Do all lung adenocarcinomas follow a stepwise progression?

Abstract

Similar to the adenoma-carcinoma sequence of colorectal cancer, lung adenocarcinoma is thought to follow a linear multistep progression, in which a precursor lesion progresses to adenocarcinoma in situ, which is followed by invasive adenocarcinoma. However, lung adenocarcinoma can no longer be considered as a single type of tumor but rather a group of distinct subsets of tumors that arise from different molecular pathways. Consistent with this concept, recent findings revealed that this linear progression might not occur in all lung adenocarcinomas. First, according to the molecular classification based on expression profiling, lung cancer can be divided into at least two subsets; precancerous and in-situ lesions share characteristics of molecular expression and clinical features with only one of the two subsets, suggesting that the linear progression is only applicable to the subset in the molecular classification. Second, when EGFR and KRAS were examined based on the progression steps, the mutation rate of KRAS was disproportionally distributed; however, according to the progression schema, gene alterations should be evenly accumulated along the entire progression. Third, by means of comparative genomic hybridization analysis, some adenocarcinoma in situ revealed gene alterations discontinuous to invasive adenocarcinoma. Finally, there were some clinical observations that support that some lesions retain the progression. In this review, we hypothesize a novel scenario for the progression of lung adenocarcinoma, which does not support a linear progression schema.

Introduction

Despite significant advances in clinical treatment and in the understanding of its biology, lung cancer remains the most common cancer in men and the leading cause of cancer-related death in men and women worldwide [1]. Understanding the molecular pathogenesis of lung cancer is crucial for developing effective prevention and therapeutic strategies. Currently, lung adenocarcinoma is thought to follow a linear multistep progression, similar to the adenoma-carcinoma sequence of colorectal cancer. A precursor or premalignant lesion, known as atypical adenomatous hyperplasia (AAH), progresses to a preinvasive adenocarcinoma in situ, formerly known as bronchioloalveolar carcinoma (BAC) [2], which is followed by invasive adenocarcinoma (Fig. 1). Ultimately, some of the invasive adenocarcinoma acquires the potential to metastasize to other organs. This idea is supported by the finding that an adenocarcinoma commonly has a low-grade, in-situ lesion in the periphery, which is morphologically identical to adenocarcinoma in-situ (BAC). However, in contrast to colorectal cancer, the low-grade lesion gradually progresses to invasive carcinoma without a clear boundary between the adenoma and carcinoma components. In terms of lepidic growth (growth along alveolar walls), as metastatic cancers show such growth pattern, it would be reasonable that invasive lung cancers grow along the alveolar wall. Indeed, it is reported that peripheral squamous cell carcinoma could follow such growth pattern. Furthermore, although such low-grade lesions morphologically mimic precursor or in-situ-like lesions, expression profiles revealed no significant difference between peripheral low-grade and central invasive component [3]. In addition, with the exception of EGFR amplification [4, 5], no genetic alterations that can differentiate between the two lesions have yet been identified. In this review, we propose a possible novel scenario for the progression of lung adenocarcinoma, which depends on the acquired genetic changes.

Figure 1.

Current linear progression schema for lung adenocarcinoma. Lung glandular neoplasia initiates as a precursor lesion, which is referred to as atypical adenomatous hyperplasia (AAH) and progresses to adenocarcinoma in situ (bronchioloalveolar carcinoma, BAC), followed by invasive adenocarcinoma.

Linear progression schema is only applicable to a subset of lung adenocarcinomas

Genome-wide expression profiles can be used to describe the characteristic expression patterns of various molecules in individual cancers. Unsupervised hierarchical clustering based on expression profiling illustrates the molecular classification of tumors, which is based on the similarity of genome-wide expression patterns. According to its molecular classification, lung cancer is largely divided into two subsets [6, 7]. One subset is comprised of adenocarcinoma, whereas the other includes all four of the histological subtypes. The former subset is further divided into two clusters and is referred to as the “terminal respiratory unit” and non-terminal respiratory unit branches. Hayes et al. analyzed the expression profile of three independent cohorts and concluded that this configuration of molecular-defined subsets of lung adenocarcinoma is quite robust [8]. Within this molecular schema, all of the precursor lesions and in-situ adenocarcinomas are associated with the “terminal respiratory unit” branch in terms of their expression profiles and clinical characteristics. Vast majority of the precursor lesions and adenocarcinoma in situ express thyroid transcription factor-1 and surfactant proteins [9], and adenocarcinomas in the “terminal respiratory unit” branch are characterized by high expression of these molecules [10]. Because thyroid transcription factor-1 is a lineage marker of the “terminal respiratory unit” [11], shared expression of this gene suggests that these lesions belong to the same cellular lineage and are related. Furthermore, because these lesions are developed preferentially in females and non-smokers, these patient characteristics are also associated with the same group. In terms of genetic changes, EGFR mutations, which can be detected in the precursor and in-situ lesions, are nearly specific to the “terminal respiratory unit” branch [10, 12]. This finding also supports the applicability of the current linear progression schema only to a subset of adenocarcinomas, namely those of the “terminal respiratory unit” branch.

Disproportionate distribution of EGFR and KRAS mutations in the linear progression schema

Among the RAS gene family members, KRAS is most frequently mutated in human tumors. In the lung, 10–15% of cancers harbor KRAS mutations, which are specific to adenocarcinomas [13]. These mutations have been detected in precursor lesions (AAHs), which suggests a very early involvement of this gene in the progression of lung adenocarcinoma [14, 15]. Indeed, in KRAS transgenic mice, mutant KRAS causes non-invasive lung adenomas [16]; only the selective cellular lineages, including the lung peripheral cells, are affected by the expression of this oncogene. Guerra et al. reported that the expression of KRAS^V12 throughout the body does not cause early-phase proliferative lesions, but rather only a percentage of KRAS^V12-expressing lung bronchiolo-alveolar cells lead to the formation of multiple adenomas [17]. Furthermore, a recent study by Collado et al. suggested that the resultant adenomas in KRAS^V12-expressing mice had a biological significance that was different from the role of the mutant KRAS in advanced adenocarcinomas [18]. Adenomas, but not adenocarcinomas, underwent KRAS-induced senescence as shown by their expression of β-galactosidase, p16, p15 and HP1-γ. This finding suggested that cells in premalignant tumors undergo oncogene-induced senescence, whereas cells in malignant tumors overcome the barrier of senescence. Indeed, the induction of KRAS^G12D in murine intestinal epithelial cells induces the development of serrated hyperplasia showing a hallmark of cellular senescence [19]. However, deletion of p16 enabled the hyperplastic cells to overcome senescence, and consequently progress and metastasize. Taken together, these studies suggest KRAS can play dual roles in the molecular pathogenesis of cancer.

It has been reported that EGFR mutations are also detected in AAH and adenocarcinoma in situ [12]. Together with the above findings, both EGFR and KRAS are involved in the early phase of lung adenocarcinoma development. Therefore, the frequencies of the EGFR and KRAS mutations were examined along the progression schema from AAH to adenocarcinoma in-situ to adenocarcinoma. Strikingly, KRAS mutations decreased along the progression, detectable in 33% of AAH, 12% of carcinomas in situ, 8% of minimally invasive adenocarcinomas and 0% of well-differentiated adenocarcinomas[20]. In contrast, EGFR-mutated tumors were evenly distributed along each progression step, although the incidence in adenocarcinoma with a predominant BAC pattern was higher [20]. The distribution of KRAS mutations suggests that KRAS-mutated AAH rarely progresses to more advanced tumors because the overall frequency of KRAS mutations in lung adenocarcinoma was limited to 10–15% [21, 22], with most of them being different types of adenocarcinoma, i.e., mucinous/goblet cell adenocarcinoma[23, 24]. According to the results of the mouse model, senescence may occur in AAH. Indeed, decreased expression and epigenetic silencing of p16 were significantly infrequent in AAH [25, 26]. The putative progression schema based on these findings is illustrated in Figure 2. As mentioned above, AAH, BAC and invasive cancer belong to in a group of TRU lineage, thus two progression schemes are described within the group. Regarding non-TRU lineage, as shown “the other cells” in Figure 2, precursor lesions have not been well documented. UllmannR et al. reported “bronchiolar columnar cell dysplasia” as a putative preneoplastic lesion of bronchiolar epithelium, which was morphologically different form AAH [27]. On the other hand, it has also been reported that the degree and incidence of aneuploidy increased with progressive severity of morphological change in the preinvasive area near the invasive cancers, although altered ploidy was more frequent in the peripheral parenchyma (bronchioles or alveoli) than in the central bronchi [28].

Figure 2.

Comparison between current linear progression schema and novel non-linear progression schema. Current model follows a single linear progression regardless of the genes involved, whereas the novel model has several pathways to invasive adenocarcinoma, some of which may be terminated prior to invasive adenocarcinoma.

Discontinuous genetic changes along the progression model revealed by CGH analysis

The molecular analysis of lung adenocarcinoma has identified distinct morphological subtypes as well as a small number of discrete molecular mutations that are therapeutically significant but account for a small proportion of tumors. More recent work has identified a discrete chromosomal translocation of the EML4/ALK locus in adenocarcinoma that may account for 10% of tumors and signify sensitivity to ALK inhibitor therapy [29]. The EML4/ALK studies in lung adenocarcinoma taken together with work showing the importance of amplification of pathognomonic translocation in driving the progression of sarcoma [30] indicate the potential important role for discontinuous genetic amplification events in promoting lung adenocarcinoma progression and metastasis.

Recent unpublished observations of gene expression signatures in microdissected adenocarcinoma tumor cells support the hypothesis that focal genetic alterations are important for lung adenocarcinoma invasiveness. Three hundred forty genes were differentially expressed between the two subclasses of adenocarcinoma in-situ and lepidic predominant invasive adenocarcinoma. We examined the chromosomal distribution of the 340-gene signature and identified significant overrepresentation of genes from chromosomes 7, 8, 9, 13 with the greatest percentage of differentially expressed genes located on chromosome 7 (Figure 3). To verify structural copy number changes, we performed comparative genomic hybridization (CGH) on metaphase spreads using whole-genome amplified DNA. These studies confirmed 7q amplification in all mixed tumors and showed less frequent amplification of loci on chromosomes 8, 9, and 13 that coincided with mRNA transcriptional data. In addition, CGH studies indicated uniform copy number increase of the 7p EGFR locus in adenocarcinoma in situ. Taken together, these findings suggest the following paradigm: 1). EGFR amplification drives tumor proliferation in these subtypes; 2). Amplification of 7q loci promotes invasion in adenocarcinoma; 3). copy number changes of some loci was discontinuous to invasive adenocarcinoma.

Figure 3.

Chromosomal Distribution of 340 gene Lung Adenocarcinoma Invasion Classifier. Significant overrepresentation of genes in the classifier is shown from chromosomes 7, 8, 9, 13 with the greatest percentage of differentially expressed genes located on chromosome 7.

Non-linear progression of lung adenocarcinoma supported by clinical observations

Frequently, it is assumed that lung cancer patients with AAH that has been histologically detected in resected non-malignant lung tissue, would have AAHs in the other lobes that could portend the subsequent development of secondary invasive adenocarcinomas and poor prognosis. However, Tanigawa et al. reported that the prognosis of patients without AAHs was almost same or slightly better than that of patients with AAH, even when adjusted for the pathological stages of the primary cancers[31]. Similar results were reported by Chapman et al. [32]. Further, it is widely accepted that small lesions with ground-glass opacity (GGO) appearance frequently persist for a decade without changes in size (Figure 4). The GGO lesions correspond to lepidic growth morphology, which are comprised mostly of adenocarcinoma in situ and occasionally of precancerous lesions (AAH), although inflammatory lesions may also be included. Hiramatsu et al. reported that a repeat high resolution CT obtained with an average of 2.8 years of observation revealed increases in size or increased in density in only 26 (21%) of 125 GGO lesions [33]. They also noted that the time to such progression was significantly different between small and large GGO lesions, and only 14% of small GGOs (10 mm or less in diameter) developing such progression. This finding also suggests that many preinvasive lesions maintain their size for long periods, which is suggestive of cellular senescence. Further research is required to define the factors predictive of progression or senescence in these GGO lesions.

Figure 4.

Long-standing small lesion with ground-glass opacity (GGO) appearance. The patient with this lesion was observed for 5 years, but the size and radiographic features of the lesion did not significantly change.

Conclusion

In this review, we describe a possible scenario for the development of lung adenocarcinoma. The existing paradigm of stepwise progression in terminal respiratory unit adenocarcinoma is exemplified by EGFR gene amplification that is seen in central invasive foci but not in peripheral lepidic foci, despite both sites harboring identical EGFR mutations [4, 5]. In other words, identical EGFR mutations suggests a clonal relationship between peripheral low-grade and central invasive components, while the finding of gene amplification that is restricted to the central invasive components suggests that amplification is acquired in association with tumor progression. This finding clearly shows linear progression of lung adenocarcinoma. However, presence of linear progression schema does not preclude other schema and nonlinear progression model could be adapted partly to lung adenocarcinoma development as shown here (Figure 2). Notably, under the proposed hypothesis of lung adenocarcinoma progression, since the fate of a tumor is influenced by the initial oncogenic driver, treatment strategies for small GGO lesions could be determined by the mutation.