Discovery of New Membrane-Associated Proteins Overexpressed in Small-Cell Lung Cancer

Abstract

Introduction

Small-cell lung cancer (SCLC) is the most aggressive subtype of lung cancer, with no early detection strategy or targeted therapy currently available. We hypothesized that difference gel electrophoresis (DIGE) may identify membrane-associated proteins (MAPs) specific to SCLC, advance our understanding of SCLC biology, and discover new biomarkers of SCLC.

Methods

MAP lysates were prepared from three SCLCs, three non–small-cell lung cancers, and three immortalized normal bronchial epithelial cell lines and coanalyzed by DIGE. Subsequent protein identification was performed by mass spectrometry. Proteins were submitted to Ingenuity Pathway Analysis. Candidate biomarkers were validated by Western blotting (WB) and immunohistochemistry (IHC).

Results

Principal component analysis on the global DIGE data set demonstrated that the four replicates derived from each of the nine cell lines clustered closely, as did samples within the same histological group. One hundred thirty-seven proteins were differentially expressed in SCLC compared with non–small-cell lung cancer and immortalized normal bronchial epithelial cells. These proteins were overrepresented in cellular/tissue morphology networks. Dihydropyrimidinase-related protein 2, guanine nucleotide–binding protein alpha-q, laminin receptor 1, pontin, and stathmin 1 were selected as candidate biomarkers among MAPs overexpressed in SCLC. Overexpression of all candidates but RSSA in SCLC was verified by WB and/or IHC on tissue microarrays. These proteins were significantly associated with SCLC histology and survival in univariables analyses.

Conclusion

DIGE analysis of a membrane-associated subproteome discovered overexpression of dihydropyrimidinase-related protein 2, guanine nucleotide–binding protein alpha-q, RUVB1, and stathmin 1 in SCLC. Results were verified by WB and/or IHC in primary tumors, suggesting that investigating their functional relevance in SCLC progression is warranted. Association with survival requires further validation in larger clinical data sets.

Small-cell lung cancer (SCLC) represents 10% to 15% of lung cancers and is clinically the most aggressive subtype, with a 5-year overall survival (OS) as low as 5%.¹ Molecular steps leading to SCLC are still poorly understood, and this has translated into the absence of efficient early detection strategies or targeted therapies. In-depth proteomic analysis is therefore needed to improve our understanding of SCLC tumorigenesis and may have implications in the discovery of novel early detection and therapeutic strategies for this aggressive cancer.

Although (epi)genetic alterations are driving carcinogenesis and genomic studies have provided valuable information on cancer biology, the cellular phenotype is determined by proteins and cannot be predicted by genomics alone. Proteomic approaches are therefore powerful tools to study cancer biology.^2,3 Among them, difference gel electrophoresis (DIGE) coupled with protein identification by mass spectrometry (MS) is widely used in comparative proteomics for the detection of protein differences with high reproducibility and reliability.^4–6 However, these studies have not identified key underpinnings of tumor progression or specific molecular targets for SCLC diagnostics or therapeutics.

Membrane-associated proteins (MAPs) are excellent targets for diagnostic and therapeutic interventions. Although membrane proteins constitute only 20% to 30% of human proteome, they represent more than 60% of all drug targets.⁷ In this study, we hypothesized that the identification of SCLC-specific MAPs by DIGE may advance our understanding of SCLC biology and lead to the discovery of new candidate diagnostic or therapeutic biomarkers. We therefore analyzed MAPs from three SCLC, three non–small-cell lung cancer (NSCLC), and three immortalized normal bronchial epithelial (INBE) cell lines by DIGE.

MATERIALS AND METHODS

Cell Lines and Tissues

Nine cell lines were used for this study: National Cancer Institute (NCI)-H69, NCI-H82, NCI-H209, A549, NCI-H23, NCI-H520, BEAS-2B, HBEC-3KT, and 16-HBE. All were purchased from ATCC (Manassas, VA), except HBEC-3KT (Dr. Minna’s gift, University of Texas Southwestern) and 16-HBE (Dr. Gruenert’s gift, Children’s Hospital, Oakland Research Institute). They were cultured under recommended conditions.

Tissue microarrays (TMAs) made of SCLC and NSCLC specimens were prepared from formalin-fixed, paraffin-embedded tissue blocks⁸ retrieved from the Pathology Department’s archives at Vanderbilt University Medical Center, Nashville VA Medical Center, and St Thomas Hospital in Nashville, TN. They were obtained between 1996 and 2008 from 136 patients who had surgery or bronchoscopy. Samples were annotated with clinical data elements. The study was approved by the Institutional Review Board at each institution.

MAP Extraction

MAP extraction was performed four independent times for each cell line with the ProteoExtract Native Membrane Protein Extraction Kit (EMD Chemicals, San Diego, CA) per manufacturer’s protocol.

DIGE and MS

Minimal labeling was performed with N-hydroxysuccinimide–ester dyes Cy2/3/5 using the mixed internal standard methodology as described previously,⁹ with 150 μg of protein of each of the 36 experimental samples. DIGE-associated instrumentation was manufactured by GE Healthcare (Piscataway, NJ). Isoelectric focusing (pH 4–7, 24 cm) was performed after passive rehydration with sample and then focused using a manifold-equipped IPGphor per manufacturer’s protocol. Second-dimensional sodium dodecyl sulfate polyacrylamide gel electrophoresis was performed using handcast gels for which one plate was pre-silanized using an Ettan DALT 12U per manufacturer’s protocol. Cy2/3/5-specific 16-bit data files were acquired at 100-μm resolution separately by dye-specific excitation and emission wavelengths using a Typhoon 9400 Variable Mode Imager and analyzed using DeCyder_6.5 software. Gels were counterstained for total protein content with SyproRuby (Molecular Probes/Invitrogen, Grand Island, NY), followed by robot excision of proteins of interest and subsequent processing for trypsin digestion and MS using an automated Spot Handling Workstation. Resulting peptides were subjected to C18 reverse-phase liquid chromatography coupled in-line with tandem MS using an LTQ Orbitrap tandem MS equipped with MicroAS autosampler and Surveyor HPLC pump, nanospray source, and Xcalibur_2.0 instrument control (Thermo Scientific, San Jose, CA). Tandem MS/MS data were searched against proteins extracted from UniProtKB database (www.uniprot.org), with taxonomy tag “Homo sapiens (human)” acquired on July 2009 (173,736 entries). This database was concatenated with protein sequences in reverse to enable false discovery rate calculations and also contained common laboratory contaminants. Searches were performed using Sequest and X! Tandem algorithms, allowing for cysteine carbamidomethylation and partial methionine oxidation. Results were validated and assembled into protein identifications using Scaffold_3.1.2 (Proteome Software Inc., Portland, OR). Peptide identifications were accepted if they could be established at more than 95.0% probability as specified by Peptide Prophet algorithm.¹⁰ Protein identifications were accepted if they could be established at more than 99.0% probability and contained two or more identified peptides. Protein probabilities were assigned by Protein Prophet algorithm.¹¹ Proteins containing similar peptides and not differentiated based on MS/MS alone were grouped to satisfy parsimony principles.

Ingenuity Pathway Analysis

Biological processes, molecular functions, and genetic/protein networks of identified proteins were analyzed using Ingenuity Pathway Analysis (IPA) tools (Ingenuity Systems, Mountain View, CA). The score associated with a particular network was the likehood of differentially expressed proteins being found together due to chance.

Western Blotting

Blots were probed with antibodies against human v-Akt murine thymoma viral oncogene homolog 1 (AKT) (1:1000; Cell Signaling, Danvers, MA), dihydropyrimidinase-related protein 2 (DPYL2) (1:1000; LSBio, Seattle, WA), phospho-DPYL2 (Thr⁵¹⁴/Ser⁵¹⁸) (1:1000; LSBio), E-cadherin (1:1000; BD Transduction Laboratories, San Jose, CA), endothelial growth factor receptor (EGFR) (1:750; Santa Cruz Biotechnology, Santa Cruz, CA), guanine nucleotide–binding protein alpha-q (GNAQ) (1:900; Abcam, Cambridge, MA), neural cell adhesion molecule (NCAM) (1:500; Sigma, St. Louis, MO), laminin receptor 1 (RSSA) (1:1000; Sigma), pontin (RUVB1) (1:1000; Atlas Antibodies, Stockholm, Sweden), stathmin 1 (STMN1) (1:1000; Cell Signaling), and phospho-STMN1 (Ser¹⁶) (1:1000; Cell Signaling). Protein expression was normalized to Actin (1:5000; Sigma).

Immunohistochemistry

Tissue sections were stained following previously reported protocol,¹² using antibodies against human DPYL2 (1:500; LSBio), GNAQ (1:100; Abcam), RSSA (1:150; LSBio), RUVB1 (1:100; Atlas Antibodies), and STMN1 (1:50; Cell Signaling). Staining intensity of cases represented in triplicate was evaluated by two independent observers (B.W., S.O.) as follows: 0, no staining; 1, weak; 2, moderate; and 3, strong. Staining intensity was then multiplied by stained tumor cell percentage to obtain the final staining score (range, 0–300).

Statistics

DIGE

In DeCyder_6.5, the normalized volume ratio of each protein spot feature from a Cy3- or Cy5-labeled sample was quantified relative to the Cy2 signal from the pooled-sample internal standard corresponding to the same spot feature, and all ratios were normalized across all 18 DIGE gels using the individual signals from the Cy2-labeled standard. Univariable analyses (Student’s t test and analysis of variants) were used to establish confidence associated with expression changes. Multivariate principal components analysis and hierarchical clustering (using Euclidean correlation and average linkage) were used to assess global variation arising from individual samples.

iPA

One-side right-tailed Fisher’s exact test (α = 0.05) was used to calculate the significance of the values assigned to the analyzed functions of proteins.

Immunohistochemistry

Wilcoxon rank-sum or Kruskal-Wallis tests were used to correlate average staining scores with categorical clinical outcomes, and Spearman’s rank correlation was used for continuous outcomes. OS was calculated from the date of diagnosis to the date of death or last date of contact for those alive at the time of analysis. Descriptive statistics, including median and interquartile ranges for continuous and percentages or frequencies for categorical variables, were reported. Multivariable analyses were performed using Cox proportional hazards model to adjust for age, pack years, histology, and stage. All tests were two-sided, and p values less than 0.05 were considered statistically significant. Analyses were performed using R_2.15.1.

RESULTS

Identification by DIGE of 137 Proteins Differentially Expressed in SCLC

We extracted MAPs from three SCLC (NCI-H69, NCI-H82, NCI-H209), three NSCLC (A549, NCI-H23, NCI-H520), and three INBE (BEAS-2B, HBEC-3KT, 16-HBE) cell lines in quadriplicate, obtaining 36 extracts. By Western blotting (WB), membrane proteins E-cadherin, NCAM, and EGFR were mainly expressed in membrane-associated, while cytosolic protein AKT was mainly expressed in cytosolic extracts (Fig. 1). This confirmed enrichment for MAPs in MAP extracts.

FIGURE 1.

Isolation of membrane-associated proteins. Membrane-associated (mbr) and cytosolic (cyto) extracts were resolved with sodium dodecyl sulfate polyacrylamide gel electrophoresis, and blots were incubated with anti-E-cadherin, NCAM, EGFR, AKT, or actin antibodies. E-cadherin, NCAM, EGFR, and AKT levels were normalized to actin. In each quadruplicate analysis, membrane proteins E-cadherin, NCAM, and EGFR were mainly expressed in membrane-associated extracts, whereas cytosolic protein AKT was mainly expressed in cytosolic extracts. Blots are representative of three independent experiments. AKT, v-Akt murine thymoma viral oncogene homolog 1; EGFR, epidermal growth factor receptor; NCAM, neural cell adhesion molecule.

MAP extracts were analyzed by DIGE according to the experimental design summarized in Supplementary table 1 (Supplemental Digital Content 1, http://links.lww.com/JTO/A529). On each of the 18 gels used, many differentially expressed protein spots were observed (representative DIGE image in Supplementary figure 1, Supplemental Digital Content 2, http://links.lww.com/JTO/A530). Principal components analysis on the global data set demonstrated that the four replicates derived from each of the nine cell lines clustered very closely by global expression patterns, as did samples within each classification (12 samples per classification) using the first three principle components (cumulative variance = 62%; Fig. 2A).

FIGURE 2.

DIGE results. A, PCA of the DIGE global data set. The axes correspond to the first three principal components. The apparent groups yielded by PCA are enclosed in circles: blue circle for immortalized normal bronchial epithelial cell lines (normal), green circle for NSCLC cell lines, and red circle for SCLC cell lines. The four replicates from each of the nine cell lines clustered very closely, as did samples within each histological subgroup. B, C, Quantitative analysis of protein expression in the nine cell lines analyzed by DIGE. Representative graphs of protein features significantly overexpressed (B) and underexpressed (C) in SCLC compared with immortalized normal bronchial epithelial and NSCLC cell lines (p < 10⁻³). The Y axis represents the standardized log protein abundance generated by the DeCyder analysis. DIGE, difference gel electrophoresis; NSCLC, non small-cell lung cancer; SCLC, small-cell lung cancer; PC, principle component; PCA, principal component analysis.

For further analyses, we selected 135 DIGE protein spots significantly over/underexpressed in SCLC compared with NSCLC and INBE cells (p < 10⁻³) (Fig. 2B). These protein spots were digested into peptides for tandem MS analysis, and 137 unique proteins were identified. Among them, 83 were from overexpressed (Table 1) and 54 from underexpressed spots (Supplementary table 2, Supplemental Digital Content 1, http://links.lww.com/JTO/A529).

TABLE 1.

Proteins Overexpressed in Small-Cell Lung Cancer, Membrane-Associated, and with Functions Relevant to Cancer Progression and/or Neuroendocrine Phenotype and/or Part of Networks Related to Cell/Tissue Morphology

Protein Symbol	Gene Symbol	Overexpressed in SCLC (N = 83)		Membrane-Associated (N = 32)	Interesting Functions for SCLC Progression (N = 17)			Final Candidate Biomarkers (N = 5)
		Av. Ratio	t test		Relevant to Cancer Progression	Relevant to Neuroendocrine Phenotype	Part of Cell/Tissue Morphology Networks
PRDX2	PRDX2	7.23	2.90E–11	x	x		x
STMN1	STMN1	5.71	1.30E–12	x	x	x	x	x
HNRPC	HNRNPC	5.38	8.90E–10
FPPS	FDPS	4.68	1.80E–09
STRAP	STRAP	4.68	1.80E–09	x	x
PSA1	PSMA1	4.18	3.10E–12
ATD3A	ATAD3A	3.92	4.90E–09	x
ERO1A	ERO1L	3.92	4.90E–09	x
G3BP1	G3BP1	3.92	4.90E–09	x
HSP71	HSPA1A-1B	3.92	4.90E–09
	MIR7-1	3.92	4.90E–09
HMGB1	HMGB1	3.67	6.90E–07
CALD1	CALD1	3.55	1.80E–09	x	x		x
SPF27	BCAS2	3.44	8.50E–13
ROAA	HNRNPAB	3.31	1.70E–08
CAPG	CAPG	3.06	2.40E–09	x			x
RS10	RPS10	3.06	5.90E–05
DPYL2	DPYSL2	2.73	4.10E–08	x	x	x	x	x
TCPG	CCT3	2.73	4.10E–08
AL9A1	ALDH9	2.72	4.50E–10
EFR1	ERVFRDE1	2.72	4.50E–10	x
IF2B	EIF2S2	2.72	4.50E–10
ODO2	DLST	2.72	4.50E–10	x
PTBP1	PTBP1	2.72	4.50E–10
RUVB2	RUVBL2	2.72	4.50E–10	x	x
SPF45	RBM17	2.72	4.50E–10
UAP56	BAT1	2.72	4.50E–10
BUB3	BUB3	2.71	2.50E–08
PCBP1	PCBP1	2.71	2.50E–08
RBM4B	RBM4B	2.71	2.50E–08
RSSA	RPSA	2.71	2.50E–08	x	x	x	x	x
EIF3H	EIF3H	2.69	3.10E–10
AHNK	AHNAK	2.64	2.40E–07
HSP7E	HSPA14	2.64	2.40E–07
IF5	EIF5	2.64	2.40E–07
SFRS6	SFRS6	2.64	2.40E–07
TCPQ	CCT8	2.64	2.40E–07
ARP2	ACTR2	2.62	1.40E–06
PURA	PURA	2.62	1.40E–06
AMPL	LAP3	2.61	4.20E–09
CSTF1	CSTF1	2.61	4.20E–09
ODB2	DBT	2.61	4.20E–09
RUVB1	RUVBL1	2.61	4.20E–09	x	x			x
CT007	C20orf7	2.56	6.00E–05	x
OCAD1	OCIAD1	2.42	2.80E–05
SYYC	YARS	2.39	2.80E–05
FUBP2	KHSRP	2.29	0.00059
DNJC3	DNAJC3	2.14	6.20E–07
GLSK	GLS	2.14	6.20E–07
PRDX4	PRDX4	2.14	9.50E–06	x	x
PRP19	PRPF19	2.14	6.20E–07
ARSA1	ASNA1	2.1	3.80E–08	x
FKBP8	FKBP8	2.03	6.30E–08	x		x	x
HNRH1	HNRNPH1	2.03	6.30E–08
TBB5	TUBB	2.03	6.30E–08	x			x
CH60	HSPD1	1.97	2.60E–05	x	x		x
NONO	NONO	1.97	2.60E–05
SF3A3	SF3A3	1.97	2.60E–05
TBL1R	TBL1XR1	1.97	2.60E–05
TPR	TPR	1.92	0.0006	x			x
ACTB	ACTB	1.9	2.60E–06
RANG	RANBP1	1.9	2.60E–06
ACON	ACO2	1.83	3.40E–05
EFG2	GFM2	1.83	3.40E–05
IMMT	IMMT	1.83	3.40E–05	x
MIRO2	RHOT2	1.83	3.40E–05	x
PLOD3	PLOD3	1.83	3.40E–05	x			x
XRCC5	XRCC5	1.83	3.40E–05	x
CORO1A	CORO1A	1.82	2.80E–05	x	x		x
DLDH	DLD	1.82	2.80E–05
SCOT1	OXCT1	1.82	2.80E–05
EIF3G	EIF3G	1.81	6.00E–07
SCMC1	SLC25A24	1.81	6.00E–07	x
TADBP	TARDBP	1.81	6.00E–07
GNAQ	GNAQ	1.75	2.50E–05	x	x	x	x	x
IDH3B	IDH3B	1.75	2.50E–05
UBFD1	UBFD1	1.75	2.50E–05
RBBP7	RBBP7	1.73	0.0056
ATPB	ATP5B	1.65	4.90E–05	x
TPD52	TPD52	1.65	0.00054	x
KAP2	PRKAR2A	1.62	0.00062	x
PDIA6	PDIA6	1.62	0.00062	x
PCBP2	PCBP2	1.56	3.80E–06

SCLC, small-cell lung cancer.

Bold indicates candidate biomarkers.

Overrepresentation in SCLC of Proteins Involved in Cellular/Tissue Morphology Networks

Using IPA, we found that the 137 proteins differentially expressed in SCLC were overrepresented in five specific networks: (1) gene expression, cellular response to therapeutics, cellular assembly, and organization (score: 60); (2) protein synthesis, RNA post-transcriptional modification, gene expression (score: 49); (3) cell death, cellular assembly and organization, cellular function, and maintenance (score: 47); (4) carbohydrate metabolism, small molecule biochemistry, DNA replication, recombination, and repair (score: 29); and (5) hematological system development/function, humoral immune response, tissue morphology (score: 23). Proteins related to cellular assembly, organization, and morphology and tissue morphology (Table 3) were overrepresented in three of these five networks, the most interesting one being the cell death, cellular assembly, organization, function, and maintenance network (Supplementary figure 2, Supplemental Digital Content 3, http://links.lww.com/JTO/A531). Many proteins represented in these three networks were underexpressed in SCLC (Table 4). Among proteins represented in these top five networks and overexpressed in SCLC, we found proteins related to the nervous system (Table 5) and involved in inhibition of apoptosis and induction of proliferation and motility (Supplementary table 3, Supplemental Digital Content 1, http://links.lww.com/JTO/A529).

TABLE 3.

Proteins Related to Cellular Assembly, Organization, Morphology, and Tissue Morphology (Including Proteins Related to Nervous System) in the Top Five Networks in Ingenuity Pathway Analysis

Protein Symbol	Function	p	Number of Proteins
ACTN4, HMGB1, HSP71, K1C18, PRDX2, PSA, PTBP1, ROAA, STMN1, TMOD3, TPD52, TPR, VIME, VINC	Cell morphology	0.000079	14
ACTB, ACTN4, ARP2, CALD1, CH60, CORO1A, CRC8, HMGB1, NPM, STMN1, TBB5, TMOD3, VIME, VINC, SYYC	Cell motility	0.0362	15
CAPG, CORO1A, CRC8, DPYL2, GNAQ, PTBP1, VINC	Shape change of cells	0.0439	7
CRC8, DPYL2	Cell blebbing	0.0123	2
CAPG, VINC	Biogenesis of plasma membrane projections	0.0273	2
CALD1	Quantity of podosomes	0.00633	2
CALD1, CRC8	Formation of podosomes	0.00633	2
PTBP1, VINC	Cell spreading	0.0412	2
CALD1, GNAQ, VIME	Contraction of cells	0.0314	3
STMN1, VIME	Quantity of microtubules	0.0123	2
DPYL2, STMN1, TBB5	Assembly of microtubules	0.00468	3
STMN1, TBB5	Polymerization of microtubules	0.0199	2
RRBP1, TBB5	Organization of microtubules	0.0159	2
CORO1A, K1C18, RRBP1, TBB5	Organization of filaments	0.0299	4
CRC8, K1C18, STMN1	Stabilization of filaments	0.0131	3
ACTN4	Elongation of actin filaments	0.0397	1
K1C18, VIME	Quantity of intermediate filaments	0.0397	1
VIME	Assembly of intermediate filaments	0.0397	1
K1C18	Organization and stabilization of keratin filaments	0.0161	1
VINC	Organization of zonula adherens, junctions and complexes integrity	0.0161	1
CALD1, PTBP1	Size of focal adhesions	0.00175	2
HSP71	Reformation of focal adhesions	0.00806	1
CRC8	Formation of cell-matrix contacts, (de)branching of actin filaments	0.0161	1
RSSA	Cell adhesion	0.0474	1
GNAQ	Remodeling of neurons	0.0474	1
DPYL2	Induction of axons and formation of supernumerary axons	0.00806	1
VIME	Loss of synapse	0.0474	1
FKBP8	Morphology of neural tube	0.024	1
VINC	Development of spinal nerve	0.00806	1
STIP1	Neuritogenesis of hippocampal neurons	0.0397	1

ACTB, β actin; ACTN4, actinin alpha 4; ARP2, actin-related protein 2; CALD1, caldesmon 1; CAPG, capping protein (actin filament), gelsolin-like; CH60, 60 kDa chaperonin; COR1A, coronin 1A; CRC8, cortactin; DPYL2, dihydropyrimidinase-related protein 2; FKBP8, FK506-binding protein 8; GNAQ, guanine nucleotide–binding protein alpha-q; HMGB1, high mobility group box 1; HSP71, heat shock 70 kDa protein 1A; K1C18, cytokeratin 18; NPM, nucleophosmin; PRDX2, peroxiredoxin 2; PSA, puromycin-sensitive aminopeptidase; PTBP1, polypyrimidine tract–binding protein 1; ROAA, heterogeneous nuclear ribonucleoprotein A/B; RRBP1, ribosome-binding protein 1; RSSA, laminin receptor 1; STIP1, stress-induced-phosphoprotein 1; STMN1, stathmin 1; SYYC, tyrosyl-tRNA synthetase; TBB5, tubulin β5; TMOD3, tropomodulin 3; TPD52, tumor protein D52; TPR, translocated promoter region—to activated MET oncogene; VIME, vimentin; VINC, vinculin.

Bold indicates candidate biomarkers.

TABLE 4.

Proteins Overrepresented in the Cellular Assembly, Organization, Morphology, and Tissue Morphology Networks and Downregulated in Small-Cell Lung Cancer Cell Lines

Protein Symbol	Function
ACTN4	Elongation of actin filaments Cellular adhesion
CRC8	Debranching and branching of actin filaments Cell blebbing Formation of podosomes and of cell-matrix contacts
K1C18	Organization and stabilization of keratin filaments
TMOD3	Cell morphology
VIME	Contraction of cells Quantity of microtubules Quantity and assembly of intermediate filaments
VINC	Organization of zonula adherens and adherens junctions Integrity of junctional complexes Formation of membrane projections and cell spreading

ACTN4, actinin alpha 4; CRC8, cortactin; K1C18, cytokeratin 18; TMOD3, tropomodulin 3; VIME, vimentin; VINC, vinculin.

TABLE 5.

Proteins Related to the Nervous System in the Top Five Networks in Ingenuity Pathway Analysis

Protein Symbol	Function	p
GNAQ	Remodeling of neurons	0.0474
DPYL2	Induction of axons and formation of supernumerary axons	0.00806
VIME	Loss of synapse	0.0474
FKBP8	Morphology of neural tube	0.024
VINC	Development of spinal nerve	0.00806
STIP1	Neuritogenesis of hippocampal neurons	0.0397

DPYL2, dihydropyrimidinase-related protein 2; FKBP8, FK506-binding protein 8; GNAQ, guanine nucleotide–binding protein alpha-q; STIP1, stress-induced-phosphoprotein 1; VIME, vimentin; VINC, vinculin.

Bold indicates candidate biomarkers.

The top biological functions of the proteins differentially expressed in SCLC were also determined using IPA (Supplementary table 4, Supplemental Digital Content 1, http://links.lww.com/JTO/A529). Consistent with the network analysis, we found that the functions of these proteins were related to the cytoskeleton, morphology, and interactions with the environment and to the neuroendocrine system, cell cycle, growth, apoptosis, and metabolism.

Selection of DPYL2, GNAQ, RSSA, RUVB1, and STMN1 as SCLC Candidate Biomarkers

To select SCLC candidate biomarkers among the 137 proteins differentially expressed in SCLC, we used three criteria: (1) proteins overexpressed in SCLC (N = 81), (2) confirmed membrane-associated (N = 32) (including 10 plasma membrane, 12 endomembrane, and 10 MAPs), and (3) with functions relevant to cancer progression and/or neuroendocrine phenotype and/or part of networks related to cell/tissue morphology (Table 1). Five proteins met these criteria: DPYL2, GNAQ, RSSA, RUVB1, and STMN1 (main functions reported on Table 2).

TABLE 2.

Subcellular Location and Functions of the Final Candidate Biomarkers

Protein Symbol	Subcellular Location	Interesting Functions for SCLC Progression
STMN1	Cytosolic MAP	Role in cell cycle progression and migration through microtubule depolymerization Possible role in axon formation during neurogenesis and neuron polarization Overexpression plays a role in tumor progression in various cancers, and contributes to chemoresistance
DPYL2	PMP and cytosolic MAP	Role in neuronal differentiation, axon guidance, neuronal growth, and cell migration in developing nervous system Associated with various neuropathologic or psychiatric disorders but not with cancer
RSSA	PMP (mainly), cytosolic and nuclear	Laminin receptor, role in cell adhesion to basal membrane and activation of signaling transduction pathways Overexpression correlated with invasive phenotype in various cancers
RUVB1	Nuclear (mainly) and PMP	AAA+ ATPase interacting with several transcription factors involved in carcinogenesis Role in tumor invasion by binding plasminogen at cell surface and increasing its activation into plasmin Overexpression observed in many cancers
GNAQ	PMP and cytosolic MAP	Couples cell surface transmembrane receptors to intracellular signaling pathways, including the MAPK pathway Activating mutations described in melanoma

MAP, membrane-associated protein; PMP, plasma membrane protein; SCLC, small-cell lung cancer.

To verify the expression of our candidates, we analyzed the MAP lysates initially used for DIGE by WB. Although there was no expression difference among SCLC, NSCLC, and INBE cells for RSSA and RUVB1, we confirmed DPYL2, GNAQ, and STMN1 overexpression in SCLC (Fig. 3A). Moreover, we found phospho-DPYL2 (Thr⁵¹⁴/Ser⁵¹⁸) and phospho-STMN1 (Ser¹⁶) overexpression in SCLC, suggesting the activation of DPYL2 and STMN1 in SCLC (Fig. 3).

FIGURE 3.

Verification of the candidate biomarkers’ expression by Western blotting. Membrane-associated protein lysates were resolved with sodium dodecyl sulfate polyacrylamide gel electrophoresis, and blots were incubated with an antibody against DPYL2, phospho-DPYL2 (Thr⁵¹⁴/Ser⁵¹⁸), GNAQ, RSSA, RUVB1, STMN1, or phospho-STMN1 (Ser¹⁶). Expression levels were normalized to actin. Overexpression of DPYL2, GNAQ, and STMN1 was observed in SCLC compared with immortalized normal bronchial epithelial (normal) and NSCLC cell lines, while RSSA and RUVB1 expression was similar in all the cell lines. Increased phosphorylation of DPYL2 and STMN1 was also observed in SCLC compared with normal and NSCLC cell lines, suggesting an activation of DPYL2 and STMN1 in SCLC. DPYL2, dihydropyrimidinase-related protein 2; GNAQ, guanine nucleotide binding protein alpha-q; NSCLC, non small-cell lung cancer; RSSA, laminin receptor 1; RUVB1, pontin; SCLC, small-cell lung cancer; STMN1, stathmin 1.

Next, we performed immunohistochemistry (IHC) on TMAs made of 100 NSCLCs and 36 SCLCs. Protein expression was membranous and cytosolic for DPYL2, GNAQ, and STMN1 and cytosolic for RSSA and RUVB1. Median staining scores in NSCLC and SCLC, respectively, were 180 (range, 50–300) and 300 (range, 60–300) for DPYL2 (p < 0.001), 80 (range, 0–300) and 250 (range, 30–300) for GNAQ (p < 0.001), 100 (range, 0–300) and 150 (range, 0–300) for RSSA (p = 0.051), 28 (range, 0–300) and 180 (range, 0–200) for RUVB1 (p < 0.001), and 0 (range, 0–240) and 240 (range, 0–300) for STMN1 (p < 0.001). Thus, IHC on TMAs verified overexpression in SCLC of all candidate biomarkers but RSSA. STMN1 was strongly expressed in SCLCs, whereas it was absent in most NSCLCs. Representative images of NSCLC and SCLC tissues with staining scores close to average staining scores for each biomarker in each histological subgroup are displayed in Figure 4.

FIGURE 4.

Verification of the candidate biomarkers’ expression by immunohistochemistry: representative cases of 130 studied NSCLC and 36 studied SCLC. Tissue sections were incubated with an antibody against DPYL2, GNAQ, RSSA, RUVB1, or STMN1. A J, Representative images of NSCLC and SCLC tissues with a staining score close to the average staining score for each biomarker in each histological subgroup. A, NSCLC tumor with a DPYL2 staining score of 200. B, SCLC tumor with a DPYL2 staining score of 300. C, NSCLC tumor with a GNAQ staining score of 100. D: SCLC tumor with a GNAQ staining score of 300. E, NSCLC tumor with an RSSA staining score of 100. F, SCLC tumor with an RSSA staining score of 100. G, NSCLC tumor with an RUVB1 staining score of 100. H, SCLC tumor with an RUVB1 staining score of 200. I, NSCLC tumor with an STMN1 staining score of 0. J, SCLC tumor with an STMN1 staining score of 300 (magnification: ×10 in the large windows and ×40 in the small windows; scale bars: 100 μm in the large windows and 20 μm in the small windows). DPYL2, dihydropyrimidinase-related protein 2; GNAQ, guanine nucleotide binding protein alpha-q; NSCLC, non small-cell lung cancer; RSSA, laminin receptor 1; RUVB1, pontin; SCLC, small-cell lung cancer; STMN1, stathmin 1.

Correlation of the Candidate Biomarkers’ Expression with Clinical Characteristics

Results of the correlations between the candidate biomarkers’ staining scores and various clinical characteristics in univariable analyses are presented in Table 6. In these univariable analyses, staining scores were significantly higher in SCLC than in NSCLC for DPYL2, GNAQ, RUVB1, and STMN1 (p < 0.001). OS was significantly shorter in the presence of higher staining scores for DPYL2 (p = 0.022), GNAQ (p = 0.026), RUVB1 (p = 0.024), and STMN1 (p = 0.001). However, these were not associated with OS after tumors were broken by histology.

TABLE 6.

Patient Characteristics and Statistical Significance of the Association with DPYL2, GNAQ, RSSA, RUVB1, and STMN1 Staining Scores in Univariables Analyses

	n (%)	IQR	95% CI	p
				DPYL2	GNAQ	RSSA	RUVB1	STMN1
Sex
Male	78 (57)			0.583	0.236	0.22	0.211	0.258
Female	58 (43)
Race
Caucasian	126 (93)			0.240	0.222	0.422	0.949	0.857
African American	10 (7)
Age at diagnosis
Median, yr		67 (58–73)		0.358	0.710	0.278	0.646	0.897
≤60	45 (33)			0.740	0.475	0.760	0.713	0.852
>60	91 (67)
Smoking status
Current smoker	45 (33)			0.689	0.408	0.633	0.059	0.071
Ex smoker	85 (63)
Never smoker	5 (4)
Smoking history
Median pack year		50 (35–75)		0.301	0.163	0.659	0.643	0.231
Lung cancer histology
NSCLC	100 (74)			<0.001	<0.001	0.051	<0.001	<0.001
SCLC	36 (26)
NSCLC histology
ADC	49 (36)			0.006	<0.001	0.541	0.947	0.184
SqCC	42 (31)
SCLC path stage
Limited	24 (67)			0.262	0.687	0.378	0.309	0.240
Extensive	12 (33)
NSCLC path stage
I	65 (66)			0.072	0.126	0.209	0.813	0.009
II	13 (13)
III	17 (17)
IV	3 (3)
Overall survival
Median, yr			2.7 (1.9–4.0)	0.022	0.026	0.651	0.024	0.001
Alive at 3 yr	66 (49)			0.594	0.072	0.392	0.189	0.017
SCLC overall survival
Median, yr			1.1 (0.7–2.3)	0.466	0.427	0.643	0.778	0.705
Alive at 3 yr	9 (25)			0.779	0.102	0.955	0.576	0.415
NSCLC overall survival
Median, yrs			4.0 (2.6–5.6)	0.134	0.763	0.312	0.947	0.801
Alive at 3 yrs	57 (57)			0.139	0.64	0.103	0.624	0.863
ADC overall survival
Median, yr			4.4 (2.5–5.8)	0.295	0.926	0.678	0.543	0.642
Alive at 3 yr	31 (63)			0.502	0.610	0.848	0.922	0.749
SqCC overall survival
Median, yr			3.9 (2.4–5.7)	0.147	0.855	0.146	0.504	0.862
Alive at 3 yr	23 (55)			0.082	0.345	0.017	0.176	0.946

ADC, adenocarcinoma; CI, confidence interval; DPYL2, dihydropyrimidinase-related protein 2; GNAQ, guanine-nucleotide-binding protein alpha-q; IQR, interquartile range; NSCLC, non–small-cell lung carcinoma; RSSA, laminin receptor 1; RUVB1, pontin; SCLC, smallcell lung carcinoma; SqCC, squamous cell carcinoma; STMN1, stathmin 1.

Bold indicates statistically significant P values.

In multivariable analyses including each biomarker’s average staining score, lung cancer histology, disease stage, age, sex, and smoking status, we did not confirm the association between staining scores and OS (Table 7). However, age was an independent predictor of OS (nonlinear trend) for all biomarkers. For the subset of 36 SCLCs, we did not find a significant association between staining scores and OS in multivariable analyses including each biomarker’s average staining score, disease stage, and age (Table 8).

TABLE 7.

Multivariable Analyses for the Association of Candidate Biomarkers’ Staining Scores with Overall Survival for the Whole Data Set, Including Non Small-Cell and Small-Cell Lung Cancers (n = 136)

	p
	DPYL2	GNAQ	RSSA	RUVB1	STMN1
Average staining score	0.0881	0.5177	0.1500	0.7983	0.6491
Lung cancer histology	0.0047	0.1720	0.0046	0.0540	0.1711
Disease stage	0.0962	0.1038	0.0943	0.0647	0.0760
Age	0.0375	0.0328	0.0175	0.0380	0.0280
Sex (male vs. female)	0.0604	0.2517	0.0515	0.0984	0.1114
Smoking status (ex vs. current)	0.8785	0.9432	0.7353	0.9380	0.8702

DPYL2, dihydropyrimidinase-related protein 2; GNAQ, guanine nucleotide–binding protein alpha-q; RSSA, laminin receptor 1; RUVB1, pontin; STMN1, stathmin 1.

Bold indicates statistically significant P values.

TABLE 8.

Multivariable Analyses for the Association of Candidate Biomarkers’ Staining Scores with Overall Survival for the Subset of Small-Cell Lung Cancers (n = 36)

	p
	DPYL2	GNAQ	RSSA	RUVB1	STMN1
Average staining score	0.3468	0.0795	0.9800	0.8892	0.3109
Disease stage	0.4081	0.2269	0.2836	0.2329	0.1399
Age	0.0251	0.0111	0.0349	0.0926	0.0459

DPYL2, dihydropyrimidinase-related protein 2; GNAQ, guanine nucleotide–binding protein alpha-q; RSSA, laminin receptor 1; RUVB1, pontin; STMN1, stathmin 1.

Bold indicates statistically significant P values.

DISCUSSION

The main objective of this study was to test the hypothesis that the identification of MAPs specific to SCLC by DIGE may advance our understanding of SCLC biology and lead to the discovery of new candidate biomarkers of SCLC. To do so, we analyzed a membrane-associated subproteome in three SCLC, three NSCLC, and three INBE cells by DIGE and validated these results in primary SCLCs.

DIGE is a proteomic tool widely used to study cancer biology. In a previous study, DIGE analysis of 30 lung cancer cell lines including SCLCs, SqCCs, and ADCs led to the discovery of a 32-protein spots signature able to distinguish between the three histological types, further validated in tissues (29 NSCLCs and 1 SCLC).⁶ Other DIGE studies found proteins overexpressed in SCLC when compared with control cells¹³ or normal bronchial epithelium tissues.¹⁴ Only one DIGE study specifically analyzed MAPs in the SCLC cell line H69 and its doxorubicin-resistant subcell line H69AR and found, in H69AR, upregulation of Serca2, a calcium pump located in the endoplasmic reticulum membrane.¹⁵

In our DIGE analysis, replicates from each cell line clustered very closely, demonstrating low technical noise and high reproducibility in membrane fractionation. Samples within each histological group also clustered very closely, indicating that many variations were descriptive of differences between classifications. Further analyses revealed 135 distinct protein features differentially expressed in SCLC when compared with NSCLC and INBE cells, including 137 proteins. These were overrepresented in the cellular assembly, organization, morphology, and tissue morphology networks. Among proteins represented in these networks and underexpressed in SCLC, many were involved in cell–cell and cell–extracellular matrix interactions. This was consistent with the nonadherent phenotype that many SCLC cells have in vitro (including these in this study). Among proteins part of these networks and overexpressed in SCLC, there were proteins involved in nerve/brain development and endocrine functions, according to the neuroendocrine nature of SCLC cells and previous proteomic studies.¹³ Many proteins were involved in cell cycle, growth, apoptosis, and metabolism, which were also consistent with SCLC aggressive behavior. DPYL2, GNAQ, RSSA, RUVB1, and STMN1 were selected as candidate biomarkers among MAPs overexpressed in SCLC and with cellular functions relevant to cancer progression (all of them) and/or the neuroendocrine phenotype (GNAQ, DPYL2, and STMN1), and/or part of networks related to cell/tissue morphology (all of them but RUVB1). Our ultimate goal was to find molecular tools to move our discoveries toward SCLC molecular imaging or targeted therapies. Interestingly, overexpression of DPYL2, GNAQ, RUVB1, and STMN1 in SCLC was verified by WB and/or IHC on TMAs, supporting their potential as targets in SCLC.

In our DIGE analysis, STMN1 was the most significantly overexpressed protein in SCLC cell lines. By IHC, STMN1 was also strongly expressed in most SCLC tissues, while the expression was absent in most NSCLCs and very low in the remaining minority. Moreover, in IPA, STMN1 was part of molecular networks related to cellular/tissue morphology. STMN1 is a membrane-associated phosphoprotein playing a role in cell cycle progression and migration through microtubule depolymerization.^16,17 STMN1 phosphorylation at Ser¹⁶ may also be required for axon formation during neurogenesis and neuron polarization, which is relevant for cells with neuroendocrine differentiation. STMN1 overexpression plays a role in tumor progression and metastasis in various cancers and contributes to chemoresistance.¹⁸ Therefore, STMN1 appears as an attractive therapeutic target candidate. Recent integrative genome analyses showed high STMN1 gene expression levels in 15 SCLC tumors,¹⁹ confirming our data adequately. Although STMN1 protein expression was absent in most of the NSCLC tissues we tested, previous studies demonstrated STMN1 overexpression in lung ADC and SqCC.^20,21 However, consistent with our results, other studies described higher STMN1 expression in SCLC than NSCLC.^22,23 Moreover, by WB, we showed higher STMN1 phosphorylation in SCLC than NSCLC and INBE cells, suggesting the activation of STMN1 in SCLC and a potential role in SCLC progression. STMN1 expression in NSCLC and SCLC was not found to be an independent predictor of OS. This is in contrast with observations made in other cancers¹⁸ and may be explained by the relative small number of SCLCs studied here (N = 36).

DPYL2 plays a role in neuronal differentiation, axon guidance, neuronal growth, and cell migration in the developing nervous system,²⁴ which is once again relevant for SCLC. DPYL2 has been associated with various neuropathologic or psychiatric disorders²⁴ but never with cancer, even though the gene is located in 8p, an oncogene- and tumor suppressor gene-rich region.²⁵ This is therefore the first report of its overexpression in SCLC. Moreover, we demonstrated higher DPYL2 phosphorylation in SCLC than NSCLC and INBE cells, suggesting DPYL2 activation in SCLC and its possible role in SCLC progression.

GNAQ is a membrane and membrane-bound protein from the family of guanine nucleotide–binding proteins that couple cell surface transmembrane receptors to intracellular signaling pathways, including the mitogen activated protein kinase (MAPK) pathway. Activating mutations of GNAQ have been described in melanoma, rendering them potentially susceptible to mitogen-activated protein kinase kinase kinase 1, E3 ubiquitin protein ligase (a kinase in the MAPK pathway) inhibition^26,27 but have not been found in other tumors and particularly not in lung cancer.²⁸ GNAQ overexpression in SCLC has never been reported before and deserves further investigation.

Finally, RUVB1 is an AAA+ ATPase interacting with several transcription factors involved in carcinogenesis, such as c-MYC.²⁹ RUVB1 localized at cell surface may also be able to bind plasminogen and increase its activation into plasmin,³⁰ which plays a role in tumor invasion. RUVB1 overexpression has been observed in many cancers, including NSCLC.³¹ However, this is the first report of RUVB1 overexpression in SCLC, especially when compared with NSCLC. Although RUVB1 expression has been correlated with poor outcome in colorectal carcinoma,³² we did not confirm this correlation with OS in our study. Small molecule inhibitors of RUVB1 have recently been developed, are commercially available,³³ and will facilitate testing the functional implications of RUVB1.

Among the five SCLC candidate biomarkers we selected by DIGE, the only one for which we did not verify the overexpression in SCLC when compared with NSCLC was RSSA, a laminin receptor mainly localized at the surface of the plasma membrane. RSSA plays a role in cell adhesion to the basal membrane and the consequent activation of signaling transduction pathways. Its overexpression in many cancers,^34–36 including NSCLC and SCLC,³⁷ has already been described. A correlation between RSSA upregulation in cancer cells and their invasive phenotype has also been reported.³⁸ Consistently, the expression of RSSA has been found to be higher in SCLC than in NSCLC³⁷ and to be a predictor of poor prognosis in many cancers, including NSCLC.³⁹ However, the higher expression of RSSA in SCLC than in NSCLC and the prognostic value of RSSA were not verified in our study. These discordant results may be related to differences in methodologies and antibodies used to evaluate RSSA expression or to sample size issues.

We specifically focused on MAPs because of their potential as diagnostic or therapeutic target. Yet, cell surface proteins are often poorly represented in two-dimensional gels due to their hydrophobic nature and high molecular weight.⁴⁰ Therefore, this study probably underestimated the number of MAPs important for SCLC progression. By restricting our study to MAPs, we omitted proteins potentially relevant to SCLC biology. Moreover, potentially relevant proteins with high expression level did not fulfil our three selection criteria, leaving aside other potential candidate biomarkers (available in Table 1). Another limitation of the study is the small number of cell lines used for the DIGE analysis, although the results were validated in a larger group of tumors from patients with SCLC and NSCLC. Of note, the mutation status of the cell lines may have limited the generalizability of the findings. We therefore searched the Sanger Center database (http://cancer.sanger.ac.uk/cancergenome/projects/cosmic/) for the mutation status of well-known oncogenes or tumor suppressor genes in the cell lines we used. We found that none of them had AKT1, EGFR, v-Raf Murine sarcoma viral oncogene homolog B, MAP2K4, MET, Myc, or NRAS mutation, while KRAS mutation was found in NCI-H23, PIK3CA mutation in NCI-H69, PTEN mutation in NCI-H23, retinoblastoma 1 mutation in all the three SCLC cell lines, and TP53 mutation in NCI-H69, NCI-H209, NCI-H23, and NCI-H520. To the exception of retinoblastoma 1 mutation found in all the SCLC cell lines (reported in >90% of SCLCs), no other mutation was found in every cell line of the same histological subgroup.

In summary, using DIGE coupled with tandem MS to analyze a membrane-associated subproteome, we identified 137 proteins differentially expressed in SCLC compared with NSCLC and INBE cells. These proteins were enriched for cellular/tissue morphology networks. DPYL2, GNAQ, RUVB1, and STMN1 overexpression in SCLC was verified by WB and/or IHC, suggesting that these results need to be tested for functional implication in SCLC progression. The association with survival requires further validation in larger clinical data sets.