Evaluating g-ratio weighted changes in the corpus callosum as a function of age and sex
Introduction
The study of human white matter has progressed tremendously in recent years. It is now broadly accepted that the microstructural properties of white matter change over the course of normal development (Bowley et al., 2010, Callaghan et al., 2014, Nagy et al., 2004, Yeatman et al., 2014), and correlate with cognitive functions (Moeller et al., 2015, Voineskos et al., 2012, Yeatman et al., 2011, Yeatman et al., 2012, Zatorre et al., 2012), and neuropathological states (Compston and Coles, 2008, Fields, 2008). Many of these findings were made possible by advancements in in vivo magnetic resonance imaging (MRI) techniques, including quantitative MRI (qMRI), and in particular diffusion MRI (dMRI). Recent years have seen a growing interest in relating MRI measured parameters to the structure and biophysical properties of the fibers in white matter. A recent proposal suggests that a fundamental property of the white matter, the fiber g-ratio, can be modeled by combining dMRI and other myelin sensitive qMRI signals (Dean et al., 2016, Mohammadi et al., 2015, Stikov et al., 2012, Stikov et al., 2015a, Stikov et al., 2015b, West et al., 2015).
The fiber g-ratio is defined as the ratio of the inner to outer radius of the myelin sheath wrapped around the axon. It has been analytically shown that the g-ratio affects the signal conduction velocity of the axons (Rushton, 1951), as well their energy consumption and conduction fidelity (Chomiak and Hu, 2009). Theoretical work has concluded that the optimal g-ratio is between 0.6 and 0.75. These studies, together with suggested coupling of axonal growth and myelination (de Waegh et al., 1992, Friede, 1972, Griffiths et al., 1998, Nave, 2010), have encouraged the assumption that the average g-ratio is fairly constant across the central nervous system (CNS) although, until very recently, this had not been tested in vivo in the human brain.
Studies that measure the g-ratio ex vivo in the CNS typically use electron microscopy (EM), and focus on either the corpus callosum or the optic nerve, where the axon bundles are clearly separated from their environment and their direction is known. The g-ratio values reported are around 0.67 for monkeys (Stikov et al., 2015a, Stikov et al., 2015b), 0.81 for guinea pigs (Guy et al., 1989), and 0.76 for mice (Arnett et al., 2001). These values are congruent with the optimal g-ratio values suggested by theoreticians (Chomiak and Hu, 2009, Rushton, 1951).
Claims for the significance of the g-ratio in white matter functionality highlight possible effects on cognitive functions and behavior (Paus and Toro, 2009), and may be related to abnormalities in subjects suffering from demyelinating diseases such as multiple sclerosis (Albert et al., 2007, Hampton et al., 2012, Wolswijk, 2002). Furthermore, it was suggested that the g-ratio changes during development in a sexually dimorphic fashion (Paus, 2010, Paus and Toro, 2009). Ex vivo measurements of the g-ratio show sex differences in the splenium of young rats (Pesaresi et al., 2015), however, it is still unclear whether these differences are also present in humans and whether the effect seen in rats reflects a measurable difference that can be detected in vivo.
Recent in vivo studies measured changes in the g-ratio as a function of age. Dean et al. reported a decrease in the g-ratio during early development, between 3 months and 7.5 years of age (Dean et al., 2016). Cercignani et al. measured the g-ratio of aging subjects, ages 20–80, and found an increase that was similar between males and females (Cercignani et al., 2017). Interestingly, this aging effect is in contrast with recent results in rodents that show decrease in g-ratio in the optic nerve in aging mice (Stahon et al., 2016).
A variety of qMRI approaches have been proposed to model the g-ratio based on in vivo measurements and a number of research groups (Dean et al., 2016, Mohammadi et al., 2015, Stikov et al., 2015a, Stikov et al., 2015b, West et al., 2016) have proposed methods for estimating the g-ratio by using different MRI acquisitions. These methods utilize advanced dMRI techniques to model the voxel-wise axon volume fraction (AVF), or the volume fraction of a voxel that contains fibers (axons and myelin) (also called fiber volume fraction (FVF)). Techniques used include AxCaliber (Assaf and Blumenfeld-Katzir, 2008), NODDI (Stikov et al., 2015a, Stikov et al., 2015b, Zhang et al., 2012), ActiveAx (Alexander et al., 2010, Duval et al., 2017), TFD (Mohammadi et al., 2015, Reisert et al., 2013, Ellerbrock and Mohammadi, 2016), and tensor modeling (Campbell et al., 2017, Stikov et al., 2012). The dMRI models can then be combined with other qMRI models that measure the myelin volume fraction (MVF), which refers to the volume within a voxel that is occupied by myelin sheaths. MVF estimates are not experimentally straightforward and include parameters of quantitative magnetization transfer (qMT) (Helms et al., 2008, Mohammadi et al., 2015, Sled and Pike, 2001, Stikov et al., 2015a, Stikov et al., 2015b, West et al., 2016, Yarnykh, 2007), which require scaling in order to achieve the MVF values (Campbell et al., 2017, Cercignani et al., 2017, Stikov et al., 2015a, Stikov et al., 2015b, West et al., 2016). Another potentially useful parameter is the myelin water fraction (MWF) (Mackay et al., 1994, Meyers et al., 2017, West et al., 2016). MWF is specific to myelin, but requires extended data acquisition and analysis that is rarely practical in humans (but could be of potential use in the future). Alternatively, MWF can be estimated with a faster acquisition that relies on additional modeling assumptions (Deoni et al., 2013, Deoni et al., 2008).
Recently, Duval et al. (2017) suggested the use of the lipid and macro-molecular tissue volume (MTV), derived from proton density (PD) as an estimate for MVF, when measuring the g-ratio in the spinal cord. While MTV may be less specific for myelin than some alternatives (qMT, MWF), these three measurements are correlated with each other (Dula et al., 2010, Mezer et al., 2013, West et al., 2016). This correlation is higher than the qMT scan-rescan reliability as measured in Stikov et al. (2011), and it was recently shown that PD is highly correlated with histologically measured MVF (West et al., 2016). Furthermore, the calculation of g-ratio with MTV as an estimate for MVF is comparable with other calculations (Ellerbrock and Mohammadi, 2016). Importantly, MTV, which is based on relatively simple acquisition protocols, can be measured with very a high signal-to-noise ratio in feasible scan times (Abbas et al., 2014, Abbas et al., 2015, Volz et al., 2012). Thus, estimating MVF via MTV may provide an accurate and straightforward method to obtain a measure of the g-ratio. We term the MTV-based g-ratio-weighted MRI measurement GR* since a portion of the variance in the signal might also reflect other tissue properties.
In the current study, we present the measurement of GR* in the corpus callosum of 92 subjects, and evaluate changes in GR* as a function of age and sex. Our results indicate first that the GR* values obtained in the corpus callosum are similar to the g-ratio measured previously. Further findings are that GR* is generally constant with age with no sex-related differences identified.
Section snippets
Methods
The current study utilized a large human database that was also used for our previous work (Yeatman et al., 2014) and (Gomez et al., 2017), as well as rodent data from West et al. (2016).
g-Ratio estimation across modalities
Fig. 1a shows the relationship between the histological measurement of the myelin volume fraction and the estimated MTV, derived from the multi-exponential T2 model. The high coefficient of determination (R2 = 0.74) is similar to the values found with more traditional myelin imaging methods (qMT, r = 0.84, and multi-exponential T2, r = 0.81 (West et al., 2016)). Importantly, MTV estimations are unbiased by water exchange, while it has been shown to effect MWF and to a lesser extent qMT (
Discussion
Our analysis of 92 subjects demonstrates that MTV can be used successfully to estimate the g-ratio in the corpus callosum in vivo. The results in Fig. 1 show that MTV represents a valid approximation of myelin volume to calculate GR*. Fig. 3 shows the results are in agreement with the literature reports of ex vivo and in vivo measurements and also agree with theoretical calculations of optimal g-ratio values. Although in two out of eight sub-regions in the corpus callosum we did detect a small
Conclusions
The MTV derived GR* provides a simple and reliable in vivo estimation of the g-ratio in the human brain. It was used to test the stability of g-ratio with age and between the sexes. Unlike other tissue parameters measured with MRI, GR* was shown to be mostly constant across a large age range. We therefore propose that these results support theoretical evidence suggesting that the g-ratio must remain within a narrow range of values around an optimal value for white matter function.
Acknowledgements
We acknowledge R. Schurr for helpful advice and feedback. This work was supported ISF grant 0399306, awarded to A.A.M, by NSF/SBE-BSF (NSF no. 1551330 and BSF no. 2015608) awarded to A.A.M by Research on Schizophrenia and Depression (NARSAD) Young Investigator Grant from the Brain and Behavior Research Foundation awarded to A.A.M and J.D.Y, and NIH grant R01EB019980 awarded to M.D.D. We thank J. Gomez and K. Grill-Spector for providing data for testing reproducibility, their work was supported
References (86)
- et al.
Quantitative water content mapping at clinically relevant field strengths: a comparative study at 1.5T and 3T
NeuroImage
(2015) - et al.
Fiber composition of the human corpus callosum
Brain Res.
(1992) - et al.
Orientationally invariant indices of axon diameter and density from diffusion MRI
NeuroImage
(2010) - et al.
Structural properties of the human corpus callosum: multimodal assessment and sex differences
NeuroImage
(2017) - et al.
Widespread age-related differences in the human brain microstructure revealed by quantitative magnetic resonance imaging
Neurobiol. Aging
(2014) - et al.
Characterizing axonal myelination within the healthy population: a tract-by-tract mapping of effects of age and gender on the fiber g-ratio
Neurobiol. Aging
(2017) - et al.
Multiple sclerosis
Lancet
(2008) - et al.
Mapping an index of the myelin g-ratio in infants using magnetic resonance imaging
NeuroImage
(2016) - et al.
Local modulation of neurofilament phosphorylation, axonal caliber, and slow axonal transport by myelinating Schwann cells
Cell
(1992) - et al.
g-Ratio weighted imaging of the human spinal cord in vivo
Neuroimage
(2017)