Study cohort demographics and clinical characteristics

The demographic and clinical characteristics of subjects included in the parent cohort have been published elsewhere [39, 40, 44, 45]. Characteristics specific to the now included subjects are summarized in Table 1. Briefly, controls were 5 years older compared to MCI-ADD (p < 0.001) and ADD patients (p < 0.001), and as expected, the controls exhibited the highest MMSE test scores, the highest CSF Aβ42, Aβ40 levels, and the lowest t-tau and p-tau levels (Table 1). As expected, controls exhibited a higher ratio of CSF Aβ42 over Aβ40 (Aβ42/Aβ40) compared to MCI-ADD (p = 0.001) and ADD (p < 0.001) patients. Among the controls, the frequency of the APOE ε4 allele (41%) was significantly lower compared to the MCI-MCI (63%), MCI-ADD (75%) and ADD groups (86%) (chi-square, p = 0.001) (Table 1). Based on a recent assessment of the parent cohort [40], CSF AD biomarker cut-off levels (630 pg/mL for Αβ42 [47], 66 pg/mL for p-tau and 394 pg/mL for t-tau) were established to allow for classification of the included subjects according to the A/T/N classification system (Table 2). Levels of CSF α-synuclein and KLK6 levels in both plasma and CSF have been described elsewhere [15, 43,44,45] and used only for correlation analyses in the current study.

Table 1 Demographics and clinical characteristicsTable 2 A/T/N classificationPlasma and CSF NfL and YKL-40 levels

Plasma NfL and YKL-40 and CSF YKL-40 levels are presented in Table 3, whereas CSF NfL levels were previously reported [46]. Globally, CSF NfL levels were nearly 60 times higher than in the plasma and the levels in both compartments positively correlated (Spearman’s (ρ) = 0.323, p = 0.002, n = 92) and also significantly associated with age (plasma NfL: Spearman (ρ) = 0.220, p = 0.015, n = 123, CSF NfL: Spearman’s (ρ) = 0.229, p = 0.026, n = 94). Similarly, plasma and CSF levels of YKL-40 were positively correlated (Spearman’s (ρ) = 0.404, p < 0.001, n = 120), but only CSF YKL-40 was significantly associated with age (plasma YKL-40; Spearman’s (ρ) = 0.150, p = 0.094, n = 125, and CSF YKL-40: Pearson’s (r) = 0.378, p < 0.001, n = 120). Neither plasma nor CSF NfL and YKL-40 levels were influenced by APOE ε4 status (plasma NfL: Kruskal–Wallis, p = 0.575, Plasma YKL-40: ANOVA, p = 0.794, CSF NfL: ANOVA, p = 0.175, CSF YKL-40: ANOVA, p = 0.855).

Table 3 Plasma and CSF NfL and YKL-40 levels

Accounting for the significant age difference between sexes (females: 65.0 ± 4.8, males: 67.6 ± 6.9, Student’s t-test, p = 0.014) as well as between the controls and the patients, CSF and plasma NfL as well as CSF YKL-40 group comparisons were performed with age as a covariant (ANCOVA, or Quade nonparametric ANCOVA). Between the sexes, levels of YKL-40 in plasma (Mann–Whitney U test, p = 0.656) and CSF (ANCOVA, p = 0.631), as well as NfL in plasma (Quade nonparametric ANCOVA, p = 0.290), were not different; however, males exhibited 6% higher CSF NfL levels compared to females (ANCOVA, p < 0.001).

Between the diagnostic groups, baseline ADD patients exhibited 1.2-fold higher plasma NfL levels compared to controls or MCI-MCI patients (Table 3). Comparison between controls, stable MCI, and ADD (baseline and MCI-ADD) revealed the same results (Quade nonparametric ANCOVA, p = 0.006, post hoc with Bonferroni adjustment for n = 3, ADD vs controls: p = 0.042, ADD vs stable MCI: p = 0.042). Levels of YKL-40 in plasma and CSF did not differ between the diagnostic groups (plasma: Kruskal–Wallis, p = 0.927, CSF: ANCOVA, p = 0.127) (Table 3).

ApoE phenotype confirmation

In all the samples, the apoE phenotype was assessed based on the presence or absence of endogenous variants specific to each APOE genotype (Supplementary Table 2), as previously described [38]. The acquired apoE phenotypes (Table 4) were compared to the previously assessed APOE genotype and were in 100% accordance.

Table 4 ApoE phenotype determined by mass spectrometry APOE ε4 associated with lower plasma apoE levels

Previous studies have repeatedly shown that presence of the ε4 allele is linked to lower plasma levels of apoE [20, 21]. Comparing levels of plasma total apoE across the five APOE genotypes present in the current study we also recorded APOE genotype-specific effects on plasma apoE levels (Kruskal–Wallis, p = 0.004). As shown in Fig. 1a, levels of plasma total apoE were the highest in APOE ε2/ε3 subjects and the lowest in individuals with the ε4/ε4 genotype. In more detail, plasma apoE levels in APOE ε4/ε4 carriers were 30% and 56% lower compared to individuals with the ε3/ε3 and ε2/ε3 genotype (Fig. 1a). Plasma apoE levels were directly associated with APOE ε4 allele dose (zero, one or two copies) (Kruskal–Wallis, p = 0.006) and plasma apoE in heterozygous and homozygous individuals were 16% versus 33% lower compared to ε4 non-carriers (Fig. 1b). The latter difference remained significant after accounting for multiple comparisons (n = 3).

Fig. 1

Levels of plasma apoE in subjects with different APOE genotypes. Plasma apoE levels as assessed in subjects grouped based on their APOE genotype (a), APOE ε4 status (b), and in males and females with different APOE genotype (c). Data are shown as median (minimum–maximum). Group comparisons were done using the Kruskal–Wallis test followed by Dunn’s test (a, b) before/after Bonferroni correction for multiple comparisons or ANOVA with Tukey HSD as post hoc test (c)

Plasma total apoE levels and effects of sex

With female sex as a strong risk factor for AD [48, 49], we examined whether sex was associated with variations in plasma apoE levels. Due to low sample numbers, we excluded females with the APOE ε2/ε3 (n = 1) and ε2/ε4 (n = 2) genotype, as well as males with the ε2/ε4 (n = 1) genotype from the group comparisons. Levels of total plasma apoE were different across females with APOE ε3/ε3 (n = 27), ε3/ε4 (n = 23), and ε4/ε4 (n = 18) (ANOVA, p = 0.009), with ε4/ε4 females exhibiting the lowest levels. Among males with the APOE ε2/ε3 (n = 4), ε3/ε3 (n = 13), ε3/ε4 (n = 24) and ε4/ε4 (n = 12) genotype, we also observed differences in the levels of plasma apoE (ANOVA, p = 0.038), specifically between males with the APOE ε2/ε3 and ε3/ε4 genotype (Fig. 1c). Although levels of total plasma apoE varied between females with different APOE genotypes, and similarly between males with different APOE genotypes, only female APOE ε3/ε4 subjects exhibited significantly different plasma apoE levels compared to their male counterparts. Specifically, APOE ε3/ε4 females exhibited higher plasma total apoE levels (females (n = 23) 55.9 ± 25.5 μg/mL, males (n = 24) 39.2 ± 19.1 μg/mL, p = 0.014, Student’s t-test), attributed to an increase in the levels of the apoE4 isoform (females (n = 23) 20.7 ± 8.2 μg/mL versus males (n = 24) 14.0 ± 5.5 μg/mL, p = 0.002, Student’s t-test). Importantly, only in females, plasma apoE was significantly associated with age (Spearman’s (ρ) = 0.349, p = 0.003, n = 71).

Plasma apoE isoform composition in APOE heterozygotes

With varying total apoE plasma levels between APOE genotypes we aimed to determine the contribution of the individual apoE isoforms to the total plasma apoE levels in APOE heterozygotes. In subjects with the APOE ε2/ε3 and APOE ε2/ε4 genotype, plasma total apoE consisted predominantly of the apoE2 isoform (62 ± 5% and 75 ± 2%, respectively) (Fig. 2a) with significantly different isoform levels in the APOE ε2/ε3 subjects (Fig. 2b). In subjects with the APOE ε3/ε4 genotype, the levels of apoE4 were nearly 30% lower than the apoE3 levels (Fig. 2a, b).

Fig. 2

Total plasma apoE isoform distribution in APOE heterozygous individuals. Percentage (%) (a) and actual concentrations (b) of apoE2, apoE3, and apoE4 isoforms of total plasma apoE in APOE ε2/ε3, APOE ε2/ε4, and APOE ε3/ε4 subjects. Data are presented as average (a) or median (minimum–maximum) (b). p-values for APOE ε2/ε3 (black dots for apoE2, black triangles for apoE3), for APOE ε2/ε4 (black dots for apoE2, black squares for apoE4) and APOE ε3/ε4 (black triangles for apoE3, black squares for apoE4) were acquired using the Student’s t-test

Plasma apoE levels by diagnostic group and A/T/N classification

Plasma total apoE levels varied significantly between the diagnostic groups (Fig. 3a, ANOVA, p = 0.002). In detail, plasma total apoE levels were 1.5 times lower in MCI-ADD and ADD patients compared stable MCI patients respectively (Fig. 3a) with the difference remaining when accounting for APOE genotype (Wald Chi-Square p = 0.022). When combined, the prodromal and baseline ADD patients (n = 56) had 25% and 23% lower total plasma apoE levels compared to controls (ANOVA, p < 0.001, Tukey HSD post hoc, p = 0.012, n = 39) and stable MCI (Tukey HSD, p = 0.001, n = 30) respectively. When accounting for APOE genotype, a significant difference remained only between baseline ADD patients and stable MCI (Wald Chi-Square p = 0.008, Bonferroni post hoc, p = 0.007).

Fig. 3

Plasma apoE levels per diagnostic group. Plasma apoE levels in controls, MCI-MCI, MCI-ADD and ADD patients (a) and in groups based on the A/T/N classification (b) and the Aβ1-42 status (c). Data is presented as median (minimum–maximum). Group differences were assessed using ANOVA (Tukey HSD post hoc) (a), the Kruskal–Wallis test followed by the Dunn’s test uncorrected/corrected for multiple comparisons using Bonferroni correction (b), or Mann–Whitney U test (c). Star marked p-values obtained after accounting for the APOE genotype of the studied subjects. The A-/T-/N + and A + /T + /N- groups were excluded from the statistical analysis due to low n-numbers (n = 2, in each group)

Accounting for AD brain pathology rather than clinical diagnosis, we compared plasma apoE levels among subjects classified according to the A/T/N classification system [50]. Among the resulting eight groups (A-/T-/N-, A-/T-/N + , A-/T + /N-, A-/T + /N +, A+/T-/N-, A + /T-/N + , A + /T + /N-, A + /T + /N +), we found a significant difference in the total plasma apoE levels (Kruskal Wallis, p = 0.007). The recorded difference between the “all pathology positive” (A + /T + /N +) and the “all pathology negative” (A-/T-/N-) groups remained significant when accounting for multiple comparisons (n = 15), and specifically, the Aβ1-42 positive subjects (with CSF Aβ42 levels lower than 630 pg/mL [47]) classified as A + /T-/N- and A + /T + /N +) exhibited lower plasma apoE levels compared to the “all pathology negative” subjects (A-/T-/N-) group (Fig. 3b). Irrespective of APOE genotype, Aβ1-42 positive subjects had 28% lower total plasma apoE levels compared to Aβ1-42 negative subjects (Fig. 3c). Accounting for the APOE genotype, plasma apoE levels remained different between A-/T-/N- and A + /T + /N + (Fig. 3b), as well as between Aβ1-42 positive and negative individuals (Fig. 3c). Subjects with A-/T-/N + and A + /T + /N- were excluded from the analysis due to the low sample numbers (n = 2 per group).

Correlations between plasma apoE, apoE isoforms, cognition, and CSF markers

In the whole cohort, irrespective of diagnosis and APOE genotype, higher plasma apoE levels were significantly associated with higher MMSE scores levels of CSF Aβ42 and CSF Aβ42/Aβ40 and lower levels of CSF t-tau and p-tau, as well as CSF NfL levels (Table 5), whereas no associations were found between plasma apoE, plasma Nfl, plasma, and CSF YKL-40 nor CSF Aβ40, α-synuclein and KLK6 levels (data not shown). The association between Aβ42, Aβ42/Aβ40 and t-tau with plasma apoE levels remained when accounting for APOE genotype (Table 5).

Table 5 Correlations between plasma apoE levels, cognition, and markers in plasma and CSF

Stratifying our analysis based on diagnosis the association between plasma apoE and the Aβ42/Aβ40 ratio was eliminated; however, we observed that plasma apoE levels were negatively associated with plasma NfL only in controls (Table 5). Plasma apoE was furthermore positively associated with CSF Aβ42 only in MCI-MCI patients and negatively associated with CSF t-tau and p-tau only in the aMCI patients that over the 24 months study period received an ADD diagnosis (Table 5). Furthermore, accounting for diagnostic group we found significant negative correlations between plasma apoE and both CSF α-synuclein and KLK6 levels only in converting aMCI patients (Table 5), with the significant association with KLK6 levels remaining when taking APOE genotype into account. Grouping together the MCI-ADD and baseline ADD patients, the significant correlation between plasma apoE and CSF α-synuclein (Pearson’s (r) = -0.297, p = 0.029, n = 54) was eliminated when accounting for APOE genotype (Partial correlation r(51) = -0.257, p = 0.063). No statistically significant correlations between plasma levels of apoE and KLK6 or CSF and plasma YKL-40 levels were observed.

Lastly, we assessed whether specifically the apoE3 and apoE4 isoforms in APOE ε3/ε4 individuals were linked to MMSE scores, CSF AD biomarker levels, Aβ40, and the Aβ42/Aβ40 ratio, as well as CSF α-synuclein and both CSF and plasma levels of KLK6, YKL-40, and NfL. Among these different markers, only CSF α-synuclein and NfL levels exhibited a negative association with the apoE4 (α -synuclein: Pearson’s (r) =  − 0.294, p = 0.045, n = 47, NfL: Pearson’s (r) =  − 0.333, p = 0.041, n = 47), but not apoE3 (α -synuclein: Pearson’s (r) =  − 0.269, p = 0.067, n = 47, NfL: Pearson’s (r) =  − 0.315, p = 0.054, n = 47) isoform levels. In addition, in the APOE ε3/ε4 subjects, we identified a positive association between total plasma apoE levels with both apoE3 (Pearson’s (r) = 0.961, p < 0.001, n = 47) and apoE4 (Pearson’s (r) = 0.878, p < 0.001, n = 47) isoforms, which were further shown to be positively linked to each other (Spearman’s (ρ) = 0.724, p < 0.001, n = 47).