Diagnostic performance of RASSF1A and SHOX2 methylation combined with EGFR mutations for differentiation between small pulmonary nodules

Abstract

Background and aim

Aberrant methylation of Ras association domain family 1, isoform A (RASSF1A), and short-stature homeobox gene 2 (SHOX2) promoters has been validated as a pair of valuable biomarkers for diagnosing early lung adenocarcinomas (LUADs). Epidermal growth factor receptor (EGFR) is the key driver mutation in lung carcinogenesis. This study aimed to investigate the aberrant promoter methylation of RASSF1A and SHOX2, and the genetic mutation of EGFR in 258 specimens of early LUADs.

Methods

We retrospectively selected 258 paraffin-embedded samples of pulmonary nodules measuring 2 cm or less in diameter and evaluated the diagnostic performance of individual biomarker assays and multiple panels between noninvasive (group 1) and invasive lesions (groups 2A and 2B). Then, we investigated the interaction between genetic and epigenetic alterations.

Results

The degree of RASSF1A and SHOX2 promoter methylation and EGFR mutation was significantly higher in invasive lesions than in noninvasive lesions. The three biomarkers distinguished between noninvasive and invasive lesions with reliable sensitivity and specificity: 60.9% sensitivity [95% confidence interval (CI) 52.41–68.78] and 80.0% specificity (95% CI 72.14–86.07). The novel panel biomarkers could further discriminate among three invasive pathological subtypes (area under the curve value > 0.6). The distribution of RASSF1A methylation and EGFR mutation was considerably exclusive in early LUAD (P = 0.002).

Conclusion

DNA methylation of RASSF1A and SHOX2 is a pair of promising biomarkers, which may be used in combination with other driver alterations, such as EGFR mutation, to support the differential diagnosis of LUADs, especially for stage I.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00432-023-04745-8.

Introduction

Lung cancer is the leading cause of cancer-related deaths worldwide, and its survival is poor, with the highest mortality rate among malignant tumors (approx. 20%) (Sung et al. 2021). Late diagnosis is largely responsible for the extremely high mortality rate associated with lung cancer. The outcomes can be remarkably better at an early-stage diagnosis, especially for stage I; the 5-year survival rates can rise from 15%–19% to 81%–85% (Begum et al. 2011; Blandin Knight et al. 2017). The National Lung Screening Trial recommended a low-dose chest computed tomography (LDCT) screening for the early detection of lung cancer, which has been shown to reduce the mortality rate by 20% (Aberle et al. 2011; Patz et al. 2014). However, a growing number of pulmonary nodules (PNs) are detected with the widespread application of LDCT, and some of these have been discerned as an indication of early-stage lung adenocarcinoma (LUAD). As yet, the essence of these accidental nodules remains unclear; also, the cytological and histological examinations based on tissue biopsy remain the gold standard (Osarogiagbon et al. 2019). However, the mainstream morphological diagnosis can be markedly influenced by specimen conditions and individual professional levels of pathologists (Cerfolio et al. 2010; Cucuruz et al. 2012).

Emerging molecular biomarkers can have the potential to provide a novel, highly specific, and objective approach to improve lung cancer diagnosis. DNA methylation is a relatively stable epigenetic alteration occurring frequently in the early stages of tumorigenesis, including lung cancer (Klutstein et al. 2016; Mari-Alexandre et al. 2017; Vizoso et al. 2015). As reported, DNA methylation occurs mainly in the promoter CpG islands (Sandoval et al. 2013) by inactivating the transcriptional factors of tumor suppressor genes and being involved in tumor formation (Esteller and Herman 2002). It can be examined noninvasively in tissues and in sputum, plasma, bronchoalveolar lavage fluid, and other body fluids, including lung cancer and other cancers (Hulbert et al. 2017). Consequently, various methylation biomarkers are progressively being designed for early cancer detection and diagnosis (Roy and Tiirikainen 2020).

The promotor methylation of the short-stature homeobox gene 2 (SHOX2) and the Ras association domain family 1, isoform A (RASSF1A) have been separately identified as promising and valuable biomarkers for lung cancer diagnosis and prognosis (Darwiche et al. 2013; Wei et al. 2015). However, the sensitivity of SHOX2 gene methylation seems to be poorer in stage I tumors and in LUAD compared to squamous cell carcinoma (SCC) and small cell lung cancer (SCLC) (Kneip et al. 2011; Mari-Alexandre et al. 2017). A majority of reports demonstrated that RASSF1A methylation might be a powerful indicator of non-small cell lung cancer prognosis (Buckingham et al. 2010; Ko et al. 2013), as it could contribute to the determination of a subset of lower-grade cancer that progressed to the more advanced and more invasive condition (Grawenda and O'Neill 2015).

The genetic variation in the epidermal growth factor receptor (EGFR) has been well recognized to engage in lung carcinogenesis. EGFR mutation–positive tumors exhibited a correlation with the growth of pulmonary ground-glass nodules (GGNs) as reported (Kobayashi et al. 2015). GGNs, hazy lesions on LDCT scans, tend to be diagnosed in stage I of LUAD. They can be characterized as two lesions: early pure GGNs and later mixed GGNs (part-solid) before developing into an invasive lesion. Once fully invasive, the GGNs often change into solid nodules, and the overall 5-year survival rate for LUAD is about 15% (Siegel et al. 2018). Although patients with preinvasive adenocarcinoma are identified and surgically excised, a long-term survival approach of nearly 100% (Kodama et al. 2001; Suzuki et al. 2002). Therefore, accurately differentiating the invasion patterns of mini-sized malignant PNs is imperative, which can considerably improve the prognosis of patients with early LUAD.

Substantial prospective studies have indicated that multiple gene or marker panels can function as an approach to overcome limited sensitivity and enhance cancer diagnostic performance (Dochez et al. 2019; Weiss et al. 2017). The methylation detection of both RASSF1A and SHOX2 combined has been proven to be an effective strategy to improve sensitivity and exhibits a much higher diagnostic potential for detecting early lung cancer, with a specificity of up to 90% and sensitivity of up to 70% (Ren et al. 2017; Zhang et al. 2017). However, the combined assay of two-gene methylation with EGFR alteration for early LUAD has hardly been reported, and their diagnostic performance in early LUAD remains unclear. The relationship between these genetic and epigenetic alterations in the multistep invasive progression of LUAD has not yet been fully understood.

In this study, we investigated the aberrant promoter methylation of RASSF1A and SHOX2, and the genetic mutation of EGFR in 258 specimens of early LUADs to measure the frequency and distribution. We attempted to assess whether combining DNA methylation biomarkers with EGFR gene detection could more efficiently distinguish between noninvasive and invasive tumors or the more detailed subtypes, including (1) noninvasive lesions, (2) minimally invasive lesions, and (3) invasive adenocarcinomas. Lastly, the interaction between the genetic modification of EGFR and the hypermethylation of candidate genes was also analyzed.

Methods

Participants and sample collection

All sample collection and procedures were approved by the ethics committee of Southern Medical University. In this study, we retrospectively collected all formalin-fixed and paraffin-embedded (FFPE) specimens of small PNs at the Department of Pathology in the Nanfang Hospital, Southern Medical University (Guangzhou, China) between January 2021 and June 2021. The inclusion criteria were as follows: (1) patients detected with PNs on computed tomography (CT) or positron emission tomography-CT scans; (2) the nodule diameter less than or equal to 2 cm; and (3) adequate and complete clinical information of participants. The exclusion criteria were as follows: (1) sample DNA undetected or methylation results invalid; (2) nodule lesions with a benign or non-adenocarcinoma histological diagnosis; and (3) paraffin-embedded sections containing less than 5% of tumor cells. The left-over FFPE resection specimens were enlisted for DNA methylation and EGFR mutation detection.

All FFPE samples were preserved for no longer than 2 years. The paraffin-embedded tissue samples results and categorization were diagnosed and verified by experienced pathologists with senior titles based on the histologic criteria for adenocarcinoma in the World Health Organization classification. The histological evaluation of HE-stained sections was performed under the microscope to ensure sufficient tumor tissue and tumor density available for pathological diagnosis and subsequent molecular gene detection.

RASSF1A and SHOX2 DNA methylation assay

The methylation statuses of RASSF1A and SHOX2 were examined by the methylation-specific PCR (MS-PCR) method, using the LungMe real-time polymerase chain reaction (PCR) kit (Tellgen Co., Shanghai, China). An FFPE DNA extraction kit (Tiangen Co., Beijing, China) was used for processing paraffin-embedded tissue samples. The concentration of isolated DNA was precisely quantified using a high-sensitivity Qubit assay kit (Qubit 4.0 Fluorometer, CA, USA). DNA bisulfite modification was performed using a Tellgen DNA purification kit (Tellgen Co.) following the manufacturer’s protocols. After purification, bisulfite-converted DNA was eluted, followed by a real-time PCR reaction. An SLAN-96S real-time PCR system (Hongshi Co., Shanghai, China) was used for PCR amplification and data analysis. The methylated RASSF1A and SHOX2 served as the detection targets, whereas ACTB served as an internal reference to quantify the input DNA. The Ct values of the fluorescent signals FAM (RASSF1A), VIC (SHOX2), and Cy5 (ACTB) were obtained separately. The relative concentrations of methylated RASSF1A and SHOX2 were determined using the △Ct method. (△Ct_SHOX2 = Ct_SHOX2 − Ct_β-ACTB; △Ct_RASSF1A = Ct_RASSF1A − Ct_β-ACTB). The results of methylation were interpreted following the manufacturer’s guidelines: RASSF1A methylation positive with Ct_FAM < 35 and △Ct_RASSF1A ≤ 12, SHOX2 methylation positive with Ct _VIC < 32 and △Ct_SHOX2 ≤ 7.5.

EGFR gene mutation analysis

Four exons (exons 18–21) in the tyrosine kinase domain of the EGFR were found to have EGFR mutations using the amplification-refractory mutation system PCR (ARMS-PCR) technique. These exons were selected due to the majority of the reported EGFR mutations located in these regions. PCR was performed as follows: Phase I: 95℃ for 10 min, 1 cycle; Phase II: 95℃ for 15 s and 60℃ for 30 s, 10 cycles; and Phase III: 93℃ for 15 s and 57℃ for 30 s, 35 cycles. The FAM and VIC signals were collected at 57 °C in Phase III to obtain the Ct values of the corresponding fluorescent signals. EGFR mutations were quantitatively evaluated by the △Ct method using VIC as the internal standard and FAM as the mutation and external standard. △Ct = (mutant tube FAM channel Ct value)—(external standard tube FAM channel Ct value). The positive criteria for specimens were judged as follows: mutation tube △Ct ≤ cutoff value, regarding the set threshold value for each mutation type to determine the mutation site result of the sample. The △Ct cutoff value of each mutation locus was as follows: 8 for exon 18 (G719X) and exon 20 (T790M), and 10 for exon 19 (DEL19), exon 20 (S768I and INS20), and exon 21 (L858R and L861Q).

Patient classification and data definition

The patients were assigned to three groups based on their histological results: patients with pathologically confirmed noninvasive tumors were classified as group 1, and patients with histopathologically confirmed minimally invasive and invasive tumors were classified as groups 2A and 2B, respectively (Fig. 1). All the demographic, radiologic, and clinicopathological characteristics were obtained from surgical and pathological records in the hospital information system. The total methylation (TM) status was defined as the methylation levels of SHOX2 plus methylation levels of RASSF1A. A positive methylation result for either RASSF1A or SHOX2 indicated a positive TM (TM +), whereas a negative methylation result for both genes indicated a negative TM (TM-). The definition of the combined testing using the methylation assay together with EGFR mutation was as follows: either a positive RASSF1A or SHOX2 methylation or EGFR mutation denoted a positive combined testing result, which was shown as “RASSF1A + EGFR + ,” “SHOX2 + EGFR + ,” and “TM + EGFR + ,” respectively. Both the two-gene methylation and EGFR mutation results were negative, and the combined testing was judged as negative.

Fig. 1.

Flowchart of the study design. Based on the histopathological diagnosis, participants with pulmonary nodules were divided into three subgroups: group 1 (preinvasive lesions), group 2A (minimally invasive lesions), and group 2B (invasive adenocarcinoma). FFPE formalin-fixed and paraffin-embedded tissues, AAH atypical adenomatous hyperplasia, AIS adenocarcinoma in situ, IA invasive adenocarcinoma, MIA minimally invasive adenocarcinoma

Statistical analysis

All statistical analyses were carried out with GraphPad Prism 8.0 software (GraphPad Software Inc., CA, USA) and SPSS 25.0 (SPSS Inc., IL, USA). The continuous clinical variables were compared using the Kruskal–Wallis H test and Mann–Whitney U test, whereas the categorical clinical variables were compared using the Pearson chi-square test or Fisher exact test. Spearman correlation analysis was used to analyze the correlation between methylation levels and EGFR gene alteration. The McNemar test and kappa agreement index were used to examine the diagnosis consistency between the two-gene detection assays. The Cochran–Mantel–Haenszel test was used to analyze the association between methylation and EGFR mutations after controlling pathological subtypes as covariates. Receiver operating characteristic (ROC) curves were designed to evaluate the diagnostic performance of the panel of multiple biomarkers among different pathological subtypes and the area under the curve (AUC). The sensitivity (Sen), specificity (Spe), positive predictive value (PPV), and negative predictive value (NPV) were also calculated. Statistical significance was assumed at a two-tailed P value of less than 0.05.

Results

Clinicopathological characteristics

We identified 258 patients that fulfilled the inclusion criteria, including 106 male (41.1%) and 152 female specimens (58.9%). The patients ranged in age from 20 to 77 years (average, 52.9 years). Approximately two-thirds of the patients were nonsmokers (75.6%), and roughly half of the nodules were GGNs (49.2%). The size of these nodules was usually within 1 cm. Among 258 patients, 125 were diagnosed with preinvasive lesions, including 2 histological subtypes: atypical adenomatous hyperplasia (AAH) and adenocarcinoma in situ (AIS). Further, 59 were diagnosed with minimally invasive adenocarcinoma (MIA), and 74 with invasive adenocarcinoma (IA). The representative morphologic features and the distribution of PNs are shown in Figs. 2 and 3A. More details of demographics and clinical characteristics of patients are described in Table 1. As the pathological subtype advanced from AAH/AIS to MIA to IA, the proportion of male patients, older age, smokers, solid component, and larger lesion size significantly increased (P < 0.05). No association was found between the nodule number, location, and pathological subtype (Table 1).

Fig. 2.

Representative examples of pathological subtypes and radiological appearance in malignant pulmonary nodules. A Morphology characteristics of pathological subtypes in early lung adenocarcinoma (LUAD) detected by H&E staining. Photographs were taken at 200 × magnification. a Atypical adenomatous hyperplasia (AAH); b adenocarcinoma in situ (AIS); c minimally invasive adenocarcinoma (MIA); and d invasive pulmonary adenocarcinoma (IPA). B Imaging features of pulmonary nodules in early LUAD with computed tomography (CT). a Pure ground-glass nodule (PGGN); b part-solid nodule (PSN); and c solid nodule (SN)

Fig. 3.

A Distribution of the histopathological subtypes of all specimens. B Distribution of EGFR mutation locus variety in 258 specimens. C Venn diagram depicting the relationship among RASSF1A, SHOX2 methylation, and EGFR mutation in early lung adenocarcinoma (n = 258)

Table 1.

Demographics and clinicopathological characteristics of patients

Clinicopathological characteristics			Noninvasive (AAH + AIS)	Minimally invasive (MIA)	Invasive (IA)	P value (χ2)	P value (Linear correlation)
N		258	125 (48.4)	59 (22.9)	74 (28.7)
Sex	Male Female	106 (41.1) 152 (58.9)	40 (32.0) 85 (68.0)	25 (42.4%) 34 (57.6%)	41 (55.4%) 33 (44.6%)	0.005**	0.001***
Age, year	Median Range	53 20–77	50 20–74	53 27–77	58 37–74	< 0.001***	< 0.001***
Smoking	Yes Never	63 (24.4) 195 (75.6)	21 (16.8) 104 (83.2)	10 (16.9) 49 (83.1)	32 (43.2) 42 (56.8)	< 0.001***	< 0.001***
Nodule number	Single Multiple	163 (63.2) 95 (36.8)	84 (67.2) 41 (32.8)	37 (62.7) 22 (37.3)	42 (56.8) 32 (43.2)	0.335	0.145
Nodule location	Left upper lobe Left lower lobe Right upper lobe Right middle lobe Right lower lobe	75 (29.1) 44 (17.1) 74 (28.7) 17 (6.6) 48 (18.6)	39 (31.2) 21 (16.8) 35 (28.0) 8 (6.4) 22 (17.6)	17 (28.8) 9 (15.3) 16 (27.1) 4 (6.8) 13 (22.0)	19 (25.7) 14 (18.9) 23 (31.1) 5 (6.8) 13 (17.6)	0.993	0.612
Nodule type	Nonsolid or ground-glass opacity (GGNs) Part-solid (PSNs) Solid or Soft tissue density (SNs) Cystic lesions	126 (49.2) 81 (31.6) 46 (18.0) 3 (1.2)	85 (68.5) 31 (25.0) 6 (4.8) 2 (1.6)	30 (50.8) 22 (37.3) 7 (11.9) 0	11 (15.1) 28 (38.4) 33 (45.2) 1 (1.4)	< 0.001***	< 0.001***
Lesion diameter (cm)	≤ 1 1.1–2	174 (67.4) 84 (32.6)	101 (80.8) 24 (19.2)	49 (83.1) 10 (16.9)	24 (32.4) 50 (67.6)	< 0.001***	< 0.001***
Nodule size (cm)	≤ 1 1.1–2 2 <	147 (57.0) 86 (33.3) 25 (9.7)	89 (71.2) 36 (28.8) 0	42 (71.2) 16 (27.1) 1 (1.7)	16 (21.6) 34 (45.9) 24 (32.4)	< 0.001***	< 0.001***

AAH atypical adenomatous hyperplasia, AIS adenocarcinoma in situ, IPA invasive pulmonary adenocarcinoma, MIA minimally invasive adenocarcinoma

∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001

The methylation and mutation analysis revealed that 79 cases (30.6%) were positive for RASSF1A methylation and 61 (23.6%) were positive for SHOX2 methylation. The TM positivity of both genes was 41.5% (107/258). EGFR mutation type accounted for 38.8% (100/258) compared with the wild type (61.2%, 158/258). The relationships between clinicopathological features and RASSF1A, SHOX2, TM, and EGFR mutation status were separately analyzed, and are shown in more detail in Tables S1 and S2. The patients with a positive RASSF1A or positive SHOX2 methylation result, or those in the TM + group, were characterized by older age, smokers, part-solid or solid nodules, and larger nodule size (> 1 cm) (P < 0.05). EGFR mutation was associated with several clinicopathological features such as age (P < 0.001), lesion diameter (P = 0.002), nodule size, and solid components (P < 0.001, P = 0.004). However, EGFR mutation was not correlated with other clinicopathological characteristics such as sex and smoking status (Table S1).

Diagnostic efficacy of the three biomarkers and combined assay for identifying post-invasive PNs

We analyzed the diagnostic performance of the individual RASSF1A, SHOX2 methylation, EGFR mutation, and several combined assay panels in FFPE between groups 1 and 2 (Table 2). The ROC curve analysis was performed in our study (Fig. 4). Regarding the identification of the post-invasive lesions, individual RASSF1A promoter methylation was marginally more sensitive than individual SHOX2 or EGFR mutations (43.61 vs 35.34 or 34.59, respectively). Generally, the diagnostic accuracy was greatly improved for the multi-gene assay panels. The RASSF1A and SHOX2 combined promoter methylation showed a slightly higher AUC value of 0.7011 (95% CI 0.6375–0.7647), compared with individual RASSF1A (AUC value: 0.6635, 95%CI 0.5972–0.7299) and SHOX2 assays (AUC value: 0.6533, 95% CI 0.5866–0.7200). Besides, the sensitivity was enhanced by the combined EGFR mutation detection compared with the combined assay of RASSF1A and SHOX2 methylation. Notably, for the RASSF1A and SHOX2 methylation panel with EGFR mutation, the AUC value was the highest at 0.7294 (95% CI: 0.6674–0.7913), with a sensitivity of 60.9% and a specificity of 80.0%. Therefore, it suggested that the three biomarker panels could be an effective tool for discriminating invasive lesions from noninvasive lesions.

Table 2.

Diagnostic efficacy of different markers and combined assay panels in identifying groups 1 and 2

	AUC		Sensitivity% (95% CI)	Specificity% (95% CI)	PPV	NPV
	Value	95% CI
RASSF1A	0.6635	0.5972–0.7299	43.61 (35.48–52.10)	88.0 (81.14–92.59)	0.747	0.586
SHOX2	0.6533	0.5866–0.7200	35.34 (27.73–43.77)	88.0 (81.14–92.59)	0.754	0.558
EGFR	0.6102	0.5413–0.6792	34.59 (27.04–43.00)	88.0 (81.14–92.59)	0.660	0.576
TM (RASSF1A + SHOX2)	0.7011	0.6375–0.7647	43.61 (35.48–52.10)	88.0 (81.14–92.59)	0.738	0.642
			51.88 (43.46–60.20)	80.0 (72.14–86.07)
RASSF1A + EGFR	0.7022	0.6382–0.7662	40.60 (32.63–49.10)	88.0 (81.14–92.59)	0.706	0.697
			52.63 (44.19–60.92)	80.0 (72.14–86.07)
SHOX2 + EGFR	0.6864	0.6216–0.7513	43.61 (35.48–52.10)	88.0 (81.14–92.59)	0.681	0.667
			54.89 (45.67–62.37)	80.0 (72.14–86.07)
TM + EGFR	0.7294	0.6674–0.7913	48.12 (39.80–56.54)	88.0 (81.14–92.59)	0.690	0.748
			60.90 (52.41–68.78)	80.0 (72.14–86.07)

Fig. 4.

Receiver operating characteristic (ROC) curve analysis of individual RASSF1A, SHOX2, EGFR, and multiple gene panel detection. A Individual gene assay; and B multi-gene panel detection. The area under the ROC curve (AUC) for each combination shows the higher predictive power for identifying invasive lesions in early lung adenocarcinoma (group 1 vs group 2)

Subtype analysis of RASSF1A, SHOX2, and TM level in early LUAD

The methylation statuses of RASSF1A and SHOX2 were quantified by real-time PCR according to the delta cycle threshold (△Ct) method. According to the previous research, the LungMe assay approved the reliable diagnosis of methylated RASSF1A and SHOX2 DNA with the methylation cutoff values of △Ct_RASSF1A = 12 and △Ct_SHOX2 = 7.5, allowing for higher sensitivity (76.6%) and specificity (92.1%) of SHOX2 in FFPE compared with sensitivity and specificity of RASSF1A (61.3–95.6%, respectively) (Shi et al. 2020). Based on this criterion, the methylation level of RASSF1A and SHOX2 was displayed as 12.1—△Ct_RASSF1A and 7.6—△Ct_SHOX2, respectively, to illustrate the variation in methylation level more evidently. The quantitative RASSF1A and SHOX2 methylation levels in the three groups were assayed and presented in Fig. 5. We further investigated the distribution and association of RASSF1A, SHOX2, and TM in different pathological subgroups. As shown in Table 3, a statistically significant difference and correlation were observed between methylation and pathological subtypes, irrespective of individual RASSF1A and SHOX2, or combined promoter methylation assays (all P < 0.001). Besides, our data showed that the methylation-positive detection rate gradually increased as preinvasive lesions progressed to minimally invasive lesions followed by IA. When combining RASSF1A with SHOX2 methylation, the positive detection rate of MIA and IA was improved to 50.8% and 66.2%, respectively. In addition, significant differences in the methylation level were determined between group 1 and group 2B (P < 0.001) and between group 1 and group 2A (P < 0.01, except for SHOX2) using paired comparisons. No statistically significant difference between group 2A and 2B was observed except for SHOX2 methylation (adjusted P value, Table S3).

Fig. 5.

Quantitative comparison of RASSF1A, SHOX2 methylation, and EGFR mutation levels in 3 groups comprising 258 specimens. To depict the difference in methylation and mutation levels more intuitively, the methylation level of RASSF1A and SHOX2 was shown as 12.1 –△Ct_RASSF1A and 7.6 –△Ct_SHOX2 respectively. The EGFR mutation status was displayed as a cutoff value-△Ct_EGFR according to corresponding mutation sites

Table 3.

Pathological subtype performance of RASSF1A, SHOX2 methylation, EGFR mutation, and combined assay panels using FFPE samples from patients with suspected malignant lung nodules

Methylation positivity (%)	Pathological subtypes			Total	P for chi-square	P for linear correlation
	Group 1 (AAH + AIS)	Group 2A (MIA)	Group 2B (IA)
N	125	59	74	258
RASSF1A	20a (16.0)	24b (40.7)	35b (47.3)	79 (30.6)	< 0.001	< 0.001
SHOX2 +	15a (12.0)	14a, b (23.7)	32b (43.2)	61 (23.6)	< 0.001	< 0.001
TM +	28a (22.4)	30b (50.8)	49b (66.2)	107 (41.5)	< 0.001	< 0.001
EGFR +	34a (34.0)	25a, b (25.0)	41b (41.0)	100 (38.9)	< 0.001	< 0.001
RASSF1A + EGFR +	40a (32.0)	38b (64.4)	58b (78.4)	136 (52.7)	< 0.001	< 0.001
SHOX2 + EGFR +	43a (34.4)	33b (55.9)	59c (79.7)	135 (52.3)	< 0.001	< 0.001
TM + EGFR +	48a (38.4)	41b (69.5)	66c (89.2)	155 (60.1)	< 0.001	< 0.001

TM +  total methylation, which refers to the combined methylation detection of both the RASSF1A and SHOX2 genes

Each subscript letter denotes a subset of pathological subtype categories whose column proportions do not differ significantly from each other at the 0.05 level. The same subscript letters represent no statistically significant difference at the 0.05 level

Subtype analysis of EGFR mutation in early LUAD

EGFR mutations were detected in 258 early LUAD FFPE samples with 100 mutations spread in 4 exons (exons 18–21), based on amplification-refractory mutation system PCR (ARMS-PCR). The EGFR mutation exon locations were as shown in Fig. 3B, including 2 G719X in exon 18, 32 DEL19 in exon 19, 3 S768I, 1 T790M, and 2 INS20 in exon 20, and 54 L858R and 1 L861Q in exon 21. Besides, four cases of multi-locus mutation were present (including one L858R + DEL19, and three L858R + T790M). Among the 100 mutated specimens, 86.8% contained either the in-frame deletion in exon 19 or L858R point in exon 21, which are the most prevalent EGFR mutations in NSCLC and are proven to predict responsiveness to the epidermal growth factor receptor–tyrosine kinase inhibitors. The aforementioned mutation sites were mostly drug sensitive except for one case of T790M point mutation and two cases of insertion mutation in exon 20, which are commonly drug resistant.

The relationship between pathological subtypes (groups 1, 2A, and 2B) and quantitative EGFR mutation status (cutoff value △Ct_EGFR) was analyzed in more detail in Table 3 and displayed in Fig. 5. As shown in Table 3, EGFR mutation was significantly associated with the pathological subtype of early LUAD (P < 0.001). The prevalence of EGFR mutation was significantly higher in invasive lesions than in noninvasive lesions, but significant differences were not found between the two invasive subgroups (group 2A vs group 2B) (P = 0.126) (Table S3). No statistically significant difference regarding diverse mutation sites with pathological characteristics was observed (P = 0.959) (Table S4).

Performance on methylation assay combined with EGFR mutation in identifying invasive subtypes in early LUAD

Although the use of the individual RASSF1A, SHOX2 methylation, and EGFR mutation showed significant differences and correlation with pathological subtypes (P < 0.001, Table 3), identifying detailed invasive subtypes of early LUAD using paired comparisons was insufficient: group 2A compared with group 2B (P = 1.000 for RASSF1A and 0.378 for EGFR) and group 1 compared with group 2A (P = 0.243 for SHOX2) (Table S3). The performance of multi-gene panels for DNA methylation assay combined with EGFR detection was separately evaluated as shown in Table 3. The data showed that the multi-gene assays remained dramatically correlated with pathological subtypes and presented a greater positivity rate than the mono-gene assays. The positive rates for preinvasive lesions (group 1) were maintained at more than 30%, while a positivity rate of more than 50% could be achieved for two groups in post-invasive lesions (groups 2A and 2B). Interestingly, when considering both SHOX2 and EGFR among the three subgroups, the post hoc analysis allowed for identifying significant differences between groups 2A and 2B (adjusted P = 0.019). As for the three-gene assay panel, significant differences were also observed using paired comparisons between group 1 and 2A or 2B (P < 0.001), while the statistical significance was borderline between groups 2A and 2B (adjusted P = 0.064) (Table S3). The aforementioned correlation transformed into the predictive power of the two kinds of multi-gene assay panels, as reflected by the area under the ROC curve for group 1 versus group 2A (AUC = 0.6153 and 0.6779), group 1 versus group 2B (AUC = 0.7542 and 0.7858), and group 2A versus group 2B (AUC = 0.6458 and 0.6489) (Fig. 6). These findings revealed that the combination of methylation biomarkers and EGFR detection could provide a viable, effective method for identifying the classification of pathological subtypes in early LUAD.

Fig. 6.

Predictive power of the two kinds of combined assay panels in the differentiation among preinvasive, minimally invasive, and fully invasive lesions. The receiver operating characteristic (ROC) curves displayed the predictive power of the SHOX2 or RASSF1A methylation combined with EGFR mutation for both groups (2A and 2B) compared with the control group 1. The area under the curve (AUC) = 0.6153 for group 1 versus 2A, 0.7542 for group 1 versus 2B, and 0.6458 for group 2A versus 2B (left). AUC = 0.6779 for group 1 versus 2A, 0.7858 for group 1 versus 2B, and 0.6489 for group 2A versus 2B (right)

Correlation of EGFR mutation with aberrant promoter methylation of RASSF1A and SHOX2 genes in early LUAD

The interrelationships among RASSF1A and SHOX2 methylation, and EGFR mutation are shown in Fig. 3C. As shown in Table 4, the distributions of RASSF1A methylation and EGFR mutation were considerably exclusive in early LUAD (P = 0.002). The aberrant methylation of RASSF1A was slightly higher in EGFR mutation samples than in the wild type. This correlation persisted in another analysis by controlling for pathological subtypes as covariates (P = 0.038). Table 5 displays that the correlation between RASSF1A methylation and EGFR mutation was only observed in noninvasive lesions (P = 0.013), and a weak fit was observed between both assays (κ = 0.351, P < 0.001). However, no association was found between the SHOX2 methylation and EGFR mutation (P = 0.478).

Table 4.

Correlation between EGFR mutation and RASSF1A and SHOX2 methylations

Methylation		EGFR mutation		P value (Chi-square)	Rho (Spearman)	P value (Correlation)	P value (CMH test)
		Mutated (%)	Wild type (%)
RASSF1A	+	42 (53.2)	37 (46.8)	0.002	0.196	0.002	0.038
	−	58 (32.4)	121 (67.6)
SHOX2	+	26 (42.6)	35 (57.4)	0.478	0.044	0.480	0.576
	−	74 (37.6)	123 (62.4)

Cochran–Mantel–Haenszel test was used for the correlation analysis after controlling the pathological subtypes as covariates

Table 5.

Correlation and consistency between EGFR mutation and RASSF1A methylations in different pathological groups

	RASSF1A methylation	EGFR		P value (McNemar)	Rho (Spearman)	P value (Correlation)	Kappa	P value (Kappa)
		Mutated (%)	Wild type (%)
Noninvasive	+	13 (65.0)	7 (35.0)	0.013	0.371	< 0.001	0.351	< 0.001
	−	21 (20.0)	84 (80.0)
Invasive	+	29 (49.2)	30 (50.8)	0.464	− 0.008	0.923	− 0.008	0.923
	−	37 (50.0)	37 (50.0)

McNemar test: paired-sample Chi-square test was used for the comparisons between two assays in different groups

Discussion

In this study, we evaluated the predictive potential of two methylation biomarkers RASSF1A and SHOX2, and EGFR gene alteration for invasive pathological subtypes of early LUADs based on 258 malignant PNs measuring 2 cm or less. DNA methylation biomarkers RASSF1A and SHOX2 in the FFPE specimens may, in conjunction with additional genomic modifications (such as EGFR mutation), contribute to the differentiation among patients with noninvasive, minimally invasive, and invasive PNs.

On the basis of morphological characteristics, we analyzed the clinicopathological parameters in different early LUAD groups. The development of early LUAD followed a linear progression spectrum (AAH–AIS–MIA–IA). The nodule size and solid component gradually increased as preinvasive lesions progressed to minimally invasive lesions, followed by IA. The larger and higher–solid density lesions seemed more likely to grow up and progress to more advanced diseases. Besides, male sex, elderly status, and smoking were often known to promote the growth of nodules and enhance the likelihood of their solid transformation (Heller et al. 2013; Kakinuma et al. 2016; Matsuguma et al. 2013). It is well established that a strong link exists between tobacco smoking and lung carcinogenesis, and that smoking accounts for roughly 90% and 70% of male and female lung cancer cases, respectively (Shopland 1995). Continued tobacco smoking and inhalation may induce the development of new lesions. According to a study with a follow-up of 5.5 years, more than 50% of smokers were observed to have an increased number of nodules, whereas the number of nodules in never-smokers remained steady (Remy-Jardin et al. 2002). Furthermore, chronic exposure to cigarette smoke triggered aberrant methylation in tumor oncogenes and/or suppressor genes. RASSF1A, considered to be a suppressor gene, was found to have higher methylation frequency in male individuals (Vaissière et al. 2009) and was correlated with smoking (Zhang et al. 2017). As an oncogene, SHOX2 plays a negative role in LUAD development; high levels of gene expression and promoter methylation were found in lung cancer tissues. The expression of SHOX2 was higher in patients with middle age and a smoking history (Li et al. 2021). In our study, SHOX2 gene hypermethylation was associated with smoking. EGFR mutation, the key driver mutation of tumorigenesis in the early stage of LUAD, has been recognized to be associated with the female sex, nonsmokers, and adenocarcinoma subtype (Pao et al. 2004; Tsao et al. 2005). However, our results were contradictory to the current understanding; they did not identify a correlation between EGFR mutation and the aforementioned clinical factors. This discrepancy might be attributed to the variations in sample size, environmental context, frequency of genetic and epigenetic alterations in lung cancer among communities investigated, and methodology. Otherwise, a positive correlation between the radiological features, such as the total size and solid size, of PNs and aforementioned molecular abnormality was observed in the study. Therefore, the promoter methylation in RASSF1A and SHOX2, as well as EGFR mutation, may exhibit a strong association in both PN growth and PN invasiveness, but their potential for the differential diagnosis of specific invasive phenotypes in early LUAD remains unclear.

Diverse molecular abnormalities and complicated evolutionary history of adenocarcinoma are presented and underscored in the sequential progression schema from preinvasive lesions into invasive adenocarcinomas (Chung et al. 2011; Selamat et al. 2011). The complex genetic mechanism has limited our knowledge of the accurately assessed characteristics, proper management, and timely intervention of early lesions with suspected PNs. Epigenetic modifications, including DNA methylation, are essential in initiating preinvasive lesions. The progressive methylated CpG islands were observed within AIS compared with AAH in an earlier study; following the linear progression schema of LUAD, the promoter methylation was further expanded (Chung et al. 2011; Selamat et al. 2011). We demonstrated that the methylation positivity rate gradually improved as the invasive components of lesions increased, which was similar to previous findings (Gao et al. 2022). RASSF1A or SHOX2 methylation assay was highly sensitive and specific for early LUAD, and the combined methylation panel further improved the diagnostic performance with the improvement in the AUC value, obtaining more than 50% methylation positivity rate of IA in our study. RASSF1A and SHOX2 methylation played crucial roles in the invasiveness and deterioration of early-stage lung cancer development. Some researchers mentioned that the promoter hypermethylation of RASSF1A was strongly associated with lung carcinogenesis and aggressive cancer phenotype (Grawenda and O'Neill 2015). Most studies so far indicated that the hypermethylation of RASSF1A and SHOX2 exhibited poor differentiation, advanced stages, and rapid progression in early-stage disease (Dubois et al. 2016; Grawenda and O'Neill 2015; Li et al. 2020; Zhao et al. 2015). According to recently reported studies, high Ki-67 expression was correlated favorably with the combined promoter methylation level of RASSF1A and SHOX2, and Ki-67 as clinically related to advanced cancer cell proliferation, invasion, and metastasis in primary lung cancer (Gao et al. 2022; Myong 2003). Additionally, the tumor cell motility and invasion were promoted by RASSF1A deletion as a result of the reduction of β-catenin and E-cadherin expression (Bao et al. 2019). These findings confirmed that RASSF1A and SHOX2 could act as promising predictors of malignant lung diseases and might have the potential to discriminate the aggressive phenotypes of early LUAD.

Typically, small PNs with better prognoses were diagnosed as early-stage LUADs, including AIS or MIA. These “relatively benign” nodules could transform into IA after sufficient accumulation of driver mutations (Chen et al. 2019). A crucial section in the transformation of a lesion is invasion. Most prior NGS studies involving early-stage LUADs concentrated on genomic evolution during invasiveness acquisition (Chen et al. 2019; Hu et al. 2019; Zhang et al. 2019). Recent research demonstrated that genomic heterogeneity and clonal evolution were discerned in the earliest stages of LUADs (Zhang et al. 2019). Heterozygosity loss was one of the earliest alterations in preinvasive adenocarcinoma; as the lesion progressed, the subclones predominated after sufficient accumulation of driver mutations. EGFR was the key driver mutation for LUAD, which was mutated across all stages of the early invasive lesions (Izumchenko et al. 2015). We demonstrated that the incidence of EGFR mutations increased as preinvasive lesions advanced, and a relatively higher mutation frequency was observed in invasive diseases. This was consistent with Yatabe’s hypothesis that EGFR-mutated tumors followed a linear developmental pathway, in which AAH proceeded to AIS and was then followed by MIA and IA (Yatabe et al. 2011). Higher mutation rates in EGFR were proved to correlate with a more invasive radiological and pathological behavior. Besides, except for the common mutational locus of L858R and DEL19, other mutations, including S768I, G719X, INS20, T790M, L861Q, and so forth, were detected in early-stage LUADs despite at a relatively low frequency. However, the pathophysiological function of these polymorphic variations is still unknown in lung carcinogenesis. Multi-locus mutations were frequently observed in IA other than noninvasive diseases. These results endorsed Qian’s and Hao Li’s findings. They found that a fairly higher mutational burden and copy number gains were observed in lung adenocarcinoma compared with MIA and AIS, indicating that advanced and progressive LUAD possessed more somatic mutations and more complicated genomic architecture and driver profile, which might be a potential molecular interpretation for their more invasive clinicopathological characteristics (Hu et al. 2021; Hua et al. 2020).

In our study, the combined detection of RASSF1A and SHOX2 methylation showed an efficient discriminative ability between noninvasive and invasive lesions. The positive result of combined methylation assays might denote aggressive tumor growth and intensive therapeutic requirement. The combined methylation detection had a specificity of 88.0% and an extremely low sensitivity of 43.61%, with an AUC value of 0.701 in diagnosing invasive lesions. Additional biomarkers could be used as a beneficial strategy to improve low sensitivity and enhance the identification of invasive nodules compared with noninvasive nodules. Combined with EGFR mutation, three-gene panel detection showed a specificity of 80.0% and a fairly high sensitivity of 60.9% with an AUC value of 0.7294. Minimally invasive adenocarcinoma, as the intermediate lesion during the invasive progression schema, is independent and entirely different from AIS and IA, especially in morphological characterization, operative management, and patient outcome. The resection of no fully invasive lesions has achieved considerable surgical outcomes with a disease survival of approximately 100% (Kodama et al. 2001; Suzuki et al. 2002). In our study, integrating the gene methylation assay and EGFR mutation demonstrated efficient discriminatory power to distinguish lung nodules with three invasive subtypes. The subtype analysis found that two kinds of panels, SHOX2 combined with EGFR panel and three-gene panel, both yielded excellent diagnostic performance, which harbored a reliable discriminatory ability between MIA and IA (AUC = 0.6458 and 0.6489, respectively). Combined with EGFR detection, the methylation panel achieved a great improvement in the positive rate of detection. Hence, the diagnostic positivity of IA nearly reached 90% in our results. This suggested that DNA methylation, in conjunction with somatic mutations, might help define early-stage LUAD pathological subtypes more objectively and accurately, contributing to the implementation of individualized treatment. Recent investigations have revealed comparable genomic and DNA methylation evolutionary paths in lung cancer (Hu et al. 2021; Hua et al. 2020) and other malignancies (Brocks et al. 2014; Mazor et al. 2015). Even at a single-cell level, DNA methylation was found to have the potential for examining cell heterogeneity and discriminating different cell types in cancers (Karemaker and Vermeulen 2018). The relationship between DNA methylation and cancer-driver gene mutation has not been elucidated in detail, and hence requires further investigation of genetic and epigenetic interactions in lung carcinogenesis.

Subsequently, we investigated the interaction and correlation between the epigenetic modification in DNA methylation and genetic alteration in EGFR mutation. Our data suggested a significant association between RASSF1A methylation and EGFR mutation. Two potential explanations for the result were as follows. On the one hand, an overlap between EGFR mutation and RASSF1A methylation was previously found in the signaling mechanism. As the prevailing paradigm suggested, DNA hypermethylation might lead to a shift in the equilibrium toward shutting some key driver genes in the development of LUAD to upregulate carcinogenic pathways such as Ras/RAF/MAPK signaling and transform cells into malignant cells (Cheung and Nguyen 2015). As a member of the Ras family, the RASSF1A gene harbored the Ras association domain, which was related to vital signaling pathways, specifically the Ras/MAPK pathway. RASSF1A could control different protein kinases to enhance or reduce signaling activity (Ram et al. 2014; Schmidt et al. 2018). One of the most prevalent EGFR-related pathways was the Ras/RAF/MAPK signaling, a pathway that regulated cell proliferation, differentiation, migration, and metastasis in tumorigenesis (Shigematsu et al. 2005). On the other hand, our results demonstrated a similarity between RASSF1A methylation and EGFR mutation concerning clinicopathological characteristics. Smoking might be a crucial clinical factor triggering gene abnormalities in RASSF1A and EGFR. Moreover, the hypermethylation of RASSF1A and EGFR mutation occurred mainly in higher-stage tumors, implying that the synergistic abnormality of RASSF1A and EGFR alteration might drive poor outcomes and play collaborative roles in lung adenocarcinogenesis. In addition, our results showed that the probability of RASSF1A methylation was slightly higher in tumors with EGFR mutation than in those without mutation, which was contradictory to the previous findings (Liu et al. 2007; Nguyen et al. 2019; Toyooka et al. 2006). The discordance might result from specimen specificity. Further exploration is required to clarify these differences. As far as SHOX2 was concerned, no remarkable correlation was found between SHOX2 methylation and EGFR mutation in our study. However, a potential synergistic effect existed between DNA methylation and driver mutation, especially in acquiring invasiveness.

Adenocarcinoma has relatively lower levels of methylation than SCLC and SCC (Castro et al. 2010; Hawes et al. 2010; Schmidt et al. 2010; Toyooka et al. 2003) Small size, mostly peripheral types, limited location, and restricted tumor volume of the sampling are the main explanations for the lower positive detection rate and false methylation results of adenocarcinoma, especially in the early stage, by bronchoscopy biopsy. Restrictions in gene methylation detection are imposed by factors including copy number and concentration (Shi et al. 2020). In our study, we also found that the positive detection rate of methylation increased when the tumor cell content increased (Table S5 and Fig. 1). This could be due to early-stage tumors being smaller in size and still in their initial phase, releasing less tumor-specific DNA into the circulation. More tumor cells underwent necrosis and lysis as the tumor progressed, releasing more methylated genes from the cells as a result (Gao et al. 2022; Kneip et al. 2011). Tumor mass and volume might be novel indices reflecting the invasiveness of PNs (Wang et al. 2018). Occasionally, several cases of terribly malignant lung nodules presented with negative RASSF1A and SHOX2 methylation assays. The detected tumor cell amount might be another potential impact factor explaining these negative cases.

There existed several limitations in the study. First, this was a retrospective investigation at a single institution in an Asian nation. The specimens were selected from hospitalized individuals with PNs, and it might not be representative of the whole population screened by LDCT for early-stage lung cancer. Second, the sample types and quantities were inadequate. The specific number of participants in histological subgroups, such as AAH, was so limited that it could not be separated as an extra subgroup for individual analysis. Other different lung cancer subtypes, such as SCC and SCLC, have diverse genetic alterations and molecular characteristics and have demonstrated significantly high levels of methylation in RASSF1A and SHOX2. Substantial studies, including large sample size and different pathological types of lung cancer, would be required to systematically describe the differences.

Early genetic diagnosis of PNs helps in clinical decision-making and predicts nodule growth potential. For detection-negative PNs, it is feasible to extend CT scan intervals, reducing cumulative radiation exposure. Careful follow-up and timely surgical intervention should be considered for PNs presenting with molecular abnormalities. A 3-year follow-up period is a rational yardstick for evaluating the progress of nodules properly. Lesions in the preinvasive group may develop within a sufficient time of follow-up. Therefore, we aim to follow up and observe the morphological characteristics and biological behavior of these nodules in 3 years to improve the validation rate of the three-gene panel. The association between gene methylation and mutation and the mechanism of their role in the incidence and progression of LUAD warrants more investigation.

Conclusions

In conclusion, the methylation assay of SHOX2 and RASSF1A in combination with EGFR mutation detection demonstrated excellent diagnostic potential in LUAD, especially for stage I. Methylation combined with gene mutation assay could effectively identify three invasive subtypes of small malignant lung nodules and might help compensate for the subjective shortcomings of traditional pathological diagnosis. Early identification of these entities results in a more precise differential diagnosis and allows for more immediate clinical decision-making. Further prospective studies to confirm these results are required.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Jie Lin conceived and provided the research ideas for this paper. Xiangyu Ji collected samples and performed data analysis and wrote the original manuscript. Hong Li and Hui-Hui Chen are responsible for performing the experiments and evaluating the results. Xiangyu Ji# and Hong Li# contributed equally to this work and share first authorship. All authors read and approved the final manuscript.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-proft sectors.

Data availability

The data utilized and analyzed in this study are available upon reasonable request to the corresponding author. Detailed data that might endanger participant consent and confidentiality are not published.

Declarations

Conflict of interests

No grants or conflicts of interest supported this effort.