Invasive Mucinous Adenocarcinomas with Spatially Separate Lung Lesions: Analysis of Clonal Relationship by Comparative Molecular Profiling

Abstract

Introduction:

Pulmonary invasive mucinous adenocarcinomas (IMA) commonly present with spatially separate lung lesions. Clonal relationship between such lesions, particularly those involving contralateral lobes, is not well established. Here, we used comparative genomic profiling to address this question.

Methods:

Patients with genomic analysis performed on 2 IMA located in different lung regions were identified. Molecular assays included DNA-based next-generation sequencing (NGS) for 410–468 genes (MSK-IMPACT), RNA-based NGS for 62 genes (MSK-Fusion), or non-NGS assays.

Results:

Comparative genomic profiling was performed on 2 separate IMAs in 24 patients, of whom 19 had contralateral lesions. Tumors from all but one patient shared matching driver alterations, including KRAS (n=19), NRG1 (n=2), ERBB2 (n=1) or BRAF (n=1). In addition, in patients with paired tumors profiled by NGS (n=12), shared driver alterations were accompanied by up to 4 (average 2.6) other identical mutations, further supporting the clonal relationship between the tumors. Only in a single patient separate IMAs harbored entirely non-overlapping mutation profiles, supporting clonally unrelated, distinct primary tumors. Notably, in a subset of patients (n=3), molecular testing confirmed a clonal relationship between the original resected IMAs and subsequent contralateral IMA presenting after an extremely long latency (8.1–11.7 years).

Conclusions:

Comparative molecular profiling supports that nearly all separate pulmonary IMA lesions represent intrapulmonary spread arising from a single tumor and documents a subset with a remarkably protracted course of intrapulmonary progression. This study reinforces the unique biology and clinical behavior of IMAs while further highlighting the value of genomic testing for clarifying the clonal relationship between multiple lung carcinomas.

Introduction

Pulmonary invasive mucinous adenocarcinoma (IMA) – formerly mucinous bronchioalveolar carcinoma – is a highly distinct subtype of lung adenocarcinoma, representing approximately 3–5% of all lung adenocarcinomas^1–3. Histologically, the tumor cells are characterized by columnar morphology with abundant mucinous or eosinophilic cytoplasm and small basally located nuclei. These tumors frequently have abundant extracellular mucin and tend to grow along the alveolar walls in a so-called lepidic pattern⁴. Genomically, IMAs are characterized by activating KRAS mutations in the majority of cases and frequent NRG1 fusions in the remainder^5–9. Clinically, IMAs commonly arise in never or light smokers. While these tumors may present as solitary lesions, they have a tendency for multifocality which may involve different lobes^10–13. In addition, some patients have a diffuse, consolidative presentation resembling lobar or multilobar pneumonia^14–19. In contrast to non-mucinous carcinomas, nodal and distant metastases are rare⁵.

Biologically, lung carcinomas with spatially separate lesions comprise two distinct processes: one is separate, clonally unrelated primary tumors and the other is intrapulmonary spread of a single tumor, also referred to as intrapulmonary metastasis (IPM). This distinction is critical as it directly impacts tumor staging, prognosis, and treatment decisions. Specifically, separate primary tumors are staged individually, whereas IPMs are staged as T3, T4, or M1a for separate lesions involving same lobe, ipsilateral different lobe, or contralateral lobe, respectively²⁰. Recent studies have highlighted the clinical utility of genomic profiling, particularly next-generation sequencing (NGS), for establishing clonal relationships between lung carcinomas^21–25. However, these genomic efforts have been primarily focused on lung adenocarcinomas with non-mucinous histology, which represent the predominant type of lung carcinomas. In a recent study from our group, comprehensive NGS was applied to assess clonal relationships among a large set of lung adenocarcinomas; however, IMAs were entirely excluded given their distinctive clinicopathological features and their unique propensity to form additional lung lesions²⁵.

It is generally presumed that IMAs with consolidative, pneumonic lesions represent intra-pulmonary spread of a single tumor⁹. However, this has not been confirmed by molecular methods, especially for lesions involving different lung lobes. In addition, clonal relationship between IMAs presenting as discrete, oligo-nodular lesions has not been studied in detail. Here, we present the results of comparative genomic analysis performed to clarify the clonal relationship of IMAs with spatially separate lesions.

Materials and Methods

Study Design

Patients for whom genomic profiling was performed on 2 IMAs during a 14-year period (2007–2020) were identified. The study was approved by the institutional review board at Memorial Sloan Kettering Cancer Center (MSKCC).

Histologic, Radiographic, and Clinical Review

The diagnosis of IMA was confirmed by two pathologists (SY, NR) using the 2015 World Health Organization (WHO) classification⁴. Clinicopathologic and radiographic data were used to exclude metastasis from non-pulmonary mucinous tumors. Computed tomography (CT) of the chest was reviewed, and the lesions were classified as follows: (i) nodular lesion - a rounded or solid opacity with a clear, well-defined periphery; and (ii) pneumonic lesion - a diffuse consolidation that is involving at least the majority of a single lobe^26,27. Metachronous tumors from same patient were enumerated by the chronological order in which they were sampled (e.g., T1 and T2). Clinical information and outcomes data were obtained by review of the electronic medical records.

Genetic Testing

Tumor specimens were analyzed as part of prospective clinical care using a variety of clinically-validated molecular assays in the Molecular Diagnostics Service clinical laboratory as follows: (i) Sanger sequencing of KRAS codons 12 and 13, (ii) mutation screening for 92 mutations in 8 genes, by Matrix-Assisted Laser Desorption/Ionization-Time Of Flight (MALDI-TOF) mass spectrometry (MS) (Sequenom, San Diego, CA) (Supplemental Table 1)^28,29, and/or (iii) targeted DNA and RNA-based next-generation sequencing (NGS) using MSK-IMPACT³⁰ and MSK-Fusion³¹, respectively. In brief, MSK-IMPACT is a hybridization capture-based NGS assay that sequences the entire exons and select introns of 410–468 genes (Supplemental Table 2). MSK-Fusion is an RNA-based NGS assay that utilizes Anchored Multiplex PCR technology for targeted sequencing of 62 genes (Supplemental Table 3), which is performed on all tumors lacking a mitogenic driver by MSK-IMPACT. For cases that were negative for drivers by non-NGS assays, a high-sensitivity locked nucleic acid (LNA)-PCR sequencing was subsequently performed for enhanced detection of KRAS mutations, as previously described^32,33. DNA and RNA extractions were performed using standard operating procedures on formalin-fixed paraffin-embedded tissue on tumor specimens with matched blood normal control for MSK-IMPACT.

Clonality Assessment

Somatic mutations and structural variants from paired tumors were compared to determine their clonal relationship as previously described²⁵. In brief, pairs with 2 or more shared alterations by NGS were considered definitely clonal, whereas pairs with entirely non-overlapping, unique mutational profiles by NGS, including distinct driver alterations, were classified non-clonal (separate primaries). Tumors sharing only 1 hotspot driver alteration were classified as likely clonal (see Discussion). As described previously²⁵, manual review of the sequence reads was performed for all tumor pairs to confirm the presence (in clonal pairs) or absence (in non-clonal pairs) of shared mutations at low variant allele frequencies (VAF) (i.e., <5% VAF). In addition, the cases were reviewed manually for silent coding mutations and intronic mutations, and those with high quality read support were included for comparative analysis (Supplemental Table 4).

For tumors classified as clonal, we determined the odds of a full set of shared mutations occurring by chance. This probability of chance co-occurrence was calculated by using IMA-specific mutation prevalence derived from an internal database of IMAs that were sequenced using MSK-IMPACT (n=136)^8,34.

Results

Patient and tumor characteristics

We identified 24 patients in whom 2 IMAs, located in different lung regions, were analyzed molecularly. The patient and tumor characteristics are summarized in Table 1. Among the 24 tumor pairs, 11 presented synchronously and 13 metachronously. Nineteen tumor pairs were contralateral, 3 involved an ipsilateral different lobe and 2 were located in the same lobe (the latter were all anatomically separate, discrete nodules). Radiologically, tumor pairs presented as discrete nodules (n=18), pneumonic infiltrate plus an isolated contralateral nodule (n=3) or contralateral pneumonic infiltrates (n=3).

Table 1:

Summary of patient and tumor characteristics

Patient characteristics	N=24 patients
Age (years)
Mean	64
Range	30–81
Gender
Female	15 (63%)
Male	9 (37%)
Pack years
Mean	13
Range	0–55
Tumor pair characteristics	N=24 pairs
Timeline of presentation
Synchronous	11
Metachronous	13
Radiologic presentation of tumor pairs
Both nodules	18 (75%)
Contralateral	13
Ipsilateral different lobe	3
Ipsilateral same lobe	2
Pneumonic infiltrate and contralateral nodule	3 (13%)
Both pneumonic infiltrates (contralateral)	3 (13%)
Specimen type of tumor pairs
Both resections	9 (38%)
Resection + biopsy	9 (38%)
Both biopsies	6 (25%)

Molecular testing methods

The cohort included a total of 48 individual tumors, of which 29 (60%) were tested by MSK-IMPACT followed by MSK-Fusion if no driver alteration was identified (Tables 2 and 3). The remaining 19 (40%) cases were analyzed using non-NGS assays, including Sanger sequencing for KRAS mutations (n=6) or MALDI-TOF MS for hotspot mutations in major NSCLC drivers (n=13). With this approach, 13 tumor pairs from individual patients (54%) had both tumors sequenced using NGS, 8 pairs (33%) had both tumors profiled using non-NGS assays, and 3 pairs (13%) were tested using a combination of NGS and non-NGS assays.

Table 2:

Clinicopathologic and molecular features of individual tumors

Pt	Age (yrs)	Sex	Tumor	Presentation and latency (yrs)	Radiology	Size (cm)	Molecular method	MAPK driver	Additional shared alterations (for NGS)	Chance co-occurrence
1	61	F	T1	Metachronous 1.1 (C)	Nodule	1.8	MALDI-TOF MS#	KRAS G12D	N/A	0.055
			T2		Nodule	1.7	IMPACT	KRAS G12D
2	60	M	T1	Metachronous 2.2 (C)	Nodule	2	MALDI-TOF MS#	KRAS G12D	N/A	0.055
			T2		Nodule	2.5	IMPACT	KRAS G12D
3	30	M	T1	Metachronous 11.7 (C)	Nodule	3.2	IMPACT	KRAS G12D	CDKN2A H83Y, TP53 R248Q	1.5E-09
			T2		Nodule	2.8	IMPACT	KRAS G12D
4	61	F	T1	Synchronous (C)	Nodule	3.2	IMPACT	KRAS G12D	ARID1A M1361Cfs*120	3.0E-06
			T2		Nodule	2	IMPACT	KRAS G12D
5	77	M	T1	Synchronous (C)	Nodule	0.8	Sanger	KRAS G12D	N/A	0.055
			T2		Nodule	3	Sanger	KRAS G12D
6	74	F	T1	Synchronous (C)	Pneumonic	6.8	MALDI-TOF MS	KRAS G12D	N/A	0.055
			T2		Pneumonic	8.4	MALDI-TOF MS	KRAS G12D
7	46	M	T1	Synchronous (C)	Nodule	3.6	MALDI-TOF MS	KRAS G12D	N/A	0.055
			T2		Pneumonic	7.2	MALDI-TOF MS	KRAS G12D
8*	69	F	T1	Synchronous (S)	Nodule	3.2	IMPACT	KRAS G12D	N/A	N/A
			T2		Nodule	1.6	IMPACT	KRAS G12A
9	66	F	T1	Synchronous (C)	Pneumonic	5.2	IMPACT	KRAS G12V	SOX9 H169Pfs*83, RPTOR V971I, RAF1 K367Nfs*8, CDKN2A R80Pfs*65	4.7E-19
			T2		Pneumonic	5.8	IMPACT	KRAS G12V
10	70	F	T1	Synchronous (C)	Nodule	1.3	IMPACT	KRAS G12V	SMARCB1 R377C, SMAD4 N107Kfs*3, FYN G275S	8.7E-15
			T2		Nodule	1.2	IMPACT	KRAS G12V
11	66	F	T1	Metachronous 2 (ID)	Nodule	3	IMPACT	KRAS G12V	TP53 R175H, NOTCH3 A1044T, KMT2C N3388S	3.5E-14
			T2		Nodule	1.2	IMPACT	KRAS G12V
12	66	M	T1	Synchronous (C)	Pneumonic	5.8	MALDI-TOF MS	KRAS G12V	N/A	0.055
			T2		Nodule	3.4	MALDI-TOF MS	KRAS G12V
13	65	F	T1	Metachronous 8.1 (C)	Nodule	1.4	Sanger	KRAS G12V	N/A	0.055
			T2		Nodule	0.7	Sanger	KRAS G12V
14	69	F	T1	Metachronous 8.6 (C)	Nodule	3	MALDI-TOF MS	KRAS G12V	N/A	0.055
			T2		Nodule	2	MALDI-TOF MS#	KRAS G12V
15	61	M	T1	Metachronous 9.1 (S)	Nodule	1.2	MALDI-TOF MS	KRAS G12V	N/A	0.055
			T2		Nodule	0.8	IMPACT	KRAS G12V
16	75	M	T1	Metachronous 1 (C)	Nodule	3.4	IMPACT	KRAS G12C	ARID1A G276Efs* 87, BMPR1A P465A, KMT2A I1132V, NKX2–1 Q234L	1.2E-19
			T2		Nodule	4.7	IMPACT	KRAS G12C
17	68	F	T1	Metachronous 1.4 (ID)	Nodule	5.6	IMPACT	KRAS G12C	SDHB S144N	7.5E-07
			T2		Nodule	1.5	IMPACT	KRAS G12C
18	56	F	T1	Metachronous 2.9 (C)	Nodule	4.5	MALDI-TOF MS	KRAS G12C	N/A	0.014
			T2		Nodule	1.1	MALDI-TOF MS	KRAS G12C
19	64	F	T1	Metachronous 3.6 (C)	Nodule	2	Sanger#	KRAS G12A	N/A	0.0026
			T2		Nodule	1.1	Sanger	KRAS G12A
20	59	F	T1	Synchronous (C)	Nodule	1	IMPACT	KRAS G12R	PRKD1 A137T, ERBB4 T731M	2.5E-12
			T2		Nodule	1.6	IMPACT	KRAS G12R
21	81	M	T1	Metachronous 1.1 (C)	Pneumonic	15.5	IMPACT, Fusion	CD74-NRG1	CTNNB1 N387K, SMARCA4 D362N, STK11 c.374+220_c.863–1del, NOTCH3-CNTNAP4	4.2E-21
			T2		Nodule	1.9	IMPACT, Fusion	CD74-NRG1
22	74	M	T1	Metachronous 2.4 (ID)	Nodule	7.8	IMPACT, Fusion	F11R-NRG1	MAP3K1 c.2179+27A>G	2.9E-09
			T2		Nodule	1.2	IMPACT, Fusion	F11R-NRG1
23	73	F	T1	Synchronous (C)	Pneumonic	9.8	IMPACT	BRAF K483E	STK11 Q37*, NKX2–1 K214_Y215delinsN, RBM10 X630_splice	8.5E-18
			T2		Pneumonic	9.3	IMPACT	BRAF K483E
24	48	F	T1	Synchronous (C)	Nodule	3.2	IMPACT	ERBB2 V658_V659insR	SMAD4 D355Y, RBM10 R251*, NBN V210I	8.5E-18
			T2		Nodule	0.8	IMPACT	ERBB2 V658_V659insR

Pt=Patient; F=Female; M=Male; Yrs=Years; N/A=Not applicable (refers to comparisons where one or both tumors were tested by non-NGS methods, and therefore status of non-driver genes is not known); C=Contralateral; ID=Ipsilateral different lobe; S=Same lobe; IMPACT=MSK-IMPACT; FUSION=MSK-Fusion;

^#Initial testing resulted in a false-negative result; the mutation was detected only after using a locked nucleic acid-PCR sequencing.

^*Indicates the sole patient with clonally-unrelated tumors

Table 3:

Summary of molecular results

Molecular testing methods
Individual tumors	N=48 tumors
NGS assays (MSK-IMPACT, MSK-Fusion)	29 (60%)
Non-NGS assays	19 (40%)
Tumor pairs	N=24 pairs
Both tumors tested by NGS assays	13 (54%)
Both tumors tested by non-NGS assays	8 (33%)
NGS and non-NGS assays	3 (13%)
Genetic alterations and clonality assessment
Mitogenic driver alteration per patient	N=24 tumor pairs
Matching KRAS mutations	19 (79%)
Matching NRG1 fusions	2 (8%)
Matching BRAF mutations	1 (4%)
Matching ERBB2 insertions	1 (4%)
Each tumor with different KRAS mutation	1 (4%)
Number of additional shared alterations in pairs with matching drivers by NGS	N=12 pairs
Range	1–4
Mean	2.6
Clonal relationship*	N=24 pairs
Definitely clonal	12 (50%)
Likely clonal	11 (46%)
Definitely non-clonal (separate primaries)	1 (4%)
Probability of chance co-occurrence of shared alterations for pairs interpreted as clonally-related	N=23 pairs
Pairs with ≥2 shared alterations (definitely clonal)	N=12
Median	2.2 × 10−14
Range	4.2 × 10−21 − 3.0 ×10−6
Pairs with a shared driver alteration only (likely clonal)	N=11
Median	0.055
Range	0.0026 – 0.055

^*As detailed in the Methods and Results, “definitely clonal” defined as sharing ≥2 mutations, “likely clonal” defined as sharing a single hotspot mutation, and “non-clonal” defined as having distinct driver mutations and lacking any shared mutations.

Landscape of somatic alterations and tumor pair clonality assessment

Molecular testing of 48 individual tumors revealed canonical mitogenic driver alterations and other pathogenic variants predicted to activate the MAPK pathway in all cases (Figure 1). To assess tumor clonal relationships, we compared the genetic profiles of tumors in individual patients. This revealed that with single exception, all tumors in individual patients shared identical driver alterations: KRAS mutations (n=19), NRG1 fusions (n=2), BRAF p.K483E mutation (n=1) and ERBB2 in-frame insertion p.V658_V659insR (n=1). Of 12 tumor pairs with shared drivers analyzed by NGS, all pairs harbored between 1 to 4 (average 2.6) additional shared mutations. Full mutational data are provided in Supplemental Table 4.

Figure 1: Summary of genomic findings.

The top rows show the patient number and the individual tumor designations as T1 and T2. ^#Initial testing by Sanger sequencing or MALDI-TOF MS resulted in a false-negative result; the mutation was detected only after re-testing with a locked nucleic acid-PCR sequencing. *Additional shared alterations other than the shared pathogenic MAPK alterations as provided in the panel above.

In 4 samples (each from a different patient), initial testing with MALDI-TOF MS (n=3) or Sanger sequencing (n=1) was negative for mutations, while the other tumors in these patients were positive for a KRAS mutation. However, subsequent re-testing using high-sensitivity LNA-PCR revealed pathogenic KRAS mutations in all negative samples. On histologic re-review, tumor cell content was low to borderline (<20% of all nucleated cells) due the abundance of extra-cellular mucin and admixed inflammation.

Based on the approach outlined in the Methods, IMAs with ≥2 shared somatic alterations were classified as definitely clonal (n=12; 50%), tumors with a single shared alteration as likely clonal (n=11; 46%) and a single tumor pair with entirely unique profiles in each tumor lacking any overlapping alterations as non-clonal (4%).

Next, for pairs classified as clonal, we measured the level of support for the clonality classifications based on probability of chance co-occurrence of a full set of shared alterations using IMA-specific prevalence of each alteration (see Methods; Figure 1 and Supplementary Table 5). For tumors with ≥2 shared alterations (n=12; all for paired NGS assays), the probability of chance co-occurrence of a full set of alterations was virtually nil (median: 2.2 × 10⁻¹⁴, range: 4.2 × 10⁻²¹ – 3.0 × 10⁻⁶). For tumors with only a single shared alteration (n=11; all non-NGS assays), the probability of chance co-occurrence ranged from 0.0026 – 0.055 (0.055 representing a 1 in 18 chance).

In the sole patient with distinct driver mutations, one tumor harbored a KRAS p.G12D and the other, a KRAS p.G12A. In addition, one tumor harbored an additional unique mutation (PDCD1 c.76+114G>A) and the other harbored 2 additional unique mutations (NKX2–1 p.Q350Hfs*88 and IGF1R p.R753Q). Manual review of sequencing reads did not identify any overlap in mutations even at a sub-threshold level (<5% VAF). This patient was a 69-year-old never-smoker with synchronous IMAs (3.2 cm and 1.6 cm) presenting as 2 discrete nodules in the same lobe. There was no evidence of underlying interstitial lung disease or known cancer predisposition syndrome.

Representative radiologic and histologic findings in the analyzed tumor pairs are illustrated in Figure 2. Although the clonally unrelated pair showed some histologic differences (Figure 2E), morphologic heterogeneity was also seen in clonally related tumors (see Figure 2C).

Figure 2: Radiologic, histologic, and genomic findings for select tumor pairs that were analyzed using next-generation sequencing.

A-D: Tumor pairs classified as clonally related. E: The sole patient in this study with tumor pair classified as clonally unrelated (separate primary IMAs). Histologic sections (20x objective) and the molecular results are provided below. The IMA-specific prevalence is provided for each genetic alteration in parenthesis. Probability of chance co-occurrence is provided for pairs with shared genetic alterations (A-D).

Patient presentation and progression: The description of a subset of clonally related IMAs with an extremely protracted course of intrapulmonary spread.

Next, we reviewed the initial radiologic presentation and subsequent progression of patients with clonally related IMAs (n=23). Figure 3A and Figure 3B summarize the chronological relationship of the index lesions that were genotyped as well as other pulmonary lesions in patient with metachronous and synchronous presentation, respectively. Clinical course of the patients was highly heterogenous, with the presence of pneumonic infiltrates being significantly associated with poor overall survival (Log-rank test, p=0.0007; Figure 3C). Despite confirmed intrapulmonary spread, none of the patients had nodal or distant metastases on presentation or during the available follow-up.

Figure 3: Radiologic presentation and the timeline of progression of clonally-related IMAs.

Progression timeline for patients with metachronous tumors (A) and synchronous tumors (B), and (C) overall survival based on the presence of pneumonic infiltrates. Each row represents individual patients and the course of their disease. The horizontal axis represents time in years. Shading legend: red death of disease (DD), gray death of other causes (DOC), blue no evidence of disease (NED) and yellow alive with disease (AWD). T1 and T2 refer to index lesions that underwent molecular analysis. Superscript values represent additional lesions identified radiologically.

Notably, among the patients with clonally related metachronous lesions, we identified a subgroup with an unusually long time course of intrapulmonary progression. This included 3 patients (P3, P13, P14) in whom resection of the initial IMA was followed by the development of an isolated contralateral IMA 8.1 – 11.7 years later. In one additional patient (P15), an isolated IMA nodule developed after 9.1 years in same lobe away from the resection margin of the prior wedge resection. Clinicopathologically, all subsequent tumors in these 4 patients were favored to represent unrelated new primary tumors given the long time interval. However, molecular profiling supported their clonal relationship. For the patient with an 11.7-year latency between the tumors (P3), paired NGS revealed shared KRAS G12D mutation plus 2 additional shared identical mutations, providing definite support for clonal recurrence (Figure 2C). While the three remaining tumor pairs did not have paired NGS testing on both tumors, they were found to share a KRAS G12V, suggesting that these are clonal (see Discussion). Even after resection of the recurrent lesions, all 4 patients have remained free of disease during up to 4.8 years of follow-up. In another patient (P19), isolated recurrent IMA developed 3.6 years after resection of the initial IMA, and the patient remained disease-free for 8.4 years of follow-up after resection of that lesion. Altogether, the 5 aforementioned patients were thus alive with no evidence of disease 11.9 to 16.4 years after initial IMA resection despite developing molecularly-supported, limited intrapulmonary recurrence.

Discussion

This is the first comprehensive molecular study to date to examine clonal relationships between spatially separate IMA lesions. By performing comparative genomic analysis, we found that in all but one of the 24 patients in this study, these lesions represented intrapulmonary spread of a single tumor despite the contralateral location in the majority of lesions. In addition, we describe remarkable instances of extremely protracted clinical course of molecularly-supported intrapulmonary progression in a subset of patients.

The first major finding in this study is the confirmation that pneumonic IMAs involving different lobes, including contralateral sites, represent intrapulmonary spread of a single tumor. While single lobe involvement by pneumonic IMA is generally presumed to represent the spread of a single tumor, the clonal nature of pneumonic IMAs involving different lobes is a novel observation. The presumption of the clonal nature of pneumonic IMAs in a single lobe has been supported by the histologic hallmark of these tumors to form loosely-interconnected skipping foci surrounding the main tumor, which are thought to arise through aerogenous spread of tumor cells, recently described as spread through air spaces (STAS)^14–17. It can be hypothesized that multi-lobar disease, including contralateral spread of IMA, is propagated via floating mucin through major airways and across the carina, in line with the clinical symptom of bronchorrhea in some patients with IMA^18,19.

The second major finding relates to IMAs presenting as discrete, isolated synchronous or metachronous nodules, which comprised the majority of cases in this study. Here we demonstrate that such lesions, with only single exception, also represent intrapulmonary spread of the same tumor. Intrapulmonary spread manifesting as a single additional nodule or oligo-nodular disease, in some cases involving contralateral lobes, is not a well-recognized pattern of IMA progression. Here, using molecular methods, we confirm that this pattern in most cases represents a limited, localized form of IMA intrapulmonary spread.

Notably, here we document a remarkable subset of IMA patients (n=4) in whom intrapulmonary spread manifested after an exceptionally long latency (8.1 to 11.7 years). In these patients, initial IMA resection was followed by development of a single new IMA lesion in a different location. For all these patients, a new primary tumor was strongly favored based on clinicoradiologic grounds, given the (i) extremely long latency between tumors, (ii) solitary nature of the second nodule, and (iii) contralateral location in 3 of these patients. In sharp contrast to the clinicopathologic impression, genomic profiling supported tumors being clonal in all cases. We speculate that these remarkable cases represent IMAs with slow growth rate and limited aerogenous tumor spread, leading to the protracted formation of isolated additional nodules. All tumors with such progression harbored KRAS mutations and the initial tumors were fully resected nodules; however, specific genomic or clinicopathologic predictors of such slow-progressive disease are not clear from the available data. To our knowledge, this is the first documentation using molecular techniques of instances of ultra-late IMA recurrence. Importantly, this finding suggests that patients with IMAs may benefit from extended clinical surveillance for recurrence after surgical resection.

In this study, we identified a single outlier case of a patient with two unrelated primary IMAs that presented as synchronous separate nodules in the same lobe. These tumors had entirely non-overlapping mutational profiles with different KRAS drivers (G12D vs. G12A) and additional mutations that were unique to each tumor. Given that the prevalence of IMA is low (~5% of all lung adenocarcinomas), the odds of 2 independent IMAs arising in a single patient by chance is low but not nil. It is known that underlying fibrosing interstitial lung disease (ILD) may predispose to formation of multiple IMAs^35,36. However, no background ILD was identified pathologically or on imaging studies for this patient. Yet, the possibility that this patient may have an occult risk factor or subclinical predisposition for developing multiple unrelated IMAs cannot be excluded. Overall, although our data show that nearly all cases of multiple IMAs represent intrapulmonary spread of a single tumor, there are exceptions, and these cases can be clarified only by comparative genomic profiling.

Overall, the molecular landscape of IMAs with spatially separate lesions in our series is consistent with findings from prior studies on the distribution of driver alterations in IMAs in general, which includes KRAS mutations in the major of cases and NRG1 fusions in a minority^5–9. We also found activating BRAF mutation p.K483E³⁷ and an in-frame insertion in the ERBB2 transmembrane domain (p.V658_V659insR) that are likely oncogenic^8,38,39. Our results indicate that intrapulmonary spread can occur in the context of various driver alterations.

Our study highlights several specific technical challenges for comparative molecular analysis of IMAs. First is related to the low tumor cellularity that is frequently encountered in IMAs. This is typically due to the abundance of mucin, stromal cells, and admixed inflammation in the tumor microenvironment. Thus, IMAs are particularly prone to false-negative results with lower-sensitivity assays such as Sanger sequencing and other multiplexed small gene panels. In our study, false-negative results by initial non-NGS assay led to a false impression of distinct genomic profiles in 4 patients. Yet, on re-testing with higher-sensitivity LNA-PCR sequencing, KRAS mutations were identified in all samples that were negative by lower-sensitivity methods.

Another challenge relates to the use of single gene assays or small hotspot panels. Given the high prevalence of KRAS G12D mutations in IMAs (23.5%)⁸, the odds of a patient with 2 unrelated, separate primary IMAs harboring KRAS G12D entirely by chance is 5.5% (or 1 in 18). Thus, in our series, analysis by non-NGS assays provided classification only as “likely” clonal for shared prevalent KRAS mutations, whereas paired analysis by broad panel NGS detected multiple shared mutations in all clonally related cases, thus providing a definitive confirmation of clonal relationship.

Despite the clear advantages of NGS for comparative tumor profiling, a unique consideration for IMAs is their overall low prevalence among lung adenocarcinomas (3–5%). Thus, the pre-test probability of second IMA representing a related tumor rather than a new primary is high. Given this consideration and in the context of our findings showing that additional IMA lesions are in almost all cases clonal, we propose that in the setting of multiple IMAs, shared single driver alteration may represent sufficient evidence to support tumor clonality. We note that this contrasts sharply with multiple non-mucinous carcinomas in smokers, which in our recent study had a high chance of incidentally sharing the same KRAS (usually G12C) driver mutation²⁵.

This represents the first comprehensive study to date in which genomic approaches were used to establish clonal relationship in a large series of IMAs with separate pulmonary lesions. Although a number of studies have applied clinical NGS and other sequencing methods to determine clonal relationship between multiple lung cancers^21–25, IMAs have been either excluded by study design²⁵, or only a few cases were profiled given their low prevalence^40,41. Notably, one prior study has suggested independent origin of different areas of pneumonic IMAs based on discordant TP53 mutations (1/1 patient)⁴², and another found discrepant results from KRAS Sanger sequencing (1/2 patients)⁴¹. Here, we used an extended panel of >400 genes, assays with high analytical sensitivity and a strict definition of clonal relationship to show that multiple IMAs are usually clonally related. Interestingly, among the 12 patients that were classified as definitely clonal by comparative NGS, 4 patients had tumor pairs where one tumor (e.g., T1 or T2) harbored at least one private, non-driver mutation that was not present in the other tumor. The presence of these private mutations does not contradict our assessment of clonality, but rather reflects the concept of branched tumor evolution in which clonal “truncal” mutations are accompanied by divergent “branch” mutations^8,43.

Review of the overall presentation and progression in IMA patients with molecularly confirmed intrapulmonary spread in our series highlighted several unique features of this disease. First, we confirm that even in patients with confirmed intrapulmonary spread, IMAs rarely develop nodal or distant metastases, consistent with prior reports^6,11. Second, we observed a wide spectrum of disease presentations and clinical outcomes, ranging from extensive to limited lung involvement, including the aforementioned subset with a remarkably slow course of pulmonary spread. Overall, the prognosis of patients with IMAs has been a controversial subject with studies showing a wide range of clinical behaviors^5,11,44–46. While our study was not designed to investigate prognostic factors in IMA, we noted that among this cohort of patients with confirmed intra-pulmonary spread, the presence of pneumonic infiltrates was associated with poor outcome, as also suggested in prior studies^10,12,26,47.

Our results may help optimize the clinical management of patients with IMAs, who may be candidates for molecular targeted therapies, including investigational agents against tumors with NRG1, KRAS, and ERBB2 alterations^8,48. While these alterations were detected in some cases, patients in our cohort did not receive targeted therapies given the relative recency of these investigational treatments. Despite this, our findings offer important insights for future therapy selection and suggest that for patients with unresectable multifocal IMAs, driver alteration at a single site is in most cases representative of other lesions; however, confirmatory testing of an additional site could be considered in candidates for systemic targeted therapies.

Although intrapulmonary “metastasis” is currently the standard terminology to describe intrapulmonary spread of lung carcinomas²⁰, metastatic disease generally implies lymphatic or hematogenous dissemination. Given that for IMAs aerogenous spread is likely the predominant mechanism of dissemination though the lung, “intrapulmonary spread” may be more appropriate terminology, particularly given the apparent indolent nature of limited intrapulmonary progression in some patients.

In conclusion, this study reinforces the concept that IMA represents a unique variant of lung adenocarcinoma unified by a distinct molecular profile and a propensity for intrapulmonary progression. While speculated in the past, our study, for the first time, provides a comprehensive molecular evidence supporting that intrapulmonary spread represents the predominant mechanism of pneumonic interlobar spread and formation of additional isolated nodules in IMAs, including tumors that were suspected to represent separate primaries on clinical grounds. In addition, our data expand on the utility of broad NGS-based testing for clonality assessment of multiple lung carcinomas.