Neuroblastoma patient-derived xenograft cells cultured in stem-cell promoting medium retain tumorigenic and metastatic capacities but differentiate in serum

Cultured cancer cells serve as important models for preclinical testing of anti-cancer compounds. However, the optimal conditions for retaining original tumor features during in vitro culturing of cancer cells have not been investigated in detail. Here we show that serum-free conditions are critical for maintaining an immature phenotype of neuroblastoma cells isolated from orthotopic patient-derived xenografts (PDXs). PDX cells could be grown either as spheres or adherent on laminin in serum-free conditions with retained patient-specific genomic aberrations as well as tumorigenic and metastatic capabilities. However, addition of serum led to morphological changes, neuronal differentiation and reduced cell proliferation. The epidermal growth factor (EGF) and basic fibroblast growth factor (bFGF) were central for PDX cell proliferation and MYCN expression, and also hindered the serum-induced differentiation. Although serum induced a robust expression of neurotrophin receptors, stimulation with their cognate ligands did not induce further sympathetic differentiation, which likely reflects a block in PDX cell differentiation capacity coupled to their tumor genotype. Finally, PDX cells cultured as spheres or adherent on laminin responded similarly to various cytotoxic drugs, suggesting that both conditions are suitable in vitro screening models for neuroblastoma-targeting compounds.


Neuroblastoma PDX cells retain metastatic capacity and patient-specific genomic aberrations following prolonged in vitro culturing. Previously, we reported that cells isolated from neuroblastoma
PDXs can grow as neurospheres in stem-cell promoting medium (SC medium) (ref. 10 and Fig. 1a). Here we have further characterized cells isolated from PDX #2 and PDX #3, established from high-risk MYCN amplified neuroblastomas 10 . The corresponding cell lines LU-NB-2 and LU-NB-3 are thus MYCN amplified and express typical neuroblastoma markers, as shown previously 10 and here by NCAM (also known as CD56) and NSE expression as well as chromogranin A (CHGA) and tyrosine hydroxylase (TH) protein expression, thereby confirming their neuroblastoma origin (Fig. 1a-c). Orthotopic injection of different numbers of LU-NB-2 cells revealed that 10 6 and 10 4 cells gave rise to tumors in all mice, whereas 10 2 cells did not result in a visible tumor 1 year after injection (Supplementary Table S1) (due to ethical reasons, mice could not be kept longer). Orthotopic injection of 10 6 LU-NB-3 cells gave rise to tumors in 3 out of 4 mice (Supplementary Table S1). PDX cell-derived tumors exhibited strong NCAM staining and heterogeneous expression of TH and Ki67 positive tumor cells (Fig. 1d). Metastasis to lungs, liver and bone barrow were detected in tumor-bearing animals as shown by NCAM staining (Fig. 1e and Supplementary Table S1).
The capacity to seed metastasis is hypothesized to be a trait of cancer stem cells and while neuronal marker expression levels were in parity with those of the classical serum-cultured SK-N-BE(2) c neuroblastoma cells, PDX cells had higher expression of putative neuroblastoma stem-cell associated markers such as NOTCH1, NOTCH3 and C-KIT (also known as CD117 or SCFR) (Fig. 1b) 13 .
To confirm that cultured PDX cells retained patient specific genomic aberrations we performed single nucleotide polymorphism (SNP) array. This revealed that cultured PDX cells at low (passage 3), middle (passage 15) and high (passage 30) passages retained the genomic aberration profiles (e.g. 1p deletion, MYCN amplification and 17q gain) observed in the corresponding patient tumor and xenograft (Fig. 1f). No other gross genomic changes were observed at later passages, suggesting that the PDX cells are genomically stable in long-term culture. Importantly, tumor-initiating and metastasizing capacity was maintained also at high in vitro passage (passage 54) ( Supplementary Fig. S1). Thus, the neuroblastoma PDX cells retain tumor-initiating and metastatic capacity also after longer in vitro passaging.
Serum promotes differentiation and reduces proliferation. The optimal culture condition for LU-NB-2 and LU-NB-3 cells was explored by comparing short-term serum-grown PDX cells and serum-free cultured cells. Addition of serum induced cell attachment of both LU-NB-2 and LU-NB-3 cells ( Fig. 2a and Supplementary Fig. S2a). Other overt serum concentration dependent effects were morphological differentiation and reduced cell numbers (Fig. 2a,f and Supplementary Fig. S2a). The attached cells showed neurite outgrowth and a robust increase in TH expression, both at mRNA and protein level, suggesting serum-driven differentiation (Fig. 2a-c and Supplementary Fig. S2a-c). In addition, mRNA and protein expression of other sympathetic neuronal markers including CHGA, SYP and SCG2 14 , also increased (Fig. 2b,c and Supplementary Fig. S2b,c). The majority of differentiation markers was induced in a serum-concentration dependent manner in both LU-NB-2 and LU-NB-3 cells, however LU-NB-2 cells were less responsive to serum (Fig. 2a-c and Supplementary  Fig. S2a-c). This could potentially reflect the fact that PDX #2 was established from relapsed disease while PDX #3 was established prior to treatment. The expression of genes associated with an immature sympathetic nervous system phenotype, e.g. NOTCH1, NOTCH3 and HEY1 was downregulated in response to serum (Fig. 2d). Furthermore, serum led to an increased number of cells in G0/G1 phase at the expense of S/G2/M phase cells (Fig. 2e). There was no increase in the sub G0/G1 fraction as determined by flow cytometry, indicating that the μm. (f) SNP array profiles of patient tumors; patient #2 and patient #3 (upper panels), the corresponding xenografts; PDX #2 and PDX #3 (middle panels), and corresponding PDX cells LU-NB-2 and LU-NB-3 (lower panels). The SNP profiles of patient tumors and PDXs are taken from Braekeveldt et al. 10 . Error bars represent ± SEM from 3 independent analyses. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001; Student's t test. reduced cell number in serum was not due to cell death, a result corroborated by trypan blue staining (Fig. 2f). Continued growth in 10% serum for up to 6 weeks revealed that the cells survived, were still differentiated but with stagnating growth rates ( Supplementary Fig. S2d,e). Taken together, LU-NB-2 and LU-NB-3 cells cultured under serum-free conditions maintain a less differentiated, proliferative cell phenotype while serum reduces proliferation and induces neuronal differentiation.
The neurotrophins NGF and NT-3 and their cognate receptors, TrkA and TrkC, are required for terminal differentiation of proliferating, non-transformed sympathetic neuroblasts 15,16 . In neuroblastoma, high TrkA and TrkC expression is associated with low stage and less aggressive disease 17,18 . The TRKA (also known as NTRK1) and TRKC (also known as NTRK3) expression increased robustly in serum-cultured LU-NB-3 cells (Fig. 2g), further supporting our conclusion that serum induces sympathetic differentiation. The obligatory co-receptor p75 NTR subunit of high affinity neurotrophin receptors is expressed in these cells, although at low levels (Fig. 2g). Interestingly, treating serum-cultured LU-NB-3 cells with NT-3 or NGF alone or in combination did not further increase the expression of TH, DBH or SCG2 (Supplementary Fig. S2f) nor did it enhance the serum-induced morphological differentiation. Cultured neuroblastoma cells expressing TRKA or TRKC are not resistant to neurotrophin stimulation, per se, as exemplified by NGF or NT-3 treated TRKA or TRKC overexpressing SH-SY5Y cells, although neurothrophin treatment in those cells did not induce terminal differentiation 19,20 . In line with reported data obtained from established neuroblastoma cell lines 21 , we propose that also the PDX cells lack the capacity to terminally differentiate.

MYCN, telomerase complex genes and neuroblastoma PDX cell differentiation. RNA sequenc-
ing of LU-NB-3 cells cultured in either serum-free or serum-containing medium revealed a general induction of a sympathetic neuronal phenotype at serum conditions as exemplified by increased expression of DBH, TH, PHOX2A, BCL2, MAOA and NSG1 (Fig. 3a). Furthermore, gene ontology analysis revealed a significant enrichment of neuronal differentiation in the samples of serum-cultured cells (Fig. 3b,c and Supplementary File S1). Interestingly, MYCN was the top downregulated gene in serum-cultured LU-NB-3 cells, showing over 9-fold decrease in mRNA expression level (Fig. 3d). This result was further supported by decreased protein levels ( Fig. 2c and Supplementary Fig. S2c). Gene set enrichment analysis (GSEA) revealed a significant enrichment of genes regulated by MYCN in serum-free cultured cells ( Fig. 3e and Supplementary File S2).
MYCN inhibition has previously been proposed to promote neuroblastoma cell differentiation, and we therefore treated LU-NB-3 cells with the MYC-MAX inhibitor 10058-F4. Treatment with 10058-F4 led to decreased MYCN protein, verifying MYCN inhibition (Fig. 3f). At mRNA level, treatment with 60 μM 10058-F4 for 72 h led to a modest upregulation of TH but had little or no effect on SCG2, TRKA and NOTCH1 expression, indicating that MYCN inhibition promotes differentiation of neuroblastoma PDX cells to a limited extent (Fig. 3g).
Tumor cells are characterized by high proliferation and are dependent on active telomerase to circumvent shortening of chromosomal ends. Because MYC can induce TERT expression 22 we asked whether downregulation of MYCN and reduced proliferation in serum was coupled to decreased expression of TERT. Interestingly, TERT and other components of the telomerase complex, including TERC, DKC1, GAR1 and NHP2, were all downregulated in serum-grown LU-NB-3 cells (Fig. 3h). The snoRNPs DKC1, GAR1 and NHP2 have been shown to be important for telomerase activity in neuroblastoma 23 . A signature score based on these markers correlates with segmental aberrations and is an independent predictor of poor prognosis in neuroblastoma 23,24 .

The SC medium growth factors EGF and bFGF decrease serum-induced differentiation.
For serum culture conditions we excluded the growth factors EGF and bFGF. These are two important components of stem-cell promoting medium previously shown to support expansion of neural stem cells by symmetrical division 25 . Withdrawal of EGF and bFGF from SC medium led to lower cell numbers and lowered MYCN levels in LU-NB-3 cells (Fig. 4a,b). However, this did not induce differentiation as determined by TH expression levels ( Fig. 4b). Furthermore, addition of EGF and bFGF to serum-containing medium led to a less pronounced differentiating effect of added serum (Fig. 4b). Thus, our data suggest that EGF and bFGF are important for sustained proliferation of neuroblastoma PDX cells and that these factors diminish serum-induced differentiation.
Glucocorticoid receptor signaling has been reported to promote differentiation of neuroblastoma cells 26 . Since serum contains glucocorticoids we tested if 10% lipid-free, charcoal-stripped serum would induce differentiation. This condition did not promote cellular attachment or morphological differentiation, yet the expression of differentiation markers was still induced and MYCN levels decreased (Fig. 4c-e). We hypothesized that charcoal-stripped serum might still contain protein-bound corticosteroids able to induce the expression of differentiation markers and therefore measured the amount of steroids in regular and charcoal-stripped serum. Charcoal-stripped serum was principally free of steroids (Supplementary Table S2), suggesting that glucocorticoid signaling is not responsible for upregulation of differentiation markers at serum culture conditions.

Serum-induced differentiation of PDX cells is reversible.
As we did not observe further differentiation by neurothrophin treatment we asked whether the serum-induced differentiation was reversible. LU-NB-3 cells were cultured in 10% serum medium for 7 days and then transferred back to SC medium for 4 days (Fig. 4f). This led to an increased proliferation and MYCN levels, and reduced expression of TH and SYP (Fig. 4g-i). Thus, the cellular differentiation induced by serum is reversible, which is in line with the observation that the PDX cells could not be terminally differentiated.
We next asked whether neuroblastoma cells that have been cultured in serum for decades would convert to a less differentiated phenotype in serum-free medium. We explored this by growing SK-N-BE(2)c cells for 7 and 14 days in either standard serum-containing or SC medium. SK-N-BE(2)c cells cultured in SC medium presented with a less flattened morphological appearance and grew in clusters, but not as spheres ( Supplementary Fig. S3a). Furthermore, differentiation marker expression data did not uniformly suggest that these cells acquired a less differentiated phenotype (Supplementary Fig. S3b).
Laminin promotes adherent growth and maintains the tumorigenic and metastasizing phenotype. As serum-induced adherence resulted in strong phenotypic changes we sought alternative protocols to grow neuroblastoma PDX cells as a monolayer. Adherently grown glioma, as well as mouse and human neural stem cells, have successfully been established on laminin-1 (laminin α1-β1-γ1) with retained stemness features 25,27,28 . The neuroblastoma PDX cells, however, adhered poorly to this substrate. A screening of laminin isoforms synthetized by the neuroblastoma PDX cells revealed that laminin α5 chain was highly expressed in both LU-NB-2 and LU-NB-3 ( Supplementary Fig. S4a). We therefore tested recombinant human laminin 511 (LN-511) and 521 (LN-521) as substrates for the PDX cells. Though both laminin isoforms promoted adherent growth of LU-NB-2 and LU-NB-3, we used LN-521 for further experiments. Following adherent growth on LN-521, PDX cells again formed spheres when transferred back to uncoated dishes (Fig. 5a). As laminin α5 was significantly upregulated in serum-grown LU-NB-3 cells (Supplementary Fig. S4b) and PDX cells adhering to LN-511 or LN-521 showed signs of morphological differentiation, we asked whether surface adherence alone could induce a differentiated phenotype. PDX cells grown on LN-521 showed a modest increase in TH, CHGA and SYP expression ( Fig. 5b and Supplementary Fig. S4c) as compared to serum-grown PDX cells (e.g. compare 3.4-fold increase in TH expression in Fig. 5b and 17.9-fold increase in Fig. 2b). At the protein level, TH and GAP43 levels  slightly increased in LN-521-cultured LU-NB-2 and LU-NB-3 cells, while CHGA was downregulated, despite its increased mRNA expression (Fig. 5c and Supplementary Fig. S4d). In addition, the expression of genes associated with a less differentiated phenotype was not affected by laminin growth (Supplementary Fig. S4e). Monolayer culture provides uniform access to growth factors and LN-521-cultured LU-NB-3 cells were more viable than sphere-grown cells (Fig. 5d). MYCN expression did not differ between LN-521-cultured and sphere-cultured LU-NB-3 cells (Fig. 5b), although it was slightly lower in LN-521-cultured LU-NB-2 cells ( Supplementary  Fig. S4c). Importantly, TERT levels were unaffected in LN-521-cultured PDX cells (Fig. 5b and Supplementary  Fig. S4c). In conclusion, neuroblastoma PDX cells can grow adherently on laminin, which promotes the viability of the cells with little effect on differentiation. As anticipated, addition of serum to LN-521-cultured cells promoted differentiation (Supplementary Fig. S5a,b).
To determine whether PDX cells cultured on LN-521 maintained tumor-initiating and metastatic capacities, we cultured cells on LN-521 for 3 months before injecting cells orthotopically into the adrenal gland of NSG mice. All animals (n = 6) developed tumors with typical neuroblastoma morphology and examination of lungs, liver and bone marrow revealed metastatic growth in all examined animals (Fig. 5e,f). Finally, the response to various cytotoxic drugs did not differ between sphere-and LN-521-cultured cells (Fig. 5g and Supplementary Fig. S4f), suggesting that laminin adherent cells are promising models for future drug screening.

Discussion
Conventional tumor-derived cell lines established decades ago, including neuroblastoma cell lines, have traditionally been cultured in the presence of serum. Although derived from highly aggressive and metastatic tumors they show limited infiltrative growth in rodent xenograft models and generally do not metastasize. We previously established and characterized orthotopic neuroblastoma PDX models and showed that PDX tumors infiltrated adjacent tissues and seeded metastases to liver, lungs, and bone marrow. Such patterns of infiltrative growth and metastases are not observed with orthotopically xenografted conventional cell lines 10 . We further showed that neuroblastoma cells isolated from PDXs could be cultured short-term as spheres, and when re-injected they formed tumors and metastases 10,11 . Here, we show that these cells retain their proliferative capacity after long-term passaging, suggesting that they have infinite in vitro growth capacity. The PDX cell lines retain patient-specific genomic aberrations up to at least 30 passages without obtaining additional gross DNA aberrations. Furthermore, their tumorigenic and metastatic capacities are retained also at high in vitro passages. We further demonstrate that the sphere-forming PDX cell lines can be grown as a monolayer on LN-521 without compromising their capacities to form tumors and metastases. Importantly, PDX cells adhering to LN-521 did not significantly alter the cellular response to cytotoxic drugs. As the neuroblastoma PDX tumors are established from high-risk tumors and metastasize to clinically relevant sites, including the bone marrow, the neuroblastoma PDX model offers a highly promising in vivo model for pre-clinical drug testing. We further conclude that PDX cells grown as spheres or on LN-521 are suitable in vitro systems for future screening of drugs targeting aggressive neuroblastoma.
Further analysis of PDX cells grown on LN-521 revealed that they were slightly more viable and expanded readily as compared to sphere-grown cells, suggesting that laminin-integrin signaling supports neuroblastoma PDX cell survival. Intriguingly, the PDX cells readily attached to two laminin isoforms containing the α5 chain, laminin 511 and 521; the same type of laminins used for growing embryonal stem cells 29,30 . This may suggest that the neuroblastoma PDX cells share features with cells of embryonal origin. Recent work comparing glioma cells grown as neurospheres versus those grown on laminin, showed no significant differences in growth rate, apoptosis, differentiation markers or tumorigenicity 31 . These findings together with our results support that both growth conditions are suitable for stable expansion of neural-derived tumor cells in vitro.
We found that withdrawal of EGF and bFGF and addition of serum also promoted attachment of the PDX cells but led to differentiation, reduced MYCN levels and decreased proliferation. Thus, avoiding serum appears to be a key strategy to preserve the original immature features of these neuroblastoma cells which is in line with previously established primary neuroblastoma cells propagated in serum-free media 12 . Differentiation is highly associated with neuroblastoma patient outcome, and it has been proposed that malignant transformation of neuroblasts at different stages during fetal development causes the great phenotypic diversity of this disease 32 . The enigmatic spontaneous differentiation observed in stage 4S neuroblastomas, together with experimentally induced in vitro differentiation of neuroblastoma cells 21 raised hope that aggressive neuroblastomas could be treated by triggering neuronal differentiation. In the clinic, high-risk neuroblastoma patients are treated with the differentiating agent isotretinoin to combat minimal residual disease, however whether the effect of this treatment is due to induced tumor cell differentiation in patients has not been demonstrated. Furthermore, so far, there are no studies demonstrating that neuroblastoma cells in culture established from high-risk patients can terminally differentiate (i.e. acquire an irreversibly differentiated phenotype and ceased proliferation capacity). Here we show that the serum-induced overt morphological differentiation accompanied by increased and robust expression of marker genes is reversible. Furthermore, despite the serum-differentiated PDX cells expressing the neurotrophin receptors TrkC and TrkA, which mediate the key responses leading to terminally differentiated sympathetic neurons 16 , no additive effect was observed when combining serum with the neurothrophins NT-3 and NGF. We conclude that the neuroblastoma PDX cells can be pushed to a more differentiated phenotype; however, they appear to have lost the capacity to terminally differentiate (Fig. 6).
Prolonged culture of LU-NB-3 cells in serum did not result in a regained high growth rate. Whether the PDX cells could be established as classical cell lines in serum-containing medium when cultured for longer periods remains to be shown. Although it would be of academic interest to try to establish PDX cells in serum-medium to compare geno-and phenotypes with PDX cells cultured in SC medium, such cell lines would be of less value for understanding the biology of neuroblastoma. Based on our own observations and on similar findings reported in glioma 5 , we expect that such cells will phenotypically and genotypically deviate from the tumors they were derived from. Theoretically, serum-grown cells experience cellular crisis similar to what happens in tumor cells upon loss of telomerase activity. Reduced telomerase activity in neuroblastoma cells resulted in increased frequency of anaphase-bridges 23 . Thus, upon re-activation of telomerase activity and recovered proliferative capacity, long-term serum-grown cells have likely gained additional genetic aberrations, created by breakage-fusion-bridge cycles.
Considering the significance of MYCN in neuroblastoma, our observation that MYCN was the top downregulated transcript in serum-grown PDX cells is intriguing. As well as being essential for normal neurogenesis 33 , MYCN has been shown to prevent neuronal differentiation 26,34,35 , although overexpression of MYCN in non-MYCN-amplified SK-N-SH cells did not impede differentiation induced by several different protocols 36 . Treatment with the MYC-MAX inhibitor 10058-F4 promoted expression of a few, but not all of the differentiation markers tested, showing that MYCN inhibition alone cannot substantially push the PDX cells towards a more differentiated stage. We speculate that presence of EGF and bFGF hinders differentiation to some extent. It would therefore be interesting to test 10058-F4 without the addition of growth factors. However, withdrawal of growth factors in itself leads to downregulation of MYCN, making it difficult to draw conclusions from that type of experiment.
It was recently shown that MYCN-regulated miRNAs inhibit the expression of several nuclear hormone receptors, including the glucocorticoid hormone receptor, and simultaneous inhibition of MYCN and activation of glucocorticoid signaling resulted in differentiation of neuroblastoma cells 26 . Our data show that charcoal-stripped serum devoid of steroids promoted MYCN downregulation and increased the expression of differentiation markers in the PDX cells, however it did not promote attachment or morphological differentiation. In conclusion, we show that neuroblastoma PDX cells grown under serum-free conditions maintain patient-specific genomic aberrations over time and that tumorigenic and metastatic capabilities, as well as response to cytotoxic drugs, are preserved also at higher in vitro passages or when cells are cultured as monolayer on laminin. Our data and other studies showing serum-induced differentiation of tumor cells 5,37 strongly suggest that serum should be avoided in order to maintain tumor cell immaturity in culture. The phenotypic effects of serum have mostly been overlooked but they can explain why serum-established cell lines do not metastasize or grow invasively in xenograft models. This observation could account for the generally poor translation of results from pre-clinical screening based on classical cell lines to clinical testing of drugs 4 .

Materials and Methods
Neuroblastoma PDX cells, cell lines and cell culture treatments. Tissue from neuroblastoma PDXs was dissociated and digested for 45 min at 37 °C with Liberase (0.15 mg/mL, Roche), passed through a 70 µm cell strainer (BD Biosciences) and cultured in stem cell medium (SC medium) at 37 °C in 5% CO 2 10 . Spheres were dissociated using Accutase (Sigma-Aldrich, St. Louis, MO). PDX #2 had been established from cerebral metastasis from a stage 4 tumor where the patient had undergone prior treatment while PDX #3 had been established from a primary stage 3 tumor in the adrenal gland. For serum culture, SC medium excluding growth factors and B-27 supplement was supplemented with fetal bovine serum (FBS) or 10% charcoal-stripped FBS (Thermo Scientific). For monolayer culture on human recombinant laminin (Biolamina), plates were coated according to manufacturers instructions. PDX cells were routinely grown to confluence and dissociated using Accutase. The neuroblastoma cell line SK-N-BE(2)c (ATCC) was cultured as previously described 38  Cell cycle analysis. Cell cycle distribution was determined by fixating PDX cells in 70% ice-cold ethanol in −20 °C, washed in PBS and incubated on ice for 45 min in Vindelöv solution (3.5 µmol/L Tris-HCl (pH 7.6), 10 mmol/L NaCl, 50 µg/mL propidium iodide, 20 µg/mL RNase, 0.1% v/v NP40). Samples were run on a FACSVerse instrument (BD Biosciences) and data was analyzed using FlowJo software (FlowJo, LLC).
Single nucleotide polymorphism array. DNA was extracted from PDX cells at low (p3), middle (p15) and high (p30) in vitro passages using DNeasy Blood and Tissue Kit (Qiagen). For analysis, genomic DNA was hybridized to an Affymetrix CytoScan HD chip (Affymetrix) containing approximately 2.6 million markers of which almost 750,000 are SNPs. Constitutional copy number variants were removed by filtering against the Database of Genomic Variants (Oct. 2016). Copy number variation analyses were performed using ChAS software. The analyses of xenografts and the corresponding patient tumors were reported previously 10 .
RNA sequencing and data analysis. PDX cells were cultured in SC medium or 10% serum for 7 days.
Total RNA was extracted using the RNeasy Mini Kit (Qiagen) and RIN values were checked using an Agilent 2100 Bioanalyzer (Agilent). RNA-Seq analysis was performed on an Ion Proton System for next-generation sequencing (Thermo Fischer). For each of the samples, 10 ng of total RNA was reverse transcribed using the Ion AmpliSeq Transcriptome Human Gene Expression kit (Revision A.0) following the protocol of the manufacturer (Thermo Fisher) and as previously described 39 .
Sequences were aligned to hg19 AmpliSeq Transcriptome ERCC v1 and were quantified on gene level. Gene level reads were normalized according to total number of reads per sample. Data was analyzed using R statistical language (version 3.1.1). Genes with zero variance were removed from the data. The neuronal differentiation signature score was calculated as described by Fredlund and colleagues 2 . Differential expression was determined using unpaired t-test analysis in R using the 'stats' package (version 3.3.2). Gene Ontology (GO) analyses were performed on the top 250 upregulated genes in serum treated samples, using the Database for Annotation, Visualization and Integrated Discovery (DAVID) v6.8 tool 40 . A complete gene list of all genes present in the RNA sequencing data was used as the experimental background for GO analyses. Gene set enrichment analyses (GSEA) 41 was performed on a ranked list of all genes based on differential expression (t-statistic) between SC and 10% serum treatments using the c2.all.v5.0 curated gene set collection (Broad Institute MSigDB).