Introduction

ApoE functions as a component of plasma lipoproteins in the transport of lipids (e.g., cholesterol) [1, 2]. Recent studies suggested a relationship between ApoE and vitamin D by indirect synthesis or reabsorption mechanisms. These studies indicated a protective effect of ApoEε4 against vitamin D deficiency [3], higher levels of vitamin D in ApoEε4 carriers [4], and an association of vitamin D with chylomicrons or lipoproteins [5]. In addition, vitamin D is a secosteroid hormone that consists of a broken cholesterol backbone. The synthetic pathway of the precursor of vitamin D is highly conserved and is the same as that of cholesterol [6]. These data might indicate the transport of vitamin D via ApoE chylomicrons or lipoprotein particles in brain or circulating blood. If such suggestions are supported, we should be able to discern a relationship between ApoE alleles and vitamin D levels in AD just as we observe such a relationship between ApoE and cholesterol in cardiovascular disease. Thus, our aim in this study was (1) to investigate the serum levels of 25OHD in patients with AD, MCI, and in age-matched healthy controls; (2) to compare the 25OHD levels in these patients according to ApoEɛ4 allele status to investigate any correlation of ApoE-dependent genetic vulnerability to AD and circulating vitamin D levels; (3) to investigate the correlation of serum cytokine and serum 25OHD levels in regard to the ApoEɛ4 allele status to determine any potential consequences of a vitamin D-ApoE relationship on immune response.

Methods

The patients and control groups

The patients, including AD or MCI, recruited to study at the Behavioral and Movement Disorders Unit of the Istanbul Faculty of Medicine at Istanbul University. AD patients were clinically diagnosed according to DSM-IV criteria. No patient with advanced disease stage was included. Clinical Dementia Ratings (CDR) Scale scores of AD patients were 1 and 2. Thus, none of the patients were bedridden or had any significant disability. All MCI patients had mild cognitive symptoms and had a global score of 0.5 in the CDR scale [7]. Three patient groups were established according to the age and disease status after the exclusion of patients with inflammatory diseases, autoimmune disease, infectious or psychiatric diseases, non-Alzheimer’s dementia (other than MCI), chronic heart disease, patients taking antibiotics or non-steroidal anti-inflammatory drugs, and those with significant laboratory abnormalities. The patients included in the study had erythrocyte sedimentation rates within the reference values. The healthy controls were free from any neurodegenerative disorders. A neuropsychological assessment using the mini mental status examination (MMSE) was performed for the patients with AD and MCI and for all controls. The demographics of each group are given in Table 1.

Table 1 The demographics of the patient and control groups
Full size table

DNA isolation and ApoE genotyping

DNA was extracted from peripheral blood samples using a QIAamp DNA Mini Kit (Cat.No.51304, QIAGEN, Germany). Genotypes of ApoE were determined by real-time polymerase chain reaction (RT-PCR) using a LightMix® Kit ApoE-C112R-R158C (Cat.No.40-0445-16, TibMolBiol, Germany) in a LightCycler® 480 InstrumentII (Roche Diagnostics, Germany) as previously described [8]. Melting curve analyses were performed for genotyping.

CLIA and ELISA assays

The quantitative determination of 25OHD in serum samples was performed by the chemiluminescent immunoassay (CLIA) technology using the DiaSorin Liaison 25(OH)D TOTAL assay (DiaSorin, USA) in a Liaison analyzer. ELISA was used to measure the levels of BDNF, Hsp90, CFH, TNFα, IL-1α, IL-1β, IL-6, IL-10, and α2M in the serum samples, as described in our previous studies [9, 10].

Statistical analysis

The distribution of ApoE alleles in regard to the presence of the ε4 allele in the patient groups was analyzed by Chi-square (χ 2) test using SPSS 21.0 software. All data are expressed as the mean ± the standard deviation (SD). p values less than 0.05 were considered statistically significant.

Each 25OHD test run included case–control sets, two blinded quality control (QC) specimens from our study as well as two inter assay controls and six calibrators. The test detection range was 4–150 ng/ml. The intra-assay coefficient variation was 8.51 % for low controls and 9.93 % for high controls. Both the low and high controls were in the range of the calibrators in each run. 25OHD values were season-adjusted according to the cosinor model of Sachs et al. [11]. No significant difference was observed between the adjusted and unadjusted 25OHD values; thus, all comparisons were made with the adjusted values.

The raw data for each group were analyzed using GraphPad InStat DTCG 3.06. Data are presented as the mean ± SD in the text and figure legends. Group comparisons of patients with LOAD and MCI with the age-matched controls were performed by one-way ANOVA (more than 2 groups) when appropriate and by the Kruskal–Wallis test (non-parametric ANOVA) when the data were not normally distributed. Group comparisons between patients with EOAD and age-matched controls were generated by unpaired t test (with two groups), with a Welch correction (when required) or by Mann–Whitney U test when the data were not normally distributed. The comparisons were also made with respect to the presence of the ApoEɛ4 allele. Any significant result for multiple parameters within group comparisons was analyzed again with Dunn’s multiple comparisons test or with the Tukey–Kramer multiple comparisons test to avoid type I errors; in addition, p values are given under the graphs. Multiple regression analysis was used to compare the significant contribution of the serum 25OHD levels to the constant parameters (serum BDNF, Hsp90, CFH, TNFα, IL-1α, IL-1β, IL-6, IL-10, and α2M levels [9, 10]). If any positive data were obtained from the multiple regression analysis, the analysis was repeated with Pearson’s correlation coefficient to keep most of the data available, given that multiple regression analysis excludes any patients with missing data. The correlations of 25OHD with serum BDNF, Hsp90, CFH, TNFα, IL-1α, IL-1β, IL-6, IL-10, and α2M levels were also assessed with respect to the presence of the ApoEɛ4 allele. The Bonferroni correction, which lowers the significance level (α c value) to avoid a type I error, was performed for the p values of Table 3; the data were considered significant at p < α c. Retrospective power analysis was also performed.

Results

ApoE allele frequencies

The frequency of the ApoEε4 allele in groups was as follows: 89 % for ε4 non-carriers, 11 % for ε4 carriers out of all of the healthy controls; 79 % for ε4 non-carriers, 21 % for ε4 carriers in all patients with AD. The frequency of ApoEε4 allele carriers out of all AD patients was significantly higher than that of all healthy controls (p = 0.01; 95 % CI: 1.2–3.7; OR 2.1). The numbers of ApoE ε2 allele carriers were not sufficient for statistical comparison, given that these individuals were evaluated in the ε4 non-carrier group.

Serum 25OHD levels

No significant difference was observed between the unadjusted and the season-adjusted 25OHD levels in each group (p > 0.05). Given that all of the comparisons were generated with season-adjusted 25OHD levels, henceforth, all ‘25OHD levels’ reflect the adjusted values.

Comparison of the unadjusted and season-adjusted levels of serum 25OHD between the age-matched control groups and the disease groups was given in Table 2. Multiple comparisons of the season-adjusted levels of serum 25OHD between groups and the same analysis depending on the ApoEɛ4 allele presence were given in Fig. 1a, b. Retrospective power analysis of adjusted 25OHD levels of AllAD vs all control indicated the power of the study to be 83.1 %.

Table 2 Comparison of the unadjusted and season-adjusted levels of serum 25OHD between the age-matched control groups and the disease groups
Full size table
Fig. 1
figure 1

a Multiple comparisons of the season-adjusted levels of serum 25OHD between groups. Serum 25OHD levels of LOAD patients were significantly lower than that of all healthy controls (*p < 0.01) and MCI patients († p < 0.05). No difference was observed between EOAD patients and other groups in terms of 25OHD levels (p > 0.05) (Kruskal–Wallis with Dunn’s multiple comparisons). If these groups were compared again with the appropriate age-matched controls, the serum 25OHD levels were nearly significantly decreased in the EOAD group compared with the age-matched controls (p = 0.05). The serum 25OHD levels of the LOAD group were significantly lower than the levels in the corresponding age-matched controls (p < 0.001). No significant difference was observed in the 25OHD levels between patients with MCI and the age-matched control group (p > 0.05). The serum 25OHD levels were significantly decreased in all AD patients (LOAD and EOAD) compared with all healthy subjects (p < 0.0001) (Table 2). All unadjusted and season-adjusted serum 25OHD comparisons are given in greater detail in Table 2. b Multiple comparisons of the season-adjusted levels of serum 25OHD between groups depending on ApoEε4 allele status. The serum 25OHD levels of ε4 non-carrier LOAD patients were significantly lower than those of ε4 non-carrier healthy subjects and MCI patients (*p < 0.001, p < 0.01, respectively, Kruskal–Wallis with Dunn’s multiple comparisons). No significant difference was observed in the serum 25OHD levels between ε4 carrier healthy subjects and any of the patient groups (p = 0.68). The serum 25OHD levels of ε4 non-carrier LOAD patients were significantly lower than those of ε4 carrier LOAD patients (*p = 0.03). A similar difference was not observed within any other group. The numbers of ApoEε4 carriers were given in Table 1

Full size image

Correlation analysis of 25OHD with age, age of onset and MMSE with respect to ApoEɛ4 allele status

No correlation was found between the serum 25OHD levels and age or age of onset in any of the healthy controls or in patients with MCI or LOAD (p > 0.05). On the contrary, positive correlations were found between the serum 25OHD levels and age (r = 0.48, 95 % CI 0.06–0.75, p = 0.03) as well as age of onset (r = 0.67, 95 % CI 0.33–0.85, p = 0.0009) in the EOAD group. No significant difference was observed between the serum 25OHD levels and MMSE scores in LOAD patients or in healthy subjects, whereas positive correlations were observed between the serum 25OHD levels and MMSE scores in the EOAD (r = 0.51, 95 % CI 0.1–0.77, p = 0.01) and MCI groups (r = 0.38, 95 % CI 0.01–0.66, p = 0.04).

No correlation was found between the serum 25OHD levels and age, age of onset or MMSE scores of the ε4 non-carriers, the ε4 carrier healthy subjects, the MCI patients or the LOAD patients (p > 0.05). No correlation was observed between the serum 25OHD levels and age or MMSE scores of the ε4 non-carriers or the ε4 carrier EOAD patients (p > 0.05), but positive correlations were observed between the serum 25OHD levels and age of onset in both the ε4 non-carriers and the ε4 carrier EOAD patients (r = 0.57, 95 % CI 0.07–0.85, p = 0.03; r = 0.83, 95 % CI 0.20–0.97, p = 0.02, respectively).

Correlation analysis of 25OHD with certain serum parameters with respect to ApoEɛ4 allele status

We investigated the correlation of serum 25OHD levels and certain serum parameters (BDNF, HSP90, CFH, TNFα, IL-1α, IL-1β, IL-6, IL-10, α2M), which were investigated in our previous studies [9, 10], with respect to the presence of the ApoEɛ4 allele. The correlation analysis of these parameters was performed only in individuals who were included in our previous studies. The numbers of individuals who were included in the correlation analysis were as follows: all healthy subjects, nmax:47; MCI patients, nmax:29; LOAD patients, nmax:52; EOAD patients, nmax:21.

Correlation analysis of 25OHD with BDNF or Hsp90 or immune response components (CFH, TNFα, IL-1α, IL-1β, IL-6, IL-10, and α2M) with respect to ApoEɛ4 allele status

Correlation analysis of 25OHD with specified parameters with respect to ApoEɛ4 allele status was given in Table 3. The levels of serum parameters with respect to the presence of the ApoEε4 allele were also assessed for each group (Fig. 2).

Table 3 Correlation analysis of 25OHD with serum parameters with respect to ApoEε4 allele status
Full size table
Fig. 2
figure 2

Serum levels of immune response components, BDNF and HSP90 with respect to ApoEɛ4 allele status. a All healthy controls: serum α2M levels of ApoEɛ4 carrier healthy controls (n:11, 95 % CI 2804.3–3458.3) were significantly higher than those of ApoEɛ4 non-carrier healthy controls (n:36, 95 % CI 2804.3–3458.3) (*p = 0.04, Mann–Whitney U). b MCI patients: serum BDNF levels of ApoEɛ4 carrier MCI patients (n:9, 95 % CI 978.9–2082.2) were significantly higher than those of ApoEɛ4 non-carrier MCI patients (n:20, 95 % CI 701.8–1249.1) (*p = 0.019, Mann–Whitney U). c LOAD patients: serum TNFα and IL-10 levels of ApoEɛ4 carrier LOAD patients (n:17, 95 % CI 17.6–25.1; n:17, 95 % CI 4.6–6.3, respectively) were significantly higher than those of ApoEɛ4 non-carrier LOAD patients (n:35, 95 % CI 16.1–18.2; n:35, 95 % CI 3.6–4.5, respectively) (*p = 0.03 Unpaired T test with Welch, p = 0.004, Mann–Whitney U, respectively). d EOAD patients: no significant difference was seen between the ApoEɛ4 carrier EOAD patients and the non-carrier patients in terms of any of the above-mentioned parameters

Full size image

Discussion

The mechanisms of ApoE in AD are not fully understood. Likewise, whether a deficiency in vitamin D is a casual factor or is the result of the disease is a matter of debate. Surprisingly, recent studies have indicated the presence of an evolutionary juncture of these two factors, which encompasses latitude-UVB radiation, the immune system, and dietary habits [3, 4].

Data from our cohort of Turkish EOAD and LOAD patients supported the suggestion that 25OHD deficiency is frequent among patients with cognitive disorders [12–15]. In regard to the healthy subjects, we should note that several Turkish studies presented mean 25OHD values [16–18] that were similar to those in our current study, but the 25OHD levels in the Turkish population seem to be lower than the mean levels in other populations [13, 19]. Our study indicated that although EOAD and LOAD patients had lower 25OHD levels than the corresponding age-matched healthy controls, the MCI patients had similar values of circulating 25OHD as the healthy subjects and both have significantly higher MMSE scores than EOAD or LOAD patients. Furthermore, increased serum 25OHD has the potential to positively influence MMSE scores given that they are correlated in MCI and EOAD patients, but not in LOAD patients.

It has been hypothesized that the ApoEε4 allele protects against a deficiency in vitamin D in circumstances of high geographical latitude, dark skin or insufficient dietary supplementation where vitamin D synthesis is compromised [3, 4]. Higher levels of vitamin D were indicated in ApoEε4 carrier healthy subjects compared with non-carrier ones [4], and the higher 25OHD level in ɛ4 allele carriers was associated with higher memory functions [20]. This may be interpreted as the capability of these individuals to better endure low UV concentrations, which may be an advantage of the ancestral ApoEɛ4 allele compared with the evolutionally young ɛ3 allele [21]. This concept is known as the UVB radiation hypothesis of ApoE [3, 21].

The vitamin D status was also assessed in our AD patients with respect to the presence of the ApoEε4 allele. First, if the patient groups were compared with all healthy subjects independent of age discrimination, only LOAD patients had significantly low levels of 25OHD. No significant difference was observed between the serum 25OHD levels in any of the patient groups and healthy subjects who carried at least one ApoEε4 allele. Surprisingly, the 25OHD status of LOAD patients indicated that the significantly low levels seen in LOAD originated from ε4 non-carrier patients and not from ε4 carriers. At first glance, our study supports the suggestions by Huebbe [4] and Maddock [20] studies, which showed high levels of 25OHD in ε4 allele carriers. However, they demonstrated this alteration in healthy subjects, whereas we demonstrated it in LOAD patients. The differences between the 25OHD levels in the ApoEε4 allele carriers and non-carrier healthy subjects of the Maddock and Huebbe studies and those in our study might be explained by the differences in sample number or the geographical locations of the populations.

From a different point of view, our findings might also be interpreted as low levels of 25OHD in ApoEε4 allele non-carriers. This finding presents an opportunity to claim vitamin D deficiency to be a casual factor for individuals with the sporadic AD phenotype who have no ε4 allele. The biological relevance of this relationship is unclear. Yet, there are a limited number of studies indicating a possible mechanistic link between vitamin D and ApoE: ApoE-deficient mice had lower 1,25(OH)2D3 levels and VDR silencing in endothelial cells resulted in decreased ApoE expression [22]. ApoE suggested to be an upstream regulator of vitamin D metabolism via Lrp2 (megalin) which is a known vitamin D transporter and a receptor for ApoE-lipoproteins [23], or via VDBP [4]. ApoE functions in the transport of cholesterol [1, 2]. Basically, vitamin D and cholesterol originate from the same molecule (7-dehydrocholesterol), and they differ from each other only by the presence of two hydrogens [24, 25]. In addition, serum 25OHD levels are positively correlated both with ApoA-I and LpA-I [26] and the total cholesterol level after UVB radiation [25] and negatively correlated with the ratio of LDL to HDL [26]. In addition to this possible relation, one can speculate that the levels of a certain molecule may also depend on transport mechanisms: a system may exist for the simultaneous transport of vitamin D and cholesterol via ApoE- bearing lipoproteins in the CNS or plasma. This suggestion is relevant considering the limited (5 % or less) ability of 25OHD [27, 28] and cholesterol [29] to cross the blood brain barrier. If we assume that vitamin D and ApoE coexist in the same cellular mechanisms or pathways, then we have to consider that either the effect of vitamin D or the effect of ApoE, for example, on the immune response might require each other. ApoEɛ4 has been reported to be associated with an inflammatory response [30], and vitamin D has been consistently known to display anti-inflammatory effects [31]. In PBMCs, ApoE secretion is restricted to CD14+ monocytes [32] and CD14 was suggested as the most suitable marker for the description of the vitamin D status of blood samples [33]. ApoE promotes the conversion of macrophages from the proinflammatory M1 to the anti-inflammatory M2 phenotype [34]. Vitamin D can stimulate the macrophages in AD patients, so that they function in amyloid beta clearance, and as a result of this response, vitamin D stimulates cytokines, which contrasts with its usual anti-inflammatory role [35]. ApoE affects the levels of cytokines, but TNF-α and IL-1β can suppress ApoE production [36]. Due to these overlapping effects of vitamin D and ApoE on the immune system, we also investigated the correlation of serum 25OHD levels and several serum parameters (BDNF, HSP90, CFH, TNFα, IL-1α, IL-1β, IL-6, IL-10, α2M), which were investigated in our previous studies [9, 10], with respect to the presence of the ApoEɛ4 allele. We have not observed any study that has focused on 25OHD and cytokines in AD patients.

In this study, our results revealed a common pattern in groups with relatively higher 25OHD levels (healthy subjects or MCI patients) and supported the negative effect of vitamin D on the expression of TNFα, IL-1β and IL-6; however, the effects were no longer significant after Bonferroni adjustment. The negative correlation of serum 25OHD levels and IL-1β in healthy subjects originated from the ApoEɛ4 non-carrier subjects. The correlations of serum 25OHD and either TNFα or IL-6 were independent from the ApoE allele status in healthy subjects and MCI patients. These negative correlation patterns between the serum levels of 25OHD and the serum levels of TNFα, IL-1β or IL-6 were altered in groups with low 25OHD levels (EOAD and LOAD). Intriguingly, the negative correlations became positive between 25OHD, and all three parameters are in ɛ4 carrier LOAD patients. Thus, even though this is speculative, higher vitamin D levels in ApoEɛ4 carriers might be one way to compensate for the lack of an anti-inflammatory response. In addition, vitamin D might induce cytokines to suppress the production of ApoE only in ApoEɛ4 carriers to prevent the elevation of amyloid beta accumulation in the brain or other parts of the body [37] given the higher affinity of ApoEɛ4 for amyloid plaques. This suggestion may explain why we see a positive correlation of serum 25OHD levels and serum TNF-α, IL-1β, IL-6, and IL-10 levels only in ApoEɛ4 carrier LOAD patients but not in healthy controls or MCI patients.

The reason that the EOAD group did not present such correlations or any difference between 25OHD levels with respect to the presence of the ɛ4 allele might be due to the effect of age on the immune system. This may also be explained by different genetic backgrounds, the different roles of vitamin D over different periods of time throughout life [15] or the low number of EOAD samples. Our results also indicate that the correlations between 25OHD levels and age, age of onset or MMSE scores in EOAD patients, and the correlations between 25OHD levels and MMSE scores in MCI patients were independent of the ApoE allele status.

The fundamental finding of this study was that LOAD patients had very low levels of vitamin D, but that these low levels originated from only the ApoEɛ4 non-carrier patients. This finding was the first observation for ApoE and vitamin D relation in AD. Thus, the vitamin D status should be monitored in AD patients who are non-carriers of the ApoEɛ4 allele to confirm our data.

Vitamin D has existed on earth for more than 750 billion years and affects many cellular mechanisms [38–40]. Moreover, given its shared synthesis pathway with cholesterol, it would be more logical to reconsider and combine the ApoE perspective with that of vitamin D deficiency in AD.