Measuring Myelin Repair and Axonal Loss with Diffusion Tensor Imaging

Discussion

We used longitudinal evaluation of patients with MS starting anti-inflammatory therapy to evaluate the ability of DTI to characterize changes within areas of focal inflammatory injury. Within gadolinium-enhancing lesions, we observed increased FA and decreased λ_⊥ during the first 2 months. Remyelination is well-known to occur following acute demyelination injury,^15–17 and the reduction in λ_⊥ may represent an imaging correlate of remyelination. Experimental animal studies have suggested that λ_⊥ is a somewhat specific measure of myelin integrity. Demyelination without axonal injury is induced in cuprizone-treated mice, and changes in λ_⊥ paralleled pathologically confirmed demyelination and remyelination.¹⁸ DTI studies in shiverer mice (who have congenital dysmyelination with intact axons) found increased λ_⊥ but normal λ_∥.⁹ In the EAE mouse model, focal regions of myelin loss are associated with decreased λ_⊥.⁴ Animal ischemia studies also found increased λ_⊥ but no change λ_∥, and histopathology showed reduced myelination.¹⁹ DTI studies from a mouse traumatic brain injury model found reduced FA and λ_∥, which corresponded to increased staining for amyloid precursor protein and neurofilament, both indicators of axonal injury.²⁰ Altogether, these animal imaging studies suggest that the changes in λ_⊥ we observed within gadolinium-enhancing lesions possibly represent remyelination.

An alternative pathologic explanation for the longitudinal increased FA within gadolinium-enhancing lesions is resolution of edema and inflammatory infiltrates. However, DTI studies from other cerebral edema models suggest that there is little contribution of resolving edema to our DTI observations. Cerebral edema from hepatic encephalopathy is associated with increased MD and variable changes in FA.^21,22 After lactulose therapy for hepatic encephalopathy, MD decreased while FA did not change.²² Peritumoral edema shows both a decrease in FA and an increase in apparent diffusion coefficient (which is comparable with MD).^23,24 Individual eigenvalue analysis of peritumoral edema showed increase in both λ_⊥ and λ_∥. In contrast, we observed no change in either MD or λ_∥. Furthermore, we observed a change in λ_⊥ alone, with no change in λ_∥. Although the resolution of gadolinium-enhancing lesions likely involves some resolving edema, comparison of our DTI observations with those from hepatic encephalopathy and cerebral tumors suggest that this represents only a small proportion of tissue recovery.

Chronic T1 hypointensities (T1 black holes) are associated with severe tissue damage, as measured by axonal loss, axonal swelling, and demyelination.^25–27 To date, no published studies have evaluated DTI predictors of conversion from gadolinium-enhancing lesions to T1 black holes. In our study, approximately half of the gadolinium-enhancing lesions developed into T1 black holes by 12 months. λ_⊥ within gadolinium-enhancing lesions at baseline predicted evolution into T1 black holes, and a trend was observed with FA. In contrast, we observed no difference in change in DTI measures during 1 year of follow-up between gadolinium-enhancing lesions that became T1 black holes and those that did not. These observations suggest that evolution of gadolinium-enhancing lesions into T1 black holes is influenced predominantly by the degree of initial injury and not the amount of later recovery. These observations also add to the accumulating evidence that DTI is a quantitative measure of the degree of brain tissue injury.

We also applied DTI to study changes in NABT with time during MS therapy. This longitudinal analysis showed gradual decline in FA during 1 year of follow-up. In contrast to enhancing lesions, this decline was driven by a decrease in λ_∥, with no significant change in λ_⊥. Animal studies of inflammatory injury found a relative specificity of λ_∥ to represent axonal injury. DTI studies of the spinal cord of mice with EAE showed decreased λ_⊥ throughout the cord, which correlated with increased staining of antibodies against the axonal-relevant β-amyloid precursor protein.⁴ Retinal ischemia causes wallerian degeneration of the optic nerve, leaving myelin intact for a period of time. Three days after induction of retinal ischemia, DTI of the optic nerve shows decreased λ_∥ with no change in λ_⊥.⁸ Subsequent pathology confirmed axonal degeneration without demyelination. A study of patients with corpus callosotomy found an early decrease in λ_∥ in the white matter adjacent to the surgery, which was followed by a later decline in λ_⊥.⁶ This imaging time course parallels the wallerian degeneration following axonotomy, with distal axonal loss first, followed later by myelin loss.

These animal studies suggest that our observed decline in λ_∥ represents an ongoing degenerative process outside of demyelinating lesions, which preferentially affects axonal integrity. The changes we observed continued despite anti-inflammatory therapy. Continued degeneration in NABT was also observed as accelerated brain atrophy in natalizumab-treated patients compared with healthy controls.²⁸ Our studies suggest that a similar process driven by axonal loss may be occurring in the NABT in patients with MS, even in the absence of active inflammation as measured by clinical relapses and gadolinium-enhancing lesions. There was no change with time in MD within NABT, which suggests that changes in FA and λ_∥ were not secondary to resolution of generalized edema.

There are few other longitudinal DTI studies in MS, and no studies have evaluated patients after starting a specific therapy. One 18-month study of untreated patients with MS found increased MD and decreased FA with time in chronic MS lesions. MD also increased in normal-appearing gray matter, but not NAWM.²⁹ Another study of mostly patients with progressive MS found that MD (but not FA) increased during 2.4 years in NAWM.³⁰ We studied patients at an earlier stage of disease and after starting anti-inflammatory therapy, so our observations more accurately characterize a stage of disease that has greater potential for tissue recovery. In addition, we were able to characterize the continued neurodegeneration in NABT, despite cessation of active inflammation. Finally, compared with our study using 71 diffusion gradients, both previous DTI studies used an 8-direction pulse sequence. It has been shown that at least 20 and 30 directions are necessary for robust determination of anisotropy and mean diffusivity, respectively.³¹

In an open-label study with no formal control group, it may be difficult for our data to differentiate the effect of anti-inflammatory therapy from the natural course of MS. Gadolinium enhancement always resolves, so the increase in FA may be secondary simply to the natural resolution of inflammation in this disease. DTI measures vary from 1 brain region to another and from 1 patient to another, making comparison studies difficult to interpret. Our longitudinal comparison of the same brain regions within the same patients minimized this problem. Furthermore, there is no pathologic data comparing treated and untreated patients with MS, making it difficult to extrapolate imaging data to pathologic changes. Future studies comparing treated and untreated patients are needed to identify the effects of MS therapy on DTI measures of tissue integrity more clearly. This study enrolled patients with MS with variable disease duration and different prior MS therapies, irrespective of other comorbid conditions. Although this process introduces variability in the dataset, it also allows greater generalization of this dataset to general clinical practice and future clinical trials. Indeed, the subjects are representative of both the general MS population and those studied in the major natalizumab clinical trials, making these observations widely applicable to the general MS patient population.

It is well-established that natalizumab therapy reduces new lesions as measured by conventional imaging,^32–34 but less known is the impact of any immunosuppressive therapy on advanced imaging measures. The rate of progressive brain atrophy is slower in patients treated with natalizumab, but only in the second year of therapy.²⁸ Natalizumab therapy was associated with a reduction in conversion of gadolinium-enhancing lesions into T1 black holes.³⁵ No studies have evaluated the potential impact of natalizumab therapy on advanced imaging measures, such as DTI. Natalizumab prevents leukocyte adhesion to the capillary endothelial surface, thereby reducing leukocyte transmigration into the central nervous system.^36,37 Natalizumab is associated with a rapid and robust reduction in the development of new gadolinium-enhancing lesions,³² which is expected to create a less inflammatory environment, thus allowing more effective repair. Furthermore, enhancing lesions in patients started on natalizumab are less likely to later convert to T1 black holes.³⁵ These observations suggest that natalizumab therapy may have a positive impact on tissue integrity, even in regions with active inflammation.

Our observation suggests that during treatment with natalizumab, there is tissue recovery in areas of active inflammation, as measured by increased FA and reduced λ_⊥. Perhaps more important, we observed continued decline in FA and λ_∥ within NABT, despite treatment with a highly effective anti-inflammatory therapy. These observations add further evidence to the pathophysiologic separation between inflammation and degeneration.^38,39 Additional studies are needed to compare our observations in natalizumab-treated patients with MS with untreated patients with MS.

Our results imply that DTI may provide pathology-specific insights into MS, because FA was driven by different eigenvectors in gadolinium enhancing lesions and NABT. If axial and radial diffusion indeed are relatively specific for axonal injury and demyelination, respectively, then DTI may be a useful biomarker of tissue recovery and later degeneration. Additionally, this study demonstrates that DTI is a dynamic measure of tissue integrity, with the ability to measure both short-term tissue recovery and long-term degeneration. These observations suggest that DTI may be a useful metric of therapeutic trials of tissue recovery and antidegeneration. These studies should target FA and individual diffusivities (λ_⊥ for tissue recovery and λ_∥ for antidegeneration) instead of MD, because these appear to be more dynamic metrics with time.