Multimodal Magnetic Resonance Imaging Assessment of White Matter Aging Trajectories Over the Lifespan of Healthy Individuals

ARCHIVAL REPORT

Multimodal Magnetic Resonance Imaging Assessment of White Matter Aging Trajectories Over the Lifespan of Healthy Individuals George Bartzokis, Po H. Lu, Panthea Heydari, Alexander Couvrette, Grace J. Lee, Greta Kalashyan, Frank Freeman, John W. Grinstead, Pablo Villablanca, J. Paul Finn, Jim Mintz, Jeffry R. Alger, and Lori L. Altshuler Background: Postmortem and volumetric imaging data suggest that brain myelination is a dynamic lifelong process that, in vulnerable late-myelinating regions, peaks in middle age. We examined whether known regional differences in axon size and age at myelination influence the timing and rates of development and degeneration/repair trajectories of white matter (WM) microstructure biomarkers. Methods: Healthy subjects (n ⫽ 171) 14 –93 years of age were examined with transverse relaxation rate (R2) and four diffusion tensor imaging measures (fractional anisotropy [FA] and radial, axial, and mean diffusivity [RD, AxD, MD, respectively]) of frontal lobe, genu, and splenium of the corpus callosum WM (FWM, GWM, and SWM, respectively). Results: Only R2 reflected known levels of myelin content with high values in late-myelinating FWM and GWM regions and low ones in early-myelinating SWM. In FWM and GWM, all metrics except FA had significant quadratic components that peaked at different ages (R2 ⬍ RD ⬍ MD ⬍ AxD), with FWM peaking later than GWM. Factor analysis revealed that, although they defined different factors, R2 and RD were the metrics most closely associated with each other and differed from AxD, which entered into a third factor. Conclusions: The R2 and RD trajectories were most dynamic in late-myelinating regions and reflect age-related differences in myelination, whereas AxD reflects axonal size and extra-axonal space. The FA and MD had limited specificity. The data suggest that the healthy adult brain undergoes continual change driven by development and repair processes devoted to creating and maintaining synchronous function among neural networks on which optimal cognition and behavior depend.

Key Words: Aging, Alzheimer, axial diffusivity, cognition, degeneration, diffusion tensor imaging (DTI), development, fractional anisotropy, magnetic resonance imaging (MRI), myelin, oligodendrocytes, radial diffusivity, relaxation rate (R2), white matter (WM) he exceptional myelination of the human brain has imposed especially high metabolic demands that might have belated some maturational processes. Starting from minimal or no myelin at birth, late-myelinating regions such as the frontal lobes might not reach peak myelination until well into middle age. Thus, during the adult lifespan, myelination of the thinner axons of latemyelinating regions in combination with maintenance/repair and degenerative processes combine to create white matter (WM) volume trajectories that are well-approximated by quadratic (inverted U) functions (Figure 1). The interplay of developmental, degenerative, and repair processes have been hypothesized to contribute to developmental diseases such as schizophrenia and bipolar disorder as well as trigger the proteinopathies associated with the pathology of highly prevalent dementing disorders of old age such as

T

From the Department of Psychiatry (GB, PH, AC, GK, FF, LLA); Laboratory of Neuroimaging (GB); Department of Neurology (PHL, GJL, JRA); Department of Radiology (PV, JPF), The David Geffen School of Medicine at UCLA, Los Angeles, California; Greater Los Angeles Veterans Administration Healthcare System (GB), West Los Angeles, California; Siemens Healthcare (JWG), Portland, Oregon; and the University of Texas Health Science Center at San Antonio (JM), Department of Epidemiology and Biostatistics, San Antonio, Texas. Address correspondence to George Bartzokis, M.D., The David Geffen School of Medicine at UCLA, Department of Psychiatry, 300 UCLA Medical Plaza, Los Angeles, CA 90095– 6968; E-mail: [email protected]. Received Mar 3, 2012; revised Jun 8, 2012; accepted Jul 1, 2012.

0006-3223/$36.00 http://dx.doi.org/10.1016/j.biopsych.2012.07.010

Alzheimer’s disease (AD) (1,2). It is therefore of great diagnostic and therapeutic importance to define in vivo biomarkers that can track microstructural changes underlying brain myelination and axonal maturation as well as degenerative declines in myelin and axonal integrity associated with aging that result in quadratic trajectories of WM volumes (Figure 1). We and others have previously used a magnetic resonance imaging (MRI) measure called transverse relaxation rate (R2) to indirectly assess subcortical myelination in vivo. The R2 is calculated from the better-known measure transverse relaxation time (T2) as its reciprocal (R2 ⫽ 1/T2 ⫻ 1000). Calculated R2 measures are sensitive indicators of myelination, with myelination increasing R2, whereas age-related myelin breakdown decreases it (3–10). Recently, the relationship between R2 and myelin breakdown was confirmed in an animal model of genetically induced oligodendrocyte cell death that lacks the inflammatory response that can obscure the underlying myelin-based R2 changes (11) as well as in a genetically hypomyelinated mouse mutant (12) (see Supplement 1 for prior studies on this subject). Our own prior human studies showed that R2 reductions track age-related myelin breakdown/ loss and are more severe in AD (10,13) as well as healthy individuals at increased genetic risk for developing AD (14,15). Furthermore, in healthy older individuals, increased R2 is positively associated with better cognitive processing and motor speeds (13,15,16). The MRI measures derived from diffusion tensor imaging (DTI) have recently become most popular as another way to investigate WM microstructural changes. Diffusion tensor imaging measures the rate of diffusion motion of water molecules on a microscopic spatial scale in a multiplicity of directions. Diffusion tensor imaging produces three standard measures of diffusivity: axial (AxD), along the direction of axons; radial (RD), perpendicular to AxD; and the summary diffusivity measure mean diffusivity (MD), the average of BIOL PSYCHIATRY 2012;72:1026 –1034 © 2012 Society of Biological Psychiatry

G. Bartzokis et al.

BIOL PSYCHIATRY 2012;72:1026 –1034 1027

Figure 1. Postmortem assessment of relationships between axonal fiber size, myelin content, and lifespan myelination trajectories of chosen regions of interest. (A) Myelin stain of a 1-year-old human brain depicting corpus callosum (CC). Compare early-myelinating caudal end (black myelin stain on right side of the CC) where splenium (SWM) is located with the largely unmyelinated genu (GWM—light gray curved part of the CC on the left side of the panel) (reprinted with permission from Figure 1.3 on page 9 of Kemper [35], adapted from Kaes [68]). (B) Left panel: high-magnification myelin stain of adult primate GWM showing preponderance of small myelinated fibers (with some not yet myelinated marked with arrows). Right panel: SWM showing preponderance of very large axon fibers that are almost all thickly myelinated; however, because of the large volume of axonal cytoplasm, the proportion of myelin content is much lower in SWM compared with that of the GWM. Reprinted with permission from Lamantia and Rakic (36). (C) Myelination volumes (y axis) versus age (x axis) in frontal lobes of normal human brain. Top panel: postmortem myelin stain data (35) depicting the quadratic lifespan trajectory of frontal lobe intracortical myelin volume peaking at approximately age 45. Lower panel: in vivo magnetic resonance imaging data depicting the similar quadratic lifespan trajectory of frontal lobe myelinated white matter volume of healthy individuals also peaking at age 45. Reprinted with permission from Bartzokis et al. (37), copyright © 2001 American Medical Association. All rights reserved.

AxD and RD, as well as another summary variable, fractional anisotropy (FA), encapsulating directional diffusion information (see Supplement 1 for detailed description of these DTI metrics). Restriction in RD is believed to be primarily caused by axonal and myelin membranes making RD potentially sensitive to myelination. That interpretation of the RD metric has been questioned, however (17), and it has been suggested that RD might be most appropriately considered as a marker of tissue integrity (reviewed in [18]). Nevertheless, empirical experiments in animal models (19 –23), human postmortem samples of multiple sclerosis brain (24,25), and infant brain development (26) have supported the suggestion that changes in RD might be most closely associated with myelination or loss thereof. In the current study we assessed the adult lifespan trajectory of R2 and DTI metrics in a healthy population. We chose a region of interest (ROI) approach to minimize the many pitfalls associated with whole brain DTI analyses, such as nonlinear susceptibility- and eddy current-induced anatomic distortions (27,28). Furthermore, through our three ROI choices, we interrogated: 1) different “types” of myelin, 2) the impact of crossing fibers, and 3) regional differences in vulnerability to age-related myelin breakdown (10) (for details see Supplement 1). Two late-myelinating regions with different fiber orientations were contrasted with one early-myelinating region. Bilateral midfrontal white matter (FWM) regions were chosen as the latest-myelinating regions that contain primarily very thin myelinated fibers with some admixture of large fibers (29,30) connecting it to the rest of the brain with a preponderance of crossing fibers. We also chose two midline corpus callosum (CC) regions that contain essentially only parallel fibers. These two CC regions were also specifically chosen on the basis of their timing/ age at myelination and axonal fiber characteristics that produce

strikingly different myelin contents: the late-myelinating genu (GWM), which contains almost exclusively small fibers that are thinly myelinated; and the early-myelinating lower splenium (SWM) that contains a preponderance of large primary visual fibers that are thickly myelinated. As can be seen in Figure 1B, these characteristics make the proportion of myelin/voxel volume much higher in the GWM (and FWM) than SWM, because a much larger proportion of the SWM volume is occupied by axonal cytoplasm even though each of those large axons is more thickly myelinated. The anatomical characteristics of these three ROIs were used to examine the suggestion derived from postmortem and volumetric myelination data (Figure 1C) that over the lifespan there is a dynamic interplay at a microstructural level between developmental/ maturational and degenerative/repair processes. To provide crossvalidating data on myelin trajectories we examined two different myelination metrics (R2 and RD). For comparison, we also examined AxD that was expected to track different biological processes (31,32) and the two standard summary measures of overall WM integrity (FA and MD). We assessed the proposition that, although age-related myelin breakdown is a generalized phenomenon, in old age small and thinly myelinated fibers of the late-myelinating regions (FWM and GWM) are most vulnerable to breakdown and loss (10,33,34). In addition, by choosing two CC regions that contain fibers aligned in parallel yet have striking differences in fiber size as well as myelin thickness, content, and vulnerability to breakdown, we tested the hypothesis that—similar to R2—the RD metric will be most sensitive to an accelerated myelin loss in late-myelinating regions (FWM and GWM). Finally, we examined the proposition that the AxD metric will be most sensitive to axonal size and especially extra-axonal space (31,32). www.sobp.org/journal

1028 BIOL PSYCHIATRY 2012;72:1026 –1034

G. Bartzokis et al.

Table 1. Demographic Data

Age, Years Education, Years Race (C/As/AA/His)

All Subjects (n ⫽ 171) Mean (SD)

Male Subjects (n ⫽ 93) Mean (SD)

Female Subjects (n ⫽ 78) Mean (SD)

51.0 (22.0) 16.8 (2.5) 143/7/11/10

51.3 (21.5) 17.2 (2.6) 79/2/7/5

50.8 (22.7) 16.3 (2.2) 64/5/4/5

AA, African-American; As, Asian; C, Caucasian; His, Hispanic.

Methods and Materials Subjects Healthy volunteers (n ⫽ 171) between the ages of 14 and 93 were recruited from the community and hospital staff to participate in a study of healthy aging. This age range was chosen because our previous studies showed that quadratic functions are well-suited for modeling age-related differences in the adult age range (10,38) (see Supplement 1 for inclusion/exclusion details). Demographic data are presented in Table 1. MRI Protocol All subjects were scanned with the same 1.5-Tesla Siemens MRI instrument (Siemens, Malvern, Pennsylvania), and all scans used the same imaging protocol. The subjects were scanned randomly over time, irrespective of demographic (e.g., age, gender) variables. Details of the protocol have been previously published (10,38) and will only be summarized here. The axial image acquisition sequence acquired interleaved contiguous slices with a dual spin-echo sequence repetition time ⫽ 2500 msec, echo times ⫽ 20/90 msec, 3 mm slice thickness, 192 gradient steps, and 24 cm field of view.

DTI was performed with a single-shot multisection spin-echo echo-planar pulse sequence (repetition time/echo time ⫽ 5900/ 104 msec; flip angle ⫽ 90°) obtained in the axial plane with a 128 ⫻ 128 matrix size, 24 cm field of view, 3.0 mm slice thickness, 35 slices with no interslice gap, and a readout bandwidth of 1630 Hz/pixel. For each slice, diffusion gradients were applied along 12 independent orientations with b ⫽ 800 sec/mm2 after the acquisition of one b ⫽ 0 sec/mm2 (b0) set of images. Both R2 and DTI data were extracted from two contiguous slices after manuallycorrectingforanatomicmisregistrationsandremovingfromthe ROIs voxels impinging on cerebrospinal fluid and/or gray matter. Details on image and data analysis are included in Supplement 1.

Results The relationship between R2 and DTI measures of frontal WM integrity and age were examined in this population of 171 healthy individuals. Evaluation of gender differences in age-related trajectories revealed that none of the gender ⫻ age interactions was statistically significant, with p values ranging from .16 to .74 (all df ⫽ 1,165) in FWM, with similar results for GWM and SWM regions with p values ranging from .20 to .98 and .10 to .95, respectively (all df ⫽ 1,165). Thus, there was insufficient evidence to conclude that aging trajectories of these five variables differ by gender, and this variable was no longer considered in subsequent analyses. The R2 and FA data are depicted in Figure 2. It demonstrates that the differences in axonal sizes and myelination characteristics of the three ROIs are reflected in strikingly different absolute values as well as age-related trajectories. To achieve the same directionality as the R2 and FA measures, the three DTI diffusivity measures (RD, AxD, and MD) were inversed

Figure 2. Lifespan trajectories of absolute values of transverse relaxation rates (R2) and diffusion tensor imaging fractional anisotropy (FA) metric in three white matter regions. The three regions are the late-myelinating frontal white matter (FWM), composed largely of small axons with an admixture of larger axons, and two CC regions: late-myelinating GWM, composed primarily of thin axons; and early-myelinating SWM, composed primarily of large axons (see Figure 1B). Trajectories that are not statistically significantly quadratic are depicted by their linear functions. Only R2 values seem to track the absolute myelin content of the three regions GWM ⬎ FWM ⬎ SWM. In addition, in late-myelinating regions (FWM and GWM) the dynamic interplay between developmental age-related increases in myelin and degenerative changes of myelin breakdown and loss in middle and older ages produce strikingly nonlinear R2 trajectories in this age range of the lifespan. These nonlinear trajectories stand in contrast to the linear trajectory of SWM R2, which—although it contains very thick myelin sheaths— has a lower myelin content due to its large primary visual axons (see Figure 1B and Table 2 for statistical results). In contradistinction, FA regression lines replicate prior publications that demonstrated linear declines with age in the late-myelinating regions (FWM and GWM). The much lower absolute FA values in FWM likely reflect the preponderance of crossing fibers in this region, whereas both GWM and SWM (composed of parallel fibers) have considerably higher FA values. Abbreviations as in Figure 1.

www.sobp.org/journal

BIOL PSYCHIATRY 2012;72:1026 –1034 1029

G. Bartzokis et al.

Figure 3. Lifespan trajectories of absolute values of R2 and three diffusion tensor imaging metrics in three different white matter regions. The three regions are the late-myelinating FWM and two CC regions: late-myelinating GWM and early-myelinating SWM. Due to the large subject number and multiple metrics, only regression lines are depicted. The metrics are R2 (same data as depicted in Figure 2 and included here as a comparator nondiffusivity metric) and three diffusivity metrics: radial (RD), axial (AxD), and mean (MD) that were inverted by multiplying the values by ⫺1 to achieve the same directionality as the R2 and FA metrics. Thus, higher values consistently indicate better integrity and reduced diffusivity. Trajectories that are not statistically significantly quadratic are depicted by their linear functions. Note that RD follows the pattern of R2 with nonlinear trajectory in early-myelinating regions (FWM and GWM) and linear in SWM; however, in counter distinction to R2, it reaches the highest values in SWM even though that region has the lowest myelin content (see Figure 1B). Also note that the absolute values of RD and AxD are strongly affected by the crossing fiber structure of the FWM region compared with the parallel axon arrangements of GWM and SWM regions, whose values are more similar to each other. This is not the case for MD, which— unlike the other diffusion tensor imaging metrics— does not incorporate directional information, is much less influenced by the crossing fibers of FWM, and thus has similar values in all three regions of interest. Both MD and AxD have quadratic trajectories in all three regions. Abbreviations as in Figures 1 and 2.

(multiplied by ⫺1). Thus, higher values consistently indicate better integrity and reduced diffusivity (Figure 3). The means and SDs of the R2 and four DTI metrics are tabulated in Table 2. They show that only R2 values track the myelin content of the different regions with high values for the latemyelinating, thin fiber, and high-myelin-content FWM and GWM regions and low values for the early-myelinating, large-fiber, and lower myelin content of primary visual axons of the SWM (also see Figures 1 and 2). Pearson correlation analyses were performed to examine the association and the shared variance between R2 and the four DTI metrics (Table 3). The relationship between the MRI metrics and age were examined next. As can be seen in Table 4 and Figure 4, the later myelinat-

ing regions (FWM and GWM) had similar patterns that differed substantially from the early-myelinating SWM. Once again R2 and RD had similar relationships with age. Both had significant quadratic components that were most significant in the latest-myelinating FWM, followed by the GWM, which stood in contrast to the early-myelinating SWM region where only a marginally significant linear function for R2 was evident. To compare whether the peak of the quadratic age regression curves differed between FWM R2 and each of the other three DTI measures with significant quadratic trajectories, a bootstrap replication analysis was employed. One hundred bootstrap replication samples were created to serve as a sampling reference, and each replication was an individual random draw of 171 cases from the sample. The results showed that peak FWM R2 is reached at an

Table 2. Regional Mean and SDs of R2 and Four DTI Metrics

R2 RD FA MD AxD

FWM vs. SWM

GWM vs. SWM

FWM vs. GWM

FWM Mean (SD)

GWM Mean (SD)

SWM Mean (SD)

t

p

t

p

t

p

16.73 (.75) ⫺.60 (.06) .44 (.05) ⫺.80 (.05) ⫺1.19 (.08)

17.09 (.75) ⫺.34 (.08) .79 (.05) ⫺.82 (.08) ⫺1.79 (.14)

15.52 (.65) ⫺.27 (.05) .83 (.04) ⫺.76 (.05) ⫺1.76 (.14)

18.13 ⫺62.31 ⫺81.23 ⫺8.14 58.58

⬍ .0001 ⬍ .0001 ⬍ .0001 ⬍ .0001 ⬍ .0001

26.80 ⫺12.35 ⫺9.93 ⫺11.17 ⫺2.48

⬍ .0001 ⬍ .0001 ⬍ .0001 ⬍ .0001 .014

⫺8.27 ⫺47.84 ⫺81.56 6.28 65.05

⬍ .0001 ⬍ .0001 ⬍ .0001 ⬍ .0001 ⬍ .0001

The negative values of the three diffusivity measures (radial [RD], axial [AxD], and mean [MD]) were the result of inverting the measures by multiplying the values by ⫺1 to achieve the same directionality as the transverse relaxation rate (R2) and fractional anisotropy (FA) measures. Thus, higher values consistently indicate better integrity and reduced diffusivity. The mean values for R2 and the four diffusion tensor imaging (DTI) metrics were uniformly significantly different across regions, due primarily to the large sample size (all df ⫽ 170). Nevertheless, examination of the t statistic shows that: for R2, absolute differences across regions of interest vary considerably in parallel with known myelin content (see Figure 1B); and for the diffusivity measures that incorporate directional information (FA, RD, AxD), the values are markedly affected by the region of interest content of crossing fibers (frontal lobe white matter [FWM] ⬎⬎ genu white matter [GWM] and splenium white matter of corpus callosum [SWM]).

www.sobp.org/journal

1030 BIOL PSYCHIATRY 2012;72:1026 –1034

G. Bartzokis et al.

Table 3. Correlations and Shared Variance Percentage Between R2 and Four DTI Metrics FWM Correlation RD FA MD AxD

r ⫽ .731, p ⬍ .0001 r ⫽ .697, p ⬍ .0001 r ⫽ .592, p ⬍ .0001 r ⫽ .087, p ⫽ .26

GWM SV% 53 49 35 .7

SWM

Correlation

SV%

r ⫽ .453, p ⬍ .0001 r ⫽ .413, p ⬍ .0001 r ⫽ .367, p ⬍ .0001 r ⫽ .110, p ⫽ .154

21 17 13 1.2

Correlation r ⫽ .189, p ⫽ .013 r ⫽ .155, p ⫽ .043 r ⫽ .120, p ⫽ .119 r ⫽ ⫺.003, p ⫽ .974

SV% 3.6 2.4 1.4 .001

The three diffusivity measures (RD, MD, and AxD) were inverted by multiplying the values by ⫺1 to achieve the same directionality as the R2 and FA measures. Thus, higher values consistently indicate better integrity and reduced diffusivity. The R2 was not correlated with AxD in any of the three regions; however, it was significantly correlated with all other DTI measures (except MD for SWM) and especially RD. The shared variance (SV%) between R2 and RD was greatest in the latest myelinating region (FWM, 53%) followed by the later myelinating CC region (GWM, 21%) and least with the earlier-myelinating SWM (3%) that is composed primarily of large visual pathways fibers. As expected, statistically adjusting for the effects of age reduces the SV% somewhat; however, all significant relationships remained significant, and specifically in FWM region, R2 and RD retained by far the highest SV% of 40% (RD vs. R2 correlation: r ⫽ .634, p ⬍ .0001). Furthermore, adjusting for the effects of age as well as sex did not meaningfully change the results beyond adjusting for age. Abbreviations as in Table 2.

earlier age than FWM AxD, RD, and MD measures (p ⬍ .0001 for AxD and MD and p ⫽ .0002 for RD). The results for the other latemyelinating region showed that peak GWM R2 is also reached at an earlier age than GWM AxD, RD, and MD (p ⬍ .0001 for all three) measures. For the early-myelinating SWM, only AxD and MD exhibited very similar inverted-U quadratic curves, and the bootstrap replication analysis showed that their peaks did not differ (p ⫽ .28). In SWM the trajectories of R2, RD, and FA metrics were linear (Figure 4 and Table 4), and once again, R2 and RD did not differ significantly (t ⫽ 1.62, p ⫽ .107), whereas the increasing trajectory of FA differed from the decreasing R2 and horizontal RD trajectories (t ⫽ 2.46, p ⫽ .015 and t ⫽ 2.21, p ⫽ .028 respectively). We used the same bootstrap method to compare the peaks across regions for those metrics with quadratic lifespan functions in all three regions (AxD, MD). For the MD, the peaks are significantly

different across all three regions (p ⬍ .0001); however, for AxD, the peak age differs significantly between FWM and GWM and between FWM and SWM (p ⬍ .0001) and was not significantly different between GWM and SWM (p ⫽ .09). Details on additional principal components analysis and assessment of alternatives to the smooth quadratic polynomial trajectory modeling are included in Supplement 1.

Discussion This study assessed subcortical WM changes over the lifespan of healthy individuals with a combination of two types of MRI biomarkers (R2 and DTI) obtained consecutively during a single scanning session. We chose to focus on three specific regions (FWM, GWM, and SWM) to interrogate the effects of crossing fibers (FWM)

Table 4. Linear and Quadratic Components of Five MRI Metrics in Three White Matter Regions

Frontal WM Correlations with Age R2 RD AxD MD FA Genu WM Correlations with Age R2 RD AxD MD FA Splenium WM Correlations with Age R2 RD AxD MD FA

Linear

Quadratic

Peak Age

r ⫽ ⫺.666, p ⬍ .0001 r ⫽ ⫺.471, p ⬍ .0001 NS, p ⫽ .21 r ⫽ ⫺.311, p ⬍ .0001 r ⫽ ⫺.569, p ⬍ .0001

t ⫽ ⫺5.84, p ⬍ .0001 t ⫽ ⫺3.96, p ⫽ .0001 t ⫽ ⫺3.67, p ⫽ .0003 t ⫽ ⫺4.64, p ⬍ .0001 NS, p ⫽ .31

32.1 35.1 54.2 42.6

r ⫽ ⫺.628, p ⬍ .0001 r ⫽ ⫺.491, p ⬍ .0001 NS, p ⫽ .74 r ⫽ ⫺.341, p ⬍ .0001 r ⫽ ⫺.457, p ⬍ .0001

t ⫽ ⫺2.22, p ⫽ .028 t ⫽ ⫺2.80, p ⫽ .006 t ⫽ ⫺3.16, p ⫽ .002 t ⫽ ⫺3.81, p ⫽ .0002 NS, p ⫽ .23

9.4 27.8 50.1 39.9

r ⫽ ⫺.166, p ⫽ .030 NS, p ⫽ .91 NS, p ⫽ .29 NS, p ⫽ .31 NS, p ⫽ .33

NS, p ⫽ .28 NS, p ⫽ .49 t ⫽ ⫺4.74, p ⬍ .0001 t ⫽ ⫺4.74, p ⬍ .0001 NS, p ⫽ .16

49.0 49.1

The metrics tabulated are R2 and four DTI metrics: RD, AxD, and MD; and FA. The functions for RD, AxD, and MD are inverted so that the peak represents best integrity (minimum diffusivity) and thus is consistent with R2 and FA where high values are considered better/optimal than lower values. Peak age refers to the maximum age a quadratic function reaches its peak. Peak ages for the quadratic components of R2 and RD progressed from older ages in latest-myelinating FWM (32.1 and 35.1 years, respectively) to younger ages (9.4 and 27.8 years, respectively) in the earlier-myelinating GWM. The AxD peak ages also progressed in the expected development-driven direction by region (FWM ⬎ GWM ⬎ SWM). The relationships with age of R2 and RD differed from the AxD metric. The AxD was atypical in that it was the only metric that had no significant linear component in any of the regions, and its quadratic component peaked approximately 2 decades later than the R2 and RD quadratic components. The relationships of MD and FA with age were an admixture of the patterns observed with RD and AxD—as would be expected, because RD and AxD are subcomponents of and incorporated into the MD and FA metrics. Like AxD, MD had quadratic components in all three regions but lacked a clear direction for the age progression between regions. The FA was the only metric without a quadratic component in any of the regions. MRI, magnetic resonance imaging; NS, not significant; other abbreviations as in Table 2.

www.sobp.org/journal

G. Bartzokis et al.

BIOL PSYCHIATRY 2012;72:1026 –1034 1031

Figure 4. Standardized lifespan trajectories of R2 and diffusion tensor imaging metrics in three white matter regions. Age at magnetic resonance imaging was centered to compare trajectories, and the metrics were standardized with mean of 0 and SD of 1 so that the scales are comparable between all five metrics. The three white matter regions are the late-myelinating FWM and two CC regions: late-myelinating GWM, and early-myelinating SWM. The metrics depicted are R2 and four diffusion tensor imaging metrics: RD, AxD, and MD diffusivities; and FA. The three diffusivity measures (RD, MD, and AxD) were inverted by multiplying the values by ⫺1 to achieve the same directionality as the R2 and FA metrics. Thus, higher values consistently indicate better integrity and reduced diffusivity. Trajectories that are not significantly quadratic are depicted by linear functions. Unlike the other metrics examined, in this region of interest analysis FWM FA regression on age is linear and thus does not seem to detect the quadratic myelination process (see Figure 1) even in this most actively myelinating late-myelinating FWM region. Furthermore, the FA trajectory in FWM and GWM is linear even though its subcomponents (AxD and RD) are both significantly quadratic. Finally, in the early-myelinating SWM region FA has a nonsignificantly increasing trajectory that counter-intuitively seems to suggest improved SWM integrity in old age (for statistics, see Table 4). Abbreviations as in Figures 1–3.

versus parallel fibers (GWM and SWM) and late-myelinating regions (FWM and GWM) where myelin content is highest versus earlymyelinating regions (SWM) where the very large axons result in reduced myelin content despite each fiber being more thickly myelinated (Figure 1B). We also chose an ROI analysis approach to mitigate some major sources of error such as susceptibility- and eddy current-induced distortions and misregistrations associated with DTI that are difficult to correct through image post-processing (27). Such misregistrations can produce partial volume errors with nearby structures such as cerebrospinal fluid and might introduce spurious age-related effects (27). Consistent with our a priori hypothesis, both R2 and RD seem to best track nonlinear (quadratic-like) trajectories of myelination that are most striking in late-myelinating FWM where these two metrics share over 50% of their variance (Figures 3 and 4; Table 3). As postmortem data suggest (Figure 1) (35,39), for both R2 and RD the developmental process of myelination peaks in adulthood with the predicted timing (earlier-myelinating GWM peak before FWM; Table 4). The R2 and RD trajectories are also consistent with postmortem studies that revealed that increasing myelin breakdown, repair, and eventual loss are most prominent in small and thinly myelinated axons of late-myelinating regions (Figure 1) (33) (reviewed in [40]). In contrast to the R2 and RD metrics, AxD had strikingly quadratic trajectories in all three ROIs and reached its apparent “optimum” (minimal diffusivity) over 2 decades later than R2 and RD. Nevertheless, AxD shared the same developmental timing pattern by region as R2 and RD. Minimum diffusivity (i.e., the inverse of its diffusivity value peaked in Figure 3) was reached first in the earlymyelinating SWM, followed by GWM, and lastly in FWM (Table 4). Despite similarities in trajectories to R2, neither RD nor the other DTI metrics track the known differences in myelin quantity between the three ROIs. Only R2 tracked known levels of myelin content with high values for the late-myelinating FWM and GWM regions and low values for early-myelinating SWM where large axons predominate (Figures 1B, 2, and 3; Table 2). As expected, the prevalence of crossing fibers in FWM markedly lowered absolute values of the three DTI metrics that incorporate directional information (FA, RD, AxD; Figures 2 and 3). Crossing fibers are minimal in GWM and SWM regions; and in these two CC regions, these same three metrics seemed to also be sensitive to regional axonal size differences (Figure 1B) with higher absolute

diffusivities in SWM compared with GWM (see overall levels in Figure 3 and Table 2). Despite some of the similarities between metrics described in the preceding text, when directly compared in this large sample, the lifespan trajectories of each of the five metrics were all significantly different from each other in each of the regions examined. Factor analysis revealed that R2, RD, and AxD loaded into their own factors. Not surprisingly, FA and MD, which are composite measures incorporating both RD and AxD metrics, loaded into more than one factor. The data thus suggest that R2, RD, and AxD are each preferentially sensitive to a unique subset of the underlying biological parameters that differ across the age span as well as brain regions, whereas the most commonly used metric (FA) is not. When used as a metric (as opposed to its initial purpose of concisely delineating directional vectors of axonal fiber bundles), FA is commonly interpreted as a measure of WM “integrity”; however, this interpretation can sometimes be confounded (see Discussion Section in Supplement 1). The underlying biology of the dynamic and interrelated agerelated differences detected by MRI is complex. Postmortem and in vivo volumetric data suggest that late-myelinating regions such as FWM and GWM continue myelinating (as measured by an expansion of total volume of myelinated WM) into middle age, followed by a reduction of this volume in old age (Figure 1C) (35–37,39,41– 43). The microstructural changes that result in this volumetric inverted U-shaped lifespan trajectory can be divided into developmental, degenerative, and reparative processes. These all occur in adulthood and are important to briefly delineate here, because they inform the interpretation of the DTI and R2 trajectories. The developmental process of myelination increases the number of axons that acquire myelin sheaths even during adulthood in latermyelinating regions (36) (Figure 1B); axon myelination increases axonal size, and increasing axon size is closely associated with thicker myelin sheaths (Figure 1B) (reviewed in [44]); after initial myelination, myelin thickness continues increasing with increasing age by as much as 15% (45)—these three volume-expanding developmental processes are expected to result in reduced extra-axonal interstitial space. The degenerative processes: 1) break down myelin sheaths and result in myelin distortions (reviewed in [40]) and localized axonal swellings during the subsequent process of repair (reviewed in [2]); 2) cause loss of myelin sheaths that is especially www.sobp.org/journal

1032 BIOL PSYCHIATRY 2012;72:1026 –1034 prominent in late-myelinating regions (33); and 3) might cause loss of entire axons when remyelination of damaged segments fails (reviewed in [2]). The myelin repair processes: 1) decrease myelin sheath thickness while maintaining axonal size (reviewed in [44]); 2) produce shortened myelin sheaths, which increase the number of nodes of Ranvier (which lack myelin) and thus increase the proportion of axons denuded of myelin (reviewed in [40]); and 3) increase the number of oligodendrocytes as they are needed to produce the new myelin of remyelinated axons (46 – 48). Unlike the developmental processes, the three degenerative processes and the first two of three repair processes listed in the preceding text would be expected to result in increased extra-axonal interstitial space. Change in AxD is predicted to be strongly impacted by extraaxonal diffusivity (31,32). In development and early adulthood, myelination-driven increases in axonal and myelin sheath thickness could reduce extra-axonal space and thus decrease AxD, whereas degenerative and reparative processes of old age would have the opposite effect (see prior paragraph). It is thus not surprising that AxD shows a highly quadratic trajectory in all three regions examined and peaks (reaches minimal diffusivity) in the expected pattern of development— earlier-myelinating regions before latermyelinating ones (Table 4). By contrast, although changes in R2 and RD are both impacted by myelin-related changes (reviewed in the introduction), some of the age-related developmental, degenerative, and reparative changes impact these two metrics differently. During development, the R2 and RD metrics are both influenced by continued myelination of previously unmyelinated axons in latemyelinating regions (Figure 1B). The impact of membrane barriers such as myelin to water diffusion quickly reaches a maximum beyond which additional layers of membrane do not have substantial impact (49). In contrast, R2 tracks the increasing quantity of myelin and might therefore be more sensitive than RD to the small agerelated increase in myelin thickness (45). Conversely, during myelin repair processes, R2 should be most sensitive to the reduction in the quantity of myelin associated with the reduced thickness of remyelinated sheaths, whereas RD would be more sensitive to the increased number of nodes of Ranvier that are devoid of myelin (40,44). Consistent with this interpretation, the sensitivity of R2 and RD to myelin quantity is most marked in late-myelinating regions where repair processes seem to be most pronounced (90% of axonal segments are remyelinated in prefrontal cortex compared with 60% in visual cortex) (reviewed in [50]). With age, the proportion of remyelinated sheaths will continually increase, and the thinning myelin and increasing nodes of Ranvier will disproportionately impact R2 and RD, resulting in an accelerated age-related decline that is most striking in FWM (Figure 4). It is also likely that, compared with RD, the accelerated age-related R2 decline as well as its earlier decline from peak R2 values might best be explained by the greater sensitivity of this metric to the ongoing repair/remyelination process that reduces the quantity of myelin (33) (reviewed in [44]). Similarly, in the SWM where the robust myelin sheaths are less vulnerable to breakdown (reviewed in [2,34]) and a frank loss of myelin is muted compared with early-myelinating regions (33) (Figure 4 and Table 4), myelin repair should be the predominant agerelated change, and— because of the initially very thick myelin in this region (Figure 1B)—the loss of myelin quantity would be the dominant change. This is consistent with the observed linear agerelated decline in the R2 trajectory and the absence of a significant decline in RD (Figure 4) and confirms postmortem data suggesting a process dominated by remyelination with replacement of very thick myelin with thinner sheaths (44) and minimal loss of myelination of large axons (33). This interpretation is cross-validated by the observed increase in AxD, because the replacement of thicker mywww.sobp.org/journal

G. Bartzokis et al. elin with the thinner myelin sheaths of remyelination should increase extra-axonal space and AxD (note: AxD data is inverted in Figure 4 and is thus depicted as an age-related decline). Given that each MRI metric is most influenced by its own individualized subset of microstructural parameters, it is not surprising that in this large sample each and every metric has a significantly different lifespan trajectory. Nevertheless, observed similarities and differences in trajectories of the different metrics included in our multimodal imaging approach help provide cross-validation and strengthen interpretation. For example, the large proportion of shared variance between R2 and RD in FWM (⬎50%) that drops precipitously and is almost nonexistent in SWM (4%) (Table 3) suggests that these metrics share a specific sensitivity to the myelination and degenerative/reparative changes of the highly vulnerable small and thinly myelinated axons that are most prevalent in latestmyelinating FWM (29,30,33). Thus, combining the understanding of the regional differences in axons, myelin (Figure 1), and myelin vulnerability (reviewed in [2,34]) with the differential sensitivity of each metric to the multiple interdependent microstructural processes can help explain the observed trajectories. An improved understanding of healthy age-related changes might help more precisely differentiate the changes associated with pathology seen in disorders such as schizophrenia and AD. By contrast, changes tracked by the composite metrics of FA and MD, despite their popularity, would by their very nature be considerably more complex and difficult to attribute to specific microstructural changes. Bearing in mind the limitations of this study—including its cross-sectional design, indirect nature of MRI assessments of myelination, age-range examined, and quadratic modeling of aging trajectories (detailed in the Discussion section of Supplement 1)—the results have important implications for understanding human brain function. By determining the speed of action potential transmission, the status of brain myelination controls the timing of action potential arrival at their destination and thus underlies the synchronized neuronal network oscillations on which optimal cognitive and behavioral functions depend (reviewed in [51]). The breakdown of myelin integrity observed in postmortem nonhuman primate studies of aging is strongly associated with premorbid cognitive performance (52). The MRI metrics have also been indirectly cross-validated against cognitive performance in humans. For example, significant associations have been observed between reductions in motor and cognitive processing speed (common features of healthy aging) and R2 and DTI measures of WM integrity (15,53–57). Some studies found these associations to be relatively specific to WM microstructure and not related to measures of brain or gray matter volume, WM hyperintensities, or vascular disease (58,59), whereas others showed premature myelin breakdown and loss in healthy individuals at increased genetic risk for AD (14,60). These studies suggest that an underlying cause-and-effect relationship might exist between WM/myelin health and cognitive and behavioral functions (reviewed in [61]). Our trajectory data confirm the suggestion that, during the life span, the healthy brain is in a constant state of change roughly defined as periods of development and maturation followed by degenerative and reparative processes and that, biologically speaking, the societal concept of a stable or unchanging adult brain is not valid (37). Deviations from normal trajectories have been demonstrated in both developmental and degenerative disorders, sometimes preceding clinical manifestation of disease (60,62) (reviewed in [61,63]). A more precise understanding of WM imaging biomarkers through multimodal imaging combined with genetic and neurocognitive measures might help define new myelin-centered avenues of disease treatment and prevention. Serial evaluations of MRI

BIOL PSYCHIATRY 2012;72:1026 –1034 1033

G. Bartzokis et al. metrics have already been used to monitor treatment-induced WM and myelination changes (64 – 67). It might thus be possible that medication development focused on correcting myelination deficits and targeting both manifest disease/deficits as well as their primary prevention could be carried out with noninvasive in vivo neuroimaging biomarkers (2).

16.

17. 18.

This work was supported in part by National Institutes of Health Grants MH0266029, AG027342, and K23-AG028727; Risk Control Strategies Inc. Alzheimer’s Foundation; and the Department of Veterans Affairs. These funding organizations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or preparation, review, and approval of the manuscript. George Bartzokis, Po H. Lu, and Jim Mintz had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. All authors report no biomedical financial interests or potential conflicts of interest.

19.

20.

21.

22.

Supplementary material cited in this article is available online. 1. Bartzokis G (2012): Neuroglialpharmacology: myelination as a shared mechanism of action of psychotropic treatments. Neuropharmacology 62:2137–2153. 2. Bartzokis G (2011): Alzheimer’s disease as homeostatic responses to age-related myelin breakdown. Neurobiol Aging 32:1341–1371. 3. Ferrie JC, Barantin L, Saliba E, Akoka S, Tranquart F, Sirinelli D, et al. (1999): MR assessment of the brain maturation during the perinatal period: Quantitative T2 MR study in premature newborns. Magn Reson Imaging 17:1275–1288. 4. Miot E, Hoffschir D, Poncy J, Masse R, Le Pape A, Akoka S (1995): Magnetic resonance imaging in vivo monitoring of T2 relaxation time: Quantitative assessment of primate brain maturation. J. Med Primatol 24: 87–93. 5. Ono J, Kodaka R, Imai K, Itagaki Y, Tanaka J, Inui K, et al. (1993): Evaluation of myelination by means of the T2 value on magnetic resonance imaging. Brain Dev 15:433– 438. 6. Miot-Noirault E, Barantin L, Akoka S, Le Pape A (1997): T2 relaxation time as a marker of brain myelination: Experimental MR study in two neonatal animal models. J Neurosci Methods 72:5–14. 7. Fazekas F, Schmidt R, Scheltens P (1998): Pathophysiologic mechanisms in the development of age-related white matter changes of the brain. Dement Geriatr Cogn Disord 9(suppl 1):2–5. 8. Takao M, Koto A, Tanahashi N, Fukuuchi Y, Takagi M, Morinaga S (1999): Pathologic findings of silent hyperintense white matter lesions on MRI. J Neurol Sci 167:127–131. 9. Bartzokis G, Mintz J, Sultzer D, Marx P, Herzberg JS, Phelan CK, et al. (1994): In vivo MR evaluation of age-related increases in brain iron. AJNR Am J Neuroradiol 15:1129 –1138. 10. Bartzokis G, Sultzer D, Lu PH, Nuechterlein KH, Mintz J, Cummings J (2004): Heterogeneous age-related breakdown of white matter structural integrity: Implications for cortical “disconnection” in aging and Alzheimer’s disease. Neurobiol Aging 25:843– 851. 11. Mueggler T, Pohl H, Baltes C, Riethmacher D, Suter U, Rudin M (2012): MRI signature in a novel mouse model of genetically induced adult oligodendrocyte cell death. Neuroimage 59:1028 –36. 12. Dyakin VV, Chen Y, Branch CA, Veeranna, Yuan A, Rao M, et al. (2010): The contributions of myelin and axonal caliber to transverse relaxation time in shiverer and neurofilament-deficient mouse models. Neuroimage 51: 1098 –1105. 13. Bartzokis G, Lu PH, Tingus K, Mendez MF, Richard A, Peters DG, et al. (2010): Lifespan trajectory of myelin integrity and maximum motor speed. Neurobiol Aging 31:1554 –1562. 14. Bartzokis G, Lu PH, Geschwind DH, Edwards N, Mintz J, Cummings JL (2006): Apolipoprotein E genotype and age-related myelin breakdown in healthy individuals: Implications for cognitive decline and dementia. Arch Gen Psychiatry 63:63–72. 15. Bartzokis G, Lu PH, Geschwind D, Tingus K, Huang D, Mendez MF, et al. (2007): Apolipoprotein E affects both myelin breakdown and cognition:

23.

24.

25.

26.

27. 28. 29.

30. 31. 32.

33.

34.

35.

36.

37.

38.

Implications for age-related trajectories of decline into dementia. Biol Psychiatry 62:1380 –1387. Lu PH, Lee GJ, Raven EP, Tingus K, Khoo T, Thompson PM, et al. (2011): Age-related slowing in cognitive processing speed is associated with myelin integrity in a very healthy elderly sample. J Clin Exp Neuropsychol 33:1059 –1068. Wheeler-Kingshott CA, Cercignani M (2009): About “axial” and “radial” diffusivities. Magn Reson Med 61:1255–1260. Madden DJ, Bennett IJ, Burzynska A, Potter GG, Chen NK, Song AW (2012): Diffusion tensor imaging of cerebral white matter integrity in cognitive aging. Biochim Biophys Acta 1822:386 – 400. Song SK, Sun SW, Ju WK, Lin SJ, Cross AH, Neufeld AH (2003): Diffusion tensor imaging detects and differentiates axon and myelin degeneration in mouse optic nerve after retinal ischemia. Neuroimage 20:1714 – 1722. Song SK, Yoshino J, Le TQ, Lin SJ, Sun SW, Cross AH, et al. (2005): Demyelination increases radial diffusivity in corpus callosum of mouse brain. Neuroimage 26:132–140. Zhang J, Jones MV, McMahon MT, Mori S, Calabresi PA (2012): In vivo and ex vivo diffusion tensor imaging of cuprizone-induced demyelination in the mouse corpus callosum. Magn Reson Med 67:750 –759. Chandran P, Upadhyay J, Markosyan S, Lisowski A, Buck W, Chin CL, et al. (2012): Magnetic resonance imaging and histological evidence for the blockade of cuprizone-induced demyelination in C57BL/6 mice. Neuroscience 202:446 – 453. Larvaron P, Boespflug-Tanguy O, Renou JP, Bonny JM (2007): In vivo analysis of the post-natal development of normal mouse brain by DTI. NMR Biomed 20:413– 421. Klawiter EC, Schmidt RE, Trinkaus K, Liang HF, Budde MD, Naismith RT, et al. (2011): Radial diffusivity predicts demyelination in ex vivo multiple sclerosis spinal cords. Neuroimage 55:1454 –1460. Schmierer K, Wheeler-Kingshott CA, Tozer DJ, Boulby PA, Parkes HG, Yousry TA, et al. (2008): Quantitative magnetic resonance of postmortem multiple sclerosis brain before and after fixation. Magn Reson Med 59:268 –277. Gao W, Lin W, Chen Y, Gerig G, Smith JK, Jewells V, et al. (2009): Temporal and spatial development of axonal maturation and myelination of white matter in the developing brain. AJNR Am J Neuroradiol 30:290 – 296. Jones DK, Cercignani M (2010): Twenty-five pitfalls in the analysis of diffusion MRI data. NMR Biomed 23:803– 820. Tournier JD, Mori S, Leemans A (2011): Diffusion tensor imaging and beyond. Magn Reson Med 65:1532–1556. Raschke C, Schmidt S, Schwab M, Jirikowski G (2008): Effects of betamethasone treatment on central myelination in fetal sheep: An electron microscopical study. Anat Histol Embryol 37:95–100. Riise J, Pakkenberg B (2011): Stereological estimation of the total number of myelinated callosal fibers in human subjects. J Anat 218:277–284. Sen PN, Basser PJ (2005): A model for diffusion in white matter in the brain. Biophys J 89:2927–2938. Choe AS, Stepniewska I, Colvin DC, Ding Z, Anderson AW (2012): Validation of diffusion tensor MRI in the central nervous system using light microscopy: Quantitative comparison of fiber properties. NMR Biomed 25:900 –908. Marner L, Nyengaard JR, Tang Y, Pakkenberg B (2003): Marked loss of myelinated nerve fibers in the human brain with age. J Comp Neurol 462:144 –152. Bartzokis G (2004): Age-related myelin breakdown: A developmental model of cognitive decline and Alzheimer’s disease. Neurobiol Aging 25:5–18. Kemper T (1994): Neuroanatomical and neuropathological changes during aging and dementia. In: Albert M, Knoefel J, editors. Clinical Neurology of Aging, 2nd ed. New York: Oxford University Press, 3– 67. Lamantia AS, Rakic P (1990): Cytological and quantitative characteristics of four cerebral commissures in the rhesus monkey. J Comp Neurol 291:520 –537. Bartzokis G, Beckson M, Lu PH, Nuechterlein KH, Edwards N, Mintz J (2001): Age-related changes in frontal and temporal lobe volumes in men: A magnetic resonance imaging study. Arch Gen Psychiatry 58:461– 465. Bartzokis G, Cummings JL, Sultzer D, Henderson VW, Nuechterlein KH, Mintz J (2003): White matter structural integrity in healthy aging adults

www.sobp.org/journal

1034 BIOL PSYCHIATRY 2012;72:1026 –1034

39.

40. 41. 42.

43. 44.

45.

46.

47.

48.

49. 50.

51.

52.

53.

54.

and patients with Alzheimer disease: A magnetic resonance imaging study. Arch Neurol 60:393–398. Yakovlev PI, Lecours AR (1967): The myelogenetic cycles of regional maturation of the brain. In: Minkowski A, editor. Regional Development of the Brain in Early Life. Boston: Blackwell Scientific Publications, 3-70. Peters A (2009): The effects of normal aging on myelinated nerve fibers in monkey central nervous system. Front Neuroanat 3:11. Flechsig P (1901): Developmental (myelogenetic) localisation of the cerebral cortex in the human subject. Lancet 158:1027–1030. Meyer A (1981): Paul Flechsig’s system of myelogenetic cortical localization in the light of recent research in neuroanatomy and neurophysiology. Part I. Can J Neurol Sci 8:1– 6. Benes FM (1989): Myelination of cortical-hippocampal relays during late adolescence. Schizophr Bull 15:585–593. Fancy SP, Chan JR, Baranzini SE, Franklin RJ, Rowitch DH (2011): Myelin regeneration: A recapitulation of development? Annu Rev Neurosci 34: 21– 43. Peters A, Sethares C, Killiany RJ (2001): Effects of age on the thickness of myelin sheaths in monkey primary visual cortex. J Comp Neurol 435: 241–248. O’Kusky J, Colonnier M (1982): Postnatal changes in the number of astrocytes, oligodendrocytes, and microglia in the visual cortex (area 17) of the macaque monkey: A stereological analysis in normal and monocularly deprived animals. J Comp Neurol 210:307–315. Peters A, Sethares C (2004): Oligodendrocytes, their progenitors and other neuroglial cells in the aging primate cerebral cortex. Cereb Cortex 14:995–1007. Peters A, Verderosa A, Sethares C (2008): The neuroglial population in the primary visual cortex of the aging rhesus monkey. Glia 56:1151– 1161. Beaulieu C (2002): The basis of anisotropic water diffusion in the nervous system—a technical review. NMR Biomed 15:435–55. Peters A, Kemper T (2012): A review of the structural alterations in the cerebral hemispheres of the aging rhesus monkey. Neurobiol Aging 33:2357–2372. Bartzokis G (2002): Schizophrenia: breakdown in the well-regulated lifelong process of brain development and maturation. Neuropsychopharmacology 27:672– 683. Peters A, Sethares C (2002): Aging and the myelinated fibers in prefrontal cortex and corpus callosum of the monkey. J Comp Neurol 442:277– 291. Kochunov P, Coyle T, Lancaster J, Robin DA, Hardies J, Kochunov V, et al. (2009): Processing speed is correlated with cerebral health markers in the frontal lobes as quantified by neuroimaging. Neuroimage 49:1190 – 1199. Ryan L, Walther K, Bendlin BB, Liu LF, Walker DG, Glisky EL (2011): Age-related differences in white matter integrity and cognitive function are related to APOE status. Neuroimage 54:1565–1577.

www.sobp.org/journal

G. Bartzokis et al. 55. Penke L, Maniega SM, Murray C, Gow AJ, Hernandez MC, Clayden JD, et al. (2010): A general factor of brain white matter integrity predicts information processing speed in healthy older people. J Neurosci 30: 7569 –7574. 56. Vernooij MW, Ikram MA, Vrooman HA, Wielopolski PA, Krestin GP, Hofman A, et al. (2009): White matter microstructural integrity and cognitive function in a general elderly population. Arch Gen Psychiatry 66: 545–553. 57. Burgmans S, Gronenschild EH, Fandakova Y, Shing YL, van Boxtel MP, Vuurman EF, et al. (2011): Age differences in speed of processing are partially mediated by differences in axonal integrity. Neuroimage 55: 1287–1297. 58. Charlton RA, Schiavone F, Barrick TR, Morris RG, Markus HS (2011): Diffusion tensor imaging detects age related white matter change over a 2 year follow-up which is associated with working memory decline. J Neurol Neurosurg Psychiatry 81:13–19. 59. Bosch B, Arenaza-Urquijo EM, Rami L, Sala-Llonch R, Junque C, SolePadulles C, et al. (2012): Multiple DTI index analysis in normal aging, amnestic MCI and AD. Relationship with neuropsychological performance. Neurobiol Aging 33:61–74. 60. Ringman J, O’Neill J, Geschwind D, Medina L, Apostolova L, Rodriguez Y, et al. (2007): Diffusion tensor imaging in preclinical and presymptomatic carriers of familial Alzheimer’s disease mutations. Brain 130:1767–1776. 61. Bartzokis G (2011): Neuroglialpharmacology: White matter pathophysiologies and psychiatric treatments. Front Biosci 16:2695–2733. 62. Braskie MN, Jahanshad N, Stein JL, Barysheva M, McMahon KL, de Zubicaray GI, et al. (2010): Common Alzheimer’s disease risk variant within the CLU gene affects white matter microstructure in young adults. J Neurosci 31:6764 – 6770. 63. Peters BD, Blaas J, de Haan L (2010): Diffusion tensor imaging in the early phase of schizophrenia: What have we learned? J Psychiatr Res 44:993– 1004. 64. Venneri A, Lane R (2009): Effects of cholinesterase inhibition on brain white matter volume in Alzheimer’s disease. Neuroreport 20:285–288. 65. Likitjaroen Y, Meindl T, Friese U, Wagner M, Buerger K, Hampel H, et al. (2011): Longitudinal changes of fractional anisotropy in Alzheimer’s disease patients treated with galantamine: A 12-month randomized, placebo-controlled, double-blinded study. Eur Arch Psychiatry Clin Neurosci 262:341–350. 66. Berger GE, Wood SJ, Ross M, Hamer CA, Wellard RM, Pell G, et al. (2012): Neuroprotective effects of low-dose lithium in individuals at ultra-high risk for psychosis. A longitudinal MRI/MRS study. Curr Pharm Des 18: 570 –575. 67. Bartzokis G, Lu PH, Amar CP, Raven EP, DeTore NR, Altshuler LL, et al. (2011): Long acting injection versus oral risperidone in first-episode schizophrenia: Differential impact on white matter myelination trajectory. Schizophrenia Research 132:35– 41. 68. Kaes T (1907): Die Grosshirnrinde des Menschen in ihren Massen und in ihrem Fasergehalt. Jena: Gustav Fisher.