Quantitative diffusion tensor imaging analysis does not distinguish pediatric canines with mucopolysaccharidosis I from control canines

Abstract

Purpose

We compared fractional anisotropy and radial diffusivity measurements between pediatric canines affected with mucopolysaccharidosis I and pediatric control canines. We hypothesized that lower fractional anisotropy and higher radial diffusivity values, consistent with dysmyelination, would be present in the mucopolysaccharidosis I cohort.

Methods

Six canine brains, three affected with mucopolysaccharidosis I and three unaffected, were euthanized at 7 weeks and imaged using a 7T small-animal magnetic resonance imaging system. Average fractional anisotropy and radial diffusivity values were calculated for four white-matter regions based on 100 regions of interest per region per specimen. A 95% confidence interval was calculated for each mean value.

Results

No difference was seen in fractional anisotropy or radial diffusivity values between mucopolysaccharidosis affected and unaffected brains in any region. In particular, the 95% confidence intervals for mucopolysaccharidosis affected and unaffected canines frequently overlapped for both fractional anisotropy and radial diffusivity measurements. In addition, in some brain regions a large range of fractional anisotropy and radial diffusivity values were seen within the same cohort.

Conclusion

The fractional anisotropy and radial diffusivity values of white matter did not differ between pediatric mucopolysaccharidosis affected canines and pediatric control canines. Possible explanations include: (a) a lack of white matter tissue differences between mucopolysaccharidosis affected and unaffected brains at early disease stages; (b) diffusion tensor imaging does not detect any existing differences; (c) inflammatory processes such as astrogliosis produce changes that offset the decreased fractional anisotropy values and increased radial diffusivity values that are expected in dysmyelination; and (d) our sample size was insufficient to detect differences. Further studies correlating diffusion tensor imaging findings to histology are warranted.

Introduction

Diffusion tensor imaging (DTI) characterizes the microscopic diffusion of water in tissues. This technique can be used to estimate the degree of diffusion anisotropy, or directionality, in white matter tracts. Isotropic diffusion refers to microscopic diffusion of equal magnitudes in all directions, as occurs in a liquid with no diffusion barriers, or in a microstructural environment where barriers are disorganized. When these barriers are organized, they can cause non-random, or anisotropic, diffusion.

The degree of anisotropy is directly related to the geometry of the diffusion barriers, such as myelin, glial cells, cell membranes, cytoskeletal structures, and macromolecules in the extracellular space.^1,2 In particular, compact white matter structures that are composed of parallel myelinated axons, such as the corpus callosum, provide a substantial barrier to random diffusion, causing water to diffuse primarily along the long axis of the structure.

With DTI analysis, it is possible to describe this diffusion quantitatively by measuring a variety of metrics, such as fractional anisotropy (FA) and radial diffusivity (RD). FA represents the degree of diffusion along a primary axis, indicating directionality. RD represents the magnitude of diffusion along the two non-dominant diffusion axes, and has been correlated to myelination in small animals.²

The FA values of intact white matter structures composed of compact or highly parallel fibers would be expected to be relatively high and the RD values low, because myelin provides a strong diffusion barrier. By measuring FA and RD, one can estimate the structural integrity of the associated white matter region to understand better brain development and maturation,³ normal anatomy, the pathology of various white matter diseases,⁴ and the result of potential treatments for these diseases.⁵

The purpose this study was to compare the FA and RD values between pediatric control canines and pediatric canines affected with mucopolysaccharidosis I (MPS I). MPS I is a lysosomal storage disease that causes developmental delay, enlarged visceral organs, coarse facial features, hydrocephalus, and potentially dysmyelination. FA is usually decreased and RD increased in diseases that cause dysmyelination, such as MPS,^2,5–7 a trend that is also seen in human demyelinating diseases.^1,4,8 We hypothesized that DTI metrics consistent with dysmyelination would be present in a cohort of pediatric canines with MPS I.

Methods

Specimen acquisition and preparation

The brains of six post-mortem canines, three affected with MPS and three unaffected, were obtained from Iowa State University Department of Animal Science. The canines had been euthanized at 7 weeks as part of an independent research study that had been performed under the approval by the Institutional Animal Care and Use Committee at Iowa State University. All canines had been administered intravenous heparin (0.5 ml, 500 U/dog) 10 minutes before euthanasia. They were then anesthetized with Euthasol (200 mg/kg), until respiration and cardiac activity had stopped and corneal and pedal reflexes were absent. Perfusion catheters were placed in each of the carotid arteries, the arteries clamped below the catheter placement and the abdominal aorta cut. The dogs were perfused first with a 0.9% sodium chloride solution and then with a solution of 10% Magnevist (Bayer Healthcare Pharmaceuticals, Whippany, NJ, USA) in 10% formalin.

Following perfusion, most of the skin, subcutaneous tissue, and muscle were removed from the head. The heads were then placed in a 10% formalin solution and allowed to post-fix for 24 hours at room temperature. After 24 hours, the heads were transferred to a solution of 1% Magnevist in 10% formalin in order to lower the T1 of brain tissue and allow for a shorter magnetic resonance imaging (MRI) protocol and shipped to Duke University at room temperature. Upon receipt of the heads, we carefully removed the brains from the skulls using an atraumatic technique and placed them in a solution of 1% ProHance (Bracco Diagnostics Inc., Monroe Township, NJ, USA) in 10% formalin. Two weeks before imaging, the brains were immersed in a solution of 0.5% ProHance in phosphate-buffered saline (pH 7.4) in order to allow the tissue to rehydrate and to reduce the T1 relaxation time and allow for a shorter imaging acquisition.

Imaging

All brains were scanned using an identical protocol on a 7T small-animal MRI system (Magnex Scientific, Yarnton, Oxford, UK) equipped with 670 mT/m Resonance Research gradient coils (Resonance Research Inc., Billerica, MA, USA) using a 65 mm internal diameter quadrature radiofrequency coil (M2M Imaging, Cleveland, OH, USA). The system was controlled with an Agilent direct drive console (Agilent Technologies, Santa Clara, CA, USA). The images were acquired using a custom-designed six direction diffusion-weighted spin-echo pulse sequence (repetition time 100 ms, echo time 18.1 ms, number of excitations 1, b value 1506 s/mm²⁾. The acquisition matrix was optimized to fit the dimensions of each canine brain and adjusted for a field of view producing a Nyquist-limited isotropic voxel size of 100 µm and a slice thickness of 100 µm. Diffusion preparation was accomplished using a modified Tanner–Stejskal diffusion-encoding scheme with a pair of unipolar, half-sine diffusion gradient waveforms. Total acquisition time was approximately 40 hours for each brain. Following image acquisition, the data for each brain were then smoothed using the SUSAN de-noising algorithm implemented in FSL with a three-voxel-kernel radius. For each brain, maps of the three eigenvalues, the three eigenvectors, and FA were reconstructed using Diffusion Toolkit version 0.6.2. RD maps were calculated by averaging the λ₂ and λ₃ eigenvalue maps, which represent the two non-dominant diffusion axes as discussed above, using ImageJ version 1.5. The equations for FA and RD in terms of the three eigenvalues are shown in equations (1) and (2), respectively.

$$\mathrm{FA}=\sqrt{\frac{3}{2}}\frac{\sqrt{({\lambda }_1-{\lambda }_m{)}^2+({\lambda }_2-{\lambda }_m{)}^2+({\lambda }_3-{\lambda }_m{)}^2}}{\sqrt{{\lambda }_1^2+{\lambda }_2^2+{\lambda }_3^2}}$$ (1)

with ${\lambda }_m=\frac{{\lambda }_1+{\lambda }_2+{\lambda }_3}{3}$ or the mean of the eigenvalues, referred to as mean diffusivity

$$\mathrm{RD}=\frac{{\lambda }_2+{\lambda }_3}{2}$$ (2)

DTI metric acquisition and comparison

We interrogated four brain regions: the anterior and posterior internal capsule (AIC and PIC, respectively), and the anterior and posterior centrum semiovale (ACS and PCS, respectively). To determine the FA and RD values within these brain regions, regions of interest (ROIs) were placed on coronal sections of the B0 map for each brain using ImageJ. For each region, we defined the region’s anatomical boundaries in the dorsal, ventral, rostral, and caudal directions. We then selected 10 contiguous slices containing the most robust white matter volume within each of the four brain regions. These slices were defined using anatomical landmarks in order to ensure that comparable slices were sampled in all specimens and that the interrogation method would be reproducible. Once the boundaries were defined, two independent raters placed 10 non-overlapping, square ROIs (five per hemisphere) on each slice, for a total of 100 ROIs per brain region per specimen for each rater. This method of ROI sampling has previously been described in previous studies by our group.⁹

These ROIs were then superimposed on the FA, RD, and the three eigenvalue maps. In all six specimens, an average FA and RD value was calculated from measurements from 100 ROIs in each of four sampled brain regions (AIC, PIC, ACS, and PCS). A 95% confidence interval was calculated for each mean.

As a quality assessment check, we determined the inter-observer variability between two readers. Using this technique, the greatest difference in FA or RD measurements between two raters from a given brain region of the same specimen was 5.8%, and less than 2% for the majority of the regions.

Results

Example coronal B0 images of the PIC in both 7-week-old MPS affected and unaffected canines are shown in Figure 1. Note that the MPS brain has enlarged perivascular (or Virchow–Robin) spaces, and the differentiation between white matter and grey matter is relatively poor.

Figure 1.

Representative coronal B0 images (i.e. before diffusion gradients were applied) acquired from (a) a control brain and (b) an MPS affected brain. Arrows on the MPS specimen indicate some of the many dilated perivascular (or Virchow–Robin) spaces, which are not present in the control brain. Also note the relatively poor differentiation between white matter and grey matter in the MPS affected specimen. MPS: mucopolysaccharidosis.

Comparison of DTI metrics between groups

A comparison of FA and RD values between groups is presented in Figure 2. In all four regions, the 95% confidence intervals for FA measurements within a cohort frequently overlapped between MPS affected and unaffected canines. The same was true for RD measurements.

Figure 2.

Average (a) FA and (b) RD measurements over 100 ROIs in each of the four sampled brain regions for all six specimens: three control subjects compared to three MPS brains. Error bars represent the 95% confidence interval for each mean value. Dotted lines indicate the presence of overlap in the 95% confidence intervals between one or more mean values in opposing cohorts. A confidence interval that crosses or falls completely between these dotted lines indicates that it overlaps with one or more confidence intervals from the opposing cohort. FA: fractional anisotropy; RD: radial diffusivity; ROI: region of interest; MPS: mucopolysaccharidosis.

In some brain regions, the 95% confidence intervals for measurements from individual specimens within the same cohort (as opposed to the confidence intervals for an entire cohort mentioned earlier) did not show overlap (Figure 2). This lack of overlap within a cohort was observed for FA measurement in the AIC and PIC, and RD measurements in the ACS and PCS.

When measurements from the three specimens within a cohort were averaged to generate one representative FA value and one representative RD value for each region, we found that for all four regions, the average FA value was greater in the MPS cohort than in the control cohort. The average RD value in the MPS cohort was less than that of the control cohort in the AIC and PIC. The average RD value was greater in the MPS cohort than the control cohort in the ACS and PCS. These findings are contrary to what is typically seen in a dysmyelinating process. Because values within a cohort were often significantly different from each other, we elected to represent each specimen individually, rather than one average value per cohort.

Discussion

In a prior study, we found that FA values in 18-week-old MPS canine brains were consistently lower, and RD values consistently higher, than in age-matched unaffected dogs.¹⁰ However, our findings in the current study on 7-week-old canines follow a different pattern. The major findings in this study were twofold. First, in each brain region, no difference in DTI metrics was seen between MPS affected and unaffected brains, as seen by the lack of substantive overlap in confidence intervals for both FA and RD measurements between cohorts. Second, in each brain region a statistically significant difference was frequently seen in the FA between at least two specimens within the same cohort. The same was true of the RD measurements. These findings reject our hypothesis that the FA and RD values of white matter in pediatric canines with MPS would be consistently different from those in control animals and depict the typical pattern of chronic dysmyelination. However, as outlined below, we believe that the findings outlined here may be typical of the acute (as opposed to chronic) stages of dysmyelination.

One possible explanation is that during this early stage of the disease, no differences actually exist between the white matter tissue in MPS affected and unaffected brains. Because the canines in this study were all 7 weeks old, they were undergoing active myelination. According to histological work done in a study on 17 normal canines,¹¹ cerebral myelination in canines progresses from 1 to 12 weeks. Depending on how the MPS disease process interferes with the myelination process, white matter could still be quite similar in 7-week-old MPS affected and unaffected canines if the dysmyelination process was dominant later than 7 weeks. Alternatively, the onset of MPS-related white matter changes could be delayed in some canines but not others, which could be partly responsible for the overlap in FA and RD measurements between the two cohorts and lack of overlap between measurements within a cohort for some brain regions.

It is also possible that real histological differences in white matter do exist between the two cohorts, but that DTI lacks sufficient sensitivity to detect these differences at this early stage of the disease. Many factors can contribute to DTI findings in maturing brains. If these factors influence DTI metrics to a greater extent than MPS in the early stages of disease, they could prevent DTI from detecting a typical demyelination pattern even in the presence of histological changes. Although RD has been found to correlate with myelination in experiments on rats and mice,^2,12–15 other factors such as an increase in axonal diameter, decrease in extracellular water, and increased complexity of the extracellular architecture¹⁶ have also been suggested to decrease RD in normal human subjects. As these microstructural changes occur during brain maturation, they could decrease the RD values in pediatric canines to the point where RD in diseased pediatric canines does not differ from that of normal controls, even in the setting of dysmyelination. Similarly, FA probably reflects the quantity of microtubule-associated proteins within an axon, the degree of glial proliferation, and the tightness of axonal packing.^13,17–20 These microstructural factors are also likely to be in flux during brain maturation, and therefore could increase FA in the pediatric canine. If the influence of these factors on FA measurements was large enough, it could overshadow the influence of any existing dysmyelination, leading to similar FA values in the diseased brains and normal controls. This might then prevent DTI from detecting any early changes from MPS in pediatric canines. If so, such differences might be revealed by an alternative form of diffusion imaging, such as kurtosis imaging^21–24 or diffusion spectrum imaging,^25–26 or a different form of DTI analysis.²⁷

Alternatively, the findings in this study could reflect microstructural changes at early stages of MPS that elevate FA values and lower RD values, such as inflammation secondary to the disease process. Support for inflammation altering DTI metrics, specifically RD, in demyelination can be found in studies using a murine demyelination model.^28,29 In that study, the investigators found that, although RD increased in chronic demyelination, it remained stable in areas of acute demyelination and inflammatory cell infiltration. They hypothesized that the lack of change in RD during acute myelination was the result of a decrease in intra-axonal water diffusivity after acute injury and/or increased restriction arising from inflammatory cells. Similarly, another study found that RD in early demyelination was similar to that of age-matched control mice, which was attributed to reduced extracellular space resulting from axonal swelling.²⁹ The authors hypothesized that the opposing actions of axonal damage and inflammation reduced RD and demyelination increased RD offset each other, resulting in no net change in RD in acute demyelination. Notably, inflammation has previously been shown to contribute to the disease pathology of MPS. In particular, as substrates accumulate in lysosomal storage diseases, an inflammatory response is triggered that increases with storage burden.³⁰

As part of the presumed inflammatory process, the presence of astrocytic gliosis and activated microglia in the MPS cohort may lead to no substantive difference between RD values between groups. In support of this argument is the fact that in another disease process (i.e. Creutzfeldt–Jakob disease), affected white matter has been shown to exhibit patchy foci of demyelination and decreased RD in association with diffuse astrocytic gliosis and activated microglia.³¹ As such, inflammation and astrogliosis could contribute to our finding of a lack of elevated RD values in 7-week-old MPS affected canines compared to age-matched controls. In addition, studies on a murine MPS I model have found perineuronal activation microglia within the central nervous system.³² In summary, astrocytic gliosis and activated microglia may, at least in part, be the basis for a lack of elevation of RD values at the relatively early stage of the dysmyelination process.

In conclusion, although the etiology of these findings cannot be determined without histological investigation, we have found no differences in the FA or RD values of white matter regions in MPS affected and unaffected canines, as originally hypothesized. We have also shown that FA and RD measurements in 7-week-old canines are variable in both MPS affected and unaffected specimens. Future histological work and DTI analysis in older canines are needed before causal conclusions can be drawn.

Limitations

The most notable limitation of our study was the small sample size. Because only three brains were studied in each cohort, it is possible that with more canines we would have found significant differences in the two DTI metrics between MPS affected and unaffected brains. In addition, the possibility exists for sampling error when placing ROIs, particularly for the diseased specimens in which the various brain regions proved harder to delineate due to white matter damage. It was sometimes impossible to avoid sampling some of the numerous striations and enlarged Virchow–Robin spaces in the MPS brains, which could have influenced our FA and RD measurements. Finally, we could not directly compare DTI metrics between pediatric and mature canines in this study because we had no older specimens at the time. Future work on mature canines is needed to follow the changes in DTI metrics longitudinally.

Funding

This work was supported by Patricia Dickson’s following grant: National Institutes of Health grant R01 NS085381, National Institute for Neurological Disorders and Stroke. James M Provenzale and Steven Chen are paid consultants on this grant.

Conflict of interest

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.