To our knowledge, this is the first study to report a relationship between chronic psychological distress and regressive dendrite and dendritic spine changes in humans, specifically a reduction in MAP2-immunoreactive dendrites and synaptopodin-immunoreactive dendritic spines in the CA3 subfield of the hippocampal formation. Our results are reminiscent of the extensive data in jewel fish, mice, rats, and tree shrews, which have shown that chronic stress induces atrophic changes in the architecture of neuronal dendrites and spines.26,28–31,63–66
Dendrites and spines are crucial components for synaptic function and plasticity. Changes in their density or morphological features can result in significant alterations in the connectivity of neural systems67,68
and would be expected to result in changes in the neurobehavioral functions subserved by those systems. Chronic stress exposure in rodents causes reversible atrophy of dendrites and spines in the dentate gyrus and hippocampus, as well as impairments in hippocampal-based spatial memory.69–72
Whether these anatomic and functional effects of chronic stress are causally related or are parallel phenomena has not been established yet. Although a direct pathophysiological relationship has been widely assumed because of the similar time courses of anatomic and functional changes with experimental, environmental, or pharmacological stress manipulations, important exceptions to this relationship have been recognized (for review, see Conrad32
). It is also important to note that our measures of psychological distress may be analogous but not equivalent to the presumed effects of chronic stressors used in animal experiments. Although our study found a similar association between distress and our anatomic variables of interest, we cannot assume that the pathophysiological mechanisms are similar.
Human postmortem research bears many uncontrollable factors that may confound data or their interpretation. Nevertheless, such research plays an important role in translating findings from preclinical studies to humans. Intraspecific biological and environmental variability, the reliability and fidelity of clinical assessment measures, changes that may occur during the intervals between death and the clinical assessments of participants, comorbid medical illnesses and pharmacotherapy, cause and manner of death, postmortem interval between death and tissue fixation, and other factors all may affect sensitive and dynamic neuroanatomic structures and molecular integrity. With regard to these factors, our sample has some notable strengths as well as potential weaknesses.
First, the ROS cohort from which this sample was drawn differs from the general US population in terms of educational, social, cultural, and other lifestyle factors. Many potential sources of variability are reduced in this group, possibly increasing the precision of the associations, but the generalizability of findings to other populations also may be limited. For instance, ROS participants are highly educated, generally maintain a healthy lifestyle, and are largely without histories of major psychiatric illness. The “cerebral reserve” that education and other lifestyle factors purportedly confer or reflect may have allowed study participants’ brains to tolerate more distress before anatomic changes or cognitive effects were manifest. Furthermore, although the participants we studied exhibited a range of scores on the psychological distress rating scales administered, these were generally mild. Therefore, because this study showed a significant association between distress measures and changes in hippocampal anatomy, we consider that the general population may present changes that are as great or even greater. Associations of psychological distress with cognition have been found in other cohorts, including the Rush Memory and Aging Project, a lay cohort of similar study design,6,73
and the Chicago Health and Aging Project, a longitudinal population-based study in a biracial community.15
These data suggest that, despite differences in some risk factors, associations with cognition are generally replicable.
Most postmortem neuroanatomic and molecular studies in psychiatry are typically conducted in clinical samples of patients with major psychiatric illness. Our study is unique in that it investigates the neuroanatomic and cognitive correlates of anxiety and depressive symptoms in people who have no history of major psychiatric illness. It is curious that, despite CA3 and the mossy fiber terminal zone being such a prominent focus of anatomic and molecular changes in animal models of chronic stress, there have been relatively few postmortem studies of this important hippocampal subfield in psychiatric illness. In schizophrenia, findings in CA3 have included decreased densities of mossy fiber synapses,74,75
decreased spinophilin messenger RNA expression,76
decreased dysbindin 1 protein expression in the mossy fiber terminal zone,51
decreased chromogranin B and synapsin I in mossy fibers,77
and a variety of neurotransmitter receptor abnormalities.78–81
Such changes may be specific to schizophrenia or, alternatively, they may be nonspecific effects of the psychological distress that accompanies this severe mental illness. To our knowledge, there have been no similar postmortem studies in major depression or anxiety disorders that might shed light on this. Thus, it is noteworthy that, even in our nonclinical sample, we observed an association of dendrite and spine densities with trait anxiety and longitudinal symptoms of depression.
The dendrite and spine atrophy that occurs with chronic stress exposure in experimental animals is dynamic. It occurs over days and, upon discontinuation of the stress, reverts to normal.26
In our human sample, there are likely to have been various stressors unknown to us occurring in the weeks before death. Illnesses culminating in natural deaths are major stressors, and it is not clear what effects these may have had on dendritic and spinal architecture.
Also of particular interest are medications that participants may have been taking, such as antidepressants, anxiolytics, antipsychotics, and estrogens, which have well-established though complex effects on dendrites and spines.66,82–85
We considered these medication exposures specifically, as well as a variety of other potentially confounding variables such as illnesses, causes of death, acuity of death, coexistent neuropathological lesions (eg, infarction, AD lesions, and Lewy bodies), and various other medication exposures. Each of these may affect neuron, dendrite, or spine morphometry in vivo or even exert differential effects on tissue shrinkage during postmortem tissue processing.56
We statistically examined our neuroanatomic data for associations with each of these potential confounders but found none. Nonetheless, caution is advised in attributing the relative atrophy of hippocampal dendrites and spines solely to the enduring effects of trait anxiety and enduring depressive symptoms without applying other analytic approaches and replication in other samples.
In interpreting our investigation of dendrites and spines in light of studies of chronic stress in animal models, it is important to emphasize that we investigated the association of symptoms and traits of psychological distress and not stress per se. Exposure to stressors—whether experimental or naturally occurring, acute or chronic—should not necessarily be conflated with the experience of psychological distress because the neurophysiological, molecular, and anatomic correlates are diverse and complex.
We used a novel approach to estimate the densities of dendrites and spines in this study. Most morphometric studies of dendrites and spines in experimental animals and in human postmortem brain tissues have used Golgi staining with Scholl analyses or other labor-intensive approaches to estimate the number, lengths, and patterns of labeled processes and objects of interest in comparatively small samples. In the present study, we used a more high-throughput approach consisting of immunohistochemical labeling in thin sections for proteins that are highly enriched in the neuronal compartments of interest, ie, MAP2 in dendrites and synaptopodin in dendritic spines. For quantitation, we used semiautomated algorithmic image analysis to segment and estimate the densities of neuronal profiles, dendrites, and spines. There are relative advantages and disadvantages to each approach. Morphological detail with Golgi staining can be exquisite, and much of the full extent of a given neuron’s dendritic tree can be measured. However, Golgi staining labels the somata, axons, dendrites, and spines of only some neurons, and which neurons it labels is unpredictable. Finally, Golgi staining is technically lengthy and challenging, variable from case to case, and does not lend itself well to larger-scale studies. In contrast, immunohistochemistry with antibodies directed at MAP2 and synaptopodin respectively label all dendrites and spines that express those proteins, is technically more uniform and robust, can be conducted in variously fixed and processed tissues, and is suitable for high-throughput analysis of relatively large samples. On the other hand, limitations of this method can include variability in enzymatic reaction product contrast and resolution, especially for dense and overlapping dendrites and spines, different shrinkages of tissue with heat-induced or other epitope retrieval methods, and vulnerability to split-object artifact inherent in 2-dimensional quantitation. We attempted to attenuate these limitations by using computer-assisted algorithmic size, shape, optical density, and pixel contiguity filtering to delineate and then measure standard 1-pixel-wide lengths of skeletonized dendrites or count segmented synaptopodin-immunoreactive puncta. These methods allowed for more reliable measurement, although they precluded measurement of certain data that may be informative, such as thickness, shape, or staining intensity. We also applied a correction factor to moderate split-cell artifact bias in our 2-dimensional neuron density determination. Finally, we “normalized” our dendrite and spine densities to the densities of neurons in CA3, thus eliminating the potential confounder of variable tissue shrinkage.
It has been suggested that chronic stress in animals causes memory impairment via stress-induced synaptic changes. Accordingly, we initially hypothesized that atrophy of dendrites and spines associated with anxiety and depression would mediate the relationship between these distress traits and cognitive impairment that our group has previously described in the ROS and Memory and Aging Project cohorts.6–9,11,12,15,38–40,73
Our analyses did not support this. One possible explanation is that our study was underpowered to detect a relationship with cognition. Another explanation is that dendrite and spine atrophy in CA3 is only one of many factors contributing to cognitive decline. Common neuropathological lesions in older people, especially neurofibrillary tangles and amyloid plaques, may have more powerful effects on cognition, obscuring the lesser effects of distress-related neuroanatomic changes in CA3. Indeed, while there was no relationship between anxiety and depression and densities of tangles and plaques in this sample, the relationship between these lesions and cognition was strong, as expected. Finally, another possibility is that dendrite and spine atrophy does not lie in a causal chain linking chronic stress and cognitive impairment, as has been proposed.32
Chronic stress in animals may cause neuroanatomic and behavioral changes via parallel but independent mechanisms that may not be the same in humans.
One final consideration is that human postmortem studies are limited in that they can only give a snapshot of the brain as it was at the time of death. As such, postmortem studies are useful for showing associations of clinical and brain data but cannot ascertain causal relatedness with certainty. Therefore, although animal studies of stress have shown causality in one direction, the results found in this study could imply causality in the opposite direction for humans. In other words, inherently reduced numbers of dendrites and spines could be one factor that makes a person more susceptible to distress than another.