Main

Based on the absence or presence of cellular neuroendocrine differentiation, lung cancer can be grouped into non-neuroendocrine or neuroendocrine tumors. The former is also called non-small-cell lung cancer and largely comprises squamous cell carcinoma, adenocarcinoma and large-cell carcinoma. The latter is also a broad entity ranging from low-grade typical carcinoid, intermediate-grade atypical carcinoid to the most lethal large-cell neuroendocrine carcinoma and small-cell carcinoma.1, 2, 3 The neuroendocrine phenotypes may contribute to the virulence of large-cell neuroendocrine carcinoma and small-cell carcinoma via autocrine loops to promote cell growth and protect tumor cells from apoptosis.4, 5 However, the molecular mechanisms involved in the determination, regulation and maintenance of the neuroendocrine phenotypes of lung cancer remain to be further elucidated.

Basic helix–loop–helix (bHLH) proteins form a large family of transcriptional regulatory factors, which are essential in cell fate determination and tissue-specific phenotypic differentiation.6, 7, 8 Cells of the diffuse neuroendocrine system and neuronal cells share many common biological and morphological features, and they may utilize similar transcriptional mechanisms to determine cell fate and phenotypic differentiation. In Drosophila, the genes of the achaete–scute complex are neurogenic determinants and play a critical role in neuronal commitment and differentiation.9, 10 Mammalian and human achaete–scute homologs 1, MASH1 and hASH1, were cloned from a rat sympathetic neuronal precursor cell line and a human medullary thyroid carcinoma cell line, respectively.11, 12 Subsequent studies have documented that MASH1, like its Drosophila homolog, is specifically and transiently expressed in embryonic neuronal precursors and is instrumental in the development of a subset of neurons.8, 13 MASH1 is also indispensable for the development of pulmonary neuroendocrine cells, thyroid C cells and adrenal chromaffin cells in fetal mice, and homozygous MASH1 null mice lack pulmonary neuroendocrine cells, while mice deficient in HES1, a specific repressor gene of hASH1, show a marked increase in pulmonary neuroendocrine cells.14, 15, 16, 17 In vitro studies also documented restrictive hASH1 expression in tumor cell lines with neuroendocrine or neuronal differentiation such as pulmonary small-cell carcinoma, medullary thyroid carcinoma and neuroblastoma,12, 14, 18, 19, 20, 21, 22, 23 suggesting a key role for hASH1 in dictating cellular neuroendocrine differentiation in lung cancer. However, few data about in vivo hASH1 expression in lung cancer, as well as in other neuroendocrine tumors, are currently available.

Materials and methods

Tissue Samples

Formalin-fixed, paraffin-embedded tissue samples of 238 surgically resected primary lung cancers from the Kitasato University Hospital, Sagamihara; the National Cancer Center Hospital, Tokyo, and the National Cancer Center East Hospital, Kashiwa, Japan, were used. The cases comprised 50 non-small-cell lung cancers and 188 neuroendocrine tumors. The former included 21 squamous cell carcinomas, 20 adenocarcinomas and nine large-cell carcinomas, and the latter 30 typical carcinoids, 13 atypical carcinoids, 67 large-cell neuroendocrine carcinomas and 78 small-cell carcinomas. The diagnoses were based on the criteria of the 1999 WHO Histological Typing of Lung and Pleural Tumors.24

Nonradioactive mRNA Detection by In Situ Hybridization

Total RNAs were extracted from LCN1, a large-cell neuroendocrine carcinoma cell line established in our laboratory. A 499-bp hASH1 fragment corresponding to base pairs 1016–1514 of the hASH1 cDNA sequence12 was subcloned into pCR™II (Invitrogen, CA, USA). The digoxigenin-labelled sense and antisense hASH1 cRNA probes were generated by T7 or SP6 RNA polymerase (Roche Mol. Biochem., Mannheim, Germany), respectively. A modified highly sensitive RNA in situ hybridization method with catalyzed signal amplification based on biotinylated tyramines was employed, which allows messengers to be detected readily with routine formalin-fixed and paraffin-embedded samples.25 In brief, after hybridization with sense or antisense cRNA hASH1 probe at 50°C for 2½h, the sections were treated with ribonuclease and washed under highly stringent conditions followed by endogenous biotin blocking with a Biotin Blocking System (DAKO, CA, USA). Then, the sections were sequentially incubated with horseradish peroxidase conjugated anti-digoxigenin antibody (DAKO, Denmark), biotinylated tyramide and horseradish peroxidase-labelled streptavidin of the GenPoint kit (DAKO, CA, USA). Hybridization results were visualized with 3,3′-diaminobenzidine and the slides were counterstained with hematoxylin.

A nude mouse xenograft of the LCN1 in which constitutive hASH1 expression had been confirmed by RT-PCR was used as a positive sample control. Hybridization of a 340-bp antisense beta-2-microglobulin cRNA probe, when necessary, was also used as an internal control to confirm the quality of the messengers. The negative controls included usage of a sense probe and treatment of the anti-sense probe or sections with ribonuclease. Furthermore, a squamous cell carcinoma without detectable hASH1 expression by RT-PCR was also used as a negative sample control.

The results were scored as negative, or focally, patchily or diffusely positive when definite hybridization signals were detected in none, less than 10, 10–50% or more than 50% of the tumor cells in individual tumors, respectively. The positive hybridization signals were also scored as mild, moderate or intense, according to intensity.

Immunohistochemical Staining

Immunohistochemical staining of general neuroendocrine markers, chromogranin A (1000 × dilution, Dako, Denmark) and neural cell adhesion molecule (200 × dilution, Nippon Kayaku, Tokyo, Japan), and peptide hormones, gastrin-releasing peptide (2000 × dilution, Dako, Japan) and calcitonin (200 × dilution, Incstar, Stillwater, MN, USA), were performed with the standard labelled streptavidin–biotin method, as described previously.3, 26

Statistical Analysis

Correlation of categorical variables between different groups was assessed using Fisher's exact test, Mann–Whitney U-test or the χ2-test. Comparison of survival rates was performed using Kaplan–Meier survival rate curves analyzed with the log-rank test. A probability value (P) less than or equal to 0.05 was considered significant. All analyses were performed with StatView statistical software, version 5.0.

Results

hASH1 Expression in Non-small-cell Lung Cancer

No hASH1 hybridization signals could be detected in all of the 21 squamous cell carcinomas and nine large-cell carcinomas. Two (10%) of the 20 adenocarcinomas expressed hASH1 and both showed patchy and moderate hybridization signals.

hASH1 Expression in Carcinoid Tumors

Of the 30 typical carcinoids, only four (13.3%). cases showed hASH1 expression, with a diffuse pattern in two and a focal pattern in another two cases (Figure 1a). One was intense, while the other three revealed weak hybridization signals. On the other hand, 11 (84.6%) of the 13 atypical carcinoids were hASH1+ and each of them demonstrated hybridization signals in a diffuse pattern (Figure 1b). The hybridization signals were intense in three, moderate in seven and weak in one case, respectively. There was a significant difference in hASH1 expression incidence between typical carcinoids and atypical carcinoids (Fisher's exact test, P<0.0001).

Figure 1
figure 1

Representative hASH1 mRNA in situ hybridization results in pulmonary neuroendocrine tumors. No hASH1 hybridization signals can be detected in typical carcinoid (a), while atypical carcinoid (b), large-cell neuroendocrine carcinoma (c) and small-cell carcinoma (d) are strongly and diffusely positive for hASH1.

Full size image

hASH1 Expression in Large-cell Neuroendocrine Carcinoma and Small-cell Carcinoma

Of the 67 large-cell neuroendocrine carcinomas, 38 (56.7%) cases showed readily detectable hASH1 hybridization signals. A total of 30 cases showed diffuse staining, while only six and two cases showed patchy and focal staining, respectively. The hybridization signals were intense in 16, moderate in 13 and weak in nine cases (Figure 1c).

hASH1 hybridization signals were observed in 56 (71.8%) of the 78 small-cell carcinomas; diffuse in 49, patchy in five and focal in two cases. The hybridization signals were intense in 21, moderate in 23 and weak in 12 cases, respectively (Figure 1d).

There was no statistical difference in hASH1 expression incidence between large-cell neuroendocrine carcinoma and small-cell carcinoma, although hASH1 mRNA was detected in a higher percentage of the latter (Fisher's exact test, P=0.0807). The same was true for the distribution range and intensity of hASH1 hybridization signals between large-cell neuroendocrine carcinoma and small-cell carcinoma (Mann–Whitney U-test, P=0.0549 and 0.1417, respectively).

Of the 67 large-cell neuroendocrine carcinomas and 78 small-cell carcinomas, eight and 11 cases were combined tumors,24 respectively, in which an element of squamous cell carcinoma or adenocarcinoma coexisted. Of these 19 combined carcinomas, 10 expressed hASH1, and the hybridization signals were strictly restricted to the large-cell neuroendocrine carcinoma and small-cell carcinoma regions, while the squamous cell carcinoma and adenocarcinoma elements were completely negative for hASH1 (Figure 2).

Figure 2
figure 2

(a) A combined large-cell neuroendocrine carcinoma (left) and adenocarcinoma (right). (b) Only the large-cell neuroendocrine carcinoma element is positive for chromogranin A (left). (c) On a consecutive section, the hASH1 hybridization signals are also confined to the large-cell neuroendocrine carcinoma element.

Full size image

The LCN1 xenograft, the positive sample control, demonstrated intense and diffuse hASH1 hybridization signals. No signals could be observed in the squamous cell carcinoma, the negative sample control, or when a sense probe or ribonuclease pretreated antisense probe or sections were used. All the hASH1− tumors showed beta-2-microglo-bulin hybridization signals, indicating that the lack of hASH1 signals in these specimens was not due to degraded tumor mRNA.

Immunohistochemical Staining Results

As mentioned above, two adenocarcinomas showed patchily moderate hASH1 hybridization signals. Also, in the same areas of these two tumors, positive staining for chromogranin A and calcitonin was observed on consecutive sections, and one of them was also positive for gastrin-releasing peptide. The remaining 18 hASH1− adenocarcinomas, and all the 21 squamous cell carcinomas and nine large-cell carcinomas, were completely negative for the general neuroendocrine markers or peptide hormones.

Of the 43 typical carcinoids and atypical carcinoids, all and 38 cases were positive for chromogranin A and neural cell adhesion molecule, respectively, and 32 and 17 cases were positive for gastrin-releasing peptide and calcitonin, respectively, with all 17 calcitonin+ cases being also positive for gastrin-releasing peptide.

Of the 145 large-cell neuroendocrine carcinomas and small-cell carcinomas, 88 and 133 were positive for chromogranin A and neural cell adhesion molecule, respectively, with all cases expressing at least one marker and 80 cases being positive for both the markers, confirming their neuroendocrine differentiation. In all, 95 and 51 tumors also expressed gastrin-releasing peptide and calcitonin, respectively, with 45 cases expressing both the peptide hormones. When compared with hASH1 expression status, an obvious tendency for hASH1 expression in those tumors expressing chromogranin A was found. 79/94 hASH1+ large-cell neuroendocrine carcinomas and small-cell carcinomas were positive, while the majority (42/51) of hASH1− tumors were negative for chromogranin A; and there was a close positive correlation between hASH1 and chromogranin A expressions (χ2-test, P<0.0001, r=0.649). On the other hand, hASH1 expression showed no relationship to neural cell adhesion molecule expression status (Table 1, χ2-test, P=0.8892).

Table 1 Irrelative hASH1 and neural cell adhesion molecule expression in 145 large-cell neuroendocrine carcinomas and small-cell carcinomas
Full size table

Chromogranin A is a constituent of secretory granules of normal as well as neoplastic neuroendocrine cells, and its expression is closely associated with the cellular endocrine phenotype defined as peptide hormone/biogenic amine production, storage and secretion. Thus, the correlation between hASH1 expression and gastrin-releasing peptide or calcitonin expression was also analyzed. In all, 86 and 49 of the 94 hASH1+ large-cell neuroendocrine carcinomas and small-cell carcinomas were positive for gastrin-releasing peptide and calcitonin, respectively, while 42 and 49 of the 51 hASH1− tumors were negative for gastrin-releasing peptide and calcitonin, respectively. Thus, a positive correlation between hASH1 expression and gastrin-releasing peptide or calcitonin expression was also found (χ2-test, P<0.0001, r=0.742 and r=0.482, respectively). Furthermore, when the immunostaining results were taken together, as shown in Table 2, all the 94 hASH1+ large-cell neuroendocrine carcinomas and small-cell carcinomas expressed chromogranin A and/or gastrin-releasing peptide and/or calcitonin, while the vast majority of hASH1− tumors were negative for these neuroendocrine markers; and there was a very close positive correlation between hASH1 expression and chromogranin A/gastrin-releasing peptide/calcitonin expression (χ2-test, P<0.0001, r=0.852).

Table 2 hASH1 and chromogranin A/gastrin-releasing peptide/calcitonin (CGA/GRP/CT) expression in 145 large-cell neuroendocrine carcinomas and small-cell carcinomas
Full size table

Clinical Significance of hASH1 Expression in Pulmonary Neuroendocrine Tumors

Complete follow-up information was available for all cases except for one atypical carcinoid and four small-cell carcinoma patients. The Kaplan–Meier survival curves of the patients are shown in Figure 3a. There was a significant difference in overall survival for typical carcinoid patients vs atypical carcinoid patients and for atypical carcinoid patients vs large-cell neuroendocrine carcinoma patients (log-rank test, P=0.0259 and 0.0183, respectively), but not for large-cell neuroendocrine carcinoma patients vs small-cell carcinoma patients (P=0.4692). One of the four hASH1+ typical carcinoid patients died of the disease, while all 26 hASH1− typical carcinoid patients survived. Two of the 10 hASH1+ and one of the two hASH1− atypical carcinoid patients succumbed to the disease, respectively. The small numbers of hASH1+ typical carcinoid and hASH1− atypical carcinoid cases impeded statistical analysis for survival difference based on hASH1 expression status. The 2- and 5-year overall survival rates for the 38 hASH1+ and the 29 hASH1− large-cell neuroendocrine carcinoma patients were 51.0 and 31.5%, and 51.9 and 42.1% respectively. Although hASH1+ large cell neuroendocrine cases had a shorter 5-year survival than hASH1− patients, the difference was not significant (Figure 3b, log-rank test, P=0.4702). However, hASH1 expression was associated with a worse prognosis in small-cell carcinoma patients. As shown in Figure 3c, the 2- and 5-year survival for patients with hASH1+ and hASH1− small-cell carcinoma are 39.0 and 14.6%, and 55.9 and 40.0%, respectively, and there was a significant difference in prognosis (log-rank test, P=0.041).

Figure 3
figure 3

(a) Kaplan–Meier survival curves and log-rank test for patients with typical carcinoid (TC), atypical carcinoid (AC), large-cell neuroendocrine carcinoma (LCNEC) and small-cell carcinoma (SCLC). There was a significant difference in overall survival for typical carcinoid vs atypical carcinoid and for atypical carcinoid vs large-cell neuroendocrine carcinoma (P=0.0259 and P<0.0183, respectively), but not for large-cell neuroendocrine carcinoma vs small-cell carcinoma (P=0.4692). (b) Kaplan–Meier survival curves and log-rank test for large-cell neuroendocrine carcinoma patients by hASH1 expression status (P=0.4702). (c) Kaplan–Meier survival curves and log-rank test for small-cell carcinoma patients by hASH1 expression status (P=0.041).

Full size image

Discussion

It is important to define, at a molecular level, the mechanisms involved in regulating neuroendocrine differentiation of lung cancers because this may contribute to the virulence of small-cell carcinoma and large-cell neuroendocrine carcinoma.4, 5 In the present study, hASH1 expression was detected in 4/30 typical carcinoids, 11/13 atypical carcinoids, 38/67 large-cell neuroendocrine carcinomas and 56/78 small-cell carcinomas. Of the 50 non-small-cell lung cancers, only two adenocarcinomas expressed hASH1 in a patchy pattern. However, both of these two hASH1+ adenocarcinomas also expressed neuroendocrine markers. Thus, our data confirmed and extended the findings of previous in vitro studies that hASH1 was specifically expressed in lung cancers with neuroendocrine differentiation,12, 14, 18, 19 indicating that besides being instrumental in development of a subset of neurons and neuroendocrine cells,8, 13, 14, 15, 16, 17 hASH1 also plays a key role in regulating neuroendocrine differentiation of tumor cells.

The term ‘neuroendocrine differentiation’ in a broad sense is used to define cells expressing a neural and/or an endocrine phenotype. The markers of the former include cytosolic and cell membrane markers common to neurons and endocrine cells such as neuron-specific enolase, neuron-specific class III beta-tubulin and neural cell adhesion molecule. On the other hand, an endocrine phenotype should be defined as the capacity of a given cell to produce and secrete peptide hormone(s)/biogenic amine(s). Chromogranins, a group of acidic monomeric proteins, are regarded as markers of the endocrine phenotype, as they are constituents of the matrix of secretory granules where peptides/amines are stored.27, 28, 29 It is well known that some pulmonary neuroendocrine tumors express neural marker(s), but are completely negative for either chromogranin A or hormones, indicating these tumors have neural differentiation, but lack an endocrine phenotype. In the present study, most hASH1+ large-cell neuroendocrine carcinomas and small-cell carcinomas expressed chromogranin A, while the vast majority of hASH1− large-cell neuroendocrine carcinomas and small-cell carcinomas were completely negative for chromogranin A (P<0.0001). Furthermore, hASH1 expression was completely irrelative to neural cell adhesion molecule expression (P=0.8892). Thus, we suspected that a linkage existed between hASH1 expression and cellular endocrine phenotype other than ‘neuroendocrine differentiation’ in a broad sense.

However, 15 hASH1+ large-cell neuroendocrine carcinomas and small-cell carcinomas showed no chromogranin A expression. Although chromogranin A expression generally parallels hormone expression, it is not uncommon for pulmonary and other neuroendocrine tumors to express hormone(s), but be negative for chromogranin A. One possible explanation for this is that chromogranins B and/or C, but not chromogranin A, are alternatively present in these tumors.30 According to our previous studies,3, 26 gastrin-releasing peptide and calcitonin are the most frequently expressed hormones in lung cancer, and a combination of immunostainings of these two hormones and chromogranin A can discriminate most, if not all, lung cancers with an endocrine phenotype. In the present study, first, either gastrin-releasing peptide or calcitonin expression was related to hASH1 expression in large-cell neuroendocrine carcinomas and small-cell carcinomas (P<0.0001, respectively). Second, when taking the staining results together, there was a very close positive correlation between hASH1 and chromogranin A/gastrin-releasing peptide/calcitonin expressions (P<0.0001, r=0.852). Thus, the present results based on a substantial number of cases strongly suggested that hASH1 expression, at least in the context of lung cancer, is associated with a cellular endocrine phenotype. This notion is supported by a recent observation that development of most adrenal medullary cells in MASH1-deficient mice were arrested at an early stage, and the medullary cells expressed high levels of neurofilament and resembled immature sympathetic neuroblast, but failed to acquire an endocrine phenotype.17

On the other hand, hASH1 expression was virtually absent in typical carcinoid, which has the most overt and mature endocrine phenotype. Across the spectrum of pulmonary neuroendocrine tumors, typical carcinoid and small-cell carcinoma represent the two ends of the differentiation extent, with atypical carcinoid and large-cell neuroendocrine carcinoma in between. The Drosophila and mammalian homologs of hASH1 are only transiently expressed in neuronal progenitor cells and pulmonary neuroendocrine cells during early fetal development,8, 9, 11, 13 and in vitro hASH1 expression preceded and overlapped induced neuronal differentiation of neuroblastoma and P19 cells, but vanished before the overt neuronal phenotype appeared.22, 31 Silencing of hASH1 expression also parallelled the development of induced mature C cell phenotype of medullary thyroid carcinoma cells.20 In the present study, hASH1 expression also depended on differentiation extent, and was detected in the vast majority of less differentiated atypical carcinoids, as well as in large-cell neuroendocrine carcinomas and small-cell carcinomas, but was virtually absent in almost fully differentiated typical carcinoids. These data together strongly suggest that hASH1 expression in pulmonary neuroendocrine tumors imitates its early and transient expression pattern during development, and hASH1 is instrumental in establishment of a cellular endocrine phenotype, but is not necessary for maintenance of the fully developed endocrine phenotype. Thus, it is not surprising that clinically benign pheochromocytomas showed no hASH1 expression, while neural crest-derived malignant tumors expressed hASH1.12, 20, 22, 23 Our typical carcinoids and atypical carcinoids were differentially diagnosed based on mitotic count and/or presence or absence of necrosis.2, 24 As one of the four hASH1+ typical carcinoid patients died of the disease while all 26 hASH1− typical carcinoid patients survived, hASH1+ typical carcinoid diagnosed with these criteria might be biologically more like an atypical carcinoid.

hASH1 expression was associated with a significantly shortened overall survival in small-cell carcinoma patients, suggesting the potential prognostic significance of hASH1. Since peptide hormones such as gastrin-releasing peptide can promote cell growth and protect tumor cells from apoptosis through autocrine loops,4, 5 hASH1 might confer a role on the biological behavior of small-cell carcinoma via its impact on the endocrine phenotype of the tumor cells. However, hASH1 expression was not significantly associated with clinical outcome in large-cell neuroendocrine carcinoma patients, although a shorter 5-year survival was observed in patients with hASH1+ tumor. The prognostic significance of hASH1 expression in pulmonary neuroendocrine tumors remains to be clarified.