0.009	7.7

Table 2.

The number of voxels in each ROI, and mean and standard deviation (SD) MWF values averaged over all participants for each ROI and for each age-interval as well as across the whole age-range.

	Mean ± SD MWF values									Number of voxels
	20-29 yrs.	30-39 yrs.	40-49 yrs.	50-59 yrs.	60-69 yrs.	70-79 yrs.	80-89 yrs.	90-99 yrs.	20-99 yrs.
Superior cerebellar peduncle	0.131 ± 0.036	0.142 ± 0.035	0.136 ± 0.035	0.134 ± 0.039	0.121 ± 0.022	0.118 ± 0.021	0.115 ± 0.029	0.121 ± 0.016	0.128 ± 0.032	250
Middle cerebellar peduncle	0.147 ± 0.032	0.142 ± 0.025	0.147 ± 0.027	0.141 ± 0.025	0.133 ± 0.016	0.136 ± 0.019	0.136 ± 0.025	0.149 ± 0.021	0.142 ± 0.025	1934
Inferior cerebellar peduncle	0.147 ± 0.036	0.151 ± 0.023	0.153 ± 0.031	0.139 ± 0.032	0.122 ± 0.031	0.126 ± 0.022	0.128 ± 0.033	0.139 ± 0.041	0.140 ± 0.031	253
Cerebral peduncle	0.192 ± 0.033	0.206 ± 0.020	0.210 ± 0.030	0.200 ± 0.042	0.183 ± 0.025	0.187 ± 0.027	0.186 ± 0.031	0.194 ± 0.005	0.197 ± 0.030	2549
Corticospinal Tract	0.144 ± 0.038	0.145 ± 0.030	0.145 ± 0.033	0.141 ± 0.028	0.132 ± 0.024	0.128 ± 0.020	0.126 ± 0.031	0.130 ± 0.023	0.138 ± 0.031	1142
Pontine tract	0.140 ± 0.037	0.145 ± 0.033	0.152 ± 0.035	0.139 ± 0.029	0.128 ± 0.024	0.127 ± 0.022	0.128 ± 0.033	0.132 ± 0.035	0.139 ± 0.033	481
Lemniscus tract	0.109 ± 0.030	0.111 ± 0.020	0.112 ± 0.028	0.104 ± 0.021	0.094 ± 0.017	0.097 ± 0.021	0.097 ± 0.026	0.096 ± 0.008	0.105 ± 0.025	418
Whole white matter	0.129 ± 0.023	0.133 ± 0.018	0.138 ± 0.026	0.125 ± 0.018	0.118 ± 0.018	0.119 ± 0.016	0.121 ± 0.021	0.120 ± 0.014	0.128 ± 0.022	28977
Midbrain white matter	0.142 ± 0.024	0.155 ± 0.019	0.160 ± 0.025	0.140 ± 0.026	0.140 ± 0.020	0.131 ± 0.022	0.128 ± 0.025	0.125 ± 0.009	0.144 ± 0.026	8992
Pons white matter	0.137 ± 0.028	0.141 ± 0.022	0.146 ± 0.026	0.132 ± 0.022	0.123 ± 0.022	0.125 ± 0.019	0.127 ± 0.024	0.128 ± 0.018	0.135 ± 0.025	11577
Medulla white matter	0.109 ± 0.021	0.111 ± 0.023	0.116 ± 0.031	0.100 ± 0.016	0.099 ± 0.017	0.100 ± 0.016	0.099 ± 0.024	0.106 ± 0.021	0.107 ± 0.024	1838
Red nucleus	0.174 ± 0.029	0.183 ± 0.021	0.187 ± 0.030	0.174 ± 0.048	0.170 ± 0.017	0.148 ± 0.027	0.150 ± 0.024	0.172 ± 0.031	0.171 ± 0.031	646
Subthalamic nucleus	0.117 ± 0.027	0.135 ± 0.024	0.131 ± 0.029	0.106 ± 0.032	0.115 ± 0.024	0.097 ± 0.026	0.107 ± 0.028	0.077 ± 0.016	0.117 ± 0.030	280
Substantia nigra	0.125 ± 0.034	0.133 ± 0.027	0.132 ± 0.033	0.128 ± 0.033	0.128 ± 0.017	0.116 ± 0.022	0.113 ± 0.026	0.109 ± 0.026	0.125 ± 0.029	537

Linear regression analysis showed that the effect of sex was significant (p < 0.05) or close to significant (p < 0.1) in only three brainstem regions, namely, the middle cerebellar peduncle (F = 6, p = 0.006), the cerebral peduncle (F = 3,2, p = 0.075), and the corticospinal tract (F = 4.3, p = 0.04); however, significance did not survive the FDR correction (p > 0.1 after FDR correction; Table 1). In these brain regions, women showed a trend of 7 to 9 % higher myelin content than men. The interaction term between age and sex was also significant in only five brain regions before FDR correction, and in no regions after correction: the superior cerebellar peduncle (F = 7.1, p = 0.009; p > 0.05 after FDR correction), the middle cerebellar peduncle (F = 7.7, p = 0.006; p > 0.05 after FDR correction), the corticospinal tract (F = 5.7, p = 0.02; p > 0.1 after FDR correction), the pons (F = 4.7, p = 0.03; p > 0.1 after FDR correction), and the substantia nigra (F = 5.1, p = 0.023; p > 0.1 after FDR correction). In these regions, women showed a trend to a more rapid decline of MWF with age as compared to men (p > 0.05 for comparison of slopes of linear regression of MWF on age).

Finally, in accordance with literature results in the cerebrum, we evaluated a quadratic age term, age², in the linear regression model (7). The statistical analysis indicated that the effect of this term was significant (p < 0.05) or close to significant (p < 0.1) in only six brainstem regions, namely, the superior cerebral peduncle (F = 2.9, p = 0.09), the whole white matter (F = 3.3, p = 0.07), the midbrain (F = 13.5, p = 0.0003), the red nucleus (F = 3.7, p = 0.057), the subthalamic nucleus (F = 7.9, p = 0.006), and the substantia nigra (F = 5.5, p = 0.021). We note that, except the midbrain and the subthalamic nucleus, statistical significance did not survive the FDR correction. The regression plots for these regions incorporating a quadratic age term are shown in Figure 5.

Figure 5: Plots of myelin water fraction (MWF) ROI values as a function of age (N = 125) for six of the substructures also highlighted in Fig. 4, but for which quadratic trends of MWF were significant.

For each region, the coefficient of determination, R², and the significance of the overall regression model, p, are reported. These selected regions clearly exhibited a quadratic trend of age on MWF, with significant or close to significant p values. This suggests progressive myelination continuing into middle age followed by more rapid myelin loss. We note that similar trends have recently been reported in the cerebrum (7, 21).

4. DISCUSSION

Multiple studies have shown that myelin degeneration is a cardinal feature of the biological process of aging as well as of various neurodegenerative diseases (2, 49, 63-71). Myelin provides axonal insulation, essential for proper saltatory conduction, as well as contributing to axonal viability. Therefore, myelin damage or loss inhibits normal nerve impulse transmission, resulting in a variety of neurologic manifestations (2, 49, 64, 66, 67, 70, 72-78). Although several studies have investigated myelination patterns within the cerebrum and established potential mechanisms for demyelination, very little comparable work has been conducted on the brainstem, in spite of its potentially central role in aging and dementia. Indeed, this work was motivated by the emerging evidence of brainstem involvement in the early development of Alzheimer’s disease and other neurodegenerative diseases (22-26). In this cross-sectional study, conducted on a large cohort of cognitively unimpaired subjects, we showed that the brainstem exhibits myelination differences throughout the human lifespan using a stabilized measure of MWF. Our results indicate that brainstem white matter and gray matter substructures exhibit different myelination trends with normal aging.

One of our particular interests was to investigate whether the myelination pattern in the brainstem followed a progression similar to that observed in the cerebrum, that is, a progressive increase in myelination from young adulthood through middle age, followed by decrease through older age (7, 21). Therefore, in all cases, we evaluated the statistical significance of quadratic terms in the age-dependence of myelination. We found that this quadratic term was significant or close to statistical significance in six of the fourteen brainstem regions studied, with only a marginal trend seen in other regions. It is unclear whether this lack of a clear quadratic trend reflects the power of the present study given the effect size, or rather reflects the true biologic time course of myelination in the brainstem. We note that in fact, the brainstem is one of the earliest structures to demonstrate significant myelination, with significant myelin present even at the pre-term stage of development (79). In any event, in this dataset, any quadratic trend is substantially less pronounced than is observed in the cerebrum (7, 21). This likely reflects the structural and functional differences between the brainstem and the cerebrum. Indeed, the brainstem is primarily focused on integrating information for survival, proprioception, motion, and sensation, while the cerebrum integrates more complicated reactions to stimuli and processes information necessary for higher level functions such as perception, thought, learning, and memory. Therefore, it is plausible that the myelination patterns between these functionally and developmentally different regions of the brain would be different. However, it must also be recognized that the relative importance of the linear versus quadratic age trends in our model, and the specific values of the parameters and their significance, will exhibit some variability as a function of sampling density within age groups, range of ages incorporated, and consistency of data (80). In addition, the choice of a linear regression model is consistent with our visual inspection but may not describe the underlying biology. Thus, this choice must be regarded as an expedient to model the data rather than a description of underlying physiologic processes. Other models, such as piecewise linear, may serve equally well as data descriptors. Finally, our dataset is cross-sectional, so that the linear or quadratic age-associations of MWF observed here require further validation through longitudinal studies. Such work, motivated by the present results, is underway.

We observed a more rapid decrease of myelin in the deep gray matter regions of the brainstem including the red nucleus and the subthalamic nucleus as compared to most other regions evaluated; this is in agreement with Lambert et al. (28). Studies have shown that these regions are particularly susceptible to increased iron deposition during the processes of normative aging and neurodegeneration (81-84). This iron may serve to catalyze free radical reactions promoting lipid peroxidation and oxidative tissue damage, promoting myelin breakdown and consequent additional release of iron (85). Moreover, the midbrain also exhibited a rapid decline of MWF with age. Consistent with this, morphometry-based studies have shown that the midbrain specifically exhibits significant atrophy with aging, in contrast to other brainstem regions (28, 86). Accelerated demyelination of the midbrain observed here, and subsequent axonal loss could explain these consistent observations. Further, we observed progressively decreasing myelin content from the superior to inferior brainstem. We speculate that this is because the midbrain, the most superior of the main brainstem structures, contains large bundles of myelinated axons, such as the cerebral peduncle, while the medulla, in the inferior position, contains cell bodies of most of the cranial nerves in addition to gray matter nuclei with unmyelinated axons. Finally, our results exhibit relatively large MWF values in the deep gray matter regions, especially in the red-nucleus and the substantia nigra. This could be due to a bias in MWF values resulting from increased iron. Indeed, it has been shown that diffusion of iron into the extracellular space may shorten the corresponding transverse relaxation time leading to an artificial overestimation of MWF values (87). Further ex-vivo or histological experiments are required to unambiguously elucidate the origin of this observation.

The effect of sex on myelination trajectory was significant in three brainstem regions, with women exhibiting overall a rapid decrease in MWF with age as compared to men, but these observations were not statistically significant after FDR correction. It is likely that our model is underpowered to detect such sex differences; therefore, an increased cohort size could provide further insights into differences in myelination between males and females. Indeed, sexual dimorphism in myelination would be consistent with the known differences in proliferation of oligodendrocytes and myelin proteins (88-90) which are modulated by sex steroids (91). Moreover, several studies, including MRI investigations, have revealed gender differences in brain maturational processes and emphasize the importance of myelination in understanding the mechanism of neuropsychiatric disorders (92-95). However, the literature remains sparse and further advanced investigations are required (96, 97). Nonetheless, because the brainstem is critical for basic brain functions and is a locus for many phylogenetically conserved functions and anatomy throughout mammals, sex dimorphism is not expected to be as pronounced in comparison to other higher-level cortical areas (98).

To our knowledge, only one quantitative in-vivo MRI study has been conducted to date that investigates regional microstructural age-differences in the adult human brainstem (28). Indeed, based on relaxation rates, proton density, and MTS, Lambert and colleagues have shown that different gray matter brainstem substructures exhibited decreased MTS or relaxation rates and increased proton density; as indicators of axonal loss, demyelination, and increased iron accumulation (28). Building on Lambert’s et al. original study, our goal was to investigate age-differences in myelin content of the brainstem using our MWF method and to provide complementary insights regarding brainstem myelination with normative aging. While sensitive to tissue changes, conventional quantitative MRI methods are not specific to any particular microstructural process. For example, it is most often assumed that relaxation within each imaging voxel may be described by a single relaxation component, although this assumption does not capture the structural and compositional complexity of brain tissue. In fact, a number of previous studies have demonstrated the presence of multicomponent relaxation processes in brain as a signature of compartmentation (37, 41, 47, 77, 99, 100). However, conventional single-component MRI approaches provide higher precision in derived parameter estimates as compared to multicomponent parameter estimates, while requiring greatly reduced acquisition time. This permits, as in Lambert’s et al. study (28), investigation of multiple quantitative MRI parameters, in contrast to a single metric, providing complementary insights of the microstructural age-differences. The tradeoff for this is use of a model that is arguably less descriptive of tissue properties than is a multicompartment model. Nevertheless, our finding, overall, agrees with Lambert’s et al. observation of decreasing myelin content with age in various brainstem substructures (28). However, as indicated by Lambert et al., axonal loss may also occur with normal aging. It is unclear whether such axonal damage results from long-term demyelination and consequent lack of trophic support or is concurrent with it. Further investigations using multimodal approaches would provide further insights (28, 35). In fact, various metrics for myelin all have their important roles. MWF (via mcDESPOT or multiexponential T₂ or T₂* decay) provides a value that is in principle proportional to myelin content. MWF measures however do not provide any information regarding organizational structure, and may even incorporate fractionated and ineffective myelin fragments. DTI, on the other hand, is an excellent metric for architecture, providing insight into another crucial aspect of myelin health, even though the values are not proportional to myelin content. Magnetization transfer, similarly, provides insight into aspects of proximity and chemical communication between pools, as well as their mobility. Similarly, the relaxation times T₁ and T₂ provide information complementary, but not redundant with, these metrics.

Although multicomponent analysis accounting for compartmentation improves both sensitivity and specificity for myelin content (77, 99, 100), the conventional techniques for this analysis are very sensitive to noise and instable due to the small signal from the myelin water component (40, 41); this renders high-resolution imaging challenging, especially in the clinical setting. This issue is further exacerbated for brainstem MR imaging due to its small structure and complex anatomy; this could explain the lack of MR-based work investigating the myelination patterns of the brainstem with aging or pathology. In the present work, we used BMC-mcDESPOT method we have developed for MWF imaging (46-48, 101); this technique is a substantial extension of the mcDESPOT analysis introduced by Deoni and colleagues (53,102). While the original mcDESPOT work permits fast imaging and has been extensively used to characterize brain maturation and various neurodegenerative diseases (2, 49, 67, 103-105), the nonlinear least-squares analysis employed can lead to substantial inaccuracies (47, 48, 101, 106-108). In contrast, our Bayesian implementation greatly improves the accuracy and precision of the analysis, permitting much more reliable high-spatial resolution imaging (47, 48). This capability was critical in our study to accurately delineate small brainstem substructures while avoiding partial volume bias. We further note that the greatly enhanced stability of MWF mapping with BMC-mcDESPOT would permit even higher spatial resolution studies to be performed as well, opening the possibility of investigating the myelination patterns of very small substructures within the brainstem, at the expense however of increased acquisition time.

One of the main challenges of in-vivo brainstem studies is accurate segmentation of its substructures. This is mainly due to their small size, requiring high-spatial resolution imaging, as well as relatively poor contrast between various regions. In our study, we conducted a careful examination of all ROIs, but some partial volume bias could nevertheless remain in the calculated MWF values due to e.g. imperfect image registration. We note that the MWF maps provide excellent contrast between various regions, which guided us in our segmentation (Fig. 2). In fact, the MWF maps exhibit better delineation of different brainstem substructures as compared to weighted images, and could help in developing new, or refining existing, brainstem atlases. We further note that fully-automated brainstem segmentation pipelines are available providing superior spatial precision (28); however, these analyses rely on multiple parametric mapping protocols.

Our work has limitations. Our cohort, although relatively large and spanning a wide and well-sampled age range, does not include very young participants (< 20 years old); this limitation derives from the exclusion criteria of the BLSA and GESTALT studies; inclusion of younger participants may influence the shape of the MWF-age trends (80). We also note that optimal uniform sampling across all age intervals was not fully achieved in this convenience sample of participants in ongoing research protocols. In fact, the number of subjects included between 50-69 years is relatively lower as compared to the other age decades. This may also influence the overall interpretability of myelination during the process of aging, as discussed above. Nevertheless, the inclusion of a large number of participants between the extremes of age, and with fairly uniform age distribution, enabled us to explore nonlinear models as well as nonmonotonic models. Further, in contrast to Lambert’s et al. study (28), our analysis was based on ROIs which led to a loss in spatial precision. As additional issue is that mcDESPOT in general, including BMC-mcDESPOT, provides relatively higher MWF values as compared to MSE-based methods (108). This discrepancy is likely due several physiological and experimental factors that are not modeled in either mcDESPOT- or MSE-based signal formalisms. This includes exchange between water pools; studies have shown that water exchange has non-negligible effect on MWF determination (109, 110). However, incorporation of water exchange in conventional mcDESPOT signal modeling leads to very instable determination of MWF (101, 106, 107). We are currently investigating the potential of BMC-mcDESPOT analysis to improve MWF determination when water exchange is incorporated. Other parameters that may bias MWF estimation are magnetization transfer between free water protons and macromolecules, iron content, T₁ effects resulting from short TRs (in MSE), off-resonance effects, J- coupling, spin locking, internal gradients, differential and signal attenuation due to water diffusion in underlying compartments. However, this limitation applies to all existing MWF mapping methods (41, 111-113). Therefore, our MWF values would best serve as reference values only for studies based on similar MWF approaches. Finally, our dataset is cross-sectional, so that the trends observed here require further validation through longitudinal studies. Such work, motivated by the present results, is underway.

5. CONCLUSIONS

We have demonstrated the feasibility of high-resolution myelin imaging in the brainstem. We found that myelin content decreases with normal aging throughout brainstem regions, with substantial regional variation, as expected. Our work also provides reference MWF values for the main substructures of the brainstem, providing a baseline for investigations of neurodegenerative diseases, such as Alzheimer’s and Parkinson’s diseases.