Diversity of neurodegenerative pathophysiology in nondemented patients with major depressive disorder: Evidence of cerebral amyloidosis and hippocampal atrophy

Abstract Background Patients with late‐life depression may be at the preclinical stage of dementia. However, the neurodegenerative processes in late‐life depression are poorly understood. This study aimed to investigate the distribution patterns of amyloid pathology and neurodegeneration in a depressive population without dementia. Methods The study recruited 63 middle‐aged and elderly patients with major depressive disorder (MDD) and 22 control subjects. The MDD patients were further subdivided into those with mild cognitive impairment (MCI) (n = 24) and non‐MCI (n = 39) patients. We used the global standardized uptake value ratio of 18F‐florbetapir (AV‐45/Amyvid) positron emission tomography imaging as a biomarker of cerebral amyloidosis and the hippocampal volume as a biomarker for neurodegeneration. Cutoff points of brain amyloid positivity and hippocampal atrophy were determined using independent data obtained from clinically diagnosed Alzheimer's disease (AD) patients in a previous study. Results Most of the control subjects (81.8%) were biomarker‐negative, in contrast to the MCI MDD patients (37.5%). A relatively high proportion of the MCI MDD patients (12.5%) exhibited both amyloid positivity and hippocampal atrophy as compared to the control subjects (4.5%) and non‐MCI patients (5.1%). However, a considerable proportion of the MCI MDD patients (29.2%) were categorized into the group with hippocampal atrophy alone, and negative amyloid deposition, as compared to the control subjects (0%) and non‐MCI patients (5.1%). Conclusions This study highlights the expected heterogeneity of the processes of neurodegeneration in MDD patients. The diverse neurodegenerative processes may have important etiologic and therapeutic implications regarding neurodegenerative pathophysiology in late‐life depression.


| INTRODUC TI ON
Several meta-analyses (Diniz, Butters, Albert, Dew, & Reynolds, 2013;Jorm, 2001;Ownby et al., 2006) have consistently suggested that a history of depression approximately doubles an individual's risk of developing dementia later in life, including Alzheimer's disease (AD) and non-AD dementia. One pilot postmortem study (Rapp et al., 2006) showed that AD patients with a lifetime history of major depression have more pronounced amyloid plaque and neurofibrillary tangle, as compared to AD patients without a history of depression. Our previous studies (Wu et al., 2013(Wu et al., , 2016 indicated increased cerebral amyloid accumulation as measured by 18 F-florbetapir uptake in specific brain regions of nondemented patients with lifetime major depression relative to comparison subjects. These findings point toward the possibility that patients with lifetime major depression might be at an early preclinical stage of the disease in which the criteria for dementia or even mild cognitive impairment (MCI) have not yet been reached.
Insight accumulated over the years regarding dynamic change in biomarkers of AD pathology has led to the establishment of new research and diagnostic criteria (Jack et al., 2009(Jack et al., , 2010. These developments provide guidance on the early detection of underlying AD pathology and early prediction of neurocognitive degeneration. A new series of criteria was recently developed by the task force of the National Institute on Aging and the Alzheimer Association (NIA-AA), mainly for research purposes, which made specific assumptions about dynamic relationships among AD biomarkers in an ordered manner (Albert et al., 2011;Jack et al., 2012;McKhann et al., 2011). Amyloid biomarkers as assessed by positron emission tomography (PET) imaging of amyloid or cerebrospinal fluid (CSF) amyloidβ can be detected to be abnormal as early as 20 years before significant clinical symptoms appear. Neurodegenerative biomarkers such as CSF tau, 18 F-fluorodeoxyglucose ( 18 F-FDG)-PET, and hippocampal volume as assessed by magnetic resonance imaging (MRI) become abnormal later and are then followed by significant clinical symptoms of cognitive impairment (Sperling et al., 2011). Biomarkers can be classed into two categories: those of an underlying amyloid pathology (CSF amyloidβ or amyloid PET) and those of neurodegenerative features (hippocampal atrophy on MRI, CSF tau, and hypometabolism on 18 F-FDG-PET).
Several researchers have reported that up to 50% of depressed elderly subjects meet the criteria for clinical diagnosis of MCI, despite differences in methodology and the definition of cognitive impairment (Bhalla et al., 2006;Lee et al., 2007;Yeh et al., 2011). This rate is far higher than the prevalence of MCI reported in the general population, which ranges from 3% to 19% (Gauthier et al., 2006). This implies that some neurodegenerative processes might underlie the high prevalence of MCI among elderly depressed patients.
Whereas patients with late-life depression represent an etiologically heterogeneous group (i.e., different age at onset, differing severity and episodes, differing medical comorbidities), it is not surprising that cognitive impairment in late-life depression should involve different ongoing mechanisms. However, the patterns of the neurodegenerative processes underlying cognitive impairment in elderly depressed patients are poorly understood (Jellinger, 2013).
The recently published NIA-AA criteria mentioned above might provide new insight and framework to explore the patterns of neurodegenerative processes in elderly depressed patients, and may allow them to be categorized into different biomarker-based groups.
In the present study, we focused on a population of nondemented patients with major depression and aimed to apply the two categories of biomarker proposed in the NIA-AA criteria to investigate the distribution patterns of amyloid pathology and abnormal neurodegeneration in a depressed population.

| Subjects and protocol
The subjects enrolled in the present study were recruited from a longitudinal clinical cohort study launched in 2011, which was performed to investigate cerebral amyloid deposition in nondemented patients with major depressive disorder (MDD). The patients were recruited consecutively from geriatric psychiatric outpatients at Chang Gung Medical Center from August 2011 to July 2015. The control subjects were recruited through public advertisements during the same period. Every MDD patient was assessed for the presence of lifetime DSM-IV major depressive episodes by clinical interview, and medical information was obtained from medical records and attending physicians. Control subjects were confirmed as having a lifetime absence of psychiatric illness. All subjects were aged >50 years, and functioned well in activities of daily living; they did not have clinically significant medical or neurological diseases, and had not abused alcohol or other substances within the past 1 year at the time of study enrollment. None of the subjects met the NINCDS-ADRDA criteria for probable AD or the DSM-IV criteria for dementia. All eligible subjects underwent 18 F-florbetapir PET study, brain MRI, and cognitive assessment. The patients' Apolipoprotein E (ApoE) genotype was also classified by polymerase chain reaction, and vascular risk factors as defined by the Framingham stroke risk score were identified, as were clinical characteristics of lifetime major depression. Written informed consent was obtained from all subjects, and the study protocol was approved by the Institutional
Individual original scores were transformed into standardized z-scores, which were generated using regression-based norms and adjusted for age and educational level according to independent normative data for Taiwan (Yeh et al., 2011). MCI was defined in MDD patients who exhibited impairment in at least one of the cognitive domains, as shown by a score of 1.5 SD below the age-and education level-adjusted norm (Petersen, 2004;Petersen et al., 2001).
The CDR had to be only 0 or 0.5 for all subjects. We used the CDR Sum of Boxes (CDR-SB) method to characterize cognitive and functional performance.

| Image acquisition
The radiosynthesis of 18 F-florbetapir  and amyloid PET data acquisition

| Image analysis
All PET image data were processed and analyzed using PMOD image analysis software (version 3.3; PMOD Technologies Ltd, Zurich, Switzerland) . Seven volumes of interest (VOIs), the frontal, anterior cingulate, posterior cingulate, precuneus, parietal, occipital, and temporal areas, were selected , and the regional standardized uptake value ratio (SUVR) using the whole cerebellum as the reference region was calculated. Moreover, the average SUVR from 7 cerebral cortical VOIs was computed as the global cortical SUVR for further analysis.
FreeSurfer image analysis software (version 5.3.0; https://surfer. nmr.mgh.harvard.edu/) was used to measure the hippocampal and intracranial volumes. To reduce intersubject variability, hippocampal volumes were corrected for the intracranial volume (ICV). A normalization method based on linear regression between the VOI and ICV was applied (Voevodskaya et al., 2014) in order to obtain the adjusted hippocampal volume (HVa) as follows: where HV was the raw hippocampal volume and ICV indicated the intracranial volume for each subject. For correction, β was the slope of the regression line between the ICV and hippocampal volume of the controls, and ICV mean was the average ICV of the control group.

| Imaging biomarker cutoff points
As the imaging biomarkers were all continuous measures in the present study, every biomarker based on the NIA-AA criteria was required to be designated normal or abnormal (Sperling et al., 2011).
Thus, cutoff points needed to be selected to dichotomize biomarkers in order to divide the subjects into normal or abnormal groups.
As FDG-PET and CSF data were not available in our study, we employed the global 18 F-florbetapir SUVR obtained by PET and the HVa as measured by MRI as cerebral amyloidosis and neurodegenerative biomarkers, respectively, to categorize MDD patients in accordance with the NIA-AA criteria.
The results of a previous study published by our group , which included 12 clinically diagnosed AD patients and 11 cognitively normal controls who had undergone the same 18 F-florbetapir PET and MRI analyses, were used to set imaging biomarker cutoff points. The threshold for global cortical amyloid positivity was constructed by the ROC method, as previously described . The cerebral amyloid-positive cutoff point was 1.178, with a sensitivity of 92% and a specificity of 91%.
The same ROC method was applied to determine the cutoff point for hippocampal atrophy: the HVa cutoff point was 6,879 mm 3 , with a sensitivity of 88% and a specificity of 100%.

| Statistical analysis
Data were expressed as means ± SD or an absolute number with a proportion for descriptive statistics. Group comparisons between the controls, non-MCI and MCI MDD patients and across the four biomarker groups were made using nonparametric Kruskal-Wallis tests with Dunn's multiple comparison post hoc analysis for continuous variables and χ 2 tests for categorical data. A p value of 0.05 was defined as the threshold of statistical significance in each test.

| RE SULTS
The study recruited 63 nondemented MDD patients and 22 control subjects. Twenty-four (38.1%) MDD patients met the clinical criteria for MCI at the time of imaging study. MCI MDD subjects had the lowest HVa among the three groups (p < 0.001); they also had a higher global 18 F-florbetapir SUVR than the other two groups, but this was not statistically significant

| D ISCUSS I ON
The present study was a preliminary study that employed a In particular, an important finding of the present study was the high percentage of MCI MDD patients who were amyloidnegative but had hippocampal atrophy. This finding clearly provided information that conflicted with the biomarker model of AD proposed by Jack et al. (2010Jack et al. ( , 2013 in which amyloid deposition becomes apparent first, and precedes other neurodegenerative biomarkers such as hippocampal atrophy or hypometabolism according to FDG-PET.

Number of depressive episodes
Mean ± SD 1.8 ± 1.1 1.4 ± 0.5 1.5 ± 0.7 2.5 ± 2.  MDD, and 0% of control subjects, respectively. The present study provided the evidence of the heterogeneity of neurodegeneration in MDD patients. In particular, the results of this study implied large proportion of MCI MDD patients with sNAP might enter the neurodegenerative process of non-AD types of dementia. Taken together, our results provided partial support for the recent NIA-AA criteria for an AD model, but also suggested that underlying factors other than the amyloid cascade of AD pathology can drive neurocognitive degeneration in MDD patients.
Several studies identified a reduced hippocampal volume in MDD patients, which has been reported to be a consequence of repeated episodes of major depression (Hickie et al., 2005;Sheline, 2003;Sheline et al., 2003;Videbech & Ravnkilde, 2004). The mechanism behind the reduced hippocampal volume remains unclear. It has been well-documented that hypothalamic-pituitary-adrenal (HPA) axis dysfunction might lead to hypercortisolism (Arborelius, Owens, Plotsky, & Nemeroff, 1999;Checkley, 1996), which is toxic to the hippocampus and further results in hippocampal shrinkage (McEwen, 2000;Sapolsky, 2000). However, it is not known whether the hippocampal atrophy observed in MDD patients might lead to changes in dementia status in later life, nor which types of dementia may be affected (Videbech & Ravnkilde, 2004).

| Limitations
Some issues and limitations need to be raised. The hippocampal volume was selected as the neurodegenerative biomarker in the present study because it has been well-studied as a validated MRI measure, and is also one of the neurodegenerative biomarkers included in the newly published NIA-AA criteria. One limitation of this study was that only the neurodegenerative biomarker of hippocampal volume was used, and no FDG-PET imaging or other CSF biomarkers were employed. Thus, the distribution rates might have differed if other biomarkers such as FDG-PET and CSF biomarkers had been included. In addition, although the newly published NIA-AA criteria provide a conceptual framework, several operational issues remain to be resolved, including standardization methods for biomarker measures, and consensus in the definitions of cutoff points for biomarkers (Jack et al., 2012). Thus, a population-based means of defining abnormality was unavailable in this study, and some subjects at the margins of the biomarker cutoff points would inevitably have been classified into incorrect biomarker groups. Together, these operational issues limited and hampered mutual comparison of data obtained from different studies. However, in attempting to implement the NIA-AA criteria, we performed a preliminary study in a MDD population that could be used as a basis for further exploration.
One additional limitation was the small sample size influenced the distribution of subjects into the different biomarker groups. The small sample size also meant that the MCI MDD patients could not be further subdivided into subgroups according to different domains of cognitive deficit (e.g., amnestic or nonamnestic MCI). Of note, this study was a clinical-based study; thus, the control and MDD subjects differed from samples from the community or those in population-based research. Our results cannot be generalized to the general population. Future long-term studies with large sample sizes employing more neurodegenerative biomarkers are needed in order to examine in depth the neurodegenerative processes in elderly depressed patients.

| CON CLUS ION
This study highlights the expected heterogeneity of the processes of neurodegeneration in MDD patients. Some of the MCI MDD patients had entered the neurodegenerative process and were evident in the prodromal stage of AD dementia. In particular, other MCI MDD patients who were amyloid-negative but had abnormal hippocampal atrophy might represent prodromal stages of other non-AD types of dementia.

ACK N OWLED G M ENTS
We thank Avid Radiopharmaceuticals Inc. (Philadelphia, PA, USA) for providing the precursor for the preparation of 18F-florbetapir.
This study was carried out with financial support from the National Science Council and the Ministry of Science and Technology, Taiwan Hospital, Linkou, for technical support.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflicts of interest.

RO LE O F TH E FU N D E R /S P O N S O R
The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and the decision to submit the manuscript for publication.