Magnetic Resonance Imaging and Late-Life Depression: Potential Biomarkers in the Era of Personalized Medicine

Magnetic resonance imaging (MRI) has revolutionized clinical neuroscience research and has led to a more sophisticated understanding of the neurobiological substrates underlying the major mental disorders. Related techniques, such as diffusion tensor imaging, morphometry, functional MRI (fMRI), and magnetization transfer, have provided important insights into the anatomical and physiological components of brain function, thereby increasing our appreciation of normal neuronal circuitry and how its aberrations can result in behavioral and cognitive phenomena and syndromes that are widely recognized as psychiatric disorders.

Complex behavioral and cognitive functions are no longer thought to reside in circumscribed brain regions but have been shown to be mediated by neural circuits that are anatomically connected yet widely dispersed in the brain (1, 2) . These multifocal, widely distributed neural networks, with gray and white matter components, are responsible for critical functions such as mood, perception, attention, motivation, and social control of behavior. Injury to focal neural networks from vascular, demyelination, genetic, or immunologic compromise predisposes to behavioral disorders characterized by impairments in mood and cognition (1, 2) . These mechanisms are interrelated and disrupt neuronal circuitry by impairing communication between brain regions that are critical components of neural circuits.

White matter provides the primary structural substrate for connectivity and circuitry in the brain (1, 2) . In the adult human brain, white matter makes up about 40%–50% of the cerebral volume. White matter is made up of myelin-invested axons, millions of which combine to form tracts that traverse the brain within and between hemispheres as well as to more caudal brainstem and cerebellar regions. White matter is an important part of the neuronal ensemble that subserves functions elemental to everyday life. Psychiatric disorders may, at least in part, represent “higher order” disconnexion syndromes; disruptions in connectivity (structural) and conductivity (physiological) may provide important substrates to these groups of brain disorders (3) .

Diffusion tensor imaging has emerged as an important in vivo tool to understand and better characterize the microstructural abnormalities of white matter tracts in the brain (4, 5) . Diffusion tensor imaging maps water proton molecular diffusion, which is a process that is sensitive to changes in the surrounding tissue microenvironment. Water molecules typically diffuse in random, thermally driven motions called Brownian movement. When unimpeded by tissue or if the barriers are not coherently oriented, the movement of water molecules is typically isotropic—diffusion is equal in all directions. When diffusion is impeded by directionally oriented tissue, it becomes nonrandom or anisotropic (4, 5) . In the presence of organized, structurally coherent barriers, such as white matter tracts in the brain, anisotropic water diffusion becomes more pronounced parallel to the bundles as opposed to the perpendicular direction. Fractional anisotropy and apparent diffusion coefficient are measures commonly used to characterize both anisotropy and water diffusion from in vivo diffusion tensor imaging studies. Anisotropic diffusion in neural fibers is a result of the density of axonal packing and axonal membranes. Preclinical evidence suggests that while the myelin sheath may attenuate the degree of anisotropy in a tract, it does not play a primary role in determining anisotropy (4) .

In this issue of the Journal, there is an interesting report from Alexopoulos et al. demonstrating that diffusion tensor imaging-determined fractional anisotropy is a marker of poor response to 12 weeks of treatment with escitalopram in a sample of patients diagnosed with late-life major depression. Forty-eight patients who met criteria for major depressive disorder entered a 12-week, open-label trial of escitalopram, 10 mg/day, after a 2-week placebo wash-in period. All patients met standard DSM-IV criteria for major depressive disorder, were nonpsychotic, and did not show any clinical evidence of dementia at the time of the study. Remission was operationally defined as a Hamilton Depression Rating Scale score of 7 or lower for 2 consecutive weeks. The investigators observed that patients who failed to achieve remission had lower fractional anisotropy values in multiple brain regions, including the rostral and dorsal anterior cingulate, dorsolateral prefrontal cortex, genu of the corpus callosum, insula, and posterior cingulate.

The finding is novel and represents an attempt to utilize modern technology to answer both clinically relevant and biologically significant questions in a vulnerable patient population. Outside the realm of degenerative disorders, neuroimaging does not have any direct diagnostic applications in clinical psychiatry at the present time. The Alexopoulos et al. study builds on earlier reports that demonstrated changes in fractional anisotropy and apparent diffusion coefficient in well-characterized clinical samples and suggested that structural brain changes may contribute to poor clinical response in the elderly (6, 7) . In the current study, clinical research is taken to the next pragmatic level by applying MRI approaches to the identification of nonresponders to pharmacological intervention, thereby raising the intriguing possibility of identifying potential nonresponders to pharmacological intervention early in the course of illness. While it is currently premature for clinical use, consistent findings over time may open the door to more aggressive or selective interventions from the outset in patients with neuroimaging determined biomarkers of poor treatment response. Alternatively, a marker of significant white matter pathology could help identify elderly patients for whom the risk of antidepressant treatment may not be balanced by a sufficient probability of a therapeutic response. As we approach the era of “personalized medicine,” neuroimaging, together with other relevant biomarkers and genomics, could potentially provide a broad-based personalized database that could help guide treatment choices among elderly individuals with mental disorders.

A valid criticism of human neuroimaging studies is that while they provide broad insights into possible etiological mechanisms, they often lack biochemical precision and neuropathological validation. Other commonly observed limitations often include the relatively small sample sizes studied and the failure to consistently replicate findings across groups. Clinical heterogeneity and methodological differences are frequently offered as explanations for discrepant findings between groups. Nonetheless, these limitations have led to understandable skepticism of many in vivo neuroimaging reports in psychiatric research.

Every neuroimaging approach has its strengths and limitations. While diffusion tensor imaging provides information on specific white matter tracts, the precise neurobiological implications of lowered fractional anisotropy remain somewhat obscure. Plausible mechanisms include abnormalities in the structural integrity of axonal membranes and possibly the myelin sheath. However, even these assertions are overly broad and do not provide specific information on the nature of the biochemical/cellular changes in membranes and the myelin sheath. Likewise, volumetric loss identified in vivo often lacks neuropathological corroboration, raising issues about the validity of in vivo observations. Despite these limitations, human neuroimaging has helped develop a body of neuroscientific knowledge that has contributed to our understanding of the pathophysiology of major mental disorders. Other MR-related approaches, such as magnetization transfer, that have been used in psychiatric research have neurobiological correlates that are better characterized and provide complementary information on the biology of gray and white matter regions (8, 9) . Similarly, fMRI, which is commonly used as a cognitive neuroscience tool, will have more scientific impact when it is integrated with some of the aforementioned imaging probes (10) . Utilizing cutting edge, validated neuroimaging tools to answer biologically and clinically meaningful questions in large, well-characterized, clinical samples will help to advance knowledge in psychiatric neuroscience and serve to translate neuroscience information from the laboratory to the bedside relatively rapidly.