Immortalized astrocytic lines, which produce natively lipidated apoE lipoproteins were originally derived from the human APOE targeted replacement mice (Sullivan et al., 1997, 1998; Knouff et al., 1999). These lines are homozygous for ε2, ε3, or ε4 human APOE alleles. Development of immortalized astrocytic lines and characterization of apoE lipoproteins they produce have been described previously (Morikawa et al., 2005). Immortalized Apoe^–/– astrocytes derived from the Apoe^–/– mice (Zhang et al., 1992) were used to produce control conditioned media lacking human apoE lipoproteins. Cell culture media, supplements, and reagents used for isolation and culture of astrocytes and neurons were of Gibco™ brand and were purchased from Thermo Fisher Scientific (Waltham, MA, United States). Proteomic reagents were of Pierce™ brand and they also were purchased from Thermo Fisher Scientific. Kits and reagents for RNA extractions were obtained from Qiagen Sciences Inc. (Germantown, MD, United States). Recombinant human apoE was bought from Leinco Technologies (St. Louis, MO, United States). Primary and secondary antibodies were from different sources individually indicated in the text. All other chemicals and reagents were obtained either from Sigma-Aldrich (St. Louis, MO, United States) or from Thermo Fisher Scientific unless indicated otherwise.

Immortalized Apoe^–/– astrocytes and those homozygous for ε2, ε3, and ε4 human APOE alleles were maintained in the Dulbecco’s Modified Eagle Medium/Nutrient Mixture F-12 (DMEM/F-12) supplemented with 10% heat-inactivated fetal bovine serum (FBS), 1 mM sodium pyruvate, and 0.4% geneticin. To harvest the apoE lipoproteins, the astrocytes were plated in T-25 flasks and grown until they reach 75–85% confluence. Then the standard growth medium was replaced with serum free DMEM/F-12 supplemented with 1× N-2 supplement and 3 mM 25-hydroxycholesterol to enhance apoE production. The astrocytes were grown in the serum free DMEM/F-12 for 48 h. prior to harvesting the conditioned media containing apoE lipoproteins (Figure 1A).

Neat astrocytic media were 10-fold concentrated to minimize the amount of DMEM/F-12 introduced to the primary neuronal cultures. This was done by centrifugation of the media in Amicon^® Ultra-15 Centrifugal Filter Units with 10 kDa molecular weight cut off (MWCO) cellulose membranes (MilliporeSigma, Burlington, MA, United States) at 4,000 × g and 4°C (Figure 1A). To limit precipitation of apoE lipoproteins on the filter membrane the Amicon Filter Units were primed with 0.1 mg/ml solution of bovine serum albumin (BSA) in 0.01 M phosphate buffered saline (PBS) (Pankiewicz et al., 2021). The apoE level in 10-fold concentrated media was quantified using enzyme-linked immunosorbent assay (ELISA) (Pankiewicz et al., 2014) and the differences between batches were balanced out by adding 10-fold concentrated media from Apoe^–/– astrocytes (Figure 1A). Lipidation status of the apoE lipoprotein particles was examined by Tris-Glycine native gel electrophoresis followed by Western blotting (Figures 2A,B).

Dissociated cultures of primary hippocampal neurons were established from P0–P2 old pups of CD1 mice. Following rapid decapitation, the brains were extracted from the skulls and the hippocampi were dissected out in aseptic conditions (Figure 1B). Cell suspension was prepared by digesting the tissue with 30 units of Papain (Worthington Biochemical Corporation, Lakewood, NJ, United States) and 200 units of DNAse I (Worthington Biochemical Corporation) for 30 min at 37°C, triturating with fire-polished glass pipettes, and filtering through a 40 μm cell strainer (Kuszczyk et al., 2013). The neurons were counted and plated into 6-well plates in the amount of 50,000 per well for cytochemical analysis or in the amount of 500,000 per well for biochemical or RNA sequencing analyses. For the former type of analysis the neurons were seeded on poly-D-lysine coated, round, 18 mm in diameter coverslips (Ted Pella, Inc., Redding, CA, United States) placed on the bottom of the well. For the latter type of analysis, the neurons were seeded directly on the poly-D-lysine coated well bottom. Primary hippocampal neurons were cultured in the Neurobasal medium supplemented with 2% B-27 supplement, 2 mmol/L glutamine, 50 mg/mL of streptomycin, and 100 U/mL penicillin. One-third of the medium volume was replaced every third day. On the third DIV, primary hippocampal neurons were treated with 5 μmol/L cytosine-β-D-arabinoside to prevent growth of non-neuronal cells. On the 10 DIV and again on the 13 DIV apoE-enriched, 10-fold concentrated astrocytic conditioned media were added to the Neurobasal medium (1:10 vol/vol). The experiment was concluded at 17 DIV when the primary neuronal cultures were fixed for cytochemical characterization or lysed for biochemical or transcriptomic analyses. The concentration of apoE in the neuronal conditioned media at the commencement of the experiment and the amount of apoE internalized by neurons at the conclusion of the experiment were quantified by ELISA and by polyacrylamide gel electrophoresis (PAGE) performed under reducing conditions, followed by Western blotting (Figures 2C–F). The control neurons were treated on 10 and 13 DIV with 10-fold concentrated conditioned media from immortalized Apoe^–/– astrocytes (1:10 vol/vol). The absence of apoE in the conditioned media from Apoe^–/– astrocytes was confirmed using apoE ELISA and Western blotting. Likewise, the absence of apoE in the conditioned media from the control neurons at the commencement of the experiment and in their lysates at the conclusion of the experiment was confirmed (Figure 2). Contamination of dissociated neuronal cultures by native astrocytes was monitored by immunostaining against glial fibrillary acidic protein (GFAP) using a rabbit polyclonal anti-GFAP antibody (1:2000; Dako/Agilent Technologies; Santa Clara, CA, United States) and Alexa Fluor 488 conjugated goat anti-rabbit IgG secondary antibody (1:500; Jackson ImmunoResearch Laboratories Inc.; West Grove, PA, United States) followed by a nuclear counterstaining with 4′,6-diamidino-2-phenylindole (DAPI) (Pankiewicz et al., 2020). The number GFAP-labeled astrocytes was then manually enumerated in test fields 1.067 mm × 1.334 mm. Ten randomly selected test fields per coverslip from three independently established cultures per treatment condition were analyzed at 17 DIV. The degree of native astrocytic contamination ranged around 5.5% of all cells in the culture and did not vary significantly across treatment conditions (Supplementary Figure 1).

Conditioned media from astrocytic or neuronal cultures were first treated with 0.5% sodium deoxycholate for 45 min at 37°C to delipidate apoE. Lysates of primary hippocampal neuronal cultures were prepared as previously described (Kuszczyk et al., 2013) and centrifuged at 10,000 × g for 20 min at 4°C to remove cellular debris. ApoE level was determined by sandwich ELISA utilizing HJ15.7 mouse monoclonal antibody (mAb) anti human apoE (Shi et al., 2017) as the capture antibody (0.1 μg/mL) (gift of D. M. Holtzman) and biotinylated goat anti-human apoE polyclonal antibody (1:5,000) (Meridian Life Science, Inc., Memphis, TN, United States) as the detection antibody (Pankiewicz et al., 2014). The color reaction was obtained using streptavidin conjugated horseradish peroxidase (HRP) (Fitzgerald Industries, Acton, MA, United States) and 3, 3′, 5, 5′- tetramethylbenzidine Peroxidase EIA Substrate Kit (Bio-Rad Laboratories, Inc., Hercules, CA, United States). Colorimetric ELISA readouts were converted to the actual apoE concentrations using standard curves prepared from increasing dilutions of recombinant human apoE. The level of apoE in the conditioned media was expressed in ng of apoE protein per mL while that in the cell lysate in ng of apoE protein per mg of the total protein, which was assessed by the Pierce BCA Protein Assay Kit following manufacturer provided instruction.

Western blots of gels run under native conditions were used to characterize apoE lipoproteins secreted by astrocytes to the conditioned media. Samples of 10-fold concentrated media were mixed with equal amounts of sample buffer containing Tris HCl 50 mM pH 6.8, 10% Glycerol, and 0.004% Coomassie blue G250 and immediately loaded on the pre-casted 4–20% Novex Tris-Glycine native gel and electrophoresed over 4 h at room temperature using Tris-Glycine Native Running Buffer. Following the electrophoresis, native gels were incubated with 1% SDS and 50 mM Tris pH 6.8 for 30 min, and electroblotted into nitrocellulose membranes over 3 h. The membranes were blocked with 5% soy milk diluted in the SuperBlock blocking buffer and incubated overnight at 4°C with polyclonal goat anti-human apoE antibody (1:7,000; Meridian Life Science, Inc.). ApoE-antibody complexes were visualized using HRP-linked donkey anti-goat IgG antibody (1:30,000; Santa Cruz Biotechnology Inc., Dallas, TX, United States) followed by SuperSignal West Pico PLUS Chemiluminescent Substrate (Thermo Fisher Scientific). Membrane autoradiographs were then obtained and digitized as previously described (Asuni et al., 2013, 2014; Pankiewicz et al., 2019). The resulting images were densitometrically analyzed using NIH ImageJ2 (FIJI) (Bethesda, MD, United States) (Schindelin et al., 2012).

Western blots of gels run under reducing conditions were used to characterize apoE levels in the neuronal conditioned media and neuronal lysates. Samples of each were mixed with equal amounts of tricine sample buffer and beta-mercaptoethanol, boiled at 95°C for 5 min., resolved using 12.5% SDS-PAGE protein gels, and transferred into nitrocellulose membranes. Detection and quantification of apoE in the nitrocellulose membranes was performed analogously to that described for Western blots run under native conditions.

Primary hippocampal neurons on the cover slips were thrice washed with PBS, fixed with 4% paraformaldehyde (PFA) in 0.1 M phosphate buffer (PB) pH 7.4 over 15 min., thrice washed with PBS again, and permeabilized with 0.25% Triton X-100 in PBS. Then they were washed with PBS containing 0.05% Tween-20 and sequentially blocked with 10% BSA, and avidin/biotin kit (Vector Laboratories; Burlingame, CA, United States). The neuronal cytoskeleton was immunostained using a mouse mAb against microtubule-associated protein 2 (MAP2) (1:500; Sigma-Aldrich) followed by anti-mouse IgG biotinylated secondary antibody (1:500; Vector Laboratories) followed by Cyanine 3 (Cy3) conjugated streptavidin (1:1,000). Immunostained neurons were photographed under 40× objective magnification using high-sensitivity, cooled, monochrome DS-Qi1Mc camera attached to 80i Nikon fluorescent microscope (Nikon Corp., Tokyo, Japan) (Kuszczyk et al., 2013). Digitized neuronal images were analyzed using NIH ImageJ2 (FIJI) (Bethesda, MD, United States) (Schindelin et al., 2012). The dendritic arbor was traced using the simple neurite tracer plugin for NIH ImageJ2 (Longair et al., 2011) and the lengths of all dendritic branches were summed to obtain the combined dendritic length of a neuron. In addition, we quantified the arbor surface area, created by joining the tips of the most far out reaching dendrites using the Image J freehand tool (Wang et al., 2019). Dendrite branching was examined by the Sholl’s analysis (Binley et al., 2014), which enumerates the cross points between dendritic branches and the set of concentric circles spaced 3 μm apart from the center of the neuronal perikaryon (Figure 3B). Twenty randomly selected neurons per culture from three independent cultures per treatment condition were analyzed for all three parameters.

Dendritic spines were labeled with a combination of 1, 1′-Dioctadecyl-3, 3, 3′, 3′-tetramethylindocarbocyanine perchlorate (DiI) and Phalloiding-iFluor 488 (Abcam Inc., Cambridge, MA, United States), which have complementary labeling affinity toward mature and immature spine forms, respectively. Dil solution was prepared by dissolving Dil crystals in dimethyl sulfoxide (DMSO) (2 μg/μl), and then diluting the DMSO based solution with Dulbecco’s phosphate-buffered saline (DPBS) 1:500. Neurons were first fixed with 2% PFA and then stained with Dil overnight at room temperature. On the following day, after copious washing with DPBS the neurons were stained again with 1:1,000 PBS solution of Phalloidin-iFluor 488 for 90 min at room temperature (Cheng et al., 2014). Dil/Phalloidin-iFluor 488 double labeled neurons were imaged using Zeiss 880 LSCM with ZEN Black 2.3 SP1 acquisition software. Z-stack of images were acquired through the entire thickness of a neuron with a capture speed of 5, 54% overlap between slices, and a pinhole diameter of 1 Airy unit. The Z-stacks were collapsed into two-dimensional images using the maximum intensity projection function. The background was subtracted with a rolling ball radius of 35 pixels, gamma was adjusted to 1.1, and the resolution of the pictures was proportionally increased by a factor of 5 both in X- and Y-axis using the Transform J plugin in FIJI software. The total number and subtype frequency of dendritic spines were analyzed along four 20-μm long segments randomly selected along secondary independent dendrites. Four spine subtypes were recognized based on their distinct morphology: filopodium, thin, stubby, and mushroom-shaped (Cheng et al., 2014) (Figures 4A,B). Ten randomly selected neurons per culture from four independent cultures per treatment condition were analyzed.

Primary hippocampal neurons were fixed 4% PFA and immunostained against postsynaptic density protein 95 (PSD95) with a mouse mAb (1:100; MilliporeSigma), NR1 subunit of the N-methyl-D-aspartate receptor (NMDAR) with a rabbit mAb (1:100; MilliporeSigma), or against synaptophysin with a mouse mAb (1:250; Santa Cruz Biotechnology, Dallas, TX, United States). Biotinylated anti-mouse or anti-rabbit secondary antibodies (both 1:500; Vector Laboratories) followed by Cy3 conjugated streptavidin (1:1,000) were used to visualize immunolabeled synaptic proteins. Neurons were imaged using Zeiss 880 laser scanning confocal microscope (LSCM) and ZEN Black 2.3 SP1 acquisition software (v. 14.0.18.201) (Carl Zeiss AG, Oberkochen, Germany) under the 63 × 1.4 N.A. objective. Z-stack of images were acquired through the entire thickness of the neuron with a capture speed of 7, 67% overlap between slices, and a pinhole diameter of 1 Airy unit. To quantify expression of synaptic proteins Z-stacks were collapsed into a two-dimensional image, which was converted into 8-bit format. Expression of synaptic proteins was quantified along secondary independent dendrites following our previously published protocols (Kuszczyk et al., 2013, 2014). Ten to 15 neurons per culture from four independent cultures per treatment condition were randomly chosen for analysis. Five rectangular test areas (20 μm × 10 μm) covering secondary independent dendrites were selected per neuron. Integrated density of the synaptic protein immunostaining within each test area was then measured using the FIJI software. The integrated density is defined as a product of area and mean gray value for a positive immunostaining within the test area.

Neuronal RNA was isolated using RNeasy Mini Kit, following manufacturer provided protocol, which included in column DNA digestion with 2 U of DNAse I. Processing of isolated RNA and its sequencing was carried out by the Genome Technology Center of the NYU Grossman School of Medicine. Purity and integrity of isolated RNA was first determined using 2100 Bioanalyzer (Agilent Technologies Inc., Santa Clara, CA, United States). RNA concentration was measured on NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific). Ribosomal RNA was depleted using the Ribo-Zero Gold kit (Illumina, Inc., San Diego, CA, United States). cDNA libraries were generated with Illumina Truseq kit with an input of 200 ng of RNA and 10 cycles. Sequencing was performed using the NovaSeq 6000 Sequencing System on an SP100 flowcell (Illumina, Inc.).

Analysis of RNA sequencing readouts was performed with the help of the Applied Bioinformatics Laboratory of the NYU Grossman School of Medicine. Sequencing reads were mapped to the reference genome (mm10) using the STAR aligner software (v2.5.0c) (Dobin et al., 2013). Alignments were guided by a Gene Transfer Format (GTF) file. The mean read insert sizes and their standard deviations were calculated using Picard tools software (v.1.126)¹. The read count tables were generated using HTSeq software (v0.6.0) (Love et al., 2014) and normalized based on their library size factors using the DEseq2 package (Quinlan and Hall, 2010). The gene expression analysis aimed to identify genes enriched by each apoE isoform compared to the control where apoE was absent. The Read Per Million normalized BigWig files were generated using BEDTools software (v2.17.0) 33 and bedGraphToBigWig tool (v4). Sets of genes enriched by each apoE isoform were analyzed using Kyoto Encyclopedia of Genes and Genomes (KEGG) with the aid of the DAVID Functional Classification Tool² (Huang da et al., 2009). This allows for identification of pathways enriched by each apoE isoforms. Significantly up and down regulated genes, which are not included in any recognized KEGG pathway, were ontogenetically classified (gene ontology [GO] analysis) based on the function of its coded protein included in the Uniport database (UniProt, 2019) and/or by literature review for their specific roles in neuronal biology. Genes were considered differentially expressed if their RNA level changed at least two fold with p < 0.05 (Wang et al., 2021; Yamaguchi et al., 2021) between any of the apoE isoform treatment and the control neurons grown in the absence of added human apoE lipoproteins. Four independent neuronal cultures per each apoE isoform and four control cultures grown in the apoE absence were subjected to RNA sequencing and subsequent bioinformatic analysis.

All data sets characterizing primary hippocampal neurons cultured in the presence of apoE lipoproteins are expressed relative to the control neurons grown in the presence of conditioned media from Apoe^–/– astrocyte. All data sets were first assessed for normality using Shapiro–Wilk test. Differences across various treatment conditions were analyzed using one-way analysis of variance (ANOVA) except for comparison of dendritic arborization (Sholl’s analysis) where two-way ANOVA was used. Holm–Sidak test was chosen for post hoc analysis for one- and two-way ANOVA. All ANOVA analyses were performed using GraphPad Prism v8.4.3 (GraphPad Software Inc., San Diego, CA, United States). Differences in the individual gene expression between neurons treated with each of apoE lipoproteins and the control neurons were analyzed using Student’s t-test, which was performed in the R environment (v3.1.1)³. To determine the significance of KEGG pathway enrichment Fisher’s exact test was used and the analysis was performed using DAVID Functional Classification Tool. All data in the manuscript are presented as the mean and the standard error of the mean (SEM).

Western blotting analysis of gels run under native conditions confirmed production of lipidated apoE particles by immortalized astrocytes and their secretion to the conditioned media (Figures 1A, 2A,B). Most of the apoE2 and apoE3 lipoproteins are large particles, which under native conditions resolve above the 242 kDa marker: 83.3% ± 4.7 and 91% ± 2.7%, respectively. Small particles, which resolve below the 242 kDa marker constitute only 18.6% ± 3.3 and 10.7% ± 4.5% of the total apoE2 and apoE3 particle pool, respectively. The relationship between large and small particles for apoE4 is inversed. The former constitutes only 35.9% ± 6.5% of the total apoE4 particle pool, while the later 59.7% ± 8.7%. These noticeable disparities in lipidated apoE particle size previously described in the literature underscore differences in the lipid binding characteristics across various apoE isoforms, indicating relatively poorer lipidation of the apoE4 (Morikawa et al., 2005; Kim et al., 2015; Fu et al., 2016; Chen Y. et al., 2021).

We confirmed that at the commencement of the experiment, the level of apoE protein in the neuronal media is similar for all apoE isoforms (Figures 2C,E). It averages 41.6 ng/mL ± 2.6 ng/mL as measured by the apoE ELISA (Figure 2E). This apoE concentration is comparable to the apoE concentration in the brain interstitial fluid of APOE ε3/ε3 targeted replacement mice, which is estimated at 38.1 ng/mL ± 4.0 ng/mL using in vivo microdialysis zero-flow method (Ulrich et al., 2013). We also directly confirmed internalization of apoE by neurons. At the conclusion of the experiment presence of the apoE protein in the neuronal lysate is both detectable by Western blotting of gels resolved under reducing conditions and ELISA (Figures 2D,F). Neurons grown in the presence of apoE4 contain 30.68% more apoE protein than neurons grown in the presence of apoE2 (p < 0.05) and 20.84% more than those grown in the presence of apoE3 (p = 0.067) (Figure 2F). The difference between apoE2 and apoE3 cultured neurons is negligible.

We also confirmed Apoe^–/– astrocytes secrete no apoE to the conditioned media. The control neurons treated with the conditioned media from Apoe^–/– astrocytes show no detectable apoE protein both in the neuronal media and in the neuronal lysate (Figure 2).

Our experimental setup shows that apoE isoforms differentially regulate complexity and extent of the dendritic arbor in primary hippocampal neurons cultured in vitro. Neurons, which between 10 and 17 DIV were cultured in the presence of apoE2 or apoE3 lipoproteins show a significant increase in the number of dendritic branches compared to the control neurons grown in the absence of added human apoE lipoproteins (Figure 3A). Sholl’s analysis reveals the greatest difference in the number of dendritic branches in the range of 30–110 μm from the soma, thus concerning primary and secondary dendrites (p < 0.0001 apoE2 vs. control and p < 0.0001 apoE3 vs. control) (Figure 3C). ApoE2 and apoE3 lipoproteins also increase the combined dendrite length by 10.8% ± 6% (p = 0.14) and 17.2% ± 6% (p < 0.05) compared to the control, respectively; and the total arbor surface by 26% ± 8.3% (p < 0.01) and 24.7% ± 8.3% (p < 0.01), compared to the control, respectively (Figures 3D,E). There are no significant differences concerning the number of dendritic branches, the combined dendrite length, and the total arbor surface between neurons cultured in the presence of apoE2 or apoE3 lipoproteins.

Neurons grown in the presence of apoE4 lipoproteins show significantly lower number of dendritic branches than those cultured in the presence of apoE2 (p < 0.0001) or apoE3 (p < 0.0001) lipoproteins. The difference in the number of branches between apoE4 treated neurons and the control neurons also reach statistical significance (p < 0.05) but this is more likely a result of the sum of minor variances in the number of cross points at various concentric circles, rather than represents any systematic effect attributable to apoE4 (Figures 3A,C). ApoE4 treated neurons demonstrate significantly lower values of the combined dendrite length and the total arbor surface metrics compared to those treated with apoE2 and apoE3 (Figures 3D,E). In comparison to the control neurons apoE4 reduces the combined dendritic length by 16% ± 6% (p < 0.05) but does not change the total arbor surface.

Formation and maturation of the dendritic spines are the mechanisms underlying neuronal network plasticity. ApoE lipoproteins show no systematic effect on the dendritic spine density compared to the control neurons. Only apoE3 modestly increases the density of spines by 11.1% (p < 0.05) (Figures 4A,C). There is, however, apoE isoform specific effect on the subtype frequency of dendritic spines. We distinguished four forms of the dendritic spines listed in the order of advanced maturation and readiness to form the synaptic connections: filopodium, thin, stubby, and mushrooms (Figures 4A,B). In the control neurons grown in the absence of added human apoE lipoproteins the percentages of immature spines (filopodium + thin) and mature spines (stubby and mushroom) constitute 47.62% ± 1.7% and 52.1% ± 1.8%, respectively (Figures 4D–F). ApoE2 significantly increases the fraction of mature spines to 59.47% ± 1.74% (p < 0.01) at the expense of immature spine, which fraction is reduced to 40.3% ± 1.63% (p < 0.01) compared to the control. Similar trend, yet statistically insignificant vs. the control, is seen for apoE3, which increases the fraction of mature spines to 54.4% ± 1.95% (p = 0.34), while reducing the fraction of immature spines down to 44.8%, ± 2.0% (p = 0.24). In comparison, apoE4 lipoproteins produce an opposite effect increasing the fraction of immature spines to 58.2% ± 1.4% (p < 0.0001) and reducing the fraction of mature spines to 41.9% ± 1.4% (p < 0.001) compared to the control. The differences in the spine fractions between neurons treated with apoE4 lipoproteins and those treated with apoE2 or apoE3 lipoproteins also are significant (Figures 4D–F).

ApoE2 and apoE3 lipoproteins increase expression of PSD95 and NMDAR NR1 subunit in the dendrites. The integrated density of PSD95 is higher by 11.8% ± 4.9% (p = 0.064) and 5.6% ± 4.9% (p = 0.38) in apoE2 and apoE3 treated neurons, respectively; while that of NR1 by 33.2% ± 8.0% (p < 0.0001) and 11.7% ± 8.1% (p = 0.27), respectively (Figures 5A–C). In contrast, apoE4 lipoproteins reduce the integrated density of PSD95 by 12.5% ± 4.9% (p < 0.05) and show a parallel trend to reduce that of NR1 by 9.9% ± 8.2% (p = 0.27) compared to the control neurons. Due to opposing direction of changes post hoc differences in PSD-95 and NR1 integrated density values between neurons treated with apoE4 and those treated with apoE2 or apoE3 are significant (Figures 5A–C). No apoE isoform significantly alters expression of synaptophysin compared to the control neurons (Figure 5D). A statistically significant post hoc effect is observed between apoE2 and apoE4 treated neurons (p < 0.01) due to opposing direction of change in the synaptophysin integrated density values with the former showing 8.3% ± 4.7% (p = 0.20) reduction and the later 9.5% ± 4.8% (p = 0.18) increase compared to the control neurons.

Analysis of neurons grown in the presence of lipoproteins comprised of various apoE isoforms shows a substantial transcriptomic enrichment compared to the control neurons grown in the absence of added human apoE lipoproteins (Figures 6A–C). The observed effect is largely apoE isoform specific as the overlap between gene sets enriched by two or more of apoE lipoproteins is moderate (Figure 6D). The most robust effect is caused by apoE3 which upregulates expression of 398 genes and downregulate expression of 218. ApoE4 upregulates expression of 308 genes and down regulates expression of 70 genes. ApoE2 produces the least robust effect upregulating 144 genes and down regulating 130 genes. There is a moderate overlap between genes differentially enriched by apoE3 and apoE2 with 51 and 27 commonly up and down regulated, respectively; and between apoE3 and apoE4 with 71 and 3 commonly up and down regulated, respectively. In contrast, the number of genes commonly up and down regulated by apoE2 and apoE4 are only 10 and 8, respectively. Out of the pool of 1,062 genes differentially enriched by any of apoE isoforms only 15 are upregulated and 3 are downregulated by all three isoforms.

There are numerous transcription factors among genes, which are differentially enriched by apoE lipoproteins (Figures 6A–C and Supplementary Table 1). Interestingly, the most robust effect is caused by apoE4, which affects expression of 20 transcription factors. ApoE3 alters expression of 12 while apoE2 only two. Most of these differentially expressed transcription factors are upregulated. Only Egr2 is downregulated by apoE3, while apoE4 downregulates Creb3l3. There is a minimal overlap across transcription factors enriched by various apoE isoforms. Only Vdr encoding vitamin D3 receptor is upregulated by all three of them. About 40% of transcription factors, which expression is affected by various apoE lipoproteins also belong to KEGG pathways, significantly enriched by the apoE lipoproteins.

KEGG pathway analysis shows apoE3 lipoproteins upregulate the greatest number of pathways 19, while apoE4 and apoE2 lipoproteins upregulate 10 and 7 pathways, respectively (Figures 7A,B and Supplementary Table 2). There are five pathways commonly upregulated by apoE3 and apoE2 and another 5 by apoE3 and apoE4, but there is no single pathway, which is commonly upregulated by apoE2 and apoE4 or by all three apoE isoforms. Relatively few pathways are down regulated. ApoE2 down regulates two pathways, while apoE3 down regulates one (Figure 7C and Supplementary Table 2).

KEGG pathways enriched by various apoE isoforms are functionally categorized as related to (1) neurodevelopment and synaptic plasticity, (2) cell survival and growth, (3) immune response, (4) cancer, (5) lipid metabolism, and (6) others. Pathways upregulated by apoE2 and apoE3 are functionally distinct from those primarily upregulated by apoE4. Both apoE2 and apoE3 but not apoE4 significantly upregulate four signaling pathways importantly involved in neurodevelopment and synaptic plasticity: PI3-AKT pathway (p < 0.01 apoE2; p < 0.01 apoE3), neuroactive ligand-receptor pathway (p < 0.01 apoE2; p < 0.05 apoE3), ECM receptor interaction pathway (p < 0.05 apoE2; p < 0.01 apoE3) and JAK-STAT pathway (p < 0.05 apoE2; p < 0.001 apoE3). ApoE2 also exclusively upregulates the calcium signaling pathway (p < 0.05), while apoE3 the focal adhesion pathway (p < 0.001) and the actin cytoskeleton regulatory pathway (p < 0.01), and in the category of pathways involved in cell survival and growth the TGFβ signaling pathway (p < 0.01). ApoE2 has down regulating effect on a part of the neuroactive ligand-receptor pathway (p < 0.05) (Figure 7C). ApoE4 does not upregulate any pathways classified as related to neurodevelopment and synaptic plasticity either exclusively or in common with other apoE isoforms. It does upregulate commonly with apoE3 the Hippo signaling pathway, which is related to cell survival and growth (p < 0.001 apoE3; p < 0.001 apoE4). ApoE4 also upregulates pathways involved in glycerophospholipid metabolism (p < 0.05). All types of apoE isoforms upregulate, either commonly or individually, several pathways related to cancer and immune response, which implication in the biology of developing neurons is unclear (Figure 7B).

There are numerous genes differentially enriched by treatment with various apoE lipoproteins, which do not belong to any established KEGG pathway. We subjected these genes to the GO analysis according to the function their encoded proteins play in neurons. Protein function was cross-referenced against the Uniprot database and the literature search (Figure 8 and Supplementary Table 3). All apoE isoforms commonly enrich a category of genes pertinent to cytoskeleton structure and maintenance, myelin maintenance, neurodevelopment, immune response, synthesis and metabolism of retinol, cell adhesion, oxidative stress, Wnt signaling, excitatory signaling, neurological and sensory disorders, excitatory signaling, survival, and energy and glucose metabolism. In categories including neurite growth control, excitatory inhibition, and ion channels, transporters genes are predominantly enriched by apoE2 and apoE3 with few genes affected by apoE4, while genes involved in exosomal and lysosomal biology, lipid metabolism and cell cycle are largely enriched both by apoE3 and apoE4 and to lesser extent by apoE2. Genes involved in the excitatory α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR) signaling, synaptic plasticity, neurotransmitter exocytosis, chaperone proteins, and suppression of cellular and tumor growth are enriched by apoE2 and apoE3 but not by apoE4. ApoE3 and apoE4 but not apoE2 affect expression of genes involved in intracellular trafficking, apoptosis, DNA repair. There is a prominent group of genes exclusively enriched by apoE4, which includes histone structural proteins.