Structural Correlates of Depressive Symptoms in Prodromal Alzheimer’s Disease

Background/Aims: We conducted a cross-sectional study to investigate the structural magnetic resonance imaging correlates of depressive symptoms at the initial clinical stages of Alzheimer’s disease (AD). Methods: Subjects aged 65 or more were categorized as prodromal AD (n=18), mild AD (n=35), or normal cognition (n=76). Depressive symptoms were measured by means of the 15-item abridged version of the Geriatric Depression Scale. Potential gray matter correlates of depressive symptoms were analyzed using the Statistical Parametric Mapping software package. Results: Significant results were obtained in the prodromal AD group only. In that group, depressive symptoms were related to atrophy in the left precentral gyrus (Brodmann area 6) (p ≤ 0.01, FWE corrected). Conclusion: Our results, added to the existing literature, suggest that dysfunction in left-sided, cognitively and functionally salient, cortical regions along with relative preservation of deficit awareness, provided by the right hemisphere, explain depressive symptoms in the initial clinical stages of AD.


INTRODUCTION
Depressive symptoms are frequently reported in Alzheimer's disease (AD) [1,2], particularly at the initial clinical stages of the disease process [3,4]. Those symptoms, which often compromise the patient's and the caregiver's quality of life [4], pose important challenges in terms of physiopathology and treatment. Although several mechanisms have been proposed to explain the presence of depression in AD, the nature of this association remains controversial [5]. Usually interpreted as a symptom of, or a psychological reaction to, AD [6], depression has also been regarded as a trigger that lowers the threshold for dementia symptoms [7] and even as a hypothetical source of toxicity for the hippocampal neurons, thus increasing the risk of dementia [8,9]. Given that complexity of potential mechanisms, one may find discrepancies amongst results in clinical and epidemiological studies of AD and depression [5].
Structural studies can enlighten the mechanisms of depression in AD. In previous seminal research, Zubenko and collaborators reported selective damage in the locus coeruleus and the substantia nigra of patients with dementia who had also suffered from depression [10], which was accompanied by selective deficits in norepinephrine and serotonine [11]. However, those post-mortem studies were conducted in patients with major depression, hence not necessarily explaining the typically less severe depressive symptoms of the initial clinical stages of AD [12,13]. In a recent magnetic resonance imaging (MRI) study, which included patients with mild to moderate AD dementia, depression was related to atrophy in the left inferior temporal gyrus [14]. In other study, cortical thinning in the left parietal and temporal regions was observed in two independent cohorts of patients with AD and depression and a possible mediation of tau pathology was described [15]. In a 1 H-magnetic resonance spectroscopy ( 1 H-MRS) study, depressive symptoms of AD patients were associated with an increase in choline/creatine peak in the left dorsolateral prefrontal cortex, possibly indicating a neuronal membrane breakdown in that specific region [16].
We conducted a cross-sectional study to investigate the anatomical correlates of depressive symptoms in subjects with normal cognition, prodromal AD, or mild AD. The present investigation was performed in the context of a longitudinal collaborative project, which will also be presented in this paper.

Setting and Subject Characteristics
The Observational Multicenter Longitudinal Study (ELMO, as from the Spanish initials) was envisioned, designed, and conducted by the Dementia Group of the Autonomous Community of Madrid (DEMCAM) to describe and analyze the clinical, biochemical, and MRI features of the initial stages of AD. Subjects aged 65 or more from 13 memory clinics of the Autonomous Community of Madrid (Spain) were invited to sign a consent form and participate in the survey (see the ELMO centers and investigators in the Acknowledgement section). Subjects were recruited from December 2008 to March 2012. The ELMO was approved by the Ethics Committee of the La Paz University Hospital.
A stringent set of exclusion criteria was utilized for enrolment of subjects. Patients with noncontrolled medical diseases and those with poor expected survival or study compliance (e.g., patients with disseminated cancer, liver cirrhosis, lung fibrosis with severe respiratory insufficiency), were not included in the study. Potential psychiatric diseases were detected by means of patient and caregiver interview and medical records review, following the Diagnostic and Statistical Manual of Mental Diseases (DSM-IV-TR) [17]. Past major depressive or anxiety disorders were permitted, but (either past or present) psychotic, bipolar, or substancerelated/addictive disorders were cause of exclusion. Tremor at rest of any severity and moderate or severe rigidity, bradykinesia, or instability, as defined by the Unified Parkinson's Disease Rating Scale [18], were not permitted. Additional exclusion criteria were as follows: a) illiteracy; b) important neurosensory defect; c) subject unable to attend the study visits; d) lack of informant (with the exception of subjects with normal cognition); e) less than six months of symptom duration (in case of cognitive deterioration); f) clinically atypical AD [19]; g) cerebrovascular disease, either clinically recorded or observed as significant vascular lesions in neuroimaging study (i.e., cortical or strategic infarction or severe leukoaraiosis); h) any circumstance that could interfere with, or advice against, performance of MRI study, and i) presence of any systemic or neurological disorder, apart from AD, that could be responsible for the observed cognitive deficit.

Medical and Neurological Variables
A structured medical and neurological evaluation was conducted by the ELMO physicians (neurologists or geriatricians), who also reviewed all the available information from medical records and ancillary studies. The ELMO protocol included questions or items related to disease duration, memory symptoms, behavioural and psychological symptoms of dementia, motor symptoms, concomitant diseases and medications, physical exam (blood pressure, heart rate, height, weight, abdominal perimeter), neurological exam, and Hachinski score [20].

Neuropsychological Evaluation
The following tests and scales were conducted by neuropsychologists: This is the most widely used brief test of general cognition in dementia. The score ranges from 0 (worst) to 30 (best cognition) [ The FAQ is a questionnaire of instrumental ADL, which is administered to an informant. It comprises 11 items and is scored from 0 (total independence) to 33 (total dependence) [26]. -Geriatric Depression Scale (GDS). The 15item abridged version of the GDS was used to analyze the potential structural correlates of depression in the present investigation. This instrument measures depressive symptoms by means of direct interview with the subject. A score from 0 (best) to 15 (worst affective status) is obtained [27]. -Clinical Dementia Rating (CDR). This instrument assesses three cognitive and three functional domains using a semistructured interview with an informant and also some interview and mental status exam with the patient. A final score of 0 (no dementia), 0.5 (questionable dementia), 1 (mild dementia), 2 (moderate dementia), or 3 (severe dementia) is obtained [28].

ELMO Groups
After neurological and neuropsychological evaluation, the ELMO participants were classified into the following clinical groups: a) Normal cognition (NC). These subjects may or may not have had memory complaints, but performed within normality (i.e., above 1.5 standard deviation [SD]) in age-and education-adjusted neuropsychological tests. b) Mild cognitive impairment (MCI). These subjects performed below normality (i.e., 1.5 SD) in at least one score of episodic memory (amnesic type, aMCI) and could also perform below normality in some other neuropsychological domain (multidomain type, mMCI). General cognition and functional capacities were essentially preserved and dementia diagnosis was absent [29]. Subjects from this group received a Clinical Dementia Rating (CDR) score of 0.5 [28]. Only those MCI subjects who had developed AD dementia at the last available follow-up visit will be presented and analyzed in the present paper (prodromal AD group). c) Mild AD. These patients received a diagnosis of probable AD according to the NINCDS-ADRDA criteria, i.e., they suffered from deterioration in memory and at least one other neuropsychological domain, which significantly interfered with some instrumental ADL [19]. Subjects from this group received a CDR score of 1 [28].

Blood Sample
Complete blood count and determinations of glucose, ions, B12, folate, kidney, liver, and thyroid function were conducted in all the ELMO participants. Apolipoprotein E gen (APOE) haplotype was also determined in a subset of participants. In addition, blood samples were kept for potential future biological determinations.

MRI Study
The MRI study of the ELMO was acquired in a GE Healthcare 3T HDx scanner at the Neuroimaging Department of the Alzheimer Center Reina Sofía Foundation (ACRSF). The scanner was equipped with an eight-channel phased array receive coil and a gradient system capable of producing a gradient strength of 40mT/m. A comprehensive, 50-minute magnetic resonance acquisition protocol which included volumetric, perfusion, diffusion tensor imaging (DTI), echo-planar, fluid attenuated inversion recovery (FLAIR), and spectroscopical series was conducted. In the present paper, only the acquisition and post-processing procedures of the volumetric study will be described and analyzed.
A 3D-T1-weighted gradient echo volume was acquired for each subject. The 3D-T1-weighted acquisition parameters were as follows: TE = 4.152 ms, TR = 9.4 ms, TI = 650 ms, flip-angle = 12°, slice thickness =1 mm (with no gap between slices), acquisition voxel size = 0.83 × 0.83 × 1 mm, FOV = 240 mm, and sagittal acquisition with whole brain coverage. The statistical parametric mapping (SPM) technique and the diffeomorphic anatomical registration through exponentiated lie (DARTEL) algebra method [30] were used to perform voxel-based morphometry (VBM) analyses. Segmentation of the T1 weighted images in grey matter, white matter and cerebrospinal fluid [31] was followed by the elaboration of a grey matter template of an oldage, cognitively normal control group, which was created using the DARTEL procedure. The normalized volumetry maps were averaged and smoothed using a [8 mm, 8 mm, 8 mm] full width at half maximum (FWHM) isotropic Gaussian kernel. Smoothing of the normalized maps is known to reduce sensitivity, but it is required to minimize normalization errors as the anatomy between individuals is highly variable. By smoothing, we achieve more overlapping between subjects for a given cluster. Furthermore, data smoothing prepares the data to fit the SPM Gaussian random field theory working frame. The drawback will be some loss of spatial precision. All native maps were normalized to the template using the Montreal Neurological Institute (MNI) coordinates. The MNI normalization option was chosen in order to locate significant regions in a well defined space.

Statistical Analyses
The demographic and clinical variables of the three study groups at the time of the ELMO inclusion were presented and compared by means of descriptive statistics, Chi-square test, and one-way analysis of variance. As mentioned above, MRI data processing was performed using the SPM version 8 (SPM8) software package, which computes statistical parametric maps for localizing significant differences between two or more experimental groups and also allows for correlation assessment between the images and the potentially explanatory variables, using the general linear model. For the present investigation, the significant correlations in brain structures according to the GDS score were searched, firstly for the whole sample and secondly for each study group, using NC group as the reference. The design matrix was defined as multiple regression with one covariate of interest (i.e., GDS) and three confounding variables (i.e., total brain gray matter volume [32], gender, and age). Gray matter rather than intracranial volume was utilized because regional, instead of generalized, atrophy should be expected in the initial stages of AD [33]. Cluster-level, family-wise (FWE), statistical corrections for multiple comparisons were applied and the threshold for statistical significance was set at p<0.05 (two-sided, FWE corrected).

RESULTS
The general disposition of the ELMO participants and the subjects who were included in the present study are depicted in Fig. 1. From a total of 201 participants that signed the consent form, 72 subjects had to be excluded due to: a) MCI that did not develop into AD dementia at the last available follow-up visit (49 subjects), b) subject did not have a usable MRI for the present study (15 subjects), and c) other reasons (8 subjects). Thus, 129 subjects were analyzed for the present paper. The mean age was 73.4 (SD 6.1, range 65 to 90), there was predominance of women (63.3%), and mean time of education was 10.7 years (SD 6.1, range 0 to 25).
The main clinical variables across the three study groups are presented in Table 1. Subjects with mild AD were older and had less educational achievement than subjects with NC. As expected, there were differences in cognitive performance, particularly regarding memory tests, and in functional status across the three study groups. Moreover, a higher frequency of APOE ε4 allele was observed in both AD groups. The severity of depressive symptoms according to the GDS was mild (mean score 1.9, SD 2.2, range 0 to 10) and comparable across the three study groups. Only eight subjects (6.2%) surpassed the GDS cut score for clinical depression (5/6).
No statistically significant results were found in the regression analysis of the structural correlates of depression when the total sample was analyzed (all p-values >0.05, data not shown). However, in the separate analyses of the study groups, significant negative correlations emerged in the group of prodromal AD. In that group, depressive symptoms were related to atrophy in the left precentral gyrus (Brodmann area [BA] 6) ( Table 2, Fig. 2). In addition, a very small region of atrophy was detected, related to depressive symptoms, in the right postcentral gyrus (BA 2).

DISCUSSION
We conducted a cross-sectional study to determine the potential gray matter correlates of depressive symptoms in the initial clinical stages of AD. The research context was not a large but highly restrictive multicenter longitudinal study that conducted extensive clinical assessment and 3T structural MRI study. As expected, the participants displayed good physical health, with low frequency of vascular features, and virtual absence of clinical depression (Table 1). This restrictive research context, albeit compromising the generalizability of results, offered the opportunity to sensitively investigate the structural underpinnings of mild/subsyndromal depressive symptoms at AD clinical inception.      Fig. 2). To the authors' knowledge, this finding was not reported before. Brodmann area 6 participates in verbal working memory [34] (i.e., the human capacity of temporarily holding information and having that information rapidly accessible) [35] and was proposed as the substrate of neural transitions from motor to cognitive functions [36]. Several structural [14][15][16]37] and functional [38][39][40] imaging studies demonstrated associations between depressive symptoms and gray and white matter atrophy or hypoperfusion in patients with AD and depression, particularly in the left prefrontal cortex [38][39][40].
Hence, it is unlikely that our results disclosed the genuine or critical region for depressive symptoms in the initial clinical stages of AD. By contrast, consistently with the previous studies, we propose a pattern of 'asymmetrical', predominantly left-sided, brain dysfunction that may be determinant to elicit a situation of cognitive and functional difficulties, with relatively well-preserved consciousness of deficits, leading to depressive symptoms and, if other psychological and biological factors concur, t clinical depression (Fig. 3). This explanation is supported by the observation of those GDS items related to cognition and activities of daily living (ADL) performance (i.e., 'have you dropped many of your activities and interests?', 'do you feel you have more problems with memory than most?', 'it is hard for you to get started on new projects?'), which were the most frequently endorsed by subjects from the prodromal AD group (post-hoc observations, data not shown).
Giving further support to the proposed previous longitudinal study of MCI subjects found that atrophy within left perisylvian areas associated with naming and fluency decline, whereas left frontal atrophy was associated with decline in executive functions [41]. Moreover, left and right hemispheres have been associated with the processing of, respectively, positively and negatively valenced stimuli [42]. A situation of predominantly left damage could produce a selective processing of the emotionally negative stimuli, further contributing to depression.
Asymmetrical cerebral dysfunction is frequently observed in clinical practice at the beginnings of AD and is supported by neuropsychological, positron-emission tomography (PET), and pathological studies of AD [43]. To what extent this asymmetrical pattern of brain dysfunction contributes to depression in AD should be a matter of future research. Evolution towards widespread, less asymmetrical, brain damage in AD, along with the loss of deficit awareness, would explain the absence of significant results 7 symptoms in the initial clinical stages of AD. By contrast, consistently with the previous studies, we propose a pattern of 'asymmetrical', brain dysfunction that may be determinant to elicit a situation of cognitive and functional difficulties, with relatively preserved consciousness of deficits, leading to depressive symptoms and, if other psychological and biological factors concur, to clinical depression (Fig. 3). This explanation is supported by the observation of those GDS items related to cognition and activities of daily living (ADL) performance (i.e., 'have you dropped many of your activities and interests?', 'do you e more problems with memory than most?', 'it is hard for you to get started on new projects?'), which were the most frequently endorsed by subjects from the prodromal AD observations, data not shown). Giving further support to the proposed model, a previous longitudinal study of MCI subjects found that atrophy within left perisylvian areas was associated with naming and fluency decline, whereas left frontal atrophy was associated with decline in executive functions [41]. Moreover, left ight hemispheres have been associated with the processing of, respectively, positively and negatively valenced stimuli [42]. A situation of predominantly left damage could produce a selective processing of the emotionally negative ting to depression.
Asymmetrical cerebral dysfunction is frequently observed in clinical practice at the beginnings of AD and is supported by neuropsychological, emission tomography (PET), and pathological studies of AD [43]. To what extent asymmetrical pattern of brain dysfunction contributes to depression in AD should be a matter of future research. Evolution towards widespread, less asymmetrical, brain damage in AD, along with the loss of deficit awareness, gnificant results in our mild AD group, thus supporting the proposed model for depressive symptoms at the initial clinical stages of AD (Fig. 3). Beyond AD, this model could also be applied to other brain conditions. In fact, depression after stroke was more frequently reported in patients with left sided lesions [44,45]. The present research is novel in the sense that 3T MRI equipment was utilized, and within regression analysis was conducted, thus increasing the sensitivity and specificity to capturing valid results. In fact, after a rigorous control of multiple comparisons, we could identify a new region related to depressive symptoms in prodromal AD. However, the present study has important limitations. First, subjects with major depression were excluded from the study and the depressive symptoms were, in most cases, of questionable clinical relevance. Hence, although some of the results and proposed mechanisms could be contributing, our results cannot be in our mild AD group, thus supporting the proposed model for depressive symptoms at the initial clinical stages of AD (Fig. 3). Beyond AD, this model could also be applied to other brain conditions. In fact, depression after stroke was re frequently reported in patients with left-

Alzheimer's disease; FWE: family-wise error
The present research is novel in the sense that 3T MRI equipment was utilized, and within-group regression analysis was conducted, thus increasing the sensitivity and specificity to ng valid results. In fact, after a rigorous control of multiple comparisons, we could identify a new region related to depressive symptoms in prodromal AD. However, the present study has important limitations. First, subjects with major luded from the study and the depressive symptoms were, in most cases, of questionable clinical relevance. Hence, although some of the results and proposed mechanisms could be contributing, our results cannot be to depression in the initial clinical stages of AD extrapolated to patients with AD and fully developed depression. Second, structural correlates of depression were found only in the group of prodromal AD. Then, it should be assumed that depression is different or has different determinants in patients with AD and more advanced cognitive deterioration. Third, subjects with subjective cognitive impairment (SCI), a transitional stage between NC and MCI, which is usually associated with depression [46], were not specifically analyzed. Future studies involving SCI subjects that present depressive symptoms and are positive for AD biomarkers should be particularly useful to confirm the model of genesis of depressive symptoms proposed herein (Fig.  3). Fourth, the midbrain monoaminergic nuclei [10] could not be specifically investigated, since those small nuclei fall out of 3T MRI resolution. And fifth, a potential role of non-AD specific factors (e.g., sex, personality, genetic factors, vascular lesions) could not be adequately analyzed, due to small sample size [47] or lack of vascular pathology [48].

CONCLUSION
The relationship between AD and depression is complex and, probably, bidirectional. In the initial clinical stages of the disease, mild depressive symptoms are explained by dysfunction in left cortical regions along with relative preservation of deficit awareness, provided by the right hemisphere. This model of genesis of depressive symptoms may contribute to more severe depression when other genetic and social circumstances concur and is also applicable in other neurological conditions, different from AD.

FUNDING
The ELMO and the DEMCAM received support from the Spanish Ministry of Health and from the Reina Sofía Foundation.