APOE4 expression alters astrocytic excitability
To determine whether expression of the allele APOE4 alters astrocyte excitability, we recorded Ca2+ in hippocampal slices of 9–12-week-old male and female mice in which the endogenous mouse APOE gene had been replaced with human APOE3 or APOE4 genes. Recordings were made in ACSF medium supplemented with 10% FBS, in order to keep lipoprotein and lipid concentration as close as possible to physiological conditions. Slices were incubated with the Ca2+ indicator fluo-4/AM and the astrocytic marker sulforhodamine (SR101) (Fig. 1a, see Materials and methods). We analyzed Ca2+ spontaneous activity—that is, Ca2+ events at rest conditions—and neurotransmitter-induced Ca2+ responses in cells co-labeled with Fluo-4/AM and SR101. To study spontaneous activity, we recorded APOE3 and APOE4 astrocytes for 120 s. For receptor-mediated Ca2+ responses, we recorded basal peak activity for 30 s, and then we stimulated slices with 1 mM ATP. We selected purinergic stimulation because it elicits Ca2+ signals both in astrocytes in vivo and in vitro, triggering several physiological functions [34, 55], and because responses elicited by stimulation of purinergic receptors are the cause of Ca2+ hyperactivity in AD mouse models [56].
In male mice, we detected increased amplitude of spontaneous Ca2+ transients in APOE4 vs APOE3 astrocytes (Fig. 1b). Regarding induced activities, purinergic stimulation caused in both genotypes the increase in magnitude and frequency of Ca2+ transients typically seen in astrocytes. The amplitude of Ca2+ responses was greater in APOE4 cells than in APOE3 cells (Fig. 1c and d). Further, the magnitude of Ca2+ responses decreased after removal of ATP (post ATP phase) to basal levels in APOE3 astrocytes, but remained significantly increased over its own basal levels, and with respect to APOE3 cells, in APOE4 cells (Fig. 1c and d). The frequency of Ca2+ responses was similar in both genotypes, in both the ATP and post-ATP phases. It is worth stressing that the amplitude of spontaneous Ca2+ transients was also statistically increased in APOE4 compared to APOE3 cells in these set of experiments, confirming the results of the 120-s recordings (Fig. 1c and d).
Female mice differed from males in two respects. First, the amplitude of spontaneous and ATP-induced events was significantly increased (p < 0.001), by at least 2-fold, in astrocytes from APOE3 females as compared to APOE3 males (compare Fig. 1b with 2a and 1d with 2c). Second, no differences were observed between APOE3 and APOE4 astrocytes in females in the magnitude of spontaneous and induced events (Fig. 2a-c). This may suggest that Ca2+ responses in astrocytes from APOE3 females represented the maximal Ca2+ response that could not increase further. Altogether, the ex vivo observations suggest that expression of human APOE alleles modulates Ca2+ transients in astrocytes in a sex-specific manner, such that expression of the APOE3 allele in male mice results in globally decreased Ca2+ transients in astrocytes—or expression of APOE3 in females in a global increase—as compared to the APOE4 allele.
Immortalized APOE4 astrocytes show increased Ca2+ mobilization from acidic stores
To gain insight into the mechanism by which expression of different APOE alleles regulates Ca2+ transients in astrocytes, we used immortalized astrocytes that express human APOE3 or APOE4 [44] since this in vitro model allows for experimental manipulations that are not feasible in brain slices. As with slices, cells were supplemented with 10% FBS, in order to keep lipoprotein and lipid concentration as close as possible to physiological conditions. It is worth stressing that immortalized astrocytes are aneuploid so the sexual identity is lost. Thus, a key question was whether their Ca2+ phenotype is male- or female-like. Our data show that they reproduce a male-like Ca2+ signaling phenotype in the presence of lipids. First, differences in Ca2+ responses at rest showed the same trend observed in male mice. APOE4 astrocytes had significantly different Ca2+ basal levels than APOE3 astrocytes (p-value = 0.01), the fluorescence ratio being 0.28 ± 0.01 and 0.41 ± 0.03, respectively (Fig. 3a-c). Note that cultured astrocytes do not show at rest the so-called spontaneous Ca2+ oscillations observed ex vivo, but stable basal Ca2+ levels that we could compare thanks to the ratiometric fura-2/AM Ca2+ indicator. Second, 100 μM ATP stimulation resulted in greater Ca2+ responses in APOE4 than in APOE3 astrocytes (Fig. 3a). Purinergic-induced Ca2+ responses also lasted longer: the response was 64.9 ± 8.2% of the maximum peak signal after 20 s in APOE4 cells but only 27.7 ± 1.5% in APOE3 astrocytes. Moreover, altered Ca2+ signaling was not restricted to purinergic stimulation, as adrenergic and muscarinic-receptor activation also triggered greater cytosolic Ca2+ responses in APOE4 than in APOE3 astrocytes (Fig. 3b and c). Importantly, the magnitude of purinergic-induced Ca2+ responses was the same with two other FBS batches (peak responses after stimulation with 100 μM ATP were 0.36 ± 0.13 in APOE3 and 0.94 ± 0.01 in APOE4; and 0.37 ± 0.04 in APOE3 and 0.91 ± 0.05 in APOE4 astrocytes). Taken together, the data support immortalized APOE3 and APOE4 astrocytes as a model to study the mechanisms underlying the regulation of Ca2+ responses by APOE alleles in males.
Next, in order to identify which pathways are dysregulated in APOE4 astrocytes, we examined Ca2+ fluxes among the principal intracellular Ca2+ sources with organelle-specific probes and pharmacological manipulations. First, we investigated the mitochondrial Ca2+ uptake that characteristically buffers increases in cytosolic Ca2+. Since APOE4 has been described as harming mitochondria in neurons [57], we reasoned that harmed mitochondria in APOE4 astrocytes could result in deficient Ca2+ uptake, and hence in increased intracellular Ca2+ responses. However, expression of the mitochondrial Ca2+ indicator CEPIA3mt (Fig. 3d) showed higher Ca2+ uptake in APOE4 mitochondria compared to APOE3 astrocytes, consistent with the higher ATP-induced Ca2+ responses in the cytosol (Fig. 3e). Second, we studied the main Ca2+ signaling pathway in astrocytes, Ca2+ mobilization from the ER through the IP3 receptor, by transfecting cells with G-CEPIA1er (Fig. 3d), and directly measuring Ca2+ contents inside this organelle. As expected, 100 μM ATP decreased Ca2+ levels in the ER of both APOE3 and APOE4 astrocytes, indicative of Ca2+ being released to the cytosol. Since the process is, although significantly, just slightly reinforced in APOE4 astrocytes (Fig. 3f), the greater cytosolic Ca2+ responses in these cells could not rely exclusively on increased Ca2+ mobilization from the ER. Third, we explored Ca2+ mobilization from acidic stores, which are mainly lysosomes and related organelles [58] that we have shown to be involved in purinergic-induced Ca2+ responses in astrocytes [37]. Figure 4a shows the main Ca2+ fluxes in lysosomes. We recorded cytosolic Ca2+ with fura-2/AM, after inhibiting Ca2+ release from acidic stores with 100 μM of Ned-19, an inhibitor of NAADP receptors responsible for Ca2+ release from these organelles [59]. Control cells were treated with DMSO, the vehicle of Ned-19. As expected, Ned-19 reduced ATP-induced Ca2+ responses in APOE3 cells (Fig. 4b), confirming the contribution of lysosomal Ca2+ to cytosolic transients [37]. Interestingly, Ned-19 greatly reduced Ca2+ responses in APOE4 astrocytes, such that purinergic-mediated Ca2+ responses in the presence of Ned-19 were of similar magnitude in both cell types. Hence, Ca2+ release from acidic-stores appears to be responsible for the greater purinergic-elicited Ca2+ responses in APOE4 astrocytes. Altogether, our results thus far lend support to the idea that lysosomal-related Ca2+ release is the main Ca2+ signaling pathway dysregulated in APOE4 astrocytes.
Dysregulation of V-ATPase activity contributes to the alteration of Ca2+ responses in APOE4 astrocytes
Ca2+ is released from acidic stores upon stimulation of NAADP receptors, which are Ca2+ channels, the most accepted candidates being two-pore channels 1 and 2 (Tpc1, Tpc2), and transient receptor potential mucolipin (Trpml) [59]. By contrast, Ca2+ uptake by acidic stores is accomplished through an indirect mechanism whereby V-ATPase pumps H+ into the vesicles, and then H+ are exchanged with Ca2+ through the Ca2+/H+ exchanger (CAX) [60] (Fig. 4a). Thus, greater Ca2+ release from acidic stores in APOE4 astrocytes might be due to two phenomena: greater expression or activation of NAADP receptors, or greater Ca2+ stored in these organelles, due to increased activity of V-ATPase and/or CAX. First, we analyzed the expression of NAADP receptors in astrocytes with real time qPCR. Expression of Trpml and Tpc1 channels is the same in APOE3 and APOE4 cells, while expression of Tpc2 channel is lower in APOE4 cells (Fig. 4c), perhaps as a result of a negative feedback mechanism in response to the high Ca2+ signals in APOE4 astrocytes. In any event, the increased release of Ca2+ from APOE4 lysosomes cannot be attributed to increased expression of NAADP receptors.
Second, we analyzed Ca2+ contents inside acidic stores with a protocol in which they were osmotically lysed due to the accumulation of the peptide Glycyl-L-phenylalanine 2-naphthylamide (GPN), followed by its proteolysis by cathepsin C, such that the stored lysosomal Ca2+ is released into the cytoplasm. Then, the area under the curve (AUC) of the cytosolic Ca2+ increase after addition of GPN to the astrocytes was used to calculate the total amount of stored calcium in these organelles. The results showed greater AUC in APOE4 cells (Fig. 4d), supporting a higher concentration of lysosomal-related Ca2+ in APOE4 cells, as compared to APOE3. Since there are no pharmacological modulators of CAX, we relied on a pharmacological inhibitor of V-ATPase, bafilomycin A1, and on measurements of pH in acidic organelles, to establish the implication of the pump and CAX in the increased luminal Ca2+. Bafilomycin A1 increased basal cytoplasmic Ca2+ concentration in APOE4 astrocytes to 0.40 ± 0.05 (control values of astrocytes treated with vehicle were 0.30 ± 0.01; p-value = 0.05), whereas there was no effect on non-stimulated intracellular Ca2+ levels in APOE3 (0.44 ± 0.08 for cells treated with bafilomycin A1 compared to 0.39 ± 0.02 of cells treated with vehicle). This suggests that the aforementioned lower basal cytoplasmic Ca2+ concentration in APOE4 cells is due to greater V-ATPase-mediated Ca2+ uptake into acidic stores, consistent with the greater intralysosomal Ca2+ levels revealed by the GPN experiments. Accordingly, bafilomycin A1 reduced the ATP-induced Ca2+ release in APOE4 astrocytes, confirming the dependence of these Ca2+ responses on acidic stores (Fig. 4e). In contrast, bafilomycin A1 did not change purinergic-induced Ca2+ responses in APOE3 astrocytes. Thus, both basal levels and ATP responses in the presence of bafilomycin A1 might indicate the existence of low V-ATPase activity in APOE3 astrocytes. Indeed, the expression of V-ATPase subunit V0D1, which is highly expressed in astrocytes [61], was higher in APOE4 cells than in APOE3 cells (Fig. 4f). Because V-ATPase controls the flow of H+ into acidic stores, we analyzed the pH of acidic stores with the probe lysosensor and fluorometry. We found that both APOE3 and APOE4 astrocytes had standard lysosomal-related pH, with a minimal but significant difference of 0.14 pH units between genotypes, APOE4 lysosomes being more alkaline (Fig. 4g). If the pH does not decrease despite increased activity of the V-ATPase pump, it follows that CAX activity must be increased in APOE4 astrocytes, too, such that the increased number of H+ entering the acidic stores exit through CAX in exchange for Ca2+. As a result, lysosomal Ca2+ concentration and hence basal Ca2+ and purinergic-induced Ca2+ release from lysosomes, are greater in APOE4 than in APOE3 astrocytes. Note that we did not determine nor manipulate CAX expression in APOE3 and APOE4 cells because the sequence of mouse CAX is unknown.
APOE3 expression in APOE4 astrocytes reduces cytosolic Ca2+ responses
Is the effect of APOE4 due to: (1) loss-of-function due to decreased contents [62], (2) malfunction, (3) gain-of-toxic function because of its structure, or (4) misplaced intracellular localization? We first examined ApoE levels in immortalized APOE3 and APOE4 astrocytes by quantifying mRNA levels with qPCR. As shown in Fig. 5a, APOE expression is lower in APOE4 than in APOE3 astrocytes. Therefore, it is plausible that lower expression of APOE4 accounts for Ca2+ signaling alterations. To test this possibility, we modulated the quantity of APOE3 and APOE4 with the rationale that, if APOE expression matters, decreasing ApoE in APOE3 cells would increase Ca2+ responses, whereas increasing ApoE in APOE4 cells would decrease Ca2+ responses. Upon decreasing APOE expression in APOE3 astrocytes with an APOE siRNA (Fig. 5b), ATP-induced Ca2+ responses were still lower, as compared to APOE4 astrocytes, with no differences observed between scramble- and siRNA-transfected APOE3 cells (Fig. 5c). We also tried the opposite strategy: we increased APOE expression using GFP-APOE3 or GFP-APOE4 plasmids, which overexpress a GFP-ApoE fusion protein, which then allows one to identify cells with increased APOE expression (Fig. 5d). The GFP signal did not interfere with fura-2/AM (data not shown). Because transfection efficiencies were different among experiments (5 to 50% of GFP transfected cells), we avoided inter-experimental variability by normalizing peak Ca2+ responses to levels recorded in APOE3 astrocytes transfected with a GFP-expressing plasmid without APOE constructs. The over-expression of GFP-APOE3, but not of GFP-APOE4, transformed APOE4 into APOE3 astrocytes, in a statistically significant manner (Fig. 5e), suggesting that Ca2+ hyperactivity in APOE4 astrocytes is mainly due to the expression of this particular isoform, but not to lower APOE expression. The data do not support, however, a gain of a toxic function by APOE4, because, if this were the case, APOE4-associated dysfunction would have been potentiated by GFP-APOE4 over-expression, and would not have been rescued by GFP-APOE3, since the toxic element, APOE4, remained. In agreement, there were no differences between ATP-induced Ca2+ signals in APOE3 cells transfected with GFP-APOE4 or with GFP (Fig. 5f), again ruling out a toxic effect of APOE4. Taken together, our data support APOE4 malfunction rather than toxicity.
Finally, because acidic stores accounted for the differences in Ca2+ responses between genotypes, we studied whether ApoE was localized in the acidic stores. Because the quality of staining of such organelles with the fluorophore lysotracker was not optimal for confocal studies (data not shown), we resorted to using Lamp1, a marker for lysosomes, autophagosomes, and different vesicles of the endolysosomal pathway, including late-endosomes [63]. In APOE3 astrocytes, we used ApoE and Lamp-1 immunostaining (Fig. 5g). In APOE4 astrocytes, we relied on APOE4 overexpression with GFP-APOE4, since the low amounts of APOE expression in APOE4 cells precluded immunocytochemical analysis (Fig. 5h). We observed that neither ApoE3 nor ApoE4 from GFP-APOE4 colocalized with Lamp-1. These observations rule out the likelihood that the alterations in lysosomal-related Ca2+ handling in APOE4 astrocytes are due to loss of a direct interaction of ApoE with channels mediating Ca2+ fluxes, suggesting, instead, indirect actions of the apolipoprotein, perhaps through changes in lipid homeostasis.
ATP-induced Ca2+ responses are modulated by extracellular lipids in APOE3 but not APOE4 astrocytes
Considering the wealth of evidence documenting how channels and pumps involved in Ca2+ signaling are regulated by lipids, particularly by membrane lipids [64, 65], we posited that lipids regulate Ca2+ fluxes in astrocytes, too, and that aberrant lipid homeostasis accounts for the observed dysregulation of Ca2+ fluxes in APOE4 astrocytes, particularly in lysosome-related organelles. Lipid modulation of Ca2+ transients in astrocytes is uncharted territory. Thus, to gain insight into the control of astrocyte excitability by lipids in the context of APOE4, we carried out two sets of experiments. First, we aimed to obtain proof-of-concept that lipids regulate Ca2+ transients in astrocytes, by examining the effect of changing lipid contents on the excitability of immortalized APOE3 and APOE4 astrocytes. Second, we analyzed cholesterol subcellular distribution, and we performed a lipidomic analysis of lysosomal and whole-membrane to identify candidates for excitability-modulating lipids, and changes thereof in the two genotypes.
Thus far, all our Ca2+ imaging experiments had been performed in medium supplemented with FBS, rich in nutrients, including lipoproteins. Here, we replaced this medium with three different media with lower lipid composition 2–5 min prior to Ca2+ imaging: 1) Krebs medium (KH), 2) DMEM supplemented with lipoproteins-deficient FBS and 3) DMEM with B27, a supplement without lipoproteins and restricted composition of lipids: linoleic acid, linolenic acid, progesterone, and corticosteroids. In the absence of lipids, non-stimulated basal Ca2+ levels were significantly lower in immortalized APOE4 vs APOE3 astrocytes, as detected in the presence of lipids. Thus, in KH, basal levels were 0.29 ± 0.02 in APOE3 vs 0.23 ± 0.03 in APOE4 (p-value = 0.04); in DMEM supplemented with lipoprotein-deficient FBS, 0.39 ± 0.07 in APOE3 vs 0.29 ± 0.06 in APOE4 (p-value = 0.04); and in DMEM with B27, 0.30 ± 0.02 in APOE3 vs 0.22 ± 0.01 in APOE4 (p-value = 0.007). By contrast, in all three conditions, stimulation of purinergic responses elicited Ca2+ responses of similar magnitude in APOE3 and APOE4 astrocytes (Fig. 6a-c, f). It was not the case that responses in APOE4 cells had diminished, but, rather, that responses had increased in APOE3 cells. That is, APOE4 astrocytes present the same magnitude of ATP-induced Ca2+ signals regardless of the lipids presents, whereas APOE3 astrocytes adapt their Ca2+ responses to the concentration of extracellular lipids, with responses being low in media rich in lipids (presence of FBS) and high in the presence of low lipid concentration, or no lipids. Consistent with the greater ATP-triggered Ca2+ responses, the responses also lasted longer in APOE3 astrocytes in the absence of lipids: the response was 36.0 ± 4.0% of the maximum peak signal after 20 s in KH, 51.6 ± 4.3% in DMEM supplemented with lipoproteins-deficient FBS and 34.6 ± 3.0% in DMEM with B27, as compared to the previously reported 27.7 ± 1.5% in DMEM with FBS (p < 0.05). Thus, the results point to a potentiation of long-lasting amplification Ca2+ pathways, perhaps by Ca2+-induced Ca2+ release from the ER or store-operated Ca2+ entry (SOCE)—that is, Ca2+ influx across the plasma membrane in response to depletion of intracellular Ca2+ stores. Still, ATP responses were significantly shorter (p-value < 0.01) in lipid-depleted APOE3 astrocytes as compared to APOE4 astrocytes (72.57 ± 3%, 68.4 ± 6.2% and 81.2 ± 6.8% for APOE4 cells kept in KH, DMEM with lipoprotein-deficient FBS and DMEM with B27), indicating that some differences in Ca2+ fluxes remain between APOE3 and APOE4 astrocytes.
It is worth stressing that the increased ATP-induced Ca2+ response in APOE3 astrocytes is a fast-onset process, as it was observed just a few min after lipoprotein removal. It is not transient, because it persisted 12 h after replacement of DMEM supplemented with FBS for DMEM supplemented with B27 (Fig. 6d and f), and it is reversible, because it diminishes if the medium is replaced again by DMEM supplemented with FBS 5 min prior to Ca2+ recordings (Fig. 6e and f).
Altogether, three conclusions may be drawn from these results: they constitute the first demonstration that astrocyte excitability is modulated by lipids and/or lipoproteins, such regulation is reversible but stable in APOE3 astrocytes as long as lipids remain present, and is lost in APOE4 astrocytes.
The absence of lipids potentiates Ca2+ release from the ER and extracellular Ca2+ entry in APOE3 but not APOE4 astrocytes
Since the core of Ca2+-signaling dysregulation in APOE4 astrocytes lies in acidic organelles, is Ca2+ homeostasis in these organelles the target of extracellular lipids in APOE3 astrocytes? The finding that the increase in ATP-induced Ca2+ signals in APOE3 cells in the absence of lipids/lipoproteins (KH media) was abrogated by Ned-19, the inhibitor of Ca2+ release from acidic stores (Fig. 7a) might indicate that extracellular lipids decrease Ca2+ levels inside such organelles. Surprisingly, measurement of Ca2+ stored inside acidic organelles using the GPN-elicited depletion showed no increased Ca2+ loading in APOE3 cells kept in KH (Fig. 7b) compared to APOE3 astrocytes kept in the presence of FBS, whereas, as expected, no difference was observed between the two conditions in APOE4 astrocytes (Fig. 7c). An explanation is that, in the absence of extracellular lipids, the release of Ca2+ from lysosomes in APOE3 astrocytes is amplified by the activation of other Ca2+ signaling mechanisms. In fact, the amplification originates in part, from Ca2+ released from the ER, because the Ca2+ content in the ER was 1.5-fold greater in APOE3 cells kept in KH than in DMEM/FBS, as recorded with the Ca2+ probe CEPIA1er (Fig. 7d). The same amount of Ca2+ was released from the ER in APOE4 astrocytes kept in the two media, in agreement with the observation that Ca2+ signals are independent of extracellular lipid concentrations in this genotype (Fig. 7d). Another source of Ca2+ is the extracellular Ca2+ entry, because the blockage of extracellular Ca2+ with the cell-impermeable Ca2+ chelator EGTA greatly reduced (by 3.3 times) ATP-induced Ca2+ responses in APOE3 astrocytes kept in KH without lipoproteins, but not in APOE3 astrocytes kept in KH with FBS (Fig. 7e). In contrast, EGTA reduced ATP-induced Ca2+ responses in APOE4 astrocytes to the same extent in the presence or absence of FBS (Fig. 7f). Thus, the data support the idea that extracellular Ca2+ entry is secondary to lysosomal Ca2+ release, and that it is modulated by extracellular lipids in APOE3 but not in APOE4 astrocytes. We next sought morphological support for this idea by studying the localization of acidic stores inside astrocytes. Again, we resorted to Lamp1 immunostaining to be able to use confocal microscopy. Lysosome distribution was abnormal in APOE4 cells (Fig. 7g); that is, a greater number of Lamp1-positive organelles accumulated near the nucleus in APOE4 astrocytes than in APOE3 astrocytes. Specifically, 37% of lysosomes are placed at 10 μm from the center of the nucleus in APOE4 cells, as compared to 21% in APOE3 astrocytes. Plausibly, the altered localization may result in changes in the coupling of such organelles with plasma-membrane channels.
In summary, in APOE3 astrocytes, activation of extracellular Ca2+ entry secondary to intracellular Ca2+ mobilization underlies the greater Ca2+ responses induced by purinergic receptors in the absence of extracellular lipoproteins. This coordination of signaling pathways does not take place in the presence of lipoproteins, suggesting, again, that lipids have the capacity to change intracellular Ca2+ fluxes in APOE3 astrocytes. By contrast, APOE4 astrocytes present a higher content of lysosomal Ca2+, but appear to have lost the capacity to have Ca2+ fluxes regulated by lipids.
Lipidomics reveals distinct lipid composition in lysosomal and whole-cell membranes from APOE3 and APOE4 astrocytes
We posited that lysosomal dysregulation and the refractoriness to lipid-based modulation in APOE4 astrocytes might be caused by altered lipid trafficking and homeostasis due to APOE4 malfunction—as concluded in a previous section. Since ApoE is a major cholesterol carrier in the brain, we studied cellular cholesterol distribution with filipin staining. We found aberrant intracellular distribution of cholesterol in APOE4 astrocytes, which presented more cholesterol in intracellular clumps, and less in plasma membrane, than APOE3 cells (Fig. 8a). This finding points to impaired cholesterol efflux in APOE4 cells. We also carried untargeted lipidomics because we reasoned that impaired lipid trafficking would leave its mark on astrocyte membranes, such that the profiling of membrane lipids would provide information about lipid dyshomeostasis in APOE4 astrocytes. To determine whether APOE4-mediated changes were specific to lysosomes, we performed lipidomics in lysosomal and whole-membranes of APOE3 and APOE4 immortalized astrocytes, since, according to a lipid map of the mammalian cell, organelles present distinct lipid compositions [66]. The multivariate analysis PLS-DA revealed that the lipids of lysosomal membranes (Fig. 8b) and whole-membranes (Fig. 8c) clustered independently in APOE3 and APOE4 genotypes. The predictive accuracy of the analysis was robust, as the Q2 and R2Y scores were 0.616 and 0.986 for lysosome lipidome, and 0.841 and 0.995 for whole-membranes. The PLS-DA analysis thus confirms that APOE genotype influences membrane lipid composition in astrocytes.
In order to identify which lipids contributed more to the differential group clustering, we used the Variable Importance in the Projection (VIP), such that lipids with VIP > 1 were the ones with greater weight on the group change. In lysosomal membranes, there were 35 lipids with VIP > 1 in the APOE4 vs APOE3. ANOVA analysis with a Tukey correction post-test of the intensity (peak values) of these 35 metabolites revealed significant differences due to the APOE phenotype with a p-value < 0.05, confirming, again, that expression of APOE4 alters the lipidome of lysosomes. We then proceeded to identification of the particular metabolites and calculation of their fold change (FC) in APOE4 vs APOE3 astrocytes. We could identify 19 lipids: 10 phosphatidylcholines (6 increased, 4 decreased), 5 phosphatidylethanolamine (2 increased, 3 decreased), 2 lysophosphatidylcholine (2 decreased), 1 lysophosphatidylethanolamine (decreased), and 1 carnitine (increased) (Fig. 8d). However, multi t-test analysis corrected by a false discovery rate (FDR) of these 19 FC values gave no statistically significant differences (q-value < 0.05). This suggests that joint changes in the contents of lipids with VIP > 1 rather than particular lipids account for the segregation of lipidomes from APOE3 and APOE4 lysosomes.
In whole-membranes, 41 metabolites had a VIP > 1, comparing APOE4 vs APOE3 astrocytes. ANOVA analysis with a Tukey correction post-test of the intensities of these 41 metabolites confirmed significant differences due to the APOE phenotype with a p-value < 0.01, in agreement with the previous PLS-DA analysis. Twety-one lipids were identified according to their m/z, and their FC in APOE4 vs APOE3 calculated (Fig. 8e). A multi T-test statistical analysis corrected with FDR, showed that 11 of these metabolites were significantly different (q-value < 0.05): 3 lysophosphatidylcholine (1 increased, 2 decreased), 5 phosphatidylcholines (3 increased, 2 decreased), 1 phosphatidylserine (decreased) and 2 carnitines (increased). In short, a general trend is that carnitines and phosphatidylcholines are more abundant in APOE4 astrocytes, whereas APOE3 cells are richer in lysophospholipids.
Overall, this is the first demonstration that the expression of APOE4 changes the lysosomal and cellular lipidomes in astrocytes, supporting a link between altered Ca2+ fluxes and lipid dyshomeostasis. It is worth noting that the different intracellular distribution of cholesterol in APOE3 and APOE4 astrocytes is not due to differences in cholesterol contents between the two genotypes, as the VIP for cholesterol was consistently lower than 1 in the lipidomes (data not shown).