Neurodevelopment Versus Neurodegeneration
Current models of pathological change in affective illness are deeply rooted to these environmental contingencies, as functional and anatomical change in depression is often viewed as degenerative: stress-induced, glutamate-mediated excitotoxicity leads to metabolic changes or decreases in GM volume. In contrast, the role of genetically-mediated neuroplasticity is not usually explicitly addressed in contemporary models of illness-associated neuropathology. We will first discuss the neurodegenerative hypothesis, which conceptualises GM or WM volume loss as a downstream effect of an unspecified environmental pathogen, such as psychological stress.
Actually, there may be a familial or genetic component to the experience of life stress because of a phenomenon known as gene-environment correlation (Bergeman et al 1988
); (Rutter 2007
). (Rowe 1981
) first reported this counter-intuitive notion by showing that adolescent twins’ reports of their parents’ levels of accepting and rejecting behaviour were under genetic influence; a finding that has been extended to retrospective measures of family warmth and parental control (Plomin et al 1988
) as well as family cohesion and encouragement of growth (Bouchard et al 1990
There are at least three ways in which gene-environment correlation might apply in the case of the degenerative hypothesis. The parental genotype or affective disorder may exert an effect on parental behaviour such that their children are reared in a high-stress environment. Here there is a correlation between passing on “stress-provoking” genes and providing a stressful family environment.
Secondly, it can be argued that people are selecting and shaping their environmental experiences on the basis of their genetic heritage, leading to preferential exposure to significantly stressful events and depression-associated neuroplastic changes in a subset of the population. This phenomenon may be related to the way in which individuals perceive or process information in their environment; an intrinsic bias often described as temperament.
Yet a third possibility is that genetic effects play no role in influencing exposure to stressors but moderate the physiological effect of these events on neural tissue. One class of proteins potentially involved in this type of gene-environment interaction is the neurotrophins.
As reviewed by (Poo 2001
), one of these enzymes, brain-derived neurotrophic factor (BDNF
) , increases forebrain serotonin fibre density and neurogenesis, prevents spontaneous and neurotoxin-induced cell death, and modulates the formation of synaptic connections, particularly in the PFC and hippocampus.
Recent studies have suggested that the low expression (met
) allele of a functional single nucleotide polymorphism (SNP) (Val66Met) of the BDNF
gene may increase the probability of developing depression (Kaufman et al 2006
) and cognitive impairment (Savitz et al 2007c
) after exposure to childhood maltreatment. Perhaps through its reduced ability to protect against neurotoxicity, the met
allele has also been reported to increase the risk of developing depression after stroke (Kim et al 2007
Another potential moderator of the stress-response is central serotonergic activity. As reviewed by (Drevets 2000b
), the binding of serotonin to post-synaptic 5-HT1A
receptors not only enhances the negative feedback inhibition of cortisol release but also prevents dendritic cytoskeletal breakdown by catalyzing the release of the neurotrophic factor, S100β, and indirectly inhibiting protein kinase-induced apoptosis (Szatmari et al 2007
). The regulation of 5-HT1A
receptors in the raphe is at least partly controlled by functional variants of the HTR1A
 (−1019C/G; rs6295) (Lemonde et al 2003
) (Parsey et al 2006
) and the SLC6A4
 (promoter region length polymorphism) (David et al 2005
) genes, respectively.
In contradistinction to the degenerative model, the developmental model advocates that neuroanatomical changes precede the onset of affective illness. An interesting set of animal experiments has lent credence to this hypothesis: A line of rats, genetically-bred to suffer from learned helplessness display baseline hypometabolism of the amygdalae, BG, VTA, dorsal-frontal, medial-OFC and ACC, but increased metabolism of the infraradiata (sgACC), hippocampi and habenula (Shumake et al 2000
); (Shumake et al 2002
). To further control for the effects of early-life stress, (Shumake et al 2004
) examined the brains of this genetic line of rats at birth and again found hypo or hypermetabolism of most of these regions. Moreover, the midbrain and brain-stem regions were found to be disconnected from limbic forebrain regulation, suggesting to the authors that the fundamental disturbance in depression is one of top-down regulatory control (Shumake et al 2004
In humans one way of examining this issue is to compare the degree of variation in regions of interest across the life-span. (Lupien et al 2007
) found that there was just as much variability in the hippocampal volumes of healthy young adults as in older individuals, implying that volume decrements attributed to aging or stress could be reflective of neurodevelopmental differences. Specifically, a quarter of their subjects in the 18–24-year age group had hippocampal volumes as small as the average hippocampal size in their 60–75-year-old sample, and the mean difference in hippocampal volumes between the upper and lower quartiles of the young age group (12–16%) was greater than volumetric reductions typically seen in depressed samples (Lupien et al 2007
). These data are congruent with an earlier study (Gilbertson et al 2002
) which reported an association between post-traumatic stress disorder (PTSD) and smaller hippocampal volume in war veterans. Intriguingly, the MZ twin brothers of the PTSD cohort who did not serve in the military also presented with smaller hippocampi than the PTSD group, raising the possibility that reduced hippocampal volumes are a contributing cause rather than an effect of PTSD.
Recent findings from the emerging field of imaging genomics also emphasize the importance of genetic influences. A variant of the neuregulin 1 (NRG1)
 gene, which is involved in the myelination process, and has been implicated in both BD and schizophrenia, may be associated with WM density and integrity of the internal capsule (McIntosh et al 2007
). In another study, the short (s)
allele of the serotonin transporter (5-HTT) gene (SLC6A4
)  promoter polymorphism has been shown to be associated with increased resting CBF in the amygdala and decreased perfusion of the ventromedial PFC in healthy individuals (Rao et al 2007
The neurodevelopmental hypothesis can also be evaluated by searching for neural changes in unaffected family members who presumably share a genetic diathesis for the disorder with their ill relatives. (Noga et al 2001
) compared MZ pairs discordant for BD with a control group of unaffected twins, and found that the right hippocampus was smaller in the affected twins, but that both ill and well twins had larger caudate nuclei than the control pairs. (McDonald et al 2004a
) reported that genetic risk for BD was associated with reduced volume of the right anterior cingulate gyrus and ventral striatum. Similarly, a compensatory hypermetabolic response to a sadness induction paradigm was observed in the medial PFC of healthy BD relatives compared with background controls (Kruger et al 2006
). Conversely, (McIntosh et al 2006
) and (Munn et al 2007
) failed to detect any significant neural changes in at-risk BD relatives.
Concerning MDD, (Monk et al 2007
) reported that the pediatric offspring of parents with MDD displayed greater amygdala and accumbens area activity in response to fearful faces, and lower accumbens area activation in response to happy faces than a low-risk control group. A small number of serotonin depletion experiments have succeeded in inducing depressive symptoms in otherwise healthy relatives of MDD probands. Six out of 20 healthy males with a family-history of affective illness but 0 out of 19 male controls displayed a lowering of mood in response to tryptophan depletion (Benkelfat et al 1994
). More recently, (van der Veen et al 2007
) reported a greater lowering of mood, and stronger amygdala response to fearful faces under tryptophan depletion in healthy individuals with a family history of depression compared with controls. (Neumeister et al 2002
) showed that this effect may be mediated by the SLC6A4
promoter length polymorphism: family-history-positive heterozygotes showed a greater mood-lowering effect than their heterozygote counterparts without a history of depression.
These potential neuroimaging endophenotypes are most likely genetically-driven. For example, serotonergic activity not only plays an important role in modulating the impact of stressful events, but is a key regulator of neural development influencing neurogenesis, apoptosis and dendritic growth (Gaspar et al 2003
). Fluoxetine-induced suppression of the serotonin transporter during development has been shown to result in abnormalities of emotional behaviour in mice (Ansorge et al 2004
). Recent data have also shown the direct effect of the 5-HTT polymorphism on neural tissue with increased hemodynamic response of the amygdala (Hariri et al 2002
); (Heinz et al 2005
); (Pezawas et al 2005
); (Dannlowski et al 2006
) to fearful faces (versus geometric images or neutral faces) in s
The distinction between early, developmental, and later-onset stress-induced pathology may be partly artificial. Genetically-determined subtleties of brain-wiring may sensitise the individual to the effect of ubiquitous, relatively mild stressors.
allele of the above-mentioned 5-HTT gene insertion/deletion polymorphism, first associated with anxiety-related personality traits more than a decade ago (Lesch et al 1996
), is now one of the prototypical examples of a risk variant that interacts with adverse life experiences to predispose to psychopathology (Caspi et al 2003
); (Kaufman et al 2004
(Pezawas et al 2005
) investigated the neurobiology behind this process. The authors demonstrated that healthy carriers of the s
allele of the polymorphism display reduced functional connectivity between the amygdala and the sgACC. The latter structure exerts an inhibitory effect on the amygdala3
and thus a genetically-determined attenuation of this negative-feedback loop may increase sensitivity to environmental adversity and by implication, lead to maladaptive neuroplastic changes (Pezawas et al 2005
A nascent trend in the study of psychiatric disorders is a recognition of the potentially important role of epigenetic mechanisms in disease causation (see (Mill and Petronis 2007
)). Epigenetic inheritance refers to a regulated pattern of gene expression which is transmitted intact from one or other of the parents to their offspring. The process is mediated by the methylation and histone acetylation of cytosine residues and chromatin, respectively, leading to the activation or silencing of particular genes. The phenomenon is epigenetic because it results in phenotypic traits that are inherited independently of the informational content of DNA.
(Meaney and Szyf 2005
) review rodent studies which demonstrate that stress sensitivity in rat pups is modulated by parental grooming behavior that exerts its effect through a histone modification-driven regulation of glucocorticoid receptor gene expression. If these biological mechanisms generalize to humans then exposure to adversity may modify gene expression in pathways that impact neuroplasticity.
Methodological and Theoretical Issues
(1). A common confounding variable is medical treatment. As (Drevets 1998
) notes, rCBF and metabolism of various neuroanatomical regions may be reduced by antidepressants, anti-psychotics and anxiolytics. Further, psychotropic drugs have been shown to alter both the behavioral and the neurophysiological response to the environmentally valenced stimuli used as neurocognitive probes in fMRI studies.
Recent studies have detailed the neurotrophic effects of lithium and mood stabilizers on hippocampal and other neural tissue. In fact, there has been some suggestion that AD may also exert a neurotrophic effect in particular regions of the brain (Stewart and Reid 2000
) (Rocher et al 2004
) (Duman and Monteggia 2006
). These data are supported by an fMRI study which showed that depressed patients suffer from reduced functional coupling of the amygdala with diverse brain regions including the hippocampus, caudate, putamen, and ACC; an effect reversed by treatment with fluoxetine (Chen et al 2007b
). Similarly, a decrease in left amygdala activity to normal has been observed after antidepressant treatment (Drevets et al 2002a
). Interestingly, a decrement in amygdala activity in response to aversive stimuli also been observed in healthy individuals treated with citalopram (Harmer et al 2006
) and reboxetine (Norbury et al 2007
The association between imaging changes and medication is rendered even more complicated by the possibility that treatment may be a proxy variable for genetic etiology. In other words, the hippocampal or sgACC volumetric changes associated with a drug like lithium, for example, could theoretically be characteristic of a particular subtype of BD that just happens to be responsive to lithium (Moore et al In press
). This is one example of etiological heterogeneity that may be a significant contributor to inter-study variability.
(2). Another example of disease heterogeneity is the difference in the pathophysiological mechanisms underlying early-onset and late-onset depression discussed above. Other hidden etiological differences may be an even more pernicious source of bias. In genetic studies, a distinction is often made between those patients who are recurrently ill and those subjects who have experienced one life-time episode of depression (Zubenko et al 2002
), but this is rarely seen in the imaging literature. Further, the degree to which MDD or BD runs in the families of recruited subjects or whether MDD patients are recruited from BPD families is not always detailed. Differences between psychotic and non-psychotic “subtypes” of BD may also contribute to inter-study variability in findings. Approximately 50% of BD I patients experience psychosis (Keck et al 2003
) (Ketter et al 2004
) and recent evidence suggests that working memory deficits may be specific to those patients with a history of psychosis (Glahn et al 2007
) (Savitz et al in press
Information about obstetric or other developmental problems is rarely reported. The diagnosis of BD in children has increased sharply (mostly in the Unites States) in recent years, yet it is unclear as to whether children diagnosed with BD actually suffer from the same disorder as adults (Duffy 2007
). This is particularly true when the broader phenotype of severe mood dysregulation (irritability and hyperarousal), which overlaps with ADHD, is used (Duffy 2007
). A narrower pediatric clinical phenotype may more faithfully represent adult nosological categories (Brotman et al 2007
). Caution should therefore be exerted in extrapolating neuroimaging findings from these pediatric populations to adults and vice versa
, particularly when broad diagnostic criteria are used. This caveat is illustrated by (Biederman et al 2007
) who reported that pediatric BD patients with co-morbid ADHD show additive patterns of MRI-evinced volume loss characteristic of both disorders when studied independently.
(3). State versus trait outcome differences do not always receive the attention that they merit: identification of the neuroanatomical or functional changes associated with the acute effects of depression and mania may serve as a useful means of evaluating the efficacy of treatments. Conversely, mood-independent abnormalities may be endophenotypes that can be exploited for genetic studies.
(4). While functional and morphometric analyses of affective illness often receive separate treatment in the literature, these two factors cannot be divorced from each other on either theoretical or methodological grounds. As alluded to above, GM volume loss may occur secondary to glutamate-driven excitotoxicity, a phenomenon which presumably correlates with elevated glucose metabolic activity (Shulman et al 2004
). Volume loss may however, lead to an underestimation of metabolic activity because of partial voluming effects: the inclusion of hypometabolic WM or CSF in putatively GM-containing voxels.
An example of this potential confound can be found in our own work. Initially, we found evidence of reduced metabolic activity in conjunction with reduced volume of the subgenual ACC in the depressed phase of both MDD and BD (Drevets et al 1997
) (Drevets et al 2002a
). Compatible with this hypothesis, depressed BD samples treated chronically with lithium indeed showed elevated sgACC metabolism irrespective of correction for partial volume effects (Bauer et al 2005
) (Mah et al 2007
), consistent with evidence that lithium treatment is associated increase in the sgACC GM volume (Moore et al In press
Another example can be found in studies of the amygdala. Despite reports of volume reductions, hypermetabolism of the amygdala is a consistent feature of the literature, which suggests that the increase in activity, if genuine, may be substantial – on the order of 70%, even though state-of-the-art PET imaging technology detects this as only a 6% difference between depressives and controls due to spatial resolution (i.e., partial volume) limitations (Drevets et al 1992
). The partial voluming effect can be attenuated to some extent by region of interest (ROI) based analyses of a small central region of the ROI as implemented by the Drevets group, presumably facilitated their finding of substantially increased amygdalar activity (Drevets et al 2002b
Limitations in spatial resolution also have impinged on the efficacy of volumetric MRI measures. Further, most of the older and even current imaging platforms do not have the necessary spatial resolution to accurately detail volumetric changes in dimunitive anatomical structures such as individual amygdalar nuclei. This limitation may lead to the implicit reification of neuroanatomical function; that is, the notion that gross anatomical structures should be uniformally affected by mood disorders simply because they are imagable units.
If one takes this argument to its logical extreme it becomes questionable as to what constitutes a replication. In other words, how similar do the functional or structural changes reported across studies have to be in order to establish that the identical functional units are affected?
(5). A comparison of the analytical methods used in imaging analysis is beyond the scope of this review. Nevertheless, a brief discussion of the 2 primary analytical approaches to imaging data analysis bears mentioning. Voxel-based analysis (as implemented in imaging analysis software such as the Statistical Parametric Mapping [SPM] series) is based on a mass-univariate statistical approach that compares the differences between two or more subject groups at each individual voxel in the brain (Ashburner and Friston 2000
). As such, Gaussian random field theory is used to correct for the multiple dependent comparisons that would, if uncorrected, lead to false positive results (Ashburner and Friston 2000
Voxel-wise analyses, however, are also highly sensitive to Type II errors because spatial normalization algorithms cannot precisely overlay small structures like the amygdala and sgACC across subjects, exaggerating the statistical variance. Possibly therefore, voxelwise analyses should not be the method of choice when diminutive, subcortical structures are the subject of study (Bookstein 2001
) (Nugent et al 2006
). Despite this limitation, voxel-wise analysis affords the advantage of allowing agnosticism with respect to the pathophysiology of the disorder under investigation.
On the other hand, the region-of-interest (ROI) approach relies on extant empirical or theoretical data to identify candidate regions that may show inter-group differences. Thus, assuming the veridicality of the disease model, Bayesian inference suggests that the probability of a significant ROI result being a true positive, is higher than in cases where convergent a priori data are absent. Nonetheless, the ROI approach is time and resource intensive especially when subject groups are large. The strength of the method is critically dependent on the precision of ROI segmentation, which depends heavily on the reproducibility of the anatomical landmarks chosen to delineate the target structure.
To our knowledge, very few comparisons of ROI and voxel-wise approaches have been made in psychiatrically ill samples. In one such study, (Kubicki et al 2002
) that the voxel-wise approach for comparing cerebral volumes, termed “voxel-based morphometry (VBM)”, produced an analogous finding (volume of STG) to their previously published ROI study, but also showed changes in other regions, one of which had not been previously implicated in schizophrenia. Later schizophrenia studies produced reasonably convergent results across ROI and VBM analyses in cortical regions (Job et al 2002
) (Giuliani et al 2005
). (Douaud et al 2006
) extended the correspondence in VBM-ROI results to the striatum in a case-control study of Huntington’s disease; although it is unclear whether a VBM analysis would be able to detect the more subtle subcortical changes characteristic of affective illness. In at least one case, however, PTSD-associated hippocampal changes have been obtained in the same sample using both ROI and VBM approaches (Emdad et al 2006
We did not identify a clear difference in the results of VBM and ROI analyses in the BD and MDD literature ( through ), possibly because there are very few published VBM studies of subcortical regions.
(6). Genetic association studies are plagued by false positive results: with approximately 20, 000 genes, and multiple variants within each gene of interest, the a priori probability of a true association is low. Sample sizes in analyses which combine genetic and imaging data are by nature small, although this is offset to some extent by the relative precision of the phenotypic data that are collected. The problem can be partially ameliorated by targeted hypotheses.
A Heuristic Model
Top-down and bottom-up disruptions to cortico-striatal-limbic circuits are the most straightforward method of describing the pathophysiological and symptomatological changes associated with affective illness. We have not made any attempt to distinguish between MDD and BD, here. A graphical representation is provided in .
As discussed above, projections from the orbital and medial PFC to the amygdala and its associated limbic and brainstem nuclei form a “visceromotor network” that modulates endocrine, autonomic, and behavioral aspects of emotion (Ongur et al 2003
In the top-down model, impaired PFC function, or cortical-subcortical “disconnection” disinhibits downstream limbic projections, altering emotional behavior. For example, disinhibition of the amygdala projections to the bed nucleus of the stria-terminalus (BNST), hypothalamus and periaqueductal gray matter (PAG) (Behbehani 1995
) (Sah et al 2003
) may increase cortisol releasing hormone (CRH) release and anxiety symptoms. Disinhibition of projections from the amygdala to the nucleus basalis, locus ceruleus and ventral tegmental area (VTA) (Davis and Whalen 2001
) (Sah et al 2003
) could account for the alterations in cholinergic (ACh), noradrenergic (NE) and dopaminergic (DA) transmission which may affect mood and attention. Finally, disinhibition of amygdala projections to the ventral striatum (Cardinal et al 2002
) would attenuate reward-seeking and goal-directed behavior, potentially contributing to the anhedonia and amotivation characteristic of depression. According to the bottom-up model, a functional hypersensitivity of limbic nuclei such as the amygdala, raphe and parahippocampus would predispose to dysregulation of PFC-mediated regulatory mechanisms.
In reality, the distinction between bottom-up and top-down models is artifical. If MDD and BD are polygenic disorders underpinned by many genes of small effect size, affected individuals are likely to possess many different risk variants affecting multiple neuroanatomical pathways and a single point of origin is unlikely.
The etiology of the suspected hypersensitivity or lesions4
is unclear but most likely involves complex gene-environment interactions. For example, the s
allele of the 5-HTT promoter variant may predispose to reduced functional connectivity between the amygdala and perigenual PFC (Pezawas et al 2005
), which may be maladaptive under stress. Similarly, the NRG1
gene may impact the myelination process and therefore predispose to WM lesions and a disruption to cortical-subcortical connections in the presence of cardiovascular risk factors. Two nonsynonymous SNPs of another gene, proline dehydrogenase (oxidase) 1 (PRODH
)  have also been associated with frontal WM volume reductions in schizophrenia (Zinkstok et al 2007
Many other similar examples are likely to emerge over time and hold out great promise for disinterring the latent pathophysiological basis of affective illness.