The dynamic nature of nonsmall-cell lung cancer (NSCLC) resembles a Lernaean Hydra precipitating variable response to treatment due to the molecular diversity of cancer cells in space and time. Heterogeneity lies at the core of the resistance against targeted therapies and represents a remarkable roadblock to the delivery of curative treatments. Its accurate assessment is a dire need. NSCLC, which includes adenocarcinomas, squamous and large-cell carcinomas, is considered a promising model to study heterogeneity in cancer because of the plethora of distinct molecular signatures, which determine its course. In this context, next-generation sequencing (NGS) seems to be the most potent tool to investigate this phenomenon. This article provides an overview of the etiology, detection, and management methods of intratumoral heterogeneity in NSCLC and discusses their clinical implications.
Tumor heterogeneity describes the observation that different tumor cells can show distinct morphological and phenotypic profiles, including cellular morphology, gene expression, metabolism, motility, proliferation, and metastatic potential. This phenomenon occurs both between tumors (intertumoral heterogeneity) and within tumors (intratumoral heterogeneity). Intertumoral heterogeneity occurs in tumors of the same histological type, owing to the diverse population of cancer cells in a single patient. Intratumoral heterogeneity is associated with variations among the tumor cells of a single patient. It does not refer to the static coexistence of different cancer cells, but to the dynamic adaptation and constant clonal selection of cancer cells. Intratumoral heterogeneity can be further classified as spatial and temporal heterogeneity. Spatial heterogeneity arises due to the uneven distribution of genetically diverse tumor subpopulations either within the primary tumor or in different metastatic lesions. Conversely, temporal heterogeneity is a dynamic process of cancer cells differentiation occurring gradually in the course of the natural progression of the disease or as a result of the selective evolutionary pressure induced by treatment interventions[1, 2]. NSCLC entails both longitudinal and temporal heterogeneity.
Intratumoral spatial heterogeneity can be detected at four levels, namely: (a) in single cells, (b) in distinct lesions within the same organ, (c) between primary and metastatic lesions, and (d) among metastatic lesions. Molecularly distinct cellular subpopulations can be determined in primary tumors due to either the asymmetrical distribution of key molecular driver alterations or the uneven distribution of additional passenger mutations. Genomic instability seems to affect the copy-number alterations, which appear as a promising biomarker. Spatial heterogeneity in an individual tumor should be investigated by multiregion sampling, so that a better delineation of its extent and a lower misrepresentation of clonal mutations can be achieved even in early-stage malignancies. The same applies to distinct lesions in the same organ, where multiregion sampling can decipher the presence of multiple clonally distinct cancer cells in the same organ. A lower level of genetic variability has been observed among primary and metastatic sites perhaps. Although malignant cells in the same tumor are more closely related than those in different tissues, whose differentiation is subject to tissue-specific factors, distant and lymph-node metastases tend to arise from independent seeding by genetically distinct subclones originating from the primary site. To date, this has served as an explanation of the lower level of genetic variation observed[3,4,5].
On the other hand, temporal heterogeneity is related to the dynamic variation in the genetic profile of a tumor over time, especially under the selective pressure of antineoplastic treatments on oncogene-driven pathways. It has been documented that this pressure has a greater influence on the course of the disease in patients under targeted therapies, rather than in patients receiving conventional chemotherapy. Two mechanisms have been hypothesized to explain this kind of heterogeneity, namely preexisting resistant clones surviving antineoplastic therapy and the expansion of drug-tolerant cells bearing
The transition from bulk sequencing phylogeny approaches to single-cell sequencing data broadens the understanding of heterogeneity. The combination of bulk and single nucleus sequencing (SNS) data is going to lead to novel modeling frameworks, which will decipher more secrets of this cancer cells pluralism including tracing cell lineages, understanding rare tumor cell populations, and measuring mutation rates[9, 10]. A thorough understanding of a tumor's subclonal composition and its mutational history is achievable through flow-sorted nuclei, whole genome amplification (WGA), and NGS. The combined use of these methods leads to accurate quantification of the genomic copy number inside individual cancer cells’ nuclei. SNS is a particularly promising sequencing method already tested in breast cancer and glioblastoma, which allows for tumor population structure and evolution to be more thoroughly investigated[11,12,13]. The specific process includes the isolation of nuclei by flow-sorting and the amplification of DNA using WGA for massively parallel sequencing, so that low coverage (<10%) of the genome of a single cell can be accomplished, which is sufficient to quantify copy number from sequence read depth[12]. It is noteworthy that significantly more mutations can be identified with the use of multiregion whole-exome sequencing (WES) than with single-sample analysis or with the use of single samples of the Cancer Genome Atlas[14].
Furthermore, genetic insights regarding both “driver” and “passenger” genes in a wide range of cancer types are feasible through NGS. Specifically, driver genes can lead to clonal growth advantage in comparison with passenger genes. As far as the process for this molecular analysis is concerned, several hundred million short (~250 bp) sequence reads can be produced by massive parallel sequencing technology. In one day, one human genome can be affordably analyzed through the conversion of both deoxyribonucleic acid (DNA) and ribonucleic acid (RNA) into complementary DNA, followed by fragmentation and sequencing. Eventually, somatic genome alterations such as mutations, copy-number changes, and structural rearrangements can be traced on a genome-wide scale. Moreover, whole-transcriptome sequencing (WTS) seems to play a complementary role in the evaluation of expressed oncogenes including fusion genes, as well as the identification of noncoding RNA and splice variants, so that a more comprehensive approach toward the molecular classification, the functional differentiation of mutant genes and epigenetic regulators can be reached. Consequently, the abovementioned technologies significantly aid in illustrating the molecular maze of cancer[15].
NSCLC, especially adenocarcinoma, serves as an example for both the driver gene-based tumor classification and the interpretation of heterogeneity in cancer for a plethora of reasons. In total, over 60% of lung adenocarcinomas have been identified for driver mutations by genetic profiling[16]. Specifically, TP53 and KRAS were documented to be the most frequent (>30% of cases) mutated driver genes followed by four recurrent (>10% of cases) genes (KEAP1, STK11, EGFR, and NF1) in a sample of 230 resected lung adenocarcinoma. BRAF (10%), PIK3CA (7%), MET (7%), (ALK)/RET/ROS1 (4%), and HER2 (2–4%) were also included in the arsenal of targetable genes[7, 16]. It is noteworthy that in the presence of intratumor heterogeneity, single tumor-biopsy samples can lead to an underestimation of the tumor genetic variability[1]. The development of reliable noninvasive methods to monitor heterogeneity is crucial for personalized treatment in the future, given that repeated sampling biopsies might be associated with complications, poor tolerance, and compliance. Molecularly targeted therapies are applicable in the therapeutic field of lung cancer and driver gene-based tumor classification should play a determining role in their
NSCLC seems to be a promising model to study and map heterogeneity in cancer[17]. Intratumor heterogeneity may entail tumor adaptation and therapeutic failure through Darwinian selection since a tumor with a high diversity index is expected to become resistant to chemotherapy, because it is more likely than a homogenous tumor mass to harbor preexisting drug-resistant clones[10, 18]. What is more, emerging research has described an association between the total number of mutations, the frequent appearance of neo-antigens by protein-coding mutations and better response to immunotherapy based on the genetic evaluation of both tumor and host immune system. Finally, intratumoral molecular heterogeneity challenges the current “static” molecular classification of a tumor. A remarkable example is the dynamic change in clonal constitution within the tumor, which plays a crucial role in acquired drug resistance[15].
From an etiological point of view, intratumoral heterogeneity reflects a Darwinian selection pattern in tumor progression. Both genetic and nongenetic pathways can lead to intratumor heterogeneity. In the first case, a monoclonal tumor population can become heterogeneous through genomic instability caused by genome doubling, chromosomal instability, and mutational processes. In the latter, plenty of factors such as alteration of oxygen level, availability of nutrition, location within the primary tumor, interaction with neighboring cells, and presence of cancer stem cells can result in preferential outgrowth of some cell populations and, subsequently, in intratumor heterogeneity[19].
As far as the genetic model is concerned, genome doubling and chromosomal instability seem to play a crucial role in the aforementioned heterogeneity. First, whole-genome duplication seems to be an early event in NSCLC evolution. There is strong evidence of an association between genome doubling events and the propagation of chromosomal instability, which is related to poor prognosis. Genome doubling is mostly clonal and is identified in approximately 3 out of 4 NSCLCs[14]. A mirrored subclonal allelic imbalance profile has also been found in the majority of enriched with copy-number alterations in NSCLCs through a complicated mechanism, in which a maternal allele is gained by a subclone in one region originating by a paternal allele that was lost by a different subclone in another region. Furthermore, chromosomal instability might be an initiator of both copy-number and mutational heterogeneity via a second mechanism, which includes the loss of genomic segments carrying clonal mutations. Finally, when it comes to mutational processes in both squamous cell carcinomas and adenocarcinomas, it has been found that the burden of mutations related to smoking is significantly correlated with the number of early mutations, which are accumulated before genome doubling or copy-number alteration. In a prospective cohort study of multiregion WES on 100 early-stage NSCLC tumors, 20% of these tumors expressed subclonal driver mutations, which could be attributed to APOBEC activity illustrating the potential role of APOBEC mutagenesis in inducing a subclonal driver event, which may promote subclonal expansions[14]. Noteworthy is the fact that increased genomic instability seems to contribute to the emergence and spreading of more competitive subclones[6].
Clonal evolution represents another important cause of intratumor heterogeneity. This evolution can further be categorized into linear and branched. The former pattern of evolution refers to the sequential genetic alterations, which are beneficial to the successive generations in order to gain proliferative advantage and dominance against the preceding clones, while the latter model is related to multiple genetically distinct populations, which are proliferated over time, while certain subclonal populations are gradually diverging from the common ancestor. The branched model of evolution, which is adopted by many solid tumors can lead to the progression of a more heterogeneous tumor[6]. Data have shown that the selection is persistent in NSCLC evolution and that limitations determine evolutionary trajectories. Particularly, advanced NSCLC has a high percentage of nonubiquitous mutations indicative of branched evolution compared to the low degree of heterogeneity in SCLC, which suggests a parallel and linear pattern of evolution[20].
NSCLC is portrayed as an extremely promising model for studying cancer heterogeneity fueling a new era in genotype-guided approaches. Although there is well-established knowledge about intertumoral heterogeneity, there is limited knowledge about intratumor heterogeneity, mostly based on small retrospective cohorts. The distinct pattern of expression of EGFR and KRAS status has been documented between lung squamous-cell carcinomas and lung adenocarcinomas, but also inside the primary tumor
A plethora of emerging studies illustrate the spatial heterogeneity of NSCLC. A representative study of this diversity revealed that 75% of early-stage surgically resected NSCLC tumors were determined by subclonal oncogenic alterations. In the aforementioned study of Jamal-Hanjani, multiple sampling appears to be of major importance given that, otherwise, more than 75% of subclonal mutations would have been misrepresented as clonal despite the high level of intratumoral diversity, which is expressed by the fact that 30% of somatic mutations are estimated to be subclonal[14]. Although this heterogeneity can be observed even within premalignant lesions, the “mutational portrait” of these preinvasive lesions seems to be distinctly different from that of malignant lesions[18]. Therefore, this phenomenon highlights the potential role of premalignant populations in the molecular mosaic of contiguous malignant cells.
Spatial heterogeneity has also been observed between primary and metastatic lesions, as well as between multiple metastatic lesions. Data are supporting the presence of a high level of concordance in mutation status both between matched primary and metastatic tumors and among multiple metastatic lesions[24]. Specifically, EGFR mutation concordance between primary and metastatic lesions of NSCLC is 83–88%[25]. Vague remains, though, to which extent can minor clones become dominant in distinct metastasis. Regarding brain metastases, NSCLC, especially the adenocarcinoma subtype, is the most common metastasizing cancer to the brain posing an increased risk (30–50%) of developing brain metastases during the course of the disease in patients with treated stage III NSCLC[26]. In a study of 86 patients, who have been diagnosed with lung (38 patients), breast, and renal cancers, 53% harbored a potentially clinically actionable alteration in the brain metastasis, which was not detected in the clinically sampled primary tumor. These changes may have crucial clinical implications given that brain metastases are often developed even when presumably truncal mutations of the primary tumor are successfully targeted with active systemic agents[27].
Regarding temporal heterogeneity, two major distinct evolutionary pathways seem το delineate the emerging resistance, which is attributed to heterogeneity. The first option refers to preexisting treatment-resistance clones, which present
The third-generation EGFR-inhibitors, such as osimertinib, are a remarkable example of the contribution of both spatial and temporal heterogeneity to emerging resistance mechanisms against targeted therapies. The first hypothesis refers to T790 wild-type subclones, which might preexist before osimertinib administration justifying the resistance in nearly half of cases. The second mechanism includes the
Another remarkable paradigm of this kind of heterogeneity is the development of acquired resistance against crizotinib due to a mutation in CD74–ROS1. Specifically, ROS1 G2032R and L2155S mutations conferred resistance to crizotinib and it has been considered to be an early event in the clonal evolution of resistance, given that the same resistance mutation is detected in all the metastatic sites in case-report series[32,33,34,35]. These molecular alterations related to this resistance in ROS1-rearranged NSCLC seem to be heterogeneous, including ROS1 tyrosine kinase mutations, EGFR activation, and epithelial-to-mesenchymal transition[32]. Regarding the G2032R mutation, it was demonstrated to be subclonal in the aforementioned case report. A potential mechanism, which could explain this of metastatic spread, is termed “self-seeding” meaning the exchange of cells among noncontiguous tumor regions[35]. Evidently, the proper noninvasive monitoring of heterogeneity appears to be a prospective requirement for emerging treatment administration. The single genetic snapshot of a diagnostic biopsy will not be sufficient in this regard. Minimally invasive techniques such as liquid biopsies at well-tolerated sampling intervals are promising tools for the detection and management of heterogeneity.
A body of evidence based on analysis of metastatic patients’ ctDNA has demonstrated that liquid biopsies are considered a highly sensitive diagnostic method to identify clinically relevant mutations with a high grade of concordance between ctDNA and tissue biopsy sample analysis. Furthermore, this minimally invasive method appears to contribute to the early detection of relapse of NSCLC and the characterization of emerging subclones seeding metastatic sites providing insights for new therapeutic strategies and identifying adjuvant chemotherapy resistance. Nonadenocarcinoma histology, necrosis, increased proliferative biomarkers, and lymphovascular invasion are included in the predictors of ctDNA detection in early-stage NSCLC, while there is the potential of extrapolation of these results beyond NSCLC to other malignancies, such as breast cancer and
Data have demonstrated that the molecular profile of the tumor can be precisely portrayed in patients’ plasma. Particularly, Jerkins et al. designed a retrospective study to evaluate the clinical utility of the cobas plasma test for the detection of the T790M mutation in ctDNA obtained from patient plasma samples collected retrospectively during screening for the AURA extension and AURA2 studies of osimertinib as second-line treatment in NSCLC. Approximately 80% of patients with a cobas tissue test provided matched tumor tissue and plasma samples. It has, also, been documented that the size of the baseline lesions is a determining factor when it comes to T790M detection in the plasma, given that this mutation was identified in the plasma of 97% of those with primary lesions of 120 mm or larger but in only 53% of the ones with baseline lesions smaller than 40 mm. Furthermore, the positive percent agreement (PPA) of second-line therapy with osimertinib was lower than the PPA of third- or later-line therapy. It could be hypothesized that the detection of T790M in plasma ctDNA of patients with progression of the disease after first-line EGFR TKIs is more challenging due to either increased shedding of ctDNA from tumor cells because of prior chemotherapy or greater disease burden and a greater allelic fraction of T790M[38]. From a therapeutical perspective, longitudinal plasma analysis suggests that the treatment with third-generation EGFR inhibitors, such as osimertinib can decrease the plasma EGFRT790M allelic fraction over time, which is possibly associated with the negative selection of EGFRT790M-positive subclones.
Moreover, in another study regarding the emerging resistance to third-line TKIs, the dominant driver of resistance (T790M-positive or T790 wild-type) identified within plasma was confirmed on the corresponding tumor biopsy. In addition to substantiating the role of tumor heterogeneity as a major factor in osimertinib resistance, these results emphasize the clinical utility of plasma ctDNA, which could play an important role in evaluating response and indicating emerging resistance in clinical practice. The detection of novel mechanisms of resistance to third-generation EGFR TKIs is feasible through ctDNA genotyping, which could promote new therapeutic approaches for these patients. Liquid biopsy (ctDNA, CTCs, and exosomes), seems to be the most promising strategy to study the dynamic evolution of oncogene-addicted tumor cells during treatment with different EGFR TKIs being able to predict possible recurrence and detect potentially targetable molecular mutations[30, 39].
Although the emergence of resistance in NSCLC has been documented to be detectable on plasma ctDNA analysis prior to the appreciation of both radiographic and clinical progression, there are data of radiogenomic studies correlating radiographic findings with intratumoral heterogeneity[28]. Imaging, especially computed tomography texture analysis (CTTA) and conventional positron emission tomography/computed tomography (PET/CT), might be alternative methods for the identification of tumor heterogeneity. Radiogenomic studies conducted in order to identify the NSCLC patient's spatial heterogeneity, have demonstrated correlations between genomic aberrations and specific CT features in NSCLC and other tumors. Particularly, CTTA can lead to the quantitative assessment of genetic instability, given that, genetically distinct subclones within a tumor, which express different phenotypes, can contribute to depicting heterogeneity[49]. Moreover, metabolic texture analysis on FDG-PET/CT seems to be able to identify the intratumoral heterogeneity and this exam might be useful as a radiogenomic marker of global intratumoral genetic heterogeneity. Specifically, hypoxia and genomic alterations in NSCLC are associated with FDG uptake that is correlated with tumor aggressiveness and a poor prognosis of survival[34]. Therefore, the assessment of metabolic heterogeneity using textural parameters of FDG-PET/CT images has been proved to have independent predictive value regarding performance-free survival (PFS) of EGFR TKI in EGFR-mutant NSCLC. Therefore, patients with increased metabolic heterogeneity should be considered as a high-risk subpopulation for early EGFR TKI failure[41, 42].
Although current treatments can initially reduce tumor burden, relapse occurs in most cases. A deeper understanding and control of the heterogeneous landscape of lung cancer might provide the oncology community with Ariadne's thread. Notably, elevated copy-number diversity and intratumor heterogeneity mediated through chromosomal instability, have been associated with an increased risk of recurrence or death, a finding that supports the potential value of chromosome instability as a prognostic predictor according to TRACERx ClinicalTrials[14]. A high degree of tumor heterogeneity entails the increased probability of resistant clones, which will be already present at the beginning of treatment. On the other hand, a high degree of heterogeneity may also be indicative of a relatively high mutation rate within the tumor cells and, therefore, an increased chance of developing drug resistance-inducing mutations. Consequently, the inclusion of multiple samples from different geographical sites of the primary tumor appears to be of paramount importance[22]. The selection of the most optimal primary treatment should depend on the specific identification of trunk mutations that are drivers and can be targeted[20].
The onset of chromosomal instability seems to have a detrimental impact on the evolution of NSCLC, given that it is both a significant promoter of parallel evolution and a predictor of poor prognosis. According to the TRACERx Consortium, patients who have been diagnosed with early-stage tumors with increased levels of copy-number heterogeneity constitute a high-risk group with shorter relapse-free survival and they might be in dire need of close monitoring and early therapeutic intervention during their follow-up period. It is noteworthy that temporary heterogeneity also lies in the core of mixed responses (MR) to systemic therapy, provided that the overall incidence of NSCLC patients with MR has been documented to be approximately 22%. This phenomenon is predominated in stage IIIB–IV disease, EGFR mutations, and those who received TKI therapy (
The
As has already been mentioned, NSCLC is a molecularly heterogeneous disease. There are genetic signatures that are related to specific mutagenic processes. A striking example is the smoking-related NSCLCs that are enriched with C>A transversions, significantly higher mutation frequencies, and bear distinctive sets of mutations: KRAS, TP53, BRAF, JAK2, JAK3, and mismatch repair gene mutations[47]. Exposure to chemotherapy might be a subsequent mechanism of creating genomic instability through the increase of the mutational range of the tumor, which, therefore, might be highly immunogenic and more vulnerable to immunotherapy.
To begin with, immune-checkpoint inhibitors against programmed cell death (PD-1), such as nivolumab and pembrolizumab, as well as its ligand (PD-L1) have proved to bear remarkable therapeutic benefits in metastatic NSCLC. However, appropriate biomarkers have yet to be added to our screening arsenal. Currently, PD-L1 is considered to be the unique biomarker in daily clinical practice, but its predictive and prognostic role is controversial owing to the limitation of immunochemistry regarding the portrayal of individual variability and tumor heterogeneity. It has been highlighted that there is significant diversity of PD-L1 expression in tumor cells of NSCLCs and mainly in the related immune cells. Particularly, there is a study of 144 patients diagnosed with NSCLC, which has shown that 10–20% of adenocarcinomas could be misclassified based on PD-L1 expression of tumor cells and up to 27% relied on lymphocyte PD-L1 expression. Moreover, PD-L1 expression of immune cells was documented to be more heterogeneous compared to that of tumor cells in adenocarcinomas. Notably, there is evidence that PD-L1 expression of immune cells might be predictive of benefit from atezolizumab[48, 49].
Furthermore, it has been reported that PD-L1 expression is closely associated with EGFR mutations and EML4–ALK protein fusion, leading to an immunosuppressive tumor microenvironment that entails immune escape of NSCLC. Therefore, the combination therapy of immunotherapy and EGFR-TKIs seems to consist of a potential alternative strategy, which has yet to be thoroughly investigated. Rebiopsy and PD-L1 reevaluation following conventional treatments of ineligible patients for immunotherapy may finally render immunotherapy an emerging beneficial choice for them[50].
Intertumoral and intratumoral heterogeneity has a considerable contribution to treatment failure in NSCLC. The combination of Darwinian selection and the innate diversity of cancer cells and its clinical sequelae appears as a hard-to-untie Gordian knot. The spatial and temporal diversity of cancer cells within a single patient set additional challenges to personalized precision medicine, calling for continuous cellular and molecular–level surveillance and adequate adjustment of the treatment plan. This also calls for careful consideration of treatment algorithms in an attempt to decrease the likelihood of resistance-bearing mutations. Liquid biopsies combined with CT and/or PET/CT imaging and a plethora of genomic sequencing techniques have a major potential to be integrated with the standard of care and help develop from-the-bench-to-the-clinic treatment strategies.