The wild-type CG4 cell line, a rat OL precursor cell line originally obtained from primary cultures of bipotential oligodendrocyte-type-2-astrocytes (O-2A), was kindly donated by Huseyin Mehmet, Imperial College, London. CG4 cells, as primary O-2A cells, can differentiate into mature OLs by withdrawal of growth factors (bFGF and PDGF) and of B104 mitogens, or into astrocytes by addition of 20% fetal calf serum. Undifferentiated cells are bipolar; after 2 days of differentiation they acquire about 90% of multipolar morphology. Differentiated CG4 cells express myelin proteins, including MBP and MOG (15–17).

CG4 cells overexpressing EPOR (CG4-EPOR) were generated as reported in our previous study (8). Briefly, CG4-EPOR cells were obtained by transduction of CG4 cells with the mouse EPOR gene in a constitutive lentiviral vector, modified to include the V5 epitope, the mouse encephalomyocarditis internal ribosome entry site (IRES) and the enhanced green fluorescent protein (EGFP) reporter. The expression of recombinant V5-tagged EPOR in transduced CG4 cells was verified by measuring by flow cytometry the EGFP reporter expression, and by immunoblotting with the anti-V5-tag mouse monoclonal antibody (Invitrogen), as described (8, 25).

CG4-EPOR cells were seeded in poly-l-ornithine-coated 6-well plates (320,000 cells in 4 ml GM per well). They were maintained at the precursor stage by culture in growth medium (GM), consisting of Dulbecco’s modified Eagle medium (DMEM) (Sigma-Aldrich) supplemented with biotin (10 ng/ml), bFGF (5 ng/ml), PDGF (1 ng/ml), N1 supplement (all from Sigma-Aldrich), and 30% B104-conditioned medium, obtained as previously reported (8). After overnight culture, the cells were induced to differentiate into OLs by switching to differentiation medium (DM), consisting of DMEM-F12 (Invitrogen) supplemented with progesterone (3 ng/ml), putrescine (5 µg/ml), sodium selenite (4 ng/ml), insulin (12.5 µg/ml), transferrin (50 µg/ml), biotin (10 ng/ml), thyroxine (0.4 µg/ml), and glucose (3 g/l) (all from Sigma-Aldrich), as reported (8). After 3 h, some of the cells were treated with recombinant human erythropoietin (rhEPO; Creative Dynamics) at 10 ng/ml and cultured for 1 or 20 h prior to RNA extraction. There is extensive biological cross-reactivity between human EPO and the EPOs of other mammals. Human EPO is approximately 80% homologous to rodent EPO, and it has been shown to be biologically active in rodents for erythropoietic and neurotrophic functions (26, 27). Undifferentiated cells were seeded as above and cultured in GM without switching to DM for the whole length of the experiment.

To measure ERK1/2 phosphorylation by western blot, cells were plated in poly-ornithine coated 24 well plates at 200,000/well in GM for 24 h, then switched to DM and incubated overnight to starve them of growth factors present in GM. Cells were then preincubated with PD184352 (Cell Signaling, #12147) or DMSO for 1 h, then treated with or without EPO at 10 ng/ml for the indicated times. PD184352 was dissolved in DMSO at 25 mg/ml and then diluted in DM at the indicated concentrations.

Each sample was lysed with 1 ml QIAzol (QIAGEN). Total RNA was extracted by using the miRNeasy system and protocol (QIAGEN). RNA purity and integrity were determined using a NanoDrop ND-1000 (NanoDrop Technologies) and an Agilent 2100 Bioanalyzer (Agilent Technologies). All samples had a A260/A280 ratio >1.8 and RNA Integrity Number >9.

All experimental conditions were performed in quadruplicate; undifferentiated cells were cultured in quadruplicate but only three random samples were used for microarray analysis and all of the four samples for quantitative PCR (qPCR) validation. In total, 19 arrays were completed: 3 undifferentiated (und) and 16 differentiated: 4 differentiated (ctr) and 4 differentiated + EPO (EPO) at each time point (4 and 23 h of differentiation; 1 and 20 h after EPO treatment, respectively).

RNA was amplified, labeled, and hybridized onto Single Color SurePrint G3 Rat GE 8 × 60K Microarrays (AMADID:028279; Agilent) at Oxford Gene Technology, Oxford, UK. Following hybridization, the arrays were scanned to derive the array images. Feature extraction software v10.7.3.1 was used to generate the array data from the images.

Raw data in standard format from the microarray experiment have been deposited in the Gene Expression Omnibus (GEO) database of NCBI (28) and are accessible through GEO Series accession number GSE84687.¹ Raw data were normalized and analyzed using GeneSpring (Agilent) and Excel software. Transcript expression levels between the experimental groups were compared by Student’s t-test done on the log₂ of the gProcessed Signal, obtaining uncorrected p-values. Subsequent multiple comparison corrections were performed using the Benjamini-Hochberg False Discovery Rate procedure, obtaining adjusted p-values. Fold change in the expression was calculated as the ratio between the average of the gProcessed Signals of the various groups and expressed as log₂. Differences in expression with an adjusted p-value < 0.05 and an absolute fold change ≥ 1.5 (log₂ fold change ≥ 0.58) were considered statistically significant. Functional annotation and biological term enrichment was done using the Database for Annotation, Visualization, and Integrated Discovery (DAVID)² (29). DAVID calculates a modified Fisher’s exact p-value to demonstrate enrichment. Categories with p-value < 0.05 were considered significantly enriched.

Reverse transcription (RT) and real-time qPCR were carried out as reported (30) on total RNA from quadruplicate samples, using TaqMan^® gene expression assays (Applied Biosystems/Life Technologies) and Brilliant III qPCR master mix (Stratagene/Agilent Technologies). Gene expression was quantified using the ΔΔCt method, according to Applied Biosystems’ guidelines. Results were normalized to HPRT1 expression (reference gene) and expressed as log₂ of the relative expression (ratio) vs one of the control samples (as indicated), chosen as the calibrator.

Cells were lysed in RIPA buffer (Thermo Fisher Scientific) and total cellular extracts were incubated on ice for 30 min and then cleared by centrifugation (15,000 g for 20 min at 4°C). Protein concentration was measured with the BCA kit (Pierce) and 30 µg of cellular proteins were analyzed by 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred to a nitrocellulose membrane (Millipore) by electroblotting. Membranes were blocked for 1 h with 5% bovine serum albumin (BSA, Sigma-Aldrich) in TBS and then probed with rabbit polyclonal anti-phospho-ERK1/2 (Cell Signaling, #9101) followed by detection with goat-anti-rabbit-IgG-horseradish peroxidase conjugate (Sigma-Aldrich, #A0545), all in 5% BSA in TBS with 0.1% Tween 20 (Sigma-Aldrich). Membranes were then stripped with Re-Blot Plus Strong Solution (Millipore) and then blocked and reprobed with rabbit polyclonal anti-ERK1/2 (Cell Signaling, #9102). Protein bands were visualized using ECL detection reagent (GE Healthcare) and exposing the membranes to autoradiography films (Hyperfilms, GE Healthcare).

Undifferentiated CG4 cells cultured for 1 day in GM were switched to DM and treated with EPO (10 ng/ml) after 3 h; gene expression analysis was performed 1 and 20 h upon EPO stimulation (at 4 and 23 h of differentiation, respectively). The dose of EPO was chosen based on dose–response experiments published in our previous study; since in these cells EPO dose-dependently induced MOG and MBP expression with maximal induction at 8 ng/ml and no further increase at 80 and 400 ng/ml, the dose of 10 ng/ml was used throughout this study (8). EPO-treated were compared with untreated differentiating cells; the effect of differentiation was investigated by comparison with undifferentiated cells cultured for 1 day in GM in the same experimental conditions. Genes differentially expressed were identified by setting a cutoff fold change (FC) of 1.5 (log₂ FC = 0.58) and cutoff p-value < 0.05 after applying the Benjamini–Hochberg (BH) correction for multiple tests.

At 20 h, culture in DM changed the expression of 2,663 out of 24,272 transcripts. Five genes represented by duplicate probes changed in opposite directions were excluded from the analysis; when genes were represented by duplicate or triplicates probes consistently changed in the same direction, only the most significantly changed one was included, obtaining 2,436 genes, of which 99 were further changed in the same direction by EPO; 21,609 transcripts were unaffected by differentiation, of which 174, representing 150 genes, were changed by EPO (Figure 1). The relative expression changes of all genes affected by differentiation (2,436; 1,164 upregulated and 1,272 downregulated; Figure 1) and/or by EPO (278; 177 upregulated and 101 downregulated; Figure 1) and details about the redundant probes removed are reported in Files S1–S3 in Supplementary Material.

On the other hand, 1 h of EPO treatment had little effect on the gene expression profile. Culture for 4 h in DM affected 931 transcripts out of 24,272; after removing one gene represented by duplicate probes changed in opposite directions, and redundant probes, 878 genes were obtained; of these, 461 were upregulated and 417 downregulated. The addition of EPO for the last hour increased the expression of only six transcripts, corresponding to five genes (Egr1, H19, Fos, Arl4a, Irx2), of which three (Egr1, Fos, and Irx2) had been decreased, and therefore changed in the opposite direction, by previous culture in DM; the remaining two, H19 and Arl4a, had been unaffected by differentiation. The experimental design flowchart at 1 h is shown in Figure S1 in Supplementary Material and the whole list of the genes affected by differentiation and/or by EPO at 1 h, with details about the removed probes, is provided in Files S4 and S5 in Supplementary Material.

To understand the general function of the genes up- or downregulated in differentiating cells at 20 h, we used DAVID to identify the overrepresented GO categories with biological process (BP) designation (GO:BP).

Strongly proliferating OPCs need to exit the cell cycle to start differentiating and producing myelin proteins (31). Therefore, unsurprisingly, the top overrepresented GO:BP categories identified by DAVID analysis of the 1,272 genes downregulated by differentiation (full list, File S1 in Supplementary Material) included “cell division” and “mitotic nuclear division.” Thus, differentiation inhibited the expression of many genes driving cell proliferation; this confirms previous studies in CG4 cells (15, 32) and in primary OLs (33, 34).

Erythropoietin further downregulated the expression of 56 genes already decreased by differentiation, and of 37 of the unaffected ones (Figure 1; File S3 in Supplementary Material); the top enriched GO:BP terms in both groups of genes included “DNA replication”; only 8 transcripts were downregulated by EPO but upregulated by differentiation, and a PubMed search did not highlight any link with OLs or myelination. Therefore, analysis of the EPO-downregulated genes revealed a strong inhibitory action of EPO on a network of genes involved in cell proliferation, some of which were already affected by differentiation; however, it did not suggest any other mechanism that might be triggered by EPO through gene downregulation. The DAVID analysis of the genes downregulated by differentiation and/or by EPO is reported in File S6 in Supplementary Material.

Among the 1,164 genes upregulated by differentiation alone (full list, File S1 in Supplementary Material), the top enriched GO:BP categories were “nervous system development” (n = 25), which included genes involved in myelination, such as Mag and Erbb2 (35), “cell adhesion” (n = 30), including genes previously found upregulated in differentiating OLs, such as Cd9, Neo1, Ninj2, Opcml, Tnr (36), “fatty acid beta-oxidation” (n = 10), and “glycolysis” (n = 9), both required to provide intermediates for myelin synthesis and energy to support the myelination process (Table 1) (37, 38). EPO further upregulated 43 of the transcripts already increased by differentiation alone (Figure 1; File S2 in Supplementary Material), among which DAVID analysis did not highlight any enriched GO:BP category. However, by manual analysis we identified two myelin genes [Mag and Pmp2, a peripheral myelin protein which is also present in the CNS (39)], one gene belonging to the GO:BP “nervous system development” (Mag) and three genes to “cell adhesion” (Cdhr2, Mag, Ntm) but none to “fatty acid beta-oxidation” or “glycolysis.” Twenty-one transcripts were upregulated by EPO but downregulated by differentiation, including low density lipoprotein receptor (Ldlr), which is expressed in mature OLs (40) and myelin transcription factor 1 (Myt1), whose multiple functions on proliferation and differentiation of OLs, likely dependent on interacting molecules, have not been fully characterized (33).

We then looked at the transcripts specifically increased by EPO but unchanged by differentiation alone (113; Figure 1). The 15 most induced genes in this group are reported in Table 2 (full list, File S2 in Supplementary Material). Some of these genes have been described to play a role in myelination, including Igf1 and Igf2 (21), and Ptpre, a protein tyrosine phospatase whose knockouts have defects in myelination (19). Interestingly, PTPRE also inhibits ERK activation (41). DAVID analysis identified enrichment of generic GO:BP terms including “response to organic cyclic compounds” (n = 10), “response to activity” (n = 6), “response to nutrient levels” (n = 6), and positive regulation of ERK1 and ERK2 cascade (n = 7) (Table 1).

We noticed, in the list in Table 2, a strong effect of EPO (about 50-fold induction) on Cd36, which mediates long-chain fatty acid uptake and metabolism (22, 42).

We, therefore, searched manually for other genes involved in lipid transport and metabolism among all the transcripts increased by EPO, including those also changed by differentiation alone (177; Figure 1; File S2 in Supplementary Material). We specifically searched for genes annotated with GO:0006629 “lipid metabolic process,” GO:0006631 “fatty acid metabolism,” GO:0006635 “fatty acid beta-oxidation,” GO:0019395 “fatty acid oxidation,” and GO:0015909 “long-chain fatty acid transport.” Other than Cd36, our search identified Ppargc1a/Pgc1alpha, lipin 1 (Lpin1), Lpin2, Pnlip, peroxisome proliferator-activated receptor delta (Ppard), and perilipin 2 (Plin2). The effect of EPO on the expression of these genes in differentiating cells is reported in Table 3.

Microarray expression of eight genes of interest induced by EPO at 20 h, including the Igfs, Ptpre, some of the genes involved in lipid transport and metabolism and one myelin gene, Pmp2, was validated by RT-qPCR, using the same RNA used for the microarray experiment and RNA from a completely independent experiment (Figure 2). EPO-induced expression of all the eight genes tested was confirmed by RT-qPCR.

To investigate whether EPO-induced PTPRE might contribute to the increased myelin gene expression, possibly by inhibiting ERK1/2 activation, we used an inhibitor of ERK1/2 phosphorylation, PD184352. In cells incubated with DM for 1 day, EPO increased the phosphorylation of ERK1/2, which peaked at 10 min and returned to background level at 2 h (Figure 3A). Preincubation with PD184352 (0.1–2 µm) for 1 h completely inhibited EPO-induced ERK1/2 phosphorylation (Figure 3B), showing that the inhibitor was functionally active; PD184352 also potentiated EPO-induced Mog expression (Figure 3C). This result showed that inhibition of the phosphorylation of ERK1/2, achieved by EPO by inducing PTPRE, might be a mechanism by which EPO increases myelin gene expression.