Traumatic brain injury, major depression, and diffusion tensor imaging: Making connections

Abstract

It is common for depression to develop after traumatic brain injury (TBI), yet despite poorer recovery, there is a lack in our understanding of whether post-TBI brain changes involved in depression are akin to those in people with depression without TBI. Modern neuroimaging has helped recognize degrees of diffuse axonal injury (DAI) as being related to extent of TBI, but its ability to predict long-term functioning is limited and has not been considered in the context of post-TBI depression. A more recent brain imaging technique (diffusion tensor imaging; DTI) can measure the integrity of white matter by measuring the directionality or anisotropy of water molecule diffusion along the axons of nerve fibers. Aim: To review DTI results in the TBI and depression literatures to determine whether this can elucidate the etiology of the development of depression after TBI. Method: We reviewed the TBI/DTI (40 articles) and depression/DTI literatures (17 articles). No articles were found that used DTI to investigate depression post-TBI, although there were some common brain regions identified between the TBI/DTI and depression/DTI studies, including frontotemporal, corpus callosum, and structures contained within the basal ganglia. Specifically, the internal capsule was commonly reported to have significantly reduced fractional anisotropy, which agrees with deep brain stimulation studies. Conclusion: It is suggested that measuring the degree of DAI by utilizing DTI in those with or without depression post-TBI, will greatly enhance prediction of functional outcome.

Introduction

The famous case of Phineas Gage in 1868 was the first to clearly illustrate that behavior and emotion can change after traumatic brain injury (TBI) (Harlow, 1868). Major depressive disorder (MDD) is common in the subacute aftermath of mild TBI and exerts a deleterious effect on psychosocial function and certain indices of cognition (Fann, 1997, Feinstein, 2006, Jorge et al., 2004). Posttraumatic behavioral disorders (including hyperactivity, agitation, mood swings, irritability, excitation, and lack of inhibition, hostility, and distrust) constitute major problems with respect to the management, rehabilitation, and social and family reintegration of these patients (Guetin et al., 2009). Although there is a decreased probability of having an Axis I diagnosis many years after a TBI (Ashman et al., 2004), it is not uncommon for a state of disability and/or hardship involving the entire familial nucleus to persist (Inzaghi et al., 2005).

Depression and diminished life satisfaction among survivors of traumatic brain injury (TBI) are persistent problems that require the close attention of medical and rehabilitation professionals (2003). In Victoria, Australia, nearly half of the approximately 73,000 people who have experienced TBI need personal assistance or supervision (Snaith and Zigmond, 1994). In the United States, direct and indirect costs of TBI exceed $80 billion annually (NDC, 2003). The reported frequency of major depression post-TBI is variable (Whelan-Goodinson et al., 2009), but may be present in up to 77% of patients (Alfano, 2006, Ashman et al., 2004, Deb et al., 1999, Fann et al., 1995, Fedoroff et al., 1992, Hibbard et al., 1998, Holsinger et al., 2002, Jorge et al., 1993a, Jorge et al., 1993b, Jorge et al., 2004, Jorge and Starkstein, 2005, Koponen et al., 2002, Kraus and Sorenson, 1994, Pagulayan et al., 2008, van Reekum et al., 2000) and, along with related anxiety disorders, is associated with poorer recovery (Gordon et al., 2006, Jorge and Starkstein, 2005).

Contributing factors to these complications in recovery is that conceptualization and potential treatment of post-TBI depression is not well understood. The evidence for a benefit of ‘standard’ antidepressant strategies in post-TBI patients is underwhelming (Gordon et al., 2006), suggesting that these depressive syndromes are not simply the result of a ‘typical’ depressive response to a significant life stressor. Of reported TBI cases, approximately 85% are classified as mild TBI (Bazarian et al., 2005). Hence, mild TBI is one of the most common causes for admission to trauma centers (Ucar et al., 2006). Even in “mild” cases where there is generally no fracture of the skull, associated loss of consciousness, or amnesia, TBI often leads to a variety of physical, emotional, and cognitive difficulties (Duncan et al., 2005, Kraus et al., 2005). Up to 30% of mild TBI patients will suffer permanent sequelae of their injury and up to 20% will be unable to return to work (Nolin and Heroux, 2006). Suicide rates are also higher post-TBI than in the general population (Fleminger et al., 2003). Across the broader TBI population, there does not appear to be a significant relationship between Glasgow Coma Scale (GCS) scores, duration of post-traumatic amnesia (PTA), ongoing cognitive functioning, and the degree of post-TBI depression. Post-TBI depression is therefore not just related to injury severity. These observations suggest a strong likelihood of a specific pattern in brain changes that may result from TBI leading to substantially higher risk of subsequent development of depression.

Just over a decade has now passed since the 1998 National Institutes of Health (NIH) consensus conference, which identified a lack of scientifically rigorous TBI research and areas of important TBI research that remained largely unexplored. For example, while major depression is the most common and hence the most studied psychiatric disorder after TBI (Koponen et al., 2002, Pagulayan et al., 2008), it has not been often approached from a neuroimaging perspective (Koponen et al., 2006). Unfortunately, there is a lack of quality studies examining psychiatric consequences of TBI and potential related brain changes (Gordon et al., 2006). While there is a consensus that underlying the brain changes is a compromise of neural tissue referred to as diffuse axonal injury (DAI), this can be subtle and difficult to detect. The aim of this paper is to review the post-TBI and depression literatures in the context of MRI, specifically focusing on studies that have applied an advanced neuroimaging acquisition sequence (diffusion tensor imaging) which can evaluate white matter (WM) integrity as an indicator of DAI by measuring water movement and restriction (“anisotropy”). By studying anisotropy maps in detail, neural models and networks can be elucidated so that links between depression and post-TBI brain changes may be evaluated.

Traumatic brain injury commonly results from a simple blow to the head, or in biomechanical terms, a rapid ‘linear deceleration.’ In such a scenario, the brain strikes the inside surface of the skull at the point of impact. Conventional wisdom proposes that the brain then subsequently rebounds to collide with the opposite side of the skull, producing what is called the ‘contre-coup’ injury (Ommaya and Gennarelli, 1974). Some authors have disputed this theory, however, suggesting that cavitation effects due to rapid deceleration (Courville, 1942) or the relative movement of brain tissue against skull bone (Shatsky et al., 1974) are responsible for this secondary injury to the brain tissue. Nonetheless, a simple linear head-strike, although common, represents the minority of primary brain injuries, and it is rare that an impact is purely linear or for that matter rotational (King et al., 2003). When an injury involves rapid rotation of the head, the tissues of the brain and brainstem are placed under significant shear stress. In fact, in the early 1940s, it was proposed that angular rather than linear acceleration of the head would result in more serious brain injuries (Holbourne, 1943). Using the ‘bowl of porridge’ analogy, it is easy to visualize how rotational shearing might result in severe tissue damage.

Bayly et al. (2005) provided a comprehensive description of the temporal order of mechanical forces during TBI and reported clear deformation of brain matter due to occipital deceleration in humans during mild, rapid deceleration of the head, with notable compression in frontal regions and stretching in posterior regions. It is this more diffuse, microstructural damage resulting from physical deformation of brain tissue that constitutes the pathophysiological entity known as DAI.

In animal models, it has been shown that mild neurotrauma that neither breaks the dura mater nor produces a widespread hematoma is expected to produce little or no neuropathology in the hippocampus and other brain regions (Baldwin et al., 1996), when compared with the damage associated with severe TBI (Chen et al., 2003). Variation in the involvement of cortex and in the degree of hippocampal damage reported for fluid percussion models is likely to be related to severity and lateralization of injury (Floyd et al., 2002, Grady et al., 2003, Obenaus et al., 2007, Thompson et al., 2005).

In the acute phase, imaging findings can often provide insights into the mechanism of injury. Purely focal traumatic lesions such as extradural and subdural hematomas not only pinpoint the site of injury but also suggest that the injury may have arisen due to more linear acceleration without significant rotational component (King et al., 2003). DAI is typically associated with more subtle changes that simple computerized tomography (CT) scans (the first-line imaging modality in assessing TBI; Lee et al., 2008) are frequently unable to detect (Jenkins et al., 1986, Mittl et al., 1994). MRI will often demonstrate lesions at the grey matter (GM) white matter (WM) junction, with additional hemorrhagic lesions occurring in the corpus callosum (CC), rostral brainstem, and basal ganglia as the severity of injury increases (Alsop et al., 1996, Arfanakis et al., 2002, Assaf et al., 1999, Barzo et al., 1997, Han et al., 2007, Hanstock et al., 1994, Ito et al., 1996, Marmarou et al., 2000, Miles et al., 2008, Sidaros et al., 2008, Takayama et al., 2000, Teasdale and Hadley, 2005, Vorisek et al., 2002, Xu et al., 2007, Yasokawa et al., 2007). The CC, which is the largest collection of commissural fibers in the brain, is especially vulnerable to TBI because of its unique location (Gorrie et al., 2001). Not all types of the brain cells are equally vulnerable; the most susceptible cellular components are axons and the most resistant structures are blood vessels (Besenski, 2002). A comprehensive review of the neurophysiology of brain injury is presented in Gaetz (2004), although an understanding of many aspects of the mechanisms of TBI is still lacking (Ragnarsson, 2006).

Neuroimaging-based volumetric studies, the number of which has grown exponentially since the 1980s (Gordon et al., 2006), have provided many useful insights into alterations in brain functioning post TBI. Both human and animal TBI studies have shown reduced brain volumes, either as total or its primary constituents (GM and WM) or as specifically vulnerable regions of interest (ROIs). Frontal, temporal (hippocampus, amygdala, and parahippocampal gyrus), and occipital regions (e.g., Anderson and Bigler, 1994, Ariza et al., 2006, Bayly et al., 2005, Bendlin et al., 2008, Bigler et al., 1997, Ducreux et al., 2005, Gale et al., 2005, Hicks et al., 1993, Kotapka et al., 1991, Levine, 2006, Maxwell et al., 2003, Tate and Bigler, 2000, Tomaiuolo et al., 2004, Wilde et al., 2005). For example, reduced right hippocampal volume after childhood TBI is a common finding (Di Stefano et al., 2000, Tasker et al., 2005, Tomaiuolo et al., 2004). An altered CC shape or reduced cross-sectional area is also a common finding among TBI survivors (e.g., Beauchamp et al., 2008, Strich, 1970, Tomaiuolo et al., 2004). From a neuropsychological standpoint, reductions in volumes have been related to reduced cognitive performance, particularly attention and memory (e.g., Bigler et al., 1997, Bigler, 2001, Ewing-Cobbs et al., 2006, Gale et al., 1995, Gale et al., 2005, Hopkins et al., 2005, Wilde et al., 2006). Serra-Grabulosa et al. (2005) reported that severe TBI in childhood resulted in hippocampal atrophy and WM loss even 10 years after the injury and that hippocampal damage may contribute towards memory impairment post-TBI. The authors suggest that due to the limbic circuitry involving hippocampal output, a number of WM pathways may be damaged in TBI and may be responsible for TBI memory sequelae. Sugiyama et al. (2007) found in two survivors of chronic TBI that the interruption of the fornix (which involves the circuit of Papez) potentially correlates with the memory disorder. In addition to the diffuse WM volume loss, there is probably WM loss in the fornix and mammillothalamic track, as well as the anterior thalamus projections to cortex and cingulate gyrus, which may be just as disruptive to memory as a specific hippocampal volume loss (Gale et al., 1995, Tate and Bigler, 2000). The study by Avants et al. (2008) supports this suggestion.

Despite these and other findings, the relationship between volume reduction and injury severity is equally unclear. The heterogeneity of the underlying injuries, radiological findings and postinjury clinical course across the TBI population is likely to be the culprit, as it is in many other negative studies involving the general TBI population (Belanger et al., 2007). Furthermore, it is difficult to determine prognosis until 12 to 24 months or more have passed following the injury (Lew et al., 2006). Hence, a more thorough investigation of the underlying pathophysiology resulting from the TBI is warranted.

Under the microscope, DAI is characterized by disruption of the cytoskeletal network and axonal membranes in the first few hours after TBI (Arfanakis et al., 2002, Benson et al., 2007, Liu et al., 1999), which ultimately results in axonal disconnection. This progressive functional and structural deterioration is thought to be brought about by a combination of altered membrane permeability with concomitant calcium ion influx, proteolysis, mitochondrial dysfunction, and, subsequently, Caspase-mediated apoptosis. Subsequent to axonal disconnection, the formation of axonal ‘retraction balls’ is observed, which is thought to be the result of the accumulation of axoplasm at the site of axonal disconnection (Bullock and Nathoo, 2005).

Adams (1982) was among the first to report DAI in head injury from a theoretical perspective, and then perform an analysis of a number (N = 45) of cases (Adams et al., 1982). A report of 78 post-TBI patients was published a few years later (Cordobes et al., 1986), which focused on assessing DAI from CT; all patients had detectable DAI, but each had sustained a very severe TBI. When patients with less severe grades of TBI (i.e., mild or moderate) are assessed with CT, DAI is less detectable. The detection of lesions suspicious of DAI is important as their presence in the acute stage of TBI is related to outcome (Paterakis et al., 2000, Schaefer et al., 2004), and associated cognitive sequelae have been described for the chronic stage (Scheid et al., 2007, Wallesch et al., 2001).

Gentry et al., 1988, Kelly et al., 1988 were the first to report on the comparative evaluation of closed head trauma via both MRI and computerized tomography (CT); they found that MRI was more useful than CT for detecting nonhemorrhagic lesions. Many similar studies have since been published with comparable findings, e.g., Mittl et al. (1994) demonstrated that MRI shows evidence of DAI in some patients with normal head CT findings after mild head injury.

While these, and many other investigations (e.g., Anderson et al., 1996, Gale et al., 1995), consistently showed that standard structural MRI sequences, usually T1- or T2-weighted, are more sensitive to detecting DAI than CT, the pathophysiology underlying DAI resulting from TBI remained elusive regardless of injury severity. T2 relaxation times in ROIs correlate precisely with tissue water content and can provide an estimate of vasogenic edema following brain injury (Kato et al., 1986). In the neuropathology literature, DAI is typically accompanied by small hemorrhages or so-called tissue tear hemorrhages (Scheid et al., 2007). Most blood products, such as deoxyhemoglobin, methemoglobin, and hemosiderin, are often undetected on conventional MR images alone. Iron deposits can be detected with T2*-weighted sequences, and the gradient recalled echo (GRE) technique is the most commonly used (Topal et al., 2008). However, even T2*-weighted sequences (for emphasizing hemosiderin deposits) do not detect DAI in every patient (Scheid et al., 2007). Images acquired at 3 T are superior to those acquired 1.5 T in revealing DAI in the chronic phase of TBI (Scheid et al., 2007) since T2* signal intensity loss, for example, depends on the magnetic field strength (Atlas et al., 1988). However, standard T1- and T2- and T2*-weighted sequences are simply not sensitive enough to detect DAI in every patient regardless of field strength (Lee et al., 2008). Consequently, conventional MRI underestimates the extent of DAI, possibly accounting for much of the transient clinical pathologic conditions following TBI and perhaps explaining the residual neurologic and cognitive deficits that are associated with TBI (Arfanakis et al., 2002, Sugiyama et al., 2007).

Note, however, that increased susceptibility at 3 T may not only be advantageous. Firstly, due to artifacts caused by air from sinuses and mastoid bone, the 3-T gradient echo images are not suitable for the evaluation of frontobasal and temporobasal/temporopolar brain structures (Scheid et al., 2007). Secondly, the increased sensitivity of 3 T MRI also to potential nontraumatic microbleeds, possibly resulting in differential diagnostic uncertainties (Fiehler, 2006).

A sequence based on Brownian movement (diffusion) of water through WM tracts was developed in 1991 (Le Bihan et al., 2001), aptly named diffusion-weighted imaging (DWI), and permitted the study of change in the random motion of proton in water in vivo (Liu et al., 1999). Restricted proton motion results in a decrease of the apparent diffusion coefficient (ADC) of water and increased signal intensity in diffusion-weighted images (Shanmuganathan et al., 2004). In a novel study by Nakahara et al. (2001), ADC maps generated from DWI data were used in the evaluation of four deeply comatose patients with severe TBI; ADC values in GM and WM were significantly different in the one patient with fatal outcome shortly after MRI examination. In another DWI study, Shanmuganathan et al. (2004) found whole-brain ADCs to correlate significantly with GCS scores in TBI patients. While DWI can detect DAI in some patients (e.g., Galloway et al., 2008, Hergan et al., 2002, Liu et al., 1999) where CT or standard MRI sequences do not (e.g., Arfanakis et al., 2002, Assaf et al., 1999, Blamire et al., 2002, Chan et al., 2003, Huisman et al., 2003, Huisman et al., 2004, Inglese et al., 2005, Jones et al., 2000, Nakayama et al., 2006, Ptak et al., 2003, Rugg-Gunn et al., 2001, Salmond et al., 2006b, Shanmuganathan et al., 2004, Suh et al., 2001, Topal et al., 2008), it cannot detect DAI in all instances: this was a strong motivating force behind the development of diffusion tensor imaging (DTI). Unlike conventional DWI (Le Bihan et al., 2001), where diffusion-weighted images are used to calculate the scalar ADC, DTI characterizes diffusive transport of water by an effective diffusion tensor D. The eigenvalues of D are the three principal diffusivities, and the eigenvectors define the local fiber tract direction field (Basser et al., 1994) (Fig. 1). Moreover, one can derive from D rotationally invariant scalar quantities that describe the intrinsic diffusion properties of the tissue. The most commonly used (Arfanakis et al., 2002) are the trace of the tensor, which measures mean diffusivity, and fractional anisotropy (FA) and lattice index, which characterize the anisotropy of the fiber structure, i.e., how much higher the diffusivity is along some directions compared with others. FA of water molecules ranges from 0 (anisotropic, i.e., random motion) to 1 (isotropic, i.e., nonrandom motion). In the cerebral WM, FA values should be high. A decrease in FA may indicate that a certain type of brain damage is present.

It was elucidated a number of years ago that regions of hyperintense signal (hyperintensities) are associated with chronicity of mild-adult and geriatric/late-onset depression (e.g., Hickie et al., 1995, Jorm et al., 2005, Nobuhara et al., 2006, Sheline et al., 2008, Yamashita et al., 2001), suggesting abnormalities in the WM, and it has been found that WM hyperintensities also contribute to poor treatment outcome. Strictly speaking, late-life depression refers to depressive syndromes defined in the International Classification of Diseases (ICD-10; World Health Organization, 1992), which arise in adults aged over 65 years. The most commonly reported sites of WM hyperintensities in the context of late-life depression are in the frontal gyri that contain fibers of the anterior cingulate and the dorsolateral pathways (Middleton and Strick, 2001). Two medial temporal structures of the major efferent prefrontal WM tracts, the amygdala and hippocampus, are also implicated in major depressive disorder (MDD) in a plethora of volumetric and shape analysis studies (e.g., Bergouignan et al., 2009, Bouix et al., 2005, Egger et al., 2008, Frodl et al., 2002, Kronmuller et al., 2008, Mervaala et al., 2000, Posener et al., 2003, Tamburo et al., 2008, Zhao et al., 2008). Two meta-analyses (Campbell et al., 2004, McKinnon et al., 2009) found reduced amygdala and hippocampal volumes to be related to MDD. Specifically, the findings suggest that a relatively small hippocampal volume may be a vulnerability factor for a poor treatment response. For example, MacQueen et al. (2008) found that MDD patients who remitted had larger pretreatment hippocampal body/tail volumes bilaterally compared with those who were not in remission at 8 weeks. Furthermore, Maller et al. (2007) found the hippocampi in patients with MDD to be significantly smaller compared to age- and sex-matched controls, but particularly in the posterior tail. This suggests that a relatively small dorsal hippocampus may be a marker of MDD. In addition, the patients had a treatment-resistant form of MDD, which suggests not only that MDD may be related to reduced hippocampal volume but also that the tail segment is most reduced in those who are treatment-resistant. Studies have elucidated that this part of the hippocampus is rich in 5-HT (serotonin) receptors (Adams et al., 2008, Joca et al., 2003), which are primary targets for a number of antidepressant therapies (Drevets et al., 2000, Lambas-Senas et al., 2009). As the majority of patients with depression post-TBI do not respond to typical antidepressant treatment, the study raises the question of whether the post-TBI depression is developed as a result of the same activity occurring in the brains of patients with treatment-resistant MDD without a history of TBI. As both the hippocampus and prefrontal region are often implicated following TBI, it is possible that the symptoms of depression are reflective of an underlying pathophysiological change occurring at, and between, these two regions. This raises a number of possible models for treatment-resistant MDD and for the development of MDD post-TBI. For example, treatment-resistant MDD may be a result of the same underlying problems as in the brains of patients with MDD post-TBI; that is, prefrontal and hippocampal dysfunction in addition to a dysfunctional connection between them. By contrast, treatment-responsive MDD may be a consequence of a dysfunctional prefrontal and/or hippocampal regions, but the connection between them is not dysfunctional. It is of prime importance to acknowledge that neural circuits and connectivity are critical to our understanding of cognitive and emotional information processing; hence, WM provides the framework and anatomical basis of neural connectivity and circuitry in the central nervous system (Kumar and Cook, 2002).

In addition to frontotemporal studies in MDD patients, there is a growing body of literature suggesting that a smaller volume of the caudate nucleus may be related to the pathophysiology of MDD and may account for abnormalities of the corticostriatal–pallidothalamic loop in MDD. For example, Butters et al. (2009) reported greater caudate volume reduction, particularly in the anterior region, to be associated with more severe late-life depression. Kim et al. (2008) reported reduced caudate GM in women with MDD, no group differences in WM, and no significant correlations between GM volumes and symptom severity within the MDD group.

There are also a number of reports of altered CC areas in those with MDD, but results are not consistent. For example, while Husain et al. (1991) found no statistically significant group differences in measurements of CC and septum pellucidum, MDD patients had a larger CC area compared to healthy controls. Wu et al. (1993) reported that the anterior and posterior quarters of the CC were significantly larger in MDD patients, and similarly, Lacerda et al. (2005) reported that patients with familial MDD had a significantly larger middle genu area compared to healthy controls, and significantly larger middle genu, anterior splenium, and middle splenium areas compared to patients with nonfamilial MDD. This difference was more pronounced in females. Walterfang et al. (2009) reported expansions in the thickness of the CC's posterior body and isthmus in MDD patients when compared to controls; this was not seen in remitted patients. That is, the CC was expanded in regions connecting frontal, temporal, and parietal regions in currently depressed patients only, suggestive of state-related changes in WM in MDD that may reflect the effects of state-related factors on WM structure. Parashos et al. (1998), examining MDD patients and controls, found smaller callosal area in patients, not reaching statistical significance. The inconsistencies in results across these studies could be explained at least in part by distinct factors, such as (1) differences in patient sample characteristics (age, gender, handedness, length of illness, age at onset, and severity), (2) methodological differences in MRI acquisition as well as in callosal measurements, and (3) insufficient statistical power of studies presenting negative findings (Lacerda et al., 2005).