Introduction

Therapeutic strategies for lung neuroendocrine neoplasms are not completely defined and greatly vary from carcinoids to high-grade carcinomas. In the former group, surgery is the mainstay for the therapy of resectable tumors, whereas adjuvant strategies are not well established yet. In recurrent or metastatic diseases, biotherapies (i.e., somatostatin analogs and mTOR inhibitors) have been proposed, but no clinical evidence is currently available on their real impact to increase patient survival, and large prospective trials are still ongoing [1]. Chemotherapeutic strategies are the only effective treatment in small-cell lung carcinoma (SCLC) and large-cell neuroendocrine carcinoma (LCNEC), although the high response rates (especially for SCLC) are paralleled by a very high likelihood of recurrence/progression and by a generally unfavorable outcome. Chemotherapeutic strategies are also considered in moderately proliferating carcinoids [2] generalizing to the lung the experience in gastrointestinal and pancreatic neuroendocrine neoplasms, but the evidence on the efficacy of chemotherapeutic agents derives from small retrospective series rather than from specifically designed prospective clinical trials.

The present paper is aimed at summarizing the pathologist’s role in determining predictive biomarkers of response to treatment in lung neuroendocrine neoplasms, with special reference to carcinoids (see Table 1).

Table 1 Candidate biomarkers with predictive value in lung neuroendocrine neoplasms
Full size table

Morphological and Phenotypic Markers

Morphology

Neuroendocrine (NE) neoplasms of the lung have been variably labeled for the purpose of better defining the different histotypes and differentiation grades [3]. The current WHO classification [4] recognizes four entities, defined by the term “carcinoid” for low/intermediate-grade tumors and “large- or small-cell carcinoma” for high-grade tumors [5, 6].

Typical carcinoids (TCs) account for about 1–2 % of lung tumors and exhibit a classical organoid (acinar, trabecular) pattern with polygonal, minimally atypical cells. Necrosis is absent and mitoses are <2/2 mm2. Similar tumors having a size of <5 mm are labeled neuroendocrine tumorlets [7, 8].

Atypical carcinoids (ACs) are extremely rare, cigarette-smoking-related tumors, though often associated with regional and distant metastases. Their morphology overlaps that of TC, except that necrosis is present and/or the mitotic count is 2–10/2 mm2 [9, 10].

Large-cell NE carcinoma (LCNEC) partly resembles the organoid architecture of AC but is made of larger cells, necrosis is extensive and the mitotic index exceeds by far 10 in 2 mm2. In the 2004 WHO scheme [4], LCNEC is classified among non-NE large-cell carcinomas, from which it should be distinguished based on the recognition of a NE phenotype, by either morphology or immunohistochemistry for neuroendocrine markers [11]. A subgroup of undifferentiated large-cell carcinomas with NE morphology but no NE marker expression has been also described [12].

Small-cell lung carcinoma (SCLC) is the most common lung NE neoplasm and classically is characterized by small cells with scant cytoplasm and condensed chromatin, a diffuse growth pattern, extensive necrosis, and a very high mitotic index (largely exceeding 10/2 mm2).

Combined NE carcinomas are the result of a relatively uncommon association of SCLC or LCNEC with conventional squamous cell or adeno-carcinoma component. Focal NE differentiation in conventional lung carcinomas is excluded from this definition, because its relevance to clinics and tumor behavior is unclear.

Immunophenotype

Chromogranin A and synaptophysin expression, in the absence of high molecular weight cytokeratins [13] or p40 [14], are the most reliable NE markers, whereas PGP9.5, NSE, and CD56 are less specific. In high-grade NECs (but not in carcinoids), hASH-1 (a transcription factor driving NE differentiation during human development) expression has been reported [15]. NE markers are mandatory for recognizing the NE nature of a lesion, although none of them has been demonstrated to be superior to morphology in terms of prognostication or response-to-therapy prediction. Lung (and thyroid) specific marker TTF-1 is mostly expressed by high-grade NE carcinomas of both pulmonary and extra-pulmonary origins, while carcinoid tumors are either unreactive or may show variable positivity especially in peripheral lesions [4, 16, 17]. A non-NE marker has been recently proposed to help in identifying mitotic figures, namely phosphohistone 3 [18].

Prognostic and Predictive Molecular Markers

Ki67 index

The proliferation index detected by Ki67 is also a useful tool to better classify a lung neuroendocrine tumor (NET), although not included in the WHO classification criteria [4], at variance with the digestive tract NETs [19]. The reason of this difference may be related to the current lack of validation of Ki67 index in lung NETs [20]. Indeed, the proliferative activity of pulmonary NE tumors has been extensively investigated (reviewed by Pelosi et al. 2014 [21]). The diagnostic role of Ki67 is so far well established in small biopsy or cytology samples, in which artifacts may hamper the differential diagnosis between small-cell lung cancer and carcinoid tumor in individual cases [22]. Analyzing the Ki67 indexes in over 1,800 reported cases, it appears that the mean proliferation values for TC, AC, LCNEC, and SCLC are 1.6, 7.3, 48.5, and 58.9 %, respectively, and these figures are paralleled by the different survival rates reported for the single histotypes [23]. There may be, however, significant overlap of Ki67 indexes between biologically adjacent tumor categories (TC vs. AC; AC vs. LCNEC; LCNEC vs. SCLC), thus preventing its reliable diagnostic use in individual cases (Fig. 1). Some authors found a Ki67 performance equal or higher than that of mitotic count for both diagnostic (cutoff between TC and AC proposed at 4 %) and prognostic purposes [24–26].

Fig. 1
figure 1

a Pulmonary atypical carcinoid with vascular invasion and b lymph node metastasis, both having a high Ki67 proliferation index (c, d) estimated at 16 % per 10HPF (c, insert) (a, b H&E, ×100; c, d immunoperoxidase, ×100; inset ×400)

Full size image

In a recent study on 399 NE tumors of the lung [27], Ki67 index was incorporated into a newly proposed grading system, which also considered two conventional morphological parameters (mitotic count and necrosis). Adapting cutoff values for mitoses and Ki67 at 4 and 25 %, it was found that a three-grade system can be reliably obtained when at least two of the three parameters were identified which allows the stratification of NE tumors into three subgroups with significantly different survivals. In terms of prediction, Ki-67 has not been associated with the response to any specific treatment in carcinoids. By contrast, in high-grade carcinomas, the higher Ki67 values detected in SCLC (90 %) as compared to those in LCNEC (50–60 %) have been postulated to be associated with different chemotherapy responses [28].

Molecular Profile

The molecular profile of lung NE tumors has been extensively investigated to identify diagnostic, prognostic, and predictive factors, and to possibly lead to a “molecular classification” of NETs [29]. Specific chromosomal alterations (e.g., 11q22.3-q25 losses) [30], oncogene mutations, and cell cycle deregulation [31] were documented in lung NETs [32]. The mutational profile of lung NETs is still largely unknown. MEN1 gene is the most largely investigated and found to be mutated in approximately 13 % of carcinoids but very rarely in high-grade NE carcinomas [33]. In term of prognosis, MEN1 gene mutation or loss was significantly related to shorter survival in carcinoid patients, together with tumor stage [34], in keeping with its function as a tumor suppressor gene. Novel data are coming from next-generation sequencing analysis, showing a genetic similarity between pulmonary and pancreatic well-differentiated neuroendocrine tumors, but surprisingly not between TC and AC in the lung [35].

Gene and protein expression profiles were also extensively investigated. The latter was analyzed by proteomics and immunohistochemistry in carcinoid and SCLC, ultimately leading to the identification of over 300 differentially expressed proteins in each tumor subtype [36]. Gene expression profiling studies identified also novel prognostic markers in the group of lung carcinoids, independent of the histological type. Among others, three genes were found to bear prognostic implications, namely orthopedia homeobox (OTP), CD44, and RET. In particular, significant associations with reduced 20-year survival were observed in the case of low messenger RNA (mRNA) levels of CD44 (p = 0.000018) and OTP (p = 0.00054), and high RET levels (p = 0.025) [37]. The same authors also found a different gene expression profile in a small series of 10 bronchial carcinoids having a favorable or a poor outcome (five cases each). The latter had a significantly higher number of downregulated genes at chromosome 11q, a region frequently lost in carcinoids (p = 0.00017). Upregulated genes are involved in the mitotic spindle checkpoint, the chromosomal passenger complex (CPC), mitotic kinase CDC2 activity, and BRCA-Fanconi anemia pathway. The above mentioned CD44 and OTP genes, as well as others, including BIRC5 (survivin), BUB1, IL20RA, and KLK12, were found to be independent predictors of patient outcome [38].

Predictive Markers

Predictors of response to chemotherapy or biotherapy are increasingly being evaluated [39–41], but investigational studies specifically designed in lung neuroendocrine neoplasms are meagre.

Thymidylate synthase (TS) is the target of antifolate drugs and intratumoral expression levels may predict response to an antifolate-based regimen. A differential expression of TS mRNA and protein in the spectrum of pulmonary NE neoplasms was observed. TS levels were higher in poorly differentiated NE carcinomas, thus supporting the extremely poor activity of these drugs in small-cell lung cancer [42].

The expression of specific receptors or enzymes implicated in the response to biotherapies has been demonstrated [16]. Somatostatin receptors (SSTRs) have been identified in NE tumors by different techniques [43]. The immunohistochemical expression of SSTR types 2 and 3 was investigated in 218 aggressive lung NE tumors (metastatic TC, AC, LCNEC, and resected SCLC). The expression of SSTRs was progressively reduced in poorly differentiated forms and correlated with octreotide scintigraphy in 70 % of cases [44] (Fig. 2). The mTOR pathway has been explored in lung NET [45]. and a lower expression of active forms of mTOR and S6K was detected in high-grade carcinomas (of either large or small cell types) [46] (Fig. 3). Indeed, the mTOR pathway is a complex network of factors, and the potential role of its players in predicting response to mTOR inhibitors (rapalogs) is unknown. Several proteins belonging to the mTOR complexes, or downstream and upstream to mTOR, interplay in regulating such central intracellular signaling pathway, ideally all being candidate biomarkers of prediction of response to mTOR inhibition. Moreover, additional players interact with mTOR, such as nutrient transporters or somatostatin receptors themselves. Preliminary data from our laboratory (Rapa and Volante, unpublished results) indicate, among others, a significant inverse correlation between the expression of glucose transporter GLUT-1 and mTOR signaling. In addition, the expression of some of the above molecules, in particular p-mTOR and GLUT-1, as well as of the amino acid transporter LAT-1, was strongly associated with SSTR2A expression, suggesting that somatostatin receptor inhibitor effects may result from mTOR pathway control and that synergies can be obtained by combined treatments, as suggested in intestinal NETs [47]. Indeed, in in vitro models, bronchial carcinoid cells of patients responding to mTOR inhibitors were shown to have higher levels of phosphorylated mTOR [48].

Fig. 2
figure 2

Somatostatin receptor type 2 expression in a pulmonary atypical carcinoid (a, score 2 [44]) with peritumoral vascular invasion (b) and lymph node metastasis (c, score 3 [44]) (immunoperoxidase, ×200)

Full size image
Fig. 3
figure 3

mTOR pathway activation in a typical carcinoid with lymph node metastases. a Organoid and trabecular growth pattern (H&E, ×200); b phospho-mTOR protein expression, Hscore = 110 [46], and c phospho-p70S6K protein expression, Hscore = 240 [45] (c, d immunoperoxidase, ×200)

Full size image

Other synergistic combination effects (i.e., apoptosis induction) were recently reported, involving erlotinib (targeting EGFR) combined with everolimus (targeting mTOR) in AC and LCNEC [49].

In another study, significant differences in c-KIT and HER2 expression were seen between LCNEC and AC, while EGFR mutations were more common in AC than in LCNEC. A potential role for VEGF and c-KIT (and possibly HER2) targeting agents in the treatment of LCNEC was therefore suggested [50, 51].

Finally, c-MET oncogene has been investigated in pulmonary NE neoplasms. PAX5 was shown to upregulate c-MET in small-cell lung carcinoma, and PAX5 and c-MET co-inhibition produced a synergistic effect in killing tumor cells, probably related to paxillin inactivation (a downstream target of activated c-Met involved in cell motility and tumor spread) [52]. However, regardless of TKI treatment, c-Met activation (phosphorylation) was not influenced by the mutational status, which was detected in 25 % of SCLC cell lines, in 8.3 % of NETs, and in 6.5 % of SCLC cases [53].