Gene expression profiling for the diagnosis of multiple primary malignant tumors

Multiple primary malignant tumors (MPMTs) are defined as two or more histologically distinct malignancies in one individual. With the development of screening technologies as well as significant treatment advances, early detection and precise treatment have led to a dramatic increase in the population of cancer survivors. In addition to this increase in the population of cancer survivors, the incidence of MPMTs is rising due to increased aging of the population [1]. Worldwide, a meta-analysis of 12 studies revealed that the frequency of MPMTs in a cancer population varies between 2.4 and 8% and is up to 17% within 20 years of follow-up [2]. In China, two epidemiological studies reported that 0.99 to 1.09% of cancer patients could develop a second primary malignancy [3, 4]. The risk of developing a second primary tumor varies across first tumor type; bladder cancer is most common as the first primary tumor, and lung cancer is the most common second primary tumor [5]. Certain patient populations, including male patients and patients with a history of smoking or alcoholism, are also at higher risk of developing MPMTs [2, 6].

When a patient with a prior cancer history has a second malignant lesion, identifying the tumor origin of the new lesion has important prognostic and therapeutic implications and still represents a difficult problem in clinical practice. If the second lesion is a primary cancer, it could be cured by radical operation supplemented by chemotherapy and/or radiotherapy, which is similar to the treatment of a single primary cancer. In contrast, recurrent or metastatic tumors indicate that the primary tumor has progressed to advanced stages. Palliative treatment is the first choice for recurrence or metastasis of the primary tumor. In the clinic, histopathologic analysis can help characterize the tumor origin in most cases. However, tumor heterogeneity and interobserver variation between pathologists can cause confusion, especially when metastatic foci are poorly differentiated or undifferentiated [7].

In recent years, gene expression profiling has been widely studied and has become a powerful tool in distinguishing the origin of tumors. Previous studies have suggested the clinical utility of gene expression profiling in distinguishing synchronous primary malignancies of the ovary and endometrium or metastatic spread from either the ovary or the endometrium, as well as in distinguishing between second primary lung cancer and lung metastasis from head and neck tumors [8, 9]. Nevertheless, few data support the broad application of gene expression profiling for MPMTs.

Recently, Ye et al. established a pan-cancer transcriptome database comprising 5434 specimens representing 21 tumor types (as shown in Additional file 1: Table S1), and adopted the SVM-RFE algorithm (Support Vector Machine Recursive Feature Elimination) to select the Top-10 most predictive genes for each of the 21 tumor types [10]. After removing redundant genes, a list of 90 genes specific to 21 tumor types was identified. The details of 90-gene list were provided in Additional file 2: Table S2. For instance, gene ACPP was significantly over-expressed in prostate cancer, while gene GATA3 was shown to be highly expressed in breast cancer, and gene SLC3A1 was significantly over-expressed in kidney cancer (Additional file 3: Figure S1). Gene Ontology and KEGG pathway analysis reveal that a diverse group of gene families is represented in the 90-gene list [10]. The most significantly enriched gene categories are those involved in specific biological processes, including tyrosine metabolism, fat digestion and absorption, cytokine-cytokine receptor interaction, extracellular matrix-receptor interaction, and gastric acid secretion. Of interest, but not surprisingly, genes described in oncogenic pathways such as those of bladder cancer, melanoma, and prostate cancer were also significantly over-represented, reflecting their differential involvement in a range of tumor classes. Next, a 90-gene expression assay was established for the classification of 21 common tumor types using quantitative real-time polymerase chain reaction (qRT-PCR) methods with total RNA extracted from formalin-fixed, paraffin-embedded (FFPE) tissue [10]. In a validation study that included 609 clinical samples, the 90-gene expression assay demonstrated an overall accuracy of 90.2% for primary tumors (292/323) and 87.3% for metastatic tumors (255/286). In addition, Wang et al. applied the 90-gene expression assay for the differential diagnosis of metastatic triple-negative breast cancer (TNBC) [11]. The gene expression assay correctly classified 97.6% of TNBC lymph node metastases (41/42) and 96.8% of distant metastatic tumors (30/31). Zheng et al. investigated the 90-gene expression assay for diagnosing the tumor origin of brain tumors [12]. The molecular assay showed 100% accuracy for discriminating primary brain tumors from brain metastases, and correctly predicted primary sites for 89% of brain metastases (39/44). More recently, Ning et al. demonstrated the strengths of the 90-gene expression assay in distinguishing multiple primary squamous cell carcinomas in head and neck, esophageal, and lung cancers [13]. In current study, we aim to evaluate the performance of the 90-gene expression assay and explore its potential diagnostic utility for MPMTs. Our results show that this PCR-based gene expression assay might serve as a useful tool for identifying the origin of MPMTs.

Currently, with the significant development in tumor screening and diagnostic tools, the detection rate of MPMTs is higher than before. Additionally, the appearance of new treatment methods, including targeted therapy and immunotherapy, has helped cancer patients achieve longer survival. Therefore, the likelihood of cancer survivors having a second malignancy is higher than before because of the longer follow-up times. Lindsay et al. found that the occurrence of cancer survivors developing second cancers rose dramatically from 9% in 1975–1979 to 19% in 2005–2009 [1]. Identifying the tumor origin of the new lesion for patients with cancer history is crucial for selecting the treatment strategy. A second primary cancer could be cured by radical operation supplemented by chemotherapy and/or radiotherapy, which is similar to the treatment of a single primary cancer. In contrast, recurrent or metastatic tumors indicate that the primary tumor has progressed to advanced stages. Palliative treatment is the first choice for recurrence or metastasis of the primary tumor. For people who develop a second primary cancer, 13% die from their original cancer, and 55% die from their second primary cancer [5]. Thus, how to accurately diagnose tumors as metastatic or a second primary cancer in a timely manner is becoming increasingly important.

Histopathologic examinations, such as morphological and IHC analyses, are the cornerstone of traditional cancer diagnosis. However, a meta-analysis showed that IHC provides correct tissue identification in 65.6% of metastatic cancers [20], and recent studies suggest that molecular profiling outperforms classification by IHC, in particular in cases with poorly differentiated tumors [7, 21]. In addition, for patients with multiple SCC lesions, morphological and IHC analyses cannot provide a definite diagnosis of primary site given similar histologic appearances among across several types of SCCs.

Previous studies have demonstrated that gene expression profiling between primary and metastatic tumors shows a high degree of similarity [7, 10]. Efforts have been made to use gene expression profiling to distinguish the second primary tumor from metastatic recurrence. Charles et al. found that gene expression profiling could accurately identify 89% of metastatic tumors, whereas IHC achieved 83% agreement. In poorly differentiated or undifferentiated subsets, gene expression profiling correctly distinguished 91% of cases compared to 71% using the IHC method [21]. Anita et al. developed a gene expression profiling test to distinguish between synchronous primary malignancies of the ovary and endometrium or metastatic spread from either the ovary or the endometrium and achieved an accuracy of 94.7% (71/75) [8]. In addition, Anital et al. reported a new approach to determine if a lung nodule was a second primary or a lung metastasis from head & neck tumors with an overall accuracy of 82.9% (63/76) [9]. Regardless, few studies have validated the performance of gene expression profiling in MPMTs across a broad spectrum of tumor types.

Here, we describe the investigation of an effective approach for the molecular classification of MPMTs. Similar to findings from previous studies [11, 12], the success rate of the 90-gene expression assay was excellent (87%), even for samples archived five years prior to analysis, indicating high reliability of the assay with FFPE samples. This may be critical for widespread access and application in clinical practice. The 90-gene expression assay classified 73 MPMT samples into twelve tumor types and reached an overall accuracy of 93.2% (68/73) when compared with the pathological diagnosis. To the best of our knowledge, this is the first report of a gene expression assay that can be used to classify MPMTs into a wide range of cancer types. Our findings suggest that the 90-gene expression signature may serve as a useful tool for identifying the tumor origin of MPMTs.

Our results represent an encouraging primary step, but this study still has several limitations. First, MPMT patients were enrolled from two hospitals. The number of several tumor types was relatively small to obtain a solid conclusion. In the future, additional multi-center validations should be performed to further evaluate the performance of the 90-gene expression assay. Second, in current study, RNAs extracted from long-term archived FFPE samples are likely degraded, and thus might diminish the distinctness of gene expression pattern cross different cancer types. For routine diagnostic evaluation, fresh tumor or newly prepared FFPE sample reserving high quality RNAs are highly recommended for precise gene expression analysis.

In conclusion, the results of this study demonstrate the promising performance of the 90-gene expression assay in identifying the tumor origin of MPMTs. In cases which the morphological and IHC work-up cannot clearly confirm the tumor origin, the 90-gene expression assay may serve as a useful tool for identifying primary site of new lesions for cancer patients, especially for SCC patients. Future incorporation of the 90-gene expression assay in tumor origin diagnosis will assist oncologists in applying precise treatments, leading to improved care and outcomes for MPMT patients.