Pulmonary adenocarcinoma with mucin production modulates phenotype according to common genetic traits: a reappraisal of mucinous adenocarcinoma and colloid adenocarcinoma.

Whether invasive mucinous adenocarcinoma (IMA) and colloid adenocarcinoma (ICA) of the lung represent separate tumour entities, or simply lie within a spectrum of phenotypic variability, is worth investigating. Fifteen ICA, 12 IMA, 9 ALK-rearranged adenocarcinomas (ALKA), 8 non-mucinous KRAS-mutated adenocarcinomas (KRASA) and 9 mucinous breast adenocarcinomas (MBA) were assessed by immunohistochemistry for alveolar (TTF1, cytoplasmic MUC1), intestinal (CDX-2, MUC2), gastric (membrane MUC1, MUC6), bronchial (MUC5AC), mesenchymal (vimentin), neuroendocrine (chromogranin A, synaptophysin), sex steroid hormone-related (oestrogen and progesterone receptors), pan-mucinous (HNF4A) and pan-epithelial (keratin 7) lineage biomarkers and by targeted next generation sequencing (TNGS) for 50 recurrently altered cancer genes. Unsupervised clustering analysis using molecular features identified cluster 1 (IMA and ICA), cluster 2 (ALKA and KRASA) and cluster 3 (MBA) (p < 0.0001). Cluster 1 showed four histology-independent sub-clusters (S1 to S4) pooled by HFN4A and MUC5AC but diversely reacting for TTF1, MUC1, MUC2, MUC6 and CDX2. Sub-cluster S1 predominantly featured intestinal-alveolar, S2 gastrointestinal, S3 gastric and S4 alveolar differentiation. In turn, KRASA and ALKA shared alveolar lineage alongside residual MUC5AC expression, with additional focal CDX2 and diffuse vimentin, respectively. A proximal-to-distal scheme extending from terminal (TB) and respiratory (RB) bronchioles to alveolar cells was devised, where S3 originated from distal TB (cellular mucinous adenocarcinoma), S2 from proximal RB (secreting mucinous adenocarcinoma), S1 from intermediate RB (mucin lake-forming colloid adenocarcinoma), S4 from distal RB (colloid alveolar adenocarcinoma), KRASA from juxta-alveolar RB (KRAS-mutated non-mucinous adenocarcinoma) and ALKA from juxta-bronchial alveolar cells (ALK-translocated adenocarcinoma). TNGS analysis showed KRAS, LKB1, TP53, APC and CDKN2A mutation predominance. In conclusion, IMA and ICA are basket categories, which likely originate from distinct domains of stem/progenitor cells spatially distributed along bronchioles upon common molecular features and genetic alterations.

Introduction

Among mucin‐laden adenocarcinomas (MLA) of the lung, the 2015 WHO classification identifies mucinous adenocarcinoma in situ, minimally invasive mucinous adenocarcinoma, invasive mucinous adenocarcinoma (IMA) and invasive colloid adenocarcinoma (ICA) 1. Invasive enteric adenocarcinoma (IEA) resembling the homologous intestinal tumours has been included as a variant 1. Other MLA types are solid adenocarcinoma with mucin 1 and signet ring cell adenocarcinoma (SRCA) 2, although SRCA has been downgraded to mere cytological change of any pattern's adenocarcinoma 1. Lastly, there are non‐mucinous adenocarcinomas accumulating extracellular mucin, which have been reappraised as histological patterns of adenocarcinoma with mucin extravasation (HPAME) 3. IMA, ICA and IEA share many phenotypic traits, which are opposed to conventional non‐mucinous adenocarcinomas and HPAME. Expression of NK2 homeobox 1/thyroid‐specific transcription factor‐1 (NKX2‐1/TTF‐1) 4, 5, 6, 7, napsin A aspartic peptidase (NAPSA) 4, 8, 9, surfactant protein A1 (SFTPA1) 5, 8, caudal type homeobox 2 (CDX2) 5, 6, 8, 10, 11, keratin 20 (KRT20) 5, 6, 8, 12, mucin 1 cell surface associated (MUC1) 13, mucin 2 oligomeric mucus/gel‐forming (MUC2) 5, 11, 13, mucin 5AC oligomeric mucus/gel‐forming (MUC5AC) 5, 10, 13, mucin 6 oligomeric mucus/gel‐forming (MUC6) 10, 13, 14 and hepatocyte nuclear factor 4 alpha (HFN4A) 15, have been variably reported onto emphasize the inherent heterogeneity of these tumours. Furthermore, IMA may even express neuroendocrine (NE) markers 16 and sex steroid hormone receptors 17, 18.The molecular landscape of IMA, ICA, IEA and, to lesser extent, HPAME is dominated by KRAS 3, 6, 10, 19, 20, TP53 and LKB1/STK11 21, 22 over Epidermal Growth Factor Receptor (EGFR) mutations 3, 6, 19, while the CD74‐NRG1 fusion genes has been documented in 13–27% KRAS‐wild type tumours 23, 24. Most SRCA harbour ALK rearrangement 2, 25.

Recent data have indicated that synchronous KRAS mutation and Nkx2‐1/TTF1 repression cause IMA in the lung 26, with napsin‐A downregulation 10 and gastric over intestinal differentiation 10, 27. These Nkx2‐1/TTF1‐deleted IMA with KRAS mutation grew in well‐differentiated multiple foci upon reduced apoptosis from mucinous precursors in the distal bronchiole epithelium 27, developing metastases only when loss‐of‐function TP53 mutation occurred 28.

According to these assumptions, it is tempting to speculate that IMA, ICA and IEA lie with in a spectrum of relatively uniform tumours on molecular grounds 6, 19, 20, 23, which differ upon morphology and immunohistochemistry (IHC) traits 4, 5, 6, 7, 8, 10. They would therefore embody prototypical lesions resembling gastro‐entero‐pancreatic or gynaecological tumours 29, 30, 31, 32, and even primary ALK‐translocated adenocarcinoma (ALKA) when abundant extracellular mucin is accumulated 6, 11, 33, 34. Upon IHC and molecular investigation, we herein suggest that IMA, ICA and IEA make up a heterogeneous family of tumours encompassing four different subgroups, which likely arise from different cancer progenitor/stem cell niches in the distal airways.

Materials and methods

Study design

We designed and conducted a proof‐of‐concept study to challenge whether pulmonary IMA, ICA and IEA encompass separate tumour entities with distinct phenotypic profiles depending upon varying interplay of architectural complexity and lineage changes (cellular phase) or mucin production and extravasation (secretion phase), under the pooling of a common mucin‐encoding gene activation programme, or simply made up a spectrum of histological heterogeneity.

Patients and tumours

A series of 44 consecutive resection specimens of mucin‐laden lung adenocarcinoma patients from 25 males averaging 63.8 years (range 28–85 years) and 19 females averaging 61.4 years (range 38–80) were retrospectively identified in the pathology archives of the participant Institutions among diagnoses of lung adenocarcinoma, inasmuch as ICA, IMA and ALKA are less likely to be documented in surgical specimens for being either quite uncommon tumours or inoperable lesions in most instances at the time of the clinical diagnosis. All tumours fulfilling diagnostic criteria for mucin‐laden adenocarcinoma made up different comparable groups in terms of confounding factors, such as age, gender and tumour stage as detailed in supplementary material A. They included 15 ICA (5 pure and 10 mixed with non‐colloid component), 12 IMA (9 pure and 3 mixed with non‐mucinous component), 9 ALKA and 8 non‐mucinous KRAS‐mutated adenocarcinomas (KRASA). The occurrence of ALK rearrangement and KRAS mutation had been verified by fluorescence in situ hybridization 35, 36 and Sanger direct sequencing of exon 2 to 4 37, 38, respectively. As negative control group of non‐pulmonary and unrelated MLA, nine mucinous breast adenocarcinomas (MBA) entered the study, inasmuch as mucin‐laden adenocarcinomas of the lung are prototypical lesions with similarities to homologous digestive and gynaecological mucinous tumours likely due to commonalities in the embryological derivation and pathogenesis.

All tumours were classified according to the criteria of WHO classifications on lung 1 and breast 39 cancer. There were 30 specimens from (bi)‐lobectomy, 12 from segmentectomy and 2 from pneumonectomy corresponding to 21 tumours staged IA, 5 staged IB, 2 staged IIA, 11 staged IIB, 3 staged IIIA and 1 staged IV (one patient was lost to pathologic staging). Information on smoking habit was not available at the time of the present study and survival analysis was not a specific endpoint of the study. All surgical specimens had been fixed in 4% buffered formaldehyde solution for 12–24 hr and embedded in paraffin according to standard histopathological methods. All the original haematoxylin and eosin sections were jointly reviewed by four (GP, MC, AF, AS) of the authors for consistency, without knowledge of the patients’ identities or original tumour categorization. Pleural invasion was classified as PL0, PL1, PL2 and PL3 according to updated criteria 40.

The group of IMA, ICA, ALKA and KRASA was comparatively studied for age, gender, type of surgery, tumour stage, tumour diameter, number of excised lymph nodes, percentage of metastatic lymph nodes, pT factor, pN factor, pleural invasion, architectural patterns and variants, and occurrence of signet ring, goblet or floating cells (each by 10% cutoff for presence vs. absence), extracellular mucin (cutoff 50% for minor vs. prevalent) and type of mucin (basophilic, amphophilic or both), as detailed in supplementary material A.

In particular, non‐mucinous adenocarcinoma components were expressed as percentages of patterns of growth and variants in 5% numerical increments according to current classification guidelines 1. No cases of adenocarcinoma in situ or minimally invasive adenocarcinoma were documented in this study.

Immunohistochemistry

The primary antibody list entering the study and technical information on IHC procedures is available in supplementary material B and supplementary material C, respectively. The antibodies included biomarkers of alveolar (TTF‐1, MUC1), gastrointestinal (CDX‐2, keratin 20, MUC2, MUC6), bronchial (MUC5AC), NE (chromogranin A, synaptophysin), mesenchymal (vimentin), sex steroid hormone (oestrogen and progesterone receptors), pan‐mucinous (HNF4A) and pan‐epithelial (keratin 7) differentiation lineages, along with the ALK rearrangement‐associated protein. Normal lung tissue was investigated for the same IHC biomarkers.

Molecular study

The group of IMA/ICA/IEA tumours underwent targeted‐next generation sequencing (NGS). Normal lung tissue was used as negative controls as appropriate for consistency. Detailed descriptions of the molecular procedures pursued in this investigation, including DNA extraction, quantification and targeted next generation sequencing (TNGS) analysis, have been reported previously 35, 37, 38. In brief, NGS analysis was carried out using small DNA samples (10 ng/μl) from tumour paraffin sections accounting for at least 80% cellularity, with the Ion AmpliSeq Cancer Hotspot Panel v.2 (Life Technologies‐Thermo Fisher Scientific, Waltham, MA, USA) running on the Ion‐TorrentTM Personal Genome Machine platform comprising 50 oncogenes and tumour suppressor genes recurrently mutated in human cancers (Life Technologies‐Thermo Fisher Scientific), as previously detailed 37, 38.

Ethics

The study was approved by the independent ethics committee of the National Tumour Institute IRCCS Foundation, Milan, Italy (accession number INT‐188/15). All patients gave their written consent for diagnosis and research activities when they were admitted to the hospital.

Statistics

Statistical analyses were performed using JMP IN software (SAS Institute, Inc., Cary, NC) and VassarStats for Statistical Computation at http://vassarstats.net website. Qualitative or quantitative data were compared by Fisher exact t test, chi‐squared test, Mann–Whitney test and Kruskal–Wallis test as appropriate. Unsupervised and supervised hierarchical clustering analyses were performed using Cluster 3.0 software (http://www.eisenlab.org/eisen /) and visualized using Java TreeView (http://jtreeview.sourceforge.net). The defining features of the diverse clusters corresponded to the different IHC percentages used for the relevant markers. Similarity was measured using Spearman rank correlation metric and clustering was performed by centroid linkage. Contingency analysis and relative mosaic plots were performed by means of JMP IN software and calculating P values through the likelihood ratio test. Mosaic plots indicated graphically the fraction (expressed from 0.0 to 1.0) of patients belonging to the different clusters after stratifying for the different tumour subtypes. For all tests, two‐sided P values were taken into account, with a threshold < 0.05 as being statistically significant.

Results

Analysis by histology

We initially searched for either morphology or relevant IHC findings, which stratified the cohort of 44 lung adenocarcinomas. In KRASA and IMA, we found only predominant acinar‐cribriform or solid with mucin and lepidic/acinar/enteric patterns, respectively, and basophilic mucin in IMA, while in ICA, ALKA and KRASA we found the presence of extracellular mucin and SRC and the lack of goblet and freely floating cells, respectively (supplementary material A).The IHC analysis revealed that the alveolar biomarkers prevailed in mixed IMA/ICA, ALKA and KRASA, whereas gastrointestinal biomarkers were preferentially expressed in IMA and ICA alongside MUC5AC and HNF4A (supplementary material D; supplementary material, Figure S1A–F). Vimentin and ALK protein were limited to ALKA, whereas all tumours exhibited diffuse labelling for keratin 7 and at least focal NE biomarkers (supplementary material D; supplementary material, Figure S1A–F). Only the staining for sex steroid hormone receptors was erratic. For secreted mucin, MUC1 was only noted in 12/15 (80%) ICA, 5/12 (42%) IMA, 7/9 (78%) ALKA and 1/8 (13%) KRASA (supplementary material D). Further details of IHC findings, including normal lung tissue, are reported on supplementary material C, with representative pictures in supplementary material, Figure S1A–F.

Analysis by clustering

We then used hierarchical clustering analysis to perform a molecular classification of this cohort of 53 MLA, using a set of known IHC biomarkers. Such analysis revealed three main clusters and a non‐random distribution of these IHC biomarkers across the 53 MLA under evaluation. Upon histology, cluster 1 included a mixture of IMA and ICA and, unexpectedly, even one KRASA case; cluster 2 comprised all ALKA and the remaining KRASA cases; cluster 3 was limited to MBA (p < 0.0001) (Figure 1A and 1B). Upon IHC, gastrointestinal (CDX‐2, keratin 20, membrane MUC1, MUC6), bronchial (MUC5AC) and pan‐mucinous (HNF4A) biomarkers prevailed in cluster 1, whilst alveolar (TTF1, cytoplasmic MUC1), ALKA‐related (ALK protein, vimentin) and sex steroid hormone along with MUC2 were dominant in clusters 2 and 3, respectively. The remaining biomarkers were not differentially distributed, even though NE biomarkers prevailed in the cluster 3 (Table 1).

Figure 1

Unsupervised hierarchical clustering analysis of the IHC biomarkers used in this study. (A) Three distinct clusters were identified corresponding to the main tree branches: cluster 1 which includes all mucinous and colloid adenocarcinomas with intermingling histological subtypes, either pure or mixed; cluster 2 which comprises all ALK‐translocated and all but one KRAS‐mutated adenocarcinoma; and cluster 3 which groups all MBA. (B) A mosaic plot (the greater the surface of the different coloured rectangles, the greater the number of cases in the different tumour subsets) confirms the significance (p < 0.0001) of the differential distribution of tumours according to phenotype, and the inherent heterogeneity of tumour composition, especially in cluster 1. NA, IHC score not available.

Table 1

Immunohistochemistry data according to unsupervised clustering 1 to 3

Type of marker	Variable	Cluster 1	Cluster 2	Cluster 3	P value
Alveolar marker	TTF1	20.5 ± 28.6	88.1 ± 23.9	0	<0.0001
	MUC1	39.7 ± 36.4	93.6 ± 15.9	n.a.	<0.0001
Gastro‐intestinal marker	CDX2	37.8 ± 37.3	1.6 ± 5.0	0	<0.0001
	CK20	16.1 ± 28.5	2.3 ± 9.0	0	0.002
	MUC2	30.5 ± 34.8	0	90.0 ± 7.5	<0.0001
	MUC6	13.5 ± 16.5	0	n.a.	0.0001
Bronchial marker	MUC5AC	60.8 ± 34.1	4.4 ± 8.4	n.a.	<0.0001
Neuroendocrine marker	Synaptophysin	4.5 ± 14.3	0.8 ± 2.8	11.7 ± 31.4	0.645
	Chromogranin A	1.1 ± 14.5	0.2 ± 0.5	10.6 ± 21.9	0.223
Sex steroid hormone marker	ER	0.2 ± 0.8	1.6 ± 3.5	97.7 ± 6.6	<0.0001
	PgR	0	0	64.4 ± 30.9	<0.0001
ALK‐related marker	ALK	0	50.6 ± 48.6	0	<0.0001
	Vimentin	0	27.4 ± 28.8	31.7 ± 34.8	<0.0001
Pan‐mucinous marker	HNF4‐apha	74.5 ± 29.3	0	0	<0.0001
Pan‐epithelial marker	CK7	99.0 ± 4.1	100.0 ± 0.0	99.4 ± 1.7	0.458

n.a., not available.

P values in bold are considered statistically significant.

Analysis by sub‐clustering

Next, we further characterized the composition of cluster 1, which revealed the presence of four distinct sub‐clusters, named S1 to S4 (Figure 2). Upon histology, a mixture of colloid and mucinous/non‐mucinous patterns was seen in S4, mucinous pattern in S2 and S3 and colloid variant in S1. Moreover, floating cells with amphophilic extracellular mucin clustered in S1 only (supplementary material E; supplementary material, Figure S2).Upon IHC, differential distribution was found among sub‐clusters for TTF1, MUC1, CDX‐2, MUC2, MUC6 andHNF4A but no other biomarkers, which were thus excluded from further analysis (supplementary material F). Of note, MUC1 labelled cell membranes in sub‐cluster 3, while was confined to cytoplasm in sub‐clusters S1, S2 and S4, as in the normal lung tissue (Table 2; supplementary material F).

Figure 2

Detailed unsupervised clustering analysis of cluster 1 tumours shows four distinct and separate sub‐clusters (highlighted in different colours), namely S1, S2, S3 and S4, according to the IHC results. NA, IHC score not available.

Table 2

Histogenetic analysis by immunohistochemistry and stratification by sub‐clusters S1 to S4

Type of marker	Variable	S3 (n = 4)	P value S3 vs. S2	S2 (n = 7)	P value S2 vs. S1	S1 (n = 13)	P value S1 vs. S4	S4 (n = 4)	P value among all
Alveolar marker	TTF1	0	0.327	6.8 ± 14.9	0.259	20.6 ± 28.7	0.021	64.5 ± 32.5	0.006
	MUC1 c	0	n.a.	0	0.0001	53.1 ± 27.5 c/m	0.052	97.5 ± 39.6 c	0.0002
Intestinal marker	CDX2	0	0.023	40.0 ± 40.1	0.284	58.6 ± 31.2	0.004	4.2 ± 8.3	0.002
	MUC2	0	0.186	24.7 ± 39.8	0.073	47.8 ± 33.4	0.186	14.9 ± 17.9	0.012
Gastric marker	MUC1 m	38.2 ± 37.0	0.009	1.0 ± 1.9	n.a.	0	n.a.	0	0.0004
	MUC6	15.3 ± 6.7	0.154	30.0 ± 19.3	0.013	6.6 ± 11.1	0.042	0	0.009
Pan‐mucinous marker	HFN4‐apha	88.7 ± 11.8	0.214	96.8 ± 3.4	0.152	71.2 ± 30.2	0.030	32.3 ± 10.4	0.018
MUC IHC in mucin	MUC1 positive	2	0.576	2	0.062	10	0.541	4	0.054
	MUC1 negative	2		5		3		0

m, membrane labelling; c, cytoplasmic labelling; c/m, cytoplasmic and membranous; n.a., not applicable.

P values in bold are considered statistically significant.

The six biomarkers were then used for re‐classifying IMA and ICA, regardless of histology. According to the distribution of alveolar and gastrointestinal biomarkers, S3 and S4 sub‐clusters were placed at both ends and S1 and S2 sub‐clusters in the middle of a physio‐anatomical unit extending from bronchioles up to alveolar cells (supplementary material, Figure S3). Briefly, sub‐cluster S3 featured exclusive gastric differentiation (membrane MUC1, MUC6), S2 admixed gastric and intestinal lineage, S1 admixed intestinal and alveolar biomarkers and S4 escalation of alveolar traits with striking downregulation of intestinal biomarkers (Table 2). Pan‐mucinous HNF4A tracer significantly decreased from S3/S2/S1 to S4 sub‐cluster, whereas MUC1‐reactiveextra‐cellular mucin accumulation increased from S3/S2 sub‐clusters (4/11 cases, 36.4%) to S1/S4 sub‐clusters (14/17 cases, 82.4%) (Table 2). HNF4A correlated inversely with TTF1 (r = −0.839, p < 0.0001) and MUC1 (r = −0.504, p = 0.006) and directly with CDX‐2 (r = 0.462, p = 0.013) and MUC6 (r = 0.523, p = 0.004), whilst HNF4A exceeded MUC5AC median values (90.5% vs. 72%, p = 0.026). In turn, CDX‐2 associated directly to MUC2 expression (r = 0.876, p < 0.001) and inversely with TTF1 (r = −0.368, p = 0.054). Regarding tumour staging, there was a tendency for S1 and S4 sub‐clusters to show pT1a early tumours (9/17 cases: 53%) as opposite to S1 and S4 sub‐clusters (2/11 cases: 18%) (p = 0.115). Representative pictures of sub‐clusters are shown in supplementary material, Figures S4–S7. For the sake of comparison, pictures from KRASA and ALKA morphology and IHC findings are also provided (supplementary material, Figures S8 and S9).

Analysis by TNGS

Finally, we performed next‐generation sequencing of the same cohort of tumours to further look at the genetic alterations underpinning the different histological subtypes and molecular sub‐clusters. There were 22 mutated out of 28 analysed tumours (78%) totalling 32 mutations (1.45 mutation/tumour patient). The most frequent mutation turned out KRAS (16 cases) followed by LKB1 (4 cases), TP53 (4 cases), APC (2 cases), CDKN2A (2 cases) and HFN1A, c‐KIT, SMAD4 and SMO in 1 case each, regardless of histology (supplementary material, Figure S10).

There were no preferential types of KRAS mutation. The rate for any mutation was 50% (2/4)in S3, 57% (4/7) in S2, 54% (7/13) in S1, 100% (4/4) in S4 and 100% (7/7) in KRASA (p = 0.021) (supplementary material F). No differences emerged between trans‐versions (17/32 mutations) and transitions (15/32 mutations), irrespective of genes. ALKA were not subjected to TNGS analysis.

Discussion

In our study, we noticed that the visceral endoderm HFN4A master gene in Nkx2‐1/TTF1‐downregulating and KRAS‐mutated tumours 10, 26, 41 could be placed at the top of a developmental scheme (Figure 3), where HFN4A acted as inducer of mucinous differentiation via MUC5AC trans‐activation 26, 41, 42. MUC5AC was in turn significantly lower than HNF4A 41, while is normally limited to bronchiolar goblet cells. The lack of recognizable precursor lesions (bronchial columnar cell dysplasia 43 or mucinogenic growth clusters 44, 45), suggested that alternative non‐linear or de novo mechanisms were likely to play some role in tumour development since the reported genotypic and phenotypic profile of these precursors 44, 45 adapted to only a subset of lesions in our study. The subsequent tumour evolution depended on the relative activation state of Nkx2‐1/TTF1, which basically causes mucin repression in terminal respiratory unit cells 26, 46. The inverse relationship of HNF4A with Nkx2‐1/TTF1 and its direct correlation with CDX‐2 47 and MUC6 were likely to induce either mucinous gastric differentiation of antral gland/mucopeptic cell type 10, 26, 48 or secretory intestinal lineage 49. CDX‐2 in turn closely associated to MUC2 expression 50 and, marginally, was alternative to Nkx2‐1/TTF1, in keeping with the finding of TTF1 and CDX‐2 co‐expression even at the level of individual tumour cells. Depolarized MUC1 decoration we documented in varying proportions has been linked to cell detachment 51, apoptosis impairment 52 and cancer cell proliferation 53.

Figure 3

Operative flow‐chart showing the development of lung adenocarcinomas with mucin production from different stem/reserve cell niches distributed along the terminal and respiratory bronchioles up to alveolar cells. A new terminology was then devised on the basis of differential phenotypic combinations of the S1 → S4 sub‐clusters of cluster 1 and cluster 2 tumours. Profiles were considered negative if staining was completely absent in the relevant cells; focally (±) immunoreactive cases exhibited immunostaining in 1–10% neoplastic cells; + cases in 11–50% neoplastic cells; ++ cases in 50% or more neoplastic cells. The prefix ‘juxta’ stands for ‘near to’.

This peculiar and non‐random distribution of master regulators and activation products made us hypothesize that diversely regulated niches or domains of mucinous‐committed stem/progenitor cancer cells existed in bronchioles, in keeping with lung ontogenesis, as suggested by the congeries of tumours in cluster 1. These tumours recapitulated two alternative biological properties: increasing architectural complexity (cellular phase) and mucin escalation (secretion phase), both phenomena being under the control of common genetic traits. When the cellular phase predominated, multifocal tumours with more advanced stage developed, which featured antral gland/foveolar gastric (membrane MUC1, MUC6) and/or intestinal (CDX‐2, MU2) but nil or negligible alveolar (TTF1, cytoplasmic MUC1) cell lineage. When gastric differentiation prevailed upon complete TTF1 silencing (S3 sub‐cluster), there were multiple microscopic neoplastic foci of well‐differentiated and organoid cells with membrane MUC1‐related cell detachment, lepidic growth, tufting, stromal invasion and spread through air spaces. These cells preserved the secretion polarity and exhibited basophilic mucin, negligible extravasation and uncommon freely floating elements. When combined gastrointestinal trans‐differentiation occurred (S2 sub‐cluster) upon HNF4A‐induced CDX‐2 and MUC2 activation 50, 54 while persisting Nkx2‐1/TTF1 downregulation, multiple microscopic neoplastic foci of less differentiated tumour cells developed alongside a greater architectural complexity revealed by incipient enteric features, less preserved secretion polarity, amphophilic mucin extravasation and freely floating tumour cells. Mitoses and apoptotic bodies were undetectable in S3 but not enteric foci of S2 sub‐cluster, indicating that apoptosis blockage 27 and cell proliferation diversely accounted for the growth of these tumours (data not shown). S3 and S2 sub‐clusters were strikingly homologous to the gastric and pancreatobiliary subtypes of intra‐ductal papillary mucinous neoplasms (IPMN) of the pancreas, respectively, inasmuch as the gastric subtype is often multifocal and the pancreatobiliary subtype shows CDX2 expression 55, in keeping also with a common embryologic origin from the foregut.

When the secretion phase predominated, mass‐forming and localized tumours developed with either an exorbitant and grossly appreciable extracellular mucin pouring or microscopic mucin remnants encasing mucinous epithelium. If intestinal differentiation prevailed upon HNF4A‐induced CDX‐2 upregulation (S1 sub‐cluster), there were MUC1‐positive amphophilic mucin lakes encasing freely floating elements, less architectural complexity, colloid‐type stromal dissection and negligible tumour cell proliferation if not in less differentiated enteric foci (data not shown).These tumours sometimes co‐expressed CDX‐2 and TTF1 at the level of individual cells. This developmental scheme was strikingly similar to the intestinal subtype of pancreatic IPMN, which usually forms large cystic lesions involving the main duct 55. If alveolar differentiation prevailed (S4 sub‐cluster), mass‐forming lung‐type tumours were encountered with solid appearance, less differentiated enteric foci and MUC1‐positive amphophilic mucin remnants encasing freely floating elements. NE cell component was virtually limited to these two sub‐clusters only intestinal differentiation as hallmarked by garland‐like architecture, eosinophilic cytoplasm, terminal bars and dirty necrosis crossed S1, S2, S4 but not S3 sub‐cluster. These observations would suggest that IEA are phenotypic changes only in lung IMA and ICA in keeping with current defining criteria 1, which are likely to be related to pancreatic rather than colorectal cancer 56.

Overall, we organized our results stepwise according to six different categories as highlighted in Figure 4 (MBA were excluded because they were completely unrelated tumours).

Figure 4

Spatial distribution of different stem/progenitor cell domains along bronchiolar and alveolar structures according to clustering analysis results. Cellular mucinous adenocarcinoma (CMA) belonging to sub‐cluster S3 showed exclusive gastric differentiation (GD) and developed from distal terminal bronchioles (red cells), while secreting mucinous adenocarcinoma (SMA) corresponding to sub‐cluster S3 with gastro‐intestinal differentiation originated from proximal respiratory bronchioles (light green cells). Intestinal differentiation (ID) peaked in mucin lake‐forming colloid adenocarcinoma (MLFCA) of intermediate respiratory bronchioles (pink cells), whereas alveolar differentiation (AD) steadily increased from S2/S1 to S4 sub‐clusters, with colloid alveolar adenocarcinoma (CALA) arising from distal respiratory bronchioles (yellow cells), and peaked in KRAS‐mutated non‐mucinous adenocarcinoma (KNMA) and ALK‐translocated adenocarcinoma (ATA). KNMA and ATA were thought to derive from stem/progenitor cells located in alveolar structures near terminal bronchioles (light blue cells) or inside alveolar cells (dark blue cells), respectively. In turn, ID was only residual in CALA and KNMA and completely disappeared in ATA. The normal bronchial epithelium comprises ciliated and non‐ciliated cells (light grey), goblet cells (emerald green) and Clara cells (orange cells), while alveolar structures are covered by type I (flat grey cells) and type II pneumocytes (forest green).

S3 sub‐cluster tumours showed gastric‐type differentiation 16 and more advanced stage likely related to impaired apoptosis. These tumours were thought to derive from distal terminal bronchioles and termed cellular mucinous adenocarcinoma.

S2 sub‐cluster tumours maintained antral gland‐type MUC6 as seen in incomplete intestinal metaplasia 57, whilst CDX‐2 recruitment alongside negligible Nkx2‐1/TTF1 activity caused a more complete intestinal differentiation to appear 58 in less differentiated and more advanced tumours. This tumour category was thought to derive from proximal respiratory bronchioles and was termed secreting mucinous adenocarcinoma.

S1 sub‐cluster tumours showed predominance of secreting intestinal (CDX‐2/MUC2) differentiation responsible for inverted secretion polarity with freely floating cells, simplified architecture and less advanced tumours. These tumours were thought to derive from intermediate respiratory bronchioles and were termed mucin lake‐forming colloid adenocarcinoma.

S4 sub‐cluster tumours showed overwhelming alveolar lineage recruitment, with minor mucin secretion, less floating cells and less advanced tumour stage, mucinous epithelium or enteric features were seen. These tumours were thought to derive from distal respiratory bronchioles and were termed colloid alveolar adenocarcinoma.

Non‐mucinous KRASA featured Nkx2‐1/TTF1 and MUC1 expression 59, while repressed any secreting gastro‐enteric lineage 26, 27. These KRASA with poor secretion were thought to derive from the most distal or juxta‐alveolar region of the respiratory bronchiole and were termed KRAS‐mutated non‐mucinous adenocarcinoma.

ALK‐translocated tumours silenced intestinal‐type mucinous differentiation and activated depolarized MUC1‐depending cell disaggregation and vimentin‐related epithelial‐mesenchymal transition vimentin accumulation via Nkx2‐1/TTF1 and oncogenic ALK recruitment, respectively 59, 60, 61. These tumours were thought to derive from proximal, juxta‐bronchial progenitors of alveolar structures and were simply defined as ALK‐translocated adenocarcinoma.

No synchronous or metachronous metastatic tumours were available for analysis in terms of inter‐tumour heterogeneity, but the relative homogeneity of sub‐clusters we observed made marked differences in genotypic or phenotypic traits quite unlikely to occur in this tumour setting. The figure of 79% for any identified gene mutation was in keeping with the literature 3, 6, 15, 19, 20, but studies on CD74‐NRG1 fusion of KRAS‐negative tumours have been planned. Limitations of this the study were the relatively small number in each tumour category and the lack of validation sets, yet these lung tumours are in absolute terms quite rare and ALKA, KRASA and MBA were indeed validation control groups. The lack of survival analysis and the use of targeted NGS rather than whole exome/transcriptome analysis may have limited drawing definite conclusions. However, strength points were the identification of interpretative biomarkers in accordance with normal lung tissue distribution, multiple and non‐random significant differences across the diverse tumour categories, a close genotype‐phenotype relationship in the relevant tumour categories and coherence of results with current concepts on the development of MLA and tissue‐specific adult stem cells in human lung 62.

Conclusion

Despite improvements in lung cancer classification, IMA, ICA, IEA and, to some extent, ALKA and non‐mucinous KRASA continue to blur with each other for concurrent unanswered phenomena. In this scenario, six IHC biomarkers and simple molecular findings allowed an effective reappraisal of IMA, ICA, IEA, ALKA and non‐mucinous KRASA according to spatially organized and histology‐independent tumour categories.

Author contributions

GP conceived, designed and wrote the manuscript, reviewed all cases and performed immunohistochemistry analysis; AS collected materials and clinical data and contributed to write the manuscript; FB carried out cluster analysis, supervised all statistical procedures and contributed to write the manuscript; AF, MC, GR, AC, BV and UP collected materials and clinical data; ET, FP, AB and IC micro‐dissected samples and performed molecular investigations; GP, AS, FB, AF and BV drafted the manuscript; GP, AS, FB, AF, MC, GR, AC, ET, FP, AB, IC, BV and UP revised critically the manuscript. GP finalized the manuscript. All authors approved the submitted version.

SUPPLEMENTARY MATERIAL ONLINE

Clinico‐pathological details for the entire series of 44 lung adenocarcinoma samples under evaluation

Click here for additional data file.

Monoclonal and polyclonal antibody list used in the current study

Click here for additional data file.

Immunohistochemistry procedures and distribution of immunohistochemistry biomarkers in normal lung tissue

Click here for additional data file.

Immunohistochemistry details according to conventional subtyping of adenocarcinoma

Click here for additional data file.

Clinico‐pathological data by unsupervised sub‐clustering S1 to S4

Click here for additional data file.

Immunohistochemistry and molecular data by unsupervised sub‐clustering S1 to S4

Click here for additional data file.

Supplementary Figure Legends

Click here for additional data file.

Figure S1. Distribution of TTF1 (A,B), MUC5AC (C), MUC1 (D) and HNF4A (E,F) in normal lung tissue. The expression of TTF1 is confined to type II pneumocytes (A,B) and luminal/basal cells of respiratory bronchioles (B), while terminal bronchioles are negative (A, on the left). MUC5AC is found in goblet cells only of terminal bronchioles (C), while the respiratory extremity is unreactive (C, inset). MUC1 labels the apex of luminal cells in respiratory bronchioles (D), whilst it stains the cytoplasm in type II pneumocytes (D, inset). HNF4A faintly decorates luminal cells of respiratory bronchioles (E), but stains terminal bronchioles only focally (F) and never alveolar cells (E,F).

Click here for additional data file.

Figure S2. Relationship between sub‐clusters S1→ S4 and TTF1, MUC1, CDX‐2, MUC2 and MUC6 immunostaining findings resulting in gastric (MUC1 membranous and MUC6), intestinal (MUC2 and CDX‐2) or alveolar (TTF1 and MUC1 cytoplasmic/membranous) lineages.

Click here for additional data file.

Figure S3. Relationship between sub‐clusters S1→ S4 as defined by immunohistochemistry findings and histological classification of mucinous and colloid adenocarcinoma, either pure or mixed. This also includes one KRAS‐mutated case, which was thus reclassified into cluster 1.

Click here for additional data file.

Figure S4. Mucin lake‐forming colloid adenocarcinoma (sub‐cluster S1) showing the prevalence of the secreting phase to form mucin lakes over the cellular phase composed of columnar cells with goblet (A) and floating (A, small inset) cell appearance and prominent enteric differentiation (A, inset). Tumour cells reacted for CDX‐2 (B) and TTF1 also at the level of the same individual cells (B, inset), alongside MUC2 (C) and predominantly cytoplasmic decoration for MUC1, which also decorated secreted mucin (D). These findings indicated predominant intestinal differentiation with mucin‐lake formation and concurrent alveolar lineage.

Click here for additional data file.

Figure S5. Secreting mucinous adenocarcinoma (sub‐cluster S2) exhibiting escalation of the secretion phase with minor (≤50%) extracellular mucin accumulation at the expense of the cellular phase. These tumours featured tall columnar mucinous cells resembling gastric foveolar epithelium as seen in S3 (A), which focally also exhibited enteric differentiation with dirty necrosis (A, inset). Neoplastic elements reacted with MUC6 (B), CDX‐2 (C) and were basically unreactive (D) or, only erratically, positive for TTF1 (D, inset). These findings indicated the presence of combined gastro‐intestinal differentiation of pulmonary type (on the basis of the focal TTF1 positivity).

Click here for additional data file.

Figure S6. Cellular mucinous adenocarcinoma (sub‐cluster S3) presenting with a prevalence of tall columnar mucinous cells (cellular phase), also arranged in tufted cells (A, inset), which looked like gastric foveolar or endo‐cervical epithelium (as in S2), but with minimal extracellular mucin secretion (A). Tumour cells showed striking membrane labelling for MUC1, with apical‐like quality (B), marked MUC6 (C) and diffuse HNF4A (D) reactivity. In turn, CDX‐2 (C, inset) and TTF1 (D, inset) were completely unreactive. This profile indicated gastric mucinous differentiation in this tumour subset.

Click here for additional data file.

Figure S7. Alveolar colloid adenocarcinoma (sub‐cluster S4) showing prevalence of alveolar over mucin‐related lineage in the form of concurrent pneumocytic differentiation (A, on the left) and still moderate extracellular mucin production encasing floating cells (A, on the right) sustained by concurrent mucinous epithelium (A inset, on the left). Neoplastic cells showed prevalent TTF1 decoration (B), also in mucinous cells (B, inset), over CDX‐2 (C), MUC2 (D) and HNF4A reactivity (D, inset). This profile was consistent with predominant alveolar differentiation alongside mucin secretion to realize broncho‐alveolar cell lineage.

Click here for additional data file.

Figure S8. Non mucinous KRAS‐mutated adenocarcinoma exhibiting acinar‐cribriform pattern with comedo‐like necrosis (A) or lepidic pattern (A, inset). Neoplastic elements were diffusely positive for TTF1 (B), cytoplasmic MUC1 (C) and, focally and faintly, CDX‐2 (D). This profile was consistent with alveolar differentiation encasing residual intestinal differentiation, indicative of a different progenitor from ALK‐rearranged adenocarcinoma.

Click here for additional data file.

Figure S9. ALK‐rearranged adenocarcinoma featuring micro‐papillary tumour aggregates encased in abundant extracellular amphophilic mucin (A), also showing signet ring cells in turn engulfed by basophilic mucin (A, inset). Tumour cells were diffusely positive for either ALK fusion protein (B) or gene rearrangement upon fluorescence in situ hybridization (B, inset), and reacted strongly in the cytoplasm for MUC1 (C) and vimentin (D). This profile was consistent with alveolar differentiation alongside epithelial‐mesenchymal transition, as seen in ALK‐rearranged adenocarcinoma.

Click here for additional data file.

Figure S10. Mutation analysis by targeted next generation sequencing and direct Sanger sequencing indicating that KRAS mutations were largely prevalent across cluster 1 including mucinous and colloid adenocarcinoma, either pure or mixed, and partially in cluster 2 tumours with known KRAS mutation.

Click here for additional data file.