Background
Lung cancer is a most common cause of cancer-related deaths in the world. Lung adenocarcinoma (LAC) is a prevalent histological type in non-small cell lung cancer (NSCLC) [
1]. The treatment of lung cancer is individualized, and thus relied on the results of molecular biology assays and each patient’s histology [
2]. Individual responses are now suspected to tumor heterogeneity and challenge personalized medicine and biomarker development [
3]. The development of epidermal growth factor receptor tyrosine kinase inhibitors (
EGFR-TKI) and immune check-point inhibitors have led a new era in lung cancer therapy. Meanwhile, The fact that
EGFR-driven NSCLC inhibiting antitumor immunity through the activation of the PD-1/PD-L1 pathway has been demonstrated by preclinical studies. However, epidemiology studies suggest that
EGFR mutant NSCLC is more likely to decrease PD-L1 expression. To palliate these controversies, intense studies focus on tumor heterogeneity, which tends to result in mixed responses (MR) to systemic
EGFR-TKI and chemotherapy. However, the clinical significance and potential mechanisms remain to be testified. Physicians and oncologists turn on comprehensive therapies combined to targeted treatment and immune checkpoint inhibitors. Molecular discordances between primary and metastatic tumors differ among histological types [
4]. A subset of patients with
EGFR-mutant LACs fosters MRs to
EGFR-TKIs. This uncertainty is suspected with intratumor heterogeneity (ITH). An MR may be an unfavorable prognostic factor and are suspected with tumor genetic heterogeneity [
5]. Temporal/spatial heterogeneity between primary and metastatic tumors have only a limited forecast associated with markedly worsened outcomes [
6]. The new classification has been reported as an independent predictor of overall survival [
7]. LACs usually have mixed components (lepidic, acinar, papillary, solid and micropapillary) in tumor masses. Therefore, it is requisite to quantitatively evaluate histological components [
1]. It have not been ascertained whether these outcomes reflect inappropriate use of targeted therapies or greater invasiveness of tumors with increased genomic instability results in generation of multiple subclones.
Laser capture microdissection (LCM) enables researchers to recognize tissue architecture and molecular characteristics. This method that helps investigate pathological changes on a molecular, cellular, or tissue level becomes more and more precise, whereas the sample can be available in smaller and smaller sizes. This study is designed to reveal the discordance of EGFR mutation in histological subtypes and the expression of PD-L1 in AC components and to investigate the potential effectiveness on targeted therapy and chemotherapy.
Methods
Patients
261 LAC patients between 2010 and 2017 were enrolled in this study and follow up to the end of 2017. Progression-free survival of each patient was evaluated in this study. LACs were histologically diagnosed based on the WHO classification (2015). Clinical stage were evaluated according to the 7th edition of the American Joint Committee for Cancer (AJCC) staging system [
8],
EGFR mutation test were carried on and sufficient specimens were used to assess PD-L1 expression level. Clinical data were obtained from the electronic medical record database from Beijing chest hospital and all patients provided written informed consent for the use of their tumor specimens.
EGFR mutation and ALK fusion assay on heterogeneous components of LACs captured by LCM
The feature that cancer cells of the same genotype locate contiguously has been suggested on colorectal cancer via microsatellite instability [
9]. Therefore, a sample will contain a genetically identical population of cancer cells if excised small enough from a tumor tissue. All 8 μm-thick FFPE sections from
EGFR mutant patients who underwent surgical resection were stained with hematoxylin and eosin. The LMD 7000 microdissection system (Leica microsystems, Wetzlar, Germany) was used to capture pure cell subpopulations in target regions selected from
EGFR-mutant samples according to the 2015 world health organization classification. Greater than 10
4 cells in each area were captured, and 1 to 4 areas (according to the amount of tumor cells) were selectively obtained in each adenocarcinoma subtype from each section. Total DNA was extracted from each captured LCM sample via AmoyDx FFPE DNA/RNA kit (Spin Column, ADx-FF03; Amoy Diagnostics, Xiamen, China). Selected areas were tested for
EGFR mutations by AmoyDx Adx-ARMS
EGFR mutation kit (Cat. No Adx-EG01; Amoy Diagnostics, Xiamen, China).
ALK fusion gene was detected by AmoyDx Adx-ARMS
ALK fusion types (Cat. No ADX-AE02; Amoy Diagnostics, Xiamen, China).
Digital PCR detection of EGFR mutations on LCM tissues
T790M, exon 19 deletions, and L858R mutations were assessed by QX-200TM ddPCR system (Bio-Rad, Hercules, CA, USA) according to the manufacturer’s instructions. A series of EGFR T790M mutation reference standards were prepared by using Human Genomic DNA, Female (Promega, US) and NCIH1975 Cell Line genomic DNA (Research DX, US) to determine cutoffs with the following mutation allele proportion of 0, 0.1, 1, 10 and 50%. Owing to NCIH1975 cell line genomic DNA is heterozygous for EGFR T790M mutation, it was used as 50% EGFR T790M mutation reference standard. Human Genomic DNA, Female (Promega, US) is regarded as negative EGFR T790M mutation reference standard. 0.1, 1 and 10% EGFR T790M mutation reference standard contained 0.2, 2 and 20% NCIH1975 Cell Line DNA, respectively. The final concentration of the above reference is 20 ng/lL.
Twenty μl ddPCR reaction system was loaded into an 8-channel droplet generation cartridge (Biorad, Milan, Italy); Emulsion was generated with 70 μL of QX200 Droplet generation oil (Biorad, Milan, Italy) and the cartridge loaded in the QX200TM Droplet Generator (Biorad, Milan, Italy). The emulsed droplets were then transferred to a 96-well plate and amplified by standard PCR using a Mastercycler® (Eppendorf). Cycling conditions consisted of a denaturizing step at 95 °C for 5 min, followed by 40 cycles of 95 °C for 30 s and 60 °C for 1 min.
PD-L1 expression assessed by immunohistochemistry
All tumor sections were reviewed by Dr. Cai and Dr. Dong. Sections containing representative components were selected for PD-L1 immunohistochemical staining. PD-L1 (SP263) Rabbit Monoclonal Primary Antibody (Cat. No. 790–4905) and all other ancillary reagents, including VENTANA detection kits, and negative antibody (Cat. No. 790–4795) were procured from Roche Diagnostics GmbH (Mannheim, Germany). PD-L1 antibody produces membranous and/or cytoplasmic staining. PD-L1 protein was stained on the Ventana BenchMark XT with Ventana PD-L1 SP263 antibody. PD-L1 expression was evaluated on tumor cells (TC) by a three-tiered grading system on tumor proportion score (TPS): < 1, 1–49% and > =50%.
Statistical analysis
All LAC components were quantitatively diagnosed in 5% increment of tumor cells on FFPE tissue sections and each component was evaluated. Non-parameter analyses were performed on skewly distributed data. Categorical variables were compared by crosstab using chi-square test. Survival analysis by Kaplan-Meier method was performed for different groups, with the use of the log-rank test. All statistical tests were two-sided, significant level was α = 0.05. Variables included in this model were age, sex, histology, clinical stage and EGFR status. All data were analyzed by using the Statistical Package for the Social Sciences software, version 25.0 (SPSS, Chicago, IL) and GraphPad Prism (version 7.01).
Discussion
Lung adenocarcinomas frequently occur in ‘mixed pattern’ and percentages (up to 5%) of various histological components: acinar, papillary, micropapillary, lepidic and solid, are evaluated by semiquantitative assessment and should be reported according to the new WHO classification [
1]. It is crucial to adopt a practical way to address tumors comprised of a complex histological constitution, since 70 to 90% of surgically resected lung cases were diagnosed invasive adenocarcinomas. Prominent diverse patterns in morphology and heterogeneity in biology among adenocarcinomas are paid more and more attention by pathologists upon the establishment of the new classification. We commenced our study after we reviewed all the sections and renewed diagnoses based on the new classification. Articles on the topic of micropapillary AC have reported patients in early-stage with a poor prognosis [
10,
11]. It has recently been convinced that tumors classified as micropapillary also have a poor prognosis similar to adenocarcinomas with a predominant solid subtype [
12].
This study discussed the association between LAC histological subtypes, PD-L1 expression levels and primary resistance to
EGFR-TKIs. Our research demonstrated that PFS time of LAC patients exhibited a better prognosis among patients with TPS < 1%, than those with TPS > =50% (22.9 months versus 15.3 months). Since that LAC usually harbor more than one components in its parenchyma, we compared the PFS of different PD-L1 expression level with histological constitution. Patients harboring two histological components and with TPS < 1% had longer PFS time than those with TPS > =50% (25.5 months versus 13.6 months). However, prognosic significance was not evinced among LACs with single or > =three components. PD-L1 was not homogeneously expressed even in a tumor mass (Fig.
1b and Table
3). We also observed that heterogeneous PD-L1 expression, especially in the group of TPS 1–49%, was inclined to the tumor cells in the rim of tumor population against stroma (Fig.
1d). Given the heterogeneity of PD-L1 expression, it is important to understand the signals that induce the expression of PD-L1 on tumour cells. Two general mechanisms for the regulation of PD-L1 by tumour cells have emerged: innate immune resistance and adaptive immune resistance [
13]. For this reason, we observed PD-L1 expression in different LAC subtypes and compared the effect on patients’ prognosis. Acinar LACs with TPS < 1% had longer PFS time compared with those with higher PD-L1 level. Likewise, solid LACs had a similar prognosis between the groups of TPS < 1% and TPS 1–49% even if it was not significant. K. Yoshimura, et al [
14] reported that PD-L1 are also heterogeneously expressed in the same primary tumor tissue in a patchy pattern, as shown in our study (Fig.
1c and Fig.
5b). Histological subtypes were really associated with heterogeneous PD-L1 expression and patients’ prognosis.
ITH is attractive depending on recent technological advances in higher resolution and more rapid analysis on cancer genomes. Intratumoral heterogeneity is considered either the molecular features or the pathologic features of LAC. However, few studies have focused on the relationship between the two. In order to better address the current controversies and to conceive future directions, it seems apposite to investigate thoroughly the early motivations for targeting the PD-1/PD-L1 axis in
EGFR-mutated NSCLC. This concept was in part recommended by retrospective studies prompting frequent PD-L1 expression in
EGFR-mutated NSCLC [
15,
16]. The co-occurrence of PD-L1-positivity and activating
EGFR mutations in clinical NSCLC specimens was first reported based on 164 surgically resected samples [
16]. Evidence to support this notion is that PD-L1 is a downstream target of
EGFR signaling, and this is interceded through IL-6/JAK/STAT3, NFĸB and p-ERK1/2/p-c-Jun pathways [
17‐
19]. PD-L1 expression can be repressed by
EGFR-TKI [
18]. In contrast, one study found PD-L1 expression was increased following gefitinib treatment [
20]. However, the conclusions from previous studies are drawn based on the analysis on tumor lesion rather than on each histological component. What potential impact of
EGFR mutated cancer cells expressing PD-L1 in tumor cell population remains be illuminated. A striking finding of our study is the identification of intratumoral genetic heterogeneity in LAC that harbors
EGFR heterogeneous alterations and differential PD-L1 expression. In spite of
EGFR status, PD-L1 overexpression occurred in LACs with solid (47.6%) versus those without this pattern, whereas low level expression of PD-L1 in mucinous ACs. Zito Marino, F. et al reported that PD-L1 expression was more frequently in LACs with solid pattern [
21]. In addition, PD-L1 expression was not homogeneous in a tumor parenchyma and more common in heterogeneous pattern. Therefore, we further investigated and analyzed the relevance of
EGFR statuses and PD-L1 expression in the 18 LACs with sensitizing/T790M mutation. We noticed that PD-L1 expressed more frequently in those components without
EGFR mutation, especially in acinar areas, but really overexpressed in
EGFR-mutated components. We suspect that
EGFR-mutation driven PD-L1 expression are activated through the pathways as reported before [
22]. In our study, LACs with single components harboring
EGFR mutation tended to be short of PD-L1 expression. It seemed to be paradoxical with previous study that the biological association is indirectly consolidated by coexistence of PD-L1 upregulation in
EGFR-mutant NSCLC, as observed in retrospective studies [
16,
19], but not supported in subsequent pooled analysis [
22]. However, another cancer cell-centric mode of PD-L1 upregulation is considered as adaptive immune resistance [
23], which is characterized by an increased expression of PD-L1 on tumor cells and immune cell subpopulation in reaction to robust CD8+ T-cell-mediated immunosurveillance. Adaptive PD-L1 upregulation relies on effective immunorecognition, which is promoted by an increased somatic mutational and neoantigen burden. However, for hitherto indefinite reasons, mutational burden seems to be lower in
EGFR-driven tumors [
24]. We consider that this dilemma results from the ITH of LACs. The fact that cancers with
EGFR mutation partially expressed PD-L1 protein suggests that a tumor mass seems to exist as if intercoursed subpopulations of cancer cells present with different biological behavior, both in PD-L1 expression and
EGFR mutation. These attributes enable LACs to represent a heterogeneous response to clinical therapy. It is also manifested that histological subtypes really prompt the possibility of potential resistance to both
EGFR-TKI and PD-L1-related immunotherapy.
A striking finding of this study is the identification of intratumoral genetic heterogeneity in LACs harboring driver coalterations. 17 patients was identified with co-occurrence of
EGFR sensitizing and resistant mutations providing an incidence rate of 6.5%. Because it is unclear whether sensitizing and resistant mutations coexist in same or different tumor cells, we used LCM to capture pure tumor cells within both the same and different growth patterns. With this method, we found that
EGFR mutation types did not concomitantly coexist in all tumor cells. Interestingly,
EGFR mutations by ddPCR assay showed that abundance of these mutations were different in the same cell population. Therefore, we determined a difference in the driver status among spatially separated tumor areas. In particular, the relative abundances of the two altered genes were different in the same tumor areas, which suggests that oncogenic genetic profile may not be the same in all tumor cells within a same primary tumor (Fig.
5c-e). To fully recognize morphologic heterogeneity, we observed that the abundances of sensitizing and T790M mutation were not concordant in micropapillary and papillary components. We also observed similar results by using ARMS according to cycle threshold value. Intratumoral heterogeneity of
EGFR mutations is demonstrated associated with the distribution of histological components in mixed LACs [
25]. We also revealed that intratumoral heterogeneity of
EGFR mutations really existed in the same histological subtypes of LAC. Therefore, we speculate that clone evolution, not only histological heterogeneity, may be mainly responsible for molecular intratumoral heterogeneity of LAC. The findings of intratumoral heterogeneity in
EGFR sensitizing/T790M coaltered LAC may be hypothesized by Darwinian-like clonal evolutionary dynamics and the resulting complex clonal architecture of LAC as reported by Cai et al [
26]. In this study, tumor cells with dual altered driver genes occurring mainly in micropapillary and papillary components prompt that these components most likely harbor heterogeneous clone. Previous studies have shown that a substantial proportion of malignant tumors have a multiclonal signature [
27]. The fact that T790M clones were selected by the TKI therapy support this hypothesis, because T790M occurs in 60% patients upon
EGFR-TKI therapy [
28,
29]. Differential expression of PD-L1 in a tumor mass seems to testify this speculation as well. Discordant PD-L1 expression in a tumor mass is thereof suspected to reactive/adaptive expression owing that it usually occurs among the peripheral cancer cells against stroma. Resistance to TKIs is considered one of the unknowns of cancer; therapy selection may make tumors become more heterogeneous, which may be the major reason for resistance to TKIs [
30]. The complex dynamics of clonal evolution could produce unique and unpredictable patterns of clonal architecture that are spatially and temporally heterogeneous [
31,
32]. Clonal evolution underlying tumor progression probably proceeds in a branching, rather than in a linear manner, which might result in substantial clonal diversity that accordingly contributes to genetic heterogeneity within tumors [
32]. Importantly, our findings may provide a rationale more reasonable to treat patients with dual mutations with Osimertinib. In addition, any pathologic diagnosis based on a whole mass does not adequately determine the oncological heterogeneity. Analyses on the correlation of histological subtypes and
EGFR mutation show that
EGFR mutation were significantly more common in LACs with micropapillary, papillary and lepidic pattern and less common in those with mucinous pattern. However, considering the potential impact of genetic intratumoral heterogeneity on histological feature, especially in LAC with more than one histological component, the situation may be more intricate than it looks.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.