Alteration of the E-cadherin/β-catenin cell adhesion system is common in pulmonary neuroendocrine tumors and is an independent predictor of lymph node metastasis in atypical carcinoids

Authors


Abstract

BACKGROUND

To the authors' knowledge, little is known regarding the role of E-cadherin/β-catenin system dysregulation in pulmonary neuroendocrine tumors.

METHODS

E-cadherin and β-catenin immunoreactivity was evaluated in 10 hyperplastic neuroendocrine tumorlets and 210 neuroendocrine tumors, including 96 typical carcinoids (CTs), 35 atypical carcinoids (ACTs), 49 large cell neuroendocrine carcinomas (LCNECs), and 30 small cell lung carcinomas (SCLCs).

RESULTS

Normal and hyperplastic bronchial neuroendocrine cells expressed E-cadherin/β-catenin with an orderly distribution along the cell membrane. Neuroendocrine tumors retained β-catenin expression in all tumors and E-cadherin in most tumors, with the exception of 2% of LCNECs, 3% of SCLCs and 9% of ACTs. E-cadherin showed a prevalent membrane-associated, linear immunoreactivity in CTs, whereas membrane-disarrayed and cytoplasmic staining was seen in most ACTs, LCNECs, and SCLCs (P < 0.001). β-Catenin exhibited similar immunoreactivity patterns according to tumor type and a close association with E-cadherin subcellular distribution (P < 0.001). Nuclear accumulation of β-catenin was found only in seven LCNECs and in two SCLCs. In ACTs, disarrayed immunoreactivity for E-cadherin and/or β-catenin was associated with a nontrabecular growth pattern, altered expression of the cell-motility marker fascin, and lymph node metastases. Furthermore, a disarrayed E-cadherin distribution pattern was associated with the pathologic lymph node classification and the number of involved lymph nodes. Multivariate analysis confirmed that a disarrayed E-cadherin or β-catenin pattern was an independent predictor of lymph node metastases in patients with ACT.

CONCLUSIONS

The subcellular compartmentalization of the E-cadherin/β-catenin complex was altered in pulmonary neuroendocrine tumors. This likely affects the tumor growth pattern and cell motility of ACT and was correlated with the occurrence of lymph node metastases. Cancer 2005. © 2005 American Cancer Society.

Neuroendocrine tumors of the lung constitute a heterogeneous group of neoplasms, including low-grade carcinoid tumors (CTs), with long life expectancy; intermediately differentiated, atypical carcinoid tumors (ACTs), with a more aggressive clinical course; and high-grade, large cell neuroendocrine carcinomas (LCNEC) and small cell lung carcinomas (SCLC), both with a dismal prognosis.1 Several molecular prognostic factors have been proposed to date,2–6 but to our knowledge histologic typing remains the most powerful predictor of clinical course.7, 8

E-cadherin is a transmembrane glycoprotein involved in Ca2+-dependent epithelial cell-cell homophylic adhesion at the adherens junctions.9, 10 The cytoplasmic domain of E-cadherin interacts with a group of closely related proteins, termed α-catenin, β-catenin, and γ-catenin, which connect the cell membrane tightly to the cytoskeleton.11, 12 Moreover, β-catenin acts as a link between the E-cadherin and the Wnt signaling pathways, because it is the main transducer of the latter pathway to the nucleus, where it activates transcription of several reporter genes.13, 14

Dysregulation of the E-cadherin/β-catenin-dependent adhesion complex has been associated with the development and progression of many solid tumors,14 including several types of endocrine tumors, in which either down-regulation or an altered subcellular localization of these proteins has been reported.15–25 A single study that dealt specifically with changes of E-cadherin/β-catenin expression in lung neuroendocrine tumors showed a subcellular redistribution of the E-cadherin-catenin complex in the majority of 27 investigated tumors but failed to document any correlation with histologic typing or tumor lymph node metastasis (TNM) classification.26

In the current study, we assessed the prevalence and clinicopathologic implications of E-cadherin/β-catenin adhesion system immunoreactivity in a series of 210 pulmonary neuroendocrine tumors. Our data indicate that changes in E-cadherin/β-catenin expression patterns are common in lung neuroendocrine tumors, with either subcellular redistribution and/or down-regulation, according to tumor differentiation. E-cadherin/β-catenin changes also are independent predictors of lymph node metastasis in patients with ATC.

MATERIALS AND METHODS

Patients

Two hundred ten patients (130 men and 80 women) with surgically excised neuroendocrine tumors of the lung were investigated. The tumors included 96 CTs, 35 ACTs, 49 LCNECs, and 30 SCLCs. Relevant clinicopathologic data for the entire cohort of patients are summarized in Table 1. All patients had been studied by total-body computed tomography scans and underwent radical surgery with extensive mediastinal lymph node dissection (median value, nine excised lymph nodes per patient) to ensure accurate staging. The diagnosis of neuroendocrine tumors was based on morphologic criteria according to the most recent World Health Organization classification1 and on immunophenotypical findings, including reactivity for panendocrine markers (synaptophysin and chromogranin A) and respiratory tract-related hormones (gastrin-related peptide, adrenocorticotropic hormone, the α subunit of human chorionic gonadotropin, calcitonin, and serotonin).1, 27, 28

Table 1. Clinicopathologic Characteristics of 210 Patients with Neuroendocrine Tumors
VariableCT (n = 96)ACT (n = 35)P valueLCNEC (n = 49)SCLC (n = 30)P value
  1. CT: typical carcinoid; ACT: atypical carcinoid; LCNEC: large cell neuroendocrine carcinoma; SCLC: small cell lung carcinoma; NS: not significant; LN: lymph node; GRP: gastrin- releasing peptide; ACTH: corticotropic hormone; α-HCG: α-human chorionic gonadotropin; 5- HT: serotonin.

  2. a The separate comparisons between patients with typical carcinoid and atypical carcinoid and patients with large cell neuroendocrine carcinoma (LCNEC) and small cell lung carcinoma (SCLC) were based on a large body of evidence supporting the view that carcinoids, whether typical or atypical, are not precursors of high-grade tumors and that LCNEC genetically is closer to nonsmall cell lung carcinoma than to SCLC(see Leotela et al.,19 Ullmann et al. (2002),36 and Ullmann et al. (2001).37

Age (yrs)      
 ≤ 493514 42 
 50–59224 98 
 60–69279 2112 
 ≥ 70128NS158NS
Gender      
 Male4915 4412 
 Female4720NS58NS
TNM Stage      
 I7012 116 
 II–III2623< 0.00013824NS
Pathologic tumor classification      
 pT17822 1514 
 pT21813 3015 
 pT3–pT4000.03741NS
Pathologic lymph node classification      
 pN08517 2910 
 pN169 1210 
 pN259< 0.00018100.023
LN metastasis      
 No8517 2910 
 Yes1118< 0.000120200.037
Positive LNs (median no.)010.00202NS
Tumor size      
 < 2 cm316 75 
 2–4 cm4921 1716 
 > 4 cm168NS259NS
Growth pattern      
 Trabecular1049  
 Solid1927  
 Spindle312  
 Other280.04
Vital status      
 Alive9529 2518 
 Dead160.00152412NS
Recurrences      
 Yes9024 2917 
 No6110.00052013NS
Median Ki-67 index2.309.00< 0.00014662.5< 0.0001
Fascin      
 Negative4223 70 
 10–60%36 218 
 > 60%060.000819200.0017
Median GRP50NS29NS
Median calcitonin00NS500.0004
Median ACTH00NS00NS
Median α-HCG2711NS40NS
Median 5-HT56NS00NS
Median microvessel densisty22091<0.00014340NS

Accurate follow-up information was available for 200 patients: Sixty-three patients (31.5%) had recurrent disease, and 43 patients (21.8%) died of disease. The follow-up period was 77 ± 61 months (median, 61.5 months) in patients with CT, 33 ± 35 months (median, 25 months) in patients with ACT, 30 ± 34 months (median, 13 months) in patients with LCNEC, and 27 ± 24 months (median, 18 months) in patients SCLC. No patient presented with symptoms due to hormone hyperproduction.

Immunohistochemistry

Paraffin embedded tumor samples were investigated using the antibodies listed in Table 2, according to staining protocols previously reported.27 Ten samples of normal lung from patients with nonmalignant lung diseases and 10 samples of pulmonary neuroendocrine tumorlets were used as control groups. Tumorlets and normal neuroendocrine cells were highlighted by chromogranin A immunostaining. Tumor angiogenesis was evaluated by microvessel density after CD34 immunostaining of endothelial cells.29 All immunostains were evaluated without knowledge of the patient's identity or disease stage. For each marker, the percentage of immunoreactive tumor cells was assessed scanning at least 2000 neoplastic cells in randomly selected fields. Tumors were considered negative if staining was either absent or was identified in < 5% of tumor cells. The specificity of immunoreaction was assessed substituting the primary antibody with a nonrelated isotypic mouse immunoglobulin or normal serum.

Table 2. Antibody Panel Used
AntibodyAntibody typeCloneSourceDilutionPretreatmenta
  • α-HCG: α-human chorionic gonadotropin; ACTH: corticotropic hormone; MWO: microwave oven; EDTA:.ethylenediamine tetraacetic acid; CB: citrate buffer.

  • a

    MWO-EDTA: microwaved at 750 watts for 12 minutes in ethylenediamine tetraacetic acid [pH 8.0]; MWO-CB: microwaved at 750 watts for 10 minutes in citrate buffer [pH 6.0].

  • b

    Clone 4A2C7 reacts with a highly conserved cytoplasmic C-terminus fragment of E-cadherin.

  • c

    Clone HECD-1 reacts with an extracellular domain binding to E-cadherin exon 8 (see Becker et al.40).

  • d

    Antibody E-5 reacts with the amino acid sequence 680–781 at the cytoplasmic C-terminus of β-catenin.

E-cadherinMonoclonal4A2C7bZymed Laboratories, San Francisco, CA1:400MWO-EDTA
 MonoclonalHECD-1cZymed Laboratories1:400MWO-EDTA
β-cateninMonoclonalE-5dSanta Cruz Biotechnology, Santa Cruz, CA1:100MWO-EDTA
Chromogranin AMonoclonalLK2H10Signet Laboratories, Dedham, MA1:40None
α-HCGPolyclonalSignet Laboratories1:200MWO-CB
SynaptophysinMonoclonalSY 38Dakopatts, Glostrup, Denmark1:20MWO-CB
Gastrin-releasing peptidePolyclonalDakopatts1:400None
CalcitoninPolyclonalDakopatts1:20None
SerotoninMonoclonal5HT-H209Dakopatts1:10None
ACTHMonoclonal02A3Dakopatts1:100None
FascinMonoclonal55k-2Dakopatts1:400None
CD34MonoclonalQBEnd/10Novocastra, Newcastle-upon-Tyne, UK1:400MWO-CB
Ki-67 antigenMonoclonalMIB-1Immunotech, Marseille, France1:400MWO-EDTA

STATISTICAL ANALYSIS

Associations of categoric or ordinal variables were evaluated with the Fisher exact test or the Mantel–Hanszel chi-square test for trend (with 1 degree of freedom), respectively. Continuous variables were contrasted with the Student t test using the Satterthwaite method in cases of unequal variance or the pooled method in cases of equal variance. A logistic regression analysis model served to compare explanatory variables with the occurrence of lymph node metastases. Survival curves were determined according to the Kaplan–Meier method and were compared with log-rank tests. All estimates were performed using SAS statistical software (SAS Institute, Inc., Cary, NC). All P values (significant if ≤ 0.05) were based on two-sided testing.

RESULTS

All Normal Lung Cells Express E-Cadherin and β-Catenin with a Membrane-Linear Pattern

All cells in normal bronchial mucosa and lung tissue from both neoplastic and nonneoplastic specimens showed a membrane-associated, linear pattern of immunoreactivity for E-cadherin and β-catenin, which decorated entire the cell membrane. The scattered bronchial mucosa-associated neuroendocrine cells and all of the neuroendocrine tumorlets also showed E-cadherin and β-catenin membrane-associated, linear immunoreactivity.

Neuroendocrine Tumors Show Variable Staining Patterns for E-Cadherin and β-Catenin

The immunostaining pattern of tumor cells for E-cadherin and β-catenin was defined as either “membrane-linear” or “disarrayed” (Fig. 1) whenever any change in the membrane-associated staining pattern was observed in association with variable degrees of cytoplasmic or nuclear immunoreactivity. In particular, disarrayed E-cadherin staining was defined as either “membrane-disrupted” if a wrinkled staining of the membrane was observed along with variable cytoplasmic accumulation or “cytoplasmic” if a prevalent cytoplasmic staining with only minimal or absent membrane labeling was observed. Disarrayed β-catenin staining was distinguished into three patterns: “membrane-linear/cytoplasmic” if there was preservation of the membrane-linear signal associated with variable degrees of cytoplasmic staining; “membrane-disrupted” if an irregular and wrinkled decoration of the cell membrane was seen along with variable, and usually more prominent, cytoplasmic staining; and “nuclear” if immunoreactivity also was localized in the nucleus, independent of the distribution of any additional staining.

Figure 1.

Representative E-cadherin and β-catenin immunostaining patterns in neuroendocrine tumors of the lung. (A,B) Membrane-linear pattern for both markers is typically seen in well differentiated carcinoids. A progressive, cytoplasmic accumulation with disruption of the membrane-linear pattern for E-cadherin is seen in (C) an atypical carcinoid, (E) a large cell neuroendocrine carcinoma, and (G) a small cell lung carcinoma. In addition, β-catenin is localized increasingly in the cytoplasm with a (D) membrane-cytoplasmic pattern or (F) a membrane-disrupted pattern in less differentiated tumors, whereas (H) nuclear accumulation of β-catenin is present exclusively in high-grade neuroendocrine tumors (all immunostains were performed with diaminobenzidine and counterstained with hematoxylin).

E-Cadherin and β-Catenin Staining Patterns Vary According to Tumor Differentiation

The results of E-cadherin and β-catenin immunostaining are summarized in Table 3. E-cadherin immunoreactivity was seen in 205 of 210 tumors (98%), including all 96 CTs (100%), 32 of 35 ACTs (91%), 48 of 49 LCNECs (98%), and 29 of 30 SCLCs (97%). The vast majority of E-cadherin-expressing tumors displayed > 30% immunoreactive tumor cells, with only 5 CTs, 5 ACTs, 4 LCNECs, and 3 SCLC that had ≤ 30% immunoreactive cells. There was a diverse distribution of E-cadherin staining patterns among the different tumor types, with a definite prevalence of retained membrane-linear decoration in 83% CTs, a prevalence of membrane-disrupted immunoreactivity in 56% LCNECs, and a prevalence of cytoplasmic staining in 67% SCLCs (P < 0.001). ACTs showed a similar prevalence of membrane-linear (46% of ACTs) and membrane-disrupted to cytoplasmic (46% of ACTs) immunostaining patterns.

Table 3. Prevalence of E-Cadherin and β-Catenin Immunoreactivity According to Tumor Types
VariableCT (n = 96)ACT (n = 35)LCNEC (n = 49)SCLC (n = 30)P value
  • CT: typical carcinoid; ACT: atypical carcinoid; LCNEC: large cell neuroendocrine carcinoma; SCLC: small cell lung carcinoma; SD: standard deviation; NS: not significant.

  • a

    E-cadherin: typical carcinoid versus atypical carcinoid, P < 0.001; typical carcinoid versus small cell lung carcinoma, P < 0.001; atypical carcinoid versus large cell neuroendocrine carcinoma, P = 0.002; large cell neuroendocrine carcinoma versus small cell lung carcinoma, P = 0.003.

  • b

    β-catenin: typical carcinoid versus atypical carcinoid, P = 0.003; typical carcinoid versus small cell lung carcinoma, P < 0.001; atypical carcinoid versus large cell neuroendocrine carcinoma, P = 0.070; large cell neuroendocrine carcinoma versus small cell lung carcinoma, P = 0.021.

E-cadherin     
Labeling index     
 Mean ± SD82.9 ± 21.971.3 ± 33.174.9 ± 26.673.3 ± 29.3 
 Median90858585 
 Range10–1000–1000–000–100NS
Pattern: No. (%)a     
 Membrane-linear80 (83)16 (46) 7 (14) 0 (0) 
 Membrane-disrupted16 (17)14 (40)28 (56) 9 (30) 
 Cytoplasmic 0 (0) 2 (6)13 (26)20 (67) 
 Negative 0 (0) 3 (8) 1 (4) 1 (3)< 0.0001
β-catenin     
Labeling index     
 Mean ± SD95.6 ± 11.593.4 ± 14.488.9 ± 19.696.7 ± 10.9 
 Median100100100100 
 Range30–10030–10020–10050–100NS
Pattern: No. (%)b     
 Membrane-linear46 (48)13 (37) 9 (18) 1 (3) 
 Membrane-cytoplasmic39 (41) 9 (26)10 (21) 2 (7) 
 Membrane-disrupted11 (11)13 (37)30 (61)27 (90)< 0.001

β-Catenin was expressed in all tumors, and the vast majority showed > 50% immunoreactive cells, with only 1 CT, 1 ACT, and 1 LCNEC that had ≤ 30% immunoreactive cells. For β-catenin, a different staining pattern also was observed, with a strong prevalence of the membrane-linear and linear-cytoplasmic pattern in more differentiated tumor types (CTs, 89%; ACTs, 63%; LCNECs, 39%; SCLCs, 10%; CT vs. others, P < 0.001). Conversely, most SCLCs (90%) and LCNECs (61%) accumulated β-catenin in the cytoplasm along with a wrinkled membrane profile compared with either CTs (11%) or ACTs (37%).

All the differences in staining patterns observed among the diverse neuroendocrine histotypes were statistically significant for both markers and also after adjustment for confounding factors such as patient age, gender, and tumor size, with the exception of a marginal difference between ACTs and LCNECs for β-catenin (Table 3). It is interesting to note that the quantitative ratio between membrane-linear and disarrayed immunoreactivity patterns for both E-cadherin and β-catenin also was distributed differentially among the tumor types (Fig. 2).

Figure 2.

The ratio between the number of membrane-linear and disarrayed-immunoreactive tumors according to histologic typing. The membrane-linear pattern was most prevalent by far in typical carcinoids (CT), and the nonmembrane pattern was most prevalent in large cell neuroendocrine carcinomas (LCNEC) and small cell lung carcinomas (SCLC), with atypical carcinoids (AC) occupying an intermediate position with both markers. An asterisk indicates that no case of SCLC presented with any membrane labeling (ratio = 0).

Correlation between E-Cadherin and β-Catenin Staining Patterns

With the exception of SCLCs, all changes in the staining pattern for E-cadherin and β-catenin were correlated closely in all tumor types (and, in LCNECs, also for the number of immunoreactive cells), with the cytoplasmic accumulation of β-catenin associated with membrane-disarrayed E-cadherin decoration (Table 4). Moreover, in CTs and ACTs, a reduced labeling index (≤ 70%) for either E-cadherin or β-catenin inversely paralleled disarrayed membrane immunostaining with progressive cytoplasmic accumulation (P < 0.001 and P = 0.003, respectively).

Table 4. Associations between E-Cadherin and β-Catenin Staining Patterns
Tumor type and E-cadherin patternNo. of patientsβ-catenin pattern: No. (%)P value
Membrane-linearMembrane-cytoplasmicMembrane-disrupted
  1. CT: typical carcinoid; ACT: atypical carcinoid; LCNEC: large cell neuroendocrine carcinoma; SCLC: small cell lung carcinoma; SD: standard deviation; NS: not significant.

CT     
 Membrane-linear8044 (55)35 (44) 1 (1) 
 Membrane-disrupted16 2 (13) 4 (25)10 (62)< 0.001
ACT     
 Membrane-linear1612 (75) 4 (25) 0 (0) 
 Membrane-disrupted14 0 (0) 3 (21)11 (79) 
 Cytoplasmic2 0 (0) 1 (50) 1 (50)< 0.001
 Negative3 
LCNEC     
 Membrane-linear7 5 (71) 2 (29) 0 (0) 
 Membrane-disrupted28 4 (14) 7 (25)17 (61) 
 Cytoplasmic13 0 (0) 0 (0)13 (100)< 0.001
 Negative1 
SCLC     
 Membrane-disrupted9 1 (11) 1 (11) 7 (78) 
 Cytoplasmic20 0 (0) 1 (5)19 (95)NS
 Negative1 

β-Catenin Nuclear Accumulation Is an Exclusive Feature of a Subset of High-Grade Neuroendocrine Tumors

Nuclear accumulation of β-catenin was observed in only seven LCNECs and two SCLCs and always was associated with either a membrane-linear/cytoplasmic pattern (four LCNECs and two SCLCs) or a membrane-disrupted pattern (three LCNEC). Moreover, nuclear immunoreactivity for β-catenin was associated with misplaced E-cadherin decoration (six LCNECs with a membrane-disrupted pattern and two SCLCs with membrane-disrupted and cytoplasmic patterns) or E-cadherin absence (one LCNEC).

Relations between E-Cadherin/ß-Catenin Stainings and Clinicopathologic Correlates

In ACTs, a labeling index < 70% or membrane-disrupted decoration for E-cadherin was associated with a nontrabecular growth pattern (solid or spindle; P = 0.039 and P = 0.015, respectively). Moreover, membrane-disrupted or cytoplasmic E-cadherin/β-catenin patterns were associated significantly with fascin overexpression in ACTs (P = 0.017 and P < 0.001, respectively) and SCLCs (P = 0.005 and P = 0.020, respectively). No other correlations were found between E-cadherin or β-catenin expression and the other clinicopathologic variables, including patient age, gender, tumor size, TNM stage, vital status, recurrences, microvessel density, proliferative activity, and immunoreactivity for panendocrine markers (chromogranin A and synaptophysin) or respiratory tract-related hormones.

E-Cadherin or β-Catenin Immunoreactivity Is an Independent Predictor of Lymph Node Metastasis in ACTs

Lymph node metastases were significantly more common in high-grade tumors than in low-grade to intermediate-grade tumors (51% vs. 22%; P < 0.001) in both males (P = 0.003) and females (P < 0.001). Patients with ACT and SCLC had more lymph node metastases (51% and 67%, respectively) than patients with CTs or LCNECs (11% and 41%, respectively; P < 0.001) independent of gender and age.

Univariate regression analysis showed that patients who had ACT with an E-cadherin labeling index < 70%, a membrane-disrupted or cytoplasmic E-cadherin/β-catenin staining pattern, a younger age, and, marginally, fascin overexpression were associated with regional lymph node metastases. In particular, the median labeling index for E-cadherin steadily decreased according to the pathologic lymph node (pN) status, ranging from 95% in pN0 tumors to ≤ 65% in pN1–pN2 tumors (P = 0.003) and decreased marginally according to the number of involved regional lymph nodes (correlation coefficient [r] = − 0.320, P = 0.057). Patients who had tumors that exhibited a membrane-linear pattern of E-cadherin immunoreactivity had fewer metastatic lymph nodes (median, 0 lymph nodes) compared with patients who had tumors that lacked E-cadherin immunoreactivity (median, 2.0 lymph nodes) and with patients who had tumors with a membrane-disrupted immunostaining pattern (median, 2.5 lymph nodes) or a cytoplasmic immunostaining pattern (median. 4.0 lymph nodes; P < 0.001).

In multivariate analysis, a disarrayed immunostaining pattern for either E-cadherin or β-catenin (odds ratio [OR], 7.77; 95% confidence interval [95% CI], 1.39–43.60 [P = 0.020]) and younger patient age (OR, 0.95; 95% CI, 0.90–1.00 [P = 0.042]) emerged as independent risk indicators for lymph node metastases only in patients with ACT (Table 5). Neither E-cadherin immunoreactivity nor β-catenin immunoreactivity was found to have any relation with overall and disease-free survival, irrespective of the tumor type.

Table 5. Tumor Characteristics and Occurrence of Regional Lymph Node Metastasis in 35 Atypical Carcinoids
V VariableUnivariate analysis:Multivariate analysis:
OR (95% CI)P valueOR (95% CI)P value
  • OR: odds ratio; 95% CI: 95% confidence interval.

  • a

    Median value.

E-cadherin labeling index (baseline: ≥ 70%)7.33 (1.53–35.1)0.0127  
E-cadherin pattern (baseline: membrane-linear)6.50 (1.47–28.8)0.0137  
β-catenin pattern (baseline: membrane-linear)7.14 (1.48–34.4)0.0142  
E-cadherin or β-catenin pattern (baseline: membrane-linear)7.14 (1.48–34.4)0.01427.77 (1.39–43.6)0.0198
Patient age (baseline: ≥50 yrsa)0.95 (0.90–1.00)0.03320.95 (0.90–1.00)0.0422
Fascin immunoreactivity (baseline: negative)4.67 (0.99–22.0)0.0517  

Differential Distribution of Immunoreactivity for E-Cadherin Domains in ACT

The clinicopathologic implications of E-cadherin dysregulation in patients with ACT prompted us to investigate these tumors further for E-cadherin integrity. Therefore, in addition to clone 4A2C7, which recognizes the cytoplasmic C-terminus of the molecule, we used a second monoclonal antibody, clone HECD-1, which recognizes an extracellular domain of the molecule (Table 2).

There was a close correlation noted between the staining patterns of the two antibodies (P < 0.001) and between their labeling indexes (r = 0.451; P = 0.006). Membrane β-catenin was associated predominantly with membrane E-cadherin for either the intracellular domain (12 of 16 tumors) or the extracellular domain (11 of 16 tumors), whereas predominantly cytoplasmic β-catenin was associated with misplaced or absent E-cadherin, as assessed with both antibodies (18 of 22 tumors [P < 0.0001] and 17 of 22 tumors [P = 0.006], respectively).

Eight ACTs showed a discordant staining with the two antibodies. Of these, seven ACTs lacked the extracellular domain, and one lacked the intracytoplasmic portion of E-cadherin. In particular, five of the seven tumors that were negative for the extracellular E-cadherin domain were immunoreactive for intracellular E-cadherin. This latter showed a membrane-linear pattern in one tumor, a membrane-disrupted pattern in three tumors, and a cytoplasmic pattern in the one tumor. The patients with these five tumors presented with regional lymph node metastases and Stage > I disease. Only one ACT lacked intracellular E-cadherin and showed a membrane-linear staining pattern for the extracellular domain.

DISCUSSION

The results of the current study may be summarized as follows: 1) normal and hyperplastic (tumorlets), bronchial-associated neuroendocrine cells consistently express E-cadherin and β-catenin, which are distributed in an orderly manner along the cell membrane; 2) neuroendocrine tumors retain β-catenin expression in all cases and E-cadherin expression in the vast majority of cases, and the loss of the latter occurs in only 2% of LCNECs, 3% of SCLCs, and 9% of ACTs; 3) the subcellular compartmentalization of the E-cadherin/β-catenin cell adhesion system largely is preserved in well differentiated tumors (CTs), whereas it is impaired in most of the less differentiated tumor types; 4) β-Catenin nuclear accumulation is an exclusive feature of a subset of high-grade neuroendocrine tumors, with a higher prevalence in LCNECs than in SCLCs; 5) in ACTs, impairment of the E-cadherin/β-catenin complex likely participates in changes of tumor growth pattern and of cell motility and invasiveness, because it is associated with fascin expression and lymph node metastasis.

The E-cadherin/β-catenin complex is distributed in an orderly manner on the cell membrane of all normal and hyperplastic neuroendocrine cells of the lower respiratory tract, as reflected by their membrane-linear immunostaining pattern. The expression of these molecules appears to be conserved highly in pulmonary neuroendocrine tumors, because their labeling indices do not change significantly in the different tumor types. However, what is altered profoundly in these tumors is their subcellular compartmentalization,26, 27 as argued by the differential distribution of the membrane-linear/disarrayed immunostaining pattern ratio among the diverse tumor types.

The membrane-linear pattern is retained for both markers in the majority of well differentiated CTs, whereas a disarrayement of the membrane staining is observed in most ACTs and poorly differentiated tumors. Such a disrupted staining pattern includes either irregular membrane immunoreactivity along with variable cytoplasmic reinforcement or loss of membrane signal with exclusive cytoplasmic accumulation of the molecules.

The current study data also suggest a coordinated expression of the two molecules in well differentiated tumors (CT) that is impaired in most of the less differentiated tumors and is lacking completely in high-grade tumors. This interpretation is based on the observation that the labeling indices for E-cadherin and β-catenin are similar in CTs and that their down-regulation is associated with a disarrayed decoration pattern with subcellular colocalization of the two molecules within individual tumor cells. Conversely, most of the other neuroendocrine tumor types lack any relation between the labeling indices and subcellular localization of the two markers, thus supporting the existence of a more complex dysregulation mechanism in these tumors.

The close correlation between the E-cadherin and β-catenin immunoreactivity patterns in CTs, ACTs, and LCNECs suggests that most β-catenin molecules are connected tightly to E-cadherin in these tumors, not only when they are localized on the cell membrane but also when they are displaced into the cytoplasm. Indeed, only a minority of lung neuroendocrine tumors showed a nuclear translocation of β-catenin, whereas the more common feature was its cytoplasmic colocalization with E-cadherin. This phenomenon may be explained by the simultaneous cytoplasmic misplacement of both proteins, with most of β-catenin associated with E-cadherin.13, 30 This colocalization is lost only when β-catenin is translocated into the nucleus. Conversely, E-cadherin has never been found in the nuclei of pulmonary neuroendocrine tumors either in the current report or in previous series.26

The high prevalence of dysregulated E-cadherin/β-catenin expression emphasizes a possible role for impaired E-cadherin/β-catenin complex-dependent cell-cell contacts and intracellular signaling in the development and progression of lung neuroendocrine tumors. A causative role for E-cadherin/β-catenin dysregulation also has been invoked for the dedifferentation of other lung carcinoma types, such as nonsmall cell carcinomas (for review, see Bremnes et al.31).

There are significant differences in the E-cadherin/β-catenin staining patterns between LCNEC and SCLC, with a higher prevalence of membrane-associated immunoreactivity in the former and of cytoplasmic immunoreactivity in the latter. Moreover, we found that β-catenin nuclear expression was more prevalent in LCNEC than in SCLC. The findings of the current study further support the hypothesis that LCNECs are more akin to nonsmall cell lung carcinomas than to SCLC.32–35 This is not surprising in view of the striking genotypic differences between LCNEC and SCLC,36–39 especially when referring to specific chromosome aberrations of neuroendocrine tumors, namely the prevalence of 6p gains in LCNEC and the loss of 10 and 16q (the chromosome that harbors the E-cadherin gene)40 and 3q gains in SCLC.37 It also is interesting to note that the significantly different distribution of E-cadherin/β-catenin immunostaining patterns in the different neuroendocrine tumor types strongly challenges the concept of a common pathogenesis for all lung neuroendocrine proliferations, confirming the lack of definite correlations between well differentiated and poorly differentiated tumors.19, 36, 37

In the current study, we provide evidence that the E-cadherin/β-catenin complex impairment in ACT is likely to affect tumor growth pattern, cell motility, and invasiveness. In fact, altered expression of these markers is associated with a nontrabecular growth pattern, with the overexpression of fascin (a tumor cell-motility marker27), and with the occurrence of regional lymph node metastases. Multivariate analysis confirmed that disarrayed E-cadherin or β-catenin staining patterns are independent predictors of metastases in patients with ACT. These findings may help to identify patients with metastasizing ACT who have different prognoses for possible multimodality treatment approaches.41 It has been reported that down-regulation or disorganized expression of the E-cadherin/β-catenin complex is correlated with dedifferentiation, lymph node involvement, and a poor prognosis in a variety of tumors, including gastric carcinoma,42 breast carcinoma,43 colorectal carcinoma,11 head and neck carcinoma,44 and hepatocellular carcinoma,45 although its independent role in the survival of patients with either nonsmall cell carcinomas or neuroendocrine tumors of the lung continues to be debated.26, 35, 46–55 In the current study, the lack of any correlation between E-cadherin/β-catenin dysregulation and patient survival observed in either ACTs or other types of pulmonary neuroendocrine tumors depended on the fact that these alterations may be early events in neuroendocrine tumorigenesis and are relevant to the first step of metastatization rather than accounting for the ultimate fate of the patients, even though a bias due to the short follow-up in some tumors (ACTs) or the low number of events in others (CTs) cannot be excluded completely.

The use of two different antibodies to E-cadherin, recognizing either the cytoplasmic C-terminus (clone 4A2C7) or the extracellular domain (clone HECD-1),40 also supports the existence of differential alterations in the functional domains of the molecule in pulmonary ACTs.40 In fact, seven tumors lacked the extracellular domain and one tumor lacked the intracytoplasmic portion of E-cadherin; the former for the most part demonstrated regional lymph node metastases and more advanced tumor stage. This confirms that the truncation of the extracellular region may abrogate cell-cell contact and favor tumor progression,56 whereas the lack of immunostaining for the cytoplasmic domain in one tumor from the current series may suggest the existence of a different type of E-cadherin gene or splice site mutation.40 β-Catenin colocalization with either fragment of E-cadherin in our series of patients with ACT supports the view that both molecules closely parallel in ACT tumorigenesis. Targeting this pathway may be a novel therapeutic option in ACT, for which the standard treatment continues to be surgery alone.

Ancillary