In this study, we found an association between AD-like patterns of brain atrophy, quantified by the SPARE-AD index, and plasma cortisol, CgA, IGFBP-2 and MIP-1α levels. Increased cortisol levels showed a significant association with worse processing speed and short term memory scores. In addition, we found that SPARE-AD correlates with t-tau/Aβ42 ratio and that MCI and AD patients with a negative AD CSF signature have lower (more normal) SPARE-AD values.
We have previously described an association between plasma cortisol levels and PIB PET scores in human subjects 
. This previous result is in agreement with the results from several animal model studies that have found that chronic stress and chronic increased hypothalamic-pituitary-adrenal axis activity increase Aβ deposition and tau hyperphosphorylation not only in transgenic animal models 
, but also in wild type mice 
and in-vitro models 
. Although, some of the studies have described that the effect is only mediated by corticotropin-releasing factor and not by corticosterone 
other studies have described an effect of glucocorticoids 
. Some studies indicate that cortisol might independently cause brain atrophy by other pathways that are independent of the amyloid pathway; glucocorticoids have widespread effects on cells through rapid and delayed effects, which are results of no-genomic, indirect genomic and direct genomics effects 
and increased cortisol levels affect dendritic spine development and reduce neurogenesis 
. In addition, glucocorticoids and chronic stress modify glutamate transmission in hippocampus and prefrontal cortex 
and impair long term potentiation 
. This has been linked to impairment in prefrontal cortex and hippocampus dependent cognitive processes in cognitively normal subjects 
One previous study has reported the association between cortisol and MRI changes, describing temporal atrophy with increasing cortisol levels 
. We found a more widespread pattern of cortisol correlation with brain atrophy that includes dorsolateral prefrontalcortex, vmPFC and insular cortex, as well as cuneus and precuneus in addition to the medial temporal lobe. The dorsolateral prefrontal atrophy and the processing impairment we describe are in line with previous results in animal models that show changes in prefrontal areas and impairments in behavioral tests associated with these areas 
. Previous studies in aged controls and hypertensive cognitively normal controls had described impairment in prefrontal cortex dependent tasks, but did not include MRI analyses describing the presence of prefrontal cortex atrophy 
. Lesion in vmPFC impair the retention of extinction learning in animal models 
and studies in human subjects have also shown an association with fear extinction and vmPFC volume 
. Chronic exposure to glucocorticoids in animal models leads to an impaired extinction in conditional fear conditioning and has been implicated in altered glutamatergic signaling in vmPFC 
. We did not have any measures of fear conditioning in our cohort, but our results of atrophy in vmPFC associated to high cortisol levels link cortisol to atrophy in regions related to fear conditioning in human subjects and confirm results described in animal models. As detailed before glucocorticoids regulate dendritic spine development and modulate gene expression resulting in changes that are independent of Aβ deposition and glucocorticoid levels are modulated by chronic stress, therefore part of the associations we described might be related to these factors. Chronic stress and post-traumatic stress disorder studies have described atrophy affecting medial prefrontal cortex, anterior cingulate, right insula and hippocampus 
. These areas overlap partially with our findings. Therefore, it might be possible that chronic stress has an additional deleterious effect and can be a coincident disease, which is a common finding in aged population of demented subjects 
or act as an additional risk factor for dementia.
Recent GWAS studies have found susceptibility-linked SNP expressed in immune cells and related to immune response 
and microglial activation is a common finding in AD neuropathological studies and is involved in the pathological changes 
. MIP-1α is a ligand for chemokine-CC-motif-receptor 5 (CCR5), a transmembrane-domain, G protein-coupled receptor belonging to the β-chemokine receptor family, which is implicated in the migration of in the migration of monocytes, NK, dendritic and Th1 cells. A previous study has shown that T lymphocytes of AD patients overexpress MIP-1α and the expression of this cytokine is associated with increased transendothelial migration in the endothelial permeability assay and decreased integrity of the human brain microvascular endothelial cell monolayer 
. The injection of Aβ into ratś hippocampus increased the expression of CCR5 and MIP-1α in endothelial cells and peripheral T cells, respectively 
. The same researcher described how RAGE was associated to the increased expression of endothelial CCR5, which is mediated by the PI3K and JNK signaling cascade 
. There are also results that indicate that MIP-1 might be also implicated in the dendritic cell transendothelial migration 
. Our study further is the first to describe an association with MRI atrophy in human subjects and adds evidence to the involvement of this molecule in Alzheimeŕs disease.
Higher IGFBP-2 levels were associated with increased brain atrophy. This protein has been shown to modulate the actions of IGF-1; high levels result in an inhibition of IGF-1 dependent signaling while low levels increase the signaling 
. IGF-1signals downstream through insulin receptor substrate 2 (IRS-2) and this pathway is associated with the prenatal and postnatal brain growth 
. In addition, reduced IRS-2 signaling has been associated with accumulation of phosphorylated tau 
and non-diabetic AD patients have shown IGF-1 resistance associated with IRS-2 dysfunction 
. Plasma IGFBP-2 levels correlated with CSF t-tau levels (r
0.040) and showed a trend with p-tau levels (r
0.075). Therefore, higher IGFBP-2 levels could further impair the already affected IGF-1 signaling in AD cases and results in increased atrophy.
Chromogranins are prohormones, which are the major constituents of secretory large dense-core vesicles. CgA inhibits nicotinic acid transmission and is involved in the regulation of secretory granules. In AD CgA can be found in neuritic plaques and has been implicated in microglial activation 
. One previous study found low CgA in CSF of patients with early AD 
IGFBP-2 and CgA levels were associated with t-tau levels this indicates that their levels might be related to neuronal injury. However, none of these analytes showed a correlation with atrophy in specific brain regions. This might be to the fact that SPARE-AD summarizes the whole pattern of atrophy and increase the power to detect associations due to the avoidance of the multivariate comparisons and adjustments that that are needed for whole brain comparisons.
Only cortisol levels were associated with cognitive measures. There are several factors that could account for the absence of an association between the other plasma biomarkers and the cognitive scores. It is possible the cognitive deficits and the biomarkers take place at different time points, that they show different floor and ceiling effects or that SPARE-AD captures nonlinear relations and measurements from different regions.
Previous studies measuring plasma or serum analytes using similar multi-analyte profiling platforms have been mainly focused on classifying AD patients vs. cognitively normal controls 
. They have shown a lower 
or similar classification accuracy 
. than CSF Aβ42
, t- and p-tau measurements, indicating that plasma analyte panels might be useful for screening patients and as measure of disease progression 
. Only one previous study has studied the association between plasma analytes and the classical CSF biomarkers (CSF Aβ42
, t- and p-tau) finding a different set of analytes 
. These differences are not unexpected because CSF biomarkers and MRI atrophy behave differently during disease progression 
and MRI captures changes that might be independent of amyloid and tau deposition as we have proposed.
SPARE-AD also correlated with the CSF T-tau/Aβ42
ratio, which discriminates between AD patients and CN individuals 
, distinguishes patients with AD from those with FTLD 
, and predicts future cognitive impairment 
. While repeated measurements of CSF Aβ as well as PIB PET Aβ plaque burden in AD patients have not been shown to change further with the progression of AD 
, t-tau CSF levels have been reported to increase as AD progresses over time 
thereby suggesting that the CSF T-tau/Aβ42
ratio could increase with the progression of AD and serve as a biomarker of increasing disease severity. The relationship of this ratio with SPARE-AD, which is a measure of brain atrophy that likely reflects the degree of AD neurodegeneration, supports this notion. Furthermore, the moderate correlation with the ADAS-Cog is in agreement with previous findings 
and consistent with the sensitivity of SPARE-AD to disease progression. The difference in SPARE-AD values in subjects who are positive or negative for the AD CSF signature reflects the ability of the SPARE-AD to identify specific atrophy patterns that are associated with AD pathology. However, the high values in MCI and AD subjects with a negative AD CSF signature indicate that the SPARE-AD may be less specific for other neurodegenerative disorders that cause brain atrophy patterns similar to those in AD. More specific biomarkers can be built by looking specifically at patient populations with different pathologies, as shown before 
, where subtle differences in brain atrophy patterns differentiated between AD and FTD patients.
In summary, our MRI analysis describes for the first time in humans the effect of MIP-1α on cortical areas and the implication in the pathogenesis of AD in agreement with previous results in experimental animals and offers new insight of the association between cortisol and brain atrophy and cognitive changes. The association of brain atrophy with MIP-1α adds further evidence for the importance of the immune response in AD pathogenesis. Last, we also describe an association between our recognition algorithm and CgA and IGFBP-2, which need further validation. These findings shed more light to the association between inflammation markers and AD-like neurodegeneration, a topic of high importance in the aging and AD literatures.