Using molecular characteristics to distinguish multiple primary lung cancers and intrapulmonary metastases

Patient selection

We included 35 patients having more than one lung carcinoma in the study. All patients visited the Fourth Hospital of Hebei Medical University between December 2014 and April 2020 and sufficient specimens were obtained for histologic and molecular analysis. Patients who received neoadjuvant therapies and those with extrathoracic metastasis were excluded from the study. All experimental plans and protocols were submitted to the ethics/licensing committees of the participating hospitals for thorough review and approval before conducting the retrospective study. The protocols were approved by the Fourth Hospital of Hebei Medical University (2018MEC133).

Sample preparation, targeted NGS, and genetic analysis

All tumor tissues and matched blood samples were tested for genomic targeting based on NGS of 450 genes (Cancer Sequencing YS panel, CSYS; OrigiMed, Shanghai, China; (Cao et al., 2019)). Formalin-fixed paraffin-embedded (FFPE) tumor tissues and matched blood samples were collected from the Fourth Hospital of Hebei Medical University. As previously described in Cao et al. (2019), genetic testing is performed in Origimed laboratory, which is accredited both by CAP (College of American Pathologists) and CLIA (Clinical Laboratory Improvement Amendments). The histologic subtyping was reviewed by independent pathologists with 10 years of experience from Origimed. The percentage of tumor cells in each sample was ≥20%. Approximately 50 ng DNA was extracted from 40-mm FFPE tumor tissues and blood samples using the QIAamp DNA FFPE Tissue Kit (Qiagen, Hilden, Germany) for subsequent genetic analysis. All the coding exons of 450 cancer-related genes and the selected introns of the parts of the targeted genes that frequently rearranged in solid tumors were captured using the custom hybridization capture panel, and DNA sequencing libraries were constructed. The libraries were sequenced on an Illumina NextSeq-500 Platform (Illumina Incorporated, San Diego, CA, US). The sequencing depth mean coverage was 1,000× and 300× for FFPE and matched blood samples, respectively. The CSYS panel was used to determine exon sequences encoding all of the 450 genes and the TERT promoter and introns of 39 genes (Cao et al., 2019). The identified somatic alterations contained mutations and copy number changes. MuTect, Pindel, and EXCAVATOR were used to identify SNVs, InDels, and CNVs, respectively. A definite SNV/InDel required a minimum of five unique supporting reads. An in-house developed algorithm was used to screen gene rearrangements. All variants were manually viewed on Integrative Genomics Viewer to avoid misreporting (Cao et al., 2019). We emphasized that the genes included in the atlas were variants with identified or potentially important functional changes as annotated by the OncoKB and ClinVar databases (Chakravarty et al., 2017; Landrum et al., 2020). In addition, we excluded gene variants with unknown functional changes, no predicted important functional changes, or unreported undetermined changes to avoid inter-observer inconsistencies in the interpretation of gene mutations.

Study design

Our retrospective cohort study included patients with MPLC and IPM adenocarcinoma. The primary inclusion criteria were adults aged ≥18 with comprehensive clinicopathologic information and confirmed diagnosis of lung adenocarcinoma based on imaging and pathologic analysis. Sufficient tumor tissue was obtained from multiple lesions for NGS sequencing. Patients with incomplete information, lung metastases other than lung cancer, and the absence of tissue or control samples for patients with primary exclusion criteria included suspicion.

Finally, 35 patients with complete sample details and clinical information (21 MPLC and 14 IPM) were included in the study. The training and validation cohorts were established in the chronologic order of enrolment in a 2:1 ratio to avoid bias. Twenty-two patients were enrolled in the training cohort, and the remaining 13 patients were included in the validation cohort. The determination of MPLC or IPM by clinicians was based on the IASLC guidelines for the classification of lung cancers with multiple pulmonary sites of involvement (before conducting genetic tests). We analyzed the imaging findings, anatomical evidence, pathologic findings, histologic examination of the primary tumor, and histologic subtypes to differentiate MPLC from IPM. Two pathologists independently examined all biopsies, and consensus was achieved for all samples. We used the Martini & Melamed criteria to classify these samples into MPLC and IPM. Synchronous tumors located in different segments but showing similar histologic types without lymphatic or systemic metastases were classified as MPLC. In addition, samples with different histologic types, at different segmental or lobar locations, with different growth rates and biomarker patterns, and without lymph node or lung cancer metastases were classified as MPLC. Heterochronous tumors were classified as MPLC when the time interval between tumors was >2 years. Such tumors were classified as IPM when the time interval between tumors was <2 years and both tumors were located in the same lobe or when the tumor interval was <2 years and the tumors were located in different lobes of the lungs with lymph node involvement or systemic metastases. Doctors assessed the available evidence and interpreted whether the tumors were primary lesions or primary tumors with metastasis.

Statistical analysis

Student’s t-test was used to compare two groups, whereas ANOVA and post hoc tests (mainly the Bonferroni test) were used to compare three or more groups. The Chi-square and Fisher tests were used to determine the significance of rates or percentages when required. A p-value of less than 0.05 was considered statistically significant.

Clinical characteristics of patients

Patients with >2 lung lesions were divided into training (n = 22) and validation (n = 13) cohorts. The median age of the training cohort was 56.5 years (range: 43–77 years) and that of the validation cohort was 57 years (range: 27–75 years). The proportion of females was 40.9% and 53.8% in the training and validation cohorts, respectively, and the proportion of patients with adenocarcinomas was 91.7% and 100% in the training and validation cohorts, respectively. Twenty-nine patients had two tumors, five patients had three tumors, and one patient had four tumors; therefore, 77 individual tumors were analyzed. Adenocarcinoma (n = 45), invasive adenocarcinoma (IAC; n = 3), adenocarcinoma in situ (AIS; n = 10) and minimally invasive adenocarcinoma (MIA; n = 6) were the main pathologic types. Among the 77 tumors, 17, 16, 15, and 25 tumors were stage I, II, III, and IV tumors, respectively. Twenty-two patients in the training cohort included nine patients with MPLC and 13 with IPM, and 13 patients in the validation cohort included eight patients with MPLC and five with IPM (Table 1).

Establishing the criteria for distinguishing MPLC from IPM

NGS was performed to detect mutational alterations in all lesions to explore strategies for differentiating MPLC and IPM. Overall, 510 mutations were detected in 48 tumors of 22 patients from the training cohort. The genes with high mutational frequency were EGFR (54%), TP53 (50%), MUC16 (22%), KRAS (17%), LRP1B (17%), ERBB2 (15%), RBM10 (15%), OBSCN (13%) and PIK3CA (13%), and the majority of mutations observed were substitution/deletion mutations (Fig. 1A). The substitution/indel and gene amplification of EGFR mutations (e.g., EGFR p. (E709A; G719S), EGFR p. T790M, EGFR p. E746_A750del, p. L858R, and EGFR amplification) were observed in six patients. Ten tumors from these six patients had the EGFR p. L858R mutation. Gene fusion was only found in ROS1, NTRK3, and RET among 22 patients. ROS1 fusion was observed in both the tumors of one patient. NTRK3 fusion was observed in one tumor, and RET fusion was found in one tumor from two patients (Table S1; ROSi fusion was observed in both the tumors of one patientNTRK3 fusion was observed in one tumor, and RET fusion wasfound in one tumor from two patients). Eight patients (36.4%) showed inconsistent driver mutations and unique mutation profiles for each tumor, and no mutation was detected in a tumor lesion of one patient (4.5%). The remaining 13 patients (59.1%) had one or more mutational alterations in each tumor (Table S1).

The Martini & Melamed criteria and previously proposed molecular identification methods (Mansuet-Lupo et al., 2019; Pei et al., 2021; Wang et al., 2020) were used to diagnose patients in the training cohort (Table 2). We found that the diagnoses of 14/22 (63.6%) patients were consistent with all previously published criteria. These 14 patients included 10 patients with MPLC, which had one or no common mutations, and four patients with IPM, which had more than three mutations (Tables S1 and S2). Patient 2, having three tumors, was identified as MPLC using the Martini & Melamed criteria but was identified as MPLC/IPM using different molecular methods (Table 2). In this patient, tumor 1 or 3 and tumor 2 had one common mutation (NTRK1 c.288-3C > A); tumors 1 and 3 had three common mutations (NTRK1 p. G137V, NTRK1 c.288-3C > A, and PIK3CA p. K111N). Therefore, tumor 2 was the primary lesion, and tumors 1 and 3 were metastatic lesions. Patients 4 and 7, having more than three common mutations, were identified as MPLC using the Martini & Melamed criteria but were identified as IPM using molecular methods. Patients 5, 6, 8, 12, and 15 were identified using the previously proposed molecular methods but had different classifications. Patient 5 (common mutations: ERBB2 c.1022-6C > T and ROS1 fusion) and patient 8 (common mutations: EGFR p. L858R and TP53 p. W146*) were identified as IPM based on the molecular methods proposed by Mansuet-Lupo et al. (2019) (IPM can be interpreted when tumors have at least two mutations in common) and Pei et al. (2021) (IPM can be identified when EGFR p. L858R and another mutation are shared between tumors or when consistent gene hotspot mutations (EGFR, KRAS, BRAF, ERBB2, ALK, ROS1, MET, or RET) occur in tumors, but EGFR p. L858R is not shared). However, MPLC was identified in these two patients based on the molecular methods proposed by Wang et al. (2020). This classification was supported by the identification of TKI-related driver gene alterations (EGFR, ALK, ROS1, MET, RET, BRAF, and KRAS) and non-TKI-related gene alterations with the total number of common mutations ≤2. Additionally, MPLC was confirmed by pathologic analysis and molecular identification. Two tumors of patient 12 had the common mutations PIK3CG p. H199N and POLD1 p. A532T. The patient was identified as IPM based on the methods proposed by Mansuet-Lupo et al. (2019) and Pei et al. (2021). However, the same patient was interpreted as MPLC based on the classification method proposed by Wang et al. (2020). We finally identified two lesions in this patient as MPLC based on the histologic types (AIS and MIA) and the common mutations.

Mutations are divided into three categories, namely trunk, shared, and branch mutations (Wang et al., 2020). We examined the occurrence of all three types of mutations in 13 MPLC and 9 IPM specimens. The results revealed that 9/13 MPLCs exclusively had branch mutations, indicating that 69.23% of MPLC had no common mutations. This rate was similar to the previously reported rate of 66.7% (Wang et al., 2020). The remaining 4 MPLCs had <2 trunk mutations and no shared mutation, indicating a very low level of common mutations (Fig. 1B). Conversely, all IPM showed a significantly higher number of trunk or shared mutations, highlighting marked disparities among mutant components (Fig. 1C). Additionally, we demonstrated that branch mutations accounted for 98.02% of the mutations in MPLC compared with only 57.89% mutations in IPM. Trunk mutations constituted 33.08% and shared mutations accounted for 9.02% in IPM (Fig. S1A).

We created a flowchart of our criteria to distinguish MPLC from IPM based on our observations (Fig. 2). When the concordance rate is 0%, A-1) and there is no common mutation between the samples, then the patient is judged to be MPLC; A-2) and there are samples with no mutation, it is necessary to differentiate between the number of samples with no mutation; A-2.1) if all samples have no detectable mutation, then the relationship between the samples cannot be determined; A-2.2) if there are only two samples and one of the samples has no mutation and the other has a mutation, then the patient is the patient is judged to be MPLC; A-2.3) if there are >2 samples and one of the samples has no mutation and the remaining sample has a mutation, then the patient cannot be judged. When the concordance rate is >0%, B-1) and there is only one common mutation between the samples, then the patient is judged to be MPLC; B-2.1) if there are two common mutations between the samples (one of which is a driver mutation and the other is a rare mutation) then the patient is considered to be an IPM; B-2.2) if there are either two common driver mutations between the samples or one common driver mutation and one p53 mutation, then the patient is considered to be patient was MPLC; B-2.3) if it was the presence of two common rare mutations, the patient could not be determined; B-3) if there were more than three common mutations between samples or all mutations were common, the patient was considered to be IPM. Driver mutations are defined as mutations derived from EGFR, KRAS, BRAF, ERBB2, ALK, ROS1, NRAS, RET, or driver mutations in the MET gene.

Validation of interpretation criteria for the identification of MPLC and IPM

We created a validation cohort of 13 patients to validate our criteria and streamline the protocol for distinguishing MPLC and IPM. NGS was performed to detect mutational alterations in all lesions to differentiate between MPLC and IPM. Overall, 316 mutations were detected in 29 tumors. The genes with high mutational frequencies were EGFR (52%), TP53 (31%), KRAS (24%), LRP1B (24%), RBM10 (21%), CFTR (17%), and CNAS (17%). These findings were similar to those observed in the training group (Fig. 3A). Moreover, EGFR mutations (EGFR p. L858R, EGFR cn_amp, and EGFR p. E746_A750del) were observed in four patients, and eight tumors of these four patients had EGFR p. L858R mutation. RB1 fusion was observed in both tumors of one patient. The EGFR, CDK4, GNAS, NPM1, SDHA, TRIO, ZNF217, and NFKBIA gene amplifications were identified in tumor samples from paired tumors of three patients. Six patients (46.2%) showed inconsistent driver mutations and unique mutation profiles in each tumor. The other seven patients (53.8%) had one or more mutational alterations in each tumor (Table S2). Five patients were interpreted as IPM, and the remaining seven patients appeared to be MPLC (Table 3). Subsequent analysis of mutational status and genes involved supported this interpretation. Figure 3B shows that six out of eight MPLCs do not have any trunk or shared mutation but only branch mutations. Therefore, 75% of MPLC did not have any common mutations, and these findings were similar to those observed in the training cohort. The remaining two MPLCs had only one trunk mutation, indicating a very low occurrence of common mutations. In contrast, IPM had a significantly higher number of trunk mutations, indicating notable differences among mutant components (Fig. 3C). The number of trunk mutations in IPM was higher and that of branch mutations was lower compared with MPLC (Fig. S1B). Overall, these findings confirmed our criteria for distinguishing MPLC from IPM.

Comprehensive analysis of the EGFR alterations

Alterations in the EGFR gene were identified in 21/35 patients (60%) enrolled in the study i.e., 38 lesions from 21 patients showed this mutation. Sixteen MPLC lesions showed EGFR mutations. Eight of them had exon 21 L858R, two had EGFR p. A767_V769dup, and the remaining six presented with EGFR p. (E709A; G719S), EGFR p. A750_I759delinsGD, EGFR p. G719S, EGFR p. L833V, EGFR p. E421Q, EGFR p. H358L, and EGFR amplification. Twenty-three IPM lesions showed EGFR alterations. Twelve lesions showed EGFR p. L858R, eight presented with EGFR p. E746_T751delinsA, two lesions showed EGFR p. L747_T751del, four lesions showed EGFR p. T790M, and nine lesions showed EGFR amplification (Table S1). In addition, we analyzed the number of EGFR, KRAS, TP53, BRAF, ERBB2, MET, RET, NRAS, ALK, and ROS1 mutations in patients identified with MPLC and IPM and found that KRAS mutations were significantly higher in MPLC (42.86%) than in IPM (7.14%; Table S3).

Molecular methods and martini criteria for identifying MPLC and IPM

We used the Martini & Melamed criteria in addition to molecular methods to distinguish whether the ambiguous samples can be categorized as MPLC or IPM. We analyzed whether the identification results were consistent between the two modalities. The first patient was a 70-year-old Chinese woman admitted to hospital. The computed tomography (CT) examination showed multiple nodules in the right lung, and metastasis was not excluded. Two nodules were observed through the CT examination (Fig. 4A). T1 nodule, measuring 1.6 × 0.8 × 1 cm³, was observed in the apical segment of the upper lobe of the right lung, and T2 nodule, measuring 1.2 × 0.6 × 0.5 cm³, was observed in the middle lobe of the right lung. Pathologic examination of these two pulmonary nodules indicated IAC in the lung tissue. Lepidic was the dominant type in the T1 nodule, and the T2 nodule was 30% acinar and 70% lepidic. The NGS results revealed that these two lesions have only one common mutation EGFR p. L858R, indicating MPLC. The second patient, a 69-year-old Chinese lady, was initially diagnosed with only one lesion was found in the right middle lobe on CT. Later, another lesion was found on surgery, indicating metastases. T1 nodule, measuring 5 × 3.5 × 3 cm³, was observed in the middle lobe of the right lung, and T2 nodule, approximately 0.8 cm in diameter, was observed in the upper lobe of the right lung. Pathologic examination of the two pulmonary nodules indicated IAC in the lung tissue, and both nodules were 50% acinar, 30% papillary, and 20% mucinous. The NGS results revealed two common mutations CDK4 amplification and EGFR p. L858R (concordance rate >50%), indicating IPM (Fig. 4B).

In addition, another two patients were identified through both pathologic examination and molecular methods. A 77-year-old Chinese woman was diagnosed with multiple nodules in the right lung, and metastasis was not excluded. Two nodules were observed through CT examination. T1 nodule in the upper lobe of the right lung was approximately 1 cm in diameter, and T2 nodule in the lower lobe of the right lung measured 1.5 × 1 × 1 cm³. Pathologic examination of the two pulmonary nodules suggested IAC in the lung tissue. The type of T1 nodule was 80% lepidic and 20% acinar, whereas the type of T2 nodule was 80% acinar and 20% lepidic. The NGS results revealed that these two lesions had no common mutation, indicating MPLC (Fig. 5A). The fourth patient was a 66-year-old Chinese man having three nodules in the upper lobe of the right lung. The CT examination showed two nodules (T1 and T2), both measuring 1 × 0.8 × 0.5 cm³, in the upper lobe of the right lung. T3 nodule in the upper lobe of the right lung measured 1 × 1 × 0.8 cm³. Pathologic examination of the three pulmonary nodules showed IAC in the lung tissue. The type of T1 nodule was 20% lepidic and 80% acinar, the type of T2 nodule was 90% acinar and 10% lepidic, and the type of T3 nodule was 30% acinar and 70% lepidic. The NGS results revealed that these two lesions had no common mutation, indicating MPLC (Fig. 5B). The molecular identification in these four patients was consistent with the pathologic diagnosis, suggesting that our molecular identification approach can be used to distinguish MPLC from IPM in the absence of pathologic diagnosis.