Benchmarking In-Cell Proteomics for Profiling Neuroblastoma Cell Differentiation and the Ubiquitin-Proteasome System

Abstract

The human neuroblastoma SH-SY5Y cell line is a widely utilized model for studying neurodegenerative diseases, owing to its ability to differentiate into cells with a neuron-like phenotype. However, a comprehensive understanding of the cellular and molecular mechanisms underpinning the SH-SY5Y cell differentiation and maturation is still lacking from a deep proteomics perspective. We systematically benchmarked an “in-cell proteomics” strategy against the sodium dodecyl sulfate (SDS) lysate-based processing method and showed superior performance in terms of simplicity, sensitivity and quantitation accuracy while requiring minimal inputs. Next, We employed the in-cell proteomics strategy to characterize SH-SY5Y cells at undifferentiated, and partially and terminally differentiated states, respectively. Among over 9000 proteins identified in total, we were able to detect marker proteins in neuronal development and integrity, and observed increases in glutamatergic synapse-related proteins and proteins previously reported in mature neurons as well as in differentiated neuroblastoma cells. Lastly, we examined proteins involved in the ubiquitin-proteasome system and found stage-specific expression of E3 ubiquitin ligases, deubiquitinases (DUBs), and proteasome subunits, revealing an important role of protein homeostasis in neuroblastoma cell differentiation. In summary, our study presented the first benchmark data set of neuroblastoma cells using an in-cell proteomics strategy, and demonstrated its great potential in cataloging neuronal function and disease.

Graphical Abstract

INTRODUCTION

In vitro neuronal cell models are essential for studying neurodevelopment and neurodegenerative diseases, offering an easily accessible and reproducible system to explore the cellular and molecular mechanisms of the central nervous system (CNS).^1–3 Several models have been reported, including immortalized cell lines such as HT22,⁴ PC12⁵ and LUHMES (Lund Human Mesencephalic Cells),⁶ human induced pluripotent stem cells (iPSCs)^{7, 8} and cells derived from tissues such as primary neurons from rodent brain hippocampus and cortex.^{9, 10} The human neuroblastoma SH-SY5Y cell line is a widely adopted model for studying neurodegenerative diseases (NDs) such as Alzheimer’s (AD) and Parkinson’s diseases (PD).^{11, 12} However, there are limitations in using undifferentiated neuroblastoma cells for neurobiological studies as they are inherently cancerous and lack the expression of key markers found in mature neurons. Furthermore, undifferentiated SH-SY5Y cells express proliferation markers and immature neuronal markers such as proliferating cell nuclear antigen (PCNA) and nestin, thus limiting their suitability for studies requiring mature neurons.¹³ Hence, various protocols and reagents have been employed to induce SH-SY5Y differentiation into cells morphologically and functionally similar to neurons for neurodegenerative disease research and drug development.^13–20

Retinoic acid (RA), a vitamin A derivative, is widely used to induce SH-SY5Y cell differentiation. It activates the PI3K/AKT signaling pathway, increasing noradrenalin (NA) concentration and neurite outgrowth.¹⁵ It also induces the expression of neuronal markers such as neuronal nuclei (NeuN), neuron-specific enolase (NSE), synaptophysin (SYP), and synapse-associated protein-97 (SAP97) in SH-SY5Y cells.¹⁸ Another reported protocol for neuroblastoma differentiation involves initial treatment with RA, followed by BDNF (brain-derived neurotrophic factor), in a serum-free medium to improve neuronal phenotypes. BDNF is known to modulate synaptic plasticity by increasing synapsis transmission and stimulating local translation of crucial synapsis components.^{21, 22}

Quantitative proteomics analysis of undifferentiated and differentiated SH-SY5Y cells provides crucial information for our understanding of the protein factors and pathways that contribute to the SH-SY5Y neuronal transformation.^{23, 24} Barth et al. quantified around 3500 proteins after comparing undifferentiated with RA- and RA-PMA (phorbol-12-myristate-13-acetate)-differentiated SH-SY5Y cells.²⁴ The authors reported that undifferentiated cells possess cancer-like characteristics, while both differentiated cell groups showed structural reorganization and a high level of lysozyme expression. However, this study did not take into account the integral membrane proteins, which might lead to an incomplete understanding of differentiation at the molecular level. In another study, iTRAQ-based quantitative proteomics was employed to assess protein expression level in undifferentiated and RA-BDNF-differentiated SH-SY5Y cells, with the identification of more than 5500 total proteins and 1321 phosphoproteins²³. Gene Ontology (GO) enrichment analysis indicated upregulation of proteins related to neuronal development and the importance of apoptosis and attachment to the extracellular matrix as prominent biological processes due to differentiation. The authors also identified potential markers for neuronal differentiation including myristoylated alanine-rich C-kinase substrate (MARCKS), stathmin 1 (STMN1), apoptosis inducing factor mitochondria associated 1 (AIFM1), survival motor neuron protein (SMN1), agrin (AGRN) and catenin δ-1 (CTNND1).²³

To date, few proteomics analyses of neuroblastoma SH-SY5Y cells have focused on the ubiquitin-proteasome system (UPS), the impairment of which has been associated with the pathophysiology of neurodegenerative diseases.^{25, 26} It is not yet clear how the abundance of E3 ubiquitin ligases, deubiquitinases (DUBs), and proteasome subunits change upon differentiation of the SH-SY5Y neuroblastoma cells. To address these questions and unveil the important role of protein homeostasis in the differentiation of neuroblastoma cells, in this study, we carried out a deep quantitative proteomics analysis of SH-SY5Y cells at different stages of differentiation. We benchmarked an “in-cell proteomics” strategy based on a recently developed E4technology,²⁷ in combination with label-free quantitation and data-independent acquisition mass spectrometry (DIA-MS) to systematically assess the proteome-wide profiles of undifferentiated, partially and terminally differentiated SH-SY5Y cells. Upon obtaining a deep proteome coverage, we analyzed protein expression across various differentiation stages and revealed elevated levels of proteins involved in neuronal development and differentiation, particularly proteins associated with glutamatergic synapses. In addition, we observed stage-specific changes in the abundance of E3 ubiquitin ligases, DUBs and proteasome subunits. Our results also indicate that RA-treated neuroblastoma cells also exhibit an intermediate phenotype, highlighting the need for BDNF treatment to achieve terminal differentiation.

EXPERIMENTAL SECTION

Differentiation of SH-SY5Y Cells

To perform SH-SY5Y cell (ATCC, CRL-2266) differentiation, we followed previously reported protocols with additional modifications.^{24, 28} Specifically, the cells with passage number <10 were grown to 20-30% confluency in collagen I-coated plate (Gibco) and maintained in basic growth medium consisting of DMEM high glucose (Corning, with L-glutamine, 4.5 g/L glucose and sodium pyruvate), 10% (v/v) heat-inactivated fetal bovine serum (hiFBS; Sigma-Aldrich), 1% GlutaMAX-I supplement (Gibco), and 1% penicillin-streptomycin 100x solution (Cytiva, 10,000 U/mL Penicillin, 10 mg/mL Streptomycin). Cells were cultured in a humidified incubator at 37 °C with 5% CO₂. To induce partial differentiation of SH-SY5Y cells, the culture medium was replaced with Medium I, containing DMEM-high glucose, 2.5% (v/v) hiFBS, 1% GlutaMAX-I Supplement, 1% penicillin-streptomycin 100x solution and 10 μM final concentration of all-trans retinoic acid (RA; Thermo Scientific). Fresh Medium I was added on days 2 and day 4 of differentiation to replace the old medium. After 5 days, cells in some wells were collected at this step following trypsin treatment, while the cells in the remaining wells were allowed to continue to differentiate by replacing the medium with Medium II, composed of Neurobasal-A medium (Gibco) supplemented with 50 ng/mL Recombinant Human BDNF (Bio-Techne), 1x serum free B-27 supplement (Gibco), 1% GlutaMAX-I, 20 mM syringe filtered potassium chloride, and 1% penicillin-streptomycin 100x solution. The medium was changed on days 7 and day 9 to maintain medium freshness. Terminally differentiated cells were collected on day 11 following trypsin treatment. Phase contrast images of the cells at various stages of differentiation were captured using a microscope with a 20x objective.

Cellular Protein Extraction and Western Blotting (WB)

Cells in various stages of differentiation were collected through centrifugation for 10 min at 2,000 rpm. The pellet was washed three times with Dulbecco’s Phosphate-Buffered Salt Solution (DPBS, Corning), and the cells were then resuspended in cold DPBS buffer containing Halt Protease Inhibitor Cocktail (Thermo Scientific) and sonicated for 10 s to lyse the cells. The lysed cells were then centrifuged at 4 °C, 14,000 rpm for 30 min, and the total protein concentration of the supernatant was measured using a BCA assay. Cell lysates were separated using sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE, 15%) and then transferred to a methanol-activated polyvinylidene (PVDF) membrane overnight. Membrane was blocked in 5% milk in 1x Tris-buffered saline-Tween 20 (TBST), containing 10 mM Tris–HCl (pH 7.6), 150 mM NaCl, and 0.1% Tween-20, at 4 °C for 2-6 h. It was then incubated with primary antibodies (1:1,000) from Mature Neuron Marker Antibody Sampler Kit (Cell Signaling Technology) and PCNA antibody (PC10, Santa Cruz) in 5% milk at 4 °C overnight. The membrane was subsequently washed with TBST and incubated with secondary antibodies (1:3,000) for approximately 2 h at room temperature. It was treated with ECL (chemiluminescence) WB substrate and analyzed using the FluorChem R System. GAPDH was used as the loading control.

Proteomics Sample Preparation

E4tip-Based Protocol.

Cultured cells were rinsed three times with cold DPBS and then resuspended in 1 mL of cold PBS buffer. Cell numbers were counted by using a hemocytometer under a light microscope. For the initial evaluation experiment, around 60,000 undifferentiated cells were used for each sample preparation method, with five replicates per method. Cells were added to E4tips (CDS Analytical, Oxford, PA) prefilled with 200 μL of methanol, and incubated at 4 °C for 5 min. The tips were then centrifuged at 3,000 rpm for 2 min. The cells retained in the tip were washed once with 200 μL of methanol and then subjected to reduction and alkylation in 100 μL of 50 mM triethylammonium bicarbonate (TEAB) buffer containing 10 mM tris(2-carboxyethyl)phosphine (TCEP) and 40 mM 2-chloroacetamide (CAA) for 15 min at 45 °C with shaking at 200 rpm. The tips were then centrifuged at 3,000 rpm for 2 min and washed with 200 μL of TEAB buffer. Next, 0.2 μg of trypsin/Lys-C mix in 150 μL of TEAB buffer was added to the E4tip and incubated at 37 °C for 16–18 h with gentle shaking. To quench the digestion, formic acid was added to the tip at a final concentration of 1% (v/v). The tips were then centrifuged at 1,500 rpm for 10 min, followed by a wash step with 200 μL of 0.5% acetic acid in water (v/v) at 3,000 rpm for 2 min. The resulting peptides were eluted sequentially with 200 μL of a mixture of 0.5% acetic acid and 60% acetonitrile (ACN) in water, followed by 200 μL of a mixture of 0.5% acetic acid and 80% ACN in water. The eluted peptide solutions were combined and dried in a SpeedVac and stored at −80 °C until further analysis.

E3tip-Based Protocol.

For E3tip-based digestion, the procedure was the same as described above except that, following digestion, the E3tip was stacked onto another C18 StageTip pre-activated with methanol and equilibrated with 0.5% acetic acid in water. The stacked tips were then centrifuged at 1,500 rpm for 10 min, followed by a wash step using 200 μL of 0.5% acetic acid in water and a quick centrifugation at 3,000 rpm for 2 min. The resulting peptides were eluted with 200 μL of a mixture of 0.5% acetic acid and 60% acetonitrile (ACN) in water, followed by 200 μL of a mixture of 0.5% acetic acid and 80% ACN in water and stored in the same way as described above. Note that the tip-based in-cell proteomics methods are designed for low-cell and low-input proteomics. When large amounts of samples are processed, E3/E4 spin column filters, cartridges and filter plates may be used instead, which are commercially available from CDS Analytical (Oxford, PA).

SDS Lysate Digestion Protocol.

Cells were transferred to a low-binding tube and resuspended in a 100 μL Tris-HCl solution (100 mM, pH 8.0) containing 5% (w/v) sodium dodecyl sulfate (SDS), 10 mM TCEP, and 40 mM CAA. The mixture was vortexed briefly, and then boiled at 95 °C for 10 min. Next, the cell lysate was mixed with 4x volume of 80% ACN aqueous solution and transferred to an E3tip. Following a centrifugation step, the E3tip was washed with 200 μL of 80% ACN aqueous solution three times. For protein digestion, 0.2 μg of trypsin/Lys-C mix in a 150 μL of TEAB buffer solution was added to the E3tip and incubated at 37 °C for 16-18 h with gentle shaking. To stop the digestion, formic acid was added to the tips at a final concentration of 1% (v/v), then the E3tip was stacked onto a preactivated C18 StageTip. Peptides were eluted and stored as described above. For the proteomics sample preparation experiments, the E3tips, E4tips and C18 StageTip were all purchased from CDS Analytical (Oxford, PA). Trypsin-Lys-C mix was obtained from Promega (Madison, WI). Tris(2-carboxyethyl) phosphine (TCEP), chloroacetamide (CAA), and triethylammonium bicarbonate (TEAB) were all obtained from Fisher Scientific.

LC-MS/MS Analysis

The LC-MS/MS analysis was performed using an UltiMate 3000 RSLCnano system coupled to an Orbitrap Eclipse mass spectrometer and a FAIMS Pro Interface (Thermo Scientific). The dried peptides were resuspended in 15 μL of LC buffer A (0.1% formic acid in water) and 10 μL of the peptide solution was then loaded onto a trap column (PepMap100 C18, 300 μm × 2 mm, 5 μm; Thermo Scientific). Separation was performed on an analytical column (PepMap100 C18, 50 cm × 75 μm i.d., 3 μm; Thermo Scientific). A linear LC gradient of 1-25% mobile phase B (0.1% formic acid in ACN) was applied at a flow rate of 250 nL/min over 125 min, followed by an increase to 32% mobile phase B over 10 min. For the ion source, the spray voltage was set to 1.9 kV, funnel RF level at 50%, and heated capillary temperature at 275 °C. The MS data were acquired on Orbitrap in positive ion mode at a resolution of 120,000, followed by MS/MS acquisition in data-independent (DIA) mode following a protocol described previously²⁹. The MS scan range (m/z) was 380–985 with a maximum injection time of 246 ms, and normalized Automatic Gain Control (AGC), target of 100%. The MS/MS data were acquired using an Orbitrap at a resolution of 30,000. The isolation mode was set to Quadrupole, and the isolation window was 8 m/z with an overlap of 1 m/z. The collision energy was set to 30%, and the Automatic Gain Control (AGC) Target was 400,000, and the normalized AGC target was 800%. For FAIMS compensation voltages (CV), a 3-CV experiment (−40, −55, and −75 V) was applied.

Proteomics Data Analysis of Peptides Derived from SH-SY5Y Cells

The raw data were processed using Spectronaut with a library-free directDIA+ workflow against the UniProt human protein databases (20,435 sequences; reviewed only). The search parameters included protein N-terminal acetylation and methionine oxidation as variable modifications, and carbamidomethylation on cysteine as a fixed modification. The cleavage rules were set to Trypsin/P with a minimum peptide length of 7 amino acids while allowing 2 missed cleavages. FDR 0.01 was set at PSM, peptide, and protein levels. The Protein LFQ Method was set to MaxLFQ with cross-run normalization and an MS2 level for quantitation. The MS raw files associated with this study have been deposited to the MassIVE server (https://massive.ucsd.edu/) with the dataset identifier MSV0000 97787.

The proteomics data analyses of samples derived from SH-SY5Y cells at different stages of differentiation (Undiff, RA, and RA-BDNF) were performed as follows. The raw data matrix was exported from Spectronaut and imported into Perseus version 1.6. Categorical annotations were added to replicates in each treatment group for downstream statistical analysis. Coefficients of variation (CV) were calculated groupwise using the raw LFQ values. Violin plots were generated in GraphPad Prism. The same analysis was performed at the peptide level and plotted in the same manner. The data were then transformed into Log₂ scale and filtered based on valid values (3 out of 4 within at least one group). Missing values were imputed using a left-shifted Gaussian distribution (width = 0.3, downshift = 1.8). Gene lists were manually curated based on the existing literature.

For heatmaps, a one-way ANOVA test was conducted based on the assigned treatment groups, using a Benjamini-Hochberg FDR value of 0.01. The data were then filtered based on significance (indicated by a “+” categorical annotation) and Z-score normalized (unless otherwise indicated in figure legends). Hierarchical clustering was performed on rows with Euclidean distance, average linkage, and no preprocessing by k-means.

For volcano plots, following the categorial annotation step in Perseus, a new workflow branch was created for each relevant pairwise comparison (Undiff vs RA, Undiff vs RA-BDNF, RA vs RA-BDNF) by using the reorder/remove columns function. The data were then Log₂ transformed, filtered based on valid values, and missing values replaced with data imputation as described above. A volcano plot was then created for each pairwise comparison using a t-test (both sides) with the default permutation-based FDR of 0.05. The resulting matrices were exported to RStudio for plotting with the package “EnhancedVolcano”.

All Gene Ontology (GO) enrichment analyses were performed using the functional annotation tool DAVID.³⁰ The UniProt accessions corresponding to all identified proteins were submitted as queries. For each GO_CC term, the iBAQ values for the associated proteins were summed and divided by the total iBAQ sum of all proteins identified in the proteomics run to obtain a percentage. To reduce redundancy of GO_BP terms, REVIGO³¹ was utilized with the SimRel semantic similarity metric. Terms were sorted by p-value and redundancy was filtered based on dispensability < 0.05. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment of differentially upregulated proteins from each pairwise comparison was performed in RStudio using the package enrichR queried against the KEGG_Human_2021 database. Pathways were ordered based on the highest combined score and filtered by p-value < 0.05 and hit counts ≥ 5.

RESULTS AND DISCUSSION

Differentiation of SH-SY5Y Cells Using RA and BDNF

In this study, we compared SH-SY5Y cells at three different states: undifferentiated, partially differentiated with RA treatment, and terminally differentiated cells with a sequential RA and BDNF treatment. Differentiation was achieved based on recently reported protocols,^{28, 32} following a workflow shown in Figure 1A. Briefly, SH-SY5Y cells were resuspended in DMEM medium and evenly distributed into individual wells (approximately 100,000 cells per well). Cells were allowed to attach to the surface and were cultured until they reached approximately 30% confluency. Next, cells were treated with 10 μM RA in a DMEM medium containing 2.5% FBS for 5 days to induce differentiation. Cells in some of the wells were collected after this step as partially differentiated SH-SY5Y cells (denoted as RA). Cells in the rest of the wells were further treated using neurobasal medium supplemented with BDNF and the B-27 supplement to generate terminally differentiated cells (denoted as RA-BDNF).

Figure 1. Experimental workflow for SH-SY5Y cell differentiation and comparative proteomics using two digestion strategies.

(A) Schematic of the SH-SY5Y cell differentiation protocol. Undifferentiated SH-SY5Y cells were cultured in DMEM with 10% iFBS until reaching 20-30% confluency. Cells were then treated with Medium I (RA-supplemented) for 5 days to induce partial differentiation, followed by treatment with Medium II (BDNF-supplemented) for an additional five days to achieve terminal differentiation on day 11. (B) Overview of three proteomics sample preparation workflows. Top, on-filter in-cell (OFIC) processing coupled with E4tip. Intact cells from each condition were directly fixed in an E4tip prefilled with methanol. Proteins were then reduced, alkylated, and digested on-filter in the tip, followed by peptide desalting in the same tip. Middle, OFIC coupled with E3tip. Intact cell fixation was performed using methanol in an E3tip. Proteins were then reduced, alkylated, and digested on-filter in the tip. The E3tip was stacked onto a preactivated C18 StageTip for desalting and peptide elution. Bottom, SDS lysate digestion coupled with E3tip. Cells were lysed in an SDS solution and subjected to reduction and alkylation at 95 °C. The resulting lysate was mixed with 4x volume of acetonitrile and then transferred to an E3tip. Proteins were digested on-filter in the E3tip, which was then stacked onto a preactivated C18 StageTip for desalting and peptide elution. (C) Digested and desalted peptides were analyzed using data-independent acquisition (DIA) LC-MS on a FAIMS Orbitrap Eclipse mass spectrometer. Created in BioRender. Moafian, Z. (2025) https://BioRender.com/6jo0dxr

The morphology of SH-SY5Y cells was examined by using a phase contrast microscope. Undifferentiated SH-SY5Y cells showed short processes (Figure S1A). In contrast, differentiated SH-SY5Y cells possess elongated and branched processes, showing an intertwined network. Both RA- and RA/BDNF-treated SH-SY5Y cells showed outgrowth of neurites of varied lengths (Figure S1B–E). Notably, RA-treated, partially differentiated cells demonstrated smaller and fewer neurites (Figure S1B, C) compared to those treated sequentially with RA and BDNF to achieve terminal differentiation, exhibiting complex connections and a dense network of neurites (Figure S1D, E). Our observation agreed with previous reports showing more mature neuronal characteristics of SH-SY5Y cells treated with RA and BDNF compared to cells treated with RA alone.^{20, 33}

Benchmarking In-Cell Proteomics Analysis of Neuroblastoma Cells

We previously reported proof-of-principle evidence of the on-filter in-cell (OFIC) digestion of HEK cells.²⁷ However, no systematic evaluation of this “in-cell proteomics” method for deep proteome analysis of mammalian cells was performed, including side-by-side comparisons with lysate-based processing methods and further optimization of the method in neuronal cell models. In this regard, we examined two types of filter devices, E3tip and E4tip, as both E3 and E4 membranes possess submicron pore sizes that are compatible with intact SH-SY5Y cells. To demonstrate their performance, we included the conventional SDS lysis-based digestion method as a control. A flowchart of the two digestion methods in conjunction with the E3tip and E4tip is shown in Figure 1B. In the final step, the digested and desalted peptides were analyzed using data-independent acquisition (DIA) LC-MS on a FAIMS Orbitrap Eclipse mass spectrometer, and the MS data were processed using Spectronaut and Perseus as described in the Experimental Section (Figure 1C).

We found that E4tip combined with OFIC digestion (OFIC_E4) identified the largest number of proteins and peptides from undifferentiated SH-SY5Y cells among the three tested workflows (Figure 2A, B). The coefficients of variation (CV) of quantified protein groups (PG) and peptides were significantly lower for the OFIC_E4 method than for the lysate-based digestion method (Figure 2C, D). We owe the high consistency and quantitative accuracy to the simplicity and single-vessel nature of the in-cell processing method, as it bypasses cell lysis and protein extraction, and carries out all sample preparations in the same tip, achieving nearly loss-free sample processing. In contrast, the SDS method requires cell lysis, sample transfer, and then protein precipitation prior to digestion, which may have caused technical variations and a lower identification rate (Figure 2A, B). These data are in line with the findings from a proteomics study of Caenorhabditis elegans using the OFIC method.²⁹ Notably, the E3tip-based digestion method provided proteomic performance similar to that of E4tips. This is likely because in the OFIC_E3tip method, we acidified the digests, stacked the E3tip onto a C18 StageTip, and desalted peptides directly after digestion (Figure 1B). The seamless workflow of E3tip digestion combined with a direct C18 tip desalting may have reduced sample loss and improved its performance. However, the OFIC_E4 method is preferred, given the simplified workflow while maintaining a high proteomics performance.

Figure 2. Evaluation of OFIC digestion method for neuroblastoma cell proteomics analysis.

(A-B) Number of protein groups (A) and peptides (B) identified by each workflow tested. OFIC_E4 stands for the OFIC method coupled with E4tip; OFIC_E3 stands for the OFIC method coupled with E3tip; SDS_E3 stands for SDS lysis method coupled with E3tip. Error bars were calculated based on five replicates (applicable to all the panels in this figure). (C, D) Coefficient of variations (CV%) of quantification of proteins (C) and peptides (D) detected by each of the three workflows. (E, F) Pearson correlation analysis of the three workflows. Median values of protein intensity (log₂-transformed) were used for each workflow. (G) Quantitative comparison of proteins identified in various cellular compartments. (H) Venn diagram of proteins identified by each workflow. Only proteins detected in at least three replicates were considered. (I) Comparison of methionine oxidation among the three workflows. The relative intensity (%) was calculated by dividing the sum intensity of oxidized methionine-containing peptides by the total intensity of the peptides in each digestion experiment. (J) Comparison of digestion efficiency (%) by dividing the sum of the intensity of peptides that contain one or two missed cleavage sites by the total peptide intensity.

We next asked whether the proteomes obtained using the OFIC method demonstrate any systematic bias compared to the SDS lysis-based method (Figure 1B). Our data indicated that over 95% of the proteins derived from the OFIC_E4 method were also identified by the SDS lysis-based method (Figure 2H). Meanwhile, the protein intensity correlated well among the three methods (Figure 2E, F). These data suggest that OFIC digestion shows no qualitative and quantitative differences from the conventional digestion method. In the context of cellular localization, no dramatic differences were observed among the three methods in terms of the intensity and number of proteins derived from known categories, such as mitochondria, ER, Golgi, and centrosomes (Figure 2G and S3L). Furthermore, we evaluated the occurrence of methionine oxidation during sample preparation, which has been reported to hinder protein digestion and electrospray ionization in bottom-up proteomics analysis.³⁴ We observed significantly fewer oxidized peptides in the OPIC_E4 sample group (Figure 2I), suggesting an additional benefit of the in-cell digestion method. In an independent experiment, we also tested the reduction/alkylation strategies and different trypsin inputs. Our data showed that reduction/alkylation can be performed on either of the protein or peptide level, which did not affect the identification rate (Figure S2A). Meanwhile, varying the trypsin input did not affect the protein identification either, although higher input (trypsin:protein ratio = 1:20) slightly improved the number of peptide hits (Figure S2B). Lastly, we examined the missed cleavages derived from the OFIC methods and observed an equivalent digestion efficiency to the SDS lysis-based digestion (Figure 2J).

In a separate experiment, we also investigated the performance of the OFIC method against a urea-based in-solution digestion method. We achieved a largely equivalent or better identification rate at both protein and peptide levels and did not observe significant bias in the context of subcellular compartments as described above (Figure S3). Our data demonstrated good accessibility of trypsin to the proteins inside the methanol-fixed SH-SY5Y cells and highlighted the great potential of the OFIC method for the deep proteomic analysis of neuroblastoma cells. Lastly, we investigated whether any biases were observed against hydrophilic or hydrophobic peptides during the in-cell digestion, and if our method can effectively identify membrane proteins, in particular, the transmembrane proteins. Our data suggested equivalent performance of the OFIC approach for the identification of hydrophobic peptides and membrane proteins (Figure S4A). The data also indicated that the in-cell proteomics methods can indeed identify peptide sequences that are located in the transmembrane regions (Figure S4D, E).

Proteomic Analysis of Differentiated SH-SY5Y Cells Reveals High Sensitivity and Reproducibility

After validating the OFIC_E4 workflow, we performed the OFIC digestion of undifferentiated and differentiated SH-SY5Y cells using E4tips. We were able to consistently detect approximately 8000 proteins from cells obtained from a single well of a six-well plate. A slightly larger number of proteins were identified in terminally differentiated SH-SY5Y cells compared to partially and undifferentiated cells, while no significant differences were observed between undifferentiated and partially differentiated cells (Figure 3A). A Venn diagram analysis (Figure S5) indicates that the majority (>92%) of the proteins were commonly identified among the three groups, highlighting the qualitative similarity between the undifferentiated and differentiated cells. Of note, although the RA-BDNF group identified slightly more proteins than the RA group, they showed no significant difference at the peptide level. After examining the quantitation data, we did notice some low-abundance proteins (ranked below 4,500; Supporting Information Figure S6) that are specific to the RA-BDNF–treated cells, including E3 ligases (e.g., TRIM62, TRIM21, TRIM47, TRIM41), ubiquitin-specific proteases (e.g., USP38, USP25, USP42, USP1), and E2 ubiquitin-conjugating enzymes such as UBE2C.

Figure 3. Proteome profiling evaluation of undifferentiated and differentiated SH-SY5Y cells.

(A, B) Number of protein groups (A) and peptides (B) identified in SH-SY5Y cells that are undifferentiated (Undiff), partially differentiated (RA), and terminally differentiated (RA-BDNF). Error bars indicate four replicate experiments. (C) Heatmap showing the Global ANOVA significant proteins across all treatment conditions. Hierarchical clustering by row was performed with the average linkage and Euclidean distance. The heatmap was split into 5 distinct clusters and the corresponding intensity profiles for each cluster are displayed adjacently. Gene names from each cluster were used for GO biological process 5 enrichment (GO_BP5) with the functional annotation tool DAVID. (D-I) GO-term analysis of proteins identified in undifferentiated (Undiff) and differentiated (RA and RA-BDNF) SH-SY5Y cells was performed for various cellular components. The sum iBAQ values for all genes within each GO term relative to the total sum iBAQ values per sample is expressed as a percentage. Data are shown as mean ± SD (n = 4 replicates). Statistical analysis was performed using one-way ANOVA followed by Tukey’s post hoc testing. p < 0.05 (*), p < 0.01 (**), p < 0.001 (***), p < 0.0001 (****), and ns: not significant.

The heatmap with GO enrichment analysis using DAVID compares the proteome changes across different stages of differentiation (Figure 3C). The columns represent the three differentiation stages, while rows show clustered proteins. Blue indicates a lower expression level, and orange indicates a higher expression level based on the Z-score of Log₂ LFQ intensity. The dendrogram was split into five distinct clusters in Perseus by tree cutting after hierarchical clustering was performed. The genes in each cluster were queried in DAVID for GO_BP5 functional enrichment. We observed consistent expression levels of proteins among replicates in each group (undifferentiated, RA, and RA-BDNF) (Figure 3C), while distinct patterns of protein abundance in the SH-SY5Y proteomes at different stages were also observed.

GO-term Biological Process 5 (GO_BP5) analysis of the genes within each cluster revealed over-represented functional categories among proteins with similar expression patterns. Proteins in cluster 1 showed higher expression levels during partial differentiation as compared to the terminal differentiation stage. The proteins in this cluster are primarily involved in core cellular processes, such as translation and cellular respiration. The change in translational and respiration-related protein abundance agrees with the switch from differentiation to proliferative states. Proteins in clusters 2 and 4 displayed the highest expression levels at the terminal differentiation stage and showed involvement with vesicular transportation and macroautophagy. This agrees with the known function of vesicle mediated transport of neurotransmitters and the role of autophagy in synaptic vesicle turnover and receptor trafficking in mature neurons.^35–37

Proteins in clusters 3 and 5 displayed median levels of expression at terminal differentiation. The proteins in cluster 3 displayed an enrichment of metabolic and morphogenic processes, and proteins in cluster 5 exhibited an enrichment of mRNA processing and ribosome biogenesis.

Further, our analysis of the protein abundance belonging to selected cellular processes revealed a clear increase in the level of proteins involved in glutamatergic synapses, neuron projection (axons and dendrites), and presynaptic and postsynaptic processes, upon differentiation of SH-SY5Y cells (terminally, partially, or both) compared to undifferentiated cells (Figure 3D–I). The expression levels of glutamatergic synapse proteins were enhanced in both partially and terminally differentiated SH-SY5Y cells, but their levels remained largely unchanged between RA and RA-BDNF treated cells (Figure 3D). Our result is in general agreement with the observation by Targett et al. showing that consecutive treatment of SH-SY5Y cells with RA and BDNF afforded cells with a cholinergic/glutamatergic phenotype.³⁸ It should be noted that, in our experiment BDNF was used in conjunction with the B-27 supplement, which is also known to promote differentiation of SH-SY5Y cells to glutamatergic phenotype based on mRNA level and glutamatergic-related protein markers.³⁹

We observed an increase in proteins involved in neuron projection formation in terminally differentiated cells compared to partially differentiated and undifferentiated cells (Figure 3E). In addition, axon-related protein levels increased during partial differentiation but remained relatively unchanged during terminal differentiation (Figure 3F). In contrast, dendrite-associated protein levels remained unchanged in RA treated cells but showed a marked increase following treatment with RA and BDNF to induce terminal differentiation (Figure 3G). We also observed an overall increase in pre- and postsynaptic protein levels upon differentiation (Figure 3H, I), agreeing with the observation that the abundance of synaptic proteins increases during neuronal maturation and their importance in synaptic connection development.⁴⁰ These results suggest higher levels of maturation in terminally differentiated SH-SY5Y cells compared with undifferentiated and partially differentiated neuroblastoma cells.

SH-SY5Y Cell Differentiation Confirmed Using Western Blotting and Quantitative MS Analyses of Selected Marker Proteins

To ensure the proper differentiation of SH-SY5Y cells and corroborate our proteomics results, we performed Western blotting (WB) analysis of several marker proteins commonly used for neuronal cell differentiation. Proliferating cell nuclear antigen (PCNA) was reported to be downregulated upon differentiation of SH-SY5Y cells.⁴¹ As shown in Figure 4A (top panel), the PCNA level, as seen in our WB analysis, was significantly decreased in differentiated SH-SY5Y cells, particularly in terminally differentiated cells, suggesting enhanced neuronal maturation due to treatment with RA and BDNF. This observation is in alignment with our quantitative proteomics analysis of the PCNA level in undifferentiated and partially and terminally differentiated SH-SY5Y cells (Figure 4A, bottom panel). Growth associated protein (GAP43) is a commonly used marker of neuronal differentiation and neurite growth.⁴² As shown in our quantitative proteomics analysis (Figure 4B, bottom panel), the expression level of GAP43 was significantly increased in fully differentiated SH-SY5Y cells, when compared to undifferentiated and partially differentiated SH-SY5Y cells, also revealed by our WB analysis (Figure 4B, top panel), and supported by a previous report.⁴³ We observed higher protein levels of pre- and postsynaptic markers, synaptophysin (SYP) and postsynaptic density-95 (PSD95), in differentiated cells than undifferentiated cells in WB and quantification proteomics analyses (Figure 4C, D). These results are consistent with previous studies showing increased expression of these markers following RA-BDNF treatment.³⁸

Figure 4. Western blot and quantitative proteomics analyses of neuronal and proliferation markers during SH-SY5Y cell differentiation.

SH-SY5Y cells were differentiated using retinoic acid (RA) or a sequential RA-BDNF treatment. Top panels (A–D): Representative Western blot were performed for (A) PCNA (proliferation marker), (B) GAP43 (neurite outgrowth marker), (C) Synaptophysin (SYP; presynaptic marker), and (D) PSD95 (postsynaptic marker). GAPDH was used as a loading control. Bottom panels (A–D): Quantification of protein abundance as protein group (PG) quantity for undifferentiated (Undiff), RA-treated, and RA-BDNF-treated cells. Data are shown as mean ± SD (n = 4). Statistical analysis was performed using one-way ANOVA followed by Tukey’s post hoc testing. p < 0.05 (*), p < 0.01 (**), p < 0.001 (***), p < 0.0001 (****), and ns: not significant.

Pairwise Analysis of Proteins Upregulated upon RA- and RA-BDNF-Induced Differentiation in SH-SY5Y Cells

We carried out pairwise differential expression analysis of three sample groups, i.e. undifferentiated versus partially differentiated, undifferentiated versus terminally differentiated, and partially differentiated versus terminally differentiated SH-SY5Y cells (Figure 5A–C). RA and RA-BDNF treatments significantly upregulate several key proteins involved in neuronal differentiation in SH-SY5Y neuroblastoma cells. The expression of cellular retinoic acid-binding protein 2 (CRABP2) was drastically increased by more than 16-fold in RA-treated and 4-fold in RA-BDNF-treated SH-SY5Y cells (Figure 5A, B, D), agreeing with its known role in neuronal differentiation and neurodevelopment.^{24, 44–46} Integrin α-1 (ITGA1) was also upregulated by more than 4-fold in differentiated cells (Figure 5A, B, E). This observation aligns with previous reports and is in agreement with the role of ITGA1 in neurite outgrowth.^{23, 47} Cell Adhesion Molecule 3 (CADM3) is a major axonal adhesion molecule that interacts with glial CADM4. This mediates Schwann cell (glial cells) and axon interactions that are vital for axonal ensheathment and proper organization of the axonal membrane in the peripheral nervous system (PNS).⁴⁸ CADM3 was upregulated in both RA and RA-BDNF treated neuroblastoma cells Figure 5A, B, F). In agreement with previous reports, acetylcholinesterase (ACHE, a differentiation marker)⁴⁹ showed elevated levels in both partially and terminally differentiated cells (Figure 5A, B, G).

Figure 5. Quantitative comparison of differentiation-specific protein markers upon differentiation.

(A) Volcano plot comparing various protein levels between undifferentiated and partially differentiated SH-SY5Y cells. (B) Volcano plot comparing undifferentiated and terminally differentiated SH-SY5Y cells. (C) Volcano plot comparing RA and terminally differentiated cells. The X-axis cutoff was set at fold change (FC) ≥ 2.0, while the Y-axis cutoff was set at p-value ≤ 0.05. The proteins related to neuronal development are highlighted as red dots. (D-I) Label-free MS quantification of identified proteins including CRABP2 (D), ITGA1 (E), CADM3 (F), ACHE (G), DPYSL3 (H), and neuropeptide Y (NPY) (I) in three different stages of differentiation. PG stands for protein group. Statistical analysis was performed using one-way ANOVA followed by Tukey’s post hoc testing. p < 0.05 (*), p < 0.01 (**), p < 0.001 (***), p < 0.0001 (****), and ns: not significant.

γ-aminobutyric acid (GABA) type B receptor subunit 1 (GABBR1), involved in hematopoietic stem and progenitor cell (HSPC) proliferation and B-cell differentiation,⁵⁰ and Annexin A2 (ANXA2), found in GABAergic interneurons throughout cortical and subcortical areas,⁵¹ were also upregulated in both RA- and RA-BDNF treated SH-SY5Y cells, although ANXA2’s role in brain development remains unclear (Figure 5A, B). SLC30A3 (ZnT3), which plays an essential role in adult hippocampal neurogenesis,⁵² was increased to more than 16-fold upon differentiation (Figure 5A, B). In addition, microtubule-associated protein tau (MAPT), α-synuclein (SNCA), neurofilament light polypeptide (NEFL), and neurofilament medium polypeptide (NEFM), well known for their roles in the structural and functional development of mature neurons,^53–55 were upregulated upon differentiation, especially in terminally differentiated SH-SY5Y cells (Figure 5A, B).

Volcano plot analysis of RA vs. RA-BDNF neuroblastoma cells identified several neuronal development-associated proteins with higher expression levels in cells treated consecutively with RA and BDNF. DPYSL3 (also referred to as collapsing response mediator protein 4 (CRMP4)) shows strong expression in both developing and mature nervous systems,^{56, 57} exhibited significantly higher expression in RA-BDNF-treated cells compared to RA-treated cells (Figure 5C, H). Notably, DPYSL3 was reported to colocalize with GAP43, a protein known to regulate postsynaptic GABA(A) receptor clustering.^{58, 59} Consistent with these reports, we also detected higher GAP43 expression in terminally differentiated SH-SY5Y cells by WB and proteomics analysis (Figures 4B, 5B, C). Major Vault Protein (MVP) was upregulated in RA-BDNF-treated SH-SY5Y cells (Figure 5C). MVP is highly expressed in developing neurons and the nucleus–neurite axis, actively transported along the cytoskeleton, and also serves as a marker for developing microglia.^60–63 Interestingly, neuropeptide Y (NPY) was detected at a higher level in terminally differentiated SH-SY5Y cells (Figure 5C, I). This neuropeptide was found to enhance in vitro cell proliferation of postnatal rat hippocampal cultures.⁶⁴ Intracerebroventricular administration of NPY has been reported to boost DG cell proliferation and induces neuronal differentiation in C57BL/6 adult mice.⁶⁵

We also carried out a systematic KEGG pathway analysis of significantly upregulated proteins in partially and terminally differentiated SH-SY5Y cells (Figure S7). We found that steroid biosynthesis and cholinergic synapses are among the most abundant pathways in both RA- and RA-BDNF-treated neuroblastoma cells (Figure S7A, B). In addition, proteins associated with the axonal guidance pathway are more enriched in RA-BDNF-treated SH-SY5Y cells compared to RA-treated cells suggesting more advanced neuronal maturation (Figure S7A, B). In addition, a comparison of RA-treated and RA-BDNF-treated SH-SY5Y cells revealed that proteins in vesicular transport, neurotrophin signaling, and spliceosomes are enriched in the terminally differentiated cells (Figure S7C). Overall, our results revealed significant changes in the proteome of SH-SY5Y cells upon differentiation, with many neuronal development-associated proteins upregulated. Our observations support the importance of BDNF in conjunction with RA to induce SH-SY5Y neuronal differentiation, thus providing a highly relevant cell model for investigating neuronal cells.

Proteasomal and Ubiquitin Pathway Dynamics during SH-SY5Y Differentiation

Ubiquitination is a crucial post translational modification (PTM), commonly associated with the proteasomal and lysosomal degradations of unwanted and misfolded proteins in cells. Dysregulation of this process is closely linked to the development of neurological and neurodegenerative diseases.^66–68 We carried out an in-depth analysis of the ubiquitination pathway components, including E3 ubiquitin ligases, deubiquitinases (DUBs), E2 ubiquitin conjugating enzymes, and proteasome subunits, across three SH-SY5Y cell states (Figure S8–S11). Based on one-way ANOVA significance criteria as described previously, 65 of the 177 identified E3 ligases were found to be differentially expressed across the differentiation stages in SH-SY5Y cells. Notably, several E3 ubiquitin ligases that are more abundantly expressed in terminally differentiated cells, including TRIM2, TRIM3, TRIM21, TRIM46, NEDD4L, HERC2, and HERC4 (Figure 6A). Our observations agree with previous studies suggested a crucial role for E3 ligases in various neurodevelopmental stages.^{69, 70}

Figure 6. Line plots showing the relative abundance of selected E3 ubiquitin ligases (A) and DUBs (B) in SH-SY5Y cells that are highly expressed in the terminally differentiated SH-SY5Y cells.

The Y-axis denotes the Z-scored Log₂ LFQ intensities of the selected proteins.

TRIM2 and TRIM46 were reported to play important roles in neuronal development. TRIM2 mediates neuron’s axonal specification by regulating the degradation of NEFL, thus playing a critical role in axon initiation and outgrowth upon differentiation.⁷¹ TRIM46 is essential for establishing neuronal polarity by promoting parallel microtubule bundle formation at the axon initial segment.⁷² TRIM3 contributes to the regulation of synaptic plasticity by modulating the morphology of dendritic spine and the dynamics of actin cytoskeleton.^{73, 74} TRIM21 has been associated with the neuronal innate immune response via regulation of neuroinflammation, however, its role in neuronal development is poorly understood.⁷⁵ NEDD4L (or NEDD4-2) regulates neuronal excitability in brain by ubiquitinating voltage-gated sodium channels, such as Nav1.7 and 1.8.⁷⁶ Deficiency in NEDD4L has been linked epilepsy⁷⁷ and neuropathic pain.⁷⁸

HERC2 is a HECT domain-containing E3 ubiquitin ligase that interacts with RNF8 (another E3 ligase). This interaction is part of a complex that negatively regulates synapse differentiation, particularly by suppressing the formation of presynaptic boutons and functional synapses. However, there are many studies that suggest a relationship between HERC2 and autism spectrum disorder,⁷⁹ neurodevelopmental phenotype⁸⁰ and intellectual disabilities.⁸¹ Biallelic mutations of HERC2 cause intellectual disability, developmental delays and seizures, similar to the symptoms observed in Angelman syndrome.⁸² More studies are needed to elucidate the exact role of HERC2 E3 ligase in neuronal development. In addition, HERC4, also identified in the fully differentiated SH-SY5Y cells, has not been reported to function in neurological processes.

Several other E3 ligases identified, UBE3A, UBR4, UBE3C, and LRSAM1, are also known to play roles in various aspects of neurological processes and have been linked to neurological diseases, such as Angelman syndrome and autism spectrum disorders for UBE3A,^83–85 early dementia for UBR4,⁸⁶ Charcot-Marie-Tooth disease type 2P for LRSAM1,⁸⁷ and distal hereditary motor neuropathies for UBE3C.⁸⁸ In addition to several E3 ligases known to play important roles in neurodevelopment and neurodegeneration, we also identified E3 ligases that have not been previously linked to neurological processes, including TRIM21, HERC4, and RNF14. Higher expression levels of many E3 ligases in RA-BDNF-differentiated cells suggest that these cells may serve as a better model system for investigation of the E3 ligases in the context of neurological processes compared with RA-treated SH-SY5Y cells. In addition, terminally differentiated cells may be used for identifying inhibitors and modulators of E3 ligases involved in various neuropathological processes, such as AD, PD, and neuropathic pain.

Protein ubiquitination by E3 ubiquitin ligases is counteracted by DUBs. While treatment of SH-SY5Y cells with RA led to the downregulation of multiple DUBs, RA-BDNF-treated neuroblastoma cells showed the upregulation of a different set of DUBs (Figure 6B). Several DUBs belonging to the ubiquitin-specific protease (USP) family, including USP8, USP9X, USP9Y, USP14, USP15, USP24, USP46 and USP47, were highly upregulated upon terminal differentiation of SH-SY5Y. Ubiquitin-specific protease 9X (USP9X), located on the X-chromosome, has a crucial role in fetal and neuronal development and its mutation has been associated with neurodevelopmental delay, intellectual disability, and various congenital anomalies observed in female.⁸⁹ In addition, missense mutation of this DUB affects neuronal migration and development in males.⁹⁰ USP9X also directly interacts with and deubiquitinates α-synuclein.⁹¹ Ubiquitin-specific protease 8 (USP8) is localized in both somatic and dendritic regions of hippocampal neurons and is notably enriched in postsynaptic density fractions, indicating a potential role in modulating synaptic strength.⁹² The loss of ubiquitin-specific protease 14 (USP14) significantly increased phosphorylated Tau level.⁹³ It has been shown that USP14 might have a sex dependent role in PD when its inhibition promotes α-synuclein clearance and reduces inflammation in female mice.⁹⁴ In another study, treatment of mice with a USP14-specific inhibitor, reduces neuronal injury, decreases animal mortality, and improves animal functional recovery.⁹⁵

Ubiquitin-specific protease 15 (USP15) has been shown to deubiquitinate mitochondrial outer membrane proteins and inhibit mitophagy. Cell and insect model studies showed that loss of USP15 function can restore impaired mitophagy, suggesting a therapeutic strategy for PD caused by reduced Parkin levels.⁹⁶ USP15 also plays an important role in maintaining RNA metabolism balance, which is essential for proper brain function and health. It regulates a terminal uridylyl transferase enzyme TUT1, thereby affecting U6 snRNA level, a key component of the spliceosome and essential for pre-mRNA splicing.⁹⁷ Ubiquitin-specific protease 24 (USP24) was found to negatively regulate autophagy in human neuroglioma, and elevated levels of USP24 were observed in a group of patients with idiopathic PD.⁹⁸ USP46 accumulates at synaptic sites and colocalizes with other synaptic marker proteins. It promotes AMPAR accumulation, key mediators of inter-neuronal communication, thus enhancing synaptic strength and ultimately improving brain function.⁹⁹ Ubiquitin-specific protease 47 (USP47) antagonizes the E3 ubiquitin ligase activity of CHIP and stabilizes katanin-p60, thereby promoting axonal growth in cultured rat hippocampal neurons.¹⁰⁰ This observation that many DUBs are highly expressed in terminally differentiated SH-SY5Y cells and have known roles in neuronal development suggests that the terminally differentiated SH-SY5Y cells are a better-suited cellular model to investigate DUBs in neurological processes. Furthermore, several E3 ligases (UBE3A, HERC4, TRIM21, LRSAM1) and USPs (USP47, USP24, USP14, USP8, USP15) showed a downregulation in the RA-differentiated SH-SY5Y and an upregulation in the RA-BDNF differentiated cells, indicating a more active protein turnover in the fully differentiated SH-SY5Y cells, as suggested in earlier studies.^101–103

The heatmap of the E2 ubiquitin-conjugating enzymes also demonstrated different patterns across the three cellular states upon hierarchical clustering (Figure S10), with several of them being abundantly expressed in terminally differentiated SH-SY5Y cells, including UBE2A, UBE2B, UBE2L3, UBE2Z, UBE2J2, UBE2L6, UBE2G2, UBE2H, and UBE2K. Specifically, UBE2A, UBE2K, and UBE2H have been previously implicated in neurological processes. UBE2A was shown to play critical roles in learning and memory in mouse models,¹⁰⁴ and has also been linked to Nascimento syndrome pathology, an X-linked intellectual disability.¹⁰⁵ UBE2K protein levels are upregulated in both the blood and brain of Schizophrenia patients,¹⁰⁶ and UBE2K also interacts with the Huntington protein in the brains of Huntington disease patients.¹⁰⁷ In a zebrafish model, UBE2H was found to be required for normal brain development.¹⁰⁸ Notably, the involvement of UBE2B, UBE2L3, UBE2Z, UBE2J2, or UBE2L6 in neurological development are not clear, which warrants further investigation.

Our analysis of the proteasome subunits revealed that the proteasome 20S core and 19S regulatory subunits showed the highest expression level in RA-treated SH-SY5Y cells (Figure S11). In comparison, the abundance of these proteasome subunits was reduced upon terminal differentiation with RA and BDNF. Interestingly, the proteasome activator subunits PSME1 and PSME2 were expressed at the highest levels in terminally differentiated SH-SY5Y cells. Our results suggest that proteasomal activity is required for the early-stage differentiation of the neuroblastoma SH-SY5Y cells but subdued once the cells are terminally differentiated. Our observation is in alignment with earlier studies showing that the proteasome plays important roles in the development and differentiation of neurons in humans.^109–111 Lastly, to further evaluate the biological relevance of SH-SY5Y-derived neurons, we compared our data with iPSC-, iPSC/NSC-, and hESC-derived neurons^111–113 in the context of ubiquitin-proteasome system (UPS) proteins. The analysis indicated that the majority of the UPS proteins expressed by the stem cell-derived neurons are also detectable in the differentiated SH-SY5Y cells and their abundances are quantitatively comparable (Figure S12), which further validates our finding and supports induced SH-SY5Y cells as a suitable model for neurons, particularly in the investigation of the ubiquitin-proteome system in a neurological setting.

CONCLUSIONS

In this study, we systematically evaluated an in-cell sample processing strategy for proteomics analysis of neuroblastoma cells. We benchmarked its performance against conventional SDS and urea lysate-based digestion methods. In combination with the DIA mass spectrometry, we showcased the power of the “in-cell proteomics” approach by achieving a deep proteome coverage of SH-SY5Y cells from different stages of differentiation. Our quantitative proteomics analysis identified significant differences in the proteomes of RA- and RA-BDNF-treated SH-SY5Y cells, particularly proteins in the ubiquitin-proteome system, highlighting the importance of reversible protein ubiquitination in neurological processes. A future study focusing on the ubiquitinome of the RA- and RA-BDNF-treated SH-SY5Y cells will provide useful information on the ubiquitinated proteins upon the differentiation of SH-SY5Y cells. Our results support the notion that RA-BDNF-treated SH-SY5Y cells represent a more appropriate model for studying the neurological development and neurodegenerative processes.

Supplementary Material

The Supporting Information is available free of charge at: https://pubs.acs.org/doi/10.1021/acs.jproteome.5c00634

Phase contrast images of SH-SY5Y cells at different stages of differentiation (Figure S1); evaluation of OFIC digestion for neuroblastoma cell proteomics analysis (Figure S2); comparative analysis of OFIC with ureaand SDS lysate-based digestion methods (Figure S3); continued evaluation of OFIC digestion method (Figure S4); Venn diagrams of proteins derived from undifferentiated and differentiated cells (Figure S5); quantitative assessment of the three groups of differentiated cells (Figure S6); enriched KEGG pathways for differentially upregulated proteins (Figure S7); heatmap of E3 ubiquitin ligases across differentiation stages (Figure S8); heatmap of DUB, deSUMOylase and deNEDDylase expression across different stages of differentiation (Figure S9); heatmap of ubiquitin-conjugating enzyme (E2) and ubiquitin-like-conjugating enzyme expression across different stages of differentiation (Figure S10); heatmap of proteasome-associated protein expression across differentiation stages (Figure S11); comparison of UPS proteins identified in SH-SY5Y, iPSC-, iPSC/NSC-, and hESC-derived neurons (Figure S12); and full Western blotting image corresponding to those shown in Figure 4 (Figure S13) (PDF).

Data Availability Statement

The MS raw files associated with this study have been deposited to the MassIVE server (https://massive.ucsd.edu/) with the data set identifier MSV0000 97787. The data sets used for comparative analyses (Figure S12) were obtained from the supporting files that are publicly available from those studies (see details in the cited references 111–113).