We evaluated the plasma levels of 24S-OH-Chol in a sample of aging patients affected by two different forms of dementia, LOAD and VD, and in individuals with CIND. The first interesting data emerging from our study is the finding of higher circulating levels of 24S-OH-Chol in patients with LOAD compared with Controls. Our results seem to confirm the unique observation by Lutjohann et al. , while are in contrast with following studies reporting decreased levels of 24S-OH-Chol in LOAD patients [13–15]. Unlike that study , we were not able to demonstrate in LOAD patients a significant correlation between 24S-OH-Chol levels and the global cognitive function (MMSE) or functional impairment (Barthel index), perhaps because our subjects were not in the very-early stage of the disease (Global Deterioration Scale: stage 3 to 5). Our data support the hypothesis of a trend toward higher levels of plasma 24S-OH-Chol in the early stages of LOAD, when the rate of neurodegeneration is higher than normal, but the amount of cell loss and resulting brain atrophy is still small. The precise mechanisms leading to increased plasma 24S-OH-Chol levels in the early stages of LOAD are not known. It is possible that, in these specific conditions, cholesterol turnover might be increased in CNS (due to neuronal degradation) and/or that the rate of conversion of brain cholesterol to 24S-OH-Chol might be higher compared to normal.
Interestingly, it has been observed by Brown et al. that 24S-OH-Chol is an efficient inhibitor of beta-amyloid formation in vitro ; if this is true also under in vivo conditions, the increase of 24S-OH-Chol levels in CNS (and consequently in the circulation) might be considered as an early attempt to counteract beta-amyloid deposition. On the contrary, in the advanced stages of the disease low levels of 24S-OH-Chol in CNS (and consequently in the circulation), would even accelerate beta-amyloid deposition and the progression of LOAD.
In part, increased 24S-OH-Chol levels might also result from a defect in blood-brain barrier, which seems to be a frequent finding in different neurological diseases including LOAD .
Unlike the study of Lutjohann et al. , but in agreement with other following clinical observations [13–15], we found that plasma 24S-OH-Chol levels were lower in VD patients compared with controls. The difference we observed between LOAD and VD is somehow unexpected, and might be related to different mechanisms. In particular, it has to be underlined that, on average, our VD patients were in a more advanced stage of the disease (GDS: stage 4-6) compared with LOAD (GDS: stage 3-5). Actually, not only it has been reported that a decrease in plasma levels of 24S-OH-Chol would be typical of dementia [25, 8], but it has been also shown that plasma 24S-OH-Chol levels progressively decrease with the worsening of the disease . Thus, the differences between LOAD and VD might be secondary to a different stage of disease. Alternatively, the differences in levels of 24S-OH-Chol between LOAD and VD might reflect the different pathogenetic mechanisms and/or evolution of these two forms of dementia.
We also evaluated the possible relationship between 24S-OH-Chol and other available variables. Interestingly, by multivariate regression analysis we found that the 24S-OH-Chol/TC ratio independently correlated with hs.CRP levels, and this explained about 10% of its total variability. Hs.CRP is a sensitive marker of systemic inflammation. The association between the 24S-OH-Chol and inflammation has been already reported in vitro ; indeed, Alexandrov found that in primary co-culture of human neurons and glia, 24S-OH-Chol is able to induce the expression of several pro-inflammatory genes. The presence of a low-grade systemic inflammation has been reported both in LOAD and VD by several Authors, including our group . The finding of a significant association between 24S-OH-Chol and CRP levels suggest a possible link between the degree of neurodegeneration (plasma 24S-OH-Chol would be a marker of it) and the degree of peripheral systemic inflammation. As a matter of fact, the relationship between 24S-OH-Chol/TC and CRP was strong in LOAD (r: 0.39), was still present in CIND (r: 0.20), but it was pratically absent in VD patients (r: 0.08). The specificity of the relationship between LOAD and increased 24S-OH-Chol plasma concentrations might be indirectly supported by the result of univariate analysis (Table 2) showing a significant correlation of 24S-OH-Chol with memory tests impairment and brain atrophy on CT scan, both typical characteristics of LOAD but not of VD.
Finally, we evaluated the possible effect of the PPARgamma Pro12Ala polymorphism on the risk of being affected by dementia or CIND, as well as on 24S-OH-Chol plasma levels. Infact, it has been consistently reported that PPARgamma plays an important role in glucose and lipid metabolism , which in turn have been associated with LOAD [28, 29]. We found no association between the PPARgamma polymorphism and LOAD, VD or CIND. Furthermore, unlike Sauder  we didn't find any increase in 24S-OH-Chol/TC ratio among the carriers of the Ala allele both in the whole sample and in the three groups of patients.
The principal limitations of the study must be also acknowledged. First, brain morphology was evaluated by qualitative CT scan assessment and not by quantitative MRI analysis, which is the best validated method and would give important information. Second, CSF biomarkers were not available for LOAD, VD and CIND patients enrolled into this study, and this might significantly influence sensibility and specificity of our diagnoses. Third, we did not systematically evaluated Apo E polymorphism in our sample, and it is probable that apo E phenotype might influence plasma levels of 24S-OH-Chol. Nevertheless, we evaluated apo E genotype in DNA from 70% of patients (84/120); we found no differences in mean/median 24S-OH-Chol levels by comparing patients bearing or not the ε4 allele (ANOVA - Mann-Whitney test p: 0.86 and 0.96, respectively).
Fourth, although we investigated plasma 24S-OH-Chol concentrations in a much larger sample of subjects compared with previous studies, sample size might be small when investigating the possible effect of genetics. For this reason our negative data on PPARgamma polymorphism need to be replicated in larger groups of individuals.