Significant high expression of cytokeratins 7, 8, 18, 19 in pulmonary large cell neuroendocrine carcinomas, compared to small cell lung carcinomas


Yuichi Sato, PhD, Department of Molecular Diagnostics, School of Allied Health Sciences, Kitasato University, 1-15-1, Kitasato, Sagamihara, Kanagawa 228-8555, Japan. Email:


The aim of the present study was to clarify protein profiling in small cell lung carcinoma (SCLC) and pulmonary large cell neuroendocrine carcinoma (LCNEC). The proteomic approach was used, and involved cell lysate from two cell lines (N231 derived from SCLC and LCN1 derived from LCNEC), with 2-D gel electrophoresis (2-DE). In the present study, 25 protein spots with greater than twofold quantitative differences between LCN1 and N231 cells on 2-DE gels were confirmed. Within the 25 identified proteins, cytokeratins (CK) 7, 8, 18 and 19 were upregulated in LCN1 cells compared with N231 cells. The expression of CK7, 8, 18, and 19 was further studied on immunohistochemistry with 81 formalin-fixed and paraffin-embedded pulmonary carcinomas, which included 27 SCLC, 30 LCNEC, 14 adenocarcinomas, and 10 squamous cell carcinomas. Although the expression of CK7, 8, 18, and 19 was observed in all histological types, the mean immunostaining scores of CK7, 8, 18, and 19 were significantly higher in LCNEC than in SCLC (P < 0.001, P < 0.001, P < 0.01 and P < 0.001, respectively). These data suggest that the biological characteristics of LCNEC and SCLC may be different and the expression of CK may serve as differential diagnostic markers.

Neuroendocrine carcinoma is the general term for carcinomas that secrete or express various peptide hormones and biogenic amines (such as adrenocorticotropic hormone, gastrin-releasing peptide, calcitonin, and serotonin). Generally, neuroendocrine carcinomas have morphological characteristics such as organoid structures, palisading basal cell arrangement and rosette formation. In 1991, Travis et al. introduced the term ‘large cell neuroendocrine carcinoma’ (LCNEC) to describe a distinct category of high-grade neuroendocrine tumor with biological and light microscopy characteristics different from those of high-grade small cell lung carcinoma (SCLC).1

Morphologically, SCLC is composed of small (most cells less than the nuclear diameter of three small resting lymphocytes), round to fusiform cells with a high nuclear/cytoplasmic ratio, hyperchromatic nuclei with fine chromatin, and absent or inconspicuous nucleoli. The mitotic index is high. Although chemotherapy and radiotherapy are more effective against SCLC than the other histological types, the prognosis of SCLC patients is very poor because most tumors relapse after chemoradiotherapy. The 5 year survival rate of SCLC patients is approximately 35.7%.2

In contrast, however, LCNEC is also characterized by neuroendocrine morphology (rosette formation); the tumor cells are large (threefold larger in diameter than a small resting lymphocyte) and tend to be polygonal rather than fusiform, with a low nuclear/cytoplasmic ratio and prominent nucleoli. The nuclear chromatin tends to be coarse and granular. The 5 year survival rate of LCNEC patients is 40.3%, which is not significantly different to that of SCLC.2

Although LCNEC appears to fall between atypical carcinoid (AC) and SCLC, it is difficult to differentiate LCNEC from SCLC, and definitive discrimination points (except for morphological characteristics) and the details of its biological behavior, including tumor aggressiveness and degree of differentiation, remain unclear.

In the present study, to clarify the biological differences of SCLC and LCNEC, we performed protein profiling using 2-D gel electrophoresis with an agarose isoelectric focusing gel in the first dimension (agarose 2-DE). Agarose 2-DE is unique in that it can analyze much larger quantities and a wider dynamic range of proteins than 2-DE with immobilized pH gradient (IPG) gel for isoelectric focusing (conventional 2-DE), and is also able to resolve high-molecular-weight proteins >100 kDa, which are difficult to resolve on conventional 2-DE.3 We identified proteins with more than twofold quantitative differences between LCNEC cells and SCLC cells using matrix-assisted laser desorption/ionization time of flight/time of flight mass spectrometry (MALDI-TOF/TOF-MS).


Cell lines

LCN1, an LCNEC line, was established in our laboratory at Kitasato University.4 The SCLC line N231 was purchased from the American Type Culture Collection (Rockville, MD, USA). The cell lines were grown in RPMI-1640 medium (Sigma, Steinheim, Germany) supplemented with 10% fetal bovine serum (Biowest, Miami, FL, USA), 100 units/mL penicillin, and 100 µg/mL streptomycin (Gibco, Auckland, New Zealand). Subconfluented cells were harvested and washed twice with PBS without bivalent ions, and were partly fixed in 10% formalin and embedded in paraffin for immunohistochemical staining and a sample was stored at −80°C until proteome analysis.


Eighty-one cases of surgically resected lung cancer tissues at Kitasato University Hospital were used in the present study. They were divided into 27 SCLC, 30 LCNEC, 14 adenocarcinomas (AD), and 10 squamous cell carcinomas (SCC).

This study was approved by the Ethics Committee of Kitasato University School of Medicine. All patients were informed of the aim of the study and gave consent to donate their samples.

Agarose 2-DE

The solubilization of cells and quantification of cell lysates have been described in a recent study.5 The agarose-2-DE method used in the present study was previously described by Oh-Ishi et al. 3 After 2-DE, the gel was visualized on Coomassie Brilliant Blue R-350 (CBB, PhastGel Blue R; Amersham Pharmacia Biotech, Uppsala, Sweden) staining. Each agarose 2-DE was performed twice. Protein patterns in gels were recorded as digitalized images using a high-resolution scanner (GT-9800; Epson, Tokyo, Japan), and the intensity of each spot was compared. Each spot was analyzed using Scion Image Beta 4.02 (Scion, Fredrick, MD, USA) and the abundance of the same proteins was compared. Protein expression levels more than twofold different between LCN1 and N231 cells progressed to in-gel digestion.

Identification of proteins differently expressed between LCN1 and N231

In-gel digestion

In brief, protein spots were excised from a 2-DE gel, destained with 50% (v/v) acetonitrile (ACN)/50 mmol/L NH4HCO3, dehydrated with 100% (v/v) ACN, and dried under vacuum conditions. Tryptic digestion was performed for 24 h at 37°C in a minimum volume of digestion solution that contained 20 ng/µL trypsin (Trypsin Gold, Mass Spectrometry Grade; Promega, Madison, WI, USA) and 25 mmol/L NH4HCO3. After incubation, digested protein fragments eluted in solution were collected, and gels were washed once in 5% (v/v) trifluoroacetic acid/50% (v/v) ACN and collected in the same tube.

Protein identification

Tryptic peptides were spotted on a Prespotted AnchorChip 96 Set for Proteomics (Bruker Daltonik, Bremen, Germany) according to the manufacturer's recommendations. MS spectra were analyzed in an autoflex III TOF/TOF (Bruker Daltonik) in reflector mode by summarizing 1000 single spectra (5 × 200) with a 50 Hz laser in the mass range from 580 to 4000 Da applying the following instrument settings: ion source 1, 19.00 kV; ion source 2, 16.60 kV; lens, 8.55 kV; reflector 1, 21.00 kV; reflector 2, 9.70 kV; reflector detector, 1400 V; suppression up to 500 Da by deflection.

MS/MS spectra of tryptic peptides were further measured in an autoflex III TOF/TOF in MS/MS mode using the following instrument settings: ion source 1, 6.00 kV; ion source 2, 5.30 kV; lens, 3.00 kV; reflector 1, 27.00 kV; reflector 2, 11.65 kV; lift 1, 19.00 kV; lift 2, 4.20 kV; reflector detector, 1400 V.

Fragment ion spectra from MS and MS/MS were submitted to MASCOT ( for a database search and identification of the corresponding proteins using the following database: IPI human 20 081 114 (74 049 sequences; 31 194 560 residues,


Three micrometer-thick sections were made from 10% formalin-fixed and paraffin-embedded lung cancer tissues, deparaffinized in xylene, rehydrated in a descending ethanol series, and then treated with 3% hydrogen peroxide for 20 min. Antigen was retrieved by autoclaving in 0.01 mol/L citrate buffer (pH 6.0) with 0.1% Tween 20 at 121°C for 10 min. After blocking with 2% normal swine serum (Dako, Glostrup, Denmark) for 10 min, the sections were reacted with 1000-fold diluted mouse anti-human cytokeratin 7 (CK7; OV-TL 12/30; Dako), 250-fold diluted mouse anti-human CK8 (NCL-CK8-TS1; Novocastra, Newcastle, UK), 500-fold diluted mouse anti-human CK18 (NCL-CK18; Novocastra) or 250-fold diluted mouse anti-human CK19 (NCL-CK19; Novocastra) for 16–18 h at room temperature. After rinsing in 0.01 mol/L Tris-HCl pH 7.5, 150 mmol/L NaCl (TBS) three times for 5 min each, the sections were reacted with ChemMate Envision reagent (Dako) for 30 min at room temperature. Finally, the sections were visualized on Stable DAB solution (Invitrogen, Carlsbad, CA, USA) and counterstained with Mayer's hematoxylin.

Evaluation of immunohistochemistry

Immunohistochemistry (IHC) was scored by multiplying the percentage of positive tumor cells and staining intensity. The percentage of positive tumor cells was scored as 0 (0%), 1+ (1–25%), 2+ (26–50%), 3+ (51–75%), or 4+ (76–100%). Staining intensity was also scored as 0 (negative), 1+ (weakly positive), 2+ (moderately positive), or 3+ (strongly positive). The Mann–Whitney U-test and the χ2-test were used for statistical evaluation of IHC data. Statistical significance was considered when P < 0.05.


Detection and identification of proteins differently expressed between LCN1 and N231

More than 2000 spots were separated from the total protein of LCN1 and N231 cells using agarose 2-DE. We excised 25 protein spots with expression levels more than twofold different between LCN1 and N231 from the gel of LCN1 cells, using MALDI-TOF/TOF-MS, and identified them. Identified proteins were functionally classified into 10 enzymes, seven cytoskeletal proteins, three signal transduction factors, two nucleic acid binding proteins, and one each of the transporter and chaperone (Table 1). The expression levels of CK7, 8, 18 and 19 of the LCN1 cells were 4.6-fold, 27-fold, 17-fold and 3.3-fold higher, respectively, than N231 cells, and these four CK were studied further (Fig. 1).

Table 1.  Proteins differentially expressed between LCN1 and N231 cells on 2-DE
No.Protein descriptionMolecular functionMolecular weight (Da)LCN1/N231 ratio
 1TransketolaseEnzyme67 7754.0
 2Pyruvate kinase, M1 isozymeSignal transduction57 9392.1
 3Enolase 1Enzyme47 1500.4
 4UnknownUnknown101 9792.1
 5Aldehyde dehydrogenase 1A1Enzyme54 84312.2
 6Glyceraldehydes-3-phosphate dehydrogenaseEnzyme36 0352.2
 7Aldolase AEnzyme39 2703.5
 8Uridine diphosphoglucose dehydrogenaseEnzyme55 0758.7
 9Protein disulfide isomeraseEnzyme56 7782.7
10Heterogeneous nuclear ribonucleoprotein A2/B1Nucleic acid binding35 9870.5
11Phosphoglycerate kinase 1Signal transduction44 5960.4
126-Phosphogluconate dehydrogenase, decarboxylatingEnzyme53 1220.3
13Heat shock proteinChaperone69 97723.9
14Keratin 18Cytoskeletal structural protein48 01017.0
15Lamin B1Cytoskeletal structural protein66 3482.0
16Keratin 19Cytoskeletal structural protein44 0613.3
17Heterogeneous nuclear ribonucleoprotein KNucleic acid binding50 9582.6
18Neuropolypeptide h3Signal transduction22 9073.3
19Ubiquitin carboxyl-terminal esterase L1Enzyme24 8060.4
20Phosphoglycerate mutase 1Enzyme28 7852.0
21Valosin-containing proteinTransporter89 3030.2
22Ezrin (p81)(Villin 2)Cytoskeletal structural protein69 38010.2
23Villin 1Cytoskeletal structural protein92 67769.8
24Keratin, type II cytoskeletal 8 (Cytokeratin 8)Cytoskeletal structural protein53 65627.0
25Keratin, type II cytoskeletal 7 (Cytokeratin 7)Cytoskeletal structural protein51 3994.6
Figure 1.

Differentially expressed cytokeratins (CK) between LCN1 cells and N231 cells on 2-D electrophoresis (2-DE). (a) LCN1 and (b) N231 cell lysates were separated on 2-DE. Arrows, CK upregulated in LCN1 in comparison with N231 cells.

IHC of cytokeratins

The stainability of CK7, 8, 18 and 19 in N231 and LCN1 cells was generally in agreement with the results of proteome analysis (Fig. 2). In general, positive staining was observed in the cytoplasm of LCN1 cells at various levels, but not in N231 cells. To evaluate the utility of these molecules as diagnostic markers, we also stained pulmonary carcinoma tissues. CK expression was localized in the cytoplasm of carcinoma cells at various levels in each histological type, and normal bronchial cells constantly had high expression levels (Fig. 3). The stainability of normal bronchial epithelium was used as an internal control. The staining scores and positivity of CK are summarized in Table 2. CK7 was detected in 17 of 27 (63.0%) SCLC, 27 of 30 (90.0%) LCNEC, all 14 (100%) AD, and three of 10 (30.0%) SCC, and the mean staining scores of CK7 were 2.8, 6.8, 10.9, and 1.8, respectively. CK8 was detected in 26 of 27 (96.3%) SCLC, 29 of 30 (96.7%) LCNEC, all 14 (100%) AD, and all 10 (100%) SCC, and the mean staining scores of CK8 were 4.0, 7.8, 9.4, and 5.7, respectively. CK18 was detected in 26 of 27 (96.3%) SCLC, 29 of 30 (96.7%) LCNEC, all 14 (100%) AD, and seven of 10 (70.0%) SCC, and the mean staining scores of CK18 were 6.0, 8.2, 10.0, and 3.2, respectively. CK19 was also detected in 26 of 27 (96.3%) SCLC, 29 of 30 (96.7%) LCNEC, all 14 (100%) AD, and all 10 (100%) SCC, and the mean staining scores of CK19 were 4.3, 7.6, 9.4, and 7.5, respectively. Although there was no difference in the positivity, the mean staining scores of CK7, CK8, CK18, and CK19 of LCNEC were significantly higher than that of SCLC (P < 0.001, P < 0.001, P < 0.01, and P < 0.001, respectively). In generally, positivity and mean staining scores of CK in LCNEC were similar to those in AD. Furthermore, cases of more than two or three of four CK with a score of >8 for each staining were found for 18 or 14 of 28 LCNEC (64.3% or 50.0%) and only one or none of 28 SCLC (4.2% or 0%; χ2 test, P= 0.0000072 or P= 0.000052), respectively.

Figure 2.

Expression of cytokeratins (CK) (a,e) 7, (b,f) 8, (c,g) 18 and (d,h) 19 in (ad) LCN1 cells and (eh) N231 cells. Expression of all CK was observed at various levels in LCN1 cells, but not in N231 cells.

Figure 3.

Expression of cytokeratins (CK) 7, 8, 18 and 19 in pulmonary carcinomas. (ad) Normal bronchial epithelia, (eh) large cell neuroendocrine carcinomas (LCNEC), (il) small cell lung carcinomas (SCLC). (a,e,i) CK7; (b,f,j) CK8; (c,g,ki) CK18; (d,h,l) CK19. Expression of all CK was observed in bronchial epithelium. For all CK expression, high mean staining scores were recognized in LCNEC in comparison with SCLC.

Table 2.  Expression of cytokeratin 7, 8, 18 and 19 in pulmonary carcinomas Thumbnail image of


In the post-genome era, proteome research using various techniques is increasing rapidly. Such research has increased because the expression levels of mRNA and proteins in cells or tissues are not always consistent;6 intra-cellular and extra-cellular proteins usually undergo post-translational modifications, such as phosphorylation, oxidization, or the addition of carbohydrate chains,7 and information about these modifications cannot be predicted from the gene sequence. Some studies have used proteome analyses for lung cancer. Li et al. identified 40 proteins for which the expression levels differed between SCC and their non-neoplastic peripheral lung tissues on 2-DE-based analysis.8 The identified proteins included cell-cycle and signal transduction system-related proteins, but the molecular weights of most identified proteins are ≤50 000 Da because 2-D electrophoresis with IPG (immobiline 2-DE method) cannot separate high-molecular-weight proteins efficiently. Chen et al. compared pulmonary AD tissues and non-neoplastic lung tissues of the same patient, and identified upregulated proteins, including antioxidant enzymes in tumor tissues, but the molecular weights of identified proteins were also only up to approximately 50 000 Da.9 It has been reported that, compared with prokaryotes, eukaryotes have many more proteins of ≥100 000 Da that probably contain fused domains of several proteins as a result of evolution, and these proteins have come to have multiple functions.9 Therefore, proteome analysis of humans, as eukaryotes, requires a 2-DE system that can analyze proteins with high molecular weight. Compared with the conventional immobiline 2-DE method, the agarose 2-DE method improved by Oh-ishi et al., which uses agarose as a carrier in the first dimension of isoelectric focusing, can analyze high molecular weight (≥100 000 Da) and basic proteins easily with the advantage of being able to analyze 10-fold as many proteins as the conventional method.3 Using this method, Kuruma et al. identified proteins that are upregulated in androgen-independent prostatic cancer, and some proteins had a molecular weight of ≥100 000 Da.10

In the present study we conducted proteome analysis using cell lines derived from SCLC and LCNEC by the agarose 2-DE method. The total proteins of these two cell lines were separated into approximately 2000 spots on CBB staining, and these electrophoresis patterns on gels were very similar. Twenty-five proteins had different expression levels and 15 (60%) had a molecular weight >50 000 Da. Functional classification showed that the most common were enzymes, followed by cytoskeleton proteins. The present study also confirmed the usefulness of the agarose 2-DE method. Because cellular size greatly differs between SCLC and LCNEC, we thought that the expression levels of cytoskeleton-associated proteins may also differ. We therefore focused on cytoskeletal proteins CK7, 8, 18 and 19, and further studied their expression on immunohistochemistry in clinical cases of pulmonary carcinomas.

All four CK identified in the present study have been reported as having expression primarily in the simple epithelium.11 In contrast, most neuroendocrine tumors, including those of the lung, express CK8, 18 and 19,12 but the differences in stainability of these CK between LCNEC and SCLC have not been extensively evaluated.

In the present study, staining scores of CK7, 8, 18 and 19 were significantly higher in LCNEC than in SCLC, suggesting that LCNEC and SCLC can be differentiated on CK7, 8, 18 and 19 staining. Although high expression levels of CK7 are observed in most AD, both the positivity and stainability of CK7 were usually low in neuroendocrine tumors, including SCLC and carcinoid, and in non-keratinizing-type SCC.11 In the present study the staining score of CK7 was low in SCC and SCLC, in agreement with a previous report, but there are few reports on CK7 expression in LCNEC. Nitadori et al. performed tissue microarray analysis of surgically resected LCNEC and SCLC specimens using 48 antibodies, including CK. They demonstrated that the expression of four proteins, CK7, 8, E-cadherin and β-catenin, were significantly higher in LCNEC than in SCLC.13 Similarly, the present study also confirmed that the positivity and staining score of CK7 were markedly higher in LCNEC, as in AD, than in SCC.

Moreover, the present study demonstrated that LCNEC can be differentiated from SCLC with high probability in cases of more than two of four CK with a score of >8 for each stain.

Lyda and Weiss have reported that most non-SCLC were positive for high-molecular-weight CK, such as 34βE12 and CK7, while most neuroendocrine carcinomas were negative for these CK.14 According to their report, it is possible to differentiate these two types of tumor based on the stainability of CK and neuroendocrine markers, such as chromogranin A and synaptophysin,14 but the difference in CK expression between SCLC and LCNEC was not described in their study. Lyda and Weiss also reported that the antibody designated as B72.3 was useful for differential diagnosis of non-SCLC from SCLC, for which 80% of non-SCLC were positive, while only 5% of SCLC were positive. They also reported that three of six (50%) LCNEC were positive for this marker.14

Giuseppe et al. classified neuroendocrine tumors of the lung based on their proliferative activity determined by the Ki-67 labeling index, expression levels of c-kit, p53, Rb, bcl-2 and cdk4 proteins and so on,15 and suggested that the tumors could be divided approximately into three groups, typical carcinoid (TC), SCLC and LCNEC. These data show that TC and SCLC are genetically different, and AC and LCNEC may belong to another group rather than to the tumor group of intermediate progression from TC to SCLC;15 the present data confirmed that study of SCLC and LCNEC. Moreover, high-grade neuroendocrine tumors are generally considered to be characterized by loss of RB protein expression. Loss of RB protein was observed in most SCLC (91%), while it occurred in only half of LCNEC, suggesting that the mechanisms for G1 checkpoint in SCLC and LCNEC differ.15 Sturm et al. studied the expression of thyroid transcription factor-1 (TTF-1) in neuroendocrine tumors of the lung on IHC, and higher expression levels of TTF-1 were detected in SCLC and LCNEC than in TC and AC. Furthermore, the positivity of TTF-1 was 49% in LCNEC and 85.5% in SCLC, respectively, and the differences were significant.16

Recently, Ullman et al. studied chromosomal abnormalities in SCLC and LCNEC on comparative genomic hybridization and found that both tumors had many common chromosomal abnormalities, but 3q+ was observed in 66% of SCLC, while it occurred in only one of 18 LCNEC.17 Thus, a gene that determines the differences in biological character between these tumors may be located in the 3q domain.

Taken together with previous studies, the present study strongly suggests that the biological characteristics of SCLC and LCNEC may differ, and CK expression was useful for the differential diagnosis of SCLC and LCNEC. Further studies of chromosomal abnormalities, gene alterations and gene expressions will facilitate the differential diagnosis of these tumors and lead to the acquisition of new markers.


This work was supported in part by Grants-in-Aid for Scientific Research C (C19590365) from the Japan Society for the Promotion of Science, and for Research Project (No. 2008-1003) from the School of Allied Health Sciences and 2008 Project Study from the Graduate School of Medical Sciences, Kitasato University.