Bottom up proteomics reveals novel differentiation proteins in neuroblastoma cells treated with 13-cis retinoic acid

Abstract

Neuroblastoma, a cancer of the sympathetic nervous system, is the second most common pediatric cancer. A unique feature of neuroblastoma is remission in some patients due to spontaneous differentiation of metastatic tumors. 13-cis retinoic acid (13-cis RA) is currently used in the clinic to treat neuroblastoma due to its differentiation inducing effects. In this study, we used shotgun proteomics to identify proteins affected by 13-cis RA treatment in neuroblastoma SK-N-SH cells. Our results showed that 13-cis RA reduced proteins involved in extracellular matrix synthesis and organization and increased proteins involved in cell adhesion and neurofilament formation. These changes indicate that 13-cis RA induces tumor cell differentiation by decreasing extracellular matrix rigidity and increasing neurite overgrowth. Differentially-affected proteins identified in this study may be novel biomarkers of drug efficacy in the treatment of neuroblastoma.

1. Introduction

Neuroblastoma, a result of aberrant neural crest development, is the most common extracranial pediatric tumor affecting 700 children in the United States annually. Survival rates range from 90% in low-risk neuroblastoma cases to 30% in high-risk patients [2]. Half of high-risk patients relapse with no available cure, despite the development of intense and multimodal treatments [3]. Interestingly, in stage IVS neuroblastoma, where there are distant metastases, tumors regress due to spontaneous differentiation [4,5]. Localized adrenal neuroblastoma can also differentiate into a benign ganglioneuroma, even when left untreated [6]. 13-cis retinoic acid (13-cis RA), a vitamin A derivative, is currently used in the clinic post-chemotherapy due to its differentiating effects, specifically targeting minimal residual disease caused by a subset of dormant tumor-initiating cells [7,8]. Although use of 13-cis RA has improved outcomes, relapse is still seen in half of the high-risk patients. Thus, understanding development leading to differentiation is important.

Neuroblastoma is derived from neural crest cells (NCCs) and is thought of as an embryonal tumor, as the neuroblastoma tumors in the ganglia and/or adrenal gland primarily consist of neuroblasts, which have stem cell-like characteristics [9,10,11]. Neuroblastoma arises when neuroblasts are unable to proceed further into development and remain undifferentiated [12]. During normal development, trunk NCCs differentiate to develop the sympathetic nervous system. This is guided by transcription factors and environmental guidance cues. In gastrulation, single cells become more organized, forming the ectoderm, proceeding towards neurulation then delamination, next migration and finally terminal differentiation. Neurulation begins when the neural tube is formed. During delamination, a portion of the neural NCCs lose their adhesion to adjacent cells and migrate towards the dorsal aorta. During migration, NCCs proliferate and remain stem-like. Guidance cues will then trigger them to differentiate into sympathetic ganglia or chromaffin cells, which are found in the adrenal gland. The post mitotic mature neurons lose their proliferative and stem-like properties, leading to the formation of axons or neurites, the final step in differentiation [10,13]. A block in differentiation after delamination results in neuroblastoma [12].

Neuroblastoma cells are rounded, non-polarized neuroblasts with short or no projections [14]. As with other cancers, the degree of differentiation of neuroblastoma tumors directly correlates with positive clinical outcome and reduced oncogenesis [15,16,17,18,19]. In neuroblastoma, degree of differentiation is measured by length of neurites, presence of neurofilaments and synaptic vesicles, and a cone-like morphology [20]. Neurite length directly correlates with the degree of differentiation [21,22]. Growth cones at the polar end of the growing neurite are responsible for axonal elongation. This elongation is mediated by the growth cone and involves the orchestration of attractive/repellant cues, extracellular matrix (ECM), adhesion, and cytoskeletal proteins [23,24,25].

Retinoic acid (RA), a differentiation inducer, has three major isoforms: as all-trans retinoic acid (ATRA), 9-cis retinoic acid (9-cis RA), and 13-cis RA. RA is a master regulator of gene expression that is transported into the nucleus via cellular RA binding protein II (CRABP2) where it then binds to RA receptors (RARs) [26,27] to induce transcription of genes involved in differentiation [28,29,30]. ATRA and 13-cis RA treatment activates the growth cone of neuroblasts, leading to neurite outgrowth [31,32, 33]. Notably, vitamin A-deficient embryos have immature shorter axons in the central nervous system and peripheral nervous system, suggesting that RA is needed for normal axonal or neurite elongation [22]. 13-cis RA is used in the clinic to treat neuroblastoma [7]. Although ATRA has a higher RAR binding affinity than 13-cis RA [34], it is more lipophilic leading to a shorter half-life in vivo and less favorable pharmacokinetic parameters than 13-cis RA [30,35,36]. It is proposed that 13-cis RA isomerization to ATRA, the more active form of RA, provides prolonged efficacy of 13-cis RA [35,37].

Neuroblastoma cell lines can consist of S-type (substrate adherent), N-type (neuronal), and I-type (intermediate between N and S) cells. S-type cells are large flattened cells with no neuritic processes. N-type cells have short neuritic processes, tend to aggregate with each other, and are more tumorigenic than S-type cells. I-type cells share the same features as S- and N-type but have a round prominent nucleus, more cytoplasm, and most tumorigenic [38]. The SK-N-SH cell line is a good model for neuroblastoma since it contains all three types [39] and has been studied using RA to induce differentiation [26,40]. Retinoic acid has been shown to induce differentiation in N-type cells more than the less tumorigenic S-type cells, which are represented by longer neuritic processes [40]. S-type cells also have been shown to extend smaller processes than N-type [41].

The goal of this study was to obtain a comprehensive view of proteins affected in this differentiation network to better understand the relationships between prior studies that examined separate pathways of differentiation induced by 13-cis RA. The Neuroblastoma New Drug Development Strategy recommends drug combinations because the multiple aberrant pathways that lead to neuroblastoma make it unlikely that a single drug would target all or most of the aberrant pathways [42,43,44]. Creation of a comprehensive differentiation network of proteins affected by 13-cis RA should identify overlapping pathways that can be targeted to provide more effective treatments for neuroblastoma.

In this study, we used label-free tandem mass spectrometry (MS/MS) to measure the protein levels in SK-N-SH neuroblastoma cells with and without 13-cis RA treatment. Time-dependent changes in the levels of potentially important marker proteins were validated by western blotting. Pathway analysis was used to identify mechanisms that lead to differentiation. Important proteins that were differentially affected are involved in development, neurofilament bundle association, postsynaptic intermediate cytoskeleton organization, ECM/structure organization, and collagen biosynthesis.

2. Materials and methods

2.1. Neuroblastoma cell lines

SK-N-SH and IMR-32, human neuroblastoma cell lines, were obtained from the American Type Culture Collection of Rockville, MD. NB-1691 cells were a gift from Peter Houghton, St Jude's Children's Hospital, Memphis, TN. Cells were maintained in minimal essential medium supplemented with 10% heat-inactivated FBS, 1x MEM (nonessential amino acids), and 100 mM sodium pyruvate. Cells were incubated at 37°C in a humidity-saturated chamber containing 95% v/v air and 5% CO₂, and were passaged every six days upon reaching 90% confluency using 0.05% trypsin/0.53 mM EDTA. All cell washes were performed using PBS. All media components were from Corning Inc., Corning, NY.

2.2. Treatment with RA and lysis

Stock solutions of 1 mg/ml 13-cis RA (MilliporeSigma, Burlington, MA) in DMSO (Thermo Fisher Scientific, Waltham, MA) were stored at −80°C, thawed, and diluted in the growth media on the day of treatment. SK-N-SH, IMR-32, and NB1691 cells were seeded 24 hours prior to treatment, then cells were either treated with 10 μM RA in media or equivalent volume of DMSO (vehicle control). Medium-containing RA or DMSO was refreshed every three days. During sample collection, medium was removed, and cells were washed before re-suspending them into cold PBS. Cells were washed an additional three times before lysis.

Lysis was completed by performing three cycles of freezing the cell pellets on dry ice followed by 5 minutes of sonication, and manually grinding with microfuge pestles. Lysates were then centrifuged at 13,000 relative centrifugal field (RCF) for 10 minutes, and the protein concentration of each supernatant was determined via modified Bradford assay using Bio-Rad Protein Assay Reagent Concentrate (Bio-Rad Laboratories, Hercules, CA).

2.3. Microscopy analysis

SK-N-SH cells were seeded into 6-well plates at 7000 cells/well and treated as described previously. Live cell images were acquired at days 0, 3, 6, and 9 using a phase contrast microscope (Nikon Eclipse TS100, Minato, Tokyo, Japan) and OpenLab software, version 5.5.0 (Agilent, Santa Clara, CA). Neurites longer than 33 microns were quantified using three different fields of view. All neurites using three different fields of view were quantified to obtain neurite length and number of neurites at a select range.

2.4. Quantitative assessments of cell viability

SK-N-SH cells were seeded in 96-well plates at 1000 cells per well. Treatments were performed in triplicate as described previously. CellTiter-Blue Cell Viability Assay (Promega, Madison, WI) was performed according to the manufacturer’s protocol on days 3, 6, and 9. Statistical differences between samples were determined using a two-tailed Student’s t-test.

2.5. Quantitative assessments of cell growth rate and death

SK-N-SH cells were seeded into 96-well plates at 1300 cells per well. Treatments were performed in triplicates as described above. CellTox Green Cytotoxicity Assay (Promega) was performed by adding 1:500 CellTox into each well and measuring cell death on days 0-3. Next, cells were lysed at 0-3 days by adding 1:25 of lysis buffer into each well to obtain the proliferation rate. Statistical differences between samples were determined using a two-tailed Student’s t-test.

2.6. Mass spectrometry

2.6.1. Treatment and lysis for mass spectrometry

SK-N-SH cells were seeded at 10⁶ cells per 10-cm dish, and treatments were performed as described previously using five replicates per treatment group. At day 3, cells were lysed in 8 M urea, 50 mM Tris-HCl pH 8.0, containing 1x Halt Protease Inhibitor Cocktail and 1x Halt Phosphatase Inhibitor Cocktail (Thermo Fisher Scientific), as described previously.

2.6.2. Trypsinization of protein samples for mass spectrometry analysis

Protein (8 μg) was used for each digestion. Urea was diluted with 50 mM Tris-HCl pH 8.0 buffer to less than 1 M to allow for trypsin digestion. Dithiothreitol (DTT) was added to a final concentration of 5 mM, and samples incubated at 60°C for 1 hour. Iodoacetic acid (IAA) was then added to a final concentration of 10 mM and incubated for 30 minutes in the dark to alkylate cysteine side chains. An additional portion of DTT (2.5 mM final) was then added to quench the alkylation. Finally, Pierce Trypsin Protease MS Grade (Thermo Fisher Scientific) was added at a 1:20 protease-to-protein ratio, followed by overnight digestion at 37°C.

2.6.3. Purification

Resulting peptide solutions were purified via Pierce C18 Spin Columns (Thermo Fisher Scientific). 200 μL of activation solution (50% ACN and 0.2% formic acid [FA]) were added to condition the Pierce C18 Spin Columns. Columns were then centrifuged at 1500 g for 1 minute and the flow through discarded. This step was repeated to fully wet the resin. 200 μL equilibration solution (0.5% FA and 5% ACN) was then added to the spin column and centrifuged at 1500 g for 1 minute, discarding the flow through. This step was repeated to equilibrate the resin. One volume of sample diluent (2% FA and 20% ACN) was added to three volumes of protein sample and loaded onto the equilibrated resin. This was centrifuged at 1500 g for 1 minute and the flow through re-applied to the resin and re-spun. The column was then washed four times with equilibration solution (0.5% FA and 5% ACN). 20 μL of elution buffer (70% ACN and 0.5% FA) was added and centrifuged at 1500 g for 1 minute to elute digested peptides. This was repeated to complete elution. The combined eluted samples were dried in a vacuum evaporator.

2.6.4. Liquid chromatography-mass spectrometry

Peptide digestions were resuspended in 10 μL 0.1% TFA and analyzed on a Q Exactive Orbitrap interfaced with an Ultimate 3000 Nano-LC system (Thermo Fisher Scientific). The peptide samples were loaded onto an Acclaim PepMap RSLC column (75 μm x 15 cm nanoViper, Thermo Fisher Scientific) and eluted with a 150-minute gradient starting from 98% Solvent A (water containing 0.1% FA), 2% Solvent B (ACN containing 0.1% FA), and finishing at 5% Solvent A and 95% Solvent B at a flow rate of 0.3 μL/minute. Mass spectrometry data were acquired using a data-dependent top10 method, which dynamically chooses the most abundant precursor ions from the survey scan for higher-energy collisional dissociation fragmentation using a stepped normalized collision energy of 28, 30, and 35 eV. Survey scans were acquired at a resolution of 70,000 at m/z 200 on the Q Exactive.

2.6.5. Mass spectrometry data analysis

Mass spectrometry data sets were processed by MaxQuant software version 1.5.8.3 [45], using UniProt FASTA human database (version July 2017) with the built-in Andromeda search engine to identify proteins from the MS/MS peptide fragment data. The false discovery rate (FDR) for protein was set to 0.01. Search parameters included minimum peptide length of 7, a precursor mass tolerance of 20 ppm, fragment ion mass tolerance of 5 ppm, and up to two missed cleavages with trypsin. Fixed modifications included carbamidomethylation of cysteines and variable modifications included N-terminal acetylation, methionine oxidation, and phosphorylation. Label-free quantification (LFQ) [45], a built-in module in MaxQuant, was used to normalize protein hits and to quantify relative intensity.

2.6.6. LFQ statistical analysis

MaxQuant output was analyzed using Perseus software version 1.5.4.1 [46]. Reverse proteins and common contaminants were filtered. LFQ values were transformed to Log2 and missing values were then imputed from the normal distribution of the total matrix with default parameters of 0.3 width and 1.8 down shift. Two sample t-tests using permutation-based FDR were performed to identify differentially-altered proteins (DAP) comparing cell extracts from RA treated and vehicle controls. To identify the most significantly altered proteins, an FDR of 1% was used, with the parameter S0 set to equal 1. For GOrilla (Gene Ontology enRichment anaLysis and visuaLizAtion tool) analysis of pathways altered, a less stringent FDR of 5% was used, with the parameter S0 set to 0 as recommended by the Perseus software.

2.7. Bioinformatic analysis

Gene ontology enrichment analysis and visualization tool (GOrilla) [47] was utilized to classify biological processes, molecular function, and cellular compartmentalization. Gene ontology (GO) terms for DAP were identified as described previously using FDR 0.05, eliminating fold changes that were less than +/− 30%, were compared with GO terms in the total set of identified proteins to identify enriched processes.

2.8. Immunoblot

SK-N-SH cells were seeded in triplicate, treated, collected, and lysed on days 0, 1, 2, 3, 6, and 9 for vehicle control and 1, 2, 3, 6, and 9 for RA treated as discussed previously. Lysis was performed in 8 M urea, 2 M thiourea, and 1% 3-[(3-Cholamidopropyl) dimethylammonio]-1-propanesulfonate hydrate (CHAPS) 1x Halt Protease Inhibitor Cocktail and 1x Halt Phosphatase Inhibitor Cocktail as described previously. Protein concentrations were determined as described previously. SDS-PAGE was performed by loading and separating equal quantities (21 μg) of protein samples in Laemmli Sample Buffer (Bio-Rad Laboratories), 1:20 β-mercaptoethanol (BME), on precast 4% to 20% acrylamide gradient gels (Bio-Rad Laboratories). Proteins were then transferred to PVDF membranes using a wet transfer system at 60 volts for 75 minutes at 4°C. Blots were probed with antibodies against CRABP2 (R&D Systems, Minneapolis, MN) [48], NEFM (MilliporeSigma) [49], intracellular adhesion molecule 1 (ICAM1) (Cell Signaling, Danvers, MA) [50], plasminogen activator tissue type (PLAT) (Sigma-Aldrich, St. Louis, MO), and β-actin (MilliporeSigma). Membranes were then probed using goat-anti-rabbit IgG (Cell Signaling) and donkey-anti-goat IgG (R&D Systems) and incubated with SuperSignal West Femto Maximum Sensitivity Substrate or SuperSignal West Pico PLUS Substrate (Thermo Fisher Scientific). Bands were measured and analyzed using a C-DiGit Blot Scanner and Image Studio Lite Software version 5.2.5 (LI-COR Biosciences, Lincoln, NE). IMR-32 and NB1691 were treated similarly as SK-N-SH cells for three days.

2.9. Semi-quantitative PCR and quantitative RT-PCR

SK-N-SH cells were seeded on 6-well plates and treated as described above. Total RNA was isolated using the RNeasy Mini Kit (Qiagen, Germantown, MD) according to the manufacturer’s directions. First-strand cDNA was synthesized using the iScript cDNA Synthesis Kit (Bio-Rad Laboratories) according to instructions from the manufacturer.

Semi-quantitative PCR was performed using HotStart Taq polymerase (New England Biolabs, Ipswich, MA) under the following conditions: an initial denaturation of 30 seconds at 95°C, followed by 30 to 40 cycles of 30 seconds at 95°C, 30 seconds at 55°C, 1 minute at 72°C, and a final extension of 5 minutes at 72°C. Samples were then brought to 4°C. Cycle number was optimized for each target transcript and is listed as follows: CRABP2, 30; ICAM1 ,30; NEFM, 40; PLAT, 30; COL1A1, 30; COL3A1, 36; COL5A1, 35; COL18A1, 36. PCR product was run on 1% agarose gel in Tris-acetate-EDTA stained with GelRed (Biotium, Fremont, CA). GeneRuler 100bp DNA ladder (Thermo Fisher Scientific) was run alongside each sample to confirm predicted size of the amplified product. Images of each gel were acquired with the Chemidoc (Bio-Rad Laboratories).

Quantitative PCR was performed in 96-well plates on the Quant Studio 6 Flex (Applied Biosystems, Foster City, CA). Target transcripts were amplified by real-time PCR using the qMAX Green Low Rox qPCR Mix (Accuris, Edison, NJ) in 20 uL total volume under the following conditions: a 2-minute initial denaturation at 95°C, followed by 40 cycles of 5 seconds at 95°C, and 30 seconds at 60°C. The following primer sets (Integrated DNA Technologies, Coalville, IA) were used for both qRT-PCR and semi-quantitative PCR: CRABP2, (F) TTGAGGAGCAGACTGTGGATGG, (R) GTTCTCTGGTCCACGAGGTCTT; ICAM1, (F) GGCCGGCCAGCTTATACAC, (R) TAGACACTTGAGCTCGGGCA; NEFM, (F) ACAACCACGACCTCAGCAGCTA, (R) GTTGAGGAGGTCCTGGTATTCG; PLAT, (F) TGGTGCTACGTCTTTAAGGCGG, (R) GCTGACCCATTCCCAAAGTAGC; COL1A1, (F) GATTCCCTGGACCTAAAGGTGC, (R) AGCCTCTCCATCTTTGCCAGCA; COL3A1, (F) TCTTGGTCAGTCCTATGCGGATA, (R) AACGGATCCTGAGTCACAGACA; COL5A1, (F) GCATTTCCCGAGGACTTCTCC, (R) AATCTGCTGGATACCCTGCTC; COL18A1, (F) GGAGAGATTGGCTTTCCTGGAC, (R) CCTCATGCCAAATCCAAGGCTG. All samples were assayed in triplicate. Data from the target transcripts were normalized to the reference transcript, GAPD. Relative transcript levels were calculated using the comparative Ct method [51]. Error bars shown are standard error of the mean.

3. Results

3.1. 13-cis RA halted expansion of SK-N-SH cells and induced neurites

SK-N-SH cells were treated with 10 μM of 13-cis RA and viable cells quantified using CellTiter-Blue and CellTox cytotoxicity assay. Fluorescence produced using the CellTiter-Blue assay is directly related to the number of viable cells. After days 3, 6, and 9, cell numbers of viable cells in the untreated group continued to rise while the 13-cis RA treated group remained constant, indicating that cell division is arrested in the 13-cis RA group (Fig. 1A). Similarly, CellTox assay showed that untreated SK-N-SH cell numbers continue to rise over 0 to 3 days while 13-cis RA treated SK-N-SH cell numbers remain constant (Fig. 1B). Additionally, the cell death assay showed that cell death in 13-cis RA SK-N-SH cells was not significantly different from that seen in controls (Fig. 1C).Our data are consistent with previous studies showing that 13 cis-RA does not induce cell death in SK-N-SH cells but rather induces cell cycle arrest and subsequent differentiation [26,52], which is required for terminal differentiation [53].

Fig. 1.

13-cis RA inhibits proliferation and induces differentiation of SK-N-SH cells. A. Quantity of viable cells. SK-N-SH cells were treated with 10 μM 13-cis RA or vehicle control (DMSO) for up to 9 days. Relative quantity of viable cells were measured using a CellTiter-Blue assay. Quantities of treated cells relative to untreated cells are shown at each time point. Arrows represent where neurites are. Values represent the mean ± standard deviations of quadruplets over three different experiments. ** p < 0.01. B, C. Quantity of live and dead cells. Live and dead untreated and 13-cis RA treated SK-N-SH cells were quantified using CellTox assay. Quantities of treated cells relative to untreated cells are shown at each time point. Values represent the mean ± standard deviations of triplicates. D-F. 13-cis RA induces neurites in SK-N-SH cells over time. D. Phase contrast microscopy of untreated and 13-cis RA SK-N-SH cells over time. 10X magnification, bar 50 μm. E. % Neurites over 0-9 days. F. Length of neurites over 0-9 days. N- and S-type labeled (zero and one day).

Cell images also showed that 13-cis RA treatment altered the morphology of the cells, with increasing length of neurite-like extensions over time (Figs. D -F). Neuritogenesis was first evident on day 1 of treatment (Fig. 1D). Both N- and S-type 13-cis RA treated cells showed extension of processes and acquired a more cone-like appearance (Fig. 1D), consistent with differentiation [20].

3.2. MS/MS-identified proteins differentially affected by 13-cis RA in SK-N-SH cells

Liquid chromatography MS/MS MaxQuant analysis of proteins extracted from control and treated cells identified 3333 quantifiable proteins (supplemental Table S1). Levels of proteins in 13-cis RA-treated SK-N-SH were compared with proteins in untreated cells using Perseus software. Using a stringent analysis with an FDR of 1% and S0 parameter set to 1, 24 significantly DAP were identified (Fig. 2, Table 1). Using a less stringent analysis with an FDR of 5% and S0 parameter set to 0, 194 DAP were identified (supplemental Fig. S1, supplemental Table S2). The five replicates between each group were clearly separated in both 1% and 5%.

Fig. 2.

Differentially altered proteins of SK-N-SH cells treated with 13-cis RA for three days. A heat map of significantly altered proteins using 1% FDR S0=1 was generated in Perseus. Green indicates decrease and red indicates increase (Z-score log2). Euclidean hierarchal clustering designates clusters.

Table 1.

Proteins differentially altered in neuroblastoma cells treated after 3 days of 13-cis RA treatment (FDR 0.01) as identified by LC-MS.

Gene names	Protein names	Majority protein IDs	T-test q-valuea	Score	Peptides	Unique peptides	Sequence coverage, %	Mean untreated	Mean 13-cis RA	Fold change
CHGA	Chromogranin-A	P10645;G5E968	0.007	34.4	3	3	8.3	27.23 ± 0.4	24.47 ± 1	0.17
COL1A1	Collagen alpha-1(I) chain	P02452	0	323.31	54	52	54	34.39 ± 0.2	27.24 ± 0.6	0.01
COL3A1	Collagen alpha-1(III) chain	P02461	0	316.69	25	17	21.9	32.27 ± 0.1	24.79 ± 0.8	0.01
COL5A1	Collagen alpha-1(V) chain	P20908;A0A087WXW9	0.007	214.32	17	17	15.3	28.85 ± 1.1	22.64 ± 1.4	0.02
COL18A1	Collagen alpha-1(XVIII) chain	P39060;H7BXV5;H7C457	0.007	120.09	12	12	10.7	30.07 ± 0.7	25.9 ± 1	0.06
CTHRC1	Collagen triple helix repeat-containing protein 1	Q96CG8	0	178.92	11	11	54.7	31.03 ± 0.2	26.29 ± 0.5	0.04
EHD2	EH domain-containing protein 2	Q9NZN4	0.007	172.67	16	15	35.7	30.45 ± 0.3	27.04 ± 0.7	0.10
FSTL1	Follistatin-related protein 1	Q12841	0.005	45.476	5	5	21.1	28.31 ± 0.1	24.48 ± 1.2	0.09
ITGA5	Integrin alpha-5	P08648	0.01	26.866	4	4	5.2	26.53 ± 0.5	24.13 ± 0.2	0.18
MME	Neprilysin	P08473	0.005	78.435	10	10	19.9	27.93 ± 0.1	24.75 ± 1	0.13
PRRX1	Paired mesoderm homeobox protein 1	G3V2N3;P54821	0.006	39.388	5	5	26	28.79 ± 0.5	25.09 ± 1	0.09
SPARC	Osteonectin	P09486;F5GY03	0.003	72.363	9	9	46.9	28.51 ± 0.2	23.68 ± 1	0.04
SLIT2	Slit homolog 2 protein	A0A087WYV5	0.008	76.338	6	6	5.4	27.3 ± 0.6	23.82 ± 1.5	0.13
TCAF1	TRPM8 channel-associated factor 1	Q9Y4C2	0.006	32.5	5	5	7.5	26.72 ± 0.7	23.86 ± 0.6	0.13
CRABP2	Cellular retinoic acid-binding protein 2	P29373;Q5SYZ4	0.007	323.31	21	21	88.4	26.51 ± 2.5	34.93 ± 0.2	162.28
HMGA1	High mobility group protein	P17096	0.007	94.069	5	5	56.1	30.12 ± 0.5	33.18 ± 0.5	8.33
ICAM1	Intercellular adhesion molecule 1	P05362	0.003	139.06	14	14	28.2	23.52 ± 1.5	30.93 ± 0.1	114.26
ITGA1	Integrin alpha-1	P56199	0.022	110.23	10	10	11.9	25.04 ± 1.6	28.38 ± 0.2	6.76
NEFL	Neurofilament light polypeptide	P07196	0.006	99.514	14	10	25	25.61 ± 1	29.46 ± 0.5	12.74
NEFM	Neurofilament medium polypeptide	E7EMV2;E7ESP9;P07197	0	203.61	21	16	36.2	26.04 ± 0.6	31.01 ± 0.3	29.14
PLAT	Tissue-type plasminogen activator	P00750;B4DNJ1;B4DN26;E7ESF4	0.011	151.4	16	16	33.5	23.66 ± 0.8	26.45 ± 0.7	6.66
SCG2	Secretogranin-2	P13521	0	238.04	22	22	42.3	27.11 ± 0.5	31.16 ± 0.1	15.85
TGM2	Protein-glutamine gamma-glutamyltransferase 2	P21980;A2A299	0	323.31	51	51	71	33.65 ± 0.1	36.24 ± 0	5.98
TLE3	Transducin-like enhancer protein 3	F5H7D6	0	15.12	2	2	3.7	24.25 ± 0.9	27.1 ± 0.3	6.27

^aq=0 (<0.0001)

CRABP2, a protein that shuttles RA from the cytoplasm into the nucleus, was the most prominently increased protein (Fig. 2, Table 1), consistent with a known direct effect of RA treatment [26]. NEFL, NEFM, ICAM1, and SCG2, all of which have been individually shown to increase after RA-induced differentiation in carcinoma cell lines including SK-N-SH cells [54,55,56], were also increased in the stringent analysis. Interestingly, PLAT, a protein that has been proposed to facilitate neurite outgrowth [57], was also increased in RA-treated cells. A group of collagens and associated proteins were consistently reduced by RA treatment (Fig. 2, Table 1).

In the less stringent dataset, there were 109 decreased DAP (DDAP) and 85 up DAP (UDAP) (supplemental Table S2). The heat map (supplemental Fig. S1) shows that despite the lower stringency of this analysis, differences between the replicates of treated and untreated cells were consistent.

3.3. Validation of selected proteins

The proteomic data were validated by western blotting and qRT-PCR for a few key proteins. CRABP2, NEFM, and ICAM1 are important markers of neuronal differentiation [30,56,58] and all three were increased in the mass spectrometry dataset after 3 days of treatment. Western blotting showed that all three were statistically increased at 3 days (Figs. 3A and 3B) in SK-N-SH cells, consistent with the MS/MS data. Time course analysis showed that CRABP2 levels are significantly elevated by day 1 of treatment whereas elevation of ICAM1 and NEFM only becomes significant by day 2. PLAT was also increased on the protein level (Figs. 3A and 3B).PLAT is not known to function as a differentiation inducer; however, earlier studies did indicate that PLAT plays a role in neuroblastoma differentiation [59,60,61].Similarly, CRABP2, NEFM, and PLAT (NB1691) were increased in N-type neuroblastoma cell lines, IMR-32 and NB1691 (Figs. 3D and 3E). ICAM1 was not detectable in IMR-32 and NB1691 cells, with or without treatment.

Fig. 3.

Effect of 13-cis RA on protein and mRNA in SK-N-SH cells. A. Representative western blot of CRABP2, NEFM, ICAM in 10 μM of 13-cis RA-treated and control SK-N-SH cell proteins after up to 9 days of treatment. PLAT at 3 days of RA treatment. B. Quantitative analysis of relative levels of CRABP2, NEFM, PLAT, ICAM1 of three replicates of control and 13-cis RA treated SK-N-SH cells. * p< 0.05. C. Effect of 10 μM 13-cis RA on expression of mRNA levels of differentially expressed gene targets in SK-N-SH cells. SK-N-SH cells were treated with 10 μM 13-cis RA for 3 days. mRNA levels of CRABP2, PLAT, ICAM1, and NEFM were measured via quantitative RT-PCR with GAPD being used as a reference transcript. * p< 0.05, ** p< 0.01. D. Representative western blot of CRABP2, NEFM, PLAT in 10 μM of 13-cis RA treated and control IMR-32 and NB1690 cell proteins at 3 days of treatment. IMR-32 in triplicates and NB1690 in doubles. E. Quantitative analysis of relative levels of CRABP2, NEFM, PLAT of three replicates of control and 13-cis RA treated IMR-32 and two replicates of NB1690 cells. * p < 0.05.

CRABP2, ICAM1, and PLAT transcripts were also increased (Fig. 3C) but NEFM transcripts were not affected by RA treatment. Transcripts for the fibrillar collagens, COL1A1, COL3A1, COL5A1, were decreased (Fig. 4), consistent with the mass spectrometry protein data. However, transcripts for the nonfibrillar collagen, COL18A1, were not changed (Fig. 4) even though the mass spectrometry data showed a decrease in this protein. Among the proteins examined, NEFM and COL18A1 appeared to be the only ones regulated on the post-transcriptional level; whereas, CRABP2, ICAM1, PLAT, COL1A1, COL3A1, and COL5A1 are transcriptionally regulated, consistent with the known gene regulatory effects of RA.

Fig. 4.

Effect of 13-cis RA on expression of mRNA levels of differentially expressed collagens in SK-N-SH cells. mRNA levels of COL1A1, COL3A1, COL5A1, and COL18A1 were measured via quantitative RT-PCR with GAPD being used as a reference transcript. * p< 0.05

3.4. Bioinformatic analysis of altered proteins

GOrilla was used to determine potential interactions between DAP and to elucidate biological and molecular functions. To identify networks of proteins affected by 13-cis RA, the 194 proteins identified in the 5%-FDR analysis were further reduced to 138 proteins by eliminating proteins that changed by less than +/− 30%.

3.4.1. Gene ontology analysis

GOrilla was used for gene ontology analysis. DAP were analyzed against a background of all the genes in the dataset to reduce false positives. In the dataset, 3184/3333 proteins and 130/138 DAP were associated with at least one GO term. For the DAP, 60, 23, and 23 GO terms were associated with biological processes, molecular function, and cellular compartment, respectively (supplemental Table S3.1).

We then performed a second analysis in which 63 UDAP and 67 DDAP were compared with the total.

3.4.1.1. Biological processes

The most enriched biological processes were collagen biosynthesis, peptidyl-lysine hydroxylation, peptidyl-proline hydroxylation, protein hydroxylation, collagen metabolism, collagen fibril organization, regulation of cellular senescence, and regulation of cell aging, all of which were enriched more than 10-fold (Fig. 5, supplemental Table S3.1). The first six of these processes are related to ECM organization that was independently enriched 6.2-fold, with a more significant p-value when compared with these six processes (Fig. 5, supplemental Table S3.1). Developmental process was the most significantly affected process (supplemental Table S3.1, Fig. 5).Eighteen of 24 FDR 1% and 69/130 FDR 5% proteins were associated with the developmental process enrichment (supplemental Table S3.1). Directed acyclic graph (DAG) links produced in GOrilla showed that developmental processes are associated with regulation of cellular senescence and regulation of cell aging, the two highly-enriched processes not related to ECM organization (Fig. 5).

Fig. 5.

GOrilla web-server top GO biological process enrichments. 130 unranked DAP were matched to the entire data set. Top Panel: GOrilla performed DAG of top highly enriched BP. Bottom panel: A table with GO terms, description, p-value, q-value, enrichment value, and protein count.

The increased proteins were associated with processes related to development and cell maturation, with 31/63 proteins associated with development (supplemental Table S3.2). Although general differentiation was not identified as enriched in the UDAP analysis, other processes, such as postsynaptic intermediate filament cytoskeleton organization and neurofilament bundle assembly, were 50-fold enriched indicating specific upregulation of neuronal differentiation. Regulation of cellular senescence was 23-fold enriched, even greater than what was seen when all altered proteins were examined (Fig. 6, supplemental Table S3.2).

Fig. 6.

GOrilla web-server top GO UDAP and DDAP biological process enrichments. Processes enriched for 63 unranked UDAP matched to the entire data set are shown on the left and processes enriched for 67 unranked DDAP matched to entire set are shown on the right.

The reduced proteins were very significantly related to ECM organization, emphasizing the significance of the effect of 13-cis RA on these biological processes. Notably, collagen biosynthesis was 47-fold enriched (supplemental Table S3.2, Fig. 6). Unexpectedly, developmental processes were also enriched in the analysis of decreased proteins. However, one of these processes is negative regulation of cell differentiation (supplemental Table S3.2, Fig. 6), which would increase differentiation if decreased. Developmental enrichment linked to skin, epithelium, and bone trabecula formation were also linked to DDAP, indicating that 13-cis RA is not inducing differentiation into nonneuronal tissues (Fig. 6, supplemental Table S3.2).

3.4.1.2. Molecular function

Consistent with biological processes, the major increased molecular functions were related to postsynaptic intermediate filament cytoskeleton formation and decreased molecular functions were related ECM structural functions.

Structural constituent of postsynaptic intermediate filament cytoskeleton was enriched 50-fold, emphasizing the significance of this molecular function (Fig. 7, supplemental Table S3.2). Procollagen lysine-5 and proline-3-dioxygenase activity, extracellular matrix structural constituent were the most affected and collagen biosynthetic process, peptidyl and proline hydroxylation, collagen fibril organization and collagen metabolic process were all enriched more than 20-fold in the DDAP analysis (Fig. 7, supplemental Table S3.2).

Fig. 7.

GOrilla web-server top GO UDAP and DDAP molecular function enrichments. Processes enriched for 63 unranked UDAP matched to the entire data set are shown on the left and processes enriched for 67 unranked DDAP matched to entire set are shown on the right.

3.4.1.3. Cellular component

The most enriched cellular components associated with the UDAP analysis, neurofilament (37-fold), intermediate filament cytoskeleton (13.5-fold), and Schaffer collateral-CA-1 synapse (9-fold) were all associated with differentiated neuronal cells (Fig. 8, supplemental Table S3.1). All of the cellular components associated with DDAP such as fibrillar collagen, collagen trimer, and ECM organization (more than 20-fold enriched) were associated with the ECM (Fig. 8, supplemental Table S3.2).

Fig. 8.

GOrilla web-server top GO UDAP and DDAP cellular component enrichments. Processes enriched for 63 unranked UDAP matched to the entire data set are shown on the left and processes enriched for 67 unranked DDAP matched to entire set are shown on the right.

4. Discussion

In this study, we used MS/MS for global proteomic analysis of proteins affected by a 3-day 13-cis RA treatment of neuroblastoma SK-N-SH cells. This study identified 3333 different proteins, of which 24 and 194 were differentially affected when using a 1% FDR and 5% FDR, respectively.

The most changed protein, CRABP2, an RA nuclear transporter, was increased by 162 fold when compared with untreated SK-N-SH cells (Fig. 3, Table 1). CRABP2 was similarly increased in IMR-32 and NB1691. This is consistent with previous studies showing that RA upregulates CRABP2 in neuroblastoma cells [26]. CRABP2 is associated with development and differentiation [29,30,62].

More than half of the 5% DAP and 75% of the 1% DAP were linked to development (Fig 5, supplemental Tables S3.1, S3.3, and S3.4), with 27% being developmental regulators (13 UDAP/23 DDAP) and 18% involved in cell differentiation (10 UDAP/13 DDAP) (supplemental Table S3.1, S3.3 and S3.4). The FDR 1% proteins involved in development are UDAP: ICAM1, NEFL, ITGA1, TGM2, HMGA1, CRABP2, and SCG2 and DDAP: PRRX1, COL18A1, COL1A1, COL3A1, COL5A1, SPARC, CTHRC1, MME, ITGA5, SLIT2, and TCAF1. GOrilla identified COL3A1 and COL5A1 as negative regulators of cell differentiation, so lower levels would increase differentiation. All of the proteins associated with the negative regulators of cell differentiation enrichment were reduced by RA treatment.

Other biological processes enriched were regulation of cellular senescence, neurofilament bundling association (both increased), ECM organization, collagen metabolic process, and protein hydroxylation (all three reduced) (Figs. 5 and 6). All of these processes are related to development and/or differentiation. Neuronal cell morphology is controlled by coordination of ECM, cell membrane, and cytoskeleton.

Increased ECM tension is inversely related to differentiation [63], neuroblastoma survival [64], and neurite outgrowth [65,66,67,68,69], so lower levels of fibrillar collagens would support differentiation and neurite outgrowth. Interestingly, when neuroblastoma cells plated on collagen were treated with collagenase, rapid extension of neurites was observed [65]. Furthermore, collagen accumulation is correlated with tumor formation in glioblastoma [70] and in breast cancer [71,72,73], and COL1, a fibrillar collagen, is directly correlated to oncogenesis in many cancers [74]. Collagen processing is also directly correlated to oncogenesis, indicating that reduced levels of collagen could also arrest tumor development [75,76]. qPCR showed that the fibrillar collagens were transcribed less in the treated cells vs. the control (Fig. 4), indicating that regulation of these proteins occurs at the transcriptional level. Proteins involved in collagen biosynthetic/assembly processes identified by GOrilla (LEPRE1, LEPREL2, PLOD1, PLOD3, P4HA1, and SERPINH1) were all reduced by treatment (supplemental Table S2). Increased expression of PLOD3 [75], P4HA1 [76], and SERPINH1 [77] have been detected in many cancers. Additionally, 13-cis RA reduced levels of SPARC, a secreted calcium binding glycoprotein, which binds to collagen and is involved in collagen assembly [78]. SPARC also affects collagens intracellularly by influencing procollagen processing and, in turn, collagen aggregation [79]. GOrilla did not link SPARC to collagen biosynthetic/assembly processes but instead linked SPARC to ECM structure/organization. Our results indicate that tension of the ECM is reduced by RA treatment, allowing cellular flexibility for neurite outgrowth during differentiation.

The process of neurofilament bundling was highly enriched by RA treatment (Fig. 6), and is a hallmark of differentiation of neuroblasts [20,54,80,81,82,83]. Differentiation is directly correlated with the length of axons or neurites [84]. The intermediate filaments NEFL and NEFM, which co-assemble along the axon as it elongates [85], are both upregulated after 13-cis RA treatment (Figs. 2, 3A, and 3B NEFM). This is consistent with previous studies that show neuroblastoma treated with RA upregulates NEFL [54] and NEFM [81]. A protein in the 5% FDR UDAP, INA, is an intermediate filament involved in neuronal morphogenesis and is thought to be the fourth intermediate filament. INA coexists with neurofilaments as the neurite extends [86,87]. INA and NEFL are involved in postsynaptic intermediate cytoskeleton organization enrichment (Fig. 6, supplemental Table S3.2),indicating that the RA-treated cells have differentiated into a mature neuronal phenotype.

Many of the DAP responsible for development are involved in ECM remodeling and neurite outgrowth (supplemental Table S3.1). Neurite outgrowth does not only involve neurofilaments but also the coordination of neurofilaments/ECM via adhesion proteins [88], which are the primary navigators [89] at growth cones [25,90], allowing for pathfinding and elongation [91,92,93,94]. Adhesion proteins interact with ECM to gain adequate traction, which in turn allows for cytoskeletal rearrangements during differentiation [95,96]. Integrins can have a positive or negative effect on development and differentiation depending on the type of integrin [97,98]. ITGA1 was increased with RA treatment (Fig. 3). ITGA1 previously has been shown to increase after RA treatment of neuroblastoma cells leading to neurite outgrowth [99,100]. When ITGA1 binds to its preferred ligand, type IV collagen (COLIV) [97,101], it promotes neurite outgrowth in the peripheral nervous system [102]. COLIV was not detected in our dataset, so we do not know whether increased levels of ITGA1 induces differentiation by binding to this or related matrix proteins. However, we did detect increases in two intracellular adapter proteins, talin and paxillin (supplemental Table S2). Recruitment of such adaptor proteins links ligand-activated integrins such as ITGA/ITGB complexes to the actin cytoskeleton, promoting neurite outgrowth [68,103]. Conversely, ITGA5 was reduced with RA treatment (Fig. 2, Table 1). FN1, the primary ligand for ITGA5 [97,104,105], was also reduced with treatment (supplemental Table S1). Levels of ITGA5 directly correlate with neuroblastoma aggressiveness [106] and oncogenesis in general [98,107,108], so reduced levels may inhibit oncogenesis. FN1 is also directly correlated with oncogenesis [109].

ICAM1, another adhesion protein, increased with RA treatment at both the protein and transcript level (Figs. 2, 3A-C, Table 1). This increase is consistent with previous studies showing that RA increases ICAM1 in SK-N-SH neuroblastoma cells [56] and in other cancer cells [110]. ICAM1 was not detectable in N-type cells, IMR-32 and NB1691 (data not shown). This may be due to N-type cells not being overly abundant in ICAM1, consistent with other studies [56].

An intriguing discovery in this study is that cis-13 RA increases cellular levels of PLAT, both at the transcript and protein level (Figs. 2, 3A-C), which was also confirmed in N-type cell NB1691 (Figs. 3D and 3E). PLAT is secreted at the growth cone in growing neurites [111,112]. PLAT previously has been shown to increase in neuroblastoma cells treated with RA and was proposed to induce axonal elongation during neuronal differentiation [59,60]. PLAT enhances neurofilament expression during sciatic nerve repair by degrading fibrin and collagen, components of the ECM [113], indicating that it is an important player in neurodifferentiation.

13-cis RA is currently used in the clinic to target minimal residual disease, which in neuroblastoma is associated with tumor initiating cells or tumor stem cells [8,112]. Proteins that are elevated in tumor stem cells such as collagens, EHD2, FSTL1, neprilysin, PRRX1, and SPARC [113,114,115,116,117,118,119], are all decreased by 13-cis RA in our study indicating that RA can affect cancer stem cells in addition to inducing differentiation in dividing neuroblastoma cells. Indeed, RA induces differentiation of stem cell-like I-type neuroblastoma, causing increased expression of SCG2 and NEFM [125]. Both of these proteins were increased in our 1% data set. Increased levels of COL1A1, COL3A1, COL5A1, and PCOLCE are associated with cancer stem cells and increased tumor recurrence [121]; all were reduced with RA treatment in our study. CHGA, a cancer stem cell marker in neuroblastoma patients [122,123], is also reduced in our study.

In summary, our data show that multiple pathways of differentiation are induced by treatment of neuroblastoma cell lines with 13-cis RA, including neurofilament bundle association and neurite outgrowth. CRABP2, an RA nuclear transporter [26,27] was significantly increased within 24 hours of treatment, whereas NEFM, ICAM1, and PLAT gradually increase over 3 days. Increased transport of RA, either as 13-cis RA or its isomer ATRA, into the nucleus enhances its effect on induction of expression of proteins [26,28,30]. Notably, our study showed that 13-cis RA treatment reduced levels of fibrillar collagens, both at the transcript and protein level. Collagens have been positively correlated with breast cancer aggressiveness (COL1A1 and COL3A1) [71] and glioblastoma aggressiveness (fibrillar collagens) [70], but our study is the first to show reduced levels of collagens in treatment of neuroblastoma. Reduced rigidity of the ECM can allow cell adhesion membrane proteins to trigger extension of neurites during differentiation. We showed that the adhesion proteins ITGA1 and ICAM1; adapter proteins talin and paxillin; and cytoskeleton proteins NEFM, NEFL, and INA were all increased by 13-cis RA treatment, creating a network of proteins to induce differentiation. Intriguingly, many markers of cancer stem cells were reduced by 13-cis RA treatment, indicating that this therapy may have a direct effect on cancer stem cells. This proteomic study of the global effect of 13-cis RA on proteins in neuroblastoma should aid development of combinatorial therapies that are more effective in the treatment of this disease and indicates that 13-cis RA may also be of value in treatment to eradicate cancer stem cells.

Supplementary Material

fS1

Fig. S1. Differentially-altered proteins of SK-N-SH cells treated with 13-cis RA for three days. Perseus heat map of significantly-altered proteins. Green indicates decrease and red indicates increase (Z-score log2). Euclidean hierarchal clustering designates clusters. Heat map of 5% FDR S0=0.

Highlights.

• 13-cis RA increased proteins in the pathway that regulates neurite outgrowth.

• Rigid fibrillar collagens (COL1A1, COL3A1, COL5A1) are all reduced by 13-cis RA.

• Collagen biosynthetic and stabilization proteins are all reduced by 13-cis RA.

• Reduced ECM rigidity may be a novel mechanism to induce tumor differentiation.

• CRABP2, PLAT, NEFM, and ICAM1 were identified as proteins in the regulation of differentiation that were significantly increased by 13-cis RA.

Significance.

As neuroblastoma can spontaneously differentiate, determining which proteins are involved in differentiation can guide development of novel treatments. 13-cis retinoic acid is currently used in the clinic as a differentiation inducer. Here we have established a proteome map of SK-N-SH cells treated with 13-cis retinoic acid. Bioinformatic analysis revealed the involvement of development, differentiation, extracellular matrix assembly, collagen biosynthesis, and neurofilament bundle association. This proteome map provides information as to which proteins are important for differentiation and identifies networks that can be targeted by drugs to treat neuroblastoma [1].